cgroup.c 152.5 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 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, 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);
T
Tejun Heo 已提交
1368
	cgrp->dummy_css.cgroup = cgrp;
1369 1370
	INIT_LIST_HEAD(&cgrp->event_list);
	spin_lock_init(&cgrp->event_list_lock);
A
Aristeu Rozanski 已提交
1371
	simple_xattrs_init(&cgrp->xattrs);
1372
}
1373

1374 1375
static void init_cgroup_root(struct cgroupfs_root *root)
{
1376
	struct cgroup *cgrp = &root->top_cgroup;
1377

1378 1379 1380
	INIT_LIST_HEAD(&root->subsys_list);
	INIT_LIST_HEAD(&root->root_list);
	root->number_of_cgroups = 1;
1381
	cgrp->root = root;
1382
	RCU_INIT_POINTER(cgrp->name, &root_cgroup_name);
1383
	init_cgroup_housekeeping(cgrp);
1384
	idr_init(&root->cgroup_idr);
1385 1386
}

1387
static int cgroup_init_root_id(struct cgroupfs_root *root, int start, int end)
1388
{
1389
	int id;
1390

T
Tejun Heo 已提交
1391 1392 1393
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1394 1395
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, start, end,
			      GFP_KERNEL);
1396 1397 1398 1399
	if (id < 0)
		return id;

	root->hierarchy_id = id;
1400 1401 1402 1403 1404
	return 0;
}

static void cgroup_exit_root_id(struct cgroupfs_root *root)
{
T
Tejun Heo 已提交
1405 1406 1407
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&cgroup_root_mutex);

1408
	if (root->hierarchy_id) {
1409
		idr_remove(&cgroup_hierarchy_idr, root->hierarchy_id);
1410 1411
		root->hierarchy_id = 0;
	}
1412 1413
}

1414 1415
static int cgroup_test_super(struct super_block *sb, void *data)
{
1416
	struct cgroup_sb_opts *opts = data;
1417 1418
	struct cgroupfs_root *root = sb->s_fs_info;

1419 1420 1421
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1422

1423 1424 1425 1426
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
1427 1428
	if ((opts->subsys_mask || opts->none)
	    && (opts->subsys_mask != root->subsys_mask))
1429 1430 1431 1432 1433
		return 0;

	return 1;
}

1434 1435 1436 1437
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1438
	if (!opts->subsys_mask && !opts->none)
1439 1440 1441 1442 1443 1444 1445
		return NULL;

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

	init_cgroup_root(root);
1446

1447 1448 1449 1450 1451 1452 1453 1454
	/*
	 * 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.
	 */
1455
	root->subsys_mask = opts->subsys_mask;
1456 1457 1458 1459 1460
	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1461 1462
	if (opts->cpuset_clone_children)
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags);
1463 1464 1465
	return root;
}

1466
static void cgroup_free_root(struct cgroupfs_root *root)
1467
{
1468 1469 1470
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);
1471

1472
		idr_destroy(&root->cgroup_idr);
1473 1474
		kfree(root);
	}
1475 1476
}

1477 1478 1479
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1480 1481 1482 1483 1484 1485
	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;

1486
	BUG_ON(!opts->subsys_mask && !opts->none);
1487 1488 1489 1490 1491

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

1492 1493
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504

	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
1505 1506
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1507
		.d_delete = cgroup_delete,
A
Al Viro 已提交
1508 1509
	};

1510 1511 1512 1513 1514 1515 1516 1517 1518 1519
	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);
1520 1521
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1522
		return -ENOMEM;
A
Al Viro 已提交
1523 1524
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1525 1526 1527
	return 0;
}

A
Al Viro 已提交
1528
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1529
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1530
			 void *data)
1531 1532
{
	struct cgroup_sb_opts opts;
1533
	struct cgroupfs_root *root;
1534 1535
	int ret = 0;
	struct super_block *sb;
1536
	struct cgroupfs_root *new_root;
1537
	struct list_head tmp_links;
T
Tejun Heo 已提交
1538
	struct inode *inode;
1539
	const struct cred *cred;
1540 1541

	/* First find the desired set of subsystems */
B
Ben Blum 已提交
1542
	mutex_lock(&cgroup_mutex);
1543
	ret = parse_cgroupfs_options(data, &opts);
B
Ben Blum 已提交
1544
	mutex_unlock(&cgroup_mutex);
1545 1546
	if (ret)
		goto out_err;
1547

1548 1549 1550 1551 1552 1553 1554
	/*
	 * 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);
1555
		goto out_err;
1556
	}
1557
	opts.new_root = new_root;
1558

1559
	/* Locate an existing or new sb for this hierarchy */
D
David Howells 已提交
1560
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1561
	if (IS_ERR(sb)) {
1562
		ret = PTR_ERR(sb);
1563
		cgroup_free_root(opts.new_root);
1564
		goto out_err;
1565 1566
	}

1567 1568 1569 1570
	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 */
1571
		struct cgroup *root_cgrp = &root->top_cgroup;
1572
		struct cgroupfs_root *existing_root;
1573
		int i;
1574
		struct css_set *cset;
1575 1576 1577 1578 1579 1580

		BUG_ON(sb->s_root != NULL);

		ret = cgroup_get_rootdir(sb);
		if (ret)
			goto drop_new_super;
1581
		inode = sb->s_root->d_inode;
1582

1583
		mutex_lock(&inode->i_mutex);
1584
		mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1585
		mutex_lock(&cgroup_root_mutex);
1586

1587 1588 1589 1590 1591
		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
1592 1593 1594 1595 1596 1597
		/* 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;
1598

1599 1600 1601 1602 1603 1604 1605
		/*
		 * 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
		 */
1606
		ret = allocate_cgrp_cset_links(css_set_count, &tmp_links);
T
Tejun Heo 已提交
1607 1608
		if (ret)
			goto unlock_drop;
1609

1610 1611
		/* ID 0 is reserved for dummy root, 1 for unified hierarchy */
		ret = cgroup_init_root_id(root, 2, 0);
1612 1613 1614
		if (ret)
			goto unlock_drop;

1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
		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);

1627
		ret = cgroup_addrm_files(root_cgrp, cgroup_base_files, true);
1628 1629 1630
		if (ret)
			goto rm_base_files;

1631
		ret = rebind_subsystems(root, root->subsys_mask, 0);
1632 1633 1634 1635 1636
		if (ret)
			goto rm_base_files;

		revert_creds(cred);

B
Ben Blum 已提交
1637 1638 1639 1640 1641
		/*
		 * There must be no failure case after here, since rebinding
		 * takes care of subsystems' refcounts, which are explicitly
		 * dropped in the failure exit path.
		 */
1642

1643 1644
		list_add(&root->root_list, &cgroup_roots);
		cgroup_root_count++;
1645

1646 1647 1648
		/* Link the top cgroup in this hierarchy into all
		 * the css_set objects */
		write_lock(&css_set_lock);
1649
		hash_for_each(css_set_table, i, cset, hlist)
1650
			link_css_set(&tmp_links, cset, root_cgrp);
1651 1652
		write_unlock(&css_set_lock);

1653
		free_cgrp_cset_links(&tmp_links);
1654

1655
		BUG_ON(!list_empty(&root_cgrp->children));
1656 1657
		BUG_ON(root->number_of_cgroups != 1);

T
Tejun Heo 已提交
1658
		mutex_unlock(&cgroup_root_mutex);
1659
		mutex_unlock(&cgroup_mutex);
1660
		mutex_unlock(&inode->i_mutex);
1661 1662 1663 1664 1665
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1666
		cgroup_free_root(opts.new_root);
1667

1668
		if ((root->flags ^ opts.flags) & CGRP_ROOT_OPTION_MASK) {
1669 1670 1671 1672 1673 1674 1675
			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");
			}
1676
		}
1677 1678
	}

1679 1680
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1681
	return dget(sb->s_root);
1682

1683 1684
 rm_base_files:
	free_cgrp_cset_links(&tmp_links);
1685
	cgroup_addrm_files(&root->top_cgroup, cgroup_base_files, false);
1686
	revert_creds(cred);
T
Tejun Heo 已提交
1687
 unlock_drop:
1688
	cgroup_exit_root_id(root);
T
Tejun Heo 已提交
1689 1690 1691
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);
1692
 drop_new_super:
1693
	deactivate_locked_super(sb);
1694 1695 1696
 out_err:
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1697
	return ERR_PTR(ret);
1698 1699 1700 1701
}

static void cgroup_kill_sb(struct super_block *sb) {
	struct cgroupfs_root *root = sb->s_fs_info;
1702
	struct cgroup *cgrp = &root->top_cgroup;
1703
	struct cgrp_cset_link *link, *tmp_link;
1704 1705 1706 1707 1708
	int ret;

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1709
	BUG_ON(!list_empty(&cgrp->children));
1710

1711
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1712
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1713
	mutex_lock(&cgroup_root_mutex);
1714 1715

	/* Rebind all subsystems back to the default hierarchy */
1716 1717 1718 1719 1720
	if (root->flags & CGRP_ROOT_SUBSYS_BOUND) {
		ret = rebind_subsystems(root, 0, root->subsys_mask);
		/* Shouldn't be able to fail ... */
		BUG_ON(ret);
	}
1721

1722
	/*
1723
	 * Release all the links from cset_links to this hierarchy's
1724 1725 1726
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1727

1728 1729 1730
	list_for_each_entry_safe(link, tmp_link, &cgrp->cset_links, cset_link) {
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
1731 1732 1733 1734
		kfree(link);
	}
	write_unlock(&css_set_lock);

1735 1736
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
1737
		cgroup_root_count--;
1738
	}
1739

1740 1741
	cgroup_exit_root_id(root);

T
Tejun Heo 已提交
1742
	mutex_unlock(&cgroup_root_mutex);
1743
	mutex_unlock(&cgroup_mutex);
1744
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1745

A
Aristeu Rozanski 已提交
1746 1747
	simple_xattrs_free(&cgrp->xattrs);

1748
	kill_litter_super(sb);
1749
	cgroup_free_root(root);
1750 1751 1752 1753
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1754
	.mount = cgroup_mount,
1755 1756 1757
	.kill_sb = cgroup_kill_sb,
};

1758 1759
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1760 1761 1762 1763 1764 1765
/**
 * 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
 *
1766 1767 1768 1769 1770 1771
 * 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.
1772
 */
1773
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1774
{
1775
	int ret = -ENAMETOOLONG;
1776
	char *start;
1777

1778 1779 1780
	if (!cgrp->parent) {
		if (strlcpy(buf, "/", buflen) >= buflen)
			return -ENAMETOOLONG;
1781 1782 1783
		return 0;
	}

1784 1785
	start = buf + buflen - 1;
	*start = '\0';
1786

1787
	rcu_read_lock();
1788
	do {
1789 1790 1791 1792
		const char *name = cgroup_name(cgrp);
		int len;

		len = strlen(name);
1793
		if ((start -= len) < buf)
1794 1795
			goto out;
		memcpy(start, name, len);
1796

1797
		if (--start < buf)
1798
			goto out;
1799
		*start = '/';
1800 1801

		cgrp = cgrp->parent;
1802
	} while (cgrp->parent);
1803
	ret = 0;
1804
	memmove(buf, start, buf + buflen - start);
1805 1806 1807
out:
	rcu_read_unlock();
	return ret;
1808
}
B
Ben Blum 已提交
1809
EXPORT_SYMBOL_GPL(cgroup_path);
1810

