cgroup.c 151.1 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 48
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/magic.h>
#include <linux/spinlock.h>
#include <linux/string.h>
49
#include <linux/sort.h>
50
#include <linux/kmod.h>
51
#include <linux/module.h>
B
Balbir Singh 已提交
52 53
#include <linux/delayacct.h>
#include <linux/cgroupstats.h>
54
#include <linux/hashtable.h>
55
#include <linux/namei.h>
L
Li Zefan 已提交
56
#include <linux/pid_namespace.h>
57
#include <linux/idr.h>
58
#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
59 60
#include <linux/eventfd.h>
#include <linux/poll.h>
61
#include <linux/flex_array.h> /* used in cgroup_attach_task */
62
#include <linux/kthread.h>
B
Balbir Singh 已提交
63

A
Arun Sharma 已提交
64
#include <linux/atomic.h>
65

T
Tejun Heo 已提交
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
/*
 * cgroup_mutex is the master lock.  Any modification to cgroup or its
 * hierarchy must be performed while holding it.
 *
 * cgroup_root_mutex nests inside cgroup_mutex and should be held to modify
 * cgroupfs_root of any cgroup hierarchy - subsys list, flags,
 * release_agent_path and so on.  Modifying requires both cgroup_mutex and
 * cgroup_root_mutex.  Readers can acquire either of the two.  This is to
 * break the following locking order cycle.
 *
 *  A. cgroup_mutex -> cred_guard_mutex -> s_type->i_mutex_key -> namespace_sem
 *  B. namespace_sem -> cgroup_mutex
 *
 * B happens only through cgroup_show_options() and using cgroup_root_mutex
 * breaks it.
 */
T
Tejun Heo 已提交
82 83 84 85
#ifdef CONFIG_PROVE_RCU
DEFINE_MUTEX(cgroup_mutex);
EXPORT_SYMBOL_GPL(cgroup_mutex);	/* only for task_subsys_state_check() */
#else
86
static DEFINE_MUTEX(cgroup_mutex);
T
Tejun Heo 已提交
87 88
#endif

T
Tejun Heo 已提交
89
static DEFINE_MUTEX(cgroup_root_mutex);
90

B
Ben Blum 已提交
91 92
/*
 * Generate an array of cgroup subsystem pointers. At boot time, this is
93
 * populated with the built in subsystems, and modular subsystems are
B
Ben Blum 已提交
94 95 96
 * registered after that. The mutable section of this array is protected by
 * cgroup_mutex.
 */
97
#define SUBSYS(_x) [_x ## _subsys_id] = &_x ## _subsys,
98
#define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option)
99
static struct cgroup_subsys *cgroup_subsys[CGROUP_SUBSYS_COUNT] = {
100 101 102 103
#include <linux/cgroup_subsys.h>
};

/*
104 105 106
 * 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.
107
 */
108 109 110 111
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;
112

T
Tejun Heo 已提交
113 114 115 116 117 118 119
/*
 * cgroupfs file entry, pointed to from leaf dentry->d_fsdata.
 */
struct cfent {
	struct list_head		node;
	struct dentry			*dentry;
	struct cftype			*type;
L
Li Zefan 已提交
120 121 122

	/* file xattrs */
	struct simple_xattrs		xattrs;
T
Tejun Heo 已提交
123 124
};

K
KAMEZAWA Hiroyuki 已提交
125 126 127 128 129 130 131 132 133 134
/*
 * CSS ID -- ID per subsys's Cgroup Subsys State(CSS). used only when
 * cgroup_subsys->use_id != 0.
 */
#define CSS_ID_MAX	(65535)
struct css_id {
	/*
	 * The css to which this ID points. This pointer is set to valid value
	 * after cgroup is populated. If cgroup is removed, this will be NULL.
	 * This pointer is expected to be RCU-safe because destroy()
T
Tejun Heo 已提交
135 136
	 * is called after synchronize_rcu(). But for safe use, css_tryget()
	 * should be used for avoiding race.
K
KAMEZAWA Hiroyuki 已提交
137
	 */
A
Arnd Bergmann 已提交
138
	struct cgroup_subsys_state __rcu *css;
K
KAMEZAWA Hiroyuki 已提交
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
	/*
	 * ID of this css.
	 */
	unsigned short id;
	/*
	 * Depth in hierarchy which this ID belongs to.
	 */
	unsigned short depth;
	/*
	 * ID is freed by RCU. (and lookup routine is RCU safe.)
	 */
	struct rcu_head rcu_head;
	/*
	 * Hierarchy of CSS ID belongs to.
	 */
	unsigned short stack[0]; /* Array of Length (depth+1) */
};

157
/*
L
Lucas De Marchi 已提交
158
 * cgroup_event represents events which userspace want to receive.
159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185
 */
struct cgroup_event {
	/*
	 * Cgroup which the event belongs to.
	 */
	struct cgroup *cgrp;
	/*
	 * Control file which the event associated.
	 */
	struct cftype *cft;
	/*
	 * eventfd to signal userspace about the event.
	 */
	struct eventfd_ctx *eventfd;
	/*
	 * Each of these stored in a list by the cgroup.
	 */
	struct list_head list;
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
	wait_queue_t wait;
	struct work_struct remove;
};
K
KAMEZAWA Hiroyuki 已提交
186

187 188
/* The list of hierarchy roots */

189 190
static LIST_HEAD(cgroup_roots);
static int cgroup_root_count;
191

T
Tejun Heo 已提交
192 193 194 195 196
/*
 * Hierarchy ID allocation and mapping.  It follows the same exclusion
 * rules as other root ops - both cgroup_mutex and cgroup_root_mutex for
 * writes, either for reads.
 */
197
static DEFINE_IDR(cgroup_hierarchy_idr);
198

199 200
static struct cgroup_name root_cgroup_name = { .name = "/" };

201 202 203 204 205
/*
 * 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
206 207
 * the ascending serial number order on the list.  Protected by
 * cgroup_mutex.
208
 */
209
static u64 cgroup_serial_nr_next = 1;
210

211
/* This flag indicates whether tasks in the fork and exit paths should
L
Li Zefan 已提交
212 213 214
 * 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.
215
 */
216
static int need_forkexit_callback __read_mostly;
217

218
static void cgroup_offline_fn(struct work_struct *work);
219
static int cgroup_destroy_locked(struct cgroup *cgrp);
220 221
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
			      struct cftype cfts[], bool is_add);
222

223
/* convenient tests for these bits */
224
static inline bool cgroup_is_dead(const struct cgroup *cgrp)
225
{
226
	return test_bit(CGRP_DEAD, &cgrp->flags);
227 228
}

229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
/**
 * 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);
248

249
static int cgroup_is_releasable(const struct cgroup *cgrp)
250 251
{
	const int bits =
252 253 254
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
255 256
}

257
static int notify_on_release(const struct cgroup *cgrp)
258
{
259
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
260 261
}

262 263 264
/* iterate each subsystem attached to a hierarchy */
#define for_each_root_subsys(root, ss)					\
	list_for_each_entry((ss), &(root)->subsys_list, sibling)
265

266 267 268
/* iterate across the active hierarchies */
#define for_each_active_root(root)					\
	list_for_each_entry((root), &cgroup_roots, root_list)
269

270 271 272 273 274
static inline struct cgroup *__d_cgrp(struct dentry *dentry)
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
275
static inline struct cfent *__d_cfe(struct dentry *dentry)
276 277 278 279
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
280 281 282 283 284
static inline struct cftype *__d_cft(struct dentry *dentry)
{
	return __d_cfe(dentry)->type;
}

285 286 287 288
/**
 * 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
289 290
 * On success, returns true; the mutex should be later unlocked.  On
 * failure returns false with no lock held.
291
 */
292
static bool cgroup_lock_live_group(struct cgroup *cgrp)
293 294
{
	mutex_lock(&cgroup_mutex);
295
	if (cgroup_is_dead(cgrp)) {
296 297 298 299 300 301
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}

302 303 304
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
305
static DEFINE_RAW_SPINLOCK(release_list_lock);
306 307
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
308
static void check_for_release(struct cgroup *cgrp);
309

310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327
/*
 * 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;
328 329 330 331 332 333 334 335 336 337
};

/* 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;
338
static struct cgrp_cset_link init_cgrp_cset_link;
339

340 341
static int cgroup_init_idr(struct cgroup_subsys *ss,
			   struct cgroup_subsys_state *css);
K
KAMEZAWA Hiroyuki 已提交
342

343 344 345 346 347 348
/* 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 cgroup_iter_start() */
static DEFINE_RWLOCK(css_set_lock);
static int css_set_count;

349 350 351 352 353
/*
 * 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.
 */
354
#define CSS_SET_HASH_BITS	7
355
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
356

357
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
358 359
{
	int i;
360
	unsigned long key = 0UL;
361 362

	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
363 364
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
365

366
	return key;
367 368
}

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

375
static void __put_css_set(struct css_set *cset, int taskexit)
376
{
377
	struct cgrp_cset_link *link, *tmp_link;
378

379 380 381 382 383
	/*
	 * Ensure that the refcount doesn't hit zero while any readers
	 * can see it. Similar to atomic_dec_and_lock(), but for an
	 * rwlock
	 */
384
	if (atomic_add_unless(&cset->refcount, -1, 1))
385 386
		return;
	write_lock(&css_set_lock);
387
	if (!atomic_dec_and_test(&cset->refcount)) {
388 389 390
		write_unlock(&css_set_lock);
		return;
	}
391

392
	/* This css_set is dead. unlink it and release cgroup refcounts */
393
	hash_del(&cset->hlist);
394 395
	css_set_count--;

396
	list_for_each_entry_safe(link, tmp_link, &cset->cgrp_links, cgrp_link) {
397
		struct cgroup *cgrp = link->cgrp;
398

399 400
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
L
Li Zefan 已提交
401

402
		/* @cgrp can't go away while we're holding css_set_lock */
T
Tejun Heo 已提交
403
		if (list_empty(&cgrp->cset_links) && notify_on_release(cgrp)) {
404
			if (taskexit)
405 406
				set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
407
		}
408 409

		kfree(link);
410
	}
411 412

	write_unlock(&css_set_lock);
413
	kfree_rcu(cset, rcu_head);
414 415
}

416 417 418
/*
 * refcounted get/put for css_set objects
 */
419
static inline void get_css_set(struct css_set *cset)
420
{
421
	atomic_inc(&cset->refcount);
422 423
}

424
static inline void put_css_set(struct css_set *cset)
425
{
426
	__put_css_set(cset, 0);
427 428
}

429
static inline void put_css_set_taskexit(struct css_set *cset)
430
{
431
	__put_css_set(cset, 1);
432 433
}

434
/**
435
 * compare_css_sets - helper function for find_existing_css_set().
436 437
 * @cset: candidate css_set being tested
 * @old_cset: existing css_set for a task
438 439 440 441 442 443
 * @new_cgrp: cgroup that's being entered by the task
 * @template: desired set of css pointers in css_set (pre-calculated)
 *
 * Returns true if "cg" matches "old_cg" except for the hierarchy
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
444 445
static bool compare_css_sets(struct css_set *cset,
			     struct css_set *old_cset,
446 447 448 449 450
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

451
	if (memcmp(template, cset->subsys, sizeof(cset->subsys))) {
452 453 454 455 456 457 458 459 460 461 462 463 464
		/* 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.
	 */

465 466
	l1 = &cset->cgrp_links;
	l2 = &old_cset->cgrp_links;
467
	while (1) {
468
		struct cgrp_cset_link *link1, *link2;
469
		struct cgroup *cgrp1, *cgrp2;
470 471 472 473

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
474 475
		if (l1 == &cset->cgrp_links) {
			BUG_ON(l2 != &old_cset->cgrp_links);
476 477
			break;
		} else {
478
			BUG_ON(l2 == &old_cset->cgrp_links);
479 480
		}
		/* Locate the cgroups associated with these links. */
481 482 483 484
		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;
485
		/* Hierarchies should be linked in the same order. */
486
		BUG_ON(cgrp1->root != cgrp2->root);
487 488 489 490 491 492 493 494

		/*
		 * 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.
		 */
495 496
		if (cgrp1->root == new_cgrp->root) {
			if (cgrp1 != new_cgrp)
497 498
				return false;
		} else {
499
			if (cgrp1 != cgrp2)
500 501 502 503 504 505
				return false;
		}
	}
	return true;
}

506 507 508 509 510
/**
 * 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
511
 */
512 513 514
static struct css_set *find_existing_css_set(struct css_set *old_cset,
					struct cgroup *cgrp,
					struct cgroup_subsys_state *template[])
515
{
516
	struct cgroupfs_root *root = cgrp->root;
517
	struct css_set *cset;
518
	unsigned long key;
519
	int i;
520

B
Ben Blum 已提交
521 522 523 524 525
	/*
	 * 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.
	 */
526
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
527
		if (root->subsys_mask & (1UL << i)) {
528 529 530
			/* Subsystem is in this hierarchy. So we want
			 * the subsystem state from the new
			 * cgroup */
531
			template[i] = cgrp->subsys[i];
532 533 534
		} else {
			/* Subsystem is not in this hierarchy, so we
			 * don't want to change the subsystem state */
535
			template[i] = old_cset->subsys[i];
536 537 538
		}
	}

539
	key = css_set_hash(template);
540 541
	hash_for_each_possible(css_set_table, cset, hlist, key) {
		if (!compare_css_sets(cset, old_cset, cgrp, template))
542 543 544
			continue;

		/* This css_set matches what we need */
545
		return cset;
546
	}
547 548 549 550 551

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

552
static void free_cgrp_cset_links(struct list_head *links_to_free)
553
{
554
	struct cgrp_cset_link *link, *tmp_link;
555

556 557
	list_for_each_entry_safe(link, tmp_link, links_to_free, cset_link) {
		list_del(&link->cset_link);
558 559 560 561
		kfree(link);
	}
}

562 563 564 565 566 567 568
/**
 * 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.
569
 */
570
static int allocate_cgrp_cset_links(int count, struct list_head *tmp_links)
571
{
572
	struct cgrp_cset_link *link;
573
	int i;
574 575 576

	INIT_LIST_HEAD(tmp_links);

577
	for (i = 0; i < count; i++) {
578
		link = kzalloc(sizeof(*link), GFP_KERNEL);
579
		if (!link) {
580
			free_cgrp_cset_links(tmp_links);
581 582
			return -ENOMEM;
		}
583
		list_add(&link->cset_link, tmp_links);
584 585 586 587
	}
	return 0;
}

588 589
/**
 * link_css_set - a helper function to link a css_set to a cgroup
590
 * @tmp_links: cgrp_cset_link objects allocated by allocate_cgrp_cset_links()
591
 * @cset: the css_set to be linked
592 593
 * @cgrp: the destination cgroup
 */
594 595
static void link_css_set(struct list_head *tmp_links, struct css_set *cset,
			 struct cgroup *cgrp)
596
{
597
	struct cgrp_cset_link *link;
598

599 600 601
	BUG_ON(list_empty(tmp_links));
	link = list_first_entry(tmp_links, struct cgrp_cset_link, cset_link);
	link->cset = cset;
602
	link->cgrp = cgrp;
603
	list_move(&link->cset_link, &cgrp->cset_links);
604 605 606 607
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
608
	list_add_tail(&link->cgrp_link, &cset->cgrp_links);
609 610
}

611 612 613 614 615 616 617
/**
 * 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.
618
 */
619 620
static struct css_set *find_css_set(struct css_set *old_cset,
				    struct cgroup *cgrp)
621
{
622
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT] = { };
623
	struct css_set *cset;
624 625
	struct list_head tmp_links;
	struct cgrp_cset_link *link;
626
	unsigned long key;
627

628 629
	lockdep_assert_held(&cgroup_mutex);

630 631
	/* First see if we already have a cgroup group that matches
	 * the desired set */
632
	read_lock(&css_set_lock);
633 634 635
	cset = find_existing_css_set(old_cset, cgrp, template);
	if (cset)
		get_css_set(cset);
636
	read_unlock(&css_set_lock);
637

638 639
	if (cset)
		return cset;
640

641
	cset = kzalloc(sizeof(*cset), GFP_KERNEL);
642
	if (!cset)
643 644
		return NULL;

645
	/* Allocate all the cgrp_cset_link objects that we'll need */
646
	if (allocate_cgrp_cset_links(cgroup_root_count, &tmp_links) < 0) {
647
		kfree(cset);
648 649 650
		return NULL;
	}

651
	atomic_set(&cset->refcount, 1);
652
	INIT_LIST_HEAD(&cset->cgrp_links);
653 654
	INIT_LIST_HEAD(&cset->tasks);
	INIT_HLIST_NODE(&cset->hlist);
655 656 657

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
658
	memcpy(cset->subsys, template, sizeof(cset->subsys));
659 660 661

	write_lock(&css_set_lock);
	/* Add reference counts and links from the new css_set. */
662
	list_for_each_entry(link, &old_cset->cgrp_links, cgrp_link) {
663
		struct cgroup *c = link->cgrp;
664

665 666
		if (c->root == cgrp->root)
			c = cgrp;
667
		link_css_set(&tmp_links, cset, c);
668
	}
669

670
	BUG_ON(!list_empty(&tmp_links));
671 672

	css_set_count++;
673 674

	/* Add this cgroup group to the hash table */
675 676
	key = css_set_hash(cset->subsys);
	hash_add(css_set_table, &cset->hlist, key);
677

678 679
	write_unlock(&css_set_lock);

680
	return cset;
681 682
}

683 684 685 686 687 688 689
/*
 * 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)
{
690
	struct css_set *cset;
691 692 693 694 695 696 697 698 699
	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.
	 */
700 701
	cset = task->cgroups;
	if (cset == &init_css_set) {
702 703
		res = &root->top_cgroup;
	} else {
704 705 706
		struct cgrp_cset_link *link;

		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
707
			struct cgroup *c = link->cgrp;
708

709 710 711 712 713 714 715 716 717 718 719
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
	read_unlock(&css_set_lock);
	BUG_ON(!res);
	return res;
}

720 721 722 723 724 725 726 727 728 729
/*
 * 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
730
 * cgroup_attach_task() can increment it again.  Because a count of zero
731 732 733 734 735 736 737 738 739 740 741 742 743
 * 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
744 745
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
746 747 748 749 750 751 752 753 754 755 756
 *
 * 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
757
 * cgroup_attach_task(), which overwrites one task's cgroup pointer with
L
Li Zefan 已提交
758
 * another.  It does so using cgroup_mutex, however there are
759 760 761
 * 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
762
 * in cgroup_attach_task(), modifying a task's cgroup pointer we use
763 764 765 766
 * 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
767
 * update of a tasks cgroup pointer by cgroup_attach_task()
768 769 770 771 772 773 774 775 776
 */

