oom_kill.c 19.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8
/*
 *  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...
 *
 *  The routines in this file are used to kill a process when
P
Paul Jackson 已提交
9 10
 *  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
11 12 13 14 15 16 17
 *
 *  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.
 */

18
#include <linux/oom.h>
L
Linus Torvalds 已提交
19
#include <linux/mm.h>
A
Alexey Dobriyan 已提交
20
#include <linux/err.h>
21
#include <linux/gfp.h>
L
Linus Torvalds 已提交
22 23 24 25
#include <linux/sched.h>
#include <linux/swap.h>
#include <linux/timex.h>
#include <linux/jiffies.h>
26
#include <linux/cpuset.h>
27 28
#include <linux/module.h>
#include <linux/notifier.h>
29
#include <linux/memcontrol.h>
30
#include <linux/mempolicy.h>
31
#include <linux/security.h>
L
Linus Torvalds 已提交
32

33
int sysctl_panic_on_oom;
34
int sysctl_oom_kill_allocating_task;
35
int sysctl_oom_dump_tasks;
D
David Rientjes 已提交
36
static DEFINE_SPINLOCK(zone_scan_lock);
L
Linus Torvalds 已提交
37 38
/* #define DEBUG */

39 40 41 42 43 44 45 46 47
#ifdef CONFIG_NUMA
/**
 * has_intersects_mems_allowed() - check task eligiblity for kill
 * @tsk: task struct of which task to consider
 * @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.
48
 */
49 50
static bool has_intersects_mems_allowed(struct task_struct *tsk,
					const nodemask_t *mask)
51
{
52
	struct task_struct *start = tsk;
53 54

	do {
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
		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.
			 */
			if (mempolicy_nodemask_intersects(tsk, mask))
				return true;
		} else {
			/*
			 * This is not a mempolicy constrained oom, so only
			 * check the mems of tsk's cpuset.
			 */
			if (cpuset_mems_allowed_intersects(current, tsk))
				return true;
		}
		tsk = next_thread(tsk);
	} while (tsk != start);
	return false;
}
#else
static bool has_intersects_mems_allowed(struct task_struct *tsk,
					const nodemask_t *mask)
{
	return true;
81
}
82
#endif /* CONFIG_NUMA */
83

84 85 86 87 88 89
/*
 * 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.
 */
90 91 92 93 94 95 96 97 98 99 100 101 102 103
static struct task_struct *find_lock_task_mm(struct task_struct *p)
{
	struct task_struct *t = p;

	do {
		task_lock(t);
		if (likely(t->mm))
			return t;
		task_unlock(t);
	} while_each_thread(p, t);

	return NULL;
}

L
Linus Torvalds 已提交
104
/**
105
 * badness - calculate a numeric value for how bad this task has been
L
Linus Torvalds 已提交
106
 * @p: task struct of which task we should calculate
P
Paul Jackson 已提交
107
 * @uptime: current uptime in seconds
L
Linus Torvalds 已提交
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122
 *
 * The formula used is relatively simple and documented inline in the
 * function. The main rationale is that we want to select a good task
 * to kill when we run out of memory.
 *
 * Good in this context means that:
 * 1) we lose the minimum amount of work done
 * 2) we recover a large amount of memory
 * 3) we don't kill anything innocent of eating tons of memory
 * 4) we want to kill the minimum amount of processes (one)
 * 5) we try to kill the process the user expects us to kill, this
 *    algorithm has been meticulously tuned to meet the principle
 *    of least surprise ... (be careful when you change it)
 */

123
unsigned long badness(struct task_struct *p, unsigned long uptime)
L
Linus Torvalds 已提交
124
{
125
	unsigned long points, cpu_time, run_time;
A
Andrew Morton 已提交
126
	struct task_struct *child;
127
	struct task_struct *c, *t;
128
	int oom_adj = p->signal->oom_adj;
129 130 131
	struct task_cputime task_time;
	unsigned long utime;
	unsigned long stime;
132 133 134

	if (oom_adj == OOM_DISABLE)
		return 0;
L
Linus Torvalds 已提交
135

136 137
	p = find_lock_task_mm(p);
	if (!p)
L
Linus Torvalds 已提交
138 139 140 141 142
		return 0;