1811
/**
1812
 * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
1813 1814 1815 1816
 * @task: target task
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1817 1818 1819 1820 1821 1822
 * 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.
1823
 */
1824
int task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
1825 1826
{
	struct cgroupfs_root *root;
1827 1828 1829 1830 1831
	struct cgroup *cgrp;
	int hierarchy_id = 1, ret = 0;

	if (buflen < 2)
		return -ENAMETOOLONG;
1832 1833 1834

	mutex_lock(&cgroup_mutex);

1835 1836
	root = idr_get_next(&cgroup_hierarchy_idr, &hierarchy_id);

1837 1838 1839
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
		ret = cgroup_path(cgrp, buf, buflen);
1840 1841 1842
	} else {
		/* if no hierarchy exists, everyone is in "/" */
		memcpy(buf, "/", 2);
1843 1844 1845 1846 1847
	}

	mutex_unlock(&cgroup_mutex);
	return ret;
}
1848
EXPORT_SYMBOL_GPL(task_cgroup_path);
1849

1850 1851 1852
/*
 * Control Group taskset
 */
1853 1854 1855
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
L
Li Zefan 已提交
1856
	struct css_set		*cset;
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 1929
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
1930 1931 1932
/*
 * cgroup_task_migrate - move a task from one cgroup to another.
 *
1933
 * Must be called with cgroup_mutex and threadgroup locked.
B
Ben Blum 已提交
1934
 */
1935 1936 1937
static void cgroup_task_migrate(struct cgroup *old_cgrp,
				struct task_struct *tsk,
				struct css_set *new_cset)
B
Ben Blum 已提交
1938
{
1939
	struct css_set *old_cset;
B
Ben Blum 已提交
1940 1941

	/*
1942 1943 1944
	 * 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
1945
	 */
1946
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
1947
	old_cset = task_css_set(tsk);
B
Ben Blum 已提交
1948 1949

	task_lock(tsk);
1950
	rcu_assign_pointer(tsk->cgroups, new_cset);
B
Ben Blum 已提交
1951 1952 1953 1954 1955
	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))
1956
		list_move(&tsk->cg_list, &new_cset->tasks);
B
Ben Blum 已提交
1957 1958 1959
	write_unlock(&css_set_lock);

	/*
1960 1961 1962
	 * 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
1963
	 */
1964 1965
	set_bit(CGRP_RELEASABLE, &old_cgrp->flags);
	put_css_set(old_cset);
B
Ben Blum 已提交
1966 1967
}

L
Li Zefan 已提交
1968
/**
1969
 * cgroup_attach_task - attach a task or a whole threadgroup to a cgroup
B
Ben Blum 已提交
1970
 * @cgrp: the cgroup to attach to
1971 1972
 * @tsk: the task or the leader of the threadgroup to be attached
 * @threadgroup: attach the whole threadgroup?
B
Ben Blum 已提交
1973
 *
1974
 * Call holding cgroup_mutex and the group_rwsem of the leader. Will take
1975
 * task_lock of @tsk or each thread in the threadgroup individually in turn.
B
Ben Blum 已提交
1976
 */
T
Tejun Heo 已提交
1977 1978
static int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk,
			      bool threadgroup)
B
Ben Blum 已提交
1979 1980 1981 1982 1983
{
	int retval, i, group_size;
	struct cgroup_subsys *ss, *failed_ss = NULL;
	struct cgroupfs_root *root = cgrp->root;
	/* threadgroup list cursor and array */
1984
	struct task_struct *leader = tsk;
1985
	struct task_and_cgroup *tc;
1986
	struct flex_array *group;
1987
	struct cgroup_taskset tset = { };
B
Ben Blum 已提交
1988 1989 1990 1991 1992

	/*
	 * 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
1993 1994
	 * 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
1995
	 */
1996 1997 1998 1999
	if (threadgroup)
		group_size = get_nr_threads(tsk);
	else
		group_size = 1;
2000
	/* flex_array supports very large thread-groups better than kmalloc. */
2001
	group = flex_array_alloc(sizeof(*tc), group_size, GFP_KERNEL);
B
Ben Blum 已提交
2002 2003
	if (!group)
		return -ENOMEM;
2004
	/* pre-allocate to guarantee space while iterating in rcu read-side. */
2005
	retval = flex_array_prealloc(group, 0, group_size, GFP_KERNEL);
2006 2007
	if (retval)
		goto out_free_group_list;
B
Ben Blum 已提交
2008 2009

	i = 0;
2010 2011 2012 2013 2014 2015
	/*
	 * 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
2016
	do {
2017 2018
		struct task_and_cgroup ent;

2019 2020 2021 2022
		/* @tsk either already exited or can't exit until the end */
		if (tsk->flags & PF_EXITING)
			continue;

B
Ben Blum 已提交
2023 2024
		/* as per above, nr_threads may decrease, but not increase. */
		BUG_ON(i >= group_size);
2025 2026
		ent.task = tsk;
		ent.cgrp = task_cgroup_from_root(tsk, root);
2027 2028 2029
		/* nothing to do if this task is already in the cgroup */
		if (ent.cgrp == cgrp)
			continue;
2030 2031 2032 2033
		/*
		 * saying GFP_ATOMIC has no effect here because we did prealloc
		 * earlier, but it's good form to communicate our expectations.
		 */
2034
		retval = flex_array_put(group, i, &ent, GFP_ATOMIC);
2035
		BUG_ON(retval != 0);
B
Ben Blum 已提交
2036
		i++;
2037 2038 2039

		if (!threadgroup)
			break;
B
Ben Blum 已提交
2040
	} while_each_thread(leader, tsk);
2041
	rcu_read_unlock();
B
Ben Blum 已提交
2042 2043
	/* remember the number of threads in the array for later. */
	group_size = i;
2044 2045
	tset.tc_array = group;
	tset.tc_array_len = group_size;
B
Ben Blum 已提交
2046

2047 2048
	/* methods shouldn't be called if no task is actually migrating */
	retval = 0;
2049
	if (!group_size)
2050
		goto out_free_group_list;
2051

B
Ben Blum 已提交
2052 2053 2054
	/*
	 * step 1: check that we can legitimately attach to the cgroup.
	 */
2055
	for_each_root_subsys(root, ss) {
2056 2057
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

B
Ben Blum 已提交
2058
		if (ss->can_attach) {
2059
			retval = ss->can_attach(css, &tset);
B
Ben Blum 已提交
2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
			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++) {
2072 2073
		struct css_set *old_cset;

2074
		tc = flex_array_get(group, i);
2075
		old_cset = task_css_set(tc->task);
L
Li Zefan 已提交
2076 2077
		tc->cset = find_css_set(old_cset, cgrp);
		if (!tc->cset) {
2078 2079
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2080 2081 2082 2083
		}
	}

	/*
2084 2085 2086
	 * 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
2087 2088
	 */
	for (i = 0; i < group_size; i++) {
2089
		tc = flex_array_get(group, i);
L
Li Zefan 已提交
2090
		cgroup_task_migrate(tc->cgrp, tc->task, tc->cset);
B
Ben Blum 已提交
2091 2092 2093 2094
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2095
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2096
	 */
2097
	for_each_root_subsys(root, ss) {
2098 2099
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

B
Ben Blum 已提交
2100
		if (ss->attach)
2101
			ss->attach(css, &tset);
B
Ben Blum 已提交
2102 2103 2104 2105 2106 2107
	}

	/*
	 * step 5: success! and cleanup
	 */
	retval = 0;
2108 2109 2110 2111
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
2112
			if (!tc->cset)
2113
				break;
L
Li Zefan 已提交
2114
			put_css_set(tc->cset);
2115
		}
B
Ben Blum 已提交
2116 2117 2118
	}
out_cancel_attach:
	if (retval) {
2119
		for_each_root_subsys(root, ss) {
2120 2121
			struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

2122
			if (ss == failed_ss)
B
Ben Blum 已提交
2123 2124
				break;
			if (ss->cancel_attach)
2125
				ss->cancel_attach(css, &tset);
B
Ben Blum 已提交
2126 2127 2128
		}
	}
out_free_group_list:
2129
	flex_array_free(group);
B
Ben Blum 已提交
2130 2131 2132 2133 2134
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2135 2136
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2137
 */
B
Ben Blum 已提交
2138
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2139 2140
{
	struct task_struct *tsk;
2141
	const struct cred *cred = current_cred(), *tcred;
2142 2143
	int ret;

B
Ben Blum 已提交
2144 2145 2146
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2147 2148
retry_find_task:
	rcu_read_lock();
2149
	if (pid) {
2150
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2151 2152
		if (!tsk) {
			rcu_read_unlock();
2153 2154
			ret= -ESRCH;
			goto out_unlock_cgroup;
2155
		}
B
Ben Blum 已提交
2156 2157 2158 2159
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2160
		tcred = __task_cred(tsk);
2161 2162 2163
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2164
			rcu_read_unlock();
2165 2166
			ret = -EACCES;
			goto out_unlock_cgroup;
2167
		}
2168 2169
	} else
		tsk = current;
2170 2171

	if (threadgroup)
2172
		tsk = tsk->group_leader;
2173 2174

	/*
2175
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2176 2177 2178
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2179
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2180 2181 2182 2183 2184
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201
	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;
		}
2202 2203 2204 2205
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2206 2207
	threadgroup_unlock(tsk);

2208
	put_task_struct(tsk);
2209
out_unlock_cgroup:
T
Tejun Heo 已提交
2210
	mutex_unlock(&cgroup_mutex);
2211 2212 2213
	return ret;
}

2214 2215 2216 2217 2218 2219 2220 2221 2222 2223
/**
 * 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
2224
	mutex_lock(&cgroup_mutex);
2225
	for_each_active_root(root) {
L
Li Zefan 已提交
2226
		struct cgroup *from_cgrp = task_cgroup_from_root(from, root);
2227

L
Li Zefan 已提交
2228
		retval = cgroup_attach_task(from_cgrp, tsk, false);
2229 2230 2231
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2232
	mutex_unlock(&cgroup_mutex);
2233 2234 2235 2236 2237

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2238
static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2239 2240 2241 2242 2243
{
	return attach_task_by_pid(cgrp, pid, false);
}

static int cgroup_procs_write(struct cgroup *cgrp, struct cftype *cft, u64 tgid)
2244
{
2245
	return attach_task_by_pid(cgrp, tgid, true);
2246 2247
}

2248 2249 2250 2251
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);
2252 2253
	if (strlen(buffer) >= PATH_MAX)
		return -EINVAL;
2254 2255
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
T
Tejun Heo 已提交
2256
	mutex_lock(&cgroup_root_mutex);
2257
	strcpy(cgrp->root->release_agent_path, buffer);
T
Tejun Heo 已提交
2258
	mutex_unlock(&cgroup_root_mutex);
T
Tejun Heo 已提交
2259
	mutex_unlock(&cgroup_mutex);
2260 2261 2262 2263 2264 2265 2266 2267 2268 2269
	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
2270
	mutex_unlock(&cgroup_mutex);
2271 2272 2273
	return 0;
}