/*
 * 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.
 */

777
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
A
Al Viro 已提交
778
static struct dentry *cgroup_lookup(struct inode *, struct dentry *, unsigned int);
779
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
780 781
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       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

K
KAMEZAWA Hiroyuki 已提交
790 791 792
static int alloc_css_id(struct cgroup_subsys *ss,
			struct cgroup *parent, struct cgroup *child);

A
Al Viro 已提交
793
static struct inode *cgroup_new_inode(umode_t mode, struct super_block *sb)
794 795 796 797
{
	struct inode *inode = new_inode(sb);

	if (inode) {
798
		inode->i_ino = get_next_ino();
799
		inode->i_mode = mode;
800 801
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
802 803 804 805 806 807
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inode->i_mapping->backing_dev_info = &cgroup_backing_dev_info;
	}
	return inode;
}

808 809 810 811 812 813 814 815 816 817 818
static struct cgroup_name *cgroup_alloc_name(struct dentry *dentry)
{
	struct cgroup_name *name;

	name = kmalloc(sizeof(*name) + dentry->d_name.len + 1, GFP_KERNEL);
	if (!name)
		return NULL;
	strcpy(name->name, dentry->d_name.name);
	return name;
}

819 820
static void cgroup_free_fn(struct work_struct *work)
{
821
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
822 823 824 825 826 827
	struct cgroup_subsys *ss;

	mutex_lock(&cgroup_mutex);
	/*
	 * Release the subsystem state objects.
	 */
828
	for_each_root_subsys(cgrp->root, ss)
829 830 831 832 833
		ss->css_free(cgrp);

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

834 835 836 837 838 839 840
	/*
	 * 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);

841 842
	ida_simple_remove(&cgrp->root->cgroup_ida, cgrp->id);

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.
847 848 849 850 851 852 853 854 855 856 857
	 */
	deactivate_super(cgrp->root->sb);

	/*
	 * if we're getting rid of the cgroup, refcount should ensure
	 * that there are no pidlists left.
	 */
	BUG_ON(!list_empty(&cgrp->pidlists));

	simple_xattrs_free(&cgrp->xattrs);

858
	kfree(rcu_dereference_raw(cgrp->name));
859 860 861 862 863 864 865
	kfree(cgrp);
}

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

866 867
	INIT_WORK(&cgrp->destroy_work, cgroup_free_fn);
	schedule_work(&cgrp->destroy_work);
868 869
}

870 871 872 873
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)) {
874
		struct cgroup *cgrp = dentry->d_fsdata;
875

876
		BUG_ON(!(cgroup_is_dead(cgrp)));
877
		call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
T
Tejun Heo 已提交
878 879 880 881 882 883 884
	} 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
885
		simple_xattrs_free(&cfe->xattrs);
T
Tejun Heo 已提交
886
		kfree(cfe);
887 888 889 890
	}
	iput(inode);
}

891 892 893 894 895
static int cgroup_delete(const struct dentry *d)
{
	return 1;
}

896 897 898 899 900 901 902 903 904
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);
}

905
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
906 907 908 909 910 911
{
	struct cfent *cfe;

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

912 913 914 915
	/*
	 * If we're doing cleanup due to failure of cgroup_create(),
	 * the corresponding @cfe may not exist.
	 */
T
Tejun Heo 已提交
916 917 918 919 920 921 922 923
	list_for_each_entry(cfe, &cgrp->files, node) {
		struct dentry *d = cfe->dentry;

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

		dget(d);
		d_delete(d);
924
		simple_unlink(cgrp->dentry->d_inode, d);
T
Tejun Heo 已提交
925 926 927
		list_del_init(&cfe->node);
		dput(d);

928
		break;
929
	}
T
Tejun Heo 已提交
930 931
}

932 933 934 935 936 937 938 939
/**
 * cgroup_clear_directory - selective removal of base and subsystem files
 * @dir: directory containing the files
 * @base_files: true if the base files should be removed
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
static void cgroup_clear_directory(struct dentry *dir, bool base_files,
				   unsigned long subsys_mask)
T
Tejun Heo 已提交
940 941
{
	struct cgroup *cgrp = __d_cgrp(dir);
942
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
943

944
	for_each_root_subsys(cgrp->root, ss) {
945 946 947 948
		struct cftype_set *set;
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
		list_for_each_entry(set, &ss->cftsets, node)
949
			cgroup_addrm_files(cgrp, NULL, set->cfts, false);
950 951 952 953 954
	}
	if (base_files) {
		while (!list_empty(&cgrp->files))
			cgroup_rm_file(cgrp, NULL);
	}
955 956 957 958 959 960 961
}

/*
 * NOTE : the dentry must have been dget()'ed
 */
static void cgroup_d_remove_dir(struct dentry *dentry)
{
N
Nick Piggin 已提交
962
	struct dentry *parent;
963
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
N
Nick Piggin 已提交
964

965
	cgroup_clear_directory(dentry, true, root->subsys_mask);
966

N
Nick Piggin 已提交
967 968
	parent = dentry->d_parent;
	spin_lock(&parent->d_lock);
969
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
970
	list_del_init(&dentry->d_u.d_child);
N
Nick Piggin 已提交
971 972
	spin_unlock(&dentry->d_lock);
	spin_unlock(&parent->d_lock);
973 974 975
	remove_dir(dentry);
}

B
Ben Blum 已提交
976
/*
B
Ben Blum 已提交
977 978 979
 * Call with cgroup_mutex held. Drops reference counts on modules, including
 * any duplicate ones that parse_cgroupfs_options took. If this function
 * returns an error, no reference counts are touched.
B
Ben Blum 已提交
980
 */
981
static int rebind_subsystems(struct cgroupfs_root *root,
982
			     unsigned long added_mask, unsigned removed_mask)
983
{
984
	struct cgroup *cgrp = &root->top_cgroup;
985 986
	int i;

B
Ben Blum 已提交
987
	BUG_ON(!mutex_is_locked(&cgroup_mutex));
T
Tejun Heo 已提交
988
	BUG_ON(!mutex_is_locked(&cgroup_root_mutex));
B
Ben Blum 已提交
989

990 991
	/* Check that any added subsystems are currently free */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
L
Li Zefan 已提交
992
		unsigned long bit = 1UL << i;
993
		struct cgroup_subsys *ss = cgroup_subsys[i];
994
		if (!(bit & added_mask))
995
			continue;
B
Ben Blum 已提交
996 997 998 999 1000 1001
		/*
		 * Nobody should tell us to do a subsys that doesn't exist:
		 * parse_cgroupfs_options should catch that case and refcounts
		 * ensure that subsystems won't disappear once selected.
		 */
		BUG_ON(ss == NULL);
1002
		if (ss->root != &cgroup_dummy_root) {
1003 1004 1005 1006 1007 1008 1009 1010 1011
			/* Subsystem isn't free */
			return -EBUSY;
		}
	}

	/* Currently we don't handle adding/removing subsystems when
	 * any child cgroups exist. This is theoretically supportable
	 * but involves complex error handling, so it's being left until
	 * later */
1012
	if (root->number_of_cgroups > 1)
1013 1014 1015 1016
		return -EBUSY;

	/* Process each subsystem */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
1017
		struct cgroup_subsys *ss = cgroup_subsys[i];
1018
		unsigned long bit = 1UL << i;
1019
		if (bit & added_mask) {
1020
			/* We're binding this subsystem to this hierarchy */
B
Ben Blum 已提交
1021
			BUG_ON(ss == NULL);
1022
			BUG_ON(cgrp->subsys[i]);
1023 1024
			BUG_ON(!cgroup_dummy_top->subsys[i]);
			BUG_ON(cgroup_dummy_top->subsys[i]->cgroup != cgroup_dummy_top);
1025

1026
			cgrp->subsys[i] = cgroup_dummy_top->subsys[i];
1027
			cgrp->subsys[i]->cgroup = cgrp;
1028
			list_move(&ss->sibling, &root->subsys_list);
1029
			ss->root = root;
1030
			if (ss->bind)
1031
				ss->bind(cgrp);
1032

B
Ben Blum 已提交
1033
			/* refcount was already taken, and we're keeping it */
1034
			root->subsys_mask |= bit;
1035
		} else if (bit & removed_mask) {
1036
			/* We're removing this subsystem */
B
Ben Blum 已提交
1037
			BUG_ON(ss == NULL);
1038
			BUG_ON(cgrp->subsys[i] != cgroup_dummy_top->subsys[i]);
1039
			BUG_ON(cgrp->subsys[i]->cgroup != cgrp);
1040

1041
			if (ss->bind)
1042 1043
				ss->bind(cgroup_dummy_top);
			cgroup_dummy_top->subsys[i]->cgroup = cgroup_dummy_top;
1044
			cgrp->subsys[i] = NULL;
1045 1046
			cgroup_subsys[i]->root = &cgroup_dummy_root;
			list_move(&ss->sibling, &cgroup_dummy_root.subsys_list);
1047

B
Ben Blum 已提交
1048 1049
			/* subsystem is now free - drop reference on module */
			module_put(ss->module);
1050 1051
			root->subsys_mask &= ~bit;
		} else if (bit & root->subsys_mask) {
1052
			/* Subsystem state should already exist */
B
Ben Blum 已提交
1053
			BUG_ON(ss == NULL);
1054
			BUG_ON(!cgrp->subsys[i]);
B
Ben Blum 已提交
1055 1056 1057 1058 1059 1060 1061 1062
			/*
			 * a refcount was taken, but we already had one, so
			 * drop the extra reference.
			 */
			module_put(ss->module);
#ifdef CONFIG_MODULE_UNLOAD
			BUG_ON(ss->module && !module_refcount(ss->module));
#endif
1063 1064
		} else {
			/* Subsystem state shouldn't exist */
1065
			BUG_ON(cgrp->subsys[i]);
1066 1067 1068 1069 1070 1071
		}
	}

	return 0;
}

1072
static int cgroup_show_options(struct seq_file *seq, struct dentry *dentry)
1073
{
1074
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
1075 1076
	struct cgroup_subsys *ss;

T
Tejun Heo 已提交
1077
	mutex_lock(&cgroup_root_mutex);
1078
	for_each_root_subsys(root, ss)
1079
		seq_printf(seq, ",%s", ss->name);
1080 1081
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR)
		seq_puts(seq, ",sane_behavior");
1082
	if (root->flags & CGRP_ROOT_NOPREFIX)
1083
		seq_puts(seq, ",noprefix");
1084
	if (root->flags & CGRP_ROOT_XATTR)
A
Aristeu Rozanski 已提交
1085
		seq_puts(seq, ",xattr");
1086 1087
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1088
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags))
1089
		seq_puts(seq, ",clone_children");
1090 1091
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
T
Tejun Heo 已提交
1092
	mutex_unlock(&cgroup_root_mutex);
1093 1094 1095 1096
	return 0;
}

struct cgroup_sb_opts {
1097
	unsigned long subsys_mask;
1098
	unsigned long flags;
1099
	char *release_agent;
1100
	bool cpuset_clone_children;
1101
	char *name;
1102 1103
	/* User explicitly requested empty subsystem */
	bool none;
1104 1105

	struct cgroupfs_root *new_root;
1106

1107 1108
};

B
Ben Blum 已提交
1109
/*
1110 1111 1112 1113
 * 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
1114
 */
B
Ben Blum 已提交
1115
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1116
{
1117 1118
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1119
	unsigned long mask = (unsigned long)-1;
B
Ben Blum 已提交
1120 1121
	int i;
	bool module_pin_failed = false;
1122

B
Ben Blum 已提交
1123 1124
	BUG_ON(!mutex_is_locked(&cgroup_mutex));

1125 1126 1127
#ifdef CONFIG_CPUSETS
	mask = ~(1UL << cpuset_subsys_id);
#endif
1128

1129
	memset(opts, 0, sizeof(*opts));
1130 1131 1132 1133

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1134
		if (!strcmp(token, "none")) {
1135 1136
			/* Explicitly have no subsystems */
			opts->none = true;
1137 1138 1139 1140 1141 1142 1143 1144 1145
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
1146 1147 1148 1149
		if (!strcmp(token, "__DEVEL__sane_behavior")) {
			opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
			continue;
		}
1150
		if (!strcmp(token, "noprefix")) {
1151
			opts->flags |= CGRP_ROOT_NOPREFIX;
1152 1153 1154
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1155
			opts->cpuset_clone_children = true;
1156 1157
			continue;
		}
A
Aristeu Rozanski 已提交
1158
		if (!strcmp(token, "xattr")) {
1159
			opts->flags |= CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
1160 1161
			continue;
		}
1162
		if (!strncmp(token, "release_agent=", 14)) {
1163 1164 1165
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1166
			opts->release_agent =
1167
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1168 1169
			if (!opts->release_agent)
				return -ENOMEM;
1170 1171 1172
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
			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,
1190
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1191 1192 1193
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1194 1195 1196 1197 1198

			continue;
		}

		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
1199
			struct cgroup_subsys *ss = cgroup_subsys[i];
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
			if (ss == NULL)
				continue;
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1210
			set_bit(i, &opts->subsys_mask);
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
			one_ss = true;

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

	/*
	 * If the 'all' option was specified select all the subsystems,
1221 1222
	 * otherwise if 'none', 'name=' and a subsystem name options
	 * were not specified, let's default to 'all'
1223
	 */
1224
	if (all_ss || (!one_ss && !opts->none && !opts->name)) {
1225
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
1226
			struct cgroup_subsys *ss = cgroup_subsys[i];
1227 1228 1229 1230
			if (ss == NULL)
				continue;
			if (ss->disabled)
				continue;
1231
			set_bit(i, &opts->subsys_mask);
1232 1233 1234
		}
	}

1235 1236
	/* Consistency checks */

1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
	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;
		}
	}

1251 1252 1253 1254 1255
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
1256
	if ((opts->flags & CGRP_ROOT_NOPREFIX) && (opts->subsys_mask & mask))
1257 1258
		return -EINVAL;

1259 1260

	/* Can't specify "none" and some subsystems */
1261
	if (opts->subsys_mask && opts->none)
1262 1263 1264 1265 1266 1267
		return -EINVAL;

	/*
	 * We either have to specify by name or by subsystems. (So all
	 * empty hierarchies must have a name).
	 */
1268
	if (!opts->subsys_mask && !opts->name)
1269 1270
		return -EINVAL;

B
Ben Blum 已提交
1271 1272 1273 1274 1275 1276
	/*
	 * Grab references on all the modules we'll need, so the subsystems
	 * don't dance around before rebind_subsystems attaches them. This may
	 * take duplicate reference counts on a subsystem that's already used,
	 * but rebind_subsystems handles this case.
	 */
1277
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
B
Ben Blum 已提交
1278 1279
		unsigned long bit = 1UL << i;

1280
		if (!(bit & opts->subsys_mask))
B
Ben Blum 已提交
1281
			continue;
1282
		if (!try_module_get(cgroup_subsys[i]->module)) {
B
Ben Blum 已提交
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
			module_pin_failed = true;
			break;
		}
	}
	if (module_pin_failed) {
		/*
		 * oops, one of the modules was going away. this means that we
		 * raced with a module_delete call, and to the user this is
		 * essentially a "subsystem doesn't exist" case.
		 */
1293
		for (i--; i >= 0; i--) {
B
Ben Blum 已提交
1294 1295 1296
			/* drop refcounts only on the ones we took */
			unsigned long bit = 1UL << i;

1297
			if (!(bit & opts->subsys_mask))
B
Ben Blum 已提交
1298
				continue;
1299
			module_put(cgroup_subsys[i]->module);
B
Ben Blum 已提交
1300 1301 1302 1303
		}
		return -ENOENT;
	}

1304 1305 1306
	return 0;
}

1307
static void drop_parsed_module_refcounts(unsigned long subsys_mask)
B
Ben Blum 已提交
1308 1309
{
	int i;
1310
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
B
Ben Blum 已提交
1311 1312
		unsigned long bit = 1UL << i;

1313
		if (!(bit & subsys_mask))
B
Ben Blum 已提交
1314
			continue;
1315
		module_put(cgroup_subsys[i]->module);
B
Ben Blum 已提交
1316 1317 1318
	}
}

1319 1320 1321 1322
static int cgroup_remount(struct super_block *sb, int *flags, char *data)
{
	int ret = 0;
	struct cgroupfs_root *root = sb->s_fs_info;
1323
	struct cgroup *cgrp = &root->top_cgroup;
1324
	struct cgroup_sb_opts opts;
1325
	unsigned long added_mask, removed_mask;
1326

1327 1328 1329 1330 1331
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR) {
		pr_err("cgroup: sane_behavior: remount is not allowed\n");
		return -EINVAL;
	}

1332
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1333
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1334
	mutex_lock(&cgroup_root_mutex);
1335 1336 1337 1338 1339 1340

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

1341
	if (opts.subsys_mask != root->subsys_mask || opts.release_agent)
1342 1343 1344
		pr_warning("cgroup: option changes via remount are deprecated (pid=%d comm=%s)\n",
			   task_tgid_nr(current), current->comm);

1345 1346
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1347

B
Ben Blum 已提交
1348 1349 1350
	/* Don't allow flags or name to change at remount */
	if (opts.flags != root->flags ||
	    (opts.name && strcmp(opts.name, root->name))) {
1351
		ret = -EINVAL;
1352
		drop_parsed_module_refcounts(opts.subsys_mask);
1353 1354 1355
		goto out_unlock;
	}

1356 1357 1358 1359 1360 1361 1362
	/*
	 * Clear out the files of subsystems that should be removed, do
	 * this before rebind_subsystems, since rebind_subsystems may
	 * change this hierarchy's subsys_list.
	 */
	cgroup_clear_directory(cgrp->dentry, false, removed_mask);

1363
	ret = rebind_subsystems(root, added_mask, removed_mask);
B
Ben Blum 已提交
1364
	if (ret) {
1365 1366
		/* rebind_subsystems failed, re-populate the removed files */
		cgroup_populate_dir(cgrp, false, removed_mask);
1367
		drop_parsed_module_refcounts(opts.subsys_mask);
1368
		goto out_unlock;
B
Ben Blum 已提交
1369
	}
1370

1371
	/* re-populate subsystem files */
1372
	cgroup_populate_dir(cgrp, false, added_mask);
1373

