cgroup.c 151.9 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
#ifdef CONFIG_PROVE_RCU
DEFINE_MUTEX(cgroup_mutex);
84
EXPORT_SYMBOL_GPL(cgroup_mutex);	/* only for lockdep */
T
Tejun Heo 已提交
85
#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

218 219
static struct cftype cgroup_base_files[];

220
static void cgroup_offline_fn(struct work_struct *work);
221
static int cgroup_destroy_locked(struct cgroup *cgrp);
222 223
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
			      struct cftype cfts[], bool is_add);
224

225
/* convenient tests for these bits */
226
static inline bool cgroup_is_dead(const struct cgroup *cgrp)
227
{
228
	return test_bit(CGRP_DEAD, &cgrp->flags);
229 230
}

231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249
/**
 * 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);
250

251
static int cgroup_is_releasable(const struct cgroup *cgrp)
252 253
{
	const int bits =
254 255 256
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
257 258
}

259
static int notify_on_release(const struct cgroup *cgrp)
260
{
261
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
262 263
}

264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288
/**
 * for_each_subsys - iterate all loaded cgroup subsystems
 * @ss: the iteration cursor
 * @i: the index of @ss, CGROUP_SUBSYS_COUNT after reaching the end
 *
 * Should be called under cgroup_mutex.
 */
#define for_each_subsys(ss, i)						\
	for ((i) = 0; (i) < CGROUP_SUBSYS_COUNT; (i)++)			\
		if (({ lockdep_assert_held(&cgroup_mutex);		\
		       !((ss) = cgroup_subsys[i]); })) { }		\
		else

/**
 * for_each_builtin_subsys - iterate all built-in cgroup subsystems
 * @ss: the iteration cursor
 * @i: the index of @ss, CGROUP_BUILTIN_SUBSYS_COUNT after reaching the end
 *
 * Bulit-in subsystems are always present and iteration itself doesn't
 * require any synchronization.
 */
#define for_each_builtin_subsys(ss, i)					\
	for ((i) = 0; (i) < CGROUP_BUILTIN_SUBSYS_COUNT &&		\
	     (((ss) = cgroup_subsys[i]) || true); (i)++)

289 290 291
/* iterate each subsystem attached to a hierarchy */
#define for_each_root_subsys(root, ss)					\
	list_for_each_entry((ss), &(root)->subsys_list, sibling)
292

293 294 295
/* iterate across the active hierarchies */
#define for_each_active_root(root)					\
	list_for_each_entry((root), &cgroup_roots, root_list)
296

297 298 299 300 301
static inline struct cgroup *__d_cgrp(struct dentry *dentry)
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
302
static inline struct cfent *__d_cfe(struct dentry *dentry)
303 304 305 306
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
307 308 309 310 311
static inline struct cftype *__d_cft(struct dentry *dentry)
{
	return __d_cfe(dentry)->type;
}

312 313 314 315
/**
 * 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
316 317
 * On success, returns true; the mutex should be later unlocked.  On
 * failure returns false with no lock held.
318
 */
319
static bool cgroup_lock_live_group(struct cgroup *cgrp)
320 321
{
	mutex_lock(&cgroup_mutex);
322
	if (cgroup_is_dead(cgrp)) {
323 324 325 326 327 328
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}

329 330 331
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
332
static DEFINE_RAW_SPINLOCK(release_list_lock);
333 334
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
335
static void check_for_release(struct cgroup *cgrp);
336

337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
/*
 * 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;
355 356 357 358 359 360 361 362 363 364
};

/* 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;
365
static struct cgrp_cset_link init_cgrp_cset_link;
366

367 368
static int cgroup_init_idr(struct cgroup_subsys *ss,
			   struct cgroup_subsys_state *css);
K
KAMEZAWA Hiroyuki 已提交
369

370 371 372 373 374 375
/* 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;

376 377 378 379 380
/*
 * 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.
 */
381
#define CSS_SET_HASH_BITS	7
382
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
383

384
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
385
{
386
	unsigned long key = 0UL;
387 388
	struct cgroup_subsys *ss;
	int i;
389

390
	for_each_subsys(ss, i)
391 392
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
393

394
	return key;
395 396
}

397 398 399 400
/* 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 */
401
static int use_task_css_set_links __read_mostly;
402

403
static void __put_css_set(struct css_set *cset, int taskexit)
404
{
405
	struct cgrp_cset_link *link, *tmp_link;
406

407 408 409 410 411
	/*
	 * Ensure that the refcount doesn't hit zero while any readers
	 * can see it. Similar to atomic_dec_and_lock(), but for an
	 * rwlock
	 */
412
	if (atomic_add_unless(&cset->refcount, -1, 1))
413 414
		return;
	write_lock(&css_set_lock);
415
	if (!atomic_dec_and_test(&cset->refcount)) {
416 417 418
		write_unlock(&css_set_lock);
		return;
	}
419

420
	/* This css_set is dead. unlink it and release cgroup refcounts */
421
	hash_del(&cset->hlist);
422 423
	css_set_count--;

424
	list_for_each_entry_safe(link, tmp_link, &cset->cgrp_links, cgrp_link) {
425
		struct cgroup *cgrp = link->cgrp;
426

427 428
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
L
Li Zefan 已提交
429

430
		/* @cgrp can't go away while we're holding css_set_lock */
T
Tejun Heo 已提交
431
		if (list_empty(&cgrp->cset_links) && notify_on_release(cgrp)) {
432
			if (taskexit)
433 434
				set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
435
		}
436 437

		kfree(link);
438
	}
439 440

	write_unlock(&css_set_lock);
441
	kfree_rcu(cset, rcu_head);
442 443
}

444 445 446
/*
 * refcounted get/put for css_set objects
 */
447
static inline void get_css_set(struct css_set *cset)
448
{
449
	atomic_inc(&cset->refcount);
450 451
}

452
static inline void put_css_set(struct css_set *cset)
453
{
454
	__put_css_set(cset, 0);
455 456
}

457
static inline void put_css_set_taskexit(struct css_set *cset)
458
{
459
	__put_css_set(cset, 1);
460 461
}

462
/**
463
 * compare_css_sets - helper function for find_existing_css_set().
464 465
 * @cset: candidate css_set being tested
 * @old_cset: existing css_set for a task
466 467 468
 * @new_cgrp: cgroup that's being entered by the task
 * @template: desired set of css pointers in css_set (pre-calculated)
 *
L
Li Zefan 已提交
469
 * Returns true if "cset" matches "old_cset" except for the hierarchy
470 471
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
472 473
static bool compare_css_sets(struct css_set *cset,
			     struct css_set *old_cset,
474 475 476 477 478
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

479
	if (memcmp(template, cset->subsys, sizeof(cset->subsys))) {
480 481 482 483 484 485 486 487 488 489 490 491 492
		/* 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.
	 */

493 494
	l1 = &cset->cgrp_links;
	l2 = &old_cset->cgrp_links;
495
	while (1) {
496
		struct cgrp_cset_link *link1, *link2;
497
		struct cgroup *cgrp1, *cgrp2;
498 499 500 501

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
502 503
		if (l1 == &cset->cgrp_links) {
			BUG_ON(l2 != &old_cset->cgrp_links);
504 505
			break;
		} else {
506
			BUG_ON(l2 == &old_cset->cgrp_links);
507 508
		}
		/* Locate the cgroups associated with these links. */
509 510 511 512
		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;
513
		/* Hierarchies should be linked in the same order. */
514
		BUG_ON(cgrp1->root != cgrp2->root);
515 516 517 518 519 520 521 522

		/*
		 * 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.
		 */
523 524
		if (cgrp1->root == new_cgrp->root) {
			if (cgrp1 != new_cgrp)
525 526
				return false;
		} else {
527
			if (cgrp1 != cgrp2)
528 529 530 531 532 533
				return false;
		}
	}
	return true;
}

534 535 536 537 538
/**
 * 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
539
 */
540 541 542
static struct css_set *find_existing_css_set(struct css_set *old_cset,
					struct cgroup *cgrp,
					struct cgroup_subsys_state *template[])
543
{
544
	struct cgroupfs_root *root = cgrp->root;
545
	struct cgroup_subsys *ss;
546
	struct css_set *cset;
547
	unsigned long key;
548
	int i;
549

B
Ben Blum 已提交
550 551 552 553 554
	/*
	 * 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.
	 */
555
	for_each_subsys(ss, i) {
556
		if (root->subsys_mask & (1UL << i)) {
557 558 559
			/* Subsystem is in this hierarchy. So we want
			 * the subsystem state from the new
			 * cgroup */
560
			template[i] = cgrp->subsys[i];
561 562 563
		} else {
			/* Subsystem is not in this hierarchy, so we
			 * don't want to change the subsystem state */
564
			template[i] = old_cset->subsys[i];
565 566 567
		}
	}

568
	key = css_set_hash(template);
569 570
	hash_for_each_possible(css_set_table, cset, hlist, key) {
		if (!compare_css_sets(cset, old_cset, cgrp, template))
571 572 573
			continue;

		/* This css_set matches what we need */
574
		return cset;
575
	}
576 577 578 579 580

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

581
static void free_cgrp_cset_links(struct list_head *links_to_free)
582
{
583
	struct cgrp_cset_link *link, *tmp_link;
584

585 586
	list_for_each_entry_safe(link, tmp_link, links_to_free, cset_link) {
		list_del(&link->cset_link);
587 588 589 590
		kfree(link);
	}
}

591 592 593 594 595 596 597
/**
 * 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.
598
 */
599
static int allocate_cgrp_cset_links(int count, struct list_head *tmp_links)
600
{
601
	struct cgrp_cset_link *link;
602
	int i;
603 604 605

	INIT_LIST_HEAD(tmp_links);

606
	for (i = 0; i < count; i++) {
607
		link = kzalloc(sizeof(*link), GFP_KERNEL);
608
		if (!link) {
609
			free_cgrp_cset_links(tmp_links);
610 611
			return -ENOMEM;
		}
612
		list_add(&link->cset_link, tmp_links);
613 614 615 616
	}
	return 0;
}

617 618
/**
 * link_css_set - a helper function to link a css_set to a cgroup
619
 * @tmp_links: cgrp_cset_link objects allocated by allocate_cgrp_cset_links()
620
 * @cset: the css_set to be linked
621 622
 * @cgrp: the destination cgroup
 */
623 624
static void link_css_set(struct list_head *tmp_links, struct css_set *cset,
			 struct cgroup *cgrp)
625
{
626
	struct cgrp_cset_link *link;
627

628 629 630
	BUG_ON(list_empty(tmp_links));
	link = list_first_entry(tmp_links, struct cgrp_cset_link, cset_link);
	link->cset = cset;
631
	link->cgrp = cgrp;
632
	list_move(&link->cset_link, &cgrp->cset_links);
633 634 635 636
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
637
	list_add_tail(&link->cgrp_link, &cset->cgrp_links);
638 639
}

640 641 642 643 644 645 646
/**
 * 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.
647
 */
648 649
static struct css_set *find_css_set(struct css_set *old_cset,
				    struct cgroup *cgrp)
650
{
651
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT] = { };
652
	struct css_set *cset;
653 654
	struct list_head tmp_links;
	struct cgrp_cset_link *link;
655
	unsigned long key;
656

657 658
	lockdep_assert_held(&cgroup_mutex);

659 660
	/* First see if we already have a cgroup group that matches
	 * the desired set */
661
	read_lock(&css_set_lock);
662 663 664
	cset = find_existing_css_set(old_cset, cgrp, template);
	if (cset)
		get_css_set(cset);
665
	read_unlock(&css_set_lock);
666

667 668
	if (cset)
		return cset;
669

670
	cset = kzalloc(sizeof(*cset), GFP_KERNEL);
671
	if (!cset)
672 673
		return NULL;

674
	/* Allocate all the cgrp_cset_link objects that we'll need */
675
	if (allocate_cgrp_cset_links(cgroup_root_count, &tmp_links) < 0) {
676
		kfree(cset);
677 678 679
		return NULL;
	}

680
	atomic_set(&cset->refcount, 1);
681
	INIT_LIST_HEAD(&cset->cgrp_links);
682 683
	INIT_LIST_HEAD(&cset->tasks);
	INIT_HLIST_NODE(&cset->hlist);
684 685 686

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
687
	memcpy(cset->subsys, template, sizeof(cset->subsys));
688 689 690

	write_lock(&css_set_lock);
	/* Add reference counts and links from the new css_set. */
691
	list_for_each_entry(link, &old_cset->cgrp_links, cgrp_link) {
692
		struct cgroup *c = link->cgrp;
693

694 695
		if (c->root == cgrp->root)
			c = cgrp;
696
		link_css_set(&tmp_links, cset, c);
697
	}
698

699
	BUG_ON(!list_empty(&tmp_links));
700 701

	css_set_count++;
702 703

	/* Add this cgroup group to the hash table */
704 705
	key = css_set_hash(cset->subsys);
	hash_add(css_set_table, &cset->hlist, key);
706

707 708
	write_unlock(&css_set_lock);

709
	return cset;
710 711
}

712 713 714 715 716 717 718
/*
 * 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)
{
719
	struct css_set *cset;
720 721 722 723 724 725 726 727 728
	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.
	 */
729
	cset = task_css_set(task);
730
	if (cset == &init_css_set) {
731 732
		res = &root->top_cgroup;
	} else {
733 734 735
		struct cgrp_cset_link *link;

		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
736
			struct cgroup *c = link->cgrp;
737

738 739 740 741 742 743 744 745 746 747 748
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
	read_unlock(&css_set_lock);
	BUG_ON(!res);
	return res;
}

749 750 751 752 753 754 755 756 757 758
/*
 * 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
759
 * cgroup_attach_task() can increment it again.  Because a count of zero
760 761 762 763 764 765 766 767 768 769 770 771 772
 * 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
773 774
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
775 776 777 778 779 780 781 782 783 784 785
 *
 * 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
786
 * cgroup_attach_task(), which overwrites one task's cgroup pointer with
L
Li Zefan 已提交
787
 * another.  It does so using cgroup_mutex, however there are
788 789 790
 * 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
791
 * in cgroup_attach_task(), modifying a task's cgroup pointer we use
792 793 794 795
 * 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
796
 * update of a tasks cgroup pointer by cgroup_attach_task()
797 798 799 800 801 802 803 804 805
 */

