cpuset.c 74.3 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>
40
#include <linux/export.h>
L
Linus Torvalds 已提交
41 42 43 44
#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
#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>
A
Arun Sharma 已提交
58
#include <linux/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
	nodemask_t mems_allowed;	/* Memory Nodes allowed to tasks */

	struct cpuset *parent;		/* my parent */

100
	struct fmeter fmeter;		/* memory_pressure filter */
P
Paul Jackson 已提交
101 102 103

	/* partition number for rebuild_sched_domains() */
	int pn;
104

105 106 107
	/* for custom sched domain */
	int relax_domain_level;

108
	/* used for walking a cpuset hierarchy */
109
	struct list_head stack_list;
L
Linus Torvalds 已提交
110 111
};

112 113 114 115 116 117 118 119 120 121 122 123 124 125
/* 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
126 127 128 129
/* bits in struct cpuset flags field */
typedef enum {
	CS_CPU_EXCLUSIVE,
	CS_MEM_EXCLUSIVE,
130
	CS_MEM_HARDWALL,
131
	CS_MEMORY_MIGRATE,
P
Paul Jackson 已提交
132
	CS_SCHED_LOAD_BALANCE,
133 134
	CS_SPREAD_PAGE,
	CS_SPREAD_SLAB,
L
Linus Torvalds 已提交
135 136 137 138 139
} cpuset_flagbits_t;

/* convenient tests for these bits */
static inline int is_cpu_exclusive(const struct cpuset *cs)
{
140
	return test_bit(CS_CPU_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
141 142 143 144
}

static inline int is_mem_exclusive(const struct cpuset *cs)
{
145
	return test_bit(CS_MEM_EXCLUSIVE, &cs->flags);
L
Linus Torvalds 已提交
146 147
}

148 149 150 151 152
static inline int is_mem_hardwall(const struct cpuset *cs)
{
	return test_bit(CS_MEM_HARDWALL, &cs->flags);
}

P
Paul Jackson 已提交
153 154 155 156 157
static inline int is_sched_load_balance(const struct cpuset *cs)
{
	return test_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
}

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

163 164 165 166 167 168 169 170 171 172
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
173 174 175 176 177
static struct cpuset top_cpuset = {
	.flags = ((1 << CS_CPU_EXCLUSIVE) | (1 << CS_MEM_EXCLUSIVE)),
};

/*
178 179 180 181 182 183 184
 * 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.
185
 *
186
 * A task must hold both mutexes to modify cpusets.  If a task
187
 * holds cgroup_mutex, then it blocks others wanting that mutex,
188
 * ensuring that it is the only task able to also acquire callback_mutex
189 190
 * and be able to modify cpusets.  It can perform various checks on
 * the cpuset structure first, knowing nothing will change.  It can
191
 * also allocate memory while just holding cgroup_mutex.  While it is
192
 * performing these checks, various callback routines can briefly
193 194
 * acquire callback_mutex to query cpusets.  Once it is ready to make
 * the changes, it takes callback_mutex, blocking everyone else.
195 196
 *
 * Calls to the kernel memory allocator can not be made while holding
197
 * callback_mutex, as that would risk double tripping on callback_mutex
198 199 200
 * from one of the callbacks into the cpuset code from within
 * __alloc_pages().
 *
201
 * If a task is only holding callback_mutex, then it has read-only
202 203
 * access to cpusets.
 *
204 205 206
 * Now, the task_struct fields mems_allowed and mempolicy may be changed
 * by other task, we use alloc_lock in the task_struct fields to protect
 * them.
207
 *
208
 * The cpuset_common_file_read() handlers only hold callback_mutex across
209 210 211
 * small pieces of code, such as when reading out possibly multi-word
 * cpumasks and nodemasks.
 *
212 213
 * 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
214 215
 */

216
static DEFINE_MUTEX(callback_mutex);
217

218 219 220 221 222 223 224 225 226 227 228
/*
 * 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);

229 230
/*
 * This is ugly, but preserves the userspace API for existing cpuset
231
 * users. If someone tries to mount the "cpuset" filesystem, we
232 233
 * silently switch it to mount "cgroup" instead
 */
A
Al Viro 已提交
234 235
static struct dentry *cpuset_mount(struct file_system_type *fs_type,
			 int flags, const char *unused_dev_name, void *data)
L
Linus Torvalds 已提交
236
{
237
	struct file_system_type *cgroup_fs = get_fs_type("cgroup");
A
Al Viro 已提交
238
	struct dentry *ret = ERR_PTR(-ENODEV);
239 240 241 242
	if (cgroup_fs) {
		char mountopts[] =
			"cpuset,noprefix,"
			"release_agent=/sbin/cpuset_release_agent";
A
Al Viro 已提交
243 244
		ret = cgroup_fs->mount(cgroup_fs, flags,
					   unused_dev_name, mountopts);
245 246 247
		put_filesystem(cgroup_fs);
	}
	return ret;
L
Linus Torvalds 已提交
248 249 250 251
}

static struct file_system_type cpuset_fs_type = {
	.name = "cpuset",
A
Al Viro 已提交
252
	.mount = cpuset_mount,
L
Linus Torvalds 已提交
253 254 255
};

/*
256
 * Return in pmask the portion of a cpusets's cpus_allowed that
L
Linus Torvalds 已提交
257 258 259 260 261 262 263 264 265
 * 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.
 *
266
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
267 268
 */

269 270
static void guarantee_online_cpus(const struct cpuset *cs,
				  struct cpumask *pmask)
L
Linus Torvalds 已提交
271
{
272
	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
L
Linus Torvalds 已提交
273 274
		cs = cs->parent;
	if (cs)
275
		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
L
Linus Torvalds 已提交
276
	else
277 278
		cpumask_copy(pmask, cpu_online_mask);
	BUG_ON(!cpumask_intersects(pmask, cpu_online_mask));
L
Linus Torvalds 已提交
279 280 281 282
}

/*
 * Return in *pmask the portion of a cpusets's mems_allowed that
283 284 285 286
 * 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
287 288
 *
 * One way or another, we guarantee to return some non-empty subset
289
 * of node_states[N_HIGH_MEMORY].
L
Linus Torvalds 已提交
290
 *
291
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
292 293 294 295
 */

static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
{
296 297
	while (cs && !nodes_intersects(cs->mems_allowed,
					node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
298 299
		cs = cs->parent;
	if (cs)
300 301
		nodes_and(*pmask, cs->mems_allowed,
					node_states[N_HIGH_MEMORY]);
L
Linus Torvalds 已提交
302
	else
303 304
		*pmask = node_states[N_HIGH_MEMORY];
	BUG_ON(!nodes_intersects(*pmask, node_states[N_HIGH_MEMORY]));
L
Linus Torvalds 已提交
305 306
}

307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
/*
 * update task's spread flag if cpuset's page/slab spread flag is set
 *
 * Called with callback_mutex/cgroup_mutex held
 */
static void cpuset_update_task_spread_flag(struct cpuset *cs,
					struct task_struct *tsk)
{
	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;
}

L
Linus Torvalds 已提交
325 326 327 328 329
/*
 * 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
330
 * are only set if the other's are set.  Call holding cgroup_mutex.
L
Linus Torvalds 已提交
331 332 333 334
 */

static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
{
335
	return	cpumask_subset(p->cpus_allowed, q->cpus_allowed) &&
L
Linus Torvalds 已提交
336 337 338 339 340
		nodes_subset(p->mems_allowed, q->mems_allowed) &&
		is_cpu_exclusive(p) <= is_cpu_exclusive(q) &&
		is_mem_exclusive(p) <= is_mem_exclusive(q);
}

341 342 343 344 345 346
/**
 * 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)
{
347 348 349 350 351 352 353 354 355 356 357 358 359
	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;
360 361 362 363 364 365 366 367
}

/**
 * free_trial_cpuset - free the trial cpuset
 * @trial: the trial cpuset to be freed
 */
static void free_trial_cpuset(struct cpuset *trial)
{
368
	free_cpumask_var(trial->cpus_allowed);
369 370 371
	kfree(trial);
}

L
Linus Torvalds 已提交
372 373 374 375 376 377 378
/*
 * 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
379
 * cgroup_mutex held.
L
Linus Torvalds 已提交
380 381 382 383 384 385 386 387 388 389 390 391 392 393
 *
 * '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)
{
394
	struct cgroup *cont;
L
Linus Torvalds 已提交
395 396 397
	struct cpuset *c, *par;

