cpuset.c 73.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6
/*
 *  kernel/cpuset.c
 *
 *  Processor and Memory placement constraints for sets of tasks.
 *
 *  Copyright (C) 2003 BULL SA.
P
Paul Jackson 已提交
7
 *  Copyright (C) 2004-2007 Silicon Graphics, Inc.
8
 *  Copyright (C) 2006 Google, Inc
L
Linus Torvalds 已提交
9 10 11 12
 *
 *  Portions derived from Patrick Mochel's sysfs code.
 *  sysfs is Copyright (c) 2001-3 Patrick Mochel
 *
13
 *  2003-10-10 Written by Simon Derr.
L
Linus Torvalds 已提交
14
 *  2003-10-22 Updates by Stephen Hemminger.
15
 *  2004 May-July Rework by Paul Jackson.
16
 *  2006 Rework by Paul Menage to use generic cgroups
17 18
 *  2008 Rework of the scheduler domains and CPU hotplug handling
 *       by Max Krasnyansky
L
Linus Torvalds 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
 *
 *  This file is subject to the terms and conditions of the GNU General Public
 *  License.  See the file COPYING in the main directory of the Linux
 *  distribution for more details.
 */

#include <linux/cpu.h>
#include <linux/cpumask.h>
#include <linux/cpuset.h>
#include <linux/err.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/kernel.h>
#include <linux/kmod.h>
#include <linux/list.h>
37
#include <linux/mempolicy.h>
L
Linus Torvalds 已提交
38
#include <linux/mm.h>
39
#include <linux/memory.h>
L
Linus Torvalds 已提交
40 41 42 43 44
#include <linux/module.h>
#include <linux/mount.h>
#include <linux/namei.h>
#include <linux/pagemap.h>
#include <linux/proc_fs.h>
45
#include <linux/rcupdate.h>
L
Linus Torvalds 已提交
46 47
#include <linux/sched.h>
#include <linux/seq_file.h>
48
#include <linux/security.h>
L
Linus Torvalds 已提交
49 50 51 52 53 54 55 56 57 58
#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/time.h>
#include <linux/backing-dev.h>
#include <linux/sort.h>

#include <asm/uaccess.h>
#include <asm/atomic.h>
59
#include <linux/mutex.h>
60 61
#include <linux/workqueue.h>
#include <linux/cgroup.h>
L
Linus Torvalds 已提交
62

63 64 65 66 67 68 69 70
/*
 * Workqueue for cpuset related tasks.
 *
 * Using kevent workqueue may cause deadlock when memory_migrate
 * is set. So we create a separate workqueue thread for cpuset.
 */
static struct workqueue_struct *cpuset_wq;

71 72 73 74 75
/*
 * Tracks how many cpusets are currently defined in system.
 * When there is only one cpuset (the root cpuset) we can
 * short circuit some hooks.
 */
76
int number_of_cpusets __read_mostly;
77

78
/* Forward declare cgroup structures */
79 80 81
struct cgroup_subsys cpuset_subsys;
struct cpuset;

82 83 84 85 86 87 88 89 90
/* See "Frequency meter" comments, below. */

struct fmeter {
	int cnt;		/* unprocessed events count */
	int val;		/* most recent output value */
	time_t time;		/* clock (secs) when val computed */
	spinlock_t lock;	/* guards read or write of above */
};

L
Linus Torvalds 已提交
91
struct cpuset {
92 93
	struct cgroup_subsys_state css;

L
Linus Torvalds 已提交
94
	unsigned long flags;		/* "unsigned long" so bitops work */
95
	cpumask_var_t cpus_allowed;	/* CPUs allowed to tasks in cpuset */
L
Linus Torvalds 已提交
96 97 98 99 100 101 102 103
	nodemask_t mems_allowed;	/* Memory Nodes allowed to tasks */

	struct cpuset *parent;		/* my parent */

	/*
	 * Copy of global cpuset_mems_generation as of the most
	 * recent time this cpuset changed its mems_allowed.
	 */
104 105 106
	int mems_generation;

	struct fmeter fmeter;		/* memory_pressure filter */
P
Paul Jackson 已提交
107 108 109

	/* partition number for rebuild_sched_domains() */
	int pn;
110

111 112 113
	/* for custom sched domain */
	int relax_domain_level;

114 115
	/* used for walking a cpuset heirarchy */
	struct list_head stack_list;
L
Linus Torvalds 已提交
116 117
};

118 119 120 121 122 123 124 125 126 127 128 129 130
/* Retrieve the cpuset for a cgroup */
static inline struct cpuset *cgroup_cs(struct cgroup *cont)
{
	return container_of(cgroup_subsys_state(cont, cpuset_subsys_id),
			    struct cpuset, css);
}

/* Retrieve the cpuset for a task */
static inline struct cpuset *task_cs(struct task_struct *task)
{
	return container_of(task_subsys_state(task, cpuset_subsys_id),
			    struct cpuset, css);
}
131 132 133 134
struct cpuset_hotplug_scanner {
	struct cgroup_scanner scan;
	struct cgroup *to;
};
135

L
Linus Torvalds 已提交
136 137 138 139
/* bits in struct cpuset flags field */
typedef enum {
	CS_CPU_EXCLUSIVE,
	CS_MEM_EXCLUSIVE,
140
	CS_MEM_HARDWALL,
141
	CS_MEMORY_MIGRATE,
P
Paul Jackson 已提交
142
	CS_SCHED_LOAD_BALANCE,
143 144
	CS_SPREAD_PAGE,
	CS_SPREAD_SLAB,
L
Linus Torvalds 已提交
145 146 147 148 149
} cpuset_flagbits_t;

/* convenient tests for these bits */
static inline int is_cpu_exclusive(const struct cpuset *cs)
{
150
	return test_bit(CS_CPU_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
151 152 153 154
}

static inline int is_mem_exclusive(const struct cpuset *cs)
{
155
	return test_bit(CS_MEM_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
156 157
}

158 159 160 161 162
static inline int is_mem_hardwall(const struct cpuset *cs)
{
	return test_bit(CS_MEM_HARDWALL, &cs->flags);
}

P
Paul Jackson 已提交
163 164 165 166 167
static inline int is_sched_load_balance(const struct cpuset *cs)
{
	return test_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
}

168 169
static inline int is_memory_migrate(const struct cpuset *cs)
{
170
	return test_bit(CS_MEMORY_MIGRATE, &cs->flags);
171 172
}

173 174 175 176 177 178 179 180 181 182
static inline int is_spread_page(const struct cpuset *cs)
{
	return test_bit(CS_SPREAD_PAGE, &cs->flags);
}

static inline int is_spread_slab(const struct cpuset *cs)
{
	return test_bit(CS_SPREAD_SLAB, &cs->flags);
}

L
Linus Torvalds 已提交
183
/*
184
 * Increment this integer everytime any cpuset changes its
L
Linus Torvalds 已提交
185 186 187 188
 * mems_allowed value.  Users of cpusets can track this generation
 * number, and avoid having to lock and reload mems_allowed unless
 * the cpuset they're using changes generation.
 *
189
 * A single, global generation is needed because cpuset_attach_task() could
L
Linus Torvalds 已提交
190 191 192 193
 * reattach a task to a different cpuset, which must not have its
 * generation numbers aliased with those of that tasks previous cpuset.
 *
 * Generations are needed for mems_allowed because one task cannot
194
 * modify another's memory placement.  So we must enable every task,
L
Linus Torvalds 已提交
195 196 197
 * on every visit to __alloc_pages(), to efficiently check whether
 * its current->cpuset->mems_allowed has changed, requiring an update
 * of its current->mems_allowed.
198
 *
199
 * Since writes to cpuset_mems_generation are guarded by the cgroup lock
200
 * there is no need to mark it atomic.
L
Linus Torvalds 已提交
201
 */
202
static int cpuset_mems_generation;
L
Linus Torvalds 已提交
203 204 205 206 207 208

static struct cpuset top_cpuset = {
	.flags = ((1 << CS_CPU_EXCLUSIVE) | (1 << CS_MEM_EXCLUSIVE)),
};

/*
209 210 211 212 213 214 215
 * There are two global mutexes guarding cpuset structures.  The first
 * is the main control groups cgroup_mutex, accessed via
 * cgroup_lock()/cgroup_unlock().  The second is the cpuset-specific
 * callback_mutex, below. They can nest.  It is ok to first take
 * cgroup_mutex, then nest callback_mutex.  We also require taking
 * task_lock() when dereferencing a task's cpuset pointer.  See "The
 * task_lock() exception", at the end of this comment.
216
 *
217
 * A task must hold both mutexes to modify cpusets.  If a task
218
 * holds cgroup_mutex, then it blocks others wanting that mutex,
219
 * ensuring that it is the only task able to also acquire callback_mutex
220 221
 * and be able to modify cpusets.  It can perform various checks on
 * the cpuset structure first, knowing nothing will change.  It can
222
 * also allocate memory while just holding cgroup_mutex.  While it is
223
 * performing these checks, various callback routines can briefly
224 225
 * acquire callback_mutex to query cpusets.  Once it is ready to make
 * the changes, it takes callback_mutex, blocking everyone else.
226 227
 *
 * Calls to the kernel memory allocator can not be made while holding
228
 * callback_mutex, as that would risk double tripping on callback_mutex
229 230 231
 * from one of the callbacks into the cpuset code from within
 * __alloc_pages().
 *
232
 * If a task is only holding callback_mutex, then it has read-only
233 234 235 236 237
 * access to cpusets.
 *
 * The task_struct fields mems_allowed and mems_generation may only
 * be accessed in the context of that task, so require no locks.
 *
238
 * The cpuset_common_file_read() handlers only hold callback_mutex across
239 240 241
 * small pieces of code, such as when reading out possibly multi-word
 * cpumasks and nodemasks.
 *
242 243
 * Accessing a task's cpuset should be done in accordance with the
 * guidelines for accessing subsystem state in kernel/cgroup.c
L
Linus Torvalds 已提交
244 245
 */

246
static DEFINE_MUTEX(callback_mutex);
247

248 249 250 251 252 253 254 255 256 257 258
/*
 * cpuset_buffer_lock protects both the cpuset_name and cpuset_nodelist
 * buffers.  They are statically allocated to prevent using excess stack
 * when calling cpuset_print_task_mems_allowed().
 */
#define CPUSET_NAME_LEN		(128)
#define	CPUSET_NODELIST_LEN	(256)
static char cpuset_name[CPUSET_NAME_LEN];
static char cpuset_nodelist[CPUSET_NODELIST_LEN];
static DEFINE_SPINLOCK(cpuset_buffer_lock);

259 260
/*
 * This is ugly, but preserves the userspace API for existing cpuset
261
 * users. If someone tries to mount the "cpuset" filesystem, we
262 263
 * silently switch it to mount "cgroup" instead
 */
264 265 266
static int cpuset_get_sb(struct file_system_type *fs_type,
			 int flags, const char *unused_dev_name,
			 void *data, struct vfsmount *mnt)
L
Linus Torvalds 已提交
267
{
268 269 270 271 272 273 274 275 276 277 278
	struct file_system_type *cgroup_fs = get_fs_type("cgroup");
	int ret = -ENODEV;
	if (cgroup_fs) {
		char mountopts[] =
			"cpuset,noprefix,"
			"release_agent=/sbin/cpuset_release_agent";
		ret = cgroup_fs->get_sb(cgroup_fs, flags,
					   unused_dev_name, mountopts, mnt);
		put_filesystem(cgroup_fs);
	}
	return ret;
L
Linus Torvalds 已提交
279 280 281 282 283 284 285 286
}

static struct file_system_type cpuset_fs_type = {
	.name = "cpuset",
	.get_sb = cpuset_get_sb,
};

/*
287
 * Return in pmask the portion of a cpusets's cpus_allowed that
L
Linus Torvalds 已提交
288 289 290 291 292 293 294 295 296
 * are online.  If none are online, walk up the cpuset hierarchy
 * until we find one that does have some online cpus.  If we get
 * all the way to the top and still haven't found any online cpus,
 * return cpu_online_map.  Or if passed a NULL cs from an exit'ing
 * task, return cpu_online_map.
 *
 * One way or another, we guarantee to return some non-empty subset
 * of cpu_online_map.
 *
297
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
298 299
 */

300 301
static void guarantee_online_cpus(const struct cpuset *cs,
				  struct cpumask *pmask)
L
Linus Torvalds 已提交
302
{
303
	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
L
Linus Torvalds 已提交
304 305
		cs = cs->parent;
	if (cs)
306
		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
L
Linus Torvalds 已提交
307
	else
308 309
		cpumask_copy(pmask, cpu_online_mask);
	BUG_ON(!cpumask_intersects(pmask, cpu_online_mask));
L
Linus Torvalds 已提交
310 311 312 313
}

