1 |
commit: 7f8cc789b776e0cf768d27cff4118be5960094c0 |
2 |
Author: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
3 |
AuthorDate: Tue Feb 11 16:20:35 2020 +0000 |
4 |
Commit: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
5 |
CommitDate: Tue Feb 11 16:20:35 2020 +0000 |
6 |
URL: https://gitweb.gentoo.org/proj/linux-patches.git/commit/?id=7f8cc789 |
7 |
|
8 |
Linux patch 4.19.103 |
9 |
|
10 |
Signed-off-by: Mike Pagano <mpagano <AT> gentoo.org> |
11 |
|
12 |
0000_README | 4 + |
13 |
1102_linux-4.19.103.patch | 17138 ++++++++++++++++++++++++++++++++++++++++++++ |
14 |
2 files changed, 17142 insertions(+) |
15 |
|
16 |
diff --git a/0000_README b/0000_README |
17 |
index 0d9cae4..8ac0cbe 100644 |
18 |
--- a/0000_README |
19 |
+++ b/0000_README |
20 |
@@ -447,6 +447,10 @@ Patch: 1101_linux-4.19.102.patch |
21 |
From: https://www.kernel.org |
22 |
Desc: Linux 4.19.102 |
23 |
|
24 |
+Patch: 1102_linux-4.19.103.patch |
25 |
+From: https://www.kernel.org |
26 |
+Desc: Linux 4.19.103 |
27 |
+ |
28 |
Patch: 1500_XATTR_USER_PREFIX.patch |
29 |
From: https://bugs.gentoo.org/show_bug.cgi?id=470644 |
30 |
Desc: Support for namespace user.pax.* on tmpfs. |
31 |
|
32 |
diff --git a/1102_linux-4.19.103.patch b/1102_linux-4.19.103.patch |
33 |
new file mode 100644 |
34 |
index 0000000..7b0adc1 |
35 |
--- /dev/null |
36 |
+++ b/1102_linux-4.19.103.patch |
37 |
@@ -0,0 +1,17138 @@ |
38 |
+diff --git a/Makefile b/Makefile |
39 |
+index 597a14e2127b..37f58becf5c2 100644 |
40 |
+--- a/Makefile |
41 |
++++ b/Makefile |
42 |
+@@ -1,7 +1,7 @@ |
43 |
+ # SPDX-License-Identifier: GPL-2.0 |
44 |
+ VERSION = 4 |
45 |
+ PATCHLEVEL = 19 |
46 |
+-SUBLEVEL = 102 |
47 |
++SUBLEVEL = 103 |
48 |
+ EXTRAVERSION = |
49 |
+ NAME = "People's Front" |
50 |
+ |
51 |
+diff --git a/arch/arm/include/asm/kvm_emulate.h b/arch/arm/include/asm/kvm_emulate.h |
52 |
+index 77121b713bef..7d2ca035d6c8 100644 |
53 |
+--- a/arch/arm/include/asm/kvm_emulate.h |
54 |
++++ b/arch/arm/include/asm/kvm_emulate.h |
55 |
+@@ -26,13 +26,25 @@ |
56 |
+ #include <asm/cputype.h> |
57 |
+ |
58 |
+ /* arm64 compatibility macros */ |
59 |
++#define PSR_AA32_MODE_FIQ FIQ_MODE |
60 |
++#define PSR_AA32_MODE_SVC SVC_MODE |
61 |
+ #define PSR_AA32_MODE_ABT ABT_MODE |
62 |
+ #define PSR_AA32_MODE_UND UND_MODE |
63 |
+ #define PSR_AA32_T_BIT PSR_T_BIT |
64 |
++#define PSR_AA32_F_BIT PSR_F_BIT |
65 |
+ #define PSR_AA32_I_BIT PSR_I_BIT |
66 |
+ #define PSR_AA32_A_BIT PSR_A_BIT |
67 |
+ #define PSR_AA32_E_BIT PSR_E_BIT |
68 |
+ #define PSR_AA32_IT_MASK PSR_IT_MASK |
69 |
++#define PSR_AA32_GE_MASK 0x000f0000 |
70 |
++#define PSR_AA32_DIT_BIT 0x00200000 |
71 |
++#define PSR_AA32_PAN_BIT 0x00400000 |
72 |
++#define PSR_AA32_SSBS_BIT 0x00800000 |
73 |
++#define PSR_AA32_Q_BIT PSR_Q_BIT |
74 |
++#define PSR_AA32_V_BIT PSR_V_BIT |
75 |
++#define PSR_AA32_C_BIT PSR_C_BIT |
76 |
++#define PSR_AA32_Z_BIT PSR_Z_BIT |
77 |
++#define PSR_AA32_N_BIT PSR_N_BIT |
78 |
+ |
79 |
+ unsigned long *vcpu_reg(struct kvm_vcpu *vcpu, u8 reg_num); |
80 |
+ |
81 |
+@@ -53,6 +65,11 @@ static inline void vcpu_write_spsr(struct kvm_vcpu *vcpu, unsigned long v) |
82 |
+ *__vcpu_spsr(vcpu) = v; |
83 |
+ } |
84 |
+ |
85 |
++static inline unsigned long host_spsr_to_spsr32(unsigned long spsr) |
86 |
++{ |
87 |
++ return spsr; |
88 |
++} |
89 |
++ |
90 |
+ static inline unsigned long vcpu_get_reg(struct kvm_vcpu *vcpu, |
91 |
+ u8 reg_num) |
92 |
+ { |
93 |
+@@ -189,6 +206,11 @@ static inline bool kvm_vcpu_dabt_issext(struct kvm_vcpu *vcpu) |
94 |
+ return kvm_vcpu_get_hsr(vcpu) & HSR_SSE; |
95 |
+ } |
96 |
+ |
97 |
++static inline bool kvm_vcpu_dabt_issf(const struct kvm_vcpu *vcpu) |
98 |
++{ |
99 |
++ return false; |
100 |
++} |
101 |
++ |
102 |
+ static inline int kvm_vcpu_dabt_get_rd(struct kvm_vcpu *vcpu) |
103 |
+ { |
104 |
+ return (kvm_vcpu_get_hsr(vcpu) & HSR_SRT_MASK) >> HSR_SRT_SHIFT; |
105 |
+diff --git a/arch/arm/include/asm/kvm_mmio.h b/arch/arm/include/asm/kvm_mmio.h |
106 |
+index f3a7de71f515..848339d76f9a 100644 |
107 |
+--- a/arch/arm/include/asm/kvm_mmio.h |
108 |
++++ b/arch/arm/include/asm/kvm_mmio.h |
109 |
+@@ -26,6 +26,8 @@ |
110 |
+ struct kvm_decode { |
111 |
+ unsigned long rt; |
112 |
+ bool sign_extend; |
113 |
++ /* Not used on 32-bit arm */ |
114 |
++ bool sixty_four; |
115 |
+ }; |
116 |
+ |
117 |
+ void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data); |
118 |
+diff --git a/arch/arm/mach-tegra/sleep-tegra30.S b/arch/arm/mach-tegra/sleep-tegra30.S |
119 |
+index dd4a67dabd91..b7cd41461e7d 100644 |
120 |
+--- a/arch/arm/mach-tegra/sleep-tegra30.S |
121 |
++++ b/arch/arm/mach-tegra/sleep-tegra30.S |
122 |
+@@ -382,6 +382,14 @@ _pll_m_c_x_done: |
123 |
+ pll_locked r1, r0, CLK_RESET_PLLC_BASE |
124 |
+ pll_locked r1, r0, CLK_RESET_PLLX_BASE |
125 |
+ |
126 |
++ tegra_get_soc_id TEGRA_APB_MISC_BASE, r1 |
127 |
++ cmp r1, #TEGRA30 |
128 |
++ beq 1f |
129 |
++ ldr r1, [r0, #CLK_RESET_PLLP_BASE] |
130 |
++ bic r1, r1, #(1<<31) @ disable PllP bypass |
131 |
++ str r1, [r0, #CLK_RESET_PLLP_BASE] |
132 |
++1: |
133 |
++ |
134 |
+ mov32 r7, TEGRA_TMRUS_BASE |
135 |
+ ldr r1, [r7] |
136 |
+ add r1, r1, #LOCK_DELAY |
137 |
+@@ -641,7 +649,10 @@ tegra30_switch_cpu_to_clk32k: |
138 |
+ str r0, [r4, #PMC_PLLP_WB0_OVERRIDE] |
139 |
+ |
140 |
+ /* disable PLLP, PLLA, PLLC and PLLX */ |
141 |
++ tegra_get_soc_id TEGRA_APB_MISC_BASE, r1 |
142 |
++ cmp r1, #TEGRA30 |
143 |
+ ldr r0, [r5, #CLK_RESET_PLLP_BASE] |
144 |
++ orrne r0, r0, #(1 << 31) @ enable PllP bypass on fast cluster |
145 |
+ bic r0, r0, #(1 << 30) |
146 |
+ str r0, [r5, #CLK_RESET_PLLP_BASE] |
147 |
+ ldr r0, [r5, #CLK_RESET_PLLA_BASE] |
148 |
+diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h |
149 |
+index 6106a85ae0be..778cb4f868d9 100644 |
150 |
+--- a/arch/arm64/include/asm/kvm_emulate.h |
151 |
++++ b/arch/arm64/include/asm/kvm_emulate.h |
152 |
+@@ -202,6 +202,38 @@ static inline void vcpu_write_spsr(struct kvm_vcpu *vcpu, unsigned long v) |
153 |
+ vcpu_gp_regs(vcpu)->spsr[KVM_SPSR_EL1] = v; |
154 |
+ } |
155 |
+ |
156 |
++/* |
157 |
++ * The layout of SPSR for an AArch32 state is different when observed from an |
158 |
++ * AArch64 SPSR_ELx or an AArch32 SPSR_*. This function generates the AArch32 |
159 |
++ * view given an AArch64 view. |
160 |
++ * |
161 |
++ * In ARM DDI 0487E.a see: |
162 |
++ * |
163 |
++ * - The AArch64 view (SPSR_EL2) in section C5.2.18, page C5-426 |
164 |
++ * - The AArch32 view (SPSR_abt) in section G8.2.126, page G8-6256 |
165 |
++ * - The AArch32 view (SPSR_und) in section G8.2.132, page G8-6280 |
166 |
++ * |
167 |
++ * Which show the following differences: |
168 |
++ * |
169 |
++ * | Bit | AA64 | AA32 | Notes | |
170 |
++ * +-----+------+------+-----------------------------| |
171 |
++ * | 24 | DIT | J | J is RES0 in ARMv8 | |
172 |
++ * | 21 | SS | DIT | SS doesn't exist in AArch32 | |
173 |
++ * |
174 |
++ * ... and all other bits are (currently) common. |
175 |
++ */ |
176 |
++static inline unsigned long host_spsr_to_spsr32(unsigned long spsr) |
177 |
++{ |
178 |
++ const unsigned long overlap = BIT(24) | BIT(21); |
179 |
++ unsigned long dit = !!(spsr & PSR_AA32_DIT_BIT); |
180 |
++ |
181 |
++ spsr &= ~overlap; |
182 |
++ |
183 |
++ spsr |= dit << 21; |
184 |
++ |
185 |
++ return spsr; |
186 |
++} |
187 |
++ |
188 |
+ static inline bool vcpu_mode_priv(const struct kvm_vcpu *vcpu) |
189 |
+ { |
190 |
+ u32 mode; |
191 |
+@@ -261,6 +293,11 @@ static inline bool kvm_vcpu_dabt_issext(const struct kvm_vcpu *vcpu) |
192 |
+ return !!(kvm_vcpu_get_hsr(vcpu) & ESR_ELx_SSE); |
193 |
+ } |
194 |
+ |
195 |
++static inline bool kvm_vcpu_dabt_issf(const struct kvm_vcpu *vcpu) |
196 |
++{ |
197 |
++ return !!(kvm_vcpu_get_hsr(vcpu) & ESR_ELx_SF); |
198 |
++} |
199 |
++ |
200 |
+ static inline int kvm_vcpu_dabt_get_rd(const struct kvm_vcpu *vcpu) |
201 |
+ { |
202 |
+ return (kvm_vcpu_get_hsr(vcpu) & ESR_ELx_SRT_MASK) >> ESR_ELx_SRT_SHIFT; |
203 |
+diff --git a/arch/arm64/include/asm/kvm_mmio.h b/arch/arm64/include/asm/kvm_mmio.h |
204 |
+index 75ea42079757..0240290cf764 100644 |
205 |
+--- a/arch/arm64/include/asm/kvm_mmio.h |
206 |
++++ b/arch/arm64/include/asm/kvm_mmio.h |
207 |
+@@ -21,13 +21,11 @@ |
208 |
+ #include <linux/kvm_host.h> |
209 |
+ #include <asm/kvm_arm.h> |
210 |
+ |
211 |
+-/* |
212 |
+- * This is annoying. The mmio code requires this, even if we don't |
213 |
+- * need any decoding. To be fixed. |
214 |
+- */ |
215 |
+ struct kvm_decode { |
216 |
+ unsigned long rt; |
217 |
+ bool sign_extend; |
218 |
++ /* Witdth of the register accessed by the faulting instruction is 64-bits */ |
219 |
++ bool sixty_four; |
220 |
+ }; |
221 |
+ |
222 |
+ void kvm_mmio_write_buf(void *buf, unsigned int len, unsigned long data); |
223 |
+diff --git a/arch/arm64/include/asm/ptrace.h b/arch/arm64/include/asm/ptrace.h |
224 |
+index 6bc43889d11e..163970d5c4b1 100644 |
225 |
+--- a/arch/arm64/include/asm/ptrace.h |
226 |
++++ b/arch/arm64/include/asm/ptrace.h |
227 |
+@@ -50,6 +50,7 @@ |
228 |
+ #define PSR_AA32_I_BIT 0x00000080 |
229 |
+ #define PSR_AA32_A_BIT 0x00000100 |
230 |
+ #define PSR_AA32_E_BIT 0x00000200 |
231 |
++#define PSR_AA32_PAN_BIT 0x00400000 |
232 |
+ #define PSR_AA32_SSBS_BIT 0x00800000 |
233 |
+ #define PSR_AA32_DIT_BIT 0x01000000 |
234 |
+ #define PSR_AA32_Q_BIT 0x08000000 |
235 |
+diff --git a/arch/arm64/include/uapi/asm/ptrace.h b/arch/arm64/include/uapi/asm/ptrace.h |
236 |
+index b0fd1d300154..978ff79fba2b 100644 |
237 |
+--- a/arch/arm64/include/uapi/asm/ptrace.h |
238 |
++++ b/arch/arm64/include/uapi/asm/ptrace.h |
239 |
+@@ -49,6 +49,7 @@ |
240 |
+ #define PSR_SSBS_BIT 0x00001000 |
241 |
+ #define PSR_PAN_BIT 0x00400000 |
242 |
+ #define PSR_UAO_BIT 0x00800000 |
243 |
++#define PSR_DIT_BIT 0x01000000 |
244 |
+ #define PSR_V_BIT 0x10000000 |
245 |
+ #define PSR_C_BIT 0x20000000 |
246 |
+ #define PSR_Z_BIT 0x40000000 |
247 |
+diff --git a/arch/arm64/kvm/inject_fault.c b/arch/arm64/kvm/inject_fault.c |
248 |
+index a55e91dfcf8f..41c80c311367 100644 |
249 |
+--- a/arch/arm64/kvm/inject_fault.c |
250 |
++++ b/arch/arm64/kvm/inject_fault.c |
251 |
+@@ -25,9 +25,6 @@ |
252 |
+ #include <asm/kvm_emulate.h> |
253 |
+ #include <asm/esr.h> |
254 |
+ |
255 |
+-#define PSTATE_FAULT_BITS_64 (PSR_MODE_EL1h | PSR_A_BIT | PSR_F_BIT | \ |
256 |
+- PSR_I_BIT | PSR_D_BIT) |
257 |
+- |
258 |
+ #define CURRENT_EL_SP_EL0_VECTOR 0x0 |
259 |
+ #define CURRENT_EL_SP_ELx_VECTOR 0x200 |
260 |
+ #define LOWER_EL_AArch64_VECTOR 0x400 |
261 |
+@@ -61,6 +58,69 @@ static u64 get_except_vector(struct kvm_vcpu *vcpu, enum exception_type type) |
262 |
+ return vcpu_read_sys_reg(vcpu, VBAR_EL1) + exc_offset + type; |
263 |
+ } |
264 |
+ |
265 |
++/* |
266 |
++ * When an exception is taken, most PSTATE fields are left unchanged in the |
267 |
++ * handler. However, some are explicitly overridden (e.g. M[4:0]). Luckily all |
268 |
++ * of the inherited bits have the same position in the AArch64/AArch32 SPSR_ELx |
269 |
++ * layouts, so we don't need to shuffle these for exceptions from AArch32 EL0. |
270 |
++ * |
271 |
++ * For the SPSR_ELx layout for AArch64, see ARM DDI 0487E.a page C5-429. |
272 |
++ * For the SPSR_ELx layout for AArch32, see ARM DDI 0487E.a page C5-426. |
273 |
++ * |
274 |
++ * Here we manipulate the fields in order of the AArch64 SPSR_ELx layout, from |
275 |
++ * MSB to LSB. |
276 |
++ */ |
277 |
++static unsigned long get_except64_pstate(struct kvm_vcpu *vcpu) |
278 |
++{ |
279 |
++ unsigned long sctlr = vcpu_read_sys_reg(vcpu, SCTLR_EL1); |
280 |
++ unsigned long old, new; |
281 |
++ |
282 |
++ old = *vcpu_cpsr(vcpu); |
283 |
++ new = 0; |
284 |
++ |
285 |
++ new |= (old & PSR_N_BIT); |
286 |
++ new |= (old & PSR_Z_BIT); |
287 |
++ new |= (old & PSR_C_BIT); |
288 |
++ new |= (old & PSR_V_BIT); |
289 |
++ |
290 |
++ // TODO: TCO (if/when ARMv8.5-MemTag is exposed to guests) |
291 |
++ |
292 |
++ new |= (old & PSR_DIT_BIT); |
293 |
++ |
294 |
++ // PSTATE.UAO is set to zero upon any exception to AArch64 |
295 |
++ // See ARM DDI 0487E.a, page D5-2579. |
296 |
++ |
297 |
++ // PSTATE.PAN is unchanged unless SCTLR_ELx.SPAN == 0b0 |
298 |
++ // SCTLR_ELx.SPAN is RES1 when ARMv8.1-PAN is not implemented |
299 |
++ // See ARM DDI 0487E.a, page D5-2578. |
300 |
++ new |= (old & PSR_PAN_BIT); |
301 |
++ if (!(sctlr & SCTLR_EL1_SPAN)) |
302 |
++ new |= PSR_PAN_BIT; |
303 |
++ |
304 |
++ // PSTATE.SS is set to zero upon any exception to AArch64 |
305 |
++ // See ARM DDI 0487E.a, page D2-2452. |
306 |
++ |
307 |
++ // PSTATE.IL is set to zero upon any exception to AArch64 |
308 |
++ // See ARM DDI 0487E.a, page D1-2306. |
309 |
++ |
310 |
++ // PSTATE.SSBS is set to SCTLR_ELx.DSSBS upon any exception to AArch64 |
311 |
++ // See ARM DDI 0487E.a, page D13-3258 |
312 |
++ if (sctlr & SCTLR_ELx_DSSBS) |
313 |
++ new |= PSR_SSBS_BIT; |
314 |
++ |
315 |
++ // PSTATE.BTYPE is set to zero upon any exception to AArch64 |
316 |
++ // See ARM DDI 0487E.a, pages D1-2293 to D1-2294. |
317 |
++ |
318 |
++ new |= PSR_D_BIT; |
319 |
++ new |= PSR_A_BIT; |
320 |
++ new |= PSR_I_BIT; |
321 |
++ new |= PSR_F_BIT; |
322 |
++ |
323 |
++ new |= PSR_MODE_EL1h; |
324 |
++ |
325 |
++ return new; |
326 |
++} |
327 |
++ |
328 |
+ static void inject_abt64(struct kvm_vcpu *vcpu, bool is_iabt, unsigned long addr) |
329 |
+ { |
330 |
+ unsigned long cpsr = *vcpu_cpsr(vcpu); |
331 |
+@@ -70,7 +130,7 @@ static void inject_abt64(struct kvm_vcpu *vcpu, bool is_iabt, unsigned long addr |
332 |
+ vcpu_write_elr_el1(vcpu, *vcpu_pc(vcpu)); |
333 |
+ *vcpu_pc(vcpu) = get_except_vector(vcpu, except_type_sync); |
334 |
+ |
335 |
+- *vcpu_cpsr(vcpu) = PSTATE_FAULT_BITS_64; |
336 |
++ *vcpu_cpsr(vcpu) = get_except64_pstate(vcpu); |
337 |
+ vcpu_write_spsr(vcpu, cpsr); |
338 |
+ |
339 |
+ vcpu_write_sys_reg(vcpu, addr, FAR_EL1); |
340 |
+@@ -105,7 +165,7 @@ static void inject_undef64(struct kvm_vcpu *vcpu) |
341 |
+ vcpu_write_elr_el1(vcpu, *vcpu_pc(vcpu)); |
342 |
+ *vcpu_pc(vcpu) = get_except_vector(vcpu, except_type_sync); |
343 |
+ |
344 |
+- *vcpu_cpsr(vcpu) = PSTATE_FAULT_BITS_64; |
345 |
++ *vcpu_cpsr(vcpu) = get_except64_pstate(vcpu); |
346 |
+ vcpu_write_spsr(vcpu, cpsr); |
347 |
+ |
348 |
+ /* |
349 |
+diff --git a/arch/mips/Makefile.postlink b/arch/mips/Makefile.postlink |
350 |
+index 4eea4188cb20..13e0beb9eee3 100644 |
351 |
+--- a/arch/mips/Makefile.postlink |
352 |
++++ b/arch/mips/Makefile.postlink |
353 |
+@@ -12,7 +12,7 @@ __archpost: |
354 |
+ include scripts/Kbuild.include |
355 |
+ |
356 |
+ CMD_RELOCS = arch/mips/boot/tools/relocs |
357 |
+-quiet_cmd_relocs = RELOCS $@ |
358 |
++quiet_cmd_relocs = RELOCS $@ |
359 |
+ cmd_relocs = $(CMD_RELOCS) $@ |
360 |
+ |
361 |
+ # `@true` prevents complaint when there is nothing to be done |
362 |
+diff --git a/arch/mips/boot/Makefile b/arch/mips/boot/Makefile |
363 |
+index 35704c28a28b..0ccc20320099 100644 |
364 |
+--- a/arch/mips/boot/Makefile |
365 |
++++ b/arch/mips/boot/Makefile |
366 |
+@@ -123,7 +123,7 @@ $(obj)/vmlinux.its.S: $(addprefix $(srctree)/arch/mips/$(PLATFORM)/,$(ITS_INPUTS |
367 |
+ targets += vmlinux.its |
368 |
+ targets += vmlinux.gz.its |
369 |
+ targets += vmlinux.bz2.its |
370 |
+-targets += vmlinux.lzmo.its |
371 |
++targets += vmlinux.lzma.its |
372 |
+ targets += vmlinux.lzo.its |
373 |
+ |
374 |
+ quiet_cmd_cpp_its_S = ITS $@ |
375 |
+diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig |
376 |
+index a80669209155..6f475dc5829b 100644 |
377 |
+--- a/arch/powerpc/Kconfig |
378 |
++++ b/arch/powerpc/Kconfig |
379 |
+@@ -230,6 +230,7 @@ config PPC |
380 |
+ select NEED_SG_DMA_LENGTH |
381 |
+ select NO_BOOTMEM |
382 |
+ select OF |
383 |
++ select OF_DMA_DEFAULT_COHERENT if !NOT_COHERENT_CACHE |
384 |
+ select OF_EARLY_FLATTREE |
385 |
+ select OF_RESERVED_MEM |
386 |
+ select OLD_SIGACTION if PPC32 |
387 |
+diff --git a/arch/powerpc/boot/4xx.c b/arch/powerpc/boot/4xx.c |
388 |
+index f7da65169124..3c8774163c7e 100644 |
389 |
+--- a/arch/powerpc/boot/4xx.c |
390 |
++++ b/arch/powerpc/boot/4xx.c |
391 |
+@@ -232,7 +232,7 @@ void ibm4xx_denali_fixup_memsize(void) |
392 |
+ dpath = 8; /* 64 bits */ |
393 |
+ |
394 |
+ /* get address pins (rows) */ |
395 |
+- val = SDRAM0_READ(DDR0_42); |
396 |
++ val = SDRAM0_READ(DDR0_42); |
397 |
+ |
398 |
+ row = DDR_GET_VAL(val, DDR_APIN, DDR_APIN_SHIFT); |
399 |
+ if (row > max_row) |
400 |
+diff --git a/arch/powerpc/kvm/book3s_hv.c b/arch/powerpc/kvm/book3s_hv.c |
401 |
+index dbfe32327212..5dc592fb4f5f 100644 |
402 |
+--- a/arch/powerpc/kvm/book3s_hv.c |
403 |
++++ b/arch/powerpc/kvm/book3s_hv.c |
404 |
+@@ -2065,7 +2065,7 @@ static struct kvm_vcpu *kvmppc_core_vcpu_create_hv(struct kvm *kvm, |
405 |
+ mutex_unlock(&kvm->lock); |
406 |
+ |
407 |
+ if (!vcore) |
408 |
+- goto free_vcpu; |
409 |
++ goto uninit_vcpu; |
410 |
+ |
411 |
+ spin_lock(&vcore->lock); |
412 |
+ ++vcore->num_threads; |
413 |
+@@ -2082,6 +2082,8 @@ static struct kvm_vcpu *kvmppc_core_vcpu_create_hv(struct kvm *kvm, |
414 |
+ |
415 |
+ return vcpu; |
416 |
+ |
417 |
++uninit_vcpu: |
418 |
++ kvm_vcpu_uninit(vcpu); |
419 |
+ free_vcpu: |
420 |
+ kmem_cache_free(kvm_vcpu_cache, vcpu); |
421 |
+ out: |
422 |
+diff --git a/arch/powerpc/kvm/book3s_pr.c b/arch/powerpc/kvm/book3s_pr.c |
423 |
+index de9702219dee..7869112a8f3c 100644 |
424 |
+--- a/arch/powerpc/kvm/book3s_pr.c |
425 |
++++ b/arch/powerpc/kvm/book3s_pr.c |
426 |
+@@ -1772,10 +1772,12 @@ static struct kvm_vcpu *kvmppc_core_vcpu_create_pr(struct kvm *kvm, |
427 |
+ |
428 |
+ err = kvmppc_mmu_init(vcpu); |
429 |
+ if (err < 0) |
430 |
+- goto uninit_vcpu; |
431 |
++ goto free_shared_page; |
432 |
+ |
433 |
+ return vcpu; |
434 |
+ |
435 |
++free_shared_page: |
436 |
++ free_page((unsigned long)vcpu->arch.shared); |
437 |
+ uninit_vcpu: |
438 |
+ kvm_vcpu_uninit(vcpu); |
439 |
+ free_shadow_vcpu: |
440 |
+diff --git a/arch/powerpc/platforms/pseries/hotplug-memory.c b/arch/powerpc/platforms/pseries/hotplug-memory.c |
441 |
+index c2c6f32848e1..fc01a2c0f8ed 100644 |
442 |
+--- a/arch/powerpc/platforms/pseries/hotplug-memory.c |
443 |
++++ b/arch/powerpc/platforms/pseries/hotplug-memory.c |
444 |
+@@ -366,8 +366,10 @@ static bool lmb_is_removable(struct drmem_lmb *lmb) |
445 |
+ |
446 |
+ for (i = 0; i < scns_per_block; i++) { |
447 |
+ pfn = PFN_DOWN(phys_addr); |
448 |
+- if (!pfn_present(pfn)) |
449 |
++ if (!pfn_present(pfn)) { |
450 |
++ phys_addr += MIN_MEMORY_BLOCK_SIZE; |
451 |
+ continue; |
452 |
++ } |
453 |
+ |
454 |
+ rc &= is_mem_section_removable(pfn, PAGES_PER_SECTION); |
455 |
+ phys_addr += MIN_MEMORY_BLOCK_SIZE; |
456 |
+diff --git a/arch/powerpc/xmon/xmon.c b/arch/powerpc/xmon/xmon.c |
457 |
+index f0fa22e7d36c..3291e5fb94bc 100644 |
458 |
+--- a/arch/powerpc/xmon/xmon.c |
459 |
++++ b/arch/powerpc/xmon/xmon.c |
460 |
+@@ -1878,15 +1878,14 @@ static void dump_300_sprs(void) |
461 |
+ |
462 |
+ printf("pidr = %.16lx tidr = %.16lx\n", |
463 |
+ mfspr(SPRN_PID), mfspr(SPRN_TIDR)); |
464 |
+- printf("asdr = %.16lx psscr = %.16lx\n", |
465 |
+- mfspr(SPRN_ASDR), hv ? mfspr(SPRN_PSSCR) |
466 |
+- : mfspr(SPRN_PSSCR_PR)); |
467 |
++ printf("psscr = %.16lx\n", |
468 |
++ hv ? mfspr(SPRN_PSSCR) : mfspr(SPRN_PSSCR_PR)); |
469 |
+ |
470 |
+ if (!hv) |
471 |
+ return; |
472 |
+ |
473 |
+- printf("ptcr = %.16lx\n", |
474 |
+- mfspr(SPRN_PTCR)); |
475 |
++ printf("ptcr = %.16lx asdr = %.16lx\n", |
476 |
++ mfspr(SPRN_PTCR), mfspr(SPRN_ASDR)); |
477 |
+ #endif |
478 |
+ } |
479 |
+ |
480 |
+diff --git a/arch/s390/include/asm/page.h b/arch/s390/include/asm/page.h |
481 |
+index 0d753291c43c..ac3c86b21d79 100644 |
482 |
+--- a/arch/s390/include/asm/page.h |
483 |
++++ b/arch/s390/include/asm/page.h |
484 |
+@@ -33,6 +33,8 @@ |
485 |
+ #define ARCH_HAS_PREPARE_HUGEPAGE |
486 |
+ #define ARCH_HAS_HUGEPAGE_CLEAR_FLUSH |
487 |
+ |
488 |
++#define HAVE_ARCH_HUGETLB_UNMAPPED_AREA |
489 |
++ |
490 |
+ #include <asm/setup.h> |
491 |
+ #ifndef __ASSEMBLY__ |
492 |
+ |
493 |
+diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c |
494 |
+index db3196aebaa1..11c3cd906ab4 100644 |
495 |
+--- a/arch/s390/kvm/kvm-s390.c |
496 |
++++ b/arch/s390/kvm/kvm-s390.c |
497 |
+@@ -2564,9 +2564,7 @@ static void kvm_s390_vcpu_initial_reset(struct kvm_vcpu *vcpu) |
498 |
+ vcpu->arch.sie_block->gcr[14] = CR14_UNUSED_32 | |
499 |
+ CR14_UNUSED_33 | |
500 |
+ CR14_EXTERNAL_DAMAGE_SUBMASK; |
501 |
+- /* make sure the new fpc will be lazily loaded */ |
502 |
+- save_fpu_regs(); |
503 |
+- current->thread.fpu.fpc = 0; |
504 |
++ vcpu->run->s.regs.fpc = 0; |
505 |
+ vcpu->arch.sie_block->gbea = 1; |
506 |
+ vcpu->arch.sie_block->pp = 0; |
507 |
+ vcpu->arch.sie_block->fpf &= ~FPF_BPBC; |
508 |
+@@ -3994,7 +3992,7 @@ long kvm_arch_vcpu_ioctl(struct file *filp, |
509 |
+ switch (ioctl) { |
510 |
+ case KVM_S390_STORE_STATUS: |
511 |
+ idx = srcu_read_lock(&vcpu->kvm->srcu); |
512 |
+- r = kvm_s390_vcpu_store_status(vcpu, arg); |
513 |
++ r = kvm_s390_store_status_unloaded(vcpu, arg); |
514 |
+ srcu_read_unlock(&vcpu->kvm->srcu, idx); |
515 |
+ break; |
516 |
+ case KVM_S390_SET_INITIAL_PSW: { |
517 |
+diff --git a/arch/s390/mm/hugetlbpage.c b/arch/s390/mm/hugetlbpage.c |
518 |
+index b0246c705a19..5674710a4841 100644 |
519 |
+--- a/arch/s390/mm/hugetlbpage.c |
520 |
++++ b/arch/s390/mm/hugetlbpage.c |
521 |
+@@ -2,7 +2,7 @@ |
522 |
+ /* |
523 |
+ * IBM System z Huge TLB Page Support for Kernel. |
524 |
+ * |
525 |
+- * Copyright IBM Corp. 2007,2016 |
526 |
++ * Copyright IBM Corp. 2007,2020 |
527 |
+ * Author(s): Gerald Schaefer <gerald.schaefer@××××××.com> |
528 |
+ */ |
529 |
+ |
530 |
+@@ -11,6 +11,9 @@ |
531 |
+ |
532 |
+ #include <linux/mm.h> |
533 |
+ #include <linux/hugetlb.h> |
534 |
++#include <linux/mman.h> |
535 |
++#include <linux/sched/mm.h> |
536 |
++#include <linux/security.h> |
537 |
+ |
538 |
+ /* |
539 |
+ * If the bit selected by single-bit bitmask "a" is set within "x", move |
540 |
+@@ -267,3 +270,98 @@ static __init int setup_hugepagesz(char *opt) |
541 |
+ return 1; |
542 |
+ } |
543 |
+ __setup("hugepagesz=", setup_hugepagesz); |
544 |
++ |
545 |
++static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *file, |
546 |
++ unsigned long addr, unsigned long len, |
547 |
++ unsigned long pgoff, unsigned long flags) |
548 |
++{ |
549 |
++ struct hstate *h = hstate_file(file); |
550 |
++ struct vm_unmapped_area_info info; |
551 |
++ |
552 |
++ info.flags = 0; |
553 |
++ info.length = len; |
554 |
++ info.low_limit = current->mm->mmap_base; |
555 |
++ info.high_limit = TASK_SIZE; |
556 |
++ info.align_mask = PAGE_MASK & ~huge_page_mask(h); |
557 |
++ info.align_offset = 0; |
558 |
++ return vm_unmapped_area(&info); |
559 |
++} |
560 |
++ |
561 |
++static unsigned long hugetlb_get_unmapped_area_topdown(struct file *file, |
562 |
++ unsigned long addr0, unsigned long len, |
563 |
++ unsigned long pgoff, unsigned long flags) |
564 |
++{ |
565 |
++ struct hstate *h = hstate_file(file); |
566 |
++ struct vm_unmapped_area_info info; |
567 |
++ unsigned long addr; |
568 |
++ |
569 |
++ info.flags = VM_UNMAPPED_AREA_TOPDOWN; |
570 |
++ info.length = len; |
571 |
++ info.low_limit = max(PAGE_SIZE, mmap_min_addr); |
572 |
++ info.high_limit = current->mm->mmap_base; |
573 |
++ info.align_mask = PAGE_MASK & ~huge_page_mask(h); |
574 |
++ info.align_offset = 0; |
575 |
++ addr = vm_unmapped_area(&info); |
576 |
++ |
577 |
++ /* |
578 |
++ * A failed mmap() very likely causes application failure, |
579 |
++ * so fall back to the bottom-up function here. This scenario |
580 |
++ * can happen with large stack limits and large mmap() |
581 |
++ * allocations. |
582 |
++ */ |
583 |
++ if (addr & ~PAGE_MASK) { |
584 |
++ VM_BUG_ON(addr != -ENOMEM); |
585 |
++ info.flags = 0; |
586 |
++ info.low_limit = TASK_UNMAPPED_BASE; |
587 |
++ info.high_limit = TASK_SIZE; |
588 |
++ addr = vm_unmapped_area(&info); |
589 |
++ } |
590 |
++ |
591 |
++ return addr; |
592 |
++} |
593 |
++ |
594 |
++unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, |
595 |
++ unsigned long len, unsigned long pgoff, unsigned long flags) |
596 |
++{ |
597 |
++ struct hstate *h = hstate_file(file); |
598 |
++ struct mm_struct *mm = current->mm; |
599 |
++ struct vm_area_struct *vma; |
600 |
++ int rc; |
601 |
++ |
602 |
++ if (len & ~huge_page_mask(h)) |
603 |
++ return -EINVAL; |
604 |
++ if (len > TASK_SIZE - mmap_min_addr) |
605 |
++ return -ENOMEM; |
606 |
++ |
607 |
++ if (flags & MAP_FIXED) { |
608 |
++ if (prepare_hugepage_range(file, addr, len)) |
609 |
++ return -EINVAL; |
610 |
++ goto check_asce_limit; |
611 |
++ } |
612 |
++ |
613 |
++ if (addr) { |
614 |
++ addr = ALIGN(addr, huge_page_size(h)); |
615 |
++ vma = find_vma(mm, addr); |
616 |
++ if (TASK_SIZE - len >= addr && addr >= mmap_min_addr && |
617 |
++ (!vma || addr + len <= vm_start_gap(vma))) |
618 |
++ goto check_asce_limit; |
619 |
++ } |
620 |
++ |
621 |
++ if (mm->get_unmapped_area == arch_get_unmapped_area) |
622 |
++ addr = hugetlb_get_unmapped_area_bottomup(file, addr, len, |
623 |
++ pgoff, flags); |
624 |
++ else |
625 |
++ addr = hugetlb_get_unmapped_area_topdown(file, addr, len, |
626 |
++ pgoff, flags); |
627 |
++ if (addr & ~PAGE_MASK) |
628 |
++ return addr; |
629 |
++ |
630 |
++check_asce_limit: |
631 |
++ if (addr + len > current->mm->context.asce_limit && |
632 |
++ addr + len <= TASK_SIZE) { |
633 |
++ rc = crst_table_upgrade(mm, addr + len); |
634 |
++ if (rc) |
635 |
++ return (unsigned long) rc; |
636 |
++ } |
637 |
++ return addr; |
638 |
++} |
639 |
+diff --git a/arch/sparc/include/uapi/asm/ipcbuf.h b/arch/sparc/include/uapi/asm/ipcbuf.h |
640 |
+index 9d0d125500e2..084b8949ddff 100644 |
641 |
+--- a/arch/sparc/include/uapi/asm/ipcbuf.h |
642 |
++++ b/arch/sparc/include/uapi/asm/ipcbuf.h |
643 |
+@@ -15,19 +15,19 @@ |
644 |
+ |
645 |
+ struct ipc64_perm |
646 |
+ { |
647 |
+- __kernel_key_t key; |
648 |
+- __kernel_uid_t uid; |
649 |
+- __kernel_gid_t gid; |
650 |
+- __kernel_uid_t cuid; |
651 |
+- __kernel_gid_t cgid; |
652 |
++ __kernel_key_t key; |
653 |
++ __kernel_uid32_t uid; |
654 |
++ __kernel_gid32_t gid; |
655 |
++ __kernel_uid32_t cuid; |
656 |
++ __kernel_gid32_t cgid; |
657 |
+ #ifndef __arch64__ |
658 |
+- unsigned short __pad0; |
659 |
++ unsigned short __pad0; |
660 |
+ #endif |
661 |
+- __kernel_mode_t mode; |
662 |
+- unsigned short __pad1; |
663 |
+- unsigned short seq; |
664 |
+- unsigned long long __unused1; |
665 |
+- unsigned long long __unused2; |
666 |
++ __kernel_mode_t mode; |
667 |
++ unsigned short __pad1; |
668 |
++ unsigned short seq; |
669 |
++ unsigned long long __unused1; |
670 |
++ unsigned long long __unused2; |
671 |
+ }; |
672 |
+ |
673 |
+ #endif /* __SPARC_IPCBUF_H */ |
674 |
+diff --git a/arch/x86/include/asm/apic.h b/arch/x86/include/asm/apic.h |
675 |
+index 050368db9d35..3c1e51ead072 100644 |
676 |
+--- a/arch/x86/include/asm/apic.h |
677 |
++++ b/arch/x86/include/asm/apic.h |
678 |
+@@ -448,6 +448,14 @@ static inline void ack_APIC_irq(void) |
679 |
+ apic_eoi(); |
680 |
+ } |
681 |
+ |
682 |
++ |
683 |
++static inline bool lapic_vector_set_in_irr(unsigned int vector) |
684 |
++{ |
685 |
++ u32 irr = apic_read(APIC_IRR + (vector / 32 * 0x10)); |
686 |
++ |
687 |
++ return !!(irr & (1U << (vector % 32))); |
688 |
++} |
689 |
++ |
690 |
+ static inline unsigned default_get_apic_id(unsigned long x) |
691 |
+ { |
692 |
+ unsigned int ver = GET_APIC_VERSION(apic_read(APIC_LVR)); |
693 |
+diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h |
694 |
+index 155be8adb934..21a58fcc3dd4 100644 |
695 |
+--- a/arch/x86/include/asm/kvm_host.h |
696 |
++++ b/arch/x86/include/asm/kvm_host.h |
697 |
+@@ -350,12 +350,12 @@ struct kvm_mmu { |
698 |
+ void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long root); |
699 |
+ unsigned long (*get_cr3)(struct kvm_vcpu *vcpu); |
700 |
+ u64 (*get_pdptr)(struct kvm_vcpu *vcpu, int index); |
701 |
+- int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err, |
702 |
++ int (*page_fault)(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 err, |
703 |
+ bool prefault); |
704 |
+ void (*inject_page_fault)(struct kvm_vcpu *vcpu, |
705 |
+ struct x86_exception *fault); |
706 |
+- gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva, u32 access, |
707 |
+- struct x86_exception *exception); |
708 |
++ gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t gva_or_gpa, |
709 |
++ u32 access, struct x86_exception *exception); |
710 |
+ gpa_t (*translate_gpa)(struct kvm_vcpu *vcpu, gpa_t gpa, u32 access, |
711 |
+ struct x86_exception *exception); |
712 |
+ int (*sync_page)(struct kvm_vcpu *vcpu, |
713 |
+@@ -1354,7 +1354,7 @@ void kvm_vcpu_deactivate_apicv(struct kvm_vcpu *vcpu); |
714 |
+ |
715 |
+ int kvm_emulate_hypercall(struct kvm_vcpu *vcpu); |
716 |
+ |
717 |
+-int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u64 error_code, |
718 |
++int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code, |
719 |
+ void *insn, int insn_len); |
720 |
+ void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva); |
721 |
+ void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid); |
722 |
+diff --git a/arch/x86/kernel/apic/msi.c b/arch/x86/kernel/apic/msi.c |
723 |
+index 72a94401f9e0..1f5df339e48f 100644 |
724 |
+--- a/arch/x86/kernel/apic/msi.c |
725 |
++++ b/arch/x86/kernel/apic/msi.c |
726 |
+@@ -26,10 +26,8 @@ |
727 |
+ |
728 |
+ static struct irq_domain *msi_default_domain; |
729 |
+ |
730 |
+-static void irq_msi_compose_msg(struct irq_data *data, struct msi_msg *msg) |
731 |
++static void __irq_msi_compose_msg(struct irq_cfg *cfg, struct msi_msg *msg) |
732 |
+ { |
733 |
+- struct irq_cfg *cfg = irqd_cfg(data); |
734 |
+- |
735 |
+ msg->address_hi = MSI_ADDR_BASE_HI; |
736 |
+ |
737 |
+ if (x2apic_enabled()) |
738 |
+@@ -50,6 +48,127 @@ static void irq_msi_compose_msg(struct irq_data *data, struct msi_msg *msg) |
739 |
+ MSI_DATA_VECTOR(cfg->vector); |
740 |
+ } |
741 |
+ |
742 |
++static void irq_msi_compose_msg(struct irq_data *data, struct msi_msg *msg) |
743 |
++{ |
744 |
++ __irq_msi_compose_msg(irqd_cfg(data), msg); |
745 |
++} |
746 |
++ |
747 |
++static void irq_msi_update_msg(struct irq_data *irqd, struct irq_cfg *cfg) |
748 |
++{ |
749 |
++ struct msi_msg msg[2] = { [1] = { }, }; |
750 |
++ |
751 |
++ __irq_msi_compose_msg(cfg, msg); |
752 |
++ irq_data_get_irq_chip(irqd)->irq_write_msi_msg(irqd, msg); |
753 |
++} |
754 |
++ |
755 |
++static int |
756 |
++msi_set_affinity(struct irq_data *irqd, const struct cpumask *mask, bool force) |
757 |
++{ |
758 |
++ struct irq_cfg old_cfg, *cfg = irqd_cfg(irqd); |
759 |
++ struct irq_data *parent = irqd->parent_data; |
760 |
++ unsigned int cpu; |
761 |
++ int ret; |
762 |
++ |
763 |
++ /* Save the current configuration */ |
764 |
++ cpu = cpumask_first(irq_data_get_effective_affinity_mask(irqd)); |
765 |
++ old_cfg = *cfg; |
766 |
++ |
767 |
++ /* Allocate a new target vector */ |
768 |
++ ret = parent->chip->irq_set_affinity(parent, mask, force); |
769 |
++ if (ret < 0 || ret == IRQ_SET_MASK_OK_DONE) |
770 |
++ return ret; |
771 |
++ |
772 |
++ /* |
773 |
++ * For non-maskable and non-remapped MSI interrupts the migration |
774 |
++ * to a different destination CPU and a different vector has to be |
775 |
++ * done careful to handle the possible stray interrupt which can be |
776 |
++ * caused by the non-atomic update of the address/data pair. |
777 |
++ * |
778 |
++ * Direct update is possible when: |
779 |
++ * - The MSI is maskable (remapped MSI does not use this code path)). |
780 |
++ * The quirk bit is not set in this case. |
781 |
++ * - The new vector is the same as the old vector |
782 |
++ * - The old vector is MANAGED_IRQ_SHUTDOWN_VECTOR (interrupt starts up) |
783 |
++ * - The new destination CPU is the same as the old destination CPU |
784 |
++ */ |
785 |
++ if (!irqd_msi_nomask_quirk(irqd) || |
786 |
++ cfg->vector == old_cfg.vector || |
787 |
++ old_cfg.vector == MANAGED_IRQ_SHUTDOWN_VECTOR || |
788 |
++ cfg->dest_apicid == old_cfg.dest_apicid) { |
789 |
++ irq_msi_update_msg(irqd, cfg); |
790 |
++ return ret; |
791 |
++ } |
792 |
++ |
793 |
++ /* |
794 |
++ * Paranoia: Validate that the interrupt target is the local |
795 |
++ * CPU. |
796 |
++ */ |
797 |
++ if (WARN_ON_ONCE(cpu != smp_processor_id())) { |
798 |
++ irq_msi_update_msg(irqd, cfg); |
799 |
++ return ret; |
800 |
++ } |
801 |
++ |
802 |
++ /* |
803 |
++ * Redirect the interrupt to the new vector on the current CPU |
804 |
++ * first. This might cause a spurious interrupt on this vector if |
805 |
++ * the device raises an interrupt right between this update and the |
806 |
++ * update to the final destination CPU. |
807 |
++ * |
808 |
++ * If the vector is in use then the installed device handler will |
809 |
++ * denote it as spurious which is no harm as this is a rare event |
810 |
++ * and interrupt handlers have to cope with spurious interrupts |
811 |
++ * anyway. If the vector is unused, then it is marked so it won't |
812 |
++ * trigger the 'No irq handler for vector' warning in do_IRQ(). |
813 |
++ * |
814 |
++ * This requires to hold vector lock to prevent concurrent updates to |
815 |
++ * the affected vector. |
816 |
++ */ |
817 |
++ lock_vector_lock(); |
818 |
++ |
819 |
++ /* |
820 |
++ * Mark the new target vector on the local CPU if it is currently |
821 |
++ * unused. Reuse the VECTOR_RETRIGGERED state which is also used in |
822 |
++ * the CPU hotplug path for a similar purpose. This cannot be |
823 |
++ * undone here as the current CPU has interrupts disabled and |
824 |
++ * cannot handle the interrupt before the whole set_affinity() |
825 |
++ * section is done. In the CPU unplug case, the current CPU is |
826 |
++ * about to vanish and will not handle any interrupts anymore. The |
827 |
++ * vector is cleaned up when the CPU comes online again. |
828 |
++ */ |
829 |
++ if (IS_ERR_OR_NULL(this_cpu_read(vector_irq[cfg->vector]))) |
830 |
++ this_cpu_write(vector_irq[cfg->vector], VECTOR_RETRIGGERED); |
831 |
++ |
832 |
++ /* Redirect it to the new vector on the local CPU temporarily */ |
833 |
++ old_cfg.vector = cfg->vector; |
834 |
++ irq_msi_update_msg(irqd, &old_cfg); |
835 |
++ |
836 |
++ /* Now transition it to the target CPU */ |
837 |
++ irq_msi_update_msg(irqd, cfg); |
838 |
++ |
839 |
++ /* |
840 |
++ * All interrupts after this point are now targeted at the new |
841 |
++ * vector/CPU. |
842 |
++ * |
843 |
++ * Drop vector lock before testing whether the temporary assignment |
844 |
++ * to the local CPU was hit by an interrupt raised in the device, |
845 |
++ * because the retrigger function acquires vector lock again. |
846 |
++ */ |
847 |
++ unlock_vector_lock(); |
848 |
++ |
849 |
++ /* |
850 |
++ * Check whether the transition raced with a device interrupt and |
851 |
++ * is pending in the local APICs IRR. It is safe to do this outside |
852 |
++ * of vector lock as the irq_desc::lock of this interrupt is still |
853 |
++ * held and interrupts are disabled: The check is not accessing the |
854 |
++ * underlying vector store. It's just checking the local APIC's |
855 |
++ * IRR. |
856 |
++ */ |
857 |
++ if (lapic_vector_set_in_irr(cfg->vector)) |
858 |
++ irq_data_get_irq_chip(irqd)->irq_retrigger(irqd); |
859 |
++ |
860 |
++ return ret; |
861 |
++} |
862 |
++ |
863 |
+ /* |
864 |
+ * IRQ Chip for MSI PCI/PCI-X/PCI-Express Devices, |
865 |
+ * which implement the MSI or MSI-X Capability Structure. |
866 |
+@@ -61,6 +180,7 @@ static struct irq_chip pci_msi_controller = { |
867 |
+ .irq_ack = irq_chip_ack_parent, |
868 |
+ .irq_retrigger = irq_chip_retrigger_hierarchy, |
869 |
+ .irq_compose_msi_msg = irq_msi_compose_msg, |
870 |
++ .irq_set_affinity = msi_set_affinity, |
871 |
+ .flags = IRQCHIP_SKIP_SET_WAKE, |
872 |
+ }; |
873 |
+ |
874 |
+@@ -149,6 +269,8 @@ void __init arch_init_msi_domain(struct irq_domain *parent) |
875 |
+ } |
876 |
+ if (!msi_default_domain) |
877 |
+ pr_warn("failed to initialize irqdomain for MSI/MSI-x.\n"); |
878 |
++ else |
879 |
++ msi_default_domain->flags |= IRQ_DOMAIN_MSI_NOMASK_QUIRK; |
880 |
+ } |
881 |
+ |
882 |
+ #ifdef CONFIG_IRQ_REMAP |
883 |
+diff --git a/arch/x86/kernel/cpu/tsx.c b/arch/x86/kernel/cpu/tsx.c |
884 |
+index 3e20d322bc98..032509adf9de 100644 |
885 |
+--- a/arch/x86/kernel/cpu/tsx.c |
886 |
++++ b/arch/x86/kernel/cpu/tsx.c |
887 |
+@@ -115,11 +115,12 @@ void __init tsx_init(void) |
888 |
+ tsx_disable(); |
889 |
+ |
890 |
+ /* |
891 |
+- * tsx_disable() will change the state of the |
892 |
+- * RTM CPUID bit. Clear it here since it is now |
893 |
+- * expected to be not set. |
894 |
++ * tsx_disable() will change the state of the RTM and HLE CPUID |
895 |
++ * bits. Clear them here since they are now expected to be not |
896 |
++ * set. |
897 |
+ */ |
898 |
+ setup_clear_cpu_cap(X86_FEATURE_RTM); |
899 |
++ setup_clear_cpu_cap(X86_FEATURE_HLE); |
900 |
+ } else if (tsx_ctrl_state == TSX_CTRL_ENABLE) { |
901 |
+ |
902 |
+ /* |
903 |
+@@ -131,10 +132,10 @@ void __init tsx_init(void) |
904 |
+ tsx_enable(); |
905 |
+ |
906 |
+ /* |
907 |
+- * tsx_enable() will change the state of the |
908 |
+- * RTM CPUID bit. Force it here since it is now |
909 |
+- * expected to be set. |
910 |
++ * tsx_enable() will change the state of the RTM and HLE CPUID |
911 |
++ * bits. Force them here since they are now expected to be set. |
912 |
+ */ |
913 |
+ setup_force_cpu_cap(X86_FEATURE_RTM); |
914 |
++ setup_force_cpu_cap(X86_FEATURE_HLE); |
915 |
+ } |
916 |
+ } |
917 |
+diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c |
918 |
+index e699f4d2a450..c91431bc476e 100644 |
919 |
+--- a/arch/x86/kvm/emulate.c |
920 |
++++ b/arch/x86/kvm/emulate.c |
921 |
+@@ -5164,16 +5164,28 @@ int x86_decode_insn(struct x86_emulate_ctxt *ctxt, void *insn, int insn_len) |
922 |
+ ctxt->ad_bytes = def_ad_bytes ^ 6; |
923 |
+ break; |
924 |
+ case 0x26: /* ES override */ |
925 |
++ has_seg_override = true; |
926 |
++ ctxt->seg_override = VCPU_SREG_ES; |
927 |
++ break; |
928 |
+ case 0x2e: /* CS override */ |
929 |
++ has_seg_override = true; |
930 |
++ ctxt->seg_override = VCPU_SREG_CS; |
931 |
++ break; |
932 |
+ case 0x36: /* SS override */ |
933 |
++ has_seg_override = true; |
934 |
++ ctxt->seg_override = VCPU_SREG_SS; |
935 |
++ break; |
936 |
+ case 0x3e: /* DS override */ |
937 |
+ has_seg_override = true; |
938 |
+- ctxt->seg_override = (ctxt->b >> 3) & 3; |
939 |
++ ctxt->seg_override = VCPU_SREG_DS; |
940 |
+ break; |
941 |
+ case 0x64: /* FS override */ |
942 |
++ has_seg_override = true; |
943 |
++ ctxt->seg_override = VCPU_SREG_FS; |
944 |
++ break; |
945 |
+ case 0x65: /* GS override */ |
946 |
+ has_seg_override = true; |
947 |
+- ctxt->seg_override = ctxt->b & 7; |
948 |
++ ctxt->seg_override = VCPU_SREG_GS; |
949 |
+ break; |
950 |
+ case 0x40 ... 0x4f: /* REX */ |
951 |
+ if (mode != X86EMUL_MODE_PROT64) |
952 |
+@@ -5257,10 +5269,15 @@ done_prefixes: |
953 |
+ } |
954 |
+ break; |
955 |
+ case Escape: |
956 |
+- if (ctxt->modrm > 0xbf) |
957 |
+- opcode = opcode.u.esc->high[ctxt->modrm - 0xc0]; |
958 |
+- else |
959 |
++ if (ctxt->modrm > 0xbf) { |
960 |
++ size_t size = ARRAY_SIZE(opcode.u.esc->high); |
961 |
++ u32 index = array_index_nospec( |
962 |
++ ctxt->modrm - 0xc0, size); |
963 |
++ |
964 |
++ opcode = opcode.u.esc->high[index]; |
965 |
++ } else { |
966 |
+ opcode = opcode.u.esc->op[(ctxt->modrm >> 3) & 7]; |
967 |
++ } |
968 |
+ break; |
969 |
+ case InstrDual: |
970 |
+ if ((ctxt->modrm >> 6) == 3) |
971 |
+diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c |
972 |
+index 5842c5f587fe..3fd6c4b2c2b7 100644 |
973 |
+--- a/arch/x86/kvm/hyperv.c |
974 |
++++ b/arch/x86/kvm/hyperv.c |
975 |
+@@ -792,11 +792,12 @@ static int kvm_hv_msr_get_crash_data(struct kvm_vcpu *vcpu, |
976 |
+ u32 index, u64 *pdata) |
977 |
+ { |
978 |
+ struct kvm_hv *hv = &vcpu->kvm->arch.hyperv; |
979 |
++ size_t size = ARRAY_SIZE(hv->hv_crash_param); |
980 |
+ |
981 |
+- if (WARN_ON_ONCE(index >= ARRAY_SIZE(hv->hv_crash_param))) |
982 |
++ if (WARN_ON_ONCE(index >= size)) |
983 |
+ return -EINVAL; |
984 |
+ |
985 |
+- *pdata = hv->hv_crash_param[index]; |
986 |
++ *pdata = hv->hv_crash_param[array_index_nospec(index, size)]; |
987 |
+ return 0; |
988 |
+ } |
989 |
+ |
990 |
+@@ -835,11 +836,12 @@ static int kvm_hv_msr_set_crash_data(struct kvm_vcpu *vcpu, |
991 |
+ u32 index, u64 data) |
992 |
+ { |
993 |
+ struct kvm_hv *hv = &vcpu->kvm->arch.hyperv; |
994 |
++ size_t size = ARRAY_SIZE(hv->hv_crash_param); |
995 |
+ |
996 |
+- if (WARN_ON_ONCE(index >= ARRAY_SIZE(hv->hv_crash_param))) |
997 |
++ if (WARN_ON_ONCE(index >= size)) |
998 |
+ return -EINVAL; |
999 |
+ |
1000 |
+- hv->hv_crash_param[index] = data; |
1001 |
++ hv->hv_crash_param[array_index_nospec(index, size)] = data; |
1002 |
+ return 0; |
1003 |
+ } |
1004 |
+ |
1005 |
+diff --git a/arch/x86/kvm/i8259.c b/arch/x86/kvm/i8259.c |
1006 |
+index bdcd4139eca9..38a36a1cc87f 100644 |
1007 |
+--- a/arch/x86/kvm/i8259.c |
1008 |
++++ b/arch/x86/kvm/i8259.c |
1009 |
+@@ -460,10 +460,14 @@ static int picdev_write(struct kvm_pic *s, |
1010 |
+ switch (addr) { |
1011 |
+ case 0x20: |
1012 |
+ case 0x21: |
1013 |
++ pic_lock(s); |
1014 |
++ pic_ioport_write(&s->pics[0], addr, data); |
1015 |
++ pic_unlock(s); |
1016 |
++ break; |
1017 |
+ case 0xa0: |
1018 |
+ case 0xa1: |
1019 |
+ pic_lock(s); |
1020 |
+- pic_ioport_write(&s->pics[addr >> 7], addr, data); |
1021 |
++ pic_ioport_write(&s->pics[1], addr, data); |
1022 |
+ pic_unlock(s); |
1023 |
+ break; |
1024 |
+ case 0x4d0: |
1025 |
+diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c |
1026 |
+index 4e822ad363f3..bac2ec9b4443 100644 |
1027 |
+--- a/arch/x86/kvm/ioapic.c |
1028 |
++++ b/arch/x86/kvm/ioapic.c |
1029 |
+@@ -36,6 +36,7 @@ |
1030 |
+ #include <linux/io.h> |
1031 |
+ #include <linux/slab.h> |
1032 |
+ #include <linux/export.h> |
1033 |
++#include <linux/nospec.h> |
1034 |
+ #include <asm/processor.h> |
1035 |
+ #include <asm/page.h> |
1036 |
+ #include <asm/current.h> |
1037 |
+@@ -73,13 +74,14 @@ static unsigned long ioapic_read_indirect(struct kvm_ioapic *ioapic, |
1038 |
+ default: |
1039 |
+ { |
1040 |
+ u32 redir_index = (ioapic->ioregsel - 0x10) >> 1; |
1041 |
+- u64 redir_content; |
1042 |
++ u64 redir_content = ~0ULL; |
1043 |
+ |
1044 |
+- if (redir_index < IOAPIC_NUM_PINS) |
1045 |
+- redir_content = |
1046 |
+- ioapic->redirtbl[redir_index].bits; |
1047 |
+- else |
1048 |
+- redir_content = ~0ULL; |
1049 |
++ if (redir_index < IOAPIC_NUM_PINS) { |
1050 |
++ u32 index = array_index_nospec( |
1051 |
++ redir_index, IOAPIC_NUM_PINS); |
1052 |
++ |
1053 |
++ redir_content = ioapic->redirtbl[index].bits; |
1054 |
++ } |
1055 |
+ |
1056 |
+ result = (ioapic->ioregsel & 0x1) ? |
1057 |
+ (redir_content >> 32) & 0xffffffff : |
1058 |
+@@ -297,6 +299,7 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val) |
1059 |
+ ioapic_debug("change redir index %x val %x\n", index, val); |
1060 |
+ if (index >= IOAPIC_NUM_PINS) |
1061 |
+ return; |
1062 |
++ index = array_index_nospec(index, IOAPIC_NUM_PINS); |
1063 |
+ e = &ioapic->redirtbl[index]; |
1064 |
+ mask_before = e->fields.mask; |
1065 |
+ /* Preserve read-only fields */ |
1066 |
+diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c |
1067 |
+index 262e49301cae..05905961ecca 100644 |
1068 |
+--- a/arch/x86/kvm/lapic.c |
1069 |
++++ b/arch/x86/kvm/lapic.c |
1070 |
+@@ -1862,15 +1862,20 @@ int kvm_lapic_reg_write(struct kvm_lapic *apic, u32 reg, u32 val) |
1071 |
+ case APIC_LVTTHMR: |
1072 |
+ case APIC_LVTPC: |
1073 |
+ case APIC_LVT1: |
1074 |
+- case APIC_LVTERR: |
1075 |
++ case APIC_LVTERR: { |
1076 |
+ /* TODO: Check vector */ |
1077 |
++ size_t size; |
1078 |
++ u32 index; |
1079 |
++ |
1080 |
+ if (!kvm_apic_sw_enabled(apic)) |
1081 |
+ val |= APIC_LVT_MASKED; |
1082 |
+- |
1083 |
+- val &= apic_lvt_mask[(reg - APIC_LVTT) >> 4]; |
1084 |
++ size = ARRAY_SIZE(apic_lvt_mask); |
1085 |
++ index = array_index_nospec( |
1086 |
++ (reg - APIC_LVTT) >> 4, size); |
1087 |
++ val &= apic_lvt_mask[index]; |
1088 |
+ kvm_lapic_set_reg(apic, reg, val); |
1089 |
+- |
1090 |
+ break; |
1091 |
++ } |
1092 |
+ |
1093 |
+ case APIC_LVTT: |
1094 |
+ if (!kvm_apic_sw_enabled(apic)) |
1095 |
+diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c |
1096 |
+index eddf91a0e363..62f1e4663bc3 100644 |
1097 |
+--- a/arch/x86/kvm/mmu.c |
1098 |
++++ b/arch/x86/kvm/mmu.c |
1099 |
+@@ -1184,12 +1184,12 @@ static bool mmu_gfn_lpage_is_disallowed(struct kvm_vcpu *vcpu, gfn_t gfn, |
1100 |
+ return __mmu_gfn_lpage_is_disallowed(gfn, level, slot); |
1101 |
+ } |
1102 |
+ |
1103 |
+-static int host_mapping_level(struct kvm *kvm, gfn_t gfn) |
1104 |
++static int host_mapping_level(struct kvm_vcpu *vcpu, gfn_t gfn) |
1105 |
+ { |
1106 |
+ unsigned long page_size; |
1107 |
+ int i, ret = 0; |
1108 |
+ |
1109 |
+- page_size = kvm_host_page_size(kvm, gfn); |
1110 |
++ page_size = kvm_host_page_size(vcpu, gfn); |
1111 |
+ |
1112 |
+ for (i = PT_PAGE_TABLE_LEVEL; i <= PT_MAX_HUGEPAGE_LEVEL; ++i) { |
1113 |
+ if (page_size >= KVM_HPAGE_SIZE(i)) |
1114 |
+@@ -1239,7 +1239,7 @@ static int mapping_level(struct kvm_vcpu *vcpu, gfn_t large_gfn, |
1115 |
+ if (unlikely(*force_pt_level)) |
1116 |
+ return PT_PAGE_TABLE_LEVEL; |
1117 |
+ |
1118 |
+- host_level = host_mapping_level(vcpu->kvm, large_gfn); |
1119 |
++ host_level = host_mapping_level(vcpu, large_gfn); |
1120 |
+ |
1121 |
+ if (host_level == PT_PAGE_TABLE_LEVEL) |
1122 |
+ return host_level; |
1123 |
+@@ -3390,7 +3390,7 @@ static bool is_access_allowed(u32 fault_err_code, u64 spte) |
1124 |
+ * - true: let the vcpu to access on the same address again. |
1125 |
+ * - false: let the real page fault path to fix it. |
1126 |
+ */ |
1127 |
+-static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level, |
1128 |
++static bool fast_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, int level, |
1129 |
+ u32 error_code) |
1130 |
+ { |
1131 |
+ struct kvm_shadow_walk_iterator iterator; |
1132 |
+@@ -3410,7 +3410,7 @@ static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level, |
1133 |
+ do { |
1134 |
+ u64 new_spte; |
1135 |
+ |
1136 |
+- for_each_shadow_entry_lockless(vcpu, gva, iterator, spte) |
1137 |
++ for_each_shadow_entry_lockless(vcpu, cr2_or_gpa, iterator, spte) |
1138 |
+ if (!is_shadow_present_pte(spte) || |
1139 |
+ iterator.level < level) |
1140 |
+ break; |
1141 |
+@@ -3488,7 +3488,7 @@ static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level, |
1142 |
+ |
1143 |
+ } while (true); |
1144 |
+ |
1145 |
+- trace_fast_page_fault(vcpu, gva, error_code, iterator.sptep, |
1146 |
++ trace_fast_page_fault(vcpu, cr2_or_gpa, error_code, iterator.sptep, |
1147 |
+ spte, fault_handled); |
1148 |
+ walk_shadow_page_lockless_end(vcpu); |
1149 |
+ |
1150 |
+@@ -3496,10 +3496,11 @@ static bool fast_page_fault(struct kvm_vcpu *vcpu, gva_t gva, int level, |
1151 |
+ } |
1152 |
+ |
1153 |
+ static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn, |
1154 |
+- gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable); |
1155 |
++ gpa_t cr2_or_gpa, kvm_pfn_t *pfn, bool write, |
1156 |
++ bool *writable); |
1157 |
+ static int make_mmu_pages_available(struct kvm_vcpu *vcpu); |
1158 |
+ |
1159 |
+-static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code, |
1160 |
++static int nonpaging_map(struct kvm_vcpu *vcpu, gpa_t gpa, u32 error_code, |
1161 |
+ gfn_t gfn, bool prefault) |
1162 |
+ { |
1163 |
+ int r; |
1164 |
+@@ -3525,16 +3526,16 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code, |
1165 |
+ gfn &= ~(KVM_PAGES_PER_HPAGE(level) - 1); |
1166 |
+ } |
1167 |
+ |
1168 |
+- if (fast_page_fault(vcpu, v, level, error_code)) |
1169 |
++ if (fast_page_fault(vcpu, gpa, level, error_code)) |
1170 |
+ return RET_PF_RETRY; |
1171 |
+ |
1172 |
+ mmu_seq = vcpu->kvm->mmu_notifier_seq; |
1173 |
+ smp_rmb(); |
1174 |
+ |
1175 |
+- if (try_async_pf(vcpu, prefault, gfn, v, &pfn, write, &map_writable)) |
1176 |
++ if (try_async_pf(vcpu, prefault, gfn, gpa, &pfn, write, &map_writable)) |
1177 |
+ return RET_PF_RETRY; |
1178 |
+ |
1179 |
+- if (handle_abnormal_pfn(vcpu, v, gfn, pfn, ACC_ALL, &r)) |
1180 |
++ if (handle_abnormal_pfn(vcpu, gpa, gfn, pfn, ACC_ALL, &r)) |
1181 |
+ return r; |
1182 |
+ |
1183 |
+ r = RET_PF_RETRY; |
1184 |
+@@ -3545,7 +3546,7 @@ static int nonpaging_map(struct kvm_vcpu *vcpu, gva_t v, u32 error_code, |
1185 |
+ goto out_unlock; |
1186 |
+ if (likely(!force_pt_level)) |
1187 |
+ transparent_hugepage_adjust(vcpu, gfn, &pfn, &level); |
1188 |
+- r = __direct_map(vcpu, v, write, map_writable, level, pfn, |
1189 |
++ r = __direct_map(vcpu, gpa, write, map_writable, level, pfn, |
1190 |
+ prefault, false); |
1191 |
+ out_unlock: |
1192 |
+ spin_unlock(&vcpu->kvm->mmu_lock); |
1193 |
+@@ -3838,7 +3839,7 @@ void kvm_mmu_sync_roots(struct kvm_vcpu *vcpu) |
1194 |
+ } |
1195 |
+ EXPORT_SYMBOL_GPL(kvm_mmu_sync_roots); |
1196 |
+ |
1197 |
+-static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr, |
1198 |
++static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gpa_t vaddr, |
1199 |
+ u32 access, struct x86_exception *exception) |
1200 |
+ { |
1201 |
+ if (exception) |
1202 |
+@@ -3846,7 +3847,7 @@ static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr, |
1203 |
+ return vaddr; |
1204 |
+ } |
1205 |
+ |
1206 |
+-static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gva_t vaddr, |
1207 |
++static gpa_t nonpaging_gva_to_gpa_nested(struct kvm_vcpu *vcpu, gpa_t vaddr, |
1208 |
+ u32 access, |
1209 |
+ struct x86_exception *exception) |
1210 |
+ { |
1211 |
+@@ -4006,13 +4007,14 @@ static void shadow_page_table_clear_flood(struct kvm_vcpu *vcpu, gva_t addr) |
1212 |
+ walk_shadow_page_lockless_end(vcpu); |
1213 |
+ } |
1214 |
+ |
1215 |
+-static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva, |
1216 |
++static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gpa_t gpa, |
1217 |
+ u32 error_code, bool prefault) |
1218 |
+ { |
1219 |
+- gfn_t gfn = gva >> PAGE_SHIFT; |
1220 |
++ gfn_t gfn = gpa >> PAGE_SHIFT; |
1221 |
+ int r; |
1222 |
+ |
1223 |
+- pgprintk("%s: gva %lx error %x\n", __func__, gva, error_code); |
1224 |
++ /* Note, paging is disabled, ergo gva == gpa. */ |
1225 |
++ pgprintk("%s: gva %lx error %x\n", __func__, gpa, error_code); |
1226 |
+ |
1227 |
+ if (page_fault_handle_page_track(vcpu, error_code, gfn)) |
1228 |
+ return RET_PF_EMULATE; |
1229 |
+@@ -4024,11 +4026,12 @@ static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva, |
1230 |
+ MMU_WARN_ON(!VALID_PAGE(vcpu->arch.mmu.root_hpa)); |
1231 |
+ |
1232 |
+ |
1233 |
+- return nonpaging_map(vcpu, gva & PAGE_MASK, |
1234 |
++ return nonpaging_map(vcpu, gpa & PAGE_MASK, |
1235 |
+ error_code, gfn, prefault); |
1236 |
+ } |
1237 |
+ |
1238 |
+-static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn) |
1239 |
++static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
1240 |
++ gfn_t gfn) |
1241 |
+ { |
1242 |
+ struct kvm_arch_async_pf arch; |
1243 |
+ |
1244 |
+@@ -4037,7 +4040,8 @@ static int kvm_arch_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn) |
1245 |
+ arch.direct_map = vcpu->arch.mmu.direct_map; |
1246 |
+ arch.cr3 = vcpu->arch.mmu.get_cr3(vcpu); |
1247 |
+ |
1248 |
+- return kvm_setup_async_pf(vcpu, gva, kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch); |
1249 |
++ return kvm_setup_async_pf(vcpu, cr2_or_gpa, |
1250 |
++ kvm_vcpu_gfn_to_hva(vcpu, gfn), &arch); |
1251 |
+ } |
1252 |
+ |
1253 |
+ bool kvm_can_do_async_pf(struct kvm_vcpu *vcpu) |
1254 |
+@@ -4054,7 +4058,8 @@ bool kvm_can_do_async_pf(struct kvm_vcpu *vcpu) |
1255 |
+ } |
1256 |
+ |
1257 |
+ static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn, |
1258 |
+- gva_t gva, kvm_pfn_t *pfn, bool write, bool *writable) |
1259 |
++ gpa_t cr2_or_gpa, kvm_pfn_t *pfn, bool write, |
1260 |
++ bool *writable) |
1261 |
+ { |
1262 |
+ struct kvm_memory_slot *slot; |
1263 |
+ bool async; |
1264 |
+@@ -4074,12 +4079,12 @@ static bool try_async_pf(struct kvm_vcpu *vcpu, bool prefault, gfn_t gfn, |
1265 |
+ return false; /* *pfn has correct page already */ |
1266 |
+ |
1267 |
+ if (!prefault && kvm_can_do_async_pf(vcpu)) { |
1268 |
+- trace_kvm_try_async_get_page(gva, gfn); |
1269 |
++ trace_kvm_try_async_get_page(cr2_or_gpa, gfn); |
1270 |
+ if (kvm_find_async_pf_gfn(vcpu, gfn)) { |
1271 |
+- trace_kvm_async_pf_doublefault(gva, gfn); |
1272 |
++ trace_kvm_async_pf_doublefault(cr2_or_gpa, gfn); |
1273 |
+ kvm_make_request(KVM_REQ_APF_HALT, vcpu); |
1274 |
+ return true; |
1275 |
+- } else if (kvm_arch_setup_async_pf(vcpu, gva, gfn)) |
1276 |
++ } else if (kvm_arch_setup_async_pf(vcpu, cr2_or_gpa, gfn)) |
1277 |
+ return true; |
1278 |
+ } |
1279 |
+ |
1280 |
+@@ -4092,6 +4097,12 @@ int kvm_handle_page_fault(struct kvm_vcpu *vcpu, u64 error_code, |
1281 |
+ { |
1282 |
+ int r = 1; |
1283 |
+ |
1284 |
++#ifndef CONFIG_X86_64 |
1285 |
++ /* A 64-bit CR2 should be impossible on 32-bit KVM. */ |
1286 |
++ if (WARN_ON_ONCE(fault_address >> 32)) |
1287 |
++ return -EFAULT; |
1288 |
++#endif |
1289 |
++ |
1290 |
+ vcpu->arch.l1tf_flush_l1d = true; |
1291 |
+ switch (vcpu->arch.apf.host_apf_reason) { |
1292 |
+ default: |
1293 |
+@@ -4129,7 +4140,7 @@ check_hugepage_cache_consistency(struct kvm_vcpu *vcpu, gfn_t gfn, int level) |
1294 |
+ return kvm_mtrr_check_gfn_range_consistency(vcpu, gfn, page_num); |
1295 |
+ } |
1296 |
+ |
1297 |
+-static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code, |
1298 |
++static int tdp_page_fault(struct kvm_vcpu *vcpu, gpa_t gpa, u32 error_code, |
1299 |
+ bool prefault) |
1300 |
+ { |
1301 |
+ kvm_pfn_t pfn; |
1302 |
+@@ -5307,7 +5318,7 @@ static int make_mmu_pages_available(struct kvm_vcpu *vcpu) |
1303 |
+ return 0; |
1304 |
+ } |
1305 |
+ |
1306 |
+-int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code, |
1307 |
++int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u64 error_code, |
1308 |
+ void *insn, int insn_len) |
1309 |
+ { |
1310 |
+ int r, emulation_type = 0; |
1311 |
+@@ -5317,19 +5328,20 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code, |
1312 |
+ /* With shadow page tables, fault_address contains a GVA or nGPA. */ |
1313 |
+ if (vcpu->arch.mmu.direct_map) { |
1314 |
+ vcpu->arch.gpa_available = true; |
1315 |
+- vcpu->arch.gpa_val = cr2; |
1316 |
++ vcpu->arch.gpa_val = cr2_or_gpa; |
1317 |
+ } |
1318 |
+ |
1319 |
+ r = RET_PF_INVALID; |
1320 |
+ if (unlikely(error_code & PFERR_RSVD_MASK)) { |
1321 |
+- r = handle_mmio_page_fault(vcpu, cr2, direct); |
1322 |
++ r = handle_mmio_page_fault(vcpu, cr2_or_gpa, direct); |
1323 |
+ if (r == RET_PF_EMULATE) |
1324 |
+ goto emulate; |
1325 |
+ } |
1326 |
+ |
1327 |
+ if (r == RET_PF_INVALID) { |
1328 |
+- r = vcpu->arch.mmu.page_fault(vcpu, cr2, lower_32_bits(error_code), |
1329 |
+- false); |
1330 |
++ r = vcpu->arch.mmu.page_fault(vcpu, cr2_or_gpa, |
1331 |
++ lower_32_bits(error_code), |
1332 |
++ false); |
1333 |
+ WARN_ON(r == RET_PF_INVALID); |
1334 |
+ } |
1335 |
+ |
1336 |
+@@ -5347,7 +5359,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code, |
1337 |
+ */ |
1338 |
+ if (vcpu->arch.mmu.direct_map && |
1339 |
+ (error_code & PFERR_NESTED_GUEST_PAGE) == PFERR_NESTED_GUEST_PAGE) { |
1340 |
+- kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(cr2)); |
1341 |
++ kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(cr2_or_gpa)); |
1342 |
+ return 1; |
1343 |
+ } |
1344 |
+ |
1345 |
+@@ -5362,7 +5374,7 @@ int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t cr2, u64 error_code, |
1346 |
+ * explicitly shadowing L1's page tables, i.e. unprotecting something |
1347 |
+ * for L1 isn't going to magically fix whatever issue cause L2 to fail. |
1348 |
+ */ |
1349 |
+- if (!mmio_info_in_cache(vcpu, cr2, direct) && !is_guest_mode(vcpu)) |
1350 |
++ if (!mmio_info_in_cache(vcpu, cr2_or_gpa, direct) && !is_guest_mode(vcpu)) |
1351 |
+ emulation_type = EMULTYPE_ALLOW_RETRY; |
1352 |
+ emulate: |
1353 |
+ /* |
1354 |
+@@ -5375,7 +5387,7 @@ emulate: |
1355 |
+ if (unlikely(insn && !insn_len)) |
1356 |
+ return 1; |
1357 |
+ |
1358 |
+- er = x86_emulate_instruction(vcpu, cr2, emulation_type, insn, insn_len); |
1359 |
++ er = x86_emulate_instruction(vcpu, cr2_or_gpa, emulation_type, insn, insn_len); |
1360 |
+ |
1361 |
+ switch (er) { |
1362 |
+ case EMULATE_DONE: |
1363 |
+diff --git a/arch/x86/kvm/mmutrace.h b/arch/x86/kvm/mmutrace.h |
1364 |
+index 918b0d5bf272..cb41b036eb26 100644 |
1365 |
+--- a/arch/x86/kvm/mmutrace.h |
1366 |
++++ b/arch/x86/kvm/mmutrace.h |
1367 |
+@@ -249,13 +249,13 @@ TRACE_EVENT( |
1368 |
+ |
1369 |
+ TRACE_EVENT( |
1370 |
+ fast_page_fault, |
1371 |
+- TP_PROTO(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code, |
1372 |
++ TP_PROTO(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, u32 error_code, |
1373 |
+ u64 *sptep, u64 old_spte, bool retry), |
1374 |
+- TP_ARGS(vcpu, gva, error_code, sptep, old_spte, retry), |
1375 |
++ TP_ARGS(vcpu, cr2_or_gpa, error_code, sptep, old_spte, retry), |
1376 |
+ |
1377 |
+ TP_STRUCT__entry( |
1378 |
+ __field(int, vcpu_id) |
1379 |
+- __field(gva_t, gva) |
1380 |
++ __field(gpa_t, cr2_or_gpa) |
1381 |
+ __field(u32, error_code) |
1382 |
+ __field(u64 *, sptep) |
1383 |
+ __field(u64, old_spte) |
1384 |
+@@ -265,7 +265,7 @@ TRACE_EVENT( |
1385 |
+ |
1386 |
+ TP_fast_assign( |
1387 |
+ __entry->vcpu_id = vcpu->vcpu_id; |
1388 |
+- __entry->gva = gva; |
1389 |
++ __entry->cr2_or_gpa = cr2_or_gpa; |
1390 |
+ __entry->error_code = error_code; |
1391 |
+ __entry->sptep = sptep; |
1392 |
+ __entry->old_spte = old_spte; |
1393 |
+@@ -273,9 +273,9 @@ TRACE_EVENT( |
1394 |
+ __entry->retry = retry; |
1395 |
+ ), |
1396 |
+ |
1397 |
+- TP_printk("vcpu %d gva %lx error_code %s sptep %p old %#llx" |
1398 |
++ TP_printk("vcpu %d gva %llx error_code %s sptep %p old %#llx" |
1399 |
+ " new %llx spurious %d fixed %d", __entry->vcpu_id, |
1400 |
+- __entry->gva, __print_flags(__entry->error_code, "|", |
1401 |
++ __entry->cr2_or_gpa, __print_flags(__entry->error_code, "|", |
1402 |
+ kvm_mmu_trace_pferr_flags), __entry->sptep, |
1403 |
+ __entry->old_spte, __entry->new_spte, |
1404 |
+ __spte_satisfied(old_spte), __spte_satisfied(new_spte) |
1405 |
+diff --git a/arch/x86/kvm/mtrr.c b/arch/x86/kvm/mtrr.c |
1406 |
+index 9f72cc427158..fabce87697e5 100644 |
1407 |
+--- a/arch/x86/kvm/mtrr.c |
1408 |
++++ b/arch/x86/kvm/mtrr.c |
1409 |
+@@ -194,11 +194,15 @@ static bool fixed_msr_to_seg_unit(u32 msr, int *seg, int *unit) |
1410 |
+ break; |
1411 |
+ case MSR_MTRRfix16K_80000 ... MSR_MTRRfix16K_A0000: |
1412 |
+ *seg = 1; |
1413 |
+- *unit = msr - MSR_MTRRfix16K_80000; |
1414 |
++ *unit = array_index_nospec( |
1415 |
++ msr - MSR_MTRRfix16K_80000, |
1416 |
++ MSR_MTRRfix16K_A0000 - MSR_MTRRfix16K_80000 + 1); |
1417 |
+ break; |
1418 |
+ case MSR_MTRRfix4K_C0000 ... MSR_MTRRfix4K_F8000: |
1419 |
+ *seg = 2; |
1420 |
+- *unit = msr - MSR_MTRRfix4K_C0000; |
1421 |
++ *unit = array_index_nospec( |
1422 |
++ msr - MSR_MTRRfix4K_C0000, |
1423 |
++ MSR_MTRRfix4K_F8000 - MSR_MTRRfix4K_C0000 + 1); |
1424 |
+ break; |
1425 |
+ default: |
1426 |
+ return false; |
1427 |
+diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h |
1428 |
+index adf42dc8d38b..100ae4fabf17 100644 |
1429 |
+--- a/arch/x86/kvm/paging_tmpl.h |
1430 |
++++ b/arch/x86/kvm/paging_tmpl.h |
1431 |
+@@ -273,11 +273,11 @@ static inline unsigned FNAME(gpte_pkeys)(struct kvm_vcpu *vcpu, u64 gpte) |
1432 |
+ } |
1433 |
+ |
1434 |
+ /* |
1435 |
+- * Fetch a guest pte for a guest virtual address |
1436 |
++ * Fetch a guest pte for a guest virtual address, or for an L2's GPA. |
1437 |
+ */ |
1438 |
+ static int FNAME(walk_addr_generic)(struct guest_walker *walker, |
1439 |
+ struct kvm_vcpu *vcpu, struct kvm_mmu *mmu, |
1440 |
+- gva_t addr, u32 access) |
1441 |
++ gpa_t addr, u32 access) |
1442 |
+ { |
1443 |
+ int ret; |
1444 |
+ pt_element_t pte; |
1445 |
+@@ -478,7 +478,7 @@ error: |
1446 |
+ } |
1447 |
+ |
1448 |
+ static int FNAME(walk_addr)(struct guest_walker *walker, |
1449 |
+- struct kvm_vcpu *vcpu, gva_t addr, u32 access) |
1450 |
++ struct kvm_vcpu *vcpu, gpa_t addr, u32 access) |
1451 |
+ { |
1452 |
+ return FNAME(walk_addr_generic)(walker, vcpu, &vcpu->arch.mmu, addr, |
1453 |
+ access); |
1454 |
+@@ -593,7 +593,7 @@ static void FNAME(pte_prefetch)(struct kvm_vcpu *vcpu, struct guest_walker *gw, |
1455 |
+ * If the guest tries to write a write-protected page, we need to |
1456 |
+ * emulate this operation, return 1 to indicate this case. |
1457 |
+ */ |
1458 |
+-static int FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr, |
1459 |
++static int FNAME(fetch)(struct kvm_vcpu *vcpu, gpa_t addr, |
1460 |
+ struct guest_walker *gw, |
1461 |
+ int write_fault, int hlevel, |
1462 |
+ kvm_pfn_t pfn, bool map_writable, bool prefault, |
1463 |
+@@ -747,7 +747,7 @@ FNAME(is_self_change_mapping)(struct kvm_vcpu *vcpu, |
1464 |
+ * Returns: 1 if we need to emulate the instruction, 0 otherwise, or |
1465 |
+ * a negative value on error. |
1466 |
+ */ |
1467 |
+-static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr, u32 error_code, |
1468 |
++static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gpa_t addr, u32 error_code, |
1469 |
+ bool prefault) |
1470 |
+ { |
1471 |
+ int write_fault = error_code & PFERR_WRITE_MASK; |
1472 |
+@@ -926,18 +926,19 @@ static void FNAME(invlpg)(struct kvm_vcpu *vcpu, gva_t gva, hpa_t root_hpa) |
1473 |
+ spin_unlock(&vcpu->kvm->mmu_lock); |
1474 |
+ } |
1475 |
+ |
1476 |
+-static gpa_t FNAME(gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t vaddr, u32 access, |
1477 |
++/* Note, @addr is a GPA when gva_to_gpa() translates an L2 GPA to an L1 GPA. */ |
1478 |
++static gpa_t FNAME(gva_to_gpa)(struct kvm_vcpu *vcpu, gpa_t addr, u32 access, |
1479 |
+ struct x86_exception *exception) |
1480 |
+ { |
1481 |
+ struct guest_walker walker; |
1482 |
+ gpa_t gpa = UNMAPPED_GVA; |
1483 |
+ int r; |
1484 |
+ |
1485 |
+- r = FNAME(walk_addr)(&walker, vcpu, vaddr, access); |
1486 |
++ r = FNAME(walk_addr)(&walker, vcpu, addr, access); |
1487 |
+ |
1488 |
+ if (r) { |
1489 |
+ gpa = gfn_to_gpa(walker.gfn); |
1490 |
+- gpa |= vaddr & ~PAGE_MASK; |
1491 |
++ gpa |= addr & ~PAGE_MASK; |
1492 |
+ } else if (exception) |
1493 |
+ *exception = walker.fault; |
1494 |
+ |
1495 |
+@@ -945,7 +946,8 @@ static gpa_t FNAME(gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t vaddr, u32 access, |
1496 |
+ } |
1497 |
+ |
1498 |
+ #if PTTYPE != PTTYPE_EPT |
1499 |
+-static gpa_t FNAME(gva_to_gpa_nested)(struct kvm_vcpu *vcpu, gva_t vaddr, |
1500 |
++/* Note, gva_to_gpa_nested() is only used to translate L2 GVAs. */ |
1501 |
++static gpa_t FNAME(gva_to_gpa_nested)(struct kvm_vcpu *vcpu, gpa_t vaddr, |
1502 |
+ u32 access, |
1503 |
+ struct x86_exception *exception) |
1504 |
+ { |
1505 |
+@@ -953,6 +955,11 @@ static gpa_t FNAME(gva_to_gpa_nested)(struct kvm_vcpu *vcpu, gva_t vaddr, |
1506 |
+ gpa_t gpa = UNMAPPED_GVA; |
1507 |
+ int r; |
1508 |
+ |
1509 |
++#ifndef CONFIG_X86_64 |
1510 |
++ /* A 64-bit GVA should be impossible on 32-bit KVM. */ |
1511 |
++ WARN_ON_ONCE(vaddr >> 32); |
1512 |
++#endif |
1513 |
++ |
1514 |
+ r = FNAME(walk_addr_nested)(&walker, vcpu, vaddr, access); |
1515 |
+ |
1516 |
+ if (r) { |
1517 |
+diff --git a/arch/x86/kvm/pmu.h b/arch/x86/kvm/pmu.h |
1518 |
+index 22dff661145a..7b4828e50ab3 100644 |
1519 |
+--- a/arch/x86/kvm/pmu.h |
1520 |
++++ b/arch/x86/kvm/pmu.h |
1521 |
+@@ -2,6 +2,8 @@ |
1522 |
+ #ifndef __KVM_X86_PMU_H |
1523 |
+ #define __KVM_X86_PMU_H |
1524 |
+ |
1525 |
++#include <linux/nospec.h> |
1526 |
++ |
1527 |
+ #define vcpu_to_pmu(vcpu) (&(vcpu)->arch.pmu) |
1528 |
+ #define pmu_to_vcpu(pmu) (container_of((pmu), struct kvm_vcpu, arch.pmu)) |
1529 |
+ #define pmc_to_pmu(pmc) (&(pmc)->vcpu->arch.pmu) |
1530 |
+@@ -86,8 +88,12 @@ static inline bool pmc_is_enabled(struct kvm_pmc *pmc) |
1531 |
+ static inline struct kvm_pmc *get_gp_pmc(struct kvm_pmu *pmu, u32 msr, |
1532 |
+ u32 base) |
1533 |
+ { |
1534 |
+- if (msr >= base && msr < base + pmu->nr_arch_gp_counters) |
1535 |
+- return &pmu->gp_counters[msr - base]; |
1536 |
++ if (msr >= base && msr < base + pmu->nr_arch_gp_counters) { |
1537 |
++ u32 index = array_index_nospec(msr - base, |
1538 |
++ pmu->nr_arch_gp_counters); |
1539 |
++ |
1540 |
++ return &pmu->gp_counters[index]; |
1541 |
++ } |
1542 |
+ |
1543 |
+ return NULL; |
1544 |
+ } |
1545 |
+@@ -97,8 +103,12 @@ static inline struct kvm_pmc *get_fixed_pmc(struct kvm_pmu *pmu, u32 msr) |
1546 |
+ { |
1547 |
+ int base = MSR_CORE_PERF_FIXED_CTR0; |
1548 |
+ |
1549 |
+- if (msr >= base && msr < base + pmu->nr_arch_fixed_counters) |
1550 |
+- return &pmu->fixed_counters[msr - base]; |
1551 |
++ if (msr >= base && msr < base + pmu->nr_arch_fixed_counters) { |
1552 |
++ u32 index = array_index_nospec(msr - base, |
1553 |
++ pmu->nr_arch_fixed_counters); |
1554 |
++ |
1555 |
++ return &pmu->fixed_counters[index]; |
1556 |
++ } |
1557 |
+ |
1558 |
+ return NULL; |
1559 |
+ } |
1560 |
+diff --git a/arch/x86/kvm/pmu_intel.c b/arch/x86/kvm/pmu_intel.c |
1561 |
+index c3f103e2b08e..2ab8c20c8bfa 100644 |
1562 |
+--- a/arch/x86/kvm/pmu_intel.c |
1563 |
++++ b/arch/x86/kvm/pmu_intel.c |
1564 |
+@@ -87,10 +87,14 @@ static unsigned intel_find_arch_event(struct kvm_pmu *pmu, |
1565 |
+ |
1566 |
+ static unsigned intel_find_fixed_event(int idx) |
1567 |
+ { |
1568 |
+- if (idx >= ARRAY_SIZE(fixed_pmc_events)) |
1569 |
++ u32 event; |
1570 |
++ size_t size = ARRAY_SIZE(fixed_pmc_events); |
1571 |
++ |
1572 |
++ if (idx >= size) |
1573 |
+ return PERF_COUNT_HW_MAX; |
1574 |
+ |
1575 |
+- return intel_arch_events[fixed_pmc_events[idx]].event_type; |
1576 |
++ event = fixed_pmc_events[array_index_nospec(idx, size)]; |
1577 |
++ return intel_arch_events[event].event_type; |
1578 |
+ } |
1579 |
+ |
1580 |
+ /* check if a PMC is enabled by comparing it with globl_ctrl bits. */ |
1581 |
+@@ -131,16 +135,20 @@ static struct kvm_pmc *intel_msr_idx_to_pmc(struct kvm_vcpu *vcpu, |
1582 |
+ struct kvm_pmu *pmu = vcpu_to_pmu(vcpu); |
1583 |
+ bool fixed = idx & (1u << 30); |
1584 |
+ struct kvm_pmc *counters; |
1585 |
++ unsigned int num_counters; |
1586 |
+ |
1587 |
+ idx &= ~(3u << 30); |
1588 |
+- if (!fixed && idx >= pmu->nr_arch_gp_counters) |
1589 |
+- return NULL; |
1590 |
+- if (fixed && idx >= pmu->nr_arch_fixed_counters) |
1591 |
++ if (fixed) { |
1592 |
++ counters = pmu->fixed_counters; |
1593 |
++ num_counters = pmu->nr_arch_fixed_counters; |
1594 |
++ } else { |
1595 |
++ counters = pmu->gp_counters; |
1596 |
++ num_counters = pmu->nr_arch_gp_counters; |
1597 |
++ } |
1598 |
++ if (idx >= num_counters) |
1599 |
+ return NULL; |
1600 |
+- counters = fixed ? pmu->fixed_counters : pmu->gp_counters; |
1601 |
+ *mask &= pmu->counter_bitmask[fixed ? KVM_PMC_FIXED : KVM_PMC_GP]; |
1602 |
+- |
1603 |
+- return &counters[idx]; |
1604 |
++ return &counters[array_index_nospec(idx, num_counters)]; |
1605 |
+ } |
1606 |
+ |
1607 |
+ static bool intel_is_valid_msr(struct kvm_vcpu *vcpu, u32 msr) |
1608 |
+diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c |
1609 |
+index fa2abed1a14d..2660c01eadae 100644 |
1610 |
+--- a/arch/x86/kvm/vmx.c |
1611 |
++++ b/arch/x86/kvm/vmx.c |
1612 |
+@@ -8793,8 +8793,10 @@ static int handle_vmread(struct kvm_vcpu *vcpu) |
1613 |
+ /* _system ok, nested_vmx_check_permission has verified cpl=0 */ |
1614 |
+ if (kvm_write_guest_virt_system(vcpu, gva, &field_value, |
1615 |
+ (is_long_mode(vcpu) ? 8 : 4), |
1616 |
+- &e)) |
1617 |
++ &e)) { |
1618 |
+ kvm_inject_page_fault(vcpu, &e); |
1619 |
++ return 1; |
1620 |
++ } |
1621 |
+ } |
1622 |
+ |
1623 |
+ nested_vmx_succeed(vcpu); |
1624 |
+diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c |
1625 |
+new file mode 100644 |
1626 |
+index 000000000000..3791ce8d269e |
1627 |
+--- /dev/null |
1628 |
++++ b/arch/x86/kvm/vmx/vmx.c |
1629 |
+@@ -0,0 +1,8033 @@ |
1630 |
++// SPDX-License-Identifier: GPL-2.0-only |
1631 |
++/* |
1632 |
++ * Kernel-based Virtual Machine driver for Linux |
1633 |
++ * |
1634 |
++ * This module enables machines with Intel VT-x extensions to run virtual |
1635 |
++ * machines without emulation or binary translation. |
1636 |
++ * |
1637 |
++ * Copyright (C) 2006 Qumranet, Inc. |
1638 |
++ * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
1639 |
++ * |
1640 |
++ * Authors: |
1641 |
++ * Avi Kivity <avi@××××××××.com> |
1642 |
++ * Yaniv Kamay <yaniv@××××××××.com> |
1643 |
++ */ |
1644 |
++ |
1645 |
++#include <linux/frame.h> |
1646 |
++#include <linux/highmem.h> |
1647 |
++#include <linux/hrtimer.h> |
1648 |
++#include <linux/kernel.h> |
1649 |
++#include <linux/kvm_host.h> |
1650 |
++#include <linux/module.h> |
1651 |
++#include <linux/moduleparam.h> |
1652 |
++#include <linux/mod_devicetable.h> |
1653 |
++#include <linux/mm.h> |
1654 |
++#include <linux/sched.h> |
1655 |
++#include <linux/sched/smt.h> |
1656 |
++#include <linux/slab.h> |
1657 |
++#include <linux/tboot.h> |
1658 |
++#include <linux/trace_events.h> |
1659 |
++ |
1660 |
++#include <asm/apic.h> |
1661 |
++#include <asm/asm.h> |
1662 |
++#include <asm/cpu.h> |
1663 |
++#include <asm/debugreg.h> |
1664 |
++#include <asm/desc.h> |
1665 |
++#include <asm/fpu/internal.h> |
1666 |
++#include <asm/io.h> |
1667 |
++#include <asm/irq_remapping.h> |
1668 |
++#include <asm/kexec.h> |
1669 |
++#include <asm/perf_event.h> |
1670 |
++#include <asm/mce.h> |
1671 |
++#include <asm/mmu_context.h> |
1672 |
++#include <asm/mshyperv.h> |
1673 |
++#include <asm/spec-ctrl.h> |
1674 |
++#include <asm/virtext.h> |
1675 |
++#include <asm/vmx.h> |
1676 |
++ |
1677 |
++#include "capabilities.h" |
1678 |
++#include "cpuid.h" |
1679 |
++#include "evmcs.h" |
1680 |
++#include "irq.h" |
1681 |
++#include "kvm_cache_regs.h" |
1682 |
++#include "lapic.h" |
1683 |
++#include "mmu.h" |
1684 |
++#include "nested.h" |
1685 |
++#include "ops.h" |
1686 |
++#include "pmu.h" |
1687 |
++#include "trace.h" |
1688 |
++#include "vmcs.h" |
1689 |
++#include "vmcs12.h" |
1690 |
++#include "vmx.h" |
1691 |
++#include "x86.h" |
1692 |
++ |
1693 |
++MODULE_AUTHOR("Qumranet"); |
1694 |
++MODULE_LICENSE("GPL"); |
1695 |
++ |
1696 |
++static const struct x86_cpu_id vmx_cpu_id[] = { |
1697 |
++ X86_FEATURE_MATCH(X86_FEATURE_VMX), |
1698 |
++ {} |
1699 |
++}; |
1700 |
++MODULE_DEVICE_TABLE(x86cpu, vmx_cpu_id); |
1701 |
++ |
1702 |
++bool __read_mostly enable_vpid = 1; |
1703 |
++module_param_named(vpid, enable_vpid, bool, 0444); |
1704 |
++ |
1705 |
++static bool __read_mostly enable_vnmi = 1; |
1706 |
++module_param_named(vnmi, enable_vnmi, bool, S_IRUGO); |
1707 |
++ |
1708 |
++bool __read_mostly flexpriority_enabled = 1; |
1709 |
++module_param_named(flexpriority, flexpriority_enabled, bool, S_IRUGO); |
1710 |
++ |
1711 |
++bool __read_mostly enable_ept = 1; |
1712 |
++module_param_named(ept, enable_ept, bool, S_IRUGO); |
1713 |
++ |
1714 |
++bool __read_mostly enable_unrestricted_guest = 1; |
1715 |
++module_param_named(unrestricted_guest, |
1716 |
++ enable_unrestricted_guest, bool, S_IRUGO); |
1717 |
++ |
1718 |
++bool __read_mostly enable_ept_ad_bits = 1; |
1719 |
++module_param_named(eptad, enable_ept_ad_bits, bool, S_IRUGO); |
1720 |
++ |
1721 |
++static bool __read_mostly emulate_invalid_guest_state = true; |
1722 |
++module_param(emulate_invalid_guest_state, bool, S_IRUGO); |
1723 |
++ |
1724 |
++static bool __read_mostly fasteoi = 1; |
1725 |
++module_param(fasteoi, bool, S_IRUGO); |
1726 |
++ |
1727 |
++static bool __read_mostly enable_apicv = 1; |
1728 |
++module_param(enable_apicv, bool, S_IRUGO); |
1729 |
++ |
1730 |
++/* |
1731 |
++ * If nested=1, nested virtualization is supported, i.e., guests may use |
1732 |
++ * VMX and be a hypervisor for its own guests. If nested=0, guests may not |
1733 |
++ * use VMX instructions. |
1734 |
++ */ |
1735 |
++static bool __read_mostly nested = 1; |
1736 |
++module_param(nested, bool, S_IRUGO); |
1737 |
++ |
1738 |
++bool __read_mostly enable_pml = 1; |
1739 |
++module_param_named(pml, enable_pml, bool, S_IRUGO); |
1740 |
++ |
1741 |
++static bool __read_mostly dump_invalid_vmcs = 0; |
1742 |
++module_param(dump_invalid_vmcs, bool, 0644); |
1743 |
++ |
1744 |
++#define MSR_BITMAP_MODE_X2APIC 1 |
1745 |
++#define MSR_BITMAP_MODE_X2APIC_APICV 2 |
1746 |
++ |
1747 |
++#define KVM_VMX_TSC_MULTIPLIER_MAX 0xffffffffffffffffULL |
1748 |
++ |
1749 |
++/* Guest_tsc -> host_tsc conversion requires 64-bit division. */ |
1750 |
++static int __read_mostly cpu_preemption_timer_multi; |
1751 |
++static bool __read_mostly enable_preemption_timer = 1; |
1752 |
++#ifdef CONFIG_X86_64 |
1753 |
++module_param_named(preemption_timer, enable_preemption_timer, bool, S_IRUGO); |
1754 |
++#endif |
1755 |
++ |
1756 |
++#define KVM_VM_CR0_ALWAYS_OFF (X86_CR0_NW | X86_CR0_CD) |
1757 |
++#define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST X86_CR0_NE |
1758 |
++#define KVM_VM_CR0_ALWAYS_ON \ |
1759 |
++ (KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | \ |
1760 |
++ X86_CR0_WP | X86_CR0_PG | X86_CR0_PE) |
1761 |
++#define KVM_CR4_GUEST_OWNED_BITS \ |
1762 |
++ (X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR \ |
1763 |
++ | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_TSD) |
1764 |
++ |
1765 |
++#define KVM_VM_CR4_ALWAYS_ON_UNRESTRICTED_GUEST X86_CR4_VMXE |
1766 |
++#define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE) |
1767 |
++#define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE) |
1768 |
++ |
1769 |
++#define RMODE_GUEST_OWNED_EFLAGS_BITS (~(X86_EFLAGS_IOPL | X86_EFLAGS_VM)) |
1770 |
++ |
1771 |
++#define MSR_IA32_RTIT_STATUS_MASK (~(RTIT_STATUS_FILTEREN | \ |
1772 |
++ RTIT_STATUS_CONTEXTEN | RTIT_STATUS_TRIGGEREN | \ |
1773 |
++ RTIT_STATUS_ERROR | RTIT_STATUS_STOPPED | \ |
1774 |
++ RTIT_STATUS_BYTECNT)) |
1775 |
++ |
1776 |
++#define MSR_IA32_RTIT_OUTPUT_BASE_MASK \ |
1777 |
++ (~((1UL << cpuid_query_maxphyaddr(vcpu)) - 1) | 0x7f) |
1778 |
++ |
1779 |
++/* |
1780 |
++ * These 2 parameters are used to config the controls for Pause-Loop Exiting: |
1781 |
++ * ple_gap: upper bound on the amount of time between two successive |
1782 |
++ * executions of PAUSE in a loop. Also indicate if ple enabled. |
1783 |
++ * According to test, this time is usually smaller than 128 cycles. |
1784 |
++ * ple_window: upper bound on the amount of time a guest is allowed to execute |
1785 |
++ * in a PAUSE loop. Tests indicate that most spinlocks are held for |
1786 |
++ * less than 2^12 cycles |
1787 |
++ * Time is measured based on a counter that runs at the same rate as the TSC, |
1788 |
++ * refer SDM volume 3b section 21.6.13 & 22.1.3. |
1789 |
++ */ |
1790 |
++static unsigned int ple_gap = KVM_DEFAULT_PLE_GAP; |
1791 |
++module_param(ple_gap, uint, 0444); |
1792 |
++ |
1793 |
++static unsigned int ple_window = KVM_VMX_DEFAULT_PLE_WINDOW; |
1794 |
++module_param(ple_window, uint, 0444); |
1795 |
++ |
1796 |
++/* Default doubles per-vcpu window every exit. */ |
1797 |
++static unsigned int ple_window_grow = KVM_DEFAULT_PLE_WINDOW_GROW; |
1798 |
++module_param(ple_window_grow, uint, 0444); |
1799 |
++ |
1800 |
++/* Default resets per-vcpu window every exit to ple_window. */ |
1801 |
++static unsigned int ple_window_shrink = KVM_DEFAULT_PLE_WINDOW_SHRINK; |
1802 |
++module_param(ple_window_shrink, uint, 0444); |
1803 |
++ |
1804 |
++/* Default is to compute the maximum so we can never overflow. */ |
1805 |
++static unsigned int ple_window_max = KVM_VMX_DEFAULT_PLE_WINDOW_MAX; |
1806 |
++module_param(ple_window_max, uint, 0444); |
1807 |
++ |
1808 |
++/* Default is SYSTEM mode, 1 for host-guest mode */ |
1809 |
++int __read_mostly pt_mode = PT_MODE_SYSTEM; |
1810 |
++module_param(pt_mode, int, S_IRUGO); |
1811 |
++ |
1812 |
++static DEFINE_STATIC_KEY_FALSE(vmx_l1d_should_flush); |
1813 |
++static DEFINE_STATIC_KEY_FALSE(vmx_l1d_flush_cond); |
1814 |
++static DEFINE_MUTEX(vmx_l1d_flush_mutex); |
1815 |
++ |
1816 |
++/* Storage for pre module init parameter parsing */ |
1817 |
++static enum vmx_l1d_flush_state __read_mostly vmentry_l1d_flush_param = VMENTER_L1D_FLUSH_AUTO; |
1818 |
++ |
1819 |
++static const struct { |
1820 |
++ const char *option; |
1821 |
++ bool for_parse; |
1822 |
++} vmentry_l1d_param[] = { |
1823 |
++ [VMENTER_L1D_FLUSH_AUTO] = {"auto", true}, |
1824 |
++ [VMENTER_L1D_FLUSH_NEVER] = {"never", true}, |
1825 |
++ [VMENTER_L1D_FLUSH_COND] = {"cond", true}, |
1826 |
++ [VMENTER_L1D_FLUSH_ALWAYS] = {"always", true}, |
1827 |
++ [VMENTER_L1D_FLUSH_EPT_DISABLED] = {"EPT disabled", false}, |
1828 |
++ [VMENTER_L1D_FLUSH_NOT_REQUIRED] = {"not required", false}, |
1829 |
++}; |
1830 |
++ |
1831 |
++#define L1D_CACHE_ORDER 4 |
1832 |
++static void *vmx_l1d_flush_pages; |
1833 |
++ |
1834 |
++static int vmx_setup_l1d_flush(enum vmx_l1d_flush_state l1tf) |
1835 |
++{ |
1836 |
++ struct page *page; |
1837 |
++ unsigned int i; |
1838 |
++ |
1839 |
++ if (!boot_cpu_has_bug(X86_BUG_L1TF)) { |
1840 |
++ l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED; |
1841 |
++ return 0; |
1842 |
++ } |
1843 |
++ |
1844 |
++ if (!enable_ept) { |
1845 |
++ l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_EPT_DISABLED; |
1846 |
++ return 0; |
1847 |
++ } |
1848 |
++ |
1849 |
++ if (boot_cpu_has(X86_FEATURE_ARCH_CAPABILITIES)) { |
1850 |
++ u64 msr; |
1851 |
++ |
1852 |
++ rdmsrl(MSR_IA32_ARCH_CAPABILITIES, msr); |
1853 |
++ if (msr & ARCH_CAP_SKIP_VMENTRY_L1DFLUSH) { |
1854 |
++ l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED; |
1855 |
++ return 0; |
1856 |
++ } |
1857 |
++ } |
1858 |
++ |
1859 |
++ /* If set to auto use the default l1tf mitigation method */ |
1860 |
++ if (l1tf == VMENTER_L1D_FLUSH_AUTO) { |
1861 |
++ switch (l1tf_mitigation) { |
1862 |
++ case L1TF_MITIGATION_OFF: |
1863 |
++ l1tf = VMENTER_L1D_FLUSH_NEVER; |
1864 |
++ break; |
1865 |
++ case L1TF_MITIGATION_FLUSH_NOWARN: |
1866 |
++ case L1TF_MITIGATION_FLUSH: |
1867 |
++ case L1TF_MITIGATION_FLUSH_NOSMT: |
1868 |
++ l1tf = VMENTER_L1D_FLUSH_COND; |
1869 |
++ break; |
1870 |
++ case L1TF_MITIGATION_FULL: |
1871 |
++ case L1TF_MITIGATION_FULL_FORCE: |
1872 |
++ l1tf = VMENTER_L1D_FLUSH_ALWAYS; |
1873 |
++ break; |
1874 |
++ } |
1875 |
++ } else if (l1tf_mitigation == L1TF_MITIGATION_FULL_FORCE) { |
1876 |
++ l1tf = VMENTER_L1D_FLUSH_ALWAYS; |
1877 |
++ } |
1878 |
++ |
1879 |
++ if (l1tf != VMENTER_L1D_FLUSH_NEVER && !vmx_l1d_flush_pages && |
1880 |
++ !boot_cpu_has(X86_FEATURE_FLUSH_L1D)) { |
1881 |
++ /* |
1882 |
++ * This allocation for vmx_l1d_flush_pages is not tied to a VM |
1883 |
++ * lifetime and so should not be charged to a memcg. |
1884 |
++ */ |
1885 |
++ page = alloc_pages(GFP_KERNEL, L1D_CACHE_ORDER); |
1886 |
++ if (!page) |
1887 |
++ return -ENOMEM; |
1888 |
++ vmx_l1d_flush_pages = page_address(page); |
1889 |
++ |
1890 |
++ /* |
1891 |
++ * Initialize each page with a different pattern in |
1892 |
++ * order to protect against KSM in the nested |
1893 |
++ * virtualization case. |
1894 |
++ */ |
1895 |
++ for (i = 0; i < 1u << L1D_CACHE_ORDER; ++i) { |
1896 |
++ memset(vmx_l1d_flush_pages + i * PAGE_SIZE, i + 1, |
1897 |
++ PAGE_SIZE); |
1898 |
++ } |
1899 |
++ } |
1900 |
++ |
1901 |
++ l1tf_vmx_mitigation = l1tf; |
1902 |
++ |
1903 |
++ if (l1tf != VMENTER_L1D_FLUSH_NEVER) |
1904 |
++ static_branch_enable(&vmx_l1d_should_flush); |
1905 |
++ else |
1906 |
++ static_branch_disable(&vmx_l1d_should_flush); |
1907 |
++ |
1908 |
++ if (l1tf == VMENTER_L1D_FLUSH_COND) |
1909 |
++ static_branch_enable(&vmx_l1d_flush_cond); |
1910 |
++ else |
1911 |
++ static_branch_disable(&vmx_l1d_flush_cond); |
1912 |
++ return 0; |
1913 |
++} |
1914 |
++ |
1915 |
++static int vmentry_l1d_flush_parse(const char *s) |
1916 |
++{ |
1917 |
++ unsigned int i; |
1918 |
++ |
1919 |
++ if (s) { |
1920 |
++ for (i = 0; i < ARRAY_SIZE(vmentry_l1d_param); i++) { |
1921 |
++ if (vmentry_l1d_param[i].for_parse && |
1922 |
++ sysfs_streq(s, vmentry_l1d_param[i].option)) |
1923 |
++ return i; |
1924 |
++ } |
1925 |
++ } |
1926 |
++ return -EINVAL; |
1927 |
++} |
1928 |
++ |
1929 |
++static int vmentry_l1d_flush_set(const char *s, const struct kernel_param *kp) |
1930 |
++{ |
1931 |
++ int l1tf, ret; |
1932 |
++ |
1933 |
++ l1tf = vmentry_l1d_flush_parse(s); |
1934 |
++ if (l1tf < 0) |
1935 |
++ return l1tf; |
1936 |
++ |
1937 |
++ if (!boot_cpu_has(X86_BUG_L1TF)) |
1938 |
++ return 0; |
1939 |
++ |
1940 |
++ /* |
1941 |
++ * Has vmx_init() run already? If not then this is the pre init |
1942 |
++ * parameter parsing. In that case just store the value and let |
1943 |
++ * vmx_init() do the proper setup after enable_ept has been |
1944 |
++ * established. |
1945 |
++ */ |
1946 |
++ if (l1tf_vmx_mitigation == VMENTER_L1D_FLUSH_AUTO) { |
1947 |
++ vmentry_l1d_flush_param = l1tf; |
1948 |
++ return 0; |
1949 |
++ } |
1950 |
++ |
1951 |
++ mutex_lock(&vmx_l1d_flush_mutex); |
1952 |
++ ret = vmx_setup_l1d_flush(l1tf); |
1953 |
++ mutex_unlock(&vmx_l1d_flush_mutex); |
1954 |
++ return ret; |
1955 |
++} |
1956 |
++ |
1957 |
++static int vmentry_l1d_flush_get(char *s, const struct kernel_param *kp) |
1958 |
++{ |
1959 |
++ if (WARN_ON_ONCE(l1tf_vmx_mitigation >= ARRAY_SIZE(vmentry_l1d_param))) |
1960 |
++ return sprintf(s, "???\n"); |
1961 |
++ |
1962 |
++ return sprintf(s, "%s\n", vmentry_l1d_param[l1tf_vmx_mitigation].option); |
1963 |
++} |
1964 |
++ |
1965 |
++static const struct kernel_param_ops vmentry_l1d_flush_ops = { |
1966 |
++ .set = vmentry_l1d_flush_set, |
1967 |
++ .get = vmentry_l1d_flush_get, |
1968 |
++}; |
1969 |
++module_param_cb(vmentry_l1d_flush, &vmentry_l1d_flush_ops, NULL, 0644); |
1970 |
++ |
1971 |
++static bool guest_state_valid(struct kvm_vcpu *vcpu); |
1972 |
++static u32 vmx_segment_access_rights(struct kvm_segment *var); |
1973 |
++static __always_inline void vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, |
1974 |
++ u32 msr, int type); |
1975 |
++ |
1976 |
++void vmx_vmexit(void); |
1977 |
++ |
1978 |
++#define vmx_insn_failed(fmt...) \ |
1979 |
++do { \ |
1980 |
++ WARN_ONCE(1, fmt); \ |
1981 |
++ pr_warn_ratelimited(fmt); \ |
1982 |
++} while (0) |
1983 |
++ |
1984 |
++asmlinkage void vmread_error(unsigned long field, bool fault) |
1985 |
++{ |
1986 |
++ if (fault) |
1987 |
++ kvm_spurious_fault(); |
1988 |
++ else |
1989 |
++ vmx_insn_failed("kvm: vmread failed: field=%lx\n", field); |
1990 |
++} |
1991 |
++ |
1992 |
++noinline void vmwrite_error(unsigned long field, unsigned long value) |
1993 |
++{ |
1994 |
++ vmx_insn_failed("kvm: vmwrite failed: field=%lx val=%lx err=%d\n", |
1995 |
++ field, value, vmcs_read32(VM_INSTRUCTION_ERROR)); |
1996 |
++} |
1997 |
++ |
1998 |
++noinline void vmclear_error(struct vmcs *vmcs, u64 phys_addr) |
1999 |
++{ |
2000 |
++ vmx_insn_failed("kvm: vmclear failed: %p/%llx\n", vmcs, phys_addr); |
2001 |
++} |
2002 |
++ |
2003 |
++noinline void vmptrld_error(struct vmcs *vmcs, u64 phys_addr) |
2004 |
++{ |
2005 |
++ vmx_insn_failed("kvm: vmptrld failed: %p/%llx\n", vmcs, phys_addr); |
2006 |
++} |
2007 |
++ |
2008 |
++noinline void invvpid_error(unsigned long ext, u16 vpid, gva_t gva) |
2009 |
++{ |
2010 |
++ vmx_insn_failed("kvm: invvpid failed: ext=0x%lx vpid=%u gva=0x%lx\n", |
2011 |
++ ext, vpid, gva); |
2012 |
++} |
2013 |
++ |
2014 |
++noinline void invept_error(unsigned long ext, u64 eptp, gpa_t gpa) |
2015 |
++{ |
2016 |
++ vmx_insn_failed("kvm: invept failed: ext=0x%lx eptp=%llx gpa=0x%llx\n", |
2017 |
++ ext, eptp, gpa); |
2018 |
++} |
2019 |
++ |
2020 |
++static DEFINE_PER_CPU(struct vmcs *, vmxarea); |
2021 |
++DEFINE_PER_CPU(struct vmcs *, current_vmcs); |
2022 |
++/* |
2023 |
++ * We maintain a per-CPU linked-list of VMCS loaded on that CPU. This is needed |
2024 |
++ * when a CPU is brought down, and we need to VMCLEAR all VMCSs loaded on it. |
2025 |
++ */ |
2026 |
++static DEFINE_PER_CPU(struct list_head, loaded_vmcss_on_cpu); |
2027 |
++ |
2028 |
++/* |
2029 |
++ * We maintian a per-CPU linked-list of vCPU, so in wakeup_handler() we |
2030 |
++ * can find which vCPU should be waken up. |
2031 |
++ */ |
2032 |
++static DEFINE_PER_CPU(struct list_head, blocked_vcpu_on_cpu); |
2033 |
++static DEFINE_PER_CPU(spinlock_t, blocked_vcpu_on_cpu_lock); |
2034 |
++ |
2035 |
++static DECLARE_BITMAP(vmx_vpid_bitmap, VMX_NR_VPIDS); |
2036 |
++static DEFINE_SPINLOCK(vmx_vpid_lock); |
2037 |
++ |
2038 |
++struct vmcs_config vmcs_config; |
2039 |
++struct vmx_capability vmx_capability; |
2040 |
++ |
2041 |
++#define VMX_SEGMENT_FIELD(seg) \ |
2042 |
++ [VCPU_SREG_##seg] = { \ |
2043 |
++ .selector = GUEST_##seg##_SELECTOR, \ |
2044 |
++ .base = GUEST_##seg##_BASE, \ |
2045 |
++ .limit = GUEST_##seg##_LIMIT, \ |
2046 |
++ .ar_bytes = GUEST_##seg##_AR_BYTES, \ |
2047 |
++ } |
2048 |
++ |
2049 |
++static const struct kvm_vmx_segment_field { |
2050 |
++ unsigned selector; |
2051 |
++ unsigned base; |
2052 |
++ unsigned limit; |
2053 |
++ unsigned ar_bytes; |
2054 |
++} kvm_vmx_segment_fields[] = { |
2055 |
++ VMX_SEGMENT_FIELD(CS), |
2056 |
++ VMX_SEGMENT_FIELD(DS), |
2057 |
++ VMX_SEGMENT_FIELD(ES), |
2058 |
++ VMX_SEGMENT_FIELD(FS), |
2059 |
++ VMX_SEGMENT_FIELD(GS), |
2060 |
++ VMX_SEGMENT_FIELD(SS), |
2061 |
++ VMX_SEGMENT_FIELD(TR), |
2062 |
++ VMX_SEGMENT_FIELD(LDTR), |
2063 |
++}; |
2064 |
++ |
2065 |
++u64 host_efer; |
2066 |
++static unsigned long host_idt_base; |
2067 |
++ |
2068 |
++/* |
2069 |
++ * Though SYSCALL is only supported in 64-bit mode on Intel CPUs, kvm |
2070 |
++ * will emulate SYSCALL in legacy mode if the vendor string in guest |
2071 |
++ * CPUID.0:{EBX,ECX,EDX} is "AuthenticAMD" or "AMDisbetter!" To |
2072 |
++ * support this emulation, IA32_STAR must always be included in |
2073 |
++ * vmx_msr_index[], even in i386 builds. |
2074 |
++ */ |
2075 |
++const u32 vmx_msr_index[] = { |
2076 |
++#ifdef CONFIG_X86_64 |
2077 |
++ MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, |
2078 |
++#endif |
2079 |
++ MSR_EFER, MSR_TSC_AUX, MSR_STAR, |
2080 |
++ MSR_IA32_TSX_CTRL, |
2081 |
++}; |
2082 |
++ |
2083 |
++#if IS_ENABLED(CONFIG_HYPERV) |
2084 |
++static bool __read_mostly enlightened_vmcs = true; |
2085 |
++module_param(enlightened_vmcs, bool, 0444); |
2086 |
++ |
2087 |
++/* check_ept_pointer() should be under protection of ept_pointer_lock. */ |
2088 |
++static void check_ept_pointer_match(struct kvm *kvm) |
2089 |
++{ |
2090 |
++ struct kvm_vcpu *vcpu; |
2091 |
++ u64 tmp_eptp = INVALID_PAGE; |
2092 |
++ int i; |
2093 |
++ |
2094 |
++ kvm_for_each_vcpu(i, vcpu, kvm) { |
2095 |
++ if (!VALID_PAGE(tmp_eptp)) { |
2096 |
++ tmp_eptp = to_vmx(vcpu)->ept_pointer; |
2097 |
++ } else if (tmp_eptp != to_vmx(vcpu)->ept_pointer) { |
2098 |
++ to_kvm_vmx(kvm)->ept_pointers_match |
2099 |
++ = EPT_POINTERS_MISMATCH; |
2100 |
++ return; |
2101 |
++ } |
2102 |
++ } |
2103 |
++ |
2104 |
++ to_kvm_vmx(kvm)->ept_pointers_match = EPT_POINTERS_MATCH; |
2105 |
++} |
2106 |
++ |
2107 |
++static int kvm_fill_hv_flush_list_func(struct hv_guest_mapping_flush_list *flush, |
2108 |
++ void *data) |
2109 |
++{ |
2110 |
++ struct kvm_tlb_range *range = data; |
