cgroup.c 141.2 KB
Newer Older
1 2 3 4 5 6
/*
 *  Generic process-grouping system.
 *
 *  Based originally on the cpuset system, extracted by Paul Menage
 *  Copyright (C) 2006 Google, Inc
 *
7 8 9 10
 *  Notifications support
 *  Copyright (C) 2009 Nokia Corporation
 *  Author: Kirill A. Shutemov
 *
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
 *  Copyright notices from the original cpuset code:
 *  --------------------------------------------------
 *  Copyright (C) 2003 BULL SA.
 *  Copyright (C) 2004-2006 Silicon Graphics, Inc.
 *
 *  Portions derived from Patrick Mochel's sysfs code.
 *  sysfs is Copyright (c) 2001-3 Patrick Mochel
 *
 *  2003-10-10 Written by Simon Derr.
 *  2003-10-22 Updates by Stephen Hemminger.
 *  2004 May-July Rework by Paul Jackson.
 *  ---------------------------------------------------
 *
 *  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/cgroup.h>
30
#include <linux/cred.h>
31
#include <linux/ctype.h>
32
#include <linux/errno.h>
33
#include <linux/init_task.h>
34 35 36 37 38 39
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
40
#include <linux/proc_fs.h>
41 42
#include <linux/rcupdate.h>
#include <linux/sched.h>
43
#include <linux/backing-dev.h>
44 45 46 47
#include <linux/slab.h>
#include <linux/magic.h>
#include <linux/spinlock.h>
#include <linux/string.h>
48
#include <linux/sort.h>
49
#include <linux/kmod.h>
B
Balbir Singh 已提交
50 51
#include <linux/delayacct.h>
#include <linux/cgroupstats.h>
52
#include <linux/hashtable.h>
53
#include <linux/namei.h>
L
Li Zefan 已提交
54
#include <linux/pid_namespace.h>
55
#include <linux/idr.h>
56
#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
57
#include <linux/flex_array.h> /* used in cgroup_attach_task */
58
#include <linux/kthread.h>
B
Balbir Singh 已提交
59

A
Arun Sharma 已提交
60
#include <linux/atomic.h>
61

62 63 64 65 66 67 68 69
/*
 * pidlists linger the following amount before being destroyed.  The goal
 * is avoiding frequent destruction in the middle of consecutive read calls
 * Expiring in the middle is a performance problem not a correctness one.
 * 1 sec should be enough.
 */
#define CGROUP_PIDLIST_DESTROY_DELAY	HZ

T
Tejun Heo 已提交
70 71 72
#define CGROUP_FILE_NAME_MAX		(MAX_CGROUP_TYPE_NAMELEN +	\
					 MAX_CFTYPE_NAME + 2)

T
Tejun Heo 已提交
73 74 75 76 77 78 79 80 81
/*
 * cgroup_tree_mutex nests above cgroup_mutex and protects cftypes, file
 * creation/removal and hierarchy changing operations including cgroup
 * creation, removal, css association and controller rebinding.  This outer
 * lock is needed mainly to resolve the circular dependency between kernfs
 * active ref and cgroup_mutex.  cgroup_tree_mutex nests above both.
 */
static DEFINE_MUTEX(cgroup_tree_mutex);

T
Tejun Heo 已提交
82 83 84 85
/*
 * cgroup_mutex is the master lock.  Any modification to cgroup or its
 * hierarchy must be performed while holding it.
 */
T
Tejun Heo 已提交
86 87
#ifdef CONFIG_PROVE_RCU
DEFINE_MUTEX(cgroup_mutex);
88
EXPORT_SYMBOL_GPL(cgroup_mutex);	/* only for lockdep */
T
Tejun Heo 已提交
89
#else
90
static DEFINE_MUTEX(cgroup_mutex);
T
Tejun Heo 已提交
91 92
#endif

93 94 95 96 97 98
/*
 * Protects cgroup_subsys->release_agent_path.  Modifying it also requires
 * cgroup_mutex.  Reading requires either cgroup_mutex or this spinlock.
 */
static DEFINE_SPINLOCK(release_agent_path_lock);

T
Tejun Heo 已提交
99
#define cgroup_assert_mutexes_or_rcu_locked()				\
100
	rcu_lockdep_assert(rcu_read_lock_held() ||			\
T
Tejun Heo 已提交
101
			   lockdep_is_held(&cgroup_tree_mutex) ||	\
102
			   lockdep_is_held(&cgroup_mutex),		\
T
Tejun Heo 已提交
103
			   "cgroup_[tree_]mutex or RCU read lock required");
104

105 106 107 108 109 110 111 112
/*
 * cgroup destruction makes heavy use of work items and there can be a lot
 * of concurrent destructions.  Use a separate workqueue so that cgroup
 * destruction work items don't end up filling up max_active of system_wq
 * which may lead to deadlock.
 */
static struct workqueue_struct *cgroup_destroy_wq;

113 114 115 116 117 118
/*
 * pidlist destructions need to be flushed on cgroup destruction.  Use a
 * separate workqueue as flush domain.
 */
static struct workqueue_struct *cgroup_pidlist_destroy_wq;

T
Tejun Heo 已提交
119
/* generate an array of cgroup subsystem pointers */
120
#define SUBSYS(_x) [_x ## _cgrp_id] = &_x ## _cgrp_subsys,
T
Tejun Heo 已提交
121
static struct cgroup_subsys *cgroup_subsys[] = {
122 123
#include <linux/cgroup_subsys.h>
};
124 125 126 127 128 129 130 131
#undef SUBSYS

/* array of cgroup subsystem names */
#define SUBSYS(_x) [_x ## _cgrp_id] = #_x,
static const char *cgroup_subsys_name[] = {
#include <linux/cgroup_subsys.h>
};
#undef SUBSYS
132 133

/*
134 135 136
 * The dummy hierarchy, reserved for the subsystems that are otherwise
 * unattached - it never has more than a single cgroup, and all tasks are
 * part of that cgroup.
137
 */
138 139 140 141
static struct cgroupfs_root cgroup_dummy_root;

/* dummy_top is a shorthand for the dummy hierarchy's top cgroup */
static struct cgroup * const cgroup_dummy_top = &cgroup_dummy_root.top_cgroup;
142 143 144

/* The list of hierarchy roots */

145 146
static LIST_HEAD(cgroup_roots);
static int cgroup_root_count;
147

T
Tejun Heo 已提交
148
/* hierarchy ID allocation and mapping, protected by cgroup_mutex */
149
static DEFINE_IDR(cgroup_hierarchy_idr);
150

151 152
static struct cgroup_name root_cgroup_name = { .name = "/" };

153 154 155 156 157
/*
 * Assign a monotonically increasing serial number to cgroups.  It
 * guarantees cgroups with bigger numbers are newer than those with smaller
 * numbers.  Also, as cgroups are always appended to the parent's
 * ->children list, it guarantees that sibling cgroups are always sorted in
158 159
 * the ascending serial number order on the list.  Protected by
 * cgroup_mutex.
160
 */
161
static u64 cgroup_serial_nr_next = 1;
162

163
/* This flag indicates whether tasks in the fork and exit paths should
L
Li Zefan 已提交
164 165 166
 * check for fork/exit handlers to call. This avoids us having to do
 * extra work in the fork/exit path if none of the subsystems need to
 * be called.
167
 */
168
static int need_forkexit_callback __read_mostly;
169

170 171
static struct cftype cgroup_base_files[];

172
static void cgroup_destroy_css_killed(struct cgroup *cgrp);
173
static int cgroup_destroy_locked(struct cgroup *cgrp);
174 175
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add);
176
static int cgroup_file_release(struct inode *inode, struct file *file);
177
static void cgroup_pidlist_destroy_all(struct cgroup *cgrp);
178

T
Tejun Heo 已提交
179 180 181
/**
 * cgroup_css - obtain a cgroup's css for the specified subsystem
 * @cgrp: the cgroup of interest
182
 * @ss: the subsystem of interest (%NULL returns the dummy_css)
T
Tejun Heo 已提交
183
 *
184 185 186 187 188
 * Return @cgrp's css (cgroup_subsys_state) associated with @ss.  This
 * function must be called either under cgroup_mutex or rcu_read_lock() and
 * the caller is responsible for pinning the returned css if it wants to
 * keep accessing it outside the said locks.  This function may return
 * %NULL if @cgrp doesn't have @subsys_id enabled.
T
Tejun Heo 已提交
189 190
 */
static struct cgroup_subsys_state *cgroup_css(struct cgroup *cgrp,
191
					      struct cgroup_subsys *ss)
T
Tejun Heo 已提交
192
{
193
	if (ss)
194
		return rcu_dereference_check(cgrp->subsys[ss->id],
T
Tejun Heo 已提交
195 196
					lockdep_is_held(&cgroup_tree_mutex) ||
					lockdep_is_held(&cgroup_mutex));
197 198
	else
		return &cgrp->dummy_css;
T
Tejun Heo 已提交
199
}
200

201
/* convenient tests for these bits */
202
static inline bool cgroup_is_dead(const struct cgroup *cgrp)
203
{
204
	return test_bit(CGRP_DEAD, &cgrp->flags);
205 206
}

207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
/**
 * cgroup_is_descendant - test ancestry
 * @cgrp: the cgroup to be tested
 * @ancestor: possible ancestor of @cgrp
 *
 * Test whether @cgrp is a descendant of @ancestor.  It also returns %true
 * if @cgrp == @ancestor.  This function is safe to call as long as @cgrp
 * and @ancestor are accessible.
 */
bool cgroup_is_descendant(struct cgroup *cgrp, struct cgroup *ancestor)
{
	while (cgrp) {
		if (cgrp == ancestor)
			return true;
		cgrp = cgrp->parent;
	}
	return false;
}
EXPORT_SYMBOL_GPL(cgroup_is_descendant);
226

227
static int cgroup_is_releasable(const struct cgroup *cgrp)
228 229
{
	const int bits =
230 231 232
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
233 234
}

235
static int notify_on_release(const struct cgroup *cgrp)
236
{
237
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
238 239
}

T
Tejun Heo 已提交
240 241 242 243 244 245 246 247 248 249 250 251
/**
 * for_each_css - iterate all css's of a cgroup
 * @css: the iteration cursor
 * @ssid: the index of the subsystem, CGROUP_SUBSYS_COUNT after reaching the end
 * @cgrp: the target cgroup to iterate css's of
 *
 * Should be called under cgroup_mutex.
 */
#define for_each_css(css, ssid, cgrp)					\
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT; (ssid)++)	\
		if (!((css) = rcu_dereference_check(			\
				(cgrp)->subsys[(ssid)],			\
T
Tejun Heo 已提交
252
				lockdep_is_held(&cgroup_tree_mutex) ||	\
T
Tejun Heo 已提交
253 254 255
				lockdep_is_held(&cgroup_mutex)))) { }	\
		else

256
/**
T
Tejun Heo 已提交
257
 * for_each_subsys - iterate all enabled cgroup subsystems
258
 * @ss: the iteration cursor
259
 * @ssid: the index of @ss, CGROUP_SUBSYS_COUNT after reaching the end
260
 */
261
#define for_each_subsys(ss, ssid)					\
T
Tejun Heo 已提交
262 263
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT &&		\
	     (((ss) = cgroup_subsys[ssid]) || true); (ssid)++)
264

265 266 267
/* iterate across the active hierarchies */
#define for_each_active_root(root)					\
	list_for_each_entry((root), &cgroup_roots, root_list)
268

269 270 271 272 273
static inline struct cgroup *__d_cgrp(struct dentry *dentry)
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
274
static inline struct cfent *__d_cfe(struct dentry *dentry)
275 276 277 278
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
279 280 281 282 283
static inline struct cftype *__d_cft(struct dentry *dentry)
{
	return __d_cfe(dentry)->type;
}

284 285 286 287
/**
 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive.
 * @cgrp: the cgroup to be checked for liveness
 *
T
Tejun Heo 已提交
288 289
 * On success, returns true; the mutex should be later unlocked.  On
 * failure returns false with no lock held.
290
 */
291
static bool cgroup_lock_live_group(struct cgroup *cgrp)
292 293
{
	mutex_lock(&cgroup_mutex);
294
	if (cgroup_is_dead(cgrp)) {
295 296 297 298 299 300
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}

301 302 303
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
304
static DEFINE_RAW_SPINLOCK(release_list_lock);
305 306
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
307
static void check_for_release(struct cgroup *cgrp);
308

309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
/*
 * A cgroup can be associated with multiple css_sets as different tasks may
 * belong to different cgroups on different hierarchies.  In the other
 * direction, a css_set is naturally associated with multiple cgroups.
 * This M:N relationship is represented by the following link structure
 * which exists for each association and allows traversing the associations
 * from both sides.
 */
struct cgrp_cset_link {
	/* the cgroup and css_set this link associates */
	struct cgroup		*cgrp;
	struct css_set		*cset;

	/* list of cgrp_cset_links anchored at cgrp->cset_links */
	struct list_head	cset_link;

	/* list of cgrp_cset_links anchored at css_set->cgrp_links */
	struct list_head	cgrp_link;
327 328 329 330 331 332 333 334 335 336
};

/* The default css_set - used by init and its children prior to any
 * hierarchies being mounted. It contains a pointer to the root state
 * for each subsystem. Also used to anchor the list of css_sets. Not
 * reference-counted, to improve performance when child cgroups
 * haven't been created.
 */

static struct css_set init_css_set;
337
static struct cgrp_cset_link init_cgrp_cset_link;
338

339 340 341
/*
 * css_set_lock protects the list of css_set objects, and the chain of
 * tasks off each css_set.  Nests outside task->alloc_lock due to
342
 * css_task_iter_start().
343
 */
344 345 346
static DEFINE_RWLOCK(css_set_lock);
static int css_set_count;

347 348 349 350 351
/*
 * hash table for cgroup groups. This improves the performance to find
 * an existing css_set. This hash doesn't (currently) take into
 * account cgroups in empty hierarchies.
 */
352
#define CSS_SET_HASH_BITS	7
353
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
354

355
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
356
{
357
	unsigned long key = 0UL;
358 359
	struct cgroup_subsys *ss;
	int i;
360

361
	for_each_subsys(ss, i)
362 363
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
364

365
	return key;
366 367
}

368 369
/*
 * We don't maintain the lists running through each css_set to its task
370 371 372
 * until after the first call to css_task_iter_start().  This reduces the
 * fork()/exit() overhead for people who have cgroups compiled into their
 * kernel but not actually in use.
373
 */
374
static int use_task_css_set_links __read_mostly;
375

376
static void __put_css_set(struct css_set *cset, int taskexit)
377
{
378
	struct cgrp_cset_link *link, *tmp_link;
379

380 381 382 383 384
	/*
	 * Ensure that the refcount doesn't hit zero while any readers
	 * can see it. Similar to atomic_dec_and_lock(), but for an
	 * rwlock
	 */
385
	if (atomic_add_unless(&cset->refcount, -1, 1))
386 387
		return;
	write_lock(&css_set_lock);
388
	if (!atomic_dec_and_test(&cset->refcount)) {
389 390 391
		write_unlock(&css_set_lock);
		return;
	}
392

393
	/* This css_set is dead. unlink it and release cgroup refcounts */
394
	hash_del(&cset->hlist);
395 396
	css_set_count--;

397
	list_for_each_entry_safe(link, tmp_link, &cset->cgrp_links, cgrp_link) {
398
		struct cgroup *cgrp = link->cgrp;
399

400 401
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
L
Li Zefan 已提交
402

403
		/* @cgrp can't go away while we're holding css_set_lock */
T
Tejun Heo 已提交
404
		if (list_empty(&cgrp->cset_links) && notify_on_release(cgrp)) {
405
			if (taskexit)
406 407
				set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
408
		}
409 410

		kfree(link);
411
	}
412 413

	write_unlock(&css_set_lock);
414
	kfree_rcu(cset, rcu_head);
415 416
}

417 418 419
/*
 * refcounted get/put for css_set objects
 */
420
static inline void get_css_set(struct css_set *cset)
421
{
422
	atomic_inc(&cset->refcount);
423 424
}

425
static inline void put_css_set(struct css_set *cset)
426
{
427
	__put_css_set(cset, 0);
428 429
}

430
static inline void put_css_set_taskexit(struct css_set *cset)
431
{
432
	__put_css_set(cset, 1);
433 434
}

435
/**
436
 * compare_css_sets - helper function for find_existing_css_set().
437 438
 * @cset: candidate css_set being tested
 * @old_cset: existing css_set for a task
439 440 441
 * @new_cgrp: cgroup that's being entered by the task
 * @template: desired set of css pointers in css_set (pre-calculated)
 *
L
Li Zefan 已提交
442
 * Returns true if "cset" matches "old_cset" except for the hierarchy
443 444
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
445 446
static bool compare_css_sets(struct css_set *cset,
			     struct css_set *old_cset,
447 448 449 450 451
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

452
	if (memcmp(template, cset->subsys, sizeof(cset->subsys))) {
453 454 455 456 457 458 459 460 461 462 463 464 465
		/* Not all subsystems matched */
		return false;
	}

	/*
	 * Compare cgroup pointers in order to distinguish between
	 * different cgroups in heirarchies with no subsystems. We
	 * could get by with just this check alone (and skip the
	 * memcmp above) but on most setups the memcmp check will
	 * avoid the need for this more expensive check on almost all
	 * candidates.
	 */

466 467
	l1 = &cset->cgrp_links;
	l2 = &old_cset->cgrp_links;
468
	while (1) {
469
		struct cgrp_cset_link *link1, *link2;
470
		struct cgroup *cgrp1, *cgrp2;
471 472 473 474

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
475 476
		if (l1 == &cset->cgrp_links) {
			BUG_ON(l2 != &old_cset->cgrp_links);
477 478
			break;
		} else {
479
			BUG_ON(l2 == &old_cset->cgrp_links);
480 481
		}
		/* Locate the cgroups associated with these links. */
482 483 484 485
		link1 = list_entry(l1, struct cgrp_cset_link, cgrp_link);
		link2 = list_entry(l2, struct cgrp_cset_link, cgrp_link);
		cgrp1 = link1->cgrp;
		cgrp2 = link2->cgrp;
486
		/* Hierarchies should be linked in the same order. */
487
		BUG_ON(cgrp1->root != cgrp2->root);
488 489 490 491 492 493 494 495

		/*
		 * If this hierarchy is the hierarchy of the cgroup
		 * that's changing, then we need to check that this
		 * css_set points to the new cgroup; if it's any other
		 * hierarchy, then this css_set should point to the
		 * same cgroup as the old css_set.
		 */
496 497
		if (cgrp1->root == new_cgrp->root) {
			if (cgrp1 != new_cgrp)
498 499
				return false;
		} else {
500
			if (cgrp1 != cgrp2)
501 502 503 504 505 506
				return false;
		}
	}
	return true;
}

507 508 509 510 511
/**
 * find_existing_css_set - init css array and find the matching css_set
 * @old_cset: the css_set that we're using before the cgroup transition
 * @cgrp: the cgroup that we're moving into
 * @template: out param for the new set of csses, should be clear on entry
512
 */
513 514 515
static struct css_set *find_existing_css_set(struct css_set *old_cset,
					struct cgroup *cgrp,
					struct cgroup_subsys_state *template[])
516
{
517
	struct cgroupfs_root *root = cgrp->root;
518
	struct cgroup_subsys *ss;
519
	struct css_set *cset;
520
	unsigned long key;
521
	int i;
522

B
Ben Blum 已提交
523 524 525 526 527
	/*
	 * Build the set of subsystem state objects that we want to see in the
	 * new css_set. while subsystems can change globally, the entries here
	 * won't change, so no need for locking.
	 */
528
	for_each_subsys(ss, i) {
529
		if (root->subsys_mask & (1UL << i)) {
530 531 532
			/* Subsystem is in this hierarchy. So we want
			 * the subsystem state from the new
			 * cgroup */
533
			template[i] = cgroup_css(cgrp, ss);
534 535 536
		} else {
			/* Subsystem is not in this hierarchy, so we
			 * don't want to change the subsystem state */
537
			template[i] = old_cset->subsys[i];
538 539 540
		}
	}

541
	key = css_set_hash(template);
542 543
	hash_for_each_possible(css_set_table, cset, hlist, key) {
		if (!compare_css_sets(cset, old_cset, cgrp, template))
544 545 546
			continue;

		/* This css_set matches what we need */
547
		return cset;
548
	}
549 550 551 552 553

	/* No existing cgroup group matched */
	return NULL;
}

554
static void free_cgrp_cset_links(struct list_head *links_to_free)
555
{
556
	struct cgrp_cset_link *link, *tmp_link;
557

558 559
	list_for_each_entry_safe(link, tmp_link, links_to_free, cset_link) {
		list_del(&link->cset_link);
560 561 562 563
		kfree(link);
	}
}

564 565 566 567 568 569 570
/**
 * allocate_cgrp_cset_links - allocate cgrp_cset_links
 * @count: the number of links to allocate
 * @tmp_links: list_head the allocated links are put on
 *
 * Allocate @count cgrp_cset_link structures and chain them on @tmp_links
 * through ->cset_link.  Returns 0 on success or -errno.
571
 */
572
static int allocate_cgrp_cset_links(int count, struct list_head *tmp_links)
573
{
574
	struct cgrp_cset_link *link;
575
	int i;
576 577 578

	INIT_LIST_HEAD(tmp_links);

579
	for (i = 0; i < count; i++) {
580
		link = kzalloc(sizeof(*link), GFP_KERNEL);
581
		if (!link) {
582
			free_cgrp_cset_links(tmp_links);
583 584
			return -ENOMEM;
		}
585
		list_add(&link->cset_link, tmp_links);
586 587 588 589
	}
	return 0;
}

590 591
/**
 * link_css_set - a helper function to link a css_set to a cgroup
592
 * @tmp_links: cgrp_cset_link objects allocated by allocate_cgrp_cset_links()
593
 * @cset: the css_set to be linked
594 595
 * @cgrp: the destination cgroup
 */
596 597
static void link_css_set(struct list_head *tmp_links, struct css_set *cset,
			 struct cgroup *cgrp)
598
{
599
	struct cgrp_cset_link *link;
600

601 602 603
	BUG_ON(list_empty(tmp_links));
	link = list_first_entry(tmp_links, struct cgrp_cset_link, cset_link);
	link->cset = cset;
604
	link->cgrp = cgrp;
605
	list_move(&link->cset_link, &cgrp->cset_links);
606 607 608 609
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
610
	list_add_tail(&link->cgrp_link, &cset->cgrp_links);
611 612
}

