cpuset.c 71.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6
/*
 *  kernel/cpuset.c
 *
 *  Processor and Memory placement constraints for sets of tasks.
 *
 *  Copyright (C) 2003 BULL SA.
P
Paul Jackson 已提交
7
 *  Copyright (C) 2004-2007 Silicon Graphics, Inc.
8
 *  Copyright (C) 2006 Google, Inc
L
Linus Torvalds 已提交
9 10 11 12
 *
 *  Portions derived from Patrick Mochel's sysfs code.
 *  sysfs is Copyright (c) 2001-3 Patrick Mochel
 *
13
 *  2003-10-10 Written by Simon Derr.
L
Linus Torvalds 已提交
14
 *  2003-10-22 Updates by Stephen Hemminger.
15
 *  2004 May-July Rework by Paul Jackson.
16
 *  2006 Rework by Paul Menage to use generic cgroups
17 18
 *  2008 Rework of the scheduler domains and CPU hotplug handling
 *       by Max Krasnyansky
L
Linus Torvalds 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
 *
 *  This file is subject to the terms and conditions of the GNU General Public
 *  License.  See the file COPYING in the main directory of the Linux
 *  distribution for more details.
 */

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

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

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

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

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

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

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

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

L
Linus Torvalds 已提交
94
	unsigned long flags;		/* "unsigned long" so bitops work */
95
	cpumask_var_t cpus_allowed;	/* CPUs allowed to tasks in cpuset */
L
Linus Torvalds 已提交
96 97 98 99
	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;

108 109
	/* used for walking a cpuset heirarchy */
	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
 */
234 235 236
static int cpuset_get_sb(struct file_system_type *fs_type,
			 int flags, const char *unused_dev_name,
			 void *data, struct vfsmount *mnt)
L
Linus Torvalds 已提交
237
{
238 239 240 241 242 243 244 245 246 247 248
	struct file_system_type *cgroup_fs = get_fs_type("cgroup");
	int ret = -ENODEV;
	if (cgroup_fs) {
		char mountopts[] =
			"cpuset,noprefix,"
			"release_agent=/sbin/cpuset_release_agent";
		ret = cgroup_fs->get_sb(cgroup_fs, flags,
					   unused_dev_name, mountopts, mnt);
		put_filesystem(cgroup_fs);
	}
	return ret;
L
Linus Torvalds 已提交
249 250 251 252 253 254 255 256
}

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

/*
257
 * Return in pmask the portion of a cpusets's cpus_allowed that
L
Linus Torvalds 已提交
258 259 260 261 262 263 264 265 266
 * 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.
 *
267
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
268 269
 */

270 271
static void guarantee_online_cpus(const struct cpuset *cs,
				  struct cpumask *pmask)
L
Linus Torvalds 已提交
272
{
273
	while (cs && !cpumask_intersects(cs->cpus_allowed, cpu_online_mask))
L
Linus Torvalds 已提交
274 275
		cs = cs->parent;
	if (cs)
276
		cpumask_and(pmask, cs->cpus_allowed, cpu_online_mask);
L
Linus Torvalds 已提交
277
	else
278 279
		cpumask_copy(pmask, cpu_online_mask);
	BUG_ON(!cpumask_intersects(pmask, cpu_online_mask));
L
Linus Torvalds 已提交
280 281 282 283
}

/*
 * Return in *pmask the portion of a cpusets's mems_allowed that
284 285 286 287
 * 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
288 289
 *
 * One way or another, we guarantee to return some non-empty subset
290
 * of node_states[N_HIGH_MEMORY].
L
Linus Torvalds 已提交
291
 *
292
 * Call with callback_mutex held.
L
Linus Torvalds 已提交
293 294 295 296
 */

static void guarantee_online_mems(const struct cpuset *cs, nodemask_t *pmask)
{
297 298
	while (cs && !nodes_intersects(cs->mems_allowed,
					node_states[N_HIGH_MEMORY]))
L
Linus Torvalds 已提交
299 300
		cs = cs->parent;
	if (cs)
301 302
		nodes_and(*pmask, cs->mems_allowed,
					node_states[N_HIGH_MEMORY]);
L
Linus Torvalds 已提交
303
	else
304 305
		*pmask = node_states[N_HIGH_MEMORY];
	BUG_ON(!nodes_intersects(*pmask, node_states[N_HIGH_MEMORY]));
L
Linus Torvalds 已提交
306 307
}

308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
/*
 * 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
326 327 328 329 330
/*
 * 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
331
 * are only set if the other's are set.  Call holding cgroup_mutex.
L
Linus Torvalds 已提交
332 333 334 335
 */

static int is_cpuset_subset(const struct cpuset *p, const struct cpuset *q)
{
336
	return	cpumask_subset(p->cpus_allowed, q->cpus_allowed) &&
L
Linus Torvalds 已提交
337 338 339 340 341
		nodes_subset(p->mems_allowed, q->mems_allowed) &&
		is_cpu_exclusive(p) <= is_cpu_exclusive(q) &&
		is_mem_exclusive(p) <= is_mem_exclusive(q);
}

342 343 344 345 346 347
/**
 * 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)
{
348 349 350 351 352 353 354 355 356 357 358 359 360
	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;
361 362 363 364 365 366 367 368
}

/**
 * free_trial_cpuset - free the trial cpuset
 * @trial: the trial cpuset to be freed
 */
static void free_trial_cpuset(struct cpuset *trial)
{
369
	free_cpumask_var(trial->cpus_allowed);
370 371 372
	kfree(trial);
}

L
Linus Torvalds 已提交
373 374 375 376 377 378 379
/*
 * 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
380
 * cgroup_mutex held.
L
Linus Torvalds 已提交
381 382 383 384 385 386 387 388 389 390 391 392 393 394
 *
 * '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)
{
395
	struct cgroup *cont;
L
Linus Torvalds 已提交
396 397 398
	struct cpuset *c, *par;

	/* Each of our child cpusets must be a subset of us */
399 400
	list_for_each_entry(cont, &cur->css.cgroup->children, sibling) {
		if (!is_cpuset_subset(cgroup_cs(cont), trial))
L
Linus Torvalds 已提交
401 402 403 404
			return -EBUSY;
	}

	/* Remaining checks don't apply to root cpuset */
405
	if (cur == &top_cpuset)
L
Linus Torvalds 已提交
406 407
		return 0;

408 409
	par = cur->parent;

L
Linus Torvalds 已提交
410 411 412 413
	/* We must be a subset of our parent cpuset */
	if (!is_cpuset_subset(trial, par))
		return -EACCES;

414 415 416 417
	/*
	 * If either I or some sibling (!= me) is exclusive, we can't
	 * overlap
	 */
418 419
	list_for_each_entry(cont, &par->css.cgroup->children, sibling) {
		c = cgroup_cs(cont);
L
Linus Torvalds 已提交
420 421
		if ((is_cpu_exclusive(trial) || is_cpu_exclusive(c)) &&
		    c != cur &&
422
		    cpumask_intersects(trial->cpus_allowed, c->cpus_allowed))
L
Linus Torvalds 已提交
423 424 425 426 427 428 429
			return -EINVAL;
		if ((is_mem_exclusive(trial) || is_mem_exclusive(c)) &&
		    c != cur &&
		    nodes_intersects(trial->mems_allowed, c->mems_allowed))
			return -EINVAL;
	}

430 431
	/* Cpusets with tasks can't have empty cpus_allowed or mems_allowed */
	if (cgroup_task_count(cur->css.cgroup)) {
432
		if (cpumask_empty(trial->cpus_allowed) ||
433 434 435 436 437
		    nodes_empty(trial->mems_allowed)) {
			return -ENOSPC;
		}
	}

