cgroup.c 144.6 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
 *  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.
 */

29 30
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

31
#include <linux/cgroup.h>
32
#include <linux/cred.h>
33
#include <linux/ctype.h>
34
#include <linux/errno.h>
35
#include <linux/init_task.h>
36 37 38 39 40 41
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
42
#include <linux/proc_fs.h>
43 44 45 46
#include <linux/rcupdate.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/spinlock.h>
47
#include <linux/rwsem.h>
48
#include <linux/string.h>
49
#include <linux/sort.h>
50
#include <linux/kmod.h>
B
Balbir Singh 已提交
51 52
#include <linux/delayacct.h>
#include <linux/cgroupstats.h>
53
#include <linux/hashtable.h>
L
Li Zefan 已提交
54
#include <linux/pid_namespace.h>
55
#include <linux/idr.h>
56
#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
57
#include <linux/kthread.h>
T
Tejun Heo 已提交
58
#include <linux/delay.h>
B
Balbir Singh 已提交
59

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

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

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

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

T
Tejun Heo 已提交
82 83 84 85
/*
 * cgroup_mutex is the master lock.  Any modification to cgroup or its
 * hierarchy must be performed while holding it.
 *
86 87
 * css_set_rwsem protects task->cgroups pointer, the list of css_set
 * objects, and the chain of tasks off each css_set.
T
Tejun Heo 已提交
88
 *
89 90
 * These locks are exported if CONFIG_PROVE_RCU so that accessors in
 * cgroup.h can use them for lockdep annotations.
T
Tejun Heo 已提交
91
 */
T
Tejun Heo 已提交
92 93
#ifdef CONFIG_PROVE_RCU
DEFINE_MUTEX(cgroup_mutex);
94 95 96
DECLARE_RWSEM(css_set_rwsem);
EXPORT_SYMBOL_GPL(cgroup_mutex);
EXPORT_SYMBOL_GPL(css_set_rwsem);
T
Tejun Heo 已提交
97
#else
98
static DEFINE_MUTEX(cgroup_mutex);
99
static DECLARE_RWSEM(css_set_rwsem);
T
Tejun Heo 已提交
100 101
#endif

102
/*
103 104
 * Protects cgroup_idr and css_idr so that IDs can be released without
 * grabbing cgroup_mutex.
105 106 107
 */
static DEFINE_SPINLOCK(cgroup_idr_lock);

108 109 110 111 112
/*
 * Protects cgroup_subsys->release_agent_path.  Modifying it also requires
 * cgroup_mutex.  Reading requires either cgroup_mutex or this spinlock.
 */
static DEFINE_SPINLOCK(release_agent_path_lock);
113

T
Tejun Heo 已提交
114
#define cgroup_assert_mutexes_or_rcu_locked()				\
115
	rcu_lockdep_assert(rcu_read_lock_held() ||			\
T
Tejun Heo 已提交
116
			   lockdep_is_held(&cgroup_tree_mutex) ||	\
117
			   lockdep_is_held(&cgroup_mutex),		\
T
Tejun Heo 已提交
118
			   "cgroup_[tree_]mutex or RCU read lock required");
119

120 121 122 123 124 125 126 127
/*
 * cgroup destruction makes heavy use of work items and there can be a lot
 * of concurrent destructions.  Use a separate workqueue so that cgroup
 * destruction work items don't end up filling up max_active of system_wq
 * which may lead to deadlock.
 */
static struct workqueue_struct *cgroup_destroy_wq;

128 129 130 131 132 133
/*
 * pidlist destructions need to be flushed on cgroup destruction.  Use a
 * separate workqueue as flush domain.
 */
static struct workqueue_struct *cgroup_pidlist_destroy_wq;

T
Tejun Heo 已提交
134
/* generate an array of cgroup subsystem pointers */
135
#define SUBSYS(_x) [_x ## _cgrp_id] = &_x ## _cgrp_subsys,
T
Tejun Heo 已提交
136
static struct cgroup_subsys *cgroup_subsys[] = {
137 138
#include <linux/cgroup_subsys.h>
};
139 140 141 142 143
#undef SUBSYS

/* array of cgroup subsystem names */
#define SUBSYS(_x) [_x ## _cgrp_id] = #_x,
static const char *cgroup_subsys_name[] = {
144 145
#include <linux/cgroup_subsys.h>
};
146
#undef SUBSYS
147 148

/*
149
 * The default hierarchy, reserved for the subsystems that are otherwise
150 151
 * unattached - it never has more than a single cgroup, and all tasks are
 * part of that cgroup.
152
 */
T
Tejun Heo 已提交
153
struct cgroup_root cgrp_dfl_root;
154

T
Tejun Heo 已提交
155 156 157 158 159
/*
 * The default hierarchy always exists but is hidden until mounted for the
 * first time.  This is for backward compatibility.
 */
static bool cgrp_dfl_root_visible;
160 161 162

/* The list of hierarchy roots */

163 164
static LIST_HEAD(cgroup_roots);
static int cgroup_root_count;
165

T
Tejun Heo 已提交
166
/* hierarchy ID allocation and mapping, protected by cgroup_mutex */
167
static DEFINE_IDR(cgroup_hierarchy_idr);
168

169 170 171 172 173
/*
 * 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
174 175
 * the ascending serial number order on the list.  Protected by
 * cgroup_mutex.
176
 */
177
static u64 cgroup_serial_nr_next = 1;
178

179
/* This flag indicates whether tasks in the fork and exit paths should
L
Li Zefan 已提交
180 181 182
 * 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.
183
 */
184
static int need_forkexit_callback __read_mostly;
185

186 187
static struct cftype cgroup_base_files[];

188
static void cgroup_put(struct cgroup *cgrp);
189
static int rebind_subsystems(struct cgroup_root *dst_root,
190
			     unsigned int ss_mask);
191
static void cgroup_destroy_css_killed(struct cgroup *cgrp);
192
static int cgroup_destroy_locked(struct cgroup *cgrp);
193 194
static int create_css(struct cgroup *cgrp, struct cgroup_subsys *ss);
static void kill_css(struct cgroup_subsys_state *css);
195 196
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add);
197
static void cgroup_pidlist_destroy_all(struct cgroup *cgrp);
198

199 200 201 202 203 204 205
/* IDR wrappers which synchronize using cgroup_idr_lock */
static int cgroup_idr_alloc(struct idr *idr, void *ptr, int start, int end,
			    gfp_t gfp_mask)
{
	int ret;

	idr_preload(gfp_mask);
T
Tejun Heo 已提交
206
	spin_lock_bh(&cgroup_idr_lock);
207
	ret = idr_alloc(idr, ptr, start, end, gfp_mask);
T
Tejun Heo 已提交
208
	spin_unlock_bh(&cgroup_idr_lock);
209 210 211 212 213 214 215 216
	idr_preload_end();
	return ret;
}

static void *cgroup_idr_replace(struct idr *idr, void *ptr, int id)
{
	void *ret;

T
Tejun Heo 已提交
217
	spin_lock_bh(&cgroup_idr_lock);
218
	ret = idr_replace(idr, ptr, id);
T
Tejun Heo 已提交
219
	spin_unlock_bh(&cgroup_idr_lock);
220 221 222 223 224
	return ret;
}

static void cgroup_idr_remove(struct idr *idr, int id)
{
T
Tejun Heo 已提交
225
	spin_lock_bh(&cgroup_idr_lock);
226
	idr_remove(idr, id);
T
Tejun Heo 已提交
227
	spin_unlock_bh(&cgroup_idr_lock);
228 229
}