/*
 * Return in *pmask the portion of a cpusets's mems_allowed that
314 315 316 317
 * are online, with memory.  If none are online with memory, walk
 * up the cpuset hierarchy until we find one that does have some
 * online mems.  If we get all the way to the top and still haven't
 * found any online mems, return node_states[N_HIGH_MEMORY].
L
Linus Torvalds 已提交
318 319
 *
 * One way or another, we guarantee to return some non-empty subset
320
 * of node_states[N_HIGH_MEMORY].
L
Linus Torvalds 已提交
321
 *
322
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
323 324 325 326
 */

static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
{
327 328
	while (cs && !nodes_intersects(cs->mems_allowed,
					node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
329 330
		cs = cs->parent;
	if (cs)
331 332
		nodes_and(*pmask, cs->mems_allowed,
					node_states[N_HIGH_MEMORY]);
L
Linus Torvalds 已提交
333
	else
334 335
		*pmask = node_states[N_HIGH_MEMORY];
	BUG_ON(!nodes_intersects(*pmask, node_states[N_HIGH_MEMORY]));
L
Linus Torvalds 已提交
336 337
}

338 339 340 341 342 343
/**
 * cpuset_update_task_memory_state - update task memory placement
 *
 * If the current tasks cpusets mems_allowed changed behind our
 * backs, update current->mems_allowed, mems_generation and task NUMA
 * mempolicy to the new value.
344
 *
345 346 347 348
 * Task mempolicy is updated by rebinding it relative to the
 * current->cpuset if a task has its memory placement changed.
 * Do not call this routine if in_interrupt().
 *
349
 * Call without callback_mutex or task_lock() held.  May be
350 351
 * called with or without cgroup_mutex held.  Thanks in part to
 * 'the_top_cpuset_hack', the task's cpuset pointer will never
D
David Rientjes 已提交
352 353
 * be NULL.  This routine also might acquire callback_mutex during
 * call.
354
 *
355 356
 * Reading current->cpuset->mems_generation doesn't need task_lock
 * to guard the current->cpuset derefence, because it is guarded
357
 * from concurrent freeing of current->cpuset using RCU.
358 359 360 361 362 363 364 365 366 367 368 369 370 371
 *
 * The rcu_dereference() is technically probably not needed,
 * as I don't actually mind if I see a new cpuset pointer but
 * an old value of mems_generation.  However this really only
 * matters on alpha systems using cpusets heavily.  If I dropped
 * that rcu_dereference(), it would save them a memory barrier.
 * For all other arch's, rcu_dereference is a no-op anyway, and for
 * alpha systems not using cpusets, another planned optimization,
 * avoiding the rcu critical section for tasks in the root cpuset
 * which is statically allocated, so can't vanish, will make this
 * irrelevant.  Better to use RCU as intended, than to engage in
 * some cute trick to save a memory barrier that is impossible to
 * test, for alpha systems using cpusets heavily, which might not
 * even exist.
372 373 374 375 376
 *
 * This routine is needed to update the per-task mems_allowed data,
 * within the tasks context, when it is trying to allocate memory
 * (in various mm/mempolicy.c routines) and notices that some other
 * task has been modifying its cpuset.
L
Linus Torvalds 已提交
377 378
 */

379
void cpuset_update_task_memory_state(void)
L
Linus Torvalds 已提交
380
{
381
	int my_cpusets_mem_gen;
382
	struct task_struct *tsk = current;
383
	struct cpuset *cs;
384

385 386 387
	rcu_read_lock();
	my_cpusets_mem_gen = task_cs(tsk)->mems_generation;
	rcu_read_unlock();
L
Linus Torvalds 已提交
388

389
	if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) {
390
		mutex_lock(&callback_mutex);
391
		task_lock(tsk);
392
		cs = task_cs(tsk); /* Maybe changed when task not locked */
393 394
		guarantee_online_mems(cs, &tsk->mems_allowed);
		tsk->cpuset_mems_generation = cs->mems_generation;
395 396 397 398 399 400 401 402
		if (is_spread_page(cs))
			tsk->flags |= PF_SPREAD_PAGE;
		else
			tsk->flags &= ~PF_SPREAD_PAGE;
		if (is_spread_slab(cs))
			tsk->flags |= PF_SPREAD_SLAB;
		else
			tsk->flags &= ~PF_SPREAD_SLAB;
403
		task_unlock(tsk);
404
		mutex_unlock(&callback_mutex);
405
		mpol_rebind_task(tsk, &tsk->mems_allowed);
L
Linus Torvalds 已提交
406 407 408 409 410 411 412 413
	}
}

/*
 * is_cpuset_subset(p, q) - Is cpuset p a subset of cpuset q?
 *
 * One cpuset is a subset of another if all its allowed CPUs and
 * Memory Nodes are a subset of the other, and its exclusive flags
414
 * are only set if the other's are set.  Call holding cgroup_mutex.
L
Linus Torvalds 已提交
415 416 417 418
 */

static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
{
419
	return	cpumask_subset(p->cpus_allowed, q->cpus_allowed) &&
L
Linus Torvalds 已提交
420 421 422 423 424
		nodes_subset(p->mems_allowed, q->mems_allowed) &&
		is_cpu_exclusive(p) <= is_cpu_exclusive(q) &&
		is_mem_exclusive(p) <= is_mem_exclusive(q);
}

425 426 427 428 429 430
/**
 * alloc_trial_cpuset - allocate a trial cpuset
 * @cs: the cpuset that the trial cpuset duplicates
 */
static struct cpuset *alloc_trial_cpuset(const struct cpuset *cs)
{
431 432 433 434 435 436 437 438 439 440 441 442 443
	struct cpuset *trial;

	trial = kmemdup(cs, sizeof(*cs), GFP_KERNEL);
	if (!trial)
		return NULL;

	if (!alloc_cpumask_var(&trial->cpus_allowed, GFP_KERNEL)) {
		kfree(trial);
		return NULL;
	}
	cpumask_copy(trial->cpus_allowed, cs->cpus_allowed);

	return trial;
444 445 446 447 448 449 450 451
}

/**
 * free_trial_cpuset - free the trial cpuset
 * @trial: the trial cpuset to be freed
 */
static void free_trial_cpuset(struct cpuset *trial)
{
452
	free_cpumask_var(trial->cpus_allowed);
453 454 455
	kfree(trial);
}

L
Linus Torvalds 已提交
456 457 458 459 460 461 462
/*
 * validate_change() - Used to validate that any proposed cpuset change
 *		       follows the structural rules for cpusets.
 *
 * If we replaced the flag and mask values of the current cpuset
 * (cur) with those values in the trial cpuset (trial), would
 * our various subset and exclusive rules still be valid?  Presumes
463
 * cgroup_mutex held.
L
Linus Torvalds 已提交
464 465 466 467 468 469 470 471 472 473 474 475 476 477
 *
 * 'cur' is the address of an actual, in-use cpuset.  Operations
 * such as list traversal that depend on the actual address of the
 * cpuset in the list must use cur below, not trial.
 *
 * 'trial' is the address of bulk structure copy of cur, with
 * perhaps one or more of the fields cpus_allowed, mems_allowed,
 * or flags changed to new, trial values.
 *
 * Return 0 if valid, -errno if not.
 */

static int validate_change(const struct cpuset *cur, const struct cpuset *trial)
{
478
	struct cgroup *cont;
L
Linus Torvalds 已提交
479 480 481
	struct cpuset *c, *par;

	/* Each of our child cpusets must be a subset of us */
482 483
	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
		if (!is_cpuset_subset(cgroup_cs(cont), trial))
L
Linus Torvalds 已提交
484 485 486 487
			return -EBUSY;
	}

	/* Remaining checks don't apply to root cpuset */
488
	if (cur == &top_cpuset)
L
Linus Torvalds 已提交
489 490
		return 0;

491 492
	par = cur->parent;

L
Linus Torvalds 已提交
493 494 495 496
	/* We must be a subset of our parent cpuset */
	if (!is_cpuset_subset(trial, par))
		return -EACCES;

497 498 499 500
	/*
	 * If either I or some sibling (!= me) is exclusive, we can't
	 * overlap
	 */
501 502
	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
		c = cgroup_cs(cont);
L
Linus Torvalds 已提交
503 504
		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
		    c != cur &&
505
		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
L
Linus Torvalds 已提交
506 507 508 509 510 511 512
			return -EINVAL;
		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
		    c != cur &&
		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
			return -EINVAL;
	}

513 514
	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
	if (cgroup_task_count(cur->css.cgroup)) {
515
		if (cpumask_empty(trial->cpus_allowed) ||
516 517 518 519 520
		    nodes_empty(trial->mems_allowed)) {
			return -ENOSPC;
		}
	}

L
Linus Torvalds 已提交
521 522 523
	return 0;
}

P
Paul Jackson 已提交
524
/*
525
 * Helper routine for generate_sched_domains().
P
Paul Jackson 已提交
526 527 528 529
 * Do cpusets a, b have overlapping cpus_allowed masks?
 */
static int cpusets_overlap(struct cpuset *a, struct cpuset *b)
{
530
	return cpumask_intersects(a->cpus_allowed, b->cpus_allowed);
P
Paul Jackson 已提交
531 532
}

533 534 535 536 537 538 539 540
static void
update_domain_attr(struct sched_domain_attr *dattr, struct cpuset *c)
{
	if (dattr->relax_domain_level < c->relax_domain_level)
		dattr->relax_domain_level = c->relax_domain_level;
	return;
}

541 542 543 544 545 546 547 548 549 550 551 552 553 554
static void
update_domain_attr_tree(struct sched_domain_attr *dattr, struct cpuset *c)
{
	LIST_HEAD(q);

	list_add(&c->stack_list, &q);
	while (!list_empty(&q)) {
		struct cpuset *cp;
		struct cgroup *cont;
		struct cpuset *child;

		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

555
		if (cpumask_empty(cp->cpus_allowed))
556 557 558 559 560 561 562 563 564 565 566 567
			continue;

		if (is_sched_load_balance(cp))
			update_domain_attr(dattr, cp);

		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
			list_add_tail(&child->stack_list, &q);
		}
	}
}

P
Paul Jackson 已提交
568
/*
569 570 571 572 573 574 575 576 577
 * generate_sched_domains()
 *
 * This function builds a partial partition of the systems CPUs
 * A 'partial partition' is a set of non-overlapping subsets whose
 * union is a subset of that set.
 * The output of this function needs to be passed to kernel/sched.c
 * partition_sched_domains() routine, which will rebuild the scheduler's
 * load balancing domains (sched domains) as specified by that partial
 * partition.
P
Paul Jackson 已提交
578
 *
L
Li Zefan 已提交
579
 * See "What is sched_load_balance" in Documentation/cgroups/cpusets.txt
P
Paul Jackson 已提交
580 581 582 583 584 585 586
 * for a background explanation of this.
 *
 * Does not return errors, on the theory that the callers of this
 * routine would rather not worry about failures to rebuild sched
 * domains when operating in the severe memory shortage situations
 * that could cause allocation failures below.
 *
587
 * Must be called with cgroup_lock held.
P
Paul Jackson 已提交
588 589
 *
 * The three key local variables below are:
590
 *    q  - a linked-list queue of cpuset pointers, used to implement a
P
Paul Jackson 已提交
591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621
 *	   top-down scan of all cpusets.  This scan loads a pointer
 *	   to each cpuset marked is_sched_load_balance into the
 *	   array 'csa'.  For our purposes, rebuilding the schedulers
 *	   sched domains, we can ignore !is_sched_load_balance cpusets.
 *  csa  - (for CpuSet Array) Array of pointers to all the cpusets
 *	   that need to be load balanced, for convenient iterative
 *	   access by the subsequent code that finds the best partition,
 *	   i.e the set of domains (subsets) of CPUs such that the
 *	   cpus_allowed of every cpuset marked is_sched_load_balance
 *	   is a subset of one of these domains, while there are as
 *	   many such domains as possible, each as small as possible.
 * doms  - Conversion of 'csa' to an array of cpumasks, for passing to
 *	   the kernel/sched.c routine partition_sched_domains() in a
 *	   convenient format, that can be easily compared to the prior
 *	   value to determine what partition elements (sched domains)
 *	   were changed (added or removed.)
 *
 * Finding the best partition (set of domains):
 *	The triple nested loops below over i, j, k scan over the
 *	load balanced cpusets (using the array of cpuset pointers in
 *	csa[]) looking for pairs of cpusets that have overlapping
 *	cpus_allowed, but which don't have the same 'pn' partition
 *	number and gives them in the same partition number.  It keeps
 *	looping on the 'restart' label until it can no longer find
 *	any such pairs.
 *
 *	The union of the cpus_allowed masks from the set of
 *	all cpusets having the same 'pn' value then form the one
 *	element of the partition (one sched domain) to be passed to
 *	partition_sched_domains().
 */
622 623
/* FIXME: see the FIXME in partition_sched_domains() */
static int generate_sched_domains(struct cpumask **domains,
624
			struct sched_domain_attr **attributes)
P
Paul Jackson 已提交
625
{
626
	LIST_HEAD(q);		/* queue of cpusets to be scanned */
P
Paul Jackson 已提交
627 628 629 630
	struct cpuset *cp;	/* scans q */
	struct cpuset **csa;	/* array of all cpuset ptrs */
	int csn;		/* how many cpuset ptrs in csa so far */
	int i, j, k;		/* indices for partition finding loops */
631
	struct cpumask *doms;	/* resulting partition; i.e. sched domains */
632
	struct sched_domain_attr *dattr;  /* attributes for custom domains */
633
	int ndoms = 0;		/* number of sched domains in result */
634
	int nslot;		/* next empty doms[] struct cpumask slot */
P
Paul Jackson 已提交
635 636

	doms = NULL;
637
	dattr = NULL;
638
	csa = NULL;
P
Paul Jackson 已提交
639 640 641

	/* Special case for the 99% of systems with one, full, sched domain */
	if (is_sched_load_balance(&top_cpuset)) {
642
		doms = kmalloc(cpumask_size(), GFP_KERNEL);
P
Paul Jackson 已提交
643
		if (!doms)
644 645
			goto done;