2274 2275 2276 2277
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));
2278 2279 2280
	return 0;
}

2281 2282 2283 2284 2285 2286 2287 2288
/* return the css for the given cgroup file */
static struct cgroup_subsys_state *cgroup_file_css(struct cfent *cfe)
{
	struct cftype *cft = cfe->type;
	struct cgroup *cgrp = __d_cgrp(cfe->dentry->d_parent);

	if (cft->ss)
		return cgrp->subsys[cft->ss->subsys_id];
T
Tejun Heo 已提交
2289
	return &cgrp->dummy_css;
2290 2291
}

2292 2293 2294
/* A buffer size big enough for numbers or short strings */
#define CGROUP_LOCAL_BUFFER_SIZE 64

2295
static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
2296 2297 2298
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
2299
{
2300
	char buffer[CGROUP_LOCAL_BUFFER_SIZE];
2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311
	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 */
2312
	if (cft->write_u64) {
K
KOSAKI Motohiro 已提交
2313
		u64 val = simple_strtoull(strstrip(buffer), &end, 0);
2314 2315 2316 2317
		if (*end)
			return -EINVAL;
		retval = cft->write_u64(cgrp, cft, val);
	} else {
K
KOSAKI Motohiro 已提交
2318
		s64 val = simple_strtoll(strstrip(buffer), &end, 0);
2319 2320 2321 2322
		if (*end)
			return -EINVAL;
		retval = cft->write_s64(cgrp, cft, val);
	}
2323 2324 2325 2326 2327
	if (!retval)
		retval = nbytes;
	return retval;
}

2328 2329 2330 2331 2332
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)
{
2333
	char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
	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
2348 2349 2350 2351
	if (nbytes && copy_from_user(buffer, userbuf, nbytes)) {
		retval = -EFAULT;
		goto out;
	}
2352 2353

	buffer[nbytes] = 0;     /* nul-terminate */
K
KOSAKI Motohiro 已提交
2354
	retval = cft->write_string(cgrp, cft, strstrip(buffer));
2355 2356
	if (!retval)
		retval = nbytes;
L
Li Zefan 已提交
2357
out:
2358 2359 2360 2361 2362
	if (buffer != local_buffer)
		kfree(buffer);
	return retval;
}

2363 2364 2365 2366
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);
2367
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2368

2369
	if (cft->write)
2370
		return cft->write(cgrp, cft, file, buf, nbytes, ppos);
2371 2372
	if (cft->write_u64 || cft->write_s64)
		return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
2373 2374
	if (cft->write_string)
		return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
2375 2376 2377 2378
	if (cft->trigger) {
		int ret = cft->trigger(cgrp, (unsigned int)cft->private);
		return ret ? ret : nbytes;
	}
2379
	return -EINVAL;
2380 2381
}

2382 2383 2384 2385
static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
2386
{
2387
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2388
	u64 val = cft->read_u64(cgrp, cft);
2389 2390 2391 2392 2393
	int len = sprintf(tmp, "%llu\n", (unsigned long long) val);

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

2394 2395 2396 2397 2398
static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
{
2399
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2400 2401 2402 2403 2404 2405
	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);
}

2406 2407 2408 2409
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);
2410
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2411 2412

	if (cft->read)
2413
		return cft->read(cgrp, cft, file, buf, nbytes, ppos);
2414 2415
	if (cft->read_u64)
		return cgroup_read_u64(cgrp, cft, file, buf, nbytes, ppos);
2416 2417
	if (cft->read_s64)
		return cgroup_read_s64(cgrp, cft, file, buf, nbytes, ppos);
2418 2419 2420
	return -EINVAL;
}

2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433
/*
 * 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)
{
2434 2435 2436 2437
	struct cfent *cfe = m->private;
	struct cftype *cft = cfe->type;
	struct cgroup *cgrp = __d_cgrp(cfe->dentry->d_parent);

2438 2439 2440 2441 2442
	if (cft->read_map) {
		struct cgroup_map_cb cb = {
			.fill = cgroup_map_add,
			.state = m,
		};
2443
		return cft->read_map(cgrp, cft, &cb);
2444
	}
2445
	return cft->read_seq_string(cgrp, cft, m);
2446 2447
}

2448
static const struct file_operations cgroup_seqfile_operations = {
2449
	.read = seq_read,
2450
	.write = cgroup_file_write,
2451
	.llseek = seq_lseek,
2452
	.release = single_release,
2453 2454
};

2455 2456
static int cgroup_file_open(struct inode *inode, struct file *file)
{
2457 2458 2459
	struct cfent *cfe = __d_cfe(file->f_dentry);
	struct cftype *cft = __d_cft(file->f_dentry);
	struct cgroup_subsys_state *css = cgroup_file_css(cfe);
2460 2461 2462 2463 2464
	int err;

	err = generic_file_open(inode, file);
	if (err)
		return err;
2465 2466 2467 2468 2469 2470

	/*
	 * If the file belongs to a subsystem, pin the css.  Will be
	 * unpinned either on open failure or release.  This ensures that
	 * @css stays alive for all file operations.
	 */
T
Tejun Heo 已提交
2471
	if (css->ss && !css_tryget(css))
2472
		return -ENODEV;
2473

2474
	if (cft->read_map || cft->read_seq_string) {
2475
		file->f_op = &cgroup_seqfile_operations;
2476 2477
		err = single_open(file, cgroup_seqfile_show, cfe);
	} else if (cft->open) {
2478
		err = cft->open(inode, file);
2479
	}
2480

T
Tejun Heo 已提交
2481
	if (css->ss && err)
2482
		css_put(css);
2483 2484 2485 2486 2487
	return err;
}

static int cgroup_file_release(struct inode *inode, struct file *file)
{
2488
	struct cfent *cfe = __d_cfe(file->f_dentry);
2489
	struct cftype *cft = __d_cft(file->f_dentry);
2490 2491 2492
	struct cgroup_subsys_state *css = cgroup_file_css(cfe);
	int ret = 0;

2493
	if (cft->release)
2494
		ret = cft->release(inode, file);
T
Tejun Heo 已提交
2495
	if (css->ss)
2496 2497
		css_put(css);
	return ret;
2498 2499 2500 2501 2502 2503 2504 2505
}

/*
 * 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)
{
2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
	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);

2516 2517 2518 2519 2520 2521
	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;
2522 2523 2524

	cgrp = __d_cgrp(old_dentry);

2525 2526 2527 2528 2529 2530 2531
	/*
	 * This isn't a proper migration and its usefulness is very
	 * limited.  Disallow if sane_behavior.
	 */
	if (cgroup_sane_behavior(cgrp))
		return -EPERM;

2532 2533 2534 2535 2536 2537 2538 2539 2540 2541
	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;
	}

2542
	old_name = rcu_dereference_protected(cgrp->name, true);
2543 2544 2545 2546
	rcu_assign_pointer(cgrp->name, name);

	kfree_rcu(old_name, rcu_head);
	return 0;
2547 2548
}

A
Aristeu Rozanski 已提交
2549 2550 2551 2552 2553
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
2554
		return &__d_cfe(dentry)->xattrs;
A
Aristeu Rozanski 已提交
2555 2556 2557 2558 2559
}

static inline int xattr_enabled(struct dentry *dentry)
{
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
2560
	return root->flags & CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606
}

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);
}

2607
static const struct file_operations cgroup_file_operations = {
2608 2609 2610 2611 2612 2613 2614
	.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
2615 2616 2617 2618 2619 2620 2621
static const struct inode_operations cgroup_file_inode_operations = {
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
};

2622
static const struct inode_operations cgroup_dir_inode_operations = {
2623
	.lookup = cgroup_lookup,
2624 2625 2626
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
A
Aristeu Rozanski 已提交
2627 2628 2629 2630
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
2631 2632
};

A
Al Viro 已提交
2633
static struct dentry *cgroup_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2634 2635 2636 2637 2638 2639 2640
{
	if (dentry->d_name.len > NAME_MAX)
		return ERR_PTR(-ENAMETOOLONG);
	d_add(dentry, NULL);
	return NULL;
}

2641 2642 2643 2644 2645
/*
 * Check if a file is a control file
 */
static inline struct cftype *__file_cft(struct file *file)
{
A
Al Viro 已提交
2646
	if (file_inode(file)->i_fop != &cgroup_file_operations)
2647 2648 2649 2650
		return ERR_PTR(-EINVAL);
	return __d_cft(file->f_dentry);
}

A
Al Viro 已提交
2651
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2652 2653
				struct super_block *sb)
{
2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670
	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
2671
		inc_nlink(dentry->d_parent->d_inode);
2672

2673 2674 2675 2676 2677 2678 2679 2680 2681
		/*
		 * 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));
2682 2683 2684
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
A
Aristeu Rozanski 已提交
2685
		inode->i_op = &cgroup_file_inode_operations;
2686 2687 2688 2689 2690 2691
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

L
Li Zefan 已提交
2692 2693 2694 2695 2696 2697 2698 2699 2700
/**
 * 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
2701
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2702
{
A
Al Viro 已提交
2703
	umode_t mode = 0;
L
Li Zefan 已提交
2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718

	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;
}

2719
static int cgroup_add_file(struct cgroup *cgrp, struct cftype *cft)
2720
{
2721
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2722
	struct cgroup *parent = __d_cgrp(dir);
2723
	struct dentry *dentry;
T
Tejun Heo 已提交
2724
	struct cfent *cfe;
2725
	int error;
A
Al Viro 已提交
2726
	umode_t mode;
2727
	char name[MAX_CGROUP_TYPE_NAMELEN + MAX_CFTYPE_NAME + 2] = { 0 };
2728

2729 2730
	if (cft->ss && !(cgrp->root->flags & CGRP_ROOT_NOPREFIX)) {
		strcpy(name, cft->ss->name);
2731 2732 2733
		strcat(name, ".");
	}
	strcat(name, cft->name);
T
Tejun Heo 已提交
2734

2735
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2736 2737 2738 2739 2740

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

2741
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2742
	if (IS_ERR(dentry)) {
2743
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2744 2745 2746
		goto out;
	}

2747 2748 2749 2750 2751
	cfe->type = (void *)cft;
	cfe->dentry = dentry;
	dentry->d_fsdata = cfe;
	simple_xattrs_init(&cfe->xattrs);

T
Tejun Heo 已提交
2752 2753 2754 2755 2756 2757 2758 2759 2760
	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);
2761 2762 2763
	return error;
}

2764 2765 2766 2767 2768 2769 2770
/**
 * cgroup_addrm_files - add or remove files to a cgroup directory
 * @cgrp: the target cgroup
 * @cfts: array of cftypes to be added
 * @is_add: whether to add or remove
 *
 * Depending on @is_add, add or remove files defined by @cfts on @cgrp.
2771 2772 2773
 * For removals, this function never fails.  If addition fails, this
 * function doesn't remove files already added.  The caller is responsible
 * for cleaning up.
2774
 */
2775 2776
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add)
2777
{
A
Aristeu Rozanski 已提交
2778
	struct cftype *cft;
2779 2780 2781 2782
	int ret;

	lockdep_assert_held(&cgrp->dentry->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);
T
Tejun Heo 已提交
2783 2784

