1 |
commit: 52f0bee9c888fcd51ddadcf3fa91689ea0a2da70 |
2 |
Author: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
3 |
AuthorDate: Thu May 2 10:13:36 2019 +0000 |
4 |
Commit: Mike Pagano <mpagano <AT> gentoo <DOT> org> |
5 |
CommitDate: Thu May 2 10:13:36 2019 +0000 |
6 |
URL: https://gitweb.gentoo.org/proj/linux-patches.git/commit/?id=52f0bee9 |
7 |
|
8 |
Linux patch 4.19.38 |
9 |
|
10 |
Signed-off-by: Mike Pagano <mpagano <AT> gentoo.org> |
11 |
|
12 |
0000_README | 4 + |
13 |
1037_linux-4.19.38.patch | 4023 ++++++++++++++++++++++++++++++++++++++++++++++ |
14 |
2 files changed, 4027 insertions(+) |
15 |
|
16 |
diff --git a/0000_README b/0000_README |
17 |
index 2acba8b..ad3e29c 100644 |
18 |
--- a/0000_README |
19 |
+++ b/0000_README |
20 |
@@ -191,6 +191,10 @@ Patch: 1036_linux-4.19.37.patch |
21 |
From: http://www.kernel.org |
22 |
Desc: Linux 4.19.37 |
23 |
|
24 |
+Patch: 1037_linux-4.19.38.patch |
25 |
+From: http://www.kernel.org |
26 |
+Desc: Linux 4.19.38 |
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/1037_linux-4.19.38.patch b/1037_linux-4.19.38.patch |
33 |
new file mode 100644 |
34 |
index 0000000..71a6d7d |
35 |
--- /dev/null |
36 |
+++ b/1037_linux-4.19.38.patch |
37 |
@@ -0,0 +1,4023 @@ |
38 |
+diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt |
39 |
+index f5acf35c712f..8b6567f7cb9b 100644 |
40 |
+--- a/Documentation/admin-guide/kernel-parameters.txt |
41 |
++++ b/Documentation/admin-guide/kernel-parameters.txt |
42 |
+@@ -2805,7 +2805,7 @@ |
43 |
+ check bypass). With this option data leaks are possible |
44 |
+ in the system. |
45 |
+ |
46 |
+- nospectre_v2 [X86] Disable all mitigations for the Spectre variant 2 |
47 |
++ nospectre_v2 [X86,PPC_FSL_BOOK3E] Disable all mitigations for the Spectre variant 2 |
48 |
+ (indirect branch prediction) vulnerability. System may |
49 |
+ allow data leaks with this option, which is equivalent |
50 |
+ to spectre_v2=off. |
51 |
+diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt |
52 |
+index 960de8fe3f40..2c31208528d5 100644 |
53 |
+--- a/Documentation/networking/ip-sysctl.txt |
54 |
++++ b/Documentation/networking/ip-sysctl.txt |
55 |
+@@ -410,6 +410,7 @@ tcp_min_rtt_wlen - INTEGER |
56 |
+ minimum RTT when it is moved to a longer path (e.g., due to traffic |
57 |
+ engineering). A longer window makes the filter more resistant to RTT |
58 |
+ inflations such as transient congestion. The unit is seconds. |
59 |
++ Possible values: 0 - 86400 (1 day) |
60 |
+ Default: 300 |
61 |
+ |
62 |
+ tcp_moderate_rcvbuf - BOOLEAN |
63 |
+diff --git a/Makefile b/Makefile |
64 |
+index 7b495cad8c2e..14d4aeb48907 100644 |
65 |
+--- a/Makefile |
66 |
++++ b/Makefile |
67 |
+@@ -1,7 +1,7 @@ |
68 |
+ # SPDX-License-Identifier: GPL-2.0 |
69 |
+ VERSION = 4 |
70 |
+ PATCHLEVEL = 19 |
71 |
+-SUBLEVEL = 37 |
72 |
++SUBLEVEL = 38 |
73 |
+ EXTRAVERSION = |
74 |
+ NAME = "People's Front" |
75 |
+ |
76 |
+diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S |
77 |
+index 517e0e18f0b8..e205bbbe2794 100644 |
78 |
+--- a/arch/arm/boot/compressed/head.S |
79 |
++++ b/arch/arm/boot/compressed/head.S |
80 |
+@@ -1395,7 +1395,21 @@ ENTRY(efi_stub_entry) |
81 |
+ |
82 |
+ @ Preserve return value of efi_entry() in r4 |
83 |
+ mov r4, r0 |
84 |
+- bl cache_clean_flush |
85 |
++ |
86 |
++ @ our cache maintenance code relies on CP15 barrier instructions |
87 |
++ @ but since we arrived here with the MMU and caches configured |
88 |
++ @ by UEFI, we must check that the CP15BEN bit is set in SCTLR. |
89 |
++ @ Note that this bit is RAO/WI on v6 and earlier, so the ISB in |
90 |
++ @ the enable path will be executed on v7+ only. |
91 |
++ mrc p15, 0, r1, c1, c0, 0 @ read SCTLR |
92 |
++ tst r1, #(1 << 5) @ CP15BEN bit set? |
93 |
++ bne 0f |
94 |
++ orr r1, r1, #(1 << 5) @ CP15 barrier instructions |
95 |
++ mcr p15, 0, r1, c1, c0, 0 @ write SCTLR |
96 |
++ ARM( .inst 0xf57ff06f @ v7+ isb ) |
97 |
++ THUMB( isb ) |
98 |
++ |
99 |
++0: bl cache_clean_flush |
100 |
+ bl cache_off |
101 |
+ |
102 |
+ @ Set parameters for booting zImage according to boot protocol |
103 |
+diff --git a/arch/mips/kernel/scall64-o32.S b/arch/mips/kernel/scall64-o32.S |
104 |
+index 73913f072e39..579608342ac6 100644 |
105 |
+--- a/arch/mips/kernel/scall64-o32.S |
106 |
++++ b/arch/mips/kernel/scall64-o32.S |
107 |
+@@ -125,7 +125,7 @@ trace_a_syscall: |
108 |
+ subu t1, v0, __NR_O32_Linux |
109 |
+ move a1, v0 |
110 |
+ bnez t1, 1f /* __NR_syscall at offset 0 */ |
111 |
+- lw a1, PT_R4(sp) /* Arg1 for __NR_syscall case */ |
112 |
++ ld a1, PT_R4(sp) /* Arg1 for __NR_syscall case */ |
113 |
+ .set pop |
114 |
+ |
115 |
+ 1: jal syscall_trace_enter |
116 |
+diff --git a/arch/powerpc/configs/skiroot_defconfig b/arch/powerpc/configs/skiroot_defconfig |
117 |
+index 6bd5e7261335..ffeaed63675b 100644 |
118 |
+--- a/arch/powerpc/configs/skiroot_defconfig |
119 |
++++ b/arch/powerpc/configs/skiroot_defconfig |
120 |
+@@ -195,6 +195,7 @@ CONFIG_UDF_FS=m |
121 |
+ CONFIG_MSDOS_FS=m |
122 |
+ CONFIG_VFAT_FS=m |
123 |
+ CONFIG_PROC_KCORE=y |
124 |
++CONFIG_HUGETLBFS=y |
125 |
+ CONFIG_TMPFS=y |
126 |
+ CONFIG_TMPFS_POSIX_ACL=y |
127 |
+ # CONFIG_MISC_FILESYSTEMS is not set |
128 |
+diff --git a/arch/powerpc/kernel/vdso32/gettimeofday.S b/arch/powerpc/kernel/vdso32/gettimeofday.S |
129 |
+index 769c2624e0a6..75cff3f336b3 100644 |
130 |
+--- a/arch/powerpc/kernel/vdso32/gettimeofday.S |
131 |
++++ b/arch/powerpc/kernel/vdso32/gettimeofday.S |
132 |
+@@ -98,7 +98,7 @@ V_FUNCTION_BEGIN(__kernel_clock_gettime) |
133 |
+ * can be used, r7 contains NSEC_PER_SEC. |
134 |
+ */ |
135 |
+ |
136 |
+- lwz r5,WTOM_CLOCK_SEC(r9) |
137 |
++ lwz r5,(WTOM_CLOCK_SEC+LOPART)(r9) |
138 |
+ lwz r6,WTOM_CLOCK_NSEC(r9) |
139 |
+ |
140 |
+ /* We now have our offset in r5,r6. We create a fake dependency |
141 |
+diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype |
142 |
+index 6c6a7c72cae4..ad0216c41d2c 100644 |
143 |
+--- a/arch/powerpc/platforms/Kconfig.cputype |
144 |
++++ b/arch/powerpc/platforms/Kconfig.cputype |
145 |
+@@ -330,7 +330,7 @@ config ARCH_ENABLE_SPLIT_PMD_PTLOCK |
146 |
+ |
147 |
+ config PPC_RADIX_MMU |
148 |
+ bool "Radix MMU Support" |
149 |
+- depends on PPC_BOOK3S_64 |
150 |
++ depends on PPC_BOOK3S_64 && HUGETLB_PAGE |
151 |
+ select ARCH_HAS_GIGANTIC_PAGE if (MEMORY_ISOLATION && COMPACTION) || CMA |
152 |
+ default y |
153 |
+ help |
154 |
+diff --git a/arch/x86/Makefile b/arch/x86/Makefile |
155 |
+index b84f61bc5e7a..ffc823a8312f 100644 |
156 |
+--- a/arch/x86/Makefile |
157 |
++++ b/arch/x86/Makefile |
158 |
+@@ -224,6 +224,15 @@ KBUILD_CFLAGS += -fno-asynchronous-unwind-tables |
159 |
+ # Avoid indirect branches in kernel to deal with Spectre |
160 |
+ ifdef CONFIG_RETPOLINE |
161 |
+ KBUILD_CFLAGS += $(RETPOLINE_CFLAGS) |
162 |
++ # Additionally, avoid generating expensive indirect jumps which |
163 |
++ # are subject to retpolines for small number of switch cases. |
164 |
++ # clang turns off jump table generation by default when under |
165 |
++ # retpoline builds, however, gcc does not for x86. This has |
166 |
++ # only been fixed starting from gcc stable version 8.4.0 and |
167 |
++ # onwards, but not for older ones. See gcc bug #86952. |
168 |
++ ifndef CONFIG_CC_IS_CLANG |
169 |
++ KBUILD_CFLAGS += $(call cc-option,-fno-jump-tables) |
170 |
++ endif |
171 |
+ endif |
172 |
+ |
173 |
+ archscripts: scripts_basic |
174 |
+diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c |
175 |
+index 9f8084f18d58..6eb76106c469 100644 |
176 |
+--- a/arch/x86/events/intel/cstate.c |
177 |
++++ b/arch/x86/events/intel/cstate.c |
178 |
+@@ -76,15 +76,15 @@ |
179 |
+ * Scope: Package (physical package) |
180 |
+ * MSR_PKG_C8_RESIDENCY: Package C8 Residency Counter. |
181 |
+ * perf code: 0x04 |
182 |
+- * Available model: HSW ULT,CNL |
183 |
++ * Available model: HSW ULT,KBL,CNL |
184 |
+ * Scope: Package (physical package) |
185 |
+ * MSR_PKG_C9_RESIDENCY: Package C9 Residency Counter. |
186 |
+ * perf code: 0x05 |
187 |
+- * Available model: HSW ULT,CNL |
188 |
++ * Available model: HSW ULT,KBL,CNL |
189 |
+ * Scope: Package (physical package) |
190 |
+ * MSR_PKG_C10_RESIDENCY: Package C10 Residency Counter. |
191 |
+ * perf code: 0x06 |
192 |
+- * Available model: HSW ULT,GLM,CNL |
193 |
++ * Available model: HSW ULT,KBL,GLM,CNL |
194 |
+ * Scope: Package (physical package) |
195 |
+ * |
196 |
+ */ |
197 |
+@@ -572,8 +572,8 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = { |
198 |
+ X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_DESKTOP, snb_cstates), |
199 |
+ X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_X, snb_cstates), |
200 |
+ |
201 |
+- X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE, snb_cstates), |
202 |
+- X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, snb_cstates), |
203 |
++ X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE, hswult_cstates), |
204 |
++ X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, hswult_cstates), |
205 |
+ |
206 |
+ X86_CSTATES_MODEL(INTEL_FAM6_CANNONLAKE_MOBILE, cnl_cstates), |
207 |
+ |
208 |
+diff --git a/arch/x86/include/asm/efi.h b/arch/x86/include/asm/efi.h |
209 |
+index cec5fae23eb3..baa549f8e918 100644 |
210 |
+--- a/arch/x86/include/asm/efi.h |
211 |
++++ b/arch/x86/include/asm/efi.h |
212 |
+@@ -82,8 +82,7 @@ struct efi_scratch { |
213 |
+ #define arch_efi_call_virt_setup() \ |
214 |
+ ({ \ |
215 |
+ efi_sync_low_kernel_mappings(); \ |
216 |
+- preempt_disable(); \ |
217 |
+- __kernel_fpu_begin(); \ |
218 |
++ kernel_fpu_begin(); \ |
219 |
+ firmware_restrict_branch_speculation_start(); \ |
220 |
+ \ |
221 |
+ if (!efi_enabled(EFI_OLD_MEMMAP)) \ |
222 |
+@@ -99,8 +98,7 @@ struct efi_scratch { |
223 |
+ efi_switch_mm(efi_scratch.prev_mm); \ |
224 |
+ \ |
225 |
+ firmware_restrict_branch_speculation_end(); \ |
226 |
+- __kernel_fpu_end(); \ |
227 |
+- preempt_enable(); \ |
228 |
++ kernel_fpu_end(); \ |
229 |
+ }) |
230 |
+ |
231 |
+ extern void __iomem *__init efi_ioremap(unsigned long addr, unsigned long size, |
232 |
+diff --git a/arch/x86/include/asm/fpu/api.h b/arch/x86/include/asm/fpu/api.h |
233 |
+index a9caac9d4a72..b56d504af654 100644 |
234 |
+--- a/arch/x86/include/asm/fpu/api.h |
235 |
++++ b/arch/x86/include/asm/fpu/api.h |
236 |
+@@ -12,17 +12,12 @@ |
237 |
+ #define _ASM_X86_FPU_API_H |
238 |
+ |
239 |
+ /* |
240 |
+- * Careful: __kernel_fpu_begin/end() must be called with preempt disabled |
241 |
+- * and they don't touch the preempt state on their own. |
242 |
+- * If you enable preemption after __kernel_fpu_begin(), preempt notifier |
243 |
+- * should call the __kernel_fpu_end() to prevent the kernel/user FPU |
244 |
+- * state from getting corrupted. KVM for example uses this model. |
245 |
+- * |
246 |
+- * All other cases use kernel_fpu_begin/end() which disable preemption |
247 |
+- * during kernel FPU usage. |
248 |
++ * Use kernel_fpu_begin/end() if you intend to use FPU in kernel context. It |
249 |
++ * disables preemption so be careful if you intend to use it for long periods |
250 |
++ * of time. |
251 |
++ * If you intend to use the FPU in softirq you need to check first with |
252 |
++ * irq_fpu_usable() if it is possible. |
253 |
+ */ |
254 |
+-extern void __kernel_fpu_begin(void); |
255 |
+-extern void __kernel_fpu_end(void); |
256 |
+ extern void kernel_fpu_begin(void); |
257 |
+ extern void kernel_fpu_end(void); |
258 |
+ extern bool irq_fpu_usable(void); |
259 |
+diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c |
260 |
+index 2ea85b32421a..2e5003fef51a 100644 |
261 |
+--- a/arch/x86/kernel/fpu/core.c |
262 |
++++ b/arch/x86/kernel/fpu/core.c |
263 |
+@@ -93,7 +93,7 @@ bool irq_fpu_usable(void) |
264 |
+ } |
265 |
+ EXPORT_SYMBOL(irq_fpu_usable); |
266 |
+ |
267 |
+-void __kernel_fpu_begin(void) |
268 |
++static void __kernel_fpu_begin(void) |
269 |
+ { |
270 |
+ struct fpu *fpu = ¤t->thread.fpu; |
271 |
+ |
272 |
+@@ -111,9 +111,8 @@ void __kernel_fpu_begin(void) |
273 |
+ __cpu_invalidate_fpregs_state(); |
274 |
+ } |
275 |
+ } |
276 |
+-EXPORT_SYMBOL(__kernel_fpu_begin); |
277 |
+ |
278 |
+-void __kernel_fpu_end(void) |
279 |
++static void __kernel_fpu_end(void) |
280 |
+ { |
281 |
+ struct fpu *fpu = ¤t->thread.fpu; |
282 |
+ |
283 |
+@@ -122,7 +121,6 @@ void __kernel_fpu_end(void) |
284 |
+ |
285 |
+ kernel_fpu_enable(); |
286 |
+ } |
287 |
+-EXPORT_SYMBOL(__kernel_fpu_end); |
288 |
+ |
289 |
+ void kernel_fpu_begin(void) |
290 |
+ { |
291 |
+diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c |
292 |
+index 030c98f35cca..a654ccfd1a22 100644 |
293 |
+--- a/drivers/android/binder_alloc.c |
294 |
++++ b/drivers/android/binder_alloc.c |
295 |
+@@ -958,14 +958,13 @@ enum lru_status binder_alloc_free_page(struct list_head *item, |
296 |
+ |
297 |
+ index = page - alloc->pages; |
298 |
+ page_addr = (uintptr_t)alloc->buffer + index * PAGE_SIZE; |
299 |
++ |
300 |
++ mm = alloc->vma_vm_mm; |
301 |
++ if (!mmget_not_zero(mm)) |
302 |
++ goto err_mmget; |
303 |
++ if (!down_write_trylock(&mm->mmap_sem)) |
304 |
++ goto err_down_write_mmap_sem_failed; |
305 |
+ vma = binder_alloc_get_vma(alloc); |
306 |
+- if (vma) { |
307 |
+- if (!mmget_not_zero(alloc->vma_vm_mm)) |
308 |
+- goto err_mmget; |
309 |
+- mm = alloc->vma_vm_mm; |
310 |
+- if (!down_write_trylock(&mm->mmap_sem)) |
311 |
+- goto err_down_write_mmap_sem_failed; |
312 |
+- } |
313 |
+ |
314 |
+ list_lru_isolate(lru, item); |
315 |
+ spin_unlock(lock); |
316 |
+@@ -978,10 +977,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item, |
317 |
+ PAGE_SIZE); |
318 |
+ |
319 |
+ trace_binder_unmap_user_end(alloc, index); |
320 |
+- |
321 |
+- up_write(&mm->mmap_sem); |
322 |
+- mmput(mm); |
323 |
+ } |
324 |
++ up_write(&mm->mmap_sem); |
325 |
++ mmput(mm); |
326 |
+ |
327 |
+ trace_binder_unmap_kernel_start(alloc, index); |
328 |
+ |
329 |
+diff --git a/drivers/block/loop.c b/drivers/block/loop.c |
330 |
+index a63da9e07341..f1e63eb7cbca 100644 |
331 |
+--- a/drivers/block/loop.c |
332 |
++++ b/drivers/block/loop.c |
333 |
+@@ -1112,8 +1112,9 @@ out_unlock: |
334 |
+ err = __blkdev_reread_part(bdev); |
335 |
+ else |
336 |
+ err = blkdev_reread_part(bdev); |
337 |
+- pr_warn("%s: partition scan of loop%d failed (rc=%d)\n", |
338 |
+- __func__, lo_number, err); |
339 |
++ if (err) |
340 |
++ pr_warn("%s: partition scan of loop%d failed (rc=%d)\n", |
341 |
++ __func__, lo_number, err); |
342 |
+ /* Device is gone, no point in returning error */ |
343 |
+ err = 0; |
344 |
+ } |
345 |
+diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c |
346 |
+index a65505db09e5..70cbd0ee1b07 100644 |
347 |
+--- a/drivers/block/zram/zram_drv.c |
348 |
++++ b/drivers/block/zram/zram_drv.c |
349 |
+@@ -515,18 +515,18 @@ struct zram_work { |
350 |
+ struct zram *zram; |
351 |
+ unsigned long entry; |
352 |
+ struct bio *bio; |
353 |
++ struct bio_vec bvec; |
354 |
+ }; |
355 |
+ |
356 |
+ #if PAGE_SIZE != 4096 |
357 |
+ static void zram_sync_read(struct work_struct *work) |
358 |
+ { |
359 |
+- struct bio_vec bvec; |
360 |
+ struct zram_work *zw = container_of(work, struct zram_work, work); |
361 |
+ struct zram *zram = zw->zram; |
362 |
+ unsigned long entry = zw->entry; |
363 |
+ struct bio *bio = zw->bio; |
364 |
+ |
365 |
+- read_from_bdev_async(zram, &bvec, entry, bio); |
366 |
++ read_from_bdev_async(zram, &zw->bvec, entry, bio); |
367 |
+ } |
368 |
+ |
369 |
+ /* |
370 |
+@@ -539,6 +539,7 @@ static int read_from_bdev_sync(struct zram *zram, struct bio_vec *bvec, |
371 |
+ { |
372 |
+ struct zram_work work; |
373 |
+ |
374 |
++ work.bvec = *bvec; |
375 |
+ work.zram = zram; |
376 |
+ work.entry = entry; |
377 |
+ work.bio = bio; |
378 |
+diff --git a/drivers/dma/sh/rcar-dmac.c b/drivers/dma/sh/rcar-dmac.c |
379 |
+index 48ee35e2bce6..0b05a1e08d21 100644 |
380 |
+--- a/drivers/dma/sh/rcar-dmac.c |
381 |
++++ b/drivers/dma/sh/rcar-dmac.c |
382 |
+@@ -1281,6 +1281,9 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan, |
383 |
+ enum dma_status status; |
384 |
+ unsigned int residue = 0; |
385 |
+ unsigned int dptr = 0; |
386 |
++ unsigned int chcrb; |
387 |
++ unsigned int tcrb; |
388 |
++ unsigned int i; |
389 |
+ |
390 |
+ if (!desc) |
391 |
+ return 0; |
392 |
+@@ -1328,6 +1331,24 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan, |
393 |
+ return 0; |
394 |
+ } |
395 |
+ |
396 |
++ /* |
397 |
++ * We need to read two registers. |
398 |
++ * Make sure the control register does not skip to next chunk |
399 |
++ * while reading the counter. |
400 |
++ * Trying it 3 times should be enough: Initial read, retry, retry |
401 |
++ * for the paranoid. |
402 |
++ */ |
403 |
++ for (i = 0; i < 3; i++) { |
404 |
++ chcrb = rcar_dmac_chan_read(chan, RCAR_DMACHCRB) & |
405 |
++ RCAR_DMACHCRB_DPTR_MASK; |
406 |
++ tcrb = rcar_dmac_chan_read(chan, RCAR_DMATCRB); |
407 |
++ /* Still the same? */ |
408 |
++ if (chcrb == (rcar_dmac_chan_read(chan, RCAR_DMACHCRB) & |
409 |
++ RCAR_DMACHCRB_DPTR_MASK)) |
410 |
++ break; |
411 |
++ } |
412 |
++ WARN_ONCE(i >= 3, "residue might be not continuous!"); |
413 |
++ |
414 |
+ /* |
415 |
+ * In descriptor mode the descriptor running pointer is not maintained |
416 |
+ * by the interrupt handler, find the running descriptor from the |
417 |
+@@ -1335,8 +1356,7 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan, |
418 |
+ * mode just use the running descriptor pointer. |
419 |
+ */ |
420 |
+ if (desc->hwdescs.use) { |
421 |
+- dptr = (rcar_dmac_chan_read(chan, RCAR_DMACHCRB) & |
422 |
+- RCAR_DMACHCRB_DPTR_MASK) >> RCAR_DMACHCRB_DPTR_SHIFT; |
423 |
++ dptr = chcrb >> RCAR_DMACHCRB_DPTR_SHIFT; |
424 |
+ if (dptr == 0) |
425 |
+ dptr = desc->nchunks; |
426 |
+ dptr--; |
427 |
+@@ -1354,7 +1374,7 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan, |
428 |
+ } |
429 |
+ |
430 |
+ /* Add the residue for the current chunk. */ |
431 |
+- residue += rcar_dmac_chan_read(chan, RCAR_DMATCRB) << desc->xfer_shift; |
432 |
++ residue += tcrb << desc->xfer_shift; |
433 |
+ |
434 |
+ return residue; |
435 |
+ } |
436 |
+@@ -1367,6 +1387,7 @@ static enum dma_status rcar_dmac_tx_status(struct dma_chan *chan, |
437 |
+ enum dma_status status; |
438 |
+ unsigned long flags; |
439 |
+ unsigned int residue; |
440 |
++ bool cyclic; |
441 |
+ |
442 |
+ status = dma_cookie_status(chan, cookie, txstate); |
443 |
+ if (status == DMA_COMPLETE || !txstate) |
444 |
+@@ -1374,10 +1395,11 @@ static enum dma_status rcar_dmac_tx_status(struct dma_chan *chan, |
445 |
+ |
446 |
+ spin_lock_irqsave(&rchan->lock, flags); |
447 |
+ residue = rcar_dmac_chan_get_residue(rchan, cookie); |
448 |
++ cyclic = rchan->desc.running ? rchan->desc.running->cyclic : false; |
449 |
+ spin_unlock_irqrestore(&rchan->lock, flags); |
450 |
+ |
451 |
+ /* if there's no residue, the cookie is complete */ |
452 |
+- if (!residue) |
453 |
++ if (!residue && !cyclic) |
454 |
+ return DMA_COMPLETE; |
455 |
+ |
456 |
+ dma_set_residue(txstate, residue); |
457 |
+diff --git a/drivers/gpio/gpio-eic-sprd.c b/drivers/gpio/gpio-eic-sprd.c |
458 |
+index e41223c05f6e..6cf2e2ce4093 100644 |
459 |
+--- a/drivers/gpio/gpio-eic-sprd.c |
460 |
++++ b/drivers/gpio/gpio-eic-sprd.c |
461 |
+@@ -414,6 +414,7 @@ static int sprd_eic_irq_set_type(struct irq_data *data, unsigned int flow_type) |
462 |
+ irq_set_handler_locked(data, handle_edge_irq); |
463 |
+ break; |
464 |
+ case IRQ_TYPE_EDGE_BOTH: |
465 |
++ sprd_eic_update(chip, offset, SPRD_EIC_SYNC_INTMODE, 0); |
466 |
+ sprd_eic_update(chip, offset, SPRD_EIC_SYNC_INTBOTH, 1); |
467 |
+ irq_set_handler_locked(data, handle_edge_irq); |
468 |
+ break; |
469 |
+diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c |
470 |
+index 6f91634880aa..2d6506c08bf7 100644 |
471 |
+--- a/drivers/gpu/drm/i915/intel_fbdev.c |
472 |
++++ b/drivers/gpu/drm/i915/intel_fbdev.c |
473 |
+@@ -334,8 +334,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, |
474 |
+ bool *enabled, int width, int height) |
475 |
+ { |
476 |
+ struct drm_i915_private *dev_priv = to_i915(fb_helper->dev); |
477 |
++ unsigned long conn_configured, conn_seq, mask; |
478 |
+ unsigned int count = min(fb_helper->connector_count, BITS_PER_LONG); |
479 |
+- unsigned long conn_configured, conn_seq; |
480 |
+ int i, j; |
481 |
+ bool *save_enabled; |
482 |
+ bool fallback = true, ret = true; |
483 |