613 614 615 616 617 618 619
/**
 * find_css_set - return a new css_set with one cgroup updated
 * @old_cset: the baseline css_set
 * @cgrp: the cgroup to be updated
 *
 * Return a new css_set that's equivalent to @old_cset, but with @cgrp
 * substituted into the appropriate hierarchy.
620
 */
621 622
static struct css_set *find_css_set(struct css_set *old_cset,
				    struct cgroup *cgrp)
623
{
624
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT] = { };
625
	struct css_set *cset;
626 627
	struct list_head tmp_links;
	struct cgrp_cset_link *link;
628
	unsigned long key;
629

630 631
	lockdep_assert_held(&cgroup_mutex);

632 633
	/* First see if we already have a cgroup group that matches
	 * the desired set */
634
	read_lock(&css_set_lock);
635 636 637
	cset = find_existing_css_set(old_cset, cgrp, template);
	if (cset)
		get_css_set(cset);
638
	read_unlock(&css_set_lock);
639

640 641
	if (cset)
		return cset;
642

643
	cset = kzalloc(sizeof(*cset), GFP_KERNEL);
644
	if (!cset)
645 646
		return NULL;

647
	/* Allocate all the cgrp_cset_link objects that we'll need */
648
	if (allocate_cgrp_cset_links(cgroup_root_count, &tmp_links) < 0) {
649
		kfree(cset);
650 651 652
		return NULL;
	}

653
	atomic_set(&cset->refcount, 1);
654
	INIT_LIST_HEAD(&cset->cgrp_links);
655 656
	INIT_LIST_HEAD(&cset->tasks);
	INIT_HLIST_NODE(&cset->hlist);
657 658 659

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
660
	memcpy(cset->subsys, template, sizeof(cset->subsys));
661 662 663

	write_lock(&css_set_lock);
	/* Add reference counts and links from the new css_set. */
664
	list_for_each_entry(link, &old_cset->cgrp_links, cgrp_link) {
665
		struct cgroup *c = link->cgrp;
666

667 668
		if (c->root == cgrp->root)
			c = cgrp;
669
		link_css_set(&tmp_links, cset, c);
670
	}
671

672
	BUG_ON(!list_empty(&tmp_links));
673 674

	css_set_count++;
675 676

	/* Add this cgroup group to the hash table */
677 678
	key = css_set_hash(cset->subsys);
	hash_add(css_set_table, &cset->hlist, key);
679

680 681
	write_unlock(&css_set_lock);

682
	return cset;
683 684
}

685 686 687 688 689 690 691
/*
 * Return the cgroup for "task" from the given hierarchy. Must be
 * called with cgroup_mutex held.
 */
static struct cgroup *task_cgroup_from_root(struct task_struct *task,
					    struct cgroupfs_root *root)
{
692
	struct css_set *cset;
693 694 695 696 697 698 699 700 701
	struct cgroup *res = NULL;

	BUG_ON(!mutex_is_locked(&cgroup_mutex));
	read_lock(&css_set_lock);
	/*
	 * No need to lock the task - since we hold cgroup_mutex the
	 * task can't change groups, so the only thing that can happen
	 * is that it exits and its css is set back to init_css_set.
	 */
702
	cset = task_css_set(task);
703
	if (cset == &init_css_set) {
704 705
		res = &root->top_cgroup;
	} else {
706 707 708
		struct cgrp_cset_link *link;

		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
709
			struct cgroup *c = link->cgrp;
710

711 712 713 714 715 716 717 718 719 720 721
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
	read_unlock(&css_set_lock);
	BUG_ON(!res);
	return res;
}

722 723 724 725 726 727 728 729 730 731
/*
 * There is one global cgroup mutex. We also require taking
 * task_lock() when dereferencing a task's cgroup subsys pointers.
 * See "The task_lock() exception", at the end of this comment.
 *
 * A task must hold cgroup_mutex to modify cgroups.
 *
 * Any task can increment and decrement the count field without lock.
 * So in general, code holding cgroup_mutex can't rely on the count
 * field not changing.  However, if the count goes to zero, then only
732
 * cgroup_attach_task() can increment it again.  Because a count of zero
733 734 735 736 737 738 739 740 741 742 743 744 745
 * means that no tasks are currently attached, therefore there is no
 * way a task attached to that cgroup can fork (the other way to
 * increment the count).  So code holding cgroup_mutex can safely
 * assume that if the count is zero, it will stay zero. Similarly, if
 * a task holds cgroup_mutex on a cgroup with zero count, it
 * knows that the cgroup won't be removed, as cgroup_rmdir()
 * needs that mutex.
 *
 * The fork and exit callbacks cgroup_fork() and cgroup_exit(), don't
 * (usually) take cgroup_mutex.  These are the two most performance
 * critical pieces of code here.  The exception occurs on cgroup_exit(),
 * when a task in a notify_on_release cgroup exits.  Then cgroup_mutex
 * is taken, and if the cgroup count is zero, a usermode call made
L
Li Zefan 已提交
746 747
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
748 749 750 751 752 753 754 755 756 757 758
 *
 * A cgroup can only be deleted if both its 'count' of using tasks
 * is zero, and its list of 'children' cgroups is empty.  Since all
 * tasks in the system use _some_ cgroup, and since there is always at
 * least one task in the system (init, pid == 1), therefore, top_cgroup
 * always has either children cgroups and/or using tasks.  So we don't
 * need a special hack to ensure that top_cgroup cannot be deleted.
 *
 *	The task_lock() exception
 *
 * The need for this exception arises from the action of
759
 * cgroup_attach_task(), which overwrites one task's cgroup pointer with
L
Li Zefan 已提交
760
 * another.  It does so using cgroup_mutex, however there are
761 762 763
 * several performance critical places that need to reference
 * task->cgroup without the expense of grabbing a system global
 * mutex.  Therefore except as noted below, when dereferencing or, as
764
 * in cgroup_attach_task(), modifying a task's cgroup pointer we use
765 766 767 768
 * task_lock(), which acts on a spinlock (task->alloc_lock) already in
 * the task_struct routinely used for such matters.
 *
 * P.S.  One more locking exception.  RCU is used to guard the
769
 * update of a tasks cgroup pointer by cgroup_attach_task()
770 771 772 773 774 775 776 777 778
 */

/*
 * A couple of forward declarations required, due to cyclic reference loop:
 * cgroup_mkdir -> cgroup_create -> cgroup_populate_dir ->
 * cgroup_add_file -> cgroup_create_file -> cgroup_dir_inode_operations
 * -> cgroup_mkdir.
 */

779
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
780
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
781
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned long subsys_mask);
782
static const struct inode_operations cgroup_dir_inode_operations;
783
static const struct file_operations proc_cgroupstats_operations;
784 785

static struct backing_dev_info cgroup_backing_dev_info = {
786
	.name		= "cgroup",
787
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
788
};
789

A
Al Viro 已提交
790
static struct inode *cgroup_new_inode(umode_t mode, struct super_block *sb)
791 792 793 794
{
	struct inode *inode = new_inode(sb);

	if (inode) {
795
		inode->i_ino = get_next_ino();
796
		inode->i_mode = mode;
797 798
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
799 800 801 802 803 804
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inode->i_mapping->backing_dev_info = &cgroup_backing_dev_info;
	}
	return inode;
}

T
Tejun Heo 已提交
805
static struct cgroup_name *cgroup_alloc_name(const char *name_str)
806 807 808
{
	struct cgroup_name *name;

T
Tejun Heo 已提交
809
	name = kmalloc(sizeof(*name) + strlen(name_str) + 1, GFP_KERNEL);
810 811
	if (!name)
		return NULL;
T
Tejun Heo 已提交
812
	strcpy(name->name, name_str);
813 814 815
	return name;
}

T
Tejun Heo 已提交
816 817 818 819 820 821 822 823 824 825 826 827
static char *cgroup_file_name(struct cgroup *cgrp, const struct cftype *cft,
			      char *buf)
{
	if (cft->ss && !(cft->flags & CFTYPE_NO_PREFIX) &&
	    !(cgrp->root->flags & CGRP_ROOT_NOPREFIX))
		snprintf(buf, CGROUP_FILE_NAME_MAX, "%s.%s",
			 cft->ss->name, cft->name);
	else
		strncpy(buf, cft->name, CGROUP_FILE_NAME_MAX);
	return buf;
}

828 829
static void cgroup_free_fn(struct work_struct *work)
{
830
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
831 832 833 834 835

	mutex_lock(&cgroup_mutex);
	cgrp->root->number_of_cgroups--;
	mutex_unlock(&cgroup_mutex);

836 837 838 839 840 841 842
	/*
	 * We get a ref to the parent's dentry, and put the ref when
	 * this cgroup is being freed, so it's guaranteed that the
	 * parent won't be destroyed before its children.
	 */
	dput(cgrp->parent->dentry);

843 844
	/*
	 * Drop the active superblock reference that we took when we
845 846
	 * created the cgroup. This will free cgrp->root, if we are
	 * holding the last reference to @sb.
847 848 849
	 */
	deactivate_super(cgrp->root->sb);

850
	cgroup_pidlist_destroy_all(cgrp);
851 852 853

	simple_xattrs_free(&cgrp->xattrs);

854
	kfree(rcu_dereference_raw(cgrp->name));
855 856 857 858 859 860 861
	kfree(cgrp);
}

static void cgroup_free_rcu(struct rcu_head *head)
{
	struct cgroup *cgrp = container_of(head, struct cgroup, rcu_head);

862
	INIT_WORK(&cgrp->destroy_work, cgroup_free_fn);
863
	queue_work(cgroup_destroy_wq, &cgrp->destroy_work);
864 865
}

866 867 868 869
static void cgroup_diput(struct dentry *dentry, struct inode *inode)
{
	/* is dentry a directory ? if so, kfree() associated cgroup */
	if (S_ISDIR(inode->i_mode)) {
870
		struct cgroup *cgrp = dentry->d_fsdata;
871

872
		BUG_ON(!(cgroup_is_dead(cgrp)));
873 874 875 876 877 878 879

		/*
		 * XXX: cgrp->id is only used to look up css's.  As cgroup
		 * and css's lifetimes will be decoupled, it should be made
		 * per-subsystem and moved to css->id so that lookups are
		 * successful until the target css is released.
		 */
880
		mutex_lock(&cgroup_mutex);
881
		idr_remove(&cgrp->root->cgroup_idr, cgrp->id);
882
		mutex_unlock(&cgroup_mutex);
883 884
		cgrp->id = -1;

885
		call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
T
Tejun Heo 已提交
886 887 888 889 890 891 892
	} else {
		struct cfent *cfe = __d_cfe(dentry);
		struct cgroup *cgrp = dentry->d_parent->d_fsdata;

		WARN_ONCE(!list_empty(&cfe->node) &&
			  cgrp != &cgrp->root->top_cgroup,
			  "cfe still linked for %s\n", cfe->type->name);
L
Li Zefan 已提交
893
		simple_xattrs_free(&cfe->xattrs);
T
Tejun Heo 已提交
894
		kfree(cfe);
895 896 897 898 899 900 901 902 903 904 905 906 907
	}
	iput(inode);
}

static void remove_dir(struct dentry *d)
{
	struct dentry *parent = dget(d->d_parent);

	d_delete(d);
	simple_rmdir(parent->d_inode, d);
	dput(parent);
}

908
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
909 910 911 912
{
	struct cfent *cfe;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
T
Tejun Heo 已提交
913
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
914

915 916 917 918
	/*
	 * If we're doing cleanup due to failure of cgroup_create(),
	 * the corresponding @cfe may not exist.
	 */
T
Tejun Heo 已提交
919 920 921 922 923 924 925 926
	list_for_each_entry(cfe, &cgrp->files, node) {
		struct dentry *d = cfe->dentry;

		if (cft && cfe->type != cft)
			continue;

		dget(d);
		d_delete(d);
927
		simple_unlink(cgrp->dentry->d_inode, d);
T
Tejun Heo 已提交
928 929 930
		list_del_init(&cfe->node);
		dput(d);

931
		break;
932
	}
T
Tejun Heo 已提交
933 934
}

935
/**
936
 * cgroup_clear_dir - remove subsys files in a cgroup directory
937
 * @cgrp: target cgroup
938 939
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
940
static void cgroup_clear_dir(struct cgroup *cgrp, unsigned long subsys_mask)
T
Tejun Heo 已提交
941
{
942
	struct cgroup_subsys *ss;
943
	int i;
T
Tejun Heo 已提交
944

945
	for_each_subsys(ss, i) {
946
		struct cftype_set *set;
947 948

		if (!test_bit(i, &subsys_mask))
949 950
			continue;
		list_for_each_entry(set, &ss->cftsets, node)
951
			cgroup_addrm_files(cgrp, set->cfts, false);
952
	}
953 954 955 956 957 958 959
}

/*
 * NOTE : the dentry must have been dget()'ed
 */
static void cgroup_d_remove_dir(struct dentry *dentry)
{
N
Nick Piggin 已提交
960
	struct dentry *parent;
961

N
Nick Piggin 已提交
962 963
	parent = dentry->d_parent;
	spin_lock(&parent->d_lock);
964
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
965
	list_del_init(&dentry->d_u.d_child);
N
Nick Piggin 已提交
966 967
	spin_unlock(&dentry->d_lock);
	spin_unlock(&parent->d_lock);
968 969 970 971
	remove_dir(dentry);
}

static int rebind_subsystems(struct cgroupfs_root *root,
972
			     unsigned long added_mask, unsigned removed_mask)
973
{
974
	struct cgroup *cgrp = &root->top_cgroup;
975
	struct cgroup_subsys *ss;
976
	int i, ret;
977

T
Tejun Heo 已提交
978 979
	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
B
Ben Blum 已提交
980

981
	/* Check that any added subsystems are currently free */
T
Tejun Heo 已提交
982 983 984
	for_each_subsys(ss, i)
		if ((added_mask & (1 << i)) && ss->root != &cgroup_dummy_root)
			return -EBUSY;
985

986 987
	ret = cgroup_populate_dir(cgrp, added_mask);
	if (ret)
T
Tejun Heo 已提交
988
		return ret;
989 990 991 992 993

	/*
	 * Nothing can fail from this point on.  Remove files for the
	 * removed subsystems and rebind each subsystem.
	 */
994
	mutex_unlock(&cgroup_mutex);
995
	cgroup_clear_dir(cgrp, removed_mask);
996
	mutex_lock(&cgroup_mutex);
997

998
	for_each_subsys(ss, i) {
999
		unsigned long bit = 1UL << i;
1000

1001
		if (bit & added_mask) {
1002
			/* We're binding this subsystem to this hierarchy */
1003 1004 1005
			BUG_ON(cgroup_css(cgrp, ss));
			BUG_ON(!cgroup_css(cgroup_dummy_top, ss));
			BUG_ON(cgroup_css(cgroup_dummy_top, ss)->cgroup != cgroup_dummy_top);
1006

1007
			rcu_assign_pointer(cgrp->subsys[i],
1008 1009
					   cgroup_css(cgroup_dummy_top, ss));
			cgroup_css(cgrp, ss)->cgroup = cgrp;
1010

1011
			ss->root = root;
1012
			if (ss->bind)
1013
				ss->bind(cgroup_css(cgrp, ss));
1014

B
Ben Blum 已提交
1015
			/* refcount was already taken, and we're keeping it */
1016
			root->subsys_mask |= bit;
1017
		} else if (bit & removed_mask) {
1018
			/* We're removing this subsystem */
1019 1020
			BUG_ON(cgroup_css(cgrp, ss) != cgroup_css(cgroup_dummy_top, ss));
			BUG_ON(cgroup_css(cgrp, ss)->cgroup != cgrp);
1021

1022
			if (ss->bind)
1023
				ss->bind(cgroup_css(cgroup_dummy_top, ss));
1024

1025
			cgroup_css(cgroup_dummy_top, ss)->cgroup = cgroup_dummy_top;
1026 1027
			RCU_INIT_POINTER(cgrp->subsys[i], NULL);

1028
			cgroup_subsys[i]->root = &cgroup_dummy_root;
1029
			root->subsys_mask &= ~bit;
1030 1031 1032
		}
	}

1033 1034 1035 1036 1037 1038
	/*
	 * Mark @root has finished binding subsystems.  @root->subsys_mask
	 * now matches the bound subsystems.
	 */
	root->flags |= CGRP_ROOT_SUBSYS_BOUND;

1039 1040 1041
	return 0;
}

1042
static int cgroup_show_options(struct seq_file *seq, struct dentry *dentry)
1043
{
1044
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
1045
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
1046
	int ssid;
1047

T
Tejun Heo 已提交
1048 1049 1050
	for_each_subsys(ss, ssid)
		if (root->subsys_mask & (1 << ssid))
			seq_printf(seq, ",%s", ss->name);
1051 1052
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR)
		seq_puts(seq, ",sane_behavior");
1053
	if (root->flags & CGRP_ROOT_NOPREFIX)
1054
		seq_puts(seq, ",noprefix");
1055
	if (root->flags & CGRP_ROOT_XATTR)
A
Aristeu Rozanski 已提交
1056
		seq_puts(seq, ",xattr");
1057 1058

	spin_lock(&release_agent_path_lock);
1059 1060
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1061 1062
	spin_unlock(&release_agent_path_lock);

1063
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags))
1064
		seq_puts(seq, ",clone_children");
1065 1066
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
1067 1068 1069 1070
	return 0;
}

struct cgroup_sb_opts {
1071
	unsigned long subsys_mask;
1072
	unsigned long flags;
1073
	char *release_agent;
1074
	bool cpuset_clone_children;
1075
	char *name;
1076 1077
	/* User explicitly requested empty subsystem */
	bool none;
1078 1079

	struct cgroupfs_root *new_root;
1080

1081 1082
};

B
Ben Blum 已提交
1083
/*
1084 1085 1086 1087
 * Convert a hierarchy specifier into a bitmask of subsystems and
 * flags. Call with cgroup_mutex held to protect the cgroup_subsys[]
 * array. This function takes refcounts on subsystems to be used, unless it
 * returns error, in which case no refcounts are taken.
B
Ben Blum 已提交
1088
 */
B
Ben Blum 已提交
1089
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1090
{
1091 1092
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1093
	unsigned long mask = (unsigned long)-1;
1094 1095
	struct cgroup_subsys *ss;
	int i;
1096

B
Ben Blum 已提交
1097 1098
	BUG_ON(!mutex_is_locked(&cgroup_mutex));

1099
#ifdef CONFIG_CPUSETS
1100
	mask = ~(1UL << cpuset_cgrp_id);
1101
#endif
1102

1103
	memset(opts, 0, sizeof(*opts));
1104 1105 1106 1107

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1108
		if (!strcmp(token, "none")) {
1109 1110
			/* Explicitly have no subsystems */
			opts->none = true;
1111 1112 1113 1114 1115 1116 1117 1118 1119
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
1120 1121 1122 1123
		if (!strcmp(token, "__DEVEL__sane_behavior")) {
			opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
			continue;
		}
1124
		if (!strcmp(token, "noprefix")) {
1125
			opts->flags |= CGRP_ROOT_NOPREFIX;
1126 1127 1128
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1129
			opts->cpuset_clone_children = true;
1130 1131
			continue;
		}
A
Aristeu Rozanski 已提交
1132
		if (!strcmp(token, "xattr")) {
1133
			opts->flags |= CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
1134 1135
			continue;
		}
1136
		if (!strncmp(token, "release_agent=", 14)) {
1137 1138 1139
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1140
			opts->release_agent =
1141
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1142 1143
			if (!opts->release_agent)
				return -ENOMEM;
1144 1145 1146
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
			const char *name = token + 5;
			/* Can't specify an empty name */
			if (!strlen(name))
				return -EINVAL;
			/* Must match [\w.-]+ */
			for (i = 0; i < strlen(name); i++) {
				char c = name[i];
				if (isalnum(c))
					continue;
				if ((c == '.') || (c == '-') || (c == '_'))
					continue;
				return -EINVAL;
			}
			/* Specifying two names is forbidden */
			if (opts->name)
				return -EINVAL;
			opts->name = kstrndup(name,
1164
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1165 1166 1167
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1168 1169 1170 1171

			continue;
		}

1172
		for_each_subsys(ss, i) {
1173 1174 1175 1176 1177 1178 1179 1180
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1181
			set_bit(i, &opts->subsys_mask);
1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
			one_ss = true;

			break;
		}
		if (i == CGROUP_SUBSYS_COUNT)
			return -ENOENT;
	}

	/*
	 * If the 'all' option was specified select all the subsystems,
1192 1193
	 * otherwise if 'none', 'name=' and a subsystem name options
	 * were not specified, let's default to 'all'
1194
	 */
1195 1196 1197 1198
	if (all_ss || (!one_ss && !opts->none && !opts->name))
		for_each_subsys(ss, i)
			if (!ss->disabled)
				set_bit(i, &opts->subsys_mask);
1199

1200 1201
	/* Consistency checks */

1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
	if (opts->flags & CGRP_ROOT_SANE_BEHAVIOR) {
		pr_warning("cgroup: sane_behavior: this is still under development and its behaviors will change, proceed at your own risk\n");

		if (opts->flags & CGRP_ROOT_NOPREFIX) {
			pr_err("cgroup: sane_behavior: noprefix is not allowed\n");
			return -EINVAL;
		}

		if (opts->cpuset_clone_children) {
			pr_err("cgroup: sane_behavior: clone_children is not allowed\n");
			return -EINVAL;
		}
	}

1216 1217 1218 1219 1220
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
1221
	if ((opts->flags & CGRP_ROOT_NOPREFIX) && (opts->subsys_mask & mask))
1222 1223
		return -EINVAL;

1224 1225

	/* Can't specify "none" and some subsystems */
1226
	if (opts->subsys_mask && opts->none)
1227 1228 1229 1230 1231 1232
		return -EINVAL;

	/*
	 * We either have to specify by name or by subsystems. (So all
	 * empty hierarchies must have a name).
	 */
1233
	if (!opts->subsys_mask && !opts->name)
1234 1235 1236 1237 1238 1239 1240 1241 1242
		return -EINVAL;

	return 0;
}

static int cgroup_remount(struct super_block *sb, int *flags, char *data)
{
	int ret = 0;
	struct cgroupfs_root *root = sb->s_fs_info;
1243
	struct cgroup *cgrp = &root->top_cgroup;
1244
	struct cgroup_sb_opts opts;
1245
	unsigned long added_mask, removed_mask;
1246

1247 1248 1249 1250 1251
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR) {
		pr_err("cgroup: sane_behavior: remount is not allowed\n");
		return -EINVAL;
	}