2111 |
++ |
2112 |
++ return hyperv_fill_flush_guest_mapping_list(flush, range->start_gfn, |
2113 |
++ range->pages); |
2114 |
++} |
2115 |
++ |
2116 |
++static inline int __hv_remote_flush_tlb_with_range(struct kvm *kvm, |
2117 |
++ struct kvm_vcpu *vcpu, struct kvm_tlb_range *range) |
2118 |
++{ |
2119 |
++ u64 ept_pointer = to_vmx(vcpu)->ept_pointer; |
2120 |
++ |
2121 |
++ /* |
2122 |
++ * FLUSH_GUEST_PHYSICAL_ADDRESS_SPACE hypercall needs address |
2123 |
++ * of the base of EPT PML4 table, strip off EPT configuration |
2124 |
++ * information. |
2125 |
++ */ |
2126 |
++ if (range) |
2127 |
++ return hyperv_flush_guest_mapping_range(ept_pointer & PAGE_MASK, |
2128 |
++ kvm_fill_hv_flush_list_func, (void *)range); |
2129 |
++ else |
2130 |
++ return hyperv_flush_guest_mapping(ept_pointer & PAGE_MASK); |
2131 |
++} |
2132 |
++ |
2133 |
++static int hv_remote_flush_tlb_with_range(struct kvm *kvm, |
2134 |
++ struct kvm_tlb_range *range) |
2135 |
++{ |
2136 |
++ struct kvm_vcpu *vcpu; |
2137 |
++ int ret = 0, i; |
2138 |
++ |
2139 |
++ spin_lock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
2140 |
++ |
2141 |
++ if (to_kvm_vmx(kvm)->ept_pointers_match == EPT_POINTERS_CHECK) |
2142 |
++ check_ept_pointer_match(kvm); |
2143 |
++ |
2144 |
++ if (to_kvm_vmx(kvm)->ept_pointers_match != EPT_POINTERS_MATCH) { |
2145 |
++ kvm_for_each_vcpu(i, vcpu, kvm) { |
2146 |
++ /* If ept_pointer is invalid pointer, bypass flush request. */ |
2147 |
++ if (VALID_PAGE(to_vmx(vcpu)->ept_pointer)) |
2148 |
++ ret |= __hv_remote_flush_tlb_with_range( |
2149 |
++ kvm, vcpu, range); |
2150 |
++ } |
2151 |
++ } else { |
2152 |
++ ret = __hv_remote_flush_tlb_with_range(kvm, |
2153 |
++ kvm_get_vcpu(kvm, 0), range); |
2154 |
++ } |
2155 |
++ |
2156 |
++ spin_unlock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
2157 |
++ return ret; |
2158 |
++} |
2159 |
++static int hv_remote_flush_tlb(struct kvm *kvm) |
2160 |
++{ |
2161 |
++ return hv_remote_flush_tlb_with_range(kvm, NULL); |
2162 |
++} |
2163 |
++ |
2164 |
++static int hv_enable_direct_tlbflush(struct kvm_vcpu *vcpu) |
2165 |
++{ |
2166 |
++ struct hv_enlightened_vmcs *evmcs; |
2167 |
++ struct hv_partition_assist_pg **p_hv_pa_pg = |
2168 |
++ &vcpu->kvm->arch.hyperv.hv_pa_pg; |
2169 |
++ /* |
2170 |
++ * Synthetic VM-Exit is not enabled in current code and so All |
2171 |
++ * evmcs in singe VM shares same assist page. |
2172 |
++ */ |
2173 |
++ if (!*p_hv_pa_pg) |
2174 |
++ *p_hv_pa_pg = kzalloc(PAGE_SIZE, GFP_KERNEL); |
2175 |
++ |
2176 |
++ if (!*p_hv_pa_pg) |
2177 |
++ return -ENOMEM; |
2178 |
++ |
2179 |
++ evmcs = (struct hv_enlightened_vmcs *)to_vmx(vcpu)->loaded_vmcs->vmcs; |
2180 |
++ |
2181 |
++ evmcs->partition_assist_page = |
2182 |
++ __pa(*p_hv_pa_pg); |
2183 |
++ evmcs->hv_vm_id = (unsigned long)vcpu->kvm; |
2184 |
++ evmcs->hv_enlightenments_control.nested_flush_hypercall = 1; |
2185 |
++ |
2186 |
++ return 0; |
2187 |
++} |
2188 |
++ |
2189 |
++#endif /* IS_ENABLED(CONFIG_HYPERV) */ |
2190 |
++ |
2191 |
++/* |
2192 |
++ * Comment's format: document - errata name - stepping - processor name. |
2193 |
++ * Refer from |
2194 |
++ * https://www.virtualbox.org/svn/vbox/trunk/src/VBox/VMM/VMMR0/HMR0.cpp |
2195 |
++ */ |
2196 |
++static u32 vmx_preemption_cpu_tfms[] = { |
2197 |
++/* 323344.pdf - BA86 - D0 - Xeon 7500 Series */ |
2198 |
++0x000206E6, |
2199 |
++/* 323056.pdf - AAX65 - C2 - Xeon L3406 */ |
2200 |
++/* 322814.pdf - AAT59 - C2 - i7-600, i5-500, i5-400 and i3-300 Mobile */ |
2201 |
++/* 322911.pdf - AAU65 - C2 - i5-600, i3-500 Desktop and Pentium G6950 */ |
2202 |
++0x00020652, |
2203 |
++/* 322911.pdf - AAU65 - K0 - i5-600, i3-500 Desktop and Pentium G6950 */ |
2204 |
++0x00020655, |
2205 |
++/* 322373.pdf - AAO95 - B1 - Xeon 3400 Series */ |
2206 |
++/* 322166.pdf - AAN92 - B1 - i7-800 and i5-700 Desktop */ |
2207 |
++/* |
2208 |
++ * 320767.pdf - AAP86 - B1 - |
2209 |
++ * i7-900 Mobile Extreme, i7-800 and i7-700 Mobile |
2210 |
++ */ |
2211 |
++0x000106E5, |
2212 |
++/* 321333.pdf - AAM126 - C0 - Xeon 3500 */ |
2213 |
++0x000106A0, |
2214 |
++/* 321333.pdf - AAM126 - C1 - Xeon 3500 */ |
2215 |
++0x000106A1, |
2216 |
++/* 320836.pdf - AAJ124 - C0 - i7-900 Desktop Extreme and i7-900 Desktop */ |
2217 |
++0x000106A4, |
2218 |
++ /* 321333.pdf - AAM126 - D0 - Xeon 3500 */ |
2219 |
++ /* 321324.pdf - AAK139 - D0 - Xeon 5500 */ |
2220 |
++ /* 320836.pdf - AAJ124 - D0 - i7-900 Extreme and i7-900 Desktop */ |
2221 |
++0x000106A5, |
2222 |
++ /* Xeon E3-1220 V2 */ |
2223 |
++0x000306A8, |
2224 |
++}; |
2225 |
++ |
2226 |
++static inline bool cpu_has_broken_vmx_preemption_timer(void) |
2227 |
++{ |
2228 |
++ u32 eax = cpuid_eax(0x00000001), i; |
2229 |
++ |
2230 |
++ /* Clear the reserved bits */ |
2231 |
++ eax &= ~(0x3U << 14 | 0xfU << 28); |
2232 |
++ for (i = 0; i < ARRAY_SIZE(vmx_preemption_cpu_tfms); i++) |
2233 |
++ if (eax == vmx_preemption_cpu_tfms[i]) |
2234 |
++ return true; |
2235 |
++ |
2236 |
++ return false; |
2237 |
++} |
2238 |
++ |
2239 |
++static inline bool cpu_need_virtualize_apic_accesses(struct kvm_vcpu *vcpu) |
2240 |
++{ |
2241 |
++ return flexpriority_enabled && lapic_in_kernel(vcpu); |
2242 |
++} |
2243 |
++ |
2244 |
++static inline bool report_flexpriority(void) |
2245 |
++{ |
2246 |
++ return flexpriority_enabled; |
2247 |
++} |
2248 |
++ |
2249 |
++static inline int __find_msr_index(struct vcpu_vmx *vmx, u32 msr) |
2250 |
++{ |
2251 |
++ int i; |
2252 |
++ |
2253 |
++ for (i = 0; i < vmx->nmsrs; ++i) |
2254 |
++ if (vmx_msr_index[vmx->guest_msrs[i].index] == msr) |
2255 |
++ return i; |
2256 |
++ return -1; |
2257 |
++} |
2258 |
++ |
2259 |
++struct shared_msr_entry *find_msr_entry(struct vcpu_vmx *vmx, u32 msr) |
2260 |
++{ |
2261 |
++ int i; |
2262 |
++ |
2263 |
++ i = __find_msr_index(vmx, msr); |
2264 |
++ if (i >= 0) |
2265 |
++ return &vmx->guest_msrs[i]; |
2266 |
++ return NULL; |
2267 |
++} |
2268 |
++ |
2269 |
++static int vmx_set_guest_msr(struct vcpu_vmx *vmx, struct shared_msr_entry *msr, u64 data) |
2270 |
++{ |
2271 |
++ int ret = 0; |
2272 |
++ |
2273 |
++ u64 old_msr_data = msr->data; |
2274 |
++ msr->data = data; |
2275 |
++ if (msr - vmx->guest_msrs < vmx->save_nmsrs) { |
2276 |
++ preempt_disable(); |
2277 |
++ ret = kvm_set_shared_msr(msr->index, msr->data, |
2278 |
++ msr->mask); |
2279 |
++ preempt_enable(); |
2280 |
++ if (ret) |
2281 |
++ msr->data = old_msr_data; |
2282 |
++ } |
2283 |
++ return ret; |
2284 |
++} |
2285 |
++ |
2286 |
++void loaded_vmcs_init(struct loaded_vmcs *loaded_vmcs) |
2287 |
++{ |
2288 |
++ vmcs_clear(loaded_vmcs->vmcs); |
2289 |
++ if (loaded_vmcs->shadow_vmcs && loaded_vmcs->launched) |
2290 |
++ vmcs_clear(loaded_vmcs->shadow_vmcs); |
2291 |
++ loaded_vmcs->cpu = -1; |
2292 |
++ loaded_vmcs->launched = 0; |
2293 |
++} |
2294 |
++ |
2295 |
++#ifdef CONFIG_KEXEC_CORE |
2296 |
++/* |
2297 |
++ * This bitmap is used to indicate whether the vmclear |
2298 |
++ * operation is enabled on all cpus. All disabled by |
2299 |
++ * default. |
2300 |
++ */ |
2301 |
++static cpumask_t crash_vmclear_enabled_bitmap = CPU_MASK_NONE; |
2302 |
++ |
2303 |
++static inline void crash_enable_local_vmclear(int cpu) |
2304 |
++{ |
2305 |
++ cpumask_set_cpu(cpu, &crash_vmclear_enabled_bitmap); |
2306 |
++} |
2307 |
++ |
2308 |
++static inline void crash_disable_local_vmclear(int cpu) |
2309 |
++{ |
2310 |
++ cpumask_clear_cpu(cpu, &crash_vmclear_enabled_bitmap); |
2311 |
++} |
2312 |
++ |
2313 |
++static inline int crash_local_vmclear_enabled(int cpu) |
2314 |
++{ |
2315 |
++ return cpumask_test_cpu(cpu, &crash_vmclear_enabled_bitmap); |
2316 |
++} |
2317 |
++ |
2318 |
++static void crash_vmclear_local_loaded_vmcss(void) |
2319 |
++{ |
2320 |
++ int cpu = raw_smp_processor_id(); |
2321 |
++ struct loaded_vmcs *v; |
2322 |
++ |
2323 |
++ if (!crash_local_vmclear_enabled(cpu)) |
2324 |
++ return; |
2325 |
++ |
2326 |
++ list_for_each_entry(v, &per_cpu(loaded_vmcss_on_cpu, cpu), |
2327 |
++ loaded_vmcss_on_cpu_link) |
2328 |
++ vmcs_clear(v->vmcs); |
2329 |
++} |
2330 |
++#else |
2331 |
++static inline void crash_enable_local_vmclear(int cpu) { } |
2332 |
++static inline void crash_disable_local_vmclear(int cpu) { } |
2333 |
++#endif /* CONFIG_KEXEC_CORE */ |
2334 |
++ |
2335 |
++static void __loaded_vmcs_clear(void *arg) |
2336 |
++{ |
2337 |
++ struct loaded_vmcs *loaded_vmcs = arg; |
2338 |
++ int cpu = raw_smp_processor_id(); |
2339 |
++ |
2340 |
++ if (loaded_vmcs->cpu != cpu) |
2341 |
++ return; /* vcpu migration can race with cpu offline */ |
2342 |
++ if (per_cpu(current_vmcs, cpu) == loaded_vmcs->vmcs) |
2343 |
++ per_cpu(current_vmcs, cpu) = NULL; |
2344 |
++ crash_disable_local_vmclear(cpu); |
2345 |
++ list_del(&loaded_vmcs->loaded_vmcss_on_cpu_link); |
2346 |
++ |
2347 |
++ /* |
2348 |
++ * we should ensure updating loaded_vmcs->loaded_vmcss_on_cpu_link |
2349 |
++ * is before setting loaded_vmcs->vcpu to -1 which is done in |
2350 |
++ * loaded_vmcs_init. Otherwise, other cpu can see vcpu = -1 fist |
2351 |
++ * then adds the vmcs into percpu list before it is deleted. |
2352 |
++ */ |
2353 |
++ smp_wmb(); |
2354 |
++ |
2355 |
++ loaded_vmcs_init(loaded_vmcs); |
2356 |
++ crash_enable_local_vmclear(cpu); |
2357 |
++} |
2358 |
++ |
2359 |
++void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs) |
2360 |
++{ |
2361 |
++ int cpu = loaded_vmcs->cpu; |
2362 |
++ |
2363 |
++ if (cpu != -1) |
2364 |
++ smp_call_function_single(cpu, |
2365 |
++ __loaded_vmcs_clear, loaded_vmcs, 1); |
2366 |
++} |
2367 |
++ |
2368 |
++static bool vmx_segment_cache_test_set(struct vcpu_vmx *vmx, unsigned seg, |
2369 |
++ unsigned field) |
2370 |
++{ |
2371 |
++ bool ret; |
2372 |
++ u32 mask = 1 << (seg * SEG_FIELD_NR + field); |
2373 |
++ |
2374 |
++ if (!kvm_register_is_available(&vmx->vcpu, VCPU_EXREG_SEGMENTS)) { |
2375 |
++ kvm_register_mark_available(&vmx->vcpu, VCPU_EXREG_SEGMENTS); |
2376 |
++ vmx->segment_cache.bitmask = 0; |
2377 |
++ } |
2378 |
++ ret = vmx->segment_cache.bitmask & mask; |
2379 |
++ vmx->segment_cache.bitmask |= mask; |
2380 |
++ return ret; |
2381 |
++} |
2382 |
++ |
2383 |
++static u16 vmx_read_guest_seg_selector(struct vcpu_vmx *vmx, unsigned seg) |
2384 |
++{ |
2385 |
++ u16 *p = &vmx->segment_cache.seg[seg].selector; |
2386 |
++ |
2387 |
++ if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_SEL)) |
2388 |
++ *p = vmcs_read16(kvm_vmx_segment_fields[seg].selector); |
2389 |
++ return *p; |
2390 |
++} |
2391 |
++ |
2392 |
++static ulong vmx_read_guest_seg_base(struct vcpu_vmx *vmx, unsigned seg) |
2393 |
++{ |
2394 |
++ ulong *p = &vmx->segment_cache.seg[seg].base; |
2395 |
++ |
2396 |
++ if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_BASE)) |
2397 |
++ *p = vmcs_readl(kvm_vmx_segment_fields[seg].base); |
2398 |
++ return *p; |
2399 |
++} |
2400 |
++ |
2401 |
++static u32 vmx_read_guest_seg_limit(struct vcpu_vmx *vmx, unsigned seg) |
2402 |
++{ |
2403 |
++ u32 *p = &vmx->segment_cache.seg[seg].limit; |
2404 |
++ |
2405 |
++ if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_LIMIT)) |
2406 |
++ *p = vmcs_read32(kvm_vmx_segment_fields[seg].limit); |
2407 |
++ return *p; |
2408 |
++} |
2409 |
++ |
2410 |
++static u32 vmx_read_guest_seg_ar(struct vcpu_vmx *vmx, unsigned seg) |
2411 |
++{ |
2412 |
++ u32 *p = &vmx->segment_cache.seg[seg].ar; |
2413 |
++ |
2414 |
++ if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_AR)) |
2415 |
++ *p = vmcs_read32(kvm_vmx_segment_fields[seg].ar_bytes); |
2416 |
++ return *p; |
2417 |
++} |
2418 |
++ |
2419 |
++void update_exception_bitmap(struct kvm_vcpu *vcpu) |
2420 |
++{ |
2421 |
++ u32 eb; |
2422 |
++ |
2423 |
++ eb = (1u << PF_VECTOR) | (1u << UD_VECTOR) | (1u << MC_VECTOR) | |
2424 |
++ (1u << DB_VECTOR) | (1u << AC_VECTOR); |
2425 |
++ /* |
2426 |
++ * Guest access to VMware backdoor ports could legitimately |
2427 |
++ * trigger #GP because of TSS I/O permission bitmap. |
2428 |
++ * We intercept those #GP and allow access to them anyway |
2429 |
++ * as VMware does. |
2430 |
++ */ |
2431 |
++ if (enable_vmware_backdoor) |
2432 |
++ eb |= (1u << GP_VECTOR); |
2433 |
++ if ((vcpu->guest_debug & |
2434 |
++ (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) == |
2435 |
++ (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) |
2436 |
++ eb |= 1u << BP_VECTOR; |
2437 |
++ if (to_vmx(vcpu)->rmode.vm86_active) |
2438 |
++ eb = ~0; |
2439 |
++ if (enable_ept) |
2440 |
++ eb &= ~(1u << PF_VECTOR); /* bypass_guest_pf = 0 */ |
2441 |
++ |
2442 |
++ /* When we are running a nested L2 guest and L1 specified for it a |
2443 |
++ * certain exception bitmap, we must trap the same exceptions and pass |
2444 |
++ * them to L1. When running L2, we will only handle the exceptions |
2445 |
++ * specified above if L1 did not want them. |
2446 |
++ */ |
2447 |
++ if (is_guest_mode(vcpu)) |
2448 |
++ eb |= get_vmcs12(vcpu)->exception_bitmap; |
2449 |
++ |
2450 |
++ vmcs_write32(EXCEPTION_BITMAP, eb); |
2451 |
++} |
2452 |
++ |
2453 |
++/* |
2454 |
++ * Check if MSR is intercepted for currently loaded MSR bitmap. |
2455 |
++ */ |
2456 |
++static bool msr_write_intercepted(struct kvm_vcpu *vcpu, u32 msr) |
2457 |
++{ |
2458 |
++ unsigned long *msr_bitmap; |
2459 |
++ int f = sizeof(unsigned long); |
2460 |
++ |
2461 |
++ if (!cpu_has_vmx_msr_bitmap()) |
2462 |
++ return true; |
2463 |
++ |
2464 |
++ msr_bitmap = to_vmx(vcpu)->loaded_vmcs->msr_bitmap; |
2465 |
++ |
2466 |
++ if (msr <= 0x1fff) { |
2467 |
++ return !!test_bit(msr, msr_bitmap + 0x800 / f); |
2468 |
++ } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
2469 |
++ msr &= 0x1fff; |
2470 |
++ return !!test_bit(msr, msr_bitmap + 0xc00 / f); |
2471 |
++ } |
2472 |
++ |
2473 |
++ return true; |
2474 |
++} |
2475 |
++ |
2476 |
++static void clear_atomic_switch_msr_special(struct vcpu_vmx *vmx, |
2477 |
++ unsigned long entry, unsigned long exit) |
2478 |
++{ |
2479 |
++ vm_entry_controls_clearbit(vmx, entry); |
2480 |
++ vm_exit_controls_clearbit(vmx, exit); |
2481 |
++} |
2482 |
++ |
2483 |
++int vmx_find_msr_index(struct vmx_msrs *m, u32 msr) |
2484 |
++{ |
2485 |
++ unsigned int i; |
2486 |
++ |
2487 |
++ for (i = 0; i < m->nr; ++i) { |
2488 |
++ if (m->val[i].index == msr) |
2489 |
++ return i; |
2490 |
++ } |
2491 |
++ return -ENOENT; |
2492 |
++} |
2493 |
++ |
2494 |
++static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) |
2495 |
++{ |
2496 |
++ int i; |
2497 |
++ struct msr_autoload *m = &vmx->msr_autoload; |
2498 |
++ |
2499 |
++ switch (msr) { |
2500 |
++ case MSR_EFER: |
2501 |
++ if (cpu_has_load_ia32_efer()) { |
2502 |
++ clear_atomic_switch_msr_special(vmx, |
2503 |
++ VM_ENTRY_LOAD_IA32_EFER, |
2504 |
++ VM_EXIT_LOAD_IA32_EFER); |
2505 |
++ return; |
2506 |
++ } |
2507 |
++ break; |
2508 |
++ case MSR_CORE_PERF_GLOBAL_CTRL: |
2509 |
++ if (cpu_has_load_perf_global_ctrl()) { |
2510 |
++ clear_atomic_switch_msr_special(vmx, |
2511 |
++ VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, |
2512 |
++ VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); |
2513 |
++ return; |
2514 |
++ } |
2515 |
++ break; |
2516 |
++ } |
2517 |
++ i = vmx_find_msr_index(&m->guest, msr); |
2518 |
++ if (i < 0) |
2519 |
++ goto skip_guest; |
2520 |
++ --m->guest.nr; |
2521 |
++ m->guest.val[i] = m->guest.val[m->guest.nr]; |
2522 |
++ vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->guest.nr); |
2523 |
++ |
2524 |
++skip_guest: |
2525 |
++ i = vmx_find_msr_index(&m->host, msr); |
2526 |
++ if (i < 0) |
2527 |
++ return; |
2528 |
++ |
2529 |
++ --m->host.nr; |
2530 |
++ m->host.val[i] = m->host.val[m->host.nr]; |
2531 |
++ vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->host.nr); |
2532 |
++} |
2533 |
++ |
2534 |
++static void add_atomic_switch_msr_special(struct vcpu_vmx *vmx, |
2535 |
++ unsigned long entry, unsigned long exit, |
2536 |
++ unsigned long guest_val_vmcs, unsigned long host_val_vmcs, |
2537 |
++ u64 guest_val, u64 host_val) |
2538 |
++{ |
2539 |
++ vmcs_write64(guest_val_vmcs, guest_val); |
2540 |
++ if (host_val_vmcs != HOST_IA32_EFER) |
2541 |
++ vmcs_write64(host_val_vmcs, host_val); |
2542 |
++ vm_entry_controls_setbit(vmx, entry); |
2543 |
++ vm_exit_controls_setbit(vmx, exit); |
2544 |
++} |
2545 |
++ |
2546 |
++static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, |
2547 |
++ u64 guest_val, u64 host_val, bool entry_only) |
2548 |
++{ |
2549 |
++ int i, j = 0; |
2550 |
++ struct msr_autoload *m = &vmx->msr_autoload; |
2551 |
++ |
2552 |
++ switch (msr) { |
2553 |
++ case MSR_EFER: |
2554 |
++ if (cpu_has_load_ia32_efer()) { |
2555 |
++ add_atomic_switch_msr_special(vmx, |
2556 |
++ VM_ENTRY_LOAD_IA32_EFER, |
2557 |
++ VM_EXIT_LOAD_IA32_EFER, |
2558 |
++ GUEST_IA32_EFER, |
2559 |
++ HOST_IA32_EFER, |
2560 |
++ guest_val, host_val); |
2561 |
++ return; |
2562 |
++ } |
2563 |
++ break; |
2564 |
++ case MSR_CORE_PERF_GLOBAL_CTRL: |
2565 |
++ if (cpu_has_load_perf_global_ctrl()) { |
2566 |
++ add_atomic_switch_msr_special(vmx, |
2567 |
++ VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, |
2568 |
++ VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL, |
2569 |
++ GUEST_IA32_PERF_GLOBAL_CTRL, |
2570 |
++ HOST_IA32_PERF_GLOBAL_CTRL, |
2571 |
++ guest_val, host_val); |
2572 |
++ return; |
2573 |
++ } |
2574 |
++ break; |
2575 |
++ case MSR_IA32_PEBS_ENABLE: |
2576 |
++ /* PEBS needs a quiescent period after being disabled (to write |
2577 |
++ * a record). Disabling PEBS through VMX MSR swapping doesn't |
2578 |
++ * provide that period, so a CPU could write host's record into |
2579 |
++ * guest's memory. |
2580 |
++ */ |
2581 |
++ wrmsrl(MSR_IA32_PEBS_ENABLE, 0); |
2582 |
++ } |
2583 |
++ |
2584 |
++ i = vmx_find_msr_index(&m->guest, msr); |
2585 |
++ if (!entry_only) |
2586 |
++ j = vmx_find_msr_index(&m->host, msr); |
2587 |
++ |
2588 |
++ if ((i < 0 && m->guest.nr == NR_LOADSTORE_MSRS) || |
2589 |
++ (j < 0 && m->host.nr == NR_LOADSTORE_MSRS)) { |
2590 |
++ printk_once(KERN_WARNING "Not enough msr switch entries. " |
2591 |
++ "Can't add msr %x\n", msr); |
2592 |
++ return; |
2593 |
++ } |
2594 |
++ if (i < 0) { |
2595 |
++ i = m->guest.nr++; |
2596 |
++ vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->guest.nr); |
2597 |
++ } |
2598 |
++ m->guest.val[i].index = msr; |
2599 |
++ m->guest.val[i].value = guest_val; |
2600 |
++ |
2601 |
++ if (entry_only) |
2602 |
++ return; |
2603 |
++ |
2604 |
++ if (j < 0) { |
2605 |
++ j = m->host.nr++; |
2606 |
++ vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->host.nr); |
2607 |
++ } |
2608 |
++ m->host.val[j].index = msr; |
2609 |
++ m->host.val[j].value = host_val; |
2610 |
++} |
2611 |
++ |
2612 |
++static bool update_transition_efer(struct vcpu_vmx *vmx, int efer_offset) |
2613 |
++{ |
2614 |
++ u64 guest_efer = vmx->vcpu.arch.efer; |
2615 |
++ u64 ignore_bits = 0; |
2616 |
++ |
2617 |
++ /* Shadow paging assumes NX to be available. */ |
2618 |
++ if (!enable_ept) |
2619 |
++ guest_efer |= EFER_NX; |
2620 |
++ |
2621 |
++ /* |
2622 |
++ * LMA and LME handled by hardware; SCE meaningless outside long mode. |
2623 |
++ */ |
2624 |
++ ignore_bits |= EFER_SCE; |
2625 |
++#ifdef CONFIG_X86_64 |
2626 |
++ ignore_bits |= EFER_LMA | EFER_LME; |
2627 |
++ /* SCE is meaningful only in long mode on Intel */ |
2628 |
++ if (guest_efer & EFER_LMA) |
2629 |
++ ignore_bits &= ~(u64)EFER_SCE; |
2630 |
++#endif |
2631 |
++ |
2632 |
++ /* |
2633 |
++ * On EPT, we can't emulate NX, so we must switch EFER atomically. |
2634 |
++ * On CPUs that support "load IA32_EFER", always switch EFER |
2635 |
++ * atomically, since it's faster than switching it manually. |
2636 |
++ */ |
2637 |
++ if (cpu_has_load_ia32_efer() || |
2638 |
++ (enable_ept && ((vmx->vcpu.arch.efer ^ host_efer) & EFER_NX))) { |
2639 |
++ if (!(guest_efer & EFER_LMA)) |
2640 |
++ guest_efer &= ~EFER_LME; |
2641 |
++ if (guest_efer != host_efer) |
2642 |
++ add_atomic_switch_msr(vmx, MSR_EFER, |
2643 |
++ guest_efer, host_efer, false); |
2644 |
++ else |
2645 |
++ clear_atomic_switch_msr(vmx, MSR_EFER); |
2646 |
++ return false; |
2647 |
++ } else { |
2648 |
++ clear_atomic_switch_msr(vmx, MSR_EFER); |
2649 |
++ |
2650 |
++ guest_efer &= ~ignore_bits; |
2651 |
++ guest_efer |= host_efer & ignore_bits; |
2652 |
++ |
2653 |
++ vmx->guest_msrs[efer_offset].data = guest_efer; |
2654 |
++ vmx->guest_msrs[efer_offset].mask = ~ignore_bits; |
2655 |
++ |
2656 |
++ return true; |
2657 |
++ } |
2658 |
++} |
2659 |
++ |
2660 |
++#ifdef CONFIG_X86_32 |
2661 |
++/* |
2662 |
++ * On 32-bit kernels, VM exits still load the FS and GS bases from the |
2663 |
++ * VMCS rather than the segment table. KVM uses this helper to figure |
2664 |
++ * out the current bases to poke them into the VMCS before entry. |
2665 |
++ */ |
2666 |
++static unsigned long segment_base(u16 selector) |
2667 |
++{ |
2668 |
++ struct desc_struct *table; |
2669 |
++ unsigned long v; |
2670 |
++ |
2671 |
++ if (!(selector & ~SEGMENT_RPL_MASK)) |
2672 |
++ return 0; |
2673 |
++ |
2674 |
++ table = get_current_gdt_ro(); |
2675 |
++ |
2676 |
++ if ((selector & SEGMENT_TI_MASK) == SEGMENT_LDT) { |
2677 |
++ u16 ldt_selector = kvm_read_ldt(); |
2678 |
++ |
2679 |
++ if (!(ldt_selector & ~SEGMENT_RPL_MASK)) |
2680 |
++ return 0; |
2681 |
++ |
2682 |
++ table = (struct desc_struct *)segment_base(ldt_selector); |
2683 |
++ } |
2684 |
++ v = get_desc_base(&table[selector >> 3]); |
2685 |
++ return v; |
2686 |
++} |
2687 |
++#endif |
2688 |
++ |
2689 |
++static inline void pt_load_msr(struct pt_ctx *ctx, u32 addr_range) |
2690 |
++{ |
2691 |
++ u32 i; |
2692 |
++ |
2693 |
++ wrmsrl(MSR_IA32_RTIT_STATUS, ctx->status); |
2694 |
++ wrmsrl(MSR_IA32_RTIT_OUTPUT_BASE, ctx->output_base); |
2695 |
++ wrmsrl(MSR_IA32_RTIT_OUTPUT_MASK, ctx->output_mask); |
2696 |
++ wrmsrl(MSR_IA32_RTIT_CR3_MATCH, ctx->cr3_match); |
2697 |
++ for (i = 0; i < addr_range; i++) { |
2698 |
++ wrmsrl(MSR_IA32_RTIT_ADDR0_A + i * 2, ctx->addr_a[i]); |
2699 |
++ wrmsrl(MSR_IA32_RTIT_ADDR0_B + i * 2, ctx->addr_b[i]); |
2700 |
++ } |
2701 |
++} |
2702 |
++ |
2703 |
++static inline void pt_save_msr(struct pt_ctx *ctx, u32 addr_range) |
2704 |
++{ |
2705 |
++ u32 i; |
2706 |
++ |
2707 |
++ rdmsrl(MSR_IA32_RTIT_STATUS, ctx->status); |
2708 |
++ rdmsrl(MSR_IA32_RTIT_OUTPUT_BASE, ctx->output_base); |
2709 |
++ rdmsrl(MSR_IA32_RTIT_OUTPUT_MASK, ctx->output_mask); |
2710 |
++ rdmsrl(MSR_IA32_RTIT_CR3_MATCH, ctx->cr3_match); |
2711 |
++ for (i = 0; i < addr_range; i++) { |
2712 |
++ rdmsrl(MSR_IA32_RTIT_ADDR0_A + i * 2, ctx->addr_a[i]); |
2713 |
++ rdmsrl(MSR_IA32_RTIT_ADDR0_B + i * 2, ctx->addr_b[i]); |
2714 |
++ } |
2715 |
++} |
2716 |
++ |
2717 |
++static void pt_guest_enter(struct vcpu_vmx *vmx) |
2718 |
++{ |
2719 |
++ if (pt_mode == PT_MODE_SYSTEM) |
2720 |
++ return; |
2721 |
++ |
2722 |
++ /* |
2723 |
++ * GUEST_IA32_RTIT_CTL is already set in the VMCS. |
2724 |
++ * Save host state before VM entry. |
2725 |
++ */ |
2726 |
++ rdmsrl(MSR_IA32_RTIT_CTL, vmx->pt_desc.host.ctl); |
2727 |
++ if (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) { |
2728 |
++ wrmsrl(MSR_IA32_RTIT_CTL, 0); |
2729 |
++ pt_save_msr(&vmx->pt_desc.host, vmx->pt_desc.addr_range); |
2730 |
++ pt_load_msr(&vmx->pt_desc.guest, vmx->pt_desc.addr_range); |
2731 |
++ } |
2732 |
++} |
2733 |
++ |
2734 |
++static void pt_guest_exit(struct vcpu_vmx *vmx) |
2735 |
++{ |
2736 |
++ if (pt_mode == PT_MODE_SYSTEM) |
2737 |
++ return; |
2738 |
++ |
2739 |
++ if (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) { |
2740 |
++ pt_save_msr(&vmx->pt_desc.guest, vmx->pt_desc.addr_range); |
2741 |
++ pt_load_msr(&vmx->pt_desc.host, vmx->pt_desc.addr_range); |
2742 |
++ } |
2743 |
++ |
2744 |
++ /* Reload host state (IA32_RTIT_CTL will be cleared on VM exit). */ |
2745 |
++ wrmsrl(MSR_IA32_RTIT_CTL, vmx->pt_desc.host.ctl); |
2746 |
++} |
2747 |
++ |
2748 |
++void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel, |
2749 |
++ unsigned long fs_base, unsigned long gs_base) |
2750 |
++{ |
2751 |
++ if (unlikely(fs_sel != host->fs_sel)) { |
2752 |
++ if (!(fs_sel & 7)) |
2753 |
++ vmcs_write16(HOST_FS_SELECTOR, fs_sel); |
2754 |
++ else |
2755 |
++ vmcs_write16(HOST_FS_SELECTOR, 0); |
2756 |
++ host->fs_sel = fs_sel; |
2757 |
++ } |
2758 |
++ if (unlikely(gs_sel != host->gs_sel)) { |
2759 |
++ if (!(gs_sel & 7)) |
2760 |
++ vmcs_write16(HOST_GS_SELECTOR, gs_sel); |
2761 |
++ else |
2762 |
++ vmcs_write16(HOST_GS_SELECTOR, 0); |
2763 |
++ host->gs_sel = gs_sel; |
2764 |
++ } |
2765 |
++ if (unlikely(fs_base != host->fs_base)) { |
2766 |
++ vmcs_writel(HOST_FS_BASE, fs_base); |
2767 |
++ host->fs_base = fs_base; |
2768 |
++ } |
2769 |
++ if (unlikely(gs_base != host->gs_base)) { |
2770 |
++ vmcs_writel(HOST_GS_BASE, gs_base); |
2771 |
++ host->gs_base = gs_base; |
2772 |
++ } |
2773 |
++} |
2774 |
++ |
2775 |
++void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu) |
2776 |
++{ |
2777 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
2778 |
++ struct vmcs_host_state *host_state; |
2779 |
++#ifdef CONFIG_X86_64 |
2780 |
++ int cpu = raw_smp_processor_id(); |
2781 |
++#endif |
2782 |
++ unsigned long fs_base, gs_base; |
2783 |
++ u16 fs_sel, gs_sel; |
2784 |
++ int i; |
2785 |
++ |
2786 |
++ vmx->req_immediate_exit = false; |
2787 |
++ |
2788 |
++ /* |
2789 |
++ * Note that guest MSRs to be saved/restored can also be changed |
2790 |
++ * when guest state is loaded. This happens when guest transitions |
2791 |
++ * to/from long-mode by setting MSR_EFER.LMA. |
2792 |
++ */ |
2793 |
++ if (!vmx->guest_msrs_ready) { |
2794 |
++ vmx->guest_msrs_ready = true; |
2795 |
++ for (i = 0; i < vmx->save_nmsrs; ++i) |
2796 |
++ kvm_set_shared_msr(vmx->guest_msrs[i].index, |
2797 |
++ vmx->guest_msrs[i].data, |
2798 |
++ vmx->guest_msrs[i].mask); |
2799 |
++ |
2800 |
++ } |
2801 |
++ if (vmx->guest_state_loaded) |
2802 |
++ return; |
2803 |
++ |
2804 |
++ host_state = &vmx->loaded_vmcs->host_state; |
2805 |
++ |
2806 |
++ /* |
2807 |
++ * Set host fs and gs selectors. Unfortunately, 22.2.3 does not |
2808 |
++ * allow segment selectors with cpl > 0 or ti == 1. |
2809 |
++ */ |
2810 |
++ host_state->ldt_sel = kvm_read_ldt(); |
2811 |
++ |
2812 |
++#ifdef CONFIG_X86_64 |
2813 |
++ savesegment(ds, host_state->ds_sel); |
2814 |
++ savesegment(es, host_state->es_sel); |
2815 |
++ |
2816 |
++ gs_base = cpu_kernelmode_gs_base(cpu); |
2817 |
++ if (likely(is_64bit_mm(current->mm))) { |
2818 |
++ save_fsgs_for_kvm(); |
2819 |
++ fs_sel = current->thread.fsindex; |
2820 |
++ gs_sel = current->thread.gsindex; |
2821 |
++ fs_base = current->thread.fsbase; |
2822 |
++ vmx->msr_host_kernel_gs_base = current->thread.gsbase; |
2823 |
++ } else { |
2824 |
++ savesegment(fs, fs_sel); |
2825 |
++ savesegment(gs, gs_sel); |
2826 |
++ fs_base = read_msr(MSR_FS_BASE); |
2827 |
++ vmx->msr_host_kernel_gs_base = read_msr(MSR_KERNEL_GS_BASE); |
2828 |
++ } |
2829 |
++ |
2830 |
++ wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2831 |
++#else |
2832 |
++ savesegment(fs, fs_sel); |
2833 |
++ savesegment(gs, gs_sel); |
2834 |
++ fs_base = segment_base(fs_sel); |
2835 |
++ gs_base = segment_base(gs_sel); |
2836 |
++#endif |
2837 |
++ |
2838 |
++ vmx_set_host_fs_gs(host_state, fs_sel, gs_sel, fs_base, gs_base); |
2839 |
++ vmx->guest_state_loaded = true; |
2840 |
++} |
2841 |
++ |
2842 |
++static void vmx_prepare_switch_to_host(struct vcpu_vmx *vmx) |
2843 |
++{ |
2844 |
++ struct vmcs_host_state *host_state; |
2845 |
++ |
2846 |
++ if (!vmx->guest_state_loaded) |
2847 |
++ return; |
2848 |
++ |
2849 |
++ host_state = &vmx->loaded_vmcs->host_state; |
2850 |
++ |
2851 |
++ ++vmx->vcpu.stat.host_state_reload; |
2852 |
++ |
2853 |
++#ifdef CONFIG_X86_64 |
2854 |
++ rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2855 |
++#endif |
2856 |
++ if (host_state->ldt_sel || (host_state->gs_sel & 7)) { |
2857 |
++ kvm_load_ldt(host_state->ldt_sel); |
2858 |
++#ifdef CONFIG_X86_64 |
2859 |
++ load_gs_index(host_state->gs_sel); |
2860 |
++#else |
2861 |
++ loadsegment(gs, host_state->gs_sel); |
2862 |
++#endif |
2863 |
++ } |
2864 |
++ if (host_state->fs_sel & 7) |
2865 |
++ loadsegment(fs, host_state->fs_sel); |
2866 |
++#ifdef CONFIG_X86_64 |
2867 |
++ if (unlikely(host_state->ds_sel | host_state->es_sel)) { |
2868 |
++ loadsegment(ds, host_state->ds_sel); |
2869 |
++ loadsegment(es, host_state->es_sel); |
2870 |
++ } |
2871 |
++#endif |
2872 |
++ invalidate_tss_limit(); |
2873 |
++#ifdef CONFIG_X86_64 |
2874 |
++ wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); |
2875 |
++#endif |
2876 |
++ load_fixmap_gdt(raw_smp_processor_id()); |
2877 |
++ vmx->guest_state_loaded = false; |
2878 |
++ vmx->guest_msrs_ready = false; |
2879 |
++} |
2880 |
++ |
2881 |
++#ifdef CONFIG_X86_64 |
2882 |
++static u64 vmx_read_guest_kernel_gs_base(struct vcpu_vmx *vmx) |
2883 |
++{ |
2884 |
++ preempt_disable(); |
2885 |
++ if (vmx->guest_state_loaded) |
2886 |
++ rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2887 |
++ preempt_enable(); |
2888 |
++ return vmx->msr_guest_kernel_gs_base; |
2889 |
++} |
2890 |
++ |
2891 |
++static void vmx_write_guest_kernel_gs_base(struct vcpu_vmx *vmx, u64 data) |
2892 |
++{ |
2893 |
++ preempt_disable(); |
2894 |
++ if (vmx->guest_state_loaded) |
2895 |
++ wrmsrl(MSR_KERNEL_GS_BASE, data); |
2896 |
++ preempt_enable(); |
2897 |
++ vmx->msr_guest_kernel_gs_base = data; |
2898 |
++} |
2899 |
++#endif |
2900 |
++ |
2901 |
++static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu) |
2902 |
++{ |
2903 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
2904 |
++ struct pi_desc old, new; |
2905 |
++ unsigned int dest; |
2906 |
++ |
2907 |
++ /* |
2908 |
++ * In case of hot-plug or hot-unplug, we may have to undo |
2909 |
++ * vmx_vcpu_pi_put even if there is no assigned device. And we |
2910 |
++ * always keep PI.NDST up to date for simplicity: it makes the |
2911 |
++ * code easier, and CPU migration is not a fast path. |
2912 |
++ */ |
2913 |
++ if (!pi_test_sn(pi_desc) && vcpu->cpu == cpu) |
2914 |
++ return; |
2915 |
++ |
2916 |
++ /* |
2917 |
++ * If the 'nv' field is POSTED_INTR_WAKEUP_VECTOR, do not change |
2918 |
++ * PI.NDST: pi_post_block is the one expected to change PID.NDST and the |
2919 |
++ * wakeup handler expects the vCPU to be on the blocked_vcpu_list that |
2920 |
++ * matches PI.NDST. Otherwise, a vcpu may not be able to be woken up |
2921 |
++ * correctly. |
2922 |
++ */ |
2923 |
++ if (pi_desc->nv == POSTED_INTR_WAKEUP_VECTOR || vcpu->cpu == cpu) { |
2924 |
++ pi_clear_sn(pi_desc); |
2925 |
++ goto after_clear_sn; |
2926 |
++ } |
2927 |
++ |
2928 |
++ /* The full case. */ |
2929 |
++ do { |
2930 |
++ old.control = new.control = pi_desc->control; |
2931 |
++ |
2932 |
++ dest = cpu_physical_id(cpu); |
2933 |
++ |
2934 |
++ if (x2apic_enabled()) |
2935 |
++ new.ndst = dest; |
2936 |
++ else |
2937 |
++ new.ndst = (dest << 8) & 0xFF00; |
2938 |
++ |
2939 |
++ new.sn = 0; |
2940 |
++ } while (cmpxchg64(&pi_desc->control, old.control, |
2941 |
++ new.control) != old.control); |
2942 |
++ |
2943 |
++after_clear_sn: |
2944 |
++ |
2945 |
++ /* |
2946 |
++ * Clear SN before reading the bitmap. The VT-d firmware |
2947 |
++ * writes the bitmap and reads SN atomically (5.2.3 in the |
2948 |
++ * spec), so it doesn't really have a memory barrier that |
2949 |
++ * pairs with this, but we cannot do that and we need one. |
2950 |
++ */ |
2951 |
++ smp_mb__after_atomic(); |
2952 |
++ |
2953 |
++ if (!pi_is_pir_empty(pi_desc)) |
2954 |
++ pi_set_on(pi_desc); |
2955 |
++} |
2956 |
++ |
2957 |
++void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu) |
2958 |
++{ |
2959 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
2960 |
++ bool already_loaded = vmx->loaded_vmcs->cpu == cpu; |
2961 |
++ |
2962 |
++ if (!already_loaded) { |
2963 |
++ loaded_vmcs_clear(vmx->loaded_vmcs); |
2964 |
++ local_irq_disable(); |
2965 |
++ crash_disable_local_vmclear(cpu); |
2966 |
++ |
2967 |
++ /* |
2968 |
++ * Read loaded_vmcs->cpu should be before fetching |
2969 |
++ * loaded_vmcs->loaded_vmcss_on_cpu_link. |
2970 |
++ * See the comments in __loaded_vmcs_clear(). |
2971 |
++ */ |
2972 |
++ smp_rmb(); |
2973 |
++ |
2974 |
++ list_add(&vmx->loaded_vmcs->loaded_vmcss_on_cpu_link, |
2975 |
++ &per_cpu(loaded_vmcss_on_cpu, cpu)); |
2976 |
++ crash_enable_local_vmclear(cpu); |
2977 |
++ local_irq_enable(); |
2978 |
++ } |
2979 |
++ |
2980 |
++ if (per_cpu(current_vmcs, cpu) != vmx->loaded_vmcs->vmcs) { |
2981 |
++ per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs; |
2982 |
++ vmcs_load(vmx->loaded_vmcs->vmcs); |
2983 |
++ indirect_branch_prediction_barrier(); |
2984 |
++ } |
2985 |
++ |
2986 |
++ if (!already_loaded) { |
2987 |
++ void *gdt = get_current_gdt_ro(); |
2988 |
++ unsigned long sysenter_esp; |
2989 |
++ |
2990 |
++ kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
2991 |
++ |
2992 |
++ /* |
2993 |
++ * Linux uses per-cpu TSS and GDT, so set these when switching |
2994 |
++ * processors. See 22.2.4. |
2995 |
++ */ |
2996 |
++ vmcs_writel(HOST_TR_BASE, |
2997 |
++ (unsigned long)&get_cpu_entry_area(cpu)->tss.x86_tss); |
2998 |
++ vmcs_writel(HOST_GDTR_BASE, (unsigned long)gdt); /* 22.2.4 */ |
2999 |
++ |
3000 |
++ rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp); |
3001 |
++ vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */ |
3002 |
++ |
3003 |
++ vmx->loaded_vmcs->cpu = cpu; |
3004 |
++ } |
3005 |
++ |
3006 |
++ /* Setup TSC multiplier */ |
3007 |
++ if (kvm_has_tsc_control && |
3008 |
++ vmx->current_tsc_ratio != vcpu->arch.tsc_scaling_ratio) |
3009 |
++ decache_tsc_multiplier(vmx); |
3010 |
++} |
3011 |
++ |
3012 |
++/* |
3013 |
++ * Switches to specified vcpu, until a matching vcpu_put(), but assumes |
3014 |
++ * vcpu mutex is already taken. |
3015 |
++ */ |
3016 |
++void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
3017 |
++{ |
3018 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3019 |
++ |
3020 |
++ vmx_vcpu_load_vmcs(vcpu, cpu); |
3021 |
++ |
3022 |
++ vmx_vcpu_pi_load(vcpu, cpu); |
3023 |
++ |
3024 |
++ vmx->host_pkru = read_pkru(); |
3025 |
++ vmx->host_debugctlmsr = get_debugctlmsr(); |
3026 |
++} |
3027 |
++ |
3028 |
++static void vmx_vcpu_pi_put(struct kvm_vcpu *vcpu) |
3029 |
++{ |
3030 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
3031 |
++ |
3032 |
++ if (!kvm_arch_has_assigned_device(vcpu->kvm) || |
3033 |
++ !irq_remapping_cap(IRQ_POSTING_CAP) || |
3034 |
++ !kvm_vcpu_apicv_active(vcpu)) |
3035 |
++ return; |
3036 |
++ |
3037 |
++ /* Set SN when the vCPU is preempted */ |
3038 |
++ if (vcpu->preempted) |
3039 |
++ pi_set_sn(pi_desc); |
3040 |
++} |
3041 |
++ |
3042 |
++static void vmx_vcpu_put(struct kvm_vcpu *vcpu) |
3043 |
++{ |
3044 |
++ vmx_vcpu_pi_put(vcpu); |
3045 |
++ |
3046 |
++ vmx_prepare_switch_to_host(to_vmx(vcpu)); |
3047 |
++} |
3048 |
++ |
3049 |
++static bool emulation_required(struct kvm_vcpu *vcpu) |
3050 |
++{ |
3051 |
++ return emulate_invalid_guest_state && !guest_state_valid(vcpu); |
3052 |
++} |
3053 |
++ |
3054 |
++static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu); |
3055 |
++ |
3056 |
++unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) |
3057 |
++{ |
3058 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3059 |
++ unsigned long rflags, save_rflags; |
3060 |
++ |
3061 |
++ if (!kvm_register_is_available(vcpu, VCPU_EXREG_RFLAGS)) { |
3062 |
++ kvm_register_mark_available(vcpu, VCPU_EXREG_RFLAGS); |
3063 |
++ rflags = vmcs_readl(GUEST_RFLAGS); |
3064 |
++ if (vmx->rmode.vm86_active) { |
3065 |
++ rflags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
3066 |
++ save_rflags = vmx->rmode.save_rflags; |
3067 |
++ rflags |= save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; |
3068 |
++ } |
3069 |
++ vmx->rflags = rflags; |
3070 |
++ } |
3071 |
++ return vmx->rflags; |
3072 |
++} |
3073 |
++ |
3074 |
++void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) |
3075 |
++{ |
3076 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3077 |
++ unsigned long old_rflags; |
3078 |
++ |
3079 |
++ if (enable_unrestricted_guest) { |
3080 |
++ kvm_register_mark_available(vcpu, VCPU_EXREG_RFLAGS); |
3081 |
++ vmx->rflags = rflags; |
3082 |
++ vmcs_writel(GUEST_RFLAGS, rflags); |
3083 |
++ return; |
3084 |
++ } |
3085 |
++ |
3086 |
++ old_rflags = vmx_get_rflags(vcpu); |
3087 |
++ vmx->rflags = rflags; |
3088 |
++ if (vmx->rmode.vm86_active) { |
3089 |
++ vmx->rmode.save_rflags = rflags; |
3090 |
++ rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
3091 |
++ } |
3092 |
++ vmcs_writel(GUEST_RFLAGS, rflags); |
3093 |
++ |
3094 |
++ if ((old_rflags ^ vmx->rflags) & X86_EFLAGS_VM) |
3095 |
++ vmx->emulation_required = emulation_required(vcpu); |
3096 |
++} |
3097 |
++ |
3098 |
++u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu) |
3099 |
++{ |
3100 |
++ u32 interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); |
3101 |
++ int ret = 0; |
3102 |
++ |
3103 |
++ if (interruptibility & GUEST_INTR_STATE_STI) |
3104 |
++ ret |= KVM_X86_SHADOW_INT_STI; |
3105 |
++ if (interruptibility & GUEST_INTR_STATE_MOV_SS) |
3106 |
++ ret |= KVM_X86_SHADOW_INT_MOV_SS; |
3107 |
++ |
3108 |
++ return ret; |
3109 |
++} |
3110 |
++ |
3111 |
++void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) |
3112 |
++{ |
3113 |
++ u32 interruptibility_old = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); |
3114 |
++ u32 interruptibility = interruptibility_old; |
3115 |
++ |
3116 |
++ interruptibility &= ~(GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS); |
3117 |
++ |
3118 |
++ if (mask & KVM_X86_SHADOW_INT_MOV_SS) |
3119 |
++ interruptibility |= GUEST_INTR_STATE_MOV_SS; |
3120 |
++ else if (mask & KVM_X86_SHADOW_INT_STI) |
3121 |
++ interruptibility |= GUEST_INTR_STATE_STI; |
3122 |
++ |
3123 |
++ if ((interruptibility != interruptibility_old)) |
3124 |
++ vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, interruptibility); |
3125 |
++} |
3126 |
++ |
3127 |
++static int vmx_rtit_ctl_check(struct kvm_vcpu *vcpu, u64 data) |
3128 |
++{ |
3129 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3130 |
++ unsigned long value; |
3131 |
++ |
3132 |
++ /* |
3133 |
++ * Any MSR write that attempts to change bits marked reserved will |
3134 |
++ * case a #GP fault. |
3135 |
++ */ |
3136 |
++ if (data & vmx->pt_desc.ctl_bitmask) |
3137 |
++ return 1; |
3138 |
++ |
3139 |
++ /* |
3140 |
++ * Any attempt to modify IA32_RTIT_CTL while TraceEn is set will |
3141 |
++ * result in a #GP unless the same write also clears TraceEn. |
3142 |
++ */ |
3143 |
++ if ((vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) && |
3144 |
++ ((vmx->pt_desc.guest.ctl ^ data) & ~RTIT_CTL_TRACEEN)) |
3145 |
++ return 1; |
3146 |
++ |
3147 |
++ /* |
3148 |
++ * WRMSR to IA32_RTIT_CTL that sets TraceEn but clears this bit |
3149 |
++ * and FabricEn would cause #GP, if |
3150 |
++ * CPUID.(EAX=14H, ECX=0):ECX.SNGLRGNOUT[bit 2] = 0 |
3151 |
++ */ |
3152 |
++ if ((data & RTIT_CTL_TRACEEN) && !(data & RTIT_CTL_TOPA) && |
3153 |
++ !(data & RTIT_CTL_FABRIC_EN) && |
3154 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3155 |
++ PT_CAP_single_range_output)) |
3156 |
++ return 1; |
3157 |
++ |
3158 |
++ /* |
3159 |
++ * MTCFreq, CycThresh and PSBFreq encodings check, any MSR write that |
3160 |
++ * utilize encodings marked reserved will casue a #GP fault. |
3161 |
++ */ |
3162 |
++ value = intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc_periods); |
3163 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc) && |
3164 |
++ !test_bit((data & RTIT_CTL_MTC_RANGE) >> |
3165 |
++ RTIT_CTL_MTC_RANGE_OFFSET, &value)) |
3166 |
++ return 1; |
3167 |
++ value = intel_pt_validate_cap(vmx->pt_desc.caps, |
3168 |
++ PT_CAP_cycle_thresholds); |
3169 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc) && |
3170 |
++ !test_bit((data & RTIT_CTL_CYC_THRESH) >> |
3171 |
++ RTIT_CTL_CYC_THRESH_OFFSET, &value)) |
3172 |
++ return 1; |
3173 |
++ value = intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_periods); |
3174 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc) && |
3175 |
++ !test_bit((data & RTIT_CTL_PSB_FREQ) >> |
3176 |
++ RTIT_CTL_PSB_FREQ_OFFSET, &value)) |
3177 |
++ return 1; |
3178 |
++ |
3179 |
++ /* |
3180 |
++ * If ADDRx_CFG is reserved or the encodings is >2 will |
3181 |
++ * cause a #GP fault. |
3182 |
++ */ |
3183 |
++ value = (data & RTIT_CTL_ADDR0) >> RTIT_CTL_ADDR0_OFFSET; |
3184 |
++ if ((value && (vmx->pt_desc.addr_range < 1)) || (value > 2)) |
3185 |
++ return 1; |
3186 |
++ value = (data & RTIT_CTL_ADDR1) >> RTIT_CTL_ADDR1_OFFSET; |
3187 |
++ if ((value && (vmx->pt_desc.addr_range < 2)) || (value > 2)) |
3188 |
++ return 1; |
3189 |
++ value = (data & RTIT_CTL_ADDR2) >> RTIT_CTL_ADDR2_OFFSET; |
3190 |
++ if ((value && (vmx->pt_desc.addr_range < 3)) || (value > 2)) |
3191 |
++ return 1; |
3192 |
++ value = (data & RTIT_CTL_ADDR3) >> RTIT_CTL_ADDR3_OFFSET; |
3193 |
++ if ((value && (vmx->pt_desc.addr_range < 4)) || (value > 2)) |
3194 |
++ return 1; |
3195 |
++ |
3196 |
++ return 0; |
3197 |
++} |
3198 |
++ |
3199 |
++static int skip_emulated_instruction(struct kvm_vcpu *vcpu) |
3200 |
++{ |
3201 |
++ unsigned long rip; |
3202 |
++ |
3203 |
++ /* |
3204 |
++ * Using VMCS.VM_EXIT_INSTRUCTION_LEN on EPT misconfig depends on |
3205 |
++ * undefined behavior: Intel's SDM doesn't mandate the VMCS field be |
3206 |
++ * set when EPT misconfig occurs. In practice, real hardware updates |
3207 |
++ * VM_EXIT_INSTRUCTION_LEN on EPT misconfig, but other hypervisors |
3208 |
++ * (namely Hyper-V) don't set it due to it being undefined behavior, |
3209 |
++ * i.e. we end up advancing IP with some random value. |
3210 |
++ */ |
3211 |
++ if (!static_cpu_has(X86_FEATURE_HYPERVISOR) || |
3212 |
++ to_vmx(vcpu)->exit_reason != EXIT_REASON_EPT_MISCONFIG) { |
3213 |
++ rip = kvm_rip_read(vcpu); |
3214 |
++ rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
3215 |
++ kvm_rip_write(vcpu, rip); |
3216 |
++ } else { |
3217 |
++ if (!kvm_emulate_instruction(vcpu, EMULTYPE_SKIP)) |
3218 |
++ return 0; |
3219 |
++ } |
3220 |
++ |
3221 |
++ /* skipping an emulated instruction also counts */ |
3222 |
++ vmx_set_interrupt_shadow(vcpu, 0); |
3223 |
++ |
3224 |
++ return 1; |
3225 |
++} |
3226 |
++ |
3227 |
++static void vmx_clear_hlt(struct kvm_vcpu *vcpu) |
3228 |
++{ |
3229 |
++ /* |
3230 |
++ * Ensure that we clear the HLT state in the VMCS. We don't need to |
3231 |
++ * explicitly skip the instruction because if the HLT state is set, |
3232 |
++ * then the instruction is already executing and RIP has already been |
3233 |
++ * advanced. |
3234 |
++ */ |
3235 |
++ if (kvm_hlt_in_guest(vcpu->kvm) && |
3236 |
++ vmcs_read32(GUEST_ACTIVITY_STATE) == GUEST_ACTIVITY_HLT) |
3237 |
++ vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE); |
3238 |
++} |
3239 |
++ |
3240 |
++static void vmx_queue_exception(struct kvm_vcpu *vcpu) |
3241 |
++{ |
3242 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3243 |
++ unsigned nr = vcpu->arch.exception.nr; |
3244 |
++ bool has_error_code = vcpu->arch.exception.has_error_code; |
3245 |
++ u32 error_code = vcpu->arch.exception.error_code; |
3246 |
++ u32 intr_info = nr | INTR_INFO_VALID_MASK; |
3247 |
++ |
3248 |
++ kvm_deliver_exception_payload(vcpu); |
3249 |
++ |
3250 |
++ if (has_error_code) { |
3251 |
++ vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code); |
3252 |
++ intr_info |= INTR_INFO_DELIVER_CODE_MASK; |
3253 |
++ } |
3254 |
++ |
3255 |
++ if (vmx->rmode.vm86_active) { |
3256 |
++ int inc_eip = 0; |
3257 |
++ if (kvm_exception_is_soft(nr)) |
3258 |
++ inc_eip = vcpu->arch.event_exit_inst_len; |
3259 |
++ kvm_inject_realmode_interrupt(vcpu, nr, inc_eip); |
3260 |
++ return; |
3261 |
++ } |
3262 |
++ |
3263 |
++ WARN_ON_ONCE(vmx->emulation_required); |
3264 |
++ |
3265 |
++ if (kvm_exception_is_soft(nr)) { |
3266 |
++ vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, |
3267 |
++ vmx->vcpu.arch.event_exit_inst_len); |
3268 |
++ intr_info |= INTR_TYPE_SOFT_EXCEPTION; |
3269 |
++ } else |
3270 |
++ intr_info |= INTR_TYPE_HARD_EXCEPTION; |
3271 |
++ |
3272 |
++ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info); |
3273 |
++ |
3274 |
++ vmx_clear_hlt(vcpu); |
3275 |
++} |
3276 |
++ |
3277 |
++static bool vmx_rdtscp_supported(void) |
3278 |
++{ |
3279 |
++ return cpu_has_vmx_rdtscp(); |
3280 |
++} |
3281 |
++ |
3282 |
++static bool vmx_invpcid_supported(void) |
3283 |
++{ |
3284 |
++ return cpu_has_vmx_invpcid(); |
3285 |
++} |
3286 |
++ |
3287 |
++/* |
3288 |
++ * Swap MSR entry in host/guest MSR entry array. |
3289 |
++ */ |
3290 |
++static void move_msr_up(struct vcpu_vmx *vmx, int from, int to) |
3291 |
++{ |
3292 |
++ struct shared_msr_entry tmp; |
3293 |
++ |
3294 |
++ tmp = vmx->guest_msrs[to]; |
3295 |
++ vmx->guest_msrs[to] = vmx->guest_msrs[from]; |
3296 |
++ vmx->guest_msrs[from] = tmp; |
3297 |
++} |
3298 |
++ |
3299 |
++/* |
3300 |
++ * Set up the vmcs to automatically save and restore system |
3301 |
++ * msrs. Don't touch the 64-bit msrs if the guest is in legacy |
3302 |
++ * mode, as fiddling with msrs is very expensive. |
3303 |
++ */ |
3304 |
++static void setup_msrs(struct vcpu_vmx *vmx) |
3305 |
++{ |
3306 |
++ int save_nmsrs, index; |
3307 |
++ |
3308 |
++ save_nmsrs = 0; |
3309 |
++#ifdef CONFIG_X86_64 |
3310 |
++ /* |
3311 |
++ * The SYSCALL MSRs are only needed on long mode guests, and only |
3312 |
++ * when EFER.SCE is set. |
3313 |
++ */ |
3314 |
++ if (is_long_mode(&vmx->vcpu) && (vmx->vcpu.arch.efer & EFER_SCE)) { |
3315 |
++ index = __find_msr_index(vmx, MSR_STAR); |
3316 |
++ if (index >= 0) |
3317 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3318 |
++ index = __find_msr_index(vmx, MSR_LSTAR); |
3319 |
++ if (index >= 0) |
3320 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3321 |
++ index = __find_msr_index(vmx, MSR_SYSCALL_MASK); |
3322 |
++ if (index >= 0) |
3323 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3324 |
++ } |
3325 |
++#endif |
3326 |
++ index = __find_msr_index(vmx, MSR_EFER); |
3327 |
++ if (index >= 0 && update_transition_efer(vmx, index)) |
3328 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3329 |
++ index = __find_msr_index(vmx, MSR_TSC_AUX); |
3330 |
++ if (index >= 0 && guest_cpuid_has(&vmx->vcpu, X86_FEATURE_RDTSCP)) |
3331 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3332 |
++ index = __find_msr_index(vmx, MSR_IA32_TSX_CTRL); |
3333 |
++ if (index >= 0) |
3334 |
++ move_msr_up(vmx, index, save_nmsrs++); |
3335 |
++ |
3336 |
++ vmx->save_nmsrs = save_nmsrs; |
3337 |
++ vmx->guest_msrs_ready = false; |
3338 |
++ |
3339 |
++ if (cpu_has_vmx_msr_bitmap()) |
3340 |
++ vmx_update_msr_bitmap(&vmx->vcpu); |
3341 |
++} |
3342 |
++ |
3343 |
++static u64 vmx_read_l1_tsc_offset(struct kvm_vcpu *vcpu) |
3344 |
++{ |
3345 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
3346 |
++ |
3347 |
++ if (is_guest_mode(vcpu) && |
3348 |
++ (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)) |
3349 |
++ return vcpu->arch.tsc_offset - vmcs12->tsc_offset; |
3350 |
++ |
3351 |
++ return vcpu->arch.tsc_offset; |
3352 |
++} |
3353 |
++ |
3354 |
++static u64 vmx_write_l1_tsc_offset(struct kvm_vcpu *vcpu, u64 offset) |
3355 |
++{ |
3356 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
3357 |
++ u64 g_tsc_offset = 0; |
3358 |
++ |
3359 |
++ /* |
3360 |
++ * We're here if L1 chose not to trap WRMSR to TSC. According |
3361 |
++ * to the spec, this should set L1's TSC; The offset that L1 |
3362 |
++ * set for L2 remains unchanged, and still needs to be added |
3363 |
++ * to the newly set TSC to get L2's TSC. |
3364 |
++ */ |
3365 |
++ if (is_guest_mode(vcpu) && |
3366 |
++ (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)) |
3367 |
++ g_tsc_offset = vmcs12->tsc_offset; |
3368 |
++ |
3369 |
++ trace_kvm_write_tsc_offset(vcpu->vcpu_id, |
3370 |
++ vcpu->arch.tsc_offset - g_tsc_offset, |
3371 |
++ offset); |
3372 |
++ vmcs_write64(TSC_OFFSET, offset + g_tsc_offset); |
3373 |
++ return offset + g_tsc_offset; |
3374 |
++} |
3375 |
++ |
3376 |
++/* |
3377 |
++ * nested_vmx_allowed() checks whether a guest should be allowed to use VMX |
3378 |
++ * instructions and MSRs (i.e., nested VMX). Nested VMX is disabled for |
3379 |
++ * all guests if the "nested" module option is off, and can also be disabled |
3380 |
++ * for a single guest by disabling its VMX cpuid bit. |
3381 |
++ */ |
3382 |
++bool nested_vmx_allowed(struct kvm_vcpu *vcpu) |
3383 |
++{ |
3384 |
++ return nested && guest_cpuid_has(vcpu, X86_FEATURE_VMX); |
3385 |
++} |
3386 |
++ |
3387 |
++static inline bool vmx_feature_control_msr_valid(struct kvm_vcpu *vcpu, |
3388 |
++ uint64_t val) |
3389 |
++{ |
3390 |
++ uint64_t valid_bits = to_vmx(vcpu)->msr_ia32_feature_control_valid_bits; |
3391 |
++ |
3392 |
++ return !(val & ~valid_bits); |
3393 |
++} |
3394 |
++ |
3395 |
++static int vmx_get_msr_feature(struct kvm_msr_entry *msr) |
3396 |
++{ |
3397 |
++ switch (msr->index) { |
3398 |
++ case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
3399 |
++ if (!nested) |
3400 |
++ return 1; |
3401 |
++ return vmx_get_vmx_msr(&vmcs_config.nested, msr->index, &msr->data); |
3402 |
++ default: |
3403 |
++ return 1; |
3404 |
++ } |
3405 |
++} |
3406 |
++ |
3407 |
++/* |
3408 |
++ * Reads an msr value (of 'msr_index') into 'pdata'. |
3409 |
++ * Returns 0 on success, non-0 otherwise. |
3410 |
++ * Assumes vcpu_load() was already called. |
3411 |
++ */ |
3412 |
++static int vmx_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) |
3413 |
++{ |
3414 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3415 |
++ struct shared_msr_entry *msr; |
3416 |
++ u32 index; |
3417 |
++ |
3418 |
++ switch (msr_info->index) { |
3419 |
++#ifdef CONFIG_X86_64 |
3420 |
++ case MSR_FS_BASE: |
3421 |
++ msr_info->data = vmcs_readl(GUEST_FS_BASE); |
3422 |
++ break; |
3423 |
++ case MSR_GS_BASE: |
3424 |
++ msr_info->data = vmcs_readl(GUEST_GS_BASE); |
3425 |
++ break; |
3426 |
++ case MSR_KERNEL_GS_BASE: |
3427 |
++ msr_info->data = vmx_read_guest_kernel_gs_base(vmx); |
3428 |
++ break; |
3429 |
++#endif |
3430 |
++ case MSR_EFER: |
3431 |
++ return kvm_get_msr_common(vcpu, msr_info); |
3432 |
++ case MSR_IA32_TSX_CTRL: |
3433 |
++ if (!msr_info->host_initiated && |
3434 |
++ !(vcpu->arch.arch_capabilities & ARCH_CAP_TSX_CTRL_MSR)) |
3435 |
++ return 1; |
3436 |
++ goto find_shared_msr; |
3437 |
++ case MSR_IA32_UMWAIT_CONTROL: |
3438 |
++ if (!msr_info->host_initiated && !vmx_has_waitpkg(vmx)) |
3439 |
++ return 1; |
3440 |
++ |
3441 |
++ msr_info->data = vmx->msr_ia32_umwait_control; |
3442 |
++ break; |
3443 |
++ case MSR_IA32_SPEC_CTRL: |
3444 |
++ if (!msr_info->host_initiated && |
3445 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
3446 |
++ return 1; |
3447 |
++ |
3448 |
++ msr_info->data = to_vmx(vcpu)->spec_ctrl; |
3449 |
++ break; |
3450 |
++ case MSR_IA32_SYSENTER_CS: |
3451 |
++ msr_info->data = vmcs_read32(GUEST_SYSENTER_CS); |
3452 |
++ break; |
3453 |
++ case MSR_IA32_SYSENTER_EIP: |
3454 |
++ msr_info->data = vmcs_readl(GUEST_SYSENTER_EIP); |
3455 |
++ break; |
3456 |
++ case MSR_IA32_SYSENTER_ESP: |
3457 |
++ msr_info->data = vmcs_readl(GUEST_SYSENTER_ESP); |
3458 |
++ break; |
3459 |
++ case MSR_IA32_BNDCFGS: |
3460 |
++ if (!kvm_mpx_supported() || |
3461 |
++ (!msr_info->host_initiated && |
3462 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_MPX))) |
3463 |
++ return 1; |
3464 |
++ msr_info->data = vmcs_read64(GUEST_BNDCFGS); |
3465 |
++ break; |
3466 |
++ case MSR_IA32_MCG_EXT_CTL: |
3467 |
++ if (!msr_info->host_initiated && |
3468 |
++ !(vmx->msr_ia32_feature_control & |
3469 |
++ FEATURE_CONTROL_LMCE)) |
3470 |
++ return 1; |
3471 |
++ msr_info->data = vcpu->arch.mcg_ext_ctl; |
3472 |
++ break; |
3473 |
++ case MSR_IA32_FEATURE_CONTROL: |
3474 |
++ msr_info->data = vmx->msr_ia32_feature_control; |
3475 |
++ break; |
3476 |
++ case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
3477 |
++ if (!nested_vmx_allowed(vcpu)) |
3478 |
++ return 1; |
3479 |
++ return vmx_get_vmx_msr(&vmx->nested.msrs, msr_info->index, |
3480 |
++ &msr_info->data); |
3481 |
++ case MSR_IA32_RTIT_CTL: |
3482 |
++ if (pt_mode != PT_MODE_HOST_GUEST) |
3483 |
++ return 1; |
3484 |
++ msr_info->data = vmx->pt_desc.guest.ctl; |
3485 |
++ break; |
3486 |
++ case MSR_IA32_RTIT_STATUS: |
3487 |
++ if (pt_mode != PT_MODE_HOST_GUEST) |
3488 |
++ return 1; |
3489 |
++ msr_info->data = vmx->pt_desc.guest.status; |
3490 |
++ break; |
3491 |
++ case MSR_IA32_RTIT_CR3_MATCH: |
3492 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3493 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3494 |
++ PT_CAP_cr3_filtering)) |
3495 |
++ return 1; |
3496 |
++ msr_info->data = vmx->pt_desc.guest.cr3_match; |
3497 |
++ break; |
3498 |
++ case MSR_IA32_RTIT_OUTPUT_BASE: |
3499 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3500 |
++ (!intel_pt_validate_cap(vmx->pt_desc.caps, |
3501 |
++ PT_CAP_topa_output) && |
3502 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3503 |
++ PT_CAP_single_range_output))) |
3504 |
++ return 1; |
3505 |
++ msr_info->data = vmx->pt_desc.guest.output_base; |
3506 |
++ break; |
3507 |
++ case MSR_IA32_RTIT_OUTPUT_MASK: |
3508 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3509 |
++ (!intel_pt_validate_cap(vmx->pt_desc.caps, |
3510 |
++ PT_CAP_topa_output) && |
3511 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3512 |
++ PT_CAP_single_range_output))) |
3513 |
++ return 1; |
3514 |
++ msr_info->data = vmx->pt_desc.guest.output_mask; |
3515 |
++ break; |
3516 |
++ case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B: |
3517 |
++ index = msr_info->index - MSR_IA32_RTIT_ADDR0_A; |
3518 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3519 |
++ (index >= 2 * intel_pt_validate_cap(vmx->pt_desc.caps, |
3520 |
++ PT_CAP_num_address_ranges))) |
3521 |
++ return 1; |
3522 |
++ if (is_noncanonical_address(data, vcpu)) |
3523 |
++ return 1; |
3524 |
++ if (index % 2) |
3525 |
++ msr_info->data = vmx->pt_desc.guest.addr_b[index / 2]; |
3526 |
++ else |
3527 |
++ msr_info->data = vmx->pt_desc.guest.addr_a[index / 2]; |
3528 |
++ break; |
3529 |
++ case MSR_TSC_AUX: |
3530 |
++ if (!msr_info->host_initiated && |
3531 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP)) |
3532 |
++ return 1; |
3533 |
++ goto find_shared_msr; |
3534 |
++ default: |
3535 |
++ find_shared_msr: |
3536 |
++ msr = find_msr_entry(vmx, msr_info->index); |
3537 |
++ if (msr) { |
3538 |
++ msr_info->data = msr->data; |
3539 |
++ break; |
3540 |
++ } |
3541 |
++ return kvm_get_msr_common(vcpu, msr_info); |
3542 |
++ } |
3543 |
++ |
3544 |
++ return 0; |
3545 |
++} |
3546 |
++ |
3547 |
++/* |
3548 |
++ * Writes msr value into the appropriate "register". |
3549 |
++ * Returns 0 on success, non-0 otherwise. |
3550 |
++ * Assumes vcpu_load() was already called. |
3551 |
++ */ |
3552 |
++static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) |
3553 |
++{ |
3554 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
3555 |
++ struct shared_msr_entry *msr; |
3556 |
++ int ret = 0; |
3557 |
++ u32 msr_index = msr_info->index; |
3558 |
++ u64 data = msr_info->data; |
3559 |
++ u32 index; |
3560 |
++ |
3561 |
++ switch (msr_index) { |
3562 |
++ case MSR_EFER: |
3563 |
++ ret = kvm_set_msr_common(vcpu, msr_info); |
3564 |
++ break; |
3565 |
++#ifdef CONFIG_X86_64 |
3566 |
++ case MSR_FS_BASE: |
3567 |
++ vmx_segment_cache_clear(vmx); |
3568 |
++ vmcs_writel(GUEST_FS_BASE, data); |
3569 |
++ break; |
3570 |
++ case MSR_GS_BASE: |
3571 |
++ vmx_segment_cache_clear(vmx); |
3572 |
++ vmcs_writel(GUEST_GS_BASE, data); |
3573 |
++ break; |
3574 |
++ case MSR_KERNEL_GS_BASE: |
3575 |
++ vmx_write_guest_kernel_gs_base(vmx, data); |
3576 |
++ break; |
3577 |
++#endif |
3578 |
++ case MSR_IA32_SYSENTER_CS: |
3579 |
++ if (is_guest_mode(vcpu)) |
3580 |
++ get_vmcs12(vcpu)->guest_sysenter_cs = data; |
3581 |
++ vmcs_write32(GUEST_SYSENTER_CS, data); |
3582 |
++ break; |
3583 |
++ case MSR_IA32_SYSENTER_EIP: |
3584 |
++ if (is_guest_mode(vcpu)) |
3585 |
++ get_vmcs12(vcpu)->guest_sysenter_eip = data; |
3586 |
++ vmcs_writel(GUEST_SYSENTER_EIP, data); |
3587 |
++ break; |
3588 |
++ case MSR_IA32_SYSENTER_ESP: |
3589 |
++ if (is_guest_mode(vcpu)) |
3590 |
++ get_vmcs12(vcpu)->guest_sysenter_esp = data; |
3591 |
++ vmcs_writel(GUEST_SYSENTER_ESP, data); |
3592 |
++ break; |
3593 |
++ case MSR_IA32_DEBUGCTLMSR: |
3594 |
++ if (is_guest_mode(vcpu) && get_vmcs12(vcpu)->vm_exit_controls & |
3595 |
++ VM_EXIT_SAVE_DEBUG_CONTROLS) |
3596 |
++ get_vmcs12(vcpu)->guest_ia32_debugctl = data; |
3597 |
++ |
3598 |
++ ret = kvm_set_msr_common(vcpu, msr_info); |
3599 |
++ break; |
3600 |
++ |
3601 |
++ case MSR_IA32_BNDCFGS: |
3602 |
++ if (!kvm_mpx_supported() || |
3603 |
++ (!msr_info->host_initiated && |
3604 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_MPX))) |
3605 |
++ return 1; |
3606 |
++ if (is_noncanonical_address(data & PAGE_MASK, vcpu) || |
3607 |
++ (data & MSR_IA32_BNDCFGS_RSVD)) |
3608 |
++ return 1; |
3609 |
++ vmcs_write64(GUEST_BNDCFGS, data); |
3610 |
++ break; |
3611 |
++ case MSR_IA32_UMWAIT_CONTROL: |
3612 |
++ if (!msr_info->host_initiated && !vmx_has_waitpkg(vmx)) |
3613 |
++ return 1; |
3614 |
++ |
3615 |
++ /* The reserved bit 1 and non-32 bit [63:32] should be zero */ |
3616 |
++ if (data & (BIT_ULL(1) | GENMASK_ULL(63, 32))) |
3617 |
++ return 1; |
3618 |
++ |
3619 |
++ vmx->msr_ia32_umwait_control = data; |
3620 |
++ break; |
3621 |
++ case MSR_IA32_SPEC_CTRL: |
3622 |
++ if (!msr_info->host_initiated && |
3623 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
3624 |
++ return 1; |
3625 |
++ |
3626 |
++ /* The STIBP bit doesn't fault even if it's not advertised */ |
3627 |
++ if (data & ~(SPEC_CTRL_IBRS | SPEC_CTRL_STIBP | SPEC_CTRL_SSBD)) |
3628 |
++ return 1; |
3629 |
++ |
3630 |
++ vmx->spec_ctrl = data; |
3631 |
++ |
3632 |
++ if (!data) |
3633 |
++ break; |
3634 |
++ |
3635 |
++ /* |
3636 |
++ * For non-nested: |
3637 |
++ * When it's written (to non-zero) for the first time, pass |
3638 |
++ * it through. |
3639 |
++ * |
3640 |
++ * For nested: |
3641 |
++ * The handling of the MSR bitmap for L2 guests is done in |
3642 |
++ * nested_vmx_prepare_msr_bitmap. We should not touch the |
3643 |
++ * vmcs02.msr_bitmap here since it gets completely overwritten |
3644 |
++ * in the merging. We update the vmcs01 here for L1 as well |
3645 |
++ * since it will end up touching the MSR anyway now. |
3646 |
++ */ |
3647 |
++ vmx_disable_intercept_for_msr(vmx->vmcs01.msr_bitmap, |
3648 |
++ MSR_IA32_SPEC_CTRL, |
3649 |
++ MSR_TYPE_RW); |
3650 |
++ break; |
3651 |
++ case MSR_IA32_TSX_CTRL: |
3652 |
++ if (!msr_info->host_initiated && |
3653 |
++ !(vcpu->arch.arch_capabilities & ARCH_CAP_TSX_CTRL_MSR)) |
3654 |
++ return 1; |
3655 |
++ if (data & ~(TSX_CTRL_RTM_DISABLE | TSX_CTRL_CPUID_CLEAR)) |
3656 |
++ return 1; |
3657 |
++ goto find_shared_msr; |
3658 |
++ case MSR_IA32_PRED_CMD: |
3659 |
++ if (!msr_info->host_initiated && |
3660 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
3661 |
++ return 1; |
3662 |
++ |
3663 |
++ if (data & ~PRED_CMD_IBPB) |
3664 |
++ return 1; |
3665 |
++ |
3666 |
++ if (!data) |
3667 |
++ break; |
3668 |
++ |
3669 |
++ wrmsrl(MSR_IA32_PRED_CMD, PRED_CMD_IBPB); |
3670 |
++ |
3671 |
++ /* |
3672 |
++ * For non-nested: |
3673 |
++ * When it's written (to non-zero) for the first time, pass |
3674 |
++ * it through. |
3675 |
++ * |
3676 |
++ * For nested: |
3677 |
++ * The handling of the MSR bitmap for L2 guests is done in |
3678 |
++ * nested_vmx_prepare_msr_bitmap. We should not touch the |
3679 |
++ * vmcs02.msr_bitmap here since it gets completely overwritten |
3680 |
++ * in the merging. |
3681 |
++ */ |
3682 |
++ vmx_disable_intercept_for_msr(vmx->vmcs01.msr_bitmap, MSR_IA32_PRED_CMD, |
3683 |
++ MSR_TYPE_W); |
3684 |
++ break; |
3685 |
++ case MSR_IA32_CR_PAT: |
3686 |
++ if (!kvm_pat_valid(data)) |
3687 |
++ return 1; |
3688 |
++ |
3689 |
++ if (is_guest_mode(vcpu) && |
3690 |
++ get_vmcs12(vcpu)->vm_exit_controls & VM_EXIT_SAVE_IA32_PAT) |
3691 |
++ get_vmcs12(vcpu)->guest_ia32_pat = data; |
3692 |
++ |
3693 |
++ if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { |
3694 |
++ vmcs_write64(GUEST_IA32_PAT, data); |
3695 |
++ vcpu->arch.pat = data; |
3696 |
++ break; |
3697 |
++ } |
3698 |
++ ret = kvm_set_msr_common(vcpu, msr_info); |
3699 |
++ break; |
3700 |
++ case MSR_IA32_TSC_ADJUST: |
3701 |
++ ret = kvm_set_msr_common(vcpu, msr_info); |
3702 |
++ break; |
3703 |
++ case MSR_IA32_MCG_EXT_CTL: |
3704 |
++ if ((!msr_info->host_initiated && |
3705 |
++ !(to_vmx(vcpu)->msr_ia32_feature_control & |
3706 |
++ FEATURE_CONTROL_LMCE)) || |
3707 |
++ (data & ~MCG_EXT_CTL_LMCE_EN)) |
3708 |
++ return 1; |
3709 |
++ vcpu->arch.mcg_ext_ctl = data; |
3710 |
++ break; |
3711 |
++ case MSR_IA32_FEATURE_CONTROL: |
3712 |
++ if (!vmx_feature_control_msr_valid(vcpu, data) || |
3713 |
++ (to_vmx(vcpu)->msr_ia32_feature_control & |
3714 |
++ FEATURE_CONTROL_LOCKED && !msr_info->host_initiated)) |
3715 |
++ return 1; |
3716 |
++ vmx->msr_ia32_feature_control = data; |
3717 |
++ if (msr_info->host_initiated && data == 0) |
3718 |
++ vmx_leave_nested(vcpu); |
3719 |
++ break; |
3720 |
++ case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
3721 |
++ if (!msr_info->host_initiated) |
3722 |
++ return 1; /* they are read-only */ |
3723 |
++ if (!nested_vmx_allowed(vcpu)) |
3724 |
++ return 1; |
3725 |
++ return vmx_set_vmx_msr(vcpu, msr_index, data); |
3726 |
++ case MSR_IA32_RTIT_CTL: |
3727 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3728 |
++ vmx_rtit_ctl_check(vcpu, data) || |
3729 |
++ vmx->nested.vmxon) |
3730 |
++ return 1; |
3731 |
++ vmcs_write64(GUEST_IA32_RTIT_CTL, data); |
3732 |
++ vmx->pt_desc.guest.ctl = data; |
3733 |
++ pt_update_intercept_for_msr(vmx); |
3734 |
++ break; |
3735 |
++ case MSR_IA32_RTIT_STATUS: |
3736 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3737 |
++ (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
3738 |
++ (data & MSR_IA32_RTIT_STATUS_MASK)) |
3739 |
++ return 1; |
3740 |
++ vmx->pt_desc.guest.status = data; |
3741 |
++ break; |
3742 |
++ case MSR_IA32_RTIT_CR3_MATCH: |
3743 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3744 |
++ (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
3745 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3746 |
++ PT_CAP_cr3_filtering)) |
3747 |
++ return 1; |
3748 |
++ vmx->pt_desc.guest.cr3_match = data; |
3749 |
++ break; |
3750 |
++ case MSR_IA32_RTIT_OUTPUT_BASE: |
3751 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3752 |
++ (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
3753 |
++ (!intel_pt_validate_cap(vmx->pt_desc.caps, |
3754 |
++ PT_CAP_topa_output) && |
3755 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3756 |
++ PT_CAP_single_range_output)) || |
3757 |
++ (data & MSR_IA32_RTIT_OUTPUT_BASE_MASK)) |
3758 |
++ return 1; |
3759 |
++ vmx->pt_desc.guest.output_base = data; |
3760 |
++ break; |
3761 |
++ case MSR_IA32_RTIT_OUTPUT_MASK: |
3762 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3763 |
++ (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
3764 |
++ (!intel_pt_validate_cap(vmx->pt_desc.caps, |
3765 |
++ PT_CAP_topa_output) && |
3766 |
++ !intel_pt_validate_cap(vmx->pt_desc.caps, |
3767 |
++ PT_CAP_single_range_output))) |
3768 |
++ return 1; |
3769 |
++ vmx->pt_desc.guest.output_mask = data; |
3770 |
++ break; |
3771 |
++ case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B: |
3772 |
++ index = msr_info->index - MSR_IA32_RTIT_ADDR0_A; |
3773 |
++ if ((pt_mode != PT_MODE_HOST_GUEST) || |
3774 |
++ (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
3775 |
++ (index >= 2 * intel_pt_validate_cap(vmx->pt_desc.caps, |
3776 |
++ PT_CAP_num_address_ranges))) |
3777 |
++ return 1; |
3778 |
++ if (is_noncanonical_address(data, vcpu)) |
3779 |
++ return 1; |
3780 |
++ if (index % 2) |
3781 |
++ vmx->pt_desc.guest.addr_b[index / 2] = data; |
3782 |
++ else |
3783 |
++ vmx->pt_desc.guest.addr_a[index / 2] = data; |
3784 |
++ break; |
3785 |
++ case MSR_TSC_AUX: |
3786 |
++ if (!msr_info->host_initiated && |
3787 |
++ !guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP)) |
3788 |
++ return 1; |
3789 |
++ /* Check reserved bit, higher 32 bits should be zero */ |
3790 |
++ if ((data >> 32) != 0) |
3791 |
++ return 1; |
3792 |
++ goto find_shared_msr; |
3793 |
++ |
3794 |
++ default: |
3795 |
++ find_shared_msr: |
3796 |
++ msr = find_msr_entry(vmx, msr_index); |
3797 |
++ if (msr) |
3798 |
++ ret = vmx_set_guest_msr(vmx, msr, data); |
3799 |
++ else |
3800 |
++ ret = kvm_set_msr_common(vcpu, msr_info); |
3801 |
++ } |
3802 |
++ |
3803 |
++ return ret; |
3804 |
++} |
3805 |
++ |
3806 |
++static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg) |
3807 |
++{ |
3808 |
++ kvm_register_mark_available(vcpu, reg); |
3809 |
++ |
3810 |
++ switch (reg) { |
3811 |
++ case VCPU_REGS_RSP: |
3812 |
++ vcpu->arch.regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP); |
3813 |
++ break; |
3814 |
++ case VCPU_REGS_RIP: |
3815 |
++ vcpu->arch.regs[VCPU_REGS_RIP] = vmcs_readl(GUEST_RIP); |
3816 |
++ break; |
3817 |
++ case VCPU_EXREG_PDPTR: |
3818 |
++ if (enable_ept) |
3819 |
++ ept_save_pdptrs(vcpu); |
3820 |
++ break; |
3821 |
++ case VCPU_EXREG_CR3: |
3822 |
++ if (enable_unrestricted_guest || (enable_ept && is_paging(vcpu))) |
3823 |
++ vcpu->arch.cr3 = vmcs_readl(GUEST_CR3); |
3824 |
++ break; |
3825 |
++ default: |
3826 |
++ WARN_ON_ONCE(1); |
3827 |
++ break; |
3828 |
++ } |
3829 |
++} |
3830 |
++ |
3831 |
++static __init int cpu_has_kvm_support(void) |
3832 |
++{ |
3833 |
++ return cpu_has_vmx(); |
3834 |
++} |
3835 |
++ |
3836 |
++static __init int vmx_disabled_by_bios(void) |
3837 |
++{ |
3838 |
++ u64 msr; |
3839 |
++ |
3840 |
++ rdmsrl(MSR_IA32_FEATURE_CONTROL, msr); |
3841 |
++ if (msr & FEATURE_CONTROL_LOCKED) { |
3842 |
++ /* launched w/ TXT and VMX disabled */ |
3843 |
++ if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
3844 |
++ && tboot_enabled()) |
3845 |
++ return 1; |
3846 |
++ /* launched w/o TXT and VMX only enabled w/ TXT */ |
3847 |
++ if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) |
3848 |
++ && (msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
3849 |
++ && !tboot_enabled()) { |
3850 |
++ printk(KERN_WARNING "kvm: disable TXT in the BIOS or " |
3851 |
++ "activate TXT before enabling KVM\n"); |
3852 |
++ return 1; |
3853 |
++ } |
3854 |
++ /* launched w/o TXT and VMX disabled */ |
3855 |
++ if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) |
3856 |
++ && !tboot_enabled()) |
3857 |
++ return 1; |
3858 |
++ } |
3859 |
++ |
3860 |
++ return 0; |
3861 |
++} |
3862 |
++ |
3863 |
++static void kvm_cpu_vmxon(u64 addr) |
3864 |
++{ |
3865 |
++ cr4_set_bits(X86_CR4_VMXE); |
3866 |
++ intel_pt_handle_vmx(1); |
3867 |
++ |
3868 |
++ asm volatile ("vmxon %0" : : "m"(addr)); |
3869 |
++} |
3870 |
++ |
3871 |
++static int hardware_enable(void) |
3872 |
++{ |
3873 |
++ int cpu = raw_smp_processor_id(); |
3874 |
++ u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); |
3875 |
++ u64 old, test_bits; |
3876 |
++ |
3877 |
++ if (cr4_read_shadow() & X86_CR4_VMXE) |
3878 |
++ return -EBUSY; |
3879 |
++ |
3880 |
++ /* |
3881 |
++ * This can happen if we hot-added a CPU but failed to allocate |
3882 |
++ * VP assist page for it. |
3883 |
++ */ |
3884 |
++ if (static_branch_unlikely(&enable_evmcs) && |
3885 |
++ !hv_get_vp_assist_page(cpu)) |
3886 |
++ return -EFAULT; |
3887 |
++ |
3888 |
++ INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu)); |
3889 |
++ INIT_LIST_HEAD(&per_cpu(blocked_vcpu_on_cpu, cpu)); |
3890 |
++ spin_lock_init(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
3891 |
++ |
3892 |
++ /* |
3893 |
++ * Now we can enable the vmclear operation in kdump |
3894 |
++ * since the loaded_vmcss_on_cpu list on this cpu |
3895 |
++ * has been initialized. |
3896 |
++ * |
3897 |
++ * Though the cpu is not in VMX operation now, there |
3898 |
++ * is no problem to enable the vmclear operation |
3899 |
++ * for the loaded_vmcss_on_cpu list is empty! |
3900 |
++ */ |
3901 |
++ crash_enable_local_vmclear(cpu); |
3902 |
++ |
3903 |
++ rdmsrl(MSR_IA32_FEATURE_CONTROL, old); |
3904 |
++ |
3905 |
++ test_bits = FEATURE_CONTROL_LOCKED; |
3906 |
++ test_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; |
3907 |
++ if (tboot_enabled()) |
3908 |
++ test_bits |= FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX; |
3909 |
++ |
3910 |
++ if ((old & test_bits) != test_bits) { |
3911 |
++ /* enable and lock */ |
3912 |
++ wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits); |
3913 |
++ } |
3914 |
++ kvm_cpu_vmxon(phys_addr); |
3915 |
++ if (enable_ept) |
3916 |
++ ept_sync_global(); |
3917 |
++ |
3918 |
++ return 0; |
3919 |
++} |
3920 |
++ |
3921 |
++static void vmclear_local_loaded_vmcss(void) |
3922 |
++{ |
3923 |
++ int cpu = raw_smp_processor_id(); |
3924 |
++ struct loaded_vmcs *v, *n; |
3925 |
++ |
3926 |
++ list_for_each_entry_safe(v, n, &per_cpu(loaded_vmcss_on_cpu, cpu), |
3927 |
++ loaded_vmcss_on_cpu_link) |
3928 |
++ __loaded_vmcs_clear(v); |
3929 |
++} |
3930 |
++ |
3931 |
++ |
3932 |
++/* Just like cpu_vmxoff(), but with the __kvm_handle_fault_on_reboot() |
3933 |
++ * tricks. |
3934 |
++ */ |
3935 |
++static void kvm_cpu_vmxoff(void) |
3936 |
++{ |
3937 |
++ asm volatile (__ex("vmxoff")); |
3938 |
++ |
3939 |
++ intel_pt_handle_vmx(0); |
3940 |
++ cr4_clear_bits(X86_CR4_VMXE); |
3941 |
++} |
3942 |
++ |
3943 |
++static void hardware_disable(void) |
3944 |
++{ |
3945 |
++ vmclear_local_loaded_vmcss(); |
3946 |
++ kvm_cpu_vmxoff(); |
3947 |
++} |
3948 |
++ |
3949 |
++static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt, |
3950 |
++ u32 msr, u32 *result) |
3951 |
++{ |
3952 |
++ u32 vmx_msr_low, vmx_msr_high; |
3953 |
++ u32 ctl = ctl_min | ctl_opt; |
3954 |
++ |
3955 |
++ rdmsr(msr, vmx_msr_low, vmx_msr_high); |
3956 |
++ |
3957 |
++ ctl &= vmx_msr_high; /* bit == 0 in high word ==> must be zero */ |
3958 |
++ ctl |= vmx_msr_low; /* bit == 1 in low word ==> must be one */ |
3959 |
++ |
3960 |
++ /* Ensure minimum (required) set of control bits are supported. */ |
3961 |
++ if (ctl_min & ~ctl) |
3962 |
++ return -EIO; |
3963 |
++ |
3964 |
++ *result = ctl; |
3965 |
++ return 0; |
3966 |
++} |
3967 |
++ |
3968 |
++static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf, |
3969 |
++ struct vmx_capability *vmx_cap) |
3970 |
++{ |
3971 |
++ u32 vmx_msr_low, vmx_msr_high; |
3972 |
++ u32 min, opt, min2, opt2; |
3973 |
++ u32 _pin_based_exec_control = 0; |
3974 |
++ u32 _cpu_based_exec_control = 0; |
3975 |
++ u32 _cpu_based_2nd_exec_control = 0; |
3976 |
++ u32 _vmexit_control = 0; |
3977 |
++ u32 _vmentry_control = 0; |
3978 |
++ |
3979 |
++ memset(vmcs_conf, 0, sizeof(*vmcs_conf)); |
3980 |
++ min = CPU_BASED_HLT_EXITING | |
3981 |
++#ifdef CONFIG_X86_64 |
3982 |
++ CPU_BASED_CR8_LOAD_EXITING | |
3983 |
++ CPU_BASED_CR8_STORE_EXITING | |
3984 |
++#endif |
3985 |
++ CPU_BASED_CR3_LOAD_EXITING | |
3986 |
++ CPU_BASED_CR3_STORE_EXITING | |
3987 |
++ CPU_BASED_UNCOND_IO_EXITING | |
3988 |
++ CPU_BASED_MOV_DR_EXITING | |
3989 |
++ CPU_BASED_USE_TSC_OFFSETTING | |
3990 |
++ CPU_BASED_MWAIT_EXITING | |
3991 |
++ CPU_BASED_MONITOR_EXITING | |
3992 |
++ CPU_BASED_INVLPG_EXITING | |
3993 |
++ CPU_BASED_RDPMC_EXITING; |
3994 |
++ |
3995 |
++ opt = CPU_BASED_TPR_SHADOW | |
3996 |
++ CPU_BASED_USE_MSR_BITMAPS | |
3997 |
++ CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; |
3998 |
++ if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PROCBASED_CTLS, |
3999 |
++ &_cpu_based_exec_control) < 0) |
4000 |
++ return -EIO; |
4001 |
++#ifdef CONFIG_X86_64 |
4002 |
++ if ((_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) |
4003 |
++ _cpu_based_exec_control &= ~CPU_BASED_CR8_LOAD_EXITING & |
4004 |
++ ~CPU_BASED_CR8_STORE_EXITING; |
4005 |
++#endif |
4006 |
++ if (_cpu_based_exec_control & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) { |
4007 |
++ min2 = 0; |
4008 |
++ opt2 = SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
4009 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
4010 |
++ SECONDARY_EXEC_WBINVD_EXITING | |
4011 |
++ SECONDARY_EXEC_ENABLE_VPID | |
4012 |
++ SECONDARY_EXEC_ENABLE_EPT | |
4013 |
++ SECONDARY_EXEC_UNRESTRICTED_GUEST | |
4014 |
++ SECONDARY_EXEC_PAUSE_LOOP_EXITING | |
4015 |
++ SECONDARY_EXEC_DESC | |
4016 |
++ SECONDARY_EXEC_RDTSCP | |
4017 |
++ SECONDARY_EXEC_ENABLE_INVPCID | |
4018 |
++ SECONDARY_EXEC_APIC_REGISTER_VIRT | |
4019 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY | |
4020 |
++ SECONDARY_EXEC_SHADOW_VMCS | |
4021 |
++ SECONDARY_EXEC_XSAVES | |
4022 |
++ SECONDARY_EXEC_RDSEED_EXITING | |
4023 |
++ SECONDARY_EXEC_RDRAND_EXITING | |
4024 |
++ SECONDARY_EXEC_ENABLE_PML | |
4025 |
++ SECONDARY_EXEC_TSC_SCALING | |
4026 |
++ SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE | |
4027 |
++ SECONDARY_EXEC_PT_USE_GPA | |
4028 |
++ SECONDARY_EXEC_PT_CONCEAL_VMX | |
4029 |
++ SECONDARY_EXEC_ENABLE_VMFUNC | |
4030 |
++ SECONDARY_EXEC_ENCLS_EXITING; |
4031 |
++ if (adjust_vmx_controls(min2, opt2, |
4032 |
++ MSR_IA32_VMX_PROCBASED_CTLS2, |
4033 |
++ &_cpu_based_2nd_exec_control) < 0) |
4034 |
++ return -EIO; |
4035 |
++ } |
4036 |
++#ifndef CONFIG_X86_64 |
4037 |
++ if (!(_cpu_based_2nd_exec_control & |
4038 |
++ SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) |
4039 |
++ _cpu_based_exec_control &= ~CPU_BASED_TPR_SHADOW; |
4040 |
++#endif |
4041 |
++ |
4042 |
++ if (!(_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) |
4043 |
++ _cpu_based_2nd_exec_control &= ~( |
4044 |
++ SECONDARY_EXEC_APIC_REGISTER_VIRT | |
4045 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
4046 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
4047 |
++ |
4048 |
++ rdmsr_safe(MSR_IA32_VMX_EPT_VPID_CAP, |
4049 |
++ &vmx_cap->ept, &vmx_cap->vpid); |
4050 |
++ |
4051 |
++ if (_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_EPT) { |
4052 |
++ /* CR3 accesses and invlpg don't need to cause VM Exits when EPT |
4053 |
++ enabled */ |
4054 |
++ _cpu_based_exec_control &= ~(CPU_BASED_CR3_LOAD_EXITING | |
4055 |
++ CPU_BASED_CR3_STORE_EXITING | |
4056 |
++ CPU_BASED_INVLPG_EXITING); |
4057 |
++ } else if (vmx_cap->ept) { |
4058 |
++ vmx_cap->ept = 0; |
4059 |
++ pr_warn_once("EPT CAP should not exist if not support " |
4060 |
++ "1-setting enable EPT VM-execution control\n"); |
4061 |
++ } |
4062 |
++ if (!(_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_VPID) && |
4063 |
++ vmx_cap->vpid) { |
4064 |
++ vmx_cap->vpid = 0; |
4065 |
++ pr_warn_once("VPID CAP should not exist if not support " |
4066 |
++ "1-setting enable VPID VM-execution control\n"); |
4067 |
++ } |
4068 |
++ |
4069 |
++ min = VM_EXIT_SAVE_DEBUG_CONTROLS | VM_EXIT_ACK_INTR_ON_EXIT; |
4070 |
++#ifdef CONFIG_X86_64 |
4071 |
++ min |= VM_EXIT_HOST_ADDR_SPACE_SIZE; |
4072 |
++#endif |
4073 |
++ opt = VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL | |
4074 |
++ VM_EXIT_LOAD_IA32_PAT | |
4075 |
++ VM_EXIT_LOAD_IA32_EFER | |
4076 |
++ VM_EXIT_CLEAR_BNDCFGS | |
4077 |
++ VM_EXIT_PT_CONCEAL_PIP | |
4078 |
++ VM_EXIT_CLEAR_IA32_RTIT_CTL; |
4079 |
++ if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_EXIT_CTLS, |
4080 |
++ &_vmexit_control) < 0) |
4081 |
++ return -EIO; |
4082 |
++ |
4083 |
++ min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING; |
4084 |
++ opt = PIN_BASED_VIRTUAL_NMIS | PIN_BASED_POSTED_INTR | |
4085 |
++ PIN_BASED_VMX_PREEMPTION_TIMER; |
4086 |
++ if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PINBASED_CTLS, |
4087 |
++ &_pin_based_exec_control) < 0) |
4088 |
++ return -EIO; |
4089 |
++ |
4090 |
++ if (cpu_has_broken_vmx_preemption_timer()) |
4091 |
++ _pin_based_exec_control &= ~PIN_BASED_VMX_PREEMPTION_TIMER; |
4092 |
++ if (!(_cpu_based_2nd_exec_control & |
4093 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY)) |
4094 |
++ _pin_based_exec_control &= ~PIN_BASED_POSTED_INTR; |
4095 |
++ |
4096 |
++ min = VM_ENTRY_LOAD_DEBUG_CONTROLS; |
4097 |
++ opt = VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL | |
4098 |
++ VM_ENTRY_LOAD_IA32_PAT | |
4099 |
++ VM_ENTRY_LOAD_IA32_EFER | |
4100 |
++ VM_ENTRY_LOAD_BNDCFGS | |
4101 |
++ VM_ENTRY_PT_CONCEAL_PIP | |
4102 |
++ VM_ENTRY_LOAD_IA32_RTIT_CTL; |
4103 |
++ if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_ENTRY_CTLS, |
4104 |
++ &_vmentry_control) < 0) |
4105 |
++ return -EIO; |
4106 |
++ |
4107 |
++ /* |
4108 |
++ * Some cpus support VM_{ENTRY,EXIT}_IA32_PERF_GLOBAL_CTRL but they |
4109 |
++ * can't be used due to an errata where VM Exit may incorrectly clear |
4110 |
++ * IA32_PERF_GLOBAL_CTRL[34:32]. Workaround the errata by using the |
4111 |
++ * MSR load mechanism to switch IA32_PERF_GLOBAL_CTRL. |
4112 |
++ */ |
4113 |
++ if (boot_cpu_data.x86 == 0x6) { |
4114 |
++ switch (boot_cpu_data.x86_model) { |
4115 |
++ case 26: /* AAK155 */ |
4116 |
++ case 30: /* AAP115 */ |
4117 |
++ case 37: /* AAT100 */ |
4118 |
++ case 44: /* BC86,AAY89,BD102 */ |
4119 |
++ case 46: /* BA97 */ |
4120 |
++ _vmentry_control &= ~VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL; |
4121 |
++ _vmexit_control &= ~VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL; |
4122 |
++ pr_warn_once("kvm: VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL " |
4123 |
++ "does not work properly. Using workaround\n"); |
4124 |
++ break; |
4125 |
++ default: |
4126 |
++ break; |
4127 |
++ } |
4128 |
++ } |
4129 |
++ |
4130 |
++ |
4131 |
++ rdmsr(MSR_IA32_VMX_BASIC, vmx_msr_low, vmx_msr_high); |
4132 |
++ |
4133 |
++ /* IA-32 SDM Vol 3B: VMCS size is never greater than 4kB. */ |
4134 |
++ if ((vmx_msr_high & 0x1fff) > PAGE_SIZE) |
4135 |
++ return -EIO; |
4136 |
++ |
4137 |
++#ifdef CONFIG_X86_64 |
4138 |
++ /* IA-32 SDM Vol 3B: 64-bit CPUs always have VMX_BASIC_MSR[48]==0. */ |
4139 |
++ if (vmx_msr_high & (1u<<16)) |
4140 |
++ return -EIO; |
4141 |
++#endif |
4142 |
++ |
4143 |
++ /* Require Write-Back (WB) memory type for VMCS accesses. */ |
4144 |
++ if (((vmx_msr_high >> 18) & 15) != 6) |
4145 |
++ return -EIO; |
4146 |
++ |
4147 |
++ vmcs_conf->size = vmx_msr_high & 0x1fff; |
4148 |
++ vmcs_conf->order = get_order(vmcs_conf->size); |
4149 |
++ vmcs_conf->basic_cap = vmx_msr_high & ~0x1fff; |
4150 |
++ |
4151 |
++ vmcs_conf->revision_id = vmx_msr_low; |
4152 |
++ |
4153 |
++ vmcs_conf->pin_based_exec_ctrl = _pin_based_exec_control; |
4154 |
++ vmcs_conf->cpu_based_exec_ctrl = _cpu_based_exec_control; |
4155 |
++ vmcs_conf->cpu_based_2nd_exec_ctrl = _cpu_based_2nd_exec_control; |
4156 |
++ vmcs_conf->vmexit_ctrl = _vmexit_control; |
4157 |
++ vmcs_conf->vmentry_ctrl = _vmentry_control; |
4158 |
++ |
4159 |
++ if (static_branch_unlikely(&enable_evmcs)) |
4160 |
++ evmcs_sanitize_exec_ctrls(vmcs_conf); |
4161 |
++ |
4162 |
++ return 0; |
4163 |
++} |
4164 |
++ |
4165 |
++struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags) |
4166 |
++{ |
4167 |
++ int node = cpu_to_node(cpu); |
4168 |
++ struct page *pages; |
4169 |
++ struct vmcs *vmcs; |
4170 |
++ |
4171 |
++ pages = __alloc_pages_node(node, flags, vmcs_config.order); |
4172 |
++ if (!pages) |
4173 |
++ return NULL; |
4174 |
++ vmcs = page_address(pages); |
4175 |
++ memset(vmcs, 0, vmcs_config.size); |
4176 |
++ |
4177 |
++ /* KVM supports Enlightened VMCS v1 only */ |
4178 |
++ if (static_branch_unlikely(&enable_evmcs)) |
4179 |
++ vmcs->hdr.revision_id = KVM_EVMCS_VERSION; |
4180 |
++ else |
4181 |
++ vmcs->hdr.revision_id = vmcs_config.revision_id; |
4182 |
++ |
4183 |
++ if (shadow) |
4184 |
++ vmcs->hdr.shadow_vmcs = 1; |
4185 |
++ return vmcs; |
4186 |
++} |
4187 |
++ |
4188 |
++void free_vmcs(struct vmcs *vmcs) |
4189 |
++{ |
4190 |
++ free_pages((unsigned long)vmcs, vmcs_config.order); |
4191 |
++} |
4192 |
++ |
4193 |
++/* |
4194 |
++ * Free a VMCS, but before that VMCLEAR it on the CPU where it was last loaded |
4195 |
++ */ |
4196 |
++void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs) |
4197 |
++{ |
4198 |
++ if (!loaded_vmcs->vmcs) |
4199 |
++ return; |
4200 |
++ loaded_vmcs_clear(loaded_vmcs); |
4201 |
++ free_vmcs(loaded_vmcs->vmcs); |
4202 |
++ loaded_vmcs->vmcs = NULL; |
4203 |
++ if (loaded_vmcs->msr_bitmap) |
4204 |
++ free_page((unsigned long)loaded_vmcs->msr_bitmap); |
4205 |
++ WARN_ON(loaded_vmcs->shadow_vmcs != NULL); |
4206 |
++} |
4207 |
++ |
4208 |
++int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs) |
4209 |
++{ |
4210 |
++ loaded_vmcs->vmcs = alloc_vmcs(false); |
4211 |
++ if (!loaded_vmcs->vmcs) |
4212 |
++ return -ENOMEM; |
4213 |
++ |
4214 |
++ loaded_vmcs->shadow_vmcs = NULL; |
4215 |
++ loaded_vmcs->hv_timer_soft_disabled = false; |
4216 |
++ loaded_vmcs_init(loaded_vmcs); |
4217 |
++ |
4218 |
++ if (cpu_has_vmx_msr_bitmap()) { |
4219 |
++ loaded_vmcs->msr_bitmap = (unsigned long *) |
4220 |
++ __get_free_page(GFP_KERNEL_ACCOUNT); |
4221 |
++ if (!loaded_vmcs->msr_bitmap) |
4222 |
++ goto out_vmcs; |
4223 |
++ memset(loaded_vmcs->msr_bitmap, 0xff, PAGE_SIZE); |
4224 |
++ |
4225 |
++ if (IS_ENABLED(CONFIG_HYPERV) && |
4226 |
++ static_branch_unlikely(&enable_evmcs) && |
4227 |
++ (ms_hyperv.nested_features & HV_X64_NESTED_MSR_BITMAP)) { |
4228 |
++ struct hv_enlightened_vmcs *evmcs = |
4229 |
++ (struct hv_enlightened_vmcs *)loaded_vmcs->vmcs; |
4230 |
++ |
4231 |
++ evmcs->hv_enlightenments_control.msr_bitmap = 1; |
4232 |
++ } |
4233 |
++ } |
4234 |
++ |
4235 |
++ memset(&loaded_vmcs->host_state, 0, sizeof(struct vmcs_host_state)); |
4236 |
++ memset(&loaded_vmcs->controls_shadow, 0, |
4237 |
++ sizeof(struct vmcs_controls_shadow)); |
4238 |
++ |
4239 |
++ return 0; |
4240 |
++ |
4241 |
++out_vmcs: |
4242 |
++ free_loaded_vmcs(loaded_vmcs); |
4243 |
++ return -ENOMEM; |
4244 |
++} |
4245 |
++ |
4246 |
++static void free_kvm_area(void) |
4247 |
++{ |
4248 |
++ int cpu; |
4249 |
++ |
4250 |
++ for_each_possible_cpu(cpu) { |
4251 |
++ free_vmcs(per_cpu(vmxarea, cpu)); |
4252 |
++ per_cpu(vmxarea, cpu) = NULL; |
4253 |
++ } |
4254 |
++} |
4255 |
++ |
4256 |
++static __init int alloc_kvm_area(void) |
4257 |
++{ |
4258 |
++ int cpu; |
4259 |
++ |
4260 |
++ for_each_possible_cpu(cpu) { |
4261 |
++ struct vmcs *vmcs; |
4262 |
++ |
4263 |
++ vmcs = alloc_vmcs_cpu(false, cpu, GFP_KERNEL); |
4264 |
++ if (!vmcs) { |
4265 |
++ free_kvm_area(); |
4266 |
++ return -ENOMEM; |
4267 |
++ } |
4268 |
++ |
4269 |
++ /* |
4270 |
++ * When eVMCS is enabled, alloc_vmcs_cpu() sets |
4271 |
++ * vmcs->revision_id to KVM_EVMCS_VERSION instead of |
4272 |
++ * revision_id reported by MSR_IA32_VMX_BASIC. |
4273 |
++ * |
4274 |
++ * However, even though not explicitly documented by |
4275 |
++ * TLFS, VMXArea passed as VMXON argument should |
4276 |
++ * still be marked with revision_id reported by |
4277 |
++ * physical CPU. |
4278 |
++ */ |
4279 |
++ if (static_branch_unlikely(&enable_evmcs)) |
4280 |
++ vmcs->hdr.revision_id = vmcs_config.revision_id; |
4281 |
++ |
4282 |
++ per_cpu(vmxarea, cpu) = vmcs; |
4283 |
++ } |
4284 |
++ return 0; |
4285 |
++} |
4286 |
++ |
4287 |
++static void fix_pmode_seg(struct kvm_vcpu *vcpu, int seg, |
4288 |
++ struct kvm_segment *save) |
4289 |
++{ |
4290 |
++ if (!emulate_invalid_guest_state) { |
4291 |
++ /* |
4292 |
++ * CS and SS RPL should be equal during guest entry according |
4293 |
++ * to VMX spec, but in reality it is not always so. Since vcpu |
4294 |
++ * is in the middle of the transition from real mode to |
4295 |
++ * protected mode it is safe to assume that RPL 0 is a good |
4296 |
++ * default value. |
4297 |
++ */ |
4298 |
++ if (seg == VCPU_SREG_CS || seg == VCPU_SREG_SS) |
4299 |
++ save->selector &= ~SEGMENT_RPL_MASK; |
4300 |
++ save->dpl = save->selector & SEGMENT_RPL_MASK; |
4301 |
++ save->s = 1; |
4302 |
++ } |
4303 |
++ vmx_set_segment(vcpu, save, seg); |
4304 |
++} |
4305 |
++ |
4306 |
++static void enter_pmode(struct kvm_vcpu *vcpu) |
4307 |
++{ |
4308 |
++ unsigned long flags; |
4309 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4310 |
++ |
4311 |
++ /* |
4312 |
++ * Update real mode segment cache. It may be not up-to-date if sement |
4313 |
++ * register was written while vcpu was in a guest mode. |
4314 |
++ */ |
4315 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES); |
4316 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS); |
4317 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_FS], VCPU_SREG_FS); |
4318 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_GS], VCPU_SREG_GS); |
4319 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_SS], VCPU_SREG_SS); |
4320 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_CS], VCPU_SREG_CS); |
4321 |
++ |
4322 |
++ vmx->rmode.vm86_active = 0; |
4323 |
++ |
4324 |
++ vmx_segment_cache_clear(vmx); |
4325 |
++ |
4326 |
++ vmx_set_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
4327 |
++ |
4328 |
++ flags = vmcs_readl(GUEST_RFLAGS); |
4329 |
++ flags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
4330 |
++ flags |= vmx->rmode.save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; |
4331 |
++ vmcs_writel(GUEST_RFLAGS, flags); |
4332 |
++ |
4333 |
++ vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~X86_CR4_VME) | |
4334 |
++ (vmcs_readl(CR4_READ_SHADOW) & X86_CR4_VME)); |
4335 |
++ |
4336 |
++ update_exception_bitmap(vcpu); |
4337 |
++ |
4338 |
++ fix_pmode_seg(vcpu, VCPU_SREG_CS, &vmx->rmode.segs[VCPU_SREG_CS]); |
4339 |
++ fix_pmode_seg(vcpu, VCPU_SREG_SS, &vmx->rmode.segs[VCPU_SREG_SS]); |
4340 |
++ fix_pmode_seg(vcpu, VCPU_SREG_ES, &vmx->rmode.segs[VCPU_SREG_ES]); |
4341 |
++ fix_pmode_seg(vcpu, VCPU_SREG_DS, &vmx->rmode.segs[VCPU_SREG_DS]); |
4342 |
++ fix_pmode_seg(vcpu, VCPU_SREG_FS, &vmx->rmode.segs[VCPU_SREG_FS]); |
4343 |
++ fix_pmode_seg(vcpu, VCPU_SREG_GS, &vmx->rmode.segs[VCPU_SREG_GS]); |
4344 |
++} |
4345 |
++ |
4346 |
++static void fix_rmode_seg(int seg, struct kvm_segment *save) |
4347 |
++{ |
4348 |
++ const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
4349 |
++ struct kvm_segment var = *save; |
4350 |
++ |
4351 |
++ var.dpl = 0x3; |
4352 |
++ if (seg == VCPU_SREG_CS) |
4353 |
++ var.type = 0x3; |
4354 |
++ |
4355 |
++ if (!emulate_invalid_guest_state) { |
4356 |
++ var.selector = var.base >> 4; |
4357 |
++ var.base = var.base & 0xffff0; |
4358 |
++ var.limit = 0xffff; |
4359 |
++ var.g = 0; |
4360 |
++ var.db = 0; |
4361 |
++ var.present = 1; |
4362 |
++ var.s = 1; |
4363 |
++ var.l = 0; |
4364 |
++ var.unusable = 0; |
4365 |
++ var.type = 0x3; |
4366 |
++ var.avl = 0; |
4367 |
++ if (save->base & 0xf) |
4368 |
++ printk_once(KERN_WARNING "kvm: segment base is not " |
4369 |
++ "paragraph aligned when entering " |
4370 |
++ "protected mode (seg=%d)", seg); |
4371 |
++ } |
4372 |
++ |
4373 |
++ vmcs_write16(sf->selector, var.selector); |
4374 |
++ vmcs_writel(sf->base, var.base); |
4375 |
++ vmcs_write32(sf->limit, var.limit); |