+@@ -353,9 +353,10 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper, |
484 |
+ drm_modeset_backoff(&ctx); |
485 |
+ |
486 |
+ memcpy(save_enabled, enabled, count); |
487 |
+- conn_seq = GENMASK(count - 1, 0); |
488 |
++ mask = GENMASK(count - 1, 0); |
489 |
+ conn_configured = 0; |
490 |
+ retry: |
491 |
++ conn_seq = conn_configured; |
492 |
+ for (i = 0; i < count; i++) { |
493 |
+ struct drm_fb_helper_connector *fb_conn; |
494 |
+ struct drm_connector *connector; |
495 |
+@@ -368,8 +369,7 @@ retry: |
496 |
+ if (conn_configured & BIT(i)) |
497 |
+ continue; |
498 |
+ |
499 |
+- /* First pass, only consider tiled connectors */ |
500 |
+- if (conn_seq == GENMASK(count - 1, 0) && !connector->has_tile) |
501 |
++ if (conn_seq == 0 && !connector->has_tile) |
502 |
+ continue; |
503 |
+ |
504 |
+ if (connector->status == connector_status_connected) |
505 |
+@@ -473,10 +473,8 @@ retry: |
506 |
+ conn_configured |= BIT(i); |
507 |
+ } |
508 |
+ |
509 |
+- if (conn_configured != conn_seq) { /* repeat until no more are found */ |
510 |
+- conn_seq = conn_configured; |
511 |
++ if ((conn_configured & mask) != mask && conn_configured != conn_seq) |
512 |
+ goto retry; |
513 |
+- } |
514 |
+ |
515 |
+ /* |
516 |
+ * If the BIOS didn't enable everything it could, fall back to have the |
517 |
+diff --git a/drivers/gpu/drm/rockchip/cdn-dp-reg.c b/drivers/gpu/drm/rockchip/cdn-dp-reg.c |
518 |
+index 5a485489a1e2..6c8b14fb1d2f 100644 |
519 |
+--- a/drivers/gpu/drm/rockchip/cdn-dp-reg.c |
520 |
++++ b/drivers/gpu/drm/rockchip/cdn-dp-reg.c |
521 |
+@@ -113,7 +113,7 @@ static int cdp_dp_mailbox_write(struct cdn_dp_device *dp, u8 val) |
522 |
+ |
523 |
+ static int cdn_dp_mailbox_validate_receive(struct cdn_dp_device *dp, |
524 |
+ u8 module_id, u8 opcode, |
525 |
+- u8 req_size) |
526 |
++ u16 req_size) |
527 |
+ { |
528 |
+ u32 mbox_size, i; |
529 |
+ u8 header[4]; |
530 |
+diff --git a/drivers/gpu/drm/vc4/vc4_crtc.c b/drivers/gpu/drm/vc4/vc4_crtc.c |
531 |
+index 0e6a121858d1..5615ceb15708 100644 |
532 |
+--- a/drivers/gpu/drm/vc4/vc4_crtc.c |
533 |
++++ b/drivers/gpu/drm/vc4/vc4_crtc.c |
534 |
+@@ -998,7 +998,7 @@ static void |
535 |
+ vc4_crtc_reset(struct drm_crtc *crtc) |
536 |
+ { |
537 |
+ if (crtc->state) |
538 |
+- __drm_atomic_helper_crtc_destroy_state(crtc->state); |
539 |
++ vc4_crtc_destroy_state(crtc, crtc->state); |
540 |
+ |
541 |
+ crtc->state = kzalloc(sizeof(struct vc4_crtc_state), GFP_KERNEL); |
542 |
+ if (crtc->state) |
543 |
+diff --git a/drivers/hwtracing/intel_th/gth.c b/drivers/hwtracing/intel_th/gth.c |
544 |
+index cc287cf6eb29..edc52d75e6bd 100644 |
545 |
+--- a/drivers/hwtracing/intel_th/gth.c |
546 |
++++ b/drivers/hwtracing/intel_th/gth.c |
547 |
+@@ -616,7 +616,7 @@ static void intel_th_gth_unassign(struct intel_th_device *thdev, |
548 |
+ othdev->output.port = -1; |
549 |
+ othdev->output.active = false; |
550 |
+ gth->output[port].output = NULL; |
551 |
+- for (master = 0; master < TH_CONFIGURABLE_MASTERS; master++) |
552 |
++ for (master = 0; master <= TH_CONFIGURABLE_MASTERS; master++) |
553 |
+ if (gth->master[master] == port) |
554 |
+ gth->master[master] = -1; |
555 |
+ spin_unlock(>h->gth_lock); |
556 |
+diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c |
557 |
+index 50be240df331..8cc4da62f050 100644 |
558 |
+--- a/drivers/infiniband/hw/mlx5/main.c |
559 |
++++ b/drivers/infiniband/hw/mlx5/main.c |
560 |
+@@ -2014,6 +2014,7 @@ static int mlx5_ib_mmap_clock_info_page(struct mlx5_ib_dev *dev, |
561 |
+ |
562 |
+ if (vma->vm_flags & VM_WRITE) |
563 |
+ return -EPERM; |
564 |
++ vma->vm_flags &= ~VM_MAYWRITE; |
565 |
+ |
566 |
+ if (!dev->mdev->clock_info_page) |
567 |
+ return -EOPNOTSUPP; |
568 |
+@@ -2197,6 +2198,7 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm |
569 |
+ |
570 |
+ if (vma->vm_flags & VM_WRITE) |
571 |
+ return -EPERM; |
572 |
++ vma->vm_flags &= ~VM_MAYWRITE; |
573 |
+ |
574 |
+ /* Don't expose to user-space information it shouldn't have */ |
575 |
+ if (PAGE_SIZE > 4096) |
576 |
+diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c |
577 |
+index 49c9541050d4..5819c9d6ffdc 100644 |
578 |
+--- a/drivers/infiniband/sw/rdmavt/mr.c |
579 |
++++ b/drivers/infiniband/sw/rdmavt/mr.c |
580 |
+@@ -611,11 +611,6 @@ static int rvt_set_page(struct ib_mr *ibmr, u64 addr) |
581 |
+ if (unlikely(mapped_segs == mr->mr.max_segs)) |
582 |
+ return -ENOMEM; |
583 |
+ |
584 |
+- if (mr->mr.length == 0) { |
585 |
+- mr->mr.user_base = addr; |
586 |
+- mr->mr.iova = addr; |
587 |
+- } |
588 |
+- |
589 |
+ m = mapped_segs / RVT_SEGSZ; |
590 |
+ n = mapped_segs % RVT_SEGSZ; |
591 |
+ mr->mr.map[m]->segs[n].vaddr = (void *)addr; |
592 |
+@@ -633,17 +628,24 @@ static int rvt_set_page(struct ib_mr *ibmr, u64 addr) |
593 |
+ * @sg_nents: number of entries in sg |
594 |
+ * @sg_offset: offset in bytes into sg |
595 |
+ * |
596 |
++ * Overwrite rvt_mr length with mr length calculated by ib_sg_to_pages. |
597 |
++ * |
598 |
+ * Return: number of sg elements mapped to the memory region |
599 |
+ */ |
600 |
+ int rvt_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, |
601 |
+ int sg_nents, unsigned int *sg_offset) |
602 |
+ { |
603 |
+ struct rvt_mr *mr = to_imr(ibmr); |
604 |
++ int ret; |
605 |
+ |
606 |
+ mr->mr.length = 0; |
607 |
+ mr->mr.page_shift = PAGE_SHIFT; |
608 |
+- return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, |
609 |
+- rvt_set_page); |
610 |
++ ret = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, rvt_set_page); |
611 |
++ mr->mr.user_base = ibmr->iova; |
612 |
++ mr->mr.iova = ibmr->iova; |
613 |
++ mr->mr.offset = ibmr->iova - (u64)mr->mr.map[0]->segs[0].vaddr; |
614 |
++ mr->mr.length = (size_t)ibmr->length; |
615 |
++ return ret; |
616 |
+ } |
617 |
+ |
618 |
+ /** |
619 |
+@@ -674,6 +676,7 @@ int rvt_fast_reg_mr(struct rvt_qp *qp, struct ib_mr *ibmr, u32 key, |
620 |
+ ibmr->rkey = key; |
621 |
+ mr->mr.lkey = key; |
622 |
+ mr->mr.access_flags = access; |
623 |
++ mr->mr.iova = ibmr->iova; |
624 |
+ atomic_set(&mr->mr.lkey_invalid, 0); |
625 |
+ |
626 |
+ return 0; |
627 |
+diff --git a/drivers/input/rmi4/rmi_f11.c b/drivers/input/rmi4/rmi_f11.c |
628 |
+index df64d6aed4f7..93901ebd122a 100644 |
629 |
+--- a/drivers/input/rmi4/rmi_f11.c |
630 |
++++ b/drivers/input/rmi4/rmi_f11.c |
631 |
+@@ -1230,7 +1230,7 @@ static int rmi_f11_initialize(struct rmi_function *fn) |
632 |
+ } |
633 |
+ |
634 |
+ rc = f11_write_control_regs(fn, &f11->sens_query, |
635 |
+- &f11->dev_controls, fn->fd.query_base_addr); |
636 |
++ &f11->dev_controls, fn->fd.control_base_addr); |
637 |
+ if (rc) |
638 |
+ dev_warn(&fn->dev, "Failed to write control registers\n"); |
639 |
+ |
640 |
+diff --git a/drivers/net/dsa/mv88e6xxx/chip.c b/drivers/net/dsa/mv88e6xxx/chip.c |
641 |
+index dabe89968a78..2caa5c0c2bc4 100644 |
642 |
+--- a/drivers/net/dsa/mv88e6xxx/chip.c |
643 |
++++ b/drivers/net/dsa/mv88e6xxx/chip.c |
644 |
+@@ -4821,6 +4821,7 @@ static int mv88e6xxx_probe(struct mdio_device *mdiodev) |
645 |
+ if (err) |
646 |
+ goto out; |
647 |
+ |
648 |
++ mv88e6xxx_ports_cmode_init(chip); |
649 |
+ mv88e6xxx_phy_init(chip); |
650 |
+ |
651 |
+ if (chip->info->ops->get_eeprom) { |
652 |
+diff --git a/drivers/net/ethernet/hisilicon/hns/hns_enet.c b/drivers/net/ethernet/hisilicon/hns/hns_enet.c |
653 |
+index b043370c2685..cc84133c184d 100644 |
654 |
+--- a/drivers/net/ethernet/hisilicon/hns/hns_enet.c |
655 |
++++ b/drivers/net/ethernet/hisilicon/hns/hns_enet.c |
656 |
+@@ -1169,6 +1169,12 @@ int hns_nic_init_phy(struct net_device *ndev, struct hnae_handle *h) |
657 |
+ if (!h->phy_dev) |
658 |
+ return 0; |
659 |
+ |
660 |
++ phy_dev->supported &= h->if_support; |
661 |
++ phy_dev->advertising = phy_dev->supported; |
662 |
++ |
663 |
++ if (h->phy_if == PHY_INTERFACE_MODE_XGMII) |
664 |
++ phy_dev->autoneg = false; |
665 |
++ |
666 |
+ if (h->phy_if != PHY_INTERFACE_MODE_XGMII) { |
667 |
+ phy_dev->dev_flags = 0; |
668 |
+ |
669 |
+@@ -1180,15 +1186,6 @@ int hns_nic_init_phy(struct net_device *ndev, struct hnae_handle *h) |
670 |
+ if (unlikely(ret)) |
671 |
+ return -ENODEV; |
672 |
+ |
673 |
+- phy_dev->supported &= h->if_support; |
674 |
+- phy_dev->advertising = phy_dev->supported; |
675 |
+- |
676 |
+- if (h->phy_if == PHY_INTERFACE_MODE_XGMII) |
677 |
+- phy_dev->autoneg = false; |
678 |
+- |
679 |
+- if (h->phy_if == PHY_INTERFACE_MODE_SGMII) |
680 |
+- phy_stop(phy_dev); |
681 |
+- |
682 |
+ return 0; |
683 |
+ } |
684 |
+ |
685 |
+diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c |
686 |
+index a475f36ddf8c..426789e2c23d 100644 |
687 |
+--- a/drivers/net/ethernet/ibm/ibmvnic.c |
688 |
++++ b/drivers/net/ethernet/ibm/ibmvnic.c |
689 |
+@@ -1859,7 +1859,7 @@ static int do_reset(struct ibmvnic_adapter *adapter, |
690 |
+ |
691 |
+ if (adapter->reset_reason != VNIC_RESET_FAILOVER && |
692 |
+ adapter->reset_reason != VNIC_RESET_CHANGE_PARAM) |
693 |
+- netdev_notify_peers(netdev); |
694 |
++ call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev); |
695 |
+ |
696 |
+ netif_carrier_on(netdev); |
697 |
+ |
698 |
+diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_main.c b/drivers/net/ethernet/intel/fm10k/fm10k_main.c |
699 |
+index 3f536541f45f..78a43d688cb1 100644 |
700 |
+--- a/drivers/net/ethernet/intel/fm10k/fm10k_main.c |
701 |
++++ b/drivers/net/ethernet/intel/fm10k/fm10k_main.c |
702 |
+@@ -41,6 +41,8 @@ static int __init fm10k_init_module(void) |
703 |
+ /* create driver workqueue */ |
704 |
+ fm10k_workqueue = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, |
705 |
+ fm10k_driver_name); |
706 |
++ if (!fm10k_workqueue) |
707 |
++ return -ENOMEM; |
708 |
+ |
709 |
+ fm10k_dbg_init(); |
710 |
+ |
711 |
+diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c |
712 |
+index 9988c89ed9fd..9b10abb604cb 100644 |
713 |
+--- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c |
714 |
++++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c |
715 |
+@@ -4272,7 +4272,7 @@ static void mvpp2_phylink_validate(struct net_device *dev, |
716 |
+ case PHY_INTERFACE_MODE_RGMII_ID: |
717 |
+ case PHY_INTERFACE_MODE_RGMII_RXID: |
718 |
+ case PHY_INTERFACE_MODE_RGMII_TXID: |
719 |
+- if (port->gop_id == 0) |
720 |
++ if (port->priv->hw_version == MVPP22 && port->gop_id == 0) |
721 |
+ goto empty_set; |
722 |
+ break; |
723 |
+ default: |
724 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c |
725 |
+index 4a33c9a7cac7..599114ab7821 100644 |
726 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c |
727 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c |
728 |
+@@ -33,6 +33,26 @@ |
729 |
+ #include <linux/bpf_trace.h> |
730 |
+ #include "en/xdp.h" |
731 |
+ |
732 |
++int mlx5e_xdp_max_mtu(struct mlx5e_params *params) |
733 |
++{ |
734 |
++ int hr = NET_IP_ALIGN + XDP_PACKET_HEADROOM; |
735 |
++ |
736 |
++ /* Let S := SKB_DATA_ALIGN(sizeof(struct skb_shared_info)). |
737 |
++ * The condition checked in mlx5e_rx_is_linear_skb is: |
738 |
++ * SKB_DATA_ALIGN(sw_mtu + hard_mtu + hr) + S <= PAGE_SIZE (1) |
739 |
++ * (Note that hw_mtu == sw_mtu + hard_mtu.) |
740 |
++ * What is returned from this function is: |
741 |
++ * max_mtu = PAGE_SIZE - S - hr - hard_mtu (2) |
742 |
++ * After assigning sw_mtu := max_mtu, the left side of (1) turns to |
743 |
++ * SKB_DATA_ALIGN(PAGE_SIZE - S) + S, which is equal to PAGE_SIZE, |
744 |
++ * because both PAGE_SIZE and S are already aligned. Any number greater |
745 |
++ * than max_mtu would make the left side of (1) greater than PAGE_SIZE, |
746 |
++ * so max_mtu is the maximum MTU allowed. |
747 |
++ */ |
748 |
++ |
749 |
++ return MLX5E_HW2SW_MTU(params, SKB_MAX_HEAD(hr)); |
750 |
++} |
751 |
++ |
752 |
+ static inline bool |
753 |
+ mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_dma_info *di, |
754 |
+ struct xdp_buff *xdp) |
755 |
+@@ -207,9 +227,9 @@ bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq) |
756 |
+ sqcc++; |
757 |
+ |
758 |
+ if (is_redirect) { |
759 |
+- xdp_return_frame(xdpi->xdpf); |
760 |
+ dma_unmap_single(sq->pdev, xdpi->dma_addr, |
761 |
+ xdpi->xdpf->len, DMA_TO_DEVICE); |
762 |
++ xdp_return_frame(xdpi->xdpf); |
763 |
+ } else { |
764 |
+ /* Recycle RX page */ |
765 |
+ mlx5e_page_release(rq, &xdpi->di, true); |
766 |
+@@ -243,9 +263,9 @@ void mlx5e_free_xdpsq_descs(struct mlx5e_xdpsq *sq) |
767 |
+ sq->cc++; |
768 |
+ |
769 |
+ if (is_redirect) { |
770 |
+- xdp_return_frame(xdpi->xdpf); |
771 |
+ dma_unmap_single(sq->pdev, xdpi->dma_addr, |
772 |
+ xdpi->xdpf->len, DMA_TO_DEVICE); |
773 |
++ xdp_return_frame(xdpi->xdpf); |
774 |
+ } else { |
775 |
+ /* Recycle RX page */ |
776 |
+ mlx5e_page_release(rq, &xdpi->di, false); |
777 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h |
778 |
+index 4d096623178b..827ceef5fa93 100644 |
779 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h |
780 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h |
781 |
+@@ -34,12 +34,11 @@ |
782 |
+ |
783 |
+ #include "en.h" |
784 |
+ |
785 |
+-#define MLX5E_XDP_MAX_MTU ((int)(PAGE_SIZE - \ |
786 |
+- MLX5_SKB_FRAG_SZ(XDP_PACKET_HEADROOM))) |
787 |
+ #define MLX5E_XDP_MIN_INLINE (ETH_HLEN + VLAN_HLEN) |
788 |
+ #define MLX5E_XDP_TX_DS_COUNT \ |
789 |
+ ((sizeof(struct mlx5e_tx_wqe) / MLX5_SEND_WQE_DS) + 1 /* SG DS */) |
790 |
+ |
791 |
++int mlx5e_xdp_max_mtu(struct mlx5e_params *params); |
792 |
+ bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di, |
793 |
+ void *va, u16 *rx_headroom, u32 *len); |
794 |
+ bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq); |
795 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c |
796 |
+index 5e5423076b03..9ca4f88d7cf6 100644 |
797 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c |
798 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c |
799 |
+@@ -1317,7 +1317,7 @@ static int mlx5e_get_module_info(struct net_device *netdev, |
800 |
+ break; |
801 |
+ case MLX5_MODULE_ID_SFP: |
802 |
+ modinfo->type = ETH_MODULE_SFF_8472; |
803 |
+- modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN; |
804 |
++ modinfo->eeprom_len = MLX5_EEPROM_PAGE_LENGTH; |
805 |
+ break; |
806 |
+ default: |
807 |
+ netdev_err(priv->netdev, "%s: cable type not recognized:0x%x\n", |
808 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
809 |
+index 637d59c01fe5..b190c447aeb0 100644 |
810 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
811 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c |
812 |
+@@ -3761,7 +3761,7 @@ int mlx5e_change_mtu(struct net_device *netdev, int new_mtu, |
813 |
+ if (params->xdp_prog && |
814 |
+ !mlx5e_rx_is_linear_skb(priv->mdev, &new_channels.params)) { |
815 |
+ netdev_err(netdev, "MTU(%d) > %d is not allowed while XDP enabled\n", |
816 |
+- new_mtu, MLX5E_XDP_MAX_MTU); |
817 |
++ new_mtu, mlx5e_xdp_max_mtu(params)); |
818 |
+ err = -EINVAL; |
819 |
+ goto out; |
820 |
+ } |
821 |
+@@ -4227,7 +4227,8 @@ static int mlx5e_xdp_allowed(struct mlx5e_priv *priv, struct bpf_prog *prog) |
822 |
+ |
823 |
+ if (!mlx5e_rx_is_linear_skb(priv->mdev, &new_channels.params)) { |
824 |
+ netdev_warn(netdev, "XDP is not allowed with MTU(%d) > %d\n", |
825 |
+- new_channels.params.sw_mtu, MLX5E_XDP_MAX_MTU); |
826 |
++ new_channels.params.sw_mtu, |
827 |
++ mlx5e_xdp_max_mtu(&new_channels.params)); |
828 |
+ return -EINVAL; |
829 |
+ } |
830 |
+ |
831 |
+diff --git a/drivers/net/ethernet/mellanox/mlx5/core/port.c b/drivers/net/ethernet/mellanox/mlx5/core/port.c |
832 |
+index 31a9cbd85689..09b6b1bfbfa8 100644 |
833 |
+--- a/drivers/net/ethernet/mellanox/mlx5/core/port.c |
834 |
++++ b/drivers/net/ethernet/mellanox/mlx5/core/port.c |
835 |
+@@ -404,10 +404,6 @@ int mlx5_query_module_eeprom(struct mlx5_core_dev *dev, |
836 |
+ size -= offset + size - MLX5_EEPROM_PAGE_LENGTH; |
837 |
+ |
838 |
+ i2c_addr = MLX5_I2C_ADDR_LOW; |
839 |
+- if (offset >= MLX5_EEPROM_PAGE_LENGTH) { |
840 |
+- i2c_addr = MLX5_I2C_ADDR_HIGH; |
841 |
+- offset -= MLX5_EEPROM_PAGE_LENGTH; |
842 |
+- } |
843 |
+ |
844 |
+ MLX5_SET(mcia_reg, in, l, 0); |
845 |
+ MLX5_SET(mcia_reg, in, module, module_num); |
846 |
+diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h |
847 |
+index 72cdaa01d56d..100618531021 100644 |
848 |
+--- a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h |
849 |
++++ b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h |
850 |
+@@ -27,7 +27,7 @@ |
851 |
+ |
852 |
+ #define MLXSW_PCI_SW_RESET 0xF0010 |
853 |
+ #define MLXSW_PCI_SW_RESET_RST_BIT BIT(0) |
854 |
+-#define MLXSW_PCI_SW_RESET_TIMEOUT_MSECS 13000 |
855 |
++#define MLXSW_PCI_SW_RESET_TIMEOUT_MSECS 20000 |
856 |
+ #define MLXSW_PCI_SW_RESET_WAIT_MSECS 100 |
857 |
+ #define MLXSW_PCI_FW_READY 0xA1844 |
858 |
+ #define MLXSW_PCI_FW_READY_MASK 0xFFFF |
859 |
+diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c |
860 |
+index f9bef030ee05..c5b82e283d13 100644 |
861 |
+--- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c |
862 |
++++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c |
863 |
+@@ -2504,11 +2504,11 @@ mlxsw_sp_port_set_link_ksettings(struct net_device *dev, |
864 |
+ if (err) |
865 |
+ return err; |
866 |
+ |
867 |
++ mlxsw_sp_port->link.autoneg = autoneg; |
868 |
++ |
869 |
+ if (!netif_running(dev)) |
870 |
+ return 0; |
871 |
+ |
872 |
+- mlxsw_sp_port->link.autoneg = autoneg; |
873 |
+- |
874 |
+ mlxsw_sp_port_admin_status_set(mlxsw_sp_port, false); |
875 |
+ mlxsw_sp_port_admin_status_set(mlxsw_sp_port, true); |
876 |
+ |
877 |
+@@ -2783,7 +2783,7 @@ static int mlxsw_sp_port_ets_init(struct mlxsw_sp_port *mlxsw_sp_port) |
878 |
+ err = mlxsw_sp_port_ets_set(mlxsw_sp_port, |
879 |
+ MLXSW_REG_QEEC_HIERARCY_TC, |
880 |
+ i + 8, i, |
881 |
+- false, 0); |
882 |
++ true, 100); |
883 |
+ if (err) |
884 |
+ return err; |
885 |
+ } |
886 |
+diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
887 |
+index 886176be818e..62460a5b4ad9 100644 |
888 |
+--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
889 |
++++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c |
890 |
+@@ -2595,8 +2595,6 @@ static int stmmac_open(struct net_device *dev) |
891 |
+ u32 chan; |
892 |
+ int ret; |
893 |
+ |
894 |
+- stmmac_check_ether_addr(priv); |
895 |
+- |
896 |
+ if (priv->hw->pcs != STMMAC_PCS_RGMII && |
897 |
+ priv->hw->pcs != STMMAC_PCS_TBI && |
898 |
+ priv->hw->pcs != STMMAC_PCS_RTBI) { |
899 |
+@@ -4296,6 +4294,8 @@ int stmmac_dvr_probe(struct device *device, |
900 |
+ if (ret) |
901 |
+ goto error_hw_init; |
902 |
+ |
903 |
++ stmmac_check_ether_addr(priv); |
904 |
++ |
905 |
+ /* Configure real RX and TX queues */ |
906 |
+ netif_set_real_num_rx_queues(ndev, priv->plat->rx_queues_to_use); |
907 |
+ netif_set_real_num_tx_queues(ndev, priv->plat->tx_queues_to_use); |
908 |
+diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c |
909 |
+index d819e8eaba12..cc1e887e47b5 100644 |
910 |
+--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c |
911 |
++++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c |
912 |
+@@ -159,6 +159,12 @@ static const struct dmi_system_id quark_pci_dmi[] = { |
913 |
+ }, |
914 |
+ .driver_data = (void *)&galileo_stmmac_dmi_data, |
915 |
+ }, |
916 |
++ /* |
917 |
++ * There are 2 types of SIMATIC IOT2000: IOT20202 and IOT2040. |
918 |
++ * The asset tag "6ES7647-0AA00-0YA2" is only for IOT2020 which |
919 |
++ * has only one pci network device while other asset tags are |
920 |
++ * for IOT2040 which has two. |
921 |
++ */ |
922 |
+ { |
923 |
+ .matches = { |
924 |
+ DMI_EXACT_MATCH(DMI_BOARD_NAME, "SIMATIC IOT2000"), |
925 |
+@@ -170,8 +176,6 @@ static const struct dmi_system_id quark_pci_dmi[] = { |
926 |
+ { |
927 |
+ .matches = { |
928 |
+ DMI_EXACT_MATCH(DMI_BOARD_NAME, "SIMATIC IOT2000"), |
929 |
+- DMI_EXACT_MATCH(DMI_BOARD_ASSET_TAG, |
930 |
+- "6ES7647-0AA00-1YA2"), |
931 |
+ }, |
932 |
+ .driver_data = (void *)&iot2040_stmmac_dmi_data, |
933 |
+ }, |
934 |
+diff --git a/drivers/net/slip/slhc.c b/drivers/net/slip/slhc.c |
935 |
+index f4e93f5fc204..ea90db3c7705 100644 |
936 |
+--- a/drivers/net/slip/slhc.c |
937 |
++++ b/drivers/net/slip/slhc.c |
938 |
+@@ -153,7 +153,7 @@ out_fail: |
939 |
+ void |
940 |
+ slhc_free(struct slcompress *comp) |
941 |
+ { |
942 |
+- if ( comp == NULLSLCOMPR ) |
943 |
++ if ( IS_ERR_OR_NULL(comp) ) |
944 |
+ return; |
945 |
+ |
946 |
+ if ( comp->tstate != NULLSLSTATE ) |
947 |
+diff --git a/drivers/net/team/team.c b/drivers/net/team/team.c |