1252
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
T
Tejun Heo 已提交
1253
	mutex_lock(&cgroup_tree_mutex);
1254 1255 1256 1257 1258 1259 1260
	mutex_lock(&cgroup_mutex);

	/* See what subsystems are wanted */
	ret = parse_cgroupfs_options(data, &opts);
	if (ret)
		goto out_unlock;

1261
	if (opts.subsys_mask != root->subsys_mask || opts.release_agent)
1262 1263 1264
		pr_warning("cgroup: option changes via remount are deprecated (pid=%d comm=%s)\n",
			   task_tgid_nr(current), current->comm);

1265 1266
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1267

B
Ben Blum 已提交
1268
	/* Don't allow flags or name to change at remount */
1269
	if (((opts.flags ^ root->flags) & CGRP_ROOT_OPTION_MASK) ||
B
Ben Blum 已提交
1270
	    (opts.name && strcmp(opts.name, root->name))) {
1271 1272 1273
		pr_err("cgroup: option or name mismatch, new: 0x%lx \"%s\", old: 0x%lx \"%s\"\n",
		       opts.flags & CGRP_ROOT_OPTION_MASK, opts.name ?: "",
		       root->flags & CGRP_ROOT_OPTION_MASK, root->name);
1274 1275 1276 1277
		ret = -EINVAL;
		goto out_unlock;
	}

1278 1279 1280
	/* remounting is not allowed for populated hierarchies */
	if (root->number_of_cgroups > 1) {
		ret = -EBUSY;
1281
		goto out_unlock;
B
Ben Blum 已提交
1282
	}
1283

1284
	ret = rebind_subsystems(root, added_mask, removed_mask);
1285
	if (ret)
1286
		goto out_unlock;
1287

1288 1289
	if (opts.release_agent) {
		spin_lock(&release_agent_path_lock);
1290
		strcpy(root->release_agent_path, opts.release_agent);
1291 1292
		spin_unlock(&release_agent_path_lock);
	}
1293
 out_unlock:
1294
	kfree(opts.release_agent);
1295
	kfree(opts.name);
1296
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1297
	mutex_unlock(&cgroup_tree_mutex);
1298
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1299 1300 1301
	return ret;
}

1302
static const struct super_operations cgroup_ops = {
1303 1304 1305 1306 1307 1308
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
	.show_options = cgroup_show_options,
	.remount_fs = cgroup_remount,
};

1309 1310 1311 1312
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
T
Tejun Heo 已提交
1313
	INIT_LIST_HEAD(&cgrp->files);
1314
	INIT_LIST_HEAD(&cgrp->cset_links);
1315
	INIT_LIST_HEAD(&cgrp->release_list);
1316 1317
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
T
Tejun Heo 已提交
1318
	cgrp->dummy_css.cgroup = cgrp;
A
Aristeu Rozanski 已提交
1319
	simple_xattrs_init(&cgrp->xattrs);
1320
}
1321

1322 1323
static void init_cgroup_root(struct cgroupfs_root *root)
{
1324
	struct cgroup *cgrp = &root->top_cgroup;
1325

1326 1327
	INIT_LIST_HEAD(&root->root_list);
	root->number_of_cgroups = 1;
1328
	cgrp->root = root;
1329
	RCU_INIT_POINTER(cgrp->name, &root_cgroup_name);
1330
	init_cgroup_housekeeping(cgrp);
1331
	idr_init(&root->cgroup_idr);
1332 1333
}

1334
static int cgroup_init_root_id(struct cgroupfs_root *root, int start, int end)
1335
{
1336
	int id;
1337

T
Tejun Heo 已提交
1338 1339
	lockdep_assert_held(&cgroup_mutex);

1340 1341
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, start, end,
			      GFP_KERNEL);
1342 1343 1344 1345
	if (id < 0)
		return id;

	root->hierarchy_id = id;
1346 1347 1348 1349 1350
	return 0;
}

static void cgroup_exit_root_id(struct cgroupfs_root *root)
{
T
Tejun Heo 已提交
1351 1352
	lockdep_assert_held(&cgroup_mutex);

1353
	if (root->hierarchy_id) {
1354
		idr_remove(&cgroup_hierarchy_idr, root->hierarchy_id);
1355 1356
		root->hierarchy_id = 0;
	}
1357 1358
}

1359 1360
static int cgroup_test_super(struct super_block *sb, void *data)
{
1361
	struct cgroup_sb_opts *opts = data;
1362 1363
	struct cgroupfs_root *root = sb->s_fs_info;

1364 1365 1366
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1367

1368 1369 1370 1371
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
1372 1373
	if ((opts->subsys_mask || opts->none)
	    && (opts->subsys_mask != root->subsys_mask))
1374 1375 1376 1377 1378
		return 0;

	return 1;
}

1379 1380 1381 1382
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1383
	if (!opts->subsys_mask && !opts->none)
1384 1385 1386 1387 1388 1389 1390
		return NULL;

	root = kzalloc(sizeof(*root), GFP_KERNEL);
	if (!root)
		return ERR_PTR(-ENOMEM);

	init_cgroup_root(root);
1391

1392 1393 1394 1395 1396 1397 1398 1399
	/*
	 * We need to set @root->subsys_mask now so that @root can be
	 * matched by cgroup_test_super() before it finishes
	 * initialization; otherwise, competing mounts with the same
	 * options may try to bind the same subsystems instead of waiting
	 * for the first one leading to unexpected mount errors.
	 * SUBSYS_BOUND will be set once actual binding is complete.
	 */
1400
	root->subsys_mask = opts->subsys_mask;
1401 1402 1403 1404 1405
	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1406 1407
	if (opts->cpuset_clone_children)
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags);
1408 1409 1410
	return root;
}

1411
static void cgroup_free_root(struct cgroupfs_root *root)
1412
{
1413 1414 1415
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);
1416

1417
		idr_destroy(&root->cgroup_idr);
1418 1419
		kfree(root);
	}
1420 1421
}

1422 1423 1424
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1425 1426 1427 1428 1429 1430
	struct cgroup_sb_opts *opts = data;

	/* If we don't have a new root, we can't set up a new sb */
	if (!opts->new_root)
		return -EINVAL;

1431
	BUG_ON(!opts->subsys_mask && !opts->none);
1432 1433 1434 1435 1436

	ret = set_anon_super(sb, NULL);
	if (ret)
		return ret;

1437 1438
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449

	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = CGROUP_SUPER_MAGIC;
	sb->s_op = &cgroup_ops;

	return 0;
}

static int cgroup_get_rootdir(struct super_block *sb)
{
A
Al Viro 已提交
1450 1451
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1452
		.d_delete = always_delete_dentry,
A
Al Viro 已提交
1453 1454
	};

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
	struct inode *inode =
		cgroup_new_inode(S_IFDIR | S_IRUGO | S_IXUGO | S_IWUSR, sb);

	if (!inode)
		return -ENOMEM;

	inode->i_fop = &simple_dir_operations;
	inode->i_op = &cgroup_dir_inode_operations;
	/* directories start off with i_nlink == 2 (for "." entry) */
	inc_nlink(inode);
1465 1466
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1467
		return -ENOMEM;
A
Al Viro 已提交
1468 1469
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1470 1471 1472
	return 0;
}

1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
static int cgroup_setup_root(struct cgroupfs_root *root)
{
	LIST_HEAD(tmp_links);
	struct super_block *sb = root->sb;
	struct cgroup *root_cgrp = &root->top_cgroup;
	struct cgroupfs_root *existing_root;
	struct css_set *cset;
	struct inode *inode;
	const struct cred *cred;
	int i, ret;

	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
	BUG_ON(sb->s_root != NULL);

	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&cgroup_tree_mutex);

	ret = cgroup_get_rootdir(sb);
	if (ret) {
		mutex_lock(&cgroup_tree_mutex);
		mutex_lock(&cgroup_mutex);
		return ret;
	}
	inode = sb->s_root->d_inode;

	mutex_lock(&inode->i_mutex);
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);

	ret = idr_alloc(&root->cgroup_idr, root_cgrp, 0, 1, GFP_KERNEL);
	if (ret < 0)
		goto out_unlock;
	root_cgrp->id = ret;

	/* check for name clashes with existing mounts */
	ret = -EBUSY;
	if (strlen(root->name))
		for_each_active_root(existing_root)
			if (!strcmp(existing_root->name, root->name))
				goto out_unlock;

	/*
	 * We're accessing css_set_count without locking css_set_lock here,
	 * but that's OK - it can only be increased by someone holding
	 * cgroup_lock, and that's us. The worst that can happen is that we
	 * have some link structures left over
	 */
	ret = allocate_cgrp_cset_links(css_set_count, &tmp_links);
	if (ret)
		goto out_unlock;

	/* ID 0 is reserved for dummy root, 1 for unified hierarchy */
	ret = cgroup_init_root_id(root, 2, 0);
	if (ret)
		goto out_unlock;

	sb->s_root->d_fsdata = root_cgrp;
	root_cgrp->dentry = sb->s_root;

	/*
	 * We're inside get_sb() and will call lookup_one_len() to create
	 * the root files, which doesn't work if SELinux is in use.  The
	 * following cred dancing somehow works around it.  See 2ce9738ba
	 * ("cgroupfs: use init_cred when populating new cgroupfs mount")
	 * for more details.
	 */
	cred = override_creds(&init_cred);

	ret = cgroup_addrm_files(root_cgrp, cgroup_base_files, true);
	if (ret)
		goto rm_base_files;

	ret = rebind_subsystems(root, root->subsys_mask, 0);
	if (ret)
		goto rm_base_files;

	revert_creds(cred);

	/*
	 * There must be no failure case after here, since rebinding takes
	 * care of subsystems' refcounts, which are explicitly dropped in
	 * the failure exit path.
	 */
	list_add(&root->root_list, &cgroup_roots);
	cgroup_root_count++;

	/*
	 * Link the top cgroup in this hierarchy into all the css_set
	 * objects.
	 */
	write_lock(&css_set_lock);
	hash_for_each(css_set_table, i, cset, hlist)
		link_css_set(&tmp_links, cset, root_cgrp);
	write_unlock(&css_set_lock);

	BUG_ON(!list_empty(&root_cgrp->children));
	BUG_ON(root->number_of_cgroups != 1);

	ret = 0;
	goto out_unlock;

rm_base_files:
	cgroup_addrm_files(&root->top_cgroup, cgroup_base_files, false);
	revert_creds(cred);
	cgroup_exit_root_id(root);
out_unlock:
	mutex_unlock(&inode->i_mutex);
	free_cgrp_cset_links(&tmp_links);
	return ret;
}

A
Al Viro 已提交
1585
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1586
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1587
			 void *data)
1588
{
1589 1590
	struct super_block *sb = NULL;
	struct cgroupfs_root *root = NULL;
1591
	struct cgroup_sb_opts opts;
1592
	struct cgroupfs_root *new_root;
1593
	int ret;
1594

1595
	mutex_lock(&cgroup_tree_mutex);
B
Ben Blum 已提交
1596
	mutex_lock(&cgroup_mutex);
1597 1598

	/* First find the desired set of subsystems */
1599
	ret = parse_cgroupfs_options(data, &opts);
1600
	if (ret)
1601
		goto out_unlock;
1602

1603 1604 1605 1606 1607 1608 1609
	/*
	 * Allocate a new cgroup root. We may not need it if we're
	 * reusing an existing hierarchy.
	 */
	new_root = cgroup_root_from_opts(&opts);
	if (IS_ERR(new_root)) {
		ret = PTR_ERR(new_root);
1610
		goto out_unlock;
1611
	}
1612
	opts.new_root = new_root;
1613

1614
	/* Locate an existing or new sb for this hierarchy */
1615 1616
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&cgroup_tree_mutex);
D
David Howells 已提交
1617
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1618 1619
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);
1620
	if (IS_ERR(sb)) {
1621
		ret = PTR_ERR(sb);
1622
		cgroup_free_root(opts.new_root);
1623
		goto out_unlock;
1624 1625
	}

1626 1627 1628
	root = sb->s_fs_info;
	BUG_ON(!root);
	if (root == opts.new_root) {
1629
		ret = cgroup_setup_root(root);
1630
		if (ret)
1631
			goto out_unlock;
1632 1633 1634 1635 1636
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1637
		cgroup_free_root(opts.new_root);
1638

1639
		if ((root->flags ^ opts.flags) & CGRP_ROOT_OPTION_MASK) {
1640 1641 1642
			if ((root->flags | opts.flags) & CGRP_ROOT_SANE_BEHAVIOR) {
				pr_err("cgroup: sane_behavior: new mount options should match the existing superblock\n");
				ret = -EINVAL;
1643
				goto out_unlock;
1644 1645 1646
			} else {
				pr_warning("cgroup: new mount options do not match the existing superblock, will be ignored\n");
			}
1647
		}
1648 1649
	}

1650 1651
	ret = 0;
out_unlock:
T
Tejun Heo 已提交
1652
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1653
	mutex_unlock(&cgroup_tree_mutex);
1654 1655 1656 1657

	if (ret && !IS_ERR_OR_NULL(sb))
		deactivate_locked_super(sb);

1658 1659
	kfree(opts.release_agent);
	kfree(opts.name);
1660 1661 1662 1663 1664

	if (!ret)
		return dget(sb->s_root);
	else
		return ERR_PTR(ret);
1665 1666
}

S
SeongJae Park 已提交
1667 1668
static void cgroup_kill_sb(struct super_block *sb)
{
1669
	struct cgroupfs_root *root = sb->s_fs_info;
1670
	struct cgroup *cgrp = &root->top_cgroup;
1671
	struct cgrp_cset_link *link, *tmp_link;
1672 1673 1674 1675 1676
	int ret;

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1677
	BUG_ON(!list_empty(&cgrp->children));
1678

1679
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
T
Tejun Heo 已提交
1680
	mutex_lock(&cgroup_tree_mutex);
1681 1682 1683
	mutex_lock(&cgroup_mutex);

	/* Rebind all subsystems back to the default hierarchy */
1684 1685 1686 1687 1688
	if (root->flags & CGRP_ROOT_SUBSYS_BOUND) {
		ret = rebind_subsystems(root, 0, root->subsys_mask);
		/* Shouldn't be able to fail ... */
		BUG_ON(ret);
	}
1689

1690
	/*
1691
	 * Release all the links from cset_links to this hierarchy's
1692 1693 1694
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1695

1696 1697 1698
	list_for_each_entry_safe(link, tmp_link, &cgrp->cset_links, cset_link) {
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
1699 1700 1701 1702
		kfree(link);
	}
	write_unlock(&css_set_lock);

1703 1704
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
1705
		cgroup_root_count--;
1706
	}
1707

1708 1709
	cgroup_exit_root_id(root);

1710
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1711
	mutex_unlock(&cgroup_tree_mutex);
1712
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1713

A
Aristeu Rozanski 已提交
1714 1715
	simple_xattrs_free(&cgrp->xattrs);

1716
	kill_litter_super(sb);
1717
	cgroup_free_root(root);
1718 1719 1720 1721
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1722
	.mount = cgroup_mount,
1723 1724 1725
	.kill_sb = cgroup_kill_sb,
};

1726 1727
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1728 1729 1730 1731 1732 1733
/**
 * cgroup_path - generate the path of a cgroup
 * @cgrp: the cgroup in question
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1734 1735 1736 1737 1738 1739
 * Writes path of cgroup into buf.  Returns 0 on success, -errno on error.
 *
 * We can't generate cgroup path using dentry->d_name, as accessing
 * dentry->name must be protected by irq-unsafe dentry->d_lock or parent
 * inode's i_mutex, while on the other hand cgroup_path() can be called
 * with some irq-safe spinlocks held.
1740
 */
1741
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1742
{
1743
	int ret = -ENAMETOOLONG;
1744
	char *start;
1745

1746 1747 1748
	if (!cgrp->parent) {
		if (strlcpy(buf, "/", buflen) >= buflen)
			return -ENAMETOOLONG;
1749 1750 1751
		return 0;
	}

1752 1753
	start = buf + buflen - 1;
	*start = '\0';
1754

1755
	rcu_read_lock();
1756
	do {
1757 1758 1759 1760
		const char *name = cgroup_name(cgrp);
		int len;

		len = strlen(name);
1761
		if ((start -= len) < buf)
1762 1763
			goto out;
		memcpy(start, name, len);
1764

1765
		if (--start < buf)
1766
			goto out;
1767
		*start = '/';
1768 1769

		cgrp = cgrp->parent;
1770
	} while (cgrp->parent);
1771
	ret = 0;
1772
	memmove(buf, start, buf + buflen - start);
1773 1774 1775
out:
	rcu_read_unlock();
	return ret;
1776
}
B
Ben Blum 已提交
1777
EXPORT_SYMBOL_GPL(cgroup_path);
1778

1779
/**
1780
 * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
1781 1782 1783 1784
 * @task: target task
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1785 1786 1787 1788 1789 1790
 * Determine @task's cgroup on the first (the one with the lowest non-zero
 * hierarchy_id) cgroup hierarchy and copy its path into @buf.  This
 * function grabs cgroup_mutex and shouldn't be used inside locks used by
 * cgroup controller callbacks.
 *
 * Returns 0 on success, fails with -%ENAMETOOLONG if @buflen is too short.
1791
 */
1792
int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
1793 1794
{
	struct cgroupfs_root *root;
1795 1796 1797 1798 1799
	struct cgroup *cgrp;
	int hierarchy_id = 1, ret = 0;

	if (buflen < 2)
		return -ENAMETOOLONG;
1800 1801 1802

	mutex_lock(&cgroup_mutex);

1803 1804
	root = idr_get_next(&cgroup_hierarchy_idr, &hierarchy_id);

1805 1806 1807
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
		ret = cgroup_path(cgrp, buf, buflen);
1808 1809 1810
	} else {
		/* if no hierarchy exists, everyone is in "/" */
		memcpy(buf, "/", 2);
1811 1812 1813 1814 1815
	}

	mutex_unlock(&cgroup_mutex);
	return ret;
}
1816
EXPORT_SYMBOL_GPL(task_cgroup_path);
1817

1818 1819 1820
/*
 * Control Group taskset
 */
1821 1822 1823
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
L
Li Zefan 已提交
1824
	struct css_set		*cset;
1825 1826
};

1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873
struct cgroup_taskset {
	struct task_and_cgroup	single;
	struct flex_array	*tc_array;
	int			tc_array_len;
	int			idx;
	struct cgroup		*cur_cgrp;
};

/**
 * cgroup_taskset_first - reset taskset and return the first task
 * @tset: taskset of interest
 *
 * @tset iteration is initialized and the first task is returned.
 */
struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset)
{
	if (tset->tc_array) {
		tset->idx = 0;
		return cgroup_taskset_next(tset);
	} else {
		tset->cur_cgrp = tset->single.cgrp;
		return tset->single.task;
	}
}
EXPORT_SYMBOL_GPL(cgroup_taskset_first);

/**
 * cgroup_taskset_next - iterate to the next task in taskset
 * @tset: taskset of interest
 *
 * Return the next task in @tset.  Iteration must have been initialized
 * with cgroup_taskset_first().
 */
struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset)
{
	struct task_and_cgroup *tc;

	if (!tset->tc_array || tset->idx >= tset->tc_array_len)
		return NULL;

	tc = flex_array_get(tset->tc_array, tset->idx++);
	tset->cur_cgrp = tc->cgrp;
	return tc->task;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_next);

/**
1874
 * cgroup_taskset_cur_css - return the matching css for the current task
1875
 * @tset: taskset of interest
1876
 * @subsys_id: the ID of the target subsystem
1877
 *
1878 1879 1880
 * Return the css for the current (last returned) task of @tset for
 * subsystem specified by @subsys_id.  This function must be preceded by
 * either cgroup_taskset_first() or cgroup_taskset_next().
1881
 */
1882 1883
struct cgroup_subsys_state *cgroup_taskset_cur_css(struct cgroup_taskset *tset,
						   int subsys_id)
1884
{
1885
	return cgroup_css(tset->cur_cgrp, cgroup_subsys[subsys_id]);
1886
}
1887
EXPORT_SYMBOL_GPL(cgroup_taskset_cur_css);
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899

/**
 * cgroup_taskset_size - return the number of tasks in taskset
 * @tset: taskset of interest
 */
int cgroup_taskset_size(struct cgroup_taskset *tset)
{
	return tset->tc_array ? tset->tc_array_len : 1;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_size);


B
Ben Blum 已提交
1900 1901 1902
/*
 * cgroup_task_migrate - move a task from one cgroup to another.
 *
1903
 * Must be called with cgroup_mutex and threadgroup locked.
B
Ben Blum 已提交
1904
 */
1905 1906 1907
static void cgroup_task_migrate(struct cgroup *old_cgrp,
				struct task_struct *tsk,
				struct css_set *new_cset)
B
Ben Blum 已提交
1908
{
1909
	struct css_set *old_cset;
B
Ben Blum 已提交
1910 1911

	/*
1912 1913 1914
	 * We are synchronized through threadgroup_lock() against PF_EXITING
	 * setting such that we can't race against cgroup_exit() changing the
	 * css_set to init_css_set and dropping the old one.
B
Ben Blum 已提交
1915
	 */
1916
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
1917
	old_cset = task_css_set(tsk);
B
Ben Blum 已提交
1918 1919

	task_lock(tsk);
1920
	rcu_assign_pointer(tsk->cgroups, new_cset);
B
Ben Blum 已提交
1921 1922 1923 1924 1925
	task_unlock(tsk);

	/* Update the css_set linked lists if we're using them */
	write_lock(&css_set_lock);
	if (!list_empty(&tsk->cg_list))
1926
		list_move(&tsk->cg_list, &new_cset->tasks);
B
Ben Blum 已提交
1927 1928 1929
	write_unlock(&css_set_lock);

	/*
1930 1931 1932
	 * We just gained a reference on old_cset by taking it from the
	 * task. As trading it for new_cset is protected by cgroup_mutex,
	 * we're safe to drop it here; it will be freed under RCU.
B
Ben Blum 已提交
1933
	 */
1934 1935
	set_bit(CGRP_RELEASABLE, &old_cgrp->flags);
	put_css_set(old_cset);
B
Ben Blum 已提交
1936 1937
}

