]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * linux/mm/oom_kill.c | |
3 | * | |
4 | * Copyright (C) 1998,2000 Rik van Riel | |
5 | * Thanks go out to Claus Fischer for some serious inspiration and | |
6 | * for goading me into coding this file... | |
7 | * Copyright (C) 2010 Google, Inc. | |
8 | * Rewritten by David Rientjes | |
9 | * | |
10 | * The routines in this file are used to kill a process when | |
11 | * we're seriously out of memory. This gets called from __alloc_pages() | |
12 | * in mm/page_alloc.c when we really run out of memory. | |
13 | * | |
14 | * Since we won't call these routines often (on a well-configured | |
15 | * machine) this file will double as a 'coding guide' and a signpost | |
16 | * for newbie kernel hackers. It features several pointers to major | |
17 | * kernel subsystems and hints as to where to find out what things do. | |
18 | */ | |
19 | ||
20 | #include <linux/oom.h> | |
21 | #include <linux/mm.h> | |
22 | #include <linux/err.h> | |
23 | #include <linux/gfp.h> | |
24 | #include <linux/sched.h> | |
25 | #include <linux/swap.h> | |
26 | #include <linux/timex.h> | |
27 | #include <linux/jiffies.h> | |
28 | #include <linux/cpuset.h> | |
29 | #include <linux/export.h> | |
30 | #include <linux/notifier.h> | |
31 | #include <linux/memcontrol.h> | |
32 | #include <linux/mempolicy.h> | |
33 | #include <linux/security.h> | |
34 | #include <linux/ptrace.h> | |
35 | #include <linux/freezer.h> | |
36 | #include <linux/ftrace.h> | |
37 | #include <linux/ratelimit.h> | |
38 | ||
39 | #define CREATE_TRACE_POINTS | |
40 | #include <trace/events/oom.h> | |
41 | ||
42 | int sysctl_panic_on_oom; | |
43 | int sysctl_oom_kill_allocating_task; | |
44 | int sysctl_oom_dump_tasks = 1; | |
45 | ||
46 | DEFINE_MUTEX(oom_lock); | |
47 | ||
48 | #ifdef CONFIG_NUMA | |
49 | /** | |
50 | * has_intersects_mems_allowed() - check task eligiblity for kill | |
51 | * @start: task struct of which task to consider | |
52 | * @mask: nodemask passed to page allocator for mempolicy ooms | |
53 | * | |
54 | * Task eligibility is determined by whether or not a candidate task, @tsk, | |
55 | * shares the same mempolicy nodes as current if it is bound by such a policy | |
56 | * and whether or not it has the same set of allowed cpuset nodes. | |
57 | */ | |
58 | static bool has_intersects_mems_allowed(struct task_struct *start, | |
59 | const nodemask_t *mask) | |
60 | { | |
61 | struct task_struct *tsk; | |
62 | bool ret = false; | |
63 | ||
64 | rcu_read_lock(); | |
65 | for_each_thread(start, tsk) { | |
66 | if (mask) { | |
67 | /* | |
68 | * If this is a mempolicy constrained oom, tsk's | |
69 | * cpuset is irrelevant. Only return true if its | |
70 | * mempolicy intersects current, otherwise it may be | |
71 | * needlessly killed. | |
72 | */ | |
73 | ret = mempolicy_nodemask_intersects(tsk, mask); | |
74 | } else { | |
75 | /* | |
76 | * This is not a mempolicy constrained oom, so only | |
77 | * check the mems of tsk's cpuset. | |
78 | */ | |
79 | ret = cpuset_mems_allowed_intersects(current, tsk); | |
80 | } | |
81 | if (ret) | |
82 | break; | |
83 | } | |
84 | rcu_read_unlock(); | |
85 | ||
86 | return ret; | |
87 | } | |
88 | #else | |
89 | static bool has_intersects_mems_allowed(struct task_struct *tsk, | |
90 | const nodemask_t *mask) | |
91 | { | |
