1 |
commit: 86fa72f29ab4a860912cb34ce60fbd2053ddf2d3 |
2 |
Author: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
3 |
AuthorDate: Mon Feb 24 11:06:35 2020 +0000 |
4 |
Commit: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
5 |
CommitDate: Mon Feb 24 11:06:35 2020 +0000 |
6 |
URL: https://gitweb.gentoo.org/proj/linux-patches.git/commit/?id=86fa72f2 |
7 |
|
8 |
Linux patch 4.19.106 |
9 |
|
10 |
Signed-off-by: Mike Pagano <mpagano <AT> gentoo.org> |
11 |
|
12 |
0000_README | 4 + |
13 |
1105_linux-4.19.106.patch | 13713 ++++++++++++++++++++++++++++++++++++++++++++ |
14 |
2 files changed, 13717 insertions(+) |
15 |
|
16 |
diff --git a/0000_README b/0000_README |
17 |
index f83fdcc..3213eab 100644 |
18 |
--- a/0000_README |
19 |
+++ b/0000_README |
20 |
@@ -459,6 +459,10 @@ Patch: 1104_linux-4.19.105.patch |
21 |
From: https://www.kernel.org |
22 |
Desc: Linux 4.19.105 |
23 |
|
24 |
+Patch: 1105_linux-4.19.106.patch |
25 |
+From: https://www.kernel.org |
26 |
+Desc: Linux 4.19.106 |
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/1105_linux-4.19.106.patch b/1105_linux-4.19.106.patch |
33 |
new file mode 100644 |
34 |
index 0000000..ea910af |
35 |
--- /dev/null |
36 |
+++ b/1105_linux-4.19.106.patch |
37 |
@@ -0,0 +1,13713 @@ |
38 |
+diff --git a/Makefile b/Makefile |
39 |
+index eef7de60cd94..c010fd4a3286 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 = 105 |
47 |
++SUBLEVEL = 106 |
48 |
+ EXTRAVERSION = |
49 |
+ NAME = "People's Front" |
50 |
+ |
51 |
+diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig |
52 |
+index 185e552f1461..e2f7c50dbace 100644 |
53 |
+--- a/arch/arm/Kconfig |
54 |
++++ b/arch/arm/Kconfig |
55 |
+@@ -61,7 +61,7 @@ config ARM |
56 |
+ select HAVE_EBPF_JIT if !CPU_ENDIAN_BE32 |
57 |
+ select HAVE_CONTEXT_TRACKING |
58 |
+ select HAVE_C_RECORDMCOUNT |
59 |
+- select HAVE_DEBUG_KMEMLEAK |
60 |
++ select HAVE_DEBUG_KMEMLEAK if !XIP_KERNEL |
61 |
+ select HAVE_DMA_CONTIGUOUS if MMU |
62 |
+ select HAVE_DYNAMIC_FTRACE if (!XIP_KERNEL) && !CPU_ENDIAN_BE32 && MMU |
63 |
+ select HAVE_DYNAMIC_FTRACE_WITH_REGS if HAVE_DYNAMIC_FTRACE |
64 |
+@@ -2008,7 +2008,7 @@ config XIP_DEFLATED_DATA |
65 |
+ config KEXEC |
66 |
+ bool "Kexec system call (EXPERIMENTAL)" |
67 |
+ depends on (!SMP || PM_SLEEP_SMP) |
68 |
+- depends on !CPU_V7M |
69 |
++ depends on MMU |
70 |
+ select KEXEC_CORE |
71 |
+ help |
72 |
+ kexec is a system call that implements the ability to shutdown your |
73 |
+diff --git a/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi b/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi |
74 |
+index 315d0e7615f3..bc5f2de02d43 100644 |
75 |
+--- a/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi |
76 |
++++ b/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi |
77 |
+@@ -657,7 +657,7 @@ |
78 |
+ pinctrl-0 = <&pinctrl_usdhc2>; |
79 |
+ bus-width = <4>; |
80 |
+ cd-gpios = <&gpio2 2 GPIO_ACTIVE_LOW>; |
81 |
+- wp-gpios = <&gpio2 3 GPIO_ACTIVE_HIGH>; |
82 |
++ disable-wp; |
83 |
+ vmmc-supply = <®_3p3v_sd>; |
84 |
+ vqmmc-supply = <®_3p3v>; |
85 |
+ no-1-8-v; |
86 |
+@@ -670,7 +670,7 @@ |
87 |
+ pinctrl-0 = <&pinctrl_usdhc3>; |
88 |
+ bus-width = <4>; |
89 |
+ cd-gpios = <&gpio2 0 GPIO_ACTIVE_LOW>; |
90 |
+- wp-gpios = <&gpio2 1 GPIO_ACTIVE_HIGH>; |
91 |
++ disable-wp; |
92 |
+ vmmc-supply = <®_3p3v_sd>; |
93 |
+ vqmmc-supply = <®_3p3v>; |
94 |
+ no-1-8-v; |
95 |
+@@ -804,6 +804,7 @@ |
96 |
+ &usbh1 { |
97 |
+ vbus-supply = <®_5p0v_main>; |
98 |
+ disable-over-current; |
99 |
++ maximum-speed = "full-speed"; |
100 |
+ status = "okay"; |
101 |
+ }; |
102 |
+ |
103 |
+@@ -1081,7 +1082,6 @@ |
104 |
+ MX6QDL_PAD_SD2_DAT1__SD2_DATA1 0x17059 |
105 |
+ MX6QDL_PAD_SD2_DAT2__SD2_DATA2 0x17059 |
106 |
+ MX6QDL_PAD_SD2_DAT3__SD2_DATA3 0x17059 |
107 |
+- MX6QDL_PAD_NANDF_D3__GPIO2_IO03 0x40010040 |
108 |
+ MX6QDL_PAD_NANDF_D2__GPIO2_IO02 0x40010040 |
109 |
+ >; |
110 |
+ }; |
111 |
+@@ -1094,7 +1094,6 @@ |
112 |
+ MX6QDL_PAD_SD3_DAT1__SD3_DATA1 0x17059 |
113 |
+ MX6QDL_PAD_SD3_DAT2__SD3_DATA2 0x17059 |
114 |
+ MX6QDL_PAD_SD3_DAT3__SD3_DATA3 0x17059 |
115 |
+- MX6QDL_PAD_NANDF_D1__GPIO2_IO01 0x40010040 |
116 |
+ MX6QDL_PAD_NANDF_D0__GPIO2_IO00 0x40010040 |
117 |
+ |
118 |
+ >; |
119 |
+diff --git a/arch/arm/boot/dts/r8a7779.dtsi b/arch/arm/boot/dts/r8a7779.dtsi |
120 |
+index 03919714645a..f1c9b2bc542c 100644 |
121 |
+--- a/arch/arm/boot/dts/r8a7779.dtsi |
122 |
++++ b/arch/arm/boot/dts/r8a7779.dtsi |
123 |
+@@ -68,6 +68,14 @@ |
124 |
+ <0xf0000100 0x100>; |
125 |
+ }; |
126 |
+ |
127 |
++ timer@f0000200 { |
128 |
++ compatible = "arm,cortex-a9-global-timer"; |
129 |
++ reg = <0xf0000200 0x100>; |
130 |
++ interrupts = <GIC_PPI 11 |
131 |
++ (GIC_CPU_MASK_SIMPLE(4) | IRQ_TYPE_EDGE_RISING)>; |
132 |
++ clocks = <&cpg_clocks R8A7779_CLK_ZS>; |
133 |
++ }; |
134 |
++ |
135 |
+ timer@f0000600 { |
136 |
+ compatible = "arm,cortex-a9-twd-timer"; |
137 |
+ reg = <0xf0000600 0x20>; |
138 |
+diff --git a/arch/arm/boot/dts/stm32f469-disco.dts b/arch/arm/boot/dts/stm32f469-disco.dts |
139 |
+index 3ee768cb86fc..eea979ef5512 100644 |
140 |
+--- a/arch/arm/boot/dts/stm32f469-disco.dts |
141 |
++++ b/arch/arm/boot/dts/stm32f469-disco.dts |
142 |
+@@ -75,6 +75,13 @@ |
143 |
+ regulator-max-microvolt = <3300000>; |
144 |
+ }; |
145 |
+ |
146 |
++ vdd_dsi: vdd-dsi { |
147 |
++ compatible = "regulator-fixed"; |
148 |
++ regulator-name = "vdd_dsi"; |
149 |
++ regulator-min-microvolt = <3300000>; |
150 |
++ regulator-max-microvolt = <3300000>; |
151 |
++ }; |
152 |
++ |
153 |
+ soc { |
154 |
+ dma-ranges = <0xc0000000 0x0 0x10000000>; |
155 |
+ }; |
156 |
+@@ -154,6 +161,7 @@ |
157 |
+ compatible = "orisetech,otm8009a"; |
158 |
+ reg = <0>; /* dsi virtual channel (0..3) */ |
159 |
+ reset-gpios = <&gpioh 7 GPIO_ACTIVE_LOW>; |
160 |
++ power-supply = <&vdd_dsi>; |
161 |
+ status = "okay"; |
162 |
+ |
163 |
+ port { |
164 |
+diff --git a/arch/arm/boot/dts/sun8i-h3.dtsi b/arch/arm/boot/dts/sun8i-h3.dtsi |
165 |
+index 9233ba30a857..11172fbdc03a 100644 |
166 |
+--- a/arch/arm/boot/dts/sun8i-h3.dtsi |
167 |
++++ b/arch/arm/boot/dts/sun8i-h3.dtsi |
168 |
+@@ -80,7 +80,7 @@ |
169 |
+ #cooling-cells = <2>; |
170 |
+ }; |
171 |
+ |
172 |
+- cpu@1 { |
173 |
++ cpu1: cpu@1 { |
174 |
+ compatible = "arm,cortex-a7"; |
175 |
+ device_type = "cpu"; |
176 |
+ reg = <1>; |
177 |
+@@ -90,7 +90,7 @@ |
178 |
+ #cooling-cells = <2>; |
179 |
+ }; |
180 |
+ |
181 |
+- cpu@2 { |
182 |
++ cpu2: cpu@2 { |
183 |
+ compatible = "arm,cortex-a7"; |
184 |
+ device_type = "cpu"; |
185 |
+ reg = <2>; |
186 |
+@@ -100,7 +100,7 @@ |
187 |
+ #cooling-cells = <2>; |
188 |
+ }; |
189 |
+ |
190 |
+- cpu@3 { |
191 |
++ cpu3: cpu@3 { |
192 |
+ compatible = "arm,cortex-a7"; |
193 |
+ device_type = "cpu"; |
194 |
+ reg = <3>; |
195 |
+@@ -111,6 +111,15 @@ |
196 |
+ }; |
197 |
+ }; |
198 |
+ |
199 |
++ pmu { |
200 |
++ compatible = "arm,cortex-a7-pmu"; |
201 |
++ interrupts = <GIC_SPI 120 IRQ_TYPE_LEVEL_HIGH>, |
202 |
++ <GIC_SPI 121 IRQ_TYPE_LEVEL_HIGH>, |
203 |
++ <GIC_SPI 122 IRQ_TYPE_LEVEL_HIGH>, |
204 |
++ <GIC_SPI 123 IRQ_TYPE_LEVEL_HIGH>; |
205 |
++ interrupt-affinity = <&cpu0>, <&cpu1>, <&cpu2>, <&cpu3>; |
206 |
++ }; |
207 |
++ |
208 |
+ timer { |
209 |
+ compatible = "arm,armv7-timer"; |
210 |
+ interrupts = <GIC_PPI 13 (GIC_CPU_MASK_SIMPLE(4) | IRQ_TYPE_LEVEL_LOW)>, |
211 |
+diff --git a/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi b/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi |
212 |
+index 72813e7aefb8..bd4391269611 100644 |
213 |
+--- a/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi |
214 |
++++ b/arch/arm64/boot/dts/allwinner/sun50i-h6.dtsi |
215 |
+@@ -69,6 +69,16 @@ |
216 |
+ clock-output-names = "osc32k"; |
217 |
+ }; |
218 |
+ |
219 |
++ pmu { |
220 |
++ compatible = "arm,cortex-a53-pmu", |
221 |
++ "arm,armv8-pmuv3"; |
222 |
++ interrupts = <GIC_SPI 140 IRQ_TYPE_LEVEL_HIGH>, |
223 |
++ <GIC_SPI 141 IRQ_TYPE_LEVEL_HIGH>, |
224 |
++ <GIC_SPI 142 IRQ_TYPE_LEVEL_HIGH>, |
225 |
++ <GIC_SPI 143 IRQ_TYPE_LEVEL_HIGH>; |
226 |
++ interrupt-affinity = <&cpu0>, <&cpu1>, <&cpu2>, <&cpu3>; |
227 |
++ }; |
228 |
++ |
229 |
+ psci { |
230 |
+ compatible = "arm,psci-0.2"; |
231 |
+ method = "smc"; |
232 |
+diff --git a/arch/arm64/boot/dts/qcom/msm8996.dtsi b/arch/arm64/boot/dts/qcom/msm8996.dtsi |
233 |
+index 8c86c41a0d25..3e7baabf6450 100644 |
234 |
+--- a/arch/arm64/boot/dts/qcom/msm8996.dtsi |
235 |
++++ b/arch/arm64/boot/dts/qcom/msm8996.dtsi |
236 |
+@@ -918,6 +918,8 @@ |
237 |
+ interrupts = <0 138 IRQ_TYPE_LEVEL_HIGH>; |
238 |
+ phys = <&hsusb_phy2>; |
239 |
+ phy-names = "usb2-phy"; |
240 |
++ snps,dis_u2_susphy_quirk; |
241 |
++ snps,dis_enblslpm_quirk; |
242 |
+ }; |
243 |
+ }; |
244 |
+ |
245 |
+@@ -947,6 +949,8 @@ |
246 |
+ interrupts = <0 131 IRQ_TYPE_LEVEL_HIGH>; |
247 |
+ phys = <&hsusb_phy1>, <&ssusb_phy_0>; |
248 |
+ phy-names = "usb2-phy", "usb3-phy"; |
249 |
++ snps,dis_u2_susphy_quirk; |
250 |
++ snps,dis_enblslpm_quirk; |
251 |
+ }; |
252 |
+ }; |
253 |
+ |
254 |
+diff --git a/arch/arm64/include/asm/alternative.h b/arch/arm64/include/asm/alternative.h |
255 |
+index 4b650ec1d7dd..887a8512bf10 100644 |
256 |
+--- a/arch/arm64/include/asm/alternative.h |
257 |
++++ b/arch/arm64/include/asm/alternative.h |
258 |
+@@ -35,13 +35,16 @@ void apply_alternatives_module(void *start, size_t length); |
259 |
+ static inline void apply_alternatives_module(void *start, size_t length) { } |
260 |
+ #endif |
261 |
+ |
262 |
+-#define ALTINSTR_ENTRY(feature,cb) \ |
263 |
++#define ALTINSTR_ENTRY(feature) \ |
264 |
+ " .word 661b - .\n" /* label */ \ |
265 |
+- " .if " __stringify(cb) " == 0\n" \ |
266 |
+ " .word 663f - .\n" /* new instruction */ \ |
267 |
+- " .else\n" \ |
268 |
++ " .hword " __stringify(feature) "\n" /* feature bit */ \ |
269 |
++ " .byte 662b-661b\n" /* source len */ \ |
270 |
++ " .byte 664f-663f\n" /* replacement len */ |
271 |
++ |
272 |
++#define ALTINSTR_ENTRY_CB(feature, cb) \ |
273 |
++ " .word 661b - .\n" /* label */ \ |
274 |
+ " .word " __stringify(cb) "- .\n" /* callback */ \ |
275 |
+- " .endif\n" \ |
276 |
+ " .hword " __stringify(feature) "\n" /* feature bit */ \ |
277 |
+ " .byte 662b-661b\n" /* source len */ \ |
278 |
+ " .byte 664f-663f\n" /* replacement len */ |
279 |
+@@ -62,15 +65,14 @@ static inline void apply_alternatives_module(void *start, size_t length) { } |
280 |
+ * |
281 |
+ * Alternatives with callbacks do not generate replacement instructions. |
282 |
+ */ |
283 |
+-#define __ALTERNATIVE_CFG(oldinstr, newinstr, feature, cfg_enabled, cb) \ |
284 |
++#define __ALTERNATIVE_CFG(oldinstr, newinstr, feature, cfg_enabled) \ |
285 |
+ ".if "__stringify(cfg_enabled)" == 1\n" \ |
286 |
+ "661:\n\t" \ |
287 |
+ oldinstr "\n" \ |
288 |
+ "662:\n" \ |
289 |
+ ".pushsection .altinstructions,\"a\"\n" \ |
290 |
+- ALTINSTR_ENTRY(feature,cb) \ |
291 |
++ ALTINSTR_ENTRY(feature) \ |
292 |
+ ".popsection\n" \ |
293 |
+- " .if " __stringify(cb) " == 0\n" \ |
294 |
+ ".pushsection .altinstr_replacement, \"a\"\n" \ |
295 |
+ "663:\n\t" \ |
296 |
+ newinstr "\n" \ |
297 |
+@@ -78,17 +80,25 @@ static inline void apply_alternatives_module(void *start, size_t length) { } |
298 |
+ ".popsection\n\t" \ |
299 |
+ ".org . - (664b-663b) + (662b-661b)\n\t" \ |
300 |
+ ".org . - (662b-661b) + (664b-663b)\n" \ |
301 |
+- ".else\n\t" \ |
302 |
++ ".endif\n" |
303 |
++ |
304 |
++#define __ALTERNATIVE_CFG_CB(oldinstr, feature, cfg_enabled, cb) \ |
305 |
++ ".if "__stringify(cfg_enabled)" == 1\n" \ |
306 |
++ "661:\n\t" \ |
307 |
++ oldinstr "\n" \ |
308 |
++ "662:\n" \ |
309 |
++ ".pushsection .altinstructions,\"a\"\n" \ |
310 |
++ ALTINSTR_ENTRY_CB(feature, cb) \ |
311 |
++ ".popsection\n" \ |
312 |
+ "663:\n\t" \ |
313 |
+ "664:\n\t" \ |
314 |
+- ".endif\n" \ |
315 |
+ ".endif\n" |
316 |
+ |
317 |
+ #define _ALTERNATIVE_CFG(oldinstr, newinstr, feature, cfg, ...) \ |
318 |
+- __ALTERNATIVE_CFG(oldinstr, newinstr, feature, IS_ENABLED(cfg), 0) |
319 |
++ __ALTERNATIVE_CFG(oldinstr, newinstr, feature, IS_ENABLED(cfg)) |
320 |
+ |
321 |
+ #define ALTERNATIVE_CB(oldinstr, cb) \ |
322 |
+- __ALTERNATIVE_CFG(oldinstr, "NOT_AN_INSTRUCTION", ARM64_CB_PATCH, 1, cb) |
323 |
++ __ALTERNATIVE_CFG_CB(oldinstr, ARM64_CB_PATCH, 1, cb) |
324 |
+ #else |
325 |
+ |
326 |
+ #include <asm/assembler.h> |
327 |
+diff --git a/arch/microblaze/kernel/cpu/cache.c b/arch/microblaze/kernel/cpu/cache.c |
328 |
+index 0bde47e4fa69..dcba53803fa5 100644 |
329 |
+--- a/arch/microblaze/kernel/cpu/cache.c |
330 |
++++ b/arch/microblaze/kernel/cpu/cache.c |
331 |
+@@ -92,7 +92,8 @@ static inline void __disable_dcache_nomsr(void) |
332 |
+ #define CACHE_LOOP_LIMITS(start, end, cache_line_length, cache_size) \ |
333 |
+ do { \ |
334 |
+ int align = ~(cache_line_length - 1); \ |
335 |
+- end = min(start + cache_size, end); \ |
336 |
++ if (start < UINT_MAX - cache_size) \ |
337 |
++ end = min(start + cache_size, end); \ |
338 |
+ start &= align; \ |
339 |
+ } while (0) |
340 |
+ |
341 |
+diff --git a/arch/mips/loongson64/loongson-3/platform.c b/arch/mips/loongson64/loongson-3/platform.c |
342 |
+index 25a97cc0ee33..0db4cc3196eb 100644 |
343 |
+--- a/arch/mips/loongson64/loongson-3/platform.c |
344 |
++++ b/arch/mips/loongson64/loongson-3/platform.c |
345 |
+@@ -31,6 +31,9 @@ static int __init loongson3_platform_init(void) |
346 |
+ continue; |
347 |
+ |
348 |
+ pdev = kzalloc(sizeof(struct platform_device), GFP_KERNEL); |
349 |
++ if (!pdev) |
350 |
++ return -ENOMEM; |
351 |
++ |
352 |
+ pdev->name = loongson_sysconf.sensors[i].name; |
353 |
+ pdev->id = loongson_sysconf.sensors[i].id; |
354 |
+ pdev->dev.platform_data = &loongson_sysconf.sensors[i]; |
355 |
+diff --git a/arch/powerpc/kernel/eeh_driver.c b/arch/powerpc/kernel/eeh_driver.c |
356 |
+index af1f3d5f9a0f..377d23f58197 100644 |
357 |
+--- a/arch/powerpc/kernel/eeh_driver.c |
358 |
++++ b/arch/powerpc/kernel/eeh_driver.c |
359 |
+@@ -554,12 +554,6 @@ static void *eeh_rmv_device(struct eeh_dev *edev, void *userdata) |
360 |
+ |
361 |
+ pci_iov_remove_virtfn(edev->physfn, pdn->vf_index); |
362 |
+ edev->pdev = NULL; |
363 |
+- |
364 |
+- /* |
365 |
+- * We have to set the VF PE number to invalid one, which is |
366 |
+- * required to plug the VF successfully. |
367 |
+- */ |
368 |
+- pdn->pe_number = IODA_INVALID_PE; |
369 |
+ #endif |
370 |
+ if (rmv_data) |
371 |
+ list_add(&edev->rmv_list, &rmv_data->edev_list); |
372 |
+diff --git a/arch/powerpc/kernel/pci_dn.c b/arch/powerpc/kernel/pci_dn.c |
373 |
+index ab147a1909c8..7cecc3bd953b 100644 |
374 |
+--- a/arch/powerpc/kernel/pci_dn.c |
375 |
++++ b/arch/powerpc/kernel/pci_dn.c |
376 |
+@@ -257,9 +257,22 @@ void remove_dev_pci_data(struct pci_dev *pdev) |
377 |
+ continue; |
378 |
+ |
379 |
+ #ifdef CONFIG_EEH |
380 |
+- /* Release EEH device for the VF */ |
381 |
++ /* |
382 |
++ * Release EEH state for this VF. The PCI core |
383 |
++ * has already torn down the pci_dev for this VF, but |
384 |
++ * we're responsible to removing the eeh_dev since it |
385 |
++ * has the same lifetime as the pci_dn that spawned it. |
386 |
++ */ |
387 |
+ edev = pdn_to_eeh_dev(pdn); |
388 |
+ if (edev) { |
389 |
++ /* |
390 |
++ * We allocate pci_dn's for the totalvfs count, |
391 |
++ * but only only the vfs that were activated |
392 |
++ * have a configured PE. |
393 |
++ */ |
394 |
++ if (edev->pe) |
395 |
++ eeh_rmv_from_parent_pe(edev); |
396 |
++ |
397 |
+ pdn->edev = NULL; |
398 |
+ kfree(edev); |
399 |
+ } |
400 |
+diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c |
401 |
+index ee63749a2d47..ecd211c5f24a 100644 |
402 |
+--- a/arch/powerpc/platforms/powernv/pci-ioda.c |
403 |
++++ b/arch/powerpc/platforms/powernv/pci-ioda.c |
404 |
+@@ -1552,6 +1552,10 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs) |
405 |
+ |
406 |
+ /* Reserve PE for each VF */ |
407 |
+ for (vf_index = 0; vf_index < num_vfs; vf_index++) { |
408 |
++ int vf_devfn = pci_iov_virtfn_devfn(pdev, vf_index); |
409 |
++ int vf_bus = pci_iov_virtfn_bus(pdev, vf_index); |
410 |
++ struct pci_dn *vf_pdn; |
411 |
++ |
412 |
+ if (pdn->m64_single_mode) |
413 |
+ pe_num = pdn->pe_num_map[vf_index]; |
414 |
+ else |
415 |
+@@ -1564,13 +1568,11 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs) |
416 |
+ pe->pbus = NULL; |
417 |
+ pe->parent_dev = pdev; |
418 |
+ pe->mve_number = -1; |
419 |
+- pe->rid = (pci_iov_virtfn_bus(pdev, vf_index) << 8) | |
420 |
+- pci_iov_virtfn_devfn(pdev, vf_index); |
421 |
++ pe->rid = (vf_bus << 8) | vf_devfn; |
422 |
+ |
423 |
+ pe_info(pe, "VF %04d:%02d:%02d.%d associated with PE#%x\n", |
424 |
+ hose->global_number, pdev->bus->number, |
425 |
+- PCI_SLOT(pci_iov_virtfn_devfn(pdev, vf_index)), |
426 |
+- PCI_FUNC(pci_iov_virtfn_devfn(pdev, vf_index)), pe_num); |
427 |
++ PCI_SLOT(vf_devfn), PCI_FUNC(vf_devfn), pe_num); |
428 |
+ |
429 |
+ if (pnv_ioda_configure_pe(phb, pe)) { |
430 |
+ /* XXX What do we do here ? */ |
431 |
+@@ -1584,6 +1586,15 @@ static void pnv_ioda_setup_vf_PE(struct pci_dev *pdev, u16 num_vfs) |
432 |
+ list_add_tail(&pe->list, &phb->ioda.pe_list); |
433 |
+ mutex_unlock(&phb->ioda.pe_list_mutex); |
434 |
+ |
435 |
++ /* associate this pe to it's pdn */ |
436 |
++ list_for_each_entry(vf_pdn, &pdn->parent->child_list, list) { |
437 |
++ if (vf_pdn->busno == vf_bus && |
438 |
++ vf_pdn->devfn == vf_devfn) { |
439 |
++ vf_pdn->pe_number = pe_num; |
440 |
++ break; |
441 |
++ } |
442 |
++ } |
443 |
++ |
444 |
+ pnv_pci_ioda2_setup_dma_pe(phb, pe); |
445 |
+ } |
446 |
+ } |
447 |
+@@ -3004,9 +3015,6 @@ static void pnv_pci_ioda_fixup_iov_resources(struct pci_dev *pdev) |
448 |
+ struct pci_dn *pdn; |
449 |
+ int mul, total_vfs; |
450 |
+ |
451 |
+- if (!pdev->is_physfn || pci_dev_is_added(pdev)) |
452 |
+- return; |
453 |
+- |
454 |
+ pdn = pci_get_pdn(pdev); |
455 |
+ pdn->vfs_expanded = 0; |
456 |
+ pdn->m64_single_mode = false; |
457 |
+@@ -3081,6 +3089,30 @@ truncate_iov: |
458 |
+ res->end = res->start - 1; |
459 |
+ } |
460 |
+ } |
461 |
++ |
462 |
++static void pnv_pci_ioda_fixup_iov(struct pci_dev *pdev) |
463 |
++{ |
464 |
++ if (WARN_ON(pci_dev_is_added(pdev))) |
465 |
++ return; |
466 |
++ |
467 |
++ if (pdev->is_virtfn) { |
468 |
++ struct pnv_ioda_pe *pe = pnv_ioda_get_pe(pdev); |
469 |
++ |
470 |
++ /* |
471 |
++ * VF PEs are single-device PEs so their pdev pointer needs to |
472 |
++ * be set. The pdev doesn't exist when the PE is allocated (in |
473 |
++ * (pcibios_sriov_enable()) so we fix it up here. |
474 |
++ */ |
475 |
++ pe->pdev = pdev; |
476 |
++ WARN_ON(!(pe->flags & PNV_IODA_PE_VF)); |
477 |
++ } else if (pdev->is_physfn) { |
478 |
++ /* |
479 |
++ * For PFs adjust their allocated IOV resources to match what |
480 |
++ * the PHB can support using it's M64 BAR table. |
481 |
++ */ |
482 |
++ pnv_pci_ioda_fixup_iov_resources(pdev); |
483 |
++ } |
484 |
++} |
485 |
+ #endif /* CONFIG_PCI_IOV */ |
486 |
+ |
487 |
+ static void pnv_ioda_setup_pe_res(struct pnv_ioda_pe *pe, |
488 |
+@@ -3974,7 +4006,7 @@ static void __init pnv_pci_init_ioda_phb(struct device_node *np, |
489 |
+ ppc_md.pcibios_default_alignment = pnv_pci_default_alignment; |
490 |
+ |
491 |
+ #ifdef CONFIG_PCI_IOV |
492 |
+- ppc_md.pcibios_fixup_sriov = pnv_pci_ioda_fixup_iov_resources; |
493 |
++ ppc_md.pcibios_fixup_sriov = pnv_pci_ioda_fixup_iov; |
494 |
+ ppc_md.pcibios_iov_resource_alignment = pnv_pci_iov_resource_alignment; |
495 |
+ ppc_md.pcibios_sriov_enable = pnv_pcibios_sriov_enable; |
496 |
+ ppc_md.pcibios_sriov_disable = pnv_pcibios_sriov_disable; |
497 |
+diff --git a/arch/powerpc/platforms/powernv/pci.c b/arch/powerpc/platforms/powernv/pci.c |
498 |
+index c846300b7836..b6fa900af5da 100644 |
499 |
+--- a/arch/powerpc/platforms/powernv/pci.c |
500 |
++++ b/arch/powerpc/platforms/powernv/pci.c |
501 |
+@@ -820,24 +820,6 @@ void pnv_pci_dma_dev_setup(struct pci_dev *pdev) |
502 |
+ { |
503 |
+ struct pci_controller *hose = pci_bus_to_host(pdev->bus); |
504 |
+ struct pnv_phb *phb = hose->private_data; |
505 |
+-#ifdef CONFIG_PCI_IOV |
506 |
+- struct pnv_ioda_pe *pe; |
507 |
+- struct pci_dn *pdn; |
508 |
+- |
509 |
+- /* Fix the VF pdn PE number */ |
510 |
+- if (pdev->is_virtfn) { |
511 |
+- pdn = pci_get_pdn(pdev); |
512 |
+- WARN_ON(pdn->pe_number != IODA_INVALID_PE); |
513 |
+- list_for_each_entry(pe, &phb->ioda.pe_list, list) { |
514 |
+- if (pe->rid == ((pdev->bus->number << 8) | |
515 |
+- (pdev->devfn & 0xff))) { |
516 |
+- pdn->pe_number = pe->pe_number; |
517 |
+- pe->pdev = pdev; |
518 |
+- break; |
519 |
+- } |
520 |
+- } |
521 |
+- } |
522 |
+-#endif /* CONFIG_PCI_IOV */ |
523 |
+ |
524 |
+ if (phb && phb->dma_dev_setup) |
525 |
+ phb->dma_dev_setup(phb, pdev); |
526 |
+diff --git a/arch/s390/Makefile b/arch/s390/Makefile |
527 |
+index e6c2e8925fef..4bccde36cb16 100644 |
528 |
+--- a/arch/s390/Makefile |
529 |
++++ b/arch/s390/Makefile |
530 |
+@@ -63,7 +63,7 @@ cflags-y += -Wa,-I$(srctree)/arch/$(ARCH)/include |
531 |
+ # |
532 |
+ cflags-$(CONFIG_FRAME_POINTER) += -fno-optimize-sibling-calls |
533 |
+ |
534 |
+-ifeq ($(call cc-option-yn,-mpacked-stack),y) |
535 |
++ifeq ($(call cc-option-yn,-mpacked-stack -mbackchain -msoft-float),y) |
536 |
+ cflags-$(CONFIG_PACK_STACK) += -mpacked-stack -D__PACK_STACK |
537 |
+ aflags-$(CONFIG_PACK_STACK) += -D__PACK_STACK |
538 |
+ endif |
539 |
+diff --git a/arch/s390/kernel/mcount.S b/arch/s390/kernel/mcount.S |
540 |
+index e93fbf02490c..83afd5b78e16 100644 |
541 |
+--- a/arch/s390/kernel/mcount.S |
542 |
++++ b/arch/s390/kernel/mcount.S |
543 |
+@@ -25,6 +25,12 @@ ENTRY(ftrace_stub) |
544 |
+ #define STACK_PTREGS (STACK_FRAME_OVERHEAD) |
545 |
+ #define STACK_PTREGS_GPRS (STACK_PTREGS + __PT_GPRS) |
546 |
+ #define STACK_PTREGS_PSW (STACK_PTREGS + __PT_PSW) |
547 |
++#ifdef __PACK_STACK |
548 |
++/* allocate just enough for r14, r15 and backchain */ |
549 |
++#define TRACED_FUNC_FRAME_SIZE 24 |
550 |
++#else |
551 |
++#define TRACED_FUNC_FRAME_SIZE STACK_FRAME_OVERHEAD |
552 |
++#endif |
553 |
+ |
554 |
+ ENTRY(_mcount) |
555 |
+ BR_EX %r14 |
556 |
+@@ -38,9 +44,16 @@ ENTRY(ftrace_caller) |
557 |
+ #if !(defined(CC_USING_HOTPATCH) || defined(CC_USING_NOP_MCOUNT)) |
558 |
+ aghi %r0,MCOUNT_RETURN_FIXUP |
559 |
+ #endif |
560 |
+- aghi %r15,-STACK_FRAME_SIZE |
561 |
++ # allocate stack frame for ftrace_caller to contain traced function |
562 |
++ aghi %r15,-TRACED_FUNC_FRAME_SIZE |
563 |
+ stg %r1,__SF_BACKCHAIN(%r15) |
564 |
++ stg %r0,(__SF_GPRS+8*8)(%r15) |
565 |
++ stg %r15,(__SF_GPRS+9*8)(%r15) |
566 |
++ # allocate pt_regs and stack frame for ftrace_trace_function |
567 |
++ aghi %r15,-STACK_FRAME_SIZE |
568 |
+ stg %r1,(STACK_PTREGS_GPRS+15*8)(%r15) |
569 |
++ aghi %r1,-TRACED_FUNC_FRAME_SIZE |
570 |
++ stg %r1,__SF_BACKCHAIN(%r15) |
571 |
+ stg %r0,(STACK_PTREGS_PSW+8)(%r15) |
572 |
+ stmg %r2,%r14,(STACK_PTREGS_GPRS+2*8)(%r15) |
573 |
+ #ifdef CONFIG_HAVE_MARCH_Z196_FEATURES |
574 |
+diff --git a/arch/s390/kvm/interrupt.c b/arch/s390/kvm/interrupt.c |
575 |
+index 05ea466b9e40..3515f2b55eb9 100644 |
576 |
+--- a/arch/s390/kvm/interrupt.c |
577 |
++++ b/arch/s390/kvm/interrupt.c |
578 |
+@@ -2109,7 +2109,7 @@ static int flic_ais_mode_get_all(struct kvm *kvm, struct kvm_device_attr *attr) |
579 |
+ return -EINVAL; |
580 |
+ |
581 |
+ if (!test_kvm_facility(kvm, 72)) |
582 |
+- return -ENOTSUPP; |
583 |
++ return -EOPNOTSUPP; |
584 |
+ |
585 |
+ mutex_lock(&fi->ais_lock); |
586 |
+ ais.simm = fi->simm; |
587 |
+@@ -2412,7 +2412,7 @@ static int modify_ais_mode(struct kvm *kvm, struct kvm_device_attr *attr) |
588 |
+ int ret = 0; |
589 |
+ |
590 |
+ if (!test_kvm_facility(kvm, 72)) |
591 |
+- return -ENOTSUPP; |
592 |
++ return -EOPNOTSUPP; |
593 |
+ |
594 |
+ if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req))) |
595 |
+ return -EFAULT; |
596 |
+@@ -2492,7 +2492,7 @@ static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr) |
597 |
+ struct kvm_s390_ais_all ais; |
598 |
+ |
599 |
+ if (!test_kvm_facility(kvm, 72)) |
600 |
+- return -ENOTSUPP; |
601 |
++ return -EOPNOTSUPP; |
602 |
+ |
603 |
+ if (copy_from_user(&ais, (void __user *)attr->addr, sizeof(ais))) |
604 |
+ return -EFAULT; |
605 |
+diff --git a/arch/s390/pci/pci_sysfs.c b/arch/s390/pci/pci_sysfs.c |
606 |
+index 430c14b006d1..0e11fc023fe7 100644 |
607 |
+--- a/arch/s390/pci/pci_sysfs.c |
608 |
++++ b/arch/s390/pci/pci_sysfs.c |
609 |
+@@ -13,6 +13,8 @@ |
610 |
+ #include <linux/stat.h> |
611 |
+ #include <linux/pci.h> |
612 |
+ |
613 |
++#include "../../../drivers/pci/pci.h" |
614 |
++ |
615 |
+ #include <asm/sclp.h> |
616 |
+ |
617 |
+ #define zpci_attr(name, fmt, member) \ |
618 |
+@@ -40,31 +42,50 @@ zpci_attr(segment3, "0x%02x\n", pfip[3]); |
619 |
+ static ssize_t recover_store(struct device *dev, struct device_attribute *attr, |
620 |
+ const char *buf, size_t count) |
621 |
+ { |
622 |
++ struct kernfs_node *kn; |
623 |
+ struct pci_dev *pdev = to_pci_dev(dev); |
624 |
+ struct zpci_dev *zdev = to_zpci(pdev); |
625 |
+- int ret; |
626 |
+- |
627 |
+- if (!device_remove_file_self(dev, attr)) |
628 |
+- return count; |
629 |
+- |
630 |
++ int ret = 0; |
631 |
++ |
632 |
++ /* Can't use device_remove_self() here as that would lead us to lock |
633 |
++ * the pci_rescan_remove_lock while holding the device' kernfs lock. |
634 |
++ * This would create a possible deadlock with disable_slot() which is |
635 |
++ * not directly protected by the device' kernfs lock but takes it |
636 |
++ * during the device removal which happens under |
637 |
++ * pci_rescan_remove_lock. |
638 |
++ * |
639 |
++ * This is analogous to sdev_store_delete() in |
640 |
++ * drivers/scsi/scsi_sysfs.c |
641 |
++ */ |
642 |
++ kn = sysfs_break_active_protection(&dev->kobj, &attr->attr); |
643 |
++ WARN_ON_ONCE(!kn); |
644 |
++ /* device_remove_file() serializes concurrent calls ignoring all but |
645 |
++ * the first |
646 |
++ */ |
647 |
++ device_remove_file(dev, attr); |
648 |
++ |
649 |
++ /* A concurrent call to recover_store() may slip between |
650 |
++ * sysfs_break_active_protection() and the sysfs file removal. |
651 |
++ * Once it unblocks from pci_lock_rescan_remove() the original pdev |
652 |
++ * will already be removed. |
653 |
++ */ |
654 |
+ pci_lock_rescan_remove(); |
655 |
+- pci_stop_and_remove_bus_device(pdev); |
656 |
+- ret = zpci_disable_device(zdev); |
657 |
+- if (ret) |
658 |
+- goto error; |
659 |
+- |
660 |
+- ret = zpci_enable_device(zdev); |
661 |
+- if (ret) |
662 |
+- goto error; |
663 |
+- |
664 |
+- pci_rescan_bus(zdev->bus); |
665 |
++ if (pci_dev_is_added(pdev)) { |
666 |
++ pci_stop_and_remove_bus_device(pdev); |
667 |
++ ret = zpci_disable_device(zdev); |
668 |
++ if (ret) |
669 |
++ goto out; |
670 |
++ |
671 |
++ ret = zpci_enable_device(zdev); |
672 |
++ if (ret) |
673 |
++ goto out; |
674 |
++ pci_rescan_bus(zdev->bus); |
675 |
++ } |
676 |
++out: |
677 |
+ pci_unlock_rescan_remove(); |
678 |
+- |
679 |
+- return count; |
680 |
+- |
681 |
+-error: |
682 |
+- pci_unlock_rescan_remove(); |
683 |
+- return ret; |
684 |
++ if (kn) |
685 |
++ sysfs_unbreak_active_protection(kn); |
686 |
++ return ret ? ret : count; |
687 |
+ } |
688 |
+ static DEVICE_ATTR_WO(recover); |
689 |
+ |
690 |
+diff --git a/arch/sh/include/cpu-sh2a/cpu/sh7269.h b/arch/sh/include/cpu-sh2a/cpu/sh7269.h |
691 |
+index d516e5d48818..b887cc402b71 100644 |
692 |
+--- a/arch/sh/include/cpu-sh2a/cpu/sh7269.h |
693 |
++++ b/arch/sh/include/cpu-sh2a/cpu/sh7269.h |
694 |
+@@ -78,8 +78,15 @@ enum { |
695 |
+ GPIO_FN_WDTOVF, |
696 |
+ |
697 |
+ /* CAN */ |
698 |
+- GPIO_FN_CTX1, GPIO_FN_CRX1, GPIO_FN_CTX0, GPIO_FN_CTX0_CTX1, |
699 |
+- GPIO_FN_CRX0, GPIO_FN_CRX0_CRX1, GPIO_FN_CRX0_CRX1_CRX2, |
700 |
++ GPIO_FN_CTX2, GPIO_FN_CRX2, |
701 |
++ GPIO_FN_CTX1, GPIO_FN_CRX1, |
702 |
++ GPIO_FN_CTX0, GPIO_FN_CRX0, |
703 |
++ GPIO_FN_CTX0_CTX1, GPIO_FN_CRX0_CRX1, |
704 |
++ GPIO_FN_CTX0_CTX1_CTX2, GPIO_FN_CRX0_CRX1_CRX2, |
705 |
++ GPIO_FN_CTX2_PJ21, GPIO_FN_CRX2_PJ20, |
706 |
++ GPIO_FN_CTX1_PJ23, GPIO_FN_CRX1_PJ22, |
707 |
++ GPIO_FN_CTX0_CTX1_PJ23, GPIO_FN_CRX0_CRX1_PJ22, |
708 |
++ GPIO_FN_CTX0_CTX1_CTX2_PJ21, GPIO_FN_CRX0_CRX1_CRX2_PJ20, |
709 |
+ |
710 |
+ /* DMAC */ |
711 |
+ GPIO_FN_TEND0, GPIO_FN_DACK0, GPIO_FN_DREQ0, |
712 |
+diff --git a/arch/sparc/kernel/vmlinux.lds.S b/arch/sparc/kernel/vmlinux.lds.S |
713 |
+index 61afd787bd0c..59b6df13ddea 100644 |
714 |
+--- a/arch/sparc/kernel/vmlinux.lds.S |
715 |
++++ b/arch/sparc/kernel/vmlinux.lds.S |
716 |
+@@ -172,12 +172,14 @@ SECTIONS |
717 |
+ } |
718 |
+ PERCPU_SECTION(SMP_CACHE_BYTES) |
719 |
+ |
720 |
+-#ifdef CONFIG_JUMP_LABEL |
721 |
+ . = ALIGN(PAGE_SIZE); |
722 |
+ .exit.text : { |
723 |
+ EXIT_TEXT |
724 |
+ } |
725 |
+-#endif |
726 |
++ |
727 |
++ .exit.data : { |
728 |
++ EXIT_DATA |
729 |
++ } |
730 |
+ |
731 |
+ . = ALIGN(PAGE_SIZE); |
732 |
+ __init_end = .; |
733 |
+diff --git a/arch/x86/entry/vdso/vdso32-setup.c b/arch/x86/entry/vdso/vdso32-setup.c |
734 |
+index 42d4c89f990e..ddff0ca6f509 100644 |
735 |
+--- a/arch/x86/entry/vdso/vdso32-setup.c |
736 |
++++ b/arch/x86/entry/vdso/vdso32-setup.c |
737 |
+@@ -11,6 +11,7 @@ |
738 |
+ #include <linux/smp.h> |
739 |
+ #include <linux/kernel.h> |
740 |
+ #include <linux/mm_types.h> |
741 |
++#include <linux/elf.h> |
742 |
+ |
743 |
+ #include <asm/processor.h> |
744 |
+ #include <asm/vdso.h> |
745 |
+diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h |
746 |
+index 75ded1d13d98..9d5d949e662e 100644 |
747 |
+--- a/arch/x86/include/asm/nmi.h |
748 |
++++ b/arch/x86/include/asm/nmi.h |
749 |
+@@ -41,7 +41,6 @@ struct nmiaction { |
750 |
+ struct list_head list; |
751 |
+ nmi_handler_t handler; |
752 |
+ u64 max_duration; |
753 |
+- struct irq_work irq_work; |
754 |
+ unsigned long flags; |
755 |
+ const char *name; |
756 |
+ }; |
757 |
+diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c |
758 |
+index 086cf1d1d71d..0f8b9b900b0e 100644 |
759 |
+--- a/arch/x86/kernel/nmi.c |
760 |
++++ b/arch/x86/kernel/nmi.c |
761 |
+@@ -102,18 +102,22 @@ static int __init nmi_warning_debugfs(void) |
762 |
+ } |
763 |
+ fs_initcall(nmi_warning_debugfs); |
764 |
+ |
765 |
+-static void nmi_max_handler(struct irq_work *w) |
766 |
++static void nmi_check_duration(struct nmiaction *action, u64 duration) |
767 |
+ { |
768 |
+- struct nmiaction *a = container_of(w, struct nmiaction, irq_work); |
769 |
++ u64 whole_msecs = READ_ONCE(action->max_duration); |
770 |
+ int remainder_ns, decimal_msecs; |
771 |
+- u64 whole_msecs = READ_ONCE(a->max_duration); |
772 |
++ |
773 |
++ if (duration < nmi_longest_ns || duration < action->max_duration) |
774 |
++ return; |
775 |
++ |
776 |
++ action->max_duration = duration; |
777 |
+ |
778 |
+ remainder_ns = do_div(whole_msecs, (1000 * 1000)); |
779 |
+ decimal_msecs = remainder_ns / 1000; |
780 |
+ |
781 |
+ printk_ratelimited(KERN_INFO |
782 |
+ "INFO: NMI handler (%ps) took too long to run: %lld.%03d msecs\n", |
783 |
+- a->handler, whole_msecs, decimal_msecs); |
784 |
++ action->handler, whole_msecs, decimal_msecs); |
785 |
+ } |
786 |
+ |
787 |
+ static int nmi_handle(unsigned int type, struct pt_regs *regs) |
788 |
+@@ -140,11 +144,7 @@ static int nmi_handle(unsigned int type, struct pt_regs *regs) |
789 |
+ delta = sched_clock() - delta; |
790 |
+ trace_nmi_handler(a->handler, (int)delta, thishandled); |
791 |
+ |
792 |
+- if (delta < nmi_longest_ns || delta < a->max_duration) |
793 |
+- continue; |
794 |
+- |
795 |
+- a->max_duration = delta; |
796 |
+- irq_work_queue(&a->irq_work); |
797 |
++ nmi_check_duration(a, delta); |
798 |
+ } |
799 |
+ |
800 |
+ rcu_read_unlock(); |
801 |
+@@ -162,8 +162,6 @@ int __register_nmi_handler(unsigned int type, struct nmiaction *action) |
802 |
+ if (!action->handler) |
803 |
+ return -EINVAL; |
804 |
+ |
805 |
+- init_irq_work(&action->irq_work, nmi_max_handler); |
806 |
+- |
807 |
+ raw_spin_lock_irqsave(&desc->lock, flags); |
808 |
+ |
809 |
+ /* |
810 |
+diff --git a/arch/x86/kernel/sysfb_simplefb.c b/arch/x86/kernel/sysfb_simplefb.c |
811 |
+index 85195d447a92..f3215346e47f 100644 |
812 |
+--- a/arch/x86/kernel/sysfb_simplefb.c |
813 |
++++ b/arch/x86/kernel/sysfb_simplefb.c |
814 |
+@@ -94,11 +94,11 @@ __init int create_simplefb(const struct screen_info *si, |
815 |
+ if (si->orig_video_isVGA == VIDEO_TYPE_VLFB) |
816 |
+ size <<= 16; |
817 |
+ length = mode->height * mode->stride; |
818 |
+- length = PAGE_ALIGN(length); |
819 |
+ if (length > size) { |
820 |
+ printk(KERN_WARNING "sysfb: VRAM smaller than advertised\n"); |
821 |
+ return -EINVAL; |
822 |
+ } |
823 |
++ length = PAGE_ALIGN(length); |
824 |
+ |
825 |
+ /* setup IORESOURCE_MEM as framebuffer memory */ |
826 |
+ memset(&res, 0, sizeof(res)); |
827 |
+diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c |
828 |
+index 2660c01eadae..aead984d89ad 100644 |
829 |
+--- a/arch/x86/kvm/vmx.c |
830 |
++++ b/arch/x86/kvm/vmx.c |
831 |
+@@ -5302,6 +5302,9 @@ static void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
832 |
+ |
833 |
+ static int get_ept_level(struct kvm_vcpu *vcpu) |
834 |
+ { |
835 |
++ /* Nested EPT currently only supports 4-level walks. */ |
836 |
++ if (is_guest_mode(vcpu) && nested_cpu_has_ept(get_vmcs12(vcpu))) |
837 |
++ return 4; |
838 |
+ if (cpu_has_vmx_ept_5levels() && (cpuid_maxphyaddr(vcpu) > 48)) |
839 |
+ return 5; |
840 |
+ return 4; |
841 |
+diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c |
842 |
+deleted file mode 100644 |
843 |
+index 997926a9121c..000000000000 |
844 |
+--- a/arch/x86/kvm/vmx/vmx.c |
845 |
++++ /dev/null |
846 |
+@@ -1,8036 +0,0 @@ |
847 |
+-// SPDX-License-Identifier: GPL-2.0-only |
848 |
+-/* |
849 |
+- * Kernel-based Virtual Machine driver for Linux |
850 |
+- * |
851 |
+- * This module enables machines with Intel VT-x extensions to run virtual |
852 |
+- * machines without emulation or binary translation. |
853 |
+- * |
854 |
+- * Copyright (C) 2006 Qumranet, Inc. |
855 |
+- * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
856 |
+- * |
857 |
+- * Authors: |
858 |
+- * Avi Kivity <avi@××××××××.com> |
859 |
+- * Yaniv Kamay <yaniv@××××××××.com> |
860 |
+- */ |
861 |
+- |
862 |
+-#include <linux/frame.h> |
863 |
+-#include <linux/highmem.h> |
864 |
+-#include <linux/hrtimer.h> |
865 |
+-#include <linux/kernel.h> |
866 |
+-#include <linux/kvm_host.h> |
867 |
+-#include <linux/module.h> |
868 |
+-#include <linux/moduleparam.h> |
869 |
+-#include <linux/mod_devicetable.h> |
870 |
+-#include <linux/mm.h> |
871 |
+-#include <linux/sched.h> |
872 |
+-#include <linux/sched/smt.h> |
873 |
+-#include <linux/slab.h> |
874 |
+-#include <linux/tboot.h> |
875 |
+-#include <linux/trace_events.h> |
876 |
+- |
877 |
+-#include <asm/apic.h> |
878 |
+-#include <asm/asm.h> |
879 |
+-#include <asm/cpu.h> |
880 |
+-#include <asm/debugreg.h> |
881 |
+-#include <asm/desc.h> |
882 |
+-#include <asm/fpu/internal.h> |
883 |
+-#include <asm/io.h> |
884 |
+-#include <asm/irq_remapping.h> |
885 |
+-#include <asm/kexec.h> |
886 |
+-#include <asm/perf_event.h> |
887 |
+-#include <asm/mce.h> |
888 |
+-#include <asm/mmu_context.h> |
889 |
+-#include <asm/mshyperv.h> |
890 |
+-#include <asm/spec-ctrl.h> |
891 |
+-#include <asm/virtext.h> |
892 |
+-#include <asm/vmx.h> |
893 |
+- |
894 |
+-#include "capabilities.h" |
895 |
+-#include "cpuid.h" |
896 |
+-#include "evmcs.h" |
897 |
+-#include "irq.h" |
898 |
+-#include "kvm_cache_regs.h" |
899 |
+-#include "lapic.h" |
900 |
+-#include "mmu.h" |
901 |
+-#include "nested.h" |
902 |
+-#include "ops.h" |
903 |
+-#include "pmu.h" |
904 |
+-#include "trace.h" |
905 |
+-#include "vmcs.h" |
906 |
+-#include "vmcs12.h" |
907 |
+-#include "vmx.h" |
908 |
+-#include "x86.h" |
909 |
+- |
910 |
+-MODULE_AUTHOR("Qumranet"); |
911 |
+-MODULE_LICENSE("GPL"); |
912 |
+- |
913 |
+-static const struct x86_cpu_id vmx_cpu_id[] = { |
914 |
+- X86_FEATURE_MATCH(X86_FEATURE_VMX), |
915 |
+- {} |
916 |
+-}; |
917 |
+-MODULE_DEVICE_TABLE(x86cpu, vmx_cpu_id); |
918 |
+- |
919 |
+-bool __read_mostly enable_vpid = 1; |
920 |
+-module_param_named(vpid, enable_vpid, bool, 0444); |
921 |
+- |
922 |
+-static bool __read_mostly enable_vnmi = 1; |
923 |
+-module_param_named(vnmi, enable_vnmi, bool, S_IRUGO); |
924 |
+- |
925 |
+-bool __read_mostly flexpriority_enabled = 1; |
926 |
+-module_param_named(flexpriority, flexpriority_enabled, bool, S_IRUGO); |
927 |
+- |
928 |
+-bool __read_mostly enable_ept = 1; |
929 |
+-module_param_named(ept, enable_ept, bool, S_IRUGO); |
930 |
+- |
931 |
+-bool __read_mostly enable_unrestricted_guest = 1; |
932 |
+-module_param_named(unrestricted_guest, |
933 |
+- enable_unrestricted_guest, bool, S_IRUGO); |
934 |
+- |
935 |
+-bool __read_mostly enable_ept_ad_bits = 1; |
936 |
+-module_param_named(eptad, enable_ept_ad_bits, bool, S_IRUGO); |
937 |
+- |
938 |
+-static bool __read_mostly emulate_invalid_guest_state = true; |
939 |
+-module_param(emulate_invalid_guest_state, bool, S_IRUGO); |
940 |
+- |
941 |
+-static bool __read_mostly fasteoi = 1; |
942 |
+-module_param(fasteoi, bool, S_IRUGO); |
943 |
+- |
944 |
+-static bool __read_mostly enable_apicv = 1; |
945 |
+-module_param(enable_apicv, bool, S_IRUGO); |
946 |
+- |
947 |
+-/* |
948 |
+- * If nested=1, nested virtualization is supported, i.e., guests may use |
949 |
+- * VMX and be a hypervisor for its own guests. If nested=0, guests may not |
950 |
+- * use VMX instructions. |
951 |
+- */ |
952 |
+-static bool __read_mostly nested = 1; |
953 |
+-module_param(nested, bool, S_IRUGO); |
954 |
+- |
955 |
+-bool __read_mostly enable_pml = 1; |
956 |
+-module_param_named(pml, enable_pml, bool, S_IRUGO); |
957 |
+- |
958 |
+-static bool __read_mostly dump_invalid_vmcs = 0; |
959 |
+-module_param(dump_invalid_vmcs, bool, 0644); |
960 |
+- |
961 |
+-#define MSR_BITMAP_MODE_X2APIC 1 |
962 |
+-#define MSR_BITMAP_MODE_X2APIC_APICV 2 |
963 |
+- |
964 |
+-#define KVM_VMX_TSC_MULTIPLIER_MAX 0xffffffffffffffffULL |
965 |
+- |
966 |
+-/* Guest_tsc -> host_tsc conversion requires 64-bit division. */ |
967 |
+-static int __read_mostly cpu_preemption_timer_multi; |
968 |
+-static bool __read_mostly enable_preemption_timer = 1; |
969 |
+-#ifdef CONFIG_X86_64 |
970 |
+-module_param_named(preemption_timer, enable_preemption_timer, bool, S_IRUGO); |
971 |
+-#endif |
972 |
+- |
973 |
+-#define KVM_VM_CR0_ALWAYS_OFF (X86_CR0_NW | X86_CR0_CD) |
974 |
+-#define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST X86_CR0_NE |
975 |
+-#define KVM_VM_CR0_ALWAYS_ON \ |
976 |
+- (KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | \ |
977 |
+- X86_CR0_WP | X86_CR0_PG | X86_CR0_PE) |
978 |
+-#define KVM_CR4_GUEST_OWNED_BITS \ |
979 |
+- (X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR \ |
980 |
+- | X86_CR4_OSXMMEXCPT | X86_CR4_LA57 | X86_CR4_TSD) |
981 |
+- |
982 |
+-#define KVM_VM_CR4_ALWAYS_ON_UNRESTRICTED_GUEST X86_CR4_VMXE |
983 |
+-#define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE) |
984 |
+-#define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE) |
985 |
+- |
986 |
+-#define RMODE_GUEST_OWNED_EFLAGS_BITS (~(X86_EFLAGS_IOPL | X86_EFLAGS_VM)) |
987 |
+- |
988 |
+-#define MSR_IA32_RTIT_STATUS_MASK (~(RTIT_STATUS_FILTEREN | \ |
989 |
+- RTIT_STATUS_CONTEXTEN | RTIT_STATUS_TRIGGEREN | \ |
990 |
+- RTIT_STATUS_ERROR | RTIT_STATUS_STOPPED | \ |
991 |
+- RTIT_STATUS_BYTECNT)) |
992 |
+- |
993 |
+-#define MSR_IA32_RTIT_OUTPUT_BASE_MASK \ |
994 |
+- (~((1UL << cpuid_query_maxphyaddr(vcpu)) - 1) | 0x7f) |
995 |
+- |
996 |
+-/* |
997 |
+- * These 2 parameters are used to config the controls for Pause-Loop Exiting: |
998 |
+- * ple_gap: upper bound on the amount of time between two successive |
999 |
+- * executions of PAUSE in a loop. Also indicate if ple enabled. |
1000 |
+- * According to test, this time is usually smaller than 128 cycles. |
1001 |
+- * ple_window: upper bound on the amount of time a guest is allowed to execute |
1002 |
+- * in a PAUSE loop. Tests indicate that most spinlocks are held for |
1003 |
+- * less than 2^12 cycles |
1004 |
+- * Time is measured based on a counter that runs at the same rate as the TSC, |
1005 |
+- * refer SDM volume 3b section 21.6.13 & 22.1.3. |
1006 |
+- */ |
1007 |
+-static unsigned int ple_gap = KVM_DEFAULT_PLE_GAP; |
1008 |
+-module_param(ple_gap, uint, 0444); |
1009 |
+- |
1010 |
+-static unsigned int ple_window = KVM_VMX_DEFAULT_PLE_WINDOW; |
1011 |
+-module_param(ple_window, uint, 0444); |
1012 |
+- |
1013 |
+-/* Default doubles per-vcpu window every exit. */ |
1014 |
+-static unsigned int ple_window_grow = KVM_DEFAULT_PLE_WINDOW_GROW; |
1015 |
+-module_param(ple_window_grow, uint, 0444); |
1016 |
+- |
1017 |
+-/* Default resets per-vcpu window every exit to ple_window. */ |
1018 |
+-static unsigned int ple_window_shrink = KVM_DEFAULT_PLE_WINDOW_SHRINK; |
1019 |
+-module_param(ple_window_shrink, uint, 0444); |
1020 |
+- |
1021 |
+-/* Default is to compute the maximum so we can never overflow. */ |
1022 |
+-static unsigned int ple_window_max = KVM_VMX_DEFAULT_PLE_WINDOW_MAX; |
1023 |
+-module_param(ple_window_max, uint, 0444); |
1024 |
+- |
1025 |
+-/* Default is SYSTEM mode, 1 for host-guest mode */ |
1026 |
+-int __read_mostly pt_mode = PT_MODE_SYSTEM; |
1027 |
+-module_param(pt_mode, int, S_IRUGO); |
1028 |
+- |
1029 |
+-static DEFINE_STATIC_KEY_FALSE(vmx_l1d_should_flush); |
1030 |
+-static DEFINE_STATIC_KEY_FALSE(vmx_l1d_flush_cond); |
1031 |
+-static DEFINE_MUTEX(vmx_l1d_flush_mutex); |
1032 |
+- |
1033 |
+-/* Storage for pre module init parameter parsing */ |
1034 |
+-static enum vmx_l1d_flush_state __read_mostly vmentry_l1d_flush_param = VMENTER_L1D_FLUSH_AUTO; |
1035 |
+- |
1036 |
+-static const struct { |
1037 |
+- const char *option; |
1038 |
+- bool for_parse; |
1039 |
+-} vmentry_l1d_param[] = { |
1040 |
+- [VMENTER_L1D_FLUSH_AUTO] = {"auto", true}, |
1041 |
+- [VMENTER_L1D_FLUSH_NEVER] = {"never", true}, |
1042 |
+- [VMENTER_L1D_FLUSH_COND] = {"cond", true}, |
1043 |
+- [VMENTER_L1D_FLUSH_ALWAYS] = {"always", true}, |
1044 |
+- [VMENTER_L1D_FLUSH_EPT_DISABLED] = {"EPT disabled", false}, |
1045 |
+- [VMENTER_L1D_FLUSH_NOT_REQUIRED] = {"not required", false}, |
1046 |
+-}; |
1047 |
+- |
1048 |
+-#define L1D_CACHE_ORDER 4 |
1049 |
+-static void *vmx_l1d_flush_pages; |
1050 |
+- |
1051 |
+-static int vmx_setup_l1d_flush(enum vmx_l1d_flush_state l1tf) |
1052 |
+-{ |
1053 |
+- struct page *page; |
1054 |
+- unsigned int i; |
1055 |
+- |
1056 |
+- if (!boot_cpu_has_bug(X86_BUG_L1TF)) { |
1057 |
+- l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED; |
1058 |
+- return 0; |
1059 |
+- } |
1060 |
+- |
1061 |
+- if (!enable_ept) { |
1062 |
+- l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_EPT_DISABLED; |
1063 |
+- return 0; |
1064 |
+- } |
1065 |
+- |
1066 |
+- if (boot_cpu_has(X86_FEATURE_ARCH_CAPABILITIES)) { |
1067 |
+- u64 msr; |
1068 |
+- |
1069 |
+- rdmsrl(MSR_IA32_ARCH_CAPABILITIES, msr); |
1070 |
+- if (msr & ARCH_CAP_SKIP_VMENTRY_L1DFLUSH) { |
1071 |
+- l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED; |
1072 |
+- return 0; |
1073 |
+- } |
1074 |
+- } |
1075 |
+- |
1076 |
+- /* If set to auto use the default l1tf mitigation method */ |
1077 |
+- if (l1tf == VMENTER_L1D_FLUSH_AUTO) { |
1078 |
+- switch (l1tf_mitigation) { |
1079 |
+- case L1TF_MITIGATION_OFF: |
1080 |
+- l1tf = VMENTER_L1D_FLUSH_NEVER; |
1081 |
+- break; |
1082 |
+- case L1TF_MITIGATION_FLUSH_NOWARN: |
1083 |
+- case L1TF_MITIGATION_FLUSH: |
1084 |
+- case L1TF_MITIGATION_FLUSH_NOSMT: |
1085 |
+- l1tf = VMENTER_L1D_FLUSH_COND; |
1086 |
+- break; |
1087 |
+- case L1TF_MITIGATION_FULL: |
1088 |
+- case L1TF_MITIGATION_FULL_FORCE: |
1089 |
+- l1tf = VMENTER_L1D_FLUSH_ALWAYS; |
1090 |
+- break; |
1091 |
+- } |
1092 |
+- } else if (l1tf_mitigation == L1TF_MITIGATION_FULL_FORCE) { |
1093 |
+- l1tf = VMENTER_L1D_FLUSH_ALWAYS; |
1094 |
+- } |
1095 |
+- |
1096 |
+- if (l1tf != VMENTER_L1D_FLUSH_NEVER && !vmx_l1d_flush_pages && |
1097 |
+- !boot_cpu_has(X86_FEATURE_FLUSH_L1D)) { |
1098 |
+- /* |
1099 |
+- * This allocation for vmx_l1d_flush_pages is not tied to a VM |
1100 |
+- * lifetime and so should not be charged to a memcg. |
1101 |
+- */ |
1102 |
+- page = alloc_pages(GFP_KERNEL, L1D_CACHE_ORDER); |
1103 |
+- if (!page) |
1104 |
+- return -ENOMEM; |
1105 |
+- vmx_l1d_flush_pages = page_address(page); |
1106 |
+- |
1107 |
+- /* |
1108 |
+- * Initialize each page with a different pattern in |
1109 |
+- * order to protect against KSM in the nested |
1110 |
+- * virtualization case. |
1111 |
+- */ |
1112 |
+- for (i = 0; i < 1u << L1D_CACHE_ORDER; ++i) { |
1113 |
+- memset(vmx_l1d_flush_pages + i * PAGE_SIZE, i + 1, |
1114 |
+- PAGE_SIZE); |
1115 |
+- } |
1116 |
+- } |
1117 |
+- |
1118 |
+- l1tf_vmx_mitigation = l1tf; |
1119 |
+- |
1120 |
+- if (l1tf != VMENTER_L1D_FLUSH_NEVER) |
1121 |
+- static_branch_enable(&vmx_l1d_should_flush); |
1122 |
+- else |
1123 |
+- static_branch_disable(&vmx_l1d_should_flush); |
1124 |
+- |
1125 |
+- if (l1tf == VMENTER_L1D_FLUSH_COND) |
1126 |
+- static_branch_enable(&vmx_l1d_flush_cond); |
1127 |
+- else |
1128 |
+- static_branch_disable(&vmx_l1d_flush_cond); |
1129 |
+- return 0; |
1130 |
+-} |
1131 |
+- |
1132 |
+-static int vmentry_l1d_flush_parse(const char *s) |
1133 |
+-{ |
1134 |
+- unsigned int i; |
1135 |
+- |
1136 |
+- if (s) { |
1137 |
+- for (i = 0; i < ARRAY_SIZE(vmentry_l1d_param); i++) { |
1138 |
+- if (vmentry_l1d_param[i].for_parse && |
1139 |
+- sysfs_streq(s, vmentry_l1d_param[i].option)) |
1140 |
+- return i; |
1141 |
+- } |
1142 |
+- } |
1143 |
+- return -EINVAL; |
1144 |
+-} |
1145 |
+- |
1146 |
+-static int vmentry_l1d_flush_set(const char *s, const struct kernel_param *kp) |
1147 |
+-{ |
1148 |
+- int l1tf, ret; |
1149 |
+- |
1150 |
+- l1tf = vmentry_l1d_flush_parse(s); |
1151 |
+- if (l1tf < 0) |
1152 |
+- return l1tf; |
1153 |
+- |
1154 |
+- if (!boot_cpu_has(X86_BUG_L1TF)) |
1155 |
+- return 0; |
1156 |
+- |
1157 |
+- /* |
1158 |
+- * Has vmx_init() run already? If not then this is the pre init |
1159 |
+- * parameter parsing. In that case just store the value and let |
1160 |
+- * vmx_init() do the proper setup after enable_ept has been |
1161 |
+- * established. |
1162 |
+- */ |
1163 |
+- if (l1tf_vmx_mitigation == VMENTER_L1D_FLUSH_AUTO) { |
1164 |
+- vmentry_l1d_flush_param = l1tf; |
1165 |
+- return 0; |
1166 |
+- } |
1167 |
+- |
1168 |
+- mutex_lock(&vmx_l1d_flush_mutex); |
1169 |
+- ret = vmx_setup_l1d_flush(l1tf); |
1170 |
+- mutex_unlock(&vmx_l1d_flush_mutex); |
1171 |
+- return ret; |
1172 |
+-} |
1173 |
+- |
1174 |
+-static int vmentry_l1d_flush_get(char *s, const struct kernel_param *kp) |
1175 |
+-{ |
1176 |
+- if (WARN_ON_ONCE(l1tf_vmx_mitigation >= ARRAY_SIZE(vmentry_l1d_param))) |
1177 |
+- return sprintf(s, "???\n"); |
1178 |
+- |
1179 |
+- return sprintf(s, "%s\n", vmentry_l1d_param[l1tf_vmx_mitigation].option); |
1180 |
+-} |
1181 |
+- |
1182 |
+-static const struct kernel_param_ops vmentry_l1d_flush_ops = { |
1183 |
+- .set = vmentry_l1d_flush_set, |
1184 |
+- .get = vmentry_l1d_flush_get, |
1185 |
+-}; |
1186 |
+-module_param_cb(vmentry_l1d_flush, &vmentry_l1d_flush_ops, NULL, 0644); |
1187 |
+- |
1188 |
+-static bool guest_state_valid(struct kvm_vcpu *vcpu); |
1189 |
+-static u32 vmx_segment_access_rights(struct kvm_segment *var); |
1190 |
+-static __always_inline void vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, |
1191 |
+- u32 msr, int type); |
1192 |
+- |
1193 |
+-void vmx_vmexit(void); |
1194 |
+- |
1195 |
+-#define vmx_insn_failed(fmt...) \ |
1196 |
+-do { \ |
1197 |
+- WARN_ONCE(1, fmt); \ |
1198 |
+- pr_warn_ratelimited(fmt); \ |
1199 |
+-} while (0) |
1200 |
+- |
1201 |
+-asmlinkage void vmread_error(unsigned long field, bool fault) |
1202 |
+-{ |
1203 |
+- if (fault) |
1204 |
+- kvm_spurious_fault(); |
1205 |
+- else |
1206 |
+- vmx_insn_failed("kvm: vmread failed: field=%lx\n", field); |
1207 |
+-} |
1208 |
+- |
1209 |
+-noinline void vmwrite_error(unsigned long field, unsigned long value) |
1210 |
+-{ |
1211 |
+- vmx_insn_failed("kvm: vmwrite failed: field=%lx val=%lx err=%d\n", |
1212 |
+- field, value, vmcs_read32(VM_INSTRUCTION_ERROR)); |
1213 |
+-} |
1214 |
+- |
1215 |
+-noinline void vmclear_error(struct vmcs *vmcs, u64 phys_addr) |
1216 |
+-{ |
1217 |
+- vmx_insn_failed("kvm: vmclear failed: %p/%llx\n", vmcs, phys_addr); |
1218 |
+-} |
1219 |
+- |
1220 |
+-noinline void vmptrld_error(struct vmcs *vmcs, u64 phys_addr) |
1221 |
+-{ |
1222 |
+- vmx_insn_failed("kvm: vmptrld failed: %p/%llx\n", vmcs, phys_addr); |
1223 |
+-} |
1224 |
+- |
1225 |
+-noinline void invvpid_error(unsigned long ext, u16 vpid, gva_t gva) |
1226 |
+-{ |
1227 |
+- vmx_insn_failed("kvm: invvpid failed: ext=0x%lx vpid=%u gva=0x%lx\n", |
1228 |
+- ext, vpid, gva); |
1229 |
+-} |
1230 |
+- |
1231 |
+-noinline void invept_error(unsigned long ext, u64 eptp, gpa_t gpa) |
1232 |
+-{ |
1233 |
+- vmx_insn_failed("kvm: invept failed: ext=0x%lx eptp=%llx gpa=0x%llx\n", |
1234 |
+- ext, eptp, gpa); |
1235 |
+-} |
1236 |
+- |
1237 |
+-static DEFINE_PER_CPU(struct vmcs *, vmxarea); |
1238 |
+-DEFINE_PER_CPU(struct vmcs *, current_vmcs); |
1239 |
+-/* |
1240 |
+- * We maintain a per-CPU linked-list of VMCS loaded on that CPU. This is needed |
1241 |
+- * when a CPU is brought down, and we need to VMCLEAR all VMCSs loaded on it. |
1242 |
+- */ |
1243 |
+-static DEFINE_PER_CPU(struct list_head, loaded_vmcss_on_cpu); |
1244 |
+- |
1245 |
+-/* |
1246 |
+- * We maintian a per-CPU linked-list of vCPU, so in wakeup_handler() we |
1247 |
+- * can find which vCPU should be waken up. |
1248 |
+- */ |
1249 |
+-static DEFINE_PER_CPU(struct list_head, blocked_vcpu_on_cpu); |
1250 |
+-static DEFINE_PER_CPU(spinlock_t, blocked_vcpu_on_cpu_lock); |
1251 |
+- |
1252 |
+-static DECLARE_BITMAP(vmx_vpid_bitmap, VMX_NR_VPIDS); |
1253 |
+-static DEFINE_SPINLOCK(vmx_vpid_lock); |
1254 |
+- |
1255 |
+-struct vmcs_config vmcs_config; |
1256 |
+-struct vmx_capability vmx_capability; |
1257 |
+- |
1258 |
+-#define VMX_SEGMENT_FIELD(seg) \ |
1259 |
+- [VCPU_SREG_##seg] = { \ |
1260 |
+- .selector = GUEST_##seg##_SELECTOR, \ |
1261 |
+- .base = GUEST_##seg##_BASE, \ |
1262 |
+- .limit = GUEST_##seg##_LIMIT, \ |
1263 |
+- .ar_bytes = GUEST_##seg##_AR_BYTES, \ |
1264 |
+- } |
1265 |
+- |
1266 |
+-static const struct kvm_vmx_segment_field { |
1267 |
+- unsigned selector; |
1268 |
+- unsigned base; |
1269 |
+- unsigned limit; |
1270 |
+- unsigned ar_bytes; |
1271 |
+-} kvm_vmx_segment_fields[] = { |
1272 |
+- VMX_SEGMENT_FIELD(CS), |
1273 |
+- VMX_SEGMENT_FIELD(DS), |
1274 |
+- VMX_SEGMENT_FIELD(ES), |
1275 |
+- VMX_SEGMENT_FIELD(FS), |
1276 |
+- VMX_SEGMENT_FIELD(GS), |
1277 |
+- VMX_SEGMENT_FIELD(SS), |
1278 |
+- VMX_SEGMENT_FIELD(TR), |
1279 |
+- VMX_SEGMENT_FIELD(LDTR), |
1280 |
+-}; |
1281 |
+- |
1282 |
+-u64 host_efer; |
1283 |
+-static unsigned long host_idt_base; |
1284 |
+- |
1285 |
+-/* |
1286 |
+- * Though SYSCALL is only supported in 64-bit mode on Intel CPUs, kvm |
1287 |
+- * will emulate SYSCALL in legacy mode if the vendor string in guest |
1288 |
+- * CPUID.0:{EBX,ECX,EDX} is "AuthenticAMD" or "AMDisbetter!" To |
1289 |
+- * support this emulation, IA32_STAR must always be included in |
1290 |
+- * vmx_msr_index[], even in i386 builds. |
1291 |
+- */ |
1292 |
+-const u32 vmx_msr_index[] = { |
1293 |
+-#ifdef CONFIG_X86_64 |
1294 |
+- MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, |
1295 |
+-#endif |
1296 |
+- MSR_EFER, MSR_TSC_AUX, MSR_STAR, |
1297 |
+- MSR_IA32_TSX_CTRL, |
1298 |
+-}; |
1299 |
+- |
1300 |
+-#if IS_ENABLED(CONFIG_HYPERV) |
1301 |
+-static bool __read_mostly enlightened_vmcs = true; |
1302 |
+-module_param(enlightened_vmcs, bool, 0444); |
1303 |
+- |
1304 |
+-/* check_ept_pointer() should be under protection of ept_pointer_lock. */ |
1305 |
+-static void check_ept_pointer_match(struct kvm *kvm) |
1306 |
+-{ |
1307 |
+- struct kvm_vcpu *vcpu; |
1308 |
+- u64 tmp_eptp = INVALID_PAGE; |
1309 |
+- int i; |
1310 |
+- |
1311 |
+- kvm_for_each_vcpu(i, vcpu, kvm) { |
1312 |
+- if (!VALID_PAGE(tmp_eptp)) { |
1313 |
+- tmp_eptp = to_vmx(vcpu)->ept_pointer; |
1314 |
+- } else if (tmp_eptp != to_vmx(vcpu)->ept_pointer) { |
1315 |
+- to_kvm_vmx(kvm)->ept_pointers_match |
1316 |
+- = EPT_POINTERS_MISMATCH; |
1317 |
+- return; |
1318 |
+- } |
1319 |
+- } |
1320 |
+- |
1321 |
+- to_kvm_vmx(kvm)->ept_pointers_match = EPT_POINTERS_MATCH; |
1322 |
+-} |
1323 |
+- |
1324 |
+-static int kvm_fill_hv_flush_list_func(struct hv_guest_mapping_flush_list *flush, |
1325 |
+- void *data) |
1326 |
+-{ |
1327 |
+- struct kvm_tlb_range *range = data; |
1328 |
+- |
1329 |
+- return hyperv_fill_flush_guest_mapping_list(flush, range->start_gfn, |
1330 |
+- range->pages); |
1331 |
+-} |
1332 |
+- |
1333 |
+-static inline int __hv_remote_flush_tlb_with_range(struct kvm *kvm, |
1334 |
+- struct kvm_vcpu *vcpu, struct kvm_tlb_range *range) |
1335 |
+-{ |
1336 |
+- u64 ept_pointer = to_vmx(vcpu)->ept_pointer; |
1337 |
+- |
1338 |
+- /* |
1339 |
+- * FLUSH_GUEST_PHYSICAL_ADDRESS_SPACE hypercall needs address |
1340 |
+- * of the base of EPT PML4 table, strip off EPT configuration |
1341 |
+- * information. |
1342 |
+- */ |
1343 |
+- if (range) |
1344 |
+- return hyperv_flush_guest_mapping_range(ept_pointer & PAGE_MASK, |
1345 |
+- kvm_fill_hv_flush_list_func, (void *)range); |
1346 |
+- else |
1347 |
+- return hyperv_flush_guest_mapping(ept_pointer & PAGE_MASK); |
1348 |
+-} |
1349 |
+- |
1350 |
+-static int hv_remote_flush_tlb_with_range(struct kvm *kvm, |
1351 |
+- struct kvm_tlb_range *range) |
1352 |
+-{ |
1353 |
+- struct kvm_vcpu *vcpu; |
1354 |
+- int ret = 0, i; |
1355 |
+- |
1356 |
+- spin_lock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
1357 |
+- |
1358 |
+- if (to_kvm_vmx(kvm)->ept_pointers_match == EPT_POINTERS_CHECK) |
1359 |
+- check_ept_pointer_match(kvm); |
1360 |
+- |
1361 |
+- if (to_kvm_vmx(kvm)->ept_pointers_match != EPT_POINTERS_MATCH) { |
1362 |
+- kvm_for_each_vcpu(i, vcpu, kvm) { |
1363 |
+- /* If ept_pointer is invalid pointer, bypass flush request. */ |
1364 |
+- if (VALID_PAGE(to_vmx(vcpu)->ept_pointer)) |
1365 |
+- ret |= __hv_remote_flush_tlb_with_range( |
1366 |
+- kvm, vcpu, range); |
1367 |
+- } |
1368 |
+- } else { |
1369 |
+- ret = __hv_remote_flush_tlb_with_range(kvm, |
1370 |
+- kvm_get_vcpu(kvm, 0), range); |
1371 |
+- } |
1372 |
+- |
1373 |
+- spin_unlock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
1374 |
+- return ret; |
1375 |
+-} |
1376 |
+-static int hv_remote_flush_tlb(struct kvm *kvm) |
1377 |
+-{ |
1378 |
+- return hv_remote_flush_tlb_with_range(kvm, NULL); |
1379 |
+-} |
1380 |
+- |
1381 |
+-static int hv_enable_direct_tlbflush(struct kvm_vcpu *vcpu) |
1382 |
+-{ |
1383 |
+- struct hv_enlightened_vmcs *evmcs; |
1384 |
+- struct hv_partition_assist_pg **p_hv_pa_pg = |
1385 |
+- &vcpu->kvm->arch.hyperv.hv_pa_pg; |
1386 |
+- /* |
1387 |
+- * Synthetic VM-Exit is not enabled in current code and so All |
1388 |
+- * evmcs in singe VM shares same assist page. |
1389 |
+- */ |
1390 |
+- if (!*p_hv_pa_pg) |
1391 |
+- *p_hv_pa_pg = kzalloc(PAGE_SIZE, GFP_KERNEL); |
1392 |
+- |
1393 |
+- if (!*p_hv_pa_pg) |
1394 |
+- return -ENOMEM; |
1395 |
+- |
1396 |
+- evmcs = (struct hv_enlightened_vmcs *)to_vmx(vcpu)->loaded_vmcs->vmcs; |
1397 |
+- |
1398 |
+- evmcs->partition_assist_page = |
1399 |
+- __pa(*p_hv_pa_pg); |
1400 |
+- evmcs->hv_vm_id = (unsigned long)vcpu->kvm; |
1401 |
+- evmcs->hv_enlightenments_control.nested_flush_hypercall = 1; |
1402 |
+- |
1403 |
+- return 0; |
1404 |
+-} |
1405 |
+- |
1406 |
+-#endif /* IS_ENABLED(CONFIG_HYPERV) */ |
1407 |
+- |
1408 |
+-/* |
1409 |
+- * Comment's format: document - errata name - stepping - processor name. |
1410 |
+- * Refer from |
1411 |
+- * https://www.virtualbox.org/svn/vbox/trunk/src/VBox/VMM/VMMR0/HMR0.cpp |
1412 |
+- */ |
1413 |
+-static u32 vmx_preemption_cpu_tfms[] = { |
1414 |
+-/* 323344.pdf - BA86 - D0 - Xeon 7500 Series */ |
1415 |
+-0x000206E6, |
1416 |
+-/* 323056.pdf - AAX65 - C2 - Xeon L3406 */ |
1417 |
+-/* 322814.pdf - AAT59 - C2 - i7-600, i5-500, i5-400 and i3-300 Mobile */ |
1418 |
+-/* 322911.pdf - AAU65 - C2 - i5-600, i3-500 Desktop and Pentium G6950 */ |
1419 |
+-0x00020652, |
1420 |
+-/* 322911.pdf - AAU65 - K0 - i5-600, i3-500 Desktop and Pentium G6950 */ |
1421 |
+-0x00020655, |
1422 |
+-/* 322373.pdf - AAO95 - B1 - Xeon 3400 Series */ |
1423 |
+-/* 322166.pdf - AAN92 - B1 - i7-800 and i5-700 Desktop */ |
1424 |
+-/* |
1425 |
+- * 320767.pdf - AAP86 - B1 - |
1426 |
+- * i7-900 Mobile Extreme, i7-800 and i7-700 Mobile |
1427 |
+- */ |
1428 |
+-0x000106E5, |
1429 |
+-/* 321333.pdf - AAM126 - C0 - Xeon 3500 */ |
1430 |
+-0x000106A0, |
1431 |
+-/* 321333.pdf - AAM126 - C1 - Xeon 3500 */ |
1432 |
+-0x000106A1, |
1433 |
+-/* 320836.pdf - AAJ124 - C0 - i7-900 Desktop Extreme and i7-900 Desktop */ |
1434 |
+-0x000106A4, |
1435 |
+- /* 321333.pdf - AAM126 - D0 - Xeon 3500 */ |
1436 |
+- /* 321324.pdf - AAK139 - D0 - Xeon 5500 */ |
1437 |
+- /* 320836.pdf - AAJ124 - D0 - i7-900 Extreme and i7-900 Desktop */ |
1438 |
+-0x000106A5, |
1439 |
+- /* Xeon E3-1220 V2 */ |
1440 |
+-0x000306A8, |
1441 |
+-}; |
1442 |
+- |
1443 |
+-static inline bool cpu_has_broken_vmx_preemption_timer(void) |
1444 |
+-{ |
1445 |
+- u32 eax = cpuid_eax(0x00000001), i; |
1446 |
+- |
1447 |
+- /* Clear the reserved bits */ |
1448 |
+- eax &= ~(0x3U << 14 | 0xfU << 28); |
1449 |
+- for (i = 0; i < ARRAY_SIZE(vmx_preemption_cpu_tfms); i++) |
1450 |
+- if (eax == vmx_preemption_cpu_tfms[i]) |
1451 |
+- return true; |
1452 |
+- |
1453 |
+- return false; |
1454 |
+-} |
1455 |
+- |
1456 |
+-static inline bool cpu_need_virtualize_apic_accesses(struct kvm_vcpu *vcpu) |
1457 |
+-{ |
1458 |
+- return flexpriority_enabled && lapic_in_kernel(vcpu); |
1459 |
+-} |
1460 |
+- |
1461 |
+-static inline bool report_flexpriority(void) |
1462 |
+-{ |
1463 |
+- return flexpriority_enabled; |
1464 |
+-} |
1465 |
+- |
1466 |
+-static inline int __find_msr_index(struct vcpu_vmx *vmx, u32 msr) |
1467 |
+-{ |
1468 |
+- int i; |
1469 |
+- |
1470 |
+- for (i = 0; i < vmx->nmsrs; ++i) |
1471 |
+- if (vmx_msr_index[vmx->guest_msrs[i].index] == msr) |
1472 |
+- return i; |
1473 |
+- return -1; |
1474 |
+-} |
1475 |
+- |
1476 |
+-struct shared_msr_entry *find_msr_entry(struct vcpu_vmx *vmx, u32 msr) |
1477 |
+-{ |
1478 |
+- int i; |
1479 |
+- |
1480 |
+- i = __find_msr_index(vmx, msr); |
1481 |
+- if (i >= 0) |
1482 |
+- return &vmx->guest_msrs[i]; |
1483 |
+- return NULL; |
1484 |
+-} |
1485 |
+- |
1486 |
+-static int vmx_set_guest_msr(struct vcpu_vmx *vmx, struct shared_msr_entry *msr, u64 data) |
1487 |
+-{ |
1488 |
+- int ret = 0; |
1489 |
+- |
1490 |
+- u64 old_msr_data = msr->data; |
1491 |
+- msr->data = data; |
1492 |
+- if (msr - vmx->guest_msrs < vmx->save_nmsrs) { |
1493 |
+- preempt_disable(); |
1494 |
+- ret = kvm_set_shared_msr(msr->index, msr->data, |
1495 |
+- msr->mask); |
1496 |
+- preempt_enable(); |
1497 |
+- if (ret) |
1498 |
+- msr->data = old_msr_data; |
1499 |
+- } |
1500 |
+- return ret; |
1501 |
+-} |
1502 |
+- |
1503 |
+-void loaded_vmcs_init(struct loaded_vmcs *loaded_vmcs) |
1504 |
+-{ |
1505 |
+- vmcs_clear(loaded_vmcs->vmcs); |
1506 |
+- if (loaded_vmcs->shadow_vmcs && loaded_vmcs->launched) |
1507 |
+- vmcs_clear(loaded_vmcs->shadow_vmcs); |
1508 |
+- loaded_vmcs->cpu = -1; |
1509 |
+- loaded_vmcs->launched = 0; |
1510 |
+-} |
1511 |
+- |
1512 |
+-#ifdef CONFIG_KEXEC_CORE |
1513 |
+-/* |
1514 |
+- * This bitmap is used to indicate whether the vmclear |
1515 |
+- * operation is enabled on all cpus. All disabled by |
1516 |
+- * default. |
1517 |
+- */ |
1518 |
+-static cpumask_t crash_vmclear_enabled_bitmap = CPU_MASK_NONE; |
1519 |
+- |
1520 |
+-static inline void crash_enable_local_vmclear(int cpu) |
1521 |
+-{ |
1522 |
+- cpumask_set_cpu(cpu, &crash_vmclear_enabled_bitmap); |
1523 |
+-} |
1524 |
+- |
1525 |
+-static inline void crash_disable_local_vmclear(int cpu) |
1526 |
+-{ |
1527 |
+- cpumask_clear_cpu(cpu, &crash_vmclear_enabled_bitmap); |
1528 |
+-} |
1529 |
+- |
1530 |
+-static inline int crash_local_vmclear_enabled(int cpu) |
1531 |
+-{ |
1532 |
+- return cpumask_test_cpu(cpu, &crash_vmclear_enabled_bitmap); |
1533 |
+-} |
1534 |
+- |
1535 |
+-static void crash_vmclear_local_loaded_vmcss(void) |
1536 |
+-{ |
1537 |
+- int cpu = raw_smp_processor_id(); |
1538 |
+- struct loaded_vmcs *v; |
1539 |
+- |
1540 |
+- if (!crash_local_vmclear_enabled(cpu)) |
1541 |
+- return; |
1542 |
+- |
1543 |
+- list_for_each_entry(v, &per_cpu(loaded_vmcss_on_cpu, cpu), |
1544 |
+- loaded_vmcss_on_cpu_link) |
1545 |
+- vmcs_clear(v->vmcs); |
1546 |
+-} |
1547 |
+-#else |
1548 |
+-static inline void crash_enable_local_vmclear(int cpu) { } |
1549 |
+-static inline void crash_disable_local_vmclear(int cpu) { } |
1550 |
+-#endif /* CONFIG_KEXEC_CORE */ |
1551 |
+- |
1552 |
+-static void __loaded_vmcs_clear(void *arg) |
1553 |
+-{ |
1554 |
+- struct loaded_vmcs *loaded_vmcs = arg; |
1555 |
+- int cpu = raw_smp_processor_id(); |
1556 |
+- |
1557 |
+- if (loaded_vmcs->cpu != cpu) |
1558 |
+- return; /* vcpu migration can race with cpu offline */ |
1559 |
+- if (per_cpu(current_vmcs, cpu) == loaded_vmcs->vmcs) |
1560 |
+- per_cpu(current_vmcs, cpu) = NULL; |
1561 |
+- crash_disable_local_vmclear(cpu); |
1562 |
+- list_del(&loaded_vmcs->loaded_vmcss_on_cpu_link); |
1563 |
+- |
1564 |
+- /* |
1565 |
+- * we should ensure updating loaded_vmcs->loaded_vmcss_on_cpu_link |
1566 |
+- * is before setting loaded_vmcs->vcpu to -1 which is done in |
1567 |
+- * loaded_vmcs_init. Otherwise, other cpu can see vcpu = -1 fist |
1568 |
+- * then adds the vmcs into percpu list before it is deleted. |
1569 |
+- */ |
1570 |
+- smp_wmb(); |
1571 |
+- |
1572 |
+- loaded_vmcs_init(loaded_vmcs); |
1573 |
+- crash_enable_local_vmclear(cpu); |
1574 |
+-} |
1575 |
+- |
1576 |
+-void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs) |
1577 |
+-{ |
1578 |
+- int cpu = loaded_vmcs->cpu; |
1579 |
+- |
1580 |
+- if (cpu != -1) |
1581 |
+- smp_call_function_single(cpu, |
1582 |
+- __loaded_vmcs_clear, loaded_vmcs, 1); |
1583 |
+-} |
1584 |
+- |
1585 |
+-static bool vmx_segment_cache_test_set(struct vcpu_vmx *vmx, unsigned seg, |
1586 |
+- unsigned field) |
1587 |
+-{ |
1588 |
+- bool ret; |
1589 |
+- u32 mask = 1 << (seg * SEG_FIELD_NR + field); |
1590 |
+- |
1591 |
+- if (!kvm_register_is_available(&vmx->vcpu, VCPU_EXREG_SEGMENTS)) { |
1592 |
+- kvm_register_mark_available(&vmx->vcpu, VCPU_EXREG_SEGMENTS); |
1593 |
+- vmx->segment_cache.bitmask = 0; |
1594 |
+- } |
1595 |
+- ret = vmx->segment_cache.bitmask & mask; |
1596 |
+- vmx->segment_cache.bitmask |= mask; |
1597 |
+- return ret; |
1598 |
+-} |
1599 |
+- |
1600 |
+-static u16 vmx_read_guest_seg_selector(struct vcpu_vmx *vmx, unsigned seg) |
1601 |
+-{ |
1602 |
+- u16 *p = &vmx->segment_cache.seg[seg].selector; |
1603 |
+- |
1604 |
+- if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_SEL)) |
1605 |
+- *p = vmcs_read16(kvm_vmx_segment_fields[seg].selector); |
1606 |
+- return *p; |
1607 |
+-} |
1608 |
+- |
1609 |
+-static ulong vmx_read_guest_seg_base(struct vcpu_vmx *vmx, unsigned seg) |
1610 |
+-{ |
1611 |
+- ulong *p = &vmx->segment_cache.seg[seg].base; |
1612 |
+- |
1613 |
+- if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_BASE)) |
1614 |
+- *p = vmcs_readl(kvm_vmx_segment_fields[seg].base); |
1615 |
+- return *p; |
1616 |
+-} |
1617 |
+- |
1618 |
+-static u32 vmx_read_guest_seg_limit(struct vcpu_vmx *vmx, unsigned seg) |
1619 |
+-{ |
1620 |
+- u32 *p = &vmx->segment_cache.seg[seg].limit; |
1621 |
+- |
1622 |
+- if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_LIMIT)) |
1623 |
+- *p = vmcs_read32(kvm_vmx_segment_fields[seg].limit); |
1624 |
+- return *p; |
1625 |
+-} |
1626 |
+- |
1627 |
+-static u32 vmx_read_guest_seg_ar(struct vcpu_vmx *vmx, unsigned seg) |
1628 |
+-{ |
1629 |
+- u32 *p = &vmx->segment_cache.seg[seg].ar; |
1630 |
+- |
1631 |
+- if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_AR)) |
1632 |
+- *p = vmcs_read32(kvm_vmx_segment_fields[seg].ar_bytes); |
1633 |
+- return *p; |
1634 |
+-} |
1635 |
+- |
1636 |
+-void update_exception_bitmap(struct kvm_vcpu *vcpu) |
1637 |
+-{ |
1638 |
+- u32 eb; |
1639 |
+- |
1640 |
+- eb = (1u << PF_VECTOR) | (1u << UD_VECTOR) | (1u << MC_VECTOR) | |
1641 |
+- (1u << DB_VECTOR) | (1u << AC_VECTOR); |
1642 |
+- /* |
1643 |
+- * Guest access to VMware backdoor ports could legitimately |
1644 |
+- * trigger #GP because of TSS I/O permission bitmap. |
1645 |
+- * We intercept those #GP and allow access to them anyway |
1646 |
+- * as VMware does. |
1647 |
+- */ |
1648 |
+- if (enable_vmware_backdoor) |
1649 |
+- eb |= (1u << GP_VECTOR); |
1650 |
+- if ((vcpu->guest_debug & |
1651 |
+- (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) == |
1652 |
+- (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) |
1653 |
+- eb |= 1u << BP_VECTOR; |
1654 |
+- if (to_vmx(vcpu)->rmode.vm86_active) |
1655 |
+- eb = ~0; |
1656 |
+- if (enable_ept) |
1657 |
+- eb &= ~(1u << PF_VECTOR); /* bypass_guest_pf = 0 */ |
1658 |
+- |
1659 |
+- /* When we are running a nested L2 guest and L1 specified for it a |
1660 |
+- * certain exception bitmap, we must trap the same exceptions and pass |
1661 |
+- * them to L1. When running L2, we will only handle the exceptions |
1662 |
+- * specified above if L1 did not want them. |
1663 |
+- */ |
1664 |
+- if (is_guest_mode(vcpu)) |
1665 |
+- eb |= get_vmcs12(vcpu)->exception_bitmap; |
1666 |
+- |
1667 |
+- vmcs_write32(EXCEPTION_BITMAP, eb); |
1668 |
+-} |
1669 |
+- |
1670 |
+-/* |
1671 |
+- * Check if MSR is intercepted for currently loaded MSR bitmap. |
1672 |
+- */ |
1673 |
+-static bool msr_write_intercepted(struct kvm_vcpu *vcpu, u32 msr) |
1674 |
+-{ |
1675 |
+- unsigned long *msr_bitmap; |
1676 |
+- int f = sizeof(unsigned long); |
1677 |
+- |
1678 |
+- if (!cpu_has_vmx_msr_bitmap()) |
1679 |
+- return true; |
1680 |
+- |
1681 |
+- msr_bitmap = to_vmx(vcpu)->loaded_vmcs->msr_bitmap; |
1682 |
+- |
1683 |
+- if (msr <= 0x1fff) { |
1684 |
+- return !!test_bit(msr, msr_bitmap + 0x800 / f); |
1685 |
+- } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
1686 |
+- msr &= 0x1fff; |
1687 |
+- return !!test_bit(msr, msr_bitmap + 0xc00 / f); |
1688 |
+- } |
1689 |
+- |
1690 |
+- return true; |
1691 |
+-} |
1692 |
+- |
1693 |
+-static void clear_atomic_switch_msr_special(struct vcpu_vmx *vmx, |
1694 |
+- unsigned long entry, unsigned long exit) |
1695 |
+-{ |
1696 |
+- vm_entry_controls_clearbit(vmx, entry); |
1697 |
+- vm_exit_controls_clearbit(vmx, exit); |
1698 |
+-} |
1699 |
+- |
1700 |
+-int vmx_find_msr_index(struct vmx_msrs *m, u32 msr) |
1701 |
+-{ |
1702 |
+- unsigned int i; |
1703 |
+- |
1704 |
+- for (i = 0; i < m->nr; ++i) { |
1705 |
+- if (m->val[i].index == msr) |
1706 |
+- return i; |
1707 |
+- } |
1708 |
+- return -ENOENT; |
1709 |
+-} |
1710 |
+- |
1711 |
+-static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) |
1712 |
+-{ |
1713 |
+- int i; |
1714 |
+- struct msr_autoload *m = &vmx->msr_autoload; |
1715 |
+- |
1716 |
+- switch (msr) { |
1717 |
+- case MSR_EFER: |
1718 |
+- if (cpu_has_load_ia32_efer()) { |
1719 |
+- clear_atomic_switch_msr_special(vmx, |
1720 |
+- VM_ENTRY_LOAD_IA32_EFER, |
1721 |
+- VM_EXIT_LOAD_IA32_EFER); |
1722 |
+- return; |
1723 |
+- } |
1724 |
+- break; |
1725 |
+- case MSR_CORE_PERF_GLOBAL_CTRL: |
1726 |
+- if (cpu_has_load_perf_global_ctrl()) { |
1727 |
+- clear_atomic_switch_msr_special(vmx, |
1728 |
+- VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, |
1729 |
+- VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); |
1730 |
+- return; |
1731 |
+- } |
1732 |
+- break; |
1733 |
+- } |
1734 |
+- i = vmx_find_msr_index(&m->guest, msr); |
1735 |
+- if (i < 0) |
1736 |
+- goto skip_guest; |
1737 |
+- --m->guest.nr; |
1738 |
+- m->guest.val[i] = m->guest.val[m->guest.nr]; |
1739 |
+- vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->guest.nr); |
1740 |
+- |
1741 |
+-skip_guest: |
1742 |
+- i = vmx_find_msr_index(&m->host, msr); |
1743 |
+- if (i < 0) |
1744 |
+- return; |
1745 |
+- |
1746 |
+- --m->host.nr; |
1747 |
+- m->host.val[i] = m->host.val[m->host.nr]; |
1748 |
+- vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->host.nr); |
1749 |
+-} |
1750 |
+- |
1751 |
+-static void add_atomic_switch_msr_special(struct vcpu_vmx *vmx, |
1752 |
+- unsigned long entry, unsigned long exit, |
1753 |
+- unsigned long guest_val_vmcs, unsigned long host_val_vmcs, |
1754 |
+- u64 guest_val, u64 host_val) |
1755 |
+-{ |
1756 |
+- vmcs_write64(guest_val_vmcs, guest_val); |
1757 |
+- if (host_val_vmcs != HOST_IA32_EFER) |
1758 |
+- vmcs_write64(host_val_vmcs, host_val); |
1759 |
+- vm_entry_controls_setbit(vmx, entry); |
1760 |
+- vm_exit_controls_setbit(vmx, exit); |
1761 |
+-} |
1762 |
+- |
1763 |
+-static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, |
1764 |
+- u64 guest_val, u64 host_val, bool entry_only) |
1765 |
+-{ |
1766 |
+- int i, j = 0; |
1767 |
+- struct msr_autoload *m = &vmx->msr_autoload; |
1768 |
+- |
1769 |
+- switch (msr) { |
1770 |
+- case MSR_EFER: |
1771 |
+- if (cpu_has_load_ia32_efer()) { |
1772 |
+- add_atomic_switch_msr_special(vmx, |
1773 |
+- VM_ENTRY_LOAD_IA32_EFER, |
1774 |
+- VM_EXIT_LOAD_IA32_EFER, |
1775 |
+- GUEST_IA32_EFER, |
1776 |
+- HOST_IA32_EFER, |
1777 |
+- guest_val, host_val); |
1778 |
+- return; |
1779 |
+- } |
1780 |
+- break; |
1781 |
+- case MSR_CORE_PERF_GLOBAL_CTRL: |
1782 |
+- if (cpu_has_load_perf_global_ctrl()) { |
1783 |
+- add_atomic_switch_msr_special(vmx, |
1784 |
+- VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, |
1785 |
+- VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL, |
1786 |
+- GUEST_IA32_PERF_GLOBAL_CTRL, |
1787 |
+- HOST_IA32_PERF_GLOBAL_CTRL, |
1788 |
+- guest_val, host_val); |
1789 |
+- return; |
1790 |
+- } |
1791 |
+- break; |
1792 |
+- case MSR_IA32_PEBS_ENABLE: |
1793 |
+- /* PEBS needs a quiescent period after being disabled (to write |
1794 |