L
Li Zefan 已提交
1938
/**
1939
 * cgroup_attach_task - attach a task or a whole threadgroup to a cgroup
B
Ben Blum 已提交
1940
 * @cgrp: the cgroup to attach to
1941 1942
 * @tsk: the task or the leader of the threadgroup to be attached
 * @threadgroup: attach the whole threadgroup?
B
Ben Blum 已提交
1943
 *
1944
 * Call holding cgroup_mutex and the group_rwsem of the leader. Will take
1945
 * task_lock of @tsk or each thread in the threadgroup individually in turn.
B
Ben Blum 已提交
1946
 */
T
Tejun Heo 已提交
1947 1948
static int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk,
			      bool threadgroup)
B
Ben Blum 已提交
1949 1950 1951
{
	int retval, i, group_size;
	struct cgroupfs_root *root = cgrp->root;
T
Tejun Heo 已提交
1952
	struct cgroup_subsys_state *css, *failed_css = NULL;
B
Ben Blum 已提交
1953
	/* threadgroup list cursor and array */
1954
	struct task_struct *leader = tsk;
1955
	struct task_and_cgroup *tc;
1956
	struct flex_array *group;
1957
	struct cgroup_taskset tset = { };
B
Ben Blum 已提交
1958 1959 1960 1961 1962

	/*
	 * step 0: in order to do expensive, possibly blocking operations for
	 * every thread, we cannot iterate the thread group list, since it needs
	 * rcu or tasklist locked. instead, build an array of all threads in the
1963 1964
	 * group - group_rwsem prevents new threads from appearing, and if
	 * threads exit, this will just be an over-estimate.
B
Ben Blum 已提交
1965
	 */
1966 1967 1968 1969
	if (threadgroup)
		group_size = get_nr_threads(tsk);
	else
		group_size = 1;
1970
	/* flex_array supports very large thread-groups better than kmalloc. */
1971
	group = flex_array_alloc(sizeof(*tc), group_size, GFP_KERNEL);
B
Ben Blum 已提交
1972 1973
	if (!group)
		return -ENOMEM;
1974
	/* pre-allocate to guarantee space while iterating in rcu read-side. */
1975
	retval = flex_array_prealloc(group, 0, group_size, GFP_KERNEL);
1976 1977
	if (retval)
		goto out_free_group_list;
B
Ben Blum 已提交
1978 1979

	i = 0;
1980 1981 1982 1983 1984 1985
	/*
	 * Prevent freeing of tasks while we take a snapshot. Tasks that are
	 * already PF_EXITING could be freed from underneath us unless we
	 * take an rcu_read_lock.
	 */
	rcu_read_lock();
B
Ben Blum 已提交
1986
	do {
1987 1988
		struct task_and_cgroup ent;

1989 1990
		/* @tsk either already exited or can't exit until the end */
		if (tsk->flags & PF_EXITING)
1991
			goto next;
1992

B
Ben Blum 已提交
1993 1994
		/* as per above, nr_threads may decrease, but not increase. */
		BUG_ON(i >= group_size);
1995 1996
		ent.task = tsk;
		ent.cgrp = task_cgroup_from_root(tsk, root);
1997 1998
		/* nothing to do if this task is already in the cgroup */
		if (ent.cgrp == cgrp)
1999
			goto next;
2000 2001 2002 2003
		/*
		 * saying GFP_ATOMIC has no effect here because we did prealloc
		 * earlier, but it's good form to communicate our expectations.
		 */
2004
		retval = flex_array_put(group, i, &ent, GFP_ATOMIC);
2005
		BUG_ON(retval != 0);
B
Ben Blum 已提交
2006
		i++;
2007
	next:
2008 2009
		if (!threadgroup)
			break;
B
Ben Blum 已提交
2010
	} while_each_thread(leader, tsk);
2011
	rcu_read_unlock();
B
Ben Blum 已提交
2012 2013
	/* remember the number of threads in the array for later. */
	group_size = i;
2014 2015
	tset.tc_array = group;
	tset.tc_array_len = group_size;
B
Ben Blum 已提交
2016

2017 2018
	/* methods shouldn't be called if no task is actually migrating */
	retval = 0;
2019
	if (!group_size)
2020
		goto out_free_group_list;
2021

B
Ben Blum 已提交
2022 2023 2024
	/*
	 * step 1: check that we can legitimately attach to the cgroup.
	 */
T
Tejun Heo 已提交
2025 2026 2027
	for_each_css(css, i, cgrp) {
		if (css->ss->can_attach) {
			retval = css->ss->can_attach(css, &tset);
B
Ben Blum 已提交
2028
			if (retval) {
T
Tejun Heo 已提交
2029
				failed_css = css;
B
Ben Blum 已提交
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039
				goto out_cancel_attach;
			}
		}
	}

	/*
	 * step 2: make sure css_sets exist for all threads to be migrated.
	 * we use find_css_set, which allocates a new one if necessary.
	 */
	for (i = 0; i < group_size; i++) {
2040 2041
		struct css_set *old_cset;

2042
		tc = flex_array_get(group, i);
2043
		old_cset = task_css_set(tc->task);
L
Li Zefan 已提交
2044 2045
		tc->cset = find_css_set(old_cset, cgrp);
		if (!tc->cset) {
2046 2047
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2048 2049 2050 2051
		}
	}

	/*
2052 2053 2054
	 * step 3: now that we're guaranteed success wrt the css_sets,
	 * proceed to move all tasks to the new cgroup.  There are no
	 * failure cases after here, so this is the commit point.
B
Ben Blum 已提交
2055 2056
	 */
	for (i = 0; i < group_size; i++) {
2057
		tc = flex_array_get(group, i);
L
Li Zefan 已提交
2058
		cgroup_task_migrate(tc->cgrp, tc->task, tc->cset);
B
Ben Blum 已提交
2059 2060 2061 2062
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2063
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2064
	 */
T
Tejun Heo 已提交
2065 2066 2067
	for_each_css(css, i, cgrp)
		if (css->ss->attach)
			css->ss->attach(css, &tset);
B
Ben Blum 已提交
2068 2069 2070 2071 2072

	/*
	 * step 5: success! and cleanup
	 */
	retval = 0;
2073 2074 2075 2076
out_put_css_set_refs:
	if (retval) {
		for (i = 0; i < group_size; i++) {
			tc = flex_array_get(group, i);
L
Li Zefan 已提交
2077
			if (!tc->cset)
2078
				break;
L
Li Zefan 已提交
2079
			put_css_set(tc->cset);
2080
		}
B
Ben Blum 已提交
2081 2082 2083
	}
out_cancel_attach:
	if (retval) {
T
Tejun Heo 已提交
2084 2085
		for_each_css(css, i, cgrp) {
			if (css == failed_css)
B
Ben Blum 已提交
2086
				break;
T
Tejun Heo 已提交
2087 2088
			if (css->ss->cancel_attach)
				css->ss->cancel_attach(css, &tset);
B
Ben Blum 已提交
2089 2090 2091
		}
	}
out_free_group_list:
2092
	flex_array_free(group);
B
Ben Blum 已提交
2093 2094 2095 2096 2097
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2098 2099
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2100
 */
B
Ben Blum 已提交
2101
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2102 2103
{
	struct task_struct *tsk;
2104
	const struct cred *cred = current_cred(), *tcred;
2105 2106
	int ret;

B
Ben Blum 已提交
2107 2108 2109
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2110 2111
retry_find_task:
	rcu_read_lock();
2112
	if (pid) {
2113
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2114 2115
		if (!tsk) {
			rcu_read_unlock();
S
SeongJae Park 已提交
2116
			ret = -ESRCH;
2117
			goto out_unlock_cgroup;
2118
		}
B
Ben Blum 已提交
2119 2120 2121 2122
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2123
		tcred = __task_cred(tsk);
2124 2125 2126
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2127
			rcu_read_unlock();
2128 2129
			ret = -EACCES;
			goto out_unlock_cgroup;
2130
		}
2131 2132
	} else
		tsk = current;
2133 2134

	if (threadgroup)
2135
		tsk = tsk->group_leader;
2136 2137

	/*
2138
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2139 2140 2141
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2142
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2143 2144 2145 2146 2147
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164
	get_task_struct(tsk);
	rcu_read_unlock();

	threadgroup_lock(tsk);
	if (threadgroup) {
		if (!thread_group_leader(tsk)) {
			/*
			 * a race with de_thread from another thread's exec()
			 * may strip us of our leadership, if this happens,
			 * there is no choice but to throw this task away and
			 * try again; this is
			 * "double-double-toil-and-trouble-check locking".
			 */
			threadgroup_unlock(tsk);
			put_task_struct(tsk);
			goto retry_find_task;
		}
2165 2166 2167 2168
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2169 2170
	threadgroup_unlock(tsk);

2171
	put_task_struct(tsk);
2172
out_unlock_cgroup:
T
Tejun Heo 已提交
2173
	mutex_unlock(&cgroup_mutex);
2174 2175 2176
	return ret;
}

2177 2178 2179 2180 2181 2182 2183 2184 2185 2186
/**
 * cgroup_attach_task_all - attach task 'tsk' to all cgroups of task 'from'
 * @from: attach to all cgroups of a given task
 * @tsk: the task to be attached
 */
int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
{
	struct cgroupfs_root *root;
	int retval = 0;

T
Tejun Heo 已提交
2187
	mutex_lock(&cgroup_mutex);
2188
	for_each_active_root(root) {
L
Li Zefan 已提交
2189
		struct cgroup *from_cgrp = task_cgroup_from_root(from, root);
2190

L
Li Zefan 已提交
2191
		retval = cgroup_attach_task(from_cgrp, tsk, false);
2192 2193 2194
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2195
	mutex_unlock(&cgroup_mutex);
2196 2197 2198 2199 2200

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2201 2202
static int cgroup_tasks_write(struct cgroup_subsys_state *css,
			      struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2203
{
2204
	return attach_task_by_pid(css->cgroup, pid, false);
B
Ben Blum 已提交
2205 2206
}

2207 2208
static int cgroup_procs_write(struct cgroup_subsys_state *css,
			      struct cftype *cft, u64 tgid)
2209
{
2210
	return attach_task_by_pid(css->cgroup, tgid, true);
2211 2212
}

2213 2214
static int cgroup_release_agent_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, const char *buffer)
2215
{
2216 2217 2218
	struct cgroupfs_root *root = css->cgroup->root;

	BUILD_BUG_ON(sizeof(root->release_agent_path) < PATH_MAX);
2219
	if (!cgroup_lock_live_group(css->cgroup))
2220
		return -ENODEV;
2221
	spin_lock(&release_agent_path_lock);
2222 2223
	strlcpy(root->release_agent_path, buffer,
		sizeof(root->release_agent_path));
2224
	spin_unlock(&release_agent_path_lock);
T
Tejun Heo 已提交
2225
	mutex_unlock(&cgroup_mutex);
2226 2227 2228
	return 0;
}

2229
static int cgroup_release_agent_show(struct seq_file *seq, void *v)
2230
{
2231
	struct cgroup *cgrp = seq_css(seq)->cgroup;
2232

2233 2234 2235 2236
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
	seq_puts(seq, cgrp->root->release_agent_path);
	seq_putc(seq, '\n');
T
Tejun Heo 已提交
2237
	mutex_unlock(&cgroup_mutex);
2238 2239 2240
	return 0;
}

2241
static int cgroup_sane_behavior_show(struct seq_file *seq, void *v)
2242
{
2243 2244 2245
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	seq_printf(seq, "%d\n", cgroup_sane_behavior(cgrp));
2246 2247 2248
	return 0;
}

2249 2250
static ssize_t cgroup_file_write(struct file *file, const char __user *userbuf,
				 size_t nbytes, loff_t *ppos)
2251
{
2252 2253 2254
	struct cfent *cfe = __d_cfe(file->f_dentry);
	struct cftype *cft = __d_cft(file->f_dentry);
	struct cgroup_subsys_state *css = cfe->css;
2255
	size_t max_bytes = max(cft->max_write_len, PAGE_SIZE);
2256 2257
	char *buf;
	int ret;
2258

2259
	if (nbytes > max_bytes)
2260 2261
		return -E2BIG;

2262 2263 2264
	buf = kmalloc(nbytes + 1, GFP_KERNEL);
	if (!buf)
		return -ENOMEM;
2265

2266 2267 2268 2269
	if (copy_from_user(buf, userbuf, nbytes)) {
		ret = -EFAULT;
		goto out_free;
	}
2270

2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
	buf[nbytes] = '\0';

	if (cft->write_string) {
		ret = cft->write_string(css, cft, strstrip(buf));
	} else if (cft->write_u64) {
		unsigned long long v;
		ret = kstrtoull(buf, 0, &v);
		if (!ret)
			ret = cft->write_u64(css, cft, v);
	} else if (cft->write_s64) {
		long long v;
		ret = kstrtoll(buf, 0, &v);
		if (!ret)
			ret = cft->write_s64(css, cft, v);
	} else if (cft->trigger) {
		ret = cft->trigger(css, (unsigned int)cft->private);
2287
	} else {
2288
		ret = -EINVAL;
2289
	}
2290 2291 2292
out_free:
	kfree(buf);
	return ret ?: nbytes;
2293 2294
}

2295 2296 2297 2298 2299
/*
 * seqfile ops/methods for returning structured data. Currently just
 * supports string->u64 maps, but can be extended in future.
 */

2300
static void *cgroup_seqfile_start(struct seq_file *seq, loff_t *ppos)
2301
{
2302
	struct cftype *cft = seq_cft(seq);
2303

2304 2305 2306 2307 2308 2309 2310 2311
	if (cft->seq_start) {
		return cft->seq_start(seq, ppos);
	} else {
		/*
		 * The same behavior and code as single_open().  Returns
		 * !NULL if pos is at the beginning; otherwise, NULL.
		 */
		return NULL + !*ppos;
L
Li Zefan 已提交
2312
	}
2313 2314
}

2315
static void *cgroup_seqfile_next(struct seq_file *seq, void *v, loff_t *ppos)
2316
{
2317
	struct cftype *cft = seq_cft(seq);
2318

2319 2320 2321 2322 2323 2324 2325 2326 2327
	if (cft->seq_next) {
		return cft->seq_next(seq, v, ppos);
	} else {
		/*
		 * The same behavior and code as single_open(), always
		 * terminate after the initial read.
		 */
		++*ppos;
		return NULL;
2328
	}
2329 2330
}

2331
static void cgroup_seqfile_stop(struct seq_file *seq, void *v)
2332
{
2333
	struct cftype *cft = seq_cft(seq);
2334

2335 2336
	if (cft->seq_stop)
		cft->seq_stop(seq, v);
2337 2338
}

2339
static int cgroup_seqfile_show(struct seq_file *m, void *arg)
2340
{
2341 2342
	struct cftype *cft = seq_cft(m);
	struct cgroup_subsys_state *css = seq_css(m);
2343

2344 2345
	if (cft->seq_show)
		return cft->seq_show(m, arg);
2346

2347
	if (cft->read_u64)
2348 2349 2350 2351 2352 2353
		seq_printf(m, "%llu\n", cft->read_u64(css, cft));
	else if (cft->read_s64)
		seq_printf(m, "%lld\n", cft->read_s64(css, cft));
	else
		return -EINVAL;
	return 0;
2354 2355
}

2356 2357 2358 2359 2360
static struct seq_operations cgroup_seq_operations = {
	.start		= cgroup_seqfile_start,
	.next		= cgroup_seqfile_next,
	.stop		= cgroup_seqfile_stop,
	.show		= cgroup_seqfile_show,
2361 2362
};

2363 2364
static int cgroup_file_open(struct inode *inode, struct file *file)
{
2365 2366
	struct cfent *cfe = __d_cfe(file->f_dentry);
	struct cftype *cft = __d_cft(file->f_dentry);
2367 2368
	struct cgroup *cgrp = __d_cgrp(cfe->dentry->d_parent);
	struct cgroup_subsys_state *css;
2369
	struct cgroup_open_file *of;
2370 2371 2372 2373 2374
	int err;

	err = generic_file_open(inode, file);
	if (err)
		return err;
2375

2376 2377 2378 2379 2380
	/*
	 * If the file belongs to a subsystem, pin the css.  Will be
	 * unpinned either on open failure or release.  This ensures that
	 * @css stays alive for all file operations.
	 */
2381
	rcu_read_lock();
2382 2383 2384
	css = cgroup_css(cgrp, cft->ss);
	if (cft->ss && !css_tryget(css))
		css = NULL;
2385
	rcu_read_unlock();
2386

2387
	if (!css)
2388
		return -ENODEV;
2389

2390 2391 2392 2393 2394 2395 2396 2397
	/*
	 * @cfe->css is used by read/write/close to determine the
	 * associated css.  @file->private_data would be a better place but
	 * that's already used by seqfile.  Multiple accessors may use it
	 * simultaneously which is okay as the association never changes.
	 */
	WARN_ON_ONCE(cfe->css && cfe->css != css);
	cfe->css = css;
2398

2399 2400 2401 2402 2403
	of = __seq_open_private(file, &cgroup_seq_operations,
				sizeof(struct cgroup_open_file));
	if (of) {
		of->cfe = cfe;
		return 0;
2404
	}
2405

2406
	if (css->ss)
2407
		css_put(css);
2408
	return -ENOMEM;
2409 2410 2411 2412
}

static int cgroup_file_release(struct inode *inode, struct file *file)
{
2413
	struct cfent *cfe = __d_cfe(file->f_dentry);
2414
	struct cgroup_subsys_state *css = cfe->css;
2415

T
Tejun Heo 已提交
2416
	if (css->ss)
2417
		css_put(css);
2418
	return seq_release_private(inode, file);
2419 2420 2421 2422 2423 2424 2425 2426
}

/*
 * cgroup_rename - Only allow simple rename of directories in place.
 */
static int cgroup_rename(struct inode *old_dir, struct dentry *old_dentry,
			    struct inode *new_dir, struct dentry *new_dentry)
{
2427 2428 2429 2430 2431 2432 2433 2434 2435 2436
	int ret;
	struct cgroup_name *name, *old_name;
	struct cgroup *cgrp;

	/*
	 * It's convinient to use parent dir's i_mutex to protected
	 * cgrp->name.
	 */
	lockdep_assert_held(&old_dir->i_mutex);

2437 2438 2439 2440 2441 2442
	if (!S_ISDIR(old_dentry->d_inode->i_mode))
		return -ENOTDIR;
	if (new_dentry->d_inode)
		return -EEXIST;
	if (old_dir != new_dir)
		return -EIO;
2443 2444 2445

	cgrp = __d_cgrp(old_dentry);

2446 2447 2448 2449 2450 2451 2452
	/*
	 * This isn't a proper migration and its usefulness is very
	 * limited.  Disallow if sane_behavior.
	 */
	if (cgroup_sane_behavior(cgrp))
		return -EPERM;

T
Tejun Heo 已提交
2453
	name = cgroup_alloc_name(new_dentry->d_name.name);
2454 2455 2456 2457 2458 2459 2460 2461 2462
	if (!name)
		return -ENOMEM;

	ret = simple_rename(old_dir, old_dentry, new_dir, new_dentry);
	if (ret) {
		kfree(name);
		return ret;
	}

2463
	old_name = rcu_dereference_protected(cgrp->name, true);
2464 2465 2466 2467
	rcu_assign_pointer(cgrp->name, name);

	kfree_rcu(old_name, rcu_head);
	return 0;
2468 2469
}

A
Aristeu Rozanski 已提交
2470 2471 2472 2473 2474
static struct simple_xattrs *__d_xattrs(struct dentry *dentry)
{
	if (S_ISDIR(dentry->d_inode->i_mode))
		return &__d_cgrp(dentry)->xattrs;
	else
L
Li Zefan 已提交
2475
		return &__d_cfe(dentry)->xattrs;
A
Aristeu Rozanski 已提交
2476 2477 2478 2479 2480
}

static inline int xattr_enabled(struct dentry *dentry)
{
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
2481
	return root->flags & CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 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 2524 2525 2526 2527
}

static bool is_valid_xattr(const char *name)
{
	if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
	    !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
		return true;
	return false;
}

static int cgroup_setxattr(struct dentry *dentry, const char *name,
			   const void *val, size_t size, int flags)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_set(__d_xattrs(dentry), name, val, size, flags);
}

static int cgroup_removexattr(struct dentry *dentry, const char *name)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_remove(__d_xattrs(dentry), name);
}

static ssize_t cgroup_getxattr(struct dentry *dentry, const char *name,
			       void *buf, size_t size)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_get(__d_xattrs(dentry), name, buf, size);
}

static ssize_t cgroup_listxattr(struct dentry *dentry, char *buf, size_t size)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	return simple_xattr_list(__d_xattrs(dentry), buf, size);
}

2528
static const struct file_operations cgroup_file_operations = {
2529
	.read = seq_read,
2530 2531 2532 2533 2534 2535
	.write = cgroup_file_write,
	.llseek = generic_file_llseek,
	.open = cgroup_file_open,
	.release = cgroup_file_release,
};

A
Aristeu Rozanski 已提交
2536 2537 2538 2539 2540 2541 2542
static const struct inode_operations cgroup_file_inode_operations = {
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
};

2543
static const struct inode_operations cgroup_dir_inode_operations = {
A
Al Viro 已提交
2544
	.lookup = simple_lookup,
2545 2546 2547
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
A
Aristeu Rozanski 已提交
2548 2549 2550 2551
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
2552 2553
};

A
Al Viro 已提交
2554
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2555 2556
				struct super_block *sb)
{
2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
	struct inode *inode;

	if (!dentry)
		return -ENOENT;
	if (dentry->d_inode)
		return -EEXIST;

	inode = cgroup_new_inode(mode, sb);
	if (!inode)
		return -ENOMEM;

	if (S_ISDIR(mode)) {
		inode->i_op = &cgroup_dir_inode_operations;
		inode->i_fop = &simple_dir_operations;

		/* start off with i_nlink == 2 (for "." entry) */
		inc_nlink(inode);
T
Tejun Heo 已提交
2574
		inc_nlink(dentry->d_parent->d_inode);
2575

2576 2577 2578 2579 2580 2581 2582 2583 2584
		/*
		 * Control reaches here with cgroup_mutex held.
		 * @inode->i_mutex should nest outside cgroup_mutex but we
		 * want to populate it immediately without releasing
		 * cgroup_mutex.  As @inode isn't visible to anyone else
		 * yet, trylock will always succeed without affecting
		 * lockdep checks.
		 */
		WARN_ON_ONCE(!mutex_trylock(&inode->i_mutex));
2585 2586 2587
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
A
Aristeu Rozanski 已提交
2588
		inode->i_op = &cgroup_file_inode_operations;
2589 2590 2591 2592 2593 2594
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