1374 1375
	if (opts.release_agent)
		strcpy(root->release_agent_path, opts.release_agent);
1376
 out_unlock:
1377
	kfree(opts.release_agent);
1378
	kfree(opts.name);
T
Tejun Heo 已提交
1379
	mutex_unlock(&cgroup_root_mutex);
1380
	mutex_unlock(&cgroup_mutex);
1381
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1382 1383 1384
	return ret;
}

1385
static const struct super_operations cgroup_ops = {
1386 1387 1388 1389 1390 1391
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
	.show_options = cgroup_show_options,
	.remount_fs = cgroup_remount,
};

1392 1393 1394 1395
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
T
Tejun Heo 已提交
1396
	INIT_LIST_HEAD(&cgrp->files);
1397
	INIT_LIST_HEAD(&cgrp->cset_links);
1398
	INIT_LIST_HEAD(&cgrp->release_list);
1399 1400
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
1401 1402
	INIT_LIST_HEAD(&cgrp->event_list);
	spin_lock_init(&cgrp->event_list_lock);
A
Aristeu Rozanski 已提交
1403
	simple_xattrs_init(&cgrp->xattrs);
1404
}
1405

1406 1407
static void init_cgroup_root(struct cgroupfs_root *root)
{
1408
	struct cgroup *cgrp = &root->top_cgroup;
1409

1410 1411 1412
	INIT_LIST_HEAD(&root->subsys_list);
	INIT_LIST_HEAD(&root->root_list);
	root->number_of_cgroups = 1;
1413
	cgrp->root = root;
1414
	cgrp->name = &root_cgroup_name;
1415
	init_cgroup_housekeeping(cgrp);
1416 1417
}

1418
static int cgroup_init_root_id(struct cgroupfs_root *root)
1419
{
1420
	int id;
1421

T
Tejun Heo 已提交
1422 1423 1424
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1425 1426 1427 1428 1429
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, 2, 0, GFP_KERNEL);
	if (id < 0)
		return id;

	root->hierarchy_id = id;
1430 1431 1432 1433 1434
	return 0;
}

static void cgroup_exit_root_id(struct cgroupfs_root *root)
{
T
Tejun Heo 已提交
1435 1436 1437
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1438
	if (root->hierarchy_id) {
1439
		idr_remove(&cgroup_hierarchy_idr, root->hierarchy_id);
1440 1441
		root->hierarchy_id = 0;
	}
1442 1443
}

1444 1445
static int cgroup_test_super(struct super_block *sb, void *data)
{
1446
	struct cgroup_sb_opts *opts = data;
1447 1448
	struct cgroupfs_root *root = sb->s_fs_info;

1449 1450 1451
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1452

1453 1454 1455 1456
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
1457 1458
	if ((opts->subsys_mask || opts->none)
	    && (opts->subsys_mask != root->subsys_mask))
1459 1460 1461 1462 1463
		return 0;

	return 1;
}

1464 1465 1466 1467
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1468
	if (!opts->subsys_mask && !opts->none)
1469 1470 1471 1472 1473 1474 1475
		return NULL;

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

	init_cgroup_root(root);
1476

1477
	root->subsys_mask = opts->subsys_mask;
1478
	root->flags = opts->flags;
T
Tejun Heo 已提交
1479
	ida_init(&root->cgroup_ida);
1480 1481 1482 1483
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1484 1485
	if (opts->cpuset_clone_children)
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags);
1486 1487 1488
	return root;
}

1489
static void cgroup_free_root(struct cgroupfs_root *root)
1490
{
1491 1492 1493
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);
1494

1495 1496 1497
		ida_destroy(&root->cgroup_ida);
		kfree(root);
	}
1498 1499
}

1500 1501 1502
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1503 1504 1505 1506 1507 1508
	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;

1509
	BUG_ON(!opts->subsys_mask && !opts->none);
1510 1511 1512 1513 1514

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

1515 1516
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527

	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
1528 1529
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1530
		.d_delete = cgroup_delete,
A
Al Viro 已提交
1531 1532
	};

1533 1534 1535 1536 1537 1538 1539 1540 1541 1542
	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);
1543 1544
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1545
		return -ENOMEM;
A
Al Viro 已提交
1546 1547
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1548 1549 1550
	return 0;
}

A
Al Viro 已提交
1551
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1552
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1553
			 void *data)
1554 1555
{
	struct cgroup_sb_opts opts;
1556
	struct cgroupfs_root *root;
1557 1558
	int ret = 0;
	struct super_block *sb;
1559
	struct cgroupfs_root *new_root;
T
Tejun Heo 已提交
1560
	struct inode *inode;
1561 1562

	/* First find the desired set of subsystems */
B
Ben Blum 已提交
1563
	mutex_lock(&cgroup_mutex);
1564
	ret = parse_cgroupfs_options(data, &opts);
B
Ben Blum 已提交
1565
	mutex_unlock(&cgroup_mutex);
1566 1567
	if (ret)
		goto out_err;
1568

1569 1570 1571 1572 1573 1574 1575
	/*
	 * 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);
B
Ben Blum 已提交
1576
		goto drop_modules;
1577
	}
1578
	opts.new_root = new_root;
1579

1580
	/* Locate an existing or new sb for this hierarchy */
D
David Howells 已提交
1581
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1582
	if (IS_ERR(sb)) {
1583
		ret = PTR_ERR(sb);
1584
		cgroup_free_root(opts.new_root);
B
Ben Blum 已提交
1585
		goto drop_modules;
1586 1587
	}

1588 1589 1590 1591
	root = sb->s_fs_info;
	BUG_ON(!root);
	if (root == opts.new_root) {
		/* We used the new root structure, so this is a new hierarchy */
1592
		struct list_head tmp_links;
1593
		struct cgroup *root_cgrp = &root->top_cgroup;
1594
		struct cgroupfs_root *existing_root;
1595
		const struct cred *cred;
1596
		int i;
1597
		struct css_set *cset;
1598 1599 1600 1601 1602 1603

		BUG_ON(sb->s_root != NULL);

		ret = cgroup_get_rootdir(sb);
		if (ret)
			goto drop_new_super;
1604
		inode = sb->s_root->d_inode;
1605

1606
		mutex_lock(&inode->i_mutex);
1607
		mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1608
		mutex_lock(&cgroup_root_mutex);
1609

T
Tejun Heo 已提交
1610 1611 1612 1613 1614 1615
		/* 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 unlock_drop;
1616

1617 1618 1619 1620 1621 1622 1623
		/*
		 * 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
		 */
1624
		ret = allocate_cgrp_cset_links(css_set_count, &tmp_links);
T
Tejun Heo 已提交
1625 1626
		if (ret)
			goto unlock_drop;
1627

1628 1629 1630 1631
		ret = cgroup_init_root_id(root);
		if (ret)
			goto unlock_drop;

1632
		ret = rebind_subsystems(root, root->subsys_mask, 0);
1633
		if (ret == -EBUSY) {
1634
			free_cgrp_cset_links(&tmp_links);
T
Tejun Heo 已提交
1635
			goto unlock_drop;
1636
		}
B
Ben Blum 已提交
1637 1638 1639 1640 1641
		/*
		 * There must be no failure case after here, since rebinding
		 * takes care of subsystems' refcounts, which are explicitly
		 * dropped in the failure exit path.
		 */
1642 1643 1644 1645

		/* EBUSY should be the only error here */
		BUG_ON(ret);

1646 1647
		list_add(&root->root_list, &cgroup_roots);
		cgroup_root_count++;
1648

1649
		sb->s_root->d_fsdata = root_cgrp;
1650 1651
		root->top_cgroup.dentry = sb->s_root;

1652 1653 1654
		/* Link the top cgroup in this hierarchy into all
		 * the css_set objects */
		write_lock(&css_set_lock);
1655
		hash_for_each(css_set_table, i, cset, hlist)
1656
			link_css_set(&tmp_links, cset, root_cgrp);
1657 1658
		write_unlock(&css_set_lock);

1659
		free_cgrp_cset_links(&tmp_links);
1660

1661
		BUG_ON(!list_empty(&root_cgrp->children));
1662 1663
		BUG_ON(root->number_of_cgroups != 1);

1664
		cred = override_creds(&init_cred);
1665
		cgroup_populate_dir(root_cgrp, true, root->subsys_mask);
1666
		revert_creds(cred);
T
Tejun Heo 已提交
1667
		mutex_unlock(&cgroup_root_mutex);
1668
		mutex_unlock(&cgroup_mutex);
1669
		mutex_unlock(&inode->i_mutex);
1670 1671 1672 1673 1674
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1675
		cgroup_free_root(opts.new_root);
1676

1677 1678 1679 1680 1681 1682 1683 1684
		if (root->flags != opts.flags) {
			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;
				goto drop_new_super;
			} else {
				pr_warning("cgroup: new mount options do not match the existing superblock, will be ignored\n");
			}
1685 1686
		}

B
Ben Blum 已提交
1687
		/* no subsys rebinding, so refcounts don't change */
1688
		drop_parsed_module_refcounts(opts.subsys_mask);
1689 1690
	}

1691 1692
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1693
	return dget(sb->s_root);
1694

T
Tejun Heo 已提交
1695
 unlock_drop:
1696
	cgroup_exit_root_id(root);
T
Tejun Heo 已提交
1697 1698 1699
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);
1700
 drop_new_super:
1701
	deactivate_locked_super(sb);
B
Ben Blum 已提交
1702
 drop_modules:
1703
	drop_parsed_module_refcounts(opts.subsys_mask);
1704 1705 1706
 out_err:
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1707
	return ERR_PTR(ret);
1708 1709 1710 1711
}

static void cgroup_kill_sb(struct super_block *sb) {
	struct cgroupfs_root *root = sb->s_fs_info;
1712
	struct cgroup *cgrp = &root->top_cgroup;
1713
	struct cgrp_cset_link *link, *tmp_link;
1714 1715 1716 1717 1718
	int ret;

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1719
	BUG_ON(!list_empty(&cgrp->children));
1720 1721

	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1722
	mutex_lock(&cgroup_root_mutex);
1723 1724

	/* Rebind all subsystems back to the default hierarchy */
1725
	ret = rebind_subsystems(root, 0, root->subsys_mask);
1726 1727 1728
	/* Shouldn't be able to fail ... */
	BUG_ON(ret);

1729
	/*
1730
	 * Release all the links from cset_links to this hierarchy's
1731 1732 1733
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1734

1735 1736 1737
	list_for_each_entry_safe(link, tmp_link, &cgrp->cset_links, cset_link) {
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
1738 1739 1740 1741
		kfree(link);
	}
	write_unlock(&css_set_lock);

1742 1743
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
1744
		cgroup_root_count--;
1745
	}
1746

1747 1748
	cgroup_exit_root_id(root);

T
Tejun Heo 已提交
1749
	mutex_unlock(&cgroup_root_mutex);
1750 1751
	mutex_unlock(&cgroup_mutex);

A
Aristeu Rozanski 已提交
1752 1753
	simple_xattrs_free(&cgrp->xattrs);

1754
	kill_litter_super(sb);
1755
	cgroup_free_root(root);
1756 1757 1758 1759
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1760
	.mount = cgroup_mount,
1761 1762 1763
	.kill_sb = cgroup_kill_sb,
};

1764 1765
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1766 1767 1768 1769 1770 1771
/**
 * 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
 *
1772 1773 1774 1775 1776 1777
 * 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.
1778
 */
1779
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1780
{
1781
	int ret = -ENAMETOOLONG;
1782
	char *start;
1783

1784 1785 1786
	if (!cgrp->parent) {
		if (strlcpy(buf, "/", buflen) >= buflen)
			return -ENAMETOOLONG;
1787 1788 1789
		return 0;
	}

1790 1791
	start = buf + buflen - 1;
	*start = '\0';
1792

1793
	rcu_read_lock();
1794
	do {
1795 1796 1797 1798
		const char *name = cgroup_name(cgrp);
		int len;

		len = strlen(name);
1799
		if ((start -= len) < buf)
1800 1801
			goto out;
		memcpy(start, name, len);
1802

1803
		if (--start < buf)
1804
			goto out;
1805
		*start = '/';
1806 1807

		cgrp = cgrp->parent;
1808
	} while (cgrp->parent);
1809
	ret = 0;
1810
	memmove(buf, start, buf + buflen - start);
1811 1812 1813
out:
	rcu_read_unlock();
	return ret;
1814
}
B
Ben Blum 已提交
1815
EXPORT_SYMBOL_GPL(cgroup_path);
1816

1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
/**
 * task_cgroup_path_from_hierarchy - cgroup path of a task on a hierarchy
 * @task: target task
 * @hierarchy_id: the hierarchy to look up @task's cgroup from
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
 * Determine @task's cgroup on the hierarchy specified by @hierarchy_id and
 * copy its path into @buf.  This function grabs cgroup_mutex and shouldn't
 * be used inside locks used by cgroup controller callbacks.
 */
int task_cgroup_path_from_hierarchy(struct task_struct *task, int hierarchy_id,
				    char *buf, size_t buflen)
{
	struct cgroupfs_root *root;
	struct cgroup *cgrp = NULL;
	int ret = -ENOENT;

	mutex_lock(&cgroup_mutex);

	root = idr_find(&cgroup_hierarchy_idr, hierarchy_id);
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
		ret = cgroup_path(cgrp, buf, buflen);
	}

	mutex_unlock(&cgroup_mutex);

	return ret;
}
EXPORT_SYMBOL_GPL(task_cgroup_path_from_hierarchy);

1849 1850 1851
/*
 * Control Group taskset
 */
1852 1853 1854
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
1855
	struct css_set		*cg;
1856 1857
};

1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928
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);

/**
 * cgroup_taskset_cur_cgroup - return the matching cgroup for the current task
 * @tset: taskset of interest
 *
 * Return the cgroup for the current (last returned) task of @tset.  This
 * function must be preceded by either cgroup_taskset_first() or
 * cgroup_taskset_next().
 */
struct cgroup *cgroup_taskset_cur_cgroup(struct cgroup_taskset *tset)
{
	return tset->cur_cgrp;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_cur_cgroup);

/**
 * 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
1929 1930 1931
/*
 * cgroup_task_migrate - move a task from one cgroup to another.
 *
1932
 * Must be called with cgroup_mutex and threadgroup locked.
B
Ben Blum 已提交
1933
 */
1934 1935 1936
static void cgroup_task_migrate(struct cgroup *old_cgrp,
				struct task_struct *tsk,
				struct css_set *new_cset)
B
Ben Blum 已提交
1937
{
1938
	struct css_set *old_cset;
B
Ben Blum 已提交
1939 1940

	/*
1941 1942 1943
	 * 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
1944
	 */
1945
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
1946
	old_cset = tsk->cgroups;
B
Ben Blum 已提交
1947 1948

	task_lock(tsk);
1949
	rcu_assign_pointer(tsk->cgroups, new_cset);
B
Ben Blum 已提交
1950 1951 1952 1953 1954
	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))
1955
		list_move(&tsk->cg_list, &new_cset->tasks);
B
Ben Blum 已提交
1956 1957 1958
	write_unlock(&css_set_lock);

	/*
1959 1960 1961
	 * 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
1962
	 */
1963 1964
	set_bit(CGRP_RELEASABLE, &old_cgrp->flags);
	put_css_set(old_cset);
B
Ben Blum 已提交
1965 1966
}

L
Li Zefan 已提交
1967
/**
1968
 * cgroup_attach_task - attach a task or a whole threadgroup to a cgroup
B
Ben Blum 已提交
1969
 * @cgrp: the cgroup to attach to
1970 1971
 * @tsk: the task or the leader of the threadgroup to be attached
 * @threadgroup: attach the whole threadgroup?
B
Ben Blum 已提交
1972
 *
1973
 * Call holding cgroup_mutex and the group_rwsem of the leader. Will take
1974
 * task_lock of @tsk or each thread in the threadgroup individually in turn.
B
Ben Blum 已提交
1975
 */
T
Tejun Heo 已提交
1976 1977
static int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk,
			      bool threadgroup)
B
Ben Blum 已提交
1978 1979 1980 1981 1982
{
	int retval, i, group_size;
	struct cgroup_subsys *ss, *failed_ss = NULL;
	struct cgroupfs_root *root = cgrp->root;
	/* threadgroup list cursor and array */
1983
	struct task_struct *leader = tsk;
1984
	struct task_and_cgroup *tc;
1985
	struct flex_array *group;
1986
	struct cgroup_taskset tset = { };
B
Ben Blum 已提交
1987 1988 1989 1990 1991

	/*
	 * 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
1992 1993
	 * 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
1994
	 */
1995 1996 1997 1998
	if (threadgroup)
		group_size = get_nr_threads(tsk);
	else
		group_size = 1;
1999
	/* flex_array supports very large thread-groups better than kmalloc. */
2000
	group = flex_array_alloc(sizeof(*tc), group_size, GFP_KERNEL);
B
Ben Blum 已提交
2001 2002
	if (!group)
		return -ENOMEM;
2003
	/* pre-allocate to guarantee space while iterating in rcu read-side. */
2004
	retval = flex_array_prealloc(group, 0, group_size, GFP_KERNEL);
2005 2006
	if (retval)
		goto out_free_group_list;
B
Ben Blum 已提交
2007 2008

	i = 0;
2009 2010 2011 2012 2013 2014
	/*
	 * 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
2015
	do {
2016 2017
		struct task_and_cgroup ent;

2018 2019 2020 2021
		/* @tsk either already exited or can't exit until the end */
		if (tsk->flags & PF_EXITING)
			continue;

B
Ben Blum 已提交
2022 2023
		/* as per above, nr_threads may decrease, but not increase. */
		BUG_ON(i >= group_size);
2024 2025
		ent.task = tsk;
		ent.cgrp = task_cgroup_from_root(tsk, root);
2026 2027 2028
		/* nothing to do if this task is already in the cgroup */
		if (ent.cgrp == cgrp)
			continue;
2029 2030 2031 2032
		/*
		 * saying GFP_ATOMIC has no effect here because we did prealloc
		 * earlier, but it's good form to communicate our expectations.
		 */
2033
		retval = flex_array_put(group, i, &ent, GFP_ATOMIC);
2034
		BUG_ON(retval != 0);
B
Ben Blum 已提交
2035
		i++;
2036 2037 2038

		if (!threadgroup)
			break;
B
Ben Blum 已提交
2039
	} while_each_thread(leader, tsk);
2040
	rcu_read_unlock();
B
Ben Blum 已提交
2041 2042
	/* remember the number of threads in the array for later. */
	group_size = i;
2043 2044
	tset.tc_array = group;
	tset.tc_array_len = group_size;
B
Ben Blum 已提交
2045

