cpuset.c 73 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6
/*
 *  kernel/cpuset.c
 *
 *  Processor and Memory placement constraints for sets of tasks.
 *
 *  Copyright (C) 2003 BULL SA.
Paul Jackson's avatar
Paul Jackson committed
7
 *  Copyright (C) 2004-2007 Silicon Graphics, Inc.
8
 *  Copyright (C) 2006 Google, Inc
Linus Torvalds's avatar
Linus Torvalds committed
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.
Linus Torvalds's avatar
Linus Torvalds committed
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
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
38
#include <linux/mm.h>
39
#include <linux/memory.h>
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
46 47
#include <linux/sched.h>
#include <linux/seq_file.h>
48
#include <linux/security.h>
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
62

63 64 65 66 67
/*
 * Tracks how many cpusets are currently defined in system.
 * When there is only one cpuset (the root cpuset) we can
 * short circuit some hooks.
 */
68
int number_of_cpusets __read_mostly;
69

70
/* Forward declare cgroup structures */
71 72 73
struct cgroup_subsys cpuset_subsys;
struct cpuset;

74 75 76 77 78 79 80 81 82
/* 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 */
};

Linus Torvalds's avatar
Linus Torvalds committed
83
struct cpuset {
84 85
	struct cgroup_subsys_state css;

Linus Torvalds's avatar
Linus Torvalds committed
86
	unsigned long flags;		/* "unsigned long" so bitops work */
87
	cpumask_var_t cpus_allowed;	/* CPUs allowed to tasks in cpuset */
Linus Torvalds's avatar
Linus Torvalds committed
88 89 90 91 92 93 94 95
	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.
	 */
96 97 98
	int mems_generation;

	struct fmeter fmeter;		/* memory_pressure filter */
Paul Jackson's avatar
Paul Jackson committed
99 100 101

	/* partition number for rebuild_sched_domains() */
	int pn;
102

103 104 105
	/* for custom sched domain */
	int relax_domain_level;

106 107
	/* used for walking a cpuset heirarchy */
	struct list_head stack_list;
Linus Torvalds's avatar
Linus Torvalds committed
108 109
};

110 111 112 113 114 115 116 117 118 119 120 121 122
/* 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);
}
123 124 125 126
struct cpuset_hotplug_scanner {
	struct cgroup_scanner scan;
	struct cgroup *to;
};
127

Linus Torvalds's avatar
Linus Torvalds committed
128 129 130 131
/* bits in struct cpuset flags field */
typedef enum {
	CS_CPU_EXCLUSIVE,
	CS_MEM_EXCLUSIVE,
132
	CS_MEM_HARDWALL,
133
	CS_MEMORY_MIGRATE,
Paul Jackson's avatar
Paul Jackson committed
134
	CS_SCHED_LOAD_BALANCE,
135 136
	CS_SPREAD_PAGE,
	CS_SPREAD_SLAB,
Linus Torvalds's avatar
Linus Torvalds committed
137 138 139 140 141
} cpuset_flagbits_t;

/* convenient tests for these bits */
static inline int is_cpu_exclusive(const struct cpuset *cs)
{
142
	return test_bit(CS_CPU_EXCLUSIVE, &cs->flags);
Linus Torvalds's avatar
Linus Torvalds committed
143 144 145 146
}

static inline int is_mem_exclusive(const struct cpuset *cs)
{
147
	return test_bit(CS_MEM_EXCLUSIVE, &cs->flags);
Linus Torvalds's avatar
Linus Torvalds committed
148 149
}

150 151 152 153 154
static inline int is_mem_hardwall(const struct cpuset *cs)
{
	return test_bit(CS_MEM_HARDWALL, &cs->flags);
}

Paul Jackson's avatar
Paul Jackson committed
155 156 157 158 159
static inline int is_sched_load_balance(const struct cpuset *cs)
{
	return test_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
}

160 161
static inline int is_memory_migrate(const struct cpuset *cs)
{
162
	return test_bit(CS_MEMORY_MIGRATE, &cs->flags);
163 164
}

165 166 167 168 169 170 171 172 173 174
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);
}