	/*
	 * The memory size of the process is the basis for the badness.
	 */
143
	points = p->mm->total_vm;
A
Andrew Morton 已提交
144
	task_unlock(p);
L
Linus Torvalds 已提交
145

146 147 148
	/*
	 * swapoff can easily use up all memory, so kill those first.
	 */
H
Hugh Dickins 已提交
149
	if (p->flags & PF_OOM_ORIGIN)
150 151
		return ULONG_MAX;

L
Linus Torvalds 已提交
152 153
	/*
	 * Processes which fork a lot of child processes are likely
154
	 * a good choice. We add half the vmsize of the children if they
L
Linus Torvalds 已提交
155
	 * have an own mm. This prevents forking servers to flood the
156 157 158
	 * machine with an endless amount of children. In case a single
	 * child is eating the vast majority of memory, adding only half
	 * to the parents will make the child our kill candidate of choice.
L
Linus Torvalds 已提交
159
	 */
160 161 162 163 164 165 166 167 168 169 170
	t = p;
	do {
		list_for_each_entry(c, &t->children, sibling) {
			child = find_lock_task_mm(c);
			if (child) {
				if (child->mm != p->mm)
					points += child->mm->total_vm/2 + 1;
				task_unlock(child);
			}
		}
	} while_each_thread(p, t);
L
Linus Torvalds 已提交
171 172 173 174 175 176

	/*
	 * CPU time is in tens of seconds and run time is in thousands
         * of seconds. There is no particular reason for this other than
         * that it turned out to work very well in practice.
	 */
177 178 179 180 181
	thread_group_cputime(p, &task_time);
	utime = cputime_to_jiffies(task_time.utime);
	stime = cputime_to_jiffies(task_time.stime);
	cpu_time = (utime + stime) >> (SHIFT_HZ + 3);

L
Linus Torvalds 已提交
182 183 184 185 186 187

	if (uptime >= p->start_time.tv_sec)
		run_time = (uptime - p->start_time.tv_sec) >> 10;
	else
		run_time = 0;

188 189 190 191
	if (cpu_time)
		points /= int_sqrt(cpu_time);
	if (run_time)
		points /= int_sqrt(int_sqrt(run_time));
L
Linus Torvalds 已提交
192 193 194 195 196 197 198 199 200 201 202 203

	/*
	 * Niced processes are most likely less important, so double
	 * their badness points.
	 */
	if (task_nice(p) > 0)
		points *= 2;

	/*
	 * Superuser processes are usually more important, so we make it
	 * less likely that we kill those.
	 */
204 205
	if (has_capability_noaudit(p, CAP_SYS_ADMIN) ||
	    has_capability_noaudit(p, CAP_SYS_RESOURCE))
L
Linus Torvalds 已提交
206 207 208 209 210 211 212 213
		points /= 4;

	/*
	 * We don't want to kill a process with direct hardware access.
	 * Not only could that mess up the hardware, but usually users
	 * tend to only have this flag set on applications they think
	 * of as important.
	 */
214
	if (has_capability_noaudit(p, CAP_SYS_RAWIO))
L
Linus Torvalds 已提交
215 216 217
		points /= 4;

	/*
218
	 * Adjust the score by oom_adj.
L
Linus Torvalds 已提交
219
	 */
220 221
	if (oom_adj) {
		if (oom_adj > 0) {
222 223
			if (!points)
				points = 1;
224
			points <<= oom_adj;
225
		} else
226
			points >>= -(oom_adj);
L
Linus Torvalds 已提交
227 228 229
	}

#ifdef DEBUG
230
	printk(KERN_DEBUG "OOMkill: task %d (%s) got %lu points\n",
L
Linus Torvalds 已提交
231 232 233 234 235
	p->pid, p->comm, points);
#endif
	return points;
}