	/* Each of our child cpusets must be a subset of us */
398 399
	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
		if (!is_cpuset_subset(cgroup_cs(cont), trial))
L
Linus Torvalds 已提交
400 401 402 403
			return -EBUSY;
	}

	/* Remaining checks don't apply to root cpuset */
404
	if (cur == &top_cpuset)
L
Linus Torvalds 已提交
405 406
		return 0;

407 408
	par = cur->parent;

L
Linus Torvalds 已提交
409 410 411 412
	/* We must be a subset of our parent cpuset */
	if (!is_cpuset_subset(trial, par))
		return -EACCES;

413 414 415 416
	/*
	 * If either I or some sibling (!= me) is exclusive, we can't
	 * overlap
	 */
417 418
	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
		c = cgroup_cs(cont);
L
Linus Torvalds 已提交
419 420
		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
		    c != cur &&
421
		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
L
Linus Torvalds 已提交
422 423 424 425 426 427 428
			return -EINVAL;
		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
		    c != cur &&
		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
			return -EINVAL;
	}

429 430
	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
	if (cgroup_task_count(cur->css.cgroup)) {
431
		if (cpumask_empty(trial->cpus_allowed) ||
432 433 434 435 436
		    nodes_empty(trial->mems_allowed)) {
			return -ENOSPC;
		}
	}

L
Linus Torvalds 已提交
437 438 439
	return 0;
}

440
#ifdef CONFIG_SMP
P
Paul Jackson 已提交
441
/*
442
 * Helper routine for generate_sched_domains().
P
Paul Jackson 已提交
443 444 445 446
 * Do cpusets a, b have overlapping cpus_allowed masks?
 */
static int cpusets_overlap(struct cpuset *a, struct cpuset *b)
{
447
	return cpumask_intersects(a->cpus_allowed, b->cpus_allowed);
P
Paul Jackson 已提交
448 449
}

450 451 452 453 454 455 456 457
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;
}

458 459 460 461 462 463 464 465 466 467 468 469 470 471
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);

472
		if (cpumask_empty(cp->cpus_allowed))
473 474 475 476 477 478 479 480 481 482 483 484
			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
485
/*
486 487 488 489 490 491 492 493 494
 * 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
495
 *
L
Li Zefan 已提交
496
 * See "What is sched_load_balance" in Documentation/cgroups/cpusets.txt
P
Paul Jackson 已提交
497 498 499 500 501 502 503
 * 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.
 *
504
 * Must be called with cgroup_lock held.
P
Paul Jackson 已提交
505 506
 *
 * The three key local variables below are:
507
 *    q  - a linked-list queue of cpuset pointers, used to implement a
P
Paul Jackson 已提交
508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538
 *	   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().
 */
539
static int generate_sched_domains(cpumask_var_t **domains,
540
			struct sched_domain_attr **attributes)
P
Paul Jackson 已提交
541
{
542
	LIST_HEAD(q);		/* queue of cpusets to be scanned */
P
Paul Jackson 已提交
543 544 545 546
	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 */
547
	cpumask_var_t *doms;	/* resulting partition; i.e. sched domains */
548
	struct sched_domain_attr *dattr;  /* attributes for custom domains */
549
	int ndoms = 0;		/* number of sched domains in result */
550
	int nslot;		/* next empty doms[] struct cpumask slot */
P
Paul Jackson 已提交
551 552

	doms = NULL;
553
	dattr = NULL;
554
	csa = NULL;
P
Paul Jackson 已提交
555 556 557

	/* Special case for the 99% of systems with one, full, sched domain */
	if (is_sched_load_balance(&top_cpuset)) {
558 559
		ndoms = 1;
		doms = alloc_sched_domains(ndoms);
P
Paul Jackson 已提交
560
		if (!doms)
561 562
			goto done;

563 564 565
		dattr = kmalloc(sizeof(struct sched_domain_attr), GFP_KERNEL);
		if (dattr) {
			*dattr = SD_ATTR_INIT;
566
			update_domain_attr_tree(dattr, &top_cpuset);
567
		}
568
		cpumask_copy(doms[0], top_cpuset.cpus_allowed);
569 570

		goto done;
P
Paul Jackson 已提交
571 572 573 574 575 576 577
	}

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

578 579
	list_add(&top_cpuset.stack_list, &q);
	while (!list_empty(&q)) {
P
Paul Jackson 已提交
580 581
		struct cgroup *cont;
		struct cpuset *child;   /* scans child cpusets of cp */
582

583 584 585
		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

586
		if (cpumask_empty(cp->cpus_allowed))
587 588
			continue;

589 590 591 592 593 594 595
		/*
		 * 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
596
			csa[csn++] = cp;
597 598
			continue;
		}
599

P
Paul Jackson 已提交
600 601
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
602
			list_add_tail(&child->stack_list, &q);
P
Paul Jackson 已提交
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
		}
  	}

	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;
			}
		}
	}

633 634 635 636
	/*
	 * Now we know how many domains to create.
	 * Convert <csn, csa> to <ndoms, doms> and populate cpu masks.
	 */
637
	doms = alloc_sched_domains(ndoms);
638
	if (!doms)
639 640 641 642 643 644
		goto done;

	/*
	 * The rest of the code, including the scheduler, can deal with
	 * dattr==NULL case. No need to abort if alloc fails.
	 */
645
	dattr = kmalloc(ndoms * sizeof(struct sched_domain_attr), GFP_KERNEL);
P
Paul Jackson 已提交
646 647 648

	for (nslot = 0, i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
649
		struct cpumask *dp;
P
Paul Jackson 已提交
650 651
		int apn = a->pn;

652 653 654 655 656
		if (apn < 0) {
			/* Skip completed partitions */
			continue;
		}

657
		dp = doms[nslot];
658 659 660 661 662 663 664 665 666 667

		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
668
			}
669 670
			continue;
		}
P
Paul Jackson 已提交
671

672
		cpumask_clear(dp);
673 674 675 676 677 678
		if (dattr)
			*(dattr + nslot) = SD_ATTR_INIT;
		for (j = i; j < csn; j++) {
			struct cpuset *b = csa[j];

			if (apn == b->pn) {
679
				cpumask_or(dp, dp, b->cpus_allowed);
680 681 682 683 684
				if (dattr)
					update_domain_attr_tree(dattr + nslot, b);

				/* Done with this partition */
				b->pn = -1;
P
Paul Jackson 已提交
685 686
			}
		}
687
		nslot++;
P
Paul Jackson 已提交
688 689 690
	}
	BUG_ON(nslot != ndoms);

691 692 693
done:
	kfree(csa);

694 695 696 697 698 699 700
	/*
	 * Fallback to the default domain if kmalloc() failed.
	 * See comments in partition_sched_domains().
	 */
	if (doms == NULL)
		ndoms = 1;

701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
	*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;
719
	cpumask_var_t *doms;
720 721
	int ndoms;

722
	get_online_cpus();
723 724 725 726 727 728 729 730 731

	/* 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);

732
	put_online_cpus();
733
}
734 735 736 737 738
#else /* !CONFIG_SMP */
static void do_rebuild_sched_domains(struct work_struct *unused)
{
}

739
static int generate_sched_domains(cpumask_var_t **domains,
740 741 742 743 744 745
			struct sched_domain_attr **attributes)
{
	*domains = NULL;
	return 1;
}
#endif /* CONFIG_SMP */
P
Paul Jackson 已提交
746

747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
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)
{
770
	queue_work(cpuset_wq, &rebuild_sched_domains_work);
771 772 773 774 775 776 777 778 779 780 781 782 783 784
}

/*
 * 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
785 786
}

C
Cliff Wickman 已提交
787 788 789 790 791
/**
 * 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
 *
792
 * Call with cgroup_mutex held.  May take callback_mutex during call.
C
Cliff Wickman 已提交
793 794 795
 * 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).
796
 */
797 798
static int cpuset_test_cpumask(struct task_struct *tsk,
			       struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
799
{
800
	return !cpumask_equal(&tsk->cpus_allowed,
C
Cliff Wickman 已提交
801 802
			(cgroup_cs(scan->cg))->cpus_allowed);
}
803