L
Linus Torvalds 已提交
438 439 440
	return 0;
}

441
#ifdef CONFIG_SMP
P
Paul Jackson 已提交
442
/*
443
 * Helper routine for generate_sched_domains().
P
Paul Jackson 已提交
444 445 446 447
 * Do cpusets a, b have overlapping cpus_allowed masks?
 */
static int cpusets_overlap(struct cpuset *a, struct cpuset *b)
{
448
	return cpumask_intersects(a->cpus_allowed, b->cpus_allowed);
P
Paul Jackson 已提交
449 450
}

451 452 453 454 455 456 457 458
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;
}

459 460 461 462 463 464 465 466 467 468 469 470 471 472
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);

473
		if (cpumask_empty(cp->cpus_allowed))
474 475 476 477 478 479 480 481 482 483 484 485
			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
486
/*
487 488 489 490 491 492 493 494 495
 * 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
496
 *
L
Li Zefan 已提交
497
 * See "What is sched_load_balance" in Documentation/cgroups/cpusets.txt
P
Paul Jackson 已提交
498 499 500 501 502 503 504
 * 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.
 *
505
 * Must be called with cgroup_lock held.
P
Paul Jackson 已提交
506 507
 *
 * The three key local variables below are:
508
 *    q  - a linked-list queue of cpuset pointers, used to implement a
P
Paul Jackson 已提交
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
 *	   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().
 */
540
static int generate_sched_domains(cpumask_var_t **domains,
541
			struct sched_domain_attr **attributes)
P
Paul Jackson 已提交
542
{
543
	LIST_HEAD(q);		/* queue of cpusets to be scanned */
P
Paul Jackson 已提交
544 545 546 547
	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 */
548
	cpumask_var_t *doms;	/* resulting partition; i.e. sched domains */
549
	struct sched_domain_attr *dattr;  /* attributes for custom domains */
550
	int ndoms = 0;		/* number of sched domains in result */
551
	int nslot;		/* next empty doms[] struct cpumask slot */
P
Paul Jackson 已提交
552 553

	doms = NULL;
554
	dattr = NULL;
555
	csa = NULL;
P
Paul Jackson 已提交
556 557 558

	/* Special case for the 99% of systems with one, full, sched domain */
	if (is_sched_load_balance(&top_cpuset)) {
559 560
		ndoms = 1;
		doms = alloc_sched_domains(ndoms);
P
Paul Jackson 已提交
561
		if (!doms)
562 563
			goto done;

564 565 566
		dattr = kmalloc(sizeof(struct sched_domain_attr), GFP_KERNEL);
		if (dattr) {
			*dattr = SD_ATTR_INIT;
567
			update_domain_attr_tree(dattr, &top_cpuset);
568
		}
569
		cpumask_copy(doms[0], top_cpuset.cpus_allowed);
570 571

		goto done;
P
Paul Jackson 已提交
572 573 574 575 576 577 578
	}

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

579 580
	list_add(&top_cpuset.stack_list, &q);
	while (!list_empty(&q)) {
P
Paul Jackson 已提交
581 582
		struct cgroup *cont;
		struct cpuset *child;   /* scans child cpusets of cp */
583

584 585 586
		cp = list_first_entry(&q, struct cpuset, stack_list);
		list_del(q.next);

587
		if (cpumask_empty(cp->cpus_allowed))
588 589
			continue;

590 591 592 593 594 595 596
		/*
		 * 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
597
			csa[csn++] = cp;
598 599
			continue;
		}
600

P
Paul Jackson 已提交
601 602
		list_for_each_entry(cont, &cp->css.cgroup->children, sibling) {
			child = cgroup_cs(cont);
603
			list_add_tail(&child->stack_list, &q);
P
Paul Jackson 已提交
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
		}
  	}

	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;
			}
		}
	}

634 635 636 637
	/*
	 * Now we know how many domains to create.
	 * Convert <csn, csa> to <ndoms, doms> and populate cpu masks.
	 */
638
	doms = alloc_sched_domains(ndoms);
639
	if (!doms)
640 641 642 643 644 645
		goto done;

	/*
	 * The rest of the code, including the scheduler, can deal with
	 * dattr==NULL case. No need to abort if alloc fails.
	 */
646
	dattr = kmalloc(ndoms * sizeof(struct sched_domain_attr), GFP_KERNEL);
P
Paul Jackson 已提交
647 648 649

	for (nslot = 0, i = 0; i < csn; i++) {
		struct cpuset *a = csa[i];
650
		struct cpumask *dp;
P
Paul Jackson 已提交
651 652
		int apn = a->pn;

653 654 655 656 657
		if (apn < 0) {
			/* Skip completed partitions */
			continue;
		}

658
		dp = doms[nslot];
659 660 661 662 663 664 665 666 667 668

		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
669
			}
670 671
			continue;
		}
P
Paul Jackson 已提交
672

673
		cpumask_clear(dp);
674 675 676 677 678 679
		if (dattr)
			*(dattr + nslot) = SD_ATTR_INIT;
		for (j = i; j < csn; j++) {
			struct cpuset *b = csa[j];

			if (apn == b->pn) {
680
				cpumask_or(dp, dp, b->cpus_allowed);
681 682 683 684 685
				if (dattr)
					update_domain_attr_tree(dattr + nslot, b);

				/* Done with this partition */
				b->pn = -1;
P
Paul Jackson 已提交
686 687
			}
		}
688
		nslot++;
P
Paul Jackson 已提交
689 690 691
	}
	BUG_ON(nslot != ndoms);

692 693 694
done:
	kfree(csa);

695 696 697 698 699 700 701
	/*
	 * Fallback to the default domain if kmalloc() failed.
	 * See comments in partition_sched_domains().
	 */
	if (doms == NULL)
		ndoms = 1;

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719
	*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;
720
	cpumask_var_t *doms;
721 722
	int ndoms;

723
	get_online_cpus();
724 725 726 727 728 729 730 731 732

	/* 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);

733
	put_online_cpus();
734
}
735 736 737 738 739
#else /* !CONFIG_SMP */
static void do_rebuild_sched_domains(struct work_struct *unused)
{
}

740
static int generate_sched_domains(cpumask_var_t **domains,
741 742 743 744 745 746
			struct sched_domain_attr **attributes)
{
	*domains = NULL;
	return 1;
}
#endif /* CONFIG_SMP */
P
Paul Jackson 已提交
747

748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770
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)
{
771
	queue_work(cpuset_wq, &rebuild_sched_domains_work);
772 773 774 775 776 777 778 779 780 781 782 783 784 785
}

/*
 * 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
786 787
}

C
Cliff Wickman 已提交
788 789 790 791 792
/**
 * 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
 *
793
 * Call with cgroup_mutex held.  May take callback_mutex during call.
C
Cliff Wickman 已提交
794 795 796
 * 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).
797
 */
798 799
static int cpuset_test_cpumask(struct task_struct *tsk,
			       struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
800
{
801
	return !cpumask_equal(&tsk->cpus_allowed,
C
Cliff Wickman 已提交
802 803
			(cgroup_cs(scan->cg))->cpus_allowed);
}
804

C
Cliff Wickman 已提交
805 806 807 808 809 810 811 812 813 814 815
/**
 * 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.
 */
816 817
static void cpuset_change_cpumask(struct task_struct *tsk,
				  struct cgroup_scanner *scan)
C
Cliff Wickman 已提交
818
{
819
	set_cpus_allowed_ptr(tsk, ((cgroup_cs(scan->cg))->cpus_allowed));
C
Cliff Wickman 已提交
820 821
}

822 823 824
/**
 * 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
825
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
826 827 828 829 830 831
 *
 * Called with cgroup_mutex held
 *
 * The cgroup_scan_tasks() function will scan all the tasks in a cgroup,
 * calling callback functions for each.
 *
832 833
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
834
 */
835
static void update_tasks_cpumask(struct cpuset *cs, struct ptr_heap *heap)
836 837 838 839 840 841
{
	struct cgroup_scanner scan;

	scan.cg = cs->css.cgroup;
	scan.test_task = cpuset_test_cpumask;
	scan.process_task = cpuset_change_cpumask;
842 843
	scan.heap = heap;
	cgroup_scan_tasks(&scan);
844 845
}

