Magellan Linux

Contents of /trunk/kernel-alx/patches-4.19/0137-4.19.38-all-fixes.patch

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3416 - (show annotations) (download)
Fri Aug 2 11:47:46 2019 UTC (4 years, 9 months ago) by niro
File size: 127326 byte(s)
-linux-4.19.38
1 diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
2 index f5acf35c712f..8b6567f7cb9b 100644
3 --- a/Documentation/admin-guide/kernel-parameters.txt
4 +++ b/Documentation/admin-guide/kernel-parameters.txt
5 @@ -2805,7 +2805,7 @@
6 check bypass). With this option data leaks are possible
7 in the system.
8
9 - nospectre_v2 [X86] Disable all mitigations for the Spectre variant 2
10 + nospectre_v2 [X86,PPC_FSL_BOOK3E] Disable all mitigations for the Spectre variant 2
11 (indirect branch prediction) vulnerability. System may
12 allow data leaks with this option, which is equivalent
13 to spectre_v2=off.
14 diff --git a/Documentation/networking/ip-sysctl.txt b/Documentation/networking/ip-sysctl.txt
15 index 960de8fe3f40..2c31208528d5 100644
16 --- a/Documentation/networking/ip-sysctl.txt
17 +++ b/Documentation/networking/ip-sysctl.txt
18 @@ -410,6 +410,7 @@ tcp_min_rtt_wlen - INTEGER
19 minimum RTT when it is moved to a longer path (e.g., due to traffic
20 engineering). A longer window makes the filter more resistant to RTT
21 inflations such as transient congestion. The unit is seconds.
22 + Possible values: 0 - 86400 (1 day)
23 Default: 300
24
25 tcp_moderate_rcvbuf - BOOLEAN
26 diff --git a/Makefile b/Makefile
27 index 7b495cad8c2e..14d4aeb48907 100644
28 --- a/Makefile
29 +++ b/Makefile
30 @@ -1,7 +1,7 @@
31 # SPDX-License-Identifier: GPL-2.0
32 VERSION = 4
33 PATCHLEVEL = 19
34 -SUBLEVEL = 37
35 +SUBLEVEL = 38
36 EXTRAVERSION =
37 NAME = "People's Front"
38
39 diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S
40 index 517e0e18f0b8..e205bbbe2794 100644
41 --- a/arch/arm/boot/compressed/head.S
42 +++ b/arch/arm/boot/compressed/head.S
43 @@ -1395,7 +1395,21 @@ ENTRY(efi_stub_entry)
44
45 @ Preserve return value of efi_entry() in r4
46 mov r4, r0
47 - bl cache_clean_flush
48 +
49 + @ our cache maintenance code relies on CP15 barrier instructions
50 + @ but since we arrived here with the MMU and caches configured
51 + @ by UEFI, we must check that the CP15BEN bit is set in SCTLR.
52 + @ Note that this bit is RAO/WI on v6 and earlier, so the ISB in
53 + @ the enable path will be executed on v7+ only.
54 + mrc p15, 0, r1, c1, c0, 0 @ read SCTLR
55 + tst r1, #(1 << 5) @ CP15BEN bit set?
56 + bne 0f
57 + orr r1, r1, #(1 << 5) @ CP15 barrier instructions
58 + mcr p15, 0, r1, c1, c0, 0 @ write SCTLR
59 + ARM( .inst 0xf57ff06f @ v7+ isb )
60 + THUMB( isb )
61 +
62 +0: bl cache_clean_flush
63 bl cache_off
64
65 @ Set parameters for booting zImage according to boot protocol
66 diff --git a/arch/mips/kernel/scall64-o32.S b/arch/mips/kernel/scall64-o32.S
67 index 73913f072e39..579608342ac6 100644
68 --- a/arch/mips/kernel/scall64-o32.S
69 +++ b/arch/mips/kernel/scall64-o32.S
70 @@ -125,7 +125,7 @@ trace_a_syscall:
71 subu t1, v0, __NR_O32_Linux
72 move a1, v0
73 bnez t1, 1f /* __NR_syscall at offset 0 */
74 - lw a1, PT_R4(sp) /* Arg1 for __NR_syscall case */
75 + ld a1, PT_R4(sp) /* Arg1 for __NR_syscall case */
76 .set pop
77
78 1: jal syscall_trace_enter
79 diff --git a/arch/powerpc/configs/skiroot_defconfig b/arch/powerpc/configs/skiroot_defconfig
80 index 6bd5e7261335..ffeaed63675b 100644
81 --- a/arch/powerpc/configs/skiroot_defconfig
82 +++ b/arch/powerpc/configs/skiroot_defconfig
83 @@ -195,6 +195,7 @@ CONFIG_UDF_FS=m
84 CONFIG_MSDOS_FS=m
85 CONFIG_VFAT_FS=m
86 CONFIG_PROC_KCORE=y
87 +CONFIG_HUGETLBFS=y
88 CONFIG_TMPFS=y
89 CONFIG_TMPFS_POSIX_ACL=y
90 # CONFIG_MISC_FILESYSTEMS is not set
91 diff --git a/arch/powerpc/kernel/vdso32/gettimeofday.S b/arch/powerpc/kernel/vdso32/gettimeofday.S
92 index 769c2624e0a6..75cff3f336b3 100644
93 --- a/arch/powerpc/kernel/vdso32/gettimeofday.S
94 +++ b/arch/powerpc/kernel/vdso32/gettimeofday.S
95 @@ -98,7 +98,7 @@ V_FUNCTION_BEGIN(__kernel_clock_gettime)
96 * can be used, r7 contains NSEC_PER_SEC.
97 */
98
99 - lwz r5,WTOM_CLOCK_SEC(r9)
100 + lwz r5,(WTOM_CLOCK_SEC+LOPART)(r9)
101 lwz r6,WTOM_CLOCK_NSEC(r9)
102
103 /* We now have our offset in r5,r6. We create a fake dependency
104 diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
105 index 6c6a7c72cae4..ad0216c41d2c 100644
106 --- a/arch/powerpc/platforms/Kconfig.cputype
107 +++ b/arch/powerpc/platforms/Kconfig.cputype
108 @@ -330,7 +330,7 @@ config ARCH_ENABLE_SPLIT_PMD_PTLOCK
109
110 config PPC_RADIX_MMU
111 bool "Radix MMU Support"
112 - depends on PPC_BOOK3S_64
113 + depends on PPC_BOOK3S_64 && HUGETLB_PAGE
114 select ARCH_HAS_GIGANTIC_PAGE if (MEMORY_ISOLATION && COMPACTION) || CMA
115 default y
116 help
117 diff --git a/arch/x86/Makefile b/arch/x86/Makefile
118 index b84f61bc5e7a..ffc823a8312f 100644
119 --- a/arch/x86/Makefile
120 +++ b/arch/x86/Makefile
121 @@ -224,6 +224,15 @@ KBUILD_CFLAGS += -fno-asynchronous-unwind-tables
122 # Avoid indirect branches in kernel to deal with Spectre
123 ifdef CONFIG_RETPOLINE
124 KBUILD_CFLAGS += $(RETPOLINE_CFLAGS)
125 + # Additionally, avoid generating expensive indirect jumps which
126 + # are subject to retpolines for small number of switch cases.
127 + # clang turns off jump table generation by default when under
128 + # retpoline builds, however, gcc does not for x86. This has
129 + # only been fixed starting from gcc stable version 8.4.0 and
130 + # onwards, but not for older ones. See gcc bug #86952.
131 + ifndef CONFIG_CC_IS_CLANG
132 + KBUILD_CFLAGS += $(call cc-option,-fno-jump-tables)
133 + endif
134 endif
135
136 archscripts: scripts_basic
137 diff --git a/arch/x86/events/intel/cstate.c b/arch/x86/events/intel/cstate.c
138 index 9f8084f18d58..6eb76106c469 100644
139 --- a/arch/x86/events/intel/cstate.c
140 +++ b/arch/x86/events/intel/cstate.c
141 @@ -76,15 +76,15 @@
142 * Scope: Package (physical package)
143 * MSR_PKG_C8_RESIDENCY: Package C8 Residency Counter.
144 * perf code: 0x04
145 - * Available model: HSW ULT,CNL
146 + * Available model: HSW ULT,KBL,CNL
147 * Scope: Package (physical package)
148 * MSR_PKG_C9_RESIDENCY: Package C9 Residency Counter.
149 * perf code: 0x05
150 - * Available model: HSW ULT,CNL
151 + * Available model: HSW ULT,KBL,CNL
152 * Scope: Package (physical package)
153 * MSR_PKG_C10_RESIDENCY: Package C10 Residency Counter.
154 * perf code: 0x06
155 - * Available model: HSW ULT,GLM,CNL
156 + * Available model: HSW ULT,KBL,GLM,CNL
157 * Scope: Package (physical package)
158 *
159 */
160 @@ -572,8 +572,8 @@ static const struct x86_cpu_id intel_cstates_match[] __initconst = {
161 X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_DESKTOP, snb_cstates),
162 X86_CSTATES_MODEL(INTEL_FAM6_SKYLAKE_X, snb_cstates),
163
164 - X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE, snb_cstates),
165 - X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, snb_cstates),
166 + X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_MOBILE, hswult_cstates),
167 + X86_CSTATES_MODEL(INTEL_FAM6_KABYLAKE_DESKTOP, hswult_cstates),
168
169 X86_CSTATES_MODEL(INTEL_FAM6_CANNONLAKE_MOBILE, cnl_cstates),
170
171 diff --git a/arch/x86/include/asm/efi.h b/arch/x86/include/asm/efi.h
172 index cec5fae23eb3..baa549f8e918 100644
173 --- a/arch/x86/include/asm/efi.h
174 +++ b/arch/x86/include/asm/efi.h
175 @@ -82,8 +82,7 @@ struct efi_scratch {
176 #define arch_efi_call_virt_setup() \
177 ({ \
178 efi_sync_low_kernel_mappings(); \
179 - preempt_disable(); \
180 - __kernel_fpu_begin(); \
181 + kernel_fpu_begin(); \
182 firmware_restrict_branch_speculation_start(); \
183 \
184 if (!efi_enabled(EFI_OLD_MEMMAP)) \
185 @@ -99,8 +98,7 @@ struct efi_scratch {
186 efi_switch_mm(efi_scratch.prev_mm); \
187 \
188 firmware_restrict_branch_speculation_end(); \
189 - __kernel_fpu_end(); \
190 - preempt_enable(); \
191 + kernel_fpu_end(); \
192 })
193
194 extern void __iomem *__init efi_ioremap(unsigned long addr, unsigned long size,
195 diff --git a/arch/x86/include/asm/fpu/api.h b/arch/x86/include/asm/fpu/api.h
196 index a9caac9d4a72..b56d504af654 100644
197 --- a/arch/x86/include/asm/fpu/api.h
198 +++ b/arch/x86/include/asm/fpu/api.h
199 @@ -12,17 +12,12 @@
200 #define _ASM_X86_FPU_API_H
201
202 /*
203 - * Careful: __kernel_fpu_begin/end() must be called with preempt disabled
204 - * and they don't touch the preempt state on their own.
205 - * If you enable preemption after __kernel_fpu_begin(), preempt notifier
206 - * should call the __kernel_fpu_end() to prevent the kernel/user FPU
207 - * state from getting corrupted. KVM for example uses this model.
208 - *
209 - * All other cases use kernel_fpu_begin/end() which disable preemption
210 - * during kernel FPU usage.
211 + * Use kernel_fpu_begin/end() if you intend to use FPU in kernel context. It
212 + * disables preemption so be careful if you intend to use it for long periods
213 + * of time.
214 + * If you intend to use the FPU in softirq you need to check first with
215 + * irq_fpu_usable() if it is possible.
216 */
217 -extern void __kernel_fpu_begin(void);
218 -extern void __kernel_fpu_end(void);
219 extern void kernel_fpu_begin(void);
220 extern void kernel_fpu_end(void);
221 extern bool irq_fpu_usable(void);
222 diff --git a/arch/x86/kernel/fpu/core.c b/arch/x86/kernel/fpu/core.c
223 index 2ea85b32421a..2e5003fef51a 100644
224 --- a/arch/x86/kernel/fpu/core.c
225 +++ b/arch/x86/kernel/fpu/core.c
226 @@ -93,7 +93,7 @@ bool irq_fpu_usable(void)
227 }
228 EXPORT_SYMBOL(irq_fpu_usable);
229
230 -void __kernel_fpu_begin(void)
231 +static void __kernel_fpu_begin(void)
232 {
233 struct fpu *fpu = &current->thread.fpu;
234
235 @@ -111,9 +111,8 @@ void __kernel_fpu_begin(void)
236 __cpu_invalidate_fpregs_state();
237 }
238 }
239 -EXPORT_SYMBOL(__kernel_fpu_begin);
240
241 -void __kernel_fpu_end(void)
242 +static void __kernel_fpu_end(void)
243 {
244 struct fpu *fpu = &current->thread.fpu;
245
246 @@ -122,7 +121,6 @@ void __kernel_fpu_end(void)
247
248 kernel_fpu_enable();
249 }
250 -EXPORT_SYMBOL(__kernel_fpu_end);
251
252 void kernel_fpu_begin(void)
253 {
254 diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
255 index 030c98f35cca..a654ccfd1a22 100644
256 --- a/drivers/android/binder_alloc.c
257 +++ b/drivers/android/binder_alloc.c
258 @@ -958,14 +958,13 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
259
260 index = page - alloc->pages;
261 page_addr = (uintptr_t)alloc->buffer + index * PAGE_SIZE;
262 +
263 + mm = alloc->vma_vm_mm;
264 + if (!mmget_not_zero(mm))
265 + goto err_mmget;
266 + if (!down_write_trylock(&mm->mmap_sem))
267 + goto err_down_write_mmap_sem_failed;
268 vma = binder_alloc_get_vma(alloc);
269 - if (vma) {
270 - if (!mmget_not_zero(alloc->vma_vm_mm))
271 - goto err_mmget;
272 - mm = alloc->vma_vm_mm;
273 - if (!down_write_trylock(&mm->mmap_sem))
274 - goto err_down_write_mmap_sem_failed;
275 - }
276
277 list_lru_isolate(lru, item);
278 spin_unlock(lock);
279 @@ -978,10 +977,9 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
280 PAGE_SIZE);
281
282 trace_binder_unmap_user_end(alloc, index);
283 -
284 - up_write(&mm->mmap_sem);
285 - mmput(mm);
286 }
287 + up_write(&mm->mmap_sem);
288 + mmput(mm);
289
290 trace_binder_unmap_kernel_start(alloc, index);
291
292 diff --git a/drivers/block/loop.c b/drivers/block/loop.c
293 index a63da9e07341..f1e63eb7cbca 100644
294 --- a/drivers/block/loop.c
295 +++ b/drivers/block/loop.c
296 @@ -1112,8 +1112,9 @@ out_unlock:
297 err = __blkdev_reread_part(bdev);
298 else
299 err = blkdev_reread_part(bdev);
300 - pr_warn("%s: partition scan of loop%d failed (rc=%d)\n",
301 - __func__, lo_number, err);
302 + if (err)
303 + pr_warn("%s: partition scan of loop%d failed (rc=%d)\n",
304 + __func__, lo_number, err);
305 /* Device is gone, no point in returning error */
306 err = 0;
307 }
308 diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
309 index a65505db09e5..70cbd0ee1b07 100644
310 --- a/drivers/block/zram/zram_drv.c
311 +++ b/drivers/block/zram/zram_drv.c
312 @@ -515,18 +515,18 @@ struct zram_work {
313 struct zram *zram;
314 unsigned long entry;
315 struct bio *bio;
316 + struct bio_vec bvec;
317 };
318
319 #if PAGE_SIZE != 4096
320 static void zram_sync_read(struct work_struct *work)
321 {
322 - struct bio_vec bvec;
323 struct zram_work *zw = container_of(work, struct zram_work, work);
324 struct zram *zram = zw->zram;
325 unsigned long entry = zw->entry;
326 struct bio *bio = zw->bio;
327
328 - read_from_bdev_async(zram, &bvec, entry, bio);
329 + read_from_bdev_async(zram, &zw->bvec, entry, bio);
330 }
331
332 /*
333 @@ -539,6 +539,7 @@ static int read_from_bdev_sync(struct zram *zram, struct bio_vec *bvec,
334 {
335 struct zram_work work;
336
337 + work.bvec = *bvec;
338 work.zram = zram;
339 work.entry = entry;
340 work.bio = bio;
341 diff --git a/drivers/dma/sh/rcar-dmac.c b/drivers/dma/sh/rcar-dmac.c
342 index 48ee35e2bce6..0b05a1e08d21 100644
343 --- a/drivers/dma/sh/rcar-dmac.c
344 +++ b/drivers/dma/sh/rcar-dmac.c
345 @@ -1281,6 +1281,9 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan,
346 enum dma_status status;
347 unsigned int residue = 0;
348 unsigned int dptr = 0;
349 + unsigned int chcrb;
350 + unsigned int tcrb;
351 + unsigned int i;
352
353 if (!desc)
354 return 0;
355 @@ -1328,6 +1331,24 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan,
356 return 0;
357 }
358
359 + /*
360 + * We need to read two registers.
361 + * Make sure the control register does not skip to next chunk
362 + * while reading the counter.
363 + * Trying it 3 times should be enough: Initial read, retry, retry
364 + * for the paranoid.
365 + */
366 + for (i = 0; i < 3; i++) {
367 + chcrb = rcar_dmac_chan_read(chan, RCAR_DMACHCRB) &
368 + RCAR_DMACHCRB_DPTR_MASK;
369 + tcrb = rcar_dmac_chan_read(chan, RCAR_DMATCRB);
370 + /* Still the same? */
371 + if (chcrb == (rcar_dmac_chan_read(chan, RCAR_DMACHCRB) &
372 + RCAR_DMACHCRB_DPTR_MASK))
373 + break;
374 + }
375 + WARN_ONCE(i >= 3, "residue might be not continuous!");
376 +
377 /*
378 * In descriptor mode the descriptor running pointer is not maintained
379 * by the interrupt handler, find the running descriptor from the
380 @@ -1335,8 +1356,7 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan,
381 * mode just use the running descriptor pointer.