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2785
		/* does cft->flags tell us to skip this file on @cgrp? */
2786 2787
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2788 2789 2790 2791 2792
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2793
		if (is_add) {
2794
			ret = cgroup_add_file(cgrp, cft);
2795
			if (ret) {
2796
				pr_warn("cgroup_addrm_files: failed to add %s, err=%d\n",
2797 2798 2799
					cft->name, ret);
				return ret;
			}
2800 2801
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2802
		}
2803
	}
2804
	return 0;
2805 2806
}

2807
static void cgroup_cfts_prepare(void)
2808
	__acquires(&cgroup_mutex)
2809 2810 2811 2812
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
2813 2814
	 * Instead, we use cgroup_for_each_descendant_pre() and drop RCU
	 * read lock before calling cgroup_addrm_files().
2815 2816 2817 2818
	 */
	mutex_lock(&cgroup_mutex);
}

2819
static int cgroup_cfts_commit(struct cftype *cfts, bool is_add)
2820
	__releases(&cgroup_mutex)
2821 2822
{
	LIST_HEAD(pending);
2823
	struct cgroup_subsys *ss = cfts[0].ss;
2824
	struct cgroup *cgrp, *root = &ss->root->top_cgroup;
2825
	struct super_block *sb = ss->root->sb;
2826 2827
	struct dentry *prev = NULL;
	struct inode *inode;
2828
	u64 update_before;
2829
	int ret = 0;
2830 2831

	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
2832
	if (!cfts || ss->root == &cgroup_dummy_root ||
2833 2834
	    !atomic_inc_not_zero(&sb->s_active)) {
		mutex_unlock(&cgroup_mutex);
2835
		return 0;
2836 2837 2838
	}

	/*
2839 2840
	 * All cgroups which are created after we drop cgroup_mutex will
	 * have the updated set of files, so we only need to update the
2841
	 * cgroups created before the current @cgroup_serial_nr_next.
2842
	 */
2843
	update_before = cgroup_serial_nr_next;
2844 2845 2846 2847 2848 2849 2850

	mutex_unlock(&cgroup_mutex);

	/* @root always needs to be updated */
	inode = root->dentry->d_inode;
	mutex_lock(&inode->i_mutex);
	mutex_lock(&cgroup_mutex);
2851
	ret = cgroup_addrm_files(root, cfts, is_add);
2852 2853 2854
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);

2855 2856 2857
	if (ret)
		goto out_deact;

2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
	/* 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;
2870 2871 2872

		mutex_lock(&inode->i_mutex);
		mutex_lock(&cgroup_mutex);
2873
		if (cgrp->serial_nr < update_before && !cgroup_is_dead(cgrp))
2874
			ret = cgroup_addrm_files(cgrp, cfts, is_add);
2875 2876 2877
		mutex_unlock(&cgroup_mutex);
		mutex_unlock(&inode->i_mutex);

2878
		rcu_read_lock();
2879 2880
		if (ret)
			break;
2881
	}
2882 2883
	rcu_read_unlock();
	dput(prev);
2884
out_deact:
2885
	deactivate_super(sb);
2886
	return ret;
2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902
}

/**
 * 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
2903
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2904 2905
{
	struct cftype_set *set;
2906
	struct cftype *cft;
2907
	int ret;
2908 2909 2910 2911 2912

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

2913 2914 2915
	for (cft = cfts; cft->name[0] != '\0'; cft++)
		cft->ss = ss;

2916 2917 2918
	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2919
	ret = cgroup_cfts_commit(cfts, true);
2920
	if (ret)
2921
		cgroup_rm_cftypes(cfts);
2922
	return ret;
2923 2924 2925
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2926 2927 2928 2929
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
2930 2931 2932
 * Unregister @cfts.  Files described by @cfts are removed from all
 * existing cgroups and all future cgroups won't have them either.  This
 * function can be called anytime whether @cfts' subsys is attached or not.
2933 2934
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
2935
 * registered.
2936
 */
2937
int cgroup_rm_cftypes(struct cftype *cfts)
2938 2939 2940
{
	struct cftype_set *set;

2941 2942 2943
	if (!cfts || !cfts[0].ss)
		return -ENOENT;

2944 2945
	cgroup_cfts_prepare();

2946
	list_for_each_entry(set, &cfts[0].ss->cftsets, node) {
2947
		if (set->cfts == cfts) {
2948 2949
			list_del(&set->node);
			kfree(set);
2950
			cgroup_cfts_commit(cfts, false);
2951 2952 2953 2954
			return 0;
		}
	}

2955
	cgroup_cfts_commit(NULL, false);
2956 2957 2958
	return -ENOENT;
}

L
Li Zefan 已提交
2959 2960 2961 2962 2963 2964
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2965
int cgroup_task_count(const struct cgroup *cgrp)
2966 2967
{
	int count = 0;
2968
	struct cgrp_cset_link *link;
2969 2970

	read_lock(&css_set_lock);
2971 2972
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
2973
	read_unlock(&css_set_lock);
2974 2975 2976
	return count;
}

2977 2978 2979 2980
/*
 * Advance a list_head iterator.  The iterator should be positioned at
 * the start of a css_set
 */
2981
static void cgroup_advance_iter(struct cgroup *cgrp, struct cgroup_iter *it)
2982
{
2983 2984
	struct list_head *l = it->cset_link;
	struct cgrp_cset_link *link;
2985
	struct css_set *cset;
2986 2987 2988 2989

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
2990 2991
		if (l == &cgrp->cset_links) {
			it->cset_link = NULL;
2992 2993
			return;
		}
2994 2995
		link = list_entry(l, struct cgrp_cset_link, cset_link);
		cset = link->cset;
2996
	} while (list_empty(&cset->tasks));
2997
	it->cset_link = l;
2998
	it->task = cset->tasks.next;
2999 3000
}

3001 3002 3003 3004 3005 3006
/*
 * 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().
 */
3007
static void cgroup_enable_task_cg_lists(void)
3008 3009 3010 3011
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
3012 3013 3014 3015 3016 3017 3018 3019
	/*
	 * 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);
3020 3021
	do_each_thread(g, p) {
		task_lock(p);
3022 3023 3024 3025 3026 3027
		/*
		 * 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))
3028
			list_add(&p->cg_list, &task_css_set(p)->tasks);
3029 3030
		task_unlock(p);
	} while_each_thread(g, p);
3031
	read_unlock(&tasklist_lock);
3032 3033 3034
	write_unlock(&css_set_lock);
}

3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052
/**
 * 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
3053 3054 3055 3056 3057 3058 3059
	 * 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.
3060
	 */
3061
	if (likely(!cgroup_is_dead(pos))) {
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084
		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);

3085 3086 3087 3088 3089 3090 3091
/**
 * 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.
3092 3093 3094 3095 3096
 *
 * While this function requires RCU read locking, it doesn't require the
 * whole traversal to be contained in a single RCU critical section.  This
 * function will return the correct next descendant as long as both @pos
 * and @cgroup are accessible and @pos is a descendant of @cgroup.
3097 3098 3099 3100 3101 3102 3103 3104 3105
 */
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 */
3106
	if (!pos)
3107 3108 3109 3110 3111 3112 3113 3114
		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 */
3115
	while (pos != cgroup) {
3116 3117
		next = cgroup_next_sibling(pos);
		if (next)
3118 3119
			return next;
		pos = pos->parent;
3120
	}
3121 3122 3123 3124 3125

	return NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_descendant_pre);

3126 3127 3128 3129 3130 3131 3132
/**
 * 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.
3133 3134 3135 3136 3137
 *
 * 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.
3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156
 */
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);

3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176
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.
3177 3178 3179 3180 3181
 *
 * 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.
3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196
 */
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 */
3197 3198
	next = cgroup_next_sibling(pos);
	if (next)
3199 3200 3201 3202 3203 3204 3205 3206
		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);

3207
void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it)
3208
	__acquires(css_set_lock)
3209 3210 3211 3212 3213 3214
{
	/*
	 * 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.
	 */
3215 3216 3217
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

3218
	read_lock(&css_set_lock);
3219
	it->cset_link = &cgrp->cset_links;
3220
	cgroup_advance_iter(cgrp, it);
3221 3222
}

3223
struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
3224 3225 3226 3227
					struct cgroup_iter *it)
{
	struct task_struct *res;
	struct list_head *l = it->task;
3228
	struct cgrp_cset_link *link;
3229 3230

	/* If the iterator cg is NULL, we have no tasks */
3231
	if (!it->cset_link)
3232 3233 3234 3235
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
3236 3237
	link = list_entry(it->cset_link, struct cgrp_cset_link, cset_link);
	if (l == &link->cset->tasks) {
3238 3239
		/* We reached the end of this task list - move on to
		 * the next cg_cgroup_link */
3240
		cgroup_advance_iter(cgrp, it);
3241 3242 3243 3244 3245 3246
	} else {
		it->task = l;
	}
	return res;
}

3247
void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it)
3248
	__releases(css_set_lock)
3249 3250 3251 3252
{
	read_unlock(&css_set_lock);
}

3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351
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
3352 3353
	cgroup_iter_start(scan->cgrp, &it);
	while ((p = cgroup_iter_next(scan->cgrp, &it))) {
3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385
		/*
		 * 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
3386
	cgroup_iter_end(scan->cgrp, &it);
3387 3388 3389

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3390
			struct task_struct *q = heap->ptrs[i];
3391
			if (i == 0) {
3392 3393
				latest_time = q->start_time;
				latest_task = q;
3394 3395
			}
			/* Process the task per the caller's callback */
3396 3397
			scan->process_task(q, scan);
			put_task_struct(q);
3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
		}
		/*
		 * 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;
}

3413 3414 3415 3416 3417
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
3418
	mutex_lock(&cgroup_mutex);
3419
	cgroup_attach_task(new_cgroup, task, false);
T
Tejun Heo 已提交
3420
	mutex_unlock(&cgroup_mutex);
3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431
}

/**
 * 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
3432
	scan.cgrp = from;
3433 3434 3435 3436 3437 3438 3439 3440
	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);
}

3441
/*
3442
 * Stuff for reading the 'tasks'/'procs' files.
3443 3444 3445 3446 3447 3448 3449 3450
 *
 * 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.
 *
 */

3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479
/* 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
3480
	struct rw_semaphore rwsem;
3481 3482
};

3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503
/*
 * 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);
}

3504
/*
3505
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3506
 * Returns the number of unique elements.
3507
 */
3508
static int pidlist_uniq(pid_t *list, int length)
3509
{
3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538
	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;
}

3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
/*
 * 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 */
3550
	struct pid_namespace *ns = task_active_pid_ns(current);
3551

3552
	/*
L
Li Zefan 已提交
3553
	 * We can't drop the pidlist_mutex before taking the l->rwsem in case
3554 3555 3556 3557 3558 3559 3560 3561
	 * 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
3562
			down_write(&l->rwsem);
3563 3564 3565 3566 3567
			mutex_unlock(&cgrp->pidlist_mutex);
			return l;
		}
	}
	/* entry not found; create a new one */
3568
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
3569 3570 3571 3572
	if (!l) {
		mutex_unlock(&cgrp->pidlist_mutex);
		return l;
	}
L
Li Zefan 已提交
3573 3574
	init_rwsem(&l->rwsem);
	down_write(&l->rwsem);
3575
	l->key.type = type;
3576
	l->key.ns = get_pid_ns(ns);
3577 3578 3579 3580 3581 3582
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	mutex_unlock(&cgrp->pidlist_mutex);
	return l;
}