Linus Torvalds's avatar
Linus Torvalds committed
175
/*
176
 * Increment this integer everytime any cpuset changes its
Linus Torvalds's avatar
Linus Torvalds committed
177 178 179 180
 * 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.
 *
181
 * A single, global generation is needed because cpuset_attach_task() could
Linus Torvalds's avatar
Linus Torvalds committed
182 183 184 185
 * 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
186
 * modify another's memory placement.  So we must enable every task,
Linus Torvalds's avatar
Linus Torvalds committed
187 188 189
 * 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.
190
 *
191
 * Since writes to cpuset_mems_generation are guarded by the cgroup lock
192
 * there is no need to mark it atomic.
Linus Torvalds's avatar
Linus Torvalds committed
193
 */
194
static int cpuset_mems_generation;
Linus Torvalds's avatar
Linus Torvalds committed
195 196 197 198 199 200

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

/*
201 202 203 204 205 206 207
 * 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.
208
 *
209
 * A task must hold both mutexes to modify cpusets.  If a task
210
 * holds cgroup_mutex, then it blocks others wanting that mutex,
211
 * ensuring that it is the only task able to also acquire callback_mutex
212 213
 * and be able to modify cpusets.  It can perform various checks on
 * the cpuset structure first, knowing nothing will change.  It can
214
 * also allocate memory while just holding cgroup_mutex.  While it is
215
 * performing these checks, various callback routines can briefly
216 217
 * acquire callback_mutex to query cpusets.  Once it is ready to make
 * the changes, it takes callback_mutex, blocking everyone else.
218 219
 *
 * Calls to the kernel memory allocator can not be made while holding
220
 * callback_mutex, as that would risk double tripping on callback_mutex
221 222 223
 * from one of the callbacks into the cpuset code from within
 * __alloc_pages().
 *
224
 * If a task is only holding callback_mutex, then it has read-only
225 226 227 228 229
 * 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.
 *
230
 * The cpuset_common_file_read() handlers only hold callback_mutex across
231 232 233
 * small pieces of code, such as when reading out possibly multi-word
 * cpumasks and nodemasks.
 *
234 235
 * Accessing a task's cpuset should be done in accordance with the
 * guidelines for accessing subsystem state in kernel/cgroup.c
Linus Torvalds's avatar
Linus Torvalds committed
236 237
 */

238
static DEFINE_MUTEX(callback_mutex);
239

240 241 242 243 244 245 246 247 248 249 250
/*
 * 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);

251 252
/*
 * This is ugly, but preserves the userspace API for existing cpuset
253
 * users. If someone tries to mount the "cpuset" filesystem, we
254 255
 * silently switch it to mount "cgroup" instead
 */
256 257 258
static int cpuset_get_sb(struct file_system_type *fs_type,
			 int flags, const char *unused_dev_name,
			 void *data, struct vfsmount *mnt)
Linus Torvalds's avatar
Linus Torvalds committed
259
{
260 261 262 263 264 265 266 267 268 269 270
	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;
Linus Torvalds's avatar
Linus Torvalds committed
271 272 273 274 275 276 277 278
}

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

/*
279
 * Return in pmask the portion of a cpusets's cpus_allowed that
Linus Torvalds's avatar
Linus Torvalds committed
280 281 282 283 284 285 286 287 288
 * 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.
 *
289
 * Call with callback_mutex held.
Linus Torvalds's avatar
Linus Torvalds committed
290 291 292 293
 */

static void guarantee_online_cpus(const struct cpuset *cs, cpumask_t *pmask)
{
294
	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
Linus Torvalds's avatar
Linus Torvalds committed
295 296
		cs = cs->parent;
	if (cs)
297
		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
Linus Torvalds's avatar
Linus Torvalds committed
298
	else
299 300
		cpumask_copy(pmask, cpu_online_mask);
	BUG_ON(!cpumask_intersects(pmask, cpu_online_mask));
Linus Torvalds's avatar
Linus Torvalds committed
301 302 303 304
}