382 */
383 if (desc->hwdescs.use) {
384 - dptr = (rcar_dmac_chan_read(chan, RCAR_DMACHCRB) &
385 - RCAR_DMACHCRB_DPTR_MASK) >> RCAR_DMACHCRB_DPTR_SHIFT;
386 + dptr = chcrb >> RCAR_DMACHCRB_DPTR_SHIFT;
387 if (dptr == 0)
388 dptr = desc->nchunks;
389 dptr--;
390 @@ -1354,7 +1374,7 @@ static unsigned int rcar_dmac_chan_get_residue(struct rcar_dmac_chan *chan,
391 }
392
393 /* Add the residue for the current chunk. */
394 - residue += rcar_dmac_chan_read(chan, RCAR_DMATCRB) << desc->xfer_shift;
395 + residue += tcrb << desc->xfer_shift;
396
397 return residue;
398 }
399 @@ -1367,6 +1387,7 @@ static enum dma_status rcar_dmac_tx_status(struct dma_chan *chan,
400 enum dma_status status;
401 unsigned long flags;
402 unsigned int residue;
403 + bool cyclic;
404
405 status = dma_cookie_status(chan, cookie, txstate);
406 if (status == DMA_COMPLETE || !txstate)
407 @@ -1374,10 +1395,11 @@ static enum dma_status rcar_dmac_tx_status(struct dma_chan *chan,
408
409 spin_lock_irqsave(&rchan->lock, flags);
410 residue = rcar_dmac_chan_get_residue(rchan, cookie);
411 + cyclic = rchan->desc.running ? rchan->desc.running->cyclic : false;
412 spin_unlock_irqrestore(&rchan->lock, flags);
413
414 /* if there's no residue, the cookie is complete */
415 - if (!residue)
416 + if (!residue && !cyclic)
417 return DMA_COMPLETE;
418
419 dma_set_residue(txstate, residue);
420 diff --git a/drivers/gpio/gpio-eic-sprd.c b/drivers/gpio/gpio-eic-sprd.c
421 index e41223c05f6e..6cf2e2ce4093 100644
422 --- a/drivers/gpio/gpio-eic-sprd.c
423 +++ b/drivers/gpio/gpio-eic-sprd.c
424 @@ -414,6 +414,7 @@ static int sprd_eic_irq_set_type(struct irq_data *data, unsigned int flow_type)
425 irq_set_handler_locked(data, handle_edge_irq);
426 break;
427 case IRQ_TYPE_EDGE_BOTH:
428 + sprd_eic_update(chip, offset, SPRD_EIC_SYNC_INTMODE, 0);
429 sprd_eic_update(chip, offset, SPRD_EIC_SYNC_INTBOTH, 1);
430 irq_set_handler_locked(data, handle_edge_irq);
431 break;
432 diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c
433 index 6f91634880aa..2d6506c08bf7 100644
434 --- a/drivers/gpu/drm/i915/intel_fbdev.c
435 +++ b/drivers/gpu/drm/i915/intel_fbdev.c
436 @@ -334,8 +334,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
437 bool *enabled, int width, int height)
438 {
439 struct drm_i915_private *dev_priv = to_i915(fb_helper->dev);
440 + unsigned long conn_configured, conn_seq, mask;
441 unsigned int count = min(fb_helper->connector_count, BITS_PER_LONG);
442 - unsigned long conn_configured, conn_seq;
443 int i, j;
444 bool *save_enabled;
445 bool fallback = true, ret = true;
446 @@ -353,9 +353,10 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
447 drm_modeset_backoff(&ctx);
448
449 memcpy(save_enabled, enabled, count);
450 - conn_seq = GENMASK(count - 1, 0);
451 + mask = GENMASK(count - 1, 0);
452 conn_configured = 0;
453 retry:
454 + conn_seq = conn_configured;
455 for (i = 0; i < count; i++) {
456 struct drm_fb_helper_connector *fb_conn;
457 struct drm_connector *connector;
458 @@ -368,8 +369,7 @@ retry:
459 if (conn_configured & BIT(i))
460 continue;
461
462 - /* First pass, only consider tiled connectors */
463 - if (conn_seq == GENMASK(count - 1, 0) && !connector->has_tile)
464 + if (conn_seq == 0 && !connector->has_tile)
465 continue;
466
467 if (connector->status == connector_status_connected)
468 @@ -473,10 +473,8 @@ retry:
469 conn_configured |= BIT(i);
470 }
471
472 - if (conn_configured != conn_seq) { /* repeat until no more are found */
473 - conn_seq = conn_configured;
474 + if ((conn_configured & mask) != mask && conn_configured != conn_seq)
475 goto retry;
476 - }
477
478 /*
479 * If the BIOS didn't enable everything it could, fall back to have the
480 diff --git a/drivers/gpu/drm/rockchip/cdn-dp-reg.c b/drivers/gpu/drm/rockchip/cdn-dp-reg.c
481 index 5a485489a1e2..6c8b14fb1d2f 100644
482 --- a/drivers/gpu/drm/rockchip/cdn-dp-reg.c
483 +++ b/drivers/gpu/drm/rockchip/cdn-dp-reg.c
484 @@ -113,7 +113,7 @@ static int cdp_dp_mailbox_write(struct cdn_dp_device *dp, u8 val)
485
486 static int cdn_dp_mailbox_validate_receive(struct cdn_dp_device *dp,
487 u8 module_id, u8 opcode,
488 - u8 req_size)
489 + u16 req_size)
490 {
491 u32 mbox_size, i;
492 u8 header[4];
493 diff --git a/drivers/gpu/drm/vc4/vc4_crtc.c b/drivers/gpu/drm/vc4/vc4_crtc.c
494 index 0e6a121858d1..5615ceb15708 100644
495 --- a/drivers/gpu/drm/vc4/vc4_crtc.c
496 +++ b/drivers/gpu/drm/vc4/vc4_crtc.c
497 @@ -998,7 +998,7 @@ static void
498 vc4_crtc_reset(struct drm_crtc *crtc)
499 {
500 if (crtc->state)
501 - __drm_atomic_helper_crtc_destroy_state(crtc->state);
502 + vc4_crtc_destroy_state(crtc, crtc->state);
503
504 crtc->state = kzalloc(sizeof(struct vc4_crtc_state), GFP_KERNEL);
505 if (crtc->state)
506 diff --git a/drivers/hwtracing/intel_th/gth.c b/drivers/hwtracing/intel_th/gth.c
507 index cc287cf6eb29..edc52d75e6bd 100644
508 --- a/drivers/hwtracing/intel_th/gth.c
509 +++ b/drivers/hwtracing/intel_th/gth.c
510 @@ -616,7 +616,7 @@ static void intel_th_gth_unassign(struct intel_th_device *thdev,
511 othdev->output.port = -1;
512 othdev->output.active = false;
513 gth->output[port].output = NULL;
514 - for (master = 0; master < TH_CONFIGURABLE_MASTERS; master++)
515 + for (master = 0; master <= TH_CONFIGURABLE_MASTERS; master++)
516 if (gth->master[master] == port)
517 gth->master[master] = -1;
518 spin_unlock(&gth->gth_lock);
519 diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
520 index 50be240df331..8cc4da62f050 100644
521 --- a/drivers/infiniband/hw/mlx5/main.c
522 +++ b/drivers/infiniband/hw/mlx5/main.c
523 @@ -2014,6 +2014,7 @@ static int mlx5_ib_mmap_clock_info_page(struct mlx5_ib_dev *dev,
524
525 if (vma->vm_flags & VM_WRITE)
526 return -EPERM;
527 + vma->vm_flags &= ~VM_MAYWRITE;
528
529 if (!dev->mdev->clock_info_page)
530 return -EOPNOTSUPP;
531 @@ -2197,6 +2198,7 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
532
533 if (vma->vm_flags & VM_WRITE)
534 return -EPERM;
535 + vma->vm_flags &= ~VM_MAYWRITE;
536
537 /* Don't expose to user-space information it shouldn't have */
538 if (PAGE_SIZE > 4096)
539 diff --git a/drivers/infiniband/sw/rdmavt/mr.c b/drivers/infiniband/sw/rdmavt/mr.c
540 index 49c9541050d4..5819c9d6ffdc 100644
541 --- a/drivers/infiniband/sw/rdmavt/mr.c
542 +++ b/drivers/infiniband/sw/rdmavt/mr.c
543 @@ -611,11 +611,6 @@ static int rvt_set_page(struct ib_mr *ibmr, u64 addr)
544 if (unlikely(mapped_segs == mr->mr.max_segs))
545 return -ENOMEM;
546
547 - if (mr->mr.length == 0) {
548 - mr->mr.user_base = addr;
549 - mr->mr.iova = addr;
550 - }
551 -
552 m = mapped_segs / RVT_SEGSZ;
553 n = mapped_segs % RVT_SEGSZ;
554 mr->mr.map[m]->segs[n].vaddr = (void *)addr;
555 @@ -633,17 +628,24 @@ static int rvt_set_page(struct ib_mr *ibmr, u64 addr)
556 * @sg_nents: number of entries in sg
557 * @sg_offset: offset in bytes into sg
558 *
559 + * Overwrite rvt_mr length with mr length calculated by ib_sg_to_pages.
560 + *
561 * Return: number of sg elements mapped to the memory region
562 */
563 int rvt_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg,
564 int sg_nents, unsigned int *sg_offset)
565 {
566 struct rvt_mr *mr = to_imr(ibmr);
567 + int ret;
568
569 mr->mr.length = 0;
570 mr->mr.page_shift = PAGE_SHIFT;
571 - return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset,
572 - rvt_set_page);
573 + ret = ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, rvt_set_page);
574 + mr->mr.user_base = ibmr->iova;
575 + mr->mr.iova = ibmr->iova;
576 + mr->mr.offset = ibmr->iova - (u64)mr->mr.map[0]->segs[0].vaddr;
577 + mr->mr.length = (size_t)ibmr->length;
578 + return ret;
579 }
580
581 /**
582 @@ -674,6 +676,7 @@ int rvt_fast_reg_mr(struct rvt_qp *qp, struct ib_mr *ibmr, u32 key,
583 ibmr->rkey = key;
584 mr->mr.lkey = key;
585 mr->mr.access_flags = access;
586 + mr->mr.iova = ibmr->iova;
587 atomic_set(&mr->mr.lkey_invalid, 0);
588
589 return 0;
590 diff --git a/drivers/input/rmi4/rmi_f11.c b/drivers/input/rmi4/rmi_f11.c
591 index df64d6aed4f7..93901ebd122a 100644
592 --- a/drivers/input/rmi4/rmi_f11.c
593 +++ b/drivers/input/rmi4/rmi_f11.c
594 @@ -1230,7 +1230,7 @@ static int rmi_f11_initialize(struct rmi_function *fn)
595 }
596
597 rc = f11_write_control_regs(fn, &f11->sens_query,
598 - &f11->dev_controls, fn->fd.query_base_addr);
599 + &f11->dev_controls, fn->fd.control_base_addr);
600 if (rc)
601 dev_warn(&fn->dev, "Failed to write control registers\n");
602
603 diff --git a/drivers/net/dsa/mv88e6xxx/chip.c b/drivers/net/dsa/mv88e6xxx/chip.c
604 index dabe89968a78..2caa5c0c2bc4 100644
605 --- a/drivers/net/dsa/mv88e6xxx/chip.c
606 +++ b/drivers/net/dsa/mv88e6xxx/chip.c
607 @@ -4821,6 +4821,7 @@ static int mv88e6xxx_probe(struct mdio_device *mdiodev)
608 if (err)
609 goto out;
610
611 + mv88e6xxx_ports_cmode_init(chip);
612 mv88e6xxx_phy_init(chip);
613
614 if (chip->info->ops->get_eeprom) {
615 diff --git a/drivers/net/ethernet/hisilicon/hns/hns_enet.c b/drivers/net/ethernet/hisilicon/hns/hns_enet.c
616 index b043370c2685..cc84133c184d 100644
617 --- a/drivers/net/ethernet/hisilicon/hns/hns_enet.c
618 +++ b/drivers/net/ethernet/hisilicon/hns/hns_enet.c
619 @@ -1169,6 +1169,12 @@ int hns_nic_init_phy(struct net_device *ndev, struct hnae_handle *h)
620 if (!h->phy_dev)
621 return 0;
622
623 + phy_dev->supported &= h->if_support;
624 + phy_dev->advertising = phy_dev->supported;
625 +
626 + if (h->phy_if == PHY_INTERFACE_MODE_XGMII)
627 + phy_dev->autoneg = false;
628 +
629 if (h->phy_if != PHY_INTERFACE_MODE_XGMII) {
630 phy_dev->dev_flags = 0;
631
632 @@ -1180,15 +1186,6 @@ int hns_nic_init_phy(struct net_device *ndev, struct hnae_handle *h)
633 if (unlikely(ret))
634 return -ENODEV;
635
636 - phy_dev->supported &= h->if_support;
637 - phy_dev->advertising = phy_dev->supported;
638 -
639 - if (h->phy_if == PHY_INTERFACE_MODE_XGMII)
640 - phy_dev->autoneg = false;
641 -
642 - if (h->phy_if == PHY_INTERFACE_MODE_SGMII)
643 - phy_stop(phy_dev);
644 -
645 return 0;
646 }
647
648 diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c
649 index a475f36ddf8c..426789e2c23d 100644
650 --- a/drivers/net/ethernet/ibm/ibmvnic.c
651 +++ b/drivers/net/ethernet/ibm/ibmvnic.c
652 @@ -1859,7 +1859,7 @@ static int do_reset(struct ibmvnic_adapter *adapter,
653
654 if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
655 adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
656 - netdev_notify_peers(netdev);
657 + call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
658
659 netif_carrier_on(netdev);
660
661 diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_main.c b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
662 index 3f536541f45f..78a43d688cb1 100644
663 --- a/drivers/net/ethernet/intel/fm10k/fm10k_main.c
664 +++ b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
665 @@ -41,6 +41,8 @@ static int __init fm10k_init_module(void)
666 /* create driver workqueue */
667 fm10k_workqueue = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0,
668 fm10k_driver_name);
669 + if (!fm10k_workqueue)
670 + return -ENOMEM;
671
672 fm10k_dbg_init();
673
674 diff --git a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
675 index 9988c89ed9fd..9b10abb604cb 100644
676 --- a/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
677 +++ b/drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c
678 @@ -4272,7 +4272,7 @@ static void mvpp2_phylink_validate(struct net_device *dev,
679 case PHY_INTERFACE_MODE_RGMII_ID:
680 case PHY_INTERFACE_MODE_RGMII_RXID:
681 case PHY_INTERFACE_MODE_RGMII_TXID:
682 - if (port->gop_id == 0)
683 + if (port->priv->hw_version == MVPP22 && port->gop_id == 0)
684 goto empty_set;
685 break;
686 default:
687 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
688 index 4a33c9a7cac7..599114ab7821 100644
689 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
690 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c
691 @@ -33,6 +33,26 @@
692 #include <linux/bpf_trace.h>
693 #include "en/xdp.h"
694
695 +int mlx5e_xdp_max_mtu(struct mlx5e_params *params)
696 +{
697 + int hr = NET_IP_ALIGN + XDP_PACKET_HEADROOM;
698 +
699 + /* Let S := SKB_DATA_ALIGN(sizeof(struct skb_shared_info)).
700 + * The condition checked in mlx5e_rx_is_linear_skb is:
701 + * SKB_DATA_ALIGN(sw_mtu + hard_mtu + hr) + S <= PAGE_SIZE (1)
702 + * (Note that hw_mtu == sw_mtu + hard_mtu.)
703 + * What is returned from this function is:
704 + * max_mtu = PAGE_SIZE - S - hr - hard_mtu (2)
705 + * After assigning sw_mtu := max_mtu, the left side of (1) turns to
706 + * SKB_DATA_ALIGN(PAGE_SIZE - S) + S, which is equal to PAGE_SIZE,
707 + * because both PAGE_SIZE and S are already aligned. Any number greater
708 + * than max_mtu would make the left side of (1) greater than PAGE_SIZE,
709 + * so max_mtu is the maximum MTU allowed.
710 + */
711 +
712 + return MLX5E_HW2SW_MTU(params, SKB_MAX_HEAD(hr));
713 +}
714 +
715 static inline bool
716 mlx5e_xmit_xdp_buff(struct mlx5e_xdpsq *sq, struct mlx5e_dma_info *di,
717 struct xdp_buff *xdp)
718 @@ -207,9 +227,9 @@ bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq)
719 sqcc++;
720
721 if (is_redirect) {
722 - xdp_return_frame(xdpi->xdpf);
723 dma_unmap_single(sq->pdev, xdpi->dma_addr,
724 xdpi->xdpf->len, DMA_TO_DEVICE);
725 + xdp_return_frame(xdpi->xdpf);
726 } else {
727 /* Recycle RX page */
728 mlx5e_page_release(rq, &xdpi->di, true);
729 @@ -243,9 +263,9 @@ void mlx5e_free_xdpsq_descs(struct mlx5e_xdpsq *sq)
730 sq->cc++;
731
732 if (is_redirect) {
733 - xdp_return_frame(xdpi->xdpf);
734 dma_unmap_single(sq->pdev, xdpi->dma_addr,
735 xdpi->xdpf->len, DMA_TO_DEVICE);
736 + xdp_return_frame(xdpi->xdpf);
737 } else {
738 /* Recycle RX page */
739 mlx5e_page_release(rq, &xdpi->di, false);
740 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
741 index 4d096623178b..827ceef5fa93 100644
742 --- a/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
743 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h
744 @@ -34,12 +34,11 @@
745
746 #include "en.h"
747
748 -#define MLX5E_XDP_MAX_MTU ((int)(PAGE_SIZE - \
749 - MLX5_SKB_FRAG_SZ(XDP_PACKET_HEADROOM)))
750 #define MLX5E_XDP_MIN_INLINE (ETH_HLEN + VLAN_HLEN)
751 #define MLX5E_XDP_TX_DS_COUNT \
752 ((sizeof(struct mlx5e_tx_wqe) / MLX5_SEND_WQE_DS) + 1 /* SG DS */)
753
754 +int mlx5e_xdp_max_mtu(struct mlx5e_params *params);
755 bool mlx5e_xdp_handle(struct mlx5e_rq *rq, struct mlx5e_dma_info *di,
756 void *va, u16 *rx_headroom, u32 *len);
757 bool mlx5e_poll_xdpsq_cq(struct mlx5e_cq *cq);
758 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
759 index 5e5423076b03..9ca4f88d7cf6 100644
760 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
761 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c
762 @@ -1317,7 +1317,7 @@ static int mlx5e_get_module_info(struct net_device *netdev,
763 break;
764 case MLX5_MODULE_ID_SFP:
765 modinfo->type = ETH_MODULE_SFF_8472;
766 - modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
767 + modinfo->eeprom_len = MLX5_EEPROM_PAGE_LENGTH;
768 break;
769 default:
770 netdev_err(priv->netdev, "%s: cable type not recognized:0x%x\n",
771 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
772 index 637d59c01fe5..b190c447aeb0 100644
773 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
774 +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
775 @@ -3761,7 +3761,7 @@ int mlx5e_change_mtu(struct net_device *netdev, int new_mtu,
776 if (params->xdp_prog &&
777 !mlx5e_rx_is_linear_skb(priv->mdev, &new_channels.params)) {
778 netdev_err(netdev, "MTU(%d) > %d is not allowed while XDP enabled\n",
779 - new_mtu, MLX5E_XDP_MAX_MTU);
780 + new_mtu, mlx5e_xdp_max_mtu(params));
781 err = -EINVAL;
782 goto out;
783 }
784 @@ -4227,7 +4227,8 @@ static int mlx5e_xdp_allowed(struct mlx5e_priv *priv, struct bpf_prog *prog)
785
786 if (!mlx5e_rx_is_linear_skb(priv->mdev, &new_channels.params)) {
787 netdev_warn(netdev, "XDP is not allowed with MTU(%d) > %d\n",
788 - new_channels.params.sw_mtu, MLX5E_XDP_MAX_MTU);
789 + new_channels.params.sw_mtu,
790 + mlx5e_xdp_max_mtu(&new_channels.params));
791 return -EINVAL;
792 }
793
794 diff --git a/drivers/net/ethernet/mellanox/mlx5/core/port.c b/drivers/net/ethernet/mellanox/mlx5/core/port.c
795 index 31a9cbd85689..09b6b1bfbfa8 100644
796 --- a/drivers/net/ethernet/mellanox/mlx5/core/port.c
797 +++ b/drivers/net/ethernet/mellanox/mlx5/core/port.c
798 @@ -404,10 +404,6 @@ int mlx5_query_module_eeprom(struct mlx5_core_dev *dev,
799 size -= offset + size - MLX5_EEPROM_PAGE_LENGTH;
800
801 i2c_addr = MLX5_I2C_ADDR_LOW;
802 - if (offset >= MLX5_EEPROM_PAGE_LENGTH) {
803 - i2c_addr = MLX5_I2C_ADDR_HIGH;
804 - offset -= MLX5_EEPROM_PAGE_LENGTH;
805 - }
806
807 MLX5_SET(mcia_reg, in, l, 0);
808 MLX5_SET(mcia_reg, in, module, module_num);
809 diff --git a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
810 index 72cdaa01d56d..100618531021 100644
811 --- a/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
812 +++ b/drivers/net/ethernet/mellanox/mlxsw/pci_hw.h
813 @@ -27,7 +27,7 @@
814
815 #define MLXSW_PCI_SW_RESET 0xF0010
816 #define MLXSW_PCI_SW_RESET_RST_BIT BIT(0)
817 -#define MLXSW_PCI_SW_RESET_TIMEOUT_MSECS 13000
818 +#define MLXSW_PCI_SW_RESET_TIMEOUT_MSECS 20000
819 #define MLXSW_PCI_SW_RESET_WAIT_MSECS 100
820 #define MLXSW_PCI_FW_READY 0xA1844
821 #define MLXSW_PCI_FW_READY_MASK 0xFFFF
822 diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
823 index f9bef030ee05..c5b82e283d13 100644
824 --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
825 +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum.c
826 @@ -2504,11 +2504,11 @@ mlxsw_sp_port_set_link_ksettings(struct net_device *dev,
827 if (err)
828 return err;
829
830 + mlxsw_sp_port->link.autoneg = autoneg;
831 +
832 if (!netif_running(dev))
833 return 0;
834
835 - mlxsw_sp_port->link.autoneg = autoneg;
836 -
837 mlxsw_sp_port_admin_status_set(mlxsw_sp_port, false);
838 mlxsw_sp_port_admin_status_set(mlxsw_sp_port, true);
839
840 @@ -2783,7 +2783,7 @@ static int mlxsw_sp_port_ets_init(struct mlxsw_sp_port *mlxsw_sp_port)
841 err = mlxsw_sp_port_ets_set(mlxsw_sp_port,
842 MLXSW_REG_QEEC_HIERARCY_TC,
843 i + 8, i,
844 - false, 0);
845 + true, 100);
846 if (err)
847 return err;
848 }
849 diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
850 index 886176be818e..62460a5b4ad9 100644
851 --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
852 +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
853 @@ -2595,8 +2595,6 @@ static int stmmac_open(struct net_device *dev)
854 u32 chan;
855 int ret;
856
857 - stmmac_check_ether_addr(priv);
858 -
859 if (priv->hw->pcs != STMMAC_PCS_RGMII &&
860 priv->hw->pcs != STMMAC_PCS_TBI &&
861 priv->hw->pcs != STMMAC_PCS_RTBI) {
862 @@ -4296,6 +4294,8 @@ int stmmac_dvr_probe(struct device *device,
863 if (ret)
864 goto error_hw_init;
865
866 + stmmac_check_ether_addr(priv);
867 +
868 /* Configure real RX and TX queues */
869 netif_set_real_num_rx_queues(ndev, priv->plat->rx_queues_to_use);
870 netif_set_real_num_tx_queues(ndev, priv->plat->tx_queues_to_use);
871 diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
872 index d819e8eaba12..cc1e887e47b5 100644
873 --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
874 +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c
875 @@ -159,6 +159,12 @@ static const struct dmi_system_id quark_pci_dmi[] = {
876 },
877 .driver_data = (void *)&galileo_stmmac_dmi_data,
878 },
879 + /*
880 + * There are 2 types of SIMATIC IOT2000: IOT20202 and IOT2040.
881 + * The asset tag "6ES7647-0AA00-0YA2" is only for IOT2020 which
882 + * has only one pci network device while other asset tags are
883 + * for IOT2040 which has two.