92 | return true; | |
93 | } | |
94 | #endif /* CONFIG_NUMA */ | |
95 | ||
96 | /* | |
97 | * The process p may have detached its own ->mm while exiting or through | |
98 | * use_mm(), but one or more of its subthreads may still have a valid | |
99 | * pointer. Return p, or any of its subthreads with a valid ->mm, with | |
100 | * task_lock() held. | |
101 | */ | |
102 | struct task_struct *find_lock_task_mm(struct task_struct *p) | |
103 | { | |
104 | struct task_struct *t; | |
105 | ||
106 | rcu_read_lock(); | |
107 | ||
108 | for_each_thread(p, t) { | |
109 | task_lock(t); | |
110 | if (likely(t->mm)) | |
111 | goto found; | |
112 | task_unlock(t); | |
113 | } | |
114 | t = NULL; | |
115 | found: | |
116 | rcu_read_unlock(); | |
117 | ||
118 | return t; | |
119 | } | |
120 | ||
121 | /* return true if the task is not adequate as candidate victim task. */ | |
122 | static bool oom_unkillable_task(struct task_struct *p, | |
123 | struct mem_cgroup *memcg, const nodemask_t *nodemask) | |
124 | { | |
125 | if (is_global_init(p)) | |
126 | return true; | |
127 | if (p->flags & PF_KTHREAD) | |
128 | return true; | |
129 | ||
130 | /* When mem_cgroup_out_of_memory() and p is not member of the group */ | |
131 | if (memcg && !task_in_mem_cgroup(p, memcg)) | |
132 | return true; | |
133 | ||
134 | /* p may not have freeable memory in nodemask */ | |
135 | if (!has_intersects_mems_allowed(p, nodemask)) | |
136 | return true; | |
137 | ||
138 | return false; | |
139 | } | |
140 | ||
141 | /** | |
142 | * oom_badness - heuristic function to determine which candidate task to kill | |
143 | * @p: task struct of which task we should calculate | |
144 | * @totalpages: total present RAM allowed for page allocation | |
145 | * | |
146 | * The heuristic for determining which task to kill is made to be as simple and | |
147 | * predictable as possible. The goal is to return the highest value for the | |
148 | * task consuming the most memory to avoid subsequent oom failures. | |
149 | */ | |
150 | unsigned long oom_badness(struct task_struct *p, struct mem_cgroup *memcg, | |
151 | const nodemask_t *nodemask, unsigned long totalpages) | |
152 | { | |
153 | long points; | |
154 | long adj; | |
155 | ||
156 | if (oom_unkillable_task(p, memcg, nodemask)) | |
157 | return 0; | |
158 | ||
159 | p = find_lock_task_mm(p); | |
160 | if (!p) | |
161 | return 0; | |
162 | ||
163 | adj = (long)p->signal->oom_score_adj; | |
164 | if (adj == OOM_SCORE_ADJ_MIN) { | |
165 | task_unlock(p); | |
166 | return 0; | |
167 | } | |
168 | ||
169 | /* | |
170 | * The baseline for the badness score is the proportion of RAM that each | |
171 | * task's rss, pagetable and swap space use. | |
172 | */ | |
173 | points = get_mm_rss(p->mm) + get_mm_counter(p->mm, MM_SWAPENTS) + | |
174 | atomic_long_read(&p->mm->nr_ptes) + mm_nr_pmds(p->mm); | |
175 | task_unlock(p); | |
176 | ||
177 | /* | |
178 | * Root processes get 3% bonus, just like the __vm_enough_memory() | |
179 | * implementation used by LSMs. | |
180 | */ | |
181 | if (has_capability_noaudit(p, CAP_SYS_ADMIN)) | |
182 | points -= (points * 3) / 100; | |
183 | ||
184 | /* Normalize to oom_score_adj units */ | |
185 | adj *= totalpages / 1000; | |
186 | points += adj; | |
187 | ||
188 | /* | |
189 | * Never return 0 for an eligible task regardless of the root bonus and | |