+- * a record). Disabling PEBS through VMX MSR swapping doesn't |
1795 |
+- * provide that period, so a CPU could write host's record into |
1796 |
+- * guest's memory. |
1797 |
+- */ |
1798 |
+- wrmsrl(MSR_IA32_PEBS_ENABLE, 0); |
1799 |
+- } |
1800 |
+- |
1801 |
+- i = vmx_find_msr_index(&m->guest, msr); |
1802 |
+- if (!entry_only) |
1803 |
+- j = vmx_find_msr_index(&m->host, msr); |
1804 |
+- |
1805 |
+- if ((i < 0 && m->guest.nr == NR_LOADSTORE_MSRS) || |
1806 |
+- (j < 0 && m->host.nr == NR_LOADSTORE_MSRS)) { |
1807 |
+- printk_once(KERN_WARNING "Not enough msr switch entries. " |
1808 |
+- "Can't add msr %x\n", msr); |
1809 |
+- return; |
1810 |
+- } |
1811 |
+- if (i < 0) { |
1812 |
+- i = m->guest.nr++; |
1813 |
+- vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->guest.nr); |
1814 |
+- } |
1815 |
+- m->guest.val[i].index = msr; |
1816 |
+- m->guest.val[i].value = guest_val; |
1817 |
+- |
1818 |
+- if (entry_only) |
1819 |
+- return; |
1820 |
+- |
1821 |
+- if (j < 0) { |
1822 |
+- j = m->host.nr++; |
1823 |
+- vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->host.nr); |
1824 |
+- } |
1825 |
+- m->host.val[j].index = msr; |
1826 |
+- m->host.val[j].value = host_val; |
1827 |
+-} |
1828 |
+- |
1829 |
+-static bool update_transition_efer(struct vcpu_vmx *vmx, int efer_offset) |
1830 |
+-{ |
1831 |
+- u64 guest_efer = vmx->vcpu.arch.efer; |
1832 |
+- u64 ignore_bits = 0; |
1833 |
+- |
1834 |
+- /* Shadow paging assumes NX to be available. */ |
1835 |
+- if (!enable_ept) |
1836 |
+- guest_efer |= EFER_NX; |
1837 |
+- |
1838 |
+- /* |
1839 |
+- * LMA and LME handled by hardware; SCE meaningless outside long mode. |
1840 |
+- */ |
1841 |
+- ignore_bits |= EFER_SCE; |
1842 |
+-#ifdef CONFIG_X86_64 |
1843 |
+- ignore_bits |= EFER_LMA | EFER_LME; |
1844 |
+- /* SCE is meaningful only in long mode on Intel */ |
1845 |
+- if (guest_efer & EFER_LMA) |
1846 |
+- ignore_bits &= ~(u64)EFER_SCE; |
1847 |
+-#endif |
1848 |
+- |
1849 |
+- /* |
1850 |
+- * On EPT, we can't emulate NX, so we must switch EFER atomically. |
1851 |
+- * On CPUs that support "load IA32_EFER", always switch EFER |
1852 |
+- * atomically, since it's faster than switching it manually. |
1853 |
+- */ |
1854 |
+- if (cpu_has_load_ia32_efer() || |
1855 |
+- (enable_ept && ((vmx->vcpu.arch.efer ^ host_efer) & EFER_NX))) { |
1856 |
+- if (!(guest_efer & EFER_LMA)) |
1857 |
+- guest_efer &= ~EFER_LME; |
1858 |
+- if (guest_efer != host_efer) |
1859 |
+- add_atomic_switch_msr(vmx, MSR_EFER, |
1860 |
+- guest_efer, host_efer, false); |
1861 |
+- else |
1862 |
+- clear_atomic_switch_msr(vmx, MSR_EFER); |
1863 |
+- return false; |
1864 |
+- } else { |
1865 |
+- clear_atomic_switch_msr(vmx, MSR_EFER); |
1866 |
+- |
1867 |
+- guest_efer &= ~ignore_bits; |
1868 |
+- guest_efer |= host_efer & ignore_bits; |
1869 |
+- |
1870 |
+- vmx->guest_msrs[efer_offset].data = guest_efer; |
1871 |
+- vmx->guest_msrs[efer_offset].mask = ~ignore_bits; |
1872 |
+- |
1873 |
+- return true; |
1874 |
+- } |
1875 |
+-} |
1876 |
+- |
1877 |
+-#ifdef CONFIG_X86_32 |
1878 |
+-/* |
1879 |
+- * On 32-bit kernels, VM exits still load the FS and GS bases from the |
1880 |
+- * VMCS rather than the segment table. KVM uses this helper to figure |
1881 |
+- * out the current bases to poke them into the VMCS before entry. |
1882 |
+- */ |
1883 |
+-static unsigned long segment_base(u16 selector) |
1884 |
+-{ |
1885 |
+- struct desc_struct *table; |
1886 |
+- unsigned long v; |
1887 |
+- |
1888 |
+- if (!(selector & ~SEGMENT_RPL_MASK)) |
1889 |
+- return 0; |
1890 |
+- |
1891 |
+- table = get_current_gdt_ro(); |
1892 |
+- |
1893 |
+- if ((selector & SEGMENT_TI_MASK) == SEGMENT_LDT) { |
1894 |
+- u16 ldt_selector = kvm_read_ldt(); |
1895 |
+- |
1896 |
+- if (!(ldt_selector & ~SEGMENT_RPL_MASK)) |
1897 |
+- return 0; |
1898 |
+- |
1899 |
+- table = (struct desc_struct *)segment_base(ldt_selector); |
1900 |
+- } |
1901 |
+- v = get_desc_base(&table[selector >> 3]); |
1902 |
+- return v; |
1903 |
+-} |
1904 |
+-#endif |
1905 |
+- |
1906 |
+-static inline void pt_load_msr(struct pt_ctx *ctx, u32 addr_range) |
1907 |
+-{ |
1908 |
+- u32 i; |
1909 |
+- |
1910 |
+- wrmsrl(MSR_IA32_RTIT_STATUS, ctx->status); |
1911 |
+- wrmsrl(MSR_IA32_RTIT_OUTPUT_BASE, ctx->output_base); |
1912 |
+- wrmsrl(MSR_IA32_RTIT_OUTPUT_MASK, ctx->output_mask); |
1913 |
+- wrmsrl(MSR_IA32_RTIT_CR3_MATCH, ctx->cr3_match); |
1914 |
+- for (i = 0; i < addr_range; i++) { |
1915 |
+- wrmsrl(MSR_IA32_RTIT_ADDR0_A + i * 2, ctx->addr_a[i]); |
1916 |
+- wrmsrl(MSR_IA32_RTIT_ADDR0_B + i * 2, ctx->addr_b[i]); |
1917 |
+- } |
1918 |
+-} |
1919 |
+- |
1920 |
+-static inline void pt_save_msr(struct pt_ctx *ctx, u32 addr_range) |
1921 |
+-{ |
1922 |
+- u32 i; |
1923 |
+- |
1924 |
+- rdmsrl(MSR_IA32_RTIT_STATUS, ctx->status); |
1925 |
+- rdmsrl(MSR_IA32_RTIT_OUTPUT_BASE, ctx->output_base); |
1926 |
+- rdmsrl(MSR_IA32_RTIT_OUTPUT_MASK, ctx->output_mask); |
1927 |
+- rdmsrl(MSR_IA32_RTIT_CR3_MATCH, ctx->cr3_match); |
1928 |
+- for (i = 0; i < addr_range; i++) { |
1929 |
+- rdmsrl(MSR_IA32_RTIT_ADDR0_A + i * 2, ctx->addr_a[i]); |
1930 |
+- rdmsrl(MSR_IA32_RTIT_ADDR0_B + i * 2, ctx->addr_b[i]); |
1931 |
+- } |
1932 |
+-} |
1933 |
+- |
1934 |
+-static void pt_guest_enter(struct vcpu_vmx *vmx) |
1935 |
+-{ |
1936 |
+- if (pt_mode == PT_MODE_SYSTEM) |
1937 |
+- return; |
1938 |
+- |
1939 |
+- /* |
1940 |
+- * GUEST_IA32_RTIT_CTL is already set in the VMCS. |
1941 |
+- * Save host state before VM entry. |
1942 |
+- */ |
1943 |
+- rdmsrl(MSR_IA32_RTIT_CTL, vmx->pt_desc.host.ctl); |
1944 |
+- if (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) { |
1945 |
+- wrmsrl(MSR_IA32_RTIT_CTL, 0); |
1946 |
+- pt_save_msr(&vmx->pt_desc.host, vmx->pt_desc.addr_range); |
1947 |
+- pt_load_msr(&vmx->pt_desc.guest, vmx->pt_desc.addr_range); |
1948 |
+- } |
1949 |
+-} |
1950 |
+- |
1951 |
+-static void pt_guest_exit(struct vcpu_vmx *vmx) |
1952 |
+-{ |
1953 |
+- if (pt_mode == PT_MODE_SYSTEM) |
1954 |
+- return; |
1955 |
+- |
1956 |
+- if (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) { |
1957 |
+- pt_save_msr(&vmx->pt_desc.guest, vmx->pt_desc.addr_range); |
1958 |
+- pt_load_msr(&vmx->pt_desc.host, vmx->pt_desc.addr_range); |
1959 |
+- } |
1960 |
+- |
1961 |
+- /* Reload host state (IA32_RTIT_CTL will be cleared on VM exit). */ |
1962 |
+- wrmsrl(MSR_IA32_RTIT_CTL, vmx->pt_desc.host.ctl); |
1963 |
+-} |
1964 |
+- |
1965 |
+-void vmx_set_host_fs_gs(struct vmcs_host_state *host, u16 fs_sel, u16 gs_sel, |
1966 |
+- unsigned long fs_base, unsigned long gs_base) |
1967 |
+-{ |
1968 |
+- if (unlikely(fs_sel != host->fs_sel)) { |
1969 |
+- if (!(fs_sel & 7)) |
1970 |
+- vmcs_write16(HOST_FS_SELECTOR, fs_sel); |
1971 |
+- else |
1972 |
+- vmcs_write16(HOST_FS_SELECTOR, 0); |
1973 |
+- host->fs_sel = fs_sel; |
1974 |
+- } |
1975 |
+- if (unlikely(gs_sel != host->gs_sel)) { |
1976 |
+- if (!(gs_sel & 7)) |
1977 |
+- vmcs_write16(HOST_GS_SELECTOR, gs_sel); |
1978 |
+- else |
1979 |
+- vmcs_write16(HOST_GS_SELECTOR, 0); |
1980 |
+- host->gs_sel = gs_sel; |
1981 |
+- } |
1982 |
+- if (unlikely(fs_base != host->fs_base)) { |
1983 |
+- vmcs_writel(HOST_FS_BASE, fs_base); |
1984 |
+- host->fs_base = fs_base; |
1985 |
+- } |
1986 |
+- if (unlikely(gs_base != host->gs_base)) { |
1987 |
+- vmcs_writel(HOST_GS_BASE, gs_base); |
1988 |
+- host->gs_base = gs_base; |
1989 |
+- } |
1990 |
+-} |
1991 |
+- |
1992 |
+-void vmx_prepare_switch_to_guest(struct kvm_vcpu *vcpu) |
1993 |
+-{ |
1994 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
1995 |
+- struct vmcs_host_state *host_state; |
1996 |
+-#ifdef CONFIG_X86_64 |
1997 |
+- int cpu = raw_smp_processor_id(); |
1998 |
+-#endif |
1999 |
+- unsigned long fs_base, gs_base; |
2000 |
+- u16 fs_sel, gs_sel; |
2001 |
+- int i; |
2002 |
+- |
2003 |
+- vmx->req_immediate_exit = false; |
2004 |
+- |
2005 |
+- /* |
2006 |
+- * Note that guest MSRs to be saved/restored can also be changed |
2007 |
+- * when guest state is loaded. This happens when guest transitions |
2008 |
+- * to/from long-mode by setting MSR_EFER.LMA. |
2009 |
+- */ |
2010 |
+- if (!vmx->guest_msrs_ready) { |
2011 |
+- vmx->guest_msrs_ready = true; |
2012 |
+- for (i = 0; i < vmx->save_nmsrs; ++i) |
2013 |
+- kvm_set_shared_msr(vmx->guest_msrs[i].index, |
2014 |
+- vmx->guest_msrs[i].data, |
2015 |
+- vmx->guest_msrs[i].mask); |
2016 |
+- |
2017 |
+- } |
2018 |
+- if (vmx->guest_state_loaded) |
2019 |
+- return; |
2020 |
+- |
2021 |
+- host_state = &vmx->loaded_vmcs->host_state; |
2022 |
+- |
2023 |
+- /* |
2024 |
+- * Set host fs and gs selectors. Unfortunately, 22.2.3 does not |
2025 |
+- * allow segment selectors with cpl > 0 or ti == 1. |
2026 |
+- */ |
2027 |
+- host_state->ldt_sel = kvm_read_ldt(); |
2028 |
+- |
2029 |
+-#ifdef CONFIG_X86_64 |
2030 |
+- savesegment(ds, host_state->ds_sel); |
2031 |
+- savesegment(es, host_state->es_sel); |
2032 |
+- |
2033 |
+- gs_base = cpu_kernelmode_gs_base(cpu); |
2034 |
+- if (likely(is_64bit_mm(current->mm))) { |
2035 |
+- save_fsgs_for_kvm(); |
2036 |
+- fs_sel = current->thread.fsindex; |
2037 |
+- gs_sel = current->thread.gsindex; |
2038 |
+- fs_base = current->thread.fsbase; |
2039 |
+- vmx->msr_host_kernel_gs_base = current->thread.gsbase; |
2040 |
+- } else { |
2041 |
+- savesegment(fs, fs_sel); |
2042 |
+- savesegment(gs, gs_sel); |
2043 |
+- fs_base = read_msr(MSR_FS_BASE); |
2044 |
+- vmx->msr_host_kernel_gs_base = read_msr(MSR_KERNEL_GS_BASE); |
2045 |
+- } |
2046 |
+- |
2047 |
+- wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2048 |
+-#else |
2049 |
+- savesegment(fs, fs_sel); |
2050 |
+- savesegment(gs, gs_sel); |
2051 |
+- fs_base = segment_base(fs_sel); |
2052 |
+- gs_base = segment_base(gs_sel); |
2053 |
+-#endif |
2054 |
+- |
2055 |
+- vmx_set_host_fs_gs(host_state, fs_sel, gs_sel, fs_base, gs_base); |
2056 |
+- vmx->guest_state_loaded = true; |
2057 |
+-} |
2058 |
+- |
2059 |
+-static void vmx_prepare_switch_to_host(struct vcpu_vmx *vmx) |
2060 |
+-{ |
2061 |
+- struct vmcs_host_state *host_state; |
2062 |
+- |
2063 |
+- if (!vmx->guest_state_loaded) |
2064 |
+- return; |
2065 |
+- |
2066 |
+- host_state = &vmx->loaded_vmcs->host_state; |
2067 |
+- |
2068 |
+- ++vmx->vcpu.stat.host_state_reload; |
2069 |
+- |
2070 |
+-#ifdef CONFIG_X86_64 |
2071 |
+- rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2072 |
+-#endif |
2073 |
+- if (host_state->ldt_sel || (host_state->gs_sel & 7)) { |
2074 |
+- kvm_load_ldt(host_state->ldt_sel); |
2075 |
+-#ifdef CONFIG_X86_64 |
2076 |
+- load_gs_index(host_state->gs_sel); |
2077 |
+-#else |
2078 |
+- loadsegment(gs, host_state->gs_sel); |
2079 |
+-#endif |
2080 |
+- } |
2081 |
+- if (host_state->fs_sel & 7) |
2082 |
+- loadsegment(fs, host_state->fs_sel); |
2083 |
+-#ifdef CONFIG_X86_64 |
2084 |
+- if (unlikely(host_state->ds_sel | host_state->es_sel)) { |
2085 |
+- loadsegment(ds, host_state->ds_sel); |
2086 |
+- loadsegment(es, host_state->es_sel); |
2087 |
+- } |
2088 |
+-#endif |
2089 |
+- invalidate_tss_limit(); |
2090 |
+-#ifdef CONFIG_X86_64 |
2091 |
+- wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); |
2092 |
+-#endif |
2093 |
+- load_fixmap_gdt(raw_smp_processor_id()); |
2094 |
+- vmx->guest_state_loaded = false; |
2095 |
+- vmx->guest_msrs_ready = false; |
2096 |
+-} |
2097 |
+- |
2098 |
+-#ifdef CONFIG_X86_64 |
2099 |
+-static u64 vmx_read_guest_kernel_gs_base(struct vcpu_vmx *vmx) |
2100 |
+-{ |
2101 |
+- preempt_disable(); |
2102 |
+- if (vmx->guest_state_loaded) |
2103 |
+- rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
2104 |
+- preempt_enable(); |
2105 |
+- return vmx->msr_guest_kernel_gs_base; |
2106 |
+-} |
2107 |
+- |
2108 |
+-static void vmx_write_guest_kernel_gs_base(struct vcpu_vmx *vmx, u64 data) |
2109 |
+-{ |
2110 |
+- preempt_disable(); |
2111 |
+- if (vmx->guest_state_loaded) |
2112 |
+- wrmsrl(MSR_KERNEL_GS_BASE, data); |
2113 |
+- preempt_enable(); |
2114 |
+- vmx->msr_guest_kernel_gs_base = data; |
2115 |
+-} |
2116 |
+-#endif |
2117 |
+- |
2118 |
+-static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu) |
2119 |
+-{ |
2120 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
2121 |
+- struct pi_desc old, new; |
2122 |
+- unsigned int dest; |
2123 |
+- |
2124 |
+- /* |
2125 |
+- * In case of hot-plug or hot-unplug, we may have to undo |
2126 |
+- * vmx_vcpu_pi_put even if there is no assigned device. And we |
2127 |
+- * always keep PI.NDST up to date for simplicity: it makes the |
2128 |
+- * code easier, and CPU migration is not a fast path. |
2129 |
+- */ |
2130 |
+- if (!pi_test_sn(pi_desc) && vcpu->cpu == cpu) |
2131 |
+- return; |
2132 |
+- |
2133 |
+- /* |
2134 |
+- * If the 'nv' field is POSTED_INTR_WAKEUP_VECTOR, do not change |
2135 |
+- * PI.NDST: pi_post_block is the one expected to change PID.NDST and the |
2136 |
+- * wakeup handler expects the vCPU to be on the blocked_vcpu_list that |
2137 |
+- * matches PI.NDST. Otherwise, a vcpu may not be able to be woken up |
2138 |
+- * correctly. |
2139 |
+- */ |
2140 |
+- if (pi_desc->nv == POSTED_INTR_WAKEUP_VECTOR || vcpu->cpu == cpu) { |
2141 |
+- pi_clear_sn(pi_desc); |
2142 |
+- goto after_clear_sn; |
2143 |
+- } |
2144 |
+- |
2145 |
+- /* The full case. */ |
2146 |
+- do { |
2147 |
+- old.control = new.control = pi_desc->control; |
2148 |
+- |
2149 |
+- dest = cpu_physical_id(cpu); |
2150 |
+- |
2151 |
+- if (x2apic_enabled()) |
2152 |
+- new.ndst = dest; |
2153 |
+- else |
2154 |
+- new.ndst = (dest << 8) & 0xFF00; |
2155 |
+- |
2156 |
+- new.sn = 0; |
2157 |
+- } while (cmpxchg64(&pi_desc->control, old.control, |
2158 |
+- new.control) != old.control); |
2159 |
+- |
2160 |
+-after_clear_sn: |
2161 |
+- |
2162 |
+- /* |
2163 |
+- * Clear SN before reading the bitmap. The VT-d firmware |
2164 |
+- * writes the bitmap and reads SN atomically (5.2.3 in the |
2165 |
+- * spec), so it doesn't really have a memory barrier that |
2166 |
+- * pairs with this, but we cannot do that and we need one. |
2167 |
+- */ |
2168 |
+- smp_mb__after_atomic(); |
2169 |
+- |
2170 |
+- if (!pi_is_pir_empty(pi_desc)) |
2171 |
+- pi_set_on(pi_desc); |
2172 |
+-} |
2173 |
+- |
2174 |
+-void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu) |
2175 |
+-{ |
2176 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2177 |
+- bool already_loaded = vmx->loaded_vmcs->cpu == cpu; |
2178 |
+- |
2179 |
+- if (!already_loaded) { |
2180 |
+- loaded_vmcs_clear(vmx->loaded_vmcs); |
2181 |
+- local_irq_disable(); |
2182 |
+- crash_disable_local_vmclear(cpu); |
2183 |
+- |
2184 |
+- /* |
2185 |
+- * Read loaded_vmcs->cpu should be before fetching |
2186 |
+- * loaded_vmcs->loaded_vmcss_on_cpu_link. |
2187 |
+- * See the comments in __loaded_vmcs_clear(). |
2188 |
+- */ |
2189 |
+- smp_rmb(); |
2190 |
+- |
2191 |
+- list_add(&vmx->loaded_vmcs->loaded_vmcss_on_cpu_link, |
2192 |
+- &per_cpu(loaded_vmcss_on_cpu, cpu)); |
2193 |
+- crash_enable_local_vmclear(cpu); |
2194 |
+- local_irq_enable(); |
2195 |
+- } |
2196 |
+- |
2197 |
+- if (per_cpu(current_vmcs, cpu) != vmx->loaded_vmcs->vmcs) { |
2198 |
+- per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs; |
2199 |
+- vmcs_load(vmx->loaded_vmcs->vmcs); |
2200 |
+- indirect_branch_prediction_barrier(); |
2201 |
+- } |
2202 |
+- |
2203 |
+- if (!already_loaded) { |
2204 |
+- void *gdt = get_current_gdt_ro(); |
2205 |
+- unsigned long sysenter_esp; |
2206 |
+- |
2207 |
+- kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
2208 |
+- |
2209 |
+- /* |
2210 |
+- * Linux uses per-cpu TSS and GDT, so set these when switching |
2211 |
+- * processors. See 22.2.4. |
2212 |
+- */ |
2213 |
+- vmcs_writel(HOST_TR_BASE, |
2214 |
+- (unsigned long)&get_cpu_entry_area(cpu)->tss.x86_tss); |
2215 |
+- vmcs_writel(HOST_GDTR_BASE, (unsigned long)gdt); /* 22.2.4 */ |
2216 |
+- |
2217 |
+- rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp); |
2218 |
+- vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */ |
2219 |
+- |
2220 |
+- vmx->loaded_vmcs->cpu = cpu; |
2221 |
+- } |
2222 |
+- |
2223 |
+- /* Setup TSC multiplier */ |
2224 |
+- if (kvm_has_tsc_control && |
2225 |
+- vmx->current_tsc_ratio != vcpu->arch.tsc_scaling_ratio) |
2226 |
+- decache_tsc_multiplier(vmx); |
2227 |
+-} |
2228 |
+- |
2229 |
+-/* |
2230 |
+- * Switches to specified vcpu, until a matching vcpu_put(), but assumes |
2231 |
+- * vcpu mutex is already taken. |
2232 |
+- */ |
2233 |
+-void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
2234 |
+-{ |
2235 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2236 |
+- |
2237 |
+- vmx_vcpu_load_vmcs(vcpu, cpu); |
2238 |
+- |
2239 |
+- vmx_vcpu_pi_load(vcpu, cpu); |
2240 |
+- |
2241 |
+- vmx->host_pkru = read_pkru(); |
2242 |
+- vmx->host_debugctlmsr = get_debugctlmsr(); |
2243 |
+-} |
2244 |
+- |
2245 |
+-static void vmx_vcpu_pi_put(struct kvm_vcpu *vcpu) |
2246 |
+-{ |
2247 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
2248 |
+- |
2249 |
+- if (!kvm_arch_has_assigned_device(vcpu->kvm) || |
2250 |
+- !irq_remapping_cap(IRQ_POSTING_CAP) || |
2251 |
+- !kvm_vcpu_apicv_active(vcpu)) |
2252 |
+- return; |
2253 |
+- |
2254 |
+- /* Set SN when the vCPU is preempted */ |
2255 |
+- if (vcpu->preempted) |
2256 |
+- pi_set_sn(pi_desc); |
2257 |
+-} |
2258 |
+- |
2259 |
+-static void vmx_vcpu_put(struct kvm_vcpu *vcpu) |
2260 |
+-{ |
2261 |
+- vmx_vcpu_pi_put(vcpu); |
2262 |
+- |
2263 |
+- vmx_prepare_switch_to_host(to_vmx(vcpu)); |
2264 |
+-} |
2265 |
+- |
2266 |
+-static bool emulation_required(struct kvm_vcpu *vcpu) |
2267 |
+-{ |
2268 |
+- return emulate_invalid_guest_state && !guest_state_valid(vcpu); |
2269 |
+-} |
2270 |
+- |
2271 |
+-static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu); |
2272 |
+- |
2273 |
+-unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) |
2274 |
+-{ |
2275 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2276 |
+- unsigned long rflags, save_rflags; |
2277 |
+- |
2278 |
+- if (!kvm_register_is_available(vcpu, VCPU_EXREG_RFLAGS)) { |
2279 |
+- kvm_register_mark_available(vcpu, VCPU_EXREG_RFLAGS); |
2280 |
+- rflags = vmcs_readl(GUEST_RFLAGS); |
2281 |
+- if (vmx->rmode.vm86_active) { |
2282 |
+- rflags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
2283 |
+- save_rflags = vmx->rmode.save_rflags; |
2284 |
+- rflags |= save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; |
2285 |
+- } |
2286 |
+- vmx->rflags = rflags; |
2287 |
+- } |
2288 |
+- return vmx->rflags; |
2289 |
+-} |
2290 |
+- |
2291 |
+-void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) |
2292 |
+-{ |
2293 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2294 |
+- unsigned long old_rflags; |
2295 |
+- |
2296 |
+- if (enable_unrestricted_guest) { |
2297 |
+- kvm_register_mark_available(vcpu, VCPU_EXREG_RFLAGS); |
2298 |
+- vmx->rflags = rflags; |
2299 |
+- vmcs_writel(GUEST_RFLAGS, rflags); |
2300 |
+- return; |
2301 |
+- } |
2302 |
+- |
2303 |
+- old_rflags = vmx_get_rflags(vcpu); |
2304 |
+- vmx->rflags = rflags; |
2305 |
+- if (vmx->rmode.vm86_active) { |
2306 |
+- vmx->rmode.save_rflags = rflags; |
2307 |
+- rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
2308 |
+- } |
2309 |
+- vmcs_writel(GUEST_RFLAGS, rflags); |
2310 |
+- |
2311 |
+- if ((old_rflags ^ vmx->rflags) & X86_EFLAGS_VM) |
2312 |
+- vmx->emulation_required = emulation_required(vcpu); |
2313 |
+-} |
2314 |
+- |
2315 |
+-u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu) |
2316 |
+-{ |
2317 |
+- u32 interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); |
2318 |
+- int ret = 0; |
2319 |
+- |
2320 |
+- if (interruptibility & GUEST_INTR_STATE_STI) |
2321 |
+- ret |= KVM_X86_SHADOW_INT_STI; |
2322 |
+- if (interruptibility & GUEST_INTR_STATE_MOV_SS) |
2323 |
+- ret |= KVM_X86_SHADOW_INT_MOV_SS; |
2324 |
+- |
2325 |
+- return ret; |
2326 |
+-} |
2327 |
+- |
2328 |
+-void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) |
2329 |
+-{ |
2330 |
+- u32 interruptibility_old = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); |
2331 |
+- u32 interruptibility = interruptibility_old; |
2332 |
+- |
2333 |
+- interruptibility &= ~(GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS); |
2334 |
+- |
2335 |
+- if (mask & KVM_X86_SHADOW_INT_MOV_SS) |
2336 |
+- interruptibility |= GUEST_INTR_STATE_MOV_SS; |
2337 |
+- else if (mask & KVM_X86_SHADOW_INT_STI) |
2338 |
+- interruptibility |= GUEST_INTR_STATE_STI; |
2339 |
+- |
2340 |
+- if ((interruptibility != interruptibility_old)) |
2341 |
+- vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, interruptibility); |
2342 |
+-} |
2343 |
+- |
2344 |
+-static int vmx_rtit_ctl_check(struct kvm_vcpu *vcpu, u64 data) |
2345 |
+-{ |
2346 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2347 |
+- unsigned long value; |
2348 |
+- |
2349 |
+- /* |
2350 |
+- * Any MSR write that attempts to change bits marked reserved will |
2351 |
+- * case a #GP fault. |
2352 |
+- */ |
2353 |
+- if (data & vmx->pt_desc.ctl_bitmask) |
2354 |
+- return 1; |
2355 |
+- |
2356 |
+- /* |
2357 |
+- * Any attempt to modify IA32_RTIT_CTL while TraceEn is set will |
2358 |
+- * result in a #GP unless the same write also clears TraceEn. |
2359 |
+- */ |
2360 |
+- if ((vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) && |
2361 |
+- ((vmx->pt_desc.guest.ctl ^ data) & ~RTIT_CTL_TRACEEN)) |
2362 |
+- return 1; |
2363 |
+- |
2364 |
+- /* |
2365 |
+- * WRMSR to IA32_RTIT_CTL that sets TraceEn but clears this bit |
2366 |
+- * and FabricEn would cause #GP, if |
2367 |
+- * CPUID.(EAX=14H, ECX=0):ECX.SNGLRGNOUT[bit 2] = 0 |
2368 |
+- */ |
2369 |
+- if ((data & RTIT_CTL_TRACEEN) && !(data & RTIT_CTL_TOPA) && |
2370 |
+- !(data & RTIT_CTL_FABRIC_EN) && |
2371 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2372 |
+- PT_CAP_single_range_output)) |
2373 |
+- return 1; |
2374 |
+- |
2375 |
+- /* |
2376 |
+- * MTCFreq, CycThresh and PSBFreq encodings check, any MSR write that |
2377 |
+- * utilize encodings marked reserved will casue a #GP fault. |
2378 |
+- */ |
2379 |
+- value = intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc_periods); |
2380 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc) && |
2381 |
+- !test_bit((data & RTIT_CTL_MTC_RANGE) >> |
2382 |
+- RTIT_CTL_MTC_RANGE_OFFSET, &value)) |
2383 |
+- return 1; |
2384 |
+- value = intel_pt_validate_cap(vmx->pt_desc.caps, |
2385 |
+- PT_CAP_cycle_thresholds); |
2386 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc) && |
2387 |
+- !test_bit((data & RTIT_CTL_CYC_THRESH) >> |
2388 |
+- RTIT_CTL_CYC_THRESH_OFFSET, &value)) |
2389 |
+- return 1; |
2390 |
+- value = intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_periods); |
2391 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc) && |
2392 |
+- !test_bit((data & RTIT_CTL_PSB_FREQ) >> |
2393 |
+- RTIT_CTL_PSB_FREQ_OFFSET, &value)) |
2394 |
+- return 1; |
2395 |
+- |
2396 |
+- /* |
2397 |
+- * If ADDRx_CFG is reserved or the encodings is >2 will |
2398 |
+- * cause a #GP fault. |
2399 |
+- */ |
2400 |
+- value = (data & RTIT_CTL_ADDR0) >> RTIT_CTL_ADDR0_OFFSET; |
2401 |
+- if ((value && (vmx->pt_desc.addr_range < 1)) || (value > 2)) |
2402 |
+- return 1; |
2403 |
+- value = (data & RTIT_CTL_ADDR1) >> RTIT_CTL_ADDR1_OFFSET; |
2404 |
+- if ((value && (vmx->pt_desc.addr_range < 2)) || (value > 2)) |
2405 |
+- return 1; |
2406 |
+- value = (data & RTIT_CTL_ADDR2) >> RTIT_CTL_ADDR2_OFFSET; |
2407 |
+- if ((value && (vmx->pt_desc.addr_range < 3)) || (value > 2)) |
2408 |
+- return 1; |
2409 |
+- value = (data & RTIT_CTL_ADDR3) >> RTIT_CTL_ADDR3_OFFSET; |
2410 |
+- if ((value && (vmx->pt_desc.addr_range < 4)) || (value > 2)) |
2411 |
+- return 1; |
2412 |
+- |
2413 |
+- return 0; |
2414 |
+-} |
2415 |
+- |
2416 |
+-static int skip_emulated_instruction(struct kvm_vcpu *vcpu) |
2417 |
+-{ |
2418 |
+- unsigned long rip; |
2419 |
+- |
2420 |
+- /* |
2421 |
+- * Using VMCS.VM_EXIT_INSTRUCTION_LEN on EPT misconfig depends on |
2422 |
+- * undefined behavior: Intel's SDM doesn't mandate the VMCS field be |
2423 |
+- * set when EPT misconfig occurs. In practice, real hardware updates |
2424 |
+- * VM_EXIT_INSTRUCTION_LEN on EPT misconfig, but other hypervisors |
2425 |
+- * (namely Hyper-V) don't set it due to it being undefined behavior, |
2426 |
+- * i.e. we end up advancing IP with some random value. |
2427 |
+- */ |
2428 |
+- if (!static_cpu_has(X86_FEATURE_HYPERVISOR) || |
2429 |
+- to_vmx(vcpu)->exit_reason != EXIT_REASON_EPT_MISCONFIG) { |
2430 |
+- rip = kvm_rip_read(vcpu); |
2431 |
+- rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
2432 |
+- kvm_rip_write(vcpu, rip); |
2433 |
+- } else { |
2434 |
+- if (!kvm_emulate_instruction(vcpu, EMULTYPE_SKIP)) |
2435 |
+- return 0; |
2436 |
+- } |
2437 |
+- |
2438 |
+- /* skipping an emulated instruction also counts */ |
2439 |
+- vmx_set_interrupt_shadow(vcpu, 0); |
2440 |
+- |
2441 |
+- return 1; |
2442 |
+-} |
2443 |
+- |
2444 |
+-static void vmx_clear_hlt(struct kvm_vcpu *vcpu) |
2445 |
+-{ |
2446 |
+- /* |
2447 |
+- * Ensure that we clear the HLT state in the VMCS. We don't need to |
2448 |
+- * explicitly skip the instruction because if the HLT state is set, |
2449 |
+- * then the instruction is already executing and RIP has already been |
2450 |
+- * advanced. |
2451 |
+- */ |
2452 |
+- if (kvm_hlt_in_guest(vcpu->kvm) && |
2453 |
+- vmcs_read32(GUEST_ACTIVITY_STATE) == GUEST_ACTIVITY_HLT) |
2454 |
+- vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE); |
2455 |
+-} |
2456 |
+- |
2457 |
+-static void vmx_queue_exception(struct kvm_vcpu *vcpu) |
2458 |
+-{ |
2459 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2460 |
+- unsigned nr = vcpu->arch.exception.nr; |
2461 |
+- bool has_error_code = vcpu->arch.exception.has_error_code; |
2462 |
+- u32 error_code = vcpu->arch.exception.error_code; |
2463 |
+- u32 intr_info = nr | INTR_INFO_VALID_MASK; |
2464 |
+- |
2465 |
+- kvm_deliver_exception_payload(vcpu); |
2466 |
+- |
2467 |
+- if (has_error_code) { |
2468 |
+- vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code); |
2469 |
+- intr_info |= INTR_INFO_DELIVER_CODE_MASK; |
2470 |
+- } |
2471 |
+- |
2472 |
+- if (vmx->rmode.vm86_active) { |
2473 |
+- int inc_eip = 0; |
2474 |
+- if (kvm_exception_is_soft(nr)) |
2475 |
+- inc_eip = vcpu->arch.event_exit_inst_len; |
2476 |
+- kvm_inject_realmode_interrupt(vcpu, nr, inc_eip); |
2477 |
+- return; |
2478 |
+- } |
2479 |
+- |
2480 |
+- WARN_ON_ONCE(vmx->emulation_required); |
2481 |
+- |
2482 |
+- if (kvm_exception_is_soft(nr)) { |
2483 |
+- vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, |
2484 |
+- vmx->vcpu.arch.event_exit_inst_len); |
2485 |
+- intr_info |= INTR_TYPE_SOFT_EXCEPTION; |
2486 |
+- } else |
2487 |
+- intr_info |= INTR_TYPE_HARD_EXCEPTION; |
2488 |
+- |
2489 |
+- vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info); |
2490 |
+- |
2491 |
+- vmx_clear_hlt(vcpu); |
2492 |
+-} |
2493 |
+- |
2494 |
+-static bool vmx_rdtscp_supported(void) |
2495 |
+-{ |
2496 |
+- return cpu_has_vmx_rdtscp(); |
2497 |
+-} |
2498 |
+- |
2499 |
+-static bool vmx_invpcid_supported(void) |
2500 |
+-{ |
2501 |
+- return cpu_has_vmx_invpcid(); |
2502 |
+-} |
2503 |
+- |
2504 |
+-/* |
2505 |
+- * Swap MSR entry in host/guest MSR entry array. |
2506 |
+- */ |
2507 |
+-static void move_msr_up(struct vcpu_vmx *vmx, int from, int to) |
2508 |
+-{ |
2509 |
+- struct shared_msr_entry tmp; |
2510 |
+- |
2511 |
+- tmp = vmx->guest_msrs[to]; |
2512 |
+- vmx->guest_msrs[to] = vmx->guest_msrs[from]; |
2513 |
+- vmx->guest_msrs[from] = tmp; |
2514 |
+-} |
2515 |
+- |
2516 |
+-/* |
2517 |
+- * Set up the vmcs to automatically save and restore system |
2518 |
+- * msrs. Don't touch the 64-bit msrs if the guest is in legacy |
2519 |
+- * mode, as fiddling with msrs is very expensive. |
2520 |
+- */ |
2521 |
+-static void setup_msrs(struct vcpu_vmx *vmx) |
2522 |
+-{ |
2523 |
+- int save_nmsrs, index; |
2524 |
+- |
2525 |
+- save_nmsrs = 0; |
2526 |
+-#ifdef CONFIG_X86_64 |
2527 |
+- /* |
2528 |
+- * The SYSCALL MSRs are only needed on long mode guests, and only |
2529 |
+- * when EFER.SCE is set. |
2530 |
+- */ |
2531 |
+- if (is_long_mode(&vmx->vcpu) && (vmx->vcpu.arch.efer & EFER_SCE)) { |
2532 |
+- index = __find_msr_index(vmx, MSR_STAR); |
2533 |
+- if (index >= 0) |
2534 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2535 |
+- index = __find_msr_index(vmx, MSR_LSTAR); |
2536 |
+- if (index >= 0) |
2537 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2538 |
+- index = __find_msr_index(vmx, MSR_SYSCALL_MASK); |
2539 |
+- if (index >= 0) |
2540 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2541 |
+- } |
2542 |
+-#endif |
2543 |
+- index = __find_msr_index(vmx, MSR_EFER); |
2544 |
+- if (index >= 0 && update_transition_efer(vmx, index)) |
2545 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2546 |
+- index = __find_msr_index(vmx, MSR_TSC_AUX); |
2547 |
+- if (index >= 0 && guest_cpuid_has(&vmx->vcpu, X86_FEATURE_RDTSCP)) |
2548 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2549 |
+- index = __find_msr_index(vmx, MSR_IA32_TSX_CTRL); |
2550 |
+- if (index >= 0) |
2551 |
+- move_msr_up(vmx, index, save_nmsrs++); |
2552 |
+- |
2553 |
+- vmx->save_nmsrs = save_nmsrs; |
2554 |
+- vmx->guest_msrs_ready = false; |
2555 |
+- |
2556 |
+- if (cpu_has_vmx_msr_bitmap()) |
2557 |
+- vmx_update_msr_bitmap(&vmx->vcpu); |
2558 |
+-} |
2559 |
+- |
2560 |
+-static u64 vmx_read_l1_tsc_offset(struct kvm_vcpu *vcpu) |
2561 |
+-{ |
2562 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
2563 |
+- |
2564 |
+- if (is_guest_mode(vcpu) && |
2565 |
+- (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)) |
2566 |
+- return vcpu->arch.tsc_offset - vmcs12->tsc_offset; |
2567 |
+- |
2568 |
+- return vcpu->arch.tsc_offset; |
2569 |
+-} |
2570 |
+- |
2571 |
+-static u64 vmx_write_l1_tsc_offset(struct kvm_vcpu *vcpu, u64 offset) |
2572 |
+-{ |
2573 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
2574 |
+- u64 g_tsc_offset = 0; |
2575 |
+- |
2576 |
+- /* |
2577 |
+- * We're here if L1 chose not to trap WRMSR to TSC. According |
2578 |
+- * to the spec, this should set L1's TSC; The offset that L1 |
2579 |
+- * set for L2 remains unchanged, and still needs to be added |
2580 |
+- * to the newly set TSC to get L2's TSC. |
2581 |
+- */ |
2582 |
+- if (is_guest_mode(vcpu) && |
2583 |
+- (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETTING)) |
2584 |
+- g_tsc_offset = vmcs12->tsc_offset; |
2585 |
+- |
2586 |
+- trace_kvm_write_tsc_offset(vcpu->vcpu_id, |
2587 |
+- vcpu->arch.tsc_offset - g_tsc_offset, |
2588 |
+- offset); |
2589 |
+- vmcs_write64(TSC_OFFSET, offset + g_tsc_offset); |
2590 |
+- return offset + g_tsc_offset; |
2591 |
+-} |
2592 |
+- |
2593 |
+-/* |
2594 |
+- * nested_vmx_allowed() checks whether a guest should be allowed to use VMX |
2595 |
+- * instructions and MSRs (i.e., nested VMX). Nested VMX is disabled for |
2596 |
+- * all guests if the "nested" module option is off, and can also be disabled |
2597 |
+- * for a single guest by disabling its VMX cpuid bit. |
2598 |
+- */ |
2599 |
+-bool nested_vmx_allowed(struct kvm_vcpu *vcpu) |
2600 |
+-{ |
2601 |
+- return nested && guest_cpuid_has(vcpu, X86_FEATURE_VMX); |
2602 |
+-} |
2603 |
+- |
2604 |
+-static inline bool vmx_feature_control_msr_valid(struct kvm_vcpu *vcpu, |
2605 |
+- uint64_t val) |
2606 |
+-{ |
2607 |
+- uint64_t valid_bits = to_vmx(vcpu)->msr_ia32_feature_control_valid_bits; |
2608 |
+- |
2609 |
+- return !(val & ~valid_bits); |
2610 |
+-} |
2611 |
+- |
2612 |
+-static int vmx_get_msr_feature(struct kvm_msr_entry *msr) |
2613 |
+-{ |
2614 |
+- switch (msr->index) { |
2615 |
+- case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
2616 |
+- if (!nested) |
2617 |
+- return 1; |
2618 |
+- return vmx_get_vmx_msr(&vmcs_config.nested, msr->index, &msr->data); |
2619 |
+- default: |
2620 |
+- return 1; |
2621 |
+- } |
2622 |
+-} |
2623 |
+- |
2624 |
+-/* |
2625 |
+- * Reads an msr value (of 'msr_index') into 'pdata'. |
2626 |
+- * Returns 0 on success, non-0 otherwise. |
2627 |
+- * Assumes vcpu_load() was already called. |
2628 |
+- */ |
2629 |
+-static int vmx_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) |
2630 |
+-{ |
2631 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2632 |
+- struct shared_msr_entry *msr; |
2633 |
+- u32 index; |
2634 |
+- |
2635 |
+- switch (msr_info->index) { |
2636 |
+-#ifdef CONFIG_X86_64 |
2637 |
+- case MSR_FS_BASE: |
2638 |
+- msr_info->data = vmcs_readl(GUEST_FS_BASE); |
2639 |
+- break; |
2640 |
+- case MSR_GS_BASE: |
2641 |
+- msr_info->data = vmcs_readl(GUEST_GS_BASE); |
2642 |
+- break; |
2643 |
+- case MSR_KERNEL_GS_BASE: |
2644 |
+- msr_info->data = vmx_read_guest_kernel_gs_base(vmx); |
2645 |
+- break; |
2646 |
+-#endif |
2647 |
+- case MSR_EFER: |
2648 |
+- return kvm_get_msr_common(vcpu, msr_info); |
2649 |
+- case MSR_IA32_TSX_CTRL: |
2650 |
+- if (!msr_info->host_initiated && |
2651 |
+- !(vcpu->arch.arch_capabilities & ARCH_CAP_TSX_CTRL_MSR)) |
2652 |
+- return 1; |
2653 |
+- goto find_shared_msr; |
2654 |
+- case MSR_IA32_UMWAIT_CONTROL: |
2655 |
+- if (!msr_info->host_initiated && !vmx_has_waitpkg(vmx)) |
2656 |
+- return 1; |
2657 |
+- |
2658 |
+- msr_info->data = vmx->msr_ia32_umwait_control; |
2659 |
+- break; |
2660 |
+- case MSR_IA32_SPEC_CTRL: |
2661 |
+- if (!msr_info->host_initiated && |
2662 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
2663 |
+- return 1; |
2664 |
+- |
2665 |
+- msr_info->data = to_vmx(vcpu)->spec_ctrl; |
2666 |
+- break; |
2667 |
+- case MSR_IA32_SYSENTER_CS: |
2668 |
+- msr_info->data = vmcs_read32(GUEST_SYSENTER_CS); |
2669 |
+- break; |
2670 |
+- case MSR_IA32_SYSENTER_EIP: |
2671 |
+- msr_info->data = vmcs_readl(GUEST_SYSENTER_EIP); |
2672 |
+- break; |
2673 |
+- case MSR_IA32_SYSENTER_ESP: |
2674 |
+- msr_info->data = vmcs_readl(GUEST_SYSENTER_ESP); |
2675 |
+- break; |
2676 |
+- case MSR_IA32_BNDCFGS: |
2677 |
+- if (!kvm_mpx_supported() || |
2678 |
+- (!msr_info->host_initiated && |
2679 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_MPX))) |
2680 |
+- return 1; |
2681 |
+- msr_info->data = vmcs_read64(GUEST_BNDCFGS); |
2682 |
+- break; |
2683 |
+- case MSR_IA32_MCG_EXT_CTL: |
2684 |
+- if (!msr_info->host_initiated && |
2685 |
+- !(vmx->msr_ia32_feature_control & |
2686 |
+- FEATURE_CONTROL_LMCE)) |
2687 |
+- return 1; |
2688 |
+- msr_info->data = vcpu->arch.mcg_ext_ctl; |
2689 |
+- break; |
2690 |
+- case MSR_IA32_FEATURE_CONTROL: |
2691 |
+- msr_info->data = vmx->msr_ia32_feature_control; |
2692 |
+- break; |
2693 |
+- case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
2694 |
+- if (!nested_vmx_allowed(vcpu)) |
2695 |
+- return 1; |
2696 |
+- return vmx_get_vmx_msr(&vmx->nested.msrs, msr_info->index, |
2697 |
+- &msr_info->data); |
2698 |
+- case MSR_IA32_RTIT_CTL: |
2699 |
+- if (pt_mode != PT_MODE_HOST_GUEST) |
2700 |
+- return 1; |
2701 |
+- msr_info->data = vmx->pt_desc.guest.ctl; |
2702 |
+- break; |
2703 |
+- case MSR_IA32_RTIT_STATUS: |
2704 |
+- if (pt_mode != PT_MODE_HOST_GUEST) |
2705 |
+- return 1; |
2706 |
+- msr_info->data = vmx->pt_desc.guest.status; |
2707 |
+- break; |
2708 |
+- case MSR_IA32_RTIT_CR3_MATCH: |
2709 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2710 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2711 |
+- PT_CAP_cr3_filtering)) |
2712 |
+- return 1; |
2713 |
+- msr_info->data = vmx->pt_desc.guest.cr3_match; |
2714 |
+- break; |
2715 |
+- case MSR_IA32_RTIT_OUTPUT_BASE: |
2716 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2717 |
+- (!intel_pt_validate_cap(vmx->pt_desc.caps, |
2718 |
+- PT_CAP_topa_output) && |
2719 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2720 |
+- PT_CAP_single_range_output))) |
2721 |
+- return 1; |
2722 |
+- msr_info->data = vmx->pt_desc.guest.output_base; |
2723 |
+- break; |
2724 |
+- case MSR_IA32_RTIT_OUTPUT_MASK: |
2725 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2726 |
+- (!intel_pt_validate_cap(vmx->pt_desc.caps, |
2727 |
+- PT_CAP_topa_output) && |
2728 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2729 |
+- PT_CAP_single_range_output))) |
2730 |
+- return 1; |
2731 |
+- msr_info->data = vmx->pt_desc.guest.output_mask; |
2732 |
+- break; |
2733 |
+- case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B: |
2734 |
+- index = msr_info->index - MSR_IA32_RTIT_ADDR0_A; |
2735 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2736 |
+- (index >= 2 * intel_pt_validate_cap(vmx->pt_desc.caps, |
2737 |
+- PT_CAP_num_address_ranges))) |
2738 |
+- return 1; |
2739 |
+- if (is_noncanonical_address(data, vcpu)) |
2740 |
+- return 1; |
2741 |
+- if (index % 2) |
2742 |
+- msr_info->data = vmx->pt_desc.guest.addr_b[index / 2]; |
2743 |
+- else |
2744 |
+- msr_info->data = vmx->pt_desc.guest.addr_a[index / 2]; |
2745 |
+- break; |
2746 |
+- case MSR_TSC_AUX: |
2747 |
+- if (!msr_info->host_initiated && |
2748 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP)) |
2749 |
+- return 1; |
2750 |
+- goto find_shared_msr; |
2751 |
+- default: |
2752 |
+- find_shared_msr: |
2753 |
+- msr = find_msr_entry(vmx, msr_info->index); |
2754 |
+- if (msr) { |
2755 |
+- msr_info->data = msr->data; |
2756 |
+- break; |
2757 |
+- } |
2758 |
+- return kvm_get_msr_common(vcpu, msr_info); |
2759 |
+- } |
2760 |
+- |
2761 |
+- return 0; |
2762 |
+-} |
2763 |
+- |
2764 |
+-/* |
2765 |
+- * Writes msr value into the appropriate "register". |
2766 |
+- * Returns 0 on success, non-0 otherwise. |
2767 |
+- * Assumes vcpu_load() was already called. |
2768 |
+- */ |
2769 |
+-static int vmx_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info) |
2770 |
+-{ |
2771 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
2772 |
+- struct shared_msr_entry *msr; |
2773 |
+- int ret = 0; |
2774 |
+- u32 msr_index = msr_info->index; |
2775 |
+- u64 data = msr_info->data; |
2776 |
+- u32 index; |
2777 |
+- |
2778 |
+- switch (msr_index) { |
2779 |
+- case MSR_EFER: |
2780 |
+- ret = kvm_set_msr_common(vcpu, msr_info); |
2781 |
+- break; |
2782 |
+-#ifdef CONFIG_X86_64 |
2783 |
+- case MSR_FS_BASE: |
2784 |
+- vmx_segment_cache_clear(vmx); |
2785 |
+- vmcs_writel(GUEST_FS_BASE, data); |
2786 |
+- break; |
2787 |
+- case MSR_GS_BASE: |
2788 |
+- vmx_segment_cache_clear(vmx); |
2789 |
+- vmcs_writel(GUEST_GS_BASE, data); |
2790 |
+- break; |
2791 |
+- case MSR_KERNEL_GS_BASE: |
2792 |
+- vmx_write_guest_kernel_gs_base(vmx, data); |
2793 |
+- break; |
2794 |
+-#endif |
2795 |
+- case MSR_IA32_SYSENTER_CS: |
2796 |
+- if (is_guest_mode(vcpu)) |
2797 |
+- get_vmcs12(vcpu)->guest_sysenter_cs = data; |
2798 |
+- vmcs_write32(GUEST_SYSENTER_CS, data); |
2799 |
+- break; |
2800 |
+- case MSR_IA32_SYSENTER_EIP: |
2801 |
+- if (is_guest_mode(vcpu)) |
2802 |
+- get_vmcs12(vcpu)->guest_sysenter_eip = data; |
2803 |
+- vmcs_writel(GUEST_SYSENTER_EIP, data); |
2804 |
+- break; |
2805 |
+- case MSR_IA32_SYSENTER_ESP: |
2806 |
+- if (is_guest_mode(vcpu)) |
2807 |
+- get_vmcs12(vcpu)->guest_sysenter_esp = data; |
2808 |
+- vmcs_writel(GUEST_SYSENTER_ESP, data); |
2809 |
+- break; |
2810 |
+- case MSR_IA32_DEBUGCTLMSR: |
2811 |
+- if (is_guest_mode(vcpu) && get_vmcs12(vcpu)->vm_exit_controls & |
2812 |
+- VM_EXIT_SAVE_DEBUG_CONTROLS) |
2813 |
+- get_vmcs12(vcpu)->guest_ia32_debugctl = data; |
2814 |
+- |
2815 |
+- ret = kvm_set_msr_common(vcpu, msr_info); |
2816 |
+- break; |
2817 |
+- |
2818 |
+- case MSR_IA32_BNDCFGS: |
2819 |
+- if (!kvm_mpx_supported() || |
2820 |
+- (!msr_info->host_initiated && |
2821 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_MPX))) |
2822 |
+- return 1; |
2823 |
+- if (is_noncanonical_address(data & PAGE_MASK, vcpu) || |
2824 |
+- (data & MSR_IA32_BNDCFGS_RSVD)) |
2825 |
+- return 1; |
2826 |
+- vmcs_write64(GUEST_BNDCFGS, data); |
2827 |
+- break; |
2828 |
+- case MSR_IA32_UMWAIT_CONTROL: |
2829 |
+- if (!msr_info->host_initiated && !vmx_has_waitpkg(vmx)) |
2830 |
+- return 1; |
2831 |
+- |
2832 |
+- /* The reserved bit 1 and non-32 bit [63:32] should be zero */ |
2833 |
+- if (data & (BIT_ULL(1) | GENMASK_ULL(63, 32))) |
2834 |
+- return 1; |
2835 |
+- |
2836 |
+- vmx->msr_ia32_umwait_control = data; |
2837 |
+- break; |
2838 |
+- case MSR_IA32_SPEC_CTRL: |
2839 |
+- if (!msr_info->host_initiated && |
2840 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
2841 |
+- return 1; |
2842 |
+- |
2843 |
+- /* The STIBP bit doesn't fault even if it's not advertised */ |
2844 |
+- if (data & ~(SPEC_CTRL_IBRS | SPEC_CTRL_STIBP | SPEC_CTRL_SSBD)) |
2845 |
+- return 1; |
2846 |
+- |
2847 |
+- vmx->spec_ctrl = data; |
2848 |
+- |
2849 |
+- if (!data) |
2850 |
+- break; |
2851 |
+- |
2852 |
+- /* |
2853 |
+- * For non-nested: |
2854 |
+- * When it's written (to non-zero) for the first time, pass |
2855 |
+- * it through. |
2856 |
+- * |
2857 |
+- * For nested: |
2858 |
+- * The handling of the MSR bitmap for L2 guests is done in |
2859 |
+- * nested_vmx_prepare_msr_bitmap. We should not touch the |
2860 |
+- * vmcs02.msr_bitmap here since it gets completely overwritten |
2861 |
+- * in the merging. We update the vmcs01 here for L1 as well |
2862 |
+- * since it will end up touching the MSR anyway now. |
2863 |
+- */ |
2864 |
+- vmx_disable_intercept_for_msr(vmx->vmcs01.msr_bitmap, |
2865 |
+- MSR_IA32_SPEC_CTRL, |
2866 |
+- MSR_TYPE_RW); |
2867 |
+- break; |
2868 |
+- case MSR_IA32_TSX_CTRL: |
2869 |
+- if (!msr_info->host_initiated && |
2870 |
+- !(vcpu->arch.arch_capabilities & ARCH_CAP_TSX_CTRL_MSR)) |
2871 |
+- return 1; |
2872 |
+- if (data & ~(TSX_CTRL_RTM_DISABLE | TSX_CTRL_CPUID_CLEAR)) |
2873 |
+- return 1; |
2874 |
+- goto find_shared_msr; |
2875 |
+- case MSR_IA32_PRED_CMD: |
2876 |
+- if (!msr_info->host_initiated && |
2877 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_SPEC_CTRL)) |
2878 |
+- return 1; |
2879 |
+- |
2880 |
+- if (data & ~PRED_CMD_IBPB) |
2881 |
+- return 1; |
2882 |
+- |
2883 |
+- if (!data) |
2884 |
+- break; |
2885 |
+- |
2886 |
+- wrmsrl(MSR_IA32_PRED_CMD, PRED_CMD_IBPB); |
2887 |
+- |
2888 |
+- /* |
2889 |
+- * For non-nested: |
2890 |
+- * When it's written (to non-zero) for the first time, pass |
2891 |
+- * it through. |
2892 |
+- * |
2893 |
+- * For nested: |
2894 |
+- * The handling of the MSR bitmap for L2 guests is done in |
2895 |
+- * nested_vmx_prepare_msr_bitmap. We should not touch the |
2896 |
+- * vmcs02.msr_bitmap here since it gets completely overwritten |
2897 |
+- * in the merging. |
2898 |
+- */ |
2899 |
+- vmx_disable_intercept_for_msr(vmx->vmcs01.msr_bitmap, MSR_IA32_PRED_CMD, |
2900 |
+- MSR_TYPE_W); |
2901 |
+- break; |
2902 |
+- case MSR_IA32_CR_PAT: |
2903 |
+- if (!kvm_pat_valid(data)) |
2904 |
+- return 1; |
2905 |
+- |
2906 |
+- if (is_guest_mode(vcpu) && |
2907 |
+- get_vmcs12(vcpu)->vm_exit_controls & VM_EXIT_SAVE_IA32_PAT) |
2908 |
+- get_vmcs12(vcpu)->guest_ia32_pat = data; |
2909 |
+- |
2910 |
+- if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { |
2911 |
+- vmcs_write64(GUEST_IA32_PAT, data); |
2912 |
+- vcpu->arch.pat = data; |
2913 |
+- break; |
2914 |
+- } |
2915 |
+- ret = kvm_set_msr_common(vcpu, msr_info); |
2916 |
+- break; |
2917 |
+- case MSR_IA32_TSC_ADJUST: |
2918 |
+- ret = kvm_set_msr_common(vcpu, msr_info); |
2919 |
+- break; |
2920 |
+- case MSR_IA32_MCG_EXT_CTL: |
2921 |
+- if ((!msr_info->host_initiated && |
2922 |
+- !(to_vmx(vcpu)->msr_ia32_feature_control & |
2923 |
+- FEATURE_CONTROL_LMCE)) || |
2924 |
+- (data & ~MCG_EXT_CTL_LMCE_EN)) |
2925 |
+- return 1; |
2926 |
+- vcpu->arch.mcg_ext_ctl = data; |
2927 |
+- break; |
2928 |
+- case MSR_IA32_FEATURE_CONTROL: |
2929 |
+- if (!vmx_feature_control_msr_valid(vcpu, data) || |
2930 |
+- (to_vmx(vcpu)->msr_ia32_feature_control & |
2931 |
+- FEATURE_CONTROL_LOCKED && !msr_info->host_initiated)) |
2932 |
+- return 1; |
2933 |
+- vmx->msr_ia32_feature_control = data; |
2934 |
+- if (msr_info->host_initiated && data == 0) |
2935 |
+- vmx_leave_nested(vcpu); |
2936 |
+- break; |
2937 |
+- case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
2938 |
+- if (!msr_info->host_initiated) |
2939 |
+- return 1; /* they are read-only */ |
2940 |
+- if (!nested_vmx_allowed(vcpu)) |
2941 |
+- return 1; |
2942 |
+- return vmx_set_vmx_msr(vcpu, msr_index, data); |
2943 |
+- case MSR_IA32_RTIT_CTL: |
2944 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2945 |
+- vmx_rtit_ctl_check(vcpu, data) || |
2946 |
+- vmx->nested.vmxon) |
2947 |
+- return 1; |
2948 |
+- vmcs_write64(GUEST_IA32_RTIT_CTL, data); |
2949 |
+- vmx->pt_desc.guest.ctl = data; |
2950 |
+- pt_update_intercept_for_msr(vmx); |
2951 |
+- break; |
2952 |
+- case MSR_IA32_RTIT_STATUS: |
2953 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2954 |
+- (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
2955 |
+- (data & MSR_IA32_RTIT_STATUS_MASK)) |
2956 |
+- return 1; |
2957 |
+- vmx->pt_desc.guest.status = data; |
2958 |
+- break; |
2959 |
+- case MSR_IA32_RTIT_CR3_MATCH: |
2960 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2961 |
+- (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
2962 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2963 |
+- PT_CAP_cr3_filtering)) |
2964 |
+- return 1; |
2965 |
+- vmx->pt_desc.guest.cr3_match = data; |
2966 |
+- break; |
2967 |
+- case MSR_IA32_RTIT_OUTPUT_BASE: |
2968 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2969 |
+- (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
2970 |
+- (!intel_pt_validate_cap(vmx->pt_desc.caps, |
2971 |
+- PT_CAP_topa_output) && |
2972 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2973 |
+- PT_CAP_single_range_output)) || |
2974 |
+- (data & MSR_IA32_RTIT_OUTPUT_BASE_MASK)) |
2975 |
+- return 1; |
2976 |
+- vmx->pt_desc.guest.output_base = data; |
2977 |
+- break; |
2978 |
+- case MSR_IA32_RTIT_OUTPUT_MASK: |
2979 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2980 |
+- (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
2981 |
+- (!intel_pt_validate_cap(vmx->pt_desc.caps, |
2982 |
+- PT_CAP_topa_output) && |
2983 |
+- !intel_pt_validate_cap(vmx->pt_desc.caps, |
2984 |
+- PT_CAP_single_range_output))) |
2985 |
+- return 1; |
2986 |
+- vmx->pt_desc.guest.output_mask = data; |
2987 |
+- break; |
2988 |
+- case MSR_IA32_RTIT_ADDR0_A ... MSR_IA32_RTIT_ADDR3_B: |
2989 |
+- index = msr_info->index - MSR_IA32_RTIT_ADDR0_A; |
2990 |
+- if ((pt_mode != PT_MODE_HOST_GUEST) || |
2991 |
+- (vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN) || |
2992 |
+- (index >= 2 * intel_pt_validate_cap(vmx->pt_desc.caps, |
2993 |
+- PT_CAP_num_address_ranges))) |
2994 |
+- return 1; |
2995 |
+- if (is_noncanonical_address(data, vcpu)) |
2996 |
+- return 1; |
2997 |
+- if (index % 2) |
2998 |
+- vmx->pt_desc.guest.addr_b[index / 2] = data; |
2999 |
+- else |
3000 |
+- vmx->pt_desc.guest.addr_a[index / 2] = data; |
3001 |
+- break; |
3002 |
+- case MSR_TSC_AUX: |
3003 |
+- if (!msr_info->host_initiated && |
3004 |
+- !guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP)) |
3005 |
+- return 1; |
3006 |
+- /* Check reserved bit, higher 32 bits should be zero */ |
3007 |
+- if ((data >> 32) != 0) |
3008 |
+- return 1; |
3009 |
+- goto find_shared_msr; |
3010 |
+- |
3011 |
+- default: |
3012 |
+- find_shared_msr: |
3013 |
+- msr = find_msr_entry(vmx, msr_index); |
3014 |
+- if (msr) |
3015 |
+- ret = vmx_set_guest_msr(vmx, msr, data); |
3016 |
+- else |
3017 |
+- ret = kvm_set_msr_common(vcpu, msr_info); |
3018 |
+- } |
3019 |
+- |
3020 |
+- return ret; |
3021 |
+-} |
3022 |
+- |
3023 |
+-static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg) |
3024 |
+-{ |
3025 |
+- kvm_register_mark_available(vcpu, reg); |
3026 |
+- |
3027 |
+- switch (reg) { |
3028 |
+- case VCPU_REGS_RSP: |
3029 |
+- vcpu->arch.regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP); |
3030 |
+- break; |
3031 |
+- case VCPU_REGS_RIP: |
3032 |
+- vcpu->arch.regs[VCPU_REGS_RIP] = vmcs_readl(GUEST_RIP); |
3033 |
+- break; |
3034 |
+- case VCPU_EXREG_PDPTR: |
3035 |
+- if (enable_ept) |
3036 |
+- ept_save_pdptrs(vcpu); |
3037 |
+- break; |
3038 |
+- case VCPU_EXREG_CR3: |
3039 |
+- if (enable_unrestricted_guest || (enable_ept && is_paging(vcpu))) |
3040 |
+- vcpu->arch.cr3 = vmcs_readl(GUEST_CR3); |
3041 |
+- break; |
3042 |
+- default: |
3043 |
+- WARN_ON_ONCE(1); |
3044 |
+- break; |
3045 |
+- } |
3046 |
+-} |
3047 |
+- |
3048 |
+-static __init int cpu_has_kvm_support(void) |
3049 |
+-{ |
3050 |
+- return cpu_has_vmx(); |
3051 |
+-} |
3052 |
+- |
3053 |
+-static __init int vmx_disabled_by_bios(void) |
3054 |
+-{ |
3055 |
+- u64 msr; |
3056 |
+- |
3057 |
+- rdmsrl(MSR_IA32_FEATURE_CONTROL, msr); |
3058 |
+- if (msr & FEATURE_CONTROL_LOCKED) { |
3059 |
+- /* launched w/ TXT and VMX disabled */ |
3060 |
+- if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
3061 |
+- && tboot_enabled()) |
3062 |
+- return 1; |
3063 |
+- /* launched w/o TXT and VMX only enabled w/ TXT */ |
3064 |
+- if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) |
3065 |
+- && (msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
3066 |
+- && !tboot_enabled()) { |
3067 |
+- printk(KERN_WARNING "kvm: disable TXT in the BIOS or " |
3068 |
+- "activate TXT before enabling KVM\n"); |
3069 |
+- return 1; |
3070 |
+- } |
3071 |
+- /* launched w/o TXT and VMX disabled */ |
3072 |
+- if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) |
3073 |
+- && !tboot_enabled()) |
3074 |
+- return 1; |
3075 |
+- } |
3076 |
+- |
3077 |
+- return 0; |
3078 |
+-} |
3079 |
+- |
3080 |
+-static void kvm_cpu_vmxon(u64 addr) |
3081 |
+-{ |
3082 |
+- cr4_set_bits(X86_CR4_VMXE); |
3083 |
+- intel_pt_handle_vmx(1); |
3084 |
+- |
3085 |
+- asm volatile ("vmxon %0" : : "m"(addr)); |
3086 |
+-} |
3087 |
+- |
3088 |
+-static int hardware_enable(void) |
3089 |
+-{ |
3090 |
+- int cpu = raw_smp_processor_id(); |
3091 |
+- u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); |
3092 |
+- u64 old, test_bits; |
3093 |
+- |
3094 |
+- if (cr4_read_shadow() & X86_CR4_VMXE) |
3095 |
+- return -EBUSY; |
3096 |
+- |
3097 |
+- /* |
3098 |
+- * This can happen if we hot-added a CPU but failed to allocate |
3099 |
+- * VP assist page for it. |
3100 |
+- */ |
3101 |
+- if (static_branch_unlikely(&enable_evmcs) && |
3102 |
+- !hv_get_vp_assist_page(cpu)) |
3103 |
+- return -EFAULT; |
3104 |
+- |
3105 |
+- INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu)); |
3106 |
+- INIT_LIST_HEAD(&per_cpu(blocked_vcpu_on_cpu, cpu)); |
3107 |
+- spin_lock_init(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
3108 |
+- |
3109 |
+- /* |
3110 |
+- * Now we can enable the vmclear operation in kdump |
3111 |
+- * since the loaded_vmcss_on_cpu list on this cpu |
3112 |
+- * has been initialized. |
3113 |
+- * |
3114 |
+- * Though the cpu is not in VMX operation now, there |
3115 |
+- * is no problem to enable the vmclear operation |
3116 |
+- * for the loaded_vmcss_on_cpu list is empty! |
3117 |
+- */ |
3118 |
+- crash_enable_local_vmclear(cpu); |
3119 |
+- |
3120 |
+- rdmsrl(MSR_IA32_FEATURE_CONTROL, old); |
3121 |
+- |
3122 |
+- test_bits = FEATURE_CONTROL_LOCKED; |
3123 |
+- test_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; |
3124 |
+- if (tboot_enabled()) |
3125 |
+- test_bits |= FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX; |
3126 |
+- |
3127 |
+- if ((old & test_bits) != test_bits) { |
3128 |
+- /* enable and lock */ |
3129 |
+- wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits); |
3130 |
+- } |
3131 |
+- kvm_cpu_vmxon(phys_addr); |
3132 |
+- if (enable_ept) |
3133 |
+- ept_sync_global(); |
3134 |
+- |
3135 |
+- return 0; |
3136 |
+-} |
3137 |
+- |
3138 |
+-static void vmclear_local_loaded_vmcss(void) |
3139 |
+-{ |
3140 |
+- int cpu = raw_smp_processor_id(); |
3141 |
+- struct loaded_vmcs *v, *n; |
3142 |
+- |
3143 |
+- list_for_each_entry_safe(v, n, &per_cpu(loaded_vmcss_on_cpu, cpu), |
3144 |
+- loaded_vmcss_on_cpu_link) |
3145 |
+- __loaded_vmcs_clear(v); |
3146 |
+-} |
3147 |
+- |
3148 |
+- |
3149 |
+-/* Just like cpu_vmxoff(), but with the __kvm_handle_fault_on_reboot() |
3150 |
+- * tricks. |
3151 |
+- */ |
3152 |
+-static void kvm_cpu_vmxoff(void) |
3153 |
+-{ |
3154 |
+- asm volatile (__ex("vmxoff")); |
3155 |
+- |
3156 |
+- intel_pt_handle_vmx(0); |
3157 |
+- cr4_clear_bits(X86_CR4_VMXE); |
3158 |
+-} |
3159 |
+- |
3160 |
+-static void hardware_disable(void) |
3161 |
+-{ |
3162 |
+- vmclear_local_loaded_vmcss(); |
3163 |
+- kvm_cpu_vmxoff(); |
3164 |
+-} |
3165 |
+- |
3166 |
+-static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt, |
3167 |
+- u32 msr, u32 *result) |
3168 |
+-{ |
3169 |
+- u32 vmx_msr_low, vmx_msr_high; |
3170 |
+- u32 ctl = ctl_min | ctl_opt; |
3171 |
+- |
3172 |
+- rdmsr(msr, vmx_msr_low, vmx_msr_high); |
3173 |
+- |
3174 |
+- ctl &= vmx_msr_high; /* bit == 0 in high word ==> must be zero */ |
3175 |
+- ctl |= vmx_msr_low; /* bit == 1 in low word ==> must be one */ |
3176 |
+- |
3177 |
+- /* Ensure minimum (required) set of control bits are supported. */ |
3178 |
+- if (ctl_min & ~ctl) |
3179 |
+- return -EIO; |
3180 |
+- |
3181 |
+- *result = ctl; |
3182 |
+- return 0; |
3183 |
+-} |
3184 |
+- |
3185 |
+-static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf, |
3186 |
+- struct vmx_capability *vmx_cap) |
3187 |
+-{ |
3188 |
+- u32 vmx_msr_low, vmx_msr_high; |
3189 |
+- u32 min, opt, min2, opt2; |
3190 |
+- u32 _pin_based_exec_control = 0; |
3191 |
+- u32 _cpu_based_exec_control = 0; |
3192 |
+- u32 _cpu_based_2nd_exec_control = 0; |
3193 |
+- u32 _vmexit_control = 0; |
3194 |
+- u32 _vmentry_control = 0; |
3195 |
+- |
3196 |
+- memset(vmcs_conf, 0, sizeof(*vmcs_conf)); |
3197 |
+- min = CPU_BASED_HLT_EXITING | |
3198 |
+-#ifdef CONFIG_X86_64 |
3199 |
+- CPU_BASED_CR8_LOAD_EXITING | |
3200 |
+- CPU_BASED_CR8_STORE_EXITING | |
3201 |
+-#endif |
3202 |
+- CPU_BASED_CR3_LOAD_EXITING | |
3203 |
+- CPU_BASED_CR3_STORE_EXITING | |
3204 |
+- CPU_BASED_UNCOND_IO_EXITING | |
3205 |
+- CPU_BASED_MOV_DR_EXITING | |
3206 |
+- CPU_BASED_USE_TSC_OFFSETTING | |
3207 |
+- CPU_BASED_MWAIT_EXITING | |
3208 |
+- CPU_BASED_MONITOR_EXITING | |
3209 |
+- CPU_BASED_INVLPG_EXITING | |
3210 |
+- CPU_BASED_RDPMC_EXITING; |
3211 |
+- |
3212 |
+- opt = CPU_BASED_TPR_SHADOW | |
3213 |
+- CPU_BASED_USE_MSR_BITMAPS | |
3214 |
+- CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; |
3215 |
+- if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PROCBASED_CTLS, |
3216 |
+- &_cpu_based_exec_control) < 0) |
3217 |
+- return -EIO; |
3218 |
+-#ifdef CONFIG_X86_64 |
3219 |
+- if ((_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) |
3220 |
+- _cpu_based_exec_control &= ~CPU_BASED_CR8_LOAD_EXITING & |
3221 |
+- ~CPU_BASED_CR8_STORE_EXITING; |
3222 |
+-#endif |
3223 |
+- if (_cpu_based_exec_control & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) { |
3224 |
+- min2 = 0; |
3225 |
+- opt2 = SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
3226 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
3227 |
+- SECONDARY_EXEC_WBINVD_EXITING | |
3228 |
+- SECONDARY_EXEC_ENABLE_VPID | |
3229 |
+- SECONDARY_EXEC_ENABLE_EPT | |
3230 |
+- SECONDARY_EXEC_UNRESTRICTED_GUEST | |
3231 |
+- SECONDARY_EXEC_PAUSE_LOOP_EXITING | |
3232 |
+- SECONDARY_EXEC_DESC | |
3233 |
+- SECONDARY_EXEC_RDTSCP | |
3234 |
+- SECONDARY_EXEC_ENABLE_INVPCID | |
3235 |
+- SECONDARY_EXEC_APIC_REGISTER_VIRT | |
3236 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY | |
3237 |
+- SECONDARY_EXEC_SHADOW_VMCS | |
3238 |
+- SECONDARY_EXEC_XSAVES | |
3239 |
+- SECONDARY_EXEC_RDSEED_EXITING | |
3240 |
+- SECONDARY_EXEC_RDRAND_EXITING | |
3241 |
+- SECONDARY_EXEC_ENABLE_PML | |
3242 |
+- SECONDARY_EXEC_TSC_SCALING | |
3243 |
+- SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE | |
3244 |
+- SECONDARY_EXEC_PT_USE_GPA | |
3245 |
+- SECONDARY_EXEC_PT_CONCEAL_VMX | |
3246 |
+- SECONDARY_EXEC_ENABLE_VMFUNC | |
3247 |
+- SECONDARY_EXEC_ENCLS_EXITING; |
3248 |
+- if (adjust_vmx_controls(min2, opt2, |
3249 |
+- MSR_IA32_VMX_PROCBASED_CTLS2, |
3250 |
+- &_cpu_based_2nd_exec_control) < 0) |
3251 |
+- return -EIO; |
3252 |
+- } |
3253 |
+-#ifndef CONFIG_X86_64 |
3254 |
+- if (!(_cpu_based_2nd_exec_control & |
3255 |
+- SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) |
3256 |
+- _cpu_based_exec_control &= ~CPU_BASED_TPR_SHADOW; |
3257 |
+-#endif |
3258 |
+- |
3259 |
+- if (!(_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) |
3260 |
+- _cpu_based_2nd_exec_control &= ~( |
3261 |
+- SECONDARY_EXEC_APIC_REGISTER_VIRT | |
3262 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
3263 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
3264 |
+- |
3265 |
+- rdmsr_safe(MSR_IA32_VMX_EPT_VPID_CAP, |
3266 |
+- &vmx_cap->ept, &vmx_cap->vpid); |
3267 |
+- |
3268 |
+- if (_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_EPT) { |
3269 |
+- /* CR3 accesses and invlpg don't need to cause VM Exits when EPT |
3270 |
+- enabled */ |
3271 |
+- _cpu_based_exec_control &= ~(CPU_BASED_CR3_LOAD_EXITING | |
3272 |
+- CPU_BASED_CR3_STORE_EXITING | |
3273 |
+- CPU_BASED_INVLPG_EXITING); |
3274 |
+- } else if (vmx_cap->ept) { |
3275 |
+- vmx_cap->ept = 0; |
3276 |
+- pr_warn_once("EPT CAP should not exist if not support " |
3277 |
+- "1-setting enable EPT VM-execution control\n"); |
3278 |
+- } |
3279 |
+- if (!(_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_VPID) && |
3280 |
+- vmx_cap->vpid) { |
3281 |
+- vmx_cap->vpid = 0; |
3282 |
+- pr_warn_once("VPID CAP should not exist if not support " |
3283 |
+- "1-setting enable VPID VM-execution control\n"); |
3284 |
+- } |
3285 |
+- |
3286 |
+- min = VM_EXIT_SAVE_DEBUG_CONTROLS | VM_EXIT_ACK_INTR_ON_EXIT; |
3287 |
+-#ifdef CONFIG_X86_64 |
3288 |
+- min |= VM_EXIT_HOST_ADDR_SPACE_SIZE; |
3289 |
+-#endif |
3290 |
+- opt = VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL | |
3291 |
+- VM_EXIT_LOAD_IA32_PAT | |
3292 |
+- VM_EXIT_LOAD_IA32_EFER | |
3293 |
+- VM_EXIT_CLEAR_BNDCFGS | |
3294 |
+- VM_EXIT_PT_CONCEAL_PIP | |
3295 |
+- VM_EXIT_CLEAR_IA32_RTIT_CTL; |
3296 |
+- if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_EXIT_CTLS, |
3297 |
+- &_vmexit_control) < 0) |
3298 |
+- return -EIO; |
3299 |
+- |
3300 |
+- min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING; |
3301 |
+- opt = PIN_BASED_VIRTUAL_NMIS | PIN_BASED_POSTED_INTR | |
3302 |
+- PIN_BASED_VMX_PREEMPTION_TIMER; |
3303 |
+- if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PINBASED_CTLS, |
3304 |
+- &_pin_based_exec_control) < 0) |
3305 |
+- return -EIO; |
3306 |
+- |
3307 |
+- if (cpu_has_broken_vmx_preemption_timer()) |
3308 |
+- _pin_based_exec_control &= ~PIN_BASED_VMX_PREEMPTION_TIMER; |
3309 |
+- if (!(_cpu_based_2nd_exec_control & |
3310 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY)) |
3311 |
+- _pin_based_exec_control &= ~PIN_BASED_POSTED_INTR; |
3312 |
+- |
3313 |
+- min = VM_ENTRY_LOAD_DEBUG_CONTROLS; |
3314 |
+- opt = VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL | |
3315 |
+- VM_ENTRY_LOAD_IA32_PAT | |
3316 |
+- VM_ENTRY_LOAD_IA32_EFER | |
3317 |
+- VM_ENTRY_LOAD_BNDCFGS | |
3318 |
+- VM_ENTRY_PT_CONCEAL_PIP | |
3319 |
+- VM_ENTRY_LOAD_IA32_RTIT_CTL; |
3320 |
+- if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_ENTRY_CTLS, |
3321 |
+- &_vmentry_control) < 0) |
3322 |
+- return -EIO; |
3323 |
+- |
3324 |
+- /* |
3325 |
+- * Some cpus support VM_{ENTRY,EXIT}_IA32_PERF_GLOBAL_CTRL but they |
3326 |
+- * can't be used due to an errata where VM Exit may incorrectly clear |
3327 |
+- * IA32_PERF_GLOBAL_CTRL[34:32]. Workaround the errata by using the |
3328 |
+- * MSR load mechanism to switch IA32_PERF_GLOBAL_CTRL. |
3329 |
+- */ |
3330 |
+- if (boot_cpu_data.x86 == 0x6) { |
3331 |
+- switch (boot_cpu_data.x86_model) { |
3332 |
+- case 26: /* AAK155 */ |
3333 |
+- case 30: /* AAP115 */ |
3334 |
+- case 37: /* AAT100 */ |
3335 |
+- case 44: /* BC86,AAY89,BD102 */ |
3336 |
+- case 46: /* BA97 */ |
3337 |
+- _vmentry_control &= ~VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL; |
3338 |
+- _vmexit_control &= ~VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL; |
3339 |
+- pr_warn_once("kvm: VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL " |
3340 |
+- "does not work properly. Using workaround\n"); |
3341 |
+- break; |
3342 |
+- default: |
3343 |
+- break; |
3344 |
+- } |
3345 |
+- } |
3346 |
+- |
3347 |
+- |
3348 |
+- rdmsr(MSR_IA32_VMX_BASIC, vmx_msr_low, vmx_msr_high); |
3349 |
+- |
3350 |
+- /* IA-32 SDM Vol 3B: VMCS size is never greater than 4kB. */ |
3351 |
+- if ((vmx_msr_high & 0x1fff) > PAGE_SIZE) |
3352 |
+- return -EIO; |
3353 |
+- |
3354 |
+-#ifdef CONFIG_X86_64 |
3355 |
+- /* IA-32 SDM Vol 3B: 64-bit CPUs always have VMX_BASIC_MSR[48]==0. */ |
3356 |
+- if (vmx_msr_high & (1u<<16)) |
3357 |
+- return -EIO; |
3358 |
+-#endif |
3359 |
+- |
3360 |
+- /* Require Write-Back (WB) memory type for VMCS accesses. */ |
3361 |
+- if (((vmx_msr_high >> 18) & 15) != 6) |
3362 |
+- return -EIO; |
3363 |
+- |
3364 |
+- vmcs_conf->size = vmx_msr_high & 0x1fff; |
3365 |
+- vmcs_conf->order = get_order(vmcs_conf->size); |
3366 |
+- vmcs_conf->basic_cap = vmx_msr_high & ~0x1fff; |
3367 |
+- |
3368 |
+- vmcs_conf->revision_id = vmx_msr_low; |
3369 |
+- |
3370 |
+- vmcs_conf->pin_based_exec_ctrl = _pin_based_exec_control; |
3371 |
+- vmcs_conf->cpu_based_exec_ctrl = _cpu_based_exec_control; |
3372 |
+- vmcs_conf->cpu_based_2nd_exec_ctrl = _cpu_based_2nd_exec_control; |
3373 |
+- vmcs_conf->vmexit_ctrl = _vmexit_control; |
3374 |
+- vmcs_conf->vmentry_ctrl = _vmentry_control; |
3375 |
+- |
3376 |
+- if (static_branch_unlikely(&enable_evmcs)) |
3377 |
+- evmcs_sanitize_exec_ctrls(vmcs_conf); |
3378 |
+- |
3379 |
+- return 0; |
3380 |
+-} |
3381 |
+- |
3382 |
+-struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags) |
3383 |
+-{ |
3384 |
+- int node = cpu_to_node(cpu); |
3385 |
+- struct page *pages; |
3386 |
+- struct vmcs *vmcs; |
3387 |
+- |
3388 |
+- pages = __alloc_pages_node(node, flags, vmcs_config.order); |
3389 |
+- if (!pages) |
3390 |
+- return NULL; |
3391 |
+- vmcs = page_address(pages); |
3392 |
+- memset(vmcs, 0, vmcs_config.size); |
3393 |
+- |
3394 |
+- /* KVM supports Enlightened VMCS v1 only */ |
3395 |
+- if (static_branch_unlikely(&enable_evmcs)) |
3396 |
+- vmcs->hdr.revision_id = KVM_EVMCS_VERSION; |
3397 |
+- else |
3398 |
+- vmcs->hdr.revision_id = vmcs_config.revision_id; |
3399 |
+- |
3400 |
+- if (shadow) |
3401 |
+- vmcs->hdr.shadow_vmcs = 1; |
3402 |
+- return vmcs; |
3403 |
+-} |
3404 |
+- |
3405 |
+-void free_vmcs(struct vmcs *vmcs) |
3406 |
+-{ |
3407 |
+- free_pages((unsigned long)vmcs, vmcs_config.order); |
3408 |
+-} |
3409 |
+- |
3410 |
+-/* |
3411 |
+- * Free a VMCS, but before that VMCLEAR it on the CPU where it was last loaded |
3412 |
+- */ |
3413 |
+-void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs) |
3414 |
+-{ |
3415 |
+- if (!loaded_vmcs->vmcs) |
3416 |
+- return; |
3417 |
+- loaded_vmcs_clear(loaded_vmcs); |
3418 |
+- free_vmcs(loaded_vmcs->vmcs); |
3419 |
+- loaded_vmcs->vmcs = NULL; |
3420 |
+- if (loaded_vmcs->msr_bitmap) |
3421 |
+- free_page((unsigned long)loaded_vmcs->msr_bitmap); |
3422 |
+- WARN_ON(loaded_vmcs->shadow_vmcs != NULL); |
3423 |
+-} |
3424 |
+- |
3425 |
+-int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs) |
3426 |
+-{ |
3427 |
+- loaded_vmcs->vmcs = alloc_vmcs(false); |
3428 |
+- if (!loaded_vmcs->vmcs) |
3429 |
+- return -ENOMEM; |
3430 |
+- |
3431 |
+- loaded_vmcs->shadow_vmcs = NULL; |
3432 |
+- loaded_vmcs->hv_timer_soft_disabled = false; |
3433 |
+- loaded_vmcs_init(loaded_vmcs); |
3434 |
+- |
3435 |
+- if (cpu_has_vmx_msr_bitmap()) { |
3436 |
+- loaded_vmcs->msr_bitmap = (unsigned long *) |
3437 |
+- __get_free_page(GFP_KERNEL_ACCOUNT); |
3438 |
+- if (!loaded_vmcs->msr_bitmap) |
3439 |
+- goto out_vmcs; |
3440 |
+- memset(loaded_vmcs->msr_bitmap, 0xff, PAGE_SIZE); |
3441 |
+- |
3442 |
+- if (IS_ENABLED(CONFIG_HYPERV) && |
3443 |
+- static_branch_unlikely(&enable_evmcs) && |
3444 |
+- (ms_hyperv.nested_features & HV_X64_NESTED_MSR_BITMAP)) { |
3445 |
+- struct hv_enlightened_vmcs *evmcs = |
3446 |
+- (struct hv_enlightened_vmcs *)loaded_vmcs->vmcs; |
3447 |
+- |
3448 |
+- evmcs->hv_enlightenments_control.msr_bitmap = 1; |
3449 |
+- } |
3450 |
+- } |
3451 |
+- |
3452 |
+- memset(&loaded_vmcs->host_state, 0, sizeof(struct vmcs_host_state)); |
3453 |
+- memset(&loaded_vmcs->controls_shadow, 0, |
3454 |
+- sizeof(struct vmcs_controls_shadow)); |
3455 |
+- |
3456 |
+- return 0; |
3457 |
+- |
3458 |
+-out_vmcs: |
3459 |
+- free_loaded_vmcs(loaded_vmcs); |
3460 |
+- return -ENOMEM; |
3461 |
+-} |
3462 |
+- |
3463 |
+-static void free_kvm_area(void) |
3464 |
+-{ |
3465 |
+- int cpu; |
3466 |
+- |
3467 |
+- for_each_possible_cpu(cpu) { |
3468 |
+- free_vmcs(per_cpu(vmxarea, cpu)); |
3469 |
+- per_cpu(vmxarea, cpu) = NULL; |
3470 |
+- } |
3471 |
+-} |
3472 |
+- |
3473 |
+-static __init int alloc_kvm_area(void) |
3474 |
+-{ |
3475 |
+- int cpu; |
3476 |
+- |
3477 |
+- for_each_possible_cpu(cpu) { |
3478 |
+- struct vmcs *vmcs; |
3479 |
+- |
3480 |
+- vmcs = alloc_vmcs_cpu(false, cpu, GFP_KERNEL); |
3481 |
+- if (!vmcs) { |
3482 |
+- free_kvm_area(); |
3483 |
+- return -ENOMEM; |
3484 |
+- } |
3485 |
+- |
3486 |
+- /* |
3487 |
+- * When eVMCS is enabled, alloc_vmcs_cpu() sets |
3488 |
+- * vmcs->revision_id to KVM_EVMCS_VERSION instead of |
3489 |
+- * revision_id reported by MSR_IA32_VMX_BASIC. |
3490 |
+- * |
3491 |
+- * However, even though not explicitly documented by |
3492 |
+- * TLFS, VMXArea passed as VMXON argument should |
3493 |
+- * still be marked with revision_id reported by |
3494 |
+- * physical CPU. |
3495 |
+- */ |
3496 |
+- if (static_branch_unlikely(&enable_evmcs)) |
3497 |
+- vmcs->hdr.revision_id = vmcs_config.revision_id; |
3498 |
+- |
3499 |
+- per_cpu(vmxarea, cpu) = vmcs; |
3500 |
+- } |
3501 |
+- return 0; |
3502 |
+-} |
3503 |
+- |
3504 |
+-static void fix_pmode_seg(struct kvm_vcpu *vcpu, int seg, |
3505 |
+- struct kvm_segment *save) |
3506 |
+-{ |
3507 |
+- if (!emulate_invalid_guest_state) { |
3508 |
+- /* |
3509 |
+- * CS and SS RPL should be equal during guest entry according |
3510 |
+- * to VMX spec, but in reality it is not always so. Since vcpu |
3511 |
+- * is in the middle of the transition from real mode to |
3512 |
+- * protected mode it is safe to assume that RPL 0 is a good |
3513 |
+- * default value. |
3514 |
+- */ |
3515 |
+- if (seg == VCPU_SREG_CS || seg == VCPU_SREG_SS) |
3516 |
+- save->selector &= ~SEGMENT_RPL_MASK; |
3517 |
+- save->dpl = save->selector & SEGMENT_RPL_MASK; |
3518 |
+- save->s = 1; |
3519 |
+- } |
3520 |
+- vmx_set_segment(vcpu, save, seg); |
3521 |
+-} |
3522 |
+- |
3523 |
+-static void enter_pmode(struct kvm_vcpu *vcpu) |
3524 |
+-{ |
3525 |
+- unsigned long flags; |
3526 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3527 |
+- |
3528 |
+- /* |
3529 |
+- * Update real mode segment cache. It may be not up-to-date if sement |
3530 |
+- * register was written while vcpu was in a guest mode. |
3531 |
+- */ |
3532 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES); |
3533 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS); |
3534 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_FS], VCPU_SREG_FS); |
3535 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_GS], VCPU_SREG_GS); |
3536 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_SS], VCPU_SREG_SS); |
3537 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_CS], VCPU_SREG_CS); |
3538 |
+- |
3539 |
+- vmx->rmode.vm86_active = 0; |
3540 |
+- |
3541 |
+- vmx_segment_cache_clear(vmx); |
3542 |
+- |
3543 |
+- vmx_set_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
3544 |
+- |
3545 |
+- flags = vmcs_readl(GUEST_RFLAGS); |
3546 |
+- flags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
3547 |
+- flags |= vmx->rmode.save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; |
3548 |
+- vmcs_writel(GUEST_RFLAGS, flags); |
3549 |
+- |
3550 |
+- vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~X86_CR4_VME) | |
3551 |
+- (vmcs_readl(CR4_READ_SHADOW) & X86_CR4_VME)); |
3552 |
+- |
3553 |
+- update_exception_bitmap(vcpu); |
3554 |
+- |
3555 |
+- fix_pmode_seg(vcpu, VCPU_SREG_CS, &vmx->rmode.segs[VCPU_SREG_CS]); |
3556 |
+- fix_pmode_seg(vcpu, VCPU_SREG_SS, &vmx->rmode.segs[VCPU_SREG_SS]); |
3557 |
+- fix_pmode_seg(vcpu, VCPU_SREG_ES, &vmx->rmode.segs[VCPU_SREG_ES]); |
3558 |
+- fix_pmode_seg(vcpu, VCPU_SREG_DS, &vmx->rmode.segs[VCPU_SREG_DS]); |
3559 |
+- fix_pmode_seg(vcpu, VCPU_SREG_FS, &vmx->rmode.segs[VCPU_SREG_FS]); |
3560 |
+- fix_pmode_seg(vcpu, VCPU_SREG_GS, &vmx->rmode.segs[VCPU_SREG_GS]); |
3561 |
+-} |
3562 |
+- |
3563 |
+-static void fix_rmode_seg(int seg, struct kvm_segment *save) |
3564 |
+-{ |
3565 |
+- const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
3566 |
+- struct kvm_segment var = *save; |
3567 |
+- |
3568 |
+- var.dpl = 0x3; |
3569 |
+- if (seg == VCPU_SREG_CS) |
3570 |
+- var.type = 0x3; |
3571 |
+- |
3572 |
+- if (!emulate_invalid_guest_state) { |
3573 |
+- var.selector = var.base >> 4; |
3574 |
+- var.base = var.base & 0xffff0; |
3575 |
+- var.limit = 0xffff; |
3576 |
+- var.g = 0; |
3577 |
+- var.db = 0; |
3578 |
+- var.present = 1; |
3579 |
+- var.s = 1; |
3580 |
+- var.l = 0; |
3581 |
+- var.unusable = 0; |
3582 |
+- var.type = 0x3; |
3583 |
+- var.avl = 0; |
3584 |
+- if (save->base & 0xf) |
3585 |
+- printk_once(KERN_WARNING "kvm: segment base is not " |
3586 |
+- "paragraph aligned when entering " |
3587 |
+- "protected mode (seg=%d)", seg); |
3588 |
+- } |
3589 |
+- |
3590 |
+- vmcs_write16(sf->selector, var.selector); |
3591 |
+- vmcs_writel(sf->base, var.base); |
3592 |
+- vmcs_write32(sf->limit, var.limit); |
3593 |
+- vmcs_write32(sf->ar_bytes, vmx_segment_access_rights(&var)); |
3594 |
+-} |
3595 |
+- |
3596 |
+-static void enter_rmode(struct kvm_vcpu *vcpu) |
3597 |
+-{ |
3598 |
+- unsigned long flags; |
3599 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3600 |
