oom_kill.c 20.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6
/*
 *  linux/mm/oom_kill.c
 * 
 *  Copyright (C)  1998,2000  Rik van Riel
 *	Thanks go out to Claus Fischer for some serious inspiration and
 *	for goading me into coding this file...
D
David Rientjes 已提交
7 8
 *  Copyright (C)  2010  Google, Inc.
 *	Rewritten by David Rientjes
L
Linus Torvalds 已提交
9 10
 *
 *  The routines in this file are used to kill a process when
P
Paul Jackson 已提交
11 12
 *  we're seriously out of memory. This gets called from __alloc_pages()
 *  in mm/page_alloc.c when we really run out of memory.
L
Linus Torvalds 已提交
13 14 15 16 17 18 19
 *
 *  Since we won't call these routines often (on a well-configured
 *  machine) this file will double as a 'coding guide' and a signpost
 *  for newbie kernel hackers. It features several pointers to major
 *  kernel subsystems and hints as to where to find out what things do.
 */

20
#include <linux/oom.h>
L
Linus Torvalds 已提交
21
#include <linux/mm.h>
A
Alexey Dobriyan 已提交
22
#include <linux/err.h>
23
#include <linux/gfp.h>
L
Linus Torvalds 已提交
24 25 26 27
#include <linux/sched.h>
#include <linux/swap.h>
#include <linux/timex.h>
#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 40

#define CREATE_TRACE_POINTS
#include <trace/events/oom.h>
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
47

48 49 50
#ifdef CONFIG_NUMA
/**
 * has_intersects_mems_allowed() - check task eligiblity for kill
51
 * @start: task struct of which task to consider
52 53 54 55 56
 * @mask: nodemask passed to page allocator for mempolicy ooms
 *
 * Task eligibility is determined by whether or not a candidate task, @tsk,
 * shares the same mempolicy nodes as current if it is bound by such a policy
 * 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 62
	struct task_struct *tsk;
	bool ret = false;
63

64
	rcu_read_lock();
65
	for_each_thread(start, tsk) {
66 67 68 69 70 71 72
		if (mask) {
			/*
			 * If this is a mempolicy constrained oom, tsk's
			 * cpuset is irrelevant.  Only return true if its
			 * mempolicy intersects current, otherwise it may be
			 * needlessly killed.
			 */
73
			ret = mempolicy_nodemask_intersects(tsk, mask);
74 75 76 77 78
		} else {
			/*
			 * This is not a mempolicy constrained oom, so only
			 * check the mems of tsk's cpuset.
			 */
79
			ret = cpuset_mems_allowed_intersects(current, tsk);
80
		}
81 82
		if (ret)
			break;
83
	}
84
	rcu_read_unlock();
85

86
	return ret;
87 88 89 90 91 92
}
#else
static bool has_intersects_mems_allowed(struct task_struct *tsk,
					const nodemask_t *mask)
{
	return true;
93
}
94
#endif /* CONFIG_NUMA */
95

96 97 98 99 100 101
/*
 * The process p may have detached its own ->mm while exiting or through
 * use_mm(), but one or more of its subthreads may still have a valid
 * pointer.  Return p, or any of its subthreads with a valid ->mm, with
 * task_lock() held.
 */
102
struct task_struct *find_lock_task_mm(struct task_struct *p)
103
{
104
	struct task_struct *t;
105

106 107
	rcu_read_lock();

108
	for_each_thread(p, t) {
109 110
		task_lock(t);
		if (likely(t->mm))
111
			goto found;
112
		task_unlock(t);
113
	}
114 115 116
	t = NULL;
found:
	rcu_read_unlock();
117

118
	return t;
119 120
}

121 122 123 124 125 126 127 128 129
/*
 * order == -1 means the oom kill is required by sysrq, otherwise only
 * for display purposes.
 */
static inline bool is_sysrq_oom(struct oom_control *oc)
{
	return oc->order == -1;
}

130
/* return true if the task is not adequate as candidate victim task. */
131
static bool oom_unkillable_task(struct task_struct *p,
132
		struct mem_cgroup *memcg, const nodemask_t *nodemask)
133 134 135 136 137 138 139
{
	if (is_global_init(p))
		return true;
	if (p->flags & PF_KTHREAD)
		return true;