T
Tejun Heo 已提交
230 231 232
/**
 * cgroup_css - obtain a cgroup's css for the specified subsystem
 * @cgrp: the cgroup of interest
233
 * @ss: the subsystem of interest (%NULL returns the dummy_css)
T
Tejun Heo 已提交
234
 *
235 236 237 238 239
 * Return @cgrp's css (cgroup_subsys_state) associated with @ss.  This
 * function must be called either under cgroup_mutex or rcu_read_lock() and
 * the caller is responsible for pinning the returned css if it wants to
 * keep accessing it outside the said locks.  This function may return
 * %NULL if @cgrp doesn't have @subsys_id enabled.
T
Tejun Heo 已提交
240 241
 */
static struct cgroup_subsys_state *cgroup_css(struct cgroup *cgrp,
242
					      struct cgroup_subsys *ss)
T
Tejun Heo 已提交
243
{
244
	if (ss)
245
		return rcu_dereference_check(cgrp->subsys[ss->id],
T
Tejun Heo 已提交
246 247
					lockdep_is_held(&cgroup_tree_mutex) ||
					lockdep_is_held(&cgroup_mutex));
248 249
	else
		return &cgrp->dummy_css;
T
Tejun Heo 已提交
250
}
251

252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
/**
 * cgroup_e_css - obtain a cgroup's effective css for the specified subsystem
 * @cgrp: the cgroup of interest
 * @ss: the subsystem of interest (%NULL returns the dummy_css)
 *
 * Similar to cgroup_css() but returns the effctive css, which is defined
 * as the matching css of the nearest ancestor including self which has @ss
 * enabled.  If @ss is associated with the hierarchy @cgrp is on, this
 * function is guaranteed to return non-NULL css.
 */
static struct cgroup_subsys_state *cgroup_e_css(struct cgroup *cgrp,
						struct cgroup_subsys *ss)
{
	lockdep_assert_held(&cgroup_mutex);

	if (!ss)
		return &cgrp->dummy_css;

	if (!(cgrp->root->subsys_mask & (1 << ss->id)))
		return NULL;

	while (cgrp->parent &&
	       !(cgrp->parent->child_subsys_mask & (1 << ss->id)))
		cgrp = cgrp->parent;

	return cgroup_css(cgrp, ss);
}

280
/* convenient tests for these bits */
281
static inline bool cgroup_is_dead(const struct cgroup *cgrp)
282
{
283
	return test_bit(CGRP_DEAD, &cgrp->flags);
284 285
}

286 287
struct cgroup_subsys_state *seq_css(struct seq_file *seq)
{
T
Tejun Heo 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
	struct kernfs_open_file *of = seq->private;
	struct cgroup *cgrp = of->kn->parent->priv;
	struct cftype *cft = seq_cft(seq);

	/*
	 * This is open and unprotected implementation of cgroup_css().
	 * seq_css() is only called from a kernfs file operation which has
	 * an active reference on the file.  Because all the subsystem
	 * files are drained before a css is disassociated with a cgroup,
	 * the matching css from the cgroup's subsys table is guaranteed to
	 * be and stay valid until the enclosing operation is complete.
	 */
	if (cft->ss)
		return rcu_dereference_raw(cgrp->subsys[cft->ss->id]);
	else
		return &cgrp->dummy_css;
304 305 306
}
EXPORT_SYMBOL_GPL(seq_css);

307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
/**
 * 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;
}
325

326
static int cgroup_is_releasable(const struct cgroup *cgrp)
327 328
{
	const int bits =
329 330 331
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
332 333
}

334
static int notify_on_release(const struct cgroup *cgrp)
335
{
336
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
337 338
}

T
Tejun Heo 已提交
339 340 341 342 343 344
/**
 * for_each_css - iterate all css's of a cgroup
 * @css: the iteration cursor
 * @ssid: the index of the subsystem, CGROUP_SUBSYS_COUNT after reaching the end
 * @cgrp: the target cgroup to iterate css's of
 *
345
 * Should be called under cgroup_[tree_]mutex.
T
Tejun Heo 已提交
346 347 348 349 350
 */
#define for_each_css(css, ssid, cgrp)					\
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT; (ssid)++)	\
		if (!((css) = rcu_dereference_check(			\
				(cgrp)->subsys[(ssid)],			\
T
Tejun Heo 已提交
351
				lockdep_is_held(&cgroup_tree_mutex) ||	\
T
Tejun Heo 已提交
352 353 354
				lockdep_is_held(&cgroup_mutex)))) { }	\
		else

355 356 357 358 359 360 361 362 363 364 365 366 367 368
/**
 * for_each_e_css - iterate all effective css's of a cgroup
 * @css: the iteration cursor
 * @ssid: the index of the subsystem, CGROUP_SUBSYS_COUNT after reaching the end
 * @cgrp: the target cgroup to iterate css's of
 *
 * Should be called under cgroup_[tree_]mutex.
 */
#define for_each_e_css(css, ssid, cgrp)					\
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT; (ssid)++)	\
		if (!((css) = cgroup_e_css(cgrp, cgroup_subsys[(ssid)]))) \
			;						\
		else

369
/**
T
Tejun Heo 已提交
370
 * for_each_subsys - iterate all enabled cgroup subsystems
371
 * @ss: the iteration cursor
372
 * @ssid: the index of @ss, CGROUP_SUBSYS_COUNT after reaching the end
373
 */
374
#define for_each_subsys(ss, ssid)					\
T
Tejun Heo 已提交
375 376
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT &&		\
	     (((ss) = cgroup_subsys[ssid]) || true); (ssid)++)
377

378 379
/* iterate across the hierarchies */
#define for_each_root(root)						\
380
	list_for_each_entry((root), &cgroup_roots, root_list)
381

382 383 384 385 386 387 388 389
/* iterate over child cgrps, lock should be held throughout iteration */
#define cgroup_for_each_live_child(child, cgrp)				\
	list_for_each_entry((child), &(cgrp)->children, sibling)	\
		if (({ lockdep_assert_held(&cgroup_tree_mutex);		\
		       cgroup_is_dead(child); }))			\
			;						\
		else

390 391 392 393
/**
 * 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
394 395
 * On success, returns true; the mutex should be later unlocked.  On
 * failure returns false with no lock held.
396
 */
397
static bool cgroup_lock_live_group(struct cgroup *cgrp)
398 399
{
	mutex_lock(&cgroup_mutex);
400
	if (cgroup_is_dead(cgrp)) {
401 402 403 404 405 406
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}

407 408 409
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
410
static DEFINE_RAW_SPINLOCK(release_list_lock);
411 412
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
413
static void check_for_release(struct cgroup *cgrp);
414

415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
/*
 * 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;
433 434
};

435 436
/*
 * The default css_set - used by init and its children prior to any
437 438 439 440 441
 * 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.
 */
T
Tejun Heo 已提交
442
struct css_set init_css_set = {
443 444 445 446 447 448 449
	.refcount		= ATOMIC_INIT(1),
	.cgrp_links		= LIST_HEAD_INIT(init_css_set.cgrp_links),
	.tasks			= LIST_HEAD_INIT(init_css_set.tasks),
	.mg_tasks		= LIST_HEAD_INIT(init_css_set.mg_tasks),
	.mg_preload_node	= LIST_HEAD_INIT(init_css_set.mg_preload_node),
	.mg_node		= LIST_HEAD_INIT(init_css_set.mg_node),
};
450

451
static int css_set_count	= 1;	/* 1 for init_css_set */
452

453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
/**
 * cgroup_update_populated - updated populated count of a cgroup
 * @cgrp: the target cgroup
 * @populated: inc or dec populated count
 *
 * @cgrp is either getting the first task (css_set) or losing the last.
 * Update @cgrp->populated_cnt accordingly.  The count is propagated
 * towards root so that a given cgroup's populated_cnt is zero iff the
 * cgroup and all its descendants are empty.
 *
 * @cgrp's interface file "cgroup.populated" is zero if
 * @cgrp->populated_cnt is zero and 1 otherwise.  When @cgrp->populated_cnt
 * changes from or to zero, userland is notified that the content of the
 * interface file has changed.  This can be used to detect when @cgrp and
 * its descendants become populated or empty.
 */
static void cgroup_update_populated(struct cgroup *cgrp, bool populated)
{
	lockdep_assert_held(&css_set_rwsem);

	do {
		bool trigger;

		if (populated)
			trigger = !cgrp->populated_cnt++;
		else
			trigger = !--cgrp->populated_cnt;

		if (!trigger)
			break;

		if (cgrp->populated_kn)
			kernfs_notify(cgrp->populated_kn);
		cgrp = cgrp->parent;
	} while (cgrp);
}