948 |
+index e23eaf3f6d03..6c6230b44bcd 100644 |
949 |
+--- a/drivers/net/team/team.c |
950 |
++++ b/drivers/net/team/team.c |
951 |
+@@ -1160,6 +1160,13 @@ static int team_port_add(struct team *team, struct net_device *port_dev, |
952 |
+ return -EINVAL; |
953 |
+ } |
954 |
+ |
955 |
++ if (netdev_has_upper_dev(dev, port_dev)) { |
956 |
++ NL_SET_ERR_MSG(extack, "Device is already an upper device of the team interface"); |
957 |
++ netdev_err(dev, "Device %s is already an upper device of the team interface\n", |
958 |
++ portname); |
959 |
++ return -EBUSY; |
960 |
++ } |
961 |
++ |
962 |
+ if (port_dev->features & NETIF_F_VLAN_CHALLENGED && |
963 |
+ vlan_uses_dev(dev)) { |
964 |
+ NL_SET_ERR_MSG(extack, "Device is VLAN challenged and team device has VLAN set up"); |
965 |
+diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c |
966 |
+index a1f225f077cd..ef47c226e1d2 100644 |
967 |
+--- a/drivers/usb/core/driver.c |
968 |
++++ b/drivers/usb/core/driver.c |
969 |
+@@ -1899,14 +1899,11 @@ int usb_runtime_idle(struct device *dev) |
970 |
+ return -EBUSY; |
971 |
+ } |
972 |
+ |
973 |
+-int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable) |
974 |
++static int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable) |
975 |
+ { |
976 |
+ struct usb_hcd *hcd = bus_to_hcd(udev->bus); |
977 |
+ int ret = -EPERM; |
978 |
+ |
979 |
+- if (enable && !udev->usb2_hw_lpm_allowed) |
980 |
+- return 0; |
981 |
+- |
982 |
+ if (hcd->driver->set_usb2_hw_lpm) { |
983 |
+ ret = hcd->driver->set_usb2_hw_lpm(hcd, udev, enable); |
984 |
+ if (!ret) |
985 |
+@@ -1916,6 +1913,24 @@ int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable) |
986 |
+ return ret; |
987 |
+ } |
988 |
+ |
989 |
++int usb_enable_usb2_hardware_lpm(struct usb_device *udev) |
990 |
++{ |
991 |
++ if (!udev->usb2_hw_lpm_capable || |
992 |
++ !udev->usb2_hw_lpm_allowed || |
993 |
++ udev->usb2_hw_lpm_enabled) |
994 |
++ return 0; |
995 |
++ |
996 |
++ return usb_set_usb2_hardware_lpm(udev, 1); |
997 |
++} |
998 |
++ |
999 |
++int usb_disable_usb2_hardware_lpm(struct usb_device *udev) |
1000 |
++{ |
1001 |
++ if (!udev->usb2_hw_lpm_enabled) |
1002 |
++ return 0; |
1003 |
++ |
1004 |
++ return usb_set_usb2_hardware_lpm(udev, 0); |
1005 |
++} |
1006 |
++ |
1007 |
+ #endif /* CONFIG_PM */ |
1008 |
+ |
1009 |
+ struct bus_type usb_bus_type = { |
1010 |
+diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c |
1011 |
+index 3adff4da2ee1..bbcfa63d0233 100644 |
1012 |
+--- a/drivers/usb/core/hub.c |
1013 |
++++ b/drivers/usb/core/hub.c |
1014 |
+@@ -3217,8 +3217,7 @@ int usb_port_suspend(struct usb_device *udev, pm_message_t msg) |
1015 |
+ } |
1016 |
+ |
1017 |
+ /* disable USB2 hardware LPM */ |
1018 |
+- if (udev->usb2_hw_lpm_enabled == 1) |
1019 |
+- usb_set_usb2_hardware_lpm(udev, 0); |
1020 |
++ usb_disable_usb2_hardware_lpm(udev); |
1021 |
+ |
1022 |
+ if (usb_disable_ltm(udev)) { |
1023 |
+ dev_err(&udev->dev, "Failed to disable LTM before suspend\n"); |
1024 |
+@@ -3256,8 +3255,7 @@ int usb_port_suspend(struct usb_device *udev, pm_message_t msg) |
1025 |
+ usb_enable_ltm(udev); |
1026 |
+ err_ltm: |
1027 |
+ /* Try to enable USB2 hardware LPM again */ |
1028 |
+- if (udev->usb2_hw_lpm_capable == 1) |
1029 |
+- usb_set_usb2_hardware_lpm(udev, 1); |
1030 |
++ usb_enable_usb2_hardware_lpm(udev); |
1031 |
+ |
1032 |
+ if (udev->do_remote_wakeup) |
1033 |
+ (void) usb_disable_remote_wakeup(udev); |
1034 |
+@@ -3540,8 +3538,7 @@ int usb_port_resume(struct usb_device *udev, pm_message_t msg) |
1035 |
+ hub_port_logical_disconnect(hub, port1); |
1036 |
+ } else { |
1037 |
+ /* Try to enable USB2 hardware LPM */ |
1038 |
+- if (udev->usb2_hw_lpm_capable == 1) |
1039 |
+- usb_set_usb2_hardware_lpm(udev, 1); |
1040 |
++ usb_enable_usb2_hardware_lpm(udev); |
1041 |
+ |
1042 |
+ /* Try to enable USB3 LTM */ |
1043 |
+ usb_enable_ltm(udev); |
1044 |
+@@ -4432,7 +4429,7 @@ static void hub_set_initial_usb2_lpm_policy(struct usb_device *udev) |
1045 |
+ if ((udev->bos->ext_cap->bmAttributes & cpu_to_le32(USB_BESL_SUPPORT)) || |
1046 |
+ connect_type == USB_PORT_CONNECT_TYPE_HARD_WIRED) { |
1047 |
+ udev->usb2_hw_lpm_allowed = 1; |
1048 |
+- usb_set_usb2_hardware_lpm(udev, 1); |
1049 |
++ usb_enable_usb2_hardware_lpm(udev); |
1050 |
+ } |
1051 |
+ } |
1052 |
+ |
1053 |
+@@ -5608,8 +5605,7 @@ static int usb_reset_and_verify_device(struct usb_device *udev) |
1054 |
+ /* Disable USB2 hardware LPM. |
1055 |
+ * It will be re-enabled by the enumeration process. |
1056 |
+ */ |
1057 |
+- if (udev->usb2_hw_lpm_enabled == 1) |
1058 |
+- usb_set_usb2_hardware_lpm(udev, 0); |
1059 |
++ usb_disable_usb2_hardware_lpm(udev); |
1060 |
+ |
1061 |
+ /* Disable LPM while we reset the device and reinstall the alt settings. |
1062 |
+ * Device-initiated LPM, and system exit latency settings are cleared |
1063 |
+@@ -5712,7 +5708,7 @@ static int usb_reset_and_verify_device(struct usb_device *udev) |
1064 |
+ |
1065 |
+ done: |
1066 |
+ /* Now that the alt settings are re-installed, enable LTM and LPM. */ |
1067 |
+- usb_set_usb2_hardware_lpm(udev, 1); |
1068 |
++ usb_enable_usb2_hardware_lpm(udev); |
1069 |
+ usb_unlocked_enable_lpm(udev); |
1070 |
+ usb_enable_ltm(udev); |
1071 |
+ usb_release_bos_descriptor(udev); |
1072 |
+diff --git a/drivers/usb/core/message.c b/drivers/usb/core/message.c |
1073 |
+index bfa5eda0cc26..4f33eb632a88 100644 |
1074 |
+--- a/drivers/usb/core/message.c |
1075 |
++++ b/drivers/usb/core/message.c |
1076 |
+@@ -1243,8 +1243,7 @@ void usb_disable_device(struct usb_device *dev, int skip_ep0) |
1077 |
+ dev->actconfig->interface[i] = NULL; |
1078 |
+ } |
1079 |
+ |
1080 |
+- if (dev->usb2_hw_lpm_enabled == 1) |
1081 |
+- usb_set_usb2_hardware_lpm(dev, 0); |
1082 |
++ usb_disable_usb2_hardware_lpm(dev); |
1083 |
+ usb_unlocked_disable_lpm(dev); |
1084 |
+ usb_disable_ltm(dev); |
1085 |
+ |
1086 |
+diff --git a/drivers/usb/core/sysfs.c b/drivers/usb/core/sysfs.c |
1087 |
+index ea18284dfa9a..7e88fdfe3cf5 100644 |
1088 |
+--- a/drivers/usb/core/sysfs.c |
1089 |
++++ b/drivers/usb/core/sysfs.c |
1090 |
+@@ -528,7 +528,10 @@ static ssize_t usb2_hardware_lpm_store(struct device *dev, |
1091 |
+ |
1092 |
+ if (!ret) { |
1093 |
+ udev->usb2_hw_lpm_allowed = value; |
1094 |
+- ret = usb_set_usb2_hardware_lpm(udev, value); |
1095 |
++ if (value) |
1096 |
++ ret = usb_enable_usb2_hardware_lpm(udev); |
1097 |
++ else |
1098 |
++ ret = usb_disable_usb2_hardware_lpm(udev); |
1099 |
+ } |
1100 |
+ |
1101 |
+ usb_unlock_device(udev); |
1102 |
+diff --git a/drivers/usb/core/usb.h b/drivers/usb/core/usb.h |
1103 |
+index 546a2219454b..d95a5358f73d 100644 |
1104 |
+--- a/drivers/usb/core/usb.h |
1105 |
++++ b/drivers/usb/core/usb.h |
1106 |
+@@ -92,7 +92,8 @@ extern int usb_remote_wakeup(struct usb_device *dev); |
1107 |
+ extern int usb_runtime_suspend(struct device *dev); |
1108 |
+ extern int usb_runtime_resume(struct device *dev); |
1109 |
+ extern int usb_runtime_idle(struct device *dev); |
1110 |
+-extern int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable); |
1111 |
++extern int usb_enable_usb2_hardware_lpm(struct usb_device *udev); |
1112 |
++extern int usb_disable_usb2_hardware_lpm(struct usb_device *udev); |
1113 |
+ |
1114 |
+ #else |
1115 |
+ |
1116 |
+@@ -112,7 +113,12 @@ static inline int usb_autoresume_device(struct usb_device *udev) |
1117 |
+ return 0; |
1118 |
+ } |
1119 |
+ |
1120 |
+-static inline int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable) |
1121 |
++static inline int usb_enable_usb2_hardware_lpm(struct usb_device *udev) |
1122 |
++{ |
1123 |
++ return 0; |
1124 |
++} |
1125 |
++ |
1126 |
++static inline int usb_disable_usb2_hardware_lpm(struct usb_device *udev) |
1127 |
+ { |
1128 |
+ return 0; |
1129 |
+ } |
1130 |
+diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c |
1131 |
+index 64cbc2d007c9..c36275754086 100644 |
1132 |
+--- a/drivers/vfio/vfio_iommu_type1.c |
1133 |
++++ b/drivers/vfio/vfio_iommu_type1.c |
1134 |
+@@ -58,12 +58,18 @@ module_param_named(disable_hugepages, |
1135 |
+ MODULE_PARM_DESC(disable_hugepages, |
1136 |
+ "Disable VFIO IOMMU support for IOMMU hugepages."); |
1137 |
+ |
1138 |
++static unsigned int dma_entry_limit __read_mostly = U16_MAX; |
1139 |
++module_param_named(dma_entry_limit, dma_entry_limit, uint, 0644); |
1140 |
++MODULE_PARM_DESC(dma_entry_limit, |
1141 |
++ "Maximum number of user DMA mappings per container (65535)."); |
1142 |
++ |
1143 |
+ struct vfio_iommu { |
1144 |
+ struct list_head domain_list; |
1145 |
+ struct vfio_domain *external_domain; /* domain for external user */ |
1146 |
+ struct mutex lock; |
1147 |
+ struct rb_root dma_list; |
1148 |
+ struct blocking_notifier_head notifier; |
1149 |
++ unsigned int dma_avail; |
1150 |
+ bool v2; |
1151 |
+ bool nesting; |
1152 |
+ }; |
1153 |
+@@ -836,6 +842,7 @@ static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *dma) |
1154 |
+ vfio_unlink_dma(iommu, dma); |
1155 |
+ put_task_struct(dma->task); |
1156 |
+ kfree(dma); |
1157 |
++ iommu->dma_avail++; |
1158 |
+ } |
1159 |
+ |
1160 |
+ static unsigned long vfio_pgsize_bitmap(struct vfio_iommu *iommu) |
1161 |
+@@ -1110,12 +1117,18 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu, |
1162 |
+ goto out_unlock; |
1163 |
+ } |
1164 |
+ |
1165 |
++ if (!iommu->dma_avail) { |
1166 |
++ ret = -ENOSPC; |
1167 |
++ goto out_unlock; |
1168 |
++ } |
1169 |
++ |
1170 |
+ dma = kzalloc(sizeof(*dma), GFP_KERNEL); |
1171 |
+ if (!dma) { |
1172 |
+ ret = -ENOMEM; |
1173 |
+ goto out_unlock; |
1174 |
+ } |
1175 |
+ |
1176 |
++ iommu->dma_avail--; |
1177 |
+ dma->iova = iova; |
1178 |
+ dma->vaddr = vaddr; |
1179 |
+ dma->prot = prot; |
1180 |
+@@ -1612,6 +1625,7 @@ static void *vfio_iommu_type1_open(unsigned long arg) |
1181 |
+ |
1182 |
+ INIT_LIST_HEAD(&iommu->domain_list); |
1183 |
+ iommu->dma_list = RB_ROOT; |
1184 |
++ iommu->dma_avail = dma_entry_limit; |
1185 |
+ mutex_init(&iommu->lock); |
1186 |
+ BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier); |
1187 |
+ |
1188 |
+diff --git a/fs/aio.c b/fs/aio.c |
1189 |
+index 45d5ef8dd0a8..911e23087dfb 100644 |
1190 |
+--- a/fs/aio.c |
1191 |
++++ b/fs/aio.c |
1192 |
+@@ -161,9 +161,13 @@ struct kioctx { |
1193 |
+ unsigned id; |
1194 |
+ }; |
1195 |
+ |
1196 |
++/* |
1197 |
++ * First field must be the file pointer in all the |
1198 |
++ * iocb unions! See also 'struct kiocb' in <linux/fs.h> |
1199 |
++ */ |
1200 |
+ struct fsync_iocb { |
1201 |
+- struct work_struct work; |
1202 |
+ struct file *file; |
1203 |
++ struct work_struct work; |
1204 |
+ bool datasync; |
1205 |
+ }; |
1206 |
+ |
1207 |
+@@ -171,14 +175,21 @@ struct poll_iocb { |
1208 |
+ struct file *file; |
1209 |
+ struct wait_queue_head *head; |
1210 |
+ __poll_t events; |
1211 |
+- bool woken; |
1212 |
++ bool done; |
1213 |
+ bool cancelled; |
1214 |
+ struct wait_queue_entry wait; |
1215 |
+ struct work_struct work; |
1216 |
+ }; |
1217 |
+ |
1218 |
++/* |
1219 |
++ * NOTE! Each of the iocb union members has the file pointer |
1220 |
++ * as the first entry in their struct definition. So you can |
1221 |
++ * access the file pointer through any of the sub-structs, |
1222 |
++ * or directly as just 'ki_filp' in this struct. |
1223 |
++ */ |
1224 |
+ struct aio_kiocb { |
1225 |
+ union { |
1226 |
++ struct file *ki_filp; |
1227 |
+ struct kiocb rw; |
1228 |
+ struct fsync_iocb fsync; |
1229 |
+ struct poll_iocb poll; |
1230 |
+@@ -187,8 +198,7 @@ struct aio_kiocb { |
1231 |
+ struct kioctx *ki_ctx; |
1232 |
+ kiocb_cancel_fn *ki_cancel; |
1233 |
+ |
1234 |
+- struct iocb __user *ki_user_iocb; /* user's aiocb */ |
1235 |
+- __u64 ki_user_data; /* user's data for completion */ |
1236 |
++ struct io_event ki_res; |
1237 |
+ |
1238 |
+ struct list_head ki_list; /* the aio core uses this |
1239 |
+ * for cancellation */ |
1240 |
+@@ -902,7 +912,7 @@ static void put_reqs_available(struct kioctx *ctx, unsigned nr) |
1241 |
+ local_irq_restore(flags); |
1242 |
+ } |
1243 |
+ |
1244 |
+-static bool get_reqs_available(struct kioctx *ctx) |
1245 |
++static bool __get_reqs_available(struct kioctx *ctx) |
1246 |
+ { |
1247 |
+ struct kioctx_cpu *kcpu; |
1248 |
+ bool ret = false; |
1249 |
+@@ -994,32 +1004,35 @@ static void user_refill_reqs_available(struct kioctx *ctx) |
1250 |
+ spin_unlock_irq(&ctx->completion_lock); |
1251 |
+ } |
1252 |
+ |
1253 |
++static bool get_reqs_available(struct kioctx *ctx) |
1254 |
++{ |
1255 |
++ if (__get_reqs_available(ctx)) |
1256 |
++ return true; |
1257 |
++ user_refill_reqs_available(ctx); |
1258 |
++ return __get_reqs_available(ctx); |
1259 |
++} |
1260 |
++ |
1261 |
+ /* aio_get_req |
1262 |
+ * Allocate a slot for an aio request. |
1263 |
+ * Returns NULL if no requests are free. |
1264 |
++ * |
1265 |
++ * The refcount is initialized to 2 - one for the async op completion, |
1266 |
++ * one for the synchronous code that does this. |
1267 |
+ */ |
1268 |
+ static inline struct aio_kiocb *aio_get_req(struct kioctx *ctx) |
1269 |
+ { |
1270 |
+ struct aio_kiocb *req; |
1271 |
+ |
1272 |
+- if (!get_reqs_available(ctx)) { |
1273 |
+- user_refill_reqs_available(ctx); |
1274 |
+- if (!get_reqs_available(ctx)) |
1275 |
+- return NULL; |
1276 |
+- } |
1277 |
+- |
1278 |
+- req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO); |
1279 |
++ req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL); |
1280 |
+ if (unlikely(!req)) |
1281 |
+- goto out_put; |
1282 |
++ return NULL; |
1283 |
+ |
1284 |
+ percpu_ref_get(&ctx->reqs); |
1285 |
+- INIT_LIST_HEAD(&req->ki_list); |
1286 |
+- refcount_set(&req->ki_refcnt, 0); |
1287 |
+ req->ki_ctx = ctx; |
1288 |
++ INIT_LIST_HEAD(&req->ki_list); |
1289 |
++ refcount_set(&req->ki_refcnt, 2); |
1290 |
++ req->ki_eventfd = NULL; |
1291 |
+ return req; |
1292 |
+-out_put: |
1293 |
+- put_reqs_available(ctx, 1); |
1294 |
+- return NULL; |
1295 |
+ } |
1296 |
+ |
1297 |
+ static struct kioctx *lookup_ioctx(unsigned long ctx_id) |
1298 |
+@@ -1050,19 +1063,18 @@ out: |
1299 |
+ return ret; |
1300 |
+ } |
1301 |
+ |
1302 |
+-static inline void iocb_put(struct aio_kiocb *iocb) |
1303 |
++static inline void iocb_destroy(struct aio_kiocb *iocb) |
1304 |
+ { |
1305 |
+- if (refcount_read(&iocb->ki_refcnt) == 0 || |
1306 |
+- refcount_dec_and_test(&iocb->ki_refcnt)) { |
1307 |
+- percpu_ref_put(&iocb->ki_ctx->reqs); |
1308 |
+- kmem_cache_free(kiocb_cachep, iocb); |
1309 |
+- } |
1310 |
++ if (iocb->ki_filp) |
1311 |
++ fput(iocb->ki_filp); |
1312 |
++ percpu_ref_put(&iocb->ki_ctx->reqs); |
1313 |
++ kmem_cache_free(kiocb_cachep, iocb); |
1314 |
+ } |
1315 |
+ |
1316 |
+ /* aio_complete |
1317 |
+ * Called when the io request on the given iocb is complete. |
1318 |
+ */ |
1319 |
+-static void aio_complete(struct aio_kiocb *iocb, long res, long res2) |
1320 |
++static void aio_complete(struct aio_kiocb *iocb) |
1321 |
+ { |
1322 |
+ struct kioctx *ctx = iocb->ki_ctx; |
1323 |
+ struct aio_ring *ring; |
1324 |
+@@ -1086,17 +1098,14 @@ static void aio_complete(struct aio_kiocb *iocb, long res, long res2) |
1325 |
+ ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]); |
1326 |
+ event = ev_page + pos % AIO_EVENTS_PER_PAGE; |
1327 |
+ |
1328 |
+- event->obj = (u64)(unsigned long)iocb->ki_user_iocb; |
1329 |
+- event->data = iocb->ki_user_data; |
1330 |
+- event->res = res; |
1331 |
+- event->res2 = res2; |
1332 |
++ *event = iocb->ki_res; |
1333 |
+ |
1334 |
+ kunmap_atomic(ev_page); |
1335 |
+ flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]); |
1336 |
+ |
1337 |
+- pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n", |
1338 |
+- ctx, tail, iocb, iocb->ki_user_iocb, iocb->ki_user_data, |
1339 |
+- res, res2); |
1340 |
++ pr_debug("%p[%u]: %p: %p %Lx %Lx %Lx\n", ctx, tail, iocb, |
1341 |
++ (void __user *)(unsigned long)iocb->ki_res.obj, |
1342 |
++ iocb->ki_res.data, iocb->ki_res.res, iocb->ki_res.res2); |
1343 |
+ |
1344 |
+ /* after flagging the request as done, we |
1345 |
+ * must never even look at it again |
1346 |
+@@ -1138,7 +1147,14 @@ static void aio_complete(struct aio_kiocb *iocb, long res, long res2) |
1347 |
+ |
1348 |
+ if (waitqueue_active(&ctx->wait)) |
1349 |
+ wake_up(&ctx->wait); |
1350 |
+- iocb_put(iocb); |
1351 |
++} |
1352 |
++ |
1353 |
++static inline void iocb_put(struct aio_kiocb *iocb) |
1354 |
++{ |
1355 |
++ if (refcount_dec_and_test(&iocb->ki_refcnt)) { |
1356 |
++ aio_complete(iocb); |
1357 |
++ iocb_destroy(iocb); |
1358 |
++ } |
1359 |
+ } |
1360 |
+ |
1361 |
+ /* aio_read_events_ring |
1362 |
+@@ -1412,18 +1428,17 @@ static void aio_complete_rw(struct kiocb *kiocb, long res, long res2) |
1363 |
+ file_end_write(kiocb->ki_filp); |
1364 |
+ } |
1365 |
+ |
1366 |
+- fput(kiocb->ki_filp); |
1367 |
+- aio_complete(iocb, res, res2); |
1368 |
++ iocb->ki_res.res = res; |
1369 |
++ iocb->ki_res.res2 = res2; |
1370 |
++ iocb_put(iocb); |
1371 |
+ } |
1372 |
+ |
1373 |
+-static int aio_prep_rw(struct kiocb *req, struct iocb *iocb) |
1374 |
++static int aio_prep_rw(struct kiocb *req, const struct iocb *iocb) |
1375 |
+ { |
1376 |
+ int ret; |
1377 |
+ |
1378 |
+- req->ki_filp = fget(iocb->aio_fildes); |
1379 |
+- if (unlikely(!req->ki_filp)) |
1380 |
+- return -EBADF; |
1381 |
+ req->ki_complete = aio_complete_rw; |
1382 |
++ req->private = NULL; |
1383 |
+ req->ki_pos = iocb->aio_offset; |
1384 |
+ req->ki_flags = iocb_flags(req->ki_filp); |
1385 |
+ if (iocb->aio_flags & IOCB_FLAG_RESFD) |
1386 |
+@@ -1438,7 +1453,6 @@ static int aio_prep_rw(struct kiocb *req, struct iocb *iocb) |
1387 |
+ ret = ioprio_check_cap(iocb->aio_reqprio); |
1388 |
+ if (ret) { |
1389 |
+ pr_debug("aio ioprio check cap error: %d\n", ret); |
1390 |
+- fput(req->ki_filp); |
1391 |
+ return ret; |
1392 |
+ } |
1393 |
+ |
1394 |
+@@ -1448,11 +1462,13 @@ static int aio_prep_rw(struct kiocb *req, struct iocb *iocb) |
1395 |
+ |
1396 |
+ ret = kiocb_set_rw_flags(req, iocb->aio_rw_flags); |
1397 |
+ if (unlikely(ret)) |
1398 |
+- fput(req->ki_filp); |
1399 |
+- return ret; |
1400 |
++ return ret; |
1401 |
++ |
1402 |
++ req->ki_flags &= ~IOCB_HIPRI; /* no one is going to poll for this I/O */ |
1403 |
++ return 0; |
1404 |
+ } |
1405 |
+ |
1406 |
+-static int aio_setup_rw(int rw, struct iocb *iocb, struct iovec **iovec, |
1407 |
++static int aio_setup_rw(int rw, const struct iocb *iocb, struct iovec **iovec, |
1408 |
+ bool vectored, bool compat, struct iov_iter *iter) |
1409 |
+ { |
1410 |
+ void __user *buf = (void __user *)(uintptr_t)iocb->aio_buf; |
1411 |
+@@ -1487,12 +1503,12 @@ static inline void aio_rw_done(struct kiocb *req, ssize_t ret) |
1412 |
+ ret = -EINTR; |
1413 |
+ /*FALLTHRU*/ |
1414 |
+ default: |
1415 |
+- aio_complete_rw(req, ret, 0); |
1416 |
++ req->ki_complete(req, ret, 0); |
1417 |
+ } |
1418 |
+ } |
1419 |
+ |
1420 |
+-static ssize_t aio_read(struct kiocb *req, struct iocb *iocb, bool vectored, |
1421 |
+- bool compat) |
1422 |
++static ssize_t aio_read(struct kiocb *req, const struct iocb *iocb, |
1423 |
++ bool vectored, bool compat) |
1424 |
+ { |
1425 |
+ struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; |
1426 |
+ struct iov_iter iter; |
1427 |
+@@ -1503,29 +1519,24 @@ static ssize_t aio_read(struct kiocb *req, struct iocb *iocb, bool vectored, |
1428 |
+ if (ret) |
1429 |
+ return ret; |
1430 |
+ file = req->ki_filp; |
1431 |
+- |
1432 |
+- ret = -EBADF; |
1433 |
+ if (unlikely(!(file->f_mode & FMODE_READ))) |
1434 |
+- goto out_fput; |
1435 |
++ return -EBADF; |
1436 |
+ ret = -EINVAL; |
1437 |
+ if (unlikely(!file->f_op->read_iter)) |
1438 |
+- goto out_fput; |
1439 |
++ return -EINVAL; |
1440 |
+ |
1441 |
+ ret = aio_setup_rw(READ, iocb, &iovec, vectored, compat, &iter); |
1442 |
+ if (ret) |
1443 |
+- goto out_fput; |
1444 |
++ return ret; |
1445 |
+ ret = rw_verify_area(READ, file, &req->ki_pos, iov_iter_count(&iter)); |
1446 |
+ if (!ret) |
1447 |
+ aio_rw_done(req, call_read_iter(file, req, &iter)); |
1448 |
+ kfree(iovec); |
1449 |
+-out_fput: |
1450 |
+- if (unlikely(ret)) |
1451 |
+- fput(file); |
1452 |
+ return ret; |
1453 |
+ } |
1454 |
+ |
1455 |
+-static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored, |
1456 |
+- bool compat) |
1457 |
++static ssize_t aio_write(struct kiocb *req, const struct iocb *iocb, |
1458 |
++ bool vectored, bool compat) |
1459 |
+ { |
1460 |
+ struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs; |
1461 |
+ struct iov_iter iter; |
1462 |
+@@ -1537,16 +1548,14 @@ static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored, |
1463 |
+ return ret; |
1464 |
+ file = req->ki_filp; |
1465 |
+ |
1466 |
+- ret = -EBADF; |
1467 |
+ if (unlikely(!(file->f_mode & FMODE_WRITE))) |
1468 |
+- goto out_fput; |
1469 |
+- ret = -EINVAL; |
1470 |
++ return -EBADF; |
1471 |
+ if (unlikely(!file->f_op->write_iter)) |
1472 |
+- goto out_fput; |
1473 |
++ return -EINVAL; |
1474 |
+ |
1475 |
+ ret = aio_setup_rw(WRITE, iocb, &iovec, vectored, compat, &iter); |
1476 |
+ if (ret) |
1477 |
+- goto out_fput; |
1478 |
++ return ret; |
1479 |
+ ret = rw_verify_area(WRITE, file, &req->ki_pos, iov_iter_count(&iter)); |
1480 |
+ if (!ret) { |
1481 |
+ /* |
1482 |
+@@ -1564,35 +1573,26 @@ static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored, |
1483 |
+ aio_rw_done(req, call_write_iter(file, req, &iter)); |
1484 |
+ } |
1485 |
+ kfree(iovec); |
1486 |
+-out_fput: |
1487 |
+- if (unlikely(ret)) |
1488 |
+- fput(file); |
1489 |
+ return ret; |
1490 |
+ } |
1491 |
+ |
1492 |
+ static void aio_fsync_work(struct work_struct *work) |
1493 |
+ { |
1494 |
+- struct fsync_iocb *req = container_of(work, struct fsync_iocb, work); |
1495 |
+- int ret; |
1496 |
++ struct aio_kiocb *iocb = container_of(work, struct aio_kiocb, fsync.work); |
1497 |
+ |
1498 |
+- ret = vfs_fsync(req->file, req->datasync); |
1499 |
+- fput(req->file); |
1500 |
+- aio_complete(container_of(req, struct aio_kiocb, fsync), ret, 0); |
1501 |
++ iocb->ki_res.res = vfs_fsync(iocb->fsync.file, iocb->fsync.datasync); |
1502 |
++ iocb_put(iocb); |
1503 |
+ } |
1504 |
+ |
1505 |
+-static int aio_fsync(struct fsync_iocb *req, struct iocb *iocb, bool datasync) |
1506 |
++static int aio_fsync(struct fsync_iocb *req, const struct iocb *iocb, |
1507 |
++ bool datasync) |
1508 |
+ { |
1509 |
+ if (unlikely(iocb->aio_buf || iocb->aio_offset || iocb->aio_nbytes || |
1510 |
+ iocb->aio_rw_flags)) |
1511 |
+ return -EINVAL; |
1512 |
+ |
1513 |
+- req->file = fget(iocb->aio_fildes); |
1514 |
+- if (unlikely(!req->file)) |
1515 |
+- return -EBADF; |
1516 |
+- if (unlikely(!req->file->f_op->fsync)) { |
1517 |
+- fput(req->file); |
1518 |
++ if (unlikely(!req->file->f_op->fsync)) |
1519 |
+ return -EINVAL; |
1520 |
+- } |
1521 |
+ |
1522 |
+ req->datasync = datasync; |
1523 |
+ INIT_WORK(&req->work, aio_fsync_work); |
1524 |
+@@ -1600,14 +1600,6 @@ static int aio_fsync(struct fsync_iocb *req, struct iocb *iocb, bool datasync) |
1525 |
+ return 0; |
1526 |
+ } |
1527 |
+ |
1528 |
+-static inline void aio_poll_complete(struct aio_kiocb *iocb, __poll_t mask) |
1529 |
+-{ |
1530 |
+- struct file *file = iocb->poll.file; |
1531 |
+- |
1532 |
+- aio_complete(iocb, mangle_poll(mask), 0); |
1533 |
+- fput(file); |
1534 |
+-} |
1535 |
+- |
1536 |
+ static void aio_poll_complete_work(struct work_struct *work) |
1537 |
+ { |
1538 |
+ struct poll_iocb *req = container_of(work, struct poll_iocb, work); |
1539 |
+@@ -1633,9 +1625,11 @@ static void aio_poll_complete_work(struct work_struct *work) |
1540 |
+ return; |
1541 |
+ } |
1542 |
+ list_del_init(&iocb->ki_list); |
1543 |
++ iocb->ki_res.res = mangle_poll(mask); |
1544 |
++ req->done = true; |
1545 |
+ spin_unlock_irq(&ctx->ctx_lock); |
1546 |
+ |
1547 |
+- aio_poll_complete(iocb, mask); |
1548 |
++ iocb_put(iocb); |
1549 |
+ } |
1550 |
+ |
1551 |
+ /* assumes we are called with irqs disabled */ |
1552 |
+@@ -1663,31 +1657,27 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync, |
1553 |
+ __poll_t mask = key_to_poll(key); |
1554 |
+ unsigned long flags; |
1555 |
+ |
1556 |
+- req->woken = true; |
1557 |
+- |
1558 |
+ /* for instances that support it check for an event match first: */ |
1559 |
+- if (mask) { |
1560 |
+- if (!(mask & req->events)) |
1561 |
+- return 0; |
1562 |
++ if (mask && !(mask & req->events)) |
1563 |
++ return 0; |
1564 |
++ |
1565 |
++ list_del_init(&req->wait.entry); |
1566 |
+ |
1567 |
++ if (mask && spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) { |
1568 |
+ /* |
1569 |
+ * Try to complete the iocb inline if we can. Use |
1570 |
+ * irqsave/irqrestore because not all filesystems (e.g. fuse) |
1571 |
+ * call this function with IRQs disabled and because IRQs |
1572 |
+ * have to be disabled before ctx_lock is obtained. |
1573 |
+ */ |
1574 |
+- if (spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) { |
1575 |
+- list_del(&iocb->ki_list); |
1576 |
+- spin_unlock_irqrestore(&iocb->ki_ctx->ctx_lock, flags); |
1577 |
+- |
1578 |
+- list_del_init(&req->wait.entry); |
1579 |
+- aio_poll_complete(iocb, mask); |
1580 |
+- return 1; |
1581 |
+- } |
1582 |
++ list_del(&iocb->ki_list); |
1583 |
++ iocb->ki_res.res = mangle_poll(mask); |
1584 |
++ req->done = true; |
1585 |
++ spin_unlock_irqrestore(&iocb->ki_ctx->ctx_lock, flags); |
1586 |
++ iocb_put(iocb); |
1587 |
++ } else { |
1588 |
++ schedule_work(&req->work); |
1589 |
+ } |
1590 |
+- |
1591 |
+- list_del_init(&req->wait.entry); |
1592 |
+- schedule_work(&req->work); |
1593 |
+ return 1; |
1594 |
+ } |
1595 |
+ |
1596 |
+@@ -1714,11 +1704,12 @@ aio_poll_queue_proc(struct file *file, struct wait_queue_head *head, |
1597 |
+ add_wait_queue(head, &pt->iocb->poll.wait); |
1598 |
+ } |
1599 |
+ |
1600 |
+-static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb) |
1601 |
++static ssize_t aio_poll(struct aio_kiocb *aiocb, const struct iocb *iocb) |
1602 |
+ { |
1603 |
+ struct kioctx *ctx = aiocb->ki_ctx; |
1604 |
+ struct poll_iocb *req = &aiocb->poll; |
1605 |
+ struct aio_poll_table apt; |
1606 |
++ bool cancel = false; |
1607 |
+ __poll_t mask; |
1608 |
+ |
1609 |
+ /* reject any unknown events outside the normal event mask. */ |
1610 |
+@@ -1730,9 +1721,10 @@ static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb) |
1611 |
+ |
1612 |
+ INIT_WORK(&req->work, aio_poll_complete_work); |
1613 |
+ req->events = demangle_poll(iocb->aio_buf) | EPOLLERR | EPOLLHUP; |
1614 |
+- req->file = fget(iocb->aio_fildes); |
1615 |
+- if (unlikely(!req->file)) |
1616 |
+- return -EBADF; |
1617 |
++ |
1618 |
++ req->head = NULL; |
1619 |
++ req->done = false; |
1620 |
++ req->cancelled = false; |
1621 |
+ |
1622 |
+ apt.pt._qproc = aio_poll_queue_proc; |
1623 |
+ apt.pt._key = req->events; |
1624 |
+@@ -1743,83 +1735,79 @@ static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb) |
1625 |
+ INIT_LIST_HEAD(&req->wait.entry); |
1626 |
+ init_waitqueue_func_entry(&req->wait, aio_poll_wake); |
1627 |
+ |
1628 |
+- /* one for removal from waitqueue, one for this function */ |
1629 |
+- refcount_set(&aiocb->ki_refcnt, 2); |
1630 |
+- |
1631 |
+ mask = vfs_poll(req->file, &apt.pt) & req->events; |
1632 |
+- if (unlikely(!req->head)) { |
1633 |
+- /* we did not manage to set up a waitqueue, done */ |
1634 |
+- goto out; |
1635 |
+- } |
1636 |
+- |
1637 |
+ spin_lock_irq(&ctx->ctx_lock); |
1638 |
+- spin_lock(&req->head->lock); |
1639 |
+- if (req->woken) { |
1640 |
+- /* wake_up context handles the rest */ |
1641 |
+- mask = 0; |
1642 |
++ if (likely(req->head)) { |
1643 |
++ spin_lock(&req->head->lock); |
1644 |
++ if (unlikely(list_empty(&req->wait.entry))) { |
1645 |
++ if (apt.error) |
1646 |
++ cancel = true; |
1647 |
++ apt.error = 0; |
1648 |
++ mask = 0; |
1649 |
++ } |
1650 |
++ if (mask || apt.error) { |
1651 |
++ list_del_init(&req->wait.entry); |
1652 |
++ } else if (cancel) { |
1653 |
++ WRITE_ONCE(req->cancelled, true); |
1654 |
++ } else if (!req->done) { /* actually waiting for an event */ |
1655 |
++ list_add_tail(&aiocb->ki_list, &ctx->active_reqs); |
1656 |
++ aiocb->ki_cancel = aio_poll_cancel; |
1657 |
++ } |
1658 |
++ spin_unlock(&req->head->lock); |
1659 |
++ } |
1660 |
++ if (mask) { /* no async, we'd stolen it */ |
1661 |
++ aiocb->ki_res.res = mangle_poll(mask); |
1662 |
+ apt.error = 0; |
1663 |
+- } else if (mask || apt.error) { |
1664 |
+- /* if we get an error or a mask we are done */ |
1665 |
+- WARN_ON_ONCE(list_empty(&req->wait.entry)); |
1666 |
+- list_del_init(&req->wait.entry); |
1667 |
+- } else { |
1668 |
+- /* actually waiting for an event */ |
1669 |
+- list_add_tail(&aiocb->ki_list, &ctx->active_reqs); |
1670 |
+- aiocb->ki_cancel = aio_poll_cancel; |
1671 |
+ } |
1672 |
+- spin_unlock(&req->head->lock); |
1673 |
+ spin_unlock_irq(&ctx->ctx_lock); |
1674 |
+- |
1675 |
+-out: |
1676 |
+- if (unlikely(apt.error)) { |
1677 |
+- fput(req->file); |
1678 |
+- return apt.error; |
1679 |
+- } |
1680 |
+- |
1681 |
+ if (mask) |
1682 |
+- aio_poll_complete(aiocb, mask); |
1683 |
+- iocb_put(aiocb); |
1684 |
+- return 0; |
1685 |
++ iocb_put(aiocb); |
1686 |
++ return apt.error; |
1687 |
+ } |
1688 |
+ |
1689 |
+-static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, |
1690 |
+- bool compat) |
1691 |
++static int __io_submit_one(struct kioctx *ctx, const struct iocb *iocb, |
1692 |
++ struct iocb __user *user_iocb, bool compat) |
1693 |
+ { |
1694 |
+ struct aio_kiocb *req; |
1695 |
+- struct iocb iocb; |
1696 |
+ ssize_t ret; |
1697 |
+ |
1698 |
+- if (unlikely(copy_from_user(&iocb, user_iocb, sizeof(iocb)))) |
1699 |
+- return -EFAULT; |
1700 |
+- |
1701 |
+ /* enforce forwards compatibility on users */ |
1702 |
+- if (unlikely(iocb.aio_reserved2)) { |
1703 |
++ if (unlikely(iocb->aio_reserved2)) { |
1704 |
+ pr_debug("EINVAL: reserve field set\n"); |
1705 |
+ return -EINVAL; |
1706 |
+ } |
1707 |
+ |
1708 |
+ /* prevent overflows */ |
1709 |
+ if (unlikely( |
1710 |
+- (iocb.aio_buf != (unsigned long)iocb.aio_buf) || |
1711 |
+- (iocb.aio_nbytes != (size_t)iocb.aio_nbytes) || |
1712 |
+- ((ssize_t)iocb.aio_nbytes < 0) |
1713 |
++ (iocb->aio_buf != (unsigned long)iocb->aio_buf) || |
1714 |
++ (iocb->aio_nbytes != (size_t)iocb->aio_nbytes) || |
1715 |
++ ((ssize_t)iocb->aio_nbytes < 0) |
1716 |
+ )) { |
1717 |
+ pr_debug("EINVAL: overflow check\n"); |
1718 |
+ return -EINVAL; |
1719 |
+ } |
1720 |
+ |
1721 |
++ if (!get_reqs_available(ctx)) |
1722 |
++ return -EAGAIN; |
1723 |
++ |
1724 |
++ ret = -EAGAIN; |
1725 |
+ req = aio_get_req(ctx); |
1726 |
+ if (unlikely(!req)) |
1727 |
+- return -EAGAIN; |
1728 |
++ goto out_put_reqs_available; |
1729 |
++ |
1730 |
++ req->ki_filp = fget(iocb->aio_fildes); |
1731 |
++ ret = -EBADF; |
1732 |
++ if (unlikely(!req->ki_filp)) |
1733 |
++ goto out_put_req; |
1734 |
+ |
1735 |
+- if (iocb.aio_flags & IOCB_FLAG_RESFD) { |
1736 |
++ if (iocb->aio_flags & IOCB_FLAG_RESFD) { |
1737 |
+ /* |
1738 |
+ * If the IOCB_FLAG_RESFD flag of aio_flags is set, get an |
1739 |
+ * instance of the file* now. The file descriptor must be |
1740 |
+ * an eventfd() fd, and will be signaled for each completed |
1741 |
+ * event using the eventfd_signal() function. |
1742 |
+ */ |
1743 |
+- req->ki_eventfd = eventfd_ctx_fdget((int) iocb.aio_resfd); |
1744 |
++ req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd); |
1745 |
+ if (IS_ERR(req->ki_eventfd)) { |
1746 |
+ ret = PTR_ERR(req->ki_eventfd); |
1747 |
+ req->ki_eventfd = NULL; |
1748 |
+@@ -1833,54 +1821,70 @@ static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, |
1749 |
+ goto out_put_req; |
1750 |
+ } |
1751 |
+ |
1752 |
+- req->ki_user_iocb = user_iocb; |
1753 |
+- req->ki_user_data = iocb.aio_data; |
1754 |
++ req->ki_res.obj = (u64)(unsigned long)user_iocb; |
1755 |
++ req->ki_res.data = iocb->aio_data; |
1756 |
++ req->ki_res.res = 0; |
1757 |
++ req->ki_res.res2 = 0; |
1758 |
+ |
1759 |
+- switch (iocb.aio_lio_opcode) { |
1760 |
++ switch (iocb->aio_lio_opcode) { |
1761 |
+ case IOCB_CMD_PREAD: |
1762 |
+- ret = aio_read(&req->rw, &iocb, false, compat); |
1763 |
++ ret = aio_read(&req->rw, iocb, false, compat); |
1764 |
+ break; |
1765 |
+ case IOCB_CMD_PWRITE: |
1766 |
+- ret = aio_write(&req->rw, &iocb, false, compat); |
1767 |
++ ret = aio_write(&req->rw, iocb, false, compat); |
1768 |
+ break; |
1769 |
+ case IOCB_CMD_PREADV: |
1770 |
+- ret = aio_read(&req->rw, &iocb, true, compat); |
1771 |
++ ret = aio_read(&req->rw, iocb, true, compat); |
1772 |
+ break; |
1773 |
+ case IOCB_CMD_PWRITEV: |
1774 |
+- ret = aio_write(&req->rw, &iocb, true, compat); |
1775 |
++ ret = aio_write(&req->rw, iocb, true, compat); |
1776 |
+ break; |
1777 |
+ case IOCB_CMD_FSYNC: |
1778 |
+- ret = aio_fsync(&req->fsync, &iocb, false); |
1779 |
++ ret = aio_fsync(&req->fsync, iocb, false); |
1780 |
+ break; |
1781 |
+ case IOCB_CMD_FDSYNC: |
1782 |
+- ret = aio_fsync(&req->fsync, &iocb, true); |
1783 |
++ ret = aio_fsync(&req->fsync, iocb, true); |
1784 |
+ break; |
1785 |
+ case IOCB_CMD_POLL: |
1786 |
+- ret = aio_poll(req, &iocb); |
1787 |
++ ret = aio_poll(req, iocb); |
1788 |
+ break; |
1789 |
+ default: |
1790 |
+- pr_debug("invalid aio operation %d\n", iocb.aio_lio_opcode); |
1791 |
++ pr_debug("invalid aio operation %d\n", iocb->aio_lio_opcode); |
1792 |
+ ret = -EINVAL; |
1793 |
+ break; |
1794 |
+ } |
1795 |
+ |
1796 |
++ /* Done with the synchronous reference */ |
1797 |
++ iocb_put(req); |
1798 |
++ |
1799 |
+ /* |
1800 |
+ * If ret is 0, we'd either done aio_complete() ourselves or have |
1801 |
+ * arranged for that to be done asynchronously. Anything non-zero |
1802 |
+ * means that we need to destroy req ourselves. |
1803 |
+ */ |
1804 |
+- if (ret) |
1805 |
+- goto out_put_req; |
1806 |
+- return 0; |
1807 |
++ if (!ret) |
1808 |
++ return 0; |
1809 |
++ |
1810 |
+ out_put_req: |
1811 |
+- put_reqs_available(ctx, 1); |
1812 |
+- percpu_ref_put(&ctx->reqs); |
1813 |
+ if (req->ki_eventfd) |
1814 |
+ eventfd_ctx_put(req->ki_eventfd); |
1815 |
+- kmem_cache_free(kiocb_cachep, req); |
1816 |
++ iocb_destroy(req); |
1817 |
++out_put_reqs_available: |
1818 |
++ put_reqs_available(ctx, 1); |
1819 |
+ return ret; |
1820 |
+ } |
1821 |
+ |
1822 |
++static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, |
1823 |
++ bool compat) |
1824 |
++{ |
1825 |
++ struct iocb iocb; |
1826 |
++ |
1827 |
++ if (unlikely(copy_from_user(&iocb, user_iocb, sizeof(iocb)))) |
1828 |
++ return -EFAULT; |
1829 |
++ |
1830 |
++ return __io_submit_one(ctx, &iocb, user_iocb, compat); |
1831 |
++} |
1832 |
++ |
1833 |
+ /* sys_io_submit: |
1834 |
+ * Queue the nr iocbs pointed to by iocbpp for processing. Returns |
1835 |
+ * the number of iocbs queued. May return -EINVAL if the aio_context |
1836 |
+@@ -1973,24 +1977,6 @@ COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id, |
1837 |
+ } |
1838 |
+ #endif |
1839 |
+ |
1840 |
+-/* lookup_kiocb |
1841 |
+- * Finds a given iocb for cancellation. |
1842 |
+- */ |
1843 |
+-static struct aio_kiocb * |
1844 |
+-lookup_kiocb(struct kioctx *ctx, struct iocb __user *iocb) |
1845 |
+-{ |
1846 |
+- struct aio_kiocb *kiocb; |
1847 |
+- |
1848 |
+- assert_spin_locked(&ctx->ctx_lock); |
1849 |
+- |
1850 |
+- /* TODO: use a hash or array, this sucks. */ |
1851 |
+- list_for_each_entry(kiocb, &ctx->active_reqs, ki_list) { |
1852 |
+- if (kiocb->ki_user_iocb == iocb) |
1853 |
+- return kiocb; |
1854 |
+- } |
1855 |
+- return NULL; |
1856 |
+-} |
1857 |
+- |
1858 |
+ /* sys_io_cancel: |
1859 |
+ * Attempts to cancel an iocb previously passed to io_submit. If |
1860 |
+ * the operation is successfully cancelled, the resulting event is |
1861 |
+@@ -2008,6 +1994,7 @@ SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb, |
1862 |
+ struct aio_kiocb *kiocb; |
1863 |
+ int ret = -EINVAL; |
1864 |
+ u32 key; |
1865 |
++ u64 obj = (u64)(unsigned long)iocb; |
1866 |
+ |
1867 |
+ if (unlikely(get_user(key, &iocb->aio_key))) |
1868 |
+ return -EFAULT; |
1869 |
+@@ -2019,10 +2006,13 @@ SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb, |
1870 |
+ return -EINVAL; |
1871 |
+ |
1872 |
+ spin_lock_irq(&ctx->ctx_lock); |
1873 |
+- kiocb = lookup_kiocb(ctx, iocb); |
1874 |
+- if (kiocb) { |
1875 |
+- ret = kiocb->ki_cancel(&kiocb->rw); |
1876 |
+- list_del_init(&kiocb->ki_list); |
1877 |
++ /* TODO: use a hash or array, this sucks. */ |
1878 |
++ list_for_each_entry(kiocb, &ctx->active_reqs, ki_list) { |
1879 |
++ if (kiocb->ki_res.obj == obj) { |
1880 |
++ ret = kiocb->ki_cancel(&kiocb->rw); |
1881 |
++ list_del_init(&kiocb->ki_list); |
1882 |
++ break; |
1883 |
++ } |
1884 |
+ } |
1885 |
+ spin_unlock_irq(&ctx->ctx_lock); |
1886 |
+ |
1887 |
+diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c |
1888 |
+index 82928cea0209..7f3f64ba464f 100644 |
1889 |
+--- a/fs/ceph/dir.c |
1890 |
++++ b/fs/ceph/dir.c |
1891 |
+@@ -1470,6 +1470,7 @@ void ceph_dentry_lru_del(struct dentry *dn) |
1892 |
+ unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn) |
1893 |
+ { |
1894 |
+ struct ceph_inode_info *dci = ceph_inode(dir); |
1895 |
++ unsigned hash; |
1896 |
+ |
1897 |
+ switch (dci->i_dir_layout.dl_dir_hash) { |
1898 |
+ case 0: /* for backward compat */ |
1899 |
+@@ -1477,8 +1478,11 @@ unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn) |
1900 |
+ return dn->d_name.hash; |
1901 |
+ |
1902 |
+ default: |
1903 |
+- return ceph_str_hash(dci->i_dir_layout.dl_dir_hash, |
1904 |
++ spin_lock(&dn->d_lock); |
1905 |
++ hash = ceph_str_hash(dci->i_dir_layout.dl_dir_hash, |
1906 |
+ dn->d_name.name, dn->d_name.len); |
1907 |
++ spin_unlock(&dn->d_lock); |
1908 |
++ return hash; |
1909 |
+ } |
1910 |
+ } |
1911 |
+ |
1912 |
+diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c |
1913 |
+index bc43c822426a..bfcf11c70bfa 100644 |
1914 |
+--- a/fs/ceph/mds_client.c |
1915 |
++++ b/fs/ceph/mds_client.c |
1916 |
+@@ -1290,6 +1290,15 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap, |
1917 |
+ list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove); |
1918 |
+ ci->i_prealloc_cap_flush = NULL; |
1919 |
+ } |
1920 |
++ |
1921 |
++ if (drop && |
1922 |
++ ci->i_wrbuffer_ref_head == 0 && |
1923 |
++ ci->i_wr_ref == 0 && |
1924 |
++ ci->i_dirty_caps == 0 && |
1925 |
++ ci->i_flushing_caps == 0) { |
1926 |
++ ceph_put_snap_context(ci->i_head_snapc); |
1927 |
++ ci->i_head_snapc = NULL; |
1928 |
++ } |
1929 |
+ } |
1930 |
+ spin_unlock(&ci->i_ceph_lock); |
1931 |
+ while (!list_empty(&to_remove)) { |
1932 |
+@@ -1945,10 +1954,39 @@ retry: |
1933 |
+ return path; |
1934 |
+ } |
1935 |
+ |
1936 |
++/* Duplicate the dentry->d_name.name safely */ |
1937 |
++static int clone_dentry_name(struct dentry *dentry, const char **ppath, |
1938 |
++ int *ppathlen) |
1939 |
++{ |
1940 |
++ u32 len; |
1941 |
++ char *name; |
1942 |
++ |
1943 |
++retry: |
1944 |
++ len = READ_ONCE(dentry->d_name.len); |
1945 |
++ name = kmalloc(len + 1, GFP_NOFS); |
1946 |
++ if (!name) |
1947 |
++ return -ENOMEM; |
1948 |
++ |
1949 |
++ spin_lock(&dentry->d_lock); |
1950 |
++ if (dentry->d_name.len != len) { |
1951 |
++ spin_unlock(&dentry->d_lock); |
1952 |
++ kfree(name); |
1953 |
++ goto retry; |
1954 |
++ } |
1955 |
++ memcpy(name, dentry->d_name.name, len); |
1956 |
++ spin_unlock(&dentry->d_lock); |
1957 |
++ |
1958 |
++ name[len] = '\0'; |
1959 |
++ *ppath = name; |
1960 |
++ *ppathlen = len; |
1961 |
++ return 0; |
1962 |
++} |
1963 |
++ |
1964 |
+ static int build_dentry_path(struct dentry *dentry, struct inode *dir, |
1965 |
+ const char **ppath, int *ppathlen, u64 *pino, |
1966 |
+- int *pfreepath) |
1967 |
++ bool *pfreepath, bool parent_locked) |
1968 |
+ { |
1969 |
++ int ret; |
1970 |
+ char *path; |
1971 |
+ |
1972 |
+ rcu_read_lock(); |
1973 |
+@@ -1957,8 +1995,15 @@ static int build_dentry_path(struct dentry *dentry, struct inode *dir, |
1974 |
+ if (dir && ceph_snap(dir) == CEPH_NOSNAP) { |
1975 |
+ *pino = ceph_ino(dir); |
1976 |
+ rcu_read_unlock(); |
1977 |
+- *ppath = dentry->d_name.name; |
1978 |
+- *ppathlen = dentry->d_name.len; |
1979 |
++ if (parent_locked) { |
1980 |
++ *ppath = dentry->d_name.name; |
1981 |
++ *ppathlen = dentry->d_name.len; |
1982 |
++ } else { |
1983 |
++ ret = clone_dentry_name(dentry, ppath, ppathlen); |
1984 |
++ if (ret) |
1985 |
++ return ret; |
1986 |
++ *pfreepath = true; |
1987 |
++ } |
1988 |
+ return 0; |
1989 |
+ } |
1990 |
+ rcu_read_unlock(); |
1991 |
+@@ -1966,13 +2011,13 @@ static int build_dentry_path(struct dentry *dentry, struct inode *dir, |
1992 |
+ if (IS_ERR(path)) |
1993 |
+ return PTR_ERR(path); |
1994 |
+ *ppath = path; |
1995 |
+- *pfreepath = 1; |
1996 |
++ *pfreepath = true; |
1997 |
+ return 0; |
1998 |
+ } |
1999 |
+ |
2000 |
+ static int build_inode_path(struct inode *inode, |
2001 |