+- struct kvm_vmx *kvm_vmx = to_kvm_vmx(vcpu->kvm); |
3601 |
+- |
3602 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
3603 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES); |
3604 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS); |
3605 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_FS], VCPU_SREG_FS); |
3606 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_GS], VCPU_SREG_GS); |
3607 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_SS], VCPU_SREG_SS); |
3608 |
+- vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_CS], VCPU_SREG_CS); |
3609 |
+- |
3610 |
+- vmx->rmode.vm86_active = 1; |
3611 |
+- |
3612 |
+- /* |
3613 |
+- * Very old userspace does not call KVM_SET_TSS_ADDR before entering |
3614 |
+- * vcpu. Warn the user that an update is overdue. |
3615 |
+- */ |
3616 |
+- if (!kvm_vmx->tss_addr) |
3617 |
+- printk_once(KERN_WARNING "kvm: KVM_SET_TSS_ADDR need to be " |
3618 |
+- "called before entering vcpu\n"); |
3619 |
+- |
3620 |
+- vmx_segment_cache_clear(vmx); |
3621 |
+- |
3622 |
+- vmcs_writel(GUEST_TR_BASE, kvm_vmx->tss_addr); |
3623 |
+- vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1); |
3624 |
+- vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
3625 |
+- |
3626 |
+- flags = vmcs_readl(GUEST_RFLAGS); |
3627 |
+- vmx->rmode.save_rflags = flags; |
3628 |
+- |
3629 |
+- flags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
3630 |
+- |
3631 |
+- vmcs_writel(GUEST_RFLAGS, flags); |
3632 |
+- vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | X86_CR4_VME); |
3633 |
+- update_exception_bitmap(vcpu); |
3634 |
+- |
3635 |
+- fix_rmode_seg(VCPU_SREG_SS, &vmx->rmode.segs[VCPU_SREG_SS]); |
3636 |
+- fix_rmode_seg(VCPU_SREG_CS, &vmx->rmode.segs[VCPU_SREG_CS]); |
3637 |
+- fix_rmode_seg(VCPU_SREG_ES, &vmx->rmode.segs[VCPU_SREG_ES]); |
3638 |
+- fix_rmode_seg(VCPU_SREG_DS, &vmx->rmode.segs[VCPU_SREG_DS]); |
3639 |
+- fix_rmode_seg(VCPU_SREG_GS, &vmx->rmode.segs[VCPU_SREG_GS]); |
3640 |
+- fix_rmode_seg(VCPU_SREG_FS, &vmx->rmode.segs[VCPU_SREG_FS]); |
3641 |
+- |
3642 |
+- kvm_mmu_reset_context(vcpu); |
3643 |
+-} |
3644 |
+- |
3645 |
+-void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer) |
3646 |
+-{ |
3647 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3648 |
+- struct shared_msr_entry *msr = find_msr_entry(vmx, MSR_EFER); |
3649 |
+- |
3650 |
+- if (!msr) |
3651 |
+- return; |
3652 |
+- |
3653 |
+- vcpu->arch.efer = efer; |
3654 |
+- if (efer & EFER_LMA) { |
3655 |
+- vm_entry_controls_setbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
3656 |
+- msr->data = efer; |
3657 |
+- } else { |
3658 |
+- vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
3659 |
+- |
3660 |
+- msr->data = efer & ~EFER_LME; |
3661 |
+- } |
3662 |
+- setup_msrs(vmx); |
3663 |
+-} |
3664 |
+- |
3665 |
+-#ifdef CONFIG_X86_64 |
3666 |
+- |
3667 |
+-static void enter_lmode(struct kvm_vcpu *vcpu) |
3668 |
+-{ |
3669 |
+- u32 guest_tr_ar; |
3670 |
+- |
3671 |
+- vmx_segment_cache_clear(to_vmx(vcpu)); |
3672 |
+- |
3673 |
+- guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES); |
3674 |
+- if ((guest_tr_ar & VMX_AR_TYPE_MASK) != VMX_AR_TYPE_BUSY_64_TSS) { |
3675 |
+- pr_debug_ratelimited("%s: tss fixup for long mode. \n", |
3676 |
+- __func__); |
3677 |
+- vmcs_write32(GUEST_TR_AR_BYTES, |
3678 |
+- (guest_tr_ar & ~VMX_AR_TYPE_MASK) |
3679 |
+- | VMX_AR_TYPE_BUSY_64_TSS); |
3680 |
+- } |
3681 |
+- vmx_set_efer(vcpu, vcpu->arch.efer | EFER_LMA); |
3682 |
+-} |
3683 |
+- |
3684 |
+-static void exit_lmode(struct kvm_vcpu *vcpu) |
3685 |
+-{ |
3686 |
+- vm_entry_controls_clearbit(to_vmx(vcpu), VM_ENTRY_IA32E_MODE); |
3687 |
+- vmx_set_efer(vcpu, vcpu->arch.efer & ~EFER_LMA); |
3688 |
+-} |
3689 |
+- |
3690 |
+-#endif |
3691 |
+- |
3692 |
+-static void vmx_flush_tlb_gva(struct kvm_vcpu *vcpu, gva_t addr) |
3693 |
+-{ |
3694 |
+- int vpid = to_vmx(vcpu)->vpid; |
3695 |
+- |
3696 |
+- if (!vpid_sync_vcpu_addr(vpid, addr)) |
3697 |
+- vpid_sync_context(vpid); |
3698 |
+- |
3699 |
+- /* |
3700 |
+- * If VPIDs are not supported or enabled, then the above is a no-op. |
3701 |
+- * But we don't really need a TLB flush in that case anyway, because |
3702 |
+- * each VM entry/exit includes an implicit flush when VPID is 0. |
3703 |
+- */ |
3704 |
+-} |
3705 |
+- |
3706 |
+-static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu) |
3707 |
+-{ |
3708 |
+- ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits; |
3709 |
+- |
3710 |
+- vcpu->arch.cr0 &= ~cr0_guest_owned_bits; |
3711 |
+- vcpu->arch.cr0 |= vmcs_readl(GUEST_CR0) & cr0_guest_owned_bits; |
3712 |
+-} |
3713 |
+- |
3714 |
+-static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu) |
3715 |
+-{ |
3716 |
+- ulong cr4_guest_owned_bits = vcpu->arch.cr4_guest_owned_bits; |
3717 |
+- |
3718 |
+- vcpu->arch.cr4 &= ~cr4_guest_owned_bits; |
3719 |
+- vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & cr4_guest_owned_bits; |
3720 |
+-} |
3721 |
+- |
3722 |
+-static void ept_load_pdptrs(struct kvm_vcpu *vcpu) |
3723 |
+-{ |
3724 |
+- struct kvm_mmu *mmu = vcpu->arch.walk_mmu; |
3725 |
+- |
3726 |
+- if (!kvm_register_is_dirty(vcpu, VCPU_EXREG_PDPTR)) |
3727 |
+- return; |
3728 |
+- |
3729 |
+- if (is_pae_paging(vcpu)) { |
3730 |
+- vmcs_write64(GUEST_PDPTR0, mmu->pdptrs[0]); |
3731 |
+- vmcs_write64(GUEST_PDPTR1, mmu->pdptrs[1]); |
3732 |
+- vmcs_write64(GUEST_PDPTR2, mmu->pdptrs[2]); |
3733 |
+- vmcs_write64(GUEST_PDPTR3, mmu->pdptrs[3]); |
3734 |
+- } |
3735 |
+-} |
3736 |
+- |
3737 |
+-void ept_save_pdptrs(struct kvm_vcpu *vcpu) |
3738 |
+-{ |
3739 |
+- struct kvm_mmu *mmu = vcpu->arch.walk_mmu; |
3740 |
+- |
3741 |
+- if (is_pae_paging(vcpu)) { |
3742 |
+- mmu->pdptrs[0] = vmcs_read64(GUEST_PDPTR0); |
3743 |
+- mmu->pdptrs[1] = vmcs_read64(GUEST_PDPTR1); |
3744 |
+- mmu->pdptrs[2] = vmcs_read64(GUEST_PDPTR2); |
3745 |
+- mmu->pdptrs[3] = vmcs_read64(GUEST_PDPTR3); |
3746 |
+- } |
3747 |
+- |
3748 |
+- kvm_register_mark_dirty(vcpu, VCPU_EXREG_PDPTR); |
3749 |
+-} |
3750 |
+- |
3751 |
+-static void ept_update_paging_mode_cr0(unsigned long *hw_cr0, |
3752 |
+- unsigned long cr0, |
3753 |
+- struct kvm_vcpu *vcpu) |
3754 |
+-{ |
3755 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3756 |
+- |
3757 |
+- if (!kvm_register_is_available(vcpu, VCPU_EXREG_CR3)) |
3758 |
+- vmx_cache_reg(vcpu, VCPU_EXREG_CR3); |
3759 |
+- if (!(cr0 & X86_CR0_PG)) { |
3760 |
+- /* From paging/starting to nonpaging */ |
3761 |
+- exec_controls_setbit(vmx, CPU_BASED_CR3_LOAD_EXITING | |
3762 |
+- CPU_BASED_CR3_STORE_EXITING); |
3763 |
+- vcpu->arch.cr0 = cr0; |
3764 |
+- vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
3765 |
+- } else if (!is_paging(vcpu)) { |
3766 |
+- /* From nonpaging to paging */ |
3767 |
+- exec_controls_clearbit(vmx, CPU_BASED_CR3_LOAD_EXITING | |
3768 |
+- CPU_BASED_CR3_STORE_EXITING); |
3769 |
+- vcpu->arch.cr0 = cr0; |
3770 |
+- vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
3771 |
+- } |
3772 |
+- |
3773 |
+- if (!(cr0 & X86_CR0_WP)) |
3774 |
+- *hw_cr0 &= ~X86_CR0_WP; |
3775 |
+-} |
3776 |
+- |
3777 |
+-void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
3778 |
+-{ |
3779 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3780 |
+- unsigned long hw_cr0; |
3781 |
+- |
3782 |
+- hw_cr0 = (cr0 & ~KVM_VM_CR0_ALWAYS_OFF); |
3783 |
+- if (enable_unrestricted_guest) |
3784 |
+- hw_cr0 |= KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST; |
3785 |
+- else { |
3786 |
+- hw_cr0 |= KVM_VM_CR0_ALWAYS_ON; |
3787 |
+- |
3788 |
+- if (vmx->rmode.vm86_active && (cr0 & X86_CR0_PE)) |
3789 |
+- enter_pmode(vcpu); |
3790 |
+- |
3791 |
+- if (!vmx->rmode.vm86_active && !(cr0 & X86_CR0_PE)) |
3792 |
+- enter_rmode(vcpu); |
3793 |
+- } |
3794 |
+- |
3795 |
+-#ifdef CONFIG_X86_64 |
3796 |
+- if (vcpu->arch.efer & EFER_LME) { |
3797 |
+- if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) |
3798 |
+- enter_lmode(vcpu); |
3799 |
+- if (is_paging(vcpu) && !(cr0 & X86_CR0_PG)) |
3800 |
+- exit_lmode(vcpu); |
3801 |
+- } |
3802 |
+-#endif |
3803 |
+- |
3804 |
+- if (enable_ept && !enable_unrestricted_guest) |
3805 |
+- ept_update_paging_mode_cr0(&hw_cr0, cr0, vcpu); |
3806 |
+- |
3807 |
+- vmcs_writel(CR0_READ_SHADOW, cr0); |
3808 |
+- vmcs_writel(GUEST_CR0, hw_cr0); |
3809 |
+- vcpu->arch.cr0 = cr0; |
3810 |
+- |
3811 |
+- /* depends on vcpu->arch.cr0 to be set to a new value */ |
3812 |
+- vmx->emulation_required = emulation_required(vcpu); |
3813 |
+-} |
3814 |
+- |
3815 |
+-static int get_ept_level(struct kvm_vcpu *vcpu) |
3816 |
+-{ |
3817 |
+- /* Nested EPT currently only supports 4-level walks. */ |
3818 |
+- if (is_guest_mode(vcpu) && nested_cpu_has_ept(get_vmcs12(vcpu))) |
3819 |
+- return 4; |
3820 |
+- if (cpu_has_vmx_ept_5levels() && (cpuid_maxphyaddr(vcpu) > 48)) |
3821 |
+- return 5; |
3822 |
+- return 4; |
3823 |
+-} |
3824 |
+- |
3825 |
+-u64 construct_eptp(struct kvm_vcpu *vcpu, unsigned long root_hpa) |
3826 |
+-{ |
3827 |
+- u64 eptp = VMX_EPTP_MT_WB; |
3828 |
+- |
3829 |
+- eptp |= (get_ept_level(vcpu) == 5) ? VMX_EPTP_PWL_5 : VMX_EPTP_PWL_4; |
3830 |
+- |
3831 |
+- if (enable_ept_ad_bits && |
3832 |
+- (!is_guest_mode(vcpu) || nested_ept_ad_enabled(vcpu))) |
3833 |
+- eptp |= VMX_EPTP_AD_ENABLE_BIT; |
3834 |
+- eptp |= (root_hpa & PAGE_MASK); |
3835 |
+- |
3836 |
+- return eptp; |
3837 |
+-} |
3838 |
+- |
3839 |
+-void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) |
3840 |
+-{ |
3841 |
+- struct kvm *kvm = vcpu->kvm; |
3842 |
+- bool update_guest_cr3 = true; |
3843 |
+- unsigned long guest_cr3; |
3844 |
+- u64 eptp; |
3845 |
+- |
3846 |
+- guest_cr3 = cr3; |
3847 |
+- if (enable_ept) { |
3848 |
+- eptp = construct_eptp(vcpu, cr3); |
3849 |
+- vmcs_write64(EPT_POINTER, eptp); |
3850 |
+- |
3851 |
+- if (kvm_x86_ops->tlb_remote_flush) { |
3852 |
+- spin_lock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
3853 |
+- to_vmx(vcpu)->ept_pointer = eptp; |
3854 |
+- to_kvm_vmx(kvm)->ept_pointers_match |
3855 |
+- = EPT_POINTERS_CHECK; |
3856 |
+- spin_unlock(&to_kvm_vmx(kvm)->ept_pointer_lock); |
3857 |
+- } |
3858 |
+- |
3859 |
+- /* Loading vmcs02.GUEST_CR3 is handled by nested VM-Enter. */ |
3860 |
+- if (is_guest_mode(vcpu)) |
3861 |
+- update_guest_cr3 = false; |
3862 |
+- else if (!enable_unrestricted_guest && !is_paging(vcpu)) |
3863 |
+- guest_cr3 = to_kvm_vmx(kvm)->ept_identity_map_addr; |
3864 |
+- else if (test_bit(VCPU_EXREG_CR3, (ulong *)&vcpu->arch.regs_avail)) |
3865 |
+- guest_cr3 = vcpu->arch.cr3; |
3866 |
+- else /* vmcs01.GUEST_CR3 is already up-to-date. */ |
3867 |
+- update_guest_cr3 = false; |
3868 |
+- ept_load_pdptrs(vcpu); |
3869 |
+- } |
3870 |
+- |
3871 |
+- if (update_guest_cr3) |
3872 |
+- vmcs_writel(GUEST_CR3, guest_cr3); |
3873 |
+-} |
3874 |
+- |
3875 |
+-int vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
3876 |
+-{ |
3877 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3878 |
+- /* |
3879 |
+- * Pass through host's Machine Check Enable value to hw_cr4, which |
3880 |
+- * is in force while we are in guest mode. Do not let guests control |
3881 |
+- * this bit, even if host CR4.MCE == 0. |
3882 |
+- */ |
3883 |
+- unsigned long hw_cr4; |
3884 |
+- |
3885 |
+- hw_cr4 = (cr4_read_shadow() & X86_CR4_MCE) | (cr4 & ~X86_CR4_MCE); |
3886 |
+- if (enable_unrestricted_guest) |
3887 |
+- hw_cr4 |= KVM_VM_CR4_ALWAYS_ON_UNRESTRICTED_GUEST; |
3888 |
+- else if (vmx->rmode.vm86_active) |
3889 |
+- hw_cr4 |= KVM_RMODE_VM_CR4_ALWAYS_ON; |
3890 |
+- else |
3891 |
+- hw_cr4 |= KVM_PMODE_VM_CR4_ALWAYS_ON; |
3892 |
+- |
3893 |
+- if (!boot_cpu_has(X86_FEATURE_UMIP) && vmx_umip_emulated()) { |
3894 |
+- if (cr4 & X86_CR4_UMIP) { |
3895 |
+- secondary_exec_controls_setbit(vmx, SECONDARY_EXEC_DESC); |
3896 |
+- hw_cr4 &= ~X86_CR4_UMIP; |
3897 |
+- } else if (!is_guest_mode(vcpu) || |
3898 |
+- !nested_cpu_has2(get_vmcs12(vcpu), SECONDARY_EXEC_DESC)) { |
3899 |
+- secondary_exec_controls_clearbit(vmx, SECONDARY_EXEC_DESC); |
3900 |
+- } |
3901 |
+- } |
3902 |
+- |
3903 |
+- if (cr4 & X86_CR4_VMXE) { |
3904 |
+- /* |
3905 |
+- * To use VMXON (and later other VMX instructions), a guest |
3906 |
+- * must first be able to turn on cr4.VMXE (see handle_vmon()). |
3907 |
+- * So basically the check on whether to allow nested VMX |
3908 |
+- * is here. We operate under the default treatment of SMM, |
3909 |
+- * so VMX cannot be enabled under SMM. |
3910 |
+- */ |
3911 |
+- if (!nested_vmx_allowed(vcpu) || is_smm(vcpu)) |
3912 |
+- return 1; |
3913 |
+- } |
3914 |
+- |
3915 |
+- if (vmx->nested.vmxon && !nested_cr4_valid(vcpu, cr4)) |
3916 |
+- return 1; |
3917 |
+- |
3918 |
+- vcpu->arch.cr4 = cr4; |
3919 |
+- |
3920 |
+- if (!enable_unrestricted_guest) { |
3921 |
+- if (enable_ept) { |
3922 |
+- if (!is_paging(vcpu)) { |
3923 |
+- hw_cr4 &= ~X86_CR4_PAE; |
3924 |
+- hw_cr4 |= X86_CR4_PSE; |
3925 |
+- } else if (!(cr4 & X86_CR4_PAE)) { |
3926 |
+- hw_cr4 &= ~X86_CR4_PAE; |
3927 |
+- } |
3928 |
+- } |
3929 |
+- |
3930 |
+- /* |
3931 |
+- * SMEP/SMAP/PKU is disabled if CPU is in non-paging mode in |
3932 |
+- * hardware. To emulate this behavior, SMEP/SMAP/PKU needs |
3933 |
+- * to be manually disabled when guest switches to non-paging |
3934 |
+- * mode. |
3935 |
+- * |
3936 |
+- * If !enable_unrestricted_guest, the CPU is always running |
3937 |
+- * with CR0.PG=1 and CR4 needs to be modified. |
3938 |
+- * If enable_unrestricted_guest, the CPU automatically |
3939 |
+- * disables SMEP/SMAP/PKU when the guest sets CR0.PG=0. |
3940 |
+- */ |
3941 |
+- if (!is_paging(vcpu)) |
3942 |
+- hw_cr4 &= ~(X86_CR4_SMEP | X86_CR4_SMAP | X86_CR4_PKE); |
3943 |
+- } |
3944 |
+- |
3945 |
+- vmcs_writel(CR4_READ_SHADOW, cr4); |
3946 |
+- vmcs_writel(GUEST_CR4, hw_cr4); |
3947 |
+- return 0; |
3948 |
+-} |
3949 |
+- |
3950 |
+-void vmx_get_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg) |
3951 |
+-{ |
3952 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
3953 |
+- u32 ar; |
3954 |
+- |
3955 |
+- if (vmx->rmode.vm86_active && seg != VCPU_SREG_LDTR) { |
3956 |
+- *var = vmx->rmode.segs[seg]; |
3957 |
+- if (seg == VCPU_SREG_TR |
3958 |
+- || var->selector == vmx_read_guest_seg_selector(vmx, seg)) |
3959 |
+- return; |
3960 |
+- var->base = vmx_read_guest_seg_base(vmx, seg); |
3961 |
+- var->selector = vmx_read_guest_seg_selector(vmx, seg); |
3962 |
+- return; |
3963 |
+- } |
3964 |
+- var->base = vmx_read_guest_seg_base(vmx, seg); |
3965 |
+- var->limit = vmx_read_guest_seg_limit(vmx, seg); |
3966 |
+- var->selector = vmx_read_guest_seg_selector(vmx, seg); |
3967 |
+- ar = vmx_read_guest_seg_ar(vmx, seg); |
3968 |
+- var->unusable = (ar >> 16) & 1; |
3969 |
+- var->type = ar & 15; |
3970 |
+- var->s = (ar >> 4) & 1; |
3971 |
+- var->dpl = (ar >> 5) & 3; |
3972 |
+- /* |
3973 |
+- * Some userspaces do not preserve unusable property. Since usable |
3974 |
+- * segment has to be present according to VMX spec we can use present |
3975 |
+- * property to amend userspace bug by making unusable segment always |
3976 |
+- * nonpresent. vmx_segment_access_rights() already marks nonpresent |
3977 |
+- * segment as unusable. |
3978 |
+- */ |
3979 |
+- var->present = !var->unusable; |
3980 |
+- var->avl = (ar >> 12) & 1; |
3981 |
+- var->l = (ar >> 13) & 1; |
3982 |
+- var->db = (ar >> 14) & 1; |
3983 |
+- var->g = (ar >> 15) & 1; |
3984 |
+-} |
3985 |
+- |
3986 |
+-static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg) |
3987 |
+-{ |
3988 |
+- struct kvm_segment s; |
3989 |
+- |
3990 |
+- if (to_vmx(vcpu)->rmode.vm86_active) { |
3991 |
+- vmx_get_segment(vcpu, &s, seg); |
3992 |
+- return s.base; |
3993 |
+- } |
3994 |
+- return vmx_read_guest_seg_base(to_vmx(vcpu), seg); |
3995 |
+-} |
3996 |
+- |
3997 |
+-int vmx_get_cpl(struct kvm_vcpu *vcpu) |
3998 |
+-{ |
3999 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4000 |
+- |
4001 |
+- if (unlikely(vmx->rmode.vm86_active)) |
4002 |
+- return 0; |
4003 |
+- else { |
4004 |
+- int ar = vmx_read_guest_seg_ar(vmx, VCPU_SREG_SS); |
4005 |
+- return VMX_AR_DPL(ar); |
4006 |
+- } |
4007 |
+-} |
4008 |
+- |
4009 |
+-static u32 vmx_segment_access_rights(struct kvm_segment *var) |
4010 |
+-{ |
4011 |
+- u32 ar; |
4012 |
+- |
4013 |
+- if (var->unusable || !var->present) |
4014 |
+- ar = 1 << 16; |
4015 |
+- else { |
4016 |
+- ar = var->type & 15; |
4017 |
+- ar |= (var->s & 1) << 4; |
4018 |
+- ar |= (var->dpl & 3) << 5; |
4019 |
+- ar |= (var->present & 1) << 7; |
4020 |
+- ar |= (var->avl & 1) << 12; |
4021 |
+- ar |= (var->l & 1) << 13; |
4022 |
+- ar |= (var->db & 1) << 14; |
4023 |
+- ar |= (var->g & 1) << 15; |
4024 |
+- } |
4025 |
+- |
4026 |
+- return ar; |
4027 |
+-} |
4028 |
+- |
4029 |
+-void vmx_set_segment(struct kvm_vcpu *vcpu, struct kvm_segment *var, int seg) |
4030 |
+-{ |
4031 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4032 |
+- const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
4033 |
+- |
4034 |
+- vmx_segment_cache_clear(vmx); |
4035 |
+- |
4036 |
+- if (vmx->rmode.vm86_active && seg != VCPU_SREG_LDTR) { |
4037 |
+- vmx->rmode.segs[seg] = *var; |
4038 |
+- if (seg == VCPU_SREG_TR) |
4039 |
+- vmcs_write16(sf->selector, var->selector); |
4040 |
+- else if (var->s) |
4041 |
+- fix_rmode_seg(seg, &vmx->rmode.segs[seg]); |
4042 |
+- goto out; |
4043 |
+- } |
4044 |
+- |
4045 |
+- vmcs_writel(sf->base, var->base); |
4046 |
+- vmcs_write32(sf->limit, var->limit); |
4047 |
+- vmcs_write16(sf->selector, var->selector); |
4048 |
+- |
4049 |
+- /* |
4050 |
+- * Fix the "Accessed" bit in AR field of segment registers for older |
4051 |
+- * qemu binaries. |
4052 |
+- * IA32 arch specifies that at the time of processor reset the |
4053 |
+- * "Accessed" bit in the AR field of segment registers is 1. And qemu |
4054 |
+- * is setting it to 0 in the userland code. This causes invalid guest |
4055 |
+- * state vmexit when "unrestricted guest" mode is turned on. |
4056 |
+- * Fix for this setup issue in cpu_reset is being pushed in the qemu |
4057 |
+- * tree. Newer qemu binaries with that qemu fix would not need this |
4058 |
+- * kvm hack. |
4059 |
+- */ |
4060 |
+- if (enable_unrestricted_guest && (seg != VCPU_SREG_LDTR)) |
4061 |
+- var->type |= 0x1; /* Accessed */ |
4062 |
+- |
4063 |
+- vmcs_write32(sf->ar_bytes, vmx_segment_access_rights(var)); |
4064 |
+- |
4065 |
+-out: |
4066 |
+- vmx->emulation_required = emulation_required(vcpu); |
4067 |
+-} |
4068 |
+- |
4069 |
+-static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
4070 |
+-{ |
4071 |
+- u32 ar = vmx_read_guest_seg_ar(to_vmx(vcpu), VCPU_SREG_CS); |
4072 |
+- |
4073 |
+- *db = (ar >> 14) & 1; |
4074 |
+- *l = (ar >> 13) & 1; |
4075 |
+-} |
4076 |
+- |
4077 |
+-static void vmx_get_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4078 |
+-{ |
4079 |
+- dt->size = vmcs_read32(GUEST_IDTR_LIMIT); |
4080 |
+- dt->address = vmcs_readl(GUEST_IDTR_BASE); |
4081 |
+-} |
4082 |
+- |
4083 |
+-static void vmx_set_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4084 |
+-{ |
4085 |
+- vmcs_write32(GUEST_IDTR_LIMIT, dt->size); |
4086 |
+- vmcs_writel(GUEST_IDTR_BASE, dt->address); |
4087 |
+-} |
4088 |
+- |
4089 |
+-static void vmx_get_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4090 |
+-{ |
4091 |
+- dt->size = vmcs_read32(GUEST_GDTR_LIMIT); |
4092 |
+- dt->address = vmcs_readl(GUEST_GDTR_BASE); |
4093 |
+-} |
4094 |
+- |
4095 |
+-static void vmx_set_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
4096 |
+-{ |
4097 |
+- vmcs_write32(GUEST_GDTR_LIMIT, dt->size); |
4098 |
+- vmcs_writel(GUEST_GDTR_BASE, dt->address); |
4099 |
+-} |
4100 |
+- |
4101 |
+-static bool rmode_segment_valid(struct kvm_vcpu *vcpu, int seg) |
4102 |
+-{ |
4103 |
+- struct kvm_segment var; |
4104 |
+- u32 ar; |
4105 |
+- |
4106 |
+- vmx_get_segment(vcpu, &var, seg); |
4107 |
+- var.dpl = 0x3; |
4108 |
+- if (seg == VCPU_SREG_CS) |
4109 |
+- var.type = 0x3; |
4110 |
+- ar = vmx_segment_access_rights(&var); |
4111 |
+- |
4112 |
+- if (var.base != (var.selector << 4)) |
4113 |
+- return false; |
4114 |
+- if (var.limit != 0xffff) |
4115 |
+- return false; |
4116 |
+- if (ar != 0xf3) |
4117 |
+- return false; |
4118 |
+- |
4119 |
+- return true; |
4120 |
+-} |
4121 |
+- |
4122 |
+-static bool code_segment_valid(struct kvm_vcpu *vcpu) |
4123 |
+-{ |
4124 |
+- struct kvm_segment cs; |
4125 |
+- unsigned int cs_rpl; |
4126 |
+- |
4127 |
+- vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); |
4128 |
+- cs_rpl = cs.selector & SEGMENT_RPL_MASK; |
4129 |
+- |
4130 |
+- if (cs.unusable) |
4131 |
+- return false; |
4132 |
+- if (~cs.type & (VMX_AR_TYPE_CODE_MASK|VMX_AR_TYPE_ACCESSES_MASK)) |
4133 |
+- return false; |
4134 |
+- if (!cs.s) |
4135 |
+- return false; |
4136 |
+- if (cs.type & VMX_AR_TYPE_WRITEABLE_MASK) { |
4137 |
+- if (cs.dpl > cs_rpl) |
4138 |
+- return false; |
4139 |
+- } else { |
4140 |
+- if (cs.dpl != cs_rpl) |
4141 |
+- return false; |
4142 |
+- } |
4143 |
+- if (!cs.present) |
4144 |
+- return false; |
4145 |
+- |
4146 |
+- /* TODO: Add Reserved field check, this'll require a new member in the kvm_segment_field structure */ |
4147 |
+- return true; |
4148 |
+-} |
4149 |
+- |
4150 |
+-static bool stack_segment_valid(struct kvm_vcpu *vcpu) |
4151 |
+-{ |
4152 |
+- struct kvm_segment ss; |
4153 |
+- unsigned int ss_rpl; |
4154 |
+- |
4155 |
+- vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); |
4156 |
+- ss_rpl = ss.selector & SEGMENT_RPL_MASK; |
4157 |
+- |
4158 |
+- if (ss.unusable) |
4159 |
+- return true; |
4160 |
+- if (ss.type != 3 && ss.type != 7) |
4161 |
+- return false; |
4162 |
+- if (!ss.s) |
4163 |
+- return false; |
4164 |
+- if (ss.dpl != ss_rpl) /* DPL != RPL */ |
4165 |
+- return false; |
4166 |
+- if (!ss.present) |
4167 |
+- return false; |
4168 |
+- |
4169 |
+- return true; |
4170 |
+-} |
4171 |
+- |
4172 |
+-static bool data_segment_valid(struct kvm_vcpu *vcpu, int seg) |
4173 |
+-{ |
4174 |
+- struct kvm_segment var; |
4175 |
+- unsigned int rpl; |
4176 |
+- |
4177 |
+- vmx_get_segment(vcpu, &var, seg); |
4178 |
+- rpl = var.selector & SEGMENT_RPL_MASK; |
4179 |
+- |
4180 |
+- if (var.unusable) |
4181 |
+- return true; |
4182 |
+- if (!var.s) |
4183 |
+- return false; |
4184 |
+- if (!var.present) |
4185 |
+- return false; |
4186 |
+- if (~var.type & (VMX_AR_TYPE_CODE_MASK|VMX_AR_TYPE_WRITEABLE_MASK)) { |
4187 |
+- if (var.dpl < rpl) /* DPL < RPL */ |
4188 |
+- return false; |
4189 |
+- } |
4190 |
+- |
4191 |
+- /* TODO: Add other members to kvm_segment_field to allow checking for other access |
4192 |
+- * rights flags |
4193 |
+- */ |
4194 |
+- return true; |
4195 |
+-} |
4196 |
+- |
4197 |
+-static bool tr_valid(struct kvm_vcpu *vcpu) |
4198 |
+-{ |
4199 |
+- struct kvm_segment tr; |
4200 |
+- |
4201 |
+- vmx_get_segment(vcpu, &tr, VCPU_SREG_TR); |
4202 |
+- |
4203 |
+- if (tr.unusable) |
4204 |
+- return false; |
4205 |
+- if (tr.selector & SEGMENT_TI_MASK) /* TI = 1 */ |
4206 |
+- return false; |
4207 |
+- if (tr.type != 3 && tr.type != 11) /* TODO: Check if guest is in IA32e mode */ |
4208 |
+- return false; |
4209 |
+- if (!tr.present) |
4210 |
+- return false; |
4211 |
+- |
4212 |
+- return true; |
4213 |
+-} |
4214 |
+- |
4215 |
+-static bool ldtr_valid(struct kvm_vcpu *vcpu) |
4216 |
+-{ |
4217 |
+- struct kvm_segment ldtr; |
4218 |
+- |
4219 |
+- vmx_get_segment(vcpu, &ldtr, VCPU_SREG_LDTR); |
4220 |
+- |
4221 |
+- if (ldtr.unusable) |
4222 |
+- return true; |
4223 |
+- if (ldtr.selector & SEGMENT_TI_MASK) /* TI = 1 */ |
4224 |
+- return false; |
4225 |
+- if (ldtr.type != 2) |
4226 |
+- return false; |
4227 |
+- if (!ldtr.present) |
4228 |
+- return false; |
4229 |
+- |
4230 |
+- return true; |
4231 |
+-} |
4232 |
+- |
4233 |
+-static bool cs_ss_rpl_check(struct kvm_vcpu *vcpu) |
4234 |
+-{ |
4235 |
+- struct kvm_segment cs, ss; |
4236 |
+- |
4237 |
+- vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); |
4238 |
+- vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); |
4239 |
+- |
4240 |
+- return ((cs.selector & SEGMENT_RPL_MASK) == |
4241 |
+- (ss.selector & SEGMENT_RPL_MASK)); |
4242 |
+-} |
4243 |
+- |
4244 |
+-/* |
4245 |
+- * Check if guest state is valid. Returns true if valid, false if |
4246 |
+- * not. |
4247 |
+- * We assume that registers are always usable |
4248 |
+- */ |
4249 |
+-static bool guest_state_valid(struct kvm_vcpu *vcpu) |
4250 |
+-{ |
4251 |
+- if (enable_unrestricted_guest) |
4252 |
+- return true; |
4253 |
+- |
4254 |
+- /* real mode guest state checks */ |
4255 |
+- if (!is_protmode(vcpu) || (vmx_get_rflags(vcpu) & X86_EFLAGS_VM)) { |
4256 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_CS)) |
4257 |
+- return false; |
4258 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_SS)) |
4259 |
+- return false; |
4260 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_DS)) |
4261 |
+- return false; |
4262 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_ES)) |
4263 |
+- return false; |
4264 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_FS)) |
4265 |
+- return false; |
4266 |
+- if (!rmode_segment_valid(vcpu, VCPU_SREG_GS)) |
4267 |
+- return false; |
4268 |
+- } else { |
4269 |
+- /* protected mode guest state checks */ |
4270 |
+- if (!cs_ss_rpl_check(vcpu)) |
4271 |
+- return false; |
4272 |
+- if (!code_segment_valid(vcpu)) |
4273 |
+- return false; |
4274 |
+- if (!stack_segment_valid(vcpu)) |
4275 |
+- return false; |
4276 |
+- if (!data_segment_valid(vcpu, VCPU_SREG_DS)) |
4277 |
+- return false; |
4278 |
+- if (!data_segment_valid(vcpu, VCPU_SREG_ES)) |
4279 |
+- return false; |
4280 |
+- if (!data_segment_valid(vcpu, VCPU_SREG_FS)) |
4281 |
+- return false; |
4282 |
+- if (!data_segment_valid(vcpu, VCPU_SREG_GS)) |
4283 |
+- return false; |
4284 |
+- if (!tr_valid(vcpu)) |
4285 |
+- return false; |
4286 |
+- if (!ldtr_valid(vcpu)) |
4287 |
+- return false; |
4288 |
+- } |
4289 |
+- /* TODO: |
4290 |
+- * - Add checks on RIP |
4291 |
+- * - Add checks on RFLAGS |
4292 |
+- */ |
4293 |
+- |
4294 |
+- return true; |
4295 |
+-} |
4296 |
+- |
4297 |
+-static int init_rmode_tss(struct kvm *kvm) |
4298 |
+-{ |
4299 |
+- gfn_t fn; |
4300 |
+- u16 data = 0; |
4301 |
+- int idx, r; |
4302 |
+- |
4303 |
+- idx = srcu_read_lock(&kvm->srcu); |
4304 |
+- fn = to_kvm_vmx(kvm)->tss_addr >> PAGE_SHIFT; |
4305 |
+- r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
4306 |
+- if (r < 0) |
4307 |
+- goto out; |
4308 |
+- data = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE; |
4309 |
+- r = kvm_write_guest_page(kvm, fn++, &data, |
4310 |
+- TSS_IOPB_BASE_OFFSET, sizeof(u16)); |
4311 |
+- if (r < 0) |
4312 |
+- goto out; |
4313 |
+- r = kvm_clear_guest_page(kvm, fn++, 0, PAGE_SIZE); |
4314 |
+- if (r < 0) |
4315 |
+- goto out; |
4316 |
+- r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
4317 |
+- if (r < 0) |
4318 |
+- goto out; |
4319 |
+- data = ~0; |
4320 |
+- r = kvm_write_guest_page(kvm, fn, &data, |
4321 |
+- RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1, |
4322 |
+- sizeof(u8)); |
4323 |
+-out: |
4324 |
+- srcu_read_unlock(&kvm->srcu, idx); |
4325 |
+- return r; |
4326 |
+-} |
4327 |
+- |
4328 |
+-static int init_rmode_identity_map(struct kvm *kvm) |
4329 |
+-{ |
4330 |
+- struct kvm_vmx *kvm_vmx = to_kvm_vmx(kvm); |
4331 |
+- int i, idx, r = 0; |
4332 |
+- kvm_pfn_t identity_map_pfn; |
4333 |
+- u32 tmp; |
4334 |
+- |
4335 |
+- /* Protect kvm_vmx->ept_identity_pagetable_done. */ |
4336 |
+- mutex_lock(&kvm->slots_lock); |
4337 |
+- |
4338 |
+- if (likely(kvm_vmx->ept_identity_pagetable_done)) |
4339 |
+- goto out2; |
4340 |
+- |
4341 |
+- if (!kvm_vmx->ept_identity_map_addr) |
4342 |
+- kvm_vmx->ept_identity_map_addr = VMX_EPT_IDENTITY_PAGETABLE_ADDR; |
4343 |
+- identity_map_pfn = kvm_vmx->ept_identity_map_addr >> PAGE_SHIFT; |
4344 |
+- |
4345 |
+- r = __x86_set_memory_region(kvm, IDENTITY_PAGETABLE_PRIVATE_MEMSLOT, |
4346 |
+- kvm_vmx->ept_identity_map_addr, PAGE_SIZE); |
4347 |
+- if (r < 0) |
4348 |
+- goto out2; |
4349 |
+- |
4350 |
+- idx = srcu_read_lock(&kvm->srcu); |
4351 |
+- r = kvm_clear_guest_page(kvm, identity_map_pfn, 0, PAGE_SIZE); |
4352 |
+- if (r < 0) |
4353 |
+- goto out; |
4354 |
+- /* Set up identity-mapping pagetable for EPT in real mode */ |
4355 |
+- for (i = 0; i < PT32_ENT_PER_PAGE; i++) { |
4356 |
+- tmp = (i << 22) + (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | |
4357 |
+- _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_PSE); |
4358 |
+- r = kvm_write_guest_page(kvm, identity_map_pfn, |
4359 |
+- &tmp, i * sizeof(tmp), sizeof(tmp)); |
4360 |
+- if (r < 0) |
4361 |
+- goto out; |
4362 |
+- } |
4363 |
+- kvm_vmx->ept_identity_pagetable_done = true; |
4364 |
+- |
4365 |
+-out: |
4366 |
+- srcu_read_unlock(&kvm->srcu, idx); |
4367 |
+- |
4368 |
+-out2: |
4369 |
+- mutex_unlock(&kvm->slots_lock); |
4370 |
+- return r; |
4371 |
+-} |
4372 |
+- |
4373 |
+-static void seg_setup(int seg) |
4374 |
+-{ |
4375 |
+- const struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
4376 |
+- unsigned int ar; |
4377 |
+- |
4378 |
+- vmcs_write16(sf->selector, 0); |
4379 |
+- vmcs_writel(sf->base, 0); |
4380 |
+- vmcs_write32(sf->limit, 0xffff); |
4381 |
+- ar = 0x93; |
4382 |
+- if (seg == VCPU_SREG_CS) |
4383 |
+- ar |= 0x08; /* code segment */ |
4384 |
+- |
4385 |
+- vmcs_write32(sf->ar_bytes, ar); |
4386 |
+-} |
4387 |
+- |
4388 |
+-static int alloc_apic_access_page(struct kvm *kvm) |
4389 |
+-{ |
4390 |
+- struct page *page; |
4391 |
+- int r = 0; |
4392 |
+- |
4393 |
+- mutex_lock(&kvm->slots_lock); |
4394 |
+- if (kvm->arch.apic_access_page_done) |
4395 |
+- goto out; |
4396 |
+- r = __x86_set_memory_region(kvm, APIC_ACCESS_PAGE_PRIVATE_MEMSLOT, |
4397 |
+- APIC_DEFAULT_PHYS_BASE, PAGE_SIZE); |
4398 |
+- if (r) |
4399 |
+- goto out; |
4400 |
+- |
4401 |
+- page = gfn_to_page(kvm, APIC_DEFAULT_PHYS_BASE >> PAGE_SHIFT); |
4402 |
+- if (is_error_page(page)) { |
4403 |
+- r = -EFAULT; |
4404 |
+- goto out; |
4405 |
+- } |
4406 |
+- |
4407 |
+- /* |
4408 |
+- * Do not pin the page in memory, so that memory hot-unplug |
4409 |
+- * is able to migrate it. |
4410 |
+- */ |
4411 |
+- put_page(page); |
4412 |
+- kvm->arch.apic_access_page_done = true; |
4413 |
+-out: |
4414 |
+- mutex_unlock(&kvm->slots_lock); |
4415 |
+- return r; |
4416 |
+-} |
4417 |
+- |
4418 |
+-int allocate_vpid(void) |
4419 |
+-{ |
4420 |
+- int vpid; |
4421 |
+- |
4422 |
+- if (!enable_vpid) |
4423 |
+- return 0; |
4424 |
+- spin_lock(&vmx_vpid_lock); |
4425 |
+- vpid = find_first_zero_bit(vmx_vpid_bitmap, VMX_NR_VPIDS); |
4426 |
+- if (vpid < VMX_NR_VPIDS) |
4427 |
+- __set_bit(vpid, vmx_vpid_bitmap); |
4428 |
+- else |
4429 |
+- vpid = 0; |
4430 |
+- spin_unlock(&vmx_vpid_lock); |
4431 |
+- return vpid; |
4432 |
+-} |
4433 |
+- |
4434 |
+-void free_vpid(int vpid) |
4435 |
+-{ |
4436 |
+- if (!enable_vpid || vpid == 0) |
4437 |
+- return; |
4438 |
+- spin_lock(&vmx_vpid_lock); |
4439 |
+- __clear_bit(vpid, vmx_vpid_bitmap); |
4440 |
+- spin_unlock(&vmx_vpid_lock); |
4441 |
+-} |
4442 |
+- |
4443 |
+-static __always_inline void vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, |
4444 |
+- u32 msr, int type) |
4445 |
+-{ |
4446 |
+- int f = sizeof(unsigned long); |
4447 |
+- |
4448 |
+- if (!cpu_has_vmx_msr_bitmap()) |
4449 |
+- return; |
4450 |
+- |
4451 |
+- if (static_branch_unlikely(&enable_evmcs)) |
4452 |
+- evmcs_touch_msr_bitmap(); |
4453 |
+- |
4454 |
+- /* |
4455 |
+- * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals |
4456 |
+- * have the write-low and read-high bitmap offsets the wrong way round. |
4457 |
+- * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. |
4458 |
+- */ |
4459 |
+- if (msr <= 0x1fff) { |
4460 |
+- if (type & MSR_TYPE_R) |
4461 |
+- /* read-low */ |
4462 |
+- __clear_bit(msr, msr_bitmap + 0x000 / f); |
4463 |
+- |
4464 |
+- if (type & MSR_TYPE_W) |
4465 |
+- /* write-low */ |
4466 |
+- __clear_bit(msr, msr_bitmap + 0x800 / f); |
4467 |
+- |
4468 |
+- } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
4469 |
+- msr &= 0x1fff; |
4470 |
+- if (type & MSR_TYPE_R) |
4471 |
+- /* read-high */ |
4472 |
+- __clear_bit(msr, msr_bitmap + 0x400 / f); |
4473 |
+- |
4474 |
+- if (type & MSR_TYPE_W) |
4475 |
+- /* write-high */ |
4476 |
+- __clear_bit(msr, msr_bitmap + 0xc00 / f); |
4477 |
+- |
4478 |
+- } |
4479 |
+-} |
4480 |
+- |
4481 |
+-static __always_inline void vmx_enable_intercept_for_msr(unsigned long *msr_bitmap, |
4482 |
+- u32 msr, int type) |
4483 |
+-{ |
4484 |
+- int f = sizeof(unsigned long); |
4485 |
+- |
4486 |
+- if (!cpu_has_vmx_msr_bitmap()) |
4487 |
+- return; |
4488 |
+- |
4489 |
+- if (static_branch_unlikely(&enable_evmcs)) |
4490 |
+- evmcs_touch_msr_bitmap(); |
4491 |
+- |
4492 |
+- /* |
4493 |
+- * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals |
4494 |
+- * have the write-low and read-high bitmap offsets the wrong way round. |
4495 |
+- * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. |
4496 |
+- */ |
4497 |
+- if (msr <= 0x1fff) { |
4498 |
+- if (type & MSR_TYPE_R) |
4499 |
+- /* read-low */ |
4500 |
+- __set_bit(msr, msr_bitmap + 0x000 / f); |
4501 |
+- |
4502 |
+- if (type & MSR_TYPE_W) |
4503 |
+- /* write-low */ |
4504 |
+- __set_bit(msr, msr_bitmap + 0x800 / f); |
4505 |
+- |
4506 |
+- } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
4507 |
+- msr &= 0x1fff; |
4508 |
+- if (type & MSR_TYPE_R) |
4509 |
+- /* read-high */ |
4510 |
+- __set_bit(msr, msr_bitmap + 0x400 / f); |
4511 |
+- |
4512 |
+- if (type & MSR_TYPE_W) |
4513 |
+- /* write-high */ |
4514 |
+- __set_bit(msr, msr_bitmap + 0xc00 / f); |
4515 |
+- |
4516 |
+- } |
4517 |
+-} |
4518 |
+- |
4519 |
+-static __always_inline void vmx_set_intercept_for_msr(unsigned long *msr_bitmap, |
4520 |
+- u32 msr, int type, bool value) |
4521 |
+-{ |
4522 |
+- if (value) |
4523 |
+- vmx_enable_intercept_for_msr(msr_bitmap, msr, type); |
4524 |
+- else |
4525 |
+- vmx_disable_intercept_for_msr(msr_bitmap, msr, type); |
4526 |
+-} |
4527 |
+- |
4528 |
+-static u8 vmx_msr_bitmap_mode(struct kvm_vcpu *vcpu) |
4529 |
+-{ |
4530 |
+- u8 mode = 0; |
4531 |
+- |
4532 |
+- if (cpu_has_secondary_exec_ctrls() && |
4533 |
+- (secondary_exec_controls_get(to_vmx(vcpu)) & |
4534 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE)) { |
4535 |
+- mode |= MSR_BITMAP_MODE_X2APIC; |
4536 |
+- if (enable_apicv && kvm_vcpu_apicv_active(vcpu)) |
4537 |
+- mode |= MSR_BITMAP_MODE_X2APIC_APICV; |
4538 |
+- } |
4539 |
+- |
4540 |
+- return mode; |
4541 |
+-} |
4542 |
+- |
4543 |
+-static void vmx_update_msr_bitmap_x2apic(unsigned long *msr_bitmap, |
4544 |
+- u8 mode) |
4545 |
+-{ |
4546 |
+- int msr; |
4547 |
+- |
4548 |
+- for (msr = 0x800; msr <= 0x8ff; msr += BITS_PER_LONG) { |
4549 |
+- unsigned word = msr / BITS_PER_LONG; |
4550 |
+- msr_bitmap[word] = (mode & MSR_BITMAP_MODE_X2APIC_APICV) ? 0 : ~0; |
4551 |
+- msr_bitmap[word + (0x800 / sizeof(long))] = ~0; |
4552 |
+- } |
4553 |
+- |
4554 |
+- if (mode & MSR_BITMAP_MODE_X2APIC) { |
4555 |
+- /* |
4556 |
+- * TPR reads and writes can be virtualized even if virtual interrupt |
4557 |
+- * delivery is not in use. |
4558 |
+- */ |
4559 |
+- vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_TASKPRI), MSR_TYPE_RW); |
4560 |
+- if (mode & MSR_BITMAP_MODE_X2APIC_APICV) { |
4561 |
+- vmx_enable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_TMCCT), MSR_TYPE_R); |
4562 |
+- vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_EOI), MSR_TYPE_W); |
4563 |
+- vmx_disable_intercept_for_msr(msr_bitmap, X2APIC_MSR(APIC_SELF_IPI), MSR_TYPE_W); |
4564 |
+- } |
4565 |
+- } |
4566 |
+-} |
4567 |
+- |
4568 |
+-void vmx_update_msr_bitmap(struct kvm_vcpu *vcpu) |
4569 |
+-{ |
4570 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4571 |
+- unsigned long *msr_bitmap = vmx->vmcs01.msr_bitmap; |
4572 |
+- u8 mode = vmx_msr_bitmap_mode(vcpu); |
4573 |
+- u8 changed = mode ^ vmx->msr_bitmap_mode; |
4574 |
+- |
4575 |
+- if (!changed) |
4576 |
+- return; |
4577 |
+- |
4578 |
+- if (changed & (MSR_BITMAP_MODE_X2APIC | MSR_BITMAP_MODE_X2APIC_APICV)) |
4579 |
+- vmx_update_msr_bitmap_x2apic(msr_bitmap, mode); |
4580 |
+- |
4581 |
+- vmx->msr_bitmap_mode = mode; |
4582 |
+-} |
4583 |
+- |
4584 |
+-void pt_update_intercept_for_msr(struct vcpu_vmx *vmx) |
4585 |
+-{ |
4586 |
+- unsigned long *msr_bitmap = vmx->vmcs01.msr_bitmap; |
4587 |
+- bool flag = !(vmx->pt_desc.guest.ctl & RTIT_CTL_TRACEEN); |
4588 |
+- u32 i; |
4589 |
+- |
4590 |
+- vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_STATUS, |
4591 |
+- MSR_TYPE_RW, flag); |
4592 |
+- vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_OUTPUT_BASE, |
4593 |
+- MSR_TYPE_RW, flag); |
4594 |
+- vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_OUTPUT_MASK, |
4595 |
+- MSR_TYPE_RW, flag); |
4596 |
+- vmx_set_intercept_for_msr(msr_bitmap, MSR_IA32_RTIT_CR3_MATCH, |
4597 |
+- MSR_TYPE_RW, flag); |
4598 |
+- for (i = 0; i < vmx->pt_desc.addr_range; i++) { |
4599 |
+- vmx_set_intercept_for_msr(msr_bitmap, |
4600 |
+- MSR_IA32_RTIT_ADDR0_A + i * 2, MSR_TYPE_RW, flag); |
4601 |
+- vmx_set_intercept_for_msr(msr_bitmap, |
4602 |
+- MSR_IA32_RTIT_ADDR0_B + i * 2, MSR_TYPE_RW, flag); |
4603 |
+- } |
4604 |
+-} |
4605 |
+- |
4606 |
+-static bool vmx_get_enable_apicv(struct kvm *kvm) |
4607 |
+-{ |
4608 |
+- return enable_apicv; |
4609 |
+-} |
4610 |
+- |
4611 |
+-static bool vmx_guest_apic_has_interrupt(struct kvm_vcpu *vcpu) |
4612 |
+-{ |
4613 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4614 |
+- void *vapic_page; |
4615 |
+- u32 vppr; |
4616 |
+- int rvi; |
4617 |
+- |
4618 |
+- if (WARN_ON_ONCE(!is_guest_mode(vcpu)) || |
4619 |
+- !nested_cpu_has_vid(get_vmcs12(vcpu)) || |
4620 |
+- WARN_ON_ONCE(!vmx->nested.virtual_apic_map.gfn)) |
4621 |
+- return false; |
4622 |
+- |
4623 |
+- rvi = vmx_get_rvi(); |
4624 |
+- |
4625 |
+- vapic_page = vmx->nested.virtual_apic_map.hva; |
4626 |
+- vppr = *((u32 *)(vapic_page + APIC_PROCPRI)); |
4627 |
+- |
4628 |
+- return ((rvi & 0xf0) > (vppr & 0xf0)); |
4629 |
+-} |
4630 |
+- |
4631 |
+-static inline bool kvm_vcpu_trigger_posted_interrupt(struct kvm_vcpu *vcpu, |
4632 |
+- bool nested) |
4633 |
+-{ |
4634 |
+-#ifdef CONFIG_SMP |
4635 |
+- int pi_vec = nested ? POSTED_INTR_NESTED_VECTOR : POSTED_INTR_VECTOR; |
4636 |
+- |
4637 |
+- if (vcpu->mode == IN_GUEST_MODE) { |
4638 |
+- /* |
4639 |
+- * The vector of interrupt to be delivered to vcpu had |
4640 |
+- * been set in PIR before this function. |
4641 |
+- * |
4642 |
+- * Following cases will be reached in this block, and |
4643 |
+- * we always send a notification event in all cases as |
4644 |
+- * explained below. |
4645 |
+- * |
4646 |
+- * Case 1: vcpu keeps in non-root mode. Sending a |
4647 |
+- * notification event posts the interrupt to vcpu. |
4648 |
+- * |
4649 |
+- * Case 2: vcpu exits to root mode and is still |
4650 |
+- * runnable. PIR will be synced to vIRR before the |
4651 |
+- * next vcpu entry. Sending a notification event in |
4652 |
+- * this case has no effect, as vcpu is not in root |
4653 |
+- * mode. |
4654 |
+- * |
4655 |
+- * Case 3: vcpu exits to root mode and is blocked. |
4656 |
+- * vcpu_block() has already synced PIR to vIRR and |
4657 |
+- * never blocks vcpu if vIRR is not cleared. Therefore, |
4658 |
+- * a blocked vcpu here does not wait for any requested |
4659 |
+- * interrupts in PIR, and sending a notification event |
4660 |
+- * which has no effect is safe here. |
4661 |
+- */ |
4662 |
+- |
4663 |
+- apic->send_IPI_mask(get_cpu_mask(vcpu->cpu), pi_vec); |
4664 |
+- return true; |
4665 |
+- } |
4666 |
+-#endif |
4667 |
+- return false; |
4668 |
+-} |
4669 |
+- |
4670 |
+-static int vmx_deliver_nested_posted_interrupt(struct kvm_vcpu *vcpu, |
4671 |
+- int vector) |
4672 |
+-{ |
4673 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4674 |
+- |
4675 |
+- if (is_guest_mode(vcpu) && |
4676 |
+- vector == vmx->nested.posted_intr_nv) { |
4677 |
+- /* |
4678 |
+- * If a posted intr is not recognized by hardware, |
4679 |
+- * we will accomplish it in the next vmentry. |
4680 |
+- */ |
4681 |
+- vmx->nested.pi_pending = true; |
4682 |
+- kvm_make_request(KVM_REQ_EVENT, vcpu); |
4683 |
+- /* the PIR and ON have been set by L1. */ |
4684 |
+- if (!kvm_vcpu_trigger_posted_interrupt(vcpu, true)) |
4685 |
+- kvm_vcpu_kick(vcpu); |
4686 |
+- return 0; |
4687 |
+- } |
4688 |
+- return -1; |
4689 |
+-} |
4690 |
+-/* |
4691 |
+- * Send interrupt to vcpu via posted interrupt way. |
4692 |
+- * 1. If target vcpu is running(non-root mode), send posted interrupt |
4693 |
+- * notification to vcpu and hardware will sync PIR to vIRR atomically. |
4694 |
+- * 2. If target vcpu isn't running(root mode), kick it to pick up the |
4695 |
+- * interrupt from PIR in next vmentry. |
4696 |
+- */ |
4697 |
+-static void vmx_deliver_posted_interrupt(struct kvm_vcpu *vcpu, int vector) |
4698 |
+-{ |
4699 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4700 |
+- int r; |
4701 |
+- |
4702 |
+- r = vmx_deliver_nested_posted_interrupt(vcpu, vector); |
4703 |
+- if (!r) |
4704 |
+- return; |
4705 |
+- |
4706 |
+- if (pi_test_and_set_pir(vector, &vmx->pi_desc)) |
4707 |
+- return; |
4708 |
+- |
4709 |
+- /* If a previous notification has sent the IPI, nothing to do. */ |
4710 |
+- if (pi_test_and_set_on(&vmx->pi_desc)) |
4711 |
+- return; |
4712 |
+- |
4713 |
+- if (!kvm_vcpu_trigger_posted_interrupt(vcpu, false)) |
4714 |
+- kvm_vcpu_kick(vcpu); |
4715 |
+-} |
4716 |
+- |
4717 |
+-/* |
4718 |
+- * Set up the vmcs's constant host-state fields, i.e., host-state fields that |
4719 |
+- * will not change in the lifetime of the guest. |
4720 |
+- * Note that host-state that does change is set elsewhere. E.g., host-state |
4721 |
+- * that is set differently for each CPU is set in vmx_vcpu_load(), not here. |
4722 |
+- */ |
4723 |
+-void vmx_set_constant_host_state(struct vcpu_vmx *vmx) |
4724 |
+-{ |
4725 |
+- u32 low32, high32; |
4726 |
+- unsigned long tmpl; |
4727 |
+- unsigned long cr0, cr3, cr4; |
4728 |
+- |
4729 |
+- cr0 = read_cr0(); |
4730 |
+- WARN_ON(cr0 & X86_CR0_TS); |
4731 |
+- vmcs_writel(HOST_CR0, cr0); /* 22.2.3 */ |
4732 |
+- |
4733 |
+- /* |
4734 |
+- * Save the most likely value for this task's CR3 in the VMCS. |
4735 |
+- * We can't use __get_current_cr3_fast() because we're not atomic. |
4736 |
+- */ |
4737 |
+- cr3 = __read_cr3(); |
4738 |
+- vmcs_writel(HOST_CR3, cr3); /* 22.2.3 FIXME: shadow tables */ |
4739 |
+- vmx->loaded_vmcs->host_state.cr3 = cr3; |
4740 |
+- |
4741 |
+- /* Save the most likely value for this task's CR4 in the VMCS. */ |
4742 |
+- cr4 = cr4_read_shadow(); |
4743 |
+- vmcs_writel(HOST_CR4, cr4); /* 22.2.3, 22.2.5 */ |
4744 |
+- vmx->loaded_vmcs->host_state.cr4 = cr4; |
4745 |
+- |
4746 |
+- vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS); /* 22.2.4 */ |
4747 |
+-#ifdef CONFIG_X86_64 |
4748 |
+- /* |
4749 |
+- * Load null selectors, so we can avoid reloading them in |
4750 |
+- * vmx_prepare_switch_to_host(), in case userspace uses |
4751 |
+- * the null selectors too (the expected case). |
4752 |
+- */ |
4753 |
+- vmcs_write16(HOST_DS_SELECTOR, 0); |
4754 |
+- vmcs_write16(HOST_ES_SELECTOR, 0); |
4755 |
+-#else |
4756 |
+- vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
4757 |
+- vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
4758 |
+-#endif |
4759 |
+- vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
4760 |
+- vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8); /* 22.2.4 */ |
4761 |
+- |
4762 |
+- vmcs_writel(HOST_IDTR_BASE, host_idt_base); /* 22.2.4 */ |
4763 |
+- |
4764 |
+- vmcs_writel(HOST_RIP, (unsigned long)vmx_vmexit); /* 22.2.5 */ |
4765 |
+- |
4766 |
+- rdmsr(MSR_IA32_SYSENTER_CS, low32, high32); |
4767 |
+- vmcs_write32(HOST_IA32_SYSENTER_CS, low32); |
4768 |
+- rdmsrl(MSR_IA32_SYSENTER_EIP, tmpl); |
4769 |
+- vmcs_writel(HOST_IA32_SYSENTER_EIP, tmpl); /* 22.2.3 */ |
4770 |
+- |
4771 |
+- if (vmcs_config.vmexit_ctrl & VM_EXIT_LOAD_IA32_PAT) { |
4772 |
+- rdmsr(MSR_IA32_CR_PAT, low32, high32); |
4773 |
+- vmcs_write64(HOST_IA32_PAT, low32 | ((u64) high32 << 32)); |
4774 |
+- } |
4775 |
+- |
4776 |
+- if (cpu_has_load_ia32_efer()) |
4777 |
+- vmcs_write64(HOST_IA32_EFER, host_efer); |
4778 |
+-} |
4779 |
+- |
4780 |
+-void set_cr4_guest_host_mask(struct vcpu_vmx *vmx) |
4781 |
+-{ |
4782 |
+- vmx->vcpu.arch.cr4_guest_owned_bits = KVM_CR4_GUEST_OWNED_BITS; |
4783 |
+- if (enable_ept) |
4784 |
+- vmx->vcpu.arch.cr4_guest_owned_bits |= X86_CR4_PGE; |
4785 |
+- if (is_guest_mode(&vmx->vcpu)) |
4786 |
+- vmx->vcpu.arch.cr4_guest_owned_bits &= |
4787 |
+- ~get_vmcs12(&vmx->vcpu)->cr4_guest_host_mask; |
4788 |
+- vmcs_writel(CR4_GUEST_HOST_MASK, ~vmx->vcpu.arch.cr4_guest_owned_bits); |
4789 |
+-} |
4790 |
+- |
4791 |
+-u32 vmx_pin_based_exec_ctrl(struct vcpu_vmx *vmx) |
4792 |
+-{ |
4793 |
+- u32 pin_based_exec_ctrl = vmcs_config.pin_based_exec_ctrl; |
4794 |
+- |
4795 |
+- if (!kvm_vcpu_apicv_active(&vmx->vcpu)) |
4796 |
+- pin_based_exec_ctrl &= ~PIN_BASED_POSTED_INTR; |
4797 |
+- |
4798 |
+- if (!enable_vnmi) |
4799 |
+- pin_based_exec_ctrl &= ~PIN_BASED_VIRTUAL_NMIS; |
4800 |
+- |
4801 |
+- if (!enable_preemption_timer) |
4802 |
+- pin_based_exec_ctrl &= ~PIN_BASED_VMX_PREEMPTION_TIMER; |
4803 |
+- |
4804 |
+- return pin_based_exec_ctrl; |
4805 |
+-} |
4806 |
+- |
4807 |
+-static void vmx_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu) |
4808 |
+-{ |
4809 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
4810 |
+- |
4811 |
+- pin_controls_set(vmx, vmx_pin_based_exec_ctrl(vmx)); |
4812 |
+- if (cpu_has_secondary_exec_ctrls()) { |
4813 |
+- if (kvm_vcpu_apicv_active(vcpu)) |
4814 |
+- secondary_exec_controls_setbit(vmx, |
4815 |
+- SECONDARY_EXEC_APIC_REGISTER_VIRT | |
4816 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
4817 |
+- else |
4818 |
+- secondary_exec_controls_clearbit(vmx, |
4819 |
+- SECONDARY_EXEC_APIC_REGISTER_VIRT | |
4820 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
4821 |
+- } |
4822 |
+- |
4823 |
+- if (cpu_has_vmx_msr_bitmap()) |
4824 |
+- vmx_update_msr_bitmap(vcpu); |
4825 |
+-} |
4826 |
+- |
4827 |
+-u32 vmx_exec_control(struct vcpu_vmx *vmx) |
4828 |
+-{ |
4829 |
+- u32 exec_control = vmcs_config.cpu_based_exec_ctrl; |
4830 |
+- |
4831 |
+- if (vmx->vcpu.arch.switch_db_regs & KVM_DEBUGREG_WONT_EXIT) |
4832 |
+- exec_control &= ~CPU_BASED_MOV_DR_EXITING; |
4833 |
+- |
4834 |
+- if (!cpu_need_tpr_shadow(&vmx->vcpu)) { |
4835 |
+- exec_control &= ~CPU_BASED_TPR_SHADOW; |
4836 |
+-#ifdef CONFIG_X86_64 |
4837 |
+- exec_control |= CPU_BASED_CR8_STORE_EXITING | |
4838 |
+- CPU_BASED_CR8_LOAD_EXITING; |
4839 |
+-#endif |
4840 |
+- } |
4841 |
+- if (!enable_ept) |
4842 |
+- exec_control |= CPU_BASED_CR3_STORE_EXITING | |
4843 |
+- CPU_BASED_CR3_LOAD_EXITING | |
4844 |
+- CPU_BASED_INVLPG_EXITING; |
4845 |
+- if (kvm_mwait_in_guest(vmx->vcpu.kvm)) |
4846 |
+- exec_control &= ~(CPU_BASED_MWAIT_EXITING | |
4847 |
+- CPU_BASED_MONITOR_EXITING); |
4848 |
+- if (kvm_hlt_in_guest(vmx->vcpu.kvm)) |
4849 |
+- exec_control &= ~CPU_BASED_HLT_EXITING; |
4850 |
+- return exec_control; |
4851 |
+-} |
4852 |
+- |
4853 |
+- |
4854 |
+-static void vmx_compute_secondary_exec_control(struct vcpu_vmx *vmx) |
4855 |
+-{ |
4856 |
+- struct kvm_vcpu *vcpu = &vmx->vcpu; |
4857 |
+- |
4858 |
+- u32 exec_control = vmcs_config.cpu_based_2nd_exec_ctrl; |
4859 |
+- |
4860 |
+- if (pt_mode == PT_MODE_SYSTEM) |
4861 |
+- exec_control &= ~(SECONDARY_EXEC_PT_USE_GPA | SECONDARY_EXEC_PT_CONCEAL_VMX); |
4862 |
+- if (!cpu_need_virtualize_apic_accesses(vcpu)) |
4863 |
+- exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; |
4864 |
+- if (vmx->vpid == 0) |
4865 |
+- exec_control &= ~SECONDARY_EXEC_ENABLE_VPID; |
4866 |
+- if (!enable_ept) { |
4867 |
+- exec_control &= ~SECONDARY_EXEC_ENABLE_EPT; |
4868 |
+- enable_unrestricted_guest = 0; |
4869 |
+- } |
4870 |
+- if (!enable_unrestricted_guest) |
4871 |
+- exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST; |
4872 |
+- if (kvm_pause_in_guest(vmx->vcpu.kvm)) |
4873 |
+- exec_control &= ~SECONDARY_EXEC_PAUSE_LOOP_EXITING; |
4874 |
+- if (!kvm_vcpu_apicv_active(vcpu)) |
4875 |
+- exec_control &= ~(SECONDARY_EXEC_APIC_REGISTER_VIRT | |
4876 |
+- SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY); |
4877 |
+- exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE; |
4878 |
+- |
4879 |
+- /* SECONDARY_EXEC_DESC is enabled/disabled on writes to CR4.UMIP, |
4880 |
+- * in vmx_set_cr4. */ |
4881 |
+- exec_control &= ~SECONDARY_EXEC_DESC; |
4882 |
+- |
4883 |
+- /* SECONDARY_EXEC_SHADOW_VMCS is enabled when L1 executes VMPTRLD |
4884 |
+- (handle_vmptrld). |
4885 |
+- We can NOT enable shadow_vmcs here because we don't have yet |
4886 |
+- a current VMCS12 |
4887 |
+- */ |
4888 |
+- exec_control &= ~SECONDARY_EXEC_SHADOW_VMCS; |
4889 |
+- |
4890 |
+- if (!enable_pml) |
4891 |
+- exec_control &= ~SECONDARY_EXEC_ENABLE_PML; |
4892 |
+- |
4893 |
+- if (vmx_xsaves_supported()) { |
4894 |
+- /* Exposing XSAVES only when XSAVE is exposed */ |
4895 |
+- bool xsaves_enabled = |
4896 |
+- guest_cpuid_has(vcpu, X86_FEATURE_XSAVE) && |
4897 |
+- guest_cpuid_has(vcpu, X86_FEATURE_XSAVES); |
4898 |
+- |
4899 |
+- vcpu->arch.xsaves_enabled = xsaves_enabled; |
4900 |
+- |
4901 |
+- if (!xsaves_enabled) |
4902 |
+- exec_control &= ~SECONDARY_EXEC_XSAVES; |
4903 |
+- |
4904 |
+- if (nested) { |
4905 |
+- if (xsaves_enabled) |
4906 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4907 |
+- SECONDARY_EXEC_XSAVES; |
4908 |
+- else |
4909 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4910 |
+- ~SECONDARY_EXEC_XSAVES; |
4911 |
+- } |
4912 |
+- } |
4913 |
+- |
4914 |
+- if (vmx_rdtscp_supported()) { |
4915 |
+- bool rdtscp_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDTSCP); |
4916 |
+- if (!rdtscp_enabled) |
4917 |
+- exec_control &= ~SECONDARY_EXEC_RDTSCP; |
4918 |
+- |
4919 |
+- if (nested) { |
4920 |
+- if (rdtscp_enabled) |
4921 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4922 |
+- SECONDARY_EXEC_RDTSCP; |
4923 |
+- else |
4924 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4925 |
+- ~SECONDARY_EXEC_RDTSCP; |
4926 |
+- } |
4927 |
+- } |
4928 |
+- |
4929 |
+- if (vmx_invpcid_supported()) { |
4930 |
+- /* Exposing INVPCID only when PCID is exposed */ |
4931 |
+- bool invpcid_enabled = |
4932 |
+- guest_cpuid_has(vcpu, X86_FEATURE_INVPCID) && |
4933 |
+- guest_cpuid_has(vcpu, X86_FEATURE_PCID); |
4934 |
+- |
4935 |
+- if (!invpcid_enabled) { |
4936 |
+- exec_control &= ~SECONDARY_EXEC_ENABLE_INVPCID; |
4937 |
+- guest_cpuid_clear(vcpu, X86_FEATURE_INVPCID); |
4938 |
+- } |
4939 |
+- |
4940 |
+- if (nested) { |
4941 |
+- if (invpcid_enabled) |
4942 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4943 |
+- SECONDARY_EXEC_ENABLE_INVPCID; |
4944 |
+- else |
4945 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4946 |
+- ~SECONDARY_EXEC_ENABLE_INVPCID; |
4947 |
+- } |
4948 |
+- } |
4949 |
+- |
4950 |
+- if (vmx_rdrand_supported()) { |
4951 |
+- bool rdrand_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDRAND); |
4952 |
+- if (rdrand_enabled) |
4953 |
+- exec_control &= ~SECONDARY_EXEC_RDRAND_EXITING; |
4954 |
+- |
4955 |
+- if (nested) { |
4956 |
+- if (rdrand_enabled) |
4957 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4958 |
+- SECONDARY_EXEC_RDRAND_EXITING; |
4959 |
+- else |
4960 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4961 |
+- ~SECONDARY_EXEC_RDRAND_EXITING; |
4962 |
+- } |
4963 |
+- } |
4964 |
+- |
4965 |
+- if (vmx_rdseed_supported()) { |
4966 |
+- bool rdseed_enabled = guest_cpuid_has(vcpu, X86_FEATURE_RDSEED); |
4967 |
+- if (rdseed_enabled) |
4968 |
+- exec_control &= ~SECONDARY_EXEC_RDSEED_EXITING; |
4969 |
+- |
4970 |
+- if (nested) { |
4971 |
+- if (rdseed_enabled) |
4972 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4973 |
+- SECONDARY_EXEC_RDSEED_EXITING; |
4974 |
+- else |
4975 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4976 |
+- ~SECONDARY_EXEC_RDSEED_EXITING; |
4977 |
+- } |
4978 |
+- } |
4979 |
+- |
4980 |
+- if (vmx_waitpkg_supported()) { |
4981 |
+- bool waitpkg_enabled = |
4982 |
+- guest_cpuid_has(vcpu, X86_FEATURE_WAITPKG); |
4983 |
+- |
4984 |
+- if (!waitpkg_enabled) |
4985 |
+- exec_control &= ~SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
4986 |
+- |
4987 |
+- if (nested) { |
4988 |
+- if (waitpkg_enabled) |
4989 |
+- vmx->nested.msrs.secondary_ctls_high |= |
4990 |
+- SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
4991 |
+- else |
4992 |
+- vmx->nested.msrs.secondary_ctls_high &= |
4993 |
+- ~SECONDARY_EXEC_ENABLE_USR_WAIT_PAUSE; |
4994 |
+- } |
4995 |
+- } |
4996 |
+- |
4997 |
+- vmx->secondary_exec_control = exec_control; |
4998 |
+-} |
4999 |
+- |
5000 |
+-static void ept_set_mmio_spte_mask(void) |
5001 |
+-{ |
5002 |
+- /* |
5003 |
+- * EPT Misconfigurations can be generated if the value of bits 2:0 |
5004 |
+- * of an EPT paging-structure entry is 110b (write/execute). |
5005 |
+- */ |
5006 |
+- kvm_mmu_set_mmio_spte_mask(VMX_EPT_RWX_MASK, |
5007 |
+- VMX_EPT_MISCONFIG_WX_VALUE, 0); |
5008 |
+-} |
5009 |
+- |
5010 |
+-#define VMX_XSS_EXIT_BITMAP 0 |
5011 |
+- |
5012 |
+-/* |
5013 |
+- * Noting that the initialization of Guest-state Area of VMCS is in |
5014 |
+- * vmx_vcpu_reset(). |
5015 |
+- */ |
5016 |
+-static void init_vmcs(struct vcpu_vmx *vmx) |
5017 |
+-{ |
5018 |
+- if (nested) |
5019 |
+- nested_vmx_set_vmcs_shadowing_bitmap(); |
5020 |
+- |
5021 |
+- if (cpu_has_vmx_msr_bitmap()) |
5022 |
+- vmcs_write64(MSR_BITMAP, __pa(vmx->vmcs01.msr_bitmap)); |
5023 |
+- |
5024 |
+- vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */ |
5025 |
+- |
5026 |
+- /* Control */ |
5027 |
+- pin_controls_set(vmx, vmx_pin_based_exec_ctrl(vmx)); |
5028 |
+- |
5029 |
+- exec_controls_set(vmx, vmx_exec_control(vmx)); |
5030 |
+- |
5031 |
+- if (cpu_has_secondary_exec_ctrls()) { |
5032 |
+- vmx_compute_secondary_exec_control(vmx); |
5033 |
+- secondary_exec_controls_set(vmx, vmx->secondary_exec_control); |
5034 |
+- } |
5035 |
+- |
5036 |
+- if (kvm_vcpu_apicv_active(&vmx->vcpu)) { |
5037 |
+- vmcs_write64(EOI_EXIT_BITMAP0, 0); |
5038 |
+- vmcs_write64(EOI_EXIT_BITMAP1, 0); |
5039 |
+- vmcs_write64(EOI_EXIT_BITMAP2, 0); |
5040 |
+- vmcs_write64(EOI_EXIT_BITMAP3, 0); |
5041 |
+- |
5042 |
+- vmcs_write16(GUEST_INTR_STATUS, 0); |
5043 |
+- |
5044 |
+- vmcs_write16(POSTED_INTR_NV, POSTED_INTR_VECTOR); |
5045 |
+- vmcs_write64(POSTED_INTR_DESC_ADDR, __pa((&vmx->pi_desc))); |
5046 |
+- } |
5047 |
+- |
5048 |
+- if (!kvm_pause_in_guest(vmx->vcpu.kvm)) { |
5049 |
+- vmcs_write32(PLE_GAP, ple_gap); |
5050 |
+- vmx->ple_window = ple_window; |
5051 |
+- vmx->ple_window_dirty = true; |
5052 |
+- } |
5053 |
+- |
5054 |
+- vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0); |
5055 |
+- vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0); |
5056 |
+- vmcs_write32(CR3_TARGET_COUNT, 0); /* 22.2.1 */ |
5057 |
+- |
5058 |
+- vmcs_write16(HOST_FS_SELECTOR, 0); /* 22.2.4 */ |
5059 |
+- vmcs_write16(HOST_GS_SELECTOR, 0); /* 22.2.4 */ |
5060 |
+- vmx_set_constant_host_state(vmx); |
5061 |
+- vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */ |
5062 |
+- vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */ |
5063 |
+- |
5064 |
+- if (cpu_has_vmx_vmfunc()) |
5065 |
+- vmcs_write64(VM_FUNCTION_CONTROL, 0); |
5066 |
+- |
5067 |
+- vmcs_write32(VM_EXIT_MSR_STORE_COUNT, 0); |
5068 |
+- vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, 0); |
5069 |
+- vmcs_write64(VM_EXIT_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.host.val)); |
5070 |
+- vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, 0); |
5071 |
+- vmcs_write64(VM_ENTRY_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.guest.val)); |
5072 |
+- |
5073 |
+- if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) |
5074 |
+- vmcs_write64(GUEST_IA32_PAT, vmx->vcpu.arch.pat); |
5075 |
+- |
5076 |
+- vm_exit_controls_set(vmx, vmx_vmexit_ctrl()); |
5077 |
+- |
5078 |
+- /* 22.2.1, 20.8.1 */ |
5079 |
+- vm_entry_controls_set(vmx, vmx_vmentry_ctrl()); |
5080 |
+- |
5081 |
+- vmx->vcpu.arch.cr0_guest_owned_bits = X86_CR0_TS; |
5082 |
+- vmcs_writel(CR0_GUEST_HOST_MASK, ~X86_CR0_TS); |
5083 |
+- |
5084 |
+- set_cr4_guest_host_mask(vmx); |
5085 |
+- |
5086 |
+- if (vmx->vpid != 0) |
5087 |
+- vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid); |
5088 |
+- |
5089 |
+- if (vmx_xsaves_supported()) |
5090 |
+- vmcs_write64(XSS_EXIT_BITMAP, VMX_XSS_EXIT_BITMAP); |
5091 |
+- |
5092 |
+- if (enable_pml) { |
5093 |
+- vmcs_write64(PML_ADDRESS, page_to_phys(vmx->pml_pg)); |
5094 |
+- vmcs_write16(GUEST_PML_INDEX, PML_ENTITY_NUM - 1); |
5095 |
+- } |
5096 |
+- |
5097 |
+- if (cpu_has_vmx_encls_vmexit()) |
5098 |
+- vmcs_write64(ENCLS_EXITING_BITMAP, -1ull); |
5099 |
+- |
5100 |
+- if (pt_mode == PT_MODE_HOST_GUEST) { |
5101 |
+- memset(&vmx->pt_desc, 0, sizeof(vmx->pt_desc)); |
5102 |
+- /* Bit[6~0] are forced to 1, writes are ignored. */ |
5103 |
+- vmx->pt_desc.guest.output_mask = 0x7F; |
5104 |
+- vmcs_write64(GUEST_IA32_RTIT_CTL, 0); |
5105 |
+- } |
5106 |
+-} |
5107 |
+- |
5108 |
+-static void vmx_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event) |
5109 |
+-{ |
5110 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5111 |
+- struct msr_data apic_base_msr; |
5112 |
+- u64 cr0; |
5113 |
+- |
5114 |
+- vmx->rmode.vm86_active = 0; |
5115 |
+- vmx->spec_ctrl = 0; |
5116 |
+- |
5117 |
+- vmx->msr_ia32_umwait_control = 0; |
5118 |
+- |
5119 |
+- vcpu->arch.microcode_version = 0x100000000ULL; |
5120 |
+- vmx->vcpu.arch.regs[VCPU_REGS_RDX] = get_rdx_init_val(); |
5121 |
+- vmx->hv_deadline_tsc = -1; |
5122 |
+- kvm_set_cr8(vcpu, 0); |
5123 |
+- |
5124 |
+- if (!init_event) { |
5125 |
+- apic_base_msr.data = APIC_DEFAULT_PHYS_BASE | |
5126 |
+- MSR_IA32_APICBASE_ENABLE; |
5127 |
+- if (kvm_vcpu_is_reset_bsp(vcpu)) |
5128 |
+- apic_base_msr.data |= MSR_IA32_APICBASE_BSP; |
5129 |
+- apic_base_msr.host_initiated = true; |
5130 |
+- kvm_set_apic_base(vcpu, &apic_base_msr); |
5131 |
+- } |
5132 |
+- |
5133 |
+- vmx_segment_cache_clear(vmx); |
5134 |
+- |
5135 |
+- seg_setup(VCPU_SREG_CS); |
5136 |
+- vmcs_write16(GUEST_CS_SELECTOR, 0xf000); |
5137 |
+- vmcs_writel(GUEST_CS_BASE, 0xffff0000ul); |
5138 |
+- |
5139 |
+- seg_setup(VCPU_SREG_DS); |
5140 |
+- seg_setup(VCPU_SREG_ES); |
5141 |
+- seg_setup(VCPU_SREG_FS); |
5142 |
+- seg_setup(VCPU_SREG_GS); |
5143 |
+- seg_setup(VCPU_SREG_SS); |
5144 |
+- |
5145 |
+- vmcs_write16(GUEST_TR_SELECTOR, 0); |
5146 |
+- vmcs_writel(GUEST_TR_BASE, 0); |
5147 |
+- vmcs_write32(GUEST_TR_LIMIT, 0xffff); |
5148 |
+- vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
5149 |
+- |
5150 |
+- vmcs_write16(GUEST_LDTR_SELECTOR, 0); |
5151 |
+- vmcs_writel(GUEST_LDTR_BASE, 0); |
5152 |
+- vmcs_write32(GUEST_LDTR_LIMIT, 0xffff); |
5153 |
+- vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082); |
5154 |
+- |
5155 |
+- if (!init_event) { |
5156 |
+- vmcs_write32(GUEST_SYSENTER_CS, 0); |
5157 |
+- vmcs_writel(GUEST_SYSENTER_ESP, 0); |
5158 |
+- vmcs_writel(GUEST_SYSENTER_EIP, 0); |
5159 |
+- vmcs_write64(GUEST_IA32_DEBUGCTL, 0); |
5160 |
+- } |
5161 |
+- |
5162 |
+- kvm_set_rflags(vcpu, X86_EFLAGS_FIXED); |
5163 |
+- kvm_rip_write(vcpu, 0xfff0); |
5164 |
+- |
5165 |
+- vmcs_writel(GUEST_GDTR_BASE, 0); |
5166 |
+- vmcs_write32(GUEST_GDTR_LIMIT, 0xffff); |
5167 |
+- |
5168 |
+- vmcs_writel(GUEST_IDTR_BASE, 0); |
5169 |
+- vmcs_write32(GUEST_IDTR_LIMIT, 0xffff); |
5170 |
+- |
5171 |
+- vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE); |
5172 |
+- vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0); |
5173 |
+- vmcs_writel(GUEST_PENDING_DBG_EXCEPTIONS, 0); |
5174 |
+- if (kvm_mpx_supported()) |
5175 |
+- vmcs_write64(GUEST_BNDCFGS, 0); |
5176 |
+- |
5177 |
+- setup_msrs(vmx); |
5178 |
+- |
5179 |
+- vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); /* 22.2.1 */ |
5180 |
+- |
5181 |
+- if (cpu_has_vmx_tpr_shadow() && !init_event) { |
5182 |
+- vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, 0); |
5183 |
+- if (cpu_need_tpr_shadow(vcpu)) |
5184 |
+- vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, |
5185 |
+- __pa(vcpu->arch.apic->regs)); |
5186 |
+- vmcs_write32(TPR_THRESHOLD, 0); |
5187 |
+- } |
5188 |
+- |
5189 |
+- kvm_make_request(KVM_REQ_APIC_PAGE_RELOAD, vcpu); |
5190 |
+- |
5191 |
+- cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET; |
5192 |
+- vmx->vcpu.arch.cr0 = cr0; |
5193 |
+- vmx_set_cr0(vcpu, cr0); /* enter rmode */ |
5194 |
+- vmx_set_cr4(vcpu, 0); |
5195 |
+- vmx_set_efer(vcpu, 0); |
5196 |
+- |
5197 |
+- update_exception_bitmap(vcpu); |
5198 |
+- |
5199 |
+- vpid_sync_context(vmx->vpid); |
5200 |
+- if (init_event) |
5201 |
+- vmx_clear_hlt(vcpu); |
5202 |
+-} |
5203 |
+- |
5204 |
+-static void enable_irq_window(struct kvm_vcpu *vcpu) |
5205 |
+-{ |
5206 |
+- exec_controls_setbit(to_vmx(vcpu), CPU_BASED_INTR_WINDOW_EXITING); |
5207 |
+-} |
5208 |
+- |
5209 |
+-static void enable_nmi_window(struct kvm_vcpu *vcpu) |
5210 |
+-{ |
5211 |
+- if (!enable_vnmi || |
5212 |
+- vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) { |
5213 |
+- enable_irq_window(vcpu); |
5214 |
+- return; |
5215 |
+- } |
5216 |
+- |
5217 |
+- exec_controls_setbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING); |
5218 |
+-} |
5219 |
+- |
5220 |
+-static void vmx_inject_irq(struct kvm_vcpu *vcpu) |
5221 |
+-{ |
5222 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5223 |
+- uint32_t intr; |
5224 |
+- int irq = vcpu->arch.interrupt.nr; |
5225 |
+- |
5226 |
+- trace_kvm_inj_virq(irq); |
5227 |
+- |
5228 |
+- ++vcpu->stat.irq_injections; |
5229 |
+- if (vmx->rmode.vm86_active) { |
5230 |
+- int inc_eip = 0; |
5231 |
+- if (vcpu->arch.interrupt.soft) |
5232 |
+- inc_eip = vcpu->arch.event_exit_inst_len; |
5233 |
+- kvm_inject_realmode_interrupt(vcpu, irq, inc_eip); |
5234 |
+- return; |
5235 |
+- } |
5236 |
+- intr = irq | INTR_INFO_VALID_MASK; |
5237 |
+- if (vcpu->arch.interrupt.soft) { |
5238 |
+- intr |= INTR_TYPE_SOFT_INTR; |
5239 |
+- vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, |
5240 |
+- vmx->vcpu.arch.event_exit_inst_len); |
5241 |
+- } else |
5242 |
+- intr |= INTR_TYPE_EXT_INTR; |
5243 |
+- vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr); |
5244 |
+- |
5245 |
+- vmx_clear_hlt(vcpu); |
5246 |
+-} |
5247 |
+- |
5248 |
+-static void vmx_inject_nmi(struct kvm_vcpu *vcpu) |
5249 |
+-{ |
5250 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5251 |
+- |
5252 |
+- if (!enable_vnmi) { |
5253 |
+- /* |
5254 |
+- * Tracking the NMI-blocked state in software is built upon |
5255 |
+- * finding the next open IRQ window. This, in turn, depends on |
5256 |
+- * well-behaving guests: They have to keep IRQs disabled at |
5257 |
+- * least as long as the NMI handler runs. Otherwise we may |
5258 |
+- * cause NMI nesting, maybe breaking the guest. But as this is |
5259 |
+- * highly unlikely, we can live with the residual risk. |
5260 |
+- */ |
5261 |
+- vmx->loaded_vmcs->soft_vnmi_blocked = 1; |
5262 |
+- vmx->loaded_vmcs->vnmi_blocked_time = 0; |
5263 |
+- } |
5264 |
+- |
5265 |
+- ++vcpu->stat.nmi_injections; |
5266 |
+- vmx->loaded_vmcs->nmi_known_unmasked = false; |
5267 |
+- |
5268 |
+- if (vmx->rmode.vm86_active) { |
5269 |
+- kvm_inject_realmode_interrupt(vcpu, NMI_VECTOR, 0); |
5270 |
+- return; |
5271 |
+- } |
5272 |
+- |
5273 |
+- vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, |
5274 |
+- INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR); |
5275 |
+- |
5276 |
+- vmx_clear_hlt(vcpu); |
5277 |
+-} |
5278 |
+- |
5279 |
+-bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu) |
5280 |
+-{ |
5281 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5282 |
+- bool masked; |
5283 |
+- |
5284 |
+- if (!enable_vnmi) |
5285 |
+- return vmx->loaded_vmcs->soft_vnmi_blocked; |
5286 |
+- if (vmx->loaded_vmcs->nmi_known_unmasked) |
5287 |
+- return false; |
5288 |
+- masked = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_NMI; |
5289 |
+- vmx->loaded_vmcs->nmi_known_unmasked = !masked; |
5290 |
+- return masked; |
5291 |
+-} |
5292 |
+- |
5293 |
+-void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked) |
5294 |
+-{ |
5295 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5296 |
+- |
5297 |
+- if (!enable_vnmi) { |
5298 |
+- if (vmx->loaded_vmcs->soft_vnmi_blocked != masked) { |
5299 |
+- vmx->loaded_vmcs->soft_vnmi_blocked = masked; |
5300 |
+- vmx->loaded_vmcs->vnmi_blocked_time = 0; |
5301 |
+- } |
5302 |
+- } else { |
5303 |
+- vmx->loaded_vmcs->nmi_known_unmasked = !masked; |
5304 |
+- if (masked) |
5305 |
+- vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
5306 |
+- GUEST_INTR_STATE_NMI); |
5307 |
+- else |
5308 |
+- vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO, |
5309 |
+- GUEST_INTR_STATE_NMI); |
5310 |
+- } |
5311 |
+-} |
5312 |
+- |
5313 |
+-static int vmx_nmi_allowed(struct kvm_vcpu *vcpu) |
5314 |
+-{ |
5315 |
+- if (to_vmx(vcpu)->nested.nested_run_pending) |
5316 |
+- return 0; |
5317 |
+- |
5318 |
+- if (!enable_vnmi && |
5319 |
+- to_vmx(vcpu)->loaded_vmcs->soft_vnmi_blocked) |
5320 |
+- return 0; |
5321 |
+- |
5322 |
+- return !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
5323 |
+- (GUEST_INTR_STATE_MOV_SS | GUEST_INTR_STATE_STI |
5324 |
+- | GUEST_INTR_STATE_NMI)); |
5325 |
+-} |
5326 |
+- |
5327 |
+-static int vmx_interrupt_allowed(struct kvm_vcpu *vcpu) |
5328 |
+-{ |
5329 |
+- return (!to_vmx(vcpu)->nested.nested_run_pending && |
5330 |
+- vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF) && |
5331 |
+- !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
5332 |
+- (GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS)); |
5333 |
+-} |
5334 |
+- |
5335 |
+-static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr) |
5336 |
+-{ |
5337 |
+- int ret; |
5338 |
+- |
5339 |
+- if (enable_unrestricted_guest) |
5340 |
+- return 0; |
5341 |
+- |
5342 |
+- ret = x86_set_memory_region(kvm, TSS_PRIVATE_MEMSLOT, addr, |
5343 |
+- PAGE_SIZE * 3); |
5344 |
+- if (ret) |
5345 |
+- return ret; |
5346 |
+- to_kvm_vmx(kvm)->tss_addr = addr; |
5347 |
+- return init_rmode_tss(kvm); |
5348 |
+-} |
5349 |
+- |
5350 |
+-static int vmx_set_identity_map_addr(struct kvm *kvm, u64 ident_addr) |
5351 |
+-{ |
5352 |
+- to_kvm_vmx(kvm)->ept_identity_map_addr = ident_addr; |
5353 |
+- return 0; |
5354 |
+-} |
5355 |
+- |
5356 |
+-static bool rmode_exception(struct kvm_vcpu *vcpu, int vec) |
5357 |
+-{ |
5358 |
+- switch (vec) { |
5359 |
+- case BP_VECTOR: |
5360 |
+- /* |
5361 |
+- * Update instruction length as we may reinject the exception |
5362 |
+- * from user space while in guest debugging mode. |
5363 |
+- */ |
5364 |
+- to_vmx(vcpu)->vcpu.arch.event_exit_inst_len = |
5365 |
+- vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
5366 |
+- if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) |
5367 |
+- return false; |
5368 |
+- /* fall through */ |
5369 |
+- case DB_VECTOR: |
5370 |
+- if (vcpu->guest_debug & |
5371 |
+- (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP)) |
5372 |
+- return false; |
5373 |
+- /* fall through */ |
5374 |
+- case DE_VECTOR: |
5375 |
+- case OF_VECTOR: |
5376 |
+- case BR_VECTOR: |
5377 |
+- case UD_VECTOR: |
5378 |
+- case DF_VECTOR: |
5379 |
+- case SS_VECTOR: |
5380 |
+- case GP_VECTOR: |
5381 |
+- case MF_VECTOR: |
5382 |
+- return true; |
5383 |
+- break; |
5384 |
+- } |
5385 |
+- return false; |
5386 |
+-} |
5387 |
+- |
5388 |
+-static int handle_rmode_exception(struct kvm_vcpu *vcpu, |
5389 |
+- int vec, u32 err_code) |
5390 |
+-{ |
5391 |
+- /* |
5392 |
+- * Instruction with address size override prefix opcode 0x67 |
5393 |
+- * Cause the #SS fault with 0 error code in VM86 mode. |
5394 |
+- */ |
5395 |
+- if (((vec == GP_VECTOR) || (vec == SS_VECTOR)) && err_code == 0) { |
5396 |
+- if (kvm_emulate_instruction(vcpu, 0)) { |
5397 |
+- if (vcpu->arch.halt_request) { |
5398 |
+- vcpu->arch.halt_request = 0; |
5399 |
+- return kvm_vcpu_halt(vcpu); |
5400 |
+- } |
5401 |
+- return 1; |
5402 |
+- } |
5403 |
+- return 0; |
5404 |
+- } |
5405 |
+- |
5406 |
+- /* |
5407 |
+- * Forward all other exceptions that are valid in real mode. |
5408 |
+- * FIXME: Breaks guest debugging in real mode, needs to be fixed with |
5409 |
+- * the required debugging infrastructure rework. |
5410 |
+- */ |
5411 |
+- kvm_queue_exception(vcpu, vec); |
5412 |
+- return 1; |
5413 |
+-} |
5414 |
+- |
5415 |
+-/* |
5416 |
+- * Trigger machine check on the host. We assume all the MSRs are already set up |
5417 |
+- * by the CPU and that we still run on the same CPU as the MCE occurred on. |
5418 |
+- * We pass a fake environment to the machine check handler because we want |
5419 |
+- * the guest to be always treated like user space, no matter what context |
5420 |
+- * it used internally. |
5421 |
+- */ |
5422 |
+-static void kvm_machine_check(void) |
5423 |
+-{ |
5424 |
+-#if defined(CONFIG_X86_MCE) && defined(CONFIG_X86_64) |
5425 |
+- struct pt_regs regs = { |
5426 |
+- .cs = 3, /* Fake ring 3 no matter what the guest ran on */ |
5427 |
+- .flags = X86_EFLAGS_IF, |
5428 |
+- }; |
5429 |
+- |
5430 |
+- do_machine_check(®s, 0); |
5431 |
+-#endif |
5432 |
+-} |
5433 |
+- |
5434 |
+-static int handle_machine_check(struct kvm_vcpu *vcpu) |
5435 |
+-{ |
5436 |
+- /* handled by vmx_vcpu_run() */ |
5437 |
+- return 1; |
5438 |
+-} |
5439 |
+- |
5440 |
+-static int handle_exception_nmi(struct kvm_vcpu *vcpu) |
5441 |
+-{ |
5442 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5443 |
+- struct kvm_run *kvm_run = vcpu->run; |
5444 |
+- u32 intr_info, ex_no, error_code; |
5445 |
+- unsigned long cr2, rip, dr6; |
5446 |
+- u32 vect_info; |
5447 |
+- |
5448 |
+- vect_info = vmx->idt_vectoring_info; |
5449 |
+- intr_info = vmx->exit_intr_info; |
5450 |
+- |
5451 |
+- if (is_machine_check(intr_info) || is_nmi(intr_info)) |
5452 |