884 + */
885 {
886 .matches = {
887 DMI_EXACT_MATCH(DMI_BOARD_NAME, "SIMATIC IOT2000"),
888 @@ -170,8 +176,6 @@ static const struct dmi_system_id quark_pci_dmi[] = {
889 {
890 .matches = {
891 DMI_EXACT_MATCH(DMI_BOARD_NAME, "SIMATIC IOT2000"),
892 - DMI_EXACT_MATCH(DMI_BOARD_ASSET_TAG,
893 - "6ES7647-0AA00-1YA2"),
894 },
895 .driver_data = (void *)&iot2040_stmmac_dmi_data,
896 },
897 diff --git a/drivers/net/slip/slhc.c b/drivers/net/slip/slhc.c
898 index f4e93f5fc204..ea90db3c7705 100644
899 --- a/drivers/net/slip/slhc.c
900 +++ b/drivers/net/slip/slhc.c
901 @@ -153,7 +153,7 @@ out_fail:
902 void
903 slhc_free(struct slcompress *comp)
904 {
905 - if ( comp == NULLSLCOMPR )
906 + if ( IS_ERR_OR_NULL(comp) )
907 return;
908
909 if ( comp->tstate != NULLSLSTATE )
910 diff --git a/drivers/net/team/team.c b/drivers/net/team/team.c
911 index e23eaf3f6d03..6c6230b44bcd 100644
912 --- a/drivers/net/team/team.c
913 +++ b/drivers/net/team/team.c
914 @@ -1160,6 +1160,13 @@ static int team_port_add(struct team *team, struct net_device *port_dev,
915 return -EINVAL;
916 }
917
918 + if (netdev_has_upper_dev(dev, port_dev)) {
919 + NL_SET_ERR_MSG(extack, "Device is already an upper device of the team interface");
920 + netdev_err(dev, "Device %s is already an upper device of the team interface\n",
921 + portname);
922 + return -EBUSY;
923 + }
924 +
925 if (port_dev->features & NETIF_F_VLAN_CHALLENGED &&
926 vlan_uses_dev(dev)) {
927 NL_SET_ERR_MSG(extack, "Device is VLAN challenged and team device has VLAN set up");
928 diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c
929 index a1f225f077cd..ef47c226e1d2 100644
930 --- a/drivers/usb/core/driver.c
931 +++ b/drivers/usb/core/driver.c
932 @@ -1899,14 +1899,11 @@ int usb_runtime_idle(struct device *dev)
933 return -EBUSY;
934 }
935
936 -int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable)
937 +static int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable)
938 {
939 struct usb_hcd *hcd = bus_to_hcd(udev->bus);
940 int ret = -EPERM;
941
942 - if (enable && !udev->usb2_hw_lpm_allowed)
943 - return 0;
944 -
945 if (hcd->driver->set_usb2_hw_lpm) {
946 ret = hcd->driver->set_usb2_hw_lpm(hcd, udev, enable);
947 if (!ret)
948 @@ -1916,6 +1913,24 @@ int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable)
949 return ret;
950 }
951
952 +int usb_enable_usb2_hardware_lpm(struct usb_device *udev)
953 +{
954 + if (!udev->usb2_hw_lpm_capable ||
955 + !udev->usb2_hw_lpm_allowed ||
956 + udev->usb2_hw_lpm_enabled)
957 + return 0;
958 +
959 + return usb_set_usb2_hardware_lpm(udev, 1);
960 +}
961 +
962 +int usb_disable_usb2_hardware_lpm(struct usb_device *udev)
963 +{
964 + if (!udev->usb2_hw_lpm_enabled)
965 + return 0;
966 +
967 + return usb_set_usb2_hardware_lpm(udev, 0);
968 +}
969 +
970 #endif /* CONFIG_PM */
971
972 struct bus_type usb_bus_type = {
973 diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
974 index 3adff4da2ee1..bbcfa63d0233 100644
975 --- a/drivers/usb/core/hub.c
976 +++ b/drivers/usb/core/hub.c
977 @@ -3217,8 +3217,7 @@ int usb_port_suspend(struct usb_device *udev, pm_message_t msg)
978 }
979
980 /* disable USB2 hardware LPM */
981 - if (udev->usb2_hw_lpm_enabled == 1)
982 - usb_set_usb2_hardware_lpm(udev, 0);
983 + usb_disable_usb2_hardware_lpm(udev);
984
985 if (usb_disable_ltm(udev)) {
986 dev_err(&udev->dev, "Failed to disable LTM before suspend\n");
987 @@ -3256,8 +3255,7 @@ int usb_port_suspend(struct usb_device *udev, pm_message_t msg)
988 usb_enable_ltm(udev);
989 err_ltm:
990 /* Try to enable USB2 hardware LPM again */
991 - if (udev->usb2_hw_lpm_capable == 1)
992 - usb_set_usb2_hardware_lpm(udev, 1);
993 + usb_enable_usb2_hardware_lpm(udev);
994
995 if (udev->do_remote_wakeup)
996 (void) usb_disable_remote_wakeup(udev);
997 @@ -3540,8 +3538,7 @@ int usb_port_resume(struct usb_device *udev, pm_message_t msg)
998 hub_port_logical_disconnect(hub, port1);
999 } else {
1000 /* Try to enable USB2 hardware LPM */
1001 - if (udev->usb2_hw_lpm_capable == 1)
1002 - usb_set_usb2_hardware_lpm(udev, 1);
1003 + usb_enable_usb2_hardware_lpm(udev);
1004
1005 /* Try to enable USB3 LTM */
1006 usb_enable_ltm(udev);
1007 @@ -4432,7 +4429,7 @@ static void hub_set_initial_usb2_lpm_policy(struct usb_device *udev)
1008 if ((udev->bos->ext_cap->bmAttributes & cpu_to_le32(USB_BESL_SUPPORT)) ||
1009 connect_type == USB_PORT_CONNECT_TYPE_HARD_WIRED) {
1010 udev->usb2_hw_lpm_allowed = 1;
1011 - usb_set_usb2_hardware_lpm(udev, 1);
1012 + usb_enable_usb2_hardware_lpm(udev);
1013 }
1014 }
1015
1016 @@ -5608,8 +5605,7 @@ static int usb_reset_and_verify_device(struct usb_device *udev)
1017 /* Disable USB2 hardware LPM.
1018 * It will be re-enabled by the enumeration process.
1019 */
1020 - if (udev->usb2_hw_lpm_enabled == 1)
1021 - usb_set_usb2_hardware_lpm(udev, 0);
1022 + usb_disable_usb2_hardware_lpm(udev);
1023
1024 /* Disable LPM while we reset the device and reinstall the alt settings.
1025 * Device-initiated LPM, and system exit latency settings are cleared
1026 @@ -5712,7 +5708,7 @@ static int usb_reset_and_verify_device(struct usb_device *udev)
1027
1028 done:
1029 /* Now that the alt settings are re-installed, enable LTM and LPM. */
1030 - usb_set_usb2_hardware_lpm(udev, 1);
1031 + usb_enable_usb2_hardware_lpm(udev);
1032 usb_unlocked_enable_lpm(udev);
1033 usb_enable_ltm(udev);
1034 usb_release_bos_descriptor(udev);
1035 diff --git a/drivers/usb/core/message.c b/drivers/usb/core/message.c
1036 index bfa5eda0cc26..4f33eb632a88 100644
1037 --- a/drivers/usb/core/message.c
1038 +++ b/drivers/usb/core/message.c
1039 @@ -1243,8 +1243,7 @@ void usb_disable_device(struct usb_device *dev, int skip_ep0)
1040 dev->actconfig->interface[i] = NULL;
1041 }
1042
1043 - if (dev->usb2_hw_lpm_enabled == 1)
1044 - usb_set_usb2_hardware_lpm(dev, 0);
1045 + usb_disable_usb2_hardware_lpm(dev);
1046 usb_unlocked_disable_lpm(dev);
1047 usb_disable_ltm(dev);
1048
1049 diff --git a/drivers/usb/core/sysfs.c b/drivers/usb/core/sysfs.c
1050 index ea18284dfa9a..7e88fdfe3cf5 100644
1051 --- a/drivers/usb/core/sysfs.c
1052 +++ b/drivers/usb/core/sysfs.c
1053 @@ -528,7 +528,10 @@ static ssize_t usb2_hardware_lpm_store(struct device *dev,
1054
1055 if (!ret) {
1056 udev->usb2_hw_lpm_allowed = value;
1057 - ret = usb_set_usb2_hardware_lpm(udev, value);
1058 + if (value)
1059 + ret = usb_enable_usb2_hardware_lpm(udev);
1060 + else
1061 + ret = usb_disable_usb2_hardware_lpm(udev);
1062 }
1063
1064 usb_unlock_device(udev);
1065 diff --git a/drivers/usb/core/usb.h b/drivers/usb/core/usb.h
1066 index 546a2219454b..d95a5358f73d 100644
1067 --- a/drivers/usb/core/usb.h
1068 +++ b/drivers/usb/core/usb.h
1069 @@ -92,7 +92,8 @@ extern int usb_remote_wakeup(struct usb_device *dev);
1070 extern int usb_runtime_suspend(struct device *dev);
1071 extern int usb_runtime_resume(struct device *dev);
1072 extern int usb_runtime_idle(struct device *dev);
1073 -extern int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable);
1074 +extern int usb_enable_usb2_hardware_lpm(struct usb_device *udev);
1075 +extern int usb_disable_usb2_hardware_lpm(struct usb_device *udev);
1076
1077 #else
1078
1079 @@ -112,7 +113,12 @@ static inline int usb_autoresume_device(struct usb_device *udev)
1080 return 0;
1081 }
1082
1083 -static inline int usb_set_usb2_hardware_lpm(struct usb_device *udev, int enable)
1084 +static inline int usb_enable_usb2_hardware_lpm(struct usb_device *udev)
1085 +{
1086 + return 0;
1087 +}
1088 +
1089 +static inline int usb_disable_usb2_hardware_lpm(struct usb_device *udev)
1090 {
1091 return 0;
1092 }
1093 diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
1094 index 64cbc2d007c9..c36275754086 100644
1095 --- a/drivers/vfio/vfio_iommu_type1.c
1096 +++ b/drivers/vfio/vfio_iommu_type1.c
1097 @@ -58,12 +58,18 @@ module_param_named(disable_hugepages,
1098 MODULE_PARM_DESC(disable_hugepages,
1099 "Disable VFIO IOMMU support for IOMMU hugepages.");
1100
1101 +static unsigned int dma_entry_limit __read_mostly = U16_MAX;
1102 +module_param_named(dma_entry_limit, dma_entry_limit, uint, 0644);
1103 +MODULE_PARM_DESC(dma_entry_limit,
1104 + "Maximum number of user DMA mappings per container (65535).");
1105 +
1106 struct vfio_iommu {
1107 struct list_head domain_list;
1108 struct vfio_domain *external_domain; /* domain for external user */
1109 struct mutex lock;
1110 struct rb_root dma_list;
1111 struct blocking_notifier_head notifier;
1112 + unsigned int dma_avail;
1113 bool v2;
1114 bool nesting;
1115 };
1116 @@ -836,6 +842,7 @@ static void vfio_remove_dma(struct vfio_iommu *iommu, struct vfio_dma *dma)
1117 vfio_unlink_dma(iommu, dma);
1118 put_task_struct(dma->task);
1119 kfree(dma);
1120 + iommu->dma_avail++;
1121 }
1122
1123 static unsigned long vfio_pgsize_bitmap(struct vfio_iommu *iommu)
1124 @@ -1110,12 +1117,18 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu,
1125 goto out_unlock;
1126 }
1127
1128 + if (!iommu->dma_avail) {
1129 + ret = -ENOSPC;
1130 + goto out_unlock;
1131 + }
1132 +
1133 dma = kzalloc(sizeof(*dma), GFP_KERNEL);
1134 if (!dma) {
1135 ret = -ENOMEM;
1136 goto out_unlock;
1137 }
1138
1139 + iommu->dma_avail--;
1140 dma->iova = iova;
1141 dma->vaddr = vaddr;
1142 dma->prot = prot;
1143 @@ -1612,6 +1625,7 @@ static void *vfio_iommu_type1_open(unsigned long arg)
1144
1145 INIT_LIST_HEAD(&iommu->domain_list);
1146 iommu->dma_list = RB_ROOT;
1147 + iommu->dma_avail = dma_entry_limit;
1148 mutex_init(&iommu->lock);
1149 BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier);
1150
1151 diff --git a/fs/aio.c b/fs/aio.c
1152 index 45d5ef8dd0a8..911e23087dfb 100644
1153 --- a/fs/aio.c
1154 +++ b/fs/aio.c
1155 @@ -161,9 +161,13 @@ struct kioctx {
1156 unsigned id;
1157 };
1158
1159 +/*
1160 + * First field must be the file pointer in all the
1161 + * iocb unions! See also 'struct kiocb' in <linux/fs.h>
1162 + */
1163 struct fsync_iocb {
1164 - struct work_struct work;
1165 struct file *file;
1166 + struct work_struct work;
1167 bool datasync;
1168 };
1169
1170 @@ -171,14 +175,21 @@ struct poll_iocb {
1171 struct file *file;
1172 struct wait_queue_head *head;
1173 __poll_t events;
1174 - bool woken;
1175 + bool done;
1176 bool cancelled;
1177 struct wait_queue_entry wait;
1178 struct work_struct work;
1179 };
1180
1181 +/*
1182 + * NOTE! Each of the iocb union members has the file pointer
1183 + * as the first entry in their struct definition. So you can
1184 + * access the file pointer through any of the sub-structs,
1185 + * or directly as just 'ki_filp' in this struct.
1186 + */
1187 struct aio_kiocb {
1188 union {
1189 + struct file *ki_filp;
1190 struct kiocb rw;
1191 struct fsync_iocb fsync;
1192 struct poll_iocb poll;
1193 @@ -187,8 +198,7 @@ struct aio_kiocb {
1194 struct kioctx *ki_ctx;
1195 kiocb_cancel_fn *ki_cancel;
1196
1197 - struct iocb __user *ki_user_iocb; /* user's aiocb */
1198 - __u64 ki_user_data; /* user's data for completion */
1199 + struct io_event ki_res;
1200
1201 struct list_head ki_list; /* the aio core uses this
1202 * for cancellation */
1203 @@ -902,7 +912,7 @@ static void put_reqs_available(struct kioctx *ctx, unsigned nr)
1204 local_irq_restore(flags);
1205 }
1206
1207 -static bool get_reqs_available(struct kioctx *ctx)
1208 +static bool __get_reqs_available(struct kioctx *ctx)
1209 {
1210 struct kioctx_cpu *kcpu;
1211 bool ret = false;
1212 @@ -994,32 +1004,35 @@ static void user_refill_reqs_available(struct kioctx *ctx)
1213 spin_unlock_irq(&ctx->completion_lock);
1214 }
1215
1216 +static bool get_reqs_available(struct kioctx *ctx)
1217 +{
1218 + if (__get_reqs_available(ctx))
1219 + return true;
1220 + user_refill_reqs_available(ctx);
1221 + return __get_reqs_available(ctx);
1222 +}
1223 +
1224 /* aio_get_req
1225 * Allocate a slot for an aio request.
1226 * Returns NULL if no requests are free.
1227 + *
1228 + * The refcount is initialized to 2 - one for the async op completion,
1229 + * one for the synchronous code that does this.
1230 */
1231 static inline struct aio_kiocb *aio_get_req(struct kioctx *ctx)
1232 {
1233 struct aio_kiocb *req;
1234
1235 - if (!get_reqs_available(ctx)) {
1236 - user_refill_reqs_available(ctx);
1237 - if (!get_reqs_available(ctx))
1238 - return NULL;
1239 - }
1240 -
1241 - req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL|__GFP_ZERO);
1242 + req = kmem_cache_alloc(kiocb_cachep, GFP_KERNEL);
1243 if (unlikely(!req))
1244 - goto out_put;
1245 + return NULL;
1246
1247 percpu_ref_get(&ctx->reqs);
1248 - INIT_LIST_HEAD(&req->ki_list);
1249 - refcount_set(&req->ki_refcnt, 0);
1250 req->ki_ctx = ctx;
1251 + INIT_LIST_HEAD(&req->ki_list);
1252 + refcount_set(&req->ki_refcnt, 2);
1253 + req->ki_eventfd = NULL;
1254 return req;
1255 -out_put:
1256 - put_reqs_available(ctx, 1);
1257 - return NULL;
1258 }
1259
1260 static struct kioctx *lookup_ioctx(unsigned long ctx_id)
1261 @@ -1050,19 +1063,18 @@ out:
1262 return ret;
1263 }
1264
1265 -static inline void iocb_put(struct aio_kiocb *iocb)
1266 +static inline void iocb_destroy(struct aio_kiocb *iocb)
1267 {
1268 - if (refcount_read(&iocb->ki_refcnt) == 0 ||
1269 - refcount_dec_and_test(&iocb->ki_refcnt)) {
1270 - percpu_ref_put(&iocb->ki_ctx->reqs);
1271 - kmem_cache_free(kiocb_cachep, iocb);
1272 - }
1273 + if (iocb->ki_filp)
1274 + fput(iocb->ki_filp);
1275 + percpu_ref_put(&iocb->ki_ctx->reqs);
1276 + kmem_cache_free(kiocb_cachep, iocb);
1277 }
1278
1279 /* aio_complete
1280 * Called when the io request on the given iocb is complete.
1281 */
1282 -static void aio_complete(struct aio_kiocb *iocb, long res, long res2)
1283 +static void aio_complete(struct aio_kiocb *iocb)
1284 {
1285 struct kioctx *ctx = iocb->ki_ctx;
1286 struct aio_ring *ring;
1287 @@ -1086,17 +1098,14 @@ static void aio_complete(struct aio_kiocb *iocb, long res, long res2)
1288 ev_page = kmap_atomic(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
1289 event = ev_page + pos % AIO_EVENTS_PER_PAGE;
1290
1291 - event->obj = (u64)(unsigned long)iocb->ki_user_iocb;
1292 - event->data = iocb->ki_user_data;
1293 - event->res = res;
1294 - event->res2 = res2;
1295 + *event = iocb->ki_res;
1296
1297 kunmap_atomic(ev_page);
1298 flush_dcache_page(ctx->ring_pages[pos / AIO_EVENTS_PER_PAGE]);
1299
1300 - pr_debug("%p[%u]: %p: %p %Lx %lx %lx\n",
1301 - ctx, tail, iocb, iocb->ki_user_iocb, iocb->ki_user_data,
1302 - res, res2);
1303 + pr_debug("%p[%u]: %p: %p %Lx %Lx %Lx\n", ctx, tail, iocb,
1304 + (void __user *)(unsigned long)iocb->ki_res.obj,
1305 + iocb->ki_res.data, iocb->ki_res.res, iocb->ki_res.res2);
1306
1307 /* after flagging the request as done, we
1308 * must never even look at it again
1309 @@ -1138,7 +1147,14 @@ static void aio_complete(struct aio_kiocb *iocb, long res, long res2)
1310
1311 if (waitqueue_active(&ctx->wait))
1312 wake_up(&ctx->wait);
1313 - iocb_put(iocb);
1314 +}
1315 +
1316 +static inline void iocb_put(struct aio_kiocb *iocb)
1317 +{
1318 + if (refcount_dec_and_test(&iocb->ki_refcnt)) {
1319 + aio_complete(iocb);
1320 + iocb_destroy(iocb);
1321 + }
1322 }
1323
1324 /* aio_read_events_ring
1325 @@ -1412,18 +1428,17 @@ static void aio_complete_rw(struct kiocb *kiocb, long res, long res2)
1326 file_end_write(kiocb->ki_filp);
1327 }
1328
1329 - fput(kiocb->ki_filp);
1330 - aio_complete(iocb, res, res2);
1331 + iocb->ki_res.res = res;
1332 + iocb->ki_res.res2 = res2;
1333 + iocb_put(iocb);
1334 }
1335
1336 -static int aio_prep_rw(struct kiocb *req, struct iocb *iocb)
1337 +static int aio_prep_rw(struct kiocb *req, const struct iocb *iocb)
1338 {
1339 int ret;
1340
1341 - req->ki_filp = fget(iocb->aio_fildes);
1342 - if (unlikely(!req->ki_filp))
1343 - return -EBADF;
1344 req->ki_complete = aio_complete_rw;
1345 + req->private = NULL;
1346 req->ki_pos = iocb->aio_offset;
1347 req->ki_flags = iocb_flags(req->ki_filp);
1348 if (iocb->aio_flags & IOCB_FLAG_RESFD)
1349 @@ -1438,7 +1453,6 @@ static int aio_prep_rw(struct kiocb *req, struct iocb *iocb)
1350 ret = ioprio_check_cap(iocb->aio_reqprio);
1351 if (ret) {
1352 pr_debug("aio ioprio check cap error: %d\n", ret);
1353 - fput(req->ki_filp);
1354 return ret;
1355 }
1356
1357 @@ -1448,11 +1462,13 @@ static int aio_prep_rw(struct kiocb *req, struct iocb *iocb)
1358
1359 ret = kiocb_set_rw_flags(req, iocb->aio_rw_flags);
1360 if (unlikely(ret))
1361 - fput(req->ki_filp);
1362 - return ret;
1363 + return ret;
1364 +
1365 + req->ki_flags &= ~IOCB_HIPRI; /* no one is going to poll for this I/O */
1366 + return 0;
1367 }
1368
1369 -static int aio_setup_rw(int rw, struct iocb *iocb, struct iovec **iovec,
1370 +static int aio_setup_rw(int rw, const struct iocb *iocb, struct iovec **iovec,
1371 bool vectored, bool compat, struct iov_iter *iter)
1372 {
1373 void __user *buf = (void __user *)(uintptr_t)iocb->aio_buf;
1374 @@ -1487,12 +1503,12 @@ static inline void aio_rw_done(struct kiocb *req, ssize_t ret)
1375 ret = -EINTR;
1376 /*FALLTHRU*/
1377 default:
1378 - aio_complete_rw(req, ret, 0);
1379 + req->ki_complete(req, ret, 0);
1380 }
1381 }
1382
1383 -static ssize_t aio_read(struct kiocb *req, struct iocb *iocb, bool vectored,
1384 - bool compat)
1385 +static ssize_t aio_read(struct kiocb *req, const struct iocb *iocb,
1386 + bool vectored, bool compat)
1387 {
1388 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1389 struct iov_iter iter;
1390 @@ -1503,29 +1519,24 @@ static ssize_t aio_read(struct kiocb *req, struct iocb *iocb, bool vectored,
1391 if (ret)
1392 return ret;
1393 file = req->ki_filp;
1394 -
1395 - ret = -EBADF;
1396 if (unlikely(!(file->f_mode & FMODE_READ)))
1397 - goto out_fput;
1398 + return -EBADF;
1399 ret = -EINVAL;
1400 if (unlikely(!file->f_op->read_iter))
1401 - goto out_fput;
1402 + return -EINVAL;
1403
1404 ret = aio_setup_rw(READ, iocb, &iovec, vectored, compat, &iter);
1405 if (ret)
1406 - goto out_fput;
1407 + return ret;
1408 ret = rw_verify_area(READ, file, &req->ki_pos, iov_iter_count(&iter));
1409 if (!ret)
1410 aio_rw_done(req, call_read_iter(file, req, &iter));
1411 kfree(iovec);
1412 -out_fput:
1413 - if (unlikely(ret))
1414 - fput(file);
1415 return ret;
1416 }
1417
1418 -static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored,
1419 - bool compat)
1420 +static ssize_t aio_write(struct kiocb *req, const struct iocb *iocb,
1421 + bool vectored, bool compat)
1422 {
1423 struct iovec inline_vecs[UIO_FASTIOV], *iovec = inline_vecs;
1424 struct iov_iter iter;
1425 @@ -1537,16 +1548,14 @@ static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored,
1426 return ret;
1427 file = req->ki_filp;
1428
1429 - ret = -EBADF;
1430 if (unlikely(!(file->f_mode & FMODE_WRITE)))
1431 - goto out_fput;
1432 - ret = -EINVAL;
1433 + return -EBADF;
1434 if (unlikely(!file->f_op->write_iter))
1435 - goto out_fput;
1436 + return -EINVAL;
1437
1438 ret = aio_setup_rw(WRITE, iocb, &iovec, vectored, compat, &iter);
1439 if (ret)
1440 - goto out_fput;
1441 + return ret;
1442 ret = rw_verify_area(WRITE, file, &req->ki_pos, iov_iter_count(&iter));
1443 if (!ret) {
1444 /*
1445 @@ -1564,35 +1573,26 @@ static ssize_t aio_write(struct kiocb *req, struct iocb *iocb, bool vectored,
1446 aio_rw_done(req, call_write_iter(file, req, &iter));
1447 }
1448 kfree(iovec);
1449 -out_fput:
1450 - if (unlikely(ret))
1451 - fput(file);
1452 return ret;
1453 }
1454
1455 static void aio_fsync_work(struct work_struct *work)
1456 {
1457 - struct fsync_iocb *req = container_of(work, struct fsync_iocb, work);
1458 - int ret;
1459 + struct aio_kiocb *iocb = container_of(work, struct aio_kiocb, fsync.work);
1460
1461 - ret = vfs_fsync(req->file, req->datasync);
1462 - fput(req->file);
1463 - aio_complete(container_of(req, struct aio_kiocb, fsync), ret, 0);
1464 + iocb->ki_res.res = vfs_fsync(iocb->fsync.file, iocb->fsync.datasync);
1465 + iocb_put(iocb);
1466 }
1467
1468 -static int aio_fsync(struct fsync_iocb *req, struct iocb *iocb, bool datasync)
1469 +static int aio_fsync(struct fsync_iocb *req, const struct iocb *iocb,
1470 + bool datasync)
1471 {
1472 if (unlikely(iocb->aio_buf || iocb->aio_offset || iocb->aio_nbytes ||
1473 iocb->aio_rw_flags))
1474 return -EINVAL;
1475
1476 - req->file = fget(iocb->aio_fildes);
1477 - if (unlikely(!req->file))
1478 - return -EBADF;
1479 - if (unlikely(!req->file->f_op->fsync)) {
1480 - fput(req->file);
1481 + if (unlikely(!req->file->f_op->fsync))
1482 return -EINVAL;
1483 - }
1484
1485 req->datasync = datasync;
1486 INIT_WORK(&req->work, aio_fsync_work);
1487 @@ -1600,14 +1600,6 @@ static int aio_fsync(struct fsync_iocb *req, struct iocb *iocb, bool datasync)
1488 return 0;
1489 }
1490
1491 -static inline void aio_poll_complete(struct aio_kiocb *iocb, __poll_t mask)
1492 -{
1493 - struct file *file = iocb->poll.file;
1494 -
1495 - aio_complete(iocb, mangle_poll(mask), 0);
1496 - fput(file);
1497 -}
1498 -
1499 static void aio_poll_complete_work(struct work_struct *work)
1500 {
1501 struct poll_iocb *req = container_of(work, struct poll_iocb, work);
1502 @@ -1633,9 +1625,11 @@ static void aio_poll_complete_work(struct work_struct *work)
1503 return;
1504 }
1505 list_del_init(&iocb->ki_list);
1506 + iocb->ki_res.res = mangle_poll(mask);
1507 + req->done = true;
1508 spin_unlock_irq(&ctx->ctx_lock);
1509
1510 - aio_poll_complete(iocb, mask);
1511 + iocb_put(iocb);
1512 }
1513
1514 /* assumes we are called with irqs disabled */
1515 @@ -1663,31 +1657,27 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
1516 __poll_t mask = key_to_poll(key);
1517 unsigned long flags;
1518
1519 - req->woken = true;
1520 -
1521 /* for instances that support it check for an event match first: */
1522 - if (mask) {
1523 - if (!(mask & req->events))
1524 - return 0;
1525 + if (mask && !(mask & req->events))
1526 + return 0;
1527 +
1528 + list_del_init(&req->wait.entry);
1529
1530 + if (mask && spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) {
1531 /*
1532 * Try to complete the iocb inline if we can. Use
1533 * irqsave/irqrestore because not all filesystems (e.g. fuse)
1534 * call this function with IRQs disabled and because IRQs
1535 * have to be disabled before ctx_lock is obtained.