2046 2047
	/* methods shouldn't be called if no task is actually migrating */
	retval = 0;
2048
	if (!group_size)
2049
		goto out_free_group_list;
2050

B
Ben Blum 已提交
2051 2052 2053
	/*
	 * step 1: check that we can legitimately attach to the cgroup.
	 */
2054
	for_each_root_subsys(root, ss) {
B
Ben Blum 已提交
2055
		if (ss->can_attach) {
2056
			retval = ss->can_attach(cgrp, &tset);
B
Ben Blum 已提交
2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068
			if (retval) {
				failed_ss = ss;
				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++) {
2069
		tc = flex_array_get(group, i);
2070 2071 2072 2073
		tc->cg = find_css_set(tc->task->cgroups, cgrp);
		if (!tc->cg) {
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2074 2075 2076 2077
		}
	}

	/*
2078 2079 2080
	 * 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
2081 2082
	 */
	for (i = 0; i < group_size; i++) {
2083
		tc = flex_array_get(group, i);
2084
		cgroup_task_migrate(tc->cgrp, tc->task, tc->cg);
B
Ben Blum 已提交
2085 2086 2087 2088
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2089
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2090
	 */
2091
	for_each_root_subsys(root, ss) {
B
Ben Blum 已提交
2092
		if (ss->attach)
2093
			ss->attach(cgrp, &tset);
B
Ben Blum 已提交
2094 2095 2096 2097 2098 2099
	}

	/*
	 * step 5: success! and cleanup
	 */
	retval = 0;
2100 2101 2102 2103 2104 2105 2106 2107
out_put_css_set_refs:
	if (retval) {
		for (i = 0; i < group_size; i++) {
			tc = flex_array_get(group, i);
			if (!tc->cg)
				break;
			put_css_set(tc->cg);
		}
B
Ben Blum 已提交
2108 2109 2110
	}
out_cancel_attach:
	if (retval) {
2111
		for_each_root_subsys(root, ss) {
2112
			if (ss == failed_ss)
B
Ben Blum 已提交
2113 2114
				break;
			if (ss->cancel_attach)
2115
				ss->cancel_attach(cgrp, &tset);
B
Ben Blum 已提交
2116 2117 2118
		}
	}
out_free_group_list:
2119
	flex_array_free(group);
B
Ben Blum 已提交
2120 2121 2122 2123 2124
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2125 2126
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2127
 */
B
Ben Blum 已提交
2128
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2129 2130
{
	struct task_struct *tsk;
2131
	const struct cred *cred = current_cred(), *tcred;
2132 2133
	int ret;

B
Ben Blum 已提交
2134 2135 2136
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2137 2138
retry_find_task:
	rcu_read_lock();
2139
	if (pid) {
2140
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2141 2142
		if (!tsk) {
			rcu_read_unlock();
2143 2144
			ret= -ESRCH;
			goto out_unlock_cgroup;
2145
		}
B
Ben Blum 已提交
2146 2147 2148 2149
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2150
		tcred = __task_cred(tsk);
2151 2152 2153
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2154
			rcu_read_unlock();
2155 2156
			ret = -EACCES;
			goto out_unlock_cgroup;
2157
		}
2158 2159
	} else
		tsk = current;
2160 2161

	if (threadgroup)
2162
		tsk = tsk->group_leader;
2163 2164

	/*
2165
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2166 2167 2168
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2169
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2170 2171 2172 2173 2174
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
	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;
		}
2192 2193 2194 2195
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2196 2197
	threadgroup_unlock(tsk);

2198
	put_task_struct(tsk);
2199
out_unlock_cgroup:
T
Tejun Heo 已提交
2200
	mutex_unlock(&cgroup_mutex);
2201 2202 2203
	return ret;
}

2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
/**
 * 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
2214
	mutex_lock(&cgroup_mutex);
2215 2216 2217 2218 2219 2220 2221
	for_each_active_root(root) {
		struct cgroup *from_cg = task_cgroup_from_root(from, root);

		retval = cgroup_attach_task(from_cg, tsk, false);
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2222
	mutex_unlock(&cgroup_mutex);
2223 2224 2225 2226 2227

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2228
static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2229 2230 2231 2232 2233
{
	return attach_task_by_pid(cgrp, pid, false);
}

static int cgroup_procs_write(struct cgroup *cgrp, struct cftype *cft, u64 tgid)
2234
{
2235
	return attach_task_by_pid(cgrp, tgid, true);
2236 2237
}

2238 2239 2240 2241
static int cgroup_release_agent_write(struct cgroup *cgrp, struct cftype *cft,
				      const char *buffer)
{
	BUILD_BUG_ON(sizeof(cgrp->root->release_agent_path) < PATH_MAX);
2242 2243
	if (strlen(buffer) >= PATH_MAX)
		return -EINVAL;
2244 2245
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
T
Tejun Heo 已提交
2246
	mutex_lock(&cgroup_root_mutex);
2247
	strcpy(cgrp->root->release_agent_path, buffer);
T
Tejun Heo 已提交
2248
	mutex_unlock(&cgroup_root_mutex);
T
Tejun Heo 已提交
2249
	mutex_unlock(&cgroup_mutex);
2250 2251 2252 2253 2254 2255 2256 2257 2258 2259
	return 0;
}

static int cgroup_release_agent_show(struct cgroup *cgrp, struct cftype *cft,
				     struct seq_file *seq)
{
	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
2260
	mutex_unlock(&cgroup_mutex);
2261 2262 2263
	return 0;
}

2264 2265 2266 2267
static int cgroup_sane_behavior_show(struct cgroup *cgrp, struct cftype *cft,
				     struct seq_file *seq)
{
	seq_printf(seq, "%d\n", cgroup_sane_behavior(cgrp));
2268 2269 2270
	return 0;
}

2271 2272 2273
/* A buffer size big enough for numbers or short strings */
#define CGROUP_LOCAL_BUFFER_SIZE 64

2274
static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
2275 2276 2277
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
2278
{
2279
	char buffer[CGROUP_LOCAL_BUFFER_SIZE];
2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290
	int retval = 0;
	char *end;

	if (!nbytes)
		return -EINVAL;
	if (nbytes >= sizeof(buffer))
		return -E2BIG;
	if (copy_from_user(buffer, userbuf, nbytes))
		return -EFAULT;

	buffer[nbytes] = 0;     /* nul-terminate */
2291
	if (cft->write_u64) {
K
KOSAKI Motohiro 已提交
2292
		u64 val = simple_strtoull(strstrip(buffer), &end, 0);
2293 2294 2295 2296
		if (*end)
			return -EINVAL;
		retval = cft->write_u64(cgrp, cft, val);
	} else {
K
KOSAKI Motohiro 已提交
2297
		s64 val = simple_strtoll(strstrip(buffer), &end, 0);
2298 2299 2300 2301
		if (*end)
			return -EINVAL;
		retval = cft->write_s64(cgrp, cft, val);
	}
2302 2303 2304 2305 2306
	if (!retval)
		retval = nbytes;
	return retval;
}

2307 2308 2309 2310 2311
static ssize_t cgroup_write_string(struct cgroup *cgrp, struct cftype *cft,
				   struct file *file,
				   const char __user *userbuf,
				   size_t nbytes, loff_t *unused_ppos)
{
2312
	char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326
	int retval = 0;
	size_t max_bytes = cft->max_write_len;
	char *buffer = local_buffer;

	if (!max_bytes)
		max_bytes = sizeof(local_buffer) - 1;
	if (nbytes >= max_bytes)
		return -E2BIG;
	/* Allocate a dynamic buffer if we need one */
	if (nbytes >= sizeof(local_buffer)) {
		buffer = kmalloc(nbytes + 1, GFP_KERNEL);
		if (buffer == NULL)
			return -ENOMEM;
	}
L
Li Zefan 已提交
2327 2328 2329 2330
	if (nbytes && copy_from_user(buffer, userbuf, nbytes)) {
		retval = -EFAULT;
		goto out;
	}
2331 2332

	buffer[nbytes] = 0;     /* nul-terminate */
K
KOSAKI Motohiro 已提交
2333
	retval = cft->write_string(cgrp, cft, strstrip(buffer));
2334 2335
	if (!retval)
		retval = nbytes;
L
Li Zefan 已提交
2336
out:
2337 2338 2339 2340 2341
	if (buffer != local_buffer)
		kfree(buffer);
	return retval;
}

2342 2343 2344 2345
static ssize_t cgroup_file_write(struct file *file, const char __user *buf,
						size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2346
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2347

2348
	if (cgroup_is_dead(cgrp))
2349
		return -ENODEV;
2350
	if (cft->write)
2351
		return cft->write(cgrp, cft, file, buf, nbytes, ppos);
2352 2353
	if (cft->write_u64 || cft->write_s64)
		return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
2354 2355
	if (cft->write_string)
		return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
2356 2357 2358 2359
	if (cft->trigger) {
		int ret = cft->trigger(cgrp, (unsigned int)cft->private);
		return ret ? ret : nbytes;
	}
2360
	return -EINVAL;
2361 2362
}

2363 2364 2365 2366
static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
2367
{
2368
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2369
	u64 val = cft->read_u64(cgrp, cft);
2370 2371 2372 2373 2374
	int len = sprintf(tmp, "%llu\n", (unsigned long long) val);

	return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
}

2375 2376 2377 2378 2379
static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
{
2380
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2381 2382 2383 2384 2385 2386
	s64 val = cft->read_s64(cgrp, cft);
	int len = sprintf(tmp, "%lld\n", (long long) val);

	return simple_read_from_buffer(buf, nbytes, ppos, tmp, len);
}

2387 2388 2389 2390
static ssize_t cgroup_file_read(struct file *file, char __user *buf,
				   size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2391
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2392

2393
	if (cgroup_is_dead(cgrp))
2394 2395 2396
		return -ENODEV;

	if (cft->read)
2397
		return cft->read(cgrp, cft, file, buf, nbytes, ppos);
2398 2399
	if (cft->read_u64)
		return cgroup_read_u64(cgrp, cft, file, buf, nbytes, ppos);
2400 2401
	if (cft->read_s64)
		return cgroup_read_s64(cgrp, cft, file, buf, nbytes, ppos);
2402 2403 2404
	return -EINVAL;
}

2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424
/*
 * seqfile ops/methods for returning structured data. Currently just
 * supports string->u64 maps, but can be extended in future.
 */

struct cgroup_seqfile_state {
	struct cftype *cft;
	struct cgroup *cgroup;
};

static int cgroup_map_add(struct cgroup_map_cb *cb, const char *key, u64 value)
{
	struct seq_file *sf = cb->state;
	return seq_printf(sf, "%s %llu\n", key, (unsigned long long)value);
}

static int cgroup_seqfile_show(struct seq_file *m, void *arg)
{
	struct cgroup_seqfile_state *state = m->private;
	struct cftype *cft = state->cft;
2425 2426 2427 2428 2429 2430 2431 2432
	if (cft->read_map) {
		struct cgroup_map_cb cb = {
			.fill = cgroup_map_add,
			.state = m,
		};
		return cft->read_map(state->cgroup, cft, &cb);
	}
	return cft->read_seq_string(state->cgroup, cft, m);
2433 2434
}

2435
static int cgroup_seqfile_release(struct inode *inode, struct file *file)
2436 2437 2438 2439 2440 2441
{
	struct seq_file *seq = file->private_data;
	kfree(seq->private);
	return single_release(inode, file);
}

2442
static const struct file_operations cgroup_seqfile_operations = {
2443
	.read = seq_read,
2444
	.write = cgroup_file_write,
2445 2446 2447 2448
	.llseek = seq_lseek,
	.release = cgroup_seqfile_release,
};

2449 2450 2451 2452 2453 2454 2455 2456 2457
static int cgroup_file_open(struct inode *inode, struct file *file)
{
	int err;
	struct cftype *cft;

	err = generic_file_open(inode, file);
	if (err)
		return err;
	cft = __d_cft(file->f_dentry);
2458

2459
	if (cft->read_map || cft->read_seq_string) {
2460 2461 2462
		struct cgroup_seqfile_state *state;

		state = kzalloc(sizeof(*state), GFP_USER);
2463 2464
		if (!state)
			return -ENOMEM;
2465

2466 2467 2468 2469 2470 2471 2472
		state->cft = cft;
		state->cgroup = __d_cgrp(file->f_dentry->d_parent);
		file->f_op = &cgroup_seqfile_operations;
		err = single_open(file, cgroup_seqfile_show, state);
		if (err < 0)
			kfree(state);
	} else if (cft->open)
2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
		err = cft->open(inode, file);
	else
		err = 0;

	return err;
}

static int cgroup_file_release(struct inode *inode, struct file *file)
{
	struct cftype *cft = __d_cft(file->f_dentry);
	if (cft->release)
		return cft->release(inode, file);
	return 0;
}

/*
 * 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)
{
2494 2495 2496 2497 2498 2499 2500 2501 2502 2503
	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);

2504 2505 2506 2507 2508 2509
	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;
2510 2511 2512

	cgrp = __d_cgrp(old_dentry);

2513 2514 2515 2516 2517 2518 2519
	/*
	 * This isn't a proper migration and its usefulness is very
	 * limited.  Disallow if sane_behavior.
	 */
	if (cgroup_sane_behavior(cgrp))
		return -EPERM;

2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534
	name = cgroup_alloc_name(new_dentry);
	if (!name)
		return -ENOMEM;

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

	old_name = cgrp->name;
	rcu_assign_pointer(cgrp->name, name);

	kfree_rcu(old_name, rcu_head);
	return 0;
2535 2536
}

A
Aristeu Rozanski 已提交
2537 2538 2539 2540 2541
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
2542
		return &__d_cfe(dentry)->xattrs;
A
Aristeu Rozanski 已提交
2543 2544 2545 2546 2547
}

static inline int xattr_enabled(struct dentry *dentry)
{
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
2548
	return root->flags & CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594
}

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);
}

2595
static const struct file_operations cgroup_file_operations = {
2596 2597 2598 2599 2600 2601 2602
	.read = cgroup_file_read,
	.write = cgroup_file_write,
	.llseek = generic_file_llseek,
	.open = cgroup_file_open,
	.release = cgroup_file_release,
};

A
Aristeu Rozanski 已提交
2603 2604 2605 2606 2607 2608 2609
static const struct inode_operations cgroup_file_inode_operations = {
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
};

2610
static const struct inode_operations cgroup_dir_inode_operations = {
2611
	.lookup = cgroup_lookup,
2612 2613 2614
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
A
Aristeu Rozanski 已提交
2615 2616 2617 2618
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
2619 2620
};

A
Al Viro 已提交
2621
static struct dentry *cgroup_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2622 2623 2624 2625 2626 2627 2628
{
	if (dentry->d_name.len > NAME_MAX)
		return ERR_PTR(-ENAMETOOLONG);
	d_add(dentry, NULL);
	return NULL;
}

2629 2630 2631 2632 2633
/*
 * Check if a file is a control file
 */
static inline struct cftype *__file_cft(struct file *file)
{
A
Al Viro 已提交
2634
	if (file_inode(file)->i_fop != &cgroup_file_operations)
2635 2636 2637 2638
		return ERR_PTR(-EINVAL);
	return __d_cft(file->f_dentry);
}

A
Al Viro 已提交
2639
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2640 2641
				struct super_block *sb)
{
2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658
	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
2659
		inc_nlink(dentry->d_parent->d_inode);
2660

2661 2662 2663 2664 2665 2666 2667 2668 2669
		/*
		 * 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));
2670 2671 2672
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
A
Aristeu Rozanski 已提交
2673
		inode->i_op = &cgroup_file_inode_operations;
2674 2675 2676 2677 2678 2679
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

L
Li Zefan 已提交
2680 2681 2682 2683 2684 2685 2686 2687 2688
/**
 * 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
2689
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2690
{
A
Al Viro 已提交
2691
	umode_t mode = 0;
L
Li Zefan 已提交
2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706

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

	if (cft->read || cft->read_u64 || cft->read_s64 ||
	    cft->read_map || cft->read_seq_string)
		mode |= S_IRUGO;

	if (cft->write || cft->write_u64 || cft->write_s64 ||
	    cft->write_string || cft->trigger)
		mode |= S_IWUSR;

	return mode;
}

T
Tejun Heo 已提交
2707
static int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2708
			   struct cftype *cft)
2709
{
2710
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2711
	struct cgroup *parent = __d_cgrp(dir);
2712
	struct dentry *dentry;
T
Tejun Heo 已提交
2713
	struct cfent *cfe;
2714
	int error;
A
Al Viro 已提交
2715
	umode_t mode;
2716
	char name[MAX_CGROUP_TYPE_NAMELEN + MAX_CFTYPE_NAME + 2] = { 0 };
2717

2718
	if (subsys && !(cgrp->root->flags & CGRP_ROOT_NOPREFIX)) {
2719 2720 2721 2722
		strcpy(name, subsys->name);
		strcat(name, ".");
	}
	strcat(name, cft->name);
T
Tejun Heo 已提交
2723

2724
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2725 2726 2727 2728 2729

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

2730
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2731
	if (IS_ERR(dentry)) {
2732
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2733 2734 2735
		goto out;
	}

2736 2737 2738 2739 2740
	cfe->type = (void *)cft;
	cfe->dentry = dentry;
	dentry->d_fsdata = cfe;
	simple_xattrs_init(&cfe->xattrs);

T
Tejun Heo 已提交
2741 2742 2743 2744 2745 2746 2747 2748 2749
	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);
2750 2751 2752
	return error;
}

2753
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2754
			      struct cftype cfts[], bool is_add)
2755
{
A
Aristeu Rozanski 已提交
2756
	struct cftype *cft;
T
Tejun Heo 已提交
2757 2758 2759
	int err, ret = 0;

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2760
		/* does cft->flags tell us to skip this file on @cgrp? */
2761 2762
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2763 2764 2765 2766 2767
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2768
		if (is_add) {
2769
			err = cgroup_add_file(cgrp, subsys, cft);
2770 2771 2772
			if (err)
				pr_warn("cgroup_addrm_files: failed to add %s, err=%d\n",
					cft->name, err);
T
Tejun Heo 已提交
2773
			ret = err;
2774 2775
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2776
		}
2777
	}
T
Tejun Heo 已提交
2778
	return ret;
2779 2780
}