490 491 492 493 494
/*
 * 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.
 */
495
#define CSS_SET_HASH_BITS	7
496
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
497

498
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
499
{
500
	unsigned long key = 0UL;
501 502
	struct cgroup_subsys *ss;
	int i;
503

504
	for_each_subsys(ss, i)
505 506
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
507

508
	return key;
509 510
}

511
static void put_css_set_locked(struct css_set *cset, bool taskexit)
512
{
513
	struct cgrp_cset_link *link, *tmp_link;
T
Tejun Heo 已提交
514 515
	struct cgroup_subsys *ss;
	int ssid;
516

517 518 519
	lockdep_assert_held(&css_set_rwsem);

	if (!atomic_dec_and_test(&cset->refcount))
520
		return;
521

522
	/* This css_set is dead. unlink it and release cgroup refcounts */
T
Tejun Heo 已提交
523 524
	for_each_subsys(ss, ssid)
		list_del(&cset->e_cset_node[ssid]);
525
	hash_del(&cset->hlist);
526 527
	css_set_count--;

528
	list_for_each_entry_safe(link, tmp_link, &cset->cgrp_links, cgrp_link) {
529
		struct cgroup *cgrp = link->cgrp;
530

531 532
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
L
Li Zefan 已提交
533

534
		/* @cgrp can't go away while we're holding css_set_rwsem */
535 536 537 538 539 540 541
		if (list_empty(&cgrp->cset_links)) {
			cgroup_update_populated(cgrp, false);
			if (notify_on_release(cgrp)) {
				if (taskexit)
					set_bit(CGRP_RELEASABLE, &cgrp->flags);
				check_for_release(cgrp);
			}
542
		}
543 544

		kfree(link);
545
	}
546

547
	kfree_rcu(cset, rcu_head);
548 549
}

550 551 552 553 554 555 556 557 558 559 560 561 562 563 564
static void put_css_set(struct css_set *cset, bool taskexit)
{
	/*
	 * Ensure that the refcount doesn't hit zero while any readers
	 * can see it. Similar to atomic_dec_and_lock(), but for an
	 * rwlock
	 */
	if (atomic_add_unless(&cset->refcount, -1, 1))
		return;

	down_write(&css_set_rwsem);
	put_css_set_locked(cset, taskexit);
	up_write(&css_set_rwsem);
}

565 566 567
/*
 * refcounted get/put for css_set objects
 */
568
static inline void get_css_set(struct css_set *cset)
569
{
570
	atomic_inc(&cset->refcount);
571 572
}

573
/**
574
 * compare_css_sets - helper function for find_existing_css_set().
575 576
 * @cset: candidate css_set being tested
 * @old_cset: existing css_set for a task
577 578 579
 * @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
580
 * Returns true if "cset" matches "old_cset" except for the hierarchy
581 582
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
583 584
static bool compare_css_sets(struct css_set *cset,
			     struct css_set *old_cset,
585 586 587 588 589
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

590 591 592 593 594 595
	/*
	 * On the default hierarchy, there can be csets which are
	 * associated with the same set of cgroups but different csses.
	 * Let's first ensure that csses match.
	 */
	if (memcmp(template, cset->subsys, sizeof(cset->subsys)))
596 597 598 599
		return false;

	/*
	 * Compare cgroup pointers in order to distinguish between
600 601 602
	 * different cgroups in hierarchies.  As different cgroups may
	 * share the same effective css, this comparison is always
	 * necessary.
603
	 */
604 605
	l1 = &cset->cgrp_links;
	l2 = &old_cset->cgrp_links;
606
	while (1) {
607
		struct cgrp_cset_link *link1, *link2;
608
		struct cgroup *cgrp1, *cgrp2;
609 610 611 612

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
613 614
		if (l1 == &cset->cgrp_links) {
			BUG_ON(l2 != &old_cset->cgrp_links);
615 616
			break;
		} else {
617
			BUG_ON(l2 == &old_cset->cgrp_links);
618 619
		}
		/* Locate the cgroups associated with these links. */
620 621 622 623
		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;
624
		/* Hierarchies should be linked in the same order. */
625
		BUG_ON(cgrp1->root != cgrp2->root);
626 627 628 629 630 631 632 633

		/*
		 * 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.
		 */
634 635
		if (cgrp1->root == new_cgrp->root) {
			if (cgrp1 != new_cgrp)
636 637
				return false;
		} else {
638
			if (cgrp1 != cgrp2)
639 640 641 642 643 644
				return false;
		}
	}
	return true;
}

645 646 647 648 649
/**
 * 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
650
 */
651 652 653
static struct css_set *find_existing_css_set(struct css_set *old_cset,
					struct cgroup *cgrp,
					struct cgroup_subsys_state *template[])
654
{
655
	struct cgroup_root *root = cgrp->root;
656
	struct cgroup_subsys *ss;
657
	struct css_set *cset;
658
	unsigned long key;
659
	int i;
660

B
Ben Blum 已提交
661 662 663 664 665
	/*
	 * 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.
	 */
666
	for_each_subsys(ss, i) {
667
		if (root->subsys_mask & (1UL << i)) {
668 669 670 671 672
			/*
			 * @ss is in this hierarchy, so we want the
			 * effective css from @cgrp.
			 */
			template[i] = cgroup_e_css(cgrp, ss);
673
		} else {
674 675 676 677
			/*
			 * @ss is not in this hierarchy, so we don't want
			 * to change the css.
			 */
678
			template[i] = old_cset->subsys[i];
679 680 681
		}
	}

682
	key = css_set_hash(template);
683 684
	hash_for_each_possible(css_set_table, cset, hlist, key) {
		if (!compare_css_sets(cset, old_cset, cgrp, template))
685 686 687
			continue;

		/* This css_set matches what we need */
688
		return cset;
689
	}
690 691 692 693 694

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

695
static void free_cgrp_cset_links(struct list_head *links_to_free)
696
{
697
	struct cgrp_cset_link *link, *tmp_link;
698

699 700
	list_for_each_entry_safe(link, tmp_link, links_to_free, cset_link) {
		list_del(&link->cset_link);
701 702 703 704
		kfree(link);
	}
}

705 706 707 708 709 710 711
/**
 * 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.
712
 */
713
static int allocate_cgrp_cset_links(int count, struct list_head *tmp_links)
714
{
715
	struct cgrp_cset_link *link;
716
	int i;
717 718 719

	INIT_LIST_HEAD(tmp_links);

720
	for (i = 0; i < count; i++) {
721
		link = kzalloc(sizeof(*link), GFP_KERNEL);
722
		if (!link) {
723
			free_cgrp_cset_links(tmp_links);
724 725
			return -ENOMEM;
		}
726
		list_add(&link->cset_link, tmp_links);
727 728 729 730
	}
	return 0;
}

731 732
/**
 * link_css_set - a helper function to link a css_set to a cgroup
733
 * @tmp_links: cgrp_cset_link objects allocated by allocate_cgrp_cset_links()
734
 * @cset: the css_set to be linked
735 736
 * @cgrp: the destination cgroup
 */
737 738
static void link_css_set(struct list_head *tmp_links, struct css_set *cset,
			 struct cgroup *cgrp)
739
{
740
	struct cgrp_cset_link *link;
741

742
	BUG_ON(list_empty(tmp_links));
T
Tejun Heo 已提交
743 744 745 746

	if (cgroup_on_dfl(cgrp))
		cset->dfl_cgrp = cgrp;

747 748
	link = list_first_entry(tmp_links, struct cgrp_cset_link, cset_link);
	link->cset = cset;
749
	link->cgrp = cgrp;
750 751 752

	if (list_empty(&cgrp->cset_links))
		cgroup_update_populated(cgrp, true);
753
	list_move(&link->cset_link, &cgrp->cset_links);
754

