Contents of /trunk/kernel-alx-legacy/patches-4.9/0343-4.9.244-all-fixes.patch
Parent Directory | Revision Log
Revision 3645 -
(show annotations)
(download)
Mon Oct 24 14:07:25 2022 UTC (23 months ago) by niro
File size: 130164 byte(s)
Mon Oct 24 14:07:25 2022 UTC (23 months ago) by niro
File size: 130164 byte(s)
-linux-4.9.244
1 | diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt |
2 | index 40602517ca528..e51e42d9f6460 100644 |
3 | --- a/Documentation/kernel-parameters.txt |
4 | +++ b/Documentation/kernel-parameters.txt |
5 | @@ -5020,6 +5020,14 @@ bytes respectively. Such letter suffixes can also be entirely omitted. |
6 | Disables the PV optimizations forcing the HVM guest to |
7 | run as generic HVM guest with no PV drivers. |
8 | |
9 | + xen.event_eoi_delay= [XEN] |
10 | + How long to delay EOI handling in case of event |
11 | + storms (jiffies). Default is 10. |
12 | + |
13 | + xen.event_loop_timeout= [XEN] |
14 | + After which time (jiffies) the event handling loop |
15 | + should start to delay EOI handling. Default is 2. |
16 | + |
17 | xirc2ps_cs= [NET,PCMCIA] |
18 | Format: |
19 | <irq>,<irq_mask>,<io>,<full_duplex>,<do_sound>,<lockup_hack>[,<irq2>[,<irq3>[,<irq4>]]] |
20 | diff --git a/Makefile b/Makefile |
21 | index c6fcfe4bfeed5..27314b9f0fe67 100644 |
22 | --- a/Makefile |
23 | +++ b/Makefile |
24 | @@ -1,6 +1,6 @@ |
25 | VERSION = 4 |
26 | PATCHLEVEL = 9 |
27 | -SUBLEVEL = 243 |
28 | +SUBLEVEL = 244 |
29 | EXTRAVERSION = |
30 | NAME = Roaring Lionus |
31 | |
32 | diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c |
33 | index df60b58691e7c..1808c57ce1614 100644 |
34 | --- a/arch/x86/events/intel/pt.c |
35 | +++ b/arch/x86/events/intel/pt.c |
36 | @@ -1117,7 +1117,7 @@ static int pt_event_addr_filters_validate(struct list_head *filters) |
37 | if (!filter->range || !filter->size) |
38 | return -EOPNOTSUPP; |
39 | |
40 | - if (!filter->inode) { |
41 | + if (!filter->path.dentry) { |
42 | if (!valid_kernel_ip(filter->offset)) |
43 | return -EINVAL; |
44 | |
45 | @@ -1144,7 +1144,7 @@ static void pt_event_addr_filters_sync(struct perf_event *event) |
46 | return; |
47 | |
48 | list_for_each_entry(filter, &head->list, entry) { |
49 | - if (filter->inode && !offs[range]) { |
50 | + if (filter->path.dentry && !offs[range]) { |
51 | msr_a = msr_b = 0; |
52 | } else { |
53 | /* apply the offset */ |
54 | diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c |
55 | index 85c1cc0305f39..f8a7aba4b0959 100644 |
56 | --- a/arch/x86/kernel/cpu/bugs.c |
57 | +++ b/arch/x86/kernel/cpu/bugs.c |
58 | @@ -1248,6 +1248,14 @@ static int ssb_prctl_set(struct task_struct *task, unsigned long ctrl) |
59 | return 0; |
60 | } |
61 | |
62 | +static bool is_spec_ib_user_controlled(void) |
63 | +{ |
64 | + return spectre_v2_user_ibpb == SPECTRE_V2_USER_PRCTL || |
65 | + spectre_v2_user_ibpb == SPECTRE_V2_USER_SECCOMP || |
66 | + spectre_v2_user_stibp == SPECTRE_V2_USER_PRCTL || |
67 | + spectre_v2_user_stibp == SPECTRE_V2_USER_SECCOMP; |
68 | +} |
69 | + |
70 | static int ib_prctl_set(struct task_struct *task, unsigned long ctrl) |
71 | { |
72 | switch (ctrl) { |
73 | @@ -1255,17 +1263,26 @@ static int ib_prctl_set(struct task_struct *task, unsigned long ctrl) |
74 | if (spectre_v2_user_ibpb == SPECTRE_V2_USER_NONE && |
75 | spectre_v2_user_stibp == SPECTRE_V2_USER_NONE) |
76 | return 0; |
77 | - /* |
78 | - * Indirect branch speculation is always disabled in strict |
79 | - * mode. It can neither be enabled if it was force-disabled |
80 | - * by a previous prctl call. |
81 | |
82 | + /* |
83 | + * With strict mode for both IBPB and STIBP, the instruction |
84 | + * code paths avoid checking this task flag and instead, |
85 | + * unconditionally run the instruction. However, STIBP and IBPB |
86 | + * are independent and either can be set to conditionally |
87 | + * enabled regardless of the mode of the other. |
88 | + * |
89 | + * If either is set to conditional, allow the task flag to be |
90 | + * updated, unless it was force-disabled by a previous prctl |
91 | + * call. Currently, this is possible on an AMD CPU which has the |
92 | + * feature X86_FEATURE_AMD_STIBP_ALWAYS_ON. In this case, if the |
93 | + * kernel is booted with 'spectre_v2_user=seccomp', then |
94 | + * spectre_v2_user_ibpb == SPECTRE_V2_USER_SECCOMP and |
95 | + * spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT_PREFERRED. |
96 | */ |
97 | - if (spectre_v2_user_ibpb == SPECTRE_V2_USER_STRICT || |
98 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT || |
99 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT_PREFERRED || |
100 | + if (!is_spec_ib_user_controlled() || |
101 | task_spec_ib_force_disable(task)) |
102 | return -EPERM; |
103 | + |
104 | task_clear_spec_ib_disable(task); |
105 | task_update_spec_tif(task); |
106 | break; |
107 | @@ -1278,10 +1295,10 @@ static int ib_prctl_set(struct task_struct *task, unsigned long ctrl) |
108 | if (spectre_v2_user_ibpb == SPECTRE_V2_USER_NONE && |
109 | spectre_v2_user_stibp == SPECTRE_V2_USER_NONE) |
110 | return -EPERM; |
111 | - if (spectre_v2_user_ibpb == SPECTRE_V2_USER_STRICT || |
112 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT || |
113 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT_PREFERRED) |
114 | + |
115 | + if (!is_spec_ib_user_controlled()) |
116 | return 0; |
117 | + |
118 | task_set_spec_ib_disable(task); |
119 | if (ctrl == PR_SPEC_FORCE_DISABLE) |
120 | task_set_spec_ib_force_disable(task); |
121 | @@ -1344,20 +1361,17 @@ static int ib_prctl_get(struct task_struct *task) |
122 | if (spectre_v2_user_ibpb == SPECTRE_V2_USER_NONE && |
123 | spectre_v2_user_stibp == SPECTRE_V2_USER_NONE) |
124 | return PR_SPEC_ENABLE; |
125 | - else if (spectre_v2_user_ibpb == SPECTRE_V2_USER_STRICT || |
126 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT || |
127 | - spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT_PREFERRED) |
128 | - return PR_SPEC_DISABLE; |
129 | - else if (spectre_v2_user_ibpb == SPECTRE_V2_USER_PRCTL || |
130 | - spectre_v2_user_ibpb == SPECTRE_V2_USER_SECCOMP || |
131 | - spectre_v2_user_stibp == SPECTRE_V2_USER_PRCTL || |
132 | - spectre_v2_user_stibp == SPECTRE_V2_USER_SECCOMP) { |
133 | + else if (is_spec_ib_user_controlled()) { |
134 | if (task_spec_ib_force_disable(task)) |
135 | return PR_SPEC_PRCTL | PR_SPEC_FORCE_DISABLE; |
136 | if (task_spec_ib_disable(task)) |
137 | return PR_SPEC_PRCTL | PR_SPEC_DISABLE; |
138 | return PR_SPEC_PRCTL | PR_SPEC_ENABLE; |
139 | - } else |
140 | + } else if (spectre_v2_user_ibpb == SPECTRE_V2_USER_STRICT || |
141 | + spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT || |
142 | + spectre_v2_user_stibp == SPECTRE_V2_USER_STRICT_PREFERRED) |
143 | + return PR_SPEC_DISABLE; |
144 | + else |
145 | return PR_SPEC_NOT_AFFECTED; |
146 | } |
147 | |
148 | diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c |
149 | index a700e525535ce..4f643a87f9c7d 100644 |
150 | --- a/drivers/block/xen-blkback/blkback.c |
151 | +++ b/drivers/block/xen-blkback/blkback.c |
152 | @@ -183,7 +183,7 @@ static inline void shrink_free_pagepool(struct xen_blkif_ring *ring, int num) |
153 | |
154 | #define vaddr(page) ((unsigned long)pfn_to_kaddr(page_to_pfn(page))) |
155 | |
156 | -static int do_block_io_op(struct xen_blkif_ring *ring); |
157 | +static int do_block_io_op(struct xen_blkif_ring *ring, unsigned int *eoi_flags); |
158 | static int dispatch_rw_block_io(struct xen_blkif_ring *ring, |
159 | struct blkif_request *req, |
160 | struct pending_req *pending_req); |
161 | @@ -608,6 +608,8 @@ int xen_blkif_schedule(void *arg) |
162 | struct xen_vbd *vbd = &blkif->vbd; |
163 | unsigned long timeout; |
164 | int ret; |
165 | + bool do_eoi; |
166 | + unsigned int eoi_flags = XEN_EOI_FLAG_SPURIOUS; |
167 | |
168 | set_freezable(); |
169 | while (!kthread_should_stop()) { |
170 | @@ -632,16 +634,23 @@ int xen_blkif_schedule(void *arg) |
171 | if (timeout == 0) |
172 | goto purge_gnt_list; |
173 | |
174 | + do_eoi = ring->waiting_reqs; |
175 | + |
176 | ring->waiting_reqs = 0; |
177 | smp_mb(); /* clear flag *before* checking for work */ |
178 | |
179 | - ret = do_block_io_op(ring); |
180 | + ret = do_block_io_op(ring, &eoi_flags); |
181 | if (ret > 0) |
182 | ring->waiting_reqs = 1; |
183 | if (ret == -EACCES) |
184 | wait_event_interruptible(ring->shutdown_wq, |
185 | kthread_should_stop()); |
186 | |
187 | + if (do_eoi && !ring->waiting_reqs) { |
188 | + xen_irq_lateeoi(ring->irq, eoi_flags); |
189 | + eoi_flags |= XEN_EOI_FLAG_SPURIOUS; |
190 | + } |
191 | + |
192 | purge_gnt_list: |
193 | if (blkif->vbd.feature_gnt_persistent && |
194 | time_after(jiffies, ring->next_lru)) { |
195 | @@ -1117,7 +1126,7 @@ static void end_block_io_op(struct bio *bio) |
196 | * and transmute it to the block API to hand it over to the proper block disk. |
197 | */ |
198 | static int |
199 | -__do_block_io_op(struct xen_blkif_ring *ring) |
200 | +__do_block_io_op(struct xen_blkif_ring *ring, unsigned int *eoi_flags) |
201 | { |
202 | union blkif_back_rings *blk_rings = &ring->blk_rings; |
203 | struct blkif_request req; |
204 | @@ -1140,6 +1149,9 @@ __do_block_io_op(struct xen_blkif_ring *ring) |
205 | if (RING_REQUEST_CONS_OVERFLOW(&blk_rings->common, rc)) |
206 | break; |
207 | |
208 | + /* We've seen a request, so clear spurious eoi flag. */ |
209 | + *eoi_flags &= ~XEN_EOI_FLAG_SPURIOUS; |
210 | + |
211 | if (kthread_should_stop()) { |
212 | more_to_do = 1; |
213 | break; |
214 | @@ -1198,13 +1210,13 @@ done: |
215 | } |
216 | |
217 | static int |
218 | -do_block_io_op(struct xen_blkif_ring *ring) |
219 | +do_block_io_op(struct xen_blkif_ring *ring, unsigned int *eoi_flags) |
220 | { |
221 | union blkif_back_rings *blk_rings = &ring->blk_rings; |
222 | int more_to_do; |
223 | |
224 | do { |
225 | - more_to_do = __do_block_io_op(ring); |
226 | + more_to_do = __do_block_io_op(ring, eoi_flags); |
227 | if (more_to_do) |
228 | break; |
229 | |
230 | diff --git a/drivers/block/xen-blkback/xenbus.c b/drivers/block/xen-blkback/xenbus.c |
231 | index 1d1f866579679..702ebfc4face5 100644 |
232 | --- a/drivers/block/xen-blkback/xenbus.c |
233 | +++ b/drivers/block/xen-blkback/xenbus.c |
234 | @@ -236,9 +236,8 @@ static int xen_blkif_map(struct xen_blkif_ring *ring, grant_ref_t *gref, |
235 | BUG(); |
236 | } |
237 | |
238 | - err = bind_interdomain_evtchn_to_irqhandler(blkif->domid, evtchn, |
239 | - xen_blkif_be_int, 0, |
240 | - "blkif-backend", ring); |
241 | + err = bind_interdomain_evtchn_to_irqhandler_lateeoi(blkif->domid, |
242 | + evtchn, xen_blkif_be_int, 0, "blkif-backend", ring); |
243 | if (err < 0) { |
244 | xenbus_unmap_ring_vfree(blkif->be->dev, ring->blk_ring); |
245 | ring->blk_rings.common.sring = NULL; |
246 | diff --git a/drivers/char/random.c b/drivers/char/random.c |
247 | index c417aa19f9962..4cbc73173701d 100644 |
248 | --- a/drivers/char/random.c |
249 | +++ b/drivers/char/random.c |
250 | @@ -1211,7 +1211,6 @@ void add_interrupt_randomness(int irq, int irq_flags) |
251 | |
252 | fast_mix(fast_pool); |
253 | add_interrupt_bench(cycles); |
254 | - this_cpu_add(net_rand_state.s1, fast_pool->pool[cycles & 3]); |
255 | |
256 | if (unlikely(crng_init == 0)) { |
257 | if ((fast_pool->count >= 64) && |
258 | diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c |
259 | index cb952acc71339..2934443fbd4dc 100644 |
260 | --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c |
261 | +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c |
262 | @@ -1053,22 +1053,19 @@ static int cik_sdma_soft_reset(void *handle) |
263 | { |
264 | u32 srbm_soft_reset = 0; |
265 | struct amdgpu_device *adev = (struct amdgpu_device *)handle; |
266 | - u32 tmp = RREG32(mmSRBM_STATUS2); |
267 | + u32 tmp; |
268 | |
269 | - if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) { |
270 | - /* sdma0 */ |
271 | - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET); |
272 | - tmp |= SDMA0_F32_CNTL__HALT_MASK; |
273 | - WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp); |
274 | - srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK; |
275 | - } |
276 | - if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK) { |
277 | - /* sdma1 */ |
278 | - tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET); |
279 | - tmp |= SDMA0_F32_CNTL__HALT_MASK; |
280 | - WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp); |
281 | - srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK; |
282 | - } |
283 | + /* sdma0 */ |
284 | + tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET); |
285 | + tmp |= SDMA0_F32_CNTL__HALT_MASK; |
286 | + WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp); |
287 | + srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK; |
288 | + |
289 | + /* sdma1 */ |
290 | + tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET); |
291 | + tmp |= SDMA0_F32_CNTL__HALT_MASK; |
292 | + WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp); |
293 | + srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK; |
294 | |
295 | if (srbm_soft_reset) { |
296 | tmp = RREG32(mmSRBM_SOFT_RESET); |
297 | diff --git a/drivers/gpu/drm/gma500/psb_irq.c b/drivers/gpu/drm/gma500/psb_irq.c |
298 | index 78eb109028091..076b6da44f461 100644 |
299 | --- a/drivers/gpu/drm/gma500/psb_irq.c |
300 | +++ b/drivers/gpu/drm/gma500/psb_irq.c |
301 | @@ -350,6 +350,7 @@ int psb_irq_postinstall(struct drm_device *dev) |
302 | { |
303 | struct drm_psb_private *dev_priv = dev->dev_private; |
304 | unsigned long irqflags; |
305 | + unsigned int i; |
306 | |
307 | spin_lock_irqsave(&dev_priv->irqmask_lock, irqflags); |
308 | |
309 | @@ -362,20 +363,12 @@ int psb_irq_postinstall(struct drm_device *dev) |
310 | PSB_WVDC32(dev_priv->vdc_irq_mask, PSB_INT_ENABLE_R); |
311 | PSB_WVDC32(0xFFFFFFFF, PSB_HWSTAM); |
312 | |
313 | - if (dev->vblank[0].enabled) |
314 | - psb_enable_pipestat(dev_priv, 0, PIPE_VBLANK_INTERRUPT_ENABLE); |
315 | - else |
316 | - psb_disable_pipestat(dev_priv, 0, PIPE_VBLANK_INTERRUPT_ENABLE); |
317 | - |
318 | - if (dev->vblank[1].enabled) |
319 | - psb_enable_pipestat(dev_priv, 1, PIPE_VBLANK_INTERRUPT_ENABLE); |
320 | - else |
321 | - psb_disable_pipestat(dev_priv, 1, PIPE_VBLANK_INTERRUPT_ENABLE); |
322 | - |
323 | - if (dev->vblank[2].enabled) |
324 | - psb_enable_pipestat(dev_priv, 2, PIPE_VBLANK_INTERRUPT_ENABLE); |
325 | - else |
326 | - psb_disable_pipestat(dev_priv, 2, PIPE_VBLANK_INTERRUPT_ENABLE); |
327 | + for (i = 0; i < dev->num_crtcs; ++i) { |
328 | + if (dev->vblank[i].enabled) |
329 | + psb_enable_pipestat(dev_priv, i, PIPE_VBLANK_INTERRUPT_ENABLE); |
330 | + else |
331 | + psb_disable_pipestat(dev_priv, i, PIPE_VBLANK_INTERRUPT_ENABLE); |
332 | + } |
333 | |
334 | if (dev_priv->ops->hotplug_enable) |
335 | dev_priv->ops->hotplug_enable(dev, true); |
336 | @@ -388,6 +381,7 @@ void psb_irq_uninstall(struct drm_device *dev) |
337 | { |
338 | struct drm_psb_private *dev_priv = dev->dev_private; |
339 | unsigned long irqflags; |
340 | + unsigned int i; |
341 | |
342 | spin_lock_irqsave(&dev_priv->irqmask_lock, irqflags); |
343 | |
344 | @@ -396,14 +390,10 @@ void psb_irq_uninstall(struct drm_device *dev) |
345 | |
346 | PSB_WVDC32(0xFFFFFFFF, PSB_HWSTAM); |
347 | |
348 | - if (dev->vblank[0].enabled) |
349 | - psb_disable_pipestat(dev_priv, 0, PIPE_VBLANK_INTERRUPT_ENABLE); |
350 | - |
351 | - if (dev->vblank[1].enabled) |
352 | - psb_disable_pipestat(dev_priv, 1, PIPE_VBLANK_INTERRUPT_ENABLE); |
353 | - |
354 | - if (dev->vblank[2].enabled) |
355 | - psb_disable_pipestat(dev_priv, 2, PIPE_VBLANK_INTERRUPT_ENABLE); |
356 | + for (i = 0; i < dev->num_crtcs; ++i) { |
357 | + if (dev->vblank[i].enabled) |
358 | + psb_disable_pipestat(dev_priv, i, PIPE_VBLANK_INTERRUPT_ENABLE); |
359 | + } |
360 | |
361 | dev_priv->vdc_irq_mask &= _PSB_IRQ_SGX_FLAG | |
362 | _PSB_IRQ_MSVDX_FLAG | |
363 | diff --git a/drivers/iommu/amd_iommu_types.h b/drivers/iommu/amd_iommu_types.h |
364 | index da3fbf82d1cf4..e19c05d9e84ba 100644 |
365 | --- a/drivers/iommu/amd_iommu_types.h |
366 | +++ b/drivers/iommu/amd_iommu_types.h |
367 | @@ -383,7 +383,11 @@ extern bool amd_iommu_np_cache; |
368 | /* Only true if all IOMMUs support device IOTLBs */ |
369 | extern bool amd_iommu_iotlb_sup; |
370 | |
371 | -#define MAX_IRQS_PER_TABLE 256 |
372 | +/* |
373 | + * AMD IOMMU hardware only support 512 IRTEs despite |
374 | + * the architectural limitation of 2048 entries. |
375 | + */ |
376 | +#define MAX_IRQS_PER_TABLE 512 |
377 | #define IRQ_TABLE_ALIGNMENT 128 |
378 | |
379 | struct irq_remap_table { |
380 | diff --git a/drivers/misc/mei/client.h b/drivers/misc/mei/client.h |
381 | index d2bfabecd882c..f9d3211ce5994 100644 |
382 | --- a/drivers/misc/mei/client.h |
383 | +++ b/drivers/misc/mei/client.h |
384 | @@ -152,11 +152,11 @@ static inline u8 mei_cl_me_id(const struct mei_cl *cl) |
385 | * |
386 | * @cl: host client |
387 | * |
388 | - * Return: mtu |
389 | + * Return: mtu or 0 if client is not connected |
390 | */ |
391 | static inline size_t mei_cl_mtu(const struct mei_cl *cl) |
392 | { |
393 | - return cl->me_cl->props.max_msg_length; |
394 | + return cl->me_cl ? cl->me_cl->props.max_msg_length : 0; |
395 | } |
396 | |
397 | /** |
398 | diff --git a/drivers/net/can/dev.c b/drivers/net/can/dev.c |
399 | index ffc5467a1ec2b..617eb75c7c0ce 100644 |
400 | --- a/drivers/net/can/dev.c |
401 | +++ b/drivers/net/can/dev.c |
402 | @@ -469,9 +469,13 @@ struct sk_buff *__can_get_echo_skb(struct net_device *dev, unsigned int idx, u8 |
403 | */ |
404 | struct sk_buff *skb = priv->echo_skb[idx]; |
405 | struct canfd_frame *cf = (struct canfd_frame *)skb->data; |
406 | - u8 len = cf->len; |
407 | |
408 | - *len_ptr = len; |
409 | + /* get the real payload length for netdev statistics */ |
410 | + if (cf->can_id & CAN_RTR_FLAG) |
411 | + *len_ptr = 0; |
412 | + else |
413 | + *len_ptr = cf->len; |
414 | + |
415 | priv->echo_skb[idx] = NULL; |
416 | |
417 | return skb; |
418 | @@ -496,7 +500,11 @@ unsigned int can_get_echo_skb(struct net_device *dev, unsigned int idx) |
419 | if (!skb) |
420 | return 0; |
421 | |
422 | - netif_rx(skb); |
423 | + skb_get(skb); |
424 | + if (netif_rx(skb) == NET_RX_SUCCESS) |
425 | + dev_consume_skb_any(skb); |
426 | + else |
427 | + dev_kfree_skb_any(skb); |
428 | |
429 | return len; |
430 | } |
431 | diff --git a/drivers/net/can/usb/peak_usb/pcan_usb_core.c b/drivers/net/can/usb/peak_usb/pcan_usb_core.c |
432 | index 6cd4317fe94df..74b37309efab7 100644 |
433 | --- a/drivers/net/can/usb/peak_usb/pcan_usb_core.c |
434 | +++ b/drivers/net/can/usb/peak_usb/pcan_usb_core.c |
435 | @@ -152,14 +152,55 @@ void peak_usb_get_ts_tv(struct peak_time_ref *time_ref, u32 ts, |
436 | /* protect from getting timeval before setting now */ |
437 | if (time_ref->tv_host.tv_sec > 0) { |
438 | u64 delta_us; |
439 | + s64 delta_ts = 0; |
440 | + |
441 | + /* General case: dev_ts_1 < dev_ts_2 < ts, with: |
442 | + * |
443 | + * - dev_ts_1 = previous sync timestamp |
444 | + * - dev_ts_2 = last sync timestamp |
445 | + * - ts = event timestamp |
446 | + * - ts_period = known sync period (theoretical) |
447 | + * ~ dev_ts2 - dev_ts1 |
448 | + * *but*: |
449 | + * |
450 | + * - time counters wrap (see adapter->ts_used_bits) |
451 | + * - sometimes, dev_ts_1 < ts < dev_ts2 |
452 | + * |
453 | + * "normal" case (sync time counters increase): |
454 | + * must take into account case when ts wraps (tsw) |
455 | + * |
456 | + * < ts_period > < > |
457 | + * | | | |
458 | + * ---+--------+----+-------0-+--+--> |
459 | + * ts_dev_1 | ts_dev_2 | |
460 | + * ts tsw |
461 | + */ |
462 | + if (time_ref->ts_dev_1 < time_ref->ts_dev_2) { |
463 | + /* case when event time (tsw) wraps */ |
464 | + if (ts < time_ref->ts_dev_1) |
465 | + delta_ts = 1 << time_ref->adapter->ts_used_bits; |
466 | + |
467 | + /* Otherwise, sync time counter (ts_dev_2) has wrapped: |
468 | + * handle case when event time (tsn) hasn't. |
469 | + * |
470 | + * < ts_period > < > |
471 | + * | | | |
472 | + * ---+--------+--0-+---------+--+--> |
473 | + * ts_dev_1 | ts_dev_2 | |