190 | * oom_score_adj (oom_score_adj can't be OOM_SCORE_ADJ_MIN here). | |
191 | */ | |
192 | return points > 0 ? points : 1; | |
193 | } | |
194 | ||
195 | /* | |
196 | * Determine the type of allocation constraint. | |
197 | */ | |
198 | #ifdef CONFIG_NUMA | |
199 | static enum oom_constraint constrained_alloc(struct oom_control *oc, | |
200 | unsigned long *totalpages) | |
201 | { | |
202 | struct zone *zone; | |
203 | struct zoneref *z; | |
204 | enum zone_type high_zoneidx = gfp_zone(oc->gfp_mask); | |
205 | bool cpuset_limited = false; | |
206 | int nid; | |
207 | ||
208 | /* Default to all available memory */ | |
209 | *totalpages = totalram_pages + total_swap_pages; | |
210 | ||
211 | if (!oc->zonelist) | |
212 | return CONSTRAINT_NONE; | |
213 | /* | |
214 | * Reach here only when __GFP_NOFAIL is used. So, we should avoid | |
215 | * to kill current.We have to random task kill in this case. | |
216 | * Hopefully, CONSTRAINT_THISNODE...but no way to handle it, now. | |
217 | */ | |
218 | if (oc->gfp_mask & __GFP_THISNODE) | |
219 | return CONSTRAINT_NONE; | |
220 | ||
221 | /* | |
222 | * This is not a __GFP_THISNODE allocation, so a truncated nodemask in | |
223 | * the page allocator means a mempolicy is in effect. Cpuset policy | |
224 | * is enforced in get_page_from_freelist(). | |
225 | */ | |
226 | if (oc->nodemask && | |
227 | !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) { | |
228 | *totalpages = total_swap_pages; | |
229 | for_each_node_mask(nid, *oc->nodemask) | |
230 | *totalpages += node_spanned_pages(nid); | |
231 | return CONSTRAINT_MEMORY_POLICY; | |
232 | } | |
233 | ||
234 | /* Check this allocation failure is caused by cpuset's wall function */ | |
235 | for_each_zone_zonelist_nodemask(zone, z, oc->zonelist, | |
236 | high_zoneidx, oc->nodemask) | |
237 | if (!cpuset_zone_allowed(zone, oc->gfp_mask)) | |
238 | cpuset_limited = true; | |
239 | ||
240 | if (cpuset_limited) { | |
241 | *totalpages = total_swap_pages; | |
242 | for_each_node_mask(nid, cpuset_current_mems_allowed) | |
243 | *totalpages += node_spanned_pages(nid); | |
244 | return CONSTRAINT_CPUSET; | |
245 | } | |
246 | return CONSTRAINT_NONE; | |
247 | } | |
248 | #else | |
249 | static enum oom_constraint constrained_alloc(struct oom_control *oc, | |
250 | unsigned long *totalpages) | |
251 | { | |
252 | *totalpages = totalram_pages + total_swap_pages; | |
253 | return CONSTRAINT_NONE; | |
254 | } | |
255 | #endif | |
256 | ||
257 | enum oom_scan_t oom_scan_process_thread(struct oom_control *oc, | |
258 | struct task_struct *task, unsigned long totalpages) | |
259 | { | |
260 | if (oom_unkillable_task(task, NULL, oc->nodemask)) | |
261 | return OOM_SCAN_CONTINUE; | |
262 | ||
263 | /* | |
264 | * This task already has access to memory reserves and is being killed. | |
265 | * Don't allow any other task to have access to the reserves. | |
266 | */ | |
267 | if (test_tsk_thread_flag(task, TIF_MEMDIE)) { | |
268 | if (!oc->force_kill) | |
269 | return OOM_SCAN_ABORT; | |
270 | } | |
271 | if (!task->mm) | |
272 | return OOM_SCAN_CONTINUE; | |
273 | ||
274 | /* | |
275 | * If task is allocating a lot of memory and has been marked to be | |
276 | * killed first if it triggers an oom, then select it. | |
277 | */ | |
278 | if (oom_task_origin(task)) | |