/*
 * Return in *pmask the portion of a cpusets's mems_allowed that
305 306 307 308
 * 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].
Linus Torvalds's avatar
Linus Torvalds committed
309 310
 *
 * One way or another, we guarantee to return some non-empty subset
311
 * of node_states[N_HIGH_MEMORY].
Linus Torvalds's avatar
Linus Torvalds committed
312
 *
313
 * Call with callback_mutex held.
Linus Torvalds's avatar
Linus Torvalds committed
314 315 316 317
 */

static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
{
318 319
	while (cs && !nodes_intersects(cs->mems_allowed,
					node_states[N_HIGH_MEMORY]))
Linus Torvalds's avatar
Linus Torvalds committed
320 321
		cs = cs->parent;
	if (cs)
322 323
		nodes_and(*pmask, cs->mems_allowed,
					node_states[N_HIGH_MEMORY]);
Linus Torvalds's avatar
Linus Torvalds committed
324
	else
325 326
		*pmask = node_states[N_HIGH_MEMORY];
	BUG_ON(!nodes_intersects(*pmask, node_states[N_HIGH_MEMORY]));
Linus Torvalds's avatar
Linus Torvalds committed
327 328
}

329 330 331 332 333 334
/**
 * 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.
335
 *
336 337 338 339
 * 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().
 *
340
 * Call without callback_mutex or task_lock() held.  May be
341 342
 * called with or without cgroup_mutex held.  Thanks in part to
 * 'the_top_cpuset_hack', the task's cpuset pointer will never
David Rientjes's avatar
David Rientjes committed
343 344
 * be NULL.  This routine also might acquire callback_mutex during
 * call.
345
 *
346 347
 * Reading current->cpuset->mems_generation doesn't need task_lock
 * to guard the current->cpuset derefence, because it is guarded
348
 * from concurrent freeing of current->cpuset using RCU.
349 350 351 352 353 354 355 356 357 358 359 360 361 362
 *
 * 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.
363 364 365 366 367
 *
 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
368 369
 */

370
void cpuset_update_task_memory_state(void)
Linus Torvalds's avatar
Linus Torvalds committed
371
{
372
	int my_cpusets_mem_gen;
373
	struct task_struct *tsk = current;
374
	struct cpuset *cs;
375

376 377 378
	rcu_read_lock();
	my_cpusets_mem_gen = task_cs(tsk)->mems_generation;
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
379

380
	if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) {
381
		mutex_lock(&callback_mutex);
382
		task_lock(tsk);
383
		cs = task_cs(tsk); /* Maybe changed when task not locked */
384 385
		guarantee_online_mems(cs, &tsk->mems_allowed);
		tsk->cpuset_mems_generation = cs->mems_generation;
386 387 388 389 390 391 392 393
		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;
394
		task_unlock(tsk);
395
		mutex_unlock(&callback_mutex);
396
		mpol_rebind_task(tsk, &tsk->mems_allowed);
Linus Torvalds's avatar
Linus Torvalds committed
397 398 399 400 401 402 403 404
	}
}

/*
 * 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
405
 * are only set if the other's are set.  Call holding cgroup_mutex.
Linus Torvalds's avatar
Linus Torvalds committed
406 407 408 409
 */

static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
{
410
	return	cpumask_subset(p->cpus_allowed, q->cpus_allowed) &&
Linus Torvalds's avatar
Linus Torvalds committed
411 412 413 414 415
		nodes_subset(p->mems_allowed, q->mems_allowed) &&
		is_cpu_exclusive(p) <= is_cpu_exclusive(q) &&
		is_mem_exclusive(p) <= is_mem_exclusive(q);
}

416 417 418 419 420 421
/**
 * 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)
{
422 423 424 425 426 427 428 429 430 431 432 433 434
	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;
435 436 437 438 439 440 441 442
}

/**
 * free_trial_cpuset - free the trial cpuset
 * @trial: the trial cpuset to be freed
 */
static void free_trial_cpuset(struct cpuset *trial)
{
443
	free_cpumask_var(trial->cpus_allowed);
444 445 446
	kfree(trial);
}

Linus Torvalds's avatar
Linus Torvalds committed
447 448 449 450 451 452 453
/*
 * 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
454
 * cgroup_mutex held.
Linus Torvalds's avatar
Linus Torvalds committed
455 456 457 458 459 460 461 462 463 464 465 466 467 468
 *
 * '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)
{
469
	struct cgroup *cont;
Linus Torvalds's avatar
Linus Torvalds committed
470 471 472
	struct cpuset *c, *par;