/*
 * 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.
 */

806
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
A
Al Viro 已提交
807
static struct dentry *cgroup_lookup(struct inode *, struct dentry *, unsigned int);
808
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
809
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned long subsys_mask);
810
static const struct inode_operations cgroup_dir_inode_operations;
811
static const struct file_operations proc_cgroupstats_operations;
812 813

static struct backing_dev_info cgroup_backing_dev_info = {
814
	.name		= "cgroup",
815
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
816
};
817

K
KAMEZAWA Hiroyuki 已提交
818 819 820
static int alloc_css_id(struct cgroup_subsys *ss,
			struct cgroup *parent, struct cgroup *child);

A
Al Viro 已提交
821
static struct inode *cgroup_new_inode(umode_t mode, struct super_block *sb)
822 823 824 825
{
	struct inode *inode = new_inode(sb);

	if (inode) {
826
		inode->i_ino = get_next_ino();
827
		inode->i_mode = mode;
828 829
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
830 831 832 833 834 835
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inode->i_mapping->backing_dev_info = &cgroup_backing_dev_info;
	}
	return inode;
}

836 837 838 839 840 841 842 843 844 845 846
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;
}

847 848
static void cgroup_free_fn(struct work_struct *work)
{
849
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
850 851 852 853 854 855
	struct cgroup_subsys *ss;

	mutex_lock(&cgroup_mutex);
	/*
	 * Release the subsystem state objects.
	 */
856 857 858 859 860
	for_each_root_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

		ss->css_free(css);
	}
861 862 863 864

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

865 866 867 868 869 870 871
	/*
	 * 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);

872 873
	/*
	 * Drop the active superblock reference that we took when we
874 875
	 * created the cgroup. This will free cgrp->root, if we are
	 * holding the last reference to @sb.
876 877 878 879 880 881 882 883 884 885 886
	 */
	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);

887
	kfree(rcu_dereference_raw(cgrp->name));
888 889 890 891 892 893 894
	kfree(cgrp);
}

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

895 896
	INIT_WORK(&cgrp->destroy_work, cgroup_free_fn);
	schedule_work(&cgrp->destroy_work);
897 898
}

899 900 901 902
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)) {
903
		struct cgroup *cgrp = dentry->d_fsdata;
904

905
		BUG_ON(!(cgroup_is_dead(cgrp)));
906
		call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
T
Tejun Heo 已提交
907 908 909 910 911 912 913
	} 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
914
		simple_xattrs_free(&cfe->xattrs);
T
Tejun Heo 已提交
915
		kfree(cfe);
916 917 918 919
	}
	iput(inode);
}

920 921 922 923 924
static int cgroup_delete(const struct dentry *d)
{
	return 1;
}

925 926 927 928 929 930 931 932 933
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);
}

934
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
935 936 937 938 939 940
{
	struct cfent *cfe;

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

941 942 943 944
	/*
	 * If we're doing cleanup due to failure of cgroup_create(),
	 * the corresponding @cfe may not exist.
	 */
T
Tejun Heo 已提交
945 946 947 948 949 950 951 952
	list_for_each_entry(cfe, &cgrp->files, node) {
		struct dentry *d = cfe->dentry;

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

		dget(d);
		d_delete(d);
953
		simple_unlink(cgrp->dentry->d_inode, d);
T
Tejun Heo 已提交
954 955 956
		list_del_init(&cfe->node);
		dput(d);

957
		break;
958
	}
T
Tejun Heo 已提交
959 960
}

961
/**
962
 * cgroup_clear_dir - remove subsys files in a cgroup directory
963
 * @cgrp: target cgroup
964 965
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
966
static void cgroup_clear_dir(struct cgroup *cgrp, unsigned long subsys_mask)
T
Tejun Heo 已提交
967
{
968
	struct cgroup_subsys *ss;
969
	int i;
T
Tejun Heo 已提交
970

971
	for_each_subsys(ss, i) {
972
		struct cftype_set *set;
973 974

		if (!test_bit(i, &subsys_mask))
975 976
			continue;
		list_for_each_entry(set, &ss->cftsets, node)
977
			cgroup_addrm_files(cgrp, NULL, set->cfts, false);
978
	}
979 980 981 982 983 984 985
}

/*
 * NOTE : the dentry must have been dget()'ed
 */
static void cgroup_d_remove_dir(struct dentry *dentry)
{
N
Nick Piggin 已提交
986
	struct dentry *parent;
987

N
Nick Piggin 已提交
988 989
	parent = dentry->d_parent;
	spin_lock(&parent->d_lock);
990
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
991
	list_del_init(&dentry->d_u.d_child);
N
Nick Piggin 已提交
992 993
	spin_unlock(&dentry->d_lock);
	spin_unlock(&parent->d_lock);
994 995 996
	remove_dir(dentry);
}

B
Ben Blum 已提交
997
/*
B
Ben Blum 已提交
998 999 1000
 * 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
1001
 */
1002
static int rebind_subsystems(struct cgroupfs_root *root,
1003
			     unsigned long added_mask, unsigned removed_mask)
1004
{
1005
	struct cgroup *cgrp = &root->top_cgroup;
1006
	struct cgroup_subsys *ss;
1007
	unsigned long pinned = 0;
1008
	int i, ret;
1009

B
Ben Blum 已提交
1010
	BUG_ON(!mutex_is_locked(&cgroup_mutex));
T
Tejun Heo 已提交
1011
	BUG_ON(!mutex_is_locked(&cgroup_root_mutex));
B
Ben Blum 已提交
1012

1013
	/* Check that any added subsystems are currently free */
1014
	for_each_subsys(ss, i) {
1015
		if (!(added_mask & (1 << i)))
1016
			continue;
1017

1018
		/* is the subsystem mounted elsewhere? */
1019
		if (ss->root != &cgroup_dummy_root) {
1020 1021 1022 1023 1024 1025 1026 1027
			ret = -EBUSY;
			goto out_put;
		}

		/* pin the module */
		if (!try_module_get(ss->module)) {
			ret = -ENOENT;
			goto out_put;
1028
		}
1029 1030 1031 1032 1033 1034 1035
		pinned |= 1 << i;
	}

	/* subsys could be missing if unloaded between parsing and here */
	if (added_mask != pinned) {
		ret = -ENOENT;
		goto out_put;
1036 1037
	}

1038 1039
	ret = cgroup_populate_dir(cgrp, added_mask);
	if (ret)
1040
		goto out_put;
1041 1042 1043 1044 1045 1046 1047

	/*
	 * Nothing can fail from this point on.  Remove files for the
	 * removed subsystems and rebind each subsystem.
	 */
	cgroup_clear_dir(cgrp, removed_mask);

1048
	for_each_subsys(ss, i) {
1049
		unsigned long bit = 1UL << i;
1050

1051
		if (bit & added_mask) {
1052
			/* We're binding this subsystem to this hierarchy */
1053
			BUG_ON(cgrp->subsys[i]);
1054 1055
			BUG_ON(!cgroup_dummy_top->subsys[i]);
			BUG_ON(cgroup_dummy_top->subsys[i]->cgroup != cgroup_dummy_top);
1056

1057
			cgrp->subsys[i] = cgroup_dummy_top->subsys[i];
1058
			cgrp->subsys[i]->cgroup = cgrp;
1059
			list_move(&ss->sibling, &root->subsys_list);
1060
			ss->root = root;
1061
			if (ss->bind)
1062
				ss->bind(cgrp->subsys[i]);
1063

B
Ben Blum 已提交
1064
			/* refcount was already taken, and we're keeping it */
1065
			root->subsys_mask |= bit;
1066
		} else if (bit & removed_mask) {
1067
			/* We're removing this subsystem */
1068
			BUG_ON(cgrp->subsys[i] != cgroup_dummy_top->subsys[i]);
1069
			BUG_ON(cgrp->subsys[i]->cgroup != cgrp);
1070

1071
			if (ss->bind)
1072
				ss->bind(cgroup_dummy_top->subsys[i]);
1073
			cgroup_dummy_top->subsys[i]->cgroup = cgroup_dummy_top;
1074
			cgrp->subsys[i] = NULL;
1075 1076
			cgroup_subsys[i]->root = &cgroup_dummy_root;
			list_move(&ss->sibling, &cgroup_dummy_root.subsys_list);
1077

B
Ben Blum 已提交
1078 1079
			/* subsystem is now free - drop reference on module */
			module_put(ss->module);
1080
			root->subsys_mask &= ~bit;
1081 1082 1083
		}
	}

1084 1085 1086 1087 1088 1089
	/*
	 * Mark @root has finished binding subsystems.  @root->subsys_mask
	 * now matches the bound subsystems.
	 */
	root->flags |= CGRP_ROOT_SUBSYS_BOUND;

1090
	return 0;
1091 1092 1093 1094 1095 1096

out_put:
	for_each_subsys(ss, i)
		if (pinned & (1 << i))
			module_put(ss->module);
	return ret;
1097 1098
}

1099
static int cgroup_show_options(struct seq_file *seq, struct dentry *dentry)
1100
{
1101
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
1102 1103
	struct cgroup_subsys *ss;

T
Tejun Heo 已提交
1104
	mutex_lock(&cgroup_root_mutex);
1105
	for_each_root_subsys(root, ss)
1106
		seq_printf(seq, ",%s", ss->name);
1107 1108
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR)
		seq_puts(seq, ",sane_behavior");
1109
	if (root->flags & CGRP_ROOT_NOPREFIX)
1110
		seq_puts(seq, ",noprefix");
1111
	if (root->flags & CGRP_ROOT_XATTR)
A
Aristeu Rozanski 已提交
1112
		seq_puts(seq, ",xattr");
1113 1114
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1115
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags))
1116
		seq_puts(seq, ",clone_children");
1117 1118
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
T
Tejun Heo 已提交
1119
	mutex_unlock(&cgroup_root_mutex);
1120 1121 1122 1123
	return 0;
}

struct cgroup_sb_opts {
1124
	unsigned long subsys_mask;
1125
	unsigned long flags;
1126
	char *release_agent;
1127
	bool cpuset_clone_children;
1128
	char *name;
1129 1130
	/* User explicitly requested empty subsystem */
	bool none;
1131 1132

	struct cgroupfs_root *new_root;
1133

1134 1135
};

B
Ben Blum 已提交
1136
/*
1137 1138 1139 1140
 * 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
1141
 */
B
Ben Blum 已提交
1142
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1143
{
1144 1145
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1146
	unsigned long mask = (unsigned long)-1;
1147 1148
	struct cgroup_subsys *ss;
	int i;
1149

B
Ben Blum 已提交
1150 1151
	BUG_ON(!mutex_is_locked(&cgroup_mutex));

1152 1153 1154
#ifdef CONFIG_CPUSETS
	mask = ~(1UL << cpuset_subsys_id);
#endif
1155

1156
	memset(opts, 0, sizeof(*opts));
1157 1158 1159 1160

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1161
		if (!strcmp(token, "none")) {
1162 1163
			/* Explicitly have no subsystems */
			opts->none = true;
1164 1165 1166 1167 1168 1169 1170 1171 1172
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
1173 1174 1175 1176
		if (!strcmp(token, "__DEVEL__sane_behavior")) {
			opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
			continue;
		}
1177
		if (!strcmp(token, "noprefix")) {
1178
			opts->flags |= CGRP_ROOT_NOPREFIX;
1179 1180 1181
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1182
			opts->cpuset_clone_children = true;
1183 1184
			continue;
		}
A
Aristeu Rozanski 已提交
1185
		if (!strcmp(token, "xattr")) {
1186
			opts->flags |= CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
1187 1188
			continue;
		}
1189
		if (!strncmp(token, "release_agent=", 14)) {
1190 1191 1192
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1193
			opts->release_agent =
1194
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1195 1196
			if (!opts->release_agent)
				return -ENOMEM;
1197 1198 1199
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
			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,
1217
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1218 1219 1220
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1221 1222 1223 1224

			continue;
		}

1225
		for_each_subsys(ss, i) {
1226 1227 1228 1229 1230 1231 1232 1233
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1234
			set_bit(i, &opts->subsys_mask);
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
			one_ss = true;

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

	/*
	 * If the 'all' option was specified select all the subsystems,
1245 1246
	 * otherwise if 'none', 'name=' and a subsystem name options
	 * were not specified, let's default to 'all'
1247
	 */
1248 1249 1250 1251
	if (all_ss || (!one_ss && !opts->none && !opts->name))
		for_each_subsys(ss, i)
			if (!ss->disabled)
				set_bit(i, &opts->subsys_mask);
1252

1253 1254
	/* Consistency checks */

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
	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;
		}
	}

1269 1270 1271 1272 1273
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
1274
	if ((opts->flags & CGRP_ROOT_NOPREFIX) && (opts->subsys_mask & mask))
1275 1276
		return -EINVAL;

1277 1278

	/* Can't specify "none" and some subsystems */
1279
	if (opts->subsys_mask && opts->none)
1280 1281 1282 1283 1284 1285
		return -EINVAL;

	/*
	 * We either have to specify by name or by subsystems. (So all
	 * empty hierarchies must have a name).
	 */
1286
	if (!opts->subsys_mask && !opts->name)
1287 1288 1289 1290 1291 1292 1293 1294 1295
		return -EINVAL;

	return 0;
}

static int cgroup_remount(struct super_block *sb, int *flags, char *data)
{
	int ret = 0;
	struct cgroupfs_root *root = sb->s_fs_info;
1296
	struct cgroup *cgrp = &root->top_cgroup;
1297
	struct cgroup_sb_opts opts;
1298
	unsigned long added_mask, removed_mask;
1299

1300 1301 1302 1303 1304
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR) {
		pr_err("cgroup: sane_behavior: remount is not allowed\n");
		return -EINVAL;
	}

1305
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1306
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1307
	mutex_lock(&cgroup_root_mutex);
1308 1309 1310 1311 1312 1313

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

1314
	if (opts.subsys_mask != root->subsys_mask || opts.release_agent)
1315 1316 1317
		pr_warning("cgroup: option changes via remount are deprecated (pid=%d comm=%s)\n",
			   task_tgid_nr(current), current->comm);