4376 |
++ vmcs_write32(sf->ar_bytes, vmx_segment_access_rights(&var)); |
4377 |
++} |
4378 |
++ |
4379 |
++static void enter_rmode(struct kvm_vcpu *vcpu) |
4380 |
++{ |
4381 |
++ unsigned long flags; |
4382 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4383 |
++ struct kvm_vmx *kvm_vmx = to_kvm_vmx(vcpu->kvm); |
4384 |
++ |
4385 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
4386 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES); |
4387 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS); |
4388 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_FS], VCPU_SREG_FS); |
4389 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_GS], VCPU_SREG_GS); |
4390 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_SS], VCPU_SREG_SS); |
4391 |
++ vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_CS], VCPU_SREG_CS); |
4392 |
++ |
4393 |
++ vmx->rmode.vm86_active = 1; |
4394 |
++ |
4395 |
++ /* |
4396 |
++ * Very old userspace does not call KVM_SET_TSS_ADDR before entering |
4397 |
++ * vcpu. Warn the user that an update is overdue. |
4398 |
++ */ |
4399 |
++ if (!kvm_vmx->tss_addr) |
4400 |
++ printk_once(KERN_WARNING "kvm: KVM_SET_TSS_ADDR need to be " |
4401 |
++ "called before entering vcpu\n"); |
4402 |
++ |
4403 |
++ vmx_segment_cache_clear(vmx); |
4404 |
++ |
4405 |
++ vmcs_writel(GUEST_TR_BASE, kvm_vmx->tss_addr); |
4406 |
++ vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1); |
4407 |
++ vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
4408 |
++ |
4409 |
++ flags = vmcs_readl(GUEST_RFLAGS); |
4410 |
++ vmx->rmode.save_rflags = flags; |
4411 |
++ |
4412 |
++ flags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
4413 |
++ |
4414 |
++ vmcs_writel(GUEST_RFLAGS, flags); |
4415 |
++ vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | X86_CR4_VME); |
4416 |
++ update_exception_bitmap(vcpu); |
4417 |
++ |
4418 |
++ fix_rmode_seg(VCPU_SREG_SS, &vmx->rmode.segs[VCPU_SREG_SS]); |
4419 |
++ fix_rmode_seg(VCPU_SREG_CS, &vmx->rmode.segs[VCPU_SREG_CS]); |
4420 |
++ fix_rmode_seg(VCPU_SREG_ES, &vmx->rmode.segs[VCPU_SREG_ES]); |
4421 |
++ fix_rmode_seg(VCPU_SREG_DS, &vmx->rmode.segs[VCPU_SREG_DS]); |
4422 |
++ fix_rmode_seg(VCPU_SREG_GS, &vmx->rmode.segs[VCPU_SREG_GS]); |
4423 |
++ fix_rmode_seg(VCPU_SREG_FS, &vmx->rmode.segs[VCPU_SREG_FS]); |
4424 |
++ |
4425 |
++ kvm_mmu_reset_context(vcpu); |
4426 |
++} |
4427 |
++ |
4428 |
++void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer) |
4429 |
++{ |
4430 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4431 |
++ struct shared_msr_entry *msr = find_msr_entry(vmx, MSR_EFER); |
4432 |
++ |
4433 |
++ if (!msr) |
4434 |
++ return; |
4435 |
++ |
4436 |
++ vcpu->arch.efer = efer; |
4437 |
++ if (efer & EFER_LMA) { |
4438 |
++ vm_entry_controls_setbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
4439 |
++ msr->data = efer; |
4440 |
++ } else { |
4441 |
++ vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
4442 |
++ |
4443 |
++ msr->data = efer & ~EFER_LME; |
4444 |
++ } |
4445 |
++ setup_msrs(vmx); |
4446 |
++} |
4447 |
++ |
4448 |
++#ifdef CONFIG_X86_64 |
4449 |
++ |
4450 |
++static void enter_lmode(struct kvm_vcpu *vcpu) |
4451 |
++{ |
4452 |
++ u32 guest_tr_ar; |
4453 |
++ |
4454 |
++ vmx_segment_cache_clear(to_vmx(vcpu)); |
4455 |
++ |
4456 |
++ guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES); |
4457 |
++ if ((guest_tr_ar & VMX_AR_TYPE_MASK) != VMX_AR_TYPE_BUSY_64_TSS) { |
4458 |
++ pr_debug_ratelimited("%s: tss fixup for long mode. \n", |
4459 |
++ __func__); |
4460 |
++ vmcs_write32(GUEST_TR_AR_BYTES, |
4461 |
++ (guest_tr_ar & ~VMX_AR_TYPE_MASK) |
4462 |
++ | VMX_AR_TYPE_BUSY_64_TSS); |
4463 |
++ } |
4464 |
++ vmx_set_efer(vcpu, vcpu->arch.efer | EFER_LMA); |
4465 |
++} |
4466 |
++ |
4467 |
++static void exit_lmode(struct kvm_vcpu *vcpu) |
4468 |
++{ |
4469 |
++ vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
4470 |
++ vmx_set_efer(vcpu, vcpu->arch.efer & ~EFER_LMA); |
4471 |
++} |
4472 |
++ |
4473 |
++#endif |
4474 |
++ |
4475 |
++static void vmx_flush_tlb_gva(struct kvm_vcpu *vcpu, gva_t addr) |
4476 |
++{ |
4477 |
++ int vpid = to_vmx(vcpu)->vpid; |
4478 |
++ |
4479 |
++ if (!vpid_sync_vcpu_addr(vpid, addr)) |
4480 |
++ vpid_sync_context(vpid); |
4481 |
++ |
4482 |
++ /* |
4483 |
++ * If VPIDs are not supported or enabled, then the above is a no-op. |
4484 |
++ * But we don't really need a TLB flush in that case anyway, because |
4485 |
++ * each VM entry/exit includes an implicit flush when VPID is 0. |
4486 |
++ */ |
4487 |
++} |
4488 |
++ |
4489 |
++static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu) |
4490 |
++{ |
4491 |
++ ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits; |
4492 |
++ |
4493 |
++ vcpu->arch.cr0 &= ~cr0_guest_owned_bits; |
4494 |
++ vcpu->arch.cr0 |= vmcs_readl(GUEST_CR0) & cr0_guest_owned_bits; |
4495 |
++} |
4496 |
++ |
4497 |
++static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu) |
4498 |
++{ |
4499 |
++ ulong cr4_guest_owned_bits = vcpu->arch.cr4_guest_owned_bits; |
4500 |
++ |
4501 |
++ vcpu->arch.cr4 &= ~cr4_guest_owned_bits; |
4502 |
++ vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & cr4_guest_owned_bits; |
4503 |
++} |
4504 |
++ |
4505 |
++static void ept_load_pdptrs(struct kvm_vcpu *vcpu) |
4506 |
++{ |
4507 |
++ struct kvm_mmu *mmu = vcpu->arch.walk_mmu; |
4508 |
++ |
4509 |
++ if (!kvm_register_is_dirty(vcpu, VCPU_EXREG_PDPTR)) |
4510 |
++ return; |
4511 |
++ |
4512 |
++ if (is_pae_paging(vcpu)) { |
4513 |
++ vmcs_write64(GUEST_PDPTR0, mmu->pdptrs[0]); |
4514 |
++ vmcs_write64(GUEST_PDPTR1, mmu->pdptrs[1]); |
4515 |
++ vmcs_write64(GUEST_PDPTR2, mmu->pdptrs[2]); |
4516 |
++ vmcs_write64(GUEST_PDPTR3, mmu->pdptrs[3]); |
4517 |
++ } |
4518 |
++} |
4519 |
++ |
4520 |
++void ept_save_pdptrs(struct kvm_vcpu *vcpu) |
4521 |
++{ |
4522 |
++ struct kvm_mmu *mmu = vcpu->arch.walk_mmu; |
4523 |
++ |
4524 |
++ if (is_pae_paging(vcpu)) { |
4525 |
++ mmu->pdptrs[0] = vmcs_read64(GUEST_PDPTR0); |
4526 |
++ mmu->pdptrs[1] = vmcs_read64(GUEST_PDPTR1); |
4527 |
++ mmu->pdptrs[2] = vmcs_read64(GUEST_PDPTR2); |
4528 |
++ mmu->pdptrs[3] = vmcs_read64(GUEST_PDPTR3); |
4529 |
++ } |
4530 |
++ |
4531 |
++ kvm_register_mark_dirty(vcpu, VCPU_EXREG_PDPTR); |
4532 |
++} |
4533 |
++ |
4534 |
++static void ept_update_paging_mode_cr0(unsigned long *hw_cr0, |
4535 |
++ unsigned long cr0, |
4536 |
++ struct kvm_vcpu *vcpu) |
4537 |
++{ |
4538 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4539 |
++ |
4540 |
++ if (!kvm_register_is_available(vcpu, VCPU_EXREG_CR3)) |
4541 |
++ vmx_cache_reg(vcpu, VCPU_EXREG_CR3); |
4542 |
++ if (!(cr0 & X86_CR0_PG)) { |
4543 |
++ /* From paging/starting to nonpaging */ |
4544 |
++ exec_controls_setbit(vmx, CPU_BASED_CR3_LOAD_EXITING | |
4545 |
++ CPU_BASED_CR3_STORE_EXITING); |
4546 |
++ vcpu->arch.cr0 = cr0; |
4547 |
++ vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
4548 |
++ } else if (!is_paging(vcpu)) { |
4549 |
++ /* From nonpaging to paging */ |
4550 |
++ exec_controls_clearbit(vmx, CPU_BASED_CR3_LOAD_EXITING | |
4551 |
++ CPU_BASED_CR3_STORE_EXITING); |
4552 |
++ vcpu->arch.cr0 = cr0; |
4553 |
++ vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
4554 |
++ } |
4555 |
++ |
4556 |
++ if (!(cr0 & X86_CR0_WP)) |
4557 |
++ *hw_cr0 &= ~X86_CR0_WP; |
4558 |
++} |
4559 |
++ |
4560 |
++void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
4561 |
++{ |
4562 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4563 |
++ unsigned long hw_cr0; |
4564 |
++ |
4565 |
++ hw_cr0 = (cr0 & ~KVM_VM_CR0_ALWAYS_OFF); |
4566 |
++ if (enable_unrestricted_guest) |
4567 |
++ hw_cr0 |= KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST; |
4568 |
++ else { |
4569 |
++ hw_cr0 |= KVM_VM_CR0_ALWAYS_ON; |
4570 |
++ |
4571 |
++ if (vmx->rmode.vm86_active && (cr0 & X86_CR0_PE)) |
4572 |
++ enter_pmode(vcpu); |
4573 |
++ |
4574 |
++ if (!vmx->rmode.vm86_active && !(cr0 & X86_CR0_PE)) |
4575 |
++ enter_rmode(vcpu); |
4576 |
++ } |
4577 |
++ |
4578 |
++#ifdef CONFIG_X86_64 |
4579 |
++ if (vcpu->arch.efer & EFER_LME) { |
4580 |
++ if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) |
4581 |
++ enter_lmode(vcpu); |
4582 |
++ if (is_paging(vcpu) && !(cr0 & X86_CR0_PG)) |
4583 |
++ exit_lmode(vcpu); |
4584 |
++ } |
4585 |
++#endif |
4586 |
++ |
4587 |
++ if (enable_ept && !enable_unrestricted_guest) |
4588 |
++ ept_update_paging_mode_cr0(&hw_cr0, cr0, vcpu); |
4589 |
++ |
4590 |
++ vmcs_writel(CR0_READ_SHADOW, cr0); |
4591 |
++ vmcs_writel(GUEST_CR0, hw_cr0); |
4592 |
++ vcpu->arch.cr0 = cr0; |
4593 |
++ |
4594 |
++ /* depends on vcpu->arch.cr0 to be set to a new value */ |
4595 |
++ vmx->emulation_required = emulation_required(vcpu); |
4596 |
++} |
4597 |
++ |
4598 |
++static int get_ept_level(struct kvm_vcpu *vcpu) |
4599 |
++{ |
4600 |
++ if (cpu_has_vmx_ept_5levels() && (cpuid_maxphyaddr(vcpu) > 48)) |
4601 |
++ return 5; |
4602 |
++ return 4; |
4603 |
++} |
4604 |
++ |
4605 |
++u64 construct_eptp(struct kvm_vcpu *vcpu, unsigned long root_hpa) |
4606 |
++{ |
4607 |
++ u64 eptp = VMX_EPTP_MT_WB; |
4608 |
++ |
4609 |
++ eptp |= (get_ept_level(vcpu) == 5) ? VMX_EPTP_PWL_5 : VMX_EPTP_PWL_4; |
4610 |
++ |
4611 |
++ if (enable_ept_ad_bits && |
4612 |
++ (!is_guest_mode(vcpu) || nested_ept_ad_enabled(vcpu))) |
4613 |
++ eptp |= VMX_EPTP_AD_ENABLE_BIT; |
4614 |
++ eptp |= (root_hpa & PAGE_MASK); |
4615 |
++ |
4616 |
++ return eptp; |
4617 |
++} |
4618 |
++ |
4619 |
++void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) |
4620 |
++{ |
4621 |
++ struct kvm *kvm = vcpu->kvm; |
4622 |
++ bool update_guest_cr3 = true; |
4623 |
++ unsigned long guest_cr3; |
4624 |
++ u64 eptp; |
4625 |
++ |
4626 |
++ guest_cr3 = cr3; |
4627 |
++ if (enable_ept) { |
4628 |
++ eptp = construct_eptp(vcpu, cr3); |
4629 |
++ vmcs_write64(EPT_POINTER, eptp); |
4630 |
++ |
4631 |
++ if (kvm_x86_ops->tlb_remote_flush) { |
4632 |
++ spin_lock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
4633 |
++ to_vmx(vcpu)->ept_pointer = eptp; |
4634 |
++ to_kvm_vmx(kvm)->ept_pointers_match |
4635 |
++ = EPT_POINTERS_CHECK; |
4636 |
++ spin_unlock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
4637 |
++ } |
4638 |
++ |
4639 |
++ /* Loading vmcs02.GUEST_CR3 is handled by nested VM-Enter. */ |
4640 |
++ if (is_guest_mode(vcpu)) |
4641 |
++ update_guest_cr3 = false; |
4642 |
++ else if (!enable_unrestricted_guest && !is_paging(vcpu)) |
4643 |
++ guest_cr3 = to_kvm_vmx(kvm)->ept_identity_map_addr; |
4644 |
++ else if (test_bit(VCPU_EXREG_CR3, (ulong *)&vcpu->arch.regs_avail)) |
4645 |
++ guest_cr3 = vcpu->arch.cr3; |
4646 |
++ else /* vmcs01.GUEST_CR3 is already up-to-date. */ |
4647 |
++ update_guest_cr3 = false; |
4648 |
++ ept_load_pdptrs(vcpu); |
4649 |
++ } |
4650 |
++ |
4651 |
++ if (update_guest_cr3) |
4652 |
++ vmcs_writel(GUEST_CR3, guest_cr3); |
4653 |
++} |
4654 |
++ |
4655 |
++int vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
4656 |
++{ |
4657 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4658 |
++ /* |
4659 |
++ * Pass through host's Machine Check Enable value to hw_cr4, which |
4660 |
++ * is in force while we are in guest mode. Do not let guests control |
4661 |
++ * this bit, even if host CR4.MCE == 0. |
4662 |
++ */ |
4663 |
++ unsigned long hw_cr4; |
4664 |
++ |
4665 |
++ hw_cr4 = (cr4_read_shadow() & X86_CR4_MCE) | (cr4 & ~X86_CR4_MCE); |
4666 |
++ if (enable_unrestricted_guest) |
4667 |
++ hw_cr4 |= KVM_VM_CR4_ALWAYS_ON_UNRESTRICTED_GUEST; |
4668 |
++ else if (vmx->rmode.vm86_active) |
4669 |
++ hw_cr4 |= KVM_RMODE_VM_CR4_ALWAYS_ON; |
4670 |
++ else |
4671 |
++ hw_cr4 |= KVM_PMODE_VM_CR4_ALWAYS_ON; |
4672 |
++ |
4673 |
++ if (!boot_cpu_has(X86_FEATURE_UMIP) && vmx_umip_emulated()) { |
4674 |
++ if (cr4 & X86_CR4_UMIP) { |
4675 |
++ secondary_exec_controls_setbit(vmx, SECONDARY_EXEC_DESC); |
4676 |
++ hw_cr4 &= ~X86_CR4_UMIP; |
4677 |
++ } else if (!is_guest_mode(vcpu) || |
4678 |
++ !nested_cpu_has2(get_vmcs12(vcpu), SECONDARY_EXEC_DESC)) { |
4679 |
++ secondary_exec_controls_clearbit(vmx, SECONDARY_EXEC_DESC); |
4680 |
++ } |
4681 |
++ } |
4682 |
++ |
4683 |
++ if (cr4 & X86_CR4_VMXE) { |
4684 |
++ /* |
4685 |
++ * To use VMXON (and later other VMX instructions), a guest |
4686 |
++ * must first be able to turn on cr4.VMXE (see handle_vmon()). |
4687 |
++ * So basically the check on whether to allow nested VMX |
4688 |
++ * is here. We operate under the default treatment of SMM, |
4689 |
++ * so VMX cannot be enabled under SMM. |
4690 |
++ */ |
4691 |
++ if (!nested_vmx_allowed(vcpu) || is_smm(vcpu)) |
4692 |
++ return 1; |
4693 |
++ } |
4694 |
++ |
4695 |
++ if (vmx->nested.vmxon && !nested_cr4_valid(vcpu, cr4)) |
4696 |
++ return 1; |
4697 |
++ |
4698 |
++ vcpu->arch.cr4 = cr4; |
4699 |
++ |
4700 |
++ if (!enable_unrestricted_guest) { |
4701 |
++ if (enable_ept) { |
4702 |
++ if (!is_paging(vcpu)) { |
4703 |
++ hw_cr4 &= ~X86_CR4_PAE; |
4704 |
++ hw_cr4 |= X86_CR4_PSE; |
4705 |
++ } else if (!(cr4 & X86_CR4_PAE)) { |
4706 |
++ hw_cr4 &= ~X86_CR4_PAE; |
4707 |
++ } |
4708 |
++ } |
4709 |
++ |
4710 |
++ /* |
4711 |
++ * SMEP/SMAP/PKU is disabled if CPU is in non-paging mode in |
4712 |
++ * hardware. To emulate this behavior, SMEP/SMAP/PKU needs |
4713 |
++ * to be manually disabled when guest switches to non-paging |
4714 |
++ * mode. |
4715 |
++ * |
4716 |
++ * If !enable_unrestricted_guest, the CPU is always running |
4717 |
++ * with CR0.PG=1 and CR4 needs to be modified. |
4718 |
++ * If enable_unrestricted_guest, the CPU automatically |
4719 |
++ * disables SMEP/SMAP/PKU when the guest sets CR0.PG=0. |
4720 |
++ */ |
4721 |
++ if (!is_paging(vcpu)) |
4722 |
++ hw_cr4 &= ~(X86_CR4_SMEP | X86_CR4_SMAP | X86_CR4_PKE); |
4723 |
++ } |
4724 |
++ |
4725 |
++ vmcs_writel(CR4_READ_SHADOW, cr4); |
4726 |
++ vmcs_writel(GUEST_CR4, hw_cr4); |
4727 |
++ return 0; |
4728 |
++} |
4729 |
++ |
4730 |
++void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg) |
4731 |
++{ |
4732 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4733 |
++ u32 ar; |
4734 |
++ |
4735 |
++ if (vmx->rmode.vm86_active && seg != VCPU_SREG_LDTR) { |
4736 |
++ *var = vmx->rmode.segs[seg]; |
4737 |
++ if (seg == VCPU_SREG_TR |
4738 |
++ || var->selector == vmx_read_guest_seg_selector(vmx, seg)) |
4739 |
++ return; |
4740 |
++ var->base = vmx_read_guest_seg_base(vmx, seg); |
4741 |
++ var->selector = vmx_read_guest_seg_selector(vmx, seg); |
4742 |
++ return; |
4743 |
++ } |
4744 |
++ var->base = vmx_read_guest_seg_base(vmx, seg); |
4745 |
++ var->limit = vmx_read_guest_seg_limit(vmx, seg); |
4746 |
++ var->selector = vmx_read_guest_seg_selector(vmx, seg); |
4747 |
++ ar = vmx_read_guest_seg_ar(vmx, seg); |
4748 |
++ var->unusable = (ar >> 16) & 1; |
4749 |
++ var->type = ar & 15; |
4750 |
++ var->s = (ar >> 4) & 1; |
4751 |
++ var->dpl = (ar >> 5) & 3; |
4752 |
++ /* |
4753 |
++ * Some userspaces do not preserve unusable property. Since usable |
4754 |
++ * segment has to be present according to VMX spec we can use present |
4755 |
++ * property to amend userspace bug by making unusable segment always |
4756 |
++ * nonpresent. vmx_segment_access_rights() already marks nonpresent |
4757 |
++ * segment as unusable. |
4758 |
++ */ |
4759 |
++ var->present = !var->unusable; |
4760 |
++ var->avl = (ar >> 12) & 1; |
4761 |
++ var->l = (ar >> 13) & 1; |
4762 |
++ var->db = (ar >> 14) & 1; |
4763 |
++ var->g = (ar >> 15) & 1; |
4764 |
++} |
4765 |
++ |
4766 |
++static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg) |
4767 |
++{ |
4768 |
++ struct kvm_segment s; |
4769 |
++ |
4770 |
++ if (to_vmx(vcpu)->rmode.vm86_active) { |
4771 |
++ vmx_get_segment(vcpu, &s, seg); |
4772 |
++ return s.base; |
4773 |
++ } |
4774 |
++ return vmx_read_guest_seg_base(to_vmx(vcpu), seg); |
4775 |
++} |
4776 |
++ |
4777 |
++int vmx_get_cpl(struct kvm_vcpu *vcpu) |
4778 |
++{ |
4779 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4780 |
++ |
4781 |
++ if (unlikely(vmx->rmode.vm86_active)) |
4782 |
++ return 0; |
4783 |
++ else { |
4784 |
++ int ar = vmx_read_guest_seg_ar(vmx, VCPU_SREG_SS); |
4785 |
++ return VMX_AR_DPL(ar); |
4786 |
++ } |
4787 |
++} |
4788 |
++ |
4789 |
++static u32 vmx_segment_access_rights(struct kvm_segment *var) |
4790 |
++{ |
4791 |
++ u32 ar; |
4792 |
++ |
4793 |
++ if (var->unusable || !var->present) |
4794 |
++ ar = 1 << 16; |
4795 |
++ else { |
4796 |
++ ar = var->type & 15; |
4797 |
++ ar |= (var->s & 1) << 4; |
4798 |
++ ar |= (var->dpl & 3) << 5; |
4799 |
++ ar |= (var->present & 1) << 7; |
4800 |
++ ar |= (var->avl & 1) << 12; |
4801 |
++ ar |= (var->l & 1) << 13; |
4802 |
++ ar |= (var->db & 1) << 14; |
4803 |
++ ar |= (var->g & 1) << 15; |
4804 |
++ } |
4805 |
++ |
4806 |
++ return ar; |
4807 |
++} |
4808 |
++ |
4809 |
++void vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg) |
4810 |
++{ |
4811 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
4812 |
++ const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
4813 |
++ |
4814 |
++ vmx_segment_cache_clear(vmx); |
4815 |
++ |
4816 |
++ if (vmx->rmode.vm86_active && seg != VCPU_SREG_LDTR) { |
4817 |
++ vmx->rmode.segs[seg] = *var; |
4818 |
++ if (seg == VCPU_SREG_TR) |
4819 |
++ vmcs_write16(sf->selector, var->selector); |
4820 |
++ else if (var->s) |
4821 |
++ fix_rmode_seg(seg, &vmx->rmode.segs[seg]); |
4822 |
++ goto out; |
4823 |
++ } |
4824 |
++ |
4825 |
++ vmcs_writel(sf->base, var->base); |
4826 |
++ vmcs_write32(sf->limit, var->limit); |
4827 |
++ vmcs_write16(sf->selector, var->selector); |
4828 |
++ |
4829 |
++ /* |
4830 |
++ * Fix the "Accessed" bit in AR field of segment registers for older |
4831 |
++ * qemu binaries. |
4832 |
++ * IA32 arch specifies that at the time of processor reset the |
4833 |
++ * "Accessed" bit in the AR field of segment registers is 1. And qemu |
4834 |
++ * is setting it to 0 in the userland code. This causes invalid guest |
4835 |
++ * state vmexit when "unrestricted guest" mode is turned on. |
4836 |
++ * Fix for this setup issue in cpu_reset is being pushed in the qemu |
4837 |
++ * tree. Newer qemu binaries with that qemu fix would not need this |
4838 |
++ * kvm hack. |
4839 |
++ */ |
4840 |
++ if (enable_unrestricted_guest && (seg != VCPU_SREG_LDTR)) |
4841 |
++ var->type |= 0x1; /* Accessed */ |
4842 |
++ |
4843 |
++ vmcs_write32(sf->ar_bytes, vmx_segment_access_rights(var)); |
4844 |
++ |
4845 |
++out: |
4846 |
++ vmx->emulation_required = emulation_required(vcpu); |
4847 |
++} |
4848 |
++ |
4849 |
++static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
4850 |
++{ |
4851 |
++ u32 ar = vmx_read_guest_seg_ar(to_vmx(vcpu), VCPU_SREG_CS); |
4852 |
++ |
4853 |
++ *db = (ar >> 14) & 1; |
4854 |
++ *l = (ar >> 13) & 1; |
4855 |
++} |
4856 |
++ |
4857 |
++static void vmx_get_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4858 |
++{ |
4859 |
++ dt->size = vmcs_read32(GUEST_IDTR_LIMIT); |
4860 |
++ dt->address = vmcs_readl(GUEST_IDTR_BASE); |
4861 |
++} |
4862 |
++ |
4863 |
++static void vmx_set_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4864 |
++{ |
4865 |
++ vmcs_write32(GUEST_IDTR_LIMIT, dt->size); |
4866 |
++ vmcs_writel(GUEST_IDTR_BASE, dt->address); |
4867 |
++} |
4868 |
++ |
4869 |
++static void vmx_get_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4870 |
++{ |
4871 |
++ dt->size = vmcs_read32(GUEST_GDTR_LIMIT); |
4872 |
++ dt->address = vmcs_readl(GUEST_GDTR_BASE); |
4873 |
++} |
4874 |
++ |
4875 |
++static void vmx_set_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4876 |
++{ |
4877 |
++ vmcs_write32(GUEST_GDTR_LIMIT, dt->size); |
4878 |
++ vmcs_writel(GUEST_GDTR_BASE, dt->address); |
4879 |
++} |
4880 |
++ |
4881 |
++static bool rmode_segment_valid(struct kvm_vcpu *vcpu, int seg) |
4882 |
++{ |
4883 |
++ struct kvm_segment var; |
4884 |
++ u32 ar; |
4885 |
++ |
4886 |
++ vmx_get_segment(vcpu, &var, seg); |
4887 |
++ var.dpl = 0x3; |
4888 |
++ if (seg == VCPU_SREG_CS) |
4889 |
++ var.type = 0x3; |
4890 |
++ ar = vmx_segment_access_rights(&var); |
4891 |
++ |
4892 |
++ if (var.base != (var.selector << 4)) |
4893 |
++ return false; |
4894 |
++ if (var.limit != 0xffff) |
4895 |
++ return false; |
4896 |
++ if (ar != 0xf3) |
4897 |
++ return false; |
4898 |
++ |
4899 |
++ return true; |
4900 |
++} |
4901 |
++ |
4902 |
++static bool code_segment_valid(struct kvm_vcpu *vcpu) |
4903 |
++{ |
4904 |
++ struct kvm_segment cs; |
4905 |
++ unsigned int cs_rpl; |
4906 |
++ |
4907 |
++ vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); |
4908 |
++ cs_rpl = cs.selector & SEGMENT_RPL_MASK; |
4909 |
++ |
4910 |
++ if (cs.unusable) |
4911 |
++ return false; |
4912 |
++ if (~cs.type & (VMX_AR_TYPE_CODE_MASK|VMX_AR_TYPE_ACCESSES_MASK)) |
4913 |
++ return false; |
4914 |
++ if (!cs.s) |
4915 |
++ return false; |
4916 |
++ if (cs.type & VMX_AR_TYPE_WRITEABLE_MASK) { |
4917 |
++ if (cs.dpl > cs_rpl) |
4918 |
++ return false; |
4919 |
++ } else { |
4920 |
++ if (cs.dpl != cs_rpl) |
4921 |
++ return false; |
4922 |
++ } |
4923 |
++ if (!cs.present) |
4924 |
++ return false; |
4925 |
++ |
4926 |
++ /* TODO: Add Reserved field check, this'll require a new member in the kvm_segment_field structure */ |
4927 |
++ return true; |
4928 |
++} |
4929 |
++ |
4930 |
++static bool stack_segment_valid(struct kvm_vcpu *vcpu) |
4931 |
++{ |
4932 |
++ struct kvm_segment ss; |
4933 |
++ unsigned int ss_rpl; |
4934 |
++ |
4935 |
++ vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); |
4936 |
++ ss_rpl = ss.selector & SEGMENT_RPL_MASK; |
4937 |
++ |
4938 |
++ if (ss.unusable) |
4939 |
++ return true; |
4940 |
++ if (ss.type != 3 && ss.type != 7) |
4941 |
++ return false; |
4942 |
++ if (!ss.s) |
4943 |
++ return false; |
4944 |
++ if (ss.dpl != ss_rpl) /* DPL != RPL */ |
4945 |
++ return false; |
4946 |
++ if (!ss.present) |
4947 |
++ return false; |
4948 |
++ |
4949 |
++ return true; |
4950 |
++} |
4951 |
++ |
4952 |
++static bool data_segment_valid(struct kvm_vcpu *vcpu, int seg) |
4953 |
++{ |
4954 |
++ struct kvm_segment var; |
4955 |
++ unsigned int rpl; |
4956 |
++ |
4957 |
++ vmx_get_segment(vcpu, &var, seg); |
4958 |
++ rpl = var.selector & SEGMENT_RPL_MASK; |
4959 |
++ |
4960 |
++ if (var.unusable) |
4961 |
++ return true; |
4962 |
++ if (!var.s) |
4963 |
++ return false; |
4964 |
++ if (!var.present) |
4965 |
++ return false; |
4966 |
++ if (~var.type & (VMX_AR_TYPE_CODE_MASK|VMX_AR_TYPE_WRITEABLE_MASK)) { |
4967 |
++ if (var.dpl < rpl) /* DPL < RPL */ |
4968 |
++ return false; |
4969 |
++ } |
4970 |
++ |
4971 |
++ /* TODO: Add other members to kvm_segment_field to allow checking for other access |
4972 |
++ * rights flags |
4973 |
++ */ |
4974 |
++ return true; |
4975 |
++} |
4976 |
++ |
4977 |
++static bool tr_valid(struct kvm_vcpu *vcpu) |
4978 |
++{ |
4979 |
++ struct kvm_segment tr; |
4980 |
++ |
4981 |
++ vmx_get_segment(vcpu, &tr, VCPU_SREG_TR); |
4982 |
++ |
4983 |
++ if (tr.unusable) |
4984 |
++ return false; |
4985 |
++ if (tr.selector & SEGMENT_TI_MASK) /* TI = 1 */ |
4986 |
++ return false; |
4987 |
++ if (tr.type != 3 && tr.type != 11) /* TODO: Check if guest is in IA32e mode */ |
4988 |
++ return false; |
4989 |
++ if (!tr.present) |
4990 |
++ return false; |
4991 |
++ |
4992 |
++ return true; |
4993 |
++} |
4994 |
++ |
4995 |
++static bool ldtr_valid(struct kvm_vcpu *vcpu) |
4996 |
++{ |
4997 |
++ struct kvm_segment ldtr; |
4998 |
++ |
4999 |
++ vmx_get_segment(vcpu, &ldtr, VCPU_SREG_LDTR); |
5000 |
++ |
5001 |
++ if (ldtr.unusable) |
5002 |
++ return true; |
5003 |
++ if (ldtr.selector & SEGMENT_TI_MASK) /* TI = 1 */ |
5004 |
++ return false; |
5005 |
++ if (ldtr.type != 2) |
5006 |
++ return false; |
5007 |
++ if (!ldtr.present) |
5008 |
++ return false; |
5009 |
++ |
5010 |
++ return true; |
5011 |
++} |
5012 |
++ |
5013 |
++static bool cs_ss_rpl_check(struct kvm_vcpu *vcpu) |
5014 |
++{ |
5015 |
++ struct kvm_segment cs, ss; |
5016 |
++ |
5017 |
++ vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); |
5018 |
++ vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); |
5019 |
++ |
5020 |
++ return ((cs.selector & SEGMENT_RPL_MASK) == |
5021 |
++ (ss.selector & SEGMENT_RPL_MASK)); |
5022 |
++} |
5023 |
++ |
5024 |
++/* |
5025 |
++ * Check if guest state is valid. Returns true if valid, false if |
5026 |
++ * not. |
5027 |
++ * We assume that registers are always usable |
5028 |
++ */ |
5029 |
++static bool guest_state_valid(struct kvm_vcpu *vcpu) |
5030 |
++{ |
5031 |
++ if (enable_unrestricted_guest) |
5032 |
++ return true; |
5033 |
++ |
5034 |
++ /* real mode guest state checks */ |
5035 |
++ if (!is_protmode(vcpu) || (vmx_get_rflags(vcpu) & X86_EFLAGS_VM)) { |
5036 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_CS)) |
5037 |
++ return false; |
5038 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_SS)) |
5039 |
++ return false; |
5040 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_DS)) |
5041 |
++ return false; |
5042 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_ES)) |
5043 |
++ return false; |
5044 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_FS)) |
5045 |
++ return false; |
5046 |
++ if (!rmode_segment_valid(vcpu, VCPU_SREG_GS)) |
5047 |
++ return false; |
5048 |
++ } else { |
5049 |
++ /* protected mode guest state checks */ |
5050 |
++ if (!cs_ss_rpl_check(vcpu)) |
5051 |
++ return false; |
5052 |
++ if (!code_segment_valid(vcpu)) |
5053 |
++ return false; |
5054 |
++ if (!stack_segment_valid(vcpu)) |
5055 |
++ return false; |
5056 |
++ if (!data_segment_valid(vcpu, VCPU_SREG_DS)) |
5057 |
++ return false; |
5058 |
++ if (!data_segment_valid(vcpu, VCPU_SREG_ES)) |
5059 |
++ return false; |
5060 |
++ if (!data_segment_valid(vcpu, VCPU_SREG_FS)) |
5061 |
++ return false; |
5062 |
++ if (!data_segment_valid(vcpu, VCPU_SREG_GS)) |
5063 |
++ return false; |
5064 |
++ if (!tr_valid(vcpu)) |
5065 |
++ return false; |
5066 |
++ if (!ldtr_valid(vcpu)) |
5067 |
++ return false; |
5068 |
++ } |
5069 |
++ /* TODO: |
5070 |
++ * - Add checks on RIP |
5071 |
++ * - Add checks on RFLAGS |
5072 |
++ */ |
5073 |
++ |
5074 |
++ return true; |
5075 |
++} |
5076 |
++ |
5077 |
++static int init_rmode_tss(struct kvm *kvm) |
5078 |
++{ |
5079 |
++ gfn_t fn; |
5080 |
++ u16 data = 0; |
5081 |
++ int idx, r; |
5082 |
++ |
5083 |
++ idx = srcu_read_lock(&kvm->srcu); |
5084 |
++ fn = to_kvm_vmx(kvm)->tss_addr >> PAGE_SHIFT; |
5085 |
++ r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
5086 |
++ if (r < 0) |
5087 |
++ goto out; |
5088 |
++ data = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE; |
5089 |
++ r = kvm_write_guest_page(kvm, fn++, &data, |
5090 |
++ TSS_IOPB_BASE_OFFSET, sizeof(u16)); |
5091 |
++ if (r < 0) |
5092 |
++ goto out; |
5093 |
++ r = kvm_clear_guest_page(kvm, fn++, 0, PAGE_SIZE); |
5094 |
++ if (r < 0) |
5095 |
++ goto out; |
5096 |
++ r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
5097 |
++ if (r < 0) |
5098 |
++ goto out; |
5099 |
++ data = ~0; |
5100 |
++ r = kvm_write_guest_page(kvm, fn, &data, |
5101 |
++ RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1, |
5102 |
++ sizeof(u8)); |
5103 |
++out: |
5104 |
++ srcu_read_unlock(&kvm->srcu, idx); |
5105 |
++ return r; |
5106 |
++} |
5107 |
++ |
5108 |
++static int init_rmode_identity_map(struct kvm *kvm) |
5109 |
++{ |
5110 |
++ struct kvm_vmx *kvm_vmx = to_kvm_vmx(kvm); |
5111 |
++ int i, idx, r = 0; |
5112 |
++ kvm_pfn_t identity_map_pfn; |
5113 |
++ u32 tmp; |
5114 |
++ |
5115 |
++ /* Protect kvm_vmx->ept_identity_pagetable_done. */ |
5116 |
++ mutex_lock(&kvm->slots_lock); |
5117 |
++ |
5118 |
++ if (likely(kvm_vmx->ept_identity_pagetable_done)) |
5119 |
++ goto out2; |
5120 |
++ |
5121 |
++ if (!kvm_vmx->ept_identity_map_addr) |
5122 |
++ kvm_vmx->ept_identity_map_addr = VMX_EPT_IDENTITY_PAGETABLE_ADDR; |
5123 |
++ identity_map_pfn = kvm_vmx->ept_identity_map_addr >> PAGE_SHIFT; |
5124 |
++ |
5125 |
++ r = __x86_set_memory_region(kvm, IDENTITY_PAGETABLE_PRIVATE_MEMSLOT, |
5126 |
++ kvm_vmx->ept_identity_map_addr, PAGE_SIZE); |
5127 |
++ if (r < 0) |
5128 |
++ goto out2; |
5129 |
++ |
5130 |
++ idx = srcu_read_lock(&kvm->srcu); |
5131 |
++ r = kvm_clear_guest_page(kvm, identity_map_pfn, 0, PAGE_SIZE); |
5132 |
++ if (r < 0) |
5133 |
++ goto out; |
5134 |
++ /* Set up identity-mapping pagetable for EPT in real mode */ |
5135 |
++ for (i = 0; i < PT32_ENT_PER_PAGE; i++) { |
5136 |
++ tmp = (i << 22) + (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | |
5137 |
++ _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_PSE); |
5138 |
++ r = kvm_write_guest_page(kvm, identity_map_pfn, |
5139 |
++ &tmp, i * sizeof(tmp), sizeof(tmp)); |
5140 |
++ if (r < 0) |
5141 |
++ goto out; |
5142 |
++ } |
5143 |
++ kvm_vmx->ept_identity_pagetable_done = true; |
5144 |
++ |
5145 |
++out: |
5146 |
++ srcu_read_unlock(&kvm->srcu, idx); |
5147 |
++ |
5148 |
++out2: |
5149 |
++ mutex_unlock(&kvm->slots_lock); |
5150 |
++ return r; |
5151 |
++} |
5152 |
++ |
5153 |
++static void seg_setup(int seg) |
5154 |
++{ |
5155 |
++ const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
5156 |
++ unsigned int ar; |
5157 |
++ |
5158 |
++ vmcs_write16(sf->selector, 0); |
5159 |
++ vmcs_writel(sf->base, 0); |
5160 |
++ vmcs_write32(sf->limit, 0xffff); |
5161 |
++ ar = 0x93; |
5162 |
++ if (seg == VCPU_SREG_CS) |
5163 |
++ ar |= 0x08; /* code segment */ |
5164 |
++ |
5165 |
++ vmcs_write32(sf->ar_bytes, ar); |
5166 |
++} |
5167 |
++ |
5168 |
++static int alloc_apic_access_page(struct kvm *kvm) |
5169 |
++{ |
5170 |
++ struct page *page; |
5171 |
++ int r = 0; |
5172 |
++ |
5173 |
++ mutex_lock(&kvm->slots_lock); |
5174 |
++ if (kvm->arch.apic_access_page_done) |
5175 |
++ goto out; |
5176 |
++ r = __x86_set_memory_region(kvm, APIC_ACCESS_PAGE_PRIVATE_MEMSLOT, |
5177 |
++ APIC_DEFAULT_PHYS_BASE, PAGE_SIZE); |
5178 |
++ if (r) |
5179 |
++ goto out; |
5180 |
++ |
5181 |
++ page = gfn_to_page(kvm, APIC_DEFAULT_PHYS_BASE >> PAGE_SHIFT); |
5182 |
++ if (is_error_page(page)) { |
5183 |
++ r = -EFAULT; |
5184 |
++ goto out; |
5185 |
++ } |
5186 |
++ |
5187 |
++ /* |
5188 |
++ * Do not pin the page in memory, so that memory hot-unplug |
5189 |
++ * is able to migrate it. |
5190 |
++ */ |
5191 |
++ put_page(page); |
5192 |
++ kvm->arch.apic_access_page_done = true; |
5193 |
++out: |
5194 |
++ mutex_unlock(&kvm->slots_lock); |
5195 |
++ return r; |
5196 |
++} |
5197 |
++ |
5198 |
++int allocate_vpid(void) |
5199 |
++{ |
5200 |
++ int vpid; |
5201 |
++ |
5202 |
++ if (!enable_vpid) |
5203 |
++ return 0; |
5204 |
++ spin_lock(&vmx_vpid_lock); |
5205 |
++ vpid = find_first_zero_bit(vmx_vpid_bitmap, VMX_NR_VPIDS); |
5206 |
++ if (vpid < VMX_NR_VPIDS) |
5207 |
++ __set_bit(vpid, vmx_vpid_bitmap); |
5208 |
++ else |
5209 |
++ vpid = 0; |
5210 |
++ spin_unlock(&vmx_vpid_lock); |
5211 |
++ return vpid; |
5212 |
++} |
5213 |
++ |
5214 |
++void free_vpid(int vpid) |
5215 |
++{ |
5216 |
++ if (!enable_vpid || vpid == 0) |
5217 |
++ return; |
5218 |
++ spin_lock(&vmx_vpid_lock); |
5219 |
++ __clear_bit(vpid, vmx_vpid_bitmap); |
5220 |
++ spin_unlock(&vmx_vpid_lock); |
5221 |
++} |
5222 |
++ |
5223 |
++static __always_inline void vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, |
5224 |
++ u32 msr, int type) |
5225 |
++{ |
5226 |
++ int f = sizeof(unsigned long); |
5227 |
++ |
5228 |
++ if (!cpu_has_vmx_msr_bitmap()) |
5229 |
++ return; |
5230 |
++ |
5231 |
++ if (static_branch_unlikely(&enable_evmcs)) |
5232 |
++ evmcs_touch_msr_bitmap(); |
5233 |
++ |
5234 |
++ /* |
5235 |
++ * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals |
5236 |
++ * have the write-low and read-high bitmap offsets the wrong way round. |
5237 |
++ * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. |
5238 |
++ */ |
5239 |
++ if (msr <= 0x1fff) { |
5240 |
++ if (type & MSR_TYPE_R) |
5241 |
++ /* read-low */ |
5242 |
++ __clear_bit(msr, msr_bitmap + 0x000 / f); |
5243 |
++ |
5244 |
++ if (type & MSR_TYPE_W) |
5245 |
++ /* write-low */ |
5246 |
++ __clear_bit(msr, msr_bitmap + 0x800 / f); |
5247 |
++ |
5248 |
++ } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
5249 |
++ msr &= 0x1fff; |
5250 |
++ if (type & MSR_TYPE_R) |
5251 |
++ /* read-high */ |
5252 |
++ __clear_bit(msr, msr_bitmap + 0x400 / f); |
5253 |
++ |
5254 |
++ if (type & MSR_TYPE_W) |
5255 |
++ /* write-high */ |
5256 |
++ __clear_bit(msr, msr_bitmap + 0xc00 / f); |
5257 |
++ |
5258 |
++ } |
5259 |
++} |
5260 |
++ |
5261 |
++static __always_inline void vmx_enable_intercept_for_msr(unsigned long *msr_bitmap, |
5262 |
++ u32 msr, int type) |
5263 |
++{ |
5264 |
++ int f = sizeof(unsigned long); |
5265 |
++ |
5266 |
++ if (!cpu_has_vmx_msr_bitmap()) |
5267 |
++ return; |
5268 |
++ |
5269 |
++ if (static_branch_unlikely(&enable_evmcs)) |
5270 |
++ evmcs_touch_msr_bitmap(); |
5271 |
++ |
5272 |
++ /* |
5273 |
++ * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals |
5274 |
++ * have the write-low and read-high bitmap offsets the wrong way round. |
5275 |
++ * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. |
5276 |
++ */ |
5277 |
++ if (msr <= 0x1fff) { |
5278 |
++ if (type & MSR_TYPE_R) |
5279 |
++ /* read-low */ |
5280 |
++ __set_bit(msr, msr_bitmap + 0x000 / f); |
5281 |
++ |
5282 |
++ if (type & MSR_TYPE_W) |
5283 |
++ /* write-low */ |
5284 |
++ __set_bit(msr, msr_bitmap + 0x800 / f); |
5285 |
++ |
5286 |
++ } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
5287 |
++ msr &= 0x1fff; |
5288 |
++ if (type & MSR_TYPE_R) |
5289 |
++ /* read-high */ |
5290 |
++ __set_bit(msr, msr_bitmap + 0x400 / f); |
5291 |
++ |
5292 |
++ if (type & MSR_TYPE_W) |
5293 |
++ /* write-high */ |
5294 |
++ __set_bit(msr, msr_bitmap + 0xc00 / f); |
5295 |
++ |
5296 |
++ } |
5297 |
++} |
5298 |
++ |
5299 |
++static __always_inline void vmx_set_intercept_for_msr(unsigned long *msr_bitmap, |
5300 |
++ u32 msr, int type, bool value) |
5301 |
++{ |
5302 |
++ if (value) |
5303 |
++ vmx_enable_intercept_for_msr(msr_bitmap, msr, type); |
5304 |
++ else |
5305 |
++ vmx_disable_intercept_for_msr(msr_bitmap, msr, type); |
5306 |
++} |
5307 |
++ |
5308 |
++static u8 vmx_msr_bitmap_mode(struct kvm_vcpu *vcpu) |
5309 |
++{ |
5310 |
++ u8 mode = 0; |
5311 |
++ |
5312 |
++ if (cpu_has_secondary_exec_ctrls() && |
5313 |
++ (secondary_exec_controls_get(to_vmx(vcpu)) & |
5314 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE)) { |
5315 |
++ mode |= MSR_BITMAP_MODE_X2APIC; |
5316 |
++ if (enable_apicv && kvm_vcpu_apicv_active(vcpu)) |
5317 |
++ mode |= MSR_BITMAP_MODE_X2APIC_APICV; |
5318 |
++ } |
5319 |
++ |
5320 |
++ return mode; |
5321 |
++} |
5322 |
++ |
5323 |
++static void vmx_update_msr_bitmap_x2apic(unsigned long *msr_bitmap, |
5324 |
++ u8 mode) |
5325 |
++{ |
5326 |
++ int msr; |
5327 |
++ |
5328 |
++ for (msr = 0x800; msr <= 0x8ff; msr += BITS_PER_LONG) { |
5329 |
++ unsigned word = msr / BITS_PER_LONG; |
5330 |
++ msr_bitmap[word] = (mode & MSR_BITMAP_MODE_X2APIC_APICV) ? 0 : ~0; |
5331 |
++ msr_bitmap[word + (0x800 / sizeof(long))] = ~0; |
5332 |
++ } |
5333 |
++ |
5334 |
++ if (mode & MSR_BITMAP_MODE_X2APIC) { |
5335 |
++ /* |
5336 |
++ * TPR reads and writes can be virtualized even if virtual interrupt |
5337 |
++ * delivery is not in use. |
5338 |
++ */ |
5339 |
++ vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_TASKPRI), MSR_TYPE_RW); |
5340 |
++ if (mode & MSR_BITMAP_MODE_X2APIC_APICV) { |
5341 |
++ vmx_enable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_TMCCT), MSR_TYPE_R); |
5342 |
++ vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_EOI), MSR_TYPE_W); |
5343 |
++ vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_SELF_IPI), MSR_TYPE_W); |
5344 |
++ } |
5345 |
++ } |
5346 |
++} |
5347 |
++ |
5348 |
++void vmx_update_msr_bitmap(struct kvm_vcpu *vcpu) |
5349 |
++{ |
5350 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5351 |
++ unsigned long *msr_bitmap = vmx->vmcs01.msr_bitmap; |
5352 |
++ u8 mode = vmx_msr_bitmap_mode(vcpu); |
5353 |
++ u8 changed = mode ^ vmx->msr_bitmap_mode; |
5354 |
++ |
5355 |
++ if (!changed) |
5356 |
++ return; |
5357 |
++ |
5358 |
++ if (changed & (MSR_BITMAP_MODE_X2APIC | MSR_BITMAP_MODE_X2APIC_APICV)) |
5359 |
++ vmx_update_msr_bitmap_x2apic(msr_bitmap, mode); |
5360 |
++ |
5361 |
++ vmx->msr_bitmap_mode = mode; |
5362 |
++} |
5363 |
++ |
5364 |
++void pt_update_intercept_for_msr(struct vcpu_vmx *vmx) |
5365 |
++{ |
5366 |
++ unsigned long *msr_bitmap = vmx->vmcs01.msr_bitmap; |
5367 |
++ bool flag = !(vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN); |
5368 |
++ u32 i; |
5369 |
++ |
5370 |
++ vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_STATUS, |
5371 |
++ MSR_TYPE_RW, flag); |
5372 |
++ vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_OUTPUT_BASE, |
5373 |
++ MSR_TYPE_RW, flag); |
5374 |
++ vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_OUTPUT_MASK, |
5375 |
++ MSR_TYPE_RW, flag); |
5376 |
++ vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_CR3_MATCH, |
5377 |
++ MSR_TYPE_RW, flag); |
5378 |
++ for (i = 0; i < vmx->pt_desc.addr_range; i++) { |
5379 |
++ vmx_set_intercept_for_msr(msr_bitmap, |
5380 |
++ MSR_IA32_RTIT_ADDR0_A + i * 2, MSR_TYPE_RW, flag); |
5381 |
++ vmx_set_intercept_for_msr(msr_bitmap, |
5382 |
++ MSR_IA32_RTIT_ADDR0_B + i * 2, MSR_TYPE_RW, flag); |
5383 |
++ } |
5384 |
++} |
5385 |
++ |
5386 |
++static bool vmx_get_enable_apicv(struct kvm *kvm) |
5387 |
++{ |
5388 |
++ return enable_apicv; |
5389 |
++} |
5390 |
++ |
5391 |
++static bool vmx_guest_apic_has_interrupt(struct kvm_vcpu *vcpu) |
5392 |
++{ |
5393 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5394 |
++ void *vapic_page; |
5395 |
++ u32 vppr; |
5396 |
++ int rvi; |
5397 |
++ |
5398 |
++ if (WARN_ON_ONCE(!is_guest_mode(vcpu)) || |
5399 |
++ !nested_cpu_has_vid(get_vmcs12(vcpu)) || |
5400 |
++ WARN_ON_ONCE(!vmx->nested.virtual_apic_map.gfn)) |
5401 |
++ return false; |
5402 |
++ |
5403 |
++ rvi = vmx_get_rvi(); |
5404 |
++ |
5405 |
++ vapic_page = vmx->nested.virtual_apic_map.hva; |
5406 |
++ vppr = *((u32 *)(vapic_page + APIC_PROCPRI)); |
5407 |
++ |
5408 |
++ return ((rvi & 0xf0) > (vppr & 0xf0)); |
5409 |
++} |
5410 |
++ |
5411 |
++static inline bool kvm_vcpu_trigger_posted_interrupt(struct kvm_vcpu *vcpu, |
5412 |
++ bool nested) |
5413 |
++{ |
5414 |
++#ifdef CONFIG_SMP |
5415 |
++ int pi_vec = nested ? POSTED_INTR_NESTED_VECTOR : POSTED_INTR_VECTOR; |
5416 |
++ |
5417 |
++ if (vcpu->mode == IN_GUEST_MODE) { |
5418 |
++ /* |
5419 |
++ * The vector of interrupt to be delivered to vcpu had |
5420 |
++ * been set in PIR before this function. |
5421 |
++ * |
5422 |
++ * Following cases will be reached in this block, and |
5423 |
++ * we always send a notification event in all cases as |
5424 |
++ * explained below. |
5425 |
++ * |
5426 |
++ * Case 1: vcpu keeps in non-root mode. Sending a |
5427 |
++ * notification event posts the interrupt to vcpu. |
5428 |
++ * |
5429 |
++ * Case 2: vcpu exits to root mode and is still |
5430 |
++ * runnable. PIR will be synced to vIRR before the |
5431 |
++ * next vcpu entry. Sending a notification event in |
5432 |
++ * this case has no effect, as vcpu is not in root |
5433 |
++ * mode. |
5434 |
++ * |
5435 |
++ * Case 3: vcpu exits to root mode and is blocked. |
5436 |
++ * vcpu_block() has already synced PIR to vIRR and |
5437 |
++ * never blocks vcpu if vIRR is not cleared. Therefore, |
5438 |
++ * a blocked vcpu here does not wait for any requested |
5439 |
++ * interrupts in PIR, and sending a notification event |
5440 |
++ * which has no effect is safe here. |
5441 |
++ */ |
5442 |
++ |
5443 |
++ apic->send_IPI_mask(get_cpu_mask(vcpu->cpu), pi_vec); |
5444 |
++ return true; |
5445 |
++ } |
5446 |
++#endif |
5447 |
++ return false; |
5448 |
++} |
5449 |
++ |
5450 |
++static int vmx_deliver_nested_posted_interrupt(struct kvm_vcpu *vcpu, |
5451 |
++ int vector) |
5452 |
++{ |
5453 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5454 |
++ |
5455 |
++ if (is_guest_mode(vcpu) && |
5456 |
++ vector == vmx->nested.posted_intr_nv) { |
5457 |
++ /* |
5458 |
++ * If a posted intr is not recognized by hardware, |
5459 |
++ * we will accomplish it in the next vmentry. |
5460 |
++ */ |
5461 |
++ vmx->nested.pi_pending = true; |
5462 |
++ kvm_make_request(KVM_REQ_EVENT, vcpu); |
5463 |
++ /* the PIR and ON have been set by L1. */ |
5464 |
++ if (!kvm_vcpu_trigger_posted_interrupt(vcpu, true)) |
5465 |
++ kvm_vcpu_kick(vcpu); |
5466 |
++ return 0; |
5467 |
++ } |
5468 |
++ return -1; |
5469 |
++} |
5470 |
++/* |
5471 |
++ * Send interrupt to vcpu via posted interrupt way. |
5472 |
++ * 1. If target vcpu is running(non-root mode), send posted interrupt |
5473 |
++ * notification to vcpu and hardware will sync PIR to vIRR atomically. |
5474 |
++ * 2. If target vcpu isn't running(root mode), kick it to pick up the |
5475 |
++ * interrupt from PIR in next vmentry. |
5476 |
++ */ |
5477 |
++static void vmx_deliver_posted_interrupt(struct kvm_vcpu *vcpu, int vector) |
5478 |
++{ |
5479 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5480 |
++ int r; |
5481 |
++ |
5482 |
++ r = vmx_deliver_nested_posted_interrupt(vcpu, vector); |
5483 |
++ if (!r) |
5484 |
++ return; |
5485 |
++ |
5486 |
++ if (pi_test_and_set_pir(vector, &vmx->pi_desc)) |
5487 |
++ return; |
5488 |
++ |
5489 |
++ /* If a previous notification has sent the IPI, nothing to do. */ |
5490 |
++ if (pi_test_and_set_on(&vmx->pi_desc)) |
5491 |
++ return; |
5492 |
++ |
5493 |
++ if (!kvm_vcpu_trigger_posted_interrupt(vcpu, false)) |
5494 |
++ kvm_vcpu_kick(vcpu); |
5495 |
++} |
5496 |
++ |
5497 |
++/* |
5498 |
++ * Set up the vmcs's constant host-state fields, i.e., host-state fields that |
5499 |
++ * will not change in the lifetime of the guest. |
5500 |
++ * Note that host-state that does change is set elsewhere. E.g., host-state |
5501 |
++ * that is set differently for each CPU is set in vmx_vcpu_load(), not here. |
5502 |
++ */ |
5503 |
++void vmx_set_constant_host_state(struct vcpu_vmx *vmx) |
5504 |
++{ |
5505 |
++ u32 low32, high32; |
5506 |
++ unsigned long tmpl; |
5507 |
++ unsigned long cr0, cr3, cr4; |
5508 |
++ |
5509 |
++ cr0 = read_cr0(); |
5510 |
++ WARN_ON(cr0 & X86_CR0_TS); |
5511 |
++ vmcs_writel(HOST_CR0, cr0); /* 22.2.3 */ |
5512 |
++ |
5513 |
++ /* |
5514 |
++ * Save the most likely value for this task's CR3 in the VMCS. |
5515 |
++ * We can't use __get_current_cr3_fast() because we're not atomic. |
5516 |
++ */ |
5517 |
++ cr3 = __read_cr3(); |
5518 |
++ vmcs_writel(HOST_CR3, cr3); /* 22.2.3 FIXME: shadow tables */ |
5519 |
++ vmx->loaded_vmcs->host_state.cr3 = cr3; |
5520 |
++ |
5521 |
++ /* Save the most likely value for this task's CR4 in the VMCS. */ |
5522 |
++ cr4 = cr4_read_shadow(); |
5523 |
++ vmcs_writel(HOST_CR4, cr4); /* 22.2.3, 22.2.5 */ |
5524 |
++ vmx->loaded_vmcs->host_state.cr4 = cr4; |
5525 |
++ |
5526 |
++ vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS); /* 22.2.4 */ |
5527 |
++#ifdef CONFIG_X86_64 |
5528 |
++ /* |
5529 |
++ * Load null selectors, so we can avoid reloading them in |
5530 |
++ * vmx_prepare_switch_to_host(), in case userspace uses |
5531 |
++ * the null selectors too (the expected case). |
5532 |
++ */ |
5533 |
++ vmcs_write16(HOST_DS_SELECTOR, 0); |
5534 |
++ vmcs_write16(HOST_ES_SELECTOR, 0); |
5535 |
++#else |
5536 |
++ vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
5537 |
++ vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
5538 |
++#endif |
5539 |
++ vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
5540 |
++ vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8); /* 22.2.4 */ |
5541 |
++ |
5542 |
++ vmcs_writel(HOST_IDTR_BASE, host_idt_base); /* 22.2.4 */ |
5543 |
++ |
5544 |
++ vmcs_writel(HOST_RIP, (unsigned long)vmx_vmexit); /* 22.2.5 */ |
5545 |
++ |
5546 |
++ rdmsr(MSR_IA32_SYSENTER_CS, low32, high32); |
5547 |
++ vmcs_write32(HOST_IA32_SYSENTER_CS, low32); |
5548 |
++ rdmsrl(MSR_IA32_SYSENTER_EIP, tmpl); |
5549 |
++ vmcs_writel(HOST_IA32_SYSENTER_EIP, tmpl); /* 22.2.3 */ |
5550 |
++ |
5551 |
++ if (vmcs_config.vmexit_ctrl & VM_EXIT_LOAD_IA32_PAT) { |
5552 |
++ rdmsr(MSR_IA32_CR_PAT, low32, high32); |
5553 |
++ vmcs_write64(HOST_IA32_PAT, low32 | ((u64) high32 << 32)); |
5554 |
++ } |
5555 |
++ |
5556 |
++ if (cpu_has_load_ia32_efer()) |
5557 |
++ vmcs_write64(HOST_IA32_EFER, host_efer); |
5558 |
++} |
5559 |
++ |
5560 |
++void set_cr4_guest_host_mask(struct vcpu_vmx *vmx) |
5561 |
++{ |
5562 |
++ vmx->vcpu.arch.cr4_guest_owned_bits = KVM_CR4_GUEST_OWNED_BITS; |
5563 |
++ if (enable_ept) |
5564 |
++ vmx->vcpu.arch.cr4_guest_owned_bits |= X86_CR4_PGE; |
5565 |
++ if (is_guest_mode(&vmx->vcpu)) |
5566 |
++ vmx->vcpu.arch.cr4_guest_owned_bits &= |
5567 |
++ ~get_vmcs12(&vmx->vcpu)->cr4_guest_host_mask; |
5568 |
++ vmcs_writel(CR4_GUEST_HOST_MASK, ~vmx->vcpu.arch.cr4_guest_owned_bits); |
5569 |
++} |
5570 |
++ |
5571 |
++u32 vmx_pin_based_exec_ctrl(struct vcpu_vmx *vmx) |
5572 |
++{ |
5573 |
++ u32 pin_based_exec_ctrl = vmcs_config.pin_based_exec_ctrl; |
5574 |
++ |
5575 |
++ if (!kvm_vcpu_apicv_active(&vmx->vcpu)) |
5576 |
++ pin_based_exec_ctrl &= ~PIN_BASED_POSTED_INTR; |
5577 |
++ |
5578 |
++ if (!enable_vnmi) |
5579 |
++ pin_based_exec_ctrl &= ~PIN_BASED_VIRTUAL_NMIS; |
5580 |
++ |
5581 |
++ if (!enable_preemption_timer) |
5582 |
++ pin_based_exec_ctrl &= ~PIN_BASED_VMX_PREEMPTION_TIMER; |
5583 |
++ |
5584 |
++ return pin_based_exec_ctrl; |
5585 |
++} |
5586 |
++ |
5587 |
++static void vmx_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu) |
5588 |
++{ |
5589 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5590 |
++ |
5591 |
++ pin_controls_set(vmx, vmx_pin_based_exec_ctrl(vmx)); |
5592 |
++ if (cpu_has_secondary_exec_ctrls()) { |
5593 |
++ if (kvm_vcpu_apicv_active(vcpu)) |
5594 |
++ secondary_exec_controls_setbit(vmx, |
5595 |
++ SECONDARY_EXEC_APIC_REGISTER_VIRT | |
5596 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
5597 |
++ else |
5598 |
++ secondary_exec_controls_clearbit(vmx, |
5599 |
++ SECONDARY_EXEC_APIC_REGISTER_VIRT | |
5600 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
5601 |
++ } |
5602 |
++ |
5603 |
++ if (cpu_has_vmx_msr_bitmap()) |
5604 |
++ vmx_update_msr_bitmap(vcpu); |
5605 |
++} |
5606 |
++ |
5607 |
++u32 vmx_exec_control(struct vcpu_vmx *vmx) |
5608 |
++{ |
5609 |
++ u32 exec_control = vmcs_config.cpu_based_exec_ctrl; |
5610 |
++ |
5611 |
++ if (vmx->vcpu.arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT) |
5612 |
++ exec_control &= ~CPU_BASED_MOV_DR_EXITING; |
5613 |
++ |
5614 |
++ if (!cpu_need_tpr_shadow(&vmx->vcpu)) { |
5615 |
++ exec_control &= ~CPU_BASED_TPR_SHADOW; |
5616 |
++#ifdef CONFIG_X86_64 |
5617 |
++ exec_control |= CPU_BASED_CR8_STORE_EXITING | |
5618 |
++ CPU_BASED_CR8_LOAD_EXITING; |
5619 |
++#endif |
5620 |
++ } |
5621 |
++ if (!enable_ept) |
5622 |
++ exec_control |= CPU_BASED_CR3_STORE_EXITING | |
5623 |
++ CPU_BASED_CR3_LOAD_EXITING | |
5624 |
++ CPU_BASED_INVLPG_EXITING; |
5625 |
++ if (kvm_mwait_in_guest(vmx->vcpu.kvm)) |
5626 |
++ exec_control &= ~(CPU_BASED_MWAIT_EXITING | |
5627 |
++ CPU_BASED_MONITOR_EXITING); |
5628 |
++ if (kvm_hlt_in_guest(vmx->vcpu.kvm)) |
5629 |
++ exec_control &= ~CPU_BASED_HLT_EXITING; |
5630 |
++ return exec_control; |
5631 |
++} |
5632 |
++ |
5633 |
++ |
5634 |
++static void vmx_compute_secondary_exec_control(struct vcpu_vmx *vmx) |
5635 |
++{ |
5636 |
++ struct kvm_vcpu *vcpu = &vmx->vcpu; |
5637 |
++ |
5638 |
++ u32 exec_control = vmcs_config.cpu_based_2nd_exec_ctrl; |
5639 |
++ |
5640 |
++ if (pt_mode == PT_MODE_SYSTEM) |
5641 |
++ exec_control &= ~(SECONDARY_EXEC_PT_USE_GPA | SECONDARY_EXEC_PT_CONCEAL_VMX); |
5642 |
++ if (!cpu_need_virtualize_apic_accesses(vcpu)) |
5643 |
++ exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; |
5644 |
++ if (vmx->vpid == 0) |
5645 |
++ exec_control &= ~SECONDARY_EXEC_ENABLE_VPID; |
5646 |
++ if (!enable_ept) { |
5647 |
++ exec_control &= ~SECONDARY_EXEC_ENABLE_EPT; |
5648 |
++ enable_unrestricted_guest = 0; |
5649 |
++ } |
5650 |
++ if (!enable_unrestricted_guest) |
5651 |
++ exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST; |
5652 |
++ if (kvm_pause_in_guest(vmx->vcpu.kvm)) |
5653 |
++ exec_control &= ~SECONDARY_EXEC_PAUSE_LOOP_EXITING; |
5654 |
++ if (!kvm_vcpu_apicv_active(vcpu)) |
5655 |
++ exec_control &= ~(SECONDARY_EXEC_APIC_REGISTER_VIRT | |
5656 |
++ SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
5657 |
++ exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE; |
5658 |
++ |
5659 |
++ /* SECONDARY_EXEC_DESC is enabled/disabled on writes to CR4.UMIP, |
5660 |
++ * in vmx_set_cr4. */ |
5661 |
++ exec_control &= ~SECONDARY_EXEC_DESC; |
5662 |
++ |
5663 |
++ /* SECONDARY_EXEC_SHADOW_VMCS is enabled when L1 executes VMPTRLD |
5664 |
++ (handle_vmptrld). |
5665 |
++ We can NOT enable shadow_vmcs here because we don't have yet |
5666 |
++ a current VMCS12 |
5667 |
++ */ |
5668 |
++ exec_control &= ~SECONDARY_EXEC_SHADOW_VMCS; |
5669 |
++ |
5670 |
++ if (!enable_pml) |
5671 |
++ exec_control &= ~SECONDARY_EXEC_ENABLE_PML; |
5672 |
++ |
5673 |
++ if (vmx_xsaves_supported()) { |
5674 |
++ /* Exposing XSAVES only when XSAVE is exposed */ |
5675 |
++ bool xsaves_enabled = |
5676 |
++ guest_cpuid_has(vcpu, X86_FEATURE_XSAVE) && |
5677 |
++ guest_cpuid_has(vcpu, X86_FEATURE_XSAVES); |
5678 |
++ |
5679 |
++ vcpu->arch.xsaves_enabled = xsaves_enabled; |
5680 |
++ |
5681 |
++ if (!xsaves_enabled) |
5682 |
++ exec_control &= ~SECONDARY_EXEC_XSAVES; |
5683 |
++ |
5684 |
++ if (nested) { |
5685 |
++ if (xsaves_enabled) |
5686 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5687 |
++ SECONDARY_EXEC_XSAVES; |
5688 |
++ else |
5689 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5690 |
++ ~SECONDARY_EXEC_XSAVES; |
5691 |
++ } |
5692 |
++ } |
5693 |
++ |
5694 |
++ if (vmx_rdtscp_supported()) { |
5695 |
++ bool rdtscp_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP); |
5696 |
++ if (!rdtscp_enabled) |
5697 |
++ exec_control &= ~SECONDARY_EXEC_RDTSCP; |
5698 |
++ |
5699 |
++ if (nested) { |
5700 |
++ if (rdtscp_enabled) |
5701 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5702 |
++ SECONDARY_EXEC_RDTSCP; |
5703 |
++ else |
5704 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5705 |
++ ~SECONDARY_EXEC_RDTSCP; |
5706 |
++ } |
5707 |
++ } |
5708 |
++ |
5709 |
++ if (vmx_invpcid_supported()) { |
5710 |
++ /* Exposing INVPCID only when PCID is exposed */ |
5711 |
++ bool invpcid_enabled = |
5712 |
++ guest_cpuid_has(vcpu, X86_FEATURE_INVPCID) && |
5713 |
++ guest_cpuid_has(vcpu, X86_FEATURE_PCID); |
5714 |
++ |
5715 |
++ if (!invpcid_enabled) { |
5716 |
++ exec_control &= ~SECONDARY_EXEC_ENABLE_INVPCID; |
5717 |
++ guest_cpuid_clear(vcpu, X86_FEATURE_INVPCID); |
5718 |
++ } |
5719 |
++ |
5720 |
++ if (nested) { |
5721 |
++ if (invpcid_enabled) |
5722 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5723 |
++ SECONDARY_EXEC_ENABLE_INVPCID; |
5724 |
++ else |
5725 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5726 |
++ ~SECONDARY_EXEC_ENABLE_INVPCID; |
5727 |
++ } |
5728 |
++ } |
5729 |
++ |
5730 |
++ if (vmx_rdrand_supported()) { |
5731 |
++ bool rdrand_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDRAND); |
5732 |
++ if (rdrand_enabled) |
5733 |
++ exec_control &= ~SECONDARY_EXEC_RDRAND_EXITING; |
5734 |
++ |
5735 |
++ if (nested) { |
5736 |
++ if (rdrand_enabled) |
5737 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5738 |
++ SECONDARY_EXEC_RDRAND_EXITING; |
5739 |
++ else |
5740 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5741 |
++ ~SECONDARY_EXEC_RDRAND_EXITING; |
5742 |
++ } |
5743 |
++ } |
5744 |
++ |
5745 |
++ if (vmx_rdseed_supported()) { |
5746 |
++ bool rdseed_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDSEED); |
5747 |
++ if (rdseed_enabled) |
5748 |
++ exec_control &= ~SECONDARY_EXEC_RDSEED_EXITING; |
5749 |
++ |
5750 |
++ if (nested) { |
5751 |
++ if (rdseed_enabled) |
5752 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5753 |
++ SECONDARY_EXEC_RDSEED_EXITING; |
5754 |
++ else |
5755 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5756 |
++ ~SECONDARY_EXEC_RDSEED_EXITING; |
5757 |
++ } |
5758 |
++ } |
5759 |
++ |
5760 |
++ if (vmx_waitpkg_supported()) { |
5761 |
++ bool waitpkg_enabled = |
5762 |
++ guest_cpuid_has(vcpu, X86_FEATURE_WAITPKG); |
5763 |
++ |
5764 |
++ if (!waitpkg_enabled) |
5765 |
++ exec_control &= ~SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
5766 |
++ |
5767 |
++ if (nested) { |
5768 |
++ if (waitpkg_enabled) |
5769 |
++ vmx->nested.msrs.secondary_ctls_high |= |
5770 |
++ SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
5771 |
++ else |
5772 |
++ vmx->nested.msrs.secondary_ctls_high &= |
5773 |
++ ~SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
5774 |
++ } |
5775 |
++ } |
5776 |
++ |
5777 |
++ vmx->secondary_exec_control = exec_control; |
5778 |
++} |
5779 |
++ |
5780 |
++static void ept_set_mmio_spte_mask(void) |
5781 |
++{ |
5782 |
++ /* |
5783 |
++ * EPT Misconfigurations can be generated if the value of bits 2:0 |
5784 |
++ * of an EPT paging-structure entry is 110b (write/execute). |
5785 |
++ */ |
5786 |
++ kvm_mmu_set_mmio_spte_mask(VMX_EPT_RWX_MASK, |
5787 |
++ VMX_EPT_MISCONFIG_WX_VALUE, 0); |
5788 |
++} |
5789 |
++ |
5790 |
++#define VMX_XSS_EXIT_BITMAP 0 |
5791 |
++ |
5792 |
++/* |
5793 |
++ * Noting that the initialization of Guest-state Area of VMCS is in |
5794 |
++ * vmx_vcpu_reset(). |
5795 |
++ */ |
5796 |
++static void init_vmcs(struct vcpu_vmx *vmx) |
5797 |
++{ |
5798 |
++ if (nested) |
5799 |
++ nested_vmx_set_vmcs_shadowing_bitmap(); |
5800 |
++ |
5801 |
++ if (cpu_has_vmx_msr_bitmap()) |
5802 |
++ vmcs_write64(MSR_BITMAP, __pa(vmx->vmcs01.msr_bitmap)); |
5803 |
++ |
5804 |
++ vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */ |
5805 |
++ |
5806 |
++ /* Control */ |
5807 |
++ pin_controls_set(vmx, vmx_pin_based_exec_ctrl(vmx)); |
5808 |
++ |
5809 |
++ exec_controls_set(vmx, vmx_exec_control(vmx)); |
5810 |
++ |
5811 |
++ if (cpu_has_secondary_exec_ctrls()) { |
5812 |
++ vmx_compute_secondary_exec_control(vmx); |
5813 |
++ secondary_exec_controls_set(vmx, vmx->secondary_exec_control); |
5814 |
++ } |
5815 |
++ |
5816 |
++ if (kvm_vcpu_apicv_active(&vmx->vcpu)) { |
5817 |
++ vmcs_write64(EOI_EXIT_BITMAP0, 0); |
5818 |
++ vmcs_write64(EOI_EXIT_BITMAP1, 0); |
5819 |
++ vmcs_write64(EOI_EXIT_BITMAP2, 0); |
5820 |
++ vmcs_write64(EOI_EXIT_BITMAP3, 0); |
5821 |
++ |
5822 |
++ vmcs_write16(GUEST_INTR_STATUS, 0); |
5823 |
++ |
5824 |
++ vmcs_write16(POSTED_INTR_NV, POSTED_INTR_VECTOR); |
5825 |
++ vmcs_write64(POSTED_INTR_DESC_ADDR, __pa((&vmx->pi_desc))); |
5826 |
++ } |
5827 |
++ |
5828 |
++ if (!kvm_pause_in_guest(vmx->vcpu.kvm)) { |
5829 |
++ vmcs_write32(PLE_GAP, ple_gap); |
5830 |
++ vmx->ple_window = ple_window; |
5831 |
++ vmx->ple_window_dirty = true; |
5832 |
++ } |
5833 |
++ |
5834 |
++ vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0); |
5835 |
++ vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0); |
5836 |
++ vmcs_write32(CR3_TARGET_COUNT, 0); /* 22.2.1 */ |
5837 |
++ |
5838 |
++ vmcs_write16(HOST_FS_SELECTOR, 0); /* 22.2.4 */ |
5839 |
++ vmcs_write16(HOST_GS_SELECTOR, 0); /* 22.2.4 */ |
5840 |
++ vmx_set_constant_host_state(vmx); |
5841 |
++ vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */ |
5842 |
++ vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */ |
5843 |
++ |
5844 |
++ if (cpu_has_vmx_vmfunc()) |
5845 |
++ vmcs_write64(VM_FUNCTION_CONTROL, 0); |
5846 |
++ |
5847 |
++ vmcs_write32(VM_EXIT_MSR_STORE_COUNT, 0); |
5848 |
++ vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, 0); |
5849 |
++ vmcs_write64(VM_EXIT_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.host.val)); |
5850 |
++ vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, 0); |
5851 |
++ vmcs_write64(VM_ENTRY_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.guest.val)); |
5852 |
++ |
5853 |
++ if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) |
5854 |
++ vmcs_write64(GUEST_IA32_PAT, vmx->vcpu.arch.pat); |
5855 |
++ |
5856 |
++ vm_exit_controls_set(vmx, vmx_vmexit_ctrl()); |
5857 |
++ |
5858 |
++ /* 22.2.1, 20.8.1 */ |
5859 |
++ vm_entry_controls_set(vmx, vmx_vmentry_ctrl()); |
5860 |
++ |
5861 |
++ vmx->vcpu.arch.cr0_guest_owned_bits = X86_CR0_TS; |
5862 |
++ vmcs_writel(CR0_GUEST_HOST_MASK, ~X86_CR0_TS); |
5863 |
++ |
5864 |
++ set_cr4_guest_host_mask(vmx); |
5865 |
++ |
5866 |
++ if (vmx->vpid != 0) |
5867 |
++ vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid); |
5868 |
++ |
5869 |
++ if (vmx_xsaves_supported()) |
5870 |
++ vmcs_write64(XSS_EXIT_BITMAP, VMX_XSS_EXIT_BITMAP); |
5871 |
++ |
5872 |
++ if (enable_pml) { |
5873 |
++ vmcs_write64(PML_ADDRESS, page_to_phys(vmx->pml_pg)); |
5874 |
++ vmcs_write16(GUEST_PML_INDEX, PML_ENTITY_NUM - 1); |
5875 |
++ } |
5876 |
++ |
5877 |
++ if (cpu_has_vmx_encls_vmexit()) |
5878 |
++ vmcs_write64(ENCLS_EXITING_BITMAP, -1ull); |
5879 |
++ |
5880 |
++ if (pt_mode == PT_MODE_HOST_GUEST) { |
5881 |
++ memset(&vmx->pt_desc, 0, sizeof(vmx->pt_desc)); |
5882 |
++ /* Bit[6~0] are forced to 1, writes are ignored. */ |
5883 |
++ vmx->pt_desc.guest.output_mask = 0x7F; |
5884 |
++ vmcs_write64(GUEST_IA32_RTIT_CTL, 0); |
5885 |
++ } |
5886 |
++} |
5887 |
++ |
5888 |
++static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event) |
5889 |
++{ |
5890 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
5891 |
++ struct msr_data apic_base_msr; |
5892 |
++ u64 cr0; |
5893 |
++ |
5894 |
++ vmx->rmode.vm86_active = 0; |
5895 |
++ vmx->spec_ctrl = 0; |
5896 |
++ |
5897 |
++ vmx->msr_ia32_umwait_control = 0; |
5898 |
++ |
5899 |
++ vcpu->arch.microcode_version = 0x100000000ULL; |
5900 |
++ vmx->vcpu.arch.regs[VCPU_REGS_RDX] = get_rdx_init_val(); |
5901 |
++ vmx->hv_deadline_tsc = -1; |
5902 |
++ kvm_set_cr8(vcpu, 0); |
5903 |
++ |
5904 |
++ if (!init_event) { |
5905 |
++ apic_base_msr.data = APIC_DEFAULT_PHYS_BASE | |
5906 |
++ MSR_IA32_APICBASE_ENABLE; |
5907 |
++ if (kvm_vcpu_is_reset_bsp(vcpu)) |
5908 |
++ apic_base_msr.data |= MSR_IA32_APICBASE_BSP; |
5909 |
++ apic_base_msr.host_initiated = true; |
5910 |
++ kvm_set_apic_base(vcpu, &apic_base_msr); |
5911 |
++ } |
5912 |
++ |
5913 |
++ vmx_segment_cache_clear(vmx); |
5914 |
++ |
5915 |
++ seg_setup(VCPU_SREG_CS); |
5916 |
++ vmcs_write16(GUEST_CS_SELECTOR, 0xf000); |
5917 |
++ vmcs_writel(GUEST_CS_BASE, 0xffff0000ul); |
5918 |
++ |
5919 |
++ seg_setup(VCPU_SREG_DS); |
5920 |
++ seg_setup(VCPU_SREG_ES); |
5921 |
++ seg_setup(VCPU_SREG_FS); |
5922 |
++ seg_setup(VCPU_SREG_GS); |
5923 |
++ seg_setup(VCPU_SREG_SS); |
5924 |
++ |
5925 |
++ vmcs_write16(GUEST_TR_SELECTOR, 0); |
5926 |
++ vmcs_writel(GUEST_TR_BASE, 0); |
5927 |
++ vmcs_write32(GUEST_TR_LIMIT, 0xffff); |
5928 |
++ vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
5929 |
++ |
5930 |
++ vmcs_write16(GUEST_LDTR_SELECTOR, 0); |
5931 |
++ vmcs_writel(GUEST_LDTR_BASE, 0); |
5932 |
++ vmcs_write32(GUEST_LDTR_LIMIT, 0xffff); |
5933 |
++ vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082); |
5934 |
++ |
5935 |
++ if (!init_event) { |
5936 |
++ vmcs_write32(GUEST_SYSENTER_CS, 0); |
5937 |
++ vmcs_writel(GUEST_SYSENTER_ESP, 0); |
5938 |
++ vmcs_writel(GUEST_SYSENTER_EIP, 0); |
5939 |
++ vmcs_write64(GUEST_IA32_DEBUGCTL, 0); |
5940 |
++ } |
5941 |
++ |
5942 |
++ kvm_set_rflags(vcpu, X86_EFLAGS_FIXED); |
5943 |
++ kvm_rip_write(vcpu, 0xfff0); |
5944 |
++ |
5945 |
++ vmcs_writel(GUEST_GDTR_BASE, 0); |
5946 |
++ vmcs_write32(GUEST_GDTR_LIMIT, 0xffff); |
5947 |
++ |
5948 |
++ vmcs_writel(GUEST_IDTR_BASE, 0); |
5949 |
++ vmcs_write32(GUEST_IDTR_LIMIT, 0xffff); |
5950 |
++ |
5951 |
++ vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE); |
5952 |
++ vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0); |
5953 |
++ vmcs_writel(GUEST_PENDING_DBG_EXCEPTIONS, 0); |
5954 |
++ if (kvm_mpx_supported()) |
5955 |
++ vmcs_write64(GUEST_BNDCFGS, 0); |
5956 |
++ |
5957 |
++ setup_msrs(vmx); |
5958 |
++ |
5959 |
++ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); /* 22.2.1 */ |
5960 |
++ |
5961 |
++ if (cpu_has_vmx_tpr_shadow() && !init_event) { |
5962 |
++ vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, 0); |
5963 |
++ if (cpu_need_tpr_shadow(vcpu)) |
5964 |
++ vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, |
5965 |
++ __pa(vcpu->arch.apic->regs)); |
5966 |
++ vmcs_write32(TPR_THRESHOLD, 0); |
5967 |
++ } |
5968 |
++ |
5969 |
++ kvm_make_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu); |
5970 |
++ |
5971 |
++ cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET; |
5972 |
++ vmx->vcpu.arch.cr0 = cr0; |
5973 |
++ vmx_set_cr0(vcpu, cr0); /* enter rmode */ |
5974 |
++ vmx_set_cr4(vcpu, 0); |
5975 |
++ vmx_set_efer(vcpu, 0); |
5976 |
++ |
5977 |
++ update_exception_bitmap(vcpu); |
5978 |
++ |
5979 |
++ vpid_sync_context(vmx->vpid); |
5980 |
++ if (init_event) |
5981 |
++ vmx_clear_hlt(vcpu); |
5982 |
++} |
5983 |
++ |
5984 |
++static void enable_irq_window(struct kvm_vcpu *vcpu) |
5985 |
++{ |
5986 |
++ exec_controls_setbit(to_vmx(vcpu), CPU_BASED_INTR_WINDOW_EXITING); |
5987 |
++} |
5988 |
++ |
5989 |
++static void enable_nmi_window(struct kvm_vcpu *vcpu) |
5990 |
++{ |
5991 |
++ if (!enable_vnmi || |
5992 |
++ vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) { |
5993 |
++ enable_irq_window(vcpu); |
5994 |
++ return; |
5995 |
++ } |
5996 |
++ |
5997 |
++ exec_controls_setbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING); |
5998 |
++} |
5999 |
++ |
6000 |
++static void vmx_inject_irq(struct kvm_vcpu *vcpu) |
6001 |
++{ |
6002 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6003 |
++ uint32_t intr; |
6004 |
++ int irq = vcpu->arch.interrupt.nr; |
6005 |
++ |
6006 |
++ trace_kvm_inj_virq(irq); |
6007 |
++ |
6008 |
++ ++vcpu->stat.irq_injections; |
6009 |
++ if (vmx->rmode.vm86_active) { |
6010 |
++ int inc_eip = 0; |
6011 |
++ if (vcpu->arch.interrupt.soft) |
6012 |
++ inc_eip = vcpu->arch.event_exit_inst_len; |
6013 |
++ kvm_inject_realmode_interrupt(vcpu, irq, inc_eip); |
6014 |
++ return; |
6015 |
++ } |
6016 |
++ intr = irq | INTR_INFO_VALID_MASK; |
6017 |
++ if (vcpu->arch.interrupt.soft) { |
6018 |
++ intr |= INTR_TYPE_SOFT_INTR; |
6019 |
++ vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, |
6020 |
++ vmx->vcpu.arch.event_exit_inst_len); |
6021 |
++ } else |
6022 |
++ intr |= INTR_TYPE_EXT_INTR; |
6023 |
++ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr); |
6024 |
++ |
6025 |
++ vmx_clear_hlt(vcpu); |
6026 |
++} |
6027 |
++ |
6028 |
++static void vmx_inject_nmi(struct kvm_vcpu *vcpu) |
6029 |
++{ |
6030 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6031 |
++ |
6032 |
++ if (!enable_vnmi) { |
6033 |
++ /* |
6034 |
++ * Tracking the NMI-blocked state in software is built upon |
6035 |
++ * finding the next open IRQ window. This, in turn, depends on |
6036 |
++ * well-behaving guests: They have to keep IRQs disabled at |
6037 |
++ * least as long as the NMI handler runs. Otherwise we may |
6038 |
++ * cause NMI nesting, maybe breaking the guest. But as this is |
6039 |
++ * highly unlikely, we can live with the residual risk. |
6040 |
++ */ |
6041 |
++ vmx->loaded_vmcs->soft_vnmi_blocked = 1; |
6042 |
++ vmx->loaded_vmcs->vnmi_blocked_time = 0; |
6043 |
++ } |
6044 |
++ |
6045 |
++ ++vcpu->stat.nmi_injections; |
6046 |
++ vmx->loaded_vmcs->nmi_known_unmasked = false; |
6047 |
++ |
6048 |
++ if (vmx->rmode.vm86_active) { |
6049 |
++ kvm_inject_realmode_interrupt(vcpu, NMI_VECTOR, 0); |
6050 |
++ return; |
6051 |
++ } |
6052 |
++ |
6053 |
++ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, |
6054 |
++ INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR); |
6055 |
++ |
6056 |
++ vmx_clear_hlt(vcpu); |
6057 |
++} |
6058 |
++ |
6059 |
++bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu) |
6060 |
++{ |
6061 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6062 |
++ bool masked; |
6063 |
++ |
6064 |
++ if (!enable_vnmi) |
6065 |
++ return vmx->loaded_vmcs->soft_vnmi_blocked; |
6066 |
++ if (vmx->loaded_vmcs->nmi_known_unmasked) |
6067 |
++ return false; |
6068 |
++ masked = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_NMI; |
6069 |
++ vmx->loaded_vmcs->nmi_known_unmasked = !masked; |
6070 |
++ return masked; |
6071 |
++} |
6072 |
++ |
6073 |
++void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked) |
6074 |
++{ |
6075 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6076 |
++ |
6077 |
++ if (!enable_vnmi) { |
6078 |
++ if (vmx->loaded_vmcs->soft_vnmi_blocked != masked) { |
6079 |
++ vmx->loaded_vmcs->soft_vnmi_blocked = masked; |
6080 |
++ vmx->loaded_vmcs->vnmi_blocked_time = 0; |
6081 |
++ } |
6082 |
++ } else { |
6083 |
++ vmx->loaded_vmcs->nmi_known_unmasked = !masked; |
6084 |
++ if (masked) |
6085 |
++ vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
6086 |
++ GUEST_INTR_STATE_NMI); |
6087 |
++ else |
6088 |
++ vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO, |
6089 |
++ GUEST_INTR_STATE_NMI); |
6090 |
++ } |
6091 |
++} |
6092 |
++ |
6093 |
++static int vmx_nmi_allowed(struct kvm_vcpu *vcpu) |
6094 |
++{ |
6095 |
++ if (to_vmx(vcpu)->nested.nested_run_pending) |
6096 |
++ return 0; |
6097 |
++ |
6098 |
++ if (!enable_vnmi && |
6099 |
++ to_vmx(vcpu)->loaded_vmcs->soft_vnmi_blocked) |
6100 |
++ return 0; |
6101 |
++ |
6102 |
++ return !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
6103 |
++ (GUEST_INTR_STATE_MOV_SS | GUEST_INTR_STATE_STI |
6104 |
++ | GUEST_INTR_STATE_NMI)); |
6105 |
++} |
6106 |
++ |
6107 |
++static int vmx_interrupt_allowed(struct kvm_vcpu *vcpu) |
6108 |
++{ |
6109 |
++ return (!to_vmx(vcpu)->nested.nested_run_pending && |
6110 |
++ vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF) && |
6111 |
++ !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
6112 |
++ (GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS)); |
6113 |
++} |
6114 |
++ |
6115 |
++static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr) |
6116 |
++{ |
6117 |
++ int ret; |
6118 |
++ |
6119 |
++ if (enable_unrestricted_guest) |
6120 |
++ return 0; |
6121 |
++ |
6122 |
++ ret = x86_set_memory_region(kvm, TSS_PRIVATE_MEMSLOT, addr, |
6123 |
++ PAGE_SIZE * 3); |
6124 |
++ if (ret) |
6125 |
++ return ret; |
6126 |
++ to_kvm_vmx(kvm)->tss_addr = addr; |
6127 |
++ return init_rmode_tss(kvm); |
6128 |
++} |
6129 |
++ |
6130 |
++static int vmx_set_identity_map_addr(struct kvm *kvm, u64 ident_addr) |
6131 |
++{ |
6132 |
++ to_kvm_vmx(kvm)->ept_identity_map_addr = ident_addr; |
6133 |
++ return 0; |
6134 |
++} |
6135 |
++ |
6136 |
++static bool rmode_exception(struct kvm_vcpu *vcpu, int vec) |
6137 |
++{ |
6138 |
++ switch (vec) { |
6139 |
++ case BP_VECTOR: |
6140 |
++ /* |
6141 |
++ * Update instruction length as we may reinject the exception |
6142 |
++ * from user space while in guest debugging mode. |
6143 |
++ */ |
6144 |
++ to_vmx(vcpu)->vcpu.arch.event_exit_inst_len = |
6145 |
++ vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
6146 |
++ if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) |
6147 |
++ return false; |
6148 |
++ /* fall through */ |
6149 |
++ case DB_VECTOR: |
6150 |
++ if (vcpu->guest_debug & |
6151 |
++ (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP)) |
6152 |
++ return false; |
6153 |
++ /* fall through */ |
6154 |
++ case DE_VECTOR: |
6155 |
++ case OF_VECTOR: |
6156 |
++ case BR_VECTOR: |
6157 |
++ case UD_VECTOR: |
6158 |
++ case DF_VECTOR: |
6159 |
++ case SS_VECTOR: |
6160 |
++ case GP_VECTOR: |
6161 |
++ case MF_VECTOR: |
6162 |
++ return true; |
6163 |
++ break; |
6164 |
++ } |
6165 |
++ return false; |
6166 |
++} |
6167 |
++ |
6168 |
++static int handle_rmode_exception(struct kvm_vcpu *vcpu, |
6169 |
++ int vec, u32 err_code) |
6170 |
++{ |
6171 |
++ /* |
6172 |
++ * Instruction with address size override prefix opcode 0x67 |
6173 |
++ * Cause the #SS fault with 0 error code in VM86 mode. |
6174 |
++ */ |
6175 |
++ if (((vec == GP_VECTOR) || (vec == SS_VECTOR)) && err_code == 0) { |
6176 |
++ if (kvm_emulate_instruction(vcpu, 0)) { |
6177 |
++ if (vcpu->arch.halt_request) { |
6178 |
++ vcpu->arch.halt_request = 0; |
6179 |
++ return kvm_vcpu_halt(vcpu); |
6180 |
++ } |
6181 |
++ return 1; |
6182 |
++ } |
6183 |
++ return 0; |
6184 |
++ } |
6185 |
++ |
6186 |
++ /* |
6187 |
++ * Forward all other exceptions that are valid in real mode. |
6188 |
++ * FIXME: Breaks guest debugging in real mode, needs to be fixed with |
6189 |
++ * the required debugging infrastructure rework. |
6190 |
++ */ |
6191 |
++ kvm_queue_exception(vcpu, vec); |
6192 |
++ return 1; |
6193 |
++} |
6194 |
++ |
6195 |
++/* |
6196 |
++ * Trigger machine check on the host. We assume all the MSRs are already set up |
6197 |
++ * by the CPU and that we still run on the same CPU as the MCE occurred on. |
6198 |
++ * We pass a fake environment to the machine check handler because we want |
6199 |
++ * the guest to be always treated like user space, no matter what context |
6200 |
++ * it used internally. |
6201 |
++ */ |
6202 |
++static void kvm_machine_check(void) |
6203 |
++{ |
6204 |
++#if defined(CONFIG_X86_MCE) && defined(CONFIG_X86_64) |
6205 |
++ struct pt_regs regs = { |
6206 |
++ .cs = 3, /* Fake ring 3 no matter what the guest ran on */ |
6207 |
++ .flags = X86_EFLAGS_IF, |
6208 |
++ }; |
6209 |
++ |
6210 |
++ do_machine_check(®s, 0); |
6211 |
++#endif |
6212 |
++} |
6213 |
++ |
6214 |
++static int handle_machine_check(struct kvm_vcpu *vcpu) |
6215 |
++{ |
6216 |
++ /* handled by vmx_vcpu_run() */ |
6217 |
++ return 1; |
6218 |
++} |
6219 |
++ |
6220 |
++static int handle_exception_nmi(struct kvm_vcpu *vcpu) |
6221 |
++{ |
6222 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6223 |
++ struct kvm_run *kvm_run = vcpu->run; |
6224 |
++ u32 intr_info, ex_no, error_code; |
6225 |
++ unsigned long cr2, rip, dr6; |
6226 |
++ u32 vect_info; |
6227 |
++ |
6228 |
++ vect_info = vmx->idt_vectoring_info; |
6229 |
++ intr_info = vmx->exit_intr_info; |
6230 |
++ |
6231 |
++ if (is_machine_check(intr_info) || is_nmi(intr_info)) |
6232 |
++ return 1; /* handled by handle_exception_nmi_irqoff() */ |
6233 |
++ |
6234 |
++ if (is_invalid_opcode(intr_info)) |
6235 |
++ return handle_ud(vcpu); |
6236 |
++ |
6237 |
++ error_code = 0; |
6238 |
++ if (intr_info & INTR_INFO_DELIVER_CODE_MASK) |
6239 |
++ error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE); |
6240 |
++ |
6241 |
++ if (!vmx->rmode.vm86_active && is_gp_fault(intr_info)) { |
6242 |
++ WARN_ON_ONCE(!enable_vmware_backdoor); |
6243 |
++ |
6244 |
++ /* |
6245 |
++ * VMware backdoor emulation on #GP interception only handles |
6246 |
++ * IN{S}, OUT{S}, and RDPMC, none of which generate a non-zero |
6247 |
++ * error code on #GP. |
6248 |
++ */ |
6249 |
++ if (error_code) { |
6250 |
++ kvm_queue_exception_e(vcpu, GP_VECTOR, error_code); |
6251 |
++ return 1; |
6252 |
++ } |
6253 |
++ return kvm_emulate_instruction(vcpu, EMULTYPE_VMWARE_GP); |
6254 |
++ } |
6255 |
++ |
6256 |
++ /* |
6257 |
++ * The #PF with PFEC.RSVD = 1 indicates the guest is accessing |
6258 |
++ * MMIO, it is better to report an internal error. |
6259 |
++ * See the comments in vmx_handle_exit. |
6260 |
++ */ |
6261 |
++ if ((vect_info & VECTORING_INFO_VALID_MASK) && |
6262 |
++ !(is_page_fault(intr_info) && !(error_code & PFERR_RSVD_MASK))) { |
6263 |
++ vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
6264 |
++ vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_SIMUL_EX; |
6265 |
++ vcpu->run->internal.ndata = 3; |
6266 |
++ vcpu->run->internal.data[0] = vect_info; |
6267 |
++ vcpu->run->internal.data[1] = intr_info; |
6268 |
++ vcpu->run->internal.data[2] = error_code; |
6269 |
++ return 0; |
6270 |
++ } |
6271 |
++ |
6272 |
++ if (is_page_fault(intr_info)) { |
6273 |
++ cr2 = vmcs_readl(EXIT_QUALIFICATION); |
6274 |
++ /* EPT won't cause page fault directly */ |
6275 |
++ WARN_ON_ONCE(!vcpu->arch.apf.host_apf_reason && enable_ept); |
6276 |
++ return kvm_handle_page_fault(vcpu, error_code, cr2, NULL, 0); |
6277 |
++ } |
6278 |
++ |
6279 |
++ ex_no = intr_info & INTR_INFO_VECTOR_MASK; |
6280 |
++ |
6281 |
++ if (vmx->rmode.vm86_active && rmode_exception(vcpu, ex_no)) |
6282 |
++ return handle_rmode_exception(vcpu, ex_no, error_code); |
6283 |
++ |
6284 |
++ switch (ex_no) { |
6285 |
++ case AC_VECTOR: |
6286 |
++ kvm_queue_exception_e(vcpu, AC_VECTOR, error_code); |
6287 |
++ return 1; |
6288 |
++ case DB_VECTOR: |
6289 |
++ dr6 = vmcs_readl(EXIT_QUALIFICATION); |
6290 |
++ if (!(vcpu->guest_debug & |
6291 |
++ (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP))) { |
6292 |
++ vcpu->arch.dr6 &= ~DR_TRAP_BITS; |
6293 |
++ vcpu->arch.dr6 |= dr6 | DR6_RTM; |
6294 |
++ if (is_icebp(intr_info)) |
6295 |
++ WARN_ON(!skip_emulated_instruction(vcpu)); |
6296 |
++ |
6297 |
++ kvm_queue_exception(vcpu, DB_VECTOR); |
6298 |
++ return 1; |
6299 |
++ } |
6300 |
++ kvm_run->debug.arch.dr6 = dr6 | DR6_FIXED_1; |
6301 |
++ kvm_run->debug.arch.dr7 = vmcs_readl(GUEST_DR7); |
6302 |
++ /* fall through */ |
6303 |
++ case BP_VECTOR: |
6304 |
++ /* |
6305 |
++ * Update instruction length as we may reinject #BP from |
6306 |
++ * user space while in guest debugging mode. Reading it for |
6307 |
++ * #DB as well causes no harm, it is not used in that case. |
6308 |
++ */ |
6309 |
++ vmx->vcpu.arch.event_exit_inst_len = |
6310 |
++ vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
6311 |
++ kvm_run->exit_reason = KVM_EXIT_DEBUG; |
6312 |
++ rip = kvm_rip_read(vcpu); |
6313 |
++ kvm_run->debug.arch.pc = vmcs_readl(GUEST_CS_BASE) + rip; |
6314 |
++ kvm_run->debug.arch.exception = ex_no; |
6315 |
++ break; |
6316 |
++ default: |
6317 |
++ kvm_run->exit_reason = KVM_EXIT_EXCEPTION; |
6318 |
++ kvm_run->ex.exception = ex_no; |
6319 |
++ kvm_run->ex.error_code = error_code; |
6320 |
++ break; |
6321 |
++ } |
6322 |
++ return 0; |
6323 |
++} |
6324 |
++ |
6325 |
++static __always_inline int handle_external_interrupt(struct kvm_vcpu *vcpu) |
6326 |
++{ |
6327 |
++ ++vcpu->stat.irq_exits; |
6328 |
++ return 1; |
6329 |
++} |
6330 |
++ |
6331 |
++static int handle_triple_fault(struct kvm_vcpu *vcpu) |
6332 |
++{ |
6333 |
++ vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; |
6334 |
++ vcpu->mmio_needed = 0; |
6335 |
++ return 0; |
6336 |
++} |
6337 |
++ |
6338 |
++static int handle_io(struct kvm_vcpu *vcpu) |
6339 |
++{ |
6340 |
++ unsigned long exit_qualification; |
6341 |
++ int size, in, string; |
6342 |
++ unsigned port; |
6343 |
++ |
6344 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6345 |
++ string = (exit_qualification & 16) != 0; |
6346 |
++ |
6347 |
++ ++vcpu->stat.io_exits; |
6348 |
++ |
6349 |
++ if (string) |
6350 |
++ return kvm_emulate_instruction(vcpu, 0); |
6351 |
++ |
6352 |
++ port = exit_qualification >> 16; |
6353 |
++ size = (exit_qualification & 7) + 1; |
6354 |
++ in = (exit_qualification & 8) != 0; |
6355 |
++ |
6356 |
++ return kvm_fast_pio(vcpu, size, port, in); |
6357 |
++} |
6358 |
++ |
6359 |
++static void |
6360 |
++vmx_patch_hypercall(struct kvm_vcpu *vcpu, unsigned char *hypercall) |
6361 |
++{ |
6362 |
++ /* |
6363 |
++ * Patch in the VMCALL instruction: |
6364 |
++ */ |
6365 |
++ hypercall[0] = 0x0f; |
6366 |
++ hypercall[1] = 0x01; |
6367 |
++ hypercall[2] = 0xc1; |
6368 |
++} |
6369 |
++ |
6370 |
++/* called to set cr0 as appropriate for a mov-to-cr0 exit. */ |
6371 |
++static int handle_set_cr0(struct kvm_vcpu *vcpu, unsigned long val) |
6372 |
++{ |
6373 |
++ if (is_guest_mode(vcpu)) { |
6374 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
6375 |
++ unsigned long orig_val = val; |
6376 |
++ |
6377 |
++ /* |
6378 |
++ * We get here when L2 changed cr0 in a way that did not change |
6379 |
++ * any of L1's shadowed bits (see nested_vmx_exit_handled_cr), |
6380 |
++ * but did change L0 shadowed bits. So we first calculate the |
6381 |
++ * effective cr0 value that L1 would like to write into the |
6382 |
++ * hardware. It consists of the L2-owned bits from the new |
6383 |
++ * value combined with the L1-owned bits from L1's guest_cr0. |
6384 |
++ */ |
6385 |
++ val = (val & ~vmcs12->cr0_guest_host_mask) | |
6386 |
++ (vmcs12->guest_cr0 & vmcs12->cr0_guest_host_mask); |
6387 |
++ |
6388 |
++ if (!nested_guest_cr0_valid(vcpu, val)) |
6389 |
++ return 1; |
6390 |
++ |
6391 |
++ if (kvm_set_cr0(vcpu, val)) |
6392 |
++ return 1; |
6393 |
++ vmcs_writel(CR0_READ_SHADOW, orig_val); |
6394 |
++ return 0; |
6395 |
++ } else { |
6396 |
++ if (to_vmx(vcpu)->nested.vmxon && |
6397 |
++ !nested_host_cr0_valid(vcpu, val)) |
6398 |
++ return 1; |
6399 |
++ |
6400 |
++ return kvm_set_cr0(vcpu, val); |
6401 |
++ } |
6402 |
++} |
6403 |
++ |
6404 |
++static int handle_set_cr4(struct kvm_vcpu *vcpu, unsigned long val) |
6405 |
++{ |
6406 |
++ if (is_guest_mode(vcpu)) { |
6407 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
6408 |
++ unsigned long orig_val = val; |
6409 |
++ |
6410 |
++ /* analogously to handle_set_cr0 */ |
6411 |
++ val = (val & ~vmcs12->cr4_guest_host_mask) | |
6412 |
++ (vmcs12->guest_cr4 & vmcs12->cr4_guest_host_mask); |
6413 |
++ if (kvm_set_cr4(vcpu, val)) |
6414 |
++ return 1; |
6415 |
++ vmcs_writel(CR4_READ_SHADOW, orig_val); |
6416 |
++ return 0; |
6417 |
++ } else |
6418 |
++ return kvm_set_cr4(vcpu, val); |
6419 |
++} |
6420 |
++ |
6421 |
++static int handle_desc(struct kvm_vcpu *vcpu) |
6422 |
++{ |
6423 |
++ WARN_ON(!(vcpu->arch.cr4 & X86_CR4_UMIP)); |
6424 |
++ return kvm_emulate_instruction(vcpu, 0); |
6425 |
++} |
6426 |
++ |
6427 |
++static int handle_cr(struct kvm_vcpu *vcpu) |
6428 |
++{ |
6429 |
++ unsigned long exit_qualification, val; |
6430 |
++ int cr; |
6431 |
++ int reg; |
6432 |
++ int err; |
6433 |
++ int ret; |
6434 |
++ |
6435 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6436 |
++ cr = exit_qualification & 15; |
6437 |
++ reg = (exit_qualification >> 8) & 15; |
6438 |
++ switch ((exit_qualification >> 4) & 3) { |
6439 |
++ case 0: /* mov to cr */ |
6440 |
++ val = kvm_register_readl(vcpu, reg); |
6441 |
++ trace_kvm_cr_write(cr, val); |
6442 |
++ switch (cr) { |
6443 |
++ case 0: |
6444 |
++ err = handle_set_cr0(vcpu, val); |
6445 |
++ return kvm_complete_insn_gp(vcpu, err); |
6446 |
++ case 3: |
6447 |
++ WARN_ON_ONCE(enable_unrestricted_guest); |
6448 |
++ err = kvm_set_cr3(vcpu, val); |
6449 |
++ return kvm_complete_insn_gp(vcpu, err); |
6450 |
++ case 4: |
6451 |
++ err = handle_set_cr4(vcpu, val); |
6452 |
++ return kvm_complete_insn_gp(vcpu, err); |
6453 |
++ case 8: { |
6454 |
++ u8 cr8_prev = kvm_get_cr8(vcpu); |
6455 |
++ u8 cr8 = (u8)val; |
6456 |
++ err = kvm_set_cr8(vcpu, cr8); |
6457 |
++ ret = kvm_complete_insn_gp(vcpu, err); |
6458 |
++ if (lapic_in_kernel(vcpu)) |
6459 |
++ return ret; |
6460 |
++ if (cr8_prev <= cr8) |
6461 |
++ return ret; |
6462 |
++ /* |
6463 |
++ * TODO: we might be squashing a |
6464 |
++ * KVM_GUESTDBG_SINGLESTEP-triggered |
6465 |
++ * KVM_EXIT_DEBUG here. |
6466 |
++ */ |
6467 |
++ vcpu->run->exit_reason = KVM_EXIT_SET_TPR; |
6468 |
++ return 0; |
6469 |
++ } |
6470 |
++ } |
6471 |
++ break; |
6472 |
++ case 2: /* clts */ |
6473 |
++ WARN_ONCE(1, "Guest should always own CR0.TS"); |
6474 |
++ vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); |
6475 |
++ trace_kvm_cr_write(0, kvm_read_cr0(vcpu)); |
6476 |
++ return kvm_skip_emulated_instruction(vcpu); |
6477 |
++ case 1: /*mov from cr*/ |
6478 |
++ switch (cr) { |
6479 |
++ case 3: |
6480 |
++ WARN_ON_ONCE(enable_unrestricted_guest); |
6481 |
++ val = kvm_read_cr3(vcpu); |
6482 |
++ kvm_register_write(vcpu, reg, val); |
6483 |
++ trace_kvm_cr_read(cr, val); |
6484 |
++ return kvm_skip_emulated_instruction(vcpu); |
6485 |
++ case 8: |
6486 |
++ val = kvm_get_cr8(vcpu); |
6487 |
++ kvm_register_write(vcpu, reg, val); |
6488 |
++ trace_kvm_cr_read(cr, val); |
6489 |
++ return kvm_skip_emulated_instruction(vcpu); |
6490 |
++ } |
6491 |
++ break; |
6492 |
++ case 3: /* lmsw */ |
6493 |
++ val = (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f; |
6494 |
++ trace_kvm_cr_write(0, (kvm_read_cr0(vcpu) & ~0xful) | val); |
6495 |
++ kvm_lmsw(vcpu, val); |
6496 |
++ |
6497 |
++ return kvm_skip_emulated_instruction(vcpu); |
6498 |
++ default: |
6499 |
++ break; |
6500 |
++ } |
6501 |
++ vcpu->run->exit_reason = 0; |
6502 |
++ vcpu_unimpl(vcpu, "unhandled control register: op %d cr %d\n", |
6503 |
++ (int)(exit_qualification >> 4) & 3, cr); |
6504 |
++ return 0; |
6505 |
++} |
6506 |
++ |
6507 |
++static int handle_dr(struct kvm_vcpu *vcpu) |
6508 |
++{ |
6509 |
++ unsigned long exit_qualification; |
6510 |
++ int dr, dr7, reg; |
6511 |
++ |
6512 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6513 |
++ dr = exit_qualification & DEBUG_REG_ACCESS_NUM; |
6514 |
++ |
6515 |
++ /* First, if DR does not exist, trigger UD */ |
6516 |
++ if (!kvm_require_dr(vcpu, dr)) |
6517 |
++ return 1; |
6518 |
++ |
6519 |
++ /* Do not handle if the CPL > 0, will trigger GP on re-entry */ |
6520 |
++ if (!kvm_require_cpl(vcpu, 0)) |
6521 |
++ return 1; |
6522 |
++ dr7 = vmcs_readl(GUEST_DR7); |
6523 |
++ if (dr7 & DR7_GD) { |
6524 |
++ /* |
6525 |
++ * As the vm-exit takes precedence over the debug trap, we |
6526 |
++ * need to emulate the latter, either for the host or the |
6527 |
++ * guest debugging itself. |
6528 |
++ */ |
6529 |
++ if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { |
6530 |
++ vcpu->run->debug.arch.dr6 = vcpu->arch.dr6; |
6531 |
++ vcpu->run->debug.arch.dr7 = dr7; |
6532 |
++ vcpu->run->debug.arch.pc = kvm_get_linear_rip(vcpu); |
6533 |
++ vcpu->run->debug.arch.exception = DB_VECTOR; |
6534 |
++ vcpu->run->exit_reason = KVM_EXIT_DEBUG; |
6535 |
++ return 0; |
6536 |
++ } else { |
6537 |
++ vcpu->arch.dr6 &= ~DR_TRAP_BITS; |
6538 |
++ vcpu->arch.dr6 |= DR6_BD | DR6_RTM; |
6539 |
++ kvm_queue_exception(vcpu, DB_VECTOR); |
6540 |
++ return 1; |
6541 |
++ } |
6542 |
++ } |
6543 |
++ |
6544 |
++ if (vcpu->guest_debug == 0) { |
6545 |
++ exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_MOV_DR_EXITING); |
6546 |
++ |
6547 |
++ /* |
6548 |
++ * No more DR vmexits; force a reload of the debug registers |
6549 |
++ * and reenter on this instruction. The next vmexit will |
6550 |
++ * retrieve the full state of the debug registers. |
6551 |
++ */ |
6552 |
++ vcpu->arch.switch_db_regs |= KVM_DEBUGREG_WONT_EXIT; |
6553 |
++ return 1; |
6554 |
++ } |
6555 |
++ |
6556 |
++ reg = DEBUG_REG_ACCESS_REG(exit_qualification); |
6557 |
++ if (exit_qualification & TYPE_MOV_FROM_DR) { |
6558 |
++ unsigned long val; |
6559 |
++ |
6560 |
++ if (kvm_get_dr(vcpu, dr, &val)) |
6561 |
++ return 1; |
6562 |
++ kvm_register_write(vcpu, reg, val); |
6563 |
++ } else |
6564 |
++ if (kvm_set_dr(vcpu, dr, kvm_register_readl(vcpu, reg))) |
6565 |
++ return 1; |
6566 |
++ |
6567 |
++ return kvm_skip_emulated_instruction(vcpu); |
6568 |
++} |
6569 |
++ |
6570 |
++static u64 vmx_get_dr6(struct kvm_vcpu *vcpu) |
6571 |
++{ |
6572 |
++ return vcpu->arch.dr6; |
6573 |
++} |
6574 |
++ |
6575 |
++static void vmx_set_dr6(struct kvm_vcpu *vcpu, unsigned long val) |
6576 |
++{ |
6577 |
++} |
6578 |
++ |
6579 |
++static void vmx_sync_dirty_debug_regs(struct kvm_vcpu *vcpu) |
6580 |
++{ |
6581 |
++ get_debugreg(vcpu->arch.db[0], 0); |
6582 |
++ get_debugreg(vcpu->arch.db[1], 1); |
6583 |
++ get_debugreg(vcpu->arch.db[2], 2); |
6584 |
++ get_debugreg(vcpu->arch.db[3], 3); |
6585 |
++ get_debugreg(vcpu->arch.dr6, 6); |
6586 |
++ vcpu->arch.dr7 = vmcs_readl(GUEST_DR7); |
6587 |
++ |
6588 |
++ vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_WONT_EXIT; |
6589 |
++ exec_controls_setbit(to_vmx(vcpu), CPU_BASED_MOV_DR_EXITING); |
6590 |
++} |
6591 |
++ |
6592 |
++static void vmx_set_dr7(struct kvm_vcpu *vcpu, unsigned long val) |
6593 |
++{ |
6594 |
++ vmcs_writel(GUEST_DR7, val); |
6595 |
++} |
6596 |
++ |
6597 |
++static int handle_tpr_below_threshold(struct kvm_vcpu *vcpu) |
6598 |
++{ |
6599 |
++ kvm_apic_update_ppr(vcpu); |
6600 |
++ return 1; |
6601 |
++} |
6602 |
++ |
6603 |
++static int handle_interrupt_window(struct kvm_vcpu *vcpu) |
6604 |
++{ |
6605 |
++ exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_INTR_WINDOW_EXITING); |
6606 |
++ |
6607 |
++ kvm_make_request(KVM_REQ_EVENT, vcpu); |
6608 |
++ |
6609 |
++ ++vcpu->stat.irq_window_exits; |
6610 |
++ return 1; |
6611 |
++} |
6612 |
++ |
6613 |
++static int handle_vmcall(struct kvm_vcpu *vcpu) |
6614 |
++{ |
6615 |
++ return kvm_emulate_hypercall(vcpu); |
6616 |
++} |
6617 |
++ |
6618 |
++static int handle_invd(struct kvm_vcpu *vcpu) |
6619 |
++{ |
6620 |
++ return kvm_emulate_instruction(vcpu, 0); |
6621 |
++} |
6622 |
++ |
6623 |
++static int handle_invlpg(struct kvm_vcpu *vcpu) |
6624 |
++{ |
6625 |
++ unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6626 |
++ |
6627 |
++ kvm_mmu_invlpg(vcpu, exit_qualification); |
6628 |
++ return kvm_skip_emulated_instruction(vcpu); |
6629 |
++} |
6630 |
++ |
6631 |
++static int handle_rdpmc(struct kvm_vcpu *vcpu) |
6632 |
++{ |
6633 |
++ int err; |
6634 |
++ |
6635 |
++ err = kvm_rdpmc(vcpu); |
6636 |
++ return kvm_complete_insn_gp(vcpu, err); |
6637 |
++} |
6638 |
++ |
6639 |
++static int handle_wbinvd(struct kvm_vcpu *vcpu) |
6640 |
++{ |
6641 |
++ return kvm_emulate_wbinvd(vcpu); |
6642 |
++} |
6643 |
++ |
6644 |
++static int handle_xsetbv(struct kvm_vcpu *vcpu) |
6645 |
++{ |
6646 |
++ u64 new_bv = kvm_read_edx_eax(vcpu); |
6647 |
++ u32 index = kvm_rcx_read(vcpu); |
6648 |
++ |
6649 |
++ if (kvm_set_xcr(vcpu, index, new_bv) == 0) |
6650 |
++ return kvm_skip_emulated_instruction(vcpu); |
6651 |
++ return 1; |
6652 |
++} |
6653 |
++ |
6654 |
++static int handle_apic_access(struct kvm_vcpu *vcpu) |
6655 |
++{ |
6656 |
++ if (likely(fasteoi)) { |
6657 |
++ unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6658 |
++ int access_type, offset; |
6659 |
++ |
6660 |
++ access_type = exit_qualification & APIC_ACCESS_TYPE; |
6661 |
++ offset = exit_qualification & APIC_ACCESS_OFFSET; |
6662 |
++ /* |
6663 |
++ * Sane guest uses MOV to write EOI, with written value |
6664 |
++ * not cared. So make a short-circuit here by avoiding |
6665 |
++ * heavy instruction emulation. |
6666 |
++ */ |
6667 |
++ if ((access_type == TYPE_LINEAR_APIC_INST_WRITE) && |
6668 |
++ (offset == APIC_EOI)) { |
6669 |
++ kvm_lapic_set_eoi(vcpu); |
6670 |
++ return kvm_skip_emulated_instruction(vcpu); |
6671 |
++ } |
6672 |
++ } |
6673 |
++ return kvm_emulate_instruction(vcpu, 0); |
6674 |
++} |
6675 |
++ |
6676 |
++static int handle_apic_eoi_induced(struct kvm_vcpu *vcpu) |
6677 |
++{ |
6678 |
++ unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6679 |
++ int vector = exit_qualification & 0xff; |
6680 |
++ |
6681 |
++ /* EOI-induced VM exit is trap-like and thus no need to adjust IP */ |
6682 |
++ kvm_apic_set_eoi_accelerated(vcpu, vector); |
6683 |
++ return 1; |
6684 |
++} |
6685 |
++ |
6686 |
++static int handle_apic_write(struct kvm_vcpu *vcpu) |
6687 |
++{ |
6688 |
++ unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6689 |
++ u32 offset = exit_qualification & 0xfff; |
6690 |
++ |
6691 |