3583 3584 3585
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3586 3587
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3588 3589 3590 3591
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3592 3593
	struct cgroup_iter it;
	struct task_struct *tsk;
3594 3595 3596 3597 3598 3599 3600 3601 3602
	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);
3603
	array = pidlist_allocate(length);
3604 3605 3606
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3607 3608
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
3609
		if (unlikely(n == length))
3610
			break;
3611
		/* get tgid or pid for procs or tasks file respectively */
3612 3613 3614 3615
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3616 3617
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3618
	}
3619
	cgroup_iter_end(cgrp, &it);
3620 3621 3622
	length = n;
	/* now sort & (if procs) strip out duplicates */
	sort(array, length, sizeof(pid_t), cmppid, NULL);
3623
	if (type == CGROUP_FILE_PROCS)
3624
		length = pidlist_uniq(array, length);
3625 3626
	l = cgroup_pidlist_find(cgrp, type);
	if (!l) {
3627
		pidlist_free(array);
3628
		return -ENOMEM;
3629
	}
3630
	/* store array, freeing old if necessary - lock already held */
3631
	pidlist_free(l->list);
3632 3633 3634
	l->list = array;
	l->length = length;
	l->use_count++;
L
Li Zefan 已提交
3635
	up_write(&l->rwsem);
3636
	*lp = l;
3637
	return 0;
3638 3639
}

B
Balbir Singh 已提交
3640
/**
L
Li Zefan 已提交
3641
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3642 3643 3644
 * @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
3645 3646 3647
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3648 3649 3650 3651
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3652
	struct cgroup *cgrp;
B
Balbir Singh 已提交
3653 3654
	struct cgroup_iter it;
	struct task_struct *tsk;
3655

B
Balbir Singh 已提交
3656
	/*
3657 3658
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3659
	 */
3660 3661
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3662 3663 3664
		 goto err;

	ret = 0;
3665
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3666

3667 3668
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
B
Balbir Singh 已提交
3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687
		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;
		}
	}
3688
	cgroup_iter_end(cgrp, &it);
B
Balbir Singh 已提交
3689 3690 3691 3692 3693

err:
	return ret;
}

3694

3695
/*
3696
 * seq_file methods for the tasks/procs files. The seq_file position is the
3697
 * next pid to display; the seq_file iterator is a pointer to the pid
3698
 * in the cgroup->l->list array.
3699
 */
3700

3701
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3702
{
3703 3704 3705 3706 3707 3708
	/*
	 * 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
	 */
3709
	struct cgroup_pidlist *l = s->private;
3710 3711 3712
	int index = 0, pid = *pos;
	int *iter;

L
Li Zefan 已提交
3713
	down_read(&l->rwsem);
3714
	if (pid) {
3715
		int end = l->length;
S
Stephen Rothwell 已提交
3716

3717 3718
		while (index < end) {
			int mid = (index + end) / 2;
3719
			if (l->list[mid] == pid) {
3720 3721
				index = mid;
				break;
3722
			} else if (l->list[mid] <= pid)
3723 3724 3725 3726 3727 3728
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3729
	if (index >= l->length)
3730 3731
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3732
	iter = l->list + index;
3733 3734 3735 3736
	*pos = *iter;
	return iter;
}

3737
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3738
{
3739
	struct cgroup_pidlist *l = s->private;
L
Li Zefan 已提交
3740
	up_read(&l->rwsem);
3741 3742
}

3743
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3744
{
3745 3746 3747
	struct cgroup_pidlist *l = s->private;
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760
	/*
	 * 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;
	}
}

3761
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3762 3763 3764
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3765

3766 3767 3768 3769 3770 3771 3772 3773 3774
/*
 * 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,
3775 3776
};

3777
static void cgroup_release_pid_array(struct cgroup_pidlist *l)
3778
{
3779 3780 3781 3782 3783 3784 3785
	/*
	 * 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
3786
	down_write(&l->rwsem);
3787 3788
	BUG_ON(!l->use_count);
	if (!--l->use_count) {
3789 3790 3791
		/* we're the last user if refcount is 0; remove and free */
		list_del(&l->links);
		mutex_unlock(&l->owner->pidlist_mutex);
3792
		pidlist_free(l->list);
3793
		put_pid_ns(l->key.ns);
L
Li Zefan 已提交
3794
		up_write(&l->rwsem);
3795 3796
		kfree(l);
		return;
3797
	}
3798
	mutex_unlock(&l->owner->pidlist_mutex);
L
Li Zefan 已提交
3799
	up_write(&l->rwsem);
3800 3801
}

3802
static int cgroup_pidlist_release(struct inode *inode, struct file *file)
3803
{
3804
	struct cgroup_pidlist *l;
3805 3806
	if (!(file->f_mode & FMODE_READ))
		return 0;
3807 3808 3809 3810 3811 3812
	/*
	 * 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);
3813 3814 3815
	return seq_release(inode, file);
}

3816
static const struct file_operations cgroup_pidlist_operations = {
3817 3818 3819
	.read = seq_read,
	.llseek = seq_lseek,
	.write = cgroup_file_write,
3820
	.release = cgroup_pidlist_release,
3821 3822
};

3823
/*
3824 3825 3826
 * 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.
3827
 */
3828
/* helper function for the two below it */
3829
static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
3830
{
3831
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
3832
	struct cgroup_pidlist *l;
3833
	int retval;
3834

3835
	/* Nothing to do for write-only files */
3836 3837 3838
	if (!(file->f_mode & FMODE_READ))
		return 0;

3839
	/* have the array populated */
3840
	retval = pidlist_array_load(cgrp, type, &l);
3841 3842 3843 3844
	if (retval)
		return retval;
	/* configure file information */
	file->f_op = &cgroup_pidlist_operations;
3845

3846
	retval = seq_open(file, &cgroup_pidlist_seq_operations);
3847
	if (retval) {
3848
		cgroup_release_pid_array(l);
3849
		return retval;
3850
	}
3851
	((struct seq_file *)file->private_data)->private = l;
3852 3853
	return 0;
}
3854 3855
static int cgroup_tasks_open(struct inode *unused, struct file *file)
{
3856
	return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
3857 3858 3859
}
static int cgroup_procs_open(struct inode *unused, struct file *file)
{
3860
	return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
3861
}
3862

3863
static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
3864 3865
					    struct cftype *cft)
{
3866
	return notify_on_release(cgrp);
3867 3868
}

3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880
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;
}

3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897
/*
 * 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);
}

3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908
/*
 * 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;

3909 3910
	remove_wait_queue(event->wqh, &event->wait);

3911 3912
	event->cft->unregister_event(cgrp, event->cft, event->eventfd);

3913 3914 3915
	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

3916 3917
	eventfd_ctx_put(event->eventfd);
	kfree(event);
3918
	cgroup_dput(cgrp);
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935
}

/*
 * 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) {
		/*
3936 3937 3938 3939 3940 3941 3942
		 * 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.
3943
		 */
3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
		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);
3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977
	}

	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)
{
3978
	struct cgroup_event *event;
3979
	struct cgroup *cgrp_cfile;
3980
	unsigned int efd, cfd;
3981 3982
	struct file *efile;
	struct file *cfile;
3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007
	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);
4008
		goto out_kfree;
4009 4010 4011 4012 4013
	}

	event->eventfd = eventfd_ctx_fileget(efile);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
4014
		goto out_put_efile;
4015 4016 4017 4018 4019
	}

	cfile = fget(cfd);
	if (!cfile) {
		ret = -EBADF;
4020
		goto out_put_eventfd;
4021 4022 4023
	}

	/* the process need read permission on control file */
A
Al Viro 已提交
4024
	/* AV: shouldn't we check that it's been opened for read instead? */
A
Al Viro 已提交
4025
	ret = inode_permission(file_inode(cfile), MAY_READ);
4026
	if (ret < 0)
4027
		goto out_put_cfile;
4028 4029 4030 4031

	event->cft = __file_cft(cfile);
	if (IS_ERR(event->cft)) {
		ret = PTR_ERR(event->cft);
4032
		goto out_put_cfile;
4033 4034
	}

4035 4036 4037 4038 4039 4040 4041
	/*
	 * 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;
4042
		goto out_put_cfile;
4043 4044
	}

4045 4046
	if (!event->cft->register_event || !event->cft->unregister_event) {
		ret = -EINVAL;
4047
		goto out_put_cfile;
4048 4049 4050 4051 4052
	}

	ret = event->cft->register_event(cgrp, event->cft,
			event->eventfd, buffer);
	if (ret)
4053
		goto out_put_cfile;
4054

4055
	efile->f_op->poll(efile, &event->pt);
4056

4057 4058 4059 4060 4061 4062 4063
	/*
	 * 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);

4064 4065 4066 4067 4068 4069 4070 4071 4072
	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;

4073 4074 4075 4076 4077 4078 4079
out_put_cfile:
	fput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fput(efile);
out_kfree:
4080 4081 4082 4083 4084
	kfree(event);

	return ret;
}

4085 4086 4087
static u64 cgroup_clone_children_read(struct cgroup *cgrp,
				    struct cftype *cft)
{
4088
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4089 4090 4091 4092 4093 4094 4095
}

static int cgroup_clone_children_write(struct cgroup *cgrp,
				     struct cftype *cft,
				     u64 val)
{
	if (val)
4096
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4097
	else
4098
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4099 4100 4101
	return 0;
}

4102
static struct cftype cgroup_base_files[] = {
4103
	{
4104
		.name = "cgroup.procs",
4105
		.open = cgroup_procs_open,
B
Ben Blum 已提交
4106
		.write_u64 = cgroup_procs_write,
4107
		.release = cgroup_pidlist_release,
B
Ben Blum 已提交
4108
		.mode = S_IRUGO | S_IWUSR,
4109
	},
4110
	{
4111
		.name = "cgroup.event_control",
4112 4113 4114
		.write_string = cgroup_write_event_control,
		.mode = S_IWUGO,
	},
4115 4116
	{
		.name = "cgroup.clone_children",
4117
		.flags = CFTYPE_INSANE,
4118 4119 4120
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
4121 4122 4123 4124 4125
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
		.read_seq_string = cgroup_sane_behavior_show,
	},
4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145

	/*
	 * 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,
	},
4146 4147
	{
		.name = "release_agent",
4148
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
4149 4150 4151 4152
		.read_seq_string = cgroup_release_agent_show,
		.write_string = cgroup_release_agent_write,
		.max_write_len = PATH_MAX,
	},
T
Tejun Heo 已提交
4153
	{ }	/* terminate */
4154 4155
};

4156
/**
4157
 * cgroup_populate_dir - create subsys files in a cgroup directory
4158 4159
 * @cgrp: target cgroup
 * @subsys_mask: mask of the subsystem ids whose files should be added
4160 4161
 *
 * On failure, no file is added.
4162
 */
4163
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned long subsys_mask)
4164 4165
{
	struct cgroup_subsys *ss;
4166
	int i, ret = 0;
4167