2781
static void cgroup_cfts_prepare(void)
2782
	__acquires(&cgroup_mutex)
2783 2784 2785 2786
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
2787 2788
	 * Instead, we use cgroup_for_each_descendant_pre() and drop RCU
	 * read lock before calling cgroup_addrm_files().
2789 2790 2791 2792 2793
	 */
	mutex_lock(&cgroup_mutex);
}

static void cgroup_cfts_commit(struct cgroup_subsys *ss,
A
Aristeu Rozanski 已提交
2794
			       struct cftype *cfts, bool is_add)
2795
	__releases(&cgroup_mutex)
2796 2797
{
	LIST_HEAD(pending);
2798
	struct cgroup *cgrp, *root = &ss->root->top_cgroup;
2799
	struct super_block *sb = ss->root->sb;
2800 2801
	struct dentry *prev = NULL;
	struct inode *inode;
2802
	u64 update_before;
2803 2804

	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
2805
	if (!cfts || ss->root == &cgroup_dummy_root ||
2806 2807 2808
	    !atomic_inc_not_zero(&sb->s_active)) {
		mutex_unlock(&cgroup_mutex);
		return;
2809 2810 2811
	}

	/*
2812 2813
	 * All cgroups which are created after we drop cgroup_mutex will
	 * have the updated set of files, so we only need to update the
2814
	 * cgroups created before the current @cgroup_serial_nr_next.
2815
	 */
2816
	update_before = cgroup_serial_nr_next;
2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839

	mutex_unlock(&cgroup_mutex);

	/* @root always needs to be updated */
	inode = root->dentry->d_inode;
	mutex_lock(&inode->i_mutex);
	mutex_lock(&cgroup_mutex);
	cgroup_addrm_files(root, ss, cfts, is_add);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);

	/* add/rm files for all cgroups created before */
	rcu_read_lock();
	cgroup_for_each_descendant_pre(cgrp, root) {
		if (cgroup_is_dead(cgrp))
			continue;

		inode = cgrp->dentry->d_inode;
		dget(cgrp->dentry);
		rcu_read_unlock();

		dput(prev);
		prev = cgrp->dentry;
2840 2841 2842

		mutex_lock(&inode->i_mutex);
		mutex_lock(&cgroup_mutex);
2843
		if (cgrp->serial_nr < update_before && !cgroup_is_dead(cgrp))
2844
			cgroup_addrm_files(cgrp, ss, cfts, is_add);
2845 2846 2847
		mutex_unlock(&cgroup_mutex);
		mutex_unlock(&inode->i_mutex);

2848
		rcu_read_lock();
2849
	}
2850 2851 2852
	rcu_read_unlock();
	dput(prev);
	deactivate_super(sb);
2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868
}

/**
 * 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
2869
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
{
	struct cftype_set *set;

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

	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2880
	cgroup_cfts_commit(ss, cfts, true);
2881 2882 2883 2884 2885

	return 0;
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @ss: target cgroup subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
 * Unregister @cfts from @ss.  Files described by @cfts are removed from
 * all existing cgroups to which @ss is attached and all future cgroups
 * won't have them either.  This function can be called anytime whether @ss
 * is attached or not.
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
 * registered with @ss.
 */
A
Aristeu Rozanski 已提交
2899
int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2900 2901 2902 2903 2904 2905 2906
{
	struct cftype_set *set;

	cgroup_cfts_prepare();

	list_for_each_entry(set, &ss->cftsets, node) {
		if (set->cfts == cfts) {
2907 2908
			list_del(&set->node);
			kfree(set);
2909 2910 2911 2912 2913 2914 2915 2916 2917
			cgroup_cfts_commit(ss, cfts, false);
			return 0;
		}
	}

	cgroup_cfts_commit(ss, NULL, false);
	return -ENOENT;
}

L
Li Zefan 已提交
2918 2919 2920 2921 2922 2923
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2924
int cgroup_task_count(const struct cgroup *cgrp)
2925 2926
{
	int count = 0;
2927
	struct cgrp_cset_link *link;
2928 2929

	read_lock(&css_set_lock);
2930 2931
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
2932
	read_unlock(&css_set_lock);
2933 2934 2935
	return count;
}

2936 2937 2938 2939
/*
 * Advance a list_head iterator.  The iterator should be positioned at
 * the start of a css_set
 */
2940
static void cgroup_advance_iter(struct cgroup *cgrp, struct cgroup_iter *it)
2941
{
2942 2943
	struct list_head *l = it->cset_link;
	struct cgrp_cset_link *link;
2944
	struct css_set *cset;
2945 2946 2947 2948

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
2949 2950
		if (l == &cgrp->cset_links) {
			it->cset_link = NULL;
2951 2952
			return;
		}
2953 2954
		link = list_entry(l, struct cgrp_cset_link, cset_link);
		cset = link->cset;
2955
	} while (list_empty(&cset->tasks));
2956
	it->cset_link = l;
2957
	it->task = cset->tasks.next;
2958 2959
}

2960 2961 2962 2963 2964 2965
/*
 * 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 words after the first call to cgroup_iter_start().
 */
2966
static void cgroup_enable_task_cg_lists(void)
2967 2968 2969 2970
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
2971 2972 2973 2974 2975 2976 2977 2978
	/*
	 * 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);
2979 2980
	do_each_thread(g, p) {
		task_lock(p);
2981 2982 2983 2984 2985 2986
		/*
		 * 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))
2987 2988 2989
			list_add(&p->cg_list, &p->cgroups->tasks);
		task_unlock(p);
	} while_each_thread(g, p);
2990
	read_unlock(&tasklist_lock);
2991 2992 2993
	write_unlock(&css_set_lock);
}

2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011
/**
 * cgroup_next_sibling - find the next sibling of a given cgroup
 * @pos: the current cgroup
 *
 * This function returns the next sibling of @pos and should be called
 * under RCU read lock.  The only requirement is that @pos is accessible.
 * The next sibling is guaranteed to be returned regardless of @pos's
 * state.
 */
struct cgroup *cgroup_next_sibling(struct cgroup *pos)
{
	struct cgroup *next;

	WARN_ON_ONCE(!rcu_read_lock_held());

	/*
	 * @pos could already have been removed.  Once a cgroup is removed,
	 * its ->sibling.next is no longer updated when its next sibling
3012 3013 3014 3015 3016 3017 3018
	 * 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.
3019
	 */
3020
	if (likely(!cgroup_is_dead(pos))) {
3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043
		next = list_entry_rcu(pos->sibling.next, struct cgroup, sibling);
		if (&next->sibling != &pos->parent->children)
			return next;
		return NULL;
	}

	/*
	 * Can't dereference the next pointer.  Each cgroup is given a
	 * monotonically increasing unique serial number and always
	 * appended to the sibling list, so 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 slow, it's taken only when either the
	 * current cgroup is removed or iteration and removal race.
	 */
	list_for_each_entry_rcu(next, &pos->parent->children, sibling)
		if (next->serial_nr > pos->serial_nr)
			return next;
	return NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_sibling);

3044 3045 3046 3047 3048 3049 3050
/**
 * cgroup_next_descendant_pre - find the next descendant for pre-order walk
 * @pos: the current position (%NULL to initiate traversal)
 * @cgroup: cgroup whose descendants to walk
 *
 * To be used by cgroup_for_each_descendant_pre().  Find the next
 * descendant to visit for pre-order traversal of @cgroup's descendants.
3051 3052 3053 3054 3055
 *
 * While this function requires RCU read locking, it doesn't require the
 * whole traversal to be contained in a single RCU 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.
3056 3057 3058 3059 3060 3061 3062 3063 3064
 */
struct cgroup *cgroup_next_descendant_pre(struct cgroup *pos,
					  struct cgroup *cgroup)
{
	struct cgroup *next;

	WARN_ON_ONCE(!rcu_read_lock_held());

	/* if first iteration, pretend we just visited @cgroup */
3065
	if (!pos)
3066 3067 3068 3069 3070 3071 3072 3073
		pos = cgroup;

	/* visit the first child if exists */
	next = list_first_or_null_rcu(&pos->children, struct cgroup, sibling);
	if (next)
		return next;

	/* no child, visit my or the closest ancestor's next sibling */
3074
	while (pos != cgroup) {
3075 3076
		next = cgroup_next_sibling(pos);
		if (next)
3077 3078
			return next;
		pos = pos->parent;
3079
	}
3080 3081 3082 3083 3084

	return NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_descendant_pre);

3085 3086 3087 3088 3089 3090 3091
/**
 * cgroup_rightmost_descendant - return the rightmost descendant of a cgroup
 * @pos: cgroup of interest
 *
 * Return the rightmost descendant of @pos.  If there's no descendant,
 * @pos is returned.  This can be used during pre-order traversal to skip
 * subtree of @pos.
3092 3093 3094 3095 3096
 *
 * While this function requires RCU read locking, it doesn't require the
 * whole traversal to be contained in a single RCU critical section.  This
 * function will return the correct rightmost descendant as long as @pos is
 * accessible.
3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115
 */
struct cgroup *cgroup_rightmost_descendant(struct cgroup *pos)
{
	struct cgroup *last, *tmp;

	WARN_ON_ONCE(!rcu_read_lock_held());

	do {
		last = pos;
		/* ->prev isn't RCU safe, walk ->next till the end */
		pos = NULL;
		list_for_each_entry_rcu(tmp, &last->children, sibling)
			pos = tmp;
	} while (pos);

	return last;
}
EXPORT_SYMBOL_GPL(cgroup_rightmost_descendant);

3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135
static struct cgroup *cgroup_leftmost_descendant(struct cgroup *pos)
{
	struct cgroup *last;

	do {
		last = pos;
		pos = list_first_or_null_rcu(&pos->children, struct cgroup,
					     sibling);
	} while (pos);

	return last;
}

/**
 * cgroup_next_descendant_post - find the next descendant for post-order walk
 * @pos: the current position (%NULL to initiate traversal)
 * @cgroup: cgroup whose descendants to walk
 *
 * To be used by cgroup_for_each_descendant_post().  Find the next
 * descendant to visit for post-order traversal of @cgroup's descendants.
3136 3137 3138 3139 3140
 *
 * While this function requires RCU read locking, it doesn't require the
 * whole traversal to be contained in a single RCU 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.
3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155
 */
struct cgroup *cgroup_next_descendant_post(struct cgroup *pos,
					   struct cgroup *cgroup)
{
	struct cgroup *next;

	WARN_ON_ONCE(!rcu_read_lock_held());

	/* if first iteration, visit the leftmost descendant */
	if (!pos) {
		next = cgroup_leftmost_descendant(cgroup);
		return next != cgroup ? next : NULL;
	}

	/* if there's an unvisited sibling, visit its leftmost descendant */
3156 3157
	next = cgroup_next_sibling(pos);
	if (next)
3158 3159 3160 3161 3162 3163 3164 3165
		return cgroup_leftmost_descendant(next);

	/* no sibling left, visit parent */
	next = pos->parent;
	return next != cgroup ? next : NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_descendant_post);

3166
void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it)
3167
	__acquires(css_set_lock)
3168 3169 3170 3171 3172 3173
{
	/*
	 * The first time anyone tries to iterate across a cgroup,
	 * we need to enable the list linking each css_set to its
	 * tasks, and fix up all existing tasks.
	 */
3174 3175 3176
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

3177
	read_lock(&css_set_lock);
3178
	it->cset_link = &cgrp->cset_links;
3179
	cgroup_advance_iter(cgrp, it);
3180 3181
}

3182
struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
3183 3184 3185 3186
					struct cgroup_iter *it)
{
	struct task_struct *res;
	struct list_head *l = it->task;
3187
	struct cgrp_cset_link *link;
3188 3189

	/* If the iterator cg is NULL, we have no tasks */
3190
	if (!it->cset_link)
3191 3192 3193 3194
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
3195 3196
	link = list_entry(it->cset_link, struct cgrp_cset_link, cset_link);
	if (l == &link->cset->tasks) {
3197 3198
		/* We reached the end of this task list - move on to
		 * the next cg_cgroup_link */
3199
		cgroup_advance_iter(cgrp, it);
3200 3201 3202 3203 3204 3205
	} else {
		it->task = l;
	}
	return res;
}

3206
void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it)
3207
	__releases(css_set_lock)
3208 3209 3210 3211
{
	read_unlock(&css_set_lock);
}

3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 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 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
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);
}

/**
 * cgroup_scan_tasks - iterate though all the tasks in a cgroup
 * @scan: struct cgroup_scanner containing arguments for the scan
 *
 * Arguments include pointers to callback functions test_task() and
 * process_task().
 * Iterate through all the tasks in a cgroup, calling test_task() for each,
 * and if it returns true, call process_task() for it also.
 * The test_task pointer may be NULL, meaning always true (select all tasks).
 * Effectively duplicates cgroup_iter_{start,next,end}()
 * but does not lock css_set_lock for the call to process_task().
 * The struct cgroup_scanner may be embedded in any structure of the caller's
 * creation.
 * It is guaranteed that process_task() will act on every task that
 * is a member of the cgroup for the duration of this call. This
 * function may or may not call process_task() for tasks that exit
 * or move to a different cgroup during the call, or are forked or
 * move into the cgroup during the call.
 *
 * Note that test_task() may be called with locks held, and may in some
 * situations be called multiple times for the same task, so it should
 * be cheap.
 * If the heap pointer in the struct cgroup_scanner 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).
 */
int cgroup_scan_tasks(struct cgroup_scanner *scan)
{
	int retval, i;
	struct cgroup_iter it;
	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 ptr_heap *heap;
	struct timespec latest_time = { 0, 0 };

	if (scan->heap) {
		/* The caller supplied our heap and pre-allocated its memory */
		heap = scan->heap;
		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:
	/*
	 * Scan tasks in the cgroup, using the scanner's "test_task" callback
	 * to determine which are of interest, and using the scanner's
	 * "process_task" callback to process any of them that 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
	 * guarantees forward progress and that we don't miss any tasks.
	 */
	heap->size = 0;
	cgroup_iter_start(scan->cg, &it);
	while ((p = cgroup_iter_next(scan->cg, &it))) {
		/*
		 * Only affect tasks that qualify per the caller's callback,
		 * if he provided one
		 */
		if (scan->test_task && !scan->test_task(p, scan))
			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
		 */
	}
	cgroup_iter_end(scan->cg, &it);

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3349
			struct task_struct *q = heap->ptrs[i];
3350
			if (i == 0) {
3351 3352
				latest_time = q->start_time;
				latest_task = q;
3353 3354
			}
			/* Process the task per the caller's callback */
3355 3356
			scan->process_task(q, scan);
			put_task_struct(q);
3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371
		}
		/*
		 * 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;
}

3372 3373 3374 3375 3376
static void cgroup_transfer_one_task(struct task_struct *task,
				     struct cgroup_scanner *scan)
{
	struct cgroup *new_cgroup = scan->data;

T
Tejun Heo 已提交
3377
	mutex_lock(&cgroup_mutex);
3378
	cgroup_attach_task(new_cgroup, task, false);
T
Tejun Heo 已提交
3379
	mutex_unlock(&cgroup_mutex);
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399
}

/**
 * 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)
{
	struct cgroup_scanner scan;

	scan.cg = from;
	scan.test_task = NULL; /* select all tasks in cgroup */
	scan.process_task = cgroup_transfer_one_task;
	scan.heap = NULL;
	scan.data = to;

	return cgroup_scan_tasks(&scan);
}

3400
/*
3401
 * Stuff for reading the 'tasks'/'procs' files.
3402 3403 3404 3405 3406 3407 3408 3409
 *
 * 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.
 *
 */

3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441
/* 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;
	/* how many files are using the current array */
	int use_count;
	/* 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;
	/* protects the other fields */
	struct rw_semaphore mutex;
};

3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462
/*
 * 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);
}
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}

3463
/*
3464
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3465
 * Returns the number of unique elements.
3466
 */
3467
static int pidlist_uniq(pid_t *list, int length)
3468
{
3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497
	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;
}

static int cmppid(const void *a, const void *b)
{
	return *(pid_t *)a - *(pid_t *)b;
}

3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508
/*
 * 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.
 */
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 */
3509
	struct pid_namespace *ns = task_active_pid_ns(current);
3510

3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526
	/*
	 * We can't drop the pidlist_mutex before taking the l->mutex in case
	 * the last ref-holder is trying to remove l from the list at the same
	 * time. Holding the pidlist_mutex precludes somebody taking whichever
	 * list we find out from under us - compare release_pid_array().
	 */
	mutex_lock(&cgrp->pidlist_mutex);
	list_for_each_entry(l, &cgrp->pidlists, links) {
		if (l->key.type == type && l->key.ns == ns) {
			/* make sure l doesn't vanish out from under us */
			down_write(&l->mutex);
			mutex_unlock(&cgrp->pidlist_mutex);
			return l;
		}
	}
	/* entry not found; create a new one */
3527
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
3528 3529 3530 3531 3532 3533 3534
	if (!l) {
		mutex_unlock(&cgrp->pidlist_mutex);
		return l;
	}
	init_rwsem(&l->mutex);
	down_write(&l->mutex);
	l->key.type = type;
3535
	l->key.ns = get_pid_ns(ns);
3536 3537 3538 3539 3540 3541
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	mutex_unlock(&cgrp->pidlist_mutex);
	return l;
}

3542 3543 3544
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3545 3546
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3547 3548 3549 3550
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3551 3552
	struct cgroup_iter it;
	struct task_struct *tsk;
3553 3554 3555 3556 3557 3558 3559 3560 3561
	struct cgroup_pidlist *l;

	/*
	 * 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);
3562
	array = pidlist_allocate(length);
3563 3564 3565
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3566 3567
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
3568
		if (unlikely(n == length))
3569
			break;
3570
		/* get tgid or pid for procs or tasks file respectively */
3571 3572 3573 3574
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3575 3576
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3577
	}
3578
	cgroup_iter_end(cgrp, &it);
3579 3580 3581
	length = n;
	/* now sort & (if procs) strip out duplicates */
	sort(array, length, sizeof(pid_t), cmppid, NULL);
3582
	if (type == CGROUP_FILE_PROCS)
3583
		length = pidlist_uniq(array, length);
3584 3585
	l = cgroup_pidlist_find(cgrp, type);
	if (!l) {
3586
		pidlist_free(array);
3587
		return -ENOMEM;
3588
	}
3589
	/* store array, freeing old if necessary - lock already held */
3590
	pidlist_free(l->list);
3591 3592 3593 3594
	l->list = array;
	l->length = length;
	l->use_count++;
	up_write(&l->mutex);
3595
	*lp = l;
3596
	return 0;
3597 3598
}