+ const char **ppath, int *ppathlen, u64 *pino, |
2002 |
+- int *pfreepath) |
2003 |
++ bool *pfreepath) |
2004 |
+ { |
2005 |
+ struct dentry *dentry; |
2006 |
+ char *path; |
2007 |
+@@ -1988,7 +2033,7 @@ static int build_inode_path(struct inode *inode, |
2008 |
+ if (IS_ERR(path)) |
2009 |
+ return PTR_ERR(path); |
2010 |
+ *ppath = path; |
2011 |
+- *pfreepath = 1; |
2012 |
++ *pfreepath = true; |
2013 |
+ return 0; |
2014 |
+ } |
2015 |
+ |
2016 |
+@@ -1999,7 +2044,7 @@ static int build_inode_path(struct inode *inode, |
2017 |
+ static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry, |
2018 |
+ struct inode *rdiri, const char *rpath, |
2019 |
+ u64 rino, const char **ppath, int *pathlen, |
2020 |
+- u64 *ino, int *freepath) |
2021 |
++ u64 *ino, bool *freepath, bool parent_locked) |
2022 |
+ { |
2023 |
+ int r = 0; |
2024 |
+ |
2025 |
+@@ -2009,7 +2054,7 @@ static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry, |
2026 |
+ ceph_snap(rinode)); |
2027 |
+ } else if (rdentry) { |
2028 |
+ r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino, |
2029 |
+- freepath); |
2030 |
++ freepath, parent_locked); |
2031 |
+ dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen, |
2032 |
+ *ppath); |
2033 |
+ } else if (rpath || rino) { |
2034 |
+@@ -2035,7 +2080,7 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc, |
2035 |
+ const char *path2 = NULL; |
2036 |
+ u64 ino1 = 0, ino2 = 0; |
2037 |
+ int pathlen1 = 0, pathlen2 = 0; |
2038 |
+- int freepath1 = 0, freepath2 = 0; |
2039 |
++ bool freepath1 = false, freepath2 = false; |
2040 |
+ int len; |
2041 |
+ u16 releases; |
2042 |
+ void *p, *end; |
2043 |
+@@ -2043,16 +2088,19 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc, |
2044 |
+ |
2045 |
+ ret = set_request_path_attr(req->r_inode, req->r_dentry, |
2046 |
+ req->r_parent, req->r_path1, req->r_ino1.ino, |
2047 |
+- &path1, &pathlen1, &ino1, &freepath1); |
2048 |
++ &path1, &pathlen1, &ino1, &freepath1, |
2049 |
++ test_bit(CEPH_MDS_R_PARENT_LOCKED, |
2050 |
++ &req->r_req_flags)); |
2051 |
+ if (ret < 0) { |
2052 |
+ msg = ERR_PTR(ret); |
2053 |
+ goto out; |
2054 |
+ } |
2055 |
+ |
2056 |
++ /* If r_old_dentry is set, then assume that its parent is locked */ |
2057 |
+ ret = set_request_path_attr(NULL, req->r_old_dentry, |
2058 |
+ req->r_old_dentry_dir, |
2059 |
+ req->r_path2, req->r_ino2.ino, |
2060 |
+- &path2, &pathlen2, &ino2, &freepath2); |
2061 |
++ &path2, &pathlen2, &ino2, &freepath2, true); |
2062 |
+ if (ret < 0) { |
2063 |
+ msg = ERR_PTR(ret); |
2064 |
+ goto out_free1; |
2065 |
+diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c |
2066 |
+index f74193da0e09..1f46b02f7314 100644 |
2067 |
+--- a/fs/ceph/snap.c |
2068 |
++++ b/fs/ceph/snap.c |
2069 |
+@@ -568,7 +568,12 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci) |
2070 |
+ old_snapc = NULL; |
2071 |
+ |
2072 |
+ update_snapc: |
2073 |
+- if (ci->i_head_snapc) { |
2074 |
++ if (ci->i_wrbuffer_ref_head == 0 && |
2075 |
++ ci->i_wr_ref == 0 && |
2076 |
++ ci->i_dirty_caps == 0 && |
2077 |
++ ci->i_flushing_caps == 0) { |
2078 |
++ ci->i_head_snapc = NULL; |
2079 |
++ } else { |
2080 |
+ ci->i_head_snapc = ceph_get_snap_context(new_snapc); |
2081 |
+ dout(" new snapc is %p\n", new_snapc); |
2082 |
+ } |
2083 |
+diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c |
2084 |
+index b59ebed4f615..1fadd314ae7f 100644 |
2085 |
+--- a/fs/cifs/inode.c |
2086 |
++++ b/fs/cifs/inode.c |
2087 |
+@@ -1735,6 +1735,10 @@ cifs_do_rename(const unsigned int xid, struct dentry *from_dentry, |
2088 |
+ if (rc == 0 || rc != -EBUSY) |
2089 |
+ goto do_rename_exit; |
2090 |
+ |
2091 |
++ /* Don't fall back to using SMB on SMB 2+ mount */ |
2092 |
++ if (server->vals->protocol_id != 0) |
2093 |
++ goto do_rename_exit; |
2094 |
++ |
2095 |
+ /* open-file renames don't work across directories */ |
2096 |
+ if (to_dentry->d_parent != from_dentry->d_parent) |
2097 |
+ goto do_rename_exit; |
2098 |
+diff --git a/fs/cifs/smb2pdu.c b/fs/cifs/smb2pdu.c |
2099 |
+index c6fd3acc5560..33afb637e6f8 100644 |
2100 |
+--- a/fs/cifs/smb2pdu.c |
2101 |
++++ b/fs/cifs/smb2pdu.c |
2102 |
+@@ -3285,6 +3285,7 @@ SMB2_read(const unsigned int xid, struct cifs_io_parms *io_parms, |
2103 |
+ rc); |
2104 |
+ } |
2105 |
+ free_rsp_buf(resp_buftype, rsp_iov.iov_base); |
2106 |
++ cifs_small_buf_release(req); |
2107 |
+ return rc == -ENODATA ? 0 : rc; |
2108 |
+ } else |
2109 |
+ trace_smb3_read_done(xid, req->PersistentFileId, |
2110 |
+diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c |
2111 |
+index c0ba5206cd9d..006c277dc22e 100644 |
2112 |
+--- a/fs/ext4/xattr.c |
2113 |
++++ b/fs/ext4/xattr.c |
2114 |
+@@ -829,6 +829,7 @@ int ext4_get_inode_usage(struct inode *inode, qsize_t *usage) |
2115 |
+ bh = ext4_sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl, REQ_PRIO); |
2116 |
+ if (IS_ERR(bh)) { |
2117 |
+ ret = PTR_ERR(bh); |
2118 |
++ bh = NULL; |
2119 |
+ goto out; |
2120 |
+ } |
2121 |
+ |
2122 |
+@@ -2907,6 +2908,7 @@ int ext4_xattr_delete_inode(handle_t *handle, struct inode *inode, |
2123 |
+ if (error == -EIO) |
2124 |
+ EXT4_ERROR_INODE(inode, "block %llu read error", |
2125 |
+ EXT4_I(inode)->i_file_acl); |
2126 |
++ bh = NULL; |
2127 |
+ goto cleanup; |
2128 |
+ } |
2129 |
+ error = ext4_xattr_check_block(inode, bh); |
2130 |
+@@ -3063,6 +3065,7 @@ ext4_xattr_block_cache_find(struct inode *inode, |
2131 |
+ if (IS_ERR(bh)) { |
2132 |
+ if (PTR_ERR(bh) == -ENOMEM) |
2133 |
+ return NULL; |
2134 |
++ bh = NULL; |
2135 |
+ EXT4_ERROR_INODE(inode, "block %lu read error", |
2136 |
+ (unsigned long)ce->e_value); |
2137 |
+ } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) { |
2138 |
+diff --git a/fs/nfs/super.c b/fs/nfs/super.c |
2139 |
+index 6b666d187907..6df9b85caf20 100644 |
2140 |
+--- a/fs/nfs/super.c |
2141 |
++++ b/fs/nfs/super.c |
2142 |
+@@ -2052,7 +2052,8 @@ static int nfs23_validate_mount_data(void *options, |
2143 |
+ memcpy(sap, &data->addr, sizeof(data->addr)); |
2144 |
+ args->nfs_server.addrlen = sizeof(data->addr); |
2145 |
+ args->nfs_server.port = ntohs(data->addr.sin_port); |
2146 |
+- if (!nfs_verify_server_address(sap)) |
2147 |
++ if (sap->sa_family != AF_INET || |
2148 |
++ !nfs_verify_server_address(sap)) |
2149 |
+ goto out_no_address; |
2150 |
+ |
2151 |
+ if (!(data->flags & NFS_MOUNT_TCP)) |
2152 |
+diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c |
2153 |
+index 601bf33c26a0..ebbb0285addb 100644 |
2154 |
+--- a/fs/nfsd/nfs4callback.c |
2155 |
++++ b/fs/nfsd/nfs4callback.c |
2156 |
+@@ -926,8 +926,9 @@ static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata) |
2157 |
+ cb->cb_seq_status = 1; |
2158 |
+ cb->cb_status = 0; |
2159 |
+ if (minorversion) { |
2160 |
+- if (!nfsd41_cb_get_slot(clp, task)) |
2161 |
++ if (!cb->cb_holds_slot && !nfsd41_cb_get_slot(clp, task)) |
2162 |
+ return; |
2163 |
++ cb->cb_holds_slot = true; |
2164 |
+ } |
2165 |
+ rpc_call_start(task); |
2166 |
+ } |
2167 |
+@@ -954,6 +955,9 @@ static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback |
2168 |
+ return true; |
2169 |
+ } |
2170 |
+ |
2171 |
++ if (!cb->cb_holds_slot) |
2172 |
++ goto need_restart; |
2173 |
++ |
2174 |
+ switch (cb->cb_seq_status) { |
2175 |
+ case 0: |
2176 |
+ /* |
2177 |
+@@ -992,6 +996,7 @@ static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback |
2178 |
+ cb->cb_seq_status); |
2179 |
+ } |
2180 |
+ |
2181 |
++ cb->cb_holds_slot = false; |
2182 |
+ clear_bit(0, &clp->cl_cb_slot_busy); |
2183 |
+ rpc_wake_up_next(&clp->cl_cb_waitq); |
2184 |
+ dprintk("%s: freed slot, new seqid=%d\n", __func__, |
2185 |
+@@ -1199,6 +1204,7 @@ void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp, |
2186 |
+ cb->cb_seq_status = 1; |
2187 |
+ cb->cb_status = 0; |
2188 |
+ cb->cb_need_restart = false; |
2189 |
++ cb->cb_holds_slot = false; |
2190 |
+ } |
2191 |
+ |
2192 |
+ void nfsd4_run_cb(struct nfsd4_callback *cb) |
2193 |
+diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h |
2194 |
+index 0b15dac7e609..0f07ad6dc1ef 100644 |
2195 |
+--- a/fs/nfsd/state.h |
2196 |
++++ b/fs/nfsd/state.h |
2197 |
+@@ -70,6 +70,7 @@ struct nfsd4_callback { |
2198 |
+ int cb_seq_status; |
2199 |
+ int cb_status; |
2200 |
+ bool cb_need_restart; |
2201 |
++ bool cb_holds_slot; |
2202 |
+ }; |
2203 |
+ |
2204 |
+ struct nfsd4_callback_ops { |
2205 |
+diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c |
2206 |
+index d65390727541..7325baa8f9d4 100644 |
2207 |
+--- a/fs/proc/proc_sysctl.c |
2208 |
++++ b/fs/proc/proc_sysctl.c |
2209 |
+@@ -1626,9 +1626,11 @@ static void drop_sysctl_table(struct ctl_table_header *header) |
2210 |
+ if (--header->nreg) |
2211 |
+ return; |
2212 |
+ |
2213 |
+- if (parent) |
2214 |
++ if (parent) { |
2215 |
+ put_links(header); |
2216 |
+- start_unregistering(header); |
2217 |
++ start_unregistering(header); |
2218 |
++ } |
2219 |
++ |
2220 |
+ if (!--header->count) |
2221 |
+ kfree_rcu(header, rcu); |
2222 |
+ |
2223 |
+diff --git a/fs/splice.c b/fs/splice.c |
2224 |
+index 29e92b506394..c78e0e3ff6c4 100644 |
2225 |
+--- a/fs/splice.c |
2226 |
++++ b/fs/splice.c |
2227 |
+@@ -333,8 +333,8 @@ const struct pipe_buf_operations default_pipe_buf_ops = { |
2228 |
+ .get = generic_pipe_buf_get, |
2229 |
+ }; |
2230 |
+ |
2231 |
+-static int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe, |
2232 |
+- struct pipe_buffer *buf) |
2233 |
++int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe, |
2234 |
++ struct pipe_buffer *buf) |
2235 |
+ { |
2236 |
+ return 1; |
2237 |
+ } |
2238 |
+diff --git a/include/linux/fs.h b/include/linux/fs.h |
2239 |
+index 7b6084854bfe..111c94c4baa1 100644 |
2240 |
+--- a/include/linux/fs.h |
2241 |
++++ b/include/linux/fs.h |
2242 |
+@@ -304,13 +304,19 @@ enum rw_hint { |
2243 |
+ |
2244 |
+ struct kiocb { |
2245 |
+ struct file *ki_filp; |
2246 |
++ |
2247 |
++ /* The 'ki_filp' pointer is shared in a union for aio */ |
2248 |
++ randomized_struct_fields_start |
2249 |
++ |
2250 |
+ loff_t ki_pos; |
2251 |
+ void (*ki_complete)(struct kiocb *iocb, long ret, long ret2); |
2252 |
+ void *private; |
2253 |
+ int ki_flags; |
2254 |
+ u16 ki_hint; |
2255 |
+ u16 ki_ioprio; /* See linux/ioprio.h */ |
2256 |
+-} __randomize_layout; |
2257 |
++ |
2258 |
++ randomized_struct_fields_end |
2259 |
++}; |
2260 |
+ |
2261 |
+ static inline bool is_sync_kiocb(struct kiocb *kiocb) |
2262 |
+ { |
2263 |
+diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h |
2264 |
+index 3ecd7ea212ae..66ee63cd5968 100644 |
2265 |
+--- a/include/linux/pipe_fs_i.h |
2266 |
++++ b/include/linux/pipe_fs_i.h |
2267 |
+@@ -181,6 +181,7 @@ void free_pipe_info(struct pipe_inode_info *); |
2268 |
+ void generic_pipe_buf_get(struct pipe_inode_info *, struct pipe_buffer *); |
2269 |
+ int generic_pipe_buf_confirm(struct pipe_inode_info *, struct pipe_buffer *); |
2270 |
+ int generic_pipe_buf_steal(struct pipe_inode_info *, struct pipe_buffer *); |
2271 |
++int generic_pipe_buf_nosteal(struct pipe_inode_info *, struct pipe_buffer *); |
2272 |
+ void generic_pipe_buf_release(struct pipe_inode_info *, struct pipe_buffer *); |
2273 |
+ void pipe_buf_mark_unmergeable(struct pipe_buffer *buf); |
2274 |
+ |
2275 |
+diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h |
2276 |
+index 0f39ac487012..f2be5d041ba3 100644 |
2277 |
+--- a/include/net/netfilter/nf_tables.h |
2278 |
++++ b/include/net/netfilter/nf_tables.h |
2279 |
+@@ -382,6 +382,7 @@ void nft_unregister_set(struct nft_set_type *type); |
2280 |
+ * @dtype: data type (verdict or numeric type defined by userspace) |
2281 |
+ * @objtype: object type (see NFT_OBJECT_* definitions) |
2282 |
+ * @size: maximum set size |
2283 |
++ * @use: number of rules references to this set |
2284 |
+ * @nelems: number of elements |
2285 |
+ * @ndeact: number of deactivated elements queued for removal |
2286 |
+ * @timeout: default timeout value in jiffies |
2287 |
+@@ -407,6 +408,7 @@ struct nft_set { |
2288 |
+ u32 dtype; |
2289 |
+ u32 objtype; |
2290 |
+ u32 size; |
2291 |
++ u32 use; |
2292 |
+ atomic_t nelems; |
2293 |
+ u32 ndeact; |
2294 |
+ u64 timeout; |
2295 |
+@@ -416,7 +418,8 @@ struct nft_set { |
2296 |
+ unsigned char *udata; |
2297 |
+ /* runtime data below here */ |
2298 |
+ const struct nft_set_ops *ops ____cacheline_aligned; |
2299 |
+- u16 flags:14, |
2300 |
++ u16 flags:13, |
2301 |
++ bound:1, |
2302 |
+ genmask:2; |
2303 |
+ u8 klen; |
2304 |
+ u8 dlen; |
2305 |
+@@ -466,10 +469,15 @@ struct nft_set_binding { |
2306 |
+ u32 flags; |
2307 |
+ }; |
2308 |
+ |
2309 |
++enum nft_trans_phase; |
2310 |
++void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set, |
2311 |
++ struct nft_set_binding *binding, |
2312 |
++ enum nft_trans_phase phase); |
2313 |
+ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set, |
2314 |
+ struct nft_set_binding *binding); |
2315 |
+ void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set, |
2316 |
+- struct nft_set_binding *binding); |
2317 |
++ struct nft_set_binding *binding, bool commit); |
2318 |
++void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set); |
2319 |
+ |
2320 |
+ /** |
2321 |
+ * enum nft_set_extensions - set extension type IDs |
2322 |
+@@ -689,10 +697,12 @@ static inline void nft_set_gc_batch_add(struct nft_set_gc_batch *gcb, |
2323 |
+ gcb->elems[gcb->head.cnt++] = elem; |
2324 |
+ } |
2325 |
+ |
2326 |
++struct nft_expr_ops; |
2327 |
+ /** |
2328 |
+ * struct nft_expr_type - nf_tables expression type |
2329 |
+ * |
2330 |
+ * @select_ops: function to select nft_expr_ops |
2331 |
++ * @release_ops: release nft_expr_ops |
2332 |
+ * @ops: default ops, used when no select_ops functions is present |
2333 |
+ * @list: used internally |
2334 |
+ * @name: Identifier |
2335 |
+@@ -705,6 +715,7 @@ static inline void nft_set_gc_batch_add(struct nft_set_gc_batch *gcb, |
2336 |
+ struct nft_expr_type { |
2337 |
+ const struct nft_expr_ops *(*select_ops)(const struct nft_ctx *, |
2338 |
+ const struct nlattr * const tb[]); |
2339 |
++ void (*release_ops)(const struct nft_expr_ops *ops); |
2340 |
+ const struct nft_expr_ops *ops; |
2341 |
+ struct list_head list; |
2342 |
+ const char *name; |
2343 |
+@@ -718,13 +729,22 @@ struct nft_expr_type { |
2344 |
+ #define NFT_EXPR_STATEFUL 0x1 |
2345 |
+ #define NFT_EXPR_GC 0x2 |
2346 |
+ |
2347 |
++enum nft_trans_phase { |
2348 |
++ NFT_TRANS_PREPARE, |
2349 |
++ NFT_TRANS_ABORT, |
2350 |
++ NFT_TRANS_COMMIT, |
2351 |
++ NFT_TRANS_RELEASE |
2352 |
++}; |
2353 |
++ |
2354 |
+ /** |
2355 |
+ * struct nft_expr_ops - nf_tables expression operations |
2356 |
+ * |
2357 |
+ * @eval: Expression evaluation function |
2358 |
+ * @size: full expression size, including private data size |
2359 |
+ * @init: initialization function |
2360 |
+- * @destroy: destruction function |
2361 |
++ * @activate: activate expression in the next generation |
2362 |
++ * @deactivate: deactivate expression in next generation |
2363 |
++ * @destroy: destruction function, called after synchronize_rcu |
2364 |
+ * @dump: function to dump parameters |
2365 |
+ * @type: expression type |
2366 |
+ * @validate: validate expression, called during loop detection |
2367 |
+@@ -745,7 +765,8 @@ struct nft_expr_ops { |
2368 |
+ void (*activate)(const struct nft_ctx *ctx, |
2369 |
+ const struct nft_expr *expr); |
2370 |
+ void (*deactivate)(const struct nft_ctx *ctx, |
2371 |
+- const struct nft_expr *expr); |
2372 |
++ const struct nft_expr *expr, |
2373 |
++ enum nft_trans_phase phase); |
2374 |
+ void (*destroy)(const struct nft_ctx *ctx, |
2375 |
+ const struct nft_expr *expr); |
2376 |
+ void (*destroy_clone)(const struct nft_ctx *ctx, |
2377 |
+diff --git a/include/net/netrom.h b/include/net/netrom.h |
2378 |
+index 5a0714ff500f..80f15b1c1a48 100644 |
2379 |
+--- a/include/net/netrom.h |
2380 |
++++ b/include/net/netrom.h |
2381 |
+@@ -266,7 +266,7 @@ void nr_stop_idletimer(struct sock *); |
2382 |
+ int nr_t1timer_running(struct sock *); |
2383 |
+ |
2384 |
+ /* sysctl_net_netrom.c */ |
2385 |
+-void nr_register_sysctl(void); |
2386 |
++int nr_register_sysctl(void); |
2387 |
+ void nr_unregister_sysctl(void); |
2388 |
+ |
2389 |
+ #endif |
2390 |
+diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c |
2391 |
+index 91e4202b0634..72c07059ef37 100644 |
2392 |
+--- a/kernel/sched/deadline.c |
2393 |
++++ b/kernel/sched/deadline.c |
2394 |
+@@ -252,7 +252,6 @@ static void task_non_contending(struct task_struct *p) |
2395 |
+ if (dl_entity_is_special(dl_se)) |
2396 |
+ return; |
2397 |
+ |
2398 |
+- WARN_ON(hrtimer_active(&dl_se->inactive_timer)); |
2399 |
+ WARN_ON(dl_se->dl_non_contending); |
2400 |
+ |
2401 |
+ zerolag_time = dl_se->deadline - |
2402 |
+@@ -269,7 +268,7 @@ static void task_non_contending(struct task_struct *p) |
2403 |
+ * If the "0-lag time" already passed, decrease the active |
2404 |
+ * utilization now, instead of starting a timer |
2405 |
+ */ |
2406 |
+- if (zerolag_time < 0) { |
2407 |
++ if ((zerolag_time < 0) || hrtimer_active(&dl_se->inactive_timer)) { |
2408 |
+ if (dl_task(p)) |
2409 |
+ sub_running_bw(dl_se, dl_rq); |
2410 |
+ if (!dl_task(p) || p->state == TASK_DEAD) { |
2411 |
+diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c |
2412 |
+index 4aa8e7d90c25..d31916366d39 100644 |
2413 |
+--- a/kernel/sched/fair.c |
2414 |
++++ b/kernel/sched/fair.c |
2415 |
+@@ -2016,6 +2016,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period) |
2416 |
+ if (p->last_task_numa_placement) { |
2417 |
+ delta = runtime - p->last_sum_exec_runtime; |
2418 |
+ *period = now - p->last_task_numa_placement; |
2419 |
++ |
2420 |
++ /* Avoid time going backwards, prevent potential divide error: */ |
2421 |
++ if (unlikely((s64)*period < 0)) |
2422 |
++ *period = 0; |
2423 |
+ } else { |
2424 |
+ delta = p->se.avg.load_sum; |
2425 |
+ *period = LOAD_AVG_MAX; |
2426 |
+diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c |
2427 |
+index 34b4c32b0692..805aef83b5cf 100644 |
2428 |
+--- a/kernel/trace/ring_buffer.c |
2429 |
++++ b/kernel/trace/ring_buffer.c |
2430 |
+@@ -730,7 +730,7 @@ u64 ring_buffer_time_stamp(struct ring_buffer *buffer, int cpu) |
2431 |
+ |
2432 |
+ preempt_disable_notrace(); |
2433 |
+ time = rb_time_stamp(buffer); |
2434 |
+- preempt_enable_no_resched_notrace(); |
2435 |
++ preempt_enable_notrace(); |
2436 |
+ |
2437 |
+ return time; |
2438 |
+ } |
2439 |
+diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c |
2440 |
+index c65cea71d1ee..5455ee05bc3b 100644 |
2441 |
+--- a/kernel/trace/trace.c |
2442 |
++++ b/kernel/trace/trace.c |
2443 |
+@@ -496,8 +496,10 @@ int trace_pid_write(struct trace_pid_list *filtered_pids, |
2444 |
+ * not modified. |
2445 |
+ */ |
2446 |
+ pid_list = kmalloc(sizeof(*pid_list), GFP_KERNEL); |
2447 |
+- if (!pid_list) |
2448 |
++ if (!pid_list) { |
2449 |
++ trace_parser_put(&parser); |
2450 |
+ return -ENOMEM; |
2451 |
++ } |
2452 |
+ |
2453 |
+ pid_list->pid_max = READ_ONCE(pid_max); |
2454 |
+ |
2455 |
+@@ -507,6 +509,7 @@ int trace_pid_write(struct trace_pid_list *filtered_pids, |
2456 |
+ |
2457 |
+ pid_list->pids = vzalloc((pid_list->pid_max + 7) >> 3); |
2458 |
+ if (!pid_list->pids) { |
2459 |
++ trace_parser_put(&parser); |
2460 |
+ kfree(pid_list); |
2461 |
+ return -ENOMEM; |
2462 |
+ } |
2463 |
+@@ -6800,19 +6803,23 @@ struct buffer_ref { |
2464 |
+ struct ring_buffer *buffer; |
2465 |
+ void *page; |
2466 |
+ int cpu; |
2467 |
+- int ref; |
2468 |
++ refcount_t refcount; |
2469 |
+ }; |
2470 |
+ |
2471 |
++static void buffer_ref_release(struct buffer_ref *ref) |
2472 |
++{ |
2473 |
++ if (!refcount_dec_and_test(&ref->refcount)) |
2474 |
++ return; |
2475 |
++ ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page); |
2476 |
++ kfree(ref); |
2477 |
++} |
2478 |
++ |
2479 |
+ static void buffer_pipe_buf_release(struct pipe_inode_info *pipe, |
2480 |
+ struct pipe_buffer *buf) |
2481 |
+ { |
2482 |
+ struct buffer_ref *ref = (struct buffer_ref *)buf->private; |
2483 |
+ |
2484 |
+- if (--ref->ref) |
2485 |
+- return; |
2486 |
+- |
2487 |
+- ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page); |
2488 |
+- kfree(ref); |
2489 |
++ buffer_ref_release(ref); |
2490 |
+ buf->private = 0; |
2491 |
+ } |
2492 |
+ |
2493 |
+@@ -6821,7 +6828,7 @@ static void buffer_pipe_buf_get(struct pipe_inode_info *pipe, |
2494 |
+ { |
2495 |