1318 1319
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1320

B
Ben Blum 已提交
1321
	/* Don't allow flags or name to change at remount */
1322
	if (((opts.flags ^ root->flags) & CGRP_ROOT_OPTION_MASK) ||
B
Ben Blum 已提交
1323
	    (opts.name && strcmp(opts.name, root->name))) {
1324 1325 1326
		pr_err("cgroup: option or name mismatch, new: 0x%lx \"%s\", old: 0x%lx \"%s\"\n",
		       opts.flags & CGRP_ROOT_OPTION_MASK, opts.name ?: "",
		       root->flags & CGRP_ROOT_OPTION_MASK, root->name);
1327 1328 1329 1330
		ret = -EINVAL;
		goto out_unlock;
	}

1331 1332 1333 1334 1335 1336
	/* remounting is not allowed for populated hierarchies */
	if (root->number_of_cgroups > 1) {
		ret = -EBUSY;
		goto out_unlock;
	}

1337
	ret = rebind_subsystems(root, added_mask, removed_mask);
1338
	if (ret)
1339
		goto out_unlock;
1340

1341 1342
	if (opts.release_agent)
		strcpy(root->release_agent_path, opts.release_agent);
1343
 out_unlock:
1344
	kfree(opts.release_agent);
1345
	kfree(opts.name);
T
Tejun Heo 已提交
1346
	mutex_unlock(&cgroup_root_mutex);
1347
	mutex_unlock(&cgroup_mutex);
1348
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1349 1350 1351
	return ret;
}

1352
static const struct super_operations cgroup_ops = {
1353 1354 1355 1356 1357 1358
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
	.show_options = cgroup_show_options,
	.remount_fs = cgroup_remount,
};

1359 1360 1361 1362
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
T
Tejun Heo 已提交
1363
	INIT_LIST_HEAD(&cgrp->files);
1364
	INIT_LIST_HEAD(&cgrp->cset_links);
1365
	INIT_LIST_HEAD(&cgrp->release_list);
1366 1367
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
1368 1369
	INIT_LIST_HEAD(&cgrp->event_list);
	spin_lock_init(&cgrp->event_list_lock);
A
Aristeu Rozanski 已提交
1370
	simple_xattrs_init(&cgrp->xattrs);
1371
}
1372

1373 1374
static void init_cgroup_root(struct cgroupfs_root *root)
{
1375
	struct cgroup *cgrp = &root->top_cgroup;
1376

1377 1378 1379
	INIT_LIST_HEAD(&root->subsys_list);
	INIT_LIST_HEAD(&root->root_list);
	root->number_of_cgroups = 1;
1380
	cgrp->root = root;
1381
	RCU_INIT_POINTER(cgrp->name, &root_cgroup_name);
1382
	init_cgroup_housekeeping(cgrp);
1383
	idr_init(&root->cgroup_idr);
1384 1385
}

1386
static int cgroup_init_root_id(struct cgroupfs_root *root, int start, int end)
1387
{
1388
	int id;
1389

T
Tejun Heo 已提交
1390 1391 1392
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1393 1394
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, start, end,
			      GFP_KERNEL);
1395 1396 1397 1398
	if (id < 0)
		return id;

	root->hierarchy_id = id;
1399 1400 1401 1402 1403
	return 0;
}

static void cgroup_exit_root_id(struct cgroupfs_root *root)
{
T
Tejun Heo 已提交
1404 1405 1406
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1407
	if (root->hierarchy_id) {
1408
		idr_remove(&cgroup_hierarchy_idr, root->hierarchy_id);
1409 1410
		root->hierarchy_id = 0;
	}
1411 1412
}

1413 1414
static int cgroup_test_super(struct super_block *sb, void *data)
{
1415
	struct cgroup_sb_opts *opts = data;
1416 1417
	struct cgroupfs_root *root = sb->s_fs_info;

1418 1419 1420
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1421

1422 1423 1424 1425
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
1426 1427
	if ((opts->subsys_mask || opts->none)
	    && (opts->subsys_mask != root->subsys_mask))
1428 1429 1430 1431 1432
		return 0;

	return 1;
}

1433 1434 1435 1436
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1437
	if (!opts->subsys_mask && !opts->none)
1438 1439 1440 1441 1442 1443 1444
		return NULL;

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

	init_cgroup_root(root);
1445

1446 1447 1448 1449 1450 1451 1452 1453
	/*
	 * We need to set @root->subsys_mask now so that @root can be
	 * matched by cgroup_test_super() before it finishes
	 * initialization; otherwise, competing mounts with the same
	 * options may try to bind the same subsystems instead of waiting
	 * for the first one leading to unexpected mount errors.
	 * SUBSYS_BOUND will be set once actual binding is complete.
	 */
1454
	root->subsys_mask = opts->subsys_mask;
1455 1456 1457 1458 1459
	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1460 1461
	if (opts->cpuset_clone_children)
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags);
1462 1463 1464
	return root;
}

1465
static void cgroup_free_root(struct cgroupfs_root *root)
1466
{
1467 1468 1469
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);
1470

1471
		idr_destroy(&root->cgroup_idr);
1472 1473
		kfree(root);
	}
1474 1475
}

1476 1477 1478
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1479 1480 1481 1482 1483 1484
	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;

1485
	BUG_ON(!opts->subsys_mask && !opts->none);
1486 1487 1488 1489 1490

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

1491 1492
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503

	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
1504 1505
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1506
		.d_delete = cgroup_delete,
A
Al Viro 已提交
1507 1508
	};

1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
	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);
1519 1520
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1521
		return -ENOMEM;
A
Al Viro 已提交
1522 1523
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1524 1525 1526
	return 0;
}

A
Al Viro 已提交
1527
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1528
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1529
			 void *data)
1530 1531
{
	struct cgroup_sb_opts opts;
1532
	struct cgroupfs_root *root;
1533 1534
	int ret = 0;
	struct super_block *sb;
1535
	struct cgroupfs_root *new_root;
1536
	struct list_head tmp_links;
T
Tejun Heo 已提交
1537
	struct inode *inode;
1538
	const struct cred *cred;
1539 1540

	/* First find the desired set of subsystems */
B
Ben Blum 已提交
1541
	mutex_lock(&cgroup_mutex);
1542
	ret = parse_cgroupfs_options(data, &opts);
B
Ben Blum 已提交
1543
	mutex_unlock(&cgroup_mutex);
1544 1545
	if (ret)
		goto out_err;
1546

1547 1548 1549 1550 1551 1552 1553
	/*
	 * 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);
1554
		goto out_err;
1555
	}
1556
	opts.new_root = new_root;
1557

1558
	/* Locate an existing or new sb for this hierarchy */
D
David Howells 已提交
1559
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1560
	if (IS_ERR(sb)) {
1561
		ret = PTR_ERR(sb);
1562
		cgroup_free_root(opts.new_root);
1563
		goto out_err;
1564 1565
	}

1566 1567 1568 1569
	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 */
1570
		struct cgroup *root_cgrp = &root->top_cgroup;
1571
		struct cgroupfs_root *existing_root;
1572
		int i;
1573
		struct css_set *cset;
1574 1575 1576 1577 1578 1579

		BUG_ON(sb->s_root != NULL);

		ret = cgroup_get_rootdir(sb);
		if (ret)
			goto drop_new_super;
1580
		inode = sb->s_root->d_inode;
1581

1582
		mutex_lock(&inode->i_mutex);
1583
		mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1584
		mutex_lock(&cgroup_root_mutex);
1585

1586 1587 1588 1589 1590
		root_cgrp->id = idr_alloc(&root->cgroup_idr, root_cgrp,
					   0, 1, GFP_KERNEL);
		if (root_cgrp->id < 0)
			goto unlock_drop;

T
Tejun Heo 已提交
1591 1592 1593 1594 1595 1596
		/* 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;
1597

1598 1599 1600 1601 1602 1603 1604
		/*
		 * 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
		 */
1605
		ret = allocate_cgrp_cset_links(css_set_count, &tmp_links);
T
Tejun Heo 已提交
1606 1607
		if (ret)
			goto unlock_drop;
1608

1609 1610
		/* ID 0 is reserved for dummy root, 1 for unified hierarchy */
		ret = cgroup_init_root_id(root, 2, 0);
1611 1612 1613
		if (ret)
			goto unlock_drop;

1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
		sb->s_root->d_fsdata = root_cgrp;
		root_cgrp->dentry = sb->s_root;

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

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

1630
		ret = rebind_subsystems(root, root->subsys_mask, 0);
1631 1632 1633 1634 1635
		if (ret)
			goto rm_base_files;

		revert_creds(cred);

B
Ben Blum 已提交
1636 1637 1638 1639 1640
		/*
		 * There must be no failure case after here, since rebinding
		 * takes care of subsystems' refcounts, which are explicitly
		 * dropped in the failure exit path.
		 */
1641

1642 1643
		list_add(&root->root_list, &cgroup_roots);
		cgroup_root_count++;
1644

1645 1646 1647
		/* Link the top cgroup in this hierarchy into all
		 * the css_set objects */
		write_lock(&css_set_lock);
1648
		hash_for_each(css_set_table, i, cset, hlist)
1649
			link_css_set(&tmp_links, cset, root_cgrp);
1650 1651
		write_unlock(&css_set_lock);

1652
		free_cgrp_cset_links(&tmp_links);
1653

1654
		BUG_ON(!list_empty(&root_cgrp->children));
1655 1656
		BUG_ON(root->number_of_cgroups != 1);

T
Tejun Heo 已提交
1657
		mutex_unlock(&cgroup_root_mutex);
1658
		mutex_unlock(&cgroup_mutex);
1659
		mutex_unlock(&inode->i_mutex);
1660 1661 1662 1663 1664
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1665
		cgroup_free_root(opts.new_root);
1666

1667
		if ((root->flags ^ opts.flags) & CGRP_ROOT_OPTION_MASK) {
1668 1669 1670 1671 1672 1673 1674
			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");
			}
1675
		}
1676 1677
	}

1678 1679
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1680
	return dget(sb->s_root);
1681

1682 1683 1684 1685
 rm_base_files:
	free_cgrp_cset_links(&tmp_links);
	cgroup_addrm_files(&root->top_cgroup, NULL, cgroup_base_files, false);
	revert_creds(cred);
T
Tejun Heo 已提交
1686
 unlock_drop:
1687
	cgroup_exit_root_id(root);
T
Tejun Heo 已提交
1688 1689 1690
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);
1691
 drop_new_super:
1692
	deactivate_locked_super(sb);
1693 1694 1695
 out_err:
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1696
	return ERR_PTR(ret);
1697 1698 1699 1700
}

static void cgroup_kill_sb(struct super_block *sb) {
	struct cgroupfs_root *root = sb->s_fs_info;
1701
	struct cgroup *cgrp = &root->top_cgroup;
1702
	struct cgrp_cset_link *link, *tmp_link;
1703 1704 1705 1706 1707
	int ret;

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1708
	BUG_ON(!list_empty(&cgrp->children));
1709

1710
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1711
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1712
	mutex_lock(&cgroup_root_mutex);
1713 1714

	/* Rebind all subsystems back to the default hierarchy */
1715 1716 1717 1718 1719
	if (root->flags & CGRP_ROOT_SUBSYS_BOUND) {
		ret = rebind_subsystems(root, 0, root->subsys_mask);
		/* Shouldn't be able to fail ... */
		BUG_ON(ret);
	}
1720

1721
	/*
1722
	 * Release all the links from cset_links to this hierarchy's
1723 1724 1725
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1726

1727 1728 1729
	list_for_each_entry_safe(link, tmp_link, &cgrp->cset_links, cset_link) {
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
1730 1731 1732 1733
		kfree(link);
	}
	write_unlock(&css_set_lock);

1734 1735
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
1736
		cgroup_root_count--;
1737
	}
1738

1739 1740
	cgroup_exit_root_id(root);

T
Tejun Heo 已提交
1741
	mutex_unlock(&cgroup_root_mutex);
1742
	mutex_unlock(&cgroup_mutex);
1743
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1744

A
Aristeu Rozanski 已提交
1745 1746
	simple_xattrs_free(&cgrp->xattrs);

1747
	kill_litter_super(sb);
1748
	cgroup_free_root(root);
1749 1750 1751 1752
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1753
	.mount = cgroup_mount,
1754 1755 1756
	.kill_sb = cgroup_kill_sb,
};

1757 1758
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1759 1760 1761 1762 1763 1764
/**
 * 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
 *
1765 1766 1767 1768 1769 1770
 * 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.
1771
 */
1772
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1773
{
1774
	int ret = -ENAMETOOLONG;
1775
	char *start;
1776

1777 1778 1779
	if (!cgrp->parent) {
		if (strlcpy(buf, "/", buflen) >= buflen)
			return -ENAMETOOLONG;
1780 1781 1782
		return 0;
	}

1783 1784
	start = buf + buflen - 1;
	*start = '\0';
1785

1786
	rcu_read_lock();
1787
	do {
1788 1789 1790 1791
		const char *name = cgroup_name(cgrp);
		int len;

		len = strlen(name);
1792
		if ((start -= len) < buf)
1793 1794
			goto out;
		memcpy(start, name, len);
1795

1796
		if (--start < buf)
1797
			goto out;
1798
		*start = '/';
1799 1800

		cgrp = cgrp->parent;
1801
	} while (cgrp->parent);
1802
	ret = 0;
1803
	memmove(buf, start, buf + buflen - start);
1804 1805 1806
out:
	rcu_read_unlock();
	return ret;
1807
}
B
Ben Blum 已提交
1808
EXPORT_SYMBOL_GPL(cgroup_path);
1809

1810
/**
1811
 * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
1812 1813 1814 1815
 * @task: target task
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1816 1817 1818 1819 1820 1821
 * Determine @task's cgroup on the first (the one with the lowest non-zero
 * hierarchy_id) cgroup hierarchy and copy its path into @buf.  This
 * function grabs cgroup_mutex and shouldn't be used inside locks used by
 * cgroup controller callbacks.
 *
 * Returns 0 on success, fails with -%ENAMETOOLONG if @buflen is too short.
1822
 */
1823
int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
1824 1825
{
	struct cgroupfs_root *root;
1826 1827 1828 1829 1830
	struct cgroup *cgrp;
	int hierarchy_id = 1, ret = 0;

	if (buflen < 2)
		return -ENAMETOOLONG;
1831 1832 1833

	mutex_lock(&cgroup_mutex);