B
Balbir Singh 已提交
3599
/**
L
Li Zefan 已提交
3600
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3601 3602 3603
 * @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
3604 3605 3606
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3607 3608 3609 3610
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3611
	struct cgroup *cgrp;
B
Balbir Singh 已提交
3612 3613
	struct cgroup_iter it;
	struct task_struct *tsk;
3614

B
Balbir Singh 已提交
3615
	/*
3616 3617
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3618
	 */
3619 3620
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3621 3622 3623
		 goto err;

	ret = 0;
3624
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3625

3626 3627
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
B
Balbir Singh 已提交
3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
		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;
		}
	}
3647
	cgroup_iter_end(cgrp, &it);
B
Balbir Singh 已提交
3648 3649 3650 3651 3652

err:
	return ret;
}

3653

3654
/*
3655
 * seq_file methods for the tasks/procs files. The seq_file position is the
3656
 * next pid to display; the seq_file iterator is a pointer to the pid
3657
 * in the cgroup->l->list array.
3658
 */
3659

3660
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3661
{
3662 3663 3664 3665 3666 3667
	/*
	 * 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
	 */
3668
	struct cgroup_pidlist *l = s->private;
3669 3670 3671
	int index = 0, pid = *pos;
	int *iter;

3672
	down_read(&l->mutex);
3673
	if (pid) {
3674
		int end = l->length;
S
Stephen Rothwell 已提交
3675

3676 3677
		while (index < end) {
			int mid = (index + end) / 2;
3678
			if (l->list[mid] == pid) {
3679 3680
				index = mid;
				break;
3681
			} else if (l->list[mid] <= pid)
3682 3683 3684 3685 3686 3687
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3688
	if (index >= l->length)
3689 3690
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3691
	iter = l->list + index;
3692 3693 3694 3695
	*pos = *iter;
	return iter;
}

3696
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3697
{
3698 3699
	struct cgroup_pidlist *l = s->private;
	up_read(&l->mutex);
3700 3701
}

3702
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3703
{
3704 3705 3706
	struct cgroup_pidlist *l = s->private;
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
		*pos = *p;
		return p;
	}
}

3720
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3721 3722 3723
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3724

3725 3726 3727 3728 3729 3730 3731 3732 3733
/*
 * 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,
3734 3735
};

3736
static void cgroup_release_pid_array(struct cgroup_pidlist *l)
3737
{
3738 3739 3740 3741 3742 3743 3744
	/*
	 * the case where we're the last user of this particular pidlist will
	 * have us remove it from the cgroup's list, which entails taking the
	 * mutex. since in pidlist_find the pidlist->lock depends on cgroup->
	 * pidlist_mutex, we have to take pidlist_mutex first.
	 */
	mutex_lock(&l->owner->pidlist_mutex);
3745 3746 3747
	down_write(&l->mutex);
	BUG_ON(!l->use_count);
	if (!--l->use_count) {
3748 3749 3750
		/* we're the last user if refcount is 0; remove and free */
		list_del(&l->links);
		mutex_unlock(&l->owner->pidlist_mutex);
3751
		pidlist_free(l->list);
3752 3753 3754 3755
		put_pid_ns(l->key.ns);
		up_write(&l->mutex);
		kfree(l);
		return;
3756
	}
3757
	mutex_unlock(&l->owner->pidlist_mutex);
3758
	up_write(&l->mutex);
3759 3760
}

3761
static int cgroup_pidlist_release(struct inode *inode, struct file *file)
3762
{
3763
	struct cgroup_pidlist *l;
3764 3765
	if (!(file->f_mode & FMODE_READ))
		return 0;
3766 3767 3768 3769 3770 3771
	/*
	 * the seq_file will only be initialized if the file was opened for
	 * reading; hence we check if it's not null only in that case.
	 */
	l = ((struct seq_file *)file->private_data)->private;
	cgroup_release_pid_array(l);
3772 3773 3774
	return seq_release(inode, file);
}

3775
static const struct file_operations cgroup_pidlist_operations = {
3776 3777 3778
	.read = seq_read,
	.llseek = seq_lseek,
	.write = cgroup_file_write,
3779
	.release = cgroup_pidlist_release,
3780 3781
};

3782
/*
3783 3784 3785
 * The following functions handle opens on a file that displays a pidlist
 * (tasks or procs). Prepare an array of the process/thread IDs of whoever's
 * in the cgroup.
3786
 */
3787
/* helper function for the two below it */
3788
static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
3789
{
3790
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
3791
	struct cgroup_pidlist *l;
3792
	int retval;
3793

3794
	/* Nothing to do for write-only files */
3795 3796 3797
	if (!(file->f_mode & FMODE_READ))
		return 0;

3798
	/* have the array populated */
3799
	retval = pidlist_array_load(cgrp, type, &l);
3800 3801 3802 3803
	if (retval)
		return retval;
	/* configure file information */
	file->f_op = &cgroup_pidlist_operations;
3804

3805
	retval = seq_open(file, &cgroup_pidlist_seq_operations);
3806
	if (retval) {
3807
		cgroup_release_pid_array(l);
3808
		return retval;
3809
	}
3810
	((struct seq_file *)file->private_data)->private = l;
3811 3812
	return 0;
}
3813 3814
static int cgroup_tasks_open(struct inode *unused, struct file *file)
{
3815
	return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
3816 3817 3818
}
static int cgroup_procs_open(struct inode *unused, struct file *file)
{
3819
	return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
3820
}
3821

3822
static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
3823 3824
					    struct cftype *cft)
{
3825
	return notify_on_release(cgrp);
3826 3827
}

3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
static int cgroup_write_notify_on_release(struct cgroup *cgrp,
					  struct cftype *cft,
					  u64 val)
{
	clear_bit(CGRP_RELEASABLE, &cgrp->flags);
	if (val)
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	else
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	return 0;
}

3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856
/*
 * 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);
}

3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
static void cgroup_event_remove(struct work_struct *work)
{
	struct cgroup_event *event = container_of(work, struct cgroup_event,
			remove);
	struct cgroup *cgrp = event->cgrp;

3868 3869
	remove_wait_queue(event->wqh, &event->wait);

3870 3871
	event->cft->unregister_event(cgrp, event->cft, event->eventfd);

3872 3873 3874
	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

3875 3876
	eventfd_ctx_put(event->eventfd);
	kfree(event);
3877
	cgroup_dput(cgrp);
3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
static int cgroup_event_wake(wait_queue_t *wait, unsigned mode,
		int sync, void *key)
{
	struct cgroup_event *event = container_of(wait,
			struct cgroup_event, wait);
	struct cgroup *cgrp = event->cgrp;
	unsigned long flags = (unsigned long)key;

	if (flags & POLLHUP) {
		/*
3895 3896 3897 3898 3899 3900 3901
		 * If the event has been detached at cgroup removal, we
		 * can simply return knowing the other side will cleanup
		 * for us.
		 *
		 * We can't race against event freeing since the other
		 * side will require wqh->lock via remove_wait_queue(),
		 * which we hold.
3902
		 */
3903 3904 3905 3906 3907 3908 3909 3910 3911 3912
		spin_lock(&cgrp->event_list_lock);
		if (!list_empty(&event->list)) {
			list_del_init(&event->list);
			/*
			 * We are in atomic context, but cgroup_event_remove()
			 * may sleep, so we have to call it in workqueue.
			 */
			schedule_work(&event->remove);
		}
		spin_unlock(&cgrp->event_list_lock);
3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
	}

	return 0;
}

static void cgroup_event_ptable_queue_proc(struct file *file,
		wait_queue_head_t *wqh, poll_table *pt)
{
	struct cgroup_event *event = container_of(pt,
			struct cgroup_event, pt);

	event->wqh = wqh;
	add_wait_queue(wqh, &event->wait);
}

/*
 * Parse input and register new cgroup event handler.
 *
 * Input must be in format '<event_fd> <control_fd> <args>'.
 * Interpretation of args is defined by control file implementation.
 */
static int cgroup_write_event_control(struct cgroup *cgrp, struct cftype *cft,
				      const char *buffer)
{
	struct cgroup_event *event = NULL;
3938
	struct cgroup *cgrp_cfile;
3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982
	unsigned int efd, cfd;
	struct file *efile = NULL;
	struct file *cfile = NULL;
	char *endp;
	int ret;

	efd = simple_strtoul(buffer, &endp, 10);
	if (*endp != ' ')
		return -EINVAL;
	buffer = endp + 1;

	cfd = simple_strtoul(buffer, &endp, 10);
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
	buffer = endp + 1;

	event = kzalloc(sizeof(*event), GFP_KERNEL);
	if (!event)
		return -ENOMEM;
	event->cgrp = cgrp;
	INIT_LIST_HEAD(&event->list);
	init_poll_funcptr(&event->pt, cgroup_event_ptable_queue_proc);
	init_waitqueue_func_entry(&event->wait, cgroup_event_wake);
	INIT_WORK(&event->remove, cgroup_event_remove);

	efile = eventfd_fget(efd);
	if (IS_ERR(efile)) {
		ret = PTR_ERR(efile);
		goto fail;
	}

	event->eventfd = eventfd_ctx_fileget(efile);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
		goto fail;
	}

	cfile = fget(cfd);
	if (!cfile) {
		ret = -EBADF;
		goto fail;
	}

	/* the process need read permission on control file */
A
Al Viro 已提交
3983
	/* AV: shouldn't we check that it's been opened for read instead? */
A
Al Viro 已提交
3984
	ret = inode_permission(file_inode(cfile), MAY_READ);
3985 3986 3987 3988 3989 3990 3991 3992 3993
	if (ret < 0)
		goto fail;

	event->cft = __file_cft(cfile);
	if (IS_ERR(event->cft)) {
		ret = PTR_ERR(event->cft);
		goto fail;
	}

3994 3995 3996 3997 3998 3999 4000 4001 4002 4003
	/*
	 * The file to be monitored must be in the same cgroup as
	 * cgroup.event_control is.
	 */
	cgrp_cfile = __d_cgrp(cfile->f_dentry->d_parent);
	if (cgrp_cfile != cgrp) {
		ret = -EINVAL;
		goto fail;
	}

4004 4005 4006 4007 4008 4009 4010 4011 4012 4013
	if (!event->cft->register_event || !event->cft->unregister_event) {
		ret = -EINVAL;
		goto fail;
	}

	ret = event->cft->register_event(cgrp, event->cft,
			event->eventfd, buffer);
	if (ret)
		goto fail;

4014
	efile->f_op->poll(efile, &event->pt);
4015

4016 4017 4018 4019 4020 4021 4022
	/*
	 * Events should be removed after rmdir of cgroup directory, but before
	 * destroying subsystem state objects. Let's take reference to cgroup
	 * directory dentry to do that.
	 */
	dget(cgrp->dentry);

4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046
	spin_lock(&cgrp->event_list_lock);
	list_add(&event->list, &cgrp->event_list);
	spin_unlock(&cgrp->event_list_lock);

	fput(cfile);
	fput(efile);

	return 0;

fail:
	if (cfile)
		fput(cfile);

	if (event && event->eventfd && !IS_ERR(event->eventfd))
		eventfd_ctx_put(event->eventfd);

	if (!IS_ERR_OR_NULL(efile))
		fput(efile);

	kfree(event);

	return ret;
}

4047 4048 4049
static u64 cgroup_clone_children_read(struct cgroup *cgrp,
				    struct cftype *cft)
{
4050
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4051 4052 4053 4054 4055 4056 4057
}

static int cgroup_clone_children_write(struct cgroup *cgrp,
				     struct cftype *cft,
				     u64 val)
{
	if (val)
4058
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4059
	else
4060
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4061 4062 4063
	return 0;
}

4064
static struct cftype cgroup_base_files[] = {
4065
	{
4066
		.name = "cgroup.procs",
4067
		.open = cgroup_procs_open,
B
Ben Blum 已提交
4068
		.write_u64 = cgroup_procs_write,
4069
		.release = cgroup_pidlist_release,
B
Ben Blum 已提交
4070
		.mode = S_IRUGO | S_IWUSR,
4071
	},
4072
	{
4073
		.name = "cgroup.event_control",
4074 4075 4076
		.write_string = cgroup_write_event_control,
		.mode = S_IWUGO,
	},
4077 4078
	{
		.name = "cgroup.clone_children",
4079
		.flags = CFTYPE_INSANE,
4080 4081 4082
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
4083 4084 4085 4086 4087
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
		.read_seq_string = cgroup_sane_behavior_show,
	},
4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107

	/*
	 * 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 */
		.open = cgroup_tasks_open,
		.write_u64 = cgroup_tasks_write,
		.release = cgroup_pidlist_release,
		.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,
	},
4108 4109
	{
		.name = "release_agent",
4110
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
4111 4112 4113 4114
		.read_seq_string = cgroup_release_agent_show,
		.write_string = cgroup_release_agent_write,
		.max_write_len = PATH_MAX,
	},
T
Tejun Heo 已提交
4115
	{ }	/* terminate */
4116 4117
};

4118 4119 4120 4121 4122 4123 4124 4125
/**
 * cgroup_populate_dir - selectively creation of files in a directory
 * @cgrp: target cgroup
 * @base_files: true if the base files should be added
 * @subsys_mask: mask of the subsystem ids whose files should be added
 */
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       unsigned long subsys_mask)
4126 4127 4128 4129
{
	int err;
	struct cgroup_subsys *ss;

4130
	if (base_files) {
4131
		err = cgroup_addrm_files(cgrp, NULL, cgroup_base_files, true);
4132 4133 4134
		if (err < 0)
			return err;
	}
4135

4136
	/* process cftsets of each subsystem */
4137
	for_each_root_subsys(cgrp->root, ss) {
4138
		struct cftype_set *set;
4139 4140
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
4141

T
Tejun Heo 已提交
4142
		list_for_each_entry(set, &ss->cftsets, node)
4143
			cgroup_addrm_files(cgrp, ss, set->cfts, true);
4144
	}
4145

K
KAMEZAWA Hiroyuki 已提交
4146
	/* This cgroup is ready now */
4147
	for_each_root_subsys(cgrp->root, ss) {
K
KAMEZAWA Hiroyuki 已提交
4148 4149 4150 4151 4152 4153 4154 4155 4156
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
		/*
		 * Update id->css pointer and make this css visible from
		 * CSS ID functions. This pointer will be dereferened
		 * from RCU-read-side without locks.
		 */
		if (css->id)
			rcu_assign_pointer(css->id->css, css);
	}
4157 4158 4159 4160

	return 0;
}

4161 4162 4163 4164 4165
static void css_dput_fn(struct work_struct *work)
{
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, dput_work);

4166
	cgroup_dput(css->cgroup);
4167 4168
}

4169 4170 4171 4172 4173 4174 4175 4176
static void css_release(struct percpu_ref *ref)
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

	schedule_work(&css->dput_work);
}

4177 4178
static void init_cgroup_css(struct cgroup_subsys_state *css,
			       struct cgroup_subsys *ss,
4179
			       struct cgroup *cgrp)
4180
{
4181
	css->cgroup = cgrp;
4182
	css->flags = 0;
K
KAMEZAWA Hiroyuki 已提交
4183
	css->id = NULL;
4184
	if (cgrp == cgroup_dummy_top)
4185
		css->flags |= CSS_ROOT;
4186 4187
	BUG_ON(cgrp->subsys[ss->subsys_id]);
	cgrp->subsys[ss->subsys_id] = css;
4188 4189

	/*
4190 4191 4192 4193
	 * css holds an extra ref to @cgrp->dentry which is put on the last
	 * css_put().  dput() requires process context, which css_put() may
	 * be called without.  @css->dput_work will be used to invoke
	 * dput() asynchronously from css_put().
4194 4195
	 */
	INIT_WORK(&css->dput_work, css_dput_fn);
4196 4197
}

T
Tejun Heo 已提交
4198 4199
/* invoke ->post_create() on a new CSS and mark it online if successful */
static int online_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
4200
{
T
Tejun Heo 已提交
4201 4202
	int ret = 0;

4203 4204
	lockdep_assert_held(&cgroup_mutex);

4205 4206
	if (ss->css_online)
		ret = ss->css_online(cgrp);
T
Tejun Heo 已提交
4207 4208 4209
	if (!ret)
		cgrp->subsys[ss->subsys_id]->flags |= CSS_ONLINE;
	return ret;
4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222
}

/* if the CSS is online, invoke ->pre_destory() on it and mark it offline */
static void offline_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
{
	struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

	lockdep_assert_held(&cgroup_mutex);

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

4223
	if (ss->css_offline)
4224
		ss->css_offline(cgrp);
4225 4226 4227 4228

	cgrp->subsys[ss->subsys_id]->flags &= ~CSS_ONLINE;
}

4229
/*
L
Li Zefan 已提交
4230 4231 4232 4233
 * 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
4234
 *
L
Li Zefan 已提交
4235
 * Must be called with the mutex on the parent inode held
4236 4237
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
4238
			     umode_t mode)
4239
{
4240
	struct cgroup *cgrp;
4241
	struct cgroup_name *name;
4242 4243 4244 4245 4246
	struct cgroupfs_root *root = parent->root;
	int err = 0;
	struct cgroup_subsys *ss;
	struct super_block *sb = root->sb;

T
Tejun Heo 已提交
4247
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4248 4249
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4250 4251
		return -ENOMEM;

4252 4253 4254 4255 4256
	name = cgroup_alloc_name(dentry);
	if (!name)
		goto err_free_cgrp;
	rcu_assign_pointer(cgrp->name, name);

T
Tejun Heo 已提交
4257 4258
	cgrp->id = ida_simple_get(&root->cgroup_ida, 1, 0, GFP_KERNEL);
	if (cgrp->id < 0)
4259
		goto err_free_name;
T
Tejun Heo 已提交
4260

4261 4262 4263 4264 4265 4266 4267 4268 4269
	/*
	 * 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
4270
		goto err_free_id;
4271 4272
	}

4273 4274 4275 4276 4277 4278 4279
	/* 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);

4280
	init_cgroup_housekeeping(cgrp);
4281

4282 4283 4284
	dentry->d_fsdata = cgrp;
	cgrp->dentry = dentry;

4285 4286
	cgrp->parent = parent;
	cgrp->root = parent->root;
4287

4288 4289 4290
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4291 4292
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4293

4294
	for_each_root_subsys(root, ss) {
4295
		struct cgroup_subsys_state *css;
4296

4297
		css = ss->css_alloc(cgrp);
4298 4299
		if (IS_ERR(css)) {
			err = PTR_ERR(css);
4300
			goto err_free_all;
4301
		}
4302 4303 4304 4305 4306