755 756 757 758
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
759
	list_add_tail(&link->cgrp_link, &cset->cgrp_links);
760 761
}

762 763 764 765 766 767 768
/**
 * 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.
769
 */
770 771
static struct css_set *find_css_set(struct css_set *old_cset,
				    struct cgroup *cgrp)
772
{
773
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT] = { };
774
	struct css_set *cset;
775 776
	struct list_head tmp_links;
	struct cgrp_cset_link *link;
T
Tejun Heo 已提交
777
	struct cgroup_subsys *ss;
778
	unsigned long key;
T
Tejun Heo 已提交
779
	int ssid;
780

781 782
	lockdep_assert_held(&cgroup_mutex);

783 784
	/* First see if we already have a cgroup group that matches
	 * the desired set */
785
	down_read(&css_set_rwsem);
786 787 788
	cset = find_existing_css_set(old_cset, cgrp, template);
	if (cset)
		get_css_set(cset);
789
	up_read(&css_set_rwsem);
790

791 792
	if (cset)
		return cset;
793

794
	cset = kzalloc(sizeof(*cset), GFP_KERNEL);
795
	if (!cset)
796 797
		return NULL;

798
	/* Allocate all the cgrp_cset_link objects that we'll need */
799
	if (allocate_cgrp_cset_links(cgroup_root_count, &tmp_links) < 0) {
800
		kfree(cset);
801 802 803
		return NULL;
	}

804
	atomic_set(&cset->refcount, 1);
805
	INIT_LIST_HEAD(&cset->cgrp_links);
806
	INIT_LIST_HEAD(&cset->tasks);
T
Tejun Heo 已提交
807
	INIT_LIST_HEAD(&cset->mg_tasks);
808
	INIT_LIST_HEAD(&cset->mg_preload_node);
809
	INIT_LIST_HEAD(&cset->mg_node);
810
	INIT_HLIST_NODE(&cset->hlist);
811 812 813

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
814
	memcpy(cset->subsys, template, sizeof(cset->subsys));
815

816
	down_write(&css_set_rwsem);
817
	/* Add reference counts and links from the new css_set. */
818
	list_for_each_entry(link, &old_cset->cgrp_links, cgrp_link) {
819
		struct cgroup *c = link->cgrp;
820

821 822
		if (c->root == cgrp->root)
			c = cgrp;
823
		link_css_set(&tmp_links, cset, c);
824
	}
825

826
	BUG_ON(!list_empty(&tmp_links));
827 828

	css_set_count++;
829

T
Tejun Heo 已提交
830
	/* Add @cset to the hash table */
831 832
	key = css_set_hash(cset->subsys);
	hash_add(css_set_table, &cset->hlist, key);
833

T
Tejun Heo 已提交
834 835 836 837
	for_each_subsys(ss, ssid)
		list_add_tail(&cset->e_cset_node[ssid],
			      &cset->subsys[ssid]->cgroup->e_csets[ssid]);

838
	up_write(&css_set_rwsem);
839

840
	return cset;
841 842
}

843
static struct cgroup_root *cgroup_root_from_kf(struct kernfs_root *kf_root)
844
{
845
	struct cgroup *root_cgrp = kf_root->kn->priv;
T
Tejun Heo 已提交
846

847
	return root_cgrp->root;
T
Tejun Heo 已提交
848 849
}

850
static int cgroup_init_root_id(struct cgroup_root *root)
851 852 853 854 855
{
	int id;

	lockdep_assert_held(&cgroup_mutex);

856
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, 0, 0, GFP_KERNEL);
857 858 859 860 861 862 863
	if (id < 0)
		return id;

	root->hierarchy_id = id;
	return 0;
}

864
static void cgroup_exit_root_id(struct cgroup_root *root)
865 866 867 868 869 870 871 872 873
{
	lockdep_assert_held(&cgroup_mutex);

	if (root->hierarchy_id) {
		idr_remove(&cgroup_hierarchy_idr, root->hierarchy_id);
		root->hierarchy_id = 0;
	}
}

874
static void cgroup_free_root(struct cgroup_root *root)
875 876 877 878 879 880 881 882 883 884
{
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);

		idr_destroy(&root->cgroup_idr);
		kfree(root);
	}
}

885
static void cgroup_destroy_root(struct cgroup_root *root)
886
{
887
	struct cgroup *cgrp = &root->cgrp;
888 889
	struct cgrp_cset_link *link, *tmp_link;

T
Tejun Heo 已提交
890 891
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);
892

T
Tejun Heo 已提交
893
	BUG_ON(atomic_read(&root->nr_cgrps));
894 895 896
	BUG_ON(!list_empty(&cgrp->children));

	/* Rebind all subsystems back to the default hierarchy */
897
	rebind_subsystems(&cgrp_dfl_root, root->subsys_mask);
898 899

	/*
900 901
	 * Release all the links from cset_links to this hierarchy's
	 * root cgroup
902
	 */
903
	down_write(&css_set_rwsem);
904 905 906 907 908 909

	list_for_each_entry_safe(link, tmp_link, &cgrp->cset_links, cset_link) {
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
		kfree(link);
	}
910
	up_write(&css_set_rwsem);
911 912 913 914 915 916 917 918 919 920 921

	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
		cgroup_root_count--;
	}

	cgroup_exit_root_id(root);

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

T
Tejun Heo 已提交
922
	kernfs_destroy_root(root->kf_root);
923 924 925
	cgroup_free_root(root);
}

926 927
/* look up cgroup associated with given css_set on the specified hierarchy */
static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
928
					    struct cgroup_root *root)
929 930 931
{
	struct cgroup *res = NULL;

932 933 934
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&css_set_rwsem);

935
	if (cset == &init_css_set) {
936
		res = &root->cgrp;
937
	} else {
938 939 940
		struct cgrp_cset_link *link;

		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
941
			struct cgroup *c = link->cgrp;
942

943 944 945 946 947 948
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
949

950 951 952 953
	BUG_ON(!res);
	return res;
}

954
/*
955 956 957 958
 * Return the cgroup for "task" from the given hierarchy. Must be
 * called with cgroup_mutex and css_set_rwsem held.
 */
static struct cgroup *task_cgroup_from_root(struct task_struct *task,
959
					    struct cgroup_root *root)
960 961 962 963 964 965 966 967 968
{
	/*
	 * 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.
	 */
	return cset_cgroup_from_root(task_css_set(task), root);
}

969 970 971 972 973 974
/*
 * 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
975
 * cgroup_attach_task() can increment it again.  Because a count of zero
976 977 978 979 980 981 982 983 984 985 986 987 988
 * 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
989 990
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
991 992 993 994
 *
 * 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
995
 * least one task in the system (init, pid == 1), therefore, root cgroup
996
 * always has either children cgroups and/or using tasks.  So we don't
997
 * need a special hack to ensure that root cgroup cannot be deleted.
998 999
 *
 * P.S.  One more locking exception.  RCU is used to guard the
1000
 * update of a tasks cgroup pointer by cgroup_attach_task()
1001 1002
 */

1003
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned int subsys_mask);
T
Tejun Heo 已提交
1004
static struct kernfs_syscall_ops cgroup_kf_syscall_ops;
1005
static const struct file_operations proc_cgroupstats_operations;
1006

T
Tejun Heo 已提交
1007 1008
static char *cgroup_file_name(struct cgroup *cgrp, const struct cftype *cft,
			      char *buf)
1009
{
T
Tejun Heo 已提交
1010 1011 1012 1013 1014 1015 1016
	if (cft->ss && !(cft->flags & CFTYPE_NO_PREFIX) &&
	    !(cgrp->root->flags & CGRP_ROOT_NOPREFIX))
		snprintf(buf, CGROUP_FILE_NAME_MAX, "%s.%s",
			 cft->ss->name, cft->name);
	else
		strncpy(buf, cft->name, CGROUP_FILE_NAME_MAX);
	return buf;
1017 1018
}