1834 1835
	root = idr_get_next(&cgroup_hierarchy_idr, &hierarchy_id);

1836 1837 1838
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
		ret = cgroup_path(cgrp, buf, buflen);
1839 1840 1841
	} else {
		/* if no hierarchy exists, everyone is in "/" */
		memcpy(buf, "/", 2);
1842 1843 1844 1845 1846
	}

	mutex_unlock(&cgroup_mutex);
	return ret;
}
1847
EXPORT_SYMBOL_GPL(task_cgroup_path);
1848

1849 1850 1851
/*
 * Control Group taskset
 */
1852 1853 1854
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
L
Li Zefan 已提交
1855
	struct css_set		*cset;
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 = task_css_set(tsk);
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) {
2055 2056
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

B
Ben Blum 已提交
2057
		if (ss->can_attach) {
2058
			retval = ss->can_attach(css, &tset);
B
Ben Blum 已提交
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
			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++) {
2071 2072
		struct css_set *old_cset;

2073
		tc = flex_array_get(group, i);
2074
		old_cset = task_css_set(tc->task);
L
Li Zefan 已提交
2075 2076
		tc->cset = find_css_set(old_cset, cgrp);
		if (!tc->cset) {
2077 2078
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2079 2080 2081 2082
		}
	}

	/*
2083 2084 2085
	 * 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
2086 2087
	 */
	for (i = 0; i < group_size; i++) {
2088
		tc = flex_array_get(group, i);
L
Li Zefan 已提交
2089
		cgroup_task_migrate(tc->cgrp, tc->task, tc->cset);
B
Ben Blum 已提交
2090 2091 2092 2093
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2094
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2095
	 */
2096
	for_each_root_subsys(root, ss) {
2097 2098
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

B
Ben Blum 已提交
2099
		if (ss->attach)
2100
			ss->attach(css, &tset);
B
Ben Blum 已提交
2101 2102 2103 2104 2105 2106
	}

	/*
	 * step 5: success! and cleanup
	 */
	retval = 0;
2107 2108 2109 2110
out_put_css_set_refs:
	if (retval) {
		for (i = 0; i < group_size; i++) {
			tc = flex_array_get(group, i);
L
Li Zefan 已提交
2111
			if (!tc->cset)
2112
				break;
L
Li Zefan 已提交
2113
			put_css_set(tc->cset);
2114
		}
B
Ben Blum 已提交
2115 2116 2117
	}
out_cancel_attach:
	if (retval) {
2118
		for_each_root_subsys(root, ss) {
2119 2120
			struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

2121
			if (ss == failed_ss)
B
Ben Blum 已提交
2122 2123
				break;
			if (ss->cancel_attach)
2124
				ss->cancel_attach(css, &tset);
B
Ben Blum 已提交
2125 2126 2127
		}
	}
out_free_group_list:
2128
	flex_array_free(group);
B
Ben Blum 已提交
2129 2130 2131 2132 2133
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2134 2135
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2136
 */
B
Ben Blum 已提交
2137
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2138 2139
{
	struct task_struct *tsk;
2140
	const struct cred *cred = current_cred(), *tcred;
2141 2142
	int ret;

B
Ben Blum 已提交
2143 2144 2145
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2146 2147
retry_find_task:
	rcu_read_lock();
2148
	if (pid) {
2149
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2150 2151
		if (!tsk) {
			rcu_read_unlock();
2152 2153
			ret= -ESRCH;
			goto out_unlock_cgroup;
2154
		}
B
Ben Blum 已提交
2155 2156 2157 2158
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2159
		tcred = __task_cred(tsk);
2160 2161 2162
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2163
			rcu_read_unlock();
2164 2165
			ret = -EACCES;
			goto out_unlock_cgroup;
2166
		}
2167 2168
	} else
		tsk = current;
2169 2170

	if (threadgroup)
2171
		tsk = tsk->group_leader;
2172 2173

	/*
2174
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2175 2176 2177
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2178
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2179 2180 2181 2182 2183
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200
	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;
		}
2201 2202 2203 2204
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2205 2206
	threadgroup_unlock(tsk);

2207
	put_task_struct(tsk);
2208
out_unlock_cgroup:
T
Tejun Heo 已提交
2209
	mutex_unlock(&cgroup_mutex);
2210 2211 2212
	return ret;
}

2213 2214 2215 2216 2217 2218 2219 2220 2221 2222
/**
 * 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
2223
	mutex_lock(&cgroup_mutex);
2224
	for_each_active_root(root) {
L
Li Zefan 已提交
2225
		struct cgroup *from_cgrp = task_cgroup_from_root(from, root);
2226

L
Li Zefan 已提交
2227
		retval = cgroup_attach_task(from_cgrp, tsk, false);
2228 2229 2230
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2231
	mutex_unlock(&cgroup_mutex);
2232 2233 2234 2235 2236

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2237
static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2238 2239 2240 2241 2242
{
	return attach_task_by_pid(cgrp, pid, false);
}

static int cgroup_procs_write(struct cgroup *cgrp, struct cftype *cft, u64 tgid)
2243
{
2244
	return attach_task_by_pid(cgrp, tgid, true);
2245 2246
}

2247 2248 2249 2250
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);
2251 2252
	if (strlen(buffer) >= PATH_MAX)
		return -EINVAL;
2253 2254
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
T
Tejun Heo 已提交
2255
	mutex_lock(&cgroup_root_mutex);
2256
	strcpy(cgrp->root->release_agent_path, buffer);
T
Tejun Heo 已提交
2257
	mutex_unlock(&cgroup_root_mutex);
T
Tejun Heo 已提交
2258
	mutex_unlock(&cgroup_mutex);
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268
	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
2269
	mutex_unlock(&cgroup_mutex);
2270 2271 2272
	return 0;
}

2273 2274 2275 2276
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));
2277 2278 2279
	return 0;
}

2280 2281 2282
/* A buffer size big enough for numbers or short strings */
#define CGROUP_LOCAL_BUFFER_SIZE 64

2283
static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
2284 2285 2286
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
2287
{
2288
	char buffer[CGROUP_LOCAL_BUFFER_SIZE];
2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
	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 */
2300
	if (cft->write_u64) {
K
KOSAKI Motohiro 已提交
2301
		u64 val = simple_strtoull(strstrip(buffer), &end, 0);
2302 2303 2304 2305
		if (*end)
			return -EINVAL;
		retval = cft->write_u64(cgrp, cft, val);
	} else {
K
KOSAKI Motohiro 已提交
2306
		s64 val = simple_strtoll(strstrip(buffer), &end, 0);
2307 2308 2309 2310
		if (*end)
			return -EINVAL;
		retval = cft->write_s64(cgrp, cft, val);
	}
2311 2312 2313 2314 2315
	if (!retval)
		retval = nbytes;
	return retval;
}

2316 2317 2318 2319 2320
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)
{
2321
	char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
	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
2336 2337 2338 2339
	if (nbytes && copy_from_user(buffer, userbuf, nbytes)) {
		retval = -EFAULT;
		goto out;
	}
2340 2341

	buffer[nbytes] = 0;     /* nul-terminate */
K
KOSAKI Motohiro 已提交
2342
	retval = cft->write_string(cgrp, cft, strstrip(buffer));
2343 2344
	if (!retval)
		retval = nbytes;
L
Li Zefan 已提交
2345
out:
2346 2347 2348 2349 2350
	if (buffer != local_buffer)
		kfree(buffer);
	return retval;
}

2351 2352 2353 2354
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);
2355
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2356

2357
	if (cgroup_is_dead(cgrp))
2358
		return -ENODEV;
2359
	if (cft->write)
2360
		return cft->write(cgrp, cft, file, buf, nbytes, ppos);
2361 2362
	if (cft->write_u64 || cft->write_s64)
		return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
2363 2364
	if (cft->write_string)
		return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
2365 2366 2367 2368
	if (cft->trigger) {
		int ret = cft->trigger(cgrp, (unsigned int)cft->private);
		return ret ? ret : nbytes;
	}
2369
	return -EINVAL;
2370 2371
}

2372 2373 2374 2375
static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
2376
{
2377
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2378
	u64 val = cft->read_u64(cgrp, cft);
2379 2380 2381 2382 2383
	int len = sprintf(tmp, "%llu\n", (unsigned long long) val);

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

2384 2385 2386 2387 2388
static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
{
2389
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2390 2391 2392 2393 2394 2395
	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);
}

2396 2397 2398 2399
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);
2400
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2401

2402
	if (cgroup_is_dead(cgrp))
2403 2404 2405
		return -ENODEV;

	if (cft->read)
2406
		return cft->read(cgrp, cft, file, buf, nbytes, ppos);
2407 2408
	if (cft->read_u64)
		return cgroup_read_u64(cgrp, cft, file, buf, nbytes, ppos);
2409 2410
	if (cft->read_s64)
		return cgroup_read_s64(cgrp, cft, file, buf, nbytes, ppos);
2411 2412 2413
	return -EINVAL;
}

2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426
/*
 * seqfile ops/methods for returning structured data. Currently just
 * supports string->u64 maps, but can be extended in future.
 */

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)
{
2427 2428 2429 2430
	struct cfent *cfe = m->private;
	struct cftype *cft = cfe->type;
	struct cgroup *cgrp = __d_cgrp(cfe->dentry->d_parent);

2431 2432 2433 2434 2435
	if (cft->read_map) {
		struct cgroup_map_cb cb = {
			.fill = cgroup_map_add,
			.state = m,
		};
2436
		return cft->read_map(cgrp, cft, &cb);
2437
	}
2438
	return cft->read_seq_string(cgrp, cft, m);
2439 2440
}

2441
static const struct file_operations cgroup_seqfile_operations = {
2442
	.read = seq_read,
2443
	.write = cgroup_file_write,
2444
	.llseek = seq_lseek,
2445
	.release = single_release,
2446 2447
};

2448 2449 2450
static int cgroup_file_open(struct inode *inode, struct file *file)
{
	int err;
2451
	struct cfent *cfe;
2452 2453 2454 2455 2456
	struct cftype *cft;

	err = generic_file_open(inode, file);
	if (err)
		return err;
2457 2458
	cfe = __d_cfe(file->f_dentry);
	cft = cfe->type;
2459

2460
	if (cft->read_map || cft->read_seq_string) {
2461
		file->f_op = &cgroup_seqfile_operations;
2462 2463
		err = single_open(file, cgroup_seqfile_show, cfe);
	} else if (cft->open) {
2464
		err = cft->open(inode, file);
2465
	}
2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483

	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)
{
2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
	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);

2494 2495 2496 2497 2498 2499
	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;
2500 2501 2502

	cgrp = __d_cgrp(old_dentry);

2503 2504 2505 2506 2507 2508 2509
	/*
	 * This isn't a proper migration and its usefulness is very
	 * limited.  Disallow if sane_behavior.
	 */
	if (cgroup_sane_behavior(cgrp))
		return -EPERM;

2510 2511 2512 2513 2514 2515 2516 2517 2518 2519
	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;
	}

2520
	old_name = rcu_dereference_protected(cgrp->name, true);
2521 2522 2523 2524
	rcu_assign_pointer(cgrp->name, name);

	kfree_rcu(old_name, rcu_head);
	return 0;
2525 2526
}

A
Aristeu Rozanski 已提交
2527 2528 2529 2530 2531
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
2532
		return &__d_cfe(dentry)->xattrs;
A
Aristeu Rozanski 已提交
2533 2534 2535 2536 2537
}

static inline int xattr_enabled(struct dentry *dentry)
{
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
2538
	return root->flags & CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 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
}

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);
}

2585
static const struct file_operations cgroup_file_operations = {
2586 2587 2588 2589 2590 2591 2592
	.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
2593 2594 2595 2596 2597 2598 2599
static const struct inode_operations cgroup_file_inode_operations = {
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
};

2600
static const struct inode_operations cgroup_dir_inode_operations = {
2601
	.lookup = cgroup_lookup,
2602 2603 2604
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
A
Aristeu Rozanski 已提交
2605 2606 2607 2608
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
2609 2610
};

A
Al Viro 已提交
2611
static struct dentry *cgroup_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2612 2613 2614 2615 2616 2617 2618
{
	if (dentry->d_name.len > NAME_MAX)
		return ERR_PTR(-ENAMETOOLONG);
	d_add(dentry, NULL);
	return NULL;
}

2619 2620 2621 2622 2623
/*
 * Check if a file is a control file
 */
static inline struct cftype *__file_cft(struct file *file)
{
A
Al Viro 已提交
2624
	if (file_inode(file)->i_fop != &cgroup_file_operations)
2625 2626 2627 2628
		return ERR_PTR(-EINVAL);
	return __d_cft(file->f_dentry);
}

A
Al Viro 已提交
2629
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2630 2631
				struct super_block *sb)
{
2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
	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
2649
		inc_nlink(dentry->d_parent->d_inode);
2650

2651 2652 2653 2654 2655 2656 2657 2658 2659
		/*
		 * 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));
2660 2661 2662
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
A
Aristeu Rozanski 已提交
2663
		inode->i_op = &cgroup_file_inode_operations;
2664 2665 2666 2667 2668 2669
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

L
Li Zefan 已提交
2670 2671 2672 2673 2674 2675 2676 2677 2678
/**
 * 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
2679
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2680
{
A
Al Viro 已提交
2681
	umode_t mode = 0;
L
Li Zefan 已提交
2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696

	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
2697
static int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2698
			   struct cftype *cft)
2699
{
2700
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2701
	struct cgroup *parent = __d_cgrp(dir);
2702
	struct dentry *dentry;
T
Tejun Heo 已提交
2703
	struct cfent *cfe;
2704
	int error;
A
Al Viro 已提交
2705
	umode_t mode;
2706
	char name[MAX_CGROUP_TYPE_NAMELEN + MAX_CFTYPE_NAME + 2] = { 0 };
2707

2708
	if (subsys && !(cgrp->root->flags & CGRP_ROOT_NOPREFIX)) {
2709 2710 2711 2712
		strcpy(name, subsys->name);
		strcat(name, ".");
	}
	strcat(name, cft->name);
T
Tejun Heo 已提交
2713

2714
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2715 2716 2717 2718 2719

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

2720
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2721
	if (IS_ERR(dentry)) {
2722
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2723 2724 2725
		goto out;
	}