4168
	/* process cftsets of each subsystem */
4169
	for_each_subsys(ss, i) {
4170
		struct cftype_set *set;
4171 4172

		if (!test_bit(i, &subsys_mask))
4173
			continue;
4174

4175
		list_for_each_entry(set, &ss->cftsets, node) {
4176
			ret = cgroup_addrm_files(cgrp, set->cfts, true);
4177 4178 4179
			if (ret < 0)
				goto err;
		}
4180
	}
4181

K
KAMEZAWA Hiroyuki 已提交
4182
	/* This cgroup is ready now */
4183
	for_each_root_subsys(cgrp->root, ss) {
K
KAMEZAWA Hiroyuki 已提交
4184
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4185 4186
		struct css_id *id = rcu_dereference_protected(css->id, true);

K
KAMEZAWA Hiroyuki 已提交
4187 4188 4189 4190 4191
		/*
		 * Update id->css pointer and make this css visible from
		 * CSS ID functions. This pointer will be dereferened
		 * from RCU-read-side without locks.
		 */
4192 4193
		if (id)
			rcu_assign_pointer(id->css, css);
K
KAMEZAWA Hiroyuki 已提交
4194
	}
4195 4196

	return 0;
4197 4198 4199
err:
	cgroup_clear_dir(cgrp, subsys_mask);
	return ret;
4200 4201
}

4202 4203 4204 4205 4206
static void css_dput_fn(struct work_struct *work)
{
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, dput_work);

4207
	cgroup_dput(css->cgroup);
4208 4209
}

4210 4211 4212 4213 4214 4215 4216 4217
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);
}

4218 4219
static void init_cgroup_css(struct cgroup_subsys_state *css,
			       struct cgroup_subsys *ss,
4220
			       struct cgroup *cgrp)
4221
{
4222
	css->cgroup = cgrp;
4223
	css->ss = ss;
4224
	css->flags = 0;
K
KAMEZAWA Hiroyuki 已提交
4225
	css->id = NULL;
4226
	if (cgrp == cgroup_dummy_top)
4227
		css->flags |= CSS_ROOT;
4228 4229
	BUG_ON(cgrp->subsys[ss->subsys_id]);
	cgrp->subsys[ss->subsys_id] = css;
4230 4231

	/*
4232 4233 4234 4235
	 * 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().
4236 4237
	 */
	INIT_WORK(&css->dput_work, css_dput_fn);
4238 4239
}

4240
/* invoke ->css_online() on a new CSS and mark it online if successful */
T
Tejun Heo 已提交
4241
static int online_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
4242
{
4243
	struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
T
Tejun Heo 已提交
4244 4245
	int ret = 0;

4246 4247
	lockdep_assert_held(&cgroup_mutex);

4248
	if (ss->css_online)
4249
		ret = ss->css_online(css);
T
Tejun Heo 已提交
4250
	if (!ret)
4251
		css->flags |= CSS_ONLINE;
T
Tejun Heo 已提交
4252
	return ret;
4253 4254
}

4255
/* if the CSS is online, invoke ->css_offline() on it and mark it offline */
4256 4257 4258 4259 4260 4261 4262 4263 4264
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;

4265
	if (ss->css_offline)
4266
		ss->css_offline(css);
4267

4268
	css->flags &= ~CSS_ONLINE;
4269 4270
}

4271
/*
L
Li Zefan 已提交
4272 4273 4274 4275
 * 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
4276
 *
L
Li Zefan 已提交
4277
 * Must be called with the mutex on the parent inode held
4278 4279
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
4280
			     umode_t mode)
4281
{
4282
	struct cgroup *cgrp;
4283
	struct cgroup_name *name;
4284 4285 4286 4287 4288
	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
4289
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4290 4291
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4292 4293
		return -ENOMEM;

4294 4295 4296 4297 4298
	name = cgroup_alloc_name(dentry);
	if (!name)
		goto err_free_cgrp;
	rcu_assign_pointer(cgrp->name, name);

4299 4300 4301 4302 4303
	/*
	 * 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
4304
	if (cgrp->id < 0)
4305
		goto err_free_name;
T
Tejun Heo 已提交
4306

4307 4308 4309 4310 4311 4312 4313 4314 4315
	/*
	 * 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
4316
		goto err_free_id;
4317 4318
	}

4319 4320 4321 4322 4323 4324 4325
	/* 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);

4326
	init_cgroup_housekeeping(cgrp);
4327

4328 4329 4330
	dentry->d_fsdata = cgrp;
	cgrp->dentry = dentry;

4331 4332
	cgrp->parent = parent;
	cgrp->root = parent->root;
4333

4334 4335 4336
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4337 4338
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4339

4340
	for_each_root_subsys(root, ss) {
4341
		struct cgroup_subsys_state *css;
4342

4343
		css = ss->css_alloc(parent->subsys[ss->subsys_id]);
4344 4345
		if (IS_ERR(css)) {
			err = PTR_ERR(css);
4346
			goto err_free_all;
4347
		}
4348 4349

		err = percpu_ref_init(&css->refcnt, css_release);
4350
		if (err) {
4351
			ss->css_free(css);
4352
			goto err_free_all;
4353
		}
4354

4355
		init_cgroup_css(css, ss, cgrp);
4356

4357 4358 4359
		if (ss->use_id) {
			err = alloc_css_id(ss, parent, cgrp);
			if (err)
4360
				goto err_free_all;
4361
		}
4362 4363
	}

4364 4365 4366 4367 4368
	/*
	 * 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
4369
	err = cgroup_create_file(dentry, S_IFDIR | mode, sb);
4370
	if (err < 0)
4371
		goto err_free_all;
4372
	lockdep_assert_held(&dentry->d_inode->i_mutex);
4373

4374
	cgrp->serial_nr = cgroup_serial_nr_next++;
4375

4376 4377 4378
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
	root->number_of_cgroups++;
T
Tejun Heo 已提交
4379

T
Tejun Heo 已提交
4380
	/* each css holds a ref to the cgroup's dentry */
4381
	for_each_root_subsys(root, ss)
4382
		dget(dentry);
4383

4384 4385 4386
	/* hold a ref to the parent's dentry */
	dget(parent->dentry);

T
Tejun Heo 已提交
4387
	/* creation succeeded, notify subsystems */
4388
	for_each_root_subsys(root, ss) {
T
Tejun Heo 已提交
4389 4390 4391
		err = online_css(ss, cgrp);
		if (err)
			goto err_destroy;
4392 4393 4394 4395 4396 4397 4398 4399 4400

		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;
		}
4401 4402
	}

4403 4404
	idr_replace(&root->cgroup_idr, cgrp, cgrp->id);

4405
	err = cgroup_addrm_files(cgrp, cgroup_base_files, true);
4406 4407 4408 4409
	if (err)
		goto err_destroy;

	err = cgroup_populate_dir(cgrp, root->subsys_mask);
4410 4411
	if (err)
		goto err_destroy;
4412 4413

	mutex_unlock(&cgroup_mutex);
4414
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4415 4416 4417

	return 0;

4418
err_free_all:
4419
	for_each_root_subsys(root, ss) {
4420 4421 4422 4423
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];

		if (css) {
			percpu_ref_cancel_init(&css->refcnt);
4424
			ss->css_free(css);
4425
		}
4426 4427 4428 4429
	}
	mutex_unlock(&cgroup_mutex);
	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);
T
Tejun Heo 已提交
4430
err_free_id:
4431
	idr_remove(&root->cgroup_idr, cgrp->id);
4432 4433
err_free_name:
	kfree(rcu_dereference_raw(cgrp->name));
4434
err_free_cgrp:
4435
	kfree(cgrp);
4436
	return err;
4437 4438 4439 4440 4441 4442

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&dentry->d_inode->i_mutex);
	return err;
4443 4444
}

4445
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4446 4447 4448 4449 4450 4451 4452
{
	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);
}

4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494
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.
 */
4495 4496
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4497
{
4498
	struct dentry *d = cgrp->dentry;
4499
	struct cgroup_event *event, *tmp;
4500
	struct cgroup_subsys *ss;
4501
	bool empty;
4502

4503 4504 4505
	lockdep_assert_held(&d->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

4506
	/*
T
Tejun Heo 已提交
4507 4508
	 * css_set_lock synchronizes access to ->cset_links and prevents
	 * @cgrp from being removed while __put_css_set() is in progress.
4509 4510
	 */
	read_lock(&css_set_lock);
T
Tejun Heo 已提交
4511
	empty = list_empty(&cgrp->cset_links) && list_empty(&cgrp->children);
4512 4513
	read_unlock(&css_set_lock);
	if (!empty)
4514
		return -EBUSY;
L
Li Zefan 已提交
4515

4516
	/*
4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529
	 * 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.
4530
	 */
4531
	atomic_set(&cgrp->css_kill_cnt, 1);
4532
	for_each_root_subsys(cgrp->root, ss) {
4533
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4534

4535 4536 4537 4538 4539 4540 4541 4542
		/*
		 * 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);
4543
	}
4544
	cgroup_css_killed(cgrp);
4545 4546 4547 4548 4549 4550 4551 4552

	/*
	 * 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.
	 */
4553
	set_bit(CGRP_DEAD, &cgrp->flags);
4554

4555 4556 4557 4558 4559 4560 4561
	/* 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);

	/*
4562 4563
	 * Clear and remove @cgrp directory.  The removal puts the base ref
	 * but we aren't quite done with @cgrp yet, so hold onto it.
4564
	 */
4565
	cgroup_clear_dir(cgrp, cgrp->root->subsys_mask);
4566
	cgroup_addrm_files(cgrp, cgroup_base_files, false);
4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581
	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);

4582 4583 4584
	return 0;
};

4585 4586 4587 4588 4589 4590 4591 4592 4593 4594
/**
 * 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().
 */
4595 4596 4597 4598 4599 4600 4601 4602 4603
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);

4604 4605 4606 4607
	/*
	 * css_tryget() is guaranteed to fail now.  Tell subsystems to
	 * initate destruction.
	 */
4608
	for_each_root_subsys(cgrp->root, ss)
4609
		offline_css(ss, cgrp);
4610 4611

	/*
4612 4613 4614 4615 4616
	 * 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.
4617
	 */
4618
	for_each_root_subsys(cgrp->root, ss)
T
Tejun Heo 已提交
4619
		css_put(cgrp->subsys[ss->subsys_id]);
4620

4621
	/* delete this cgroup from parent->children */
4622
	list_del_rcu(&cgrp->sibling);
4623

4624 4625 4626 4627 4628 4629 4630 4631
	/*
	 * 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;

4632 4633
	dput(d);

4634
	set_bit(CGRP_RELEASABLE, &parent->flags);
4635 4636
	check_for_release(parent);

4637
	mutex_unlock(&cgroup_mutex);
4638 4639
}

4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650
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;
}

4651 4652 4653 4654 4655 4656 4657 4658 4659
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) {
4660 4661 4662 4663 4664
		struct cftype *cft;

		for (cft = ss->base_cftypes; cft->name[0] != '\0'; cft++)
			cft->ss = ss;

4665 4666 4667 4668 4669
		ss->base_cftset.cfts = ss->base_cftypes;
		list_add_tail(&ss->base_cftset.node, &ss->cftsets);
	}
}

4670
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4671 4672
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4673 4674

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4675

4676 4677
	mutex_lock(&cgroup_mutex);