474 | + * tsn ts |
475 | + */ |
476 | + } else if (time_ref->ts_dev_1 < ts) { |
477 | + delta_ts = -(1 << time_ref->adapter->ts_used_bits); |
478 | + } |
479 | |
480 | - delta_us = ts - time_ref->ts_dev_2; |
481 | - if (ts < time_ref->ts_dev_2) |
482 | - delta_us &= (1 << time_ref->adapter->ts_used_bits) - 1; |
483 | + /* add delay between last sync and event timestamps */ |
484 | + delta_ts += (signed int)(ts - time_ref->ts_dev_2); |
485 | |
486 | - delta_us += time_ref->ts_total; |
487 | + /* add time from beginning to last sync */ |
488 | + delta_ts += time_ref->ts_total; |
489 | |
490 | - delta_us *= time_ref->adapter->us_per_ts_scale; |
491 | + /* convert ticks number into microseconds */ |
492 | + delta_us = delta_ts * time_ref->adapter->us_per_ts_scale; |
493 | delta_us >>= time_ref->adapter->us_per_ts_shift; |
494 | |
495 | *tv = time_ref->tv_host_0; |
496 | diff --git a/drivers/net/can/usb/peak_usb/pcan_usb_fd.c b/drivers/net/can/usb/peak_usb/pcan_usb_fd.c |
497 | index 40647b837b31f..d314e73f3d061 100644 |
498 | --- a/drivers/net/can/usb/peak_usb/pcan_usb_fd.c |
499 | +++ b/drivers/net/can/usb/peak_usb/pcan_usb_fd.c |
500 | @@ -475,12 +475,18 @@ static int pcan_usb_fd_decode_canmsg(struct pcan_usb_fd_if *usb_if, |
501 | struct pucan_msg *rx_msg) |
502 | { |
503 | struct pucan_rx_msg *rm = (struct pucan_rx_msg *)rx_msg; |
504 | - struct peak_usb_device *dev = usb_if->dev[pucan_msg_get_channel(rm)]; |
505 | - struct net_device *netdev = dev->netdev; |
506 | + struct peak_usb_device *dev; |
507 | + struct net_device *netdev; |
508 | struct canfd_frame *cfd; |
509 | struct sk_buff *skb; |
510 | const u16 rx_msg_flags = le16_to_cpu(rm->flags); |
511 | |
512 | + if (pucan_msg_get_channel(rm) >= ARRAY_SIZE(usb_if->dev)) |
513 | + return -ENOMEM; |
514 | + |
515 | + dev = usb_if->dev[pucan_msg_get_channel(rm)]; |
516 | + netdev = dev->netdev; |
517 | + |
518 | if (rx_msg_flags & PUCAN_MSG_EXT_DATA_LEN) { |
519 | /* CANFD frame case */ |
520 | skb = alloc_canfd_skb(netdev, &cfd); |
521 | @@ -527,15 +533,21 @@ static int pcan_usb_fd_decode_status(struct pcan_usb_fd_if *usb_if, |
522 | struct pucan_msg *rx_msg) |
523 | { |
524 | struct pucan_status_msg *sm = (struct pucan_status_msg *)rx_msg; |
525 | - struct peak_usb_device *dev = usb_if->dev[pucan_stmsg_get_channel(sm)]; |
526 | - struct pcan_usb_fd_device *pdev = |
527 | - container_of(dev, struct pcan_usb_fd_device, dev); |
528 | + struct pcan_usb_fd_device *pdev; |
529 | enum can_state new_state = CAN_STATE_ERROR_ACTIVE; |
530 | enum can_state rx_state, tx_state; |
531 | - struct net_device *netdev = dev->netdev; |
532 | + struct peak_usb_device *dev; |
533 | + struct net_device *netdev; |
534 | struct can_frame *cf; |
535 | struct sk_buff *skb; |
536 | |
537 | + if (pucan_stmsg_get_channel(sm) >= ARRAY_SIZE(usb_if->dev)) |
538 | + return -ENOMEM; |
539 | + |
540 | + dev = usb_if->dev[pucan_stmsg_get_channel(sm)]; |
541 | + pdev = container_of(dev, struct pcan_usb_fd_device, dev); |
542 | + netdev = dev->netdev; |
543 | + |
544 | /* nothing should be sent while in BUS_OFF state */ |
545 | if (dev->can.state == CAN_STATE_BUS_OFF) |
546 | return 0; |
547 | @@ -588,9 +600,14 @@ static int pcan_usb_fd_decode_error(struct pcan_usb_fd_if *usb_if, |
548 | struct pucan_msg *rx_msg) |
549 | { |
550 | struct pucan_error_msg *er = (struct pucan_error_msg *)rx_msg; |
551 | - struct peak_usb_device *dev = usb_if->dev[pucan_ermsg_get_channel(er)]; |
552 | - struct pcan_usb_fd_device *pdev = |
553 | - container_of(dev, struct pcan_usb_fd_device, dev); |
554 | + struct pcan_usb_fd_device *pdev; |
555 | + struct peak_usb_device *dev; |
556 | + |
557 | + if (pucan_ermsg_get_channel(er) >= ARRAY_SIZE(usb_if->dev)) |
558 | + return -EINVAL; |
559 | + |
560 | + dev = usb_if->dev[pucan_ermsg_get_channel(er)]; |
561 | + pdev = container_of(dev, struct pcan_usb_fd_device, dev); |
562 | |
563 | /* keep a trace of tx and rx error counters for later use */ |
564 | pdev->bec.txerr = er->tx_err_cnt; |
565 | @@ -604,11 +621,17 @@ static int pcan_usb_fd_decode_overrun(struct pcan_usb_fd_if *usb_if, |
566 | struct pucan_msg *rx_msg) |
567 | { |
568 | struct pcan_ufd_ovr_msg *ov = (struct pcan_ufd_ovr_msg *)rx_msg; |
569 | - struct peak_usb_device *dev = usb_if->dev[pufd_omsg_get_channel(ov)]; |
570 | - struct net_device *netdev = dev->netdev; |
571 | + struct peak_usb_device *dev; |
572 | + struct net_device *netdev; |
573 | struct can_frame *cf; |
574 | struct sk_buff *skb; |
575 | |
576 | + if (pufd_omsg_get_channel(ov) >= ARRAY_SIZE(usb_if->dev)) |
577 | + return -EINVAL; |
578 | + |
579 | + dev = usb_if->dev[pufd_omsg_get_channel(ov)]; |
580 | + netdev = dev->netdev; |
581 | + |
582 | /* allocate an skb to store the error frame */ |
583 | skb = alloc_can_err_skb(netdev, &cf); |
584 | if (!skb) |
585 | @@ -726,6 +749,9 @@ static int pcan_usb_fd_encode_msg(struct peak_usb_device *dev, |
586 | u16 tx_msg_size, tx_msg_flags; |
587 | u8 can_dlc; |
588 | |
589 | + if (cfd->len > CANFD_MAX_DLEN) |
590 | + return -EINVAL; |
591 | + |
592 | tx_msg_size = ALIGN(sizeof(struct pucan_tx_msg) + cfd->len, 4); |
593 | tx_msg->size = cpu_to_le16(tx_msg_size); |
594 | tx_msg->type = cpu_to_le16(PUCAN_MSG_CAN_TX); |
595 | diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c |
596 | index 7484ad3c955db..e3e02ec8f1498 100644 |
597 | --- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c |
598 | +++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c |
599 | @@ -188,7 +188,7 @@ static inline bool i40e_vc_isvalid_vsi_id(struct i40e_vf *vf, u16 vsi_id) |
600 | * check for the valid queue id |
601 | **/ |
602 | static inline bool i40e_vc_isvalid_queue_id(struct i40e_vf *vf, u16 vsi_id, |
603 | - u8 qid) |
604 | + u16 qid) |
605 | { |
606 | struct i40e_pf *pf = vf->pf; |
607 | struct i40e_vsi *vsi = i40e_find_vsi_from_id(pf, vsi_id); |
608 | @@ -203,7 +203,7 @@ static inline bool i40e_vc_isvalid_queue_id(struct i40e_vf *vf, u16 vsi_id, |
609 | * |
610 | * check for the valid vector id |
611 | **/ |
612 | -static inline bool i40e_vc_isvalid_vector_id(struct i40e_vf *vf, u8 vector_id) |
613 | +static inline bool i40e_vc_isvalid_vector_id(struct i40e_vf *vf, u32 vector_id) |
614 | { |
615 | struct i40e_pf *pf = vf->pf; |
616 | |
617 | @@ -417,11 +417,28 @@ static int i40e_config_iwarp_qvlist(struct i40e_vf *vf, |
618 | u32 v_idx, i, reg_idx, reg; |
619 | u32 next_q_idx, next_q_type; |
620 | u32 msix_vf, size; |
621 | + int ret = 0; |
622 | + |
623 | + msix_vf = pf->hw.func_caps.num_msix_vectors_vf; |
624 | + |
625 | + if (qvlist_info->num_vectors > msix_vf) { |
626 | + dev_warn(&pf->pdev->dev, |
627 | + "Incorrect number of iwarp vectors %u. Maximum %u allowed.\n", |
628 | + qvlist_info->num_vectors, |
629 | + msix_vf); |
630 | + ret = -EINVAL; |
631 | + goto err_out; |
632 | + } |
633 | |
634 | size = sizeof(struct i40e_virtchnl_iwarp_qvlist_info) + |
635 | (sizeof(struct i40e_virtchnl_iwarp_qv_info) * |
636 | (qvlist_info->num_vectors - 1)); |
637 | + kfree(vf->qvlist_info); |
638 | vf->qvlist_info = kzalloc(size, GFP_KERNEL); |
639 | + if (!vf->qvlist_info) { |
640 | + ret = -ENOMEM; |
641 | + goto err_out; |
642 | + } |
643 | vf->qvlist_info->num_vectors = qvlist_info->num_vectors; |
644 | |
645 | msix_vf = pf->hw.func_caps.num_msix_vectors_vf; |
646 | @@ -432,8 +449,10 @@ static int i40e_config_iwarp_qvlist(struct i40e_vf *vf, |
647 | v_idx = qv_info->v_idx; |
648 | |
649 | /* Validate vector id belongs to this vf */ |
650 | - if (!i40e_vc_isvalid_vector_id(vf, v_idx)) |
651 | - goto err; |
652 | + if (!i40e_vc_isvalid_vector_id(vf, v_idx)) { |
653 | + ret = -EINVAL; |
654 | + goto err_free; |
655 | + } |
656 | |
657 | vf->qvlist_info->qv_info[i] = *qv_info; |
658 | |
659 | @@ -475,10 +494,11 @@ static int i40e_config_iwarp_qvlist(struct i40e_vf *vf, |
660 | } |
661 | |
662 | return 0; |
663 | -err: |
664 | +err_free: |
665 | kfree(vf->qvlist_info); |
666 | vf->qvlist_info = NULL; |
667 | - return -EINVAL; |
668 | +err_out: |
669 | + return ret; |
670 | } |
671 | |
672 | /** |
673 | diff --git a/drivers/net/geneve.c b/drivers/net/geneve.c |
674 | index 35d8c636de123..d89995f4bd433 100644 |
675 | --- a/drivers/net/geneve.c |
676 | +++ b/drivers/net/geneve.c |
677 | @@ -732,7 +732,8 @@ free_dst: |
678 | static struct rtable *geneve_get_v4_rt(struct sk_buff *skb, |
679 | struct net_device *dev, |
680 | struct flowi4 *fl4, |
681 | - struct ip_tunnel_info *info) |
682 | + struct ip_tunnel_info *info, |
683 | + __be16 dport, __be16 sport) |
684 | { |
685 | bool use_cache = ip_tunnel_dst_cache_usable(skb, info); |
686 | struct geneve_dev *geneve = netdev_priv(dev); |
687 | @@ -746,6 +747,8 @@ static struct rtable *geneve_get_v4_rt(struct sk_buff *skb, |
688 | memset(fl4, 0, sizeof(*fl4)); |
689 | fl4->flowi4_mark = skb->mark; |
690 | fl4->flowi4_proto = IPPROTO_UDP; |
691 | + fl4->fl4_dport = dport; |
692 | + fl4->fl4_sport = sport; |
693 | |
694 | if (info) { |
695 | fl4->daddr = info->key.u.ipv4.dst; |
696 | @@ -791,7 +794,8 @@ static struct rtable *geneve_get_v4_rt(struct sk_buff *skb, |
697 | static struct dst_entry *geneve_get_v6_dst(struct sk_buff *skb, |
698 | struct net_device *dev, |
699 | struct flowi6 *fl6, |
700 | - struct ip_tunnel_info *info) |
701 | + struct ip_tunnel_info *info, |
702 | + __be16 dport, __be16 sport) |
703 | { |
704 | bool use_cache = ip_tunnel_dst_cache_usable(skb, info); |
705 | struct geneve_dev *geneve = netdev_priv(dev); |
706 | @@ -807,6 +811,8 @@ static struct dst_entry *geneve_get_v6_dst(struct sk_buff *skb, |
707 | memset(fl6, 0, sizeof(*fl6)); |
708 | fl6->flowi6_mark = skb->mark; |
709 | fl6->flowi6_proto = IPPROTO_UDP; |
710 | + fl6->fl6_dport = dport; |
711 | + fl6->fl6_sport = sport; |
712 | |
713 | if (info) { |
714 | fl6->daddr = info->key.u.ipv6.dst; |
715 | @@ -894,13 +900,14 @@ static netdev_tx_t geneve_xmit_skb(struct sk_buff *skb, struct net_device *dev, |
716 | goto tx_error; |
717 | } |
718 | |
719 | - rt = geneve_get_v4_rt(skb, dev, &fl4, info); |
720 | + sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true); |
721 | + rt = geneve_get_v4_rt(skb, dev, &fl4, info, |
722 | + geneve->dst_port, sport); |
723 | if (IS_ERR(rt)) { |
724 | err = PTR_ERR(rt); |
725 | goto tx_error; |
726 | } |
727 | |
728 | - sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true); |
729 | skb_reset_mac_header(skb); |
730 | |
731 | if (info) { |
732 | @@ -983,13 +990,14 @@ static netdev_tx_t geneve6_xmit_skb(struct sk_buff *skb, struct net_device *dev, |
733 | } |
734 | } |
735 | |
736 | - dst = geneve_get_v6_dst(skb, dev, &fl6, info); |
737 | + sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true); |
738 | + dst = geneve_get_v6_dst(skb, dev, &fl6, info, |
739 | + geneve->dst_port, sport); |
740 | if (IS_ERR(dst)) { |
741 | err = PTR_ERR(dst); |
742 | goto tx_error; |
743 | } |
744 | |
745 | - sport = udp_flow_src_port(geneve->net, skb, 1, USHRT_MAX, true); |
746 | skb_reset_mac_header(skb); |
747 | |
748 | if (info) { |
749 | @@ -1114,9 +1122,14 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
750 | struct dst_entry *dst; |
751 | struct flowi6 fl6; |
752 | #endif |
753 | + __be16 sport; |
754 | |
755 | if (ip_tunnel_info_af(info) == AF_INET) { |
756 | - rt = geneve_get_v4_rt(skb, dev, &fl4, info); |
757 | + sport = udp_flow_src_port(geneve->net, skb, |
758 | + 1, USHRT_MAX, true); |
759 | + |
760 | + rt = geneve_get_v4_rt(skb, dev, &fl4, info, |
761 | + geneve->dst_port, sport); |
762 | if (IS_ERR(rt)) |
763 | return PTR_ERR(rt); |
764 | |
765 | @@ -1124,7 +1137,11 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
766 | info->key.u.ipv4.src = fl4.saddr; |
767 | #if IS_ENABLED(CONFIG_IPV6) |
768 | } else if (ip_tunnel_info_af(info) == AF_INET6) { |
769 | - dst = geneve_get_v6_dst(skb, dev, &fl6, info); |
770 | + sport = udp_flow_src_port(geneve->net, skb, |
771 | + 1, USHRT_MAX, true); |
772 | + |
773 | + dst = geneve_get_v6_dst(skb, dev, &fl6, info, |
774 | + geneve->dst_port, sport); |
775 | if (IS_ERR(dst)) |
776 | return PTR_ERR(dst); |
777 | |
778 | @@ -1135,8 +1152,7 @@ static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
779 | return -EINVAL; |
780 | } |
781 | |
782 | - info->key.tp_src = udp_flow_src_port(geneve->net, skb, |
783 | - 1, USHRT_MAX, true); |
784 | + info->key.tp_src = sport; |
785 | info->key.tp_dst = geneve->dst_port; |
786 | return 0; |
787 | } |
788 | diff --git a/drivers/net/wan/cosa.c b/drivers/net/wan/cosa.c |
789 | index b87fe0a01c69f..3c02473a20f21 100644 |
790 | --- a/drivers/net/wan/cosa.c |
791 | +++ b/drivers/net/wan/cosa.c |
792 | @@ -903,6 +903,7 @@ static ssize_t cosa_write(struct file *file, |
793 | chan->tx_status = 1; |
794 | spin_unlock_irqrestore(&cosa->lock, flags); |
795 | up(&chan->wsem); |
796 | + kfree(kbuf); |
797 | return -ERESTARTSYS; |
798 | } |
799 | } |
800 | diff --git a/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c b/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c |
801 | index 2eb169b204f8d..6a9c9b4ef2c92 100644 |
802 | --- a/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c |
803 | +++ b/drivers/net/wireless/ath/ath9k/htc_drv_txrx.c |
804 | @@ -972,7 +972,7 @@ static bool ath9k_rx_prepare(struct ath9k_htc_priv *priv, |
805 | struct ath_htc_rx_status *rxstatus; |
806 | struct ath_rx_status rx_stats; |
807 | bool decrypt_error = false; |
808 | - __be16 rs_datalen; |
809 | + u16 rs_datalen; |
810 | bool is_phyerr; |
811 | |
812 | if (skb->len < HTC_RX_FRAME_HEADER_SIZE) { |
813 | diff --git a/drivers/net/xen-netback/common.h b/drivers/net/xen-netback/common.h |
814 | index 5b1d2e8402d9d..347c796afd4ed 100644 |
815 | --- a/drivers/net/xen-netback/common.h |
816 | +++ b/drivers/net/xen-netback/common.h |
817 | @@ -140,6 +140,20 @@ struct xenvif_queue { /* Per-queue data for xenvif */ |
818 | char name[QUEUE_NAME_SIZE]; /* DEVNAME-qN */ |
819 | struct xenvif *vif; /* Parent VIF */ |
820 | |
821 | + /* |
822 | + * TX/RX common EOI handling. |
823 | + * When feature-split-event-channels = 0, interrupt handler sets |
824 | + * NETBK_COMMON_EOI, otherwise NETBK_RX_EOI and NETBK_TX_EOI are set |
825 | + * by the RX and TX interrupt handlers. |
826 | + * RX and TX handler threads will issue an EOI when either |
827 | + * NETBK_COMMON_EOI or their specific bits (NETBK_RX_EOI or |
828 | + * NETBK_TX_EOI) are set and they will reset those bits. |
829 | + */ |
830 | + atomic_t eoi_pending; |
831 | +#define NETBK_RX_EOI 0x01 |
832 | +#define NETBK_TX_EOI 0x02 |
833 | +#define NETBK_COMMON_EOI 0x04 |
834 | + |
835 | /* Use NAPI for guest TX */ |
836 | struct napi_struct napi; |
837 | /* When feature-split-event-channels = 0, tx_irq = rx_irq. */ |
838 | @@ -356,6 +370,7 @@ int xenvif_dealloc_kthread(void *data); |
839 | |
840 | irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data); |
841 | |
842 | +bool xenvif_have_rx_work(struct xenvif_queue *queue, bool test_kthread); |
843 | void xenvif_rx_action(struct xenvif_queue *queue); |
844 | void xenvif_rx_queue_tail(struct xenvif_queue *queue, struct sk_buff *skb); |
845 | |
846 | diff --git a/drivers/net/xen-netback/interface.c b/drivers/net/xen-netback/interface.c |
847 | index 46008f2845502..e61073c751428 100644 |
848 | --- a/drivers/net/xen-netback/interface.c |
849 | +++ b/drivers/net/xen-netback/interface.c |
850 | @@ -76,12 +76,28 @@ int xenvif_schedulable(struct xenvif *vif) |
851 | !vif->disabled; |
852 | } |
853 | |
854 | +static bool xenvif_handle_tx_interrupt(struct xenvif_queue *queue) |
855 | +{ |
856 | + bool rc; |
857 | + |
858 | + rc = RING_HAS_UNCONSUMED_REQUESTS(&queue->tx); |
859 | + if (rc) |
860 | + napi_schedule(&queue->napi); |
861 | + return rc; |
862 | +} |
863 | + |
864 | static irqreturn_t xenvif_tx_interrupt(int irq, void *dev_id) |
865 | { |
866 | struct xenvif_queue *queue = dev_id; |
867 | + int old; |
868 | |
869 | - if (RING_HAS_UNCONSUMED_REQUESTS(&queue->tx)) |
870 | - napi_schedule(&queue->napi); |
871 | + old = atomic_fetch_or(NETBK_TX_EOI, &queue->eoi_pending); |
872 | + WARN(old & NETBK_TX_EOI, "Interrupt while EOI pending\n"); |
873 | + |
874 | + if (!xenvif_handle_tx_interrupt(queue)) { |
875 | + atomic_andnot(NETBK_TX_EOI, &queue->eoi_pending); |
876 | + xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); |
877 | + } |
878 | |
879 | return IRQ_HANDLED; |
880 | } |
881 | @@ -115,19 +131,46 @@ static int xenvif_poll(struct napi_struct *napi, int budget) |
882 | return work_done; |
883 | } |
884 | |
885 | +static bool xenvif_handle_rx_interrupt(struct xenvif_queue *queue) |
886 | +{ |
887 | + bool rc; |
888 | + |
889 | + rc = xenvif_have_rx_work(queue, false); |
890 | + if (rc) |
891 | + xenvif_kick_thread(queue); |
892 | + return rc; |
893 | +} |
894 | + |
895 | static irqreturn_t xenvif_rx_interrupt(int irq, void *dev_id) |
896 | { |
897 | struct xenvif_queue *queue = dev_id; |
898 | + int old; |
899 | |
900 | - xenvif_kick_thread(queue); |
901 | + old = atomic_fetch_or(NETBK_RX_EOI, &queue->eoi_pending); |
902 | + WARN(old & NETBK_RX_EOI, "Interrupt while EOI pending\n"); |
903 | + |
904 | + if (!xenvif_handle_rx_interrupt(queue)) { |
905 | + atomic_andnot(NETBK_RX_EOI, &queue->eoi_pending); |
906 | + xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); |
907 | + } |
908 | |
909 | return IRQ_HANDLED; |
910 | } |
911 | |
912 | irqreturn_t xenvif_interrupt(int irq, void *dev_id) |
913 | { |
914 | - xenvif_tx_interrupt(irq, dev_id); |
915 | - xenvif_rx_interrupt(irq, dev_id); |
916 | + struct xenvif_queue *queue = dev_id; |
917 | + int old; |
918 | + |
919 | + old = atomic_fetch_or(NETBK_COMMON_EOI, &queue->eoi_pending); |
920 | + WARN(old, "Interrupt while EOI pending\n"); |
921 | + |
922 | + /* Use bitwise or as we need to call both functions. */ |
923 | + if ((!xenvif_handle_tx_interrupt(queue) | |
924 | + !xenvif_handle_rx_interrupt(queue))) { |
925 | + atomic_andnot(NETBK_COMMON_EOI, &queue->eoi_pending); |
926 | + xen_irq_lateeoi(irq, XEN_EOI_FLAG_SPURIOUS); |
927 | + } |
928 | |
929 | return IRQ_HANDLED; |
930 | } |
931 | @@ -583,7 +626,7 @@ int xenvif_connect_ctrl(struct xenvif *vif, grant_ref_t ring_ref, |
932 | shared = (struct xen_netif_ctrl_sring *)addr; |
933 | BACK_RING_INIT(&vif->ctrl, shared, XEN_PAGE_SIZE); |
934 | |
935 | - err = bind_interdomain_evtchn_to_irq(vif->domid, evtchn); |
936 | + err = bind_interdomain_evtchn_to_irq_lateeoi(vif->domid, evtchn); |
937 | if (err < 0) |
938 | goto err_unmap; |
939 | |
940 | @@ -641,7 +684,7 @@ int xenvif_connect_data(struct xenvif_queue *queue, |
941 | |
942 | if (tx_evtchn == rx_evtchn) { |
943 | /* feature-split-event-channels == 0 */ |
944 | - err = bind_interdomain_evtchn_to_irqhandler( |
945 | + err = bind_interdomain_evtchn_to_irqhandler_lateeoi( |
946 | queue->vif->domid, tx_evtchn, xenvif_interrupt, 0, |
947 | queue->name, queue); |
948 | if (err < 0) |
949 | @@ -652,7 +695,7 @@ int xenvif_connect_data(struct xenvif_queue *queue, |
950 | /* feature-split-event-channels == 1 */ |
951 | snprintf(queue->tx_irq_name, sizeof(queue->tx_irq_name), |
952 | "%s-tx", queue->name); |
953 | - err = bind_interdomain_evtchn_to_irqhandler( |
954 | + err = bind_interdomain_evtchn_to_irqhandler_lateeoi( |
955 | queue->vif->domid, tx_evtchn, xenvif_tx_interrupt, 0, |
956 | queue->tx_irq_name, queue); |
957 | if (err < 0) |
958 | @@ -662,7 +705,7 @@ int xenvif_connect_data(struct xenvif_queue *queue, |
959 | |
960 | snprintf(queue->rx_irq_name, sizeof(queue->rx_irq_name), |
961 | "%s-rx", queue->name); |
962 | - err = bind_interdomain_evtchn_to_irqhandler( |
963 | + err = bind_interdomain_evtchn_to_irqhandler_lateeoi( |
964 | queue->vif->domid, rx_evtchn, xenvif_rx_interrupt, 0, |
965 | queue->rx_irq_name, queue); |
966 | if (err < 0) |
967 | diff --git a/drivers/net/xen-netback/netback.c b/drivers/net/xen-netback/netback.c |
968 | index a469fbe1abaf3..fd2ac6cd0c691 100644 |
969 | --- a/drivers/net/xen-netback/netback.c |
970 | +++ b/drivers/net/xen-netback/netback.c |
971 | @@ -162,6 +162,10 @@ void xenvif_napi_schedule_or_enable_events(struct xenvif_queue *queue) |