2726 2727 2728 2729 2730
	cfe->type = (void *)cft;
	cfe->dentry = dentry;
	dentry->d_fsdata = cfe;
	simple_xattrs_init(&cfe->xattrs);

T
Tejun Heo 已提交
2731 2732 2733 2734 2735 2736 2737 2738 2739
	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);
2740 2741 2742
	return error;
}

2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754
/**
 * cgroup_addrm_files - add or remove files to a cgroup directory
 * @cgrp: the target cgroup
 * @subsys: the subsystem of files to be added
 * @cfts: array of cftypes to be added
 * @is_add: whether to add or remove
 *
 * Depending on @is_add, add or remove files defined by @cfts on @cgrp.
 * All @cfts should belong to @subsys.  For removals, this function never
 * fails.  If addition fails, this function doesn't remove files already
 * added.  The caller is responsible for cleaning up.
 */
2755
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2756
			      struct cftype cfts[], bool is_add)
2757
{
A
Aristeu Rozanski 已提交
2758
	struct cftype *cft;
2759 2760 2761 2762
	int ret;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);
T
Tejun Heo 已提交
2763 2764

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2765
		/* does cft->flags tell us to skip this file on @cgrp? */
2766 2767
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2768 2769 2770 2771 2772
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2773
		if (is_add) {
2774 2775
			ret = cgroup_add_file(cgrp, subsys, cft);
			if (ret) {
2776
				pr_warn("cgroup_addrm_files: failed to add %s, err=%d\n",
2777 2778 2779
					cft->name, ret);
				return ret;
			}
2780 2781
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2782
		}
2783
	}
2784
	return 0;
2785 2786
}

2787
static void cgroup_cfts_prepare(void)
2788
	__acquires(&cgroup_mutex)
2789 2790 2791 2792
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
2793 2794
	 * Instead, we use cgroup_for_each_descendant_pre() and drop RCU
	 * read lock before calling cgroup_addrm_files().
2795 2796 2797 2798
	 */
	mutex_lock(&cgroup_mutex);
}

2799 2800
static int cgroup_cfts_commit(struct cgroup_subsys *ss,
			      struct cftype *cfts, bool is_add)
2801
	__releases(&cgroup_mutex)
2802 2803
{
	LIST_HEAD(pending);
2804
	struct cgroup *cgrp, *root = &ss->root->top_cgroup;
2805
	struct super_block *sb = ss->root->sb;
2806 2807
	struct dentry *prev = NULL;
	struct inode *inode;
2808
	u64 update_before;
2809
	int ret = 0;
2810 2811

	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
2812
	if (!cfts || ss->root == &cgroup_dummy_root ||
2813 2814
	    !atomic_inc_not_zero(&sb->s_active)) {
		mutex_unlock(&cgroup_mutex);
2815
		return 0;
2816 2817 2818
	}

	/*
2819 2820
	 * All cgroups which are created after we drop cgroup_mutex will
	 * have the updated set of files, so we only need to update the
2821
	 * cgroups created before the current @cgroup_serial_nr_next.
2822
	 */
2823
	update_before = cgroup_serial_nr_next;
2824 2825 2826 2827 2828 2829 2830

	mutex_unlock(&cgroup_mutex);

	/* @root always needs to be updated */
	inode = root->dentry->d_inode;
	mutex_lock(&inode->i_mutex);
	mutex_lock(&cgroup_mutex);
2831
	ret = cgroup_addrm_files(root, ss, cfts, is_add);
2832 2833 2834
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);

2835 2836 2837
	if (ret)
		goto out_deact;

2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849
	/* 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;
2850 2851 2852

		mutex_lock(&inode->i_mutex);
		mutex_lock(&cgroup_mutex);
2853
		if (cgrp->serial_nr < update_before && !cgroup_is_dead(cgrp))
2854
			ret = cgroup_addrm_files(cgrp, ss, cfts, is_add);
2855 2856 2857
		mutex_unlock(&cgroup_mutex);
		mutex_unlock(&inode->i_mutex);

2858
		rcu_read_lock();
2859 2860
		if (ret)
			break;
2861
	}
2862 2863
	rcu_read_unlock();
	dput(prev);
2864
out_deact:
2865
	deactivate_super(sb);
2866
	return ret;
2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882
}

/**
 * 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
2883
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2884 2885
{
	struct cftype_set *set;
2886
	int ret;
2887 2888 2889 2890 2891 2892 2893 2894

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

	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2895 2896 2897 2898
	ret = cgroup_cfts_commit(ss, cfts, true);
	if (ret)
		cgroup_rm_cftypes(ss, cfts);
	return ret;
2899 2900 2901
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914
/**
 * 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
2915
int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2916 2917 2918 2919 2920 2921 2922
{
	struct cftype_set *set;

	cgroup_cfts_prepare();

	list_for_each_entry(set, &ss->cftsets, node) {
		if (set->cfts == cfts) {
2923 2924
			list_del(&set->node);
			kfree(set);
2925 2926 2927 2928 2929 2930 2931 2932 2933
			cgroup_cfts_commit(ss, cfts, false);
			return 0;
		}
	}

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

L
Li Zefan 已提交
2934 2935 2936 2937 2938 2939
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2940
int cgroup_task_count(const struct cgroup *cgrp)
2941 2942
{
	int count = 0;
2943
	struct cgrp_cset_link *link;
2944 2945

	read_lock(&css_set_lock);
2946 2947
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
2948
	read_unlock(&css_set_lock);
2949 2950 2951
	return count;
}

2952 2953 2954 2955
/*
 * Advance a list_head iterator.  The iterator should be positioned at
 * the start of a css_set
 */
2956
static void cgroup_advance_iter(struct cgroup *cgrp, struct cgroup_iter *it)
2957
{
2958 2959
	struct list_head *l = it->cset_link;
	struct cgrp_cset_link *link;
2960
	struct css_set *cset;
2961 2962 2963 2964

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
2965 2966
		if (l == &cgrp->cset_links) {
			it->cset_link = NULL;
2967 2968
			return;
		}
2969 2970
		link = list_entry(l, struct cgrp_cset_link, cset_link);
		cset = link->cset;
2971
	} while (list_empty(&cset->tasks));
2972
	it->cset_link = l;
2973
	it->task = cset->tasks.next;
2974 2975
}

2976 2977 2978 2979 2980 2981
/*
 * 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().
 */
2982
static void cgroup_enable_task_cg_lists(void)
2983 2984 2985 2986
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
2987 2988 2989 2990 2991 2992 2993 2994
	/*
	 * 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);
2995 2996
	do_each_thread(g, p) {
		task_lock(p);
2997 2998 2999 3000 3001 3002
		/*
		 * 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))
3003
			list_add(&p->cg_list, &task_css_set(p)->tasks);
3004 3005
		task_unlock(p);
	} while_each_thread(g, p);
3006
	read_unlock(&tasklist_lock);
3007 3008 3009
	write_unlock(&css_set_lock);
}

3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027
/**
 * 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
3028 3029 3030 3031 3032 3033 3034
	 * 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.
3035
	 */
3036
	if (likely(!cgroup_is_dead(pos))) {
3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
		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);

3060 3061 3062 3063 3064 3065 3066
/**
 * 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.
3067 3068 3069 3070 3071
 *
 * 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.
3072 3073 3074 3075 3076 3077 3078 3079 3080
 */
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 */
3081
	if (!pos)
3082 3083 3084 3085 3086 3087 3088 3089
		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 */
3090
	while (pos != cgroup) {
3091 3092
		next = cgroup_next_sibling(pos);
		if (next)
3093 3094
			return next;
		pos = pos->parent;
3095
	}
3096 3097 3098 3099 3100

	return NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_descendant_pre);

3101 3102 3103 3104 3105 3106 3107
/**
 * 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.
3108 3109 3110 3111 3112
 *
 * 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.
3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131
 */
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);

3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151
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.
3152 3153 3154 3155 3156
 *
 * 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.
3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171
 */
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 */
3172 3173
	next = cgroup_next_sibling(pos);
	if (next)
3174 3175 3176 3177 3178 3179 3180 3181
		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);

3182
void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it)
3183
	__acquires(css_set_lock)
3184 3185 3186 3187 3188 3189
{
	/*
	 * 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.
	 */
3190 3191 3192
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

3193
	read_lock(&css_set_lock);
3194
	it->cset_link = &cgrp->cset_links;
3195
	cgroup_advance_iter(cgrp, it);
3196 3197
}

3198
struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
3199 3200 3201 3202
					struct cgroup_iter *it)
{
	struct task_struct *res;
	struct list_head *l = it->task;
3203
	struct cgrp_cset_link *link;
3204 3205

	/* If the iterator cg is NULL, we have no tasks */
3206
	if (!it->cset_link)
3207 3208 3209 3210
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
3211 3212
	link = list_entry(it->cset_link, struct cgrp_cset_link, cset_link);
	if (l == &link->cset->tasks) {
3213 3214
		/* We reached the end of this task list - move on to
		 * the next cg_cgroup_link */
3215
		cgroup_advance_iter(cgrp, it);
3216 3217 3218 3219 3220 3221
	} else {
		it->task = l;
	}
	return res;
}

3222
void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it)
3223
	__releases(css_set_lock)
3224 3225 3226 3227
{
	read_unlock(&css_set_lock);
}

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
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;
L
Li Zefan 已提交
3327 3328
	cgroup_iter_start(scan->cgrp, &it);
	while ((p = cgroup_iter_next(scan->cgrp, &it))) {
3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360
		/*
		 * 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
		 */
	}
L
Li Zefan 已提交
3361
	cgroup_iter_end(scan->cgrp, &it);
3362 3363 3364

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3365
			struct task_struct *q = heap->ptrs[i];
3366
			if (i == 0) {
3367 3368
				latest_time = q->start_time;
				latest_task = q;
3369 3370
			}
			/* Process the task per the caller's callback */
3371 3372
			scan->process_task(q, scan);
			put_task_struct(q);
3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387
		}
		/*
		 * 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;
}

3388 3389 3390 3391 3392
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
3393
	mutex_lock(&cgroup_mutex);
3394
	cgroup_attach_task(new_cgroup, task, false);
T
Tejun Heo 已提交
3395
	mutex_unlock(&cgroup_mutex);
3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406
}

/**
 * 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;

L
Li Zefan 已提交
3407
	scan.cgrp = from;
3408 3409 3410 3411 3412 3413 3414 3415
	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);
}

3416
/*
3417
 * Stuff for reading the 'tasks'/'procs' files.
3418 3419 3420 3421 3422 3423 3424 3425
 *
 * 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.
 *
 */

3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454
/* 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 */
L
Li Zefan 已提交
3455
	struct rw_semaphore rwsem;
3456 3457
};

3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478
/*
 * 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);
}

3479
/*
3480
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3481
 * Returns the number of unique elements.
3482
 */
3483
static int pidlist_uniq(pid_t *list, int length)
3484
{
3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513
	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;
}

3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524
/*
 * 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 */
3525
	struct pid_namespace *ns = task_active_pid_ns(current);
3526

3527
	/*
L
Li Zefan 已提交
3528
	 * We can't drop the pidlist_mutex before taking the l->rwsem in case
3529 3530 3531 3532 3533 3534 3535 3536
	 * 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 */
L
Li Zefan 已提交
3537
			down_write(&l->rwsem);
3538 3539 3540 3541 3542
			mutex_unlock(&cgrp->pidlist_mutex);
			return l;
		}
	}
	/* entry not found; create a new one */
3543
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
3544 3545 3546 3547
	if (!l) {
		mutex_unlock(&cgrp->pidlist_mutex);
		return l;
	}
L
Li Zefan 已提交
3548 3549
	init_rwsem(&l->rwsem);
	down_write(&l->rwsem);
3550
	l->key.type = type;
3551
	l->key.ns = get_pid_ns(ns);
3552 3553 3554 3555 3556 3557
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	mutex_unlock(&cgrp->pidlist_mutex);
	return l;
}

3558 3559 3560
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3561 3562
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3563 3564 3565 3566
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3567 3568
	struct cgroup_iter it;
	struct task_struct *tsk;
3569 3570 3571 3572 3573 3574 3575 3576 3577
	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);
3578
	array = pidlist_allocate(length);
3579 3580 3581
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3582 3583
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
3584
		if (unlikely(n == length))
3585
			break;
3586
		/* get tgid or pid for procs or tasks file respectively */
3587 3588 3589 3590
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3591 3592
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3593
	}
3594
	cgroup_iter_end(cgrp, &it);
3595 3596 3597
	length = n;
	/* now sort & (if procs) strip out duplicates */
	sort(array, length, sizeof(pid_t), cmppid, NULL);
3598
	if (type == CGROUP_FILE_PROCS)
3599
		length = pidlist_uniq(array, length);
3600 3601
	l = cgroup_pidlist_find(cgrp, type);
	if (!l) {
3602
		pidlist_free(array);
3603
		return -ENOMEM;
3604
	}
3605
	/* store array, freeing old if necessary - lock already held */
3606
	pidlist_free(l->list);
3607 3608 3609
	l->list = array;
	l->length = length;
	l->use_count++;
L
Li Zefan 已提交
3610
	up_write(&l->rwsem);
3611
	*lp = l;
3612
	return 0;
3613 3614
}

B
Balbir Singh 已提交
3615
/**
L
Li Zefan 已提交
3616
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3617 3618 3619
 * @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
3620 3621 3622
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3623 3624 3625 3626
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3627
	struct cgroup *cgrp;
B
Balbir Singh 已提交
3628 3629
	struct cgroup_iter it;
	struct task_struct *tsk;
3630

B
Balbir Singh 已提交
3631
	/*
3632 3633
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3634
	 */
3635 3636
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3637 3638 3639
		 goto err;

	ret = 0;
3640
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3641

3642 3643
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
B
Balbir Singh 已提交
3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662
		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;
		}
	}
3663
	cgroup_iter_end(cgrp, &it);
B
Balbir Singh 已提交
3664 3665 3666 3667 3668

err:
	return ret;
}

3669

3670
/*
3671
 * seq_file methods for the tasks/procs files. The seq_file position is the
3672
 * next pid to display; the seq_file iterator is a pointer to the pid
3673
 * in the cgroup->l->list array.
3674
 */
3675