C
Cliff Wickman 已提交
846 847 848 849 850
/**
 * 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
 */
851 852
static int update_cpumask(struct cpuset *cs, struct cpuset *trialcs,
			  const char *buf)
L
Linus Torvalds 已提交
853
{
854
	struct ptr_heap heap;
C
Cliff Wickman 已提交
855 856
	int retval;
	int is_load_balanced;
L
Linus Torvalds 已提交
857

858 859 860 861
	/* top_cpuset.cpus_allowed tracks cpu_online_map; it's read-only */
	if (cs == &top_cpuset)
		return -EACCES;

862
	/*
863
	 * An empty cpus_allowed is ok only if the cpuset has no tasks.
864 865 866
	 * Since cpulist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have cpus.
867
	 */
868
	if (!*buf) {
869
		cpumask_clear(trialcs->cpus_allowed);
870
	} else {
871
		retval = cpulist_parse(buf, trialcs->cpus_allowed);
872 873
		if (retval < 0)
			return retval;
874

875
		if (!cpumask_subset(trialcs->cpus_allowed, cpu_active_mask))
876
			return -EINVAL;
877
	}
878
	retval = validate_change(cs, trialcs);
879 880
	if (retval < 0)
		return retval;
P
Paul Jackson 已提交
881

P
Paul Menage 已提交
882
	/* Nothing to do if the cpus didn't change */
883
	if (cpumask_equal(cs->cpus_allowed, trialcs->cpus_allowed))
P
Paul Menage 已提交
884
		return 0;
C
Cliff Wickman 已提交
885

886 887 888 889
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval)
		return retval;

890
	is_load_balanced = is_sched_load_balance(trialcs);
P
Paul Jackson 已提交
891

892
	mutex_lock(&callback_mutex);
893
	cpumask_copy(cs->cpus_allowed, trialcs->cpus_allowed);
894
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
895

P
Paul Menage 已提交
896 897
	/*
	 * Scan tasks in the cpuset, and update the cpumasks of any
C
Cliff Wickman 已提交
898
	 * that need an update.
P
Paul Menage 已提交
899
	 */
900 901 902
	update_tasks_cpumask(cs, &heap);

	heap_free(&heap);
C
Cliff Wickman 已提交
903

P
Paul Menage 已提交
904
	if (is_load_balanced)
905
		async_rebuild_sched_domains();
906
	return 0;
L
Linus Torvalds 已提交
907 908
}

909 910 911 912 913 914 915 916
/*
 * 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.
 *
917
 *    Call holding cgroup_mutex, so current's cpuset won't change
918
 *    during this call, as manage_mutex holds off any cpuset_attach()
919 920
 *    calls.  Therefore we don't need to take task_lock around the
 *    call to guarantee_online_mems(), as we know no one is changing
921
 *    our task's cpuset.
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937
 *
 *    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);

938
	guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
939 940
}

941
/*
942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964
 * 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.
 *
 * Called with task's alloc_lock held
 */
static void cpuset_change_task_nodemask(struct task_struct *tsk,
					nodemask_t *newmems)
{
	nodes_or(tsk->mems_allowed, tsk->mems_allowed, *newmems);
	mpol_rebind_task(tsk, &tsk->mems_allowed);
	mpol_rebind_task(tsk, newmems);
	tsk->mems_allowed = *newmems;
}

/*
 * 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.
965 966 967 968 969 970 971 972
 */
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;
973 974 975 976 977 978 979 980
	nodemask_t newmems;

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

	task_lock(p);
	cpuset_change_task_nodemask(p, &newmems);
	task_unlock(p);
981 982 983 984 985 986 987 988 989 990 991 992 993

	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);
}

994 995
static void *cpuset_being_rebound;

996 997 998 999
/**
 * 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
1000
 * @heap: if NULL, defer allocating heap memory to cgroup_scan_tasks()
1001 1002
 *
 * Called with cgroup_mutex held
1003 1004
 * No return value. It's guaranteed that cgroup_scan_tasks() always returns 0
 * if @heap != NULL.
1005
 */
1006 1007
static void update_tasks_nodemask(struct cpuset *cs, const nodemask_t *oldmem,
				 struct ptr_heap *heap)
L
Linus Torvalds 已提交
1008
{
1009
	struct cgroup_scanner scan;
1010

1011
	cpuset_being_rebound = cs;		/* causes mpol_dup() rebind */
1012

1013 1014 1015
	scan.cg = cs->css.cgroup;
	scan.test_task = NULL;
	scan.process_task = cpuset_change_nodemask;
1016
	scan.heap = heap;
1017
	scan.data = (nodemask_t *)oldmem;
1018 1019

	/*
1020 1021 1022 1023 1024 1025
	 * 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.
1026
	 * It's ok if we rebind the same mm twice; mpol_rebind_mm()
1027
	 * is idempotent.  Also migrate pages in each mm to new nodes.
1028
	 */
1029
	cgroup_scan_tasks(&scan);
1030

1031
	/* We're done rebinding vmas to this cpuset's new mems_allowed. */
1032
	cpuset_being_rebound = NULL;
L
Linus Torvalds 已提交
1033 1034
}

1035 1036 1037
/*
 * Handle user request to change the 'mems' memory placement
 * of a cpuset.  Needs to validate the request, update the
1038 1039 1040 1041
 * 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.
1042 1043 1044 1045 1046 1047
 *
 * 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.
 */
1048 1049
static int update_nodemask(struct cpuset *cs, struct cpuset *trialcs,
			   const char *buf)
1050 1051 1052
{
	nodemask_t oldmem;
	int retval;
1053
	struct ptr_heap heap;
1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068

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

	/*
	 * An empty mems_allowed is ok iff there are no tasks in the cpuset.
	 * Since nodelist_parse() fails on an empty mask, we special case
	 * that parsing.  The validate_change() call ensures that cpusets
	 * with tasks have memory.
	 */
	if (!*buf) {
1069
		nodes_clear(trialcs->mems_allowed);
1070
	} else {
1071
		retval = nodelist_parse(buf, trialcs->mems_allowed);
1072 1073 1074
		if (retval < 0)
			goto done;

1075
		if (!nodes_subset(trialcs->mems_allowed,
1076 1077 1078 1079
				node_states[N_HIGH_MEMORY]))
			return -EINVAL;
	}
	oldmem = cs->mems_allowed;
1080
	if (nodes_equal(oldmem, trialcs->mems_allowed)) {
1081 1082 1083
		retval = 0;		/* Too easy - nothing to do */
		goto done;
	}
1084
	retval = validate_change(cs, trialcs);
1085 1086 1087
	if (retval < 0)
		goto done;

1088 1089 1090 1091
	retval = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (retval < 0)
		goto done;

1092
	mutex_lock(&callback_mutex);
1093
	cs->mems_allowed = trialcs->mems_allowed;
1094 1095
	mutex_unlock(&callback_mutex);

1096 1097 1098
	update_tasks_nodemask(cs, &oldmem, &heap);

	heap_free(&heap);
1099 1100 1101 1102
done:
	return retval;
}

1103 1104 1105 1106 1107
int current_cpuset_is_being_rebound(void)
{
	return task_cs(current) == cpuset_being_rebound;
}

1108
static int update_relax_domain_level(struct cpuset *cs, s64 val)
1109
{
1110
#ifdef CONFIG_SMP
1111 1112
	if (val < -1 || val >= SD_LV_MAX)
		return -EINVAL;
1113
#endif
1114 1115 1116

	if (val != cs->relax_domain_level) {
		cs->relax_domain_level = val;
1117 1118
		if (!cpumask_empty(cs->cpus_allowed) &&
		    is_sched_load_balance(cs))
1119
			async_rebuild_sched_domains();
1120 1121 1122 1123 1124
	}

	return 0;
}

1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164
/*
 * 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
1165 1166
/*
 * update_flag - read a 0 or a 1 in a file and update associated flag
1167 1168 1169
 * bit:		the bit to update (see cpuset_flagbits_t)
 * cs:		the cpuset to update
 * turning_on: 	whether the flag is being set or cleared
1170
 *
1171
 * Call with cgroup_mutex held.
L
Linus Torvalds 已提交
1172 1173
 */