972 | |
973 | if (more_to_do) |
974 | napi_schedule(&queue->napi); |
975 | + else if (atomic_fetch_andnot(NETBK_TX_EOI | NETBK_COMMON_EOI, |
976 | + &queue->eoi_pending) & |
977 | + (NETBK_TX_EOI | NETBK_COMMON_EOI)) |
978 | + xen_irq_lateeoi(queue->tx_irq, 0); |
979 | } |
980 | |
981 | static void tx_add_credit(struct xenvif_queue *queue) |
982 | @@ -1615,9 +1619,14 @@ static bool xenvif_ctrl_work_todo(struct xenvif *vif) |
983 | irqreturn_t xenvif_ctrl_irq_fn(int irq, void *data) |
984 | { |
985 | struct xenvif *vif = data; |
986 | + unsigned int eoi_flag = XEN_EOI_FLAG_SPURIOUS; |
987 | |
988 | - while (xenvif_ctrl_work_todo(vif)) |
989 | + while (xenvif_ctrl_work_todo(vif)) { |
990 | xenvif_ctrl_action(vif); |
991 | + eoi_flag = 0; |
992 | + } |
993 | + |
994 | + xen_irq_lateeoi(irq, eoi_flag); |
995 | |
996 | return IRQ_HANDLED; |
997 | } |
998 | diff --git a/drivers/net/xen-netback/rx.c b/drivers/net/xen-netback/rx.c |
999 | index b1cf7c6f407a9..f152246c7dfb7 100644 |
1000 | --- a/drivers/net/xen-netback/rx.c |
1001 | +++ b/drivers/net/xen-netback/rx.c |
1002 | @@ -490,13 +490,13 @@ static bool xenvif_rx_queue_ready(struct xenvif_queue *queue) |
1003 | return queue->stalled && prod - cons >= 1; |
1004 | } |
1005 | |
1006 | -static bool xenvif_have_rx_work(struct xenvif_queue *queue) |
1007 | +bool xenvif_have_rx_work(struct xenvif_queue *queue, bool test_kthread) |
1008 | { |
1009 | return xenvif_rx_ring_slots_available(queue) || |
1010 | (queue->vif->stall_timeout && |
1011 | (xenvif_rx_queue_stalled(queue) || |
1012 | xenvif_rx_queue_ready(queue))) || |
1013 | - kthread_should_stop() || |
1014 | + (test_kthread && kthread_should_stop()) || |
1015 | queue->vif->disabled; |
1016 | } |
1017 | |
1018 | @@ -527,15 +527,20 @@ static void xenvif_wait_for_rx_work(struct xenvif_queue *queue) |
1019 | { |
1020 | DEFINE_WAIT(wait); |
1021 | |
1022 | - if (xenvif_have_rx_work(queue)) |
1023 | + if (xenvif_have_rx_work(queue, true)) |
1024 | return; |
1025 | |
1026 | for (;;) { |
1027 | long ret; |
1028 | |
1029 | prepare_to_wait(&queue->wq, &wait, TASK_INTERRUPTIBLE); |
1030 | - if (xenvif_have_rx_work(queue)) |
1031 | + if (xenvif_have_rx_work(queue, true)) |
1032 | break; |
1033 | + if (atomic_fetch_andnot(NETBK_RX_EOI | NETBK_COMMON_EOI, |
1034 | + &queue->eoi_pending) & |
1035 | + (NETBK_RX_EOI | NETBK_COMMON_EOI)) |
1036 | + xen_irq_lateeoi(queue->rx_irq, 0); |
1037 | + |
1038 | ret = schedule_timeout(xenvif_rx_queue_timeout(queue)); |
1039 | if (!ret) |
1040 | break; |
1041 | diff --git a/drivers/of/address.c b/drivers/of/address.c |
1042 | index 37619bb2c97ad..d188eacbd3b80 100644 |
1043 | --- a/drivers/of/address.c |
1044 | +++ b/drivers/of/address.c |
1045 | @@ -901,11 +901,13 @@ EXPORT_SYMBOL_GPL(of_dma_get_range); |
1046 | */ |
1047 | bool of_dma_is_coherent(struct device_node *np) |
1048 | { |
1049 | - struct device_node *node = of_node_get(np); |
1050 | + struct device_node *node; |
1051 | |
1052 | if (IS_ENABLED(CONFIG_OF_DMA_DEFAULT_COHERENT)) |
1053 | return true; |
1054 | |
1055 | + node = of_node_get(np); |
1056 | + |
1057 | while (node) { |
1058 | if (of_property_read_bool(node, "dma-coherent")) { |
1059 | of_node_put(node); |
1060 | diff --git a/drivers/pinctrl/aspeed/pinctrl-aspeed.c b/drivers/pinctrl/aspeed/pinctrl-aspeed.c |
1061 | index 49aeba9125319..23d2f0ba12db5 100644 |
1062 | --- a/drivers/pinctrl/aspeed/pinctrl-aspeed.c |
1063 | +++ b/drivers/pinctrl/aspeed/pinctrl-aspeed.c |
1064 | @@ -387,13 +387,14 @@ int aspeed_pinmux_set_mux(struct pinctrl_dev *pctldev, unsigned int function, |
1065 | static bool aspeed_expr_is_gpio(const struct aspeed_sig_expr *expr) |
1066 | { |
1067 | /* |
1068 | - * The signal type is GPIO if the signal name has "GPIO" as a prefix. |
1069 | + * The signal type is GPIO if the signal name has "GPI" as a prefix. |
1070 | * strncmp (rather than strcmp) is used to implement the prefix |
1071 | * requirement. |
1072 | * |
1073 | - * expr->signal might look like "GPIOT3" in the GPIO case. |
1074 | + * expr->signal might look like "GPIOB1" in the GPIO case. |
1075 | + * expr->signal might look like "GPIT0" in the GPI case. |
1076 | */ |
1077 | - return strncmp(expr->signal, "GPIO", 4) == 0; |
1078 | + return strncmp(expr->signal, "GPI", 3) == 0; |
1079 | } |
1080 | |
1081 | static bool aspeed_gpio_in_exprs(const struct aspeed_sig_expr **exprs) |
1082 | diff --git a/drivers/pinctrl/devicetree.c b/drivers/pinctrl/devicetree.c |
1083 | index 54dad89fc9bfe..d32aedfc6dd03 100644 |
1084 | --- a/drivers/pinctrl/devicetree.c |
1085 | +++ b/drivers/pinctrl/devicetree.c |
1086 | @@ -40,6 +40,13 @@ struct pinctrl_dt_map { |
1087 | static void dt_free_map(struct pinctrl_dev *pctldev, |
1088 | struct pinctrl_map *map, unsigned num_maps) |
1089 | { |
1090 | + int i; |
1091 | + |
1092 | + for (i = 0; i < num_maps; ++i) { |
1093 | + kfree_const(map[i].dev_name); |
1094 | + map[i].dev_name = NULL; |
1095 | + } |
1096 | + |
1097 | if (pctldev) { |
1098 | const struct pinctrl_ops *ops = pctldev->desc->pctlops; |
1099 | ops->dt_free_map(pctldev, map, num_maps); |
1100 | @@ -73,7 +80,13 @@ static int dt_remember_or_free_map(struct pinctrl *p, const char *statename, |
1101 | |
1102 | /* Initialize common mapping table entry fields */ |
1103 | for (i = 0; i < num_maps; i++) { |
1104 | - map[i].dev_name = dev_name(p->dev); |
1105 | + const char *devname; |
1106 | + |
1107 | + devname = kstrdup_const(dev_name(p->dev), GFP_KERNEL); |
1108 | + if (!devname) |
1109 | + goto err_free_map; |
1110 | + |
1111 | + map[i].dev_name = devname; |
1112 | map[i].name = statename; |
1113 | if (pctldev) |
1114 | map[i].ctrl_dev_name = dev_name(pctldev->dev); |
1115 | @@ -81,11 +94,8 @@ static int dt_remember_or_free_map(struct pinctrl *p, const char *statename, |
1116 | |
1117 | /* Remember the converted mapping table entries */ |
1118 | dt_map = kzalloc(sizeof(*dt_map), GFP_KERNEL); |
1119 | - if (!dt_map) { |
1120 | - dev_err(p->dev, "failed to alloc struct pinctrl_dt_map\n"); |
1121 | - dt_free_map(pctldev, map, num_maps); |
1122 | - return -ENOMEM; |
1123 | - } |
1124 | + if (!dt_map) |
1125 | + goto err_free_map; |
1126 | |
1127 | dt_map->pctldev = pctldev; |
1128 | dt_map->map = map; |
1129 | @@ -93,6 +103,10 @@ static int dt_remember_or_free_map(struct pinctrl *p, const char *statename, |
1130 | list_add_tail(&dt_map->node, &p->dt_maps); |
1131 | |
1132 | return pinctrl_register_map(map, num_maps, false); |
1133 | + |
1134 | +err_free_map: |
1135 | + dt_free_map(pctldev, map, num_maps); |
1136 | + return -ENOMEM; |
1137 | } |
1138 | |
1139 | struct pinctrl_dev *of_pinctrl_get(struct device_node *np) |
1140 | diff --git a/drivers/pinctrl/pinctrl-amd.c b/drivers/pinctrl/pinctrl-amd.c |
1141 | index a5b7bd3c9bac3..82fac6261efa1 100644 |
1142 | --- a/drivers/pinctrl/pinctrl-amd.c |
1143 | +++ b/drivers/pinctrl/pinctrl-amd.c |
1144 | @@ -140,7 +140,7 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset, |
1145 | pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF); |
1146 | pin_reg &= ~BIT(DB_TMR_LARGE_OFF); |
1147 | } else if (debounce < 250000) { |
1148 | - time = debounce / 15600; |
1149 | + time = debounce / 15625; |
1150 | pin_reg |= time & DB_TMR_OUT_MASK; |
1151 | pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF); |
1152 | pin_reg |= BIT(DB_TMR_LARGE_OFF); |
1153 | @@ -150,14 +150,14 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset, |
1154 | pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF); |
1155 | pin_reg |= BIT(DB_TMR_LARGE_OFF); |
1156 | } else { |
1157 | - pin_reg &= ~DB_CNTRl_MASK; |
1158 | + pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF); |
1159 | ret = -EINVAL; |
1160 | } |
1161 | } else { |
1162 | pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF); |
1163 | pin_reg &= ~BIT(DB_TMR_LARGE_OFF); |
1164 | pin_reg &= ~DB_TMR_OUT_MASK; |
1165 | - pin_reg &= ~DB_CNTRl_MASK; |
1166 | + pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF); |
1167 | } |
1168 | writel(pin_reg, gpio_dev->base + offset * 4); |
1169 | spin_unlock_irqrestore(&gpio_dev->lock, flags); |
1170 | diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c |
1171 | index 0f730e4bf6bcb..0caf751d85ded 100644 |
1172 | --- a/drivers/regulator/core.c |
1173 | +++ b/drivers/regulator/core.c |
1174 | @@ -3185,6 +3185,8 @@ static int _regulator_get_voltage(struct regulator_dev *rdev) |
1175 | ret = rdev->desc->fixed_uV; |
1176 | } else if (rdev->supply) { |
1177 | ret = _regulator_get_voltage(rdev->supply->rdev); |
1178 | + } else if (rdev->supply_name) { |
1179 | + return -EPROBE_DEFER; |
1180 | } else { |
1181 | return -EINVAL; |
1182 | } |
1183 | diff --git a/drivers/scsi/device_handler/scsi_dh_alua.c b/drivers/scsi/device_handler/scsi_dh_alua.c |
1184 | index 60c288526355a..2bc3dc6244a5e 100644 |
1185 | --- a/drivers/scsi/device_handler/scsi_dh_alua.c |
1186 | +++ b/drivers/scsi/device_handler/scsi_dh_alua.c |
1187 | @@ -657,8 +657,8 @@ static int alua_rtpg(struct scsi_device *sdev, struct alua_port_group *pg) |
1188 | rcu_read_lock(); |
1189 | list_for_each_entry_rcu(h, |
1190 | &tmp_pg->dh_list, node) { |
1191 | - /* h->sdev should always be valid */ |
1192 | - BUG_ON(!h->sdev); |
1193 | + if (!h->sdev) |
1194 | + continue; |
1195 | h->sdev->access_state = desc[0]; |
1196 | } |
1197 | rcu_read_unlock(); |
1198 | @@ -704,7 +704,8 @@ static int alua_rtpg(struct scsi_device *sdev, struct alua_port_group *pg) |
1199 | pg->expiry = 0; |
1200 | rcu_read_lock(); |
1201 | list_for_each_entry_rcu(h, &pg->dh_list, node) { |
1202 | - BUG_ON(!h->sdev); |
1203 | + if (!h->sdev) |
1204 | + continue; |
1205 | h->sdev->access_state = |
1206 | (pg->state & SCSI_ACCESS_STATE_MASK); |
1207 | if (pg->pref) |
1208 | @@ -1149,7 +1150,6 @@ static void alua_bus_detach(struct scsi_device *sdev) |
1209 | spin_lock(&h->pg_lock); |
1210 | pg = h->pg; |
1211 | rcu_assign_pointer(h->pg, NULL); |
1212 | - h->sdev = NULL; |
1213 | spin_unlock(&h->pg_lock); |
1214 | if (pg) { |
1215 | spin_lock_irq(&pg->lock); |
1216 | @@ -1158,6 +1158,7 @@ static void alua_bus_detach(struct scsi_device *sdev) |
1217 | kref_put(&pg->kref, release_port_group); |
1218 | } |
1219 | sdev->handler_data = NULL; |
1220 | + synchronize_rcu(); |
1221 | kfree(h); |
1222 | } |
1223 | |
1224 | diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c |
1225 | index b82df8cdf9626..7f1d6d52d48bd 100644 |
1226 | --- a/drivers/scsi/hpsa.c |
1227 | +++ b/drivers/scsi/hpsa.c |
1228 | @@ -8937,7 +8937,7 @@ reinit_after_soft_reset: |
1229 | /* hook into SCSI subsystem */ |
1230 | rc = hpsa_scsi_add_host(h); |
1231 | if (rc) |
1232 | - goto clean7; /* perf, sg, cmd, irq, shost, pci, lu, aer/h */ |
1233 | + goto clean8; /* lastlogicals, perf, sg, cmd, irq, shost, pci, lu, aer/h */ |
1234 | |
1235 | /* Monitor the controller for firmware lockups */ |
1236 | h->heartbeat_sample_interval = HEARTBEAT_SAMPLE_INTERVAL; |
1237 | @@ -8949,6 +8949,8 @@ reinit_after_soft_reset: |
1238 | h->heartbeat_sample_interval); |
1239 | return 0; |
1240 | |
1241 | +clean8: /* lastlogicals, perf, sg, cmd, irq, shost, pci, lu, aer/h */ |
1242 | + kfree(h->lastlogicals); |
1243 | clean7: /* perf, sg, cmd, irq, shost, pci, lu, aer/h */ |
1244 | hpsa_free_performant_mode(h); |
1245 | h->access.set_intr_mask(h, HPSA_INTR_OFF); |
1246 | diff --git a/drivers/usb/class/cdc-acm.c b/drivers/usb/class/cdc-acm.c |
1247 | index 726291c5562da..90f4a5c8012ca 100644 |
1248 | --- a/drivers/usb/class/cdc-acm.c |
1249 | +++ b/drivers/usb/class/cdc-acm.c |
1250 | @@ -1648,6 +1648,15 @@ static const struct usb_device_id acm_ids[] = { |
1251 | { USB_DEVICE(0x0870, 0x0001), /* Metricom GS Modem */ |
1252 | .driver_info = NO_UNION_NORMAL, /* has no union descriptor */ |
1253 | }, |
1254 | + { USB_DEVICE(0x045b, 0x023c), /* Renesas USB Download mode */ |
1255 | + .driver_info = DISABLE_ECHO, /* Don't echo banner */ |
1256 | + }, |
1257 | + { USB_DEVICE(0x045b, 0x0248), /* Renesas USB Download mode */ |
1258 | + .driver_info = DISABLE_ECHO, /* Don't echo banner */ |
1259 | + }, |
1260 | + { USB_DEVICE(0x045b, 0x024D), /* Renesas USB Download mode */ |
1261 | + .driver_info = DISABLE_ECHO, /* Don't echo banner */ |
1262 | + }, |
1263 | { USB_DEVICE(0x0e8d, 0x0003), /* FIREFLY, MediaTek Inc; andrey.arapov@gmail.com */ |
1264 | .driver_info = NO_UNION_NORMAL, /* has no union descriptor */ |
1265 | }, |
1266 | diff --git a/drivers/usb/gadget/udc/goku_udc.c b/drivers/usb/gadget/udc/goku_udc.c |
1267 | index 5107987bd3538..d363224dce6f5 100644 |
1268 | --- a/drivers/usb/gadget/udc/goku_udc.c |
1269 | +++ b/drivers/usb/gadget/udc/goku_udc.c |
1270 | @@ -1772,6 +1772,7 @@ static int goku_probe(struct pci_dev *pdev, const struct pci_device_id *id) |
1271 | goto err; |
1272 | } |
1273 | |
1274 | + pci_set_drvdata(pdev, dev); |
1275 | spin_lock_init(&dev->lock); |
1276 | dev->pdev = pdev; |
1277 | dev->gadget.ops = &goku_ops; |
1278 | @@ -1805,7 +1806,6 @@ static int goku_probe(struct pci_dev *pdev, const struct pci_device_id *id) |
1279 | } |
1280 | dev->regs = (struct goku_udc_regs __iomem *) base; |
1281 | |
1282 | - pci_set_drvdata(pdev, dev); |
1283 | INFO(dev, "%s\n", driver_desc); |
1284 | INFO(dev, "version: " DRIVER_VERSION " %s\n", dmastr()); |
1285 | INFO(dev, "irq %d, pci mem %p\n", pdev->irq, base); |
1286 | diff --git a/drivers/xen/events/events_2l.c b/drivers/xen/events/events_2l.c |
1287 | index bdff01095f549..c31c08a708be5 100644 |
1288 | --- a/drivers/xen/events/events_2l.c |
1289 | +++ b/drivers/xen/events/events_2l.c |
1290 | @@ -90,6 +90,8 @@ static void evtchn_2l_unmask(unsigned port) |
1291 | |
1292 | BUG_ON(!irqs_disabled()); |
1293 | |
1294 | + smp_wmb(); /* All writes before unmask must be visible. */ |
1295 | + |
1296 | if (unlikely((cpu != cpu_from_evtchn(port)))) |
1297 | do_hypercall = 1; |
1298 | else { |
1299 | @@ -158,7 +160,7 @@ static inline xen_ulong_t active_evtchns(unsigned int cpu, |
1300 | * a bitset of words which contain pending event bits. The second |
1301 | * level is a bitset of pending events themselves. |
1302 | */ |
1303 | -static void evtchn_2l_handle_events(unsigned cpu) |
1304 | +static void evtchn_2l_handle_events(unsigned cpu, struct evtchn_loop_ctrl *ctrl) |
1305 | { |
1306 | int irq; |
1307 | xen_ulong_t pending_words; |
1308 | @@ -239,10 +241,7 @@ static void evtchn_2l_handle_events(unsigned cpu) |
1309 | |
1310 | /* Process port. */ |
1311 | port = (word_idx * BITS_PER_EVTCHN_WORD) + bit_idx; |
1312 | - irq = get_evtchn_to_irq(port); |
1313 | - |
1314 | - if (irq != -1) |
1315 | - generic_handle_irq(irq); |
1316 | + handle_irq_for_port(port, ctrl); |
1317 | |
1318 | bit_idx = (bit_idx + 1) % BITS_PER_EVTCHN_WORD; |
1319 | |
1320 | diff --git a/drivers/xen/events/events_base.c b/drivers/xen/events/events_base.c |
1321 | index 83bb9fdbadc6a..5308bc1e0189d 100644 |
1322 | --- a/drivers/xen/events/events_base.c |
1323 | +++ b/drivers/xen/events/events_base.c |
1324 | @@ -32,6 +32,10 @@ |
1325 | #include <linux/slab.h> |
1326 | #include <linux/irqnr.h> |
1327 | #include <linux/pci.h> |
1328 | +#include <linux/spinlock.h> |
1329 | +#include <linux/cpuhotplug.h> |
1330 | +#include <linux/atomic.h> |
1331 | +#include <linux/ktime.h> |
1332 | |
1333 | #ifdef CONFIG_X86 |
1334 | #include <asm/desc.h> |
1335 | @@ -62,6 +66,15 @@ |
1336 | |
1337 | #include "events_internal.h" |
1338 | |
1339 | +#undef MODULE_PARAM_PREFIX |
1340 | +#define MODULE_PARAM_PREFIX "xen." |
1341 | + |
1342 | +static uint __read_mostly event_loop_timeout = 2; |
1343 | +module_param(event_loop_timeout, uint, 0644); |
1344 | + |
1345 | +static uint __read_mostly event_eoi_delay = 10; |
1346 | +module_param(event_eoi_delay, uint, 0644); |
1347 | + |
1348 | const struct evtchn_ops *evtchn_ops; |
1349 | |
1350 | /* |
1351 | @@ -70,6 +83,24 @@ const struct evtchn_ops *evtchn_ops; |
1352 | */ |
1353 | static DEFINE_MUTEX(irq_mapping_update_lock); |
1354 | |
1355 | +/* |
1356 | + * Lock protecting event handling loop against removing event channels. |
1357 | + * Adding of event channels is no issue as the associated IRQ becomes active |
1358 | + * only after everything is setup (before request_[threaded_]irq() the handler |
1359 | + * can't be entered for an event, as the event channel will be unmasked only |
1360 | + * then). |
1361 | + */ |
1362 | +static DEFINE_RWLOCK(evtchn_rwlock); |
1363 | + |
1364 | +/* |
1365 | + * Lock hierarchy: |
1366 | + * |
1367 | + * irq_mapping_update_lock |
1368 | + * evtchn_rwlock |
1369 | + * IRQ-desc lock |
1370 | + * percpu eoi_list_lock |
1371 | + */ |
1372 | + |
1373 | static LIST_HEAD(xen_irq_list_head); |
1374 | |
1375 | /* IRQ <-> VIRQ mapping. */ |
1376 | @@ -94,17 +125,20 @@ static bool (*pirq_needs_eoi)(unsigned irq); |
1377 | static struct irq_info *legacy_info_ptrs[NR_IRQS_LEGACY]; |
1378 | |
1379 | static struct irq_chip xen_dynamic_chip; |
1380 | +static struct irq_chip xen_lateeoi_chip; |
1381 | static struct irq_chip xen_percpu_chip; |
1382 | static struct irq_chip xen_pirq_chip; |
1383 | static void enable_dynirq(struct irq_data *data); |
1384 | static void disable_dynirq(struct irq_data *data); |
1385 | |
1386 | +static DEFINE_PER_CPU(unsigned int, irq_epoch); |
1387 | + |
1388 | static void clear_evtchn_to_irq_row(unsigned row) |
1389 | { |
1390 | unsigned col; |
1391 | |
1392 | for (col = 0; col < EVTCHN_PER_ROW; col++) |
1393 | - evtchn_to_irq[row][col] = -1; |
1394 | + WRITE_ONCE(evtchn_to_irq[row][col], -1); |
1395 | } |
1396 | |
1397 | static void clear_evtchn_to_irq_all(void) |
1398 | @@ -141,7 +175,7 @@ static int set_evtchn_to_irq(unsigned evtchn, unsigned irq) |
1399 | clear_evtchn_to_irq_row(row); |
1400 | } |
1401 | |
1402 | - evtchn_to_irq[row][col] = irq; |
1403 | + WRITE_ONCE(evtchn_to_irq[row][col], irq); |
1404 | return 0; |
1405 | } |
1406 | |
1407 | @@ -151,7 +185,7 @@ int get_evtchn_to_irq(unsigned evtchn) |
1408 | return -1; |
1409 | if (evtchn_to_irq[EVTCHN_ROW(evtchn)] == NULL) |
1410 | return -1; |
1411 | - return evtchn_to_irq[EVTCHN_ROW(evtchn)][EVTCHN_COL(evtchn)]; |
1412 | + return READ_ONCE(evtchn_to_irq[EVTCHN_ROW(evtchn)][EVTCHN_COL(evtchn)]); |
1413 | } |
1414 | |
1415 | /* Get info for IRQ */ |
1416 | @@ -260,10 +294,14 @@ static void xen_irq_info_cleanup(struct irq_info *info) |
1417 | */ |
1418 | unsigned int evtchn_from_irq(unsigned irq) |
1419 | { |
1420 | - if (unlikely(WARN(irq >= nr_irqs, "Invalid irq %d!\n", irq))) |
1421 | + const struct irq_info *info = NULL; |
1422 | + |
1423 | + if (likely(irq < nr_irqs)) |
1424 | + info = info_for_irq(irq); |
1425 | + if (!info) |
1426 | return 0; |
1427 | |
1428 | - return info_for_irq(irq)->evtchn; |
1429 | + return info->evtchn; |
1430 | } |
1431 | |
1432 | unsigned irq_from_evtchn(unsigned int evtchn) |
1433 | @@ -382,9 +420,157 @@ void notify_remote_via_irq(int irq) |
1434 | } |
1435 | EXPORT_SYMBOL_GPL(notify_remote_via_irq); |
1436 | |
1437 | +struct lateeoi_work { |
1438 | + struct delayed_work delayed; |
1439 | + spinlock_t eoi_list_lock; |
1440 | + struct list_head eoi_list; |
1441 | +}; |
1442 | + |
1443 | +static DEFINE_PER_CPU(struct lateeoi_work, lateeoi); |
1444 | + |
1445 | +static void lateeoi_list_del(struct irq_info *info) |
1446 | +{ |
1447 | + struct lateeoi_work *eoi = &per_cpu(lateeoi, info->eoi_cpu); |
1448 | + unsigned long flags; |
1449 | + |
1450 | + spin_lock_irqsave(&eoi->eoi_list_lock, flags); |
1451 | + list_del_init(&info->eoi_list); |
1452 | + spin_unlock_irqrestore(&eoi->eoi_list_lock, flags); |
1453 | +} |
1454 | + |
1455 | +static void lateeoi_list_add(struct irq_info *info) |
1456 | +{ |
1457 | + struct lateeoi_work *eoi = &per_cpu(lateeoi, info->eoi_cpu); |
1458 | + struct irq_info *elem; |
1459 | + u64 now = get_jiffies_64(); |
1460 | + unsigned long delay; |
1461 | + unsigned long flags; |
1462 | + |
1463 | + if (now < info->eoi_time) |
1464 | + delay = info->eoi_time - now; |
1465 | + else |
1466 | + delay = 1; |
1467 | + |
1468 | + spin_lock_irqsave(&eoi->eoi_list_lock, flags); |
1469 | + |