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
/**
 * 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
 */
static umode_t cgroup_file_mode(const struct cftype *cft)
1029
{
1030
	umode_t mode = 0;
1031

1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
	if (cft->mode)
		return cft->mode;

	if (cft->read_u64 || cft->read_s64 || cft->seq_show)
		mode |= S_IRUGO;

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

	return mode;
1043 1044
}

1045 1046
static void cgroup_free_fn(struct work_struct *work)
{
1047
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
1048

1049
	atomic_dec(&cgrp->root->nr_cgrps);
1050
	cgroup_pidlist_destroy_all(cgrp);
1051

T
Tejun Heo 已提交
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
	if (cgrp->parent) {
		/*
		 * We get a ref to the parent, and put the ref when this
		 * cgroup is being freed, so it's guaranteed that the
		 * parent won't be destroyed before its children.
		 */
		cgroup_put(cgrp->parent);
		kernfs_put(cgrp->kn);
		kfree(cgrp);
	} else {
		/*
1063
		 * This is root cgroup's refcnt reaching zero, which
T
Tejun Heo 已提交
1064 1065 1066 1067
		 * indicates that the root should be released.
		 */
		cgroup_destroy_root(cgrp->root);
	}
1068 1069 1070 1071 1072 1073
}

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

1074
	INIT_WORK(&cgrp->destroy_work, cgroup_free_fn);
1075
	queue_work(cgroup_destroy_wq, &cgrp->destroy_work);
1076 1077
}

1078
static void cgroup_get(struct cgroup *cgrp)
1079
{
T
Tejun Heo 已提交
1080 1081 1082
	WARN_ON_ONCE(cgroup_is_dead(cgrp));
	WARN_ON_ONCE(atomic_read(&cgrp->refcnt) <= 0);
	atomic_inc(&cgrp->refcnt);
1083 1084
}

1085
static void cgroup_put(struct cgroup *cgrp)
T
Tejun Heo 已提交
1086
{
T
Tejun Heo 已提交
1087 1088
	if (!atomic_dec_and_test(&cgrp->refcnt))
		return;
T
Tejun Heo 已提交
1089
	if (WARN_ON_ONCE(cgrp->parent && !cgroup_is_dead(cgrp)))
T
Tejun Heo 已提交
1090
		return;
T
Tejun Heo 已提交
1091

1092
	cgroup_idr_remove(&cgrp->root->cgroup_idr, cgrp->id);
T
Tejun Heo 已提交
1093
	cgrp->id = -1;
T
Tejun Heo 已提交
1094

T
Tejun Heo 已提交
1095
	call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
1096
}
T
Tejun Heo 已提交
1097

1098
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
1099
{
T
Tejun Heo 已提交
1100
	char name[CGROUP_FILE_NAME_MAX];
T
Tejun Heo 已提交
1101

T
Tejun Heo 已提交
1102
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
1103
	kernfs_remove_by_name(cgrp->kn, cgroup_file_name(cgrp, cft, name));
T
Tejun Heo 已提交
1104 1105
}

1106
/**
1107
 * cgroup_clear_dir - remove subsys files in a cgroup directory
1108
 * @cgrp: target cgroup
1109 1110
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
1111
static void cgroup_clear_dir(struct cgroup *cgrp, unsigned int subsys_mask)
T
Tejun Heo 已提交
1112
{
1113
	struct cgroup_subsys *ss;
1114
	int i;
T
Tejun Heo 已提交
1115

1116
	for_each_subsys(ss, i) {
T
Tejun Heo 已提交
1117
		struct cftype *cfts;
1118

1119
		if (!(subsys_mask & (1 << i)))
1120
			continue;
T
Tejun Heo 已提交
1121 1122
		list_for_each_entry(cfts, &ss->cfts, node)
			cgroup_addrm_files(cgrp, cfts, false);
1123
	}
1124 1125
}

1126
static int rebind_subsystems(struct cgroup_root *dst_root, unsigned int ss_mask)
1127
{
1128
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
1129
	int ssid, i, ret;
1130

T
Tejun Heo 已提交
1131 1132
	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
1133

1134 1135 1136
	for_each_subsys(ss, ssid) {
		if (!(ss_mask & (1 << ssid)))
			continue;
B
Ben Blum 已提交
1137

1138 1139
		/* if @ss has non-root csses attached to it, can't move */
		if (css_next_child(NULL, cgroup_css(&ss->root->cgrp, ss)))
T
Tejun Heo 已提交
1140
			return -EBUSY;
1141

1142
		/* can't move between two non-dummy roots either */
1143
		if (ss->root != &cgrp_dfl_root && dst_root != &cgrp_dfl_root)
1144
			return -EBUSY;
1145 1146
	}

T
Tejun Heo 已提交
1147 1148 1149
	ret = cgroup_populate_dir(&dst_root->cgrp, ss_mask);
	if (ret) {
		if (dst_root != &cgrp_dfl_root)
1150
			return ret;
1151

T
Tejun Heo 已提交
1152 1153 1154 1155 1156 1157 1158
		/*
		 * Rebinding back to the default root is not allowed to
		 * fail.  Using both default and non-default roots should
		 * be rare.  Moving subsystems back and forth even more so.
		 * Just warn about it and continue.
		 */
		if (cgrp_dfl_root_visible) {
1159
			pr_warn("failed to create files (%d) while rebinding 0x%x to default root\n",
1160
				ret, ss_mask);
1161
			pr_warn("you may retry by moving them to a different hierarchy and unbinding\n");
T
Tejun Heo 已提交
1162
		}
1163
	}
1164 1165 1166 1167 1168

	/*
	 * Nothing can fail from this point on.  Remove files for the
	 * removed subsystems and rebind each subsystem.
	 */
1169
	mutex_unlock(&cgroup_mutex);
1170
	for_each_subsys(ss, ssid)
T
Tejun Heo 已提交
1171
		if (ss_mask & (1 << ssid))
1172
			cgroup_clear_dir(&ss->root->cgrp, 1 << ssid);
1173
	mutex_lock(&cgroup_mutex);
1174

1175
	for_each_subsys(ss, ssid) {
1176
		struct cgroup_root *src_root;
1177
		struct cgroup_subsys_state *css;
T
Tejun Heo 已提交
1178
		struct css_set *cset;
1179

1180 1181
		if (!(ss_mask & (1 << ssid)))
			continue;
1182

1183
		src_root = ss->root;
1184
		css = cgroup_css(&src_root->cgrp, ss);
1185

1186
		WARN_ON(!css || cgroup_css(&dst_root->cgrp, ss));
1187

1188 1189
		RCU_INIT_POINTER(src_root->cgrp.subsys[ssid], NULL);
		rcu_assign_pointer(dst_root->cgrp.subsys[ssid], css);
1190
		ss->root = dst_root;
1191
		css->cgroup = &dst_root->cgrp;
1192

T
Tejun Heo 已提交
1193 1194 1195 1196 1197 1198
		down_write(&css_set_rwsem);
		hash_for_each(css_set_table, i, cset, hlist)
			list_move_tail(&cset->e_cset_node[ss->id],
				       &dst_root->cgrp.e_csets[ss->id]);
		up_write(&css_set_rwsem);

1199 1200 1201
		src_root->subsys_mask &= ~(1 << ssid);
		src_root->cgrp.child_subsys_mask &= ~(1 << ssid);

1202
		/* default hierarchy doesn't enable controllers by default */
1203
		dst_root->subsys_mask |= 1 << ssid;
1204 1205
		if (dst_root != &cgrp_dfl_root)
			dst_root->cgrp.child_subsys_mask |= 1 << ssid;
1206

1207 1208
		if (ss->bind)
			ss->bind(css);
1209 1210
	}

T
Tejun Heo 已提交
1211
	kernfs_activate(dst_root->cgrp.kn);
1212 1213 1214
	return 0;
}

T
Tejun Heo 已提交
1215 1216
static int cgroup_show_options(struct seq_file *seq,
			       struct kernfs_root *kf_root)
1217
{
1218
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
1219
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
1220
	int ssid;
1221