1174 1175
static int update_flag(cpuset_flagbits_t bit, struct cpuset *cs,
		       int turning_on)
L
Linus Torvalds 已提交
1176
{
1177
	struct cpuset *trialcs;
R
Rakib Mullick 已提交
1178
	int balance_flag_changed;
1179 1180 1181
	int spread_flag_changed;
	struct ptr_heap heap;
	int err;
L
Linus Torvalds 已提交
1182

1183 1184 1185 1186
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

L
Linus Torvalds 已提交
1187
	if (turning_on)
1188
		set_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1189
	else
1190
		clear_bit(bit, &trialcs->flags);
L
Linus Torvalds 已提交
1191

1192
	err = validate_change(cs, trialcs);
1193
	if (err < 0)
1194
		goto out;
P
Paul Jackson 已提交
1195

1196 1197 1198 1199
	err = heap_init(&heap, PAGE_SIZE, GFP_KERNEL, NULL);
	if (err < 0)
		goto out;

P
Paul Jackson 已提交
1200
	balance_flag_changed = (is_sched_load_balance(cs) !=
1201
				is_sched_load_balance(trialcs));
P
Paul Jackson 已提交
1202

1203 1204 1205
	spread_flag_changed = ((is_spread_slab(cs) != is_spread_slab(trialcs))
			|| (is_spread_page(cs) != is_spread_page(trialcs)));

1206
	mutex_lock(&callback_mutex);
1207
	cs->flags = trialcs->flags;
1208
	mutex_unlock(&callback_mutex);
1209

1210
	if (!cpumask_empty(trialcs->cpus_allowed) && balance_flag_changed)
1211
		async_rebuild_sched_domains();
P
Paul Jackson 已提交
1212

1213 1214 1215
	if (spread_flag_changed)
		update_tasks_flags(cs, &heap);
	heap_free(&heap);
1216 1217 1218
out:
	free_trial_cpuset(trialcs);
	return err;
L
Linus Torvalds 已提交
1219 1220
}

1221
/*
A
Adrian Bunk 已提交
1222
 * Frequency meter - How fast is some event occurring?
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
 *
 * 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;
}

1319 1320 1321
/* Protected by cgroup_lock */
static cpumask_var_t cpus_attach;

1322
/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
1323 1324
static int cpuset_can_attach(struct cgroup_subsys *ss, struct cgroup *cont,
			     struct task_struct *tsk, bool threadgroup)
L
Linus Torvalds 已提交
1325
{
1326
	int ret;
1327
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1328

1329
	if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
L
Linus Torvalds 已提交
1330
		return -ENOSPC;
1331

1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
	/*
	 * 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
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 1373 1374 1375 1376 1377
	ret = security_task_setscheduler(tsk, 0, NULL);
	if (ret)
		return ret;
	if (threadgroup) {
		struct task_struct *c;

		rcu_read_lock();
		list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
			ret = security_task_setscheduler(c, 0, NULL);
			if (ret) {
				rcu_read_unlock();
				return ret;
			}
		}
		rcu_read_unlock();
	}
	return 0;
}

static void cpuset_attach_task(struct task_struct *tsk, nodemask_t *to,
			       struct cpuset *cs)
{
	int err;
	/*
	 * 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);

	task_lock(tsk);
	cpuset_change_task_nodemask(tsk, to);
	task_unlock(tsk);
	cpuset_update_task_spread_flag(cs, tsk);

1378
}
L
Linus Torvalds 已提交
1379

1380 1381 1382
static void cpuset_attach(struct cgroup_subsys *ss, struct cgroup *cont,
			  struct cgroup *oldcont, struct task_struct *tsk,
			  bool threadgroup)
1383 1384 1385 1386 1387
{
	nodemask_t from, to;
	struct mm_struct *mm;
	struct cpuset *cs = cgroup_cs(cont);
	struct cpuset *oldcs = cgroup_cs(oldcont);
1388

1389
	if (cs == &top_cpuset) {
1390
		cpumask_copy(cpus_attach, cpu_possible_mask);
1391
	} else {
1392
		guarantee_online_cpus(cs, cpus_attach);
1393
	}
1394
	guarantee_online_mems(cs, &to);
L
Linus Torvalds 已提交
1395

1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
	/* do per-task migration stuff possibly for each in the threadgroup */
	cpuset_attach_task(tsk, &to, cs);
	if (threadgroup) {
		struct task_struct *c;
		rcu_read_lock();
		list_for_each_entry_rcu(c, &tsk->thread_group, thread_group) {
			cpuset_attach_task(c, &to, cs);
		}
		rcu_read_unlock();
	}
1406

1407
	/* change mm; only needs to be done once even if threadgroup */
1408 1409
	from = oldcs->mems_allowed;
	to = cs->mems_allowed;
1410 1411 1412
	mm = get_task_mm(tsk);
	if (mm) {
		mpol_rebind_mm(mm, &to);
1413
		if (is_memory_migrate(cs))
1414
			cpuset_migrate_mm(mm, &from, &to);
1415 1416
		mmput(mm);
	}
L
Linus Torvalds 已提交
1417 1418 1419 1420 1421
}

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

typedef enum {
1422
	FILE_MEMORY_MIGRATE,
L
Linus Torvalds 已提交
1423 1424 1425 1426
	FILE_CPULIST,
	FILE_MEMLIST,
	FILE_CPU_EXCLUSIVE,
	FILE_MEM_EXCLUSIVE,
1427
	FILE_MEM_HARDWALL,
P
Paul Jackson 已提交
1428
	FILE_SCHED_LOAD_BALANCE,
1429
	FILE_SCHED_RELAX_DOMAIN_LEVEL,
1430 1431
	FILE_MEMORY_PRESSURE_ENABLED,
	FILE_MEMORY_PRESSURE,
1432 1433
	FILE_SPREAD_PAGE,
	FILE_SPREAD_SLAB,
L
Linus Torvalds 已提交
1434 1435
} cpuset_filetype_t;

1436 1437 1438 1439 1440 1441
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;

1442
	if (!cgroup_lock_live_group(cgrp))
1443 1444 1445
		return -ENODEV;

	switch (type) {
L
Linus Torvalds 已提交
1446
	case FILE_CPU_EXCLUSIVE:
1447
		retval = update_flag(CS_CPU_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1448 1449
		break;
	case FILE_MEM_EXCLUSIVE:
1450
		retval = update_flag(CS_MEM_EXCLUSIVE, cs, val);
L
Linus Torvalds 已提交
1451
		break;
1452 1453 1454
	case FILE_MEM_HARDWALL:
		retval = update_flag(CS_MEM_HARDWALL, cs, val);
		break;
P
Paul Jackson 已提交
1455
	case FILE_SCHED_LOAD_BALANCE:
1456
		retval = update_flag(CS_SCHED_LOAD_BALANCE, cs, val);
1457
		break;
1458
	case FILE_MEMORY_MIGRATE:
1459
		retval = update_flag(CS_MEMORY_MIGRATE, cs, val);
1460
		break;
1461
	case FILE_MEMORY_PRESSURE_ENABLED:
1462
		cpuset_memory_pressure_enabled = !!val;
1463 1464 1465 1466
		break;
	case FILE_MEMORY_PRESSURE:
		retval = -EACCES;
		break;
1467
	case FILE_SPREAD_PAGE:
1468
		retval = update_flag(CS_SPREAD_PAGE, cs, val);
1469 1470
		break;
	case FILE_SPREAD_SLAB:
1471
		retval = update_flag(CS_SPREAD_SLAB, cs, val);
1472
		break;
L
Linus Torvalds 已提交
1473 1474
	default:
		retval = -EINVAL;
1475
		break;
L
Linus Torvalds 已提交
1476
	}
1477
	cgroup_unlock();
L
Linus Torvalds 已提交
1478 1479 1480
	return retval;
}