1470 | + if (list_empty(&eoi->eoi_list)) { |
1471 | + list_add(&info->eoi_list, &eoi->eoi_list); |
1472 | + mod_delayed_work_on(info->eoi_cpu, system_wq, |
1473 | + &eoi->delayed, delay); |
1474 | + } else { |
1475 | + list_for_each_entry_reverse(elem, &eoi->eoi_list, eoi_list) { |
1476 | + if (elem->eoi_time <= info->eoi_time) |
1477 | + break; |
1478 | + } |
1479 | + list_add(&info->eoi_list, &elem->eoi_list); |
1480 | + } |
1481 | + |
1482 | + spin_unlock_irqrestore(&eoi->eoi_list_lock, flags); |
1483 | +} |
1484 | + |
1485 | +static void xen_irq_lateeoi_locked(struct irq_info *info, bool spurious) |
1486 | +{ |
1487 | + evtchn_port_t evtchn; |
1488 | + unsigned int cpu; |
1489 | + unsigned int delay = 0; |
1490 | + |
1491 | + evtchn = info->evtchn; |
1492 | + if (!VALID_EVTCHN(evtchn) || !list_empty(&info->eoi_list)) |
1493 | + return; |
1494 | + |
1495 | + if (spurious) { |
1496 | + if ((1 << info->spurious_cnt) < (HZ << 2)) |
1497 | + info->spurious_cnt++; |
1498 | + if (info->spurious_cnt > 1) { |
1499 | + delay = 1 << (info->spurious_cnt - 2); |
1500 | + if (delay > HZ) |
1501 | + delay = HZ; |
1502 | + if (!info->eoi_time) |
1503 | + info->eoi_cpu = smp_processor_id(); |
1504 | + info->eoi_time = get_jiffies_64() + delay; |
1505 | + } |
1506 | + } else { |
1507 | + info->spurious_cnt = 0; |
1508 | + } |
1509 | + |
1510 | + cpu = info->eoi_cpu; |
1511 | + if (info->eoi_time && |
1512 | + (info->irq_epoch == per_cpu(irq_epoch, cpu) || delay)) { |
1513 | + lateeoi_list_add(info); |
1514 | + return; |
1515 | + } |
1516 | + |
1517 | + info->eoi_time = 0; |
1518 | + unmask_evtchn(evtchn); |
1519 | +} |
1520 | + |
1521 | +static void xen_irq_lateeoi_worker(struct work_struct *work) |
1522 | +{ |
1523 | + struct lateeoi_work *eoi; |
1524 | + struct irq_info *info; |
1525 | + u64 now = get_jiffies_64(); |
1526 | + unsigned long flags; |
1527 | + |
1528 | + eoi = container_of(to_delayed_work(work), struct lateeoi_work, delayed); |
1529 | + |
1530 | + read_lock_irqsave(&evtchn_rwlock, flags); |
1531 | + |
1532 | + while (true) { |
1533 | + spin_lock(&eoi->eoi_list_lock); |
1534 | + |
1535 | + info = list_first_entry_or_null(&eoi->eoi_list, struct irq_info, |
1536 | + eoi_list); |
1537 | + |
1538 | + if (info == NULL || now < info->eoi_time) { |
1539 | + spin_unlock(&eoi->eoi_list_lock); |
1540 | + break; |
1541 | + } |
1542 | + |
1543 | + list_del_init(&info->eoi_list); |
1544 | + |
1545 | + spin_unlock(&eoi->eoi_list_lock); |
1546 | + |
1547 | + info->eoi_time = 0; |
1548 | + |
1549 | + xen_irq_lateeoi_locked(info, false); |
1550 | + } |
1551 | + |
1552 | + if (info) |
1553 | + mod_delayed_work_on(info->eoi_cpu, system_wq, |
1554 | + &eoi->delayed, info->eoi_time - now); |
1555 | + |
1556 | + read_unlock_irqrestore(&evtchn_rwlock, flags); |
1557 | +} |
1558 | + |
1559 | +static void xen_cpu_init_eoi(unsigned int cpu) |
1560 | +{ |
1561 | + struct lateeoi_work *eoi = &per_cpu(lateeoi, cpu); |
1562 | + |
1563 | + INIT_DELAYED_WORK(&eoi->delayed, xen_irq_lateeoi_worker); |
1564 | + spin_lock_init(&eoi->eoi_list_lock); |
1565 | + INIT_LIST_HEAD(&eoi->eoi_list); |
1566 | +} |
1567 | + |
1568 | +void xen_irq_lateeoi(unsigned int irq, unsigned int eoi_flags) |
1569 | +{ |
1570 | + struct irq_info *info; |
1571 | + unsigned long flags; |
1572 | + |
1573 | + read_lock_irqsave(&evtchn_rwlock, flags); |
1574 | + |
1575 | + info = info_for_irq(irq); |
1576 | + |
1577 | + if (info) |
1578 | + xen_irq_lateeoi_locked(info, eoi_flags & XEN_EOI_FLAG_SPURIOUS); |
1579 | + |
1580 | + read_unlock_irqrestore(&evtchn_rwlock, flags); |
1581 | +} |
1582 | +EXPORT_SYMBOL_GPL(xen_irq_lateeoi); |
1583 | + |
1584 | static void xen_irq_init(unsigned irq) |
1585 | { |
1586 | struct irq_info *info; |
1587 | + |
1588 | #ifdef CONFIG_SMP |
1589 | /* By default all event channels notify CPU#0. */ |
1590 | cpumask_copy(irq_get_affinity_mask(irq), cpumask_of(0)); |
1591 | @@ -399,6 +585,7 @@ static void xen_irq_init(unsigned irq) |
1592 | |
1593 | set_info_for_irq(irq, info); |
1594 | |
1595 | + INIT_LIST_HEAD(&info->eoi_list); |
1596 | list_add_tail(&info->list, &xen_irq_list_head); |
1597 | } |
1598 | |
1599 | @@ -447,16 +634,24 @@ static int __must_check xen_allocate_irq_gsi(unsigned gsi) |
1600 | static void xen_free_irq(unsigned irq) |
1601 | { |
1602 | struct irq_info *info = info_for_irq(irq); |
1603 | + unsigned long flags; |
1604 | |
1605 | if (WARN_ON(!info)) |
1606 | return; |
1607 | |
1608 | + write_lock_irqsave(&evtchn_rwlock, flags); |
1609 | + |
1610 | + if (!list_empty(&info->eoi_list)) |
1611 | + lateeoi_list_del(info); |
1612 | + |
1613 | list_del(&info->list); |
1614 | |
1615 | set_info_for_irq(irq, NULL); |
1616 | |
1617 | WARN_ON(info->refcnt > 0); |
1618 | |
1619 | + write_unlock_irqrestore(&evtchn_rwlock, flags); |
1620 | + |
1621 | kfree(info); |
1622 | |
1623 | /* Legacy IRQ descriptors are managed by the arch. */ |
1624 | @@ -848,7 +1043,7 @@ int xen_pirq_from_irq(unsigned irq) |
1625 | } |
1626 | EXPORT_SYMBOL_GPL(xen_pirq_from_irq); |
1627 | |
1628 | -int bind_evtchn_to_irq(unsigned int evtchn) |
1629 | +static int bind_evtchn_to_irq_chip(evtchn_port_t evtchn, struct irq_chip *chip) |
1630 | { |
1631 | int irq; |
1632 | int ret; |
1633 | @@ -865,7 +1060,7 @@ int bind_evtchn_to_irq(unsigned int evtchn) |
1634 | if (irq < 0) |
1635 | goto out; |
1636 | |
1637 | - irq_set_chip_and_handler_name(irq, &xen_dynamic_chip, |
1638 | + irq_set_chip_and_handler_name(irq, chip, |
1639 | handle_edge_irq, "event"); |
1640 | |
1641 | ret = xen_irq_info_evtchn_setup(irq, evtchn); |
1642 | @@ -886,8 +1081,19 @@ out: |
1643 | |
1644 | return irq; |
1645 | } |
1646 | + |
1647 | +int bind_evtchn_to_irq(evtchn_port_t evtchn) |
1648 | +{ |
1649 | + return bind_evtchn_to_irq_chip(evtchn, &xen_dynamic_chip); |
1650 | +} |
1651 | EXPORT_SYMBOL_GPL(bind_evtchn_to_irq); |
1652 | |
1653 | +int bind_evtchn_to_irq_lateeoi(evtchn_port_t evtchn) |
1654 | +{ |
1655 | + return bind_evtchn_to_irq_chip(evtchn, &xen_lateeoi_chip); |
1656 | +} |
1657 | +EXPORT_SYMBOL_GPL(bind_evtchn_to_irq_lateeoi); |
1658 | + |
1659 | static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu) |
1660 | { |
1661 | struct evtchn_bind_ipi bind_ipi; |
1662 | @@ -929,8 +1135,9 @@ static int bind_ipi_to_irq(unsigned int ipi, unsigned int cpu) |
1663 | return irq; |
1664 | } |
1665 | |
1666 | -int bind_interdomain_evtchn_to_irq(unsigned int remote_domain, |
1667 | - unsigned int remote_port) |
1668 | +static int bind_interdomain_evtchn_to_irq_chip(unsigned int remote_domain, |
1669 | + evtchn_port_t remote_port, |
1670 | + struct irq_chip *chip) |
1671 | { |
1672 | struct evtchn_bind_interdomain bind_interdomain; |
1673 | int err; |
1674 | @@ -941,10 +1148,26 @@ int bind_interdomain_evtchn_to_irq(unsigned int remote_domain, |
1675 | err = HYPERVISOR_event_channel_op(EVTCHNOP_bind_interdomain, |
1676 | &bind_interdomain); |
1677 | |
1678 | - return err ? : bind_evtchn_to_irq(bind_interdomain.local_port); |
1679 | + return err ? : bind_evtchn_to_irq_chip(bind_interdomain.local_port, |
1680 | + chip); |
1681 | +} |
1682 | + |
1683 | +int bind_interdomain_evtchn_to_irq(unsigned int remote_domain, |
1684 | + evtchn_port_t remote_port) |
1685 | +{ |
1686 | + return bind_interdomain_evtchn_to_irq_chip(remote_domain, remote_port, |
1687 | + &xen_dynamic_chip); |
1688 | } |
1689 | EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irq); |
1690 | |
1691 | +int bind_interdomain_evtchn_to_irq_lateeoi(unsigned int remote_domain, |
1692 | + evtchn_port_t remote_port) |
1693 | +{ |
1694 | + return bind_interdomain_evtchn_to_irq_chip(remote_domain, remote_port, |
1695 | + &xen_lateeoi_chip); |
1696 | +} |
1697 | +EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irq_lateeoi); |
1698 | + |
1699 | static int find_virq(unsigned int virq, unsigned int cpu) |
1700 | { |
1701 | struct evtchn_status status; |
1702 | @@ -1040,14 +1263,15 @@ static void unbind_from_irq(unsigned int irq) |
1703 | mutex_unlock(&irq_mapping_update_lock); |
1704 | } |
1705 | |
1706 | -int bind_evtchn_to_irqhandler(unsigned int evtchn, |
1707 | - irq_handler_t handler, |
1708 | - unsigned long irqflags, |
1709 | - const char *devname, void *dev_id) |
1710 | +static int bind_evtchn_to_irqhandler_chip(evtchn_port_t evtchn, |
1711 | + irq_handler_t handler, |
1712 | + unsigned long irqflags, |
1713 | + const char *devname, void *dev_id, |
1714 | + struct irq_chip *chip) |
1715 | { |
1716 | int irq, retval; |
1717 | |
1718 | - irq = bind_evtchn_to_irq(evtchn); |
1719 | + irq = bind_evtchn_to_irq_chip(evtchn, chip); |
1720 | if (irq < 0) |
1721 | return irq; |
1722 | retval = request_irq(irq, handler, irqflags, devname, dev_id); |
1723 | @@ -1058,18 +1282,38 @@ int bind_evtchn_to_irqhandler(unsigned int evtchn, |
1724 | |
1725 | return irq; |
1726 | } |
1727 | + |
1728 | +int bind_evtchn_to_irqhandler(evtchn_port_t evtchn, |
1729 | + irq_handler_t handler, |
1730 | + unsigned long irqflags, |
1731 | + const char *devname, void *dev_id) |
1732 | +{ |
1733 | + return bind_evtchn_to_irqhandler_chip(evtchn, handler, irqflags, |
1734 | + devname, dev_id, |
1735 | + &xen_dynamic_chip); |
1736 | +} |
1737 | EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler); |
1738 | |
1739 | -int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain, |
1740 | - unsigned int remote_port, |
1741 | - irq_handler_t handler, |
1742 | - unsigned long irqflags, |
1743 | - const char *devname, |
1744 | - void *dev_id) |
1745 | +int bind_evtchn_to_irqhandler_lateeoi(evtchn_port_t evtchn, |
1746 | + irq_handler_t handler, |
1747 | + unsigned long irqflags, |
1748 | + const char *devname, void *dev_id) |
1749 | +{ |
1750 | + return bind_evtchn_to_irqhandler_chip(evtchn, handler, irqflags, |
1751 | + devname, dev_id, |
1752 | + &xen_lateeoi_chip); |
1753 | +} |
1754 | +EXPORT_SYMBOL_GPL(bind_evtchn_to_irqhandler_lateeoi); |
1755 | + |
1756 | +static int bind_interdomain_evtchn_to_irqhandler_chip( |
1757 | + unsigned int remote_domain, evtchn_port_t remote_port, |
1758 | + irq_handler_t handler, unsigned long irqflags, |
1759 | + const char *devname, void *dev_id, struct irq_chip *chip) |
1760 | { |
1761 | int irq, retval; |
1762 | |
1763 | - irq = bind_interdomain_evtchn_to_irq(remote_domain, remote_port); |
1764 | + irq = bind_interdomain_evtchn_to_irq_chip(remote_domain, remote_port, |
1765 | + chip); |
1766 | if (irq < 0) |
1767 | return irq; |
1768 | |
1769 | @@ -1081,8 +1325,33 @@ int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain, |
1770 | |
1771 | return irq; |
1772 | } |
1773 | + |
1774 | +int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain, |
1775 | + evtchn_port_t remote_port, |
1776 | + irq_handler_t handler, |
1777 | + unsigned long irqflags, |
1778 | + const char *devname, |
1779 | + void *dev_id) |
1780 | +{ |
1781 | + return bind_interdomain_evtchn_to_irqhandler_chip(remote_domain, |
1782 | + remote_port, handler, irqflags, devname, |
1783 | + dev_id, &xen_dynamic_chip); |
1784 | +} |
1785 | EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irqhandler); |
1786 | |
1787 | +int bind_interdomain_evtchn_to_irqhandler_lateeoi(unsigned int remote_domain, |
1788 | + evtchn_port_t remote_port, |
1789 | + irq_handler_t handler, |
1790 | + unsigned long irqflags, |
1791 | + const char *devname, |
1792 | + void *dev_id) |
1793 | +{ |
1794 | + return bind_interdomain_evtchn_to_irqhandler_chip(remote_domain, |
1795 | + remote_port, handler, irqflags, devname, |
1796 | + dev_id, &xen_lateeoi_chip); |
1797 | +} |
1798 | +EXPORT_SYMBOL_GPL(bind_interdomain_evtchn_to_irqhandler_lateeoi); |
1799 | + |
1800 | int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu, |
1801 | irq_handler_t handler, |
1802 | unsigned long irqflags, const char *devname, void *dev_id) |
1803 | @@ -1195,7 +1464,7 @@ int evtchn_get(unsigned int evtchn) |
1804 | goto done; |
1805 | |
1806 | err = -EINVAL; |
1807 | - if (info->refcnt <= 0) |
1808 | + if (info->refcnt <= 0 || info->refcnt == SHRT_MAX) |
1809 | goto done; |
1810 | |
1811 | info->refcnt++; |
1812 | @@ -1234,6 +1503,54 @@ void xen_send_IPI_one(unsigned int cpu, enum ipi_vector vector) |
1813 | notify_remote_via_irq(irq); |
1814 | } |
1815 | |
1816 | +struct evtchn_loop_ctrl { |
1817 | + ktime_t timeout; |
1818 | + unsigned count; |
1819 | + bool defer_eoi; |
1820 | +}; |
1821 | + |
1822 | +void handle_irq_for_port(evtchn_port_t port, struct evtchn_loop_ctrl *ctrl) |
1823 | +{ |
1824 | + int irq; |
1825 | + struct irq_info *info; |
1826 | + |
1827 | + irq = get_evtchn_to_irq(port); |
1828 | + if (irq == -1) |
1829 | + return; |
1830 | + |
1831 | + /* |
1832 | + * Check for timeout every 256 events. |
1833 | + * We are setting the timeout value only after the first 256 |
1834 | + * events in order to not hurt the common case of few loop |
1835 | + * iterations. The 256 is basically an arbitrary value. |
1836 | + * |
1837 | + * In case we are hitting the timeout we need to defer all further |
1838 | + * EOIs in order to ensure to leave the event handling loop rather |
1839 | + * sooner than later. |
1840 | + */ |
1841 | + if (!ctrl->defer_eoi && !(++ctrl->count & 0xff)) { |
1842 | + ktime_t kt = ktime_get(); |
1843 | + |
1844 | + if (!ctrl->timeout.tv64) { |
1845 | + kt = ktime_add_ms(kt, |
1846 | + jiffies_to_msecs(event_loop_timeout)); |
1847 | + ctrl->timeout = kt; |
1848 | + } else if (kt.tv64 > ctrl->timeout.tv64) { |
1849 | + ctrl->defer_eoi = true; |
1850 | + } |
1851 | + } |
1852 | + |
1853 | + info = info_for_irq(irq); |
1854 | + |
1855 | + if (ctrl->defer_eoi) { |
1856 | + info->eoi_cpu = smp_processor_id(); |
1857 | + info->irq_epoch = __this_cpu_read(irq_epoch); |
1858 | + info->eoi_time = get_jiffies_64() + event_eoi_delay; |
1859 | + } |
1860 | + |
1861 | + generic_handle_irq(irq); |
1862 | +} |
1863 | + |
1864 | static DEFINE_PER_CPU(unsigned, xed_nesting_count); |
1865 | |
1866 | static void __xen_evtchn_do_upcall(void) |
1867 | @@ -1241,6 +1558,9 @@ static void __xen_evtchn_do_upcall(void) |
1868 | struct vcpu_info *vcpu_info = __this_cpu_read(xen_vcpu); |
1869 | int cpu = get_cpu(); |
1870 | unsigned count; |
1871 | + struct evtchn_loop_ctrl ctrl = { 0 }; |
1872 | + |
1873 | + read_lock(&evtchn_rwlock); |
1874 | |
1875 | do { |
1876 | vcpu_info->evtchn_upcall_pending = 0; |
1877 | @@ -1248,7 +1568,7 @@ static void __xen_evtchn_do_upcall(void) |
1878 | if (__this_cpu_inc_return(xed_nesting_count) - 1) |
1879 | goto out; |
1880 | |
1881 | - xen_evtchn_handle_events(cpu); |
1882 | + xen_evtchn_handle_events(cpu, &ctrl); |
1883 | |
1884 | BUG_ON(!irqs_disabled()); |
1885 | |
1886 | @@ -1257,6 +1577,14 @@ static void __xen_evtchn_do_upcall(void) |
1887 | } while (count != 1 || vcpu_info->evtchn_upcall_pending); |
1888 | |
1889 | out: |
1890 | + read_unlock(&evtchn_rwlock); |
1891 | + |
1892 | + /* |
1893 | + * Increment irq_epoch only now to defer EOIs only for |
1894 | + * xen_irq_lateeoi() invocations occurring from inside the loop |
1895 | + * above. |
1896 | + */ |
1897 | + __this_cpu_inc(irq_epoch); |
1898 | |
1899 | put_cpu(); |
1900 | } |
1901 | @@ -1613,6 +1941,21 @@ static struct irq_chip xen_dynamic_chip __read_mostly = { |
1902 | .irq_retrigger = retrigger_dynirq, |
1903 | }; |
1904 | |
1905 | +static struct irq_chip xen_lateeoi_chip __read_mostly = { |
1906 | + /* The chip name needs to contain "xen-dyn" for irqbalance to work. */ |
1907 | + .name = "xen-dyn-lateeoi", |
1908 | + |
1909 | + .irq_disable = disable_dynirq, |
1910 | + .irq_mask = disable_dynirq, |
1911 | + .irq_unmask = enable_dynirq, |
1912 | + |
1913 | + .irq_ack = mask_ack_dynirq, |
1914 | + .irq_mask_ack = mask_ack_dynirq, |
1915 | + |
1916 | + .irq_set_affinity = set_affinity_irq, |
1917 | + .irq_retrigger = retrigger_dynirq, |
1918 | +}; |
1919 | + |
1920 | static struct irq_chip xen_pirq_chip __read_mostly = { |
1921 | .name = "xen-pirq", |
1922 | |
1923 | @@ -1680,12 +2023,31 @@ void xen_callback_vector(void) |
1924 | void xen_callback_vector(void) {} |
1925 | #endif |
1926 | |
1927 | -#undef MODULE_PARAM_PREFIX |
1928 | -#define MODULE_PARAM_PREFIX "xen." |
1929 | - |
1930 | static bool fifo_events = true; |
1931 | module_param(fifo_events, bool, 0); |
1932 | |
1933 | +static int xen_evtchn_cpu_prepare(unsigned int cpu) |
1934 | +{ |
1935 | + int ret = 0; |
1936 | + |
1937 | + xen_cpu_init_eoi(cpu); |
1938 | + |
1939 | + if (evtchn_ops->percpu_init) |
1940 | + ret = evtchn_ops->percpu_init(cpu); |
1941 | + |
1942 | + return ret; |
1943 | +} |
1944 | + |
1945 | +static int xen_evtchn_cpu_dead(unsigned int cpu) |
1946 | +{ |
1947 | + int ret = 0; |
1948 | + |
1949 | + if (evtchn_ops->percpu_deinit) |
1950 | + ret = evtchn_ops->percpu_deinit(cpu); |
1951 | + |
1952 | + return ret; |
1953 | +} |
1954 | + |
1955 | void __init xen_init_IRQ(void) |
1956 | { |
1957 | int ret = -EINVAL; |
1958 | @@ -1695,6 +2057,12 @@ void __init xen_init_IRQ(void) |
1959 | if (ret < 0) |
1960 | xen_evtchn_2l_init(); |
1961 | |
1962 | + xen_cpu_init_eoi(smp_processor_id()); |
1963 | + |
1964 | + cpuhp_setup_state_nocalls(CPUHP_XEN_EVTCHN_PREPARE, |
1965 | + "CPUHP_XEN_EVTCHN_PREPARE", |
1966 | + xen_evtchn_cpu_prepare, xen_evtchn_cpu_dead); |
1967 | + |
1968 | evtchn_to_irq = kcalloc(EVTCHN_ROW(xen_evtchn_max_channels()), |
1969 | sizeof(*evtchn_to_irq), GFP_KERNEL); |
1970 | BUG_ON(!evtchn_to_irq); |
1971 | diff --git a/drivers/xen/events/events_fifo.c b/drivers/xen/events/events_fifo.c |
1972 | index 7ef27c6ed72fb..0a4fece5fd8d3 100644 |
1973 | --- a/drivers/xen/events/events_fifo.c |
1974 | +++ b/drivers/xen/events/events_fifo.c |
1975 | @@ -227,19 +227,25 @@ static bool evtchn_fifo_is_masked(unsigned port) |
1976 | return sync_test_bit(EVTCHN_FIFO_BIT(MASKED, word), BM(word)); |
1977 | } |
1978 | /* |
1979 | - * Clear MASKED, spinning if BUSY is set. |
1980 | + * Clear MASKED if not PENDING, spinning if BUSY is set. |
1981 | + * Return true if mask was cleared. |
1982 | */ |
1983 | -static void clear_masked(volatile event_word_t *word) |
1984 | +static bool clear_masked_cond(volatile event_word_t *word) |
1985 | { |
1986 | event_word_t new, old, w; |
1987 | |
1988 | w = *word; |
1989 | |
1990 | do { |
1991 | + if (w & (1 << EVTCHN_FIFO_PENDING)) |
1992 | + return false; |
1993 | + |
1994 | old = w & ~(1 << EVTCHN_FIFO_BUSY); |
1995 | new = old & ~(1 << EVTCHN_FIFO_MASKED); |
1996 | w = sync_cmpxchg(word, old, new); |
1997 | } while (w != old); |
1998 | + |
1999 | + return true; |
2000 | } |
2001 | |
2002 | static void evtchn_fifo_unmask(unsigned port) |
2003 | @@ -248,8 +254,7 @@ static void evtchn_fifo_unmask(unsigned port) |
2004 | |
2005 | BUG_ON(!irqs_disabled()); |
2006 | |
2007 | - clear_masked(word); |
2008 | - if (evtchn_fifo_is_pending(port)) { |
2009 | + if (!clear_masked_cond(word)) { |
2010 | struct evtchn_unmask unmask = { .port = port }; |
2011 | (void)HYPERVISOR_event_channel_op(EVTCHNOP_unmask, &unmask); |
2012 | } |
2013 | @@ -270,19 +275,9 @@ static uint32_t clear_linked(volatile event_word_t *word) |
2014 | return w & EVTCHN_FIFO_LINK_MASK; |
2015 | } |
2016 | |
2017 | -static void handle_irq_for_port(unsigned port) |
2018 | -{ |
2019 | - int irq; |
2020 | - |
2021 | - irq = get_evtchn_to_irq(port); |
2022 | - if (irq != -1) |
2023 | - generic_handle_irq(irq); |
2024 | -} |
2025 | - |
2026 | -static void consume_one_event(unsigned cpu, |
2027 | +static void consume_one_event(unsigned cpu, struct evtchn_loop_ctrl *ctrl, |
2028 | struct evtchn_fifo_control_block *control_block, |
2029 | - unsigned priority, unsigned long *ready, |
2030 | - bool drop) |
2031 | + unsigned priority, unsigned long *ready) |
2032 | { |
2033 | struct evtchn_fifo_queue *q = &per_cpu(cpu_queue, cpu); |
2034 | uint32_t head; |
2035 | @@ -315,16 +310,17 @@ static void consume_one_event(unsigned cpu, |
2036 | clear_bit(priority, ready); |
2037 | |
2038 | if (evtchn_fifo_is_pending(port) && !evtchn_fifo_is_masked(port)) { |
2039 | - if (unlikely(drop)) |
2040 | + if (unlikely(!ctrl)) |
2041 | pr_warn("Dropping pending event for port %u\n", port); |