279 | return OOM_SCAN_SELECT; | |
280 | ||
281 | if (task_will_free_mem(task) && !oc->force_kill) | |
282 | return OOM_SCAN_ABORT; | |
283 | ||
284 | return OOM_SCAN_OK; | |
285 | } | |
286 | ||
287 | /* | |
288 | * Simple selection loop. We chose the process with the highest | |
289 | * number of 'points'. Returns -1 on scan abort. | |
290 | */ | |
291 | static struct task_struct *select_bad_process(struct oom_control *oc, | |
292 | unsigned int *ppoints, unsigned long totalpages) | |
293 | { | |
294 | struct task_struct *g, *p; | |
295 | struct task_struct *chosen = NULL; | |
296 | unsigned long chosen_points = 0; | |
297 | ||
298 | rcu_read_lock(); | |
299 | for_each_process_thread(g, p) { | |
300 | unsigned int points; | |
301 | ||
302 | switch (oom_scan_process_thread(oc, p, totalpages)) { | |
303 | case OOM_SCAN_SELECT: | |
304 | chosen = p; | |
305 | chosen_points = ULONG_MAX; | |
306 | /* fall through */ | |
307 | case OOM_SCAN_CONTINUE: | |
308 | continue; | |
309 | case OOM_SCAN_ABORT: | |
310 | rcu_read_unlock(); | |
311 | return (struct task_struct *)(-1UL); | |
312 | case OOM_SCAN_OK: | |
313 | break; | |
314 | }; | |
315 | points = oom_badness(p, NULL, oc->nodemask, totalpages); | |
316 | if (!points || points < chosen_points) | |
317 | continue; | |
318 | /* Prefer thread group leaders for display purposes */ | |
319 | if (points == chosen_points && thread_group_leader(chosen)) | |
320 | continue; | |
321 | ||
322 | chosen = p; | |
323 | chosen_points = points; | |
324 | } | |
325 | if (chosen) | |
326 | get_task_struct(chosen); | |
327 | rcu_read_unlock(); | |
328 | ||
329 | *ppoints = chosen_points * 1000 / totalpages; | |
330 | return chosen; | |
331 | } | |
332 | ||
333 | /** | |
334 | * dump_tasks - dump current memory state of all system tasks | |
335 | * @memcg: current's memory controller, if constrained | |
336 | * @nodemask: nodemask passed to page allocator for mempolicy ooms | |
337 | * | |
338 | * Dumps the current memory state of all eligible tasks. Tasks not in the same | |
339 | * memcg, not in the same cpuset, or bound to a disjoint set of mempolicy nodes | |
340 | * are not shown. | |
341 | * State information includes task's pid, uid, tgid, vm size, rss, nr_ptes, | |
342 | * swapents, oom_score_adj value, and name. | |
343 | */ | |
344 | static void dump_tasks(struct mem_cgroup *memcg, const nodemask_t *nodemask) | |
345 | { | |
346 | struct task_struct *p; | |
347 | struct task_struct *task; | |
348 | ||
349 | pr_info("[ pid ] uid tgid total_vm rss nr_ptes nr_pmds swapents oom_score_adj name\n"); | |
350 | rcu_read_lock(); | |
351 | for_each_process(p) { | |
352 | if (oom_unkillable_task(p, memcg, nodemask)) | |
353 | continue; | |
354 | ||
355 | task = find_lock_task_mm(p); | |
356 | if (!task) { | |
357 | /* | |
358 | * This is a kthread or all of p's threads have already | |
359 | * detached their mm's. There's no need to report | |
360 | * them; they can't be oom killed anyway. | |
361 | */ | |
362 | continue; | |
363 | } | |
364 | ||
365 | pr_info("[%5d] %5d %5d %8lu %8lu %7ld %7ld %8lu %5hd %s\n", | |
366 | task->pid, from_kuid(&init_user_ns, task_uid(task)), | |
367 | task->tgid, task->mm->total_vm, get_mm_rss(task->mm), | |
368 | atomic_long_read(&task->mm->nr_ptes), | |
369 | mm_nr_pmds(task->mm), | |
370 | get_mm_counter(task->mm, MM_SWAPENTS), | |