+- return 1; /* handled by handle_exception_nmi_irqoff() */ |
5453 |
+- |
5454 |
+- if (is_invalid_opcode(intr_info)) |
5455 |
+- return handle_ud(vcpu); |
5456 |
+- |
5457 |
+- error_code = 0; |
5458 |
+- if (intr_info & INTR_INFO_DELIVER_CODE_MASK) |
5459 |
+- error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE); |
5460 |
+- |
5461 |
+- if (!vmx->rmode.vm86_active && is_gp_fault(intr_info)) { |
5462 |
+- WARN_ON_ONCE(!enable_vmware_backdoor); |
5463 |
+- |
5464 |
+- /* |
5465 |
+- * VMware backdoor emulation on #GP interception only handles |
5466 |
+- * IN{S}, OUT{S}, and RDPMC, none of which generate a non-zero |
5467 |
+- * error code on #GP. |
5468 |
+- */ |
5469 |
+- if (error_code) { |
5470 |
+- kvm_queue_exception_e(vcpu, GP_VECTOR, error_code); |
5471 |
+- return 1; |
5472 |
+- } |
5473 |
+- return kvm_emulate_instruction(vcpu, EMULTYPE_VMWARE_GP); |
5474 |
+- } |
5475 |
+- |
5476 |
+- /* |
5477 |
+- * The #PF with PFEC.RSVD = 1 indicates the guest is accessing |
5478 |
+- * MMIO, it is better to report an internal error. |
5479 |
+- * See the comments in vmx_handle_exit. |
5480 |
+- */ |
5481 |
+- if ((vect_info & VECTORING_INFO_VALID_MASK) && |
5482 |
+- !(is_page_fault(intr_info) && !(error_code & PFERR_RSVD_MASK))) { |
5483 |
+- vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
5484 |
+- vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_SIMUL_EX; |
5485 |
+- vcpu->run->internal.ndata = 3; |
5486 |
+- vcpu->run->internal.data[0] = vect_info; |
5487 |
+- vcpu->run->internal.data[1] = intr_info; |
5488 |
+- vcpu->run->internal.data[2] = error_code; |
5489 |
+- return 0; |
5490 |
+- } |
5491 |
+- |
5492 |
+- if (is_page_fault(intr_info)) { |
5493 |
+- cr2 = vmcs_readl(EXIT_QUALIFICATION); |
5494 |
+- /* EPT won't cause page fault directly */ |
5495 |
+- WARN_ON_ONCE(!vcpu->arch.apf.host_apf_reason && enable_ept); |
5496 |
+- return kvm_handle_page_fault(vcpu, error_code, cr2, NULL, 0); |
5497 |
+- } |
5498 |
+- |
5499 |
+- ex_no = intr_info & INTR_INFO_VECTOR_MASK; |
5500 |
+- |
5501 |
+- if (vmx->rmode.vm86_active && rmode_exception(vcpu, ex_no)) |
5502 |
+- return handle_rmode_exception(vcpu, ex_no, error_code); |
5503 |
+- |
5504 |
+- switch (ex_no) { |
5505 |
+- case AC_VECTOR: |
5506 |
+- kvm_queue_exception_e(vcpu, AC_VECTOR, error_code); |
5507 |
+- return 1; |
5508 |
+- case DB_VECTOR: |
5509 |
+- dr6 = vmcs_readl(EXIT_QUALIFICATION); |
5510 |
+- if (!(vcpu->guest_debug & |
5511 |
+- (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP))) { |
5512 |
+- vcpu->arch.dr6 &= ~DR_TRAP_BITS; |
5513 |
+- vcpu->arch.dr6 |= dr6 | DR6_RTM; |
5514 |
+- if (is_icebp(intr_info)) |
5515 |
+- WARN_ON(!skip_emulated_instruction(vcpu)); |
5516 |
+- |
5517 |
+- kvm_queue_exception(vcpu, DB_VECTOR); |
5518 |
+- return 1; |
5519 |
+- } |
5520 |
+- kvm_run->debug.arch.dr6 = dr6 | DR6_FIXED_1; |
5521 |
+- kvm_run->debug.arch.dr7 = vmcs_readl(GUEST_DR7); |
5522 |
+- /* fall through */ |
5523 |
+- case BP_VECTOR: |
5524 |
+- /* |
5525 |
+- * Update instruction length as we may reinject #BP from |
5526 |
+- * user space while in guest debugging mode. Reading it for |
5527 |
+- * #DB as well causes no harm, it is not used in that case. |
5528 |
+- */ |
5529 |
+- vmx->vcpu.arch.event_exit_inst_len = |
5530 |
+- vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
5531 |
+- kvm_run->exit_reason = KVM_EXIT_DEBUG; |
5532 |
+- rip = kvm_rip_read(vcpu); |
5533 |
+- kvm_run->debug.arch.pc = vmcs_readl(GUEST_CS_BASE) + rip; |
5534 |
+- kvm_run->debug.arch.exception = ex_no; |
5535 |
+- break; |
5536 |
+- default: |
5537 |
+- kvm_run->exit_reason = KVM_EXIT_EXCEPTION; |
5538 |
+- kvm_run->ex.exception = ex_no; |
5539 |
+- kvm_run->ex.error_code = error_code; |
5540 |
+- break; |
5541 |
+- } |
5542 |
+- return 0; |
5543 |
+-} |
5544 |
+- |
5545 |
+-static __always_inline int handle_external_interrupt(struct kvm_vcpu *vcpu) |
5546 |
+-{ |
5547 |
+- ++vcpu->stat.irq_exits; |
5548 |
+- return 1; |
5549 |
+-} |
5550 |
+- |
5551 |
+-static int handle_triple_fault(struct kvm_vcpu *vcpu) |
5552 |
+-{ |
5553 |
+- vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; |
5554 |
+- vcpu->mmio_needed = 0; |
5555 |
+- return 0; |
5556 |
+-} |
5557 |
+- |
5558 |
+-static int handle_io(struct kvm_vcpu *vcpu) |
5559 |
+-{ |
5560 |
+- unsigned long exit_qualification; |
5561 |
+- int size, in, string; |
5562 |
+- unsigned port; |
5563 |
+- |
5564 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5565 |
+- string = (exit_qualification & 16) != 0; |
5566 |
+- |
5567 |
+- ++vcpu->stat.io_exits; |
5568 |
+- |
5569 |
+- if (string) |
5570 |
+- return kvm_emulate_instruction(vcpu, 0); |
5571 |
+- |
5572 |
+- port = exit_qualification >> 16; |
5573 |
+- size = (exit_qualification & 7) + 1; |
5574 |
+- in = (exit_qualification & 8) != 0; |
5575 |
+- |
5576 |
+- return kvm_fast_pio(vcpu, size, port, in); |
5577 |
+-} |
5578 |
+- |
5579 |
+-static void |
5580 |
+-vmx_patch_hypercall(struct kvm_vcpu *vcpu, unsigned char *hypercall) |
5581 |
+-{ |
5582 |
+- /* |
5583 |
+- * Patch in the VMCALL instruction: |
5584 |
+- */ |
5585 |
+- hypercall[0] = 0x0f; |
5586 |
+- hypercall[1] = 0x01; |
5587 |
+- hypercall[2] = 0xc1; |
5588 |
+-} |
5589 |
+- |
5590 |
+-/* called to set cr0 as appropriate for a mov-to-cr0 exit. */ |
5591 |
+-static int handle_set_cr0(struct kvm_vcpu *vcpu, unsigned long val) |
5592 |
+-{ |
5593 |
+- if (is_guest_mode(vcpu)) { |
5594 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
5595 |
+- unsigned long orig_val = val; |
5596 |
+- |
5597 |
+- /* |
5598 |
+- * We get here when L2 changed cr0 in a way that did not change |
5599 |
+- * any of L1's shadowed bits (see nested_vmx_exit_handled_cr), |
5600 |
+- * but did change L0 shadowed bits. So we first calculate the |
5601 |
+- * effective cr0 value that L1 would like to write into the |
5602 |
+- * hardware. It consists of the L2-owned bits from the new |
5603 |
+- * value combined with the L1-owned bits from L1's guest_cr0. |
5604 |
+- */ |
5605 |
+- val = (val & ~vmcs12->cr0_guest_host_mask) | |
5606 |
+- (vmcs12->guest_cr0 & vmcs12->cr0_guest_host_mask); |
5607 |
+- |
5608 |
+- if (!nested_guest_cr0_valid(vcpu, val)) |
5609 |
+- return 1; |
5610 |
+- |
5611 |
+- if (kvm_set_cr0(vcpu, val)) |
5612 |
+- return 1; |
5613 |
+- vmcs_writel(CR0_READ_SHADOW, orig_val); |
5614 |
+- return 0; |
5615 |
+- } else { |
5616 |
+- if (to_vmx(vcpu)->nested.vmxon && |
5617 |
+- !nested_host_cr0_valid(vcpu, val)) |
5618 |
+- return 1; |
5619 |
+- |
5620 |
+- return kvm_set_cr0(vcpu, val); |
5621 |
+- } |
5622 |
+-} |
5623 |
+- |
5624 |
+-static int handle_set_cr4(struct kvm_vcpu *vcpu, unsigned long val) |
5625 |
+-{ |
5626 |
+- if (is_guest_mode(vcpu)) { |
5627 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
5628 |
+- unsigned long orig_val = val; |
5629 |
+- |
5630 |
+- /* analogously to handle_set_cr0 */ |
5631 |
+- val = (val & ~vmcs12->cr4_guest_host_mask) | |
5632 |
+- (vmcs12->guest_cr4 & vmcs12->cr4_guest_host_mask); |
5633 |
+- if (kvm_set_cr4(vcpu, val)) |
5634 |
+- return 1; |
5635 |
+- vmcs_writel(CR4_READ_SHADOW, orig_val); |
5636 |
+- return 0; |
5637 |
+- } else |
5638 |
+- return kvm_set_cr4(vcpu, val); |
5639 |
+-} |
5640 |
+- |
5641 |
+-static int handle_desc(struct kvm_vcpu *vcpu) |
5642 |
+-{ |
5643 |
+- WARN_ON(!(vcpu->arch.cr4 & X86_CR4_UMIP)); |
5644 |
+- return kvm_emulate_instruction(vcpu, 0); |
5645 |
+-} |
5646 |
+- |
5647 |
+-static int handle_cr(struct kvm_vcpu *vcpu) |
5648 |
+-{ |
5649 |
+- unsigned long exit_qualification, val; |
5650 |
+- int cr; |
5651 |
+- int reg; |
5652 |
+- int err; |
5653 |
+- int ret; |
5654 |
+- |
5655 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5656 |
+- cr = exit_qualification & 15; |
5657 |
+- reg = (exit_qualification >> 8) & 15; |
5658 |
+- switch ((exit_qualification >> 4) & 3) { |
5659 |
+- case 0: /* mov to cr */ |
5660 |
+- val = kvm_register_readl(vcpu, reg); |
5661 |
+- trace_kvm_cr_write(cr, val); |
5662 |
+- switch (cr) { |
5663 |
+- case 0: |
5664 |
+- err = handle_set_cr0(vcpu, val); |
5665 |
+- return kvm_complete_insn_gp(vcpu, err); |
5666 |
+- case 3: |
5667 |
+- WARN_ON_ONCE(enable_unrestricted_guest); |
5668 |
+- err = kvm_set_cr3(vcpu, val); |
5669 |
+- return kvm_complete_insn_gp(vcpu, err); |
5670 |
+- case 4: |
5671 |
+- err = handle_set_cr4(vcpu, val); |
5672 |
+- return kvm_complete_insn_gp(vcpu, err); |
5673 |
+- case 8: { |
5674 |
+- u8 cr8_prev = kvm_get_cr8(vcpu); |
5675 |
+- u8 cr8 = (u8)val; |
5676 |
+- err = kvm_set_cr8(vcpu, cr8); |
5677 |
+- ret = kvm_complete_insn_gp(vcpu, err); |
5678 |
+- if (lapic_in_kernel(vcpu)) |
5679 |
+- return ret; |
5680 |
+- if (cr8_prev <= cr8) |
5681 |
+- return ret; |
5682 |
+- /* |
5683 |
+- * TODO: we might be squashing a |
5684 |
+- * KVM_GUESTDBG_SINGLESTEP-triggered |
5685 |
+- * KVM_EXIT_DEBUG here. |
5686 |
+- */ |
5687 |
+- vcpu->run->exit_reason = KVM_EXIT_SET_TPR; |
5688 |
+- return 0; |
5689 |
+- } |
5690 |
+- } |
5691 |
+- break; |
5692 |
+- case 2: /* clts */ |
5693 |
+- WARN_ONCE(1, "Guest should always own CR0.TS"); |
5694 |
+- vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); |
5695 |
+- trace_kvm_cr_write(0, kvm_read_cr0(vcpu)); |
5696 |
+- return kvm_skip_emulated_instruction(vcpu); |
5697 |
+- case 1: /*mov from cr*/ |
5698 |
+- switch (cr) { |
5699 |
+- case 3: |
5700 |
+- WARN_ON_ONCE(enable_unrestricted_guest); |
5701 |
+- val = kvm_read_cr3(vcpu); |
5702 |
+- kvm_register_write(vcpu, reg, val); |
5703 |
+- trace_kvm_cr_read(cr, val); |
5704 |
+- return kvm_skip_emulated_instruction(vcpu); |
5705 |
+- case 8: |
5706 |
+- val = kvm_get_cr8(vcpu); |
5707 |
+- kvm_register_write(vcpu, reg, val); |
5708 |
+- trace_kvm_cr_read(cr, val); |
5709 |
+- return kvm_skip_emulated_instruction(vcpu); |
5710 |
+- } |
5711 |
+- break; |
5712 |
+- case 3: /* lmsw */ |
5713 |
+- val = (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f; |
5714 |
+- trace_kvm_cr_write(0, (kvm_read_cr0(vcpu) & ~0xful) | val); |
5715 |
+- kvm_lmsw(vcpu, val); |
5716 |
+- |
5717 |
+- return kvm_skip_emulated_instruction(vcpu); |
5718 |
+- default: |
5719 |
+- break; |
5720 |
+- } |
5721 |
+- vcpu->run->exit_reason = 0; |
5722 |
+- vcpu_unimpl(vcpu, "unhandled control register: op %d cr %d\n", |
5723 |
+- (int)(exit_qualification >> 4) & 3, cr); |
5724 |
+- return 0; |
5725 |
+-} |
5726 |
+- |
5727 |
+-static int handle_dr(struct kvm_vcpu *vcpu) |
5728 |
+-{ |
5729 |
+- unsigned long exit_qualification; |
5730 |
+- int dr, dr7, reg; |
5731 |
+- |
5732 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5733 |
+- dr = exit_qualification & DEBUG_REG_ACCESS_NUM; |
5734 |
+- |
5735 |
+- /* First, if DR does not exist, trigger UD */ |
5736 |
+- if (!kvm_require_dr(vcpu, dr)) |
5737 |
+- return 1; |
5738 |
+- |
5739 |
+- /* Do not handle if the CPL > 0, will trigger GP on re-entry */ |
5740 |
+- if (!kvm_require_cpl(vcpu, 0)) |
5741 |
+- return 1; |
5742 |
+- dr7 = vmcs_readl(GUEST_DR7); |
5743 |
+- if (dr7 & DR7_GD) { |
5744 |
+- /* |
5745 |
+- * As the vm-exit takes precedence over the debug trap, we |
5746 |
+- * need to emulate the latter, either for the host or the |
5747 |
+- * guest debugging itself. |
5748 |
+- */ |
5749 |
+- if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { |
5750 |
+- vcpu->run->debug.arch.dr6 = vcpu->arch.dr6; |
5751 |
+- vcpu->run->debug.arch.dr7 = dr7; |
5752 |
+- vcpu->run->debug.arch.pc = kvm_get_linear_rip(vcpu); |
5753 |
+- vcpu->run->debug.arch.exception = DB_VECTOR; |
5754 |
+- vcpu->run->exit_reason = KVM_EXIT_DEBUG; |
5755 |
+- return 0; |
5756 |
+- } else { |
5757 |
+- vcpu->arch.dr6 &= ~DR_TRAP_BITS; |
5758 |
+- vcpu->arch.dr6 |= DR6_BD | DR6_RTM; |
5759 |
+- kvm_queue_exception(vcpu, DB_VECTOR); |
5760 |
+- return 1; |
5761 |
+- } |
5762 |
+- } |
5763 |
+- |
5764 |
+- if (vcpu->guest_debug == 0) { |
5765 |
+- exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_MOV_DR_EXITING); |
5766 |
+- |
5767 |
+- /* |
5768 |
+- * No more DR vmexits; force a reload of the debug registers |
5769 |
+- * and reenter on this instruction. The next vmexit will |
5770 |
+- * retrieve the full state of the debug registers. |
5771 |
+- */ |
5772 |
+- vcpu->arch.switch_db_regs |= KVM_DEBUGREG_WONT_EXIT; |
5773 |
+- return 1; |
5774 |
+- } |
5775 |
+- |
5776 |
+- reg = DEBUG_REG_ACCESS_REG(exit_qualification); |
5777 |
+- if (exit_qualification & TYPE_MOV_FROM_DR) { |
5778 |
+- unsigned long val; |
5779 |
+- |
5780 |
+- if (kvm_get_dr(vcpu, dr, &val)) |
5781 |
+- return 1; |
5782 |
+- kvm_register_write(vcpu, reg, val); |
5783 |
+- } else |
5784 |
+- if (kvm_set_dr(vcpu, dr, kvm_register_readl(vcpu, reg))) |
5785 |
+- return 1; |
5786 |
+- |
5787 |
+- return kvm_skip_emulated_instruction(vcpu); |
5788 |
+-} |
5789 |
+- |
5790 |
+-static u64 vmx_get_dr6(struct kvm_vcpu *vcpu) |
5791 |
+-{ |
5792 |
+- return vcpu->arch.dr6; |
5793 |
+-} |
5794 |
+- |
5795 |
+-static void vmx_set_dr6(struct kvm_vcpu *vcpu, unsigned long val) |
5796 |
+-{ |
5797 |
+-} |
5798 |
+- |
5799 |
+-static void vmx_sync_dirty_debug_regs(struct kvm_vcpu *vcpu) |
5800 |
+-{ |
5801 |
+- get_debugreg(vcpu->arch.db[0], 0); |
5802 |
+- get_debugreg(vcpu->arch.db[1], 1); |
5803 |
+- get_debugreg(vcpu->arch.db[2], 2); |
5804 |
+- get_debugreg(vcpu->arch.db[3], 3); |
5805 |
+- get_debugreg(vcpu->arch.dr6, 6); |
5806 |
+- vcpu->arch.dr7 = vmcs_readl(GUEST_DR7); |
5807 |
+- |
5808 |
+- vcpu->arch.switch_db_regs &= ~KVM_DEBUGREG_WONT_EXIT; |
5809 |
+- exec_controls_setbit(to_vmx(vcpu), CPU_BASED_MOV_DR_EXITING); |
5810 |
+-} |
5811 |
+- |
5812 |
+-static void vmx_set_dr7(struct kvm_vcpu *vcpu, unsigned long val) |
5813 |
+-{ |
5814 |
+- vmcs_writel(GUEST_DR7, val); |
5815 |
+-} |
5816 |
+- |
5817 |
+-static int handle_tpr_below_threshold(struct kvm_vcpu *vcpu) |
5818 |
+-{ |
5819 |
+- kvm_apic_update_ppr(vcpu); |
5820 |
+- return 1; |
5821 |
+-} |
5822 |
+- |
5823 |
+-static int handle_interrupt_window(struct kvm_vcpu *vcpu) |
5824 |
+-{ |
5825 |
+- exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_INTR_WINDOW_EXITING); |
5826 |
+- |
5827 |
+- kvm_make_request(KVM_REQ_EVENT, vcpu); |
5828 |
+- |
5829 |
+- ++vcpu->stat.irq_window_exits; |
5830 |
+- return 1; |
5831 |
+-} |
5832 |
+- |
5833 |
+-static int handle_vmcall(struct kvm_vcpu *vcpu) |
5834 |
+-{ |
5835 |
+- return kvm_emulate_hypercall(vcpu); |
5836 |
+-} |
5837 |
+- |
5838 |
+-static int handle_invd(struct kvm_vcpu *vcpu) |
5839 |
+-{ |
5840 |
+- return kvm_emulate_instruction(vcpu, 0); |
5841 |
+-} |
5842 |
+- |
5843 |
+-static int handle_invlpg(struct kvm_vcpu *vcpu) |
5844 |
+-{ |
5845 |
+- unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5846 |
+- |
5847 |
+- kvm_mmu_invlpg(vcpu, exit_qualification); |
5848 |
+- return kvm_skip_emulated_instruction(vcpu); |
5849 |
+-} |
5850 |
+- |
5851 |
+-static int handle_rdpmc(struct kvm_vcpu *vcpu) |
5852 |
+-{ |
5853 |
+- int err; |
5854 |
+- |
5855 |
+- err = kvm_rdpmc(vcpu); |
5856 |
+- return kvm_complete_insn_gp(vcpu, err); |
5857 |
+-} |
5858 |
+- |
5859 |
+-static int handle_wbinvd(struct kvm_vcpu *vcpu) |
5860 |
+-{ |
5861 |
+- return kvm_emulate_wbinvd(vcpu); |
5862 |
+-} |
5863 |
+- |
5864 |
+-static int handle_xsetbv(struct kvm_vcpu *vcpu) |
5865 |
+-{ |
5866 |
+- u64 new_bv = kvm_read_edx_eax(vcpu); |
5867 |
+- u32 index = kvm_rcx_read(vcpu); |
5868 |
+- |
5869 |
+- if (kvm_set_xcr(vcpu, index, new_bv) == 0) |
5870 |
+- return kvm_skip_emulated_instruction(vcpu); |
5871 |
+- return 1; |
5872 |
+-} |
5873 |
+- |
5874 |
+-static int handle_apic_access(struct kvm_vcpu *vcpu) |
5875 |
+-{ |
5876 |
+- if (likely(fasteoi)) { |
5877 |
+- unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5878 |
+- int access_type, offset; |
5879 |
+- |
5880 |
+- access_type = exit_qualification & APIC_ACCESS_TYPE; |
5881 |
+- offset = exit_qualification & APIC_ACCESS_OFFSET; |
5882 |
+- /* |
5883 |
+- * Sane guest uses MOV to write EOI, with written value |
5884 |
+- * not cared. So make a short-circuit here by avoiding |
5885 |
+- * heavy instruction emulation. |
5886 |
+- */ |
5887 |
+- if ((access_type == TYPE_LINEAR_APIC_INST_WRITE) && |
5888 |
+- (offset == APIC_EOI)) { |
5889 |
+- kvm_lapic_set_eoi(vcpu); |
5890 |
+- return kvm_skip_emulated_instruction(vcpu); |
5891 |
+- } |
5892 |
+- } |
5893 |
+- return kvm_emulate_instruction(vcpu, 0); |
5894 |
+-} |
5895 |
+- |
5896 |
+-static int handle_apic_eoi_induced(struct kvm_vcpu *vcpu) |
5897 |
+-{ |
5898 |
+- unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5899 |
+- int vector = exit_qualification & 0xff; |
5900 |
+- |
5901 |
+- /* EOI-induced VM exit is trap-like and thus no need to adjust IP */ |
5902 |
+- kvm_apic_set_eoi_accelerated(vcpu, vector); |
5903 |
+- return 1; |
5904 |
+-} |
5905 |
+- |
5906 |
+-static int handle_apic_write(struct kvm_vcpu *vcpu) |
5907 |
+-{ |
5908 |
+- unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5909 |
+- u32 offset = exit_qualification & 0xfff; |
5910 |
+- |
5911 |
+- /* APIC-write VM exit is trap-like and thus no need to adjust IP */ |
5912 |
+- kvm_apic_write_nodecode(vcpu, offset); |
5913 |
+- return 1; |
5914 |
+-} |
5915 |
+- |
5916 |
+-static int handle_task_switch(struct kvm_vcpu *vcpu) |
5917 |
+-{ |
5918 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
5919 |
+- unsigned long exit_qualification; |
5920 |
+- bool has_error_code = false; |
5921 |
+- u32 error_code = 0; |
5922 |
+- u16 tss_selector; |
5923 |
+- int reason, type, idt_v, idt_index; |
5924 |
+- |
5925 |
+- idt_v = (vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK); |
5926 |
+- idt_index = (vmx->idt_vectoring_info & VECTORING_INFO_VECTOR_MASK); |
5927 |
+- type = (vmx->idt_vectoring_info & VECTORING_INFO_TYPE_MASK); |
5928 |
+- |
5929 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5930 |
+- |
5931 |
+- reason = (u32)exit_qualification >> 30; |
5932 |
+- if (reason == TASK_SWITCH_GATE && idt_v) { |
5933 |
+- switch (type) { |
5934 |
+- case INTR_TYPE_NMI_INTR: |
5935 |
+- vcpu->arch.nmi_injected = false; |
5936 |
+- vmx_set_nmi_mask(vcpu, true); |
5937 |
+- break; |
5938 |
+- case INTR_TYPE_EXT_INTR: |
5939 |
+- case INTR_TYPE_SOFT_INTR: |
5940 |
+- kvm_clear_interrupt_queue(vcpu); |
5941 |
+- break; |
5942 |
+- case INTR_TYPE_HARD_EXCEPTION: |
5943 |
+- if (vmx->idt_vectoring_info & |
5944 |
+- VECTORING_INFO_DELIVER_CODE_MASK) { |
5945 |
+- has_error_code = true; |
5946 |
+- error_code = |
5947 |
+- vmcs_read32(IDT_VECTORING_ERROR_CODE); |
5948 |
+- } |
5949 |
+- /* fall through */ |
5950 |
+- case INTR_TYPE_SOFT_EXCEPTION: |
5951 |
+- kvm_clear_exception_queue(vcpu); |
5952 |
+- break; |
5953 |
+- default: |
5954 |
+- break; |
5955 |
+- } |
5956 |
+- } |
5957 |
+- tss_selector = exit_qualification; |
5958 |
+- |
5959 |
+- if (!idt_v || (type != INTR_TYPE_HARD_EXCEPTION && |
5960 |
+- type != INTR_TYPE_EXT_INTR && |
5961 |
+- type != INTR_TYPE_NMI_INTR)) |
5962 |
+- WARN_ON(!skip_emulated_instruction(vcpu)); |
5963 |
+- |
5964 |
+- /* |
5965 |
+- * TODO: What about debug traps on tss switch? |
5966 |
+- * Are we supposed to inject them and update dr6? |
5967 |
+- */ |
5968 |
+- return kvm_task_switch(vcpu, tss_selector, |
5969 |
+- type == INTR_TYPE_SOFT_INTR ? idt_index : -1, |
5970 |
+- reason, has_error_code, error_code); |
5971 |
+-} |
5972 |
+- |
5973 |
+-static int handle_ept_violation(struct kvm_vcpu *vcpu) |
5974 |
+-{ |
5975 |
+- unsigned long exit_qualification; |
5976 |
+- gpa_t gpa; |
5977 |
+- u64 error_code; |
5978 |
+- |
5979 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
5980 |
+- |
5981 |
+- /* |
5982 |
+- * EPT violation happened while executing iret from NMI, |
5983 |
+- * "blocked by NMI" bit has to be set before next VM entry. |
5984 |
+- * There are errata that may cause this bit to not be set: |
5985 |
+- * AAK134, BY25. |
5986 |
+- */ |
5987 |
+- if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) && |
5988 |
+- enable_vnmi && |
5989 |
+- (exit_qualification & INTR_INFO_UNBLOCK_NMI)) |
5990 |
+- vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, GUEST_INTR_STATE_NMI); |
5991 |
+- |
5992 |
+- gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
5993 |
+- trace_kvm_page_fault(gpa, exit_qualification); |
5994 |
+- |
5995 |
+- /* Is it a read fault? */ |
5996 |
+- error_code = (exit_qualification & EPT_VIOLATION_ACC_READ) |
5997 |
+- ? PFERR_USER_MASK : 0; |
5998 |
+- /* Is it a write fault? */ |
5999 |
+- error_code |= (exit_qualification & EPT_VIOLATION_ACC_WRITE) |
6000 |
+- ? PFERR_WRITE_MASK : 0; |
6001 |
+- /* Is it a fetch fault? */ |
6002 |
+- error_code |= (exit_qualification & EPT_VIOLATION_ACC_INSTR) |
6003 |
+- ? PFERR_FETCH_MASK : 0; |
6004 |
+- /* ept page table entry is present? */ |
6005 |
+- error_code |= (exit_qualification & |
6006 |
+- (EPT_VIOLATION_READABLE | EPT_VIOLATION_WRITABLE | |
6007 |
+- EPT_VIOLATION_EXECUTABLE)) |
6008 |
+- ? PFERR_PRESENT_MASK : 0; |
6009 |
+- |
6010 |
+- error_code |= (exit_qualification & 0x100) != 0 ? |
6011 |
+- PFERR_GUEST_FINAL_MASK : PFERR_GUEST_PAGE_MASK; |
6012 |
+- |
6013 |
+- vcpu->arch.exit_qualification = exit_qualification; |
6014 |
+- return kvm_mmu_page_fault(vcpu, gpa, error_code, NULL, 0); |
6015 |
+-} |
6016 |
+- |
6017 |
+-static int handle_ept_misconfig(struct kvm_vcpu *vcpu) |
6018 |
+-{ |
6019 |
+- gpa_t gpa; |
6020 |
+- |
6021 |
+- /* |
6022 |
+- * A nested guest cannot optimize MMIO vmexits, because we have an |
6023 |
+- * nGPA here instead of the required GPA. |
6024 |
+- */ |
6025 |
+- gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
6026 |
+- if (!is_guest_mode(vcpu) && |
6027 |
+- !kvm_io_bus_write(vcpu, KVM_FAST_MMIO_BUS, gpa, 0, NULL)) { |
6028 |
+- trace_kvm_fast_mmio(gpa); |
6029 |
+- return kvm_skip_emulated_instruction(vcpu); |
6030 |
+- } |
6031 |
+- |
6032 |
+- return kvm_mmu_page_fault(vcpu, gpa, PFERR_RSVD_MASK, NULL, 0); |
6033 |
+-} |
6034 |
+- |
6035 |
+-static int handle_nmi_window(struct kvm_vcpu *vcpu) |
6036 |
+-{ |
6037 |
+- WARN_ON_ONCE(!enable_vnmi); |
6038 |
+- exec_controls_clearbit(to_vmx(vcpu), CPU_BASED_NMI_WINDOW_EXITING); |
6039 |
+- ++vcpu->stat.nmi_window_exits; |
6040 |
+- kvm_make_request(KVM_REQ_EVENT, vcpu); |
6041 |
+- |
6042 |
+- return 1; |
6043 |
+-} |
6044 |
+- |
6045 |
+-static int handle_invalid_guest_state(struct kvm_vcpu *vcpu) |
6046 |
+-{ |
6047 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6048 |
+- bool intr_window_requested; |
6049 |
+- unsigned count = 130; |
6050 |
+- |
6051 |
+- /* |
6052 |
+- * We should never reach the point where we are emulating L2 |
6053 |
+- * due to invalid guest state as that means we incorrectly |
6054 |
+- * allowed a nested VMEntry with an invalid vmcs12. |
6055 |
+- */ |
6056 |
+- WARN_ON_ONCE(vmx->emulation_required && vmx->nested.nested_run_pending); |
6057 |
+- |
6058 |
+- intr_window_requested = exec_controls_get(vmx) & |
6059 |
+- CPU_BASED_INTR_WINDOW_EXITING; |
6060 |
+- |
6061 |
+- while (vmx->emulation_required && count-- != 0) { |
6062 |
+- if (intr_window_requested && vmx_interrupt_allowed(vcpu)) |
6063 |
+- return handle_interrupt_window(&vmx->vcpu); |
6064 |
+- |
6065 |
+- if (kvm_test_request(KVM_REQ_EVENT, vcpu)) |
6066 |
+- return 1; |
6067 |
+- |
6068 |
+- if (!kvm_emulate_instruction(vcpu, 0)) |
6069 |
+- return 0; |
6070 |
+- |
6071 |
+- if (vmx->emulation_required && !vmx->rmode.vm86_active && |
6072 |
+- vcpu->arch.exception.pending) { |
6073 |
+- vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
6074 |
+- vcpu->run->internal.suberror = |
6075 |
+- KVM_INTERNAL_ERROR_EMULATION; |
6076 |
+- vcpu->run->internal.ndata = 0; |
6077 |
+- return 0; |
6078 |
+- } |
6079 |
+- |
6080 |
+- if (vcpu->arch.halt_request) { |
6081 |
+- vcpu->arch.halt_request = 0; |
6082 |
+- return kvm_vcpu_halt(vcpu); |
6083 |
+- } |
6084 |
+- |
6085 |
+- /* |
6086 |
+- * Note, return 1 and not 0, vcpu_run() is responsible for |
6087 |
+- * morphing the pending signal into the proper return code. |
6088 |
+- */ |
6089 |
+- if (signal_pending(current)) |
6090 |
+- return 1; |
6091 |
+- |
6092 |
+- if (need_resched()) |
6093 |
+- schedule(); |
6094 |
+- } |
6095 |
+- |
6096 |
+- return 1; |
6097 |
+-} |
6098 |
+- |
6099 |
+-static void grow_ple_window(struct kvm_vcpu *vcpu) |
6100 |
+-{ |
6101 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6102 |
+- unsigned int old = vmx->ple_window; |
6103 |
+- |
6104 |
+- vmx->ple_window = __grow_ple_window(old, ple_window, |
6105 |
+- ple_window_grow, |
6106 |
+- ple_window_max); |
6107 |
+- |
6108 |
+- if (vmx->ple_window != old) { |
6109 |
+- vmx->ple_window_dirty = true; |
6110 |
+- trace_kvm_ple_window_update(vcpu->vcpu_id, |
6111 |
+- vmx->ple_window, old); |
6112 |
+- } |
6113 |
+-} |
6114 |
+- |
6115 |
+-static void shrink_ple_window(struct kvm_vcpu *vcpu) |
6116 |
+-{ |
6117 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6118 |
+- unsigned int old = vmx->ple_window; |
6119 |
+- |
6120 |
+- vmx->ple_window = __shrink_ple_window(old, ple_window, |
6121 |
+- ple_window_shrink, |
6122 |
+- ple_window); |
6123 |
+- |
6124 |
+- if (vmx->ple_window != old) { |
6125 |
+- vmx->ple_window_dirty = true; |
6126 |
+- trace_kvm_ple_window_update(vcpu->vcpu_id, |
6127 |
+- vmx->ple_window, old); |
6128 |
+- } |
6129 |
+-} |
6130 |
+- |
6131 |
+-/* |
6132 |
+- * Handler for POSTED_INTERRUPT_WAKEUP_VECTOR. |
6133 |
+- */ |
6134 |
+-static void wakeup_handler(void) |
6135 |
+-{ |
6136 |
+- struct kvm_vcpu *vcpu; |
6137 |
+- int cpu = smp_processor_id(); |
6138 |
+- |
6139 |
+- spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
6140 |
+- list_for_each_entry(vcpu, &per_cpu(blocked_vcpu_on_cpu, cpu), |
6141 |
+- blocked_vcpu_list) { |
6142 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
6143 |
+- |
6144 |
+- if (pi_test_on(pi_desc) == 1) |
6145 |
+- kvm_vcpu_kick(vcpu); |
6146 |
+- } |
6147 |
+- spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, cpu)); |
6148 |
+-} |
6149 |
+- |
6150 |
+-static void vmx_enable_tdp(void) |
6151 |
+-{ |
6152 |
+- kvm_mmu_set_mask_ptes(VMX_EPT_READABLE_MASK, |
6153 |
+- enable_ept_ad_bits ? VMX_EPT_ACCESS_BIT : 0ull, |
6154 |
+- enable_ept_ad_bits ? VMX_EPT_DIRTY_BIT : 0ull, |
6155 |
+- 0ull, VMX_EPT_EXECUTABLE_MASK, |
6156 |
+- cpu_has_vmx_ept_execute_only() ? 0ull : VMX_EPT_READABLE_MASK, |
6157 |
+- VMX_EPT_RWX_MASK, 0ull); |
6158 |
+- |
6159 |
+- ept_set_mmio_spte_mask(); |
6160 |
+- kvm_enable_tdp(); |
6161 |
+-} |
6162 |
+- |
6163 |
+-/* |
6164 |
+- * Indicate a busy-waiting vcpu in spinlock. We do not enable the PAUSE |
6165 |
+- * exiting, so only get here on cpu with PAUSE-Loop-Exiting. |
6166 |
+- */ |
6167 |
+-static int handle_pause(struct kvm_vcpu *vcpu) |
6168 |
+-{ |
6169 |
+- if (!kvm_pause_in_guest(vcpu->kvm)) |
6170 |
+- grow_ple_window(vcpu); |
6171 |
+- |
6172 |
+- /* |
6173 |
+- * Intel sdm vol3 ch-25.1.3 says: The "PAUSE-loop exiting" |
6174 |
+- * VM-execution control is ignored if CPL > 0. OTOH, KVM |
6175 |
+- * never set PAUSE_EXITING and just set PLE if supported, |
6176 |
+- * so the vcpu must be CPL=0 if it gets a PAUSE exit. |
6177 |
+- */ |
6178 |
+- kvm_vcpu_on_spin(vcpu, true); |
6179 |
+- return kvm_skip_emulated_instruction(vcpu); |
6180 |
+-} |
6181 |
+- |
6182 |
+-static int handle_nop(struct kvm_vcpu *vcpu) |
6183 |
+-{ |
6184 |
+- return kvm_skip_emulated_instruction(vcpu); |
6185 |
+-} |
6186 |
+- |
6187 |
+-static int handle_mwait(struct kvm_vcpu *vcpu) |
6188 |
+-{ |
6189 |
+- printk_once(KERN_WARNING "kvm: MWAIT instruction emulated as NOP!\n"); |
6190 |
+- return handle_nop(vcpu); |
6191 |
+-} |
6192 |
+- |
6193 |
+-static int handle_invalid_op(struct kvm_vcpu *vcpu) |
6194 |
+-{ |
6195 |
+- kvm_queue_exception(vcpu, UD_VECTOR); |
6196 |
+- return 1; |
6197 |
+-} |
6198 |
+- |
6199 |
+-static int handle_monitor_trap(struct kvm_vcpu *vcpu) |
6200 |
+-{ |
6201 |
+- return 1; |
6202 |
+-} |
6203 |
+- |
6204 |
+-static int handle_monitor(struct kvm_vcpu *vcpu) |
6205 |
+-{ |
6206 |
+- printk_once(KERN_WARNING "kvm: MONITOR instruction emulated as NOP!\n"); |
6207 |
+- return handle_nop(vcpu); |
6208 |
+-} |
6209 |
+- |
6210 |
+-static int handle_invpcid(struct kvm_vcpu *vcpu) |
6211 |
+-{ |
6212 |
+- u32 vmx_instruction_info; |
6213 |
+- unsigned long type; |
6214 |
+- bool pcid_enabled; |
6215 |
+- gva_t gva; |
6216 |
+- struct x86_exception e; |
6217 |
+- unsigned i; |
6218 |
+- unsigned long roots_to_free = 0; |
6219 |
+- struct { |
6220 |
+- u64 pcid; |
6221 |
+- u64 gla; |
6222 |
+- } operand; |
6223 |
+- |
6224 |
+- if (!guest_cpuid_has(vcpu, X86_FEATURE_INVPCID)) { |
6225 |
+- kvm_queue_exception(vcpu, UD_VECTOR); |
6226 |
+- return 1; |
6227 |
+- } |
6228 |
+- |
6229 |
+- vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); |
6230 |
+- type = kvm_register_readl(vcpu, (vmx_instruction_info >> 28) & 0xf); |
6231 |
+- |
6232 |
+- if (type > 3) { |
6233 |
+- kvm_inject_gp(vcpu, 0); |
6234 |
+- return 1; |
6235 |
+- } |
6236 |
+- |
6237 |
+- /* According to the Intel instruction reference, the memory operand |
6238 |
+- * is read even if it isn't needed (e.g., for type==all) |
6239 |
+- */ |
6240 |
+- if (get_vmx_mem_address(vcpu, vmcs_readl(EXIT_QUALIFICATION), |
6241 |
+- vmx_instruction_info, false, |
6242 |
+- sizeof(operand), &gva)) |
6243 |
+- return 1; |
6244 |
+- |
6245 |
+- if (kvm_read_guest_virt(vcpu, gva, &operand, sizeof(operand), &e)) { |
6246 |
+- kvm_inject_page_fault(vcpu, &e); |
6247 |
+- return 1; |
6248 |
+- } |
6249 |
+- |
6250 |
+- if (operand.pcid >> 12 != 0) { |
6251 |
+- kvm_inject_gp(vcpu, 0); |
6252 |
+- return 1; |
6253 |
+- } |
6254 |
+- |
6255 |
+- pcid_enabled = kvm_read_cr4_bits(vcpu, X86_CR4_PCIDE); |
6256 |
+- |
6257 |
+- switch (type) { |
6258 |
+- case INVPCID_TYPE_INDIV_ADDR: |
6259 |
+- if ((!pcid_enabled && (operand.pcid != 0)) || |
6260 |
+- is_noncanonical_address(operand.gla, vcpu)) { |
6261 |
+- kvm_inject_gp(vcpu, 0); |
6262 |
+- return 1; |
6263 |
+- } |
6264 |
+- kvm_mmu_invpcid_gva(vcpu, operand.gla, operand.pcid); |
6265 |
+- return kvm_skip_emulated_instruction(vcpu); |
6266 |
+- |
6267 |
+- case INVPCID_TYPE_SINGLE_CTXT: |
6268 |
+- if (!pcid_enabled && (operand.pcid != 0)) { |
6269 |
+- kvm_inject_gp(vcpu, 0); |
6270 |
+- return 1; |
6271 |
+- } |
6272 |
+- |
6273 |
+- if (kvm_get_active_pcid(vcpu) == operand.pcid) { |
6274 |
+- kvm_mmu_sync_roots(vcpu); |
6275 |
+- kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
6276 |
+- } |
6277 |
+- |
6278 |
+- for (i = 0; i < KVM_MMU_NUM_PREV_ROOTS; i++) |
6279 |
+- if (kvm_get_pcid(vcpu, vcpu->arch.mmu->prev_roots[i].cr3) |
6280 |
+- == operand.pcid) |
6281 |
+- roots_to_free |= KVM_MMU_ROOT_PREVIOUS(i); |
6282 |
+- |
6283 |
+- kvm_mmu_free_roots(vcpu, vcpu->arch.mmu, roots_to_free); |
6284 |
+- /* |
6285 |
+- * If neither the current cr3 nor any of the prev_roots use the |
6286 |
+- * given PCID, then nothing needs to be done here because a |
6287 |
+- * resync will happen anyway before switching to any other CR3. |
6288 |
+- */ |
6289 |
+- |
6290 |
+- return kvm_skip_emulated_instruction(vcpu); |
6291 |
+- |
6292 |
+- case INVPCID_TYPE_ALL_NON_GLOBAL: |
6293 |
+- /* |
6294 |
+- * Currently, KVM doesn't mark global entries in the shadow |
6295 |
+- * page tables, so a non-global flush just degenerates to a |
6296 |
+- * global flush. If needed, we could optimize this later by |
6297 |
+- * keeping track of global entries in shadow page tables. |
6298 |
+- */ |
6299 |
+- |
6300 |
+- /* fall-through */ |
6301 |
+- case INVPCID_TYPE_ALL_INCL_GLOBAL: |
6302 |
+- kvm_mmu_unload(vcpu); |
6303 |
+- return kvm_skip_emulated_instruction(vcpu); |
6304 |
+- |
6305 |
+- default: |
6306 |
+- BUG(); /* We have already checked above that type <= 3 */ |
6307 |
+- } |
6308 |
+-} |
6309 |
+- |
6310 |
+-static int handle_pml_full(struct kvm_vcpu *vcpu) |
6311 |
+-{ |
6312 |
+- unsigned long exit_qualification; |
6313 |
+- |
6314 |
+- trace_kvm_pml_full(vcpu->vcpu_id); |
6315 |
+- |
6316 |
+- exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6317 |
+- |
6318 |
+- /* |
6319 |
+- * PML buffer FULL happened while executing iret from NMI, |
6320 |
+- * "blocked by NMI" bit has to be set before next VM entry. |
6321 |
+- */ |
6322 |
+- if (!(to_vmx(vcpu)->idt_vectoring_info & VECTORING_INFO_VALID_MASK) && |
6323 |
+- enable_vnmi && |
6324 |
+- (exit_qualification & INTR_INFO_UNBLOCK_NMI)) |
6325 |
+- vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
6326 |
+- GUEST_INTR_STATE_NMI); |
6327 |
+- |
6328 |
+- /* |
6329 |
+- * PML buffer already flushed at beginning of VMEXIT. Nothing to do |
6330 |
+- * here.., and there's no userspace involvement needed for PML. |
6331 |
+- */ |
6332 |
+- return 1; |
6333 |
+-} |
6334 |
+- |
6335 |
+-static int handle_preemption_timer(struct kvm_vcpu *vcpu) |
6336 |
+-{ |
6337 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6338 |
+- |
6339 |
+- if (!vmx->req_immediate_exit && |
6340 |
+- !unlikely(vmx->loaded_vmcs->hv_timer_soft_disabled)) |
6341 |
+- kvm_lapic_expired_hv_timer(vcpu); |
6342 |
+- |
6343 |
+- return 1; |
6344 |
+-} |
6345 |
+- |
6346 |
+-/* |
6347 |
+- * When nested=0, all VMX instruction VM Exits filter here. The handlers |
6348 |
+- * are overwritten by nested_vmx_setup() when nested=1. |
6349 |
+- */ |
6350 |
+-static int handle_vmx_instruction(struct kvm_vcpu *vcpu) |
6351 |
+-{ |
6352 |
+- kvm_queue_exception(vcpu, UD_VECTOR); |
6353 |
+- return 1; |
6354 |
+-} |
6355 |
+- |
6356 |
+-static int handle_encls(struct kvm_vcpu *vcpu) |
6357 |
+-{ |
6358 |
+- /* |
6359 |
+- * SGX virtualization is not yet supported. There is no software |
6360 |
+- * enable bit for SGX, so we have to trap ENCLS and inject a #UD |
6361 |
+- * to prevent the guest from executing ENCLS. |
6362 |
+- */ |
6363 |
+- kvm_queue_exception(vcpu, UD_VECTOR); |
6364 |
+- return 1; |
6365 |
+-} |
6366 |
+- |
6367 |
+-/* |
6368 |
+- * The exit handlers return 1 if the exit was handled fully and guest execution |
6369 |
+- * may resume. Otherwise they set the kvm_run parameter to indicate what needs |
6370 |
+- * to be done to userspace and return 0. |
6371 |
+- */ |
6372 |
+-static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = { |
6373 |
+- [EXIT_REASON_EXCEPTION_NMI] = handle_exception_nmi, |
6374 |
+- [EXIT_REASON_EXTERNAL_INTERRUPT] = handle_external_interrupt, |
6375 |
+- [EXIT_REASON_TRIPLE_FAULT] = handle_triple_fault, |
6376 |
+- [EXIT_REASON_NMI_WINDOW] = handle_nmi_window, |
6377 |
+- [EXIT_REASON_IO_INSTRUCTION] = handle_io, |
6378 |
+- [EXIT_REASON_CR_ACCESS] = handle_cr, |
6379 |
+- [EXIT_REASON_DR_ACCESS] = handle_dr, |
6380 |
+- [EXIT_REASON_CPUID] = kvm_emulate_cpuid, |
6381 |
+- [EXIT_REASON_MSR_READ] = kvm_emulate_rdmsr, |
6382 |
+- [EXIT_REASON_MSR_WRITE] = kvm_emulate_wrmsr, |
6383 |
+- [EXIT_REASON_INTERRUPT_WINDOW] = handle_interrupt_window, |
6384 |
+- [EXIT_REASON_HLT] = kvm_emulate_halt, |
6385 |
+- [EXIT_REASON_INVD] = handle_invd, |
6386 |
+- [EXIT_REASON_INVLPG] = handle_invlpg, |
6387 |
+- [EXIT_REASON_RDPMC] = handle_rdpmc, |
6388 |
+- [EXIT_REASON_VMCALL] = handle_vmcall, |
6389 |
+- [EXIT_REASON_VMCLEAR] = handle_vmx_instruction, |
6390 |
+- [EXIT_REASON_VMLAUNCH] = handle_vmx_instruction, |
6391 |
+- [EXIT_REASON_VMPTRLD] = handle_vmx_instruction, |
6392 |
+- [EXIT_REASON_VMPTRST] = handle_vmx_instruction, |
6393 |
+- [EXIT_REASON_VMREAD] = handle_vmx_instruction, |
6394 |
+- [EXIT_REASON_VMRESUME] = handle_vmx_instruction, |
6395 |
+- [EXIT_REASON_VMWRITE] = handle_vmx_instruction, |
6396 |
+- [EXIT_REASON_VMOFF] = handle_vmx_instruction, |
6397 |
+- [EXIT_REASON_VMON] = handle_vmx_instruction, |
6398 |
+- [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, |
6399 |
+- [EXIT_REASON_APIC_ACCESS] = handle_apic_access, |
6400 |
+- [EXIT_REASON_APIC_WRITE] = handle_apic_write, |
6401 |
+- [EXIT_REASON_EOI_INDUCED] = handle_apic_eoi_induced, |
6402 |
+- [EXIT_REASON_WBINVD] = handle_wbinvd, |
6403 |
+- [EXIT_REASON_XSETBV] = handle_xsetbv, |
6404 |
+- [EXIT_REASON_TASK_SWITCH] = handle_task_switch, |
6405 |
+- [EXIT_REASON_MCE_DURING_VMENTRY] = handle_machine_check, |
6406 |
+- [EXIT_REASON_GDTR_IDTR] = handle_desc, |
6407 |
+- [EXIT_REASON_LDTR_TR] = handle_desc, |
6408 |
+- [EXIT_REASON_EPT_VIOLATION] = handle_ept_violation, |
6409 |
+- [EXIT_REASON_EPT_MISCONFIG] = handle_ept_misconfig, |
6410 |
+- [EXIT_REASON_PAUSE_INSTRUCTION] = handle_pause, |
6411 |
+- [EXIT_REASON_MWAIT_INSTRUCTION] = handle_mwait, |
6412 |
+- [EXIT_REASON_MONITOR_TRAP_FLAG] = handle_monitor_trap, |
6413 |
+- [EXIT_REASON_MONITOR_INSTRUCTION] = handle_monitor, |
6414 |
+- [EXIT_REASON_INVEPT] = handle_vmx_instruction, |
6415 |
+- [EXIT_REASON_INVVPID] = handle_vmx_instruction, |
6416 |
+- [EXIT_REASON_RDRAND] = handle_invalid_op, |
6417 |
+- [EXIT_REASON_RDSEED] = handle_invalid_op, |
6418 |
+- [EXIT_REASON_PML_FULL] = handle_pml_full, |
6419 |
+- [EXIT_REASON_INVPCID] = handle_invpcid, |
6420 |
+- [EXIT_REASON_VMFUNC] = handle_vmx_instruction, |
6421 |
+- [EXIT_REASON_PREEMPTION_TIMER] = handle_preemption_timer, |
6422 |
+- [EXIT_REASON_ENCLS] = handle_encls, |
6423 |
+-}; |
6424 |
+- |
6425 |
+-static const int kvm_vmx_max_exit_handlers = |
6426 |
+- ARRAY_SIZE(kvm_vmx_exit_handlers); |
6427 |
+- |
6428 |
+-static void vmx_get_exit_info(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2) |
6429 |
+-{ |
6430 |
+- *info1 = vmcs_readl(EXIT_QUALIFICATION); |
6431 |
+- *info2 = vmcs_read32(VM_EXIT_INTR_INFO); |
6432 |
+-} |
6433 |
+- |
6434 |
+-static void vmx_destroy_pml_buffer(struct vcpu_vmx *vmx) |
6435 |
+-{ |
6436 |
+- if (vmx->pml_pg) { |
6437 |
+- __free_page(vmx->pml_pg); |
6438 |
+- vmx->pml_pg = NULL; |
6439 |
+- } |
6440 |
+-} |
6441 |
+- |
6442 |
+-static void vmx_flush_pml_buffer(struct kvm_vcpu *vcpu) |
6443 |
+-{ |
6444 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6445 |
+- u64 *pml_buf; |
6446 |
+- u16 pml_idx; |
6447 |
+- |
6448 |
+- pml_idx = vmcs_read16(GUEST_PML_INDEX); |
6449 |
+- |
6450 |
+- /* Do nothing if PML buffer is empty */ |
6451 |
+- if (pml_idx == (PML_ENTITY_NUM - 1)) |
6452 |
+- return; |
6453 |
+- |
6454 |
+- /* PML index always points to next available PML buffer entity */ |
6455 |
+- if (pml_idx >= PML_ENTITY_NUM) |
6456 |
+- pml_idx = 0; |
6457 |
+- else |
6458 |
+- pml_idx++; |
6459 |
+- |
6460 |
+- pml_buf = page_address(vmx->pml_pg); |
6461 |
+- for (; pml_idx < PML_ENTITY_NUM; pml_idx++) { |
6462 |
+- u64 gpa; |
6463 |
+- |
6464 |
+- gpa = pml_buf[pml_idx]; |
6465 |
+- WARN_ON(gpa & (PAGE_SIZE - 1)); |
6466 |
+- kvm_vcpu_mark_page_dirty(vcpu, gpa >> PAGE_SHIFT); |
6467 |
+- } |
6468 |
+- |
6469 |
+- /* reset PML index */ |
6470 |
+- vmcs_write16(GUEST_PML_INDEX, PML_ENTITY_NUM - 1); |
6471 |
+-} |
6472 |
+- |
6473 |
+-/* |
6474 |
+- * Flush all vcpus' PML buffer and update logged GPAs to dirty_bitmap. |
6475 |
+- * Called before reporting dirty_bitmap to userspace. |
6476 |
+- */ |
6477 |
+-static void kvm_flush_pml_buffers(struct kvm *kvm) |
6478 |
+-{ |
6479 |
+- int i; |
6480 |
+- struct kvm_vcpu *vcpu; |
6481 |
+- /* |
6482 |
+- * We only need to kick vcpu out of guest mode here, as PML buffer |
6483 |
+- * is flushed at beginning of all VMEXITs, and it's obvious that only |
6484 |
+- * vcpus running in guest are possible to have unflushed GPAs in PML |
6485 |
+- * buffer. |
6486 |
+- */ |
6487 |
+- kvm_for_each_vcpu(i, vcpu, kvm) |
6488 |
+- kvm_vcpu_kick(vcpu); |
6489 |
+-} |
6490 |
+- |
6491 |
+-static void vmx_dump_sel(char *name, uint32_t sel) |
6492 |
+-{ |
6493 |
+- pr_err("%s sel=0x%04x, attr=0x%05x, limit=0x%08x, base=0x%016lx\n", |
6494 |
+- name, vmcs_read16(sel), |
6495 |
+- vmcs_read32(sel + GUEST_ES_AR_BYTES - GUEST_ES_SELECTOR), |
6496 |
+- vmcs_read32(sel + GUEST_ES_LIMIT - GUEST_ES_SELECTOR), |
6497 |
+- vmcs_readl(sel + GUEST_ES_BASE - GUEST_ES_SELECTOR)); |
6498 |
+-} |
6499 |
+- |
6500 |
+-static void vmx_dump_dtsel(char *name, uint32_t limit) |
6501 |
+-{ |
6502 |
+- pr_err("%s limit=0x%08x, base=0x%016lx\n", |
6503 |
+- name, vmcs_read32(limit), |
6504 |
+- vmcs_readl(limit + GUEST_GDTR_BASE - GUEST_GDTR_LIMIT)); |
6505 |
+-} |
6506 |
+- |
6507 |
+-void dump_vmcs(void) |
6508 |
+-{ |
6509 |
+- u32 vmentry_ctl, vmexit_ctl; |
6510 |
+- u32 cpu_based_exec_ctrl, pin_based_exec_ctrl, secondary_exec_control; |
6511 |
+- unsigned long cr4; |
6512 |
+- u64 efer; |
6513 |
+- int i, n; |
6514 |
+- |
6515 |
+- if (!dump_invalid_vmcs) { |
6516 |
+- pr_warn_ratelimited("set kvm_intel.dump_invalid_vmcs=1 to dump internal KVM state.\n"); |
6517 |
+- return; |
6518 |
+- } |
6519 |
+- |
6520 |
+- vmentry_ctl = vmcs_read32(VM_ENTRY_CONTROLS); |
6521 |
+- vmexit_ctl = vmcs_read32(VM_EXIT_CONTROLS); |
6522 |
+- cpu_based_exec_ctrl = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); |
6523 |
+- pin_based_exec_ctrl = vmcs_read32(PIN_BASED_VM_EXEC_CONTROL); |
6524 |
+- cr4 = vmcs_readl(GUEST_CR4); |
6525 |
+- efer = vmcs_read64(GUEST_IA32_EFER); |
6526 |
+- secondary_exec_control = 0; |
6527 |
+- if (cpu_has_secondary_exec_ctrls()) |
6528 |
+- secondary_exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL); |
6529 |
+- |
6530 |
+- pr_err("*** Guest State ***\n"); |
6531 |
+- pr_err("CR0: actual=0x%016lx, shadow=0x%016lx, gh_mask=%016lx\n", |
6532 |
+- vmcs_readl(GUEST_CR0), vmcs_readl(CR0_READ_SHADOW), |
6533 |
+- vmcs_readl(CR0_GUEST_HOST_MASK)); |
6534 |
+- pr_err("CR4: actual=0x%016lx, shadow=0x%016lx, gh_mask=%016lx\n", |
6535 |
+- cr4, vmcs_readl(CR4_READ_SHADOW), vmcs_readl(CR4_GUEST_HOST_MASK)); |
6536 |
+- pr_err("CR3 = 0x%016lx\n", vmcs_readl(GUEST_CR3)); |
6537 |
+- if ((secondary_exec_control & SECONDARY_EXEC_ENABLE_EPT) && |
6538 |
+- (cr4 & X86_CR4_PAE) && !(efer & EFER_LMA)) |
6539 |
+- { |
6540 |
+- pr_err("PDPTR0 = 0x%016llx PDPTR1 = 0x%016llx\n", |
6541 |
+- vmcs_read64(GUEST_PDPTR0), vmcs_read64(GUEST_PDPTR1)); |
6542 |
+- pr_err("PDPTR2 = 0x%016llx PDPTR3 = 0x%016llx\n", |
6543 |
+- vmcs_read64(GUEST_PDPTR2), vmcs_read64(GUEST_PDPTR3)); |
6544 |
+- } |
6545 |
+- pr_err("RSP = 0x%016lx RIP = 0x%016lx\n", |
6546 |
+- vmcs_readl(GUEST_RSP), vmcs_readl(GUEST_RIP)); |
6547 |
+- pr_err("RFLAGS=0x%08lx DR7 = 0x%016lx\n", |
6548 |
+- vmcs_readl(GUEST_RFLAGS), vmcs_readl(GUEST_DR7)); |
6549 |
+- pr_err("Sysenter RSP=%016lx CS:RIP=%04x:%016lx\n", |
6550 |
+- vmcs_readl(GUEST_SYSENTER_ESP), |
6551 |
+- vmcs_read32(GUEST_SYSENTER_CS), vmcs_readl(GUEST_SYSENTER_EIP)); |
6552 |
+- vmx_dump_sel("CS: ", GUEST_CS_SELECTOR); |
6553 |
+- vmx_dump_sel("DS: ", GUEST_DS_SELECTOR); |
6554 |
+- vmx_dump_sel("SS: ", GUEST_SS_SELECTOR); |
6555 |
+- vmx_dump_sel("ES: ", GUEST_ES_SELECTOR); |
6556 |
+- vmx_dump_sel("FS: ", GUEST_FS_SELECTOR); |
6557 |
+- vmx_dump_sel("GS: ", GUEST_GS_SELECTOR); |
6558 |
+- vmx_dump_dtsel("GDTR:", GUEST_GDTR_LIMIT); |
6559 |
+- vmx_dump_sel("LDTR:", GUEST_LDTR_SELECTOR); |
6560 |
+- vmx_dump_dtsel("IDTR:", GUEST_IDTR_LIMIT); |
6561 |
+- vmx_dump_sel("TR: ", GUEST_TR_SELECTOR); |
6562 |
+- if ((vmexit_ctl & (VM_EXIT_SAVE_IA32_PAT | VM_EXIT_SAVE_IA32_EFER)) || |
6563 |
+- (vmentry_ctl & (VM_ENTRY_LOAD_IA32_PAT | VM_ENTRY_LOAD_IA32_EFER))) |
6564 |
+- pr_err("EFER = 0x%016llx PAT = 0x%016llx\n", |
6565 |
+- efer, vmcs_read64(GUEST_IA32_PAT)); |
6566 |
+- pr_err("DebugCtl = 0x%016llx DebugExceptions = 0x%016lx\n", |
6567 |
+- vmcs_read64(GUEST_IA32_DEBUGCTL), |
6568 |
+- vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS)); |
6569 |
+- if (cpu_has_load_perf_global_ctrl() && |
6570 |
+- vmentry_ctl & VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) |
6571 |
+- pr_err("PerfGlobCtl = 0x%016llx\n", |
6572 |
+- vmcs_read64(GUEST_IA32_PERF_GLOBAL_CTRL)); |
6573 |
+- if (vmentry_ctl & VM_ENTRY_LOAD_BNDCFGS) |
6574 |
+- pr_err("BndCfgS = 0x%016llx\n", vmcs_read64(GUEST_BNDCFGS)); |
6575 |
+- pr_err("Interruptibility = %08x ActivityState = %08x\n", |
6576 |
+- vmcs_read32(GUEST_INTERRUPTIBILITY_INFO), |
6577 |
+- vmcs_read32(GUEST_ACTIVITY_STATE)); |
6578 |
+- if (secondary_exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY) |
6579 |
+- pr_err("InterruptStatus = %04x\n", |
6580 |
+- vmcs_read16(GUEST_INTR_STATUS)); |
6581 |
+- |
6582 |
+- pr_err("*** Host State ***\n"); |
6583 |
+- pr_err("RIP = 0x%016lx RSP = 0x%016lx\n", |
6584 |
+- vmcs_readl(HOST_RIP), vmcs_readl(HOST_RSP)); |
6585 |
+- pr_err("CS=%04x SS=%04x DS=%04x ES=%04x FS=%04x GS=%04x TR=%04x\n", |
6586 |
+- vmcs_read16(HOST_CS_SELECTOR), vmcs_read16(HOST_SS_SELECTOR), |
6587 |
+- vmcs_read16(HOST_DS_SELECTOR), vmcs_read16(HOST_ES_SELECTOR), |
6588 |
+- vmcs_read16(HOST_FS_SELECTOR), vmcs_read16(HOST_GS_SELECTOR), |
6589 |
+- vmcs_read16(HOST_TR_SELECTOR)); |
6590 |
+- pr_err("FSBase=%016lx GSBase=%016lx TRBase=%016lx\n", |
6591 |
+- vmcs_readl(HOST_FS_BASE), vmcs_readl(HOST_GS_BASE), |
6592 |
+- vmcs_readl(HOST_TR_BASE)); |
6593 |
+- pr_err("GDTBase=%016lx IDTBase=%016lx\n", |
6594 |
+- vmcs_readl(HOST_GDTR_BASE), vmcs_readl(HOST_IDTR_BASE)); |
6595 |
+- pr_err("CR0=%016lx CR3=%016lx CR4=%016lx\n", |
6596 |
+- vmcs_readl(HOST_CR0), vmcs_readl(HOST_CR3), |
6597 |
+- vmcs_readl(HOST_CR4)); |
6598 |
+- pr_err("Sysenter RSP=%016lx CS:RIP=%04x:%016lx\n", |
6599 |
+- vmcs_readl(HOST_IA32_SYSENTER_ESP), |
6600 |
+- vmcs_read32(HOST_IA32_SYSENTER_CS), |
6601 |
+- vmcs_readl(HOST_IA32_SYSENTER_EIP)); |
6602 |
+- if (vmexit_ctl & (VM_EXIT_LOAD_IA32_PAT | VM_EXIT_LOAD_IA32_EFER)) |
6603 |
+- pr_err("EFER = 0x%016llx PAT = 0x%016llx\n", |
6604 |
+- vmcs_read64(HOST_IA32_EFER), |
6605 |
+- vmcs_read64(HOST_IA32_PAT)); |
6606 |
+- if (cpu_has_load_perf_global_ctrl() && |
6607 |
+- vmexit_ctl & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) |
6608 |
+- pr_err("PerfGlobCtl = 0x%016llx\n", |
6609 |
+- vmcs_read64(HOST_IA32_PERF_GLOBAL_CTRL)); |
6610 |
+- |
6611 |
+- pr_err("*** Control State ***\n"); |
6612 |
+- pr_err("PinBased=%08x CPUBased=%08x SecondaryExec=%08x\n", |
6613 |
+- pin_based_exec_ctrl, cpu_based_exec_ctrl, secondary_exec_control); |
6614 |
+- pr_err("EntryControls=%08x ExitControls=%08x\n", vmentry_ctl, vmexit_ctl); |
6615 |
+- pr_err("ExceptionBitmap=%08x PFECmask=%08x PFECmatch=%08x\n", |
6616 |
+- vmcs_read32(EXCEPTION_BITMAP), |
6617 |
+- vmcs_read32(PAGE_FAULT_ERROR_CODE_MASK), |
6618 |
+- vmcs_read32(PAGE_FAULT_ERROR_CODE_MATCH)); |
6619 |
+- pr_err("VMEntry: intr_info=%08x errcode=%08x ilen=%08x\n", |
6620 |
+- vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), |
6621 |
+- vmcs_read32(VM_ENTRY_EXCEPTION_ERROR_CODE), |
6622 |
+- vmcs_read32(VM_ENTRY_INSTRUCTION_LEN)); |
6623 |
+- pr_err("VMExit: intr_info=%08x errcode=%08x ilen=%08x\n", |
6624 |
+- vmcs_read32(VM_EXIT_INTR_INFO), |
6625 |
+- vmcs_read32(VM_EXIT_INTR_ERROR_CODE), |
6626 |
+- vmcs_read32(VM_EXIT_INSTRUCTION_LEN)); |
6627 |
+- pr_err(" reason=%08x qualification=%016lx\n", |
6628 |
+- vmcs_read32(VM_EXIT_REASON), vmcs_readl(EXIT_QUALIFICATION)); |
6629 |
+- pr_err("IDTVectoring: info=%08x errcode=%08x\n", |
6630 |
+- vmcs_read32(IDT_VECTORING_INFO_FIELD), |
6631 |
+- vmcs_read32(IDT_VECTORING_ERROR_CODE)); |
6632 |
+- pr_err("TSC Offset = 0x%016llx\n", vmcs_read64(TSC_OFFSET)); |
6633 |
+- if (secondary_exec_control & SECONDARY_EXEC_TSC_SCALING) |
6634 |
+- pr_err("TSC Multiplier = 0x%016llx\n", |
6635 |
+- vmcs_read64(TSC_MULTIPLIER)); |
6636 |
+- if (cpu_based_exec_ctrl & CPU_BASED_TPR_SHADOW) { |
6637 |
+- if (secondary_exec_control & SECONDARY_EXEC_VIRTUAL_INTR_DELIVERY) { |
6638 |
+- u16 status = vmcs_read16(GUEST_INTR_STATUS); |
6639 |
+- pr_err("SVI|RVI = %02x|%02x ", status >> 8, status & 0xff); |
6640 |
+- } |
6641 |
+- pr_cont("TPR Threshold = 0x%02x\n", vmcs_read32(TPR_THRESHOLD)); |
6642 |
+- if (secondary_exec_control & SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES) |
6643 |
+- pr_err("APIC-access addr = 0x%016llx ", vmcs_read64(APIC_ACCESS_ADDR)); |
6644 |
+- pr_cont("virt-APIC addr = 0x%016llx\n", vmcs_read64(VIRTUAL_APIC_PAGE_ADDR)); |
6645 |
+- } |
6646 |
+- if (pin_based_exec_ctrl & PIN_BASED_POSTED_INTR) |
6647 |
+- pr_err("PostedIntrVec = 0x%02x\n", vmcs_read16(POSTED_INTR_NV)); |
6648 |
+- if ((secondary_exec_control & SECONDARY_EXEC_ENABLE_EPT)) |
6649 |
+- pr_err("EPT pointer = 0x%016llx\n", vmcs_read64(EPT_POINTER)); |
6650 |
+- n = vmcs_read32(CR3_TARGET_COUNT); |
6651 |
+- for (i = 0; i + 1 < n; i += 4) |
6652 |
+- pr_err("CR3 target%u=%016lx target%u=%016lx\n", |
6653 |
+- i, vmcs_readl(CR3_TARGET_VALUE0 + i * 2), |
6654 |
+- i + 1, vmcs_readl(CR3_TARGET_VALUE0 + i * 2 + 2)); |
6655 |
+- if (i < n) |
6656 |
+- pr_err("CR3 target%u=%016lx\n", |
6657 |
+- i, vmcs_readl(CR3_TARGET_VALUE0 + i * 2)); |
6658 |
+- if (secondary_exec_control & SECONDARY_EXEC_PAUSE_LOOP_EXITING) |
6659 |
+- pr_err("PLE Gap=%08x Window=%08x\n", |
6660 |
+- vmcs_read32(PLE_GAP), vmcs_read32(PLE_WINDOW)); |
6661 |
+- if (secondary_exec_control & SECONDARY_EXEC_ENABLE_VPID) |
6662 |
+- pr_err("Virtual processor ID = 0x%04x\n", |
6663 |
+- vmcs_read16(VIRTUAL_PROCESSOR_ID)); |
6664 |
+-} |
6665 |
+- |
6666 |
+-/* |
6667 |
+- * The guest has exited. See if we can fix it or if we need userspace |
6668 |
+- * assistance. |
6669 |
+- */ |
6670 |
+-static int vmx_handle_exit(struct kvm_vcpu *vcpu, |
6671 |
+- enum exit_fastpath_completion exit_fastpath) |
6672 |
+-{ |
6673 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6674 |
+- u32 exit_reason = vmx->exit_reason; |
6675 |
+- u32 vectoring_info = vmx->idt_vectoring_info; |
6676 |
+- |
6677 |
+- trace_kvm_exit(exit_reason, vcpu, KVM_ISA_VMX); |
6678 |
+- |
6679 |
+- /* |
6680 |
+- * Flush logged GPAs PML buffer, this will make dirty_bitmap more |
6681 |
+- * updated. Another good is, in kvm_vm_ioctl_get_dirty_log, before |
6682 |
+- * querying dirty_bitmap, we only need to kick all vcpus out of guest |
6683 |
+- * mode as if vcpus is in root mode, the PML buffer must has been |
6684 |
+- * flushed already. |
6685 |
+- */ |
6686 |
+- if (enable_pml) |
6687 |
+- vmx_flush_pml_buffer(vcpu); |
6688 |
+- |
6689 |
+- /* If guest state is invalid, start emulating */ |
6690 |
+- if (vmx->emulation_required) |
6691 |
+- return handle_invalid_guest_state(vcpu); |
6692 |
+- |
6693 |
+- if (is_guest_mode(vcpu) && nested_vmx_exit_reflected(vcpu, exit_reason)) |
6694 |
+- return nested_vmx_reflect_vmexit(vcpu, exit_reason); |
6695 |
+- |
6696 |
+- if (exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) { |
6697 |
+- dump_vmcs(); |
6698 |
+- vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
6699 |
+- vcpu->run->fail_entry.hardware_entry_failure_reason |
6700 |
+- = exit_reason; |
6701 |
+- return 0; |
6702 |
+- } |
6703 |
+- |
6704 |
+- if (unlikely(vmx->fail)) { |
6705 |
+- dump_vmcs(); |
6706 |
+- vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
6707 |
+- vcpu->run->fail_entry.hardware_entry_failure_reason |
6708 |
+- = vmcs_read32(VM_INSTRUCTION_ERROR); |
6709 |
+- return 0; |
6710 |
+- } |
6711 |
+- |
6712 |
+- /* |
6713 |
+- * Note: |
6714 |
+- * Do not try to fix EXIT_REASON_EPT_MISCONFIG if it caused by |
6715 |
+- * delivery event since it indicates guest is accessing MMIO. |
6716 |
+- * The vm-exit can be triggered again after return to guest that |
6717 |
+- * will cause infinite loop. |
6718 |
+- */ |
6719 |
+- if ((vectoring_info & VECTORING_INFO_VALID_MASK) && |
6720 |
+- (exit_reason != EXIT_REASON_EXCEPTION_NMI && |
6721 |
+- exit_reason != EXIT_REASON_EPT_VIOLATION && |
6722 |
+- exit_reason != EXIT_REASON_PML_FULL && |
6723 |
+- exit_reason != EXIT_REASON_TASK_SWITCH)) { |
6724 |
+- vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
6725 |
+- vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_DELIVERY_EV; |
6726 |
+- vcpu->run->internal.ndata = 3; |
6727 |
+- vcpu->run->internal.data[0] = vectoring_info; |
6728 |
+- vcpu->run->internal.data[1] = exit_reason; |
6729 |
+- vcpu->run->internal.data[2] = vcpu->arch.exit_qualification; |
6730 |
+- if (exit_reason == EXIT_REASON_EPT_MISCONFIG) { |
6731 |
+- vcpu->run->internal.ndata++; |
6732 |
+- vcpu->run->internal.data[3] = |
6733 |
+- vmcs_read64(GUEST_PHYSICAL_ADDRESS); |
6734 |
+- } |
6735 |
+- return 0; |
6736 |
+- } |
6737 |
+- |
6738 |
+- if (unlikely(!enable_vnmi && |
6739 |
+- vmx->loaded_vmcs->soft_vnmi_blocked)) { |
6740 |
+- if (vmx_interrupt_allowed(vcpu)) { |
6741 |
+- vmx->loaded_vmcs->soft_vnmi_blocked = 0; |
6742 |
+- } else if (vmx->loaded_vmcs->vnmi_blocked_time > 1000000000LL && |
6743 |
+- vcpu->arch.nmi_pending) { |
6744 |
+- /* |
6745 |
+- * This CPU don't support us in finding the end of an |
6746 |
+- * NMI-blocked window if the guest runs with IRQs |
6747 |
+- * disabled. So we pull the trigger after 1 s of |
6748 |
+- * futile waiting, but inform the user about this. |
6749 |
+- */ |
6750 |
+- printk(KERN_WARNING "%s: Breaking out of NMI-blocked " |
6751 |
+- "state on VCPU %d after 1 s timeout\n", |
6752 |
+- __func__, vcpu->vcpu_id); |
6753 |
+- vmx->loaded_vmcs->soft_vnmi_blocked = 0; |
6754 |
+- } |
6755 |
+- } |
6756 |
+- |
6757 |
+- if (exit_fastpath == EXIT_FASTPATH_SKIP_EMUL_INS) { |
6758 |
+- kvm_skip_emulated_instruction(vcpu); |
6759 |
+- return 1; |
6760 |
+- } else if (exit_reason < kvm_vmx_max_exit_handlers |
6761 |
+- && kvm_vmx_exit_handlers[exit_reason]) { |
6762 |
+-#ifdef CONFIG_RETPOLINE |
6763 |
+- if (exit_reason == EXIT_REASON_MSR_WRITE) |
6764 |
+- return kvm_emulate_wrmsr(vcpu); |
6765 |
+- else if (exit_reason == EXIT_REASON_PREEMPTION_TIMER) |
6766 |
+- return handle_preemption_timer(vcpu); |
6767 |
+- else if (exit_reason == EXIT_REASON_INTERRUPT_WINDOW) |
6768 |
+- return handle_interrupt_window(vcpu); |
6769 |
+- else if (exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT) |
6770 |
+- return handle_external_interrupt(vcpu); |
6771 |
+- else if (exit_reason == EXIT_REASON_HLT) |
6772 |
+- return kvm_emulate_halt(vcpu); |
6773 |
+- else if (exit_reason == EXIT_REASON_EPT_MISCONFIG) |
6774 |
+- return handle_ept_misconfig(vcpu); |
6775 |
+-#endif |
6776 |
+- return kvm_vmx_exit_handlers[exit_reason](vcpu); |
6777 |
+- } else { |
6778 |
+- vcpu_unimpl(vcpu, "vmx: unexpected exit reason 0x%x\n", |
6779 |
+- exit_reason); |
6780 |
+- dump_vmcs(); |
6781 |
+- vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
6782 |
+- vcpu->run->internal.suberror = |
6783 |
+- KVM_INTERNAL_ERROR_UNEXPECTED_EXIT_REASON; |
6784 |
+- vcpu->run->internal.ndata = 1; |
6785 |
+- vcpu->run->internal.data[0] = exit_reason; |
6786 |
+- return 0; |
6787 |
+- } |
6788 |
+-} |
6789 |
+- |
6790 |
+-/* |
6791 |
+- * Software based L1D cache flush which is used when microcode providing |
6792 |
+- * the cache control MSR is not loaded. |
6793 |
+- * |
6794 |
+- * The L1D cache is 32 KiB on Nehalem and later microarchitectures, but to |
6795 |
+- * flush it is required to read in 64 KiB because the replacement algorithm |
6796 |
+- * is not exactly LRU. This could be sized at runtime via topology |
6797 |
+- * information but as all relevant affected CPUs have 32KiB L1D cache size |
6798 |
+- * there is no point in doing so. |
6799 |
+- */ |
6800 |
+-static void vmx_l1d_flush(struct kvm_vcpu *vcpu) |
6801 |
+-{ |
6802 |
+- int size = PAGE_SIZE << L1D_CACHE_ORDER; |
6803 |
+- |
6804 |
+- /* |
6805 |
+- * This code is only executed when the the flush mode is 'cond' or |
6806 |
+- * 'always' |
6807 |
+- */ |
6808 |
+- if (static_branch_likely(&vmx_l1d_flush_cond)) { |
6809 |
+- bool flush_l1d; |
6810 |
+- |
6811 |
+- /* |
6812 |
+- * Clear the per-vcpu flush bit, it gets set again |
6813 |
+- * either from vcpu_run() or from one of the unsafe |
6814 |
+- * VMEXIT handlers. |
6815 |
+- */ |
6816 |
+- flush_l1d = vcpu->arch.l1tf_flush_l1d; |
6817 |
+- vcpu->arch.l1tf_flush_l1d = false; |
6818 |
+- |
6819 |
+- /* |
6820 |
+- * Clear the per-cpu flush bit, it gets set again from |
6821 |
+- * the interrupt handlers. |
6822 |
+- */ |
6823 |
+- flush_l1d |= kvm_get_cpu_l1tf_flush_l1d(); |
6824 |
+- kvm_clear_cpu_l1tf_flush_l1d(); |
6825 |
+- |
6826 |
+- if (!flush_l1d) |
6827 |
+- return; |
6828 |
+- } |
6829 |
+- |
6830 |
+- vcpu->stat.l1d_flush++; |
6831 |
+- |
6832 |
+- if (static_cpu_has(X86_FEATURE_FLUSH_L1D)) { |
6833 |
+- wrmsrl(MSR_IA32_FLUSH_CMD, L1D_FLUSH); |
6834 |
+- return; |
6835 |
+- } |
6836 |
+- |
6837 |
+- asm volatile( |
6838 |
+- /* First ensure the pages are in the TLB */ |
6839 |
+- "xorl %%eax, %%eax\n" |
6840 |
+- ".Lpopulate_tlb:\n\t" |
6841 |
+- "movzbl (%[flush_pages], %%" _ASM_AX "), %%ecx\n\t" |
6842 |
+- "addl $4096, %%eax\n\t" |
6843 |
+- "cmpl %%eax, %[size]\n\t" |
6844 |
+- "jne .Lpopulate_tlb\n\t" |
6845 |
+- "xorl %%eax, %%eax\n\t" |
6846 |
+- "cpuid\n\t" |
6847 |
+- /* Now fill the cache */ |
6848 |
+- "xorl %%eax, %%eax\n" |
6849 |
+- ".Lfill_cache:\n" |
6850 |
+- "movzbl (%[flush_pages], %%" _ASM_AX "), %%ecx\n\t" |
6851 |
+- "addl $64, %%eax\n\t" |
6852 |
+- "cmpl %%eax, %[size]\n\t" |
6853 |
+- "jne .Lfill_cache\n\t" |
6854 |
+- "lfence\n" |
6855 |
+- :: [flush_pages] "r" (vmx_l1d_flush_pages), |
6856 |
+- [size] "r" (size) |
6857 |
+- : "eax", "ebx", "ecx", "edx"); |
6858 |
+-} |
6859 |
+- |
6860 |
+-static void update_cr8_intercept(struct kvm_vcpu *vcpu, int tpr, int irr) |
6861 |
+-{ |
6862 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
6863 |
+- int tpr_threshold; |
6864 |
+- |
6865 |
+- if (is_guest_mode(vcpu) && |
6866 |
+- nested_cpu_has(vmcs12, CPU_BASED_TPR_SHADOW)) |
6867 |
+- return; |
6868 |
+- |
6869 |
+- tpr_threshold = (irr == -1 || tpr < irr) ? 0 : irr; |
6870 |
+- if (is_guest_mode(vcpu)) |
6871 |
+- to_vmx(vcpu)->nested.l1_tpr_threshold = tpr_threshold; |
6872 |
+- else |
6873 |
+- vmcs_write32(TPR_THRESHOLD, tpr_threshold); |
6874 |
+-} |
6875 |
+- |
6876 |
+-void vmx_set_virtual_apic_mode(struct kvm_vcpu *vcpu) |
6877 |
+-{ |
6878 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6879 |
+- u32 sec_exec_control; |
6880 |
+- |
6881 |
+- if (!lapic_in_kernel(vcpu)) |
6882 |
+- return; |
6883 |
+- |
6884 |
+- if (!flexpriority_enabled && |
6885 |
+- !cpu_has_vmx_virtualize_x2apic_mode()) |
6886 |
+- return; |
6887 |
+- |
6888 |
+- /* Postpone execution until vmcs01 is the current VMCS. */ |
6889 |
+- if (is_guest_mode(vcpu)) { |
6890 |
+- vmx->nested.change_vmcs01_virtual_apic_mode = true; |
6891 |
+- return; |
6892 |
+- } |
6893 |
+- |
6894 |
+- sec_exec_control = secondary_exec_controls_get(vmx); |
6895 |
+- sec_exec_control &= ~(SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
6896 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE); |
6897 |
+- |
6898 |
+- switch (kvm_get_apic_mode(vcpu)) { |
6899 |
+- case LAPIC_MODE_INVALID: |
6900 |
+- WARN_ONCE(true, "Invalid local APIC state"); |
6901 |
+- case LAPIC_MODE_DISABLED: |
6902 |
+- break; |
6903 |
+- case LAPIC_MODE_XAPIC: |
6904 |
+- if (flexpriority_enabled) { |
6905 |
+- sec_exec_control |= |
6906 |
+- SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; |
6907 |
+- vmx_flush_tlb(vcpu, true); |
6908 |
+- } |
6909 |
+- break; |
6910 |
+- case LAPIC_MODE_X2APIC: |
6911 |
+- if (cpu_has_vmx_virtualize_x2apic_mode()) |
6912 |
+- sec_exec_control |= |
6913 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE; |
6914 |
+- break; |
6915 |
+- } |
6916 |
+- secondary_exec_controls_set(vmx, sec_exec_control); |
6917 |
+- |
6918 |
+- vmx_update_msr_bitmap(vcpu); |
6919 |
+-} |
6920 |
+- |
6921 |
+-static void vmx_set_apic_access_page_addr(struct kvm_vcpu *vcpu, hpa_t hpa) |
6922 |
+-{ |
6923 |
+- if (!is_guest_mode(vcpu)) { |
6924 |
+- vmcs_write64(APIC_ACCESS_ADDR, hpa); |
6925 |
+- vmx_flush_tlb(vcpu, true); |
6926 |
+- } |
6927 |
+-} |
6928 |
+- |
6929 |
+-static void vmx_hwapic_isr_update(struct kvm_vcpu *vcpu, int max_isr) |
6930 |
+-{ |
6931 |
+- u16 status; |
6932 |
+- u8 old; |
6933 |
+- |
6934 |
+- if (max_isr == -1) |
6935 |
+- max_isr = 0; |
6936 |
+- |
6937 |
+- status = vmcs_read16(GUEST_INTR_STATUS); |
6938 |
+- old = status >> 8; |
6939 |
+- if (max_isr != old) { |
6940 |
+- status &= 0xff; |
6941 |
+- status |= max_isr << 8; |
6942 |
+- vmcs_write16(GUEST_INTR_STATUS, status); |
6943 |
+- } |
6944 |
+-} |
6945 |
+- |
6946 |
+-static void vmx_set_rvi(int vector) |
6947 |
+-{ |
6948 |
+- u16 status; |
6949 |
+- u8 old; |
6950 |
+- |
6951 |
+- if (vector == -1) |
6952 |
+- vector = 0; |
6953 |
+- |
6954 |
+- status = vmcs_read16(GUEST_INTR_STATUS); |
6955 |
+- old = (u8)status & 0xff; |
6956 |
+- if ((u8)vector != old) { |
6957 |
+- status &= ~0xff; |
6958 |
+- status |= (u8)vector; |
6959 |
+- vmcs_write16(GUEST_INTR_STATUS, status); |
6960 |
+- } |
6961 |
+-} |
6962 |
+- |
6963 |
+-static void vmx_hwapic_irr_update(struct kvm_vcpu *vcpu, int max_irr) |
6964 |
+-{ |
6965 |
+- /* |
6966 |
+- * When running L2, updating RVI is only relevant when |
6967 |
+- * vmcs12 virtual-interrupt-delivery enabled. |
6968 |
+- * However, it can be enabled only when L1 also |
6969 |
+- * intercepts external-interrupts and in that case |
6970 |
+- * we should not update vmcs02 RVI but instead intercept |
6971 |
+- * interrupt. Therefore, do nothing when running L2. |
6972 |
+- */ |
6973 |
+- if (!is_guest_mode(vcpu)) |
6974 |
+- vmx_set_rvi(max_irr); |
6975 |
+-} |
6976 |
+- |
6977 |
+-static int vmx_sync_pir_to_irr(struct kvm_vcpu *vcpu) |
6978 |
+-{ |
6979 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
6980 |
+- int max_irr; |
6981 |
+- bool max_irr_updated; |
6982 |
+- |
6983 |
+- WARN_ON(!vcpu->arch.apicv_active); |
6984 |
+- if (pi_test_on(&vmx->pi_desc)) { |
6985 |
+- pi_clear_on(&vmx->pi_desc); |
6986 |
+- /* |
6987 |
+- * IOMMU can write to PID.ON, so the barrier matters even on UP. |
6988 |
+- * But on x86 this is just a compiler barrier anyway. |
6989 |
+- */ |
6990 |
+- smp_mb__after_atomic(); |
6991 |
+- max_irr_updated = |
6992 |
+- kvm_apic_update_irr(vcpu, vmx->pi_desc.pir, &max_irr); |
6993 |
+- |
6994 |
+- /* |
6995 |
+- * If we are running L2 and L1 has a new pending interrupt |
6996 |
+- * which can be injected, we should re-evaluate |
6997 |
+- * what should be done with this new L1 interrupt. |
6998 |
+- * If L1 intercepts external-interrupts, we should |
6999 |
+- * exit from L2 to L1. Otherwise, interrupt should be |
7000 |
+- * delivered directly to L2. |
7001 |
+- */ |
7002 |
+- if (is_guest_mode(vcpu) && max_irr_updated) { |
7003 |
+- if (nested_exit_on_intr(vcpu)) |
7004 |
+- kvm_vcpu_exiting_guest_mode(vcpu); |
7005 |
+- else |
7006 |
+- kvm_make_request(KVM_REQ_EVENT, vcpu); |
7007 |
+- } |
7008 |
+- } else { |
7009 |
+- max_irr = kvm_lapic_find_highest_irr(vcpu); |
7010 |
+- } |
7011 |
+- vmx_hwapic_irr_update(vcpu, max_irr); |
7012 |
+- return max_irr; |
7013 |
+-} |
7014 |
+- |
7015 |
+-static bool vmx_dy_apicv_has_pending_interrupt(struct kvm_vcpu *vcpu) |
7016 |
+-{ |
7017 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
7018 |
+- |
7019 |
+- return pi_test_on(pi_desc) || |
7020 |
+- (pi_test_sn(pi_desc) && !pi_is_pir_empty(pi_desc)); |
7021 |
+-} |
7022 |
+- |
7023 |
+-static void vmx_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap) |
7024 |
+-{ |
7025 |
+- if (!kvm_vcpu_apicv_active(vcpu)) |
7026 |
+- return; |
7027 |
+- |
7028 |
+- vmcs_write64(EOI_EXIT_BITMAP0, eoi_exit_bitmap[0]); |
7029 |
+- vmcs_write64(EOI_EXIT_BITMAP1, eoi_exit_bitmap[1]); |
7030 |
+- vmcs_write64(EOI_EXIT_BITMAP2, eoi_exit_bitmap[2]); |
7031 |
+- vmcs_write64(EOI_EXIT_BITMAP3, eoi_exit_bitmap[3]); |
7032 |
+-} |
7033 |
+- |
7034 |
+-static void vmx_apicv_post_state_restore(struct kvm_vcpu *vcpu) |
7035 |
+-{ |
7036 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7037 |
+- |
7038 |
+- pi_clear_on(&vmx->pi_desc); |
7039 |
+- memset(vmx->pi_desc.pir, 0, sizeof(vmx->pi_desc.pir)); |
7040 |
+-} |
7041 |
+- |
7042 |
+-static void handle_exception_nmi_irqoff(struct vcpu_vmx *vmx) |
7043 |
+-{ |
7044 |
+- vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7045 |
+- |
7046 |
+- /* if exit due to PF check for async PF */ |
7047 |
+- if (is_page_fault(vmx->exit_intr_info)) |
7048 |
+- vmx->vcpu.arch.apf.host_apf_reason = kvm_read_and_reset_pf_reason(); |
7049 |
+- |
7050 |
+- /* Handle machine checks before interrupts are enabled */ |
7051 |
+- if (is_machine_check(vmx->exit_intr_info)) |
7052 |
+- kvm_machine_check(); |
7053 |
+- |
7054 |
+- /* We need to handle NMIs before interrupts are enabled */ |
7055 |
+- if (is_nmi(vmx->exit_intr_info)) { |
7056 |
+- kvm_before_interrupt(&vmx->vcpu); |
7057 |
+- asm("int $2"); |
7058 |
+- kvm_after_interrupt(&vmx->vcpu); |
7059 |
+- } |
7060 |
+-} |
7061 |
+- |
7062 |
+-static void handle_external_interrupt_irqoff(struct kvm_vcpu *vcpu) |
7063 |
+-{ |
7064 |
+- unsigned int vector; |
7065 |
+- unsigned long entry; |
7066 |
+-#ifdef CONFIG_X86_64 |
7067 |
+- unsigned long tmp; |
7068 |
+-#endif |
7069 |
+- gate_desc *desc; |
7070 |
+- u32 intr_info; |
7071 |
+- |
7072 |
+- intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7073 |
+- if (WARN_ONCE(!is_external_intr(intr_info), |
7074 |
+- "KVM: unexpected VM-Exit interrupt info: 0x%x", intr_info)) |
7075 |
+- return; |
7076 |
+- |
7077 |
+- vector = intr_info & INTR_INFO_VECTOR_MASK; |
7078 |
+- desc = (gate_desc *)host_idt_base + vector; |
7079 |
+- entry = gate_offset(desc); |
7080 |
+- |
7081 |
+- kvm_before_interrupt(vcpu); |
7082 |
+- |
7083 |
+- asm volatile( |
7084 |
+-#ifdef CONFIG_X86_64 |
7085 |
+- "mov %%" _ASM_SP ", %[sp]\n\t" |
7086 |
+- "and $0xfffffffffffffff0, %%" _ASM_SP "\n\t" |
7087 |
+- "push $%c[ss]\n\t" |
7088 |
+- "push %[sp]\n\t" |
7089 |
+-#endif |
7090 |
+- "pushf\n\t" |
7091 |
+- __ASM_SIZE(push) " $%c[cs]\n\t" |
7092 |
+- CALL_NOSPEC |
7093 |
+- : |
7094 |
+-#ifdef CONFIG_X86_64 |
7095 |
+- [sp]"=&r"(tmp), |
7096 |
+-#endif |
7097 |
+- ASM_CALL_CONSTRAINT |
7098 |
+- : |
7099 |
+- THUNK_TARGET(entry), |
7100 |
+- [ss]"i"(__KERNEL_DS), |
7101 |
+- [cs]"i"(__KERNEL_CS) |
7102 |
+- ); |
7103 |
+- |
7104 |
+- kvm_after_interrupt(vcpu); |
7105 |
+-} |
7106 |
+-STACK_FRAME_NON_STANDARD(handle_external_interrupt_irqoff); |
7107 |
+- |
7108 |
+-static void vmx_handle_exit_irqoff(struct kvm_vcpu *vcpu, |
7109 |
+- enum exit_fastpath_completion *exit_fastpath) |
7110 |
+-{ |
7111 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7112 |
+- |
7113 |
+- if (vmx->exit_reason == EXIT_REASON_EXTERNAL_INTERRUPT) |
7114 |
+- handle_external_interrupt_irqoff(vcpu); |
7115 |
+- else if (vmx->exit_reason == EXIT_REASON_EXCEPTION_NMI) |
7116 |
+- handle_exception_nmi_irqoff(vmx); |
7117 |
+- else if (!is_guest_mode(vcpu) && |
7118 |
+- vmx->exit_reason == EXIT_REASON_MSR_WRITE) |
7119 |
+- *exit_fastpath = handle_fastpath_set_msr_irqoff(vcpu); |
7120 |
+-} |
7121 |
+- |
7122 |
+-static bool vmx_has_emulated_msr(int index) |
7123 |
+-{ |
7124 |
+- switch (index) { |
7125 |
+- case MSR_IA32_SMBASE: |
7126 |
+- /* |
7127 |
+- * We cannot do SMM unless we can run the guest in big |
7128 |
+- * real mode. |
7129 |
+- */ |
7130 |
+- return enable_unrestricted_guest || emulate_invalid_guest_state; |
7131 |
+- case MSR_IA32_VMX_BASIC ... MSR_IA32_VMX_VMFUNC: |
7132 |
+- return nested; |
7133 |
+- case MSR_AMD64_VIRT_SPEC_CTRL: |
7134 |
+- /* This is AMD only. */ |
7135 |
+- return false; |
7136 |
+- default: |
7137 |
+- return true; |
7138 |
+- } |
7139 |
+-} |
7140 |
+- |
7141 |
+-static bool vmx_pt_supported(void) |
7142 |
+-{ |
7143 |
+- return pt_mode == PT_MODE_HOST_GUEST; |
7144 |
+-} |
7145 |
+- |
7146 |
+-static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx) |
7147 |
+-{ |
7148 |
+- u32 exit_intr_info; |
7149 |
+- bool unblock_nmi; |
7150 |
+- u8 vector; |
7151 |
+- bool idtv_info_valid; |
7152 |
+- |
7153 |
+- idtv_info_valid = vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK; |
7154 |
+- |
7155 |
+- if (enable_vnmi) { |
7156 |
+- if (vmx->loaded_vmcs->nmi_known_unmasked) |
7157 |
+- return; |
7158 |
+- /* |
7159 |
+- * Can't use vmx->exit_intr_info since we're not sure what |
7160 |
+- * the exit reason is. |
7161 |
+- */ |
7162 |
+- exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
7163 |
+- unblock_nmi = (exit_intr_info & INTR_INFO_UNBLOCK_NMI) != 0; |
7164 |
+- vector = exit_intr_info & INTR_INFO_VECTOR_MASK; |
7165 |
+- /* |
7166 |
+- * SDM 3: 27.7.1.2 (September 2008) |
7167 |
+- * Re-set bit "block by NMI" before VM entry if vmexit caused by |
7168 |
+- * a guest IRET fault. |
7169 |
+- * SDM 3: 23.2.2 (September 2008) |
7170 |
+- * Bit 12 is undefined in any of the following cases: |
7171 |
+- * If the VM exit sets the valid bit in the IDT-vectoring |
7172 |
+- * information field. |
7173 |
+- * If the VM exit is due to a double fault. |
7174 |
+- */ |
7175 |
+- if ((exit_intr_info & INTR_INFO_VALID_MASK) && unblock_nmi && |
7176 |
+- vector != DF_VECTOR && !idtv_info_valid) |
7177 |
+- vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
7178 |
+- GUEST_INTR_STATE_NMI); |
7179 |
+- else |
7180 |
+- vmx->loaded_vmcs->nmi_known_unmasked = |
7181 |
+- !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) |
7182 |
+- & GUEST_INTR_STATE_NMI); |
7183 |
+- } else if (unlikely(vmx->loaded_vmcs->soft_vnmi_blocked)) |
7184 |
+- vmx->loaded_vmcs->vnmi_blocked_time += |
7185 |
+- ktime_to_ns(ktime_sub(ktime_get(), |
7186 |
+- vmx->loaded_vmcs->entry_time)); |
7187 |
+-} |
7188 |
+- |
7189 |
+-static void __vmx_complete_interrupts(struct kvm_vcpu *vcpu, |
7190 |
+- u32 idt_vectoring_info, |
7191 |
+- int instr_len_field, |
7192 |
+- int error_code_field) |
7193 |
+-{ |
7194 |
+- u8 vector; |
7195 |
+- int type; |
7196 |
+- bool idtv_info_valid; |
7197 |
+- |
7198 |
+- idtv_info_valid = idt_vectoring_info & VECTORING_INFO_VALID_MASK; |
7199 |
+- |
7200 |
+- vcpu->arch.nmi_injected = false; |
7201 |
+- kvm_clear_exception_queue(vcpu); |
7202 |
+- kvm_clear_interrupt_queue(vcpu); |
7203 |
+- |
7204 |
+- if (!idtv_info_valid) |
7205 |
+- return; |
7206 |
+- |
7207 |
+- kvm_make_request(KVM_REQ_EVENT, vcpu); |
7208 |
+- |
7209 |
+- vector = idt_vectoring_info & VECTORING_INFO_VECTOR_MASK; |
7210 |
+- type = idt_vectoring_info & VECTORING_INFO_TYPE_MASK; |
7211 |
+- |
7212 |
+- switch (type) { |
7213 |
+- case INTR_TYPE_NMI_INTR: |
7214 |