++ /* APIC-write VM exit is trap-like and thus no need to adjust IP */ |
6692 |
++ kvm_apic_write_nodecode(vcpu, offset); |
6693 |
++ return 1; |
6694 |
++} |
6695 |
++ |
6696 |
++static int handle_task_switch(struct kvm_vcpu *vcpu) |
6697 |
++{ |
6698 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6699 |
++ unsigned long exit_qualification; |
6700 |
++ bool has_error_code = false; |
6701 |
++ u32 error_code = 0; |
6702 |
++ u16 tss_selector; |
6703 |
++ int reason, type, idt_v, idt_index; |
6704 |
++ |
6705 |
++ idt_v = (vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK); |
6706 |
++ idt_index = (vmx->idt_vectoring_info & VECTORING_INFO_VECTOR_MASK); |
6707 |
++ type = (vmx->idt_vectoring_info & VECTORING_INFO_TYPE_MASK); |
6708 |
++ |
6709 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6710 |
++ |
6711 |
++ reason = (u32)exit_qualification >> 30; |
6712 |
++ if (reason == TASK_SWITCH_GATE && idt_v) { |
6713 |
++ switch (type) { |
6714 |
++ case INTR_TYPE_NMI_INTR: |
6715 |
++ vcpu->arch.nmi_injected = false; |
6716 |
++ vmx_set_nmi_mask(vcpu, true); |
6717 |
++ break; |
6718 |
++ case INTR_TYPE_EXT_INTR: |
6719 |
++ case INTR_TYPE_SOFT_INTR: |
6720 |
++ kvm_clear_interrupt_queue(vcpu); |
6721 |
++ break; |
6722 |
++ case INTR_TYPE_HARD_EXCEPTION: |
6723 |
++ if (vmx->idt_vectoring_info & |
6724 |
++ VECTORING_INFO_DELIVER_CODE_MASK) { |
6725 |
++ has_error_code = true; |
6726 |
++ error_code = |
6727 |
++ vmcs_read32(IDT_VECTORING_ERROR_CODE); |
6728 |
++ } |
6729 |
++ /* fall through */ |
6730 |
++ case INTR_TYPE_SOFT_EXCEPTION: |
6731 |
++ kvm_clear_exception_queue(vcpu); |
6732 |
++ break; |
6733 |
++ default: |
6734 |
++ break; |
6735 |
++ } |
6736 |
++ } |
6737 |
++ tss_selector = exit_qualification; |
6738 |
++ |
6739 |
++ if (!idt_v || (type != INTR_TYPE_HARD_EXCEPTION && |
6740 |
++ type != INTR_TYPE_EXT_INTR && |
6741 |
++ type != INTR_TYPE_NMI_INTR)) |
6742 |
++ WARN_ON(!skip_emulated_instruction(vcpu)); |
6743 |
++ |
6744 |
++ /* |
6745 |
++ * TODO: What about debug traps on tss switch? |
6746 |
++ * Are we supposed to inject them and update dr6? |
6747 |
++ */ |
6748 |
++ return kvm_task_switch(vcpu, tss_selector, |
6749 |
++ type == INTR_TYPE_SOFT_INTR ? idt_index : -1, |
6750 |
++ reason, has_error_code, error_code); |
6751 |
++} |
6752 |
++ |
6753 |
++static int handle_ept_violation(struct kvm_vcpu *vcpu) |
6754 |
++{ |
6755 |
++ unsigned long exit_qualification; |
6756 |
++ gpa_t gpa; |
6757 |
++ u64 error_code; |
6758 |
++ |
6759 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6760 |
++ |
6761 |
++ /* |
6762 |
++ * EPT violation happened while executing iret from NMI, |
6763 |
++ * "blocked by NMI" bit has to be set before next VM entry. |
6764 |
++ * There are errata that may cause this bit to not be set: |
6765 |
++ * AAK134, BY25. |
6766 |
++ */ |
6767 |
++ if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) && |
6768 |
++ enable_vnmi && |
6769 |
++ (exit_qualification & INTR_INFO_UNBLOCK_NMI)) |
6770 |
++ vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, GUEST_INTR_STATE_NMI); |
6771 |
++ |
6772 |
++ gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
6773 |
++ trace_kvm_page_fault(gpa, exit_qualification); |
6774 |
++ |
6775 |
++ /* Is it a read fault? */ |
6776 |
++ error_code = (exit_qualification & EPT_VIOLATION_ACC_READ) |
6777 |
++ ? PFERR_USER_MASK : 0; |
6778 |
++ /* Is it a write fault? */ |
6779 |
++ error_code |= (exit_qualification & EPT_VIOLATION_ACC_WRITE) |
6780 |
++ ? PFERR_WRITE_MASK : 0; |
6781 |
++ /* Is it a fetch fault? */ |
6782 |
++ error_code |= (exit_qualification & EPT_VIOLATION_ACC_INSTR) |
6783 |
++ ? PFERR_FETCH_MASK : 0; |
6784 |
++ /* ept page table entry is present? */ |
6785 |
++ error_code |= (exit_qualification & |
6786 |
++ (EPT_VIOLATION_READABLE | EPT_VIOLATION_WRITABLE | |
6787 |
++ EPT_VIOLATION_EXECUTABLE)) |
6788 |
++ ? PFERR_PRESENT_MASK : 0; |
6789 |
++ |
6790 |
++ error_code |= (exit_qualification & 0x100) != 0 ? |
6791 |
++ PFERR_GUEST_FINAL_MASK : PFERR_GUEST_PAGE_MASK; |
6792 |
++ |
6793 |
++ vcpu->arch.exit_qualification = exit_qualification; |
6794 |
++ return kvm_mmu_page_fault(vcpu, gpa, error_code, NULL, 0); |
6795 |
++} |
6796 |
++ |
6797 |
++static int handle_ept_misconfig(struct kvm_vcpu *vcpu) |
6798 |
++{ |
6799 |
++ gpa_t gpa; |
6800 |
++ |
6801 |
++ /* |
6802 |
++ * A nested guest cannot optimize MMIO vmexits, because we have an |
6803 |
++ * nGPA here instead of the required GPA. |
6804 |
++ */ |
6805 |
++ gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
6806 |
++ if (!is_guest_mode(vcpu) && |
6807 |
++ !kvm_io_bus_write(vcpu, KVM_FAST_MMIO_BUS, gpa, 0, NULL)) { |
6808 |
++ trace_kvm_fast_mmio(gpa); |
6809 |
++ return kvm_skip_emulated_instruction(vcpu); |
6810 |
++ } |
6811 |
++ |
6812 |
++ return kvm_mmu_page_fault(vcpu, gpa, PFERR_RSVD_MASK, NULL, 0); |
6813 |
++} |
6814 |
++ |
6815 |
++static int handle_nmi_window(struct kvm_vcpu *vcpu) |
6816 |
++{ |
6817 |
++ WARN_ON_ONCE(!enable_vnmi); |
6818 |
++ exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING); |
6819 |
++ ++vcpu->stat.nmi_window_exits; |
6820 |
++ kvm_make_request(KVM_REQ_EVENT, vcpu); |
6821 |
++ |
6822 |
++ return 1; |
6823 |
++} |
6824 |
++ |
6825 |
++static int handle_invalid_guest_state(struct kvm_vcpu *vcpu) |
6826 |
++{ |
6827 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6828 |
++ bool intr_window_requested; |
6829 |
++ unsigned count = 130; |
6830 |
++ |
6831 |
++ /* |
6832 |
++ * We should never reach the point where we are emulating L2 |
6833 |
++ * due to invalid guest state as that means we incorrectly |
6834 |
++ * allowed a nested VMEntry with an invalid vmcs12. |
6835 |
++ */ |
6836 |
++ WARN_ON_ONCE(vmx->emulation_required && vmx->nested.nested_run_pending); |
6837 |
++ |
6838 |
++ intr_window_requested = exec_controls_get(vmx) & |
6839 |
++ CPU_BASED_INTR_WINDOW_EXITING; |
6840 |
++ |
6841 |
++ while (vmx->emulation_required && count-- != 0) { |
6842 |
++ if (intr_window_requested && vmx_interrupt_allowed(vcpu)) |
6843 |
++ return handle_interrupt_window(&vmx->vcpu); |
6844 |
++ |
6845 |
++ if (kvm_test_request(KVM_REQ_EVENT, vcpu)) |
6846 |
++ return 1; |
6847 |
++ |
6848 |
++ if (!kvm_emulate_instruction(vcpu, 0)) |
6849 |
++ return 0; |
6850 |
++ |
6851 |
++ if (vmx->emulation_required && !vmx->rmode.vm86_active && |
6852 |
++ vcpu->arch.exception.pending) { |
6853 |
++ vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
6854 |
++ vcpu->run->internal.suberror = |
6855 |
++ KVM_INTERNAL_ERROR_EMULATION; |
6856 |
++ vcpu->run->internal.ndata = 0; |
6857 |
++ return 0; |
6858 |
++ } |
6859 |
++ |
6860 |
++ if (vcpu->arch.halt_request) { |
6861 |
++ vcpu->arch.halt_request = 0; |
6862 |
++ return kvm_vcpu_halt(vcpu); |
6863 |
++ } |
6864 |
++ |
6865 |
++ /* |
6866 |
++ * Note, return 1 and not 0, vcpu_run() is responsible for |
6867 |
++ * morphing the pending signal into the proper return code. |
6868 |
++ */ |
6869 |
++ if (signal_pending(current)) |
6870 |
++ return 1; |
6871 |
++ |
6872 |
++ if (need_resched()) |
6873 |
++ schedule(); |
6874 |
++ } |
6875 |
++ |
6876 |
++ return 1; |
6877 |
++} |
6878 |
++ |
6879 |
++static void grow_ple_window(struct kvm_vcpu *vcpu) |
6880 |
++{ |
6881 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6882 |
++ unsigned int old = vmx->ple_window; |
6883 |
++ |
6884 |
++ vmx->ple_window = __grow_ple_window(old, ple_window, |
6885 |
++ ple_window_grow, |
6886 |
++ ple_window_max); |
6887 |
++ |
6888 |
++ if (vmx->ple_window != old) { |
6889 |
++ vmx->ple_window_dirty = true; |
6890 |
++ trace_kvm_ple_window_update(vcpu->vcpu_id, |
6891 |
++ vmx->ple_window, old); |
6892 |
++ } |
6893 |
++} |
6894 |
++ |
6895 |
++static void shrink_ple_window(struct kvm_vcpu *vcpu) |
6896 |
++{ |
6897 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
6898 |
++ unsigned int old = vmx->ple_window; |
6899 |
++ |
6900 |
++ vmx->ple_window = __shrink_ple_window(old, ple_window, |
6901 |
++ ple_window_shrink, |
6902 |
++ ple_window); |
6903 |
++ |
6904 |
++ if (vmx->ple_window != old) { |
6905 |
++ vmx->ple_window_dirty = true; |
6906 |
++ trace_kvm_ple_window_update(vcpu->vcpu_id, |
6907 |
++ vmx->ple_window, old); |
6908 |
++ } |
6909 |
++} |
6910 |
++ |
6911 |
++/* |
6912 |
++ * Handler for POSTED_INTERRUPT_WAKEUP_VECTOR. |
6913 |
++ */ |
6914 |
++static void wakeup_handler(void) |
6915 |
++{ |
6916 |
++ struct kvm_vcpu *vcpu; |
6917 |
++ int cpu = smp_processor_id(); |
6918 |
++ |
6919 |
++ spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
6920 |
++ list_for_each_entry(vcpu, &per_cpu(blocked_vcpu_on_cpu, cpu), |
6921 |
++ blocked_vcpu_list) { |
6922 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
6923 |
++ |
6924 |
++ if (pi_test_on(pi_desc) == 1) |
6925 |
++ kvm_vcpu_kick(vcpu); |
6926 |
++ } |
6927 |
++ spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
6928 |
++} |
6929 |
++ |
6930 |
++static void vmx_enable_tdp(void) |
6931 |
++{ |
6932 |
++ kvm_mmu_set_mask_ptes(VMX_EPT_READABLE_MASK, |
6933 |
++ enable_ept_ad_bits ? VMX_EPT_ACCESS_BIT : 0ull, |
6934 |
++ enable_ept_ad_bits ? VMX_EPT_DIRTY_BIT : 0ull, |
6935 |
++ 0ull, VMX_EPT_EXECUTABLE_MASK, |
6936 |
++ cpu_has_vmx_ept_execute_only() ? 0ull : VMX_EPT_READABLE_MASK, |
6937 |
++ VMX_EPT_RWX_MASK, 0ull); |
6938 |
++ |
6939 |
++ ept_set_mmio_spte_mask(); |
6940 |
++ kvm_enable_tdp(); |
6941 |
++} |
6942 |
++ |
6943 |
++/* |
6944 |
++ * Indicate a busy-waiting vcpu in spinlock. We do not enable the PAUSE |
6945 |
++ * exiting, so only get here on cpu with PAUSE-Loop-Exiting. |
6946 |
++ */ |
6947 |
++static int handle_pause(struct kvm_vcpu *vcpu) |
6948 |
++{ |
6949 |
++ if (!kvm_pause_in_guest(vcpu->kvm)) |
6950 |
++ grow_ple_window(vcpu); |
6951 |
++ |
6952 |
++ /* |
6953 |
++ * Intel sdm vol3 ch-25.1.3 says: The "PAUSE-loop exiting" |
6954 |
++ * VM-execution control is ignored if CPL > 0. OTOH, KVM |
6955 |
++ * never set PAUSE_EXITING and just set PLE if supported, |
6956 |
++ * so the vcpu must be CPL=0 if it gets a PAUSE exit. |
6957 |
++ */ |
6958 |
++ kvm_vcpu_on_spin(vcpu, true); |
6959 |
++ return kvm_skip_emulated_instruction(vcpu); |
6960 |
++} |
6961 |
++ |
6962 |
++static int handle_nop(struct kvm_vcpu *vcpu) |
6963 |
++{ |
6964 |
++ return kvm_skip_emulated_instruction(vcpu); |
6965 |
++} |
6966 |
++ |
6967 |
++static int handle_mwait(struct kvm_vcpu *vcpu) |
6968 |
++{ |
6969 |
++ printk_once(KERN_WARNING "kvm: MWAIT instruction emulated as NOP!\n"); |
6970 |
++ return handle_nop(vcpu); |
6971 |
++} |
6972 |
++ |
6973 |
++static int handle_invalid_op(struct kvm_vcpu *vcpu) |
6974 |
++{ |
6975 |
++ kvm_queue_exception(vcpu, UD_VECTOR); |
6976 |
++ return 1; |
6977 |
++} |
6978 |
++ |
6979 |
++static int handle_monitor_trap(struct kvm_vcpu *vcpu) |
6980 |
++{ |
6981 |
++ return 1; |
6982 |
++} |
6983 |
++ |
6984 |
++static int handle_monitor(struct kvm_vcpu *vcpu) |
6985 |
++{ |
6986 |
++ printk_once(KERN_WARNING "kvm: MONITOR instruction emulated as NOP!\n"); |
6987 |
++ return handle_nop(vcpu); |
6988 |
++} |
6989 |
++ |
6990 |
++static int handle_invpcid(struct kvm_vcpu *vcpu) |
6991 |
++{ |
6992 |
++ u32 vmx_instruction_info; |
6993 |
++ unsigned long type; |
6994 |
++ bool pcid_enabled; |
6995 |
++ gva_t gva; |
6996 |
++ struct x86_exception e; |
6997 |
++ unsigned i; |
6998 |
++ unsigned long roots_to_free = 0; |
6999 |
++ struct { |
7000 |
++ u64 pcid; |
7001 |
++ u64 gla; |
7002 |
++ } operand; |
7003 |
++ |
7004 |
++ if (!guest_cpuid_has(vcpu, X86_FEATURE_INVPCID)) { |
7005 |
++ kvm_queue_exception(vcpu, UD_VECTOR); |
7006 |
++ return 1; |
7007 |
++ } |
7008 |
++ |
7009 |
++ vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); |
7010 |
++ type = kvm_register_readl(vcpu, (vmx_instruction_info >> 28) & 0xf); |
7011 |
++ |
7012 |
++ if (type > 3) { |
7013 |
++ kvm_inject_gp(vcpu, 0); |
7014 |
++ return 1; |
7015 |
++ } |
7016 |
++ |
7017 |
++ /* According to the Intel instruction reference, the memory operand |
7018 |
++ * is read even if it isn't needed (e.g., for type==all) |
7019 |
++ */ |
7020 |
++ if (get_vmx_mem_address(vcpu, vmcs_readl(EXIT_QUALIFICATION), |
7021 |
++ vmx_instruction_info, false, |
7022 |
++ sizeof(operand), &gva)) |
7023 |
++ return 1; |
7024 |
++ |
7025 |
++ if (kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e)) { |
7026 |
++ kvm_inject_page_fault(vcpu, &e); |
7027 |
++ return 1; |
7028 |
++ } |
7029 |
++ |
7030 |
++ if (operand.pcid >> 12 != 0) { |
7031 |
++ kvm_inject_gp(vcpu, 0); |
7032 |
++ return 1; |
7033 |
++ } |
7034 |
++ |
7035 |
++ pcid_enabled = kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE); |
7036 |
++ |
7037 |
++ switch (type) { |
7038 |
++ case INVPCID_TYPE_INDIV_ADDR: |
7039 |
++ if ((!pcid_enabled && (operand.pcid != 0)) || |
7040 |
++ is_noncanonical_address(operand.gla, vcpu)) { |
7041 |
++ kvm_inject_gp(vcpu, 0); |
7042 |
++ return 1; |
7043 |
++ } |
7044 |
++ kvm_mmu_invpcid_gva(vcpu, operand.gla, operand.pcid); |
7045 |
++ return kvm_skip_emulated_instruction(vcpu); |
7046 |
++ |
7047 |
++ case INVPCID_TYPE_SINGLE_CTXT: |
7048 |
++ if (!pcid_enabled && (operand.pcid != 0)) { |
7049 |
++ kvm_inject_gp(vcpu, 0); |
7050 |
++ return 1; |
7051 |
++ } |
7052 |
++ |
7053 |
++ if (kvm_get_active_pcid(vcpu) == operand.pcid) { |
7054 |
++ kvm_mmu_sync_roots(vcpu); |
7055 |
++ kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
7056 |
++ } |
7057 |
++ |
7058 |
++ for (i = 0; i < KVM_MMU_NUM_PREV_ROOTS; i++) |
7059 |
++ if (kvm_get_pcid(vcpu, vcpu->arch.mmu->prev_roots[i].cr3) |
7060 |
++ == operand.pcid) |
7061 |
++ roots_to_free |= KVM_MMU_ROOT_PREVIOUS(i); |
7062 |
++ |
7063 |
++ kvm_mmu_free_roots(vcpu, vcpu->arch.mmu, roots_to_free); |
7064 |
++ /* |
7065 |
++ * If neither the current cr3 nor any of the prev_roots use the |
7066 |
++ * given PCID, then nothing needs to be done here because a |
7067 |
++ * resync will happen anyway before switching to any other CR3. |
7068 |
++ */ |
7069 |
++ |
7070 |
++ return kvm_skip_emulated_instruction(vcpu); |
7071 |
++ |
7072 |
++ case INVPCID_TYPE_ALL_NON_GLOBAL: |
7073 |
++ /* |
7074 |
++ * Currently, KVM doesn't mark global entries in the shadow |
7075 |
++ * page tables, so a non-global flush just degenerates to a |
7076 |
++ * global flush. If needed, we could optimize this later by |
7077 |
++ * keeping track of global entries in shadow page tables. |
7078 |
++ */ |
7079 |
++ |
7080 |
++ /* fall-through */ |
7081 |
++ case INVPCID_TYPE_ALL_INCL_GLOBAL: |
7082 |
++ kvm_mmu_unload(vcpu); |
7083 |
++ return kvm_skip_emulated_instruction(vcpu); |
7084 |
++ |
7085 |
++ default: |
7086 |
++ BUG(); /* We have already checked above that type <= 3 */ |
7087 |
++ } |
7088 |
++} |
7089 |
++ |
7090 |
++static int handle_pml_full(struct kvm_vcpu *vcpu) |
7091 |
++{ |
7092 |
++ unsigned long exit_qualification; |
7093 |
++ |
7094 |
++ trace_kvm_pml_full(vcpu->vcpu_id); |
7095 |
++ |
7096 |
++ exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
7097 |
++ |
7098 |
++ /* |
7099 |
++ * PML buffer FULL happened while executing iret from NMI, |
7100 |
++ * "blocked by NMI" bit has to be set before next VM entry. |
7101 |
++ */ |
7102 |
++ if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) && |
7103 |
++ enable_vnmi && |
7104 |
++ (exit_qualification & INTR_INFO_UNBLOCK_NMI)) |
7105 |
++ vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
7106 |
++ GUEST_INTR_STATE_NMI); |
7107 |
++ |
7108 |
++ /* |
7109 |
++ * PML buffer already flushed at beginning of VMEXIT. Nothing to do |
7110 |
++ * here.., and there's no userspace involvement needed for PML. |
7111 |
++ */ |
7112 |
++ return 1; |
7113 |
++} |
7114 |
++ |
7115 |
++static int handle_preemption_timer(struct kvm_vcpu *vcpu) |
7116 |
++{ |
7117 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7118 |
++ |
7119 |
++ if (!vmx->req_immediate_exit && |
7120 |
++ !unlikely(vmx->loaded_vmcs->hv_timer_soft_disabled)) |
7121 |
++ kvm_lapic_expired_hv_timer(vcpu); |
7122 |
++ |
7123 |
++ return 1; |
7124 |
++} |
7125 |
++ |
7126 |
++/* |
7127 |
++ * When nested=0, all VMX instruction VM Exits filter here. The handlers |
7128 |
++ * are overwritten by nested_vmx_setup() when nested=1. |
7129 |
++ */ |
7130 |
++static int handle_vmx_instruction(struct kvm_vcpu *vcpu) |
7131 |
++{ |
7132 |
++ kvm_queue_exception(vcpu, UD_VECTOR); |
7133 |
++ return 1; |
7134 |
++} |
7135 |
++ |
7136 |
++static int handle_encls(struct kvm_vcpu *vcpu) |
7137 |
++{ |
7138 |
++ /* |
7139 |
++ * SGX virtualization is not yet supported. There is no software |
7140 |
++ * enable bit for SGX, so we have to trap ENCLS and inject a #UD |
7141 |
++ * to prevent the guest from executing ENCLS. |
7142 |
++ */ |
7143 |
++ kvm_queue_exception(vcpu, UD_VECTOR); |
7144 |
++ return 1; |
7145 |
++} |
7146 |
++ |
7147 |
++/* |
7148 |
++ * The exit handlers return 1 if the exit was handled fully and guest execution |
7149 |
++ * may resume. Otherwise they set the kvm_run parameter to indicate what needs |
7150 |
++ * to be done to userspace and return 0. |
7151 |
++ */ |
7152 |
++static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = { |
7153 |
++ [EXIT_REASON_EXCEPTION_NMI] = handle_exception_nmi, |
7154 |
++ [EXIT_REASON_EXTERNAL_INTERRUPT] = handle_external_interrupt, |
7155 |
++ [EXIT_REASON_TRIPLE_FAULT] = handle_triple_fault, |
7156 |
++ [EXIT_REASON_NMI_WINDOW] = handle_nmi_window, |
7157 |
++ [EXIT_REASON_IO_INSTRUCTION] = handle_io, |
7158 |
++ [EXIT_REASON_CR_ACCESS] = handle_cr, |
7159 |
++ [EXIT_REASON_DR_ACCESS] = handle_dr, |
7160 |
++ [EXIT_REASON_CPUID] = kvm_emulate_cpuid, |
7161 |
++ [EXIT_REASON_MSR_READ] = kvm_emulate_rdmsr, |
7162 |
++ [EXIT_REASON_MSR_WRITE] = kvm_emulate_wrmsr, |
7163 |
++ [EXIT_REASON_INTERRUPT_WINDOW] = handle_interrupt_window, |
7164 |
++ [EXIT_REASON_HLT] = kvm_emulate_halt, |
7165 |
++ [EXIT_REASON_INVD] = handle_invd, |
7166 |
++ [EXIT_REASON_INVLPG] = handle_invlpg, |
7167 |
++ [EXIT_REASON_RDPMC] = handle_rdpmc, |
7168 |
++ [EXIT_REASON_VMCALL] = handle_vmcall, |
7169 |
++ [EXIT_REASON_VMCLEAR] = handle_vmx_instruction, |
7170 |
++ [EXIT_REASON_VMLAUNCH] = handle_vmx_instruction, |
7171 |
++ [EXIT_REASON_VMPTRLD] = handle_vmx_instruction, |
7172 |
++ [EXIT_REASON_VMPTRST] = handle_vmx_instruction, |
7173 |
++ [EXIT_REASON_VMREAD] = handle_vmx_instruction, |
7174 |
++ [EXIT_REASON_VMRESUME] = handle_vmx_instruction, |
7175 |
++ [EXIT_REASON_VMWRITE] = handle_vmx_instruction, |
7176 |
++ [EXIT_REASON_VMOFF] = handle_vmx_instruction, |
7177 |
++ [EXIT_REASON_VMON] = handle_vmx_instruction, |
7178 |
++ [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, |
7179 |
++ [EXIT_REASON_APIC_ACCESS] = handle_apic_access, |
7180 |
++ [EXIT_REASON_APIC_WRITE] = handle_apic_write, |
7181 |
++ [EXIT_REASON_EOI_INDUCED] = handle_apic_eoi_induced, |
7182 |
++ [EXIT_REASON_WBINVD] = handle_wbinvd, |
7183 |
++ [EXIT_REASON_XSETBV] = handle_xsetbv, |
7184 |
++ [EXIT_REASON_TASK_SWITCH] = handle_task_switch, |
7185 |
++ [EXIT_REASON_MCE_DURING_VMENTRY] = handle_machine_check, |
7186 |
++ [EXIT_REASON_GDTR_IDTR] = handle_desc, |
7187 |
++ [EXIT_REASON_LDTR_TR] = handle_desc, |
7188 |
++ [EXIT_REASON_EPT_VIOLATION] = handle_ept_violation, |
7189 |
++ [EXIT_REASON_EPT_MISCONFIG] = handle_ept_misconfig, |
7190 |
++ [EXIT_REASON_PAUSE_INSTRUCTION] = handle_pause, |
7191 |
++ [EXIT_REASON_MWAIT_INSTRUCTION] = handle_mwait, |
7192 |
++ [EXIT_REASON_MONITOR_TRAP_FLAG] = handle_monitor_trap, |
7193 |
++ [EXIT_REASON_MONITOR_INSTRUCTION] = handle_monitor, |
7194 |
++ [EXIT_REASON_INVEPT] = handle_vmx_instruction, |
7195 |
++ [EXIT_REASON_INVVPID] = handle_vmx_instruction, |
7196 |
++ [EXIT_REASON_RDRAND] = handle_invalid_op, |
7197 |
++ [EXIT_REASON_RDSEED] = handle_invalid_op, |
7198 |
++ [EXIT_REASON_PML_FULL] = handle_pml_full, |
7199 |
++ [EXIT_REASON_INVPCID] = handle_invpcid, |
7200 |
++ [EXIT_REASON_VMFUNC] = handle_vmx_instruction, |
7201 |
++ [EXIT_REASON_PREEMPTION_TIMER] = handle_preemption_timer, |
7202 |
++ [EXIT_REASON_ENCLS] = handle_encls, |
7203 |
++}; |
7204 |
++ |
7205 |
++static const int kvm_vmx_max_exit_handlers = |
7206 |
++ ARRAY_SIZE(kvm_vmx_exit_handlers); |
7207 |
++ |
7208 |
++static void vmx_get_exit_info(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2) |
7209 |
++{ |
7210 |
++ *info1 = vmcs_readl(EXIT_QUALIFICATION); |
7211 |
++ *info2 = vmcs_read32(VM_EXIT_INTR_INFO); |
7212 |
++} |
7213 |
++ |
7214 |
++static void vmx_destroy_pml_buffer(struct vcpu_vmx *vmx) |
7215 |
++{ |
7216 |
++ if (vmx->pml_pg) { |
7217 |
++ __free_page(vmx->pml_pg); |
7218 |
++ vmx->pml_pg = NULL; |
7219 |
++ } |
7220 |
++} |
7221 |
++ |
7222 |
++static void vmx_flush_pml_buffer(struct kvm_vcpu *vcpu) |
7223 |
++{ |
7224 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7225 |
++ u64 *pml_buf; |
7226 |
++ u16 pml_idx; |
7227 |
++ |
7228 |
++ pml_idx = vmcs_read16(GUEST_PML_INDEX); |
7229 |
++ |
7230 |
++ /* Do nothing if PML buffer is empty */ |
7231 |
++ if (pml_idx == (PML_ENTITY_NUM - 1)) |
7232 |
++ return; |
7233 |
++ |
7234 |
++ /* PML index always points to next available PML buffer entity */ |
7235 |
++ if (pml_idx >= PML_ENTITY_NUM) |
7236 |
++ pml_idx = 0; |
7237 |
++ else |
7238 |
++ pml_idx++; |
7239 |
++ |
7240 |
++ pml_buf = page_address(vmx->pml_pg); |
7241 |
++ for (; pml_idx < PML_ENTITY_NUM; pml_idx++) { |
7242 |
++ u64 gpa; |
7243 |
++ |
7244 |
++ gpa = pml_buf[pml_idx]; |
7245 |
++ WARN_ON(gpa & (PAGE_SIZE - 1)); |
7246 |
++ kvm_vcpu_mark_page_dirty(vcpu, gpa >> PAGE_SHIFT); |
7247 |
++ } |
7248 |
++ |
7249 |
++ /* reset PML index */ |
7250 |
++ vmcs_write16(GUEST_PML_INDEX, PML_ENTITY_NUM - 1); |
7251 |
++} |
7252 |
++ |
7253 |
++/* |
7254 |
++ * Flush all vcpus' PML buffer and update logged GPAs to dirty_bitmap. |
7255 |
++ * Called before reporting dirty_bitmap to userspace. |
7256 |
++ */ |
7257 |
++static void kvm_flush_pml_buffers(struct kvm *kvm) |
7258 |
++{ |
7259 |
++ int i; |
7260 |
++ struct kvm_vcpu *vcpu; |
7261 |
++ /* |
7262 |
++ * We only need to kick vcpu out of guest mode here, as PML buffer |
7263 |
++ * is flushed at beginning of all VMEXITs, and it's obvious that only |
7264 |
++ * vcpus running in guest are possible to have unflushed GPAs in PML |
7265 |
++ * buffer. |
7266 |
++ */ |
7267 |
++ kvm_for_each_vcpu(i, vcpu, kvm) |
7268 |
++ kvm_vcpu_kick(vcpu); |
7269 |
++} |
7270 |
++ |
7271 |
++static void vmx_dump_sel(char *name, uint32_t sel) |
7272 |
++{ |
7273 |
++ pr_err("%s sel=0x%04x, attr=0x%05x, limit=0x%08x, base=0x%016lx\n", |
7274 |
++ name, vmcs_read16(sel), |
7275 |
++ vmcs_read32(sel + GUEST_ES_AR_BYTES - GUEST_ES_SELECTOR), |
7276 |
++ vmcs_read32(sel + GUEST_ES_LIMIT - GUEST_ES_SELECTOR), |
7277 |
++ vmcs_readl(sel + GUEST_ES_BASE - GUEST_ES_SELECTOR)); |
7278 |
++} |
7279 |
++ |
7280 |
++static void vmx_dump_dtsel(char *name, uint32_t limit) |
7281 |
++{ |
7282 |
++ pr_err("%s limit=0x%08x, base=0x%016lx\n", |
7283 |
++ name, vmcs_read32(limit), |
7284 |
++ vmcs_readl(limit + GUEST_GDTR_BASE - GUEST_GDTR_LIMIT)); |
7285 |
++} |
7286 |
++ |
7287 |
++void dump_vmcs(void) |
7288 |
++{ |
7289 |
++ u32 vmentry_ctl, vmexit_ctl; |
7290 |
++ u32 cpu_based_exec_ctrl, pin_based_exec_ctrl, secondary_exec_control; |
7291 |
++ unsigned long cr4; |
7292 |
++ u64 efer; |
7293 |
++ int i, n; |
7294 |
++ |
7295 |
++ if (!dump_invalid_vmcs) { |
7296 |
++ pr_warn_ratelimited("set kvm_intel.dump_invalid_vmcs=1 to dump internal KVM state.\n"); |
7297 |
++ return; |
7298 |
++ } |
7299 |
++ |
7300 |
++ vmentry_ctl = vmcs_read32(VM_ENTRY_CONTROLS); |
7301 |
++ vmexit_ctl = vmcs_read32(VM_EXIT_CONTROLS); |
7302 |
++ cpu_based_exec_ctrl = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); |
7303 |
++ pin_based_exec_ctrl = vmcs_read32(PIN_BASED_VM_EXEC_CONTROL); |
7304 |
++ cr4 = vmcs_readl(GUEST_CR4); |
7305 |
++ efer = vmcs_read64(GUEST_IA32_EFER); |
7306 |
++ secondary_exec_control = 0; |
7307 |
++ if (cpu_has_secondary_exec_ctrls()) |
7308 |
++ secondary_exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL); |
7309 |
++ |
7310 |
++ pr_err("*** Guest State ***\n"); |
7311 |
++ pr_err("CR0: actual=0x%016lx, shadow=0x%016lx, gh_mask=%016lx\n", |
7312 |
++ vmcs_readl(GUEST_CR0), vmcs_readl(CR0_READ_SHADOW), |
7313 |
++ vmcs_readl(CR0_GUEST_HOST_MASK)); |
7314 |
++ pr_err("CR4: actual=0x%016lx, shadow=0x%016lx, gh_mask=%016lx\n", |
7315 |
++ cr4, vmcs_readl(CR4_READ_SHADOW), vmcs_readl(CR4_GUEST_HOST_MASK)); |
7316 |
++ pr_err("CR3 = 0x%016lx\n", vmcs_readl(GUEST_CR3)); |
7317 |
++ if ((secondary_exec_control & SECONDARY_EXEC_ENABLE_EPT) && |
7318 |
++ (cr4 & X86_CR4_PAE) && !(efer & EFER_LMA)) |
7319 |
++ { |
7320 |
++ pr_err("PDPTR0 = 0x%016llx PDPTR1 = 0x%016llx\n", |
7321 |
++ vmcs_read64(GUEST_PDPTR0), vmcs_read64(GUEST_PDPTR1)); |
7322 |
++ pr_err("PDPTR2 = 0x%016llx PDPTR3 = 0x%016llx\n", |
7323 |
++ vmcs_read64(GUEST_PDPTR2), vmcs_read64(GUEST_PDPTR3)); |
7324 |
++ } |
7325 |
++ pr_err("RSP = 0x%016lx RIP = 0x%016lx\n", |
7326 |
++ vmcs_readl(GUEST_RSP), vmcs_readl(GUEST_RIP)); |
7327 |
++ pr_err("RFLAGS=0x%08lx DR7 = 0x%016lx\n", |
7328 |
++ vmcs_readl(GUEST_RFLAGS), vmcs_readl(GUEST_DR7)); |
7329 |
++ pr_err("Sysenter RSP=%016lx CS:RIP=%04x:%016lx\n", |
7330 |
++ vmcs_readl(GUEST_SYSENTER_ESP), |
7331 |
++ vmcs_read32(GUEST_SYSENTER_CS), vmcs_readl(GUEST_SYSENTER_EIP)); |
7332 |
++ vmx_dump_sel("CS: ", GUEST_CS_SELECTOR); |
7333 |
++ vmx_dump_sel("DS: ", GUEST_DS_SELECTOR); |
7334 |
++ vmx_dump_sel("SS: ", GUEST_SS_SELECTOR); |
7335 |
++ vmx_dump_sel("ES: ", GUEST_ES_SELECTOR); |
7336 |
++ vmx_dump_sel("FS: ", GUEST_FS_SELECTOR); |
7337 |
++ vmx_dump_sel("GS: ", GUEST_GS_SELECTOR); |
7338 |
++ vmx_dump_dtsel("GDTR:", GUEST_GDTR_LIMIT); |
7339 |
++ vmx_dump_sel("LDTR:", GUEST_LDTR_SELECTOR); |
7340 |
++ vmx_dump_dtsel("IDTR:", GUEST_IDTR_LIMIT); |
7341 |
++ vmx_dump_sel("TR: ", GUEST_TR_SELECTOR); |
7342 |
++ if ((vmexit_ctl & (VM_EXIT_SAVE_IA32_PAT | VM_EXIT_SAVE_IA32_EFER)) || |
7343 |
++ (vmentry_ctl & (VM_ENTRY_LOAD_IA32_PAT | VM_ENTRY_LOAD_IA32_EFER))) |
7344 |
++ pr_err("EFER = 0x%016llx PAT = 0x%016llx\n", |
7345 |
++ efer, vmcs_read64(GUEST_IA32_PAT)); |
7346 |
++ pr_err("DebugCtl = 0x%016llx DebugExceptions = 0x%016lx\n", |
7347 |
++ vmcs_read64(GUEST_IA32_DEBUGCTL), |
7348 |
++ vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS)); |
7349 |
++ if (cpu_has_load_perf_global_ctrl() && |
7350 |
++ vmentry_ctl & VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) |
7351 |
++ pr_err("PerfGlobCtl = 0x%016llx\n", |
7352 |
++ vmcs_read64(GUEST_IA32_PERF_GLOBAL_CTRL)); |
7353 |
++ if (vmentry_ctl & VM_ENTRY_LOAD_BNDCFGS) |
7354 |
++ pr_err("BndCfgS = 0x%016llx\n", vmcs_read64(GUEST_BNDCFGS)); |
7355 |
++ pr_err("Interruptibility = %08x ActivityState = %08x\n", |
7356 |
++ vmcs_read32(GUEST_INTERRUPTIBILITY_INFO), |
7357 |
++ vmcs_read32(GUEST_ACTIVITY_STATE)); |
7358 |
++ if (secondary_exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY) |
7359 |
++ pr_err("InterruptStatus = %04x\n", |
7360 |
++ vmcs_read16(GUEST_INTR_STATUS)); |
7361 |
++ |
7362 |
++ pr_err("*** Host State ***\n"); |
7363 |
++ pr_err("RIP = 0x%016lx RSP = 0x%016lx\n", |
7364 |
++ vmcs_readl(HOST_RIP), vmcs_readl(HOST_RSP)); |
7365 |
++ pr_err("CS=%04x SS=%04x DS=%04x ES=%04x FS=%04x GS=%04x TR=%04x\n", |
7366 |
++ vmcs_read16(HOST_CS_SELECTOR), vmcs_read16(HOST_SS_SELECTOR), |
7367 |
++ vmcs_read16(HOST_DS_SELECTOR), vmcs_read16(HOST_ES_SELECTOR), |
7368 |
++ vmcs_read16(HOST_FS_SELECTOR), vmcs_read16(HOST_GS_SELECTOR), |
7369 |
++ vmcs_read16(HOST_TR_SELECTOR)); |
7370 |
++ pr_err("FSBase=%016lx GSBase=%016lx TRBase=%016lx\n", |
7371 |
++ vmcs_readl(HOST_FS_BASE), vmcs_readl(HOST_GS_BASE), |
7372 |
++ vmcs_readl(HOST_TR_BASE)); |
7373 |
++ pr_err("GDTBase=%016lx IDTBase=%016lx\n", |
7374 |
++ vmcs_readl(HOST_GDTR_BASE), vmcs_readl(HOST_IDTR_BASE)); |
7375 |
++ pr_err("CR0=%016lx CR3=%016lx CR4=%016lx\n", |
7376 |
++ vmcs_readl(HOST_CR0), vmcs_readl(HOST_CR3), |
7377 |
++ vmcs_readl(HOST_CR4)); |
7378 |
++ pr_err("Sysenter RSP=%016lx CS:RIP=%04x:%016lx\n", |
7379 |
++ vmcs_readl(HOST_IA32_SYSENTER_ESP), |
7380 |
++ vmcs_read32(HOST_IA32_SYSENTER_CS), |
7381 |
++ vmcs_readl(HOST_IA32_SYSENTER_EIP)); |
7382 |
++ if (vmexit_ctl & (VM_EXIT_LOAD_IA32_PAT | VM_EXIT_LOAD_IA32_EFER)) |
7383 |
++ pr_err("EFER = 0x%016llx PAT = 0x%016llx\n", |
7384 |
++ vmcs_read64(HOST_IA32_EFER), |
7385 |
++ vmcs_read64(HOST_IA32_PAT)); |
7386 |
++ if (cpu_has_load_perf_global_ctrl() && |
7387 |
++ vmexit_ctl & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) |
7388 |
++ pr_err("PerfGlobCtl = 0x%016llx\n", |
7389 |
++ vmcs_read64(HOST_IA32_PERF_GLOBAL_CTRL)); |
7390 |
++ |
7391 |
++ pr_err("*** Control State ***\n"); |
7392 |
++ pr_err("PinBased=%08x CPUBased=%08x SecondaryExec=%08x\n", |
7393 |
++ pin_based_exec_ctrl, cpu_based_exec_ctrl, secondary_exec_control); |
7394 |
++ pr_err("EntryControls=%08x ExitControls=%08x\n", vmentry_ctl, vmexit_ctl); |
7395 |
++ pr_err("ExceptionBitmap=%08x PFECmask=%08x PFECmatch=%08x\n", |
7396 |
++ vmcs_read32(EXCEPTION_BITMAP), |
7397 |
++ vmcs_read32(PAGE_FAULT_ERROR_CODE_MASK), |
7398 |
++ vmcs_read32(PAGE_FAULT_ERROR_CODE_MATCH)); |
7399 |
++ pr_err("VMEntry: intr_info=%08x errcode=%08x ilen=%08x\n", |
7400 |
++ vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), |
7401 |
++ vmcs_read32(VM_ENTRY_EXCEPTION_ERROR_CODE), |
7402 |
++ vmcs_read32(VM_ENTRY_INSTRUCTION_LEN)); |
7403 |
++ pr_err("VMExit: intr_info=%08x errcode=%08x ilen=%08x\n", |
7404 |
++ vmcs_read32(VM_EXIT_INTR_INFO), |
7405 |
++ vmcs_read32(VM_EXIT_INTR_ERROR_CODE), |
7406 |
++ vmcs_read32(VM_EXIT_INSTRUCTION_LEN)); |
7407 |
++ pr_err(" reason=%08x qualification=%016lx\n", |
7408 |
++ vmcs_read32(VM_EXIT_REASON), vmcs_readl(EXIT_QUALIFICATION)); |
7409 |
++ pr_err("IDTVectoring: info=%08x errcode=%08x\n", |
7410 |
++ vmcs_read32(IDT_VECTORING_INFO_FIELD), |
7411 |
++ vmcs_read32(IDT_VECTORING_ERROR_CODE)); |
7412 |
++ pr_err("TSC Offset = 0x%016llx\n", vmcs_read64(TSC_OFFSET)); |
7413 |
++ if (secondary_exec_control & SECONDARY_EXEC_TSC_SCALING) |
7414 |
++ pr_err("TSC Multiplier = 0x%016llx\n", |
7415 |
++ vmcs_read64(TSC_MULTIPLIER)); |
7416 |
++ if (cpu_based_exec_ctrl & CPU_BASED_TPR_SHADOW) { |
7417 |
++ if (secondary_exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY) { |
7418 |
++ u16 status = vmcs_read16(GUEST_INTR_STATUS); |
7419 |
++ pr_err("SVI|RVI = %02x|%02x ", status >> 8, status & 0xff); |
7420 |
++ } |
7421 |
++ pr_cont("TPR Threshold = 0x%02x\n", vmcs_read32(TPR_THRESHOLD)); |
7422 |
++ if (secondary_exec_control & SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES) |
7423 |
++ pr_err("APIC-access addr = 0x%016llx ", vmcs_read64(APIC_ACCESS_ADDR)); |
7424 |
++ pr_cont("virt-APIC addr = 0x%016llx\n", vmcs_read64(VIRTUAL_APIC_PAGE_ADDR)); |
7425 |
++ } |
7426 |
++ if (pin_based_exec_ctrl & PIN_BASED_POSTED_INTR) |
7427 |
++ pr_err("PostedIntrVec = 0x%02x\n", vmcs_read16(POSTED_INTR_NV)); |
7428 |
++ if ((secondary_exec_control & SECONDARY_EXEC_ENABLE_EPT)) |
7429 |
++ pr_err("EPT pointer = 0x%016llx\n", vmcs_read64(EPT_POINTER)); |
7430 |
++ n = vmcs_read32(CR3_TARGET_COUNT); |
7431 |
++ for (i = 0; i + 1 < n; i += 4) |
7432 |
++ pr_err("CR3 target%u=%016lx target%u=%016lx\n", |
7433 |
++ i, vmcs_readl(CR3_TARGET_VALUE0 + i * 2), |
7434 |
++ i + 1, vmcs_readl(CR3_TARGET_VALUE0 + i * 2 + 2)); |
7435 |
++ if (i < n) |
7436 |
++ pr_err("CR3 target%u=%016lx\n", |
7437 |
++ i, vmcs_readl(CR3_TARGET_VALUE0 + i * 2)); |
7438 |
++ if (secondary_exec_control & SECONDARY_EXEC_PAUSE_LOOP_EXITING) |
7439 |
++ pr_err("PLE Gap=%08x Window=%08x\n", |
7440 |
++ vmcs_read32(PLE_GAP), vmcs_read32(PLE_WINDOW)); |
7441 |
++ if (secondary_exec_control & SECONDARY_EXEC_ENABLE_VPID) |
7442 |
++ pr_err("Virtual processor ID = 0x%04x\n", |
7443 |
++ vmcs_read16(VIRTUAL_PROCESSOR_ID)); |
7444 |
++} |
7445 |
++ |
7446 |
++/* |
7447 |
++ * The guest has exited. See if we can fix it or if we need userspace |
7448 |
++ * assistance. |
7449 |
++ */ |
7450 |
++static int vmx_handle_exit(struct kvm_vcpu *vcpu, |
7451 |
++ enum exit_fastpath_completion exit_fastpath) |
7452 |
++{ |
7453 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7454 |
++ u32 exit_reason = vmx->exit_reason; |
7455 |
++ u32 vectoring_info = vmx->idt_vectoring_info; |
7456 |
++ |
7457 |
++ trace_kvm_exit(exit_reason, vcpu, KVM_ISA_VMX); |
7458 |
++ |
7459 |
++ /* |
7460 |
++ * Flush logged GPAs PML buffer, this will make dirty_bitmap more |
7461 |
++ * updated. Another good is, in kvm_vm_ioctl_get_dirty_log, before |
7462 |
++ * querying dirty_bitmap, we only need to kick all vcpus out of guest |
7463 |
++ * mode as if vcpus is in root mode, the PML buffer must has been |
7464 |
++ * flushed already. |
7465 |
++ */ |
7466 |
++ if (enable_pml) |
7467 |
++ vmx_flush_pml_buffer(vcpu); |
7468 |
++ |
7469 |
++ /* If guest state is invalid, start emulating */ |
7470 |
++ if (vmx->emulation_required) |
7471 |
++ return handle_invalid_guest_state(vcpu); |
7472 |
++ |
7473 |
++ if (is_guest_mode(vcpu) && nested_vmx_exit_reflected(vcpu, exit_reason)) |
7474 |
++ return nested_vmx_reflect_vmexit(vcpu, exit_reason); |
7475 |
++ |
7476 |
++ if (exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) { |
7477 |
++ dump_vmcs(); |
7478 |
++ vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
7479 |
++ vcpu->run->fail_entry.hardware_entry_failure_reason |
7480 |
++ = exit_reason; |
7481 |
++ return 0; |
7482 |
++ } |
7483 |
++ |
7484 |
++ if (unlikely(vmx->fail)) { |
7485 |
++ dump_vmcs(); |
7486 |
++ vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
7487 |
++ vcpu->run->fail_entry.hardware_entry_failure_reason |
7488 |
++ = vmcs_read32(VM_INSTRUCTION_ERROR); |
7489 |
++ return 0; |
7490 |
++ } |
7491 |
++ |
7492 |
++ /* |
7493 |
++ * Note: |
7494 |
++ * Do not try to fix EXIT_REASON_EPT_MISCONFIG if it caused by |
7495 |
++ * delivery event since it indicates guest is accessing MMIO. |
7496 |
++ * The vm-exit can be triggered again after return to guest that |
7497 |
++ * will cause infinite loop. |
7498 |
++ */ |
7499 |
++ if ((vectoring_info & VECTORING_INFO_VALID_MASK) && |
7500 |
++ (exit_reason != EXIT_REASON_EXCEPTION_NMI && |
7501 |
++ exit_reason != EXIT_REASON_EPT_VIOLATION && |
7502 |
++ exit_reason != EXIT_REASON_PML_FULL && |
7503 |
++ exit_reason != EXIT_REASON_TASK_SWITCH)) { |
7504 |
++ vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
7505 |
++ vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_DELIVERY_EV; |
7506 |
++ vcpu->run->internal.ndata = 3; |
7507 |
++ vcpu->run->internal.data[0] = vectoring_info; |
7508 |
++ vcpu->run->internal.data[1] = exit_reason; |
7509 |
++ vcpu->run->internal.data[2] = vcpu->arch.exit_qualification; |
7510 |
++ if (exit_reason == EXIT_REASON_EPT_MISCONFIG) { |
7511 |
++ vcpu->run->internal.ndata++; |
7512 |
++ vcpu->run->internal.data[3] = |
7513 |
++ vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
7514 |
++ } |
7515 |
++ return 0; |
7516 |
++ } |
7517 |
++ |
7518 |
++ if (unlikely(!enable_vnmi && |
7519 |
++ vmx->loaded_vmcs->soft_vnmi_blocked)) { |
7520 |
++ if (vmx_interrupt_allowed(vcpu)) { |
7521 |
++ vmx->loaded_vmcs->soft_vnmi_blocked = 0; |
7522 |
++ } else if (vmx->loaded_vmcs->vnmi_blocked_time > 1000000000LL && |
7523 |
++ vcpu->arch.nmi_pending) { |
7524 |
++ /* |
7525 |
++ * This CPU don't support us in finding the end of an |
7526 |
++ * NMI-blocked window if the guest runs with IRQs |
7527 |
++ * disabled. So we pull the trigger after 1 s of |
7528 |
++ * futile waiting, but inform the user about this. |
7529 |
++ */ |
7530 |
++ printk(KERN_WARNING "%s: Breaking out of NMI-blocked " |
7531 |
++ "state on VCPU %d after 1 s timeout\n", |
7532 |
++ __func__, vcpu->vcpu_id); |
7533 |
++ vmx->loaded_vmcs->soft_vnmi_blocked = 0; |
7534 |
++ } |
7535 |
++ } |
7536 |
++ |
7537 |
++ if (exit_fastpath == EXIT_FASTPATH_SKIP_EMUL_INS) { |
7538 |
++ kvm_skip_emulated_instruction(vcpu); |
7539 |
++ return 1; |
7540 |
++ } else if (exit_reason < kvm_vmx_max_exit_handlers |
7541 |
++ && kvm_vmx_exit_handlers[exit_reason]) { |
7542 |
++#ifdef CONFIG_RETPOLINE |
7543 |
++ if (exit_reason == EXIT_REASON_MSR_WRITE) |
7544 |
++ return kvm_emulate_wrmsr(vcpu); |
7545 |
++ else if (exit_reason == EXIT_REASON_PREEMPTION_TIMER) |
7546 |
++ return handle_preemption_timer(vcpu); |
7547 |
++ else if (exit_reason == EXIT_REASON_INTERRUPT_WINDOW) |
7548 |
++ return handle_interrupt_window(vcpu); |
7549 |
++ else if (exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT) |
7550 |
++ return handle_external_interrupt(vcpu); |
7551 |
++ else if (exit_reason == EXIT_REASON_HLT) |
7552 |
++ return kvm_emulate_halt(vcpu); |
7553 |
++ else if (exit_reason == EXIT_REASON_EPT_MISCONFIG) |
7554 |
++ return handle_ept_misconfig(vcpu); |
7555 |
++#endif |
7556 |
++ return kvm_vmx_exit_handlers[exit_reason](vcpu); |
7557 |
++ } else { |
7558 |
++ vcpu_unimpl(vcpu, "vmx: unexpected exit reason 0x%x\n", |
7559 |
++ exit_reason); |
7560 |
++ dump_vmcs(); |
7561 |
++ vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
7562 |
++ vcpu->run->internal.suberror = |
7563 |
++ KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; |
7564 |
++ vcpu->run->internal.ndata = 1; |
7565 |
++ vcpu->run->internal.data[0] = exit_reason; |
7566 |
++ return 0; |
7567 |
++ } |
7568 |
++} |
7569 |
++ |
7570 |
++/* |
7571 |
++ * Software based L1D cache flush which is used when microcode providing |
7572 |
++ * the cache control MSR is not loaded. |
7573 |
++ * |
7574 |
++ * The L1D cache is 32 KiB on Nehalem and later microarchitectures, but to |
7575 |
++ * flush it is required to read in 64 KiB because the replacement algorithm |
7576 |
++ * is not exactly LRU. This could be sized at runtime via topology |
7577 |
++ * information but as all relevant affected CPUs have 32KiB L1D cache size |
7578 |
++ * there is no point in doing so. |
7579 |
++ */ |
7580 |
++static void vmx_l1d_flush(struct kvm_vcpu *vcpu) |
7581 |
++{ |
7582 |
++ int size = PAGE_SIZE << L1D_CACHE_ORDER; |
7583 |
++ |
7584 |
++ /* |
7585 |
++ * This code is only executed when the the flush mode is 'cond' or |
7586 |
++ * 'always' |
7587 |
++ */ |
7588 |
++ if (static_branch_likely(&vmx_l1d_flush_cond)) { |
7589 |
++ bool flush_l1d; |
7590 |
++ |
7591 |
++ /* |
7592 |
++ * Clear the per-vcpu flush bit, it gets set again |
7593 |
++ * either from vcpu_run() or from one of the unsafe |
7594 |
++ * VMEXIT handlers. |
7595 |
++ */ |
7596 |
++ flush_l1d = vcpu->arch.l1tf_flush_l1d; |
7597 |
++ vcpu->arch.l1tf_flush_l1d = false; |
7598 |
++ |
7599 |
++ /* |
7600 |
++ * Clear the per-cpu flush bit, it gets set again from |
7601 |
++ * the interrupt handlers. |
7602 |
++ */ |
7603 |
++ flush_l1d |= kvm_get_cpu_l1tf_flush_l1d(); |
7604 |
++ kvm_clear_cpu_l1tf_flush_l1d(); |
7605 |
++ |
7606 |
++ if (!flush_l1d) |
7607 |
++ return; |
7608 |
++ } |
7609 |
++ |
7610 |
++ vcpu->stat.l1d_flush++; |
7611 |
++ |
7612 |
++ if (static_cpu_has(X86_FEATURE_FLUSH_L1D)) { |
7613 |
++ wrmsrl(MSR_IA32_FLUSH_CMD, L1D_FLUSH); |
7614 |
++ return; |
7615 |
++ } |
7616 |
++ |
7617 |
++ asm volatile( |
7618 |
++ /* First ensure the pages are in the TLB */ |
7619 |
++ "xorl %%eax, %%eax\n" |
7620 |
++ ".Lpopulate_tlb:\n\t" |
7621 |
++ "movzbl (%[flush_pages], %%" _ASM_AX "), %%ecx\n\t" |
7622 |
++ "addl $4096, %%eax\n\t" |
7623 |
++ "cmpl %%eax, %[size]\n\t" |
7624 |
++ "jne .Lpopulate_tlb\n\t" |
7625 |
++ "xorl %%eax, %%eax\n\t" |
7626 |
++ "cpuid\n\t" |
7627 |
++ /* Now fill the cache */ |
7628 |
++ "xorl %%eax, %%eax\n" |
7629 |
++ ".Lfill_cache:\n" |
7630 |
++ "movzbl (%[flush_pages], %%" _ASM_AX "), %%ecx\n\t" |
7631 |
++ "addl $64, %%eax\n\t" |
7632 |
++ "cmpl %%eax, %[size]\n\t" |
7633 |
++ "jne .Lfill_cache\n\t" |
7634 |
++ "lfence\n" |
7635 |
++ :: [flush_pages] "r" (vmx_l1d_flush_pages), |
7636 |
++ [size] "r" (size) |
7637 |
++ : "eax", "ebx", "ecx", "edx"); |
7638 |
++} |
7639 |
++ |
7640 |
++static void update_cr8_intercept(struct kvm_vcpu *vcpu, int tpr, int irr) |
7641 |
++{ |
7642 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
7643 |
++ int tpr_threshold; |
7644 |
++ |
7645 |
++ if (is_guest_mode(vcpu) && |
7646 |
++ nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW)) |
7647 |
++ return; |
7648 |
++ |
7649 |
++ tpr_threshold = (irr == -1 || tpr < irr) ? 0 : irr; |
7650 |
++ if (is_guest_mode(vcpu)) |
7651 |
++ to_vmx(vcpu)->nested.l1_tpr_threshold = tpr_threshold; |
7652 |
++ else |
7653 |
++ vmcs_write32(TPR_THRESHOLD, tpr_threshold); |
7654 |
++} |
7655 |
++ |
7656 |
++void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu) |
7657 |
++{ |
7658 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7659 |
++ u32 sec_exec_control; |
7660 |
++ |
7661 |
++ if (!lapic_in_kernel(vcpu)) |
7662 |
++ return; |
7663 |
++ |
7664 |
++ if (!flexpriority_enabled && |
7665 |
++ !cpu_has_vmx_virtualize_x2apic_mode()) |
7666 |
++ return; |
7667 |
++ |
7668 |
++ /* Postpone execution until vmcs01 is the current VMCS. */ |
7669 |
++ if (is_guest_mode(vcpu)) { |
7670 |
++ vmx->nested.change_vmcs01_virtual_apic_mode = true; |
7671 |
++ return; |
7672 |
++ } |
7673 |
++ |
7674 |
++ sec_exec_control = secondary_exec_controls_get(vmx); |
7675 |
++ sec_exec_control &= ~(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
7676 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE); |
7677 |
++ |
7678 |
++ switch (kvm_get_apic_mode(vcpu)) { |
7679 |
++ case LAPIC_MODE_INVALID: |
7680 |
++ WARN_ONCE(true, "Invalid local APIC state"); |
7681 |
++ case LAPIC_MODE_DISABLED: |
7682 |
++ break; |
7683 |
++ case LAPIC_MODE_XAPIC: |
7684 |
++ if (flexpriority_enabled) { |
7685 |
++ sec_exec_control |= |
7686 |
++ SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; |
7687 |
++ vmx_flush_tlb(vcpu, true); |
7688 |
++ } |
7689 |
++ break; |
7690 |
++ case LAPIC_MODE_X2APIC: |
7691 |
++ if (cpu_has_vmx_virtualize_x2apic_mode()) |
7692 |
++ sec_exec_control |= |
7693 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE; |
7694 |
++ break; |
7695 |
++ } |
7696 |
++ secondary_exec_controls_set(vmx, sec_exec_control); |
7697 |
++ |
7698 |
++ vmx_update_msr_bitmap(vcpu); |
7699 |
++} |
7700 |
++ |
7701 |
++static void vmx_set_apic_access_page_addr(struct kvm_vcpu *vcpu, hpa_t hpa) |
7702 |
++{ |
7703 |
++ if (!is_guest_mode(vcpu)) { |
7704 |
++ vmcs_write64(APIC_ACCESS_ADDR, hpa); |
7705 |
++ vmx_flush_tlb(vcpu, true); |
7706 |
++ } |
7707 |
++} |
7708 |
++ |
7709 |
++static void vmx_hwapic_isr_update(struct kvm_vcpu *vcpu, int max_isr) |
7710 |
++{ |
7711 |
++ u16 status; |
7712 |
++ u8 old; |
7713 |
++ |
7714 |
++ if (max_isr == -1) |
7715 |
++ max_isr = 0; |
7716 |
++ |
7717 |
++ status = vmcs_read16(GUEST_INTR_STATUS); |
7718 |
++ old = status >> 8; |
7719 |
++ if (max_isr != old) { |
7720 |
++ status &= 0xff; |
7721 |
++ status |= max_isr << 8; |
7722 |
++ vmcs_write16(GUEST_INTR_STATUS, status); |
7723 |
++ } |
7724 |
++} |
7725 |
++ |
7726 |
++static void vmx_set_rvi(int vector) |
7727 |
++{ |
7728 |
++ u16 status; |
7729 |
++ u8 old; |
7730 |
++ |
7731 |
++ if (vector == -1) |
7732 |
++ vector = 0; |
7733 |
++ |
7734 |
++ status = vmcs_read16(GUEST_INTR_STATUS); |
7735 |
++ old = (u8)status & 0xff; |
7736 |
++ if ((u8)vector != old) { |
7737 |
++ status &= ~0xff; |
7738 |
++ status |= (u8)vector; |
7739 |
++ vmcs_write16(GUEST_INTR_STATUS, status); |
7740 |
++ } |
7741 |
++} |
7742 |
++ |
7743 |
++static void vmx_hwapic_irr_update(struct kvm_vcpu *vcpu, int max_irr) |
7744 |
++{ |
7745 |
++ /* |
7746 |
++ * When running L2, updating RVI is only relevant when |
7747 |
++ * vmcs12 virtual-interrupt-delivery enabled. |
7748 |
++ * However, it can be enabled only when L1 also |
7749 |
++ * intercepts external-interrupts and in that case |
7750 |
++ * we should not update vmcs02 RVI but instead intercept |
7751 |
++ * interrupt. Therefore, do nothing when running L2. |
7752 |
++ */ |
7753 |
++ if (!is_guest_mode(vcpu)) |
7754 |
++ vmx_set_rvi(max_irr); |
7755 |
++} |
7756 |
++ |
7757 |
++static int vmx_sync_pir_to_irr(struct kvm_vcpu *vcpu) |
7758 |
++{ |
7759 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7760 |
++ int max_irr; |
7761 |
++ bool max_irr_updated; |
7762 |
++ |
7763 |
++ WARN_ON(!vcpu->arch.apicv_active); |
7764 |
++ if (pi_test_on(&vmx->pi_desc)) { |
7765 |
++ pi_clear_on(&vmx->pi_desc); |
7766 |
++ /* |
7767 |
++ * IOMMU can write to PID.ON, so the barrier matters even on UP. |
7768 |
++ * But on x86 this is just a compiler barrier anyway. |
7769 |
++ */ |
7770 |
++ smp_mb__after_atomic(); |
7771 |
++ max_irr_updated = |
7772 |
++ kvm_apic_update_irr(vcpu, vmx->pi_desc.pir, &max_irr); |
7773 |
++ |
7774 |
++ /* |
7775 |
++ * If we are running L2 and L1 has a new pending interrupt |
7776 |
++ * which can be injected, we should re-evaluate |
7777 |
++ * what should be done with this new L1 interrupt. |
7778 |
++ * If L1 intercepts external-interrupts, we should |
7779 |
++ * exit from L2 to L1. Otherwise, interrupt should be |
7780 |
++ * delivered directly to L2. |
7781 |
++ */ |
7782 |
++ if (is_guest_mode(vcpu) && max_irr_updated) { |
7783 |
++ if (nested_exit_on_intr(vcpu)) |
7784 |
++ kvm_vcpu_exiting_guest_mode(vcpu); |
7785 |
++ else |
7786 |
++ kvm_make_request(KVM_REQ_EVENT, vcpu); |
7787 |
++ } |
7788 |
++ } else { |
7789 |
++ max_irr = kvm_lapic_find_highest_irr(vcpu); |
7790 |
++ } |
7791 |
++ vmx_hwapic_irr_update(vcpu, max_irr); |
7792 |
++ return max_irr; |
7793 |
++} |
7794 |
++ |
7795 |
++static bool vmx_dy_apicv_has_pending_interrupt(struct kvm_vcpu *vcpu) |
7796 |
++{ |
7797 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
7798 |
++ |
7799 |
++ return pi_test_on(pi_desc) || |
7800 |
++ (pi_test_sn(pi_desc) && !pi_is_pir_empty(pi_desc)); |
7801 |
++} |
7802 |
++ |
7803 |
++static void vmx_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap) |
7804 |
++{ |
7805 |
++ if (!kvm_vcpu_apicv_active(vcpu)) |
7806 |
++ return; |
7807 |
++ |
7808 |
++ vmcs_write64(EOI_EXIT_BITMAP0, eoi_exit_bitmap[0]); |
7809 |
++ vmcs_write64(EOI_EXIT_BITMAP1, eoi_exit_bitmap[1]); |
7810 |
++ vmcs_write64(EOI_EXIT_BITMAP2, eoi_exit_bitmap[2]); |
7811 |
++ vmcs_write64(EOI_EXIT_BITMAP3, eoi_exit_bitmap[3]); |
7812 |
++} |
7813 |
++ |
7814 |
++static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu) |
7815 |
++{ |
7816 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7817 |
++ |
7818 |
++ pi_clear_on(&vmx->pi_desc); |
7819 |
++ memset(vmx->pi_desc.pir, 0, sizeof(vmx->pi_desc.pir)); |
7820 |
++} |
7821 |
++ |
7822 |
++static void handle_exception_nmi_irqoff(struct vcpu_vmx *vmx) |
7823 |
++{ |
7824 |
++ vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7825 |
++ |
7826 |
++ /* if exit due to PF check for async PF */ |
7827 |
++ if (is_page_fault(vmx->exit_intr_info)) |
7828 |
++ vmx->vcpu.arch.apf.host_apf_reason = kvm_read_and_reset_pf_reason(); |
7829 |
++ |
7830 |
++ /* Handle machine checks before interrupts are enabled */ |
7831 |
++ if (is_machine_check(vmx->exit_intr_info)) |
7832 |
++ kvm_machine_check(); |
7833 |
++ |
7834 |
++ /* We need to handle NMIs before interrupts are enabled */ |
7835 |
++ if (is_nmi(vmx->exit_intr_info)) { |
7836 |
++ kvm_before_interrupt(&vmx->vcpu); |
7837 |
++ asm("int $2"); |
7838 |
++ kvm_after_interrupt(&vmx->vcpu); |
7839 |
++ } |
7840 |
++} |
7841 |
++ |
7842 |
++static void handle_external_interrupt_irqoff(struct kvm_vcpu *vcpu) |
7843 |
++{ |
7844 |
++ unsigned int vector; |
7845 |
++ unsigned long entry; |
7846 |
++#ifdef CONFIG_X86_64 |
7847 |
++ unsigned long tmp; |
7848 |
++#endif |
7849 |
++ gate_desc *desc; |
7850 |
++ u32 intr_info; |
7851 |
++ |
7852 |
++ intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7853 |
++ if (WARN_ONCE(!is_external_intr(intr_info), |
7854 |
++ "KVM: unexpected VM-Exit interrupt info: 0x%x", intr_info)) |
7855 |
++ return; |
7856 |
++ |
7857 |
++ vector = intr_info & INTR_INFO_VECTOR_MASK; |
7858 |
++ desc = (gate_desc *)host_idt_base + vector; |
7859 |
++ entry = gate_offset(desc); |
7860 |
++ |
7861 |
++ kvm_before_interrupt(vcpu); |
7862 |
++ |
7863 |
++ asm volatile( |
7864 |
++#ifdef CONFIG_X86_64 |
7865 |
++ "mov %%" _ASM_SP ", %[sp]\n\t" |
7866 |
++ "and $0xfffffffffffffff0, %%" _ASM_SP "\n\t" |
7867 |
++ "push $%c[ss]\n\t" |
7868 |
++ "push %[sp]\n\t" |
7869 |
++#endif |
7870 |
++ "pushf\n\t" |
7871 |
++ __ASM_SIZE(push) " $%c[cs]\n\t" |
7872 |
++ CALL_NOSPEC |
7873 |
++ : |
7874 |
++#ifdef CONFIG_X86_64 |
7875 |
++ [sp]"=&r"(tmp), |
7876 |
++#endif |
7877 |
++ ASM_CALL_CONSTRAINT |
7878 |
++ : |
7879 |
++ THUNK_TARGET(entry), |
7880 |
++ [ss]"i"(__KERNEL_DS), |
7881 |
++ [cs]"i"(__KERNEL_CS) |
7882 |
++ ); |
7883 |
++ |
7884 |
++ kvm_after_interrupt(vcpu); |
7885 |
++} |
7886 |
++STACK_FRAME_NON_STANDARD(handle_external_interrupt_irqoff); |
7887 |
++ |
7888 |
++static void vmx_handle_exit_irqoff(struct kvm_vcpu *vcpu, |
7889 |
++ enum exit_fastpath_completion *exit_fastpath) |
7890 |
++{ |
7891 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
7892 |
++ |
7893 |
++ if (vmx->exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT) |
7894 |
++ handle_external_interrupt_irqoff(vcpu); |
7895 |
++ else if (vmx->exit_reason == EXIT_REASON_EXCEPTION_NMI) |
7896 |
++ handle_exception_nmi_irqoff(vmx); |
7897 |
++ else if (!is_guest_mode(vcpu) && |
7898 |
++ vmx->exit_reason == EXIT_REASON_MSR_WRITE) |
7899 |
++ *exit_fastpath = handle_fastpath_set_msr_irqoff(vcpu); |
7900 |
++} |
7901 |
++ |
7902 |
++static bool vmx_has_emulated_msr(int index) |
7903 |
++{ |
7904 |
++ switch (index) { |
7905 |
++ case MSR_IA32_SMBASE: |
7906 |
++ /* |
7907 |
++ * We cannot do SMM unless we can run the guest in big |
7908 |
++ * real mode. |
7909 |
++ */ |
7910 |
++ return enable_unrestricted_guest || emulate_invalid_guest_state; |
7911 |
++ case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
7912 |
++ return nested; |
7913 |
++ case MSR_AMD64_VIRT_SPEC_CTRL: |
7914 |
++ /* This is AMD only. */ |
7915 |
++ return false; |
7916 |
++ default: |
7917 |
++ return true; |
7918 |
++ } |
7919 |
++} |
7920 |
++ |
7921 |
++static bool vmx_pt_supported(void) |
7922 |
++{ |
7923 |
++ return pt_mode == PT_MODE_HOST_GUEST; |
7924 |
++} |
7925 |
++ |
7926 |
++static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx) |
7927 |
++{ |
7928 |
++ u32 exit_intr_info; |
7929 |
++ bool unblock_nmi; |
7930 |
++ u8 vector; |
7931 |
++ bool idtv_info_valid; |
7932 |
++ |
7933 |
++ idtv_info_valid = vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK; |
7934 |
++ |
7935 |
++ if (enable_vnmi) { |
7936 |
++ if (vmx->loaded_vmcs->nmi_known_unmasked) |
7937 |
++ return; |
7938 |
++ /* |
7939 |
++ * Can't use vmx->exit_intr_info since we're not sure what |
7940 |
++ * the exit reason is. |
7941 |
++ */ |
7942 |
++ exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7943 |
++ unblock_nmi = (exit_intr_info & INTR_INFO_UNBLOCK_NMI) != 0; |
7944 |
++ vector = exit_intr_info & INTR_INFO_VECTOR_MASK; |
7945 |
++ /* |
7946 |
++ * SDM 3: 27.7.1.2 (September 2008) |
7947 |
++ * Re-set bit "block by NMI" before VM entry if vmexit caused by |
7948 |
++ * a guest IRET fault. |
7949 |
++ * SDM 3: 23.2.2 (September 2008) |
7950 |
++ * Bit 12 is undefined in any of the following cases: |
7951 |
++ * If the VM exit sets the valid bit in the IDT-vectoring |
7952 |
++ * information field. |
7953 |
++ * If the VM exit is due to a double fault. |
7954 |
++ */ |
7955 |
++ if ((exit_intr_info & INTR_INFO_VALID_MASK) && unblock_nmi && |
7956 |
++ vector != DF_VECTOR && !idtv_info_valid) |
7957 |
++ vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
7958 |
++ GUEST_INTR_STATE_NMI); |
7959 |
++ else |
7960 |
++ vmx->loaded_vmcs->nmi_known_unmasked = |
7961 |
++ !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) |
7962 |
++ & GUEST_INTR_STATE_NMI); |
7963 |
++ } else if (unlikely(vmx->loaded_vmcs->soft_vnmi_blocked)) |
7964 |
++ vmx->loaded_vmcs->vnmi_blocked_time += |
7965 |
++ ktime_to_ns(ktime_sub(ktime_get(), |
7966 |
++ vmx->loaded_vmcs->entry_time)); |
7967 |
++} |
7968 |
++ |
7969 |
++static void __vmx_complete_interrupts(struct kvm_vcpu *vcpu, |
7970 |
++ u32 idt_vectoring_info, |
7971 |
++ int instr_len_field, |
7972 |
++ int error_code_field) |
7973 |
++{ |
7974 |
++ u8 vector; |
7975 |
++ int type; |
7976 |
++ bool idtv_info_valid; |
7977 |
++ |
7978 |
++ idtv_info_valid = idt_vectoring_info & VECTORING_INFO_VALID_MASK; |
7979 |
++ |
7980 |
++ vcpu->arch.nmi_injected = false; |
7981 |
++ kvm_clear_exception_queue(vcpu); |
7982 |
++ kvm_clear_interrupt_queue(vcpu); |
7983 |
++ |
7984 |
++ if (!idtv_info_valid) |
7985 |
++ return; |
7986 |
++ |
7987 |
++ kvm_make_request(KVM_REQ_EVENT, vcpu); |
7988 |
++ |
7989 |
++ vector = idt_vectoring_info & VECTORING_INFO_VECTOR_MASK; |
7990 |
++ type = idt_vectoring_info & VECTORING_INFO_TYPE_MASK; |
7991 |
++ |
7992 |
++ switch (type) { |
7993 |
++ case INTR_TYPE_NMI_INTR: |
7994 |
++ vcpu->arch.nmi_injected = true; |
7995 |
++ /* |
7996 |
++ * SDM 3: 27.7.1.2 (September 2008) |
7997 |
++ * Clear bit "block by NMI" before VM entry if a NMI |
7998 |
++ * delivery faulted. |
7999 |
++ */ |
8000 |
++ vmx_set_nmi_mask(vcpu, false); |
8001 |
++ break; |
8002 |
++ case INTR_TYPE_SOFT_EXCEPTION: |
8003 |
++ vcpu->arch.event_exit_inst_len = vmcs_read32(instr_len_field); |
8004 |
++ /* fall through */ |
8005 |
++ case INTR_TYPE_HARD_EXCEPTION: |
8006 |
++ if (idt_vectoring_info & VECTORING_INFO_DELIVER_CODE_MASK) { |
8007 |
++ u32 err = vmcs_read32(error_code_field); |
8008 |
++ kvm_requeue_exception_e(vcpu, vector, err); |
8009 |
++ } else |
8010 |
++ kvm_requeue_exception(vcpu, vector); |
8011 |
++ break; |
8012 |
++ case INTR_TYPE_SOFT_INTR: |
8013 |
++ vcpu->arch.event_exit_inst_len = vmcs_read32(instr_len_field); |
8014 |
++ /* fall through */ |
8015 |
++ case INTR_TYPE_EXT_INTR: |
8016 |
++ kvm_queue_interrupt(vcpu, vector, type == INTR_TYPE_SOFT_INTR); |
8017 |
++ break; |
8018 |
++ default: |
8019 |
++ break; |
8020 |
++ } |
8021 |
++} |
8022 |
++ |
8023 |
++static void vmx_complete_interrupts(struct vcpu_vmx *vmx) |
8024 |
++{ |
8025 |
++ __vmx_complete_interrupts(&vmx->vcpu, vmx->idt_vectoring_info, |
8026 |
++ VM_EXIT_INSTRUCTION_LEN, |
8027 |
++ IDT_VECTORING_ERROR_CODE); |
8028 |
++} |
8029 |
++ |
8030 |
++static void vmx_cancel_injection(struct kvm_vcpu *vcpu) |
8031 |
++{ |
8032 |
++ __vmx_complete_interrupts(vcpu, |
8033 |
++ vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), |
8034 |
++ VM_ENTRY_INSTRUCTION_LEN, |
8035 |
++ VM_ENTRY_EXCEPTION_ERROR_CODE); |
8036 |
++ |
8037 |
++ vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); |
8038 |
++} |
8039 |
++ |
8040 |
++static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) |
8041 |
++{ |
8042 |
++ int i, nr_msrs; |
8043 |
++ struct perf_guest_switch_msr *msrs; |
8044 |
++ |
8045 |
++ msrs = perf_guest_get_msrs(&nr_msrs); |
8046 |
++ |
8047 |
++ if (!msrs) |
8048 |
++ return; |
8049 |
++ |
8050 |
++ for (i = 0; i < nr_msrs; i++) |
8051 |
++ if (msrs[i].host == msrs[i].guest) |
8052 |
++ clear_atomic_switch_msr(vmx, msrs[i].msr); |
8053 |
++ else |
8054 |
++ add_atomic_switch_msr(vmx, msrs[i].msr, msrs[i].guest, |
8055 |
++ msrs[i].host, false); |
8056 |
++} |
8057 |
++ |
8058 |
++static void atomic_switch_umwait_control_msr(struct vcpu_vmx *vmx) |
8059 |
++{ |
8060 |
++ u32 host_umwait_control; |
8061 |
++ |
8062 |
++ if (!vmx_has_waitpkg(vmx)) |
8063 |
++ return; |
8064 |
++ |
8065 |
++ host_umwait_control = get_umwait_control_msr(); |
8066 |
++ |
8067 |
++ if (vmx->msr_ia32_umwait_control != host_umwait_control) |
8068 |
++ add_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL, |
8069 |
++ vmx->msr_ia32_umwait_control, |
8070 |
++ host_umwait_control, false); |
8071 |
++ else |
8072 |
++ clear_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL); |
8073 |
++} |
8074 |
++ |
8075 |
++static void vmx_update_hv_timer(struct kvm_vcpu *vcpu) |
8076 |
++{ |
8077 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8078 |
++ u64 tscl; |
8079 |
++ u32 delta_tsc; |
8080 |
++ |
8081 |
++ if (vmx->req_immediate_exit) { |
8082 |
++ vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, 0); |
8083 |
++ vmx->loaded_vmcs->hv_timer_soft_disabled = false; |
8084 |
++ } else if (vmx->hv_deadline_tsc != -1) { |
8085 |
++ tscl = rdtsc(); |
8086 |
++ if (vmx->hv_deadline_tsc > tscl) |
8087 |
++ /* set_hv_timer ensures the delta fits in 32-bits */ |
8088 |
++ delta_tsc = (u32)((vmx->hv_deadline_tsc - tscl) >> |
8089 |
++ cpu_preemption_timer_multi); |
8090 |
++ else |
8091 |
++ delta_tsc = 0; |
8092 |
++ |
8093 |
++ vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, delta_tsc); |
8094 |
++ vmx->loaded_vmcs->hv_timer_soft_disabled = false; |
8095 |
++ } else if (!vmx->loaded_vmcs->hv_timer_soft_disabled) { |
8096 |
++ vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, -1); |
8097 |
++ vmx->loaded_vmcs->hv_timer_soft_disabled = true; |
8098 |
++ } |
8099 |
++} |
8100 |
++ |
8101 |
++void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp) |
8102 |
++{ |
8103 |
++ if (unlikely(host_rsp != vmx->loaded_vmcs->host_state.rsp)) { |
8104 |
++ vmx->loaded_vmcs->host_state.rsp = host_rsp; |
8105 |
++ vmcs_writel(HOST_RSP, host_rsp); |
8106 |
++ } |
8107 |
++} |
8108 |
++ |
8109 |
++bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs, bool launched); |
8110 |
++ |
8111 |
++static void vmx_vcpu_run(struct kvm_vcpu *vcpu) |
8112 |
++{ |
8113 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8114 |
++ unsigned long cr3, cr4; |
8115 |
++ |
8116 |
++ /* Record the guest's net vcpu time for enforced NMI injections. */ |
8117 |
++ if (unlikely(!enable_vnmi && |
8118 |
++ vmx->loaded_vmcs->soft_vnmi_blocked)) |
8119 |
++ vmx->loaded_vmcs->entry_time = ktime_get(); |
8120 |
++ |
8121 |
++ /* Don't enter VMX if guest state is invalid, let the exit handler |
8122 |
++ start emulation until we arrive back to a valid state */ |
8123 |
++ if (vmx->emulation_required) |
8124 |
++ return; |
8125 |
++ |
8126 |
++ if (vmx->ple_window_dirty) { |
8127 |
++ vmx->ple_window_dirty = false; |
8128 |
++ vmcs_write32(PLE_WINDOW, vmx->ple_window); |
8129 |
++ } |
8130 |
++ |
8131 |
++ if (vmx->nested.need_vmcs12_to_shadow_sync) |
8132 |
++ nested_sync_vmcs12_to_shadow(vcpu); |
8133 |
++ |
8134 |
++ if (kvm_register_is_dirty(vcpu, VCPU_REGS_RSP)) |
8135 |
++ vmcs_writel(GUEST_RSP, vcpu->arch.regs[VCPU_REGS_RSP]); |
8136 |
++ if (kvm_register_is_dirty(vcpu, VCPU_REGS_RIP)) |
8137 |
++ vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]); |
8138 |
++ |
8139 |
++ cr3 = __get_current_cr3_fast(); |
8140 |
++ if (unlikely(cr3 != vmx->loaded_vmcs->host_state.cr3)) { |
8141 |
++ vmcs_writel(HOST_CR3, cr3); |
8142 |
++ vmx->loaded_vmcs->host_state.cr3 = cr3; |
8143 |
++ } |
8144 |
++ |
8145 |
++ cr4 = cr4_read_shadow(); |
8146 |
++ if (unlikely(cr4 != vmx->loaded_vmcs->host_state.cr4)) { |
8147 |
++ vmcs_writel(HOST_CR4, cr4); |
8148 |
++ vmx->loaded_vmcs->host_state.cr4 = cr4; |
8149 |
++ } |
8150 |
++ |
8151 |
++ /* When single-stepping over STI and MOV SS, we must clear the |
8152 |
++ * corresponding interruptibility bits in the guest state. Otherwise |
8153 |
++ * vmentry fails as it then expects bit 14 (BS) in pending debug |
8154 |
++ * exceptions being set, but that's not correct for the guest debugging |
8155 |
++ * case. */ |
8156 |
++ if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) |
8157 |
++ vmx_set_interrupt_shadow(vcpu, 0); |
8158 |
++ |
8159 |
++ kvm_load_guest_xsave_state(vcpu); |
8160 |
++ |
8161 |
++ if (static_cpu_has(X86_FEATURE_PKU) && |
8162 |
++ kvm_read_cr4_bits(vcpu, X86_CR4_PKE) && |
8163 |
++ vcpu->arch.pkru != vmx->host_pkru) |
8164 |
++ __write_pkru(vcpu->arch.pkru); |
8165 |
++ |
8166 |
++ pt_guest_enter(vmx); |
8167 |
++ |
8168 |
++ atomic_switch_perf_msrs(vmx); |
8169 |
++ atomic_switch_umwait_control_msr(vmx); |
8170 |
++ |
8171 |
++ if (enable_preemption_timer) |
8172 |
++ vmx_update_hv_timer(vcpu); |
8173 |
++ |
8174 |
++ if (lapic_in_kernel(vcpu) && |
8175 |
++ vcpu->arch.apic->lapic_timer.timer_advance_ns) |
8176 |
++ kvm_wait_lapic_expire(vcpu); |
8177 |
++ |
8178 |
++ /* |
8179 |
++ * If this vCPU has touched SPEC_CTRL, restore the guest's value if |
8180 |
++ * it's non-zero. Since vmentry is serialising on affected CPUs, there |
8181 |
++ * is no need to worry about the conditional branch over the wrmsr |
8182 |
++ * being speculatively taken. |
8183 |
++ */ |
8184 |
++ x86_spec_ctrl_set_guest(vmx->spec_ctrl, 0); |
8185 |
++ |
8186 |
++ /* L1D Flush includes CPU buffer clear to mitigate MDS */ |
8187 |
++ if (static_branch_unlikely(&vmx_l1d_should_flush)) |
8188 |
++ vmx_l1d_flush(vcpu); |
8189 |
++ else if (static_branch_unlikely(&mds_user_clear)) |
8190 |
++ mds_clear_cpu_buffers(); |
8191 |
++ |
8192 |
++ if (vcpu->arch.cr2 != read_cr2()) |
8193 |
++ write_cr2(vcpu->arch.cr2); |
8194 |
++ |
8195 |
++ vmx->fail = __vmx_vcpu_run(vmx, (unsigned long *)&vcpu->arch.regs, |
8196 |
++ vmx->loaded_vmcs->launched); |
8197 |
++ |
8198 |
++ vcpu->arch.cr2 = read_cr2(); |
8199 |
++ |
8200 |
++ /* |
8201 |
++ * We do not use IBRS in the kernel. If this vCPU has used the |
8202 |
++ * SPEC_CTRL MSR it may have left it on; save the value and |
8203 |
++ * turn it off. This is much more efficient than blindly adding |
8204 |
++ * it to the atomic save/restore list. Especially as the former |
8205 |
++ * (Saving guest MSRs on vmexit) doesn't even exist in KVM. |
8206 |
++ * |
8207 |
++ * For non-nested case: |
8208 |
++ * If the L01 MSR bitmap does not intercept the MSR, then we need to |
8209 |
++ * save it. |
8210 |
++ * |
8211 |
++ * For nested case: |
8212 |
++ * If the L02 MSR bitmap does not intercept the MSR, then we need to |
8213 |
++ * save it. |
8214 |
++ */ |
8215 |
++ if (unlikely(!msr_write_intercepted(vcpu, MSR_IA32_SPEC_CTRL))) |
8216 |
++ vmx->spec_ctrl = native_read_msr(MSR_IA32_SPEC_CTRL); |
8217 |
++ |
8218 |
++ x86_spec_ctrl_restore_host(vmx->spec_ctrl, 0); |
8219 |
++ |
8220 |
++ /* All fields are clean at this point */ |
8221 |
++ if (static_branch_unlikely(&enable_evmcs)) |
8222 |
++ current_evmcs->hv_clean_fields |= |
8223 |
++ HV_VMX_ENLIGHTENED_CLEAN_FIELD_ALL; |
8224 |
++ |
8225 |
++ if (static_branch_unlikely(&enable_evmcs)) |
8226 |
++ current_evmcs->hv_vp_id = vcpu->arch.hyperv.vp_index; |
8227 |
++ |
8228 |
++ /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */ |
8229 |
++ if (vmx->host_debugctlmsr) |
8230 |
++ update_debugctlmsr(vmx->host_debugctlmsr); |
8231 |
++ |
8232 |
++#ifndef CONFIG_X86_64 |
8233 |
++ /* |
8234 |
++ * The sysexit path does not restore ds/es, so we must set them to |
8235 |
++ * a reasonable value ourselves. |
8236 |
++ * |
8237 |
++ * We can't defer this to vmx_prepare_switch_to_host() since that |
8238 |
++ * function may be executed in interrupt context, which saves and |
8239 |
++ * restore segments around it, nullifying its effect. |
8240 |
++ */ |
8241 |
++ loadsegment(ds, __USER_DS); |
8242 |
++ loadsegment(es, __USER_DS); |
8243 |
++#endif |
8244 |
++ |
8245 |
++ vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP) |
8246 |
++ | (1 << VCPU_EXREG_RFLAGS) |
8247 |
++ | (1 << VCPU_EXREG_PDPTR) |
8248 |
++ | (1 << VCPU_EXREG_SEGMENTS) |
8249 |
++ | (1 << VCPU_EXREG_CR3)); |
8250 |
++ vcpu->arch.regs_dirty = 0; |
8251 |
++ |
8252 |
++ pt_guest_exit(vmx); |
8253 |
++ |
8254 |
++ /* |
8255 |
++ * eager fpu is enabled if PKEY is supported and CR4 is switched |
8256 |
++ * back on host, so it is safe to read guest PKRU from current |
8257 |
++ * XSAVE. |
8258 |
++ */ |
8259 |
++ if (static_cpu_has(X86_FEATURE_PKU) && |
8260 |
++ kvm_read_cr4_bits(vcpu, X86_CR4_PKE)) { |
8261 |
++ vcpu->arch.pkru = rdpkru(); |
8262 |
++ if (vcpu->arch.pkru != vmx->host_pkru) |
8263 |
++ __write_pkru(vmx->host_pkru); |
8264 |
++ } |
8265 |
++ |
8266 |
++ kvm_load_host_xsave_state(vcpu); |
8267 |
++ |
8268 |
++ vmx->nested.nested_run_pending = 0; |
8269 |
++ vmx->idt_vectoring_info = 0; |
8270 |
++ |
8271 |
++ vmx->exit_reason = vmx->fail ? 0xdead : vmcs_read32(VM_EXIT_REASON); |
8272 |
++ if ((u16)vmx->exit_reason == EXIT_REASON_MCE_DURING_VMENTRY) |
8273 |
++ kvm_machine_check(); |
8274 |
++ |
8275 |
++ if (vmx->fail || (vmx->exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY)) |
8276 |
++ return; |
8277 |
++ |
8278 |
++ vmx->loaded_vmcs->launched = 1; |
8279 |
++ vmx->idt_vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD); |
8280 |
++ |
8281 |
++ vmx_recover_nmi_blocking(vmx); |
8282 |
++ vmx_complete_interrupts(vmx); |
8283 |
++} |
8284 |
++ |
8285 |
++static struct kvm *vmx_vm_alloc(void) |
8286 |
++{ |
8287 |
++ struct kvm_vmx *kvm_vmx = __vmalloc(sizeof(struct kvm_vmx), |
8288 |
++ GFP_KERNEL_ACCOUNT | __GFP_ZERO, |
8289 |
++ PAGE_KERNEL); |
8290 |
++ return &kvm_vmx->kvm; |
8291 |
++} |
8292 |
++ |
8293 |
++static void vmx_vm_free(struct kvm *kvm) |
8294 |
++{ |
8295 |
++ kfree(kvm->arch.hyperv.hv_pa_pg); |
8296 |
++ vfree(to_kvm_vmx(kvm)); |
8297 |
++} |
8298 |
++ |
8299 |
++static void vmx_free_vcpu(struct kvm_vcpu *vcpu) |
8300 |
++{ |
8301 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8302 |
++ |
8303 |
++ if (enable_pml) |
8304 |
++ vmx_destroy_pml_buffer(vmx); |
8305 |
++ free_vpid(vmx->vpid); |
8306 |
++ nested_vmx_free_vcpu(vcpu); |
8307 |
++ free_loaded_vmcs(vmx->loaded_vmcs); |
8308 |
++ kvm_vcpu_uninit(vcpu); |
8309 |
++ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu); |
8310 |
++ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu); |
8311 |
++ kmem_cache_free(kvm_vcpu_cache, vmx); |
8312 |
++} |
8313 |
++ |
8314 |
++static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id) |
8315 |
++{ |
8316 |
++ int err; |
8317 |
++ struct vcpu_vmx *vmx; |
8318 |
++ unsigned long *msr_bitmap; |
8319 |
++ int i, cpu; |
8320 |
++ |
8321 |
++ BUILD_BUG_ON_MSG(offsetof(struct vcpu_vmx, vcpu) != 0, |
8322 |
++ "struct kvm_vcpu must be at offset 0 for arch usercopy region"); |
8323 |
++ |
8324 |
++ vmx = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL_ACCOUNT); |
8325 |
++ if (!vmx) |
8326 |
++ return ERR_PTR(-ENOMEM); |
8327 |
++ |
8328 |
++ vmx->vcpu.arch.user_fpu = kmem_cache_zalloc(x86_fpu_cache, |
8329 |
++ GFP_KERNEL_ACCOUNT); |
8330 |
++ if (!vmx->vcpu.arch.user_fpu) { |
8331 |
++ printk(KERN_ERR "kvm: failed to allocate kvm userspace's fpu\n"); |
8332 |
++ err = -ENOMEM; |
8333 |
++ goto free_partial_vcpu; |
8334 |
++ } |
8335 |
++ |
8336 |
++ vmx->vcpu.arch.guest_fpu = kmem_cache_zalloc(x86_fpu_cache, |
8337 |
++ GFP_KERNEL_ACCOUNT); |
8338 |
++ if (!vmx->vcpu.arch.guest_fpu) { |
8339 |
++ printk(KERN_ERR "kvm: failed to allocate vcpu's fpu\n"); |
8340 |
++ err = -ENOMEM; |
8341 |
++ goto free_user_fpu; |
8342 |
++ } |
8343 |
++ |
8344 |
++ vmx->vpid = allocate_vpid(); |
8345 |
++ |
8346 |
++ err = kvm_vcpu_init(&vmx->vcpu, kvm, id); |
8347 |
++ if (err) |
8348 |
++ goto free_vcpu; |
8349 |
++ |
8350 |
++ err = -ENOMEM; |
8351 |
++ |
8352 |
++ /* |
8353 |
++ * If PML is turned on, failure on enabling PML just results in failure |
8354 |
++ * of creating the vcpu, therefore we can simplify PML logic (by |
8355 |
++ * avoiding dealing with cases, such as enabling PML partially on vcpus |
8356 |
++ * for the guest), etc. |
8357 |
++ */ |
8358 |
++ if (enable_pml) { |
8359 |
++ vmx->pml_pg = alloc_page(GFP_KERNEL_ACCOUNT | __GFP_ZERO); |
8360 |
++ if (!vmx->pml_pg) |
8361 |
++ goto uninit_vcpu; |
8362 |
++ } |
8363 |
++ |
8364 |
++ BUILD_BUG_ON(ARRAY_SIZE(vmx_msr_index) != NR_SHARED_MSRS); |
8365 |
++ |
8366 |
++ for (i = 0; i < ARRAY_SIZE(vmx_msr_index); ++i) { |
8367 |
++ u32 index = vmx_msr_index[i]; |
8368 |
++ u32 data_low, data_high; |
8369 |
++ int j = vmx->nmsrs; |
8370 |
++ |
8371 |
++ if (rdmsr_safe(index, &data_low, &data_high) < 0) |
8372 |
++ continue; |
8373 |
++ if (wrmsr_safe(index, data_low, data_high) < 0) |
8374 |
++ continue; |
8375 |
++ |
8376 |
++ vmx->guest_msrs[j].index = i; |
8377 |
++ vmx->guest_msrs[j].data = 0; |
8378 |
++ switch (index) { |
8379 |
++ case MSR_IA32_TSX_CTRL: |
8380 |
++ /* |
8381 |
++ * No need to pass TSX_CTRL_CPUID_CLEAR through, so |
8382 |
++ * let's avoid changing CPUID bits under the host |
8383 |
++ * kernel's feet. |
8384 |
++ */ |
8385 |
++ vmx->guest_msrs[j].mask = ~(u64)TSX_CTRL_CPUID_CLEAR; |
8386 |
++ break; |
8387 |
++ default: |
8388 |
++ vmx->guest_msrs[j].mask = -1ull; |
8389 |
++ break; |
8390 |
++ } |
8391 |
++ ++vmx->nmsrs; |
8392 |
++ } |
8393 |
++ |
8394 |
++ err = alloc_loaded_vmcs(&vmx->vmcs01); |
8395 |
++ if (err < 0) |
8396 |
++ goto free_pml; |
8397 |
++ |
8398 |
++ msr_bitmap = vmx->vmcs01.msr_bitmap; |
8399 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_TSC, MSR_TYPE_R); |
8400 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_FS_BASE, MSR_TYPE_RW); |
8401 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_GS_BASE, MSR_TYPE_RW); |
8402 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_KERNEL_GS_BASE, MSR_TYPE_RW); |
8403 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_CS, MSR_TYPE_RW); |
8404 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_ESP, MSR_TYPE_RW); |
8405 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_EIP, MSR_TYPE_RW); |
8406 |
++ if (kvm_cstate_in_guest(kvm)) { |
8407 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C1_RES, MSR_TYPE_R); |
8408 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C3_RESIDENCY, MSR_TYPE_R); |
8409 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C6_RESIDENCY, MSR_TYPE_R); |
8410 |
++ vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C7_RESIDENCY, MSR_TYPE_R); |
8411 |
++ } |
8412 |
++ vmx->msr_bitmap_mode = 0; |
8413 |
++ |
8414 |
++ vmx->loaded_vmcs = &vmx->vmcs01; |
8415 |
++ cpu = get_cpu(); |
8416 |
++ vmx_vcpu_load(&vmx->vcpu, cpu); |
8417 |
++ vmx->vcpu.cpu = cpu; |
8418 |
++ init_vmcs(vmx); |
8419 |
++ vmx_vcpu_put(&vmx->vcpu); |
8420 |
++ put_cpu(); |
8421 |
++ if (cpu_need_virtualize_apic_accesses(&vmx->vcpu)) { |
8422 |
++ err = alloc_apic_access_page(kvm); |
8423 |
++ if (err) |
8424 |
++ goto free_vmcs; |
8425 |
++ } |
8426 |
++ |
8427 |
++ if (enable_ept && !enable_unrestricted_guest) { |
8428 |
++ err = init_rmode_identity_map(kvm); |
8429 |
++ if (err) |
8430 |
++ goto free_vmcs; |
8431 |
++ } |
8432 |
++ |
8433 |
++ if (nested) |
8434 |
++ nested_vmx_setup_ctls_msrs(&vmx->nested.msrs, |
8435 |
++ vmx_capability.ept, |
8436 |
++ kvm_vcpu_apicv_active(&vmx->vcpu)); |
8437 |
++ else |
8438 |
++ memset(&vmx->nested.msrs, 0, sizeof(vmx->nested.msrs)); |
8439 |
++ |
8440 |
++ vmx->nested.posted_intr_nv = -1; |
8441 |
++ vmx->nested.current_vmptr = -1ull; |
8442 |
++ |
8443 |
++ vmx->msr_ia32_feature_control_valid_bits = FEATURE_CONTROL_LOCKED; |
8444 |
++ |
8445 |
++ /* |
8446 |
++ * Enforce invariant: pi_desc.nv is always either POSTED_INTR_VECTOR |
8447 |
++ * or POSTED_INTR_WAKEUP_VECTOR. |
8448 |
++ */ |
8449 |
++ vmx->pi_desc.nv = POSTED_INTR_VECTOR; |
8450 |
++ vmx->pi_desc.sn = 1; |
8451 |
++ |
8452 |
++ vmx->ept_pointer = INVALID_PAGE; |
8453 |
++ |
8454 |
++ return &vmx->vcpu; |
8455 |
++ |
8456 |
++free_vmcs: |
8457 |
++ free_loaded_vmcs(vmx->loaded_vmcs); |
8458 |
++free_pml: |
8459 |
++ vmx_destroy_pml_buffer(vmx); |
8460 |
++uninit_vcpu: |
8461 |
++ kvm_vcpu_uninit(&vmx->vcpu); |
8462 |
++free_vcpu: |
8463 |
++ free_vpid(vmx->vpid); |
8464 |
++ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu); |
8465 |
++free_user_fpu: |
8466 |
++ kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu); |
8467 |
++free_partial_vcpu: |
8468 |
++ kmem_cache_free(kvm_vcpu_cache, vmx); |
8469 |
++ return ERR_PTR(err); |
8470 |
++} |
8471 |
++ |
8472 |
++#define L1TF_MSG_SMT "L1TF CPU bug present and SMT on, data leak possible. See CVE-2018-3646 and https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/l1tf.html for details.\n" |
8473 |
++#define L1TF_MSG_L1D "L1TF CPU bug present and virtualization mitigation disabled, data leak possible. See CVE-2018-3646 and https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/l1tf.html for details.\n" |
8474 |
++ |
8475 |
++static int vmx_vm_init(struct kvm *kvm) |
8476 |
++{ |
8477 |
++ spin_lock_init(&to_kvm_vmx(kvm)->ept_pointer_lock); |
8478 |
++ |
8479 |
++ if (!ple_gap) |
8480 |
++ kvm->arch.pause_in_guest = true; |
8481 |
++ |
8482 |
++ if (boot_cpu_has(X86_BUG_L1TF) && enable_ept) { |
8483 |
++ switch (l1tf_mitigation) { |
8484 |
++ case L1TF_MITIGATION_OFF: |
8485 |
++ case L1TF_MITIGATION_FLUSH_NOWARN: |
8486 |
++ /* 'I explicitly don't care' is set */ |
8487 |
++ break; |
8488 |
++ case L1TF_MITIGATION_FLUSH: |
8489 |
++ case L1TF_MITIGATION_FLUSH_NOSMT: |
8490 |
++ case L1TF_MITIGATION_FULL: |
8491 |
++ /* |
8492 |
++ * Warn upon starting the first VM in a potentially |
8493 |
++ * insecure environment. |
8494 |
++ */ |
8495 |
++ if (sched_smt_active()) |
8496 |
++ pr_warn_once(L1TF_MSG_SMT); |
8497 |
++ if (l1tf_vmx_mitigation == VMENTER_L1D_FLUSH_NEVER) |
8498 |
++ pr_warn_once(L1TF_MSG_L1D); |
8499 |
++ break; |
8500 |
++ case L1TF_MITIGATION_FULL_FORCE: |
8501 |
++ /* Flush is enforced */ |
8502 |
++ break; |
8503 |
++ } |
8504 |
++ } |
8505 |
++ return 0; |
8506 |
++} |
8507 |
++ |
8508 |
++static int __init vmx_check_processor_compat(void) |
8509 |
++{ |
8510 |
++ struct vmcs_config vmcs_conf; |
8511 |
++ struct vmx_capability vmx_cap; |
8512 |
++ |
8513 |
++ if (setup_vmcs_config(&vmcs_conf, &vmx_cap) < 0) |
8514 |
++ return -EIO; |
8515 |
++ if (nested) |
8516 |
++ nested_vmx_setup_ctls_msrs(&vmcs_conf.nested, vmx_cap.ept, |
8517 |
++ enable_apicv); |
8518 |
++ if (memcmp(&vmcs_config, &vmcs_conf, sizeof(struct vmcs_config)) != 0) { |
8519 |
++ printk(KERN_ERR "kvm: CPU %d feature inconsistency!\n", |
8520 |
++ smp_processor_id()); |
8521 |
++ return -EIO; |
8522 |
++ } |
8523 |
++ return 0; |
8524 |
++} |
8525 |
++ |
8526 |
++static u64 vmx_get_mt_mask(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio) |
8527 |
++{ |
8528 |
++ u8 cache; |
8529 |
++ u64 ipat = 0; |
8530 |
++ |
8531 |
++ /* For VT-d and EPT combination |
8532 |
++ * 1. MMIO: always map as UC |
8533 |
++ * 2. EPT with VT-d: |
8534 |
++ * a. VT-d without snooping control feature: can't guarantee the |
8535 |
++ * result, try to trust guest. |
8536 |
++ * b. VT-d with snooping control feature: snooping control feature of |
8537 |
++ * VT-d engine can guarantee the cache correctness. Just set it |
8538 |
++ * to WB to keep consistent with host. So the same as item 3. |
8539 |
++ * 3. EPT without VT-d: always map as WB and set IPAT=1 to keep |
8540 |
++ * consistent with host MTRR |
8541 |
++ */ |
8542 |
++ if (is_mmio) { |
8543 |
++ cache = MTRR_TYPE_UNCACHABLE; |
8544 |
++ goto exit; |
8545 |
++ } |
8546 |
++ |
8547 |
++ if (!kvm_arch_has_noncoherent_dma(vcpu->kvm)) { |
8548 |
++ ipat = VMX_EPT_IPAT_BIT; |
8549 |
++ cache = MTRR_TYPE_WRBACK; |
8550 |
++ goto exit; |
8551 |
++ } |
8552 |
++ |
8553 |
++ if (kvm_read_cr0(vcpu) & X86_CR0_CD) { |
8554 |
++ ipat = VMX_EPT_IPAT_BIT; |
8555 |
++ if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED)) |
8556 |
++ cache = MTRR_TYPE_WRBACK; |
8557 |
++ else |
8558 |
++ cache = MTRR_TYPE_UNCACHABLE; |
8559 |
++ goto exit; |
8560 |
++ } |
8561 |
++ |
8562 |
++ cache = kvm_mtrr_get_guest_memory_type(vcpu, gfn); |
8563 |
++ |
8564 |
++exit: |
8565 |
++ return (cache << VMX_EPT_MT_EPTE_SHIFT) | ipat; |
8566 |
++} |
8567 |
++ |
8568 |
++static int vmx_get_lpage_level(void) |
8569 |
++{ |
8570 |
++ if (enable_ept && !cpu_has_vmx_ept_1g_page()) |
8571 |
++ return PT_DIRECTORY_LEVEL; |
8572 |
++ else |
8573 |
++ /* For shadow and EPT supported 1GB page */ |
8574 |
++ return PT_PDPE_LEVEL; |
8575 |
++} |
8576 |
++ |
8577 |
++static void vmcs_set_secondary_exec_control(struct vcpu_vmx *vmx) |
8578 |
++{ |
8579 |
++ /* |
8580 |
++ * These bits in the secondary execution controls field |
8581 |
++ * are dynamic, the others are mostly based on the hypervisor |
8582 |
++ * architecture and the guest's CPUID. Do not touch the |
8583 |
++ * dynamic bits. |
8584 |
++ */ |
8585 |
++ u32 mask = |
8586 |
++ SECONDARY_EXEC_SHADOW_VMCS | |
8587 |
++ SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
8588 |
++ SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
8589 |
++ SECONDARY_EXEC_DESC; |
8590 |
++ |
8591 |
++ u32 new_ctl = vmx->secondary_exec_control; |
8592 |
++ u32 cur_ctl = secondary_exec_controls_get(vmx); |
8593 |
++ |
8594 |
++ secondary_exec_controls_set(vmx, (new_ctl & ~mask) | (cur_ctl & mask)); |
8595 |
++} |
8596 |
++ |
8597 |
++/* |
8598 |
++ * Generate MSR_IA32_VMX_CR{0,4}_FIXED1 according to CPUID. Only set bits |
8599 |
++ * (indicating "allowed-1") if they are supported in the guest's CPUID. |
8600 |
++ */ |
8601 |
++static void nested_vmx_cr_fixed1_bits_update(struct kvm_vcpu *vcpu) |
8602 |
++{ |
8603 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8604 |
++ struct kvm_cpuid_entry2 *entry; |
8605 |
++ |
8606 |
++ vmx->nested.msrs.cr0_fixed1 = 0xffffffff; |
8607 |
++ vmx->nested.msrs.cr4_fixed1 = X86_CR4_PCE; |
8608 |
++ |
8609 |
++#define cr4_fixed1_update(_cr4_mask, _reg, _cpuid_mask) do { \ |
8610 |
++ if (entry && (entry->_reg & (_cpuid_mask))) \ |
8611 |
++ vmx->nested.msrs.cr4_fixed1 |= (_cr4_mask); \ |
8612 |
++} while (0) |
8613 |
++ |
8614 |
++ entry = kvm_find_cpuid_entry(vcpu, 0x1, 0); |
8615 |
++ cr4_fixed1_update(X86_CR4_VME, edx, bit(X86_FEATURE_VME)); |
8616 |
++ cr4_fixed1_update(X86_CR4_PVI, edx, bit(X86_FEATURE_VME)); |
8617 |
++ cr4_fixed1_update(X86_CR4_TSD, edx, bit(X86_FEATURE_TSC)); |
8618 |
++ cr4_fixed1_update(X86_CR4_DE, edx, bit(X86_FEATURE_DE)); |
8619 |
++ cr4_fixed1_update(X86_CR4_PSE, edx, bit(X86_FEATURE_PSE)); |
8620 |
++ cr4_fixed1_update(X86_CR4_PAE, edx, bit(X86_FEATURE_PAE)); |
8621 |
++ cr4_fixed1_update(X86_CR4_MCE, edx, bit(X86_FEATURE_MCE)); |
8622 |
++ cr4_fixed1_update(X86_CR4_PGE, edx, bit(X86_FEATURE_PGE)); |
8623 |
++ cr4_fixed1_update(X86_CR4_OSFXSR, edx, bit(X86_FEATURE_FXSR)); |
8624 |
++ cr4_fixed1_update(X86_CR4_OSXMMEXCPT, edx, bit(X86_FEATURE_XMM)); |
8625 |
++ cr4_fixed1_update(X86_CR4_VMXE, ecx, bit(X86_FEATURE_VMX)); |
8626 |
++ cr4_fixed1_update(X86_CR4_SMXE, ecx, bit(X86_FEATURE_SMX)); |
8627 |
++ cr4_fixed1_update(X86_CR4_PCIDE, ecx, bit(X86_FEATURE_PCID)); |
8628 |
++ cr4_fixed1_update(X86_CR4_OSXSAVE, ecx, bit(X86_FEATURE_XSAVE)); |
8629 |
++ |
8630 |
++ entry = kvm_find_cpuid_entry(vcpu, 0x7, 0); |
8631 |
++ cr4_fixed1_update(X86_CR4_FSGSBASE, ebx, bit(X86_FEATURE_FSGSBASE)); |
8632 |
++ cr4_fixed1_update(X86_CR4_SMEP, ebx, bit(X86_FEATURE_SMEP)); |
8633 |
++ cr4_fixed1_update(X86_CR4_SMAP, ebx, bit(X86_FEATURE_SMAP)); |
8634 |
++ cr4_fixed1_update(X86_CR4_PKE, ecx, bit(X86_FEATURE_PKU)); |
8635 |
++ cr4_fixed1_update(X86_CR4_UMIP, ecx, bit(X86_FEATURE_UMIP)); |
8636 |
++ cr4_fixed1_update(X86_CR4_LA57, ecx, bit(X86_FEATURE_LA57)); |
8637 |
++ |
8638 |
++#undef cr4_fixed1_update |
8639 |
++} |
8640 |
++ |
8641 |
++static void nested_vmx_entry_exit_ctls_update(struct kvm_vcpu *vcpu) |
8642 |
++{ |
8643 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8644 |
++ |
8645 |
++ if (kvm_mpx_supported()) { |
8646 |
++ bool mpx_enabled = guest_cpuid_has(vcpu, X86_FEATURE_MPX); |
8647 |
++ |
8648 |
++ if (mpx_enabled) { |
8649 |
++ vmx->nested.msrs.entry_ctls_high |= VM_ENTRY_LOAD_BNDCFGS; |
8650 |
++ vmx->nested.msrs.exit_ctls_high |= VM_EXIT_CLEAR_BNDCFGS; |
8651 |
++ } else { |
8652 |
++ vmx->nested.msrs.entry_ctls_high &= ~VM_ENTRY_LOAD_BNDCFGS; |
8653 |
++ vmx->nested.msrs.exit_ctls_high &= ~VM_EXIT_CLEAR_BNDCFGS; |
8654 |
++ } |
8655 |
++ } |
8656 |
++} |
8657 |
++ |
8658 |
++static void update_intel_pt_cfg(struct kvm_vcpu *vcpu) |
8659 |
++{ |
8660 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8661 |
++ struct kvm_cpuid_entry2 *best = NULL; |
8662 |
++ int i; |
8663 |
++ |
8664 |
++ for (i = 0; i < PT_CPUID_LEAVES; i++) { |
8665 |
++ best = kvm_find_cpuid_entry(vcpu, 0x14, i); |
8666 |
++ if (!best) |
8667 |
++ return; |
8668 |
++ vmx->pt_desc.caps[CPUID_EAX + i*PT_CPUID_REGS_NUM] = best->eax; |
8669 |
++ vmx->pt_desc.caps[CPUID_EBX + i*PT_CPUID_REGS_NUM] = best->ebx; |
8670 |
++ vmx->pt_desc.caps[CPUID_ECX + i*PT_CPUID_REGS_NUM] = best->ecx; |
8671 |
++ vmx->pt_desc.caps[CPUID_EDX + i*PT_CPUID_REGS_NUM] = best->edx; |
8672 |
++ } |
8673 |
++ |
8674 |
++ /* Get the number of configurable Address Ranges for filtering */ |
8675 |
++ vmx->pt_desc.addr_range = intel_pt_validate_cap(vmx->pt_desc.caps, |
8676 |
++ PT_CAP_num_address_ranges); |
8677 |
++ |
8678 |
++ /* Initialize and clear the no dependency bits */ |
8679 |
++ vmx->pt_desc.ctl_bitmask = ~(RTIT_CTL_TRACEEN | RTIT_CTL_OS | |
8680 |
++ RTIT_CTL_USR | RTIT_CTL_TSC_EN | RTIT_CTL_DISRETC); |
8681 |
++ |
8682 |
++ /* |
8683 |
++ * If CPUID.(EAX=14H,ECX=0):EBX[0]=1 CR3Filter can be set otherwise |
8684 |
++ * will inject an #GP |
8685 |
++ */ |
8686 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_cr3_filtering)) |
8687 |
++ vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_CR3EN; |
8688 |
++ |
8689 |
++ /* |
8690 |
++ * If CPUID.(EAX=14H,ECX=0):EBX[1]=1 CYCEn, CycThresh and |
8691 |
++ * PSBFreq can be set |
8692 |
++ */ |
8693 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc)) |
8694 |
++ vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_CYCLEACC | |
8695 |
++ RTIT_CTL_CYC_THRESH | RTIT_CTL_PSB_FREQ); |
8696 |
++ |
8697 |
++ /* |
8698 |
++ * If CPUID.(EAX=14H,ECX=0):EBX[3]=1 MTCEn BranchEn and |
8699 |
++ * MTCFreq can be set |
8700 |
++ */ |
8701 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc)) |
8702 |
++ vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_MTC_EN | |
8703 |
++ RTIT_CTL_BRANCH_EN | RTIT_CTL_MTC_RANGE); |
8704 |
++ |
8705 |
++ /* If CPUID.(EAX=14H,ECX=0):EBX[4]=1 FUPonPTW and PTWEn can be set */ |
8706 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_ptwrite)) |
8707 |
++ vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_FUP_ON_PTW | |
8708 |
++ RTIT_CTL_PTW_EN); |
8709 |
++ |
8710 |
++ /* If CPUID.(EAX=14H,ECX=0):EBX[5]=1 PwrEvEn can be set */ |
8711 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_power_event_trace)) |
8712 |
++ vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_PWR_EVT_EN; |
8713 |
++ |
8714 |
++ /* If CPUID.(EAX=14H,ECX=0):ECX[0]=1 ToPA can be set */ |
8715 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_topa_output)) |
8716 |
++ vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_TOPA; |
8717 |
++ |
8718 |
++ /* If CPUID.(EAX=14H,ECX=0):ECX[3]=1 FabircEn can be set */ |
8719 |
++ if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_output_subsys)) |
8720 |
++ vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_FABRIC_EN; |
8721 |
++ |
8722 |
++ /* unmask address range configure area */ |
8723 |
++ for (i = 0; i < vmx->pt_desc.addr_range; i++) |
8724 |
++ vmx->pt_desc.ctl_bitmask &= ~(0xfULL << (32 + i * 4)); |
8725 |
++} |
8726 |
++ |
8727 |
++static void vmx_cpuid_update(struct kvm_vcpu *vcpu) |
8728 |
++{ |
8729 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8730 |
++ |
8731 |
++ /* xsaves_enabled is recomputed in vmx_compute_secondary_exec_control(). */ |
8732 |
++ vcpu->arch.xsaves_enabled = false; |
8733 |
++ |
8734 |
++ if (cpu_has_secondary_exec_ctrls()) { |
8735 |
++ vmx_compute_secondary_exec_control(vmx); |
8736 |
++ vmcs_set_secondary_exec_control(vmx); |
8737 |
++ } |
8738 |
++ |
8739 |
++ if (nested_vmx_allowed(vcpu)) |
8740 |
++ to_vmx(vcpu)->msr_ia32_feature_control_valid_bits |= |
8741 |
++ FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX | |
8742 |
++ FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; |
8743 |
++ else |
8744 |
++ to_vmx(vcpu)->msr_ia32_feature_control_valid_bits &= |
8745 |
++ ~(FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX | |
8746 |
++ FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX); |
8747 |
++ |
8748 |
++ if (nested_vmx_allowed(vcpu)) { |
8749 |
++ nested_vmx_cr_fixed1_bits_update(vcpu); |
8750 |
++ nested_vmx_entry_exit_ctls_update(vcpu); |
8751 |
++ } |
8752 |
++ |
8753 |
++ if (boot_cpu_has(X86_FEATURE_INTEL_PT) && |
8754 |
++ guest_cpuid_has(vcpu, X86_FEATURE_INTEL_PT)) |
8755 |
++ update_intel_pt_cfg(vcpu); |
8756 |
++ |
8757 |
++ if (boot_cpu_has(X86_FEATURE_RTM)) { |
8758 |
++ struct shared_msr_entry *msr; |
8759 |
++ msr = find_msr_entry(vmx, MSR_IA32_TSX_CTRL); |
8760 |
++ if (msr) { |
8761 |
++ bool enabled = guest_cpuid_has(vcpu, X86_FEATURE_RTM); |
8762 |
++ vmx_set_guest_msr(vmx, msr, enabled ? 0 : TSX_CTRL_RTM_DISABLE); |
8763 |
++ } |
8764 |
++ } |
8765 |
++} |
8766 |
++ |
8767 |
++static void vmx_set_supported_cpuid(u32 func, struct kvm_cpuid_entry2 *entry) |
8768 |
++{ |
8769 |
++ if (func == 1 && nested) |
8770 |
++ entry->ecx |= bit(X86_FEATURE_VMX); |
8771 |
++} |
8772 |
++ |
8773 |
++static void vmx_request_immediate_exit(struct kvm_vcpu *vcpu) |
8774 |
++{ |
8775 |
++ to_vmx(vcpu)->req_immediate_exit = true; |
8776 |
++} |
8777 |
++ |
8778 |
++static int vmx_check_intercept(struct kvm_vcpu *vcpu, |
8779 |
++ struct x86_instruction_info *info, |
8780 |
++ enum x86_intercept_stage stage) |
8781 |
++{ |
8782 |
++ struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
8783 |
++ struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt; |
8784 |
++ |
8785 |
++ /* |
8786 |
++ * RDPID causes #UD if disabled through secondary execution controls. |
8787 |
++ * Because it is marked as EmulateOnUD, we need to intercept it here. |
8788 |
++ */ |
8789 |
++ if (info->intercept == x86_intercept_rdtscp && |
8790 |
++ !nested_cpu_has2(vmcs12, SECONDARY_EXEC_RDTSCP)) { |
8791 |
++ ctxt->exception.vector = UD_VECTOR; |
8792 |
++ ctxt->exception.error_code_valid = false; |
8793 |
++ return X86EMUL_PROPAGATE_FAULT; |
8794 |
++ } |
8795 |
++ |
8796 |
++ /* TODO: check more intercepts... */ |
8797 |
++ return X86EMUL_CONTINUE; |
8798 |
++} |
8799 |
++ |
8800 |
++#ifdef CONFIG_X86_64 |
8801 |
++/* (a << shift) / divisor, return 1 if overflow otherwise 0 */ |
8802 |
++static inline int u64_shl_div_u64(u64 a, unsigned int shift, |
8803 |
++ u64 divisor, u64 *result) |
8804 |
++{ |
8805 |
++ u64 low = a << shift, high = a >> (64 - shift); |
8806 |
++ |
8807 |
++ /* To avoid the overflow on divq */ |
8808 |
++ if (high >= divisor) |
8809 |
++ return 1; |
8810 |
++ |
8811 |
++ /* Low hold the result, high hold rem which is discarded */ |
8812 |
++ asm("divq %2\n\t" : "=a" (low), "=d" (high) : |
8813 |
++ "rm" (divisor), "0" (low), "1" (high)); |
8814 |
++ *result = low; |
8815 |
++ |
8816 |
++ return 0; |
8817 |
++} |
8818 |
++ |
8819 |
++static int vmx_set_hv_timer(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc, |
8820 |
++ bool *expired) |
8821 |
++{ |
8822 |
++ struct vcpu_vmx *vmx; |
8823 |
++ u64 tscl, guest_tscl, delta_tsc, lapic_timer_advance_cycles; |
8824 |
++ struct kvm_timer *ktimer = &vcpu->arch.apic->lapic_timer; |
8825 |
++ |
8826 |
++ if (kvm_mwait_in_guest(vcpu->kvm) || |
8827 |
++ kvm_can_post_timer_interrupt(vcpu)) |