646 647 648
		dattr = kmalloc(sizeof(struct sched_domain_attr), GFP_KERNEL);
		if (dattr) {
			*dattr = SD_ATTR_INIT;
649
			update_domain_attr_tree(dattr, &top_cpuset);
650
		}
651
		cpumask_copy(doms, top_cpuset.cpus_allowed);
652 653 654

		ndoms = 1;
		goto done;
P
Paul Jackson 已提交
655 656 657 658 659 660 661
	}

	csa = kmalloc(number_of_cpusets * sizeof(cp), GFP_KERNEL);
	if (!csa)
		goto done;
	csn = 0;

662 663
	list_add(&top_cpuset.stack_list, &q);
	while (!list_empty(&q)) {
P
Paul Jackson 已提交
664 665
		struct cgroup *cont;
		struct cpuset *child;   /* scans child cpusets of cp */
666

667 668 669
		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

670
		if (cpumask_empty(cp->cpus_allowed))
671 672
			continue;

673 674 675 676 677 678 679
		/*
		 * All child cpusets contain a subset of the parent's cpus, so
		 * just skip them, and then we call update_domain_attr_tree()
		 * to calc relax_domain_level of the corresponding sched
		 * domain.
		 */
		if (is_sched_load_balance(cp)) {
P
Paul Jackson 已提交
680
			csa[csn++] = cp;
681 682
			continue;
		}
683

P
Paul Jackson 已提交
684 685
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
686
			list_add_tail(&child->stack_list, &q);
P
Paul Jackson 已提交
687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716
		}
  	}

	for (i = 0; i < csn; i++)
		csa[i]->pn = i;
	ndoms = csn;

restart:
	/* Find the best partition (set of sched domains) */
	for (i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
		int apn = a->pn;

		for (j = 0; j < csn; j++) {
			struct cpuset *b = csa[j];
			int bpn = b->pn;

			if (apn != bpn && cpusets_overlap(a, b)) {
				for (k = 0; k < csn; k++) {
					struct cpuset *c = csa[k];

					if (c->pn == bpn)
						c->pn = apn;
				}
				ndoms--;	/* one less element */
				goto restart;
			}
		}
	}

717 718 719 720
	/*
	 * Now we know how many domains to create.
	 * Convert <csn, csa> to <ndoms, doms> and populate cpu masks.
	 */
721
	doms = kmalloc(ndoms * cpumask_size(), GFP_KERNEL);
722
	if (!doms)
723 724 725 726 727 728
		goto done;

	/*
	 * The rest of the code, including the scheduler, can deal with
	 * dattr==NULL case. No need to abort if alloc fails.
	 */
729
	dattr = kmalloc(ndoms * sizeof(struct sched_domain_attr), GFP_KERNEL);
P
Paul Jackson 已提交
730 731 732

	for (nslot = 0, i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
733
		struct cpumask *dp;
P
Paul Jackson 已提交
734 735
		int apn = a->pn;

736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751
		if (apn < 0) {
			/* Skip completed partitions */
			continue;
		}

		dp = doms + nslot;

		if (nslot == ndoms) {
			static int warnings = 10;
			if (warnings) {
				printk(KERN_WARNING
				 "rebuild_sched_domains confused:"
				  " nslot %d, ndoms %d, csn %d, i %d,"
				  " apn %d\n",
				  nslot, ndoms, csn, i, apn);
				warnings--;
P
Paul Jackson 已提交
752
			}
753 754
			continue;
		}
P
Paul Jackson 已提交
755

756
		cpumask_clear(dp);
757 758 759 760 761 762
		if (dattr)
			*(dattr + nslot) = SD_ATTR_INIT;
		for (j = i; j < csn; j++) {
			struct cpuset *b = csa[j];

			if (apn == b->pn) {
763
				cpumask_or(dp, dp, b->cpus_allowed);
764 765 766 767 768
				if (dattr)
					update_domain_attr_tree(dattr + nslot, b);

				/* Done with this partition */
				b->pn = -1;
P
Paul Jackson 已提交
769 770
			}
		}
771
		nslot++;
P
Paul Jackson 已提交
772 773 774
	}
	BUG_ON(nslot != ndoms);

775 776 777
done:
	kfree(csa);

778 779 780 781 782 783 784
	/*
	 * Fallback to the default domain if kmalloc() failed.
	 * See comments in partition_sched_domains().
	 */
	if (doms == NULL)
		ndoms = 1;

785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802
	*domains    = doms;
	*attributes = dattr;
	return ndoms;
}

/*
 * Rebuild scheduler domains.
 *
 * Call with neither cgroup_mutex held nor within get_online_cpus().
 * Takes both cgroup_mutex and get_online_cpus().
 *
 * Cannot be directly called from cpuset code handling changes
 * to the cpuset pseudo-filesystem, because it cannot be called
 * from code that already holds cgroup_mutex.
 */
static void do_rebuild_sched_domains(struct work_struct *unused)
{
	struct sched_domain_attr *attr;
803
	struct cpumask *doms;
804 805
	int ndoms;

806
	get_online_cpus();
807 808 809 810 811 812 813 814 815

	/* Generate domain masks and attrs */
	cgroup_lock();
	ndoms = generate_sched_domains(&doms, &attr);
	cgroup_unlock();

	/* Have scheduler rebuild the domains */
	partition_sched_domains(ndoms, doms, attr);

816
	put_online_cpus();
817
}
P
Paul Jackson 已提交
818

819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
static DECLARE_WORK(rebuild_sched_domains_work, do_rebuild_sched_domains);

/*
 * Rebuild scheduler domains, asynchronously via workqueue.
 *
 * If the flag 'sched_load_balance' of any cpuset with non-empty
 * 'cpus' changes, or if the 'cpus' allowed changes in any cpuset
 * which has that flag enabled, or if any cpuset with a non-empty
 * 'cpus' is removed, then call this routine to rebuild the
 * scheduler's dynamic sched domains.
 *
 * The rebuild_sched_domains() and partition_sched_domains()
 * routines must nest cgroup_lock() inside get_online_cpus(),
 * but such cpuset changes as these must nest that locking the
 * other way, holding cgroup_lock() for much of the code.
 *
 * So in order to avoid an ABBA deadlock, the cpuset code handling
 * these user changes delegates the actual sched domain rebuilding
 * to a separate workqueue thread, which ends up processing the
 * above do_rebuild_sched_domains() function.
 */
static void async_rebuild_sched_domains(void)
{
842
	queue_work(cpuset_wq, &rebuild_sched_domains_work);
843 844 845 846 847 848 849 850 851 852 853 854 855 856
}

/*
 * Accomplishes the same scheduler domain rebuild as the above
 * async_rebuild_sched_domains(), however it directly calls the
 * rebuild routine synchronously rather than calling it via an
 * asynchronous work thread.
 *
 * This can only be called from code that is not holding
 * cgroup_mutex (not nested in a cgroup_lock() call.)
 */
void rebuild_sched_domains(void)
{
	do_rebuild_sched_domains(NULL);
P
Paul Jackson 已提交
857 858
}

C
Cliff Wickman 已提交
859 860 861 862 863
/**
 * cpuset_test_cpumask - test a task's cpus_allowed versus its cpuset's
 * @tsk: task to test
 * @scan: struct cgroup_scanner contained in its struct cpuset_hotplug_scanner
 *
864
 * Call with cgroup_mutex held.  May take callback_mutex during call.
C
Cliff Wickman 已提交
865 866 867
 * Called for each task in a cgroup by cgroup_scan_tasks().
 * Return nonzero if this tasks's cpus_allowed mask should be changed (in other
 * words, if its mask is not equal to its cpuset's mask).
868
 */
869 870
static int cpuset_test_cpumask(struct task_struct *tsk,
			       struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
871
{
872
	return !cpumask_equal(&tsk->cpus_allowed,
C
Cliff Wickman 已提交
873 874
			(cgroup_cs(scan->cg))->cpus_allowed);
}
875

C
Cliff Wickman 已提交
876 877 878 879 880 881 882 883 884 885 886
/**
 * cpuset_change_cpumask - make a task's cpus_allowed the same as its cpuset's
 * @tsk: task to test
 * @scan: struct cgroup_scanner containing the cgroup of the task
 *
 * Called by cgroup_scan_tasks() for each task in a cgroup whose
 * cpus_allowed mask needs to be changed.
 *
 * We don't need to re-check for the cgroup/cpuset membership, since we're
 * holding cgroup_lock() at this point.
 */
887 888
static void cpuset_change_cpumask(struct task_struct *tsk,
				  struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
889
{
890
	set_cpus_allowed_ptr(tsk, ((cgroup_cs(scan->cg))->cpus_allowed));
C
Cliff Wickman 已提交
891 892
}

893 894 895
/**
 * update_tasks_cpumask - Update the cpumasks of tasks in the cpuset.
 * @cs: the cpuset in which each task's cpus_allowed mask needs to be changed
896
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
897 898 899 900 901 902
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
903 904
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
905
 */
906
static void update_tasks_cpumask(struct cpuset *cs, struct ptr_heap *heap)
907 908 909 910 911 912
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = cpuset_test_cpumask;
	scan.process_task = cpuset_change_cpumask;
913 914
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
915 916
}

C
Cliff Wickman 已提交
917 918 919 920 921
/**
 * update_cpumask - update the cpus_allowed mask of a cpuset and all tasks in it
 * @cs: the cpuset to consider
 * @buf: buffer of cpu numbers written to this cpuset
 */
922 923
static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
			  const char *buf)
L
Linus Torvalds 已提交
924
{
925
	struct ptr_heap heap;
C
Cliff Wickman 已提交
926 927
	int retval;
	int is_load_balanced;
L
Linus Torvalds 已提交
928

929 930 931 932
	/* top_cpuset.cpus_allowed tracks cpu_online_map; it's read-only */
	if (cs == &top_cpuset)
		return -EACCES;

933
	/*
934
	 * An empty cpus_allowed is ok only if the cpuset has no tasks.
935 936 937
	 * Since cpulist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have cpus.
938
	 */
939
	if (!*buf) {
940
		cpumask_clear(trialcs->cpus_allowed);
941
	} else {
942
		retval = cpulist_parse(buf, trialcs->cpus_allowed);
943 944
		if (retval < 0)
			return retval;
945

946
		if (!cpumask_subset(trialcs->cpus_allowed, cpu_online_mask))
947
			return -EINVAL;
948
	}
949
	retval = validate_change(cs, trialcs);
950 951
	if (retval < 0)
		return retval;
P
Paul Jackson 已提交
952

P
Paul Menage 已提交
953
	/* Nothing to do if the cpus didn't change */
954
	if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed))
P
Paul Menage 已提交
955
		return 0;
C
Cliff Wickman 已提交
956