371 | task->signal->oom_score_adj, task->comm); | |
372 | task_unlock(task); | |
373 | } | |
374 | rcu_read_unlock(); | |
375 | } | |
376 | ||
377 | static void dump_header(struct oom_control *oc, struct task_struct *p, | |
378 | struct mem_cgroup *memcg) | |
379 | { | |
380 | task_lock(current); | |
381 | pr_warning("%s invoked oom-killer: gfp_mask=0x%x, order=%d, " | |
382 | "oom_score_adj=%hd\n", | |
383 | current->comm, oc->gfp_mask, oc->order, | |
384 | current->signal->oom_score_adj); | |
385 | cpuset_print_task_mems_allowed(current); | |
386 | task_unlock(current); | |
387 | dump_stack(); | |
388 | if (memcg) | |
389 | mem_cgroup_print_oom_info(memcg, p); | |
390 | else | |
391 | show_mem(SHOW_MEM_FILTER_NODES); | |
392 | if (sysctl_oom_dump_tasks) | |
393 | dump_tasks(memcg, oc->nodemask); | |
394 | } | |
395 | ||
396 | /* | |
397 | * Number of OOM victims in flight | |
398 | */ | |
399 | static atomic_t oom_victims = ATOMIC_INIT(0); | |
400 | static DECLARE_WAIT_QUEUE_HEAD(oom_victims_wait); | |
401 | ||
402 | bool oom_killer_disabled __read_mostly; | |
403 | ||
404 | /** | |
405 | * mark_oom_victim - mark the given task as OOM victim | |
406 | * @tsk: task to mark | |
407 | * | |
408 | * Has to be called with oom_lock held and never after | |
409 | * oom has been disabled already. | |
410 | */ | |
411 | void mark_oom_victim(struct task_struct *tsk) | |
412 | { | |
413 | WARN_ON(oom_killer_disabled); | |
414 | /* OOM killer might race with memcg OOM */ | |
415 | if (test_and_set_tsk_thread_flag(tsk, TIF_MEMDIE)) | |
416 | return; | |
417 | /* | |
418 | * Make sure that the task is woken up from uninterruptible sleep | |
419 | * if it is frozen because OOM killer wouldn't be able to free | |
420 | * any memory and livelock. freezing_slow_path will tell the freezer | |
421 | * that TIF_MEMDIE tasks should be ignored. | |
422 | */ | |
423 | __thaw_task(tsk); | |
424 | atomic_inc(&oom_victims); | |
425 | } | |
426 | ||
427 | /** | |
428 | * exit_oom_victim - note the exit of an OOM victim | |
429 | */ | |
430 | void exit_oom_victim(void) | |
431 | { | |
432 | clear_thread_flag(TIF_MEMDIE); | |
433 | ||
434 | if (!atomic_dec_return(&oom_victims)) | |
435 | wake_up_all(&oom_victims_wait); | |
436 | } | |
437 | ||
438 | /** | |
439 | * oom_killer_disable - disable OOM killer | |
440 | * | |
441 | * Forces all page allocations to fail rather than trigger OOM killer. | |
442 | * Will block and wait until all OOM victims are killed. | |
443 | * | |
444 | * The function cannot be called when there are runnable user tasks because | |
445 | * the userspace would see unexpected allocation failures as a result. Any | |
446 | * new usage of this function should be consulted with MM people. | |
447 | * | |
448 | * Returns true if successful and false if the OOM killer cannot be | |
449 | * disabled. | |
450 | */ | |
451 | bool oom_killer_disable(void) | |
452 | { | |
453 | /* | |
454 | * Make sure to not race with an ongoing OOM killer | |
455 | * and that the current is not the victim. | |
456 | */ | |
457 | mutex_lock(&oom_lock); | |
458 | if (test_thread_flag(TIF_MEMDIE)) { | |
459 | mutex_unlock(&oom_lock); | |
460 | return false; | |
461 | } | |
462 | ||
463 | oom_killer_disabled = true; | |
464 | mutex_unlock(&oom_lock); | |
465 | ||