8828 |
++ return -EOPNOTSUPP; |
8829 |
++ |
8830 |
++ vmx = to_vmx(vcpu); |
8831 |
++ tscl = rdtsc(); |
8832 |
++ guest_tscl = kvm_read_l1_tsc(vcpu, tscl); |
8833 |
++ delta_tsc = max(guest_deadline_tsc, guest_tscl) - guest_tscl; |
8834 |
++ lapic_timer_advance_cycles = nsec_to_cycles(vcpu, |
8835 |
++ ktimer->timer_advance_ns); |
8836 |
++ |
8837 |
++ if (delta_tsc > lapic_timer_advance_cycles) |
8838 |
++ delta_tsc -= lapic_timer_advance_cycles; |
8839 |
++ else |
8840 |
++ delta_tsc = 0; |
8841 |
++ |
8842 |
++ /* Convert to host delta tsc if tsc scaling is enabled */ |
8843 |
++ if (vcpu->arch.tsc_scaling_ratio != kvm_default_tsc_scaling_ratio && |
8844 |
++ delta_tsc && u64_shl_div_u64(delta_tsc, |
8845 |
++ kvm_tsc_scaling_ratio_frac_bits, |
8846 |
++ vcpu->arch.tsc_scaling_ratio, &delta_tsc)) |
8847 |
++ return -ERANGE; |
8848 |
++ |
8849 |
++ /* |
8850 |
++ * If the delta tsc can't fit in the 32 bit after the multi shift, |
8851 |
++ * we can't use the preemption timer. |
8852 |
++ * It's possible that it fits on later vmentries, but checking |
8853 |
++ * on every vmentry is costly so we just use an hrtimer. |
8854 |
++ */ |
8855 |
++ if (delta_tsc >> (cpu_preemption_timer_multi + 32)) |
8856 |
++ return -ERANGE; |
8857 |
++ |
8858 |
++ vmx->hv_deadline_tsc = tscl + delta_tsc; |
8859 |
++ *expired = !delta_tsc; |
8860 |
++ return 0; |
8861 |
++} |
8862 |
++ |
8863 |
++static void vmx_cancel_hv_timer(struct kvm_vcpu *vcpu) |
8864 |
++{ |
8865 |
++ to_vmx(vcpu)->hv_deadline_tsc = -1; |
8866 |
++} |
8867 |
++#endif |
8868 |
++ |
8869 |
++static void vmx_sched_in(struct kvm_vcpu *vcpu, int cpu) |
8870 |
++{ |
8871 |
++ if (!kvm_pause_in_guest(vcpu->kvm)) |
8872 |
++ shrink_ple_window(vcpu); |
8873 |
++} |
8874 |
++ |
8875 |
++static void vmx_slot_enable_log_dirty(struct kvm *kvm, |
8876 |
++ struct kvm_memory_slot *slot) |
8877 |
++{ |
8878 |
++ kvm_mmu_slot_leaf_clear_dirty(kvm, slot); |
8879 |
++ kvm_mmu_slot_largepage_remove_write_access(kvm, slot); |
8880 |
++} |
8881 |
++ |
8882 |
++static void vmx_slot_disable_log_dirty(struct kvm *kvm, |
8883 |
++ struct kvm_memory_slot *slot) |
8884 |
++{ |
8885 |
++ kvm_mmu_slot_set_dirty(kvm, slot); |
8886 |
++} |
8887 |
++ |
8888 |
++static void vmx_flush_log_dirty(struct kvm *kvm) |
8889 |
++{ |
8890 |
++ kvm_flush_pml_buffers(kvm); |
8891 |
++} |
8892 |
++ |
8893 |
++static int vmx_write_pml_buffer(struct kvm_vcpu *vcpu) |
8894 |
++{ |
8895 |
++ struct vmcs12 *vmcs12; |
8896 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
8897 |
++ gpa_t gpa, dst; |
8898 |
++ |
8899 |
++ if (is_guest_mode(vcpu)) { |
8900 |
++ WARN_ON_ONCE(vmx->nested.pml_full); |
8901 |
++ |
8902 |
++ /* |
8903 |
++ * Check if PML is enabled for the nested guest. |
8904 |
++ * Whether eptp bit 6 is set is already checked |
8905 |
++ * as part of A/D emulation. |
8906 |
++ */ |
8907 |
++ vmcs12 = get_vmcs12(vcpu); |
8908 |
++ if (!nested_cpu_has_pml(vmcs12)) |
8909 |
++ return 0; |
8910 |
++ |
8911 |
++ if (vmcs12->guest_pml_index >= PML_ENTITY_NUM) { |
8912 |
++ vmx->nested.pml_full = true; |
8913 |
++ return 1; |
8914 |
++ } |
8915 |
++ |
8916 |
++ gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS) & ~0xFFFull; |
8917 |
++ dst = vmcs12->pml_address + sizeof(u64) * vmcs12->guest_pml_index; |
8918 |
++ |
8919 |
++ if (kvm_write_guest_page(vcpu->kvm, gpa_to_gfn(dst), &gpa, |
8920 |
++ offset_in_page(dst), sizeof(gpa))) |
8921 |
++ return 0; |
8922 |
++ |
8923 |
++ vmcs12->guest_pml_index--; |
8924 |
++ } |
8925 |
++ |
8926 |
++ return 0; |
8927 |
++} |
8928 |
++ |
8929 |
++static void vmx_enable_log_dirty_pt_masked(struct kvm *kvm, |
8930 |
++ struct kvm_memory_slot *memslot, |
8931 |
++ gfn_t offset, unsigned long mask) |
8932 |
++{ |
8933 |
++ kvm_mmu_clear_dirty_pt_masked(kvm, memslot, offset, mask); |
8934 |
++} |
8935 |
++ |
8936 |
++static void __pi_post_block(struct kvm_vcpu *vcpu) |
8937 |
++{ |
8938 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
8939 |
++ struct pi_desc old, new; |
8940 |
++ unsigned int dest; |
8941 |
++ |
8942 |
++ do { |
8943 |
++ old.control = new.control = pi_desc->control; |
8944 |
++ WARN(old.nv != POSTED_INTR_WAKEUP_VECTOR, |
8945 |
++ "Wakeup handler not enabled while the VCPU is blocked\n"); |
8946 |
++ |
8947 |
++ dest = cpu_physical_id(vcpu->cpu); |
8948 |
++ |
8949 |
++ if (x2apic_enabled()) |
8950 |
++ new.ndst = dest; |
8951 |
++ else |
8952 |
++ new.ndst = (dest << 8) & 0xFF00; |
8953 |
++ |
8954 |
++ /* set 'NV' to 'notification vector' */ |
8955 |
++ new.nv = POSTED_INTR_VECTOR; |
8956 |
++ } while (cmpxchg64(&pi_desc->control, old.control, |
8957 |
++ new.control) != old.control); |
8958 |
++ |
8959 |
++ if (!WARN_ON_ONCE(vcpu->pre_pcpu == -1)) { |
8960 |
++ spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8961 |
++ list_del(&vcpu->blocked_vcpu_list); |
8962 |
++ spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8963 |
++ vcpu->pre_pcpu = -1; |
8964 |
++ } |
8965 |
++} |
8966 |
++ |
8967 |
++/* |
8968 |
++ * This routine does the following things for vCPU which is going |
8969 |
++ * to be blocked if VT-d PI is enabled. |
8970 |
++ * - Store the vCPU to the wakeup list, so when interrupts happen |
8971 |
++ * we can find the right vCPU to wake up. |
8972 |
++ * - Change the Posted-interrupt descriptor as below: |
8973 |
++ * 'NDST' <-- vcpu->pre_pcpu |
8974 |
++ * 'NV' <-- POSTED_INTR_WAKEUP_VECTOR |
8975 |
++ * - If 'ON' is set during this process, which means at least one |
8976 |
++ * interrupt is posted for this vCPU, we cannot block it, in |
8977 |
++ * this case, return 1, otherwise, return 0. |
8978 |
++ * |
8979 |
++ */ |
8980 |
++static int pi_pre_block(struct kvm_vcpu *vcpu) |
8981 |
++{ |
8982 |
++ unsigned int dest; |
8983 |
++ struct pi_desc old, new; |
8984 |
++ struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
8985 |
++ |
8986 |
++ if (!kvm_arch_has_assigned_device(vcpu->kvm) || |
8987 |
++ !irq_remapping_cap(IRQ_POSTING_CAP) || |
8988 |
++ !kvm_vcpu_apicv_active(vcpu)) |
8989 |
++ return 0; |
8990 |
++ |
8991 |
++ WARN_ON(irqs_disabled()); |
8992 |
++ local_irq_disable(); |
8993 |
++ if (!WARN_ON_ONCE(vcpu->pre_pcpu != -1)) { |
8994 |
++ vcpu->pre_pcpu = vcpu->cpu; |
8995 |
++ spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8996 |
++ list_add_tail(&vcpu->blocked_vcpu_list, |
8997 |
++ &per_cpu(blocked_vcpu_on_cpu, |
8998 |
++ vcpu->pre_pcpu)); |
8999 |
++ spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
9000 |
++ } |
9001 |
++ |
9002 |
++ do { |
9003 |
++ old.control = new.control = pi_desc->control; |
9004 |
++ |
9005 |
++ WARN((pi_desc->sn == 1), |
9006 |
++ "Warning: SN field of posted-interrupts " |
9007 |
++ "is set before blocking\n"); |
9008 |
++ |
9009 |
++ /* |
9010 |
++ * Since vCPU can be preempted during this process, |
9011 |
++ * vcpu->cpu could be different with pre_pcpu, we |
9012 |
++ * need to set pre_pcpu as the destination of wakeup |
9013 |
++ * notification event, then we can find the right vCPU |
9014 |
++ * to wakeup in wakeup handler if interrupts happen |
9015 |
++ * when the vCPU is in blocked state. |
9016 |
++ */ |
9017 |
++ dest = cpu_physical_id(vcpu->pre_pcpu); |
9018 |
++ |
9019 |
++ if (x2apic_enabled()) |
9020 |
++ new.ndst = dest; |
9021 |
++ else |
9022 |
++ new.ndst = (dest << 8) & 0xFF00; |
9023 |
++ |
9024 |
++ /* set 'NV' to 'wakeup vector' */ |
9025 |
++ new.nv = POSTED_INTR_WAKEUP_VECTOR; |
9026 |
++ } while (cmpxchg64(&pi_desc->control, old.control, |
9027 |
++ new.control) != old.control); |
9028 |
++ |
9029 |
++ /* We should not block the vCPU if an interrupt is posted for it. */ |
9030 |
++ if (pi_test_on(pi_desc) == 1) |
9031 |
++ __pi_post_block(vcpu); |
9032 |
++ |
9033 |
++ local_irq_enable(); |
9034 |
++ return (vcpu->pre_pcpu == -1); |
9035 |
++} |
9036 |
++ |
9037 |
++static int vmx_pre_block(struct kvm_vcpu *vcpu) |
9038 |
++{ |
9039 |
++ if (pi_pre_block(vcpu)) |
9040 |
++ return 1; |
9041 |
++ |
9042 |
++ if (kvm_lapic_hv_timer_in_use(vcpu)) |
9043 |
++ kvm_lapic_switch_to_sw_timer(vcpu); |
9044 |
++ |
9045 |
++ return 0; |
9046 |
++} |
9047 |
++ |
9048 |
++static void pi_post_block(struct kvm_vcpu *vcpu) |
9049 |
++{ |
9050 |
++ if (vcpu->pre_pcpu == -1) |
9051 |
++ return; |
9052 |
++ |
9053 |
++ WARN_ON(irqs_disabled()); |
9054 |
++ local_irq_disable(); |
9055 |
++ __pi_post_block(vcpu); |
9056 |
++ local_irq_enable(); |
9057 |
++} |
9058 |
++ |
9059 |
++static void vmx_post_block(struct kvm_vcpu *vcpu) |
9060 |
++{ |
9061 |
++ if (kvm_x86_ops->set_hv_timer) |
9062 |
++ kvm_lapic_switch_to_hv_timer(vcpu); |
9063 |
++ |
9064 |
++ pi_post_block(vcpu); |
9065 |
++} |
9066 |
++ |
9067 |
++/* |
9068 |
++ * vmx_update_pi_irte - set IRTE for Posted-Interrupts |
9069 |
++ * |
9070 |
++ * @kvm: kvm |
9071 |
++ * @host_irq: host irq of the interrupt |
9072 |
++ * @guest_irq: gsi of the interrupt |
9073 |
++ * @set: set or unset PI |
9074 |
++ * returns 0 on success, < 0 on failure |
9075 |
++ */ |
9076 |
++static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, |
9077 |
++ uint32_t guest_irq, bool set) |
9078 |
++{ |
9079 |
++ struct kvm_kernel_irq_routing_entry *e; |
9080 |
++ struct kvm_irq_routing_table *irq_rt; |
9081 |
++ struct kvm_lapic_irq irq; |
9082 |
++ struct kvm_vcpu *vcpu; |
9083 |
++ struct vcpu_data vcpu_info; |
9084 |
++ int idx, ret = 0; |
9085 |
++ |
9086 |
++ if (!kvm_arch_has_assigned_device(kvm) || |
9087 |
++ !irq_remapping_cap(IRQ_POSTING_CAP) || |
9088 |
++ !kvm_vcpu_apicv_active(kvm->vcpus[0])) |
9089 |
++ return 0; |
9090 |
++ |
9091 |
++ idx = srcu_read_lock(&kvm->irq_srcu); |
9092 |
++ irq_rt = srcu_dereference(kvm->irq_routing, &kvm->irq_srcu); |
9093 |
++ if (guest_irq >= irq_rt->nr_rt_entries || |
9094 |
++ hlist_empty(&irq_rt->map[guest_irq])) { |
9095 |
++ pr_warn_once("no route for guest_irq %u/%u (broken user space?)\n", |
9096 |
++ guest_irq, irq_rt->nr_rt_entries); |
9097 |
++ goto out; |
9098 |
++ } |
9099 |
++ |
9100 |
++ hlist_for_each_entry(e, &irq_rt->map[guest_irq], link) { |
9101 |
++ if (e->type != KVM_IRQ_ROUTING_MSI) |
9102 |
++ continue; |
9103 |
++ /* |
9104 |
++ * VT-d PI cannot support posting multicast/broadcast |
9105 |
++ * interrupts to a vCPU, we still use interrupt remapping |
9106 |
++ * for these kind of interrupts. |
9107 |
++ * |
9108 |
++ * For lowest-priority interrupts, we only support |
9109 |
++ * those with single CPU as the destination, e.g. user |
9110 |
++ * configures the interrupts via /proc/irq or uses |
9111 |
++ * irqbalance to make the interrupts single-CPU. |
9112 |
++ * |
9113 |
++ * We will support full lowest-priority interrupt later. |
9114 |
++ * |
9115 |
++ * In addition, we can only inject generic interrupts using |
9116 |
++ * the PI mechanism, refuse to route others through it. |
9117 |
++ */ |
9118 |
++ |
9119 |
++ kvm_set_msi_irq(kvm, e, &irq); |
9120 |
++ if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) || |
9121 |
++ !kvm_irq_is_postable(&irq)) { |
9122 |
++ /* |
9123 |
++ * Make sure the IRTE is in remapped mode if |
9124 |
++ * we don't handle it in posted mode. |
9125 |
++ */ |
9126 |
++ ret = irq_set_vcpu_affinity(host_irq, NULL); |
9127 |
++ if (ret < 0) { |
9128 |
++ printk(KERN_INFO |
9129 |
++ "failed to back to remapped mode, irq: %u\n", |
9130 |
++ host_irq); |
9131 |
++ goto out; |
9132 |
++ } |
9133 |
++ |
9134 |
++ continue; |
9135 |
++ } |
9136 |
++ |
9137 |
++ vcpu_info.pi_desc_addr = __pa(vcpu_to_pi_desc(vcpu)); |
9138 |
++ vcpu_info.vector = irq.vector; |
9139 |
++ |
9140 |
++ trace_kvm_pi_irte_update(host_irq, vcpu->vcpu_id, e->gsi, |
9141 |
++ vcpu_info.vector, vcpu_info.pi_desc_addr, set); |
9142 |
++ |
9143 |
++ if (set) |
9144 |
++ ret = irq_set_vcpu_affinity(host_irq, &vcpu_info); |
9145 |
++ else |
9146 |
++ ret = irq_set_vcpu_affinity(host_irq, NULL); |
9147 |
++ |
9148 |
++ if (ret < 0) { |
9149 |
++ printk(KERN_INFO "%s: failed to update PI IRTE\n", |
9150 |
++ __func__); |
9151 |
++ goto out; |
9152 |
++ } |
9153 |
++ } |
9154 |
++ |
9155 |
++ ret = 0; |
9156 |
++out: |
9157 |
++ srcu_read_unlock(&kvm->irq_srcu, idx); |
9158 |
++ return ret; |
9159 |
++} |
9160 |
++ |
9161 |
++static void vmx_setup_mce(struct kvm_vcpu *vcpu) |
9162 |
++{ |
9163 |
++ if (vcpu->arch.mcg_cap & MCG_LMCE_P) |
9164 |
++ to_vmx(vcpu)->msr_ia32_feature_control_valid_bits |= |
9165 |
++ FEATURE_CONTROL_LMCE; |
9166 |
++ else |
9167 |
++ to_vmx(vcpu)->msr_ia32_feature_control_valid_bits &= |
9168 |
++ ~FEATURE_CONTROL_LMCE; |
9169 |
++} |
9170 |
++ |
9171 |
++static int vmx_smi_allowed(struct kvm_vcpu *vcpu) |
9172 |
++{ |
9173 |
++ /* we need a nested vmexit to enter SMM, postpone if run is pending */ |
9174 |
++ if (to_vmx(vcpu)->nested.nested_run_pending) |
9175 |
++ return 0; |
9176 |
++ return 1; |
9177 |
++} |
9178 |
++ |
9179 |
++static int vmx_pre_enter_smm(struct kvm_vcpu *vcpu, char *smstate) |
9180 |
++{ |
9181 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
9182 |
++ |
9183 |
++ vmx->nested.smm.guest_mode = is_guest_mode(vcpu); |
9184 |
++ if (vmx->nested.smm.guest_mode) |
9185 |
++ nested_vmx_vmexit(vcpu, -1, 0, 0); |
9186 |
++ |
9187 |
++ vmx->nested.smm.vmxon = vmx->nested.vmxon; |
9188 |
++ vmx->nested.vmxon = false; |
9189 |
++ vmx_clear_hlt(vcpu); |
9190 |
++ return 0; |
9191 |
++} |
9192 |
++ |
9193 |
++static int vmx_pre_leave_smm(struct kvm_vcpu *vcpu, const char *smstate) |
9194 |
++{ |
9195 |
++ struct vcpu_vmx *vmx = to_vmx(vcpu); |
9196 |
++ int ret; |
9197 |
++ |
9198 |
++ if (vmx->nested.smm.vmxon) { |
9199 |
++ vmx->nested.vmxon = true; |
9200 |
++ vmx->nested.smm.vmxon = false; |
9201 |
++ } |
9202 |
++ |
9203 |
++ if (vmx->nested.smm.guest_mode) { |
9204 |
++ ret = nested_vmx_enter_non_root_mode(vcpu, false); |
9205 |
++ if (ret) |
9206 |
++ return ret; |
9207 |
++ |
9208 |
++ vmx->nested.smm.guest_mode = false; |
9209 |
++ } |
9210 |
++ return 0; |
9211 |
++} |
9212 |
++ |
9213 |
++static int enable_smi_window(struct kvm_vcpu *vcpu) |
9214 |
++{ |
9215 |
++ return 0; |
9216 |
++} |
9217 |
++ |
9218 |
++static bool vmx_need_emulation_on_page_fault(struct kvm_vcpu *vcpu) |
9219 |
++{ |
9220 |
++ return false; |
9221 |
++} |
9222 |
++ |
9223 |
++static bool vmx_apic_init_signal_blocked(struct kvm_vcpu *vcpu) |
9224 |
++{ |
9225 |
++ return to_vmx(vcpu)->nested.vmxon; |
9226 |
++} |
9227 |
++ |
9228 |
++static __init int hardware_setup(void) |
9229 |
++{ |
9230 |
++ unsigned long host_bndcfgs; |
9231 |
++ struct desc_ptr dt; |
9232 |
++ int r, i; |
9233 |
++ |
9234 |
++ rdmsrl_safe(MSR_EFER, &host_efer); |
9235 |
++ |
9236 |
++ store_idt(&dt); |
9237 |
++ host_idt_base = dt.address; |
9238 |
++ |
9239 |
++ for (i = 0; i < ARRAY_SIZE(vmx_msr_index); ++i) |
9240 |
++ kvm_define_shared_msr(i, vmx_msr_index[i]); |
9241 |
++ |
9242 |
++ if (setup_vmcs_config(&vmcs_config, &vmx_capability) < 0) |
9243 |
++ return -EIO; |
9244 |
++ |
9245 |
++ if (boot_cpu_has(X86_FEATURE_NX)) |
9246 |
++ kvm_enable_efer_bits(EFER_NX); |
9247 |
++ |
9248 |
++ if (boot_cpu_has(X86_FEATURE_MPX)) { |
9249 |
++ rdmsrl(MSR_IA32_BNDCFGS, host_bndcfgs); |
9250 |
++ WARN_ONCE(host_bndcfgs, "KVM: BNDCFGS in host will be lost"); |
9251 |
++ } |
9252 |
++ |
9253 |
++ if (!cpu_has_vmx_vpid() || !cpu_has_vmx_invvpid() || |
9254 |
++ !(cpu_has_vmx_invvpid_single() || cpu_has_vmx_invvpid_global())) |
9255 |
++ enable_vpid = 0; |
9256 |
++ |
9257 |
++ if (!cpu_has_vmx_ept() || |
9258 |
++ !cpu_has_vmx_ept_4levels() || |
9259 |
++ !cpu_has_vmx_ept_mt_wb() || |
9260 |
++ !cpu_has_vmx_invept_global()) |
9261 |
++ enable_ept = 0; |
9262 |
++ |
9263 |
++ if (!cpu_has_vmx_ept_ad_bits() || !enable_ept) |
9264 |
++ enable_ept_ad_bits = 0; |
9265 |
++ |
9266 |
++ if (!cpu_has_vmx_unrestricted_guest() || !enable_ept) |
9267 |
++ enable_unrestricted_guest = 0; |
9268 |
++ |
9269 |
++ if (!cpu_has_vmx_flexpriority()) |
9270 |
++ flexpriority_enabled = 0; |
9271 |
++ |
9272 |
++ if (!cpu_has_virtual_nmis()) |
9273 |
++ enable_vnmi = 0; |
9274 |
++ |
9275 |
++ /* |
9276 |
++ * set_apic_access_page_addr() is used to reload apic access |
9277 |
++ * page upon invalidation. No need to do anything if not |
9278 |
++ * using the APIC_ACCESS_ADDR VMCS field. |
9279 |
++ */ |
9280 |
++ if (!flexpriority_enabled) |
9281 |
++ kvm_x86_ops->set_apic_access_page_addr = NULL; |
9282 |
++ |
9283 |
++ if (!cpu_has_vmx_tpr_shadow()) |
9284 |
++ kvm_x86_ops->update_cr8_intercept = NULL; |
9285 |
++ |
9286 |
++ if (enable_ept && !cpu_has_vmx_ept_2m_page()) |
9287 |
++ kvm_disable_largepages(); |
9288 |
++ |
9289 |
++#if IS_ENABLED(CONFIG_HYPERV) |
9290 |
++ if (ms_hyperv.nested_features & HV_X64_NESTED_GUEST_MAPPING_FLUSH |
9291 |
++ && enable_ept) { |
9292 |
++ kvm_x86_ops->tlb_remote_flush = hv_remote_flush_tlb; |
9293 |
++ kvm_x86_ops->tlb_remote_flush_with_range = |
9294 |
++ hv_remote_flush_tlb_with_range; |
9295 |
++ } |
9296 |
++#endif |
9297 |
++ |
9298 |
++ if (!cpu_has_vmx_ple()) { |
9299 |
++ ple_gap = 0; |
9300 |
++ ple_window = 0; |
9301 |
++ ple_window_grow = 0; |
9302 |
++ ple_window_max = 0; |
9303 |
++ ple_window_shrink = 0; |
9304 |
++ } |
9305 |
++ |
9306 |
++ if (!cpu_has_vmx_apicv()) { |
9307 |
++ enable_apicv = 0; |
9308 |
++ kvm_x86_ops->sync_pir_to_irr = NULL; |
9309 |
++ } |
9310 |
++ |
9311 |
++ if (cpu_has_vmx_tsc_scaling()) { |
9312 |
++ kvm_has_tsc_control = true; |
9313 |
++ kvm_max_tsc_scaling_ratio = KVM_VMX_TSC_MULTIPLIER_MAX; |
9314 |
++ kvm_tsc_scaling_ratio_frac_bits = 48; |
9315 |
++ } |
9316 |
++ |
9317 |
++ set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */ |
9318 |
++ |
9319 |
++ if (enable_ept) |
9320 |
++ vmx_enable_tdp(); |
9321 |
++ else |
9322 |
++ kvm_disable_tdp(); |
9323 |
++ |
9324 |
++ /* |
9325 |
++ * Only enable PML when hardware supports PML feature, and both EPT |
9326 |
++ * and EPT A/D bit features are enabled -- PML depends on them to work. |
9327 |
++ */ |
9328 |
++ if (!enable_ept || !enable_ept_ad_bits || !cpu_has_vmx_pml()) |
9329 |
++ enable_pml = 0; |
9330 |
++ |
9331 |
++ if (!enable_pml) { |
9332 |
++ kvm_x86_ops->slot_enable_log_dirty = NULL; |
9333 |
++ kvm_x86_ops->slot_disable_log_dirty = NULL; |
9334 |
++ kvm_x86_ops->flush_log_dirty = NULL; |
9335 |
++ kvm_x86_ops->enable_log_dirty_pt_masked = NULL; |
9336 |
++ } |
9337 |
++ |
9338 |
++ if (!cpu_has_vmx_preemption_timer()) |
9339 |
++ enable_preemption_timer = false; |
9340 |
++ |
9341 |
++ if (enable_preemption_timer) { |
9342 |
++ u64 use_timer_freq = 5000ULL * 1000 * 1000; |
9343 |
++ u64 vmx_msr; |
9344 |
++ |
9345 |
++ rdmsrl(MSR_IA32_VMX_MISC, vmx_msr); |
9346 |
++ cpu_preemption_timer_multi = |
9347 |
++ vmx_msr & VMX_MISC_PREEMPTION_TIMER_RATE_MASK; |
9348 |
++ |
9349 |
++ if (tsc_khz) |
9350 |
++ use_timer_freq = (u64)tsc_khz * 1000; |
9351 |
++ use_timer_freq >>= cpu_preemption_timer_multi; |
9352 |
++ |
9353 |
++ /* |
9354 |
++ * KVM "disables" the preemption timer by setting it to its max |
9355 |
++ * value. Don't use the timer if it might cause spurious exits |
9356 |
++ * at a rate faster than 0.1 Hz (of uninterrupted guest time). |
9357 |
++ */ |
9358 |
++ if (use_timer_freq > 0xffffffffu / 10) |
9359 |
++ enable_preemption_timer = false; |
9360 |
++ } |
9361 |
++ |
9362 |
++ if (!enable_preemption_timer) { |
9363 |
++ kvm_x86_ops->set_hv_timer = NULL; |
9364 |
++ kvm_x86_ops->cancel_hv_timer = NULL; |
9365 |
++ kvm_x86_ops->request_immediate_exit = __kvm_request_immediate_exit; |
9366 |
++ } |
9367 |
++ |
9368 |
++ kvm_set_posted_intr_wakeup_handler(wakeup_handler); |
9369 |
++ |
9370 |
++ kvm_mce_cap_supported |= MCG_LMCE_P; |
9371 |
++ |
9372 |
++ if (pt_mode != PT_MODE_SYSTEM && pt_mode != PT_MODE_HOST_GUEST) |
9373 |
++ return -EINVAL; |
9374 |
++ if (!enable_ept || !cpu_has_vmx_intel_pt()) |
9375 |
++ pt_mode = PT_MODE_SYSTEM; |
9376 |
++ |
9377 |
++ if (nested) { |
9378 |
++ nested_vmx_setup_ctls_msrs(&vmcs_config.nested, |
9379 |
++ vmx_capability.ept, enable_apicv); |
9380 |
++ |
9381 |
++ r = nested_vmx_hardware_setup(kvm_vmx_exit_handlers); |
9382 |
++ if (r) |
9383 |
++ return r; |
9384 |
++ } |
9385 |
++ |
9386 |
++ r = alloc_kvm_area(); |
9387 |
++ if (r) |
9388 |
++ nested_vmx_hardware_unsetup(); |
9389 |
++ return r; |
9390 |
++} |
9391 |
++ |
9392 |
++static __exit void hardware_unsetup(void) |
9393 |
++{ |
9394 |
++ if (nested) |
9395 |
++ nested_vmx_hardware_unsetup(); |
9396 |
++ |
9397 |
++ free_kvm_area(); |
9398 |
++} |
9399 |
++ |
9400 |
++static struct kvm_x86_ops vmx_x86_ops __ro_after_init = { |
9401 |
++ .cpu_has_kvm_support = cpu_has_kvm_support, |
9402 |
++ .disabled_by_bios = vmx_disabled_by_bios, |
9403 |
++ .hardware_setup = hardware_setup, |
9404 |
++ .hardware_unsetup = hardware_unsetup, |
9405 |
++ .check_processor_compatibility = vmx_check_processor_compat, |
9406 |
++ .hardware_enable = hardware_enable, |
9407 |
++ .hardware_disable = hardware_disable, |
9408 |
++ .cpu_has_accelerated_tpr = report_flexpriority, |
9409 |
++ .has_emulated_msr = vmx_has_emulated_msr, |
9410 |
++ |
9411 |
++ .vm_init = vmx_vm_init, |
9412 |
++ .vm_alloc = vmx_vm_alloc, |
9413 |
++ .vm_free = vmx_vm_free, |
9414 |
++ |
9415 |
++ .vcpu_create = vmx_create_vcpu, |
9416 |
++ .vcpu_free = vmx_free_vcpu, |
9417 |
++ .vcpu_reset = vmx_vcpu_reset, |
9418 |
++ |
9419 |
++ .prepare_guest_switch = vmx_prepare_switch_to_guest, |
9420 |
++ .vcpu_load = vmx_vcpu_load, |
9421 |
++ .vcpu_put = vmx_vcpu_put, |
9422 |
++ |
9423 |
++ .update_bp_intercept = update_exception_bitmap, |
9424 |
++ .get_msr_feature = vmx_get_msr_feature, |
9425 |
++ .get_msr = vmx_get_msr, |
9426 |
++ .set_msr = vmx_set_msr, |
9427 |
++ .get_segment_base = vmx_get_segment_base, |
9428 |
++ .get_segment = vmx_get_segment, |
9429 |
++ .set_segment = vmx_set_segment, |
9430 |
++ .get_cpl = vmx_get_cpl, |
9431 |
++ .get_cs_db_l_bits = vmx_get_cs_db_l_bits, |
9432 |
++ .decache_cr0_guest_bits = vmx_decache_cr0_guest_bits, |
9433 |
++ .decache_cr4_guest_bits = vmx_decache_cr4_guest_bits, |
9434 |
++ .set_cr0 = vmx_set_cr0, |
9435 |
++ .set_cr3 = vmx_set_cr3, |
9436 |
++ .set_cr4 = vmx_set_cr4, |
9437 |
++ .set_efer = vmx_set_efer, |
9438 |
++ .get_idt = vmx_get_idt, |
9439 |
++ .set_idt = vmx_set_idt, |
9440 |
++ .get_gdt = vmx_get_gdt, |
9441 |
++ .set_gdt = vmx_set_gdt, |
9442 |
++ .get_dr6 = vmx_get_dr6, |
9443 |
++ .set_dr6 = vmx_set_dr6, |
9444 |
++ .set_dr7 = vmx_set_dr7, |
9445 |
++ .sync_dirty_debug_regs = vmx_sync_dirty_debug_regs, |
9446 |
++ .cache_reg = vmx_cache_reg, |
9447 |
++ .get_rflags = vmx_get_rflags, |
9448 |
++ .set_rflags = vmx_set_rflags, |
9449 |
++ |
9450 |
++ .tlb_flush = vmx_flush_tlb, |
9451 |
++ .tlb_flush_gva = vmx_flush_tlb_gva, |
9452 |
++ |
9453 |
++ .run = vmx_vcpu_run, |
9454 |
++ .handle_exit = vmx_handle_exit, |
9455 |
++ .skip_emulated_instruction = skip_emulated_instruction, |
9456 |
++ .set_interrupt_shadow = vmx_set_interrupt_shadow, |
9457 |
++ .get_interrupt_shadow = vmx_get_interrupt_shadow, |
9458 |
++ .patch_hypercall = vmx_patch_hypercall, |
9459 |
++ .set_irq = vmx_inject_irq, |
9460 |
++ .set_nmi = vmx_inject_nmi, |
9461 |
++ .queue_exception = vmx_queue_exception, |
9462 |
++ .cancel_injection = vmx_cancel_injection, |
9463 |
++ .interrupt_allowed = vmx_interrupt_allowed, |
9464 |
++ .nmi_allowed = vmx_nmi_allowed, |
9465 |
++ .get_nmi_mask = vmx_get_nmi_mask, |
9466 |
++ .set_nmi_mask = vmx_set_nmi_mask, |
9467 |
++ .enable_nmi_window = enable_nmi_window, |
9468 |
++ .enable_irq_window = enable_irq_window, |
9469 |
++ .update_cr8_intercept = update_cr8_intercept, |
9470 |
++ .set_virtual_apic_mode = vmx_set_virtual_apic_mode, |
9471 |
++ .set_apic_access_page_addr = vmx_set_apic_access_page_addr, |
9472 |
++ .get_enable_apicv = vmx_get_enable_apicv, |
9473 |
++ .refresh_apicv_exec_ctrl = vmx_refresh_apicv_exec_ctrl, |
9474 |
++ .load_eoi_exitmap = vmx_load_eoi_exitmap, |
9475 |
++ .apicv_post_state_restore = vmx_apicv_post_state_restore, |
9476 |
++ .hwapic_irr_update = vmx_hwapic_irr_update, |
9477 |
++ .hwapic_isr_update = vmx_hwapic_isr_update, |
9478 |
++ .guest_apic_has_interrupt = vmx_guest_apic_has_interrupt, |
9479 |
++ .sync_pir_to_irr = vmx_sync_pir_to_irr, |
9480 |
++ .deliver_posted_interrupt = vmx_deliver_posted_interrupt, |
9481 |
++ .dy_apicv_has_pending_interrupt = vmx_dy_apicv_has_pending_interrupt, |
9482 |
++ |
9483 |
++ .set_tss_addr = vmx_set_tss_addr, |
9484 |
++ .set_identity_map_addr = vmx_set_identity_map_addr, |
9485 |
++ .get_tdp_level = get_ept_level, |
9486 |
++ .get_mt_mask = vmx_get_mt_mask, |
9487 |
++ |
9488 |
++ .get_exit_info = vmx_get_exit_info, |
9489 |
++ |
9490 |
++ .get_lpage_level = vmx_get_lpage_level, |
9491 |
++ |
9492 |
++ .cpuid_update = vmx_cpuid_update, |
9493 |
++ |
9494 |
++ .rdtscp_supported = vmx_rdtscp_supported, |
9495 |
++ .invpcid_supported = vmx_invpcid_supported, |
9496 |
++ |
9497 |
++ .set_supported_cpuid = vmx_set_supported_cpuid, |
9498 |
++ |
9499 |
++ .has_wbinvd_exit = cpu_has_vmx_wbinvd_exit, |
9500 |
++ |
9501 |
++ .read_l1_tsc_offset = vmx_read_l1_tsc_offset, |
9502 |
++ .write_l1_tsc_offset = vmx_write_l1_tsc_offset, |
9503 |
++ |
9504 |
++ .set_tdp_cr3 = vmx_set_cr3, |
9505 |
++ |
9506 |
++ .check_intercept = vmx_check_intercept, |
9507 |
++ .handle_exit_irqoff = vmx_handle_exit_irqoff, |
9508 |
++ .mpx_supported = vmx_mpx_supported, |
9509 |
++ .xsaves_supported = vmx_xsaves_supported, |
9510 |
++ .umip_emulated = vmx_umip_emulated, |
9511 |
++ .pt_supported = vmx_pt_supported, |
9512 |
++ |
9513 |
++ .request_immediate_exit = vmx_request_immediate_exit, |
9514 |
++ |
9515 |
++ .sched_in = vmx_sched_in, |
9516 |
++ |
9517 |
++ .slot_enable_log_dirty = vmx_slot_enable_log_dirty, |
9518 |
++ .slot_disable_log_dirty = vmx_slot_disable_log_dirty, |
9519 |
++ .flush_log_dirty = vmx_flush_log_dirty, |
9520 |
++ .enable_log_dirty_pt_masked = vmx_enable_log_dirty_pt_masked, |
9521 |
++ .write_log_dirty = vmx_write_pml_buffer, |
9522 |
++ |
9523 |
++ .pre_block = vmx_pre_block, |
9524 |
++ .post_block = vmx_post_block, |
9525 |
++ |
9526 |
++ .pmu_ops = &intel_pmu_ops, |
9527 |
++ |
9528 |
++ .update_pi_irte = vmx_update_pi_irte, |
9529 |
++ |
9530 |
++#ifdef CONFIG_X86_64 |
9531 |
++ .set_hv_timer = vmx_set_hv_timer, |
9532 |
++ .cancel_hv_timer = vmx_cancel_hv_timer, |
9533 |
++#endif |
9534 |
++ |
9535 |
++ .setup_mce = vmx_setup_mce, |
9536 |
++ |
9537 |
++ .smi_allowed = vmx_smi_allowed, |
9538 |
++ .pre_enter_smm = vmx_pre_enter_smm, |
9539 |
++ .pre_leave_smm = vmx_pre_leave_smm, |
9540 |
++ .enable_smi_window = enable_smi_window, |
9541 |
++ |
9542 |
++ .check_nested_events = NULL, |
9543 |
++ .get_nested_state = NULL, |
9544 |
++ .set_nested_state = NULL, |
9545 |
++ .get_vmcs12_pages = NULL, |
9546 |
++ .nested_enable_evmcs = NULL, |
9547 |
++ .nested_get_evmcs_version = NULL, |
9548 |
++ .need_emulation_on_page_fault = vmx_need_emulation_on_page_fault, |
9549 |
++ .apic_init_signal_blocked = vmx_apic_init_signal_blocked, |
9550 |
++}; |
9551 |
++ |
9552 |
++static void vmx_cleanup_l1d_flush(void) |
9553 |
++{ |
9554 |
++ if (vmx_l1d_flush_pages) { |
9555 |
++ free_pages((unsigned long)vmx_l1d_flush_pages, L1D_CACHE_ORDER); |
9556 |
++ vmx_l1d_flush_pages = NULL; |
9557 |
++ } |
9558 |
++ /* Restore state so sysfs ignores VMX */ |
9559 |
++ l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_AUTO; |
9560 |
++} |
9561 |
++ |
9562 |
++static void vmx_exit(void) |
9563 |
++{ |
9564 |
++#ifdef CONFIG_KEXEC_CORE |
9565 |
++ RCU_INIT_POINTER(crash_vmclear_loaded_vmcss, NULL); |
9566 |
++ synchronize_rcu(); |
9567 |
++#endif |
9568 |
++ |
9569 |
++ kvm_exit(); |
9570 |
++ |
9571 |
++#if IS_ENABLED(CONFIG_HYPERV) |
9572 |
++ if (static_branch_unlikely(&enable_evmcs)) { |
9573 |
++ int cpu; |
9574 |
++ struct hv_vp_assist_page *vp_ap; |
9575 |
++ /* |
9576 |
++ * Reset everything to support using non-enlightened VMCS |
9577 |
++ * access later (e.g. when we reload the module with |
9578 |
++ * enlightened_vmcs=0) |
9579 |
++ */ |
9580 |
++ for_each_online_cpu(cpu) { |
9581 |
++ vp_ap = hv_get_vp_assist_page(cpu); |
9582 |
++ |
9583 |
++ if (!vp_ap) |
9584 |
++ continue; |
9585 |
++ |
9586 |
++ vp_ap->nested_control.features.directhypercall = 0; |
9587 |
++ vp_ap->current_nested_vmcs = 0; |
9588 |
++ vp_ap->enlighten_vmentry = 0; |
9589 |
++ } |
9590 |
++ |
9591 |
++ static_branch_disable(&enable_evmcs); |
9592 |
++ } |
9593 |
++#endif |
9594 |
++ vmx_cleanup_l1d_flush(); |
9595 |
++} |
9596 |
++module_exit(vmx_exit); |
9597 |
++ |
9598 |
++static int __init vmx_init(void) |
9599 |
++{ |
9600 |
++ int r; |
9601 |
++ |
9602 |
++#if IS_ENABLED(CONFIG_HYPERV) |
9603 |
++ /* |
9604 |
++ * Enlightened VMCS usage should be recommended and the host needs |
9605 |
++ * to support eVMCS v1 or above. We can also disable eVMCS support |
9606 |
++ * with module parameter. |
9607 |
++ */ |
9608 |
++ if (enlightened_vmcs && |
9609 |
++ ms_hyperv.hints & HV_X64_ENLIGHTENED_VMCS_RECOMMENDED && |
9610 |
++ (ms_hyperv.nested_features & HV_X64_ENLIGHTENED_VMCS_VERSION) >= |
9611 |
++ KVM_EVMCS_VERSION) { |
9612 |
++ int cpu; |
9613 |
++ |
9614 |
++ /* Check that we have assist pages on all online CPUs */ |
9615 |
++ for_each_online_cpu(cpu) { |
9616 |
++ if (!hv_get_vp_assist_page(cpu)) { |
9617 |
++ enlightened_vmcs = false; |
9618 |
++ break; |
9619 |
++ } |
9620 |
++ } |
9621 |
++ |
9622 |
++ if (enlightened_vmcs) { |
9623 |
++ pr_info("KVM: vmx: using Hyper-V Enlightened VMCS\n"); |
9624 |
++ static_branch_enable(&enable_evmcs); |
9625 |
++ } |
9626 |
++ |
9627 |
++ if (ms_hyperv.nested_features & HV_X64_NESTED_DIRECT_FLUSH) |
9628 |
++ vmx_x86_ops.enable_direct_tlbflush |
9629 |
++ = hv_enable_direct_tlbflush; |
9630 |
++ |
9631 |
++ } else { |
9632 |
++ enlightened_vmcs = false; |
9633 |
++ } |
9634 |
++#endif |
9635 |
++ |
9636 |
++ r = kvm_init(&vmx_x86_ops, sizeof(struct vcpu_vmx), |
9637 |
++ __alignof__(struct vcpu_vmx), THIS_MODULE); |
9638 |
++ if (r) |
9639 |
++ return r; |
9640 |
++ |
9641 |
++ /* |
9642 |
++ * Must be called after kvm_init() so enable_ept is properly set |
9643 |
++ * up. Hand the parameter mitigation value in which was stored in |
9644 |
++ * the pre module init parser. If no parameter was given, it will |
9645 |
++ * contain 'auto' which will be turned into the default 'cond' |
9646 |
++ * mitigation mode. |
9647 |
++ */ |
9648 |
++ r = vmx_setup_l1d_flush(vmentry_l1d_flush_param); |
9649 |
++ if (r) { |
9650 |
++ vmx_exit(); |
9651 |
++ return r; |
9652 |
++ } |
9653 |
++ |
9654 |
++#ifdef CONFIG_KEXEC_CORE |
9655 |
++ rcu_assign_pointer(crash_vmclear_loaded_vmcss, |
9656 |
++ crash_vmclear_local_loaded_vmcss); |
9657 |
++#endif |
9658 |
++ vmx_check_vmcs12_offsets(); |
9659 |
++ |
9660 |
++ return 0; |
9661 |
++} |
9662 |
++module_init(vmx_init); |
9663 |
+diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c |
9664 |
+index 353f63f3b262..ade694f94a49 100644 |
9665 |
+--- a/arch/x86/kvm/x86.c |
9666 |
++++ b/arch/x86/kvm/x86.c |
9667 |
+@@ -92,6 +92,8 @@ u64 __read_mostly efer_reserved_bits = ~((u64)(EFER_SCE | EFER_LME | EFER_LMA)); |
9668 |
+ static u64 __read_mostly efer_reserved_bits = ~((u64)EFER_SCE); |
9669 |
+ #endif |
9670 |
+ |
9671 |
++static u64 __read_mostly cr4_reserved_bits = CR4_RESERVED_BITS; |
9672 |
++ |
9673 |
+ #define VM_STAT(x, ...) offsetof(struct kvm, stat.x), KVM_STAT_VM, ## __VA_ARGS__ |
9674 |
+ #define VCPU_STAT(x, ...) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU, ## __VA_ARGS__ |
9675 |
+ |
9676 |
+@@ -793,9 +795,38 @@ int kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr) |
9677 |
+ } |
9678 |
+ EXPORT_SYMBOL_GPL(kvm_set_xcr); |
9679 |
+ |
9680 |
++static u64 kvm_host_cr4_reserved_bits(struct cpuinfo_x86 *c) |
9681 |
++{ |
9682 |
++ u64 reserved_bits = CR4_RESERVED_BITS; |
9683 |
++ |
9684 |
++ if (!cpu_has(c, X86_FEATURE_XSAVE)) |
9685 |
++ reserved_bits |= X86_CR4_OSXSAVE; |
9686 |
++ |
9687 |
++ if (!cpu_has(c, X86_FEATURE_SMEP)) |
9688 |
++ reserved_bits |= X86_CR4_SMEP; |
9689 |
++ |
9690 |
++ if (!cpu_has(c, X86_FEATURE_SMAP)) |
9691 |
++ reserved_bits |= X86_CR4_SMAP; |
9692 |
++ |
9693 |
++ if (!cpu_has(c, X86_FEATURE_FSGSBASE)) |
9694 |
++ reserved_bits |= X86_CR4_FSGSBASE; |
9695 |
++ |
9696 |
++ if (!cpu_has(c, X86_FEATURE_PKU)) |
9697 |
++ reserved_bits |= X86_CR4_PKE; |
9698 |
++ |
9699 |
++ if (!cpu_has(c, X86_FEATURE_LA57) && |
9700 |
++ !(cpuid_ecx(0x7) & bit(X86_FEATURE_LA57))) |
9701 |
++ reserved_bits |= X86_CR4_LA57; |
9702 |
++ |
9703 |
++ if (!cpu_has(c, X86_FEATURE_UMIP) && !kvm_x86_ops->umip_emulated()) |
9704 |
++ reserved_bits |= X86_CR4_UMIP; |
9705 |
++ |
9706 |
++ return reserved_bits; |
9707 |
++} |
9708 |
++ |
9709 |
+ static int kvm_valid_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
9710 |
+ { |
9711 |
+- if (cr4 & CR4_RESERVED_BITS) |
9712 |
++ if (cr4 & cr4_reserved_bits) |
9713 |
+ return -EINVAL; |
9714 |
+ |
9715 |
+ if (!guest_cpuid_has(vcpu, X86_FEATURE_XSAVE) && (cr4 & X86_CR4_OSXSAVE)) |
9716 |
+@@ -961,9 +992,11 @@ static u64 kvm_dr6_fixed(struct kvm_vcpu *vcpu) |
9717 |
+ |
9718 |
+ static int __kvm_set_dr(struct kvm_vcpu *vcpu, int dr, unsigned long val) |
9719 |
+ { |
9720 |
++ size_t size = ARRAY_SIZE(vcpu->arch.db); |
9721 |
++ |
9722 |
+ switch (dr) { |
9723 |
+ case 0 ... 3: |
9724 |
+- vcpu->arch.db[dr] = val; |
9725 |
++ vcpu->arch.db[array_index_nospec(dr, size)] = val; |
9726 |
+ if (!(vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP)) |
9727 |
+ vcpu->arch.eff_db[dr] = val; |
9728 |
+ break; |
9729 |
+@@ -1000,9 +1033,11 @@ EXPORT_SYMBOL_GPL(kvm_set_dr); |
9730 |
+ |
9731 |
+ int kvm_get_dr(struct kvm_vcpu *vcpu, int dr, unsigned long *val) |
9732 |
+ { |
9733 |
++ size_t size = ARRAY_SIZE(vcpu->arch.db); |
9734 |
++ |
9735 |
+ switch (dr) { |
9736 |
+ case 0 ... 3: |
9737 |
+- *val = vcpu->arch.db[dr]; |
9738 |
++ *val = vcpu->arch.db[array_index_nospec(dr, size)]; |
9739 |
+ break; |
9740 |
+ case 4: |
9741 |
+ /* fall through */ |
9742 |
+@@ -2269,7 +2304,10 @@ static int set_msr_mce(struct kvm_vcpu *vcpu, struct msr_data *msr_info) |
9743 |
+ default: |
9744 |
+ if (msr >= MSR_IA32_MC0_CTL && |
9745 |
+ msr < MSR_IA32_MCx_CTL(bank_num)) { |
9746 |
+- u32 offset = msr - MSR_IA32_MC0_CTL; |
9747 |
++ u32 offset = array_index_nospec( |
9748 |
++ msr - MSR_IA32_MC0_CTL, |
9749 |
++ MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL); |
9750 |
++ |
9751 |
+ /* only 0 or all 1s can be written to IA32_MCi_CTL |
9752 |
+ * some Linux kernels though clear bit 10 in bank 4 to |
9753 |
+ * workaround a BIOS/GART TBL issue on AMD K8s, ignore |
9754 |
+@@ -2681,7 +2719,10 @@ static int get_msr_mce(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata, bool host) |
9755 |
+ default: |
9756 |
+ if (msr >= MSR_IA32_MC0_CTL && |
9757 |
+ msr < MSR_IA32_MCx_CTL(bank_num)) { |
9758 |
+- u32 offset = msr - MSR_IA32_MC0_CTL; |
9759 |
++ u32 offset = array_index_nospec( |
9760 |
++ msr - MSR_IA32_MC0_CTL, |
9761 |
++ MSR_IA32_MCx_CTL(bank_num) - MSR_IA32_MC0_CTL); |
9762 |
++ |
9763 |
+ data = vcpu->arch.mce_banks[offset]; |
9764 |
+ break; |
9765 |
+ } |
9766 |
+@@ -3234,6 +3275,9 @@ static void kvm_steal_time_set_preempted(struct kvm_vcpu *vcpu) |
9767 |
+ if (!(vcpu->arch.st.msr_val & KVM_MSR_ENABLED)) |
9768 |
+ return; |
9769 |
+ |
9770 |
++ if (vcpu->arch.st.steal.preempted) |
9771 |
++ return; |
9772 |
++ |
9773 |
+ vcpu->arch.st.steal.preempted = KVM_VCPU_PREEMPTED; |
9774 |
+ |
9775 |
+ kvm_write_guest_offset_cached(vcpu->kvm, &vcpu->arch.st.stime, |
9776 |
+@@ -5977,11 +6021,11 @@ static int handle_emulation_failure(struct kvm_vcpu *vcpu, int emulation_type) |
9777 |
+ return r; |
9778 |
+ } |
9779 |
+ |
9780 |
+-static bool reexecute_instruction(struct kvm_vcpu *vcpu, gva_t cr2, |
9781 |
++static bool reexecute_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
9782 |
+ bool write_fault_to_shadow_pgtable, |
9783 |
+ int emulation_type) |
9784 |
+ { |
9785 |
+- gpa_t gpa = cr2; |
9786 |
++ gpa_t gpa = cr2_or_gpa; |
9787 |
+ kvm_pfn_t pfn; |
9788 |
+ |
9789 |
+ if (!(emulation_type & EMULTYPE_ALLOW_RETRY)) |
9790 |
+@@ -5995,7 +6039,7 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gva_t cr2, |
9791 |
+ * Write permission should be allowed since only |
9792 |
+ * write access need to be emulated. |
9793 |
+ */ |
9794 |
+- gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2, NULL); |
9795 |
++ gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL); |
9796 |
+ |
9797 |
+ /* |
9798 |
+ * If the mapping is invalid in guest, let cpu retry |
9799 |
+@@ -6052,10 +6096,10 @@ static bool reexecute_instruction(struct kvm_vcpu *vcpu, gva_t cr2, |
9800 |
+ } |
9801 |
+ |
9802 |
+ static bool retry_instruction(struct x86_emulate_ctxt *ctxt, |
9803 |
+- unsigned long cr2, int emulation_type) |
9804 |
++ gpa_t cr2_or_gpa, int emulation_type) |
9805 |
+ { |
9806 |
+ struct kvm_vcpu *vcpu = emul_to_vcpu(ctxt); |
9807 |
+- unsigned long last_retry_eip, last_retry_addr, gpa = cr2; |
9808 |
++ unsigned long last_retry_eip, last_retry_addr, gpa = cr2_or_gpa; |
9809 |
+ |
9810 |
+ last_retry_eip = vcpu->arch.last_retry_eip; |
9811 |
+ last_retry_addr = vcpu->arch.last_retry_addr; |
9812 |
+@@ -6084,14 +6128,14 @@ static bool retry_instruction(struct x86_emulate_ctxt *ctxt, |
9813 |
+ if (x86_page_table_writing_insn(ctxt)) |
9814 |
+ return false; |
9815 |
+ |
9816 |
+- if (ctxt->eip == last_retry_eip && last_retry_addr == cr2) |
9817 |
++ if (ctxt->eip == last_retry_eip && last_retry_addr == cr2_or_gpa) |
9818 |
+ return false; |
9819 |
+ |
9820 |
+ vcpu->arch.last_retry_eip = ctxt->eip; |
9821 |
+- vcpu->arch.last_retry_addr = cr2; |
9822 |
++ vcpu->arch.last_retry_addr = cr2_or_gpa; |
9823 |
+ |
9824 |
+ if (!vcpu->arch.mmu.direct_map) |
9825 |
+- gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2, NULL); |
9826 |
++ gpa = kvm_mmu_gva_to_gpa_write(vcpu, cr2_or_gpa, NULL); |
9827 |
+ |
9828 |
+ kvm_mmu_unprotect_page(vcpu->kvm, gpa_to_gfn(gpa)); |
9829 |
+ |
9830 |
+@@ -6252,11 +6296,8 @@ static bool is_vmware_backdoor_opcode(struct x86_emulate_ctxt *ctxt) |
9831 |
+ return false; |
9832 |
+ } |
9833 |
+ |
9834 |
+-int x86_emulate_instruction(struct kvm_vcpu *vcpu, |
9835 |
+- unsigned long cr2, |
9836 |
+- int emulation_type, |
9837 |
+- void *insn, |
9838 |
+- int insn_len) |
9839 |
++int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
9840 |
++ int emulation_type, void *insn, int insn_len) |
9841 |
+ { |
9842 |
+ int r; |
9843 |
+ struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt; |
9844 |
+@@ -6299,7 +6340,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, |
9845 |
+ if (r != EMULATION_OK) { |
9846 |
+ if (emulation_type & EMULTYPE_TRAP_UD) |
9847 |
+ return EMULATE_FAIL; |
9848 |
+- if (reexecute_instruction(vcpu, cr2, write_fault_to_spt, |
9849 |
++ if (reexecute_instruction(vcpu, cr2_or_gpa, write_fault_to_spt, |
9850 |
+ emulation_type)) |
9851 |
+ return EMULATE_DONE; |
9852 |
+ if (ctxt->have_exception) { |
9853 |
+@@ -6329,7 +6370,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, |
9854 |
+ return EMULATE_DONE; |
9855 |
+ } |
9856 |
+ |
9857 |
+- if (retry_instruction(ctxt, cr2, emulation_type)) |
9858 |
++ if (retry_instruction(ctxt, cr2_or_gpa, emulation_type)) |
9859 |
+ return EMULATE_DONE; |
9860 |
+ |
9861 |
+ /* this is needed for vmware backdoor interface to work since it |
9862 |
+@@ -6341,7 +6382,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu, |
9863 |
+ |
9864 |
+ restart: |
9865 |
+ /* Save the faulting GPA (cr2) in the address field */ |
9866 |
+- ctxt->exception.address = cr2; |
9867 |
++ ctxt->exception.address = cr2_or_gpa; |
9868 |
+ |
9869 |
+ r = x86_emulate_insn(ctxt); |
9870 |
+ |
9871 |
+@@ -6349,7 +6390,7 @@ restart: |
9872 |
+ return EMULATE_DONE; |
9873 |
+ |
9874 |
+ if (r == EMULATION_FAILED) { |
9875 |
+- if (reexecute_instruction(vcpu, cr2, write_fault_to_spt, |
9876 |
++ if (reexecute_instruction(vcpu, cr2_or_gpa, write_fault_to_spt, |
9877 |
+ emulation_type)) |
9878 |
+ return EMULATE_DONE; |
9879 |
+ |
9880 |
+@@ -6753,7 +6794,7 @@ static void kvm_set_mmio_spte_mask(void) |
9881 |
+ * If reserved bit is not supported, clear the present bit to disable |
9882 |
+ * mmio page fault. |
9883 |
+ */ |
9884 |
+- if (IS_ENABLED(CONFIG_X86_64) && maxphyaddr == 52) |
9885 |
++ if (maxphyaddr == 52) |
9886 |
+ mask &= ~1ull; |
9887 |
+ |
9888 |
+ kvm_mmu_set_mmio_spte_mask(mask, mask); |
9889 |
+@@ -8225,6 +8266,8 @@ int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
9890 |
+ struct kvm_mp_state *mp_state) |
9891 |
+ { |
9892 |
+ vcpu_load(vcpu); |
9893 |
++ if (kvm_mpx_supported()) |
9894 |
++ kvm_load_guest_fpu(vcpu); |
9895 |
+ |
9896 |
+ kvm_apic_accept_events(vcpu); |
9897 |
+ if (vcpu->arch.mp_state == KVM_MP_STATE_HALTED && |
9898 |
+@@ -8233,6 +8276,8 @@ int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
9899 |
+ else |
9900 |
+ mp_state->mp_state = vcpu->arch.mp_state; |
9901 |
+ |
9902 |
++ if (kvm_mpx_supported()) |
9903 |
++ kvm_put_guest_fpu(vcpu); |
9904 |
+ vcpu_put(vcpu); |
9905 |
+ return 0; |
9906 |
+ } |
9907 |
+@@ -8654,7 +8699,7 @@ void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) |
9908 |
+ kvm_mmu_unload(vcpu); |
9909 |
+ vcpu_put(vcpu); |
9910 |
+ |
9911 |
+- kvm_x86_ops->vcpu_free(vcpu); |
9912 |
++ kvm_arch_vcpu_free(vcpu); |
9913 |
+ } |
9914 |
+ |
9915 |
+ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event) |
9916 |
+@@ -8847,6 +8892,8 @@ int kvm_arch_hardware_setup(void) |
9917 |
+ if (r != 0) |
9918 |
+ return r; |
9919 |
+ |
9920 |
++ cr4_reserved_bits = kvm_host_cr4_reserved_bits(&boot_cpu_data); |
9921 |
++ |
9922 |
+ if (kvm_has_tsc_control) { |
9923 |
+ /* |
9924 |
+ * Make sure the user can only configure tsc_khz values that |
9925 |
+@@ -9505,7 +9552,7 @@ void kvm_arch_async_page_ready(struct kvm_vcpu *vcpu, struct kvm_async_pf *work) |
9926 |
+ work->arch.cr3 != vcpu->arch.mmu.get_cr3(vcpu)) |
9927 |
+ return; |
9928 |
+ |
9929 |
+- vcpu->arch.mmu.page_fault(vcpu, work->gva, 0, true); |
9930 |
++ vcpu->arch.mmu.page_fault(vcpu, work->cr2_or_gpa, 0, true); |
9931 |
+ } |
9932 |
+ |
9933 |
+ static inline u32 kvm_async_pf_hash_fn(gfn_t gfn) |
9934 |
+@@ -9588,7 +9635,7 @@ void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu, |
9935 |
+ { |
9936 |
+ struct x86_exception fault; |
9937 |
+ |
9938 |
+- trace_kvm_async_pf_not_present(work->arch.token, work->gva); |
9939 |
++ trace_kvm_async_pf_not_present(work->arch.token, work->cr2_or_gpa); |
9940 |
+ kvm_add_async_pf_gfn(vcpu, work->arch.gfn); |
9941 |
+ |
9942 |
+ if (!(vcpu->arch.apf.msr_val & KVM_ASYNC_PF_ENABLED) || |
9943 |
+@@ -9616,7 +9663,7 @@ void kvm_arch_async_page_present(struct kvm_vcpu *vcpu, |
9944 |
+ work->arch.token = ~0; /* broadcast wakeup */ |
9945 |
+ else |
9946 |
+ kvm_del_async_pf_gfn(vcpu, work->arch.gfn); |
9947 |
+- trace_kvm_async_pf_ready(work->arch.token, work->gva); |
9948 |
++ trace_kvm_async_pf_ready(work->arch.token, work->cr2_or_gpa); |
9949 |
+ |
9950 |
+ if (vcpu->arch.apf.msr_val & KVM_ASYNC_PF_ENABLED && |
9951 |
+ !apf_get_user(vcpu, &val)) { |
9952 |
+diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h |
9953 |
+index 608e5f8c5d0a..422331b257d3 100644 |
9954 |
+--- a/arch/x86/kvm/x86.h |
9955 |
++++ b/arch/x86/kvm/x86.h |
9956 |
+@@ -284,7 +284,7 @@ int kvm_mtrr_get_msr(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata); |
9957 |
+ bool kvm_mtrr_check_gfn_range_consistency(struct kvm_vcpu *vcpu, gfn_t gfn, |
9958 |
+ int page_num); |
9959 |
+ bool kvm_vector_hashing_enabled(void); |
9960 |
+-int x86_emulate_instruction(struct kvm_vcpu *vcpu, unsigned long cr2, |
9961 |
++int x86_emulate_instruction(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
9962 |
+ int emulation_type, void *insn, int insn_len); |
9963 |
+ |
9964 |
+ #define KVM_SUPPORTED_XCR0 (XFEATURE_MASK_FP | XFEATURE_MASK_SSE \ |
9965 |
+diff --git a/crypto/algapi.c b/crypto/algapi.c |
9966 |
+index c0755cf4f53f..346557a3fc0b 100644 |
9967 |
+--- a/crypto/algapi.c |
9968 |
++++ b/crypto/algapi.c |
9969 |
+@@ -649,11 +649,9 @@ EXPORT_SYMBOL_GPL(crypto_grab_spawn); |
9970 |
+ |
9971 |
+ void crypto_drop_spawn(struct crypto_spawn *spawn) |
9972 |
+ { |
9973 |
+- if (!spawn->alg) |
9974 |
+- return; |
9975 |
+- |
9976 |
+ down_write(&crypto_alg_sem); |
9977 |
+- list_del(&spawn->list); |
9978 |
++ if (spawn->alg) |
9979 |
++ list_del(&spawn->list); |
9980 |
+ up_write(&crypto_alg_sem); |
9981 |
+ } |
9982 |
+ EXPORT_SYMBOL_GPL(crypto_drop_spawn); |
9983 |
+@@ -661,22 +659,16 @@ EXPORT_SYMBOL_GPL(crypto_drop_spawn); |
9984 |
+ static struct crypto_alg *crypto_spawn_alg(struct crypto_spawn *spawn) |
9985 |
+ { |
9986 |
+ struct crypto_alg *alg; |
9987 |
+- struct crypto_alg *alg2; |
9988 |
+ |
9989 |
+ down_read(&crypto_alg_sem); |
9990 |
+ alg = spawn->alg; |
9991 |
+- alg2 = alg; |
9992 |
+- if (alg2) |
9993 |
+- alg2 = crypto_mod_get(alg2); |
9994 |
+- up_read(&crypto_alg_sem); |
9995 |
+- |
9996 |
+- if (!alg2) { |
9997 |
+- if (alg) |
9998 |
+- crypto_shoot_alg(alg); |
9999 |
+- return ERR_PTR(-EAGAIN); |
10000 |
++ if (alg && !crypto_mod_get(alg)) { |
10001 |
++ alg->cra_flags |= CRYPTO_ALG_DYING; |
10002 |
++ alg = NULL; |
10003 |
+ } |
10004 |
++ up_read(&crypto_alg_sem); |
10005 |
+ |
10006 |
+- return alg; |
10007 |
++ return alg ?: ERR_PTR(-EAGAIN); |
10008 |
+ } |
10009 |
+ |
10010 |
+ struct crypto_tfm *crypto_spawn_tfm(struct crypto_spawn *spawn, u32 type, |
10011 |
+diff --git a/crypto/api.c b/crypto/api.c |
10012 |
+index 7aca9f86c5f3..1909195b2c70 100644 |
10013 |
+--- a/crypto/api.c |
10014 |
++++ b/crypto/api.c |
10015 |
+@@ -349,13 +349,12 @@ static unsigned int crypto_ctxsize(struct crypto_alg *alg, u32 type, u32 mask) |
10016 |
+ return len; |
10017 |
+ } |
10018 |
+ |
10019 |
+-void crypto_shoot_alg(struct crypto_alg *alg) |
10020 |
++static void crypto_shoot_alg(struct crypto_alg *alg) |
10021 |
+ { |
10022 |
+ down_write(&crypto_alg_sem); |
10023 |
+ alg->cra_flags |= CRYPTO_ALG_DYING; |
10024 |
+ up_write(&crypto_alg_sem); |
10025 |
+ } |
10026 |
+-EXPORT_SYMBOL_GPL(crypto_shoot_alg); |
10027 |
+ |
10028 |
+ struct crypto_tfm *__crypto_alloc_tfm(struct crypto_alg *alg, u32 type, |
10029 |
+ u32 mask) |
10030 |
+diff --git a/crypto/internal.h b/crypto/internal.h |
10031 |
+index 9a3f39939fba..f8d6efaffef9 100644 |
10032 |
+--- a/crypto/internal.h |
10033 |
++++ b/crypto/internal.h |
10034 |
+@@ -79,7 +79,6 @@ void crypto_alg_tested(const char *name, int err); |
10035 |
+ void crypto_remove_spawns(struct crypto_alg *alg, struct list_head *list, |
10036 |
+ struct crypto_alg *nalg); |
10037 |
+ void crypto_remove_final(struct list_head *list); |
10038 |
+-void crypto_shoot_alg(struct crypto_alg *alg); |
10039 |
+ struct crypto_tfm *__crypto_alloc_tfm(struct crypto_alg *alg, u32 type, |
10040 |
+ u32 mask); |
10041 |
+ void *crypto_create_tfm(struct crypto_alg *alg, |
10042 |
+diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c |
10043 |
+index 1348541da463..85082574c515 100644 |
10044 |
+--- a/crypto/pcrypt.c |
10045 |
++++ b/crypto/pcrypt.c |
10046 |
+@@ -130,7 +130,6 @@ static void pcrypt_aead_done(struct crypto_async_request *areq, int err) |
10047 |
+ struct padata_priv *padata = pcrypt_request_padata(preq); |
10048 |
+ |
10049 |
+ padata->info = err; |
10050 |
+- req->base.flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; |
10051 |
+ |
10052 |
+ padata_do_serial(padata); |
10053 |
+ } |
10054 |
+diff --git a/drivers/acpi/battery.c b/drivers/acpi/battery.c |
10055 |
+index cb97b6105f52..674a0e92b798 100644 |
10056 |
+--- a/drivers/acpi/battery.c |
10057 |
++++ b/drivers/acpi/battery.c |
10058 |
+@@ -51,6 +51,8 @@ |
10059 |
+ #define PREFIX "ACPI: " |
10060 |
+ |
10061 |
+ #define ACPI_BATTERY_VALUE_UNKNOWN 0xFFFFFFFF |
10062 |
++#define ACPI_BATTERY_CAPACITY_VALID(capacity) \ |
10063 |
++ ((capacity) != 0 && (capacity) != ACPI_BATTERY_VALUE_UNKNOWN) |
10064 |
+ |
10065 |
+ #define ACPI_BATTERY_DEVICE_NAME "Battery" |
10066 |
+ |
10067 |
+@@ -205,7 +207,8 @@ static int acpi_battery_is_charged(struct acpi_battery *battery) |
10068 |
+ |
10069 |
+ static bool acpi_battery_is_degraded(struct acpi_battery *battery) |
10070 |
+ { |
10071 |
+- return battery->full_charge_capacity && battery->design_capacity && |
10072 |
++ return ACPI_BATTERY_CAPACITY_VALID(battery->full_charge_capacity) && |
10073 |
++ ACPI_BATTERY_CAPACITY_VALID(battery->design_capacity) && |
10074 |
+ battery->full_charge_capacity < battery->design_capacity; |
10075 |
+ } |
10076 |
+ |
10077 |
+@@ -227,7 +230,7 @@ static int acpi_battery_get_property(struct power_supply *psy, |
10078 |
+ enum power_supply_property psp, |
10079 |
+ union power_supply_propval *val) |
10080 |
+ { |
10081 |
+- int ret = 0; |
10082 |
++ int full_capacity = ACPI_BATTERY_VALUE_UNKNOWN, ret = 0; |
10083 |
+ struct acpi_battery *battery = to_acpi_battery(psy); |
10084 |
+ |
10085 |
+ if (acpi_battery_present(battery)) { |
10086 |
+@@ -276,14 +279,14 @@ static int acpi_battery_get_property(struct power_supply *psy, |
10087 |
+ break; |
10088 |
+ case POWER_SUPPLY_PROP_CHARGE_FULL_DESIGN: |
10089 |
+ case POWER_SUPPLY_PROP_ENERGY_FULL_DESIGN: |
10090 |
+- if (battery->design_capacity == ACPI_BATTERY_VALUE_UNKNOWN) |
10091 |
++ if (!ACPI_BATTERY_CAPACITY_VALID(battery->design_capacity)) |
10092 |
+ ret = -ENODEV; |
10093 |
+ else |
10094 |
+ val->intval = battery->design_capacity * 1000; |
10095 |
+ break; |
10096 |
+ case POWER_SUPPLY_PROP_CHARGE_FULL: |
10097 |
+ case POWER_SUPPLY_PROP_ENERGY_FULL: |
10098 |
+- if (battery->full_charge_capacity == ACPI_BATTERY_VALUE_UNKNOWN) |
10099 |
++ if (!ACPI_BATTERY_CAPACITY_VALID(battery->full_charge_capacity)) |
10100 |
+ ret = -ENODEV; |
10101 |
+ else |
10102 |
+ val->intval = battery->full_charge_capacity * 1000; |
10103 |
+@@ -296,11 +299,17 @@ static int acpi_battery_get_property(struct power_supply *psy, |
10104 |
+ val->intval = battery->capacity_now * 1000; |
10105 |
+ break; |
10106 |
+ case POWER_SUPPLY_PROP_CAPACITY: |
10107 |
+- if (battery->capacity_now && battery->full_charge_capacity) |
10108 |
+- val->intval = battery->capacity_now * 100/ |
10109 |
+- battery->full_charge_capacity; |
10110 |
++ if (ACPI_BATTERY_CAPACITY_VALID(battery->full_charge_capacity)) |
10111 |
++ full_capacity = battery->full_charge_capacity; |
10112 |
++ else if (ACPI_BATTERY_CAPACITY_VALID(battery->design_capacity)) |
10113 |
++ full_capacity = battery->design_capacity; |
10114 |
++ |
10115 |
++ if (battery->capacity_now == ACPI_BATTERY_VALUE_UNKNOWN || |
10116 |
++ full_capacity == ACPI_BATTERY_VALUE_UNKNOWN) |
10117 |
++ ret = -ENODEV; |
10118 |
+ else |
10119 |
+- val->intval = 0; |
10120 |
++ val->intval = battery->capacity_now * 100/ |
10121 |
++ full_capacity; |
10122 |
+ break; |
10123 |
+ case POWER_SUPPLY_PROP_CAPACITY_LEVEL: |
10124 |
+ if (battery->state & ACPI_BATTERY_STATE_CRITICAL) |
10125 |
+@@ -346,6 +355,20 @@ static enum power_supply_property charge_battery_props[] = { |
10126 |
+ POWER_SUPPLY_PROP_SERIAL_NUMBER, |
10127 |
+ }; |
10128 |
+ |
10129 |
++static enum power_supply_property charge_battery_full_cap_broken_props[] = { |
10130 |
++ POWER_SUPPLY_PROP_STATUS, |
10131 |
++ POWER_SUPPLY_PROP_PRESENT, |
10132 |
++ POWER_SUPPLY_PROP_TECHNOLOGY, |
10133 |
++ POWER_SUPPLY_PROP_CYCLE_COUNT, |
10134 |
++ POWER_SUPPLY_PROP_VOLTAGE_MIN_DESIGN, |
10135 |
++ POWER_SUPPLY_PROP_VOLTAGE_NOW, |
10136 |
++ POWER_SUPPLY_PROP_CURRENT_NOW, |
10137 |
++ POWER_SUPPLY_PROP_CHARGE_NOW, |
10138 |
++ POWER_SUPPLY_PROP_MODEL_NAME, |
10139 |
++ POWER_SUPPLY_PROP_MANUFACTURER, |
10140 |
++ POWER_SUPPLY_PROP_SERIAL_NUMBER, |
10141 |
++}; |
10142 |
++ |
10143 |
+ static enum power_supply_property energy_battery_props[] = { |
10144 |
+ POWER_SUPPLY_PROP_STATUS, |
10145 |
+ POWER_SUPPLY_PROP_PRESENT, |
10146 |
+@@ -807,20 +830,34 @@ static void __exit battery_hook_exit(void) |
10147 |
+ static int sysfs_add_battery(struct acpi_battery *battery) |
10148 |
+ { |
10149 |
+ struct power_supply_config psy_cfg = { .drv_data = battery, }; |
10150 |
++ bool full_cap_broken = false; |
10151 |
++ |
10152 |
++ if (!ACPI_BATTERY_CAPACITY_VALID(battery->full_charge_capacity) && |
10153 |
++ !ACPI_BATTERY_CAPACITY_VALID(battery->design_capacity)) |
10154 |
++ full_cap_broken = true; |
10155 |
+ |
10156 |
+ if (battery->power_unit == ACPI_BATTERY_POWER_UNIT_MA) { |
10157 |
+- battery->bat_desc.properties = charge_battery_props; |
10158 |
+- battery->bat_desc.num_properties = |
10159 |
+- ARRAY_SIZE(charge_battery_props); |
10160 |
+- } else if (battery->full_charge_capacity == 0) { |
10161 |
+- battery->bat_desc.properties = |
10162 |
+- energy_battery_full_cap_broken_props; |
10163 |
+- battery->bat_desc.num_properties = |
10164 |
+- ARRAY_SIZE(energy_battery_full_cap_broken_props); |
10165 |
++ if (full_cap_broken) { |
10166 |
++ battery->bat_desc.properties = |
10167 |
++ charge_battery_full_cap_broken_props; |
10168 |
++ battery->bat_desc.num_properties = |
10169 |
++ ARRAY_SIZE(charge_battery_full_cap_broken_props); |
10170 |
++ } else { |
10171 |
++ battery->bat_desc.properties = charge_battery_props; |
10172 |
++ battery->bat_desc.num_properties = |
10173 |
++ ARRAY_SIZE(charge_battery_props); |
10174 |
++ } |
10175 |
+ } else { |
10176 |
+- battery->bat_desc.properties = energy_battery_props; |
10177 |
+- battery->bat_desc.num_properties = |
10178 |
+- ARRAY_SIZE(energy_battery_props); |
10179 |
++ if (full_cap_broken) { |
10180 |
++ battery->bat_desc.properties = |
10181 |
++ energy_battery_full_cap_broken_props; |
10182 |
++ battery->bat_desc.num_properties = |
10183 |
++ ARRAY_SIZE(energy_battery_full_cap_broken_props); |
10184 |
++ } else { |
10185 |
++ battery->bat_desc.properties = energy_battery_props; |
10186 |
++ battery->bat_desc.num_properties = |
10187 |
++ ARRAY_SIZE(energy_battery_props); |
10188 |
++ } |
10189 |
+ } |
10190 |
+ |
10191 |
+ battery->bat_desc.name = acpi_device_bid(battery->device); |
10192 |
+diff --git a/drivers/acpi/video_detect.c b/drivers/acpi/video_detect.c |
10193 |
+index 43587ac680e4..214c4e2e8ade 100644 |
10194 |
+--- a/drivers/acpi/video_detect.c |
10195 |
++++ b/drivers/acpi/video_detect.c |
10196 |
+@@ -328,6 +328,11 @@ static const struct dmi_system_id video_detect_dmi_table[] = { |
10197 |
+ DMI_MATCH(DMI_PRODUCT_NAME, "Precision 7510"), |
10198 |
+ }, |
10199 |
+ }, |
10200 |
++ |
10201 |
++ /* |
10202 |
++ * Desktops which falsely report a backlight and which our heuristics |
10203 |
++ * for this do not catch. |
10204 |
++ */ |
10205 |
+ { |
10206 |
+ .callback = video_detect_force_none, |
10207 |
+ .ident = "Dell OptiPlex 9020M", |
10208 |
+@@ -336,6 +341,14 @@ static const struct dmi_system_id video_detect_dmi_table[] = { |
10209 |
+ DMI_MATCH(DMI_PRODUCT_NAME, "OptiPlex 9020M"), |
10210 |
+ }, |
10211 |
+ }, |
10212 |
++ { |
10213 |
++ .callback = video_detect_force_none, |
10214 |
++ .ident = "MSI MS-7721", |
10215 |
++ .matches = { |
10216 |
++ DMI_MATCH(DMI_SYS_VENDOR, "MSI"), |
10217 |
++ DMI_MATCH(DMI_PRODUCT_NAME, "MS-7721"), |
10218 |
++ }, |
10219 |
++ }, |
10220 |
+ { }, |
10221 |
+ }; |
10222 |
+ |
10223 |
+diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c |
10224 |
+index 4abd7c6531d9..3b382a7e07b2 100644 |
10225 |
+--- a/drivers/base/power/main.c |
10226 |
++++ b/drivers/base/power/main.c |
10227 |
+@@ -265,10 +265,38 @@ static void dpm_wait_for_suppliers(struct device *dev, bool async) |
10228 |
+ device_links_read_unlock(idx); |
10229 |
+ } |
10230 |
+ |
10231 |
+-static void dpm_wait_for_superior(struct device *dev, bool async) |
10232 |
++static bool dpm_wait_for_superior(struct device *dev, bool async) |
10233 |
+ { |
10234 |
+- dpm_wait(dev->parent, async); |
10235 |
++ struct device *parent; |
10236 |
++ |
10237 |
++ /* |
10238 |
++ * If the device is resumed asynchronously and the parent's callback |
10239 |
++ * deletes both the device and the parent itself, the parent object may |
10240 |
++ * be freed while this function is running, so avoid that by reference |
10241 |
++ * counting the parent once more unless the device has been deleted |
10242 |
++ * already (in which case return right away). |
10243 |
++ */ |
10244 |
++ mutex_lock(&dpm_list_mtx); |
10245 |
++ |
10246 |
++ if (!device_pm_initialized(dev)) { |
10247 |
++ mutex_unlock(&dpm_list_mtx); |
10248 |
++ return false; |
10249 |
++ } |
10250 |
++ |
10251 |
++ parent = get_device(dev->parent); |
10252 |
++ |
10253 |
++ mutex_unlock(&dpm_list_mtx); |
10254 |
++ |
10255 |
++ dpm_wait(parent, async); |
10256 |
++ put_device(parent); |
10257 |
++ |
10258 |
+ dpm_wait_for_suppliers(dev, async); |
10259 |
++ |
10260 |
++ /* |
10261 |
++ * If the parent's callback has deleted the device, attempting to resume |
10262 |
++ * it would be invalid, so avoid doing that then. |
10263 |
++ */ |
10264 |
++ return device_pm_initialized(dev); |
10265 |
+ } |
10266 |
+ |
10267 |
+ static void dpm_wait_for_consumers(struct device *dev, bool async) |
10268 |
+@@ -628,7 +656,8 @@ static int device_resume_noirq(struct device *dev, pm_message_t state, bool asyn |
10269 |
+ if (!dev->power.is_noirq_suspended) |
10270 |
+ goto Out; |
10271 |
+ |
10272 |
+- dpm_wait_for_superior(dev, async); |
10273 |
++ if (!dpm_wait_for_superior(dev, async)) |
10274 |
++ goto Out; |
10275 |
+ |
10276 |
+ skip_resume = dev_pm_may_skip_resume(dev); |
10277 |
+ |
10278 |
+@@ -829,7 +858,8 @@ static int device_resume_early(struct device *dev, pm_message_t state, bool asyn |
10279 |
+ if (!dev->power.is_late_suspended) |
10280 |
+ goto Out; |
10281 |
+ |
10282 |
+- dpm_wait_for_superior(dev, async); |
10283 |
++ if (!dpm_wait_for_superior(dev, async)) |
10284 |
++ goto Out; |
10285 |
+ |
10286 |
+ callback = dpm_subsys_resume_early_cb(dev, state, &info); |
10287 |
+ |
10288 |
+@@ -949,7 +979,9 @@ static int device_resume(struct device *dev, pm_message_t state, bool async) |
10289 |
+ goto Complete; |
10290 |
+ } |
10291 |
+ |
10292 |
+- dpm_wait_for_superior(dev, async); |
10293 |
++ if (!dpm_wait_for_superior(dev, async)) |
10294 |
++ goto Complete; |
10295 |
++ |
10296 |
+ dpm_watchdog_set(&wd, dev); |
10297 |
+ device_lock(dev); |
10298 |
+ |
10299 |
+diff --git a/drivers/clk/tegra/clk-tegra-periph.c b/drivers/clk/tegra/clk-tegra-periph.c |
10300 |
+index 38c4eb28c8bf..b137c5d34eec 100644 |
10301 |
+--- a/drivers/clk/tegra/clk-tegra-periph.c |
10302 |
++++ b/drivers/clk/tegra/clk-tegra-periph.c |
10303 |
+@@ -799,7 +799,11 @@ static struct tegra_periph_init_data gate_clks[] = { |
10304 |
+ GATE("ahbdma", "hclk", 33, 0, tegra_clk_ahbdma, 0), |
10305 |
+ GATE("apbdma", "pclk", 34, 0, tegra_clk_apbdma, 0), |
10306 |
+ GATE("kbc", "clk_32k", 36, TEGRA_PERIPH_ON_APB | TEGRA_PERIPH_NO_RESET, tegra_clk_kbc, 0), |
10307 |
+- GATE("fuse", "clk_m", 39, TEGRA_PERIPH_ON_APB, tegra_clk_fuse, 0), |
10308 |
++ /* |
10309 |
++ * Critical for RAM re-repair operation, which must occur on resume |
10310 |
++ * from LP1 system suspend and as part of CCPLEX cluster switching. |
10311 |
++ */ |
10312 |
++ GATE("fuse", "clk_m", 39, TEGRA_PERIPH_ON_APB, tegra_clk_fuse, CLK_IS_CRITICAL), |
10313 |
+ GATE("fuse_burn", "clk_m", 39, TEGRA_PERIPH_ON_APB, tegra_clk_fuse_burn, 0), |
10314 |
+ GATE("kfuse", "clk_m", 40, TEGRA_PERIPH_ON_APB, tegra_clk_kfuse, 0), |
10315 |
+ GATE("apbif", "clk_m", 107, TEGRA_PERIPH_ON_APB, tegra_clk_apbif, 0), |
10316 |
+diff --git a/drivers/crypto/atmel-aes.c b/drivers/crypto/atmel-aes.c |
10317 |
+index 53a78035381d..92060864e356 100644 |
10318 |
+--- a/drivers/crypto/atmel-aes.c |
10319 |
++++ b/drivers/crypto/atmel-aes.c |
10320 |
+@@ -91,7 +91,6 @@ |
10321 |
+ struct atmel_aes_caps { |
10322 |
+ bool has_dualbuff; |
10323 |
+ bool has_cfb64; |
10324 |
+- bool has_ctr32; |
10325 |
+ bool has_gcm; |
10326 |
+ bool has_xts; |
10327 |
+ bool has_authenc; |
10328 |
+@@ -1016,8 +1015,9 @@ static int atmel_aes_ctr_transfer(struct atmel_aes_dev *dd) |
10329 |
+ struct atmel_aes_ctr_ctx *ctx = atmel_aes_ctr_ctx_cast(dd->ctx); |
10330 |
+ struct ablkcipher_request *req = ablkcipher_request_cast(dd->areq); |
10331 |
+ struct scatterlist *src, *dst; |
10332 |
+- u32 ctr, blocks; |
10333 |
+ size_t datalen; |
10334 |
++ u32 ctr; |
10335 |
++ u16 blocks, start, end; |
10336 |
+ bool use_dma, fragmented = false; |
10337 |
+ |
10338 |
+ /* Check for transfer completion. */ |
10339 |
+@@ -1029,27 +1029,17 @@ static int atmel_aes_ctr_transfer(struct atmel_aes_dev *dd) |
10340 |
+ datalen = req->nbytes - ctx->offset; |
10341 |
+ blocks = DIV_ROUND_UP(datalen, AES_BLOCK_SIZE); |
10342 |
+ ctr = be32_to_cpu(ctx->iv[3]); |
10343 |
+- if (dd->caps.has_ctr32) { |
10344 |
+- /* Check 32bit counter overflow. */ |
10345 |
+- u32 start = ctr; |
10346 |
+- u32 end = start + blocks - 1; |
10347 |
+- |
10348 |
+- if (end < start) { |
10349 |
+- ctr |= 0xffffffff; |
10350 |
+- datalen = AES_BLOCK_SIZE * -start; |
10351 |
+- fragmented = true; |
10352 |
+- } |
10353 |
+- } else { |
10354 |
+- /* Check 16bit counter overflow. */ |
10355 |
+- u16 start = ctr & 0xffff; |
10356 |
+- u16 end = start + (u16)blocks - 1; |
10357 |
+- |
10358 |
+- if (blocks >> 16 || end < start) { |
10359 |
+- ctr |= 0xffff; |
10360 |
+- datalen = AES_BLOCK_SIZE * (0x10000-start); |
10361 |
+- fragmented = true; |
10362 |
+- } |
10363 |
++ |
10364 |
++ /* Check 16bit counter overflow. */ |
10365 |
++ start = ctr & 0xffff; |
10366 |
++ end = start + blocks - 1; |
10367 |
++ |
10368 |
++ if (blocks >> 16 || end < start) { |
10369 |
++ ctr |= 0xffff; |
10370 |
++ datalen = AES_BLOCK_SIZE * (0x10000 - start); |
10371 |
++ fragmented = true; |
10372 |
+ } |
10373 |
++ |
10374 |
+ use_dma = (datalen >= ATMEL_AES_DMA_THRESHOLD); |
10375 |
+ |
10376 |
+ /* Jump to offset. */ |
10377 |
+@@ -2553,7 +2543,6 @@ static void atmel_aes_get_cap(struct atmel_aes_dev *dd) |
10378 |
+ { |
10379 |
+ dd->caps.has_dualbuff = 0; |
10380 |
+ dd->caps.has_cfb64 = 0; |
10381 |
+- dd->caps.has_ctr32 = 0; |
10382 |
+ dd->caps.has_gcm = 0; |
10383 |
+ dd->caps.has_xts = 0; |
10384 |
+ dd->caps.has_authenc = 0; |
10385 |
+@@ -2564,7 +2553,6 @@ static void atmel_aes_get_cap(struct atmel_aes_dev *dd) |
10386 |
+ case 0x500: |
10387 |
+ dd->caps.has_dualbuff = 1; |
10388 |
+ dd->caps.has_cfb64 = 1; |
10389 |
+- dd->caps.has_ctr32 = 1; |
10390 |
+ dd->caps.has_gcm = 1; |
10391 |
+ dd->caps.has_xts = 1; |
10392 |
+ dd->caps.has_authenc = 1; |
10393 |
+@@ -2573,7 +2561,6 @@ static void atmel_aes_get_cap(struct atmel_aes_dev *dd) |
10394 |
+ case 0x200: |
10395 |
+ dd->caps.has_dualbuff = 1; |
10396 |
+ dd->caps.has_cfb64 = 1; |
10397 |
+- dd->caps.has_ctr32 = 1; |
10398 |
+ dd->caps.has_gcm = 1; |
10399 |
+ dd->caps.max_burst_size = 4; |
10400 |
+ break; |
10401 |
+diff --git a/drivers/crypto/ccp/ccp-dev-v3.c b/drivers/crypto/ccp/ccp-dev-v3.c |
10402 |
+index 240bebbcb8ac..ae0cc0a4dc5c 100644 |
10403 |
+--- a/drivers/crypto/ccp/ccp-dev-v3.c |
10404 |
++++ b/drivers/crypto/ccp/ccp-dev-v3.c |
10405 |
+@@ -590,6 +590,7 @@ const struct ccp_vdata ccpv3_platform = { |
10406 |
+ .setup = NULL, |
10407 |
+ .perform = &ccp3_actions, |
10408 |
+ .offset = 0, |
10409 |
++ .rsamax = CCP_RSA_MAX_WIDTH, |
10410 |
+ }; |
10411 |
+ |
10412 |
+ const struct ccp_vdata ccpv3 = { |
10413 |
+diff --git a/drivers/crypto/ccree/cc_driver.h b/drivers/crypto/ccree/cc_driver.h |
10414 |
+index be7f9bd5c559..d41193932207 100644 |
10415 |
+--- a/drivers/crypto/ccree/cc_driver.h |
10416 |
++++ b/drivers/crypto/ccree/cc_driver.h |
10417 |
+@@ -131,6 +131,7 @@ struct cc_drvdata { |
10418 |
+ u32 axim_mon_offset; |
10419 |
+ u32 sig_offset; |
10420 |
+ u32 ver_offset; |
10421 |
++ bool pm_on; |
10422 |
+ }; |
10423 |
+ |
10424 |
+ struct cc_crypto_alg { |
10425 |
+diff --git a/drivers/crypto/ccree/cc_pm.c b/drivers/crypto/ccree/cc_pm.c |
10426 |
+index 638082dff183..2df2c2ca8aae 100644 |
10427 |
+--- a/drivers/crypto/ccree/cc_pm.c |
10428 |
++++ b/drivers/crypto/ccree/cc_pm.c |
10429 |
+@@ -23,14 +23,8 @@ const struct dev_pm_ops ccree_pm = { |
10430 |
+ int cc_pm_suspend(struct device *dev) |
10431 |
+ { |
10432 |
+ struct cc_drvdata *drvdata = dev_get_drvdata(dev); |
10433 |
+- int rc; |
10434 |
+ |
10435 |
+ dev_dbg(dev, "set HOST_POWER_DOWN_EN\n"); |
10436 |
+- rc = cc_suspend_req_queue(drvdata); |
10437 |
+- if (rc) { |
10438 |
+- dev_err(dev, "cc_suspend_req_queue (%x)\n", rc); |
10439 |
+- return rc; |
10440 |
+- } |
10441 |
+ fini_cc_regs(drvdata); |
10442 |
+ cc_iowrite(drvdata, CC_REG(HOST_POWER_DOWN_EN), POWER_DOWN_ENABLE); |
10443 |
+ cc_clk_off(drvdata); |
10444 |
+@@ -59,13 +53,6 @@ int cc_pm_resume(struct device *dev) |
10445 |
+ /* check if tee fips error occurred during power down */ |
10446 |
+ cc_tee_handle_fips_error(drvdata); |
10447 |
+ |
10448 |
+- rc = cc_resume_req_queue(drvdata); |
10449 |
+- if (rc) { |
10450 |
+- dev_err(dev, "cc_resume_req_queue (%x)\n", rc); |
10451 |
+- return rc; |
10452 |
+- } |
10453 |
+- |
10454 |
+- /* must be after the queue resuming as it uses the HW queue*/ |
10455 |
+ cc_init_hash_sram(drvdata); |
10456 |
+ |
10457 |
+ cc_init_iv_sram(drvdata); |
10458 |
+@@ -77,12 +64,10 @@ int cc_pm_get(struct device *dev) |
10459 |
+ int rc = 0; |
10460 |
+ struct cc_drvdata *drvdata = dev_get_drvdata(dev); |
10461 |
+ |
10462 |
+- if (cc_req_queue_suspended(drvdata)) |
10463 |
++ if (drvdata->pm_on) |
10464 |
+ rc = pm_runtime_get_sync(dev); |
10465 |
+- else |
10466 |
+- pm_runtime_get_noresume(dev); |
10467 |
+ |
10468 |
+- return rc; |
10469 |
++ return (rc == 1 ? 0 : rc); |
10470 |
+ } |
10471 |
+ |
10472 |
+ int cc_pm_put_suspend(struct device *dev) |
10473 |
+@@ -90,14 +75,11 @@ int cc_pm_put_suspend(struct device *dev) |
10474 |
+ int rc = 0; |
10475 |
+ struct cc_drvdata *drvdata = dev_get_drvdata(dev); |
10476 |
+ |
10477 |
+- if (!cc_req_queue_suspended(drvdata)) { |
10478 |
++ if (drvdata->pm_on) { |
10479 |
+ pm_runtime_mark_last_busy(dev); |
10480 |
+ rc = pm_runtime_put_autosuspend(dev); |
10481 |
+- } else { |
10482 |
+- /* Something wrong happens*/ |
10483 |
+- dev_err(dev, "request to suspend already suspended queue"); |
10484 |
+- rc = -EBUSY; |
10485 |
+ } |
10486 |
++ |
10487 |
+ return rc; |
10488 |
+ } |
10489 |
+ |
10490 |
+@@ -108,7 +90,7 @@ int cc_pm_init(struct cc_drvdata *drvdata) |
10491 |
+ /* must be before the enabling to avoid resdundent suspending */ |
10492 |
+ pm_runtime_set_autosuspend_delay(dev, CC_SUSPEND_TIMEOUT); |
10493 |
+ pm_runtime_use_autosuspend(dev); |
10494 |
+- /* activate the PM module */ |
10495 |
++ /* set us as active - note we won't do PM ops until cc_pm_go()! */ |
10496 |
+ return pm_runtime_set_active(dev); |
10497 |
+ } |
10498 |
+ |
10499 |
+@@ -116,9 +98,11 @@ int cc_pm_init(struct cc_drvdata *drvdata) |
10500 |
+ void cc_pm_go(struct cc_drvdata *drvdata) |
10501 |
+ { |
10502 |
+ pm_runtime_enable(drvdata_to_dev(drvdata)); |
10503 |
++ drvdata->pm_on = true; |
10504 |
+ } |
10505 |
+ |
10506 |
+ void cc_pm_fini(struct cc_drvdata *drvdata) |
10507 |
+ { |
10508 |
+ pm_runtime_disable(drvdata_to_dev(drvdata)); |
10509 |
++ drvdata->pm_on = false; |
10510 |
+ } |
10511 |
+diff --git a/drivers/crypto/ccree/cc_request_mgr.c b/drivers/crypto/ccree/cc_request_mgr.c |
10512 |
+index 83a8aaae61c7..1d88abc6d230 100644 |
10513 |
+--- a/drivers/crypto/ccree/cc_request_mgr.c |
10514 |
++++ b/drivers/crypto/ccree/cc_request_mgr.c |
10515 |
+@@ -41,7 +41,6 @@ struct cc_req_mgr_handle { |
10516 |
+ #else |
10517 |
+ struct tasklet_struct comptask; |
10518 |
+ #endif |
10519 |
+- bool is_runtime_suspended; |
10520 |
+ }; |
10521 |
+ |
10522 |
+ struct cc_bl_item { |
10523 |
+@@ -403,6 +402,7 @@ static void cc_proc_backlog(struct cc_drvdata *drvdata) |
10524 |
+ spin_lock(&mgr->bl_lock); |
10525 |
+ list_del(&bli->list); |
10526 |
+ --mgr->bl_len; |
10527 |
++ kfree(bli); |
10528 |
+ } |
10529 |
+ |
10530 |
+ spin_unlock(&mgr->bl_lock); |
10531 |
+@@ -660,52 +660,3 @@ static void comp_handler(unsigned long devarg) |
10532 |
+ |
10533 |
+ cc_proc_backlog(drvdata); |
10534 |
+ } |
10535 |
+- |
10536 |
+-/* |
10537 |
+- * resume the queue configuration - no need to take the lock as this happens |
10538 |
+- * inside the spin lock protection |
10539 |
+- */ |
10540 |
+-#if defined(CONFIG_PM) |
10541 |
+-int cc_resume_req_queue(struct cc_drvdata *drvdata) |
10542 |
+-{ |
10543 |
+- struct cc_req_mgr_handle *request_mgr_handle = |
10544 |
+- drvdata->request_mgr_handle; |
10545 |
+- |
10546 |
+- spin_lock_bh(&request_mgr_handle->hw_lock); |
10547 |
+- request_mgr_handle->is_runtime_suspended = false; |
10548 |
+- spin_unlock_bh(&request_mgr_handle->hw_lock); |
10549 |
+- |
10550 |
+- return 0; |
10551 |
+-} |
10552 |
+- |
10553 |
+-/* |
10554 |
+- * suspend the queue configuration. Since it is used for the runtime suspend |
10555 |
+- * only verify that the queue can be suspended. |
10556 |
+- */ |
10557 |
+-int cc_suspend_req_queue(struct cc_drvdata *drvdata) |
10558 |
+-{ |
10559 |
+- struct cc_req_mgr_handle *request_mgr_handle = |
10560 |
+- drvdata->request_mgr_handle; |
10561 |
+- |
10562 |
+- /* lock the send_request */ |
10563 |
+- spin_lock_bh(&request_mgr_handle->hw_lock); |
10564 |
+- if (request_mgr_handle->req_queue_head != |
10565 |
+- request_mgr_handle->req_queue_tail) { |
10566 |
+- spin_unlock_bh(&request_mgr_handle->hw_lock); |
10567 |
+- return -EBUSY; |
10568 |
+- } |
10569 |
+- request_mgr_handle->is_runtime_suspended = true; |
10570 |
+- spin_unlock_bh(&request_mgr_handle->hw_lock); |
10571 |
+- |
10572 |
+- return 0; |
10573 |
+-} |
10574 |
+- |
10575 |
+-bool cc_req_queue_suspended(struct cc_drvdata *drvdata) |
10576 |
+-{ |
10577 |
+- struct cc_req_mgr_handle *request_mgr_handle = |
10578 |
+- drvdata->request_mgr_handle; |
10579 |
+- |
10580 |
+- return request_mgr_handle->is_runtime_suspended; |
10581 |
+-} |
10582 |
+- |
10583 |
+-#endif |
10584 |
+diff --git a/drivers/crypto/ccree/cc_request_mgr.h b/drivers/crypto/ccree/cc_request_mgr.h |
10585 |
+index 573cb97af085..ae96abce25c9 100644 |
10586 |
+--- a/drivers/crypto/ccree/cc_request_mgr.h |
10587 |
++++ b/drivers/crypto/ccree/cc_request_mgr.h |
10588 |
+@@ -40,12 +40,4 @@ void complete_request(struct cc_drvdata *drvdata); |
10589 |
+ |
10590 |
+ void cc_req_mgr_fini(struct cc_drvdata *drvdata); |
10591 |
+ |
10592 |
+-#if defined(CONFIG_PM) |
10593 |
+-int cc_resume_req_queue(struct cc_drvdata *drvdata); |
10594 |
+- |
10595 |
+-int cc_suspend_req_queue(struct cc_drvdata *drvdata); |
10596 |
+- |
10597 |
+-bool cc_req_queue_suspended(struct cc_drvdata *drvdata); |
10598 |
+-#endif |
10599 |
+- |
10600 |
+ #endif /*__REQUEST_MGR_H__*/ |
10601 |
+diff --git a/drivers/crypto/geode-aes.c b/drivers/crypto/geode-aes.c |
10602 |
+index d670f7000cbb..0bd99c0decf5 100644 |
10603 |
+--- a/drivers/crypto/geode-aes.c |
10604 |
++++ b/drivers/crypto/geode-aes.c |
10605 |
+@@ -14,7 +14,7 @@ |
10606 |
+ #include <linux/spinlock.h> |
10607 |
+ #include <crypto/algapi.h> |
10608 |
+ #include <crypto/aes.h> |
10609 |
+-#include <crypto/skcipher.h> |
10610 |
++#include <crypto/internal/skcipher.h> |
10611 |
+ |
10612 |
+ #include <linux/io.h> |
10613 |
+ #include <linux/delay.h> |
10614 |
+@@ -28,12 +28,12 @@ static spinlock_t lock; |
10615 |
+ |
10616 |
+ /* Write a 128 bit field (either a writable key or IV) */ |
10617 |
+ static inline void |
10618 |
+-_writefield(u32 offset, void *value) |
10619 |
++_writefield(u32 offset, const void *value) |
10620 |
+ { |
10621 |
+ int i; |
10622 |
+ |
10623 |
+ for (i = 0; i < 4; i++) |
10624 |
+- iowrite32(((u32 *) value)[i], _iobase + offset + (i * 4)); |
10625 |
++ iowrite32(((const u32 *) value)[i], _iobase + offset + (i * 4)); |
10626 |
+ } |
10627 |
+ |
10628 |
+ /* Read a 128 bit field (either a writable key or IV) */ |
10629 |
+@@ -47,12 +47,12 @@ _readfield(u32 offset, void *value) |
10630 |
+ } |
10631 |
+ |
10632 |
+ static int |
10633 |
+-do_crypt(void *src, void *dst, int len, u32 flags) |
10634 |
++do_crypt(const void *src, void *dst, u32 len, u32 flags) |
10635 |
+ { |
10636 |
+ u32 status; |
10637 |
+ u32 counter = AES_OP_TIMEOUT; |
10638 |
+ |
10639 |
+- iowrite32(virt_to_phys(src), _iobase + AES_SOURCEA_REG); |
10640 |
++ iowrite32(virt_to_phys((void *)src), _iobase + AES_SOURCEA_REG); |
10641 |
+ iowrite32(virt_to_phys(dst), _iobase + AES_DSTA_REG); |
10642 |
+ iowrite32(len, _iobase + AES_LENA_REG); |
10643 |
+ |
10644 |
+@@ -69,16 +69,14 @@ do_crypt(void *src, void *dst, int len, u32 flags) |
10645 |
+ return counter ? 0 : 1; |
10646 |
+ } |
10647 |
+ |
10648 |
+-static unsigned int |
10649 |
+-geode_aes_crypt(struct geode_aes_op *op) |
10650 |
++static void |
10651 |
++geode_aes_crypt(const struct geode_aes_tfm_ctx *tctx, const void *src, |
10652 |
++ void *dst, u32 len, u8 *iv, int mode, int dir) |
10653 |
+ { |
10654 |
+ u32 flags = 0; |
10655 |
+ unsigned long iflags; |
10656 |
+ int ret; |
10657 |
+ |
10658 |
+- if (op->len == 0) |
10659 |
+- return 0; |
10660 |
+- |
10661 |
+ /* If the source and destination is the same, then |
10662 |
+ * we need to turn on the coherent flags, otherwise |
10663 |
+ * we don't need to worry |
10664 |
+@@ -86,32 +84,28 @@ geode_aes_crypt(struct geode_aes_op *op) |
10665 |
+ |
10666 |
+ flags |= (AES_CTRL_DCA | AES_CTRL_SCA); |
10667 |
+ |
10668 |
+- if (op->dir == AES_DIR_ENCRYPT) |
10669 |
++ if (dir == AES_DIR_ENCRYPT) |
10670 |
+ flags |= AES_CTRL_ENCRYPT; |
10671 |
+ |
10672 |
+ /* Start the critical section */ |
10673 |
+ |
10674 |
+ spin_lock_irqsave(&lock, iflags); |
10675 |
+ |
10676 |
+- if (op->mode == AES_MODE_CBC) { |
10677 |
++ if (mode == AES_MODE_CBC) { |
10678 |
+ flags |= AES_CTRL_CBC; |
10679 |
+- _writefield(AES_WRITEIV0_REG, op->iv); |
10680 |
++ _writefield(AES_WRITEIV0_REG, iv); |
10681 |
+ } |
10682 |
+ |
10683 |
+- if (!(op->flags & AES_FLAGS_HIDDENKEY)) { |
10684 |
+- flags |= AES_CTRL_WRKEY; |
10685 |
+- _writefield(AES_WRITEKEY0_REG, op->key); |
10686 |
+- } |
10687 |
++ flags |= AES_CTRL_WRKEY; |
10688 |
++ _writefield(AES_WRITEKEY0_REG, tctx->key); |
10689 |
+ |
10690 |
+- ret = do_crypt(op->src, op->dst, op->len, flags); |
10691 |
++ ret = do_crypt(src, dst, len, flags); |
10692 |
+ BUG_ON(ret); |
10693 |
+ |
10694 |
+- if (op->mode == AES_MODE_CBC) |
10695 |
+- _readfield(AES_WRITEIV0_REG, op->iv); |
10696 |
++ if (mode == AES_MODE_CBC) |
10697 |
++ _readfield(AES_WRITEIV0_REG, iv); |
10698 |
+ |
10699 |
+ spin_unlock_irqrestore(&lock, iflags); |
10700 |
+- |
10701 |
+- return op->len; |
10702 |
+ } |
10703 |
+ |
10704 |
+ /* CRYPTO-API Functions */ |
10705 |
+@@ -119,13 +113,13 @@ geode_aes_crypt(struct geode_aes_op *op) |
10706 |
+ static int geode_setkey_cip(struct crypto_tfm *tfm, const u8 *key, |
10707 |
+ unsigned int len) |
10708 |
+ { |
10709 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10710 |
++ struct geode_aes_tfm_ctx *tctx = crypto_tfm_ctx(tfm); |
10711 |
+ unsigned int ret; |
10712 |
+ |
10713 |
+- op->keylen = len; |
10714 |
++ tctx->keylen = len; |
10715 |
+ |
10716 |
+ if (len == AES_KEYSIZE_128) { |
10717 |
+- memcpy(op->key, key, len); |
10718 |
++ memcpy(tctx->key, key, len); |
10719 |
+ return 0; |
10720 |
+ } |
10721 |
+ |
10722 |
+@@ -138,132 +132,93 @@ static int geode_setkey_cip(struct crypto_tfm *tfm, const u8 *key, |
10723 |
+ /* |
10724 |
+ * The requested key size is not supported by HW, do a fallback |
10725 |
+ */ |
10726 |
+- op->fallback.cip->base.crt_flags &= ~CRYPTO_TFM_REQ_MASK; |
10727 |
+- op->fallback.cip->base.crt_flags |= (tfm->crt_flags & CRYPTO_TFM_REQ_MASK); |
10728 |
++ tctx->fallback.cip->base.crt_flags &= ~CRYPTO_TFM_REQ_MASK; |
10729 |
++ tctx->fallback.cip->base.crt_flags |= |
10730 |
++ (tfm->crt_flags & CRYPTO_TFM_REQ_MASK); |
10731 |
+ |
10732 |
+- ret = crypto_cipher_setkey(op->fallback.cip, key, len); |
10733 |
++ ret = crypto_cipher_setkey(tctx->fallback.cip, key, len); |
10734 |
+ if (ret) { |
10735 |
+ tfm->crt_flags &= ~CRYPTO_TFM_RES_MASK; |
10736 |
+- tfm->crt_flags |= (op->fallback.cip->base.crt_flags & CRYPTO_TFM_RES_MASK); |
10737 |
++ tfm->crt_flags |= (tctx->fallback.cip->base.crt_flags & |
10738 |
++ CRYPTO_TFM_RES_MASK); |
10739 |
+ } |
10740 |
+ return ret; |
10741 |
+ } |
10742 |
+ |
10743 |
+-static int geode_setkey_blk(struct crypto_tfm *tfm, const u8 *key, |
10744 |
+- unsigned int len) |
10745 |
++static int geode_setkey_skcipher(struct crypto_skcipher *tfm, const u8 *key, |
10746 |
++ unsigned int len) |
10747 |
+ { |
10748 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10749 |
++ struct geode_aes_tfm_ctx *tctx = crypto_skcipher_ctx(tfm); |
10750 |
+ unsigned int ret; |
10751 |
+ |
10752 |
+- op->keylen = len; |
10753 |
++ tctx->keylen = len; |
10754 |
+ |
10755 |
+ if (len == AES_KEYSIZE_128) { |
10756 |
+- memcpy(op->key, key, len); |
10757 |
++ memcpy(tctx->key, key, len); |
10758 |
+ return 0; |
10759 |
+ } |
10760 |
+ |
10761 |
+ if (len != AES_KEYSIZE_192 && len != AES_KEYSIZE_256) { |
10762 |
+ /* not supported at all */ |
10763 |
+- tfm->crt_flags |= CRYPTO_TFM_RES_BAD_KEY_LEN; |
10764 |
++ crypto_skcipher_set_flags(tfm, CRYPTO_TFM_RES_BAD_KEY_LEN); |
10765 |
+ return -EINVAL; |
10766 |
+ } |
10767 |
+ |
10768 |
+ /* |
10769 |
+ * The requested key size is not supported by HW, do a fallback |
10770 |
+ */ |
10771 |
+- crypto_skcipher_clear_flags(op->fallback.blk, CRYPTO_TFM_REQ_MASK); |
10772 |
+- crypto_skcipher_set_flags(op->fallback.blk, |
10773 |
+- tfm->crt_flags & CRYPTO_TFM_REQ_MASK); |
10774 |
+- |
10775 |
+- ret = crypto_skcipher_setkey(op->fallback.blk, key, len); |
10776 |
+- if (ret) { |
10777 |
+- tfm->crt_flags &= ~CRYPTO_TFM_RES_MASK; |
10778 |
+- tfm->crt_flags |= crypto_skcipher_get_flags(op->fallback.blk) & |
10779 |
+- CRYPTO_TFM_RES_MASK; |
10780 |
+- } |
10781 |
++ crypto_skcipher_clear_flags(tctx->fallback.skcipher, |
10782 |
++ CRYPTO_TFM_REQ_MASK); |
10783 |
++ crypto_skcipher_set_flags(tctx->fallback.skcipher, |
10784 |
++ crypto_skcipher_get_flags(tfm) & |
10785 |
++ CRYPTO_TFM_REQ_MASK); |
10786 |
++ ret = crypto_skcipher_setkey(tctx->fallback.skcipher, key, len); |
10787 |
++ crypto_skcipher_set_flags(tfm, |
10788 |
++ crypto_skcipher_get_flags(tctx->fallback.skcipher) & |
10789 |
++ CRYPTO_TFM_RES_MASK); |
10790 |
+ return ret; |
10791 |
+ } |
10792 |
+ |
10793 |
+-static int fallback_blk_dec(struct blkcipher_desc *desc, |
10794 |
+- struct scatterlist *dst, struct scatterlist *src, |
10795 |
+- unsigned int nbytes) |
10796 |
+-{ |
10797 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
10798 |
+- SKCIPHER_REQUEST_ON_STACK(req, op->fallback.blk); |
10799 |
+- |
10800 |
+- skcipher_request_set_tfm(req, op->fallback.blk); |
10801 |
+- skcipher_request_set_callback(req, 0, NULL, NULL); |
10802 |
+- skcipher_request_set_crypt(req, src, dst, nbytes, desc->info); |
10803 |
+- |
10804 |
+- return crypto_skcipher_decrypt(req); |
10805 |
+-} |
10806 |
+- |
10807 |
+-static int fallback_blk_enc(struct blkcipher_desc *desc, |
10808 |
+- struct scatterlist *dst, struct scatterlist *src, |
10809 |
+- unsigned int nbytes) |
10810 |
+-{ |
10811 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
10812 |
+- SKCIPHER_REQUEST_ON_STACK(req, op->fallback.blk); |
10813 |
+- |
10814 |
+- skcipher_request_set_tfm(req, op->fallback.blk); |
10815 |
+- skcipher_request_set_callback(req, 0, NULL, NULL); |
10816 |
+- skcipher_request_set_crypt(req, src, dst, nbytes, desc->info); |
10817 |
+- |
10818 |
+- return crypto_skcipher_encrypt(req); |
10819 |
+-} |
10820 |
+- |
10821 |
+ static void |
10822 |
+ geode_encrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) |
10823 |
+ { |
10824 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10825 |
++ const struct geode_aes_tfm_ctx *tctx = crypto_tfm_ctx(tfm); |
10826 |
+ |
10827 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) { |
10828 |
+- crypto_cipher_encrypt_one(op->fallback.cip, out, in); |
10829 |
++ if (unlikely(tctx->keylen != AES_KEYSIZE_128)) { |
10830 |
++ crypto_cipher_encrypt_one(tctx->fallback.cip, out, in); |
10831 |
+ return; |
10832 |
+ } |
10833 |
+ |
10834 |
+- op->src = (void *) in; |
10835 |
+- op->dst = (void *) out; |
10836 |
+- op->mode = AES_MODE_ECB; |
10837 |
+- op->flags = 0; |
10838 |
+- op->len = AES_BLOCK_SIZE; |
10839 |
+- op->dir = AES_DIR_ENCRYPT; |
10840 |
+- |
10841 |
+- geode_aes_crypt(op); |
10842 |
++ geode_aes_crypt(tctx, in, out, AES_BLOCK_SIZE, NULL, |
10843 |
++ AES_MODE_ECB, AES_DIR_ENCRYPT); |
10844 |
+ } |
10845 |
+ |
10846 |
+ |
10847 |
+ static void |
10848 |
+ geode_decrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) |
10849 |
+ { |
10850 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10851 |
++ const struct geode_aes_tfm_ctx *tctx = crypto_tfm_ctx(tfm); |
10852 |
+ |
10853 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) { |
10854 |
+- crypto_cipher_decrypt_one(op->fallback.cip, out, in); |
10855 |
++ if (unlikely(tctx->keylen != AES_KEYSIZE_128)) { |
10856 |
++ crypto_cipher_decrypt_one(tctx->fallback.cip, out, in); |
10857 |
+ return; |
10858 |
+ } |
10859 |
+ |
10860 |
+- op->src = (void *) in; |
10861 |
+- op->dst = (void *) out; |
10862 |
+- op->mode = AES_MODE_ECB; |
10863 |
+- op->flags = 0; |
10864 |
+- op->len = AES_BLOCK_SIZE; |
10865 |
+- op->dir = AES_DIR_DECRYPT; |
10866 |
+- |
10867 |
+- geode_aes_crypt(op); |
10868 |
++ geode_aes_crypt(tctx, in, out, AES_BLOCK_SIZE, NULL, |
10869 |
++ AES_MODE_ECB, AES_DIR_DECRYPT); |
10870 |
+ } |
10871 |
+ |
10872 |
+ static int fallback_init_cip(struct crypto_tfm *tfm) |
10873 |
+ { |
10874 |
+ const char *name = crypto_tfm_alg_name(tfm); |
10875 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10876 |
++ struct geode_aes_tfm_ctx *tctx = crypto_tfm_ctx(tfm); |
10877 |
+ |
10878 |
+- op->fallback.cip = crypto_alloc_cipher(name, 0, |
10879 |
+- CRYPTO_ALG_ASYNC | CRYPTO_ALG_NEED_FALLBACK); |
10880 |
++ tctx->fallback.cip = crypto_alloc_cipher(name, 0, |
10881 |
++ CRYPTO_ALG_NEED_FALLBACK); |
10882 |
+ |
10883 |
+- if (IS_ERR(op->fallback.cip)) { |
10884 |
++ if (IS_ERR(tctx->fallback.cip)) { |
10885 |
+ printk(KERN_ERR "Error allocating fallback algo %s\n", name); |
10886 |
+- return PTR_ERR(op->fallback.cip); |
10887 |
++ return PTR_ERR(tctx->fallback.cip); |
10888 |
+ } |
10889 |
+ |
10890 |
+ return 0; |
10891 |
+@@ -271,10 +226,9 @@ static int fallback_init_cip(struct crypto_tfm *tfm) |
10892 |
+ |
10893 |
+ static void fallback_exit_cip(struct crypto_tfm *tfm) |
10894 |
+ { |
10895 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
10896 |
++ struct geode_aes_tfm_ctx *tctx = crypto_tfm_ctx(tfm); |
10897 |
+ |
10898 |
+- crypto_free_cipher(op->fallback.cip); |
10899 |
+- op->fallback.cip = NULL; |
10900 |
++ crypto_free_cipher(tctx->fallback.cip); |
10901 |
+ } |
10902 |
+ |
10903 |
+ static struct crypto_alg geode_alg = { |
10904 |
+@@ -287,7 +241,7 @@ static struct crypto_alg geode_alg = { |
10905 |
+ .cra_init = fallback_init_cip, |
10906 |
+ .cra_exit = fallback_exit_cip, |
10907 |
+ .cra_blocksize = AES_BLOCK_SIZE, |
10908 |
+- .cra_ctxsize = sizeof(struct geode_aes_op), |
10909 |
++ .cra_ctxsize = sizeof(struct geode_aes_tfm_ctx), |
10910 |
+ .cra_module = THIS_MODULE, |
10911 |
+ .cra_u = { |
10912 |
+ .cipher = { |
10913 |
+@@ -300,222 +254,126 @@ static struct crypto_alg geode_alg = { |
10914 |
+ } |
10915 |
+ }; |
10916 |
+ |
10917 |
+-static int |
10918 |
+-geode_cbc_decrypt(struct blkcipher_desc *desc, |
10919 |
+- struct scatterlist *dst, struct scatterlist *src, |
10920 |
+- unsigned int nbytes) |
10921 |
++static int geode_init_skcipher(struct crypto_skcipher *tfm) |
10922 |
+ { |
10923 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
10924 |
+- struct blkcipher_walk walk; |
10925 |
+- int err, ret; |
10926 |
+- |
10927 |
+- if (nbytes % AES_BLOCK_SIZE) |
10928 |
+- return -EINVAL; |
10929 |
+- |
10930 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) |
10931 |
+- return fallback_blk_dec(desc, dst, src, nbytes); |
10932 |
++ const char *name = crypto_tfm_alg_name(&tfm->base); |
10933 |
++ struct geode_aes_tfm_ctx *tctx = crypto_skcipher_ctx(tfm); |
10934 |
+ |
10935 |
+- blkcipher_walk_init(&walk, dst, src, nbytes); |
10936 |
+- err = blkcipher_walk_virt(desc, &walk); |
10937 |
+- op->iv = walk.iv; |
10938 |
+- |
10939 |
+- while ((nbytes = walk.nbytes)) { |
10940 |
+- op->src = walk.src.virt.addr, |
10941 |
+- op->dst = walk.dst.virt.addr; |
10942 |
+- op->mode = AES_MODE_CBC; |
10943 |
+- op->len = nbytes - (nbytes % AES_BLOCK_SIZE); |
10944 |
+- op->dir = AES_DIR_DECRYPT; |
10945 |
+- |
10946 |
+- ret = geode_aes_crypt(op); |
10947 |
+- |
10948 |
+- nbytes -= ret; |
10949 |
+- err = blkcipher_walk_done(desc, &walk, nbytes); |
10950 |
++ tctx->fallback.skcipher = |
10951 |
++ crypto_alloc_skcipher(name, 0, CRYPTO_ALG_NEED_FALLBACK | |
10952 |
++ CRYPTO_ALG_ASYNC); |
10953 |
++ if (IS_ERR(tctx->fallback.skcipher)) { |
10954 |
++ printk(KERN_ERR "Error allocating fallback algo %s\n", name); |
10955 |
++ return PTR_ERR(tctx->fallback.skcipher); |
10956 |
+ } |
10957 |
+ |
10958 |
+- return err; |
10959 |
++ crypto_skcipher_set_reqsize(tfm, sizeof(struct skcipher_request) + |
10960 |
++ crypto_skcipher_reqsize(tctx->fallback.skcipher)); |
10961 |
++ return 0; |
10962 |
+ } |
10963 |
+ |
10964 |
+-static int |
10965 |
+-geode_cbc_encrypt(struct blkcipher_desc *desc, |
10966 |
+- struct scatterlist *dst, struct scatterlist *src, |
10967 |
+- unsigned int nbytes) |
10968 |
++static void geode_exit_skcipher(struct crypto_skcipher *tfm) |
10969 |
+ { |
10970 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
10971 |
+- struct blkcipher_walk walk; |
10972 |
+- int err, ret; |
10973 |
++ struct geode_aes_tfm_ctx *tctx = crypto_skcipher_ctx(tfm); |
10974 |
+ |
10975 |
+- if (nbytes % AES_BLOCK_SIZE) |
10976 |
+- return -EINVAL; |
10977 |
+- |
10978 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) |
10979 |
+- return fallback_blk_enc(desc, dst, src, nbytes); |
10980 |
++ crypto_free_skcipher(tctx->fallback.skcipher); |
10981 |
++} |
10982 |
+ |
10983 |
+- blkcipher_walk_init(&walk, dst, src, nbytes); |
10984 |
+- err = blkcipher_walk_virt(desc, &walk); |
10985 |
+- op->iv = walk.iv; |
10986 |
++static int geode_skcipher_crypt(struct skcipher_request *req, int mode, int dir) |
10987 |
++{ |
10988 |
++ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); |
10989 |
++ const struct geode_aes_tfm_ctx *tctx = crypto_skcipher_ctx(tfm); |
10990 |
++ struct skcipher_walk walk; |
10991 |
++ unsigned int nbytes; |
10992 |
++ int err; |
10993 |
++ |
10994 |
++ if (unlikely(tctx->keylen != AES_KEYSIZE_128)) { |
10995 |
++ struct skcipher_request *subreq = skcipher_request_ctx(req); |
10996 |
++ |
10997 |
++ *subreq = *req; |
10998 |
++ skcipher_request_set_tfm(subreq, tctx->fallback.skcipher); |
10999 |
++ if (dir == AES_DIR_DECRYPT) |
11000 |
++ return crypto_skcipher_decrypt(subreq); |
11001 |
++ else |
11002 |
++ return crypto_skcipher_encrypt(subreq); |
11003 |
++ } |
11004 |
+ |
11005 |
+- while ((nbytes = walk.nbytes)) { |
11006 |
+- op->src = walk.src.virt.addr, |
11007 |
+- op->dst = walk.dst.virt.addr; |
11008 |