L
Li Zefan 已提交
2595 2596 2597 2598 2599 2600 2601 2602 2603
/**
 * cgroup_file_mode - deduce file mode of a control file
 * @cft: the control file in question
 *
 * returns cft->mode if ->mode is not 0
 * returns S_IRUGO|S_IWUSR if it has both a read and a write handler
 * returns S_IRUGO if it has only a read handler
 * returns S_IWUSR if it has only a write hander
 */
A
Al Viro 已提交
2604
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2605
{
A
Al Viro 已提交
2606
	umode_t mode = 0;
L
Li Zefan 已提交
2607 2608 2609 2610

	if (cft->mode)
		return cft->mode;

2611
	if (cft->read_u64 || cft->read_s64 || cft->seq_show)
L
Li Zefan 已提交
2612 2613
		mode |= S_IRUGO;

2614 2615
	if (cft->write_u64 || cft->write_s64 || cft->write_string ||
	    cft->trigger)
L
Li Zefan 已提交
2616 2617 2618 2619 2620
		mode |= S_IWUSR;

	return mode;
}

2621
static int cgroup_add_file(struct cgroup *cgrp, struct cftype *cft)
2622
{
2623
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2624
	struct cgroup *parent = __d_cgrp(dir);
2625
	struct dentry *dentry;
T
Tejun Heo 已提交
2626
	struct cfent *cfe;
2627
	int error;
A
Al Viro 已提交
2628
	umode_t mode;
T
Tejun Heo 已提交
2629
	char name[CGROUP_FILE_NAME_MAX];
T
Tejun Heo 已提交
2630

2631
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2632 2633 2634 2635 2636

	cfe = kzalloc(sizeof(*cfe), GFP_KERNEL);
	if (!cfe)
		return -ENOMEM;

T
Tejun Heo 已提交
2637
	cgroup_file_name(cgrp, cft, name);
2638
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2639
	if (IS_ERR(dentry)) {
2640
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2641 2642 2643
		goto out;
	}

2644 2645 2646 2647 2648
	cfe->type = (void *)cft;
	cfe->dentry = dentry;
	dentry->d_fsdata = cfe;
	simple_xattrs_init(&cfe->xattrs);

T
Tejun Heo 已提交
2649 2650 2651 2652 2653 2654 2655 2656 2657
	mode = cgroup_file_mode(cft);
	error = cgroup_create_file(dentry, mode | S_IFREG, cgrp->root->sb);
	if (!error) {
		list_add_tail(&cfe->node, &parent->files);
		cfe = NULL;
	}
	dput(dentry);
out:
	kfree(cfe);
2658 2659 2660
	return error;
}

2661 2662 2663 2664 2665 2666 2667
/**
 * cgroup_addrm_files - add or remove files to a cgroup directory
 * @cgrp: the target cgroup
 * @cfts: array of cftypes to be added
 * @is_add: whether to add or remove
 *
 * Depending on @is_add, add or remove files defined by @cfts on @cgrp.
2668 2669 2670
 * For removals, this function never fails.  If addition fails, this
 * function doesn't remove files already added.  The caller is responsible
 * for cleaning up.
2671
 */
2672 2673
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add)
2674
{
A
Aristeu Rozanski 已提交
2675
	struct cftype *cft;
2676 2677 2678
	int ret;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
T
Tejun Heo 已提交
2679
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
2680 2681

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2682
		/* does cft->flags tell us to skip this file on @cgrp? */
2683 2684
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2685 2686 2687 2688 2689
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2690
		if (is_add) {
2691
			ret = cgroup_add_file(cgrp, cft);
2692
			if (ret) {
2693
				pr_warn("cgroup_addrm_files: failed to add %s, err=%d\n",
2694 2695 2696
					cft->name, ret);
				return ret;
			}
2697 2698
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2699
		}
2700
	}
2701
	return 0;
2702 2703
}

2704
static void cgroup_cfts_prepare(void)
2705
	__acquires(&cgroup_mutex)
2706 2707 2708 2709
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
2710 2711
	 * Instead, we use css_for_each_descendant_pre() and drop RCU read
	 * lock before calling cgroup_addrm_files().
2712
	 */
T
Tejun Heo 已提交
2713
	mutex_lock(&cgroup_tree_mutex);
2714 2715 2716
	mutex_lock(&cgroup_mutex);
}

2717
static int cgroup_cfts_commit(struct cftype *cfts, bool is_add)
2718
	__releases(&cgroup_mutex)
2719 2720
{
	LIST_HEAD(pending);
2721
	struct cgroup_subsys *ss = cfts[0].ss;
2722
	struct cgroup *root = &ss->root->top_cgroup;
2723
	struct super_block *sb = ss->root->sb;
2724 2725
	struct dentry *prev = NULL;
	struct inode *inode;
2726
	struct cgroup_subsys_state *css;
2727
	u64 update_before;
2728
	int ret = 0;
2729

2730 2731
	mutex_unlock(&cgroup_mutex);

2732
	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
2733
	if (!cfts || ss->root == &cgroup_dummy_root ||
2734
	    !atomic_inc_not_zero(&sb->s_active)) {
T
Tejun Heo 已提交
2735
		mutex_unlock(&cgroup_tree_mutex);
2736
		return 0;
2737 2738 2739
	}

	/*
2740 2741
	 * All cgroups which are created after we drop cgroup_mutex will
	 * have the updated set of files, so we only need to update the
2742
	 * cgroups created before the current @cgroup_serial_nr_next.
2743
	 */
2744
	update_before = cgroup_serial_nr_next;
2745 2746

	/* add/rm files for all cgroups created before */
2747
	css_for_each_descendant_pre(css, cgroup_css(root, ss)) {
2748 2749
		struct cgroup *cgrp = css->cgroup;

2750 2751 2752 2753 2754 2755 2756
		if (cgroup_is_dead(cgrp))
			continue;

		inode = cgrp->dentry->d_inode;
		dget(cgrp->dentry);
		dput(prev);
		prev = cgrp->dentry;
2757

T
Tejun Heo 已提交
2758
		mutex_unlock(&cgroup_tree_mutex);
2759
		mutex_lock(&inode->i_mutex);
T
Tejun Heo 已提交
2760
		mutex_lock(&cgroup_tree_mutex);
2761
		if (cgrp->serial_nr < update_before && !cgroup_is_dead(cgrp))
2762
			ret = cgroup_addrm_files(cgrp, cfts, is_add);
2763
		mutex_unlock(&inode->i_mutex);
2764 2765
		if (ret)
			break;
2766
	}
T
Tejun Heo 已提交
2767
	mutex_unlock(&cgroup_tree_mutex);
2768 2769
	dput(prev);
	deactivate_super(sb);
2770
	return ret;
2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786
}

/**
 * cgroup_add_cftypes - add an array of cftypes to a subsystem
 * @ss: target cgroup subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
 * Register @cfts to @ss.  Files described by @cfts are created for all
 * existing cgroups to which @ss is attached and all future cgroups will
 * have them too.  This function can be called anytime whether @ss is
 * attached or not.
 *
 * Returns 0 on successful registration, -errno on failure.  Note that this
 * function currently returns 0 as long as @cfts registration is successful
 * even if some file creation attempts on existing cgroups fail.
 */
A
Aristeu Rozanski 已提交
2787
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2788 2789
{
	struct cftype_set *set;
2790
	struct cftype *cft;
2791
	int ret;
2792 2793 2794 2795 2796

	set = kzalloc(sizeof(*set), GFP_KERNEL);
	if (!set)
		return -ENOMEM;

2797 2798 2799
	for (cft = cfts; cft->name[0] != '\0'; cft++)
		cft->ss = ss;

2800 2801 2802
	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2803
	ret = cgroup_cfts_commit(cfts, true);
2804
	if (ret)
2805
		cgroup_rm_cftypes(cfts);
2806
	return ret;
2807 2808 2809
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2810 2811 2812 2813
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
2814 2815 2816
 * Unregister @cfts.  Files described by @cfts are removed from all
 * existing cgroups and all future cgroups won't have them either.  This
 * function can be called anytime whether @cfts' subsys is attached or not.
2817 2818
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
2819
 * registered.
2820
 */
2821
int cgroup_rm_cftypes(struct cftype *cfts)
2822 2823 2824
{
	struct cftype_set *set;

2825 2826 2827
	if (!cfts || !cfts[0].ss)
		return -ENOENT;

2828 2829
	cgroup_cfts_prepare();

2830
	list_for_each_entry(set, &cfts[0].ss->cftsets, node) {
2831
		if (set->cfts == cfts) {
2832 2833
			list_del(&set->node);
			kfree(set);
2834
			cgroup_cfts_commit(cfts, false);
2835 2836 2837 2838
			return 0;
		}
	}

2839
	cgroup_cfts_commit(NULL, false);
2840 2841 2842
	return -ENOENT;
}

L
Li Zefan 已提交
2843 2844 2845 2846 2847 2848
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2849
int cgroup_task_count(const struct cgroup *cgrp)
2850 2851
{
	int count = 0;
2852
	struct cgrp_cset_link *link;
2853 2854

	read_lock(&css_set_lock);
2855 2856
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
2857
	read_unlock(&css_set_lock);
2858 2859 2860
	return count;
}

2861
/*
2862 2863 2864
 * To reduce the fork() overhead for systems that are not actually using
 * their cgroups capability, we don't maintain the lists running through
 * each css_set to its tasks until we see the list actually used - in other
2865
 * words after the first call to css_task_iter_start().
2866
 */
2867
static void cgroup_enable_task_cg_lists(void)
2868 2869 2870 2871
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
2872 2873 2874 2875 2876 2877 2878 2879
	/*
	 * We need tasklist_lock because RCU is not safe against
	 * while_each_thread(). Besides, a forking task that has passed
	 * cgroup_post_fork() without seeing use_task_css_set_links = 1
	 * is not guaranteed to have its child immediately visible in the
	 * tasklist if we walk through it with RCU.
	 */
	read_lock(&tasklist_lock);
2880 2881
	do_each_thread(g, p) {
		task_lock(p);
2882 2883 2884 2885 2886 2887
		/*
		 * We should check if the process is exiting, otherwise
		 * it will race with cgroup_exit() in that the list
		 * entry won't be deleted though the process has exited.
		 */
		if (!(p->flags & PF_EXITING) && list_empty(&p->cg_list))
2888
			list_add(&p->cg_list, &task_css_set(p)->tasks);
2889 2890
		task_unlock(p);
	} while_each_thread(g, p);
2891
	read_unlock(&tasklist_lock);
2892 2893 2894
	write_unlock(&css_set_lock);
}

2895
/**
2896 2897 2898
 * css_next_child - find the next child of a given css
 * @pos_css: the current position (%NULL to initiate traversal)
 * @parent_css: css whose children to walk
2899
 *
2900
 * This function returns the next child of @parent_css and should be called
2901 2902 2903
 * under either cgroup_mutex or RCU read lock.  The only requirement is
 * that @parent_css and @pos_css are accessible.  The next sibling is
 * guaranteed to be returned regardless of their states.
2904
 */
2905 2906 2907
struct cgroup_subsys_state *
css_next_child(struct cgroup_subsys_state *pos_css,
	       struct cgroup_subsys_state *parent_css)
2908
{
2909 2910
	struct cgroup *pos = pos_css ? pos_css->cgroup : NULL;
	struct cgroup *cgrp = parent_css->cgroup;
2911 2912
	struct cgroup *next;

T
Tejun Heo 已提交
2913
	cgroup_assert_mutexes_or_rcu_locked();
2914 2915 2916 2917

	/*
	 * @pos could already have been removed.  Once a cgroup is removed,
	 * its ->sibling.next is no longer updated when its next sibling
2918 2919 2920 2921 2922 2923 2924
	 * changes.  As CGRP_DEAD assertion is serialized and happens
	 * before the cgroup is taken off the ->sibling list, if we see it
	 * unasserted, it's guaranteed that the next sibling hasn't
	 * finished its grace period even if it's already removed, and thus
	 * safe to dereference from this RCU critical section.  If
	 * ->sibling.next is inaccessible, cgroup_is_dead() is guaranteed
	 * to be visible as %true here.
2925 2926 2927 2928 2929 2930 2931 2932
	 *
	 * If @pos is dead, its next pointer can't be dereferenced;
	 * however, as each cgroup is given a monotonically increasing
	 * unique serial number and always appended to the sibling list,
	 * the next one can be found by walking the parent's children until
	 * we see a cgroup with higher serial number than @pos's.  While
	 * this path can be slower, it's taken only when either the current
	 * cgroup is removed or iteration and removal race.
2933
	 */
2934 2935 2936
	if (!pos) {
		next = list_entry_rcu(cgrp->children.next, struct cgroup, sibling);
	} else if (likely(!cgroup_is_dead(pos))) {
2937
		next = list_entry_rcu(pos->sibling.next, struct cgroup, sibling);
2938 2939 2940 2941
	} else {
		list_for_each_entry_rcu(next, &cgrp->children, sibling)
			if (next->serial_nr > pos->serial_nr)
				break;
2942 2943
	}

2944 2945 2946
	if (&next->sibling == &cgrp->children)
		return NULL;

2947
	return cgroup_css(next, parent_css->ss);
2948
}
2949
EXPORT_SYMBOL_GPL(css_next_child);
2950

2951
/**
2952
 * css_next_descendant_pre - find the next descendant for pre-order walk
2953
 * @pos: the current position (%NULL to initiate traversal)
2954
 * @root: css whose descendants to walk
2955
 *
2956
 * To be used by css_for_each_descendant_pre().  Find the next descendant
2957 2958
 * to visit for pre-order traversal of @root's descendants.  @root is
 * included in the iteration and the first node to be visited.
2959
 *
2960 2961 2962 2963
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct next descendant as long
 * as both @pos and @root are accessible and @pos is a descendant of @root.
2964
 */
2965 2966 2967
struct cgroup_subsys_state *
css_next_descendant_pre(struct cgroup_subsys_state *pos,
			struct cgroup_subsys_state *root)
2968
{
2969
	struct cgroup_subsys_state *next;
2970

T
Tejun Heo 已提交
2971
	cgroup_assert_mutexes_or_rcu_locked();
2972

2973
	/* if first iteration, visit @root */
2974
	if (!pos)
2975
		return root;
2976 2977

	/* visit the first child if exists */
2978
	next = css_next_child(NULL, pos);
2979 2980 2981 2982
	if (next)
		return next;

	/* no child, visit my or the closest ancestor's next sibling */
2983 2984
	while (pos != root) {
		next = css_next_child(pos, css_parent(pos));
2985
		if (next)
2986
			return next;
2987
		pos = css_parent(pos);
2988
	}
2989 2990 2991

	return NULL;
}
2992
EXPORT_SYMBOL_GPL(css_next_descendant_pre);
2993

2994
/**
2995 2996
 * css_rightmost_descendant - return the rightmost descendant of a css
 * @pos: css of interest
2997
 *
2998 2999
 * Return the rightmost descendant of @pos.  If there's no descendant, @pos
 * is returned.  This can be used during pre-order traversal to skip
3000
 * subtree of @pos.
3001
 *
3002 3003 3004 3005
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct rightmost descendant as
 * long as @pos is accessible.
3006
 */
3007 3008
struct cgroup_subsys_state *
css_rightmost_descendant(struct cgroup_subsys_state *pos)
3009
{
3010
	struct cgroup_subsys_state *last, *tmp;
3011

T
Tejun Heo 已提交
3012
	cgroup_assert_mutexes_or_rcu_locked();
3013 3014 3015 3016 3017

	do {
		last = pos;
		/* ->prev isn't RCU safe, walk ->next till the end */
		pos = NULL;
3018
		css_for_each_child(tmp, last)
3019 3020 3021 3022 3023
			pos = tmp;
	} while (pos);

	return last;
}
3024
EXPORT_SYMBOL_GPL(css_rightmost_descendant);
3025

3026 3027
static struct cgroup_subsys_state *
css_leftmost_descendant(struct cgroup_subsys_state *pos)
3028
{
3029
	struct cgroup_subsys_state *last;
3030 3031 3032

	do {
		last = pos;
3033
		pos = css_next_child(NULL, pos);
3034 3035 3036 3037 3038 3039
	} while (pos);

	return last;
}

/**
3040
 * css_next_descendant_post - find the next descendant for post-order walk
3041
 * @pos: the current position (%NULL to initiate traversal)
3042
 * @root: css whose descendants to walk
3043
 *
3044
 * To be used by css_for_each_descendant_post().  Find the next descendant
3045 3046
 * to visit for post-order traversal of @root's descendants.  @root is
 * included in the iteration and the last node to be visited.
3047
 *
3048 3049 3050 3051 3052
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct next descendant as long
 * as both @pos and @cgroup are accessible and @pos is a descendant of
 * @cgroup.
3053
 */
3054 3055 3056
struct cgroup_subsys_state *
css_next_descendant_post(struct cgroup_subsys_state *pos,
			 struct cgroup_subsys_state *root)
3057
{
3058
	struct cgroup_subsys_state *next;
3059

T
Tejun Heo 已提交
3060
	cgroup_assert_mutexes_or_rcu_locked();
3061

3062 3063 3064
	/* if first iteration, visit leftmost descendant which may be @root */
	if (!pos)
		return css_leftmost_descendant(root);
3065

3066 3067 3068 3069
	/* if we visited @root, we're done */
	if (pos == root)
		return NULL;

3070
	/* if there's an unvisited sibling, visit its leftmost descendant */
3071
	next = css_next_child(pos, css_parent(pos));
3072
	if (next)
3073
		return css_leftmost_descendant(next);
3074 3075

	/* no sibling left, visit parent */
3076
	return css_parent(pos);
3077
}
3078
EXPORT_SYMBOL_GPL(css_next_descendant_post);
3079

3080
/**
3081
 * css_advance_task_iter - advance a task itererator to the next css_set
3082 3083 3084
 * @it: the iterator to advance
 *
 * Advance @it to the next css_set to walk.
3085
 */
3086
static void css_advance_task_iter(struct css_task_iter *it)
3087 3088 3089 3090 3091 3092 3093 3094
{
	struct list_head *l = it->cset_link;
	struct cgrp_cset_link *link;
	struct css_set *cset;

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
3095
		if (l == &it->origin_css->cgroup->cset_links) {
3096 3097 3098 3099 3100 3101 3102 3103 3104 3105
			it->cset_link = NULL;
			return;
		}
		link = list_entry(l, struct cgrp_cset_link, cset_link);
		cset = link->cset;
	} while (list_empty(&cset->tasks));
	it->cset_link = l;
	it->task = cset->tasks.next;
}

3106
/**
3107 3108
 * css_task_iter_start - initiate task iteration
 * @css: the css to walk tasks of
3109 3110
 * @it: the task iterator to use
 *
3111 3112 3113 3114
 * Initiate iteration through the tasks of @css.  The caller can call
 * css_task_iter_next() to walk through the tasks until the function
 * returns NULL.  On completion of iteration, css_task_iter_end() must be
 * called.
3115 3116 3117 3118 3119
 *
 * Note that this function acquires a lock which is released when the
 * iteration finishes.  The caller can't sleep while iteration is in
 * progress.
 */
3120 3121
void css_task_iter_start(struct cgroup_subsys_state *css,
			 struct css_task_iter *it)
3122
	__acquires(css_set_lock)
3123 3124
{
	/*
3125 3126 3127
	 * The first time anyone tries to iterate across a css, we need to
	 * enable the list linking each css_set to its tasks, and fix up
	 * all existing tasks.
3128
	 */
3129 3130 3131
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

3132
	read_lock(&css_set_lock);
3133

3134 3135
	it->origin_css = css;
	it->cset_link = &css->cgroup->cset_links;
3136

3137
	css_advance_task_iter(it);
3138 3139
}

3140
/**
3141
 * css_task_iter_next - return the next task for the iterator
3142 3143 3144
 * @it: the task iterator being iterated
 *
 * The "next" function for task iteration.  @it should have been
3145 3146
 * initialized via css_task_iter_start().  Returns NULL when the iteration
 * reaches the end.
3147
 */
3148
struct task_struct *css_task_iter_next(struct css_task_iter *it)
3149 3150 3151
{
	struct task_struct *res;
	struct list_head *l = it->task;
3152
	struct cgrp_cset_link *link;
3153 3154

	/* If the iterator cg is NULL, we have no tasks */
3155
	if (!it->cset_link)
3156 3157 3158 3159
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
3160 3161
	link = list_entry(it->cset_link, struct cgrp_cset_link, cset_link);
	if (l == &link->cset->tasks) {
3162 3163 3164 3165
		/*
		 * We reached the end of this task list - move on to the
		 * next cgrp_cset_link.
		 */
3166
		css_advance_task_iter(it);
3167 3168 3169 3170 3171 3172
	} else {
		it->task = l;
	}
	return res;
}

3173
/**
3174
 * css_task_iter_end - finish task iteration
3175 3176
 * @it: the task iterator to finish
 *
3177
 * Finish task iteration started by css_task_iter_start().
3178
 */
3179
void css_task_iter_end(struct css_task_iter *it)
3180
	__releases(css_set_lock)
3181 3182 3183 3184
{
	read_unlock(&css_set_lock);
}

3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219
static inline int started_after_time(struct task_struct *t1,
				     struct timespec *time,
				     struct task_struct *t2)
{
	int start_diff = timespec_compare(&t1->start_time, time);
	if (start_diff > 0) {
		return 1;
	} else if (start_diff < 0) {
		return 0;
	} else {
		/*
		 * Arbitrarily, if two processes started at the same
		 * time, we'll say that the lower pointer value
		 * started first. Note that t2 may have exited by now
		 * so this may not be a valid pointer any longer, but
		 * that's fine - it still serves to distinguish
		 * between two tasks started (effectively) simultaneously.
		 */
		return t1 > t2;
	}
}

/*
 * This function is a callback from heap_insert() and is used to order
 * the heap.
 * In this case we order the heap in descending task start time.
 */
static inline int started_after(void *p1, void *p2)
{
	struct task_struct *t1 = p1;
	struct task_struct *t2 = p2;
	return started_after_time(t1, &t2->start_time, t2);
}