236 237 238 239
/*
 * Determine the type of allocation constraint.
 */
#ifdef CONFIG_NUMA
240 241 242
static enum oom_constraint constrained_alloc(struct zonelist *zonelist,
				    gfp_t gfp_mask, nodemask_t *nodemask)
{
243
	struct zone *zone;
244
	struct zoneref *z;
245
	enum zone_type high_zoneidx = gfp_zone(gfp_mask);
246

247 248 249 250 251 252 253
	/*
	 * 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.
	 */
	if (gfp_mask & __GFP_THISNODE)
		return CONSTRAINT_NONE;
254

255 256 257 258 259 260 261
	/*
	 * The nodemask here is a nodemask passed to alloc_pages(). Now,
	 * cpuset doesn't use this nodemask for its hardwall/softwall/hierarchy
	 * feature. mempolicy is an only user of nodemask here.
	 * check mempolicy's nodemask contains all N_HIGH_MEMORY
	 */
	if (nodemask && !nodes_subset(node_states[N_HIGH_MEMORY], *nodemask))
262
		return CONSTRAINT_MEMORY_POLICY;
263 264 265 266 267 268

	/* Check this allocation failure is caused by cpuset's wall function */
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
			high_zoneidx, nodemask)
		if (!cpuset_zone_allowed_softwall(zone, gfp_mask))
			return CONSTRAINT_CPUSET;
269 270 271

	return CONSTRAINT_NONE;
}
272 273 274 275 276 277 278
#else
static enum oom_constraint constrained_alloc(struct zonelist *zonelist,
				gfp_t gfp_mask, nodemask_t *nodemask)
{
	return CONSTRAINT_NONE;
}
#endif
279

L
Linus Torvalds 已提交
280 281 282 283 284 285
/*
 * Simple selection loop. We chose the process with the highest
 * number of 'points'. We expect the caller will lock the tasklist.
 *
 * (not docbooked, we don't want this one cluttering up the manual)
 */
286
static struct task_struct *select_bad_process(unsigned long *ppoints,
287 288
		struct mem_cgroup *mem, enum oom_constraint constraint,
		const nodemask_t *mask)
L
Linus Torvalds 已提交
289
{
290
	struct task_struct *p;
L
Linus Torvalds 已提交
291 292
	struct task_struct *chosen = NULL;
	struct timespec uptime;
293
	*ppoints = 0;
L
Linus Torvalds 已提交
294 295

	do_posix_clock_monotonic_gettime(&uptime);
296
	for_each_process(p) {
P
Paul Jackson 已提交
297 298
		unsigned long points;

299 300
		/* skip the init task and kthreads */
		if (is_global_init(p) || (p->flags & PF_KTHREAD))
P
Paul Jackson 已提交
301
			continue;
302 303
		if (mem && !task_in_mem_cgroup(p, mem))
			continue;
304 305 306
		if (!has_intersects_mems_allowed(p,
				constraint == CONSTRAINT_MEMORY_POLICY ? mask :
									 NULL))
307
			continue;
308

309 310 311 312 313 314 315 316 317 318 319 320
		/*
		 * This task already has access to memory reserves and is
		 * being killed. Don't allow any other task access to the
		 * memory reserve.
		 *
		 * Note: this may have a chance of deadlock if it gets
		 * blocked waiting for another task which itself is waiting
		 * for memory. Is there a better alternative?
		 */
		if (test_tsk_thread_flag(p, TIF_MEMDIE))
			return ERR_PTR(-1UL);

P
Paul Jackson 已提交
321
		/*
322
		 * This is in the process of releasing memory so wait for it
P
Paul Jackson 已提交
323
		 * to finish before killing some other task by mistake.
324 325 326 327 328
		 *
		 * However, if p is the current task, we allow the 'kill' to
		 * go ahead if it is exiting: this will simply set TIF_MEMDIE,
		 * which will allow it to gain access to memory reserves in
		 * the process of exiting and releasing its resources.
329
		 * Otherwise we could get an easy OOM deadlock.
P
Paul Jackson 已提交
330
		 */
331
		if ((p->flags & PF_EXITING) && p->mm) {
332 333 334
			if (p != current)
				return ERR_PTR(-1UL);