1481 1482 1483 1484 1485 1486
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;

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

1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501
	switch (type) {
	case FILE_SCHED_RELAX_DOMAIN_LEVEL:
		retval = update_relax_domain_level(cs, val);
		break;
	default:
		retval = -EINVAL;
		break;
	}
	cgroup_unlock();
	return retval;
}

1502 1503 1504 1505 1506 1507 1508
/*
 * 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;
1509 1510
	struct cpuset *cs = cgroup_cs(cgrp);
	struct cpuset *trialcs;
1511 1512 1513 1514

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

1515 1516 1517 1518
	trialcs = alloc_trial_cpuset(cs);
	if (!trialcs)
		return -ENOMEM;

1519 1520
	switch (cft->private) {
	case FILE_CPULIST:
1521
		retval = update_cpumask(cs, trialcs, buf);
1522 1523
		break;
	case FILE_MEMLIST:
1524
		retval = update_nodemask(cs, trialcs, buf);
1525 1526 1527 1528 1529
		break;
	default:
		retval = -EINVAL;
		break;
	}
1530 1531

	free_trial_cpuset(trialcs);
1532 1533 1534 1535
	cgroup_unlock();
	return retval;
}

L
Linus Torvalds 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
/*
 * These ascii lists should be read in a single call, by using a user
 * buffer large enough to hold the entire map.  If read in smaller
 * chunks, there is no guarantee of atomicity.  Since the display format
 * used, list of ranges of sequential numbers, is variable length,
 * and since these maps can change value dynamically, one could read
 * gibberish by doing partial reads while a list was changing.
 * A single large read to a buffer that crosses a page boundary is
 * ok, because the result being copied to user land is not recomputed
 * across a page fault.
 */

static int cpuset_sprintf_cpulist(char *page, struct cpuset *cs)
{
1550
	int ret;
L
Linus Torvalds 已提交
1551

1552
	mutex_lock(&callback_mutex);
1553
	ret = cpulist_scnprintf(page, PAGE_SIZE, cs->cpus_allowed);
1554
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1555

1556
	return ret;
L
Linus Torvalds 已提交
1557 1558 1559 1560 1561 1562
}

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

1563
	mutex_lock(&callback_mutex);
L
Linus Torvalds 已提交
1564
	mask = cs->mems_allowed;
1565
	mutex_unlock(&callback_mutex);
L
Linus Torvalds 已提交
1566 1567 1568 1569

	return nodelist_scnprintf(page, PAGE_SIZE, mask);
}

1570 1571 1572 1573 1574
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
1575
{
1576
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1577 1578 1579 1580 1581
	cpuset_filetype_t type = cft->private;
	char *page;
	ssize_t retval = 0;
	char *s;

1582
	if (!(page = (char *)__get_free_page(GFP_TEMPORARY)))
L
Linus Torvalds 已提交
1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599
		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
1600
	retval = simple_read_from_buffer(buf, nbytes, ppos, page, s - page);
L
Linus Torvalds 已提交
1601 1602 1603 1604 1605
out:
	free_page((unsigned long)page);
	return retval;
}

1606 1607 1608 1609 1610 1611 1612 1613 1614
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);
1615 1616
	case FILE_MEM_HARDWALL:
		return is_mem_hardwall(cs);
1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
	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();
	}
1632 1633 1634

	/* Unreachable but makes gcc happy */
	return 0;
1635
}
L
Linus Torvalds 已提交
1636

1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
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();
	}
1647 1648 1649

	/* Unrechable but makes gcc happy */
	return 0;
1650 1651
}

L
Linus Torvalds 已提交
1652 1653 1654 1655 1656

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

1657 1658 1659 1660
static struct cftype files[] = {
	{
		.name = "cpus",
		.read = cpuset_common_file_read,
1661 1662
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * NR_CPUS),
1663 1664 1665 1666 1667 1668
		.private = FILE_CPULIST,
	},

	{
		.name = "mems",
		.read = cpuset_common_file_read,
1669 1670
		.write_string = cpuset_write_resmask,
		.max_write_len = (100U + 6 * MAX_NUMNODES),
1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687
		.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,
	},

1688 1689 1690 1691 1692 1693 1694
	{
		.name = "mem_hardwall",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_MEM_HARDWALL,
	},

1695 1696 1697 1698 1699 1700 1701 1702 1703
	{
		.name = "sched_load_balance",
		.read_u64 = cpuset_read_u64,
		.write_u64 = cpuset_write_u64,
		.private = FILE_SCHED_LOAD_BALANCE,
	},

	{
		.name = "sched_relax_domain_level",
1704 1705
		.read_s64 = cpuset_read_s64,
		.write_s64 = cpuset_write_s64,
1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720
		.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
1721
		.mode = S_IRUGO,
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736
	},

	{
		.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,
	},
1737 1738
};

1739 1740
static struct cftype cft_memory_pressure_enabled = {
	.name = "memory_pressure_enabled",
1741 1742
	.read_u64 = cpuset_read_u64,
	.write_u64 = cpuset_write_u64,
1743 1744 1745
	.private = FILE_MEMORY_PRESSURE_ENABLED,
};

1746
static int cpuset_populate(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1747 1748 1749
{
	int err;

1750 1751
	err = cgroup_add_files(cont, ss, files, ARRAY_SIZE(files));
	if (err)
L
Linus Torvalds 已提交
1752
		return err;
1753
	/* memory_pressure_enabled is in root cpuset only */
1754
	if (!cont->parent)
1755
		err = cgroup_add_file(cont, ss,
1756 1757
				      &cft_memory_pressure_enabled);
	return err;
L
Linus Torvalds 已提交
1758 1759
}

1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
/*
 * post_clone() is called at the end of cgroup_clone().
 * 'cgroup' was just created automatically as a result of
 * a cgroup_clone(), and the current task is about to
 * be moved into 'cgroup'.
 *
 * Currently we refuse to set up the cgroup - thereby
 * refusing the task to be entered, and as a result refusing
 * the sys_unshare() or clone() which initiated it - if any
 * sibling cpusets have exclusive cpus or mem.
 *
 * If this becomes a problem for some users who wish to
 * allow that scenario, then cpuset_post_clone() could be
 * changed to grant parent->cpus_allowed-sibling_cpus_exclusive
1774 1775
 * (and likewise for mems) to the new cgroup. Called with cgroup_mutex
 * held.
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
 */
static void cpuset_post_clone(struct cgroup_subsys *ss,
			      struct cgroup *cgroup)
{
	struct cgroup *parent, *child;
	struct cpuset *cs, *parent_cs;

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

	cs->mems_allowed = parent_cs->mems_allowed;
1793
	cpumask_copy(cs->cpus_allowed, parent_cs->cpus_allowed);
1794 1795 1796
	return;
}

L
Linus Torvalds 已提交
1797 1798
/*
 *	cpuset_create - create a cpuset
1799 1800
 *	ss:	cpuset cgroup subsystem
 *	cont:	control group that the new cpuset will be part of
L
Linus Torvalds 已提交
1801 1802
 */

1803 1804 1805
static struct cgroup_subsys_state *cpuset_create(
	struct cgroup_subsys *ss,
	struct cgroup *cont)
L
Linus Torvalds 已提交
1806 1807
{
	struct cpuset *cs;
1808
	struct cpuset *parent;
L
Linus Torvalds 已提交
1809