/**
3220 3221
 * css_scan_tasks - iterate though all the tasks in a css
 * @css: the css to iterate tasks of
T
Tejun Heo 已提交
3222 3223 3224 3225
 * @test: optional test callback
 * @process: process callback
 * @data: data passed to @test and @process
 * @heap: optional pre-allocated heap used for task iteration
3226
 *
3227 3228
 * Iterate through all the tasks in @css, calling @test for each, and if it
 * returns %true, call @process for it also.
3229
 *
T
Tejun Heo 已提交
3230
 * @test may be NULL, meaning always true (select all tasks), which
3231
 * effectively duplicates css_task_iter_{start,next,end}() but does not
T
Tejun Heo 已提交
3232 3233 3234
 * lock css_set_lock for the call to @process.
 *
 * It is guaranteed that @process will act on every task that is a member
3235 3236 3237
 * of @css for the duration of this call.  This function may or may not
 * call @process for tasks that exit or move to a different css during the
 * call, or are forked or move into the css during the call.
3238
 *
T
Tejun Heo 已提交
3239 3240 3241
 * Note that @test may be called with locks held, and may in some
 * situations be called multiple times for the same task, so it should be
 * cheap.
3242
 *
T
Tejun Heo 已提交
3243 3244 3245 3246
 * If @heap is non-NULL, a heap has been pre-allocated and will be used for
 * heap operations (and its "gt" member will be overwritten), else a
 * temporary heap will be used (allocation of which may cause this function
 * to fail).
3247
 */
3248 3249 3250 3251
int css_scan_tasks(struct cgroup_subsys_state *css,
		   bool (*test)(struct task_struct *, void *),
		   void (*process)(struct task_struct *, void *),
		   void *data, struct ptr_heap *heap)
3252 3253
{
	int retval, i;
3254
	struct css_task_iter it;
3255 3256 3257 3258 3259 3260
	struct task_struct *p, *dropped;
	/* Never dereference latest_task, since it's not refcounted */
	struct task_struct *latest_task = NULL;
	struct ptr_heap tmp_heap;
	struct timespec latest_time = { 0, 0 };

T
Tejun Heo 已提交
3261
	if (heap) {
3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274
		/* The caller supplied our heap and pre-allocated its memory */
		heap->gt = &started_after;
	} else {
		/* We need to allocate our own heap memory */
		heap = &tmp_heap;
		retval = heap_init(heap, PAGE_SIZE, GFP_KERNEL, &started_after);
		if (retval)
			/* cannot allocate the heap */
			return retval;
	}

 again:
	/*
3275
	 * Scan tasks in the css, using the @test callback to determine
T
Tejun Heo 已提交
3276 3277 3278 3279 3280 3281 3282
	 * which are of interest, and invoking @process callback on the
	 * ones which need an update.  Since we don't want to hold any
	 * locks during the task updates, gather tasks to be processed in a
	 * heap structure.  The heap is sorted by descending task start
	 * time.  If the statically-sized heap fills up, we overflow tasks
	 * that started later, and in future iterations only consider tasks
	 * that started after the latest task in the previous pass. This
3283 3284 3285
	 * guarantees forward progress and that we don't miss any tasks.
	 */
	heap->size = 0;
3286 3287
	css_task_iter_start(css, &it);
	while ((p = css_task_iter_next(&it))) {
3288 3289 3290 3291
		/*
		 * Only affect tasks that qualify per the caller's callback,
		 * if he provided one
		 */
T
Tejun Heo 已提交
3292
		if (test && !test(p, data))
3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319
			continue;
		/*
		 * Only process tasks that started after the last task
		 * we processed
		 */
		if (!started_after_time(p, &latest_time, latest_task))
			continue;
		dropped = heap_insert(heap, p);
		if (dropped == NULL) {
			/*
			 * The new task was inserted; the heap wasn't
			 * previously full
			 */
			get_task_struct(p);
		} else if (dropped != p) {
			/*
			 * The new task was inserted, and pushed out a
			 * different task
			 */
			get_task_struct(p);
			put_task_struct(dropped);
		}
		/*
		 * Else the new task was newer than anything already in
		 * the heap and wasn't inserted
		 */
	}
3320
	css_task_iter_end(&it);
3321 3322 3323

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3324
			struct task_struct *q = heap->ptrs[i];
3325
			if (i == 0) {
3326 3327
				latest_time = q->start_time;
				latest_task = q;
3328 3329
			}
			/* Process the task per the caller's callback */
T
Tejun Heo 已提交
3330
			process(q, data);
3331
			put_task_struct(q);
3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
		}
		/*
		 * If we had to process any tasks at all, scan again
		 * in case some of them were in the middle of forking
		 * children that didn't get processed.
		 * Not the most efficient way to do it, but it avoids
		 * having to take callback_mutex in the fork path
		 */
		goto again;
	}
	if (heap == &tmp_heap)
		heap_free(&tmp_heap);
	return 0;
}

T
Tejun Heo 已提交
3347
static void cgroup_transfer_one_task(struct task_struct *task, void *data)
3348
{
T
Tejun Heo 已提交
3349
	struct cgroup *new_cgroup = data;
3350

T
Tejun Heo 已提交
3351
	mutex_lock(&cgroup_mutex);
3352
	cgroup_attach_task(new_cgroup, task, false);
T
Tejun Heo 已提交
3353
	mutex_unlock(&cgroup_mutex);
3354 3355 3356 3357 3358 3359 3360 3361 3362
}

/**
 * cgroup_trasnsfer_tasks - move tasks from one cgroup to another
 * @to: cgroup to which the tasks will be moved
 * @from: cgroup in which the tasks currently reside
 */
int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from)
{
3363 3364
	return css_scan_tasks(&from->dummy_css, NULL, cgroup_transfer_one_task,
			      to, NULL);
3365 3366
}

3367
/*
3368
 * Stuff for reading the 'tasks'/'procs' files.
3369 3370 3371 3372 3373 3374 3375 3376
 *
 * Reading this file can return large amounts of data if a cgroup has
 * *lots* of attached tasks. So it may need several calls to read(),
 * but we cannot guarantee that the information we produce is correct
 * unless we produce it entirely atomically.
 *
 */

3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402
/* which pidlist file are we talking about? */
enum cgroup_filetype {
	CGROUP_FILE_PROCS,
	CGROUP_FILE_TASKS,
};

/*
 * A pidlist is a list of pids that virtually represents the contents of one
 * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists,
 * a pair (one each for procs, tasks) for each pid namespace that's relevant
 * to the cgroup.
 */
struct cgroup_pidlist {
	/*
	 * used to find which pidlist is wanted. doesn't change as long as
	 * this particular list stays in the list.
	*/
	struct { enum cgroup_filetype type; struct pid_namespace *ns; } key;
	/* array of xids */
	pid_t *list;
	/* how many elements the above list has */
	int length;
	/* each of these stored in a list by its cgroup */
	struct list_head links;
	/* pointer to the cgroup we belong to, for list removal purposes */
	struct cgroup *owner;
3403 3404
	/* for delayed destruction */
	struct delayed_work destroy_dwork;
3405 3406
};

3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419
/*
 * The following two functions "fix" the issue where there are more pids
 * than kmalloc will give memory for; in such cases, we use vmalloc/vfree.
 * TODO: replace with a kernel-wide solution to this problem
 */
#define PIDLIST_TOO_LARGE(c) ((c) * sizeof(pid_t) > (PAGE_SIZE * 2))
static void *pidlist_allocate(int count)
{
	if (PIDLIST_TOO_LARGE(count))
		return vmalloc(count * sizeof(pid_t));
	else
		return kmalloc(count * sizeof(pid_t), GFP_KERNEL);
}
3420

3421 3422 3423 3424 3425 3426 3427 3428
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}

3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455
/*
 * Used to destroy all pidlists lingering waiting for destroy timer.  None
 * should be left afterwards.
 */
static void cgroup_pidlist_destroy_all(struct cgroup *cgrp)
{
	struct cgroup_pidlist *l, *tmp_l;

	mutex_lock(&cgrp->pidlist_mutex);
	list_for_each_entry_safe(l, tmp_l, &cgrp->pidlists, links)
		mod_delayed_work(cgroup_pidlist_destroy_wq, &l->destroy_dwork, 0);
	mutex_unlock(&cgrp->pidlist_mutex);

	flush_workqueue(cgroup_pidlist_destroy_wq);
	BUG_ON(!list_empty(&cgrp->pidlists));
}

static void cgroup_pidlist_destroy_work_fn(struct work_struct *work)
{
	struct delayed_work *dwork = to_delayed_work(work);
	struct cgroup_pidlist *l = container_of(dwork, struct cgroup_pidlist,
						destroy_dwork);
	struct cgroup_pidlist *tofree = NULL;

	mutex_lock(&l->owner->pidlist_mutex);

	/*
3456 3457
	 * Destroy iff we didn't get queued again.  The state won't change
	 * as destroy_dwork can only be queued while locked.
3458
	 */
3459
	if (!delayed_work_pending(dwork)) {
3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
		list_del(&l->links);
		pidlist_free(l->list);
		put_pid_ns(l->key.ns);
		tofree = l;
	}

	mutex_unlock(&l->owner->pidlist_mutex);
	kfree(tofree);
}

3470
/*
3471
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3472
 * Returns the number of unique elements.
3473
 */
3474
static int pidlist_uniq(pid_t *list, int length)
3475
{
3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
	int src, dest = 1;

	/*
	 * we presume the 0th element is unique, so i starts at 1. trivial
	 * edge cases first; no work needs to be done for either
	 */
	if (length == 0 || length == 1)
		return length;
	/* src and dest walk down the list; dest counts unique elements */
	for (src = 1; src < length; src++) {
		/* find next unique element */
		while (list[src] == list[src-1]) {
			src++;
			if (src == length)
				goto after;
		}
		/* dest always points to where the next unique element goes */
		list[dest] = list[src];
		dest++;
	}
after:
	return dest;
}

3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532
/*
 * The two pid files - task and cgroup.procs - guaranteed that the result
 * is sorted, which forced this whole pidlist fiasco.  As pid order is
 * different per namespace, each namespace needs differently sorted list,
 * making it impossible to use, for example, single rbtree of member tasks
 * sorted by task pointer.  As pidlists can be fairly large, allocating one
 * per open file is dangerous, so cgroup had to implement shared pool of
 * pidlists keyed by cgroup and namespace.
 *
 * All this extra complexity was caused by the original implementation
 * committing to an entirely unnecessary property.  In the long term, we
 * want to do away with it.  Explicitly scramble sort order if
 * sane_behavior so that no such expectation exists in the new interface.
 *
 * Scrambling is done by swapping every two consecutive bits, which is
 * non-identity one-to-one mapping which disturbs sort order sufficiently.
 */
static pid_t pid_fry(pid_t pid)
{
	unsigned a = pid & 0x55555555;
	unsigned b = pid & 0xAAAAAAAA;

	return (a << 1) | (b >> 1);
}

static pid_t cgroup_pid_fry(struct cgroup *cgrp, pid_t pid)
{
	if (cgroup_sane_behavior(cgrp))
		return pid_fry(pid);
	else
		return pid;
}

3533 3534 3535 3536 3537
static int cmppid(const void *a, const void *b)
{
	return *(pid_t *)a - *(pid_t *)b;
}

3538 3539 3540 3541 3542
static int fried_cmppid(const void *a, const void *b)
{
	return pid_fry(*(pid_t *)a) - pid_fry(*(pid_t *)b);
}

T
Tejun Heo 已提交
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557
static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
						  enum cgroup_filetype type)
{
	struct cgroup_pidlist *l;
	/* don't need task_nsproxy() if we're looking at ourself */
	struct pid_namespace *ns = task_active_pid_ns(current);

	lockdep_assert_held(&cgrp->pidlist_mutex);

	list_for_each_entry(l, &cgrp->pidlists, links)
		if (l->key.type == type && l->key.ns == ns)
			return l;
	return NULL;
}

3558 3559 3560 3561 3562 3563
/*
 * find the appropriate pidlist for our purpose (given procs vs tasks)
 * returns with the lock on that pidlist already held, and takes care
 * of the use count, or returns NULL with no locks held if we're out of
 * memory.
 */
T
Tejun Heo 已提交
3564 3565
static struct cgroup_pidlist *cgroup_pidlist_find_create(struct cgroup *cgrp,
						enum cgroup_filetype type)
3566 3567
{
	struct cgroup_pidlist *l;
3568

T
Tejun Heo 已提交
3569 3570 3571 3572 3573 3574
	lockdep_assert_held(&cgrp->pidlist_mutex);

	l = cgroup_pidlist_find(cgrp, type);
	if (l)
		return l;

3575
	/* entry not found; create a new one */
3576
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
T
Tejun Heo 已提交
3577
	if (!l)
3578
		return l;
T
Tejun Heo 已提交
3579

3580
	INIT_DELAYED_WORK(&l->destroy_dwork, cgroup_pidlist_destroy_work_fn);
3581
	l->key.type = type;
T
Tejun Heo 已提交
3582 3583
	/* don't need task_nsproxy() if we're looking at ourself */
	l->key.ns = get_pid_ns(task_active_pid_ns(current));
3584 3585 3586 3587 3588
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	return l;
}

3589 3590 3591
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3592 3593
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3594 3595 3596 3597
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3598
	struct css_task_iter it;
3599
	struct task_struct *tsk;
3600 3601
	struct cgroup_pidlist *l;

3602 3603
	lockdep_assert_held(&cgrp->pidlist_mutex);

3604 3605 3606 3607 3608 3609 3610
	/*
	 * If cgroup gets more users after we read count, we won't have
	 * enough space - tough.  This race is indistinguishable to the
	 * caller from the case that the additional cgroup users didn't
	 * show up until sometime later on.
	 */
	length = cgroup_task_count(cgrp);
3611
	array = pidlist_allocate(length);
3612 3613 3614
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3615 3616
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
3617
		if (unlikely(n == length))
3618
			break;
3619
		/* get tgid or pid for procs or tasks file respectively */
3620 3621 3622 3623
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3624 3625
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3626
	}
3627
	css_task_iter_end(&it);
3628 3629
	length = n;
	/* now sort & (if procs) strip out duplicates */
3630 3631 3632 3633
	if (cgroup_sane_behavior(cgrp))
		sort(array, length, sizeof(pid_t), fried_cmppid, NULL);
	else
		sort(array, length, sizeof(pid_t), cmppid, NULL);
3634
	if (type == CGROUP_FILE_PROCS)
3635
		length = pidlist_uniq(array, length);
T
Tejun Heo 已提交
3636 3637

	l = cgroup_pidlist_find_create(cgrp, type);
3638
	if (!l) {
T
Tejun Heo 已提交
3639
		mutex_unlock(&cgrp->pidlist_mutex);
3640
		pidlist_free(array);
3641
		return -ENOMEM;
3642
	}
T
Tejun Heo 已提交
3643 3644

	/* store array, freeing old if necessary */
3645
	pidlist_free(l->list);
3646 3647
	l->list = array;
	l->length = length;
3648
	*lp = l;
3649
	return 0;
3650 3651
}

B
Balbir Singh 已提交
3652
/**
L
Li Zefan 已提交
3653
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3654 3655 3656
 * @stats: cgroupstats to fill information into
 * @dentry: A dentry entry belonging to the cgroup for which stats have
 * been requested.
L
Li Zefan 已提交
3657 3658 3659
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3660 3661 3662 3663
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3664
	struct cgroup *cgrp;
3665
	struct css_task_iter it;
B
Balbir Singh 已提交
3666
	struct task_struct *tsk;
3667

B
Balbir Singh 已提交
3668
	/*
3669 3670
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3671
	 */
3672 3673
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3674 3675 3676
		 goto err;

	ret = 0;
3677
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3678

3679 3680
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
B
Balbir Singh 已提交
3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699
		switch (tsk->state) {
		case TASK_RUNNING:
			stats->nr_running++;
			break;
		case TASK_INTERRUPTIBLE:
			stats->nr_sleeping++;
			break;
		case TASK_UNINTERRUPTIBLE:
			stats->nr_uninterruptible++;
			break;
		case TASK_STOPPED:
			stats->nr_stopped++;
			break;
		default:
			if (delayacct_is_task_waiting_on_io(tsk))
				stats->nr_io_wait++;
			break;
		}
	}
3700
	css_task_iter_end(&it);
B
Balbir Singh 已提交
3701 3702 3703 3704 3705

err:
	return ret;
}

3706

3707
/*
3708
 * seq_file methods for the tasks/procs files. The seq_file position is the
3709
 * next pid to display; the seq_file iterator is a pointer to the pid
3710
 * in the cgroup->l->list array.
3711
 */
3712

3713
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3714
{
3715 3716 3717 3718 3719 3720
	/*
	 * Initially we receive a position value that corresponds to
	 * one more than the last pid shown (or 0 on the first call or
	 * after a seek to the start). Use a binary-search to find the
	 * next pid to display, if any
	 */
3721
	struct cgroup_open_file *of = s->private;
3722
	struct cgroup *cgrp = seq_css(s)->cgroup;
3723
	struct cgroup_pidlist *l;
3724
	enum cgroup_filetype type = seq_cft(s)->private;
3725
	int index = 0, pid = *pos;
3726 3727 3728 3729 3730
	int *iter, ret;

	mutex_lock(&cgrp->pidlist_mutex);

	/*
3731
	 * !NULL @of->priv indicates that this isn't the first start()
3732
	 * after open.  If the matching pidlist is around, we can use that.
3733
	 * Look for it.  Note that @of->priv can't be used directly.  It
3734 3735
	 * could already have been destroyed.
	 */
3736 3737
	if (of->priv)
		of->priv = cgroup_pidlist_find(cgrp, type);
3738 3739 3740 3741 3742

	/*
	 * Either this is the first start() after open or the matching
	 * pidlist has been destroyed inbetween.  Create a new one.
	 */
3743 3744 3745
	if (!of->priv) {
		ret = pidlist_array_load(cgrp, type,
					 (struct cgroup_pidlist **)&of->priv);
3746 3747 3748
		if (ret)
			return ERR_PTR(ret);
	}
3749
	l = of->priv;
3750 3751

	if (pid) {
3752
		int end = l->length;
S
Stephen Rothwell 已提交
3753

3754 3755
		while (index < end) {
			int mid = (index + end) / 2;
3756
			if (cgroup_pid_fry(cgrp, l->list[mid]) == pid) {
3757 3758
				index = mid;
				break;
3759
			} else if (cgroup_pid_fry(cgrp, l->list[mid]) <= pid)
3760 3761 3762 3763 3764 3765
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3766
	if (index >= l->length)
3767 3768
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3769
	iter = l->list + index;
3770
	*pos = cgroup_pid_fry(cgrp, *iter);
3771 3772 3773
	return iter;
}

3774
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3775
{
3776 3777
	struct cgroup_open_file *of = s->private;
	struct cgroup_pidlist *l = of->priv;
3778

3779 3780
	if (l)
		mod_delayed_work(cgroup_pidlist_destroy_wq, &l->destroy_dwork,
3781
				 CGROUP_PIDLIST_DESTROY_DELAY);
3782
	mutex_unlock(&seq_css(s)->cgroup->pidlist_mutex);
3783 3784
}

3785
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3786
{
3787 3788
	struct cgroup_open_file *of = s->private;
	struct cgroup_pidlist *l = of->priv;
3789 3790
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3791 3792 3793 3794 3795 3796 3797 3798
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
3799
		*pos = cgroup_pid_fry(seq_css(s)->cgroup, *p);
3800 3801 3802 3803
		return p;
	}
}

3804
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3805 3806 3807
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3808

3809 3810 3811 3812 3813 3814 3815 3816 3817
/*
 * seq_operations functions for iterating on pidlists through seq_file -
 * independent of whether it's tasks or procs
 */
static const struct seq_operations cgroup_pidlist_seq_operations = {
	.start = cgroup_pidlist_start,
	.stop = cgroup_pidlist_stop,
	.next = cgroup_pidlist_next,
	.show = cgroup_pidlist_show,
3818 3819
};

3820 3821
static u64 cgroup_read_notify_on_release(struct cgroup_subsys_state *css,
					 struct cftype *cft)
3822
{
3823
	return notify_on_release(css->cgroup);
3824 3825
}

3826 3827
static int cgroup_write_notify_on_release(struct cgroup_subsys_state *css,
					  struct cftype *cft, u64 val)
3828
{
3829
	clear_bit(CGRP_RELEASABLE, &css->cgroup->flags);
3830
	if (val)
3831
		set_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3832
	else
3833
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3834 3835 3836
	return 0;
}

3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
/*
 * When dput() is called asynchronously, if umount has been done and
 * then deactivate_super() in cgroup_free_fn() kills the superblock,
 * there's a small window that vfs will see the root dentry with non-zero
 * refcnt and trigger BUG().
 *
 * That's why we hold a reference before dput() and drop it right after.
 */
static void cgroup_dput(struct cgroup *cgrp)
{
	struct super_block *sb = cgrp->root->sb;

	atomic_inc(&sb->s_active);
	dput(cgrp->dentry);
	deactivate_super(sb);
}

3854 3855
static u64 cgroup_clone_children_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
3856
{
3857
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3858 3859
}

3860 3861
static int cgroup_clone_children_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
3862 3863
{
	if (val)
3864
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3865
	else
3866
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3867 3868 3869
	return 0;
}

3870
static struct cftype cgroup_base_files[] = {
3871
	{
3872
		.name = "cgroup.procs",
3873 3874 3875 3876
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
3877
		.private = CGROUP_FILE_PROCS,
B
Ben Blum 已提交
3878 3879
		.write_u64 = cgroup_procs_write,
		.mode = S_IRUGO | S_IWUSR,
3880
	},
3881 3882
	{
		.name = "cgroup.clone_children",
3883
		.flags = CFTYPE_INSANE,
3884 3885 3886
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
3887 3888 3889
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
3890
		.seq_show = cgroup_sane_behavior_show,
3891
	},
3892 3893 3894 3895 3896 3897 3898 3899 3900

	/*
	 * Historical crazy stuff.  These don't have "cgroup."  prefix and
	 * don't exist if sane_behavior.  If you're depending on these, be
	 * prepared to be burned.
	 */
	{
		.name = "tasks",
		.flags = CFTYPE_INSANE,		/* use "procs" instead */
3901 3902 3903 3904
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
3905
		.private = CGROUP_FILE_TASKS,
3906 3907 3908 3909 3910 3911 3912 3913 3914
		.write_u64 = cgroup_tasks_write,
		.mode = S_IRUGO | S_IWUSR,
	},
	{
		.name = "notify_on_release",
		.flags = CFTYPE_INSANE,
		.read_u64 = cgroup_read_notify_on_release,
		.write_u64 = cgroup_write_notify_on_release,
	},
3915 3916
	{
		.name = "release_agent",
3917
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
3918
		.seq_show = cgroup_release_agent_show,
3919
		.write_string = cgroup_release_agent_write,
3920
		.max_write_len = PATH_MAX - 1,
3921
	},
T
Tejun Heo 已提交
3922
	{ }	/* terminate */
3923 3924
};

3925
/**
3926
 * cgroup_populate_dir - create subsys files in a cgroup directory
3927 3928
 * @cgrp: target cgroup
 * @subsys_mask: mask of the subsystem ids whose files should be added
3929 3930
 *
 * On failure, no file is added.
3931
 */
3932
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned long subsys_mask)
3933 3934
{
	struct cgroup_subsys *ss;
3935
	int i, ret = 0;
3936