335 336
			chosen = p;
			*ppoints = ULONG_MAX;
337
		}
338

339
		if (p->signal->oom_adj == OOM_DISABLE)
340 341
			continue;

342
		points = badness(p, uptime.tv_sec);
343
		if (points > *ppoints || !chosen) {
P
Paul Jackson 已提交
344
			chosen = p;
345
			*ppoints = points;
L
Linus Torvalds 已提交
346
		}
347
	}
348

L
Linus Torvalds 已提交
349 350 351
	return chosen;
}

352
/**
R
Randy Dunlap 已提交
353
 * dump_tasks - dump current memory state of all system tasks
354
 * @mem: current's memory controller, if constrained
R
Randy Dunlap 已提交
355
 *
356 357 358 359 360 361 362 363 364 365 366
 * Dumps the current memory state of all system tasks, excluding kernel threads.
 * State information includes task's pid, uid, tgid, vm size, rss, cpu, oom_adj
 * score, and name.
 *
 * If the actual is non-NULL, only tasks that are a member of the mem_cgroup are
 * shown.
 *
 * Call with tasklist_lock read-locked.
 */
static void dump_tasks(const struct mem_cgroup *mem)
{
367 368
	struct task_struct *p;
	struct task_struct *task;
369 370 371

	printk(KERN_INFO "[ pid ]   uid  tgid total_vm      rss cpu oom_adj "
	       "name\n");
372 373
	for_each_process(p) {
		if (p->flags & PF_KTHREAD)
374
			continue;
375
		if (mem && !task_in_mem_cgroup(p, mem))
376
			continue;
377

378 379
		task = find_lock_task_mm(p);
		if (!task) {
380
			/*
381 382
			 * This is a kthread or all of p's threads have already
			 * detached their mm's.  There's no need to report
383
			 * them; they can't be oom killed anyway.
384 385 386
			 */
			continue;
		}
387

388
		printk(KERN_INFO "[%5d] %5d %5d %8lu %8lu %3u     %3d %s\n",
389 390
		       task->pid, __task_cred(task)->uid, task->tgid,
		       task->mm->total_vm, get_mm_rss(task->mm),
391
		       task_cpu(task), task->signal->oom_adj, task->comm);
392 393
		task_unlock(task);
	}
394 395
}

396 397
static void dump_header(struct task_struct *p, gfp_t gfp_mask, int order,
							struct mem_cgroup *mem)
398
{
399
	task_lock(current);
400 401 402 403 404 405
	pr_warning("%s invoked oom-killer: gfp_mask=0x%x, order=%d, "
		"oom_adj=%d\n",
		current->comm, gfp_mask, order, current->signal->oom_adj);
	cpuset_print_task_mems_allowed(current);
	task_unlock(current);
	dump_stack();
406
	mem_cgroup_print_oom_info(mem, p);
407 408 409 410 411
	show_mem();
	if (sysctl_oom_dump_tasks)
		dump_tasks(mem);
}

412 413
#define K(x) ((x) << (PAGE_SHIFT-10))

R
Randy Dunlap 已提交
414
/*
415 416 417
 * Send SIGKILL to the selected  process irrespective of  CAP_SYS_RAW_IO
 * flag though it's unlikely that  we select a process with CAP_SYS_RAW_IO
 * set.
L
Linus Torvalds 已提交
418
 */
N
Nick Piggin 已提交
419
static void __oom_kill_task(struct task_struct *p, int verbose)
L
Linus Torvalds 已提交
420
{
421
	if (is_global_init(p)) {
L
Linus Torvalds 已提交
422 423 424 425 426
		WARN_ON(1);
		printk(KERN_WARNING "tried to kill init!\n");
		return;
	}

427 428
	p = find_lock_task_mm(p);
	if (!p)
L
Linus Torvalds 已提交
429
		return;
430