	/* When mem_cgroup_out_of_memory() and p is not member of the group */
140
	if (memcg && !task_in_mem_cgroup(p, memcg))
141 142 143 144 145 146 147 148 149
		return true;

	/* p may not have freeable memory in nodemask */
	if (!has_intersects_mems_allowed(p, nodemask))
		return true;

	return false;
}

L
Linus Torvalds 已提交
150
/**
D
David Rientjes 已提交
151
 * oom_badness - heuristic function to determine which candidate task to kill
L
Linus Torvalds 已提交
152
 * @p: task struct of which task we should calculate
D
David Rientjes 已提交
153
 * @totalpages: total present RAM allowed for page allocation
L
Linus Torvalds 已提交
154
 *
D
David Rientjes 已提交
155 156 157
 * The heuristic for determining which task to kill is made to be as simple and
 * predictable as possible.  The goal is to return the highest value for the
 * task consuming the most memory to avoid subsequent oom failures.
L
Linus Torvalds 已提交
158
 */
159 160
unsigned long oom_badness(struct task_struct *p, struct mem_cgroup *memcg,
			  const nodemask_t *nodemask, unsigned long totalpages)
L
Linus Torvalds 已提交
161
{
162
	long points;
163
	long adj;
164

165
	if (oom_unkillable_task(p, memcg, nodemask))
166
		return 0;
L
Linus Torvalds 已提交
167

168 169
	p = find_lock_task_mm(p);
	if (!p)
L
Linus Torvalds 已提交
170 171
		return 0;

172
	adj = (long)p->signal->oom_score_adj;
173
	if (adj == OOM_SCORE_ADJ_MIN) {
174 175 176 177
		task_unlock(p);
		return 0;
	}

L
Linus Torvalds 已提交
178
	/*
D
David Rientjes 已提交
179
	 * The baseline for the badness score is the proportion of RAM that each
180
	 * task's rss, pagetable and swap space use.
L
Linus Torvalds 已提交
181
	 */
182 183
	points = get_mm_rss(p->mm) + get_mm_counter(p->mm, MM_SWAPENTS) +
		atomic_long_read(&p->mm->nr_ptes) + mm_nr_pmds(p->mm);
D
David Rientjes 已提交
184
	task_unlock(p);
L
Linus Torvalds 已提交
185 186

	/*
D
David Rientjes 已提交
187 188
	 * Root processes get 3% bonus, just like the __vm_enough_memory()
	 * implementation used by LSMs.
L
Linus Torvalds 已提交
189
	 */
D
David Rientjes 已提交
190
	if (has_capability_noaudit(p, CAP_SYS_ADMIN))
191
		points -= (points * 3) / 100;
L
Linus Torvalds 已提交
192

193 194 195
	/* Normalize to oom_score_adj units */
	adj *= totalpages / 1000;
	points += adj;
L
Linus Torvalds 已提交
196

197
	/*
198 199
	 * Never return 0 for an eligible task regardless of the root bonus and
	 * oom_score_adj (oom_score_adj can't be OOM_SCORE_ADJ_MIN here).
200
	 */
201
	return points > 0 ? points : 1;
L
Linus Torvalds 已提交
202 203
}

204 205 206 207
/*
 * Determine the type of allocation constraint.
 */
#ifdef CONFIG_NUMA
208 209
static enum oom_constraint constrained_alloc(struct oom_control *oc,
					     unsigned long *totalpages)
210
{
211
	struct zone *zone;
212
	struct zoneref *z;
213
	enum zone_type high_zoneidx = gfp_zone(oc->gfp_mask);
D
David Rientjes 已提交
214 215
	bool cpuset_limited = false;
	int nid;
216

D
David Rientjes 已提交
217 218 219
	/* Default to all available memory */
	*totalpages = totalram_pages + total_swap_pages;

220
	if (!oc->zonelist)
D
David Rientjes 已提交
221
		return CONSTRAINT_NONE;
222 223 224 225 226
	/*
	 * Reach here only when __GFP_NOFAIL is used. So, we should avoid
	 * to kill current.We have to random task kill in this case.
	 * Hopefully, CONSTRAINT_THISNODE...but no way to handle it, now.
	 */
227
	if (oc->gfp_mask & __GFP_THISNODE)
228
		return CONSTRAINT_NONE;
229