3676
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3677
{
3678 3679 3680 3681 3682 3683
	/*
	 * 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
	 */
3684
	struct cgroup_pidlist *l = s->private;
3685 3686 3687
	int index = 0, pid = *pos;
	int *iter;

L
Li Zefan 已提交
3688
	down_read(&l->rwsem);
3689
	if (pid) {
3690
		int end = l->length;
S
Stephen Rothwell 已提交
3691

3692 3693
		while (index < end) {
			int mid = (index + end) / 2;
3694
			if (l->list[mid] == pid) {
3695 3696
				index = mid;
				break;
3697
			} else if (l->list[mid] <= pid)
3698 3699 3700 3701 3702 3703
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3704
	if (index >= l->length)
3705 3706
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3707
	iter = l->list + index;
3708 3709 3710 3711
	*pos = *iter;
	return iter;
}

3712
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3713
{
3714
	struct cgroup_pidlist *l = s->private;
L
Li Zefan 已提交
3715
	up_read(&l->rwsem);
3716 3717
}

3718
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3719
{
3720 3721 3722
	struct cgroup_pidlist *l = s->private;
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735
	/*
	 * 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;
	}
}

3736
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3737 3738 3739
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3740

3741 3742 3743 3744 3745 3746 3747 3748 3749
/*
 * 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,
3750 3751
};

3752
static void cgroup_release_pid_array(struct cgroup_pidlist *l)
3753
{
3754 3755 3756 3757 3758 3759 3760
	/*
	 * 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);
L
Li Zefan 已提交
3761
	down_write(&l->rwsem);
3762 3763
	BUG_ON(!l->use_count);
	if (!--l->use_count) {
3764 3765 3766
		/* we're the last user if refcount is 0; remove and free */
		list_del(&l->links);
		mutex_unlock(&l->owner->pidlist_mutex);
3767
		pidlist_free(l->list);
3768
		put_pid_ns(l->key.ns);
L
Li Zefan 已提交
3769
		up_write(&l->rwsem);
3770 3771
		kfree(l);
		return;
3772
	}
3773
	mutex_unlock(&l->owner->pidlist_mutex);
L
Li Zefan 已提交
3774
	up_write(&l->rwsem);
3775 3776
}

3777
static int cgroup_pidlist_release(struct inode *inode, struct file *file)
3778
{
3779
	struct cgroup_pidlist *l;
3780 3781
	if (!(file->f_mode & FMODE_READ))
		return 0;
3782 3783 3784 3785 3786 3787
	/*
	 * 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);
3788 3789 3790
	return seq_release(inode, file);
}

3791
static const struct file_operations cgroup_pidlist_operations = {
3792 3793 3794
	.read = seq_read,
	.llseek = seq_lseek,
	.write = cgroup_file_write,
3795
	.release = cgroup_pidlist_release,
3796 3797
};

3798
/*
3799 3800 3801
 * 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.
3802
 */
3803
/* helper function for the two below it */
3804
static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
3805
{
3806
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
3807
	struct cgroup_pidlist *l;
3808
	int retval;
3809

3810
	/* Nothing to do for write-only files */
3811 3812 3813
	if (!(file->f_mode & FMODE_READ))
		return 0;

3814
	/* have the array populated */
3815
	retval = pidlist_array_load(cgrp, type, &l);
3816 3817 3818 3819
	if (retval)
		return retval;
	/* configure file information */
	file->f_op = &cgroup_pidlist_operations;
3820

3821
	retval = seq_open(file, &cgroup_pidlist_seq_operations);
3822
	if (retval) {
3823
		cgroup_release_pid_array(l);
3824
		return retval;
3825
	}
3826
	((struct seq_file *)file->private_data)->private = l;
3827 3828
	return 0;
}
3829 3830
static int cgroup_tasks_open(struct inode *unused, struct file *file)
{
3831
	return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
3832 3833 3834
}
static int cgroup_procs_open(struct inode *unused, struct file *file)
{
3835
	return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
3836
}
3837

3838
static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
3839 3840
					    struct cftype *cft)
{
3841
	return notify_on_release(cgrp);
3842 3843
}

3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855
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;
}

3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872
/*
 * 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);
}

3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883
/*
 * 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;

3884 3885
	remove_wait_queue(event->wqh, &event->wait);

3886 3887
	event->cft->unregister_event(cgrp, event->cft, event->eventfd);

3888 3889 3890
	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

3891 3892
	eventfd_ctx_put(event->eventfd);
	kfree(event);
3893
	cgroup_dput(cgrp);
3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910
}

/*
 * 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) {
		/*
3911 3912 3913 3914 3915 3916 3917
		 * 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.
3918
		 */
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928
		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);
3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952
	}

	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)
{
3953
	struct cgroup_event *event;
3954
	struct cgroup *cgrp_cfile;
3955
	unsigned int efd, cfd;
3956 3957
	struct file *efile;
	struct file *cfile;
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
	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);
3983
		goto out_kfree;
3984 3985 3986 3987 3988
	}

	event->eventfd = eventfd_ctx_fileget(efile);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
3989
		goto out_put_efile;
3990 3991 3992 3993 3994
	}

	cfile = fget(cfd);
	if (!cfile) {
		ret = -EBADF;
3995
		goto out_put_eventfd;
3996 3997 3998
	}

	/* the process need read permission on control file */
A
Al Viro 已提交
3999
	/* AV: shouldn't we check that it's been opened for read instead? */
A
Al Viro 已提交
4000
	ret = inode_permission(file_inode(cfile), MAY_READ);
4001
	if (ret < 0)
4002
		goto out_put_cfile;
4003 4004 4005 4006

	event->cft = __file_cft(cfile);
	if (IS_ERR(event->cft)) {
		ret = PTR_ERR(event->cft);
4007
		goto out_put_cfile;
4008 4009
	}

4010 4011 4012 4013 4014 4015 4016
	/*
	 * 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;
4017
		goto out_put_cfile;
4018 4019
	}

4020 4021
	if (!event->cft->register_event || !event->cft->unregister_event) {
		ret = -EINVAL;
4022
		goto out_put_cfile;
4023 4024 4025 4026 4027
	}

	ret = event->cft->register_event(cgrp, event->cft,
			event->eventfd, buffer);
	if (ret)
4028
		goto out_put_cfile;
4029

4030
	efile->f_op->poll(efile, &event->pt);
4031

4032 4033 4034 4035 4036 4037 4038
	/*
	 * 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);

4039 4040 4041 4042 4043 4044 4045 4046 4047
	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;

4048 4049 4050 4051 4052 4053 4054
out_put_cfile:
	fput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fput(efile);
out_kfree:
4055 4056 4057 4058 4059
	kfree(event);

	return ret;
}

4060 4061 4062
static u64 cgroup_clone_children_read(struct cgroup *cgrp,
				    struct cftype *cft)
{
4063
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4064 4065 4066 4067 4068 4069 4070
}

static int cgroup_clone_children_write(struct cgroup *cgrp,
				     struct cftype *cft,
				     u64 val)
{
	if (val)
4071
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4072
	else
4073
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4074 4075 4076
	return 0;
}

4077
static struct cftype cgroup_base_files[] = {
4078
	{
4079
		.name = "cgroup.procs",
4080
		.open = cgroup_procs_open,
B
Ben Blum 已提交
4081
		.write_u64 = cgroup_procs_write,
4082
		.release = cgroup_pidlist_release,
B
Ben Blum 已提交
4083
		.mode = S_IRUGO | S_IWUSR,
4084
	},
4085
	{
4086
		.name = "cgroup.event_control",
4087 4088 4089
		.write_string = cgroup_write_event_control,
		.mode = S_IWUGO,
	},
4090 4091
	{
		.name = "cgroup.clone_children",
4092
		.flags = CFTYPE_INSANE,
4093 4094 4095
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
4096 4097 4098 4099 4100
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
		.read_seq_string = cgroup_sane_behavior_show,
	},
4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120

	/*
	 * 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,
	},
4121 4122
	{
		.name = "release_agent",
4123
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
4124 4125 4126 4127
		.read_seq_string = cgroup_release_agent_show,
		.write_string = cgroup_release_agent_write,
		.max_write_len = PATH_MAX,
	},
T
Tejun Heo 已提交
4128
	{ }	/* terminate */
4129 4130
};

4131
/**
4132
 * cgroup_populate_dir - create subsys files in a cgroup directory
4133 4134
 * @cgrp: target cgroup
 * @subsys_mask: mask of the subsystem ids whose files should be added
4135 4136
 *
 * On failure, no file is added.
4137
 */
4138
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned long subsys_mask)
4139 4140
{
	struct cgroup_subsys *ss;
4141
	int i, ret = 0;
4142

4143
	/* process cftsets of each subsystem */
4144
	for_each_subsys(ss, i) {
4145
		struct cftype_set *set;
4146 4147

		if (!test_bit(i, &subsys_mask))
4148
			continue;
4149

4150 4151 4152 4153 4154
		list_for_each_entry(set, &ss->cftsets, node) {
			ret = cgroup_addrm_files(cgrp, ss, set->cfts, true);
			if (ret < 0)
				goto err;
		}
4155
	}
4156

K
KAMEZAWA Hiroyuki 已提交
4157
	/* This cgroup is ready now */
4158
	for_each_root_subsys(cgrp->root, ss) {
K
KAMEZAWA Hiroyuki 已提交
4159
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4160 4161
		struct css_id *id = rcu_dereference_protected(css->id, true);

K
KAMEZAWA Hiroyuki 已提交
4162 4163 4164 4165 4166
		/*
		 * Update id->css pointer and make this css visible from
		 * CSS ID functions. This pointer will be dereferened
		 * from RCU-read-side without locks.
		 */
4167 4168
		if (id)
			rcu_assign_pointer(id->css, css);
K
KAMEZAWA Hiroyuki 已提交
4169
	}
4170 4171

	return 0;
4172 4173 4174
err:
	cgroup_clear_dir(cgrp, subsys_mask);
	return ret;
4175 4176
}

4177 4178 4179 4180 4181
static void css_dput_fn(struct work_struct *work)
{
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, dput_work);

4182
	cgroup_dput(css->cgroup);
4183 4184
}

4185 4186 4187 4188 4189 4190 4191 4192
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);
}

4193 4194
static void init_cgroup_css(struct cgroup_subsys_state *css,
			       struct cgroup_subsys *ss,
4195
			       struct cgroup *cgrp)
4196
{
4197
	css->cgroup = cgrp;
4198
	css->ss = ss;
4199
	css->flags = 0;
K
KAMEZAWA Hiroyuki 已提交
4200
	css->id = NULL;
4201
	if (cgrp == cgroup_dummy_top)
4202
		css->flags |= CSS_ROOT;
4203 4204
	BUG_ON(cgrp->subsys[ss->subsys_id]);
	cgrp->subsys[ss->subsys_id] = css;
4205 4206

	/*
4207 4208 4209 4210
	 * 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().
4211 4212
	 */
	INIT_WORK(&css->dput_work, css_dput_fn);
4213 4214
}

4215
/* invoke ->css_online() on a new CSS and mark it online if successful */
T
Tejun Heo 已提交
4216
static int online_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
4217
{
4218
	struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
T
Tejun Heo 已提交
4219 4220
	int ret = 0;

4221 4222
	lockdep_assert_held(&cgroup_mutex);

4223
	if (ss->css_online)
4224
		ret = ss->css_online(css);
T
Tejun Heo 已提交
4225
	if (!ret)
4226
		css->flags |= CSS_ONLINE;
T
Tejun Heo 已提交
4227
	return ret;
4228 4229
}

4230
/* if the CSS is online, invoke ->css_offline() on it and mark it offline */
4231 4232 4233 4234 4235 4236 4237 4238 4239
static void offline_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
{
	struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

	lockdep_assert_held(&cgroup_mutex);

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

4240
	if (ss->css_offline)
4241
		ss->css_offline(css);
4242

4243
	css->flags &= ~CSS_ONLINE;
4244 4245
}

4246
/*
L
Li Zefan 已提交
4247 4248 4249 4250
 * 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
4251
 *
L
Li Zefan 已提交
4252
 * Must be called with the mutex on the parent inode held
4253 4254
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
4255
			     umode_t mode)
4256
{
4257
	struct cgroup *cgrp;
4258
	struct cgroup_name *name;
4259 4260 4261 4262 4263
	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
4264
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4265 4266
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4267 4268
		return -ENOMEM;

4269 4270 4271 4272 4273
	name = cgroup_alloc_name(dentry);
	if (!name)
		goto err_free_cgrp;
	rcu_assign_pointer(cgrp->name, name);

4274 4275 4276 4277 4278
	/*
	 * Temporarily set the pointer to NULL, so idr_find() won't return
	 * a half-baked cgroup.
	 */
	cgrp->id = idr_alloc(&root->cgroup_idr, NULL, 1, 0, GFP_KERNEL);
T
Tejun Heo 已提交
4279
	if (cgrp->id < 0)
4280
		goto err_free_name;
T
Tejun Heo 已提交
4281

4282 4283 4284 4285 4286 4287 4288 4289 4290
	/*
	 * 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
4291
		goto err_free_id;
4292 4293
	}

4294 4295 4296 4297 4298 4299 4300
	/* 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);

4301
	init_cgroup_housekeeping(cgrp);
4302

4303 4304 4305
	dentry->d_fsdata = cgrp;
	cgrp->dentry = dentry;

4306 4307
	cgrp->parent = parent;
	cgrp->root = parent->root;
4308

4309 4310 4311
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4312 4313
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4314

4315
	for_each_root_subsys(root, ss) {
4316
		struct cgroup_subsys_state *css;
4317

4318
		css = ss->css_alloc(parent->subsys[ss->subsys_id]);
4319 4320
		if (IS_ERR(css)) {
			err = PTR_ERR(css);
4321
			goto err_free_all;
4322
		}
4323 4324

		err = percpu_ref_init(&css->refcnt, css_release);
4325
		if (err) {
4326
			ss->css_free(css);
4327
			goto err_free_all;
4328
		}
4329

4330
		init_cgroup_css(css, ss, cgrp);
4331

4332 4333 4334
		if (ss->use_id) {
			err = alloc_css_id(ss, parent, cgrp);
			if (err)
4335
				goto err_free_all;
4336
		}
4337 4338
	}

