cpuset.c 73.0 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
};

118 119 120 121 122 123 124 125 126 127 128 129 130 131
/* 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);
}

L
Linus Torvalds 已提交
132 133 134 135
/* bits in struct cpuset flags field */
typedef enum {
	CS_CPU_EXCLUSIVE,
	CS_MEM_EXCLUSIVE,
136
	CS_MEM_HARDWALL,
137
	CS_MEMORY_MIGRATE,
P
Paul Jackson 已提交
138
	CS_SCHED_LOAD_BALANCE,
139 140
	CS_SPREAD_PAGE,
	CS_SPREAD_SLAB,
L
Linus Torvalds 已提交
141 142 143 144 145
} cpuset_flagbits_t;

/* convenient tests for these bits */
static inline int is_cpu_exclusive(const struct cpuset *cs)
{
146
	return test_bit(CS_CPU_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
147 148 149 150
}

static inline int is_mem_exclusive(const struct cpuset *cs)
{
151
	return test_bit(CS_MEM_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
152 153
}

154 155 156 157 158
static inline int is_mem_hardwall(const struct cpuset *cs)
{
	return test_bit(CS_MEM_HARDWALL, &cs->flags);
}

P
Paul Jackson 已提交
159 160 161 162 163
static inline int is_sched_load_balance(const struct cpuset *cs)
{
	return test_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
}

164 165
static inline int is_memory_migrate(const struct cpuset *cs)
{
166
	return test_bit(CS_MEMORY_MIGRATE, &cs->flags);
167 168
}

169 170 171 172 173 174 175 176 177 178
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
179
/*
180
 * Increment this integer everytime any cpuset changes its
L
Linus Torvalds 已提交
181 182 183 184
 * 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.
 *
185
 * A single, global generation is needed because cpuset_attach_task() could
L
Linus Torvalds 已提交
186 187 188 189
 * 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
190
 * modify another's memory placement.  So we must enable every task,
L
Linus Torvalds 已提交
191 192 193
 * 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.
194
 *
195
 * Since writes to cpuset_mems_generation are guarded by the cgroup lock
196
 * there is no need to mark it atomic.
L
Linus Torvalds 已提交
197
 */
198
static int cpuset_mems_generation;
L
Linus Torvalds 已提交
199 200 201 202 203 204

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

/*
205 206 207 208 209 210 211
 * 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.
212
 *
213
 * A task must hold both mutexes to modify cpusets.  If a task
214
 * holds cgroup_mutex, then it blocks others wanting that mutex,
215
 * ensuring that it is the only task able to also acquire callback_mutex
216 217
 * and be able to modify cpusets.  It can perform various checks on
 * the cpuset structure first, knowing nothing will change.  It can
218
 * also allocate memory while just holding cgroup_mutex.  While it is
219
 * performing these checks, various callback routines can briefly
220 221
 * acquire callback_mutex to query cpusets.  Once it is ready to make
 * the changes, it takes callback_mutex, blocking everyone else.
222 223
 *
 * Calls to the kernel memory allocator can not be made while holding
224
 * callback_mutex, as that would risk double tripping on callback_mutex
225 226 227
 * from one of the callbacks into the cpuset code from within
 * __alloc_pages().
 *
228
 * If a task is only holding callback_mutex, then it has read-only
229 230 231 232 233
 * 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.
 *
234
 * The cpuset_common_file_read() handlers only hold callback_mutex across
235 236 237
 * small pieces of code, such as when reading out possibly multi-word
 * cpumasks and nodemasks.
 *
238 239
 * 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
240 241
 */

242
static DEFINE_MUTEX(callback_mutex);
243

244 245 246 247 248 249 250 251 252 253 254
/*
 * 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);

255 256
/*
 * This is ugly, but preserves the userspace API for existing cpuset
257
 * users. If someone tries to mount the "cpuset" filesystem, we
258 259
 * silently switch it to mount "cgroup" instead
 */
260 261 262
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
263
{
264 265 266 267 268 269 270 271 272 273 274
	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
275 276 277 278 279 280 281 282
}

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

/*
283
 * Return in pmask the portion of a cpusets's cpus_allowed that
L
Linus Torvalds 已提交
284 285 286 287 288 289 290 291 292
 * 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.
 *
293
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
294 295
 */

296 297
static void guarantee_online_cpus(const struct cpuset *cs,
				  struct cpumask *pmask)
L
Linus Torvalds 已提交
298
{
299
	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
L
Linus Torvalds 已提交
300 301
		cs = cs->parent;
	if (cs)
302
		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
L
Linus Torvalds 已提交
303
	else
304 305
		cpumask_copy(pmask, cpu_online_mask);
	BUG_ON(!cpumask_intersects(pmask, cpu_online_mask));
L
Linus Torvalds 已提交
306 307 308 309
}

/*
 * Return in *pmask the portion of a cpusets's mems_allowed that
310 311 312 313
 * 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
314 315
 *
 * One way or another, we guarantee to return some non-empty subset
316
 * of node_states[N_HIGH_MEMORY].
L
Linus Torvalds 已提交
317
 *
318
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
319 320 321 322
 */

static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
{
323 324
	while (cs && !nodes_intersects(cs->mems_allowed,
					node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
325 326
		cs = cs->parent;
	if (cs)
327 328
		nodes_and(*pmask, cs->mems_allowed,
					node_states[N_HIGH_MEMORY]);
L
Linus Torvalds 已提交
329
	else
330 331
		*pmask = node_states[N_HIGH_MEMORY];
	BUG_ON(!nodes_intersects(*pmask, node_states[N_HIGH_MEMORY]));
L
Linus Torvalds 已提交
332 333
}

334 335 336 337 338 339
/**
 * 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.
340
 *
341 342 343 344
 * 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().
 *
345
 * Call without callback_mutex or task_lock() held.  May be
346 347
 * 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
348 349
 * be NULL.  This routine also might acquire callback_mutex during
 * call.
350
 *
351 352
 * Reading current->cpuset->mems_generation doesn't need task_lock
 * to guard the current->cpuset derefence, because it is guarded
353
 * from concurrent freeing of current->cpuset using RCU.
354 355 356 357 358 359 360 361 362 363 364 365 366 367
 *
 * 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.
368 369 370 371 372
 *
 * 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
373 374
 */

375
void cpuset_update_task_memory_state(void)
L
Linus Torvalds 已提交
376
{
377
	int my_cpusets_mem_gen;
378
	struct task_struct *tsk = current;
379
	struct cpuset *cs;
380

381 382 383
	rcu_read_lock();
	my_cpusets_mem_gen = task_cs(tsk)->mems_generation;
	rcu_read_unlock();
L
Linus Torvalds 已提交
384

385
	if (my_cpusets_mem_gen != tsk->cpuset_mems_generation) {
386
		mutex_lock(&callback_mutex);
387
		task_lock(tsk);
388
		cs = task_cs(tsk); /* Maybe changed when task not locked */
389 390
		guarantee_online_mems(cs, &tsk->mems_allowed);
		tsk->cpuset_mems_generation = cs->mems_generation;
391 392 393 394 395 396 397 398
		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;
399
		task_unlock(tsk);
400
		mutex_unlock(&callback_mutex);
401
		mpol_rebind_task(tsk, &tsk->mems_allowed);
L
Linus Torvalds 已提交
402 403 404 405 406 407 408 409
	}
}