1536 */
1537 - if (spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) {
1538 - list_del(&iocb->ki_list);
1539 - spin_unlock_irqrestore(&iocb->ki_ctx->ctx_lock, flags);
1540 -
1541 - list_del_init(&req->wait.entry);
1542 - aio_poll_complete(iocb, mask);
1543 - return 1;
1544 - }
1545 + list_del(&iocb->ki_list);
1546 + iocb->ki_res.res = mangle_poll(mask);
1547 + req->done = true;
1548 + spin_unlock_irqrestore(&iocb->ki_ctx->ctx_lock, flags);
1549 + iocb_put(iocb);
1550 + } else {
1551 + schedule_work(&req->work);
1552 }
1553 -
1554 - list_del_init(&req->wait.entry);
1555 - schedule_work(&req->work);
1556 return 1;
1557 }
1558
1559 @@ -1714,11 +1704,12 @@ aio_poll_queue_proc(struct file *file, struct wait_queue_head *head,
1560 add_wait_queue(head, &pt->iocb->poll.wait);
1561 }
1562
1563 -static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb)
1564 +static ssize_t aio_poll(struct aio_kiocb *aiocb, const struct iocb *iocb)
1565 {
1566 struct kioctx *ctx = aiocb->ki_ctx;
1567 struct poll_iocb *req = &aiocb->poll;
1568 struct aio_poll_table apt;
1569 + bool cancel = false;
1570 __poll_t mask;
1571
1572 /* reject any unknown events outside the normal event mask. */
1573 @@ -1730,9 +1721,10 @@ static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb)
1574
1575 INIT_WORK(&req->work, aio_poll_complete_work);
1576 req->events = demangle_poll(iocb->aio_buf) | EPOLLERR | EPOLLHUP;
1577 - req->file = fget(iocb->aio_fildes);
1578 - if (unlikely(!req->file))
1579 - return -EBADF;
1580 +
1581 + req->head = NULL;
1582 + req->done = false;
1583 + req->cancelled = false;
1584
1585 apt.pt._qproc = aio_poll_queue_proc;
1586 apt.pt._key = req->events;
1587 @@ -1743,83 +1735,79 @@ static ssize_t aio_poll(struct aio_kiocb *aiocb, struct iocb *iocb)
1588 INIT_LIST_HEAD(&req->wait.entry);
1589 init_waitqueue_func_entry(&req->wait, aio_poll_wake);
1590
1591 - /* one for removal from waitqueue, one for this function */
1592 - refcount_set(&aiocb->ki_refcnt, 2);
1593 -
1594 mask = vfs_poll(req->file, &apt.pt) & req->events;
1595 - if (unlikely(!req->head)) {
1596 - /* we did not manage to set up a waitqueue, done */
1597 - goto out;
1598 - }
1599 -
1600 spin_lock_irq(&ctx->ctx_lock);
1601 - spin_lock(&req->head->lock);
1602 - if (req->woken) {
1603 - /* wake_up context handles the rest */
1604 - mask = 0;
1605 + if (likely(req->head)) {
1606 + spin_lock(&req->head->lock);
1607 + if (unlikely(list_empty(&req->wait.entry))) {
1608 + if (apt.error)
1609 + cancel = true;
1610 + apt.error = 0;
1611 + mask = 0;
1612 + }
1613 + if (mask || apt.error) {
1614 + list_del_init(&req->wait.entry);
1615 + } else if (cancel) {
1616 + WRITE_ONCE(req->cancelled, true);
1617 + } else if (!req->done) { /* actually waiting for an event */
1618 + list_add_tail(&aiocb->ki_list, &ctx->active_reqs);
1619 + aiocb->ki_cancel = aio_poll_cancel;
1620 + }
1621 + spin_unlock(&req->head->lock);
1622 + }
1623 + if (mask) { /* no async, we'd stolen it */
1624 + aiocb->ki_res.res = mangle_poll(mask);
1625 apt.error = 0;
1626 - } else if (mask || apt.error) {
1627 - /* if we get an error or a mask we are done */
1628 - WARN_ON_ONCE(list_empty(&req->wait.entry));
1629 - list_del_init(&req->wait.entry);
1630 - } else {
1631 - /* actually waiting for an event */
1632 - list_add_tail(&aiocb->ki_list, &ctx->active_reqs);
1633 - aiocb->ki_cancel = aio_poll_cancel;
1634 }
1635 - spin_unlock(&req->head->lock);
1636 spin_unlock_irq(&ctx->ctx_lock);
1637 -
1638 -out:
1639 - if (unlikely(apt.error)) {
1640 - fput(req->file);
1641 - return apt.error;
1642 - }
1643 -
1644 if (mask)
1645 - aio_poll_complete(aiocb, mask);
1646 - iocb_put(aiocb);
1647 - return 0;
1648 + iocb_put(aiocb);
1649 + return apt.error;
1650 }
1651
1652 -static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
1653 - bool compat)
1654 +static int __io_submit_one(struct kioctx *ctx, const struct iocb *iocb,
1655 + struct iocb __user *user_iocb, bool compat)
1656 {
1657 struct aio_kiocb *req;
1658 - struct iocb iocb;
1659 ssize_t ret;
1660
1661 - if (unlikely(copy_from_user(&iocb, user_iocb, sizeof(iocb))))
1662 - return -EFAULT;
1663 -
1664 /* enforce forwards compatibility on users */
1665 - if (unlikely(iocb.aio_reserved2)) {
1666 + if (unlikely(iocb->aio_reserved2)) {
1667 pr_debug("EINVAL: reserve field set\n");
1668 return -EINVAL;
1669 }
1670
1671 /* prevent overflows */
1672 if (unlikely(
1673 - (iocb.aio_buf != (unsigned long)iocb.aio_buf) ||
1674 - (iocb.aio_nbytes != (size_t)iocb.aio_nbytes) ||
1675 - ((ssize_t)iocb.aio_nbytes < 0)
1676 + (iocb->aio_buf != (unsigned long)iocb->aio_buf) ||
1677 + (iocb->aio_nbytes != (size_t)iocb->aio_nbytes) ||
1678 + ((ssize_t)iocb->aio_nbytes < 0)
1679 )) {
1680 pr_debug("EINVAL: overflow check\n");
1681 return -EINVAL;
1682 }
1683
1684 + if (!get_reqs_available(ctx))
1685 + return -EAGAIN;
1686 +
1687 + ret = -EAGAIN;
1688 req = aio_get_req(ctx);
1689 if (unlikely(!req))
1690 - return -EAGAIN;
1691 + goto out_put_reqs_available;
1692 +
1693 + req->ki_filp = fget(iocb->aio_fildes);
1694 + ret = -EBADF;
1695 + if (unlikely(!req->ki_filp))
1696 + goto out_put_req;
1697
1698 - if (iocb.aio_flags & IOCB_FLAG_RESFD) {
1699 + if (iocb->aio_flags & IOCB_FLAG_RESFD) {
1700 /*
1701 * If the IOCB_FLAG_RESFD flag of aio_flags is set, get an
1702 * instance of the file* now. The file descriptor must be
1703 * an eventfd() fd, and will be signaled for each completed
1704 * event using the eventfd_signal() function.
1705 */
1706 - req->ki_eventfd = eventfd_ctx_fdget((int) iocb.aio_resfd);
1707 + req->ki_eventfd = eventfd_ctx_fdget((int) iocb->aio_resfd);
1708 if (IS_ERR(req->ki_eventfd)) {
1709 ret = PTR_ERR(req->ki_eventfd);
1710 req->ki_eventfd = NULL;
1711 @@ -1833,54 +1821,70 @@ static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
1712 goto out_put_req;
1713 }
1714
1715 - req->ki_user_iocb = user_iocb;
1716 - req->ki_user_data = iocb.aio_data;
1717 + req->ki_res.obj = (u64)(unsigned long)user_iocb;
1718 + req->ki_res.data = iocb->aio_data;
1719 + req->ki_res.res = 0;
1720 + req->ki_res.res2 = 0;
1721
1722 - switch (iocb.aio_lio_opcode) {
1723 + switch (iocb->aio_lio_opcode) {
1724 case IOCB_CMD_PREAD:
1725 - ret = aio_read(&req->rw, &iocb, false, compat);
1726 + ret = aio_read(&req->rw, iocb, false, compat);
1727 break;
1728 case IOCB_CMD_PWRITE:
1729 - ret = aio_write(&req->rw, &iocb, false, compat);
1730 + ret = aio_write(&req->rw, iocb, false, compat);
1731 break;
1732 case IOCB_CMD_PREADV:
1733 - ret = aio_read(&req->rw, &iocb, true, compat);
1734 + ret = aio_read(&req->rw, iocb, true, compat);
1735 break;
1736 case IOCB_CMD_PWRITEV:
1737 - ret = aio_write(&req->rw, &iocb, true, compat);
1738 + ret = aio_write(&req->rw, iocb, true, compat);
1739 break;
1740 case IOCB_CMD_FSYNC:
1741 - ret = aio_fsync(&req->fsync, &iocb, false);
1742 + ret = aio_fsync(&req->fsync, iocb, false);
1743 break;
1744 case IOCB_CMD_FDSYNC:
1745 - ret = aio_fsync(&req->fsync, &iocb, true);
1746 + ret = aio_fsync(&req->fsync, iocb, true);
1747 break;
1748 case IOCB_CMD_POLL:
1749 - ret = aio_poll(req, &iocb);
1750 + ret = aio_poll(req, iocb);
1751 break;
1752 default:
1753 - pr_debug("invalid aio operation %d\n", iocb.aio_lio_opcode);
1754 + pr_debug("invalid aio operation %d\n", iocb->aio_lio_opcode);
1755 ret = -EINVAL;
1756 break;
1757 }
1758
1759 + /* Done with the synchronous reference */
1760 + iocb_put(req);
1761 +
1762 /*
1763 * If ret is 0, we'd either done aio_complete() ourselves or have
1764 * arranged for that to be done asynchronously. Anything non-zero
1765 * means that we need to destroy req ourselves.
1766 */
1767 - if (ret)
1768 - goto out_put_req;
1769 - return 0;
1770 + if (!ret)
1771 + return 0;
1772 +
1773 out_put_req:
1774 - put_reqs_available(ctx, 1);
1775 - percpu_ref_put(&ctx->reqs);
1776 if (req->ki_eventfd)
1777 eventfd_ctx_put(req->ki_eventfd);
1778 - kmem_cache_free(kiocb_cachep, req);
1779 + iocb_destroy(req);
1780 +out_put_reqs_available:
1781 + put_reqs_available(ctx, 1);
1782 return ret;
1783 }
1784
1785 +static int io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb,
1786 + bool compat)
1787 +{
1788 + struct iocb iocb;
1789 +
1790 + if (unlikely(copy_from_user(&iocb, user_iocb, sizeof(iocb))))
1791 + return -EFAULT;
1792 +
1793 + return __io_submit_one(ctx, &iocb, user_iocb, compat);
1794 +}
1795 +
1796 /* sys_io_submit:
1797 * Queue the nr iocbs pointed to by iocbpp for processing. Returns
1798 * the number of iocbs queued. May return -EINVAL if the aio_context
1799 @@ -1973,24 +1977,6 @@ COMPAT_SYSCALL_DEFINE3(io_submit, compat_aio_context_t, ctx_id,
1800 }
1801 #endif
1802
1803 -/* lookup_kiocb
1804 - * Finds a given iocb for cancellation.
1805 - */
1806 -static struct aio_kiocb *
1807 -lookup_kiocb(struct kioctx *ctx, struct iocb __user *iocb)
1808 -{
1809 - struct aio_kiocb *kiocb;
1810 -
1811 - assert_spin_locked(&ctx->ctx_lock);
1812 -
1813 - /* TODO: use a hash or array, this sucks. */
1814 - list_for_each_entry(kiocb, &ctx->active_reqs, ki_list) {
1815 - if (kiocb->ki_user_iocb == iocb)
1816 - return kiocb;
1817 - }
1818 - return NULL;
1819 -}
1820 -
1821 /* sys_io_cancel:
1822 * Attempts to cancel an iocb previously passed to io_submit. If
1823 * the operation is successfully cancelled, the resulting event is
1824 @@ -2008,6 +1994,7 @@ SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb,
1825 struct aio_kiocb *kiocb;
1826 int ret = -EINVAL;
1827 u32 key;
1828 + u64 obj = (u64)(unsigned long)iocb;
1829
1830 if (unlikely(get_user(key, &iocb->aio_key)))
1831 return -EFAULT;
1832 @@ -2019,10 +2006,13 @@ SYSCALL_DEFINE3(io_cancel, aio_context_t, ctx_id, struct iocb __user *, iocb,
1833 return -EINVAL;
1834
1835 spin_lock_irq(&ctx->ctx_lock);
1836 - kiocb = lookup_kiocb(ctx, iocb);
1837 - if (kiocb) {
1838 - ret = kiocb->ki_cancel(&kiocb->rw);
1839 - list_del_init(&kiocb->ki_list);
1840 + /* TODO: use a hash or array, this sucks. */
1841 + list_for_each_entry(kiocb, &ctx->active_reqs, ki_list) {
1842 + if (kiocb->ki_res.obj == obj) {
1843 + ret = kiocb->ki_cancel(&kiocb->rw);
1844 + list_del_init(&kiocb->ki_list);
1845 + break;
1846 + }
1847 }
1848 spin_unlock_irq(&ctx->ctx_lock);
1849
1850 diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
1851 index 82928cea0209..7f3f64ba464f 100644
1852 --- a/fs/ceph/dir.c
1853 +++ b/fs/ceph/dir.c
1854 @@ -1470,6 +1470,7 @@ void ceph_dentry_lru_del(struct dentry *dn)
1855 unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn)
1856 {
1857 struct ceph_inode_info *dci = ceph_inode(dir);
1858 + unsigned hash;
1859
1860 switch (dci->i_dir_layout.dl_dir_hash) {
1861 case 0: /* for backward compat */
1862 @@ -1477,8 +1478,11 @@ unsigned ceph_dentry_hash(struct inode *dir, struct dentry *dn)
1863 return dn->d_name.hash;
1864
1865 default:
1866 - return ceph_str_hash(dci->i_dir_layout.dl_dir_hash,
1867 + spin_lock(&dn->d_lock);
1868 + hash = ceph_str_hash(dci->i_dir_layout.dl_dir_hash,
1869 dn->d_name.name, dn->d_name.len);
1870 + spin_unlock(&dn->d_lock);
1871 + return hash;
1872 }
1873 }
1874
1875 diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
1876 index bc43c822426a..bfcf11c70bfa 100644
1877 --- a/fs/ceph/mds_client.c
1878 +++ b/fs/ceph/mds_client.c
1879 @@ -1290,6 +1290,15 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1880 list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
1881 ci->i_prealloc_cap_flush = NULL;
1882 }
1883 +
1884 + if (drop &&
1885 + ci->i_wrbuffer_ref_head == 0 &&
1886 + ci->i_wr_ref == 0 &&
1887 + ci->i_dirty_caps == 0 &&
1888 + ci->i_flushing_caps == 0) {
1889 + ceph_put_snap_context(ci->i_head_snapc);
1890 + ci->i_head_snapc = NULL;
1891 + }
1892 }
1893 spin_unlock(&ci->i_ceph_lock);
1894 while (!list_empty(&to_remove)) {
1895 @@ -1945,10 +1954,39 @@ retry:
1896 return path;
1897 }
1898
1899 +/* Duplicate the dentry->d_name.name safely */
1900 +static int clone_dentry_name(struct dentry *dentry, const char **ppath,
1901 + int *ppathlen)
1902 +{
1903 + u32 len;
1904 + char *name;
1905 +
1906 +retry:
1907 + len = READ_ONCE(dentry->d_name.len);
1908 + name = kmalloc(len + 1, GFP_NOFS);
1909 + if (!name)
1910 + return -ENOMEM;
1911 +
1912 + spin_lock(&dentry->d_lock);
1913 + if (dentry->d_name.len != len) {
1914 + spin_unlock(&dentry->d_lock);
1915 + kfree(name);
1916 + goto retry;
1917 + }
1918 + memcpy(name, dentry->d_name.name, len);
1919 + spin_unlock(&dentry->d_lock);
1920 +
1921 + name[len] = '\0';
1922 + *ppath = name;
1923 + *ppathlen = len;
1924 + return 0;
1925 +}
1926 +
1927 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
1928 const char **ppath, int *ppathlen, u64 *pino,
1929 - int *pfreepath)
1930 + bool *pfreepath, bool parent_locked)
1931 {
1932 + int ret;
1933 char *path;
1934
1935 rcu_read_lock();
1936 @@ -1957,8 +1995,15 @@ static int build_dentry_path(struct dentry *dentry, struct inode *dir,
1937 if (dir && ceph_snap(dir) == CEPH_NOSNAP) {
1938 *pino = ceph_ino(dir);
1939 rcu_read_unlock();
1940 - *ppath = dentry->d_name.name;
1941 - *ppathlen = dentry->d_name.len;
1942 + if (parent_locked) {
1943 + *ppath = dentry->d_name.name;
1944 + *ppathlen = dentry->d_name.len;
1945 + } else {
1946 + ret = clone_dentry_name(dentry, ppath, ppathlen);
1947 + if (ret)
1948 + return ret;
1949 + *pfreepath = true;
1950 + }
1951 return 0;
1952 }
1953 rcu_read_unlock();
1954 @@ -1966,13 +2011,13 @@ static int build_dentry_path(struct dentry *dentry, struct inode *dir,
1955 if (IS_ERR(path))
1956 return PTR_ERR(path);
1957 *ppath = path;
1958 - *pfreepath = 1;
1959 + *pfreepath = true;
1960 return 0;
1961 }
1962
1963 static int build_inode_path(struct inode *inode,
1964 const char **ppath, int *ppathlen, u64 *pino,
1965 - int *pfreepath)
1966 + bool *pfreepath)
1967 {
1968 struct dentry *dentry;
1969 char *path;
1970 @@ -1988,7 +2033,7 @@ static int build_inode_path(struct inode *inode,
1971 if (IS_ERR(path))
1972 return PTR_ERR(path);
1973 *ppath = path;
1974 - *pfreepath = 1;
1975 + *pfreepath = true;
1976 return 0;
1977 }
1978
1979 @@ -1999,7 +2044,7 @@ static int build_inode_path(struct inode *inode,
1980 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
1981 struct inode *rdiri, const char *rpath,
1982 u64 rino, const char **ppath, int *pathlen,
1983 - u64 *ino, int *freepath)
1984 + u64 *ino, bool *freepath, bool parent_locked)
1985 {
1986 int r = 0;
1987
1988 @@ -2009,7 +2054,7 @@ static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
1989 ceph_snap(rinode));
1990 } else if (rdentry) {
1991 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
1992 - freepath);
1993 + freepath, parent_locked);
1994 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
1995 *ppath);
1996 } else if (rpath || rino) {
1997 @@ -2035,7 +2080,7 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
1998 const char *path2 = NULL;
1999 u64 ino1 = 0, ino2 = 0;
2000 int pathlen1 = 0, pathlen2 = 0;
2001 - int freepath1 = 0, freepath2 = 0;
2002 + bool freepath1 = false, freepath2 = false;
2003 int len;
2004 u16 releases;
2005 void *p, *end;
2006 @@ -2043,16 +2088,19 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
2007
2008 ret = set_request_path_attr(req->r_inode, req->r_dentry,
2009 req->r_parent, req->r_path1, req->r_ino1.ino,
2010 - &path1, &pathlen1, &ino1, &freepath1);
2011 + &path1, &pathlen1, &ino1, &freepath1,
2012 + test_bit(CEPH_MDS_R_PARENT_LOCKED,
2013 + &req->r_req_flags));
2014 if (ret < 0) {
2015 msg = ERR_PTR(ret);
2016 goto out;
2017 }
2018
2019 + /* If r_old_dentry is set, then assume that its parent is locked */
2020 ret = set_request_path_attr(NULL, req->r_old_dentry,
2021 req->r_old_dentry_dir,
2022 req->r_path2, req->r_ino2.ino,
2023 - &path2, &pathlen2, &ino2, &freepath2);
2024 + &path2, &pathlen2, &ino2, &freepath2, true);
2025 if (ret < 0) {
2026 msg = ERR_PTR(ret);
2027 goto out_free1;
2028 diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
2029 index f74193da0e09..1f46b02f7314 100644
2030 --- a/fs/ceph/snap.c
2031 +++ b/fs/ceph/snap.c
2032 @@ -568,7 +568,12 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
2033 old_snapc = NULL;
2034
2035 update_snapc:
2036 - if (ci->i_head_snapc) {
2037 + if (ci->i_wrbuffer_ref_head == 0 &&
2038 + ci->i_wr_ref == 0 &&
2039 + ci->i_dirty_caps == 0 &&
2040 + ci->i_flushing_caps == 0) {
2041 + ci->i_head_snapc = NULL;
2042 + } else {
2043 ci->i_head_snapc = ceph_get_snap_context(new_snapc);
2044 dout(" new snapc is %p\n", new_snapc);
2045 }
2046 diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
2047 index b59ebed4f615..1fadd314ae7f 100644
2048 --- a/fs/cifs/inode.c
2049 +++ b/fs/cifs/inode.c
2050 @@ -1735,6 +1735,10 @@ cifs_do_rename(const unsigned int xid, struct dentry *from_dentry,
2051 if (rc == 0 || rc != -EBUSY)
2052 goto do_rename_exit;
2053
2054 + /* Don't fall back to using SMB on SMB 2+ mount */
2055 + if (server->vals->protocol_id != 0)
2056 + goto do_rename_exit;
2057 +