	/* Each of our child cpusets must be a subset of us */
473 474
	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
		if (!is_cpuset_subset(cgroup_cs(cont), trial))
Linus Torvalds's avatar
Linus Torvalds committed
475 476 477 478
			return -EBUSY;
	}

	/* Remaining checks don't apply to root cpuset */
479
	if (cur == &top_cpuset)
Linus Torvalds's avatar
Linus Torvalds committed
480 481
		return 0;

482 483
	par = cur->parent;

Linus Torvalds's avatar
Linus Torvalds committed
484 485 486 487
	/* We must be a subset of our parent cpuset */
	if (!is_cpuset_subset(trial, par))
		return -EACCES;

488 489 490 491
	/*
	 * If either I or some sibling (!= me) is exclusive, we can't
	 * overlap
	 */
492 493
	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
		c = cgroup_cs(cont);
Linus Torvalds's avatar
Linus Torvalds committed
494 495
		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
		    c != cur &&
496
		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
Linus Torvalds's avatar
Linus Torvalds committed
497 498 499 500 501 502 503
			return -EINVAL;
		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
		    c != cur &&
		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
			return -EINVAL;
	}

504 505
	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
	if (cgroup_task_count(cur->css.cgroup)) {
506
		if (cpumask_empty(trial->cpus_allowed) ||
507 508 509 510 511
		    nodes_empty(trial->mems_allowed)) {
			return -ENOSPC;
		}
	}

Linus Torvalds's avatar
Linus Torvalds committed
512 513 514
	return 0;
}

Paul Jackson's avatar
Paul Jackson committed
515
/*
516
 * Helper routine for generate_sched_domains().
Paul Jackson's avatar
Paul Jackson committed
517 518 519 520
 * Do cpusets a, b have overlapping cpus_allowed masks?
 */
static int cpusets_overlap(struct cpuset *a, struct cpuset *b)
{
521
	return cpumask_intersects(a->cpus_allowed, b->cpus_allowed);
Paul Jackson's avatar
Paul Jackson committed
522 523
}

524 525 526 527 528 529 530 531
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;
}

532 533 534 535 536 537 538 539 540 541 542 543 544 545
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);

546
		if (cpumask_empty(cp->cpus_allowed))
547 548 549 550 551 552 553 554 555 556 557 558
			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);
		}
	}
}

Paul Jackson's avatar
Paul Jackson committed
559
/*
560 561 562 563 564 565 566 567 568
 * 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.
Paul Jackson's avatar
Paul Jackson committed
569 570 571 572 573 574 575 576 577
 *
 * See "What is sched_load_balance" in Documentation/cpusets.txt
 * 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.
 *
578
 * Must be called with cgroup_lock held.
Paul Jackson's avatar
Paul Jackson committed
579 580
 *
 * The three key local variables below are:
581
 *    q  - a linked-list queue of cpuset pointers, used to implement a
Paul Jackson's avatar
Paul Jackson committed
582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
 *	   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().
 */
613 614
static int generate_sched_domains(cpumask_t **domains,
			struct sched_domain_attr **attributes)
Paul Jackson's avatar
Paul Jackson committed
615
{
616
	LIST_HEAD(q);		/* queue of cpusets to be scanned */
Paul Jackson's avatar
Paul Jackson committed
617 618 619 620 621
	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 */
	cpumask_t *doms;	/* resulting partition; i.e. sched domains */
622
	struct sched_domain_attr *dattr;  /* attributes for custom domains */
623
	int ndoms = 0;		/* number of sched domains in result */
Paul Jackson's avatar
Paul Jackson committed
624 625 626
	int nslot;		/* next empty doms[] cpumask_t slot */

	doms = NULL;
627
	dattr = NULL;
628
	csa = NULL;
Paul Jackson's avatar
Paul Jackson committed
629 630 631 632 633

	/* Special case for the 99% of systems with one, full, sched domain */
	if (is_sched_load_balance(&top_cpuset)) {
		doms = kmalloc(sizeof(cpumask_t), GFP_KERNEL);
		if (!doms)
634 635
			goto done;