/*
 * 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
410
 * are only set if the other's are set.  Call holding cgroup_mutex.
L
Linus Torvalds 已提交
411 412 413 414
 */

static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
{
415
	return	cpumask_subset(p->cpus_allowed, q->cpus_allowed) &&
L
Linus Torvalds 已提交
416 417 418 419 420
		nodes_subset(p->mems_allowed, q->mems_allowed) &&
		is_cpu_exclusive(p) <= is_cpu_exclusive(q) &&
		is_mem_exclusive(p) <= is_mem_exclusive(q);
}

421 422 423 424 425 426
/**
 * 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)
{
427 428 429 430 431 432 433 434 435 436 437 438 439
	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;
440 441 442 443 444 445 446 447
}

/**
 * free_trial_cpuset - free the trial cpuset
 * @trial: the trial cpuset to be freed
 */
static void free_trial_cpuset(struct cpuset *trial)
{
448
	free_cpumask_var(trial->cpus_allowed);
449 450 451
	kfree(trial);
}

L
Linus Torvalds 已提交
452 453 454 455 456 457 458
/*
 * 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
459
 * cgroup_mutex held.
L
Linus Torvalds 已提交
460 461 462 463 464 465 466 467 468 469 470 471 472 473
 *
 * '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)
{
474
	struct cgroup *cont;
L
Linus Torvalds 已提交
475 476 477
	struct cpuset *c, *par;

	/* Each of our child cpusets must be a subset of us */
478 479
	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
		if (!is_cpuset_subset(cgroup_cs(cont), trial))
L
Linus Torvalds 已提交
480 481 482 483
			return -EBUSY;
	}

	/* Remaining checks don't apply to root cpuset */
484
	if (cur == &top_cpuset)
L
Linus Torvalds 已提交
485 486
		return 0;

487 488
	par = cur->parent;

L
Linus Torvalds 已提交
489 490 491 492
	/* We must be a subset of our parent cpuset */
	if (!is_cpuset_subset(trial, par))
		return -EACCES;

493 494 495 496
	/*
	 * If either I or some sibling (!= me) is exclusive, we can't
	 * overlap
	 */
497 498
	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
		c = cgroup_cs(cont);
L
Linus Torvalds 已提交
499 500
		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
		    c != cur &&
501
		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
L
Linus Torvalds 已提交
502 503 504 505 506 507 508
			return -EINVAL;
		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
		    c != cur &&
		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
			return -EINVAL;
	}

509 510
	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
	if (cgroup_task_count(cur->css.cgroup)) {
511
		if (cpumask_empty(trial->cpus_allowed) ||
512 513 514 515 516
		    nodes_empty(trial->mems_allowed)) {
			return -ENOSPC;
		}
	}

L
Linus Torvalds 已提交
517 518 519
	return 0;
}

520
#ifdef CONFIG_SMP
P
Paul Jackson 已提交
521
/*
522
 * Helper routine for generate_sched_domains().
P
Paul Jackson 已提交
523 524 525 526
 * Do cpusets a, b have overlapping cpus_allowed masks?
 */
static int cpusets_overlap(struct cpuset *a, struct cpuset *b)
{
527
	return cpumask_intersects(a->cpus_allowed, b->cpus_allowed);
P
Paul Jackson 已提交
528 529
}

530 531 532 533 534 535 536 537
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;
}

538 539 540 541 542 543 544 545 546 547 548 549 550 551
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);

552
		if (cpumask_empty(cp->cpus_allowed))
553 554 555 556 557 558 559 560 561 562 563 564
			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
565
/*
566 567 568 569 570 571 572 573 574
 * 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
575
 *
L
Li Zefan 已提交
576
 * See "What is sched_load_balance" in Documentation/cgroups/cpusets.txt
P
Paul Jackson 已提交
577 578 579 580 581 582 583
 * 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.
 *
584
 * Must be called with cgroup_lock held.
P
Paul Jackson 已提交
585 586
 *
 * The three key local variables below are:
587
 *    q  - a linked-list queue of cpuset pointers, used to implement a
P
Paul Jackson 已提交
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
 *	   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().
 */
619 620
/* FIXME: see the FIXME in partition_sched_domains() */
static int generate_sched_domains(struct cpumask **domains,
621
			struct sched_domain_attr **attributes)
P
Paul Jackson 已提交
622
{
623
	LIST_HEAD(q);		/* queue of cpusets to be scanned */
P
Paul Jackson 已提交
624 625 626 627
	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 */
628
	struct cpumask *doms;	/* resulting partition; i.e. sched domains */
629
	struct sched_domain_attr *dattr;  /* attributes for custom domains */
630
	int ndoms = 0;		/* number of sched domains in result */
631
	int nslot;		/* next empty doms[] struct cpumask slot */
P
Paul Jackson 已提交
632 633

	doms = NULL;
634
	dattr = NULL;
635
	csa = NULL;
P
Paul Jackson 已提交
636 637 638

	/* Special case for the 99% of systems with one, full, sched domain */
	if (is_sched_load_balance(&top_cpuset)) {
639
		doms = kmalloc(cpumask_size(), GFP_KERNEL);
P
Paul Jackson 已提交
640
		if (!doms)
641 642
			goto done;

643 644 645
		dattr = kmalloc(sizeof(struct sched_domain_attr), GFP_KERNEL);
		if (dattr) {
			*dattr = SD_ATTR_INIT;
646
			update_domain_attr_tree(dattr, &top_cpuset);
647
		}
648
		cpumask_copy(doms, top_cpuset.cpus_allowed);
649 650 651

		ndoms = 1;
		goto done;
P
Paul Jackson 已提交
652 653 654 655 656 657 658
	}

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

659 660
	list_add(&top_cpuset.stack_list, &q);
	while (!list_empty(&q)) {
P
Paul Jackson 已提交
661 662
		struct cgroup *cont;
		struct cpuset *child;   /* scans child cpusets of cp */
663

664 665 666
		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

667
		if (cpumask_empty(cp->cpus_allowed))
668 669
			continue;

670 671 672 673 674 675 676
		/*
		 * 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
677
			csa[csn++] = cp;
678 679
			continue;
		}
680

P
Paul Jackson 已提交
681 682
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
683
			list_add_tail(&child->stack_list, &q);
P
Paul Jackson 已提交
684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713
		}
  	}

	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;
			}
		}
	}

714 715 716 717
	/*
	 * Now we know how many domains to create.
	 * Convert <csn, csa> to <ndoms, doms> and populate cpu masks.
	 */
718
	doms = kmalloc(ndoms * cpumask_size(), GFP_KERNEL);
719
	if (!doms)
720 721 722 723 724 725
		goto done;

	/*
	 * The rest of the code, including the scheduler, can deal with
	 * dattr==NULL case. No need to abort if alloc fails.
	 */
726
	dattr = kmalloc(ndoms * sizeof(struct sched_domain_attr), GFP_KERNEL);
P
Paul Jackson 已提交
727 728 729

	for (nslot = 0, i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
730
		struct cpumask *dp;
P
Paul Jackson 已提交
731 732
		int apn = a->pn;

733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748
		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
749
			}
750 751
			continue;
		}
P
Paul Jackson 已提交
752

753
		cpumask_clear(dp);
754 755 756 757 758 759
		if (dattr)
			*(dattr + nslot) = SD_ATTR_INIT;
		for (j = i; j < csn; j++) {
			struct cpuset *b = csa[j];

			if (apn == b->pn) {
760
				cpumask_or(dp, dp, b->cpus_allowed);
761 762 763 764 765
				if (dattr)
					update_domain_attr_tree(dattr + nslot, b);