+- vcpu->arch.nmi_injected = true; |
7215 |
+- /* |
7216 |
+- * SDM 3: 27.7.1.2 (September 2008) |
7217 |
+- * Clear bit "block by NMI" before VM entry if a NMI |
7218 |
+- * delivery faulted. |
7219 |
+- */ |
7220 |
+- vmx_set_nmi_mask(vcpu, false); |
7221 |
+- break; |
7222 |
+- case INTR_TYPE_SOFT_EXCEPTION: |
7223 |
+- vcpu->arch.event_exit_inst_len = vmcs_read32(instr_len_field); |
7224 |
+- /* fall through */ |
7225 |
+- case INTR_TYPE_HARD_EXCEPTION: |
7226 |
+- if (idt_vectoring_info & VECTORING_INFO_DELIVER_CODE_MASK) { |
7227 |
+- u32 err = vmcs_read32(error_code_field); |
7228 |
+- kvm_requeue_exception_e(vcpu, vector, err); |
7229 |
+- } else |
7230 |
+- kvm_requeue_exception(vcpu, vector); |
7231 |
+- break; |
7232 |
+- case INTR_TYPE_SOFT_INTR: |
7233 |
+- vcpu->arch.event_exit_inst_len = vmcs_read32(instr_len_field); |
7234 |
+- /* fall through */ |
7235 |
+- case INTR_TYPE_EXT_INTR: |
7236 |
+- kvm_queue_interrupt(vcpu, vector, type == INTR_TYPE_SOFT_INTR); |
7237 |
+- break; |
7238 |
+- default: |
7239 |
+- break; |
7240 |
+- } |
7241 |
+-} |
7242 |
+- |
7243 |
+-static void vmx_complete_interrupts(struct vcpu_vmx *vmx) |
7244 |
+-{ |
7245 |
+- __vmx_complete_interrupts(&vmx->vcpu, vmx->idt_vectoring_info, |
7246 |
+- VM_EXIT_INSTRUCTION_LEN, |
7247 |
+- IDT_VECTORING_ERROR_CODE); |
7248 |
+-} |
7249 |
+- |
7250 |
+-static void vmx_cancel_injection(struct kvm_vcpu *vcpu) |
7251 |
+-{ |
7252 |
+- __vmx_complete_interrupts(vcpu, |
7253 |
+- vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), |
7254 |
+- VM_ENTRY_INSTRUCTION_LEN, |
7255 |
+- VM_ENTRY_EXCEPTION_ERROR_CODE); |
7256 |
+- |
7257 |
+- vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); |
7258 |
+-} |
7259 |
+- |
7260 |
+-static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) |
7261 |
+-{ |
7262 |
+- int i, nr_msrs; |
7263 |
+- struct perf_guest_switch_msr *msrs; |
7264 |
+- |
7265 |
+- msrs = perf_guest_get_msrs(&nr_msrs); |
7266 |
+- |
7267 |
+- if (!msrs) |
7268 |
+- return; |
7269 |
+- |
7270 |
+- for (i = 0; i < nr_msrs; i++) |
7271 |
+- if (msrs[i].host == msrs[i].guest) |
7272 |
+- clear_atomic_switch_msr(vmx, msrs[i].msr); |
7273 |
+- else |
7274 |
+- add_atomic_switch_msr(vmx, msrs[i].msr, msrs[i].guest, |
7275 |
+- msrs[i].host, false); |
7276 |
+-} |
7277 |
+- |
7278 |
+-static void atomic_switch_umwait_control_msr(struct vcpu_vmx *vmx) |
7279 |
+-{ |
7280 |
+- u32 host_umwait_control; |
7281 |
+- |
7282 |
+- if (!vmx_has_waitpkg(vmx)) |
7283 |
+- return; |
7284 |
+- |
7285 |
+- host_umwait_control = get_umwait_control_msr(); |
7286 |
+- |
7287 |
+- if (vmx->msr_ia32_umwait_control != host_umwait_control) |
7288 |
+- add_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL, |
7289 |
+- vmx->msr_ia32_umwait_control, |
7290 |
+- host_umwait_control, false); |
7291 |
+- else |
7292 |
+- clear_atomic_switch_msr(vmx, MSR_IA32_UMWAIT_CONTROL); |
7293 |
+-} |
7294 |
+- |
7295 |
+-static void vmx_update_hv_timer(struct kvm_vcpu *vcpu) |
7296 |
+-{ |
7297 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7298 |
+- u64 tscl; |
7299 |
+- u32 delta_tsc; |
7300 |
+- |
7301 |
+- if (vmx->req_immediate_exit) { |
7302 |
+- vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, 0); |
7303 |
+- vmx->loaded_vmcs->hv_timer_soft_disabled = false; |
7304 |
+- } else if (vmx->hv_deadline_tsc != -1) { |
7305 |
+- tscl = rdtsc(); |
7306 |
+- if (vmx->hv_deadline_tsc > tscl) |
7307 |
+- /* set_hv_timer ensures the delta fits in 32-bits */ |
7308 |
+- delta_tsc = (u32)((vmx->hv_deadline_tsc - tscl) >> |
7309 |
+- cpu_preemption_timer_multi); |
7310 |
+- else |
7311 |
+- delta_tsc = 0; |
7312 |
+- |
7313 |
+- vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, delta_tsc); |
7314 |
+- vmx->loaded_vmcs->hv_timer_soft_disabled = false; |
7315 |
+- } else if (!vmx->loaded_vmcs->hv_timer_soft_disabled) { |
7316 |
+- vmcs_write32(VMX_PREEMPTION_TIMER_VALUE, -1); |
7317 |
+- vmx->loaded_vmcs->hv_timer_soft_disabled = true; |
7318 |
+- } |
7319 |
+-} |
7320 |
+- |
7321 |
+-void vmx_update_host_rsp(struct vcpu_vmx *vmx, unsigned long host_rsp) |
7322 |
+-{ |
7323 |
+- if (unlikely(host_rsp != vmx->loaded_vmcs->host_state.rsp)) { |
7324 |
+- vmx->loaded_vmcs->host_state.rsp = host_rsp; |
7325 |
+- vmcs_writel(HOST_RSP, host_rsp); |
7326 |
+- } |
7327 |
+-} |
7328 |
+- |
7329 |
+-bool __vmx_vcpu_run(struct vcpu_vmx *vmx, unsigned long *regs, bool launched); |
7330 |
+- |
7331 |
+-static void vmx_vcpu_run(struct kvm_vcpu *vcpu) |
7332 |
+-{ |
7333 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7334 |
+- unsigned long cr3, cr4; |
7335 |
+- |
7336 |
+- /* Record the guest's net vcpu time for enforced NMI injections. */ |
7337 |
+- if (unlikely(!enable_vnmi && |
7338 |
+- vmx->loaded_vmcs->soft_vnmi_blocked)) |
7339 |
+- vmx->loaded_vmcs->entry_time = ktime_get(); |
7340 |
+- |
7341 |
+- /* Don't enter VMX if guest state is invalid, let the exit handler |
7342 |
+- start emulation until we arrive back to a valid state */ |
7343 |
+- if (vmx->emulation_required) |
7344 |
+- return; |
7345 |
+- |
7346 |
+- if (vmx->ple_window_dirty) { |
7347 |
+- vmx->ple_window_dirty = false; |
7348 |
+- vmcs_write32(PLE_WINDOW, vmx->ple_window); |
7349 |
+- } |
7350 |
+- |
7351 |
+- if (vmx->nested.need_vmcs12_to_shadow_sync) |
7352 |
+- nested_sync_vmcs12_to_shadow(vcpu); |
7353 |
+- |
7354 |
+- if (kvm_register_is_dirty(vcpu, VCPU_REGS_RSP)) |
7355 |
+- vmcs_writel(GUEST_RSP, vcpu->arch.regs[VCPU_REGS_RSP]); |
7356 |
+- if (kvm_register_is_dirty(vcpu, VCPU_REGS_RIP)) |
7357 |
+- vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]); |
7358 |
+- |
7359 |
+- cr3 = __get_current_cr3_fast(); |
7360 |
+- if (unlikely(cr3 != vmx->loaded_vmcs->host_state.cr3)) { |
7361 |
+- vmcs_writel(HOST_CR3, cr3); |
7362 |
+- vmx->loaded_vmcs->host_state.cr3 = cr3; |
7363 |
+- } |
7364 |
+- |
7365 |
+- cr4 = cr4_read_shadow(); |
7366 |
+- if (unlikely(cr4 != vmx->loaded_vmcs->host_state.cr4)) { |
7367 |
+- vmcs_writel(HOST_CR4, cr4); |
7368 |
+- vmx->loaded_vmcs->host_state.cr4 = cr4; |
7369 |
+- } |
7370 |
+- |
7371 |
+- /* When single-stepping over STI and MOV SS, we must clear the |
7372 |
+- * corresponding interruptibility bits in the guest state. Otherwise |
7373 |
+- * vmentry fails as it then expects bit 14 (BS) in pending debug |
7374 |
+- * exceptions being set, but that's not correct for the guest debugging |
7375 |
+- * case. */ |
7376 |
+- if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) |
7377 |
+- vmx_set_interrupt_shadow(vcpu, 0); |
7378 |
+- |
7379 |
+- kvm_load_guest_xsave_state(vcpu); |
7380 |
+- |
7381 |
+- if (static_cpu_has(X86_FEATURE_PKU) && |
7382 |
+- kvm_read_cr4_bits(vcpu, X86_CR4_PKE) && |
7383 |
+- vcpu->arch.pkru != vmx->host_pkru) |
7384 |
+- __write_pkru(vcpu->arch.pkru); |
7385 |
+- |
7386 |
+- pt_guest_enter(vmx); |
7387 |
+- |
7388 |
+- atomic_switch_perf_msrs(vmx); |
7389 |
+- atomic_switch_umwait_control_msr(vmx); |
7390 |
+- |
7391 |
+- if (enable_preemption_timer) |
7392 |
+- vmx_update_hv_timer(vcpu); |
7393 |
+- |
7394 |
+- if (lapic_in_kernel(vcpu) && |
7395 |
+- vcpu->arch.apic->lapic_timer.timer_advance_ns) |
7396 |
+- kvm_wait_lapic_expire(vcpu); |
7397 |
+- |
7398 |
+- /* |
7399 |
+- * If this vCPU has touched SPEC_CTRL, restore the guest's value if |
7400 |
+- * it's non-zero. Since vmentry is serialising on affected CPUs, there |
7401 |
+- * is no need to worry about the conditional branch over the wrmsr |
7402 |
+- * being speculatively taken. |
7403 |
+- */ |
7404 |
+- x86_spec_ctrl_set_guest(vmx->spec_ctrl, 0); |
7405 |
+- |
7406 |
+- /* L1D Flush includes CPU buffer clear to mitigate MDS */ |
7407 |
+- if (static_branch_unlikely(&vmx_l1d_should_flush)) |
7408 |
+- vmx_l1d_flush(vcpu); |
7409 |
+- else if (static_branch_unlikely(&mds_user_clear)) |
7410 |
+- mds_clear_cpu_buffers(); |
7411 |
+- |
7412 |
+- if (vcpu->arch.cr2 != read_cr2()) |
7413 |
+- write_cr2(vcpu->arch.cr2); |
7414 |
+- |
7415 |
+- vmx->fail = __vmx_vcpu_run(vmx, (unsigned long *)&vcpu->arch.regs, |
7416 |
+- vmx->loaded_vmcs->launched); |
7417 |
+- |
7418 |
+- vcpu->arch.cr2 = read_cr2(); |
7419 |
+- |
7420 |
+- /* |
7421 |
+- * We do not use IBRS in the kernel. If this vCPU has used the |
7422 |
+- * SPEC_CTRL MSR it may have left it on; save the value and |
7423 |
+- * turn it off. This is much more efficient than blindly adding |
7424 |
+- * it to the atomic save/restore list. Especially as the former |
7425 |
+- * (Saving guest MSRs on vmexit) doesn't even exist in KVM. |
7426 |
+- * |
7427 |
+- * For non-nested case: |
7428 |
+- * If the L01 MSR bitmap does not intercept the MSR, then we need to |
7429 |
+- * save it. |
7430 |
+- * |
7431 |
+- * For nested case: |
7432 |
+- * If the L02 MSR bitmap does not intercept the MSR, then we need to |
7433 |
+- * save it. |
7434 |
+- */ |
7435 |
+- if (unlikely(!msr_write_intercepted(vcpu, MSR_IA32_SPEC_CTRL))) |
7436 |
+- vmx->spec_ctrl = native_read_msr(MSR_IA32_SPEC_CTRL); |
7437 |
+- |
7438 |
+- x86_spec_ctrl_restore_host(vmx->spec_ctrl, 0); |
7439 |
+- |
7440 |
+- /* All fields are clean at this point */ |
7441 |
+- if (static_branch_unlikely(&enable_evmcs)) |
7442 |
+- current_evmcs->hv_clean_fields |= |
7443 |
+- HV_VMX_ENLIGHTENED_CLEAN_FIELD_ALL; |
7444 |
+- |
7445 |
+- if (static_branch_unlikely(&enable_evmcs)) |
7446 |
+- current_evmcs->hv_vp_id = vcpu->arch.hyperv.vp_index; |
7447 |
+- |
7448 |
+- /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */ |
7449 |
+- if (vmx->host_debugctlmsr) |
7450 |
+- update_debugctlmsr(vmx->host_debugctlmsr); |
7451 |
+- |
7452 |
+-#ifndef CONFIG_X86_64 |
7453 |
+- /* |
7454 |
+- * The sysexit path does not restore ds/es, so we must set them to |
7455 |
+- * a reasonable value ourselves. |
7456 |
+- * |
7457 |
+- * We can't defer this to vmx_prepare_switch_to_host() since that |
7458 |
+- * function may be executed in interrupt context, which saves and |
7459 |
+- * restore segments around it, nullifying its effect. |
7460 |
+- */ |
7461 |
+- loadsegment(ds, __USER_DS); |
7462 |
+- loadsegment(es, __USER_DS); |
7463 |
+-#endif |
7464 |
+- |
7465 |
+- vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP) |
7466 |
+- | (1 << VCPU_EXREG_RFLAGS) |
7467 |
+- | (1 << VCPU_EXREG_PDPTR) |
7468 |
+- | (1 << VCPU_EXREG_SEGMENTS) |
7469 |
+- | (1 << VCPU_EXREG_CR3)); |
7470 |
+- vcpu->arch.regs_dirty = 0; |
7471 |
+- |
7472 |
+- pt_guest_exit(vmx); |
7473 |
+- |
7474 |
+- /* |
7475 |
+- * eager fpu is enabled if PKEY is supported and CR4 is switched |
7476 |
+- * back on host, so it is safe to read guest PKRU from current |
7477 |
+- * XSAVE. |
7478 |
+- */ |
7479 |
+- if (static_cpu_has(X86_FEATURE_PKU) && |
7480 |
+- kvm_read_cr4_bits(vcpu, X86_CR4_PKE)) { |
7481 |
+- vcpu->arch.pkru = rdpkru(); |
7482 |
+- if (vcpu->arch.pkru != vmx->host_pkru) |
7483 |
+- __write_pkru(vmx->host_pkru); |
7484 |
+- } |
7485 |
+- |
7486 |
+- kvm_load_host_xsave_state(vcpu); |
7487 |
+- |
7488 |
+- vmx->nested.nested_run_pending = 0; |
7489 |
+- vmx->idt_vectoring_info = 0; |
7490 |
+- |
7491 |
+- vmx->exit_reason = vmx->fail ? 0xdead : vmcs_read32(VM_EXIT_REASON); |
7492 |
+- if ((u16)vmx->exit_reason == EXIT_REASON_MCE_DURING_VMENTRY) |
7493 |
+- kvm_machine_check(); |
7494 |
+- |
7495 |
+- if (vmx->fail || (vmx->exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY)) |
7496 |
+- return; |
7497 |
+- |
7498 |
+- vmx->loaded_vmcs->launched = 1; |
7499 |
+- vmx->idt_vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD); |
7500 |
+- |
7501 |
+- vmx_recover_nmi_blocking(vmx); |
7502 |
+- vmx_complete_interrupts(vmx); |
7503 |
+-} |
7504 |
+- |
7505 |
+-static struct kvm *vmx_vm_alloc(void) |
7506 |
+-{ |
7507 |
+- struct kvm_vmx *kvm_vmx = __vmalloc(sizeof(struct kvm_vmx), |
7508 |
+- GFP_KERNEL_ACCOUNT | __GFP_ZERO, |
7509 |
+- PAGE_KERNEL); |
7510 |
+- return &kvm_vmx->kvm; |
7511 |
+-} |
7512 |
+- |
7513 |
+-static void vmx_vm_free(struct kvm *kvm) |
7514 |
+-{ |
7515 |
+- kfree(kvm->arch.hyperv.hv_pa_pg); |
7516 |
+- vfree(to_kvm_vmx(kvm)); |
7517 |
+-} |
7518 |
+- |
7519 |
+-static void vmx_free_vcpu(struct kvm_vcpu *vcpu) |
7520 |
+-{ |
7521 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7522 |
+- |
7523 |
+- if (enable_pml) |
7524 |
+- vmx_destroy_pml_buffer(vmx); |
7525 |
+- free_vpid(vmx->vpid); |
7526 |
+- nested_vmx_free_vcpu(vcpu); |
7527 |
+- free_loaded_vmcs(vmx->loaded_vmcs); |
7528 |
+- kvm_vcpu_uninit(vcpu); |
7529 |
+- kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu); |
7530 |
+- kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu); |
7531 |
+- kmem_cache_free(kvm_vcpu_cache, vmx); |
7532 |
+-} |
7533 |
+- |
7534 |
+-static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id) |
7535 |
+-{ |
7536 |
+- int err; |
7537 |
+- struct vcpu_vmx *vmx; |
7538 |
+- unsigned long *msr_bitmap; |
7539 |
+- int i, cpu; |
7540 |
+- |
7541 |
+- BUILD_BUG_ON_MSG(offsetof(struct vcpu_vmx, vcpu) != 0, |
7542 |
+- "struct kvm_vcpu must be at offset 0 for arch usercopy region"); |
7543 |
+- |
7544 |
+- vmx = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL_ACCOUNT); |
7545 |
+- if (!vmx) |
7546 |
+- return ERR_PTR(-ENOMEM); |
7547 |
+- |
7548 |
+- vmx->vcpu.arch.user_fpu = kmem_cache_zalloc(x86_fpu_cache, |
7549 |
+- GFP_KERNEL_ACCOUNT); |
7550 |
+- if (!vmx->vcpu.arch.user_fpu) { |
7551 |
+- printk(KERN_ERR "kvm: failed to allocate kvm userspace's fpu\n"); |
7552 |
+- err = -ENOMEM; |
7553 |
+- goto free_partial_vcpu; |
7554 |
+- } |
7555 |
+- |
7556 |
+- vmx->vcpu.arch.guest_fpu = kmem_cache_zalloc(x86_fpu_cache, |
7557 |
+- GFP_KERNEL_ACCOUNT); |
7558 |
+- if (!vmx->vcpu.arch.guest_fpu) { |
7559 |
+- printk(KERN_ERR "kvm: failed to allocate vcpu's fpu\n"); |
7560 |
+- err = -ENOMEM; |
7561 |
+- goto free_user_fpu; |
7562 |
+- } |
7563 |
+- |
7564 |
+- vmx->vpid = allocate_vpid(); |
7565 |
+- |
7566 |
+- err = kvm_vcpu_init(&vmx->vcpu, kvm, id); |
7567 |
+- if (err) |
7568 |
+- goto free_vcpu; |
7569 |
+- |
7570 |
+- err = -ENOMEM; |
7571 |
+- |
7572 |
+- /* |
7573 |
+- * If PML is turned on, failure on enabling PML just results in failure |
7574 |
+- * of creating the vcpu, therefore we can simplify PML logic (by |
7575 |
+- * avoiding dealing with cases, such as enabling PML partially on vcpus |
7576 |
+- * for the guest), etc. |
7577 |
+- */ |
7578 |
+- if (enable_pml) { |
7579 |
+- vmx->pml_pg = alloc_page(GFP_KERNEL_ACCOUNT | __GFP_ZERO); |
7580 |
+- if (!vmx->pml_pg) |
7581 |
+- goto uninit_vcpu; |
7582 |
+- } |
7583 |
+- |
7584 |
+- BUILD_BUG_ON(ARRAY_SIZE(vmx_msr_index) != NR_SHARED_MSRS); |
7585 |
+- |
7586 |
+- for (i = 0; i < ARRAY_SIZE(vmx_msr_index); ++i) { |
7587 |
+- u32 index = vmx_msr_index[i]; |
7588 |
+- u32 data_low, data_high; |
7589 |
+- int j = vmx->nmsrs; |
7590 |
+- |
7591 |
+- if (rdmsr_safe(index, &data_low, &data_high) < 0) |
7592 |
+- continue; |
7593 |
+- if (wrmsr_safe(index, data_low, data_high) < 0) |
7594 |
+- continue; |
7595 |
+- |
7596 |
+- vmx->guest_msrs[j].index = i; |
7597 |
+- vmx->guest_msrs[j].data = 0; |
7598 |
+- switch (index) { |
7599 |
+- case MSR_IA32_TSX_CTRL: |
7600 |
+- /* |
7601 |
+- * No need to pass TSX_CTRL_CPUID_CLEAR through, so |
7602 |
+- * let's avoid changing CPUID bits under the host |
7603 |
+- * kernel's feet. |
7604 |
+- */ |
7605 |
+- vmx->guest_msrs[j].mask = ~(u64)TSX_CTRL_CPUID_CLEAR; |
7606 |
+- break; |
7607 |
+- default: |
7608 |
+- vmx->guest_msrs[j].mask = -1ull; |
7609 |
+- break; |
7610 |
+- } |
7611 |
+- ++vmx->nmsrs; |
7612 |
+- } |
7613 |
+- |
7614 |
+- err = alloc_loaded_vmcs(&vmx->vmcs01); |
7615 |
+- if (err < 0) |
7616 |
+- goto free_pml; |
7617 |
+- |
7618 |
+- msr_bitmap = vmx->vmcs01.msr_bitmap; |
7619 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_TSC, MSR_TYPE_R); |
7620 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_FS_BASE, MSR_TYPE_RW); |
7621 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_GS_BASE, MSR_TYPE_RW); |
7622 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_KERNEL_GS_BASE, MSR_TYPE_RW); |
7623 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_CS, MSR_TYPE_RW); |
7624 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_ESP, MSR_TYPE_RW); |
7625 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_IA32_SYSENTER_EIP, MSR_TYPE_RW); |
7626 |
+- if (kvm_cstate_in_guest(kvm)) { |
7627 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C1_RES, MSR_TYPE_R); |
7628 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C3_RESIDENCY, MSR_TYPE_R); |
7629 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C6_RESIDENCY, MSR_TYPE_R); |
7630 |
+- vmx_disable_intercept_for_msr(msr_bitmap, MSR_CORE_C7_RESIDENCY, MSR_TYPE_R); |
7631 |
+- } |
7632 |
+- vmx->msr_bitmap_mode = 0; |
7633 |
+- |
7634 |
+- vmx->loaded_vmcs = &vmx->vmcs01; |
7635 |
+- cpu = get_cpu(); |
7636 |
+- vmx_vcpu_load(&vmx->vcpu, cpu); |
7637 |
+- vmx->vcpu.cpu = cpu; |
7638 |
+- init_vmcs(vmx); |
7639 |
+- vmx_vcpu_put(&vmx->vcpu); |
7640 |
+- put_cpu(); |
7641 |
+- if (cpu_need_virtualize_apic_accesses(&vmx->vcpu)) { |
7642 |
+- err = alloc_apic_access_page(kvm); |
7643 |
+- if (err) |
7644 |
+- goto free_vmcs; |
7645 |
+- } |
7646 |
+- |
7647 |
+- if (enable_ept && !enable_unrestricted_guest) { |
7648 |
+- err = init_rmode_identity_map(kvm); |
7649 |
+- if (err) |
7650 |
+- goto free_vmcs; |
7651 |
+- } |
7652 |
+- |
7653 |
+- if (nested) |
7654 |
+- nested_vmx_setup_ctls_msrs(&vmx->nested.msrs, |
7655 |
+- vmx_capability.ept, |
7656 |
+- kvm_vcpu_apicv_active(&vmx->vcpu)); |
7657 |
+- else |
7658 |
+- memset(&vmx->nested.msrs, 0, sizeof(vmx->nested.msrs)); |
7659 |
+- |
7660 |
+- vmx->nested.posted_intr_nv = -1; |
7661 |
+- vmx->nested.current_vmptr = -1ull; |
7662 |
+- |
7663 |
+- vmx->msr_ia32_feature_control_valid_bits = FEATURE_CONTROL_LOCKED; |
7664 |
+- |
7665 |
+- /* |
7666 |
+- * Enforce invariant: pi_desc.nv is always either POSTED_INTR_VECTOR |
7667 |
+- * or POSTED_INTR_WAKEUP_VECTOR. |
7668 |
+- */ |
7669 |
+- vmx->pi_desc.nv = POSTED_INTR_VECTOR; |
7670 |
+- vmx->pi_desc.sn = 1; |
7671 |
+- |
7672 |
+- vmx->ept_pointer = INVALID_PAGE; |
7673 |
+- |
7674 |
+- return &vmx->vcpu; |
7675 |
+- |
7676 |
+-free_vmcs: |
7677 |
+- free_loaded_vmcs(vmx->loaded_vmcs); |
7678 |
+-free_pml: |
7679 |
+- vmx_destroy_pml_buffer(vmx); |
7680 |
+-uninit_vcpu: |
7681 |
+- kvm_vcpu_uninit(&vmx->vcpu); |
7682 |
+-free_vcpu: |
7683 |
+- free_vpid(vmx->vpid); |
7684 |
+- kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.guest_fpu); |
7685 |
+-free_user_fpu: |
7686 |
+- kmem_cache_free(x86_fpu_cache, vmx->vcpu.arch.user_fpu); |
7687 |
+-free_partial_vcpu: |
7688 |
+- kmem_cache_free(kvm_vcpu_cache, vmx); |
7689 |
+- return ERR_PTR(err); |
7690 |
+-} |
7691 |
+- |
7692 |
+-#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" |
7693 |
+-#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" |
7694 |
+- |
7695 |
+-static int vmx_vm_init(struct kvm *kvm) |
7696 |
+-{ |
7697 |
+- spin_lock_init(&to_kvm_vmx(kvm)->ept_pointer_lock); |
7698 |
+- |
7699 |
+- if (!ple_gap) |
7700 |
+- kvm->arch.pause_in_guest = true; |
7701 |
+- |
7702 |
+- if (boot_cpu_has(X86_BUG_L1TF) && enable_ept) { |
7703 |
+- switch (l1tf_mitigation) { |
7704 |
+- case L1TF_MITIGATION_OFF: |
7705 |
+- case L1TF_MITIGATION_FLUSH_NOWARN: |
7706 |
+- /* 'I explicitly don't care' is set */ |
7707 |
+- break; |
7708 |
+- case L1TF_MITIGATION_FLUSH: |
7709 |
+- case L1TF_MITIGATION_FLUSH_NOSMT: |
7710 |
+- case L1TF_MITIGATION_FULL: |
7711 |
+- /* |
7712 |
+- * Warn upon starting the first VM in a potentially |
7713 |
+- * insecure environment. |
7714 |
+- */ |
7715 |
+- if (sched_smt_active()) |
7716 |
+- pr_warn_once(L1TF_MSG_SMT); |
7717 |
+- if (l1tf_vmx_mitigation == VMENTER_L1D_FLUSH_NEVER) |
7718 |
+- pr_warn_once(L1TF_MSG_L1D); |
7719 |
+- break; |
7720 |
+- case L1TF_MITIGATION_FULL_FORCE: |
7721 |
+- /* Flush is enforced */ |
7722 |
+- break; |
7723 |
+- } |
7724 |
+- } |
7725 |
+- return 0; |
7726 |
+-} |
7727 |
+- |
7728 |
+-static int __init vmx_check_processor_compat(void) |
7729 |
+-{ |
7730 |
+- struct vmcs_config vmcs_conf; |
7731 |
+- struct vmx_capability vmx_cap; |
7732 |
+- |
7733 |
+- if (setup_vmcs_config(&vmcs_conf, &vmx_cap) < 0) |
7734 |
+- return -EIO; |
7735 |
+- if (nested) |
7736 |
+- nested_vmx_setup_ctls_msrs(&vmcs_conf.nested, vmx_cap.ept, |
7737 |
+- enable_apicv); |
7738 |
+- if (memcmp(&vmcs_config, &vmcs_conf, sizeof(struct vmcs_config)) != 0) { |
7739 |
+- printk(KERN_ERR "kvm: CPU %d feature inconsistency!\n", |
7740 |
+- smp_processor_id()); |
7741 |
+- return -EIO; |
7742 |
+- } |
7743 |
+- return 0; |
7744 |
+-} |
7745 |
+- |
7746 |
+-static u64 vmx_get_mt_mask(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio) |
7747 |
+-{ |
7748 |
+- u8 cache; |
7749 |
+- u64 ipat = 0; |
7750 |
+- |
7751 |
+- /* For VT-d and EPT combination |
7752 |
+- * 1. MMIO: always map as UC |
7753 |
+- * 2. EPT with VT-d: |
7754 |
+- * a. VT-d without snooping control feature: can't guarantee the |
7755 |
+- * result, try to trust guest. |
7756 |
+- * b. VT-d with snooping control feature: snooping control feature of |
7757 |
+- * VT-d engine can guarantee the cache correctness. Just set it |
7758 |
+- * to WB to keep consistent with host. So the same as item 3. |
7759 |
+- * 3. EPT without VT-d: always map as WB and set IPAT=1 to keep |
7760 |
+- * consistent with host MTRR |
7761 |
+- */ |
7762 |
+- if (is_mmio) { |
7763 |
+- cache = MTRR_TYPE_UNCACHABLE; |
7764 |
+- goto exit; |
7765 |
+- } |
7766 |
+- |
7767 |
+- if (!kvm_arch_has_noncoherent_dma(vcpu->kvm)) { |
7768 |
+- ipat = VMX_EPT_IPAT_BIT; |
7769 |
+- cache = MTRR_TYPE_WRBACK; |
7770 |
+- goto exit; |
7771 |
+- } |
7772 |
+- |
7773 |
+- if (kvm_read_cr0(vcpu) & X86_CR0_CD) { |
7774 |
+- ipat = VMX_EPT_IPAT_BIT; |
7775 |
+- if (kvm_check_has_quirk(vcpu->kvm, KVM_X86_QUIRK_CD_NW_CLEARED)) |
7776 |
+- cache = MTRR_TYPE_WRBACK; |
7777 |
+- else |
7778 |
+- cache = MTRR_TYPE_UNCACHABLE; |
7779 |
+- goto exit; |
7780 |
+- } |
7781 |
+- |
7782 |
+- cache = kvm_mtrr_get_guest_memory_type(vcpu, gfn); |
7783 |
+- |
7784 |
+-exit: |
7785 |
+- return (cache << VMX_EPT_MT_EPTE_SHIFT) | ipat; |
7786 |
+-} |
7787 |
+- |
7788 |
+-static int vmx_get_lpage_level(void) |
7789 |
+-{ |
7790 |
+- if (enable_ept && !cpu_has_vmx_ept_1g_page()) |
7791 |
+- return PT_DIRECTORY_LEVEL; |
7792 |
+- else |
7793 |
+- /* For shadow and EPT supported 1GB page */ |
7794 |
+- return PT_PDPE_LEVEL; |
7795 |
+-} |
7796 |
+- |
7797 |
+-static void vmcs_set_secondary_exec_control(struct vcpu_vmx *vmx) |
7798 |
+-{ |
7799 |
+- /* |
7800 |
+- * These bits in the secondary execution controls field |
7801 |
+- * are dynamic, the others are mostly based on the hypervisor |
7802 |
+- * architecture and the guest's CPUID. Do not touch the |
7803 |
+- * dynamic bits. |
7804 |
+- */ |
7805 |
+- u32 mask = |
7806 |
+- SECONDARY_EXEC_SHADOW_VMCS | |
7807 |
+- SECONDARY_EXEC_VIRTUALIZE_X2APIC_MODE | |
7808 |
+- SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | |
7809 |
+- SECONDARY_EXEC_DESC; |
7810 |
+- |
7811 |
+- u32 new_ctl = vmx->secondary_exec_control; |
7812 |
+- u32 cur_ctl = secondary_exec_controls_get(vmx); |
7813 |
+- |
7814 |
+- secondary_exec_controls_set(vmx, (new_ctl & ~mask) | (cur_ctl & mask)); |
7815 |
+-} |
7816 |
+- |
7817 |
+-/* |
7818 |
+- * Generate MSR_IA32_VMX_CR{0,4}_FIXED1 according to CPUID. Only set bits |
7819 |
+- * (indicating "allowed-1") if they are supported in the guest's CPUID. |
7820 |
+- */ |
7821 |
+-static void nested_vmx_cr_fixed1_bits_update(struct kvm_vcpu *vcpu) |
7822 |
+-{ |
7823 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7824 |
+- struct kvm_cpuid_entry2 *entry; |
7825 |
+- |
7826 |
+- vmx->nested.msrs.cr0_fixed1 = 0xffffffff; |
7827 |
+- vmx->nested.msrs.cr4_fixed1 = X86_CR4_PCE; |
7828 |
+- |
7829 |
+-#define cr4_fixed1_update(_cr4_mask, _reg, _cpuid_mask) do { \ |
7830 |
+- if (entry && (entry->_reg & (_cpuid_mask))) \ |
7831 |
+- vmx->nested.msrs.cr4_fixed1 |= (_cr4_mask); \ |
7832 |
+-} while (0) |
7833 |
+- |
7834 |
+- entry = kvm_find_cpuid_entry(vcpu, 0x1, 0); |
7835 |
+- cr4_fixed1_update(X86_CR4_VME, edx, bit(X86_FEATURE_VME)); |
7836 |
+- cr4_fixed1_update(X86_CR4_PVI, edx, bit(X86_FEATURE_VME)); |
7837 |
+- cr4_fixed1_update(X86_CR4_TSD, edx, bit(X86_FEATURE_TSC)); |
7838 |
+- cr4_fixed1_update(X86_CR4_DE, edx, bit(X86_FEATURE_DE)); |
7839 |
+- cr4_fixed1_update(X86_CR4_PSE, edx, bit(X86_FEATURE_PSE)); |
7840 |
+- cr4_fixed1_update(X86_CR4_PAE, edx, bit(X86_FEATURE_PAE)); |
7841 |
+- cr4_fixed1_update(X86_CR4_MCE, edx, bit(X86_FEATURE_MCE)); |
7842 |
+- cr4_fixed1_update(X86_CR4_PGE, edx, bit(X86_FEATURE_PGE)); |
7843 |
+- cr4_fixed1_update(X86_CR4_OSFXSR, edx, bit(X86_FEATURE_FXSR)); |
7844 |
+- cr4_fixed1_update(X86_CR4_OSXMMEXCPT, edx, bit(X86_FEATURE_XMM)); |
7845 |
+- cr4_fixed1_update(X86_CR4_VMXE, ecx, bit(X86_FEATURE_VMX)); |
7846 |
+- cr4_fixed1_update(X86_CR4_SMXE, ecx, bit(X86_FEATURE_SMX)); |
7847 |
+- cr4_fixed1_update(X86_CR4_PCIDE, ecx, bit(X86_FEATURE_PCID)); |
7848 |
+- cr4_fixed1_update(X86_CR4_OSXSAVE, ecx, bit(X86_FEATURE_XSAVE)); |
7849 |
+- |
7850 |
+- entry = kvm_find_cpuid_entry(vcpu, 0x7, 0); |
7851 |
+- cr4_fixed1_update(X86_CR4_FSGSBASE, ebx, bit(X86_FEATURE_FSGSBASE)); |
7852 |
+- cr4_fixed1_update(X86_CR4_SMEP, ebx, bit(X86_FEATURE_SMEP)); |
7853 |
+- cr4_fixed1_update(X86_CR4_SMAP, ebx, bit(X86_FEATURE_SMAP)); |
7854 |
+- cr4_fixed1_update(X86_CR4_PKE, ecx, bit(X86_FEATURE_PKU)); |
7855 |
+- cr4_fixed1_update(X86_CR4_UMIP, ecx, bit(X86_FEATURE_UMIP)); |
7856 |
+- cr4_fixed1_update(X86_CR4_LA57, ecx, bit(X86_FEATURE_LA57)); |
7857 |
+- |
7858 |
+-#undef cr4_fixed1_update |
7859 |
+-} |
7860 |
+- |
7861 |
+-static void nested_vmx_entry_exit_ctls_update(struct kvm_vcpu *vcpu) |
7862 |
+-{ |
7863 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7864 |
+- |
7865 |
+- if (kvm_mpx_supported()) { |
7866 |
+- bool mpx_enabled = guest_cpuid_has(vcpu, X86_FEATURE_MPX); |
7867 |
+- |
7868 |
+- if (mpx_enabled) { |
7869 |
+- vmx->nested.msrs.entry_ctls_high |= VM_ENTRY_LOAD_BNDCFGS; |
7870 |
+- vmx->nested.msrs.exit_ctls_high |= VM_EXIT_CLEAR_BNDCFGS; |
7871 |
+- } else { |
7872 |
+- vmx->nested.msrs.entry_ctls_high &= ~VM_ENTRY_LOAD_BNDCFGS; |
7873 |
+- vmx->nested.msrs.exit_ctls_high &= ~VM_EXIT_CLEAR_BNDCFGS; |
7874 |
+- } |
7875 |
+- } |
7876 |
+-} |
7877 |
+- |
7878 |
+-static void update_intel_pt_cfg(struct kvm_vcpu *vcpu) |
7879 |
+-{ |
7880 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7881 |
+- struct kvm_cpuid_entry2 *best = NULL; |
7882 |
+- int i; |
7883 |
+- |
7884 |
+- for (i = 0; i < PT_CPUID_LEAVES; i++) { |
7885 |
+- best = kvm_find_cpuid_entry(vcpu, 0x14, i); |
7886 |
+- if (!best) |
7887 |
+- return; |
7888 |
+- vmx->pt_desc.caps[CPUID_EAX + i*PT_CPUID_REGS_NUM] = best->eax; |
7889 |
+- vmx->pt_desc.caps[CPUID_EBX + i*PT_CPUID_REGS_NUM] = best->ebx; |
7890 |
+- vmx->pt_desc.caps[CPUID_ECX + i*PT_CPUID_REGS_NUM] = best->ecx; |
7891 |
+- vmx->pt_desc.caps[CPUID_EDX + i*PT_CPUID_REGS_NUM] = best->edx; |
7892 |
+- } |
7893 |
+- |
7894 |
+- /* Get the number of configurable Address Ranges for filtering */ |
7895 |
+- vmx->pt_desc.addr_range = intel_pt_validate_cap(vmx->pt_desc.caps, |
7896 |
+- PT_CAP_num_address_ranges); |
7897 |
+- |
7898 |
+- /* Initialize and clear the no dependency bits */ |
7899 |
+- vmx->pt_desc.ctl_bitmask = ~(RTIT_CTL_TRACEEN | RTIT_CTL_OS | |
7900 |
+- RTIT_CTL_USR | RTIT_CTL_TSC_EN | RTIT_CTL_DISRETC); |
7901 |
+- |
7902 |
+- /* |
7903 |
+- * If CPUID.(EAX=14H,ECX=0):EBX[0]=1 CR3Filter can be set otherwise |
7904 |
+- * will inject an #GP |
7905 |
+- */ |
7906 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_cr3_filtering)) |
7907 |
+- vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_CR3EN; |
7908 |
+- |
7909 |
+- /* |
7910 |
+- * If CPUID.(EAX=14H,ECX=0):EBX[1]=1 CYCEn, CycThresh and |
7911 |
+- * PSBFreq can be set |
7912 |
+- */ |
7913 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_psb_cyc)) |
7914 |
+- vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_CYCLEACC | |
7915 |
+- RTIT_CTL_CYC_THRESH | RTIT_CTL_PSB_FREQ); |
7916 |
+- |
7917 |
+- /* |
7918 |
+- * If CPUID.(EAX=14H,ECX=0):EBX[3]=1 MTCEn BranchEn and |
7919 |
+- * MTCFreq can be set |
7920 |
+- */ |
7921 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_mtc)) |
7922 |
+- vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_MTC_EN | |
7923 |
+- RTIT_CTL_BRANCH_EN | RTIT_CTL_MTC_RANGE); |
7924 |
+- |
7925 |
+- /* If CPUID.(EAX=14H,ECX=0):EBX[4]=1 FUPonPTW and PTWEn can be set */ |
7926 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_ptwrite)) |
7927 |
+- vmx->pt_desc.ctl_bitmask &= ~(RTIT_CTL_FUP_ON_PTW | |
7928 |
+- RTIT_CTL_PTW_EN); |
7929 |
+- |
7930 |
+- /* If CPUID.(EAX=14H,ECX=0):EBX[5]=1 PwrEvEn can be set */ |
7931 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_power_event_trace)) |
7932 |
+- vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_PWR_EVT_EN; |
7933 |
+- |
7934 |
+- /* If CPUID.(EAX=14H,ECX=0):ECX[0]=1 ToPA can be set */ |
7935 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_topa_output)) |
7936 |
+- vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_TOPA; |
7937 |
+- |
7938 |
+- /* If CPUID.(EAX=14H,ECX=0):ECX[3]=1 FabircEn can be set */ |
7939 |
+- if (intel_pt_validate_cap(vmx->pt_desc.caps, PT_CAP_output_subsys)) |
7940 |
+- vmx->pt_desc.ctl_bitmask &= ~RTIT_CTL_FABRIC_EN; |
7941 |
+- |
7942 |
+- /* unmask address range configure area */ |
7943 |
+- for (i = 0; i < vmx->pt_desc.addr_range; i++) |
7944 |
+- vmx->pt_desc.ctl_bitmask &= ~(0xfULL << (32 + i * 4)); |
7945 |
+-} |
7946 |
+- |
7947 |
+-static void vmx_cpuid_update(struct kvm_vcpu *vcpu) |
7948 |
+-{ |
7949 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
7950 |
+- |
7951 |
+- /* xsaves_enabled is recomputed in vmx_compute_secondary_exec_control(). */ |
7952 |
+- vcpu->arch.xsaves_enabled = false; |
7953 |
+- |
7954 |
+- if (cpu_has_secondary_exec_ctrls()) { |
7955 |
+- vmx_compute_secondary_exec_control(vmx); |
7956 |
+- vmcs_set_secondary_exec_control(vmx); |
7957 |
+- } |
7958 |
+- |
7959 |
+- if (nested_vmx_allowed(vcpu)) |
7960 |
+- to_vmx(vcpu)->msr_ia32_feature_control_valid_bits |= |
7961 |
+- FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX | |
7962 |
+- FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; |
7963 |
+- else |
7964 |
+- to_vmx(vcpu)->msr_ia32_feature_control_valid_bits &= |
7965 |
+- ~(FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX | |
7966 |
+- FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX); |
7967 |
+- |
7968 |
+- if (nested_vmx_allowed(vcpu)) { |
7969 |
+- nested_vmx_cr_fixed1_bits_update(vcpu); |
7970 |
+- nested_vmx_entry_exit_ctls_update(vcpu); |
7971 |
+- } |
7972 |
+- |
7973 |
+- if (boot_cpu_has(X86_FEATURE_INTEL_PT) && |
7974 |
+- guest_cpuid_has(vcpu, X86_FEATURE_INTEL_PT)) |
7975 |
+- update_intel_pt_cfg(vcpu); |
7976 |
+- |
7977 |
+- if (boot_cpu_has(X86_FEATURE_RTM)) { |
7978 |
+- struct shared_msr_entry *msr; |
7979 |
+- msr = find_msr_entry(vmx, MSR_IA32_TSX_CTRL); |
7980 |
+- if (msr) { |
7981 |
+- bool enabled = guest_cpuid_has(vcpu, X86_FEATURE_RTM); |
7982 |
+- vmx_set_guest_msr(vmx, msr, enabled ? 0 : TSX_CTRL_RTM_DISABLE); |
7983 |
+- } |
7984 |
+- } |
7985 |
+-} |
7986 |
+- |
7987 |
+-static void vmx_set_supported_cpuid(u32 func, struct kvm_cpuid_entry2 *entry) |
7988 |
+-{ |
7989 |
+- if (func == 1 && nested) |
7990 |
+- entry->ecx |= bit(X86_FEATURE_VMX); |
7991 |
+-} |
7992 |
+- |
7993 |
+-static void vmx_request_immediate_exit(struct kvm_vcpu *vcpu) |
7994 |
+-{ |
7995 |
+- to_vmx(vcpu)->req_immediate_exit = true; |
7996 |
+-} |
7997 |
+- |
7998 |
+-static int vmx_check_intercept(struct kvm_vcpu *vcpu, |
7999 |
+- struct x86_instruction_info *info, |
8000 |
+- enum x86_intercept_stage stage) |
8001 |
+-{ |
8002 |
+- struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
8003 |
+- struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt; |
8004 |
+- |
8005 |
+- /* |
8006 |
+- * RDPID causes #UD if disabled through secondary execution controls. |
8007 |
+- * Because it is marked as EmulateOnUD, we need to intercept it here. |
8008 |
+- */ |
8009 |
+- if (info->intercept == x86_intercept_rdtscp && |
8010 |
+- !nested_cpu_has2(vmcs12, SECONDARY_EXEC_RDTSCP)) { |
8011 |
+- ctxt->exception.vector = UD_VECTOR; |
8012 |
+- ctxt->exception.error_code_valid = false; |
8013 |
+- return X86EMUL_PROPAGATE_FAULT; |
8014 |
+- } |
8015 |
+- |
8016 |
+- /* TODO: check more intercepts... */ |
8017 |
+- return X86EMUL_CONTINUE; |
8018 |
+-} |
8019 |
+- |
8020 |
+-#ifdef CONFIG_X86_64 |
8021 |
+-/* (a << shift) / divisor, return 1 if overflow otherwise 0 */ |
8022 |
+-static inline int u64_shl_div_u64(u64 a, unsigned int shift, |
8023 |
+- u64 divisor, u64 *result) |
8024 |
+-{ |
8025 |
+- u64 low = a << shift, high = a >> (64 - shift); |
8026 |
+- |
8027 |
+- /* To avoid the overflow on divq */ |
8028 |
+- if (high >= divisor) |
8029 |
+- return 1; |
8030 |
+- |
8031 |
+- /* Low hold the result, high hold rem which is discarded */ |
8032 |
+- asm("divq %2\n\t" : "=a" (low), "=d" (high) : |
8033 |
+- "rm" (divisor), "0" (low), "1" (high)); |
8034 |
+- *result = low; |
8035 |
+- |
8036 |
+- return 0; |
8037 |
+-} |
8038 |
+- |
8039 |
+-static int vmx_set_hv_timer(struct kvm_vcpu *vcpu, u64 guest_deadline_tsc, |
8040 |
+- bool *expired) |
8041 |
+-{ |
8042 |
+- struct vcpu_vmx *vmx; |
8043 |
+- u64 tscl, guest_tscl, delta_tsc, lapic_timer_advance_cycles; |
8044 |
+- struct kvm_timer *ktimer = &vcpu->arch.apic->lapic_timer; |
8045 |
+- |
8046 |
+- if (kvm_mwait_in_guest(vcpu->kvm) || |
8047 |
+- kvm_can_post_timer_interrupt(vcpu)) |
8048 |
+- return -EOPNOTSUPP; |
8049 |
+- |
8050 |
+- vmx = to_vmx(vcpu); |
8051 |
+- tscl = rdtsc(); |
8052 |
+- guest_tscl = kvm_read_l1_tsc(vcpu, tscl); |
8053 |
+- delta_tsc = max(guest_deadline_tsc, guest_tscl) - guest_tscl; |
8054 |
+- lapic_timer_advance_cycles = nsec_to_cycles(vcpu, |
8055 |
+- ktimer->timer_advance_ns); |
8056 |
+- |
8057 |
+- if (delta_tsc > lapic_timer_advance_cycles) |
8058 |
+- delta_tsc -= lapic_timer_advance_cycles; |
8059 |
+- else |
8060 |
+- delta_tsc = 0; |
8061 |
+- |
8062 |
+- /* Convert to host delta tsc if tsc scaling is enabled */ |
8063 |
+- if (vcpu->arch.tsc_scaling_ratio != kvm_default_tsc_scaling_ratio && |
8064 |
+- delta_tsc && u64_shl_div_u64(delta_tsc, |
8065 |
+- kvm_tsc_scaling_ratio_frac_bits, |
8066 |
+- vcpu->arch.tsc_scaling_ratio, &delta_tsc)) |
8067 |
+- return -ERANGE; |
8068 |
+- |
8069 |
+- /* |
8070 |
+- * If the delta tsc can't fit in the 32 bit after the multi shift, |
8071 |
+- * we can't use the preemption timer. |
8072 |
+- * It's possible that it fits on later vmentries, but checking |
8073 |
+- * on every vmentry is costly so we just use an hrtimer. |
8074 |
+- */ |
8075 |
+- if (delta_tsc >> (cpu_preemption_timer_multi + 32)) |
8076 |
+- return -ERANGE; |
8077 |
+- |
8078 |
+- vmx->hv_deadline_tsc = tscl + delta_tsc; |
8079 |
+- *expired = !delta_tsc; |
8080 |
+- return 0; |
8081 |
+-} |
8082 |
+- |
8083 |
+-static void vmx_cancel_hv_timer(struct kvm_vcpu *vcpu) |
8084 |
+-{ |
8085 |
+- to_vmx(vcpu)->hv_deadline_tsc = -1; |
8086 |
+-} |
8087 |
+-#endif |
8088 |
+- |
8089 |
+-static void vmx_sched_in(struct kvm_vcpu *vcpu, int cpu) |
8090 |
+-{ |
8091 |
+- if (!kvm_pause_in_guest(vcpu->kvm)) |
8092 |
+- shrink_ple_window(vcpu); |
8093 |
+-} |
8094 |
+- |
8095 |
+-static void vmx_slot_enable_log_dirty(struct kvm *kvm, |
8096 |
+- struct kvm_memory_slot *slot) |
8097 |
+-{ |
8098 |
+- kvm_mmu_slot_leaf_clear_dirty(kvm, slot); |
8099 |
+- kvm_mmu_slot_largepage_remove_write_access(kvm, slot); |
8100 |
+-} |
8101 |
+- |
8102 |
+-static void vmx_slot_disable_log_dirty(struct kvm *kvm, |
8103 |
+- struct kvm_memory_slot *slot) |
8104 |
+-{ |
8105 |
+- kvm_mmu_slot_set_dirty(kvm, slot); |
8106 |
+-} |
8107 |
+- |
8108 |
+-static void vmx_flush_log_dirty(struct kvm *kvm) |
8109 |
+-{ |
8110 |
+- kvm_flush_pml_buffers(kvm); |
8111 |
+-} |
8112 |
+- |
8113 |
+-static int vmx_write_pml_buffer(struct kvm_vcpu *vcpu) |
8114 |
+-{ |
8115 |
+- struct vmcs12 *vmcs12; |
8116 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
8117 |
+- gpa_t gpa, dst; |
8118 |
+- |
8119 |
+- if (is_guest_mode(vcpu)) { |
8120 |
+- WARN_ON_ONCE(vmx->nested.pml_full); |
8121 |
+- |
8122 |
+- /* |
8123 |
+- * Check if PML is enabled for the nested guest. |
8124 |
+- * Whether eptp bit 6 is set is already checked |
8125 |
+- * as part of A/D emulation. |
8126 |
+- */ |
8127 |
+- vmcs12 = get_vmcs12(vcpu); |
8128 |
+- if (!nested_cpu_has_pml(vmcs12)) |
8129 |
+- return 0; |
8130 |
+- |
8131 |
+- if (vmcs12->guest_pml_index >= PML_ENTITY_NUM) { |
8132 |
+- vmx->nested.pml_full = true; |
8133 |
+- return 1; |
8134 |
+- } |
8135 |
+- |
8136 |
+- gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS) & ~0xFFFull; |
8137 |
+- dst = vmcs12->pml_address + sizeof(u64) * vmcs12->guest_pml_index; |
8138 |
+- |
8139 |
+- if (kvm_write_guest_page(vcpu->kvm, gpa_to_gfn(dst), &gpa, |
8140 |
+- offset_in_page(dst), sizeof(gpa))) |
8141 |
+- return 0; |
8142 |
+- |
8143 |
+- vmcs12->guest_pml_index--; |
8144 |
+- } |
8145 |
+- |
8146 |
+- return 0; |
8147 |
+-} |
8148 |
+- |
8149 |
+-static void vmx_enable_log_dirty_pt_masked(struct kvm *kvm, |
8150 |
+- struct kvm_memory_slot *memslot, |
8151 |
+- gfn_t offset, unsigned long mask) |
8152 |
+-{ |
8153 |
+- kvm_mmu_clear_dirty_pt_masked(kvm, memslot, offset, mask); |
8154 |
+-} |
8155 |
+- |
8156 |
+-static void __pi_post_block(struct kvm_vcpu *vcpu) |
8157 |
+-{ |
8158 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
8159 |
+- struct pi_desc old, new; |
8160 |
+- unsigned int dest; |
8161 |
+- |
8162 |
+- do { |
8163 |
+- old.control = new.control = pi_desc->control; |
8164 |
+- WARN(old.nv != POSTED_INTR_WAKEUP_VECTOR, |
8165 |
+- "Wakeup handler not enabled while the VCPU is blocked\n"); |
8166 |
+- |
8167 |
+- dest = cpu_physical_id(vcpu->cpu); |
8168 |
+- |
8169 |
+- if (x2apic_enabled()) |
8170 |
+- new.ndst = dest; |
8171 |
+- else |
8172 |
+- new.ndst = (dest << 8) & 0xFF00; |
8173 |
+- |
8174 |
+- /* set 'NV' to 'notification vector' */ |
8175 |
+- new.nv = POSTED_INTR_VECTOR; |
8176 |
+- } while (cmpxchg64(&pi_desc->control, old.control, |
8177 |
+- new.control) != old.control); |
8178 |
+- |
8179 |
+- if (!WARN_ON_ONCE(vcpu->pre_pcpu == -1)) { |
8180 |
+- spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8181 |
+- list_del(&vcpu->blocked_vcpu_list); |
8182 |
+- spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8183 |
+- vcpu->pre_pcpu = -1; |
8184 |
+- } |
8185 |
+-} |
8186 |
+- |
8187 |
+-/* |
8188 |
+- * This routine does the following things for vCPU which is going |
8189 |
+- * to be blocked if VT-d PI is enabled. |
8190 |
+- * - Store the vCPU to the wakeup list, so when interrupts happen |
8191 |
+- * we can find the right vCPU to wake up. |
8192 |
+- * - Change the Posted-interrupt descriptor as below: |
8193 |
+- * 'NDST' <-- vcpu->pre_pcpu |
8194 |
+- * 'NV' <-- POSTED_INTR_WAKEUP_VECTOR |
8195 |
+- * - If 'ON' is set during this process, which means at least one |
8196 |
+- * interrupt is posted for this vCPU, we cannot block it, in |
8197 |
+- * this case, return 1, otherwise, return 0. |
8198 |
+- * |
8199 |
+- */ |
8200 |
+-static int pi_pre_block(struct kvm_vcpu *vcpu) |
8201 |
+-{ |
8202 |
+- unsigned int dest; |
8203 |
+- struct pi_desc old, new; |
8204 |
+- struct pi_desc *pi_desc = vcpu_to_pi_desc(vcpu); |
8205 |
+- |
8206 |
+- if (!kvm_arch_has_assigned_device(vcpu->kvm) || |
8207 |
+- !irq_remapping_cap(IRQ_POSTING_CAP) || |
8208 |
+- !kvm_vcpu_apicv_active(vcpu)) |
8209 |
+- return 0; |
8210 |
+- |
8211 |
+- WARN_ON(irqs_disabled()); |
8212 |
+- local_irq_disable(); |
8213 |
+- if (!WARN_ON_ONCE(vcpu->pre_pcpu != -1)) { |
8214 |
+- vcpu->pre_pcpu = vcpu->cpu; |
8215 |
+- spin_lock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8216 |
+- list_add_tail(&vcpu->blocked_vcpu_list, |
8217 |
+- &per_cpu(blocked_vcpu_on_cpu, |
8218 |
+- vcpu->pre_pcpu)); |
8219 |
+- spin_unlock(&per_cpu(blocked_vcpu_on_cpu_lock, vcpu->pre_pcpu)); |
8220 |
+- } |
8221 |
+- |
8222 |
+- do { |
8223 |
+- old.control = new.control = pi_desc->control; |
8224 |
+- |
8225 |
+- WARN((pi_desc->sn == 1), |
8226 |
+- "Warning: SN field of posted-interrupts " |
8227 |
+- "is set before blocking\n"); |
8228 |
+- |
8229 |
+- /* |
8230 |
+- * Since vCPU can be preempted during this process, |
8231 |
+- * vcpu->cpu could be different with pre_pcpu, we |
8232 |
+- * need to set pre_pcpu as the destination of wakeup |
8233 |
+- * notification event, then we can find the right vCPU |
8234 |
+- * to wakeup in wakeup handler if interrupts happen |
8235 |
+- * when the vCPU is in blocked state. |
8236 |
+- */ |
8237 |
+- dest = cpu_physical_id(vcpu->pre_pcpu); |
8238 |
+- |
8239 |
+- if (x2apic_enabled()) |
8240 |
+- new.ndst = dest; |
8241 |
+- else |
8242 |
+- new.ndst = (dest << 8) & 0xFF00; |
8243 |
+- |
8244 |
+- /* set 'NV' to 'wakeup vector' */ |
8245 |
+- new.nv = POSTED_INTR_WAKEUP_VECTOR; |
8246 |
+- } while (cmpxchg64(&pi_desc->control, old.control, |
8247 |
+- new.control) != old.control); |
8248 |
+- |
8249 |
+- /* We should not block the vCPU if an interrupt is posted for it. */ |
8250 |
+- if (pi_test_on(pi_desc) == 1) |
8251 |
+- __pi_post_block(vcpu); |
8252 |
+- |
8253 |
+- local_irq_enable(); |
8254 |
+- return (vcpu->pre_pcpu == -1); |
8255 |
+-} |
8256 |
+- |
8257 |
+-static int vmx_pre_block(struct kvm_vcpu *vcpu) |
8258 |
+-{ |
8259 |
+- if (pi_pre_block(vcpu)) |
8260 |
+- return 1; |
8261 |
+- |
8262 |
+- if (kvm_lapic_hv_timer_in_use(vcpu)) |
8263 |
+- kvm_lapic_switch_to_sw_timer(vcpu); |
8264 |
+- |
8265 |
+- return 0; |
8266 |
+-} |
8267 |
+- |
8268 |
+-static void pi_post_block(struct kvm_vcpu *vcpu) |
8269 |
+-{ |
8270 |
+- if (vcpu->pre_pcpu == -1) |
8271 |
+- return; |
8272 |
+- |
8273 |
+- WARN_ON(irqs_disabled()); |
8274 |
+- local_irq_disable(); |
8275 |
+- __pi_post_block(vcpu); |
8276 |
+- local_irq_enable(); |
8277 |
+-} |
8278 |
+- |
8279 |
+-static void vmx_post_block(struct kvm_vcpu *vcpu) |
8280 |
+-{ |
8281 |
+- if (kvm_x86_ops->set_hv_timer) |
8282 |
+- kvm_lapic_switch_to_hv_timer(vcpu); |
8283 |
+- |
8284 |
+- pi_post_block(vcpu); |
8285 |
+-} |
8286 |
+- |
8287 |
+-/* |
8288 |
+- * vmx_update_pi_irte - set IRTE for Posted-Interrupts |
8289 |
+- * |
8290 |
+- * @kvm: kvm |
8291 |
+- * @host_irq: host irq of the interrupt |
8292 |
+- * @guest_irq: gsi of the interrupt |
8293 |
+- * @set: set or unset PI |
8294 |
+- * returns 0 on success, < 0 on failure |
8295 |
+- */ |
8296 |
+-static int vmx_update_pi_irte(struct kvm *kvm, unsigned int host_irq, |
8297 |
+- uint32_t guest_irq, bool set) |
8298 |
+-{ |
8299 |
+- struct kvm_kernel_irq_routing_entry *e; |
8300 |
+- struct kvm_irq_routing_table *irq_rt; |
8301 |
+- struct kvm_lapic_irq irq; |
8302 |
+- struct kvm_vcpu *vcpu; |
8303 |
+- struct vcpu_data vcpu_info; |
8304 |
+- int idx, ret = 0; |
8305 |
+- |
8306 |
+- if (!kvm_arch_has_assigned_device(kvm) || |
8307 |
+- !irq_remapping_cap(IRQ_POSTING_CAP) || |
8308 |
+- !kvm_vcpu_apicv_active(kvm->vcpus[0])) |
8309 |
+- return 0; |
8310 |
+- |
8311 |
+- idx = srcu_read_lock(&kvm->irq_srcu); |
8312 |
+- irq_rt = srcu_dereference(kvm->irq_routing, &kvm->irq_srcu); |
8313 |
+- if (guest_irq >= irq_rt->nr_rt_entries || |
8314 |
+- hlist_empty(&irq_rt->map[guest_irq])) { |
8315 |
+- pr_warn_once("no route for guest_irq %u/%u (broken user space?)\n", |
8316 |
+- guest_irq, irq_rt->nr_rt_entries); |
8317 |
+- goto out; |
8318 |
+- } |
8319 |
+- |
8320 |
+- hlist_for_each_entry(e, &irq_rt->map[guest_irq], link) { |
8321 |
+- if (e->type != KVM_IRQ_ROUTING_MSI) |
8322 |
+- continue; |
8323 |
+- /* |
8324 |
+- * VT-d PI cannot support posting multicast/broadcast |
8325 |
+- * interrupts to a vCPU, we still use interrupt remapping |
8326 |
+- * for these kind of interrupts. |
8327 |
+- * |
8328 |
+- * For lowest-priority interrupts, we only support |
8329 |
+- * those with single CPU as the destination, e.g. user |
8330 |
+- * configures the interrupts via /proc/irq or uses |
8331 |
+- * irqbalance to make the interrupts single-CPU. |
8332 |
+- * |
8333 |
+- * We will support full lowest-priority interrupt later. |
8334 |
+- * |
8335 |
+- * In addition, we can only inject generic interrupts using |
8336 |
+- * the PI mechanism, refuse to route others through it. |
8337 |
+- */ |
8338 |
+- |
8339 |
+- kvm_set_msi_irq(kvm, e, &irq); |
8340 |
+- if (!kvm_intr_is_single_vcpu(kvm, &irq, &vcpu) || |
8341 |
+- !kvm_irq_is_postable(&irq)) { |
8342 |
+- /* |
8343 |
+- * Make sure the IRTE is in remapped mode if |
8344 |
+- * we don't handle it in posted mode. |
8345 |
+- */ |
8346 |
+- ret = irq_set_vcpu_affinity(host_irq, NULL); |
8347 |
+- if (ret < 0) { |
8348 |
+- printk(KERN_INFO |
8349 |
+- "failed to back to remapped mode, irq: %u\n", |
8350 |
+- host_irq); |
8351 |
+- goto out; |
8352 |
+- } |
8353 |
+- |
8354 |
+- continue; |
8355 |
+- } |
8356 |
+- |
8357 |
+- vcpu_info.pi_desc_addr = __pa(vcpu_to_pi_desc(vcpu)); |
8358 |
+- vcpu_info.vector = irq.vector; |
8359 |
+- |
8360 |
+- trace_kvm_pi_irte_update(host_irq, vcpu->vcpu_id, e->gsi, |
8361 |
+- vcpu_info.vector, vcpu_info.pi_desc_addr, set); |
8362 |
+- |
8363 |
+- if (set) |
8364 |
+- ret = irq_set_vcpu_affinity(host_irq, &vcpu_info); |
8365 |
+- else |
8366 |
+- ret = irq_set_vcpu_affinity(host_irq, NULL); |
8367 |
+- |
8368 |
+- if (ret < 0) { |
8369 |
+- printk(KERN_INFO "%s: failed to update PI IRTE\n", |
8370 |
+- __func__); |
8371 |
+- goto out; |
8372 |
+- } |
8373 |
+- } |
8374 |
+- |
8375 |
+- ret = 0; |
8376 |
+-out: |
8377 |
+- srcu_read_unlock(&kvm->irq_srcu, idx); |
8378 |
+- return ret; |
8379 |
+-} |
8380 |
+- |
8381 |
+-static void vmx_setup_mce(struct kvm_vcpu *vcpu) |
8382 |
+-{ |
8383 |
+- if (vcpu->arch.mcg_cap & MCG_LMCE_P) |
8384 |
+- to_vmx(vcpu)->msr_ia32_feature_control_valid_bits |= |
8385 |
+- FEATURE_CONTROL_LMCE; |
8386 |
+- else |
8387 |
+- to_vmx(vcpu)->msr_ia32_feature_control_valid_bits &= |
8388 |
+- ~FEATURE_CONTROL_LMCE; |
8389 |
+-} |
8390 |
+- |
8391 |
+-static int vmx_smi_allowed(struct kvm_vcpu *vcpu) |
8392 |
+-{ |
8393 |
+- /* we need a nested vmexit to enter SMM, postpone if run is pending */ |
8394 |
+- if (to_vmx(vcpu)->nested.nested_run_pending) |
8395 |
+- return 0; |
8396 |
+- return 1; |
8397 |
+-} |
8398 |
+- |
8399 |
+-static int vmx_pre_enter_smm(struct kvm_vcpu *vcpu, char *smstate) |
8400 |
+-{ |
8401 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
8402 |
+- |
8403 |
+- vmx->nested.smm.guest_mode = is_guest_mode(vcpu); |
8404 |
+- if (vmx->nested.smm.guest_mode) |
8405 |
+- nested_vmx_vmexit(vcpu, -1, 0, 0); |
8406 |
+- |
8407 |
+- vmx->nested.smm.vmxon = vmx->nested.vmxon; |
8408 |
+- vmx->nested.vmxon = false; |
8409 |
+- vmx_clear_hlt(vcpu); |
8410 |
+- return 0; |
8411 |
+-} |
8412 |
+- |
8413 |
+-static int vmx_pre_leave_smm(struct kvm_vcpu *vcpu, const char *smstate) |
8414 |
+-{ |
8415 |
+- struct vcpu_vmx *vmx = to_vmx(vcpu); |
8416 |
+- int ret; |
8417 |
+- |
8418 |
+- if (vmx->nested.smm.vmxon) { |
8419 |
+- vmx->nested.vmxon = true; |
8420 |
+- vmx->nested.smm.vmxon = false; |
8421 |
+- } |
8422 |
+- |
8423 |
+- if (vmx->nested.smm.guest_mode) { |
8424 |
+- ret = nested_vmx_enter_non_root_mode(vcpu, false); |
8425 |
+- if (ret) |
8426 |
+- return ret; |
8427 |
+- |
8428 |
+- vmx->nested.smm.guest_mode = false; |
8429 |
+- } |
8430 |
+- return 0; |
8431 |
+-} |
8432 |
+- |
8433 |
+-static int enable_smi_window(struct kvm_vcpu *vcpu) |
8434 |
+-{ |
8435 |
+- return 0; |
8436 |
+-} |
8437 |
+- |
8438 |
+-static bool vmx_need_emulation_on_page_fault(struct kvm_vcpu *vcpu) |
8439 |
+-{ |
8440 |
+- return false; |
8441 |
+-} |
8442 |
+- |
8443 |
+-static bool vmx_apic_init_signal_blocked(struct kvm_vcpu *vcpu) |
8444 |
+-{ |
8445 |
+- return to_vmx(vcpu)->nested.vmxon; |
8446 |
+-} |
8447 |
+- |
8448 |
+-static __init int hardware_setup(void) |
8449 |
+-{ |
8450 |
+- unsigned long host_bndcfgs; |
8451 |
+- struct desc_ptr dt; |
8452 |
+- int r, i; |
8453 |
+- |
8454 |
+- rdmsrl_safe(MSR_EFER, &host_efer); |
8455 |
+- |
8456 |
+- store_idt(&dt); |
8457 |
+- host_idt_base = dt.address; |
8458 |
+- |
8459 |
+- for (i = 0; i < ARRAY_SIZE(vmx_msr_index); ++i) |
8460 |
+- kvm_define_shared_msr(i, vmx_msr_index[i]); |
8461 |
+- |
8462 |
+- if (setup_vmcs_config(&vmcs_config, &vmx_capability) < 0) |
8463 |
+- return -EIO; |
8464 |
+- |
8465 |
+- if (boot_cpu_has(X86_FEATURE_NX)) |
8466 |
+- kvm_enable_efer_bits(EFER_NX); |
8467 |
+- |
8468 |
+- if (boot_cpu_has(X86_FEATURE_MPX)) { |
8469 |
+- rdmsrl(MSR_IA32_BNDCFGS, host_bndcfgs); |
8470 |
+- WARN_ONCE(host_bndcfgs, "KVM: BNDCFGS in host will be lost"); |
8471 |
+- } |
8472 |
+- |
8473 |
+- if (!cpu_has_vmx_vpid() || !cpu_has_vmx_invvpid() || |
8474 |
+- !(cpu_has_vmx_invvpid_single() || cpu_has_vmx_invvpid_global())) |
8475 |
+- enable_vpid = 0; |
8476 |
+- |
8477 |
+- if (!cpu_has_vmx_ept() || |
8478 |
+- !cpu_has_vmx_ept_4levels() || |
8479 |
+- !cpu_has_vmx_ept_mt_wb() || |
8480 |
+- !cpu_has_vmx_invept_global()) |
8481 |
+- enable_ept = 0; |
8482 |
+- |
8483 |
+- if (!cpu_has_vmx_ept_ad_bits() || !enable_ept) |
8484 |
+- enable_ept_ad_bits = 0; |
8485 |
+- |
8486 |
+- if (!cpu_has_vmx_unrestricted_guest() || !enable_ept) |
8487 |
+- enable_unrestricted_guest = 0; |
8488 |
+- |
8489 |
+- if (!cpu_has_vmx_flexpriority()) |
8490 |
+- flexpriority_enabled = 0; |
8491 |
+- |
8492 |
+- if (!cpu_has_virtual_nmis()) |
8493 |
+- enable_vnmi = 0; |
8494 |
+- |
8495 |
+- /* |
8496 |
+- * set_apic_access_page_addr() is used to reload apic access |
8497 |
+- * page upon invalidation. No need to do anything if not |
8498 |
+- * using the APIC_ACCESS_ADDR VMCS field. |
8499 |
+- */ |
8500 |
+- if (!flexpriority_enabled) |
8501 |
+- kvm_x86_ops->set_apic_access_page_addr = NULL; |
8502 |
+- |
8503 |
+- if (!cpu_has_vmx_tpr_shadow()) |
8504 |
+- kvm_x86_ops->update_cr8_intercept = NULL; |
8505 |
+- |
8506 |
+- if (enable_ept && !cpu_has_vmx_ept_2m_page()) |
8507 |
+- kvm_disable_largepages(); |
8508 |
+- |
8509 |
+-#if IS_ENABLED(CONFIG_HYPERV) |
8510 |
+- if (ms_hyperv.nested_features & HV_X64_NESTED_GUEST_MAPPING_FLUSH |
8511 |
+- && enable_ept) { |
8512 |
+- kvm_x86_ops->tlb_remote_flush = hv_remote_flush_tlb; |
8513 |
+- kvm_x86_ops->tlb_remote_flush_with_range = |
8514 |
+- hv_remote_flush_tlb_with_range; |
8515 |
+- } |
8516 |
+-#endif |
8517 |
+- |
8518 |
+- if (!cpu_has_vmx_ple()) { |
8519 |
+- ple_gap = 0; |
8520 |
+- ple_window = 0; |
8521 |
+- ple_window_grow = 0; |
8522 |
+- ple_window_max = 0; |
8523 |
+- ple_window_shrink = 0; |
8524 |
+- } |
8525 |
+- |
8526 |
+- if (!cpu_has_vmx_apicv()) { |
8527 |
+- enable_apicv = 0; |
8528 |
+- kvm_x86_ops->sync_pir_to_irr = NULL; |
8529 |
+- } |
8530 |
+- |
8531 |
+- if (cpu_has_vmx_tsc_scaling()) { |
8532 |
+- kvm_has_tsc_control = true; |
8533 |
+- kvm_max_tsc_scaling_ratio = KVM_VMX_TSC_MULTIPLIER_MAX; |
8534 |
+- kvm_tsc_scaling_ratio_frac_bits = 48; |
8535 |
+- } |
8536 |
+- |
8537 |
+- set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */ |
8538 |
+- |
8539 |
+- if (enable_ept) |
8540 |
+- vmx_enable_tdp(); |
8541 |
+- else |
8542 |
+- kvm_disable_tdp(); |
8543 |
+- |
8544 |
+- /* |
8545 |
+- * Only enable PML when hardware supports PML feature, and both EPT |
8546 |
+- * and EPT A/D bit features are enabled -- PML depends on them to work. |
8547 |
+- */ |
8548 |
+- if (!enable_ept || !enable_ept_ad_bits || !cpu_has_vmx_pml()) |
8549 |
+- enable_pml = 0; |
8550 |
+- |
8551 |
+- if (!enable_pml) { |
8552 |
+- kvm_x86_ops->slot_enable_log_dirty = NULL; |
8553 |
+- kvm_x86_ops->slot_disable_log_dirty = NULL; |
8554 |
+- kvm_x86_ops->flush_log_dirty = NULL; |
8555 |
+- kvm_x86_ops->enable_log_dirty_pt_masked = NULL; |
8556 |
+- } |
8557 |
+- |
8558 |
+- if (!cpu_has_vmx_preemption_timer()) |
8559 |
+- enable_preemption_timer = false; |
8560 |
+- |
8561 |
+- if (enable_preemption_timer) { |
8562 |
+- u64 use_timer_freq = 5000ULL * 1000 * 1000; |
8563 |
+- u64 vmx_msr; |
8564 |
+- |
8565 |
+- rdmsrl(MSR_IA32_VMX_MISC, vmx_msr); |
8566 |
+- cpu_preemption_timer_multi = |
8567 |
+- vmx_msr & VMX_MISC_PREEMPTION_TIMER_RATE_MASK; |
8568 |
+- |
8569 |
+- if (tsc_khz) |
8570 |
+- use_timer_freq = (u64)tsc_khz * 1000; |
8571 |
+- use_timer_freq >>= cpu_preemption_timer_multi; |
8572 |
+- |
8573 |
+- /* |
8574 |
+- * KVM "disables" the preemption timer by setting it to its max |
8575 |
+- * value. Don't use the timer if it might cause spurious exits |
8576 |
+- * at a rate faster than 0.1 Hz (of uninterrupted guest time). |
8577 |
+- */ |
8578 |
+- if (use_timer_freq > 0xffffffffu / 10) |
8579 |
+- enable_preemption_timer = false; |
8580 |
+- } |
8581 |
+- |
8582 |
+- if (!enable_preemption_timer) { |
8583 |
+- kvm_x86_ops->set_hv_timer = NULL; |
8584 |
+- kvm_x86_ops->cancel_hv_timer = NULL; |
8585 |
+- kvm_x86_ops->request_immediate_exit = __kvm_request_immediate_exit; |
8586 |
+- } |
8587 |
+- |
8588 |
+- kvm_set_posted_intr_wakeup_handler(wakeup_handler); |
8589 |
+- |
8590 |
+- kvm_mce_cap_supported |= MCG_LMCE_P; |
8591 |
+- |
8592 |
+- if (pt_mode != PT_MODE_SYSTEM && pt_mode != PT_MODE_HOST_GUEST) |
8593 |
+- return -EINVAL; |
8594 |
+- if (!enable_ept || !cpu_has_vmx_intel_pt()) |
8595 |
+- pt_mode = PT_MODE_SYSTEM; |
8596 |
+- |
8597 |
+- if (nested) { |
8598 |
+- nested_vmx_setup_ctls_msrs(&vmcs_config.nested, |
8599 |
+- vmx_capability.ept, enable_apicv); |
8600 |
+- |
8601 |
+- r = nested_vmx_hardware_setup(kvm_vmx_exit_handlers); |
8602 |
+- if (r) |
8603 |
+- return r; |
8604 |
+- } |
8605 |
+- |
8606 |
+- r = alloc_kvm_area(); |
8607 |
+- if (r) |
8608 |
+- nested_vmx_hardware_unsetup(); |
8609 |
+- return r; |
8610 |
+-} |
8611 |
+- |
8612 |
+-static __exit void hardware_unsetup(void) |
8613 |
+-{ |
8614 |
+- if (nested) |
8615 |
+- nested_vmx_hardware_unsetup(); |
8616 |
+- |
8617 |
+- free_kvm_area(); |
8618 |
+-} |
8619 |
+- |
8620 |
+-static struct kvm_x86_ops vmx_x86_ops __ro_after_init = { |
8621 |
+- .cpu_has_kvm_support = cpu_has_kvm_support, |
8622 |
+- .disabled_by_bios = vmx_disabled_by_bios, |
8623 |
+- .hardware_setup = hardware_setup, |
8624 |
+- .hardware_unsetup = hardware_unsetup, |
8625 |
+- .check_processor_compatibility = vmx_check_processor_compat, |
8626 |
+- .hardware_enable = hardware_enable, |
8627 |
+- .hardware_disable = hardware_disable, |
8628 |
+- .cpu_has_accelerated_tpr = report_flexpriority, |
8629 |
+- .has_emulated_msr = vmx_has_emulated_msr, |
8630 |
+- |
8631 |
+- .vm_init = vmx_vm_init, |
8632 |
+- .vm_alloc = vmx_vm_alloc, |
8633 |
+- .vm_free = vmx_vm_free, |
8634 |
+- |
8635 |
+- .vcpu_create = vmx_create_vcpu, |
8636 |
+- .vcpu_free = vmx_free_vcpu, |
8637 |
+- .vcpu_reset = vmx_vcpu_reset, |
8638 |
+- |
8639 |
+- .prepare_guest_switch = vmx_prepare_switch_to_guest, |
8640 |
+- .vcpu_load = vmx_vcpu_load, |
8641 |
+- .vcpu_put = vmx_vcpu_put, |
8642 |
+- |
8643 |
+- .update_bp_intercept = update_exception_bitmap, |
8644 |
+- .get_msr_feature = vmx_get_msr_feature, |
8645 |
+- .get_msr = vmx_get_msr, |
8646 |
+- .set_msr = vmx_set_msr, |
8647 |
+- .get_segment_base = vmx_get_segment_base, |
8648 |
+- .get_segment = vmx_get_segment, |
8649 |
+- .set_segment = vmx_set_segment, |
8650 |
+- .get_cpl = vmx_get_cpl, |
8651 |
+- .get_cs_db_l_bits = vmx_get_cs_db_l_bits, |
8652 |
+- .decache_cr0_guest_bits = vmx_decache_cr0_guest_bits, |
8653 |
+- .decache_cr4_guest_bits = vmx_decache_cr4_guest_bits, |
8654 |
+- .set_cr0 = vmx_set_cr0, |
8655 |
+- .set_cr3 = vmx_set_cr3, |
8656 |
+- .set_cr4 = vmx_set_cr4, |
8657 |
+- .set_efer = vmx_set_efer, |
8658 |
+- .get_idt = vmx_get_idt, |
8659 |
+- .set_idt = vmx_set_idt, |
8660 |
+- .get_gdt = vmx_get_gdt, |
8661 |
+- .set_gdt = vmx_set_gdt, |
8662 |
+- .get_dr6 = vmx_get_dr6, |
8663 |
+- .set_dr6 = vmx_set_dr6, |
8664 |
+- .set_dr7 = vmx_set_dr7, |
8665 |
+- .sync_dirty_debug_regs = vmx_sync_dirty_debug_regs, |
8666 |
+- .cache_reg = vmx_cache_reg, |
8667 |
+- .get_rflags = vmx_get_rflags, |
8668 |
+- .set_rflags = vmx_set_rflags, |
8669 |
+- |
8670 |
+- .tlb_flush = vmx_flush_tlb, |
8671 |
+- .tlb_flush_gva = vmx_flush_tlb_gva, |
8672 |
+- |
8673 |
+- .run = vmx_vcpu_run, |
8674 |
+- .handle_exit = vmx_handle_exit, |
8675 |
+- .skip_emulated_instruction = skip_emulated_instruction, |
8676 |
+- .set_interrupt_shadow = vmx_set_interrupt_shadow, |
8677 |
+- .get_interrupt_shadow = vmx_get_interrupt_shadow, |
8678 |
+- .patch_hypercall = vmx_patch_hypercall, |
8679 |
+- .set_irq = vmx_inject_irq, |
8680 |
+- .set_nmi = vmx_inject_nmi, |
8681 |
+- .queue_exception = vmx_queue_exception, |
8682 |
+- .cancel_injection = vmx_cancel_injection, |
8683 |
+- .interrupt_allowed = vmx_interrupt_allowed, |
8684 |
+- .nmi_allowed = vmx_nmi_allowed, |
8685 |
+- .get_nmi_mask = vmx_get_nmi_mask, |
8686 |
+- .set_nmi_mask = vmx_set_nmi_mask, |
8687 |
+- .enable_nmi_window = enable_nmi_window, |
8688 |
+- .enable_irq_window = enable_irq_window, |
8689 |
+- .update_cr8_intercept = update_cr8_intercept, |
8690 |
+- .set_virtual_apic_mode = vmx_set_virtual_apic_mode, |
8691 |
+- .set_apic_access_page_addr = vmx_set_apic_access_page_addr, |
8692 |
+- .get_enable_apicv = vmx_get_enable_apicv, |
8693 |
+- .refresh_apicv_exec_ctrl = vmx_refresh_apicv_exec_ctrl, |
8694 |
+- .load_eoi_exitmap = vmx_load_eoi_exitmap, |
8695 |
+- .apicv_post_state_restore = vmx_apicv_post_state_restore, |
8696 |
+- .hwapic_irr_update = vmx_hwapic_irr_update, |
8697 |
+- .hwapic_isr_update = vmx_hwapic_isr_update, |
8698 |
+- .guest_apic_has_interrupt = vmx_guest_apic_has_interrupt, |
8699 |
+- .sync_pir_to_irr = vmx_sync_pir_to_irr, |
8700 |
+- .deliver_posted_interrupt = vmx_deliver_posted_interrupt, |
8701 |
+- .dy_apicv_has_pending_interrupt = vmx_dy_apicv_has_pending_interrupt, |
8702 |
+- |
8703 |
+- .set_tss_addr = vmx_set_tss_addr, |
8704 |
+- .set_identity_map_addr = vmx_set_identity_map_addr, |
8705 |
+- .get_tdp_level = get_ept_level, |
8706 |
+- .get_mt_mask = vmx_get_mt_mask, |
8707 |
+- |
8708 |
+- .get_exit_info = vmx_get_exit_info, |
8709 |
+- |
8710 |
+- .get_lpage_level = vmx_get_lpage_level, |
8711 |
+- |
8712 |
+- .cpuid_update = vmx_cpuid_update, |
8713 |
+- |
8714 |
+- .rdtscp_supported = vmx_rdtscp_supported, |
8715 |
+- .invpcid_supported = vmx_invpcid_supported, |
8716 |
+- |
8717 |
+- .set_supported_cpuid = vmx_set_supported_cpuid, |
8718 |
+- |
8719 |
+- .has_wbinvd_exit = cpu_has_vmx_wbinvd_exit, |
8720 |
+- |
8721 |
+- .read_l1_tsc_offset = vmx_read_l1_tsc_offset, |
8722 |
+- .write_l1_tsc_offset = vmx_write_l1_tsc_offset, |
8723 |
+- |
8724 |
+- .set_tdp_cr3 = vmx_set_cr3, |
8725 |
+- |
8726 |
+- .check_intercept = vmx_check_intercept, |
8727 |
+- .handle_exit_irqoff = vmx_handle_exit_irqoff, |
8728 |
+- .mpx_supported = vmx_mpx_supported, |
8729 |
+- .xsaves_supported = vmx_xsaves_supported, |
8730 |
+- .umip_emulated = vmx_umip_emulated, |
8731 |
+- .pt_supported = vmx_pt_supported, |
8732 |
+- |
8733 |
+- .request_immediate_exit = vmx_request_immediate_exit, |
8734 |
+- |
8735 |
+- .sched_in = vmx_sched_in, |
8736 |
+- |
8737 |
+- .slot_enable_log_dirty = vmx_slot_enable_log_dirty, |
8738 |
+- .slot_disable_log_dirty = vmx_slot_disable_log_dirty, |
8739 |
+- .flush_log_dirty = vmx_flush_log_dirty, |
8740 |
+- .enable_log_dirty_pt_masked = vmx_enable_log_dirty_pt_masked, |
8741 |
+- .write_log_dirty = vmx_write_pml_buffer, |
8742 |
+- |
8743 |
+- .pre_block = vmx_pre_block, |
8744 |
+- .post_block = vmx_post_block, |
8745 |
+- |
8746 |
+- .pmu_ops = &intel_pmu_ops, |
8747 |
+- |
8748 |
+- .update_pi_irte = vmx_update_pi_irte, |
8749 |
+- |
8750 |
+-#ifdef CONFIG_X86_64 |
8751 |
+- .set_hv_timer = vmx_set_hv_timer, |
8752 |
+- .cancel_hv_timer = vmx_cancel_hv_timer, |
8753 |
+-#endif |
8754 |
+- |
8755 |
+- .setup_mce = vmx_setup_mce, |
8756 |
+- |
8757 |
+- .smi_allowed = vmx_smi_allowed, |
8758 |
+- .pre_enter_smm = vmx_pre_enter_smm, |
8759 |
+- .pre_leave_smm = vmx_pre_leave_smm, |
8760 |
+- .enable_smi_window = enable_smi_window, |
8761 |
+- |
8762 |
+- .check_nested_events = NULL, |
8763 |
+- .get_nested_state = NULL, |
8764 |
+- .set_nested_state = NULL, |
8765 |
+- .get_vmcs12_pages = NULL, |
8766 |
+- .nested_enable_evmcs = NULL, |
8767 |
+- .nested_get_evmcs_version = NULL, |
8768 |
+- .need_emulation_on_page_fault = vmx_need_emulation_on_page_fault, |
8769 |
+- .apic_init_signal_blocked = vmx_apic_init_signal_blocked, |
8770 |
+-}; |
8771 |
+- |
8772 |
+-static void vmx_cleanup_l1d_flush(void) |
8773 |
+-{ |
8774 |
+- if (vmx_l1d_flush_pages) { |
8775 |
+- free_pages((unsigned long)vmx_l1d_flush_pages, L1D_CACHE_ORDER); |
8776 |
+- vmx_l1d_flush_pages = NULL; |
8777 |
+- } |
8778 |
+- /* Restore state so sysfs ignores VMX */ |
8779 |
+- l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_AUTO; |
8780 |
+-} |
8781 |
+- |
8782 |
+-static void vmx_exit(void) |
8783 |
+-{ |
8784 |
+-#ifdef CONFIG_KEXEC_CORE |
8785 |
+- RCU_INIT_POINTER(crash_vmclear_loaded_vmcss, NULL); |
8786 |
+- synchronize_rcu(); |
8787 |
+-#endif |
8788 |
+- |
8789 |
+- kvm_exit(); |
8790 |
+- |
8791 |
+-#if IS_ENABLED(CONFIG_HYPERV) |
8792 |
+- if (static_branch_unlikely(&enable_evmcs)) { |
8793 |
+- int cpu; |
8794 |
+- struct hv_vp_assist_page *vp_ap; |
8795 |
+- /* |
8796 |
+- * Reset everything to support using non-enlightened VMCS |
8797 |
+- * access later (e.g. when we reload the module with |
8798 |
+- * enlightened_vmcs=0) |
8799 |
+- */ |
8800 |
+- for_each_online_cpu(cpu) { |
8801 |
+- vp_ap = hv_get_vp_assist_page(cpu); |
8802 |
+- |
8803 |
+- if (!vp_ap) |
8804 |
+- continue; |
8805 |
+- |
8806 |
+- vp_ap->nested_control.features.directhypercall = 0; |
8807 |
+- vp_ap->current_nested_vmcs = 0; |
8808 |
+- vp_ap->enlighten_vmentry = 0; |
8809 |
+- } |
8810 |
+- |
8811 |
+- static_branch_disable(&enable_evmcs); |
8812 |
+- } |
8813 |
+-#endif |
8814 |
+- vmx_cleanup_l1d_flush(); |
8815 |
+-} |
8816 |
+-module_exit(vmx_exit); |
8817 |
+- |
8818 |
+-static int __init vmx_init(void) |
8819 |
+-{ |
8820 |
+- int r; |
8821 |
+- |
8822 |
+-#if IS_ENABLED(CONFIG_HYPERV) |
8823 |
+- /* |
8824 |
+- * Enlightened VMCS usage should be recommended and the host needs |
8825 |
+- * to support eVMCS v1 or above. We can also disable eVMCS support |
8826 |
+- * with module parameter. |
8827 |
+- */ |
8828 |
+- if (enlightened_vmcs && |
8829 |
+- ms_hyperv.hints & HV_X64_ENLIGHTENED_VMCS_RECOMMENDED && |
8830 |
+- (ms_hyperv.nested_features & HV_X64_ENLIGHTENED_VMCS_VERSION) >= |
8831 |
+- KVM_EVMCS_VERSION) { |
8832 |
+- int cpu; |
8833 |
+- |
8834 |
+- /* Check that we have assist pages on all online CPUs */ |
8835 |
+- for_each_online_cpu(cpu) { |
8836 |
+- if (!hv_get_vp_assist_page(cpu)) { |
8837 |
+- enlightened_vmcs = false; |
8838 |
+- break; |
8839 |
+- } |
8840 |
+- } |
8841 |
+- |
8842 |
+- if (enlightened_vmcs) { |
8843 |
+- pr_info("KVM: vmx: using Hyper-V Enlightened VMCS\n"); |
8844 |
+- static_branch_enable(&enable_evmcs); |
8845 |
+- } |
8846 |
+- |
8847 |
+- if (ms_hyperv.nested_features & HV_X64_NESTED_DIRECT_FLUSH) |
8848 |
+- vmx_x86_ops.enable_direct_tlbflush |
8849 |
+- = hv_enable_direct_tlbflush; |
8850 |
+- |
8851 |
+- } else { |
8852 |
+- enlightened_vmcs = false; |
8853 |
+- } |
8854 |
+-#endif |
8855 |
+- |
8856 |
+- r = kvm_init(&vmx_x86_ops, sizeof(struct vcpu_vmx), |
8857 |
+- __alignof__(struct vcpu_vmx), THIS_MODULE); |
8858 |
+- if (r) |
8859 |
+- return r; |
8860 |
+- |
8861 |
+- /* |
8862 |
+- * Must be called after kvm_init() so enable_ept is properly set |
8863 |
+- * up. Hand the parameter mitigation value in which was stored in |
8864 |
+- * the pre module init parser. If no parameter was given, it will |
8865 |
+- * contain 'auto' which will be turned into the default 'cond' |
8866 |
+- * mitigation mode. |
8867 |
+- */ |
8868 |
+- r = vmx_setup_l1d_flush(vmentry_l1d_flush_param); |
8869 |
+- if (r) { |
8870 |
+- vmx_exit(); |
8871 |
+- return r; |
8872 |
+- } |
8873 |
+- |
8874 |
+-#ifdef CONFIG_KEXEC_CORE |
8875 |
+- rcu_assign_pointer(crash_vmclear_loaded_vmcss, |
8876 |
+- crash_vmclear_local_loaded_vmcss); |
8877 |
+-#endif |
8878 |
+- vmx_check_vmcs12_offsets(); |
8879 |
+- |
8880 |
+- return 0; |
8881 |
+-} |
8882 |
+-module_init(vmx_init); |
8883 |
+diff --git a/arch/x86/lib/x86-opcode-map.txt b/arch/x86/lib/x86-opcode-map.txt |
8884 |
+index 0a0e9112f284..5cb9f009f2be 100644 |
8885 |
+--- a/arch/x86/lib/x86-opcode-map.txt |
8886 |
++++ b/arch/x86/lib/x86-opcode-map.txt |
8887 |
+@@ -909,7 +909,7 @@ EndTable |
8888 |
+ |
8889 |
+ GrpTable: Grp3_2 |
8890 |
+ 0: TEST Ev,Iz |
8891 |
+-1: |
8892 |
++1: TEST Ev,Iz |
8893 |
+ 2: NOT Ev |
8894 |
+ 3: NEG Ev |
8895 |
+ 4: MUL rAX,Ev |
8896 |
+diff --git a/arch/x86/mm/pageattr.c b/arch/x86/mm/pageattr.c |
8897 |
+index e2d4b25c7aa4..101f3ad0d6ad 100644 |
8898 |
+--- a/arch/x86/mm/pageattr.c |
8899 |
++++ b/arch/x86/mm/pageattr.c |
8900 |
+@@ -2126,19 +2126,13 @@ int kernel_map_pages_in_pgd(pgd_t *pgd, u64 pfn, unsigned long address, |
8901 |
+ .pgd = pgd, |
8902 |
+ .numpages = numpages, |
8903 |
+ .mask_set = __pgprot(0), |
8904 |
+- .mask_clr = __pgprot(0), |
8905 |
++ .mask_clr = __pgprot(~page_flags & (_PAGE_NX|_PAGE_RW)), |
8906 |
+ .flags = 0, |
8907 |
+ }; |
8908 |
+ |
8909 |
+ if (!(__supported_pte_mask & _PAGE_NX)) |
8910 |
+ goto out; |
8911 |
+ |
8912 |
+- if (!(page_flags & _PAGE_NX)) |
8913 |
+- cpa.mask_clr = __pgprot(_PAGE_NX); |
8914 |
+- |
8915 |
+- if (!(page_flags & _PAGE_RW)) |
8916 |
+- cpa.mask_clr = __pgprot(_PAGE_RW); |
8917 |
+- |
8918 |
+ if (!(page_flags & _PAGE_ENC)) |
8919 |
+ cpa.mask_clr = pgprot_encrypted(cpa.mask_clr); |
8920 |
+ |
8921 |
+diff --git a/arch/x86/platform/efi/efi.c b/arch/x86/platform/efi/efi.c |
8922 |
+index 335a62e74a2e..e7f19dec16b9 100644 |
8923 |
+--- a/arch/x86/platform/efi/efi.c |
8924 |
++++ b/arch/x86/platform/efi/efi.c |
8925 |
+@@ -480,7 +480,6 @@ void __init efi_init(void) |
8926 |
+ efi_char16_t *c16; |
8927 |
+ char vendor[100] = "unknown"; |
8928 |
+ int i = 0; |
8929 |
+- void *tmp; |
8930 |
+ |
8931 |
+ #ifdef CONFIG_X86_32 |
8932 |
+ if (boot_params.efi_info.efi_systab_hi || |
8933 |
+@@ -505,14 +504,16 @@ void __init efi_init(void) |
8934 |
+ /* |
8935 |
+ * Show what we know for posterity |
8936 |
+ */ |
8937 |
+- c16 = tmp = early_memremap(efi.systab->fw_vendor, 2); |
8938 |
++ c16 = early_memremap_ro(efi.systab->fw_vendor, |
8939 |
++ sizeof(vendor) * sizeof(efi_char16_t)); |
8940 |
+ if (c16) { |
8941 |
+- for (i = 0; i < sizeof(vendor) - 1 && *c16; ++i) |
8942 |
+- vendor[i] = *c16++; |
8943 |
++ for (i = 0; i < sizeof(vendor) - 1 && c16[i]; ++i) |
8944 |
++ vendor[i] = c16[i]; |
8945 |
+ vendor[i] = '\0'; |
8946 |
+- } else |
8947 |
++ early_memunmap(c16, sizeof(vendor) * sizeof(efi_char16_t)); |
8948 |
++ } else { |
8949 |
+ pr_err("Could not map the firmware vendor!\n"); |
8950 |
+- early_memunmap(tmp, 2); |
8951 |
++ } |
8952 |
+ |
8953 |
+ pr_info("EFI v%u.%.02u by %s\n", |
8954 |
+ efi.systab->hdr.revision >> 16, |
8955 |
+@@ -929,16 +930,14 @@ static void __init __efi_enter_virtual_mode(void) |
8956 |
+ |
8957 |
+ if (efi_alloc_page_tables()) { |
8958 |
+ pr_err("Failed to allocate EFI page tables\n"); |
8959 |
+- clear_bit(EFI_RUNTIME_SERVICES, &efi.flags); |
8960 |
+- return; |
8961 |
++ goto err; |
8962 |
+ } |
8963 |
+ |
8964 |
+ efi_merge_regions(); |
8965 |
+ new_memmap = efi_map_regions(&count, &pg_shift); |
8966 |
+ if (!new_memmap) { |
8967 |
+ pr_err("Error reallocating memory, EFI runtime non-functional!\n"); |
8968 |
+- clear_bit(EFI_RUNTIME_SERVICES, &efi.flags); |
8969 |
+- return; |
8970 |
++ goto err; |
8971 |
+ } |
8972 |
+ |
8973 |
+ pa = __pa(new_memmap); |
8974 |
+@@ -952,8 +951,7 @@ static void __init __efi_enter_virtual_mode(void) |
8975 |
+ |
8976 |
+ if (efi_memmap_init_late(pa, efi.memmap.desc_size * count)) { |
8977 |
+ pr_err("Failed to remap late EFI memory map\n"); |
8978 |
+- clear_bit(EFI_RUNTIME_SERVICES, &efi.flags); |
8979 |
+- return; |
8980 |
++ goto err; |
8981 |
+ } |
8982 |
+ |
8983 |
+ if (efi_enabled(EFI_DBG)) { |
8984 |
+@@ -961,12 +959,11 @@ static void __init __efi_enter_virtual_mode(void) |
8985 |
+ efi_print_memmap(); |
8986 |
+ } |
8987 |
+ |
8988 |
+- BUG_ON(!efi.systab); |
8989 |
++ if (WARN_ON(!efi.systab)) |
8990 |
++ goto err; |
8991 |
+ |
8992 |
+- if (efi_setup_page_tables(pa, 1 << pg_shift)) { |
8993 |
+- clear_bit(EFI_RUNTIME_SERVICES, &efi.flags); |
8994 |
+- return; |
8995 |
+- } |
8996 |
++ if (efi_setup_page_tables(pa, 1 << pg_shift)) |
8997 |
++ goto err; |
8998 |
+ |
8999 |
+ efi_sync_low_kernel_mappings(); |
9000 |
+ |
9001 |
+@@ -986,9 +983,9 @@ static void __init __efi_enter_virtual_mode(void) |
9002 |
+ } |
9003 |
+ |
9004 |
+ if (status != EFI_SUCCESS) { |
9005 |
+- pr_alert("Unable to switch EFI into virtual mode (status=%lx)!\n", |
9006 |
+- status); |
9007 |
+- panic("EFI call to SetVirtualAddressMap() failed!"); |
9008 |
++ pr_err("Unable to switch EFI into virtual mode (status=%lx)!\n", |
9009 |
++ status); |
9010 |
++ goto err; |
9011 |
+ } |
9012 |
+ |
9013 |
+ /* |
9014 |
+@@ -1015,6 +1012,10 @@ static void __init __efi_enter_virtual_mode(void) |
9015 |
+ |
9016 |
+ /* clean DUMMY object */ |
9017 |
+ efi_delete_dummy_variable(); |
9018 |
++ return; |
9019 |
++ |
9020 |
++err: |
9021 |
++ clear_bit(EFI_RUNTIME_SERVICES, &efi.flags); |
9022 |
+ } |
9023 |
+ |
9024 |
+ void __init efi_enter_virtual_mode(void) |
9025 |
+diff --git a/arch/x86/platform/efi/efi_64.c b/arch/x86/platform/efi/efi_64.c |
9026 |
+index ee5d08f25ce4..6db8f3598c80 100644 |
9027 |
+--- a/arch/x86/platform/efi/efi_64.c |
9028 |
++++ b/arch/x86/platform/efi/efi_64.c |
9029 |