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

4307
		init_cgroup_css(css, ss, cgrp);
4308

4309 4310 4311
		if (ss->use_id) {
			err = alloc_css_id(ss, parent, cgrp);
			if (err)
4312
				goto err_free_all;
4313
		}
4314 4315
	}

4316 4317 4318 4319 4320
	/*
	 * 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
4321
	err = cgroup_create_file(dentry, S_IFDIR | mode, sb);
4322
	if (err < 0)
4323
		goto err_free_all;
4324
	lockdep_assert_held(&dentry->d_inode->i_mutex);
4325

4326
	cgrp->serial_nr = cgroup_serial_nr_next++;
4327

4328 4329 4330
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
	root->number_of_cgroups++;
T
Tejun Heo 已提交
4331

T
Tejun Heo 已提交
4332
	/* each css holds a ref to the cgroup's dentry */
4333
	for_each_root_subsys(root, ss)
4334
		dget(dentry);
4335

4336 4337 4338
	/* hold a ref to the parent's dentry */
	dget(parent->dentry);

T
Tejun Heo 已提交
4339
	/* creation succeeded, notify subsystems */
4340
	for_each_root_subsys(root, ss) {
T
Tejun Heo 已提交
4341 4342 4343
		err = online_css(ss, cgrp);
		if (err)
			goto err_destroy;
4344 4345 4346 4347 4348 4349 4350 4351 4352

		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;
		}
4353 4354
	}

4355
	err = cgroup_populate_dir(cgrp, true, root->subsys_mask);
4356 4357
	if (err)
		goto err_destroy;
4358 4359

	mutex_unlock(&cgroup_mutex);
4360
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4361 4362 4363

	return 0;

4364
err_free_all:
4365
	for_each_root_subsys(root, ss) {
4366 4367 4368 4369
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

		if (css) {
			percpu_ref_cancel_init(&css->refcnt);
4370
			ss->css_free(cgrp);
4371
		}
4372 4373 4374 4375
	}
	mutex_unlock(&cgroup_mutex);
	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);
T
Tejun Heo 已提交
4376 4377
err_free_id:
	ida_simple_remove(&root->cgroup_ida, cgrp->id);
4378 4379
err_free_name:
	kfree(rcu_dereference_raw(cgrp->name));
4380
err_free_cgrp:
4381
	kfree(cgrp);
4382
	return err;
4383 4384 4385 4386 4387 4388

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&dentry->d_inode->i_mutex);
	return err;
4389 4390
}

4391
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4392 4393 4394 4395 4396 4397 4398
{
	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);
}

4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440
static void cgroup_css_killed(struct cgroup *cgrp)
{
	if (!atomic_dec_and_test(&cgrp->css_kill_cnt))
		return;

	/* percpu ref's of all css's are killed, kick off the next step */
	INIT_WORK(&cgrp->destroy_work, cgroup_offline_fn);
	schedule_work(&cgrp->destroy_work);
}

static void css_ref_killed_fn(struct percpu_ref *ref)
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

	cgroup_css_killed(css->cgroup);
}

/**
 * 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.
 */
4441 4442
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4443
{
4444
	struct dentry *d = cgrp->dentry;
4445
	struct cgroup_event *event, *tmp;
4446
	struct cgroup_subsys *ss;
4447
	bool empty;
4448

4449 4450 4451
	lockdep_assert_held(&d->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

4452
	/*
T
Tejun Heo 已提交
4453 4454
	 * css_set_lock synchronizes access to ->cset_links and prevents
	 * @cgrp from being removed while __put_css_set() is in progress.
4455 4456
	 */
	read_lock(&css_set_lock);
T
Tejun Heo 已提交
4457
	empty = list_empty(&cgrp->cset_links) && list_empty(&cgrp->children);
4458 4459
	read_unlock(&css_set_lock);
	if (!empty)
4460
		return -EBUSY;
L
Li Zefan 已提交
4461

4462
	/*
4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475
	 * Block new css_tryget() by killing css refcnts.  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.  The
	 * notification callback keeps track of the number of css's to be
	 * killed and schedules cgroup_offline_fn() to perform the rest of
	 * destruction once the percpu refs of all css's are confirmed to
	 * be killed.
4476
	 */
4477
	atomic_set(&cgrp->css_kill_cnt, 1);
4478
	for_each_root_subsys(cgrp->root, ss) {
4479
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4480

4481 4482 4483 4484 4485 4486 4487 4488
		/*
		 * Killing would put the base ref, but we need to keep it
		 * alive until after ->css_offline.
		 */
		percpu_ref_get(&css->refcnt);

		atomic_inc(&cgrp->css_kill_cnt);
		percpu_ref_kill_and_confirm(&css->refcnt, css_ref_killed_fn);
4489
	}
4490
	cgroup_css_killed(cgrp);
4491 4492 4493 4494 4495 4496 4497 4498

	/*
	 * Mark @cgrp dead.  This prevents further task migration and child
	 * creation by disabling cgroup_lock_live_group().  Note that
	 * CGRP_DEAD assertion is depended upon by cgroup_next_sibling() to
	 * resume iteration after dropping RCU read lock.  See
	 * cgroup_next_sibling() for details.
	 */
4499
	set_bit(CGRP_DEAD, &cgrp->flags);
4500

4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525
	/* 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);

	/*
	 * Remove @cgrp directory.  The removal puts the base ref but we
	 * aren't quite done with @cgrp yet, so hold onto it.
	 */
	dget(d);
	cgroup_d_remove_dir(d);

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
	spin_lock(&cgrp->event_list_lock);
	list_for_each_entry_safe(event, tmp, &cgrp->event_list, list) {
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
	spin_unlock(&cgrp->event_list_lock);

4526 4527 4528
	return 0;
};

4529 4530 4531 4532 4533 4534 4535 4536 4537 4538
/**
 * cgroup_offline_fn - the second step of cgroup destruction
 * @work: cgroup->destroy_free_work
 *
 * This function is invoked from a work item for a cgroup which is being
 * destroyed after the percpu refcnts of all css's are guaranteed to be
 * seen as killed on all CPUs, and performs the rest of destruction.  This
 * is the second step of destruction described in the comment above
 * cgroup_destroy_locked().
 */
4539 4540 4541 4542 4543 4544 4545 4546 4547
static void cgroup_offline_fn(struct work_struct *work)
{
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
	struct cgroup *parent = cgrp->parent;
	struct dentry *d = cgrp->dentry;
	struct cgroup_subsys *ss;

	mutex_lock(&cgroup_mutex);

4548 4549 4550 4551
	/*
	 * css_tryget() is guaranteed to fail now.  Tell subsystems to
	 * initate destruction.
	 */
4552
	for_each_root_subsys(cgrp->root, ss)
4553
		offline_css(ss, cgrp);
4554 4555

	/*
4556 4557 4558 4559 4560
	 * Put the css refs from cgroup_destroy_locked().  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.
4561
	 */
4562
	for_each_root_subsys(cgrp->root, ss)
T
Tejun Heo 已提交
4563
		css_put(cgrp->subsys[ss->subsys_id]);
4564

4565
	/* delete this cgroup from parent->children */
4566
	list_del_rcu(&cgrp->sibling);
4567

4568 4569
	dput(d);

4570
	set_bit(CGRP_RELEASABLE, &parent->flags);
4571 4572
	check_for_release(parent);

4573
	mutex_unlock(&cgroup_mutex);
4574 4575
}

4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry)
{
	int ret;

	mutex_lock(&cgroup_mutex);
	ret = cgroup_destroy_locked(dentry->d_fsdata);
	mutex_unlock(&cgroup_mutex);

	return ret;
}

4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600
static void __init_or_module cgroup_init_cftsets(struct cgroup_subsys *ss)
{
	INIT_LIST_HEAD(&ss->cftsets);

	/*
	 * base_cftset is embedded in subsys itself, no need to worry about
	 * deregistration.
	 */
	if (ss->base_cftypes) {
		ss->base_cftset.cfts = ss->base_cftypes;
		list_add_tail(&ss->base_cftset.node, &ss->cftsets);
	}
}

4601
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4602 4603
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4604 4605

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4606

4607 4608
	mutex_lock(&cgroup_mutex);

4609 4610 4611
	/* init base cftset */
	cgroup_init_cftsets(ss);

4612
	/* Create the top cgroup state for this subsystem */
4613 4614 4615
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
	css = ss->css_alloc(cgroup_dummy_top);
4616 4617
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4618
	init_cgroup_css(css, ss, cgroup_dummy_top);
4619

L
Li Zefan 已提交
4620
	/* Update the init_css_set to contain a subsys
4621
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4622 4623
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
4624
	init_css_set.subsys[ss->subsys_id] = css;
4625 4626 4627

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

L
Li Zefan 已提交
4628 4629 4630 4631 4632
	/* 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));

4633
	BUG_ON(online_css(ss, cgroup_dummy_top));
4634

4635 4636
	mutex_unlock(&cgroup_mutex);

4637 4638 4639 4640 4641 4642 4643 4644 4645 4646
	/* this function shouldn't be used with modular subsystems, since they
	 * need to register a subsys_id, among other things */
	BUG_ON(ss->module);
}

/**
 * cgroup_load_subsys: load and register a modular subsystem at runtime
 * @ss: the subsystem to load
 *
 * This function should be called in a modular subsystem's initcall. If the
T
Thomas Weber 已提交
4647
 * subsystem is built as a module, it will be assigned a new subsys_id and set
4648 4649 4650 4651 4652 4653
 * up for use. If the subsystem is built-in anyway, work is delegated to the
 * simpler cgroup_init_subsys.
 */
int __init_or_module cgroup_load_subsys(struct cgroup_subsys *ss)
{
	struct cgroup_subsys_state *css;
4654
	int i, ret;
4655
	struct hlist_node *tmp;
4656
	struct css_set *cset;
4657
	unsigned long key;
4658 4659 4660

	/* check name and function validity */
	if (ss->name == NULL || strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN ||
4661
	    ss->css_alloc == NULL || ss->css_free == NULL)
4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677
		return -EINVAL;

	/*
	 * we don't support callbacks in modular subsystems. this check is
	 * before the ss->module check for consistency; a subsystem that could
	 * be a module should still have no callbacks even if the user isn't
	 * compiling it as one.
	 */
	if (ss->fork || ss->exit)
		return -EINVAL;

	/*
	 * an optionally modular subsystem is built-in: we want to do nothing,
	 * since cgroup_init_subsys will have already taken care of it.
	 */
	if (ss->module == NULL) {
4678
		/* a sanity check */
4679
		BUG_ON(cgroup_subsys[ss->subsys_id] != ss);
4680 4681 4682
		return 0;
	}

4683 4684 4685
	/* init base cftset */
	cgroup_init_cftsets(ss);

4686
	mutex_lock(&cgroup_mutex);
4687
	cgroup_subsys[ss->subsys_id] = ss;
4688 4689

	/*
4690
	 * no ss->css_alloc seems to need anything important in the ss
4691
	 * struct, so this can happen first (i.e. before the dummy root
4692
	 * attachment).
4693
	 */
4694
	css = ss->css_alloc(cgroup_dummy_top);
4695
	if (IS_ERR(css)) {
4696 4697
		/* failure case - need to deassign the cgroup_subsys[] slot. */
		cgroup_subsys[ss->subsys_id] = NULL;
4698 4699 4700 4701
		mutex_unlock(&cgroup_mutex);
		return PTR_ERR(css);
	}

4702 4703
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
4704 4705

	/* our new subsystem will be attached to the dummy hierarchy. */
4706
	init_cgroup_css(css, ss, cgroup_dummy_top);
4707 4708
	/* init_idr must be after init_cgroup_css because it sets css->id. */
	if (ss->use_id) {
4709 4710 4711
		ret = cgroup_init_idr(ss, css);
		if (ret)
			goto err_unload;
4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722
	}

	/*
	 * Now we need to entangle the css into the existing css_sets. unlike
	 * in cgroup_init_subsys, there are now multiple css_sets, so each one
	 * will need a new pointer to it; done by iterating the css_set_table.
	 * furthermore, modifying the existing css_sets will corrupt the hash
	 * table state, so each changed css_set will need its hash recomputed.
	 * this is all done under the css_set_lock.
	 */
	write_lock(&css_set_lock);
4723
	hash_for_each_safe(css_set_table, i, tmp, cset, hlist) {
4724
		/* skip entries that we already rehashed */
4725
		if (cset->subsys[ss->subsys_id])
4726 4727
			continue;
		/* remove existing entry */
4728
		hash_del(&cset->hlist);
4729
		/* set new value */
4730
		cset->subsys[ss->subsys_id] = css;
4731
		/* recompute hash and restore entry */
4732 4733
		key = css_set_hash(cset->subsys);
		hash_add(css_set_table, &cset->hlist, key);
4734 4735 4736
	}
	write_unlock(&css_set_lock);

4737
	ret = online_css(ss, cgroup_dummy_top);
T
Tejun Heo 已提交
4738 4739
	if (ret)
		goto err_unload;
4740

4741 4742 4743
	/* success! */
	mutex_unlock(&cgroup_mutex);
	return 0;
4744 4745 4746 4747 4748 4749

err_unload:
	mutex_unlock(&cgroup_mutex);
	/* @ss can't be mounted here as try_module_get() would fail */
	cgroup_unload_subsys(ss);
	return ret;
4750
}
4751
EXPORT_SYMBOL_GPL(cgroup_load_subsys);
4752

B
Ben Blum 已提交
4753 4754 4755 4756 4757 4758 4759 4760 4761 4762
/**
 * cgroup_unload_subsys: unload a modular subsystem
 * @ss: the subsystem to unload
 *
 * This function should be called in a modular subsystem's exitcall. When this
 * function is invoked, the refcount on the subsystem's module will be 0, so
 * the subsystem will not be attached to any hierarchy.
 */
void cgroup_unload_subsys(struct cgroup_subsys *ss)
{
4763
	struct cgrp_cset_link *link;
B
Ben Blum 已提交
4764 4765 4766 4767 4768 4769 4770 4771

	BUG_ON(ss->module == NULL);

	/*
	 * we shouldn't be called if the subsystem is in use, and the use of
	 * try_module_get in parse_cgroupfs_options should ensure that it
	 * doesn't start being used while we're killing it off.
	 */
4772
	BUG_ON(ss->root != &cgroup_dummy_root);
B
Ben Blum 已提交
4773 4774

	mutex_lock(&cgroup_mutex);
4775

4776
	offline_css(ss, cgroup_dummy_top);
4777

T
Tejun Heo 已提交
4778
	if (ss->use_id)
4779 4780
		idr_destroy(&ss->idr);

B
Ben Blum 已提交
4781
	/* deassign the subsys_id */
4782
	cgroup_subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4783

4784
	/* remove subsystem from the dummy root's list of subsystems */
4785
	list_del_init(&ss->sibling);
B
Ben Blum 已提交
4786 4787

	/*
4788 4789 4790
	 * disentangle the css from all css_sets attached to the dummy
	 * top. as in loading, we need to pay our respects to the hashtable
	 * gods.
B
Ben Blum 已提交
4791 4792
	 */
	write_lock(&css_set_lock);
4793
	list_for_each_entry(link, &cgroup_dummy_top->cset_links, cset_link) {
4794
		struct css_set *cset = link->cset;
4795
		unsigned long key;
B
Ben Blum 已提交
4796

4797 4798 4799 4800
		hash_del(&cset->hlist);
		cset->subsys[ss->subsys_id] = NULL;
		key = css_set_hash(cset->subsys);
		hash_add(css_set_table, &cset->hlist, key);
B
Ben Blum 已提交
4801 4802 4803 4804
	}
	write_unlock(&css_set_lock);

	/*
4805 4806 4807 4808
	 * remove subsystem's css from the cgroup_dummy_top and free it -
	 * need to free before marking as null because ss->css_free needs
	 * the cgrp->subsys pointer to find their state. note that this
	 * also takes care of freeing the css_id.
B
Ben Blum 已提交
4809
	 */
4810 4811
	ss->css_free(cgroup_dummy_top);
	cgroup_dummy_top->subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4812 4813 4814 4815 4816

	mutex_unlock(&cgroup_mutex);
}
EXPORT_SYMBOL_GPL(cgroup_unload_subsys);

4817
/**
L
Li Zefan 已提交
4818 4819 4820 4821
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4822 4823 4824 4825
 */
int __init cgroup_init_early(void)
{
	int i;
4826
	atomic_set(&init_css_set.refcount, 1);
4827
	INIT_LIST_HEAD(&init_css_set.cgrp_links);
4828
	INIT_LIST_HEAD(&init_css_set.tasks);
4829
	INIT_HLIST_NODE(&init_css_set.hlist);
4830
	css_set_count = 1;
4831 4832
	init_cgroup_root(&cgroup_dummy_root);
	cgroup_root_count = 1;
4833 4834
	init_task.cgroups = &init_css_set;

4835
	init_cgrp_cset_link.cset = &init_css_set;
4836 4837
	init_cgrp_cset_link.cgrp = cgroup_dummy_top;
	list_add(&init_cgrp_cset_link.cset_link, &cgroup_dummy_top->cset_links);
4838
	list_add(&init_cgrp_cset_link.cgrp_link, &init_css_set.cgrp_links);
4839

4840
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
4841
		struct cgroup_subsys *ss = cgroup_subsys[i];
4842

4843 4844 4845 4846
		/* at bootup time, we don't worry about modular subsystems */
		if (!ss || ss->module)
			continue;

4847 4848
		BUG_ON(!ss->name);
		BUG_ON(strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN);
4849 4850
		BUG_ON(!ss->css_alloc);
		BUG_ON(!ss->css_free);
4851
		if (ss->subsys_id != i) {
D
Diego Calleja 已提交
4852
			printk(KERN_ERR "cgroup: Subsys %s id == %d\n",
4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863
			       ss->name, ss->subsys_id);
			BUG();
		}

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

/**
L
Li Zefan 已提交
4864 4865 4866 4867
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4868 4869 4870 4871 4872
 */
int __init cgroup_init(void)
{
	int err;
	int i;
4873
	unsigned long key;
4874 4875 4876 4877

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4878

4879
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
4880
		struct cgroup_subsys *ss = cgroup_subsys[i];
4881 4882 4883 4884

		/* at bootup time, we don't worry about modular subsystems */
		if (!ss || ss->module)
			continue;
4885 4886
		if (!ss->early_init)
			cgroup_init_subsys(ss);
K
KAMEZAWA Hiroyuki 已提交
4887
		if (ss->use_id)
4888
			cgroup_init_idr(ss, init_css_set.subsys[ss->subsys_id]);
4889 4890
	}