3937
	/* process cftsets of each subsystem */
3938
	for_each_subsys(ss, i) {
3939
		struct cftype_set *set;
3940 3941

		if (!test_bit(i, &subsys_mask))
3942
			continue;
3943

3944
		list_for_each_entry(set, &ss->cftsets, node) {
3945
			ret = cgroup_addrm_files(cgrp, set->cfts, true);
3946 3947 3948
			if (ret < 0)
				goto err;
		}
3949 3950
	}
	return 0;
3951 3952 3953
err:
	cgroup_clear_dir(cgrp, subsys_mask);
	return ret;
3954 3955
}

3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977
/*
 * css destruction is four-stage process.
 *
 * 1. Destruction starts.  Killing of the percpu_ref is initiated.
 *    Implemented in kill_css().
 *
 * 2. When the percpu_ref is confirmed to be visible as killed on all CPUs
 *    and thus css_tryget() is guaranteed to fail, the css can be offlined
 *    by invoking offline_css().  After offlining, the base ref is put.
 *    Implemented in css_killed_work_fn().
 *
 * 3. When the percpu_ref reaches zero, the only possible remaining
 *    accessors are inside RCU read sections.  css_release() schedules the
 *    RCU callback.
 *
 * 4. After the grace period, the css can be freed.  Implemented in
 *    css_free_work_fn().
 *
 * It is actually hairier because both step 2 and 4 require process context
 * and thus involve punting to css->destroy_work adding two additional
 * steps to the already complex sequence.
 */
3978
static void css_free_work_fn(struct work_struct *work)
3979 3980
{
	struct cgroup_subsys_state *css =
3981
		container_of(work, struct cgroup_subsys_state, destroy_work);
3982
	struct cgroup *cgrp = css->cgroup;
3983

3984 3985 3986
	if (css->parent)
		css_put(css->parent);

3987 3988
	css->ss->css_free(css);
	cgroup_dput(cgrp);
3989 3990
}

3991
static void css_free_rcu_fn(struct rcu_head *rcu_head)
3992 3993
{
	struct cgroup_subsys_state *css =
3994
		container_of(rcu_head, struct cgroup_subsys_state, rcu_head);
3995

3996 3997
	/*
	 * css holds an extra ref to @cgrp->dentry which is put on the last
3998
	 * css_put().  dput() requires process context which we don't have.
3999 4000
	 */
	INIT_WORK(&css->destroy_work, css_free_work_fn);
4001
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4002 4003
}

4004 4005 4006 4007 4008
static void css_release(struct percpu_ref *ref)
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

4009
	rcu_assign_pointer(css->cgroup->subsys[css->ss->id], NULL);
4010
	call_rcu(&css->rcu_head, css_free_rcu_fn);
4011 4012
}

4013 4014
static void init_css(struct cgroup_subsys_state *css, struct cgroup_subsys *ss,
		     struct cgroup *cgrp)
4015
{
4016
	css->cgroup = cgrp;
4017
	css->ss = ss;
4018
	css->flags = 0;
4019 4020

	if (cgrp->parent)
4021
		css->parent = cgroup_css(cgrp->parent, ss);
4022
	else
4023
		css->flags |= CSS_ROOT;
4024

4025
	BUG_ON(cgroup_css(cgrp, ss));
4026 4027
}

4028
/* invoke ->css_online() on a new CSS and mark it online if successful */
4029
static int online_css(struct cgroup_subsys_state *css)
4030
{
4031
	struct cgroup_subsys *ss = css->ss;
T
Tejun Heo 已提交
4032 4033
	int ret = 0;

T
Tejun Heo 已提交
4034
	lockdep_assert_held(&cgroup_tree_mutex);
4035 4036
	lockdep_assert_held(&cgroup_mutex);

4037
	if (ss->css_online)
4038
		ret = ss->css_online(css);
4039
	if (!ret) {
4040
		css->flags |= CSS_ONLINE;
4041
		css->cgroup->nr_css++;
4042
		rcu_assign_pointer(css->cgroup->subsys[ss->id], css);
4043
	}
T
Tejun Heo 已提交
4044
	return ret;
4045 4046
}

4047
/* if the CSS is online, invoke ->css_offline() on it and mark it offline */
4048
static void offline_css(struct cgroup_subsys_state *css)
4049
{
4050
	struct cgroup_subsys *ss = css->ss;
4051

T
Tejun Heo 已提交
4052
	lockdep_assert_held(&cgroup_tree_mutex);
4053 4054 4055 4056 4057
	lockdep_assert_held(&cgroup_mutex);

	if (!(css->flags & CSS_ONLINE))
		return;

4058
	if (ss->css_offline)
4059
		ss->css_offline(css);
4060

4061
	css->flags &= ~CSS_ONLINE;
4062
	css->cgroup->nr_css--;
4063
	RCU_INIT_POINTER(css->cgroup->subsys[ss->id], css);
4064 4065
}

4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
/**
 * create_css - create a cgroup_subsys_state
 * @cgrp: the cgroup new css will be associated with
 * @ss: the subsys of new css
 *
 * Create a new css associated with @cgrp - @ss pair.  On success, the new
 * css is online and installed in @cgrp with all interface files created.
 * Returns 0 on success, -errno on failure.
 */
static int create_css(struct cgroup *cgrp, struct cgroup_subsys *ss)
{
	struct cgroup *parent = cgrp->parent;
	struct cgroup_subsys_state *css;
	int err;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

	css = ss->css_alloc(cgroup_css(parent, ss));
	if (IS_ERR(css))
		return PTR_ERR(css);

	err = percpu_ref_init(&css->refcnt, css_release);
	if (err)
		goto err_free;

	init_css(css, ss, cgrp);

4094
	err = cgroup_populate_dir(cgrp, 1 << ss->id);
4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121
	if (err)
		goto err_free;

	err = online_css(css);
	if (err)
		goto err_free;

	dget(cgrp->dentry);
	css_get(css->parent);

	if (ss->broken_hierarchy && !ss->warned_broken_hierarchy &&
	    parent->parent) {
		pr_warning("cgroup: %s (%d) created nested cgroup for controller \"%s\" which has incomplete hierarchy support. Nested cgroups may change behavior in the future.\n",
			   current->comm, current->pid, ss->name);
		if (!strcmp(ss->name, "memory"))
			pr_warning("cgroup: \"memory\" requires setting use_hierarchy to 1 on the root.\n");
		ss->warned_broken_hierarchy = true;
	}

	return 0;

err_free:
	percpu_ref_cancel_init(&css->refcnt);
	ss->css_free(css);
	return err;
}

4122
/*
L
Li Zefan 已提交
4123 4124 4125 4126
 * cgroup_create - create a cgroup
 * @parent: cgroup that will be parent of the new cgroup
 * @dentry: dentry of the new cgroup
 * @mode: mode to set on new inode
4127
 *
L
Li Zefan 已提交
4128
 * Must be called with the mutex on the parent inode held
4129 4130
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
4131
			     umode_t mode)
4132
{
4133
	struct cgroup *cgrp;
4134
	struct cgroup_name *name;
4135
	struct cgroupfs_root *root = parent->root;
4136
	int ssid, err;
4137 4138 4139
	struct cgroup_subsys *ss;
	struct super_block *sb = root->sb;

T
Tejun Heo 已提交
4140
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4141 4142
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4143 4144
		return -ENOMEM;

T
Tejun Heo 已提交
4145
	name = cgroup_alloc_name(dentry->d_name.name);
4146 4147
	if (!name) {
		err = -ENOMEM;
4148
		goto err_free_cgrp;
4149
	}
4150 4151
	rcu_assign_pointer(cgrp->name, name);

T
Tejun Heo 已提交
4152 4153
	mutex_lock(&cgroup_tree_mutex);

4154 4155 4156 4157 4158 4159 4160 4161 4162
	/*
	 * Only live parents can have children.  Note that the liveliness
	 * check isn't strictly necessary because cgroup_mkdir() and
	 * cgroup_rmdir() are fully synchronized by i_mutex; however, do it
	 * anyway so that locking is contained inside cgroup proper and we
	 * don't get nasty surprises if we ever grow another caller.
	 */
	if (!cgroup_lock_live_group(parent)) {
		err = -ENODEV;
T
Tejun Heo 已提交
4163
		goto err_unlock_tree;
4164 4165 4166 4167 4168 4169 4170 4171 4172 4173
	}

	/*
	 * Temporarily set the pointer to NULL, so idr_find() won't return
	 * a half-baked cgroup.
	 */
	cgrp->id = idr_alloc(&root->cgroup_idr, NULL, 1, 0, GFP_KERNEL);
	if (cgrp->id < 0) {
		err = -ENOMEM;
		goto err_unlock;
4174 4175
	}

4176 4177 4178 4179 4180 4181 4182
	/* Grab a reference on the superblock so the hierarchy doesn't
	 * get deleted on unmount if there are child cgroups.  This
	 * can be done outside cgroup_mutex, since the sb can't
	 * disappear while someone has an open control file on the
	 * fs */
	atomic_inc(&sb->s_active);

4183
	init_cgroup_housekeeping(cgrp);
4184

4185 4186 4187
	dentry->d_fsdata = cgrp;
	cgrp->dentry = dentry;

4188
	cgrp->parent = parent;
4189
	cgrp->dummy_css.parent = &parent->dummy_css;
4190
	cgrp->root = parent->root;
4191

4192 4193 4194
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4195 4196
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4197

4198 4199 4200 4201 4202
	/*
	 * Create directory.  cgroup_create_file() returns with the new
	 * directory locked on success so that it can be populated without
	 * dropping cgroup_mutex.
	 */
T
Tejun Heo 已提交
4203
	err = cgroup_create_file(dentry, S_IFDIR | mode, sb);
4204
	if (err < 0)
4205
		goto err_free_id;
4206
	lockdep_assert_held(&dentry->d_inode->i_mutex);
4207

4208
	cgrp->serial_nr = cgroup_serial_nr_next++;
4209

4210 4211 4212
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
	root->number_of_cgroups++;
T
Tejun Heo 已提交
4213

4214 4215 4216
	/* hold a ref to the parent's dentry */
	dget(parent->dentry);

4217 4218 4219 4220
	/*
	 * @cgrp is now fully operational.  If something fails after this
	 * point, it'll be released via the normal destruction path.
	 */
4221 4222
	idr_replace(&root->cgroup_idr, cgrp, cgrp->id);

4223
	err = cgroup_addrm_files(cgrp, cgroup_base_files, true);
4224 4225 4226
	if (err)
		goto err_destroy;

4227
	/* let's create and online css's */
T
Tejun Heo 已提交
4228 4229 4230 4231 4232 4233
	for_each_subsys(ss, ssid) {
		if (root->subsys_mask & (1 << ssid)) {
			err = create_css(cgrp, ss);
			if (err)
				goto err_destroy;
		}
4234
	}
4235 4236

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4237
	mutex_unlock(&cgroup_tree_mutex);
4238
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4239 4240 4241

	return 0;

T
Tejun Heo 已提交
4242
err_free_id:
4243
	idr_remove(&root->cgroup_idr, cgrp->id);
4244 4245 4246 4247
	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);
err_unlock:
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4248 4249
err_unlock_tree:
	mutex_unlock(&cgroup_tree_mutex);
4250
	kfree(rcu_dereference_raw(cgrp->name));
4251
err_free_cgrp:
4252
	kfree(cgrp);
4253
	return err;
4254 4255 4256 4257

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4258
	mutex_unlock(&cgroup_tree_mutex);
4259 4260
	mutex_unlock(&dentry->d_inode->i_mutex);
	return err;
4261 4262
}

4263
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4264 4265 4266 4267 4268 4269 4270
{
	struct cgroup *c_parent = dentry->d_parent->d_fsdata;

	/* the vfs holds inode->i_mutex already */
	return cgroup_create(c_parent, dentry, mode | S_IFDIR);
}

4271 4272 4273 4274 4275
/*
 * This is called when the refcnt of a css is confirmed to be killed.
 * css_tryget() is now guaranteed to fail.
 */
static void css_killed_work_fn(struct work_struct *work)
4276
{
4277 4278 4279
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, destroy_work);
	struct cgroup *cgrp = css->cgroup;
4280

T
Tejun Heo 已提交
4281
	mutex_lock(&cgroup_tree_mutex);
4282 4283
	mutex_lock(&cgroup_mutex);

4284 4285 4286 4287 4288 4289
	/*
	 * css_tryget() is guaranteed to fail now.  Tell subsystems to
	 * initate destruction.
	 */
	offline_css(css);

4290 4291 4292 4293 4294
	/*
	 * If @cgrp is marked dead, it's waiting for refs of all css's to
	 * be disabled before proceeding to the second phase of cgroup
	 * destruction.  If we are the last one, kick it off.
	 */
4295
	if (!cgrp->nr_css && cgroup_is_dead(cgrp))
4296 4297 4298
		cgroup_destroy_css_killed(cgrp);

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4299
	mutex_unlock(&cgroup_tree_mutex);
4300 4301 4302 4303 4304 4305 4306 4307 4308

	/*
	 * Put the css refs from kill_css().  Each css holds an extra
	 * reference to the cgroup's dentry and cgroup removal proceeds
	 * regardless of css refs.  On the last put of each css, whenever
	 * that may be, the extra dentry ref is put so that dentry
	 * destruction happens only after all css's are released.
	 */
	css_put(css);
4309 4310
}

4311 4312
/* css kill confirmation processing requires process context, bounce */
static void css_killed_ref_fn(struct percpu_ref *ref)
4313 4314 4315 4316
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

4317
	INIT_WORK(&css->destroy_work, css_killed_work_fn);
4318
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4319 4320
}

T
Tejun Heo 已提交
4321 4322 4323 4324
/**
 * kill_css - destroy a css
 * @css: css to destroy
 *
4325 4326 4327 4328
 * This function initiates destruction of @css by removing cgroup interface
 * files and putting its base reference.  ->css_offline() will be invoked
 * asynchronously once css_tryget() is guaranteed to fail and when the
 * reference count reaches zero, @css will be released.
T
Tejun Heo 已提交
4329 4330 4331
 */
static void kill_css(struct cgroup_subsys_state *css)
{
4332
	cgroup_clear_dir(css->cgroup, 1 << css->ss->id);
4333

T
Tejun Heo 已提交
4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350
	/*
	 * Killing would put the base ref, but we need to keep it alive
	 * until after ->css_offline().
	 */
	css_get(css);

	/*
	 * cgroup core guarantees that, by the time ->css_offline() is
	 * invoked, no new css reference will be given out via
	 * css_tryget().  We can't simply call percpu_ref_kill() and
	 * proceed to offlining css's because percpu_ref_kill() doesn't
	 * guarantee that the ref is seen as killed on all CPUs on return.
	 *
	 * Use percpu_ref_kill_and_confirm() to get notifications as each
	 * css is confirmed to be seen as killed on all CPUs.
	 */
	percpu_ref_kill_and_confirm(&css->refcnt, css_killed_ref_fn);
4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376
}

/**
 * cgroup_destroy_locked - the first stage of cgroup destruction
 * @cgrp: cgroup to be destroyed
 *
 * css's make use of percpu refcnts whose killing latency shouldn't be
 * exposed to userland and are RCU protected.  Also, cgroup core needs to
 * guarantee that css_tryget() won't succeed by the time ->css_offline() is
 * invoked.  To satisfy all the requirements, destruction is implemented in
 * the following two steps.
 *
 * s1. Verify @cgrp can be destroyed and mark it dying.  Remove all
 *     userland visible parts and start killing the percpu refcnts of
 *     css's.  Set up so that the next stage will be kicked off once all
 *     the percpu refcnts are confirmed to be killed.
 *
 * s2. Invoke ->css_offline(), mark the cgroup dead and proceed with the
 *     rest of destruction.  Once all cgroup references are gone, the
 *     cgroup is RCU-freed.
 *
 * This function implements s1.  After this step, @cgrp is gone as far as
 * the userland is concerned and a new cgroup with the same name may be
 * created.  As cgroup doesn't care about the names internally, this
 * doesn't cause any problem.
 */
4377 4378
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4379
{
4380
	struct dentry *d = cgrp->dentry;
T
Tejun Heo 已提交
4381
	struct cgroup_subsys_state *css;
4382
	struct cgroup *child;
4383
	bool empty;
T
Tejun Heo 已提交
4384
	int ssid;
4385

4386
	lockdep_assert_held(&d->d_inode->i_mutex);
T
Tejun Heo 已提交
4387
	lockdep_assert_held(&cgroup_tree_mutex);
4388 4389
	lockdep_assert_held(&cgroup_mutex);

4390
	/*
T
Tejun Heo 已提交
4391 4392
	 * css_set_lock synchronizes access to ->cset_links and prevents
	 * @cgrp from being removed while __put_css_set() is in progress.
4393 4394
	 */
	read_lock(&css_set_lock);
4395
	empty = list_empty(&cgrp->cset_links);
4396 4397
	read_unlock(&css_set_lock);
	if (!empty)
4398
		return -EBUSY;
L
Li Zefan 已提交
4399

4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415
	/*
	 * Make sure there's no live children.  We can't test ->children
	 * emptiness as dead children linger on it while being destroyed;
	 * otherwise, "rmdir parent/child parent" may fail with -EBUSY.
	 */
	empty = true;
	rcu_read_lock();
	list_for_each_entry_rcu(child, &cgrp->children, sibling) {
		empty = cgroup_is_dead(child);
		if (!empty)
			break;
	}
	rcu_read_unlock();
	if (!empty)
		return -EBUSY;

4416
	/*
T
Tejun Heo 已提交
4417 4418
	 * Initiate massacre of all css's.  cgroup_destroy_css_killed()
	 * will be invoked to perform the rest of destruction once the
4419 4420
	 * percpu refs of all css's are confirmed to be killed.  This
	 * involves removing the subsystem's files, drop cgroup_mutex.
4421
	 */
4422
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4423 4424
	for_each_css(css, ssid, cgrp)
		kill_css(css);
4425
	mutex_lock(&cgroup_mutex);
4426 4427 4428 4429

	/*
	 * Mark @cgrp dead.  This prevents further task migration and child
	 * creation by disabling cgroup_lock_live_group().  Note that
4430
	 * CGRP_DEAD assertion is depended upon by css_next_child() to
4431
	 * resume iteration after dropping RCU read lock.  See
4432
	 * css_next_child() for details.
4433
	 */
4434
	set_bit(CGRP_DEAD, &cgrp->flags);
4435

4436 4437 4438 4439 4440 4441 4442
	/* CGRP_DEAD is set, remove from ->release_list for the last time */
	raw_spin_lock(&release_list_lock);
	if (!list_empty(&cgrp->release_list))
		list_del_init(&cgrp->release_list);
	raw_spin_unlock(&release_list_lock);

	/*
4443 4444 4445 4446 4447 4448 4449 4450
	 * If @cgrp has css's attached, the second stage of cgroup
	 * destruction is kicked off from css_killed_work_fn() after the
	 * refs of all attached css's are killed.  If @cgrp doesn't have
	 * any css, we kick it off here.
	 */
	if (!cgrp->nr_css)
		cgroup_destroy_css_killed(cgrp);

4451
	/*
4452 4453 4454
	 * Clear the base files and remove @cgrp directory.  The removal
	 * puts the base ref but we aren't quite done with @cgrp yet, so
	 * hold onto it.
4455
	 */
4456
	mutex_unlock(&cgroup_mutex);
4457
	cgroup_addrm_files(cgrp, cgroup_base_files, false);
4458 4459
	dget(d);
	cgroup_d_remove_dir(d);
4460
	mutex_lock(&cgroup_mutex);
4461

4462 4463 4464
	return 0;
};

4465
/**
4466
 * cgroup_destroy_css_killed - the second step of cgroup destruction
4467 4468 4469
 * @work: cgroup->destroy_free_work
 *
 * This function is invoked from a work item for a cgroup which is being
4470 4471 4472
 * destroyed after all css's are offlined and performs the rest of
 * destruction.  This is the second step of destruction described in the
 * comment above cgroup_destroy_locked().
4473
 */
4474
static void cgroup_destroy_css_killed(struct cgroup *cgrp)
4475 4476 4477 4478
{
	struct cgroup *parent = cgrp->parent;
	struct dentry *d = cgrp->dentry;

T
Tejun Heo 已提交
4479
	lockdep_assert_held(&cgroup_tree_mutex);
4480
	lockdep_assert_held(&cgroup_mutex);
4481

4482
	/* delete this cgroup from parent->children */
4483
	list_del_rcu(&cgrp->sibling);
4484

4485 4486
	dput(d);

4487
	set_bit(CGRP_RELEASABLE, &parent->flags);
4488
	check_for_release(parent);
4489 4490
}

4491 4492 4493 4494
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry)
{
	int ret;

T
Tejun Heo 已提交
4495
	mutex_lock(&cgroup_tree_mutex);
4496 4497 4498
	mutex_lock(&cgroup_mutex);
	ret = cgroup_destroy_locked(dentry->d_fsdata);
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4499
	mutex_unlock(&cgroup_tree_mutex);
4500 4501 4502 4503

	return ret;
}

4504
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4505 4506
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4507 4508

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4509

T
Tejun Heo 已提交
4510
	mutex_lock(&cgroup_tree_mutex);
4511 4512
	mutex_lock(&cgroup_mutex);

4513
	INIT_LIST_HEAD(&ss->cftsets);
4514

4515
	/* Create the top cgroup state for this subsystem */
4516
	ss->root = &cgroup_dummy_root;
4517
	css = ss->css_alloc(cgroup_css(cgroup_dummy_top, ss));
4518 4519
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4520
	init_css(css, ss, cgroup_dummy_top);
4521

L
Li Zefan 已提交
4522
	/* Update the init_css_set to contain a subsys
4523
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4524 4525
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
4526
	init_css_set.subsys[ss->id] = css;
4527 4528 4529

	need_forkexit_callback |= ss->fork || ss->exit;

L
Li Zefan 已提交
4530 4531 4532 4533 4534
	/* At system boot, before all subsystems have been
	 * registered, no tasks have been forked, so we don't
	 * need to invoke fork callbacks here. */
	BUG_ON(!list_empty(&init_task.tasks));

4535
	BUG_ON(online_css(css));
4536

4537
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4538
	mutex_unlock(&cgroup_tree_mutex);
4539 4540
}

4541
/**
L
Li Zefan 已提交
4542 4543 4544 4545
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4546 4547 4548
 */
int __init cgroup_init_early(void)
{
4549
	struct cgroup_subsys *ss;
4550
	int i;
4551