C
Cliff Wickman 已提交
804 805 806 807 808 809 810 811 812 813 814
/**
 * 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.
 */
815 816
static void cpuset_change_cpumask(struct task_struct *tsk,
				  struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
817
{
818
	set_cpus_allowed_ptr(tsk, ((cgroup_cs(scan->cg))->cpus_allowed));
C
Cliff Wickman 已提交
819 820
}

821 822 823
/**
 * 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
824
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
825 826 827 828 829 830
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
831 832
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
833
 */
834
static void update_tasks_cpumask(struct cpuset *cs, struct ptr_heap *heap)
835 836 837 838 839 840
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = cpuset_test_cpumask;
	scan.process_task = cpuset_change_cpumask;
841 842
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
843 844
}

C
Cliff Wickman 已提交
845 846 847 848 849
/**
 * 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
 */
850 851
static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
			  const char *buf)
L
Linus Torvalds 已提交
852
{
853
	struct ptr_heap heap;
C
Cliff Wickman 已提交
854 855
	int retval;
	int is_load_balanced;
L
Linus Torvalds 已提交
856

857 858 859 860
	/* top_cpuset.cpus_allowed tracks cpu_online_map; it's read-only */
	if (cs == &top_cpuset)
		return -EACCES;

861
	/*
862
	 * An empty cpus_allowed is ok only if the cpuset has no tasks.
863 864 865
	 * Since cpulist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have cpus.
866
	 */
867
	if (!*buf) {
868
		cpumask_clear(trialcs->cpus_allowed);
869
	} else {
870
		retval = cpulist_parse(buf, trialcs->cpus_allowed);
871 872
		if (retval < 0)
			return retval;
873

874
		if (!cpumask_subset(trialcs->cpus_allowed, cpu_active_mask))
875
			return -EINVAL;
876
	}
877
	retval = validate_change(cs, trialcs);
878 879
	if (retval < 0)
		return retval;
P
Paul Jackson 已提交
880

P
Paul Menage 已提交
881
	/* Nothing to do if the cpus didn't change */
882
	if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed))
P
Paul Menage 已提交
883
		return 0;
C
Cliff Wickman 已提交
884

885 886 887 888
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval)
		return retval;

889
	is_load_balanced = is_sched_load_balance(trialcs);
P
Paul Jackson 已提交
890

891
	mutex_lock(&callback_mutex);
892
	cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed);
893
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
894

P
Paul Menage 已提交
895 896
	/*
	 * Scan tasks in the cpuset, and update the cpumasks of any
C
Cliff Wickman 已提交
897
	 * that need an update.
P
Paul Menage 已提交
898
	 */
899 900 901
	update_tasks_cpumask(cs, &heap);

	heap_free(&heap);
C
Cliff Wickman 已提交
902

P
Paul Menage 已提交
903
	if (is_load_balanced)
904
		async_rebuild_sched_domains();
905
	return 0;
L
Linus Torvalds 已提交
906 907
}

908 909 910 911 912 913 914 915
/*
 * 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.
 *
916
 *    Call holding cgroup_mutex, so current's cpuset won't change
917
 *    during this call, as manage_mutex holds off any cpuset_attach()
918 919
 *    calls.  Therefore we don't need to take task_lock around the
 *    call to guarantee_online_mems(), as we know no one is changing
920
 *    our task's cpuset.
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936
 *
 *    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.
 */

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

	tsk->mems_allowed = *to;

	do_migrate_pages(mm, from, to, MPOL_MF_MOVE_ALL);

937
	guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
938 939
}

940
/*
941 942 943 944 945 946 947 948 949 950 951
 * cpuset_change_task_nodemask - change task's mems_allowed and mempolicy
 * @tsk: the task to change
 * @newmems: new nodes that the task will be set
 *
 * In order to avoid seeing no nodes if the old and new nodes are disjoint,
 * we structure updates as setting all new allowed nodes, then clearing newly
 * disallowed ones.
 */
static void cpuset_change_task_nodemask(struct task_struct *tsk,
					nodemask_t *newmems)
{
952 953
	bool masks_disjoint = !nodes_intersects(*newmems, tsk->mems_allowed);

954 955 956 957 958 959 960 961 962 963 964
repeat:
	/*
	 * 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;
	if (current->flags & PF_EXITING) /* Let dying task have memory */
		return;

	task_lock(tsk);
965
	nodes_or(tsk->mems_allowed, tsk->mems_allowed, *newmems);
966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
	mpol_rebind_task(tsk, newmems, MPOL_REBIND_STEP1);

	/*
	 * ensure checking ->mems_allowed_change_disable after setting all new
	 * allowed nodes.
	 *
	 * the read-side task can see an nodemask with new allowed nodes and
	 * old allowed nodes. and if it allocates page when cpuset clears newly
	 * disallowed ones continuous, it can see the new allowed bits.
	 *
	 * And if setting all new allowed nodes is after the checking, setting
	 * all new allowed nodes and clearing newly disallowed ones will be done
	 * continuous, and the read-side task may find no node to alloc page.
	 */
	smp_mb();

	/*
	 * Allocation of memory is very fast, we needn't sleep when waiting
984 985
	 * for the read-side.  No wait is necessary, however, if at least one
	 * node remains unchanged.
986
	 */
987 988
	while (masks_disjoint &&
			ACCESS_ONCE(tsk->mems_allowed_change_disable)) {
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
		task_unlock(tsk);
		if (!task_curr(tsk))
			yield();
		goto repeat;
	}

	/*
	 * ensure checking ->mems_allowed_change_disable before clearing all new
	 * disallowed nodes.
	 *
	 * if clearing newly disallowed bits before the checking, the read-side
	 * task may find no node to alloc page.
	 */
	smp_mb();

	mpol_rebind_task(tsk, newmems, MPOL_REBIND_STEP2);
1005
	tsk->mems_allowed = *newmems;
1006
	task_unlock(tsk);
1007 1008 1009 1010 1011 1012
}

/*
 * Update task's mems_allowed and rebind its mempolicy and vmas' mempolicy
 * of it to cpuset's new mems_allowed, and migrate pages to new nodes if
 * memory_migrate flag is set. Called with cgroup_mutex held.
1013 1014 1015 1016 1017 1018 1019 1020
 */
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;
1021
	static nodemask_t newmems;	/* protected by cgroup_mutex */
1022 1023

	cs = cgroup_cs(scan->cg);
1024
	guarantee_online_mems(cs, &newmems);
1025

1026
	cpuset_change_task_nodemask(p, &newmems);
1027

1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
	mm = get_task_mm(p);
	if (!mm)
		return;

	migrate = is_memory_migrate(cs);

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

1040 1041
static void *cpuset_being_rebound;

1042 1043 1044 1045
/**
 * 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
1046
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
1047 1048
 *
 * Called with cgroup_mutex held
1049 1050
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
1051
 */
1052 1053
static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem,
				 struct ptr_heap *heap)
L
Linus Torvalds 已提交
1054
{
1055
	struct cgroup_scanner scan;
1056

1057
	cpuset_being_rebound = cs;		/* causes mpol_dup() rebind */
1058

1059 1060 1061
	scan.cg = cs->css.cgroup;
	scan.test_task = NULL;
	scan.process_task = cpuset_change_nodemask;
1062
	scan.heap = heap;
1063
	scan.data = (nodemask_t *)oldmem;
1064 1065

	/*
1066 1067 1068 1069 1070 1071
	 * 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.
1072
	 * It's ok if we rebind the same mm twice; mpol_rebind_mm()
1073
	 * is idempotent.  Also migrate pages in each mm to new nodes.
1074
	 */
1075
	cgroup_scan_tasks(&scan);
1076

1077
	/* We're done rebinding vmas to this cpuset's new mems_allowed. */
1078
	cpuset_being_rebound = NULL;
L
Linus Torvalds 已提交
1079 1080
}

1081 1082 1083
/*
 * Handle user request to change the 'mems' memory placement
 * of a cpuset.  Needs to validate the request, update the
1084 1085 1086 1087
 * cpusets mems_allowed, and for each task in the cpuset,
 * update mems_allowed and rebind task's mempolicy and any vma
 * mempolicies and if the cpuset is marked 'memory_migrate',
 * migrate the tasks pages to the new memory.
1088 1089 1090 1091 1092 1093
 *
 * 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.
 */
1094 1095
static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs,
			   const char *buf)
1096
{
1097
	NODEMASK_ALLOC(nodemask_t, oldmem, GFP_KERNEL);
1098
	int retval;
1099
	struct ptr_heap heap;
1100