230
	/*
D
David Rientjes 已提交
231 232 233
	 * This is not a __GFP_THISNODE allocation, so a truncated nodemask in
	 * the page allocator means a mempolicy is in effect.  Cpuset policy
	 * is enforced in get_page_from_freelist().
234
	 */
235 236
	if (oc->nodemask &&
	    !nodes_subset(node_states[N_MEMORY], *oc->nodemask)) {
D
David Rientjes 已提交
237
		*totalpages = total_swap_pages;
238
		for_each_node_mask(nid, *oc->nodemask)
D
David Rientjes 已提交
239
			*totalpages += node_spanned_pages(nid);
240
		return CONSTRAINT_MEMORY_POLICY;
D
David Rientjes 已提交
241
	}
242 243

	/* Check this allocation failure is caused by cpuset's wall function */
244 245 246
	for_each_zone_zonelist_nodemask(zone, z, oc->zonelist,
			high_zoneidx, oc->nodemask)
		if (!cpuset_zone_allowed(zone, oc->gfp_mask))
D
David Rientjes 已提交
247
			cpuset_limited = true;
248

D
David Rientjes 已提交
249 250 251 252 253 254
	if (cpuset_limited) {
		*totalpages = total_swap_pages;
		for_each_node_mask(nid, cpuset_current_mems_allowed)
			*totalpages += node_spanned_pages(nid);
		return CONSTRAINT_CPUSET;
	}
255 256
	return CONSTRAINT_NONE;
}
257
#else
258 259
static enum oom_constraint constrained_alloc(struct oom_control *oc,
					     unsigned long *totalpages)
260
{
D
David Rientjes 已提交
261
	*totalpages = totalram_pages + total_swap_pages;
262 263 264
	return CONSTRAINT_NONE;
}
#endif
265

266 267
enum oom_scan_t oom_scan_process_thread(struct oom_control *oc,
			struct task_struct *task, unsigned long totalpages)
268
{
269
	if (oom_unkillable_task(task, NULL, oc->nodemask))
270 271 272 273 274 275 276
		return OOM_SCAN_CONTINUE;

	/*
	 * This task already has access to memory reserves and is being killed.
	 * Don't allow any other task to have access to the reserves.
	 */
	if (test_tsk_thread_flag(task, TIF_MEMDIE)) {
277
		if (!is_sysrq_oom(oc))
278 279 280 281 282
			return OOM_SCAN_ABORT;
	}
	if (!task->mm)
		return OOM_SCAN_CONTINUE;

283 284 285 286 287 288 289
	/*
	 * If task is allocating a lot of memory and has been marked to be
	 * killed first if it triggers an oom, then select it.
	 */
	if (oom_task_origin(task))
		return OOM_SCAN_SELECT;

290 291 292
	return OOM_SCAN_OK;
}

L
Linus Torvalds 已提交
293 294
/*
 * Simple selection loop. We chose the process with the highest
295
 * number of 'points'.  Returns -1 on scan abort.
L
Linus Torvalds 已提交
296
 */
297 298
static struct task_struct *select_bad_process(struct oom_control *oc,
		unsigned int *ppoints, unsigned long totalpages)
L
Linus Torvalds 已提交
299
{
300
	struct task_struct *g, *p;
L
Linus Torvalds 已提交
301
	struct task_struct *chosen = NULL;
302
	unsigned long chosen_points = 0;
L
Linus Torvalds 已提交
303

304
	rcu_read_lock();
305
	for_each_process_thread(g, p) {
D
David Rientjes 已提交
306
		unsigned int points;
P
Paul Jackson 已提交
307

308
		switch (oom_scan_process_thread(oc, p, totalpages)) {
309 310 311 312 313
		case OOM_SCAN_SELECT:
			chosen = p;
			chosen_points = ULONG_MAX;
			/* fall through */
		case OOM_SCAN_CONTINUE:
314
			continue;
315
		case OOM_SCAN_ABORT:
316
			rcu_read_unlock();
317
			return (struct task_struct *)(-1UL);
318 319 320
		case OOM_SCAN_OK:
			break;
		};
321
		points = oom_badness(p, NULL, oc->nodemask, totalpages);
322 323 324 325 326 327 328 329
		if (!points || points < chosen_points)
			continue;
		/* Prefer thread group leaders for display purposes */
		if (points == chosen_points && thread_group_leader(chosen))
			continue;

		chosen = p;
		chosen_points = points;
330
	}
331 332 333
	if (chosen)
		get_task_struct(chosen);
	rcu_read_unlock();
334