1810 1811 1812 1813
	if (!cont->parent) {
		return &top_cpuset.css;
	}
	parent = cgroup_cs(cont->parent);
L
Linus Torvalds 已提交
1814 1815
	cs = kmalloc(sizeof(*cs), GFP_KERNEL);
	if (!cs)
1816
		return ERR_PTR(-ENOMEM);
1817 1818 1819 1820
	if (!alloc_cpumask_var(&cs->cpus_allowed, GFP_KERNEL)) {
		kfree(cs);
		return ERR_PTR(-ENOMEM);
	}
L
Linus Torvalds 已提交
1821 1822

	cs->flags = 0;
1823 1824 1825 1826
	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
1827
	set_bit(CS_SCHED_LOAD_BALANCE, &cs->flags);
1828
	cpumask_clear(cs->cpus_allowed);
1829
	nodes_clear(cs->mems_allowed);
1830
	fmeter_init(&cs->fmeter);
1831
	cs->relax_domain_level = -1;
L
Linus Torvalds 已提交
1832 1833

	cs->parent = parent;
1834
	number_of_cpusets++;
1835
	return &cs->css ;
L
Linus Torvalds 已提交
1836 1837
}

P
Paul Jackson 已提交
1838 1839 1840
/*
 * If the cpuset being removed has its flag 'sched_load_balance'
 * enabled, then simulate turning sched_load_balance off, which
1841
 * will call async_rebuild_sched_domains().
P
Paul Jackson 已提交
1842 1843
 */

1844
static void cpuset_destroy(struct cgroup_subsys *ss, struct cgroup *cont)
L
Linus Torvalds 已提交
1845
{
1846
	struct cpuset *cs = cgroup_cs(cont);
L
Linus Torvalds 已提交
1847

P
Paul Jackson 已提交
1848
	if (is_sched_load_balance(cs))
1849
		update_flag(CS_SCHED_LOAD_BALANCE, cs, 0);
P
Paul Jackson 已提交
1850

1851
	number_of_cpusets--;
1852
	free_cpumask_var(cs->cpus_allowed);
1853
	kfree(cs);
L
Linus Torvalds 已提交
1854 1855
}

1856 1857 1858
struct cgroup_subsys cpuset_subsys = {
	.name = "cpuset",
	.create = cpuset_create,
1859
	.destroy = cpuset_destroy,
1860 1861 1862 1863 1864 1865 1866 1867
	.can_attach = cpuset_can_attach,
	.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
1868 1869 1870 1871 1872 1873 1874 1875
/**
 * cpuset_init - initialize cpusets at system boot
 *
 * Description: Initialize top_cpuset and the cpuset internal file system,
 **/

int __init cpuset_init(void)
{
1876
	int err = 0;
L
Linus Torvalds 已提交
1877

1878 1879 1880
	if (!alloc_cpumask_var(&top_cpuset.cpus_allowed, GFP_KERNEL))
		BUG();

1881
	cpumask_setall(top_cpuset.cpus_allowed);
1882
	nodes_setall(top_cpuset.mems_allowed);
L
Linus Torvalds 已提交
1883

1884
	fmeter_init(&top_cpuset.fmeter);
P
Paul Jackson 已提交
1885
	set_bit(CS_SCHED_LOAD_BALANCE, &top_cpuset.flags);
1886
	top_cpuset.relax_domain_level = -1;
L
Linus Torvalds 已提交
1887 1888 1889

	err = register_filesystem(&cpuset_fs_type);
	if (err < 0)
1890 1891
		return err;

1892 1893 1894
	if (!alloc_cpumask_var(&cpus_attach, GFP_KERNEL))
		BUG();

1895
	number_of_cpusets = 1;
1896
	return 0;
L
Linus Torvalds 已提交
1897 1898
}

1899 1900 1901 1902 1903 1904 1905 1906
/**
 * 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.
 */
1907 1908
static void cpuset_do_move_task(struct task_struct *tsk,
				struct cgroup_scanner *scan)
1909
{
1910
	struct cgroup *new_cgroup = scan->data;
1911

1912
	cgroup_attach_task(new_cgroup, tsk);
1913 1914 1915 1916 1917 1918 1919
}

/**
 * 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
 *
1920 1921
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1922 1923 1924 1925 1926 1927
 *
 * 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)
{
1928
	struct cgroup_scanner scan;
1929

1930 1931 1932 1933 1934
	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;
1935

1936
	if (cgroup_scan_tasks(&scan))
1937 1938 1939 1940
		printk(KERN_ERR "move_member_tasks_to_cpuset: "
				"cgroup_scan_tasks failed\n");
}

1941
/*
1942
 * If CPU and/or memory hotplug handlers, below, unplug any CPUs
1943 1944
 * or memory nodes, we need to walk over the cpuset hierarchy,
 * removing that CPU or node from all cpusets.  If this removes the
1945 1946
 * last CPU or node from a cpuset, then move the tasks in the empty
 * cpuset to its next-highest non-empty parent.
1947
 *
1948 1949
 * Called with cgroup_mutex held
 * callback_mutex must not be held, as cpuset_attach() will take it.
1950
 */
1951 1952 1953 1954
static void remove_tasks_in_empty_cpuset(struct cpuset *cs)
{
	struct cpuset *parent;

1955 1956 1957 1958 1959
	/*
	 * 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.
	 */
1960 1961
	if (list_empty(&cs->css.cgroup->css_sets))
		return;
1962

1963 1964 1965 1966 1967
	/*
	 * Find its next-highest non-empty parent, (top cpuset
	 * has online cpus, so can't be empty).
	 */
	parent = cs->parent;
1968
	while (cpumask_empty(parent->cpus_allowed) ||
1969
			nodes_empty(parent->mems_allowed))
1970 1971 1972 1973 1974 1975 1976 1977 1978
		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.
 *
1979
 * Called with cgroup_mutex held.  We take callback_mutex to modify
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
 * 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.
 */
1990
static void scan_for_empty_cpusets(struct cpuset *root)
1991
{
1992
	LIST_HEAD(queue);
1993 1994
	struct cpuset *cp;	/* scans cpusets being updated */
	struct cpuset *child;	/* scans child cpusets of cp */
1995
	struct cgroup *cont;
1996
	nodemask_t oldmems;
1997

1998 1999 2000
	list_add_tail((struct list_head *)&root->stack_list, &queue);

	while (!list_empty(&queue)) {
2001
		cp = list_first_entry(&queue, struct cpuset, stack_list);
2002 2003 2004 2005 2006
		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);
		}
2007 2008

		/* Continue past cpusets with all cpus, mems online */
2009
		if (cpumask_subset(cp->cpus_allowed, cpu_active_mask) &&
2010 2011 2012
		    nodes_subset(cp->mems_allowed, node_states[N_HIGH_MEMORY]))
			continue;

2013 2014
		oldmems = cp->mems_allowed;

2015
		/* Remove offline cpus and mems from this cpuset. */
2016
		mutex_lock(&callback_mutex);
2017
		cpumask_and(cp->cpus_allowed, cp->cpus_allowed,
2018
			    cpu_active_mask);
2019 2020
		nodes_and(cp->mems_allowed, cp->mems_allowed,
						node_states[N_HIGH_MEMORY]);
2021 2022 2023
		mutex_unlock(&callback_mutex);

		/* Move tasks from the empty cpuset to a parent */
2024
		if (cpumask_empty(cp->cpus_allowed) ||
2025
		     nodes_empty(cp->mems_allowed))
2026
			remove_tasks_in_empty_cpuset(cp);
2027
		else {
2028
			update_tasks_cpumask(cp, NULL);
2029
			update_tasks_nodemask(cp, &oldmems, NULL);
2030
		}
2031 2032 2033
	}
}

2034 2035 2036 2037 2038 2039
/*
 * 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.
 *
2040 2041
 * This routine ensures that top_cpuset.cpus_allowed tracks
 * cpu_online_map on each CPU hotplug (cpuhp) event.
2042 2043 2044
 *
 * Called within get_online_cpus().  Needs to call cgroup_lock()
 * before calling generate_sched_domains().
2045
 */
2046
static int cpuset_track_online_cpus(struct notifier_block *unused_nb,
P
Paul Jackson 已提交
2047
				unsigned long phase, void *unused_cpu)
2048
{
2049
	struct sched_domain_attr *attr;
2050
	cpumask_var_t *doms;
2051 2052
	int ndoms;