T
Tejun Heo 已提交
1222
	for_each_subsys(ss, ssid)
1223
		if (root->subsys_mask & (1 << ssid))
T
Tejun Heo 已提交
1224
			seq_printf(seq, ",%s", ss->name);
1225 1226
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR)
		seq_puts(seq, ",sane_behavior");
1227
	if (root->flags & CGRP_ROOT_NOPREFIX)
1228
		seq_puts(seq, ",noprefix");
1229
	if (root->flags & CGRP_ROOT_XATTR)
A
Aristeu Rozanski 已提交
1230
		seq_puts(seq, ",xattr");
1231 1232

	spin_lock(&release_agent_path_lock);
1233 1234
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1235 1236
	spin_unlock(&release_agent_path_lock);

1237
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->cgrp.flags))
1238
		seq_puts(seq, ",clone_children");
1239 1240
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
1241 1242 1243 1244
	return 0;
}

struct cgroup_sb_opts {
1245 1246
	unsigned int subsys_mask;
	unsigned int flags;
1247
	char *release_agent;
1248
	bool cpuset_clone_children;
1249
	char *name;
1250 1251
	/* User explicitly requested empty subsystem */
	bool none;
1252 1253
};

B
Ben Blum 已提交
1254
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1255
{
1256 1257
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1258
	unsigned int mask = -1U;
1259 1260
	struct cgroup_subsys *ss;
	int i;
1261 1262

#ifdef CONFIG_CPUSETS
1263
	mask = ~(1U << cpuset_cgrp_id);
1264
#endif
1265

1266
	memset(opts, 0, sizeof(*opts));
1267 1268 1269 1270

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1271
		if (!strcmp(token, "none")) {
1272 1273
			/* Explicitly have no subsystems */
			opts->none = true;
1274 1275 1276 1277 1278 1279 1280 1281 1282
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
1283 1284 1285 1286
		if (!strcmp(token, "__DEVEL__sane_behavior")) {
			opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
			continue;
		}
1287
		if (!strcmp(token, "noprefix")) {
1288
			opts->flags |= CGRP_ROOT_NOPREFIX;
1289 1290 1291
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1292
			opts->cpuset_clone_children = true;
1293 1294
			continue;
		}
A
Aristeu Rozanski 已提交
1295
		if (!strcmp(token, "xattr")) {
1296
			opts->flags |= CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
1297 1298
			continue;
		}
1299
		if (!strncmp(token, "release_agent=", 14)) {
1300 1301 1302
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1303
			opts->release_agent =
1304
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1305 1306
			if (!opts->release_agent)
				return -ENOMEM;
1307 1308 1309
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
			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,
1327
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1328 1329 1330
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1331 1332 1333 1334

			continue;
		}

1335
		for_each_subsys(ss, i) {
1336 1337 1338 1339 1340 1341 1342 1343
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1344
			opts->subsys_mask |= (1 << i);
1345 1346 1347 1348 1349 1350 1351 1352
			one_ss = true;

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

1353 1354
	/* Consistency checks */

1355
	if (opts->flags & CGRP_ROOT_SANE_BEHAVIOR) {
1356
		pr_warn("sane_behavior: this is still under development and its behaviors will change, proceed at your own risk\n");
1357

1358 1359 1360
		if ((opts->flags & (CGRP_ROOT_NOPREFIX | CGRP_ROOT_XATTR)) ||
		    opts->cpuset_clone_children || opts->release_agent ||
		    opts->name) {
1361
			pr_err("sane_behavior: noprefix, xattr, clone_children, release_agent and name are not allowed\n");
1362 1363
			return -EINVAL;
		}
T
Tejun Heo 已提交
1364 1365 1366 1367 1368 1369 1370 1371 1372
	} else {
		/*
		 * If the 'all' option was specified select all the
		 * subsystems, otherwise if 'none', 'name=' and a subsystem
		 * name options were not specified, let's default to 'all'
		 */
		if (all_ss || (!one_ss && !opts->none && !opts->name))
			for_each_subsys(ss, i)
				if (!ss->disabled)
1373
					opts->subsys_mask |= (1 << i);
1374

T
Tejun Heo 已提交
1375 1376 1377 1378 1379
		/*
		 * We either have to specify by name or by subsystems. (So
		 * all empty hierarchies must have a name).
		 */
		if (!opts->subsys_mask && !opts->name)
1380 1381 1382
			return -EINVAL;
	}

1383 1384 1385 1386 1387
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
1388
	if ((opts->flags & CGRP_ROOT_NOPREFIX) && (opts->subsys_mask & mask))
1389 1390
		return -EINVAL;

1391 1392

	/* Can't specify "none" and some subsystems */
1393
	if (opts->subsys_mask && opts->none)
1394 1395
		return -EINVAL;

1396 1397 1398
	return 0;
}

T
Tejun Heo 已提交
1399
static int cgroup_remount(struct kernfs_root *kf_root, int *flags, char *data)
1400 1401
{
	int ret = 0;
1402
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
1403
	struct cgroup_sb_opts opts;
1404
	unsigned int added_mask, removed_mask;
1405

1406
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR) {
1407
		pr_err("sane_behavior: remount is not allowed\n");
1408 1409 1410
		return -EINVAL;
	}

T
Tejun Heo 已提交
1411
	mutex_lock(&cgroup_tree_mutex);
1412 1413 1414 1415 1416 1417 1418
	mutex_lock(&cgroup_mutex);

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

1419
	if (opts.subsys_mask != root->subsys_mask || opts.release_agent)
1420
		pr_warn("option changes via remount are deprecated (pid=%d comm=%s)\n",
1421
			task_tgid_nr(current), current->comm);
1422

1423 1424
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1425

B
Ben Blum 已提交
1426
	/* Don't allow flags or name to change at remount */
1427
	if (((opts.flags ^ root->flags) & CGRP_ROOT_OPTION_MASK) ||
B
Ben Blum 已提交
1428
	    (opts.name && strcmp(opts.name, root->name))) {
1429
		pr_err("option or name mismatch, new: 0x%x \"%s\", old: 0x%x \"%s\"\n",
1430 1431
		       opts.flags & CGRP_ROOT_OPTION_MASK, opts.name ?: "",
		       root->flags & CGRP_ROOT_OPTION_MASK, root->name);
1432 1433 1434 1435
		ret = -EINVAL;
		goto out_unlock;
	}

1436
	/* remounting is not allowed for populated hierarchies */
1437
	if (!list_empty(&root->cgrp.children)) {
1438
		ret = -EBUSY;
1439
		goto out_unlock;
B
Ben Blum 已提交
1440
	}
1441

1442
	ret = rebind_subsystems(root, added_mask);
1443
	if (ret)
1444
		goto out_unlock;
1445

1446
	rebind_subsystems(&cgrp_dfl_root, removed_mask);
1447

1448 1449
	if (opts.release_agent) {
		spin_lock(&release_agent_path_lock);
1450
		strcpy(root->release_agent_path, opts.release_agent);
1451 1452
		spin_unlock(&release_agent_path_lock);
	}
1453
 out_unlock:
1454
	kfree(opts.release_agent);
1455
	kfree(opts.name);
1456
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1457
	mutex_unlock(&cgroup_tree_mutex);
1458 1459 1460
	return ret;
}

1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472
/*
 * 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 mount.
 */
static bool use_task_css_set_links __read_mostly;

static void cgroup_enable_task_cg_lists(void)
{
	struct task_struct *p, *g;

1473
	down_write(&css_set_rwsem);
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495

	if (use_task_css_set_links)
		goto out_unlock;

	use_task_css_set_links = true;

	/*
	 * 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);
	do_each_thread(g, p) {
		WARN_ON_ONCE(!list_empty(&p->cg_list) ||
			     task_css_set(p) != &init_css_set);