+ struct buffer_ref *ref = (struct buffer_ref *)buf->private; |
2496 |
+ |
2497 |
+- ref->ref++; |
2498 |
++ refcount_inc(&ref->refcount); |
2499 |
+ } |
2500 |
+ |
2501 |
+ /* Pipe buffer operations for a buffer. */ |
2502 |
+@@ -6829,7 +6836,7 @@ static const struct pipe_buf_operations buffer_pipe_buf_ops = { |
2503 |
+ .can_merge = 0, |
2504 |
+ .confirm = generic_pipe_buf_confirm, |
2505 |
+ .release = buffer_pipe_buf_release, |
2506 |
+- .steal = generic_pipe_buf_steal, |
2507 |
++ .steal = generic_pipe_buf_nosteal, |
2508 |
+ .get = buffer_pipe_buf_get, |
2509 |
+ }; |
2510 |
+ |
2511 |
+@@ -6842,11 +6849,7 @@ static void buffer_spd_release(struct splice_pipe_desc *spd, unsigned int i) |
2512 |
+ struct buffer_ref *ref = |
2513 |
+ (struct buffer_ref *)spd->partial[i].private; |
2514 |
+ |
2515 |
+- if (--ref->ref) |
2516 |
+- return; |
2517 |
+- |
2518 |
+- ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page); |
2519 |
+- kfree(ref); |
2520 |
++ buffer_ref_release(ref); |
2521 |
+ spd->partial[i].private = 0; |
2522 |
+ } |
2523 |
+ |
2524 |
+@@ -6901,7 +6904,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos, |
2525 |
+ break; |
2526 |
+ } |
2527 |
+ |
2528 |
+- ref->ref = 1; |
2529 |
++ refcount_set(&ref->refcount, 1); |
2530 |
+ ref->buffer = iter->trace_buffer->buffer; |
2531 |
+ ref->page = ring_buffer_alloc_read_page(ref->buffer, iter->cpu_file); |
2532 |
+ if (IS_ERR(ref->page)) { |
2533 |
+diff --git a/kernel/workqueue.c b/kernel/workqueue.c |
2534 |
+index 0280deac392e..cd8b61bded78 100644 |
2535 |
+--- a/kernel/workqueue.c |
2536 |
++++ b/kernel/workqueue.c |
2537 |
+@@ -2908,6 +2908,9 @@ static bool __flush_work(struct work_struct *work, bool from_cancel) |
2538 |
+ if (WARN_ON(!wq_online)) |
2539 |
+ return false; |
2540 |
+ |
2541 |
++ if (WARN_ON(!work->func)) |
2542 |
++ return false; |
2543 |
++ |
2544 |
+ if (!from_cancel) { |
2545 |
+ lock_map_acquire(&work->lockdep_map); |
2546 |
+ lock_map_release(&work->lockdep_map); |
2547 |
+diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug |
2548 |
+index 4966c4fbe7f7..3dea52f7be9c 100644 |
2549 |
+--- a/lib/Kconfig.debug |
2550 |
++++ b/lib/Kconfig.debug |
2551 |
+@@ -1934,6 +1934,7 @@ config TEST_KMOD |
2552 |
+ depends on m |
2553 |
+ depends on BLOCK && (64BIT || LBDAF) # for XFS, BTRFS |
2554 |
+ depends on NETDEVICES && NET_CORE && INET # for TUN |
2555 |
++ depends on BLOCK |
2556 |
+ select TEST_LKM |
2557 |
+ select XFS_FS |
2558 |
+ select TUN |
2559 |
+diff --git a/mm/memory.c b/mm/memory.c |
2560 |
+index 5b3f71bcd1ae..9c69278173b7 100644 |
2561 |
+--- a/mm/memory.c |
2562 |
++++ b/mm/memory.c |
2563 |
+@@ -1787,10 +1787,15 @@ static int insert_pfn(struct vm_area_struct *vma, unsigned long addr, |
2564 |
+ * in may not match the PFN we have mapped if the |
2565 |
+ * mapped PFN is a writeable COW page. In the mkwrite |
2566 |
+ * case we are creating a writable PTE for a shared |
2567 |
+- * mapping and we expect the PFNs to match. |
2568 |
++ * mapping and we expect the PFNs to match. If they |
2569 |
++ * don't match, we are likely racing with block |
2570 |
++ * allocation and mapping invalidation so just skip the |
2571 |
++ * update. |
2572 |
+ */ |
2573 |
+- if (WARN_ON_ONCE(pte_pfn(*pte) != pfn_t_to_pfn(pfn))) |
2574 |
++ if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) { |
2575 |
++ WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte))); |
2576 |
+ goto out_unlock; |
2577 |
++ } |
2578 |
+ entry = *pte; |
2579 |
+ goto out_mkwrite; |
2580 |
+ } else |
2581 |
+diff --git a/net/bridge/netfilter/ebtables.c b/net/bridge/netfilter/ebtables.c |
2582 |
+index f77888ec93f1..0bb4d712b80c 100644 |
2583 |
+--- a/net/bridge/netfilter/ebtables.c |
2584 |
++++ b/net/bridge/netfilter/ebtables.c |
2585 |
+@@ -2032,7 +2032,8 @@ static int ebt_size_mwt(struct compat_ebt_entry_mwt *match32, |
2586 |
+ if (match_kern) |
2587 |
+ match_kern->match_size = ret; |
2588 |
+ |
2589 |
+- if (WARN_ON(type == EBT_COMPAT_TARGET && size_left)) |
2590 |
++ /* rule should have no remaining data after target */ |
2591 |
++ if (type == EBT_COMPAT_TARGET && size_left) |
2592 |
+ return -EINVAL; |
2593 |
+ |
2594 |
+ match32 = (struct compat_ebt_entry_mwt *) buf; |
2595 |
+diff --git a/net/ipv4/route.c b/net/ipv4/route.c |
2596 |
+index 98c81c21b753..8bacbcd2db90 100644 |
2597 |
+--- a/net/ipv4/route.c |
2598 |
++++ b/net/ipv4/route.c |
2599 |
+@@ -1185,25 +1185,39 @@ static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie) |
2600 |
+ return dst; |
2601 |
+ } |
2602 |
+ |
2603 |
+-static void ipv4_link_failure(struct sk_buff *skb) |
2604 |
++static void ipv4_send_dest_unreach(struct sk_buff *skb) |
2605 |
+ { |
2606 |
+ struct ip_options opt; |
2607 |
+- struct rtable *rt; |
2608 |
+ int res; |
2609 |
+ |
2610 |
+ /* Recompile ip options since IPCB may not be valid anymore. |
2611 |
++ * Also check we have a reasonable ipv4 header. |
2612 |
+ */ |
2613 |
+- memset(&opt, 0, sizeof(opt)); |
2614 |
+- opt.optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr); |
2615 |
++ if (!pskb_network_may_pull(skb, sizeof(struct iphdr)) || |
2616 |
++ ip_hdr(skb)->version != 4 || ip_hdr(skb)->ihl < 5) |
2617 |
++ return; |
2618 |
+ |
2619 |
+- rcu_read_lock(); |
2620 |
+- res = __ip_options_compile(dev_net(skb->dev), &opt, skb, NULL); |
2621 |
+- rcu_read_unlock(); |
2622 |
++ memset(&opt, 0, sizeof(opt)); |
2623 |
++ if (ip_hdr(skb)->ihl > 5) { |
2624 |
++ if (!pskb_network_may_pull(skb, ip_hdr(skb)->ihl * 4)) |
2625 |
++ return; |
2626 |
++ opt.optlen = ip_hdr(skb)->ihl * 4 - sizeof(struct iphdr); |
2627 |
+ |
2628 |
+- if (res) |
2629 |
+- return; |
2630 |
++ rcu_read_lock(); |
2631 |
++ res = __ip_options_compile(dev_net(skb->dev), &opt, skb, NULL); |
2632 |
++ rcu_read_unlock(); |
2633 |
+ |
2634 |
++ if (res) |
2635 |
++ return; |
2636 |
++ } |
2637 |
+ __icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0, &opt); |
2638 |
++} |
2639 |
++ |
2640 |
++static void ipv4_link_failure(struct sk_buff *skb) |
2641 |
++{ |
2642 |
++ struct rtable *rt; |
2643 |
++ |
2644 |
++ ipv4_send_dest_unreach(skb); |
2645 |
+ |
2646 |
+ rt = skb_rtable(skb); |
2647 |
+ if (rt) |
2648 |
+diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c |
2649 |
+index 891ed2f91467..ce64453d337d 100644 |
2650 |
+--- a/net/ipv4/sysctl_net_ipv4.c |
2651 |
++++ b/net/ipv4/sysctl_net_ipv4.c |
2652 |
+@@ -49,6 +49,7 @@ static int ip_ping_group_range_min[] = { 0, 0 }; |
2653 |
+ static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX }; |
2654 |
+ static int comp_sack_nr_max = 255; |
2655 |
+ static u32 u32_max_div_HZ = UINT_MAX / HZ; |
2656 |
++static int one_day_secs = 24 * 3600; |
2657 |
+ |
2658 |
+ /* obsolete */ |
2659 |
+ static int sysctl_tcp_low_latency __read_mostly; |
2660 |
+@@ -1140,7 +1141,9 @@ static struct ctl_table ipv4_net_table[] = { |
2661 |
+ .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen, |
2662 |
+ .maxlen = sizeof(int), |
2663 |
+ .mode = 0644, |
2664 |
+- .proc_handler = proc_dointvec |
2665 |
++ .proc_handler = proc_dointvec_minmax, |
2666 |
++ .extra1 = &zero, |
2667 |
++ .extra2 = &one_day_secs |
2668 |
+ }, |
2669 |
+ { |
2670 |
+ .procname = "tcp_autocorking", |
2671 |
+diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c |
2672 |
+index 8fd8d06454d6..2d4e048762f6 100644 |
2673 |
+--- a/net/netfilter/ipvs/ip_vs_ctl.c |
2674 |
++++ b/net/netfilter/ipvs/ip_vs_ctl.c |
2675 |
+@@ -896,12 +896,13 @@ ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest, |
2676 |
+ { |
2677 |
+ struct ip_vs_dest *dest; |
2678 |
+ unsigned int atype, i; |
2679 |
+- int ret = 0; |
2680 |
+ |
2681 |
+ EnterFunction(2); |
2682 |
+ |
2683 |
+ #ifdef CONFIG_IP_VS_IPV6 |
2684 |
+ if (udest->af == AF_INET6) { |
2685 |
++ int ret; |
2686 |
++ |
2687 |
+ atype = ipv6_addr_type(&udest->addr.in6); |
2688 |
+ if ((!(atype & IPV6_ADDR_UNICAST) || |
2689 |
+ atype & IPV6_ADDR_LINKLOCAL) && |
2690 |
+diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c |
2691 |
+index c06393fc716d..1af54119bafc 100644 |
2692 |
+--- a/net/netfilter/nf_tables_api.c |
2693 |
++++ b/net/netfilter/nf_tables_api.c |
2694 |
+@@ -112,6 +112,23 @@ static void nft_trans_destroy(struct nft_trans *trans) |
2695 |
+ kfree(trans); |
2696 |
+ } |
2697 |
+ |
2698 |
++static void nft_set_trans_bind(const struct nft_ctx *ctx, struct nft_set *set) |
2699 |
++{ |
2700 |
++ struct net *net = ctx->net; |
2701 |
++ struct nft_trans *trans; |
2702 |
++ |
2703 |
++ if (!nft_set_is_anonymous(set)) |
2704 |
++ return; |
2705 |
++ |
2706 |
++ list_for_each_entry_reverse(trans, &net->nft.commit_list, list) { |
2707 |
++ if (trans->msg_type == NFT_MSG_NEWSET && |
2708 |
++ nft_trans_set(trans) == set) { |
2709 |
++ set->bound = true; |
2710 |
++ break; |
2711 |
++ } |
2712 |
++ } |
2713 |
++} |
2714 |
++ |
2715 |
+ static int nf_tables_register_hook(struct net *net, |
2716 |
+ const struct nft_table *table, |
2717 |
+ struct nft_chain *chain) |
2718 |
+@@ -222,14 +239,15 @@ static void nft_rule_expr_activate(const struct nft_ctx *ctx, |
2719 |
+ } |
2720 |
+ |
2721 |
+ static void nft_rule_expr_deactivate(const struct nft_ctx *ctx, |
2722 |
+- struct nft_rule *rule) |
2723 |
++ struct nft_rule *rule, |
2724 |
++ enum nft_trans_phase phase) |
2725 |
+ { |
2726 |
+ struct nft_expr *expr; |
2727 |
+ |
2728 |
+ expr = nft_expr_first(rule); |
2729 |
+ while (expr != nft_expr_last(rule) && expr->ops) { |
2730 |
+ if (expr->ops->deactivate) |
2731 |
+- expr->ops->deactivate(ctx, expr); |
2732 |
++ expr->ops->deactivate(ctx, expr, phase); |
2733 |
+ |
2734 |
+ expr = nft_expr_next(expr); |
2735 |
+ } |
2736 |
+@@ -280,7 +298,7 @@ static int nft_delrule(struct nft_ctx *ctx, struct nft_rule *rule) |
2737 |
+ nft_trans_destroy(trans); |
2738 |
+ return err; |
2739 |
+ } |
2740 |
+- nft_rule_expr_deactivate(ctx, rule); |
2741 |
++ nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_PREPARE); |
2742 |
+ |
2743 |
+ return 0; |
2744 |
+ } |
2745 |
+@@ -301,7 +319,7 @@ static int nft_delrule_by_chain(struct nft_ctx *ctx) |
2746 |
+ return 0; |
2747 |
+ } |
2748 |
+ |
2749 |
+-static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type, |
2750 |
++static int nft_trans_set_add(const struct nft_ctx *ctx, int msg_type, |
2751 |
+ struct nft_set *set) |
2752 |
+ { |
2753 |
+ struct nft_trans *trans; |
2754 |
+@@ -321,7 +339,7 @@ static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type, |
2755 |
+ return 0; |
2756 |
+ } |
2757 |
+ |
2758 |
+-static int nft_delset(struct nft_ctx *ctx, struct nft_set *set) |
2759 |
++static int nft_delset(const struct nft_ctx *ctx, struct nft_set *set) |
2760 |
+ { |
2761 |
+ int err; |
2762 |
+ |
2763 |
+@@ -2105,6 +2123,7 @@ struct nft_expr *nft_expr_init(const struct nft_ctx *ctx, |
2764 |
+ { |
2765 |
+ struct nft_expr_info info; |
2766 |
+ struct nft_expr *expr; |
2767 |
++ struct module *owner; |
2768 |
+ int err; |
2769 |
+ |
2770 |
+ err = nf_tables_expr_parse(ctx, nla, &info); |
2771 |
+@@ -2124,7 +2143,11 @@ struct nft_expr *nft_expr_init(const struct nft_ctx *ctx, |
2772 |
+ err3: |
2773 |
+ kfree(expr); |
2774 |
+ err2: |
2775 |
+- module_put(info.ops->type->owner); |
2776 |
++ owner = info.ops->type->owner; |
2777 |
++ if (info.ops->type->release_ops) |
2778 |
++ info.ops->type->release_ops(info.ops); |
2779 |
++ |
2780 |
++ module_put(owner); |
2781 |
+ err1: |
2782 |
+ return ERR_PTR(err); |
2783 |
+ } |
2784 |
+@@ -2458,7 +2481,7 @@ static void nf_tables_rule_destroy(const struct nft_ctx *ctx, |
2785 |
+ static void nf_tables_rule_release(const struct nft_ctx *ctx, |
2786 |
+ struct nft_rule *rule) |
2787 |
+ { |
2788 |
+- nft_rule_expr_deactivate(ctx, rule); |
2789 |
++ nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_RELEASE); |
2790 |
+ nf_tables_rule_destroy(ctx, rule); |
2791 |
+ } |
2792 |
+ |
2793 |
+@@ -3562,19 +3585,15 @@ err1: |
2794 |
+ |
2795 |
+ static void nft_set_destroy(struct nft_set *set) |
2796 |
+ { |
2797 |
++ if (WARN_ON(set->use > 0)) |
2798 |
++ return; |
2799 |
++ |
2800 |
+ set->ops->destroy(set); |
2801 |
+ module_put(to_set_type(set->ops)->owner); |
2802 |
+ kfree(set->name); |
2803 |
+ kvfree(set); |
2804 |
+ } |
2805 |
+ |
2806 |
+-static void nf_tables_set_destroy(const struct nft_ctx *ctx, struct nft_set *set) |
2807 |
+-{ |
2808 |
+- list_del_rcu(&set->list); |
2809 |
+- nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, GFP_ATOMIC); |
2810 |
+- nft_set_destroy(set); |
2811 |
+-} |
2812 |
+- |
2813 |
+ static int nf_tables_delset(struct net *net, struct sock *nlsk, |
2814 |
+ struct sk_buff *skb, const struct nlmsghdr *nlh, |
2815 |
+ const struct nlattr * const nla[], |
2816 |
+@@ -3609,7 +3628,7 @@ static int nf_tables_delset(struct net *net, struct sock *nlsk, |
2817 |
+ NL_SET_BAD_ATTR(extack, attr); |
2818 |
+ return PTR_ERR(set); |
2819 |
+ } |
2820 |
+- if (!list_empty(&set->bindings) || |
2821 |
++ if (set->use || |
2822 |
+ (nlh->nlmsg_flags & NLM_F_NONREC && atomic_read(&set->nelems) > 0)) { |
2823 |
+ NL_SET_BAD_ATTR(extack, attr); |
2824 |
+ return -EBUSY; |
2825 |
+@@ -3639,6 +3658,9 @@ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set, |
2826 |
+ struct nft_set_binding *i; |
2827 |
+ struct nft_set_iter iter; |
2828 |
+ |
2829 |
++ if (set->use == UINT_MAX) |
2830 |
++ return -EOVERFLOW; |
2831 |
++ |
2832 |
+ if (!list_empty(&set->bindings) && nft_set_is_anonymous(set)) |
2833 |
+ return -EBUSY; |
2834 |
+ |
2835 |
+@@ -3665,21 +3687,53 @@ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set, |
2836 |
+ bind: |
2837 |
+ binding->chain = ctx->chain; |
2838 |
+ list_add_tail_rcu(&binding->list, &set->bindings); |
2839 |
++ nft_set_trans_bind(ctx, set); |
2840 |
++ set->use++; |
2841 |
++ |
2842 |
+ return 0; |
2843 |
+ } |
2844 |
+ EXPORT_SYMBOL_GPL(nf_tables_bind_set); |
2845 |
+ |
2846 |
+ void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set, |
2847 |
+- struct nft_set_binding *binding) |
2848 |
++ struct nft_set_binding *binding, bool event) |
2849 |
+ { |
2850 |
+ list_del_rcu(&binding->list); |
2851 |
+ |
2852 |
+- if (list_empty(&set->bindings) && nft_set_is_anonymous(set) && |
2853 |
+- nft_is_active(ctx->net, set)) |
2854 |
+- nf_tables_set_destroy(ctx, set); |
2855 |
++ if (list_empty(&set->bindings) && nft_set_is_anonymous(set)) { |
2856 |
++ list_del_rcu(&set->list); |
2857 |
++ if (event) |
2858 |
++ nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, |
2859 |
++ GFP_KERNEL); |
2860 |
++ } |
2861 |
+ } |
2862 |
+ EXPORT_SYMBOL_GPL(nf_tables_unbind_set); |
2863 |
+ |
2864 |
++void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set, |
2865 |
++ struct nft_set_binding *binding, |
2866 |
++ enum nft_trans_phase phase) |
2867 |
++{ |
2868 |
++ switch (phase) { |
2869 |
++ case NFT_TRANS_PREPARE: |
2870 |
++ set->use--; |
2871 |
++ return; |
2872 |
++ case NFT_TRANS_ABORT: |
2873 |
++ case NFT_TRANS_RELEASE: |
2874 |
++ set->use--; |
2875 |
++ /* fall through */ |
2876 |
++ default: |
2877 |
++ nf_tables_unbind_set(ctx, set, binding, |
2878 |
++ phase == NFT_TRANS_COMMIT); |
2879 |
++ } |
2880 |
++} |
2881 |
++EXPORT_SYMBOL_GPL(nf_tables_deactivate_set); |
2882 |
++ |
2883 |
++void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set) |
2884 |
++{ |
2885 |
++ if (list_empty(&set->bindings) && nft_set_is_anonymous(set)) |
2886 |
++ nft_set_destroy(set); |
2887 |
++} |
2888 |
++EXPORT_SYMBOL_GPL(nf_tables_destroy_set); |
2889 |
++ |
2890 |
+ const struct nft_set_ext_type nft_set_ext_types[] = { |
2891 |
+ [NFT_SET_EXT_KEY] = { |
2892 |
+ .align = __alignof__(u32), |
2893 |
+@@ -6429,6 +6483,9 @@ static int nf_tables_commit(struct net *net, struct sk_buff *skb) |
2894 |
+ nf_tables_rule_notify(&trans->ctx, |
2895 |
+ nft_trans_rule(trans), |
2896 |
+ NFT_MSG_DELRULE); |
2897 |
++ nft_rule_expr_deactivate(&trans->ctx, |
2898 |
++ nft_trans_rule(trans), |
2899 |
++ NFT_TRANS_COMMIT); |
2900 |
+ break; |
2901 |
+ case NFT_MSG_NEWSET: |
2902 |
+ nft_clear(net, nft_trans_set(trans)); |
2903 |
+@@ -6577,7 +6634,9 @@ static int __nf_tables_abort(struct net *net) |
2904 |
+ case NFT_MSG_NEWRULE: |
2905 |
+ trans->ctx.chain->use--; |
2906 |
+ list_del_rcu(&nft_trans_rule(trans)->list); |
2907 |
+- nft_rule_expr_deactivate(&trans->ctx, nft_trans_rule(trans)); |
2908 |
++ nft_rule_expr_deactivate(&trans->ctx, |
2909 |
++ nft_trans_rule(trans), |
2910 |
++ NFT_TRANS_ABORT); |
2911 |
+ break; |
2912 |
+ case NFT_MSG_DELRULE: |
2913 |
+ trans->ctx.chain->use++; |
2914 |
+@@ -6587,6 +6646,10 @@ static int __nf_tables_abort(struct net *net) |
2915 |
+ break; |
2916 |
+ case NFT_MSG_NEWSET: |
2917 |
+ trans->ctx.table->use--; |
2918 |
++ if (nft_trans_set(trans)->bound) { |
2919 |
++ nft_trans_destroy(trans); |
2920 |
++ break; |
2921 |
++ } |
2922 |
+ list_del_rcu(&nft_trans_set(trans)->list); |
2923 |
+ break; |
2924 |
+ case NFT_MSG_DELSET: |
2925 |
+@@ -6595,8 +6658,11 @@ static int __nf_tables_abort(struct net *net) |
2926 |
+ nft_trans_destroy(trans); |
2927 |
+ break; |
2928 |
+ case NFT_MSG_NEWSETELEM: |
2929 |
++ if (nft_trans_elem_set(trans)->bound) { |
2930 |
++ nft_trans_destroy(trans); |
2931 |
++ break; |
2932 |
++ } |
2933 |
+ te = (struct nft_trans_elem *)trans->data; |
2934 |
+- |
2935 |
+ te->set->ops->remove(net, te->set, &te->elem); |
2936 |
+ atomic_dec(&te->set->nelems); |
2937 |
+ break; |
2938 |
+diff --git a/net/netfilter/nft_compat.c b/net/netfilter/nft_compat.c |
2939 |
+index 38da1f5436b4..1245e02239d9 100644 |
2940 |
+--- a/net/netfilter/nft_compat.c |
2941 |
++++ b/net/netfilter/nft_compat.c |
2942 |
+@@ -23,19 +23,6 @@ |
2943 |
+ #include <linux/netfilter_arp/arp_tables.h> |
2944 |
+ #include <net/netfilter/nf_tables.h> |
2945 |
+ |
2946 |
+-struct nft_xt { |
2947 |
+- struct list_head head; |
2948 |
+- struct nft_expr_ops ops; |
2949 |
+- unsigned int refcnt; |
2950 |
+- |
2951 |
+- /* Unlike other expressions, ops doesn't have static storage duration. |
2952 |
+- * nft core assumes they do. We use kfree_rcu so that nft core can |
2953 |
+- * can check expr->ops->size even after nft_compat->destroy() frees |
2954 |
+- * the nft_xt struct that holds the ops structure. |
2955 |
+- */ |
2956 |
+- struct rcu_head rcu_head; |
2957 |
+-}; |
2958 |
+- |
2959 |
+ /* Used for matches where *info is larger than X byte */ |
2960 |
+ #define NFT_MATCH_LARGE_THRESH 192 |
2961 |
+ |
2962 |
+@@ -43,17 +30,6 @@ struct nft_xt_match_priv { |
2963 |
+ void *info; |
2964 |
+ }; |
2965 |
+ |
2966 |
+-static bool nft_xt_put(struct nft_xt *xt) |
2967 |
+-{ |
2968 |
+- if (--xt->refcnt == 0) { |
2969 |
+- list_del(&xt->head); |
2970 |
+- kfree_rcu(xt, rcu_head); |
2971 |
+- return true; |
2972 |
+- } |
2973 |
+- |
2974 |
+- return false; |
2975 |
+-} |
2976 |
+- |
2977 |
+ static int nft_compat_chain_validate_dependency(const struct nft_ctx *ctx, |
2978 |
+ const char *tablename) |
2979 |
+ { |
2980 |
+@@ -248,7 +224,6 @@ nft_target_init(const struct nft_ctx *ctx, const struct nft_expr *expr, |
2981 |
+ struct xt_target *target = expr->ops->data; |
2982 |
+ struct xt_tgchk_param par; |
2983 |
+ size_t size = XT_ALIGN(nla_len(tb[NFTA_TARGET_INFO])); |
2984 |
+- struct nft_xt *nft_xt; |
2985 |
+ u16 proto = 0; |
2986 |
+ bool inv = false; |
2987 |
+ union nft_entry e = {}; |
2988 |
+@@ -272,8 +247,6 @@ nft_target_init(const struct nft_ctx *ctx, const struct nft_expr *expr, |
2989 |
+ if (!target->target) |
2990 |
+ return -EINVAL; |
2991 |
+ |
2992 |
+- nft_xt = container_of(expr->ops, struct nft_xt, ops); |
2993 |
+- nft_xt->refcnt++; |
2994 |
+ return 0; |
2995 |
+ } |
2996 |
+ |
2997 |
+@@ -292,8 +265,8 @@ nft_target_destroy(const struct nft_ctx *ctx, const struct nft_expr *expr) |
2998 |
+ if (par.target->destroy != NULL) |
2999 |
+ par.target->destroy(&par); |
3000 |
+ |
3001 |
+- if (nft_xt_put(container_of(expr->ops, struct nft_xt, ops))) |
3002 |
+- module_put(me); |
3003 |
++ module_put(me); |
3004 |
++ kfree(expr->ops); |
3005 |
+ } |
3006 |
+ |
3007 |
+ static int nft_target_dump(struct sk_buff *skb, const struct nft_expr *expr) |
3008 |
+@@ -447,7 +420,6 @@ __nft_match_init(const struct nft_ctx *ctx, const struct nft_expr *expr, |
3009 |
+ struct xt_match *match = expr->ops->data; |
3010 |
+ struct xt_mtchk_param par; |
3011 |
+ size_t size = XT_ALIGN(nla_len(tb[NFTA_MATCH_INFO])); |
3012 |
+- struct nft_xt *nft_xt; |
3013 |
+ u16 proto = 0; |