+@@ -389,11 +389,12 @@ int __init efi_setup_page_tables(unsigned long pa_memmap, unsigned num_pages) |
9030 |
+ return 0; |
9031 |
+ |
9032 |
+ page = alloc_page(GFP_KERNEL|__GFP_DMA32); |
9033 |
+- if (!page) |
9034 |
+- panic("Unable to allocate EFI runtime stack < 4GB\n"); |
9035 |
++ if (!page) { |
9036 |
++ pr_err("Unable to allocate EFI runtime stack < 4GB\n"); |
9037 |
++ return 1; |
9038 |
++ } |
9039 |
+ |
9040 |
+- efi_scratch.phys_stack = virt_to_phys(page_address(page)); |
9041 |
+- efi_scratch.phys_stack += PAGE_SIZE; /* stack grows down */ |
9042 |
++ efi_scratch.phys_stack = page_to_phys(page + 1); /* stack grows down */ |
9043 |
+ |
9044 |
+ npages = (_etext - _text) >> PAGE_SHIFT; |
9045 |
+ text = __pa(_text); |
9046 |
+diff --git a/drivers/acpi/acpica/dsfield.c b/drivers/acpi/acpica/dsfield.c |
9047 |
+index 30fe89545d6a..bcc6a7acc576 100644 |
9048 |
+--- a/drivers/acpi/acpica/dsfield.c |
9049 |
++++ b/drivers/acpi/acpica/dsfield.c |
9050 |
+@@ -244,7 +244,7 @@ cleanup: |
9051 |
+ * FUNCTION: acpi_ds_get_field_names |
9052 |
+ * |
9053 |
+ * PARAMETERS: info - create_field info structure |
9054 |
+- * ` walk_state - Current method state |
9055 |
++ * walk_state - Current method state |
9056 |
+ * arg - First parser arg for the field name list |
9057 |
+ * |
9058 |
+ * RETURN: Status |
9059 |
+diff --git a/drivers/acpi/acpica/dswload.c b/drivers/acpi/acpica/dswload.c |
9060 |
+index d06c41446282..ba53662f1217 100644 |
9061 |
+--- a/drivers/acpi/acpica/dswload.c |
9062 |
++++ b/drivers/acpi/acpica/dswload.c |
9063 |
+@@ -412,6 +412,27 @@ acpi_status acpi_ds_load1_end_op(struct acpi_walk_state *walk_state) |
9064 |
+ ACPI_DEBUG_PRINT((ACPI_DB_DISPATCH, "Op=%p State=%p\n", op, |
9065 |
+ walk_state)); |
9066 |
+ |
9067 |
++ /* |
9068 |
++ * Disassembler: handle create field operators here. |
9069 |
++ * |
9070 |
++ * create_buffer_field is a deferred op that is typically processed in load |
9071 |
++ * pass 2. However, disassembly of control method contents walk the parse |
9072 |
++ * tree with ACPI_PARSE_LOAD_PASS1 and AML_CREATE operators are processed |
9073 |
++ * in a later walk. This is a problem when there is a control method that |
9074 |
++ * has the same name as the AML_CREATE object. In this case, any use of the |
9075 |
++ * name segment will be detected as a method call rather than a reference |
9076 |
++ * to a buffer field. |
9077 |
++ * |
9078 |
++ * This earlier creation during disassembly solves this issue by inserting |
9079 |
++ * the named object in the ACPI namespace so that references to this name |
9080 |
++ * would be a name string rather than a method call. |
9081 |
++ */ |
9082 |
++ if ((walk_state->parse_flags & ACPI_PARSE_DISASSEMBLE) && |
9083 |
++ (walk_state->op_info->flags & AML_CREATE)) { |
9084 |
++ status = acpi_ds_create_buffer_field(op, walk_state); |
9085 |
++ return_ACPI_STATUS(status); |
9086 |
++ } |
9087 |
++ |
9088 |
+ /* We are only interested in opcodes that have an associated name */ |
9089 |
+ |
9090 |
+ if (!(walk_state->op_info->flags & (AML_NAMED | AML_FIELD))) { |
9091 |
+diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c |
9092 |
+index a25d77b3a16a..d5c19e25ddf5 100644 |
9093 |
+--- a/drivers/acpi/button.c |
9094 |
++++ b/drivers/acpi/button.c |
9095 |
+@@ -102,6 +102,17 @@ static const struct dmi_system_id lid_blacklst[] = { |
9096 |
+ }, |
9097 |
+ .driver_data = (void *)(long)ACPI_BUTTON_LID_INIT_OPEN, |
9098 |
+ }, |
9099 |
++ { |
9100 |
++ /* |
9101 |
++ * Razer Blade Stealth 13 late 2019, notification of the LID device |
9102 |
++ * only happens on close, not on open and _LID always returns closed. |
9103 |
++ */ |
9104 |
++ .matches = { |
9105 |
++ DMI_MATCH(DMI_SYS_VENDOR, "Razer"), |
9106 |
++ DMI_MATCH(DMI_PRODUCT_NAME, "Razer Blade Stealth 13 Late 2019"), |
9107 |
++ }, |
9108 |
++ .driver_data = (void *)(long)ACPI_BUTTON_LID_INIT_OPEN, |
9109 |
++ }, |
9110 |
+ {} |
9111 |
+ }; |
9112 |
+ |
9113 |
+diff --git a/drivers/atm/fore200e.c b/drivers/atm/fore200e.c |
9114 |
+index 99a38115b0a8..86aab14872fd 100644 |
9115 |
+--- a/drivers/atm/fore200e.c |
9116 |
++++ b/drivers/atm/fore200e.c |
9117 |
+@@ -1504,12 +1504,14 @@ fore200e_open(struct atm_vcc *vcc) |
9118 |
+ static void |
9119 |
+ fore200e_close(struct atm_vcc* vcc) |
9120 |
+ { |
9121 |
+- struct fore200e* fore200e = FORE200E_DEV(vcc->dev); |
9122 |
+ struct fore200e_vcc* fore200e_vcc; |
9123 |
++ struct fore200e* fore200e; |
9124 |
+ struct fore200e_vc_map* vc_map; |
9125 |
+ unsigned long flags; |
9126 |
+ |
9127 |
+ ASSERT(vcc); |
9128 |
++ fore200e = FORE200E_DEV(vcc->dev); |
9129 |
++ |
9130 |
+ ASSERT((vcc->vpi >= 0) && (vcc->vpi < 1<<FORE200E_VPI_BITS)); |
9131 |
+ ASSERT((vcc->vci >= 0) && (vcc->vci < 1<<FORE200E_VCI_BITS)); |
9132 |
+ |
9133 |
+@@ -1554,10 +1556,10 @@ fore200e_close(struct atm_vcc* vcc) |
9134 |
+ static int |
9135 |
+ fore200e_send(struct atm_vcc *vcc, struct sk_buff *skb) |
9136 |
+ { |
9137 |
+- struct fore200e* fore200e = FORE200E_DEV(vcc->dev); |
9138 |
+- struct fore200e_vcc* fore200e_vcc = FORE200E_VCC(vcc); |
9139 |
++ struct fore200e* fore200e; |
9140 |
++ struct fore200e_vcc* fore200e_vcc; |
9141 |
+ struct fore200e_vc_map* vc_map; |
9142 |
+- struct host_txq* txq = &fore200e->host_txq; |
9143 |
++ struct host_txq* txq; |
9144 |
+ struct host_txq_entry* entry; |
9145 |
+ struct tpd* tpd; |
9146 |
+ struct tpd_haddr tpd_haddr; |
9147 |
+@@ -1570,9 +1572,18 @@ fore200e_send(struct atm_vcc *vcc, struct sk_buff *skb) |
9148 |
+ unsigned char* data; |
9149 |
+ unsigned long flags; |
9150 |
+ |
9151 |
+- ASSERT(vcc); |
9152 |
+- ASSERT(fore200e); |
9153 |
+- ASSERT(fore200e_vcc); |
9154 |
++ if (!vcc) |
9155 |
++ return -EINVAL; |
9156 |
++ |
9157 |
++ fore200e = FORE200E_DEV(vcc->dev); |
9158 |
++ fore200e_vcc = FORE200E_VCC(vcc); |
9159 |
++ |
9160 |
++ if (!fore200e) |
9161 |
++ return -EINVAL; |
9162 |
++ |
9163 |
++ txq = &fore200e->host_txq; |
9164 |
++ if (!fore200e_vcc) |
9165 |
++ return -EINVAL; |
9166 |
+ |
9167 |
+ if (!test_bit(ATM_VF_READY, &vcc->flags)) { |
9168 |
+ DPRINTK(1, "VC %d.%d.%d not ready for tx\n", vcc->itf, vcc->vpi, vcc->vpi); |
9169 |
+diff --git a/drivers/base/dd.c b/drivers/base/dd.c |
9170 |
+index 11d24a552ee4..5f6416e6ba96 100644 |
9171 |
+--- a/drivers/base/dd.c |
9172 |
++++ b/drivers/base/dd.c |
9173 |
+@@ -470,7 +470,10 @@ static int really_probe(struct device *dev, struct device_driver *drv) |
9174 |
+ atomic_inc(&probe_count); |
9175 |
+ pr_debug("bus: '%s': %s: probing driver %s with device %s\n", |
9176 |
+ drv->bus->name, __func__, drv->name, dev_name(dev)); |
9177 |
+- WARN_ON(!list_empty(&dev->devres_head)); |
9178 |
++ if (!list_empty(&dev->devres_head)) { |
9179 |
++ dev_crit(dev, "Resources present before probing\n"); |
9180 |
++ return -EBUSY; |
9181 |
++ } |
9182 |
+ |
9183 |
+ re_probe: |
9184 |
+ dev->driver = drv; |
9185 |
+diff --git a/drivers/base/platform.c b/drivers/base/platform.c |
9186 |
+index e9be1f56929a..d1f901b58f75 100644 |
9187 |
+--- a/drivers/base/platform.c |
9188 |
++++ b/drivers/base/platform.c |
9189 |
+@@ -27,6 +27,7 @@ |
9190 |
+ #include <linux/limits.h> |
9191 |
+ #include <linux/property.h> |
9192 |
+ #include <linux/kmemleak.h> |
9193 |
++#include <linux/types.h> |
9194 |
+ |
9195 |
+ #include "base.h" |
9196 |
+ #include "power/power.h" |
9197 |
+@@ -67,7 +68,7 @@ void __weak arch_setup_pdev_archdata(struct platform_device *pdev) |
9198 |
+ struct resource *platform_get_resource(struct platform_device *dev, |
9199 |
+ unsigned int type, unsigned int num) |
9200 |
+ { |
9201 |
+- int i; |
9202 |
++ u32 i; |
9203 |
+ |
9204 |
+ for (i = 0; i < dev->num_resources; i++) { |
9205 |
+ struct resource *r = &dev->resource[i]; |
9206 |
+@@ -162,7 +163,7 @@ struct resource *platform_get_resource_byname(struct platform_device *dev, |
9207 |
+ unsigned int type, |
9208 |
+ const char *name) |
9209 |
+ { |
9210 |
+- int i; |
9211 |
++ u32 i; |
9212 |
+ |
9213 |
+ for (i = 0; i < dev->num_resources; i++) { |
9214 |
+ struct resource *r = &dev->resource[i]; |
9215 |
+@@ -359,7 +360,8 @@ EXPORT_SYMBOL_GPL(platform_device_add_properties); |
9216 |
+ */ |
9217 |
+ int platform_device_add(struct platform_device *pdev) |
9218 |
+ { |
9219 |
+- int i, ret; |
9220 |
++ u32 i; |
9221 |
++ int ret; |
9222 |
+ |
9223 |
+ if (!pdev) |
9224 |
+ return -EINVAL; |
9225 |
+@@ -425,7 +427,7 @@ int platform_device_add(struct platform_device *pdev) |
9226 |
+ pdev->id = PLATFORM_DEVID_AUTO; |
9227 |
+ } |
9228 |
+ |
9229 |
+- while (--i >= 0) { |
9230 |
++ while (i--) { |
9231 |
+ struct resource *r = &pdev->resource[i]; |
9232 |
+ if (r->parent) |
9233 |
+ release_resource(r); |
9234 |
+@@ -446,7 +448,7 @@ EXPORT_SYMBOL_GPL(platform_device_add); |
9235 |
+ */ |
9236 |
+ void platform_device_del(struct platform_device *pdev) |
9237 |
+ { |
9238 |
+- int i; |
9239 |
++ u32 i; |
9240 |
+ |
9241 |
+ if (pdev) { |
9242 |
+ device_remove_properties(&pdev->dev); |
9243 |
+diff --git a/drivers/block/brd.c b/drivers/block/brd.c |
9244 |
+index 17defbf4f332..02e8fff3f828 100644 |
9245 |
+--- a/drivers/block/brd.c |
9246 |
++++ b/drivers/block/brd.c |
9247 |
+@@ -463,6 +463,25 @@ static struct kobject *brd_probe(dev_t dev, int *part, void *data) |
9248 |
+ return kobj; |
9249 |
+ } |
9250 |
+ |
9251 |
++static inline void brd_check_and_reset_par(void) |
9252 |
++{ |
9253 |
++ if (unlikely(!max_part)) |
9254 |
++ max_part = 1; |
9255 |
++ |
9256 |
++ /* |
9257 |
++ * make sure 'max_part' can be divided exactly by (1U << MINORBITS), |
9258 |
++ * otherwise, it is possiable to get same dev_t when adding partitions. |
9259 |
++ */ |
9260 |
++ if ((1U << MINORBITS) % max_part != 0) |
9261 |
++ max_part = 1UL << fls(max_part); |
9262 |
++ |
9263 |
++ if (max_part > DISK_MAX_PARTS) { |
9264 |
++ pr_info("brd: max_part can't be larger than %d, reset max_part = %d.\n", |
9265 |
++ DISK_MAX_PARTS, DISK_MAX_PARTS); |
9266 |
++ max_part = DISK_MAX_PARTS; |
9267 |
++ } |
9268 |
++} |
9269 |
++ |
9270 |
+ static int __init brd_init(void) |
9271 |
+ { |
9272 |
+ struct brd_device *brd, *next; |
9273 |
+@@ -486,8 +505,7 @@ static int __init brd_init(void) |
9274 |
+ if (register_blkdev(RAMDISK_MAJOR, "ramdisk")) |
9275 |
+ return -EIO; |
9276 |
+ |
9277 |
+- if (unlikely(!max_part)) |
9278 |
+- max_part = 1; |
9279 |
++ brd_check_and_reset_par(); |
9280 |
+ |
9281 |
+ for (i = 0; i < rd_nr; i++) { |
9282 |
+ brd = brd_alloc(i); |
9283 |
+diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c |
9284 |
+index b9d321bdaa8a..226103af30f0 100644 |
9285 |
+--- a/drivers/block/nbd.c |
9286 |
++++ b/drivers/block/nbd.c |
9287 |
+@@ -1216,6 +1216,16 @@ static int nbd_start_device(struct nbd_device *nbd) |
9288 |
+ args = kzalloc(sizeof(*args), GFP_KERNEL); |
9289 |
+ if (!args) { |
9290 |
+ sock_shutdown(nbd); |
9291 |
++ /* |
9292 |
++ * If num_connections is m (2 < m), |
9293 |
++ * and NO.1 ~ NO.n(1 < n < m) kzallocs are successful. |
9294 |
++ * But NO.(n + 1) failed. We still have n recv threads. |
9295 |
++ * So, add flush_workqueue here to prevent recv threads |
9296 |
++ * dropping the last config_refs and trying to destroy |
9297 |
++ * the workqueue from inside the workqueue. |
9298 |
++ */ |
9299 |
++ if (i) |
9300 |
++ flush_workqueue(nbd->recv_workq); |
9301 |
+ return -ENOMEM; |
9302 |
+ } |
9303 |
+ sk_set_memalloc(config->socks[i]->sock->sk); |
9304 |
+diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c |
9305 |
+index b942f4c8cea8..d3ad1b8c133e 100644 |
9306 |
+--- a/drivers/block/rbd.c |
9307 |
++++ b/drivers/block/rbd.c |
9308 |
+@@ -2097,7 +2097,7 @@ static int rbd_img_fill_nodata(struct rbd_img_request *img_req, |
9309 |
+ u64 off, u64 len) |
9310 |
+ { |
9311 |
+ struct ceph_file_extent ex = { off, len }; |
9312 |
+- union rbd_img_fill_iter dummy; |
9313 |
++ union rbd_img_fill_iter dummy = {}; |
9314 |
+ struct rbd_img_fill_ctx fctx = { |
9315 |
+ .pos_type = OBJ_REQUEST_NODATA, |
9316 |
+ .pos = &dummy, |
9317 |
+diff --git a/drivers/char/random.c b/drivers/char/random.c |
9318 |
+index 53e822793d46..28b110cd3977 100644 |
9319 |
+--- a/drivers/char/random.c |
9320 |
++++ b/drivers/char/random.c |
9321 |
+@@ -1609,8 +1609,9 @@ static void _warn_unseeded_randomness(const char *func_name, void *caller, |
9322 |
+ print_once = true; |
9323 |
+ #endif |
9324 |
+ if (__ratelimit(&unseeded_warning)) |
9325 |
+- pr_notice("random: %s called from %pS with crng_init=%d\n", |
9326 |
+- func_name, caller, crng_init); |
9327 |
++ printk_deferred(KERN_NOTICE "random: %s called from %pS " |
9328 |
++ "with crng_init=%d\n", func_name, caller, |
9329 |
++ crng_init); |
9330 |
+ } |
9331 |
+ |
9332 |
+ /* |
9333 |
+diff --git a/drivers/clk/qcom/clk-rcg2.c b/drivers/clk/qcom/clk-rcg2.c |
9334 |
+index 51b2388d80ac..ee693e15d9eb 100644 |
9335 |
+--- a/drivers/clk/qcom/clk-rcg2.c |
9336 |
++++ b/drivers/clk/qcom/clk-rcg2.c |
9337 |
+@@ -203,6 +203,9 @@ static int _freq_tbl_determine_rate(struct clk_hw *hw, const struct freq_tbl *f, |
9338 |
+ |
9339 |
+ clk_flags = clk_hw_get_flags(hw); |
9340 |
+ p = clk_hw_get_parent_by_index(hw, index); |
9341 |
++ if (!p) |
9342 |
++ return -EINVAL; |
9343 |
++ |
9344 |
+ if (clk_flags & CLK_SET_RATE_PARENT) { |
9345 |
+ rate = f->freq; |
9346 |
+ if (f->pre_div) { |
9347 |
+diff --git a/drivers/clk/sunxi-ng/ccu-sun50i-a64.c b/drivers/clk/sunxi-ng/ccu-sun50i-a64.c |
9348 |
+index dec4a130390a..9ac6c299e074 100644 |
9349 |
+--- a/drivers/clk/sunxi-ng/ccu-sun50i-a64.c |
9350 |
++++ b/drivers/clk/sunxi-ng/ccu-sun50i-a64.c |
9351 |
+@@ -901,11 +901,26 @@ static const struct sunxi_ccu_desc sun50i_a64_ccu_desc = { |
9352 |
+ .num_resets = ARRAY_SIZE(sun50i_a64_ccu_resets), |
9353 |
+ }; |
9354 |
+ |
9355 |
++static struct ccu_pll_nb sun50i_a64_pll_cpu_nb = { |
9356 |
++ .common = &pll_cpux_clk.common, |
9357 |
++ /* copy from pll_cpux_clk */ |
9358 |
++ .enable = BIT(31), |
9359 |
++ .lock = BIT(28), |
9360 |
++}; |
9361 |
++ |
9362 |
++static struct ccu_mux_nb sun50i_a64_cpu_nb = { |
9363 |
++ .common = &cpux_clk.common, |
9364 |
++ .cm = &cpux_clk.mux, |
9365 |
++ .delay_us = 1, /* > 8 clock cycles at 24 MHz */ |
9366 |
++ .bypass_index = 1, /* index of 24 MHz oscillator */ |
9367 |
++}; |
9368 |
++ |
9369 |
+ static int sun50i_a64_ccu_probe(struct platform_device *pdev) |
9370 |
+ { |
9371 |
+ struct resource *res; |
9372 |
+ void __iomem *reg; |
9373 |
+ u32 val; |
9374 |
++ int ret; |
9375 |
+ |
9376 |
+ res = platform_get_resource(pdev, IORESOURCE_MEM, 0); |
9377 |
+ reg = devm_ioremap_resource(&pdev->dev, res); |
9378 |
+@@ -919,7 +934,18 @@ static int sun50i_a64_ccu_probe(struct platform_device *pdev) |
9379 |
+ |
9380 |
+ writel(0x515, reg + SUN50I_A64_PLL_MIPI_REG); |
9381 |
+ |
9382 |
+- return sunxi_ccu_probe(pdev->dev.of_node, reg, &sun50i_a64_ccu_desc); |
9383 |
++ ret = sunxi_ccu_probe(pdev->dev.of_node, reg, &sun50i_a64_ccu_desc); |
9384 |
++ if (ret) |
9385 |
++ return ret; |
9386 |
++ |
9387 |
++ /* Gate then ungate PLL CPU after any rate changes */ |
9388 |
++ ccu_pll_notifier_register(&sun50i_a64_pll_cpu_nb); |
9389 |
++ |
9390 |
++ /* Reparent CPU during PLL CPU rate changes */ |
9391 |
++ ccu_mux_notifier_register(pll_cpux_clk.common.hw.clk, |
9392 |
++ &sun50i_a64_cpu_nb); |
9393 |
++ |
9394 |
++ return 0; |
9395 |
+ } |
9396 |
+ |
9397 |
+ static const struct of_device_id sun50i_a64_ccu_ids[] = { |
9398 |
+diff --git a/drivers/clk/uniphier/clk-uniphier-peri.c b/drivers/clk/uniphier/clk-uniphier-peri.c |
9399 |
+index 89b3ac378b3f..8b75dc116a98 100644 |
9400 |
+--- a/drivers/clk/uniphier/clk-uniphier-peri.c |
9401 |
++++ b/drivers/clk/uniphier/clk-uniphier-peri.c |
9402 |
+@@ -27,8 +27,8 @@ |
9403 |
+ #define UNIPHIER_PERI_CLK_FI2C(idx, ch) \ |
9404 |
+ UNIPHIER_CLK_GATE("i2c" #ch, (idx), "i2c", 0x24, 24 + (ch)) |
9405 |
+ |
9406 |
+-#define UNIPHIER_PERI_CLK_SCSSI(idx) \ |
9407 |
+- UNIPHIER_CLK_GATE("scssi", (idx), "spi", 0x20, 17) |
9408 |
++#define UNIPHIER_PERI_CLK_SCSSI(idx, ch) \ |
9409 |
++ UNIPHIER_CLK_GATE("scssi" #ch, (idx), "spi", 0x20, 17 + (ch)) |
9410 |
+ |
9411 |
+ #define UNIPHIER_PERI_CLK_MCSSI(idx) \ |
9412 |
+ UNIPHIER_CLK_GATE("mcssi", (idx), "spi", 0x24, 14) |
9413 |
+@@ -44,7 +44,7 @@ const struct uniphier_clk_data uniphier_ld4_peri_clk_data[] = { |
9414 |
+ UNIPHIER_PERI_CLK_I2C(6, 2), |
9415 |
+ UNIPHIER_PERI_CLK_I2C(7, 3), |
9416 |
+ UNIPHIER_PERI_CLK_I2C(8, 4), |
9417 |
+- UNIPHIER_PERI_CLK_SCSSI(11), |
9418 |
++ UNIPHIER_PERI_CLK_SCSSI(11, 0), |
9419 |
+ { /* sentinel */ } |
9420 |
+ }; |
9421 |
+ |
9422 |
+@@ -60,7 +60,10 @@ const struct uniphier_clk_data uniphier_pro4_peri_clk_data[] = { |
9423 |
+ UNIPHIER_PERI_CLK_FI2C(8, 4), |
9424 |
+ UNIPHIER_PERI_CLK_FI2C(9, 5), |
9425 |
+ UNIPHIER_PERI_CLK_FI2C(10, 6), |
9426 |
+- UNIPHIER_PERI_CLK_SCSSI(11), |
9427 |
+- UNIPHIER_PERI_CLK_MCSSI(12), |
9428 |
++ UNIPHIER_PERI_CLK_SCSSI(11, 0), |
9429 |
++ UNIPHIER_PERI_CLK_SCSSI(12, 1), |
9430 |
++ UNIPHIER_PERI_CLK_SCSSI(13, 2), |
9431 |
++ UNIPHIER_PERI_CLK_SCSSI(14, 3), |
9432 |
++ UNIPHIER_PERI_CLK_MCSSI(15), |
9433 |
+ { /* sentinel */ } |
9434 |
+ }; |
9435 |
+diff --git a/drivers/clocksource/bcm2835_timer.c b/drivers/clocksource/bcm2835_timer.c |
9436 |
+index 60da2537bef9..1082dcef17d1 100644 |
9437 |
+--- a/drivers/clocksource/bcm2835_timer.c |
9438 |
++++ b/drivers/clocksource/bcm2835_timer.c |
9439 |
+@@ -134,7 +134,7 @@ static int __init bcm2835_timer_init(struct device_node *node) |
9440 |
+ ret = setup_irq(irq, &timer->act); |
9441 |
+ if (ret) { |
9442 |
+ pr_err("Can't set up timer IRQ\n"); |
9443 |
+- goto err_iounmap; |
9444 |
++ goto err_timer_free; |
9445 |
+ } |
9446 |
+ |
9447 |
+ clockevents_config_and_register(&timer->evt, freq, 0xf, 0xffffffff); |
9448 |
+@@ -143,6 +143,9 @@ static int __init bcm2835_timer_init(struct device_node *node) |
9449 |
+ |
9450 |
+ return 0; |
9451 |
+ |
9452 |
++err_timer_free: |
9453 |
++ kfree(timer); |
9454 |
++ |
9455 |
+ err_iounmap: |
9456 |
+ iounmap(base); |
9457 |
+ return ret; |
9458 |
+diff --git a/drivers/crypto/chelsio/chtls/chtls_cm.c b/drivers/crypto/chelsio/chtls/chtls_cm.c |
9459 |
+index 8b749c721c87..28d24118c645 100644 |
9460 |
+--- a/drivers/crypto/chelsio/chtls/chtls_cm.c |
9461 |
++++ b/drivers/crypto/chelsio/chtls/chtls_cm.c |
9462 |
+@@ -731,6 +731,14 @@ static int chtls_close_listsrv_rpl(struct chtls_dev *cdev, struct sk_buff *skb) |
9463 |
+ return 0; |
9464 |
+ } |
9465 |
+ |
9466 |
++static void chtls_purge_wr_queue(struct sock *sk) |
9467 |
++{ |
9468 |
++ struct sk_buff *skb; |
9469 |
++ |
9470 |
++ while ((skb = dequeue_wr(sk)) != NULL) |
9471 |
++ kfree_skb(skb); |
9472 |
++} |
9473 |
++ |
9474 |
+ static void chtls_release_resources(struct sock *sk) |
9475 |
+ { |
9476 |
+ struct chtls_sock *csk = rcu_dereference_sk_user_data(sk); |
9477 |
+@@ -745,6 +753,11 @@ static void chtls_release_resources(struct sock *sk) |
9478 |
+ kfree_skb(csk->txdata_skb_cache); |
9479 |
+ csk->txdata_skb_cache = NULL; |
9480 |
+ |
9481 |
++ if (csk->wr_credits != csk->wr_max_credits) { |
9482 |
++ chtls_purge_wr_queue(sk); |
9483 |
++ chtls_reset_wr_list(csk); |
9484 |
++ } |
9485 |
++ |
9486 |
+ if (csk->l2t_entry) { |
9487 |
+ cxgb4_l2t_release(csk->l2t_entry); |
9488 |
+ csk->l2t_entry = NULL; |
9489 |
+@@ -1714,6 +1727,7 @@ static void chtls_peer_close(struct sock *sk, struct sk_buff *skb) |
9490 |
+ else |
9491 |
+ sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN); |
9492 |
+ } |
9493 |
++ kfree_skb(skb); |
9494 |
+ } |
9495 |
+ |
9496 |
+ static void chtls_close_con_rpl(struct sock *sk, struct sk_buff *skb) |
9497 |
+@@ -2041,19 +2055,6 @@ rel_skb: |
9498 |
+ return 0; |
9499 |
+ } |
9500 |
+ |
9501 |
+-static struct sk_buff *dequeue_wr(struct sock *sk) |
9502 |
+-{ |
9503 |
+- struct chtls_sock *csk = rcu_dereference_sk_user_data(sk); |
9504 |
+- struct sk_buff *skb = csk->wr_skb_head; |
9505 |
+- |
9506 |
+- if (likely(skb)) { |
9507 |
+- /* Don't bother clearing the tail */ |
9508 |
+- csk->wr_skb_head = WR_SKB_CB(skb)->next_wr; |
9509 |
+- WR_SKB_CB(skb)->next_wr = NULL; |
9510 |
+- } |
9511 |
+- return skb; |
9512 |
+-} |
9513 |
+- |
9514 |
+ static void chtls_rx_ack(struct sock *sk, struct sk_buff *skb) |
9515 |
+ { |
9516 |
+ struct cpl_fw4_ack *hdr = cplhdr(skb) + RSS_HDR; |
9517 |
+diff --git a/drivers/crypto/chelsio/chtls/chtls_cm.h b/drivers/crypto/chelsio/chtls/chtls_cm.h |
9518 |
+index 78eb3afa3a80..4282d8a4eae4 100644 |
9519 |
+--- a/drivers/crypto/chelsio/chtls/chtls_cm.h |
9520 |
++++ b/drivers/crypto/chelsio/chtls/chtls_cm.h |
9521 |
+@@ -188,6 +188,12 @@ static inline void chtls_kfree_skb(struct sock *sk, struct sk_buff *skb) |
9522 |
+ kfree_skb(skb); |
9523 |
+ } |
9524 |
+ |
9525 |
++static inline void chtls_reset_wr_list(struct chtls_sock *csk) |
9526 |
++{ |
9527 |
++ csk->wr_skb_head = NULL; |
9528 |
++ csk->wr_skb_tail = NULL; |
9529 |
++} |
9530 |
++ |
9531 |
+ static inline void enqueue_wr(struct chtls_sock *csk, struct sk_buff *skb) |
9532 |
+ { |
9533 |
+ WR_SKB_CB(skb)->next_wr = NULL; |
9534 |
+@@ -200,4 +206,19 @@ static inline void enqueue_wr(struct chtls_sock *csk, struct sk_buff *skb) |
9535 |
+ WR_SKB_CB(csk->wr_skb_tail)->next_wr = skb; |
9536 |
+ csk->wr_skb_tail = skb; |
9537 |
+ } |
9538 |
++ |
9539 |
++static inline struct sk_buff *dequeue_wr(struct sock *sk) |
9540 |
++{ |
9541 |
++ struct chtls_sock *csk = rcu_dereference_sk_user_data(sk); |
9542 |
++ struct sk_buff *skb = NULL; |
9543 |
++ |
9544 |
++ skb = csk->wr_skb_head; |
9545 |
++ |
9546 |
++ if (likely(skb)) { |
9547 |
++ /* Don't bother clearing the tail */ |
9548 |
++ csk->wr_skb_head = WR_SKB_CB(skb)->next_wr; |
9549 |
++ WR_SKB_CB(skb)->next_wr = NULL; |
9550 |
++ } |
9551 |
++ return skb; |
9552 |
++} |
9553 |
+ #endif |
9554 |
+diff --git a/drivers/crypto/chelsio/chtls/chtls_hw.c b/drivers/crypto/chelsio/chtls/chtls_hw.c |
9555 |
+index 490960755864..64d24823c65a 100644 |
9556 |
+--- a/drivers/crypto/chelsio/chtls/chtls_hw.c |
9557 |
++++ b/drivers/crypto/chelsio/chtls/chtls_hw.c |
9558 |
+@@ -361,6 +361,7 @@ int chtls_setkey(struct chtls_sock *csk, u32 keylen, u32 optname) |
9559 |
+ kwr->sc_imm.cmd_more = cpu_to_be32(ULPTX_CMD_V(ULP_TX_SC_IMM)); |
9560 |
+ kwr->sc_imm.len = cpu_to_be32(klen); |
9561 |
+ |
9562 |
++ lock_sock(sk); |
9563 |
+ /* key info */ |
9564 |
+ kctx = (struct _key_ctx *)(kwr + 1); |
9565 |
+ ret = chtls_key_info(csk, kctx, keylen, optname); |
9566 |
+@@ -399,8 +400,10 @@ int chtls_setkey(struct chtls_sock *csk, u32 keylen, u32 optname) |
9567 |
+ csk->tlshws.txkey = keyid; |
9568 |
+ } |
9569 |
+ |
9570 |
++ release_sock(sk); |
9571 |
+ return ret; |
9572 |
+ out_notcb: |
9573 |
++ release_sock(sk); |
9574 |
+ free_tls_keyid(sk); |
9575 |
+ out_nokey: |
9576 |
+ kfree_skb(skb); |
9577 |
+diff --git a/drivers/devfreq/Kconfig b/drivers/devfreq/Kconfig |
9578 |
+index 6a172d338f6d..4c4ec68b0566 100644 |
9579 |
+--- a/drivers/devfreq/Kconfig |
9580 |
++++ b/drivers/devfreq/Kconfig |
9581 |
+@@ -103,7 +103,8 @@ config ARM_TEGRA_DEVFREQ |
9582 |
+ |
9583 |
+ config ARM_RK3399_DMC_DEVFREQ |
9584 |
+ tristate "ARM RK3399 DMC DEVFREQ Driver" |
9585 |
+- depends on ARCH_ROCKCHIP |
9586 |
++ depends on (ARCH_ROCKCHIP && HAVE_ARM_SMCCC) || \ |
9587 |
++ (COMPILE_TEST && HAVE_ARM_SMCCC) |
9588 |
+ select DEVFREQ_EVENT_ROCKCHIP_DFI |
9589 |
+ select DEVFREQ_GOV_SIMPLE_ONDEMAND |
9590 |
+ select PM_DEVFREQ_EVENT |
9591 |
+diff --git a/drivers/devfreq/event/Kconfig b/drivers/devfreq/event/Kconfig |
9592 |
+index cd949800eed9..8851bc4e8e3e 100644 |
9593 |
+--- a/drivers/devfreq/event/Kconfig |
9594 |
++++ b/drivers/devfreq/event/Kconfig |
9595 |
+@@ -33,7 +33,7 @@ config DEVFREQ_EVENT_EXYNOS_PPMU |
9596 |
+ |
9597 |
+ config DEVFREQ_EVENT_ROCKCHIP_DFI |
9598 |
+ tristate "ROCKCHIP DFI DEVFREQ event Driver" |
9599 |
+- depends on ARCH_ROCKCHIP |
9600 |
++ depends on ARCH_ROCKCHIP || COMPILE_TEST |
9601 |
+ help |
9602 |
+ This add the devfreq-event driver for Rockchip SoC. It provides DFI |
9603 |
+ (DDR Monitor Module) driver to count ddr load. |
9604 |
+diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c |
9605 |
+index f1a441ab395d..8a52a5efee4f 100644 |
9606 |
+--- a/drivers/dma/dmaengine.c |
9607 |
++++ b/drivers/dma/dmaengine.c |
9608 |
+@@ -190,7 +190,7 @@ __dma_device_satisfies_mask(struct dma_device *device, |
9609 |
+ |
9610 |
+ static struct module *dma_chan_to_owner(struct dma_chan *chan) |
9611 |
+ { |
9612 |
+- return chan->device->dev->driver->owner; |
9613 |
++ return chan->device->owner; |
9614 |
+ } |
9615 |
+ |
9616 |
+ /** |
9617 |
+@@ -923,6 +923,8 @@ int dma_async_device_register(struct dma_device *device) |
9618 |
+ return -EIO; |
9619 |
+ } |
9620 |
+ |
9621 |
++ device->owner = device->dev->driver->owner; |
9622 |
++ |
9623 |
+ if (dma_has_cap(DMA_MEMCPY, device->cap_mask) && !device->device_prep_dma_memcpy) { |
9624 |
+ dev_err(device->dev, |
9625 |
+ "Device claims capability %s, but op is not defined\n", |
9626 |
+diff --git a/drivers/dma/imx-sdma.c b/drivers/dma/imx-sdma.c |
9627 |
+index ceb82e74f5b4..d66a7fdff898 100644 |
9628 |
+--- a/drivers/dma/imx-sdma.c |
9629 |
++++ b/drivers/dma/imx-sdma.c |
9630 |
+@@ -738,12 +738,8 @@ static void sdma_start_desc(struct sdma_channel *sdmac) |
9631 |
+ return; |
9632 |
+ } |
9633 |
+ sdmac->desc = desc = to_sdma_desc(&vd->tx); |
9634 |
+- /* |
9635 |
+- * Do not delete the node in desc_issued list in cyclic mode, otherwise |
9636 |
+- * the desc allocated will never be freed in vchan_dma_desc_free_list |
9637 |
+- */ |
9638 |
+- if (!(sdmac->flags & IMX_DMA_SG_LOOP)) |
9639 |
+- list_del(&vd->node); |
9640 |
++ |
9641 |
++ list_del(&vd->node); |
9642 |
+ |
9643 |
+ sdma->channel_control[channel].base_bd_ptr = desc->bd_phys; |
9644 |
+ sdma->channel_control[channel].current_bd_ptr = desc->bd_phys; |
9645 |
+@@ -1044,7 +1040,6 @@ static void sdma_channel_terminate_work(struct work_struct *work) |
9646 |
+ |
9647 |
+ spin_lock_irqsave(&sdmac->vc.lock, flags); |
9648 |
+ vchan_get_all_descriptors(&sdmac->vc, &head); |
9649 |
+- sdmac->desc = NULL; |
9650 |
+ spin_unlock_irqrestore(&sdmac->vc.lock, flags); |
9651 |
+ vchan_dma_desc_free_list(&sdmac->vc, &head); |
9652 |
+ } |
9653 |
+@@ -1052,11 +1047,19 @@ static void sdma_channel_terminate_work(struct work_struct *work) |
9654 |
+ static int sdma_disable_channel_async(struct dma_chan *chan) |
9655 |
+ { |
9656 |
+ struct sdma_channel *sdmac = to_sdma_chan(chan); |
9657 |
++ unsigned long flags; |
9658 |
++ |
9659 |
++ spin_lock_irqsave(&sdmac->vc.lock, flags); |
9660 |
+ |
9661 |
+ sdma_disable_channel(chan); |
9662 |
+ |
9663 |
+- if (sdmac->desc) |
9664 |
++ if (sdmac->desc) { |
9665 |
++ vchan_terminate_vdesc(&sdmac->desc->vd); |
9666 |
++ sdmac->desc = NULL; |
9667 |
+ schedule_work(&sdmac->terminate_worker); |
9668 |
++ } |
9669 |
++ |
9670 |
++ spin_unlock_irqrestore(&sdmac->vc.lock, flags); |
9671 |
+ |
9672 |
+ return 0; |
9673 |
+ } |
9674 |
+diff --git a/drivers/gpio/gpio-grgpio.c b/drivers/gpio/gpio-grgpio.c |
9675 |
+index 60a1556c570a..c1be299e5567 100644 |
9676 |
+--- a/drivers/gpio/gpio-grgpio.c |
9677 |
++++ b/drivers/gpio/gpio-grgpio.c |
9678 |
+@@ -258,17 +258,16 @@ static int grgpio_irq_map(struct irq_domain *d, unsigned int irq, |
9679 |
+ lirq->irq = irq; |
9680 |
+ uirq = &priv->uirqs[lirq->index]; |
9681 |
+ if (uirq->refcnt == 0) { |
9682 |
++ spin_unlock_irqrestore(&priv->gc.bgpio_lock, flags); |
9683 |
+ ret = request_irq(uirq->uirq, grgpio_irq_handler, 0, |
9684 |
+ dev_name(priv->dev), priv); |
9685 |
+ if (ret) { |
9686 |
+ dev_err(priv->dev, |
9687 |
+ "Could not request underlying irq %d\n", |
9688 |
+ uirq->uirq); |
9689 |
+- |
9690 |
+- spin_unlock_irqrestore(&priv->gc.bgpio_lock, flags); |
9691 |
+- |
9692 |
+ return ret; |
9693 |
+ } |
9694 |
++ spin_lock_irqsave(&priv->gc.bgpio_lock, flags); |
9695 |
+ } |
9696 |
+ uirq->refcnt++; |
9697 |
+ |
9698 |
+@@ -314,8 +313,11 @@ static void grgpio_irq_unmap(struct irq_domain *d, unsigned int irq) |
9699 |
+ if (index >= 0) { |
9700 |
+ uirq = &priv->uirqs[lirq->index]; |
9701 |
+ uirq->refcnt--; |
9702 |
+- if (uirq->refcnt == 0) |
9703 |
++ if (uirq->refcnt == 0) { |
9704 |
++ spin_unlock_irqrestore(&priv->gc.bgpio_lock, flags); |
9705 |
+ free_irq(uirq->uirq, priv); |
9706 |
++ return; |
9707 |
++ } |
9708 |
+ } |
9709 |
+ |
9710 |
+ spin_unlock_irqrestore(&priv->gc.bgpio_lock, flags); |
9711 |
+diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c |
9712 |
+index bf872f694f50..d1fbaea91f58 100644 |
9713 |
+--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c |
9714 |
++++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c |
9715 |
+@@ -337,17 +337,9 @@ bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device * |
9716 |
+ path_size += le16_to_cpu(path->usSize); |
9717 |
+ |
9718 |
+ if (device_support & le16_to_cpu(path->usDeviceTag)) { |
9719 |
+- uint8_t con_obj_id, con_obj_num, con_obj_type; |
9720 |
+- |
9721 |
+- con_obj_id = |
9722 |
++ uint8_t con_obj_id = |
9723 |
+ (le16_to_cpu(path->usConnObjectId) & OBJECT_ID_MASK) |
9724 |
+ >> OBJECT_ID_SHIFT; |
9725 |
+- con_obj_num = |
9726 |
+- (le16_to_cpu(path->usConnObjectId) & ENUM_ID_MASK) |
9727 |
+- >> ENUM_ID_SHIFT; |
9728 |
+- con_obj_type = |
9729 |
+- (le16_to_cpu(path->usConnObjectId) & |
9730 |
+- OBJECT_TYPE_MASK) >> OBJECT_TYPE_SHIFT; |
9731 |
+ |
9732 |
+ /* Skip TV/CV support */ |
9733 |
+ if ((le16_to_cpu(path->usDeviceTag) == |
9734 |
+@@ -372,14 +364,7 @@ bool amdgpu_atombios_get_connector_info_from_object_table(struct amdgpu_device * |
9735 |
+ router.ddc_valid = false; |
9736 |
+ router.cd_valid = false; |
9737 |
+ for (j = 0; j < ((le16_to_cpu(path->usSize) - 8) / 2); j++) { |
9738 |
+- uint8_t grph_obj_id, grph_obj_num, grph_obj_type; |
9739 |
+- |
9740 |
+- grph_obj_id = |
9741 |
+- (le16_to_cpu(path->usGraphicObjIds[j]) & |
9742 |
+- OBJECT_ID_MASK) >> OBJECT_ID_SHIFT; |
9743 |
+- grph_obj_num = |
9744 |
+- (le16_to_cpu(path->usGraphicObjIds[j]) & |
9745 |
+- ENUM_ID_MASK) >> ENUM_ID_SHIFT; |
9746 |
++ uint8_t grph_obj_type= |
9747 |
+ grph_obj_type = |
9748 |
+ (le16_to_cpu(path->usGraphicObjIds[j]) & |
9749 |
+ OBJECT_TYPE_MASK) >> OBJECT_TYPE_SHIFT; |
9750 |
+diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h |
9751 |
+index 0942f492d2e1..9d444f7fdfcd 100644 |
9752 |
+--- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h |
9753 |
++++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h |
9754 |
+@@ -51,6 +51,7 @@ |
9755 |
+ do { \ |
9756 |
+ uint32_t tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ |
9757 |
+ uint32_t loop = adev->usec_timeout; \ |
9758 |
++ ret = 0; \ |
9759 |
+ while ((tmp_ & (mask)) != (expected_value)) { \ |
9760 |
+ udelay(2); \ |
9761 |
+ tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ |
9762 |
+diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_math.h b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_math.h |
9763 |
+deleted file mode 100644 |
9764 |
+index f46ab0e24ca1..000000000000 |
9765 |
+--- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calc_math.h |
9766 |
++++ /dev/null |
9767 |
+@@ -1,40 +0,0 @@ |
9768 |
+-/* |
9769 |
+- * Copyright 2017 Advanced Micro Devices, Inc. |
9770 |
+- * |
9771 |
+- * Permission is hereby granted, free of charge, to any person obtaining a |
9772 |
+- * copy of this software and associated documentation files (the "Software"), |
9773 |
+- * to deal in the Software without restriction, including without limitation |
9774 |
+- * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
9775 |
+- * and/or sell copies of the Software, and to permit persons to whom the |
9776 |
+- * Software is furnished to do so, subject to the following conditions: |
9777 |
+- * |
9778 |
+- * The above copyright notice and this permission notice shall be included in |
9779 |
+- * all copies or substantial portions of the Software. |
9780 |
+- * |
9781 |
+- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
9782 |
+- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
9783 |
+- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
9784 |
+- * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
9785 |
+- * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
9786 |
+- * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
9787 |
+- * OTHER DEALINGS IN THE SOFTWARE. |
9788 |
+- * |
9789 |
+- * Authors: AMD |
9790 |
+- * |
9791 |
+- */ |
9792 |
+- |
9793 |
+-#ifndef _DCN_CALC_MATH_H_ |
9794 |
+-#define _DCN_CALC_MATH_H_ |
9795 |
+- |
9796 |
+-float dcn_bw_mod(const float arg1, const float arg2); |
9797 |
+-float dcn_bw_min2(const float arg1, const float arg2); |
9798 |
+-unsigned int dcn_bw_max(const unsigned int arg1, const unsigned int arg2); |
9799 |
+-float dcn_bw_max2(const float arg1, const float arg2); |
9800 |
+-float dcn_bw_floor2(const float arg, const float significance); |
9801 |
+-float dcn_bw_ceil2(const float arg, const float significance); |
9802 |
+-float dcn_bw_max3(float v1, float v2, float v3); |
9803 |
+-float dcn_bw_max5(float v1, float v2, float v3, float v4, float v5); |
9804 |
+-float dcn_bw_pow(float a, float exp); |
9805 |
+-float dcn_bw_log(float a, float b); |
9806 |
+- |
9807 |
+-#endif /* _DCN_CALC_MATH_H_ */ |
9808 |
+diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c |
9809 |
+index 6342f6499351..b0956c360393 100644 |
9810 |
+--- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c |
9811 |
++++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c |
9812 |
+@@ -1346,6 +1346,7 @@ void dcn_bw_update_from_pplib(struct dc *dc) |
9813 |
+ struct dc_context *ctx = dc->ctx; |
9814 |
+ struct dm_pp_clock_levels_with_voltage fclks = {0}, dcfclks = {0}; |
9815 |
+ bool res; |
9816 |
++ unsigned vmin0p65_idx, vmid0p72_idx, vnom0p8_idx, vmax0p9_idx; |
9817 |
+ |
9818 |
+ /* TODO: This is not the proper way to obtain fabric_and_dram_bandwidth, should be min(fclk, memclk) */ |
9819 |
+ res = dm_pp_get_clock_levels_by_type_with_voltage( |
9820 |
+@@ -1357,17 +1358,28 @@ void dcn_bw_update_from_pplib(struct dc *dc) |
9821 |
+ res = verify_clock_values(&fclks); |
9822 |
+ |
9823 |
+ if (res) { |
9824 |
+- ASSERT(fclks.num_levels >= 3); |
9825 |
+- dc->dcn_soc->fabric_and_dram_bandwidth_vmin0p65 = 32 * (fclks.data[0].clocks_in_khz / 1000.0) / 1000.0; |
9826 |
+- dc->dcn_soc->fabric_and_dram_bandwidth_vmid0p72 = dc->dcn_soc->number_of_channels * |
9827 |
+- (fclks.data[fclks.num_levels - (fclks.num_levels > 2 ? 3 : 2)].clocks_in_khz / 1000.0) |
9828 |
+- * ddr4_dram_factor_single_Channel / 1000.0; |
9829 |
+- dc->dcn_soc->fabric_and_dram_bandwidth_vnom0p8 = dc->dcn_soc->number_of_channels * |
9830 |
+- (fclks.data[fclks.num_levels - 2].clocks_in_khz / 1000.0) |
9831 |
+- * ddr4_dram_factor_single_Channel / 1000.0; |
9832 |
+- dc->dcn_soc->fabric_and_dram_bandwidth_vmax0p9 = dc->dcn_soc->number_of_channels * |
9833 |
+- (fclks.data[fclks.num_levels - 1].clocks_in_khz / 1000.0) |
9834 |
+- * ddr4_dram_factor_single_Channel / 1000.0; |
9835 |
++ ASSERT(fclks.num_levels); |
9836 |
++ |
9837 |
++ vmin0p65_idx = 0; |
9838 |
++ vmid0p72_idx = fclks.num_levels - |
9839 |
++ (fclks.num_levels > 2 ? 3 : (fclks.num_levels > 1 ? 2 : 1)); |
9840 |
++ vnom0p8_idx = fclks.num_levels - (fclks.num_levels > 1 ? 2 : 1); |
9841 |
++ vmax0p9_idx = fclks.num_levels - 1; |
9842 |
++ |
9843 |
++ dc->dcn_soc->fabric_and_dram_bandwidth_vmin0p65 = |
9844 |
++ 32 * (fclks.data[vmin0p65_idx].clocks_in_khz / 1000.0) / 1000.0; |
9845 |
++ dc->dcn_soc->fabric_and_dram_bandwidth_vmid0p72 = |
9846 |
++ dc->dcn_soc->number_of_channels * |
9847 |
++ (fclks.data[vmid0p72_idx].clocks_in_khz / 1000.0) |
9848 |
++ * ddr4_dram_factor_single_Channel / 1000.0; |
9849 |
++ dc->dcn_soc->fabric_and_dram_bandwidth_vnom0p8 = |
9850 |
++ dc->dcn_soc->number_of_channels * |
9851 |
++ (fclks.data[vnom0p8_idx].clocks_in_khz / 1000.0) |
9852 |
++ * ddr4_dram_factor_single_Channel / 1000.0; |
9853 |
++ dc->dcn_soc->fabric_and_dram_bandwidth_vmax0p9 = |
9854 |
++ dc->dcn_soc->number_of_channels * |
9855 |
++ (fclks.data[vmax0p9_idx].clocks_in_khz / 1000.0) |
9856 |
++ * ddr4_dram_factor_single_Channel / 1000.0; |
9857 |
+ } else |
9858 |
+ BREAK_TO_DEBUGGER(); |
9859 |
+ |
9860 |
+diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link.c b/drivers/gpu/drm/amd/display/dc/core/dc_link.c |
9861 |
+index 2f42964fb9f4..3abc0294c05f 100644 |
9862 |
+--- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c |
9863 |
++++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c |
9864 |
+@@ -780,8 +780,7 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) |
9865 |
+ same_edid = is_same_edid(&prev_sink->dc_edid, &sink->dc_edid); |
9866 |
+ |
9867 |
+ if (link->connector_signal == SIGNAL_TYPE_DISPLAY_PORT && |
9868 |
+- sink_caps.transaction_type == DDC_TRANSACTION_TYPE_I2C_OVER_AUX && |
9869 |
+- reason != DETECT_REASON_HPDRX) { |
9870 |
++ sink_caps.transaction_type == DDC_TRANSACTION_TYPE_I2C_OVER_AUX) { |
9871 |
+ /* |
9872 |
+ * TODO debug why Dell 2413 doesn't like |
9873 |
+ * two link trainings |
9874 |
+diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml_common_defs.c b/drivers/gpu/drm/amd/display/dc/dml/dml_common_defs.c |
9875 |
+index b953b02a1512..723af0b2dda0 100644 |
9876 |
+--- a/drivers/gpu/drm/amd/display/dc/dml/dml_common_defs.c |
9877 |
++++ b/drivers/gpu/drm/amd/display/dc/dml/dml_common_defs.c |
9878 |
+@@ -24,7 +24,7 @@ |
9879 |
+ */ |
9880 |
+ |
9881 |
+ #include "dml_common_defs.h" |
9882 |
+-#include "../calcs/dcn_calc_math.h" |
9883 |
++#include "dcn_calc_math.h" |
9884 |
+ |
9885 |
+ #include "dml_inline_defs.h" |
9886 |
+ |
9887 |
+diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h b/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h |
9888 |
+index e8ce08567cd8..e4f595a3038c 100644 |
9889 |
+--- a/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h |
9890 |
++++ b/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h |
9891 |
+@@ -27,7 +27,7 @@ |
9892 |
+ #define __DML_INLINE_DEFS_H__ |
9893 |
+ |
9894 |
+ #include "dml_common_defs.h" |
9895 |
+-#include "../calcs/dcn_calc_math.h" |
9896 |
++#include "dcn_calc_math.h" |
9897 |
+ #include "dml_logger.h" |
9898 |
+ |
9899 |
+ static inline double dml_min(double a, double b) |
9900 |
+diff --git a/drivers/gpu/drm/amd/display/dc/inc/dcn_calc_math.h b/drivers/gpu/drm/amd/display/dc/inc/dcn_calc_math.h |
9901 |
+new file mode 100644 |
9902 |
+index 000000000000..f46ab0e24ca1 |
9903 |
+--- /dev/null |
9904 |
++++ b/drivers/gpu/drm/amd/display/dc/inc/dcn_calc_math.h |
9905 |
+@@ -0,0 +1,40 @@ |
9906 |
++/* |
9907 |
++ * Copyright 2017 Advanced Micro Devices, Inc. |
9908 |
++ * |
9909 |
++ * Permission is hereby granted, free of charge, to any person obtaining a |
9910 |
++ * copy of this software and associated documentation files (the "Software"), |
9911 |
++ * to deal in the Software without restriction, including without limitation |
9912 |
++ * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
9913 |
++ * and/or sell copies of the Software, and to permit persons to whom the |
9914 |
++ * Software is furnished to do so, subject to the following conditions: |
9915 |
++ * |
9916 |
++ * The above copyright notice and this permission notice shall be included in |
9917 |
++ * all copies or substantial portions of the Software. |
9918 |
++ * |
9919 |
++ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
9920 |
++ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
9921 |
++ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
9922 |
++ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR |
9923 |
++ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, |
9924 |
++ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR |
9925 |
++ * OTHER DEALINGS IN THE SOFTWARE. |
9926 |
++ * |
9927 |
++ * Authors: AMD |
9928 |
++ * |
9929 |
++ */ |
9930 |
++ |
9931 |
++#ifndef _DCN_CALC_MATH_H_ |
9932 |
++#define _DCN_CALC_MATH_H_ |
9933 |
++ |
9934 |
++float dcn_bw_mod(const float arg1, const float arg2); |
9935 |
++float dcn_bw_min2(const float arg1, const float arg2); |
9936 |
++unsigned int dcn_bw_max(const unsigned int arg1, const unsigned int arg2); |
9937 |
++float dcn_bw_max2(const float arg1, const float arg2); |
9938 |
++float dcn_bw_floor2(const float arg, const float significance); |
9939 |
++float dcn_bw_ceil2(const float arg, const float significance); |
9940 |
++float dcn_bw_max3(float v1, float v2, float v3); |
9941 |
++float dcn_bw_max5(float v1, float v2, float v3, float v4, float v5); |
9942 |
++float dcn_bw_pow(float a, float exp); |
9943 |
++float dcn_bw_log(float a, float b); |
9944 |
++ |
9945 |
++#endif /* _DCN_CALC_MATH_H_ */ |
9946 |
+diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c |
9947 |
+index 1546bc49004f..48e31711bc68 100644 |
9948 |
+--- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c |
9949 |
++++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c |
9950 |
+@@ -994,12 +994,15 @@ static int smu10_get_clock_by_type_with_latency(struct pp_hwmgr *hwmgr, |
9951 |
+ |
9952 |
+ clocks->num_levels = 0; |
9953 |
+ for (i = 0; i < pclk_vol_table->count; i++) { |
9954 |
+- clocks->data[i].clocks_in_khz = pclk_vol_table->entries[i].clk * 10; |
9955 |
+- clocks->data[i].latency_in_us = latency_required ? |
9956 |
+- smu10_get_mem_latency(hwmgr, |
9957 |
+- pclk_vol_table->entries[i].clk) : |
9958 |
+- 0; |
9959 |
+- clocks->num_levels++; |
9960 |
++ if (pclk_vol_table->entries[i].clk) { |
9961 |
++ clocks->data[clocks->num_levels].clocks_in_khz = |
9962 |
++ pclk_vol_table->entries[i].clk * 10; |
9963 |
++ clocks->data[clocks->num_levels].latency_in_us = latency_required ? |
9964 |
++ smu10_get_mem_latency(hwmgr, |
9965 |
++ pclk_vol_table->entries[i].clk) : |
9966 |
++ 0; |
9967 |
++ clocks->num_levels++; |
9968 |
++ } |
9969 |
+ } |
9970 |
+ |
9971 |
+ return 0; |
9972 |
+@@ -1045,9 +1048,11 @@ static int smu10_get_clock_by_type_with_voltage(struct pp_hwmgr *hwmgr, |
9973 |
+ |
9974 |
+ clocks->num_levels = 0; |
9975 |
+ for (i = 0; i < pclk_vol_table->count; i++) { |
9976 |
+- clocks->data[i].clocks_in_khz = pclk_vol_table->entries[i].clk * 10; |
9977 |
+- clocks->data[i].voltage_in_mv = pclk_vol_table->entries[i].vol; |
9978 |
+- clocks->num_levels++; |
9979 |
++ if (pclk_vol_table->entries[i].clk) { |
9980 |
++ clocks->data[clocks->num_levels].clocks_in_khz = pclk_vol_table->entries[i].clk * 10; |
9981 |
++ clocks->data[clocks->num_levels].voltage_in_mv = pclk_vol_table->entries[i].vol; |
9982 |
++ clocks->num_levels++; |
9983 |
++ } |
9984 |
+ } |
9985 |
+ |
9986 |
+ return 0; |
9987 |
+diff --git a/drivers/gpu/drm/drm_debugfs_crc.c b/drivers/gpu/drm/drm_debugfs_crc.c |
9988 |
+index c88e5ff41add..a3c756710845 100644 |
9989 |
+--- a/drivers/gpu/drm/drm_debugfs_crc.c |
9990 |
++++ b/drivers/gpu/drm/drm_debugfs_crc.c |
9991 |
+@@ -101,8 +101,8 @@ static ssize_t crc_control_write(struct file *file, const char __user *ubuf, |
9992 |
+ if (IS_ERR(source)) |
9993 |
+ return PTR_ERR(source); |
9994 |
+ |
9995 |
+- if (source[len] == '\n') |
9996 |
+- source[len] = '\0'; |
9997 |
++ if (source[len - 1] == '\n') |
9998 |
++ source[len - 1] = '\0'; |
9999 |
+ |
10000 |
+ spin_lock_irq(&crc->lock); |
10001 |
+ |
10002 |
+diff --git a/drivers/gpu/drm/gma500/framebuffer.c b/drivers/gpu/drm/gma500/framebuffer.c |
10003 |
+index adefae58b5fc..b4035ef72af8 100644 |
10004 |
+--- a/drivers/gpu/drm/gma500/framebuffer.c |
10005 |
++++ b/drivers/gpu/drm/gma500/framebuffer.c |
10006 |
+@@ -480,6 +480,7 @@ static int psbfb_probe(struct drm_fb_helper *helper, |
10007 |
+ container_of(helper, struct psb_fbdev, psb_fb_helper); |
10008 |
+ struct drm_device *dev = psb_fbdev->psb_fb_helper.dev; |
10009 |
+ struct drm_psb_private *dev_priv = dev->dev_private; |
10010 |
++ unsigned int fb_size; |
10011 |
+ int bytespp; |
10012 |
+ |
10013 |
+ bytespp = sizes->surface_bpp / 8; |
10014 |
+@@ -489,8 +490,11 @@ static int psbfb_probe(struct drm_fb_helper *helper, |
10015 |
+ /* If the mode will not fit in 32bit then switch to 16bit to get |
10016 |
+ a console on full resolution. The X mode setting server will |
10017 |
+ allocate its own 32bit GEM framebuffer */ |
10018 |
+- if (ALIGN(sizes->fb_width * bytespp, 64) * sizes->fb_height > |
10019 |
+- dev_priv->vram_stolen_size) { |
10020 |
++ fb_size = ALIGN(sizes->surface_width * bytespp, 64) * |
10021 |
++ sizes->surface_height; |
10022 |
++ fb_size = ALIGN(fb_size, PAGE_SIZE); |
10023 |
++ |
10024 |
++ if (fb_size > dev_priv->vram_stolen_size) { |
10025 |
+ sizes->surface_bpp = 16; |
10026 |
+ sizes->surface_depth = 16; |
10027 |
+ } |
10028 |
+diff --git a/drivers/gpu/drm/mediatek/mtk_drm_crtc.c b/drivers/gpu/drm/mediatek/mtk_drm_crtc.c |
10029 |
+index 92ecb9bf982c..b86ee7d25af3 100644 |
10030 |
+--- a/drivers/gpu/drm/mediatek/mtk_drm_crtc.c |
10031 |
++++ b/drivers/gpu/drm/mediatek/mtk_drm_crtc.c |
10032 |
+@@ -308,6 +308,7 @@ err_pm_runtime_put: |
10033 |
+ static void mtk_crtc_ddp_hw_fini(struct mtk_drm_crtc *mtk_crtc) |
10034 |
+ { |
10035 |
+ struct drm_device *drm = mtk_crtc->base.dev; |
10036 |
++ struct drm_crtc *crtc = &mtk_crtc->base; |
10037 |
+ int i; |
10038 |
+ |
10039 |
+ DRM_DEBUG_DRIVER("%s\n", __func__); |
10040 |
+@@ -329,6 +330,13 @@ static void mtk_crtc_ddp_hw_fini(struct mtk_drm_crtc *mtk_crtc) |
10041 |
+ mtk_disp_mutex_unprepare(mtk_crtc->mutex); |
10042 |
+ |
10043 |
+ pm_runtime_put(drm->dev); |
10044 |
++ |
10045 |
++ if (crtc->state->event && !crtc->state->active) { |
10046 |
++ spin_lock_irq(&crtc->dev->event_lock); |
10047 |
++ drm_crtc_send_vblank_event(crtc, crtc->state->event); |
10048 |
++ crtc->state->event = NULL; |
10049 |
++ spin_unlock_irq(&crtc->dev->event_lock); |
10050 |
++ } |
10051 |
+ } |
10052 |
+ |
10053 |
+ static void mtk_crtc_ddp_config(struct drm_crtc *crtc) |
10054 |
+diff --git a/drivers/gpu/drm/nouveau/nouveau_fence.c b/drivers/gpu/drm/nouveau/nouveau_fence.c |
10055 |
+index 412d49bc6e56..ba3883aed456 100644 |
10056 |
+--- a/drivers/gpu/drm/nouveau/nouveau_fence.c |
10057 |
++++ b/drivers/gpu/drm/nouveau/nouveau_fence.c |
10058 |
+@@ -157,7 +157,7 @@ nouveau_fence_wait_uevent_handler(struct nvif_notify *notify) |
10059 |
+ |
10060 |
+ fence = list_entry(fctx->pending.next, typeof(*fence), head); |
10061 |
+ chan = rcu_dereference_protected(fence->channel, lockdep_is_held(&fctx->lock)); |
10062 |
+- if (nouveau_fence_update(fence->channel, fctx)) |
10063 |
++ if (nouveau_fence_update(chan, fctx)) |
10064 |
+ ret = NVIF_NOTIFY_DROP; |
10065 |
+ } |
10066 |
+ spin_unlock_irqrestore(&fctx->lock, flags); |
10067 |
+diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.c b/drivers/gpu/drm/nouveau/nouveau_ttm.c |
10068 |
+index e4b977cc8452..37715a2a2f3f 100644 |
10069 |
+--- a/drivers/gpu/drm/nouveau/nouveau_ttm.c |
10070 |
++++ b/drivers/gpu/drm/nouveau/nouveau_ttm.c |
10071 |
+@@ -63,14 +63,12 @@ nouveau_vram_manager_new(struct ttm_mem_type_manager *man, |
10072 |
+ { |
10073 |
+ struct nouveau_bo *nvbo = nouveau_bo(bo); |
10074 |
+ struct nouveau_drm *drm = nouveau_bdev(bo->bdev); |
10075 |
+- struct nouveau_mem *mem; |
10076 |
+ int ret; |
10077 |
+ |
10078 |
+ if (drm->client.device.info.ram_size == 0) |
10079 |
+ return -ENOMEM; |
10080 |
+ |
10081 |
+ ret = nouveau_mem_new(&drm->master, nvbo->kind, nvbo->comp, reg); |
10082 |
+- mem = nouveau_mem(reg); |
10083 |
+ if (ret) |
10084 |
+ return ret; |
10085 |
+ |
10086 |
+@@ -103,11 +101,9 @@ nouveau_gart_manager_new(struct ttm_mem_type_manager *man, |
10087 |
+ { |
10088 |
+ struct nouveau_bo *nvbo = nouveau_bo(bo); |
10089 |
+ struct nouveau_drm *drm = nouveau_bdev(bo->bdev); |
10090 |
+- struct nouveau_mem *mem; |
10091 |
+ int ret; |
10092 |
+ |
10093 |
+ ret = nouveau_mem_new(&drm->master, nvbo->kind, nvbo->comp, reg); |
10094 |
+- mem = nouveau_mem(reg); |
10095 |
+ if (ret) |
10096 |
+ return ret; |
10097 |
+ |
10098 |
+diff --git a/drivers/gpu/drm/nouveau/nvkm/core/memory.c b/drivers/gpu/drm/nouveau/nvkm/core/memory.c |
10099 |
+index e85a08ecd9da..4cc186262d34 100644 |
10100 |
+--- a/drivers/gpu/drm/nouveau/nvkm/core/memory.c |
10101 |
++++ b/drivers/gpu/drm/nouveau/nvkm/core/memory.c |
10102 |
+@@ -91,8 +91,8 @@ nvkm_memory_tags_get(struct nvkm_memory *memory, struct nvkm_device *device, |
10103 |
+ } |
10104 |
+ |
10105 |
+ refcount_set(&tags->refcount, 1); |
10106 |
++ *ptags = memory->tags = tags; |
10107 |
+ mutex_unlock(&fb->subdev.mutex); |
10108 |
+- *ptags = tags; |
10109 |
+ return 0; |
10110 |
+ } |
10111 |
+ |
10112 |
+diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c |
10113 |
+index bcf32d92ee5a..50e3539f33d2 100644 |
10114 |
+--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c |
10115 |
++++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/channv50.c |
10116 |
+@@ -74,6 +74,8 @@ nv50_disp_chan_mthd(struct nv50_disp_chan *chan, int debug) |
10117 |
+ |
10118 |
+ if (debug > subdev->debug) |
10119 |
+ return; |
10120 |
++ if (!mthd) |
10121 |
++ return; |
10122 |
+ |
10123 |
+ for (i = 0; (list = mthd->data[i].mthd) != NULL; i++) { |
10124 |
+ u32 base = chan->head * mthd->addr; |
10125 |
+diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c |
10126 |
+index 500cb08dd608..b57ab5cea9a1 100644 |
10127 |
+--- a/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c |
10128 |
++++ b/drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c |
10129 |
+@@ -143,23 +143,24 @@ gk20a_gr_av_to_method(struct gf100_gr *gr, const char *fw_name, |
10130 |
+ |
10131 |
+ nent = (fuc.size / sizeof(struct gk20a_fw_av)); |
10132 |
+ |
10133 |
+- pack = vzalloc((sizeof(*pack) * max_classes) + |
10134 |
+- (sizeof(*init) * (nent + 1))); |
10135 |
++ pack = vzalloc((sizeof(*pack) * (max_classes + 1)) + |
10136 |
++ (sizeof(*init) * (nent + max_classes + 1))); |
10137 |
+ if (!pack) { |
10138 |
+ ret = -ENOMEM; |
10139 |
+ goto end; |
10140 |
+ } |
10141 |
+ |
10142 |
+- init = (void *)(pack + max_classes); |
10143 |
++ init = (void *)(pack + max_classes + 1); |
10144 |
+ |
10145 |
+- for (i = 0; i < nent; i++) { |
10146 |
+- struct gf100_gr_init *ent = &init[i]; |
10147 |
++ for (i = 0; i < nent; i++, init++) { |
10148 |
+ struct gk20a_fw_av *av = &((struct gk20a_fw_av *)fuc.data)[i]; |
10149 |
+ u32 class = av->addr & 0xffff; |
10150 |
+ u32 addr = (av->addr & 0xffff0000) >> 14; |
10151 |
+ |
10152 |
+ if (prevclass != class) { |
10153 |
+- pack[classidx].init = ent; |
10154 |
++ if (prevclass) /* Add terminator to the method list. */ |
10155 |
++ init++; |
10156 |
++ pack[classidx].init = init; |
10157 |
+ pack[classidx].type = class; |
10158 |
+ prevclass = class; |
10159 |
+ if (++classidx >= max_classes) { |
10160 |
+@@ -169,10 +170,10 @@ gk20a_gr_av_to_method(struct gf100_gr *gr, const char *fw_name, |
10161 |
+ } |
10162 |
+ } |
10163 |
+ |
10164 |
+- ent->addr = addr; |
10165 |
+- ent->data = av->data; |
10166 |
+- ent->count = 1; |
10167 |
+- ent->pitch = 1; |
10168 |
++ init->addr = addr; |
10169 |
++ init->data = av->data; |
10170 |
++ init->count = 1; |
10171 |
++ init->pitch = 1; |
10172 |
+ } |
10173 |
+ |
10174 |
+ *ppack = pack; |
10175 |
+diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/fault/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/fault/base.c |
10176 |
+index 16ad91c91a7b..f18ce6ff5b7e 100644 |
10177 |
+--- a/drivers/gpu/drm/nouveau/nvkm/subdev/fault/base.c |
10178 |
++++ b/drivers/gpu/drm/nouveau/nvkm/subdev/fault/base.c |
10179 |
+@@ -150,6 +150,7 @@ nvkm_fault_dtor(struct nvkm_subdev *subdev) |
10180 |
+ struct nvkm_fault *fault = nvkm_fault(subdev); |
10181 |
+ int i; |
10182 |
+ |
10183 |
++ nvkm_notify_fini(&fault->nrpfb); |
10184 |
+ nvkm_event_fini(&fault->event); |
10185 |
+ |
10186 |
+ for (i = 0; i < fault->buffer_nr; i++) { |
10187 |
+diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c |
10188 |
+index df8b919dcf09..ace6fefba428 100644 |
10189 |
+--- a/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c |
10190 |
++++ b/drivers/gpu/drm/nouveau/nvkm/subdev/secboot/gm20b.c |
10191 |
+@@ -108,6 +108,7 @@ gm20b_secboot_new(struct nvkm_device *device, int index, |
10192 |
+ struct gm200_secboot *gsb; |
10193 |
+ struct nvkm_acr *acr; |
10194 |
+ |
10195 |
++ *psb = NULL; |
10196 |
+ acr = acr_r352_new(BIT(NVKM_SECBOOT_FALCON_FECS) | |
10197 |
+ BIT(NVKM_SECBOOT_FALCON_PMU)); |
10198 |
+ if (IS_ERR(acr)) |
10199 |
+@@ -116,10 +117,8 @@ gm20b_secboot_new(struct nvkm_device *device, int index, |
10200 |
+ acr->optional_falcons = BIT(NVKM_SECBOOT_FALCON_PMU); |
10201 |
+ |
10202 |
+ gsb = kzalloc(sizeof(*gsb), GFP_KERNEL); |
10203 |
+- if (!gsb) { |
10204 |
+- psb = NULL; |
10205 |
++ if (!gsb) |
10206 |
+ return -ENOMEM; |
10207 |
+- } |
10208 |
+ *psb = &gsb->base; |
10209 |
+ |
10210 |
+ ret = nvkm_secboot_ctor(&gm20b_secboot, acr, device, index, &gsb->base); |
10211 |
+diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c |
10212 |
+index d8e2d7b3b836..7d1e14f0140a 100644 |
10213 |
+--- a/drivers/gpu/drm/radeon/radeon_display.c |
10214 |
++++ b/drivers/gpu/drm/radeon/radeon_display.c |
10215 |
+@@ -121,6 +121,8 @@ static void dce5_crtc_load_lut(struct drm_crtc *crtc) |
10216 |
+ |
10217 |
+ DRM_DEBUG_KMS("%d\n", radeon_crtc->crtc_id); |
10218 |
+ |
10219 |
++ msleep(10); |
10220 |
++ |
10221 |
+ WREG32(NI_INPUT_CSC_CONTROL + radeon_crtc->crtc_offset, |
10222 |
+ (NI_INPUT_CSC_GRPH_MODE(NI_INPUT_CSC_BYPASS) | |
10223 |
+ NI_INPUT_CSC_OVL_MODE(NI_INPUT_CSC_BYPASS))); |
10224 |
+diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf_res.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf_res.c |
10225 |
+index 3b75af9bf85f..f27bd7cff579 100644 |
10226 |
+--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf_res.c |
10227 |
++++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf_res.c |
10228 |
+@@ -210,8 +210,10 @@ int vmw_cmdbuf_res_add(struct vmw_cmdbuf_res_manager *man, |
10229 |
+ |
10230 |
+ cres->hash.key = user_key | (res_type << 24); |
10231 |
+ ret = drm_ht_insert_item(&man->resources, &cres->hash); |
10232 |
+- if (unlikely(ret != 0)) |
10233 |
++ if (unlikely(ret != 0)) { |
10234 |
++ kfree(cres); |
10235 |
+ goto out_invalid_key; |
10236 |
++ } |
10237 |
+ |
10238 |
+ cres->state = VMW_CMDBUF_RES_ADD; |
10239 |
+ cres->res = vmw_resource_reference(res); |
10240 |
+diff --git a/drivers/ide/cmd64x.c b/drivers/ide/cmd64x.c |
10241 |
+index b127ed60c733..9dde8390da09 100644 |
10242 |
+--- a/drivers/ide/cmd64x.c |
10243 |
++++ b/drivers/ide/cmd64x.c |
10244 |
+@@ -65,6 +65,9 @@ static void cmd64x_program_timings(ide_drive_t *drive, u8 mode) |
10245 |
+ struct ide_timing t; |
10246 |
+ u8 arttim = 0; |
10247 |
+ |
10248 |
++ if (drive->dn >= ARRAY_SIZE(drwtim_regs)) |
10249 |
++ return; |
10250 |
++ |
10251 |
+ ide_timing_compute(drive, mode, &t, T, 0); |
10252 |
+ |
10253 |
+ /* |
10254 |
+diff --git a/drivers/ide/serverworks.c b/drivers/ide/serverworks.c |
10255 |
+index a97affca18ab..0f57d45484d1 100644 |
10256 |
+--- a/drivers/ide/serverworks.c |
10257 |
++++ b/drivers/ide/serverworks.c |
10258 |
+@@ -114,6 +114,9 @@ static void svwks_set_pio_mode(ide_hwif_t *hwif, ide_drive_t *drive) |
10259 |
+ struct pci_dev *dev = to_pci_dev(hwif->dev); |
10260 |
+ const u8 pio = drive->pio_mode - XFER_PIO_0; |
10261 |
+ |
10262 |
++ if (drive->dn >= ARRAY_SIZE(drive_pci)) |
10263 |
++ return; |
10264 |
++ |
10265 |
+ pci_write_config_byte(dev, drive_pci[drive->dn], pio_modes[pio]); |
10266 |
+ |
10267 |
+ if (svwks_csb_check(dev)) { |
10268 |
+@@ -140,6 +143,9 @@ static void svwks_set_dma_mode(ide_hwif_t *hwif, ide_drive_t *drive) |
10269 |
+ |
10270 |
+ u8 ultra_enable = 0, ultra_timing = 0, dma_timing = 0; |
10271 |
+ |
10272 |
++ if (drive->dn >= ARRAY_SIZE(drive_pci2)) |
10273 |
++ return; |
10274 |
++ |
10275 |
+ pci_read_config_byte(dev, (0x56|hwif->channel), &ultra_timing); |
10276 |
+ pci_read_config_byte(dev, 0x54, &ultra_enable); |
10277 |
+ |
10278 |
+diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c |
10279 |
+index b09a4b1cf397..1221faea75a6 100644 |
10280 |
+--- a/drivers/infiniband/hw/hfi1/chip.c |
10281 |
++++ b/drivers/infiniband/hw/hfi1/chip.c |
10282 |
+@@ -1687,6 +1687,14 @@ static u64 access_sw_pio_drain(const struct cntr_entry *entry, |
10283 |
+ return dd->verbs_dev.n_piodrain; |
10284 |
+ } |
10285 |
+ |
10286 |
++static u64 access_sw_ctx0_seq_drop(const struct cntr_entry *entry, |
10287 |
++ void *context, int vl, int mode, u64 data) |
10288 |
++{ |
10289 |
++ struct hfi1_devdata *dd = context; |
10290 |
++ |
10291 |
++ return dd->ctx0_seq_drop; |
10292 |
++} |
10293 |
++ |
10294 |
+ static u64 access_sw_vtx_wait(const struct cntr_entry *entry, |
10295 |
+ void *context, int vl, int mode, u64 data) |
10296 |
+ { |
10297 |
+@@ -4247,6 +4255,8 @@ static struct cntr_entry dev_cntrs[DEV_CNTR_LAST] = { |
10298 |
+ access_sw_cpu_intr), |
10299 |
+ [C_SW_CPU_RCV_LIM] = CNTR_ELEM("RcvLimit", 0, 0, CNTR_NORMAL, |
10300 |
+ access_sw_cpu_rcv_limit), |
10301 |
++[C_SW_CTX0_SEQ_DROP] = CNTR_ELEM("SeqDrop0", 0, 0, CNTR_NORMAL, |
10302 |
++ access_sw_ctx0_seq_drop), |
10303 |
+ [C_SW_VTX_WAIT] = CNTR_ELEM("vTxWait", 0, 0, CNTR_NORMAL, |
10304 |
+ access_sw_vtx_wait), |
10305 |
+ [C_SW_PIO_WAIT] = CNTR_ELEM("PioWait", 0, 0, CNTR_NORMAL, |
10306 |
+diff --git a/drivers/infiniband/hw/hfi1/chip.h b/drivers/infiniband/hw/hfi1/chip.h |
10307 |
+index 36b04d6300e5..c9a352d8a7e1 100644 |
10308 |
+--- a/drivers/infiniband/hw/hfi1/chip.h |
10309 |
++++ b/drivers/infiniband/hw/hfi1/chip.h |
10310 |
+@@ -909,6 +909,7 @@ enum { |
10311 |
+ C_DC_PG_STS_TX_MBE_CNT, |
10312 |
+ C_SW_CPU_INTR, |
10313 |
+ C_SW_CPU_RCV_LIM, |
10314 |
++ C_SW_CTX0_SEQ_DROP, |
10315 |
+ C_SW_VTX_WAIT, |
10316 |
+ C_SW_PIO_WAIT, |
10317 |
+ C_SW_PIO_DRAIN, |
10318 |
+diff --git a/drivers/infiniband/hw/hfi1/driver.c b/drivers/infiniband/hw/hfi1/driver.c |
10319 |
+index d5277c23cba6..769e114567a0 100644 |
10320 |
+--- a/drivers/infiniband/hw/hfi1/driver.c |
10321 |
++++ b/drivers/infiniband/hw/hfi1/driver.c |
10322 |
+@@ -734,6 +734,7 @@ static noinline int skip_rcv_packet(struct hfi1_packet *packet, int thread) |
10323 |
+ { |
10324 |
+ int ret; |
10325 |
+ |
10326 |
++ packet->rcd->dd->ctx0_seq_drop++; |
10327 |
+ /* Set up for the next packet */ |
10328 |
+ packet->rhqoff += packet->rsize; |
10329 |
+ if (packet->rhqoff >= packet->maxcnt) |
10330 |
+diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h |
10331 |
+index ab981874c71c..e38de547785d 100644 |
10332 |
+--- a/drivers/infiniband/hw/hfi1/hfi.h |
10333 |
++++ b/drivers/infiniband/hw/hfi1/hfi.h |
10334 |
+@@ -1093,6 +1093,8 @@ struct hfi1_devdata { |
10335 |
+ |
10336 |
+ char *boardname; /* human readable board info */ |
10337 |
+ |
10338 |
++ u64 ctx0_seq_drop; |
10339 |
++ |
10340 |
+ /* reset value */ |
10341 |
+ u64 z_int_counter; |
10342 |
+ u64 z_rcv_limit; |
10343 |
+diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.h b/drivers/infiniband/sw/rxe/rxe_verbs.h |
10344 |
+index 6a75f96b9096..b4e24362edbb 100644 |
10345 |
+--- a/drivers/infiniband/sw/rxe/rxe_verbs.h |
10346 |
++++ b/drivers/infiniband/sw/rxe/rxe_verbs.h |
10347 |
+@@ -407,7 +407,7 @@ struct rxe_dev { |
10348 |
+ struct list_head pending_mmaps; |
10349 |
+ |
10350 |
+ spinlock_t mmap_offset_lock; /* guard mmap_offset */ |
10351 |
+- int mmap_offset; |
10352 |
++ u64 mmap_offset; |
10353 |
+ |
10354 |
+ atomic64_t stats_counters[RXE_NUM_OF_COUNTERS]; |
10355 |
+ |
10356 |
+diff --git a/drivers/input/touchscreen/edt-ft5x06.c b/drivers/input/touchscreen/edt-ft5x06.c |
10357 |
+index 1e18ca0d1b4e..3fdaa644a82c 100644 |
10358 |
+--- a/drivers/input/touchscreen/edt-ft5x06.c |
10359 |
++++ b/drivers/input/touchscreen/edt-ft5x06.c |
10360 |
+@@ -968,6 +968,7 @@ static int edt_ft5x06_ts_probe(struct i2c_client *client, |
10361 |
+ { |
10362 |
+ const struct edt_i2c_chip_data *chip_data; |
10363 |
+ struct edt_ft5x06_ts_data *tsdata; |
10364 |
++ u8 buf[2] = { 0xfc, 0x00 }; |
10365 |
+ struct input_dev *input; |
10366 |
+ unsigned long irq_flags; |
10367 |
+ int error; |
10368 |
+@@ -1037,6 +1038,12 @@ static int edt_ft5x06_ts_probe(struct i2c_client *client, |
10369 |
+ return error; |
10370 |
+ } |
10371 |
+ |
10372 |
++ /* |
10373 |
++ * Dummy read access. EP0700MLP1 returns bogus data on the first |
10374 |
++ * register read access and ignores writes. |
10375 |
++ */ |
10376 |
++ edt_ft5x06_ts_readwrite(tsdata->client, 2, buf, 2, buf); |
10377 |
++ |
10378 |
+ edt_ft5x06_ts_set_regs(tsdata); |
10379 |
+ edt_ft5x06_ts_get_defaults(&client->dev, tsdata); |
10380 |
+ edt_ft5x06_ts_get_parameters(tsdata); |
10381 |
+diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c |
10382 |
+index eff1f3aa5ef4..6b7664052b5b 100644 |
10383 |
+--- a/drivers/iommu/arm-smmu-v3.c |
10384 |
++++ b/drivers/iommu/arm-smmu-v3.c |
10385 |
+@@ -1185,7 +1185,8 @@ static void arm_smmu_write_strtab_ent(struct arm_smmu_device *smmu, u32 sid, |
10386 |
+ } |
10387 |
+ |
10388 |
+ arm_smmu_sync_ste_for_sid(smmu, sid); |
10389 |
+- dst[0] = cpu_to_le64(val); |
10390 |
++ /* See comment in arm_smmu_write_ctx_desc() */ |
10391 |
++ WRITE_ONCE(dst[0], cpu_to_le64(val)); |
10392 |
+ arm_smmu_sync_ste_for_sid(smmu, sid); |
10393 |
+ |
10394 |
+ /* It's likely that we'll want to use the new STE soon */ |
10395 |
+diff --git a/drivers/iommu/dmar.c b/drivers/iommu/dmar.c |
10396 |
+index 7f9824b0609e..72994d67bc5b 100644 |
10397 |
+--- a/drivers/iommu/dmar.c |
10398 |
++++ b/drivers/iommu/dmar.c |
10399 |
+@@ -1345,7 +1345,6 @@ void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 pfsid, |
10400 |
+ struct qi_desc desc; |
10401 |
+ |
10402 |
+ if (mask) { |
10403 |
+- WARN_ON_ONCE(addr & ((1ULL << (VTD_PAGE_SHIFT + mask)) - 1)); |
10404 |
+ addr |= (1ULL << (VTD_PAGE_SHIFT + mask - 1)) - 1; |
10405 |
+ desc.high = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE; |
10406 |
+ } else |
10407 |
+diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c |
10408 |
+index fd8730b2cd46..5944d3b4dca3 100644 |
10409 |
+--- a/drivers/iommu/intel-svm.c |
10410 |
++++ b/drivers/iommu/intel-svm.c |
10411 |
+@@ -377,7 +377,7 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ |
10412 |
+ /* Do not use PASID 0 in caching mode (virtualised IOMMU) */ |
10413 |
+ ret = intel_pasid_alloc_id(svm, |
10414 |
+ !!cap_caching_mode(iommu->cap), |
10415 |
+- pasid_max - 1, GFP_KERNEL); |
10416 |
++ pasid_max, GFP_KERNEL); |
10417 |
+ if (ret < 0) { |
10418 |
+ kfree(svm); |
10419 |
+ kfree(sdev); |
10420 |
+diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c |
10421 |
+index 050d6e040128..bf7b69449b43 100644 |
10422 |
+--- a/drivers/irqchip/irq-gic-v3-its.c |
10423 |
++++ b/drivers/irqchip/irq-gic-v3-its.c |
10424 |
+@@ -578,7 +578,7 @@ static struct its_collection *its_build_invall_cmd(struct its_node *its, |
10425 |
+ struct its_cmd_desc *desc) |
10426 |
+ { |
10427 |
+ its_encode_cmd(cmd, GITS_CMD_INVALL); |
10428 |
+- its_encode_collection(cmd, desc->its_mapc_cmd.col->col_id); |
10429 |
++ its_encode_collection(cmd, desc->its_invall_cmd.col->col_id); |
10430 |
+ |
10431 |
+ its_fixup_cmd(cmd); |
10432 |
+ |
10433 |
+diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c |
10434 |
+index d5912f1ec884..ac888d7a0b00 100644 |
10435 |
+--- a/drivers/irqchip/irq-gic-v3.c |
10436 |
++++ b/drivers/irqchip/irq-gic-v3.c |
10437 |
+@@ -1347,6 +1347,7 @@ static struct |
10438 |
+ struct redist_region *redist_regs; |
10439 |
+ u32 nr_redist_regions; |
10440 |
+ bool single_redist; |
10441 |
++ int enabled_rdists; |
10442 |
+ u32 maint_irq; |
10443 |
+ int maint_irq_mode; |
10444 |
+ phys_addr_t vcpu_base; |
10445 |
+@@ -1441,8 +1442,10 @@ static int __init gic_acpi_match_gicc(struct acpi_subtable_header *header, |
10446 |
+ * If GICC is enabled and has valid gicr base address, then it means |
10447 |
+ * GICR base is presented via GICC |
10448 |
+ */ |
10449 |
+- if ((gicc->flags & ACPI_MADT_ENABLED) && gicc->gicr_base_address) |
10450 |
++ if ((gicc->flags & ACPI_MADT_ENABLED) && gicc->gicr_base_address) { |
10451 |
++ acpi_data.enabled_rdists++; |
10452 |
+ return 0; |
10453 |
++ } |
10454 |
+ |
10455 |
+ /* |
10456 |
+ * It's perfectly valid firmware can pass disabled GICC entry, driver |
10457 |
+@@ -1472,8 +1475,10 @@ static int __init gic_acpi_count_gicr_regions(void) |
10458 |
+ |
10459 |
+ count = acpi_table_parse_madt(ACPI_MADT_TYPE_GENERIC_INTERRUPT, |
10460 |
+ gic_acpi_match_gicc, 0); |
10461 |
+- if (count > 0) |
10462 |
++ if (count > 0) { |
10463 |
+ acpi_data.single_redist = true; |
10464 |
++ count = acpi_data.enabled_rdists; |
10465 |
++ } |
10466 |
+ |
10467 |
+ return count; |
10468 |
+ } |
10469 |
+diff --git a/drivers/irqchip/irq-mbigen.c b/drivers/irqchip/irq-mbigen.c |
10470 |
+index 98b6e1d4b1a6..f7fdbf5d183b 100644 |
10471 |
+--- a/drivers/irqchip/irq-mbigen.c |
10472 |
++++ b/drivers/irqchip/irq-mbigen.c |
10473 |
+@@ -381,6 +381,7 @@ static struct platform_driver mbigen_platform_driver = { |
10474 |
+ .name = "Hisilicon MBIGEN-V2", |
10475 |
+ .of_match_table = mbigen_of_match, |
10476 |
+ .acpi_match_table = ACPI_PTR(mbigen_acpi_match), |
10477 |
++ .suppress_bind_attrs = true, |
10478 |
+ }, |
10479 |
+ .probe = mbigen_device_probe, |
10480 |
+ }; |
10481 |
+diff --git a/drivers/leds/leds-pca963x.c b/drivers/leds/leds-pca963x.c |
10482 |
+index 5c0908113e38..bbcde13b77f1 100644 |
10483 |
+--- a/drivers/leds/leds-pca963x.c |
10484 |
++++ b/drivers/leds/leds-pca963x.c |
10485 |
+@@ -43,6 +43,8 @@ |
10486 |
+ #define PCA963X_LED_PWM 0x2 /* Controlled through PWM */ |
10487 |
+ #define PCA963X_LED_GRP_PWM 0x3 /* Controlled through PWM/GRPPWM */ |
10488 |
+ |
10489 |
++#define PCA963X_MODE2_OUTDRV 0x04 /* Open-drain or totem pole */ |
10490 |
++#define PCA963X_MODE2_INVRT 0x10 /* Normal or inverted direction */ |
10491 |
+ #define PCA963X_MODE2_DMBLNK 0x20 /* Enable blinking */ |
10492 |
+ |
10493 |
+ #define PCA963X_MODE1 0x00 |
10494 |
+@@ -462,12 +464,12 @@ static int pca963x_probe(struct i2c_client *client, |
10495 |
+ PCA963X_MODE2); |
10496 |
+ /* Configure output: open-drain or totem pole (push-pull) */ |
10497 |
+ if (pdata->outdrv == PCA963X_OPEN_DRAIN) |
10498 |
+- mode2 |= 0x01; |
10499 |
++ mode2 &= ~PCA963X_MODE2_OUTDRV; |
10500 |
+ else |
10501 |
+- mode2 |= 0x05; |
10502 |
++ mode2 |= PCA963X_MODE2_OUTDRV; |
10503 |
+ /* Configure direction: normal or inverted */ |
10504 |
+ if (pdata->dir == PCA963X_INVERTED) |
10505 |
+- mode2 |= 0x10; |
10506 |
++ mode2 |= PCA963X_MODE2_INVRT; |
10507 |
+ i2c_smbus_write_byte_data(pca963x->chip->client, PCA963X_MODE2, |
10508 |
+ mode2); |
10509 |
+ } |
10510 |
+diff --git a/drivers/md/bcache/bset.h b/drivers/md/bcache/bset.h |
10511 |
+index c71365e7c1fa..a50dcfda656f 100644 |
10512 |
+--- a/drivers/md/bcache/bset.h |
10513 |
++++ b/drivers/md/bcache/bset.h |
10514 |
+@@ -397,7 +397,8 @@ void bch_btree_keys_stats(struct btree_keys *b, struct bset_stats *state); |
10515 |
+ |
10516 |
+ /* Bkey utility code */ |
10517 |
+ |
10518 |
+-#define bset_bkey_last(i) bkey_idx((struct bkey *) (i)->d, (i)->keys) |
10519 |
++#define bset_bkey_last(i) bkey_idx((struct bkey *) (i)->d, \ |
10520 |
++ (unsigned int)(i)->keys) |
10521 |
+ |
10522 |
+ static inline struct bkey *bset_bkey_idx(struct bset *i, unsigned int idx) |
10523 |
+ { |
10524 |
+diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c |
10525 |
+index c45d9ad01077..5b5cbfadd003 100644 |
10526 |
+--- a/drivers/md/bcache/super.c |
10527 |
++++ b/drivers/md/bcache/super.c |
10528 |
+@@ -1226,6 +1226,9 @@ static void cached_dev_free(struct closure *cl) |
10529 |
+ |
10530 |
+ mutex_unlock(&bch_register_lock); |
10531 |
+ |
10532 |
++ if (dc->sb_bio.bi_inline_vecs[0].bv_page) |
10533 |
++ put_page(bio_first_page_all(&dc->sb_bio)); |
10534 |
++ |
10535 |
+ if (!IS_ERR_OR_NULL(dc->bdev)) |
10536 |
+ blkdev_put(dc->bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
10537 |
+ |
10538 |
+diff --git a/drivers/media/i2c/mt9v032.c b/drivers/media/i2c/mt9v032.c |
10539 |
+index f74730d24d8f..04788692c9ff 100644 |
10540 |
+--- a/drivers/media/i2c/mt9v032.c |
10541 |
++++ b/drivers/media/i2c/mt9v032.c |
10542 |
+@@ -431,10 +431,12 @@ static int mt9v032_enum_mbus_code(struct v4l2_subdev *subdev, |
10543 |
+ struct v4l2_subdev_pad_config *cfg, |
10544 |
+ struct v4l2_subdev_mbus_code_enum *code) |
10545 |
+ { |
10546 |
++ struct mt9v032 *mt9v032 = to_mt9v032(subdev); |
10547 |
++ |
10548 |
+ if (code->index > 0) |
10549 |
+ return -EINVAL; |
10550 |
+ |
10551 |
+- code->code = MEDIA_BUS_FMT_SGRBG10_1X10; |
10552 |
++ code->code = mt9v032->format.code; |
10553 |
+ return 0; |
10554 |
+ } |
10555 |
+ |
10556 |
+@@ -442,7 +444,11 @@ static int mt9v032_enum_frame_size(struct v4l2_subdev *subdev, |
10557 |
+ struct v4l2_subdev_pad_config *cfg, |
10558 |
+ struct v4l2_subdev_frame_size_enum *fse) |
10559 |
+ { |
10560 |
+- if (fse->index >= 3 || fse->code != MEDIA_BUS_FMT_SGRBG10_1X10) |
10561 |
++ struct mt9v032 *mt9v032 = to_mt9v032(subdev); |
10562 |
++ |
10563 |
++ if (fse->index >= 3) |
10564 |
++ return -EINVAL; |
10565 |
++ if (mt9v032->format.code != fse->code) |
10566 |
+ return -EINVAL; |
10567 |
+ |
10568 |
+ fse->min_width = MT9V032_WINDOW_WIDTH_DEF / (1 << fse->index); |
10569 |
+diff --git a/drivers/media/pci/cx23885/cx23885-cards.c b/drivers/media/pci/cx23885/cx23885-cards.c |
10570 |
+index ed3210dc50bc..642aefdbb7bb 100644 |
10571 |
+--- a/drivers/media/pci/cx23885/cx23885-cards.c |
10572 |
++++ b/drivers/media/pci/cx23885/cx23885-cards.c |
10573 |
+@@ -811,6 +811,25 @@ struct cx23885_board cx23885_boards[] = { |
10574 |
+ .name = "Hauppauge WinTV-Starburst2", |
10575 |
+ .portb = CX23885_MPEG_DVB, |
10576 |
+ }, |
10577 |
++ [CX23885_BOARD_AVERMEDIA_CE310B] = { |
10578 |
++ .name = "AVerMedia CE310B", |
10579 |
++ .porta = CX23885_ANALOG_VIDEO, |
10580 |
++ .force_bff = 1, |
10581 |
++ .input = {{ |
10582 |
++ .type = CX23885_VMUX_COMPOSITE1, |
10583 |
++ .vmux = CX25840_VIN1_CH1 | |
10584 |
++ CX25840_NONE_CH2 | |
10585 |
++ CX25840_NONE0_CH3, |
10586 |
++ .amux = CX25840_AUDIO7, |