		/*
		 * 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.
1496 1497
		 * Do it while holding siglock so that we don't end up
		 * racing against cgroup_exit().
1498
		 */
1499
		spin_lock_irq(&p->sighand->siglock);
1500 1501 1502 1503 1504 1505
		if (!(p->flags & PF_EXITING)) {
			struct css_set *cset = task_css_set(p);

			list_add(&p->cg_list, &cset->tasks);
			get_css_set(cset);
		}
1506
		spin_unlock_irq(&p->sighand->siglock);
1507 1508 1509
	} while_each_thread(g, p);
	read_unlock(&tasklist_lock);
out_unlock:
1510
	up_write(&css_set_rwsem);
1511
}
1512

1513 1514
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
T
Tejun Heo 已提交
1515 1516 1517
	struct cgroup_subsys *ss;
	int ssid;

T
Tejun Heo 已提交
1518
	atomic_set(&cgrp->refcnt, 1);
1519 1520
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
1521
	INIT_LIST_HEAD(&cgrp->cset_links);
1522
	INIT_LIST_HEAD(&cgrp->release_list);
1523 1524
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
T
Tejun Heo 已提交
1525
	cgrp->dummy_css.cgroup = cgrp;
T
Tejun Heo 已提交
1526 1527 1528

	for_each_subsys(ss, ssid)
		INIT_LIST_HEAD(&cgrp->e_csets[ssid]);
1529 1530

	init_waitqueue_head(&cgrp->offline_waitq);
1531
}
1532

1533
static void init_cgroup_root(struct cgroup_root *root,
1534
			     struct cgroup_sb_opts *opts)
1535
{
1536
	struct cgroup *cgrp = &root->cgrp;
1537

1538
	INIT_LIST_HEAD(&root->root_list);
1539
	atomic_set(&root->nr_cgrps, 1);
1540
	cgrp->root = root;
1541
	init_cgroup_housekeeping(cgrp);
1542
	idr_init(&root->cgroup_idr);
1543 1544 1545 1546 1547 1548

	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1549
	if (opts->cpuset_clone_children)
1550
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->cgrp.flags);
1551 1552
}

1553
static int cgroup_setup_root(struct cgroup_root *root, unsigned int ss_mask)
1554
{
1555
	LIST_HEAD(tmp_links);
1556
	struct cgroup *root_cgrp = &root->cgrp;
1557 1558
	struct css_set *cset;
	int i, ret;
1559

1560 1561
	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
1562

1563
	ret = cgroup_idr_alloc(&root->cgroup_idr, root_cgrp, 1, 2, GFP_NOWAIT);
1564
	if (ret < 0)
T
Tejun Heo 已提交
1565
		goto out;
1566
	root_cgrp->id = ret;
1567

1568
	/*
1569
	 * We're accessing css_set_count without locking css_set_rwsem here,
1570 1571 1572 1573 1574 1575
	 * but that's OK - it can only be increased by someone holding
	 * cgroup_lock, and that's us. The worst that can happen is that we
	 * have some link structures left over
	 */
	ret = allocate_cgrp_cset_links(css_set_count, &tmp_links);
	if (ret)
T
Tejun Heo 已提交
1576
		goto out;
1577

1578
	ret = cgroup_init_root_id(root);
1579
	if (ret)
T
Tejun Heo 已提交
1580
		goto out;
1581

T
Tejun Heo 已提交
1582 1583 1584 1585 1586 1587 1588 1589
	root->kf_root = kernfs_create_root(&cgroup_kf_syscall_ops,
					   KERNFS_ROOT_CREATE_DEACTIVATED,
					   root_cgrp);
	if (IS_ERR(root->kf_root)) {
		ret = PTR_ERR(root->kf_root);
		goto exit_root_id;
	}
	root_cgrp->kn = root->kf_root->kn;
1590

1591 1592
	ret = cgroup_addrm_files(root_cgrp, cgroup_base_files, true);
	if (ret)
T
Tejun Heo 已提交
1593
		goto destroy_root;
1594

1595
	ret = rebind_subsystems(root, ss_mask);
1596
	if (ret)
T
Tejun Heo 已提交
1597
		goto destroy_root;
1598

1599 1600 1601 1602 1603 1604 1605
	/*
	 * There must be no failure case after here, since rebinding takes
	 * care of subsystems' refcounts, which are explicitly dropped in
	 * the failure exit path.
	 */
	list_add(&root->root_list, &cgroup_roots);
	cgroup_root_count++;
A
Al Viro 已提交
1606

1607
	/*
1608
	 * Link the root cgroup in this hierarchy into all the css_set
1609 1610
	 * objects.
	 */
1611
	down_write(&css_set_rwsem);
1612 1613
	hash_for_each(css_set_table, i, cset, hlist)
		link_css_set(&tmp_links, cset, root_cgrp);
1614
	up_write(&css_set_rwsem);
1615

1616
	BUG_ON(!list_empty(&root_cgrp->children));
1617
	BUG_ON(atomic_read(&root->nr_cgrps) != 1);
1618

T
Tejun Heo 已提交
1619
	kernfs_activate(root_cgrp->kn);
1620
	ret = 0;
T
Tejun Heo 已提交
1621
	goto out;
1622

T
Tejun Heo 已提交
1623 1624 1625 1626
destroy_root:
	kernfs_destroy_root(root->kf_root);
	root->kf_root = NULL;
exit_root_id:
1627
	cgroup_exit_root_id(root);
T
Tejun Heo 已提交
1628
out:
1629 1630
	free_cgrp_cset_links(&tmp_links);
	return ret;
1631 1632
}

A
Al Viro 已提交
1633
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1634
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1635
			 void *data)
1636
{
1637
	struct cgroup_root *root;
1638
	struct cgroup_sb_opts opts;
T
Tejun Heo 已提交
1639
	struct dentry *dentry;
1640
	int ret;
L
Li Zefan 已提交
1641
	bool new_sb;
1642

1643 1644 1645 1646 1647 1648
	/*
	 * The first time anyone tries to mount a cgroup, enable the list
	 * linking each css_set to its tasks and fix up all existing tasks.
	 */
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();
1649

1650
	mutex_lock(&cgroup_tree_mutex);
B
Ben Blum 已提交
1651
	mutex_lock(&cgroup_mutex);
1652 1653

	/* First find the desired set of subsystems */
1654
	ret = parse_cgroupfs_options(data, &opts);
1655
	if (ret)
1656
		goto out_unlock;
1657
retry:
T
Tejun Heo 已提交
1658
	/* look for a matching existing root */
T
Tejun Heo 已提交
1659 1660 1661 1662 1663 1664
	if (!opts.subsys_mask && !opts.none && !opts.name) {
		cgrp_dfl_root_visible = true;
		root = &cgrp_dfl_root;
		cgroup_get(&root->cgrp);
		ret = 0;
		goto out_unlock;
1665 1666
	}

1667
	for_each_root(root) {
T
Tejun Heo 已提交
1668
		bool name_match = false;
1669

1670
		if (root == &cgrp_dfl_root)
1671
			continue;
1672

B
Ben Blum 已提交
1673
		/*
T
Tejun Heo 已提交
1674 1675 1676
		 * If we asked for a name then it must match.  Also, if
		 * name matches but sybsys_mask doesn't, we should fail.
		 * Remember whether name matched.
B
Ben Blum 已提交
1677
		 */
T
Tejun Heo 已提交
1678 1679 1680 1681 1682
		if (opts.name) {
			if (strcmp(opts.name, root->name))
				continue;
			name_match = true;
		}
1683

1684
		/*
T
Tejun Heo 已提交
1685 1686
		 * If we asked for subsystems (or explicitly for no
		 * subsystems) then they must match.
1687
		 */
T
Tejun Heo 已提交
1688
		if ((opts.subsys_mask || opts.none) &&
1689
		    (opts.subsys_mask != root->subsys_mask)) {
T
Tejun Heo 已提交
1690 1691 1692 1693 1694
			if (!name_match)
				continue;
			ret = -EBUSY;
			goto out_unlock;
		}
1695