N
Nick Piggin 已提交
431
	if (verbose)
432 433 434 435
		printk(KERN_ERR "Killed process %d (%s) "
		       "vsz:%lukB, anon-rss:%lukB, file-rss:%lukB\n",
		       task_pid_nr(p), p->comm,
		       K(p->mm->total_vm),
K
KAMEZAWA Hiroyuki 已提交
436 437
		       K(get_mm_counter(p->mm, MM_ANONPAGES)),
		       K(get_mm_counter(p->mm, MM_FILEPAGES)));
438
	task_unlock(p);
L
Linus Torvalds 已提交
439 440 441 442 443 444

	/*
	 * We give our sacrificial lamb high priority and access to
	 * all the memory it needs. That way it should be able to
	 * exit() and clear out its resources quickly...
	 */
P
Peter Zijlstra 已提交
445
	p->rt.time_slice = HZ;
L
Linus Torvalds 已提交
446 447 448 449 450
	set_tsk_thread_flag(p, TIF_MEMDIE);

	force_sig(SIGKILL, p);
}

N
Nick Piggin 已提交
451
static int oom_kill_task(struct task_struct *p)
L
Linus Torvalds 已提交
452
{
453 454 455 456 457 458 459 460
	/* WARNING: mm may not be dereferenced since we did not obtain its
	 * value from get_task_mm(p).  This is OK since all we need to do is
	 * compare mm to q->mm below.
	 *
	 * Furthermore, even if mm contains a non-NULL value, p->mm may
	 * change to NULL at any time since we do not hold task_lock(p).
	 * However, this is of no concern to us.
	 */
461
	if (!p->mm || p->signal->oom_adj == OOM_DISABLE)
462
		return 1;
463

N
Nick Piggin 已提交
464
	__oom_kill_task(p, 1);
465

466
	return 0;
L
Linus Torvalds 已提交
467 468
}

469
static int oom_kill_process(struct task_struct *p, gfp_t gfp_mask, int order,
470 471
			    unsigned long points, struct mem_cgroup *mem,
			    const char *message)
L
Linus Torvalds 已提交
472
{
473 474
	struct task_struct *victim = p;
	struct task_struct *child;
475
	struct task_struct *t = p;
476 477
	unsigned long victim_points = 0;
	struct timespec uptime;
L
Linus Torvalds 已提交
478

479
	if (printk_ratelimit())
480
		dump_header(p, gfp_mask, order, mem);
481

482 483 484 485
	/*
	 * 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
	 */
486
	if (p->flags & PF_EXITING) {
487
		set_tsk_thread_flag(p, TIF_MEMDIE);
488 489 490
		return 0;
	}

491 492 493 494
	task_lock(p);
	pr_err("%s: Kill process %d (%s) score %lu or sacrifice child\n",
		message, task_pid_nr(p), p->comm, points);
	task_unlock(p);
N
Nick Piggin 已提交
495

496 497 498 499 500 501 502
	/*
	 * If any of p's children has a different mm and is eligible for kill,
	 * the one with the highest badness() score is sacrificed for its
	 * parent.  This attempts to lose the minimal amount of work done while
	 * still freeing memory.
	 */
	do_posix_clock_monotonic_gettime(&uptime);
503
	do {
504 505 506 507
		list_for_each_entry(child, &t->children, sibling) {
			unsigned long child_points;

			if (child->mm == p->mm)
508
				continue;
509
			if (mem && !task_in_mem_cgroup(child, mem))
510
				continue;
511 512 513 514 515 516 517

			/* badness() returns 0 if the thread is unkillable */
			child_points = badness(child, uptime.tv_sec);
			if (child_points > victim_points) {
				victim = child;
				victim_points = child_points;
			}
518 519 520
		}
	} while_each_thread(p, t);

521
	return oom_kill_task(victim);
L
Linus Torvalds 已提交
522 523
}

524
#ifdef CONFIG_CGROUP_MEM_RES_CTLR
525 526 527 528 529
void mem_cgroup_out_of_memory(struct mem_cgroup *mem, gfp_t gfp_mask)
{
	unsigned long points = 0;
	struct task_struct *p;