4678 4679 4680
	/* init base cftset */
	cgroup_init_cftsets(ss);

4681
	/* Create the top cgroup state for this subsystem */
4682 4683
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
4684
	css = ss->css_alloc(cgroup_dummy_top->subsys[ss->subsys_id]);
4685 4686
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4687
	init_cgroup_css(css, ss, cgroup_dummy_top);
4688

L
Li Zefan 已提交
4689
	/* Update the init_css_set to contain a subsys
4690
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4691 4692
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
4693
	init_css_set.subsys[ss->subsys_id] = css;
4694 4695 4696

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

L
Li Zefan 已提交
4697 4698 4699 4700 4701
	/* 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));

4702
	BUG_ON(online_css(ss, cgroup_dummy_top));
4703

4704 4705
	mutex_unlock(&cgroup_mutex);

4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
	/* 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
4716
 * subsystem is built as a module, it will be assigned a new subsys_id and set
4717 4718 4719 4720 4721 4722
 * 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;
4723
	int i, ret;
4724
	struct hlist_node *tmp;
4725
	struct css_set *cset;
4726
	unsigned long key;
4727 4728 4729

	/* check name and function validity */
	if (ss->name == NULL || strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN ||
4730
	    ss->css_alloc == NULL || ss->css_free == NULL)
4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746
		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) {
4747
		/* a sanity check */
4748
		BUG_ON(cgroup_subsys[ss->subsys_id] != ss);
4749 4750 4751
		return 0;
	}

4752 4753 4754
	/* init base cftset */
	cgroup_init_cftsets(ss);

4755
	mutex_lock(&cgroup_mutex);
4756
	cgroup_subsys[ss->subsys_id] = ss;
4757 4758

	/*
4759
	 * no ss->css_alloc seems to need anything important in the ss
4760
	 * struct, so this can happen first (i.e. before the dummy root
4761
	 * attachment).
4762
	 */
4763
	css = ss->css_alloc(cgroup_dummy_top->subsys[ss->subsys_id]);
4764
	if (IS_ERR(css)) {
4765 4766
		/* failure case - need to deassign the cgroup_subsys[] slot. */
		cgroup_subsys[ss->subsys_id] = NULL;
4767 4768 4769 4770
		mutex_unlock(&cgroup_mutex);
		return PTR_ERR(css);
	}

4771 4772
	list_add(&ss->sibling, &cgroup_dummy_root.subsys_list);
	ss->root = &cgroup_dummy_root;
4773 4774

	/* our new subsystem will be attached to the dummy hierarchy. */
4775
	init_cgroup_css(css, ss, cgroup_dummy_top);
4776 4777
	/* init_idr must be after init_cgroup_css because it sets css->id. */
	if (ss->use_id) {
4778 4779 4780
		ret = cgroup_init_idr(ss, css);
		if (ret)
			goto err_unload;
4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791
	}

	/*
	 * 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);
4792
	hash_for_each_safe(css_set_table, i, tmp, cset, hlist) {
4793
		/* skip entries that we already rehashed */
4794
		if (cset->subsys[ss->subsys_id])
4795 4796
			continue;
		/* remove existing entry */
4797
		hash_del(&cset->hlist);
4798
		/* set new value */
4799
		cset->subsys[ss->subsys_id] = css;
4800
		/* recompute hash and restore entry */
4801 4802
		key = css_set_hash(cset->subsys);
		hash_add(css_set_table, &cset->hlist, key);
4803 4804 4805
	}
	write_unlock(&css_set_lock);

4806
	ret = online_css(ss, cgroup_dummy_top);
T
Tejun Heo 已提交
4807 4808
	if (ret)
		goto err_unload;
4809

4810 4811 4812
	/* success! */
	mutex_unlock(&cgroup_mutex);
	return 0;
4813 4814 4815 4816 4817 4818

err_unload:
	mutex_unlock(&cgroup_mutex);
	/* @ss can't be mounted here as try_module_get() would fail */
	cgroup_unload_subsys(ss);
	return ret;
4819
}
4820
EXPORT_SYMBOL_GPL(cgroup_load_subsys);
4821

B
Ben Blum 已提交
4822 4823 4824 4825 4826 4827 4828 4829 4830 4831
/**
 * 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)
{
4832
	struct cgrp_cset_link *link;
B
Ben Blum 已提交
4833 4834 4835 4836 4837

	BUG_ON(ss->module == NULL);

	/*
	 * we shouldn't be called if the subsystem is in use, and the use of
4838
	 * try_module_get() in rebind_subsystems() should ensure that it
B
Ben Blum 已提交
4839 4840
	 * doesn't start being used while we're killing it off.
	 */
4841
	BUG_ON(ss->root != &cgroup_dummy_root);
B
Ben Blum 已提交
4842 4843

	mutex_lock(&cgroup_mutex);
4844

4845
	offline_css(ss, cgroup_dummy_top);
4846

T
Tejun Heo 已提交
4847
	if (ss->use_id)
4848 4849
		idr_destroy(&ss->idr);

B
Ben Blum 已提交
4850
	/* deassign the subsys_id */
4851
	cgroup_subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4852

4853
	/* remove subsystem from the dummy root's list of subsystems */
4854
	list_del_init(&ss->sibling);
B
Ben Blum 已提交
4855 4856

	/*
4857 4858 4859
	 * 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
4860 4861
	 */
	write_lock(&css_set_lock);
4862
	list_for_each_entry(link, &cgroup_dummy_top->cset_links, cset_link) {
4863
		struct css_set *cset = link->cset;
4864
		unsigned long key;
B
Ben Blum 已提交
4865

4866 4867 4868 4869
		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
4870 4871 4872 4873
	}
	write_unlock(&css_set_lock);

	/*
4874 4875 4876 4877
	 * 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
4878
	 */
4879
	ss->css_free(cgroup_dummy_top->subsys[ss->subsys_id]);
4880
	cgroup_dummy_top->subsys[ss->subsys_id] = NULL;
B
Ben Blum 已提交
4881 4882 4883 4884 4885

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

4886
/**
L
Li Zefan 已提交
4887 4888 4889 4890
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4891 4892 4893
 */
int __init cgroup_init_early(void)
{
4894
	struct cgroup_subsys *ss;
4895
	int i;
4896

4897
	atomic_set(&init_css_set.refcount, 1);
4898
	INIT_LIST_HEAD(&init_css_set.cgrp_links);
4899
	INIT_LIST_HEAD(&init_css_set.tasks);
4900
	INIT_HLIST_NODE(&init_css_set.hlist);
4901
	css_set_count = 1;
4902 4903
	init_cgroup_root(&cgroup_dummy_root);
	cgroup_root_count = 1;
4904
	RCU_INIT_POINTER(init_task.cgroups, &init_css_set);
4905

4906
	init_cgrp_cset_link.cset = &init_css_set;
4907 4908
	init_cgrp_cset_link.cgrp = cgroup_dummy_top;
	list_add(&init_cgrp_cset_link.cset_link, &cgroup_dummy_top->cset_links);
4909
	list_add(&init_cgrp_cset_link.cgrp_link, &init_css_set.cgrp_links);
4910

4911 4912
	/* at bootup time, we don't worry about modular subsystems */
	for_each_builtin_subsys(ss, i) {
4913 4914
		BUG_ON(!ss->name);
		BUG_ON(strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN);
4915 4916
		BUG_ON(!ss->css_alloc);
		BUG_ON(!ss->css_free);
4917
		if (ss->subsys_id != i) {
D
Diego Calleja 已提交
4918
			printk(KERN_ERR "cgroup: Subsys %s id == %d\n",
4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929
			       ss->name, ss->subsys_id);
			BUG();
		}

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

/**
L
Li Zefan 已提交
4930 4931 4932 4933
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4934 4935 4936
 */
int __init cgroup_init(void)
{
4937
	struct cgroup_subsys *ss;
4938
	unsigned long key;
4939
	int i, err;
4940 4941 4942 4943

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4944

4945
	for_each_builtin_subsys(ss, i) {
4946 4947
		if (!ss->early_init)
			cgroup_init_subsys(ss);
K
KAMEZAWA Hiroyuki 已提交
4948
		if (ss->use_id)
4949
			cgroup_init_idr(ss, init_css_set.subsys[ss->subsys_id]);
4950 4951
	}

4952
	/* allocate id for the dummy hierarchy */
T
Tejun Heo 已提交
4953 4954 4955
	mutex_lock(&cgroup_mutex);
	mutex_lock(&cgroup_root_mutex);

4956 4957 4958 4959
	/* 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);

4960
	BUG_ON(cgroup_init_root_id(&cgroup_dummy_root, 0, 1));
4961

4962 4963 4964 4965
	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
4966 4967 4968
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);

4969 4970 4971 4972 4973 4974
	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4975
	err = register_filesystem(&cgroup_fs_type);
4976 4977
	if (err < 0) {
		kobject_put(cgroup_kobj);
4978
		goto out;
4979
	}
4980

L
Li Zefan 已提交
4981
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4982

4983
out:
4984 4985 4986
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4987 4988
	return err;
}
4989

4990 4991 4992 4993 4994 4995
/*
 * 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,
4996
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4997 4998 4999 5000 5001 5002
 *    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 */
5003
int proc_cgroup_show(struct seq_file *m, void *v)
5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025
{
	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);

5026
	for_each_active_root(root) {
5027
		struct cgroup_subsys *ss;
5028
		struct cgroup *cgrp;
5029 5030
		int count = 0;

5031
		seq_printf(m, "%d:", root->hierarchy_id);
5032
		for_each_root_subsys(root, ss)
5033
			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
5034 5035 5036
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
5037
		seq_putc(m, ':');
5038
		cgrp = task_cgroup_from_root(tsk, root);
5039
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057
		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)
{
5058
	struct cgroup_subsys *ss;
5059 5060
	int i;

5061
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
5062 5063 5064 5065 5066
	/*
	 * 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.
	 */
5067
	mutex_lock(&cgroup_mutex);
5068 5069

	for_each_subsys(ss, i)
5070 5071
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
5072
			   ss->root->number_of_cgroups, !ss->disabled);
5073

5074 5075 5076 5077 5078 5079
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
5080
	return single_open(file, proc_cgroupstats_show, NULL);
5081 5082
}

5083
static const struct file_operations proc_cgroupstats_operations = {
5084 5085 5086 5087 5088 5089
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

5090 5091
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
5092
 * @child: pointer to task_struct of forking parent process.
5093 5094 5095 5096 5097
 *
 * 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
5098 5099 5100 5101
 * 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.
5102 5103 5104 5105 5106 5107
 *
 * 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)
{
5108
	task_lock(current);
5109
	get_css_set(task_css_set(current));
5110
	child->cgroups = current->cgroups;
5111
	task_unlock(current);
5112
	INIT_LIST_HEAD(&child->cg_list);
5113 5114
}

5115
/**
L
Li Zefan 已提交
5116 5117 5118
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
5119 5120 5121 5122 5123
 * 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
5124
 */
5125 5126
void cgroup_post_fork(struct task_struct *child)
{
5127
	struct cgroup_subsys *ss;
5128 5129
	int i;