10587 |
++ }, { |
10588 |
++ .type = CX23885_VMUX_SVIDEO, |
10589 |
++ .vmux = CX25840_VIN8_CH1 | |
10590 |
++ CX25840_NONE_CH2 | |
10591 |
++ CX25840_VIN7_CH3 | |
10592 |
++ CX25840_SVIDEO_ON, |
10593 |
++ .amux = CX25840_AUDIO7, |
10594 |
++ } }, |
10595 |
++ }, |
10596 |
+ }; |
10597 |
+ const unsigned int cx23885_bcount = ARRAY_SIZE(cx23885_boards); |
10598 |
+ |
10599 |
+@@ -1134,6 +1153,10 @@ struct cx23885_subid cx23885_subids[] = { |
10600 |
+ .subvendor = 0x0070, |
10601 |
+ .subdevice = 0xf02a, |
10602 |
+ .card = CX23885_BOARD_HAUPPAUGE_STARBURST2, |
10603 |
++ }, { |
10604 |
++ .subvendor = 0x1461, |
10605 |
++ .subdevice = 0x3100, |
10606 |
++ .card = CX23885_BOARD_AVERMEDIA_CE310B, |
10607 |
+ }, |
10608 |
+ }; |
10609 |
+ const unsigned int cx23885_idcount = ARRAY_SIZE(cx23885_subids); |
10610 |
+@@ -2358,6 +2381,7 @@ void cx23885_card_setup(struct cx23885_dev *dev) |
10611 |
+ case CX23885_BOARD_DVBSKY_T982: |
10612 |
+ case CX23885_BOARD_VIEWCAST_260E: |
10613 |
+ case CX23885_BOARD_VIEWCAST_460E: |
10614 |
++ case CX23885_BOARD_AVERMEDIA_CE310B: |
10615 |
+ dev->sd_cx25840 = v4l2_i2c_new_subdev(&dev->v4l2_dev, |
10616 |
+ &dev->i2c_bus[2].i2c_adap, |
10617 |
+ "cx25840", 0x88 >> 1, NULL); |
10618 |
+diff --git a/drivers/media/pci/cx23885/cx23885-video.c b/drivers/media/pci/cx23885/cx23885-video.c |
10619 |
+index f8a3deadc77a..2a20c7165e1e 100644 |
10620 |
+--- a/drivers/media/pci/cx23885/cx23885-video.c |
10621 |
++++ b/drivers/media/pci/cx23885/cx23885-video.c |
10622 |
+@@ -268,7 +268,8 @@ static int cx23885_video_mux(struct cx23885_dev *dev, unsigned int input) |
10623 |
+ (dev->board == CX23885_BOARD_MYGICA_X8507) || |
10624 |
+ (dev->board == CX23885_BOARD_AVERMEDIA_HC81R) || |
10625 |
+ (dev->board == CX23885_BOARD_VIEWCAST_260E) || |
10626 |
+- (dev->board == CX23885_BOARD_VIEWCAST_460E)) { |
10627 |
++ (dev->board == CX23885_BOARD_VIEWCAST_460E) || |
10628 |
++ (dev->board == CX23885_BOARD_AVERMEDIA_CE310B)) { |
10629 |
+ /* Configure audio routing */ |
10630 |
+ v4l2_subdev_call(dev->sd_cx25840, audio, s_routing, |
10631 |
+ INPUT(input)->amux, 0, 0); |
10632 |
+diff --git a/drivers/media/pci/cx23885/cx23885.h b/drivers/media/pci/cx23885/cx23885.h |
10633 |
+index cf965efabe66..7bbd62cc993e 100644 |
10634 |
+--- a/drivers/media/pci/cx23885/cx23885.h |
10635 |
++++ b/drivers/media/pci/cx23885/cx23885.h |
10636 |
+@@ -111,6 +111,7 @@ |
10637 |
+ #define CX23885_BOARD_HAUPPAUGE_STARBURST2 59 |
10638 |
+ #define CX23885_BOARD_HAUPPAUGE_QUADHD_DVB_885 60 |
10639 |
+ #define CX23885_BOARD_HAUPPAUGE_QUADHD_ATSC_885 61 |
10640 |
++#define CX23885_BOARD_AVERMEDIA_CE310B 62 |
10641 |
+ |
10642 |
+ #define GPIO_0 0x00000001 |
10643 |
+ #define GPIO_1 0x00000002 |
10644 |
+diff --git a/drivers/media/platform/sti/bdisp/bdisp-hw.c b/drivers/media/platform/sti/bdisp/bdisp-hw.c |
10645 |
+index 26d9fa7aeb5f..d57f659d740a 100644 |
10646 |
+--- a/drivers/media/platform/sti/bdisp/bdisp-hw.c |
10647 |
++++ b/drivers/media/platform/sti/bdisp/bdisp-hw.c |
10648 |
+@@ -14,8 +14,8 @@ |
10649 |
+ #define MAX_SRC_WIDTH 2048 |
10650 |
+ |
10651 |
+ /* Reset & boot poll config */ |
10652 |
+-#define POLL_RST_MAX 50 |
10653 |
+-#define POLL_RST_DELAY_MS 20 |
10654 |
++#define POLL_RST_MAX 500 |
10655 |
++#define POLL_RST_DELAY_MS 2 |
10656 |
+ |
10657 |
+ enum bdisp_target_plan { |
10658 |
+ BDISP_RGB, |
10659 |
+@@ -382,7 +382,7 @@ int bdisp_hw_reset(struct bdisp_dev *bdisp) |
10660 |
+ for (i = 0; i < POLL_RST_MAX; i++) { |
10661 |
+ if (readl(bdisp->regs + BLT_STA1) & BLT_STA1_IDLE) |
10662 |
+ break; |
10663 |
+- msleep(POLL_RST_DELAY_MS); |
10664 |
++ udelay(POLL_RST_DELAY_MS * 1000); |
10665 |
+ } |
10666 |
+ if (i == POLL_RST_MAX) |
10667 |
+ dev_err(bdisp->dev, "Reset timeout\n"); |
10668 |
+diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c |
10669 |
+index 733d9172425b..026a3bd71204 100644 |
10670 |
+--- a/drivers/net/ethernet/cisco/enic/enic_main.c |
10671 |
++++ b/drivers/net/ethernet/cisco/enic/enic_main.c |
10672 |
+@@ -2013,10 +2013,10 @@ static int enic_stop(struct net_device *netdev) |
10673 |
+ napi_disable(&enic->napi[i]); |
10674 |
+ |
10675 |
+ netif_carrier_off(netdev); |
10676 |
+- netif_tx_disable(netdev); |
10677 |
+ if (vnic_dev_get_intr_mode(enic->vdev) == VNIC_DEV_INTR_MODE_MSIX) |
10678 |
+ for (i = 0; i < enic->wq_count; i++) |
10679 |
+ napi_disable(&enic->napi[enic_cq_wq(enic, i)]); |
10680 |
++ netif_tx_disable(netdev); |
10681 |
+ |
10682 |
+ if (!enic_is_dynamic(enic) && !enic_is_sriov_vf(enic)) |
10683 |
+ enic_dev_del_station_addr(enic); |
10684 |
+diff --git a/drivers/net/ethernet/freescale/gianfar.c b/drivers/net/ethernet/freescale/gianfar.c |
10685 |
+index c97c4edfa31b..cf2d1e846a69 100644 |
10686 |
+--- a/drivers/net/ethernet/freescale/gianfar.c |
10687 |
++++ b/drivers/net/ethernet/freescale/gianfar.c |
10688 |
+@@ -2685,13 +2685,17 @@ static void gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue) |
10689 |
+ skb_dirtytx = tx_queue->skb_dirtytx; |
10690 |
+ |
10691 |
+ while ((skb = tx_queue->tx_skbuff[skb_dirtytx])) { |
10692 |
++ bool do_tstamp; |
10693 |
++ |
10694 |
++ do_tstamp = (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) && |
10695 |
++ priv->hwts_tx_en; |
10696 |
+ |
10697 |
+ frags = skb_shinfo(skb)->nr_frags; |
10698 |
+ |
10699 |
+ /* When time stamping, one additional TxBD must be freed. |
10700 |
+ * Also, we need to dma_unmap_single() the TxPAL. |
10701 |
+ */ |
10702 |
+- if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) |
10703 |
++ if (unlikely(do_tstamp)) |
10704 |
+ nr_txbds = frags + 2; |
10705 |
+ else |
10706 |
+ nr_txbds = frags + 1; |
10707 |
+@@ -2705,7 +2709,7 @@ static void gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue) |
10708 |
+ (lstatus & BD_LENGTH_MASK)) |
10709 |
+ break; |
10710 |
+ |
10711 |
+- if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) { |
10712 |
++ if (unlikely(do_tstamp)) { |
10713 |
+ next = next_txbd(bdp, base, tx_ring_size); |
10714 |
+ buflen = be16_to_cpu(next->length) + |
10715 |
+ GMAC_FCB_LEN + GMAC_TXPAL_LEN; |
10716 |
+@@ -2715,7 +2719,7 @@ static void gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue) |
10717 |
+ dma_unmap_single(priv->dev, be32_to_cpu(bdp->bufPtr), |
10718 |
+ buflen, DMA_TO_DEVICE); |
10719 |
+ |
10720 |
+- if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) { |
10721 |
++ if (unlikely(do_tstamp)) { |
10722 |
+ struct skb_shared_hwtstamps shhwtstamps; |
10723 |
+ u64 *ns = (u64 *)(((uintptr_t)skb->data + 0x10) & |
10724 |
+ ~0x7UL); |
10725 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c |
10726 |
+index 8255d797ea94..9a68dee588c1 100644 |
10727 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c |
10728 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/en_stats.c |
10729 |
+@@ -211,6 +211,9 @@ void mlx5e_grp_sw_update_stats(struct mlx5e_priv *priv) |
10730 |
+ s->tx_tls_resync_bytes += sq_stats->tls_resync_bytes; |
10731 |
+ #endif |
10732 |
+ s->tx_cqes += sq_stats->cqes; |
10733 |
++ |
10734 |
++ /* https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92657 */ |
10735 |
++ barrier(); |
10736 |
+ } |
10737 |
+ } |
10738 |
+ |
10739 |
+diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_dpipe.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_dpipe.c |
10740 |
+index 41e607a14846..4fe193c4fa55 100644 |
10741 |
+--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_dpipe.c |
10742 |
++++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_dpipe.c |
10743 |
+@@ -215,7 +215,7 @@ mlxsw_sp_dpipe_table_erif_entries_dump(void *priv, bool counters_enabled, |
10744 |
+ start_again: |
10745 |
+ err = devlink_dpipe_entry_ctx_prepare(dump_ctx); |
10746 |
+ if (err) |
10747 |
+- return err; |
10748 |
++ goto err_ctx_prepare; |
10749 |
+ j = 0; |
10750 |
+ for (; i < rif_count; i++) { |
10751 |
+ struct mlxsw_sp_rif *rif = mlxsw_sp_rif_by_index(mlxsw_sp, i); |
10752 |
+@@ -247,6 +247,7 @@ start_again: |
10753 |
+ return 0; |
10754 |
+ err_entry_append: |
10755 |
+ err_entry_get: |
10756 |
++err_ctx_prepare: |
10757 |
+ rtnl_unlock(); |
10758 |
+ devlink_dpipe_entry_clear(&entry); |
10759 |
+ return err; |
10760 |
+diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c |
10761 |
+index 4ab87fe84542..6ea43e48d5f9 100644 |
10762 |
+--- a/drivers/net/ethernet/realtek/r8169.c |
10763 |
++++ b/drivers/net/ethernet/realtek/r8169.c |
10764 |
+@@ -7433,6 +7433,15 @@ static int rtl_init_one(struct pci_dev *pdev, const struct pci_device_id *ent) |
10765 |
+ int chipset, region, i; |
10766 |
+ int jumbo_max, rc; |
10767 |
+ |
10768 |
++ /* Some tools for creating an initramfs don't consider softdeps, then |
10769 |
++ * r8169.ko may be in initramfs, but realtek.ko not. Then the generic |
10770 |
++ * PHY driver is used that doesn't work with most chip versions. |
10771 |
++ */ |
10772 |
++ if (!driver_find("RTL8201CP Ethernet", &mdio_bus_type)) { |
10773 |
++ dev_err(&pdev->dev, "realtek.ko not loaded, maybe it needs to be added to initramfs?\n"); |
10774 |
++ return -ENOENT; |
10775 |
++ } |
10776 |
++ |
10777 |
+ dev = devm_alloc_etherdev(&pdev->dev, sizeof (*tp)); |
10778 |
+ if (!dev) |
10779 |
+ return -ENOMEM; |
10780 |
+diff --git a/drivers/net/wan/fsl_ucc_hdlc.c b/drivers/net/wan/fsl_ucc_hdlc.c |
10781 |
+index daeab33f623e..9ab04ef532f3 100644 |
10782 |
+--- a/drivers/net/wan/fsl_ucc_hdlc.c |
10783 |
++++ b/drivers/net/wan/fsl_ucc_hdlc.c |
10784 |
+@@ -242,6 +242,11 @@ static int uhdlc_init(struct ucc_hdlc_private *priv) |
10785 |
+ ret = -ENOMEM; |
10786 |
+ goto free_riptr; |
10787 |
+ } |
10788 |
++ if (riptr != (u16)riptr || tiptr != (u16)tiptr) { |
10789 |
++ dev_err(priv->dev, "MURAM allocation out of addressable range\n"); |
10790 |
++ ret = -ENOMEM; |
10791 |
++ goto free_tiptr; |
10792 |
++ } |
10793 |
+ |
10794 |
+ /* Set RIPTR, TIPTR */ |
10795 |
+ iowrite16be(riptr, &priv->ucc_pram->riptr); |
10796 |
+diff --git a/drivers/net/wan/ixp4xx_hss.c b/drivers/net/wan/ixp4xx_hss.c |
10797 |
+index 6a505c26a3e7..a269ed63d90f 100644 |
10798 |
+--- a/drivers/net/wan/ixp4xx_hss.c |
10799 |
++++ b/drivers/net/wan/ixp4xx_hss.c |
10800 |
+@@ -261,7 +261,7 @@ struct port { |
10801 |
+ struct hss_plat_info *plat; |
10802 |
+ buffer_t *rx_buff_tab[RX_DESCS], *tx_buff_tab[TX_DESCS]; |
10803 |
+ struct desc *desc_tab; /* coherent */ |
10804 |
+- u32 desc_tab_phys; |
10805 |
++ dma_addr_t desc_tab_phys; |
10806 |
+ unsigned int id; |
10807 |
+ unsigned int clock_type, clock_rate, loopback; |
10808 |
+ unsigned int initialized, carrier; |
10809 |
+@@ -861,7 +861,7 @@ static int hss_hdlc_xmit(struct sk_buff *skb, struct net_device *dev) |
10810 |
+ dev->stats.tx_dropped++; |
10811 |
+ return NETDEV_TX_OK; |
10812 |
+ } |
10813 |
+- memcpy_swab32(mem, (u32 *)((int)skb->data & ~3), bytes / 4); |
10814 |
++ memcpy_swab32(mem, (u32 *)((uintptr_t)skb->data & ~3), bytes / 4); |
10815 |
+ dev_kfree_skb(skb); |
10816 |
+ #endif |
10817 |
+ |
10818 |
+diff --git a/drivers/net/wireless/ath/ath10k/wmi.c b/drivers/net/wireless/ath/ath10k/wmi.c |
10819 |
+index 0f6ff7a78e49..3372dfa0decc 100644 |
10820 |
+--- a/drivers/net/wireless/ath/ath10k/wmi.c |
10821 |
++++ b/drivers/net/wireless/ath/ath10k/wmi.c |
10822 |
+@@ -9193,7 +9193,7 @@ static int ath10k_wmi_mgmt_tx_clean_up_pending(int msdu_id, void *ptr, |
10823 |
+ |
10824 |
+ msdu = pkt_addr->vaddr; |
10825 |
+ dma_unmap_single(ar->dev, pkt_addr->paddr, |
10826 |
+- msdu->len, DMA_FROM_DEVICE); |
10827 |
++ msdu->len, DMA_TO_DEVICE); |
10828 |
+ ieee80211_free_txskb(ar->hw, msdu); |
10829 |
+ |
10830 |
+ return 0; |
10831 |
+diff --git a/drivers/net/wireless/broadcom/b43legacy/main.c b/drivers/net/wireless/broadcom/b43legacy/main.c |
10832 |
+index 55f411925960..770cc218ca4b 100644 |
10833 |
+--- a/drivers/net/wireless/broadcom/b43legacy/main.c |
10834 |
++++ b/drivers/net/wireless/broadcom/b43legacy/main.c |
10835 |
+@@ -1304,8 +1304,9 @@ static void handle_irq_ucode_debug(struct b43legacy_wldev *dev) |
10836 |
+ } |
10837 |
+ |
10838 |
+ /* Interrupt handler bottom-half */ |
10839 |
+-static void b43legacy_interrupt_tasklet(struct b43legacy_wldev *dev) |
10840 |
++static void b43legacy_interrupt_tasklet(unsigned long data) |
10841 |
+ { |
10842 |
++ struct b43legacy_wldev *dev = (struct b43legacy_wldev *)data; |
10843 |
+ u32 reason; |
10844 |
+ u32 dma_reason[ARRAY_SIZE(dev->dma_reason)]; |
10845 |
+ u32 merged_dma_reason = 0; |
10846 |
+@@ -3775,7 +3776,7 @@ static int b43legacy_one_core_attach(struct ssb_device *dev, |
10847 |
+ b43legacy_set_status(wldev, B43legacy_STAT_UNINIT); |
10848 |
+ wldev->bad_frames_preempt = modparam_bad_frames_preempt; |
10849 |
+ tasklet_init(&wldev->isr_tasklet, |
10850 |
+- (void (*)(unsigned long))b43legacy_interrupt_tasklet, |
10851 |
++ b43legacy_interrupt_tasklet, |
10852 |
+ (unsigned long)wldev); |
10853 |
+ if (modparam_pio) |
10854 |
+ wldev->__using_pio = true; |
10855 |
+diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c |
10856 |
+index 5c3b62e61980..e0211321fe9e 100644 |
10857 |
+--- a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c |
10858 |
++++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c |
10859 |
+@@ -1934,6 +1934,7 @@ static uint brcmf_sdio_readframes(struct brcmf_sdio *bus, uint maxframes) |
10860 |
+ BRCMF_SDIO_FT_NORMAL)) { |
10861 |
+ rd->len = 0; |
10862 |
+ brcmu_pkt_buf_free_skb(pkt); |
10863 |
++ continue; |
10864 |
+ } |
10865 |
+ bus->sdcnt.rx_readahead_cnt++; |
10866 |
+ if (rd->len != roundup(rd_new.len, 16)) { |
10867 |
+diff --git a/drivers/net/wireless/intel/ipw2x00/ipw2100.c b/drivers/net/wireless/intel/ipw2x00/ipw2100.c |
10868 |
+index 910db46db6a1..a3a470976a5c 100644 |
10869 |
+--- a/drivers/net/wireless/intel/ipw2x00/ipw2100.c |
10870 |
++++ b/drivers/net/wireless/intel/ipw2x00/ipw2100.c |
10871 |
+@@ -3220,8 +3220,9 @@ static void ipw2100_tx_send_data(struct ipw2100_priv *priv) |
10872 |
+ } |
10873 |
+ } |
10874 |
+ |
10875 |
+-static void ipw2100_irq_tasklet(struct ipw2100_priv *priv) |
10876 |
++static void ipw2100_irq_tasklet(unsigned long data) |
10877 |
+ { |
10878 |
++ struct ipw2100_priv *priv = (struct ipw2100_priv *)data; |
10879 |
+ struct net_device *dev = priv->net_dev; |
10880 |
+ unsigned long flags; |
10881 |
+ u32 inta, tmp; |
10882 |
+@@ -6025,7 +6026,7 @@ static void ipw2100_rf_kill(struct work_struct *work) |
10883 |
+ spin_unlock_irqrestore(&priv->low_lock, flags); |
10884 |
+ } |
10885 |
+ |
10886 |
+-static void ipw2100_irq_tasklet(struct ipw2100_priv *priv); |
10887 |
++static void ipw2100_irq_tasklet(unsigned long data); |
10888 |
+ |
10889 |
+ static const struct net_device_ops ipw2100_netdev_ops = { |
10890 |
+ .ndo_open = ipw2100_open, |
10891 |
+@@ -6155,7 +6156,7 @@ static struct net_device *ipw2100_alloc_device(struct pci_dev *pci_dev, |
10892 |
+ INIT_DELAYED_WORK(&priv->rf_kill, ipw2100_rf_kill); |
10893 |
+ INIT_DELAYED_WORK(&priv->scan_event, ipw2100_scan_event); |
10894 |
+ |
10895 |
+- tasklet_init(&priv->irq_tasklet, (void (*)(unsigned long)) |
10896 |
++ tasklet_init(&priv->irq_tasklet, |
10897 |
+ ipw2100_irq_tasklet, (unsigned long)priv); |
10898 |
+ |
10899 |
+ /* NOTE: We do not start the deferred work for status checks yet */ |
10900 |
+diff --git a/drivers/net/wireless/intel/ipw2x00/ipw2200.c b/drivers/net/wireless/intel/ipw2x00/ipw2200.c |
10901 |
+index 9644e7b93645..04aee2fdba37 100644 |
10902 |
+--- a/drivers/net/wireless/intel/ipw2x00/ipw2200.c |
10903 |
++++ b/drivers/net/wireless/intel/ipw2x00/ipw2200.c |
10904 |
+@@ -1959,8 +1959,9 @@ static void notify_wx_assoc_event(struct ipw_priv *priv) |
10905 |
+ wireless_send_event(priv->net_dev, SIOCGIWAP, &wrqu, NULL); |
10906 |
+ } |
10907 |
+ |
10908 |
+-static void ipw_irq_tasklet(struct ipw_priv *priv) |
10909 |
++static void ipw_irq_tasklet(unsigned long data) |
10910 |
+ { |
10911 |
++ struct ipw_priv *priv = (struct ipw_priv *)data; |
10912 |
+ u32 inta, inta_mask, handled = 0; |
10913 |
+ unsigned long flags; |
10914 |
+ int rc = 0; |
10915 |
+@@ -10694,7 +10695,7 @@ static int ipw_setup_deferred_work(struct ipw_priv *priv) |
10916 |
+ INIT_WORK(&priv->qos_activate, ipw_bg_qos_activate); |
10917 |
+ #endif /* CONFIG_IPW2200_QOS */ |
10918 |
+ |
10919 |
+- tasklet_init(&priv->irq_tasklet, (void (*)(unsigned long)) |
10920 |
++ tasklet_init(&priv->irq_tasklet, |
10921 |
+ ipw_irq_tasklet, (unsigned long)priv); |
10922 |
+ |
10923 |
+ return ret; |
10924 |
+diff --git a/drivers/net/wireless/intel/iwlegacy/3945-mac.c b/drivers/net/wireless/intel/iwlegacy/3945-mac.c |
10925 |
+index 57e3b6cca234..b536ec20eacc 100644 |
10926 |
+--- a/drivers/net/wireless/intel/iwlegacy/3945-mac.c |
10927 |
++++ b/drivers/net/wireless/intel/iwlegacy/3945-mac.c |
10928 |
+@@ -1392,8 +1392,9 @@ il3945_dump_nic_error_log(struct il_priv *il) |
10929 |
+ } |
10930 |
+ |
10931 |
+ static void |
10932 |
+-il3945_irq_tasklet(struct il_priv *il) |
10933 |
++il3945_irq_tasklet(unsigned long data) |
10934 |
+ { |
10935 |
++ struct il_priv *il = (struct il_priv *)data; |
10936 |
+ u32 inta, handled = 0; |
10937 |
+ u32 inta_fh; |
10938 |
+ unsigned long flags; |
10939 |
+@@ -3419,7 +3420,7 @@ il3945_setup_deferred_work(struct il_priv *il) |
10940 |
+ timer_setup(&il->watchdog, il_bg_watchdog, 0); |
10941 |
+ |
10942 |
+ tasklet_init(&il->irq_tasklet, |
10943 |
+- (void (*)(unsigned long))il3945_irq_tasklet, |
10944 |
++ il3945_irq_tasklet, |
10945 |
+ (unsigned long)il); |
10946 |
+ } |
10947 |
+ |
10948 |
+diff --git a/drivers/net/wireless/intel/iwlegacy/4965-mac.c b/drivers/net/wireless/intel/iwlegacy/4965-mac.c |
10949 |
+index 280cd8ae1696..6fc51c74cdb8 100644 |
10950 |
+--- a/drivers/net/wireless/intel/iwlegacy/4965-mac.c |
10951 |
++++ b/drivers/net/wireless/intel/iwlegacy/4965-mac.c |
10952 |
+@@ -4360,8 +4360,9 @@ il4965_synchronize_irq(struct il_priv *il) |
10953 |
+ } |
10954 |
+ |
10955 |
+ static void |
10956 |
+-il4965_irq_tasklet(struct il_priv *il) |
10957 |
++il4965_irq_tasklet(unsigned long data) |
10958 |
+ { |
10959 |
++ struct il_priv *il = (struct il_priv *)data; |
10960 |
+ u32 inta, handled = 0; |
10961 |
+ u32 inta_fh; |
10962 |
+ unsigned long flags; |
10963 |
+@@ -6257,7 +6258,7 @@ il4965_setup_deferred_work(struct il_priv *il) |
10964 |
+ timer_setup(&il->watchdog, il_bg_watchdog, 0); |
10965 |
+ |
10966 |
+ tasklet_init(&il->irq_tasklet, |
10967 |
+- (void (*)(unsigned long))il4965_irq_tasklet, |
10968 |
++ il4965_irq_tasklet, |
10969 |
+ (unsigned long)il); |
10970 |
+ } |
10971 |
+ |
10972 |
+diff --git a/drivers/net/wireless/intel/iwlegacy/common.c b/drivers/net/wireless/intel/iwlegacy/common.c |
10973 |
+index 6514baf799fe..e16f2597c219 100644 |
10974 |
+--- a/drivers/net/wireless/intel/iwlegacy/common.c |
10975 |
++++ b/drivers/net/wireless/intel/iwlegacy/common.c |
10976 |
+@@ -717,7 +717,7 @@ il_eeprom_init(struct il_priv *il) |
10977 |
+ u32 gp = _il_rd(il, CSR_EEPROM_GP); |
10978 |
+ int sz; |
10979 |
+ int ret; |
10980 |
+- u16 addr; |
10981 |
++ int addr; |
10982 |
+ |
10983 |
+ /* allocate eeprom */ |
10984 |
+ sz = il->cfg->eeprom_size; |
10985 |
+diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/tt.c b/drivers/net/wireless/intel/iwlwifi/mvm/tt.c |
10986 |
+index 1232f63278eb..319103f4b432 100644 |
10987 |
+--- a/drivers/net/wireless/intel/iwlwifi/mvm/tt.c |
10988 |
++++ b/drivers/net/wireless/intel/iwlwifi/mvm/tt.c |
10989 |
+@@ -739,7 +739,8 @@ static struct thermal_zone_device_ops tzone_ops = { |
10990 |
+ static void iwl_mvm_thermal_zone_register(struct iwl_mvm *mvm) |
10991 |
+ { |
10992 |
+ int i; |
10993 |
+- char name[] = "iwlwifi"; |
10994 |
++ char name[16]; |
10995 |
++ static atomic_t counter = ATOMIC_INIT(0); |
10996 |
+ |
10997 |
+ if (!iwl_mvm_is_tt_in_fw(mvm)) { |
10998 |
+ mvm->tz_device.tzone = NULL; |
10999 |
+@@ -749,6 +750,7 @@ static void iwl_mvm_thermal_zone_register(struct iwl_mvm *mvm) |
11000 |
+ |
11001 |
+ BUILD_BUG_ON(ARRAY_SIZE(name) >= THERMAL_NAME_LENGTH); |
11002 |
+ |
11003 |
++ sprintf(name, "iwlwifi_%u", atomic_inc_return(&counter) & 0xFF); |
11004 |
+ mvm->tz_device.tzone = thermal_zone_device_register(name, |
11005 |
+ IWL_MAX_DTS_TRIPS, |
11006 |
+ IWL_WRITABLE_TRIPS_MSK, |
11007 |
+diff --git a/drivers/net/wireless/intersil/hostap/hostap_ap.c b/drivers/net/wireless/intersil/hostap/hostap_ap.c |
11008 |
+index 0094b1d2b577..3ec46f48cfde 100644 |
11009 |
+--- a/drivers/net/wireless/intersil/hostap/hostap_ap.c |
11010 |
++++ b/drivers/net/wireless/intersil/hostap/hostap_ap.c |
11011 |
+@@ -2508,7 +2508,7 @@ static int prism2_hostapd_add_sta(struct ap_data *ap, |
11012 |
+ sta->supported_rates[0] = 2; |
11013 |
+ if (sta->tx_supp_rates & WLAN_RATE_2M) |
11014 |
+ sta->supported_rates[1] = 4; |
11015 |
+- if (sta->tx_supp_rates & WLAN_RATE_5M5) |
11016 |
++ if (sta->tx_supp_rates & WLAN_RATE_5M5) |
11017 |
+ sta->supported_rates[2] = 11; |
11018 |
+ if (sta->tx_supp_rates & WLAN_RATE_11M) |
11019 |
+ sta->supported_rates[3] = 22; |
11020 |
+diff --git a/drivers/net/wireless/intersil/orinoco/orinoco_usb.c b/drivers/net/wireless/intersil/orinoco/orinoco_usb.c |
11021 |
+index 2c7dd2a7350c..b704e4bce171 100644 |
11022 |
+--- a/drivers/net/wireless/intersil/orinoco/orinoco_usb.c |
11023 |
++++ b/drivers/net/wireless/intersil/orinoco/orinoco_usb.c |
11024 |
+@@ -1364,7 +1364,8 @@ static int ezusb_init(struct hermes *hw) |
11025 |
+ int retval; |
11026 |
+ |
11027 |
+ BUG_ON(in_interrupt()); |
11028 |
+- BUG_ON(!upriv); |
11029 |
++ if (!upriv) |
11030 |
++ return -EINVAL; |
11031 |
+ |
11032 |
+ upriv->reply_count = 0; |
11033 |
+ /* Write the MAGIC number on the simulated registers to keep |
11034 |
+diff --git a/drivers/net/wireless/realtek/rtlwifi/pci.c b/drivers/net/wireless/realtek/rtlwifi/pci.c |
11035 |
+index 5d1fda16fc8c..83749578fa8b 100644 |
11036 |
+--- a/drivers/net/wireless/realtek/rtlwifi/pci.c |
11037 |
++++ b/drivers/net/wireless/realtek/rtlwifi/pci.c |
11038 |
+@@ -1082,13 +1082,15 @@ done: |
11039 |
+ return ret; |
11040 |
+ } |
11041 |
+ |
11042 |
+-static void _rtl_pci_irq_tasklet(struct ieee80211_hw *hw) |
11043 |
++static void _rtl_pci_irq_tasklet(unsigned long data) |
11044 |
+ { |
11045 |
++ struct ieee80211_hw *hw = (struct ieee80211_hw *)data; |
11046 |
+ _rtl_pci_tx_chk_waitq(hw); |
11047 |
+ } |
11048 |
+ |
11049 |
+-static void _rtl_pci_prepare_bcn_tasklet(struct ieee80211_hw *hw) |
11050 |
++static void _rtl_pci_prepare_bcn_tasklet(unsigned long data) |
11051 |
+ { |
11052 |
++ struct ieee80211_hw *hw = (struct ieee80211_hw *)data; |
11053 |
+ struct rtl_priv *rtlpriv = rtl_priv(hw); |
11054 |
+ struct rtl_pci *rtlpci = rtl_pcidev(rtl_pcipriv(hw)); |
11055 |
+ struct rtl_mac *mac = rtl_mac(rtl_priv(hw)); |
11056 |
+@@ -1214,10 +1216,10 @@ static void _rtl_pci_init_struct(struct ieee80211_hw *hw, |
11057 |
+ |
11058 |
+ /*task */ |
11059 |
+ tasklet_init(&rtlpriv->works.irq_tasklet, |
11060 |
+- (void (*)(unsigned long))_rtl_pci_irq_tasklet, |
11061 |
++ _rtl_pci_irq_tasklet, |
11062 |
+ (unsigned long)hw); |
11063 |
+ tasklet_init(&rtlpriv->works.irq_prepare_bcn_tasklet, |
11064 |
+- (void (*)(unsigned long))_rtl_pci_prepare_bcn_tasklet, |
11065 |
++ _rtl_pci_prepare_bcn_tasklet, |
11066 |
+ (unsigned long)hw); |
11067 |
+ INIT_WORK(&rtlpriv->works.lps_change_work, |
11068 |
+ rtl_lps_change_work_callback); |
11069 |
+diff --git a/drivers/nfc/port100.c b/drivers/nfc/port100.c |
11070 |
+index 60ae382f50da..06bb226c62ef 100644 |
11071 |
+--- a/drivers/nfc/port100.c |
11072 |
++++ b/drivers/nfc/port100.c |
11073 |
+@@ -574,7 +574,7 @@ static void port100_tx_update_payload_len(void *_frame, int len) |
11074 |
+ { |
11075 |
+ struct port100_frame *frame = _frame; |
11076 |
+ |
11077 |
+- frame->datalen = cpu_to_le16(le16_to_cpu(frame->datalen) + len); |
11078 |
++ le16_add_cpu(&frame->datalen, len); |
11079 |
+ } |
11080 |
+ |
11081 |
+ static bool port100_rx_frame_is_valid(void *_frame) |
11082 |
+diff --git a/drivers/pci/controller/pcie-iproc.c b/drivers/pci/controller/pcie-iproc.c |
11083 |
+index 9d5cbc75d5ae..ec86414216f9 100644 |
11084 |
+--- a/drivers/pci/controller/pcie-iproc.c |
11085 |
++++ b/drivers/pci/controller/pcie-iproc.c |
11086 |
+@@ -1526,6 +1526,30 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd802, |
11087 |
+ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd804, |
11088 |
+ quirk_paxc_disable_msi_parsing); |
11089 |
+ |
11090 |
++static void quirk_paxc_bridge(struct pci_dev *pdev) |
11091 |
++{ |
11092 |
++ /* |
11093 |
++ * The PCI config space is shared with the PAXC root port and the first |
11094 |
++ * Ethernet device. So, we need to workaround this by telling the PCI |
11095 |
++ * code that the bridge is not an Ethernet device. |
11096 |
++ */ |
11097 |
++ if (pdev->hdr_type == PCI_HEADER_TYPE_BRIDGE) |
11098 |
++ pdev->class = PCI_CLASS_BRIDGE_PCI << 8; |
11099 |
++ |
11100 |
++ /* |
11101 |
++ * MPSS is not being set properly (as it is currently 0). This is |
11102 |
++ * because that area of the PCI config space is hard coded to zero, and |
11103 |
++ * is not modifiable by firmware. Set this to 2 (e.g., 512 byte MPS) |
11104 |
++ * so that the MPS can be set to the real max value. |
11105 |
++ */ |
11106 |
++ pdev->pcie_mpss = 2; |
11107 |
++} |
11108 |
++DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0x16cd, quirk_paxc_bridge); |
11109 |
++DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0x16f0, quirk_paxc_bridge); |
11110 |
++DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd750, quirk_paxc_bridge); |
11111 |
++DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd802, quirk_paxc_bridge); |
11112 |
++DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd804, quirk_paxc_bridge); |
11113 |
++ |
11114 |
+ MODULE_AUTHOR("Ray Jui <rjui@××××××××.com>"); |
11115 |
+ MODULE_DESCRIPTION("Broadcom iPROC PCIe common driver"); |
11116 |
+ MODULE_LICENSE("GPL v2"); |
11117 |
+diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c |
11118 |
+index 5b4c36ab1596..419dda6dbd16 100644 |
11119 |
+--- a/drivers/pci/quirks.c |
11120 |
++++ b/drivers/pci/quirks.c |
11121 |
+@@ -1848,19 +1848,40 @@ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x2609, quirk_intel_pcie_pm); |
11122 |
+ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x260a, quirk_intel_pcie_pm); |
11123 |
+ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_INTEL, 0x260b, quirk_intel_pcie_pm); |
11124 |
+ |
11125 |
++static void quirk_d3hot_delay(struct pci_dev *dev, unsigned int delay) |
11126 |
++{ |
11127 |
++ if (dev->d3_delay >= delay) |
11128 |
++ return; |
11129 |
++ |
11130 |
++ dev->d3_delay = delay; |
11131 |
++ pci_info(dev, "extending delay after power-on from D3hot to %d msec\n", |
11132 |
++ dev->d3_delay); |
11133 |
++} |
11134 |
++ |
11135 |
+ static void quirk_radeon_pm(struct pci_dev *dev) |
11136 |
+ { |
11137 |
+ if (dev->subsystem_vendor == PCI_VENDOR_ID_APPLE && |
11138 |
+- dev->subsystem_device == 0x00e2) { |
11139 |
+- if (dev->d3_delay < 20) { |
11140 |
+- dev->d3_delay = 20; |
11141 |
+- pci_info(dev, "extending delay after power-on from D3 to %d msec\n", |
11142 |
+- dev->d3_delay); |
11143 |
+- } |
11144 |
+- } |
11145 |
++ dev->subsystem_device == 0x00e2) |
11146 |
++ quirk_d3hot_delay(dev, 20); |
11147 |
+ } |
11148 |
+ DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATI, 0x6741, quirk_radeon_pm); |
11149 |
+ |
11150 |
++/* |
11151 |
++ * Ryzen5/7 XHCI controllers fail upon resume from runtime suspend or s2idle. |
11152 |
++ * https://bugzilla.kernel.org/show_bug.cgi?id=205587 |
11153 |
++ * |
11154 |
++ * The kernel attempts to transition these devices to D3cold, but that seems |
11155 |
++ * to be ineffective on the platforms in question; the PCI device appears to |
11156 |
++ * remain on in D3hot state. The D3hot-to-D0 transition then requires an |
11157 |
++ * extended delay in order to succeed. |
11158 |
++ */ |
11159 |
++static void quirk_ryzen_xhci_d3hot(struct pci_dev *dev) |
11160 |
++{ |
11161 |
++ quirk_d3hot_delay(dev, 20); |
11162 |
++} |
11163 |
++DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, 0x15e0, quirk_ryzen_xhci_d3hot); |
11164 |
++DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_AMD, 0x15e1, quirk_ryzen_xhci_d3hot); |
11165 |
++ |
11166 |
+ #ifdef CONFIG_X86_IO_APIC |
11167 |
+ static int dmi_disable_ioapicreroute(const struct dmi_system_id *d) |
11168 |
+ { |
11169 |
+@@ -2358,32 +2379,6 @@ DECLARE_PCI_FIXUP_ENABLE(PCI_VENDOR_ID_BROADCOM, |
11170 |
+ PCI_DEVICE_ID_TIGON3_5719, |
11171 |
+ quirk_brcm_5719_limit_mrrs); |
11172 |
+ |
11173 |
+-#ifdef CONFIG_PCIE_IPROC_PLATFORM |
11174 |
+-static void quirk_paxc_bridge(struct pci_dev *pdev) |
11175 |
+-{ |
11176 |
+- /* |
11177 |
+- * The PCI config space is shared with the PAXC root port and the first |
11178 |
+- * Ethernet device. So, we need to workaround this by telling the PCI |
11179 |
+- * code that the bridge is not an Ethernet device. |
11180 |
+- */ |
11181 |
+- if (pdev->hdr_type == PCI_HEADER_TYPE_BRIDGE) |
11182 |
+- pdev->class = PCI_CLASS_BRIDGE_PCI << 8; |
11183 |
+- |
11184 |
+- /* |
11185 |
+- * MPSS is not being set properly (as it is currently 0). This is |
11186 |
+- * because that area of the PCI config space is hard coded to zero, and |
11187 |
+- * is not modifiable by firmware. Set this to 2 (e.g., 512 byte MPS) |
11188 |
+- * so that the MPS can be set to the real max value. |
11189 |
+- */ |
11190 |
+- pdev->pcie_mpss = 2; |
11191 |
+-} |
11192 |
+-DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0x16cd, quirk_paxc_bridge); |
11193 |
+-DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0x16f0, quirk_paxc_bridge); |
11194 |
+-DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd750, quirk_paxc_bridge); |
11195 |
+-DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd802, quirk_paxc_bridge); |
11196 |
+-DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_BROADCOM, 0xd804, quirk_paxc_bridge); |
11197 |
+-#endif |
11198 |
+- |
11199 |
+ /* |
11200 |
+ * Originally in EDAC sources for i82875P: Intel tells BIOS developers to |
11201 |
+ * hide device 6 which configures the overflow device access containing the |
11202 |
+diff --git a/drivers/pinctrl/intel/pinctrl-baytrail.c b/drivers/pinctrl/intel/pinctrl-baytrail.c |
11203 |
+index 021e28ff1194..a760d8bda0af 100644 |
11204 |
+--- a/drivers/pinctrl/intel/pinctrl-baytrail.c |
11205 |
++++ b/drivers/pinctrl/intel/pinctrl-baytrail.c |
11206 |
+@@ -950,7 +950,13 @@ static void byt_gpio_clear_triggering(struct byt_gpio *vg, unsigned int offset) |
11207 |
+ |
11208 |
+ raw_spin_lock_irqsave(&byt_lock, flags); |
11209 |
+ value = readl(reg); |
11210 |
+- value &= ~(BYT_TRIG_POS | BYT_TRIG_NEG | BYT_TRIG_LVL); |
11211 |
++ |
11212 |
++ /* Do not clear direct-irq enabled IRQs (from gpio_disable_free) */ |
11213 |
++ if (value & BYT_DIRECT_IRQ_EN) |
11214 |
++ /* nothing to do */ ; |
11215 |
++ else |
11216 |
++ value &= ~(BYT_TRIG_POS | BYT_TRIG_NEG | BYT_TRIG_LVL); |
11217 |
++ |
11218 |
+ writel(value, reg); |
11219 |
+ raw_spin_unlock_irqrestore(&byt_lock, flags); |
11220 |
+ } |
11221 |
+diff --git a/drivers/pinctrl/sh-pfc/pfc-sh7264.c b/drivers/pinctrl/sh-pfc/pfc-sh7264.c |
11222 |
+index e1c34e19222e..3ddb9565ed80 100644 |
11223 |
+--- a/drivers/pinctrl/sh-pfc/pfc-sh7264.c |
11224 |
++++ b/drivers/pinctrl/sh-pfc/pfc-sh7264.c |
11225 |
+@@ -500,17 +500,15 @@ enum { |
11226 |
+ SD_WP_MARK, SD_CLK_MARK, SD_CMD_MARK, |
11227 |
+ CRX0_MARK, CRX1_MARK, |
11228 |
+ CTX0_MARK, CTX1_MARK, |
11229 |
++ CRX0_CRX1_MARK, CTX0_CTX1_MARK, |
11230 |
+ |
11231 |
+ PWM1A_MARK, PWM1B_MARK, PWM1C_MARK, PWM1D_MARK, |
11232 |
+ PWM1E_MARK, PWM1F_MARK, PWM1G_MARK, PWM1H_MARK, |
11233 |
+ PWM2A_MARK, PWM2B_MARK, PWM2C_MARK, PWM2D_MARK, |
11234 |
+ PWM2E_MARK, PWM2F_MARK, PWM2G_MARK, PWM2H_MARK, |
11235 |
+ IERXD_MARK, IETXD_MARK, |
11236 |
+- CRX0_CRX1_MARK, |
11237 |
+ WDTOVF_MARK, |
11238 |
+ |
11239 |
+- CRX0X1_MARK, |
11240 |
+- |
11241 |
+ /* DMAC */ |
11242 |
+ TEND0_MARK, DACK0_MARK, DREQ0_MARK, |
11243 |
+ TEND1_MARK, DACK1_MARK, DREQ1_MARK, |
11244 |
+@@ -998,12 +996,12 @@ static const u16 pinmux_data[] = { |
11245 |
+ |
11246 |
+ PINMUX_DATA(PJ3_DATA, PJ3MD_00), |
11247 |
+ PINMUX_DATA(CRX1_MARK, PJ3MD_01), |
11248 |
+- PINMUX_DATA(CRX0X1_MARK, PJ3MD_10), |
11249 |
++ PINMUX_DATA(CRX0_CRX1_MARK, PJ3MD_10), |
11250 |
+ PINMUX_DATA(IRQ1_PJ_MARK, PJ3MD_11), |
11251 |
+ |
11252 |
+ PINMUX_DATA(PJ2_DATA, PJ2MD_000), |
11253 |
+ PINMUX_DATA(CTX1_MARK, PJ2MD_001), |
11254 |
+- PINMUX_DATA(CRX0_CRX1_MARK, PJ2MD_010), |
11255 |
++ PINMUX_DATA(CTX0_CTX1_MARK, PJ2MD_010), |
11256 |
+ PINMUX_DATA(CS2_MARK, PJ2MD_011), |
11257 |
+ PINMUX_DATA(SCK0_MARK, PJ2MD_100), |
11258 |
+ PINMUX_DATA(LCD_M_DISP_MARK, PJ2MD_101), |
11259 |
+@@ -1248,6 +1246,7 @@ static const struct pinmux_func pinmux_func_gpios[] = { |
11260 |
+ GPIO_FN(CTX1), |
11261 |
+ GPIO_FN(CRX1), |
11262 |
+ GPIO_FN(CTX0), |
11263 |
++ GPIO_FN(CTX0_CTX1), |
11264 |
+ GPIO_FN(CRX0), |
11265 |
+ GPIO_FN(CRX0_CRX1), |
11266 |
+ |
11267 |
+diff --git a/drivers/pinctrl/sh-pfc/pfc-sh7269.c b/drivers/pinctrl/sh-pfc/pfc-sh7269.c |
11268 |
+index cfdb4fc177c3..3df0c0d139d0 100644 |
11269 |
+--- a/drivers/pinctrl/sh-pfc/pfc-sh7269.c |
11270 |
++++ b/drivers/pinctrl/sh-pfc/pfc-sh7269.c |
11271 |
+@@ -740,13 +740,12 @@ enum { |
11272 |
+ CRX0_MARK, CTX0_MARK, |
11273 |
+ CRX1_MARK, CTX1_MARK, |
11274 |
+ CRX2_MARK, CTX2_MARK, |
11275 |
+- CRX0_CRX1_MARK, |
11276 |
+- CRX0_CRX1_CRX2_MARK, |
11277 |
+- CTX0CTX1CTX2_MARK, |
11278 |
++ CRX0_CRX1_MARK, CTX0_CTX1_MARK, |
11279 |
++ CRX0_CRX1_CRX2_MARK, CTX0_CTX1_CTX2_MARK, |
11280 |
+ CRX1_PJ22_MARK, CTX1_PJ23_MARK, |
11281 |
+ CRX2_PJ20_MARK, CTX2_PJ21_MARK, |
11282 |
+- CRX0CRX1_PJ22_MARK, |
11283 |
+- CRX0CRX1CRX2_PJ20_MARK, |
11284 |
++ CRX0_CRX1_PJ22_MARK, CTX0_CTX1_PJ23_MARK, |
11285 |
++ CRX0_CRX1_CRX2_PJ20_MARK, CTX0_CTX1_CTX2_PJ21_MARK, |
11286 |
+ |
11287 |
+ /* VDC */ |
11288 |
+ DV_CLK_MARK, |
11289 |
+@@ -824,6 +823,7 @@ static const u16 pinmux_data[] = { |
11290 |
+ PINMUX_DATA(CS3_MARK, PC8MD_001), |
11291 |
+ PINMUX_DATA(TXD7_MARK, PC8MD_010), |
11292 |
+ PINMUX_DATA(CTX1_MARK, PC8MD_011), |
11293 |
++ PINMUX_DATA(CTX0_CTX1_MARK, PC8MD_100), |
11294 |
+ |
11295 |
+ PINMUX_DATA(PC7_DATA, PC7MD_000), |
11296 |
+ PINMUX_DATA(CKE_MARK, PC7MD_001), |
11297 |
+@@ -836,11 +836,12 @@ static const u16 pinmux_data[] = { |
11298 |
+ PINMUX_DATA(CAS_MARK, PC6MD_001), |
11299 |
+ PINMUX_DATA(SCK7_MARK, PC6MD_010), |
11300 |
+ PINMUX_DATA(CTX0_MARK, PC6MD_011), |
11301 |
++ PINMUX_DATA(CTX0_CTX1_CTX2_MARK, PC6MD_100), |
11302 |
+ |
11303 |
+ PINMUX_DATA(PC5_DATA, PC5MD_000), |
11304 |
+ PINMUX_DATA(RAS_MARK, PC5MD_001), |
11305 |
+ PINMUX_DATA(CRX0_MARK, PC5MD_011), |
11306 |
+- PINMUX_DATA(CTX0CTX1CTX2_MARK, PC5MD_100), |
11307 |
++ PINMUX_DATA(CTX0_CTX1_CTX2_MARK, PC5MD_100), |
11308 |
+ PINMUX_DATA(IRQ0_PC_MARK, PC5MD_101), |
11309 |
+ |
11310 |
+ PINMUX_DATA(PC4_DATA, PC4MD_00), |
11311 |
+@@ -1292,30 +1293,32 @@ static const u16 pinmux_data[] = { |
11312 |
+ PINMUX_DATA(LCD_DATA23_PJ23_MARK, PJ23MD_010), |
11313 |
+ PINMUX_DATA(LCD_TCON6_MARK, PJ23MD_011), |
11314 |
+ PINMUX_DATA(IRQ3_PJ_MARK, PJ23MD_100), |
11315 |
+- PINMUX_DATA(CTX1_MARK, PJ23MD_101), |
11316 |
++ PINMUX_DATA(CTX1_PJ23_MARK, PJ23MD_101), |
11317 |
++ PINMUX_DATA(CTX0_CTX1_PJ23_MARK, PJ23MD_110), |
11318 |
+ |
11319 |
+ PINMUX_DATA(PJ22_DATA, PJ22MD_000), |
11320 |
+ PINMUX_DATA(DV_DATA22_MARK, PJ22MD_001), |
11321 |
+ PINMUX_DATA(LCD_DATA22_PJ22_MARK, PJ22MD_010), |
11322 |
+ PINMUX_DATA(LCD_TCON5_MARK, PJ22MD_011), |
11323 |
+ PINMUX_DATA(IRQ2_PJ_MARK, PJ22MD_100), |
11324 |
+- PINMUX_DATA(CRX1_MARK, PJ22MD_101), |
11325 |
+- PINMUX_DATA(CRX0_CRX1_MARK, PJ22MD_110), |
11326 |
++ PINMUX_DATA(CRX1_PJ22_MARK, PJ22MD_101), |
11327 |
++ PINMUX_DATA(CRX0_CRX1_PJ22_MARK, PJ22MD_110), |
11328 |
+ |
11329 |
+ PINMUX_DATA(PJ21_DATA, PJ21MD_000), |
11330 |
+ PINMUX_DATA(DV_DATA21_MARK, PJ21MD_001), |
11331 |
+ PINMUX_DATA(LCD_DATA21_PJ21_MARK, PJ21MD_010), |
11332 |
+ PINMUX_DATA(LCD_TCON4_MARK, PJ21MD_011), |
11333 |
+ PINMUX_DATA(IRQ1_PJ_MARK, PJ21MD_100), |
11334 |
+- PINMUX_DATA(CTX2_MARK, PJ21MD_101), |
11335 |
++ PINMUX_DATA(CTX2_PJ21_MARK, PJ21MD_101), |
11336 |
++ PINMUX_DATA(CTX0_CTX1_CTX2_PJ21_MARK, PJ21MD_110), |
11337 |
+ |
11338 |
+ PINMUX_DATA(PJ20_DATA, PJ20MD_000), |
11339 |
+ PINMUX_DATA(DV_DATA20_MARK, PJ20MD_001), |
11340 |
+ PINMUX_DATA(LCD_DATA20_PJ20_MARK, PJ20MD_010), |
11341 |
+ PINMUX_DATA(LCD_TCON3_MARK, PJ20MD_011), |
11342 |
+ PINMUX_DATA(IRQ0_PJ_MARK, PJ20MD_100), |
11343 |
+- PINMUX_DATA(CRX2_MARK, PJ20MD_101), |
11344 |
+- PINMUX_DATA(CRX0CRX1CRX2_PJ20_MARK, PJ20MD_110), |
11345 |
++ PINMUX_DATA(CRX2_PJ20_MARK, PJ20MD_101), |
11346 |
++ PINMUX_DATA(CRX0_CRX1_CRX2_PJ20_MARK, PJ20MD_110), |
11347 |
+ |
11348 |
+ PINMUX_DATA(PJ19_DATA, PJ19MD_000), |
11349 |
+ PINMUX_DATA(DV_DATA19_MARK, PJ19MD_001), |
11350 |
+@@ -1666,12 +1669,24 @@ static const struct pinmux_func pinmux_func_gpios[] = { |
11351 |
+ GPIO_FN(WDTOVF), |
11352 |
+ |
11353 |
+ /* CAN */ |
11354 |
++ GPIO_FN(CTX2), |
11355 |
++ GPIO_FN(CRX2), |
11356 |
+ GPIO_FN(CTX1), |
11357 |
+ GPIO_FN(CRX1), |
11358 |
+ GPIO_FN(CTX0), |
11359 |
+ GPIO_FN(CRX0), |
11360 |
++ GPIO_FN(CTX0_CTX1), |
11361 |
+ GPIO_FN(CRX0_CRX1), |
11362 |
++ GPIO_FN(CTX0_CTX1_CTX2), |
11363 |
+ GPIO_FN(CRX0_CRX1_CRX2), |
11364 |
++ GPIO_FN(CTX2_PJ21), |
11365 |
++ GPIO_FN(CRX2_PJ20), |
11366 |
++ GPIO_FN(CTX1_PJ23), |
11367 |
++ GPIO_FN(CRX1_PJ22), |
11368 |
++ GPIO_FN(CTX0_CTX1_PJ23), |
11369 |
++ GPIO_FN(CRX0_CRX1_PJ22), |
11370 |
++ GPIO_FN(CTX0_CTX1_CTX2_PJ21), |
11371 |
++ GPIO_FN(CRX0_CRX1_CRX2_PJ20), |
11372 |
+ |
11373 |
+ /* DMAC */ |
11374 |
+ GPIO_FN(TEND0), |
11375 |
+diff --git a/drivers/pwm/pwm-omap-dmtimer.c b/drivers/pwm/pwm-omap-dmtimer.c |
11376 |
+index f45798679e3c..c6e710a713d3 100644 |
11377 |
+--- a/drivers/pwm/pwm-omap-dmtimer.c |
11378 |
++++ b/drivers/pwm/pwm-omap-dmtimer.c |
11379 |
+@@ -301,15 +301,10 @@ static int pwm_omap_dmtimer_probe(struct platform_device *pdev) |
11380 |
+ goto put; |
11381 |
+ } |
11382 |
+ |
11383 |
+-put: |
11384 |
+- of_node_put(timer); |
11385 |
+- if (ret < 0) |
11386 |
+- return ret; |
11387 |
+- |
11388 |
+ omap = devm_kzalloc(&pdev->dev, sizeof(*omap), GFP_KERNEL); |
11389 |
+ if (!omap) { |
11390 |
+- pdata->free(dm_timer); |
11391 |
+- return -ENOMEM; |
11392 |
++ ret = -ENOMEM; |
11393 |
++ goto err_alloc_omap; |
11394 |
+ } |
11395 |
+ |
11396 |
+ omap->pdata = pdata; |
11397 |
+@@ -342,18 +337,38 @@ put: |
11398 |
+ ret = pwmchip_add(&omap->chip); |
11399 |
+ if (ret < 0) { |
11400 |
+ dev_err(&pdev->dev, "failed to register PWM\n"); |
11401 |
+- omap->pdata->free(omap->dm_timer); |
11402 |
+- return ret; |
11403 |
++ goto err_pwmchip_add; |
11404 |
+ } |
11405 |
+ |
11406 |
++ of_node_put(timer); |
11407 |
++ |
11408 |
+ platform_set_drvdata(pdev, omap); |
11409 |
+ |
11410 |
+ return 0; |
11411 |
++ |
11412 |
++err_pwmchip_add: |
11413 |
++ |
11414 |
++ /* |
11415 |
++ * *omap is allocated using devm_kzalloc, |
11416 |
++ * so no free necessary here |
11417 |
++ */ |
11418 |
++err_alloc_omap: |
11419 |
++ |
11420 |
++ pdata->free(dm_timer); |
11421 |
++put: |
11422 |
++ of_node_put(timer); |
11423 |
++ |
11424 |
++ return ret; |
11425 |
+ } |
11426 |
+ |
11427 |
+ static int pwm_omap_dmtimer_remove(struct platform_device *pdev) |
11428 |
+ { |
11429 |
+ struct pwm_omap_dmtimer_chip *omap = platform_get_drvdata(pdev); |
11430 |
++ int ret; |
11431 |
++ |
11432 |
++ ret = pwmchip_remove(&omap->chip); |
11433 |
++ if (ret) |
11434 |
++ return ret; |
11435 |
+ |
11436 |
+ if (pm_runtime_active(&omap->dm_timer_pdev->dev)) |
11437 |
+ omap->pdata->stop(omap->dm_timer); |
11438 |
+@@ -362,7 +377,7 @@ static int pwm_omap_dmtimer_remove(struct platform_device *pdev) |
11439 |
+ |
11440 |
+ mutex_destroy(&omap->mutex); |
11441 |
+ |
11442 |
+- return pwmchip_remove(&omap->chip); |
11443 |
++ return 0; |
11444 |
+ } |
11445 |
+ |
11446 |
+ static const struct of_device_id pwm_omap_dmtimer_of_match[] = { |
11447 |
+diff --git a/drivers/pwm/pwm-pca9685.c b/drivers/pwm/pwm-pca9685.c |
11448 |
+index 567f5e2771c4..e1e5dfcb16f3 100644 |
11449 |
+--- a/drivers/pwm/pwm-pca9685.c |
11450 |
++++ b/drivers/pwm/pwm-pca9685.c |
11451 |
+@@ -170,13 +170,9 @@ static void pca9685_pwm_gpio_set(struct gpio_chip *gpio, unsigned int offset, |
11452 |
+ static void pca9685_pwm_gpio_free(struct gpio_chip *gpio, unsigned int offset) |
11453 |
+ { |
11454 |
+ struct pca9685 *pca = gpiochip_get_data(gpio); |
11455 |
+- struct pwm_device *pwm; |
11456 |
+ |
11457 |
+ pca9685_pwm_gpio_set(gpio, offset, 0); |
11458 |
+ pm_runtime_put(pca->chip.dev); |
11459 |
+- mutex_lock(&pca->lock); |
11460 |
+- pwm = &pca->chip.pwms[offset]; |
11461 |
+- mutex_unlock(&pca->lock); |
11462 |
+ } |
11463 |
+ |
11464 |
+ static int pca9685_pwm_gpio_get_direction(struct gpio_chip *chip, |
11465 |
+diff --git a/drivers/regulator/rk808-regulator.c b/drivers/regulator/rk808-regulator.c |
11466 |
+index 213b68743cc8..92498ac50303 100644 |
11467 |
+--- a/drivers/regulator/rk808-regulator.c |
11468 |
++++ b/drivers/regulator/rk808-regulator.c |
11469 |
+@@ -714,7 +714,7 @@ static int rk808_regulator_dt_parse_pdata(struct device *dev, |
11470 |
+ } |
11471 |
+ |
11472 |
+ if (!pdata->dvs_gpio[i]) { |
11473 |
+- dev_warn(dev, "there is no dvs%d gpio\n", i); |
11474 |
++ dev_info(dev, "there is no dvs%d gpio\n", i); |
11475 |
+ continue; |
11476 |
+ } |
11477 |
+ |
11478 |
+diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c |
11479 |
+index aa6206706fe3..abbef17c97ee 100644 |
11480 |
+--- a/drivers/remoteproc/remoteproc_core.c |
11481 |
++++ b/drivers/remoteproc/remoteproc_core.c |
11482 |
+@@ -1786,7 +1786,7 @@ static int __init remoteproc_init(void) |
11483 |
+ |
11484 |
+ return 0; |
11485 |
+ } |
11486 |
+-module_init(remoteproc_init); |
11487 |
++subsys_initcall(remoteproc_init); |
11488 |
+ |
11489 |
+ static void __exit remoteproc_exit(void) |
11490 |
+ { |
11491 |
+diff --git a/drivers/reset/reset-uniphier.c b/drivers/reset/reset-uniphier.c |
11492 |
+index 5605745663ae..adbecb2c7cd3 100644 |
11493 |
+--- a/drivers/reset/reset-uniphier.c |
11494 |
++++ b/drivers/reset/reset-uniphier.c |
11495 |
+@@ -202,8 +202,8 @@ static const struct uniphier_reset_data uniphier_pro5_sd_reset_data[] = { |
11496 |
+ #define UNIPHIER_PERI_RESET_FI2C(id, ch) \ |
11497 |
+ UNIPHIER_RESETX((id), 0x114, 24 + (ch)) |
11498 |
+ |
11499 |
+-#define UNIPHIER_PERI_RESET_SCSSI(id) \ |
11500 |
+- UNIPHIER_RESETX((id), 0x110, 17) |
11501 |
++#define UNIPHIER_PERI_RESET_SCSSI(id, ch) \ |
11502 |
++ UNIPHIER_RESETX((id), 0x110, 17 + (ch)) |
11503 |
+ |
11504 |
+ #define UNIPHIER_PERI_RESET_MCSSI(id) \ |
11505 |
+ UNIPHIER_RESETX((id), 0x114, 14) |
11506 |
+@@ -218,7 +218,7 @@ static const struct uniphier_reset_data uniphier_ld4_peri_reset_data[] = { |
11507 |
+ UNIPHIER_PERI_RESET_I2C(6, 2), |
11508 |
+ UNIPHIER_PERI_RESET_I2C(7, 3), |
11509 |
+ UNIPHIER_PERI_RESET_I2C(8, 4), |
11510 |
+- UNIPHIER_PERI_RESET_SCSSI(11), |
11511 |
++ UNIPHIER_PERI_RESET_SCSSI(11, 0), |
11512 |
+ UNIPHIER_RESET_END, |
11513 |
+ }; |
11514 |
+ |
11515 |
+@@ -234,8 +234,11 @@ static const struct uniphier_reset_data uniphier_pro4_peri_reset_data[] = { |
11516 |
+ UNIPHIER_PERI_RESET_FI2C(8, 4), |
11517 |
+ UNIPHIER_PERI_RESET_FI2C(9, 5), |
11518 |
+ UNIPHIER_PERI_RESET_FI2C(10, 6), |
11519 |
+- UNIPHIER_PERI_RESET_SCSSI(11), |
11520 |
+- UNIPHIER_PERI_RESET_MCSSI(12), |
11521 |
++ UNIPHIER_PERI_RESET_SCSSI(11, 0), |
11522 |
++ UNIPHIER_PERI_RESET_SCSSI(12, 1), |
11523 |
++ UNIPHIER_PERI_RESET_SCSSI(13, 2), |
11524 |
++ UNIPHIER_PERI_RESET_SCSSI(14, 3), |
11525 |
++ UNIPHIER_PERI_RESET_MCSSI(15), |
11526 |
+ UNIPHIER_RESET_END, |
11527 |
+ }; |
11528 |
+ |
11529 |
+diff --git a/drivers/scsi/aic7xxx/aic7xxx_core.c b/drivers/scsi/aic7xxx/aic7xxx_core.c |
11530 |
+index 915a34f141e4..49e02e874553 100644 |
11531 |
+--- a/drivers/scsi/aic7xxx/aic7xxx_core.c |
11532 |
++++ b/drivers/scsi/aic7xxx/aic7xxx_core.c |
11533 |
+@@ -2321,7 +2321,7 @@ ahc_find_syncrate(struct ahc_softc *ahc, u_int *period, |
11534 |
+ * At some speeds, we only support |
11535 |
+ * ST transfers. |
11536 |
+ */ |
11537 |
+- if ((syncrate->sxfr_u2 & ST_SXFR) != 0) |
11538 |
++ if ((syncrate->sxfr_u2 & ST_SXFR) != 0) |
11539 |
+ *ppr_options &= ~MSG_EXT_PPR_DT_REQ; |
11540 |
+ break; |
11541 |
+ } |
11542 |
+diff --git a/drivers/scsi/iscsi_tcp.c b/drivers/scsi/iscsi_tcp.c |
11543 |
+index 55181d28291e..7212e3a13fe6 100644 |
11544 |
+--- a/drivers/scsi/iscsi_tcp.c |
11545 |
++++ b/drivers/scsi/iscsi_tcp.c |
11546 |
+@@ -892,6 +892,10 @@ free_host: |
11547 |
+ static void iscsi_sw_tcp_session_destroy(struct iscsi_cls_session *cls_session) |
11548 |
+ { |
11549 |
+ struct Scsi_Host *shost = iscsi_session_to_shost(cls_session); |
11550 |
++ struct iscsi_session *session = cls_session->dd_data; |
11551 |
++ |
11552 |
++ if (WARN_ON_ONCE(session->leadconn)) |
11553 |
++ return; |
11554 |
+ |
11555 |
+ iscsi_tcp_r2tpool_free(cls_session->dd_data); |
11556 |
+ iscsi_session_teardown(cls_session); |
11557 |
+diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c |
11558 |
+index 4c4781e5974f..c0fb9e789080 100644 |
11559 |
+--- a/drivers/scsi/scsi_transport_iscsi.c |
11560 |
++++ b/drivers/scsi/scsi_transport_iscsi.c |
11561 |
+@@ -2945,6 +2945,24 @@ iscsi_set_path(struct iscsi_transport *transport, struct iscsi_uevent *ev) |
11562 |
+ return err; |
11563 |
+ } |
11564 |
+ |
11565 |
++static int iscsi_session_has_conns(int sid) |
11566 |
++{ |
11567 |
++ struct iscsi_cls_conn *conn; |
11568 |
++ unsigned long flags; |
11569 |
++ int found = 0; |
11570 |
++ |
11571 |
++ spin_lock_irqsave(&connlock, flags); |
11572 |
++ list_for_each_entry(conn, &connlist, conn_list) { |
11573 |
++ if (iscsi_conn_get_sid(conn) == sid) { |
11574 |
++ found = 1; |
11575 |
++ break; |
11576 |
++ } |
11577 |
++ } |
11578 |
++ spin_unlock_irqrestore(&connlock, flags); |
11579 |
++ |
11580 |
++ return found; |
11581 |
++} |
11582 |
++ |
11583 |
+ static int |
11584 |
+ iscsi_set_iface_params(struct iscsi_transport *transport, |
11585 |
+ struct iscsi_uevent *ev, uint32_t len) |
11586 |
+@@ -3522,10 +3540,12 @@ iscsi_if_recv_msg(struct sk_buff *skb, struct nlmsghdr *nlh, uint32_t *group) |
11587 |
+ break; |
11588 |
+ case ISCSI_UEVENT_DESTROY_SESSION: |
11589 |
+ session = iscsi_session_lookup(ev->u.d_session.sid); |
11590 |
+- if (session) |
11591 |
+- transport->destroy_session(session); |
11592 |
+- else |
11593 |
++ if (!session) |
11594 |
+ err = -EINVAL; |
11595 |
++ else if (iscsi_session_has_conns(ev->u.d_session.sid)) |
11596 |
++ err = -EBUSY; |
11597 |
++ else |
11598 |
++ transport->destroy_session(session); |
11599 |
+ break; |
11600 |
+ case ISCSI_UEVENT_UNBIND_SESSION: |
11601 |
+ session = iscsi_session_lookup(ev->u.d_session.sid); |
11602 |
+diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c |
11603 |
+index f4fcaee41dc2..b3dee24917a8 100644 |
11604 |
+--- a/drivers/scsi/ufs/ufshcd.c |
11605 |
++++ b/drivers/scsi/ufs/ufshcd.c |
11606 |
+@@ -4809,7 +4809,7 @@ ufshcd_transfer_rsp_status(struct ufs_hba *hba, struct ufshcd_lrb *lrbp) |
11607 |
+ break; |
11608 |
+ } /* end of switch */ |
11609 |
+ |
11610 |
+- if (host_byte(result) != DID_OK) |
11611 |
++ if ((host_byte(result) != DID_OK) && !hba->silence_err_logs) |
11612 |
+ ufshcd_print_trs(hba, 1 << lrbp->task_tag, true); |
11613 |
+ return result; |
11614 |
+ } |
11615 |
+@@ -5341,8 +5341,8 @@ static void ufshcd_err_handler(struct work_struct *work) |
11616 |
+ |
11617 |
+ /* |
11618 |
+ * if host reset is required then skip clearing the pending |
11619 |
+- * transfers forcefully because they will automatically get |
11620 |
+- * cleared after link startup. |
11621 |
++ * transfers forcefully because they will get cleared during |
11622 |
++ * host reset and restore |
11623 |
+ */ |
11624 |
+ if (needs_reset) |
11625 |
+ goto skip_pending_xfer_clear; |
11626 |
+@@ -5996,9 +5996,15 @@ static int ufshcd_host_reset_and_restore(struct ufs_hba *hba) |
11627 |
+ int err; |
11628 |
+ unsigned long flags; |
11629 |
+ |
11630 |
+- /* Reset the host controller */ |
11631 |
++ /* |
11632 |
++ * Stop the host controller and complete the requests |
11633 |
++ * cleared by h/w |
11634 |
++ */ |
11635 |
+ spin_lock_irqsave(hba->host->host_lock, flags); |
11636 |
+ ufshcd_hba_stop(hba, false); |
11637 |
++ hba->silence_err_logs = true; |
11638 |
++ ufshcd_complete_requests(hba); |
11639 |
++ hba->silence_err_logs = false; |
11640 |
+ spin_unlock_irqrestore(hba->host->host_lock, flags); |
11641 |
+ |
11642 |
+ /* scale up clocks to max frequency before full reinitialization */ |
11643 |
+@@ -6032,22 +6038,12 @@ out: |
11644 |
+ static int ufshcd_reset_and_restore(struct ufs_hba *hba) |
11645 |
+ { |
11646 |
+ int err = 0; |
11647 |
+- unsigned long flags; |
11648 |
+ int retries = MAX_HOST_RESET_RETRIES; |
11649 |
+ |
11650 |
+ do { |
11651 |
+ err = ufshcd_host_reset_and_restore(hba); |
11652 |
+ } while (err && --retries); |
11653 |
+ |
11654 |
+- /* |
11655 |
+- * After reset the door-bell might be cleared, complete |
11656 |
+- * outstanding requests in s/w here. |
11657 |
+- */ |
11658 |
+- spin_lock_irqsave(hba->host->host_lock, flags); |
11659 |
+- ufshcd_transfer_req_compl(hba); |
11660 |
+- ufshcd_tmc_handler(hba); |
11661 |
+- spin_unlock_irqrestore(hba->host->host_lock, flags); |
11662 |
+- |
11663 |
+ return err; |
11664 |
+ } |
11665 |
+ |
11666 |
+diff --git a/drivers/scsi/ufs/ufshcd.h b/drivers/scsi/ufs/ufshcd.h |
11667 |
+index 33fdd3f281ae..4554a4b725b5 100644 |
11668 |
+--- a/drivers/scsi/ufs/ufshcd.h |
11669 |
++++ b/drivers/scsi/ufs/ufshcd.h |
11670 |
+@@ -489,6 +489,7 @@ struct ufs_stats { |
11671 |
+ * @uic_error: UFS interconnect layer error status |
11672 |
+ * @saved_err: sticky error mask |
11673 |
+ * @saved_uic_err: sticky UIC error mask |
11674 |
++ * @silence_err_logs: flag to silence error logs |
11675 |
+ * @dev_cmd: ufs device management command information |
11676 |
+ * @last_dme_cmd_tstamp: time stamp of the last completed DME command |
11677 |
+ * @auto_bkops_enabled: to track whether bkops is enabled in device |
11678 |
+@@ -645,6 +646,7 @@ struct ufs_hba { |
11679 |
+ u32 saved_err; |
11680 |
+ u32 saved_uic_err; |
11681 |
+ struct ufs_stats ufs_stats; |
11682 |
++ bool silence_err_logs; |
11683 |
+ |
11684 |
+ /* Device management request data */ |
11685 |
+ struct ufs_dev_cmd dev_cmd; |
11686 |
+diff --git a/drivers/soc/tegra/fuse/tegra-apbmisc.c b/drivers/soc/tegra/fuse/tegra-apbmisc.c |
11687 |
+index e5a4d8f98b10..d1cbb0fe1691 100644 |
11688 |
+--- a/drivers/soc/tegra/fuse/tegra-apbmisc.c |
11689 |
++++ b/drivers/soc/tegra/fuse/tegra-apbmisc.c |
11690 |
+@@ -135,7 +135,7 @@ void __init tegra_init_apbmisc(void) |
11691 |
+ apbmisc.flags = IORESOURCE_MEM; |
11692 |
+ |
11693 |
+ /* strapping options */ |
11694 |
+- if (tegra_get_chip_id() == TEGRA124) { |
11695 |
++ if (of_machine_is_compatible("nvidia,tegra124")) { |
11696 |
+ straps.start = 0x7000e864; |
11697 |
+ straps.end = 0x7000e867; |
11698 |
+ } else { |
11699 |
+diff --git a/drivers/staging/rtl8188eu/os_dep/ioctl_linux.c b/drivers/staging/rtl8188eu/os_dep/ioctl_linux.c |
11700 |
+index bee3c3a7a7a9..2db4444267a7 100644 |
11701 |
+--- a/drivers/staging/rtl8188eu/os_dep/ioctl_linux.c |
11702 |
++++ b/drivers/staging/rtl8188eu/os_dep/ioctl_linux.c |
11703 |
+@@ -229,18 +229,21 @@ static char *translate_scan(struct adapter *padapter, |
11704 |
+ |
11705 |
+ /* parsing WPA/WPA2 IE */ |
11706 |
+ { |
11707 |
+- u8 buf[MAX_WPA_IE_LEN]; |
11708 |
++ u8 *buf; |
11709 |
+ u8 wpa_ie[255], rsn_ie[255]; |
11710 |
+ u16 wpa_len = 0, rsn_len = 0; |
11711 |
+ u8 *p; |
11712 |
+ |
11713 |
++ buf = kzalloc(MAX_WPA_IE_LEN, GFP_ATOMIC); |
11714 |
++ if (!buf) |
11715 |
++ return start; |
11716 |
++ |
11717 |
+ rtw_get_sec_ie(pnetwork->network.ies, pnetwork->network.ie_length, rsn_ie, &rsn_len, wpa_ie, &wpa_len); |
11718 |
+ RT_TRACE(_module_rtl871x_mlme_c_, _drv_info_, ("rtw_wx_get_scan: ssid =%s\n", pnetwork->network.Ssid.Ssid)); |
11719 |
+ RT_TRACE(_module_rtl871x_mlme_c_, _drv_info_, ("rtw_wx_get_scan: wpa_len =%d rsn_len =%d\n", wpa_len, rsn_len)); |
11720 |
+ |
11721 |
+ if (wpa_len > 0) { |
11722 |
+ p = buf; |
11723 |
+- memset(buf, 0, MAX_WPA_IE_LEN); |
11724 |
+ p += sprintf(p, "wpa_ie="); |
11725 |
+ for (i = 0; i < wpa_len; i++) |
11726 |
+ p += sprintf(p, "%02x", wpa_ie[i]); |
11727 |
+@@ -257,7 +260,6 @@ static char *translate_scan(struct adapter *padapter, |
11728 |
+ } |
11729 |
+ if (rsn_len > 0) { |
11730 |
+ p = buf; |
11731 |
+- memset(buf, 0, MAX_WPA_IE_LEN); |
11732 |
+ p += sprintf(p, "rsn_ie="); |
11733 |
+ for (i = 0; i < rsn_len; i++) |
11734 |
+ p += sprintf(p, "%02x", rsn_ie[i]); |
11735 |
+@@ -271,6 +273,7 @@ static char *translate_scan(struct adapter *padapter, |
11736 |
+ iwe.u.data.length = rsn_len; |
11737 |
+ start = iwe_stream_add_point(info, start, stop, &iwe, rsn_ie); |
11738 |
+ } |
11739 |
++ kfree(buf); |
11740 |
+ } |
11741 |
+ |
11742 |
+ {/* parsing WPS IE */ |
11743 |
+diff --git a/drivers/tty/synclink_gt.c b/drivers/tty/synclink_gt.c |
11744 |
+index b88ecf102764..e9779b03ee56 100644 |
11745 |
+--- a/drivers/tty/synclink_gt.c |
11746 |
++++ b/drivers/tty/synclink_gt.c |
11747 |
+@@ -1335,10 +1335,10 @@ static void throttle(struct tty_struct * tty) |
11748 |
+ DBGINFO(("%s throttle\n", info->device_name)); |
11749 |
+ if (I_IXOFF(tty)) |
11750 |
+ send_xchar(tty, STOP_CHAR(tty)); |
11751 |
+- if (C_CRTSCTS(tty)) { |
11752 |
++ if (C_CRTSCTS(tty)) { |
11753 |
+ spin_lock_irqsave(&info->lock,flags); |
11754 |
+ info->signals &= ~SerialSignal_RTS; |
11755 |
+- set_signals(info); |
11756 |
++ set_signals(info); |
11757 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11758 |
+ } |
11759 |
+ } |
11760 |
+@@ -1360,10 +1360,10 @@ static void unthrottle(struct tty_struct * tty) |
11761 |
+ else |
11762 |
+ send_xchar(tty, START_CHAR(tty)); |
11763 |
+ } |
11764 |
+- if (C_CRTSCTS(tty)) { |
11765 |
++ if (C_CRTSCTS(tty)) { |
11766 |
+ spin_lock_irqsave(&info->lock,flags); |
11767 |
+ info->signals |= SerialSignal_RTS; |
11768 |
+- set_signals(info); |
11769 |
++ set_signals(info); |
11770 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11771 |
+ } |
11772 |
+ } |
11773 |
+@@ -2561,8 +2561,8 @@ static void change_params(struct slgt_info *info) |
11774 |
+ info->read_status_mask = IRQ_RXOVER; |
11775 |
+ if (I_INPCK(info->port.tty)) |
11776 |
+ info->read_status_mask |= MASK_PARITY | MASK_FRAMING; |
11777 |
+- if (I_BRKINT(info->port.tty) || I_PARMRK(info->port.tty)) |
11778 |
+- info->read_status_mask |= MASK_BREAK; |
11779 |
++ if (I_BRKINT(info->port.tty) || I_PARMRK(info->port.tty)) |
11780 |
++ info->read_status_mask |= MASK_BREAK; |
11781 |
+ if (I_IGNPAR(info->port.tty)) |
11782 |
+ info->ignore_status_mask |= MASK_PARITY | MASK_FRAMING; |
11783 |
+ if (I_IGNBRK(info->port.tty)) { |
11784 |
+@@ -3193,7 +3193,7 @@ static int tiocmset(struct tty_struct *tty, |
11785 |
+ info->signals &= ~SerialSignal_DTR; |
11786 |
+ |
11787 |
+ spin_lock_irqsave(&info->lock,flags); |
11788 |
+- set_signals(info); |
11789 |
++ set_signals(info); |
11790 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11791 |
+ return 0; |
11792 |
+ } |
11793 |
+@@ -3204,7 +3204,7 @@ static int carrier_raised(struct tty_port *port) |
11794 |
+ struct slgt_info *info = container_of(port, struct slgt_info, port); |
11795 |
+ |
11796 |
+ spin_lock_irqsave(&info->lock,flags); |
11797 |
+- get_signals(info); |
11798 |
++ get_signals(info); |
11799 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11800 |
+ return (info->signals & SerialSignal_DCD) ? 1 : 0; |
11801 |
+ } |
11802 |
+@@ -3219,7 +3219,7 @@ static void dtr_rts(struct tty_port *port, int on) |
11803 |
+ info->signals |= SerialSignal_RTS | SerialSignal_DTR; |
11804 |
+ else |
11805 |
+ info->signals &= ~(SerialSignal_RTS | SerialSignal_DTR); |
11806 |
+- set_signals(info); |
11807 |
++ set_signals(info); |
11808 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11809 |
+ } |
11810 |
+ |
11811 |
+diff --git a/drivers/tty/synclinkmp.c b/drivers/tty/synclinkmp.c |
11812 |
+index 1e4d5b9c981a..57c2c647af61 100644 |
11813 |
+--- a/drivers/tty/synclinkmp.c |
11814 |
++++ b/drivers/tty/synclinkmp.c |
11815 |
+@@ -1454,10 +1454,10 @@ static void throttle(struct tty_struct * tty) |
11816 |
+ if (I_IXOFF(tty)) |
11817 |
+ send_xchar(tty, STOP_CHAR(tty)); |
11818 |
+ |
11819 |
+- if (C_CRTSCTS(tty)) { |
11820 |
++ if (C_CRTSCTS(tty)) { |
11821 |
+ spin_lock_irqsave(&info->lock,flags); |
11822 |
+ info->serial_signals &= ~SerialSignal_RTS; |
11823 |
+- set_signals(info); |
11824 |
++ set_signals(info); |
11825 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11826 |
+ } |
11827 |
+ } |
11828 |
+@@ -1483,10 +1483,10 @@ static void unthrottle(struct tty_struct * tty) |
11829 |
+ send_xchar(tty, START_CHAR(tty)); |
11830 |
+ } |
11831 |
+ |
11832 |
+- if (C_CRTSCTS(tty)) { |
11833 |
++ if (C_CRTSCTS(tty)) { |
11834 |
+ spin_lock_irqsave(&info->lock,flags); |
11835 |
+ info->serial_signals |= SerialSignal_RTS; |
11836 |
+- set_signals(info); |
11837 |
++ set_signals(info); |
11838 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11839 |
+ } |
11840 |
+ } |
11841 |
+@@ -2471,7 +2471,7 @@ static void isr_io_pin( SLMP_INFO *info, u16 status ) |
11842 |
+ if (status & SerialSignal_CTS) { |
11843 |
+ if ( debug_level >= DEBUG_LEVEL_ISR ) |
11844 |
+ printk("CTS tx start..."); |
11845 |
+- info->port.tty->hw_stopped = 0; |
11846 |
++ info->port.tty->hw_stopped = 0; |
11847 |
+ tx_start(info); |
11848 |
+ info->pending_bh |= BH_TRANSMIT; |
11849 |
+ return; |
11850 |
+@@ -2480,7 +2480,7 @@ static void isr_io_pin( SLMP_INFO *info, u16 status ) |
11851 |
+ if (!(status & SerialSignal_CTS)) { |
11852 |
+ if ( debug_level >= DEBUG_LEVEL_ISR ) |
11853 |
+ printk("CTS tx stop..."); |
11854 |
+- info->port.tty->hw_stopped = 1; |
11855 |
++ info->port.tty->hw_stopped = 1; |
11856 |
+ tx_stop(info); |
11857 |
+ } |
11858 |
+ } |
11859 |
+@@ -2807,8 +2807,8 @@ static void change_params(SLMP_INFO *info) |
11860 |
+ info->read_status_mask2 = OVRN; |
11861 |
+ if (I_INPCK(info->port.tty)) |
11862 |
+ info->read_status_mask2 |= PE | FRME; |
11863 |
+- if (I_BRKINT(info->port.tty) || I_PARMRK(info->port.tty)) |
11864 |
+- info->read_status_mask1 |= BRKD; |
11865 |
++ if (I_BRKINT(info->port.tty) || I_PARMRK(info->port.tty)) |
11866 |
++ info->read_status_mask1 |= BRKD; |
11867 |
+ if (I_IGNPAR(info->port.tty)) |
11868 |
+ info->ignore_status_mask2 |= PE | FRME; |
11869 |
+ if (I_IGNBRK(info->port.tty)) { |
11870 |
+@@ -3178,7 +3178,7 @@ static int tiocmget(struct tty_struct *tty) |
11871 |
+ unsigned long flags; |
11872 |
+ |
11873 |
+ spin_lock_irqsave(&info->lock,flags); |
11874 |
+- get_signals(info); |
11875 |
++ get_signals(info); |
11876 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11877 |
+ |
11878 |
+ result = ((info->serial_signals & SerialSignal_RTS) ? TIOCM_RTS : 0) | |
11879 |
+@@ -3216,7 +3216,7 @@ static int tiocmset(struct tty_struct *tty, |
11880 |
+ info->serial_signals &= ~SerialSignal_DTR; |
11881 |
+ |
11882 |
+ spin_lock_irqsave(&info->lock,flags); |
11883 |
+- set_signals(info); |
11884 |
++ set_signals(info); |
11885 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11886 |
+ |
11887 |
+ return 0; |
11888 |
+@@ -3228,7 +3228,7 @@ static int carrier_raised(struct tty_port *port) |
11889 |
+ unsigned long flags; |
11890 |
+ |
11891 |
+ spin_lock_irqsave(&info->lock,flags); |
11892 |
+- get_signals(info); |
11893 |
++ get_signals(info); |
11894 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11895 |
+ |
11896 |
+ return (info->serial_signals & SerialSignal_DCD) ? 1 : 0; |
11897 |
+@@ -3244,7 +3244,7 @@ static void dtr_rts(struct tty_port *port, int on) |
11898 |
+ info->serial_signals |= SerialSignal_RTS | SerialSignal_DTR; |
11899 |
+ else |
11900 |
+ info->serial_signals &= ~(SerialSignal_RTS | SerialSignal_DTR); |
11901 |
+- set_signals(info); |
11902 |
++ set_signals(info); |
11903 |
+ spin_unlock_irqrestore(&info->lock,flags); |
11904 |
+ } |
11905 |
+ |
11906 |
+diff --git a/drivers/uio/uio_dmem_genirq.c b/drivers/uio/uio_dmem_genirq.c |
11907 |
+index e1134a4d97f3..a00b4aee6c79 100644 |
11908 |
+--- a/drivers/uio/uio_dmem_genirq.c |
11909 |
++++ b/drivers/uio/uio_dmem_genirq.c |
11910 |
+@@ -135,11 +135,13 @@ static int uio_dmem_genirq_irqcontrol(struct uio_info *dev_info, s32 irq_on) |
11911 |
+ if (irq_on) { |
11912 |
+ if (test_and_clear_bit(0, &priv->flags)) |
11913 |
+ enable_irq(dev_info->irq); |
11914 |
++ spin_unlock_irqrestore(&priv->lock, flags); |
11915 |
+ } else { |
11916 |
+- if (!test_and_set_bit(0, &priv->flags)) |
11917 |
++ if (!test_and_set_bit(0, &priv->flags)) { |
11918 |
++ spin_unlock_irqrestore(&priv->lock, flags); |
11919 |
+ disable_irq(dev_info->irq); |
11920 |
++ } |
11921 |
+ } |
11922 |
+- spin_unlock_irqrestore(&priv->lock, flags); |
11923 |
+ |
11924 |
+ return 0; |
11925 |
+ } |
11926 |
+diff --git a/drivers/usb/dwc2/gadget.c b/drivers/usb/dwc2/gadget.c |
11927 |
+index f64d1cd08fb6..17f3e7b4d4fe 100644 |
11928 |
+--- a/drivers/usb/dwc2/gadget.c |
11929 |
++++ b/drivers/usb/dwc2/gadget.c |
11930 |
+@@ -3918,11 +3918,12 @@ static int dwc2_hsotg_ep_enable(struct usb_ep *ep, |
11931 |
+ * a unique tx-fifo even if it is non-periodic. |
11932 |
+ */ |
11933 |
+ if (dir_in && hsotg->dedicated_fifos) { |
11934 |
++ unsigned fifo_count = dwc2_hsotg_tx_fifo_count(hsotg); |
11935 |
+ u32 fifo_index = 0; |
11936 |
+ u32 fifo_size = UINT_MAX; |
11937 |
+ |
11938 |
+ size = hs_ep->ep.maxpacket * hs_ep->mc; |
11939 |
+- for (i = 1; i < hsotg->num_of_eps; ++i) { |
11940 |
++ for (i = 1; i <= fifo_count; ++i) { |
11941 |
+ if (hsotg->fifo_map & (1 << i)) |
11942 |
+ continue; |
11943 |
+ val = dwc2_readl(hsotg, DPTXFSIZN(i)); |
11944 |
+diff --git a/drivers/usb/gadget/udc/gr_udc.c b/drivers/usb/gadget/udc/gr_udc.c |
11945 |
+index 729e60e49564..e50108f9a374 100644 |
11946 |
+--- a/drivers/usb/gadget/udc/gr_udc.c |
11947 |
++++ b/drivers/usb/gadget/udc/gr_udc.c |
11948 |
+@@ -2180,8 +2180,6 @@ static int gr_probe(struct platform_device *pdev) |
11949 |
+ return -ENOMEM; |
11950 |
+ } |
11951 |
+ |
11952 |
+- spin_lock(&dev->lock); |
11953 |
+- |
11954 |
+ /* Inside lock so that no gadget can use this udc until probe is done */ |
11955 |
+ retval = usb_add_gadget_udc(dev->dev, &dev->gadget); |
11956 |
+ if (retval) { |
11957 |
+@@ -2190,15 +2188,21 @@ static int gr_probe(struct platform_device *pdev) |
11958 |
+ } |
11959 |
+ dev->added = 1; |
11960 |
+ |
11961 |
++ spin_lock(&dev->lock); |
11962 |
++ |
11963 |
+ retval = gr_udc_init(dev); |
11964 |
+- if (retval) |
11965 |
++ if (retval) { |
11966 |
++ spin_unlock(&dev->lock); |
11967 |
+ goto out; |
11968 |
+- |
11969 |
+- gr_dfs_create(dev); |
11970 |
++ } |
11971 |
+ |
11972 |
+ /* Clear all interrupt enables that might be left on since last boot */ |
11973 |
+ gr_disable_interrupts_and_pullup(dev); |
11974 |
+ |
11975 |
++ spin_unlock(&dev->lock); |
11976 |
++ |
11977 |
++ gr_dfs_create(dev); |
11978 |
++ |
11979 |
+ retval = gr_request_irq(dev, dev->irq); |
11980 |
+ if (retval) { |
11981 |
+ dev_err(dev->dev, "Failed to request irq %d\n", dev->irq); |
11982 |
+@@ -2227,8 +2231,6 @@ static int gr_probe(struct platform_device *pdev) |
11983 |
+ dev_info(dev->dev, "regs: %p, irq %d\n", dev->regs, dev->irq); |
11984 |
+ |
11985 |
+ out: |
11986 |
+- spin_unlock(&dev->lock); |
11987 |
+- |
11988 |
+ if (retval) |
11989 |
+ gr_remove(pdev); |
11990 |
+ |
11991 |
+diff --git a/drivers/usb/musb/omap2430.c b/drivers/usb/musb/omap2430.c |
11992 |
+index b1dd81fb5f55..24e622c05638 100644 |
11993 |
+--- a/drivers/usb/musb/omap2430.c |
11994 |
++++ b/drivers/usb/musb/omap2430.c |
11995 |
+@@ -361,8 +361,6 @@ static const struct musb_platform_ops omap2430_ops = { |
11996 |
+ .init = omap2430_musb_init, |
11997 |
+ .exit = omap2430_musb_exit, |
11998 |
+ |
11999 |
+- .set_vbus = omap2430_musb_set_vbus, |
12000 |
+- |
12001 |
+ .enable = omap2430_musb_enable, |
12002 |
+ .disable = omap2430_musb_disable, |
12003 |
+ |
12004 |
+diff --git a/drivers/video/fbdev/pxa168fb.c b/drivers/video/fbdev/pxa168fb.c |
12005 |
+index d059d04c63ac..20195d3dbf08 100644 |
12006 |
+--- a/drivers/video/fbdev/pxa168fb.c |
12007 |
++++ b/drivers/video/fbdev/pxa168fb.c |
12008 |
+@@ -769,8 +769,8 @@ failed_free_cmap: |
12009 |
+ failed_free_clk: |
12010 |
+ clk_disable_unprepare(fbi->clk); |
12011 |
+ failed_free_fbmem: |
12012 |
+- dma_free_coherent(fbi->dev, info->fix.smem_len, |
12013 |
+- info->screen_base, fbi->fb_start_dma); |
12014 |
++ dma_free_wc(fbi->dev, info->fix.smem_len, |
12015 |
++ info->screen_base, fbi->fb_start_dma); |
12016 |
+ failed_free_info: |
12017 |
+ kfree(info); |
12018 |
+ |
12019 |
+@@ -804,7 +804,7 @@ static int pxa168fb_remove(struct platform_device *pdev) |
12020 |
+ |
12021 |
+ irq = platform_get_irq(pdev, 0); |
12022 |
+ |
12023 |
+- dma_free_wc(fbi->dev, PAGE_ALIGN(info->fix.smem_len), |
12024 |
++ dma_free_wc(fbi->dev, info->fix.smem_len, |
12025 |
+ info->screen_base, info->fix.smem_start); |
12026 |
+ |
12027 |
+ clk_disable_unprepare(fbi->clk); |
12028 |
+diff --git a/drivers/virtio/virtio_balloon.c b/drivers/virtio/virtio_balloon.c |
12029 |
+index 14ac36ca8fbd..1afcbef397ab 100644 |
12030 |
+--- a/drivers/virtio/virtio_balloon.c |
12031 |
++++ b/drivers/virtio/virtio_balloon.c |
12032 |
+@@ -126,6 +126,8 @@ static void set_page_pfns(struct virtio_balloon *vb, |
12033 |
+ { |
12034 |
+ unsigned int i; |
12035 |
+ |
12036 |
++ BUILD_BUG_ON(VIRTIO_BALLOON_PAGES_PER_PAGE > VIRTIO_BALLOON_ARRAY_PFNS_MAX); |
12037 |
++ |
12038 |
+ /* |
12039 |
+ * Set balloon pfns pointing at this page. |
12040 |
+ * Note that the first pfn points at start of the page. |
12041 |
+diff --git a/drivers/visorbus/visorchipset.c b/drivers/visorbus/visorchipset.c |
12042 |
+index ca752b8f495f..cb1eb7e05f87 100644 |
12043 |
+--- a/drivers/visorbus/visorchipset.c |
12044 |
++++ b/drivers/visorbus/visorchipset.c |
12045 |
+@@ -1210,14 +1210,17 @@ static void setup_crash_devices_work_queue(struct work_struct *work) |
12046 |
+ { |
12047 |
+ struct controlvm_message local_crash_bus_msg; |
12048 |
+ struct controlvm_message local_crash_dev_msg; |
12049 |
+- struct controlvm_message msg; |
12050 |
++ struct controlvm_message msg = { |
12051 |
++ .hdr.id = CONTROLVM_CHIPSET_INIT, |
12052 |
++ .cmd.init_chipset = { |
12053 |
++ .bus_count = 23, |
12054 |
++ .switch_count = 0, |
12055 |
++ }, |
12056 |
++ }; |
12057 |
+ u32 local_crash_msg_offset; |
12058 |
+ u16 local_crash_msg_count; |
12059 |
+ |
12060 |
+ /* send init chipset msg */ |
12061 |
+- msg.hdr.id = CONTROLVM_CHIPSET_INIT; |
12062 |
+- msg.cmd.init_chipset.bus_count = 23; |
12063 |
+- msg.cmd.init_chipset.switch_count = 0; |
12064 |
+ chipset_init(&msg); |
12065 |
+ /* get saved message count */ |
12066 |
+ if (visorchannel_read(chipset_dev->controlvm_channel, |
12067 |
+diff --git a/drivers/vme/bridges/vme_fake.c b/drivers/vme/bridges/vme_fake.c |
12068 |
+index 7d83691047f4..685a43bdc2a1 100644 |
12069 |
+--- a/drivers/vme/bridges/vme_fake.c |
12070 |
++++ b/drivers/vme/bridges/vme_fake.c |
12071 |
+@@ -418,8 +418,9 @@ static void fake_lm_check(struct fake_driver *bridge, unsigned long long addr, |
12072 |
+ } |
12073 |
+ } |
12074 |
+ |
12075 |
+-static u8 fake_vmeread8(struct fake_driver *bridge, unsigned long long addr, |
12076 |
+- u32 aspace, u32 cycle) |
12077 |
++static noinline_for_stack u8 fake_vmeread8(struct fake_driver *bridge, |
12078 |
++ unsigned long long addr, |
12079 |
++ u32 aspace, u32 cycle) |
12080 |
+ { |
12081 |
+ u8 retval = 0xff; |
12082 |
+ int i; |
12083 |
+@@ -450,8 +451,9 @@ static u8 fake_vmeread8(struct fake_driver *bridge, unsigned long long addr, |
12084 |
+ return retval; |
12085 |
+ } |
12086 |
+ |
12087 |
+-static u16 fake_vmeread16(struct fake_driver *bridge, unsigned long long addr, |
12088 |
+- u32 aspace, u32 cycle) |
12089 |
++static noinline_for_stack u16 fake_vmeread16(struct fake_driver *bridge, |
12090 |
++ unsigned long long addr, |
12091 |
++ u32 aspace, u32 cycle) |
12092 |
+ { |
12093 |
+ u16 retval = 0xffff; |
12094 |
+ int i; |
12095 |
+@@ -482,8 +484,9 @@ static u16 fake_vmeread16(struct fake_driver *bridge, unsigned long long addr, |
12096 |
+ return retval; |
12097 |
+ } |
12098 |
+ |
12099 |
+-static u32 fake_vmeread32(struct fake_driver *bridge, unsigned long long addr, |
12100 |
+- u32 aspace, u32 cycle) |
12101 |
++static noinline_for_stack u32 fake_vmeread32(struct fake_driver *bridge, |
12102 |
++ unsigned long long addr, |
12103 |
++ u32 aspace, u32 cycle) |
12104 |
+ { |
12105 |
+ u32 retval = 0xffffffff; |
12106 |
+ int i; |
12107 |
+@@ -613,8 +616,9 @@ out: |
12108 |
+ return retval; |
12109 |
+ } |
12110 |
+ |
12111 |
+-static void fake_vmewrite8(struct fake_driver *bridge, u8 *buf, |
12112 |
+- unsigned long long addr, u32 aspace, u32 cycle) |
12113 |
++static noinline_for_stack void fake_vmewrite8(struct fake_driver *bridge, |
12114 |
++ u8 *buf, unsigned long long addr, |
12115 |
++ u32 aspace, u32 cycle) |
12116 |
+ { |
12117 |
+ int i; |
12118 |
+ unsigned long long start, end, offset; |
12119 |
+@@ -643,8 +647,9 @@ static void fake_vmewrite8(struct fake_driver *bridge, u8 *buf, |
12120 |
+ |
12121 |
+ } |
12122 |
+ |
12123 |
+-static void fake_vmewrite16(struct fake_driver *bridge, u16 *buf, |
12124 |
+- unsigned long long addr, u32 aspace, u32 cycle) |
12125 |
++static noinline_for_stack void fake_vmewrite16(struct fake_driver *bridge, |
12126 |
++ u16 *buf, unsigned long long addr, |
12127 |
++ u32 aspace, u32 cycle) |
12128 |
+ { |
12129 |
+ int i; |
12130 |
+ unsigned long long start, end, offset; |
12131 |
+@@ -673,8 +678,9 @@ static void fake_vmewrite16(struct fake_driver *bridge, u16 *buf, |