1101 1102 1103
	if (!oldmem)
		return -ENOMEM;

1104 1105 1106 1107
	/*
	 * top_cpuset.mems_allowed tracks node_stats[N_HIGH_MEMORY];
	 * it's read-only
	 */
1108 1109 1110 1111
	if (cs == &top_cpuset) {
		retval = -EACCES;
		goto done;
	}
1112 1113 1114 1115 1116 1117 1118 1119

	/*
	 * 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) {
1120
		nodes_clear(trialcs->mems_allowed);
1121
	} else {
1122
		retval = nodelist_parse(buf, trialcs->mems_allowed);
1123 1124 1125
		if (retval < 0)
			goto done;

1126
		if (!nodes_subset(trialcs->mems_allowed,
1127 1128 1129 1130
				node_states[N_HIGH_MEMORY])) {
			retval =  -EINVAL;
			goto done;
		}
1131
	}
1132 1133
	*oldmem = cs->mems_allowed;
	if (nodes_equal(*oldmem, trialcs->mems_allowed)) {
1134 1135 1136
		retval = 0;		/* Too easy - nothing to do */
		goto done;
	}
1137
	retval = validate_change(cs, trialcs);
1138 1139 1140
	if (retval < 0)
		goto done;

1141 1142 1143 1144
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval < 0)
		goto done;

1145
	mutex_lock(&callback_mutex);
1146
	cs->mems_allowed = trialcs->mems_allowed;
1147 1148
	mutex_unlock(&callback_mutex);

1149
	update_tasks_nodemask(cs, oldmem, &heap);
1150 1151

	heap_free(&heap);
1152
done:
1153
	NODEMASK_FREE(oldmem);
1154 1155 1156
	return retval;
}

1157 1158 1159 1160 1161
int current_cpuset_is_being_rebound(void)
{
	return task_cs(current) == cpuset_being_rebound;
}

1162
static int update_relax_domain_level(struct cpuset *cs, s64 val)
1163
{
1164
#ifdef CONFIG_SMP
1165
	if (val < -1 || val >= sched_domain_level_max)
1166
		return -EINVAL;
1167
#endif
1168 1169 1170

	if (val != cs->relax_domain_level) {
		cs->relax_domain_level = val;
1171 1172
		if (!cpumask_empty(cs->cpus_allowed) &&
		    is_sched_load_balance(cs))
1173
			async_rebuild_sched_domains();
1174 1175 1176 1177 1178
	}

	return 0;
}

1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
/*
 * cpuset_change_flag - make a task's spread flags the same as its cpuset's
 * @tsk: task to be updated
 * @scan: struct cgroup_scanner containing the cgroup of the task
 *
 * Called by cgroup_scan_tasks() for each task in a cgroup.
 *
 * We don't need to re-check for the cgroup/cpuset membership, since we're
 * holding cgroup_lock() at this point.
 */
static void cpuset_change_flag(struct task_struct *tsk,
				struct cgroup_scanner *scan)
{
	cpuset_update_task_spread_flag(cgroup_cs(scan->cg), tsk);
}

/*
 * update_tasks_flags - update the spread flags of tasks in the cpuset.
 * @cs: the cpuset in which each task's spread flags needs to be changed
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
 */
static void update_tasks_flags(struct cpuset *cs, struct ptr_heap *heap)
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = NULL;
	scan.process_task = cpuset_change_flag;
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
}

L
Linus Torvalds 已提交
1219 1220
/*
 * update_flag - read a 0 or a 1 in a file and update associated flag
1221 1222 1223
 * bit:		the bit to update (see cpuset_flagbits_t)
 * cs:		the cpuset to update
 * turning_on: 	whether the flag is being set or cleared
1224
 *
1225
 * Call with cgroup_mutex held.
L
Linus Torvalds 已提交
1226 1227
 */

1228 1229
static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
		       int turning_on)
L
Linus Torvalds 已提交
1230
{
1231
	struct cpuset *trialcs;
R
Rakib Mullick 已提交
1232
	int balance_flag_changed;
1233 1234 1235
	int spread_flag_changed;
	struct ptr_heap heap;
	int err;
L
Linus Torvalds 已提交
1236

1237 1238 1239 1240
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

L
Linus Torvalds 已提交
1241
	if (turning_on)
1242
		set_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1243
	else
1244
		clear_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1245

1246
	err = validate_change(cs, trialcs);
1247
	if (err < 0)
1248
		goto out;
P
Paul Jackson 已提交
1249

1250 1251 1252 1253
	err = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (err < 0)
		goto out;

P
Paul Jackson 已提交
1254
	balance_flag_changed = (is_sched_load_balance(cs) !=
1255
				is_sched_load_balance(trialcs));
P
Paul Jackson 已提交
1256

1257 1258 1259
	spread_flag_changed = ((is_spread_slab(cs) != is_spread_slab(trialcs))
			|| (is_spread_page(cs) != is_spread_page(trialcs)));

1260
	mutex_lock(&callback_mutex);
1261
	cs->flags = trialcs->flags;
1262
	mutex_unlock(&callback_mutex);
1263

1264
	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed)
1265
		async_rebuild_sched_domains();
P
Paul Jackson 已提交
1266

1267 1268 1269
	if (spread_flag_changed)
		update_tasks_flags(cs, &heap);
	heap_free(&heap);
1270 1271 1272
out:
	free_trial_cpuset(trialcs);
	return err;
L
Linus Torvalds 已提交
1273 1274
}

1275
/*
A
Adrian Bunk 已提交
1276
 * Frequency meter - How fast is some event occurring?
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372
 *
 * 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;
}

1373
/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
1374
static int cpuset_can_attach(struct cgroup_subsys *ss, struct cgroup *cont,
1375
			     struct task_struct *tsk)
L
Linus Torvalds 已提交
1376
{
1377
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1378

1379
	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
L
Linus Torvalds 已提交
1380
		return -ENOSPC;
1381

1382 1383 1384 1385 1386 1387 1388 1389 1390 1391
	/*
	 * 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
1392

1393 1394 1395
	return 0;
}

1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
static int cpuset_can_attach_task(struct cgroup *cgrp, struct task_struct *task)
{
	return security_task_setscheduler(task);
}

/*
 * Protected by cgroup_lock. The nodemasks must be stored globally because
 * dynamically allocating them is not allowed in pre_attach, and they must
 * persist among pre_attach, attach_task, and attach.
 */
static cpumask_var_t cpus_attach;
static nodemask_t cpuset_attach_nodemask_from;
static nodemask_t cpuset_attach_nodemask_to;

/* Set-up work for before attaching each task. */
static void cpuset_pre_attach(struct cgroup *cont)
{
	struct cpuset *cs = cgroup_cs(cont);

	if (cs == &top_cpuset)
		cpumask_copy(cpus_attach, cpu_possible_mask);
	else
		guarantee_online_cpus(cs, cpus_attach);

	guarantee_online_mems(cs, &cpuset_attach_nodemask_to);
}

/* Per-thread attachment work. */
static void cpuset_attach_task(struct cgroup *cont, struct task_struct *tsk)
1425 1426
{
	int err;
1427 1428
	struct cpuset *cs = cgroup_cs(cont);

1429 1430 1431 1432 1433 1434 1435
	/*
	 * can_attach beforehand should guarantee that this doesn't fail.
	 * TODO: have a better way to handle failure here
	 */
	err = set_cpus_allowed_ptr(tsk, cpus_attach);
	WARN_ON_ONCE(err);

1436
	cpuset_change_task_nodemask(tsk, &cpuset_attach_nodemask_to);
1437
	cpuset_update_task_spread_flag(cs, tsk);
1438
}
L
Linus Torvalds 已提交
1439

1440
static void cpuset_attach(struct cgroup_subsys *ss, struct cgroup *cont,
1441
			  struct cgroup *oldcont, struct task_struct *tsk)
1442 1443 1444 1445
{
	struct mm_struct *mm;
	struct cpuset *cs = cgroup_cs(cont);
	struct cpuset *oldcs = cgroup_cs(oldcont);
1446