957 958 959 960
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval)
		return retval;

961
	is_load_balanced = is_sched_load_balance(trialcs);
P
Paul Jackson 已提交
962

963
	mutex_lock(&callback_mutex);
964
	cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed);
965
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
966

P
Paul Menage 已提交
967 968
	/*
	 * Scan tasks in the cpuset, and update the cpumasks of any
C
Cliff Wickman 已提交
969
	 * that need an update.
P
Paul Menage 已提交
970
	 */
971 972 973
	update_tasks_cpumask(cs, &heap);

	heap_free(&heap);
C
Cliff Wickman 已提交
974

P
Paul Menage 已提交
975
	if (is_load_balanced)
976
		async_rebuild_sched_domains();
977
	return 0;
L
Linus Torvalds 已提交
978 979
}

980 981 982 983 984 985 986 987
/*
 * cpuset_migrate_mm
 *
 *    Migrate memory region from one set of nodes to another.
 *
 *    Temporarilly set tasks mems_allowed to target nodes of migration,
 *    so that the migration code can allocate pages on these nodes.
 *
988
 *    Call holding cgroup_mutex, so current's cpuset won't change
989
 *    during this call, as manage_mutex holds off any cpuset_attach()
990 991
 *    calls.  Therefore we don't need to take task_lock around the
 *    call to guarantee_online_mems(), as we know no one is changing
992
 *    our task's cpuset.
993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
 *
 *    Hold callback_mutex around the two modifications of our tasks
 *    mems_allowed to synchronize with cpuset_mems_allowed().
 *
 *    While the mm_struct we are migrating is typically from some
 *    other task, the task_struct mems_allowed that we are hacking
 *    is for our current task, which must allocate new pages for that
 *    migrating memory region.
 *
 *    We call cpuset_update_task_memory_state() before hacking
 *    our tasks mems_allowed, so that we are assured of being in
 *    sync with our tasks cpuset, and in particular, callbacks to
 *    cpuset_update_task_memory_state() from nested page allocations
 *    won't see any mismatch of our cpuset and task mems_generation
 *    values, so won't overwrite our hacked tasks mems_allowed
 *    nodemask.
 */

static void cpuset_migrate_mm(struct mm_struct *mm, const nodemask_t *from,
							const nodemask_t *to)
{
	struct task_struct *tsk = current;

	cpuset_update_task_memory_state();

	mutex_lock(&callback_mutex);
	tsk->mems_allowed = *to;
	mutex_unlock(&callback_mutex);

	do_migrate_pages(mm, from, to, MPOL_MF_MOVE_ALL);

	mutex_lock(&callback_mutex);
1025
	guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
1026 1027 1028
	mutex_unlock(&callback_mutex);
}

1029 1030
static void *cpuset_being_rebound;

1031 1032 1033 1034 1035 1036 1037 1038 1039
/**
 * update_tasks_nodemask - Update the nodemasks of tasks in the cpuset.
 * @cs: the cpuset in which each task's mems_allowed mask needs to be changed
 * @oldmem: old mems_allowed of cpuset cs
 *
 * Called with cgroup_mutex held
 * Return 0 if successful, -errno if not.
 */
static int update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem)
L
Linus Torvalds 已提交
1040
{
1041
	struct task_struct *p;
1042 1043
	struct mm_struct **mmarray;
	int i, n, ntasks;
1044
	int migrate;
1045
	int fudge;
1046
	struct cgroup_iter it;
1047
	int retval;
1048

1049
	cpuset_being_rebound = cs;		/* causes mpol_dup() rebind */
1050 1051

	fudge = 10;				/* spare mmarray[] slots */
1052
	fudge += cpumask_weight(cs->cpus_allowed);/* imagine 1 fork-bomb/cpu */
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
	retval = -ENOMEM;

	/*
	 * Allocate mmarray[] to hold mm reference for each task
	 * in cpuset cs.  Can't kmalloc GFP_KERNEL while holding
	 * tasklist_lock.  We could use GFP_ATOMIC, but with a
	 * few more lines of code, we can retry until we get a big
	 * enough mmarray[] w/o using GFP_ATOMIC.
	 */
	while (1) {
1063
		ntasks = cgroup_task_count(cs->css.cgroup);  /* guess */
1064 1065 1066 1067
		ntasks += fudge;
		mmarray = kmalloc(ntasks * sizeof(*mmarray), GFP_KERNEL);
		if (!mmarray)
			goto done;
1068
		read_lock(&tasklist_lock);		/* block fork */
1069
		if (cgroup_task_count(cs->css.cgroup) <= ntasks)
1070
			break;				/* got enough */
1071
		read_unlock(&tasklist_lock);		/* try again */
1072 1073 1074 1075 1076 1077
		kfree(mmarray);
	}

	n = 0;

	/* Load up mmarray[] with mm reference for each task in cpuset. */
1078 1079
	cgroup_iter_start(cs->css.cgroup, &it);
	while ((p = cgroup_iter_next(cs->css.cgroup, &it))) {
1080 1081 1082 1083 1084
		struct mm_struct *mm;

		if (n >= ntasks) {
			printk(KERN_WARNING
				"Cpuset mempolicy rebind incomplete.\n");
1085
			break;
1086 1087 1088 1089 1090
		}
		mm = get_task_mm(p);
		if (!mm)
			continue;
		mmarray[n++] = mm;
1091 1092
	}
	cgroup_iter_end(cs->css.cgroup, &it);
1093
	read_unlock(&tasklist_lock);
1094 1095 1096 1097 1098 1099

	/*
	 * Now that we've dropped the tasklist spinlock, we can
	 * rebind the vma mempolicies of each mm in mmarray[] to their
	 * new cpuset, and release that mm.  The mpol_rebind_mm()
	 * call takes mmap_sem, which we couldn't take while holding
1100
	 * tasklist_lock.  Forks can happen again now - the mpol_dup()
1101 1102
	 * cpuset_being_rebound check will catch such forks, and rebind
	 * their vma mempolicies too.  Because we still hold the global
1103
	 * cgroup_mutex, we know that no other rebind effort will
1104 1105
	 * be contending for the global variable cpuset_being_rebound.
	 * It's ok if we rebind the same mm twice; mpol_rebind_mm()
1106
	 * is idempotent.  Also migrate pages in each mm to new nodes.
1107
	 */
1108
	migrate = is_memory_migrate(cs);
1109 1110 1111 1112
	for (i = 0; i < n; i++) {
		struct mm_struct *mm = mmarray[i];

		mpol_rebind_mm(mm, &cs->mems_allowed);
1113
		if (migrate)
1114
			cpuset_migrate_mm(mm, oldmem, &cs->mems_allowed);
1115 1116 1117
		mmput(mm);
	}

1118
	/* We're done rebinding vmas to this cpuset's new mems_allowed. */
1119
	kfree(mmarray);
1120
	cpuset_being_rebound = NULL;
1121
	retval = 0;
1122
done:
L
Linus Torvalds 已提交
1123 1124 1125
	return retval;
}

1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
/*
 * Handle user request to change the 'mems' memory placement
 * of a cpuset.  Needs to validate the request, update the
 * cpusets mems_allowed and mems_generation, and for each
 * task in the cpuset, rebind any vma mempolicies and if
 * the cpuset is marked 'memory_migrate', migrate the tasks
 * pages to the new memory.
 *
 * Call with cgroup_mutex held.  May take callback_mutex during call.
 * Will take tasklist_lock, scan tasklist for tasks in cpuset cs,
 * lock each such tasks mm->mmap_sem, scan its vma's and rebind
 * their mempolicies to the cpusets new mems_allowed.
 */
1139 1140
static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs,
			   const char *buf)
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
{
	nodemask_t oldmem;
	int retval;

	/*
	 * top_cpuset.mems_allowed tracks node_stats[N_HIGH_MEMORY];
	 * it's read-only
	 */
	if (cs == &top_cpuset)
		return -EACCES;

	/*
	 * An empty mems_allowed is ok iff there are no tasks in the cpuset.
	 * Since nodelist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have memory.
	 */
	if (!*buf) {
1159
		nodes_clear(trialcs->mems_allowed);
1160
	} else {
1161
		retval = nodelist_parse(buf, trialcs->mems_allowed);
1162 1163 1164
		if (retval < 0)
			goto done;

1165
		if (!nodes_subset(trialcs->mems_allowed,
1166 1167 1168 1169
				node_states[N_HIGH_MEMORY]))
			return -EINVAL;
	}
	oldmem = cs->mems_allowed;
1170
	if (nodes_equal(oldmem, trialcs->mems_allowed)) {
1171 1172 1173
		retval = 0;		/* Too easy - nothing to do */
		goto done;
	}
1174
	retval = validate_change(cs, trialcs);
1175 1176 1177 1178
	if (retval < 0)
		goto done;

	mutex_lock(&callback_mutex);
1179
	cs->mems_allowed = trialcs->mems_allowed;
1180 1181 1182 1183 1184 1185 1186 1187
	cs->mems_generation = cpuset_mems_generation++;
	mutex_unlock(&callback_mutex);

	retval = update_tasks_nodemask(cs, &oldmem);
done:
	return retval;
}

1188 1189 1190 1191 1192
int current_cpuset_is_being_rebound(void)
{
	return task_cs(current) == cpuset_being_rebound;
}

1193
static int update_relax_domain_level(struct cpuset *cs, s64 val)
1194
{
1195 1196
	if (val < -1 || val >= SD_LV_MAX)
		return -EINVAL;
1197 1198 1199

	if (val != cs->relax_domain_level) {
		cs->relax_domain_level = val;
1200 1201
		if (!cpumask_empty(cs->cpus_allowed) &&
		    is_sched_load_balance(cs))
1202
			async_rebuild_sched_domains();
1203 1204 1205 1206 1207
	}

	return 0;
}

L
Linus Torvalds 已提交
1208 1209
/*
 * update_flag - read a 0 or a 1 in a file and update associated flag
1210 1211 1212
 * bit:		the bit to update (see cpuset_flagbits_t)
 * cs:		the cpuset to update
 * turning_on: 	whether the flag is being set or cleared
1213
 *
1214
 * Call with cgroup_mutex held.
L
Linus Torvalds 已提交
1215 1216
 */

1217 1218
static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
		       int turning_on)
L
Linus Torvalds 已提交
1219
{
1220
	struct cpuset *trialcs;
1221
	int err;
R
Rakib Mullick 已提交
1222
	int balance_flag_changed;
L
Linus Torvalds 已提交
1223

1224 1225 1226 1227
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

L
Linus Torvalds 已提交
1228
	if (turning_on)
1229
		set_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1230
	else
1231
		clear_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1232

1233
	err = validate_change(cs, trialcs);
1234
	if (err < 0)
1235
		goto out;
P
Paul Jackson 已提交
1236 1237

	balance_flag_changed = (is_sched_load_balance(cs) !=
1238
				is_sched_load_balance(trialcs));
P
Paul Jackson 已提交
1239

1240
	mutex_lock(&callback_mutex);
1241
	cs->flags = trialcs->flags;
1242
	mutex_unlock(&callback_mutex);
1243

1244
	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed)
1245
		async_rebuild_sched_domains();
P
Paul Jackson 已提交
1246

1247 1248 1249
out:
	free_trial_cpuset(trialcs);
	return err;
L
Linus Torvalds 已提交
1250 1251
}

1252
/*
A
Adrian Bunk 已提交
1253
 * Frequency meter - How fast is some event occurring?
1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349
 *
 * These routines manage a digitally filtered, constant time based,
 * event frequency meter.  There are four routines:
 *   fmeter_init() - initialize a frequency meter.
 *   fmeter_markevent() - called each time the event happens.
 *   fmeter_getrate() - returns the recent rate of such events.
 *   fmeter_update() - internal routine used to update fmeter.
 *
 * A common data structure is passed to each of these routines,
 * which is used to keep track of the state required to manage the
 * frequency meter and its digital filter.
 *
 * The filter works on the number of events marked per unit time.
 * The filter is single-pole low-pass recursive (IIR).  The time unit
 * is 1 second.  Arithmetic is done using 32-bit integers scaled to
 * simulate 3 decimal digits of precision (multiplied by 1000).
 *
 * With an FM_COEF of 933, and a time base of 1 second, the filter
 * has a half-life of 10 seconds, meaning that if the events quit
 * happening, then the rate returned from the fmeter_getrate()
 * will be cut in half each 10 seconds, until it converges to zero.
 *
 * It is not worth doing a real infinitely recursive filter.  If more
 * than FM_MAXTICKS ticks have elapsed since the last filter event,
 * just compute FM_MAXTICKS ticks worth, by which point the level
 * will be stable.
 *
 * Limit the count of unprocessed events to FM_MAXCNT, so as to avoid
 * arithmetic overflow in the fmeter_update() routine.
 *
 * Given the simple 32 bit integer arithmetic used, this meter works
 * best for reporting rates between one per millisecond (msec) and
 * one per 32 (approx) seconds.  At constant rates faster than one
 * per msec it maxes out at values just under 1,000,000.  At constant
 * rates between one per msec, and one per second it will stabilize
 * to a value N*1000, where N is the rate of events per second.
 * At constant rates between one per second and one per 32 seconds,
 * it will be choppy, moving up on the seconds that have an event,
 * and then decaying until the next event.  At rates slower than
 * about one in 32 seconds, it decays all the way back to zero between
 * each event.
 */