335
	*ppoints = chosen_points * 1000 / totalpages;
L
Linus Torvalds 已提交
336 337 338
	return chosen;
}

339
/**
R
Randy Dunlap 已提交
340
 * dump_tasks - dump current memory state of all system tasks
W
Wanpeng Li 已提交
341
 * @memcg: current's memory controller, if constrained
342
 * @nodemask: nodemask passed to page allocator for mempolicy ooms
R
Randy Dunlap 已提交
343
 *
344 345 346
 * Dumps the current memory state of all eligible tasks.  Tasks not in the same
 * memcg, not in the same cpuset, or bound to a disjoint set of mempolicy nodes
 * are not shown.
347 348
 * State information includes task's pid, uid, tgid, vm size, rss, nr_ptes,
 * swapents, oom_score_adj value, and name.
349
 */
350
static void dump_tasks(struct mem_cgroup *memcg, const nodemask_t *nodemask)
351
{
352 353
	struct task_struct *p;
	struct task_struct *task;
354

355
	pr_info("[ pid ]   uid  tgid total_vm      rss nr_ptes nr_pmds swapents oom_score_adj name\n");
356
	rcu_read_lock();
357
	for_each_process(p) {
358
		if (oom_unkillable_task(p, memcg, nodemask))
359
			continue;
360

361 362
		task = find_lock_task_mm(p);
		if (!task) {
363
			/*
364 365
			 * This is a kthread or all of p's threads have already
			 * detached their mm's.  There's no need to report
366
			 * them; they can't be oom killed anyway.
367 368 369
			 */
			continue;
		}
370

371
		pr_info("[%5d] %5d %5d %8lu %8lu %7ld %7ld %8lu         %5hd %s\n",
372 373
			task->pid, from_kuid(&init_user_ns, task_uid(task)),
			task->tgid, task->mm->total_vm, get_mm_rss(task->mm),
374
			atomic_long_read(&task->mm->nr_ptes),
375
			mm_nr_pmds(task->mm),
376
			get_mm_counter(task->mm, MM_SWAPENTS),
D
David Rientjes 已提交
377
			task->signal->oom_score_adj, task->comm);
378 379
		task_unlock(task);
	}
380
	rcu_read_unlock();
381 382
}

383 384
static void dump_header(struct oom_control *oc, struct task_struct *p,
			struct mem_cgroup *memcg)
385
{
386 387 388
	pr_warn("%s invoked oom-killer: gfp_mask=%#x(%pGg), order=%d, "
			"oom_score_adj=%hd\n",
		current->comm, oc->gfp_mask, &oc->gfp_mask, oc->order,
D
David Rientjes 已提交
389
		current->signal->oom_score_adj);
390

391
	cpuset_print_current_mems_allowed();
392
	dump_stack();
393 394 395 396
	if (memcg)
		mem_cgroup_print_oom_info(memcg, p);
	else
		show_mem(SHOW_MEM_FILTER_NODES);
397
	if (sysctl_oom_dump_tasks)
398
		dump_tasks(memcg, oc->nodemask);
399 400
}

401
/*
402
 * Number of OOM victims in flight
403
 */
404 405
static atomic_t oom_victims = ATOMIC_INIT(0);
static DECLARE_WAIT_QUEUE_HEAD(oom_victims_wait);
406

407
bool oom_killer_disabled __read_mostly;
408

409
/**
410
 * mark_oom_victim - mark the given task as OOM victim
411
 * @tsk: task to mark
412
 *
413
 * Has to be called with oom_lock held and never after
414
 * oom has been disabled already.
415
 */
416
void mark_oom_victim(struct task_struct *tsk)
417
{
418 419 420 421
	WARN_ON(oom_killer_disabled);
	/* OOM killer might race with memcg OOM */
	if (test_and_set_tsk_thread_flag(tsk, TIF_MEMDIE))
		return;
422 423 424 425 426 427 428
	/*
	 * Make sure that the task is woken up from uninterruptible sleep
	 * if it is frozen because OOM killer wouldn't be able to free
	 * any memory and livelock. freezing_slow_path will tell the freezer
	 * that TIF_MEMDIE tasks should be ignored.
	 */
	__thaw_task(tsk);
429
	atomic_inc(&oom_victims);
430 431 432
}