				/* Done with this partition */
				b->pn = -1;
P
Paul Jackson 已提交
766 767
			}
		}
768
		nslot++;
P
Paul Jackson 已提交
769 770 771
	}
	BUG_ON(nslot != ndoms);

772 773 774
done:
	kfree(csa);

775 776 777 778 779 780 781
	/*
	 * Fallback to the default domain if kmalloc() failed.
	 * See comments in partition_sched_domains().
	 */
	if (doms == NULL)
		ndoms = 1;

782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799
	*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;
800
	struct cpumask *doms;
801 802
	int ndoms;

803
	get_online_cpus();
804 805 806 807 808 809 810 811 812

	/* 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);

813
	put_online_cpus();
814
}
815 816 817 818 819 820 821 822 823 824 825 826
#else /* !CONFIG_SMP */
static void do_rebuild_sched_domains(struct work_struct *unused)
{
}

static int generate_sched_domains(struct cpumask **domains,
			struct sched_domain_attr **attributes)
{
	*domains = NULL;
	return 1;
}
#endif /* CONFIG_SMP */
P
Paul Jackson 已提交
827

828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850
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)
{
851
	queue_work(cpuset_wq, &rebuild_sched_domains_work);
852 853 854 855 856 857 858 859 860 861 862 863 864 865
}

/*
 * 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
866 867
}

C
Cliff Wickman 已提交
868 869 870 871 872
/**
 * 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
 *
873
 * Call with cgroup_mutex held.  May take callback_mutex during call.
C
Cliff Wickman 已提交
874 875 876
 * 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).
877
 */
878 879
static int cpuset_test_cpumask(struct task_struct *tsk,
			       struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
880
{
881
	return !cpumask_equal(&tsk->cpus_allowed,
C
Cliff Wickman 已提交
882 883
			(cgroup_cs(scan->cg))->cpus_allowed);
}
884

C
Cliff Wickman 已提交
885 886 887 888 889 890 891 892 893 894 895
/**
 * 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.
 */
896 897
static void cpuset_change_cpumask(struct task_struct *tsk,
				  struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
898
{
899
	set_cpus_allowed_ptr(tsk, ((cgroup_cs(scan->cg))->cpus_allowed));
C
Cliff Wickman 已提交
900 901
}

902 903 904
/**
 * 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
905
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
906 907 908 909 910 911
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
912 913
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
914
 */
915
static void update_tasks_cpumask(struct cpuset *cs, struct ptr_heap *heap)
916 917 918 919 920 921
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = cpuset_test_cpumask;
	scan.process_task = cpuset_change_cpumask;
922 923
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
924 925
}

C
Cliff Wickman 已提交
926 927 928 929 930
/**
 * 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
 */
931 932
static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
			  const char *buf)
L
Linus Torvalds 已提交
933
{
934
	struct ptr_heap heap;
C
Cliff Wickman 已提交
935 936
	int retval;
	int is_load_balanced;
L
Linus Torvalds 已提交
937

938 939 940 941
	/* top_cpuset.cpus_allowed tracks cpu_online_map; it's read-only */
	if (cs == &top_cpuset)
		return -EACCES;

942
	/*
943
	 * An empty cpus_allowed is ok only if the cpuset has no tasks.
944 945 946
	 * Since cpulist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have cpus.
947
	 */
948
	if (!*buf) {
949
		cpumask_clear(trialcs->cpus_allowed);
950
	} else {
951
		retval = cpulist_parse(buf, trialcs->cpus_allowed);
952 953
		if (retval < 0)
			return retval;
954

955
		if (!cpumask_subset(trialcs->cpus_allowed, cpu_online_mask))
956
			return -EINVAL;
957
	}
958
	retval = validate_change(cs, trialcs);
959 960
	if (retval < 0)
		return retval;
P
Paul Jackson 已提交
961

P
Paul Menage 已提交
962
	/* Nothing to do if the cpus didn't change */
963
	if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed))
P
Paul Menage 已提交
964
		return 0;
C
Cliff Wickman 已提交
965

966 967 968 969
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval)
		return retval;

970
	is_load_balanced = is_sched_load_balance(trialcs);
P
Paul Jackson 已提交
971

972
	mutex_lock(&callback_mutex);
973
	cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed);
974
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
975

P
Paul Menage 已提交
976 977
	/*
	 * Scan tasks in the cpuset, and update the cpumasks of any
C
Cliff Wickman 已提交
978
	 * that need an update.
P
Paul Menage 已提交
979
	 */
980 981 982
	update_tasks_cpumask(cs, &heap);

	heap_free(&heap);
C
Cliff Wickman 已提交
983

P
Paul Menage 已提交
984
	if (is_load_balanced)
985
		async_rebuild_sched_domains();
986
	return 0;
L
Linus Torvalds 已提交
987 988
}

989 990 991 992 993 994 995 996
/*
 * 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.
 *
997
 *    Call holding cgroup_mutex, so current's cpuset won't change
998
 *    during this call, as manage_mutex holds off any cpuset_attach()
999 1000
 *    calls.  Therefore we don't need to take task_lock around the
 *    call to guarantee_online_mems(), as we know no one is changing
1001
 *    our task's cpuset.
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
 *
 *    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);
1034
	guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
1035 1036 1037
	mutex_unlock(&callback_mutex);
}

1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
/*
 * Rebind task's vmas to cpuset's new mems_allowed, and migrate pages to new
 * nodes if memory_migrate flag is set. Called with cgroup_mutex held.
 */
static void cpuset_change_nodemask(struct task_struct *p,
				   struct cgroup_scanner *scan)
{
	struct mm_struct *mm;
	struct cpuset *cs;
	int migrate;
	const nodemask_t *oldmem = scan->data;

	mm = get_task_mm(p);
	if (!mm)
		return;

	cs = cgroup_cs(scan->cg);
	migrate = is_memory_migrate(cs);

	mpol_rebind_mm(mm, &cs->mems_allowed);
	if (migrate)
		cpuset_migrate_mm(mm, oldmem, &cs->mems_allowed);
	mmput(mm);
}

1063 1064
static void *cpuset_being_rebound;

1065 1066 1067 1068
/**
 * 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
1069
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
1070 1071
 *
 * Called with cgroup_mutex held
1072 1073
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
1074
 */
1075 1076
static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem,
				 struct ptr_heap *heap)
L
Linus Torvalds 已提交
1077
{
1078
	struct cgroup_scanner scan;
1079

1080
	cpuset_being_rebound = cs;		/* causes mpol_dup() rebind */
1081

1082 1083 1084
	scan.cg = cs->css.cgroup;
	scan.test_task = NULL;
	scan.process_task = cpuset_change_nodemask;
1085
	scan.heap = heap;
1086
	scan.data = (nodemask_t *)oldmem;
1087 1088

	/*
1089 1090 1091 1092 1093 1094
	 * The mpol_rebind_mm() call takes mmap_sem, which we couldn't
	 * take while holding tasklist_lock.  Forks can happen - the
	 * mpol_dup() cpuset_being_rebound check will catch such forks,
	 * and rebind their vma mempolicies too.  Because we still hold
	 * the global cgroup_mutex, we know that no other rebind effort
	 * will be contending for the global variable cpuset_being_rebound.
1095
	 * 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
	cgroup_scan_tasks(&scan);
1099

1100
	/* We're done rebinding vmas to this cpuset's new mems_allowed. */
1101
	cpuset_being_rebound = NULL;
L
Linus Torvalds 已提交
1102 1103
}