1447 1448 1449 1450 1451 1452
	/*
	 * Change mm, possibly for multiple threads in a threadgroup. This is
	 * expensive and may sleep.
	 */
	cpuset_attach_nodemask_from = oldcs->mems_allowed;
	cpuset_attach_nodemask_to = cs->mems_allowed;
1453 1454
	mm = get_task_mm(tsk);
	if (mm) {
1455
		mpol_rebind_mm(mm, &cpuset_attach_nodemask_to);
1456
		if (is_memory_migrate(cs))
1457 1458
			cpuset_migrate_mm(mm, &cpuset_attach_nodemask_from,
					  &cpuset_attach_nodemask_to);
1459 1460
		mmput(mm);
	}
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465
}

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

typedef enum {
1466
	FILE_MEMORY_MIGRATE,
L
Linus Torvalds 已提交
1467 1468 1469 1470
	FILE_CPULIST,
	FILE_MEMLIST,
	FILE_CPU_EXCLUSIVE,
	FILE_MEM_EXCLUSIVE,
1471
	FILE_MEM_HARDWALL,
P
Paul Jackson 已提交
1472
	FILE_SCHED_LOAD_BALANCE,
1473
	FILE_SCHED_RELAX_DOMAIN_LEVEL,
1474 1475
	FILE_MEMORY_PRESSURE_ENABLED,
	FILE_MEMORY_PRESSURE,
1476 1477
	FILE_SPREAD_PAGE,
	FILE_SPREAD_SLAB,
L
Linus Torvalds 已提交
1478 1479
} cpuset_filetype_t;

1480 1481 1482 1483 1484 1485
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;

1486
	if (!cgroup_lock_live_group(cgrp))
1487 1488 1489
		return -ENODEV;

	switch (type) {
L
Linus Torvalds 已提交
1490
	case FILE_CPU_EXCLUSIVE:
1491
		retval = update_flag(CS_CPU_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1492 1493
		break;
	case FILE_MEM_EXCLUSIVE:
1494
		retval = update_flag(CS_MEM_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1495
		break;
1496 1497 1498
	case FILE_MEM_HARDWALL:
		retval = update_flag(CS_MEM_HARDWALL, cs, val);
		break;
P
Paul Jackson 已提交
1499
	case FILE_SCHED_LOAD_BALANCE:
1500
		retval = update_flag(CS_SCHED_LOAD_BALANCE, cs, val);
1501
		break;
1502
	case FILE_MEMORY_MIGRATE:
1503
		retval = update_flag(CS_MEMORY_MIGRATE, cs, val);
1504
		break;
1505
	case FILE_MEMORY_PRESSURE_ENABLED:
1506
		cpuset_memory_pressure_enabled = !!val;
1507 1508 1509 1510
		break;
	case FILE_MEMORY_PRESSURE:
		retval = -EACCES;
		break;
1511
	case FILE_SPREAD_PAGE:
1512
		retval = update_flag(CS_SPREAD_PAGE, cs, val);
1513 1514
		break;
	case FILE_SPREAD_SLAB:
1515
		retval = update_flag(CS_SPREAD_SLAB, cs, val);
1516
		break;
L
Linus Torvalds 已提交
1517 1518
	default:
		retval = -EINVAL;
1519
		break;
L
Linus Torvalds 已提交
1520
	}
1521
	cgroup_unlock();
L
Linus Torvalds 已提交
1522 1523 1524
	return retval;
}

1525 1526 1527 1528 1529 1530
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;

1531
	if (!cgroup_lock_live_group(cgrp))
1532
		return -ENODEV;
1533

1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545
	switch (type) {
	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
		retval = update_relax_domain_level(cs, val);
		break;
	default:
		retval = -EINVAL;
		break;
	}
	cgroup_unlock();
	return retval;
}

1546 1547 1548 1549 1550 1551 1552
/*
 * 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;
1553 1554
	struct cpuset *cs = cgroup_cs(cgrp);
	struct cpuset *trialcs;
1555 1556 1557 1558

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

1559
	trialcs = alloc_trial_cpuset(cs);
1560 1561 1562 1563
	if (!trialcs) {
		retval = -ENOMEM;
		goto out;
	}
1564

1565 1566
	switch (cft->private) {
	case FILE_CPULIST:
1567
		retval = update_cpumask(cs, trialcs, buf);
1568 1569
		break;
	case FILE_MEMLIST:
1570
		retval = update_nodemask(cs, trialcs, buf);
1571 1572 1573 1574 1575
		break;
	default:
		retval = -EINVAL;
		break;
	}
1576 1577

	free_trial_cpuset(trialcs);
1578
out:
1579 1580 1581 1582
	cgroup_unlock();
	return retval;
}

L
Linus Torvalds 已提交
1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594
/*
 * 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.
 */

1595
static size_t cpuset_sprintf_cpulist(char *page, struct cpuset *cs)
L
Linus Torvalds 已提交
1596
{
1597
	size_t count;
L
Linus Torvalds 已提交
1598

1599
	mutex_lock(&callback_mutex);
1600
	count = cpulist_scnprintf(page, PAGE_SIZE, cs->cpus_allowed);
1601
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1602

1603
	return count;
L
Linus Torvalds 已提交
1604 1605
}

1606
static size_t cpuset_sprintf_memlist(char *page, struct cpuset *cs)
L
Linus Torvalds 已提交
1607
{
1608
	size_t count;
L
Linus Torvalds 已提交
1609

1610
	mutex_lock(&callback_mutex);
1611
	count = nodelist_scnprintf(page, PAGE_SIZE, cs->mems_allowed);
1612
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1613

1614
	return count;
L
Linus Torvalds 已提交
1615 1616
}

1617 1618 1619 1620 1621
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
1622
{
1623
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1624 1625 1626 1627 1628
	cpuset_filetype_t type = cft->private;
	char *page;
	ssize_t retval = 0;
	char *s;

1629
	if (!(page = (char *)__get_free_page(GFP_TEMPORARY)))
L
Linus Torvalds 已提交
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
		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
1647
	retval = simple_read_from_buffer(buf, nbytes, ppos, page, s - page);
L
Linus Torvalds 已提交
1648 1649 1650 1651 1652
out:
	free_page((unsigned long)page);
	return retval;
}

1653 1654 1655 1656 1657 1658 1659 1660 1661
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);
1662 1663
	case FILE_MEM_HARDWALL:
		return is_mem_hardwall(cs);
1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
	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();
	}
1679 1680 1681

	/* Unreachable but makes gcc happy */
	return 0;
1682
}
L
Linus Torvalds 已提交
1683

1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
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();
	}
1694 1695 1696

	/* Unrechable but makes gcc happy */
	return 0;
1697 1698
}

L
Linus Torvalds 已提交
1699 1700 1701 1702 1703

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

1704 1705 1706 1707
static struct cftype files[] = {
	{
		.name = "cpus",
		.read = cpuset_common_file_read,
1708 1709
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * NR_CPUS),
1710 1711 1712 1713 1714 1715
		.private = FILE_CPULIST,
	},

	{
		.name = "mems",
		.read = cpuset_common_file_read,
1716 1717
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * MAX_NUMNODES),
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734
		.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,
	},

1735 1736 1737 1738 1739 1740 1741
	{
		.name = "mem_hardwall",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEM_HARDWALL,
	},

1742 1743 1744 1745 1746 1747 1748 1749 1750
	{
		.name = "sched_load_balance",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SCHED_LOAD_BALANCE,
	},

	{
		.name = "sched_relax_domain_level",
1751 1752
		.read_s64 = cpuset_read_s64,
		.write_s64 = cpuset_write_s64,
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767
		.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
1768
		.mode = S_IRUGO,
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
	},

	{
		.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,
	},
1784 1785
};

1786 1787
static struct cftype cft_memory_pressure_enabled = {
	.name = "memory_pressure_enabled",
1788 1789
	.read_u64 = cpuset_read_u64,
	.write_u64 = cpuset_write_u64,
1790 1791 1792
	.private = FILE_MEMORY_PRESSURE_ENABLED,
};

1793
static int cpuset_populate(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1794 1795 1796
{
	int err;

1797 1798
	err = cgroup_add_files(cont, ss, files, ARRAY_SIZE(files));
	if (err)
L
Linus Torvalds 已提交
1799
		return err;
1800
	/* memory_pressure_enabled is in root cpuset only */
1801
	if (!cont->parent)
1802
		err = cgroup_add_file(cont, ss,
1803 1804
				      &cft_memory_pressure_enabled);
	return err;
L
Linus Torvalds 已提交
1805 1806
}