2042 | else |
2043 | - handle_irq_for_port(port); |
2044 | + handle_irq_for_port(port, ctrl); |
2045 | } |
2046 | |
2047 | q->head[priority] = head; |
2048 | } |
2049 | |
2050 | -static void __evtchn_fifo_handle_events(unsigned cpu, bool drop) |
2051 | +static void __evtchn_fifo_handle_events(unsigned cpu, |
2052 | + struct evtchn_loop_ctrl *ctrl) |
2053 | { |
2054 | struct evtchn_fifo_control_block *control_block; |
2055 | unsigned long ready; |
2056 | @@ -336,14 +332,15 @@ static void __evtchn_fifo_handle_events(unsigned cpu, bool drop) |
2057 | |
2058 | while (ready) { |
2059 | q = find_first_bit(&ready, EVTCHN_FIFO_MAX_QUEUES); |
2060 | - consume_one_event(cpu, control_block, q, &ready, drop); |
2061 | + consume_one_event(cpu, ctrl, control_block, q, &ready); |
2062 | ready |= xchg(&control_block->ready, 0); |
2063 | } |
2064 | } |
2065 | |
2066 | -static void evtchn_fifo_handle_events(unsigned cpu) |
2067 | +static void evtchn_fifo_handle_events(unsigned cpu, |
2068 | + struct evtchn_loop_ctrl *ctrl) |
2069 | { |
2070 | - __evtchn_fifo_handle_events(cpu, false); |
2071 | + __evtchn_fifo_handle_events(cpu, ctrl); |
2072 | } |
2073 | |
2074 | static void evtchn_fifo_resume(void) |
2075 | @@ -381,21 +378,6 @@ static void evtchn_fifo_resume(void) |
2076 | event_array_pages = 0; |
2077 | } |
2078 | |
2079 | -static const struct evtchn_ops evtchn_ops_fifo = { |
2080 | - .max_channels = evtchn_fifo_max_channels, |
2081 | - .nr_channels = evtchn_fifo_nr_channels, |
2082 | - .setup = evtchn_fifo_setup, |
2083 | - .bind_to_cpu = evtchn_fifo_bind_to_cpu, |
2084 | - .clear_pending = evtchn_fifo_clear_pending, |
2085 | - .set_pending = evtchn_fifo_set_pending, |
2086 | - .is_pending = evtchn_fifo_is_pending, |
2087 | - .test_and_set_mask = evtchn_fifo_test_and_set_mask, |
2088 | - .mask = evtchn_fifo_mask, |
2089 | - .unmask = evtchn_fifo_unmask, |
2090 | - .handle_events = evtchn_fifo_handle_events, |
2091 | - .resume = evtchn_fifo_resume, |
2092 | -}; |
2093 | - |
2094 | static int evtchn_fifo_alloc_control_block(unsigned cpu) |
2095 | { |
2096 | void *control_block = NULL; |
2097 | @@ -418,19 +400,36 @@ static int evtchn_fifo_alloc_control_block(unsigned cpu) |
2098 | return ret; |
2099 | } |
2100 | |
2101 | -static int xen_evtchn_cpu_prepare(unsigned int cpu) |
2102 | +static int evtchn_fifo_percpu_init(unsigned int cpu) |
2103 | { |
2104 | if (!per_cpu(cpu_control_block, cpu)) |
2105 | return evtchn_fifo_alloc_control_block(cpu); |
2106 | return 0; |
2107 | } |
2108 | |
2109 | -static int xen_evtchn_cpu_dead(unsigned int cpu) |
2110 | +static int evtchn_fifo_percpu_deinit(unsigned int cpu) |
2111 | { |
2112 | - __evtchn_fifo_handle_events(cpu, true); |
2113 | + __evtchn_fifo_handle_events(cpu, NULL); |
2114 | return 0; |
2115 | } |
2116 | |
2117 | +static const struct evtchn_ops evtchn_ops_fifo = { |
2118 | + .max_channels = evtchn_fifo_max_channels, |
2119 | + .nr_channels = evtchn_fifo_nr_channels, |
2120 | + .setup = evtchn_fifo_setup, |
2121 | + .bind_to_cpu = evtchn_fifo_bind_to_cpu, |
2122 | + .clear_pending = evtchn_fifo_clear_pending, |
2123 | + .set_pending = evtchn_fifo_set_pending, |
2124 | + .is_pending = evtchn_fifo_is_pending, |
2125 | + .test_and_set_mask = evtchn_fifo_test_and_set_mask, |
2126 | + .mask = evtchn_fifo_mask, |
2127 | + .unmask = evtchn_fifo_unmask, |
2128 | + .handle_events = evtchn_fifo_handle_events, |
2129 | + .resume = evtchn_fifo_resume, |
2130 | + .percpu_init = evtchn_fifo_percpu_init, |
2131 | + .percpu_deinit = evtchn_fifo_percpu_deinit, |
2132 | +}; |
2133 | + |
2134 | int __init xen_evtchn_fifo_init(void) |
2135 | { |
2136 | int cpu = get_cpu(); |
2137 | @@ -444,9 +443,6 @@ int __init xen_evtchn_fifo_init(void) |
2138 | |
2139 | evtchn_ops = &evtchn_ops_fifo; |
2140 | |
2141 | - cpuhp_setup_state_nocalls(CPUHP_XEN_EVTCHN_PREPARE, |
2142 | - "CPUHP_XEN_EVTCHN_PREPARE", |
2143 | - xen_evtchn_cpu_prepare, xen_evtchn_cpu_dead); |
2144 | out: |
2145 | put_cpu(); |
2146 | return ret; |
2147 | diff --git a/drivers/xen/events/events_internal.h b/drivers/xen/events/events_internal.h |
2148 | index 50c2050a1e320..b9b4f59198930 100644 |
2149 | --- a/drivers/xen/events/events_internal.h |
2150 | +++ b/drivers/xen/events/events_internal.h |
2151 | @@ -32,11 +32,16 @@ enum xen_irq_type { |
2152 | */ |
2153 | struct irq_info { |
2154 | struct list_head list; |
2155 | - int refcnt; |
2156 | + struct list_head eoi_list; |
2157 | + short refcnt; |
2158 | + short spurious_cnt; |
2159 | enum xen_irq_type type; /* type */ |
2160 | unsigned irq; |
2161 | unsigned int evtchn; /* event channel */ |
2162 | unsigned short cpu; /* cpu bound */ |
2163 | + unsigned short eoi_cpu; /* EOI must happen on this cpu */ |
2164 | + unsigned int irq_epoch; /* If eoi_cpu valid: irq_epoch of event */ |
2165 | + u64 eoi_time; /* Time in jiffies when to EOI. */ |
2166 | |
2167 | union { |
2168 | unsigned short virq; |
2169 | @@ -55,6 +60,8 @@ struct irq_info { |
2170 | #define PIRQ_SHAREABLE (1 << 1) |
2171 | #define PIRQ_MSI_GROUP (1 << 2) |
2172 | |
2173 | +struct evtchn_loop_ctrl; |
2174 | + |
2175 | struct evtchn_ops { |
2176 | unsigned (*max_channels)(void); |
2177 | unsigned (*nr_channels)(void); |
2178 | @@ -69,14 +76,18 @@ struct evtchn_ops { |
2179 | void (*mask)(unsigned port); |
2180 | void (*unmask)(unsigned port); |
2181 | |
2182 | - void (*handle_events)(unsigned cpu); |
2183 | + void (*handle_events)(unsigned cpu, struct evtchn_loop_ctrl *ctrl); |
2184 | void (*resume)(void); |
2185 | + |
2186 | + int (*percpu_init)(unsigned int cpu); |
2187 | + int (*percpu_deinit)(unsigned int cpu); |
2188 | }; |
2189 | |
2190 | extern const struct evtchn_ops *evtchn_ops; |
2191 | |
2192 | extern int **evtchn_to_irq; |
2193 | int get_evtchn_to_irq(unsigned int evtchn); |
2194 | +void handle_irq_for_port(evtchn_port_t port, struct evtchn_loop_ctrl *ctrl); |
2195 | |
2196 | struct irq_info *info_for_irq(unsigned irq); |
2197 | unsigned cpu_from_irq(unsigned irq); |
2198 | @@ -134,9 +145,10 @@ static inline void unmask_evtchn(unsigned port) |
2199 | return evtchn_ops->unmask(port); |
2200 | } |
2201 | |
2202 | -static inline void xen_evtchn_handle_events(unsigned cpu) |
2203 | +static inline void xen_evtchn_handle_events(unsigned cpu, |
2204 | + struct evtchn_loop_ctrl *ctrl) |
2205 | { |
2206 | - return evtchn_ops->handle_events(cpu); |
2207 | + return evtchn_ops->handle_events(cpu, ctrl); |
2208 | } |
2209 | |
2210 | static inline void xen_evtchn_resume(void) |
2211 | diff --git a/drivers/xen/evtchn.c b/drivers/xen/evtchn.c |
2212 | index e8c7f09d01be8..bcf0b1e60f2c3 100644 |
2213 | --- a/drivers/xen/evtchn.c |
2214 | +++ b/drivers/xen/evtchn.c |
2215 | @@ -178,7 +178,6 @@ static irqreturn_t evtchn_interrupt(int irq, void *data) |
2216 | "Interrupt for port %d, but apparently not enabled; per-user %p\n", |
2217 | evtchn->port, u); |
2218 | |
2219 | - disable_irq_nosync(irq); |
2220 | evtchn->enabled = false; |
2221 | |
2222 | spin_lock(&u->ring_prod_lock); |
2223 | @@ -304,7 +303,7 @@ static ssize_t evtchn_write(struct file *file, const char __user *buf, |
2224 | evtchn = find_evtchn(u, port); |
2225 | if (evtchn && !evtchn->enabled) { |
2226 | evtchn->enabled = true; |
2227 | - enable_irq(irq_from_evtchn(port)); |
2228 | + xen_irq_lateeoi(irq_from_evtchn(port), 0); |
2229 | } |
2230 | } |
2231 | |
2232 | @@ -404,8 +403,8 @@ static int evtchn_bind_to_user(struct per_user_data *u, int port) |
2233 | if (rc < 0) |
2234 | goto err; |
2235 | |
2236 | - rc = bind_evtchn_to_irqhandler(port, evtchn_interrupt, 0, |
2237 | - u->name, evtchn); |
2238 | + rc = bind_evtchn_to_irqhandler_lateeoi(port, evtchn_interrupt, 0, |
2239 | + u->name, evtchn); |
2240 | if (rc < 0) |
2241 | goto err; |
2242 | |
2243 | diff --git a/drivers/xen/xen-pciback/pci_stub.c b/drivers/xen/xen-pciback/pci_stub.c |
2244 | index ee5ce9286d613..83d798d12400e 100644 |
2245 | --- a/drivers/xen/xen-pciback/pci_stub.c |
2246 | +++ b/drivers/xen/xen-pciback/pci_stub.c |
2247 | @@ -733,10 +733,17 @@ static pci_ers_result_t common_process(struct pcistub_device *psdev, |
2248 | wmb(); |
2249 | notify_remote_via_irq(pdev->evtchn_irq); |
2250 | |
2251 | + /* Enable IRQ to signal "request done". */ |
2252 | + xen_pcibk_lateeoi(pdev, 0); |
2253 | + |
2254 | ret = wait_event_timeout(xen_pcibk_aer_wait_queue, |
2255 | !(test_bit(_XEN_PCIB_active, (unsigned long *) |
2256 | &sh_info->flags)), 300*HZ); |
2257 | |
2258 | + /* Enable IRQ for pcifront request if not already active. */ |
2259 | + if (!test_bit(_PDEVF_op_active, &pdev->flags)) |
2260 | + xen_pcibk_lateeoi(pdev, 0); |
2261 | + |
2262 | if (!ret) { |
2263 | if (test_bit(_XEN_PCIB_active, |
2264 | (unsigned long *)&sh_info->flags)) { |
2265 | @@ -750,13 +757,6 @@ static pci_ers_result_t common_process(struct pcistub_device *psdev, |
2266 | } |
2267 | clear_bit(_PCIB_op_pending, (unsigned long *)&pdev->flags); |
2268 | |
2269 | - if (test_bit(_XEN_PCIF_active, |
2270 | - (unsigned long *)&sh_info->flags)) { |
2271 | - dev_dbg(&psdev->dev->dev, |
2272 | - "schedule pci_conf service in " DRV_NAME "\n"); |
2273 | - xen_pcibk_test_and_schedule_op(psdev->pdev); |
2274 | - } |
2275 | - |
2276 | res = (pci_ers_result_t)aer_op->err; |
2277 | return res; |
2278 | } |
2279 | diff --git a/drivers/xen/xen-pciback/pciback.h b/drivers/xen/xen-pciback/pciback.h |
2280 | index 7af369b6aaa29..b97cf348cdc18 100644 |
2281 | --- a/drivers/xen/xen-pciback/pciback.h |
2282 | +++ b/drivers/xen/xen-pciback/pciback.h |
2283 | @@ -13,6 +13,7 @@ |
2284 | #include <linux/spinlock.h> |
2285 | #include <linux/workqueue.h> |
2286 | #include <linux/atomic.h> |
2287 | +#include <xen/events.h> |
2288 | #include <xen/interface/io/pciif.h> |
2289 | |
2290 | #define DRV_NAME "xen-pciback" |
2291 | @@ -26,6 +27,8 @@ struct pci_dev_entry { |
2292 | #define PDEVF_op_active (1<<(_PDEVF_op_active)) |
2293 | #define _PCIB_op_pending (1) |
2294 | #define PCIB_op_pending (1<<(_PCIB_op_pending)) |
2295 | +#define _EOI_pending (2) |
2296 | +#define EOI_pending (1<<(_EOI_pending)) |
2297 | |
2298 | struct xen_pcibk_device { |
2299 | void *pci_dev_data; |
2300 | @@ -181,12 +184,17 @@ static inline void xen_pcibk_release_devices(struct xen_pcibk_device *pdev) |
2301 | irqreturn_t xen_pcibk_handle_event(int irq, void *dev_id); |
2302 | void xen_pcibk_do_op(struct work_struct *data); |
2303 | |
2304 | +static inline void xen_pcibk_lateeoi(struct xen_pcibk_device *pdev, |
2305 | + unsigned int eoi_flag) |
2306 | +{ |
2307 | + if (test_and_clear_bit(_EOI_pending, &pdev->flags)) |
2308 | + xen_irq_lateeoi(pdev->evtchn_irq, eoi_flag); |
2309 | +} |
2310 | + |
2311 | int xen_pcibk_xenbus_register(void); |
2312 | void xen_pcibk_xenbus_unregister(void); |
2313 | |
2314 | extern int verbose_request; |
2315 | - |
2316 | -void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev); |
2317 | #endif |
2318 | |
2319 | /* Handles shared IRQs that can to device domain and control domain. */ |
2320 | diff --git a/drivers/xen/xen-pciback/pciback_ops.c b/drivers/xen/xen-pciback/pciback_ops.c |
2321 | index e7fbed56c0446..eb5dd80530e74 100644 |
2322 | --- a/drivers/xen/xen-pciback/pciback_ops.c |
2323 | +++ b/drivers/xen/xen-pciback/pciback_ops.c |
2324 | @@ -296,26 +296,41 @@ int xen_pcibk_disable_msix(struct xen_pcibk_device *pdev, |
2325 | return 0; |
2326 | } |
2327 | #endif |
2328 | + |
2329 | +static inline bool xen_pcibk_test_op_pending(struct xen_pcibk_device *pdev) |
2330 | +{ |
2331 | + return test_bit(_XEN_PCIF_active, |
2332 | + (unsigned long *)&pdev->sh_info->flags) && |
2333 | + !test_and_set_bit(_PDEVF_op_active, &pdev->flags); |
2334 | +} |
2335 | + |
2336 | /* |
2337 | * Now the same evtchn is used for both pcifront conf_read_write request |
2338 | * as well as pcie aer front end ack. We use a new work_queue to schedule |
2339 | * xen_pcibk conf_read_write service for avoiding confict with aer_core |
2340 | * do_recovery job which also use the system default work_queue |
2341 | */ |
2342 | -void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev) |
2343 | +static void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev) |
2344 | { |
2345 | + bool eoi = true; |
2346 | + |
2347 | /* Check that frontend is requesting an operation and that we are not |
2348 | * already processing a request */ |
2349 | - if (test_bit(_XEN_PCIF_active, (unsigned long *)&pdev->sh_info->flags) |
2350 | - && !test_and_set_bit(_PDEVF_op_active, &pdev->flags)) { |
2351 | + if (xen_pcibk_test_op_pending(pdev)) { |
2352 | schedule_work(&pdev->op_work); |
2353 | + eoi = false; |
2354 | } |
2355 | /*_XEN_PCIB_active should have been cleared by pcifront. And also make |
2356 | sure xen_pcibk is waiting for ack by checking _PCIB_op_pending*/ |
2357 | if (!test_bit(_XEN_PCIB_active, (unsigned long *)&pdev->sh_info->flags) |
2358 | && test_bit(_PCIB_op_pending, &pdev->flags)) { |
2359 | wake_up(&xen_pcibk_aer_wait_queue); |
2360 | + eoi = false; |
2361 | } |
2362 | + |
2363 | + /* EOI if there was nothing to do. */ |
2364 | + if (eoi) |
2365 | + xen_pcibk_lateeoi(pdev, XEN_EOI_FLAG_SPURIOUS); |
2366 | } |
2367 | |
2368 | /* Performing the configuration space reads/writes must not be done in atomic |
2369 | @@ -323,10 +338,8 @@ void xen_pcibk_test_and_schedule_op(struct xen_pcibk_device *pdev) |
2370 | * use of semaphores). This function is intended to be called from a work |
2371 | * queue in process context taking a struct xen_pcibk_device as a parameter */ |
2372 | |
2373 | -void xen_pcibk_do_op(struct work_struct *data) |
2374 | +static void xen_pcibk_do_one_op(struct xen_pcibk_device *pdev) |
2375 | { |
2376 | - struct xen_pcibk_device *pdev = |
2377 | - container_of(data, struct xen_pcibk_device, op_work); |
2378 | struct pci_dev *dev; |
2379 | struct xen_pcibk_dev_data *dev_data = NULL; |
2380 | struct xen_pci_op *op = &pdev->op; |
2381 | @@ -399,16 +412,31 @@ void xen_pcibk_do_op(struct work_struct *data) |
2382 | smp_mb__before_atomic(); /* /after/ clearing PCIF_active */ |
2383 | clear_bit(_PDEVF_op_active, &pdev->flags); |
2384 | smp_mb__after_atomic(); /* /before/ final check for work */ |
2385 | +} |
2386 | |
2387 | - /* Check to see if the driver domain tried to start another request in |
2388 | - * between clearing _XEN_PCIF_active and clearing _PDEVF_op_active. |
2389 | - */ |
2390 | - xen_pcibk_test_and_schedule_op(pdev); |
2391 | +void xen_pcibk_do_op(struct work_struct *data) |
2392 | +{ |
2393 | + struct xen_pcibk_device *pdev = |
2394 | + container_of(data, struct xen_pcibk_device, op_work); |
2395 | + |
2396 | + do { |
2397 | + xen_pcibk_do_one_op(pdev); |
2398 | + } while (xen_pcibk_test_op_pending(pdev)); |
2399 | + |
2400 | + xen_pcibk_lateeoi(pdev, 0); |
2401 | } |
2402 | |
2403 | irqreturn_t xen_pcibk_handle_event(int irq, void *dev_id) |
2404 | { |
2405 | struct xen_pcibk_device *pdev = dev_id; |
2406 | + bool eoi; |
2407 | + |
2408 | + /* IRQs might come in before pdev->evtchn_irq is written. */ |
2409 | + if (unlikely(pdev->evtchn_irq != irq)) |
2410 | + pdev->evtchn_irq = irq; |
2411 | + |
2412 | + eoi = test_and_set_bit(_EOI_pending, &pdev->flags); |
2413 | + WARN(eoi, "IRQ while EOI pending\n"); |
2414 | |
2415 | xen_pcibk_test_and_schedule_op(pdev); |
2416 | |
2417 | diff --git a/drivers/xen/xen-pciback/xenbus.c b/drivers/xen/xen-pciback/xenbus.c |
2418 | index 5ce878c51d034..f33eb40cb4148 100644 |
2419 | --- a/drivers/xen/xen-pciback/xenbus.c |
2420 | +++ b/drivers/xen/xen-pciback/xenbus.c |
2421 | @@ -122,7 +122,7 @@ static int xen_pcibk_do_attach(struct xen_pcibk_device *pdev, int gnt_ref, |
2422 | |
2423 | pdev->sh_info = vaddr; |
2424 | |
2425 | - err = bind_interdomain_evtchn_to_irqhandler( |
2426 | + err = bind_interdomain_evtchn_to_irqhandler_lateeoi( |
2427 | pdev->xdev->otherend_id, remote_evtchn, xen_pcibk_handle_event, |
2428 | 0, DRV_NAME, pdev); |
2429 | if (err < 0) { |
2430 | diff --git a/drivers/xen/xen-scsiback.c b/drivers/xen/xen-scsiback.c |
2431 | index 992cb8fa272c2..3243d917651a3 100644 |
2432 | --- a/drivers/xen/xen-scsiback.c |
2433 | +++ b/drivers/xen/xen-scsiback.c |
2434 | @@ -91,7 +91,6 @@ struct vscsibk_info { |
2435 | unsigned int irq; |
2436 | |
2437 | struct vscsiif_back_ring ring; |
2438 | - int ring_error; |
2439 | |
2440 | spinlock_t ring_lock; |
2441 | atomic_t nr_unreplied_reqs; |
2442 | @@ -723,7 +722,8 @@ static struct vscsibk_pend *prepare_pending_reqs(struct vscsibk_info *info, |
2443 | return pending_req; |
2444 | } |
2445 | |
2446 | -static int scsiback_do_cmd_fn(struct vscsibk_info *info) |
2447 | +static int scsiback_do_cmd_fn(struct vscsibk_info *info, |
2448 | + unsigned int *eoi_flags) |
2449 | { |
2450 | struct vscsiif_back_ring *ring = &info->ring; |
2451 | struct vscsiif_request ring_req; |
2452 | @@ -740,11 +740,12 @@ static int scsiback_do_cmd_fn(struct vscsibk_info *info) |
2453 | rc = ring->rsp_prod_pvt; |
2454 | pr_warn("Dom%d provided bogus ring requests (%#x - %#x = %u). Halting ring processing\n", |
2455 | info->domid, rp, rc, rp - rc); |
2456 | - info->ring_error = 1; |
2457 | - return 0; |
2458 | + return -EINVAL; |
2459 | } |
2460 | |
2461 | while ((rc != rp)) { |
2462 | + *eoi_flags &= ~XEN_EOI_FLAG_SPURIOUS; |
2463 | + |
2464 | if (RING_REQUEST_CONS_OVERFLOW(ring, rc)) |
2465 | break; |
2466 | |
2467 | @@ -803,13 +804,16 @@ static int scsiback_do_cmd_fn(struct vscsibk_info *info) |
2468 | static irqreturn_t scsiback_irq_fn(int irq, void *dev_id) |
2469 | { |
2470 | struct vscsibk_info *info = dev_id; |
2471 | + int rc; |
2472 | + unsigned int eoi_flags = XEN_EOI_FLAG_SPURIOUS; |
2473 | |
2474 | - if (info->ring_error) |
2475 | - return IRQ_HANDLED; |
2476 | - |
2477 | - while (scsiback_do_cmd_fn(info)) |
2478 | + while ((rc = scsiback_do_cmd_fn(info, &eoi_flags)) > 0) |
2479 | cond_resched(); |
2480 | |
2481 | + /* In case of a ring error we keep the event channel masked. */ |
2482 | + if (!rc) |
2483 | + xen_irq_lateeoi(irq, eoi_flags); |
2484 | + |
2485 | return IRQ_HANDLED; |
2486 | } |
2487 | |
2488 | @@ -830,7 +834,7 @@ static int scsiback_init_sring(struct vscsibk_info *info, grant_ref_t ring_ref, |
2489 | sring = (struct vscsiif_sring *)area; |
2490 | BACK_RING_INIT(&info->ring, sring, PAGE_SIZE); |
2491 | |
2492 | - err = bind_interdomain_evtchn_to_irq(info->domid, evtchn); |
2493 | + err = bind_interdomain_evtchn_to_irq_lateeoi(info->domid, evtchn); |
2494 | if (err < 0) |
2495 | goto unmap_page; |
2496 | |
2497 | @@ -1253,7 +1257,6 @@ static int scsiback_probe(struct xenbus_device *dev, |
2498 | |
2499 | info->domid = dev->otherend_id; |
2500 | spin_lock_init(&info->ring_lock); |
2501 | - info->ring_error = 0; |
2502 | atomic_set(&info->nr_unreplied_reqs, 0); |
2503 | init_waitqueue_head(&info->waiting_to_free); |
2504 | info->dev = dev; |
2505 | diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c |
2506 | index d6c827a9ebc56..5c2f4f58da8ff 100644 |
2507 | --- a/fs/btrfs/extent_io.c |
2508 | +++ b/fs/btrfs/extent_io.c |
2509 | @@ -3879,6 +3879,10 @@ retry: |
2510 | if (!ret) { |
2511 | free_extent_buffer(eb); |
2512 | continue; |
2513 | + } else if (ret < 0) { |
2514 | + done = 1; |
2515 | + free_extent_buffer(eb); |
2516 | + break; |
2517 | } |
2518 | |
2519 | ret = write_one_eb(eb, fs_info, wbc, &epd); |
2520 | diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c |
2521 | index 981091bd6c3c4..ebca009030c3a 100644 |
2522 | --- a/fs/btrfs/ioctl.c |
2523 | +++ b/fs/btrfs/ioctl.c |
2524 | @@ -3854,6 +3854,8 @@ process_slot: |
2525 | ret = -EINTR; |
2526 | goto out; |
2527 | } |
2528 | + |
2529 | + cond_resched(); |
2530 | } |
2531 | ret = 0; |
2532 | |
2533 | diff --git a/fs/cifs/cifs_unicode.c b/fs/cifs/cifs_unicode.c |
2534 | index 211ac472cb9dc..942874257a092 100644 |
2535 | --- a/fs/cifs/cifs_unicode.c |
2536 | +++ b/fs/cifs/cifs_unicode.c |
2537 | @@ -493,7 +493,13 @@ cifsConvertToUTF16(__le16 *target, const char *source, int srclen, |
2538 | else if (map_chars == SFM_MAP_UNI_RSVD) { |
2539 | bool end_of_string; |
2540 | |
2541 | - if (i == srclen - 1) |
2542 | + /** |