2058 /* open-file renames don't work across directories */
2059 if (to_dentry->d_parent != from_dentry->d_parent)
2060 goto do_rename_exit;
2061 diff --git a/fs/cifs/smb2pdu.c b/fs/cifs/smb2pdu.c
2062 index c6fd3acc5560..33afb637e6f8 100644
2063 --- a/fs/cifs/smb2pdu.c
2064 +++ b/fs/cifs/smb2pdu.c
2065 @@ -3285,6 +3285,7 @@ SMB2_read(const unsigned int xid, struct cifs_io_parms *io_parms,
2066 rc);
2067 }
2068 free_rsp_buf(resp_buftype, rsp_iov.iov_base);
2069 + cifs_small_buf_release(req);
2070 return rc == -ENODATA ? 0 : rc;
2071 } else
2072 trace_smb3_read_done(xid, req->PersistentFileId,
2073 diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
2074 index c0ba5206cd9d..006c277dc22e 100644
2075 --- a/fs/ext4/xattr.c
2076 +++ b/fs/ext4/xattr.c
2077 @@ -829,6 +829,7 @@ int ext4_get_inode_usage(struct inode *inode, qsize_t *usage)
2078 bh = ext4_sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl, REQ_PRIO);
2079 if (IS_ERR(bh)) {
2080 ret = PTR_ERR(bh);
2081 + bh = NULL;
2082 goto out;
2083 }
2084
2085 @@ -2907,6 +2908,7 @@ int ext4_xattr_delete_inode(handle_t *handle, struct inode *inode,
2086 if (error == -EIO)
2087 EXT4_ERROR_INODE(inode, "block %llu read error",
2088 EXT4_I(inode)->i_file_acl);
2089 + bh = NULL;
2090 goto cleanup;
2091 }
2092 error = ext4_xattr_check_block(inode, bh);
2093 @@ -3063,6 +3065,7 @@ ext4_xattr_block_cache_find(struct inode *inode,
2094 if (IS_ERR(bh)) {
2095 if (PTR_ERR(bh) == -ENOMEM)
2096 return NULL;
2097 + bh = NULL;
2098 EXT4_ERROR_INODE(inode, "block %lu read error",
2099 (unsigned long)ce->e_value);
2100 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
2101 diff --git a/fs/nfs/super.c b/fs/nfs/super.c
2102 index 6b666d187907..6df9b85caf20 100644
2103 --- a/fs/nfs/super.c
2104 +++ b/fs/nfs/super.c
2105 @@ -2052,7 +2052,8 @@ static int nfs23_validate_mount_data(void *options,
2106 memcpy(sap, &data->addr, sizeof(data->addr));
2107 args->nfs_server.addrlen = sizeof(data->addr);
2108 args->nfs_server.port = ntohs(data->addr.sin_port);
2109 - if (!nfs_verify_server_address(sap))
2110 + if (sap->sa_family != AF_INET ||
2111 + !nfs_verify_server_address(sap))
2112 goto out_no_address;
2113
2114 if (!(data->flags & NFS_MOUNT_TCP))
2115 diff --git a/fs/nfsd/nfs4callback.c b/fs/nfsd/nfs4callback.c
2116 index 601bf33c26a0..ebbb0285addb 100644
2117 --- a/fs/nfsd/nfs4callback.c
2118 +++ b/fs/nfsd/nfs4callback.c
2119 @@ -926,8 +926,9 @@ static void nfsd4_cb_prepare(struct rpc_task *task, void *calldata)
2120 cb->cb_seq_status = 1;
2121 cb->cb_status = 0;
2122 if (minorversion) {
2123 - if (!nfsd41_cb_get_slot(clp, task))
2124 + if (!cb->cb_holds_slot && !nfsd41_cb_get_slot(clp, task))
2125 return;
2126 + cb->cb_holds_slot = true;
2127 }
2128 rpc_call_start(task);
2129 }
2130 @@ -954,6 +955,9 @@ static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback
2131 return true;
2132 }
2133
2134 + if (!cb->cb_holds_slot)
2135 + goto need_restart;
2136 +
2137 switch (cb->cb_seq_status) {
2138 case 0:
2139 /*
2140 @@ -992,6 +996,7 @@ static bool nfsd4_cb_sequence_done(struct rpc_task *task, struct nfsd4_callback
2141 cb->cb_seq_status);
2142 }
2143
2144 + cb->cb_holds_slot = false;
2145 clear_bit(0, &clp->cl_cb_slot_busy);
2146 rpc_wake_up_next(&clp->cl_cb_waitq);
2147 dprintk("%s: freed slot, new seqid=%d\n", __func__,
2148 @@ -1199,6 +1204,7 @@ void nfsd4_init_cb(struct nfsd4_callback *cb, struct nfs4_client *clp,
2149 cb->cb_seq_status = 1;
2150 cb->cb_status = 0;
2151 cb->cb_need_restart = false;
2152 + cb->cb_holds_slot = false;
2153 }
2154
2155 void nfsd4_run_cb(struct nfsd4_callback *cb)
2156 diff --git a/fs/nfsd/state.h b/fs/nfsd/state.h
2157 index 0b15dac7e609..0f07ad6dc1ef 100644
2158 --- a/fs/nfsd/state.h
2159 +++ b/fs/nfsd/state.h
2160 @@ -70,6 +70,7 @@ struct nfsd4_callback {
2161 int cb_seq_status;
2162 int cb_status;
2163 bool cb_need_restart;
2164 + bool cb_holds_slot;
2165 };
2166
2167 struct nfsd4_callback_ops {
2168 diff --git a/fs/proc/proc_sysctl.c b/fs/proc/proc_sysctl.c
2169 index d65390727541..7325baa8f9d4 100644
2170 --- a/fs/proc/proc_sysctl.c
2171 +++ b/fs/proc/proc_sysctl.c
2172 @@ -1626,9 +1626,11 @@ static void drop_sysctl_table(struct ctl_table_header *header)
2173 if (--header->nreg)
2174 return;
2175
2176 - if (parent)
2177 + if (parent) {
2178 put_links(header);
2179 - start_unregistering(header);
2180 + start_unregistering(header);
2181 + }
2182 +
2183 if (!--header->count)
2184 kfree_rcu(header, rcu);
2185
2186 diff --git a/fs/splice.c b/fs/splice.c
2187 index 29e92b506394..c78e0e3ff6c4 100644
2188 --- a/fs/splice.c
2189 +++ b/fs/splice.c
2190 @@ -333,8 +333,8 @@ const struct pipe_buf_operations default_pipe_buf_ops = {
2191 .get = generic_pipe_buf_get,
2192 };
2193
2194 -static int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe,
2195 - struct pipe_buffer *buf)
2196 +int generic_pipe_buf_nosteal(struct pipe_inode_info *pipe,
2197 + struct pipe_buffer *buf)
2198 {
2199 return 1;
2200 }
2201 diff --git a/include/linux/fs.h b/include/linux/fs.h
2202 index 7b6084854bfe..111c94c4baa1 100644
2203 --- a/include/linux/fs.h
2204 +++ b/include/linux/fs.h
2205 @@ -304,13 +304,19 @@ enum rw_hint {
2206
2207 struct kiocb {
2208 struct file *ki_filp;
2209 +
2210 + /* The 'ki_filp' pointer is shared in a union for aio */
2211 + randomized_struct_fields_start
2212 +
2213 loff_t ki_pos;
2214 void (*ki_complete)(struct kiocb *iocb, long ret, long ret2);
2215 void *private;
2216 int ki_flags;
2217 u16 ki_hint;
2218 u16 ki_ioprio; /* See linux/ioprio.h */
2219 -} __randomize_layout;
2220 +
2221 + randomized_struct_fields_end
2222 +};
2223
2224 static inline bool is_sync_kiocb(struct kiocb *kiocb)
2225 {
2226 diff --git a/include/linux/pipe_fs_i.h b/include/linux/pipe_fs_i.h
2227 index 3ecd7ea212ae..66ee63cd5968 100644
2228 --- a/include/linux/pipe_fs_i.h
2229 +++ b/include/linux/pipe_fs_i.h
2230 @@ -181,6 +181,7 @@ void free_pipe_info(struct pipe_inode_info *);
2231 void generic_pipe_buf_get(struct pipe_inode_info *, struct pipe_buffer *);
2232 int generic_pipe_buf_confirm(struct pipe_inode_info *, struct pipe_buffer *);
2233 int generic_pipe_buf_steal(struct pipe_inode_info *, struct pipe_buffer *);
2234 +int generic_pipe_buf_nosteal(struct pipe_inode_info *, struct pipe_buffer *);
2235 void generic_pipe_buf_release(struct pipe_inode_info *, struct pipe_buffer *);
2236 void pipe_buf_mark_unmergeable(struct pipe_buffer *buf);
2237
2238 diff --git a/include/net/netfilter/nf_tables.h b/include/net/netfilter/nf_tables.h
2239 index 0f39ac487012..f2be5d041ba3 100644
2240 --- a/include/net/netfilter/nf_tables.h
2241 +++ b/include/net/netfilter/nf_tables.h
2242 @@ -382,6 +382,7 @@ void nft_unregister_set(struct nft_set_type *type);
2243 * @dtype: data type (verdict or numeric type defined by userspace)
2244 * @objtype: object type (see NFT_OBJECT_* definitions)
2245 * @size: maximum set size
2246 + * @use: number of rules references to this set
2247 * @nelems: number of elements
2248 * @ndeact: number of deactivated elements queued for removal
2249 * @timeout: default timeout value in jiffies
2250 @@ -407,6 +408,7 @@ struct nft_set {
2251 u32 dtype;
2252 u32 objtype;
2253 u32 size;
2254 + u32 use;
2255 atomic_t nelems;
2256 u32 ndeact;
2257 u64 timeout;
2258 @@ -416,7 +418,8 @@ struct nft_set {
2259 unsigned char *udata;
2260 /* runtime data below here */
2261 const struct nft_set_ops *ops ____cacheline_aligned;
2262 - u16 flags:14,
2263 + u16 flags:13,
2264 + bound:1,
2265 genmask:2;
2266 u8 klen;
2267 u8 dlen;
2268 @@ -466,10 +469,15 @@ struct nft_set_binding {
2269 u32 flags;
2270 };
2271
2272 +enum nft_trans_phase;
2273 +void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set,
2274 + struct nft_set_binding *binding,
2275 + enum nft_trans_phase phase);
2276 int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
2277 struct nft_set_binding *binding);
2278 void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set,
2279 - struct nft_set_binding *binding);
2280 + struct nft_set_binding *binding, bool commit);
2281 +void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set);
2282
2283 /**
2284 * enum nft_set_extensions - set extension type IDs
2285 @@ -689,10 +697,12 @@ static inline void nft_set_gc_batch_add(struct nft_set_gc_batch *gcb,
2286 gcb->elems[gcb->head.cnt++] = elem;
2287 }
2288
2289 +struct nft_expr_ops;
2290 /**
2291 * struct nft_expr_type - nf_tables expression type
2292 *
2293 * @select_ops: function to select nft_expr_ops
2294 + * @release_ops: release nft_expr_ops
2295 * @ops: default ops, used when no select_ops functions is present
2296 * @list: used internally
2297 * @name: Identifier
2298 @@ -705,6 +715,7 @@ static inline void nft_set_gc_batch_add(struct nft_set_gc_batch *gcb,
2299 struct nft_expr_type {
2300 const struct nft_expr_ops *(*select_ops)(const struct nft_ctx *,
2301 const struct nlattr * const tb[]);
2302 + void (*release_ops)(const struct nft_expr_ops *ops);
2303 const struct nft_expr_ops *ops;
2304 struct list_head list;
2305 const char *name;
2306 @@ -718,13 +729,22 @@ struct nft_expr_type {
2307 #define NFT_EXPR_STATEFUL 0x1
2308 #define NFT_EXPR_GC 0x2
2309
2310 +enum nft_trans_phase {
2311 + NFT_TRANS_PREPARE,
2312 + NFT_TRANS_ABORT,
2313 + NFT_TRANS_COMMIT,
2314 + NFT_TRANS_RELEASE
2315 +};
2316 +
2317 /**
2318 * struct nft_expr_ops - nf_tables expression operations
2319 *
2320 * @eval: Expression evaluation function
2321 * @size: full expression size, including private data size
2322 * @init: initialization function
2323 - * @destroy: destruction function
2324 + * @activate: activate expression in the next generation
2325 + * @deactivate: deactivate expression in next generation
2326 + * @destroy: destruction function, called after synchronize_rcu
2327 * @dump: function to dump parameters
2328 * @type: expression type
2329 * @validate: validate expression, called during loop detection
2330 @@ -745,7 +765,8 @@ struct nft_expr_ops {
2331 void (*activate)(const struct nft_ctx *ctx,
2332 const struct nft_expr *expr);
2333 void (*deactivate)(const struct nft_ctx *ctx,
2334 - const struct nft_expr *expr);
2335 + const struct nft_expr *expr,
2336 + enum nft_trans_phase phase);
2337 void (*destroy)(const struct nft_ctx *ctx,
2338 const struct nft_expr *expr);
2339 void (*destroy_clone)(const struct nft_ctx *ctx,
2340 diff --git a/include/net/netrom.h b/include/net/netrom.h
2341 index 5a0714ff500f..80f15b1c1a48 100644
2342 --- a/include/net/netrom.h
2343 +++ b/include/net/netrom.h
2344 @@ -266,7 +266,7 @@ void nr_stop_idletimer(struct sock *);
2345 int nr_t1timer_running(struct sock *);
2346
2347 /* sysctl_net_netrom.c */
2348 -void nr_register_sysctl(void);
2349 +int nr_register_sysctl(void);
2350 void nr_unregister_sysctl(void);
2351
2352 #endif
2353 diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
2354 index 91e4202b0634..72c07059ef37 100644
2355 --- a/kernel/sched/deadline.c
2356 +++ b/kernel/sched/deadline.c
2357 @@ -252,7 +252,6 @@ static void task_non_contending(struct task_struct *p)
2358 if (dl_entity_is_special(dl_se))
2359 return;
2360
2361 - WARN_ON(hrtimer_active(&dl_se->inactive_timer));
2362 WARN_ON(dl_se->dl_non_contending);
2363
2364 zerolag_time = dl_se->deadline -
2365 @@ -269,7 +268,7 @@ static void task_non_contending(struct task_struct *p)
2366 * If the "0-lag time" already passed, decrease the active
2367 * utilization now, instead of starting a timer
2368 */
2369 - if (zerolag_time < 0) {
2370 + if ((zerolag_time < 0) || hrtimer_active(&dl_se->inactive_timer)) {
2371 if (dl_task(p))
2372 sub_running_bw(dl_se, dl_rq);
2373 if (!dl_task(p) || p->state == TASK_DEAD) {
2374 diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
2375 index 4aa8e7d90c25..d31916366d39 100644
2376 --- a/kernel/sched/fair.c
2377 +++ b/kernel/sched/fair.c
2378 @@ -2016,6 +2016,10 @@ static u64 numa_get_avg_runtime(struct task_struct *p, u64 *period)
2379 if (p->last_task_numa_placement) {
2380 delta = runtime - p->last_sum_exec_runtime;
2381 *period = now - p->last_task_numa_placement;
2382 +
2383 + /* Avoid time going backwards, prevent potential divide error: */
2384 + if (unlikely((s64)*period < 0))
2385 + *period = 0;
2386 } else {
2387 delta = p->se.avg.load_sum;
2388 *period = LOAD_AVG_MAX;
2389 diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
2390 index 34b4c32b0692..805aef83b5cf 100644
2391 --- a/kernel/trace/ring_buffer.c
2392 +++ b/kernel/trace/ring_buffer.c
2393 @@ -730,7 +730,7 @@ u64 ring_buffer_time_stamp(struct ring_buffer *buffer, int cpu)
2394
2395 preempt_disable_notrace();
2396 time = rb_time_stamp(buffer);
2397 - preempt_enable_no_resched_notrace();
2398 + preempt_enable_notrace();
2399
2400 return time;
2401 }
2402 diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
2403 index c65cea71d1ee..5455ee05bc3b 100644
2404 --- a/kernel/trace/trace.c
2405 +++ b/kernel/trace/trace.c
2406 @@ -496,8 +496,10 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
2407 * not modified.
2408 */
2409 pid_list = kmalloc(sizeof(*pid_list), GFP_KERNEL);
2410 - if (!pid_list)
2411 + if (!pid_list) {
2412 + trace_parser_put(&parser);
2413 return -ENOMEM;
2414 + }
2415
2416 pid_list->pid_max = READ_ONCE(pid_max);
2417
2418 @@ -507,6 +509,7 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
2419
2420 pid_list->pids = vzalloc((pid_list->pid_max + 7) >> 3);
2421 if (!pid_list->pids) {
2422 + trace_parser_put(&parser);
2423 kfree(pid_list);
2424 return -ENOMEM;
2425 }
2426 @@ -6800,19 +6803,23 @@ struct buffer_ref {
2427 struct ring_buffer *buffer;
2428 void *page;
2429 int cpu;
2430 - int ref;
2431 + refcount_t refcount;
2432 };
2433
2434 +static void buffer_ref_release(struct buffer_ref *ref)
2435 +{
2436 + if (!refcount_dec_and_test(&ref->refcount))
2437 + return;
2438 + ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
2439 + kfree(ref);
2440 +}
2441 +
2442 static void buffer_pipe_buf_release(struct pipe_inode_info *pipe,
2443 struct pipe_buffer *buf)
2444 {
2445 struct buffer_ref *ref = (struct buffer_ref *)buf->private;
2446
2447 - if (--ref->ref)
2448 - return;
2449 -
2450 - ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
2451 - kfree(ref);
2452 + buffer_ref_release(ref);
2453 buf->private = 0;
2454 }
2455
2456 @@ -6821,7 +6828,7 @@ static void buffer_pipe_buf_get(struct pipe_inode_info *pipe,
2457 {
2458 struct buffer_ref *ref = (struct buffer_ref *)buf->private;
2459
2460 - ref->ref++;
2461 + refcount_inc(&ref->refcount);
2462 }
2463
2464 /* Pipe buffer operations for a buffer. */
2465 @@ -6829,7 +6836,7 @@ static const struct pipe_buf_operations buffer_pipe_buf_ops = {
2466 .can_merge = 0,
2467 .confirm = generic_pipe_buf_confirm,
2468 .release = buffer_pipe_buf_release,
2469 - .steal = generic_pipe_buf_steal,
2470 + .steal = generic_pipe_buf_nosteal,
2471 .get = buffer_pipe_buf_get,
2472 };
2473
2474 @@ -6842,11 +6849,7 @@ static void buffer_spd_release(struct splice_pipe_desc *spd, unsigned int i)
2475 struct buffer_ref *ref =
2476 (struct buffer_ref *)spd->partial[i].private;
2477
2478 - if (--ref->ref)
2479 - return;
2480 -
2481 - ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
2482 - kfree(ref);
2483 + buffer_ref_release(ref);
2484 spd->partial[i].private = 0;
2485 }
2486
2487 @@ -6901,7 +6904,7 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos,
2488 break;
2489 }
2490
2491 - ref->ref = 1;
2492 + refcount_set(&ref->refcount, 1);
2493 ref->buffer = iter->trace_buffer->buffer;
2494 ref->page = ring_buffer_alloc_read_page(ref->buffer, iter->cpu_file);
2495 if (IS_ERR(ref->page)) {
2496 diff --git a/kernel/workqueue.c b/kernel/workqueue.c
2497 index 0280deac392e..cd8b61bded78 100644
2498 --- a/kernel/workqueue.c
2499 +++ b/kernel/workqueue.c
2500 @@ -2908,6 +2908,9 @@ static bool __flush_work(struct work_struct *work, bool from_cancel)
2501 if (WARN_ON(!wq_online))
2502 return false;
2503
2504 + if (WARN_ON(!work->func))
2505 + return false;
2506 +
2507 if (!from_cancel) {
2508 lock_map_acquire(&work->lockdep_map);
2509 lock_map_release(&work->lockdep_map);
2510 diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
2511 index 4966c4fbe7f7..3dea52f7be9c 100644
2512 --- a/lib/Kconfig.debug
2513 +++ b/lib/Kconfig.debug
2514 @@ -1934,6 +1934,7 @@ config TEST_KMOD
2515 depends on m
2516 depends on BLOCK && (64BIT || LBDAF) # for XFS, BTRFS
2517 depends on NETDEVICES && NET_CORE && INET # for TUN
2518 + depends on BLOCK
2519 select TEST_LKM
2520 select XFS_FS
2521 select TUN
2522 diff --git a/mm/memory.c b/mm/memory.c
2523 index 5b3f71bcd1ae..9c69278173b7 100644
2524 --- a/mm/memory.c
2525 +++ b/mm/memory.c
2526 @@ -1787,10 +1787,15 @@ static int insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2527 * in may not match the PFN we have mapped if the
2528 * mapped PFN is a writeable COW page. In the mkwrite
2529 * case we are creating a writable PTE for a shared
2530 - * mapping and we expect the PFNs to match.