530 531
	if (sysctl_panic_on_oom == 2)
		panic("out of memory(memcg). panic_on_oom is selected.\n");
L
Li Zefan 已提交
532
	read_lock(&tasklist_lock);
533
retry:
534
	p = select_bad_process(&points, mem, CONSTRAINT_NONE, NULL);
535
	if (!p || PTR_ERR(p) == -1UL)
536 537
		goto out;

538
	if (oom_kill_process(p, gfp_mask, 0, points, mem,
539 540 541
				"Memory cgroup out of memory"))
		goto retry;
out:
L
Li Zefan 已提交
542
	read_unlock(&tasklist_lock);
543 544 545
}
#endif

546 547 548 549 550 551 552 553 554 555 556 557 558 559
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);

D
David Rientjes 已提交
560 561 562 563 564
/*
 * Try to acquire the OOM killer lock for the zones in zonelist.  Returns zero
 * if a parallel OOM killing is already taking place that includes a zone in
 * the zonelist.  Otherwise, locks all zones in the zonelist and returns 1.
 */
565
int try_set_zone_oom(struct zonelist *zonelist, gfp_t gfp_mask)
D
David Rientjes 已提交
566
{
567 568
	struct zoneref *z;
	struct zone *zone;
D
David Rientjes 已提交
569 570
	int ret = 1;

D
David Rientjes 已提交
571
	spin_lock(&zone_scan_lock);
572 573
	for_each_zone_zonelist(zone, z, zonelist, gfp_zone(gfp_mask)) {
		if (zone_is_oom_locked(zone)) {
D
David Rientjes 已提交
574 575 576
			ret = 0;
			goto out;
		}
577 578 579 580
	}

	for_each_zone_zonelist(zone, z, zonelist, gfp_zone(gfp_mask)) {
		/*
D
David Rientjes 已提交
581
		 * Lock each zone in the zonelist under zone_scan_lock so a
582 583 584 585 586
		 * parallel invocation of try_set_zone_oom() doesn't succeed
		 * when it shouldn't.
		 */
		zone_set_flag(zone, ZONE_OOM_LOCKED);
	}
D
David Rientjes 已提交
587 588

out:
D
David Rientjes 已提交
589
	spin_unlock(&zone_scan_lock);
D
David Rientjes 已提交
590 591 592 593 594 595 596 597
	return ret;
}

/*
 * Clears the ZONE_OOM_LOCKED flag for all zones in the zonelist so that failed
 * allocation attempts with zonelists containing them may now recall the OOM
 * killer, if necessary.
 */
598
void clear_zonelist_oom(struct zonelist *zonelist, gfp_t gfp_mask)
D
David Rientjes 已提交
599
{
600 601
	struct zoneref *z;
	struct zone *zone;
D
David Rientjes 已提交
602

D
David Rientjes 已提交
603
	spin_lock(&zone_scan_lock);
604 605 606
	for_each_zone_zonelist(zone, z, zonelist, gfp_zone(gfp_mask)) {
		zone_clear_flag(zone, ZONE_OOM_LOCKED);
	}
D
David Rientjes 已提交
607
	spin_unlock(&zone_scan_lock);
D
David Rientjes 已提交
608 609
}

610 611 612
/*
 * Must be called with tasklist_lock held for read.
 */
613 614
static void __out_of_memory(gfp_t gfp_mask, int order,
			enum oom_constraint constraint, const nodemask_t *mask)
615
{
616 617
	struct task_struct *p;
	unsigned long points;
618

619 620 621
	if (sysctl_oom_kill_allocating_task)
		if (!oom_kill_process(current, gfp_mask, order, 0, NULL,
				"Out of memory (oom_kill_allocating_task)"))
622
			return;
623 624 625 626 627
retry:
	/*
	 * Rambo mode: Shoot down a process and hope it solves whatever
	 * issues we may have.
	 */
628
	p = select_bad_process(&points, NULL, constraint, mask);
629