636 637 638
		dattr = kmalloc(sizeof(struct sched_domain_attr), GFP_KERNEL);
		if (dattr) {
			*dattr = SD_ATTR_INIT;
639
			update_domain_attr_tree(dattr, &top_cpuset);
640
		}
641
		cpumask_copy(doms, top_cpuset.cpus_allowed);
642 643 644

		ndoms = 1;
		goto done;
Paul Jackson's avatar
Paul Jackson committed
645 646 647 648 649 650 651
	}

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

652 653
	list_add(&top_cpuset.stack_list, &q);
	while (!list_empty(&q)) {
Paul Jackson's avatar
Paul Jackson committed
654 655
		struct cgroup *cont;
		struct cpuset *child;   /* scans child cpusets of cp */
656

657 658 659
		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

660
		if (cpumask_empty(cp->cpus_allowed))
661 662
			continue;

663 664 665 666 667 668 669
		/*
		 * 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)) {
Paul Jackson's avatar
Paul Jackson committed
670
			csa[csn++] = cp;
671 672
			continue;
		}
673

Paul Jackson's avatar
Paul Jackson committed
674 675
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
676
			list_add_tail(&child->stack_list, &q);
Paul Jackson's avatar
Paul Jackson committed
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706
		}
  	}

	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;
			}
		}
	}

707 708 709 710
	/*
	 * Now we know how many domains to create.
	 * Convert <csn, csa> to <ndoms, doms> and populate cpu masks.
	 */
Paul Jackson's avatar
Paul Jackson committed
711
	doms = kmalloc(ndoms * sizeof(cpumask_t), GFP_KERNEL);
712
	if (!doms)
713 714 715 716 717 718
		goto done;

	/*
	 * The rest of the code, including the scheduler, can deal with
	 * dattr==NULL case. No need to abort if alloc fails.
	 */
719
	dattr = kmalloc(ndoms * sizeof(struct sched_domain_attr), GFP_KERNEL);
Paul Jackson's avatar
Paul Jackson committed
720 721 722

	for (nslot = 0, i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
723
		cpumask_t *dp;
Paul Jackson's avatar
Paul Jackson committed
724 725
		int apn = a->pn;

726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741
		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--;
Paul Jackson's avatar
Paul Jackson committed
742
			}
743 744
			continue;
		}
Paul Jackson's avatar
Paul Jackson committed
745

746 747 748 749 750 751 752
		cpus_clear(*dp);
		if (dattr)
			*(dattr + nslot) = SD_ATTR_INIT;
		for (j = i; j < csn; j++) {
			struct cpuset *b = csa[j];

			if (apn == b->pn) {
753
				cpumask_or(dp, dp, b->cpus_allowed);
754 755 756 757 758
				if (dattr)
					update_domain_attr_tree(dattr + nslot, b);

				/* Done with this partition */
				b->pn = -1;
Paul Jackson's avatar
Paul Jackson committed
759 760
			}
		}
761
		nslot++;
Paul Jackson's avatar
Paul Jackson committed
762 763 764
	}
	BUG_ON(nslot != ndoms);

765 766 767
done:
	kfree(csa);

768 769 770 771 772 773 774
	/*
	 * Fallback to the default domain if kmalloc() failed.
	 * See comments in partition_sched_domains().
	 */
	if (doms == NULL)
		ndoms = 1;

775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795
	*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;
	cpumask_t *doms;
	int ndoms;

796
	get_online_cpus();
797 798 799 800 801 802 803 804 805

	/* 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);

806
	put_online_cpus();
807
}
Paul Jackson's avatar
Paul Jackson committed
808

809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846
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)
{
	schedule_work(&rebuild_sched_domains_work);
}

/*
 * 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);
Paul Jackson's avatar
Paul Jackson committed
847 848
}

849 850 851 852 853
/**
 * 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
 *
854
 * Call with cgroup_mutex held.  May take callback_mutex during call.
855 856 857
 * 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).
858
 */
859 860
static int cpuset_test_cpumask(struct task_struct *tsk,
			       struct cgroup_scanner *scan)
861
{
862
	return !cpumask_equal(&tsk->cpus_allowed,
863 864
			(cgroup_cs(scan->cg))->cpus_allowed);
}
865

866 867 868 869 870 871 872 873 874 875 876
/**
 * 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.
 */
877 878
static void cpuset_change_cpumask(struct task_struct *tsk,
				  struct cgroup_scanner *scan)
879
{
880
	set_cpus_allowed_ptr(tsk, ((cgroup_cs(scan->cg))->cpus_allowed));
881 882
}