2531 + * mapping and we expect the PFNs to match. If they
2532 + * don't match, we are likely racing with block
2533 + * allocation and mapping invalidation so just skip the
2534 + * update.
2535 */
2536 - if (WARN_ON_ONCE(pte_pfn(*pte) != pfn_t_to_pfn(pfn)))
2537 + if (pte_pfn(*pte) != pfn_t_to_pfn(pfn)) {
2538 + WARN_ON_ONCE(!is_zero_pfn(pte_pfn(*pte)));
2539 goto out_unlock;
2540 + }
2541 entry = *pte;
2542 goto out_mkwrite;
2543 } else
2544 diff --git a/net/bridge/netfilter/ebtables.c b/net/bridge/netfilter/ebtables.c
2545 index f77888ec93f1..0bb4d712b80c 100644
2546 --- a/net/bridge/netfilter/ebtables.c
2547 +++ b/net/bridge/netfilter/ebtables.c
2548 @@ -2032,7 +2032,8 @@ static int ebt_size_mwt(struct compat_ebt_entry_mwt *match32,
2549 if (match_kern)
2550 match_kern->match_size = ret;
2551
2552 - if (WARN_ON(type == EBT_COMPAT_TARGET && size_left))
2553 + /* rule should have no remaining data after target */
2554 + if (type == EBT_COMPAT_TARGET && size_left)
2555 return -EINVAL;
2556
2557 match32 = (struct compat_ebt_entry_mwt *) buf;
2558 diff --git a/net/ipv4/route.c b/net/ipv4/route.c
2559 index 98c81c21b753..8bacbcd2db90 100644
2560 --- a/net/ipv4/route.c
2561 +++ b/net/ipv4/route.c
2562 @@ -1185,25 +1185,39 @@ static struct dst_entry *ipv4_dst_check(struct dst_entry *dst, u32 cookie)
2563 return dst;
2564 }
2565
2566 -static void ipv4_link_failure(struct sk_buff *skb)
2567 +static void ipv4_send_dest_unreach(struct sk_buff *skb)
2568 {
2569 struct ip_options opt;
2570 - struct rtable *rt;
2571 int res;
2572
2573 /* Recompile ip options since IPCB may not be valid anymore.
2574 + * Also check we have a reasonable ipv4 header.
2575 */
2576 - memset(&opt, 0, sizeof(opt));
2577 - opt.optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr);
2578 + if (!pskb_network_may_pull(skb, sizeof(struct iphdr)) ||
2579 + ip_hdr(skb)->version != 4 || ip_hdr(skb)->ihl < 5)
2580 + return;
2581
2582 - rcu_read_lock();
2583 - res = __ip_options_compile(dev_net(skb->dev), &opt, skb, NULL);
2584 - rcu_read_unlock();
2585 + memset(&opt, 0, sizeof(opt));
2586 + if (ip_hdr(skb)->ihl > 5) {
2587 + if (!pskb_network_may_pull(skb, ip_hdr(skb)->ihl * 4))
2588 + return;
2589 + opt.optlen = ip_hdr(skb)->ihl * 4 - sizeof(struct iphdr);
2590
2591 - if (res)
2592 - return;
2593 + rcu_read_lock();
2594 + res = __ip_options_compile(dev_net(skb->dev), &opt, skb, NULL);
2595 + rcu_read_unlock();
2596
2597 + if (res)
2598 + return;
2599 + }
2600 __icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_UNREACH, 0, &opt);
2601 +}
2602 +
2603 +static void ipv4_link_failure(struct sk_buff *skb)
2604 +{
2605 + struct rtable *rt;
2606 +
2607 + ipv4_send_dest_unreach(skb);
2608
2609 rt = skb_rtable(skb);
2610 if (rt)
2611 diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
2612 index 891ed2f91467..ce64453d337d 100644
2613 --- a/net/ipv4/sysctl_net_ipv4.c
2614 +++ b/net/ipv4/sysctl_net_ipv4.c
2615 @@ -49,6 +49,7 @@ static int ip_ping_group_range_min[] = { 0, 0 };
2616 static int ip_ping_group_range_max[] = { GID_T_MAX, GID_T_MAX };
2617 static int comp_sack_nr_max = 255;
2618 static u32 u32_max_div_HZ = UINT_MAX / HZ;
2619 +static int one_day_secs = 24 * 3600;
2620
2621 /* obsolete */
2622 static int sysctl_tcp_low_latency __read_mostly;
2623 @@ -1140,7 +1141,9 @@ static struct ctl_table ipv4_net_table[] = {
2624 .data = &init_net.ipv4.sysctl_tcp_min_rtt_wlen,
2625 .maxlen = sizeof(int),
2626 .mode = 0644,
2627 - .proc_handler = proc_dointvec
2628 + .proc_handler = proc_dointvec_minmax,
2629 + .extra1 = &zero,
2630 + .extra2 = &one_day_secs
2631 },
2632 {
2633 .procname = "tcp_autocorking",
2634 diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c
2635 index 8fd8d06454d6..2d4e048762f6 100644
2636 --- a/net/netfilter/ipvs/ip_vs_ctl.c
2637 +++ b/net/netfilter/ipvs/ip_vs_ctl.c
2638 @@ -896,12 +896,13 @@ ip_vs_new_dest(struct ip_vs_service *svc, struct ip_vs_dest_user_kern *udest,
2639 {
2640 struct ip_vs_dest *dest;
2641 unsigned int atype, i;
2642 - int ret = 0;
2643
2644 EnterFunction(2);
2645
2646 #ifdef CONFIG_IP_VS_IPV6
2647 if (udest->af == AF_INET6) {
2648 + int ret;
2649 +
2650 atype = ipv6_addr_type(&udest->addr.in6);
2651 if ((!(atype & IPV6_ADDR_UNICAST) ||
2652 atype & IPV6_ADDR_LINKLOCAL) &&
2653 diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
2654 index c06393fc716d..1af54119bafc 100644
2655 --- a/net/netfilter/nf_tables_api.c
2656 +++ b/net/netfilter/nf_tables_api.c
2657 @@ -112,6 +112,23 @@ static void nft_trans_destroy(struct nft_trans *trans)
2658 kfree(trans);
2659 }
2660
2661 +static void nft_set_trans_bind(const struct nft_ctx *ctx, struct nft_set *set)
2662 +{
2663 + struct net *net = ctx->net;
2664 + struct nft_trans *trans;
2665 +
2666 + if (!nft_set_is_anonymous(set))
2667 + return;
2668 +
2669 + list_for_each_entry_reverse(trans, &net->nft.commit_list, list) {
2670 + if (trans->msg_type == NFT_MSG_NEWSET &&
2671 + nft_trans_set(trans) == set) {
2672 + set->bound = true;
2673 + break;
2674 + }
2675 + }
2676 +}
2677 +
2678 static int nf_tables_register_hook(struct net *net,
2679 const struct nft_table *table,
2680 struct nft_chain *chain)
2681 @@ -222,14 +239,15 @@ static void nft_rule_expr_activate(const struct nft_ctx *ctx,
2682 }
2683
2684 static void nft_rule_expr_deactivate(const struct nft_ctx *ctx,
2685 - struct nft_rule *rule)
2686 + struct nft_rule *rule,
2687 + enum nft_trans_phase phase)
2688 {
2689 struct nft_expr *expr;
2690
2691 expr = nft_expr_first(rule);
2692 while (expr != nft_expr_last(rule) && expr->ops) {
2693 if (expr->ops->deactivate)
2694 - expr->ops->deactivate(ctx, expr);
2695 + expr->ops->deactivate(ctx, expr, phase);
2696
2697 expr = nft_expr_next(expr);
2698 }
2699 @@ -280,7 +298,7 @@ static int nft_delrule(struct nft_ctx *ctx, struct nft_rule *rule)
2700 nft_trans_destroy(trans);
2701 return err;
2702 }
2703 - nft_rule_expr_deactivate(ctx, rule);
2704 + nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_PREPARE);
2705
2706 return 0;
2707 }
2708 @@ -301,7 +319,7 @@ static int nft_delrule_by_chain(struct nft_ctx *ctx)
2709 return 0;
2710 }
2711
2712 -static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type,
2713 +static int nft_trans_set_add(const struct nft_ctx *ctx, int msg_type,
2714 struct nft_set *set)
2715 {
2716 struct nft_trans *trans;
2717 @@ -321,7 +339,7 @@ static int nft_trans_set_add(struct nft_ctx *ctx, int msg_type,
2718 return 0;
2719 }
2720
2721 -static int nft_delset(struct nft_ctx *ctx, struct nft_set *set)
2722 +static int nft_delset(const struct nft_ctx *ctx, struct nft_set *set)
2723 {
2724 int err;
2725
2726 @@ -2105,6 +2123,7 @@ struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
2727 {
2728 struct nft_expr_info info;
2729 struct nft_expr *expr;
2730 + struct module *owner;
2731 int err;
2732
2733 err = nf_tables_expr_parse(ctx, nla, &info);
2734 @@ -2124,7 +2143,11 @@ struct nft_expr *nft_expr_init(const struct nft_ctx *ctx,
2735 err3:
2736 kfree(expr);
2737 err2:
2738 - module_put(info.ops->type->owner);
2739 + owner = info.ops->type->owner;
2740 + if (info.ops->type->release_ops)
2741 + info.ops->type->release_ops(info.ops);
2742 +
2743 + module_put(owner);
2744 err1:
2745 return ERR_PTR(err);
2746 }
2747 @@ -2458,7 +2481,7 @@ static void nf_tables_rule_destroy(const struct nft_ctx *ctx,
2748 static void nf_tables_rule_release(const struct nft_ctx *ctx,
2749 struct nft_rule *rule)
2750 {
2751 - nft_rule_expr_deactivate(ctx, rule);
2752 + nft_rule_expr_deactivate(ctx, rule, NFT_TRANS_RELEASE);
2753 nf_tables_rule_destroy(ctx, rule);
2754 }
2755
2756 @@ -3562,19 +3585,15 @@ err1:
2757
2758 static void nft_set_destroy(struct nft_set *set)
2759 {
2760 + if (WARN_ON(set->use > 0))
2761 + return;
2762 +
2763 set->ops->destroy(set);
2764 module_put(to_set_type(set->ops)->owner);
2765 kfree(set->name);
2766 kvfree(set);
2767 }
2768
2769 -static void nf_tables_set_destroy(const struct nft_ctx *ctx, struct nft_set *set)
2770 -{
2771 - list_del_rcu(&set->list);
2772 - nf_tables_set_notify(ctx, set, NFT_MSG_DELSET, GFP_ATOMIC);
2773 - nft_set_destroy(set);
2774 -}
2775 -
2776 static int nf_tables_delset(struct net *net, struct sock *nlsk,
2777 struct sk_buff *skb, const struct nlmsghdr *nlh,
2778 const struct nlattr * const nla[],
2779 @@ -3609,7 +3628,7 @@ static int nf_tables_delset(struct net *net, struct sock *nlsk,
2780 NL_SET_BAD_ATTR(extack, attr);
2781 return PTR_ERR(set);
2782 }
2783 - if (!list_empty(&set->bindings) ||
2784 + if (set->use ||
2785 (nlh->nlmsg_flags & NLM_F_NONREC && atomic_read(&set->nelems) > 0)) {
2786 NL_SET_BAD_ATTR(extack, attr);
2787 return -EBUSY;
2788 @@ -3639,6 +3658,9 @@ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
2789 struct nft_set_binding *i;
2790 struct nft_set_iter iter;
2791
2792 + if (set->use == UINT_MAX)
2793 + return -EOVERFLOW;
2794 +
2795 if (!list_empty(&set->bindings) && nft_set_is_anonymous(set))
2796 return -EBUSY;
2797
2798 @@ -3665,21 +3687,53 @@ int nf_tables_bind_set(const struct nft_ctx *ctx, struct nft_set *set,
2799 bind:
2800 binding->chain = ctx->chain;
2801 list_add_tail_rcu(&binding->list, &set->bindings);
2802 + nft_set_trans_bind(ctx, set);
2803 + set->use++;
2804 +
2805 return 0;
2806 }
2807 EXPORT_SYMBOL_GPL(nf_tables_bind_set);
2808
2809 void nf_tables_unbind_set(const struct nft_ctx *ctx, struct nft_set *set,
2810 - struct nft_set_binding *binding)
2811 + struct nft_set_binding *binding, bool event)
2812 {
2813 list_del_rcu(&binding->list);
2814
2815 - if (list_empty(&set->bindings) && nft_set_is_anonymous(set) &&
2816 - nft_is_active(ctx->net, set))
2817 - nf_tables_set_destroy(ctx, set);
2818 + if (list_empty(&set->bindings) && nft_set_is_anonymous(set)) {
2819 + list_del_rcu(&set->list);
2820 + if (event)
2821 + nf_tables_set_notify(ctx, set, NFT_MSG_DELSET,
2822 + GFP_KERNEL);
2823 + }
2824 }
2825 EXPORT_SYMBOL_GPL(nf_tables_unbind_set);
2826
2827 +void nf_tables_deactivate_set(const struct nft_ctx *ctx, struct nft_set *set,
2828 + struct nft_set_binding *binding,
2829 + enum nft_trans_phase phase)
2830 +{
2831 + switch (phase) {
2832 + case NFT_TRANS_PREPARE:
2833 + set->use--;
2834 + return;
2835 + case NFT_TRANS_ABORT:
2836 + case NFT_TRANS_RELEASE:
2837 + set->use--;
2838 + /* fall through */
2839 + default:
2840 + nf_tables_unbind_set(ctx, set, binding,
2841 + phase == NFT_TRANS_COMMIT);
2842 + }
2843 +}
2844 +EXPORT_SYMBOL_GPL(nf_tables_deactivate_set);
2845 +
2846 +void nf_tables_destroy_set(const struct nft_ctx *ctx, struct nft_set *set)
2847 +{
2848 + if (list_empty(&set->bindings) && nft_set_is_anonymous(set))
2849 + nft_set_destroy(set);
2850 +}
2851 +EXPORT_SYMBOL_GPL(nf_tables_destroy_set);
2852 +
2853 const struct nft_set_ext_type nft_set_ext_types[] = {
2854 [NFT_SET_EXT_KEY] = {
2855 .align = __alignof__(u32),
2856 @@ -6429,6 +6483,9 @@ static int nf_tables_commit(struct net *net, struct sk_buff *skb)
2857 nf_tables_rule_notify(&trans->ctx,
2858 nft_trans_rule(trans),
2859 NFT_MSG_DELRULE);
2860 + nft_rule_expr_deactivate(&trans->ctx,
2861 + nft_trans_rule(trans),
2862 + NFT_TRANS_COMMIT);
2863 break;
2864 case NFT_MSG_NEWSET:
2865 nft_clear(net, nft_trans_set(trans));
2866 @@ -6577,7 +6634,9 @@ static int __nf_tables_abort(struct net *net)
2867 case NFT_MSG_NEWRULE:
2868 trans->ctx.chain->use--;
2869 list_del_rcu(&nft_trans_rule(trans)->list);
2870 - nft_rule_expr_deactivate(&trans->ctx, nft_trans_rule(trans));
2871 + nft_rule_expr_deactivate(&trans->ctx,
2872 + nft_trans_rule(trans),
2873 + NFT_TRANS_ABORT);
2874 break;
2875 case NFT_MSG_DELRULE:
2876 trans->ctx.chain->use++;
2877 @@ -6587,6 +6646,10 @@ static int __nf_tables_abort(struct net *net)
2878 break;
2879 case NFT_MSG_NEWSET:
2880 trans->ctx.table->use--;
2881 + if (nft_trans_set(trans)->bound) {
2882 + nft_trans_destroy(trans);
2883 + break;
2884 + }
2885 list_del_rcu(&nft_trans_set(trans)->list);
2886 break;
2887 case NFT_MSG_DELSET:
2888 @@ -6595,8 +6658,11 @@ static int __nf_tables_abort(struct net *net)
2889 nft_trans_destroy(trans);
2890 break;
2891 case NFT_MSG_NEWSETELEM:
2892 + if (nft_trans_elem_set(trans)->bound) {
2893 + nft_trans_destroy(trans);
2894 + break;
2895 + }
2896 te = (struct nft_trans_elem *)trans->data;
2897 -
2898 te->set->ops->remove(net, te->set, &te->elem);
2899 atomic_dec(&te->set->nelems);
2900 break;
2901 diff --git a/net/netfilter/nft_compat.c b/net/netfilter/nft_compat.c
2902 index 38da1f5436b4..1245e02239d9 100644
2903 --- a/net/netfilter/nft_compat.c
2904 +++ b/net/netfilter/nft_compat.c
2905 @@ -23,19 +23,6 @@
2906 #include <linux/netfilter_arp/arp_tables.h>
2907 #include <net/netfilter/nf_tables.h>
2908
2909 -struct nft_xt {
2910 - struct list_head head;
2911 - struct nft_expr_ops ops;
2912 - unsigned int refcnt;
2913 -
2914 - /* Unlike other expressions, ops doesn't have static storage duration.
2915 - * nft core assumes they do. We use kfree_rcu so that nft core can
2916 - * can check expr->ops->size even after nft_compat->destroy() frees
2917 - * the nft_xt struct that holds the ops structure.