1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
/*
 * 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.
 */
1117 1118
static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs,
			   const char *buf)
1119 1120 1121
{
	nodemask_t oldmem;
	int retval;
1122
	struct ptr_heap heap;
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137

	/*
	 * 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) {
1138
		nodes_clear(trialcs->mems_allowed);
1139
	} else {
1140
		retval = nodelist_parse(buf, trialcs->mems_allowed);
1141 1142 1143
		if (retval < 0)
			goto done;

1144
		if (!nodes_subset(trialcs->mems_allowed,
1145 1146 1147 1148
				node_states[N_HIGH_MEMORY]))
			return -EINVAL;
	}
	oldmem = cs->mems_allowed;
1149
	if (nodes_equal(oldmem, trialcs->mems_allowed)) {
1150 1151 1152
		retval = 0;		/* Too easy - nothing to do */
		goto done;
	}
1153
	retval = validate_change(cs, trialcs);
1154 1155 1156
	if (retval < 0)
		goto done;

1157 1158 1159 1160
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval < 0)
		goto done;

1161
	mutex_lock(&callback_mutex);
1162
	cs->mems_allowed = trialcs->mems_allowed;
1163 1164 1165
	cs->mems_generation = cpuset_mems_generation++;
	mutex_unlock(&callback_mutex);

1166 1167 1168
	update_tasks_nodemask(cs, &oldmem, &heap);

	heap_free(&heap);
1169 1170 1171 1172
done:
	return retval;
}

1173 1174 1175 1176 1177
int current_cpuset_is_being_rebound(void)
{
	return task_cs(current) == cpuset_being_rebound;
}

1178
static int update_relax_domain_level(struct cpuset *cs, s64 val)
1179
{
1180
#ifdef CONFIG_SMP
1181 1182
	if (val < -1 || val >= SD_LV_MAX)
		return -EINVAL;
1183
#endif
1184 1185 1186

	if (val != cs->relax_domain_level) {
		cs->relax_domain_level = val;
1187 1188
		if (!cpumask_empty(cs->cpus_allowed) &&
		    is_sched_load_balance(cs))
1189
			async_rebuild_sched_domains();
1190 1191 1192 1193 1194
	}

	return 0;
}

L
Linus Torvalds 已提交
1195 1196
/*
 * update_flag - read a 0 or a 1 in a file and update associated flag
1197 1198 1199
 * bit:		the bit to update (see cpuset_flagbits_t)
 * cs:		the cpuset to update
 * turning_on: 	whether the flag is being set or cleared
1200
 *
1201
 * Call with cgroup_mutex held.
L
Linus Torvalds 已提交
1202 1203
 */

1204 1205
static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
		       int turning_on)
L
Linus Torvalds 已提交
1206
{
1207
	struct cpuset *trialcs;
1208
	int err;
R
Rakib Mullick 已提交
1209
	int balance_flag_changed;
L
Linus Torvalds 已提交
1210

1211 1212 1213 1214
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

L
Linus Torvalds 已提交
1215
	if (turning_on)
1216
		set_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1217
	else
1218
		clear_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1219

1220
	err = validate_change(cs, trialcs);
1221
	if (err < 0)
1222
		goto out;
P
Paul Jackson 已提交
1223 1224

	balance_flag_changed = (is_sched_load_balance(cs) !=
1225
				is_sched_load_balance(trialcs));
P
Paul Jackson 已提交
1226

1227
	mutex_lock(&callback_mutex);
1228
	cs->flags = trialcs->flags;
1229
	mutex_unlock(&callback_mutex);
1230

1231
	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed)
1232
		async_rebuild_sched_domains();
P
Paul Jackson 已提交
1233

1234 1235 1236
out:
	free_trial_cpuset(trialcs);
	return err;
L
Linus Torvalds 已提交
1237 1238
}

1239
/*
A
Adrian Bunk 已提交
1240
 * Frequency meter - How fast is some event occurring?
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 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
 *
 * 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;
}

1337 1338 1339
/* Protected by cgroup_lock */
static cpumask_var_t cpus_attach;

1340
/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
1341 1342
static int cpuset_can_attach(struct cgroup_subsys *ss,
			     struct cgroup *cont, struct task_struct *tsk)
L
Linus Torvalds 已提交
1343
{
1344
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1345

1346
	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
L
Linus Torvalds 已提交
1347
		return -ENOSPC;
1348

1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
	/*
	 * Kthreads bound to specific cpus cannot be moved to a new cpuset; we
	 * cannot change their cpu affinity and isolating such threads by their
	 * set of allowed nodes is unnecessary.  Thus, cpusets are not
	 * applicable for such threads.  This prevents checking for success of
	 * set_cpus_allowed_ptr() on all attached tasks before cpus_allowed may
	 * be changed.
	 */
	if (tsk->flags & PF_THREAD_BOUND)
		return -EINVAL;
L
Linus Torvalds 已提交
1359

1360
	return security_task_setscheduler(tsk, 0, NULL);
1361
}
L
Linus Torvalds 已提交
1362

1363 1364 1365 1366 1367 1368 1369 1370
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);
1371
	int err;
1372

1373
	if (cs == &top_cpuset) {
1374
		cpumask_copy(cpus_attach, cpu_possible_mask);
1375 1376
	} else {
		mutex_lock(&callback_mutex);
1377
		guarantee_online_cpus(cs, cpus_attach);
1378 1379
		mutex_unlock(&callback_mutex);
	}
1380
	err = set_cpus_allowed_ptr(tsk, cpus_attach);
1381 1382
	if (err)
		return;
L
Linus Torvalds 已提交
1383

1384 1385
	from = oldcs->mems_allowed;
	to = cs->mems_allowed;
1386 1387 1388
	mm = get_task_mm(tsk);
	if (mm) {
		mpol_rebind_mm(mm, &to);
1389
		if (is_memory_migrate(cs))
1390
			cpuset_migrate_mm(mm, &from, &to);
1391 1392
		mmput(mm);
	}
L
Linus Torvalds 已提交
1393 1394 1395 1396 1397
}

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

typedef enum {
1398
	FILE_MEMORY_MIGRATE,
L
Linus Torvalds 已提交
1399 1400 1401 1402
	FILE_CPULIST,
	FILE_MEMLIST,
	FILE_CPU_EXCLUSIVE,
	FILE_MEM_EXCLUSIVE,
1403
	FILE_MEM_HARDWALL,
P
Paul Jackson 已提交
1404
	FILE_SCHED_LOAD_BALANCE,
1405
	FILE_SCHED_RELAX_DOMAIN_LEVEL,
1406 1407
	FILE_MEMORY_PRESSURE_ENABLED,
	FILE_MEMORY_PRESSURE,
1408 1409
	FILE_SPREAD_PAGE,
	FILE_SPREAD_SLAB,
L
Linus Torvalds 已提交
1410 1411
} cpuset_filetype_t;

1412 1413 1414 1415 1416 1417
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;