+- op->mode = AES_MODE_CBC; |
11009 |
+- op->len = nbytes - (nbytes % AES_BLOCK_SIZE); |
11010 |
+- op->dir = AES_DIR_ENCRYPT; |
11011 |
++ err = skcipher_walk_virt(&walk, req, false); |
11012 |
+ |
11013 |
+- ret = geode_aes_crypt(op); |
11014 |
+- nbytes -= ret; |
11015 |
+- err = blkcipher_walk_done(desc, &walk, nbytes); |
11016 |
++ while ((nbytes = walk.nbytes) != 0) { |
11017 |
++ geode_aes_crypt(tctx, walk.src.virt.addr, walk.dst.virt.addr, |
11018 |
++ round_down(nbytes, AES_BLOCK_SIZE), |
11019 |
++ walk.iv, mode, dir); |
11020 |
++ err = skcipher_walk_done(&walk, nbytes % AES_BLOCK_SIZE); |
11021 |
+ } |
11022 |
+ |
11023 |
+ return err; |
11024 |
+ } |
11025 |
+ |
11026 |
+-static int fallback_init_blk(struct crypto_tfm *tfm) |
11027 |
++static int geode_cbc_encrypt(struct skcipher_request *req) |
11028 |
+ { |
11029 |
+- const char *name = crypto_tfm_alg_name(tfm); |
11030 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
11031 |
+- |
11032 |
+- op->fallback.blk = crypto_alloc_skcipher(name, 0, |
11033 |
+- CRYPTO_ALG_ASYNC | |
11034 |
+- CRYPTO_ALG_NEED_FALLBACK); |
11035 |
+- |
11036 |
+- if (IS_ERR(op->fallback.blk)) { |
11037 |
+- printk(KERN_ERR "Error allocating fallback algo %s\n", name); |
11038 |
+- return PTR_ERR(op->fallback.blk); |
11039 |
+- } |
11040 |
+- |
11041 |
+- return 0; |
11042 |
++ return geode_skcipher_crypt(req, AES_MODE_CBC, AES_DIR_ENCRYPT); |
11043 |
+ } |
11044 |
+ |
11045 |
+-static void fallback_exit_blk(struct crypto_tfm *tfm) |
11046 |
++static int geode_cbc_decrypt(struct skcipher_request *req) |
11047 |
+ { |
11048 |
+- struct geode_aes_op *op = crypto_tfm_ctx(tfm); |
11049 |
+- |
11050 |
+- crypto_free_skcipher(op->fallback.blk); |
11051 |
+- op->fallback.blk = NULL; |
11052 |
++ return geode_skcipher_crypt(req, AES_MODE_CBC, AES_DIR_DECRYPT); |
11053 |
+ } |
11054 |
+ |
11055 |
+-static struct crypto_alg geode_cbc_alg = { |
11056 |
+- .cra_name = "cbc(aes)", |
11057 |
+- .cra_driver_name = "cbc-aes-geode", |
11058 |
+- .cra_priority = 400, |
11059 |
+- .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER | |
11060 |
+- CRYPTO_ALG_KERN_DRIVER_ONLY | |
11061 |
+- CRYPTO_ALG_NEED_FALLBACK, |
11062 |
+- .cra_init = fallback_init_blk, |
11063 |
+- .cra_exit = fallback_exit_blk, |
11064 |
+- .cra_blocksize = AES_BLOCK_SIZE, |
11065 |
+- .cra_ctxsize = sizeof(struct geode_aes_op), |
11066 |
+- .cra_alignmask = 15, |
11067 |
+- .cra_type = &crypto_blkcipher_type, |
11068 |
+- .cra_module = THIS_MODULE, |
11069 |
+- .cra_u = { |
11070 |
+- .blkcipher = { |
11071 |
+- .min_keysize = AES_MIN_KEY_SIZE, |
11072 |
+- .max_keysize = AES_MAX_KEY_SIZE, |
11073 |
+- .setkey = geode_setkey_blk, |
11074 |
+- .encrypt = geode_cbc_encrypt, |
11075 |
+- .decrypt = geode_cbc_decrypt, |
11076 |
+- .ivsize = AES_BLOCK_SIZE, |
11077 |
+- } |
11078 |
+- } |
11079 |
+-}; |
11080 |
+- |
11081 |
+-static int |
11082 |
+-geode_ecb_decrypt(struct blkcipher_desc *desc, |
11083 |
+- struct scatterlist *dst, struct scatterlist *src, |
11084 |
+- unsigned int nbytes) |
11085 |
++static int geode_ecb_encrypt(struct skcipher_request *req) |
11086 |
+ { |
11087 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
11088 |
+- struct blkcipher_walk walk; |
11089 |
+- int err, ret; |
11090 |
+- |
11091 |
+- if (nbytes % AES_BLOCK_SIZE) |
11092 |
+- return -EINVAL; |
11093 |
+- |
11094 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) |
11095 |
+- return fallback_blk_dec(desc, dst, src, nbytes); |
11096 |
+- |
11097 |
+- blkcipher_walk_init(&walk, dst, src, nbytes); |
11098 |
+- err = blkcipher_walk_virt(desc, &walk); |
11099 |
+- |
11100 |
+- while ((nbytes = walk.nbytes)) { |
11101 |
+- op->src = walk.src.virt.addr, |
11102 |
+- op->dst = walk.dst.virt.addr; |
11103 |
+- op->mode = AES_MODE_ECB; |
11104 |
+- op->len = nbytes - (nbytes % AES_BLOCK_SIZE); |
11105 |
+- op->dir = AES_DIR_DECRYPT; |
11106 |
+- |
11107 |
+- ret = geode_aes_crypt(op); |
11108 |
+- nbytes -= ret; |
11109 |
+- err = blkcipher_walk_done(desc, &walk, nbytes); |
11110 |
+- } |
11111 |
+- |
11112 |
+- return err; |
11113 |
++ return geode_skcipher_crypt(req, AES_MODE_ECB, AES_DIR_ENCRYPT); |
11114 |
+ } |
11115 |
+ |
11116 |
+-static int |
11117 |
+-geode_ecb_encrypt(struct blkcipher_desc *desc, |
11118 |
+- struct scatterlist *dst, struct scatterlist *src, |
11119 |
+- unsigned int nbytes) |
11120 |
++static int geode_ecb_decrypt(struct skcipher_request *req) |
11121 |
+ { |
11122 |
+- struct geode_aes_op *op = crypto_blkcipher_ctx(desc->tfm); |
11123 |
+- struct blkcipher_walk walk; |
11124 |
+- int err, ret; |
11125 |
+- |
11126 |
+- if (nbytes % AES_BLOCK_SIZE) |
11127 |
+- return -EINVAL; |
11128 |
+- |
11129 |
+- if (unlikely(op->keylen != AES_KEYSIZE_128)) |
11130 |
+- return fallback_blk_enc(desc, dst, src, nbytes); |
11131 |
+- |
11132 |
+- blkcipher_walk_init(&walk, dst, src, nbytes); |
11133 |
+- err = blkcipher_walk_virt(desc, &walk); |
11134 |
+- |
11135 |
+- while ((nbytes = walk.nbytes)) { |
11136 |
+- op->src = walk.src.virt.addr, |
11137 |
+- op->dst = walk.dst.virt.addr; |
11138 |
+- op->mode = AES_MODE_ECB; |
11139 |
+- op->len = nbytes - (nbytes % AES_BLOCK_SIZE); |
11140 |
+- op->dir = AES_DIR_ENCRYPT; |
11141 |
+- |
11142 |
+- ret = geode_aes_crypt(op); |
11143 |
+- nbytes -= ret; |
11144 |
+- ret = blkcipher_walk_done(desc, &walk, nbytes); |
11145 |
+- } |
11146 |
+- |
11147 |
+- return err; |
11148 |
++ return geode_skcipher_crypt(req, AES_MODE_ECB, AES_DIR_DECRYPT); |
11149 |
+ } |
11150 |
+ |
11151 |
+-static struct crypto_alg geode_ecb_alg = { |
11152 |
+- .cra_name = "ecb(aes)", |
11153 |
+- .cra_driver_name = "ecb-aes-geode", |
11154 |
+- .cra_priority = 400, |
11155 |
+- .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER | |
11156 |
+- CRYPTO_ALG_KERN_DRIVER_ONLY | |
11157 |
+- CRYPTO_ALG_NEED_FALLBACK, |
11158 |
+- .cra_init = fallback_init_blk, |
11159 |
+- .cra_exit = fallback_exit_blk, |
11160 |
+- .cra_blocksize = AES_BLOCK_SIZE, |
11161 |
+- .cra_ctxsize = sizeof(struct geode_aes_op), |
11162 |
+- .cra_alignmask = 15, |
11163 |
+- .cra_type = &crypto_blkcipher_type, |
11164 |
+- .cra_module = THIS_MODULE, |
11165 |
+- .cra_u = { |
11166 |
+- .blkcipher = { |
11167 |
+- .min_keysize = AES_MIN_KEY_SIZE, |
11168 |
+- .max_keysize = AES_MAX_KEY_SIZE, |
11169 |
+- .setkey = geode_setkey_blk, |
11170 |
+- .encrypt = geode_ecb_encrypt, |
11171 |
+- .decrypt = geode_ecb_decrypt, |
11172 |
+- } |
11173 |
+- } |
11174 |
++static struct skcipher_alg geode_skcipher_algs[] = { |
11175 |
++ { |
11176 |
++ .base.cra_name = "cbc(aes)", |
11177 |
++ .base.cra_driver_name = "cbc-aes-geode", |
11178 |
++ .base.cra_priority = 400, |
11179 |
++ .base.cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY | |
11180 |
++ CRYPTO_ALG_NEED_FALLBACK, |
11181 |
++ .base.cra_blocksize = AES_BLOCK_SIZE, |
11182 |
++ .base.cra_ctxsize = sizeof(struct geode_aes_tfm_ctx), |
11183 |
++ .base.cra_alignmask = 15, |
11184 |
++ .base.cra_module = THIS_MODULE, |
11185 |
++ .init = geode_init_skcipher, |
11186 |
++ .exit = geode_exit_skcipher, |
11187 |
++ .setkey = geode_setkey_skcipher, |
11188 |
++ .encrypt = geode_cbc_encrypt, |
11189 |
++ .decrypt = geode_cbc_decrypt, |
11190 |
++ .min_keysize = AES_MIN_KEY_SIZE, |
11191 |
++ .max_keysize = AES_MAX_KEY_SIZE, |
11192 |
++ .ivsize = AES_BLOCK_SIZE, |
11193 |
++ }, { |
11194 |
++ .base.cra_name = "ecb(aes)", |
11195 |
++ .base.cra_driver_name = "ecb-aes-geode", |
11196 |
++ .base.cra_priority = 400, |
11197 |
++ .base.cra_flags = CRYPTO_ALG_KERN_DRIVER_ONLY | |
11198 |
++ CRYPTO_ALG_NEED_FALLBACK, |
11199 |
++ .base.cra_blocksize = AES_BLOCK_SIZE, |
11200 |
++ .base.cra_ctxsize = sizeof(struct geode_aes_tfm_ctx), |
11201 |
++ .base.cra_alignmask = 15, |
11202 |
++ .base.cra_module = THIS_MODULE, |
11203 |
++ .init = geode_init_skcipher, |
11204 |
++ .exit = geode_exit_skcipher, |
11205 |
++ .setkey = geode_setkey_skcipher, |
11206 |
++ .encrypt = geode_ecb_encrypt, |
11207 |
++ .decrypt = geode_ecb_decrypt, |
11208 |
++ .min_keysize = AES_MIN_KEY_SIZE, |
11209 |
++ .max_keysize = AES_MAX_KEY_SIZE, |
11210 |
++ }, |
11211 |
+ }; |
11212 |
+ |
11213 |
+ static void geode_aes_remove(struct pci_dev *dev) |
11214 |
+ { |
11215 |
+ crypto_unregister_alg(&geode_alg); |
11216 |
+- crypto_unregister_alg(&geode_ecb_alg); |
11217 |
+- crypto_unregister_alg(&geode_cbc_alg); |
11218 |
++ crypto_unregister_skciphers(geode_skcipher_algs, |
11219 |
++ ARRAY_SIZE(geode_skcipher_algs)); |
11220 |
+ |
11221 |
+ pci_iounmap(dev, _iobase); |
11222 |
+ _iobase = NULL; |
11223 |
+@@ -553,20 +411,14 @@ static int geode_aes_probe(struct pci_dev *dev, const struct pci_device_id *id) |
11224 |
+ if (ret) |
11225 |
+ goto eiomap; |
11226 |
+ |
11227 |
+- ret = crypto_register_alg(&geode_ecb_alg); |
11228 |
++ ret = crypto_register_skciphers(geode_skcipher_algs, |
11229 |
++ ARRAY_SIZE(geode_skcipher_algs)); |
11230 |
+ if (ret) |
11231 |
+ goto ealg; |
11232 |
+ |
11233 |
+- ret = crypto_register_alg(&geode_cbc_alg); |
11234 |
+- if (ret) |
11235 |
+- goto eecb; |
11236 |
+- |
11237 |
+ dev_notice(&dev->dev, "GEODE AES engine enabled.\n"); |
11238 |
+ return 0; |
11239 |
+ |
11240 |
+- eecb: |
11241 |
+- crypto_unregister_alg(&geode_ecb_alg); |
11242 |
+- |
11243 |
+ ealg: |
11244 |
+ crypto_unregister_alg(&geode_alg); |
11245 |
+ |
11246 |
+diff --git a/drivers/crypto/geode-aes.h b/drivers/crypto/geode-aes.h |
11247 |
+index c5763a041bb8..157443dc6d8a 100644 |
11248 |
+--- a/drivers/crypto/geode-aes.h |
11249 |
++++ b/drivers/crypto/geode-aes.h |
11250 |
+@@ -50,21 +50,10 @@ |
11251 |
+ |
11252 |
+ #define AES_OP_TIMEOUT 0x50000 |
11253 |
+ |
11254 |
+-struct geode_aes_op { |
11255 |
+- |
11256 |
+- void *src; |
11257 |
+- void *dst; |
11258 |
+- |
11259 |
+- u32 mode; |
11260 |
+- u32 dir; |
11261 |
+- u32 flags; |
11262 |
+- int len; |
11263 |
+- |
11264 |
++struct geode_aes_tfm_ctx { |
11265 |
+ u8 key[AES_KEYSIZE_128]; |
11266 |
+- u8 *iv; |
11267 |
+- |
11268 |
+ union { |
11269 |
+- struct crypto_skcipher *blk; |
11270 |
++ struct crypto_skcipher *skcipher; |
11271 |
+ struct crypto_cipher *cip; |
11272 |
+ } fallback; |
11273 |
+ u32 keylen; |
11274 |
+diff --git a/drivers/crypto/picoxcell_crypto.c b/drivers/crypto/picoxcell_crypto.c |
11275 |
+index 321d5e2ac833..e2491754c468 100644 |
11276 |
+--- a/drivers/crypto/picoxcell_crypto.c |
11277 |
++++ b/drivers/crypto/picoxcell_crypto.c |
11278 |
+@@ -1616,6 +1616,11 @@ static const struct of_device_id spacc_of_id_table[] = { |
11279 |
+ MODULE_DEVICE_TABLE(of, spacc_of_id_table); |
11280 |
+ #endif /* CONFIG_OF */ |
11281 |
+ |
11282 |
++static void spacc_tasklet_kill(void *data) |
11283 |
++{ |
11284 |
++ tasklet_kill(data); |
11285 |
++} |
11286 |
++ |
11287 |
+ static int spacc_probe(struct platform_device *pdev) |
11288 |
+ { |
11289 |
+ int i, err, ret; |
11290 |
+@@ -1659,6 +1664,14 @@ static int spacc_probe(struct platform_device *pdev) |
11291 |
+ return -ENXIO; |
11292 |
+ } |
11293 |
+ |
11294 |
++ tasklet_init(&engine->complete, spacc_spacc_complete, |
11295 |
++ (unsigned long)engine); |
11296 |
++ |
11297 |
++ ret = devm_add_action(&pdev->dev, spacc_tasklet_kill, |
11298 |
++ &engine->complete); |
11299 |
++ if (ret) |
11300 |
++ return ret; |
11301 |
++ |
11302 |
+ if (devm_request_irq(&pdev->dev, irq->start, spacc_spacc_irq, 0, |
11303 |
+ engine->name, engine)) { |
11304 |
+ dev_err(engine->dev, "failed to request IRQ\n"); |
11305 |
+@@ -1716,8 +1729,6 @@ static int spacc_probe(struct platform_device *pdev) |
11306 |
+ INIT_LIST_HEAD(&engine->completed); |
11307 |
+ INIT_LIST_HEAD(&engine->in_progress); |
11308 |
+ engine->in_flight = 0; |
11309 |
+- tasklet_init(&engine->complete, spacc_spacc_complete, |
11310 |
+- (unsigned long)engine); |
11311 |
+ |
11312 |
+ platform_set_drvdata(pdev, engine); |
11313 |
+ |
11314 |
+diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c |
11315 |
+index 8403b6a9a77b..d8c53ddc23b4 100644 |
11316 |
+--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c |
11317 |
++++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c |
11318 |
+@@ -248,7 +248,8 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( |
11319 |
+ drm_dp_mst_reset_vcpi_slots(mst_mgr, mst_port); |
11320 |
+ } |
11321 |
+ |
11322 |
+- ret = drm_dp_update_payload_part1(mst_mgr); |
11323 |
++ /* It's OK for this to fail */ |
11324 |
++ drm_dp_update_payload_part1(mst_mgr); |
11325 |
+ |
11326 |
+ /* mst_mgr->->payloads are VC payload notify MST branch using DPCD or |
11327 |
+ * AUX message. The sequence is slot 1-63 allocated sequence for each |
11328 |
+@@ -257,9 +258,6 @@ bool dm_helpers_dp_mst_write_payload_allocation_table( |
11329 |
+ |
11330 |
+ get_payload_table(aconnector, proposed_table); |
11331 |
+ |
11332 |
+- if (ret) |
11333 |
+- return false; |
11334 |
+- |
11335 |
+ return true; |
11336 |
+ } |
11337 |
+ |
11338 |
+@@ -310,7 +308,6 @@ bool dm_helpers_dp_mst_send_payload_allocation( |
11339 |
+ struct amdgpu_dm_connector *aconnector; |
11340 |
+ struct drm_dp_mst_topology_mgr *mst_mgr; |
11341 |
+ struct drm_dp_mst_port *mst_port; |
11342 |
+- int ret; |
11343 |
+ |
11344 |
+ aconnector = stream->sink->priv; |
11345 |
+ |
11346 |
+@@ -324,10 +321,8 @@ bool dm_helpers_dp_mst_send_payload_allocation( |
11347 |
+ if (!mst_mgr->mst_state) |
11348 |
+ return false; |
11349 |
+ |
11350 |
+- ret = drm_dp_update_payload_part2(mst_mgr); |
11351 |
+- |
11352 |
+- if (ret) |
11353 |
+- return false; |
11354 |
++ /* It's OK for this to fail */ |
11355 |
++ drm_dp_update_payload_part2(mst_mgr); |
11356 |
+ |
11357 |
+ if (!enable) |
11358 |
+ drm_dp_mst_deallocate_vcpi(mst_mgr, mst_port); |
11359 |
+diff --git a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c |
11360 |
+index d73281095fac..976109c20d49 100644 |
11361 |
+--- a/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c |
11362 |
++++ b/drivers/gpu/drm/atmel-hlcdc/atmel_hlcdc_crtc.c |
11363 |
+@@ -79,7 +79,11 @@ static void atmel_hlcdc_crtc_mode_set_nofb(struct drm_crtc *c) |
11364 |
+ struct videomode vm; |
11365 |
+ unsigned long prate; |
11366 |
+ unsigned int cfg; |
11367 |
+- int div; |
11368 |
++ int div, ret; |
11369 |
++ |
11370 |
++ ret = clk_prepare_enable(crtc->dc->hlcdc->sys_clk); |
11371 |
++ if (ret) |
11372 |
++ return; |
11373 |
+ |
11374 |
+ vm.vfront_porch = adj->crtc_vsync_start - adj->crtc_vdisplay; |
11375 |
+ vm.vback_porch = adj->crtc_vtotal - adj->crtc_vsync_end; |
11376 |
+@@ -138,6 +142,8 @@ static void atmel_hlcdc_crtc_mode_set_nofb(struct drm_crtc *c) |
11377 |
+ ATMEL_HLCDC_VSPSU | ATMEL_HLCDC_VSPHO | |
11378 |
+ ATMEL_HLCDC_GUARDTIME_MASK | ATMEL_HLCDC_MODE_MASK, |
11379 |
+ cfg); |
11380 |
++ |
11381 |
++ clk_disable_unprepare(crtc->dc->hlcdc->sys_clk); |
11382 |
+ } |
11383 |
+ |
11384 |
+ static enum drm_mode_status |
11385 |
+diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c |
11386 |
+index 58fe3945494c..bf4eed5f6a7e 100644 |
11387 |
+--- a/drivers/gpu/drm/drm_dp_mst_topology.c |
11388 |
++++ b/drivers/gpu/drm/drm_dp_mst_topology.c |
11389 |
+@@ -2125,6 +2125,7 @@ static bool drm_dp_get_vc_payload_bw(int dp_link_bw, |
11390 |
+ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool mst_state) |
11391 |
+ { |
11392 |
+ int ret = 0; |
11393 |
++ int i = 0; |
11394 |
+ struct drm_dp_mst_branch *mstb = NULL; |
11395 |
+ |
11396 |
+ mutex_lock(&mgr->lock); |
11397 |
+@@ -2185,10 +2186,21 @@ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool ms |
11398 |
+ /* this can fail if the device is gone */ |
11399 |
+ drm_dp_dpcd_writeb(mgr->aux, DP_MSTM_CTRL, 0); |
11400 |
+ ret = 0; |
11401 |
++ mutex_lock(&mgr->payload_lock); |
11402 |
+ memset(mgr->payloads, 0, mgr->max_payloads * sizeof(struct drm_dp_payload)); |
11403 |
+ mgr->payload_mask = 0; |
11404 |
+ set_bit(0, &mgr->payload_mask); |
11405 |
++ for (i = 0; i < mgr->max_payloads; i++) { |
11406 |
++ struct drm_dp_vcpi *vcpi = mgr->proposed_vcpis[i]; |
11407 |
++ |
11408 |
++ if (vcpi) { |
11409 |
++ vcpi->vcpi = 0; |
11410 |
++ vcpi->num_slots = 0; |
11411 |
++ } |
11412 |
++ mgr->proposed_vcpis[i] = NULL; |
11413 |
++ } |
11414 |
+ mgr->vcpi_mask = 0; |
11415 |
++ mutex_unlock(&mgr->payload_lock); |
11416 |
+ } |
11417 |
+ |
11418 |
+ out_unlock: |
11419 |
+diff --git a/drivers/gpu/drm/drm_rect.c b/drivers/gpu/drm/drm_rect.c |
11420 |
+index 8c057829b804..0f5a0c64c4c4 100644 |
11421 |
+--- a/drivers/gpu/drm/drm_rect.c |
11422 |
++++ b/drivers/gpu/drm/drm_rect.c |
11423 |
+@@ -52,7 +52,12 @@ EXPORT_SYMBOL(drm_rect_intersect); |
11424 |
+ |
11425 |
+ static u32 clip_scaled(u32 src, u32 dst, u32 clip) |
11426 |
+ { |
11427 |
+- u64 tmp = mul_u32_u32(src, dst - clip); |
11428 |
++ u64 tmp; |
11429 |
++ |
11430 |
++ if (dst == 0) |
11431 |
++ return 0; |
11432 |
++ |
11433 |
++ tmp = mul_u32_u32(src, dst - clip); |
11434 |
+ |
11435 |
+ /* |
11436 |
+ * Round toward 1.0 when clipping so that we don't accidentally |
11437 |
+diff --git a/drivers/gpu/drm/msm/disp/mdp4/mdp4_dsi_encoder.c b/drivers/gpu/drm/msm/disp/mdp4/mdp4_dsi_encoder.c |
11438 |
+index 6a1ebdace391..6253717d2e0c 100644 |
11439 |
+--- a/drivers/gpu/drm/msm/disp/mdp4/mdp4_dsi_encoder.c |
11440 |
++++ b/drivers/gpu/drm/msm/disp/mdp4/mdp4_dsi_encoder.c |
11441 |
+@@ -139,7 +139,7 @@ static void mdp4_dsi_encoder_enable(struct drm_encoder *encoder) |
11442 |
+ if (mdp4_dsi_encoder->enabled) |
11443 |
+ return; |
11444 |
+ |
11445 |
+- mdp4_crtc_set_config(encoder->crtc, |
11446 |
++ mdp4_crtc_set_config(encoder->crtc, |
11447 |
+ MDP4_DMA_CONFIG_PACK_ALIGN_MSB | |
11448 |
+ MDP4_DMA_CONFIG_DEFLKR_EN | |
11449 |
+ MDP4_DMA_CONFIG_DITHER_EN | |
11450 |
+diff --git a/drivers/gpu/drm/sun4i/sun6i_mipi_dsi.c b/drivers/gpu/drm/sun4i/sun6i_mipi_dsi.c |
11451 |
+index 97a0573cc514..79eb11cd185d 100644 |
11452 |
+--- a/drivers/gpu/drm/sun4i/sun6i_mipi_dsi.c |
11453 |
++++ b/drivers/gpu/drm/sun4i/sun6i_mipi_dsi.c |
11454 |
+@@ -357,8 +357,7 @@ static void sun6i_dsi_inst_init(struct sun6i_dsi *dsi, |
11455 |
+ static u16 sun6i_dsi_get_video_start_delay(struct sun6i_dsi *dsi, |
11456 |
+ struct drm_display_mode *mode) |
11457 |
+ { |
11458 |
+- u16 start = clamp(mode->vtotal - mode->vdisplay - 10, 8, 100); |
11459 |
+- u16 delay = mode->vtotal - (mode->vsync_end - mode->vdisplay) + start; |
11460 |
++ u16 delay = mode->vtotal - (mode->vsync_end - mode->vdisplay) + 1; |
11461 |
+ |
11462 |
+ if (delay > mode->vtotal) |
11463 |
+ delay = delay % mode->vtotal; |
11464 |
+diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c |
11465 |
+index d2a735ac9ba1..9ca0706a9d40 100644 |
11466 |
+--- a/drivers/hv/hv_balloon.c |
11467 |
++++ b/drivers/hv/hv_balloon.c |
11468 |
+@@ -1215,10 +1215,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm, |
11469 |
+ unsigned int i = 0; |
11470 |
+ struct page *pg; |
11471 |
+ |
11472 |
+- if (num_pages < alloc_unit) |
11473 |
+- return 0; |
11474 |
+- |
11475 |
+- for (i = 0; (i * alloc_unit) < num_pages; i++) { |
11476 |
++ for (i = 0; i < num_pages / alloc_unit; i++) { |
11477 |
+ if (bl_resp->hdr.size + sizeof(union dm_mem_page_range) > |
11478 |
+ PAGE_SIZE) |
11479 |
+ return i * alloc_unit; |
11480 |
+@@ -1252,7 +1249,7 @@ static unsigned int alloc_balloon_pages(struct hv_dynmem_device *dm, |
11481 |
+ |
11482 |
+ } |
11483 |
+ |
11484 |
+- return num_pages; |
11485 |
++ return i * alloc_unit; |
11486 |
+ } |
11487 |
+ |
11488 |
+ static void balloon_up(struct work_struct *dummy) |
11489 |
+@@ -1267,9 +1264,6 @@ static void balloon_up(struct work_struct *dummy) |
11490 |
+ long avail_pages; |
11491 |
+ unsigned long floor; |
11492 |
+ |
11493 |
+- /* The host balloons pages in 2M granularity. */ |
11494 |
+- WARN_ON_ONCE(num_pages % PAGES_IN_2M != 0); |
11495 |
+- |
11496 |
+ /* |
11497 |
+ * We will attempt 2M allocations. However, if we fail to |
11498 |
+ * allocate 2M chunks, we will go back to 4k allocations. |
11499 |
+@@ -1279,14 +1273,13 @@ static void balloon_up(struct work_struct *dummy) |
11500 |
+ avail_pages = si_mem_available(); |
11501 |
+ floor = compute_balloon_floor(); |
11502 |
+ |
11503 |
+- /* Refuse to balloon below the floor, keep the 2M granularity. */ |
11504 |
++ /* Refuse to balloon below the floor. */ |
11505 |
+ if (avail_pages < num_pages || avail_pages - num_pages < floor) { |
11506 |
+ pr_warn("Balloon request will be partially fulfilled. %s\n", |
11507 |
+ avail_pages < num_pages ? "Not enough memory." : |
11508 |
+ "Balloon floor reached."); |
11509 |
+ |
11510 |
+ num_pages = avail_pages > floor ? (avail_pages - floor) : 0; |
11511 |
+- num_pages -= num_pages % PAGES_IN_2M; |
11512 |
+ } |
11513 |
+ |
11514 |
+ while (!done) { |
11515 |
+diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c |
11516 |
+index 6ec748eccff7..4bda1242df87 100644 |
11517 |
+--- a/drivers/infiniband/core/umem_odp.c |
11518 |
++++ b/drivers/infiniband/core/umem_odp.c |
11519 |
+@@ -689,7 +689,7 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt, |
11520 |
+ |
11521 |
+ while (bcnt > 0) { |
11522 |
+ const size_t gup_num_pages = min_t(size_t, |
11523 |
+- (bcnt + BIT(page_shift) - 1) >> page_shift, |
11524 |
++ ALIGN(bcnt, PAGE_SIZE) / PAGE_SIZE, |
11525 |
+ PAGE_SIZE / sizeof(struct page *)); |
11526 |
+ |
11527 |
+ down_read(&owning_mm->mmap_sem); |
11528 |
+diff --git a/drivers/infiniband/hw/mlx5/gsi.c b/drivers/infiniband/hw/mlx5/gsi.c |
11529 |
+index 4950df3f71b6..5c73c0a790fa 100644 |
11530 |
+--- a/drivers/infiniband/hw/mlx5/gsi.c |
11531 |
++++ b/drivers/infiniband/hw/mlx5/gsi.c |
11532 |
+@@ -507,8 +507,7 @@ int mlx5_ib_gsi_post_send(struct ib_qp *qp, const struct ib_send_wr *wr, |
11533 |
+ ret = ib_post_send(tx_qp, &cur_wr.wr, bad_wr); |
11534 |
+ if (ret) { |
11535 |
+ /* Undo the effect of adding the outstanding wr */ |
11536 |
+- gsi->outstanding_pi = (gsi->outstanding_pi - 1) % |
11537 |
+- gsi->cap.max_send_wr; |
11538 |
++ gsi->outstanding_pi--; |
11539 |
+ goto err; |
11540 |
+ } |
11541 |
+ spin_unlock_irqrestore(&gsi->lock, flags); |
11542 |
+diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h |
11543 |
+index 4677b18ac281..2a2f189dd37c 100644 |
11544 |
+--- a/drivers/md/bcache/bcache.h |
11545 |
++++ b/drivers/md/bcache/bcache.h |
11546 |
+@@ -329,6 +329,9 @@ struct cached_dev { |
11547 |
+ */ |
11548 |
+ atomic_t has_dirty; |
11549 |
+ |
11550 |
++#define BCH_CACHE_READA_ALL 0 |
11551 |
++#define BCH_CACHE_READA_META_ONLY 1 |
11552 |
++ unsigned int cache_readahead_policy; |
11553 |
+ struct bch_ratelimit writeback_rate; |
11554 |
+ struct delayed_work writeback_rate_update; |
11555 |
+ |
11556 |
+diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c |
11557 |
+index 4ca3e3d3f9c7..c1e487d1261c 100644 |
11558 |
+--- a/drivers/md/bcache/request.c |
11559 |
++++ b/drivers/md/bcache/request.c |
11560 |
+@@ -391,13 +391,20 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio) |
11561 |
+ goto skip; |
11562 |
+ |
11563 |
+ /* |
11564 |
+- * Flag for bypass if the IO is for read-ahead or background, |
11565 |
+- * unless the read-ahead request is for metadata |
11566 |
++ * If the bio is for read-ahead or background IO, bypass it or |
11567 |
++ * not depends on the following situations, |
11568 |
++ * - If the IO is for meta data, always cache it and no bypass |
11569 |
++ * - If the IO is not meta data, check dc->cache_reada_policy, |
11570 |
++ * BCH_CACHE_READA_ALL: cache it and not bypass |
11571 |
++ * BCH_CACHE_READA_META_ONLY: not cache it and bypass |
11572 |
++ * That is, read-ahead request for metadata always get cached |
11573 |
+ * (eg, for gfs2 or xfs). |
11574 |
+ */ |
11575 |
+- if (bio->bi_opf & (REQ_RAHEAD|REQ_BACKGROUND) && |
11576 |
+- !(bio->bi_opf & (REQ_META|REQ_PRIO))) |
11577 |
+- goto skip; |
11578 |
++ if ((bio->bi_opf & (REQ_RAHEAD|REQ_BACKGROUND))) { |
11579 |
++ if (!(bio->bi_opf & (REQ_META|REQ_PRIO)) && |
11580 |
++ (dc->cache_readahead_policy != BCH_CACHE_READA_ALL)) |
11581 |
++ goto skip; |
11582 |
++ } |
11583 |
+ |
11584 |
+ if (bio->bi_iter.bi_sector & (c->sb.block_size - 1) || |
11585 |
+ bio_sectors(bio) & (c->sb.block_size - 1)) { |
11586 |
+diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c |
11587 |
+index 3e8d1f1b562f..591d9c8107dd 100644 |
11588 |
+--- a/drivers/md/bcache/sysfs.c |
11589 |
++++ b/drivers/md/bcache/sysfs.c |
11590 |
+@@ -25,6 +25,12 @@ static const char * const bch_cache_modes[] = { |
11591 |
+ NULL |
11592 |
+ }; |
11593 |
+ |
11594 |
++static const char * const bch_reada_cache_policies[] = { |
11595 |
++ "all", |
11596 |
++ "meta-only", |
11597 |
++ NULL |
11598 |
++}; |
11599 |
++ |
11600 |
+ /* Default is -1; we skip past it for stop_when_cache_set_failed */ |
11601 |
+ static const char * const bch_stop_on_failure_modes[] = { |
11602 |
+ "auto", |
11603 |
+@@ -94,6 +100,7 @@ rw_attribute(congested_write_threshold_us); |
11604 |
+ rw_attribute(sequential_cutoff); |
11605 |
+ rw_attribute(data_csum); |
11606 |
+ rw_attribute(cache_mode); |
11607 |
++rw_attribute(readahead_cache_policy); |
11608 |
+ rw_attribute(stop_when_cache_set_failed); |
11609 |
+ rw_attribute(writeback_metadata); |
11610 |
+ rw_attribute(writeback_running); |
11611 |
+@@ -160,6 +167,11 @@ SHOW(__bch_cached_dev) |
11612 |
+ bch_cache_modes, |
11613 |
+ BDEV_CACHE_MODE(&dc->sb)); |
11614 |
+ |
11615 |
++ if (attr == &sysfs_readahead_cache_policy) |
11616 |
++ return bch_snprint_string_list(buf, PAGE_SIZE, |
11617 |
++ bch_reada_cache_policies, |
11618 |
++ dc->cache_readahead_policy); |
11619 |
++ |
11620 |
+ if (attr == &sysfs_stop_when_cache_set_failed) |
11621 |
+ return bch_snprint_string_list(buf, PAGE_SIZE, |
11622 |
+ bch_stop_on_failure_modes, |
11623 |
+@@ -324,6 +336,15 @@ STORE(__cached_dev) |
11624 |
+ } |
11625 |
+ } |
11626 |
+ |
11627 |
++ if (attr == &sysfs_readahead_cache_policy) { |
11628 |
++ v = __sysfs_match_string(bch_reada_cache_policies, -1, buf); |
11629 |
++ if (v < 0) |
11630 |
++ return v; |
11631 |
++ |
11632 |
++ if ((unsigned int) v != dc->cache_readahead_policy) |
11633 |
++ dc->cache_readahead_policy = v; |
11634 |
++ } |
11635 |
++ |
11636 |
+ if (attr == &sysfs_stop_when_cache_set_failed) { |
11637 |
+ v = __sysfs_match_string(bch_stop_on_failure_modes, -1, buf); |
11638 |
+ if (v < 0) |
11639 |
+@@ -417,6 +438,7 @@ static struct attribute *bch_cached_dev_files[] = { |
11640 |
+ &sysfs_data_csum, |
11641 |
+ #endif |
11642 |
+ &sysfs_cache_mode, |
11643 |
++ &sysfs_readahead_cache_policy, |
11644 |
+ &sysfs_stop_when_cache_set_failed, |
11645 |
+ &sysfs_writeback_metadata, |
11646 |
+ &sysfs_writeback_running, |
11647 |
+diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c |
11648 |
+index 34f5de13a93d..750f8b34e693 100644 |
11649 |
+--- a/drivers/md/dm-crypt.c |
11650 |
++++ b/drivers/md/dm-crypt.c |
11651 |
+@@ -482,8 +482,14 @@ static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv, |
11652 |
+ static int crypt_iv_benbi_ctr(struct crypt_config *cc, struct dm_target *ti, |
11653 |
+ const char *opts) |
11654 |
+ { |
11655 |
+- unsigned bs = crypto_skcipher_blocksize(any_tfm(cc)); |
11656 |
+- int log = ilog2(bs); |
11657 |
++ unsigned bs; |
11658 |
++ int log; |
11659 |
++ |
11660 |
++ if (test_bit(CRYPT_MODE_INTEGRITY_AEAD, &cc->cipher_flags)) |
11661 |
++ bs = crypto_aead_blocksize(any_tfm_aead(cc)); |
11662 |
++ else |
11663 |
++ bs = crypto_skcipher_blocksize(any_tfm(cc)); |
11664 |
++ log = ilog2(bs); |
11665 |
+ |
11666 |
+ /* we need to calculate how far we must shift the sector count |
11667 |
+ * to get the cipher block count, we use this shift in _gen */ |
11668 |
+diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c |
11669 |
+index 503c4265ecbe..820c2e07dadf 100644 |
11670 |
+--- a/drivers/md/dm-writecache.c |
11671 |
++++ b/drivers/md/dm-writecache.c |
11672 |
+@@ -447,7 +447,13 @@ static void writecache_notify_io(unsigned long error, void *context) |
11673 |
+ complete(&endio->c); |
11674 |
+ } |
11675 |
+ |
11676 |
+-static void ssd_commit_flushed(struct dm_writecache *wc) |
11677 |
++static void writecache_wait_for_ios(struct dm_writecache *wc, int direction) |
11678 |
++{ |
11679 |
++ wait_event(wc->bio_in_progress_wait[direction], |
11680 |
++ !atomic_read(&wc->bio_in_progress[direction])); |
11681 |
++} |
11682 |
++ |
11683 |
++static void ssd_commit_flushed(struct dm_writecache *wc, bool wait_for_ios) |
11684 |
+ { |
11685 |
+ struct dm_io_region region; |
11686 |
+ struct dm_io_request req; |
11687 |
+@@ -493,17 +499,20 @@ static void ssd_commit_flushed(struct dm_writecache *wc) |
11688 |
+ writecache_notify_io(0, &endio); |
11689 |
+ wait_for_completion_io(&endio.c); |
11690 |
+ |
11691 |
++ if (wait_for_ios) |
11692 |
++ writecache_wait_for_ios(wc, WRITE); |
11693 |
++ |
11694 |
+ writecache_disk_flush(wc, wc->ssd_dev); |
11695 |
+ |
11696 |
+ memset(wc->dirty_bitmap, 0, wc->dirty_bitmap_size); |
11697 |
+ } |
11698 |
+ |
11699 |
+-static void writecache_commit_flushed(struct dm_writecache *wc) |
11700 |
++static void writecache_commit_flushed(struct dm_writecache *wc, bool wait_for_ios) |
11701 |
+ { |
11702 |
+ if (WC_MODE_PMEM(wc)) |
11703 |
+ wmb(); |
11704 |
+ else |
11705 |
+- ssd_commit_flushed(wc); |
11706 |
++ ssd_commit_flushed(wc, wait_for_ios); |
11707 |
+ } |
11708 |
+ |
11709 |
+ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev) |
11710 |
+@@ -527,12 +536,6 @@ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev) |
11711 |
+ writecache_error(wc, r, "error flushing metadata: %d", r); |
11712 |
+ } |
11713 |
+ |
11714 |
+-static void writecache_wait_for_ios(struct dm_writecache *wc, int direction) |
11715 |
+-{ |
11716 |
+- wait_event(wc->bio_in_progress_wait[direction], |
11717 |
+- !atomic_read(&wc->bio_in_progress[direction])); |
11718 |
+-} |
11719 |
+- |
11720 |
+ #define WFE_RETURN_FOLLOWING 1 |
11721 |
+ #define WFE_LOWEST_SEQ 2 |
11722 |
+ |
11723 |
+@@ -730,14 +733,12 @@ static void writecache_flush(struct dm_writecache *wc) |
11724 |
+ e = e2; |
11725 |
+ cond_resched(); |
11726 |
+ } |
11727 |
+- writecache_commit_flushed(wc); |
11728 |
+- |
11729 |
+- writecache_wait_for_ios(wc, WRITE); |
11730 |
++ writecache_commit_flushed(wc, true); |
11731 |
+ |
11732 |
+ wc->seq_count++; |
11733 |
+ pmem_assign(sb(wc)->seq_count, cpu_to_le64(wc->seq_count)); |
11734 |
+ writecache_flush_region(wc, &sb(wc)->seq_count, sizeof sb(wc)->seq_count); |
11735 |
+- writecache_commit_flushed(wc); |
11736 |
++ writecache_commit_flushed(wc, false); |
11737 |
+ |
11738 |
+ wc->overwrote_committed = false; |
11739 |
+ |
11740 |
+@@ -761,7 +762,7 @@ static void writecache_flush(struct dm_writecache *wc) |
11741 |
+ } |
11742 |
+ |
11743 |
+ if (need_flush_after_free) |
11744 |
+- writecache_commit_flushed(wc); |
11745 |
++ writecache_commit_flushed(wc, false); |
11746 |
+ } |
11747 |
+ |
11748 |
+ static void writecache_flush_work(struct work_struct *work) |
11749 |
+@@ -814,7 +815,7 @@ static void writecache_discard(struct dm_writecache *wc, sector_t start, sector_ |
11750 |
+ } |
11751 |
+ |
11752 |
+ if (discarded_something) |
11753 |
+- writecache_commit_flushed(wc); |
11754 |
++ writecache_commit_flushed(wc, false); |
11755 |
+ } |
11756 |
+ |
11757 |
+ static bool writecache_wait_for_writeback(struct dm_writecache *wc) |
11758 |
+@@ -963,7 +964,7 @@ erase_this: |
11759 |
+ |
11760 |
+ if (need_flush) { |
11761 |
+ writecache_flush_all_metadata(wc); |
11762 |
+- writecache_commit_flushed(wc); |
11763 |
++ writecache_commit_flushed(wc, false); |
11764 |
+ } |
11765 |
+ |
11766 |
+ wc_unlock(wc); |
11767 |
+@@ -1347,7 +1348,7 @@ static void __writecache_endio_pmem(struct dm_writecache *wc, struct list_head * |
11768 |
+ wc->writeback_size--; |
11769 |
+ n_walked++; |
11770 |
+ if (unlikely(n_walked >= ENDIO_LATENCY)) { |
11771 |
+- writecache_commit_flushed(wc); |
11772 |
++ writecache_commit_flushed(wc, false); |
11773 |
+ wc_unlock(wc); |
11774 |
+ wc_lock(wc); |
11775 |
+ n_walked = 0; |
11776 |
+@@ -1428,7 +1429,7 @@ pop_from_list: |
11777 |
+ writecache_wait_for_ios(wc, READ); |
11778 |
+ } |
11779 |
+ |
11780 |
+- writecache_commit_flushed(wc); |
11781 |
++ writecache_commit_flushed(wc, false); |
11782 |
+ |
11783 |
+ wc_unlock(wc); |
11784 |
+ } |
11785 |
+@@ -1759,10 +1760,10 @@ static int init_memory(struct dm_writecache *wc) |
11786 |
+ write_original_sector_seq_count(wc, &wc->entries[b], -1, -1); |
11787 |
+ |
11788 |
+ writecache_flush_all_metadata(wc); |
11789 |
+- writecache_commit_flushed(wc); |
11790 |
++ writecache_commit_flushed(wc, false); |
11791 |
+ pmem_assign(sb(wc)->magic, cpu_to_le32(MEMORY_SUPERBLOCK_MAGIC)); |
11792 |
+ writecache_flush_region(wc, &sb(wc)->magic, sizeof sb(wc)->magic); |
11793 |
+- writecache_commit_flushed(wc); |
11794 |
++ writecache_commit_flushed(wc, false); |
11795 |
+ |
11796 |
+ return 0; |
11797 |
+ } |
11798 |
+diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c |
11799 |
+index c2c17149d968..086a870087cf 100644 |
11800 |
+--- a/drivers/md/dm-zoned-metadata.c |
11801 |
++++ b/drivers/md/dm-zoned-metadata.c |
11802 |
+@@ -132,6 +132,7 @@ struct dmz_metadata { |
11803 |
+ |
11804 |
+ sector_t zone_bitmap_size; |
11805 |
+ unsigned int zone_nr_bitmap_blocks; |
11806 |
++ unsigned int zone_bits_per_mblk; |
11807 |
+ |
11808 |
+ unsigned int nr_bitmap_blocks; |
11809 |
+ unsigned int nr_map_blocks; |
11810 |
+@@ -1165,7 +1166,10 @@ static int dmz_init_zones(struct dmz_metadata *zmd) |
11811 |
+ |
11812 |
+ /* Init */ |
11813 |
+ zmd->zone_bitmap_size = dev->zone_nr_blocks >> 3; |
11814 |
+- zmd->zone_nr_bitmap_blocks = zmd->zone_bitmap_size >> DMZ_BLOCK_SHIFT; |
11815 |
++ zmd->zone_nr_bitmap_blocks = |
11816 |
++ max_t(sector_t, 1, zmd->zone_bitmap_size >> DMZ_BLOCK_SHIFT); |
11817 |
++ zmd->zone_bits_per_mblk = min_t(sector_t, dev->zone_nr_blocks, |
11818 |
++ DMZ_BLOCK_SIZE_BITS); |
11819 |
+ |
11820 |
+ /* Allocate zone array */ |
11821 |
+ zmd->zones = kcalloc(dev->nr_zones, sizeof(struct dm_zone), GFP_KERNEL); |
11822 |
+@@ -1982,7 +1986,7 @@ int dmz_copy_valid_blocks(struct dmz_metadata *zmd, struct dm_zone *from_zone, |
11823 |
+ dmz_release_mblock(zmd, to_mblk); |
11824 |
+ dmz_release_mblock(zmd, from_mblk); |
11825 |
+ |
11826 |
+- chunk_block += DMZ_BLOCK_SIZE_BITS; |
11827 |
++ chunk_block += zmd->zone_bits_per_mblk; |
11828 |
+ } |
11829 |
+ |
11830 |
+ to_zone->weight = from_zone->weight; |
11831 |
+@@ -2043,7 +2047,7 @@ int dmz_validate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone, |
11832 |
+ |
11833 |
+ /* Set bits */ |
11834 |
+ bit = chunk_block & DMZ_BLOCK_MASK_BITS; |
11835 |
+- nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); |
11836 |
++ nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); |
11837 |
+ |
11838 |
+ count = dmz_set_bits((unsigned long *)mblk->data, bit, nr_bits); |
11839 |
+ if (count) { |
11840 |
+@@ -2122,7 +2126,7 @@ int dmz_invalidate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone, |
11841 |
+ |
11842 |
+ /* Clear bits */ |
11843 |
+ bit = chunk_block & DMZ_BLOCK_MASK_BITS; |
11844 |
+- nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); |
11845 |
++ nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); |
11846 |
+ |
11847 |
+ count = dmz_clear_bits((unsigned long *)mblk->data, |
11848 |
+ bit, nr_bits); |
11849 |
+@@ -2182,6 +2186,7 @@ static int dmz_to_next_set_block(struct dmz_metadata *zmd, struct dm_zone *zone, |
11850 |
+ { |
11851 |
+ struct dmz_mblock *mblk; |
11852 |
+ unsigned int bit, set_bit, nr_bits; |
11853 |
++ unsigned int zone_bits = zmd->zone_bits_per_mblk; |
11854 |
+ unsigned long *bitmap; |
11855 |
+ int n = 0; |
11856 |
+ |
11857 |
+@@ -2196,15 +2201,15 @@ static int dmz_to_next_set_block(struct dmz_metadata *zmd, struct dm_zone *zone, |
11858 |
+ /* Get offset */ |
11859 |
+ bitmap = (unsigned long *) mblk->data; |
11860 |
+ bit = chunk_block & DMZ_BLOCK_MASK_BITS; |
11861 |
+- nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); |
11862 |
++ nr_bits = min(nr_blocks, zone_bits - bit); |
11863 |
+ if (set) |
11864 |
+- set_bit = find_next_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit); |
11865 |
++ set_bit = find_next_bit(bitmap, zone_bits, bit); |
11866 |
+ else |
11867 |
+- set_bit = find_next_zero_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit); |
11868 |
++ set_bit = find_next_zero_bit(bitmap, zone_bits, bit); |
11869 |
+ dmz_release_mblock(zmd, mblk); |
11870 |
+ |
11871 |
+ n += set_bit - bit; |
11872 |
+- if (set_bit < DMZ_BLOCK_SIZE_BITS) |
11873 |
++ if (set_bit < zone_bits) |
11874 |
+ break; |
11875 |
+ |
11876 |
+ nr_blocks -= nr_bits; |
11877 |
+@@ -2307,7 +2312,7 @@ static void dmz_get_zone_weight(struct dmz_metadata *zmd, struct dm_zone *zone) |
11878 |
+ /* Count bits in this block */ |
11879 |
+ bitmap = mblk->data; |
11880 |
+ bit = chunk_block & DMZ_BLOCK_MASK_BITS; |
11881 |
+- nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); |
11882 |
++ nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); |
11883 |
+ n += dmz_count_bits(bitmap, bit, nr_bits); |
11884 |
+ |
11885 |
+ dmz_release_mblock(zmd, mblk); |
11886 |
+diff --git a/drivers/md/dm.c b/drivers/md/dm.c |
11887 |
+index c9860e3b04dd..3965f3cf8ea1 100644 |
11888 |
+--- a/drivers/md/dm.c |
11889 |
++++ b/drivers/md/dm.c |
11890 |
+@@ -1819,6 +1819,7 @@ static void dm_init_normal_md_queue(struct mapped_device *md) |
11891 |
+ /* |
11892 |
+ * Initialize aspects of queue that aren't relevant for blk-mq |
11893 |
+ */ |
11894 |
++ md->queue->backing_dev_info->congested_data = md; |
11895 |
+ md->queue->backing_dev_info->congested_fn = dm_any_congested; |
11896 |
+ } |
11897 |
+ |
11898 |
+@@ -1913,7 +1914,12 @@ static struct mapped_device *alloc_dev(int minor) |
11899 |
+ if (!md->queue) |
11900 |
+ goto bad; |
11901 |
+ md->queue->queuedata = md; |
11902 |
+- md->queue->backing_dev_info->congested_data = md; |
11903 |
++ /* |
11904 |
++ * default to bio-based required ->make_request_fn until DM |
11905 |
++ * table is loaded and md->type established. If request-based |
11906 |
++ * table is loaded: blk-mq will override accordingly. |
11907 |
++ */ |
11908 |
++ blk_queue_make_request(md->queue, dm_make_request); |
11909 |
+ |
11910 |
+ md->disk = alloc_disk_node(1, md->numa_node_id); |
11911 |
+ if (!md->disk) |
11912 |
+@@ -2242,7 +2248,6 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) |
11913 |
+ case DM_TYPE_BIO_BASED: |
11914 |
+ case DM_TYPE_DAX_BIO_BASED: |
11915 |
+ dm_init_normal_md_queue(md); |
11916 |
+- blk_queue_make_request(md->queue, dm_make_request); |
11917 |
+ break; |
11918 |
+ case DM_TYPE_NVME_BIO_BASED: |
11919 |
+ dm_init_normal_md_queue(md); |
11920 |
+diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c |
11921 |
+index 0a3b8ae4a29c..17aef55ed708 100644 |
11922 |
+--- a/drivers/md/persistent-data/dm-space-map-common.c |
11923 |
++++ b/drivers/md/persistent-data/dm-space-map-common.c |
11924 |
+@@ -382,6 +382,33 @@ int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin, |
11925 |
+ return -ENOSPC; |
11926 |
+ } |
11927 |
+ |
11928 |
++int sm_ll_find_common_free_block(struct ll_disk *old_ll, struct ll_disk *new_ll, |
11929 |
++ dm_block_t begin, dm_block_t end, dm_block_t *b) |
11930 |
++{ |
11931 |
++ int r; |
11932 |
++ uint32_t count; |
11933 |
++ |
11934 |
++ do { |
11935 |
++ r = sm_ll_find_free_block(new_ll, begin, new_ll->nr_blocks, b); |
11936 |
++ if (r) |
11937 |
++ break; |
11938 |
++ |
11939 |
++ /* double check this block wasn't used in the old transaction */ |
11940 |
++ if (*b >= old_ll->nr_blocks) |
11941 |
++ count = 0; |
11942 |
++ else { |
11943 |
++ r = sm_ll_lookup(old_ll, *b, &count); |
11944 |
++ if (r) |
11945 |
++ break; |
11946 |
++ |
11947 |
++ if (count) |
11948 |
++ begin = *b + 1; |
11949 |
++ } |
11950 |
++ } while (count); |
11951 |
++ |
11952 |
++ return r; |
11953 |
++} |
11954 |
++ |
11955 |
+ static int sm_ll_mutate(struct ll_disk *ll, dm_block_t b, |
11956 |
+ int (*mutator)(void *context, uint32_t old, uint32_t *new), |
11957 |
+ void *context, enum allocation_event *ev) |
11958 |
+diff --git a/drivers/md/persistent-data/dm-space-map-common.h b/drivers/md/persistent-data/dm-space-map-common.h |
11959 |
+index b3078d5eda0c..8de63ce39bdd 100644 |
11960 |
+--- a/drivers/md/persistent-data/dm-space-map-common.h |
11961 |
++++ b/drivers/md/persistent-data/dm-space-map-common.h |
11962 |
+@@ -109,6 +109,8 @@ int sm_ll_lookup_bitmap(struct ll_disk *ll, dm_block_t b, uint32_t *result); |
11963 |
+ int sm_ll_lookup(struct ll_disk *ll, dm_block_t b, uint32_t *result); |
11964 |
+ int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin, |
11965 |
+ dm_block_t end, dm_block_t *result); |
11966 |
++int sm_ll_find_common_free_block(struct ll_disk *old_ll, struct ll_disk *new_ll, |
11967 |
++ dm_block_t begin, dm_block_t end, dm_block_t *result); |
11968 |
+ int sm_ll_insert(struct ll_disk *ll, dm_block_t b, uint32_t ref_count, enum allocation_event *ev); |
11969 |
+ int sm_ll_inc(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev); |
11970 |
+ int sm_ll_dec(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev); |
11971 |
+diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c |
11972 |
+index 32adf6b4a9c7..bf4c5e2ccb6f 100644 |
11973 |
+--- a/drivers/md/persistent-data/dm-space-map-disk.c |
11974 |
++++ b/drivers/md/persistent-data/dm-space-map-disk.c |
11975 |
+@@ -167,8 +167,10 @@ static int sm_disk_new_block(struct dm_space_map *sm, dm_block_t *b) |
11976 |
+ enum allocation_event ev; |
11977 |
+ struct sm_disk *smd = container_of(sm, struct sm_disk, sm); |
11978 |
+ |
11979 |
+- /* FIXME: we should loop round a couple of times */ |
11980 |
+- r = sm_ll_find_free_block(&smd->old_ll, smd->begin, smd->old_ll.nr_blocks, b); |
11981 |
++ /* |
11982 |
++ * Any block we allocate has to be free in both the old and current ll. |
11983 |
++ */ |
11984 |
++ r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, smd->begin, smd->ll.nr_blocks, b); |
11985 |
+ if (r) |
11986 |
+ return r; |
11987 |
+ |
11988 |
+diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c |
11989 |
+index 25328582cc48..9e3c64ec2026 100644 |
11990 |
+--- a/drivers/md/persistent-data/dm-space-map-metadata.c |
11991 |
++++ b/drivers/md/persistent-data/dm-space-map-metadata.c |
11992 |
+@@ -448,7 +448,10 @@ static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b) |
11993 |
+ enum allocation_event ev; |
11994 |
+ struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm); |
11995 |
+ |
11996 |
+- r = sm_ll_find_free_block(&smm->old_ll, smm->begin, smm->old_ll.nr_blocks, b); |
11997 |
++ /* |
11998 |
++ * Any block we allocate has to be free in both the old and current ll. |
11999 |
++ */ |
12000 |
++ r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, smm->begin, smm->ll.nr_blocks, b); |
12001 |
+ if (r) |
12002 |
+ return r; |
12003 |
+ |
12004 |
+diff --git a/drivers/media/rc/iguanair.c b/drivers/media/rc/iguanair.c |
12005 |
+index 6f3030b2054d..1df9522c30fa 100644 |
12006 |
+--- a/drivers/media/rc/iguanair.c |
12007 |
++++ b/drivers/media/rc/iguanair.c |
12008 |
+@@ -424,7 +424,7 @@ static int iguanair_probe(struct usb_interface *intf, |
12009 |
+ int ret, pipein, pipeout; |
12010 |
+ struct usb_host_interface *idesc; |
12011 |
+ |
12012 |
+- idesc = intf->altsetting; |
12013 |
++ idesc = intf->cur_altsetting; |
12014 |
+ if (idesc->desc.bNumEndpoints < 2) |
12015 |
+ return -ENODEV; |
12016 |
+ |
12017 |
+diff --git a/drivers/media/rc/rc-main.c b/drivers/media/rc/rc-main.c |
12018 |
+index 0f218afdadaa..c30affbd43a9 100644 |
12019 |
+--- a/drivers/media/rc/rc-main.c |
12020 |
++++ b/drivers/media/rc/rc-main.c |
12021 |
+@@ -1874,23 +1874,28 @@ int rc_register_device(struct rc_dev *dev) |
12022 |
+ |
12023 |
+ dev->registered = true; |
12024 |
+ |
12025 |
+- if (dev->driver_type != RC_DRIVER_IR_RAW_TX) { |
12026 |
+- rc = rc_setup_rx_device(dev); |
12027 |
+- if (rc) |
12028 |
+- goto out_dev; |
12029 |
+- } |
12030 |
+- |
12031 |
+- /* Ensure that the lirc kfifo is setup before we start the thread */ |
12032 |
++ /* |
12033 |
++ * once the the input device is registered in rc_setup_rx_device, |
12034 |
++ * userspace can open the input device and rc_open() will be called |
12035 |
++ * as a result. This results in driver code being allowed to submit |
12036 |
++ * keycodes with rc_keydown, so lirc must be registered first. |
12037 |
++ */ |
12038 |
+ if (dev->allowed_protocols != RC_PROTO_BIT_CEC) { |
12039 |
+ rc = ir_lirc_register(dev); |
12040 |
+ if (rc < 0) |
12041 |
+- goto out_rx; |
12042 |
++ goto out_dev; |
12043 |
++ } |
12044 |
++ |
12045 |
++ if (dev->driver_type != RC_DRIVER_IR_RAW_TX) { |
12046 |
++ rc = rc_setup_rx_device(dev); |
12047 |
++ if (rc) |
12048 |
++ goto out_lirc; |
12049 |
+ } |
12050 |
+ |
12051 |
+ if (dev->driver_type == RC_DRIVER_IR_RAW) { |
12052 |
+ rc = ir_raw_event_register(dev); |
12053 |
+ if (rc < 0) |
12054 |
+- goto out_lirc; |
12055 |
++ goto out_rx; |
12056 |
+ } |
12057 |
+ |
12058 |
+ dev_dbg(&dev->dev, "Registered rc%u (driver: %s)\n", dev->minor, |
12059 |
+@@ -1898,11 +1903,11 @@ int rc_register_device(struct rc_dev *dev) |
12060 |
+ |
12061 |
+ return 0; |
12062 |
+ |
12063 |
++out_rx: |
12064 |
++ rc_free_rx_device(dev); |
12065 |
+ out_lirc: |
12066 |
+ if (dev->allowed_protocols != RC_PROTO_BIT_CEC) |
12067 |
+ ir_lirc_unregister(dev); |
12068 |
+-out_rx: |
12069 |
+- rc_free_rx_device(dev); |
12070 |
+ out_dev: |
12071 |
+ device_del(&dev->dev); |
12072 |
+ out_rx_free: |
12073 |
+diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c |
12074 |
+index 063e229ead5e..38c73cdbef70 100644 |
12075 |
+--- a/drivers/media/usb/uvc/uvc_driver.c |
12076 |
++++ b/drivers/media/usb/uvc/uvc_driver.c |
12077 |
+@@ -1482,6 +1482,11 @@ static int uvc_scan_chain_forward(struct uvc_video_chain *chain, |
12078 |
+ break; |
12079 |
+ if (forward == prev) |
12080 |
+ continue; |
12081 |
++ if (forward->chain.next || forward->chain.prev) { |
12082 |
++ uvc_trace(UVC_TRACE_DESCR, "Found reference to " |
12083 |
++ "entity %d already in chain.\n", forward->id); |
12084 |
++ return -EINVAL; |
12085 |
++ } |
12086 |
+ |
12087 |
+ switch (UVC_ENTITY_TYPE(forward)) { |
12088 |
+ case UVC_VC_EXTENSION_UNIT: |
12089 |
+@@ -1563,6 +1568,13 @@ static int uvc_scan_chain_backward(struct uvc_video_chain *chain, |
12090 |
+ return -1; |
12091 |
+ } |
12092 |
+ |
12093 |
++ if (term->chain.next || term->chain.prev) { |
12094 |
++ uvc_trace(UVC_TRACE_DESCR, "Found reference to " |
12095 |
++ "entity %d already in chain.\n", |
12096 |
++ term->id); |
12097 |
++ return -EINVAL; |
12098 |
++ } |
12099 |
++ |
12100 |
+ if (uvc_trace_param & UVC_TRACE_PROBE) |
12101 |
+ printk(KERN_CONT " %d", term->id); |
12102 |
+ |
12103 |
+diff --git a/drivers/media/v4l2-core/v4l2-compat-ioctl32.c b/drivers/media/v4l2-core/v4l2-compat-ioctl32.c |
12104 |
+index 6481212fda77..3efe4e0a80a4 100644 |
12105 |
+--- a/drivers/media/v4l2-core/v4l2-compat-ioctl32.c |
12106 |
++++ b/drivers/media/v4l2-core/v4l2-compat-ioctl32.c |
12107 |
+@@ -1171,36 +1171,38 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12108 |
+ u32 aux_space; |
12109 |
+ int compatible_arg = 1; |
12110 |
+ long err = 0; |
12111 |
++ unsigned int ncmd; |
12112 |
+ |
12113 |
+ /* |
12114 |
+ * 1. When struct size is different, converts the command. |
12115 |
+ */ |
12116 |
+ switch (cmd) { |
12117 |
+- case VIDIOC_G_FMT32: cmd = VIDIOC_G_FMT; break; |
12118 |
+- case VIDIOC_S_FMT32: cmd = VIDIOC_S_FMT; break; |
12119 |
+- case VIDIOC_QUERYBUF32: cmd = VIDIOC_QUERYBUF; break; |
12120 |
+- case VIDIOC_G_FBUF32: cmd = VIDIOC_G_FBUF; break; |
12121 |
+- case VIDIOC_S_FBUF32: cmd = VIDIOC_S_FBUF; break; |
12122 |
+- case VIDIOC_QBUF32: cmd = VIDIOC_QBUF; break; |
12123 |
+- case VIDIOC_DQBUF32: cmd = VIDIOC_DQBUF; break; |
12124 |
+- case VIDIOC_ENUMSTD32: cmd = VIDIOC_ENUMSTD; break; |
12125 |
+- case VIDIOC_ENUMINPUT32: cmd = VIDIOC_ENUMINPUT; break; |
12126 |
+- case VIDIOC_TRY_FMT32: cmd = VIDIOC_TRY_FMT; break; |
12127 |
+- case VIDIOC_G_EXT_CTRLS32: cmd = VIDIOC_G_EXT_CTRLS; break; |
12128 |
+- case VIDIOC_S_EXT_CTRLS32: cmd = VIDIOC_S_EXT_CTRLS; break; |
12129 |
+- case VIDIOC_TRY_EXT_CTRLS32: cmd = VIDIOC_TRY_EXT_CTRLS; break; |
12130 |
+- case VIDIOC_DQEVENT32: cmd = VIDIOC_DQEVENT; break; |
12131 |
+- case VIDIOC_OVERLAY32: cmd = VIDIOC_OVERLAY; break; |
12132 |
+- case VIDIOC_STREAMON32: cmd = VIDIOC_STREAMON; break; |
12133 |
+- case VIDIOC_STREAMOFF32: cmd = VIDIOC_STREAMOFF; break; |
12134 |
+- case VIDIOC_G_INPUT32: cmd = VIDIOC_G_INPUT; break; |
12135 |
+- case VIDIOC_S_INPUT32: cmd = VIDIOC_S_INPUT; break; |
12136 |
+- case VIDIOC_G_OUTPUT32: cmd = VIDIOC_G_OUTPUT; break; |
12137 |
+- case VIDIOC_S_OUTPUT32: cmd = VIDIOC_S_OUTPUT; break; |
12138 |
+- case VIDIOC_CREATE_BUFS32: cmd = VIDIOC_CREATE_BUFS; break; |
12139 |
+- case VIDIOC_PREPARE_BUF32: cmd = VIDIOC_PREPARE_BUF; break; |
12140 |
+- case VIDIOC_G_EDID32: cmd = VIDIOC_G_EDID; break; |
12141 |
+- case VIDIOC_S_EDID32: cmd = VIDIOC_S_EDID; break; |
12142 |
++ case VIDIOC_G_FMT32: ncmd = VIDIOC_G_FMT; break; |
12143 |
++ case VIDIOC_S_FMT32: ncmd = VIDIOC_S_FMT; break; |
12144 |
++ case VIDIOC_QUERYBUF32: ncmd = VIDIOC_QUERYBUF; break; |
12145 |
++ case VIDIOC_G_FBUF32: ncmd = VIDIOC_G_FBUF; break; |
12146 |
++ case VIDIOC_S_FBUF32: ncmd = VIDIOC_S_FBUF; break; |
12147 |
++ case VIDIOC_QBUF32: ncmd = VIDIOC_QBUF; break; |
12148 |
++ case VIDIOC_DQBUF32: ncmd = VIDIOC_DQBUF; break; |
12149 |
++ case VIDIOC_ENUMSTD32: ncmd = VIDIOC_ENUMSTD; break; |
12150 |
++ case VIDIOC_ENUMINPUT32: ncmd = VIDIOC_ENUMINPUT; break; |
12151 |
++ case VIDIOC_TRY_FMT32: ncmd = VIDIOC_TRY_FMT; break; |
12152 |
++ case VIDIOC_G_EXT_CTRLS32: ncmd = VIDIOC_G_EXT_CTRLS; break; |
12153 |
++ case VIDIOC_S_EXT_CTRLS32: ncmd = VIDIOC_S_EXT_CTRLS; break; |
12154 |
++ case VIDIOC_TRY_EXT_CTRLS32: ncmd = VIDIOC_TRY_EXT_CTRLS; break; |
12155 |
++ case VIDIOC_DQEVENT32: ncmd = VIDIOC_DQEVENT; break; |
12156 |
++ case VIDIOC_OVERLAY32: ncmd = VIDIOC_OVERLAY; break; |
12157 |
++ case VIDIOC_STREAMON32: ncmd = VIDIOC_STREAMON; break; |
12158 |
++ case VIDIOC_STREAMOFF32: ncmd = VIDIOC_STREAMOFF; break; |
12159 |
++ case VIDIOC_G_INPUT32: ncmd = VIDIOC_G_INPUT; break; |
12160 |
++ case VIDIOC_S_INPUT32: ncmd = VIDIOC_S_INPUT; break; |
12161 |
++ case VIDIOC_G_OUTPUT32: ncmd = VIDIOC_G_OUTPUT; break; |
12162 |
++ case VIDIOC_S_OUTPUT32: ncmd = VIDIOC_S_OUTPUT; break; |
12163 |
++ case VIDIOC_CREATE_BUFS32: ncmd = VIDIOC_CREATE_BUFS; break; |
12164 |
++ case VIDIOC_PREPARE_BUF32: ncmd = VIDIOC_PREPARE_BUF; break; |
12165 |
++ case VIDIOC_G_EDID32: ncmd = VIDIOC_G_EDID; break; |
12166 |
++ case VIDIOC_S_EDID32: ncmd = VIDIOC_S_EDID; break; |
12167 |
++ default: ncmd = cmd; break; |
12168 |
+ } |
12169 |
+ |
12170 |
+ /* |
12171 |
+@@ -1209,11 +1211,11 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12172 |
+ * argument into it. |
12173 |
+ */ |
12174 |
+ switch (cmd) { |
12175 |
+- case VIDIOC_OVERLAY: |
12176 |
+- case VIDIOC_STREAMON: |
12177 |
+- case VIDIOC_STREAMOFF: |
12178 |
+- case VIDIOC_S_INPUT: |
12179 |
+- case VIDIOC_S_OUTPUT: |
12180 |
++ case VIDIOC_OVERLAY32: |
12181 |
++ case VIDIOC_STREAMON32: |
12182 |
++ case VIDIOC_STREAMOFF32: |
12183 |
++ case VIDIOC_S_INPUT32: |
12184 |
++ case VIDIOC_S_OUTPUT32: |
12185 |
+ err = alloc_userspace(sizeof(unsigned int), 0, &new_p64); |
12186 |
+ if (!err && assign_in_user((unsigned int __user *)new_p64, |
12187 |
+ (compat_uint_t __user *)p32)) |
12188 |
+@@ -1221,23 +1223,23 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12189 |
+ compatible_arg = 0; |
12190 |
+ break; |
12191 |
+ |
12192 |
+- case VIDIOC_G_INPUT: |
12193 |
+- case VIDIOC_G_OUTPUT: |
12194 |
++ case VIDIOC_G_INPUT32: |
12195 |
++ case VIDIOC_G_OUTPUT32: |
12196 |
+ err = alloc_userspace(sizeof(unsigned int), 0, &new_p64); |
12197 |
+ compatible_arg = 0; |
12198 |
+ break; |
12199 |
+ |
12200 |
+- case VIDIOC_G_EDID: |
12201 |
+- case VIDIOC_S_EDID: |
12202 |
++ case VIDIOC_G_EDID32: |
12203 |
++ case VIDIOC_S_EDID32: |
12204 |
+ err = alloc_userspace(sizeof(struct v4l2_edid), 0, &new_p64); |
12205 |
+ if (!err) |
12206 |
+ err = get_v4l2_edid32(new_p64, p32); |
12207 |
+ compatible_arg = 0; |
12208 |
+ break; |
12209 |
+ |
12210 |
+- case VIDIOC_G_FMT: |
12211 |
+- case VIDIOC_S_FMT: |
12212 |
+- case VIDIOC_TRY_FMT: |
12213 |
++ case VIDIOC_G_FMT32: |
12214 |
++ case VIDIOC_S_FMT32: |
12215 |
++ case VIDIOC_TRY_FMT32: |
12216 |
+ err = bufsize_v4l2_format(p32, &aux_space); |
12217 |
+ if (!err) |
12218 |
+ err = alloc_userspace(sizeof(struct v4l2_format), |
12219 |
+@@ -1250,7 +1252,7 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12220 |
+ compatible_arg = 0; |
12221 |
+ break; |
12222 |
+ |
12223 |
+- case VIDIOC_CREATE_BUFS: |
12224 |
++ case VIDIOC_CREATE_BUFS32: |
12225 |
+ err = bufsize_v4l2_create(p32, &aux_space); |
12226 |
+ if (!err) |
12227 |
+ err = alloc_userspace(sizeof(struct v4l2_create_buffers), |
12228 |
+@@ -1263,10 +1265,10 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12229 |
+ compatible_arg = 0; |
12230 |
+ break; |
12231 |
+ |
12232 |
+- case VIDIOC_PREPARE_BUF: |
12233 |
+- case VIDIOC_QUERYBUF: |
12234 |
+- case VIDIOC_QBUF: |
12235 |
+- case VIDIOC_DQBUF: |
12236 |
++ case VIDIOC_PREPARE_BUF32: |
12237 |
++ case VIDIOC_QUERYBUF32: |
12238 |
++ case VIDIOC_QBUF32: |
12239 |
++ case VIDIOC_DQBUF32: |
12240 |
+ err = bufsize_v4l2_buffer(p32, &aux_space); |
12241 |
+ if (!err) |
12242 |
+ err = alloc_userspace(sizeof(struct v4l2_buffer), |
12243 |
+@@ -1279,7 +1281,7 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12244 |
+ compatible_arg = 0; |
12245 |
+ break; |
12246 |
+ |
12247 |
+- case VIDIOC_S_FBUF: |
12248 |
++ case VIDIOC_S_FBUF32: |
12249 |
+ err = alloc_userspace(sizeof(struct v4l2_framebuffer), 0, |
12250 |
+ &new_p64); |
12251 |
+ if (!err) |
12252 |
+@@ -1287,13 +1289,13 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12253 |
+ compatible_arg = 0; |
12254 |
+ break; |
12255 |
+ |
12256 |
+- case VIDIOC_G_FBUF: |
12257 |
++ case VIDIOC_G_FBUF32: |
12258 |
+ err = alloc_userspace(sizeof(struct v4l2_framebuffer), 0, |
12259 |
+ &new_p64); |
12260 |
+ compatible_arg = 0; |
12261 |
+ break; |
12262 |
+ |
12263 |
+- case VIDIOC_ENUMSTD: |
12264 |
++ case VIDIOC_ENUMSTD32: |
12265 |
+ err = alloc_userspace(sizeof(struct v4l2_standard), 0, |
12266 |
+ &new_p64); |
12267 |
+ if (!err) |
12268 |
+@@ -1301,16 +1303,16 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12269 |
+ compatible_arg = 0; |
12270 |
+ break; |
12271 |
+ |
12272 |
+- case VIDIOC_ENUMINPUT: |
12273 |
++ case VIDIOC_ENUMINPUT32: |
12274 |
+ err = alloc_userspace(sizeof(struct v4l2_input), 0, &new_p64); |
12275 |
+ if (!err) |
12276 |
+ err = get_v4l2_input32(new_p64, p32); |
12277 |
+ compatible_arg = 0; |
12278 |
+ break; |
12279 |
+ |
12280 |
+- case VIDIOC_G_EXT_CTRLS: |
12281 |
+- case VIDIOC_S_EXT_CTRLS: |
12282 |
+- case VIDIOC_TRY_EXT_CTRLS: |
12283 |
++ case VIDIOC_G_EXT_CTRLS32: |
12284 |
++ case VIDIOC_S_EXT_CTRLS32: |
12285 |
++ case VIDIOC_TRY_EXT_CTRLS32: |
12286 |
+ err = bufsize_v4l2_ext_controls(p32, &aux_space); |
12287 |
+ if (!err) |
12288 |
+ err = alloc_userspace(sizeof(struct v4l2_ext_controls), |
12289 |
+@@ -1322,7 +1324,7 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12290 |
+ } |
12291 |
+ compatible_arg = 0; |
12292 |
+ break; |
12293 |
+- case VIDIOC_DQEVENT: |
12294 |
++ case VIDIOC_DQEVENT32: |
12295 |
+ err = alloc_userspace(sizeof(struct v4l2_event), 0, &new_p64); |
12296 |
+ compatible_arg = 0; |
12297 |
+ break; |
12298 |
+@@ -1340,9 +1342,9 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12299 |
+ * Otherwise, it will pass the newly allocated @new_p64 argument. |
12300 |
+ */ |
12301 |
+ if (compatible_arg) |
12302 |
+- err = native_ioctl(file, cmd, (unsigned long)p32); |
12303 |
++ err = native_ioctl(file, ncmd, (unsigned long)p32); |
12304 |
+ else |
12305 |
+- err = native_ioctl(file, cmd, (unsigned long)new_p64); |
12306 |
++ err = native_ioctl(file, ncmd, (unsigned long)new_p64); |
12307 |
+ |
12308 |
+ if (err == -ENOTTY) |
12309 |
+ return err; |
12310 |
+@@ -1358,13 +1360,13 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12311 |
+ * the blocks to maximum allowed value. |
12312 |
+ */ |
12313 |
+ switch (cmd) { |
12314 |
+- case VIDIOC_G_EXT_CTRLS: |
12315 |
+- case VIDIOC_S_EXT_CTRLS: |
12316 |
+- case VIDIOC_TRY_EXT_CTRLS: |
12317 |
++ case VIDIOC_G_EXT_CTRLS32: |
12318 |
++ case VIDIOC_S_EXT_CTRLS32: |
12319 |
++ case VIDIOC_TRY_EXT_CTRLS32: |
12320 |
+ if (put_v4l2_ext_controls32(file, new_p64, p32)) |
12321 |
+ err = -EFAULT; |
12322 |
+ break; |
12323 |
+- case VIDIOC_S_EDID: |
12324 |
++ case VIDIOC_S_EDID32: |
12325 |
+ if (put_v4l2_edid32(new_p64, p32)) |
12326 |
+ err = -EFAULT; |
12327 |
+ break; |
12328 |
+@@ -1377,49 +1379,49 @@ static long do_video_ioctl(struct file *file, unsigned int cmd, unsigned long ar |
12329 |
+ * the original 32 bits structure. |
12330 |
+ */ |
12331 |
+ switch (cmd) { |
12332 |
+- case VIDIOC_S_INPUT: |
12333 |
+- case VIDIOC_S_OUTPUT: |
12334 |
+- case VIDIOC_G_INPUT: |
12335 |
+- case VIDIOC_G_OUTPUT: |
12336 |
++ case VIDIOC_S_INPUT32: |
12337 |
++ case VIDIOC_S_OUTPUT32: |
12338 |
++ case VIDIOC_G_INPUT32: |
12339 |
++ case VIDIOC_G_OUTPUT32: |
12340 |
+ if (assign_in_user((compat_uint_t __user *)p32, |
12341 |
+ ((unsigned int __user *)new_p64))) |
12342 |
+ err = -EFAULT; |
12343 |
+ break; |
12344 |
+ |
12345 |
+- case VIDIOC_G_FBUF: |
12346 |
++ case VIDIOC_G_FBUF32: |
12347 |
+ err = put_v4l2_framebuffer32(new_p64, p32); |
12348 |
+ break; |
12349 |
+ |
12350 |
+- case VIDIOC_DQEVENT: |
12351 |
++ case VIDIOC_DQEVENT32: |
12352 |
+ err = put_v4l2_event32(new_p64, p32); |
12353 |
+ break; |
12354 |
+ |
12355 |
+- case VIDIOC_G_EDID: |
12356 |
++ case VIDIOC_G_EDID32: |
12357 |
+ err = put_v4l2_edid32(new_p64, p32); |
12358 |
+ break; |
12359 |
+ |
12360 |
+- case VIDIOC_G_FMT: |
12361 |
+- case VIDIOC_S_FMT: |
12362 |
+- case VIDIOC_TRY_FMT: |
12363 |
++ case VIDIOC_G_FMT32: |
12364 |
++ case VIDIOC_S_FMT32: |
12365 |
++ case VIDIOC_TRY_FMT32: |
12366 |
+ err = put_v4l2_format32(new_p64, p32); |
12367 |
+ break; |
12368 |
+ |
12369 |
+- case VIDIOC_CREATE_BUFS: |
12370 |
++ case VIDIOC_CREATE_BUFS32: |
12371 |
+ err = put_v4l2_create32(new_p64, p32); |
12372 |
+ break; |
12373 |
+ |
12374 |
+- case VIDIOC_PREPARE_BUF: |
12375 |
+- case VIDIOC_QUERYBUF: |
12376 |
+- case VIDIOC_QBUF: |
12377 |
+- case VIDIOC_DQBUF: |
12378 |
++ case VIDIOC_PREPARE_BUF32: |
12379 |
++ case VIDIOC_QUERYBUF32: |
12380 |
++ case VIDIOC_QBUF32: |
12381 |
++ case VIDIOC_DQBUF32: |
12382 |
+ err = put_v4l2_buffer32(new_p64, p32); |
12383 |
+ break; |
12384 |
+ |
12385 |
+- case VIDIOC_ENUMSTD: |
12386 |
++ case VIDIOC_ENUMSTD32: |
12387 |
+ err = put_v4l2_standard32(new_p64, p32); |
12388 |
+ break; |
12389 |
+ |
12390 |
+- case VIDIOC_ENUMINPUT: |
12391 |
++ case VIDIOC_ENUMINPUT32: |
12392 |
+ err = put_v4l2_input32(new_p64, p32); |
12393 |
+ break; |
12394 |
+ } |
12395 |
+diff --git a/drivers/media/v4l2-core/videobuf-dma-sg.c b/drivers/media/v4l2-core/videobuf-dma-sg.c |
12396 |
+index 08929c087e27..4c396e0defa4 100644 |
12397 |
+--- a/drivers/media/v4l2-core/videobuf-dma-sg.c |
12398 |
++++ b/drivers/media/v4l2-core/videobuf-dma-sg.c |
12399 |
+@@ -352,8 +352,11 @@ int videobuf_dma_free(struct videobuf_dmabuf *dma) |
12400 |
+ BUG_ON(dma->sglen); |
12401 |
+ |
12402 |
+ if (dma->pages) { |
12403 |
+- for (i = 0; i < dma->nr_pages; i++) |
12404 |
++ for (i = 0; i < dma->nr_pages; i++) { |
12405 |
++ if (dma->direction == DMA_FROM_DEVICE) |
12406 |
++ set_page_dirty_lock(dma->pages[i]); |
12407 |
+ put_page(dma->pages[i]); |
12408 |
++ } |
12409 |
+ kfree(dma->pages); |
12410 |
+ dma->pages = NULL; |
12411 |
+ } |
12412 |
+diff --git a/drivers/mfd/axp20x.c b/drivers/mfd/axp20x.c |
12413 |
+index f8e0fa97bb31..aa65931142ba 100644 |
12414 |
+--- a/drivers/mfd/axp20x.c |
12415 |
++++ b/drivers/mfd/axp20x.c |
12416 |
+@@ -128,7 +128,7 @@ static const struct regmap_range axp288_writeable_ranges[] = { |
12417 |
+ static const struct regmap_range axp288_volatile_ranges[] = { |
12418 |
+ regmap_reg_range(AXP20X_PWR_INPUT_STATUS, AXP288_POWER_REASON), |
12419 |
+ regmap_reg_range(AXP288_BC_GLOBAL, AXP288_BC_GLOBAL), |
12420 |
+- regmap_reg_range(AXP288_BC_DET_STAT, AXP288_BC_DET_STAT), |
12421 |
++ regmap_reg_range(AXP288_BC_DET_STAT, AXP20X_VBUS_IPSOUT_MGMT), |
12422 |
+ regmap_reg_range(AXP20X_CHRG_BAK_CTRL, AXP20X_CHRG_BAK_CTRL), |
12423 |
+ regmap_reg_range(AXP20X_IRQ1_EN, AXP20X_IPSOUT_V_HIGH_L), |
12424 |
+ regmap_reg_range(AXP20X_TIMER_CTRL, AXP20X_TIMER_CTRL), |
12425 |
+diff --git a/drivers/mfd/da9062-core.c b/drivers/mfd/da9062-core.c |
12426 |
+index 9f6105906c09..83c624f6033c 100644 |
12427 |
+--- a/drivers/mfd/da9062-core.c |
12428 |
++++ b/drivers/mfd/da9062-core.c |
12429 |
+@@ -257,7 +257,7 @@ static const struct mfd_cell da9062_devs[] = { |
12430 |
+ .name = "da9062-watchdog", |
12431 |
+ .num_resources = ARRAY_SIZE(da9062_wdt_resources), |
12432 |
+ .resources = da9062_wdt_resources, |
12433 |
+- .of_compatible = "dlg,da9062-wdt", |
12434 |
++ .of_compatible = "dlg,da9062-watchdog", |
12435 |
+ }, |
12436 |
+ { |
12437 |
+ .name = "da9062-thermal", |
12438 |
+diff --git a/drivers/mfd/dln2.c b/drivers/mfd/dln2.c |
12439 |
+index 90e789943466..1476465ce803 100644 |
12440 |
+--- a/drivers/mfd/dln2.c |
12441 |
++++ b/drivers/mfd/dln2.c |
12442 |
+@@ -725,6 +725,8 @@ static int dln2_probe(struct usb_interface *interface, |
12443 |
+ const struct usb_device_id *usb_id) |
12444 |
+ { |
12445 |
+ struct usb_host_interface *hostif = interface->cur_altsetting; |
12446 |
++ struct usb_endpoint_descriptor *epin; |
12447 |
++ struct usb_endpoint_descriptor *epout; |
12448 |
+ struct device *dev = &interface->dev; |
12449 |
+ struct dln2_dev *dln2; |
12450 |
+ int ret; |
12451 |
+@@ -734,12 +736,19 @@ static int dln2_probe(struct usb_interface *interface, |
12452 |
+ hostif->desc.bNumEndpoints < 2) |
12453 |
+ return -ENODEV; |
12454 |
+ |
12455 |
++ epin = &hostif->endpoint[0].desc; |
12456 |
++ epout = &hostif->endpoint[1].desc; |
12457 |
++ if (!usb_endpoint_is_bulk_out(epout)) |
12458 |
++ return -ENODEV; |
12459 |
++ if (!usb_endpoint_is_bulk_in(epin)) |
12460 |
++ return -ENODEV; |
12461 |
++ |
12462 |
+ dln2 = kzalloc(sizeof(*dln2), GFP_KERNEL); |
12463 |
+ if (!dln2) |
12464 |
+ return -ENOMEM; |
12465 |
+ |
12466 |
+- dln2->ep_out = hostif->endpoint[0].desc.bEndpointAddress; |
12467 |
+- dln2->ep_in = hostif->endpoint[1].desc.bEndpointAddress; |
12468 |
++ dln2->ep_out = epout->bEndpointAddress; |
12469 |
++ dln2->ep_in = epin->bEndpointAddress; |
12470 |
+ dln2->usb_dev = usb_get_dev(interface_to_usbdev(interface)); |
12471 |
+ dln2->interface = interface; |
12472 |
+ usb_set_intfdata(interface, dln2); |
12473 |
+diff --git a/drivers/mfd/rn5t618.c b/drivers/mfd/rn5t618.c |
12474 |
+index f4037d42a60f..dd4251f105e0 100644 |
12475 |
+--- a/drivers/mfd/rn5t618.c |
12476 |
++++ b/drivers/mfd/rn5t618.c |
12477 |
+@@ -32,6 +32,7 @@ static bool rn5t618_volatile_reg(struct device *dev, unsigned int reg) |
12478 |
+ case RN5T618_WATCHDOGCNT: |
12479 |
+ case RN5T618_DCIRQ: |
12480 |
+ case RN5T618_ILIMDATAH ... RN5T618_AIN0DATAL: |
12481 |
++ case RN5T618_ADCCNT3: |
12482 |
+ case RN5T618_IR_ADC1 ... RN5T618_IR_ADC3: |
12483 |
+ case RN5T618_IR_GPR: |
12484 |
+ case RN5T618_IR_GPF: |
12485 |
+diff --git a/drivers/mmc/host/mmc_spi.c b/drivers/mmc/host/mmc_spi.c |
12486 |
+index ea254d00541f..24795454d106 100644 |
12487 |
+--- a/drivers/mmc/host/mmc_spi.c |
12488 |
++++ b/drivers/mmc/host/mmc_spi.c |
12489 |
+@@ -1154,17 +1154,22 @@ static void mmc_spi_initsequence(struct mmc_spi_host *host) |
12490 |
+ * SPI protocol. Another is that when chipselect is released while |
12491 |
+ * the card returns BUSY status, the clock must issue several cycles |
12492 |
+ * with chipselect high before the card will stop driving its output. |
12493 |
++ * |
12494 |
++ * SPI_CS_HIGH means "asserted" here. In some cases like when using |
12495 |
++ * GPIOs for chip select, SPI_CS_HIGH is set but this will be logically |
12496 |
++ * inverted by gpiolib, so if we want to ascertain to drive it high |
12497 |
++ * we should toggle the default with an XOR as we do here. |
12498 |
+ */ |
12499 |
+- host->spi->mode |= SPI_CS_HIGH; |
12500 |
++ host->spi->mode ^= SPI_CS_HIGH; |
12501 |
+ if (spi_setup(host->spi) != 0) { |
12502 |
+ /* Just warn; most cards work without it. */ |
12503 |
+ dev_warn(&host->spi->dev, |
12504 |
+ "can't change chip-select polarity\n"); |
12505 |
+- host->spi->mode &= ~SPI_CS_HIGH; |
12506 |
++ host->spi->mode ^= SPI_CS_HIGH; |
12507 |
+ } else { |
12508 |
+ mmc_spi_readbytes(host, 18); |
12509 |
+ |
12510 |
+- host->spi->mode &= ~SPI_CS_HIGH; |
12511 |
++ host->spi->mode ^= SPI_CS_HIGH; |
12512 |
+ if (spi_setup(host->spi) != 0) { |
12513 |
+ /* Wot, we can't get the same setup we had before? */ |
12514 |
+ dev_err(&host->spi->dev, |
12515 |
+diff --git a/drivers/mmc/host/sdhci-of-at91.c b/drivers/mmc/host/sdhci-of-at91.c |
12516 |
+index 1ebcf0eb781e..04e88d4796fa 100644 |
12517 |
+--- a/drivers/mmc/host/sdhci-of-at91.c |
12518 |
++++ b/drivers/mmc/host/sdhci-of-at91.c |
12519 |
+@@ -332,19 +332,22 @@ static int sdhci_at91_probe(struct platform_device *pdev) |
12520 |
+ priv->mainck = devm_clk_get(&pdev->dev, "baseclk"); |
12521 |
+ if (IS_ERR(priv->mainck)) { |
12522 |
+ dev_err(&pdev->dev, "failed to get baseclk\n"); |
12523 |
+- return PTR_ERR(priv->mainck); |
12524 |
++ ret = PTR_ERR(priv->mainck); |
12525 |
++ goto sdhci_pltfm_free; |
12526 |
+ } |
12527 |
+ |
12528 |
+ priv->hclock = devm_clk_get(&pdev->dev, "hclock"); |
12529 |
+ if (IS_ERR(priv->hclock)) { |
12530 |
+ dev_err(&pdev->dev, "failed to get hclock\n"); |
12531 |
+- return PTR_ERR(priv->hclock); |
12532 |
++ ret = PTR_ERR(priv->hclock); |
12533 |
++ goto sdhci_pltfm_free; |
12534 |
+ } |
12535 |
+ |
12536 |
+ priv->gck = devm_clk_get(&pdev->dev, "multclk"); |
12537 |
+ if (IS_ERR(priv->gck)) { |
12538 |
+ dev_err(&pdev->dev, "failed to get multclk\n"); |
12539 |
+- return PTR_ERR(priv->gck); |
12540 |
++ ret = PTR_ERR(priv->gck); |
12541 |
++ goto sdhci_pltfm_free; |
12542 |
+ } |
12543 |
+ |
12544 |
+ ret = sdhci_at91_set_clks_presets(&pdev->dev); |
12545 |
+diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c |
12546 |
+index 462526a10537..8e292992f84c 100644 |
12547 |
+--- a/drivers/mtd/ubi/fastmap.c |
12548 |
++++ b/drivers/mtd/ubi/fastmap.c |
12549 |
+@@ -73,7 +73,7 @@ static int self_check_seen(struct ubi_device *ubi, unsigned long *seen) |
12550 |
+ return 0; |
12551 |
+ |
12552 |
+ for (pnum = 0; pnum < ubi->peb_count; pnum++) { |
12553 |
+- if (test_bit(pnum, seen) && ubi->lookuptbl[pnum]) { |
12554 |
++ if (!test_bit(pnum, seen) && ubi->lookuptbl[pnum]) { |
12555 |
+ ubi_err(ubi, "self-check failed for PEB %d, fastmap didn't see it", pnum); |
12556 |
+ ret = -EINVAL; |
12557 |
+ } |
12558 |
+@@ -1146,7 +1146,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12559 |
+ struct rb_node *tmp_rb; |
12560 |
+ int ret, i, j, free_peb_count, used_peb_count, vol_count; |
12561 |
+ int scrub_peb_count, erase_peb_count; |
12562 |
+- unsigned long *seen_pebs = NULL; |
12563 |
++ unsigned long *seen_pebs; |
12564 |
+ |
12565 |
+ fm_raw = ubi->fm_buf; |
12566 |
+ memset(ubi->fm_buf, 0, ubi->fm_size); |
12567 |
+@@ -1160,7 +1160,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12568 |
+ dvbuf = new_fm_vbuf(ubi, UBI_FM_DATA_VOLUME_ID); |
12569 |
+ if (!dvbuf) { |
12570 |
+ ret = -ENOMEM; |
12571 |
+- goto out_kfree; |
12572 |
++ goto out_free_avbuf; |
12573 |
+ } |
12574 |
+ |
12575 |
+ avhdr = ubi_get_vid_hdr(avbuf); |
12576 |
+@@ -1169,7 +1169,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12577 |
+ seen_pebs = init_seen(ubi); |
12578 |
+ if (IS_ERR(seen_pebs)) { |
12579 |
+ ret = PTR_ERR(seen_pebs); |
12580 |
+- goto out_kfree; |
12581 |
++ goto out_free_dvbuf; |
12582 |
+ } |
12583 |
+ |
12584 |
+ spin_lock(&ubi->volumes_lock); |
12585 |
+@@ -1337,7 +1337,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12586 |
+ ret = ubi_io_write_vid_hdr(ubi, new_fm->e[0]->pnum, avbuf); |
12587 |
+ if (ret) { |
12588 |
+ ubi_err(ubi, "unable to write vid_hdr to fastmap SB!"); |
12589 |
+- goto out_kfree; |
12590 |
++ goto out_free_seen; |
12591 |
+ } |
12592 |
+ |
12593 |
+ for (i = 0; i < new_fm->used_blocks; i++) { |
12594 |
+@@ -1359,7 +1359,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12595 |
+ if (ret) { |
12596 |
+ ubi_err(ubi, "unable to write vid_hdr to PEB %i!", |
12597 |
+ new_fm->e[i]->pnum); |
12598 |
+- goto out_kfree; |
12599 |
++ goto out_free_seen; |
12600 |
+ } |
12601 |
+ } |
12602 |
+ |
12603 |
+@@ -1369,7 +1369,7 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12604 |
+ if (ret) { |
12605 |
+ ubi_err(ubi, "unable to write fastmap to PEB %i!", |
12606 |
+ new_fm->e[i]->pnum); |
12607 |
+- goto out_kfree; |
12608 |
++ goto out_free_seen; |
12609 |
+ } |
12610 |
+ } |
12611 |
+ |
12612 |
+@@ -1379,10 +1379,13 @@ static int ubi_write_fastmap(struct ubi_device *ubi, |
12613 |
+ ret = self_check_seen(ubi, seen_pebs); |
12614 |
+ dbg_bld("fastmap written!"); |
12615 |
+ |
12616 |
+-out_kfree: |
12617 |
+- ubi_free_vid_buf(avbuf); |
12618 |
+- ubi_free_vid_buf(dvbuf); |
12619 |
++out_free_seen: |
12620 |
+ free_seen(seen_pebs); |
12621 |
++out_free_dvbuf: |
12622 |
++ ubi_free_vid_buf(dvbuf); |
12623 |
++out_free_avbuf: |
12624 |
++ ubi_free_vid_buf(avbuf); |
12625 |
++ |
12626 |
+ out: |
12627 |
+ return ret; |
12628 |
+ } |
12629 |
+diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c |
12630 |
+index e82108c917a6..334e3f22d4f1 100644 |
12631 |
+--- a/drivers/net/bonding/bond_alb.c |
12632 |
++++ b/drivers/net/bonding/bond_alb.c |
12633 |
+@@ -1399,26 +1399,31 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) |
12634 |
+ bool do_tx_balance = true; |
12635 |
+ u32 hash_index = 0; |
12636 |
+ const u8 *hash_start = NULL; |
12637 |
+- struct ipv6hdr *ip6hdr; |
12638 |
+ |
12639 |
+ skb_reset_mac_header(skb); |
12640 |
+ eth_data = eth_hdr(skb); |
12641 |
+ |
12642 |
+ switch (ntohs(skb->protocol)) { |
12643 |
+ case ETH_P_IP: { |
12644 |
+- const struct iphdr *iph = ip_hdr(skb); |
12645 |
++ const struct iphdr *iph; |
12646 |
+ |
12647 |
+ if (is_broadcast_ether_addr(eth_data->h_dest) || |
12648 |
+- iph->daddr == ip_bcast || |
12649 |
+- iph->protocol == IPPROTO_IGMP) { |
12650 |
++ !pskb_network_may_pull(skb, sizeof(*iph))) { |
12651 |
++ do_tx_balance = false; |
12652 |
++ break; |
12653 |
++ } |
12654 |
++ iph = ip_hdr(skb); |
12655 |
++ if (iph->daddr == ip_bcast || iph->protocol == IPPROTO_IGMP) { |
12656 |
+ do_tx_balance = false; |
12657 |
+ break; |
12658 |
+ } |
12659 |
+ hash_start = (char *)&(iph->daddr); |
12660 |
+ hash_size = sizeof(iph->daddr); |
12661 |
+- } |
12662 |
+ break; |
12663 |
+- case ETH_P_IPV6: |
12664 |
++ } |
12665 |
++ case ETH_P_IPV6: { |
12666 |
++ const struct ipv6hdr *ip6hdr; |
12667 |
++ |
12668 |
+ /* IPv6 doesn't really use broadcast mac address, but leave |
12669 |
+ * that here just in case. |
12670 |
+ */ |
12671 |
+@@ -1435,7 +1440,11 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) |
12672 |
+ break; |
12673 |
+ } |
12674 |
+ |
12675 |
+- /* Additianally, DAD probes should not be tx-balanced as that |
12676 |
++ if (!pskb_network_may_pull(skb, sizeof(*ip6hdr))) { |
12677 |
++ do_tx_balance = false; |
12678 |
++ break; |
12679 |
++ } |
12680 |
++ /* Additionally, DAD probes should not be tx-balanced as that |
12681 |
+ * will lead to false positives for duplicate addresses and |
12682 |
+ * prevent address configuration from working. |
12683 |
+ */ |
12684 |
+@@ -1445,17 +1454,26 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) |
12685 |
+ break; |
12686 |
+ } |
12687 |
+ |
12688 |
+- hash_start = (char *)&(ipv6_hdr(skb)->daddr); |
12689 |
+- hash_size = sizeof(ipv6_hdr(skb)->daddr); |
12690 |
++ hash_start = (char *)&ip6hdr->daddr; |
12691 |
++ hash_size = sizeof(ip6hdr->daddr); |
12692 |
+ break; |
12693 |
+- case ETH_P_IPX: |
12694 |
+- if (ipx_hdr(skb)->ipx_checksum != IPX_NO_CHECKSUM) { |
12695 |
++ } |
12696 |
++ case ETH_P_IPX: { |
12697 |
++ const struct ipxhdr *ipxhdr; |
12698 |
++ |
12699 |
++ if (pskb_network_may_pull(skb, sizeof(*ipxhdr))) { |
12700 |
++ do_tx_balance = false; |
12701 |
++ break; |
12702 |
++ } |
12703 |
++ ipxhdr = (struct ipxhdr *)skb_network_header(skb); |
12704 |
++ |
12705 |
++ if (ipxhdr->ipx_checksum != IPX_NO_CHECKSUM) { |
12706 |
+ /* something is wrong with this packet */ |
12707 |
+ do_tx_balance = false; |
12708 |
+ break; |
12709 |
+ } |
12710 |
+ |
12711 |
+- if (ipx_hdr(skb)->ipx_type != IPX_TYPE_NCP) { |
12712 |
++ if (ipxhdr->ipx_type != IPX_TYPE_NCP) { |
12713 |
+ /* The only protocol worth balancing in |
12714 |
+ * this family since it has an "ARP" like |
12715 |
+ * mechanism |
12716 |
+@@ -1464,9 +1482,11 @@ netdev_tx_t bond_alb_xmit(struct sk_buff *skb, struct net_device *bond_dev) |
12717 |
+ break; |
12718 |
+ } |
12719 |
+ |
12720 |
++ eth_data = eth_hdr(skb); |
12721 |
+ hash_start = (char *)eth_data->h_dest; |
12722 |
+ hash_size = ETH_ALEN; |
12723 |
+ break; |
12724 |
++ } |
12725 |
+ case ETH_P_ARP: |
12726 |
+ do_tx_balance = false; |
12727 |
+ if (bond_info->rlb_enabled) |
12728 |
+diff --git a/drivers/net/dsa/b53/b53_common.c b/drivers/net/dsa/b53/b53_common.c |
12729 |
+index 9f21e710fc38..51436e7eae10 100644 |
12730 |
+--- a/drivers/net/dsa/b53/b53_common.c |
12731 |
++++ b/drivers/net/dsa/b53/b53_common.c |
12732 |
+@@ -655,7 +655,7 @@ int b53_configure_vlan(struct dsa_switch *ds) |
12733 |
+ b53_do_vlan_op(dev, VTA_CMD_CLEAR); |
12734 |
+ } |
12735 |
+ |
12736 |
+- b53_enable_vlan(dev, false, dev->vlan_filtering_enabled); |
12737 |
++ b53_enable_vlan(dev, dev->vlan_enabled, dev->vlan_filtering_enabled); |
12738 |
+ |
12739 |
+ b53_for_each_port(dev, i) |
12740 |
+ b53_write16(dev, B53_VLAN_PAGE, |
12741 |
+diff --git a/drivers/net/dsa/bcm_sf2.c b/drivers/net/dsa/bcm_sf2.c |
12742 |
+index c93609007670..f181a28cb452 100644 |
12743 |
+--- a/drivers/net/dsa/bcm_sf2.c |
12744 |
++++ b/drivers/net/dsa/bcm_sf2.c |
12745 |
+@@ -72,7 +72,9 @@ static void bcm_sf2_imp_setup(struct dsa_switch *ds, int port) |
12746 |
+ |
12747 |
+ /* Force link status for IMP port */ |
12748 |
+ reg = core_readl(priv, offset); |
12749 |
+- reg |= (MII_SW_OR | LINK_STS | GMII_SPEED_UP_2G); |
12750 |
++ reg |= (MII_SW_OR | LINK_STS); |
12751 |
++ if (priv->type == BCM7278_DEVICE_ID) |
12752 |
++ reg |= GMII_SPEED_UP_2G; |
12753 |
+ core_writel(priv, reg, offset); |
12754 |
+ |
12755 |
+ /* Enable Broadcast, Multicast, Unicast forwarding to IMP port */ |
12756 |
+diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c b/drivers/net/ethernet/broadcom/bcmsysport.c |
12757 |
+index 49aa3b5ea57c..6f8649376ff0 100644 |
12758 |
+--- a/drivers/net/ethernet/broadcom/bcmsysport.c |
12759 |
++++ b/drivers/net/ethernet/broadcom/bcmsysport.c |
12760 |
+@@ -2716,6 +2716,9 @@ static int __maybe_unused bcm_sysport_resume(struct device *d) |
12761 |
+ |
12762 |
+ umac_reset(priv); |
12763 |
+ |
12764 |
++ /* Disable the UniMAC RX/TX */ |
12765 |
++ umac_enable_set(priv, CMD_RX_EN | CMD_TX_EN, 0); |
12766 |
++ |
12767 |
+ /* We may have been suspended and never received a WOL event that |
12768 |
+ * would turn off MPD detection, take care of that now |
12769 |
+ */ |
12770 |
+diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c |
12771 |
+index 5cf85a89016e..c19d0eabeb52 100644 |
12772 |
+--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c |
12773 |
++++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c |
12774 |
+@@ -5861,7 +5861,7 @@ static void bnxt_setup_msix(struct bnxt *bp) |
12775 |
+ int tcs, i; |
12776 |
+ |
12777 |
+ tcs = netdev_get_num_tc(dev); |
12778 |
+- if (tcs > 1) { |
12779 |
++ if (tcs) { |
12780 |
+ int i, off, count; |
12781 |
+ |
12782 |
+ for (i = 0; i < tcs; i++) { |
12783 |
+diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c |
12784 |
+index c2eb18854794..d1ff317f3b18 100644 |
12785 |
+--- a/drivers/net/ethernet/cadence/macb_main.c |
12786 |
++++ b/drivers/net/ethernet/cadence/macb_main.c |
12787 |
+@@ -66,7 +66,11 @@ |
12788 |
+ /* Max length of transmit frame must be a multiple of 8 bytes */ |
12789 |
+ #define MACB_TX_LEN_ALIGN 8 |
12790 |
+ #define MACB_MAX_TX_LEN ((unsigned int)((1 << MACB_TX_FRMLEN_SIZE) - 1) & ~((unsigned int)(MACB_TX_LEN_ALIGN - 1))) |
12791 |
+-#define GEM_MAX_TX_LEN ((unsigned int)((1 << GEM_TX_FRMLEN_SIZE) - 1) & ~((unsigned int)(MACB_TX_LEN_ALIGN - 1))) |
12792 |
++/* Limit maximum TX length as per Cadence TSO errata. This is to avoid a |
12793 |
++ * false amba_error in TX path from the DMA assuming there is not enough |
12794 |
++ * space in the SRAM (16KB) even when there is. |
12795 |
++ */ |
12796 |
++#define GEM_MAX_TX_LEN (unsigned int)(0x3FC0) |
12797 |
+ |
12798 |
+ #define GEM_MTU_MIN_SIZE ETH_MIN_MTU |
12799 |
+ #define MACB_NETIF_LSO NETIF_F_TSO |
12800 |
+@@ -1654,16 +1658,14 @@ static netdev_features_t macb_features_check(struct sk_buff *skb, |
12801 |
+ |
12802 |
+ /* Validate LSO compatibility */ |
12803 |
+ |
12804 |
+- /* there is only one buffer */ |
12805 |
+- if (!skb_is_nonlinear(skb)) |
12806 |
++ /* there is only one buffer or protocol is not UDP */ |
12807 |
++ if (!skb_is_nonlinear(skb) || (ip_hdr(skb)->protocol != IPPROTO_UDP)) |
12808 |
+ return features; |
12809 |
+ |
12810 |
+ /* length of header */ |
12811 |
+ hdrlen = skb_transport_offset(skb); |
12812 |
+- if (ip_hdr(skb)->protocol == IPPROTO_TCP) |
12813 |
+- hdrlen += tcp_hdrlen(skb); |
12814 |
+ |
12815 |
+- /* For LSO: |
12816 |
++ /* For UFO only: |
12817 |
+ * When software supplies two or more payload buffers all payload buffers |
12818 |
+ * apart from the last must be a multiple of 8 bytes in size. |
12819 |
+ */ |
12820 |
+diff --git a/drivers/net/ethernet/dec/tulip/dmfe.c b/drivers/net/ethernet/dec/tulip/dmfe.c |
12821 |
+index 17ef7a28873d..0defd5b1212a 100644 |
12822 |
+--- a/drivers/net/ethernet/dec/tulip/dmfe.c |
12823 |
++++ b/drivers/net/ethernet/dec/tulip/dmfe.c |
12824 |
+@@ -2222,15 +2222,16 @@ static int __init dmfe_init_module(void) |
12825 |
+ if (cr6set) |
12826 |
+ dmfe_cr6_user_set = cr6set; |
12827 |
+ |
12828 |
+- switch(mode) { |
12829 |
+- case DMFE_10MHF: |
12830 |
++ switch (mode) { |
12831 |
++ case DMFE_10MHF: |
12832 |
+ case DMFE_100MHF: |
12833 |
+ case DMFE_10MFD: |
12834 |
+ case DMFE_100MFD: |
12835 |
+ case DMFE_1M_HPNA: |
12836 |
+ dmfe_media_mode = mode; |
12837 |
+ break; |
12838 |
+- default:dmfe_media_mode = DMFE_AUTO; |
12839 |
++ default: |
12840 |
++ dmfe_media_mode = DMFE_AUTO; |
12841 |
+ break; |
12842 |
+ } |
12843 |
+ |
12844 |
+diff --git a/drivers/net/ethernet/dec/tulip/uli526x.c b/drivers/net/ethernet/dec/tulip/uli526x.c |
12845 |
+index 488a744084c9..f4751a8de629 100644 |
12846 |
+--- a/drivers/net/ethernet/dec/tulip/uli526x.c |
12847 |
++++ b/drivers/net/ethernet/dec/tulip/uli526x.c |
12848 |
+@@ -1817,8 +1817,8 @@ static int __init uli526x_init_module(void) |
12849 |
+ if (cr6set) |
12850 |
+ uli526x_cr6_user_set = cr6set; |
12851 |
+ |
12852 |
+- switch (mode) { |
12853 |
+- case ULI526X_10MHF: |
12854 |
++ switch (mode) { |
12855 |
++ case ULI526X_10MHF: |
12856 |
+ case ULI526X_100MHF: |
12857 |
+ case ULI526X_10MFD: |
12858 |
+ case ULI526X_100MFD: |
12859 |
+diff --git a/drivers/net/ethernet/marvell/mvneta.c b/drivers/net/ethernet/marvell/mvneta.c |
12860 |
+index 4313bbb2396f..51885e6dec50 100644 |
12861 |
+--- a/drivers/net/ethernet/marvell/mvneta.c |
12862 |
++++ b/drivers/net/ethernet/marvell/mvneta.c |
12863 |
+@@ -385,6 +385,8 @@ struct mvneta_pcpu_stats { |
12864 |
+ struct u64_stats_sync syncp; |
12865 |
+ u64 rx_packets; |
12866 |
+ u64 rx_bytes; |
12867 |
++ u64 rx_dropped; |
12868 |
++ u64 rx_errors; |
12869 |
+ u64 tx_packets; |
12870 |
+ u64 tx_bytes; |
12871 |
+ }; |
12872 |
+@@ -701,6 +703,8 @@ mvneta_get_stats64(struct net_device *dev, |
12873 |
+ struct mvneta_pcpu_stats *cpu_stats; |
12874 |
+ u64 rx_packets; |
12875 |
+ u64 rx_bytes; |
12876 |
++ u64 rx_dropped; |
12877 |
++ u64 rx_errors; |
12878 |
+ u64 tx_packets; |
12879 |
+ u64 tx_bytes; |
12880 |
+ |
12881 |
+@@ -709,19 +713,20 @@ mvneta_get_stats64(struct net_device *dev, |
12882 |
+ start = u64_stats_fetch_begin_irq(&cpu_stats->syncp); |
12883 |
+ rx_packets = cpu_stats->rx_packets; |
12884 |
+ rx_bytes = cpu_stats->rx_bytes; |
12885 |
++ rx_dropped = cpu_stats->rx_dropped; |
12886 |
++ rx_errors = cpu_stats->rx_errors; |
12887 |
+ tx_packets = cpu_stats->tx_packets; |
12888 |
+ tx_bytes = cpu_stats->tx_bytes; |
12889 |
+ } while (u64_stats_fetch_retry_irq(&cpu_stats->syncp, start)); |
12890 |
+ |
12891 |
+ stats->rx_packets += rx_packets; |
12892 |
+ stats->rx_bytes += rx_bytes; |
12893 |
++ stats->rx_dropped += rx_dropped; |
12894 |
++ stats->rx_errors += rx_errors; |
12895 |
+ stats->tx_packets += tx_packets; |
12896 |
+ stats->tx_bytes += tx_bytes; |
12897 |
+ } |
12898 |
+ |
12899 |
+- stats->rx_errors = dev->stats.rx_errors; |
12900 |
+- stats->rx_dropped = dev->stats.rx_dropped; |
12901 |
+- |
12902 |
+ stats->tx_dropped = dev->stats.tx_dropped; |
12903 |
+ } |
12904 |
+ |
12905 |
+@@ -1698,8 +1703,14 @@ static u32 mvneta_txq_desc_csum(int l3_offs, int l3_proto, |
12906 |
+ static void mvneta_rx_error(struct mvneta_port *pp, |
12907 |
+ struct mvneta_rx_desc *rx_desc) |
12908 |
+ { |
12909 |
++ struct mvneta_pcpu_stats *stats = this_cpu_ptr(pp->stats); |
12910 |
+ u32 status = rx_desc->status; |
12911 |
+ |
12912 |
++ /* update per-cpu counter */ |
12913 |
++ u64_stats_update_begin(&stats->syncp); |
12914 |
++ stats->rx_errors++; |
12915 |
++ u64_stats_update_end(&stats->syncp); |
12916 |
++ |
12917 |
+ switch (status & MVNETA_RXD_ERR_CODE_MASK) { |
12918 |
+ case MVNETA_RXD_ERR_CRC: |
12919 |
+ netdev_err(pp->dev, "bad rx status %08x (crc error), size=%d\n", |
12920 |
+@@ -1960,7 +1971,6 @@ static int mvneta_rx_swbm(struct napi_struct *napi, |
12921 |
+ /* Check errors only for FIRST descriptor */ |
12922 |
+ if (rx_status & MVNETA_RXD_ERR_SUMMARY) { |
12923 |
+ mvneta_rx_error(pp, rx_desc); |
12924 |
+- dev->stats.rx_errors++; |
12925 |
+ /* leave the descriptor untouched */ |
12926 |
+ continue; |
12927 |
+ } |
12928 |
+@@ -1971,11 +1981,17 @@ static int mvneta_rx_swbm(struct napi_struct *napi, |
12929 |
+ skb_size = max(rx_copybreak, rx_header_size); |
12930 |
+ rxq->skb = netdev_alloc_skb_ip_align(dev, skb_size); |
12931 |
+ if (unlikely(!rxq->skb)) { |
12932 |
++ struct mvneta_pcpu_stats *stats = this_cpu_ptr(pp->stats); |
12933 |
++ |
12934 |
+ netdev_err(dev, |
12935 |
+ "Can't allocate skb on queue %d\n", |
12936 |
+ rxq->id); |
12937 |
+- dev->stats.rx_dropped++; |
12938 |
++ |
12939 |
+ rxq->skb_alloc_err++; |
12940 |
++ |
12941 |
++ u64_stats_update_begin(&stats->syncp); |
12942 |
++ stats->rx_dropped++; |
12943 |
++ u64_stats_update_end(&stats->syncp); |
12944 |
+ continue; |
12945 |
+ } |
12946 |
+ copy_size = min(skb_size, rx_bytes); |
12947 |
+@@ -2135,7 +2151,6 @@ err_drop_frame_ret_pool: |
12948 |
+ mvneta_bm_pool_put_bp(pp->bm_priv, bm_pool, |
12949 |
+ rx_desc->buf_phys_addr); |
12950 |
+ err_drop_frame: |
12951 |
+- dev->stats.rx_errors++; |
12952 |
+ mvneta_rx_error(pp, rx_desc); |
12953 |
+ /* leave the descriptor untouched */ |
12954 |
+ continue; |
12955 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fpga/ipsec.c b/drivers/net/ethernet/mellanox/mlx5/core/fpga/ipsec.c |
12956 |
+index b5a8769a5bfd..715ccafc92cd 100644 |
12957 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/fpga/ipsec.c |
12958 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/fpga/ipsec.c |
12959 |
+@@ -848,6 +848,7 @@ void mlx5_fpga_ipsec_delete_sa_ctx(void *context) |
12960 |
+ mutex_lock(&fpga_xfrm->lock); |
12961 |
+ if (!--fpga_xfrm->num_rules) { |
12962 |
+ mlx5_fpga_ipsec_release_sa_ctx(fpga_xfrm->sa_ctx); |
12963 |
++ kfree(fpga_xfrm->sa_ctx); |
12964 |
+ fpga_xfrm->sa_ctx = NULL; |
12965 |
+ } |
12966 |
+ mutex_unlock(&fpga_xfrm->lock); |
12967 |
+@@ -1472,7 +1473,7 @@ int mlx5_fpga_esp_modify_xfrm(struct mlx5_accel_esp_xfrm *xfrm, |
12968 |
+ if (!memcmp(&xfrm->attrs, attrs, sizeof(xfrm->attrs))) |
12969 |
+ return 0; |
12970 |
+ |
12971 |
+- if (!mlx5_fpga_esp_validate_xfrm_attrs(mdev, attrs)) { |
12972 |
++ if (mlx5_fpga_esp_validate_xfrm_attrs(mdev, attrs)) { |
12973 |
+ mlx5_core_warn(mdev, "Tried to create an esp with unsupported attrs\n"); |
12974 |
+ return -EOPNOTSUPP; |
12975 |
+ } |
12976 |
+diff --git a/drivers/net/ethernet/smsc/smc911x.c b/drivers/net/ethernet/smsc/smc911x.c |
12977 |
+index 8355dfbb8ec3..f97b35430c84 100644 |
12978 |
+--- a/drivers/net/ethernet/smsc/smc911x.c |
12979 |
++++ b/drivers/net/ethernet/smsc/smc911x.c |
12980 |
+@@ -947,7 +947,7 @@ static void smc911x_phy_configure(struct work_struct *work) |
12981 |
+ if (lp->ctl_rspeed != 100) |
12982 |
+ my_ad_caps &= ~(ADVERTISE_100BASE4|ADVERTISE_100FULL|ADVERTISE_100HALF); |
12983 |
+ |
12984 |
+- if (!lp->ctl_rfduplx) |
12985 |
++ if (!lp->ctl_rfduplx) |
12986 |
+ my_ad_caps &= ~(ADVERTISE_100FULL|ADVERTISE_10FULL); |
12987 |
+ |
12988 |
+ /* Update our Auto-Neg Advertisement Register */ |
12989 |
+diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
12990 |
+index f069adfc2b35..9c7b1d8e8220 100644 |
12991 |
+--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
12992 |
++++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
12993 |
+@@ -4513,6 +4513,7 @@ int stmmac_suspend(struct device *dev) |
12994 |
+ { |
12995 |
+ struct net_device *ndev = dev_get_drvdata(dev); |
12996 |
+ struct stmmac_priv *priv = netdev_priv(ndev); |
12997 |
++ u32 chan; |
12998 |
+ |
12999 |
+ if (!ndev || !netif_running(ndev)) |
13000 |
+ return 0; |
13001 |
+@@ -4527,6 +4528,9 @@ int stmmac_suspend(struct device *dev) |
13002 |
+ |
13003 |
+ stmmac_disable_all_queues(priv); |
13004 |
+ |
13005 |
++ for (chan = 0; chan < priv->plat->tx_queues_to_use; chan++) |
13006 |
++ del_timer_sync(&priv->tx_queue[chan].txtimer); |
13007 |
++ |
13008 |
+ /* Stop TX/RX DMA */ |
13009 |
+ stmmac_stop_all_dma(priv); |
13010 |
+ |
13011 |
+diff --git a/drivers/net/gtp.c b/drivers/net/gtp.c |
13012 |
+index ee086441dcbe..eab9984f73a8 100644 |
13013 |
+--- a/drivers/net/gtp.c |
13014 |
++++ b/drivers/net/gtp.c |
13015 |
+@@ -772,12 +772,12 @@ static int gtp_hashtable_new(struct gtp_dev *gtp, int hsize) |
13016 |
+ int i; |
13017 |
+ |
13018 |
+ gtp->addr_hash = kmalloc_array(hsize, sizeof(struct hlist_head), |
13019 |
+- GFP_KERNEL); |
13020 |
++ GFP_KERNEL | __GFP_NOWARN); |
13021 |
+ if (gtp->addr_hash == NULL) |
13022 |
+ return -ENOMEM; |
13023 |
+ |
13024 |
+ gtp->tid_hash = kmalloc_array(hsize, sizeof(struct hlist_head), |
13025 |
+- GFP_KERNEL); |
13026 |
++ GFP_KERNEL | __GFP_NOWARN); |
13027 |
+ if (gtp->tid_hash == NULL) |
13028 |
+ goto err1; |
13029 |
+ |
13030 |