630 631
	if (PTR_ERR(p) == -1UL)
		return;
632

633 634 635
	/* Found nothing?!?! Either we hang forever, or we panic. */
	if (!p) {
		read_unlock(&tasklist_lock);
636
		dump_header(NULL, gfp_mask, order, NULL);
637
		panic("Out of memory and no killable processes...\n");
638
	}
639 640 641 642

	if (oom_kill_process(p, gfp_mask, order, points, NULL,
			     "Out of memory"))
		goto retry;
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
}

/*
 * pagefault handler calls into here because it is out of memory but
 * doesn't know exactly how or why.
 */
void pagefault_out_of_memory(void)
{
	unsigned long freed = 0;

	blocking_notifier_call_chain(&oom_notify_list, 0, &freed);
	if (freed > 0)
		/* Got some memory back in the last second. */
		return;

	if (sysctl_panic_on_oom)
		panic("out of memory from page fault. panic_on_oom is selected.\n");

	read_lock(&tasklist_lock);
662 663
	/* unknown gfp_mask and order */
	__out_of_memory(0, 0, CONSTRAINT_NONE, NULL);
664 665 666 667 668 669 670 671 672 673
	read_unlock(&tasklist_lock);

	/*
	 * Give "p" a good chance of killing itself before we
	 * retry to allocate memory.
	 */
	if (!test_thread_flag(TIF_MEMDIE))
		schedule_timeout_uninterruptible(1);
}

L
Linus Torvalds 已提交
674
/**
675
 * out_of_memory - kill the "best" process when we run out of memory
R
Randy Dunlap 已提交
676 677 678
 * @zonelist: zonelist pointer
 * @gfp_mask: memory allocation flags
 * @order: amount of memory being requested as a power of 2
679
 * @nodemask: nodemask passed to page allocator
L
Linus Torvalds 已提交
680 681 682 683 684 685
 *
 * 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.
 */
686 687
void out_of_memory(struct zonelist *zonelist, gfp_t gfp_mask,
		int order, nodemask_t *nodemask)
L
Linus Torvalds 已提交
688
{
689
	unsigned long freed = 0;
D
David Rientjes 已提交
690
	enum oom_constraint constraint;
691 692 693 694 695

	blocking_notifier_call_chain(&oom_notify_list, 0, &freed);
	if (freed > 0)
		/* Got some memory back in the last second. */
		return;
L
Linus Torvalds 已提交
696

697 698 699 700 701 702 703 704 705 706
	/*
	 * If current has a pending SIGKILL, then automatically select it.  The
	 * goal is to allow it to allocate so that it may quickly exit and free
	 * its memory.
	 */
	if (fatal_signal_pending(current)) {
		set_thread_flag(TIF_MEMDIE);
		return;
	}

707
	if (sysctl_panic_on_oom == 2) {
708
		dump_header(NULL, gfp_mask, order, NULL);
709
		panic("out of memory. Compulsory panic_on_oom is selected.\n");
710
	}
711

712 713 714 715
	/*
	 * Check if there were limitations on the allocation (only relevant for
	 * NUMA) that may require different handling.
	 */
716
	constraint = constrained_alloc(zonelist, gfp_mask, nodemask);
D
David Rientjes 已提交
717
	read_lock(&tasklist_lock);
718 719 720 721 722 723 724
	if (unlikely(sysctl_panic_on_oom)) {
		/*
		 * panic_on_oom only affects CONSTRAINT_NONE, the kernel
		 * should not panic for cpuset or mempolicy induced memory
		 * failures.
		 */
		if (constraint == CONSTRAINT_NONE) {
725
			dump_header(NULL, gfp_mask, order, NULL);
726 727
			read_unlock(&tasklist_lock);
			panic("Out of memory: panic_on_oom is enabled\n");
728
		}
729
	}
730
	__out_of_memory(gfp_mask, order, constraint, nodemask);
731
	read_unlock(&tasklist_lock);
L
Linus Torvalds 已提交
732 733 734

	/*
	 * Give "p" a good chance of killing itself before we
735
	 * retry to allocate memory unless "p" is current
L
Linus Torvalds 已提交
736
	 */
737
	if (!test_thread_flag(TIF_MEMDIE))
738
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
739
}