2918 - */
2919 - struct rcu_head rcu_head;
2920 -};
2921 -
2922 /* Used for matches where *info is larger than X byte */
2923 #define NFT_MATCH_LARGE_THRESH 192
2924
2925 @@ -43,17 +30,6 @@ struct nft_xt_match_priv {
2926 void *info;
2927 };
2928
2929 -static bool nft_xt_put(struct nft_xt *xt)
2930 -{
2931 - if (--xt->refcnt == 0) {
2932 - list_del(&xt->head);
2933 - kfree_rcu(xt, rcu_head);
2934 - return true;
2935 - }
2936 -
2937 - return false;
2938 -}
2939 -
2940 static int nft_compat_chain_validate_dependency(const struct nft_ctx *ctx,
2941 const char *tablename)
2942 {
2943 @@ -248,7 +224,6 @@ nft_target_init(const struct nft_ctx *ctx, const struct nft_expr *expr,
2944 struct xt_target *target = expr->ops->data;
2945 struct xt_tgchk_param par;
2946 size_t size = XT_ALIGN(nla_len(tb[NFTA_TARGET_INFO]));
2947 - struct nft_xt *nft_xt;
2948 u16 proto = 0;
2949 bool inv = false;
2950 union nft_entry e = {};
2951 @@ -272,8 +247,6 @@ nft_target_init(const struct nft_ctx *ctx, const struct nft_expr *expr,
2952 if (!target->target)
2953 return -EINVAL;
2954
2955 - nft_xt = container_of(expr->ops, struct nft_xt, ops);
2956 - nft_xt->refcnt++;
2957 return 0;
2958 }
2959
2960 @@ -292,8 +265,8 @@ nft_target_destroy(const struct nft_ctx *ctx, const struct nft_expr *expr)
2961 if (par.target->destroy != NULL)
2962 par.target->destroy(&par);
2963
2964 - if (nft_xt_put(container_of(expr->ops, struct nft_xt, ops)))
2965 - module_put(me);
2966 + module_put(me);
2967 + kfree(expr->ops);
2968 }
2969
2970 static int nft_target_dump(struct sk_buff *skb, const struct nft_expr *expr)
2971 @@ -447,7 +420,6 @@ __nft_match_init(const struct nft_ctx *ctx, const struct nft_expr *expr,
2972 struct xt_match *match = expr->ops->data;
2973 struct xt_mtchk_param par;
2974 size_t size = XT_ALIGN(nla_len(tb[NFTA_MATCH_INFO]));
2975 - struct nft_xt *nft_xt;
2976 u16 proto = 0;
2977 bool inv = false;
2978 union nft_entry e = {};
2979 @@ -463,13 +435,7 @@ __nft_match_init(const struct nft_ctx *ctx, const struct nft_expr *expr,
2980
2981 nft_match_set_mtchk_param(&par, ctx, match, info, &e, proto, inv);
2982
2983 - ret = xt_check_match(&par, size, proto, inv);
2984 - if (ret < 0)
2985 - return ret;
2986 -
2987 - nft_xt = container_of(expr->ops, struct nft_xt, ops);
2988 - nft_xt->refcnt++;
2989 - return 0;
2990 + return xt_check_match(&par, size, proto, inv);
2991 }
2992
2993 static int
2994 @@ -512,8 +478,8 @@ __nft_match_destroy(const struct nft_ctx *ctx, const struct nft_expr *expr,
2995 if (par.match->destroy != NULL)
2996 par.match->destroy(&par);
2997
2998 - if (nft_xt_put(container_of(expr->ops, struct nft_xt, ops)))
2999 - module_put(me);
3000 + module_put(me);
3001 + kfree(expr->ops);
3002 }
3003
3004 static void
3005 @@ -715,22 +681,13 @@ static const struct nfnetlink_subsystem nfnl_compat_subsys = {
3006 .cb = nfnl_nft_compat_cb,
3007 };
3008
3009 -static LIST_HEAD(nft_match_list);
3010 -
3011 static struct nft_expr_type nft_match_type;
3012
3013 -static bool nft_match_cmp(const struct xt_match *match,
3014 - const char *name, u32 rev, u32 family)
3015 -{
3016 - return strcmp(match->name, name) == 0 && match->revision == rev &&
3017 - (match->family == NFPROTO_UNSPEC || match->family == family);
3018 -}
3019 -
3020 static const struct nft_expr_ops *
3021 nft_match_select_ops(const struct nft_ctx *ctx,
3022 const struct nlattr * const tb[])
3023 {
3024 - struct nft_xt *nft_match;
3025 + struct nft_expr_ops *ops;
3026 struct xt_match *match;
3027 unsigned int matchsize;
3028 char *mt_name;
3029 @@ -746,14 +703,6 @@ nft_match_select_ops(const struct nft_ctx *ctx,
3030 rev = ntohl(nla_get_be32(tb[NFTA_MATCH_REV]));
3031 family = ctx->family;
3032
3033 - /* Re-use the existing match if it's already loaded. */
3034 - list_for_each_entry(nft_match, &nft_match_list, head) {
3035 - struct xt_match *match = nft_match->ops.data;
3036 -
3037 - if (nft_match_cmp(match, mt_name, rev, family))
3038 - return &nft_match->ops;
3039 - }
3040 -
3041 match = xt_request_find_match(family, mt_name, rev);
3042 if (IS_ERR(match))
3043 return ERR_PTR(-ENOENT);
3044 @@ -763,66 +712,62 @@ nft_match_select_ops(const struct nft_ctx *ctx,
3045 goto err;
3046 }
3047
3048 - /* This is the first time we use this match, allocate operations */
3049 - nft_match = kzalloc(sizeof(struct nft_xt), GFP_KERNEL);
3050 - if (nft_match == NULL) {
3051 + ops = kzalloc(sizeof(struct nft_expr_ops), GFP_KERNEL);
3052 + if (!ops) {
3053 err = -ENOMEM;
3054 goto err;
3055 }
3056
3057 - nft_match->refcnt = 0;
3058 - nft_match->ops.type = &nft_match_type;
3059 - nft_match->ops.eval = nft_match_eval;
3060 - nft_match->ops.init = nft_match_init;
3061 - nft_match->ops.destroy = nft_match_destroy;
3062 - nft_match->ops.dump = nft_match_dump;
3063 - nft_match->ops.validate = nft_match_validate;
3064 - nft_match->ops.data = match;
3065 + ops->type = &nft_match_type;
3066 + ops->eval = nft_match_eval;
3067 + ops->init = nft_match_init;
3068 + ops->destroy = nft_match_destroy;
3069 + ops->dump = nft_match_dump;
3070 + ops->validate = nft_match_validate;
3071 + ops->data = match;
3072
3073 matchsize = NFT_EXPR_SIZE(XT_ALIGN(match->matchsize));
3074 if (matchsize > NFT_MATCH_LARGE_THRESH) {
3075 matchsize = NFT_EXPR_SIZE(sizeof(struct nft_xt_match_priv));
3076
3077 - nft_match->ops.eval = nft_match_large_eval;
3078 - nft_match->ops.init = nft_match_large_init;
3079 - nft_match->ops.destroy = nft_match_large_destroy;
3080 - nft_match->ops.dump = nft_match_large_dump;
3081 + ops->eval = nft_match_large_eval;
3082 + ops->init = nft_match_large_init;
3083 + ops->destroy = nft_match_large_destroy;
3084 + ops->dump = nft_match_large_dump;
3085 }
3086
3087 - nft_match->ops.size = matchsize;
3088 -
3089 - list_add(&nft_match->head, &nft_match_list);
3090 + ops->size = matchsize;
3091
3092 - return &nft_match->ops;
3093 + return ops;
3094 err:
3095 module_put(match->me);
3096 return ERR_PTR(err);
3097 }
3098
3099 +static void nft_match_release_ops(const struct nft_expr_ops *ops)
3100 +{
3101 + struct xt_match *match = ops->data;
3102 +
3103 + module_put(match->me);
3104 + kfree(ops);
3105 +}
3106 +
3107 static struct nft_expr_type nft_match_type __read_mostly = {
3108 .name = "match",
3109 .select_ops = nft_match_select_ops,
3110 + .release_ops = nft_match_release_ops,
3111 .policy = nft_match_policy,
3112 .maxattr = NFTA_MATCH_MAX,
3113 .owner = THIS_MODULE,
3114 };
3115
3116 -static LIST_HEAD(nft_target_list);
3117 -
3118 static struct nft_expr_type nft_target_type;
3119
3120 -static bool nft_target_cmp(const struct xt_target *tg,
3121 - const char *name, u32 rev, u32 family)
3122 -{
3123 - return strcmp(tg->name, name) == 0 && tg->revision == rev &&
3124 - (tg->family == NFPROTO_UNSPEC || tg->family == family);
3125 -}
3126 -
3127 static const struct nft_expr_ops *
3128 nft_target_select_ops(const struct nft_ctx *ctx,
3129 const struct nlattr * const tb[])
3130 {
3131 - struct nft_xt *nft_target;
3132 + struct nft_expr_ops *ops;
3133 struct xt_target *target;
3134 char *tg_name;
3135 u32 rev, family;
3136 @@ -842,17 +787,6 @@ nft_target_select_ops(const struct nft_ctx *ctx,
3137 strcmp(tg_name, "standard") == 0)
3138 return ERR_PTR(-EINVAL);
3139
3140 - /* Re-use the existing target if it's already loaded. */
3141 - list_for_each_entry(nft_target, &nft_target_list, head) {
3142 - struct xt_target *target = nft_target->ops.data;
3143 -
3144 - if (!target->target)
3145 - continue;
3146 -
3147 - if (nft_target_cmp(target, tg_name, rev, family))
3148 - return &nft_target->ops;
3149 - }
3150 -
3151 target = xt_request_find_target(family, tg_name, rev);
3152 if (IS_ERR(target))
3153 return ERR_PTR(-ENOENT);
3154 @@ -867,38 +801,43 @@ nft_target_select_ops(const struct nft_ctx *ctx,
3155 goto err;
3156 }
3157
3158 - /* This is the first time we use this target, allocate operations */
3159 - nft_target = kzalloc(sizeof(struct nft_xt), GFP_KERNEL);
3160 - if (nft_target == NULL) {
3161 + ops = kzalloc(sizeof(struct nft_expr_ops), GFP_KERNEL);
3162 + if (!ops) {
3163 err = -ENOMEM;
3164 goto err;
3165 }
3166
3167 - nft_target->refcnt = 0;
3168 - nft_target->ops.type = &nft_target_type;
3169 - nft_target->ops.size = NFT_EXPR_SIZE(XT_ALIGN(target->targetsize));
3170 - nft_target->ops.init = nft_target_init;
3171 - nft_target->ops.destroy = nft_target_destroy;
3172 - nft_target->ops.dump = nft_target_dump;
3173 - nft_target->ops.validate = nft_target_validate;
3174 - nft_target->ops.data = target;
3175 + ops->type = &nft_target_type;
3176 + ops->size = NFT_EXPR_SIZE(XT_ALIGN(target->targetsize));
3177 + ops->init = nft_target_init;
3178 + ops->destroy = nft_target_destroy;
3179 + ops->dump = nft_target_dump;
3180 + ops->validate = nft_target_validate;
3181 + ops->data = target;
3182
3183 if (family == NFPROTO_BRIDGE)
3184 - nft_target->ops.eval = nft_target_eval_bridge;
3185 + ops->eval = nft_target_eval_bridge;
3186 else
3187 - nft_target->ops.eval = nft_target_eval_xt;
3188 -
3189 - list_add(&nft_target->head, &nft_target_list);
3190 + ops->eval = nft_target_eval_xt;
3191
3192 - return &nft_target->ops;
3193 + return ops;
3194 err:
3195 module_put(target->me);
3196 return ERR_PTR(err);
3197 }
3198
3199 +static void nft_target_release_ops(const struct nft_expr_ops *ops)
3200 +{
3201 + struct xt_target *target = ops->data;
3202 +
3203 + module_put(target->me);
3204 + kfree(ops);
3205 +}
3206 +
3207 static struct nft_expr_type nft_target_type __read_mostly = {
3208 .name = "target",
3209 .select_ops = nft_target_select_ops,
3210 + .release_ops = nft_target_release_ops,
3211 .policy = nft_target_policy,
3212 .maxattr = NFTA_TARGET_MAX,
3213 .owner = THIS_MODULE,
3214 @@ -923,7 +862,6 @@ static int __init nft_compat_module_init(void)
3215 }
3216
3217 return ret;
3218 -
3219 err_target:
3220 nft_unregister_expr(&nft_target_type);
3221 err_match:
3222 @@ -933,32 +871,6 @@ err_match:
3223
3224 static void __exit nft_compat_module_exit(void)
3225 {
3226 - struct nft_xt *xt, *next;
3227 -
3228 - /* list should be empty here, it can be non-empty only in case there
3229 - * was an error that caused nft_xt expr to not be initialized fully
3230 - * and noone else requested the same expression later.
3231 - *
3232 - * In this case, the lists contain 0-refcount entries that still
3233 - * hold module reference.
3234 - */
3235 - list_for_each_entry_safe(xt, next, &nft_target_list, head) {
3236 - struct xt_target *target = xt->ops.data;
3237 -
3238 - if (WARN_ON_ONCE(xt->refcnt))
3239 - continue;
3240 - module_put(target->me);
3241 - kfree(xt);
3242 - }
3243 -
3244 - list_for_each_entry_safe(xt, next, &nft_match_list, head) {
3245 - struct xt_match *match = xt->ops.data;
3246 -
3247 - if (WARN_ON_ONCE(xt->refcnt))
3248 - continue;
3249 - module_put(match->me);
3250 - kfree(xt);
3251 - }
3252 nfnetlink_subsys_unregister(&nfnl_compat_subsys);
3253 nft_unregister_expr(&nft_target_type);
3254 nft_unregister_expr(&nft_match_type);
3255 diff --git a/net/netfilter/nft_dynset.c b/net/netfilter/nft_dynset.c
3256 index 6e91a37d57f2..eb7f9a5f2aeb 100644
3257 --- a/net/netfilter/nft_dynset.c
3258 +++ b/net/netfilter/nft_dynset.c
3259 @@ -235,14 +235,32 @@ err1:
3260 return err;
3261 }
3262
3263 +static void nft_dynset_deactivate(const struct nft_ctx *ctx,
3264 + const struct nft_expr *expr,
3265 + enum nft_trans_phase phase)
3266 +{
3267 + struct nft_dynset *priv = nft_expr_priv(expr);
3268 +
3269 + nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase);
3270 +}
3271 +
3272 +static void nft_dynset_activate(const struct nft_ctx *ctx,
3273 + const struct nft_expr *expr)
3274 +{
3275 + struct nft_dynset *priv = nft_expr_priv(expr);
3276 +
3277 + priv->set->use++;
3278 +}
3279 +
3280 static void nft_dynset_destroy(const struct nft_ctx *ctx,
3281 const struct nft_expr *expr)
3282 {
3283 struct nft_dynset *priv = nft_expr_priv(expr);
3284
3285 - nf_tables_unbind_set(ctx, priv->set, &priv->binding);
3286 if (priv->expr != NULL)
3287 nft_expr_destroy(ctx, priv->expr);
3288 +
3289 + nf_tables_destroy_set(ctx, priv->set);
3290 }
3291
3292 static int nft_dynset_dump(struct sk_buff *skb, const struct nft_expr *expr)
3293 @@ -279,6 +297,8 @@ static const struct nft_expr_ops nft_dynset_ops = {
3294 .eval = nft_dynset_eval,
3295 .init = nft_dynset_init,
3296 .destroy = nft_dynset_destroy,
3297 + .activate = nft_dynset_activate,
3298 + .deactivate = nft_dynset_deactivate,
3299 .dump = nft_dynset_dump,
3300 };
3301
3302 diff --git a/net/netfilter/nft_immediate.c b/net/netfilter/nft_immediate.c
3303 index 0777a93211e2..3f6d1d2a6281 100644
3304 --- a/net/netfilter/nft_immediate.c
3305 +++ b/net/netfilter/nft_immediate.c
3306 @@ -72,10 +72,14 @@ static void nft_immediate_activate(const struct nft_ctx *ctx,
3307 }
3308
3309 static void nft_immediate_deactivate(const struct nft_ctx *ctx,
3310 - const struct nft_expr *expr)
3311 + const struct nft_expr *expr,
3312 + enum nft_trans_phase phase)
3313 {
3314 const struct nft_immediate_expr *priv = nft_expr_priv(expr);
3315
3316 + if (phase == NFT_TRANS_COMMIT)
3317 + return;
3318 +
3319 return nft_data_release(&priv->data, nft_dreg_to_type(priv->dreg));
3320 }
3321
3322 diff --git a/net/netfilter/nft_lookup.c b/net/netfilter/nft_lookup.c
3323 index ad13e8643599..161c3451a747 100644
3324 --- a/net/netfilter/nft_lookup.c
3325 +++ b/net/netfilter/nft_lookup.c
3326 @@ -121,12 +121,29 @@ static int nft_lookup_init(const struct nft_ctx *ctx,
3327 return 0;
3328 }
3329
3330 +static void nft_lookup_deactivate(const struct nft_ctx *ctx,
3331 + const struct nft_expr *expr,
3332 + enum nft_trans_phase phase)
3333 +{
3334 + struct nft_lookup *priv = nft_expr_priv(expr);
3335 +
3336 + nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase);
3337 +}
3338 +
3339 +static void nft_lookup_activate(const struct nft_ctx *ctx,
3340 + const struct nft_expr *expr)
3341 +{
3342 + struct nft_lookup *priv = nft_expr_priv(expr);
3343 +
3344 + priv->set->use++;
3345 +}
3346 +
3347 static void nft_lookup_destroy(const struct nft_ctx *ctx,
3348 const struct nft_expr *expr)
3349 {
3350 struct nft_lookup *priv = nft_expr_priv(expr);
3351
3352 - nf_tables_unbind_set(ctx, priv->set, &priv->binding);
3353 + nf_tables_destroy_set(ctx, priv->set);
3354 }
3355
3356 static int nft_lookup_dump(struct sk_buff *skb, const struct nft_expr *expr)
3357 @@ -209,6 +226,8 @@ static const struct nft_expr_ops nft_lookup_ops = {
3358 .size = NFT_EXPR_SIZE(sizeof(struct nft_lookup)),
3359 .eval = nft_lookup_eval,
3360 .init = nft_lookup_init,
3361 + .activate = nft_lookup_activate,
3362 + .deactivate = nft_lookup_deactivate,
3363 .destroy = nft_lookup_destroy,
3364 .dump = nft_lookup_dump,
3365 .validate = nft_lookup_validate,
3366 diff --git a/net/netfilter/nft_objref.c b/net/netfilter/nft_objref.c
3367 index cdf348f751ec..bf92a40dd1b2 100644
3368 --- a/net/netfilter/nft_objref.c
3369 +++ b/net/netfilter/nft_objref.c
3370 @@ -64,21 +64,34 @@ nla_put_failure:
3371 return -1;
3372 }
3373
3374 -static void nft_objref_destroy(const struct nft_ctx *ctx,
3375 - const struct nft_expr *expr)
3376 +static void nft_objref_deactivate(const struct nft_ctx *ctx,
3377 + const struct nft_expr *expr,
3378 + enum nft_trans_phase phase)
3379 {
3380 struct nft_object *obj = nft_objref_priv(expr);
3381
3382 + if (phase == NFT_TRANS_COMMIT)
3383 + return;
3384 +
3385 obj->use--;
3386 }
3387
3388 +static void nft_objref_activate(const struct nft_ctx *ctx,
3389 + const struct nft_expr *expr)
3390 +{
3391 + struct nft_object *obj = nft_objref_priv(expr);
3392 +
3393 + obj->use++;
3394 +}
3395 +
3396 static struct nft_expr_type nft_objref_type;
3397 static const struct nft_expr_ops nft_objref_ops = {
3398 .type = &nft_objref_type,
3399 .size = NFT_EXPR_SIZE(sizeof(struct nft_object *)),
3400 .eval = nft_objref_eval,
3401 .init = nft_objref_init,
3402 - .destroy = nft_objref_destroy,
3403 + .activate = nft_objref_activate,
3404 + .deactivate = nft_objref_deactivate,
3405 .dump = nft_objref_dump,
3406 };
3407
3408 @@ -155,12 +168,29 @@ nla_put_failure:
3409 return -1;
3410 }
3411
3412 +static void nft_objref_map_deactivate(const struct nft_ctx *ctx,
3413 + const struct nft_expr *expr,
3414 + enum nft_trans_phase phase)
3415 +{
3416 + struct nft_objref_map *priv = nft_expr_priv(expr);
3417 +
3418 + nf_tables_deactivate_set(ctx, priv->set, &priv->binding, phase);
3419 +}
3420 +
3421 +static void nft_objref_map_activate(const struct nft_ctx *ctx,
3422 + const struct nft_expr *expr)
3423 +{
3424 + struct nft_objref_map *priv = nft_expr_priv(expr);
3425 +
3426 + priv->set->use++;
3427 +}
3428 +
3429 static void nft_objref_map_destroy(const struct nft_ctx *ctx,
3430 const struct nft_expr *expr)
3431 {
3432 struct nft_objref_map *priv = nft_expr_priv(expr);
3433
3434 - nf_tables_unbind_set(ctx, priv->set, &priv->binding);
3435 + nf_tables_destroy_set(ctx, priv->set);
3436 }
3437
3438 static struct nft_expr_type nft_objref_type;
3439 @@ -169,6 +199,8 @@ static const struct nft_expr_ops nft_objref_map_ops = {
3440 .size = NFT_EXPR_SIZE(sizeof(struct nft_objref_map)),
3441 .eval = nft_objref_map_eval,
3442 .init = nft_objref_map_init,
3443 + .activate = nft_objref_map_activate,
3444 + .deactivate = nft_objref_map_deactivate,
3445 .destroy = nft_objref_map_destroy,
3446 .dump = nft_objref_map_dump,
3447 };
3448 diff --git a/net/netrom/af_netrom.c b/net/netrom/af_netrom.c
3449 index 1d3144d19903..71ffd1a6dc7c 100644
3450 --- a/net/netrom/af_netrom.c
3451 +++ b/net/netrom/af_netrom.c
3452 @@ -1392,18 +1392,22 @@ static int __init nr_proto_init(void)
3453 int i;
3454 int rc = proto_register(&nr_proto, 0);
3455
3456 - if (rc != 0)
3457 - goto out;
3458 + if (rc)
3459 + return rc;
3460
3461 if (nr_ndevs > 0x7fffffff/sizeof(struct net_device *)) {
3462 - printk(KERN_ERR "NET/ROM: nr_proto_init - nr_ndevs parameter to large\n");
3463 - return -1;
3464 + pr_err("NET/ROM: %s - nr_ndevs parameter too large\n",
3465 + __func__);
3466 + rc = -EINVAL;
3467 + goto unregister_proto;
3468 }
3469
3470 dev_nr = kcalloc(nr_ndevs, sizeof(struct net_device *), GFP_KERNEL);
3471 - if (dev_nr == NULL) {
3472 - printk(KERN_ERR "NET/ROM: nr_proto_init - unable to allocate device array\n");
3473 - return -1;
3474 + if (!dev_nr) {
3475 + pr_err("NET/ROM: %s - unable to allocate device array\n",
3476 + __func__);
3477 + rc = -ENOMEM;
3478 + goto unregister_proto;
3479 }
3480
3481 for (i = 0; i < nr_ndevs; i++) {
3482 @@ -1413,13 +1417,13 @@ static int __init nr_proto_init(void)
3483 sprintf(name, "nr%d", i);
3484 dev = alloc_netdev(0, name, NET_NAME_UNKNOWN, nr_setup);
3485 if (!dev) {
3486 - printk(KERN_ERR "NET/ROM: nr_proto_init - unable to allocate device structure\n");
3487 + rc = -ENOMEM;
3488 goto fail;
3489 }
3490
3491 dev->base_addr = i;
3492 - if (register_netdev(dev)) {
3493 - printk(KERN_ERR "NET/ROM: nr_proto_init - unable to register network device\n");
3494 + rc = register_netdev(dev);
3495 + if (rc) {
3496 free_netdev(dev);
3497 goto fail;
3498 }
3499 @@ -1427,36 +1431,64 @@ static int __init nr_proto_init(void)
3500 dev_nr[i] = dev;
3501 }
3502
3503 - if (sock_register(&nr_family_ops)) {
3504 - printk(KERN_ERR "NET/ROM: nr_proto_init - unable to register socket family\n");
3505 + rc = sock_register(&nr_family_ops);
3506 + if (rc)
3507 goto fail;
3508 - }
3509
3510 - register_netdevice_notifier(&nr_dev_notifier);
3511 + rc = register_netdevice_notifier(&nr_dev_notifier);
3512 + if (rc)
3513 + goto out_sock;
3514
3515 ax25_register_pid(&nr_pid);
3516 ax25_linkfail_register(&nr_linkfail_notifier);
3517
3518 #ifdef CONFIG_SYSCTL
3519 - nr_register_sysctl();
3520 + rc = nr_register_sysctl();
3521 + if (rc)
3522 + goto out_sysctl;
3523 #endif
3524
3525 nr_loopback_init();
3526
3527 - proc_create_seq("nr", 0444, init_net.proc_net, &nr_info_seqops);
3528 - proc_create_seq("nr_neigh", 0444, init_net.proc_net, &nr_neigh_seqops);
3529 - proc_create_seq("nr_nodes", 0444, init_net.proc_net, &nr_node_seqops);
3530 -out:
3531 - return rc;
3532 + rc = -ENOMEM;
3533 + if (!proc_create_seq("nr", 0444, init_net.proc_net, &nr_info_seqops))
3534 + goto proc_remove1;
3535 + if (!proc_create_seq("nr_neigh", 0444, init_net.proc_net,
3536 + &nr_neigh_seqops))
3537 + goto proc_remove2;
3538 + if (!proc_create_seq("nr_nodes", 0444, init_net.proc_net,
3539 + &nr_node_seqops))
3540 + goto proc_remove3;
3541 +
3542 + return 0;
3543 +
3544 +proc_remove3:
3545 + remove_proc_entry("nr_neigh", init_net.proc_net);
3546 +proc_remove2:
3547 + remove_proc_entry("nr", init_net.proc_net);
3548 +proc_remove1:
3549 +
3550 + nr_loopback_clear();
3551 + nr_rt_free();
3552 +
3553 +#ifdef CONFIG_SYSCTL
3554 + nr_unregister_sysctl();
3555 +out_sysctl:
3556 +#endif
3557 + ax25_linkfail_release(&nr_linkfail_notifier);
3558 + ax25_protocol_release(AX25_P_NETROM);
3559 + unregister_netdevice_notifier(&nr_dev_notifier);
3560 +out_sock:
3561 + sock_unregister(PF_NETROM);
3562 fail:
3563 while (--i >= 0) {
3564 unregister_netdev(dev_nr[i]);
3565 free_netdev(dev_nr[i]);
3566 }
3567 kfree(dev_nr);
3568 +unregister_proto:
3569 proto_unregister(&nr_proto);
3570 - rc = -1;
3571 - goto out;
3572 + return rc;
3573 }
3574
3575 module_init(nr_proto_init);
3576 diff --git a/net/netrom/nr_loopback.c b/net/netrom/nr_loopback.c
3577 index 215ad22a9647..93d13f019981 100644
3578 --- a/net/netrom/nr_loopback.c
3579 +++ b/net/netrom/nr_loopback.c
3580 @@ -70,7 +70,7 @@ static void nr_loopback_timer(struct timer_list *unused)
3581 }
3582 }
3583
3584 -void __exit nr_loopback_clear(void)
3585 +void nr_loopback_clear(void)
3586 {
3587 del_timer_sync(&loopback_timer);
3588 skb_queue_purge(&loopback_queue);
3589 diff --git a/net/netrom/nr_route.c b/net/netrom/nr_route.c
3590 index 6485f593e2f0..b76aa668a94b 100644
3591 --- a/net/netrom/nr_route.c
3592 +++ b/net/netrom/nr_route.c
3593 @@ -953,7 +953,7 @@ const struct seq_operations nr_neigh_seqops = {
3594 /*
3595 * Free all memory associated with the nodes and routes lists.