/**
433
 * exit_oom_victim - note the exit of an OOM victim
434
 */
435
void exit_oom_victim(void)
436
{
437
	clear_thread_flag(TIF_MEMDIE);
438

439
	if (!atomic_dec_return(&oom_victims))
440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
		wake_up_all(&oom_victims_wait);
}

/**
 * oom_killer_disable - disable OOM killer
 *
 * Forces all page allocations to fail rather than trigger OOM killer.
 * Will block and wait until all OOM victims are killed.
 *
 * The function cannot be called when there are runnable user tasks because
 * the userspace would see unexpected allocation failures as a result. Any
 * new usage of this function should be consulted with MM people.
 *
 * Returns true if successful and false if the OOM killer cannot be
 * disabled.
 */
bool oom_killer_disable(void)
{
	/*
	 * Make sure to not race with an ongoing OOM killer
	 * and that the current is not the victim.
	 */
462
	mutex_lock(&oom_lock);
463
	if (test_thread_flag(TIF_MEMDIE)) {
464
		mutex_unlock(&oom_lock);
465 466 467 468
		return false;
	}

	oom_killer_disabled = true;
469
	mutex_unlock(&oom_lock);
470 471 472 473 474 475 476 477 478 479 480 481

	wait_event(oom_victims_wait, !atomic_read(&oom_victims));

	return true;
}

/**
 * oom_killer_enable - enable OOM killer
 */
void oom_killer_enable(void)
{
	oom_killer_disabled = false;
482 483
}

484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
/*
 * task->mm can be NULL if the task is the exited group leader.  So to
 * determine whether the task is using a particular mm, we examine all the
 * task's threads: if one of those is using this mm then this task was also
 * using it.
 */
static bool process_shares_mm(struct task_struct *p, struct mm_struct *mm)
{
	struct task_struct *t;

	for_each_thread(p, t) {
		struct mm_struct *t_mm = READ_ONCE(t->mm);
		if (t_mm)
			return t_mm == mm;
	}
	return false;
}

502
#define K(x) ((x) << (PAGE_SHIFT-10))
503 504 505 506
/*
 * Must be called while holding a reference to p, which will be released upon
 * returning.
 */
507
void oom_kill_process(struct oom_control *oc, struct task_struct *p,
508
		      unsigned int points, unsigned long totalpages,
509
		      struct mem_cgroup *memcg, const char *message)
L
Linus Torvalds 已提交
510
{
511
	struct task_struct *victim = p;
512
	struct task_struct *child;
513
	struct task_struct *t;
514
	struct mm_struct *mm;
515
	unsigned int victim_points = 0;
516 517
	static DEFINE_RATELIMIT_STATE(oom_rs, DEFAULT_RATELIMIT_INTERVAL,
					      DEFAULT_RATELIMIT_BURST);
L
Linus Torvalds 已提交
518

519 520 521 522
	/*
	 * If the task is already exiting, don't alarm the sysadmin or kill
	 * its children or threads, just set TIF_MEMDIE so it can die quickly
	 */
523 524
	task_lock(p);
	if (p->mm && task_will_free_mem(p)) {
525
		mark_oom_victim(p);
526
		task_unlock(p);
527
		put_task_struct(p);
528
		return;
529
	}
530
	task_unlock(p);
531

532
	if (__ratelimit(&oom_rs))
533
		dump_header(oc, p, memcg);
534

535
	pr_err("%s: Kill process %d (%s) score %u or sacrifice child\n",
536
		message, task_pid_nr(p), p->comm, points);
N
Nick Piggin 已提交
537

538 539
	/*
	 * If any of p's children has a different mm and is eligible for kill,
540
	 * the one with the highest oom_badness() score is sacrificed for its
541 542 543
	 * parent.  This attempts to lose the minimal amount of work done while
	 * still freeing memory.
	 */
544
	read_lock(&tasklist_lock);
545
	for_each_thread(p, t) {
546
		list_for_each_entry(child, &t->children, sibling) {
D
David Rientjes 已提交
547
			unsigned int child_points;
548