3014 |
+ bool inv = false; |
3015 |
+ union nft_entry e = {}; |
3016 |
+@@ -463,13 +435,7 @@ __nft_match_init(const struct nft_ctx *ctx, const struct nft_expr *expr, |
3017 |
+ |
3018 |
+ nft_match_set_mtchk_param(&par, ctx, match, info, &e, proto, inv); |
3019 |
+ |
3020 |
+- ret = xt_check_match(&par, size, proto, inv); |
3021 |
+- if (ret < 0) |
3022 |
+- return ret; |
3023 |
+- |
3024 |
+- nft_xt = container_of(expr->ops, struct nft_xt, ops); |
3025 |
+- nft_xt->refcnt++; |
3026 |
+- return 0; |
3027 |
++ return xt_check_match(&par, size, proto, inv); |
3028 |
+ } |
3029 |
+ |
3030 |
+ static int |
3031 |
+@@ -512,8 +478,8 @@ __nft_match_destroy(const struct nft_ctx *ctx, const struct nft_expr *expr, |
3032 |
+ if (par.match->destroy != NULL) |
3033 |
+ par.match->destroy(&par); |
3034 |
+ |
3035 |
+- if (nft_xt_put(container_of(expr->ops, struct nft_xt, ops))) |
3036 |
+- module_put(me); |
3037 |
++ module_put(me); |
3038 |
++ kfree(expr->ops); |
3039 |
+ } |
3040 |
+ |
3041 |
+ static void |
3042 |
+@@ -715,22 +681,13 @@ static const struct nfnetlink_subsystem nfnl_compat_subsys = { |
3043 |
+ .cb = nfnl_nft_compat_cb, |
3044 |
+ }; |
3045 |
+ |
3046 |
+-static LIST_HEAD(nft_match_list); |
3047 |
+- |
3048 |
+ static struct nft_expr_type nft_match_type; |
3049 |
+ |
3050 |
+-static bool nft_match_cmp(const struct xt_match *match, |
3051 |
+- const char *name, u32 rev, u32 family) |
3052 |
+-{ |
3053 |
+- return strcmp(match->name, name) == 0 && match->revision == rev && |
3054 |
+- (match->family == NFPROTO_UNSPEC || match->family == family); |
3055 |
+-} |
3056 |
+- |
3057 |
+ static const struct nft_expr_ops * |
3058 |
+ nft_match_select_ops(const struct nft_ctx *ctx, |
3059 |
+ const struct nlattr * const tb[]) |
3060 |
+ { |
3061 |
+- struct nft_xt *nft_match; |
3062 |
++ struct nft_expr_ops *ops; |
3063 |
+ struct xt_match *match; |
3064 |
+ unsigned int matchsize; |
3065 |
+ char *mt_name; |
3066 |
+@@ -746,14 +703,6 @@ nft_match_select_ops(const struct nft_ctx *ctx, |
3067 |
+ rev = ntohl(nla_get_be32(tb[NFTA_MATCH_REV])); |
3068 |
+ family = ctx->family; |
3069 |
+ |
3070 |
+- /* Re-use the existing match if it's already loaded. */ |
3071 |
+- list_for_each_entry(nft_match, &nft_match_list, head) { |
3072 |
+- struct xt_match *match = nft_match->ops.data; |
3073 |
+- |
3074 |
+- if (nft_match_cmp(match, mt_name, rev, family)) |
3075 |
+- return &nft_match->ops; |
3076 |
+- } |
3077 |
+- |
3078 |
+ match = xt_request_find_match(family, mt_name, rev); |
3079 |
+ if (IS_ERR(match)) |
3080 |
+ return ERR_PTR(-ENOENT); |
3081 |
+@@ -763,66 +712,62 @@ nft_match_select_ops(const struct nft_ctx *ctx, |
3082 |
+ goto err; |
3083 |
+ } |
3084 |
+ |
3085 |
+- /* This is the first time we use this match, allocate operations */ |
3086 |
+- nft_match = kzalloc(sizeof(struct nft_xt), GFP_KERNEL); |
3087 |
+- if (nft_match == NULL) { |
3088 |
++ ops = kzalloc(sizeof(struct nft_expr_ops), GFP_KERNEL); |
3089 |
++ if (!ops) { |
3090 |
+ err = -ENOMEM; |
3091 |
+ goto err; |
3092 |
+ } |
3093 |
+ |
3094 |
+- nft_match->refcnt = 0; |
3095 |
+- nft_match->ops.type = &nft_match_type; |
3096 |
+- nft_match->ops.eval = nft_match_eval; |
3097 |
+- nft_match->ops.init = nft_match_init; |
3098 |
+- nft_match->ops.destroy = nft_match_destroy; |
3099 |
+- nft_match->ops.dump = nft_match_dump; |
3100 |
+- nft_match->ops.validate = nft_match_validate; |
3101 |
+- nft_match->ops.data = match; |
3102 |
++ ops->type = &nft_match_type; |
3103 |
++ ops->eval = nft_match_eval; |
3104 |
++ ops->init = nft_match_init; |
3105 |
++ ops->destroy = nft_match_destroy; |
3106 |
++ ops->dump = nft_match_dump; |
3107 |
++ ops->validate = nft_match_validate; |
3108 |
++ ops->data = match; |
3109 |
+ |
3110 |
+ matchsize = NFT_EXPR_SIZE(XT_ALIGN(match->matchsize)); |
3111 |
+ if (matchsize > NFT_MATCH_LARGE_THRESH) { |
3112 |
+ matchsize = NFT_EXPR_SIZE(sizeof(struct nft_xt_match_priv)); |
3113 |
+ |
3114 |
+- nft_match->ops.eval = nft_match_large_eval; |
3115 |
+- nft_match->ops.init = nft_match_large_init; |
3116 |
+- nft_match->ops.destroy = nft_match_large_destroy; |
3117 |
+- nft_match->ops.dump = nft_match_large_dump; |
3118 |
++ ops->eval = nft_match_large_eval; |
3119 |
++ ops->init = nft_match_large_init; |
3120 |
++ ops->destroy = nft_match_large_destroy; |
3121 |
++ ops->dump = nft_match_large_dump; |
3122 |
+ } |
3123 |
+ |
3124 |
+- nft_match->ops.size = matchsize; |
3125 |
+- |
3126 |
+- list_add(&nft_match->head, &nft_match_list); |
3127 |
++ ops->size = matchsize; |
3128 |
+ |
3129 |
+- return &nft_match->ops; |
3130 |
++ return ops; |
3131 |
+ err: |
3132 |
+ module_put(match->me); |
3133 |
+ return ERR_PTR(err); |
3134 |
+ } |
3135 |
+ |
3136 |
++static void nft_match_release_ops(const struct nft_expr_ops *ops) |
3137 |
++{ |
3138 |
++ struct xt_match *match = ops->data; |
3139 |
++ |
3140 |
++ module_put(match->me); |
3141 |
++ kfree(ops); |
3142 |
++} |
3143 |
++ |
3144 |
+ static struct nft_expr_type nft_match_type __read_mostly = { |
3145 |
+ .name = "match", |
3146 |
+ .select_ops = nft_match_select_ops, |
3147 |
++ .release_ops = nft_match_release_ops, |
3148 |
+ .policy = nft_match_policy, |
3149 |
+ .maxattr = NFTA_MATCH_MAX, |
3150 |
+ .owner = THIS_MODULE, |
3151 |
+ }; |
3152 |
+ |
3153 |
+-static LIST_HEAD(nft_target_list); |
3154 |
+- |
3155 |
+ static struct nft_expr_type nft_target_type; |
3156 |
+ |
3157 |
+-static bool nft_target_cmp(const struct xt_target *tg, |
3158 |
+- const char *name, u32 rev, u32 family) |
3159 |
+-{ |
3160 |
+- return strcmp(tg->name, name) == 0 && tg->revision == rev && |
3161 |
+- (tg->family == NFPROTO_UNSPEC || tg->family == family); |
3162 |
+-} |
3163 |
+- |
3164 |
+ static const struct nft_expr_ops * |
3165 |
+ nft_target_select_ops(const struct nft_ctx *ctx, |
3166 |
+ const struct nlattr * const tb[]) |
3167 |
+ { |
3168 |
+- struct nft_xt *nft_target; |
3169 |
++ struct nft_expr_ops *ops; |
3170 |
+ struct xt_target *target; |
3171 |
+ char *tg_name; |
3172 |
+ u32 rev, family; |
3173 |
+@@ -842,17 +787,6 @@ nft_target_select_ops(const struct nft_ctx *ctx, |
3174 |
+ strcmp(tg_name, "standard") == 0) |
3175 |
+ return ERR_PTR(-EINVAL); |
3176 |
+ |
3177 |
+- /* Re-use the existing target if it's already loaded. */ |
3178 |
+- list_for_each_entry(nft_target, &nft_target_list, head) { |
3179 |
+- struct xt_target *target = nft_target->ops.data; |
3180 |
+- |
3181 |
+- if (!target->target) |
3182 |
+- continue; |
3183 |
+- |
3184 |
+- if (nft_target_cmp(target, tg_name, rev, family)) |
3185 |
+- return &nft_target->ops; |
3186 |
+- } |
3187 |
+- |
3188 |
+ target = xt_request_find_target(family, tg_name, rev); |
3189 |
+ if (IS_ERR(target)) |
3190 |
+ return ERR_PTR(-ENOENT); |
3191 |
+@@ -867,38 +801,43 @@ nft_target_select_ops(const struct nft_ctx *ctx, |
3192 |
+ goto err; |
3193 |
+ } |
3194 |
+ |
3195 |
+- /* This is the first time we use this target, allocate operations */ |
3196 |
+- nft_target = kzalloc(sizeof(struct nft_xt), GFP_KERNEL); |
3197 |
+- if (nft_target == NULL) { |
3198 |
++ ops = kzalloc(sizeof(struct nft_expr_ops), GFP_KERNEL); |
3199 |
++ if (!ops) { |
3200 |
+ err = -ENOMEM; |
3201 |
+ goto err; |
3202 |
+ } |
3203 |
+ |
3204 |
+- nft_target->refcnt = 0; |
3205 |
+- nft_target->ops.type = &nft_target_type; |
3206 |
+- nft_target->ops.size = NFT_EXPR_SIZE(XT_ALIGN(target->targetsize)); |
3207 |
+- nft_target->ops.init = nft_target_init; |
3208 |
+- nft_target->ops.destroy = nft_target_destroy; |
3209 |
+- nft_target->ops.dump = nft_target_dump; |
3210 |
+- nft_target->ops.validate = nft_target_validate; |
3211 |
+- nft_target->ops.data = target; |
3212 |
++ ops->type = &nft_target_type; |
3213 |
++ ops->size = NFT_EXPR_SIZE(XT_ALIGN(target->targetsize)); |
3214 |
++ ops->init = nft_target_init; |
3215 |
++ ops->destroy = nft_target_destroy; |
3216 |
++ ops->dump = nft_target_dump; |
3217 |
++ ops->validate = nft_target_validate; |
3218 |
++ ops->data = target; |
3219 |
+ |
3220 |
+ if (family == NFPROTO_BRIDGE) |
3221 |
+- nft_target->ops.eval = nft_target_eval_bridge; |
3222 |
++ ops->eval = nft_target_eval_bridge; |
3223 |
+ else |
3224 |
+- nft_target->ops.eval = nft_target_eval_xt; |
3225 |
+- |
3226 |
+- list_add(&nft_target->head, &nft_target_list); |
3227 |
++ ops->eval = nft_target_eval_xt; |
3228 |
+ |
3229 |
+- return &nft_target->ops; |
3230 |
++ return ops; |
3231 |
+ err: |
3232 |
+ module_put(target->me); |
3233 |
+ return ERR_PTR(err); |
3234 |
+ } |
3235 |
+ |
3236 |
++static void nft_target_release_ops(const struct nft_expr_ops *ops) |
3237 |
++{ |
3238 |
++ struct xt_target *target = ops->data; |
3239 |
++ |
3240 |
++ module_put(target->me); |
3241 |
++ kfree(ops); |
3242 |
++} |
3243 |
++ |
3244 |
+ static struct nft_expr_type nft_target_type __read_mostly = { |
3245 |
+ .name = "target", |
3246 |
+ .select_ops = nft_target_select_ops, |
3247 |
++ .release_ops = nft_target_release_ops, |
3248 |
+ .policy = nft_target_policy, |
3249 |
+ .maxattr = NFTA_TARGET_MAX, |
3250 |
+ .owner = THIS_MODULE, |
3251 |
+@@ -923,7 +862,6 @@ static int __init nft_compat_module_init(void) |
3252 |
+ } |
3253 |
+ |
3254 |
+ return ret; |
3255 |
+- |
3256 |
+ err_target: |
3257 |
+ nft_unregister_expr(&nft_target_type); |
3258 |
+ err_match: |
3259 |
+@@ -933,32 +871,6 @@ err_match: |
3260 |
+ |
3261 |
+ static void __exit nft_compat_module_exit(void) |
3262 |
+ { |
3263 |
+- struct nft_xt *xt, *next; |
3264 |
+- |
3265 |
+- /* list should be empty here, it can be non-empty only in case there |
3266 |
+- * was an error that caused nft_xt expr to not be initialized fully |
3267 |
+- * and noone else requested the same expression later. |
3268 |
+- * |
3269 |
+- * In this case, the lists contain 0-refcount entries that still |
3270 |
+- * hold module reference. |
3271 |
+- */ |
3272 |
+- list_for_each_entry_safe(xt, next, &nft_target_list, head) { |
3273 |
+- struct xt_target *target = xt->ops.data; |
3274 |
+- |
3275 |
+- if (WARN_ON_ONCE(xt->refcnt)) |
3276 |
+- continue; |
3277 |
+- module_put(target->me); |
3278 |
+- kfree(xt); |
3279 |
+- } |
3280 |
+- |
3281 |
+- list_for_each_entry_safe(xt, next, &nft_match_list, head) { |
3282 |
+- struct xt_match *match = xt->ops.data; |
3283 |
+- |
3284 |
+- if (WARN_ON_ONCE(xt->refcnt)) |
3285 |
+- continue; |
3286 |
+- module_put(match->me); |
3287 |
+- kfree(xt); |
3288 |
+- } |
3289 |
+ nfnetlink_subsys_unregister(&nfnl_compat_subsys); |
3290 |
+ nft_unregister_expr(&nft_target_type); |
3291 |
+ nft_unregister_expr(&nft_match_type); |
3292 |
+diff --git a/net/netfilter/nft_dynset.c b/net/netfilter/nft_dynset.c |
3293 |
+index 6e91a37d57f2..eb7f9a5f2aeb 100644 |
3294 |
+--- a/net/netfilter/nft_dynset.c |
3295 |
++++ b/net/netfilter/nft_dynset.c |
3296 |
+@@ -235,14 +235,32 @@ err1: |
3297 |
+ return err; |
3298 |
+ } |
3299 |
+ |
3300 |
++static void nft_dynset_deactivate(const struct nft_ctx *ctx, |
3301 |
++ const struct nft_expr *expr, |
3302 |
++ enum nft_trans_phase phase) |
3303 |
++{ |
3304 |
++ struct nft_dynset *priv = nft_expr_priv(expr); |
3305 |
++ |
3306 |
++ nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase); |
3307 |
++} |
3308 |
++ |
3309 |
++static void nft_dynset_activate(const struct nft_ctx *ctx, |
3310 |
++ const struct nft_expr *expr) |
3311 |
++{ |
3312 |
++ struct nft_dynset *priv = nft_expr_priv(expr); |
3313 |
++ |
3314 |
++ priv->set->use++; |
3315 |
++} |
3316 |
++ |
3317 |
+ static void nft_dynset_destroy(const struct nft_ctx *ctx, |
3318 |
+ const struct nft_expr *expr) |
3319 |
+ { |
3320 |
+ struct nft_dynset *priv = nft_expr_priv(expr); |
3321 |
+ |
3322 |
+- nf_tables_unbind_set(ctx, priv->set, &priv->binding); |
3323 |
+ if (priv->expr != NULL) |
3324 |
+ nft_expr_destroy(ctx, priv->expr); |
3325 |
++ |
3326 |
++ nf_tables_destroy_set(ctx, priv->set); |
3327 |
+ } |
3328 |
+ |
3329 |
+ static int nft_dynset_dump(struct sk_buff *skb, const struct nft_expr *expr) |
3330 |
+@@ -279,6 +297,8 @@ static const struct nft_expr_ops nft_dynset_ops = { |
3331 |
+ .eval = nft_dynset_eval, |
3332 |
+ .init = nft_dynset_init, |
3333 |
+ .destroy = nft_dynset_destroy, |
3334 |
++ .activate = nft_dynset_activate, |
3335 |
++ .deactivate = nft_dynset_deactivate, |
3336 |
+ .dump = nft_dynset_dump, |
3337 |
+ }; |
3338 |
+ |
3339 |
+diff --git a/net/netfilter/nft_immediate.c b/net/netfilter/nft_immediate.c |
3340 |
+index 0777a93211e2..3f6d1d2a6281 100644 |
3341 |
+--- a/net/netfilter/nft_immediate.c |
3342 |
++++ b/net/netfilter/nft_immediate.c |
3343 |
+@@ -72,10 +72,14 @@ static void nft_immediate_activate(const struct nft_ctx *ctx, |
3344 |
+ } |
3345 |
+ |
3346 |
+ static void nft_immediate_deactivate(const struct nft_ctx *ctx, |
3347 |
+- const struct nft_expr *expr) |
3348 |
++ const struct nft_expr *expr, |
3349 |
++ enum nft_trans_phase phase) |
3350 |
+ { |
3351 |
+ const struct nft_immediate_expr *priv = nft_expr_priv(expr); |
3352 |
+ |
3353 |
++ if (phase == NFT_TRANS_COMMIT) |
3354 |
++ return; |
3355 |
++ |
3356 |
+ return nft_data_release(&priv->data, nft_dreg_to_type(priv->dreg)); |
3357 |
+ } |
3358 |
+ |
3359 |
+diff --git a/net/netfilter/nft_lookup.c b/net/netfilter/nft_lookup.c |
3360 |
+index ad13e8643599..161c3451a747 100644 |
3361 |
+--- a/net/netfilter/nft_lookup.c |
3362 |
++++ b/net/netfilter/nft_lookup.c |
3363 |
+@@ -121,12 +121,29 @@ static int nft_lookup_init(const struct nft_ctx *ctx, |
3364 |
+ return 0; |
3365 |
+ } |
3366 |
+ |
3367 |
++static void nft_lookup_deactivate(const struct nft_ctx *ctx, |
3368 |
++ const struct nft_expr *expr, |
3369 |
++ enum nft_trans_phase phase) |
3370 |
++{ |
3371 |
++ struct nft_lookup *priv = nft_expr_priv(expr); |
3372 |
++ |
3373 |
++ nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase); |
3374 |
++} |
3375 |
++ |
3376 |
++static void nft_lookup_activate(const struct nft_ctx *ctx, |
3377 |
++ const struct nft_expr *expr) |
3378 |
++{ |
3379 |
++ struct nft_lookup *priv = nft_expr_priv(expr); |
3380 |
++ |
3381 |
++ priv->set->use++; |
3382 |
++} |
3383 |
++ |
3384 |
+ static void nft_lookup_destroy(const struct nft_ctx *ctx, |
3385 |
+ const struct nft_expr *expr) |
3386 |
+ { |
3387 |
+ struct nft_lookup *priv = nft_expr_priv(expr); |
3388 |
+ |
3389 |
+- nf_tables_unbind_set(ctx, priv->set, &priv->binding); |
3390 |
++ nf_tables_destroy_set(ctx, priv->set); |
3391 |
+ } |
3392 |
+ |
3393 |
+ static int nft_lookup_dump(struct sk_buff *skb, const struct nft_expr *expr) |
3394 |
+@@ -209,6 +226,8 @@ static const struct nft_expr_ops nft_lookup_ops = { |
3395 |
+ .size = NFT_EXPR_SIZE(sizeof(struct nft_lookup)), |
3396 |
+ .eval = nft_lookup_eval, |
3397 |
+ .init = nft_lookup_init, |
3398 |
++ .activate = nft_lookup_activate, |
3399 |
++ .deactivate = nft_lookup_deactivate, |
3400 |
+ .destroy = nft_lookup_destroy, |
3401 |
+ .dump = nft_lookup_dump, |
3402 |
+ .validate = nft_lookup_validate, |
3403 |
+diff --git a/net/netfilter/nft_objref.c b/net/netfilter/nft_objref.c |
3404 |
+index cdf348f751ec..bf92a40dd1b2 100644 |
3405 |
+--- a/net/netfilter/nft_objref.c |
3406 |
++++ b/net/netfilter/nft_objref.c |
3407 |
+@@ -64,21 +64,34 @@ nla_put_failure: |
3408 |
+ return -1; |
3409 |
+ } |
3410 |
+ |
3411 |
+-static void nft_objref_destroy(const struct nft_ctx *ctx, |
3412 |
+- const struct nft_expr *expr) |
3413 |
++static void nft_objref_deactivate(const struct nft_ctx *ctx, |
3414 |
++ const struct nft_expr *expr, |
3415 |
++ enum nft_trans_phase phase) |
3416 |
+ { |
3417 |
+ struct nft_object *obj = nft_objref_priv(expr); |
3418 |
+ |
3419 |
++ if (phase == NFT_TRANS_COMMIT) |
3420 |
++ return; |
3421 |
++ |
3422 |
+ obj->use--; |
3423 |
+ } |
3424 |
+ |
3425 |
++static void nft_objref_activate(const struct nft_ctx *ctx, |
3426 |
++ const struct nft_expr *expr) |
3427 |
++{ |
3428 |
++ struct nft_object *obj = nft_objref_priv(expr); |
3429 |
++ |
3430 |
++ obj->use++; |
3431 |
++} |
3432 |
++ |
3433 |
+ static struct nft_expr_type nft_objref_type; |
3434 |
+ static const struct nft_expr_ops nft_objref_ops = { |
3435 |
+ .type = &nft_objref_type, |
3436 |
+ .size = NFT_EXPR_SIZE(sizeof(struct nft_object *)), |
3437 |
+ .eval = nft_objref_eval, |
3438 |
+ .init = nft_objref_init, |
3439 |
+- .destroy = nft_objref_destroy, |
3440 |
++ .activate = nft_objref_activate, |
3441 |
++ .deactivate = nft_objref_deactivate, |
3442 |
+ .dump = nft_objref_dump, |
3443 |
+ }; |
3444 |
+ |
3445 |
+@@ -155,12 +168,29 @@ nla_put_failure: |
3446 |
+ return -1; |
3447 |
+ } |
3448 |
+ |
3449 |
++static void nft_objref_map_deactivate(const struct nft_ctx *ctx, |
3450 |
++ const struct nft_expr *expr, |
3451 |
++ enum nft_trans_phase phase) |
3452 |
++{ |
3453 |
++ struct nft_objref_map *priv = nft_expr_priv(expr); |
3454 |
++ |
3455 |
++ nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase); |
3456 |
++} |
3457 |
++ |
3458 |
++static void nft_objref_map_activate(const struct nft_ctx *ctx, |
3459 |
++ const struct nft_expr *expr) |
3460 |
++{ |
3461 |
++ struct nft_objref_map *priv = nft_expr_priv(expr); |
3462 |
++ |
3463 |
++ priv->set->use++; |
3464 |
++} |
3465 |
++ |
3466 |
+ static void nft_objref_map_destroy(const struct nft_ctx *ctx, |
3467 |
+ const struct nft_expr *expr) |
3468 |
+ { |
3469 |
+ struct nft_objref_map *priv = nft_expr_priv(expr); |
3470 |
+ |
3471 |
+- nf_tables_unbind_set(ctx, priv->set, &priv->binding); |
3472 |
++ nf_tables_destroy_set(ctx, priv->set); |
3473 |
+ } |
3474 |
+ |
3475 |
+ static struct nft_expr_type nft_objref_type; |
3476 |
+@@ -169,6 +199,8 @@ static const struct nft_expr_ops nft_objref_map_ops = { |
3477 |
+ .size = NFT_EXPR_SIZE(sizeof(struct nft_objref_map)), |
3478 |
+ .eval = nft_objref_map_eval, |
3479 |
+ .init = nft_objref_map_init, |
3480 |
++ .activate = nft_objref_map_activate, |
3481 |
++ .deactivate = nft_objref_map_deactivate, |
3482 |
+ .destroy = nft_objref_map_destroy, |
3483 |
+ .dump = nft_objref_map_dump, |
3484 |
+ }; |
3485 |
+diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c |
3486 |
+index 1d3144d19903..71ffd1a6dc7c 100644 |
3487 |
+--- a/net/netrom/af_netrom.c |
3488 |
++++ b/net/netrom/af_netrom.c |
3489 |
+@@ -1392,18 +1392,22 @@ static int __init nr_proto_init(void) |
3490 |
+ int i; |
3491 |
+ int rc = proto_register(&nr_proto, 0); |
3492 |
+ |
3493 |
+- if (rc != 0) |
3494 |
+- goto out; |
3495 |
++ if (rc) |
3496 |
++ return rc; |
3497 |
+ |
3498 |
+ if (nr_ndevs > 0x7fffffff/sizeof(struct net_device *)) { |
3499 |
+- printk(KERN_ERR "NET/ROM: nr_proto_init - nr_ndevs parameter to large\n"); |
3500 |
+- return -1; |
3501 |
++ pr_err("NET/ROM: %s - nr_ndevs parameter too large\n", |
3502 |
++ __func__); |
3503 |
++ rc = -EINVAL; |
3504 |
++ goto unregister_proto; |
3505 |
+ } |
3506 |
+ |
3507 |
+ dev_nr = kcalloc(nr_ndevs, sizeof(struct net_device *), GFP_KERNEL); |
3508 |
+- if (dev_nr == NULL) { |
3509 |
+- printk(KERN_ERR "NET/ROM: nr_proto_init - unable to allocate device array\n"); |
3510 |
+- return -1; |
3511 |
++ if (!dev_nr) { |
3512 |
++ pr_err("NET/ROM: %s - unable to allocate device array\n", |
3513 |
++ __func__); |
3514 |
++ rc = -ENOMEM; |
3515 |
++ goto unregister_proto; |
3516 |
+ } |
3517 |
+ |
3518 |
+ for (i = 0; i < nr_ndevs; i++) { |
3519 |
+@@ -1413,13 +1417,13 @@ static int __init nr_proto_init(void) |
3520 |
+ sprintf(name, "nr%d", i); |
3521 |
+ dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, nr_setup); |
3522 |
+ if (!dev) { |
3523 |
+- printk(KERN_ERR "NET/ROM: nr_proto_init - unable to allocate device structure\n"); |
3524 |
++ rc = -ENOMEM; |
3525 |
+ goto fail; |
3526 |
+ } |
3527 |
+ |
3528 |
+ dev->base_addr = i; |
3529 |
+- if (register_netdev(dev)) { |
3530 |
+- printk(KERN_ERR "NET/ROM: nr_proto_init - unable to register network device\n"); |
3531 |
++ rc = register_netdev(dev); |
3532 |
++ if (rc) { |
3533 |
+ free_netdev(dev); |
3534 |
+ goto fail; |
3535 |
+ } |
3536 |
+@@ -1427,36 +1431,64 @@ static int __init nr_proto_init(void) |
3537 |
+ dev_nr[i] = dev; |
3538 |
+ } |
3539 |
+ |
3540 |
+- if (sock_register(&nr_family_ops)) { |
3541 |
+- printk(KERN_ERR "NET/ROM: nr_proto_init - unable to register socket family\n"); |
3542 |
++ rc = sock_register(&nr_family_ops); |
3543 |
++ if (rc) |
3544 |
+ goto fail; |
3545 |
+- } |
3546 |
+ |
3547 |