1418
	if (!cgroup_lock_live_group(cgrp))
1419 1420 1421
		return -ENODEV;

	switch (type) {
L
Linus Torvalds 已提交
1422
	case FILE_CPU_EXCLUSIVE:
1423
		retval = update_flag(CS_CPU_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1424 1425
		break;
	case FILE_MEM_EXCLUSIVE:
1426
		retval = update_flag(CS_MEM_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1427
		break;
1428 1429 1430
	case FILE_MEM_HARDWALL:
		retval = update_flag(CS_MEM_HARDWALL, cs, val);
		break;
P
Paul Jackson 已提交
1431
	case FILE_SCHED_LOAD_BALANCE:
1432
		retval = update_flag(CS_SCHED_LOAD_BALANCE, cs, val);
1433
		break;
1434
	case FILE_MEMORY_MIGRATE:
1435
		retval = update_flag(CS_MEMORY_MIGRATE, cs, val);
1436
		break;
1437
	case FILE_MEMORY_PRESSURE_ENABLED:
1438
		cpuset_memory_pressure_enabled = !!val;
1439 1440 1441 1442
		break;
	case FILE_MEMORY_PRESSURE:
		retval = -EACCES;
		break;
1443
	case FILE_SPREAD_PAGE:
1444
		retval = update_flag(CS_SPREAD_PAGE, cs, val);
1445
		cs->mems_generation = cpuset_mems_generation++;
1446 1447
		break;
	case FILE_SPREAD_SLAB:
1448
		retval = update_flag(CS_SPREAD_SLAB, cs, val);
1449
		cs->mems_generation = cpuset_mems_generation++;
1450
		break;
L
Linus Torvalds 已提交
1451 1452
	default:
		retval = -EINVAL;
1453
		break;
L
Linus Torvalds 已提交
1454
	}
1455
	cgroup_unlock();
L
Linus Torvalds 已提交
1456 1457 1458
	return retval;
}

1459 1460 1461 1462 1463 1464
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;

1465
	if (!cgroup_lock_live_group(cgrp))
1466
		return -ENODEV;
1467

1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
	switch (type) {
	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
		retval = update_relax_domain_level(cs, val);
		break;
	default:
		retval = -EINVAL;
		break;
	}
	cgroup_unlock();
	return retval;
}

1480 1481 1482 1483 1484 1485 1486
/*
 * 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;
1487 1488
	struct cpuset *cs = cgroup_cs(cgrp);
	struct cpuset *trialcs;
1489 1490 1491 1492

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

1493 1494 1495 1496
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

1497 1498
	switch (cft->private) {
	case FILE_CPULIST:
1499
		retval = update_cpumask(cs, trialcs, buf);
1500 1501
		break;
	case FILE_MEMLIST:
1502
		retval = update_nodemask(cs, trialcs, buf);
1503 1504 1505 1506 1507
		break;
	default:
		retval = -EINVAL;
		break;
	}
1508 1509

	free_trial_cpuset(trialcs);
1510 1511 1512 1513
	cgroup_unlock();
	return retval;
}

L
Linus Torvalds 已提交
1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
/*
 * 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)
{
1528
	int ret;
L
Linus Torvalds 已提交
1529

1530
	mutex_lock(&callback_mutex);
1531
	ret = cpulist_scnprintf(page, PAGE_SIZE, cs->cpus_allowed);
1532
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1533

1534
	return ret;
L
Linus Torvalds 已提交
1535 1536 1537 1538 1539 1540
}

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

1541
	mutex_lock(&callback_mutex);
L
Linus Torvalds 已提交
1542
	mask = cs->mems_allowed;
1543
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1544 1545 1546 1547

	return nodelist_scnprintf(page, PAGE_SIZE, mask);
}

1548 1549 1550 1551 1552
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
1553
{
1554
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1555 1556 1557 1558 1559
	cpuset_filetype_t type = cft->private;
	char *page;
	ssize_t retval = 0;
	char *s;

1560
	if (!(page = (char *)__get_free_page(GFP_TEMPORARY)))
L
Linus Torvalds 已提交
1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
		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
1578
	retval = simple_read_from_buffer(buf, nbytes, ppos, page, s - page);
L
Linus Torvalds 已提交
1579 1580 1581 1582 1583
out:
	free_page((unsigned long)page);
	return retval;
}

1584 1585 1586 1587 1588 1589 1590 1591 1592
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);
1593 1594
	case FILE_MEM_HARDWALL:
		return is_mem_hardwall(cs);
1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
	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();
	}
1610 1611 1612

	/* Unreachable but makes gcc happy */
	return 0;
1613
}
L
Linus Torvalds 已提交
1614

1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
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();
	}
1625 1626 1627

	/* Unrechable but makes gcc happy */
	return 0;
1628 1629
}

L
Linus Torvalds 已提交
1630 1631 1632 1633 1634

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

1635 1636 1637 1638
static struct cftype files[] = {
	{
		.name = "cpus",
		.read = cpuset_common_file_read,
1639 1640
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * NR_CPUS),
1641 1642 1643 1644 1645 1646
		.private = FILE_CPULIST,
	},

	{
		.name = "mems",
		.read = cpuset_common_file_read,
1647 1648
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * MAX_NUMNODES),
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
		.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,
	},

1666 1667 1668 1669 1670 1671 1672
	{
		.name = "mem_hardwall",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEM_HARDWALL,
	},

1673 1674 1675 1676 1677 1678 1679 1680 1681
	{
		.name = "sched_load_balance",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SCHED_LOAD_BALANCE,
	},

	{
		.name = "sched_relax_domain_level",
1682 1683
		.read_s64 = cpuset_read_s64,
		.write_s64 = cpuset_write_s64,
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698
		.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,
L
Li Zefan 已提交
1699
		.mode = S_IRUGO,
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
	},

	{
		.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,
	},
1715 1716
};

1717 1718
static struct cftype cft_memory_pressure_enabled = {
	.name = "memory_pressure_enabled",
1719 1720
	.read_u64 = cpuset_read_u64,
	.write_u64 = cpuset_write_u64,
1721 1722 1723
	.private = FILE_MEMORY_PRESSURE_ENABLED,
};

1724
static int cpuset_populate(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1725 1726 1727
{
	int err;

1728 1729
	err = cgroup_add_files(cont, ss, files, ARRAY_SIZE(files));
	if (err)
L
Linus Torvalds 已提交
1730
		return err;
1731
	/* memory_pressure_enabled is in root cpuset only */
1732
	if (!cont->parent)
1733
		err = cgroup_add_file(cont, ss,
1734 1735
				      &cft_memory_pressure_enabled);
	return err;
L
Linus Torvalds 已提交
1736 1737
}

1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
/*
 * 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
1752 1753
 * (and likewise for mems) to the new cgroup. Called with cgroup_mutex
 * held.
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
 */
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;
1771
	cpumask_copy(cs->cpus_allowed, parent_cs->cpus_allowed);
1772 1773 1774
	return;
}

L
Linus Torvalds 已提交
1775 1776
/*
 *	cpuset_create - create a cpuset
1777 1778
 *	ss:	cpuset cgroup subsystem
 *	cont:	control group that the new cpuset will be part of
L
Linus Torvalds 已提交
1779 1780
 */

1781 1782 1783
static struct cgroup_subsys_state *cpuset_create(
	struct cgroup_subsys *ss,
	struct cgroup *cont)
L
Linus Torvalds 已提交
1784 1785
{
	struct cpuset *cs;
1786
	struct cpuset *parent;
L
Linus Torvalds 已提交
1787

1788 1789 1790 1791 1792 1793
	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
1794 1795
	cs = kmalloc(sizeof(*cs), GFP_KERNEL);
	if (!cs)
1796
		return ERR_PTR(-ENOMEM);
1797 1798 1799 1800
	if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) {
		kfree(cs);
		return ERR_PTR(-ENOMEM);
	}
L
Linus Torvalds 已提交
1801