549
			if (process_shares_mm(child, p->mm))
550
				continue;
D
David Rientjes 已提交
551 552 553
			/*
			 * oom_badness() returns 0 if the thread is unkillable
			 */
554
			child_points = oom_badness(child, memcg, oc->nodemask,
D
David Rientjes 已提交
555
								totalpages);
556
			if (child_points > victim_points) {
557
				put_task_struct(victim);
558 559
				victim = child;
				victim_points = child_points;
560
				get_task_struct(victim);
561
			}
562
		}
563
	}
564
	read_unlock(&tasklist_lock);
565

566 567 568
	p = find_lock_task_mm(victim);
	if (!p) {
		put_task_struct(victim);
569
		return;
570 571 572 573 574
	} else if (victim != p) {
		get_task_struct(p);
		put_task_struct(victim);
		victim = p;
	}
575

576
	/* Get a reference to safely compare mm after task_unlock(victim) */
577
	mm = victim->mm;
578
	atomic_inc(&mm->mm_count);
579 580 581 582 583 584
	/*
	 * We should send SIGKILL before setting TIF_MEMDIE in order to prevent
	 * the OOM victim from depleting the memory reserves from the user
	 * space under its control.
	 */
	do_send_sig_info(SIGKILL, SEND_SIG_FORCED, victim, true);
585
	mark_oom_victim(victim);
586
	pr_err("Killed process %d (%s) total-vm:%lukB, anon-rss:%lukB, file-rss:%lukB, shmem-rss:%lukB\n",
587 588
		task_pid_nr(victim), victim->comm, K(victim->mm->total_vm),
		K(get_mm_counter(victim->mm, MM_ANONPAGES)),
589 590
		K(get_mm_counter(victim->mm, MM_FILEPAGES)),
		K(get_mm_counter(victim->mm, MM_SHMEMPAGES)));
591 592 593 594 595 596 597 598 599 600 601
	task_unlock(victim);

	/*
	 * Kill all user processes sharing victim->mm in other thread groups, if
	 * any.  They don't get access to memory reserves, though, to avoid
	 * depletion of all memory.  This prevents mm->mmap_sem livelock when an
	 * oom killed thread cannot exit because it requires the semaphore and
	 * its contended by another thread trying to allocate memory itself.
	 * That thread will now get access to memory reserves since it has a
	 * pending fatal signal.
	 */
602
	rcu_read_lock();
603
	for_each_process(p) {
604
		if (!process_shares_mm(p, mm))
605 606 607 608 609
			continue;
		if (same_thread_group(p, victim))
			continue;
		if (unlikely(p->flags & PF_KTHREAD))
			continue;
610 611
		if (is_global_init(p))
			continue;
612 613
		if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
			continue;
614

615 616
		do_send_sig_info(SIGKILL, SEND_SIG_FORCED, p, true);
	}
617
	rcu_read_unlock();
618

619
	mmdrop(mm);
620
	put_task_struct(victim);
L
Linus Torvalds 已提交
621
}
622
#undef K
L
Linus Torvalds 已提交
623

624 625 626
/*
 * Determines whether the kernel must panic because of the panic_on_oom sysctl.
 */
627
void check_panic_on_oom(struct oom_control *oc, enum oom_constraint constraint,
628
			struct mem_cgroup *memcg)
629 630 631 632 633 634 635 636 637 638 639 640
{
	if (likely(!sysctl_panic_on_oom))
		return;
	if (sysctl_panic_on_oom != 2) {
		/*
		 * panic_on_oom == 1 only affects CONSTRAINT_NONE, the kernel
		 * does not panic for cpuset, mempolicy, or memcg allocation
		 * failures.
		 */
		if (constraint != CONSTRAINT_NONE)
			return;
	}
641
	/* Do not panic for oom kills triggered by sysrq */
642
	if (is_sysrq_oom(oc))
643
		return;
644
	dump_header(oc, NULL, memcg);
645 646 647 648
	panic("Out of memory: %s panic_on_oom is enabled\n",
		sysctl_panic_on_oom == 2 ? "compulsory" : "system-wide");
}

649 650 651 652 653 654 655 656 657 658 659 660 661 662
static BLOCKING_NOTIFIER_HEAD(oom_notify_list);

int register_oom_notifier(struct notifier_block *nb)
{
	return blocking_notifier_chain_register(&oom_notify_list, nb);
}
EXPORT_SYMBOL_GPL(register_oom_notifier);

int unregister_oom_notifier(struct notifier_block *nb)
{
	return blocking_notifier_chain_unregister(&oom_notify_list, nb);
}
EXPORT_SYMBOL_GPL(unregister_oom_notifier);