12132 |
+ |
12133 |
+ } |
12134 |
+ |
12135 |
+-static void fake_vmewrite32(struct fake_driver *bridge, u32 *buf, |
12136 |
+- unsigned long long addr, u32 aspace, u32 cycle) |
12137 |
++static noinline_for_stack void fake_vmewrite32(struct fake_driver *bridge, |
12138 |
++ u32 *buf, unsigned long long addr, |
12139 |
++ u32 aspace, u32 cycle) |
12140 |
+ { |
12141 |
+ int i; |
12142 |
+ unsigned long long start, end, offset; |
12143 |
+diff --git a/fs/btrfs/check-integrity.c b/fs/btrfs/check-integrity.c |
12144 |
+index 833cf3c35b4d..3b77c8ab5357 100644 |
12145 |
+--- a/fs/btrfs/check-integrity.c |
12146 |
++++ b/fs/btrfs/check-integrity.c |
12147 |
+@@ -629,7 +629,6 @@ static struct btrfsic_dev_state *btrfsic_dev_state_hashtable_lookup(dev_t dev, |
12148 |
+ static int btrfsic_process_superblock(struct btrfsic_state *state, |
12149 |
+ struct btrfs_fs_devices *fs_devices) |
12150 |
+ { |
12151 |
+- struct btrfs_fs_info *fs_info = state->fs_info; |
12152 |
+ struct btrfs_super_block *selected_super; |
12153 |
+ struct list_head *dev_head = &fs_devices->devices; |
12154 |
+ struct btrfs_device *device; |
12155 |
+@@ -700,7 +699,7 @@ static int btrfsic_process_superblock(struct btrfsic_state *state, |
12156 |
+ break; |
12157 |
+ } |
12158 |
+ |
12159 |
+- num_copies = btrfs_num_copies(fs_info, next_bytenr, |
12160 |
++ num_copies = btrfs_num_copies(state->fs_info, next_bytenr, |
12161 |
+ state->metablock_size); |
12162 |
+ if (state->print_mask & BTRFSIC_PRINT_MASK_NUM_COPIES) |
12163 |
+ pr_info("num_copies(log_bytenr=%llu) = %d\n", |
12164 |
+diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c |
12165 |
+index 4cf2817ab120..f9e280d0b44f 100644 |
12166 |
+--- a/fs/btrfs/file-item.c |
12167 |
++++ b/fs/btrfs/file-item.c |
12168 |
+@@ -275,7 +275,8 @@ found: |
12169 |
+ csum += count * csum_size; |
12170 |
+ nblocks -= count; |
12171 |
+ next: |
12172 |
+- while (count--) { |
12173 |
++ while (count > 0) { |
12174 |
++ count--; |
12175 |
+ disk_bytenr += fs_info->sectorsize; |
12176 |
+ offset += fs_info->sectorsize; |
12177 |
+ page_bytes_left -= fs_info->sectorsize; |
12178 |
+diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c |
12179 |
+index 5bbcdcff68a9..9c3b394b99fa 100644 |
12180 |
+--- a/fs/btrfs/volumes.c |
12181 |
++++ b/fs/btrfs/volumes.c |
12182 |
+@@ -7260,6 +7260,8 @@ int btrfs_get_dev_stats(struct btrfs_fs_info *fs_info, |
12183 |
+ else |
12184 |
+ btrfs_dev_stat_reset(dev, i); |
12185 |
+ } |
12186 |
++ btrfs_info(fs_info, "device stats zeroed by %s (%d)", |
12187 |
++ current->comm, task_pid_nr(current)); |
12188 |
+ } else { |
12189 |
+ for (i = 0; i < BTRFS_DEV_STAT_VALUES_MAX; i++) |
12190 |
+ if (stats->nr_items > i) |
12191 |
+diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c |
12192 |
+index 09db6d08614d..a2e903203bf9 100644 |
12193 |
+--- a/fs/ceph/mds_client.c |
12194 |
++++ b/fs/ceph/mds_client.c |
12195 |
+@@ -2343,8 +2343,7 @@ static void __do_request(struct ceph_mds_client *mdsc, |
12196 |
+ if (!(mdsc->fsc->mount_options->flags & |
12197 |
+ CEPH_MOUNT_OPT_MOUNTWAIT) && |
12198 |
+ !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) { |
12199 |
+- err = -ENOENT; |
12200 |
+- pr_info("probably no mds server is up\n"); |
12201 |
++ err = -EHOSTUNREACH; |
12202 |
+ goto finish; |
12203 |
+ } |
12204 |
+ } |
12205 |
+diff --git a/fs/ceph/super.c b/fs/ceph/super.c |
12206 |
+index 2bd0b1ed9708..c4314f449240 100644 |
12207 |
+--- a/fs/ceph/super.c |
12208 |
++++ b/fs/ceph/super.c |
12209 |
+@@ -1106,6 +1106,11 @@ static struct dentry *ceph_mount(struct file_system_type *fs_type, |
12210 |
+ return res; |
12211 |
+ |
12212 |
+ out_splat: |
12213 |
++ if (!ceph_mdsmap_is_cluster_available(fsc->mdsc->mdsmap)) { |
12214 |
++ pr_info("No mds server is up or the cluster is laggy\n"); |
12215 |
++ err = -EHOSTUNREACH; |
12216 |
++ } |
12217 |
++ |
12218 |
+ ceph_mdsc_close_sessions(fsc->mdsc); |
12219 |
+ deactivate_locked_super(sb); |
12220 |
+ goto out_final; |
12221 |
+diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c |
12222 |
+index 576cf71576da..6c62ce40608a 100644 |
12223 |
+--- a/fs/cifs/connect.c |
12224 |
++++ b/fs/cifs/connect.c |
12225 |
+@@ -3342,8 +3342,10 @@ match_prepath(struct super_block *sb, struct cifs_mnt_data *mnt_data) |
12226 |
+ { |
12227 |
+ struct cifs_sb_info *old = CIFS_SB(sb); |
12228 |
+ struct cifs_sb_info *new = mnt_data->cifs_sb; |
12229 |
+- bool old_set = old->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH; |
12230 |
+- bool new_set = new->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH; |
12231 |
++ bool old_set = (old->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH) && |
12232 |
++ old->prepath; |
12233 |
++ bool new_set = (new->mnt_cifs_flags & CIFS_MOUNT_USE_PREFIX_PATH) && |
12234 |
++ new->prepath; |
12235 |
+ |
12236 |
+ if (old_set && new_set && !strcmp(new->prepath, old->prepath)) |
12237 |
+ return 1; |
12238 |
+diff --git a/fs/cifs/smb2pdu.c b/fs/cifs/smb2pdu.c |
12239 |
+index 0d4e4d97e6cf..e2d2b749c8f3 100644 |
12240 |
+--- a/fs/cifs/smb2pdu.c |
12241 |
++++ b/fs/cifs/smb2pdu.c |
12242 |
+@@ -3425,6 +3425,9 @@ smb2_writev_callback(struct mid_q_entry *mid) |
12243 |
+ wdata->cfile->fid.persistent_fid, |
12244 |
+ tcon->tid, tcon->ses->Suid, wdata->offset, |
12245 |
+ wdata->bytes, wdata->result); |
12246 |
++ if (wdata->result == -ENOSPC) |
12247 |
++ printk_once(KERN_WARNING "Out of space writing to %s\n", |
12248 |
++ tcon->treeName); |
12249 |
+ } else |
12250 |
+ trace_smb3_write_done(0 /* no xid */, |
12251 |
+ wdata->cfile->fid.persistent_fid, |
12252 |
+diff --git a/fs/ext4/file.c b/fs/ext4/file.c |
12253 |
+index f4a24a46245e..52d155b4e733 100644 |
12254 |
+--- a/fs/ext4/file.c |
12255 |
++++ b/fs/ext4/file.c |
12256 |
+@@ -40,9 +40,10 @@ static ssize_t ext4_dax_read_iter(struct kiocb *iocb, struct iov_iter *to) |
12257 |
+ struct inode *inode = file_inode(iocb->ki_filp); |
12258 |
+ ssize_t ret; |
12259 |
+ |
12260 |
+- if (!inode_trylock_shared(inode)) { |
12261 |
+- if (iocb->ki_flags & IOCB_NOWAIT) |
12262 |
++ if (iocb->ki_flags & IOCB_NOWAIT) { |
12263 |
++ if (!inode_trylock_shared(inode)) |
12264 |
+ return -EAGAIN; |
12265 |
++ } else { |
12266 |
+ inode_lock_shared(inode); |
12267 |
+ } |
12268 |
+ /* |
12269 |
+@@ -190,9 +191,10 @@ ext4_dax_write_iter(struct kiocb *iocb, struct iov_iter *from) |
12270 |
+ struct inode *inode = file_inode(iocb->ki_filp); |
12271 |
+ ssize_t ret; |
12272 |
+ |
12273 |
+- if (!inode_trylock(inode)) { |
12274 |
+- if (iocb->ki_flags & IOCB_NOWAIT) |
12275 |
++ if (iocb->ki_flags & IOCB_NOWAIT) { |
12276 |
++ if (!inode_trylock(inode)) |
12277 |
+ return -EAGAIN; |
12278 |
++ } else { |
12279 |
+ inode_lock(inode); |
12280 |
+ } |
12281 |
+ ret = ext4_write_checks(iocb, from); |
12282 |
+diff --git a/fs/f2fs/namei.c b/fs/f2fs/namei.c |
12283 |
+index 0ace2c2e3de9..4f0cc0c79d1e 100644 |
12284 |
+--- a/fs/f2fs/namei.c |
12285 |
++++ b/fs/f2fs/namei.c |
12286 |
+@@ -769,6 +769,7 @@ static int __f2fs_tmpfile(struct inode *dir, struct dentry *dentry, |
12287 |
+ |
12288 |
+ if (whiteout) { |
12289 |
+ f2fs_i_links_write(inode, false); |
12290 |
++ inode->i_state |= I_LINKABLE; |
12291 |
+ *whiteout = inode; |
12292 |
+ } else { |
12293 |
+ d_tmpfile(dentry, inode); |
12294 |
+@@ -835,6 +836,12 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12295 |
+ F2FS_I(old_dentry->d_inode)->i_projid))) |
12296 |
+ return -EXDEV; |
12297 |
+ |
12298 |
++ if (flags & RENAME_WHITEOUT) { |
12299 |
++ err = f2fs_create_whiteout(old_dir, &whiteout); |
12300 |
++ if (err) |
12301 |
++ return err; |
12302 |
++ } |
12303 |
++ |
12304 |
+ err = dquot_initialize(old_dir); |
12305 |
+ if (err) |
12306 |
+ goto out; |
12307 |
+@@ -865,17 +872,11 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12308 |
+ } |
12309 |
+ } |
12310 |
+ |
12311 |
+- if (flags & RENAME_WHITEOUT) { |
12312 |
+- err = f2fs_create_whiteout(old_dir, &whiteout); |
12313 |
+- if (err) |
12314 |
+- goto out_dir; |
12315 |
+- } |
12316 |
+- |
12317 |
+ if (new_inode) { |
12318 |
+ |
12319 |
+ err = -ENOTEMPTY; |
12320 |
+ if (old_dir_entry && !f2fs_empty_dir(new_inode)) |
12321 |
+- goto out_whiteout; |
12322 |
++ goto out_dir; |
12323 |
+ |
12324 |
+ err = -ENOENT; |
12325 |
+ new_entry = f2fs_find_entry(new_dir, &new_dentry->d_name, |
12326 |
+@@ -883,7 +884,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12327 |
+ if (!new_entry) { |
12328 |
+ if (IS_ERR(new_page)) |
12329 |
+ err = PTR_ERR(new_page); |
12330 |
+- goto out_whiteout; |
12331 |
++ goto out_dir; |
12332 |
+ } |
12333 |
+ |
12334 |
+ f2fs_balance_fs(sbi, true); |
12335 |
+@@ -915,7 +916,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12336 |
+ err = f2fs_add_link(new_dentry, old_inode); |
12337 |
+ if (err) { |
12338 |
+ f2fs_unlock_op(sbi); |
12339 |
+- goto out_whiteout; |
12340 |
++ goto out_dir; |
12341 |
+ } |
12342 |
+ |
12343 |
+ if (old_dir_entry) |
12344 |
+@@ -939,7 +940,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12345 |
+ if (IS_ERR(old_page)) |
12346 |
+ err = PTR_ERR(old_page); |
12347 |
+ f2fs_unlock_op(sbi); |
12348 |
+- goto out_whiteout; |
12349 |
++ goto out_dir; |
12350 |
+ } |
12351 |
+ } |
12352 |
+ } |
12353 |
+@@ -958,7 +959,6 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry, |
12354 |
+ f2fs_delete_entry(old_entry, old_page, old_dir, NULL); |
12355 |
+ |
12356 |
+ if (whiteout) { |
12357 |
+- whiteout->i_state |= I_LINKABLE; |
12358 |
+ set_inode_flag(whiteout, FI_INC_LINK); |
12359 |
+ err = f2fs_add_link(old_dentry, whiteout); |
12360 |
+ if (err) |
12361 |
+@@ -992,15 +992,14 @@ put_out_dir: |
12362 |
+ f2fs_unlock_op(sbi); |
12363 |
+ if (new_page) |
12364 |
+ f2fs_put_page(new_page, 0); |
12365 |
+-out_whiteout: |
12366 |
+- if (whiteout) |
12367 |
+- iput(whiteout); |
12368 |
+ out_dir: |
12369 |
+ if (old_dir_entry) |
12370 |
+ f2fs_put_page(old_dir_page, 0); |
12371 |
+ out_old: |
12372 |
+ f2fs_put_page(old_page, 0); |
12373 |
+ out: |
12374 |
++ if (whiteout) |
12375 |
++ iput(whiteout); |
12376 |
+ return err; |
12377 |
+ } |
12378 |
+ |
12379 |
+diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c |
12380 |
+index 98887187af4c..9a59f49ba405 100644 |
12381 |
+--- a/fs/f2fs/sysfs.c |
12382 |
++++ b/fs/f2fs/sysfs.c |
12383 |
+@@ -658,10 +658,12 @@ int __init f2fs_init_sysfs(void) |
12384 |
+ |
12385 |
+ ret = kobject_init_and_add(&f2fs_feat, &f2fs_feat_ktype, |
12386 |
+ NULL, "features"); |
12387 |
+- if (ret) |
12388 |
++ if (ret) { |
12389 |
++ kobject_put(&f2fs_feat); |
12390 |
+ kset_unregister(&f2fs_kset); |
12391 |
+- else |
12392 |
++ } else { |
12393 |
+ f2fs_proc_root = proc_mkdir("fs/f2fs", NULL); |
12394 |
++ } |
12395 |
+ return ret; |
12396 |
+ } |
12397 |
+ |
12398 |
+@@ -682,8 +684,11 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi) |
12399 |
+ init_completion(&sbi->s_kobj_unregister); |
12400 |
+ err = kobject_init_and_add(&sbi->s_kobj, &f2fs_sb_ktype, NULL, |
12401 |
+ "%s", sb->s_id); |
12402 |
+- if (err) |
12403 |
++ if (err) { |
12404 |
++ kobject_put(&sbi->s_kobj); |
12405 |
++ wait_for_completion(&sbi->s_kobj_unregister); |
12406 |
+ return err; |
12407 |
++ } |
12408 |
+ |
12409 |
+ if (f2fs_proc_root) |
12410 |
+ sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); |
12411 |
+@@ -711,4 +716,5 @@ void f2fs_unregister_sysfs(struct f2fs_sb_info *sbi) |
12412 |
+ remove_proc_entry(sbi->sb->s_id, f2fs_proc_root); |
12413 |
+ } |
12414 |
+ kobject_del(&sbi->s_kobj); |
12415 |
++ kobject_put(&sbi->s_kobj); |
12416 |
+ } |
12417 |
+diff --git a/fs/jbd2/checkpoint.c b/fs/jbd2/checkpoint.c |
12418 |
+index 26f8d7e46462..66409cbd3ed5 100644 |
12419 |
+--- a/fs/jbd2/checkpoint.c |
12420 |
++++ b/fs/jbd2/checkpoint.c |
12421 |
+@@ -165,7 +165,7 @@ void __jbd2_log_wait_for_space(journal_t *journal) |
12422 |
+ "journal space in %s\n", __func__, |
12423 |
+ journal->j_devname); |
12424 |
+ WARN_ON(1); |
12425 |
+- jbd2_journal_abort(journal, 0); |
12426 |
++ jbd2_journal_abort(journal, -EIO); |
12427 |
+ } |
12428 |
+ write_lock(&journal->j_state_lock); |
12429 |
+ } else { |
12430 |
+diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c |
12431 |
+index c321fa06081c..4200a6fe9599 100644 |
12432 |
+--- a/fs/jbd2/commit.c |
12433 |
++++ b/fs/jbd2/commit.c |
12434 |
+@@ -781,7 +781,7 @@ start_journal_io: |
12435 |
+ err = journal_submit_commit_record(journal, commit_transaction, |
12436 |
+ &cbh, crc32_sum); |
12437 |
+ if (err) |
12438 |
+- __jbd2_journal_abort_hard(journal); |
12439 |
++ jbd2_journal_abort(journal, err); |
12440 |
+ } |
12441 |
+ |
12442 |
+ blk_finish_plug(&plug); |
12443 |
+@@ -874,7 +874,7 @@ start_journal_io: |
12444 |
+ err = journal_submit_commit_record(journal, commit_transaction, |
12445 |
+ &cbh, crc32_sum); |
12446 |
+ if (err) |
12447 |
+- __jbd2_journal_abort_hard(journal); |
12448 |
++ jbd2_journal_abort(journal, err); |
12449 |
+ } |
12450 |
+ if (cbh) |
12451 |
+ err = journal_wait_on_commit_record(journal, cbh); |
12452 |
+diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c |
12453 |
+index 1a2339f2cb49..a15a22d20909 100644 |
12454 |
+--- a/fs/jbd2/journal.c |
12455 |
++++ b/fs/jbd2/journal.c |
12456 |
+@@ -1701,6 +1701,11 @@ int jbd2_journal_load(journal_t *journal) |
12457 |
+ journal->j_devname); |
12458 |
+ return -EFSCORRUPTED; |
12459 |
+ } |
12460 |
++ /* |
12461 |
++ * clear JBD2_ABORT flag initialized in journal_init_common |
12462 |
++ * here to update log tail information with the newest seq. |
12463 |
++ */ |
12464 |
++ journal->j_flags &= ~JBD2_ABORT; |
12465 |
+ |
12466 |
+ /* OK, we've finished with the dynamic journal bits: |
12467 |
+ * reinitialise the dynamic contents of the superblock in memory |
12468 |
+@@ -1708,7 +1713,6 @@ int jbd2_journal_load(journal_t *journal) |
12469 |
+ if (journal_reset(journal)) |
12470 |
+ goto recovery_error; |
12471 |
+ |
12472 |
+- journal->j_flags &= ~JBD2_ABORT; |
12473 |
+ journal->j_flags |= JBD2_LOADED; |
12474 |
+ return 0; |
12475 |
+ |
12476 |
+@@ -2129,8 +2133,7 @@ static void __journal_abort_soft (journal_t *journal, int errno) |
12477 |
+ |
12478 |
+ if (journal->j_flags & JBD2_ABORT) { |
12479 |
+ write_unlock(&journal->j_state_lock); |
12480 |
+- if (!old_errno && old_errno != -ESHUTDOWN && |
12481 |
+- errno == -ESHUTDOWN) |
12482 |
++ if (old_errno != -ESHUTDOWN && errno == -ESHUTDOWN) |
12483 |
+ jbd2_journal_update_sb_errno(journal); |
12484 |
+ return; |
12485 |
+ } |
12486 |
+@@ -2138,12 +2141,10 @@ static void __journal_abort_soft (journal_t *journal, int errno) |
12487 |
+ |
12488 |
+ __jbd2_journal_abort_hard(journal); |
12489 |
+ |
12490 |
+- if (errno) { |
12491 |
+- jbd2_journal_update_sb_errno(journal); |
12492 |
+- write_lock(&journal->j_state_lock); |
12493 |
+- journal->j_flags |= JBD2_REC_ERR; |
12494 |
+- write_unlock(&journal->j_state_lock); |
12495 |
+- } |
12496 |
++ jbd2_journal_update_sb_errno(journal); |
12497 |
++ write_lock(&journal->j_state_lock); |
12498 |
++ journal->j_flags |= JBD2_REC_ERR; |
12499 |
++ write_unlock(&journal->j_state_lock); |
12500 |
+ } |
12501 |
+ |
12502 |
+ /** |
12503 |
+@@ -2185,11 +2186,6 @@ static void __journal_abort_soft (journal_t *journal, int errno) |
12504 |
+ * failure to disk. ext3_error, for example, now uses this |
12505 |
+ * functionality. |
12506 |
+ * |
12507 |
+- * Errors which originate from within the journaling layer will NOT |
12508 |
+- * supply an errno; a null errno implies that absolutely no further |
12509 |
+- * writes are done to the journal (unless there are any already in |
12510 |
+- * progress). |
12511 |
+- * |
12512 |
+ */ |
12513 |
+ |
12514 |
+ void jbd2_journal_abort(journal_t *journal, int errno) |
12515 |
+diff --git a/fs/nfs/nfs42proc.c b/fs/nfs/nfs42proc.c |
12516 |
+index 94f98e190e63..526441de89c1 100644 |
12517 |
+--- a/fs/nfs/nfs42proc.c |
12518 |
++++ b/fs/nfs/nfs42proc.c |
12519 |
+@@ -283,14 +283,14 @@ static ssize_t _nfs42_proc_copy(struct file *src, |
12520 |
+ status = handle_async_copy(res, server, src, dst, |
12521 |
+ &args->src_stateid); |
12522 |
+ if (status) |
12523 |
+- return status; |
12524 |
++ goto out; |
12525 |
+ } |
12526 |
+ |
12527 |
+ if ((!res->synchronous || !args->sync) && |
12528 |
+ res->write_res.verifier.committed != NFS_FILE_SYNC) { |
12529 |
+ status = process_copy_commit(dst, pos_dst, res); |
12530 |
+ if (status) |
12531 |
+- return status; |
12532 |
++ goto out; |
12533 |
+ } |
12534 |
+ |
12535 |
+ truncate_pagecache_range(dst_inode, pos_dst, |
12536 |
+diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h |
12537 |
+index 497a4171ef61..bfb50fc51528 100644 |
12538 |
+--- a/fs/ocfs2/journal.h |
12539 |
++++ b/fs/ocfs2/journal.h |
12540 |
+@@ -637,9 +637,11 @@ static inline void ocfs2_update_inode_fsync_trans(handle_t *handle, |
12541 |
+ { |
12542 |
+ struct ocfs2_inode_info *oi = OCFS2_I(inode); |
12543 |
+ |
12544 |
+- oi->i_sync_tid = handle->h_transaction->t_tid; |
12545 |
+- if (datasync) |
12546 |
+- oi->i_datasync_tid = handle->h_transaction->t_tid; |
12547 |
++ if (!is_handle_aborted(handle)) { |
12548 |
++ oi->i_sync_tid = handle->h_transaction->t_tid; |
12549 |
++ if (datasync) |
12550 |
++ oi->i_datasync_tid = handle->h_transaction->t_tid; |
12551 |
++ } |
12552 |
+ } |
12553 |
+ |
12554 |
+ #endif /* OCFS2_JOURNAL_H */ |
12555 |
+diff --git a/fs/orangefs/orangefs-debugfs.c b/fs/orangefs/orangefs-debugfs.c |
12556 |
+index 0732cb08173e..e24738c691f6 100644 |
12557 |
+--- a/fs/orangefs/orangefs-debugfs.c |
12558 |
++++ b/fs/orangefs/orangefs-debugfs.c |
12559 |
+@@ -305,6 +305,7 @@ static void *help_start(struct seq_file *m, loff_t *pos) |
12560 |
+ |
12561 |
+ static void *help_next(struct seq_file *m, void *v, loff_t *pos) |
12562 |
+ { |
12563 |
++ (*pos)++; |
12564 |
+ gossip_debug(GOSSIP_DEBUGFS_DEBUG, "help_next: start\n"); |
12565 |
+ |
12566 |
+ return NULL; |
12567 |
+diff --git a/fs/reiserfs/stree.c b/fs/reiserfs/stree.c |
12568 |
+index 0037aea97d39..2946713cb00d 100644 |
12569 |
+--- a/fs/reiserfs/stree.c |
12570 |
++++ b/fs/reiserfs/stree.c |
12571 |
+@@ -2250,7 +2250,8 @@ error_out: |
12572 |
+ /* also releases the path */ |
12573 |
+ unfix_nodes(&s_ins_balance); |
12574 |
+ #ifdef REISERQUOTA_DEBUG |
12575 |
+- reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE, |
12576 |
++ if (inode) |
12577 |
++ reiserfs_debug(th->t_super, REISERFS_DEBUG_CODE, |
12578 |
+ "reiserquota insert_item(): freeing %u id=%u type=%c", |
12579 |
+ quota_bytes, inode->i_uid, head2type(ih)); |
12580 |
+ #endif |
12581 |
+diff --git a/fs/reiserfs/super.c b/fs/reiserfs/super.c |
12582 |
+index 6280efeceb0a..de5eda33c92a 100644 |
12583 |
+--- a/fs/reiserfs/super.c |
12584 |
++++ b/fs/reiserfs/super.c |
12585 |
+@@ -1954,7 +1954,7 @@ static int reiserfs_fill_super(struct super_block *s, void *data, int silent) |
12586 |
+ if (!sbi->s_jdev) { |
12587 |
+ SWARN(silent, s, "", "Cannot allocate memory for " |
12588 |
+ "journal device name"); |
12589 |
+- goto error; |
12590 |
++ goto error_unlocked; |
12591 |
+ } |
12592 |
+ } |
12593 |
+ #ifdef CONFIG_QUOTA |
12594 |
+diff --git a/fs/udf/super.c b/fs/udf/super.c |
12595 |
+index 6fd0f14e9dd2..1676a175cd7a 100644 |
12596 |
+--- a/fs/udf/super.c |
12597 |
++++ b/fs/udf/super.c |
12598 |
+@@ -2469,17 +2469,29 @@ static unsigned int udf_count_free_table(struct super_block *sb, |
12599 |
+ static unsigned int udf_count_free(struct super_block *sb) |
12600 |
+ { |
12601 |
+ unsigned int accum = 0; |
12602 |
+- struct udf_sb_info *sbi; |
12603 |
++ struct udf_sb_info *sbi = UDF_SB(sb); |
12604 |
+ struct udf_part_map *map; |
12605 |
++ unsigned int part = sbi->s_partition; |
12606 |
++ int ptype = sbi->s_partmaps[part].s_partition_type; |
12607 |
++ |
12608 |
++ if (ptype == UDF_METADATA_MAP25) { |
12609 |
++ part = sbi->s_partmaps[part].s_type_specific.s_metadata. |
12610 |
++ s_phys_partition_ref; |
12611 |
++ } else if (ptype == UDF_VIRTUAL_MAP15 || ptype == UDF_VIRTUAL_MAP20) { |
12612 |
++ /* |
12613 |
++ * Filesystems with VAT are append-only and we cannot write to |
12614 |
++ * them. Let's just report 0 here. |
12615 |
++ */ |
12616 |
++ return 0; |
12617 |
++ } |
12618 |
+ |
12619 |
+- sbi = UDF_SB(sb); |
12620 |
+ if (sbi->s_lvid_bh) { |
12621 |
+ struct logicalVolIntegrityDesc *lvid = |
12622 |
+ (struct logicalVolIntegrityDesc *) |
12623 |
+ sbi->s_lvid_bh->b_data; |
12624 |
+- if (le32_to_cpu(lvid->numOfPartitions) > sbi->s_partition) { |
12625 |
++ if (le32_to_cpu(lvid->numOfPartitions) > part) { |
12626 |
+ accum = le32_to_cpu( |
12627 |
+- lvid->freeSpaceTable[sbi->s_partition]); |
12628 |
++ lvid->freeSpaceTable[part]); |
12629 |
+ if (accum == 0xFFFFFFFF) |
12630 |
+ accum = 0; |
12631 |
+ } |
12632 |
+@@ -2488,7 +2500,7 @@ static unsigned int udf_count_free(struct super_block *sb) |
12633 |
+ if (accum) |
12634 |
+ return accum; |
12635 |
+ |
12636 |
+- map = &sbi->s_partmaps[sbi->s_partition]; |
12637 |
++ map = &sbi->s_partmaps[part]; |
12638 |
+ if (map->s_partition_flags & UDF_PART_FLAG_UNALLOC_BITMAP) { |
12639 |
+ accum += udf_count_free_bitmap(sb, |
12640 |
+ map->s_uspace.s_bitmap); |
12641 |
+diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h |
12642 |
+index 0647f436f88c..50128c36f0b4 100644 |
12643 |
+--- a/include/linux/dmaengine.h |
12644 |
++++ b/include/linux/dmaengine.h |
12645 |
+@@ -686,6 +686,7 @@ struct dma_filter { |
12646 |
+ * @fill_align: alignment shift for memset operations |
12647 |
+ * @dev_id: unique device ID |
12648 |
+ * @dev: struct device reference for dma mapping api |
12649 |
++ * @owner: owner module (automatically set based on the provided dev) |
12650 |
+ * @src_addr_widths: bit mask of src addr widths the device supports |
12651 |
+ * Width is specified in bytes, e.g. for a device supporting |
12652 |
+ * a width of 4 the mask should have BIT(4) set. |
12653 |
+@@ -749,6 +750,7 @@ struct dma_device { |
12654 |
+ |
12655 |
+ int dev_id; |
12656 |
+ struct device *dev; |
12657 |
++ struct module *owner; |
12658 |
+ |
12659 |
+ u32 src_addr_widths; |
12660 |
+ u32 dst_addr_widths; |
12661 |
+diff --git a/include/linux/list_nulls.h b/include/linux/list_nulls.h |
12662 |
+index 3ef96743db8d..1ecd35664e0d 100644 |
12663 |
+--- a/include/linux/list_nulls.h |
12664 |
++++ b/include/linux/list_nulls.h |
12665 |
+@@ -72,10 +72,10 @@ static inline void hlist_nulls_add_head(struct hlist_nulls_node *n, |
12666 |
+ struct hlist_nulls_node *first = h->first; |
12667 |
+ |
12668 |
+ n->next = first; |
12669 |
+- n->pprev = &h->first; |
12670 |
++ WRITE_ONCE(n->pprev, &h->first); |
12671 |
+ h->first = n; |
12672 |
+ if (!is_a_nulls(first)) |
12673 |
+- first->pprev = &n->next; |
12674 |
++ WRITE_ONCE(first->pprev, &n->next); |
12675 |
+ } |
12676 |
+ |
12677 |
+ static inline void __hlist_nulls_del(struct hlist_nulls_node *n) |
12678 |
+@@ -85,13 +85,13 @@ static inline void __hlist_nulls_del(struct hlist_nulls_node *n) |
12679 |
+ |
12680 |
+ WRITE_ONCE(*pprev, next); |
12681 |
+ if (!is_a_nulls(next)) |
12682 |
+- next->pprev = pprev; |
12683 |
++ WRITE_ONCE(next->pprev, pprev); |
12684 |
+ } |
12685 |
+ |
12686 |
+ static inline void hlist_nulls_del(struct hlist_nulls_node *n) |
12687 |
+ { |
12688 |
+ __hlist_nulls_del(n); |
12689 |
+- n->pprev = LIST_POISON2; |
12690 |
++ WRITE_ONCE(n->pprev, LIST_POISON2); |
12691 |
+ } |
12692 |
+ |
12693 |
+ /** |
12694 |
+diff --git a/include/linux/rculist_nulls.h b/include/linux/rculist_nulls.h |
12695 |
+index 61974c4c566b..90f2e2232c6d 100644 |
12696 |
+--- a/include/linux/rculist_nulls.h |
12697 |
++++ b/include/linux/rculist_nulls.h |
12698 |
+@@ -34,7 +34,7 @@ static inline void hlist_nulls_del_init_rcu(struct hlist_nulls_node *n) |
12699 |
+ { |
12700 |
+ if (!hlist_nulls_unhashed(n)) { |
12701 |
+ __hlist_nulls_del(n); |
12702 |
+- n->pprev = NULL; |
12703 |
++ WRITE_ONCE(n->pprev, NULL); |
12704 |
+ } |
12705 |
+ } |
12706 |
+ |
12707 |
+@@ -66,7 +66,7 @@ static inline void hlist_nulls_del_init_rcu(struct hlist_nulls_node *n) |
12708 |
+ static inline void hlist_nulls_del_rcu(struct hlist_nulls_node *n) |
12709 |
+ { |
12710 |
+ __hlist_nulls_del(n); |
12711 |
+- n->pprev = LIST_POISON2; |
12712 |
++ WRITE_ONCE(n->pprev, LIST_POISON2); |
12713 |
+ } |
12714 |
+ |
12715 |
+ /** |
12716 |
+@@ -94,10 +94,10 @@ static inline void hlist_nulls_add_head_rcu(struct hlist_nulls_node *n, |
12717 |
+ struct hlist_nulls_node *first = h->first; |
12718 |
+ |
12719 |
+ n->next = first; |
12720 |
+- n->pprev = &h->first; |
12721 |
++ WRITE_ONCE(n->pprev, &h->first); |
12722 |
+ rcu_assign_pointer(hlist_nulls_first_rcu(h), n); |
12723 |
+ if (!is_a_nulls(first)) |
12724 |
+- first->pprev = &n->next; |
12725 |
++ WRITE_ONCE(first->pprev, &n->next); |
12726 |
+ } |
12727 |
+ |
12728 |
+ /** |
12729 |
+diff --git a/include/media/v4l2-device.h b/include/media/v4l2-device.h |
12730 |
+index b330e4a08a6b..40840fec337c 100644 |
12731 |
+--- a/include/media/v4l2-device.h |
12732 |
++++ b/include/media/v4l2-device.h |
12733 |
+@@ -372,7 +372,7 @@ static inline void v4l2_subdev_notify(struct v4l2_subdev *sd, |
12734 |
+ struct v4l2_subdev *__sd; \ |
12735 |
+ \ |
12736 |
+ __v4l2_device_call_subdevs_p(v4l2_dev, __sd, \ |
12737 |
+- !(grpid) || __sd->grp_id == (grpid), o, f , \ |
12738 |
++ (grpid) == 0 || __sd->grp_id == (grpid), o, f , \ |
12739 |
+ ##args); \ |
12740 |
+ } while (0) |
12741 |
+ |
12742 |
+@@ -404,7 +404,7 @@ static inline void v4l2_subdev_notify(struct v4l2_subdev *sd, |
12743 |
+ ({ \ |
12744 |
+ struct v4l2_subdev *__sd; \ |
12745 |
+ __v4l2_device_call_subdevs_until_err_p(v4l2_dev, __sd, \ |
12746 |
+- !(grpid) || __sd->grp_id == (grpid), o, f , \ |
12747 |
++ (grpid) == 0 || __sd->grp_id == (grpid), o, f , \ |
12748 |
+ ##args); \ |
12749 |
+ }) |
12750 |
+ |
12751 |
+@@ -432,8 +432,8 @@ static inline void v4l2_subdev_notify(struct v4l2_subdev *sd, |
12752 |
+ struct v4l2_subdev *__sd; \ |
12753 |
+ \ |
12754 |
+ __v4l2_device_call_subdevs_p(v4l2_dev, __sd, \ |
12755 |
+- !(grpmsk) || (__sd->grp_id & (grpmsk)), o, f , \ |
12756 |
+- ##args); \ |
12757 |
++ (grpmsk) == 0 || (__sd->grp_id & (grpmsk)), o, \ |
12758 |
++ f , ##args); \ |
12759 |
+ } while (0) |
12760 |
+ |
12761 |
+ /** |
12762 |
+@@ -463,8 +463,8 @@ static inline void v4l2_subdev_notify(struct v4l2_subdev *sd, |
12763 |
+ ({ \ |
12764 |
+ struct v4l2_subdev *__sd; \ |
12765 |
+ __v4l2_device_call_subdevs_until_err_p(v4l2_dev, __sd, \ |
12766 |
+- !(grpmsk) || (__sd->grp_id & (grpmsk)), o, f , \ |
12767 |
+- ##args); \ |
12768 |
++ (grpmsk) == 0 || (__sd->grp_id & (grpmsk)), o, \ |
12769 |
++ f , ##args); \ |
12770 |
+ }) |
12771 |
+ |
12772 |
+ |
12773 |
+diff --git a/kernel/bpf/inode.c b/kernel/bpf/inode.c |
12774 |
+index dc9d7ac8228d..c04815bb15cc 100644 |
12775 |
+--- a/kernel/bpf/inode.c |
12776 |
++++ b/kernel/bpf/inode.c |
12777 |
+@@ -198,6 +198,7 @@ static void *map_seq_next(struct seq_file *m, void *v, loff_t *pos) |
12778 |
+ void *key = map_iter(m)->key; |
12779 |
+ void *prev_key; |
12780 |
+ |
12781 |
++ (*pos)++; |
12782 |
+ if (map_iter(m)->done) |
12783 |
+ return NULL; |
12784 |
+ |
12785 |
+@@ -210,8 +211,6 @@ static void *map_seq_next(struct seq_file *m, void *v, loff_t *pos) |
12786 |
+ map_iter(m)->done = true; |
12787 |
+ return NULL; |
12788 |
+ } |
12789 |
+- |
12790 |
+- ++(*pos); |
12791 |
+ return key; |
12792 |
+ } |
12793 |
+ |
12794 |
+diff --git a/kernel/cpu.c b/kernel/cpu.c |
12795 |
+index 8d6b8b5493f9..2d850eaaf82e 100644 |
12796 |
+--- a/kernel/cpu.c |
12797 |
++++ b/kernel/cpu.c |
12798 |
+@@ -493,8 +493,7 @@ static int bringup_wait_for_ap(unsigned int cpu) |
12799 |
+ if (WARN_ON_ONCE((!cpu_online(cpu)))) |
12800 |
+ return -ECANCELED; |
12801 |
+ |
12802 |
+- /* Unpark the stopper thread and the hotplug thread of the target cpu */ |
12803 |
+- stop_machine_unpark(cpu); |
12804 |
++ /* Unpark the hotplug thread of the target cpu */ |
12805 |
+ kthread_unpark(st->thread); |
12806 |
+ |
12807 |
+ /* |
12808 |
+@@ -1048,8 +1047,8 @@ void notify_cpu_starting(unsigned int cpu) |
12809 |
+ |
12810 |
+ /* |
12811 |
+ * Called from the idle task. Wake up the controlling task which brings the |
12812 |
+- * stopper and the hotplug thread of the upcoming CPU up and then delegates |
12813 |
+- * the rest of the online bringup to the hotplug thread. |
12814 |
++ * hotplug thread of the upcoming CPU up and then delegates the rest of the |
12815 |
++ * online bringup to the hotplug thread. |
12816 |
+ */ |
12817 |
+ void cpuhp_online_idle(enum cpuhp_state state) |
12818 |
+ { |
12819 |
+@@ -1059,6 +1058,12 @@ void cpuhp_online_idle(enum cpuhp_state state) |
12820 |
+ if (state != CPUHP_AP_ONLINE_IDLE) |
12821 |
+ return; |
12822 |
+ |
12823 |
++ /* |
12824 |
++ * Unpart the stopper thread before we start the idle loop (and start |
12825 |
++ * scheduling); this ensures the stopper task is always available. |
12826 |
++ */ |
12827 |
++ stop_machine_unpark(smp_processor_id()); |
12828 |
++ |
12829 |
+ st->state = CPUHP_AP_ONLINE_IDLE; |
12830 |
+ complete_ap_thread(st, true); |
12831 |
+ } |
12832 |
+diff --git a/kernel/module.c b/kernel/module.c |
12833 |
+index 70a75a7216ab..20fc0efc679c 100644 |
12834 |
+--- a/kernel/module.c |
12835 |
++++ b/kernel/module.c |
12836 |
+@@ -2980,9 +2980,7 @@ static int setup_load_info(struct load_info *info, int flags) |
12837 |
+ |
12838 |
+ /* Try to find a name early so we can log errors with a module name */ |
12839 |
+ info->index.info = find_sec(info, ".modinfo"); |
12840 |
+- if (!info->index.info) |
12841 |
+- info->name = "(missing .modinfo section)"; |
12842 |
+- else |
12843 |
++ if (info->index.info) |
12844 |
+ info->name = get_modinfo(info, "name"); |
12845 |
+ |
12846 |
+ /* Find internal symbols and strings. */ |
12847 |
+@@ -2997,14 +2995,15 @@ static int setup_load_info(struct load_info *info, int flags) |
12848 |
+ } |
12849 |
+ |
12850 |
+ if (info->index.sym == 0) { |
12851 |
+- pr_warn("%s: module has no symbols (stripped?)\n", info->name); |
12852 |
++ pr_warn("%s: module has no symbols (stripped?)\n", |
12853 |
++ info->name ?: "(missing .modinfo section or name field)"); |
12854 |
+ return -ENOEXEC; |
12855 |
+ } |
12856 |
+ |
12857 |
+ info->index.mod = find_sec(info, ".gnu.linkonce.this_module"); |
12858 |
+ if (!info->index.mod) { |
12859 |
+ pr_warn("%s: No module found in object\n", |
12860 |
+- info->name ?: "(missing .modinfo name field)"); |
12861 |
++ info->name ?: "(missing .modinfo section or name field)"); |
12862 |
+ return -ENOEXEC; |
12863 |
+ } |
12864 |
+ /* This is temporary: point mod into copy of data. */ |
12865 |
+diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c |
12866 |
+index 53795237e975..0c379cd40bea 100644 |
12867 |
+--- a/kernel/trace/ftrace.c |
12868 |
++++ b/kernel/trace/ftrace.c |
12869 |
+@@ -6525,9 +6525,10 @@ static void *fpid_next(struct seq_file *m, void *v, loff_t *pos) |
12870 |
+ struct trace_array *tr = m->private; |
12871 |
+ struct trace_pid_list *pid_list = rcu_dereference_sched(tr->function_pids); |
12872 |
+ |
12873 |
+- if (v == FTRACE_NO_PIDS) |
12874 |
++ if (v == FTRACE_NO_PIDS) { |
12875 |
++ (*pos)++; |
12876 |
+ return NULL; |
12877 |
+- |
12878 |
++ } |
12879 |
+ return trace_pid_next(pid_list, v, pos); |
12880 |
+ } |
12881 |
+ |
12882 |
+diff --git a/kernel/trace/trace_events_trigger.c b/kernel/trace/trace_events_trigger.c |
12883 |
+index b05d1b6a6291..9300e8bbf08a 100644 |
12884 |
+--- a/kernel/trace/trace_events_trigger.c |
12885 |
++++ b/kernel/trace/trace_events_trigger.c |
12886 |
+@@ -115,9 +115,10 @@ static void *trigger_next(struct seq_file *m, void *t, loff_t *pos) |
12887 |
+ { |
12888 |
+ struct trace_event_file *event_file = event_file_data(m->private); |
12889 |
+ |
12890 |
+- if (t == SHOW_AVAILABLE_TRIGGERS) |
12891 |
++ if (t == SHOW_AVAILABLE_TRIGGERS) { |
12892 |
++ (*pos)++; |
12893 |
+ return NULL; |
12894 |
+- |
12895 |
++ } |
12896 |
+ return seq_list_next(t, &event_file->triggers, pos); |
12897 |
+ } |
12898 |
+ |
12899 |
+diff --git a/kernel/trace/trace_stat.c b/kernel/trace/trace_stat.c |
12900 |
+index 75bf1bcb4a8a..92b76f9e25ed 100644 |
12901 |
+--- a/kernel/trace/trace_stat.c |
12902 |
++++ b/kernel/trace/trace_stat.c |
12903 |
+@@ -278,18 +278,22 @@ static int tracing_stat_init(void) |
12904 |
+ |
12905 |
+ d_tracing = tracing_init_dentry(); |
12906 |
+ if (IS_ERR(d_tracing)) |
12907 |
+- return 0; |
12908 |
++ return -ENODEV; |
12909 |
+ |
12910 |
+ stat_dir = tracefs_create_dir("trace_stat", d_tracing); |
12911 |
+- if (!stat_dir) |
12912 |
++ if (!stat_dir) { |
12913 |
+ pr_warn("Could not create tracefs 'trace_stat' entry\n"); |
12914 |
++ return -ENOMEM; |
12915 |
++ } |
12916 |
+ return 0; |
12917 |
+ } |
12918 |
+ |
12919 |
+ static int init_stat_file(struct stat_session *session) |
12920 |
+ { |
12921 |
+- if (!stat_dir && tracing_stat_init()) |
12922 |
+- return -ENODEV; |
12923 |
++ int ret; |
12924 |
++ |
12925 |
++ if (!stat_dir && (ret = tracing_stat_init())) |
12926 |
++ return ret; |
12927 |
+ |
12928 |
+ session->file = tracefs_create_file(session->ts->name, 0644, |
12929 |
+ stat_dir, |
12930 |
+@@ -302,7 +306,7 @@ static int init_stat_file(struct stat_session *session) |
12931 |
+ int register_stat_tracer(struct tracer_stat *trace) |
12932 |
+ { |
12933 |
+ struct stat_session *session, *node; |
12934 |
+- int ret; |
12935 |
++ int ret = -EINVAL; |
12936 |
+ |
12937 |
+ if (!trace) |
12938 |
+ return -EINVAL; |
12939 |
+@@ -313,17 +317,15 @@ int register_stat_tracer(struct tracer_stat *trace) |
12940 |
+ /* Already registered? */ |
12941 |
+ mutex_lock(&all_stat_sessions_mutex); |
12942 |
+ list_for_each_entry(node, &all_stat_sessions, session_list) { |
12943 |
+- if (node->ts == trace) { |
12944 |
+- mutex_unlock(&all_stat_sessions_mutex); |
12945 |
+- return -EINVAL; |
12946 |
+- } |
12947 |
++ if (node->ts == trace) |
12948 |
++ goto out; |
12949 |
+ } |
12950 |
+- mutex_unlock(&all_stat_sessions_mutex); |
12951 |
+ |
12952 |
++ ret = -ENOMEM; |
12953 |
+ /* Init the session */ |
12954 |
+ session = kzalloc(sizeof(*session), GFP_KERNEL); |
12955 |
+ if (!session) |
12956 |
+- return -ENOMEM; |
12957 |
++ goto out; |
12958 |
+ |
12959 |
+ session->ts = trace; |
12960 |
+ INIT_LIST_HEAD(&session->session_list); |
12961 |
+@@ -332,15 +334,16 @@ int register_stat_tracer(struct tracer_stat *trace) |
12962 |
+ ret = init_stat_file(session); |
12963 |
+ if (ret) { |
12964 |
+ destroy_session(session); |
12965 |
+- return ret; |
12966 |
++ goto out; |
12967 |
+ } |
12968 |
+ |
12969 |
++ ret = 0; |
12970 |
+ /* Register */ |
12971 |
+- mutex_lock(&all_stat_sessions_mutex); |
12972 |
+ list_add_tail(&session->session_list, &all_stat_sessions); |
12973 |
++ out: |
12974 |
+ mutex_unlock(&all_stat_sessions_mutex); |
12975 |
+ |
12976 |
+- return 0; |
12977 |
++ return ret; |
12978 |
+ } |
12979 |
+ |
12980 |
+ void unregister_stat_tracer(struct tracer_stat *trace) |
12981 |
+diff --git a/kernel/watchdog.c b/kernel/watchdog.c |
12982 |
+index bbc4940f21af..6d60701dc636 100644 |
12983 |
+--- a/kernel/watchdog.c |
12984 |
++++ b/kernel/watchdog.c |
12985 |
+@@ -161,6 +161,8 @@ static void lockup_detector_update_enable(void) |
12986 |
+ |
12987 |
+ #ifdef CONFIG_SOFTLOCKUP_DETECTOR |
12988 |
+ |
12989 |
++#define SOFTLOCKUP_RESET ULONG_MAX |
12990 |
++ |
12991 |
+ /* Global variables, exported for sysctl */ |
12992 |
+ unsigned int __read_mostly softlockup_panic = |
12993 |
+ CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE; |
12994 |
+@@ -267,7 +269,7 @@ notrace void touch_softlockup_watchdog_sched(void) |
12995 |
+ * Preemption can be enabled. It doesn't matter which CPU's timestamp |
12996 |
+ * gets zeroed here, so use the raw_ operation. |
12997 |
+ */ |
12998 |
+- raw_cpu_write(watchdog_touch_ts, 0); |
12999 |
++ raw_cpu_write(watchdog_touch_ts, SOFTLOCKUP_RESET); |
13000 |
+ } |
13001 |
+ |
13002 |
+ notrace void touch_softlockup_watchdog(void) |
13003 |
+@@ -291,14 +293,14 @@ void touch_all_softlockup_watchdogs(void) |
13004 |
+ * the softlockup check. |
13005 |
+ */ |
13006 |
+ for_each_cpu(cpu, &watchdog_allowed_mask) |
13007 |
+- per_cpu(watchdog_touch_ts, cpu) = 0; |
13008 |
++ per_cpu(watchdog_touch_ts, cpu) = SOFTLOCKUP_RESET; |
13009 |
+ wq_watchdog_touch(-1); |
13010 |
+ } |
13011 |
+ |
13012 |
+ void touch_softlockup_watchdog_sync(void) |
13013 |
+ { |
13014 |
+ __this_cpu_write(softlockup_touch_sync, true); |
13015 |
+- __this_cpu_write(watchdog_touch_ts, 0); |
13016 |
++ __this_cpu_write(watchdog_touch_ts, SOFTLOCKUP_RESET); |
13017 |
+ } |
13018 |
+ |
13019 |
+ static int is_softlockup(unsigned long touch_ts) |
13020 |
+@@ -376,7 +378,7 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) |
13021 |
+ /* .. and repeat */ |
13022 |
+ hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period)); |
13023 |
+ |
13024 |
+- if (touch_ts == 0) { |
13025 |
++ if (touch_ts == SOFTLOCKUP_RESET) { |
13026 |
+ if (unlikely(__this_cpu_read(softlockup_touch_sync))) { |
13027 |
+ /* |
13028 |
+ * If the time stamp was touched atomically |
13029 |
+diff --git a/lib/scatterlist.c b/lib/scatterlist.c |
13030 |
+index 8c3036c37ba0..60e7eca2f4be 100644 |
13031 |
+--- a/lib/scatterlist.c |
13032 |
++++ b/lib/scatterlist.c |
13033 |
+@@ -305,7 +305,7 @@ int __sg_alloc_table(struct sg_table *table, unsigned int nents, |
13034 |
+ if (prv) |
13035 |
+ table->nents = ++table->orig_nents; |
13036 |
+ |
13037 |
+- return -ENOMEM; |
13038 |
++ return -ENOMEM; |
13039 |
+ } |
13040 |
+ |
13041 |
+ sg_init_table(sg, alloc_size); |
13042 |
+diff --git a/net/core/dev.c b/net/core/dev.c |
13043 |
+index 1c0224e8fc78..c1a3baf16957 100644 |
13044 |
+--- a/net/core/dev.c |
13045 |
++++ b/net/core/dev.c |
13046 |
+@@ -4306,14 +4306,14 @@ static u32 netif_receive_generic_xdp(struct sk_buff *skb, |
13047 |
+ /* Reinjected packets coming from act_mirred or similar should |
13048 |
+ * not get XDP generic processing. |
13049 |
+ */ |
13050 |
+- if (skb_cloned(skb) || skb_is_tc_redirected(skb)) |
13051 |
++ if (skb_is_tc_redirected(skb)) |
13052 |
+ return XDP_PASS; |
13053 |
+ |
13054 |
+ /* XDP packets must be linear and must have sufficient headroom |
13055 |
+ * of XDP_PACKET_HEADROOM bytes. This is the guarantee that also |
13056 |
+ * native XDP provides, thus we need to do it here as well. |
13057 |
+ */ |
13058 |
+- if (skb_is_nonlinear(skb) || |
13059 |
++ if (skb_cloned(skb) || skb_is_nonlinear(skb) || |
13060 |
+ skb_headroom(skb) < XDP_PACKET_HEADROOM) { |
13061 |
+ int hroom = XDP_PACKET_HEADROOM - skb_headroom(skb); |
13062 |
+ int troom = skb->tail + skb->data_len - skb->end; |
13063 |
+diff --git a/net/core/filter.c b/net/core/filter.c |
13064 |
+index 9daf1a4118b5..40b3af05c883 100644 |
13065 |
+--- a/net/core/filter.c |
13066 |
++++ b/net/core/filter.c |
13067 |
+@@ -3207,7 +3207,7 @@ static int __bpf_tx_xdp_map(struct net_device *dev_rx, void *fwd, |
13068 |
+ return err; |
13069 |
+ } |
13070 |
+ default: |
13071 |
+- break; |
13072 |
++ return -EBADRQC; |
13073 |
+ } |
13074 |
+ return 0; |
13075 |
+ } |
13076 |
+diff --git a/net/dsa/tag_qca.c b/net/dsa/tag_qca.c |
13077 |
+index 66a952118dfd..9c0dd31d4445 100644 |
13078 |
+--- a/net/dsa/tag_qca.c |
13079 |
++++ b/net/dsa/tag_qca.c |
13080 |
+@@ -41,7 +41,7 @@ static struct sk_buff *qca_tag_xmit(struct sk_buff *skb, struct net_device *dev) |
13081 |
+ struct dsa_port *dp = dsa_slave_to_port(dev); |
13082 |
+ u16 *phdr, hdr; |
13083 |
+ |
13084 |
+- if (skb_cow_head(skb, 0) < 0) |
13085 |
++ if (skb_cow_head(skb, QCA_HDR_LEN) < 0) |
13086 |
+ return NULL; |
13087 |
+ |
13088 |
+ skb_push(skb, QCA_HDR_LEN); |
13089 |
+diff --git a/net/netfilter/nft_tunnel.c b/net/netfilter/nft_tunnel.c |
13090 |
+index e5444f3ff43f..5e66042ac346 100644 |
13091 |
+--- a/net/netfilter/nft_tunnel.c |
13092 |
++++ b/net/netfilter/nft_tunnel.c |
13093 |
+@@ -218,8 +218,9 @@ static int nft_tunnel_obj_vxlan_init(const struct nlattr *attr, |
13094 |
+ } |
13095 |
+ |
13096 |
+ static const struct nla_policy nft_tunnel_opts_erspan_policy[NFTA_TUNNEL_KEY_ERSPAN_MAX + 1] = { |
13097 |
++ [NFTA_TUNNEL_KEY_ERSPAN_VERSION] = { .type = NLA_U32 }, |
13098 |
+ [NFTA_TUNNEL_KEY_ERSPAN_V1_INDEX] = { .type = NLA_U32 }, |
13099 |
+- [NFTA_TUNNEL_KEY_ERSPAN_V2_DIR] = { .type = NLA_U8 }, |
13100 |
++ [NFTA_TUNNEL_KEY_ERSPAN_V2_DIR] = { .type = NLA_U8 }, |
13101 |
+ [NFTA_TUNNEL_KEY_ERSPAN_V2_HWID] = { .type = NLA_U8 }, |
13102 |
+ }; |
13103 |
+ |
13104 |
+diff --git a/net/sched/cls_flower.c b/net/sched/cls_flower.c |
13105 |
+index 22415311f324..c006d3b89ba3 100644 |
13106 |
+--- a/net/sched/cls_flower.c |
13107 |
++++ b/net/sched/cls_flower.c |
13108 |
+@@ -486,6 +486,7 @@ static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = { |
13109 |
+ [TCA_FLOWER_KEY_ENC_IP_TTL_MASK] = { .type = NLA_U8 }, |
13110 |
+ [TCA_FLOWER_KEY_ENC_OPTS] = { .type = NLA_NESTED }, |
13111 |
+ [TCA_FLOWER_KEY_ENC_OPTS_MASK] = { .type = NLA_NESTED }, |
13112 |
++ [TCA_FLOWER_FLAGS] = { .type = NLA_U32 }, |
13113 |
+ }; |
13114 |
+ |
13115 |
+ static const struct nla_policy |
13116 |
+diff --git a/net/sched/cls_matchall.c b/net/sched/cls_matchall.c |
13117 |
+index 40be745db357..74863b0ff694 100644 |
13118 |
+--- a/net/sched/cls_matchall.c |
13119 |
++++ b/net/sched/cls_matchall.c |
13120 |
+@@ -137,6 +137,7 @@ static void *mall_get(struct tcf_proto *tp, u32 handle) |
13121 |
+ static const struct nla_policy mall_policy[TCA_MATCHALL_MAX + 1] = { |
13122 |
+ [TCA_MATCHALL_UNSPEC] = { .type = NLA_UNSPEC }, |
13123 |
+ [TCA_MATCHALL_CLASSID] = { .type = NLA_U32 }, |
13124 |
++ [TCA_MATCHALL_FLAGS] = { .type = NLA_U32 }, |
13125 |
+ }; |
13126 |
+ |
13127 |
+ static int mall_set_parms(struct net *net, struct tcf_proto *tp, |
13128 |
+diff --git a/net/smc/smc_diag.c b/net/smc/smc_diag.c |
13129 |
+index 371b4cf31fcd..2379a02c319d 100644 |
13130 |
+--- a/net/smc/smc_diag.c |
13131 |
++++ b/net/smc/smc_diag.c |
13132 |
+@@ -38,16 +38,15 @@ static void smc_diag_msg_common_fill(struct smc_diag_msg *r, struct sock *sk) |
13133 |
+ { |
13134 |
+ struct smc_sock *smc = smc_sk(sk); |
13135 |
+ |
13136 |
++ memset(r, 0, sizeof(*r)); |
13137 |
+ r->diag_family = sk->sk_family; |
13138 |
++ sock_diag_save_cookie(sk, r->id.idiag_cookie); |
13139 |
+ if (!smc->clcsock) |
13140 |
+ return; |
13141 |
+ r->id.idiag_sport = htons(smc->clcsock->sk->sk_num); |
13142 |
+ r->id.idiag_dport = smc->clcsock->sk->sk_dport; |
13143 |
+ r->id.idiag_if = smc->clcsock->sk->sk_bound_dev_if; |
13144 |
+- sock_diag_save_cookie(sk, r->id.idiag_cookie); |
13145 |
+ if (sk->sk_protocol == SMCPROTO_SMC) { |
13146 |
+- memset(&r->id.idiag_src, 0, sizeof(r->id.idiag_src)); |
13147 |
+- memset(&r->id.idiag_dst, 0, sizeof(r->id.idiag_dst)); |
13148 |
+ r->id.idiag_src[0] = smc->clcsock->sk->sk_rcv_saddr; |
13149 |
+ r->id.idiag_dst[0] = smc->clcsock->sk->sk_daddr; |
13150 |
+ #if IS_ENABLED(CONFIG_IPV6) |
13151 |
+diff --git a/scripts/Kconfig.include b/scripts/Kconfig.include |
13152 |
+index 3b2861f47709..79455ad6b386 100644 |
13153 |
+--- a/scripts/Kconfig.include |
13154 |
++++ b/scripts/Kconfig.include |
13155 |
+@@ -20,7 +20,7 @@ success = $(if-success,$(1),y,n) |
13156 |
+ |
13157 |
+ # $(cc-option,<flag>) |
13158 |
+ # Return y if the compiler supports <flag>, n otherwise |
13159 |
+-cc-option = $(success,$(CC) -Werror $(CLANG_FLAGS) $(1) -E -x c /dev/null -o /dev/null) |
13160 |
++cc-option = $(success,$(CC) -Werror $(CLANG_FLAGS) $(1) -S -x c /dev/null -o /dev/null) |
13161 |
+ |
13162 |
+ # $(ld-option,<flag>) |
13163 |
+ # Return y if the linker supports <flag>, n otherwise |
13164 |
+diff --git a/scripts/kconfig/confdata.c b/scripts/kconfig/confdata.c |
13165 |
+index 0dde19cf7486..2caf5fac102a 100644 |
13166 |
+--- a/scripts/kconfig/confdata.c |
13167 |
++++ b/scripts/kconfig/confdata.c |
13168 |
+@@ -1314,7 +1314,7 @@ bool conf_set_all_new_symbols(enum conf_def_mode mode) |
13169 |
+ |
13170 |
+ sym_calc_value(csym); |
13171 |
+ if (mode == def_random) |
13172 |
+- has_changed = randomize_choice_values(csym); |
13173 |
++ has_changed |= randomize_choice_values(csym); |
13174 |
+ else { |
13175 |
+ set_all_choice_values(csym); |
13176 |
+ has_changed = true; |
13177 |
+diff --git a/security/selinux/avc.c b/security/selinux/avc.c |
13178 |
+index 5de18a6d5c3f..d52be7b9f08c 100644 |
13179 |
+--- a/security/selinux/avc.c |
13180 |
++++ b/security/selinux/avc.c |
13181 |
+@@ -496,7 +496,7 @@ static inline int avc_xperms_audit(struct selinux_state *state, |
13182 |
+ if (likely(!audited)) |
13183 |
+ return 0; |
13184 |
+ return slow_avc_audit(state, ssid, tsid, tclass, requested, |
13185 |
+- audited, denied, result, ad, 0); |
13186 |
++ audited, denied, result, ad); |
13187 |
+ } |
13188 |
+ |
13189 |
+ static void avc_node_free(struct rcu_head *rhead) |
13190 |
+@@ -689,40 +689,37 @@ static struct avc_node *avc_insert(struct selinux_avc *avc, |
13191 |
+ struct avc_node *pos, *node = NULL; |
13192 |
+ int hvalue; |
13193 |
+ unsigned long flag; |
13194 |
++ spinlock_t *lock; |
13195 |
++ struct hlist_head *head; |
13196 |
+ |
13197 |
+ if (avc_latest_notif_update(avc, avd->seqno, 1)) |
13198 |
+- goto out; |
13199 |
++ return NULL; |
13200 |
+ |
13201 |
+ node = avc_alloc_node(avc); |
13202 |
+- if (node) { |
13203 |
+- struct hlist_head *head; |
13204 |
+- spinlock_t *lock; |
13205 |
+- int rc = 0; |
13206 |
+- |
13207 |
+- hvalue = avc_hash(ssid, tsid, tclass); |
13208 |
+- avc_node_populate(node, ssid, tsid, tclass, avd); |
13209 |
+- rc = avc_xperms_populate(node, xp_node); |
13210 |
+- if (rc) { |
13211 |
+- kmem_cache_free(avc_node_cachep, node); |
13212 |
+- return NULL; |
13213 |
+- } |
13214 |
+- head = &avc->avc_cache.slots[hvalue]; |
13215 |
+- lock = &avc->avc_cache.slots_lock[hvalue]; |
13216 |
++ if (!node) |
13217 |
++ return NULL; |
13218 |
+ |
13219 |
+- spin_lock_irqsave(lock, flag); |
13220 |
+- hlist_for_each_entry(pos, head, list) { |
13221 |
+- if (pos->ae.ssid == ssid && |
13222 |
+- pos->ae.tsid == tsid && |
13223 |
+- pos->ae.tclass == tclass) { |
13224 |
+- avc_node_replace(avc, node, pos); |
13225 |
+- goto found; |
13226 |
+- } |
13227 |
++ avc_node_populate(node, ssid, tsid, tclass, avd); |
13228 |
++ if (avc_xperms_populate(node, xp_node)) { |
13229 |
++ avc_node_kill(avc, node); |
13230 |
++ return NULL; |
13231 |
++ } |
13232 |
++ |
13233 |
++ hvalue = avc_hash(ssid, tsid, tclass); |
13234 |
++ head = &avc->avc_cache.slots[hvalue]; |
13235 |
++ lock = &avc->avc_cache.slots_lock[hvalue]; |
13236 |
++ spin_lock_irqsave(lock, flag); |
13237 |
++ hlist_for_each_entry(pos, head, list) { |
13238 |
++ if (pos->ae.ssid == ssid && |
13239 |
++ pos->ae.tsid == tsid && |
13240 |
++ pos->ae.tclass == tclass) { |
13241 |
++ avc_node_replace(avc, node, pos); |
13242 |
++ goto found; |
13243 |
+ } |
13244 |
+- hlist_add_head_rcu(&node->list, head); |
13245 |
+-found: |
13246 |
+- spin_unlock_irqrestore(lock, flag); |
13247 |
+ } |
13248 |
+-out: |
13249 |
++ hlist_add_head_rcu(&node->list, head); |
13250 |
++found: |
13251 |
++ spin_unlock_irqrestore(lock, flag); |
13252 |
+ return node; |
13253 |
+ } |
13254 |
+ |
13255 |
+@@ -766,8 +763,7 @@ static void avc_audit_post_callback(struct audit_buffer *ab, void *a) |
13256 |
+ noinline int slow_avc_audit(struct selinux_state *state, |
13257 |
+ u32 ssid, u32 tsid, u16 tclass, |
13258 |
+ u32 requested, u32 audited, u32 denied, int result, |
13259 |
+- struct common_audit_data *a, |
13260 |
+- unsigned int flags) |
13261 |
++ struct common_audit_data *a) |
13262 |
+ { |
13263 |
+ struct common_audit_data stack_data; |
13264 |
+ struct selinux_audit_data sad; |
13265 |
+@@ -777,17 +773,6 @@ noinline int slow_avc_audit(struct selinux_state *state, |
13266 |
+ a->type = LSM_AUDIT_DATA_NONE; |
13267 |
+ } |
13268 |
+ |
13269 |
+- /* |
13270 |
+- * When in a RCU walk do the audit on the RCU retry. This is because |
13271 |
+- * the collection of the dname in an inode audit message is not RCU |
13272 |
+- * safe. Note this may drop some audits when the situation changes |
13273 |
+- * during retry. However this is logically just as if the operation |
13274 |
+- * happened a little later. |
13275 |
+- */ |
13276 |
+- if ((a->type == LSM_AUDIT_DATA_INODE) && |
13277 |
+- (flags & MAY_NOT_BLOCK)) |
13278 |
+- return -ECHILD; |
13279 |
+- |
13280 |
+ sad.tclass = tclass; |
13281 |
+ sad.requested = requested; |
13282 |
+ sad.ssid = ssid; |
13283 |
+@@ -860,16 +845,14 @@ static int avc_update_node(struct selinux_avc *avc, |
13284 |
+ /* |
13285 |
+ * If we are in a non-blocking code path, e.g. VFS RCU walk, |
13286 |
+ * then we must not add permissions to a cache entry |
13287 |
+- * because we cannot safely audit the denial. Otherwise, |
13288 |
++ * because we will not audit the denial. Otherwise, |
13289 |
+ * during the subsequent blocking retry (e.g. VFS ref walk), we |
13290 |
+ * will find the permissions already granted in the cache entry |
13291 |
+ * and won't audit anything at all, leading to silent denials in |
13292 |
+ * permissive mode that only appear when in enforcing mode. |
13293 |
+ * |
13294 |
+- * See the corresponding handling in slow_avc_audit(), and the |
13295 |
+- * logic in selinux_inode_follow_link and selinux_inode_permission |
13296 |
+- * for the VFS MAY_NOT_BLOCK flag, which is transliterated into |
13297 |
+- * AVC_NONBLOCKING for avc_has_perm_noaudit(). |
13298 |
++ * See the corresponding handling of MAY_NOT_BLOCK in avc_audit() |
13299 |
++ * and selinux_inode_permission(). |
13300 |
+ */ |
13301 |
+ if (flags & AVC_NONBLOCKING) |
13302 |
+ return 0; |
13303 |
+@@ -913,7 +896,7 @@ static int avc_update_node(struct selinux_avc *avc, |
13304 |
+ if (orig->ae.xp_node) { |
13305 |
+ rc = avc_xperms_populate(node, orig->ae.xp_node); |
13306 |
+ if (rc) { |
13307 |
+- kmem_cache_free(avc_node_cachep, node); |
13308 |
++ avc_node_kill(avc, node); |
13309 |
+ goto out_unlock; |
13310 |
+ } |
13311 |
+ } |
13312 |
+diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c |
13313 |
+index 040c843968dc..c574285966f9 100644 |
13314 |
+--- a/security/selinux/hooks.c |
13315 |
++++ b/security/selinux/hooks.c |
13316 |
+@@ -3171,8 +3171,7 @@ static int selinux_inode_follow_link(struct dentry *dentry, struct inode *inode, |
13317 |
+ |
13318 |
+ static noinline int audit_inode_permission(struct inode *inode, |
13319 |
+ u32 perms, u32 audited, u32 denied, |
13320 |
+- int result, |
13321 |
+- unsigned flags) |
13322 |
++ int result) |
13323 |
+ { |
13324 |
+ struct common_audit_data ad; |
13325 |
+ struct inode_security_struct *isec = inode->i_security; |
13326 |
+@@ -3183,7 +3182,7 @@ static noinline int audit_inode_permission(struct inode *inode, |
13327 |
+ |
13328 |
+ rc = slow_avc_audit(&selinux_state, |
13329 |
+ current_sid(), isec->sid, isec->sclass, perms, |
13330 |
+- audited, denied, result, &ad, flags); |
13331 |
++ audited, denied, result, &ad); |
13332 |
+ if (rc) |
13333 |
+ return rc; |
13334 |
+ return 0; |
13335 |
+@@ -3230,7 +3229,11 @@ static int selinux_inode_permission(struct inode *inode, int mask) |
13336 |
+ if (likely(!audited)) |
13337 |
+ return rc; |
13338 |
+ |
13339 |
+- rc2 = audit_inode_permission(inode, perms, audited, denied, rc, flags); |
13340 |
++ /* fall back to ref-walk if we have to generate audit */ |
13341 |
++ if (flags & MAY_NOT_BLOCK) |
13342 |
++ return -ECHILD; |
13343 |
++ |
13344 |
++ rc2 = audit_inode_permission(inode, perms, audited, denied, rc); |
13345 |
+ if (rc2) |
13346 |
+ return rc2; |
13347 |
+ return rc; |
13348 |
+diff --git a/security/selinux/include/avc.h b/security/selinux/include/avc.h |
13349 |
+index 74ea50977c20..cf4cc3ef959b 100644 |
13350 |
+--- a/security/selinux/include/avc.h |
13351 |
++++ b/security/selinux/include/avc.h |
13352 |
+@@ -100,8 +100,7 @@ static inline u32 avc_audit_required(u32 requested, |
13353 |
+ int slow_avc_audit(struct selinux_state *state, |
13354 |
+ u32 ssid, u32 tsid, u16 tclass, |
13355 |
+ u32 requested, u32 audited, u32 denied, int result, |
13356 |
+- struct common_audit_data *a, |
13357 |
+- unsigned flags); |
13358 |
++ struct common_audit_data *a); |
13359 |
+ |
13360 |
+ /** |
13361 |
+ * avc_audit - Audit the granting or denial of permissions. |
13362 |
+@@ -135,9 +134,12 @@ static inline int avc_audit(struct selinux_state *state, |
13363 |
+ audited = avc_audit_required(requested, avd, result, 0, &denied); |
13364 |
+ if (likely(!audited)) |
13365 |
+ return 0; |
13366 |
++ /* fall back to ref-walk if we have to generate audit */ |
13367 |
++ if (flags & MAY_NOT_BLOCK) |
13368 |
++ return -ECHILD; |
13369 |
+ return slow_avc_audit(state, ssid, tsid, tclass, |
13370 |
+ requested, audited, denied, result, |
13371 |
+- a, flags); |
13372 |
++ a); |
13373 |
+ } |
13374 |
+ |
13375 |
+ #define AVC_STRICT 1 /* Ignore permissive mode. */ |
13376 |
+diff --git a/sound/core/control.c b/sound/core/control.c |
13377 |
+index 649d3217590e..d1312f14d78f 100644 |
13378 |
+--- a/sound/core/control.c |
13379 |
++++ b/sound/core/control.c |
13380 |
+@@ -1468,8 +1468,9 @@ static int call_tlv_handler(struct snd_ctl_file *file, int op_flag, |
13381 |
+ if (kctl->tlv.c == NULL) |
13382 |
+ return -ENXIO; |
13383 |
+ |
13384 |
+- /* When locked, this is unavailable. */ |
13385 |
+- if (vd->owner != NULL && vd->owner != file) |
13386 |
++ /* Write and command operations are not allowed for locked element. */ |
13387 |
++ if (op_flag != SNDRV_CTL_TLV_OP_READ && |
13388 |
++ vd->owner != NULL && vd->owner != file) |
13389 |
+ return -EPERM; |
13390 |
+ |
13391 |
+ return kctl->tlv.c(kctl, op_flag, size, buf); |
13392 |
+diff --git a/sound/pci/hda/patch_conexant.c b/sound/pci/hda/patch_conexant.c |
13393 |
+index 5500dd437b44..78bb96263bc2 100644 |
13394 |
+--- a/sound/pci/hda/patch_conexant.c |
13395 |
++++ b/sound/pci/hda/patch_conexant.c |
13396 |
+@@ -935,6 +935,7 @@ static const struct snd_pci_quirk cxt5066_fixups[] = { |
13397 |
+ SND_PCI_QUIRK(0x17aa, 0x215f, "Lenovo T510", CXT_PINCFG_LENOVO_TP410), |
13398 |
+ SND_PCI_QUIRK(0x17aa, 0x21ce, "Lenovo T420", CXT_PINCFG_LENOVO_TP410), |
13399 |
+ SND_PCI_QUIRK(0x17aa, 0x21cf, "Lenovo T520", CXT_PINCFG_LENOVO_TP410), |
13400 |
++ SND_PCI_QUIRK(0x17aa, 0x21d2, "Lenovo T420s", CXT_PINCFG_LENOVO_TP410), |
13401 |
+ SND_PCI_QUIRK(0x17aa, 0x21da, "Lenovo X220", CXT_PINCFG_LENOVO_TP410), |
13402 |
+ SND_PCI_QUIRK(0x17aa, 0x21db, "Lenovo X220-tablet", CXT_PINCFG_LENOVO_TP410), |
13403 |
+ SND_PCI_QUIRK(0x17aa, 0x38af, "Lenovo IdeaPad Z560", CXT_FIXUP_MUTE_LED_EAPD), |
13404 |
+diff --git a/sound/pci/hda/patch_hdmi.c b/sound/pci/hda/patch_hdmi.c |
13405 |
+index c827a2a89cc3..c67fadd5aae5 100644 |
13406 |
+--- a/sound/pci/hda/patch_hdmi.c |
13407 |
++++ b/sound/pci/hda/patch_hdmi.c |
13408 |
+@@ -2604,9 +2604,12 @@ static int alloc_intel_hdmi(struct hda_codec *codec) |
13409 |
+ /* parse and post-process for Intel codecs */ |
13410 |
+ static int parse_intel_hdmi(struct hda_codec *codec) |
13411 |
+ { |
13412 |
+- int err; |
13413 |
++ int err, retries = 3; |
13414 |
++ |
13415 |
++ do { |
13416 |
++ err = hdmi_parse_codec(codec); |
13417 |
++ } while (err < 0 && retries--); |
13418 |
+ |
13419 |
+- err = hdmi_parse_codec(codec); |
13420 |
+ if (err < 0) { |
13421 |
+ generic_spec_free(codec); |
13422 |
+ return err; |
13423 |
+diff --git a/sound/sh/aica.c b/sound/sh/aica.c |
13424 |
+index ad3f71358486..69ac44b33560 100644 |
13425 |
+--- a/sound/sh/aica.c |
13426 |
++++ b/sound/sh/aica.c |
13427 |
+@@ -117,10 +117,10 @@ static void spu_memset(u32 toi, u32 what, int length) |
13428 |
+ } |
13429 |
+ |
13430 |
+ /* spu_memload - write to SPU address space */ |
13431 |
+-static void spu_memload(u32 toi, void *from, int length) |
13432 |
++static void spu_memload(u32 toi, const void *from, int length) |
13433 |
+ { |
13434 |
+ unsigned long flags; |
13435 |
+- u32 *froml = from; |
13436 |
++ const u32 *froml = from; |
13437 |
+ u32 __iomem *to = (u32 __iomem *) (SPU_MEMORY_BASE + toi); |
13438 |
+ int i; |
13439 |
+ u32 val; |
13440 |
+diff --git a/sound/sh/sh_dac_audio.c b/sound/sh/sh_dac_audio.c |
13441 |
+index 834b2574786f..6251b5e1b64a 100644 |
13442 |
+--- a/sound/sh/sh_dac_audio.c |
13443 |
++++ b/sound/sh/sh_dac_audio.c |
13444 |
+@@ -190,7 +190,6 @@ static int snd_sh_dac_pcm_copy(struct snd_pcm_substream *substream, |
13445 |
+ { |
13446 |
+ /* channel is not used (interleaved data) */ |
13447 |
+ struct snd_sh_dac *chip = snd_pcm_substream_chip(substream); |
13448 |
+- struct snd_pcm_runtime *runtime = substream->runtime; |
13449 |
+ |
13450 |
+ if (copy_from_user_toio(chip->data_buffer + pos, src, count)) |
13451 |
+ return -EFAULT; |
13452 |
+@@ -210,7 +209,6 @@ static int snd_sh_dac_pcm_copy_kernel(struct snd_pcm_substream *substream, |
13453 |
+ { |
13454 |
+ /* channel is not used (interleaved data) */ |
13455 |
+ struct snd_sh_dac *chip = snd_pcm_substream_chip(substream); |
13456 |
+- struct snd_pcm_runtime *runtime = substream->runtime; |
13457 |
+ |
13458 |
+ memcpy_toio(chip->data_buffer + pos, src, count); |
13459 |
+ chip->buffer_end = chip->data_buffer + pos + count; |
13460 |
+@@ -229,7 +227,6 @@ static int snd_sh_dac_pcm_silence(struct snd_pcm_substream *substream, |
13461 |
+ { |
13462 |
+ /* channel is not used (interleaved data) */ |
13463 |
+ struct snd_sh_dac *chip = snd_pcm_substream_chip(substream); |
13464 |
+- struct snd_pcm_runtime *runtime = substream->runtime; |
13465 |
+ |
13466 |
+ memset_io(chip->data_buffer + pos, 0, count); |
13467 |
+ chip->buffer_end = chip->data_buffer + pos + count; |
13468 |
+diff --git a/sound/soc/atmel/Kconfig b/sound/soc/atmel/Kconfig |
13469 |
+index 64b784e96f84..dad778e5884b 100644 |
13470 |
+--- a/sound/soc/atmel/Kconfig |
13471 |
++++ b/sound/soc/atmel/Kconfig |
13472 |
+@@ -25,6 +25,8 @@ config SND_ATMEL_SOC_DMA |
13473 |
+ |
13474 |
+ config SND_ATMEL_SOC_SSC_DMA |
13475 |
+ tristate |
13476 |
++ select SND_ATMEL_SOC_DMA |
13477 |
++ select SND_ATMEL_SOC_PDC |
13478 |
+ |
13479 |
+ config SND_ATMEL_SOC_SSC |
13480 |
+ tristate |
13481 |
+diff --git a/sound/usb/usx2y/usX2Yhwdep.c b/sound/usb/usx2y/usX2Yhwdep.c |
13482 |
+index c1dd9a7b48df..36b345970364 100644 |
13483 |
+--- a/sound/usb/usx2y/usX2Yhwdep.c |
13484 |
++++ b/sound/usb/usx2y/usX2Yhwdep.c |
13485 |
+@@ -131,7 +131,7 @@ static int snd_usX2Y_hwdep_dsp_status(struct snd_hwdep *hw, |
13486 |
+ info->num_dsps = 2; // 0: Prepad Data, 1: FPGA Code |
13487 |
+ if (us428->chip_status & USX2Y_STAT_CHIP_INIT) |
13488 |
+ info->chip_ready = 1; |
13489 |
+- info->version = USX2Y_DRIVER_VERSION; |
13490 |
++ info->version = USX2Y_DRIVER_VERSION; |
13491 |
+ return 0; |
13492 |
+ } |
13493 |
+ |
13494 |
+diff --git a/tools/lib/api/fs/fs.c b/tools/lib/api/fs/fs.c |
13495 |
+index 7aba8243a0e7..bd021a0eeef8 100644 |
13496 |
+--- a/tools/lib/api/fs/fs.c |
13497 |
++++ b/tools/lib/api/fs/fs.c |
13498 |
+@@ -210,6 +210,7 @@ static bool fs__env_override(struct fs *fs) |
13499 |
+ size_t name_len = strlen(fs->name); |
13500 |
+ /* name + "_PATH" + '\0' */ |
13501 |
+ char upper_name[name_len + 5 + 1]; |
13502 |
++ |
13503 |
+ memcpy(upper_name, fs->name, name_len); |
13504 |
+ mem_toupper(upper_name, name_len); |
13505 |
+ strcpy(&upper_name[name_len], "_PATH"); |
13506 |
+@@ -219,7 +220,8 @@ static bool fs__env_override(struct fs *fs) |
13507 |
+ return false; |
13508 |
+ |
13509 |
+ fs->found = true; |
13510 |
+- strncpy(fs->path, override_path, sizeof(fs->path)); |
13511 |
++ strncpy(fs->path, override_path, sizeof(fs->path) - 1); |
13512 |
++ fs->path[sizeof(fs->path) - 1] = '\0'; |
13513 |
+ return true; |
13514 |
+ } |
13515 |
+ |
13516 |
+diff --git a/tools/objtool/arch/x86/lib/x86-opcode-map.txt b/tools/objtool/arch/x86/lib/x86-opcode-map.txt |
13517 |
+index 0a0e9112f284..5cb9f009f2be 100644 |
13518 |
+--- a/tools/objtool/arch/x86/lib/x86-opcode-map.txt |
13519 |
++++ b/tools/objtool/arch/x86/lib/x86-opcode-map.txt |
13520 |
+@@ -909,7 +909,7 @@ EndTable |
13521 |
+ |
13522 |
+ GrpTable: Grp3_2 |
13523 |
+ 0: TEST Ev,Iz |
13524 |
+-1: |
13525 |
++1: TEST Ev,Iz |
13526 |
+ 2: NOT Ev |
13527 |
+ 3: NEG Ev |
13528 |
+ 4: MUL rAX,Ev |
13529 |
+diff --git a/tools/testing/selftests/bpf/test_select_reuseport.c b/tools/testing/selftests/bpf/test_select_reuseport.c |
13530 |
+index 75646d9b34aa..cdbbdab2725f 100644 |
13531 |
+--- a/tools/testing/selftests/bpf/test_select_reuseport.c |
13532 |
++++ b/tools/testing/selftests/bpf/test_select_reuseport.c |
13533 |
+@@ -30,7 +30,7 @@ |
13534 |
+ #define REUSEPORT_ARRAY_SIZE 32 |
13535 |
+ |
13536 |
+ static int result_map, tmp_index_ovr_map, linum_map, data_check_map; |
13537 |
+-static enum result expected_results[NR_RESULTS]; |
13538 |
++static __u32 expected_results[NR_RESULTS]; |
13539 |
+ static int sk_fds[REUSEPORT_ARRAY_SIZE]; |
13540 |
+ static int reuseport_array, outer_map; |
13541 |
+ static int select_by_skb_data_prog; |
13542 |
+@@ -610,7 +610,19 @@ static void setup_per_test(int type, unsigned short family, bool inany) |
13543 |
+ |
13544 |
+ static void cleanup_per_test(void) |
13545 |
+ { |
13546 |
+- int i, err; |
13547 |
++ int i, err, zero = 0; |
13548 |
++ |
13549 |
++ memset(expected_results, 0, sizeof(expected_results)); |
13550 |
++ |
13551 |
++ for (i = 0; i < NR_RESULTS; i++) { |
13552 |
++ err = bpf_map_update_elem(result_map, &i, &zero, BPF_ANY); |
13553 |
++ RET_IF(err, "reset elem in result_map", |
13554 |
++ "i:%u err:%d errno:%d\n", i, err, errno); |
13555 |
++ } |
13556 |
++ |
13557 |
++ err = bpf_map_update_elem(linum_map, &zero, &zero, BPF_ANY); |
13558 |
++ RET_IF(err, "reset line number in linum_map", "err:%d errno:%d\n", |
13559 |
++ err, errno); |
13560 |
+ |
13561 |
+ for (i = 0; i < REUSEPORT_ARRAY_SIZE; i++) |
13562 |
+ close(sk_fds[i]); |
13563 |
+diff --git a/tools/testing/selftests/size/get_size.c b/tools/testing/selftests/size/get_size.c |
13564 |
+index d4b59ab979a0..f55943b6d1e2 100644 |
13565 |
+--- a/tools/testing/selftests/size/get_size.c |
13566 |
++++ b/tools/testing/selftests/size/get_size.c |
13567 |
+@@ -12,23 +12,35 @@ |
13568 |
+ * own execution. It also attempts to have as few dependencies |
13569 |
+ * on kernel features as possible. |
13570 |
+ * |
13571 |
+- * It should be statically linked, with startup libs avoided. |
13572 |
+- * It uses no library calls, and only the following 3 syscalls: |
13573 |
++ * It should be statically linked, with startup libs avoided. It uses |
13574 |
++ * no library calls except the syscall() function for the following 3 |
13575 |
++ * syscalls: |
13576 |
+ * sysinfo(), write(), and _exit() |
13577 |
+ * |
13578 |
+ * For output, it avoids printf (which in some C libraries |
13579 |
+ * has large external dependencies) by implementing it's own |
13580 |
+ * number output and print routines, and using __builtin_strlen() |
13581 |
++ * |
13582 |
++ * The test may crash if any of the above syscalls fails because in some |
13583 |
++ * libc implementations (e.g. the GNU C Library) errno is saved in |
13584 |
++ * thread-local storage, which does not get initialized due to avoiding |
13585 |
++ * startup libs. |
13586 |
+ */ |
13587 |
+ |
13588 |
+ #include <sys/sysinfo.h> |
13589 |
+ #include <unistd.h> |
13590 |
++#include <sys/syscall.h> |
13591 |
+ |
13592 |
+ #define STDOUT_FILENO 1 |
13593 |
+ |
13594 |
+ static int print(const char *s) |
13595 |
+ { |
13596 |
+- return write(STDOUT_FILENO, s, __builtin_strlen(s)); |
13597 |
++ size_t len = 0; |
13598 |
++ |
13599 |
++ while (s[len] != '\0') |
13600 |
++ len++; |
13601 |
++ |
13602 |
++ return syscall(SYS_write, STDOUT_FILENO, s, len); |
13603 |
+ } |
13604 |
+ |
13605 |
+ static inline char *num_to_str(unsigned long num, char *buf, int len) |
13606 |
+@@ -80,12 +92,12 @@ void _start(void) |
13607 |
+ print("TAP version 13\n"); |
13608 |
+ print("# Testing system size.\n"); |
13609 |
+ |
13610 |
+- ccode = sysinfo(&info); |
13611 |
++ ccode = syscall(SYS_sysinfo, &info); |
13612 |
+ if (ccode < 0) { |
13613 |
+ print("not ok 1"); |
13614 |
+ print(test_name); |
13615 |
+ print(" ---\n reason: \"could not get sysinfo\"\n ...\n"); |
13616 |
+- _exit(ccode); |
13617 |
++ syscall(SYS_exit, ccode); |
13618 |
+ } |
13619 |
+ print("ok 1"); |
13620 |
+ print(test_name); |
13621 |
+@@ -101,5 +113,5 @@ void _start(void) |
13622 |
+ print(" ...\n"); |
13623 |
+ print("1..1\n"); |
13624 |
+ |
13625 |
+- _exit(0); |
13626 |
++ syscall(SYS_exit, 0); |
13627 |
+ } |
13628 |
+diff --git a/tools/usb/usbip/src/usbip_network.c b/tools/usb/usbip/src/usbip_network.c |
13629 |
+index 8ffcd47d9638..902f55208e23 100644 |
13630 |
+--- a/tools/usb/usbip/src/usbip_network.c |
13631 |
++++ b/tools/usb/usbip/src/usbip_network.c |
13632 |
+@@ -62,39 +62,39 @@ void usbip_setup_port_number(char *arg) |
13633 |
+ info("using port %d (\"%s\")", usbip_port, usbip_port_string); |
13634 |
+ } |
13635 |
+ |
13636 |
+-void usbip_net_pack_uint32_t(int pack, uint32_t *num) |
13637 |
++uint32_t usbip_net_pack_uint32_t(int pack, uint32_t num) |
13638 |
+ { |
13639 |
+ uint32_t i; |
13640 |
+ |
13641 |
+ if (pack) |
13642 |
+- i = htonl(*num); |
13643 |
++ i = htonl(num); |
13644 |
+ else |
13645 |
+- i = ntohl(*num); |
13646 |
++ i = ntohl(num); |
13647 |
+ |
13648 |
+- *num = i; |
13649 |
++ return i; |
13650 |
+ } |
13651 |
+ |
13652 |
+-void usbip_net_pack_uint16_t(int pack, uint16_t *num) |
13653 |
++uint16_t usbip_net_pack_uint16_t(int pack, uint16_t num) |
13654 |
+ { |
13655 |
+ uint16_t i; |
13656 |
+ |
13657 |
+ if (pack) |
13658 |
+- i = htons(*num); |
13659 |
++ i = htons(num); |
13660 |
+ else |
13661 |
+- i = ntohs(*num); |
13662 |
++ i = ntohs(num); |
13663 |
+ |
13664 |
+- *num = i; |
13665 |
++ return i; |
13666 |
+ } |
13667 |
+ |
13668 |
+ void usbip_net_pack_usb_device(int pack, struct usbip_usb_device *udev) |
13669 |
+ { |
13670 |
+- usbip_net_pack_uint32_t(pack, &udev->busnum); |
13671 |
+- usbip_net_pack_uint32_t(pack, &udev->devnum); |
13672 |
+- usbip_net_pack_uint32_t(pack, &udev->speed); |
13673 |
++ udev->busnum = usbip_net_pack_uint32_t(pack, udev->busnum); |
13674 |
++ udev->devnum = usbip_net_pack_uint32_t(pack, udev->devnum); |
13675 |
++ udev->speed = usbip_net_pack_uint32_t(pack, udev->speed); |
13676 |
+ |
13677 |
+- usbip_net_pack_uint16_t(pack, &udev->idVendor); |
13678 |
+- usbip_net_pack_uint16_t(pack, &udev->idProduct); |
13679 |
+- usbip_net_pack_uint16_t(pack, &udev->bcdDevice); |
13680 |
++ udev->idVendor = usbip_net_pack_uint16_t(pack, udev->idVendor); |
13681 |
++ udev->idProduct = usbip_net_pack_uint16_t(pack, udev->idProduct); |
13682 |
++ udev->bcdDevice = usbip_net_pack_uint16_t(pack, udev->bcdDevice); |
13683 |
+ } |
13684 |
+ |
13685 |
+ void usbip_net_pack_usb_interface(int pack __attribute__((unused)), |
13686 |
+@@ -141,6 +141,14 @@ ssize_t usbip_net_send(int sockfd, void *buff, size_t bufflen) |
13687 |
+ return usbip_net_xmit(sockfd, buff, bufflen, 1); |
13688 |
+ } |
13689 |
+ |
13690 |
++static inline void usbip_net_pack_op_common(int pack, |
13691 |
++ struct op_common *op_common) |
13692 |
++{ |
13693 |
++ op_common->version = usbip_net_pack_uint16_t(pack, op_common->version); |
13694 |
++ op_common->code = usbip_net_pack_uint16_t(pack, op_common->code); |
13695 |
++ op_common->status = usbip_net_pack_uint32_t(pack, op_common->status); |
13696 |
++} |
13697 |
++ |
13698 |
+ int usbip_net_send_op_common(int sockfd, uint32_t code, uint32_t status) |
13699 |
+ { |
13700 |
+ struct op_common op_common; |
13701 |
+@@ -152,7 +160,7 @@ int usbip_net_send_op_common(int sockfd, uint32_t code, uint32_t status) |
13702 |
+ op_common.code = code; |
13703 |
+ op_common.status = status; |
13704 |
+ |
13705 |
+- PACK_OP_COMMON(1, &op_common); |
13706 |
++ usbip_net_pack_op_common(1, &op_common); |
13707 |
+ |
13708 |
+ rc = usbip_net_send(sockfd, &op_common, sizeof(op_common)); |
13709 |
+ if (rc < 0) { |
13710 |
+@@ -176,7 +184,7 @@ int usbip_net_recv_op_common(int sockfd, uint16_t *code, int *status) |
13711 |
+ goto err; |
13712 |
+ } |
13713 |
+ |
13714 |
+- PACK_OP_COMMON(0, &op_common); |
13715 |
++ usbip_net_pack_op_common(0, &op_common); |
13716 |
+ |
13717 |
+ if (op_common.version != USBIP_VERSION) { |
13718 |
+ err("USBIP Kernel and tool version mismatch: %d %d:", |
13719 |
+diff --git a/tools/usb/usbip/src/usbip_network.h b/tools/usb/usbip/src/usbip_network.h |
13720 |
+index 555215eae43e..83b4c5344f72 100644 |
13721 |
+--- a/tools/usb/usbip/src/usbip_network.h |
13722 |
++++ b/tools/usb/usbip/src/usbip_network.h |
13723 |
+@@ -32,12 +32,6 @@ struct op_common { |
13724 |
+ |
13725 |
+ } __attribute__((packed)); |
13726 |
+ |
13727 |
+-#define PACK_OP_COMMON(pack, op_common) do {\ |
13728 |
+- usbip_net_pack_uint16_t(pack, &(op_common)->version);\ |
13729 |
+- usbip_net_pack_uint16_t(pack, &(op_common)->code);\ |
13730 |
+- usbip_net_pack_uint32_t(pack, &(op_common)->status);\ |
13731 |
+-} while (0) |
13732 |
+- |
13733 |
+ /* ---------------------------------------------------------------------- */ |
13734 |
+ /* Dummy Code */ |
13735 |
+ #define OP_UNSPEC 0x00 |
13736 |
+@@ -163,11 +157,11 @@ struct op_devlist_reply_extra { |
13737 |
+ } while (0) |
13738 |
+ |
13739 |
+ #define PACK_OP_DEVLIST_REPLY(pack, reply) do {\ |
13740 |
+- usbip_net_pack_uint32_t(pack, &(reply)->ndev);\ |
13741 |
++ (reply)->ndev = usbip_net_pack_uint32_t(pack, (reply)->ndev);\ |
13742 |
+ } while (0) |
13743 |
+ |
13744 |
+-void usbip_net_pack_uint32_t(int pack, uint32_t *num); |
13745 |
+-void usbip_net_pack_uint16_t(int pack, uint16_t *num); |
13746 |
++uint32_t usbip_net_pack_uint32_t(int pack, uint32_t num); |
13747 |
++uint16_t usbip_net_pack_uint16_t(int pack, uint16_t num); |
13748 |
+ void usbip_net_pack_usb_device(int pack, struct usbip_usb_device *udev); |
13749 |
+ void usbip_net_pack_usb_interface(int pack, struct usbip_usb_interface *uinf); |
13750 |
+ |