1802
	cpuset_update_task_memory_state();
L
Linus Torvalds 已提交
1803
	cs->flags = 0;
1804 1805 1806 1807
	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
1808
	set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
1809
	cpumask_clear(cs->cpus_allowed);
1810
	nodes_clear(cs->mems_allowed);
1811
	cs->mems_generation = cpuset_mems_generation++;
1812
	fmeter_init(&cs->fmeter);
1813
	cs->relax_domain_level = -1;
L
Linus Torvalds 已提交
1814 1815

	cs->parent = parent;
1816
	number_of_cpusets++;
1817
	return &cs->css ;
L
Linus Torvalds 已提交
1818 1819
}

P
Paul Jackson 已提交
1820 1821 1822
/*
 * If the cpuset being removed has its flag 'sched_load_balance'
 * enabled, then simulate turning sched_load_balance off, which
1823
 * will call async_rebuild_sched_domains().
P
Paul Jackson 已提交
1824 1825
 */

1826
static void cpuset_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1827
{
1828
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1829

1830
	cpuset_update_task_memory_state();
P
Paul Jackson 已提交
1831 1832

	if (is_sched_load_balance(cs))
1833
		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
P
Paul Jackson 已提交
1834

1835
	number_of_cpusets--;
1836
	free_cpumask_var(cs->cpus_allowed);
1837
	kfree(cs);
L
Linus Torvalds 已提交
1838 1839
}

1840 1841 1842
struct cgroup_subsys cpuset_subsys = {
	.name = "cpuset",
	.create = cpuset_create,
1843
	.destroy = cpuset_destroy,
1844 1845 1846 1847 1848 1849 1850 1851
	.can_attach = cpuset_can_attach,
	.attach = cpuset_attach,
	.populate = cpuset_populate,
	.post_clone = cpuset_post_clone,
	.subsys_id = cpuset_subsys_id,
	.early_init = 1,
};

1852 1853 1854 1855 1856 1857 1858 1859
/*
 * 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)
{
1860 1861
	alloc_bootmem_cpumask_var(&top_cpuset.cpus_allowed);

1862
	top_cpuset.mems_generation = cpuset_mems_generation++;
1863 1864 1865
	return 0;
}

1866

L
Linus Torvalds 已提交
1867 1868 1869 1870 1871 1872 1873 1874
/**
 * cpuset_init - initialize cpusets at system boot
 *
 * Description: Initialize top_cpuset and the cpuset internal file system,
 **/

int __init cpuset_init(void)
{
1875
	int err = 0;
L
Linus Torvalds 已提交
1876

1877
	cpumask_setall(top_cpuset.cpus_allowed);
1878
	nodes_setall(top_cpuset.mems_allowed);
L
Linus Torvalds 已提交
1879

1880
	fmeter_init(&top_cpuset.fmeter);
1881
	top_cpuset.mems_generation = cpuset_mems_generation++;
P
Paul Jackson 已提交
1882
	set_bit(CS_SCHED_LOAD_BALANCE, &top_cpuset.flags);
1883
	top_cpuset.relax_domain_level = -1;
L
Linus Torvalds 已提交
1884 1885 1886

	err = register_filesystem(&cpuset_fs_type);
	if (err < 0)
1887 1888
		return err;

1889 1890 1891
	if (!alloc_cpumask_var(&cpus_attach, GFP_KERNEL))
		BUG();

1892
	number_of_cpusets = 1;
1893
	return 0;
L
Linus Torvalds 已提交
1894 1895
}

1896 1897 1898 1899 1900 1901 1902 1903
/**
 * 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.
 */
1904 1905
static void cpuset_do_move_task(struct task_struct *tsk,
				struct cgroup_scanner *scan)
1906
{
1907
	struct cgroup *new_cgroup = scan->data;
1908

1909
	cgroup_attach_task(new_cgroup, tsk);
1910 1911 1912 1913 1914 1915 1916
}

/**
 * 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
 *
1917 1918
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1919 1920 1921 1922 1923 1924
 *
 * 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)
{
1925
	struct cgroup_scanner scan;
1926

1927 1928 1929 1930 1931
	scan.cg = from->css.cgroup;
	scan.test_task = NULL; /* select all tasks in cgroup */
	scan.process_task = cpuset_do_move_task;
	scan.heap = NULL;
	scan.data = to->css.cgroup;
1932

1933
	if (cgroup_scan_tasks(&scan))
1934 1935 1936 1937
		printk(KERN_ERR "move_member_tasks_to_cpuset: "
				"cgroup_scan_tasks failed\n");
}

1938
/*
1939
 * If CPU and/or memory hotplug handlers, below, unplug any CPUs
1940 1941
 * or memory nodes, we need to walk over the cpuset hierarchy,
 * removing that CPU or node from all cpusets.  If this removes the
1942 1943
 * last CPU or node from a cpuset, then move the tasks in the empty
 * cpuset to its next-highest non-empty parent.
1944
 *
1945 1946
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1947
 */
1948 1949 1950 1951
static void remove_tasks_in_empty_cpuset(struct cpuset *cs)
{
	struct cpuset *parent;

1952 1953 1954 1955 1956
	/*
	 * 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.
	 */
1957 1958
	if (list_empty(&cs->css.cgroup->css_sets))
		return;
1959

1960 1961 1962 1963 1964
	/*
	 * Find its next-highest non-empty parent, (top cpuset
	 * has online cpus, so can't be empty).
	 */
	parent = cs->parent;
1965
	while (cpumask_empty(parent->cpus_allowed) ||
1966
			nodes_empty(parent->mems_allowed))
1967 1968 1969 1970 1971 1972 1973 1974 1975
		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.
 *
1976
 * Called with cgroup_mutex held.  We take callback_mutex to modify
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
 * 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.
 */
1987
static void scan_for_empty_cpusets(struct cpuset *root)
1988
{
1989
	LIST_HEAD(queue);
1990 1991
	struct cpuset *cp;	/* scans cpusets being updated */
	struct cpuset *child;	/* scans child cpusets of cp */
1992
	struct cgroup *cont;
1993
	nodemask_t oldmems;
1994

1995 1996 1997
	list_add_tail((struct list_head *)&root->stack_list, &queue);

	while (!list_empty(&queue)) {
1998
		cp = list_first_entry(&queue, struct cpuset, stack_list);
1999 2000 2001 2002 2003
		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);
		}
2004 2005

		/* Continue past cpusets with all cpus, mems online */
2006
		if (cpumask_subset(cp->cpus_allowed, cpu_online_mask) &&
2007 2008 2009
		    nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY]))
			continue;

2010 2011
		oldmems = cp->mems_allowed;

2012
		/* Remove offline cpus and mems from this cpuset. */
2013
		mutex_lock(&callback_mutex);
2014 2015
		cpumask_and(cp->cpus_allowed, cp->cpus_allowed,
			    cpu_online_mask);
2016 2017
		nodes_and(cp->mems_allowed, cp->mems_allowed,
						node_states[N_HIGH_MEMORY]);
2018 2019 2020
		mutex_unlock(&callback_mutex);

		/* Move tasks from the empty cpuset to a parent */
2021
		if (cpumask_empty(cp->cpus_allowed) ||
2022
		     nodes_empty(cp->mems_allowed))
2023
			remove_tasks_in_empty_cpuset(cp);
2024
		else {
2025
			update_tasks_cpumask(cp, NULL);
2026
			update_tasks_nodemask(cp, &oldmems, NULL);
2027
		}
2028 2029 2030
	}
}