L
Linus Torvalds 已提交
663
/**
664 665
 * out_of_memory - kill the "best" process when we run out of memory
 * @oc: pointer to struct oom_control
L
Linus Torvalds 已提交
666 667 668 669 670 671
 *
 * If we run out of memory, we have the choice between either
 * killing a random task (bad), letting the system crash (worse)
 * OR try to be smart about which process to kill. Note that we
 * don't have to be perfect here, we just have to be good.
 */
672
bool out_of_memory(struct oom_control *oc)
L
Linus Torvalds 已提交
673
{
674
	struct task_struct *p;
D
David Rientjes 已提交
675
	unsigned long totalpages;
676
	unsigned long freed = 0;
677
	unsigned int uninitialized_var(points);
678
	enum oom_constraint constraint = CONSTRAINT_NONE;
679

680 681 682
	if (oom_killer_disabled)
		return false;

683 684 685
	blocking_notifier_call_chain(&oom_notify_list, 0, &freed);
	if (freed > 0)
		/* Got some memory back in the last second. */
686
		return true;
L
Linus Torvalds 已提交
687

688
	/*
689 690 691
	 * If current has a pending SIGKILL or is exiting, then automatically
	 * select it.  The goal is to allow it to allocate so that it may
	 * quickly exit and free its memory.
692 693 694
	 *
	 * But don't select if current has already released its mm and cleared
	 * TIF_MEMDIE flag at exit_mm(), otherwise an OOM livelock may occur.
695
	 */
696 697
	if (current->mm &&
	    (fatal_signal_pending(current) || task_will_free_mem(current))) {
698
		mark_oom_victim(current);
699
		return true;
700 701
	}

702 703 704 705
	/*
	 * Check if there were limitations on the allocation (only relevant for
	 * NUMA) that may require different handling.
	 */
706 707 708 709
	constraint = constrained_alloc(oc, &totalpages);
	if (constraint != CONSTRAINT_MEMORY_POLICY)
		oc->nodemask = NULL;
	check_panic_on_oom(oc, constraint, NULL);
710

711
	if (sysctl_oom_kill_allocating_task && current->mm &&
712
	    !oom_unkillable_task(current, NULL, oc->nodemask) &&
713
	    current->signal->oom_score_adj != OOM_SCORE_ADJ_MIN) {
714
		get_task_struct(current);
715
		oom_kill_process(oc, current, 0, totalpages, NULL,
716
				 "Out of memory (oom_kill_allocating_task)");
717
		return true;
718 719
	}

720
	p = select_bad_process(oc, &points, totalpages);
721
	/* Found nothing?!?! Either we hang forever, or we panic. */
722
	if (!p && !is_sysrq_oom(oc)) {
723
		dump_header(oc, NULL, NULL);
724 725
		panic("Out of memory and no killable processes...\n");
	}
726
	if (p && p != (void *)-1UL) {
727 728
		oom_kill_process(oc, p, points, totalpages, NULL,
				 "Out of memory");
729 730 731 732
		/*
		 * Give the killed process a good chance to exit before trying
		 * to allocate memory again.
		 */
733
		schedule_timeout_killable(1);
734
	}
735
	return true;
736 737
}

738 739
/*
 * The pagefault handler calls here because it is out of memory, so kill a
740 741
 * memory-hogging task.  If any populated zone has ZONE_OOM_LOCKED set, a
 * parallel oom killing is already in progress so do nothing.
742 743 744
 */
void pagefault_out_of_memory(void)
{
745 746 747 748 749 750 751
	struct oom_control oc = {
		.zonelist = NULL,
		.nodemask = NULL,
		.gfp_mask = 0,
		.order = 0,
	};

752
	if (mem_cgroup_oom_synchronize(true))
753
		return;
754

755 756
	if (!mutex_trylock(&oom_lock))
		return;
757

758
	if (!out_of_memory(&oc)) {
759 760 761 762 763 764 765
		/*
		 * There shouldn't be any user tasks runnable while the
		 * OOM killer is disabled, so the current task has to
		 * be a racing OOM victim for which oom_killer_disable()
		 * is waiting for.
		 */
		WARN_ON(test_thread_flag(TIF_MEMDIE));
766
	}
767 768

	mutex_unlock(&oom_lock);
769
}