3596 */
3597 -void __exit nr_rt_free(void)
3598 +void nr_rt_free(void)
3599 {
3600 struct nr_neigh *s = NULL;
3601 struct nr_node *t = NULL;
3602 diff --git a/net/netrom/sysctl_net_netrom.c b/net/netrom/sysctl_net_netrom.c
3603 index ba1c368b3f18..771011b84270 100644
3604 --- a/net/netrom/sysctl_net_netrom.c
3605 +++ b/net/netrom/sysctl_net_netrom.c
3606 @@ -146,9 +146,12 @@ static struct ctl_table nr_table[] = {
3607 { }
3608 };
3609
3610 -void __init nr_register_sysctl(void)
3611 +int __init nr_register_sysctl(void)
3612 {
3613 nr_table_header = register_net_sysctl(&init_net, "net/netrom", nr_table);
3614 + if (!nr_table_header)
3615 + return -ENOMEM;
3616 + return 0;
3617 }
3618
3619 void nr_unregister_sysctl(void)
3620 diff --git a/net/rds/af_rds.c b/net/rds/af_rds.c
3621 index 65387e1e6964..cd7e01ea8144 100644
3622 --- a/net/rds/af_rds.c
3623 +++ b/net/rds/af_rds.c
3624 @@ -506,6 +506,9 @@ static int rds_connect(struct socket *sock, struct sockaddr *uaddr,
3625 struct rds_sock *rs = rds_sk_to_rs(sk);
3626 int ret = 0;
3627
3628 + if (addr_len < offsetofend(struct sockaddr, sa_family))
3629 + return -EINVAL;
3630 +
3631 lock_sock(sk);
3632
3633 switch (uaddr->sa_family) {
3634 diff --git a/net/rds/bind.c b/net/rds/bind.c
3635 index 17c9d9f0c848..0f4398e7f2a7 100644
3636 --- a/net/rds/bind.c
3637 +++ b/net/rds/bind.c
3638 @@ -173,6 +173,8 @@ int rds_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
3639 /* We allow an RDS socket to be bound to either IPv4 or IPv6
3640 * address.
3641 */
3642 + if (addr_len < offsetofend(struct sockaddr, sa_family))
3643 + return -EINVAL;
3644 if (uaddr->sa_family == AF_INET) {
3645 struct sockaddr_in *sin = (struct sockaddr_in *)uaddr;
3646
3647 diff --git a/net/rds/ib_fmr.c b/net/rds/ib_fmr.c
3648 index e0f70c4051b6..01e764f8f224 100644
3649 --- a/net/rds/ib_fmr.c
3650 +++ b/net/rds/ib_fmr.c
3651 @@ -44,6 +44,17 @@ struct rds_ib_mr *rds_ib_alloc_fmr(struct rds_ib_device *rds_ibdev, int npages)
3652 else
3653 pool = rds_ibdev->mr_1m_pool;
3654
3655 + if (atomic_read(&pool->dirty_count) >= pool->max_items / 10)
3656 + queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10);
3657 +
3658 + /* Switch pools if one of the pool is reaching upper limit */
3659 + if (atomic_read(&pool->dirty_count) >= pool->max_items * 9 / 10) {
3660 + if (pool->pool_type == RDS_IB_MR_8K_POOL)
3661 + pool = rds_ibdev->mr_1m_pool;
3662 + else
3663 + pool = rds_ibdev->mr_8k_pool;
3664 + }
3665 +
3666 ibmr = rds_ib_try_reuse_ibmr(pool);
3667 if (ibmr)
3668 return ibmr;
3669 diff --git a/net/rds/ib_rdma.c b/net/rds/ib_rdma.c
3670 index 63c8d107adcf..d664e9ade74d 100644
3671 --- a/net/rds/ib_rdma.c
3672 +++ b/net/rds/ib_rdma.c
3673 @@ -454,9 +454,6 @@ struct rds_ib_mr *rds_ib_try_reuse_ibmr(struct rds_ib_mr_pool *pool)
3674 struct rds_ib_mr *ibmr = NULL;
3675 int iter = 0;
3676
3677 - if (atomic_read(&pool->dirty_count) >= pool->max_items_soft / 10)
3678 - queue_delayed_work(rds_ib_mr_wq, &pool->flush_worker, 10);
3679 -
3680 while (1) {
3681 ibmr = rds_ib_reuse_mr(pool);
3682 if (ibmr)
3683 diff --git a/net/rose/rose_loopback.c b/net/rose/rose_loopback.c
3684 index 7af4f99c4a93..094a6621f8e8 100644
3685 --- a/net/rose/rose_loopback.c
3686 +++ b/net/rose/rose_loopback.c
3687 @@ -16,6 +16,7 @@
3688 #include <linux/init.h>
3689
3690 static struct sk_buff_head loopback_queue;
3691 +#define ROSE_LOOPBACK_LIMIT 1000
3692 static struct timer_list loopback_timer;
3693
3694 static void rose_set_loopback_timer(void);
3695 @@ -35,29 +36,27 @@ static int rose_loopback_running(void)
3696
3697 int rose_loopback_queue(struct sk_buff *skb, struct rose_neigh *neigh)
3698 {
3699 - struct sk_buff *skbn;
3700 + struct sk_buff *skbn = NULL;
3701
3702 - skbn = skb_clone(skb, GFP_ATOMIC);
3703 + if (skb_queue_len(&loopback_queue) < ROSE_LOOPBACK_LIMIT)
3704 + skbn = skb_clone(skb, GFP_ATOMIC);
3705
3706 - kfree_skb(skb);
3707 -
3708 - if (skbn != NULL) {
3709 + if (skbn) {
3710 + consume_skb(skb);
3711 skb_queue_tail(&loopback_queue, skbn);
3712
3713 if (!rose_loopback_running())
3714 rose_set_loopback_timer();
3715 + } else {
3716 + kfree_skb(skb);
3717 }
3718
3719 return 1;
3720 }
3721
3722 -
3723 static void rose_set_loopback_timer(void)
3724 {
3725 - del_timer(&loopback_timer);
3726 -
3727 - loopback_timer.expires = jiffies + 10;
3728 - add_timer(&loopback_timer);
3729 + mod_timer(&loopback_timer, jiffies + 10);
3730 }
3731
3732 static void rose_loopback_timer(struct timer_list *unused)
3733 @@ -68,8 +67,12 @@ static void rose_loopback_timer(struct timer_list *unused)
3734 struct sock *sk;
3735 unsigned short frametype;
3736 unsigned int lci_i, lci_o;
3737 + int count;
3738
3739 - while ((skb = skb_dequeue(&loopback_queue)) != NULL) {
3740 + for (count = 0; count < ROSE_LOOPBACK_LIMIT; count++) {
3741 + skb = skb_dequeue(&loopback_queue);
3742 + if (!skb)
3743 + return;
3744 if (skb->len < ROSE_MIN_LEN) {
3745 kfree_skb(skb);
3746 continue;
3747 @@ -106,6 +109,8 @@ static void rose_loopback_timer(struct timer_list *unused)
3748 kfree_skb(skb);
3749 }
3750 }
3751 + if (!skb_queue_empty(&loopback_queue))
3752 + mod_timer(&loopback_timer, jiffies + 1);
3753 }
3754
3755 void __exit rose_loopback_clear(void)
3756 diff --git a/net/rxrpc/input.c b/net/rxrpc/input.c
3757 index 570b49d2da42..d591f54cb91f 100644
3758 --- a/net/rxrpc/input.c
3759 +++ b/net/rxrpc/input.c
3760 @@ -1155,19 +1155,19 @@ int rxrpc_extract_header(struct rxrpc_skb_priv *sp, struct sk_buff *skb)
3761 * handle data received on the local endpoint
3762 * - may be called in interrupt context
3763 *
3764 - * The socket is locked by the caller and this prevents the socket from being
3765 - * shut down and the local endpoint from going away, thus sk_user_data will not
3766 - * be cleared until this function returns.
3767 + * [!] Note that as this is called from the encap_rcv hook, the socket is not
3768 + * held locked by the caller and nothing prevents sk_user_data on the UDP from
3769 + * being cleared in the middle of processing this function.
3770 *
3771 * Called with the RCU read lock held from the IP layer via UDP.
3772 */
3773 int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb)
3774 {
3775 + struct rxrpc_local *local = rcu_dereference_sk_user_data(udp_sk);
3776 struct rxrpc_connection *conn;
3777 struct rxrpc_channel *chan;
3778 struct rxrpc_call *call = NULL;
3779 struct rxrpc_skb_priv *sp;
3780 - struct rxrpc_local *local = udp_sk->sk_user_data;
3781 struct rxrpc_peer *peer = NULL;
3782 struct rxrpc_sock *rx = NULL;
3783 unsigned int channel;
3784 @@ -1175,6 +1175,10 @@ int rxrpc_input_packet(struct sock *udp_sk, struct sk_buff *skb)
3785
3786 _enter("%p", udp_sk);
3787
3788 + if (unlikely(!local)) {
3789 + kfree_skb(skb);
3790 + return 0;
3791 + }
3792 if (skb->tstamp == 0)
3793 skb->tstamp = ktime_get_real();
3794
3795 diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
3796 index 0906e51d3cfb..10317dbdab5f 100644
3797 --- a/net/rxrpc/local_object.c
3798 +++ b/net/rxrpc/local_object.c
3799 @@ -304,7 +304,8 @@ nomem:
3800 ret = -ENOMEM;
3801 sock_error:
3802 mutex_unlock(&rxnet->local_mutex);
3803 - kfree(local);
3804 + if (local)
3805 + call_rcu(&local->rcu, rxrpc_local_rcu);
3806 _leave(" = %d", ret);
3807 return ERR_PTR(ret);
3808
3809 diff --git a/net/sunrpc/cache.c b/net/sunrpc/cache.c
3810 index b6e8eccf2a52..214440c5b14e 100644
3811 --- a/net/sunrpc/cache.c
3812 +++ b/net/sunrpc/cache.c
3813 @@ -54,6 +54,7 @@ static void cache_init(struct cache_head *h, struct cache_detail *detail)
3814 h->last_refresh = now;
3815 }
3816
3817 +static inline int cache_is_valid(struct cache_head *h);
3818 static void cache_fresh_locked(struct cache_head *head, time_t expiry,
3819 struct cache_detail *detail);
3820 static void cache_fresh_unlocked(struct cache_head *head,
3821 @@ -100,6 +101,8 @@ struct cache_head *sunrpc_cache_lookup(struct cache_detail *detail,
3822 if (cache_is_expired(detail, tmp)) {
3823 hlist_del_init(&tmp->cache_list);
3824 detail->entries --;
3825 + if (cache_is_valid(tmp) == -EAGAIN)
3826 + set_bit(CACHE_NEGATIVE, &tmp->flags);
3827 cache_fresh_locked(tmp, 0, detail);
3828 freeme = tmp;
3829 break;
3830 diff --git a/net/tipc/netlink_compat.c b/net/tipc/netlink_compat.c
3831 index 0b21187d74df..588d5aa14c41 100644
3832 --- a/net/tipc/netlink_compat.c
3833 +++ b/net/tipc/netlink_compat.c
3834 @@ -267,8 +267,14 @@ static int tipc_nl_compat_dumpit(struct tipc_nl_compat_cmd_dump *cmd,
3835 if (msg->rep_type)
3836 tipc_tlv_init(msg->rep, msg->rep_type);
3837
3838 - if (cmd->header)
3839 - (*cmd->header)(msg);
3840 + if (cmd->header) {
3841 + err = (*cmd->header)(msg);
3842 + if (err) {
3843 + kfree_skb(msg->rep);
3844 + msg->rep = NULL;
3845 + return err;
3846 + }
3847 + }
3848
3849 arg = nlmsg_new(0, GFP_KERNEL);
3850 if (!arg) {
3851 @@ -397,7 +403,12 @@ static int tipc_nl_compat_bearer_enable(struct tipc_nl_compat_cmd_doit *cmd,
3852 if (!bearer)
3853 return -EMSGSIZE;
3854
3855 - len = min_t(int, TLV_GET_DATA_LEN(msg->req), TIPC_MAX_BEARER_NAME);
3856 + len = TLV_GET_DATA_LEN(msg->req);
3857 + len -= offsetof(struct tipc_bearer_config, name);
3858 + if (len <= 0)
3859 + return -EINVAL;
3860 +
3861 + len = min_t(int, len, TIPC_MAX_BEARER_NAME);
3862 if (!string_is_valid(b->name, len))
3863 return -EINVAL;
3864
3865 @@ -766,7 +777,12 @@ static int tipc_nl_compat_link_set(struct tipc_nl_compat_cmd_doit *cmd,
3866
3867 lc = (struct tipc_link_config *)TLV_DATA(msg->req);
3868
3869 - len = min_t(int, TLV_GET_DATA_LEN(msg->req), TIPC_MAX_LINK_NAME);
3870 + len = TLV_GET_DATA_LEN(msg->req);
3871 + len -= offsetof(struct tipc_link_config, name);
3872 + if (len <= 0)
3873 + return -EINVAL;
3874 +
3875 + len = min_t(int, len, TIPC_MAX_LINK_NAME);
3876 if (!string_is_valid(lc->name, len))
3877 return -EINVAL;
3878
3879 diff --git a/net/tls/tls_device.c b/net/tls/tls_device.c
3880 index 961b07d4d41c..c9588b682db4 100644
3881 --- a/net/tls/tls_device.c
3882 +++ b/net/tls/tls_device.c
3883 @@ -874,7 +874,9 @@ int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
3884 goto release_netdev;
3885
3886 free_sw_resources:
3887 + up_read(&device_offload_lock);
3888 tls_sw_free_resources_rx(sk);
3889 + down_read(&device_offload_lock);
3890 release_ctx:
3891 ctx->priv_ctx_rx = NULL;
3892 release_netdev:
3893 @@ -909,8 +911,6 @@ void tls_device_offload_cleanup_rx(struct sock *sk)
3894 }
3895 out:
3896 up_read(&device_offload_lock);
3897 - kfree(tls_ctx->rx.rec_seq);
3898 - kfree(tls_ctx->rx.iv);
3899 tls_sw_release_resources_rx(sk);
3900 }
3901
3902 diff --git a/net/tls/tls_device_fallback.c b/net/tls/tls_device_fallback.c
3903 index 450a6dbc5a88..ef8934fd8698 100644
3904 --- a/net/tls/tls_device_fallback.c
3905 +++ b/net/tls/tls_device_fallback.c
3906 @@ -193,6 +193,9 @@ static void update_chksum(struct sk_buff *skb, int headln)
3907
3908 static void complete_skb(struct sk_buff *nskb, struct sk_buff *skb, int headln)
3909 {
3910 + struct sock *sk = skb->sk;
3911 + int delta;
3912 +
3913 skb_copy_header(nskb, skb);
3914
3915 skb_put(nskb, skb->len);
3916 @@ -200,11 +203,15 @@ static void complete_skb(struct sk_buff *nskb, struct sk_buff *skb, int headln)
3917 update_chksum(nskb, headln);
3918
3919 nskb->destructor = skb->destructor;
3920 - nskb->sk = skb->sk;
3921 + nskb->sk = sk;
3922 skb->destructor = NULL;
3923 skb->sk = NULL;
3924 - refcount_add(nskb->truesize - skb->truesize,
3925 - &nskb->sk->sk_wmem_alloc);
3926 +
3927 + delta = nskb->truesize - skb->truesize;
3928 + if (likely(delta < 0))
3929 + WARN_ON_ONCE(refcount_sub_and_test(-delta, &sk->sk_wmem_alloc));
3930 + else if (delta)
3931 + refcount_add(delta, &sk->sk_wmem_alloc);
3932 }
3933
3934 /* This function may be called after the user socket is already
3935 diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c
3936 index a091c03abcb2..25b3fb585777 100644
3937 --- a/net/tls/tls_main.c
3938 +++ b/net/tls/tls_main.c
3939 @@ -290,11 +290,8 @@ static void tls_sk_proto_close(struct sock *sk, long timeout)
3940 tls_sw_free_resources_tx(sk);
3941 }
3942
3943 - if (ctx->rx_conf == TLS_SW) {
3944 - kfree(ctx->rx.rec_seq);
3945 - kfree(ctx->rx.iv);
3946 + if (ctx->rx_conf == TLS_SW)
3947 tls_sw_free_resources_rx(sk);
3948 - }
3949
3950 #ifdef CONFIG_TLS_DEVICE
3951 if (ctx->rx_conf == TLS_HW)
3952 diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
3953 index b9c6ecfbcfea..6848a8196711 100644
3954 --- a/net/tls/tls_sw.c
3955 +++ b/net/tls/tls_sw.c
3956 @@ -1118,6 +1118,9 @@ void tls_sw_release_resources_rx(struct sock *sk)
3957 struct tls_context *tls_ctx = tls_get_ctx(sk);
3958 struct tls_sw_context_rx *ctx = tls_sw_ctx_rx(tls_ctx);
3959
3960 + kfree(tls_ctx->rx.rec_seq);
3961 + kfree(tls_ctx->rx.iv);
3962 +
3963 if (ctx->aead_recv) {
3964 kfree_skb(ctx->recv_pkt);
3965 ctx->recv_pkt = NULL;
3966 diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
3967 index 3ae3a33da70b..602715fc9a75 100644
3968 --- a/net/vmw_vsock/virtio_transport_common.c
3969 +++ b/net/vmw_vsock/virtio_transport_common.c
3970 @@ -662,6 +662,8 @@ static int virtio_transport_reset(struct vsock_sock *vsk,
3971 */
3972 static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt)
3973 {
3974 + const struct virtio_transport *t;
3975 + struct virtio_vsock_pkt *reply;
3976 struct virtio_vsock_pkt_info info = {
3977 .op = VIRTIO_VSOCK_OP_RST,
3978 .type = le16_to_cpu(pkt->hdr.type),
3979 @@ -672,15 +674,21 @@ static int virtio_transport_reset_no_sock(struct virtio_vsock_pkt *pkt)
3980 if (le16_to_cpu(pkt->hdr.op) == VIRTIO_VSOCK_OP_RST)
3981 return 0;
3982
3983 - pkt = virtio_transport_alloc_pkt(&info, 0,
3984 - le64_to_cpu(pkt->hdr.dst_cid),
3985 - le32_to_cpu(pkt->hdr.dst_port),
3986 - le64_to_cpu(pkt->hdr.src_cid),
3987 - le32_to_cpu(pkt->hdr.src_port));
3988 - if (!pkt)
3989 + reply = virtio_transport_alloc_pkt(&info, 0,
3990 + le64_to_cpu(pkt->hdr.dst_cid),
3991 + le32_to_cpu(pkt->hdr.dst_port),
3992 + le64_to_cpu(pkt->hdr.src_cid),
3993 + le32_to_cpu(pkt->hdr.src_port));
3994 + if (!reply)
3995 return -ENOMEM;
3996
3997 - return virtio_transport_get_ops()->send_pkt(pkt);
3998 + t = virtio_transport_get_ops();
3999 + if (!t) {
4000 + virtio_transport_free_pkt(reply);
4001 + return -ENOTCONN;
4002 + }
4003 +
4004 + return t->send_pkt(reply);
4005 }
4006
4007 static void virtio_transport_wait_close(struct sock *sk, long timeout)
4008 diff --git a/sound/pci/hda/patch_ca0132.c b/sound/pci/hda/patch_ca0132.c
4009 index 80f73810b21b..0436789e7cd8 100644
4010 --- a/sound/pci/hda/patch_ca0132.c
4011 +++ b/sound/pci/hda/patch_ca0132.c
4012 @@ -7394,8 +7394,10 @@ static void ca0132_free(struct hda_codec *codec)
4013 ca0132_exit_chip(codec);
4014
4015 snd_hda_power_down(codec);
4016 - if (IS_ENABLED(CONFIG_PCI) && spec->mem_base)
4017 +#ifdef CONFIG_PCI
4018 + if (spec->mem_base)
4019 pci_iounmap(codec->bus->pci, spec->mem_base);
4020 +#endif
4021 kfree(spec->spec_init_verbs);
4022 kfree(codec->spec);
4023 }