4552
	atomic_set(&init_css_set.refcount, 1);
4553
	INIT_LIST_HEAD(&init_css_set.cgrp_links);
4554
	INIT_LIST_HEAD(&init_css_set.tasks);
4555
	INIT_HLIST_NODE(&init_css_set.hlist);
4556
	css_set_count = 1;
4557 4558
	init_cgroup_root(&cgroup_dummy_root);
	cgroup_root_count = 1;
4559
	RCU_INIT_POINTER(init_task.cgroups, &init_css_set);
4560

4561
	init_cgrp_cset_link.cset = &init_css_set;
4562 4563
	init_cgrp_cset_link.cgrp = cgroup_dummy_top;
	list_add(&init_cgrp_cset_link.cset_link, &cgroup_dummy_top->cset_links);
4564
	list_add(&init_cgrp_cset_link.cgrp_link, &init_css_set.cgrp_links);
4565

T
Tejun Heo 已提交
4566
	for_each_subsys(ss, i) {
4567
		WARN(!ss->css_alloc || !ss->css_free || ss->name || ss->id,
4568 4569
		     "invalid cgroup_subsys %d:%s css_alloc=%p css_free=%p name:id=%d:%s\n",
		     i, cgroup_subsys_name[i], ss->css_alloc, ss->css_free,
4570
		     ss->id, ss->name);
4571 4572 4573
		WARN(strlen(cgroup_subsys_name[i]) > MAX_CGROUP_TYPE_NAMELEN,
		     "cgroup_subsys_name %s too long\n", cgroup_subsys_name[i]);

4574
		ss->id = i;
4575
		ss->name = cgroup_subsys_name[i];
4576 4577 4578 4579 4580 4581 4582 4583

		if (ss->early_init)
			cgroup_init_subsys(ss);
	}
	return 0;
}

/**
L
Li Zefan 已提交
4584 4585 4586 4587
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4588 4589 4590
 */
int __init cgroup_init(void)
{
4591
	struct cgroup_subsys *ss;
4592
	unsigned long key;
4593
	int i, err;
4594 4595 4596 4597

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4598

T
Tejun Heo 已提交
4599
	for_each_subsys(ss, i) {
4600 4601
		if (!ss->early_init)
			cgroup_init_subsys(ss);
4602 4603 4604 4605 4606 4607 4608

		/*
		 * cftype registration needs kmalloc and can't be done
		 * during early_init.  Register base cftypes separately.
		 */
		if (ss->base_cftypes)
			WARN_ON(cgroup_add_cftypes(ss, ss->base_cftypes));
4609 4610
	}

4611
	/* allocate id for the dummy hierarchy */
T
Tejun Heo 已提交
4612 4613
	mutex_lock(&cgroup_mutex);

4614 4615 4616 4617
	/* Add init_css_set to the hash table */
	key = css_set_hash(init_css_set.subsys);
	hash_add(css_set_table, &init_css_set.hlist, key);

4618
	BUG_ON(cgroup_init_root_id(&cgroup_dummy_root, 0, 1));
4619

4620 4621 4622 4623
	err = idr_alloc(&cgroup_dummy_root.cgroup_idr, cgroup_dummy_top,
			0, 1, GFP_KERNEL);
	BUG_ON(err < 0);

T
Tejun Heo 已提交
4624 4625
	mutex_unlock(&cgroup_mutex);

4626 4627 4628 4629 4630 4631
	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4632
	err = register_filesystem(&cgroup_fs_type);
4633 4634
	if (err < 0) {
		kobject_put(cgroup_kobj);
4635
		goto out;
4636
	}
4637

L
Li Zefan 已提交
4638
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4639

4640
out:
4641 4642 4643
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4644 4645
	return err;
}
4646

4647 4648 4649 4650 4651
static int __init cgroup_wq_init(void)
{
	/*
	 * There isn't much point in executing destruction path in
	 * parallel.  Good chunk is serialized with cgroup_mutex anyway.
4652 4653 4654 4655 4656
	 *
	 * XXX: Must be ordered to make sure parent is offlined after
	 * children.  The ordering requirement is for memcg where a
	 * parent's offline may wait for a child's leading to deadlock.  In
	 * the long term, this should be fixed from memcg side.
4657 4658 4659 4660
	 *
	 * We would prefer to do this in cgroup_init() above, but that
	 * is called before init_workqueues(): so leave this until after.
	 */
4661
	cgroup_destroy_wq = alloc_ordered_workqueue("cgroup_destroy", 0);
4662
	BUG_ON(!cgroup_destroy_wq);
4663 4664 4665 4666 4667 4668 4669 4670 4671

	/*
	 * Used to destroy pidlists and separate to serve as flush domain.
	 * Cap @max_active to 1 too.
	 */
	cgroup_pidlist_destroy_wq = alloc_workqueue("cgroup_pidlist_destroy",
						    0, 1);
	BUG_ON(!cgroup_pidlist_destroy_wq);

4672 4673 4674 4675
	return 0;
}
core_initcall(cgroup_wq_init);

4676 4677 4678 4679 4680 4681
/*
 * proc_cgroup_show()
 *  - Print task's cgroup paths into seq_file, one line for each hierarchy
 *  - Used for /proc/<pid>/cgroup.
 *  - No need to task_lock(tsk) on this tsk->cgroup reference, as it
 *    doesn't really matter if tsk->cgroup changes after we read it,
4682
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4683 4684 4685 4686 4687 4688
 *    anyway.  No need to check that tsk->cgroup != NULL, thanks to
 *    the_top_cgroup_hack in cgroup_exit(), which sets an exiting tasks
 *    cgroup to top_cgroup.
 */

/* TODO: Use a proper seq_file iterator */
4689
int proc_cgroup_show(struct seq_file *m, void *v)
4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711
{
	struct pid *pid;
	struct task_struct *tsk;
	char *buf;
	int retval;
	struct cgroupfs_root *root;

	retval = -ENOMEM;
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
		goto out;

	retval = -ESRCH;
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
	if (!tsk)
		goto out_free;

	retval = 0;

	mutex_lock(&cgroup_mutex);

4712
	for_each_active_root(root) {
4713
		struct cgroup_subsys *ss;
4714
		struct cgroup *cgrp;
T
Tejun Heo 已提交
4715
		int ssid, count = 0;
4716

4717
		seq_printf(m, "%d:", root->hierarchy_id);
T
Tejun Heo 已提交
4718 4719 4720
		for_each_subsys(ss, ssid)
			if (root->subsys_mask & (1 << ssid))
				seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4721 4722 4723
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4724
		seq_putc(m, ':');
4725
		cgrp = task_cgroup_from_root(tsk, root);
4726
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744
		if (retval < 0)
			goto out_unlock;
		seq_puts(m, buf);
		seq_putc(m, '\n');
	}

out_unlock:
	mutex_unlock(&cgroup_mutex);
	put_task_struct(tsk);
out_free:
	kfree(buf);
out:
	return retval;
}

/* Display information about each subsystem and each hierarchy */
static int proc_cgroupstats_show(struct seq_file *m, void *v)
{
4745
	struct cgroup_subsys *ss;
4746 4747
	int i;

4748
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4749 4750 4751 4752 4753
	/*
	 * ideally we don't want subsystems moving around while we do this.
	 * cgroup_mutex is also necessary to guarantee an atomic snapshot of
	 * subsys/hierarchy state.
	 */
4754
	mutex_lock(&cgroup_mutex);
4755 4756

	for_each_subsys(ss, i)
4757 4758
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
4759
			   ss->root->number_of_cgroups, !ss->disabled);
4760

4761 4762 4763 4764 4765 4766
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
4767
	return single_open(file, proc_cgroupstats_show, NULL);
4768 4769
}

4770
static const struct file_operations proc_cgroupstats_operations = {
4771 4772 4773 4774 4775 4776
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

4777 4778
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
4779
 * @child: pointer to task_struct of forking parent process.
4780 4781 4782 4783 4784
 *
 * Description: A task inherits its parent's cgroup at fork().
 *
 * A pointer to the shared css_set was automatically copied in
 * fork.c by dup_task_struct().  However, we ignore that copy, since
4785 4786 4787 4788
 * it was not made under the protection of RCU or cgroup_mutex, so
 * might no longer be a valid cgroup pointer.  cgroup_attach_task() might
 * have already changed current->cgroups, allowing the previously
 * referenced cgroup group to be removed and freed.
4789 4790 4791 4792 4793 4794
 *
 * At the point that cgroup_fork() is called, 'current' is the parent
 * task, and the passed argument 'child' points to the child task.
 */
void cgroup_fork(struct task_struct *child)
{
4795
	task_lock(current);
4796
	get_css_set(task_css_set(current));
4797
	child->cgroups = current->cgroups;
4798
	task_unlock(current);
4799
	INIT_LIST_HEAD(&child->cg_list);
4800 4801
}

4802
/**
L
Li Zefan 已提交
4803 4804 4805
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
4806 4807 4808
 * Adds the task to the list running through its css_set if necessary and
 * call the subsystem fork() callbacks.  Has to be after the task is
 * visible on the task list in case we race with the first call to
4809
 * cgroup_task_iter_start() - to guarantee that the new task ends up on its
4810
 * list.
L
Li Zefan 已提交
4811
 */
4812 4813
void cgroup_post_fork(struct task_struct *child)
{
4814
	struct cgroup_subsys *ss;
4815 4816
	int i;

4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827
	/*
	 * use_task_css_set_links is set to 1 before we walk the tasklist
	 * under the tasklist_lock and we read it here after we added the child
	 * to the tasklist under the tasklist_lock as well. If the child wasn't
	 * yet in the tasklist when we walked through it from
	 * cgroup_enable_task_cg_lists(), then use_task_css_set_links value
	 * should be visible now due to the paired locking and barriers implied
	 * by LOCK/UNLOCK: it is written before the tasklist_lock unlock
	 * in cgroup_enable_task_cg_lists() and read here after the tasklist_lock
	 * lock on fork.
	 */
4828 4829
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
4830 4831
		task_lock(child);
		if (list_empty(&child->cg_list))
4832
			list_add(&child->cg_list, &task_css_set(child)->tasks);
4833
		task_unlock(child);
4834 4835
		write_unlock(&css_set_lock);
	}
4836 4837 4838 4839 4840 4841 4842

	/*
	 * Call ss->fork().  This must happen after @child is linked on
	 * css_set; otherwise, @child might change state between ->fork()
	 * and addition to css_set.
	 */
	if (need_forkexit_callback) {
T
Tejun Heo 已提交
4843
		for_each_subsys(ss, i)
4844 4845 4846
			if (ss->fork)
				ss->fork(child);
	}
4847
}
4848

4849 4850 4851
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
4852
 * @run_callback: run exit callbacks?
4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880
 *
 * Description: Detach cgroup from @tsk and release it.
 *
 * Note that cgroups marked notify_on_release force every task in
 * them to take the global cgroup_mutex mutex when exiting.
 * This could impact scaling on very large systems.  Be reluctant to
 * use notify_on_release cgroups where very high task exit scaling
 * is required on large systems.
 *
 * the_top_cgroup_hack:
 *
 *    Set the exiting tasks cgroup to the root cgroup (top_cgroup).
 *
 *    We call cgroup_exit() while the task is still competent to
 *    handle notify_on_release(), then leave the task attached to the
 *    root cgroup in each hierarchy for the remainder of its exit.
 *
 *    To do this properly, we would increment the reference count on
 *    top_cgroup, and near the very end of the kernel/exit.c do_exit()
 *    code we would add a second cgroup function call, to drop that
 *    reference.  This would just create an unnecessary hot spot on
 *    the top_cgroup reference count, to no avail.
 *
 *    Normally, holding a reference to a cgroup without bumping its
 *    count is unsafe.   The cgroup could go away, or someone could
 *    attach us to a different cgroup, decrementing the count on
 *    the first cgroup that we never incremented.  But in this case,
 *    top_cgroup isn't going away, and either task has PF_EXITING set,
4881 4882
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
4883 4884 4885
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
4886
	struct cgroup_subsys *ss;
4887
	struct css_set *cset;
4888
	int i;
4889 4890 4891 4892 4893 4894 4895 4896 4897

	/*
	 * Unlink from the css_set task list if necessary.
	 * Optimistically check cg_list before taking
	 * css_set_lock
	 */
	if (!list_empty(&tsk->cg_list)) {
		write_lock(&css_set_lock);
		if (!list_empty(&tsk->cg_list))
4898
			list_del_init(&tsk->cg_list);
4899 4900 4901
		write_unlock(&css_set_lock);
	}

4902 4903
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
4904 4905
	cset = task_css_set(tsk);
	RCU_INIT_POINTER(tsk->cgroups, &init_css_set);
4906 4907

	if (run_callbacks && need_forkexit_callback) {
T
Tejun Heo 已提交
4908 4909
		/* see cgroup_post_fork() for details */
		for_each_subsys(ss, i) {
4910
			if (ss->exit) {
4911 4912
				struct cgroup_subsys_state *old_css = cset->subsys[i];
				struct cgroup_subsys_state *css = task_css(tsk, i);
4913

4914
				ss->exit(css, old_css, tsk);
4915 4916 4917
			}
		}
	}
4918
	task_unlock(tsk);
4919

4920
	put_css_set_taskexit(cset);
4921
}
4922

4923
static void check_for_release(struct cgroup *cgrp)
4924
{
4925
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
4926
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
4927 4928
		/*
		 * Control Group is currently removeable. If it's not
4929
		 * already queued for a userspace notification, queue
4930 4931
		 * it now
		 */
4932
		int need_schedule_work = 0;
4933

4934
		raw_spin_lock(&release_list_lock);
4935
		if (!cgroup_is_dead(cgrp) &&
4936 4937
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
4938 4939
			need_schedule_work = 1;
		}
4940
		raw_spin_unlock(&release_list_lock);
4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972
		if (need_schedule_work)
			schedule_work(&release_agent_work);
	}
}

/*
 * Notify userspace when a cgroup is released, by running the
 * configured release agent with the name of the cgroup (path
 * relative to the root of cgroup file system) as the argument.
 *
 * Most likely, this user command will try to rmdir this cgroup.
 *
 * This races with the possibility that some other task will be
 * attached to this cgroup before it is removed, or that some other
 * user task will 'mkdir' a child cgroup of this cgroup.  That's ok.
 * The presumed 'rmdir' will fail quietly if this cgroup is no longer
 * unused, and this cgroup will be reprieved from its death sentence,
 * to continue to serve a useful existence.  Next time it's released,
 * we will get notified again, if it still has 'notify_on_release' set.
 *
 * The final arg to call_usermodehelper() is UMH_WAIT_EXEC, which
 * means only wait until the task is successfully execve()'d.  The
 * separate release agent task is forked by call_usermodehelper(),
 * then control in this thread returns here, without waiting for the
 * release agent task.  We don't bother to wait because the caller of
 * this routine has no use for the exit status of the release agent
 * task, so no sense holding our caller up for that.
 */
static void cgroup_release_agent(struct work_struct *work)
{
	BUG_ON(work != &release_agent_work);
	mutex_lock(&cgroup_mutex);
4973
	raw_spin_lock(&release_list_lock);
4974 4975 4976
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
4977
		char *pathbuf = NULL, *agentbuf = NULL;
4978
		struct cgroup *cgrp = list_entry(release_list.next,
4979 4980
						    struct cgroup,
						    release_list);
4981
		list_del_init(&cgrp->release_list);
4982
		raw_spin_unlock(&release_list_lock);
4983
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
4984 4985 4986 4987 4988 4989 4990
		if (!pathbuf)
			goto continue_free;
		if (cgroup_path(cgrp, pathbuf, PAGE_SIZE) < 0)
			goto continue_free;
		agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
		if (!agentbuf)
			goto continue_free;
4991 4992

		i = 0;
4993 4994
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008
		argv[i] = NULL;

		i = 0;
		/* minimal command environment */
		envp[i++] = "HOME=/";
		envp[i++] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
		envp[i] = NULL;

		/* Drop the lock while we invoke the usermode helper,
		 * since the exec could involve hitting disk and hence
		 * be a slow process */
		mutex_unlock(&cgroup_mutex);
		call_usermodehelper(argv[0], argv, envp, UMH_WAIT_EXEC);
		mutex_lock(&cgroup_mutex);
5009 5010 5011
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5012
		raw_spin_lock(&release_list_lock);
5013
	}
5014
	raw_spin_unlock(&release_list_lock);
5015 5016
	mutex_unlock(&cgroup_mutex);
}
5017 5018 5019

static int __init cgroup_disable(char *str)
{
5020
	struct cgroup_subsys *ss;
5021
	char *token;
5022
	int i;
5023 5024 5025 5026

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5027

T
Tejun Heo 已提交
5028
		for_each_subsys(ss, i) {
5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039
			if (!strcmp(token, ss->name)) {
				ss->disabled = 1;
				printk(KERN_INFO "Disabling %s control group"
					" subsystem\n", ss->name);
				break;
			}
		}
	}
	return 1;
}
__setup("cgroup_disable=", cgroup_disable);
K
KAMEZAWA Hiroyuki 已提交
5040

5041
/**
5042
 * css_tryget_from_dir - get corresponding css from the dentry of a cgroup dir
5043 5044
 * @dentry: directory dentry of interest
 * @ss: subsystem of interest
5045
 *
5046 5047 5048
 * If @dentry is a directory for a cgroup which has @ss enabled on it, try
 * to get the corresponding css and return it.  If such css doesn't exist
 * or can't be pinned, an ERR_PTR value is returned.
S
Stephane Eranian 已提交
5049
 */
5050 5051
struct cgroup_subsys_state *css_tryget_from_dir(struct dentry *dentry,
						struct cgroup_subsys *ss)
S
Stephane Eranian 已提交
5052 5053
{
	struct cgroup *cgrp;
5054
	struct cgroup_subsys_state *css;
5055

5056 5057 5058
	/* is @dentry a cgroup dir? */
	if (!dentry->d_inode ||
	    dentry->d_inode->i_op != &cgroup_dir_inode_operations)
S
Stephane Eranian 已提交
5059 5060
		return ERR_PTR(-EBADF);

5061 5062
	rcu_read_lock();

5063
	cgrp = __d_cgrp(dentry);
5064 5065 5066 5067 5068 5069 5070
	css = cgroup_css(cgrp, ss);

	if (!css || !css_tryget(css))
		css = ERR_PTR(-ENOENT);

	rcu_read_unlock();
	return css;
S
Stephane Eranian 已提交
5071 5072
}

5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084
/**
 * css_from_id - lookup css by id
 * @id: the cgroup id
 * @ss: cgroup subsys to be looked into
 *
 * Returns the css if there's valid one with @id, otherwise returns NULL.
 * Should be called under rcu_read_lock().
 */
struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss)
{
	struct cgroup *cgrp;

T
Tejun Heo 已提交
5085
	cgroup_assert_mutexes_or_rcu_locked();
5086 5087 5088

	cgrp = idr_find(&ss->root->cgroup_idr, id);
	if (cgrp)
5089
		return cgroup_css(cgrp, ss);
5090
	return NULL;
S
Stephane Eranian 已提交
5091 5092
}

5093
#ifdef CONFIG_CGROUP_DEBUG
5094 5095
static struct cgroup_subsys_state *
debug_css_alloc(struct cgroup_subsys_state *parent_css)
5096 5097 5098 5099 5100 5101 5102 5103 5104
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

	if (!css)
		return ERR_PTR(-ENOMEM);

	return css;
}

5105
static void debug_css_free(struct cgroup_subsys_state *css)
5106
{
5107
	kfree(css);
5108 5109
}

5110 5111
static u64 debug_taskcount_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5112
{
5113
	return cgroup_task_count(css->cgroup);
5114 5115
}

5116 5117
static u64 current_css_set_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5118 5119 5120 5121
{
	return (u64)(unsigned long)current->cgroups;
}

5122
static u64 current_css_set_refcount_read(struct cgroup_subsys_state *css,
L
Li Zefan 已提交
5123
					 struct cftype *cft)
5124 5125 5126 5127
{
	u64 count;

	rcu_read_lock();
5128
	count = atomic_read(&task_css_set(current)->refcount);
5129 5130 5131 5132
	rcu_read_unlock();
	return count;
}

5133
static int current_css_set_cg_links_read(struct seq_file *seq, void *v)
5134
{
5135
	struct cgrp_cset_link *link;
5136
	struct css_set *cset;
5137 5138 5139

	read_lock(&css_set_lock);
	rcu_read_lock();
5140
	cset = rcu_dereference(current->cgroups);
5141
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5142 5143 5144 5145 5146 5147 5148
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5149 5150
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5151 5152 5153 5154 5155 5156 5157
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
5158
static int cgroup_css_links_read(struct seq_file *seq, void *v)
5159
{
5160
	struct cgroup_subsys_state *css = seq_css(seq);
5161
	struct cgrp_cset_link *link;
5162 5163

	read_lock(&css_set_lock);
5164
	list_for_each_entry(link, &css->cgroup->cset_links, cset_link) {
5165
		struct css_set *cset = link->cset;
5166 5167
		struct task_struct *task;
		int count = 0;
5168 5169
		seq_printf(seq, "css_set %p\n", cset);
		list_for_each_entry(task, &cset->tasks, cg_list) {
5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182
			if (count++ > MAX_TASKS_SHOWN_PER_CSS) {
				seq_puts(seq, "  ...\n");
				break;
			} else {
				seq_printf(seq, "  task %d\n",
					   task_pid_vnr(task));
			}
		}
	}
	read_unlock(&css_set_lock);
	return 0;
}

5183
static u64 releasable_read(struct cgroup_subsys_state *css, struct cftype *cft)
5184
{
5185
	return test_bit(CGRP_RELEASABLE, &css->cgroup->flags);
5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203
}

static struct cftype debug_files[] =  {
	{
		.name = "taskcount",
		.read_u64 = debug_taskcount_read,
	},

	{
		.name = "current_css_set",
		.read_u64 = current_css_set_read,
	},

	{
		.name = "current_css_set_refcount",
		.read_u64 = current_css_set_refcount_read,
	},

5204 5205
	{
		.name = "current_css_set_cg_links",
5206
		.seq_show = current_css_set_cg_links_read,
5207 5208 5209 5210
	},

	{
		.name = "cgroup_css_links",
5211
		.seq_show = cgroup_css_links_read,
5212 5213
	},

5214 5215 5216 5217 5218
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5219 5220
	{ }	/* terminate */
};
5221

5222
struct cgroup_subsys debug_cgrp_subsys = {
5223 5224
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5225
	.base_cftypes = debug_files,
5226 5227
};
#endif /* CONFIG_CGROUP_DEBUG */