+- register_netdevice_notifier(&nr_dev_notifier); |
3548 |
++ rc = register_netdevice_notifier(&nr_dev_notifier); |
3549 |
++ if (rc) |
3550 |
++ goto out_sock; |
3551 |
+ |
3552 |
+ ax25_register_pid(&nr_pid); |
3553 |
+ ax25_linkfail_register(&nr_linkfail_notifier); |
3554 |
+ |
3555 |
+ #ifdef CONFIG_SYSCTL |
3556 |
+- nr_register_sysctl(); |
3557 |
++ rc = nr_register_sysctl(); |
3558 |
++ if (rc) |
3559 |
++ goto out_sysctl; |
3560 |
+ #endif |
3561 |
+ |
3562 |
+ nr_loopback_init(); |
3563 |
+ |
3564 |
+- proc_create_seq("nr", 0444, init_net.proc_net, &nr_info_seqops); |
3565 |
+- proc_create_seq("nr_neigh", 0444, init_net.proc_net, &nr_neigh_seqops); |
3566 |
+- proc_create_seq("nr_nodes", 0444, init_net.proc_net, &nr_node_seqops); |
3567 |
+-out: |
3568 |
+- return rc; |
3569 |
++ rc = -ENOMEM; |
3570 |
++ if (!proc_create_seq("nr", 0444, init_net.proc_net, &nr_info_seqops)) |
3571 |
++ goto proc_remove1; |
3572 |
++ if (!proc_create_seq("nr_neigh", 0444, init_net.proc_net, |
3573 |
++ &nr_neigh_seqops)) |
3574 |
++ goto proc_remove2; |
3575 |
++ if (!proc_create_seq("nr_nodes", 0444, init_net.proc_net, |
3576 |
++ &nr_node_seqops)) |
3577 |
++ goto proc_remove3; |
3578 |
++ |
3579 |
++ return 0; |
3580 |
++ |
3581 |
++proc_remove3: |
3582 |
++ remove_proc_entry("nr_neigh", init_net.proc_net); |
3583 |
++proc_remove2: |
3584 |
++ remove_proc_entry("nr", init_net.proc_net); |
3585 |
++proc_remove1: |
3586 |
++ |
3587 |
++ nr_loopback_clear(); |
3588 |
++ nr_rt_free(); |
3589 |
++ |
3590 |
++#ifdef CONFIG_SYSCTL |
3591 |
++ nr_unregister_sysctl(); |
3592 |
++out_sysctl: |
3593 |
++#endif |
3594 |
++ ax25_linkfail_release(&nr_linkfail_notifier); |
3595 |
++ ax25_protocol_release(AX25_P_NETROM); |
3596 |
++ unregister_netdevice_notifier(&nr_dev_notifier); |
3597 |
++out_sock: |
3598 |
++ sock_unregister(PF_NETROM); |
3599 |
+ fail: |
3600 |
+ while (--i >= 0) { |
3601 |
+ unregister_netdev(dev_nr[i]); |
3602 |
+ free_netdev(dev_nr[i]); |
3603 |
+ } |
3604 |
+ kfree(dev_nr); |
3605 |
++unregister_proto: |
3606 |
+ proto_unregister(&nr_proto); |
3607 |
+- rc = -1; |
3608 |
+- goto out; |
3609 |
++ return rc; |
3610 |
+ } |
3611 |
+ |
3612 |
+ module_init(nr_proto_init); |
3613 |
+diff --git a/net/netrom/nr_loopback.c b/net/netrom/nr_loopback.c |
3614 |
+index 215ad22a9647..93d13f019981 100644 |
3615 |
+--- a/net/netrom/nr_loopback.c |
3616 |
++++ b/net/netrom/nr_loopback.c |
3617 |
+@@ -70,7 +70,7 @@ static void nr_loopback_timer(struct timer_list *unused) |
3618 |
+ } |
3619 |
+ } |
3620 |
+ |
3621 |
+-void __exit nr_loopback_clear(void) |
3622 |
++void nr_loopback_clear(void) |
3623 |
+ { |
3624 |
+ del_timer_sync(&loopback_timer); |
3625 |
+ skb_queue_purge(&loopback_queue); |
3626 |
+diff --git a/net/netrom/nr_route.c b/net/netrom/nr_route.c |
3627 |
+index 6485f593e2f0..b76aa668a94b 100644 |
3628 |
+--- a/net/netrom/nr_route.c |
3629 |
++++ b/net/netrom/nr_route.c |
3630 |
+@@ -953,7 +953,7 @@ const struct seq_operations nr_neigh_seqops = { |
3631 |
+ /* |
3632 |
+ * Free all memory associated with the nodes and routes lists. |
3633 |
+ */ |
3634 |
+-void __exit nr_rt_free(void) |
3635 |
++void nr_rt_free(void) |
3636 |
+ { |
3637 |
+ struct nr_neigh *s = NULL; |
3638 |
+ struct nr_node *t = NULL; |
3639 |
+diff --git a/net/netrom/sysctl_net_netrom.c b/net/netrom/sysctl_net_netrom.c |
3640 |
+index ba1c368b3f18..771011b84270 100644 |
3641 |
+--- a/net/netrom/sysctl_net_netrom.c |
3642 |
++++ b/net/netrom/sysctl_net_netrom.c |
3643 |
+@@ -146,9 +146,12 @@ static struct ctl_table nr_table[] = { |
3644 |
+ { } |
3645 |
+ }; |
3646 |
+ |
3647 |
+-void __init nr_register_sysctl(void) |
3648 |
++int __init nr_register_sysctl(void) |
3649 |
+ { |
3650 |
+ nr_table_header = register_net_sysctl(&init_net, "net/netrom", nr_table); |
3651 |
++ if (!nr_table_header) |
3652 |
++ return -ENOMEM; |
3653 |
++ return 0; |
3654 |
+ } |
3655 |
+ |
3656 |
+ void nr_unregister_sysctl(void) |
3657 |
+diff --git a/net/rds/af_rds.c b/net/rds/af_rds.c |
3658 |
+index 65387e1e6964..cd7e01ea8144 100644 |
3659 |
+--- a/net/rds/af_rds.c |
3660 |
++++ b/net/rds/af_rds.c |
3661 |
+@@ -506,6 +506,9 @@ static int rds_connect(struct socket *sock, struct sockaddr *uaddr, |
3662 |
+ struct rds_sock *rs = rds_sk_to_rs(sk); |
3663 |
+ int ret = 0; |
3664 |
+ |
3665 |
++ if (addr_len < offsetofend(struct sockaddr, sa_family)) |
3666 |
++ return -EINVAL; |
3667 |
++ |
3668 |
+ lock_sock(sk); |
3669 |
+ |
3670 |
+ switch (uaddr->sa_family) { |
3671 |
+diff --git a/net/rds/bind.c b/net/rds/bind.c |
3672 |
+index 17c9d9f0c848..0f4398e7f2a7 100644 |
3673 |
+--- a/net/rds/bind.c |
3674 |
++++ b/net/rds/bind.c |
3675 |
+@@ -173,6 +173,8 @@ int rds_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len) |
3676 |
+ /* We allow an RDS socket to be bound to either IPv4 or IPv6 |
3677 |
+ * address. |
3678 |
+ */ |
3679 |
++ if (addr_len < offsetofend(struct sockaddr, sa_family)) |
3680 |
++ return -EINVAL; |
3681 |
+ if (uaddr->sa_family == AF_INET) { |
3682 |
+ struct sockaddr_in *sin = (struct sockaddr_in *)uaddr; |
3683 |
+ |
3684 |
+diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c |
3685 |
+index e0f70c4051b6..01e764f8f224 100644 |
3686 |
+--- a/net/rds/ib_fmr.c |
3687 |
++++ b/net/rds/ib_fmr.c |
3688 |
+@@ -44,6 +44,17 @@ struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev, int npages) |
3689 |
+ else |
3690 |
+ pool = rds_ibdev->mr_1m_pool; |
3691 |
+ |
3692 |
++ if (atomic_read(&pool->dirty_count) >= pool->max_items / 10) |
3693 |
++ queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10); |
3694 |
++ |
3695 |
++ /* Switch pools if one of the pool is reaching upper limit */ |
3696 |
++ if (atomic_read(&pool->dirty_count) >= pool->max_items * 9 / 10) { |
3697 |
++ if (pool->pool_type == RDS_IB_MR_8K_POOL) |
3698 |
++ pool = rds_ibdev->mr_1m_pool; |
3699 |
++ else |
3700 |
++ pool = rds_ibdev->mr_8k_pool; |
3701 |
++ } |
3702 |
++ |
3703 |
+ ibmr = rds_ib_try_reuse_ibmr(pool); |
3704 |
+ if (ibmr) |
3705 |
+ return ibmr; |
3706 |
+diff --git a/net/rds/ib_rdma.c b/net/rds/ib_rdma.c |
3707 |
+index 63c8d107adcf..d664e9ade74d 100644 |
3708 |
+--- a/net/rds/ib_rdma.c |
3709 |
++++ b/net/rds/ib_rdma.c |
3710 |
+@@ -454,9 +454,6 @@ struct rds_ib_mr *rds_ib_try_reuse_ibmr(struct rds_ib_mr_pool *pool) |
3711 |
+ struct rds_ib_mr *ibmr = NULL; |
3712 |
+ int iter = 0; |
3713 |
+ |
3714 |
+- if (atomic_read(&pool->dirty_count) >= pool->max_items_soft / 10) |
3715 |
+- queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10); |
3716 |
+- |
3717 |
+ while (1) { |
3718 |
+ ibmr = rds_ib_reuse_mr(pool); |
3719 |
+ if (ibmr) |
3720 |
+diff --git a/net/rose/rose_loopback.c b/net/rose/rose_loopback.c |
3721 |
+index 7af4f99c4a93..094a6621f8e8 100644 |
3722 |
+--- a/net/rose/rose_loopback.c |
3723 |
++++ b/net/rose/rose_loopback.c |
3724 |
+@@ -16,6 +16,7 @@ |
3725 |
+ #include <linux/init.h> |
3726 |
+ |
3727 |
+ static struct sk_buff_head loopback_queue; |
3728 |
++#define ROSE_LOOPBACK_LIMIT 1000 |
3729 |
+ static struct timer_list loopback_timer; |
3730 |
+ |
3731 |
+ static void rose_set_loopback_timer(void); |
3732 |
+@@ -35,29 +36,27 @@ static int rose_loopback_running(void) |
3733 |
+ |
3734 |
+ int rose_loopback_queue(struct sk_buff *skb, struct rose_neigh *neigh) |
3735 |
+ { |
3736 |
+- struct sk_buff *skbn; |
3737 |
++ struct sk_buff *skbn = NULL; |
3738 |
+ |
3739 |
+- skbn = skb_clone(skb, GFP_ATOMIC); |
3740 |
++ if (skb_queue_len(&loopback_queue) < ROSE_LOOPBACK_LIMIT) |
3741 |
++ skbn = skb_clone(skb, GFP_ATOMIC); |
3742 |
+ |
3743 |
+- kfree_skb(skb); |
3744 |
+- |
3745 |
+- if (skbn != NULL) { |
3746 |
++ if (skbn) { |
3747 |
++ consume_skb(skb); |
3748 |
+ skb_queue_tail(&loopback_queue, skbn); |
3749 |
+ |
3750 |
+ if (!rose_loopback_running()) |
3751 |
+ rose_set_loopback_timer(); |
3752 |
++ } else { |
3753 |
++ kfree_skb(skb); |
3754 |
+ } |
3755 |
+ |
3756 |
+ return 1; |
3757 |
+ } |
3758 |
+ |
3759 |
+- |
3760 |
+ static void rose_set_loopback_timer(void) |
3761 |
+ { |
3762 |
+- del_timer(&loopback_timer); |
3763 |
+- |
3764 |
+- loopback_timer.expires = jiffies + 10; |
3765 |
+- add_timer(&loopback_timer); |
3766 |
++ mod_timer(&loopback_timer, jiffies + 10); |
3767 |
+ } |
3768 |
+ |
3769 |
+ static void rose_loopback_timer(struct timer_list *unused) |
3770 |
+@@ -68,8 +67,12 @@ static void rose_loopback_timer(struct timer_list *unused) |
3771 |
+ struct sock *sk; |
3772 |
+ unsigned short frametype; |
3773 |
+ unsigned int lci_i, lci_o; |
3774 |
++ int count; |
3775 |
+ |
3776 |
+- while ((skb = skb_dequeue(&loopback_queue)) != NULL) { |
3777 |
++ for (count = 0; count < ROSE_LOOPBACK_LIMIT; count++) { |
3778 |
++ skb = skb_dequeue(&loopback_queue); |
3779 |
++ if (!skb) |
3780 |
++ return; |
3781 |
+ if (skb->len < ROSE_MIN_LEN) { |
3782 |
+ kfree_skb(skb); |
3783 |
+ continue; |
3784 |
+@@ -106,6 +109,8 @@ static void rose_loopback_timer(struct timer_list *unused) |
3785 |
+ kfree_skb(skb); |
3786 |
+ } |
3787 |
+ } |
3788 |
++ if (!skb_queue_empty(&loopback_queue)) |
3789 |
++ mod_timer(&loopback_timer, jiffies + 1); |
3790 |
+ } |
3791 |
+ |
3792 |
+ void __exit rose_loopback_clear(void) |
3793 |
+diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c |
3794 |
+index 570b49d2da42..d591f54cb91f 100644 |
3795 |
+--- a/net/rxrpc/input.c |
3796 |
++++ b/net/rxrpc/input.c |
3797 |
+@@ -1155,19 +1155,19 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb) |
3798 |
+ * handle data received on the local endpoint |
3799 |
+ * - may be called in interrupt context |
3800 |
+ * |
3801 |
+- * The socket is locked by the caller and this prevents the socket from being |
3802 |
+- * shut down and the local endpoint from going away, thus sk_user_data will not |
3803 |
+- * be cleared until this function returns. |
3804 |
++ * [!] Note that as this is called from the encap_rcv hook, the socket is not |
3805 |
++ * held locked by the caller and nothing prevents sk_user_data on the UDP from |
3806 |
++ * being cleared in the middle of processing this function. |
3807 |
+ * |
3808 |
+ * Called with the RCU read lock held from the IP layer via UDP. |
3809 |
+ */ |
3810 |
+ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) |
3811 |
+ { |
3812 |
++ struct rxrpc_local *local = rcu_dereference_sk_user_data(udp_sk); |
3813 |
+ struct rxrpc_connection *conn; |
3814 |
+ struct rxrpc_channel *chan; |
3815 |
+ struct rxrpc_call *call = NULL; |
3816 |
+ struct rxrpc_skb_priv *sp; |
3817 |
+- struct rxrpc_local *local = udp_sk->sk_user_data; |
3818 |
+ struct rxrpc_peer *peer = NULL; |
3819 |
+ struct rxrpc_sock *rx = NULL; |
3820 |
+ unsigned int channel; |
3821 |
+@@ -1175,6 +1175,10 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb) |
3822 |
+ |
3823 |
+ _enter("%p", udp_sk); |
3824 |
+ |
3825 |
++ if (unlikely(!local)) { |
3826 |
++ kfree_skb(skb); |
3827 |
++ return 0; |
3828 |
++ } |
3829 |
+ if (skb->tstamp == 0) |
3830 |
+ skb->tstamp = ktime_get_real(); |
3831 |
+ |
3832 |
+diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c |
3833 |
+index 0906e51d3cfb..10317dbdab5f 100644 |
3834 |
+--- a/net/rxrpc/local_object.c |
3835 |
++++ b/net/rxrpc/local_object.c |
3836 |
+@@ -304,7 +304,8 @@ nomem: |
3837 |
+ ret = -ENOMEM; |
3838 |
+ sock_error: |
3839 |
+ mutex_unlock(&rxnet->local_mutex); |
3840 |
+- kfree(local); |
3841 |
++ if (local) |
3842 |
++ call_rcu(&local->rcu, rxrpc_local_rcu); |
3843 |
+ _leave(" = %d", ret); |
3844 |
+ return ERR_PTR(ret); |
3845 |
+ |
3846 |
+diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c |
3847 |
+index b6e8eccf2a52..214440c5b14e 100644 |
3848 |
+--- a/net/sunrpc/cache.c |
3849 |
++++ b/net/sunrpc/cache.c |
3850 |
+@@ -54,6 +54,7 @@ static void cache_init(struct cache_head *h, struct cache_detail *detail) |
3851 |
+ h->last_refresh = now; |
3852 |
+ } |
3853 |
+ |
3854 |
++static inline int cache_is_valid(struct cache_head *h); |
3855 |
+ static void cache_fresh_locked(struct cache_head *head, time_t expiry, |
3856 |
+ struct cache_detail *detail); |
3857 |
+ static void cache_fresh_unlocked(struct cache_head *head, |
3858 |
+@@ -100,6 +101,8 @@ struct cache_head *sunrpc_cache_lookup(struct cache_detail *detail, |
3859 |
+ if (cache_is_expired(detail, tmp)) { |
3860 |
+ hlist_del_init(&tmp->cache_list); |
3861 |
+ detail->entries --; |
3862 |
++ if (cache_is_valid(tmp) == -EAGAIN) |
3863 |
++ set_bit(CACHE_NEGATIVE, &tmp->flags); |
3864 |
+ cache_fresh_locked(tmp, 0, detail); |
3865 |
+ freeme = tmp; |
3866 |
+ break; |
3867 |
+diff --git a/net/tipc/netlink_compat.c b/net/tipc/netlink_compat.c |
3868 |
+index 0b21187d74df..588d5aa14c41 100644 |
3869 |
+--- a/net/tipc/netlink_compat.c |
3870 |
++++ b/net/tipc/netlink_compat.c |
3871 |
+@@ -267,8 +267,14 @@ static int tipc_nl_compat_dumpit(struct tipc_nl_compat_cmd_dump *cmd, |
3872 |
+ if (msg->rep_type) |
3873 |
+ tipc_tlv_init(msg->rep, msg->rep_type); |
3874 |
+ |
3875 |
+- if (cmd->header) |
3876 |
+- (*cmd->header)(msg); |
3877 |
++ if (cmd->header) { |
3878 |
++ err = (*cmd->header)(msg); |
3879 |
++ if (err) { |
3880 |
++ kfree_skb(msg->rep); |
3881 |
++ msg->rep = NULL; |
3882 |
++ return err; |
3883 |
++ } |
3884 |
++ } |
3885 |
+ |
3886 |
+ arg = nlmsg_new(0, GFP_KERNEL); |
3887 |
+ if (!arg) { |
3888 |
+@@ -397,7 +403,12 @@ static int tipc_nl_compat_bearer_enable(struct tipc_nl_compat_cmd_doit *cmd, |
3889 |
+ if (!bearer) |
3890 |
+ return -EMSGSIZE; |
3891 |
+ |
3892 |
+- len = min_t(int, TLV_GET_DATA_LEN(msg->req), TIPC_MAX_BEARER_NAME); |
3893 |
++ len = TLV_GET_DATA_LEN(msg->req); |
3894 |
++ len -= offsetof(struct tipc_bearer_config, name); |
3895 |
++ if (len <= 0) |
3896 |
++ return -EINVAL; |
3897 |
++ |
3898 |
++ len = min_t(int, len, TIPC_MAX_BEARER_NAME); |
3899 |
+ if (!string_is_valid(b->name, len)) |
3900 |
+ return -EINVAL; |
3901 |
+ |
3902 |
+@@ -766,7 +777,12 @@ static int tipc_nl_compat_link_set(struct tipc_nl_compat_cmd_doit *cmd, |
3903 |
+ |
3904 |
+ lc = (struct tipc_link_config *)TLV_DATA(msg->req); |
3905 |
+ |
3906 |
+- len = min_t(int, TLV_GET_DATA_LEN(msg->req), TIPC_MAX_LINK_NAME); |
3907 |
++ len = TLV_GET_DATA_LEN(msg->req); |
3908 |
++ len -= offsetof(struct tipc_link_config, name); |
3909 |
++ if (len <= 0) |
3910 |
++ return -EINVAL; |
3911 |
++ |
3912 |
++ len = min_t(int, len, TIPC_MAX_LINK_NAME); |
3913 |
+ if (!string_is_valid(lc->name, len)) |
3914 |
+ return -EINVAL; |
3915 |
+ |
3916 |
+diff --git a/net/tls/tls_device.c b/net/tls/tls_device.c |
3917 |
+index 961b07d4d41c..c9588b682db4 100644 |
3918 |
+--- a/net/tls/tls_device.c |
3919 |
++++ b/net/tls/tls_device.c |
3920 |
+@@ -874,7 +874,9 @@ int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx) |
3921 |
+ goto release_netdev; |
3922 |
+ |
3923 |
+ free_sw_resources: |
3924 |
++ up_read(&device_offload_lock); |
3925 |
+ tls_sw_free_resources_rx(sk); |
3926 |
++ down_read(&device_offload_lock); |
3927 |
+ release_ctx: |
3928 |
+ ctx->priv_ctx_rx = NULL; |
3929 |
+ release_netdev: |
3930 |
+@@ -909,8 +911,6 @@ void tls_device_offload_cleanup_rx(struct sock *sk) |
3931 |
+ } |
3932 |
+ out: |
3933 |
+ up_read(&device_offload_lock); |
3934 |
+- kfree(tls_ctx->rx.rec_seq); |
3935 |
+- kfree(tls_ctx->rx.iv); |
3936 |
+ tls_sw_release_resources_rx(sk); |
3937 |
+ } |
3938 |
+ |
3939 |
+diff --git a/net/tls/tls_device_fallback.c b/net/tls/tls_device_fallback.c |
3940 |
+index 450a6dbc5a88..ef8934fd8698 100644 |
3941 |
+--- a/net/tls/tls_device_fallback.c |
3942 |
++++ b/net/tls/tls_device_fallback.c |
3943 |
+@@ -193,6 +193,9 @@ static void update_chksum(struct sk_buff *skb, int headln) |
3944 |
+ |
3945 |
+ static void complete_skb(struct sk_buff *nskb, struct sk_buff *skb, int headln) |
3946 |
+ { |
3947 |
++ struct sock *sk = skb->sk; |
3948 |
++ int delta; |
3949 |
++ |
3950 |
+ skb_copy_header(nskb, skb); |
3951 |
+ |
3952 |
+ skb_put(nskb, skb->len); |
3953 |
+@@ -200,11 +203,15 @@ static void complete_skb(struct sk_buff *nskb, struct sk_buff *skb, int headln) |
3954 |
+ update_chksum(nskb, headln); |
3955 |
+ |
3956 |
+ nskb->destructor = skb->destructor; |
3957 |
+- nskb->sk = skb->sk; |
3958 |
++ nskb->sk = sk; |
3959 |
+ skb->destructor = NULL; |
3960 |
+ skb->sk = NULL; |
3961 |
+- refcount_add(nskb->truesize - skb->truesize, |
3962 |
+- &nskb->sk->sk_wmem_alloc); |
3963 |
++ |
3964 |
++ delta = nskb->truesize - skb->truesize; |
3965 |
++ if (likely(delta < 0)) |
3966 |
++ WARN_ON_ONCE(refcount_sub_and_test(-delta, &sk->sk_wmem_alloc)); |
3967 |
++ else if (delta) |
3968 |
++ refcount_add(delta, &sk->sk_wmem_alloc); |
3969 |
+ } |
3970 |
+ |
3971 |
+ /* This function may be called after the user socket is already |
3972 |
+diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c |
3973 |
+index a091c03abcb2..25b3fb585777 100644 |
3974 |
+--- a/net/tls/tls_main.c |
3975 |
++++ b/net/tls/tls_main.c |
3976 |
+@@ -290,11 +290,8 @@ static void tls_sk_proto_close(struct sock *sk, long timeout) |
3977 |
+ tls_sw_free_resources_tx(sk); |
3978 |
+ } |
3979 |
+ |
3980 |
+- if (ctx->rx_conf == TLS_SW) { |
3981 |
+- kfree(ctx->rx.rec_seq); |
3982 |
+- kfree(ctx->rx.iv); |
3983 |
++ if (ctx->rx_conf == TLS_SW) |
3984 |
+ tls_sw_free_resources_rx(sk); |
3985 |
+- } |
3986 |
+ |
3987 |
+ #ifdef CONFIG_TLS_DEVICE |
3988 |
+ if (ctx->rx_conf == TLS_HW) |
3989 |
+diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c |
3990 |
+index b9c6ecfbcfea..6848a8196711 100644 |
3991 |
+--- a/net/tls/tls_sw.c |
3992 |
++++ b/net/tls/tls_sw.c |
3993 |
+@@ -1118,6 +1118,9 @@ void tls_sw_release_resources_rx(struct sock *sk) |
3994 |
+ struct tls_context *tls_ctx = tls_get_ctx(sk); |
3995 |
+ struct tls_sw_context_rx *ctx = tls_sw_ctx_rx(tls_ctx); |
3996 |
+ |
3997 |
++ kfree(tls_ctx->rx.rec_seq); |
3998 |
++ kfree(tls_ctx->rx.iv); |
3999 |
++ |
4000 |
+ if (ctx->aead_recv) { |
4001 |
+ kfree_skb(ctx->recv_pkt); |
4002 |
+ ctx->recv_pkt = NULL; |
4003 |
+diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c |
4004 |
+index 3ae3a33da70b..602715fc9a75 100644 |
4005 |
+--- a/net/vmw_vsock/virtio_transport_common.c |
4006 |
++++ b/net/vmw_vsock/virtio_transport_common.c |
4007 |
+@@ -662,6 +662,8 @@ static int virtio_transport_reset(struct vsock_sock *vsk, |
4008 |
+ */ |
4009 |
+ static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt) |
4010 |
+ { |
4011 |
++ const struct virtio_transport *t; |
4012 |
++ struct virtio_vsock_pkt *reply; |
4013 |
+ struct virtio_vsock_pkt_info info = { |
4014 |
+ .op = VIRTIO_VSOCK_OP_RST, |
4015 |
+ .type = le16_to_cpu(pkt->hdr.type), |
4016 |
+@@ -672,15 +674,21 @@ static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt) |
4017 |
+ if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST) |
4018 |
+ return 0; |
4019 |
+ |
4020 |
+- pkt = virtio_transport_alloc_pkt(&info, 0, |
4021 |
+- le64_to_cpu(pkt->hdr.dst_cid), |
4022 |
+- le32_to_cpu(pkt->hdr.dst_port), |
4023 |
+- le64_to_cpu(pkt->hdr.src_cid), |
4024 |
+- le32_to_cpu(pkt->hdr.src_port)); |
4025 |
+- if (!pkt) |
4026 |
++ reply = virtio_transport_alloc_pkt(&info, 0, |
4027 |
++ le64_to_cpu(pkt->hdr.dst_cid), |
4028 |
++ le32_to_cpu(pkt->hdr.dst_port), |
4029 |
++ le64_to_cpu(pkt->hdr.src_cid), |
4030 |
++ le32_to_cpu(pkt->hdr.src_port)); |
4031 |
++ if (!reply) |
4032 |
+ return -ENOMEM; |
4033 |
+ |
4034 |
+- return virtio_transport_get_ops()->send_pkt(pkt); |
4035 |
++ t = virtio_transport_get_ops(); |
4036 |
++ if (!t) { |
4037 |
++ virtio_transport_free_pkt(reply); |
4038 |
++ return -ENOTCONN; |
4039 |
++ } |
4040 |
++ |
4041 |
++ return t->send_pkt(reply); |
4042 |
+ } |
4043 |
+ |
4044 |
+ static void virtio_transport_wait_close(struct sock *sk, long timeout) |
4045 |
+diff --git a/sound/pci/hda/patch_ca0132.c b/sound/pci/hda/patch_ca0132.c |
4046 |
+index 80f73810b21b..0436789e7cd8 100644 |
4047 |
+--- a/sound/pci/hda/patch_ca0132.c |
4048 |
++++ b/sound/pci/hda/patch_ca0132.c |
4049 |
+@@ -7394,8 +7394,10 @@ static void ca0132_free(struct hda_codec *codec) |
4050 |
+ ca0132_exit_chip(codec); |
4051 |
+ |
4052 |
+ snd_hda_power_down(codec); |
4053 |
+- if (IS_ENABLED(CONFIG_PCI) && spec->mem_base) |
4054 |
++#ifdef CONFIG_PCI |
4055 |
++ if (spec->mem_base) |
4056 |
+ pci_iounmap(codec->bus->pci, spec->mem_base); |
4057 |
++#endif |
4058 |
+ kfree(spec->spec_init_verbs); |
4059 |
+ kfree(codec->spec); |
4060 |
+ } |