1807
/*
D
Daniel Lezcano 已提交
1808 1809 1810
 * post_clone() is called during cgroup_create() when the
 * clone_children mount argument was specified.  The cgroup
 * can not yet have any tasks.
1811 1812 1813 1814 1815 1816 1817 1818 1819
 *
 * 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
1820 1821
 * (and likewise for mems) to the new cgroup. Called with cgroup_mutex
 * held.
1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837
 */
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);

1838
	mutex_lock(&callback_mutex);
1839
	cs->mems_allowed = parent_cs->mems_allowed;
1840
	cpumask_copy(cs->cpus_allowed, parent_cs->cpus_allowed);
1841
	mutex_unlock(&callback_mutex);
1842 1843 1844
	return;
}

L
Linus Torvalds 已提交
1845 1846
/*
 *	cpuset_create - create a cpuset
1847 1848
 *	ss:	cpuset cgroup subsystem
 *	cont:	control group that the new cpuset will be part of
L
Linus Torvalds 已提交
1849 1850
 */

1851 1852 1853
static struct cgroup_subsys_state *cpuset_create(
	struct cgroup_subsys *ss,
	struct cgroup *cont)
L
Linus Torvalds 已提交
1854 1855
{
	struct cpuset *cs;
1856
	struct cpuset *parent;
L
Linus Torvalds 已提交
1857

1858 1859 1860 1861
	if (!cont->parent) {
		return &top_cpuset.css;
	}
	parent = cgroup_cs(cont->parent);
L
Linus Torvalds 已提交
1862 1863
	cs = kmalloc(sizeof(*cs), GFP_KERNEL);
	if (!cs)
1864
		return ERR_PTR(-ENOMEM);
1865 1866 1867 1868
	if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) {
		kfree(cs);
		return ERR_PTR(-ENOMEM);
	}
L
Linus Torvalds 已提交
1869 1870

	cs->flags = 0;
1871 1872 1873 1874
	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
1875
	set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
1876
	cpumask_clear(cs->cpus_allowed);
1877
	nodes_clear(cs->mems_allowed);
1878
	fmeter_init(&cs->fmeter);
1879
	cs->relax_domain_level = -1;
L
Linus Torvalds 已提交
1880 1881

	cs->parent = parent;
1882
	number_of_cpusets++;
1883
	return &cs->css ;
L
Linus Torvalds 已提交
1884 1885
}

P
Paul Jackson 已提交
1886 1887 1888
/*
 * If the cpuset being removed has its flag 'sched_load_balance'
 * enabled, then simulate turning sched_load_balance off, which
1889
 * will call async_rebuild_sched_domains().
P
Paul Jackson 已提交
1890 1891
 */

1892
static void cpuset_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1893
{
1894
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1895

P
Paul Jackson 已提交
1896
	if (is_sched_load_balance(cs))
1897
		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
P
Paul Jackson 已提交
1898

1899
	number_of_cpusets--;
1900
	free_cpumask_var(cs->cpus_allowed);
1901
	kfree(cs);
L
Linus Torvalds 已提交
1902 1903
}

1904 1905 1906
struct cgroup_subsys cpuset_subsys = {
	.name = "cpuset",
	.create = cpuset_create,
1907
	.destroy = cpuset_destroy,
1908
	.can_attach = cpuset_can_attach,
1909 1910 1911
	.can_attach_task = cpuset_can_attach_task,
	.pre_attach = cpuset_pre_attach,
	.attach_task = cpuset_attach_task,
1912 1913 1914 1915 1916 1917 1918
	.attach = cpuset_attach,
	.populate = cpuset_populate,
	.post_clone = cpuset_post_clone,
	.subsys_id = cpuset_subsys_id,
	.early_init = 1,
};

L
Linus Torvalds 已提交
1919 1920 1921 1922 1923 1924 1925 1926
/**
 * cpuset_init - initialize cpusets at system boot
 *
 * Description: Initialize top_cpuset and the cpuset internal file system,
 **/

int __init cpuset_init(void)
{
1927
	int err = 0;
L
Linus Torvalds 已提交
1928

1929 1930 1931
	if (!alloc_cpumask_var(&top_cpuset.cpus_allowed, GFP_KERNEL))
		BUG();

1932
	cpumask_setall(top_cpuset.cpus_allowed);
1933
	nodes_setall(top_cpuset.mems_allowed);
L
Linus Torvalds 已提交
1934

1935
	fmeter_init(&top_cpuset.fmeter);
P
Paul Jackson 已提交
1936
	set_bit(CS_SCHED_LOAD_BALANCE, &top_cpuset.flags);
1937
	top_cpuset.relax_domain_level = -1;
L
Linus Torvalds 已提交
1938 1939 1940

	err = register_filesystem(&cpuset_fs_type);
	if (err < 0)
1941 1942
		return err;

1943 1944 1945
	if (!alloc_cpumask_var(&cpus_attach, GFP_KERNEL))
		BUG();

1946
	number_of_cpusets = 1;
1947
	return 0;
L
Linus Torvalds 已提交
1948 1949
}

1950 1951 1952 1953 1954 1955 1956 1957
/**
 * 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.
 */
1958 1959
static void cpuset_do_move_task(struct task_struct *tsk,
				struct cgroup_scanner *scan)
1960
{
1961
	struct cgroup *new_cgroup = scan->data;
1962

1963
	cgroup_attach_task(new_cgroup, tsk);
1964 1965 1966 1967 1968 1969 1970
}

/**
 * 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
 *
1971 1972
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1973 1974 1975 1976 1977 1978
 *
 * 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)
{
1979
	struct cgroup_scanner scan;
1980

1981 1982 1983 1984 1985
	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;
1986

1987
	if (cgroup_scan_tasks(&scan))
1988 1989 1990 1991
		printk(KERN_ERR "move_member_tasks_to_cpuset: "
				"cgroup_scan_tasks failed\n");
}

1992
/*
1993
 * If CPU and/or memory hotplug handlers, below, unplug any CPUs
1994 1995
 * or memory nodes, we need to walk over the cpuset hierarchy,
 * removing that CPU or node from all cpusets.  If this removes the
1996 1997
 * last CPU or node from a cpuset, then move the tasks in the empty
 * cpuset to its next-highest non-empty parent.
1998
 *
1999 2000
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
2001
 */
2002 2003 2004 2005
static void remove_tasks_in_empty_cpuset(struct cpuset *cs)
{
	struct cpuset *parent;

2006 2007 2008 2009 2010
	/*
	 * 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.
	 */
2011 2012
	if (list_empty(&cs->css.cgroup->css_sets))
		return;
2013

2014 2015 2016 2017 2018
	/*
	 * Find its next-highest non-empty parent, (top cpuset
	 * has online cpus, so can't be empty).
	 */
	parent = cs->parent;
2019
	while (cpumask_empty(parent->cpus_allowed) ||
2020
			nodes_empty(parent->mems_allowed))
2021 2022 2023 2024 2025 2026 2027 2028 2029
		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.
 *
2030
 * Called with cgroup_mutex held.  We take callback_mutex to modify
2031 2032 2033 2034 2035 2036 2037 2038 2039 2040
 * 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.
 */
2041
static void scan_for_empty_cpusets(struct cpuset *root)
2042
{
2043
	LIST_HEAD(queue);
2044 2045
	struct cpuset *cp;	/* scans cpusets being updated */
	struct cpuset *child;	/* scans child cpusets of cp */
2046
	struct cgroup *cont;
2047
	static nodemask_t oldmems;	/* protected by cgroup_mutex */
2048

2049 2050 2051
	list_add_tail((struct list_head *)&root->stack_list, &queue);

	while (!list_empty(&queue)) {
2052
		cp = list_first_entry(&queue, struct cpuset, stack_list);
2053 2054 2055 2056 2057
		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);
		}
2058 2059

		/* Continue past cpusets with all cpus, mems online */
2060
		if (cpumask_subset(cp->cpus_allowed, cpu_active_mask) &&
2061 2062 2063
		    nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY]))
			continue;

2064
		oldmems = cp->mems_allowed;
2065

2066
		/* Remove offline cpus and mems from this cpuset. */
2067
		mutex_lock(&callback_mutex);
2068
		cpumask_and(cp->cpus_allowed, cp->cpus_allowed,
2069
			    cpu_active_mask);
2070 2071
		nodes_and(cp->mems_allowed, cp->mems_allowed,
						node_states[N_HIGH_MEMORY]);
2072 2073 2074
		mutex_unlock(&callback_mutex);