883 884 885
/**
 * 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
886
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
887 888 889 890 891 892
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
893 894
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
895
 */
896
static void update_tasks_cpumask(struct cpuset *cs, struct ptr_heap *heap)
897 898 899 900 901 902
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = cpuset_test_cpumask;
	scan.process_task = cpuset_change_cpumask;
903 904
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
905 906
}

907 908 909 910 911
/**
 * 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
 */
912 913
static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
			  const char *buf)
Linus Torvalds's avatar
Linus Torvalds committed
914
{
915
	struct ptr_heap heap;
916 917
	int retval;
	int is_load_balanced;
Linus Torvalds's avatar
Linus Torvalds committed
918

919 920 921 922
	/* top_cpuset.cpus_allowed tracks cpu_online_map; it's read-only */
	if (cs == &top_cpuset)
		return -EACCES;

923
	/*
924
	 * An empty cpus_allowed is ok only if the cpuset has no tasks.
925 926 927
	 * Since cpulist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have cpus.
928
	 */
929
	if (!*buf) {
930
		cpumask_clear(trialcs->cpus_allowed);
931
	} else {
932
		retval = cpulist_parse(buf, trialcs->cpus_allowed);
933 934
		if (retval < 0)
			return retval;
935

936
		if (!cpumask_subset(trialcs->cpus_allowed, cpu_online_mask))
937
			return -EINVAL;
938
	}
939
	retval = validate_change(cs, trialcs);
940 941
	if (retval < 0)
		return retval;
Paul Jackson's avatar
Paul Jackson committed
942

Paul Menage's avatar
Paul Menage committed
943
	/* Nothing to do if the cpus didn't change */
944
	if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed))
Paul Menage's avatar
Paul Menage committed
945
		return 0;
946

947 948 949 950
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval)
		return retval;

951
	is_load_balanced = is_sched_load_balance(trialcs);
Paul Jackson's avatar
Paul Jackson committed
952

953
	mutex_lock(&callback_mutex);
954
	cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed);
955
	mutex_unlock(&callback_mutex);
Paul Jackson's avatar
Paul Jackson committed
956

Paul Menage's avatar
Paul Menage committed
957 958
	/*
	 * Scan tasks in the cpuset, and update the cpumasks of any
959
	 * that need an update.
Paul Menage's avatar
Paul Menage committed
960
	 */
961 962 963
	update_tasks_cpumask(cs, &heap);

	heap_free(&heap);
964

Paul Menage's avatar
Paul Menage committed
965
	if (is_load_balanced)
966
		async_rebuild_sched_domains();
967
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
968 969
}

970 971 972 973 974 975 976 977
/*
 * 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.
 *
978
 *    Call holding cgroup_mutex, so current's cpuset won't change
979
 *    during this call, as manage_mutex holds off any cpuset_attach()
980 981
 *    calls.  Therefore we don't need to take task_lock around the
 *    call to guarantee_online_mems(), as we know no one is changing
982
 *    our task's cpuset.
983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
 *
 *    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);
1015
	guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
1016 1017 1018
	mutex_unlock(&callback_mutex);
}

1019 1020
static void *cpuset_being_rebound;

1021 1022 1023 1024 1025 1026 1027 1028 1029
/**
 * 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)
Linus Torvalds's avatar
Linus Torvalds committed
1030
{
1031
	struct task_struct *p;
1032 1033
	struct mm_struct **mmarray;
	int i, n, ntasks;
1034
	int migrate;
1035
	int fudge;
1036
	struct cgroup_iter it;
1037
	int retval;
1038

1039
	cpuset_being_rebound = cs;		/* causes mpol_dup() rebind */
1040 1041

	fudge = 10;				/* spare mmarray[] slots */
1042
	fudge += cpumask_weight(cs->cpus_allowed);/* imagine 1 fork-bomb/cpu */
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
	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) {
1053
		ntasks = cgroup_task_count(cs->css.cgroup);  /* guess */
1054 1055 1056 1057
		ntasks += fudge;
		mmarray = kmalloc(ntasks * sizeof(*mmarray), GFP_KERNEL);
		if (!mmarray)
			goto done;
1058
		read_lock(&tasklist_lock);		/* block fork */
1059
		if (cgroup_task_count(cs->css.cgroup) <= ntasks)
1060
			break;				/* got enough */
1061
		read_unlock(&tasklist_lock);		/* try again */
1062 1063 1064 1065 1066 1067
		kfree(mmarray);
	}

	n = 0;