2053 2054 2055
	switch (phase) {
	case CPU_ONLINE:
	case CPU_ONLINE_FROZEN:
2056 2057 2058 2059
	case CPU_DOWN_PREPARE:
	case CPU_DOWN_PREPARE_FROZEN:
	case CPU_DOWN_FAILED:
	case CPU_DOWN_FAILED_FROZEN:
2060
		break;
2061

2062
	default:
2063
		return NOTIFY_DONE;
2064
	}
2065

2066
	cgroup_lock();
2067
	mutex_lock(&callback_mutex);
2068
	cpumask_copy(top_cpuset.cpus_allowed, cpu_active_mask);
2069
	mutex_unlock(&callback_mutex);
2070 2071 2072 2073 2074 2075 2076
	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);

2077
	return NOTIFY_OK;
2078 2079
}

2080
#ifdef CONFIG_MEMORY_HOTPLUG
2081
/*
2082
 * Keep top_cpuset.mems_allowed tracking node_states[N_HIGH_MEMORY].
2083 2084
 * Call this routine anytime after node_states[N_HIGH_MEMORY] changes.
 * See also the previous routine cpuset_track_online_cpus().
2085
 */
2086 2087
static int cpuset_track_online_nodes(struct notifier_block *self,
				unsigned long action, void *arg)
2088
{
2089 2090
	nodemask_t oldmems;

2091
	cgroup_lock();
2092 2093
	switch (action) {
	case MEM_ONLINE:
2094
		oldmems = top_cpuset.mems_allowed;
2095
		mutex_lock(&callback_mutex);
2096
		top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2097
		mutex_unlock(&callback_mutex);
2098 2099 2100 2101 2102 2103 2104 2105
		update_tasks_nodemask(&top_cpuset, &oldmems, NULL);
		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);
2106 2107 2108 2109
		break;
	default:
		break;
	}
2110
	cgroup_unlock();
2111
	return NOTIFY_OK;
2112 2113 2114
}
#endif

L
Linus Torvalds 已提交
2115 2116 2117 2118 2119 2120 2121 2122
/**
 * cpuset_init_smp - initialize cpus_allowed
 *
 * Description: Finish top cpuset after cpu, node maps are initialized
 **/

void __init cpuset_init_smp(void)
{
2123
	cpumask_copy(top_cpuset.cpus_allowed, cpu_active_mask);
2124
	top_cpuset.mems_allowed = node_states[N_HIGH_MEMORY];
2125

2126
	hotcpu_notifier(cpuset_track_online_cpus, 0);
2127
	hotplug_memory_notifier(cpuset_track_online_nodes, 10);
2128 2129 2130

	cpuset_wq = create_singlethread_workqueue("cpuset");
	BUG_ON(!cpuset_wq);
L
Linus Torvalds 已提交
2131 2132 2133 2134 2135
}

/**
 * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset.
 * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed.
2136
 * @pmask: pointer to struct cpumask variable to receive cpus_allowed set.
L
Linus Torvalds 已提交
2137
 *
2138
 * Description: Returns the cpumask_var_t cpus_allowed of the cpuset
L
Linus Torvalds 已提交
2139 2140 2141 2142 2143
 * 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.
 **/

2144
void cpuset_cpus_allowed(struct task_struct *tsk, struct cpumask *pmask)
L
Linus Torvalds 已提交
2145
{
2146
	mutex_lock(&callback_mutex);
2147
	cpuset_cpus_allowed_locked(tsk, pmask);
2148 2149 2150 2151 2152
	mutex_unlock(&callback_mutex);
}

/**
 * cpuset_cpus_allowed_locked - return cpus_allowed mask from a tasks cpuset.
2153
 * Must be called with callback_mutex held.
2154
 **/
2155
void cpuset_cpus_allowed_locked(struct task_struct *tsk, struct cpumask *pmask)
2156
{
2157
	task_lock(tsk);
2158
	guarantee_online_cpus(task_cs(tsk), pmask);
2159
	task_unlock(tsk);
L
Linus Torvalds 已提交
2160 2161 2162 2163
}

void cpuset_init_current_mems_allowed(void)
{
2164
	nodes_setall(current->mems_allowed);
L
Linus Torvalds 已提交
2165 2166
}

2167 2168 2169 2170 2171 2172
/**
 * 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
2173
 * subset of node_states[N_HIGH_MEMORY], even if this means going outside the
2174 2175 2176 2177 2178 2179 2180
 * tasks cpuset.
 **/

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

2181
	mutex_lock(&callback_mutex);
2182
	task_lock(tsk);
2183
	guarantee_online_mems(task_cs(tsk), &mask);
2184
	task_unlock(tsk);
2185
	mutex_unlock(&callback_mutex);
2186 2187 2188 2189

	return mask;
}

2190
/**
2191 2192
 * cpuset_nodemask_valid_mems_allowed - check nodemask vs. curremt mems_allowed
 * @nodemask: the nodemask to be checked
2193
 *
2194
 * Are any of the nodes in the nodemask allowed in current->mems_allowed?
L
Linus Torvalds 已提交
2195
 */
2196
int cpuset_nodemask_valid_mems_allowed(nodemask_t *nodemask)
L
Linus Torvalds 已提交
2197
{
2198
	return nodes_intersects(*nodemask, current->mems_allowed);
L
Linus Torvalds 已提交
2199 2200
}

2201
/*
2202 2203 2204 2205
 * 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.
2206
 */
2207
static const struct cpuset *nearest_hardwall_ancestor(const struct cpuset *cs)
2208
{
2209
	while (!(is_mem_exclusive(cs) || is_mem_hardwall(cs)) && cs->parent)
2210 2211 2212 2213
		cs = cs->parent;
	return cs;
}

2214
/**
2215 2216
 * cpuset_node_allowed_softwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2217
 * @gfp_mask: memory allocation flags
2218
 *
2219 2220 2221 2222 2223 2224
 * 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.
2225 2226
 * Otherwise, no.
 *
2227 2228 2229
 * 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.
2230
 *
2231 2232
 * cpuset_node_allowed_hardwall() only handles the simpler case of hardwall
 * cpusets, and never sleeps.
2233 2234 2235 2236 2237 2238 2239
 *
 * 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'.
 *
2240
 * GFP_USER allocations are marked with the __GFP_HARDWALL bit,
2241 2242
 * and do not allow allocations outside the current tasks cpuset
 * unless the task has been OOM killed as is marked TIF_MEMDIE.
2243
 * GFP_KERNEL allocations are not so marked, so can escape to the
2244
 * nearest enclosing hardwalled ancestor cpuset.
2245
 *
2246 2247 2248 2249 2250 2251 2252
 * 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.
2253
 *
2254
 * The first call here from mm/page_alloc:get_page_from_freelist()
2255 2256 2257
 * 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).
2258 2259 2260 2261 2262 2263
 *
 * 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:
2264 2265
 *	in_interrupt - any node ok (current task context irrelevant)
 *	GFP_ATOMIC   - any node ok
2266
 *	TIF_MEMDIE   - any node ok
2267
 *	GFP_KERNEL   - any node in enclosing hardwalled cpuset ok
2268
 *	GFP_USER     - only nodes in current tasks mems allowed ok.
2269 2270
 *
 * Rule:
2271
 *    Don't call cpuset_node_allowed_softwall if you can't sleep, unless you
2272 2273
 *    pass in the __GFP_HARDWALL flag set in gfp_flag, which disables
 *    the code that might scan up ancestor cpusets and sleep.
2274
 */
2275
int __cpuset_node_allowed_softwall(int node, gfp_t gfp_mask)
L
Linus Torvalds 已提交
2276
{
2277
	const struct cpuset *cs;	/* current cpuset ancestors */
2278
	int allowed;			/* is allocation in zone z allowed? */
2279