2031 2032 2033 2034 2035 2036
/*
 * 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.
 *
2037 2038
 * This routine ensures that top_cpuset.cpus_allowed tracks
 * cpu_online_map on each CPU hotplug (cpuhp) event.
2039 2040 2041
 *
 * Called within get_online_cpus().  Needs to call cgroup_lock()
 * before calling generate_sched_domains().
2042
 */
2043
static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
P
Paul Jackson 已提交
2044
				unsigned long phase, void *unused_cpu)
2045
{
2046
	struct sched_domain_attr *attr;
2047
	struct cpumask *doms;
2048 2049
	int ndoms;

2050 2051 2052 2053 2054 2055
	switch (phase) {
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		break;
2056

2057
	default:
2058
		return NOTIFY_DONE;
2059
	}
2060

2061
	cgroup_lock();
2062
	mutex_lock(&callback_mutex);
2063
	cpumask_copy(top_cpuset.cpus_allowed, cpu_online_mask);
2064
	mutex_unlock(&callback_mutex);
2065 2066 2067 2068 2069 2070 2071
	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);

2072
	return NOTIFY_OK;
2073 2074
}

2075
#ifdef CONFIG_MEMORY_HOTPLUG
2076
/*
2077
 * Keep top_cpuset.mems_allowed tracking node_states[N_HIGH_MEMORY].
2078 2079
 * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
 * See also the previous routine cpuset_track_online_cpus().
2080
 */
2081 2082
static int cpuset_track_online_nodes(struct notifier_block *self,
				unsigned long action, void *arg)
2083
{
2084
	cgroup_lock();
2085 2086 2087
	switch (action) {
	case MEM_ONLINE:
	case MEM_OFFLINE:
2088
		mutex_lock(&callback_mutex);
2089
		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2090 2091 2092
		mutex_unlock(&callback_mutex);
		if (action == MEM_OFFLINE)
			scan_for_empty_cpusets(&top_cpuset);
2093 2094 2095 2096
		break;
	default:
		break;
	}
2097
	cgroup_unlock();
2098
	return NOTIFY_OK;
2099 2100 2101
}
#endif

L
Linus Torvalds 已提交
2102 2103 2104 2105 2106 2107 2108 2109
/**
 * cpuset_init_smp - initialize cpus_allowed
 *
 * Description: Finish top cpuset after cpu, node maps are initialized
 **/

void __init cpuset_init_smp(void)
{
2110
	cpumask_copy(top_cpuset.cpus_allowed, cpu_online_mask);
2111
	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2112

2113
	hotcpu_notifier(cpuset_track_online_cpus, 0);
2114
	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
2115 2116 2117

	cpuset_wq = create_singlethread_workqueue("cpuset");
	BUG_ON(!cpuset_wq);
L
Linus Torvalds 已提交
2118 2119 2120 2121 2122
}

/**
 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
2123
 * @pmask: pointer to struct cpumask variable to receive cpus_allowed set.
L
Linus Torvalds 已提交
2124
 *
2125
 * Description: Returns the cpumask_var_t cpus_allowed of the cpuset
L
Linus Torvalds 已提交
2126 2127 2128 2129 2130
 * 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.
 **/

2131
void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask)
L
Linus Torvalds 已提交
2132
{
2133
	mutex_lock(&callback_mutex);
2134
	cpuset_cpus_allowed_locked(tsk, pmask);
2135 2136 2137 2138 2139
	mutex_unlock(&callback_mutex);
}

/**
 * cpuset_cpus_allowed_locked - return cpus_allowed mask from a tasks cpuset.
2140
 * Must be called with callback_mutex held.
2141
 **/
2142
void cpuset_cpus_allowed_locked(struct task_struct *tsk, struct cpumask *pmask)
2143
{
2144
	task_lock(tsk);
2145
	guarantee_online_cpus(task_cs(tsk), pmask);
2146
	task_unlock(tsk);
L
Linus Torvalds 已提交
2147 2148 2149 2150
}

void cpuset_init_current_mems_allowed(void)
{
2151
	nodes_setall(current->mems_allowed);
L
Linus Torvalds 已提交
2152 2153
}

2154 2155 2156 2157 2158 2159
/**
 * 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
2160
 * subset of node_states[N_HIGH_MEMORY], even if this means going outside the
2161 2162 2163 2164 2165 2166 2167
 * tasks cpuset.
 **/

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

2168
	mutex_lock(&callback_mutex);
2169
	task_lock(tsk);
2170
	guarantee_online_mems(task_cs(tsk), &mask);
2171
	task_unlock(tsk);
2172
	mutex_unlock(&callback_mutex);
2173 2174 2175 2176

	return mask;
}

2177
/**
2178 2179
 * cpuset_nodemask_valid_mems_allowed - check nodemask vs. curremt mems_allowed
 * @nodemask: the nodemask to be checked
2180
 *
2181
 * Are any of the nodes in the nodemask allowed in current->mems_allowed?
L
Linus Torvalds 已提交
2182
 */
2183
int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask)
L
Linus Torvalds 已提交
2184
{
2185
	return nodes_intersects(*nodemask, current->mems_allowed);
L
Linus Torvalds 已提交
2186 2187
}

2188
/*
2189 2190 2191 2192
 * 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.
2193
 */
2194
static const struct cpuset *nearest_hardwall_ancestor(const struct cpuset *cs)
2195
{
2196
	while (!(is_mem_exclusive(cs) || is_mem_hardwall(cs)) && cs->parent)
2197 2198 2199 2200
		cs = cs->parent;
	return cs;
}

2201
/**
2202 2203
 * cpuset_node_allowed_softwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2204
 * @gfp_mask: memory allocation flags
2205
 *
2206 2207 2208 2209 2210 2211
 * If we're in interrupt, yes, we can always allocate.  If __GFP_THISNODE is
 * set, yes, we can always allocate.  If node is in our task's mems_allowed,
 * yes.  If it's not a __GFP_HARDWALL request and this node is in the nearest
 * hardwalled cpuset ancestor to this task's cpuset, yes.  If the task has been
 * OOM killed and has access to memory reserves as specified by the TIF_MEMDIE
 * flag, yes.
2212 2213
 * Otherwise, no.
 *
2214 2215 2216
 * If __GFP_HARDWALL is set, cpuset_node_allowed_softwall() reduces to
 * cpuset_node_allowed_hardwall().  Otherwise, cpuset_node_allowed_softwall()
 * might sleep, and might allow a node from an enclosing cpuset.
2217
 *
2218 2219
 * cpuset_node_allowed_hardwall() only handles the simpler case of hardwall
 * cpusets, and never sleeps.
2220 2221 2222 2223 2224 2225 2226
 *
 * 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'.
 *
2227
 * GFP_USER allocations are marked with the __GFP_HARDWALL bit,
2228 2229
 * and do not allow allocations outside the current tasks cpuset
 * unless the task has been OOM killed as is marked TIF_MEMDIE.
2230
 * GFP_KERNEL allocations are not so marked, so can escape to the
2231
 * nearest enclosing hardwalled ancestor cpuset.
2232
 *
2233 2234 2235 2236 2237 2238 2239
 * 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.
2240
 *
2241
 * The first call here from mm/page_alloc:get_page_from_freelist()
2242 2243 2244
 * 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).
2245 2246 2247 2248 2249 2250
 *
 * 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:
2251 2252
 *	in_interrupt - any node ok (current task context irrelevant)
 *	GFP_ATOMIC   - any node ok
2253
 *	TIF_MEMDIE   - any node ok
2254
 *	GFP_KERNEL   - any node in enclosing hardwalled cpuset ok
2255
 *	GFP_USER     - only nodes in current tasks mems allowed ok.
2256 2257
 *
 * Rule:
2258
 *    Don't call cpuset_node_allowed_softwall if you can't sleep, unless you
2259 2260
 *    pass in the __GFP_HARDWALL flag set in gfp_flag, which disables
 *    the code that might scan up ancestor cpusets and sleep.
2261
 */
2262
int __cpuset_node_allowed_softwall(int node, gfp_t gfp_mask)
L
Linus Torvalds 已提交
2263
{
2264
	const struct cpuset *cs;	/* current cpuset ancestors */
2265
	int allowed;			/* is allocation in zone z allowed? */
2266