#define FM_COEF 933		/* coefficient for half-life of 10 secs */
#define FM_MAXTICKS ((time_t)99) /* useless computing more ticks than this */
#define FM_MAXCNT 1000000	/* limit cnt to avoid overflow */
#define FM_SCALE 1000		/* faux fixed point scale */

/* Initialize a frequency meter */
static void fmeter_init(struct fmeter *fmp)
{
	fmp->cnt = 0;
	fmp->val = 0;
	fmp->time = 0;
	spin_lock_init(&fmp->lock);
}

/* Internal meter update - process cnt events and update value */
static void fmeter_update(struct fmeter *fmp)
{
	time_t now = get_seconds();
	time_t ticks = now - fmp->time;

	if (ticks == 0)
		return;

	ticks = min(FM_MAXTICKS, ticks);
	while (ticks-- > 0)
		fmp->val = (FM_COEF * fmp->val) / FM_SCALE;
	fmp->time = now;

	fmp->val += ((FM_SCALE - FM_COEF) * fmp->cnt) / FM_SCALE;
	fmp->cnt = 0;
}

/* Process any previous ticks, then bump cnt by one (times scale). */
static void fmeter_markevent(struct fmeter *fmp)
{
	spin_lock(&fmp->lock);
	fmeter_update(fmp);
	fmp->cnt = min(FM_MAXCNT, fmp->cnt + FM_SCALE);
	spin_unlock(&fmp->lock);
}

/* Process any previous ticks, then return current value. */
static int fmeter_getrate(struct fmeter *fmp)
{
	int val;

	spin_lock(&fmp->lock);
	fmeter_update(fmp);
	val = fmp->val;
	spin_unlock(&fmp->lock);
	return val;
}

1350 1351 1352
/* Protected by cgroup_lock */
static cpumask_var_t cpus_attach;

1353
/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
1354 1355
static int cpuset_can_attach(struct cgroup_subsys *ss,
			     struct cgroup *cont, struct task_struct *tsk)
L
Linus Torvalds 已提交
1356
{
1357
	struct cpuset *cs = cgroup_cs(cont);
1358
	int ret = 0;
L
Linus Torvalds 已提交
1359

1360
	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
L
Linus Torvalds 已提交
1361
		return -ENOSPC;
1362

1363
	if (tsk->flags & PF_THREAD_BOUND) {
1364
		mutex_lock(&callback_mutex);
1365
		if (!cpumask_equal(&tsk->cpus_allowed, cs->cpus_allowed))
1366
			ret = -EINVAL;
1367 1368
		mutex_unlock(&callback_mutex);
	}
L
Linus Torvalds 已提交
1369

1370
	return ret < 0 ? ret : security_task_setscheduler(tsk, 0, NULL);
1371
}
L
Linus Torvalds 已提交
1372

1373 1374 1375 1376 1377 1378 1379 1380
static void cpuset_attach(struct cgroup_subsys *ss,
			  struct cgroup *cont, struct cgroup *oldcont,
			  struct task_struct *tsk)
{
	nodemask_t from, to;
	struct mm_struct *mm;
	struct cpuset *cs = cgroup_cs(cont);
	struct cpuset *oldcs = cgroup_cs(oldcont);
1381
	int err;
1382

1383
	if (cs == &top_cpuset) {
1384
		cpumask_copy(cpus_attach, cpu_possible_mask);
1385 1386
	} else {
		mutex_lock(&callback_mutex);
1387
		guarantee_online_cpus(cs, cpus_attach);
1388 1389
		mutex_unlock(&callback_mutex);
	}
1390
	err = set_cpus_allowed_ptr(tsk, cpus_attach);
1391 1392
	if (err)
		return;
L
Linus Torvalds 已提交
1393

1394 1395
	from = oldcs->mems_allowed;
	to = cs->mems_allowed;
1396 1397 1398
	mm = get_task_mm(tsk);
	if (mm) {
		mpol_rebind_mm(mm, &to);
1399
		if (is_memory_migrate(cs))
1400
			cpuset_migrate_mm(mm, &from, &to);
1401 1402
		mmput(mm);
	}
L
Linus Torvalds 已提交
1403 1404 1405 1406 1407
}

/* The various types of files and directories in a cpuset file system */

typedef enum {
1408
	FILE_MEMORY_MIGRATE,
L
Linus Torvalds 已提交
1409 1410 1411 1412
	FILE_CPULIST,
	FILE_MEMLIST,
	FILE_CPU_EXCLUSIVE,
	FILE_MEM_EXCLUSIVE,
1413
	FILE_MEM_HARDWALL,
P
Paul Jackson 已提交
1414
	FILE_SCHED_LOAD_BALANCE,
1415
	FILE_SCHED_RELAX_DOMAIN_LEVEL,
1416 1417
	FILE_MEMORY_PRESSURE_ENABLED,
	FILE_MEMORY_PRESSURE,
1418 1419
	FILE_SPREAD_PAGE,
	FILE_SPREAD_SLAB,
L
Linus Torvalds 已提交
1420 1421
} cpuset_filetype_t;

1422 1423 1424 1425 1426 1427
static int cpuset_write_u64(struct cgroup *cgrp, struct cftype *cft, u64 val)
{
	int retval = 0;
	struct cpuset *cs = cgroup_cs(cgrp);
	cpuset_filetype_t type = cft->private;

1428
	if (!cgroup_lock_live_group(cgrp))
1429 1430 1431
		return -ENODEV;

	switch (type) {
L
Linus Torvalds 已提交
1432
	case FILE_CPU_EXCLUSIVE:
1433
		retval = update_flag(CS_CPU_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1434 1435
		break;
	case FILE_MEM_EXCLUSIVE:
1436
		retval = update_flag(CS_MEM_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1437
		break;
1438 1439 1440
	case FILE_MEM_HARDWALL:
		retval = update_flag(CS_MEM_HARDWALL, cs, val);
		break;
P
Paul Jackson 已提交
1441
	case FILE_SCHED_LOAD_BALANCE:
1442
		retval = update_flag(CS_SCHED_LOAD_BALANCE, cs, val);
1443
		break;
1444
	case FILE_MEMORY_MIGRATE:
1445
		retval = update_flag(CS_MEMORY_MIGRATE, cs, val);
1446
		break;
1447
	case FILE_MEMORY_PRESSURE_ENABLED:
1448
		cpuset_memory_pressure_enabled = !!val;
1449 1450 1451 1452
		break;
	case FILE_MEMORY_PRESSURE:
		retval = -EACCES;
		break;
1453
	case FILE_SPREAD_PAGE:
1454
		retval = update_flag(CS_SPREAD_PAGE, cs, val);
1455
		cs->mems_generation = cpuset_mems_generation++;
1456 1457
		break;
	case FILE_SPREAD_SLAB:
1458
		retval = update_flag(CS_SPREAD_SLAB, cs, val);
1459
		cs->mems_generation = cpuset_mems_generation++;
1460
		break;
L
Linus Torvalds 已提交
1461 1462
	default:
		retval = -EINVAL;
1463
		break;
L
Linus Torvalds 已提交
1464
	}
1465
	cgroup_unlock();
L
Linus Torvalds 已提交
1466 1467 1468
	return retval;
}

1469 1470 1471 1472 1473 1474
static int cpuset_write_s64(struct cgroup *cgrp, struct cftype *cft, s64 val)
{
	int retval = 0;
	struct cpuset *cs = cgroup_cs(cgrp);
	cpuset_filetype_t type = cft->private;

1475
	if (!cgroup_lock_live_group(cgrp))
1476
		return -ENODEV;
1477

1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489
	switch (type) {
	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
		retval = update_relax_domain_level(cs, val);
		break;
	default:
		retval = -EINVAL;
		break;
	}
	cgroup_unlock();
	return retval;
}

1490 1491 1492 1493 1494 1495 1496
/*
 * Common handling for a write to a "cpus" or "mems" file.
 */
static int cpuset_write_resmask(struct cgroup *cgrp, struct cftype *cft,
				const char *buf)
{
	int retval = 0;
1497 1498
	struct cpuset *cs = cgroup_cs(cgrp);
	struct cpuset *trialcs;
1499 1500 1501 1502

	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

1503 1504 1505 1506
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

1507 1508
	switch (cft->private) {
	case FILE_CPULIST:
1509
		retval = update_cpumask(cs, trialcs, buf);
1510 1511
		break;
	case FILE_MEMLIST:
1512
		retval = update_nodemask(cs, trialcs, buf);
1513 1514 1515 1516 1517
		break;
	default:
		retval = -EINVAL;
		break;
	}
1518 1519

	free_trial_cpuset(trialcs);
1520 1521 1522 1523
	cgroup_unlock();
	return retval;
}

L
Linus Torvalds 已提交
1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
/*
 * These ascii lists should be read in a single call, by using a user
 * buffer large enough to hold the entire map.  If read in smaller
 * chunks, there is no guarantee of atomicity.  Since the display format
 * used, list of ranges of sequential numbers, is variable length,
 * and since these maps can change value dynamically, one could read
 * gibberish by doing partial reads while a list was changing.
 * A single large read to a buffer that crosses a page boundary is
 * ok, because the result being copied to user land is not recomputed
 * across a page fault.
 */

static int cpuset_sprintf_cpulist(char *page, struct cpuset *cs)
{
1538
	int ret;
L
Linus Torvalds 已提交
1539

1540
	mutex_lock(&callback_mutex);
1541
	ret = cpulist_scnprintf(page, PAGE_SIZE, cs->cpus_allowed);
1542
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1543

1544
	return ret;
L
Linus Torvalds 已提交
1545 1546 1547 1548 1549 1550
}

static int cpuset_sprintf_memlist(char *page, struct cpuset *cs)
{
	nodemask_t mask;

1551
	mutex_lock(&callback_mutex);
L
Linus Torvalds 已提交
1552
	mask = cs->mems_allowed;
1553
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1554 1555 1556 1557

	return nodelist_scnprintf(page, PAGE_SIZE, mask);
}

1558 1559 1560 1561 1562
static ssize_t cpuset_common_file_read(struct cgroup *cont,
				       struct cftype *cft,
				       struct file *file,
				       char __user *buf,
				       size_t nbytes, loff_t *ppos)
L
Linus Torvalds 已提交
1563
{
1564
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1565 1566 1567 1568 1569
	cpuset_filetype_t type = cft->private;
	char *page;
	ssize_t retval = 0;
	char *s;

1570
	if (!(page = (char *)__get_free_page(GFP_TEMPORARY)))
L
Linus Torvalds 已提交
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
		return -ENOMEM;

	s = page;

	switch (type) {
	case FILE_CPULIST:
		s += cpuset_sprintf_cpulist(s, cs);
		break;
	case FILE_MEMLIST:
		s += cpuset_sprintf_memlist(s, cs);
		break;
	default:
		retval = -EINVAL;
		goto out;
	}
	*s++ = '\n';

A
Al Viro 已提交
1588
	retval = simple_read_from_buffer(buf, nbytes, ppos, page, s - page);
L
Linus Torvalds 已提交
1589 1590 1591 1592 1593
out:
	free_page((unsigned long)page);
	return retval;
}

1594 1595 1596 1597 1598 1599 1600 1601 1602
static u64 cpuset_read_u64(struct cgroup *cont, struct cftype *cft)
{
	struct cpuset *cs = cgroup_cs(cont);
	cpuset_filetype_t type = cft->private;
	switch (type) {
	case FILE_CPU_EXCLUSIVE:
		return is_cpu_exclusive(cs);
	case FILE_MEM_EXCLUSIVE:
		return is_mem_exclusive(cs);
1603 1604
	case FILE_MEM_HARDWALL:
		return is_mem_hardwall(cs);
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619
	case FILE_SCHED_LOAD_BALANCE:
		return is_sched_load_balance(cs);
	case FILE_MEMORY_MIGRATE:
		return is_memory_migrate(cs);
	case FILE_MEMORY_PRESSURE_ENABLED:
		return cpuset_memory_pressure_enabled;
	case FILE_MEMORY_PRESSURE:
		return fmeter_getrate(&cs->fmeter);
	case FILE_SPREAD_PAGE:
		return is_spread_page(cs);
	case FILE_SPREAD_SLAB:
		return is_spread_slab(cs);
	default:
		BUG();
	}
1620 1621 1622