		/* Move tasks from the empty cpuset to a parent */
2075
		if (cpumask_empty(cp->cpus_allowed) ||
2076
		     nodes_empty(cp->mems_allowed))
2077
			remove_tasks_in_empty_cpuset(cp);
2078
		else {
2079
			update_tasks_cpumask(cp, NULL);
2080
			update_tasks_nodemask(cp, &oldmems, NULL);
2081
		}
2082 2083 2084
	}
}

2085 2086 2087 2088 2089 2090
/*
 * 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.
 *
2091
 * This routine ensures that top_cpuset.cpus_allowed tracks
2092
 * cpu_active_mask on each CPU hotplug (cpuhp) event.
2093 2094 2095
 *
 * Called within get_online_cpus().  Needs to call cgroup_lock()
 * before calling generate_sched_domains().
2096
 */
2097
void cpuset_update_active_cpus(void)
2098
{
2099
	struct sched_domain_attr *attr;
2100
	cpumask_var_t *doms;
2101 2102 2103
	int ndoms;

	cgroup_lock();
2104
	mutex_lock(&callback_mutex);
2105
	cpumask_copy(top_cpuset.cpus_allowed, cpu_active_mask);
2106
	mutex_unlock(&callback_mutex);
2107 2108 2109 2110 2111 2112
	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);
2113 2114
}

2115
#ifdef CONFIG_MEMORY_HOTPLUG
2116
/*
2117
 * Keep top_cpuset.mems_allowed tracking node_states[N_HIGH_MEMORY].
2118 2119
 * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
 * See also the previous routine cpuset_track_online_cpus().
2120
 */
2121 2122
static int cpuset_track_online_nodes(struct notifier_block *self,
				unsigned long action, void *arg)
2123
{
2124
	static nodemask_t oldmems;	/* protected by cgroup_mutex */
2125

2126
	cgroup_lock();
2127 2128
	switch (action) {
	case MEM_ONLINE:
2129
		oldmems = top_cpuset.mems_allowed;
2130
		mutex_lock(&callback_mutex);
2131
		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2132
		mutex_unlock(&callback_mutex);
2133
		update_tasks_nodemask(&top_cpuset, &oldmems, NULL);
2134 2135 2136 2137 2138 2139 2140
		break;
	case MEM_OFFLINE:
		/*
		 * needn't update top_cpuset.mems_allowed explicitly because
		 * scan_for_empty_cpusets() will update it.
		 */
		scan_for_empty_cpusets(&top_cpuset);
2141 2142 2143 2144
		break;
	default:
		break;
	}
2145
	cgroup_unlock();
2146

2147
	return NOTIFY_OK;
2148 2149 2150
}
#endif

L
Linus Torvalds 已提交
2151 2152 2153 2154 2155 2156 2157 2158
/**
 * cpuset_init_smp - initialize cpus_allowed
 *
 * Description: Finish top cpuset after cpu, node maps are initialized
 **/

void __init cpuset_init_smp(void)
{
2159
	cpumask_copy(top_cpuset.cpus_allowed, cpu_active_mask);
2160
	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2161

2162
	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
2163 2164 2165

	cpuset_wq = create_singlethread_workqueue("cpuset");
	BUG_ON(!cpuset_wq);
L
Linus Torvalds 已提交
2166 2167 2168 2169 2170
}

/**
 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
2171
 * @pmask: pointer to struct cpumask variable to receive cpus_allowed set.
L
Linus Torvalds 已提交
2172
 *
2173
 * Description: Returns the cpumask_var_t cpus_allowed of the cpuset
L
Linus Torvalds 已提交
2174 2175 2176 2177 2178
 * 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.
 **/

2179
void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask)
L
Linus Torvalds 已提交
2180
{
2181
	mutex_lock(&callback_mutex);
2182
	task_lock(tsk);
2183
	guarantee_online_cpus(task_cs(tsk), pmask);
2184
	task_unlock(tsk);
2185
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
2186 2187
}

2188 2189 2190 2191 2192 2193 2194 2195
int cpuset_cpus_allowed_fallback(struct task_struct *tsk)
{
	const struct cpuset *cs;
	int cpu;

	rcu_read_lock();
	cs = task_cs(tsk);
	if (cs)
2196
		do_set_cpus_allowed(tsk, cs->cpus_allowed);
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222
	rcu_read_unlock();

	/*
	 * We own tsk->cpus_allowed, nobody can change it under us.
	 *
	 * But we used cs && cs->cpus_allowed lockless and thus can
	 * race with cgroup_attach_task() or update_cpumask() and get
	 * the wrong tsk->cpus_allowed. However, both cases imply the
	 * subsequent cpuset_change_cpumask()->set_cpus_allowed_ptr()
	 * which takes task_rq_lock().
	 *
	 * If we are called after it dropped the lock we must see all
	 * changes in tsk_cs()->cpus_allowed. Otherwise we can temporary
	 * set any mask even if it is not right from task_cs() pov,
	 * the pending set_cpus_allowed_ptr() will fix things.
	 */

	cpu = cpumask_any_and(&tsk->cpus_allowed, cpu_active_mask);
	if (cpu >= nr_cpu_ids) {
		/*
		 * Either tsk->cpus_allowed is wrong (see above) or it
		 * is actually empty. The latter case is only possible
		 * if we are racing with remove_tasks_in_empty_cpuset().
		 * Like above we can temporary set any mask and rely on
		 * set_cpus_allowed_ptr() as synchronization point.
		 */
2223
		do_set_cpus_allowed(tsk, cpu_possible_mask);
2224 2225 2226 2227 2228 2229
		cpu = cpumask_any(cpu_active_mask);
	}

	return cpu;
}

L
Linus Torvalds 已提交
2230 2231
void cpuset_init_current_mems_allowed(void)
{
2232
	nodes_setall(current->mems_allowed);
L
Linus Torvalds 已提交
2233 2234
}

2235 2236 2237 2238 2239 2240
/**
 * 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
2241
 * subset of node_states[N_HIGH_MEMORY], even if this means going outside the
2242 2243 2244 2245 2246 2247 2248
 * tasks cpuset.
 **/

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

2249
	mutex_lock(&callback_mutex);
2250
	task_lock(tsk);
2251
	guarantee_online_mems(task_cs(tsk), &mask);
2252
	task_unlock(tsk);
2253
	mutex_unlock(&callback_mutex);
2254 2255 2256 2257

	return mask;
}

2258
/**
2259 2260
 * cpuset_nodemask_valid_mems_allowed - check nodemask vs. curremt mems_allowed
 * @nodemask: the nodemask to be checked
2261
 *
2262
 * Are any of the nodes in the nodemask allowed in current->mems_allowed?
L
Linus Torvalds 已提交
2263
 */
2264
int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask)
L
Linus Torvalds 已提交
2265
{
2266
	return nodes_intersects(*nodemask, current->mems_allowed);
L
Linus Torvalds 已提交
2267 2268
}

2269
/*
2270 2271 2272 2273
 * 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.
2274
 */
2275
static const struct cpuset *nearest_hardwall_ancestor(const struct cpuset *cs)
2276
{
2277
	while (!(is_mem_exclusive(cs) || is_mem_hardwall(cs)) && cs->parent)
2278 2279 2280 2281
		cs = cs->parent;
	return cs;
}