2267
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
2268
		return 1;
2269
	might_sleep_if(!(gfp_mask & __GFP_HARDWALL));
2270 2271
	if (node_isset(node, current->mems_allowed))
		return 1;
2272 2273 2274 2275 2276 2277
	/*
	 * 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;
2278 2279 2280
	if (gfp_mask & __GFP_HARDWALL)	/* If hardwall request, stop here */
		return 0;

2281 2282 2283
	if (current->flags & PF_EXITING) /* Let dying task have memory */
		return 1;

2284
	/* Not hardwall and node outside mems_allowed: scan up cpusets */
2285
	mutex_lock(&callback_mutex);
2286 2287

	task_lock(current);
2288
	cs = nearest_hardwall_ancestor(task_cs(current));
2289 2290
	task_unlock(current);

2291
	allowed = node_isset(node, cs->mems_allowed);
2292
	mutex_unlock(&callback_mutex);
2293
	return allowed;
L
Linus Torvalds 已提交
2294 2295
}

2296
/*
2297 2298
 * cpuset_node_allowed_hardwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2299 2300
 * @gfp_mask: memory allocation flags
 *
2301 2302 2303 2304 2305
 * If we're in interrupt, yes, we can always allocate.  If __GFP_THISNODE is
 * set, yes, we can always allocate.  If node is in our task's 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.
2306 2307 2308 2309 2310 2311 2312
 *
 * 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'.
 *
2313 2314
 * Unlike the cpuset_node_allowed_softwall() variant, above,
 * this variant requires that the node be in the current task's
2315 2316 2317 2318
 * 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.
 */
2319
int __cpuset_node_allowed_hardwall(int node, gfp_t gfp_mask)
2320 2321 2322 2323 2324
{
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
		return 1;
	if (node_isset(node, current->mems_allowed))
		return 1;
D
Daniel Walker 已提交
2325 2326 2327 2328 2329 2330
	/*
	 * 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;
2331 2332 2333
	return 0;
}

P
Paul Jackson 已提交
2334 2335 2336
/**
 * cpuset_lock - lock out any changes to cpuset structures
 *
2337
 * The out of memory (oom) code needs to mutex_lock cpusets
P
Paul Jackson 已提交
2338
 * from being changed while it scans the tasklist looking for a
2339
 * task in an overlapping cpuset.  Expose callback_mutex via this
P
Paul Jackson 已提交
2340 2341
 * cpuset_lock() routine, so the oom code can lock it, before
 * locking the task list.  The tasklist_lock is a spinlock, so
2342
 * must be taken inside callback_mutex.
P
Paul Jackson 已提交
2343 2344 2345 2346
 */

void cpuset_lock(void)
{
2347
	mutex_lock(&callback_mutex);
P
Paul Jackson 已提交
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357
}

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

void cpuset_unlock(void)
{
2358
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
2359 2360
}

2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
/**
 * 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);

2399
/**
2400 2401 2402 2403 2404 2405 2406 2407
 * 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.
2408 2409
 **/

2410 2411
int cpuset_mems_allowed_intersects(const struct task_struct *tsk1,
				   const struct task_struct *tsk2)
2412
{
2413
	return nodes_intersects(tsk1->mems_allowed, tsk2->mems_allowed);
2414 2415
}

2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
/**
 * 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);
}

2439 2440 2441 2442 2443 2444
/*
 * 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.
 */

2445
int cpuset_memory_pressure_enabled __read_mostly;
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467

/**
 * 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);
2468
	fmeter_markevent(&task_cs(current)->fmeter);
2469 2470 2471
	task_unlock(current);
}

2472
#ifdef CONFIG_PROC_PID_CPUSET
L
Linus Torvalds 已提交
2473 2474 2475 2476
/*
 * proc_cpuset_show()
 *  - Print tasks cpuset path into seq_file.
 *  - Used for /proc/<pid>/cpuset.
2477 2478
 *  - 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,
2479
 *    and we take cgroup_mutex, keeping cpuset_attach() from changing it
2480
 *    anyway.
L
Linus Torvalds 已提交
2481
 */
P
Paul Jackson 已提交
2482
static int proc_cpuset_show(struct seq_file *m, void *unused_v)
L
Linus Torvalds 已提交
2483
{
2484
	struct pid *pid;
L
Linus Torvalds 已提交
2485 2486
	struct task_struct *tsk;
	char *buf;
2487
	struct cgroup_subsys_state *css;
2488
	int retval;
L
Linus Torvalds 已提交
2489

2490
	retval = -ENOMEM;
L
Linus Torvalds 已提交
2491 2492
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
2493 2494 2495
		goto out;

	retval = -ESRCH;
2496 2497
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
2498 2499
	if (!tsk)
		goto out_free;
L
Linus Torvalds 已提交
2500

2501
	retval = -EINVAL;
2502 2503 2504
	cgroup_lock();
	css = task_subsys_state(tsk, cpuset_subsys_id);
	retval = cgroup_path(css->cgroup, buf, PAGE_SIZE);
L
Linus Torvalds 已提交
2505
	if (retval < 0)
2506
		goto out_unlock;
L
Linus Torvalds 已提交
2507 2508
	seq_puts(m, buf);
	seq_putc(m, '\n');
2509
out_unlock:
2510
	cgroup_unlock();
2511 2512
	put_task_struct(tsk);
out_free:
L
Linus Torvalds 已提交
2513
	kfree(buf);
2514
out:
L
Linus Torvalds 已提交
2515 2516 2517 2518 2519
	return retval;
}

static int cpuset_open(struct inode *inode, struct file *file)
{
2520 2521
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cpuset_show, pid);
L
Linus Torvalds 已提交
2522 2523
}

2524
const struct file_operations proc_cpuset_operations = {
L
Linus Torvalds 已提交
2525 2526 2527 2528 2529
	.open		= cpuset_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};
2530
#endif /* CONFIG_PROC_PID_CPUSET */
L
Linus Torvalds 已提交
2531 2532

/* Display task cpus_allowed, mems_allowed in /proc/<pid>/status file. */
2533 2534 2535
void cpuset_task_status_allowed(struct seq_file *m, struct task_struct *task)
{
	seq_printf(m, "Cpus_allowed:\t");
2536
	seq_cpumask(m, &task->cpus_allowed);
2537
	seq_printf(m, "\n");
2538
	seq_printf(m, "Cpus_allowed_list:\t");
2539
	seq_cpumask_list(m, &task->cpus_allowed);
2540
	seq_printf(m, "\n");
2541
	seq_printf(m, "Mems_allowed:\t");
2542
	seq_nodemask(m, &task->mems_allowed);
2543
	seq_printf(m, "\n");
2544
	seq_printf(m, "Mems_allowed_list:\t");
2545
	seq_nodemask_list(m, &task->mems_allowed);
2546
	seq_printf(m, "\n");
L
Linus Torvalds 已提交
2547
}