	/* Unreachable but makes gcc happy */
	return 0;
1623
}
L
Linus Torvalds 已提交
1624

1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
static s64 cpuset_read_s64(struct cgroup *cont, struct cftype *cft)
{
	struct cpuset *cs = cgroup_cs(cont);
	cpuset_filetype_t type = cft->private;
	switch (type) {
	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
		return cs->relax_domain_level;
	default:
		BUG();
	}
1635 1636 1637

	/* Unrechable but makes gcc happy */
	return 0;
1638 1639
}

L
Linus Torvalds 已提交
1640 1641 1642 1643 1644

/*
 * for the common functions, 'private' gives the type of file
 */

1645 1646 1647 1648
static struct cftype files[] = {
	{
		.name = "cpus",
		.read = cpuset_common_file_read,
1649 1650
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * NR_CPUS),
1651 1652 1653 1654 1655 1656
		.private = FILE_CPULIST,
	},

	{
		.name = "mems",
		.read = cpuset_common_file_read,
1657 1658
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * MAX_NUMNODES),
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
		.private = FILE_MEMLIST,
	},

	{
		.name = "cpu_exclusive",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_CPU_EXCLUSIVE,
	},

	{
		.name = "mem_exclusive",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEM_EXCLUSIVE,
	},

1676 1677 1678 1679 1680 1681 1682
	{
		.name = "mem_hardwall",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEM_HARDWALL,
	},

1683 1684 1685 1686 1687 1688 1689 1690 1691
	{
		.name = "sched_load_balance",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SCHED_LOAD_BALANCE,
	},

	{
		.name = "sched_relax_domain_level",
1692 1693
		.read_s64 = cpuset_read_s64,
		.write_s64 = cpuset_write_s64,
1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723
		.private = FILE_SCHED_RELAX_DOMAIN_LEVEL,
	},

	{
		.name = "memory_migrate",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEMORY_MIGRATE,
	},

	{
		.name = "memory_pressure",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEMORY_PRESSURE,
	},

	{
		.name = "memory_spread_page",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SPREAD_PAGE,
	},

	{
		.name = "memory_spread_slab",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SPREAD_SLAB,
	},
1724 1725
};

1726 1727
static struct cftype cft_memory_pressure_enabled = {
	.name = "memory_pressure_enabled",
1728 1729
	.read_u64 = cpuset_read_u64,
	.write_u64 = cpuset_write_u64,
1730 1731 1732
	.private = FILE_MEMORY_PRESSURE_ENABLED,
};

1733
static int cpuset_populate(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1734 1735 1736
{
	int err;

1737 1738
	err = cgroup_add_files(cont, ss, files, ARRAY_SIZE(files));
	if (err)
L
Linus Torvalds 已提交
1739
		return err;
1740
	/* memory_pressure_enabled is in root cpuset only */
1741
	if (!cont->parent)
1742
		err = cgroup_add_file(cont, ss,
1743 1744
				      &cft_memory_pressure_enabled);
	return err;
L
Linus Torvalds 已提交
1745 1746
}

1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760
/*
 * post_clone() is called at the end of cgroup_clone().
 * 'cgroup' was just created automatically as a result of
 * a cgroup_clone(), and the current task is about to
 * be moved into 'cgroup'.
 *
 * Currently we refuse to set up the cgroup - thereby
 * refusing the task to be entered, and as a result refusing
 * the sys_unshare() or clone() which initiated it - if any
 * sibling cpusets have exclusive cpus or mem.
 *
 * If this becomes a problem for some users who wish to
 * allow that scenario, then cpuset_post_clone() could be
 * changed to grant parent->cpus_allowed-sibling_cpus_exclusive
1761 1762
 * (and likewise for mems) to the new cgroup. Called with cgroup_mutex
 * held.
1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779
 */
static void cpuset_post_clone(struct cgroup_subsys *ss,
			      struct cgroup *cgroup)
{
	struct cgroup *parent, *child;
	struct cpuset *cs, *parent_cs;

	parent = cgroup->parent;
	list_for_each_entry(child, &parent->children, sibling) {
		cs = cgroup_cs(child);
		if (is_mem_exclusive(cs) || is_cpu_exclusive(cs))
			return;
	}
	cs = cgroup_cs(cgroup);
	parent_cs = cgroup_cs(parent);

	cs->mems_allowed = parent_cs->mems_allowed;
1780
	cpumask_copy(cs->cpus_allowed, parent_cs->cpus_allowed);
1781 1782 1783
	return;
}

L
Linus Torvalds 已提交
1784 1785
/*
 *	cpuset_create - create a cpuset
1786 1787
 *	ss:	cpuset cgroup subsystem
 *	cont:	control group that the new cpuset will be part of
L
Linus Torvalds 已提交
1788 1789
 */

1790 1791 1792
static struct cgroup_subsys_state *cpuset_create(
	struct cgroup_subsys *ss,
	struct cgroup *cont)
L
Linus Torvalds 已提交
1793 1794
{
	struct cpuset *cs;
1795
	struct cpuset *parent;
L
Linus Torvalds 已提交
1796

1797 1798 1799 1800 1801 1802
	if (!cont->parent) {
		/* This is early initialization for the top cgroup */
		top_cpuset.mems_generation = cpuset_mems_generation++;
		return &top_cpuset.css;
	}
	parent = cgroup_cs(cont->parent);
L
Linus Torvalds 已提交
1803 1804
	cs = kmalloc(sizeof(*cs), GFP_KERNEL);
	if (!cs)
1805
		return ERR_PTR(-ENOMEM);
1806 1807 1808 1809
	if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) {
		kfree(cs);
		return ERR_PTR(-ENOMEM);
	}
L
Linus Torvalds 已提交
1810

1811
	cpuset_update_task_memory_state();
L
Linus Torvalds 已提交
1812
	cs->flags = 0;
1813 1814 1815 1816
	if (is_spread_page(parent))
		set_bit(CS_SPREAD_PAGE, &cs->flags);
	if (is_spread_slab(parent))
		set_bit(CS_SPREAD_SLAB, &cs->flags);
P
Paul Jackson 已提交
1817
	set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
1818
	cpumask_clear(cs->cpus_allowed);
1819
	nodes_clear(cs->mems_allowed);
1820
	cs->mems_generation = cpuset_mems_generation++;
1821
	fmeter_init(&cs->fmeter);
1822
	cs->relax_domain_level = -1;
L
Linus Torvalds 已提交
1823 1824

	cs->parent = parent;
1825
	number_of_cpusets++;
1826
	return &cs->css ;
L
Linus Torvalds 已提交
1827 1828
}

P
Paul Jackson 已提交
1829 1830 1831
/*
 * If the cpuset being removed has its flag 'sched_load_balance'
 * enabled, then simulate turning sched_load_balance off, which
1832
 * will call async_rebuild_sched_domains().
P
Paul Jackson 已提交
1833 1834
 */

1835
static void cpuset_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1836
{
1837
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1838

1839
	cpuset_update_task_memory_state();
P
Paul Jackson 已提交
1840 1841

	if (is_sched_load_balance(cs))
1842
		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
P
Paul Jackson 已提交
1843

1844
	number_of_cpusets--;
1845
	free_cpumask_var(cs->cpus_allowed);
1846
	kfree(cs);
L
Linus Torvalds 已提交
1847 1848
}

1849 1850 1851
struct cgroup_subsys cpuset_subsys = {
	.name = "cpuset",
	.create = cpuset_create,
1852
	.destroy = cpuset_destroy,
1853 1854 1855 1856 1857 1858 1859 1860
	.can_attach = cpuset_can_attach,
	.attach = cpuset_attach,
	.populate = cpuset_populate,
	.post_clone = cpuset_post_clone,
	.subsys_id = cpuset_subsys_id,
	.early_init = 1,
};

1861 1862 1863 1864 1865 1866 1867 1868
/*
 * cpuset_init_early - just enough so that the calls to
 * cpuset_update_task_memory_state() in early init code
 * are harmless.
 */

int __init cpuset_init_early(void)
{
1869 1870
	alloc_bootmem_cpumask_var(&top_cpuset.cpus_allowed);

1871
	top_cpuset.mems_generation = cpuset_mems_generation++;
1872 1873 1874
	return 0;
}

1875

L
Linus Torvalds 已提交
1876 1877 1878 1879 1880 1881 1882 1883
/**
 * cpuset_init - initialize cpusets at system boot
 *
 * Description: Initialize top_cpuset and the cpuset internal file system,
 **/

int __init cpuset_init(void)
{
1884
	int err = 0;
L
Linus Torvalds 已提交
1885

1886
	cpumask_setall(top_cpuset.cpus_allowed);
1887
	nodes_setall(top_cpuset.mems_allowed);
L
Linus Torvalds 已提交
1888

1889
	fmeter_init(&top_cpuset.fmeter);
1890
	top_cpuset.mems_generation = cpuset_mems_generation++;
P
Paul Jackson 已提交
1891
	set_bit(CS_SCHED_LOAD_BALANCE, &top_cpuset.flags);
1892
	top_cpuset.relax_domain_level = -1;
L
Linus Torvalds 已提交
1893 1894 1895

	err = register_filesystem(&cpuset_fs_type);
	if (err < 0)
1896 1897
		return err;

1898 1899 1900
	if (!alloc_cpumask_var(&cpus_attach, GFP_KERNEL))
		BUG();

1901
	number_of_cpusets = 1;
1902
	return 0;
L
Linus Torvalds 已提交
1903 1904
}

1905 1906 1907 1908 1909 1910 1911 1912
/**
 * cpuset_do_move_task - move a given task to another cpuset
 * @tsk: pointer to task_struct the task to move
 * @scan: struct cgroup_scanner contained in its struct cpuset_hotplug_scanner
 *
 * Called by cgroup_scan_tasks() for each task in a cgroup.
 * Return nonzero to stop the walk through the tasks.
 */
1913 1914
static void cpuset_do_move_task(struct task_struct *tsk,
				struct cgroup_scanner *scan)
1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
{
	struct cpuset_hotplug_scanner *chsp;

	chsp = container_of(scan, struct cpuset_hotplug_scanner, scan);
	cgroup_attach_task(chsp->to, tsk);
}

/**
 * move_member_tasks_to_cpuset - move tasks from one cpuset to another
 * @from: cpuset in which the tasks currently reside
 * @to: cpuset to which the tasks will be moved
 *
1927 1928
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 */
static void move_member_tasks_to_cpuset(struct cpuset *from, struct cpuset *to)
{
	struct cpuset_hotplug_scanner scan;

	scan.scan.cg = from->css.cgroup;
	scan.scan.test_task = NULL; /* select all tasks in cgroup */
	scan.scan.process_task = cpuset_do_move_task;
	scan.scan.heap = NULL;
	scan.to = to->css.cgroup;

L
Lai Jiangshan 已提交
1943
	if (cgroup_scan_tasks(&scan.scan))
1944 1945 1946 1947
		printk(KERN_ERR "move_member_tasks_to_cpuset: "
				"cgroup_scan_tasks failed\n");
}

1948
/*
1949
 * If CPU and/or memory hotplug handlers, below, unplug any CPUs
1950 1951
 * or memory nodes, we need to walk over the cpuset hierarchy,
 * removing that CPU or node from all cpusets.  If this removes the
1952 1953
 * last CPU or node from a cpuset, then move the tasks in the empty
 * cpuset to its next-highest non-empty parent.
1954
 *
1955 1956
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1957
 */
1958 1959 1960 1961
static void remove_tasks_in_empty_cpuset(struct cpuset *cs)
{
	struct cpuset *parent;

1962 1963 1964 1965 1966
	/*
	 * The cgroup's css_sets list is in use if there are tasks
	 * in the cpuset; the list is empty if there are none;
	 * the cs->css.refcnt seems always 0.
	 */
1967 1968
	if (list_empty(&cs->css.cgroup->css_sets))
		return;
1969