4339 4340 4341 4342 4343
	/*
	 * 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
4344
	err = cgroup_create_file(dentry, S_IFDIR | mode, sb);
4345
	if (err < 0)
4346
		goto err_free_all;
4347
	lockdep_assert_held(&dentry->d_inode->i_mutex);
4348

4349
	cgrp->serial_nr = cgroup_serial_nr_next++;
4350

4351 4352 4353
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
	root->number_of_cgroups++;
T
Tejun Heo 已提交
4354

T
Tejun Heo 已提交
4355
	/* each css holds a ref to the cgroup's dentry */
4356
	for_each_root_subsys(root, ss)
4357
		dget(dentry);
4358

4359 4360 4361
	/* hold a ref to the parent's dentry */
	dget(parent->dentry);

T
Tejun Heo 已提交
4362
	/* creation succeeded, notify subsystems */
4363
	for_each_root_subsys(root, ss) {
T
Tejun Heo 已提交
4364 4365 4366
		err = online_css(ss, cgrp);
		if (err)
			goto err_destroy;
4367 4368 4369 4370 4371 4372 4373 4374 4375

		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;
		}
4376 4377
	}

4378 4379
	idr_replace(&root->cgroup_idr, cgrp, cgrp->id);

4380 4381 4382 4383 4384
	err = cgroup_addrm_files(cgrp, NULL, cgroup_base_files, true);
	if (err)
		goto err_destroy;

	err = cgroup_populate_dir(cgrp, root->subsys_mask);
4385 4386
	if (err)
		goto err_destroy;
4387 4388

	mutex_unlock(&cgroup_mutex);
4389
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4390 4391 4392

	return 0;

4393
err_free_all:
4394
	for_each_root_subsys(root, ss) {
4395 4396 4397 4398
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

		if (css) {
			percpu_ref_cancel_init(&css->refcnt);
4399
			ss->css_free(css);
4400
		}
4401 4402 4403 4404
	}
	mutex_unlock(&cgroup_mutex);
	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);
T
Tejun Heo 已提交
4405
err_free_id:
4406
	idr_remove(&root->cgroup_idr, cgrp->id);
4407 4408
err_free_name:
	kfree(rcu_dereference_raw(cgrp->name));
4409
err_free_cgrp:
4410
	kfree(cgrp);
4411
	return err;
4412 4413 4414 4415 4416 4417

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&dentry->d_inode->i_mutex);
	return err;
4418 4419
}

4420
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4421 4422 4423 4424 4425 4426 4427
{
	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);
}

4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469
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.
 */
4470 4471
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4472
{
4473
	struct dentry *d = cgrp->dentry;
4474
	struct cgroup_event *event, *tmp;
4475
	struct cgroup_subsys *ss;
4476
	bool empty;
4477

4478 4479 4480
	lockdep_assert_held(&d->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

4481
	/*
T
Tejun Heo 已提交
4482 4483
	 * css_set_lock synchronizes access to ->cset_links and prevents
	 * @cgrp from being removed while __put_css_set() is in progress.
4484 4485
	 */
	read_lock(&css_set_lock);
T
Tejun Heo 已提交
4486
	empty = list_empty(&cgrp->cset_links) && list_empty(&cgrp->children);
4487 4488
	read_unlock(&css_set_lock);
	if (!empty)
4489
		return -EBUSY;
L
Li Zefan 已提交
4490

4491
	/*
4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504
	 * 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.
4505
	 */
4506
	atomic_set(&cgrp->css_kill_cnt, 1);
4507
	for_each_root_subsys(cgrp->root, ss) {
4508
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4509

4510 4511 4512 4513 4514 4515 4516 4517
		/*
		 * 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);
4518
	}
4519
	cgroup_css_killed(cgrp);
4520 4521 4522 4523 4524 4525 4526 4527

	/*
	 * 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.
	 */
4528
	set_bit(CGRP_DEAD, &cgrp->flags);
4529

4530 4531 4532 4533 4534 4535 4536
	/* 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);

	/*
4537 4538
	 * Clear and remove @cgrp directory.  The removal puts the base ref
	 * but we aren't quite done with @cgrp yet, so hold onto it.
4539
	 */
4540 4541
	cgroup_clear_dir(cgrp, cgrp->root->subsys_mask);
	cgroup_addrm_files(cgrp, NULL, cgroup_base_files, false);
4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556
	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);

4557 4558 4559
	return 0;
};

4560 4561 4562 4563 4564 4565 4566 4567 4568 4569
/**
 * 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().
 */
4570 4571 4572 4573 4574 4575 4576 4577 4578
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);

4579 4580 4581 4582
	/*
	 * css_tryget() is guaranteed to fail now.  Tell subsystems to
	 * initate destruction.
	 */
4583
	for_each_root_subsys(cgrp->root, ss)
4584
		offline_css(ss, cgrp);
4585 4586

	/*
4587 4588 4589 4590 4591
	 * 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.
4592
	 */
4593
	for_each_root_subsys(cgrp->root, ss)
T
Tejun Heo 已提交
4594
		css_put(cgrp->subsys[ss->subsys_id]);
4595

4596
	/* delete this cgroup from parent->children */
4597
	list_del_rcu(&cgrp->sibling);
4598

4599 4600 4601 4602 4603 4604 4605 4606
	/*
	 * We should remove the cgroup object from idr before its grace
	 * period starts, so we won't be looking up a cgroup while the
	 * cgroup is being freed.
	 */
	idr_remove(&cgrp->root->cgroup_idr, cgrp->id);
	cgrp->id = -1;

4607 4608
	dput(d);

4609
	set_bit(CGRP_RELEASABLE, &parent->flags);
4610 4611
	check_for_release(parent);

4612
	mutex_unlock(&cgroup_mutex);
4613 4614
}

4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625
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;
}

4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639
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);
	}
}

4640
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4641 4642
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4643 4644

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4645

4646 4647
	mutex_lock(&cgroup_mutex);

4648 4649 4650
	/* init base cftset */
	cgroup_init_cftsets(ss);

4651
	/* Create the top cgroup state for this subsystem */
4652 4653
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
4654
	css = ss->css_alloc(cgroup_dummy_top->subsys[ss->subsys_id]);
4655 4656
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4657
	init_cgroup_css(css, ss, cgroup_dummy_top);
4658

L
Li Zefan 已提交
4659
	/* Update the init_css_set to contain a subsys
4660
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4661 4662
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
4663
	init_css_set.subsys[ss->subsys_id] = css;
4664 4665 4666

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

L
Li Zefan 已提交
4667 4668 4669 4670 4671
	/* 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));

4672
	BUG_ON(online_css(ss, cgroup_dummy_top));
4673

4674 4675
	mutex_unlock(&cgroup_mutex);

4676 4677 4678 4679 4680 4681 4682 4683 4684 4685
	/* 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
4686
 * subsystem is built as a module, it will be assigned a new subsys_id and set
4687 4688 4689 4690 4691 4692
 * 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;
4693
	int i, ret;
4694
	struct hlist_node *tmp;
4695
	struct css_set *cset;
4696
	unsigned long key;
4697 4698 4699

	/* check name and function validity */
	if (ss->name == NULL || strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN ||
4700
	    ss->css_alloc == NULL || ss->css_free == NULL)
4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716
		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) {
4717
		/* a sanity check */
4718
		BUG_ON(cgroup_subsys[ss->subsys_id] != ss);
4719 4720 4721
		return 0;
	}

4722 4723 4724
	/* init base cftset */
	cgroup_init_cftsets(ss);

4725
	mutex_lock(&cgroup_mutex);
4726
	cgroup_subsys[ss->subsys_id] = ss;
4727 4728

	/*
4729
	 * no ss->css_alloc seems to need anything important in the ss
4730
	 * struct, so this can happen first (i.e. before the dummy root
4731
	 * attachment).
4732
	 */
4733
	css = ss->css_alloc(cgroup_dummy_top->subsys[ss->subsys_id]);
4734
	if (IS_ERR(css)) {
4735 4736
		/* failure case - need to deassign the cgroup_subsys[] slot. */
		cgroup_subsys[ss->subsys_id] = NULL;
4737 4738 4739 4740
		mutex_unlock(&cgroup_mutex);
		return PTR_ERR(css);
	}

4741 4742
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
4743 4744

	/* our new subsystem will be attached to the dummy hierarchy. */
4745
	init_cgroup_css(css, ss, cgroup_dummy_top);
4746 4747
	/* init_idr must be after init_cgroup_css because it sets css->id. */
	if (ss->use_id) {
4748 4749 4750
		ret = cgroup_init_idr(ss, css);
		if (ret)
			goto err_unload;
4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761
	}

	/*
	 * 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);
4762
	hash_for_each_safe(css_set_table, i, tmp, cset, hlist) {
4763
		/* skip entries that we already rehashed */
4764
		if (cset->subsys[ss->subsys_id])
4765 4766
			continue;
		/* remove existing entry */
4767
		hash_del(&cset->hlist);
4768
		/* set new value */
4769
		cset->subsys[ss->subsys_id] = css;
4770
		/* recompute hash and restore entry */
4771 4772
		key = css_set_hash(cset->subsys);
		hash_add(css_set_table, &cset->hlist, key);
4773 4774 4775
	}
	write_unlock(&css_set_lock);

4776
	ret = online_css(ss, cgroup_dummy_top);
T
Tejun Heo 已提交
4777 4778
	if (ret)
		goto err_unload;
4779

4780 4781 4782
	/* success! */
	mutex_unlock(&cgroup_mutex);
	return 0;
4783 4784 4785 4786 4787 4788

err_unload:
	mutex_unlock(&cgroup_mutex);
	/* @ss can't be mounted here as try_module_get() would fail */
	cgroup_unload_subsys(ss);
	return ret;
4789
}
4790
EXPORT_SYMBOL_GPL(cgroup_load_subsys);
4791

B
Ben Blum 已提交
4792 4793 4794 4795 4796 4797 4798 4799 4800 4801
/**
 * 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)
{
4802
	struct cgrp_cset_link *link;
B
Ben Blum 已提交
4803 4804 4805 4806 4807

	BUG_ON(ss->module == NULL);

	/*
	 * we shouldn't be called if the subsystem is in use, and the use of
4808
	 * try_module_get() in rebind_subsystems() should ensure that it
B
Ben Blum 已提交
4809 4810
	 * doesn't start being used while we're killing it off.
	 */
4811
	BUG_ON(ss->root != &cgroup_dummy_root);
B
Ben Blum 已提交
4812 4813

	mutex_lock(&cgroup_mutex);
4814

4815
	offline_css(ss, cgroup_dummy_top);
4816

T
Tejun Heo 已提交
4817
	if (ss->use_id)
4818 4819
		idr_destroy(&ss->idr);

B
Ben Blum 已提交
4820
	/* deassign the subsys_id */
4821
	cgroup_subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4822

4823
	/* remove subsystem from the dummy root's list of subsystems */
4824
	list_del_init(&ss->sibling);
B
Ben Blum 已提交
4825 4826

	/*
4827 4828 4829
	 * 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
4830 4831
	 */
	write_lock(&css_set_lock);
4832
	list_for_each_entry(link, &cgroup_dummy_top->cset_links, cset_link) {
4833
		struct css_set *cset = link->cset;
4834
		unsigned long key;
B
Ben Blum 已提交
4835

4836 4837 4838 4839
		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
4840 4841 4842 4843
	}
	write_unlock(&css_set_lock);

	/*
4844 4845 4846 4847
	 * 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
4848
	 */
4849
	ss->css_free(cgroup_dummy_top->subsys[ss->subsys_id]);
4850
	cgroup_dummy_top->subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4851 4852 4853 4854 4855

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

4856
/**
L
Li Zefan 已提交
4857 4858 4859 4860
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4861 4862 4863
 */
int __init cgroup_init_early(void)
{
4864
	struct cgroup_subsys *ss;
4865
	int i;
4866

4867
	atomic_set(&init_css_set.refcount, 1);
4868
	INIT_LIST_HEAD(&init_css_set.cgrp_links);
4869
	INIT_LIST_HEAD(&init_css_set.tasks);
4870
	INIT_HLIST_NODE(&init_css_set.hlist);
4871
	css_set_count = 1;
4872 4873
	init_cgroup_root(&cgroup_dummy_root);
	cgroup_root_count = 1;
4874
	RCU_INIT_POINTER(init_task.cgroups, &init_css_set);
4875

4876
	init_cgrp_cset_link.cset = &init_css_set;
4877 4878
	init_cgrp_cset_link.cgrp = cgroup_dummy_top;
	list_add(&init_cgrp_cset_link.cset_link, &cgroup_dummy_top->cset_links);
4879
	list_add(&init_cgrp_cset_link.cgrp_link, &init_css_set.cgrp_links);
4880

4881 4882
	/* at bootup time, we don't worry about modular subsystems */
	for_each_builtin_subsys(ss, i) {
4883 4884
		BUG_ON(!ss->name);
		BUG_ON(strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN);
4885 4886
		BUG_ON(!ss->css_alloc);
		BUG_ON(!ss->css_free);
4887
		if (ss->subsys_id != i) {
D
Diego Calleja 已提交
4888
			printk(KERN_ERR "cgroup: Subsys %s id == %d\n",
4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899
			       ss->name, ss->subsys_id);
			BUG();
		}

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

/**
L
Li Zefan 已提交
4900 4901 4902 4903
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4904 4905 4906
 */
int __init cgroup_init(void)
{
4907
	struct cgroup_subsys *ss;
4908
	unsigned long key;
4909
	int i, err;
4910 4911 4912 4913

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4914

4915
	for_each_builtin_subsys(ss, i) {
4916 4917
		if (!ss->early_init)
			cgroup_init_subsys(ss);
K
KAMEZAWA Hiroyuki 已提交
4918
		if (ss->use_id)
4919
			cgroup_init_idr(ss, init_css_set.subsys[ss->subsys_id]);
4920 4921
	}

4922
	/* allocate id for the dummy hierarchy */
T
Tejun Heo 已提交
4923 4924 4925
	mutex_lock(&cgroup_mutex);
	mutex_lock(&cgroup_root_mutex);

4926 4927 4928 4929
	/* Add init_css_set to the hash table */
	key = css_set_hash(init_css_set.subsys);
	hash_add(css_set_table, &init_css_set.hlist, key);