	/* Load up mmarray[] with mm reference for each task in cpuset. */
1068 1069
	cgroup_iter_start(cs->css.cgroup, &it);
	while ((p = cgroup_iter_next(cs->css.cgroup, &it))) {
1070 1071 1072 1073 1074
		struct mm_struct *mm;

		if (n >= ntasks) {
			printk(KERN_WARNING
				"Cpuset mempolicy rebind incomplete.\n");
1075
			break;
1076 1077 1078 1079 1080
		}
		mm = get_task_mm(p);
		if (!mm)
			continue;
		mmarray[n++] = mm;
1081 1082
	}
	cgroup_iter_end(cs->css.cgroup, &it);
1083
	read_unlock(&tasklist_lock);
1084 1085 1086 1087 1088 1089

	/*
	 * 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
1090
	 * tasklist_lock.  Forks can happen again now - the mpol_dup()
1091 1092
	 * cpuset_being_rebound check will catch such forks, and rebind
	 * their vma mempolicies too.  Because we still hold the global
1093
	 * cgroup_mutex, we know that no other rebind effort will
1094 1095
	 * be contending for the global variable cpuset_being_rebound.
	 * It's ok if we rebind the same mm twice; mpol_rebind_mm()
1096
	 * is idempotent.  Also migrate pages in each mm to new nodes.
1097
	 */
1098
	migrate = is_memory_migrate(cs);
1099 1100 1101 1102
	for (i = 0; i < n; i++) {
		struct mm_struct *mm = mmarray[i];

		mpol_rebind_mm(mm, &cs->mems_allowed);
1103
		if (migrate)
1104
			cpuset_migrate_mm(mm, oldmem, &cs->mems_allowed);
1105 1106 1107
		mmput(mm);
	}

1108
	/* We're done rebinding vmas to this cpuset's new mems_allowed. */
1109
	kfree(mmarray);
1110
	cpuset_being_rebound = NULL;
1111
	retval = 0;
1112
done:
Linus Torvalds's avatar
Linus Torvalds committed
1113 1114 1115
	return retval;
}

1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
/*
 * 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.
 */
1129 1130
static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs,
			   const char *buf)
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
{
	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) {
1149
		nodes_clear(trialcs->mems_allowed);
1150
	} else {
1151
		retval = nodelist_parse(buf, trialcs->mems_allowed);
1152 1153 1154
		if (retval < 0)
			goto done;

1155
		if (!nodes_subset(trialcs->mems_allowed,
1156 1157 1158 1159
				node_states[N_HIGH_MEMORY]))
			return -EINVAL;
	}
	oldmem = cs->mems_allowed;
1160
	if (nodes_equal(oldmem, trialcs->mems_allowed)) {
1161 1162 1163
		retval = 0;		/* Too easy - nothing to do */
		goto done;
	}
1164
	retval = validate_change(cs, trialcs);
1165 1166 1167 1168
	if (retval < 0)
		goto done;

	mutex_lock(&callback_mutex);
1169
	cs->mems_allowed = trialcs->mems_allowed;
1170 1171 1172 1173 1174 1175 1176 1177
	cs->mems_generation = cpuset_mems_generation++;
	mutex_unlock(&callback_mutex);

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

1178 1179 1180 1181 1182
int current_cpuset_is_being_rebound(void)
{
	return task_cs(current) == cpuset_being_rebound;
}

1183
static int update_relax_domain_level(struct cpuset *cs, s64 val)
1184
{
1185 1186
	if (val < -1 || val >= SD_LV_MAX)
		return -EINVAL;
1187 1188 1189

	if (val != cs->relax_domain_level) {
		cs->relax_domain_level = val;
1190 1191
		if (!cpumask_empty(cs->cpus_allowed) &&
		    is_sched_load_balance(cs))
1192
			async_rebuild_sched_domains();
1193