2543 | + * Remap spaces and periods found at the end of every |
2544 | + * component of the path. The special cases of '.' and |
2545 | + * '..' do not need to be dealt with explicitly because |
2546 | + * they are addressed in namei.c:link_path_walk(). |
2547 | + **/ |
2548 | + if ((i == srclen - 1) || (source[i+1] == '\\')) |
2549 | end_of_string = true; |
2550 | else |
2551 | end_of_string = false; |
2552 | diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c |
2553 | index 4df4d31057b36..5bf62bdd84b87 100644 |
2554 | --- a/fs/ext4/inline.c |
2555 | +++ b/fs/ext4/inline.c |
2556 | @@ -1890,6 +1890,7 @@ void ext4_inline_data_truncate(struct inode *inode, int *has_inline) |
2557 | |
2558 | ext4_write_lock_xattr(inode, &no_expand); |
2559 | if (!ext4_has_inline_data(inode)) { |
2560 | + ext4_write_unlock_xattr(inode, &no_expand); |
2561 | *has_inline = 0; |
2562 | ext4_journal_stop(handle); |
2563 | return; |
2564 | diff --git a/fs/ext4/super.c b/fs/ext4/super.c |
2565 | index 2527eb3049494..84d28d3efc602 100644 |
2566 | --- a/fs/ext4/super.c |
2567 | +++ b/fs/ext4/super.c |
2568 | @@ -1571,8 +1571,8 @@ static const struct mount_opts { |
2569 | {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | |
2570 | EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), |
2571 | MOPT_CLEAR | MOPT_Q}, |
2572 | - {Opt_usrjquota, 0, MOPT_Q}, |
2573 | - {Opt_grpjquota, 0, MOPT_Q}, |
2574 | + {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING}, |
2575 | + {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING}, |
2576 | {Opt_offusrjquota, 0, MOPT_Q}, |
2577 | {Opt_offgrpjquota, 0, MOPT_Q}, |
2578 | {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, |
2579 | @@ -4336,6 +4336,7 @@ cantfind_ext4: |
2580 | #ifdef CONFIG_QUOTA |
2581 | failed_mount8: |
2582 | ext4_unregister_sysfs(sb); |
2583 | + kobject_put(&sbi->s_kobj); |
2584 | #endif |
2585 | failed_mount7: |
2586 | ext4_unregister_li_request(sb); |
2587 | diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c |
2588 | index efd44d5645d83..f19e49a5d032b 100644 |
2589 | --- a/fs/gfs2/glock.c |
2590 | +++ b/fs/gfs2/glock.c |
2591 | @@ -758,7 +758,8 @@ again: |
2592 | } |
2593 | kfree(gl->gl_lksb.sb_lvbptr); |
2594 | kmem_cache_free(cachep, gl); |
2595 | - atomic_dec(&sdp->sd_glock_disposal); |
2596 | + if (atomic_dec_and_test(&sdp->sd_glock_disposal)) |
2597 | + wake_up(&sdp->sd_glock_wait); |
2598 | *glp = tmp; |
2599 | |
2600 | return ret; |
2601 | diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c |
2602 | index 0a80f66365492..9621badb95995 100644 |
2603 | --- a/fs/gfs2/rgrp.c |
2604 | +++ b/fs/gfs2/rgrp.c |
2605 | @@ -730,9 +730,9 @@ void gfs2_clear_rgrpd(struct gfs2_sbd *sdp) |
2606 | } |
2607 | |
2608 | gfs2_free_clones(rgd); |
2609 | + return_all_reservations(rgd); |
2610 | kfree(rgd->rd_bits); |
2611 | rgd->rd_bits = NULL; |
2612 | - return_all_reservations(rgd); |
2613 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); |
2614 | } |
2615 | } |
2616 | @@ -1371,6 +1371,9 @@ int gfs2_fitrim(struct file *filp, void __user *argp) |
2617 | if (!capable(CAP_SYS_ADMIN)) |
2618 | return -EPERM; |
2619 | |
2620 | + if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) |
2621 | + return -EROFS; |
2622 | + |
2623 | if (!blk_queue_discard(q)) |
2624 | return -EOPNOTSUPP; |
2625 | |
2626 | diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c |
2627 | index 4d6e99ea37b8e..e0fb62f5cf63d 100644 |
2628 | --- a/fs/ocfs2/super.c |
2629 | +++ b/fs/ocfs2/super.c |
2630 | @@ -1733,6 +1733,7 @@ static void ocfs2_inode_init_once(void *data) |
2631 | |
2632 | oi->ip_blkno = 0ULL; |
2633 | oi->ip_clusters = 0; |
2634 | + oi->ip_next_orphan = NULL; |
2635 | |
2636 | ocfs2_resv_init_once(&oi->ip_la_data_resv); |
2637 | |
2638 | diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c |
2639 | index 3a8cc7139912b..89fdcc641715f 100644 |
2640 | --- a/fs/xfs/libxfs/xfs_rmap.c |
2641 | +++ b/fs/xfs/libxfs/xfs_rmap.c |
2642 | @@ -1318,7 +1318,7 @@ xfs_rmap_convert_shared( |
2643 | * record for our insertion point. This will also give us the record for |
2644 | * start block contiguity tests. |
2645 | */ |
2646 | - error = xfs_rmap_lookup_le_range(cur, bno, owner, offset, flags, |
2647 | + error = xfs_rmap_lookup_le_range(cur, bno, owner, offset, oldext, |
2648 | &PREV, &i); |
2649 | XFS_WANT_CORRUPTED_GOTO(mp, i == 1, done); |
2650 | |
2651 | diff --git a/fs/xfs/libxfs/xfs_rmap_btree.c b/fs/xfs/libxfs/xfs_rmap_btree.c |
2652 | index 33a28efc3085b..c5a24b80c7f72 100644 |
2653 | --- a/fs/xfs/libxfs/xfs_rmap_btree.c |
2654 | +++ b/fs/xfs/libxfs/xfs_rmap_btree.c |
2655 | @@ -262,8 +262,8 @@ xfs_rmapbt_key_diff( |
2656 | else if (y > x) |
2657 | return -1; |
2658 | |
2659 | - x = XFS_RMAP_OFF(be64_to_cpu(kp->rm_offset)); |
2660 | - y = rec->rm_offset; |
2661 | + x = be64_to_cpu(kp->rm_offset); |
2662 | + y = xfs_rmap_irec_offset_pack(rec); |
2663 | if (x > y) |
2664 | return 1; |
2665 | else if (y > x) |
2666 | @@ -294,8 +294,8 @@ xfs_rmapbt_diff_two_keys( |
2667 | else if (y > x) |
2668 | return -1; |
2669 | |
2670 | - x = XFS_RMAP_OFF(be64_to_cpu(kp1->rm_offset)); |
2671 | - y = XFS_RMAP_OFF(be64_to_cpu(kp2->rm_offset)); |
2672 | + x = be64_to_cpu(kp1->rm_offset); |
2673 | + y = be64_to_cpu(kp2->rm_offset); |
2674 | if (x > y) |
2675 | return 1; |
2676 | else if (y > x) |
2677 | @@ -401,8 +401,8 @@ xfs_rmapbt_keys_inorder( |
2678 | return 1; |
2679 | else if (a > b) |
2680 | return 0; |
2681 | - a = XFS_RMAP_OFF(be64_to_cpu(k1->rmap.rm_offset)); |
2682 | - b = XFS_RMAP_OFF(be64_to_cpu(k2->rmap.rm_offset)); |
2683 | + a = be64_to_cpu(k1->rmap.rm_offset); |
2684 | + b = be64_to_cpu(k2->rmap.rm_offset); |
2685 | if (a <= b) |
2686 | return 1; |
2687 | return 0; |
2688 | @@ -431,8 +431,8 @@ xfs_rmapbt_recs_inorder( |
2689 | return 1; |
2690 | else if (a > b) |
2691 | return 0; |
2692 | - a = XFS_RMAP_OFF(be64_to_cpu(r1->rmap.rm_offset)); |
2693 | - b = XFS_RMAP_OFF(be64_to_cpu(r2->rmap.rm_offset)); |
2694 | + a = be64_to_cpu(r1->rmap.rm_offset); |
2695 | + b = be64_to_cpu(r2->rmap.rm_offset); |
2696 | if (a <= b) |
2697 | return 1; |
2698 | return 0; |
2699 | diff --git a/fs/xfs/xfs_iops.c b/fs/xfs/xfs_iops.c |
2700 | index 7bfddcd32d73e..0d587657056d8 100644 |
2701 | --- a/fs/xfs/xfs_iops.c |
2702 | +++ b/fs/xfs/xfs_iops.c |
2703 | @@ -864,6 +864,16 @@ xfs_setattr_size( |
2704 | if (newsize > oldsize) { |
2705 | error = xfs_zero_eof(ip, newsize, oldsize, &did_zeroing); |
2706 | } else { |
2707 | + /* |
2708 | + * iomap won't detect a dirty page over an unwritten block (or a |
2709 | + * cow block over a hole) and subsequently skips zeroing the |
2710 | + * newly post-EOF portion of the page. Flush the new EOF to |
2711 | + * convert the block before the pagecache truncate. |
2712 | + */ |
2713 | + error = filemap_write_and_wait_range(inode->i_mapping, newsize, |
2714 | + newsize); |
2715 | + if (error) |
2716 | + return error; |
2717 | error = iomap_truncate_page(inode, newsize, &did_zeroing, |
2718 | &xfs_iomap_ops); |
2719 | } |
2720 | diff --git a/fs/xfs/xfs_pnfs.c b/fs/xfs/xfs_pnfs.c |
2721 | index cecd37569ddb3..353bfe9c5cdd9 100644 |
2722 | --- a/fs/xfs/xfs_pnfs.c |
2723 | +++ b/fs/xfs/xfs_pnfs.c |
2724 | @@ -144,7 +144,7 @@ xfs_fs_map_blocks( |
2725 | goto out_unlock; |
2726 | error = invalidate_inode_pages2(inode->i_mapping); |
2727 | if (WARN_ON_ONCE(error)) |
2728 | - return error; |
2729 | + goto out_unlock; |
2730 | |
2731 | end_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)offset + length); |
2732 | offset_fsb = XFS_B_TO_FSBT(mp, offset); |
2733 | diff --git a/include/linux/can/skb.h b/include/linux/can/skb.h |
2734 | index 51bb6532785c3..1a2111c775ae1 100644 |
2735 | --- a/include/linux/can/skb.h |
2736 | +++ b/include/linux/can/skb.h |
2737 | @@ -60,21 +60,17 @@ static inline void can_skb_set_owner(struct sk_buff *skb, struct sock *sk) |
2738 | */ |
2739 | static inline struct sk_buff *can_create_echo_skb(struct sk_buff *skb) |
2740 | { |
2741 | - if (skb_shared(skb)) { |
2742 | - struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC); |
2743 | + struct sk_buff *nskb; |
2744 | |
2745 | - if (likely(nskb)) { |
2746 | - can_skb_set_owner(nskb, skb->sk); |
2747 | - consume_skb(skb); |
2748 | - return nskb; |
2749 | - } else { |
2750 | - kfree_skb(skb); |
2751 | - return NULL; |
2752 | - } |
2753 | + nskb = skb_clone(skb, GFP_ATOMIC); |
2754 | + if (unlikely(!nskb)) { |
2755 | + kfree_skb(skb); |
2756 | + return NULL; |
2757 | } |
2758 | |
2759 | - /* we can assume to have an unshared skb with proper owner */ |
2760 | - return skb; |
2761 | + can_skb_set_owner(nskb, skb->sk); |
2762 | + consume_skb(skb); |
2763 | + return nskb; |
2764 | } |
2765 | |
2766 | #endif /* !_CAN_SKB_H */ |
2767 | diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h |
2768 | index ae8ecf8210192..a7057b7726129 100644 |
2769 | --- a/include/linux/perf_event.h |
2770 | +++ b/include/linux/perf_event.h |
2771 | @@ -475,7 +475,7 @@ struct pmu { |
2772 | */ |
2773 | struct perf_addr_filter { |
2774 | struct list_head entry; |
2775 | - struct inode *inode; |
2776 | + struct path path; |
2777 | unsigned long offset; |
2778 | unsigned long size; |
2779 | unsigned int range : 1, |
2780 | diff --git a/include/linux/prandom.h b/include/linux/prandom.h |
2781 | index aa16e6468f91e..cc1e71334e53c 100644 |
2782 | --- a/include/linux/prandom.h |
2783 | +++ b/include/linux/prandom.h |
2784 | @@ -16,12 +16,44 @@ void prandom_bytes(void *buf, size_t nbytes); |
2785 | void prandom_seed(u32 seed); |
2786 | void prandom_reseed_late(void); |
2787 | |
2788 | +#if BITS_PER_LONG == 64 |
2789 | +/* |
2790 | + * The core SipHash round function. Each line can be executed in |
2791 | + * parallel given enough CPU resources. |
2792 | + */ |
2793 | +#define PRND_SIPROUND(v0, v1, v2, v3) ( \ |
2794 | + v0 += v1, v1 = rol64(v1, 13), v2 += v3, v3 = rol64(v3, 16), \ |
2795 | + v1 ^= v0, v0 = rol64(v0, 32), v3 ^= v2, \ |
2796 | + v0 += v3, v3 = rol64(v3, 21), v2 += v1, v1 = rol64(v1, 17), \ |
2797 | + v3 ^= v0, v1 ^= v2, v2 = rol64(v2, 32) \ |
2798 | +) |
2799 | + |
2800 | +#define PRND_K0 (0x736f6d6570736575 ^ 0x6c7967656e657261) |
2801 | +#define PRND_K1 (0x646f72616e646f6d ^ 0x7465646279746573) |
2802 | + |
2803 | +#elif BITS_PER_LONG == 32 |
2804 | +/* |
2805 | + * On 32-bit machines, we use HSipHash, a reduced-width version of SipHash. |
2806 | + * This is weaker, but 32-bit machines are not used for high-traffic |
2807 | + * applications, so there is less output for an attacker to analyze. |
2808 | + */ |
2809 | +#define PRND_SIPROUND(v0, v1, v2, v3) ( \ |
2810 | + v0 += v1, v1 = rol32(v1, 5), v2 += v3, v3 = rol32(v3, 8), \ |
2811 | + v1 ^= v0, v0 = rol32(v0, 16), v3 ^= v2, \ |
2812 | + v0 += v3, v3 = rol32(v3, 7), v2 += v1, v1 = rol32(v1, 13), \ |
2813 | + v3 ^= v0, v1 ^= v2, v2 = rol32(v2, 16) \ |
2814 | +) |
2815 | +#define PRND_K0 0x6c796765 |
2816 | +#define PRND_K1 0x74656462 |
2817 | + |
2818 | +#else |
2819 | +#error Unsupported BITS_PER_LONG |
2820 | +#endif |
2821 | + |
2822 | struct rnd_state { |
2823 | __u32 s1, s2, s3, s4; |
2824 | }; |
2825 | |
2826 | -DECLARE_PER_CPU(struct rnd_state, net_rand_state); |
2827 | - |
2828 | u32 prandom_u32_state(struct rnd_state *state); |
2829 | void prandom_bytes_state(struct rnd_state *state, void *buf, size_t nbytes); |
2830 | void prandom_seed_full_state(struct rnd_state __percpu *pcpu_state); |
2831 | diff --git a/include/linux/time64.h b/include/linux/time64.h |
2832 | index 980c71b3001a5..2a45b8c87edbf 100644 |
2833 | --- a/include/linux/time64.h |
2834 | +++ b/include/linux/time64.h |
2835 | @@ -188,6 +188,10 @@ static inline bool timespec64_valid_strict(const struct timespec64 *ts) |
2836 | */ |
2837 | static inline s64 timespec64_to_ns(const struct timespec64 *ts) |
2838 | { |
2839 | + /* Prevent multiplication overflow */ |
2840 | + if ((unsigned long long)ts->tv_sec >= KTIME_SEC_MAX) |
2841 | + return KTIME_MAX; |
2842 | + |
2843 | return ((s64) ts->tv_sec * NSEC_PER_SEC) + ts->tv_nsec; |
2844 | } |
2845 | |
2846 | diff --git a/include/xen/events.h b/include/xen/events.h |
2847 | index 88da2abaf5359..ad0c61cf399be 100644 |
2848 | --- a/include/xen/events.h |
2849 | +++ b/include/xen/events.h |
2850 | @@ -12,11 +12,16 @@ |
2851 | |
2852 | unsigned xen_evtchn_nr_channels(void); |
2853 | |
2854 | -int bind_evtchn_to_irq(unsigned int evtchn); |
2855 | -int bind_evtchn_to_irqhandler(unsigned int evtchn, |
2856 | +int bind_evtchn_to_irq(evtchn_port_t evtchn); |
2857 | +int bind_evtchn_to_irq_lateeoi(evtchn_port_t evtchn); |
2858 | +int bind_evtchn_to_irqhandler(evtchn_port_t evtchn, |
2859 | irq_handler_t handler, |
2860 | unsigned long irqflags, const char *devname, |
2861 | void *dev_id); |
2862 | +int bind_evtchn_to_irqhandler_lateeoi(evtchn_port_t evtchn, |
2863 | + irq_handler_t handler, |
2864 | + unsigned long irqflags, const char *devname, |
2865 | + void *dev_id); |
2866 | int bind_virq_to_irq(unsigned int virq, unsigned int cpu, bool percpu); |
2867 | int bind_virq_to_irqhandler(unsigned int virq, unsigned int cpu, |
2868 | irq_handler_t handler, |
2869 | @@ -29,13 +34,21 @@ int bind_ipi_to_irqhandler(enum ipi_vector ipi, |
2870 | const char *devname, |
2871 | void *dev_id); |
2872 | int bind_interdomain_evtchn_to_irq(unsigned int remote_domain, |
2873 | - unsigned int remote_port); |
2874 | + evtchn_port_t remote_port); |
2875 | +int bind_interdomain_evtchn_to_irq_lateeoi(unsigned int remote_domain, |
2876 | + evtchn_port_t remote_port); |
2877 | int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain, |
2878 | - unsigned int remote_port, |
2879 | + evtchn_port_t remote_port, |
2880 | irq_handler_t handler, |
2881 | unsigned long irqflags, |
2882 | const char *devname, |
2883 | void *dev_id); |
2884 | +int bind_interdomain_evtchn_to_irqhandler_lateeoi(unsigned int remote_domain, |
2885 | + evtchn_port_t remote_port, |
2886 | + irq_handler_t handler, |
2887 | + unsigned long irqflags, |
2888 | + const char *devname, |
2889 | + void *dev_id); |
2890 | |
2891 | /* |
2892 | * Common unbind function for all event sources. Takes IRQ to unbind from. |
2893 | @@ -44,6 +57,14 @@ int bind_interdomain_evtchn_to_irqhandler(unsigned int remote_domain, |
2894 | */ |
2895 | void unbind_from_irqhandler(unsigned int irq, void *dev_id); |
2896 | |
2897 | +/* |
2898 | + * Send late EOI for an IRQ bound to an event channel via one of the *_lateeoi |
2899 | + * functions above. |
2900 | + */ |
2901 | +void xen_irq_lateeoi(unsigned int irq, unsigned int eoi_flags); |
2902 | +/* Signal an event was spurious, i.e. there was no action resulting from it. */ |
2903 | +#define XEN_EOI_FLAG_SPURIOUS 0x00000001 |
2904 | + |
2905 | #define XEN_IRQ_PRIORITY_MAX EVTCHN_FIFO_PRIORITY_MAX |
2906 | #define XEN_IRQ_PRIORITY_DEFAULT EVTCHN_FIFO_PRIORITY_DEFAULT |
2907 | #define XEN_IRQ_PRIORITY_MIN EVTCHN_FIFO_PRIORITY_MIN |
2908 | diff --git a/kernel/events/core.c b/kernel/events/core.c |
2909 | index 7aad4d22b4223..9d4a71deac992 100644 |
2910 | --- a/kernel/events/core.c |
2911 | +++ b/kernel/events/core.c |
2912 | @@ -5069,11 +5069,11 @@ static void perf_pmu_output_stop(struct perf_event *event); |
2913 | static void perf_mmap_close(struct vm_area_struct *vma) |
2914 | { |
2915 | struct perf_event *event = vma->vm_file->private_data; |
2916 | - |
2917 | struct ring_buffer *rb = ring_buffer_get(event); |
2918 | struct user_struct *mmap_user = rb->mmap_user; |
2919 | int mmap_locked = rb->mmap_locked; |
2920 | unsigned long size = perf_data_size(rb); |
2921 | + bool detach_rest = false; |
2922 | |
2923 | if (event->pmu->event_unmapped) |
2924 | event->pmu->event_unmapped(event); |
2925 | @@ -5104,7 +5104,8 @@ static void perf_mmap_close(struct vm_area_struct *vma) |
2926 | mutex_unlock(&event->mmap_mutex); |
2927 | } |
2928 | |
2929 | - atomic_dec(&rb->mmap_count); |
2930 | + if (atomic_dec_and_test(&rb->mmap_count)) |
2931 | + detach_rest = true; |
2932 | |
2933 | if (!atomic_dec_and_mutex_lock(&event->mmap_count, &event->mmap_mutex)) |
2934 | goto out_put; |
2935 | @@ -5113,7 +5114,7 @@ static void perf_mmap_close(struct vm_area_struct *vma) |
2936 | mutex_unlock(&event->mmap_mutex); |
2937 | |
2938 | /* If there's still other mmap()s of this buffer, we're done. */ |
2939 | - if (atomic_read(&rb->mmap_count)) |
2940 | + if (!detach_rest) |
2941 | goto out_put; |
2942 | |
2943 | /* |
2944 | @@ -6271,7 +6272,7 @@ static void perf_event_addr_filters_exec(struct perf_event *event, void *data) |
2945 | |
2946 | raw_spin_lock_irqsave(&ifh->lock, flags); |
2947 | list_for_each_entry(filter, &ifh->list, entry) { |
2948 | - if (filter->inode) { |
2949 | + if (filter->path.dentry) { |
2950 | event->addr_filters_offs[count] = 0; |
2951 | restart++; |
2952 | } |
2953 | @@ -6814,7 +6815,11 @@ static bool perf_addr_filter_match(struct perf_addr_filter *filter, |
2954 | struct file *file, unsigned long offset, |
2955 | unsigned long size) |
2956 | { |
2957 | - if (filter->inode != file->f_inode) |
2958 | + /* d_inode(NULL) won't be equal to any mapped user-space file */ |
2959 | + if (!filter->path.dentry) |
2960 | + return false; |
2961 | + |
2962 | + if (d_inode(filter->path.dentry) != file_inode(file)) |
2963 | return false; |
2964 | |
2965 | if (filter->offset > offset + size) |
2966 | @@ -8028,8 +8033,7 @@ static void free_filters_list(struct list_head *filters) |
2967 | struct perf_addr_filter *filter, *iter; |
2968 | |
2969 | list_for_each_entry_safe(filter, iter, filters, entry) { |
2970 | - if (filter->inode) |
2971 | - iput(filter->inode); |
2972 | + path_put(&filter->path); |
2973 | list_del(&filter->entry); |
2974 | kfree(filter); |
2975 | } |
2976 | @@ -8123,7 +8127,7 @@ static void perf_event_addr_filters_apply(struct perf_event *event) |
2977 | * Adjust base offset if the filter is associated to a binary |
2978 | * that needs to be mapped: |
2979 | */ |
2980 | - if (filter->inode) |
2981 | + if (filter->path.dentry) |
2982 | event->addr_filters_offs[count] = |
2983 | perf_addr_filter_apply(filter, mm); |
2984 | |
2985 | @@ -8196,7 +8200,6 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr, |
2986 | { |
2987 | struct perf_addr_filter *filter = NULL; |
2988 | char *start, *orig, *filename = NULL; |
2989 | - struct path path; |
2990 | substring_t args[MAX_OPT_ARGS]; |
2991 | int state = IF_STATE_ACTION, token; |
2992 | unsigned int kernel = 0; |
2993 | @@ -8259,6 +8262,7 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr, |
2994 | if (token == IF_SRC_FILE || token == IF_SRC_FILEADDR) { |
2995 | int fpos = filter->range ? 2 : 1; |
2996 | |
2997 | + kfree(filename); |
2998 | filename = match_strdup(&args[fpos]); |
2999 | if (!filename) { |
3000 | ret = -ENOMEM; |
3001 | @@ -8287,19 +8291,15 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr, |
3002 | goto fail; |
3003 | |
3004 | /* look up the path and grab its inode */ |
3005 | - ret = kern_path(filename, LOOKUP_FOLLOW, &path); |
3006 | + ret = kern_path(filename, LOOKUP_FOLLOW, |
3007 | + &filter->path); |
3008 | if (ret) |
3009 | - goto fail_free_name; |
3010 | - |
3011 | - filter->inode = igrab(d_inode(path.dentry)); |
3012 | - path_put(&path); |
3013 | - kfree(filename); |
3014 | - filename = NULL; |
3015 | + goto fail; |
3016 | |
3017 | ret = -EINVAL; |
3018 | - if (!filter->inode || |
3019 | - !S_ISREG(filter->inode->i_mode)) |
3020 | - /* free_filters_list() will iput() */ |
3021 | + if (!filter->path.dentry || |
3022 | + !S_ISREG(d_inode(filter->path.dentry) |
3023 | + ->i_mode)) |
3024 | goto fail; |
3025 | } |
3026 | |
3027 | @@ -8312,13 +8312,13 @@ perf_event_parse_addr_filter(struct perf_event *event, char *fstr, |
3028 | if (state != IF_STATE_ACTION) |
3029 | goto fail; |
3030 | |
3031 | + kfree(filename); |
3032 | kfree(orig); |
3033 | |
3034 | return 0; |
3035 | |
3036 | -fail_free_name: |
3037 | - kfree(filename); |
3038 | fail: |
3039 | + kfree(filename); |
3040 | free_filters_list(filters); |
3041 | kfree(orig); |
3042 | |
3043 | diff --git a/kernel/events/internal.h b/kernel/events/internal.h |