1970 1971 1972 1973 1974
	/*
	 * Find its next-highest non-empty parent, (top cpuset
	 * has online cpus, so can't be empty).
	 */
	parent = cs->parent;
1975
	while (cpumask_empty(parent->cpus_allowed) ||
1976
			nodes_empty(parent->mems_allowed))
1977 1978 1979 1980 1981 1982 1983 1984 1985
		parent = parent->parent;

	move_member_tasks_to_cpuset(cs, parent);
}

/*
 * Walk the specified cpuset subtree and look for empty cpusets.
 * The tasks of such cpuset must be moved to a parent cpuset.
 *
1986
 * Called with cgroup_mutex held.  We take callback_mutex to modify
1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
 * cpus_allowed and mems_allowed.
 *
 * This walk processes the tree from top to bottom, completing one layer
 * before dropping down to the next.  It always processes a node before
 * any of its children.
 *
 * For now, since we lack memory hot unplug, we'll never see a cpuset
 * that has tasks along with an empty 'mems'.  But if we did see such
 * a cpuset, we'd handle it just like we do if its 'cpus' was empty.
 */
1997
static void scan_for_empty_cpusets(struct cpuset *root)
1998
{
1999
	LIST_HEAD(queue);
2000 2001
	struct cpuset *cp;	/* scans cpusets being updated */
	struct cpuset *child;	/* scans child cpusets of cp */
2002
	struct cgroup *cont;
2003
	nodemask_t oldmems;
2004

2005 2006 2007
	list_add_tail((struct list_head *)&root->stack_list, &queue);

	while (!list_empty(&queue)) {
2008
		cp = list_first_entry(&queue, struct cpuset, stack_list);
2009 2010 2011 2012 2013
		list_del(queue.next);
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
			list_add_tail(&child->stack_list, &queue);
		}
2014 2015

		/* Continue past cpusets with all cpus, mems online */
2016
		if (cpumask_subset(cp->cpus_allowed, cpu_online_mask) &&
2017 2018 2019
		    nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY]))
			continue;

2020 2021
		oldmems = cp->mems_allowed;

2022
		/* Remove offline cpus and mems from this cpuset. */
2023
		mutex_lock(&callback_mutex);
2024 2025
		cpumask_and(cp->cpus_allowed, cp->cpus_allowed,
			    cpu_online_mask);
2026 2027
		nodes_and(cp->mems_allowed, cp->mems_allowed,
						node_states[N_HIGH_MEMORY]);
2028 2029 2030
		mutex_unlock(&callback_mutex);

		/* Move tasks from the empty cpuset to a parent */
2031
		if (cpumask_empty(cp->cpus_allowed) ||
2032
		     nodes_empty(cp->mems_allowed))
2033
			remove_tasks_in_empty_cpuset(cp);
2034
		else {
2035
			update_tasks_cpumask(cp, NULL);
2036 2037
			update_tasks_nodemask(cp, &oldmems);
		}
2038 2039 2040
	}
}

2041 2042 2043 2044 2045 2046
/*
 * The top_cpuset tracks what CPUs and Memory Nodes are online,
 * period.  This is necessary in order to make cpusets transparent
 * (of no affect) on systems that are actively using CPU hotplug
 * but making no active use of cpusets.
 *
2047 2048
 * This routine ensures that top_cpuset.cpus_allowed tracks
 * cpu_online_map on each CPU hotplug (cpuhp) event.
2049 2050 2051
 *
 * Called within get_online_cpus().  Needs to call cgroup_lock()
 * before calling generate_sched_domains().
2052
 */
2053
static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
P
Paul Jackson 已提交
2054
				unsigned long phase, void *unused_cpu)
2055
{
2056
	struct sched_domain_attr *attr;
2057
	struct cpumask *doms;
2058 2059
	int ndoms;

2060 2061 2062 2063 2064 2065
	switch (phase) {
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		break;
2066

2067
	default:
2068
		return NOTIFY_DONE;
2069
	}
2070

2071
	cgroup_lock();
2072
	cpumask_copy(top_cpuset.cpus_allowed, cpu_online_mask);
2073 2074 2075 2076 2077 2078 2079
	scan_for_empty_cpusets(&top_cpuset);
	ndoms = generate_sched_domains(&doms, &attr);
	cgroup_unlock();

	/* Have scheduler rebuild the domains */
	partition_sched_domains(ndoms, doms, attr);

2080
	return NOTIFY_OK;
2081 2082
}

2083
#ifdef CONFIG_MEMORY_HOTPLUG
2084
/*
2085
 * Keep top_cpuset.mems_allowed tracking node_states[N_HIGH_MEMORY].
2086 2087
 * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
 * See also the previous routine cpuset_track_online_cpus().
2088
 */
2089 2090
static int cpuset_track_online_nodes(struct notifier_block *self,
				unsigned long action, void *arg)
2091
{
2092
	cgroup_lock();
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103
	switch (action) {
	case MEM_ONLINE:
		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
		break;
	case MEM_OFFLINE:
		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
		scan_for_empty_cpusets(&top_cpuset);
		break;
	default:
		break;
	}
2104
	cgroup_unlock();
2105
	return NOTIFY_OK;
2106 2107 2108
}
#endif

L
Linus Torvalds 已提交
2109 2110 2111 2112 2113 2114 2115 2116
/**
 * cpuset_init_smp - initialize cpus_allowed
 *
 * Description: Finish top cpuset after cpu, node maps are initialized
 **/

void __init cpuset_init_smp(void)
{
2117
	cpumask_copy(top_cpuset.cpus_allowed, cpu_online_mask);
2118
	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2119

2120
	hotcpu_notifier(cpuset_track_online_cpus, 0);
2121
	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
2122 2123 2124

	cpuset_wq = create_singlethread_workqueue("cpuset");
	BUG_ON(!cpuset_wq);
L
Linus Torvalds 已提交
2125 2126 2127 2128 2129
}

/**
 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
2130
 * @pmask: pointer to struct cpumask variable to receive cpus_allowed set.
L
Linus Torvalds 已提交
2131
 *
2132
 * Description: Returns the cpumask_var_t cpus_allowed of the cpuset
L
Linus Torvalds 已提交
2133 2134 2135 2136 2137
 * attached to the specified @tsk.  Guaranteed to return some non-empty
 * subset of cpu_online_map, even if this means going outside the
 * tasks cpuset.
 **/

2138
void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask)
L
Linus Torvalds 已提交
2139
{
2140
	mutex_lock(&callback_mutex);
2141
	cpuset_cpus_allowed_locked(tsk, pmask);
2142 2143 2144 2145 2146
	mutex_unlock(&callback_mutex);
}

/**
 * cpuset_cpus_allowed_locked - return cpus_allowed mask from a tasks cpuset.
2147
 * Must be called with callback_mutex held.
2148
 **/
2149
void cpuset_cpus_allowed_locked(struct task_struct *tsk, struct cpumask *pmask)
2150
{
2151
	task_lock(tsk);
2152
	guarantee_online_cpus(task_cs(tsk), pmask);
2153
	task_unlock(tsk);
L
Linus Torvalds 已提交
2154 2155 2156 2157
}

void cpuset_init_current_mems_allowed(void)
{
2158
	nodes_setall(current->mems_allowed);
L
Linus Torvalds 已提交
2159 2160
}

2161 2162 2163 2164 2165 2166
/**
 * cpuset_mems_allowed - return mems_allowed mask from a tasks cpuset.
 * @tsk: pointer to task_struct from which to obtain cpuset->mems_allowed.
 *
 * Description: Returns the nodemask_t mems_allowed of the cpuset
 * attached to the specified @tsk.  Guaranteed to return some non-empty
2167
 * subset of node_states[N_HIGH_MEMORY], even if this means going outside the
2168 2169 2170 2171 2172 2173 2174
 * tasks cpuset.
 **/

nodemask_t cpuset_mems_allowed(struct task_struct *tsk)
{
	nodemask_t mask;

2175
	mutex_lock(&callback_mutex);
2176
	task_lock(tsk);
2177
	guarantee_online_mems(task_cs(tsk), &mask);
2178
	task_unlock(tsk);
2179
	mutex_unlock(&callback_mutex);
2180 2181 2182 2183

	return mask;
}

2184
/**
2185 2186
 * cpuset_nodemask_valid_mems_allowed - check nodemask vs. curremt mems_allowed
 * @nodemask: the nodemask to be checked
2187
 *
2188
 * Are any of the nodes in the nodemask allowed in current->mems_allowed?
L
Linus Torvalds 已提交
2189
 */
2190
int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask)
L
Linus Torvalds 已提交
2191
{
2192
	return nodes_intersects(*nodemask, current->mems_allowed);
L
Linus Torvalds 已提交
2193 2194
}

2195
/*
2196 2197 2198 2199
 * nearest_hardwall_ancestor() - Returns the nearest mem_exclusive or
 * mem_hardwall ancestor to the specified cpuset.  Call holding
 * callback_mutex.  If no ancestor is mem_exclusive or mem_hardwall
 * (an unusual configuration), then returns the root cpuset.
2200
 */
2201
static const struct cpuset *nearest_hardwall_ancestor(const struct cpuset *cs)
2202
{
2203
	while (!(is_mem_exclusive(cs) || is_mem_hardwall(cs)) && cs->parent)
2204 2205 2206 2207
		cs = cs->parent;
	return cs;
}

2208
/**
2209
 * cpuset_zone_allowed_softwall - Can we allocate on zone z's memory node?
2210
 * @z: is this zone on an allowed node?
2211
 * @gfp_mask: memory allocation flags
2212
 *
2213 2214
 * If we're in interrupt, yes, we can always allocate.  If
 * __GFP_THISNODE is set, yes, we can always allocate.  If zone
2215 2216
 * z's node is in our tasks mems_allowed, yes.  If it's not a
 * __GFP_HARDWALL request and this zone's nodes is in the nearest
2217
 * hardwalled cpuset ancestor to this tasks cpuset, yes.
2218 2219
 * If the task has been OOM killed and has access to memory reserves
 * as specified by the TIF_MEMDIE flag, yes.
2220 2221
 * Otherwise, no.
 *
2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235
 * If __GFP_HARDWALL is set, cpuset_zone_allowed_softwall()
 * reduces to cpuset_zone_allowed_hardwall().  Otherwise,
 * cpuset_zone_allowed_softwall() might sleep, and might allow a zone
 * from an enclosing cpuset.
 *
 * cpuset_zone_allowed_hardwall() only handles the simpler case of
 * hardwall cpusets, and never sleeps.
 *
 * The __GFP_THISNODE placement logic is really handled elsewhere,
 * by forcibly using a zonelist starting at a specified node, and by
 * (in get_page_from_freelist()) refusing to consider the zones for
 * any node on the zonelist except the first.  By the time any such
 * calls get to this routine, we should just shut up and say 'yes'.
 *
2236
 * GFP_USER allocations are marked with the __GFP_HARDWALL bit,
2237 2238
 * and do not allow allocations outside the current tasks cpuset
 * unless the task has been OOM killed as is marked TIF_MEMDIE.
2239
 * GFP_KERNEL allocations are not so marked, so can escape to the
2240
 * nearest enclosing hardwalled ancestor cpuset.
2241
 *
2242 2243 2244 2245 2246 2247 2248
 * Scanning up parent cpusets requires callback_mutex.  The
 * __alloc_pages() routine only calls here with __GFP_HARDWALL bit
 * _not_ set if it's a GFP_KERNEL allocation, and all nodes in the
 * current tasks mems_allowed came up empty on the first pass over
 * the zonelist.  So only GFP_KERNEL allocations, if all nodes in the
 * cpuset are short of memory, might require taking the callback_mutex
 * mutex.
2249
 *
2250
 * The first call here from mm/page_alloc:get_page_from_freelist()
2251 2252 2253
 * has __GFP_HARDWALL set in gfp_mask, enforcing hardwall cpusets,
 * so no allocation on a node outside the cpuset is allowed (unless
 * in interrupt, of course).
2254 2255 2256 2257 2258 2259
 *
 * The second pass through get_page_from_freelist() doesn't even call
 * here for GFP_ATOMIC calls.  For those calls, the __alloc_pages()
 * variable 'wait' is not set, and the bit ALLOC_CPUSET is not set
 * in alloc_flags.  That logic and the checks below have the combined
 * affect that:
2260 2261
 *	in_interrupt - any node ok (current task context irrelevant)
 *	GFP_ATOMIC   - any node ok
2262
 *	TIF_MEMDIE   - any node ok
2263
 *	GFP_KERNEL   - any node in enclosing hardwalled cpuset ok
2264
 *	GFP_USER     - only nodes in current tasks mems allowed ok.
2265 2266
 *
 * Rule:
2267
 *    Don't call cpuset_zone_allowed_softwall if you can't sleep, unless you
2268 2269
 *    pass in the __GFP_HARDWALL flag set in gfp_flag, which disables
 *    the code that might scan up ancestor cpusets and sleep.
2270
 */
2271