466 | wait_event(oom_victims_wait, !atomic_read(&oom_victims)); | |
467 | ||
468 | return true; | |
469 | } | |
470 | ||
471 | /** | |
472 | * oom_killer_enable - enable OOM killer | |
473 | */ | |
474 | void oom_killer_enable(void) | |
475 | { | |
476 | oom_killer_disabled = false; | |
477 | } | |
478 | ||
479 | #define K(x) ((x) << (PAGE_SHIFT-10)) | |
480 | /* | |
481 | * Must be called while holding a reference to p, which will be released upon | |
482 | * returning. | |
483 | */ | |
484 | void oom_kill_process(struct oom_control *oc, struct task_struct *p, | |
485 | unsigned int points, unsigned long totalpages, | |
486 | struct mem_cgroup *memcg, const char *message) | |
487 | { | |
488 | struct task_struct *victim = p; | |
489 | struct task_struct *child; | |
490 | struct task_struct *t; | |
491 | struct mm_struct *mm; | |
492 | unsigned int victim_points = 0; | |
493 | static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL, | |
494 | DEFAULT_RATELIMIT_BURST); | |
495 | ||
496 | /* | |
497 | * If the task is already exiting, don't alarm the sysadmin or kill | |
498 | * its children or threads, just set TIF_MEMDIE so it can die quickly | |
499 | */ | |
500 | task_lock(p); | |
501 | if (p->mm && task_will_free_mem(p)) { | |
502 | mark_oom_victim(p); | |
503 | task_unlock(p); | |
504 | put_task_struct(p); | |
505 | return; | |
506 | } | |
507 | task_unlock(p); | |
508 | ||
509 | if (__ratelimit(&oom_rs)) | |
510 | dump_header(oc, p, memcg); | |
511 | ||
512 | task_lock(p); | |
513 | pr_err("%s: Kill process %d (%s) score %u or sacrifice child\n", | |
514 | message, task_pid_nr(p), p->comm, points); | |
515 | task_unlock(p); | |
516 | ||
517 | /* | |
518 | * If any of p's children has a different mm and is eligible for kill, | |
519 | * the one with the highest oom_badness() score is sacrificed for its | |
520 | * parent. This attempts to lose the minimal amount of work done while | |
521 | * still freeing memory. | |
522 | */ | |
523 | read_lock(&tasklist_lock); | |
524 | for_each_thread(p, t) { | |
525 | list_for_each_entry(child, &t->children, sibling) { | |
526 | unsigned int child_points; | |
527 | ||
528 | if (child->mm == p->mm) | |
529 | continue; | |
530 | /* | |
531 | * oom_badness() returns 0 if the thread is unkillable | |
532 | */ | |
533 | child_points = oom_badness(child, memcg, oc->nodemask, | |
534 | totalpages); | |
535 | if (child_points > victim_points) { | |
536 | put_task_struct(victim); | |
537 | victim = child; | |
538 | victim_points = child_points; | |
539 | get_task_struct(victim); | |
540 | } | |
541 | } | |
542 | } | |
543 | read_unlock(&tasklist_lock); | |
544 | ||
545 | p = find_lock_task_mm(victim); | |
546 | if (!p) { | |
547 | put_task_struct(victim); | |
548 | return; | |
549 | } else if (victim != p) { | |
550 | get_task_struct(p); | |
551 | put_task_struct(victim); | |
552 | victim = p; | |
553 | } | |
554 | ||
555 | /* mm cannot safely be dereferenced after task_unlock(victim) */ | |
556 | mm = victim->mm; | |
557 | mark_oom_victim(victim); | |
558 | pr_err("Killed process %d (%s) total-vm:%lukB, anon-rss:%lukB, file-rss:%lukB\n", | |
559 | task_pid_nr(victim), victim->comm, K(victim->mm->total_vm), | |
560 | K(get_mm_counter(victim->mm, MM_ANONPAGES)), | |
561 | K(get_mm_counter(victim->mm, MM_FILEPAGES))); | |
562 | task_unlock(victim); | |