+diff --git a/drivers/net/ppp/ppp_async.c b/drivers/net/ppp/ppp_async.c |
13031 |
+index bdc4d23627c5..bf03db40d4f0 100644 |
13032 |
+--- a/drivers/net/ppp/ppp_async.c |
13033 |
++++ b/drivers/net/ppp/ppp_async.c |
13034 |
+@@ -878,15 +878,15 @@ ppp_async_input(struct asyncppp *ap, const unsigned char *buf, |
13035 |
+ skb = dev_alloc_skb(ap->mru + PPP_HDRLEN + 2); |
13036 |
+ if (!skb) |
13037 |
+ goto nomem; |
13038 |
+- ap->rpkt = skb; |
13039 |
+- } |
13040 |
+- if (skb->len == 0) { |
13041 |
+- /* Try to get the payload 4-byte aligned. |
13042 |
+- * This should match the |
13043 |
+- * PPP_ALLSTATIONS/PPP_UI/compressed tests in |
13044 |
+- * process_input_packet, but we do not have |
13045 |
+- * enough chars here to test buf[1] and buf[2]. |
13046 |
+- */ |
13047 |
++ ap->rpkt = skb; |
13048 |
++ } |
13049 |
++ if (skb->len == 0) { |
13050 |
++ /* Try to get the payload 4-byte aligned. |
13051 |
++ * This should match the |
13052 |
++ * PPP_ALLSTATIONS/PPP_UI/compressed tests in |
13053 |
++ * process_input_packet, but we do not have |
13054 |
++ * enough chars here to test buf[1] and buf[2]. |
13055 |
++ */ |
13056 |
+ if (buf[0] != PPP_ALLSTATIONS) |
13057 |
+ skb_reserve(skb, 2 + (buf[0] & 1)); |
13058 |
+ } |
13059 |
+diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c |
13060 |
+index 6a213fe760ff..41254f04ab15 100644 |
13061 |
+--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c |
13062 |
++++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c |
13063 |
+@@ -441,6 +441,7 @@ fail: |
13064 |
+ usb_free_urb(req->urb); |
13065 |
+ list_del(q->next); |
13066 |
+ } |
13067 |
++ kfree(reqs); |
13068 |
+ return NULL; |
13069 |
+ |
13070 |
+ } |
13071 |
+diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/nvm.c b/drivers/net/wireless/intel/iwlwifi/mvm/nvm.c |
13072 |
+index f2579c94ffdb..3270faafe0bc 100644 |
13073 |
+--- a/drivers/net/wireless/intel/iwlwifi/mvm/nvm.c |
13074 |
++++ b/drivers/net/wireless/intel/iwlwifi/mvm/nvm.c |
13075 |
+@@ -286,7 +286,7 @@ iwl_parse_nvm_sections(struct iwl_mvm *mvm) |
13076 |
+ int regulatory_type; |
13077 |
+ |
13078 |
+ /* Checking for required sections */ |
13079 |
+- if (mvm->trans->cfg->nvm_type != IWL_NVM_EXT) { |
13080 |
++ if (mvm->trans->cfg->nvm_type == IWL_NVM) { |
13081 |
+ if (!mvm->nvm_sections[NVM_SECTION_TYPE_SW].data || |
13082 |
+ !mvm->nvm_sections[mvm->cfg->nvm_hw_section_num].data) { |
13083 |
+ IWL_ERR(mvm, "Can't parse empty OTP/NVM sections\n"); |
13084 |
+diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/sta.c b/drivers/net/wireless/intel/iwlwifi/mvm/sta.c |
13085 |
+index 69057701641e..373ace38edab 100644 |
13086 |
+--- a/drivers/net/wireless/intel/iwlwifi/mvm/sta.c |
13087 |
++++ b/drivers/net/wireless/intel/iwlwifi/mvm/sta.c |
13088 |
+@@ -3045,6 +3045,10 @@ static int iwl_mvm_send_sta_igtk(struct iwl_mvm *mvm, |
13089 |
+ igtk_cmd.sta_id = cpu_to_le32(sta_id); |
13090 |
+ |
13091 |
+ if (remove_key) { |
13092 |
++ /* This is a valid situation for IGTK */ |
13093 |
++ if (sta_id == IWL_MVM_INVALID_STA) |
13094 |
++ return 0; |
13095 |
++ |
13096 |
+ igtk_cmd.ctrl_flags |= cpu_to_le32(STA_KEY_NOT_VALID); |
13097 |
+ } else { |
13098 |
+ struct ieee80211_key_seq seq; |
13099 |
+@@ -3352,9 +3356,9 @@ int iwl_mvm_remove_sta_key(struct iwl_mvm *mvm, |
13100 |
+ IWL_DEBUG_WEP(mvm, "mvm remove dynamic key: idx=%d sta=%d\n", |
13101 |
+ keyconf->keyidx, sta_id); |
13102 |
+ |
13103 |
+- if (mvm_sta && (keyconf->cipher == WLAN_CIPHER_SUITE_AES_CMAC || |
13104 |
+- keyconf->cipher == WLAN_CIPHER_SUITE_BIP_GMAC_128 || |
13105 |
+- keyconf->cipher == WLAN_CIPHER_SUITE_BIP_GMAC_256)) |
13106 |
++ if (keyconf->cipher == WLAN_CIPHER_SUITE_AES_CMAC || |
13107 |
++ keyconf->cipher == WLAN_CIPHER_SUITE_BIP_GMAC_128 || |
13108 |
++ keyconf->cipher == WLAN_CIPHER_SUITE_BIP_GMAC_256) |
13109 |
+ return iwl_mvm_send_sta_igtk(mvm, keyconf, sta_id, true); |
13110 |
+ |
13111 |
+ if (!__test_and_clear_bit(keyconf->hw_key_idx, mvm->fw_key_table)) { |
13112 |
+diff --git a/drivers/net/wireless/marvell/mwifiex/sta_ioctl.c b/drivers/net/wireless/marvell/mwifiex/sta_ioctl.c |
13113 |
+index 6dd835f1efc2..fbfa0b15d0c8 100644 |
13114 |
+--- a/drivers/net/wireless/marvell/mwifiex/sta_ioctl.c |
13115 |
++++ b/drivers/net/wireless/marvell/mwifiex/sta_ioctl.c |
13116 |
+@@ -232,6 +232,7 @@ static int mwifiex_process_country_ie(struct mwifiex_private *priv, |
13117 |
+ |
13118 |
+ if (country_ie_len > |
13119 |
+ (IEEE80211_COUNTRY_STRING_LEN + MWIFIEX_MAX_TRIPLET_802_11D)) { |
13120 |
++ rcu_read_unlock(); |
13121 |
+ mwifiex_dbg(priv->adapter, ERROR, |
13122 |
+ "11D: country_ie_len overflow!, deauth AP\n"); |
13123 |
+ return -EINVAL; |
13124 |
+diff --git a/drivers/nfc/pn544/pn544.c b/drivers/nfc/pn544/pn544.c |
13125 |
+index 70e898e38b16..f30bdf95610f 100644 |
13126 |
+--- a/drivers/nfc/pn544/pn544.c |
13127 |
++++ b/drivers/nfc/pn544/pn544.c |
13128 |
+@@ -704,7 +704,7 @@ static int pn544_hci_check_presence(struct nfc_hci_dev *hdev, |
13129 |
+ target->nfcid1_len != 10) |
13130 |
+ return -EOPNOTSUPP; |
13131 |
+ |
13132 |
+- return nfc_hci_send_cmd(hdev, NFC_HCI_RF_READER_A_GATE, |
13133 |
++ return nfc_hci_send_cmd(hdev, NFC_HCI_RF_READER_A_GATE, |
13134 |
+ PN544_RF_READER_CMD_ACTIVATE_NEXT, |
13135 |
+ target->nfcid1, target->nfcid1_len, NULL); |
13136 |
+ } else if (target->supported_protocols & (NFC_PROTO_JEWEL_MASK | |
13137 |
+diff --git a/drivers/of/Kconfig b/drivers/of/Kconfig |
13138 |
+index ad3fcad4d75b..5e1315900562 100644 |
13139 |
+--- a/drivers/of/Kconfig |
13140 |
++++ b/drivers/of/Kconfig |
13141 |
+@@ -103,4 +103,8 @@ config OF_OVERLAY |
13142 |
+ config OF_NUMA |
13143 |
+ bool |
13144 |
+ |
13145 |
++config OF_DMA_DEFAULT_COHERENT |
13146 |
++ # arches should select this if DMA is coherent by default for OF devices |
13147 |
++ bool |
13148 |
++ |
13149 |
+ endif # OF |
13150 |
+diff --git a/drivers/of/address.c b/drivers/of/address.c |
13151 |
+index 7ddbf0a1ab86..c42aebba35ab 100644 |
13152 |
+--- a/drivers/of/address.c |
13153 |
++++ b/drivers/of/address.c |
13154 |
+@@ -970,12 +970,16 @@ EXPORT_SYMBOL_GPL(of_dma_get_range); |
13155 |
+ * @np: device node |
13156 |
+ * |
13157 |
+ * It returns true if "dma-coherent" property was found |
13158 |
+- * for this device in DT. |
13159 |
++ * for this device in the DT, or if DMA is coherent by |
13160 |
++ * default for OF devices on the current platform. |
13161 |
+ */ |
13162 |
+ bool of_dma_is_coherent(struct device_node *np) |
13163 |
+ { |
13164 |
+ struct device_node *node = of_node_get(np); |
13165 |
+ |
13166 |
++ if (IS_ENABLED(CONFIG_OF_DMA_DEFAULT_COHERENT)) |
13167 |
++ return true; |
13168 |
++ |
13169 |
+ while (node) { |
13170 |
+ if (of_property_read_bool(node, "dma-coherent")) { |
13171 |
+ of_node_put(node); |
13172 |
+diff --git a/drivers/pci/controller/dwc/pci-keystone-dw.c b/drivers/pci/controller/dwc/pci-keystone-dw.c |
13173 |
+index 0682213328e9..15c612e853af 100644 |
13174 |
+--- a/drivers/pci/controller/dwc/pci-keystone-dw.c |
13175 |
++++ b/drivers/pci/controller/dwc/pci-keystone-dw.c |
13176 |
+@@ -425,7 +425,7 @@ void ks_dw_pcie_initiate_link_train(struct keystone_pcie *ks_pcie) |
13177 |
+ /* Disable Link training */ |
13178 |
+ val = ks_dw_app_readl(ks_pcie, CMD_STATUS); |
13179 |
+ val &= ~LTSSM_EN_VAL; |
13180 |
+- ks_dw_app_writel(ks_pcie, CMD_STATUS, LTSSM_EN_VAL | val); |
13181 |
++ ks_dw_app_writel(ks_pcie, CMD_STATUS, val); |
13182 |
+ |
13183 |
+ /* Initiate Link Training */ |
13184 |
+ val = ks_dw_app_readl(ks_pcie, CMD_STATUS); |
13185 |
+diff --git a/drivers/pci/controller/pci-tegra.c b/drivers/pci/controller/pci-tegra.c |
13186 |
+index 58e487352853..6f86583605a4 100644 |
13187 |
+--- a/drivers/pci/controller/pci-tegra.c |
13188 |
++++ b/drivers/pci/controller/pci-tegra.c |
13189 |
+@@ -2398,7 +2398,7 @@ static int tegra_pcie_probe(struct platform_device *pdev) |
13190 |
+ |
13191 |
+ pm_runtime_enable(pcie->dev); |
13192 |
+ err = pm_runtime_get_sync(pcie->dev); |
13193 |
+- if (err) { |
13194 |
++ if (err < 0) { |
13195 |
+ dev_err(dev, "fail to enable pcie controller: %d\n", err); |
13196 |
+ goto teardown_msi; |
13197 |
+ } |
13198 |
+diff --git a/drivers/phy/qualcomm/phy-qcom-apq8064-sata.c b/drivers/phy/qualcomm/phy-qcom-apq8064-sata.c |
13199 |
+index 69ce2afac015..c6925e3e878b 100644 |
13200 |
+--- a/drivers/phy/qualcomm/phy-qcom-apq8064-sata.c |
13201 |
++++ b/drivers/phy/qualcomm/phy-qcom-apq8064-sata.c |
13202 |
+@@ -88,7 +88,7 @@ static int read_poll_timeout(void __iomem *addr, u32 mask) |
13203 |
+ if (readl_relaxed(addr) & mask) |
13204 |
+ return 0; |
13205 |
+ |
13206 |
+- usleep_range(DELAY_INTERVAL_US, DELAY_INTERVAL_US + 50); |
13207 |
++ usleep_range(DELAY_INTERVAL_US, DELAY_INTERVAL_US + 50); |
13208 |
+ } while (!time_after(jiffies, timeout)); |
13209 |
+ |
13210 |
+ return (readl_relaxed(addr) & mask) ? 0 : -ETIMEDOUT; |
13211 |
+diff --git a/drivers/platform/x86/intel_scu_ipc.c b/drivers/platform/x86/intel_scu_ipc.c |
13212 |
+index 75c8fef7a482..54f131bec192 100644 |
13213 |
+--- a/drivers/platform/x86/intel_scu_ipc.c |
13214 |
++++ b/drivers/platform/x86/intel_scu_ipc.c |
13215 |
+@@ -69,26 +69,22 @@ |
13216 |
+ struct intel_scu_ipc_pdata_t { |
13217 |
+ u32 i2c_base; |
13218 |
+ u32 i2c_len; |
13219 |
+- u8 irq_mode; |
13220 |
+ }; |
13221 |
+ |
13222 |
+ static const struct intel_scu_ipc_pdata_t intel_scu_ipc_lincroft_pdata = { |
13223 |
+ .i2c_base = 0xff12b000, |
13224 |
+ .i2c_len = 0x10, |
13225 |
+- .irq_mode = 0, |
13226 |
+ }; |
13227 |
+ |
13228 |
+ /* Penwell and Cloverview */ |
13229 |
+ static const struct intel_scu_ipc_pdata_t intel_scu_ipc_penwell_pdata = { |
13230 |
+ .i2c_base = 0xff12b000, |
13231 |
+ .i2c_len = 0x10, |
13232 |
+- .irq_mode = 1, |
13233 |
+ }; |
13234 |
+ |
13235 |
+ static const struct intel_scu_ipc_pdata_t intel_scu_ipc_tangier_pdata = { |
13236 |
+ .i2c_base = 0xff00d000, |
13237 |
+ .i2c_len = 0x10, |
13238 |
+- .irq_mode = 0, |
13239 |
+ }; |
13240 |
+ |
13241 |
+ struct intel_scu_ipc_dev { |
13242 |
+@@ -101,6 +97,9 @@ struct intel_scu_ipc_dev { |
13243 |
+ |
13244 |
+ static struct intel_scu_ipc_dev ipcdev; /* Only one for now */ |
13245 |
+ |
13246 |
++#define IPC_STATUS 0x04 |
13247 |
++#define IPC_STATUS_IRQ BIT(2) |
13248 |
++ |
13249 |
+ /* |
13250 |
+ * IPC Read Buffer (Read Only): |
13251 |
+ * 16 byte buffer for receiving data from SCU, if IPC command |
13252 |
+@@ -122,11 +121,8 @@ static DEFINE_MUTEX(ipclock); /* lock used to prevent multiple call to SCU */ |
13253 |
+ */ |
13254 |
+ static inline void ipc_command(struct intel_scu_ipc_dev *scu, u32 cmd) |
13255 |
+ { |
13256 |
+- if (scu->irq_mode) { |
13257 |
+- reinit_completion(&scu->cmd_complete); |
13258 |
+- writel(cmd | IPC_IOC, scu->ipc_base); |
13259 |
+- } |
13260 |
+- writel(cmd, scu->ipc_base); |
13261 |
++ reinit_completion(&scu->cmd_complete); |
13262 |
++ writel(cmd | IPC_IOC, scu->ipc_base); |
13263 |
+ } |
13264 |
+ |
13265 |
+ /* |
13266 |
+@@ -612,9 +608,10 @@ EXPORT_SYMBOL(intel_scu_ipc_i2c_cntrl); |
13267 |
+ static irqreturn_t ioc(int irq, void *dev_id) |
13268 |
+ { |
13269 |
+ struct intel_scu_ipc_dev *scu = dev_id; |
13270 |
++ int status = ipc_read_status(scu); |
13271 |
+ |
13272 |
+- if (scu->irq_mode) |
13273 |
+- complete(&scu->cmd_complete); |
13274 |
++ writel(status | IPC_STATUS_IRQ, scu->ipc_base + IPC_STATUS); |
13275 |
++ complete(&scu->cmd_complete); |
13276 |
+ |
13277 |
+ return IRQ_HANDLED; |
13278 |
+ } |
13279 |
+@@ -640,8 +637,6 @@ static int ipc_probe(struct pci_dev *pdev, const struct pci_device_id *id) |
13280 |
+ if (!pdata) |
13281 |
+ return -ENODEV; |
13282 |
+ |
13283 |
+- scu->irq_mode = pdata->irq_mode; |
13284 |
+- |
13285 |
+ err = pcim_enable_device(pdev); |
13286 |
+ if (err) |
13287 |
+ return err; |
13288 |
+diff --git a/drivers/power/supply/ltc2941-battery-gauge.c b/drivers/power/supply/ltc2941-battery-gauge.c |
13289 |
+index 4f129bb4c972..ff5febea1a21 100644 |
13290 |
+--- a/drivers/power/supply/ltc2941-battery-gauge.c |
13291 |
++++ b/drivers/power/supply/ltc2941-battery-gauge.c |
13292 |
+@@ -448,7 +448,7 @@ static int ltc294x_i2c_remove(struct i2c_client *client) |
13293 |
+ { |
13294 |
+ struct ltc294x_info *info = i2c_get_clientdata(client); |
13295 |
+ |
13296 |
+- cancel_delayed_work(&info->work); |
13297 |
++ cancel_delayed_work_sync(&info->work); |
13298 |
+ power_supply_unregister(info->supply); |
13299 |
+ return 0; |
13300 |
+ } |
13301 |
+diff --git a/drivers/scsi/csiostor/csio_scsi.c b/drivers/scsi/csiostor/csio_scsi.c |
13302 |
+index e09c7f360dbd..0cb585759de6 100644 |
13303 |
+--- a/drivers/scsi/csiostor/csio_scsi.c |
13304 |
++++ b/drivers/scsi/csiostor/csio_scsi.c |
13305 |
+@@ -1383,7 +1383,7 @@ csio_device_reset(struct device *dev, |
13306 |
+ return -EINVAL; |
13307 |
+ |
13308 |
+ /* Delete NPIV lnodes */ |
13309 |
+- csio_lnodes_exit(hw, 1); |
13310 |
++ csio_lnodes_exit(hw, 1); |
13311 |
+ |
13312 |
+ /* Block upper IOs */ |
13313 |
+ csio_lnodes_block_request(hw); |
13314 |
+diff --git a/drivers/scsi/qla2xxx/qla_dbg.c b/drivers/scsi/qla2xxx/qla_dbg.c |
13315 |
+index c7533fa7f46e..36871760a5d3 100644 |
13316 |
+--- a/drivers/scsi/qla2xxx/qla_dbg.c |
13317 |
++++ b/drivers/scsi/qla2xxx/qla_dbg.c |
13318 |
+@@ -2520,12 +2520,6 @@ qla83xx_fw_dump_failed: |
13319 |
+ /* Driver Debug Functions. */ |
13320 |
+ /****************************************************************************/ |
13321 |
+ |
13322 |
+-static inline int |
13323 |
+-ql_mask_match(uint32_t level) |
13324 |
+-{ |
13325 |
+- return (level & ql2xextended_error_logging) == level; |
13326 |
+-} |
13327 |
+- |
13328 |
+ /* |
13329 |
+ * This function is for formatting and logging debug information. |
13330 |
+ * It is to be used when vha is available. It formats the message |
13331 |
+diff --git a/drivers/scsi/qla2xxx/qla_dbg.h b/drivers/scsi/qla2xxx/qla_dbg.h |
13332 |
+index 8877aa97d829..ceca6dd34db1 100644 |
13333 |
+--- a/drivers/scsi/qla2xxx/qla_dbg.h |
13334 |
++++ b/drivers/scsi/qla2xxx/qla_dbg.h |
13335 |
+@@ -374,3 +374,9 @@ extern int qla24xx_dump_ram(struct qla_hw_data *, uint32_t, uint32_t *, |
13336 |
+ extern void qla24xx_pause_risc(struct device_reg_24xx __iomem *, |
13337 |
+ struct qla_hw_data *); |
13338 |
+ extern int qla24xx_soft_reset(struct qla_hw_data *); |
13339 |
++ |
13340 |
++static inline int |
13341 |
++ql_mask_match(uint level) |
13342 |
++{ |
13343 |
++ return (level & ql2xextended_error_logging) == level; |
13344 |
++} |
13345 |
+diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c |
13346 |
+index 01ded6c6ad38..f9b3151f4b10 100644 |
13347 |
+--- a/drivers/scsi/qla2xxx/qla_isr.c |
13348 |
++++ b/drivers/scsi/qla2xxx/qla_isr.c |
13349 |
+@@ -1876,6 +1876,18 @@ static void qla24xx_nvme_iocb_entry(scsi_qla_host_t *vha, struct req_que *req, |
13350 |
+ inbuf = (uint32_t *)&sts->nvme_ersp_data; |
13351 |
+ outbuf = (uint32_t *)fd->rspaddr; |
13352 |
+ iocb->u.nvme.rsp_pyld_len = le16_to_cpu(sts->nvme_rsp_pyld_len); |
13353 |
++ if (unlikely(iocb->u.nvme.rsp_pyld_len > |
13354 |
++ sizeof(struct nvme_fc_ersp_iu))) { |
13355 |
++ if (ql_mask_match(ql_dbg_io)) { |
13356 |
++ WARN_ONCE(1, "Unexpected response payload length %u.\n", |
13357 |
++ iocb->u.nvme.rsp_pyld_len); |
13358 |
++ ql_log(ql_log_warn, fcport->vha, 0x5100, |
13359 |
++ "Unexpected response payload length %u.\n", |
13360 |
++ iocb->u.nvme.rsp_pyld_len); |
13361 |
++ } |
13362 |
++ iocb->u.nvme.rsp_pyld_len = |
13363 |
++ sizeof(struct nvme_fc_ersp_iu); |
13364 |
++ } |
13365 |
+ iter = iocb->u.nvme.rsp_pyld_len >> 2; |
13366 |
+ for (; iter; iter--) |
13367 |
+ *outbuf++ = swab32(*inbuf++); |
13368 |
+diff --git a/drivers/scsi/qla2xxx/qla_mbx.c b/drivers/scsi/qla2xxx/qla_mbx.c |
13369 |
+index abef3b29fa10..bef9faea5eee 100644 |
13370 |
+--- a/drivers/scsi/qla2xxx/qla_mbx.c |
13371 |
++++ b/drivers/scsi/qla2xxx/qla_mbx.c |
13372 |
+@@ -5994,9 +5994,8 @@ qla2x00_dump_mctp_data(scsi_qla_host_t *vha, dma_addr_t req_dma, uint32_t addr, |
13373 |
+ mcp->mb[7] = LSW(MSD(req_dma)); |
13374 |
+ mcp->mb[8] = MSW(addr); |
13375 |
+ /* Setting RAM ID to valid */ |
13376 |
+- mcp->mb[10] |= BIT_7; |
13377 |
+ /* For MCTP RAM ID is 0x40 */ |
13378 |
+- mcp->mb[10] |= 0x40; |
13379 |
++ mcp->mb[10] = BIT_7 | 0x40; |
13380 |
+ |
13381 |
+ mcp->out_mb |= MBX_10|MBX_8|MBX_7|MBX_6|MBX_5|MBX_4|MBX_3|MBX_2|MBX_1| |
13382 |
+ MBX_0; |
13383 |
+diff --git a/drivers/scsi/qla2xxx/qla_nx.c b/drivers/scsi/qla2xxx/qla_nx.c |
13384 |
+index de2bc78449e7..3007eecfa509 100644 |
13385 |
+--- a/drivers/scsi/qla2xxx/qla_nx.c |
13386 |
++++ b/drivers/scsi/qla2xxx/qla_nx.c |
13387 |
+@@ -1605,8 +1605,7 @@ qla82xx_get_bootld_offset(struct qla_hw_data *ha) |
13388 |
+ return (u8 *)&ha->hablob->fw->data[offset]; |
13389 |
+ } |
13390 |
+ |
13391 |
+-static __le32 |
13392 |
+-qla82xx_get_fw_size(struct qla_hw_data *ha) |
13393 |
++static u32 qla82xx_get_fw_size(struct qla_hw_data *ha) |
13394 |
+ { |
13395 |
+ struct qla82xx_uri_data_desc *uri_desc = NULL; |
13396 |
+ |
13397 |
+@@ -1617,7 +1616,7 @@ qla82xx_get_fw_size(struct qla_hw_data *ha) |
13398 |
+ return cpu_to_le32(uri_desc->size); |
13399 |
+ } |
13400 |
+ |
13401 |
+- return cpu_to_le32(*(u32 *)&ha->hablob->fw->data[FW_SIZE_OFFSET]); |
13402 |
++ return get_unaligned_le32(&ha->hablob->fw->data[FW_SIZE_OFFSET]); |
13403 |
+ } |
13404 |
+ |
13405 |
+ static u8 * |
13406 |
+@@ -1808,7 +1807,7 @@ qla82xx_fw_load_from_blob(struct qla_hw_data *ha) |
13407 |
+ } |
13408 |
+ |
13409 |
+ flashaddr = FLASH_ADDR_START; |
13410 |
+- size = (__force u32)qla82xx_get_fw_size(ha) / 8; |
13411 |
++ size = qla82xx_get_fw_size(ha) / 8; |
13412 |
+ ptr64 = (u64 *)qla82xx_get_fw_offs(ha); |
13413 |
+ |
13414 |
+ for (i = 0; i < size; i++) { |
13415 |
+diff --git a/drivers/scsi/qla4xxx/ql4_os.c b/drivers/scsi/qla4xxx/ql4_os.c |
13416 |
+index f8acf101af3d..f59b8982b288 100644 |
13417 |
+--- a/drivers/scsi/qla4xxx/ql4_os.c |
13418 |
++++ b/drivers/scsi/qla4xxx/ql4_os.c |
13419 |
+@@ -4146,7 +4146,7 @@ static void qla4xxx_mem_free(struct scsi_qla_host *ha) |
13420 |
+ dma_free_coherent(&ha->pdev->dev, ha->queues_len, ha->queues, |
13421 |
+ ha->queues_dma); |
13422 |
+ |
13423 |
+- if (ha->fw_dump) |
13424 |
++ if (ha->fw_dump) |
13425 |
+ vfree(ha->fw_dump); |
13426 |
+ |
13427 |
+ ha->queues_len = 0; |
13428 |
+diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c |
13429 |
+index 3601e770da16..af01be59a721 100644 |
13430 |
+--- a/drivers/scsi/ufs/ufshcd.c |
13431 |
++++ b/drivers/scsi/ufs/ufshcd.c |
13432 |
+@@ -5044,6 +5044,7 @@ static int ufshcd_disable_auto_bkops(struct ufs_hba *hba) |
13433 |
+ |
13434 |
+ hba->auto_bkops_enabled = false; |
13435 |
+ trace_ufshcd_auto_bkops_state(dev_name(hba->dev), "Disabled"); |
13436 |
++ hba->is_urgent_bkops_lvl_checked = false; |
13437 |
+ out: |
13438 |
+ return err; |
13439 |
+ } |
13440 |
+@@ -5068,6 +5069,7 @@ static void ufshcd_force_reset_auto_bkops(struct ufs_hba *hba) |
13441 |
+ hba->ee_ctrl_mask &= ~MASK_EE_URGENT_BKOPS; |
13442 |
+ ufshcd_disable_auto_bkops(hba); |
13443 |
+ } |
13444 |
++ hba->is_urgent_bkops_lvl_checked = false; |
13445 |
+ } |
13446 |
+ |
13447 |
+ static inline int ufshcd_get_bkops_status(struct ufs_hba *hba, u32 *status) |
13448 |
+@@ -5114,6 +5116,7 @@ static int ufshcd_bkops_ctrl(struct ufs_hba *hba, |
13449 |
+ err = ufshcd_enable_auto_bkops(hba); |
13450 |
+ else |
13451 |
+ err = ufshcd_disable_auto_bkops(hba); |
13452 |
++ hba->urgent_bkops_lvl = curr_status; |
13453 |
+ out: |
13454 |
+ return err; |
13455 |
+ } |
13456 |
+diff --git a/drivers/usb/gadget/function/f_ecm.c b/drivers/usb/gadget/function/f_ecm.c |
13457 |
+index 460d5d7c984f..7f5cf488b2b1 100644 |
13458 |
+--- a/drivers/usb/gadget/function/f_ecm.c |
13459 |
++++ b/drivers/usb/gadget/function/f_ecm.c |
13460 |
+@@ -52,6 +52,7 @@ struct f_ecm { |
13461 |
+ struct usb_ep *notify; |
13462 |
+ struct usb_request *notify_req; |
13463 |
+ u8 notify_state; |
13464 |
++ atomic_t notify_count; |
13465 |
+ bool is_open; |
13466 |
+ |
13467 |
+ /* FIXME is_open needs some irq-ish locking |
13468 |
+@@ -380,7 +381,7 @@ static void ecm_do_notify(struct f_ecm *ecm) |
13469 |
+ int status; |
13470 |
+ |
13471 |
+ /* notification already in flight? */ |
13472 |
+- if (!req) |
13473 |
++ if (atomic_read(&ecm->notify_count)) |
13474 |
+ return; |
13475 |
+ |
13476 |
+ event = req->buf; |
13477 |
+@@ -420,10 +421,10 @@ static void ecm_do_notify(struct f_ecm *ecm) |
13478 |
+ event->bmRequestType = 0xA1; |
13479 |
+ event->wIndex = cpu_to_le16(ecm->ctrl_id); |
13480 |
+ |
13481 |
+- ecm->notify_req = NULL; |
13482 |
++ atomic_inc(&ecm->notify_count); |
13483 |
+ status = usb_ep_queue(ecm->notify, req, GFP_ATOMIC); |
13484 |
+ if (status < 0) { |
13485 |
+- ecm->notify_req = req; |
13486 |
++ atomic_dec(&ecm->notify_count); |
13487 |
+ DBG(cdev, "notify --> %d\n", status); |
13488 |
+ } |
13489 |
+ } |
13490 |
+@@ -448,17 +449,19 @@ static void ecm_notify_complete(struct usb_ep *ep, struct usb_request *req) |
13491 |
+ switch (req->status) { |
13492 |
+ case 0: |
13493 |
+ /* no fault */ |
13494 |
++ atomic_dec(&ecm->notify_count); |
13495 |
+ break; |
13496 |
+ case -ECONNRESET: |
13497 |
+ case -ESHUTDOWN: |
13498 |
++ atomic_set(&ecm->notify_count, 0); |
13499 |
+ ecm->notify_state = ECM_NOTIFY_NONE; |
13500 |
+ break; |
13501 |
+ default: |
13502 |
+ DBG(cdev, "event %02x --> %d\n", |
13503 |
+ event->bNotificationType, req->status); |
13504 |
++ atomic_dec(&ecm->notify_count); |
13505 |
+ break; |
13506 |
+ } |
13507 |
+- ecm->notify_req = req; |
13508 |
+ ecm_do_notify(ecm); |
13509 |
+ } |
13510 |
+ |
13511 |
+@@ -907,6 +910,11 @@ static void ecm_unbind(struct usb_configuration *c, struct usb_function *f) |
13512 |
+ |
13513 |
+ usb_free_all_descriptors(f); |
13514 |
+ |
13515 |
++ if (atomic_read(&ecm->notify_count)) { |
13516 |
++ usb_ep_dequeue(ecm->notify, ecm->notify_req); |
13517 |
++ atomic_set(&ecm->notify_count, 0); |
13518 |
++ } |
13519 |
++ |
13520 |
+ kfree(ecm->notify_req->buf); |
13521 |
+ usb_ep_free_request(ecm->notify, ecm->notify_req); |
13522 |
+ } |
13523 |
+diff --git a/drivers/usb/gadget/function/f_ncm.c b/drivers/usb/gadget/function/f_ncm.c |
13524 |
+index 5780fba620ab..cfca4584ae13 100644 |
13525 |
+--- a/drivers/usb/gadget/function/f_ncm.c |
13526 |
++++ b/drivers/usb/gadget/function/f_ncm.c |
13527 |
+@@ -54,6 +54,7 @@ struct f_ncm { |
13528 |
+ struct usb_ep *notify; |
13529 |
+ struct usb_request *notify_req; |
13530 |
+ u8 notify_state; |
13531 |
++ atomic_t notify_count; |
13532 |
+ bool is_open; |
13533 |
+ |
13534 |
+ const struct ndp_parser_opts *parser_opts; |
13535 |
+@@ -547,7 +548,7 @@ static void ncm_do_notify(struct f_ncm *ncm) |
13536 |
+ int status; |
13537 |
+ |
13538 |
+ /* notification already in flight? */ |
13539 |
+- if (!req) |
13540 |
++ if (atomic_read(&ncm->notify_count)) |
13541 |
+ return; |
13542 |
+ |
13543 |
+ event = req->buf; |
13544 |
+@@ -587,7 +588,8 @@ static void ncm_do_notify(struct f_ncm *ncm) |
13545 |
+ event->bmRequestType = 0xA1; |
13546 |
+ event->wIndex = cpu_to_le16(ncm->ctrl_id); |
13547 |
+ |
13548 |
+- ncm->notify_req = NULL; |
13549 |
++ atomic_inc(&ncm->notify_count); |
13550 |
++ |
13551 |
+ /* |
13552 |
+ * In double buffering if there is a space in FIFO, |
13553 |
+ * completion callback can be called right after the call, |
13554 |
+@@ -597,7 +599,7 @@ static void ncm_do_notify(struct f_ncm *ncm) |
13555 |
+ status = usb_ep_queue(ncm->notify, req, GFP_ATOMIC); |
13556 |
+ spin_lock(&ncm->lock); |
13557 |
+ if (status < 0) { |
13558 |
+- ncm->notify_req = req; |
13559 |
++ atomic_dec(&ncm->notify_count); |
13560 |
+ DBG(cdev, "notify --> %d\n", status); |
13561 |
+ } |
13562 |
+ } |
13563 |
+@@ -632,17 +634,19 @@ static void ncm_notify_complete(struct usb_ep *ep, struct usb_request *req) |
13564 |
+ case 0: |
13565 |
+ VDBG(cdev, "Notification %02x sent\n", |
13566 |
+ event->bNotificationType); |
13567 |
++ atomic_dec(&ncm->notify_count); |
13568 |
+ break; |
13569 |
+ case -ECONNRESET: |
13570 |
+ case -ESHUTDOWN: |
13571 |
++ atomic_set(&ncm->notify_count, 0); |
13572 |
+ ncm->notify_state = NCM_NOTIFY_NONE; |
13573 |
+ break; |
13574 |
+ default: |
13575 |
+ DBG(cdev, "event %02x --> %d\n", |
13576 |
+ event->bNotificationType, req->status); |
13577 |
++ atomic_dec(&ncm->notify_count); |
13578 |
+ break; |
13579 |
+ } |
13580 |
+- ncm->notify_req = req; |
13581 |
+ ncm_do_notify(ncm); |
13582 |
+ spin_unlock(&ncm->lock); |
13583 |
+ } |
13584 |
+@@ -1612,6 +1616,11 @@ static void ncm_unbind(struct usb_configuration *c, struct usb_function *f) |
13585 |
+ ncm_string_defs[0].id = 0; |
13586 |
+ usb_free_all_descriptors(f); |
13587 |
+ |
13588 |
++ if (atomic_read(&ncm->notify_count)) { |
13589 |
++ usb_ep_dequeue(ncm->notify, ncm->notify_req); |
13590 |
++ atomic_set(&ncm->notify_count, 0); |
13591 |
++ } |
13592 |
++ |
13593 |
+ kfree(ncm->notify_req->buf); |
13594 |
+ usb_ep_free_request(ncm->notify, ncm->notify_req); |
13595 |
+ } |
13596 |
+diff --git a/drivers/usb/gadget/legacy/cdc2.c b/drivers/usb/gadget/legacy/cdc2.c |
13597 |
+index da1c37933ca1..8d7a556ece30 100644 |
13598 |
+--- a/drivers/usb/gadget/legacy/cdc2.c |
13599 |
++++ b/drivers/usb/gadget/legacy/cdc2.c |
13600 |
+@@ -225,7 +225,7 @@ static struct usb_composite_driver cdc_driver = { |
13601 |
+ .name = "g_cdc", |
13602 |
+ .dev = &device_desc, |
13603 |
+ .strings = dev_strings, |
13604 |
+- .max_speed = USB_SPEED_HIGH, |
13605 |
++ .max_speed = USB_SPEED_SUPER, |
13606 |
+ .bind = cdc_bind, |
13607 |
+ .unbind = cdc_unbind, |
13608 |
+ }; |
13609 |
+diff --git a/drivers/usb/gadget/legacy/g_ffs.c b/drivers/usb/gadget/legacy/g_ffs.c |
13610 |
+index b640ed3fcf70..ae6d8f7092b8 100644 |
13611 |
+--- a/drivers/usb/gadget/legacy/g_ffs.c |
13612 |
++++ b/drivers/usb/gadget/legacy/g_ffs.c |
13613 |
+@@ -149,7 +149,7 @@ static struct usb_composite_driver gfs_driver = { |
13614 |
+ .name = DRIVER_NAME, |
13615 |
+ .dev = &gfs_dev_desc, |
13616 |
+ .strings = gfs_dev_strings, |
13617 |
+- .max_speed = USB_SPEED_HIGH, |
13618 |
++ .max_speed = USB_SPEED_SUPER, |
13619 |
+ .bind = gfs_bind, |
13620 |
+ .unbind = gfs_unbind, |
13621 |
+ }; |
13622 |
+diff --git a/drivers/usb/gadget/legacy/multi.c b/drivers/usb/gadget/legacy/multi.c |
13623 |
+index 50515f9e1022..ec9749845660 100644 |
13624 |
+--- a/drivers/usb/gadget/legacy/multi.c |
13625 |
++++ b/drivers/usb/gadget/legacy/multi.c |
13626 |
+@@ -482,7 +482,7 @@ static struct usb_composite_driver multi_driver = { |
13627 |
+ .name = "g_multi", |
13628 |
+ .dev = &device_desc, |
13629 |
+ .strings = dev_strings, |
13630 |
+- .max_speed = USB_SPEED_HIGH, |
13631 |
++ .max_speed = USB_SPEED_SUPER, |
13632 |
+ .bind = multi_bind, |
13633 |
+ .unbind = multi_unbind, |
13634 |
+ .needs_serial = 1, |
13635 |
+diff --git a/drivers/usb/gadget/legacy/ncm.c b/drivers/usb/gadget/legacy/ncm.c |
13636 |
+index 8465f081e921..c61e71ba7045 100644 |
13637 |
+--- a/drivers/usb/gadget/legacy/ncm.c |
13638 |
++++ b/drivers/usb/gadget/legacy/ncm.c |
13639 |
+@@ -197,7 +197,7 @@ static struct usb_composite_driver ncm_driver = { |
13640 |
+ .name = "g_ncm", |
13641 |
+ .dev = &device_desc, |
13642 |
+ .strings = dev_strings, |
13643 |
+- .max_speed = USB_SPEED_HIGH, |
13644 |
++ .max_speed = USB_SPEED_SUPER, |
13645 |
+ .bind = gncm_bind, |
13646 |
+ .unbind = gncm_unbind, |
13647 |
+ }; |
13648 |
+diff --git a/drivers/usb/typec/tcpci.c b/drivers/usb/typec/tcpci.c |
13649 |
+index c1f7073a56de..dfae41fe1331 100644 |
13650 |
+--- a/drivers/usb/typec/tcpci.c |
13651 |
++++ b/drivers/usb/typec/tcpci.c |
13652 |
+@@ -581,6 +581,12 @@ static int tcpci_probe(struct i2c_client *client, |
13653 |
+ static int tcpci_remove(struct i2c_client *client) |
13654 |
+ { |
13655 |
+ struct tcpci_chip *chip = i2c_get_clientdata(client); |
13656 |
++ int err; |
13657 |
++ |
13658 |
++ /* Disable chip interrupts before unregistering port */ |
13659 |
++ err = tcpci_write16(chip->tcpci, TCPC_ALERT_MASK, 0); |
13660 |
++ if (err < 0) |
13661 |
++ return err; |
13662 |
+ |
13663 |
+ tcpci_unregister_port(chip->tcpci); |
13664 |
+ |
13665 |
+diff --git a/drivers/watchdog/watchdog_core.c b/drivers/watchdog/watchdog_core.c |
13666 |
+index eb8fa25f8eb2..8b1f37ffb65a 100644 |
13667 |
+--- a/drivers/watchdog/watchdog_core.c |
13668 |
++++ b/drivers/watchdog/watchdog_core.c |
13669 |
+@@ -138,6 +138,25 @@ int watchdog_init_timeout(struct watchdog_device *wdd, |
13670 |
+ } |
13671 |
+ EXPORT_SYMBOL_GPL(watchdog_init_timeout); |
13672 |
+ |
13673 |
++static int watchdog_reboot_notifier(struct notifier_block *nb, |
13674 |
++ unsigned long code, void *data) |
13675 |
++{ |
13676 |
++ struct watchdog_device *wdd; |
13677 |
++ |
13678 |
++ wdd = container_of(nb, struct watchdog_device, reboot_nb); |
13679 |
++ if (code == SYS_DOWN || code == SYS_HALT) { |
13680 |
++ if (watchdog_active(wdd)) { |
13681 |
++ int ret; |
13682 |
++ |
13683 |
++ ret = wdd->ops->stop(wdd); |
13684 |
++ if (ret) |
13685 |
++ return NOTIFY_BAD; |
13686 |
++ } |
13687 |
++ } |
13688 |
++ |
13689 |
++ return NOTIFY_DONE; |
13690 |
++} |
13691 |
++ |
13692 |
+ static int watchdog_restart_notifier(struct notifier_block *nb, |
13693 |
+ unsigned long action, void *data) |
13694 |
+ { |
13695 |
+@@ -226,6 +245,19 @@ static int __watchdog_register_device(struct watchdog_device *wdd) |
13696 |
+ } |
13697 |
+ } |
13698 |
+ |
13699 |
++ if (test_bit(WDOG_STOP_ON_REBOOT, &wdd->status)) { |
13700 |
++ wdd->reboot_nb.notifier_call = watchdog_reboot_notifier; |
13701 |
++ |
13702 |
++ ret = register_reboot_notifier(&wdd->reboot_nb); |
13703 |
++ if (ret) { |
13704 |
++ pr_err("watchdog%d: Cannot register reboot notifier (%d)\n", |
13705 |
++ wdd->id, ret); |
13706 |
++ watchdog_dev_unregister(wdd); |
13707 |
++ ida_simple_remove(&watchdog_ida, id); |
13708 |
++ return ret; |
13709 |
++ } |
13710 |
++ } |
13711 |
++ |
13712 |
+ if (wdd->ops->restart) { |
13713 |
+ wdd->restart_nb.notifier_call = watchdog_restart_notifier; |
13714 |
+ |
13715 |
+@@ -271,6 +303,9 @@ static void __watchdog_unregister_device(struct watchdog_device *wdd) |
13716 |
+ if (wdd->ops->restart) |
13717 |
+ unregister_restart_handler(&wdd->restart_nb); |
13718 |
+ |
13719 |
++ if (test_bit(WDOG_STOP_ON_REBOOT, &wdd->status)) |
13720 |
++ unregister_reboot_notifier(&wdd->reboot_nb); |
13721 |
++ |
13722 |
+ watchdog_dev_unregister(wdd); |
13723 |
+ ida_simple_remove(&watchdog_ida, wdd->id); |
13724 |
+ } |
13725 |
+diff --git a/drivers/watchdog/watchdog_dev.c b/drivers/watchdog/watchdog_dev.c |
13726 |
+index 4b89333e8eb4..e64aa88e99da 100644 |
13727 |
+--- a/drivers/watchdog/watchdog_dev.c |
13728 |
++++ b/drivers/watchdog/watchdog_dev.c |
13729 |
+@@ -42,7 +42,6 @@ |
13730 |
+ #include <linux/miscdevice.h> /* For handling misc devices */ |
13731 |
+ #include <linux/module.h> /* For module stuff/... */ |
13732 |
+ #include <linux/mutex.h> /* For mutexes */ |
13733 |
+-#include <linux/reboot.h> /* For reboot notifier */ |
13734 |
+ #include <linux/slab.h> /* For memory functions */ |
13735 |
+ #include <linux/types.h> /* For standard types (like size_t) */ |
13736 |
+ #include <linux/watchdog.h> /* For watchdog specific items */ |
13737 |
+@@ -1048,25 +1047,6 @@ static void watchdog_cdev_unregister(struct watchdog_device *wdd) |
13738 |
+ put_device(&wd_data->dev); |
13739 |
+ } |
13740 |
+ |
13741 |
+-static int watchdog_reboot_notifier(struct notifier_block *nb, |
13742 |
+- unsigned long code, void *data) |
13743 |
+-{ |
13744 |
+- struct watchdog_device *wdd; |
13745 |
+- |
13746 |
+- wdd = container_of(nb, struct watchdog_device, reboot_nb); |
13747 |
+- if (code == SYS_DOWN || code == SYS_HALT) { |
13748 |
+- if (watchdog_active(wdd)) { |
13749 |
+- int ret; |
13750 |
+- |
13751 |
+- ret = wdd->ops->stop(wdd); |
13752 |
+- if (ret) |
13753 |
+- return NOTIFY_BAD; |
13754 |
+- } |
13755 |
+- } |
13756 |
+- |
13757 |
+- return NOTIFY_DONE; |
13758 |
+-} |
13759 |
+- |
13760 |
+ /* |
13761 |
+ * watchdog_dev_register: register a watchdog device |
13762 |
+ * @wdd: watchdog device |
13763 |
+@@ -1085,22 +1065,8 @@ int watchdog_dev_register(struct watchdog_device *wdd) |
13764 |
+ return ret; |
13765 |
+ |
13766 |
+ ret = watchdog_register_pretimeout(wdd); |
13767 |
+- if (ret) { |
13768 |
++ if (ret) |
13769 |
+ watchdog_cdev_unregister(wdd); |
13770 |
+- return ret; |
13771 |
+- } |
13772 |
+- |
13773 |
+- if (test_bit(WDOG_STOP_ON_REBOOT, &wdd->status)) { |
13774 |
+- wdd->reboot_nb.notifier_call = watchdog_reboot_notifier; |
13775 |
+- |
13776 |
+- ret = devm_register_reboot_notifier(&wdd->wd_data->dev, |
13777 |
+- &wdd->reboot_nb); |
13778 |
+- if (ret) { |
13779 |
+- pr_err("watchdog%d: Cannot register reboot notifier (%d)\n", |
13780 |
+- wdd->id, ret); |
13781 |
+- watchdog_dev_unregister(wdd); |
13782 |
+- } |
13783 |
+- } |
13784 |
+ |
13785 |
+ return ret; |
13786 |
+ } |
13787 |
+diff --git a/drivers/xen/xen-balloon.c b/drivers/xen/xen-balloon.c |
13788 |
+index 2acbfe104e46..3aab77916915 100644 |
13789 |
+--- a/drivers/xen/xen-balloon.c |
13790 |
++++ b/drivers/xen/xen-balloon.c |
13791 |
+@@ -83,7 +83,7 @@ static void watch_target(struct xenbus_watch *watch, |
13792 |
+ "%llu", &static_max) == 1)) |
13793 |
+ static_max >>= PAGE_SHIFT - 10; |
13794 |
+ else |
13795 |
+- static_max = new_target; |
13796 |
++ static_max = balloon_stats.current_pages; |
13797 |
+ |
13798 |
+ target_diff = (xen_pv_domain() || xen_initial_domain()) ? 0 |
13799 |
+ : static_max - balloon_stats.target_pages; |
13800 |
+diff --git a/fs/aio.c b/fs/aio.c |
13801 |
+index 911e23087dfb..b5fbf2061868 100644 |
13802 |
+--- a/fs/aio.c |
13803 |
++++ b/fs/aio.c |
13804 |
+@@ -1600,6 +1600,14 @@ static int aio_fsync(struct fsync_iocb *req, const struct iocb *iocb, |
13805 |
+ return 0; |
13806 |
+ } |
13807 |
+ |
13808 |
++static void aio_poll_put_work(struct work_struct *work) |
13809 |
++{ |
13810 |
++ struct poll_iocb *req = container_of(work, struct poll_iocb, work); |
13811 |
++ struct aio_kiocb *iocb = container_of(req, struct aio_kiocb, poll); |
13812 |
++ |
13813 |
++ iocb_put(iocb); |
13814 |
++} |
13815 |
++ |
13816 |
+ static void aio_poll_complete_work(struct work_struct *work) |
13817 |
+ { |
13818 |
+ struct poll_iocb *req = container_of(work, struct poll_iocb, work); |
13819 |
+@@ -1664,6 +1672,8 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, |
13820 |
+ list_del_init(&req->wait.entry); |
13821 |
+ |
13822 |
+ if (mask && spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) { |
13823 |
++ struct kioctx *ctx = iocb->ki_ctx; |
13824 |
++ |
13825 |
+ /* |
13826 |
+ * Try to complete the iocb inline if we can. Use |
13827 |
+ * irqsave/irqrestore because not all filesystems (e.g. fuse) |
13828 |
+@@ -1673,8 +1683,14 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, |
13829 |
+ list_del(&iocb->ki_list); |
13830 |
+ iocb->ki_res.res = mangle_poll(mask); |
13831 |
+ req->done = true; |
13832 |
+- spin_unlock_irqrestore(&iocb->ki_ctx->ctx_lock, flags); |
13833 |
+- iocb_put(iocb); |
13834 |
++ if (iocb->ki_eventfd && eventfd_signal_count()) { |
13835 |
++ iocb = NULL; |
13836 |
++ INIT_WORK(&req->work, aio_poll_put_work); |
13837 |
++ schedule_work(&req->work); |
13838 |
++ } |
13839 |
++ spin_unlock_irqrestore(&ctx->ctx_lock, flags); |
13840 |
++ if (iocb) |
13841 |
++ iocb_put(iocb); |
13842 |
+ } else { |
13843 |
+ schedule_work(&req->work); |
13844 |
+ } |
13845 |
+diff --git a/fs/btrfs/ctree.c b/fs/btrfs/ctree.c |
13846 |
+index 84ff398ae70b..c9943d70e2cb 100644 |
13847 |
+--- a/fs/btrfs/ctree.c |
13848 |
++++ b/fs/btrfs/ctree.c |
13849 |
+@@ -337,12 +337,10 @@ u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info, |
13850 |
+ struct seq_list *elem) |
13851 |
+ { |
13852 |
+ write_lock(&fs_info->tree_mod_log_lock); |
13853 |
+- spin_lock(&fs_info->tree_mod_seq_lock); |
13854 |
+ if (!elem->seq) { |
13855 |
+ elem->seq = btrfs_inc_tree_mod_seq(fs_info); |
13856 |
+ list_add_tail(&elem->list, &fs_info->tree_mod_seq_list); |
13857 |
+ } |
13858 |
+- spin_unlock(&fs_info->tree_mod_seq_lock); |
13859 |
+ write_unlock(&fs_info->tree_mod_log_lock); |
13860 |
+ |
13861 |
+ return elem->seq; |
13862 |
+@@ -362,7 +360,7 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info, |
13863 |
+ if (!seq_putting) |
13864 |
+ return; |
13865 |
+ |
13866 |
+- spin_lock(&fs_info->tree_mod_seq_lock); |
13867 |
++ write_lock(&fs_info->tree_mod_log_lock); |
13868 |
+ list_del(&elem->list); |
13869 |
+ elem->seq = 0; |
13870 |
+ |
13871 |
+@@ -373,19 +371,17 @@ void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info, |
13872 |
+ * blocker with lower sequence number exists, we |
13873 |
+ * cannot remove anything from the log |
13874 |
+ */ |
13875 |
+- spin_unlock(&fs_info->tree_mod_seq_lock); |
13876 |
++ write_unlock(&fs_info->tree_mod_log_lock); |
13877 |
+ return; |
13878 |
+ } |
13879 |
+ min_seq = cur_elem->seq; |
13880 |
+ } |
13881 |
+ } |
13882 |
+- spin_unlock(&fs_info->tree_mod_seq_lock); |
13883 |
+ |
13884 |
+ /* |
13885 |
+ * anything that's lower than the lowest existing (read: blocked) |
13886 |
+ * sequence number can be removed from the tree. |
13887 |
+ */ |
13888 |
+- write_lock(&fs_info->tree_mod_log_lock); |
13889 |
+ tm_root = &fs_info->tree_mod_log; |
13890 |
+ for (node = rb_first(tm_root); node; node = next) { |
13891 |
+ next = rb_next(node); |
13892 |
+diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h |
13893 |
+index d24ecbf938b6..15cb96ad15d8 100644 |
13894 |
+--- a/fs/btrfs/ctree.h |
13895 |
++++ b/fs/btrfs/ctree.h |
13896 |
+@@ -893,14 +893,12 @@ struct btrfs_fs_info { |
13897 |
+ struct list_head delayed_iputs; |
13898 |
+ struct mutex cleaner_delayed_iput_mutex; |
13899 |
+ |
13900 |
+- /* this protects tree_mod_seq_list */ |
13901 |
+- spinlock_t tree_mod_seq_lock; |
13902 |
+ atomic64_t tree_mod_seq; |
13903 |
+- struct list_head tree_mod_seq_list; |
13904 |
+ |
13905 |
+- /* this protects tree_mod_log */ |
13906 |
++ /* this protects tree_mod_log and tree_mod_seq_list */ |
13907 |
+ rwlock_t tree_mod_log_lock; |
13908 |
+ struct rb_root tree_mod_log; |
13909 |
++ struct list_head tree_mod_seq_list; |
13910 |
+ |
13911 |
+ atomic_t async_delalloc_pages; |
13912 |
+ |
13913 |
+diff --git a/fs/btrfs/delayed-ref.c b/fs/btrfs/delayed-ref.c |
13914 |
+index 7e5c81e80e15..09a12115b640 100644 |
13915 |
+--- a/fs/btrfs/delayed-ref.c |
13916 |
++++ b/fs/btrfs/delayed-ref.c |
13917 |
+@@ -301,7 +301,7 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, |
13918 |
+ if (head->is_data) |
13919 |
+ return; |
13920 |
+ |
13921 |
+- spin_lock(&fs_info->tree_mod_seq_lock); |
13922 |
++ read_lock(&fs_info->tree_mod_log_lock); |
13923 |
+ if (!list_empty(&fs_info->tree_mod_seq_list)) { |
13924 |
+ struct seq_list *elem; |
13925 |
+ |
13926 |
+@@ -309,7 +309,7 @@ void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans, |
13927 |
+ struct seq_list, list); |
13928 |
+ seq = elem->seq; |
13929 |
+ } |
13930 |
+- spin_unlock(&fs_info->tree_mod_seq_lock); |
13931 |
++ read_unlock(&fs_info->tree_mod_log_lock); |
13932 |
+ |
13933 |
+ again: |
13934 |
+ for (node = rb_first(&head->ref_tree); node; node = rb_next(node)) { |
13935 |
+@@ -326,7 +326,7 @@ int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq) |
13936 |
+ struct seq_list *elem; |
13937 |
+ int ret = 0; |
13938 |
+ |
13939 |
+- spin_lock(&fs_info->tree_mod_seq_lock); |
13940 |
++ read_lock(&fs_info->tree_mod_log_lock); |
13941 |
+ if (!list_empty(&fs_info->tree_mod_seq_list)) { |
13942 |
+ elem = list_first_entry(&fs_info->tree_mod_seq_list, |
13943 |
+ struct seq_list, list); |
13944 |
+@@ -339,7 +339,7 @@ int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq) |
13945 |
+ } |
13946 |
+ } |
13947 |
+ |
13948 |
+- spin_unlock(&fs_info->tree_mod_seq_lock); |
13949 |
++ read_unlock(&fs_info->tree_mod_log_lock); |
13950 |
+ return ret; |
13951 |
+ } |
13952 |
+ |
13953 |
+diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c |
13954 |
+index e12c37f457e0..9e467e8a8cb5 100644 |
13955 |
+--- a/fs/btrfs/disk-io.c |
13956 |
++++ b/fs/btrfs/disk-io.c |
13957 |
+@@ -2031,7 +2031,7 @@ static void free_root_extent_buffers(struct btrfs_root *root) |
13958 |
+ } |
13959 |
+ |
13960 |
+ /* helper to cleanup tree roots */ |
13961 |
+-static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) |
13962 |
++static void free_root_pointers(struct btrfs_fs_info *info, bool free_chunk_root) |
13963 |
+ { |
13964 |
+ free_root_extent_buffers(info->tree_root); |
13965 |
+ |
13966 |
+@@ -2040,7 +2040,7 @@ static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root) |
13967 |
+ free_root_extent_buffers(info->csum_root); |
13968 |
+ free_root_extent_buffers(info->quota_root); |
13969 |
+ free_root_extent_buffers(info->uuid_root); |
13970 |
+- if (chunk_root) |
13971 |
++ if (free_chunk_root) |
13972 |
+ free_root_extent_buffers(info->chunk_root); |
13973 |
+ free_root_extent_buffers(info->free_space_root); |
13974 |
+ } |
13975 |
+@@ -2645,7 +2645,6 @@ int open_ctree(struct super_block *sb, |
13976 |
+ spin_lock_init(&fs_info->fs_roots_radix_lock); |
13977 |
+ spin_lock_init(&fs_info->delayed_iput_lock); |
13978 |
+ spin_lock_init(&fs_info->defrag_inodes_lock); |
13979 |
+- spin_lock_init(&fs_info->tree_mod_seq_lock); |
13980 |
+ spin_lock_init(&fs_info->super_lock); |
13981 |
+ spin_lock_init(&fs_info->qgroup_op_lock); |
13982 |
+ spin_lock_init(&fs_info->buffer_lock); |
13983 |
+@@ -3274,7 +3273,7 @@ fail_block_groups: |
13984 |
+ btrfs_put_block_group_cache(fs_info); |
13985 |
+ |
13986 |
+ fail_tree_roots: |
13987 |
+- free_root_pointers(fs_info, 1); |
13988 |
++ free_root_pointers(fs_info, true); |
13989 |
+ invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
13990 |
+ |
13991 |
+ fail_sb_buffer: |
13992 |
+@@ -3302,7 +3301,7 @@ recovery_tree_root: |
13993 |
+ if (!btrfs_test_opt(fs_info, USEBACKUPROOT)) |
13994 |
+ goto fail_tree_roots; |
13995 |
+ |
13996 |
+- free_root_pointers(fs_info, 0); |
13997 |
++ free_root_pointers(fs_info, false); |
13998 |
+ |
13999 |
+ /* don't use the log in recovery mode, it won't be valid */ |
14000 |
+ btrfs_set_super_log_root(disk_super, 0); |
14001 |
+@@ -3984,10 +3983,17 @@ void close_ctree(struct btrfs_fs_info *fs_info) |
14002 |
+ invalidate_inode_pages2(fs_info->btree_inode->i_mapping); |
14003 |
+ btrfs_stop_all_workers(fs_info); |
14004 |
+ |
14005 |
+- btrfs_free_block_groups(fs_info); |
14006 |
+- |
14007 |
+ clear_bit(BTRFS_FS_OPEN, &fs_info->flags); |
14008 |
+- free_root_pointers(fs_info, 1); |
14009 |
++ free_root_pointers(fs_info, true); |
14010 |
++ |
14011 |
++ /* |
14012 |
++ * We must free the block groups after dropping the fs_roots as we could |
14013 |
++ * have had an IO error and have left over tree log blocks that aren't |
14014 |
++ * cleaned up until the fs roots are freed. This makes the block group |
14015 |
++ * accounting appear to be wrong because there's pending reserved bytes, |
14016 |
++ * so make sure we do the block group cleanup afterwards. |
14017 |
++ */ |
14018 |
++ btrfs_free_block_groups(fs_info); |
14019 |
+ |
14020 |
+ iput(fs_info->btree_inode); |
14021 |
+ |
14022 |
+diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c |
14023 |
+index fed44390c049..11efb4f5041c 100644 |
14024 |
+--- a/fs/btrfs/extent_io.c |
14025 |
++++ b/fs/btrfs/extent_io.c |
14026 |
+@@ -4014,6 +4014,14 @@ retry: |
14027 |
+ */ |
14028 |
+ scanned = 1; |
14029 |
+ index = 0; |
14030 |
++ |
14031 |
++ /* |
14032 |
++ * If we're looping we could run into a page that is locked by a |
14033 |
++ * writer and that writer could be waiting on writeback for a |
14034 |
++ * page in our current bio, and thus deadlock, so flush the |
14035 |
++ * write bio here. |
14036 |
++ */ |
14037 |
++ flush_write_bio(epd); |
14038 |
+ goto retry; |
14039 |
+ } |
14040 |
+ |
14041 |
+diff --git a/fs/btrfs/tests/btrfs-tests.c b/fs/btrfs/tests/btrfs-tests.c |
14042 |
+index db72b3b6209e..2eec1dd3803a 100644 |
14043 |
+--- a/fs/btrfs/tests/btrfs-tests.c |
14044 |
++++ b/fs/btrfs/tests/btrfs-tests.c |
14045 |
+@@ -102,7 +102,6 @@ struct btrfs_fs_info *btrfs_alloc_dummy_fs_info(u32 nodesize, u32 sectorsize) |
14046 |
+ spin_lock_init(&fs_info->qgroup_op_lock); |
14047 |
+ spin_lock_init(&fs_info->super_lock); |
14048 |
+ spin_lock_init(&fs_info->fs_roots_radix_lock); |
14049 |
+- spin_lock_init(&fs_info->tree_mod_seq_lock); |
14050 |
+ mutex_init(&fs_info->qgroup_ioctl_lock); |
14051 |
+ mutex_init(&fs_info->qgroup_rescan_lock); |
14052 |
+ rwlock_init(&fs_info->tree_mod_log_lock); |
14053 |
+diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c |
14054 |
+index 26317bca5649..4b1491e1b803 100644 |
14055 |
+--- a/fs/btrfs/transaction.c |
14056 |
++++ b/fs/btrfs/transaction.c |
14057 |
+@@ -1936,6 +1936,14 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans) |
14058 |
+ struct btrfs_transaction *prev_trans = NULL; |
14059 |
+ int ret; |
14060 |
+ |
14061 |
++ /* |
14062 |
++ * Some places just start a transaction to commit it. We need to make |
14063 |
++ * sure that if this commit fails that the abort code actually marks the |
14064 |
++ * transaction as failed, so set trans->dirty to make the abort code do |
14065 |
++ * the right thing. |
14066 |
++ */ |
14067 |
++ trans->dirty = true; |
14068 |
++ |
14069 |
+ /* Stop the commit early if ->aborted is set */ |
14070 |
+ if (unlikely(READ_ONCE(cur_trans->aborted))) { |
14071 |
+ ret = cur_trans->aborted; |
14072 |
+diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c |
14073 |
+index fe7165c9d875..d4c86c6cbe39 100644 |
14074 |
+--- a/fs/btrfs/tree-log.c |
14075 |
++++ b/fs/btrfs/tree-log.c |
14076 |
+@@ -3892,7 +3892,7 @@ static int log_csums(struct btrfs_trans_handle *trans, |
14077 |
+ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14078 |
+ struct btrfs_inode *inode, |
14079 |
+ struct btrfs_path *dst_path, |
14080 |
+- struct btrfs_path *src_path, u64 *last_extent, |
14081 |
++ struct btrfs_path *src_path, |
14082 |
+ int start_slot, int nr, int inode_only, |
14083 |
+ u64 logged_isize) |
14084 |
+ { |
14085 |
+@@ -3903,7 +3903,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14086 |
+ struct btrfs_file_extent_item *extent; |
14087 |
+ struct btrfs_inode_item *inode_item; |
14088 |
+ struct extent_buffer *src = src_path->nodes[0]; |
14089 |
+- struct btrfs_key first_key, last_key, key; |
14090 |
+ int ret; |
14091 |
+ struct btrfs_key *ins_keys; |
14092 |
+ u32 *ins_sizes; |
14093 |
+@@ -3911,9 +3910,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14094 |
+ int i; |
14095 |
+ struct list_head ordered_sums; |
14096 |
+ int skip_csum = inode->flags & BTRFS_INODE_NODATASUM; |
14097 |
+- bool has_extents = false; |
14098 |
+- bool need_find_last_extent = true; |
14099 |
+- bool done = false; |
14100 |
+ |
14101 |
+ INIT_LIST_HEAD(&ordered_sums); |
14102 |
+ |
14103 |
+@@ -3922,8 +3918,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14104 |
+ if (!ins_data) |
14105 |
+ return -ENOMEM; |
14106 |
+ |
14107 |
+- first_key.objectid = (u64)-1; |
14108 |
+- |
14109 |
+ ins_sizes = (u32 *)ins_data; |
14110 |
+ ins_keys = (struct btrfs_key *)(ins_data + nr * sizeof(u32)); |
14111 |
+ |
14112 |
+@@ -3944,9 +3938,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14113 |
+ |
14114 |
+ src_offset = btrfs_item_ptr_offset(src, start_slot + i); |
14115 |
+ |
14116 |
+- if (i == nr - 1) |
14117 |
+- last_key = ins_keys[i]; |
14118 |
+- |
14119 |
+ if (ins_keys[i].type == BTRFS_INODE_ITEM_KEY) { |
14120 |
+ inode_item = btrfs_item_ptr(dst_path->nodes[0], |
14121 |
+ dst_path->slots[0], |
14122 |
+@@ -3960,20 +3951,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14123 |
+ src_offset, ins_sizes[i]); |
14124 |
+ } |
14125 |
+ |
14126 |
+- /* |
14127 |
+- * We set need_find_last_extent here in case we know we were |
14128 |
+- * processing other items and then walk into the first extent in |
14129 |
+- * the inode. If we don't hit an extent then nothing changes, |
14130 |
+- * we'll do the last search the next time around. |
14131 |
+- */ |
14132 |
+- if (ins_keys[i].type == BTRFS_EXTENT_DATA_KEY) { |
14133 |
+- has_extents = true; |
14134 |
+- if (first_key.objectid == (u64)-1) |
14135 |
+- first_key = ins_keys[i]; |
14136 |
+- } else { |
14137 |
+- need_find_last_extent = false; |
14138 |
+- } |
14139 |
+- |
14140 |
+ /* take a reference on file data extents so that truncates |
14141 |
+ * or deletes of this inode don't have to relog the inode |
14142 |
+ * again |
14143 |
+@@ -4039,167 +4016,6 @@ static noinline int copy_items(struct btrfs_trans_handle *trans, |
14144 |
+ kfree(sums); |
14145 |
+ } |
14146 |
+ |
14147 |
+- if (!has_extents) |
14148 |
+- return ret; |
14149 |
+- |
14150 |
+- if (need_find_last_extent && *last_extent == first_key.offset) { |
14151 |
+- /* |
14152 |
+- * We don't have any leafs between our current one and the one |
14153 |
+- * we processed before that can have file extent items for our |
14154 |
+- * inode (and have a generation number smaller than our current |
14155 |
+- * transaction id). |
14156 |
+- */ |
14157 |
+- need_find_last_extent = false; |
14158 |
+- } |
14159 |
+- |
14160 |
+- /* |
14161 |
+- * Because we use btrfs_search_forward we could skip leaves that were |
14162 |
+- * not modified and then assume *last_extent is valid when it really |
14163 |
+- * isn't. So back up to the previous leaf and read the end of the last |
14164 |
+- * extent before we go and fill in holes. |
14165 |
+- */ |
14166 |
+- if (need_find_last_extent) { |
14167 |
+- u64 len; |
14168 |
+- |
14169 |
+- ret = btrfs_prev_leaf(inode->root, src_path); |
14170 |
+- if (ret < 0) |
14171 |
+- return ret; |
14172 |
+- if (ret) |
14173 |
+- goto fill_holes; |
14174 |
+- if (src_path->slots[0]) |
14175 |
+- src_path->slots[0]--; |
14176 |
+- src = src_path->nodes[0]; |
14177 |
+- btrfs_item_key_to_cpu(src, &key, src_path->slots[0]); |
14178 |
+- if (key.objectid != btrfs_ino(inode) || |
14179 |
+- key.type != BTRFS_EXTENT_DATA_KEY) |
14180 |
+- goto fill_holes; |
14181 |
+- extent = btrfs_item_ptr(src, src_path->slots[0], |
14182 |
+- struct btrfs_file_extent_item); |
14183 |
+- if (btrfs_file_extent_type(src, extent) == |
14184 |
+- BTRFS_FILE_EXTENT_INLINE) { |
14185 |
+- len = btrfs_file_extent_ram_bytes(src, extent); |
14186 |
+- *last_extent = ALIGN(key.offset + len, |
14187 |
+- fs_info->sectorsize); |
14188 |
+- } else { |
14189 |
+- len = btrfs_file_extent_num_bytes(src, extent); |
14190 |
+- *last_extent = key.offset + len; |
14191 |
+- } |
14192 |
+- } |
14193 |
+-fill_holes: |
14194 |
+- /* So we did prev_leaf, now we need to move to the next leaf, but a few |
14195 |
+- * things could have happened |
14196 |
+- * |
14197 |
+- * 1) A merge could have happened, so we could currently be on a leaf |
14198 |
+- * that holds what we were copying in the first place. |
14199 |
+- * 2) A split could have happened, and now not all of the items we want |
14200 |
+- * are on the same leaf. |
14201 |
+- * |
14202 |
+- * So we need to adjust how we search for holes, we need to drop the |
14203 |
+- * path and re-search for the first extent key we found, and then walk |
14204 |
+- * forward until we hit the last one we copied. |
14205 |
+- */ |
14206 |
+- if (need_find_last_extent) { |
14207 |
+- /* btrfs_prev_leaf could return 1 without releasing the path */ |
14208 |
+- btrfs_release_path(src_path); |
14209 |
+- ret = btrfs_search_slot(NULL, inode->root, &first_key, |
14210 |
+- src_path, 0, 0); |
14211 |
+- if (ret < 0) |
14212 |
+- return ret; |
14213 |
+- ASSERT(ret == 0); |
14214 |
+- src = src_path->nodes[0]; |
14215 |
+- i = src_path->slots[0]; |
14216 |
+- } else { |
14217 |
+- i = start_slot; |
14218 |
+- } |
14219 |
+- |
14220 |
+- /* |
14221 |
+- * Ok so here we need to go through and fill in any holes we may have |
14222 |
+- * to make sure that holes are punched for those areas in case they had |
14223 |
+- * extents previously. |
14224 |
+- */ |
14225 |
+- while (!done) { |
14226 |
+- u64 offset, len; |
14227 |
+- u64 extent_end; |
14228 |
+- |
14229 |
+- if (i >= btrfs_header_nritems(src_path->nodes[0])) { |
14230 |
+- ret = btrfs_next_leaf(inode->root, src_path); |
14231 |
+- if (ret < 0) |
14232 |
+- return ret; |
14233 |
+- ASSERT(ret == 0); |
14234 |
+- src = src_path->nodes[0]; |
14235 |
+- i = 0; |
14236 |
+- need_find_last_extent = true; |
14237 |
+- } |
14238 |
+- |
14239 |
+- btrfs_item_key_to_cpu(src, &key, i); |
14240 |
+- if (!btrfs_comp_cpu_keys(&key, &last_key)) |
14241 |
+- done = true; |
14242 |
+- if (key.objectid != btrfs_ino(inode) || |
14243 |
+- key.type != BTRFS_EXTENT_DATA_KEY) { |
14244 |
+- i++; |
14245 |
+- continue; |
14246 |
+- } |
14247 |
+- extent = btrfs_item_ptr(src, i, struct btrfs_file_extent_item); |
14248 |
+- if (btrfs_file_extent_type(src, extent) == |
14249 |
+- BTRFS_FILE_EXTENT_INLINE) { |
14250 |
+- len = btrfs_file_extent_ram_bytes(src, extent); |
14251 |
+- extent_end = ALIGN(key.offset + len, |
14252 |
+- fs_info->sectorsize); |
14253 |
+- } else { |
14254 |
+- len = btrfs_file_extent_num_bytes(src, extent); |
14255 |
+- extent_end = key.offset + len; |
14256 |
+- } |
14257 |
+- i++; |
14258 |
+- |
14259 |
+- if (*last_extent == key.offset) { |
14260 |
+- *last_extent = extent_end; |
14261 |
+- continue; |
14262 |
+- } |
14263 |
+- offset = *last_extent; |
14264 |
+- len = key.offset - *last_extent; |
14265 |
+- ret = btrfs_insert_file_extent(trans, log, btrfs_ino(inode), |
14266 |
+- offset, 0, 0, len, 0, len, 0, 0, 0); |
14267 |
+- if (ret) |
14268 |
+- break; |
14269 |
+- *last_extent = extent_end; |
14270 |
+- } |
14271 |
+- |
14272 |
+- /* |
14273 |
+- * Check if there is a hole between the last extent found in our leaf |
14274 |
+- * and the first extent in the next leaf. If there is one, we need to |
14275 |
+- * log an explicit hole so that at replay time we can punch the hole. |
14276 |
+- */ |
14277 |
+- if (ret == 0 && |
14278 |
+- key.objectid == btrfs_ino(inode) && |
14279 |
+- key.type == BTRFS_EXTENT_DATA_KEY && |
14280 |
+- i == btrfs_header_nritems(src_path->nodes[0])) { |
14281 |
+- ret = btrfs_next_leaf(inode->root, src_path); |
14282 |
+- need_find_last_extent = true; |
14283 |
+- if (ret > 0) { |
14284 |
+- ret = 0; |
14285 |
+- } else if (ret == 0) { |
14286 |
+- btrfs_item_key_to_cpu(src_path->nodes[0], &key, |
14287 |
+- src_path->slots[0]); |
14288 |
+- if (key.objectid == btrfs_ino(inode) && |
14289 |
+- key.type == BTRFS_EXTENT_DATA_KEY && |
14290 |
+- *last_extent < key.offset) { |
14291 |
+- const u64 len = key.offset - *last_extent; |
14292 |
+- |
14293 |
+- ret = btrfs_insert_file_extent(trans, log, |
14294 |
+- btrfs_ino(inode), |
14295 |
+- *last_extent, 0, |
14296 |
+- 0, len, 0, len, |
14297 |
+- 0, 0, 0); |
14298 |
+- *last_extent += len; |
14299 |
+- } |
14300 |
+- } |
14301 |
+- } |
14302 |
+- /* |
14303 |
+- * Need to let the callers know we dropped the path so they should |
14304 |
+- * re-search. |
14305 |
+- */ |
14306 |
+- if (!ret && need_find_last_extent) |
14307 |
+- ret = 1; |
14308 |
+ return ret; |
14309 |
+ } |
14310 |
+ |
14311 |
+@@ -4365,7 +4181,7 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans, |
14312 |
+ const u64 i_size = i_size_read(&inode->vfs_inode); |
14313 |
+ const u64 ino = btrfs_ino(inode); |
14314 |
+ struct btrfs_path *dst_path = NULL; |
14315 |
+- u64 last_extent = (u64)-1; |
14316 |
++ bool dropped_extents = false; |
14317 |
+ int ins_nr = 0; |
14318 |
+ int start_slot; |
14319 |
+ int ret; |
14320 |
+@@ -4387,8 +4203,7 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans, |
14321 |
+ if (slot >= btrfs_header_nritems(leaf)) { |
14322 |
+ if (ins_nr > 0) { |
14323 |
+ ret = copy_items(trans, inode, dst_path, path, |
14324 |
+- &last_extent, start_slot, |
14325 |
+- ins_nr, 1, 0); |
14326 |
++ start_slot, ins_nr, 1, 0); |
14327 |
+ if (ret < 0) |
14328 |
+ goto out; |
14329 |
+ ins_nr = 0; |
14330 |
+@@ -4412,8 +4227,7 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans, |
14331 |
+ path->slots[0]++; |
14332 |
+ continue; |
14333 |
+ } |
14334 |
+- if (last_extent == (u64)-1) { |
14335 |
+- last_extent = key.offset; |
14336 |
++ if (!dropped_extents) { |
14337 |
+ /* |
14338 |
+ * Avoid logging extent items logged in past fsync calls |
14339 |
+ * and leading to duplicate keys in the log tree. |
14340 |
+@@ -4427,6 +4241,7 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans, |
14341 |
+ } while (ret == -EAGAIN); |
14342 |
+ if (ret) |
14343 |
+ goto out; |
14344 |
++ dropped_extents = true; |
14345 |
+ } |
14346 |
+ if (ins_nr == 0) |
14347 |
+ start_slot = slot; |
14348 |
+@@ -4441,7 +4256,7 @@ static int btrfs_log_prealloc_extents(struct btrfs_trans_handle *trans, |
14349 |
+ } |
14350 |
+ } |
14351 |
+ if (ins_nr > 0) { |
14352 |
+- ret = copy_items(trans, inode, dst_path, path, &last_extent, |
14353 |
++ ret = copy_items(trans, inode, dst_path, path, |
14354 |
+ start_slot, ins_nr, 1, 0); |
14355 |
+ if (ret > 0) |
14356 |
+ ret = 0; |
14357 |
+@@ -4636,13 +4451,8 @@ static int btrfs_log_all_xattrs(struct btrfs_trans_handle *trans, |
14358 |
+ |
14359 |
+ if (slot >= nritems) { |
14360 |
+ if (ins_nr > 0) { |
14361 |
+- u64 last_extent = 0; |
14362 |
+- |
14363 |
+ ret = copy_items(trans, inode, dst_path, path, |
14364 |
+- &last_extent, start_slot, |
14365 |
+- ins_nr, 1, 0); |
14366 |
+- /* can't be 1, extent items aren't processed */ |
14367 |
+- ASSERT(ret <= 0); |
14368 |
++ start_slot, ins_nr, 1, 0); |
14369 |
+ if (ret < 0) |
14370 |
+ return ret; |
14371 |
+ ins_nr = 0; |
14372 |
+@@ -4666,13 +4476,8 @@ static int btrfs_log_all_xattrs(struct btrfs_trans_handle *trans, |
14373 |
+ cond_resched(); |
14374 |
+ } |
14375 |
+ if (ins_nr > 0) { |
14376 |
+- u64 last_extent = 0; |
14377 |
+- |
14378 |
+ ret = copy_items(trans, inode, dst_path, path, |
14379 |
+- &last_extent, start_slot, |
14380 |
+- ins_nr, 1, 0); |
14381 |
+- /* can't be 1, extent items aren't processed */ |
14382 |
+- ASSERT(ret <= 0); |
14383 |
++ start_slot, ins_nr, 1, 0); |
14384 |
+ if (ret < 0) |
14385 |
+ return ret; |
14386 |
+ } |
14387 |
+@@ -4681,100 +4486,119 @@ static int btrfs_log_all_xattrs(struct btrfs_trans_handle *trans, |
14388 |
+ } |
14389 |
+ |
14390 |
+ /* |
14391 |
+- * If the no holes feature is enabled we need to make sure any hole between the |
14392 |
+- * last extent and the i_size of our inode is explicitly marked in the log. This |
14393 |
+- * is to make sure that doing something like: |
14394 |
+- * |
14395 |
+- * 1) create file with 128Kb of data |
14396 |
+- * 2) truncate file to 64Kb |
14397 |
+- * 3) truncate file to 256Kb |
14398 |
+- * 4) fsync file |
14399 |
+- * 5) <crash/power failure> |
14400 |
+- * 6) mount fs and trigger log replay |
14401 |
+- * |
14402 |
+- * Will give us a file with a size of 256Kb, the first 64Kb of data match what |
14403 |
+- * the file had in its first 64Kb of data at step 1 and the last 192Kb of the |
14404 |
+- * file correspond to a hole. The presence of explicit holes in a log tree is |
14405 |
+- * what guarantees that log replay will remove/adjust file extent items in the |
14406 |
+- * fs/subvol tree. |
14407 |
+- * |
14408 |
+- * Here we do not need to care about holes between extents, that is already done |
14409 |
+- * by copy_items(). We also only need to do this in the full sync path, where we |
14410 |
+- * lookup for extents from the fs/subvol tree only. In the fast path case, we |
14411 |
+- * lookup the list of modified extent maps and if any represents a hole, we |
14412 |
+- * insert a corresponding extent representing a hole in the log tree. |
14413 |
++ * When using the NO_HOLES feature if we punched a hole that causes the |
14414 |
++ * deletion of entire leafs or all the extent items of the first leaf (the one |
14415 |
++ * that contains the inode item and references) we may end up not processing |
14416 |
++ * any extents, because there are no leafs with a generation matching the |
14417 |
++ * current transaction that have extent items for our inode. So we need to find |
14418 |
++ * if any holes exist and then log them. We also need to log holes after any |
14419 |
++ * truncate operation that changes the inode's size. |
14420 |
+ */ |
14421 |
+-static int btrfs_log_trailing_hole(struct btrfs_trans_handle *trans, |
14422 |
+- struct btrfs_root *root, |
14423 |
+- struct btrfs_inode *inode, |
14424 |
+- struct btrfs_path *path) |
14425 |
++static int btrfs_log_holes(struct btrfs_trans_handle *trans, |
14426 |
++ struct btrfs_root *root, |
14427 |
++ struct btrfs_inode *inode, |
14428 |
++ struct btrfs_path *path) |
14429 |
+ { |
14430 |
+ struct btrfs_fs_info *fs_info = root->fs_info; |
14431 |
+- int ret; |
14432 |
+ struct btrfs_key key; |
14433 |
+- u64 hole_start; |
14434 |
+- u64 hole_size; |
14435 |
+- struct extent_buffer *leaf; |
14436 |
+- struct btrfs_root *log = root->log_root; |
14437 |
+ const u64 ino = btrfs_ino(inode); |
14438 |
+ const u64 i_size = i_size_read(&inode->vfs_inode); |
14439 |
++ u64 prev_extent_end = 0; |
14440 |
++ int ret; |
14441 |
+ |
14442 |
+- if (!btrfs_fs_incompat(fs_info, NO_HOLES)) |
14443 |
++ if (!btrfs_fs_incompat(fs_info, NO_HOLES) || i_size == 0) |
14444 |
+ return 0; |
14445 |
+ |
14446 |
+ key.objectid = ino; |
14447 |
+ key.type = BTRFS_EXTENT_DATA_KEY; |
14448 |
+- key.offset = (u64)-1; |
14449 |
++ key.offset = 0; |
14450 |
+ |
14451 |
+ ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
14452 |
+- ASSERT(ret != 0); |
14453 |
+ if (ret < 0) |
14454 |
+ return ret; |
14455 |
+ |
14456 |
+- ASSERT(path->slots[0] > 0); |
14457 |
+- path->slots[0]--; |
14458 |
+- leaf = path->nodes[0]; |
14459 |
+- btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); |
14460 |
+- |
14461 |
+- if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY) { |
14462 |
+- /* inode does not have any extents */ |
14463 |
+- hole_start = 0; |
14464 |
+- hole_size = i_size; |
14465 |
+- } else { |
14466 |
++ while (true) { |
14467 |
+ struct btrfs_file_extent_item *extent; |
14468 |
++ struct extent_buffer *leaf = path->nodes[0]; |
14469 |
+ u64 len; |
14470 |
+ |
14471 |
+- /* |
14472 |
+- * If there's an extent beyond i_size, an explicit hole was |
14473 |
+- * already inserted by copy_items(). |
14474 |
+- */ |
14475 |
+- if (key.offset >= i_size) |
14476 |
+- return 0; |
14477 |
++ if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) { |
14478 |
++ ret = btrfs_next_leaf(root, path); |
14479 |
++ if (ret < 0) |
14480 |
++ return ret; |
14481 |
++ if (ret > 0) { |
14482 |
++ ret = 0; |
14483 |
++ break; |
14484 |
++ } |
14485 |
++ leaf = path->nodes[0]; |
14486 |
++ } |
14487 |
++ |
14488 |
++ btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); |
14489 |
++ if (key.objectid != ino || key.type != BTRFS_EXTENT_DATA_KEY) |
14490 |
++ break; |
14491 |
++ |
14492 |
++ /* We have a hole, log it. */ |
14493 |
++ if (prev_extent_end < key.offset) { |
14494 |
++ const u64 hole_len = key.offset - prev_extent_end; |
14495 |
++ |
14496 |
++ /* |
14497 |
++ * Release the path to avoid deadlocks with other code |
14498 |
++ * paths that search the root while holding locks on |
14499 |
++ * leafs from the log root. |
14500 |
++ */ |
14501 |
++ btrfs_release_path(path); |
14502 |
++ ret = btrfs_insert_file_extent(trans, root->log_root, |
14503 |
++ ino, prev_extent_end, 0, |
14504 |
++ 0, hole_len, 0, hole_len, |
14505 |
++ 0, 0, 0); |
14506 |
++ if (ret < 0) |
14507 |
++ return ret; |
14508 |
++ |
14509 |
++ /* |
14510 |
++ * Search for the same key again in the root. Since it's |
14511 |
++ * an extent item and we are holding the inode lock, the |
14512 |
++ * key must still exist. If it doesn't just emit warning |
14513 |
++ * and return an error to fall back to a transaction |
14514 |
++ * commit. |
14515 |
++ */ |
14516 |
++ ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); |
14517 |
++ if (ret < 0) |
14518 |
++ return ret; |
14519 |
++ if (WARN_ON(ret > 0)) |
14520 |
++ return -ENOENT; |
14521 |
++ leaf = path->nodes[0]; |
14522 |
++ } |
14523 |
+ |
14524 |
+ extent = btrfs_item_ptr(leaf, path->slots[0], |
14525 |
+ struct btrfs_file_extent_item); |
14526 |
+- |
14527 |
+ if (btrfs_file_extent_type(leaf, extent) == |
14528 |
+- BTRFS_FILE_EXTENT_INLINE) |
14529 |
+- return 0; |
14530 |
++ BTRFS_FILE_EXTENT_INLINE) { |
14531 |
++ len = btrfs_file_extent_ram_bytes(leaf, extent); |
14532 |
++ prev_extent_end = ALIGN(key.offset + len, |
14533 |
++ fs_info->sectorsize); |
14534 |
++ } else { |
14535 |
++ len = btrfs_file_extent_num_bytes(leaf, extent); |
14536 |
++ prev_extent_end = key.offset + len; |
14537 |
++ } |
14538 |
+ |
14539 |
+- len = btrfs_file_extent_num_bytes(leaf, extent); |
14540 |
+- /* Last extent goes beyond i_size, no need to log a hole. */ |
14541 |
+- if (key.offset + len > i_size) |
14542 |
+- return 0; |
14543 |
+- hole_start = key.offset + len; |
14544 |
+- hole_size = i_size - hole_start; |
14545 |
++ path->slots[0]++; |
14546 |
++ cond_resched(); |
14547 |
+ } |
14548 |
+- btrfs_release_path(path); |
14549 |
+ |
14550 |
+- /* Last extent ends at i_size. */ |
14551 |
+- if (hole_size == 0) |
14552 |
+- return 0; |
14553 |
++ if (prev_extent_end < i_size) { |
14554 |
++ u64 hole_len; |
14555 |
+ |
14556 |
+- hole_size = ALIGN(hole_size, fs_info->sectorsize); |
14557 |
+- ret = btrfs_insert_file_extent(trans, log, ino, hole_start, 0, 0, |
14558 |
+- hole_size, 0, hole_size, 0, 0, 0); |
14559 |
+- return ret; |
14560 |
++ btrfs_release_path(path); |
14561 |
++ hole_len = ALIGN(i_size - prev_extent_end, fs_info->sectorsize); |
14562 |
++ ret = btrfs_insert_file_extent(trans, root->log_root, |
14563 |
++ ino, prev_extent_end, 0, 0, |
14564 |
++ hole_len, 0, hole_len, |
14565 |
++ 0, 0, 0); |
14566 |
++ if (ret < 0) |
14567 |
++ return ret; |
14568 |
++ } |
14569 |
++ |
14570 |
++ return 0; |
14571 |
+ } |
14572 |
+ |
14573 |
+ /* |
14574 |
+@@ -4934,7 +4758,6 @@ static int btrfs_log_inode(struct btrfs_trans_handle *trans, |
14575 |
+ struct btrfs_key min_key; |
14576 |
+ struct btrfs_key max_key; |
14577 |
+ struct btrfs_root *log = root->log_root; |
14578 |
+- u64 last_extent = 0; |
14579 |
+ int err = 0; |
14580 |
+ int ret; |
14581 |
+ int nritems; |
14582 |
+@@ -5108,7 +4931,7 @@ again: |
14583 |
+ ins_start_slot = path->slots[0]; |
14584 |
+ } |
14585 |
+ ret = copy_items(trans, inode, dst_path, path, |
14586 |
+- &last_extent, ins_start_slot, |
14587 |
++ ins_start_slot, |
14588 |
+ ins_nr, inode_only, |
14589 |
+ logged_isize); |
14590 |
+ if (ret < 0) { |
14591 |
+@@ -5161,17 +4984,13 @@ again: |
14592 |
+ if (ins_nr == 0) |
14593 |
+ goto next_slot; |
14594 |
+ ret = copy_items(trans, inode, dst_path, path, |
14595 |
+- &last_extent, ins_start_slot, |
14596 |
++ ins_start_slot, |
14597 |
+ ins_nr, inode_only, logged_isize); |
14598 |
+ if (ret < 0) { |
14599 |
+ err = ret; |
14600 |
+ goto out_unlock; |
14601 |
+ } |
14602 |
+ ins_nr = 0; |
14603 |
+- if (ret) { |
14604 |
+- btrfs_release_path(path); |
14605 |
+- continue; |
14606 |
+- } |
14607 |
+ goto next_slot; |
14608 |
+ } |
14609 |
+ |
14610 |
+@@ -5184,18 +5003,13 @@ again: |
14611 |
+ goto next_slot; |
14612 |
+ } |
14613 |
+ |
14614 |
+- ret = copy_items(trans, inode, dst_path, path, &last_extent, |
14615 |
++ ret = copy_items(trans, inode, dst_path, path, |
14616 |
+ ins_start_slot, ins_nr, inode_only, |
14617 |
+ logged_isize); |
14618 |
+ if (ret < 0) { |
14619 |
+ err = ret; |
14620 |
+ goto out_unlock; |
14621 |
+ } |
14622 |
+- if (ret) { |
14623 |
+- ins_nr = 0; |
14624 |
+- btrfs_release_path(path); |
14625 |
+- continue; |
14626 |
+- } |
14627 |
+ ins_nr = 1; |
14628 |
+ ins_start_slot = path->slots[0]; |
14629 |
+ next_slot: |
14630 |
+@@ -5209,13 +5023,12 @@ next_slot: |
14631 |
+ } |
14632 |
+ if (ins_nr) { |
14633 |
+ ret = copy_items(trans, inode, dst_path, path, |
14634 |
+- &last_extent, ins_start_slot, |
14635 |
++ ins_start_slot, |
14636 |
+ ins_nr, inode_only, logged_isize); |
14637 |
+ if (ret < 0) { |
14638 |
+ err = ret; |
14639 |
+ goto out_unlock; |
14640 |
+ } |
14641 |
+- ret = 0; |
14642 |
+ ins_nr = 0; |
14643 |
+ } |
14644 |
+ btrfs_release_path(path); |
14645 |
+@@ -5230,14 +5043,13 @@ next_key: |
14646 |
+ } |
14647 |
+ } |
14648 |
+ if (ins_nr) { |
14649 |
+- ret = copy_items(trans, inode, dst_path, path, &last_extent, |
14650 |
++ ret = copy_items(trans, inode, dst_path, path, |
14651 |
+ ins_start_slot, ins_nr, inode_only, |
14652 |
+ logged_isize); |
14653 |
+ if (ret < 0) { |
14654 |
+ err = ret; |
14655 |
+ goto out_unlock; |
14656 |
+ } |
14657 |
+- ret = 0; |
14658 |
+ ins_nr = 0; |
14659 |
+ } |
14660 |
+ |
14661 |
+@@ -5250,7 +5062,7 @@ next_key: |
14662 |
+ if (max_key.type >= BTRFS_EXTENT_DATA_KEY && !fast_search) { |
14663 |
+ btrfs_release_path(path); |
14664 |
+ btrfs_release_path(dst_path); |
14665 |
+- err = btrfs_log_trailing_hole(trans, root, inode, path); |
14666 |
++ err = btrfs_log_holes(trans, root, inode, path); |
14667 |
+ if (err) |
14668 |
+ goto out_unlock; |
14669 |
+ } |
14670 |
+diff --git a/fs/cifs/smb2pdu.c b/fs/cifs/smb2pdu.c |
14671 |
+index 43f29621e51f..0d4e4d97e6cf 100644 |
14672 |
+--- a/fs/cifs/smb2pdu.c |
14673 |
++++ b/fs/cifs/smb2pdu.c |
14674 |
+@@ -259,9 +259,14 @@ smb2_reconnect(__le16 smb2_command, struct cifs_tcon *tcon) |
14675 |
+ } |
14676 |
+ |
14677 |
+ rc = cifs_negotiate_protocol(0, tcon->ses); |
14678 |
+- if (!rc && tcon->ses->need_reconnect) |
14679 |
++ if (!rc && tcon->ses->need_reconnect) { |
14680 |
+ rc = cifs_setup_session(0, tcon->ses, nls_codepage); |
14681 |
+- |
14682 |
++ if ((rc == -EACCES) && !tcon->retry) { |
14683 |
++ rc = -EHOSTDOWN; |
14684 |
++ mutex_unlock(&tcon->ses->session_mutex); |
14685 |
++ goto failed; |
14686 |
++ } |
14687 |
++ } |
14688 |
+ if (rc || !tcon->need_reconnect) { |
14689 |
+ mutex_unlock(&tcon->ses->session_mutex); |
14690 |
+ goto out; |
14691 |
+@@ -306,6 +311,7 @@ out: |
14692 |
+ case SMB2_SET_INFO: |
14693 |
+ rc = -EAGAIN; |
14694 |
+ } |
14695 |
++failed: |
14696 |
+ unload_nls(nls_codepage); |
14697 |
+ return rc; |
14698 |
+ } |
14699 |
+@@ -3130,8 +3136,8 @@ smb2_readv_callback(struct mid_q_entry *mid) |
14700 |
+ struct smb2_sync_hdr *shdr = |
14701 |
+ (struct smb2_sync_hdr *)rdata->iov[0].iov_base; |
14702 |
+ unsigned int credits_received = 0; |
14703 |
+- struct smb_rqst rqst = { .rq_iov = rdata->iov, |
14704 |
+- .rq_nvec = 2, |
14705 |
++ struct smb_rqst rqst = { .rq_iov = &rdata->iov[1], |
14706 |
++ .rq_nvec = 1, |
14707 |
+ .rq_pages = rdata->pages, |
14708 |
+ .rq_offset = rdata->page_offset, |
14709 |
+ .rq_npages = rdata->nr_pages, |
14710 |
+diff --git a/fs/eventfd.c b/fs/eventfd.c |
14711 |
+index 08d3bd602f73..ce1d1711fbba 100644 |
14712 |
+--- a/fs/eventfd.c |
14713 |
++++ b/fs/eventfd.c |
14714 |
+@@ -22,6 +22,8 @@ |
14715 |
+ #include <linux/proc_fs.h> |
14716 |
+ #include <linux/seq_file.h> |
14717 |
+ |
14718 |
++DEFINE_PER_CPU(int, eventfd_wake_count); |
14719 |
++ |
14720 |
+ struct eventfd_ctx { |
14721 |
+ struct kref kref; |
14722 |
+ wait_queue_head_t wqh; |
14723 |
+@@ -55,12 +57,25 @@ __u64 eventfd_signal(struct eventfd_ctx *ctx, __u64 n) |
14724 |
+ { |
14725 |
+ unsigned long flags; |
14726 |
+ |
14727 |
++ /* |
14728 |
++ * Deadlock or stack overflow issues can happen if we recurse here |
14729 |
++ * through waitqueue wakeup handlers. If the caller users potentially |
14730 |
++ * nested waitqueues with custom wakeup handlers, then it should |
14731 |
++ * check eventfd_signal_count() before calling this function. If |
14732 |
++ * it returns true, the eventfd_signal() call should be deferred to a |
14733 |
++ * safe context. |
14734 |
++ */ |
14735 |
++ if (WARN_ON_ONCE(this_cpu_read(eventfd_wake_count))) |
14736 |
++ return 0; |
14737 |
++ |
14738 |
+ spin_lock_irqsave(&ctx->wqh.lock, flags); |
14739 |
++ this_cpu_inc(eventfd_wake_count); |
14740 |
+ if (ULLONG_MAX - ctx->count < n) |
14741 |
+ n = ULLONG_MAX - ctx->count; |
14742 |
+ ctx->count += n; |
14743 |
+ if (waitqueue_active(&ctx->wqh)) |
14744 |
+ wake_up_locked_poll(&ctx->wqh, EPOLLIN); |
14745 |
++ this_cpu_dec(eventfd_wake_count); |
14746 |
+ spin_unlock_irqrestore(&ctx->wqh.lock, flags); |
14747 |
+ |
14748 |
+ return n; |
14749 |
+diff --git a/fs/ext2/super.c b/fs/ext2/super.c |
14750 |
+index 364e647d87c0..80a3038e0e46 100644 |
14751 |
+--- a/fs/ext2/super.c |
14752 |
++++ b/fs/ext2/super.c |
14753 |
+@@ -1093,9 +1093,9 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent) |
14754 |
+ |
14755 |
+ if (EXT2_BLOCKS_PER_GROUP(sb) == 0) |
14756 |
+ goto cantfind_ext2; |
14757 |
+- sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) - |
14758 |
+- le32_to_cpu(es->s_first_data_block) - 1) |
14759 |
+- / EXT2_BLOCKS_PER_GROUP(sb)) + 1; |
14760 |
++ sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) - |
14761 |
++ le32_to_cpu(es->s_first_data_block) - 1) |
14762 |
++ / EXT2_BLOCKS_PER_GROUP(sb)) + 1; |
14763 |
+ db_count = (sbi->s_groups_count + EXT2_DESC_PER_BLOCK(sb) - 1) / |
14764 |
+ EXT2_DESC_PER_BLOCK(sb); |
14765 |
+ sbi->s_group_desc = kmalloc_array (db_count, |
14766 |
+diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c |
14767 |
+index db7590178dfc..9cc79b7b0df1 100644 |
14768 |
+--- a/fs/ext4/page-io.c |
14769 |
++++ b/fs/ext4/page-io.c |
14770 |
+@@ -481,17 +481,26 @@ int ext4_bio_write_page(struct ext4_io_submit *io, |
14771 |
+ nr_to_submit) { |
14772 |
+ gfp_t gfp_flags = GFP_NOFS; |
14773 |
+ |
14774 |
++ /* |
14775 |
++ * Since bounce page allocation uses a mempool, we can only use |
14776 |
++ * a waiting mask (i.e. request guaranteed allocation) on the |
14777 |
++ * first page of the bio. Otherwise it can deadlock. |
14778 |
++ */ |
14779 |
++ if (io->io_bio) |
14780 |
++ gfp_flags = GFP_NOWAIT | __GFP_NOWARN; |
14781 |
+ retry_encrypt: |
14782 |
+ data_page = fscrypt_encrypt_page(inode, page, PAGE_SIZE, 0, |
14783 |
+ page->index, gfp_flags); |
14784 |
+ if (IS_ERR(data_page)) { |
14785 |
+ ret = PTR_ERR(data_page); |
14786 |
+- if (ret == -ENOMEM && wbc->sync_mode == WB_SYNC_ALL) { |
14787 |
+- if (io->io_bio) { |
14788 |
++ if (ret == -ENOMEM && |
14789 |
++ (io->io_bio || wbc->sync_mode == WB_SYNC_ALL)) { |
14790 |
++ gfp_flags = GFP_NOFS; |
14791 |
++ if (io->io_bio) |
14792 |
+ ext4_io_submit(io); |
14793 |
+- congestion_wait(BLK_RW_ASYNC, HZ/50); |
14794 |
+- } |
14795 |
+- gfp_flags |= __GFP_NOFAIL; |
14796 |
++ else |
14797 |
++ gfp_flags |= __GFP_NOFAIL; |
14798 |
++ congestion_wait(BLK_RW_ASYNC, HZ/50); |
14799 |
+ goto retry_encrypt; |
14800 |
+ } |
14801 |
+ data_page = NULL; |
14802 |
+diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c |
14803 |
+index 7a9cc64f5ca3..da348cf4ff56 100644 |
14804 |
+--- a/fs/f2fs/super.c |
14805 |
++++ b/fs/f2fs/super.c |
14806 |
+@@ -1148,9 +1148,11 @@ static int f2fs_statfs_project(struct super_block *sb, |
14807 |
+ return PTR_ERR(dquot); |
14808 |
+ spin_lock(&dquot->dq_dqb_lock); |
14809 |
+ |
14810 |
+- limit = (dquot->dq_dqb.dqb_bsoftlimit ? |
14811 |
+- dquot->dq_dqb.dqb_bsoftlimit : |
14812 |
+- dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; |
14813 |
++ limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, |
14814 |
++ dquot->dq_dqb.dqb_bhardlimit); |
14815 |
++ if (limit) |
14816 |
++ limit >>= sb->s_blocksize_bits; |
14817 |
++ |
14818 |
+ if (limit && buf->f_blocks > limit) { |
14819 |
+ curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; |
14820 |
+ buf->f_blocks = limit; |
14821 |
+@@ -1159,9 +1161,9 @@ static int f2fs_statfs_project(struct super_block *sb, |
14822 |
+ (buf->f_blocks - curblock) : 0; |
14823 |
+ } |
14824 |
+ |
14825 |
+- limit = dquot->dq_dqb.dqb_isoftlimit ? |
14826 |
+- dquot->dq_dqb.dqb_isoftlimit : |
14827 |
+- dquot->dq_dqb.dqb_ihardlimit; |
14828 |
++ limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, |
14829 |
++ dquot->dq_dqb.dqb_ihardlimit); |
14830 |
++ |
14831 |
+ if (limit && buf->f_files > limit) { |
14832 |
+ buf->f_files = limit; |
14833 |
+ buf->f_ffree = |
14834 |
+diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c |
14835 |
+index 08369c6cd127..143e7d518c5d 100644 |
14836 |
+--- a/fs/gfs2/file.c |
14837 |
++++ b/fs/gfs2/file.c |
14838 |
+@@ -780,7 +780,7 @@ static ssize_t gfs2_file_write_iter(struct kiocb *iocb, struct iov_iter *from) |
14839 |
+ struct file *file = iocb->ki_filp; |
14840 |
+ struct inode *inode = file_inode(file); |
14841 |
+ struct gfs2_inode *ip = GFS2_I(inode); |
14842 |
+- ssize_t written = 0, ret; |
14843 |
++ ssize_t ret; |
14844 |
+ |
14845 |
+ ret = gfs2_rsqa_alloc(ip); |
14846 |
+ if (ret) |
14847 |
+@@ -800,68 +800,58 @@ static ssize_t gfs2_file_write_iter(struct kiocb *iocb, struct iov_iter *from) |
14848 |
+ inode_lock(inode); |
14849 |
+ ret = generic_write_checks(iocb, from); |
14850 |
+ if (ret <= 0) |
14851 |
+- goto out; |
14852 |
+- |
14853 |
+- /* We can write back this queue in page reclaim */ |
14854 |
+- current->backing_dev_info = inode_to_bdi(inode); |
14855 |
++ goto out_unlock; |
14856 |
+ |
14857 |
+ ret = file_remove_privs(file); |
14858 |
+ if (ret) |
14859 |
+- goto out2; |
14860 |
++ goto out_unlock; |
14861 |
+ |
14862 |
+ ret = file_update_time(file); |
14863 |
+ if (ret) |
14864 |
+- goto out2; |
14865 |
++ goto out_unlock; |
14866 |
+ |
14867 |
+ if (iocb->ki_flags & IOCB_DIRECT) { |
14868 |
+ struct address_space *mapping = file->f_mapping; |
14869 |
+- loff_t pos, endbyte; |
14870 |
+- ssize_t buffered; |
14871 |
++ ssize_t buffered, ret2; |
14872 |
+ |
14873 |
+- written = gfs2_file_direct_write(iocb, from); |
14874 |
+- if (written < 0 || !iov_iter_count(from)) |
14875 |
+- goto out2; |
14876 |
++ ret = gfs2_file_direct_write(iocb, from); |
14877 |
++ if (ret < 0 || !iov_iter_count(from)) |
14878 |
++ goto out_unlock; |
14879 |
+ |
14880 |
+- ret = iomap_file_buffered_write(iocb, from, &gfs2_iomap_ops); |
14881 |
+- if (unlikely(ret < 0)) |
14882 |
+- goto out2; |
14883 |
+- buffered = ret; |
14884 |
++ iocb->ki_flags |= IOCB_DSYNC; |
14885 |
++ current->backing_dev_info = inode_to_bdi(inode); |
14886 |
++ buffered = iomap_file_buffered_write(iocb, from, &gfs2_iomap_ops); |
14887 |
++ current->backing_dev_info = NULL; |
14888 |
++ if (unlikely(buffered <= 0)) |
14889 |
++ goto out_unlock; |
14890 |
+ |
14891 |
+ /* |
14892 |
+ * We need to ensure that the page cache pages are written to |
14893 |
+ * disk and invalidated to preserve the expected O_DIRECT |
14894 |
+- * semantics. |
14895 |
++ * semantics. If the writeback or invalidate fails, only report |
14896 |
++ * the direct I/O range as we don't know if the buffered pages |
14897 |
++ * made it to disk. |
14898 |
+ */ |
14899 |
+- pos = iocb->ki_pos; |
14900 |
+- endbyte = pos + buffered - 1; |
14901 |
+- ret = filemap_write_and_wait_range(mapping, pos, endbyte); |
14902 |
+- if (!ret) { |
14903 |
+- iocb->ki_pos += buffered; |
14904 |
+- written += buffered; |
14905 |
+- invalidate_mapping_pages(mapping, |
14906 |
+- pos >> PAGE_SHIFT, |
14907 |
+- endbyte >> PAGE_SHIFT); |
14908 |
+- } else { |
14909 |
+- /* |
14910 |
+- * We don't know how much we wrote, so just return |
14911 |
+- * the number of bytes which were direct-written |
14912 |
+- */ |
14913 |
+- } |
14914 |
++ iocb->ki_pos += buffered; |
14915 |
++ ret2 = generic_write_sync(iocb, buffered); |
14916 |
++ invalidate_mapping_pages(mapping, |
14917 |
++ (iocb->ki_pos - buffered) >> PAGE_SHIFT, |
14918 |
++ (iocb->ki_pos - 1) >> PAGE_SHIFT); |
14919 |
++ if (!ret || ret2 > 0) |
14920 |
++ ret += ret2; |
14921 |
+ } else { |
14922 |
++ current->backing_dev_info = inode_to_bdi(inode); |
14923 |
+ ret = iomap_file_buffered_write(iocb, from, &gfs2_iomap_ops); |
14924 |
+- if (likely(ret > 0)) |
14925 |
++ current->backing_dev_info = NULL; |
14926 |
++ if (likely(ret > 0)) { |
14927 |
+ iocb->ki_pos += ret; |
14928 |
++ ret = generic_write_sync(iocb, ret); |
14929 |
++ } |
14930 |
+ } |
14931 |
+ |
14932 |
+-out2: |
14933 |
+- current->backing_dev_info = NULL; |
14934 |
+-out: |
14935 |
++out_unlock: |
14936 |
+ inode_unlock(inode); |
14937 |
+- if (likely(ret > 0)) { |
14938 |
+- /* Handle various SYNC-type writes */ |
14939 |
+- ret = generic_write_sync(iocb, ret); |
14940 |
+- } |
14941 |
+- return written ? written : ret; |
14942 |
++ return ret; |
14943 |
+ } |
14944 |
+ |
14945 |
+ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len, |
14946 |
+diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c |
14947 |
+index df390a69c49a..1a2339f2cb49 100644 |
14948 |
+--- a/fs/jbd2/journal.c |
14949 |
++++ b/fs/jbd2/journal.c |
14950 |
+@@ -1002,6 +1002,7 @@ static void *jbd2_seq_info_start(struct seq_file *seq, loff_t *pos) |
14951 |
+ |
14952 |
+ static void *jbd2_seq_info_next(struct seq_file *seq, void *v, loff_t *pos) |
14953 |
+ { |
14954 |
++ (*pos)++; |
14955 |
+ return NULL; |
14956 |
+ } |
14957 |
+ |
14958 |
+diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c |
14959 |
+index b8d686087952..4ae726e70d87 100644 |
14960 |
+--- a/fs/nfs/dir.c |
14961 |
++++ b/fs/nfs/dir.c |
14962 |
+@@ -162,6 +162,17 @@ typedef struct { |
14963 |
+ bool eof; |
14964 |
+ } nfs_readdir_descriptor_t; |
14965 |
+ |
14966 |
++static |
14967 |
++void nfs_readdir_init_array(struct page *page) |
14968 |
++{ |
14969 |
++ struct nfs_cache_array *array; |
14970 |
++ |
14971 |
++ array = kmap_atomic(page); |
14972 |
++ memset(array, 0, sizeof(struct nfs_cache_array)); |
14973 |
++ array->eof_index = -1; |
14974 |
++ kunmap_atomic(array); |
14975 |
++} |
14976 |
++ |
14977 |
+ /* |
14978 |
+ * we are freeing strings created by nfs_add_to_readdir_array() |
14979 |
+ */ |
14980 |
+@@ -174,6 +185,7 @@ void nfs_readdir_clear_array(struct page *page) |
14981 |
+ array = kmap_atomic(page); |
14982 |
+ for (i = 0; i < array->size; i++) |
14983 |
+ kfree(array->array[i].string.name); |
14984 |
++ array->size = 0; |
14985 |
+ kunmap_atomic(array); |
14986 |
+ } |
14987 |
+ |
14988 |
+@@ -610,6 +622,8 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, |
14989 |
+ int status = -ENOMEM; |
14990 |
+ unsigned int array_size = ARRAY_SIZE(pages); |
14991 |
+ |
14992 |
++ nfs_readdir_init_array(page); |
14993 |
++ |
14994 |
+ entry.prev_cookie = 0; |
14995 |
+ entry.cookie = desc->last_cookie; |
14996 |
+ entry.eof = 0; |
14997 |
+@@ -626,8 +640,6 @@ int nfs_readdir_xdr_to_array(nfs_readdir_descriptor_t *desc, struct page *page, |
14998 |
+ } |
14999 |
+ |
15000 |
+ array = kmap(page); |
15001 |
+- memset(array, 0, sizeof(struct nfs_cache_array)); |
15002 |
+- array->eof_index = -1; |
15003 |
+ |
15004 |
+ status = nfs_readdir_alloc_pages(pages, array_size); |
15005 |
+ if (status < 0) |
15006 |
+@@ -681,6 +693,7 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page) |
15007 |
+ unlock_page(page); |
15008 |
+ return 0; |
15009 |
+ error: |
15010 |
++ nfs_readdir_clear_array(page); |
15011 |
+ unlock_page(page); |
15012 |
+ return ret; |
15013 |
+ } |
15014 |
+@@ -688,8 +701,6 @@ int nfs_readdir_filler(nfs_readdir_descriptor_t *desc, struct page* page) |
15015 |
+ static |
15016 |
+ void cache_page_release(nfs_readdir_descriptor_t *desc) |
15017 |
+ { |
15018 |
+- if (!desc->page->mapping) |
15019 |
+- nfs_readdir_clear_array(desc->page); |
15020 |
+ put_page(desc->page); |
15021 |
+ desc->page = NULL; |
15022 |
+ } |
15023 |
+@@ -703,19 +714,28 @@ struct page *get_cache_page(nfs_readdir_descriptor_t *desc) |
15024 |
+ |
15025 |
+ /* |
15026 |
+ * Returns 0 if desc->dir_cookie was found on page desc->page_index |
15027 |
++ * and locks the page to prevent removal from the page cache. |
15028 |
+ */ |
15029 |
+ static |
15030 |
+-int find_cache_page(nfs_readdir_descriptor_t *desc) |
15031 |
++int find_and_lock_cache_page(nfs_readdir_descriptor_t *desc) |
15032 |
+ { |
15033 |
+ int res; |
15034 |
+ |
15035 |
+ desc->page = get_cache_page(desc); |
15036 |
+ if (IS_ERR(desc->page)) |
15037 |
+ return PTR_ERR(desc->page); |
15038 |
+- |
15039 |
+- res = nfs_readdir_search_array(desc); |
15040 |
++ res = lock_page_killable(desc->page); |
15041 |
+ if (res != 0) |
15042 |
+- cache_page_release(desc); |
15043 |
++ goto error; |
15044 |
++ res = -EAGAIN; |
15045 |
++ if (desc->page->mapping != NULL) { |
15046 |
++ res = nfs_readdir_search_array(desc); |
15047 |
++ if (res == 0) |
15048 |
++ return 0; |
15049 |
++ } |
15050 |
++ unlock_page(desc->page); |
15051 |
++error: |
15052 |
++ cache_page_release(desc); |
15053 |
+ return res; |
15054 |
+ } |
15055 |
+ |
15056 |
+@@ -730,7 +750,7 @@ int readdir_search_pagecache(nfs_readdir_descriptor_t *desc) |
15057 |
+ desc->last_cookie = 0; |
15058 |
+ } |
15059 |
+ do { |
15060 |
+- res = find_cache_page(desc); |
15061 |
++ res = find_and_lock_cache_page(desc); |
15062 |
+ } while (res == -EAGAIN); |
15063 |
+ return res; |
15064 |
+ } |
15065 |
+@@ -769,7 +789,6 @@ int nfs_do_filldir(nfs_readdir_descriptor_t *desc) |
15066 |
+ desc->eof = true; |
15067 |
+ |
15068 |
+ kunmap(desc->page); |
15069 |
+- cache_page_release(desc); |
15070 |
+ dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ cookie %Lu; returning = %d\n", |
15071 |
+ (unsigned long long)*desc->dir_cookie, res); |
15072 |
+ return res; |
15073 |
+@@ -815,13 +834,13 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc) |
15074 |
+ |
15075 |
+ status = nfs_do_filldir(desc); |
15076 |
+ |
15077 |
++ out_release: |
15078 |
++ nfs_readdir_clear_array(desc->page); |
15079 |
++ cache_page_release(desc); |
15080 |
+ out: |
15081 |
+ dfprintk(DIRCACHE, "NFS: %s: returns %d\n", |
15082 |
+ __func__, status); |
15083 |
+ return status; |
15084 |
+- out_release: |
15085 |
+- cache_page_release(desc); |
15086 |
+- goto out; |
15087 |
+ } |
15088 |
+ |
15089 |
+ /* The file offset position represents the dirent entry number. A |
15090 |
+@@ -886,6 +905,8 @@ static int nfs_readdir(struct file *file, struct dir_context *ctx) |
15091 |
+ break; |
15092 |
+ |
15093 |
+ res = nfs_do_filldir(desc); |
15094 |
++ unlock_page(desc->page); |
15095 |
++ cache_page_release(desc); |
15096 |
+ if (res < 0) |
15097 |
+ break; |
15098 |
+ } while (!desc->eof); |
15099 |
+diff --git a/fs/nfsd/nfs4layouts.c b/fs/nfsd/nfs4layouts.c |
15100 |
+index 2b36aa037ce0..f4cf1c0793c6 100644 |
15101 |
+--- a/fs/nfsd/nfs4layouts.c |
15102 |
++++ b/fs/nfsd/nfs4layouts.c |
15103 |
+@@ -676,7 +676,7 @@ nfsd4_cb_layout_done(struct nfsd4_callback *cb, struct rpc_task *task) |
15104 |
+ |
15105 |
+ /* Client gets 2 lease periods to return it */ |
15106 |
+ cutoff = ktime_add_ns(task->tk_start, |
15107 |
+- nn->nfsd4_lease * NSEC_PER_SEC * 2); |
15108 |
++ (u64)nn->nfsd4_lease * NSEC_PER_SEC * 2); |
15109 |
+ |
15110 |
+ if (ktime_before(now, cutoff)) { |
15111 |
+ rpc_delay(task, HZ/100); /* 10 mili-seconds */ |
15112 |
+diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c |
15113 |
+index c8ce128e0054..ed73e86194fa 100644 |
15114 |
+--- a/fs/nfsd/nfs4state.c |
15115 |
++++ b/fs/nfsd/nfs4state.c |
15116 |