2272
int __cpuset_zone_allowed_softwall(struct zone *z, gfp_t gfp_mask)
L
Linus Torvalds 已提交
2273
{
2274 2275
	int node;			/* node that zone z is on */
	const struct cpuset *cs;	/* current cpuset ancestors */
2276
	int allowed;			/* is allocation in zone z allowed? */
2277

2278
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
2279
		return 1;
2280
	node = zone_to_nid(z);
2281
	might_sleep_if(!(gfp_mask & __GFP_HARDWALL));
2282 2283
	if (node_isset(node, current->mems_allowed))
		return 1;
2284 2285 2286 2287 2288 2289
	/*
	 * Allow tasks that have access to memory reserves because they have
	 * been OOM killed to get memory anywhere.
	 */
	if (unlikely(test_thread_flag(TIF_MEMDIE)))
		return 1;
2290 2291 2292
	if (gfp_mask & __GFP_HARDWALL)	/* If hardwall request, stop here */
		return 0;

2293 2294 2295
	if (current->flags & PF_EXITING) /* Let dying task have memory */
		return 1;

2296
	/* Not hardwall and node outside mems_allowed: scan up cpusets */
2297
	mutex_lock(&callback_mutex);
2298 2299

	task_lock(current);
2300
	cs = nearest_hardwall_ancestor(task_cs(current));
2301 2302
	task_unlock(current);

2303
	allowed = node_isset(node, cs->mems_allowed);
2304
	mutex_unlock(&callback_mutex);
2305
	return allowed;
L
Linus Torvalds 已提交
2306 2307
}

2308 2309 2310 2311 2312 2313 2314
/*
 * cpuset_zone_allowed_hardwall - Can we allocate on zone z's memory node?
 * @z: is this zone on an allowed node?
 * @gfp_mask: memory allocation flags
 *
 * If we're in interrupt, yes, we can always allocate.
 * If __GFP_THISNODE is set, yes, we can always allocate.  If zone
2315 2316 2317
 * z's node is in our tasks mems_allowed, yes.   If the task has been
 * OOM killed and has access to memory reserves as specified by the
 * TIF_MEMDIE flag, yes.  Otherwise, no.
2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
 *
 * The __GFP_THISNODE placement logic is really handled elsewhere,
 * by forcibly using a zonelist starting at a specified node, and by
 * (in get_page_from_freelist()) refusing to consider the zones for
 * any node on the zonelist except the first.  By the time any such
 * calls get to this routine, we should just shut up and say 'yes'.
 *
 * Unlike the cpuset_zone_allowed_softwall() variant, above,
 * this variant requires that the zone be in the current tasks
 * mems_allowed or that we're in interrupt.  It does not scan up the
 * cpuset hierarchy for the nearest enclosing mem_exclusive cpuset.
 * It never sleeps.
 */

int __cpuset_zone_allowed_hardwall(struct zone *z, gfp_t gfp_mask)
{
	int node;			/* node that zone z is on */

	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
		return 1;
	node = zone_to_nid(z);
	if (node_isset(node, current->mems_allowed))
		return 1;
D
Daniel Walker 已提交
2341 2342 2343 2344 2345 2346
	/*
	 * Allow tasks that have access to memory reserves because they have
	 * been OOM killed to get memory anywhere.
	 */
	if (unlikely(test_thread_flag(TIF_MEMDIE)))
		return 1;
2347 2348 2349
	return 0;
}

P
Paul Jackson 已提交
2350 2351 2352
/**
 * cpuset_lock - lock out any changes to cpuset structures
 *
2353
 * The out of memory (oom) code needs to mutex_lock cpusets
P
Paul Jackson 已提交
2354
 * from being changed while it scans the tasklist looking for a
2355
 * task in an overlapping cpuset.  Expose callback_mutex via this
P
Paul Jackson 已提交
2356 2357
 * cpuset_lock() routine, so the oom code can lock it, before
 * locking the task list.  The tasklist_lock is a spinlock, so
2358
 * must be taken inside callback_mutex.
P
Paul Jackson 已提交
2359 2360 2361 2362
 */

void cpuset_lock(void)
{
2363
	mutex_lock(&callback_mutex);
P
Paul Jackson 已提交
2364 2365 2366 2367 2368 2369 2370 2371 2372 2373
}

/**
 * cpuset_unlock - release lock on cpuset changes
 *
 * Undo the lock taken in a previous cpuset_lock() call.
 */

void cpuset_unlock(void)
{
2374
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
2375 2376
}

2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414
/**
 * cpuset_mem_spread_node() - On which node to begin search for a page
 *
 * If a task is marked PF_SPREAD_PAGE or PF_SPREAD_SLAB (as for
 * tasks in a cpuset with is_spread_page or is_spread_slab set),
 * and if the memory allocation used cpuset_mem_spread_node()
 * to determine on which node to start looking, as it will for
 * certain page cache or slab cache pages such as used for file
 * system buffers and inode caches, then instead of starting on the
 * local node to look for a free page, rather spread the starting
 * node around the tasks mems_allowed nodes.
 *
 * We don't have to worry about the returned node being offline
 * because "it can't happen", and even if it did, it would be ok.
 *
 * The routines calling guarantee_online_mems() are careful to
 * only set nodes in task->mems_allowed that are online.  So it
 * should not be possible for the following code to return an
 * offline node.  But if it did, that would be ok, as this routine
 * is not returning the node where the allocation must be, only
 * the node where the search should start.  The zonelist passed to
 * __alloc_pages() will include all nodes.  If the slab allocator
 * is passed an offline node, it will fall back to the local node.
 * See kmem_cache_alloc_node().
 */

int cpuset_mem_spread_node(void)
{
	int node;

	node = next_node(current->cpuset_mem_spread_rotor, current->mems_allowed);
	if (node == MAX_NUMNODES)
		node = first_node(current->mems_allowed);
	current->cpuset_mem_spread_rotor = node;
	return node;
}
EXPORT_SYMBOL_GPL(cpuset_mem_spread_node);

2415
/**
2416 2417 2418 2419 2420 2421 2422 2423
 * cpuset_mems_allowed_intersects - Does @tsk1's mems_allowed intersect @tsk2's?
 * @tsk1: pointer to task_struct of some task.
 * @tsk2: pointer to task_struct of some other task.
 *
 * Description: Return true if @tsk1's mems_allowed intersects the
 * mems_allowed of @tsk2.  Used by the OOM killer to determine if
 * one of the task's memory usage might impact the memory available
 * to the other.
2424 2425
 **/

2426 2427
int cpuset_mems_allowed_intersects(const struct task_struct *tsk1,
				   const struct task_struct *tsk2)
2428
{
2429
	return nodes_intersects(tsk1->mems_allowed, tsk2->mems_allowed);
2430 2431
}

2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
/**
 * cpuset_print_task_mems_allowed - prints task's cpuset and mems_allowed
 * @task: pointer to task_struct of some task.
 *
 * Description: Prints @task's name, cpuset name, and cached copy of its
 * mems_allowed to the kernel log.  Must hold task_lock(task) to allow
 * dereferencing task_cs(task).
 */
void cpuset_print_task_mems_allowed(struct task_struct *tsk)
{
	struct dentry *dentry;

	dentry = task_cs(tsk)->css.cgroup->dentry;
	spin_lock(&cpuset_buffer_lock);
	snprintf(cpuset_name, CPUSET_NAME_LEN,
		 dentry ? (const char *)dentry->d_name.name : "/");
	nodelist_scnprintf(cpuset_nodelist, CPUSET_NODELIST_LEN,
			   tsk->mems_allowed);
	printk(KERN_INFO "%s cpuset=%s mems_allowed=%s\n",
	       tsk->comm, cpuset_name, cpuset_nodelist);
	spin_unlock(&cpuset_buffer_lock);
}

2455 2456 2457 2458 2459 2460
/*
 * Collection of memory_pressure is suppressed unless
 * this flag is enabled by writing "1" to the special
 * cpuset file 'memory_pressure_enabled' in the root cpuset.
 */

2461
int cpuset_memory_pressure_enabled __read_mostly;
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483

/**
 * cpuset_memory_pressure_bump - keep stats of per-cpuset reclaims.
 *
 * Keep a running average of the rate of synchronous (direct)
 * page reclaim efforts initiated by tasks in each cpuset.
 *
 * This represents the rate at which some task in the cpuset
 * ran low on memory on all nodes it was allowed to use, and
 * had to enter the kernels page reclaim code in an effort to
 * create more free memory by tossing clean pages or swapping
 * or writing dirty pages.
 *
 * Display to user space in the per-cpuset read-only file
 * "memory_pressure".  Value displayed is an integer
 * representing the recent rate of entry into the synchronous
 * (direct) page reclaim by any task attached to the cpuset.
 **/

void __cpuset_memory_pressure_bump(void)
{
	task_lock(current);
2484
	fmeter_markevent(&task_cs(current)->fmeter);
2485 2486 2487
	task_unlock(current);
}

2488
#ifdef CONFIG_PROC_PID_CPUSET
L
Linus Torvalds 已提交
2489 2490 2491 2492
/*
 * proc_cpuset_show()
 *  - Print tasks cpuset path into seq_file.
 *  - Used for /proc/<pid>/cpuset.
2493 2494
 *  - No need to task_lock(tsk) on this tsk->cpuset reference, as it
 *    doesn't really matter if tsk->cpuset changes after we read it,
2495
 *    and we take cgroup_mutex, keeping cpuset_attach() from changing it
2496
 *    anyway.
L
Linus Torvalds 已提交
2497
 */
P
Paul Jackson 已提交
2498
static int proc_cpuset_show(struct seq_file *m, void *unused_v)
L
Linus Torvalds 已提交
2499
{
2500
	struct pid *pid;
L
Linus Torvalds 已提交
2501 2502
	struct task_struct *tsk;
	char *buf;
2503
	struct cgroup_subsys_state *css;
2504
	int retval;
L
Linus Torvalds 已提交
2505

2506
	retval = -ENOMEM;
L
Linus Torvalds 已提交
2507 2508
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
2509 2510 2511
		goto out;

	retval = -ESRCH;
2512 2513
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
2514 2515
	if (!tsk)
		goto out_free;
L
Linus Torvalds 已提交
2516

2517
	retval = -EINVAL;
2518 2519 2520
	cgroup_lock();
	css = task_subsys_state(tsk, cpuset_subsys_id);
	retval = cgroup_path(css->cgroup, buf, PAGE_SIZE);
L
Linus Torvalds 已提交
2521
	if (retval < 0)
2522
		goto out_unlock;
L
Linus Torvalds 已提交
2523 2524
	seq_puts(m, buf);
	seq_putc(m, '\n');
2525
out_unlock:
2526
	cgroup_unlock();
2527 2528
	put_task_struct(tsk);
out_free:
L
Linus Torvalds 已提交
2529
	kfree(buf);
2530
out:
L
Linus Torvalds 已提交
2531 2532 2533 2534 2535
	return retval;
}

static int cpuset_open(struct inode *inode, struct file *file)
{
2536 2537
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cpuset_show, pid);
L
Linus Torvalds 已提交
2538 2539
}

2540
const struct file_operations proc_cpuset_operations = {
L
Linus Torvalds 已提交
2541 2542 2543 2544 2545
	.open		= cpuset_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};
2546
#endif /* CONFIG_PROC_PID_CPUSET */
L
Linus Torvalds 已提交
2547 2548

/* Display task cpus_allowed, mems_allowed in /proc/<pid>/status file. */
2549 2550 2551
void cpuset_task_status_allowed(struct seq_file *m, struct task_struct *task)
{
	seq_printf(m, "Cpus_allowed:\t");
2552
	seq_cpumask(m, &task->cpus_allowed);
2553
	seq_printf(m, "\n");
2554
	seq_printf(m, "Cpus_allowed_list:\t");
2555
	seq_cpumask_list(m, &task->cpus_allowed);
2556
	seq_printf(m, "\n");
2557
	seq_printf(m, "Mems_allowed:\t");
2558
	seq_nodemask(m, &task->mems_allowed);
2559
	seq_printf(m, "\n");
2560
	seq_printf(m, "Mems_allowed_list:\t");
2561
	seq_nodemask_list(m, &task->mems_allowed);
2562
	seq_printf(m, "\n");
L
Linus Torvalds 已提交
2563
}