2280
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
2281
		return 1;
2282
	might_sleep_if(!(gfp_mask & __GFP_HARDWALL));
2283 2284
	if (node_isset(node, current->mems_allowed))
		return 1;
2285 2286 2287 2288 2289 2290
	/*
	 * 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;
2291 2292 2293
	if (gfp_mask & __GFP_HARDWALL)	/* If hardwall request, stop here */
		return 0;

2294 2295 2296
	if (current->flags & PF_EXITING) /* Let dying task have memory */
		return 1;

2297
	/* Not hardwall and node outside mems_allowed: scan up cpusets */
2298
	mutex_lock(&callback_mutex);
2299 2300

	task_lock(current);
2301
	cs = nearest_hardwall_ancestor(task_cs(current));
2302 2303
	task_unlock(current);

2304
	allowed = node_isset(node, cs->mems_allowed);
2305
	mutex_unlock(&callback_mutex);
2306
	return allowed;
L
Linus Torvalds 已提交
2307 2308
}

2309
/*
2310 2311
 * cpuset_node_allowed_hardwall - Can we allocate on a memory node?
 * @node: is this an allowed node?
2312 2313
 * @gfp_mask: memory allocation flags
 *
2314 2315 2316 2317 2318
 * 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.
2319 2320 2321 2322 2323 2324 2325
 *
 * 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'.
 *
2326 2327
 * Unlike the cpuset_node_allowed_softwall() variant, above,
 * this variant requires that the node be in the current task's
2328 2329 2330 2331
 * 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.
 */
2332
int __cpuset_node_allowed_hardwall(int node, gfp_t gfp_mask)
2333 2334 2335 2336 2337
{
	if (in_interrupt() || (gfp_mask & __GFP_THISNODE))
		return 1;
	if (node_isset(node, current->mems_allowed))
		return 1;
D
Daniel Walker 已提交
2338 2339 2340 2341 2342 2343
	/*
	 * 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;
2344 2345 2346
	return 0;
}

P
Paul Jackson 已提交
2347 2348 2349
/**
 * cpuset_lock - lock out any changes to cpuset structures
 *
2350
 * The out of memory (oom) code needs to mutex_lock cpusets
P
Paul Jackson 已提交
2351
 * from being changed while it scans the tasklist looking for a
2352
 * task in an overlapping cpuset.  Expose callback_mutex via this
P
Paul Jackson 已提交
2353 2354
 * cpuset_lock() routine, so the oom code can lock it, before
 * locking the task list.  The tasklist_lock is a spinlock, so
2355
 * must be taken inside callback_mutex.
P
Paul Jackson 已提交
2356 2357 2358 2359
 */

void cpuset_lock(void)
{
2360
	mutex_lock(&callback_mutex);
P
Paul Jackson 已提交
2361 2362 2363 2364 2365 2366 2367 2368 2369 2370
}

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

void cpuset_unlock(void)
{
2371
	mutex_unlock(&callback_mutex);
P
Paul Jackson 已提交
2372 2373
}

2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
/**
 * cpuset_mem_spread_node() - On which node to begin search for a page
 *
 * If a task is marked PF_SPREAD_PAGE or PF_SPREAD_SLAB (as for
 * tasks in a cpuset with is_spread_page or is_spread_slab set),
 * and if the memory allocation used cpuset_mem_spread_node()
 * to determine on which node to start looking, as it will for
 * certain page cache or slab cache pages such as used for file
 * system buffers and inode caches, then instead of starting on the
 * local node to look for a free page, rather spread the starting
 * node around the tasks mems_allowed nodes.
 *
 * We don't have to worry about the returned node being offline
 * because "it can't happen", and even if it did, it would be ok.
 *
 * The routines calling guarantee_online_mems() are careful to
 * only set nodes in task->mems_allowed that are online.  So it
 * should not be possible for the following code to return an
 * offline node.  But if it did, that would be ok, as this routine
 * is not returning the node where the allocation must be, only
 * the node where the search should start.  The zonelist passed to
 * __alloc_pages() will include all nodes.  If the slab allocator
 * is passed an offline node, it will fall back to the local node.
 * See kmem_cache_alloc_node().
 */

int cpuset_mem_spread_node(void)
{
	int node;

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

2412
/**
2413 2414 2415 2416 2417 2418 2419 2420
 * 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.
2421 2422
 **/

2423 2424
int cpuset_mems_allowed_intersects(const struct task_struct *tsk1,
				   const struct task_struct *tsk2)
2425
{
2426
	return nodes_intersects(tsk1->mems_allowed, tsk2->mems_allowed);
2427 2428
}

2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451
/**
 * 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);
}

2452 2453 2454 2455 2456 2457
/*
 * 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.
 */

2458
int cpuset_memory_pressure_enabled __read_mostly;
2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480

/**
 * 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);
2481
	fmeter_markevent(&task_cs(current)->fmeter);
2482 2483 2484
	task_unlock(current);
}

2485
#ifdef CONFIG_PROC_PID_CPUSET
L
Linus Torvalds 已提交
2486 2487 2488 2489
/*
 * proc_cpuset_show()
 *  - Print tasks cpuset path into seq_file.
 *  - Used for /proc/<pid>/cpuset.
2490 2491
 *  - 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,
2492
 *    and we take cgroup_mutex, keeping cpuset_attach() from changing it
2493
 *    anyway.
L
Linus Torvalds 已提交
2494
 */
P
Paul Jackson 已提交
2495
static int proc_cpuset_show(struct seq_file *m, void *unused_v)
L
Linus Torvalds 已提交
2496
{
2497
	struct pid *pid;
L
Linus Torvalds 已提交
2498 2499
	struct task_struct *tsk;
	char *buf;
2500
	struct cgroup_subsys_state *css;
2501
	int retval;
L
Linus Torvalds 已提交
2502

2503
	retval = -ENOMEM;
L
Linus Torvalds 已提交
2504 2505
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
2506 2507 2508
		goto out;

	retval = -ESRCH;
2509 2510
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
2511 2512
	if (!tsk)
		goto out_free;
L
Linus Torvalds 已提交
2513

2514
	retval = -EINVAL;
2515 2516 2517
	cgroup_lock();
	css = task_subsys_state(tsk, cpuset_subsys_id);
	retval = cgroup_path(css->cgroup, buf, PAGE_SIZE);
L
Linus Torvalds 已提交
2518
	if (retval < 0)
2519
		goto out_unlock;
L
Linus Torvalds 已提交
2520 2521
	seq_puts(m, buf);
	seq_putc(m, '\n');
2522
out_unlock:
2523
	cgroup_unlock();
2524 2525
	put_task_struct(tsk);
out_free:
L
Linus Torvalds 已提交
2526
	kfree(buf);
2527
out:
L
Linus Torvalds 已提交
2528 2529 2530 2531 2532
	return retval;
}

static int cpuset_open(struct inode *inode, struct file *file)
{
2533 2534
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cpuset_show, pid);
L
Linus Torvalds 已提交
2535 2536
}

2537
const struct file_operations proc_cpuset_operations = {
L
Linus Torvalds 已提交
2538 2539 2540 2541 2542
	.open		= cpuset_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};
2543
#endif /* CONFIG_PROC_PID_CPUSET */
L
Linus Torvalds 已提交
2544

2545
/* Display task mems_allowed in /proc/<pid>/status file. */
2546 2547 2548
void cpuset_task_status_allowed(struct seq_file *m, struct task_struct *task)
{
	seq_printf(m, "Mems_allowed:\t");
2549
	seq_nodemask(m, &task->mems_allowed);
2550
	seq_printf(m, "\n");
2551
	seq_printf(m, "Mems_allowed_list:\t");
2552
	seq_nodemask_list(m, &task->mems_allowed);
2553
	seq_printf(m, "\n");
L
Linus Torvalds 已提交
2554
}