563 | ||
564 | /* | |
565 | * Kill all user processes sharing victim->mm in other thread groups, if | |
566 | * any. They don't get access to memory reserves, though, to avoid | |
567 | * depletion of all memory. This prevents mm->mmap_sem livelock when an | |
568 | * oom killed thread cannot exit because it requires the semaphore and | |
569 | * its contended by another thread trying to allocate memory itself. | |
570 | * That thread will now get access to memory reserves since it has a | |
571 | * pending fatal signal. | |
572 | */ | |
573 | rcu_read_lock(); | |
574 | for_each_process(p) | |
575 | if (p->mm == mm && !same_thread_group(p, victim) && | |
576 | !(p->flags & PF_KTHREAD)) { | |
577 | if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) | |
578 | continue; | |
579 | ||
580 | task_lock(p); /* Protect ->comm from prctl() */ | |
581 | pr_err("Kill process %d (%s) sharing same memory\n", | |
582 | task_pid_nr(p), p->comm); | |
583 | task_unlock(p); | |
584 | do_send_sig_info(SIGKILL, SEND_SIG_FORCED, p, true); | |
585 | } | |
586 | rcu_read_unlock(); | |
587 | ||
588 | do_send_sig_info(SIGKILL, SEND_SIG_FORCED, victim, true); | |
589 | put_task_struct(victim); | |
590 | } | |
591 | #undef K | |
592 | ||
593 | /* | |
594 | * Determines whether the kernel must panic because of the panic_on_oom sysctl. | |
595 | */ | |
596 | void check_panic_on_oom(struct oom_control *oc, enum oom_constraint constraint, | |
597 | struct mem_cgroup *memcg) | |
598 | { | |
599 | if (likely(!sysctl_panic_on_oom)) | |
600 | return; | |
601 | if (sysctl_panic_on_oom != 2) { | |
602 | /* | |
603 | * panic_on_oom == 1 only affects CONSTRAINT_NONE, the kernel | |
604 | * does not panic for cpuset, mempolicy, or memcg allocation | |
605 | * failures. | |
606 | */ | |
607 | if (constraint != CONSTRAINT_NONE) | |
608 | return; | |
609 | } | |
610 | dump_header(oc, NULL, memcg); | |
611 | panic("Out of memory: %s panic_on_oom is enabled\n", | |
612 | sysctl_panic_on_oom == 2 ? "compulsory" : "system-wide"); | |
613 | } | |
614 | ||
615 | static BLOCKING_NOTIFIER_HEAD(oom_notify_list); | |
616 | ||
617 | int register_oom_notifier(struct notifier_block *nb) | |
618 | { | |
619 | return blocking_notifier_chain_register(&oom_notify_list, nb); | |
620 | } | |
621 | EXPORT_SYMBOL_GPL(register_oom_notifier); | |
622 | ||
623 | int unregister_oom_notifier(struct notifier_block *nb) | |
624 | { | |
625 | return blocking_notifier_chain_unregister(&oom_notify_list, nb); | |
626 | } | |
627 | EXPORT_SYMBOL_GPL(unregister_oom_notifier); | |
628 | ||
629 | /** | |
630 | * out_of_memory - kill the "best" process when we run out of memory | |
631 | * @oc: pointer to struct oom_control | |
632 | * | |
633 | * If we run out of memory, we have the choice between either | |
634 | * killing a random task (bad), letting the system crash (worse) | |
635 | * OR try to be smart about which process to kill. Note that we | |
636 | * don't have to be perfect here, we just have to be good. | |
637 | */ | |
638 | bool out_of_memory(struct oom_control *oc) | |
639 | { | |
640 | struct task_struct *p; | |
641 | unsigned long totalpages; | |
642 | unsigned long freed = 0; | |
643 | unsigned int uninitialized_var(points); | |
644 | enum oom_constraint constraint = CONSTRAINT_NONE; | |
645 | int killed = 0; | |