5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140
	/*
	 * 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.
	 */
5141 5142
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
5143 5144
		task_lock(child);
		if (list_empty(&child->cg_list))
5145
			list_add(&child->cg_list, &task_css_set(child)->tasks);
5146
		task_unlock(child);
5147 5148
		write_unlock(&css_set_lock);
	}
5149 5150 5151 5152 5153 5154 5155

	/*
	 * 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) {
5156 5157 5158 5159 5160 5161 5162 5163
		/*
		 * 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.
		 */
5164
		for_each_builtin_subsys(ss, i)
5165 5166 5167
			if (ss->fork)
				ss->fork(child);
	}
5168
}
5169

5170 5171 5172
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
5173
 * @run_callback: run exit callbacks?
5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201
 *
 * 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,
5202 5203
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
5204 5205 5206
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
5207
	struct cgroup_subsys *ss;
5208
	struct css_set *cset;
5209
	int i;
5210 5211 5212 5213 5214 5215 5216 5217 5218

	/*
	 * 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))
5219
			list_del_init(&tsk->cg_list);
5220 5221 5222
		write_unlock(&css_set_lock);
	}

5223 5224
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
5225 5226
	cset = task_css_set(tsk);
	RCU_INIT_POINTER(tsk->cgroups, &init_css_set);
5227 5228

	if (run_callbacks && need_forkexit_callback) {
5229 5230 5231 5232
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, see cgroup_post_fork() for details.
		 */
5233
		for_each_builtin_subsys(ss, i) {
5234
			if (ss->exit) {
5235 5236
				struct cgroup_subsys_state *old_css = cset->subsys[i];
				struct cgroup_subsys_state *css = task_css(tsk, i);
5237

5238
				ss->exit(css, old_css, tsk);
5239 5240 5241
			}
		}
	}
5242
	task_unlock(tsk);
5243

5244
	put_css_set_taskexit(cset);
5245
}
5246

5247
static void check_for_release(struct cgroup *cgrp)
5248
{
5249
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
5250
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
5251 5252
		/*
		 * Control Group is currently removeable. If it's not
5253
		 * already queued for a userspace notification, queue
5254 5255
		 * it now
		 */
5256
		int need_schedule_work = 0;
5257

5258
		raw_spin_lock(&release_list_lock);
5259
		if (!cgroup_is_dead(cgrp) &&
5260 5261
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
5262 5263
			need_schedule_work = 1;
		}
5264
		raw_spin_unlock(&release_list_lock);
5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296
		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);
5297
	raw_spin_lock(&release_list_lock);
5298 5299 5300
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
5301
		char *pathbuf = NULL, *agentbuf = NULL;
5302
		struct cgroup *cgrp = list_entry(release_list.next,
5303 5304
						    struct cgroup,
						    release_list);
5305
		list_del_init(&cgrp->release_list);
5306
		raw_spin_unlock(&release_list_lock);
5307
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
5308 5309 5310 5311 5312 5313 5314
		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;
5315 5316

		i = 0;
5317 5318
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332
		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);
5333 5334 5335
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5336
		raw_spin_lock(&release_list_lock);
5337
	}
5338
	raw_spin_unlock(&release_list_lock);
5339 5340
	mutex_unlock(&cgroup_mutex);
}
5341 5342 5343

static int __init cgroup_disable(char *str)
{
5344
	struct cgroup_subsys *ss;
5345
	char *token;
5346
	int i;
5347 5348 5349 5350

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5351

5352 5353 5354 5355 5356
		/*
		 * 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) {
5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367
			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
5368 5369 5370 5371 5372

/*
 * Functons for CSS ID.
 */

5373
/* to get ID other than 0, this should be called when !cgroup_is_dead() */
K
KAMEZAWA Hiroyuki 已提交
5374 5375
unsigned short css_id(struct cgroup_subsys_state *css)
{
5376 5377 5378 5379 5380 5381 5382
	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.
	 */
5383
	cssid = rcu_dereference_raw(css->id);
K
KAMEZAWA Hiroyuki 已提交
5384 5385 5386 5387 5388

	if (cssid)
		return cssid->id;
	return 0;
}
B
Ben Blum 已提交
5389
EXPORT_SYMBOL_GPL(css_id);
K
KAMEZAWA Hiroyuki 已提交
5390

5391 5392 5393 5394 5395 5396
/**
 *  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
5397
 * this function reads css->id, the caller must hold rcu_read_lock().
5398 5399 5400 5401 5402 5403
 * 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
5404
bool css_is_ancestor(struct cgroup_subsys_state *child,
5405
		    const struct cgroup_subsys_state *root)
K
KAMEZAWA Hiroyuki 已提交
5406
{
5407 5408
	struct css_id *child_id;
	struct css_id *root_id;
K
KAMEZAWA Hiroyuki 已提交
5409

5410
	child_id  = rcu_dereference(child->id);
5411 5412
	if (!child_id)
		return false;
5413
	root_id = rcu_dereference(root->id);
5414 5415 5416 5417 5418 5419 5420
	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
5421 5422 5423 5424
}

void free_css_id(struct cgroup_subsys *ss, struct cgroup_subsys_state *css)
{
5425 5426
	struct css_id *id = rcu_dereference_protected(css->id, true);

K
KAMEZAWA Hiroyuki 已提交
5427 5428 5429 5430 5431 5432 5433 5434
	/* 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);
5435
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5436
	idr_remove(&ss->idr, id->id);
5437
	spin_unlock(&ss->id_lock);
5438
	kfree_rcu(id, rcu_head);
K
KAMEZAWA Hiroyuki 已提交
5439
}
B
Ben Blum 已提交
5440
EXPORT_SYMBOL_GPL(free_css_id);
K
KAMEZAWA Hiroyuki 已提交
5441 5442 5443 5444 5445 5446 5447 5448 5449

/*
 * 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
5450
	int ret, size;
K
KAMEZAWA Hiroyuki 已提交
5451 5452 5453 5454 5455 5456 5457

	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
5458 5459

	idr_preload(GFP_KERNEL);
5460
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5461
	/* Don't use 0. allocates an ID of 1-65535 */
T
Tejun Heo 已提交
5462
	ret = idr_alloc(&ss->idr, newid, 1, CSS_ID_MAX + 1, GFP_NOWAIT);
5463
	spin_unlock(&ss->id_lock);
T
Tejun Heo 已提交
5464
	idr_preload_end();
K
KAMEZAWA Hiroyuki 已提交
5465 5466

	/* Returns error when there are no free spaces for new ID.*/
T
Tejun Heo 已提交
5467
	if (ret < 0)
K
KAMEZAWA Hiroyuki 已提交
5468 5469
		goto err_out;

T
Tejun Heo 已提交
5470
	newid->id = ret;
K
KAMEZAWA Hiroyuki 已提交
5471 5472 5473 5474
	newid->depth = depth;
	return newid;
err_out:
	kfree(newid);
T
Tejun Heo 已提交
5475
	return ERR_PTR(ret);
K
KAMEZAWA Hiroyuki 已提交
5476 5477 5478

}

5479 5480
static int __init_or_module cgroup_init_idr(struct cgroup_subsys *ss,
					    struct cgroup_subsys_state *rootcss)
K
KAMEZAWA Hiroyuki 已提交
5481 5482 5483
{
	struct css_id *newid;

5484
	spin_lock_init(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5485 5486 5487 5488 5489 5490 5491
	idr_init(&ss->idr);

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

	newid->stack[0] = newid->id;
5492 5493
	RCU_INIT_POINTER(newid->css, rootcss);
	RCU_INIT_POINTER(rootcss->id, newid);
K
KAMEZAWA Hiroyuki 已提交
5494 5495 5496 5497 5498 5499 5500 5501
	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;
5502
	struct css_id *child_id, *parent_id;
K
KAMEZAWA Hiroyuki 已提交
5503 5504 5505 5506

	subsys_id = ss->subsys_id;
	parent_css = parent->subsys[subsys_id];
	child_css = child->subsys[subsys_id];
5507
	parent_id = rcu_dereference_protected(parent_css->id, true);
5508
	depth = parent_id->depth + 1;
K
KAMEZAWA Hiroyuki 已提交
5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545

	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
5546
EXPORT_SYMBOL_GPL(css_lookup);
K
KAMEZAWA Hiroyuki 已提交
5547

S
Stephane Eranian 已提交
5548 5549 5550 5551 5552 5553 5554 5555 5556
/*
 * 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
5557
	inode = file_inode(f);
S
Stephane Eranian 已提交
5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
	/* 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);
}

5571
#ifdef CONFIG_CGROUP_DEBUG
5572 5573
static struct cgroup_subsys_state *
debug_css_alloc(struct cgroup_subsys_state *parent_css)
5574 5575 5576 5577 5578 5579 5580 5581 5582
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

5583
static void debug_css_free(struct cgroup_subsys_state *css)
5584
{
5585
	kfree(css);
5586 5587
}

L
Li Zefan 已提交
5588
static u64 debug_taskcount_read(struct cgroup *cgrp, struct cftype *cft)
5589
{
L
Li Zefan 已提交
5590
	return cgroup_task_count(cgrp);
5591 5592
}

L
Li Zefan 已提交
5593
static u64 current_css_set_read(struct cgroup *cgrp, struct cftype *cft)
5594 5595 5596 5597
{
	return (u64)(unsigned long)current->cgroups;
}

L
Li Zefan 已提交
5598 5599
static u64 current_css_set_refcount_read(struct cgroup *cgrp,
					 struct cftype *cft)
5600 5601 5602 5603
{
	u64 count;

	rcu_read_lock();
5604
	count = atomic_read(&task_css_set(current)->refcount);
5605 5606 5607 5608
	rcu_read_unlock();
	return count;
}

L
Li Zefan 已提交
5609
static int current_css_set_cg_links_read(struct cgroup *cgrp,
5610 5611 5612
					 struct cftype *cft,
					 struct seq_file *seq)
{
5613
	struct cgrp_cset_link *link;
5614
	struct css_set *cset;
5615 5616 5617

	read_lock(&css_set_lock);
	rcu_read_lock();
5618
	cset = rcu_dereference(current->cgroups);
5619
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5620 5621 5622 5623 5624 5625 5626
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5627 5628
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5629 5630 5631 5632 5633 5634 5635
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
L
Li Zefan 已提交
5636
static int cgroup_css_links_read(struct cgroup *cgrp,
5637 5638 5639
				 struct cftype *cft,
				 struct seq_file *seq)
{
5640
	struct cgrp_cset_link *link;
5641 5642

	read_lock(&css_set_lock);
L
Li Zefan 已提交
5643
	list_for_each_entry(link, &cgrp->cset_links, cset_link) {
5644
		struct css_set *cset = link->cset;
5645 5646
		struct task_struct *task;
		int count = 0;
5647 5648
		seq_printf(seq, "css_set %p\n", cset);
		list_for_each_entry(task, &cset->tasks, cg_list) {
5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661
			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;
}

5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682
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,
	},

5683 5684 5685 5686 5687 5688 5689 5690 5691 5692
	{
		.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,
	},

5693 5694 5695 5696 5697
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5698 5699
	{ }	/* terminate */
};
5700 5701 5702

struct cgroup_subsys debug_subsys = {
	.name = "debug",
5703 5704
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5705
	.subsys_id = debug_subsys_id,
5706
	.base_cftypes = debug_files,
5707 5708
};
#endif /* CONFIG_CGROUP_DEBUG */