4891
	/* Add init_css_set to the hash table */
4892 4893
	key = css_set_hash(init_css_set.subsys);
	hash_add(css_set_table, &init_css_set.hlist, key);
4894 4895

	/* allocate id for the dummy hierarchy */
T
Tejun Heo 已提交
4896 4897 4898
	mutex_lock(&cgroup_mutex);
	mutex_lock(&cgroup_root_mutex);

4899
	BUG_ON(cgroup_init_root_id(&cgroup_dummy_root));
4900

T
Tejun Heo 已提交
4901 4902 4903
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);

4904 4905 4906 4907 4908 4909
	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4910
	err = register_filesystem(&cgroup_fs_type);
4911 4912
	if (err < 0) {
		kobject_put(cgroup_kobj);
4913
		goto out;
4914
	}
4915

L
Li Zefan 已提交
4916
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4917

4918
out:
4919 4920 4921
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4922 4923
	return err;
}
4924

4925 4926 4927 4928 4929 4930
/*
 * 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,
4931
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4932 4933 4934 4935 4936 4937
 *    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 */
4938
int proc_cgroup_show(struct seq_file *m, void *v)
4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960
{
	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);

4961
	for_each_active_root(root) {
4962
		struct cgroup_subsys *ss;
4963
		struct cgroup *cgrp;
4964 4965
		int count = 0;

4966
		seq_printf(m, "%d:", root->hierarchy_id);
4967
		for_each_root_subsys(root, ss)
4968
			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4969 4970 4971
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4972
		seq_putc(m, ':');
4973
		cgrp = task_cgroup_from_root(tsk, root);
4974
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994
		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)
{
	int i;

4995
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4996 4997 4998 4999 5000
	/*
	 * 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.
	 */
5001 5002
	mutex_lock(&cgroup_mutex);
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
5003
		struct cgroup_subsys *ss = cgroup_subsys[i];
B
Ben Blum 已提交
5004 5005
		if (ss == NULL)
			continue;
5006 5007
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
5008
			   ss->root->number_of_cgroups, !ss->disabled);
5009 5010 5011 5012 5013 5014 5015
	}
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
5016
	return single_open(file, proc_cgroupstats_show, NULL);
5017 5018
}

5019
static const struct file_operations proc_cgroupstats_operations = {
5020 5021 5022 5023 5024 5025
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

5026 5027
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
5028
 * @child: pointer to task_struct of forking parent process.
5029 5030 5031 5032 5033
 *
 * 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
5034 5035 5036 5037
 * 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.
5038 5039 5040 5041 5042 5043
 *
 * 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)
{
5044
	task_lock(current);
5045 5046
	child->cgroups = current->cgroups;
	get_css_set(child->cgroups);
5047
	task_unlock(current);
5048
	INIT_LIST_HEAD(&child->cg_list);
5049 5050
}

5051
/**
L
Li Zefan 已提交
5052 5053 5054
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
5055 5056 5057 5058 5059
 * 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
 * cgroup_iter_start() - to guarantee that the new task ends up on its
 * list.
L
Li Zefan 已提交
5060
 */
5061 5062
void cgroup_post_fork(struct task_struct *child)
{
5063 5064
	int i;

5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075
	/*
	 * 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.
	 */
5076 5077
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
5078 5079
		task_lock(child);
		if (list_empty(&child->cg_list))
5080
			list_add(&child->cg_list, &child->cgroups->tasks);
5081
		task_unlock(child);
5082 5083
		write_unlock(&css_set_lock);
	}
5084 5085 5086 5087 5088 5089 5090

	/*
	 * 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) {
5091 5092 5093 5094 5095 5096 5097 5098 5099
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, and the builtin section of the subsys
		 * array is immutable, so we don't need to lock the
		 * subsys array here. On the other hand, modular section
		 * of the array can be freed at module unload, so we
		 * can't touch that.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
5100
			struct cgroup_subsys *ss = cgroup_subsys[i];
5101 5102 5103 5104 5105

			if (ss->fork)
				ss->fork(child);
		}
	}
5106
}
5107

5108 5109 5110
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
5111
 * @run_callback: run exit callbacks?
5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139
 *
 * 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,
5140 5141
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
5142 5143 5144
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
5145
	struct css_set *cset;
5146
	int i;
5147 5148 5149 5150 5151 5152 5153 5154 5155

	/*
	 * 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))
5156
			list_del_init(&tsk->cg_list);
5157 5158 5159
		write_unlock(&css_set_lock);
	}

5160 5161
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
5162
	cset = tsk->cgroups;
5163
	tsk->cgroups = &init_css_set;
5164 5165

	if (run_callbacks && need_forkexit_callback) {
5166 5167 5168 5169 5170
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, see cgroup_post_fork() for details.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
5171
			struct cgroup_subsys *ss = cgroup_subsys[i];
5172

5173 5174
			if (ss->exit) {
				struct cgroup *old_cgrp =
5175
					rcu_dereference_raw(cset->subsys[i])->cgroup;
5176
				struct cgroup *cgrp = task_cgroup(tsk, i);
5177
				ss->exit(cgrp, old_cgrp, tsk);
5178 5179 5180
			}
		}
	}
5181
	task_unlock(tsk);
5182

5183
	put_css_set_taskexit(cset);
5184
}
5185

5186
static void check_for_release(struct cgroup *cgrp)
5187
{
5188
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
5189
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
5190 5191
		/*
		 * Control Group is currently removeable. If it's not
5192
		 * already queued for a userspace notification, queue
5193 5194
		 * it now
		 */
5195
		int need_schedule_work = 0;
5196

5197
		raw_spin_lock(&release_list_lock);
5198
		if (!cgroup_is_dead(cgrp) &&
5199 5200
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
5201 5202
			need_schedule_work = 1;
		}
5203
		raw_spin_unlock(&release_list_lock);
5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235
		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);
5236
	raw_spin_lock(&release_list_lock);
5237 5238 5239
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
5240
		char *pathbuf = NULL, *agentbuf = NULL;
5241
		struct cgroup *cgrp = list_entry(release_list.next,
5242 5243
						    struct cgroup,
						    release_list);
5244
		list_del_init(&cgrp->release_list);
5245
		raw_spin_unlock(&release_list_lock);
5246
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
5247 5248 5249 5250 5251 5252 5253
		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;
5254 5255

		i = 0;
5256 5257
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271
		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);
5272 5273 5274
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5275
		raw_spin_lock(&release_list_lock);
5276
	}
5277
	raw_spin_unlock(&release_list_lock);
5278 5279
	mutex_unlock(&cgroup_mutex);
}
5280 5281 5282 5283 5284 5285 5286 5287 5288

static int __init cgroup_disable(char *str)
{
	int i;
	char *token;

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5289
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
5290
			struct cgroup_subsys *ss = cgroup_subsys[i];
5291

5292 5293 5294 5295 5296 5297 5298 5299
			/*
			 * cgroup_disable, being at boot time, can't
			 * know about module subsystems, so we don't
			 * worry about them.
			 */
			if (!ss || ss->module)
				continue;

5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310
			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
5311 5312 5313 5314 5315

/*
 * Functons for CSS ID.
 */

5316
/* to get ID other than 0, this should be called when !cgroup_is_dead() */
K
KAMEZAWA Hiroyuki 已提交
5317 5318
unsigned short css_id(struct cgroup_subsys_state *css)
{
5319 5320 5321 5322 5323 5324 5325
	struct css_id *cssid;

	/*
	 * This css_id() can return correct value when somone has refcnt
	 * on this or this is under rcu_read_lock(). Once css->id is allocated,
	 * it's unchanged until freed.
	 */
5326
	cssid = rcu_dereference_raw(css->id);
K
KAMEZAWA Hiroyuki 已提交
5327 5328 5329 5330 5331

	if (cssid)
		return cssid->id;
	return 0;
}
B
Ben Blum 已提交
5332
EXPORT_SYMBOL_GPL(css_id);
K
KAMEZAWA Hiroyuki 已提交
5333

5334 5335 5336 5337 5338 5339
/**
 *  css_is_ancestor - test "root" css is an ancestor of "child"
 * @child: the css to be tested.
 * @root: the css supporsed to be an ancestor of the child.
 *
 * Returns true if "root" is an ancestor of "child" in its hierarchy. Because
5340
 * this function reads css->id, the caller must hold rcu_read_lock().
5341 5342 5343 5344 5345 5346
 * But, considering usual usage, the csses should be valid objects after test.
 * Assuming that the caller will do some action to the child if this returns
 * returns true, the caller must take "child";s reference count.
 * If "child" is valid object and this returns true, "root" is valid, too.
 */

K
KAMEZAWA Hiroyuki 已提交
5347
bool css_is_ancestor(struct cgroup_subsys_state *child,
5348
		    const struct cgroup_subsys_state *root)
K
KAMEZAWA Hiroyuki 已提交
5349
{
5350 5351
	struct css_id *child_id;
	struct css_id *root_id;
K
KAMEZAWA Hiroyuki 已提交
5352

5353
	child_id  = rcu_dereference(child->id);
5354 5355
	if (!child_id)
		return false;
5356
	root_id = rcu_dereference(root->id);
5357 5358 5359 5360 5361 5362 5363
	if (!root_id)
		return false;
	if (child_id->depth < root_id->depth)
		return false;
	if (child_id->stack[root_id->depth] != root_id->id)
		return false;
	return true;
K
KAMEZAWA Hiroyuki 已提交
5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376
}

void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css)
{
	struct css_id *id = css->id;
	/* When this is called before css_id initialization, id can be NULL */
	if (!id)
		return;

	BUG_ON(!ss->use_id);

	rcu_assign_pointer(id->css, NULL);
	rcu_assign_pointer(css->id, NULL);
5377
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5378
	idr_remove(&ss->idr, id->id);
5379
	spin_unlock(&ss->id_lock);
5380
	kfree_rcu(id, rcu_head);
K
KAMEZAWA Hiroyuki 已提交
5381
}
B
Ben Blum 已提交
5382
EXPORT_SYMBOL_GPL(free_css_id);
K
KAMEZAWA Hiroyuki 已提交
5383 5384 5385 5386 5387 5388 5389 5390 5391

/*
 * This is called by init or create(). Then, calls to this function are
 * always serialized (By cgroup_mutex() at create()).
 */

static struct css_id *get_new_cssid(struct cgroup_subsys *ss, int depth)
{
	struct css_id *newid;
T
Tejun Heo 已提交
5392
	int ret, size;
K
KAMEZAWA Hiroyuki 已提交
5393 5394 5395 5396 5397 5398 5399

	BUG_ON(!ss->use_id);

	size = sizeof(*newid) + sizeof(unsigned short) * (depth + 1);
	newid = kzalloc(size, GFP_KERNEL);
	if (!newid)
		return ERR_PTR(-ENOMEM);
T
Tejun Heo 已提交
5400 5401

	idr_preload(GFP_KERNEL);
5402
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5403
	/* Don't use 0. allocates an ID of 1-65535 */
T
Tejun Heo 已提交
5404
	ret = idr_alloc(&ss->idr, newid, 1, CSS_ID_MAX + 1, GFP_NOWAIT);
5405
	spin_unlock(&ss->id_lock);
T
Tejun Heo 已提交
5406
	idr_preload_end();
K
KAMEZAWA Hiroyuki 已提交
5407 5408

	/* Returns error when there are no free spaces for new ID.*/
T
Tejun Heo 已提交
5409
	if (ret < 0)
K
KAMEZAWA Hiroyuki 已提交
5410 5411
		goto err_out;

T
Tejun Heo 已提交
5412
	newid->id = ret;
K
KAMEZAWA Hiroyuki 已提交
5413 5414 5415 5416
	newid->depth = depth;
	return newid;
err_out:
	kfree(newid);
T
Tejun Heo 已提交
5417
	return ERR_PTR(ret);
K
KAMEZAWA Hiroyuki 已提交
5418 5419 5420

}

5421 5422
static int __init_or_module cgroup_init_idr(struct cgroup_subsys *ss,
					    struct cgroup_subsys_state *rootcss)
K
KAMEZAWA Hiroyuki 已提交
5423 5424 5425
{
	struct css_id *newid;

5426
	spin_lock_init(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443
	idr_init(&ss->idr);

	newid = get_new_cssid(ss, 0);
	if (IS_ERR(newid))
		return PTR_ERR(newid);

	newid->stack[0] = newid->id;
	newid->css = rootcss;
	rootcss->id = newid;
	return 0;
}

static int alloc_css_id(struct cgroup_subsys *ss, struct cgroup *parent,
			struct cgroup *child)
{
	int subsys_id, i, depth = 0;
	struct cgroup_subsys_state *parent_css, *child_css;
5444
	struct css_id *child_id, *parent_id;
K
KAMEZAWA Hiroyuki 已提交
5445 5446 5447 5448 5449

	subsys_id = ss->subsys_id;
	parent_css = parent->subsys[subsys_id];
	child_css = child->subsys[subsys_id];
	parent_id = parent_css->id;
5450
	depth = parent_id->depth + 1;
K
KAMEZAWA Hiroyuki 已提交
5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487

	child_id = get_new_cssid(ss, depth);
	if (IS_ERR(child_id))
		return PTR_ERR(child_id);

	for (i = 0; i < depth; i++)
		child_id->stack[i] = parent_id->stack[i];
	child_id->stack[depth] = child_id->id;
	/*
	 * child_id->css pointer will be set after this cgroup is available
	 * see cgroup_populate_dir()
	 */
	rcu_assign_pointer(child_css->id, child_id);

	return 0;
}

/**
 * css_lookup - lookup css by id
 * @ss: cgroup subsys to be looked into.
 * @id: the id
 *
 * Returns pointer to cgroup_subsys_state if there is valid one with id.
 * NULL if not. Should be called under rcu_read_lock()
 */
struct cgroup_subsys_state *css_lookup(struct cgroup_subsys *ss, int id)
{
	struct css_id *cssid = NULL;

	BUG_ON(!ss->use_id);
	cssid = idr_find(&ss->idr, id);

	if (unlikely(!cssid))
		return NULL;

	return rcu_dereference(cssid->css);
}
B
Ben Blum 已提交
5488
EXPORT_SYMBOL_GPL(css_lookup);
K
KAMEZAWA Hiroyuki 已提交
5489

S
Stephane Eranian 已提交
5490 5491 5492 5493 5494 5495 5496 5497 5498
/*
 * get corresponding css from file open on cgroupfs directory
 */
struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id)
{
	struct cgroup *cgrp;
	struct inode *inode;
	struct cgroup_subsys_state *css;

A
Al Viro 已提交
5499
	inode = file_inode(f);
S
Stephane Eranian 已提交
5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512
	/* check in cgroup filesystem dir */
	if (inode->i_op != &cgroup_dir_inode_operations)
		return ERR_PTR(-EBADF);

	if (id < 0 || id >= CGROUP_SUBSYS_COUNT)
		return ERR_PTR(-EINVAL);

	/* get cgroup */
	cgrp = __d_cgrp(f->f_dentry);
	css = cgrp->subsys[id];
	return css ? css : ERR_PTR(-ENOENT);
}

5513
#ifdef CONFIG_CGROUP_DEBUG
L
Li Zefan 已提交
5514
static struct cgroup_subsys_state *debug_css_alloc(struct cgroup *cgrp)
5515 5516 5517 5518 5519 5520 5521 5522 5523
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

L
Li Zefan 已提交
5524
static void debug_css_free(struct cgroup *cgrp)
5525
{
L
Li Zefan 已提交
5526
	kfree(cgrp->subsys[debug_subsys_id]);
5527 5528
}

L
Li Zefan 已提交
5529
static u64 debug_taskcount_read(struct cgroup *cgrp, struct cftype *cft)
5530
{
L
Li Zefan 已提交
5531
	return cgroup_task_count(cgrp);
5532 5533
}

L
Li Zefan 已提交
5534
static u64 current_css_set_read(struct cgroup *cgrp, struct cftype *cft)
5535 5536 5537 5538
{
	return (u64)(unsigned long)current->cgroups;
}

L
Li Zefan 已提交
5539 5540
static u64 current_css_set_refcount_read(struct cgroup *cgrp,
					 struct cftype *cft)
5541 5542 5543 5544 5545 5546 5547 5548 5549
{
	u64 count;

	rcu_read_lock();
	count = atomic_read(&current->cgroups->refcount);
	rcu_read_unlock();
	return count;
}

L
Li Zefan 已提交
5550
static int current_css_set_cg_links_read(struct cgroup *cgrp,
5551 5552 5553
					 struct cftype *cft,
					 struct seq_file *seq)
{
5554
	struct cgrp_cset_link *link;
5555
	struct css_set *cset;
5556 5557 5558

	read_lock(&css_set_lock);
	rcu_read_lock();
5559
	cset = rcu_dereference(current->cgroups);
5560
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5561 5562 5563 5564 5565 5566 5567
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5568 5569
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5570 5571 5572 5573 5574 5575 5576
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
L
Li Zefan 已提交
5577
static int cgroup_css_links_read(struct cgroup *cgrp,
5578 5579 5580
				 struct cftype *cft,
				 struct seq_file *seq)
{
5581
	struct cgrp_cset_link *link;
5582 5583

	read_lock(&css_set_lock);
L
Li Zefan 已提交
5584
	list_for_each_entry(link, &cgrp->cset_links, cset_link) {
5585
		struct css_set *cset = link->cset;
5586 5587
		struct task_struct *task;
		int count = 0;
5588 5589
		seq_printf(seq, "css_set %p\n", cset);
		list_for_each_entry(task, &cset->tasks, cg_list) {
5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602
			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;
}

5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623
static u64 releasable_read(struct cgroup *cgrp, struct cftype *cft)
{
	return test_bit(CGRP_RELEASABLE, &cgrp->flags);
}

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,
	},

5624 5625 5626 5627 5628 5629 5630 5631 5632 5633
	{
		.name = "current_css_set_cg_links",
		.read_seq_string = current_css_set_cg_links_read,
	},

	{
		.name = "cgroup_css_links",
		.read_seq_string = cgroup_css_links_read,
	},

5634 5635 5636 5637 5638
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5639 5640
	{ }	/* terminate */
};
5641 5642 5643

struct cgroup_subsys debug_subsys = {
	.name = "debug",
5644 5645
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5646
	.subsys_id = debug_subsys_id,
5647
	.base_cftypes = debug_files,
5648 5649
};
#endif /* CONFIG_CGROUP_DEBUG */