4930
	BUG_ON(cgroup_init_root_id(&cgroup_dummy_root, 0, 1));
4931

4932 4933 4934 4935
	err = idr_alloc(&cgroup_dummy_root.cgroup_idr, cgroup_dummy_top,
			0, 1, GFP_KERNEL);
	BUG_ON(err < 0);

T
Tejun Heo 已提交
4936 4937 4938
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);

4939 4940 4941 4942 4943 4944
	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4945
	err = register_filesystem(&cgroup_fs_type);
4946 4947
	if (err < 0) {
		kobject_put(cgroup_kobj);
4948
		goto out;
4949
	}
4950

L
Li Zefan 已提交
4951
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4952

4953
out:
4954 4955 4956
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4957 4958
	return err;
}
4959

4960 4961 4962 4963 4964 4965
/*
 * 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,
4966
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4967 4968 4969 4970 4971 4972
 *    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 */
4973
int proc_cgroup_show(struct seq_file *m, void *v)
4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995
{
	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);

4996
	for_each_active_root(root) {
4997
		struct cgroup_subsys *ss;
4998
		struct cgroup *cgrp;
4999 5000
		int count = 0;

5001
		seq_printf(m, "%d:", root->hierarchy_id);
5002
		for_each_root_subsys(root, ss)
5003
			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
5004 5005 5006
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
5007
		seq_putc(m, ':');
5008
		cgrp = task_cgroup_from_root(tsk, root);
5009
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027
		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)
{
5028
	struct cgroup_subsys *ss;
5029 5030
	int i;

5031
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
5032 5033 5034 5035 5036
	/*
	 * 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.
	 */
5037
	mutex_lock(&cgroup_mutex);
5038 5039

	for_each_subsys(ss, i)
5040 5041
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
5042
			   ss->root->number_of_cgroups, !ss->disabled);
5043

5044 5045 5046 5047 5048 5049
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
5050
	return single_open(file, proc_cgroupstats_show, NULL);
5051 5052
}

5053
static const struct file_operations proc_cgroupstats_operations = {
5054 5055 5056 5057 5058 5059
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

5060 5061
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
5062
 * @child: pointer to task_struct of forking parent process.
5063 5064 5065 5066 5067
 *
 * 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
5068 5069 5070 5071
 * 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.
5072 5073 5074 5075 5076 5077
 *
 * 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)
{
5078
	task_lock(current);
5079
	get_css_set(task_css_set(current));
5080
	child->cgroups = current->cgroups;
5081
	task_unlock(current);
5082
	INIT_LIST_HEAD(&child->cg_list);
5083 5084
}

5085
/**
L
Li Zefan 已提交
5086 5087 5088
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
5089 5090 5091 5092 5093
 * 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
5094
 */
5095 5096
void cgroup_post_fork(struct task_struct *child)
{
5097
	struct cgroup_subsys *ss;
5098 5099
	int i;

5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110
	/*
	 * 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.
	 */
5111 5112
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
5113 5114
		task_lock(child);
		if (list_empty(&child->cg_list))
5115
			list_add(&child->cg_list, &task_css_set(child)->tasks);
5116
		task_unlock(child);
5117 5118
		write_unlock(&css_set_lock);
	}
5119 5120 5121 5122 5123 5124 5125

	/*
	 * 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) {
5126 5127 5128 5129 5130 5131 5132 5133
		/*
		 * 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.
		 */
5134
		for_each_builtin_subsys(ss, i)
5135 5136 5137
			if (ss->fork)
				ss->fork(child);
	}
5138
}
5139

5140 5141 5142
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
5143
 * @run_callback: run exit callbacks?
5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171
 *
 * 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,
5172 5173
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
5174 5175 5176
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
5177
	struct cgroup_subsys *ss;
5178
	struct css_set *cset;
5179
	int i;
5180 5181 5182 5183 5184 5185 5186 5187 5188

	/*
	 * 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))
5189
			list_del_init(&tsk->cg_list);
5190 5191 5192
		write_unlock(&css_set_lock);
	}

5193 5194
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
5195 5196
	cset = task_css_set(tsk);
	RCU_INIT_POINTER(tsk->cgroups, &init_css_set);
5197 5198

	if (run_callbacks && need_forkexit_callback) {
5199 5200 5201 5202
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, see cgroup_post_fork() for details.
		 */
5203
		for_each_builtin_subsys(ss, i) {
5204
			if (ss->exit) {
5205 5206
				struct cgroup_subsys_state *old_css = cset->subsys[i];
				struct cgroup_subsys_state *css = task_css(tsk, i);
5207

5208
				ss->exit(css, old_css, tsk);
5209 5210 5211
			}
		}
	}
5212
	task_unlock(tsk);
5213

5214
	put_css_set_taskexit(cset);
5215
}
5216

5217
static void check_for_release(struct cgroup *cgrp)
5218
{
5219
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
5220
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
5221 5222
		/*
		 * Control Group is currently removeable. If it's not
5223
		 * already queued for a userspace notification, queue
5224 5225
		 * it now
		 */
5226
		int need_schedule_work = 0;
5227

5228
		raw_spin_lock(&release_list_lock);
5229
		if (!cgroup_is_dead(cgrp) &&
5230 5231
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
5232 5233
			need_schedule_work = 1;
		}
5234
		raw_spin_unlock(&release_list_lock);
5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266
		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);
5267
	raw_spin_lock(&release_list_lock);
5268 5269 5270
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
5271
		char *pathbuf = NULL, *agentbuf = NULL;
5272
		struct cgroup *cgrp = list_entry(release_list.next,
5273 5274
						    struct cgroup,
						    release_list);
5275
		list_del_init(&cgrp->release_list);
5276
		raw_spin_unlock(&release_list_lock);
5277
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
5278 5279 5280 5281 5282 5283 5284
		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;
5285 5286

		i = 0;
5287 5288
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302
		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);
5303 5304 5305
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5306
		raw_spin_lock(&release_list_lock);
5307
	}
5308
	raw_spin_unlock(&release_list_lock);
5309 5310
	mutex_unlock(&cgroup_mutex);
}
5311 5312 5313

static int __init cgroup_disable(char *str)
{
5314
	struct cgroup_subsys *ss;
5315
	char *token;
5316
	int i;
5317 5318 5319 5320

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5321

5322 5323 5324 5325 5326
		/*
		 * cgroup_disable, being at boot time, can't know about
		 * module subsystems, so we don't worry about them.
		 */
		for_each_builtin_subsys(ss, i) {
5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337
			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
5338 5339 5340 5341 5342

/*
 * Functons for CSS ID.
 */

5343
/* to get ID other than 0, this should be called when !cgroup_is_dead() */
K
KAMEZAWA Hiroyuki 已提交
5344 5345
unsigned short css_id(struct cgroup_subsys_state *css)
{
5346 5347 5348 5349 5350 5351 5352
	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.
	 */
5353
	cssid = rcu_dereference_raw(css->id);
K
KAMEZAWA Hiroyuki 已提交
5354 5355 5356 5357 5358

	if (cssid)
		return cssid->id;
	return 0;
}
B
Ben Blum 已提交
5359
EXPORT_SYMBOL_GPL(css_id);
K
KAMEZAWA Hiroyuki 已提交
5360

5361 5362 5363 5364 5365 5366
/**
 *  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
5367
 * this function reads css->id, the caller must hold rcu_read_lock().
5368 5369 5370 5371 5372 5373
 * 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
5374
bool css_is_ancestor(struct cgroup_subsys_state *child,
5375
		    const struct cgroup_subsys_state *root)
K
KAMEZAWA Hiroyuki 已提交
5376
{
5377 5378
	struct css_id *child_id;
	struct css_id *root_id;
K
KAMEZAWA Hiroyuki 已提交
5379

5380
	child_id  = rcu_dereference(child->id);
5381 5382
	if (!child_id)
		return false;
5383
	root_id = rcu_dereference(root->id);
5384 5385 5386 5387 5388 5389 5390
	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
5391 5392 5393 5394
}

void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css)
{
5395 5396
	struct css_id *id = rcu_dereference_protected(css->id, true);

K
KAMEZAWA Hiroyuki 已提交
5397 5398 5399 5400 5401 5402 5403 5404
	/* 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);
5405
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5406
	idr_remove(&ss->idr, id->id);
5407
	spin_unlock(&ss->id_lock);
5408
	kfree_rcu(id, rcu_head);
K
KAMEZAWA Hiroyuki 已提交
5409
}
B
Ben Blum 已提交
5410
EXPORT_SYMBOL_GPL(free_css_id);
K
KAMEZAWA Hiroyuki 已提交
5411 5412 5413 5414 5415 5416 5417 5418 5419

/*
 * 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
5420
	int ret, size;
K
KAMEZAWA Hiroyuki 已提交
5421 5422 5423 5424 5425 5426 5427

	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
5428 5429

	idr_preload(GFP_KERNEL);
5430
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5431
	/* Don't use 0. allocates an ID of 1-65535 */
T
Tejun Heo 已提交
5432
	ret = idr_alloc(&ss->idr, newid, 1, CSS_ID_MAX + 1, GFP_NOWAIT);
5433
	spin_unlock(&ss->id_lock);
T
Tejun Heo 已提交
5434
	idr_preload_end();
K
KAMEZAWA Hiroyuki 已提交
5435 5436

	/* Returns error when there are no free spaces for new ID.*/
T
Tejun Heo 已提交
5437
	if (ret < 0)
K
KAMEZAWA Hiroyuki 已提交
5438 5439
		goto err_out;

T
Tejun Heo 已提交
5440
	newid->id = ret;
K
KAMEZAWA Hiroyuki 已提交
5441 5442 5443 5444
	newid->depth = depth;
	return newid;
err_out:
	kfree(newid);
T
Tejun Heo 已提交
5445
	return ERR_PTR(ret);
K
KAMEZAWA Hiroyuki 已提交
5446 5447 5448

}

5449 5450
static int __init_or_module cgroup_init_idr(struct cgroup_subsys *ss,
					    struct cgroup_subsys_state *rootcss)
K
KAMEZAWA Hiroyuki 已提交
5451 5452 5453
{
	struct css_id *newid;

5454
	spin_lock_init(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5455 5456 5457 5458 5459 5460 5461
	idr_init(&ss->idr);

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

	newid->stack[0] = newid->id;
5462 5463
	RCU_INIT_POINTER(newid->css, rootcss);
	RCU_INIT_POINTER(rootcss->id, newid);
K
KAMEZAWA Hiroyuki 已提交
5464 5465 5466 5467 5468 5469 5470 5471
	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;
5472
	struct css_id *child_id, *parent_id;
K
KAMEZAWA Hiroyuki 已提交
5473 5474 5475 5476

	subsys_id = ss->subsys_id;
	parent_css = parent->subsys[subsys_id];
	child_css = child->subsys[subsys_id];
5477
	parent_id = rcu_dereference_protected(parent_css->id, true);
5478
	depth = parent_id->depth + 1;
K
KAMEZAWA Hiroyuki 已提交
5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515

	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
5516
EXPORT_SYMBOL_GPL(css_lookup);
K
KAMEZAWA Hiroyuki 已提交
5517

S
Stephane Eranian 已提交
5518 5519 5520 5521 5522 5523 5524 5525 5526
/*
 * 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
5527
	inode = file_inode(f);
S
Stephane Eranian 已提交
5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540
	/* 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);
}

5541
#ifdef CONFIG_CGROUP_DEBUG
5542 5543
static struct cgroup_subsys_state *
debug_css_alloc(struct cgroup_subsys_state *parent_css)
5544 5545 5546 5547 5548 5549 5550 5551 5552
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

5553
static void debug_css_free(struct cgroup_subsys_state *css)
5554
{
5555
	kfree(css);
5556 5557
}

L
Li Zefan 已提交
5558
static u64 debug_taskcount_read(struct cgroup *cgrp, struct cftype *cft)
5559
{
L
Li Zefan 已提交
5560
	return cgroup_task_count(cgrp);
5561 5562
}

L
Li Zefan 已提交
5563
static u64 current_css_set_read(struct cgroup *cgrp, struct cftype *cft)
5564 5565 5566 5567
{
	return (u64)(unsigned long)current->cgroups;
}

L
Li Zefan 已提交
5568 5569
static u64 current_css_set_refcount_read(struct cgroup *cgrp,
					 struct cftype *cft)
5570 5571 5572 5573
{
	u64 count;

	rcu_read_lock();
5574
	count = atomic_read(&task_css_set(current)->refcount);
5575 5576 5577 5578
	rcu_read_unlock();
	return count;
}

L
Li Zefan 已提交
5579
static int current_css_set_cg_links_read(struct cgroup *cgrp,
5580 5581 5582
					 struct cftype *cft,
					 struct seq_file *seq)
{
5583
	struct cgrp_cset_link *link;
5584
	struct css_set *cset;
5585 5586 5587

	read_lock(&css_set_lock);
	rcu_read_lock();
5588
	cset = rcu_dereference(current->cgroups);
5589
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5590 5591 5592 5593 5594 5595 5596
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5597 5598
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5599 5600 5601 5602 5603 5604 5605
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
L
Li Zefan 已提交
5606
static int cgroup_css_links_read(struct cgroup *cgrp,
5607 5608 5609
				 struct cftype *cft,
				 struct seq_file *seq)
{
5610
	struct cgrp_cset_link *link;
5611 5612

	read_lock(&css_set_lock);
L
Li Zefan 已提交
5613
	list_for_each_entry(link, &cgrp->cset_links, cset_link) {
5614
		struct css_set *cset = link->cset;
5615 5616
		struct task_struct *task;
		int count = 0;
5617 5618
		seq_printf(seq, "css_set %p\n", cset);
		list_for_each_entry(task, &cset->tasks, cg_list) {
5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631
			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;
}

5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652
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,
	},

5653 5654 5655 5656 5657 5658 5659 5660 5661 5662
	{
		.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,
	},

5663 5664 5665 5666 5667
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5668 5669
	{ }	/* terminate */
};
5670 5671 5672

struct cgroup_subsys debug_subsys = {
	.name = "debug",
5673 5674
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5675
	.subsys_id = debug_subsys_id,
5676
	.base_cftypes = debug_files,
5677 5678
};
#endif /* CONFIG_CGROUP_DEBUG */