646 | ||
647 | if (oom_killer_disabled) | |
648 | return false; | |
649 | ||
650 | blocking_notifier_call_chain(&oom_notify_list, 0, &freed); | |
651 | if (freed > 0) | |
652 | /* Got some memory back in the last second. */ | |
653 | goto out; | |
654 | ||
655 | /* | |
656 | * If current has a pending SIGKILL or is exiting, then automatically | |
657 | * select it. The goal is to allow it to allocate so that it may | |
658 | * quickly exit and free its memory. | |
659 | * | |
660 | * But don't select if current has already released its mm and cleared | |
661 | * TIF_MEMDIE flag at exit_mm(), otherwise an OOM livelock may occur. | |
662 | */ | |
663 | if (current->mm && | |
664 | (fatal_signal_pending(current) || task_will_free_mem(current))) { | |
665 | mark_oom_victim(current); | |
666 | goto out; | |
667 | } | |
668 | ||
669 | /* | |
670 | * Check if there were limitations on the allocation (only relevant for | |
671 | * NUMA) that may require different handling. | |
672 | */ | |
673 | constraint = constrained_alloc(oc, &totalpages); | |
674 | if (constraint != CONSTRAINT_MEMORY_POLICY) | |
675 | oc->nodemask = NULL; | |
676 | check_panic_on_oom(oc, constraint, NULL); | |
677 | ||
678 | if (sysctl_oom_kill_allocating_task && current->mm && | |
679 | !oom_unkillable_task(current, NULL, oc->nodemask) && | |
680 | current->signal->oom_score_adj != OOM_SCORE_ADJ_MIN) { | |
681 | get_task_struct(current); | |
682 | oom_kill_process(oc, current, 0, totalpages, NULL, | |
683 | "Out of memory (oom_kill_allocating_task)"); | |
684 | goto out; | |
685 | } | |
686 | ||
687 | p = select_bad_process(oc, &points, totalpages); | |
688 | /* Found nothing?!?! Either we hang forever, or we panic. */ | |
689 | if (!p) { | |
690 | dump_header(oc, NULL, NULL); | |
691 | panic("Out of memory and no killable processes...\n"); | |
692 | } | |
693 | if (p != (void *)-1UL) { | |
694 | oom_kill_process(oc, p, points, totalpages, NULL, | |
695 | "Out of memory"); | |
696 | killed = 1; | |
697 | } | |
698 | out: | |
699 | /* | |
700 | * Give the killed threads a good chance of exiting before trying to | |
701 | * allocate memory again. | |
702 | */ | |
703 | if (killed) | |
704 | schedule_timeout_killable(1); | |
705 | ||
706 | return true; | |
707 | } | |
708 | ||
709 | /* | |
710 | * The pagefault handler calls here because it is out of memory, so kill a | |
711 | * memory-hogging task. If any populated zone has ZONE_OOM_LOCKED set, a | |
712 | * parallel oom killing is already in progress so do nothing. | |
713 | */ | |
714 | void pagefault_out_of_memory(void) | |
715 | { | |
716 | struct oom_control oc = { | |
717 | .zonelist = NULL, | |
718 | .nodemask = NULL, | |
719 | .gfp_mask = 0, | |
720 | .order = 0, | |
721 | .force_kill = false, | |
722 | }; | |
723 | ||
724 | if (mem_cgroup_oom_synchronize(true)) | |
725 | return; | |
726 | ||
727 | if (!mutex_trylock(&oom_lock)) | |
728 | return; | |
729 | ||
730 | if (!out_of_memory(&oc)) { | |
731 | /* | |
732 | * There shouldn't be any user tasks runnable while the | |
733 | * OOM killer is disabled, so the current task has to | |
734 | * be a racing OOM victim for which oom_killer_disable() | |
735 | * is waiting for. | |
736 | */ | |
737 | WARN_ON(test_thread_flag(TIF_MEMDIE)); | |
738 | } | |
739 | ||
740 | mutex_unlock(&oom_lock); | |
741 | } |