3044 | index 486fd78eb8d5e..c8c1c3db5d065 100644 |
3045 | --- a/kernel/events/internal.h |
3046 | +++ b/kernel/events/internal.h |
3047 | @@ -212,7 +212,7 @@ static inline int get_recursion_context(int *recursion) |
3048 | rctx = 3; |
3049 | else if (in_irq()) |
3050 | rctx = 2; |
3051 | - else if (in_softirq()) |
3052 | + else if (in_serving_softirq()) |
3053 | rctx = 1; |
3054 | else |
3055 | rctx = 0; |
3056 | diff --git a/kernel/exit.c b/kernel/exit.c |
3057 | index 27f4168eaeb1b..f9943ef23fa82 100644 |
3058 | --- a/kernel/exit.c |
3059 | +++ b/kernel/exit.c |
3060 | @@ -483,7 +483,10 @@ static void exit_mm(struct task_struct *tsk) |
3061 | up_read(&mm->mmap_sem); |
3062 | |
3063 | self.task = tsk; |
3064 | - self.next = xchg(&core_state->dumper.next, &self); |
3065 | + if (self.task->flags & PF_SIGNALED) |
3066 | + self.next = xchg(&core_state->dumper.next, &self); |
3067 | + else |
3068 | + self.task = NULL; |
3069 | /* |
3070 | * Implies mb(), the result of xchg() must be visible |
3071 | * to core_state->dumper. |
3072 | diff --git a/kernel/irq/Kconfig b/kernel/irq/Kconfig |
3073 | index 3bbfd6a9c4756..bb3a46cbe034c 100644 |
3074 | --- a/kernel/irq/Kconfig |
3075 | +++ b/kernel/irq/Kconfig |
3076 | @@ -67,6 +67,7 @@ config IRQ_DOMAIN_HIERARCHY |
3077 | # Generic IRQ IPI support |
3078 | config GENERIC_IRQ_IPI |
3079 | bool |
3080 | + select IRQ_DOMAIN_HIERARCHY |
3081 | |
3082 | # Generic MSI interrupt support |
3083 | config GENERIC_MSI_IRQ |
3084 | diff --git a/kernel/reboot.c b/kernel/reboot.c |
3085 | index bd30a973fe946..2946ed1d99d4d 100644 |
3086 | --- a/kernel/reboot.c |
3087 | +++ b/kernel/reboot.c |
3088 | @@ -512,22 +512,22 @@ static int __init reboot_setup(char *str) |
3089 | break; |
3090 | |
3091 | case 's': |
3092 | - { |
3093 | - int rc; |
3094 | - |
3095 | - if (isdigit(*(str+1))) { |
3096 | - rc = kstrtoint(str+1, 0, &reboot_cpu); |
3097 | - if (rc) |
3098 | - return rc; |
3099 | - } else if (str[1] == 'm' && str[2] == 'p' && |
3100 | - isdigit(*(str+3))) { |
3101 | - rc = kstrtoint(str+3, 0, &reboot_cpu); |
3102 | - if (rc) |
3103 | - return rc; |
3104 | - } else |
3105 | + if (isdigit(*(str+1))) |
3106 | + reboot_cpu = simple_strtoul(str+1, NULL, 0); |
3107 | + else if (str[1] == 'm' && str[2] == 'p' && |
3108 | + isdigit(*(str+3))) |
3109 | + reboot_cpu = simple_strtoul(str+3, NULL, 0); |
3110 | + else |
3111 | reboot_mode = REBOOT_SOFT; |
3112 | + if (reboot_cpu >= num_possible_cpus()) { |
3113 | + pr_err("Ignoring the CPU number in reboot= option. " |
3114 | + "CPU %d exceeds possible cpu number %d\n", |
3115 | + reboot_cpu, num_possible_cpus()); |
3116 | + reboot_cpu = 0; |
3117 | + break; |
3118 | + } |
3119 | break; |
3120 | - } |
3121 | + |
3122 | case 'g': |
3123 | reboot_mode = REBOOT_GPIO; |
3124 | break; |
3125 | diff --git a/kernel/time/timer.c b/kernel/time/timer.c |
3126 | index d2e4698d43fec..c9325a1e30d05 100644 |
3127 | --- a/kernel/time/timer.c |
3128 | +++ b/kernel/time/timer.c |
3129 | @@ -1636,13 +1636,6 @@ void update_process_times(int user_tick) |
3130 | #endif |
3131 | scheduler_tick(); |
3132 | run_posix_cpu_timers(p); |
3133 | - |
3134 | - /* The current CPU might make use of net randoms without receiving IRQs |
3135 | - * to renew them often enough. Let's update the net_rand_state from a |
3136 | - * non-constant value that's not affine to the number of calls to make |
3137 | - * sure it's updated when there's some activity (we don't care in idle). |
3138 | - */ |
3139 | - this_cpu_add(net_rand_state.s1, rol32(jiffies, 24) + user_tick); |
3140 | } |
3141 | |
3142 | /** |
3143 | diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c |
3144 | index fb2aa2430edcc..55f60d2edc3fb 100644 |
3145 | --- a/kernel/trace/ring_buffer.c |
3146 | +++ b/kernel/trace/ring_buffer.c |
3147 | @@ -416,14 +416,16 @@ struct rb_event_info { |
3148 | |
3149 | /* |
3150 | * Used for which event context the event is in. |
3151 | - * NMI = 0 |
3152 | - * IRQ = 1 |
3153 | - * SOFTIRQ = 2 |
3154 | - * NORMAL = 3 |
3155 | + * TRANSITION = 0 |
3156 | + * NMI = 1 |
3157 | + * IRQ = 2 |
3158 | + * SOFTIRQ = 3 |
3159 | + * NORMAL = 4 |
3160 | * |
3161 | * See trace_recursive_lock() comment below for more details. |
3162 | */ |
3163 | enum { |
3164 | + RB_CTX_TRANSITION, |
3165 | RB_CTX_NMI, |
3166 | RB_CTX_IRQ, |
3167 | RB_CTX_SOFTIRQ, |
3168 | @@ -2579,10 +2581,10 @@ rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer) |
3169 | * a bit of overhead in something as critical as function tracing, |
3170 | * we use a bitmask trick. |
3171 | * |
3172 | - * bit 0 = NMI context |
3173 | - * bit 1 = IRQ context |
3174 | - * bit 2 = SoftIRQ context |
3175 | - * bit 3 = normal context. |
3176 | + * bit 1 = NMI context |
3177 | + * bit 2 = IRQ context |
3178 | + * bit 3 = SoftIRQ context |
3179 | + * bit 4 = normal context. |
3180 | * |
3181 | * This works because this is the order of contexts that can |
3182 | * preempt other contexts. A SoftIRQ never preempts an IRQ |
3183 | @@ -2605,6 +2607,30 @@ rb_wakeups(struct ring_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer) |
3184 | * The least significant bit can be cleared this way, and it |
3185 | * just so happens that it is the same bit corresponding to |
3186 | * the current context. |
3187 | + * |
3188 | + * Now the TRANSITION bit breaks the above slightly. The TRANSITION bit |
3189 | + * is set when a recursion is detected at the current context, and if |
3190 | + * the TRANSITION bit is already set, it will fail the recursion. |
3191 | + * This is needed because there's a lag between the changing of |
3192 | + * interrupt context and updating the preempt count. In this case, |
3193 | + * a false positive will be found. To handle this, one extra recursion |
3194 | + * is allowed, and this is done by the TRANSITION bit. If the TRANSITION |
3195 | + * bit is already set, then it is considered a recursion and the function |
3196 | + * ends. Otherwise, the TRANSITION bit is set, and that bit is returned. |
3197 | + * |
3198 | + * On the trace_recursive_unlock(), the TRANSITION bit will be the first |
3199 | + * to be cleared. Even if it wasn't the context that set it. That is, |
3200 | + * if an interrupt comes in while NORMAL bit is set and the ring buffer |
3201 | + * is called before preempt_count() is updated, since the check will |
3202 | + * be on the NORMAL bit, the TRANSITION bit will then be set. If an |
3203 | + * NMI then comes in, it will set the NMI bit, but when the NMI code |
3204 | + * does the trace_recursive_unlock() it will clear the TRANSTION bit |
3205 | + * and leave the NMI bit set. But this is fine, because the interrupt |
3206 | + * code that set the TRANSITION bit will then clear the NMI bit when it |
3207 | + * calls trace_recursive_unlock(). If another NMI comes in, it will |
3208 | + * set the TRANSITION bit and continue. |
3209 | + * |
3210 | + * Note: The TRANSITION bit only handles a single transition between context. |
3211 | */ |
3212 | |
3213 | static __always_inline int |
3214 | @@ -2623,8 +2649,16 @@ trace_recursive_lock(struct ring_buffer_per_cpu *cpu_buffer) |
3215 | } else |
3216 | bit = RB_CTX_NORMAL; |
3217 | |
3218 | - if (unlikely(val & (1 << bit))) |
3219 | - return 1; |
3220 | + if (unlikely(val & (1 << bit))) { |
3221 | + /* |
3222 | + * It is possible that this was called by transitioning |
3223 | + * between interrupt context, and preempt_count() has not |
3224 | + * been updated yet. In this case, use the TRANSITION bit. |
3225 | + */ |
3226 | + bit = RB_CTX_TRANSITION; |
3227 | + if (val & (1 << bit)) |
3228 | + return 1; |
3229 | + } |
3230 | |
3231 | val |= (1 << bit); |
3232 | cpu_buffer->current_context = val; |
3233 | diff --git a/lib/random32.c b/lib/random32.c |
3234 | index d5c3137d93f49..3c5b67b69cbaa 100644 |
3235 | --- a/lib/random32.c |
3236 | +++ b/lib/random32.c |
3237 | @@ -39,16 +39,6 @@ |
3238 | #include <linux/sched.h> |
3239 | #include <asm/unaligned.h> |
3240 | |
3241 | -#ifdef CONFIG_RANDOM32_SELFTEST |
3242 | -static void __init prandom_state_selftest(void); |
3243 | -#else |
3244 | -static inline void prandom_state_selftest(void) |
3245 | -{ |
3246 | -} |
3247 | -#endif |
3248 | - |
3249 | -DEFINE_PER_CPU(struct rnd_state, net_rand_state) __latent_entropy; |
3250 | - |
3251 | /** |
3252 | * prandom_u32_state - seeded pseudo-random number generator. |
3253 | * @state: pointer to state structure holding seeded state. |
3254 | @@ -68,25 +58,6 @@ u32 prandom_u32_state(struct rnd_state *state) |
3255 | } |
3256 | EXPORT_SYMBOL(prandom_u32_state); |
3257 | |
3258 | -/** |
3259 | - * prandom_u32 - pseudo random number generator |
3260 | - * |
3261 | - * A 32 bit pseudo-random number is generated using a fast |
3262 | - * algorithm suitable for simulation. This algorithm is NOT |
3263 | - * considered safe for cryptographic use. |
3264 | - */ |
3265 | -u32 prandom_u32(void) |
3266 | -{ |
3267 | - struct rnd_state *state = &get_cpu_var(net_rand_state); |
3268 | - u32 res; |
3269 | - |
3270 | - res = prandom_u32_state(state); |
3271 | - put_cpu_var(net_rand_state); |
3272 | - |
3273 | - return res; |
3274 | -} |
3275 | -EXPORT_SYMBOL(prandom_u32); |
3276 | - |
3277 | /** |
3278 | * prandom_bytes_state - get the requested number of pseudo-random bytes |
3279 | * |
3280 | @@ -118,20 +89,6 @@ void prandom_bytes_state(struct rnd_state *state, void *buf, size_t bytes) |
3281 | } |
3282 | EXPORT_SYMBOL(prandom_bytes_state); |
3283 | |
3284 | -/** |
3285 | - * prandom_bytes - get the requested number of pseudo-random bytes |
3286 | - * @buf: where to copy the pseudo-random bytes to |
3287 | - * @bytes: the requested number of bytes |
3288 | - */ |
3289 | -void prandom_bytes(void *buf, size_t bytes) |
3290 | -{ |
3291 | - struct rnd_state *state = &get_cpu_var(net_rand_state); |
3292 | - |
3293 | - prandom_bytes_state(state, buf, bytes); |
3294 | - put_cpu_var(net_rand_state); |
3295 | -} |
3296 | -EXPORT_SYMBOL(prandom_bytes); |
3297 | - |
3298 | static void prandom_warmup(struct rnd_state *state) |
3299 | { |
3300 | /* Calling RNG ten times to satisfy recurrence condition */ |
3301 | @@ -147,96 +104,6 @@ static void prandom_warmup(struct rnd_state *state) |
3302 | prandom_u32_state(state); |
3303 | } |
3304 | |
3305 | -static u32 __extract_hwseed(void) |
3306 | -{ |
3307 | - unsigned int val = 0; |
3308 | - |
3309 | - (void)(arch_get_random_seed_int(&val) || |
3310 | - arch_get_random_int(&val)); |
3311 | - |
3312 | - return val; |
3313 | -} |
3314 | - |
3315 | -static void prandom_seed_early(struct rnd_state *state, u32 seed, |
3316 | - bool mix_with_hwseed) |
3317 | -{ |
3318 | -#define LCG(x) ((x) * 69069U) /* super-duper LCG */ |
3319 | -#define HWSEED() (mix_with_hwseed ? __extract_hwseed() : 0) |
3320 | - state->s1 = __seed(HWSEED() ^ LCG(seed), 2U); |
3321 | - state->s2 = __seed(HWSEED() ^ LCG(state->s1), 8U); |
3322 | - state->s3 = __seed(HWSEED() ^ LCG(state->s2), 16U); |
3323 | - state->s4 = __seed(HWSEED() ^ LCG(state->s3), 128U); |
3324 | -} |
3325 | - |
3326 | -/** |
3327 | - * prandom_seed - add entropy to pseudo random number generator |
3328 | - * @seed: seed value |
3329 | - * |
3330 | - * Add some additional seeding to the prandom pool. |
3331 | - */ |
3332 | -void prandom_seed(u32 entropy) |
3333 | -{ |
3334 | - int i; |
3335 | - /* |
3336 | - * No locking on the CPUs, but then somewhat random results are, well, |
3337 | - * expected. |
3338 | - */ |
3339 | - for_each_possible_cpu(i) { |
3340 | - struct rnd_state *state = &per_cpu(net_rand_state, i); |
3341 | - |
3342 | - state->s1 = __seed(state->s1 ^ entropy, 2U); |
3343 | - prandom_warmup(state); |
3344 | - } |
3345 | -} |
3346 | -EXPORT_SYMBOL(prandom_seed); |
3347 | - |
3348 | -/* |
3349 | - * Generate some initially weak seeding values to allow |
3350 | - * to start the prandom_u32() engine. |
3351 | - */ |
3352 | -static int __init prandom_init(void) |
3353 | -{ |
3354 | - int i; |
3355 | - |
3356 | - prandom_state_selftest(); |
3357 | - |
3358 | - for_each_possible_cpu(i) { |
3359 | - struct rnd_state *state = &per_cpu(net_rand_state, i); |
3360 | - u32 weak_seed = (i + jiffies) ^ random_get_entropy(); |
3361 | - |
3362 | - prandom_seed_early(state, weak_seed, true); |
3363 | - prandom_warmup(state); |
3364 | - } |
3365 | - |
3366 | - return 0; |
3367 | -} |
3368 | -core_initcall(prandom_init); |
3369 | - |
3370 | -static void __prandom_timer(unsigned long dontcare); |
3371 | - |
3372 | -static DEFINE_TIMER(seed_timer, __prandom_timer, 0, 0); |
3373 | - |
3374 | -static void __prandom_timer(unsigned long dontcare) |
3375 | -{ |
3376 | - u32 entropy; |
3377 | - unsigned long expires; |
3378 | - |
3379 | - get_random_bytes(&entropy, sizeof(entropy)); |
3380 | - prandom_seed(entropy); |
3381 | - |
3382 | - /* reseed every ~60 seconds, in [40 .. 80) interval with slack */ |
3383 | - expires = 40 + prandom_u32_max(40); |
3384 | - seed_timer.expires = jiffies + msecs_to_jiffies(expires * MSEC_PER_SEC); |
3385 | - |
3386 | - add_timer(&seed_timer); |
3387 | -} |
3388 | - |
3389 | -static void __init __prandom_start_seed_timer(void) |
3390 | -{ |
3391 | - seed_timer.expires = jiffies + msecs_to_jiffies(40 * MSEC_PER_SEC); |
3392 | - add_timer(&seed_timer); |
3393 | -} |
3394 | - |
3395 | void prandom_seed_full_state(struct rnd_state __percpu *pcpu_state) |
3396 | { |
3397 | int i; |
3398 | @@ -256,51 +123,6 @@ void prandom_seed_full_state(struct rnd_state __percpu *pcpu_state) |
3399 | } |
3400 | EXPORT_SYMBOL(prandom_seed_full_state); |
3401 | |
3402 | -/* |
3403 | - * Generate better values after random number generator |
3404 | - * is fully initialized. |
3405 | - */ |
3406 | -static void __prandom_reseed(bool late) |
3407 | -{ |
3408 | - unsigned long flags; |
3409 | - static bool latch = false; |
3410 | - static DEFINE_SPINLOCK(lock); |
3411 | - |
3412 | - /* Asking for random bytes might result in bytes getting |
3413 | - * moved into the nonblocking pool and thus marking it |
3414 | - * as initialized. In this case we would double back into |
3415 | - * this function and attempt to do a late reseed. |
3416 | - * Ignore the pointless attempt to reseed again if we're |
3417 | - * already waiting for bytes when the nonblocking pool |
3418 | - * got initialized. |
3419 | - */ |
3420 | - |
3421 | - /* only allow initial seeding (late == false) once */ |
3422 | - if (!spin_trylock_irqsave(&lock, flags)) |
3423 | - return; |
3424 | - |
3425 | - if (latch && !late) |
3426 | - goto out; |
3427 | - |
3428 | - latch = true; |
3429 | - prandom_seed_full_state(&net_rand_state); |
3430 | -out: |
3431 | - spin_unlock_irqrestore(&lock, flags); |
3432 | -} |
3433 | - |
3434 | -void prandom_reseed_late(void) |
3435 | -{ |
3436 | - __prandom_reseed(true); |
3437 | -} |
3438 | - |
3439 | -static int __init prandom_reseed(void) |
3440 | -{ |
3441 | - __prandom_reseed(false); |
3442 | - __prandom_start_seed_timer(); |
3443 | - return 0; |
3444 | -} |
3445 | -late_initcall(prandom_reseed); |
3446 | - |
3447 | #ifdef CONFIG_RANDOM32_SELFTEST |
3448 | static struct prandom_test1 { |
3449 | u32 seed; |
3450 | @@ -420,7 +242,28 @@ static struct prandom_test2 { |
3451 | { 407983964U, 921U, 728767059U }, |
3452 | }; |
3453 | |
3454 | -static void __init prandom_state_selftest(void) |
3455 | +static u32 __extract_hwseed(void) |
3456 | +{ |
3457 | + unsigned int val = 0; |
3458 | + |
3459 | + (void)(arch_get_random_seed_int(&val) || |
3460 | + arch_get_random_int(&val)); |
3461 | + |
3462 | + return val; |
3463 | +} |
3464 | + |
3465 | +static void prandom_seed_early(struct rnd_state *state, u32 seed, |
3466 | + bool mix_with_hwseed) |
3467 | +{ |
3468 | +#define LCG(x) ((x) * 69069U) /* super-duper LCG */ |
3469 | +#define HWSEED() (mix_with_hwseed ? __extract_hwseed() : 0) |
3470 | + state->s1 = __seed(HWSEED() ^ LCG(seed), 2U); |
3471 | + state->s2 = __seed(HWSEED() ^ LCG(state->s1), 8U); |
3472 | + state->s3 = __seed(HWSEED() ^ LCG(state->s2), 16U); |
3473 | + state->s4 = __seed(HWSEED() ^ LCG(state->s3), 128U); |
3474 | +} |
3475 | + |
3476 | +static int __init prandom_state_selftest(void) |
3477 | { |
3478 | int i, j, errors = 0, runs = 0; |
3479 | bool error = false; |
3480 | @@ -460,5 +303,266 @@ static void __init prandom_state_selftest(void) |
3481 | pr_warn("prandom: %d/%d self tests failed\n", errors, runs); |
3482 | else |
3483 | pr_info("prandom: %d self tests passed\n", runs); |
3484 | + return 0; |
3485 | } |
3486 | +core_initcall(prandom_state_selftest); |
3487 | #endif |
3488 | + |
3489 | +/* |
3490 | + * The prandom_u32() implementation is now completely separate from the |
3491 | + * prandom_state() functions, which are retained (for now) for compatibility. |
3492 | + * |
3493 | + * Because of (ab)use in the networking code for choosing random TCP/UDP port |
3494 | + * numbers, which open DoS possibilities if guessable, we want something |
3495 | + * stronger than a standard PRNG. But the performance requirements of |
3496 | + * the network code do not allow robust crypto for this application. |
3497 | + * |
3498 | + * So this is a homebrew Junior Spaceman implementation, based on the |
3499 | + * lowest-latency trustworthy crypto primitive available, SipHash. |
3500 | + * (The authors of SipHash have not been consulted about this abuse of |
3501 | + * their work.) |
3502 | + * |
3503 | + * Standard SipHash-2-4 uses 2n+4 rounds to hash n words of input to |
3504 | + * one word of output. This abbreviated version uses 2 rounds per word |
3505 | + * of output. |
3506 | + */ |
3507 | + |
3508 | +struct siprand_state { |
3509 | + unsigned long v0; |
3510 | + unsigned long v1; |
3511 | + unsigned long v2; |
3512 | + unsigned long v3; |
3513 | +}; |
3514 | + |
3515 | +static DEFINE_PER_CPU(struct siprand_state, net_rand_state) __latent_entropy; |
3516 | + |
3517 | +/* |
3518 | + * This is the core CPRNG function. As "pseudorandom", this is not used |
3519 | + * for truly valuable things, just intended to be a PITA to guess. |
3520 | + * For maximum speed, we do just two SipHash rounds per word. This is |
3521 | + * the same rate as 4 rounds per 64 bits that SipHash normally uses, |
3522 | + * so hopefully it's reasonably secure. |
3523 | + * |
3524 | + * There are two changes from the official SipHash finalization: |
3525 | + * - We omit some constants XORed with v2 in the SipHash spec as irrelevant; |
3526 | + * they are there only to make the output rounds distinct from the input |
3527 | + * rounds, and this application has no input rounds. |
3528 | + * - Rather than returning v0^v1^v2^v3, return v1+v3. |
3529 | + * If you look at the SipHash round, the last operation on v3 is |
3530 | + * "v3 ^= v0", so "v0 ^ v3" just undoes that, a waste of time. |
3531 | + * Likewise "v1 ^= v2". (The rotate of v2 makes a difference, but |
3532 | + * it still cancels out half of the bits in v2 for no benefit.) |
3533 | + * Second, since the last combining operation was xor, continue the |
3534 | + * pattern of alternating xor/add for a tiny bit of extra non-linearity. |
3535 | + */ |
3536 | +static inline u32 siprand_u32(struct siprand_state *s) |
3537 | +{ |
3538 | + unsigned long v0 = s->v0, v1 = s->v1, v2 = s->v2, v3 = s->v3; |
3539 | + |
3540 | + PRND_SIPROUND(v0, v1, v2, v3); |
3541 | + PRND_SIPROUND(v0, v1, v2, v3); |
3542 | + s->v0 = v0; s->v1 = v1; s->v2 = v2; s->v3 = v3; |
3543 | + return v1 + v3; |
3544 | +} |
3545 | + |
3546 | + |
3547 | +/** |
3548 | + * prandom_u32 - pseudo random number generator |
3549 | + * |
3550 | + * A 32 bit pseudo-random number is generated using a fast |
3551 | + * algorithm suitable for simulation. This algorithm is NOT |
3552 | + * considered safe for cryptographic use. |
3553 | + */ |
3554 | +u32 prandom_u32(void) |
3555 | +{ |
3556 | + struct siprand_state *state = get_cpu_ptr(&net_rand_state); |
3557 | + u32 res = siprand_u32(state); |
3558 | + |
3559 | + put_cpu_ptr(&net_rand_state); |
3560 | + return res; |
3561 | +} |
3562 | +EXPORT_SYMBOL(prandom_u32); |
3563 | + |
3564 | +/** |