2282
/**
2283 2284
 * cpuset_node_allowed_softwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2285
 * @gfp_mask: memory allocation flags
2286
 *
2287 2288 2289 2290 2291 2292
 * 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.
2293 2294
 * Otherwise, no.
 *
2295 2296 2297
 * 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.
2298
 *
2299 2300
 * cpuset_node_allowed_hardwall() only handles the simpler case of hardwall
 * cpusets, and never sleeps.
2301 2302 2303 2304 2305 2306 2307
 *
 * 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'.
 *
2308
 * GFP_USER allocations are marked with the __GFP_HARDWALL bit,
2309 2310
 * and do not allow allocations outside the current tasks cpuset
 * unless the task has been OOM killed as is marked TIF_MEMDIE.
2311
 * GFP_KERNEL allocations are not so marked, so can escape to the
2312
 * nearest enclosing hardwalled ancestor cpuset.
2313
 *
2314 2315 2316 2317 2318 2319 2320
 * 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.
2321
 *
2322
 * The first call here from mm/page_alloc:get_page_from_freelist()
2323 2324 2325
 * 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).
2326 2327 2328 2329 2330 2331
 *
 * 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:
2332 2333
 *	in_interrupt - any node ok (current task context irrelevant)
 *	GFP_ATOMIC   - any node ok
2334
 *	TIF_MEMDIE   - any node ok
2335
 *	GFP_KERNEL   - any node in enclosing hardwalled cpuset ok
2336
 *	GFP_USER     - only nodes in current tasks mems allowed ok.
2337 2338
 *
 * Rule:
2339
 *    Don't call cpuset_node_allowed_softwall if you can't sleep, unless you
2340 2341
 *    pass in the __GFP_HARDWALL flag set in gfp_flag, which disables
 *    the code that might scan up ancestor cpusets and sleep.
2342
 */
2343
int __cpuset_node_allowed_softwall(int node, gfp_t gfp_mask)
L
Linus Torvalds 已提交
2344
{
2345
	const struct cpuset *cs;	/* current cpuset ancestors */
2346
	int allowed;			/* is allocation in zone z allowed? */
2347

2348
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
2349
		return 1;
2350
	might_sleep_if(!(gfp_mask & __GFP_HARDWALL));
2351 2352
	if (node_isset(node, current->mems_allowed))
		return 1;
2353 2354 2355 2356 2357 2358
	/*
	 * 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;
2359 2360 2361
	if (gfp_mask & __GFP_HARDWALL)	/* If hardwall request, stop here */
		return 0;

2362 2363 2364
	if (current->flags & PF_EXITING) /* Let dying task have memory */
		return 1;

2365
	/* Not hardwall and node outside mems_allowed: scan up cpusets */
2366
	mutex_lock(&callback_mutex);
2367 2368

	task_lock(current);
2369
	cs = nearest_hardwall_ancestor(task_cs(current));
2370 2371
	task_unlock(current);

2372
	allowed = node_isset(node, cs->mems_allowed);
2373
	mutex_unlock(&callback_mutex);
2374
	return allowed;
L
Linus Torvalds 已提交
2375 2376
}

2377
/*
2378 2379
 * cpuset_node_allowed_hardwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2380 2381
 * @gfp_mask: memory allocation flags
 *
2382 2383 2384 2385 2386
 * 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.
2387 2388 2389 2390 2391 2392 2393
 *
 * 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'.
 *
2394 2395
 * Unlike the cpuset_node_allowed_softwall() variant, above,
 * this variant requires that the node be in the current task's
2396 2397 2398 2399
 * 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.
 */
2400
int __cpuset_node_allowed_hardwall(int node, gfp_t gfp_mask)
2401 2402 2403 2404 2405
{
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
		return 1;
	if (node_isset(node, current->mems_allowed))
		return 1;
D
Daniel Walker 已提交
2406 2407 2408 2409 2410 2411
	/*
	 * 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;
2412 2413 2414
	return 0;
}

P
Paul Jackson 已提交
2415 2416 2417 2418 2419 2420 2421 2422
/**
 * cpuset_unlock - release lock on cpuset changes
 *
 * Undo the lock taken in a previous cpuset_lock() call.
 */

void cpuset_unlock(void)
{
2423
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
2424 2425
}

2426
/**
2427 2428
 * cpuset_mem_spread_node() - On which node to begin search for a file page
 * cpuset_slab_spread_node() - On which node to begin search for a slab page
2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452
 *
 * 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().
 */

2453
static int cpuset_spread_node(int *rotor)
2454 2455 2456
{
	int node;

2457
	node = next_node(*rotor, current->mems_allowed);
2458 2459
	if (node == MAX_NUMNODES)
		node = first_node(current->mems_allowed);
2460
	*rotor = node;
2461 2462
	return node;
}
2463 2464 2465

int cpuset_mem_spread_node(void)
{
2466 2467 2468 2469
	if (current->cpuset_mem_spread_rotor == NUMA_NO_NODE)
		current->cpuset_mem_spread_rotor =
			node_random(&current->mems_allowed);

2470 2471 2472 2473 2474
	return cpuset_spread_node(&current->cpuset_mem_spread_rotor);
}

int cpuset_slab_spread_node(void)
{
2475 2476 2477 2478
	if (current->cpuset_slab_spread_rotor == NUMA_NO_NODE)
		current->cpuset_slab_spread_rotor =
			node_random(&current->mems_allowed);

2479 2480 2481
	return cpuset_spread_node(&current->cpuset_slab_spread_rotor);
}

2482 2483
EXPORT_SYMBOL_GPL(cpuset_mem_spread_node);

2484
/**
2485 2486 2487 2488 2489 2490 2491 2492
 * 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.
2493 2494
 **/

2495 2496
int cpuset_mems_allowed_intersects(const struct task_struct *tsk1,
				   const struct task_struct *tsk2)
2497
{
2498
	return nodes_intersects(tsk1->mems_allowed, tsk2->mems_allowed);
2499 2500
}

2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
/**
 * 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);
}

2524 2525 2526 2527 2528 2529
/*
 * 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.
 */

2530
int cpuset_memory_pressure_enabled __read_mostly;
2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552

/**
 * 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);
2553
	fmeter_markevent(&task_cs(current)->fmeter);
2554 2555 2556
	task_unlock(current);
}

2557
#ifdef CONFIG_PROC_PID_CPUSET
L
Linus Torvalds 已提交
2558 2559 2560 2561
/*
 * proc_cpuset_show()
 *  - Print tasks cpuset path into seq_file.
 *  - Used for /proc/<pid>/cpuset.
2562 2563
 *  - 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,
2564
 *    and we take cgroup_mutex, keeping cpuset_attach() from changing it
2565
 *    anyway.
L
Linus Torvalds 已提交
2566
 */
P
Paul Jackson 已提交
2567
static int proc_cpuset_show(struct seq_file *m, void *unused_v)
L
Linus Torvalds 已提交
2568
{
2569
	struct pid *pid;
L
Linus Torvalds 已提交
2570 2571
	struct task_struct *tsk;
	char *buf;
2572
	struct cgroup_subsys_state *css;
2573
	int retval;
L
Linus Torvalds 已提交
2574

2575
	retval = -ENOMEM;
L
Linus Torvalds 已提交
2576 2577
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
2578 2579 2580
		goto out;

	retval = -ESRCH;
2581 2582
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
2583 2584
	if (!tsk)
		goto out_free;
L
Linus Torvalds 已提交
2585

2586
	retval = -EINVAL;
2587 2588 2589
	cgroup_lock();
	css = task_subsys_state(tsk, cpuset_subsys_id);
	retval = cgroup_path(css->cgroup, buf, PAGE_SIZE);
L
Linus Torvalds 已提交
2590
	if (retval < 0)
2591
		goto out_unlock;
L
Linus Torvalds 已提交
2592 2593
	seq_puts(m, buf);
	seq_putc(m, '\n');
2594
out_unlock:
2595
	cgroup_unlock();
2596 2597
	put_task_struct(tsk);
out_free:
L
Linus Torvalds 已提交
2598
	kfree(buf);
2599
out:
L
Linus Torvalds 已提交
2600 2601 2602 2603 2604
	return retval;
}

static int cpuset_open(struct inode *inode, struct file *file)
{
2605 2606
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cpuset_show, pid);
L
Linus Torvalds 已提交
2607 2608
}

2609
const struct file_operations proc_cpuset_operations = {
L
Linus Torvalds 已提交
2610 2611 2612 2613 2614
	.open		= cpuset_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};
2615
#endif /* CONFIG_PROC_PID_CPUSET */
L
Linus Torvalds 已提交
2616

2617
/* Display task mems_allowed in /proc/<pid>/status file. */
2618 2619 2620
void cpuset_task_status_allowed(struct seq_file *m, struct task_struct *task)
{
	seq_printf(m, "Mems_allowed:\t");
2621
	seq_nodemask(m, &task->mems_allowed);
2622
	seq_printf(m, "\n");
2623
	seq_printf(m, "Mems_allowed_list:\t");
2624
	seq_nodemask_list(m, &task->mems_allowed);
2625
	seq_printf(m, "\n");
L
Linus Torvalds 已提交
2626
}