+@@ -6075,7 +6075,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, |
15117 |
+ } |
15118 |
+ |
15119 |
+ if (fl_flags & FL_SLEEP) { |
15120 |
+- nbl->nbl_time = jiffies; |
15121 |
++ nbl->nbl_time = get_seconds(); |
15122 |
+ spin_lock(&nn->blocked_locks_lock); |
15123 |
+ list_add_tail(&nbl->nbl_list, &lock_sop->lo_blocked); |
15124 |
+ list_add_tail(&nbl->nbl_lru, &nn->blocked_locks_lru); |
15125 |
+diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h |
15126 |
+index 0f07ad6dc1ef..594fc1a8b672 100644 |
15127 |
+--- a/fs/nfsd/state.h |
15128 |
++++ b/fs/nfsd/state.h |
15129 |
+@@ -592,7 +592,7 @@ static inline bool nfsd4_stateid_generation_after(stateid_t *a, stateid_t *b) |
15130 |
+ struct nfsd4_blocked_lock { |
15131 |
+ struct list_head nbl_list; |
15132 |
+ struct list_head nbl_lru; |
15133 |
+- unsigned long nbl_time; |
15134 |
++ time_t nbl_time; |
15135 |
+ struct file_lock nbl_lock; |
15136 |
+ struct knfsd_fh nbl_fh; |
15137 |
+ struct nfsd4_callback nbl_cb; |
15138 |
+diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c |
15139 |
+index 4fe8db314950..80cededcd10d 100644 |
15140 |
+--- a/fs/nfsd/vfs.c |
15141 |
++++ b/fs/nfsd/vfs.c |
15142 |
+@@ -1016,6 +1016,7 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file, |
15143 |
+ host_err = vfs_iter_write(file, &iter, &pos, flags); |
15144 |
+ if (host_err < 0) |
15145 |
+ goto out_nfserr; |
15146 |
++ *cnt = host_err; |
15147 |
+ nfsdstats.io_write += *cnt; |
15148 |
+ fsnotify_modify(file); |
15149 |
+ |
15150 |
+diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c |
15151 |
+index a3e077fcfeb9..fbd70111a2f1 100644 |
15152 |
+--- a/fs/ocfs2/file.c |
15153 |
++++ b/fs/ocfs2/file.c |
15154 |
+@@ -2109,17 +2109,15 @@ static int ocfs2_is_io_unaligned(struct inode *inode, size_t count, loff_t pos) |
15155 |
+ static int ocfs2_inode_lock_for_extent_tree(struct inode *inode, |
15156 |
+ struct buffer_head **di_bh, |
15157 |
+ int meta_level, |
15158 |
+- int overwrite_io, |
15159 |
+ int write_sem, |
15160 |
+ int wait) |
15161 |
+ { |
15162 |
+ int ret = 0; |
15163 |
+ |
15164 |
+ if (wait) |
15165 |
+- ret = ocfs2_inode_lock(inode, NULL, meta_level); |
15166 |
++ ret = ocfs2_inode_lock(inode, di_bh, meta_level); |
15167 |
+ else |
15168 |
+- ret = ocfs2_try_inode_lock(inode, |
15169 |
+- overwrite_io ? NULL : di_bh, meta_level); |
15170 |
++ ret = ocfs2_try_inode_lock(inode, di_bh, meta_level); |
15171 |
+ if (ret < 0) |
15172 |
+ goto out; |
15173 |
+ |
15174 |
+@@ -2144,6 +2142,7 @@ static int ocfs2_inode_lock_for_extent_tree(struct inode *inode, |
15175 |
+ |
15176 |
+ out_unlock: |
15177 |
+ brelse(*di_bh); |
15178 |
++ *di_bh = NULL; |
15179 |
+ ocfs2_inode_unlock(inode, meta_level); |
15180 |
+ out: |
15181 |
+ return ret; |
15182 |
+@@ -2186,7 +2185,6 @@ static int ocfs2_prepare_inode_for_write(struct file *file, |
15183 |
+ ret = ocfs2_inode_lock_for_extent_tree(inode, |
15184 |
+ &di_bh, |
15185 |
+ meta_level, |
15186 |
+- overwrite_io, |
15187 |
+ write_sem, |
15188 |
+ wait); |
15189 |
+ if (ret < 0) { |
15190 |
+@@ -2244,13 +2242,13 @@ static int ocfs2_prepare_inode_for_write(struct file *file, |
15191 |
+ &di_bh, |
15192 |
+ meta_level, |
15193 |
+ write_sem); |
15194 |
++ meta_level = 1; |
15195 |
++ write_sem = 1; |
15196 |
+ ret = ocfs2_inode_lock_for_extent_tree(inode, |
15197 |
+ &di_bh, |
15198 |
+ meta_level, |
15199 |
+- overwrite_io, |
15200 |
+- 1, |
15201 |
++ write_sem, |
15202 |
+ wait); |
15203 |
+- write_sem = 1; |
15204 |
+ if (ret < 0) { |
15205 |
+ if (ret != -EAGAIN) |
15206 |
+ mlog_errno(ret); |
15207 |
+diff --git a/fs/overlayfs/file.c b/fs/overlayfs/file.c |
15208 |
+index 0bd276e4ccbe..fa5ac5de807c 100644 |
15209 |
+--- a/fs/overlayfs/file.c |
15210 |
++++ b/fs/overlayfs/file.c |
15211 |
+@@ -149,7 +149,7 @@ static loff_t ovl_llseek(struct file *file, loff_t offset, int whence) |
15212 |
+ struct inode *inode = file_inode(file); |
15213 |
+ struct fd real; |
15214 |
+ const struct cred *old_cred; |
15215 |
+- ssize_t ret; |
15216 |
++ loff_t ret; |
15217 |
+ |
15218 |
+ /* |
15219 |
+ * The two special cases below do not need to involve real fs, |
15220 |
+diff --git a/fs/overlayfs/readdir.c b/fs/overlayfs/readdir.c |
15221 |
+index cc8303a806b4..11b7941c5dbc 100644 |
15222 |
+--- a/fs/overlayfs/readdir.c |
15223 |
++++ b/fs/overlayfs/readdir.c |
15224 |
+@@ -507,7 +507,13 @@ get: |
15225 |
+ if (err) |
15226 |
+ goto fail; |
15227 |
+ |
15228 |
+- WARN_ON_ONCE(dir->d_sb->s_dev != stat.dev); |
15229 |
++ /* |
15230 |
++ * Directory inode is always on overlay st_dev. |
15231 |
++ * Non-dir with ovl_same_dev() could be on pseudo st_dev in case |
15232 |
++ * of xino bits overflow. |
15233 |
++ */ |
15234 |
++ WARN_ON_ONCE(S_ISDIR(stat.mode) && |
15235 |
++ dir->d_sb->s_dev != stat.dev); |
15236 |
+ ino = stat.ino; |
15237 |
+ } else if (xinobits && !OVL_TYPE_UPPER(type)) { |
15238 |
+ ino = ovl_remap_lower_ino(ino, xinobits, |
15239 |
+diff --git a/fs/ubifs/dir.c b/fs/ubifs/dir.c |
15240 |
+index 5767b373a8ff..d7c0aa0626cd 100644 |
15241 |
+--- a/fs/ubifs/dir.c |
15242 |
++++ b/fs/ubifs/dir.c |
15243 |
+@@ -242,6 +242,8 @@ static struct dentry *ubifs_lookup(struct inode *dir, struct dentry *dentry, |
15244 |
+ if (nm.hash) { |
15245 |
+ ubifs_assert(c, fname_len(&nm) == 0); |
15246 |
+ ubifs_assert(c, fname_name(&nm) == NULL); |
15247 |
++ if (nm.hash & ~UBIFS_S_KEY_HASH_MASK) |
15248 |
++ goto done; /* ENOENT */ |
15249 |
+ dent_key_init_hash(c, &key, dir->i_ino, nm.hash); |
15250 |
+ err = ubifs_tnc_lookup_dh(c, &key, dent, nm.minor_hash); |
15251 |
+ } else { |
15252 |
+diff --git a/fs/ubifs/file.c b/fs/ubifs/file.c |
15253 |
+index 1b78f2e09218..65b4f63349c7 100644 |
15254 |
+--- a/fs/ubifs/file.c |
15255 |
++++ b/fs/ubifs/file.c |
15256 |
+@@ -798,7 +798,9 @@ static int ubifs_do_bulk_read(struct ubifs_info *c, struct bu_info *bu, |
15257 |
+ |
15258 |
+ if (page_offset > end_index) |
15259 |
+ break; |
15260 |
+- page = find_or_create_page(mapping, page_offset, ra_gfp_mask); |
15261 |
++ page = pagecache_get_page(mapping, page_offset, |
15262 |
++ FGP_LOCK|FGP_ACCESSED|FGP_CREAT|FGP_NOWAIT, |
15263 |
++ ra_gfp_mask); |
15264 |
+ if (!page) |
15265 |
+ break; |
15266 |
+ if (!PageUptodate(page)) |
15267 |
+diff --git a/fs/ubifs/ioctl.c b/fs/ubifs/ioctl.c |
15268 |
+index 0164bcc827f8..e666459f63c7 100644 |
15269 |
+--- a/fs/ubifs/ioctl.c |
15270 |
++++ b/fs/ubifs/ioctl.c |
15271 |
+@@ -28,6 +28,11 @@ |
15272 |
+ #include <linux/mount.h> |
15273 |
+ #include "ubifs.h" |
15274 |
+ |
15275 |
++/* Need to be kept consistent with checked flags in ioctl2ubifs() */ |
15276 |
++#define UBIFS_SUPPORTED_IOCTL_FLAGS \ |
15277 |
++ (FS_COMPR_FL | FS_SYNC_FL | FS_APPEND_FL | \ |
15278 |
++ FS_IMMUTABLE_FL | FS_DIRSYNC_FL) |
15279 |
++ |
15280 |
+ /** |
15281 |
+ * ubifs_set_inode_flags - set VFS inode flags. |
15282 |
+ * @inode: VFS inode to set flags for |
15283 |
+@@ -127,7 +132,8 @@ static int setflags(struct inode *inode, int flags) |
15284 |
+ } |
15285 |
+ } |
15286 |
+ |
15287 |
+- ui->flags = ioctl2ubifs(flags); |
15288 |
++ ui->flags &= ~ioctl2ubifs(UBIFS_SUPPORTED_IOCTL_FLAGS); |
15289 |
++ ui->flags |= ioctl2ubifs(flags); |
15290 |
+ ubifs_set_inode_flags(inode); |
15291 |
+ inode->i_ctime = current_time(inode); |
15292 |
+ release = ui->dirty; |
15293 |
+@@ -169,6 +175,9 @@ long ubifs_ioctl(struct file *file, unsigned int cmd, unsigned long arg) |
15294 |
+ if (get_user(flags, (int __user *) arg)) |
15295 |
+ return -EFAULT; |
15296 |
+ |
15297 |
++ if (flags & ~UBIFS_SUPPORTED_IOCTL_FLAGS) |
15298 |
++ return -EOPNOTSUPP; |
15299 |
++ |
15300 |
+ if (!S_ISDIR(inode->i_mode)) |
15301 |
+ flags &= ~FS_DIRSYNC_FL; |
15302 |
+ |
15303 |
+diff --git a/fs/udf/super.c b/fs/udf/super.c |
15304 |
+index 7af011dc9ae8..6fd0f14e9dd2 100644 |
15305 |
+--- a/fs/udf/super.c |
15306 |
++++ b/fs/udf/super.c |
15307 |
+@@ -999,7 +999,6 @@ static int check_partition_desc(struct super_block *sb, |
15308 |
+ switch (le32_to_cpu(p->accessType)) { |
15309 |
+ case PD_ACCESS_TYPE_READ_ONLY: |
15310 |
+ case PD_ACCESS_TYPE_WRITE_ONCE: |
15311 |
+- case PD_ACCESS_TYPE_REWRITABLE: |
15312 |
+ case PD_ACCESS_TYPE_NONE: |
15313 |
+ goto force_ro; |
15314 |
+ } |
15315 |
+diff --git a/include/linux/eventfd.h b/include/linux/eventfd.h |
15316 |
+index ffcc7724ca21..dc4fd8a6644d 100644 |
15317 |
+--- a/include/linux/eventfd.h |
15318 |
++++ b/include/linux/eventfd.h |
15319 |
+@@ -12,6 +12,8 @@ |
15320 |
+ #include <linux/fcntl.h> |
15321 |
+ #include <linux/wait.h> |
15322 |
+ #include <linux/err.h> |
15323 |
++#include <linux/percpu-defs.h> |
15324 |
++#include <linux/percpu.h> |
15325 |
+ |
15326 |
+ /* |
15327 |
+ * CAREFUL: Check include/uapi/asm-generic/fcntl.h when defining |
15328 |
+@@ -40,6 +42,13 @@ __u64 eventfd_signal(struct eventfd_ctx *ctx, __u64 n); |
15329 |
+ int eventfd_ctx_remove_wait_queue(struct eventfd_ctx *ctx, wait_queue_entry_t *wait, |
15330 |
+ __u64 *cnt); |
15331 |
+ |
15332 |
++DECLARE_PER_CPU(int, eventfd_wake_count); |
15333 |
++ |
15334 |
++static inline bool eventfd_signal_count(void) |
15335 |
++{ |
15336 |
++ return this_cpu_read(eventfd_wake_count); |
15337 |
++} |
15338 |
++ |
15339 |
+ #else /* CONFIG_EVENTFD */ |
15340 |
+ |
15341 |
+ /* |
15342 |
+@@ -68,6 +77,11 @@ static inline int eventfd_ctx_remove_wait_queue(struct eventfd_ctx *ctx, |
15343 |
+ return -ENOSYS; |
15344 |
+ } |
15345 |
+ |
15346 |
++static inline bool eventfd_signal_count(void) |
15347 |
++{ |
15348 |
++ return false; |
15349 |
++} |
15350 |
++ |
15351 |
+ #endif |
15352 |
+ |
15353 |
+ #endif /* _LINUX_EVENTFD_H */ |
15354 |
+diff --git a/include/linux/irq.h b/include/linux/irq.h |
15355 |
+index c9bffda04a45..6ecaf056ab63 100644 |
15356 |
+--- a/include/linux/irq.h |
15357 |
++++ b/include/linux/irq.h |
15358 |
+@@ -208,6 +208,8 @@ struct irq_data { |
15359 |
+ * IRQD_SINGLE_TARGET - IRQ allows only a single affinity target |
15360 |
+ * IRQD_DEFAULT_TRIGGER_SET - Expected trigger already been set |
15361 |
+ * IRQD_CAN_RESERVE - Can use reservation mode |
15362 |
++ * IRQD_MSI_NOMASK_QUIRK - Non-maskable MSI quirk for affinity change |
15363 |
++ * required |
15364 |
+ */ |
15365 |
+ enum { |
15366 |
+ IRQD_TRIGGER_MASK = 0xf, |
15367 |
+@@ -230,6 +232,7 @@ enum { |
15368 |
+ IRQD_SINGLE_TARGET = (1 << 24), |
15369 |
+ IRQD_DEFAULT_TRIGGER_SET = (1 << 25), |
15370 |
+ IRQD_CAN_RESERVE = (1 << 26), |
15371 |
++ IRQD_MSI_NOMASK_QUIRK = (1 << 27), |
15372 |
+ }; |
15373 |
+ |
15374 |
+ #define __irqd_to_state(d) ACCESS_PRIVATE((d)->common, state_use_accessors) |
15375 |
+@@ -389,6 +392,21 @@ static inline bool irqd_can_reserve(struct irq_data *d) |
15376 |
+ return __irqd_to_state(d) & IRQD_CAN_RESERVE; |
15377 |
+ } |
15378 |
+ |
15379 |
++static inline void irqd_set_msi_nomask_quirk(struct irq_data *d) |
15380 |
++{ |
15381 |
++ __irqd_to_state(d) |= IRQD_MSI_NOMASK_QUIRK; |
15382 |
++} |
15383 |
++ |
15384 |
++static inline void irqd_clr_msi_nomask_quirk(struct irq_data *d) |
15385 |
++{ |
15386 |
++ __irqd_to_state(d) &= ~IRQD_MSI_NOMASK_QUIRK; |
15387 |
++} |
15388 |
++ |
15389 |
++static inline bool irqd_msi_nomask_quirk(struct irq_data *d) |
15390 |
++{ |
15391 |
++ return __irqd_to_state(d) & IRQD_MSI_NOMASK_QUIRK; |
15392 |
++} |
15393 |
++ |
15394 |
+ #undef __irqd_to_state |
15395 |
+ |
15396 |
+ static inline irq_hw_number_t irqd_to_hwirq(struct irq_data *d) |
15397 |
+diff --git a/include/linux/irqdomain.h b/include/linux/irqdomain.h |
15398 |
+index dccfa65aee96..8301f1df0682 100644 |
15399 |
+--- a/include/linux/irqdomain.h |
15400 |
++++ b/include/linux/irqdomain.h |
15401 |
+@@ -202,6 +202,13 @@ enum { |
15402 |
+ /* Irq domain implements MSI remapping */ |
15403 |
+ IRQ_DOMAIN_FLAG_MSI_REMAP = (1 << 5), |
15404 |
+ |
15405 |
++ /* |
15406 |
++ * Quirk to handle MSI implementations which do not provide |
15407 |
++ * masking. Currently known to affect x86, but partially |
15408 |
++ * handled in core code. |
15409 |
++ */ |
15410 |
++ IRQ_DOMAIN_MSI_NOMASK_QUIRK = (1 << 6), |
15411 |
++ |
15412 |
+ /* |
15413 |
+ * Flags starting from IRQ_DOMAIN_FLAG_NONCORE are reserved |
15414 |
+ * for implementation specific purposes and ignored by the |
15415 |
+diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h |
15416 |
+index 748016ae01e3..0f99ecc01bc7 100644 |
15417 |
+--- a/include/linux/kvm_host.h |
15418 |
++++ b/include/linux/kvm_host.h |
15419 |
+@@ -186,7 +186,7 @@ struct kvm_async_pf { |
15420 |
+ struct list_head queue; |
15421 |
+ struct kvm_vcpu *vcpu; |
15422 |
+ struct mm_struct *mm; |
15423 |
+- gva_t gva; |
15424 |
++ gpa_t cr2_or_gpa; |
15425 |
+ unsigned long addr; |
15426 |
+ struct kvm_arch_async_pf arch; |
15427 |
+ bool wakeup_all; |
15428 |
+@@ -194,8 +194,8 @@ struct kvm_async_pf { |
15429 |
+ |
15430 |
+ void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu); |
15431 |
+ void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu); |
15432 |
+-int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva, |
15433 |
+- struct kvm_arch_async_pf *arch); |
15434 |
++int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
15435 |
++ unsigned long hva, struct kvm_arch_async_pf *arch); |
15436 |
+ int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu); |
15437 |
+ #endif |
15438 |
+ |
15439 |
+@@ -704,7 +704,7 @@ int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len); |
15440 |
+ int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len); |
15441 |
+ struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn); |
15442 |
+ bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn); |
15443 |
+-unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn); |
15444 |
++unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn); |
15445 |
+ void mark_page_dirty(struct kvm *kvm, gfn_t gfn); |
15446 |
+ |
15447 |
+ struct kvm_memslots *kvm_vcpu_memslots(struct kvm_vcpu *vcpu); |
15448 |
+diff --git a/include/linux/memblock.h b/include/linux/memblock.h |
15449 |
+index 516920549378..2acdd046df2d 100644 |
15450 |
+--- a/include/linux/memblock.h |
15451 |
++++ b/include/linux/memblock.h |
15452 |
+@@ -265,21 +265,6 @@ void __next_mem_pfn_range(int *idx, int nid, unsigned long *out_start_pfn, |
15453 |
+ for_each_mem_range_rev(i, &memblock.memory, &memblock.reserved, \ |
15454 |
+ nid, flags, p_start, p_end, p_nid) |
15455 |
+ |
15456 |
+-/** |
15457 |
+- * for_each_resv_unavail_range - iterate through reserved and unavailable memory |
15458 |
+- * @i: u64 used as loop variable |
15459 |
+- * @p_start: ptr to phys_addr_t for start address of the range, can be %NULL |
15460 |
+- * @p_end: ptr to phys_addr_t for end address of the range, can be %NULL |
15461 |
+- * |
15462 |
+- * Walks over unavailable but reserved (reserved && !memory) areas of memblock. |
15463 |
+- * Available as soon as memblock is initialized. |
15464 |
+- * Note: because this memory does not belong to any physical node, flags and |
15465 |
+- * nid arguments do not make sense and thus not exported as arguments. |
15466 |
+- */ |
15467 |
+-#define for_each_resv_unavail_range(i, p_start, p_end) \ |
15468 |
+- for_each_mem_range(i, &memblock.reserved, &memblock.memory, \ |
15469 |
+- NUMA_NO_NODE, MEMBLOCK_NONE, p_start, p_end, NULL) |
15470 |
+- |
15471 |
+ static inline void memblock_set_region_flags(struct memblock_region *r, |
15472 |
+ enum memblock_flags flags) |
15473 |
+ { |
15474 |
+diff --git a/include/linux/percpu-defs.h b/include/linux/percpu-defs.h |
15475 |
+index 2d2096ba1cfe..90b8ce813fa6 100644 |
15476 |
+--- a/include/linux/percpu-defs.h |
15477 |
++++ b/include/linux/percpu-defs.h |
15478 |
+@@ -176,8 +176,7 @@ |
15479 |
+ * Declaration/definition used for per-CPU variables that should be accessed |
15480 |
+ * as decrypted when memory encryption is enabled in the guest. |
15481 |
+ */ |
15482 |
+-#if defined(CONFIG_VIRTUALIZATION) && defined(CONFIG_AMD_MEM_ENCRYPT) |
15483 |
+- |
15484 |
++#ifdef CONFIG_AMD_MEM_ENCRYPT |
15485 |
+ #define DECLARE_PER_CPU_DECRYPTED(type, name) \ |
15486 |
+ DECLARE_PER_CPU_SECTION(type, name, "..decrypted") |
15487 |
+ |
15488 |
+diff --git a/include/media/v4l2-rect.h b/include/media/v4l2-rect.h |
15489 |
+index 595c3ba05f23..59745e5feb4d 100644 |
15490 |
+--- a/include/media/v4l2-rect.h |
15491 |
++++ b/include/media/v4l2-rect.h |
15492 |
+@@ -63,10 +63,10 @@ static inline void v4l2_rect_map_inside(struct v4l2_rect *r, |
15493 |
+ r->left = boundary->left; |
15494 |
+ if (r->top < boundary->top) |
15495 |
+ r->top = boundary->top; |
15496 |
+- if (r->left + r->width > boundary->width) |
15497 |
+- r->left = boundary->width - r->width; |
15498 |
+- if (r->top + r->height > boundary->height) |
15499 |
+- r->top = boundary->height - r->height; |
15500 |
++ if (r->left + r->width > boundary->left + boundary->width) |
15501 |
++ r->left = boundary->left + boundary->width - r->width; |
15502 |
++ if (r->top + r->height > boundary->top + boundary->height) |
15503 |
++ r->top = boundary->top + boundary->height - r->height; |
15504 |
+ } |
15505 |
+ |
15506 |
+ /** |
15507 |
+diff --git a/include/net/ipx.h b/include/net/ipx.h |
15508 |
+index baf090390998..9d1342807b59 100644 |
15509 |
+--- a/include/net/ipx.h |
15510 |
++++ b/include/net/ipx.h |
15511 |
+@@ -47,11 +47,6 @@ struct ipxhdr { |
15512 |
+ /* From af_ipx.c */ |
15513 |
+ extern int sysctl_ipx_pprop_broadcasting; |
15514 |
+ |
15515 |
+-static __inline__ struct ipxhdr *ipx_hdr(struct sk_buff *skb) |
15516 |
+-{ |
15517 |
+- return (struct ipxhdr *)skb_transport_header(skb); |
15518 |
+-} |
15519 |
+- |
15520 |
+ struct ipx_interface { |
15521 |
+ /* IPX address */ |
15522 |
+ __be32 if_netnum; |
15523 |
+diff --git a/ipc/msg.c b/ipc/msg.c |
15524 |
+index 883642cf2b27..ac4de3f67261 100644 |
15525 |
+--- a/ipc/msg.c |
15526 |
++++ b/ipc/msg.c |
15527 |
+@@ -377,7 +377,7 @@ copy_msqid_from_user(struct msqid64_ds *out, void __user *buf, int version) |
15528 |
+ * NOTE: no locks must be held, the rwsem is taken inside this function. |
15529 |
+ */ |
15530 |
+ static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd, |
15531 |
+- struct msqid64_ds *msqid64) |
15532 |
++ struct ipc64_perm *perm, int msg_qbytes) |
15533 |
+ { |
15534 |
+ struct kern_ipc_perm *ipcp; |
15535 |
+ struct msg_queue *msq; |
15536 |
+@@ -387,7 +387,7 @@ static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd, |
15537 |
+ rcu_read_lock(); |
15538 |
+ |
15539 |
+ ipcp = ipcctl_obtain_check(ns, &msg_ids(ns), msqid, cmd, |
15540 |
+- &msqid64->msg_perm, msqid64->msg_qbytes); |
15541 |
++ perm, msg_qbytes); |
15542 |
+ if (IS_ERR(ipcp)) { |
15543 |
+ err = PTR_ERR(ipcp); |
15544 |
+ goto out_unlock1; |
15545 |
+@@ -409,18 +409,18 @@ static int msgctl_down(struct ipc_namespace *ns, int msqid, int cmd, |
15546 |
+ { |
15547 |
+ DEFINE_WAKE_Q(wake_q); |
15548 |
+ |
15549 |
+- if (msqid64->msg_qbytes > ns->msg_ctlmnb && |
15550 |
++ if (msg_qbytes > ns->msg_ctlmnb && |
15551 |
+ !capable(CAP_SYS_RESOURCE)) { |
15552 |
+ err = -EPERM; |
15553 |
+ goto out_unlock1; |
15554 |
+ } |
15555 |
+ |
15556 |
+ ipc_lock_object(&msq->q_perm); |
15557 |
+- err = ipc_update_perm(&msqid64->msg_perm, ipcp); |
15558 |
++ err = ipc_update_perm(perm, ipcp); |
15559 |
+ if (err) |
15560 |
+ goto out_unlock0; |
15561 |
+ |
15562 |
+- msq->q_qbytes = msqid64->msg_qbytes; |
15563 |
++ msq->q_qbytes = msg_qbytes; |
15564 |
+ |
15565 |
+ msq->q_ctime = ktime_get_real_seconds(); |
15566 |
+ /* |
15567 |
+@@ -603,9 +603,10 @@ long ksys_msgctl(int msqid, int cmd, struct msqid_ds __user *buf) |
15568 |
+ case IPC_SET: |
15569 |
+ if (copy_msqid_from_user(&msqid64, buf, version)) |
15570 |
+ return -EFAULT; |
15571 |
+- /* fallthru */ |
15572 |
++ return msgctl_down(ns, msqid, cmd, &msqid64.msg_perm, |
15573 |
++ msqid64.msg_qbytes); |
15574 |
+ case IPC_RMID: |
15575 |
+- return msgctl_down(ns, msqid, cmd, &msqid64); |
15576 |
++ return msgctl_down(ns, msqid, cmd, NULL, 0); |
15577 |
+ default: |
15578 |
+ return -EINVAL; |
15579 |
+ } |
15580 |
+@@ -724,9 +725,9 @@ long compat_ksys_msgctl(int msqid, int cmd, void __user *uptr) |
15581 |
+ case IPC_SET: |
15582 |
+ if (copy_compat_msqid_from_user(&msqid64, uptr, version)) |
15583 |
+ return -EFAULT; |
15584 |
+- /* fallthru */ |
15585 |
++ return msgctl_down(ns, msqid, cmd, &msqid64.msg_perm, msqid64.msg_qbytes); |
15586 |
+ case IPC_RMID: |
15587 |
+- return msgctl_down(ns, msqid, cmd, &msqid64); |
15588 |
++ return msgctl_down(ns, msqid, cmd, NULL, 0); |
15589 |
+ default: |
15590 |
+ return -EINVAL; |
15591 |
+ } |
15592 |
+diff --git a/kernel/events/core.c b/kernel/events/core.c |
15593 |
+index 16af86ab24c4..8c70ee23fbe9 100644 |
15594 |
+--- a/kernel/events/core.c |
15595 |
++++ b/kernel/events/core.c |
15596 |
+@@ -5709,7 +5709,15 @@ accounting: |
15597 |
+ */ |
15598 |
+ user_lock_limit *= num_online_cpus(); |
15599 |
+ |
15600 |
+- user_locked = atomic_long_read(&user->locked_vm) + user_extra; |
15601 |
++ user_locked = atomic_long_read(&user->locked_vm); |
15602 |
++ |
15603 |
++ /* |
15604 |
++ * sysctl_perf_event_mlock may have changed, so that |
15605 |
++ * user->locked_vm > user_lock_limit |
15606 |
++ */ |
15607 |
++ if (user_locked > user_lock_limit) |
15608 |
++ user_locked = user_lock_limit; |
15609 |
++ user_locked += user_extra; |
15610 |
+ |
15611 |
+ if (user_locked > user_lock_limit) |
15612 |
+ extra = user_locked - user_lock_limit; |
15613 |
+diff --git a/kernel/irq/debugfs.c b/kernel/irq/debugfs.c |
15614 |
+index 6f636136cccc..b3f55dd581b0 100644 |
15615 |
+--- a/kernel/irq/debugfs.c |
15616 |
++++ b/kernel/irq/debugfs.c |
15617 |
+@@ -113,6 +113,7 @@ static const struct irq_bit_descr irqdata_states[] = { |
15618 |
+ BIT_MASK_DESCR(IRQD_AFFINITY_MANAGED), |
15619 |
+ BIT_MASK_DESCR(IRQD_MANAGED_SHUTDOWN), |
15620 |
+ BIT_MASK_DESCR(IRQD_CAN_RESERVE), |
15621 |
++ BIT_MASK_DESCR(IRQD_MSI_NOMASK_QUIRK), |
15622 |
+ |
15623 |
+ BIT_MASK_DESCR(IRQD_FORWARDED_TO_VCPU), |
15624 |
+ |
15625 |
+diff --git a/kernel/irq/irqdomain.c b/kernel/irq/irqdomain.c |
15626 |
+index 6e8520a81dd8..e0eda2bd3975 100644 |
15627 |
+--- a/kernel/irq/irqdomain.c |
15628 |
++++ b/kernel/irq/irqdomain.c |
15629 |
+@@ -1421,6 +1421,7 @@ int irq_domain_push_irq(struct irq_domain *domain, int virq, void *arg) |
15630 |
+ if (rv) { |
15631 |
+ /* Restore the original irq_data. */ |
15632 |
+ *root_irq_data = *child_irq_data; |
15633 |
++ kfree(child_irq_data); |
15634 |
+ goto error; |
15635 |
+ } |
15636 |
+ |
15637 |
+diff --git a/kernel/irq/msi.c b/kernel/irq/msi.c |
15638 |
+index 4ca2fd46645d..dc1186ce3ecd 100644 |
15639 |
+--- a/kernel/irq/msi.c |
15640 |
++++ b/kernel/irq/msi.c |
15641 |
+@@ -453,8 +453,11 @@ int msi_domain_alloc_irqs(struct irq_domain *domain, struct device *dev, |
15642 |
+ continue; |
15643 |
+ |
15644 |
+ irq_data = irq_domain_get_irq_data(domain, desc->irq); |
15645 |
+- if (!can_reserve) |
15646 |
++ if (!can_reserve) { |
15647 |
+ irqd_clr_can_reserve(irq_data); |
15648 |
++ if (domain->flags & IRQ_DOMAIN_MSI_NOMASK_QUIRK) |
15649 |
++ irqd_set_msi_nomask_quirk(irq_data); |
15650 |
++ } |
15651 |
+ ret = irq_domain_activate_irq(irq_data, can_reserve); |
15652 |
+ if (ret) |
15653 |
+ goto cleanup; |
15654 |
+diff --git a/kernel/module.c b/kernel/module.c |
15655 |
+index d3aaec62c142..70a75a7216ab 100644 |
15656 |
+--- a/kernel/module.c |
15657 |
++++ b/kernel/module.c |
15658 |
+@@ -1729,6 +1729,8 @@ static int module_add_modinfo_attrs(struct module *mod) |
15659 |
+ error_out: |
15660 |
+ if (i > 0) |
15661 |
+ module_remove_modinfo_attrs(mod, --i); |
15662 |
++ else |
15663 |
++ kfree(mod->modinfo_attrs); |
15664 |
+ return error; |
15665 |
+ } |
15666 |
+ |
15667 |
+diff --git a/kernel/padata.c b/kernel/padata.c |
15668 |
+index 6c06b3039fae..11c5f9c8779e 100644 |
15669 |
+--- a/kernel/padata.c |
15670 |
++++ b/kernel/padata.c |
15671 |
+@@ -35,6 +35,8 @@ |
15672 |
+ |
15673 |
+ #define MAX_OBJ_NUM 1000 |
15674 |
+ |
15675 |
++static void padata_free_pd(struct parallel_data *pd); |
15676 |
++ |
15677 |
+ static int padata_index_to_cpu(struct parallel_data *pd, int cpu_index) |
15678 |
+ { |
15679 |
+ int cpu, target_cpu; |
15680 |
+@@ -334,6 +336,7 @@ static void padata_serial_worker(struct work_struct *serial_work) |
15681 |
+ struct padata_serial_queue *squeue; |
15682 |
+ struct parallel_data *pd; |
15683 |
+ LIST_HEAD(local_list); |
15684 |
++ int cnt; |
15685 |
+ |
15686 |
+ local_bh_disable(); |
15687 |
+ squeue = container_of(serial_work, struct padata_serial_queue, work); |
15688 |
+@@ -343,6 +346,8 @@ static void padata_serial_worker(struct work_struct *serial_work) |
15689 |
+ list_replace_init(&squeue->serial.list, &local_list); |
15690 |
+ spin_unlock(&squeue->serial.lock); |
15691 |
+ |
15692 |
++ cnt = 0; |
15693 |
++ |
15694 |
+ while (!list_empty(&local_list)) { |
15695 |
+ struct padata_priv *padata; |
15696 |
+ |
15697 |
+@@ -352,9 +357,12 @@ static void padata_serial_worker(struct work_struct *serial_work) |
15698 |
+ list_del_init(&padata->list); |
15699 |
+ |
15700 |
+ padata->serial(padata); |
15701 |
+- atomic_dec(&pd->refcnt); |
15702 |
++ cnt++; |
15703 |
+ } |
15704 |
+ local_bh_enable(); |
15705 |
++ |
15706 |
++ if (atomic_sub_and_test(cnt, &pd->refcnt)) |
15707 |
++ padata_free_pd(pd); |
15708 |
+ } |
15709 |
+ |
15710 |
+ /** |
15711 |
+@@ -501,8 +509,7 @@ static struct parallel_data *padata_alloc_pd(struct padata_instance *pinst, |
15712 |
+ timer_setup(&pd->timer, padata_reorder_timer, 0); |
15713 |
+ atomic_set(&pd->seq_nr, -1); |
15714 |
+ atomic_set(&pd->reorder_objects, 0); |
15715 |
+- atomic_set(&pd->refcnt, 0); |
15716 |
+- pd->pinst = pinst; |
15717 |
++ atomic_set(&pd->refcnt, 1); |
15718 |
+ spin_lock_init(&pd->lock); |
15719 |
+ |
15720 |
+ return pd; |
15721 |
+@@ -526,31 +533,6 @@ static void padata_free_pd(struct parallel_data *pd) |
15722 |
+ kfree(pd); |
15723 |
+ } |
15724 |
+ |
15725 |
+-/* Flush all objects out of the padata queues. */ |
15726 |
+-static void padata_flush_queues(struct parallel_data *pd) |
15727 |
+-{ |
15728 |
+- int cpu; |
15729 |
+- struct padata_parallel_queue *pqueue; |
15730 |
+- struct padata_serial_queue *squeue; |
15731 |
+- |
15732 |
+- for_each_cpu(cpu, pd->cpumask.pcpu) { |
15733 |
+- pqueue = per_cpu_ptr(pd->pqueue, cpu); |
15734 |
+- flush_work(&pqueue->work); |
15735 |
+- } |
15736 |
+- |
15737 |
+- del_timer_sync(&pd->timer); |
15738 |
+- |
15739 |
+- if (atomic_read(&pd->reorder_objects)) |
15740 |
+- padata_reorder(pd); |
15741 |
+- |
15742 |
+- for_each_cpu(cpu, pd->cpumask.cbcpu) { |
15743 |
+- squeue = per_cpu_ptr(pd->squeue, cpu); |
15744 |
+- flush_work(&squeue->work); |
15745 |
+- } |
15746 |
+- |
15747 |
+- BUG_ON(atomic_read(&pd->refcnt) != 0); |
15748 |
+-} |
15749 |
+- |
15750 |
+ static void __padata_start(struct padata_instance *pinst) |
15751 |
+ { |
15752 |
+ pinst->flags |= PADATA_INIT; |
15753 |
+@@ -564,10 +546,6 @@ static void __padata_stop(struct padata_instance *pinst) |
15754 |
+ pinst->flags &= ~PADATA_INIT; |
15755 |
+ |
15756 |
+ synchronize_rcu(); |
15757 |
+- |
15758 |
+- get_online_cpus(); |
15759 |
+- padata_flush_queues(pinst->pd); |
15760 |
+- put_online_cpus(); |
15761 |
+ } |
15762 |
+ |
15763 |
+ /* Replace the internal control structure with a new one. */ |
15764 |
+@@ -588,8 +566,8 @@ static void padata_replace(struct padata_instance *pinst, |
15765 |
+ if (!cpumask_equal(pd_old->cpumask.cbcpu, pd_new->cpumask.cbcpu)) |
15766 |
+ notification_mask |= PADATA_CPU_SERIAL; |
15767 |
+ |
15768 |
+- padata_flush_queues(pd_old); |
15769 |
+- padata_free_pd(pd_old); |
15770 |
++ if (atomic_dec_and_test(&pd_old->refcnt)) |
15771 |
++ padata_free_pd(pd_old); |
15772 |
+ |
15773 |
+ if (notification_mask) |
15774 |
+ blocking_notifier_call_chain(&pinst->cpumask_change_notifier, |
15775 |
+diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c |
15776 |
+index 845efadaf7ec..7a2fdc097c8c 100644 |
15777 |
+--- a/kernel/printk/printk.c |
15778 |
++++ b/kernel/printk/printk.c |
15779 |
+@@ -2717,8 +2717,6 @@ void register_console(struct console *newcon) |
15780 |
+ * for us. |
15781 |
+ */ |
15782 |
+ logbuf_lock_irqsave(flags); |
15783 |
+- console_seq = syslog_seq; |
15784 |
+- console_idx = syslog_idx; |
15785 |
+ /* |
15786 |
+ * We're about to replay the log buffer. Only do this to the |
15787 |
+ * just-registered console to avoid excessive message spam to |
15788 |
+@@ -2730,6 +2728,8 @@ void register_console(struct console *newcon) |
15789 |
+ */ |
15790 |
+ exclusive_console = newcon; |
15791 |
+ exclusive_console_stop_seq = console_seq; |
15792 |
++ console_seq = syslog_seq; |
15793 |
++ console_idx = syslog_idx; |
15794 |
+ logbuf_unlock_irqrestore(flags); |
15795 |
+ } |
15796 |
+ console_unlock(); |
15797 |
+diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h |
15798 |
+index a97c20ea9bce..5f6de49dc78e 100644 |
15799 |
+--- a/kernel/rcu/tree_plugin.h |
15800 |
++++ b/kernel/rcu/tree_plugin.h |
15801 |
+@@ -267,7 +267,7 @@ static void rcu_preempt_ctxt_queue(struct rcu_node *rnp, struct rcu_data *rdp) |
15802 |
+ * blocked tasks. |
15803 |
+ */ |
15804 |
+ if (!rnp->gp_tasks && (blkd_state & RCU_GP_BLKD)) { |
15805 |
+- rnp->gp_tasks = &t->rcu_node_entry; |
15806 |
++ WRITE_ONCE(rnp->gp_tasks, &t->rcu_node_entry); |
15807 |
+ WARN_ON_ONCE(rnp->completedqs == rnp->gp_seq); |
15808 |
+ } |
15809 |
+ if (!rnp->exp_tasks && (blkd_state & RCU_EXP_BLKD)) |
15810 |
+@@ -392,7 +392,7 @@ static void rcu_preempt_note_context_switch(bool preempt) |
15811 |
+ */ |
15812 |
+ static int rcu_preempt_blocked_readers_cgp(struct rcu_node *rnp) |
15813 |
+ { |
15814 |
+- return rnp->gp_tasks != NULL; |
15815 |
++ return READ_ONCE(rnp->gp_tasks) != NULL; |
15816 |
+ } |
15817 |
+ |
15818 |
+ /* |
15819 |
+@@ -557,7 +557,7 @@ static void rcu_read_unlock_special(struct task_struct *t) |
15820 |
+ trace_rcu_unlock_preempted_task(TPS("rcu_preempt"), |
15821 |
+ rnp->gp_seq, t->pid); |
15822 |
+ if (&t->rcu_node_entry == rnp->gp_tasks) |
15823 |
+- rnp->gp_tasks = np; |
15824 |
++ WRITE_ONCE(rnp->gp_tasks, np); |
15825 |
+ if (&t->rcu_node_entry == rnp->exp_tasks) |
15826 |
+ rnp->exp_tasks = np; |
15827 |
+ if (IS_ENABLED(CONFIG_RCU_BOOST)) { |
15828 |
+@@ -716,7 +716,7 @@ rcu_preempt_check_blocked_tasks(struct rcu_state *rsp, struct rcu_node *rnp) |
15829 |
+ dump_blkd_tasks(rsp, rnp, 10); |
15830 |
+ if (rcu_preempt_has_tasks(rnp) && |
15831 |
+ (rnp->qsmaskinit || rnp->wait_blkd_tasks)) { |
15832 |
+- rnp->gp_tasks = rnp->blkd_tasks.next; |
15833 |
++ WRITE_ONCE(rnp->gp_tasks, rnp->blkd_tasks.next); |
15834 |
+ t = container_of(rnp->gp_tasks, struct task_struct, |
15835 |
+ rcu_node_entry); |
15836 |
+ trace_rcu_unlock_preempted_task(TPS("rcu_preempt-GPS"), |
15837 |
+@@ -883,7 +883,8 @@ dump_blkd_tasks(struct rcu_state *rsp, struct rcu_node *rnp, int ncheck) |
15838 |
+ pr_info("%s: %d:%d ->qsmask %#lx ->qsmaskinit %#lx ->qsmaskinitnext %#lx\n", |
15839 |
+ __func__, rnp1->grplo, rnp1->grphi, rnp1->qsmask, rnp1->qsmaskinit, rnp1->qsmaskinitnext); |
15840 |
+ pr_info("%s: ->gp_tasks %p ->boost_tasks %p ->exp_tasks %p\n", |
15841 |
+- __func__, rnp->gp_tasks, rnp->boost_tasks, rnp->exp_tasks); |
15842 |
++ __func__, READ_ONCE(rnp->gp_tasks), rnp->boost_tasks, |
15843 |
++ rnp->exp_tasks); |
15844 |
+ pr_info("%s: ->blkd_tasks", __func__); |
15845 |
+ i = 0; |
15846 |
+ list_for_each(lhp, &rnp->blkd_tasks) { |
15847 |
+diff --git a/kernel/time/alarmtimer.c b/kernel/time/alarmtimer.c |
15848 |
+index f4255a65c44b..9eece67f29f3 100644 |
15849 |
+--- a/kernel/time/alarmtimer.c |
15850 |
++++ b/kernel/time/alarmtimer.c |
15851 |
+@@ -91,6 +91,7 @@ static int alarmtimer_rtc_add_device(struct device *dev, |
15852 |
+ unsigned long flags; |
15853 |
+ struct rtc_device *rtc = to_rtc_device(dev); |
15854 |
+ struct wakeup_source *__ws; |
15855 |
++ int ret = 0; |
15856 |
+ |
15857 |
+ if (rtcdev) |
15858 |
+ return -EBUSY; |
15859 |
+@@ -105,8 +106,8 @@ static int alarmtimer_rtc_add_device(struct device *dev, |
15860 |
+ spin_lock_irqsave(&rtcdev_lock, flags); |
15861 |
+ if (!rtcdev) { |
15862 |
+ if (!try_module_get(rtc->owner)) { |
15863 |
+- spin_unlock_irqrestore(&rtcdev_lock, flags); |
15864 |
+- return -1; |
15865 |
++ ret = -1; |
15866 |
++ goto unlock; |
15867 |
+ } |
15868 |
+ |
15869 |
+ rtcdev = rtc; |
15870 |
+@@ -115,11 +116,12 @@ static int alarmtimer_rtc_add_device(struct device *dev, |
15871 |
+ ws = __ws; |
15872 |
+ __ws = NULL; |
15873 |
+ } |
15874 |
++unlock: |
15875 |
+ spin_unlock_irqrestore(&rtcdev_lock, flags); |
15876 |
+ |
15877 |
+ wakeup_source_unregister(__ws); |
15878 |
+ |
15879 |
+- return 0; |
15880 |
++ return ret; |
15881 |
+ } |
15882 |
+ |
15883 |
+ static inline void alarmtimer_rtc_timer_init(void) |
15884 |
+diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c |
15885 |
+index 0e6e97a01942..f80bb104c41a 100644 |
15886 |
+--- a/kernel/time/clocksource.c |
15887 |
++++ b/kernel/time/clocksource.c |
15888 |
+@@ -311,8 +311,15 @@ static void clocksource_watchdog(struct timer_list *unused) |
15889 |
+ next_cpu = cpumask_next(raw_smp_processor_id(), cpu_online_mask); |
15890 |
+ if (next_cpu >= nr_cpu_ids) |
15891 |
+ next_cpu = cpumask_first(cpu_online_mask); |
15892 |
+- watchdog_timer.expires += WATCHDOG_INTERVAL; |
15893 |
+- add_timer_on(&watchdog_timer, next_cpu); |
15894 |
++ |
15895 |
++ /* |
15896 |
++ * Arm timer if not already pending: could race with concurrent |
15897 |
++ * pair clocksource_stop_watchdog() clocksource_start_watchdog(). |
15898 |
++ */ |
15899 |
++ if (!timer_pending(&watchdog_timer)) { |
15900 |
++ watchdog_timer.expires += WATCHDOG_INTERVAL; |
15901 |
++ add_timer_on(&watchdog_timer, next_cpu); |
15902 |
++ } |
15903 |
+ out: |
15904 |
+ spin_unlock(&watchdog_lock); |
15905 |
+ } |
15906 |
+diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c |
15907 |
+index 37a435bac161..53795237e975 100644 |
15908 |
+--- a/kernel/trace/ftrace.c |
15909 |
++++ b/kernel/trace/ftrace.c |
15910 |
+@@ -5072,8 +5072,8 @@ static const struct file_operations ftrace_notrace_fops = { |
15911 |
+ |
15912 |
+ static DEFINE_MUTEX(graph_lock); |
15913 |
+ |
15914 |
+-struct ftrace_hash *ftrace_graph_hash = EMPTY_HASH; |
15915 |
+-struct ftrace_hash *ftrace_graph_notrace_hash = EMPTY_HASH; |
15916 |
++struct ftrace_hash __rcu *ftrace_graph_hash = EMPTY_HASH; |
15917 |
++struct ftrace_hash __rcu *ftrace_graph_notrace_hash = EMPTY_HASH; |
15918 |
+ |
15919 |
+ enum graph_filter_type { |
15920 |
+ GRAPH_FILTER_NOTRACE = 0, |
15921 |
+@@ -5344,8 +5344,15 @@ ftrace_graph_release(struct inode *inode, struct file *file) |
15922 |
+ |
15923 |
+ mutex_unlock(&graph_lock); |
15924 |
+ |
15925 |
+- /* Wait till all users are no longer using the old hash */ |
15926 |
+- synchronize_sched(); |
15927 |
++ /* |
15928 |
++ * We need to do a hard force of sched synchronization. |
15929 |
++ * This is because we use preempt_disable() to do RCU, but |
15930 |
++ * the function tracers can be called where RCU is not watching |
15931 |
++ * (like before user_exit()). We can not rely on the RCU |
15932 |
++ * infrastructure to do the synchronization, thus we must do it |
15933 |
++ * ourselves. |
15934 |
++ */ |
15935 |
++ schedule_on_each_cpu(ftrace_sync); |
15936 |
+ |
15937 |
+ free_ftrace_hash(old_hash); |
15938 |
+ } |
15939 |
+diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h |
15940 |
+index d11d7bfc3fa5..ee0c6a313ed1 100644 |
15941 |
+--- a/kernel/trace/trace.h |
15942 |
++++ b/kernel/trace/trace.h |
15943 |
+@@ -872,22 +872,31 @@ extern void __trace_graph_return(struct trace_array *tr, |
15944 |
+ unsigned long flags, int pc); |
15945 |
+ |
15946 |
+ #ifdef CONFIG_DYNAMIC_FTRACE |
15947 |
+-extern struct ftrace_hash *ftrace_graph_hash; |
15948 |
+-extern struct ftrace_hash *ftrace_graph_notrace_hash; |
15949 |
++extern struct ftrace_hash __rcu *ftrace_graph_hash; |
15950 |
++extern struct ftrace_hash __rcu *ftrace_graph_notrace_hash; |
15951 |
+ |
15952 |
+ static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace) |
15953 |
+ { |
15954 |
+ unsigned long addr = trace->func; |
15955 |
+ int ret = 0; |
15956 |
++ struct ftrace_hash *hash; |
15957 |
+ |
15958 |
+ preempt_disable_notrace(); |
15959 |
+ |
15960 |
+- if (ftrace_hash_empty(ftrace_graph_hash)) { |
15961 |
++ /* |
15962 |
++ * Have to open code "rcu_dereference_sched()" because the |
15963 |
++ * function graph tracer can be called when RCU is not |
15964 |
++ * "watching". |
15965 |
++ * Protected with schedule_on_each_cpu(ftrace_sync) |
15966 |
++ */ |
15967 |
++ hash = rcu_dereference_protected(ftrace_graph_hash, !preemptible()); |
15968 |
++ |
15969 |
++ if (ftrace_hash_empty(hash)) { |
15970 |
+ ret = 1; |
15971 |
+ goto out; |
15972 |
+ } |
15973 |
+ |
15974 |
+- if (ftrace_lookup_ip(ftrace_graph_hash, addr)) { |
15975 |
++ if (ftrace_lookup_ip(hash, addr)) { |
15976 |
+ |
15977 |
+ /* |
15978 |
+ * This needs to be cleared on the return functions |
15979 |
+@@ -923,10 +932,20 @@ static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace) |
15980 |
+ static inline int ftrace_graph_notrace_addr(unsigned long addr) |
15981 |
+ { |
15982 |
+ int ret = 0; |
15983 |
++ struct ftrace_hash *notrace_hash; |
15984 |
+ |
15985 |
+ preempt_disable_notrace(); |
15986 |
+ |
15987 |
+- if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr)) |
15988 |
++ /* |
15989 |
++ * Have to open code "rcu_dereference_sched()" because the |
15990 |
++ * function graph tracer can be called when RCU is not |
15991 |
++ * "watching". |
15992 |
++ * Protected with schedule_on_each_cpu(ftrace_sync) |
15993 |
++ */ |
15994 |
++ notrace_hash = rcu_dereference_protected(ftrace_graph_notrace_hash, |
15995 |
++ !preemptible()); |
15996 |
++ |
15997 |
++ if (ftrace_lookup_ip(notrace_hash, addr)) |
15998 |
+ ret = 1; |
15999 |
+ |
16000 |
+ preempt_enable_notrace(); |
16001 |
+diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c |
16002 |
+index e288168661e1..e304196d7c28 100644 |
16003 |
+--- a/kernel/trace/trace_sched_switch.c |
16004 |
++++ b/kernel/trace/trace_sched_switch.c |
16005 |
+@@ -89,8 +89,10 @@ static void tracing_sched_unregister(void) |
16006 |
+ |
16007 |
+ static void tracing_start_sched_switch(int ops) |
16008 |
+ { |
16009 |
+- bool sched_register = (!sched_cmdline_ref && !sched_tgid_ref); |
16010 |
++ bool sched_register; |
16011 |
++ |
16012 |
+ mutex_lock(&sched_register_mutex); |
16013 |
++ sched_register = (!sched_cmdline_ref && !sched_tgid_ref); |
16014 |
+ |
16015 |
+ switch (ops) { |
16016 |
+ case RECORD_CMDLINE: |
16017 |
+diff --git a/lib/test_kasan.c b/lib/test_kasan.c |
16018 |
+index ec657105edbf..bd0e067c4895 100644 |
16019 |
+--- a/lib/test_kasan.c |
16020 |
++++ b/lib/test_kasan.c |
16021 |
+@@ -157,6 +157,7 @@ static noinline void __init kmalloc_oob_krealloc_more(void) |
16022 |
+ if (!ptr1 || !ptr2) { |
16023 |
+ pr_err("Allocation failed\n"); |
16024 |
+ kfree(ptr1); |
16025 |
++ kfree(ptr2); |
16026 |
+ return; |
16027 |
+ } |
16028 |
+ |
16029 |
+diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c |
16030 |
+index abc10dcbc9d5..aae7ff485671 100644 |
16031 |
+--- a/mm/memory_hotplug.c |
16032 |
++++ b/mm/memory_hotplug.c |
16033 |
+@@ -1902,8 +1902,6 @@ void __ref __remove_memory(int nid, u64 start, u64 size) |
16034 |
+ |
16035 |
+ BUG_ON(check_hotplug_memory_range(start, size)); |
16036 |
+ |
16037 |
+- mem_hotplug_begin(); |
16038 |
+- |
16039 |
+ /* |
16040 |
+ * All memory blocks must be offlined before removing memory. Check |
16041 |
+ * whether all memory blocks in question are offline and trigger a BUG() |
16042 |
+@@ -1919,9 +1917,14 @@ void __ref __remove_memory(int nid, u64 start, u64 size) |
16043 |
+ memblock_free(start, size); |
16044 |
+ memblock_remove(start, size); |
16045 |
+ |
16046 |
+- /* remove memory block devices before removing memory */ |
16047 |
++ /* |
16048 |
++ * Memory block device removal under the device_hotplug_lock is |
16049 |
++ * a barrier against racing online attempts. |
16050 |
++ */ |
16051 |
+ remove_memory_block_devices(start, size); |
16052 |
+ |
16053 |
++ mem_hotplug_begin(); |
16054 |
++ |
16055 |
+ arch_remove_memory(nid, start, size, NULL); |
16056 |
+ __release_memory_resource(start, size); |
16057 |
+ |
16058 |
+diff --git a/mm/migrate.c b/mm/migrate.c |
16059 |
+index 70f8ad4ade3f..a69b842f95da 100644 |
16060 |
+--- a/mm/migrate.c |
16061 |
++++ b/mm/migrate.c |
16062 |
+@@ -1623,8 +1623,19 @@ static int do_pages_move(struct mm_struct *mm, nodemask_t task_nodes, |
16063 |
+ start = i; |
16064 |
+ } else if (node != current_node) { |
16065 |
+ err = do_move_pages_to_node(mm, &pagelist, current_node); |
16066 |
+- if (err) |
16067 |
++ if (err) { |
16068 |
++ /* |
16069 |
++ * Positive err means the number of failed |
16070 |
++ * pages to migrate. Since we are going to |
16071 |
++ * abort and return the number of non-migrated |
16072 |
++ * pages, so need to incude the rest of the |
16073 |
++ * nr_pages that have not been attempted as |
16074 |
++ * well. |
16075 |
++ */ |
16076 |
++ if (err > 0) |
16077 |
++ err += nr_pages - i - 1; |
16078 |
+ goto out; |
16079 |
++ } |
16080 |
+ err = store_status(status, start, current_node, i - start); |
16081 |
+ if (err) |
16082 |
+ goto out; |
16083 |
+@@ -1655,8 +1666,11 @@ static int do_pages_move(struct mm_struct *mm, nodemask_t task_nodes, |
16084 |
+ goto out_flush; |
16085 |
+ |
16086 |
+ err = do_move_pages_to_node(mm, &pagelist, current_node); |
16087 |
+- if (err) |
16088 |
++ if (err) { |
16089 |
++ if (err > 0) |
16090 |
++ err += nr_pages - i - 1; |
16091 |
+ goto out; |
16092 |
++ } |
16093 |
+ if (i > start) { |
16094 |
+ err = store_status(status, start, current_node, i - start); |
16095 |
+ if (err) |
16096 |
+@@ -1670,6 +1684,13 @@ out_flush: |
16097 |
+ |
16098 |
+ /* Make sure we do not overwrite the existing error */ |
16099 |
+ err1 = do_move_pages_to_node(mm, &pagelist, current_node); |
16100 |
++ /* |
16101 |
++ * Don't have to report non-attempted pages here since: |
16102 |
++ * - If the above loop is done gracefully all pages have been |
16103 |
++ * attempted. |
16104 |
++ * - If the above loop is aborted it means a fatal error |
16105 |
++ * happened, should return ret. |
16106 |
++ */ |
16107 |
+ if (!err1) |
16108 |
+ err1 = store_status(status, start, current_node, i - start); |
16109 |
+ if (err >= 0) |
16110 |
+diff --git a/mm/page_alloc.c b/mm/page_alloc.c |
16111 |
+index 74fb5c338e8f..e5c610d711f3 100644 |
16112 |
+--- a/mm/page_alloc.c |
16113 |
++++ b/mm/page_alloc.c |
16114 |
+@@ -6456,45 +6456,75 @@ void __init free_area_init_node(int nid, unsigned long *zones_size, |
16115 |
+ } |
16116 |
+ |
16117 |
+ #if defined(CONFIG_HAVE_MEMBLOCK) && !defined(CONFIG_FLAT_NODE_MEM_MAP) |
16118 |
++ |
16119 |
++/* |
16120 |
++ * Zero all valid struct pages in range [spfn, epfn), return number of struct |
16121 |
++ * pages zeroed |
16122 |
++ */ |
16123 |
++static u64 zero_pfn_range(unsigned long spfn, unsigned long epfn) |
16124 |
++{ |
16125 |
++ unsigned long pfn; |
16126 |
++ u64 pgcnt = 0; |
16127 |
++ |
16128 |
++ for (pfn = spfn; pfn < epfn; pfn++) { |
16129 |
++ if (!pfn_valid(ALIGN_DOWN(pfn, pageblock_nr_pages))) { |
16130 |
++ pfn = ALIGN_DOWN(pfn, pageblock_nr_pages) |
16131 |
++ + pageblock_nr_pages - 1; |
16132 |
++ continue; |
16133 |
++ } |
16134 |
++ mm_zero_struct_page(pfn_to_page(pfn)); |
16135 |
++ pgcnt++; |
16136 |
++ } |
16137 |
++ |
16138 |
++ return pgcnt; |
16139 |
++} |
16140 |
++ |
16141 |
+ /* |
16142 |
+ * Only struct pages that are backed by physical memory are zeroed and |
16143 |
+ * initialized by going through __init_single_page(). But, there are some |
16144 |
+ * struct pages which are reserved in memblock allocator and their fields |
16145 |
+ * may be accessed (for example page_to_pfn() on some configuration accesses |
16146 |
+ * flags). We must explicitly zero those struct pages. |
16147 |
++ * |
16148 |
++ * This function also addresses a similar issue where struct pages are left |
16149 |
++ * uninitialized because the physical address range is not covered by |
16150 |
++ * memblock.memory or memblock.reserved. That could happen when memblock |
16151 |
++ * layout is manually configured via memmap=, or when the highest physical |
16152 |
++ * address (max_pfn) does not end on a section boundary. |
16153 |
+ */ |
16154 |
+ void __init zero_resv_unavail(void) |
16155 |
+ { |
16156 |
+ phys_addr_t start, end; |
16157 |
+- unsigned long pfn; |
16158 |
+ u64 i, pgcnt; |
16159 |
++ phys_addr_t next = 0; |
16160 |
+ |
16161 |
+ /* |
16162 |
+- * Loop through ranges that are reserved, but do not have reported |
16163 |
+- * physical memory backing. |
16164 |
++ * Loop through unavailable ranges not covered by memblock.memory. |
16165 |
+ */ |
16166 |
+ pgcnt = 0; |
16167 |
+- for_each_resv_unavail_range(i, &start, &end) { |
16168 |
+- for (pfn = PFN_DOWN(start); pfn < PFN_UP(end); pfn++) { |
16169 |
+- if (!pfn_valid(ALIGN_DOWN(pfn, pageblock_nr_pages))) { |
16170 |
+- pfn = ALIGN_DOWN(pfn, pageblock_nr_pages) |
16171 |
+- + pageblock_nr_pages - 1; |
16172 |
+- continue; |
16173 |
+- } |
16174 |
+- mm_zero_struct_page(pfn_to_page(pfn)); |
16175 |
+- pgcnt++; |
16176 |
+- } |
16177 |
++ for_each_mem_range(i, &memblock.memory, NULL, |
16178 |
++ NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, NULL) { |
16179 |
++ if (next < start) |
16180 |
++ pgcnt += zero_pfn_range(PFN_DOWN(next), PFN_UP(start)); |
16181 |
++ next = end; |
16182 |
+ } |
16183 |
+ |
16184 |
++ /* |
16185 |
++ * Early sections always have a fully populated memmap for the whole |
16186 |
++ * section - see pfn_valid(). If the last section has holes at the |
16187 |
++ * end and that section is marked "online", the memmap will be |
16188 |
++ * considered initialized. Make sure that memmap has a well defined |
16189 |
++ * state. |
16190 |
++ */ |
16191 |
++ pgcnt += zero_pfn_range(PFN_DOWN(next), |
16192 |
++ round_up(max_pfn, PAGES_PER_SECTION)); |
16193 |
++ |
16194 |
+ /* |
16195 |
+ * Struct pages that do not have backing memory. This could be because |
16196 |
+ * firmware is using some of this memory, or for some other reasons. |
16197 |
+- * Once memblock is changed so such behaviour is not allowed: i.e. |
16198 |
+- * list of "reserved" memory must be a subset of list of "memory", then |
16199 |
+- * this code can be removed. |
16200 |
+ */ |
16201 |
+ if (pgcnt) |
16202 |
+- pr_info("Reserved but unavailable: %lld pages", pgcnt); |
16203 |
++ pr_info("Zeroed struct page in unavailable ranges: %lld pages", pgcnt); |
16204 |
+ } |
16205 |
+ #endif /* CONFIG_HAVE_MEMBLOCK && !CONFIG_FLAT_NODE_MEM_MAP */ |
16206 |
+ |
16207 |
+diff --git a/net/hsr/hsr_slave.c b/net/hsr/hsr_slave.c |
16208 |
+index 56080da4aa77..5fee6ec7c93d 100644 |
16209 |
+--- a/net/hsr/hsr_slave.c |
16210 |
++++ b/net/hsr/hsr_slave.c |
16211 |
+@@ -32,6 +32,8 @@ static rx_handler_result_t hsr_handle_frame(struct sk_buff **pskb) |
16212 |
+ |
16213 |
+ rcu_read_lock(); /* hsr->node_db, hsr->ports */ |
16214 |
+ port = hsr_port_get_rcu(skb->dev); |
16215 |
++ if (!port) |
16216 |
++ goto finish_pass; |
16217 |
+ |
16218 |
+ if (hsr_addr_is_self(port->hsr, eth_hdr(skb)->h_source)) { |
16219 |
+ /* Directly kill frames sent by ourselves */ |
16220 |
+diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c |
16221 |
+index e80eb1788f80..34fda81c7db0 100644 |
16222 |
+--- a/net/ipv4/tcp.c |
16223 |
++++ b/net/ipv4/tcp.c |
16224 |
+@@ -2588,10 +2588,12 @@ int tcp_disconnect(struct sock *sk, int flags) |
16225 |
+ tp->snd_ssthresh = TCP_INFINITE_SSTHRESH; |
16226 |
+ tp->snd_cwnd_cnt = 0; |
16227 |
+ tp->window_clamp = 0; |
16228 |
++ tp->delivered = 0; |
16229 |
+ tp->delivered_ce = 0; |
16230 |
+ tcp_set_ca_state(sk, TCP_CA_Open); |
16231 |
+ tp->is_sack_reneg = 0; |
16232 |
+ tcp_clear_retrans(tp); |
16233 |
++ tp->total_retrans = 0; |
16234 |
+ inet_csk_delack_init(sk); |
16235 |
+ /* Initialize rcv_mss to TCP_MIN_MSS to avoid division by 0 |
16236 |
+ * issue in __tcp_select_window() |
16237 |
+@@ -2603,10 +2605,14 @@ int tcp_disconnect(struct sock *sk, int flags) |
16238 |
+ sk->sk_rx_dst = NULL; |
16239 |
+ tcp_saved_syn_free(tp); |
16240 |
+ tp->compressed_ack = 0; |
16241 |
++ tp->segs_in = 0; |
16242 |
++ tp->segs_out = 0; |
16243 |
+ tp->bytes_sent = 0; |
16244 |
+ tp->bytes_acked = 0; |
16245 |
+ tp->bytes_received = 0; |
16246 |
+ tp->bytes_retrans = 0; |
16247 |
++ tp->data_segs_in = 0; |
16248 |
++ tp->data_segs_out = 0; |
16249 |
+ tp->dsack_dups = 0; |
16250 |
+ tp->reord_seen = 0; |
16251 |
+ |
16252 |
+diff --git a/net/l2tp/l2tp_core.c b/net/l2tp/l2tp_core.c |
16253 |
+index e4dec03a19fe..d0a295cd71ef 100644 |
16254 |
+--- a/net/l2tp/l2tp_core.c |
16255 |
++++ b/net/l2tp/l2tp_core.c |
16256 |
+@@ -325,8 +325,13 @@ int l2tp_session_register(struct l2tp_session *session, |
16257 |
+ |
16258 |
+ spin_lock_bh(&pn->l2tp_session_hlist_lock); |
16259 |
+ |
16260 |
++ /* IP encap expects session IDs to be globally unique, while |
16261 |
++ * UDP encap doesn't. |
16262 |
++ */ |
16263 |
+ hlist_for_each_entry(session_walk, g_head, global_hlist) |
16264 |
+- if (session_walk->session_id == session->session_id) { |
16265 |
++ if (session_walk->session_id == session->session_id && |
16266 |
++ (session_walk->tunnel->encap == L2TP_ENCAPTYPE_IP || |
16267 |
++ tunnel->encap == L2TP_ENCAPTYPE_IP)) { |
16268 |
+ err = -EEXIST; |
16269 |
+ goto err_tlock_pnlock; |
16270 |
+ } |
16271 |
+diff --git a/net/rxrpc/af_rxrpc.c b/net/rxrpc/af_rxrpc.c |
16272 |
+index a74edb10cbfc..57f835d2442e 100644 |
16273 |
+--- a/net/rxrpc/af_rxrpc.c |
16274 |
++++ b/net/rxrpc/af_rxrpc.c |
16275 |
+@@ -196,6 +196,7 @@ static int rxrpc_bind(struct socket *sock, struct sockaddr *saddr, int len) |
16276 |
+ service_in_use: |
16277 |
+ write_unlock(&local->services_lock); |
16278 |
+ rxrpc_unuse_local(local); |
16279 |
++ rxrpc_put_local(local); |
16280 |
+ ret = -EADDRINUSE; |
16281 |
+ error_unlock: |
16282 |
+ release_sock(&rx->sk); |
16283 |
+@@ -906,6 +907,7 @@ static int rxrpc_release_sock(struct sock *sk) |
16284 |
+ rxrpc_purge_queue(&sk->sk_receive_queue); |
16285 |
+ |
16286 |
+ rxrpc_unuse_local(rx->local); |
16287 |
++ rxrpc_put_local(rx->local); |
16288 |
+ rx->local = NULL; |
16289 |
+ key_put(rx->key); |
16290 |
+ rx->key = NULL; |
16291 |
+diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h |
16292 |
+index ccef6e40e002..9c4ee7513214 100644 |
16293 |
+--- a/net/rxrpc/ar-internal.h |
16294 |
++++ b/net/rxrpc/ar-internal.h |
16295 |
+@@ -484,6 +484,7 @@ enum rxrpc_call_flag { |
16296 |
+ RXRPC_CALL_BEGAN_RX_TIMER, /* We began the expect_rx_by timer */ |
16297 |
+ RXRPC_CALL_RX_HEARD, /* The peer responded at least once to this call */ |
16298 |
+ RXRPC_CALL_RX_UNDERRUN, /* Got data underrun */ |
16299 |
++ RXRPC_CALL_DISCONNECTED, /* The call has been disconnected */ |
16300 |
+ }; |
16301 |
+ |
16302 |
+ /* |
16303 |
+@@ -1006,6 +1007,16 @@ void rxrpc_unuse_local(struct rxrpc_local *); |
16304 |
+ void rxrpc_queue_local(struct rxrpc_local *); |
16305 |
+ void rxrpc_destroy_all_locals(struct rxrpc_net *); |
16306 |
+ |
16307 |
++static inline bool __rxrpc_unuse_local(struct rxrpc_local *local) |
16308 |
++{ |
16309 |
++ return atomic_dec_return(&local->active_users) == 0; |
16310 |
++} |
16311 |
++ |
16312 |
++static inline bool __rxrpc_use_local(struct rxrpc_local *local) |
16313 |
++{ |
16314 |
++ return atomic_fetch_add_unless(&local->active_users, 1, 0) != 0; |
16315 |
++} |
16316 |
++ |
16317 |
+ /* |
16318 |
+ * misc.c |
16319 |
+ */ |
16320 |
+diff --git a/net/rxrpc/call_object.c b/net/rxrpc/call_object.c |
16321 |
+index 215f4d98baa0..17fdfce1625f 100644 |
16322 |
+--- a/net/rxrpc/call_object.c |
16323 |
++++ b/net/rxrpc/call_object.c |
16324 |
+@@ -520,7 +520,7 @@ void rxrpc_release_call(struct rxrpc_sock *rx, struct rxrpc_call *call) |
16325 |
+ |
16326 |
+ _debug("RELEASE CALL %p (%d CONN %p)", call, call->debug_id, conn); |
16327 |
+ |
16328 |
+- if (conn) |
16329 |
++ if (conn && !test_bit(RXRPC_CALL_DISCONNECTED, &call->flags)) |
16330 |
+ rxrpc_disconnect_call(call); |
16331 |
+ |
16332 |
+ for (i = 0; i < RXRPC_RXTX_BUFF_SIZE; i++) { |
16333 |
+@@ -654,6 +654,7 @@ static void rxrpc_rcu_destroy_call(struct rcu_head *rcu) |
16334 |
+ struct rxrpc_call *call = container_of(rcu, struct rxrpc_call, rcu); |
16335 |
+ struct rxrpc_net *rxnet = call->rxnet; |
16336 |
+ |
16337 |
++ rxrpc_put_connection(call->conn); |
16338 |
+ rxrpc_put_peer(call->peer); |
16339 |
+ kfree(call->rxtx_buffer); |
16340 |
+ kfree(call->rxtx_annotations); |
16341 |
+@@ -677,7 +678,6 @@ void rxrpc_cleanup_call(struct rxrpc_call *call) |
16342 |
+ |
16343 |
+ ASSERTCMP(call->state, ==, RXRPC_CALL_COMPLETE); |
16344 |
+ ASSERT(test_bit(RXRPC_CALL_RELEASED, &call->flags)); |
16345 |
+- ASSERTCMP(call->conn, ==, NULL); |
16346 |
+ |
16347 |
+ /* Clean up the Rx/Tx buffer */ |
16348 |
+ for (i = 0; i < RXRPC_RXTX_BUFF_SIZE; i++) |
16349 |
+diff --git a/net/rxrpc/conn_client.c b/net/rxrpc/conn_client.c |
16350 |
+index 38d548532024..4ffc7b87fec0 100644 |
16351 |
+--- a/net/rxrpc/conn_client.c |
16352 |
++++ b/net/rxrpc/conn_client.c |
16353 |
+@@ -786,6 +786,7 @@ void rxrpc_disconnect_client_call(struct rxrpc_call *call) |
16354 |
+ u32 cid; |
16355 |
+ |
16356 |
+ spin_lock(&conn->channel_lock); |
16357 |
++ set_bit(RXRPC_CALL_DISCONNECTED, &call->flags); |
16358 |
+ |
16359 |
+ cid = call->cid; |
16360 |
+ if (cid) { |
16361 |
+@@ -793,7 +794,6 @@ void rxrpc_disconnect_client_call(struct rxrpc_call *call) |
16362 |
+ chan = &conn->channels[channel]; |
16363 |
+ } |
16364 |
+ trace_rxrpc_client(conn, channel, rxrpc_client_chan_disconnect); |
16365 |
+- call->conn = NULL; |
16366 |
+ |
16367 |
+ /* Calls that have never actually been assigned a channel can simply be |
16368 |
+ * discarded. If the conn didn't get used either, it will follow |
16369 |
+@@ -909,7 +909,6 @@ out: |
16370 |
+ spin_unlock(&rxnet->client_conn_cache_lock); |
16371 |
+ out_2: |
16372 |
+ spin_unlock(&conn->channel_lock); |
16373 |
+- rxrpc_put_connection(conn); |
16374 |
+ _leave(""); |
16375 |
+ return; |
16376 |
+ |
16377 |
+diff --git a/net/rxrpc/conn_event.c b/net/rxrpc/conn_event.c |
16378 |
+index b6fca8ebb117..126154a97a59 100644 |
16379 |
+--- a/net/rxrpc/conn_event.c |
16380 |
++++ b/net/rxrpc/conn_event.c |
16381 |
+@@ -453,16 +453,12 @@ again: |
16382 |
+ /* |
16383 |
+ * connection-level event processor |
16384 |
+ */ |
16385 |
+-void rxrpc_process_connection(struct work_struct *work) |
16386 |
++static void rxrpc_do_process_connection(struct rxrpc_connection *conn) |
16387 |
+ { |
16388 |
+- struct rxrpc_connection *conn = |
16389 |
+- container_of(work, struct rxrpc_connection, processor); |
16390 |
+ struct sk_buff *skb; |
16391 |
+ u32 abort_code = RX_PROTOCOL_ERROR; |
16392 |
+ int ret; |
16393 |
+ |
16394 |
+- rxrpc_see_connection(conn); |
16395 |
+- |
16396 |
+ if (test_and_clear_bit(RXRPC_CONN_EV_CHALLENGE, &conn->events)) |
16397 |
+ rxrpc_secure_connection(conn); |
16398 |
+ |
16399 |
+@@ -490,18 +486,33 @@ void rxrpc_process_connection(struct work_struct *work) |
16400 |
+ } |
16401 |
+ } |
16402 |
+ |
16403 |
+-out: |
16404 |
+- rxrpc_put_connection(conn); |
16405 |
+- _leave(""); |
16406 |
+ return; |
16407 |
+ |
16408 |
+ requeue_and_leave: |
16409 |
+ skb_queue_head(&conn->rx_queue, skb); |
16410 |
+- goto out; |
16411 |
++ return; |
16412 |
+ |
16413 |
+ protocol_error: |
16414 |
+ if (rxrpc_abort_connection(conn, ret, abort_code) < 0) |
16415 |
+ goto requeue_and_leave; |
16416 |
+ rxrpc_free_skb(skb, rxrpc_skb_rx_freed); |
16417 |
+- goto out; |
16418 |
++ return; |
16419 |
++} |
16420 |
++ |
16421 |
++void rxrpc_process_connection(struct work_struct *work) |
16422 |
++{ |
16423 |
++ struct rxrpc_connection *conn = |
16424 |
++ container_of(work, struct rxrpc_connection, processor); |
16425 |
++ |
16426 |
++ rxrpc_see_connection(conn); |
16427 |
++ |
16428 |
++ if (__rxrpc_use_local(conn->params.local)) { |
16429 |
++ rxrpc_do_process_connection(conn); |
16430 |
++ rxrpc_unuse_local(conn->params.local); |
16431 |
++ } |
16432 |
++ |
16433 |
++ rxrpc_put_connection(conn); |
16434 |
++ _leave(""); |
16435 |
++ return; |
16436 |
+ } |
16437 |
++ |
16438 |
+diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c |
16439 |
+index f338efd2880a..c4c4450891e0 100644 |
16440 |
+--- a/net/rxrpc/conn_object.c |
16441 |
++++ b/net/rxrpc/conn_object.c |
16442 |
+@@ -226,9 +226,8 @@ void rxrpc_disconnect_call(struct rxrpc_call *call) |
16443 |
+ __rxrpc_disconnect_call(conn, call); |
16444 |
+ spin_unlock(&conn->channel_lock); |
16445 |
+ |
16446 |
+- call->conn = NULL; |
16447 |
++ set_bit(RXRPC_CALL_DISCONNECTED, &call->flags); |
16448 |
+ conn->idle_timestamp = jiffies; |
16449 |
+- rxrpc_put_connection(conn); |
16450 |
+ } |
16451 |
+ |
16452 |
+ /* |
16453 |
+diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c |
16454 |
+index 2f91ab909191..d9beb28fc32f 100644 |
16455 |
+--- a/net/rxrpc/input.c |
16456 |
++++ b/net/rxrpc/input.c |
16457 |
+@@ -612,10 +612,8 @@ ack: |
16458 |
+ false, true, |
16459 |
+ rxrpc_propose_ack_input_data); |
16460 |
+ |
16461 |
+- if (sp->hdr.seq == READ_ONCE(call->rx_hard_ack) + 1) { |
16462 |
+- trace_rxrpc_notify_socket(call->debug_id, serial); |
16463 |
+- rxrpc_notify_socket(call); |
16464 |
+- } |
16465 |
++ trace_rxrpc_notify_socket(call->debug_id, serial); |
16466 |
++ rxrpc_notify_socket(call); |
16467 |
+ |
16468 |
+ unlock: |
16469 |
+ spin_unlock(&call->input_lock); |
16470 |
+diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c |
16471 |
+index 04f0976841a4..4c0087a48e87 100644 |
16472 |
+--- a/net/rxrpc/local_object.c |
16473 |
++++ b/net/rxrpc/local_object.c |
16474 |
+@@ -368,11 +368,14 @@ void rxrpc_queue_local(struct rxrpc_local *local) |
16475 |
+ void rxrpc_put_local(struct rxrpc_local *local) |
16476 |
+ { |
16477 |
+ const void *here = __builtin_return_address(0); |
16478 |
++ unsigned int debug_id; |
16479 |
+ int n; |
16480 |
+ |
16481 |
+ if (local) { |
16482 |
++ debug_id = local->debug_id; |
16483 |
++ |
16484 |
+ n = atomic_dec_return(&local->usage); |
16485 |
+- trace_rxrpc_local(local->debug_id, rxrpc_local_put, n, here); |
16486 |
++ trace_rxrpc_local(debug_id, rxrpc_local_put, n, here); |
16487 |
+ |
16488 |
+ if (n == 0) |
16489 |
+ call_rcu(&local->rcu, rxrpc_local_rcu); |
16490 |
+@@ -384,14 +387,11 @@ void rxrpc_put_local(struct rxrpc_local *local) |
16491 |
+ */ |
16492 |
+ struct rxrpc_local *rxrpc_use_local(struct rxrpc_local *local) |
16493 |
+ { |
16494 |
+- unsigned int au; |
16495 |
+- |
16496 |
+ local = rxrpc_get_local_maybe(local); |
16497 |
+ if (!local) |
16498 |
+ return NULL; |
16499 |
+ |
16500 |
+- au = atomic_fetch_add_unless(&local->active_users, 1, 0); |
16501 |
+- if (au == 0) { |
16502 |
++ if (!__rxrpc_use_local(local)) { |
16503 |
+ rxrpc_put_local(local); |
16504 |
+ return NULL; |
16505 |
+ } |
16506 |
+@@ -405,14 +405,11 @@ struct rxrpc_local *rxrpc_use_local(struct rxrpc_local *local) |
16507 |
+ */ |
16508 |
+ void rxrpc_unuse_local(struct rxrpc_local *local) |
16509 |
+ { |
16510 |
+- unsigned int au; |
16511 |
+- |
16512 |
+ if (local) { |
16513 |
+- au = atomic_dec_return(&local->active_users); |
16514 |
+- if (au == 0) |
16515 |
++ if (__rxrpc_unuse_local(local)) { |
16516 |
++ rxrpc_get_local(local); |
16517 |
+ rxrpc_queue_local(local); |
16518 |
+- else |
16519 |
+- rxrpc_put_local(local); |
16520 |
++ } |
16521 |
+ } |
16522 |
+ } |
16523 |
+ |
16524 |
+@@ -469,7 +466,7 @@ static void rxrpc_local_processor(struct work_struct *work) |
16525 |
+ |
16526 |
+ do { |
16527 |
+ again = false; |
16528 |
+- if (atomic_read(&local->active_users) == 0) { |
16529 |
++ if (!__rxrpc_use_local(local)) { |
16530 |
+ rxrpc_local_destroyer(local); |
16531 |
+ break; |
16532 |
+ } |
16533 |
+@@ -483,6 +480,8 @@ static void rxrpc_local_processor(struct work_struct *work) |
16534 |
+ rxrpc_process_local_events(local); |
16535 |
+ again = true; |
16536 |
+ } |
16537 |
++ |
16538 |
++ __rxrpc_unuse_local(local); |
16539 |
+ } while (again); |
16540 |
+ |
16541 |
+ rxrpc_put_local(local); |
16542 |
+diff --git a/net/rxrpc/output.c b/net/rxrpc/output.c |
16543 |
+index 31e47cfb3e68..b0aa08e3796d 100644 |
16544 |
+--- a/net/rxrpc/output.c |
16545 |
++++ b/net/rxrpc/output.c |
16546 |
+@@ -133,7 +133,7 @@ static size_t rxrpc_fill_out_ack(struct rxrpc_connection *conn, |
16547 |
+ int rxrpc_send_ack_packet(struct rxrpc_call *call, bool ping, |
16548 |
+ rxrpc_serial_t *_serial) |
16549 |
+ { |
16550 |
+- struct rxrpc_connection *conn = NULL; |
16551 |
++ struct rxrpc_connection *conn; |
16552 |
+ struct rxrpc_ack_buffer *pkt; |
16553 |
+ struct msghdr msg; |
16554 |
+ struct kvec iov[2]; |
16555 |
+@@ -143,18 +143,14 @@ int rxrpc_send_ack_packet(struct rxrpc_call *call, bool ping, |
16556 |
+ int ret; |
16557 |
+ u8 reason; |
16558 |
+ |
16559 |
+- spin_lock_bh(&call->lock); |
16560 |
+- if (call->conn) |
16561 |
+- conn = rxrpc_get_connection_maybe(call->conn); |
16562 |
+- spin_unlock_bh(&call->lock); |
16563 |
+- if (!conn) |
16564 |
++ if (test_bit(RXRPC_CALL_DISCONNECTED, &call->flags)) |
16565 |
+ return -ECONNRESET; |
16566 |
+ |
16567 |
+ pkt = kzalloc(sizeof(*pkt), GFP_KERNEL); |
16568 |
+- if (!pkt) { |
16569 |
+- rxrpc_put_connection(conn); |
16570 |
++ if (!pkt) |
16571 |
+ return -ENOMEM; |
16572 |
+- } |
16573 |
++ |
16574 |
++ conn = call->conn; |
16575 |
+ |
16576 |
+ msg.msg_name = &call->peer->srx.transport; |
16577 |
+ msg.msg_namelen = call->peer->srx.transport_len; |
16578 |
+@@ -249,7 +245,6 @@ int rxrpc_send_ack_packet(struct rxrpc_call *call, bool ping, |
16579 |
+ } |
16580 |
+ |
16581 |
+ out: |
16582 |
+- rxrpc_put_connection(conn); |
16583 |
+ kfree(pkt); |
16584 |
+ return ret; |
16585 |
+ } |
16586 |
+@@ -259,7 +254,7 @@ out: |
16587 |
+ */ |
16588 |
+ int rxrpc_send_abort_packet(struct rxrpc_call *call) |
16589 |
+ { |
16590 |
+- struct rxrpc_connection *conn = NULL; |
16591 |
++ struct rxrpc_connection *conn; |
16592 |
+ struct rxrpc_abort_buffer pkt; |
16593 |
+ struct msghdr msg; |
16594 |
+ struct kvec iov[1]; |
16595 |
+@@ -276,13 +271,11 @@ int rxrpc_send_abort_packet(struct rxrpc_call *call) |
16596 |
+ test_bit(RXRPC_CALL_TX_LAST, &call->flags)) |
16597 |
+ return 0; |
16598 |
+ |
16599 |
+- spin_lock_bh(&call->lock); |
16600 |
+- if (call->conn) |
16601 |
+- conn = rxrpc_get_connection_maybe(call->conn); |
16602 |
+- spin_unlock_bh(&call->lock); |
16603 |
+- if (!conn) |
16604 |
++ if (test_bit(RXRPC_CALL_DISCONNECTED, &call->flags)) |
16605 |
+ return -ECONNRESET; |
16606 |
+ |
16607 |
++ conn = call->conn; |
16608 |
++ |
16609 |
+ msg.msg_name = &call->peer->srx.transport; |
16610 |
+ msg.msg_namelen = call->peer->srx.transport_len; |
16611 |
+ msg.msg_control = NULL; |
16612 |
+@@ -317,8 +310,6 @@ int rxrpc_send_abort_packet(struct rxrpc_call *call) |
16613 |
+ trace_rxrpc_tx_packet(call->debug_id, &pkt.whdr, |
16614 |
+ rxrpc_tx_point_call_abort); |
16615 |
+ rxrpc_tx_backoff(call, ret); |
16616 |
+- |
16617 |
+- rxrpc_put_connection(conn); |
16618 |
+ return ret; |
16619 |
+ } |
16620 |
+ |
16621 |
+diff --git a/net/rxrpc/peer_event.c b/net/rxrpc/peer_event.c |
16622 |
+index 42582a9ff81d..85bdc31d3dbf 100644 |
16623 |
+--- a/net/rxrpc/peer_event.c |
16624 |
++++ b/net/rxrpc/peer_event.c |
16625 |
+@@ -357,27 +357,31 @@ static void rxrpc_peer_keepalive_dispatch(struct rxrpc_net *rxnet, |
16626 |
+ if (!rxrpc_get_peer_maybe(peer)) |
16627 |
+ continue; |
16628 |
+ |
16629 |
+- spin_unlock_bh(&rxnet->peer_hash_lock); |
16630 |
+- |
16631 |
+- keepalive_at = peer->last_tx_at + RXRPC_KEEPALIVE_TIME; |
16632 |
+- slot = keepalive_at - base; |
16633 |
+- _debug("%02x peer %u t=%d {%pISp}", |
16634 |
+- cursor, peer->debug_id, slot, &peer->srx.transport); |
16635 |
++ if (__rxrpc_use_local(peer->local)) { |
16636 |
++ spin_unlock_bh(&rxnet->peer_hash_lock); |
16637 |
++ |
16638 |
++ keepalive_at = peer->last_tx_at + RXRPC_KEEPALIVE_TIME; |
16639 |
++ slot = keepalive_at - base; |
16640 |
++ _debug("%02x peer %u t=%d {%pISp}", |
16641 |
++ cursor, peer->debug_id, slot, &peer->srx.transport); |
16642 |
++ |
16643 |
++ if (keepalive_at <= base || |
16644 |
++ keepalive_at > base + RXRPC_KEEPALIVE_TIME) { |
16645 |
++ rxrpc_send_keepalive(peer); |
16646 |
++ slot = RXRPC_KEEPALIVE_TIME; |
16647 |
++ } |
16648 |
+ |
16649 |
+- if (keepalive_at <= base || |
16650 |
+- keepalive_at > base + RXRPC_KEEPALIVE_TIME) { |
16651 |
+- rxrpc_send_keepalive(peer); |
16652 |
+- slot = RXRPC_KEEPALIVE_TIME; |
16653 |
++ /* A transmission to this peer occurred since last we |
16654 |
++ * examined it so put it into the appropriate future |
16655 |
++ * bucket. |
16656 |
++ */ |
16657 |
++ slot += cursor; |
16658 |
++ slot &= mask; |
16659 |
++ spin_lock_bh(&rxnet->peer_hash_lock); |
16660 |
++ list_add_tail(&peer->keepalive_link, |
16661 |
++ &rxnet->peer_keepalive[slot & mask]); |
16662 |
++ rxrpc_unuse_local(peer->local); |
16663 |
+ } |
16664 |
+- |
16665 |
+- /* A transmission to this peer occurred since last we examined |
16666 |
+- * it so put it into the appropriate future bucket. |
16667 |
+- */ |
16668 |
+- slot += cursor; |
16669 |
+- slot &= mask; |
16670 |
+- spin_lock_bh(&rxnet->peer_hash_lock); |
16671 |
+- list_add_tail(&peer->keepalive_link, |
16672 |
+- &rxnet->peer_keepalive[slot & mask]); |
16673 |
+ rxrpc_put_peer_locked(peer); |
16674 |
+ } |
16675 |
+ |
16676 |
+diff --git a/net/sched/cls_rsvp.h b/net/sched/cls_rsvp.h |
16677 |
+index 6d30a291bcd2..eb1dd2afc5a1 100644 |
16678 |
+--- a/net/sched/cls_rsvp.h |
16679 |
++++ b/net/sched/cls_rsvp.h |
16680 |
+@@ -466,10 +466,8 @@ static u32 gen_tunnel(struct rsvp_head *data) |
16681 |
+ |
16682 |
+ static const struct nla_policy rsvp_policy[TCA_RSVP_MAX + 1] = { |
16683 |
+ [TCA_RSVP_CLASSID] = { .type = NLA_U32 }, |
16684 |
+- [TCA_RSVP_DST] = { .type = NLA_BINARY, |
16685 |
+- .len = RSVP_DST_LEN * sizeof(u32) }, |
16686 |
+- [TCA_RSVP_SRC] = { .type = NLA_BINARY, |
16687 |
+- .len = RSVP_DST_LEN * sizeof(u32) }, |
16688 |
++ [TCA_RSVP_DST] = { .len = RSVP_DST_LEN * sizeof(u32) }, |
16689 |
++ [TCA_RSVP_SRC] = { .len = RSVP_DST_LEN * sizeof(u32) }, |
16690 |
+ [TCA_RSVP_PINFO] = { .len = sizeof(struct tc_rsvp_pinfo) }, |
16691 |
+ }; |
16692 |
+ |
16693 |
+diff --git a/net/sched/cls_tcindex.c b/net/sched/cls_tcindex.c |
16694 |
+index edf27365f91c..db4b5d9ffaf7 100644 |
16695 |
+--- a/net/sched/cls_tcindex.c |
16696 |
++++ b/net/sched/cls_tcindex.c |
16697 |
+@@ -333,12 +333,31 @@ tcindex_set_parms(struct net *net, struct tcf_proto *tp, unsigned long base, |
16698 |
+ cp->fall_through = p->fall_through; |
16699 |
+ cp->tp = tp; |
16700 |
+ |
16701 |
++ if (tb[TCA_TCINDEX_HASH]) |
16702 |
++ cp->hash = nla_get_u32(tb[TCA_TCINDEX_HASH]); |
16703 |
++ |
16704 |
++ if (tb[TCA_TCINDEX_MASK]) |
16705 |
++ cp->mask = nla_get_u16(tb[TCA_TCINDEX_MASK]); |
16706 |
++ |
16707 |
++ if (tb[TCA_TCINDEX_SHIFT]) |
16708 |
++ cp->shift = nla_get_u32(tb[TCA_TCINDEX_SHIFT]); |
16709 |
++ |
16710 |
++ if (!cp->hash) { |
16711 |
++ /* Hash not specified, use perfect hash if the upper limit |
16712 |
++ * of the hashing index is below the threshold. |
16713 |
++ */ |
16714 |
++ if ((cp->mask >> cp->shift) < PERFECT_HASH_THRESHOLD) |
16715 |
++ cp->hash = (cp->mask >> cp->shift) + 1; |
16716 |
++ else |
16717 |
++ cp->hash = DEFAULT_HASH_SIZE; |
16718 |
++ } |
16719 |
++ |
16720 |
+ if (p->perfect) { |
16721 |
+ int i; |
16722 |
+ |
16723 |
+ if (tcindex_alloc_perfect_hash(net, cp) < 0) |
16724 |
+ goto errout; |
16725 |
+- for (i = 0; i < cp->hash; i++) |
16726 |
++ for (i = 0; i < min(cp->hash, p->hash); i++) |
16727 |
+ cp->perfect[i].res = p->perfect[i].res; |
16728 |
+ balloc = 1; |
16729 |
+ } |
16730 |
+@@ -346,19 +365,10 @@ tcindex_set_parms(struct net *net, struct tcf_proto *tp, unsigned long base, |
16731 |
+ |
16732 |
+ err = tcindex_filter_result_init(&new_filter_result); |
16733 |
+ if (err < 0) |
16734 |
+- goto errout1; |
16735 |
++ goto errout_alloc; |
16736 |
+ if (old_r) |
16737 |
+ cr = r->res; |
16738 |
+ |
16739 |
+- if (tb[TCA_TCINDEX_HASH]) |
16740 |
+- cp->hash = nla_get_u32(tb[TCA_TCINDEX_HASH]); |
16741 |
+- |
16742 |
+- if (tb[TCA_TCINDEX_MASK]) |
16743 |
+- cp->mask = nla_get_u16(tb[TCA_TCINDEX_MASK]); |
16744 |
+- |
16745 |
+- if (tb[TCA_TCINDEX_SHIFT]) |
16746 |
+- cp->shift = nla_get_u32(tb[TCA_TCINDEX_SHIFT]); |
16747 |
+- |
16748 |
+ err = -EBUSY; |
16749 |
+ |
16750 |
+ /* Hash already allocated, make sure that we still meet the |
16751 |
+@@ -376,16 +386,6 @@ tcindex_set_parms(struct net *net, struct tcf_proto *tp, unsigned long base, |
16752 |
+ if (tb[TCA_TCINDEX_FALL_THROUGH]) |
16753 |
+ cp->fall_through = nla_get_u32(tb[TCA_TCINDEX_FALL_THROUGH]); |
16754 |
+ |
16755 |
+- if (!cp->hash) { |
16756 |
+- /* Hash not specified, use perfect hash if the upper limit |
16757 |
+- * of the hashing index is below the threshold. |
16758 |
+- */ |
16759 |
+- if ((cp->mask >> cp->shift) < PERFECT_HASH_THRESHOLD) |
16760 |
+- cp->hash = (cp->mask >> cp->shift) + 1; |
16761 |
+- else |
16762 |
+- cp->hash = DEFAULT_HASH_SIZE; |
16763 |
+- } |
16764 |
+- |
16765 |
+ if (!cp->perfect && !cp->h) |
16766 |
+ cp->alloc_hash = cp->hash; |
16767 |
+ |
16768 |
+@@ -484,7 +484,6 @@ errout_alloc: |
16769 |
+ tcindex_free_perfect_hash(cp); |
16770 |
+ else if (balloc == 2) |
16771 |
+ kfree(cp->h); |
16772 |
+-errout1: |
16773 |
+ tcf_exts_destroy(&new_filter_result.exts); |
16774 |
+ errout: |
16775 |
+ kfree(cp); |
16776 |
+diff --git a/net/sunrpc/auth_gss/svcauth_gss.c b/net/sunrpc/auth_gss/svcauth_gss.c |
16777 |
+index 24c7a1e2bd34..68830e88b6e9 100644 |
16778 |
+--- a/net/sunrpc/auth_gss/svcauth_gss.c |
16779 |
++++ b/net/sunrpc/auth_gss/svcauth_gss.c |
16780 |
+@@ -1224,6 +1224,7 @@ static int gss_proxy_save_rsc(struct cache_detail *cd, |
16781 |
+ dprintk("RPC: No creds found!\n"); |
16782 |
+ goto out; |
16783 |
+ } else { |
16784 |
++ struct timespec64 boot; |
16785 |
+ |
16786 |
+ /* steal creds */ |
16787 |
+ rsci.cred = ud->creds; |
16788 |
+@@ -1244,6 +1245,9 @@ static int gss_proxy_save_rsc(struct cache_detail *cd, |
16789 |
+ &expiry, GFP_KERNEL); |
16790 |
+ if (status) |
16791 |
+ goto out; |
16792 |
++ |
16793 |
++ getboottime64(&boot); |
16794 |
++ expiry -= boot.tv_sec; |
16795 |
+ } |
16796 |
+ |
16797 |
+ rsci.h.expiry_time = expiry; |
16798 |
+diff --git a/samples/bpf/Makefile b/samples/bpf/Makefile |
16799 |
+index 75d4b48601aa..85a6e8f5a75d 100644 |
16800 |
+--- a/samples/bpf/Makefile |
16801 |
++++ b/samples/bpf/Makefile |
16802 |
+@@ -221,7 +221,7 @@ all: |
16803 |
+ |
16804 |
+ clean: |
16805 |
+ $(MAKE) -C ../../ M=$(CURDIR) clean |
16806 |
+- @rm -f *~ |
16807 |
++ @find $(CURDIR) -type f -name '*~' -delete |
16808 |
+ |
16809 |
+ $(LIBBPF): FORCE |
16810 |
+ # Fix up variables inherited from Kbuild that tools/ build system won't like |
16811 |
+diff --git a/scripts/find-unused-docs.sh b/scripts/find-unused-docs.sh |
16812 |
+index 3f46f8977dc4..ee6a50e33aba 100755 |
16813 |
+--- a/scripts/find-unused-docs.sh |
16814 |
++++ b/scripts/find-unused-docs.sh |
16815 |
+@@ -54,7 +54,7 @@ for file in `find $1 -name '*.c'`; do |
16816 |
+ if [[ ${FILES_INCLUDED[$file]+_} ]]; then |
16817 |
+ continue; |
16818 |
+ fi |
16819 |
+- str=$(scripts/kernel-doc -text -export "$file" 2>/dev/null) |
16820 |
++ str=$(scripts/kernel-doc -export "$file" 2>/dev/null) |
16821 |
+ if [[ -n "$str" ]]; then |
16822 |
+ echo "$file" |
16823 |
+ fi |
16824 |
+diff --git a/sound/drivers/dummy.c b/sound/drivers/dummy.c |
16825 |
+index 9af154db530a..b78cc8d86a8b 100644 |
16826 |
+--- a/sound/drivers/dummy.c |
16827 |
++++ b/sound/drivers/dummy.c |
16828 |
+@@ -929,7 +929,7 @@ static void print_formats(struct snd_dummy *dummy, |
16829 |
+ { |
16830 |
+ int i; |
16831 |
+ |
16832 |
+- for (i = 0; i < SNDRV_PCM_FORMAT_LAST; i++) { |
16833 |
++ for (i = 0; i <= SNDRV_PCM_FORMAT_LAST; i++) { |
16834 |
+ if (dummy->pcm_hw.formats & (1ULL << i)) |
16835 |
+ snd_iprintf(buffer, " %s", snd_pcm_format_name(i)); |
16836 |
+ } |
16837 |
+diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c |
16838 |
+index d63fea5d1c92..3ee5b7b9b595 100644 |
16839 |
+--- a/sound/pci/hda/hda_intel.c |
16840 |
++++ b/sound/pci/hda/hda_intel.c |
16841 |
+@@ -2324,6 +2324,8 @@ static struct snd_pci_quirk power_save_blacklist[] = { |
16842 |
+ /* https://bugzilla.redhat.com/show_bug.cgi?id=1581607 */ |
16843 |
+ SND_PCI_QUIRK(0x1558, 0x3501, "Clevo W35xSS_370SS", 0), |
16844 |
+ /* https://bugzilla.redhat.com/show_bug.cgi?id=1525104 */ |
16845 |
++ SND_PCI_QUIRK(0x1558, 0x6504, "Clevo W65_67SB", 0), |
16846 |
++ /* https://bugzilla.redhat.com/show_bug.cgi?id=1525104 */ |
16847 |
+ SND_PCI_QUIRK(0x1028, 0x0497, "Dell Precision T3600", 0), |
16848 |
+ /* https://bugzilla.redhat.com/show_bug.cgi?id=1525104 */ |
16849 |
+ /* Note the P55A-UD3 and Z87-D3HP share the subsys id for the HDA dev */ |
16850 |
+diff --git a/sound/usb/validate.c b/sound/usb/validate.c |
16851 |
+index 389e8657434a..5a3c4f7882b0 100644 |
16852 |
+--- a/sound/usb/validate.c |
16853 |
++++ b/sound/usb/validate.c |
16854 |
+@@ -110,7 +110,7 @@ static bool validate_processing_unit(const void *p, |
16855 |
+ default: |
16856 |
+ if (v->type == UAC1_EXTENSION_UNIT) |
16857 |
+ return true; /* OK */ |
16858 |
+- switch (d->wProcessType) { |
16859 |
++ switch (le16_to_cpu(d->wProcessType)) { |
16860 |
+ case UAC_PROCESS_UP_DOWNMIX: |
16861 |
+ case UAC_PROCESS_DOLBY_PROLOGIC: |
16862 |
+ if (d->bLength < len + 1) /* bNrModes */ |
16863 |
+@@ -125,7 +125,7 @@ static bool validate_processing_unit(const void *p, |
16864 |
+ case UAC_VERSION_2: |
16865 |
+ if (v->type == UAC2_EXTENSION_UNIT_V2) |
16866 |
+ return true; /* OK */ |
16867 |
+- switch (d->wProcessType) { |
16868 |
++ switch (le16_to_cpu(d->wProcessType)) { |
16869 |
+ case UAC2_PROCESS_UP_DOWNMIX: |
16870 |
+ case UAC2_PROCESS_DOLBY_PROLOCIC: /* SiC! */ |
16871 |
+ if (d->bLength < len + 1) /* bNrModes */ |
16872 |
+@@ -142,7 +142,7 @@ static bool validate_processing_unit(const void *p, |
16873 |
+ len += 2; /* wClusterDescrID */ |
16874 |
+ break; |
16875 |
+ } |
16876 |
+- switch (d->wProcessType) { |
16877 |
++ switch (le16_to_cpu(d->wProcessType)) { |
16878 |
+ case UAC3_PROCESS_UP_DOWNMIX: |
16879 |
+ if (d->bLength < len + 1) /* bNrModes */ |
16880 |
+ return false; |
16881 |
+diff --git a/tools/kvm/kvm_stat/kvm_stat b/tools/kvm/kvm_stat/kvm_stat |
16882 |
+index ba7ee74ee533..f6ca0a216f3c 100755 |
16883 |
+--- a/tools/kvm/kvm_stat/kvm_stat |
16884 |
++++ b/tools/kvm/kvm_stat/kvm_stat |
16885 |
+@@ -271,6 +271,7 @@ class ArchX86(Arch): |
16886 |
+ def __init__(self, exit_reasons): |
16887 |
+ self.sc_perf_evt_open = 298 |
16888 |
+ self.ioctl_numbers = IOCTL_NUMBERS |
16889 |
++ self.exit_reason_field = 'exit_reason' |
16890 |
+ self.exit_reasons = exit_reasons |
16891 |
+ |
16892 |
+ def debugfs_is_child(self, field): |
16893 |
+@@ -290,6 +291,7 @@ class ArchPPC(Arch): |
16894 |
+ # numbers depend on the wordsize. |
16895 |
+ char_ptr_size = ctypes.sizeof(ctypes.c_char_p) |
16896 |
+ self.ioctl_numbers['SET_FILTER'] = 0x80002406 | char_ptr_size << 16 |
16897 |
++ self.exit_reason_field = 'exit_nr' |
16898 |
+ self.exit_reasons = {} |
16899 |
+ |
16900 |
+ def debugfs_is_child(self, field): |
16901 |
+@@ -301,6 +303,7 @@ class ArchA64(Arch): |
16902 |
+ def __init__(self): |
16903 |
+ self.sc_perf_evt_open = 241 |
16904 |
+ self.ioctl_numbers = IOCTL_NUMBERS |
16905 |
++ self.exit_reason_field = 'esr_ec' |
16906 |
+ self.exit_reasons = AARCH64_EXIT_REASONS |
16907 |
+ |
16908 |
+ def debugfs_is_child(self, field): |
16909 |
+@@ -312,6 +315,7 @@ class ArchS390(Arch): |
16910 |
+ def __init__(self): |
16911 |
+ self.sc_perf_evt_open = 331 |
16912 |
+ self.ioctl_numbers = IOCTL_NUMBERS |
16913 |
++ self.exit_reason_field = None |
16914 |
+ self.exit_reasons = None |
16915 |
+ |
16916 |
+ def debugfs_is_child(self, field): |
16917 |
+@@ -542,8 +546,8 @@ class TracepointProvider(Provider): |
16918 |
+ """ |
16919 |
+ filters = {} |
16920 |
+ filters['kvm_userspace_exit'] = ('reason', USERSPACE_EXIT_REASONS) |
16921 |
+- if ARCH.exit_reasons: |
16922 |
+- filters['kvm_exit'] = ('exit_reason', ARCH.exit_reasons) |
16923 |
++ if ARCH.exit_reason_field and ARCH.exit_reasons: |
16924 |
++ filters['kvm_exit'] = (ARCH.exit_reason_field, ARCH.exit_reasons) |
16925 |
+ return filters |
16926 |
+ |
16927 |
+ def _get_available_fields(self): |
16928 |
+diff --git a/virt/kvm/arm/aarch32.c b/virt/kvm/arm/aarch32.c |
16929 |
+index 6880236974b8..d915548759a4 100644 |
16930 |
+--- a/virt/kvm/arm/aarch32.c |
16931 |
++++ b/virt/kvm/arm/aarch32.c |
16932 |
+@@ -21,6 +21,7 @@ |
16933 |
+ * along with this program. If not, see <http://www.gnu.org/licenses/>. |
16934 |
+ */ |
16935 |
+ |
16936 |
++#include <linux/bits.h> |
16937 |
+ #include <linux/kvm_host.h> |
16938 |
+ #include <asm/kvm_emulate.h> |
16939 |
+ #include <asm/kvm_hyp.h> |
16940 |
+@@ -39,25 +40,115 @@ static const u8 return_offsets[8][2] = { |
16941 |
+ [7] = { 4, 4 }, /* FIQ, unused */ |
16942 |
+ }; |
16943 |
+ |
16944 |
++/* |
16945 |
++ * When an exception is taken, most CPSR fields are left unchanged in the |
16946 |
++ * handler. However, some are explicitly overridden (e.g. M[4:0]). |
16947 |
++ * |
16948 |
++ * The SPSR/SPSR_ELx layouts differ, and the below is intended to work with |
16949 |
++ * either format. Note: SPSR.J bit doesn't exist in SPSR_ELx, but this bit was |
16950 |
++ * obsoleted by the ARMv7 virtualization extensions and is RES0. |
16951 |
++ * |
16952 |
++ * For the SPSR layout seen from AArch32, see: |
16953 |
++ * - ARM DDI 0406C.d, page B1-1148 |
16954 |
++ * - ARM DDI 0487E.a, page G8-6264 |
16955 |
++ * |
16956 |
++ * For the SPSR_ELx layout for AArch32 seen from AArch64, see: |
16957 |
++ * - ARM DDI 0487E.a, page C5-426 |
16958 |
++ * |
16959 |
++ * Here we manipulate the fields in order of the AArch32 SPSR_ELx layout, from |
16960 |
++ * MSB to LSB. |
16961 |
++ */ |
16962 |
++static unsigned long get_except32_cpsr(struct kvm_vcpu *vcpu, u32 mode) |
16963 |
++{ |
16964 |
++ u32 sctlr = vcpu_cp15(vcpu, c1_SCTLR); |
16965 |
++ unsigned long old, new; |
16966 |
++ |
16967 |
++ old = *vcpu_cpsr(vcpu); |
16968 |
++ new = 0; |
16969 |
++ |
16970 |
++ new |= (old & PSR_AA32_N_BIT); |
16971 |
++ new |= (old & PSR_AA32_Z_BIT); |
16972 |
++ new |= (old & PSR_AA32_C_BIT); |
16973 |
++ new |= (old & PSR_AA32_V_BIT); |
16974 |
++ new |= (old & PSR_AA32_Q_BIT); |
16975 |
++ |
16976 |
++ // CPSR.IT[7:0] are set to zero upon any exception |
16977 |
++ // See ARM DDI 0487E.a, section G1.12.3 |
16978 |
++ // See ARM DDI 0406C.d, section B1.8.3 |
16979 |
++ |
16980 |
++ new |= (old & PSR_AA32_DIT_BIT); |
16981 |
++ |
16982 |
++ // CPSR.SSBS is set to SCTLR.DSSBS upon any exception |
16983 |
++ // See ARM DDI 0487E.a, page G8-6244 |
16984 |
++ if (sctlr & BIT(31)) |
16985 |
++ new |= PSR_AA32_SSBS_BIT; |
16986 |
++ |
16987 |
++ // CPSR.PAN is unchanged unless SCTLR.SPAN == 0b0 |
16988 |
++ // SCTLR.SPAN is RES1 when ARMv8.1-PAN is not implemented |
16989 |
++ // See ARM DDI 0487E.a, page G8-6246 |
16990 |
++ new |= (old & PSR_AA32_PAN_BIT); |
16991 |
++ if (!(sctlr & BIT(23))) |
16992 |
++ new |= PSR_AA32_PAN_BIT; |
16993 |
++ |
16994 |
++ // SS does not exist in AArch32, so ignore |
16995 |
++ |
16996 |
++ // CPSR.IL is set to zero upon any exception |
16997 |
++ // See ARM DDI 0487E.a, page G1-5527 |
16998 |
++ |
16999 |
++ new |= (old & PSR_AA32_GE_MASK); |
17000 |
++ |
17001 |
++ // CPSR.IT[7:0] are set to zero upon any exception |
17002 |
++ // See prior comment above |
17003 |
++ |
17004 |
++ // CPSR.E is set to SCTLR.EE upon any exception |
17005 |
++ // See ARM DDI 0487E.a, page G8-6245 |
17006 |
++ // See ARM DDI 0406C.d, page B4-1701 |
17007 |
++ if (sctlr & BIT(25)) |
17008 |
++ new |= PSR_AA32_E_BIT; |
17009 |
++ |
17010 |
++ // CPSR.A is unchanged upon an exception to Undefined, Supervisor |
17011 |
++ // CPSR.A is set upon an exception to other modes |
17012 |
++ // See ARM DDI 0487E.a, pages G1-5515 to G1-5516 |
17013 |
++ // See ARM DDI 0406C.d, page B1-1182 |
17014 |
++ new |= (old & PSR_AA32_A_BIT); |
17015 |
++ if (mode != PSR_AA32_MODE_UND && mode != PSR_AA32_MODE_SVC) |
17016 |
++ new |= PSR_AA32_A_BIT; |
17017 |
++ |
17018 |
++ // CPSR.I is set upon any exception |
17019 |
++ // See ARM DDI 0487E.a, pages G1-5515 to G1-5516 |
17020 |
++ // See ARM DDI 0406C.d, page B1-1182 |
17021 |
++ new |= PSR_AA32_I_BIT; |
17022 |
++ |
17023 |
++ // CPSR.F is set upon an exception to FIQ |
17024 |
++ // CPSR.F is unchanged upon an exception to other modes |
17025 |
++ // See ARM DDI 0487E.a, pages G1-5515 to G1-5516 |
17026 |
++ // See ARM DDI 0406C.d, page B1-1182 |
17027 |
++ new |= (old & PSR_AA32_F_BIT); |
17028 |
++ if (mode == PSR_AA32_MODE_FIQ) |
17029 |
++ new |= PSR_AA32_F_BIT; |
17030 |
++ |
17031 |
++ // CPSR.T is set to SCTLR.TE upon any exception |
17032 |
++ // See ARM DDI 0487E.a, page G8-5514 |
17033 |
++ // See ARM DDI 0406C.d, page B1-1181 |
17034 |
++ if (sctlr & BIT(30)) |
17035 |
++ new |= PSR_AA32_T_BIT; |
17036 |
++ |
17037 |
++ new |= mode; |
17038 |
++ |
17039 |
++ return new; |
17040 |
++} |
17041 |
++ |
17042 |
+ static void prepare_fault32(struct kvm_vcpu *vcpu, u32 mode, u32 vect_offset) |
17043 |
+ { |
17044 |
+- unsigned long cpsr; |
17045 |
+- unsigned long new_spsr_value = *vcpu_cpsr(vcpu); |
17046 |
+- bool is_thumb = (new_spsr_value & PSR_AA32_T_BIT); |
17047 |
++ unsigned long spsr = *vcpu_cpsr(vcpu); |
17048 |
++ bool is_thumb = (spsr & PSR_AA32_T_BIT); |
17049 |
+ u32 return_offset = return_offsets[vect_offset >> 2][is_thumb]; |
17050 |
+ u32 sctlr = vcpu_cp15(vcpu, c1_SCTLR); |
17051 |
+ |
17052 |
+- cpsr = mode | PSR_AA32_I_BIT; |
17053 |
+- |
17054 |
+- if (sctlr & (1 << 30)) |
17055 |
+- cpsr |= PSR_AA32_T_BIT; |
17056 |
+- if (sctlr & (1 << 25)) |
17057 |
+- cpsr |= PSR_AA32_E_BIT; |
17058 |
+- |
17059 |
+- *vcpu_cpsr(vcpu) = cpsr; |
17060 |
++ *vcpu_cpsr(vcpu) = get_except32_cpsr(vcpu, mode); |
17061 |
+ |
17062 |
+ /* Note: These now point to the banked copies */ |
17063 |
+- vcpu_write_spsr(vcpu, new_spsr_value); |
17064 |
++ vcpu_write_spsr(vcpu, host_spsr_to_spsr32(spsr)); |
17065 |
+ *vcpu_reg32(vcpu, 14) = *vcpu_pc(vcpu) + return_offset; |
17066 |
+ |
17067 |
+ /* Branch to exception vector */ |
17068 |
+@@ -95,7 +186,7 @@ static void inject_abt32(struct kvm_vcpu *vcpu, bool is_pabt, |
17069 |
+ fsr = &vcpu_cp15(vcpu, c5_DFSR); |
17070 |
+ } |
17071 |
+ |
17072 |
+- prepare_fault32(vcpu, PSR_AA32_MODE_ABT | PSR_AA32_A_BIT, vect_offset); |
17073 |
++ prepare_fault32(vcpu, PSR_AA32_MODE_ABT, vect_offset); |
17074 |
+ |
17075 |
+ *far = addr; |
17076 |
+ |
17077 |
+diff --git a/virt/kvm/arm/mmio.c b/virt/kvm/arm/mmio.c |
17078 |
+index 3caee91bca08..878e0edb2e1b 100644 |
17079 |
+--- a/virt/kvm/arm/mmio.c |
17080 |
++++ b/virt/kvm/arm/mmio.c |
17081 |
+@@ -117,6 +117,9 @@ int kvm_handle_mmio_return(struct kvm_vcpu *vcpu, struct kvm_run *run) |
17082 |
+ data = (data ^ mask) - mask; |
17083 |
+ } |
17084 |
+ |
17085 |
++ if (!vcpu->arch.mmio_decode.sixty_four) |
17086 |
++ data = data & 0xffffffff; |
17087 |
++ |
17088 |
+ trace_kvm_mmio(KVM_TRACE_MMIO_READ, len, run->mmio.phys_addr, |
17089 |
+ &data); |
17090 |
+ data = vcpu_data_host_to_guest(vcpu, data, len); |
17091 |
+@@ -137,6 +140,7 @@ static int decode_hsr(struct kvm_vcpu *vcpu, bool *is_write, int *len) |
17092 |
+ unsigned long rt; |
17093 |
+ int access_size; |
17094 |
+ bool sign_extend; |
17095 |
++ bool sixty_four; |
17096 |
+ |
17097 |
+ if (kvm_vcpu_dabt_iss1tw(vcpu)) { |
17098 |
+ /* page table accesses IO mem: tell guest to fix its TTBR */ |
17099 |
+@@ -150,11 +154,13 @@ static int decode_hsr(struct kvm_vcpu *vcpu, bool *is_write, int *len) |
17100 |
+ |
17101 |
+ *is_write = kvm_vcpu_dabt_iswrite(vcpu); |
17102 |
+ sign_extend = kvm_vcpu_dabt_issext(vcpu); |
17103 |
++ sixty_four = kvm_vcpu_dabt_issf(vcpu); |
17104 |
+ rt = kvm_vcpu_dabt_get_rd(vcpu); |
17105 |
+ |
17106 |
+ *len = access_size; |
17107 |
+ vcpu->arch.mmio_decode.sign_extend = sign_extend; |
17108 |
+ vcpu->arch.mmio_decode.rt = rt; |
17109 |
++ vcpu->arch.mmio_decode.sixty_four = sixty_four; |
17110 |
+ |
17111 |
+ return 0; |
17112 |
+ } |
17113 |
+diff --git a/virt/kvm/async_pf.c b/virt/kvm/async_pf.c |
17114 |
+index 23c2519c5b32..c9861c2315e8 100644 |
17115 |
+--- a/virt/kvm/async_pf.c |
17116 |
++++ b/virt/kvm/async_pf.c |
17117 |
+@@ -76,7 +76,7 @@ static void async_pf_execute(struct work_struct *work) |
17118 |
+ struct mm_struct *mm = apf->mm; |
17119 |
+ struct kvm_vcpu *vcpu = apf->vcpu; |
17120 |
+ unsigned long addr = apf->addr; |
17121 |
+- gva_t gva = apf->gva; |
17122 |
++ gpa_t cr2_or_gpa = apf->cr2_or_gpa; |
17123 |
+ int locked = 1; |
17124 |
+ |
17125 |
+ might_sleep(); |
17126 |
+@@ -104,7 +104,7 @@ static void async_pf_execute(struct work_struct *work) |
17127 |
+ * this point |
17128 |
+ */ |
17129 |
+ |
17130 |
+- trace_kvm_async_pf_completed(addr, gva); |
17131 |
++ trace_kvm_async_pf_completed(addr, cr2_or_gpa); |
17132 |
+ |
17133 |
+ if (swq_has_sleeper(&vcpu->wq)) |
17134 |
+ swake_up_one(&vcpu->wq); |
17135 |
+@@ -177,8 +177,8 @@ void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu) |
17136 |
+ } |
17137 |
+ } |
17138 |
+ |
17139 |
+-int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva, |
17140 |
+- struct kvm_arch_async_pf *arch) |
17141 |
++int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gpa_t cr2_or_gpa, |
17142 |
++ unsigned long hva, struct kvm_arch_async_pf *arch) |
17143 |
+ { |
17144 |
+ struct kvm_async_pf *work; |
17145 |
+ |
17146 |
+@@ -197,7 +197,7 @@ int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, unsigned long hva, |
17147 |
+ |
17148 |
+ work->wakeup_all = false; |
17149 |
+ work->vcpu = vcpu; |
17150 |
+- work->gva = gva; |
17151 |
++ work->cr2_or_gpa = cr2_or_gpa; |
17152 |
+ work->addr = hva; |
17153 |
+ work->arch = *arch; |
17154 |
+ work->mm = current->mm; |
17155 |
+diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c |
17156 |
+index 9502b1a44232..beec19fcf8cd 100644 |
17157 |
+--- a/virt/kvm/kvm_main.c |
17158 |
++++ b/virt/kvm/kvm_main.c |
17159 |
+@@ -1294,14 +1294,14 @@ bool kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
17160 |
+ } |
17161 |
+ EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); |
17162 |
+ |
17163 |
+-unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
17164 |
++unsigned long kvm_host_page_size(struct kvm_vcpu *vcpu, gfn_t gfn) |
17165 |
+ { |
17166 |
+ struct vm_area_struct *vma; |
17167 |
+ unsigned long addr, size; |
17168 |
+ |
17169 |
+ size = PAGE_SIZE; |
17170 |
+ |
17171 |
+- addr = gfn_to_hva(kvm, gfn); |
17172 |
++ addr = kvm_vcpu_gfn_to_hva_prot(vcpu, gfn, NULL); |
17173 |
+ if (kvm_is_error_hva(addr)) |
17174 |
+ return PAGE_SIZE; |
17175 |
+ |