3565 | + * prandom_bytes - get the requested number of pseudo-random bytes |
3566 | + * @buf: where to copy the pseudo-random bytes to |
3567 | + * @bytes: the requested number of bytes |
3568 | + */ |
3569 | +void prandom_bytes(void *buf, size_t bytes) |
3570 | +{ |
3571 | + struct siprand_state *state = get_cpu_ptr(&net_rand_state); |
3572 | + u8 *ptr = buf; |
3573 | + |
3574 | + while (bytes >= sizeof(u32)) { |
3575 | + put_unaligned(siprand_u32(state), (u32 *)ptr); |
3576 | + ptr += sizeof(u32); |
3577 | + bytes -= sizeof(u32); |
3578 | + } |
3579 | + |
3580 | + if (bytes > 0) { |
3581 | + u32 rem = siprand_u32(state); |
3582 | + |
3583 | + do { |
3584 | + *ptr++ = (u8)rem; |
3585 | + rem >>= BITS_PER_BYTE; |
3586 | + } while (--bytes > 0); |
3587 | + } |
3588 | + put_cpu_ptr(&net_rand_state); |
3589 | +} |
3590 | +EXPORT_SYMBOL(prandom_bytes); |
3591 | + |
3592 | +/** |
3593 | + * prandom_seed - add entropy to pseudo random number generator |
3594 | + * @entropy: entropy value |
3595 | + * |
3596 | + * Add some additional seed material to the prandom pool. |
3597 | + * The "entropy" is actually our IP address (the only caller is |
3598 | + * the network code), not for unpredictability, but to ensure that |
3599 | + * different machines are initialized differently. |
3600 | + */ |
3601 | +void prandom_seed(u32 entropy) |
3602 | +{ |
3603 | + int i; |
3604 | + |
3605 | + add_device_randomness(&entropy, sizeof(entropy)); |
3606 | + |
3607 | + for_each_possible_cpu(i) { |
3608 | + struct siprand_state *state = per_cpu_ptr(&net_rand_state, i); |
3609 | + unsigned long v0 = state->v0, v1 = state->v1; |
3610 | + unsigned long v2 = state->v2, v3 = state->v3; |
3611 | + |
3612 | + do { |
3613 | + v3 ^= entropy; |
3614 | + PRND_SIPROUND(v0, v1, v2, v3); |
3615 | + PRND_SIPROUND(v0, v1, v2, v3); |
3616 | + v0 ^= entropy; |
3617 | + } while (unlikely(!v0 || !v1 || !v2 || !v3)); |
3618 | + |
3619 | + WRITE_ONCE(state->v0, v0); |
3620 | + WRITE_ONCE(state->v1, v1); |
3621 | + WRITE_ONCE(state->v2, v2); |
3622 | + WRITE_ONCE(state->v3, v3); |
3623 | + } |
3624 | +} |
3625 | +EXPORT_SYMBOL(prandom_seed); |
3626 | + |
3627 | +/* |
3628 | + * Generate some initially weak seeding values to allow |
3629 | + * the prandom_u32() engine to be started. |
3630 | + */ |
3631 | +static int __init prandom_init_early(void) |
3632 | +{ |
3633 | + int i; |
3634 | + unsigned long v0, v1, v2, v3; |
3635 | + |
3636 | + if (!arch_get_random_long(&v0)) |
3637 | + v0 = jiffies; |
3638 | + if (!arch_get_random_long(&v1)) |
3639 | + v1 = random_get_entropy(); |
3640 | + v2 = v0 ^ PRND_K0; |
3641 | + v3 = v1 ^ PRND_K1; |
3642 | + |
3643 | + for_each_possible_cpu(i) { |
3644 | + struct siprand_state *state; |
3645 | + |
3646 | + v3 ^= i; |
3647 | + PRND_SIPROUND(v0, v1, v2, v3); |
3648 | + PRND_SIPROUND(v0, v1, v2, v3); |
3649 | + v0 ^= i; |
3650 | + |
3651 | + state = per_cpu_ptr(&net_rand_state, i); |
3652 | + state->v0 = v0; state->v1 = v1; |
3653 | + state->v2 = v2; state->v3 = v3; |
3654 | + } |
3655 | + |
3656 | + return 0; |
3657 | +} |
3658 | +core_initcall(prandom_init_early); |
3659 | + |
3660 | + |
3661 | +/* Stronger reseeding when available, and periodically thereafter. */ |
3662 | +static void prandom_reseed(unsigned long dontcare); |
3663 | + |
3664 | +static DEFINE_TIMER(seed_timer, prandom_reseed, 0, 0); |
3665 | + |
3666 | +static void prandom_reseed(unsigned long dontcare) |
3667 | +{ |
3668 | + unsigned long expires; |
3669 | + int i; |
3670 | + |
3671 | + /* |
3672 | + * Reinitialize each CPU's PRNG with 128 bits of key. |
3673 | + * No locking on the CPUs, but then somewhat random results are, |
3674 | + * well, expected. |
3675 | + */ |
3676 | + for_each_possible_cpu(i) { |
3677 | + struct siprand_state *state; |
3678 | + unsigned long v0 = get_random_long(), v2 = v0 ^ PRND_K0; |
3679 | + unsigned long v1 = get_random_long(), v3 = v1 ^ PRND_K1; |
3680 | +#if BITS_PER_LONG == 32 |
3681 | + int j; |
3682 | + |
3683 | + /* |
3684 | + * On 32-bit machines, hash in two extra words to |
3685 | + * approximate 128-bit key length. Not that the hash |
3686 | + * has that much security, but this prevents a trivial |
3687 | + * 64-bit brute force. |
3688 | + */ |
3689 | + for (j = 0; j < 2; j++) { |
3690 | + unsigned long m = get_random_long(); |
3691 | + |
3692 | + v3 ^= m; |
3693 | + PRND_SIPROUND(v0, v1, v2, v3); |
3694 | + PRND_SIPROUND(v0, v1, v2, v3); |
3695 | + v0 ^= m; |
3696 | + } |
3697 | +#endif |
3698 | + /* |
3699 | + * Probably impossible in practice, but there is a |
3700 | + * theoretical risk that a race between this reseeding |
3701 | + * and the target CPU writing its state back could |
3702 | + * create the all-zero SipHash fixed point. |
3703 | + * |
3704 | + * To ensure that never happens, ensure the state |
3705 | + * we write contains no zero words. |
3706 | + */ |
3707 | + state = per_cpu_ptr(&net_rand_state, i); |
3708 | + WRITE_ONCE(state->v0, v0 ? v0 : -1ul); |
3709 | + WRITE_ONCE(state->v1, v1 ? v1 : -1ul); |
3710 | + WRITE_ONCE(state->v2, v2 ? v2 : -1ul); |
3711 | + WRITE_ONCE(state->v3, v3 ? v3 : -1ul); |
3712 | + } |
3713 | + |
3714 | + /* reseed every ~60 seconds, in [40 .. 80) interval with slack */ |
3715 | + expires = round_jiffies(jiffies + 40 * HZ + prandom_u32_max(40 * HZ)); |
3716 | + mod_timer(&seed_timer, expires); |
3717 | +} |
3718 | + |
3719 | +/* |
3720 | + * The random ready callback can be called from almost any interrupt. |
3721 | + * To avoid worrying about whether it's safe to delay that interrupt |
3722 | + * long enough to seed all CPUs, just schedule an immediate timer event. |
3723 | + */ |
3724 | +static void prandom_timer_start(struct random_ready_callback *unused) |
3725 | +{ |
3726 | + mod_timer(&seed_timer, jiffies); |
3727 | +} |
3728 | + |
3729 | +/* |
3730 | + * Start periodic full reseeding as soon as strong |
3731 | + * random numbers are available. |
3732 | + */ |
3733 | +static int __init prandom_init_late(void) |
3734 | +{ |
3735 | + static struct random_ready_callback random_ready = { |
3736 | + .func = prandom_timer_start |
3737 | + }; |
3738 | + int ret = add_random_ready_callback(&random_ready); |
3739 | + |
3740 | + if (ret == -EALREADY) { |
3741 | + prandom_timer_start(&random_ready); |
3742 | + ret = 0; |
3743 | + } |
3744 | + return ret; |
3745 | +} |
3746 | +late_initcall(prandom_init_late); |
3747 | diff --git a/lib/swiotlb.c b/lib/swiotlb.c |
3748 | index 7ff9dc36c2f80..74b5b88621989 100644 |
3749 | --- a/lib/swiotlb.c |
3750 | +++ b/lib/swiotlb.c |
3751 | @@ -199,6 +199,7 @@ int __init swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose) |
3752 | io_tlb_orig_addr[i] = INVALID_PHYS_ADDR; |
3753 | } |
3754 | io_tlb_index = 0; |
3755 | + no_iotlb_memory = false; |
3756 | |
3757 | if (verbose) |
3758 | swiotlb_print_info(); |
3759 | @@ -229,9 +230,11 @@ swiotlb_init(int verbose) |
3760 | if (vstart && !swiotlb_init_with_tbl(vstart, io_tlb_nslabs, verbose)) |
3761 | return; |
3762 | |
3763 | - if (io_tlb_start) |
3764 | + if (io_tlb_start) { |
3765 | memblock_free_early(io_tlb_start, |
3766 | PAGE_ALIGN(io_tlb_nslabs << IO_TLB_SHIFT)); |
3767 | + io_tlb_start = 0; |
3768 | + } |
3769 | pr_warn("Cannot allocate buffer"); |
3770 | no_iotlb_memory = true; |
3771 | } |
3772 | @@ -330,6 +333,7 @@ swiotlb_late_init_with_tbl(char *tlb, unsigned long nslabs) |
3773 | io_tlb_orig_addr[i] = INVALID_PHYS_ADDR; |
3774 | } |
3775 | io_tlb_index = 0; |
3776 | + no_iotlb_memory = false; |
3777 | |
3778 | swiotlb_print_info(); |
3779 | |
3780 | diff --git a/mm/mempolicy.c b/mm/mempolicy.c |
3781 | index a2be65bf5d8cc..2f443767fd1b4 100644 |
3782 | --- a/mm/mempolicy.c |
3783 | +++ b/mm/mempolicy.c |
3784 | @@ -487,7 +487,7 @@ static int queue_pages_pte_range(pmd_t *pmd, unsigned long addr, |
3785 | struct queue_pages *qp = walk->private; |
3786 | unsigned long flags = qp->flags; |
3787 | int nid, ret; |
3788 | - pte_t *pte; |
3789 | + pte_t *pte, *mapped_pte; |
3790 | spinlock_t *ptl; |
3791 | |
3792 | if (pmd_trans_huge(*pmd)) { |
3793 | @@ -515,7 +515,7 @@ static int queue_pages_pte_range(pmd_t *pmd, unsigned long addr, |
3794 | if (pmd_trans_unstable(pmd)) |
3795 | return 0; |
3796 | retry: |
3797 | - pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
3798 | + mapped_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl); |
3799 | for (; addr != end; pte++, addr += PAGE_SIZE) { |
3800 | if (!pte_present(*pte)) |
3801 | continue; |
3802 | @@ -554,7 +554,7 @@ retry: |
3803 | } else |
3804 | break; |
3805 | } |
3806 | - pte_unmap_unlock(pte - 1, ptl); |
3807 | + pte_unmap_unlock(mapped_pte, ptl); |
3808 | cond_resched(); |
3809 | return addr != end ? -EIO : 0; |
3810 | } |
3811 | diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c |
3812 | index b596c413d297c..bcf743432c8de 100644 |
3813 | --- a/net/ipv4/syncookies.c |
3814 | +++ b/net/ipv4/syncookies.c |
3815 | @@ -304,7 +304,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb) |
3816 | __u32 cookie = ntohl(th->ack_seq) - 1; |
3817 | struct sock *ret = sk; |
3818 | struct request_sock *req; |
3819 | - int mss; |
3820 | + int full_space, mss; |
3821 | struct rtable *rt; |
3822 | __u8 rcv_wscale; |
3823 | struct flowi4 fl4; |
3824 | @@ -388,8 +388,13 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb) |
3825 | |
3826 | /* Try to redo what tcp_v4_send_synack did. */ |
3827 | req->rsk_window_clamp = tp->window_clamp ? :dst_metric(&rt->dst, RTAX_WINDOW); |
3828 | + /* limit the window selection if the user enforce a smaller rx buffer */ |
3829 | + full_space = tcp_full_space(sk); |
3830 | + if (sk->sk_userlocks & SOCK_RCVBUF_LOCK && |
3831 | + (req->rsk_window_clamp > full_space || req->rsk_window_clamp == 0)) |
3832 | + req->rsk_window_clamp = full_space; |
3833 | |
3834 | - tcp_select_initial_window(tcp_full_space(sk), req->mss, |
3835 | + tcp_select_initial_window(full_space, req->mss, |
3836 | &req->rsk_rcv_wnd, &req->rsk_window_clamp, |
3837 | ireq->wscale_ok, &rcv_wscale, |
3838 | dst_metric(&rt->dst, RTAX_INITRWND)); |
3839 | diff --git a/net/ipv6/sit.c b/net/ipv6/sit.c |
3840 | index 16eba7b5f1a9d..df705303992e0 100644 |
3841 | --- a/net/ipv6/sit.c |
3842 | +++ b/net/ipv6/sit.c |
3843 | @@ -1072,7 +1072,6 @@ static void ipip6_tunnel_bind_dev(struct net_device *dev) |
3844 | if (tdev && !netif_is_l3_master(tdev)) { |
3845 | int t_hlen = tunnel->hlen + sizeof(struct iphdr); |
3846 | |
3847 | - dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr); |
3848 | dev->mtu = tdev->mtu - t_hlen; |
3849 | if (dev->mtu < IPV6_MIN_MTU) |
3850 | dev->mtu = IPV6_MIN_MTU; |
3851 | @@ -1372,7 +1371,6 @@ static void ipip6_tunnel_setup(struct net_device *dev) |
3852 | dev->destructor = ipip6_dev_free; |
3853 | |
3854 | dev->type = ARPHRD_SIT; |
3855 | - dev->hard_header_len = LL_MAX_HEADER + t_hlen; |
3856 | dev->mtu = ETH_DATA_LEN - t_hlen; |
3857 | dev->flags = IFF_NOARP; |
3858 | netif_keep_dst(dev); |
3859 | diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c |
3860 | index 4834015b27f49..0e368c1dd78b8 100644 |
3861 | --- a/net/ipv6/syncookies.c |
3862 | +++ b/net/ipv6/syncookies.c |
3863 | @@ -143,7 +143,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb) |
3864 | __u32 cookie = ntohl(th->ack_seq) - 1; |
3865 | struct sock *ret = sk; |
3866 | struct request_sock *req; |
3867 | - int mss; |
3868 | + int full_space, mss; |
3869 | struct dst_entry *dst; |
3870 | __u8 rcv_wscale; |
3871 | |
3872 | @@ -236,7 +236,13 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb) |
3873 | } |
3874 | |
3875 | req->rsk_window_clamp = tp->window_clamp ? :dst_metric(dst, RTAX_WINDOW); |
3876 | - tcp_select_initial_window(tcp_full_space(sk), req->mss, |
3877 | + /* limit the window selection if the user enforce a smaller rx buffer */ |
3878 | + full_space = tcp_full_space(sk); |
3879 | + if (sk->sk_userlocks & SOCK_RCVBUF_LOCK && |
3880 | + (req->rsk_window_clamp > full_space || req->rsk_window_clamp == 0)) |
3881 | + req->rsk_window_clamp = full_space; |
3882 | + |
3883 | + tcp_select_initial_window(full_space, req->mss, |
3884 | &req->rsk_rcv_wnd, &req->rsk_window_clamp, |
3885 | ireq->wscale_ok, &rcv_wscale, |
3886 | dst_metric(dst, RTAX_INITRWND)); |
3887 | diff --git a/net/iucv/af_iucv.c b/net/iucv/af_iucv.c |
3888 | index 02e10deef5b45..ba81655a584bd 100644 |
3889 | --- a/net/iucv/af_iucv.c |
3890 | +++ b/net/iucv/af_iucv.c |
3891 | @@ -1542,7 +1542,8 @@ static int iucv_sock_shutdown(struct socket *sock, int how) |
3892 | break; |
3893 | } |
3894 | |
3895 | - if (how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) { |
3896 | + if ((how == SEND_SHUTDOWN || how == SHUTDOWN_MASK) && |
3897 | + sk->sk_state == IUCV_CONNECTED) { |
3898 | if (iucv->transport == AF_IUCV_TRANS_IUCV) { |
3899 | txmsg.class = 0; |
3900 | txmsg.tag = 0; |
3901 | diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c |
3902 | index 6216279efc468..eebbddccb47b7 100644 |
3903 | --- a/net/mac80211/tx.c |
3904 | +++ b/net/mac80211/tx.c |
3905 | @@ -1847,19 +1847,24 @@ static bool ieee80211_tx(struct ieee80211_sub_if_data *sdata, |
3906 | |
3907 | /* device xmit handlers */ |
3908 | |
3909 | +enum ieee80211_encrypt { |
3910 | + ENCRYPT_NO, |
3911 | + ENCRYPT_MGMT, |
3912 | + ENCRYPT_DATA, |
3913 | +}; |
3914 | + |
3915 | static int ieee80211_skb_resize(struct ieee80211_sub_if_data *sdata, |
3916 | struct sk_buff *skb, |
3917 | - int head_need, bool may_encrypt) |
3918 | + int head_need, |
3919 | + enum ieee80211_encrypt encrypt) |
3920 | { |
3921 | struct ieee80211_local *local = sdata->local; |
3922 | - struct ieee80211_hdr *hdr; |
3923 | bool enc_tailroom; |
3924 | int tail_need = 0; |
3925 | |
3926 | - hdr = (struct ieee80211_hdr *) skb->data; |
3927 | - enc_tailroom = may_encrypt && |
3928 | - (sdata->crypto_tx_tailroom_needed_cnt || |
3929 | - ieee80211_is_mgmt(hdr->frame_control)); |
3930 | + enc_tailroom = encrypt == ENCRYPT_MGMT || |
3931 | + (encrypt == ENCRYPT_DATA && |
3932 | + sdata->crypto_tx_tailroom_needed_cnt); |
3933 | |
3934 | if (enc_tailroom) { |
3935 | tail_need = IEEE80211_ENCRYPT_TAILROOM; |
3936 | @@ -1892,21 +1897,27 @@ void ieee80211_xmit(struct ieee80211_sub_if_data *sdata, |
3937 | struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); |
3938 | struct ieee80211_hdr *hdr = (struct ieee80211_hdr *) skb->data; |
3939 | int headroom; |
3940 | - bool may_encrypt; |
3941 | + enum ieee80211_encrypt encrypt; |
3942 | |
3943 | - may_encrypt = !(info->flags & IEEE80211_TX_INTFL_DONT_ENCRYPT); |
3944 | + if (info->flags & IEEE80211_TX_INTFL_DONT_ENCRYPT) |
3945 | + encrypt = ENCRYPT_NO; |
3946 | + else if (ieee80211_is_mgmt(hdr->frame_control)) |
3947 | + encrypt = ENCRYPT_MGMT; |
3948 | + else |
3949 | + encrypt = ENCRYPT_DATA; |
3950 | |
3951 | headroom = local->tx_headroom; |
3952 | - if (may_encrypt) |
3953 | + if (encrypt != ENCRYPT_NO) |
3954 | headroom += sdata->encrypt_headroom; |
3955 | headroom -= skb_headroom(skb); |
3956 | headroom = max_t(int, 0, headroom); |
3957 | |
3958 | - if (ieee80211_skb_resize(sdata, skb, headroom, may_encrypt)) { |
3959 | + if (ieee80211_skb_resize(sdata, skb, headroom, encrypt)) { |
3960 | ieee80211_free_txskb(&local->hw, skb); |
3961 | return; |
3962 | } |
3963 | |
3964 | + /* reload after potential resize */ |
3965 | hdr = (struct ieee80211_hdr *) skb->data; |
3966 | info->control.vif = &sdata->vif; |
3967 | |
3968 | @@ -2688,7 +2699,7 @@ static struct sk_buff *ieee80211_build_hdr(struct ieee80211_sub_if_data *sdata, |
3969 | head_need += sdata->encrypt_headroom; |
3970 | head_need += local->tx_headroom; |
3971 | head_need = max_t(int, 0, head_need); |
3972 | - if (ieee80211_skb_resize(sdata, skb, head_need, true)) { |
3973 | + if (ieee80211_skb_resize(sdata, skb, head_need, ENCRYPT_DATA)) { |
3974 | ieee80211_free_txskb(&local->hw, skb); |
3975 | skb = NULL; |
3976 | return ERR_PTR(-ENOMEM); |
3977 | @@ -3313,7 +3324,7 @@ static bool ieee80211_xmit_fast(struct ieee80211_sub_if_data *sdata, |
3978 | if (unlikely(ieee80211_skb_resize(sdata, skb, |
3979 | max_t(int, extra_head + hw_headroom - |
3980 | skb_headroom(skb), 0), |
3981 | - false))) { |
3982 | + ENCRYPT_NO))) { |
3983 | kfree_skb(skb); |
3984 | return true; |
3985 | } |
3986 | diff --git a/net/wireless/reg.c b/net/wireless/reg.c |
3987 | index a649763b854d5..04da31c52d092 100644 |
3988 | --- a/net/wireless/reg.c |
3989 | +++ b/net/wireless/reg.c |
3990 | @@ -2759,7 +2759,7 @@ static void print_rd_rules(const struct ieee80211_regdomain *rd) |
3991 | power_rule = ®_rule->power_rule; |
3992 | |
3993 | if (reg_rule->flags & NL80211_RRF_AUTO_BW) |
3994 | - snprintf(bw, sizeof(bw), "%d KHz, %d KHz AUTO", |
3995 | + snprintf(bw, sizeof(bw), "%d KHz, %u KHz AUTO", |
3996 | freq_range->max_bandwidth_khz, |
3997 | reg_get_max_bandwidth(rd, reg_rule)); |
3998 | else |
3999 | diff --git a/net/x25/af_x25.c b/net/x25/af_x25.c |
4000 | index a8ca79810dcc6..3bec5c59169b5 100644 |
4001 | --- a/net/x25/af_x25.c |
4002 | +++ b/net/x25/af_x25.c |
4003 | @@ -823,7 +823,7 @@ static int x25_connect(struct socket *sock, struct sockaddr *uaddr, |
4004 | sock->state = SS_CONNECTED; |
4005 | rc = 0; |
4006 | out_put_neigh: |
4007 | - if (rc) { |
4008 | + if (rc && x25->neighbour) { |
4009 | read_lock_bh(&x25_list_lock); |
4010 | x25_neigh_put(x25->neighbour); |
4011 | x25->neighbour = NULL; |
4012 | diff --git a/net/xfrm/xfrm_state.c b/net/xfrm/xfrm_state.c |
4013 | index 0eb85765d35a1..4d19f2ff6e052 100644 |
4014 | --- a/net/xfrm/xfrm_state.c |
4015 | +++ b/net/xfrm/xfrm_state.c |
4016 | @@ -1591,6 +1591,7 @@ int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high) |
4017 | int err = -ENOENT; |
4018 | __be32 minspi = htonl(low); |
4019 | __be32 maxspi = htonl(high); |
4020 | + __be32 newspi = 0; |
4021 | u32 mark = x->mark.v & x->mark.m; |
4022 | |
4023 | spin_lock_bh(&x->lock); |
4024 | @@ -1609,21 +1610,22 @@ int xfrm_alloc_spi(struct xfrm_state *x, u32 low, u32 high) |
4025 | xfrm_state_put(x0); |
4026 | goto unlock; |
4027 | } |
4028 | - x->id.spi = minspi; |
4029 | + newspi = minspi; |
4030 | } else { |
4031 | u32 spi = 0; |
4032 | for (h = 0; h < high-low+1; h++) { |
4033 | spi = low + prandom_u32()%(high-low+1); |
4034 | x0 = xfrm_state_lookup(net, mark, &x->id.daddr, htonl(spi), x->id.proto, x->props.family); |
4035 | if (x0 == NULL) { |
4036 | - x->id.spi = htonl(spi); |
4037 | + newspi = htonl(spi); |
4038 | break; |
4039 | } |
4040 | xfrm_state_put(x0); |
4041 | } |
4042 | } |
4043 | - if (x->id.spi) { |
4044 | + if (newspi) { |
4045 | spin_lock_bh(&net->xfrm.xfrm_state_lock); |
4046 | + x->id.spi = newspi; |
4047 | h = xfrm_spi_hash(net, &x->id.daddr, x->id.spi, x->id.proto, x->props.family); |
4048 | hlist_add_head_rcu(&x->byspi, net->xfrm.state_byspi + h); |
4049 | spin_unlock_bh(&net->xfrm.xfrm_state_lock); |
4050 | diff --git a/sound/hda/ext/hdac_ext_controller.c b/sound/hda/ext/hdac_ext_controller.c |
4051 | index 261469188566c..49d42971d90da 100644 |
4052 | --- a/sound/hda/ext/hdac_ext_controller.c |
4053 | +++ b/sound/hda/ext/hdac_ext_controller.c |
4054 | @@ -155,6 +155,8 @@ struct hdac_ext_link *snd_hdac_ext_bus_get_link(struct hdac_ext_bus *ebus, |
4055 | return NULL; |
4056 | if (ebus->idx != bus_idx) |
4057 | return NULL; |
4058 | + if (addr < 0 || addr > 31) |
4059 | + return NULL; |
4060 | |
4061 | list_for_each_entry(hlink, &ebus->hlink_list, list) { |
4062 | for (i = 0; i < HDA_MAX_CODECS; i++) { |
4063 | diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c |
4064 | index 7e0573e55a356..89808ab008ad2 100644 |
4065 | --- a/tools/perf/util/session.c |
4066 | +++ b/tools/perf/util/session.c |
4067 | @@ -482,6 +482,7 @@ static void perf_event__mmap2_swap(union perf_event *event, |
4068 | event->mmap2.maj = bswap_32(event->mmap2.maj); |
4069 | event->mmap2.min = bswap_32(event->mmap2.min); |
4070 | event->mmap2.ino = bswap_64(event->mmap2.ino); |
4071 | + event->mmap2.ino_generation = bswap_64(event->mmap2.ino_generation); |
4072 | |
4073 | if (sample_id_all) { |
4074 | void *data = &event->mmap2.filename; |