1696
		if ((root->flags ^ opts.flags) & CGRP_ROOT_OPTION_MASK) {
1697
			if ((root->flags | opts.flags) & CGRP_ROOT_SANE_BEHAVIOR) {
1698
				pr_err("sane_behavior: new mount options should match the existing superblock\n");
1699
				ret = -EINVAL;
1700
				goto out_unlock;
1701
			} else {
1702
				pr_warn("new mount options do not match the existing superblock, will be ignored\n");
1703
			}
1704
		}
1705

T
Tejun Heo 已提交
1706
		/*
1707
		 * A root's lifetime is governed by its root cgroup.  Zero
T
Tejun Heo 已提交
1708 1709 1710 1711 1712
		 * ref indicate that the root is being destroyed.  Wait for
		 * destruction to complete so that the subsystems are free.
		 * We can use wait_queue for the wait but this path is
		 * super cold.  Let's just sleep for a bit and retry.
		 */
1713
		if (!atomic_inc_not_zero(&root->cgrp.refcnt)) {
T
Tejun Heo 已提交
1714 1715 1716
			mutex_unlock(&cgroup_mutex);
			mutex_unlock(&cgroup_tree_mutex);
			msleep(10);
1717 1718
			mutex_lock(&cgroup_tree_mutex);
			mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1719 1720
			goto retry;
		}
1721

T
Tejun Heo 已提交
1722
		ret = 0;
T
Tejun Heo 已提交
1723
		goto out_unlock;
1724 1725
	}

1726
	/*
1727 1728 1729
	 * No such thing, create a new one.  name= matching without subsys
	 * specification is allowed for already existing hierarchies but we
	 * can't create new one without subsys specification.
1730
	 */
1731 1732 1733
	if (!opts.subsys_mask && !opts.none) {
		ret = -EINVAL;
		goto out_unlock;
1734 1735
	}

1736 1737 1738
	root = kzalloc(sizeof(*root), GFP_KERNEL);
	if (!root) {
		ret = -ENOMEM;
T
Tejun Heo 已提交
1739
		goto out_unlock;
1740
	}
1741

1742 1743
	init_cgroup_root(root, &opts);

T
Tejun Heo 已提交
1744
	ret = cgroup_setup_root(root, opts.subsys_mask);
T
Tejun Heo 已提交
1745 1746
	if (ret)
		cgroup_free_root(root);
1747

1748
out_unlock:
1749
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1750
	mutex_unlock(&cgroup_tree_mutex);
1751

1752 1753
	kfree(opts.release_agent);
	kfree(opts.name);
A
Aristeu Rozanski 已提交
1754

T
Tejun Heo 已提交
1755
	if (ret)
1756
		return ERR_PTR(ret);
T
Tejun Heo 已提交
1757

L
Li Zefan 已提交
1758 1759
	dentry = kernfs_mount(fs_type, flags, root->kf_root, &new_sb);
	if (IS_ERR(dentry) || !new_sb)
1760
		cgroup_put(&root->cgrp);
T
Tejun Heo 已提交
1761 1762 1763 1764 1765 1766
	return dentry;
}

static void cgroup_kill_sb(struct super_block *sb)
{
	struct kernfs_root *kf_root = kernfs_root_from_sb(sb);
1767
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
T
Tejun Heo 已提交
1768

1769
	cgroup_put(&root->cgrp);
T
Tejun Heo 已提交
1770
	kernfs_kill_sb(sb);
1771 1772 1773 1774
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1775
	.mount = cgroup_mount,
1776 1777 1778
	.kill_sb = cgroup_kill_sb,
};

1779 1780
static struct kobject *cgroup_kobj;

1781
/**
1782
 * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
1783 1784 1785 1786
 * @task: target task
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1787 1788 1789 1790 1791
 * 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.
 *
T
Tejun Heo 已提交
1792
 * Return value is the same as kernfs_path().
1793
 */
T
Tejun Heo 已提交
1794
char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
1795
{
1796
	struct cgroup_root *root;
1797
	struct cgroup *cgrp;
T
Tejun Heo 已提交
1798 1799
	int hierarchy_id = 1;
	char *path = NULL;
1800 1801

	mutex_lock(&cgroup_mutex);
1802
	down_read(&css_set_rwsem);
1803

1804 1805
	root = idr_get_next(&cgroup_hierarchy_idr, &hierarchy_id);

1806 1807
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
T
Tejun Heo 已提交
1808
		path = cgroup_path(cgrp, buf, buflen);
1809 1810
	} else {
		/* if no hierarchy exists, everyone is in "/" */
T
Tejun Heo 已提交
1811 1812
		if (strlcpy(buf, "/", buflen) < buflen)
			path = buf;
1813 1814
	}

1815
	up_read(&css_set_rwsem);
1816
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1817
	return path;
1818
}
1819
EXPORT_SYMBOL_GPL(task_cgroup_path);
1820

1821
/* used to track tasks and other necessary states during migration */
1822
struct cgroup_taskset {
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840
	/* the src and dst cset list running through cset->mg_node */
	struct list_head	src_csets;
	struct list_head	dst_csets;

	/*
	 * Fields for cgroup_taskset_*() iteration.
	 *
	 * Before migration is committed, the target migration tasks are on
	 * ->mg_tasks of the csets on ->src_csets.  After, on ->mg_tasks of
	 * the csets on ->dst_csets.  ->csets point to either ->src_csets
	 * or ->dst_csets depending on whether migration is committed.
	 *
	 * ->cur_csets and ->cur_task point to the current task position
	 * during iteration.
	 */
	struct list_head	*csets;
	struct css_set		*cur_cset;
	struct task_struct	*cur_task;
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850
};

/**
 * 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)
{
1851 1852 1853 1854
	tset->cur_cset = list_first_entry(tset->csets, struct css_set, mg_node);
	tset->cur_task = NULL;

	return cgroup_taskset_next(tset);
1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865
}

/**
 * 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)
{
1866 1867
	struct css_set *cset = tset->cur_cset;
	struct task_struct *task = tset->cur_task;
1868

1869 1870 1871 1872 1873 1874
	while (&cset->mg_node != tset->csets) {
		if (!task)
			task = list_first_entry(&cset->mg_tasks,
						struct task_struct, cg_list);
		else
			task = list_next_entry(task, cg_list);
1875

1876 1877 1878 1879 1880
		if (&task->cg_list != &cset->mg_tasks) {
			tset->cur_cset = cset;
			tset->cur_task = task;
			return task;
		}
1881

1882 1883 1884
		cset = list_next_entry(cset, mg_node);
		task = NULL;
	}
1885

1886
	return NULL;
1887 1888
}

1889
/**
B
Ben Blum 已提交
1890
 * cgroup_task_migrate - move a task from one cgroup to another.
1891
 * @old_cgrp: the cgroup @tsk is being migrated from
1892 1893
 * @tsk: the task being migrated
 * @new_cset: the new css_set @tsk is being attached to
B
Ben Blum 已提交
1894
 *
1895
 * Must be called with cgroup_mutex, threadgroup and css_set_rwsem locked.
B
Ben Blum 已提交
1896
 */
1897 1898 1899
static void cgroup_task_migrate(struct cgroup *old_cgrp,
				struct task_struct *tsk,
				struct css_set *new_cset)
B
Ben Blum 已提交
1900
{
1901
	struct css_set *old_cset;
B
Ben Blum 已提交
1902

1903 1904 1905
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&css_set_rwsem);

B
Ben Blum 已提交
1906
	/*
1907 1908 1909
	 * 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
1910
	 */
1911
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
1912
	old_cset = task_css_set(tsk);
B
Ben Blum 已提交
1913

1914
	get_css_set(new_cset);
1915
	rcu_assign_pointer(tsk->cgroups, new_cset);
B
Ben Blum 已提交
1916

1917 1918 1919 1920 1921 1922 1923
	/*
	 * Use move_tail so that cgroup_taskset_first() still returns the
	 * leader after migration.  This works because cgroup_migrate()
	 * ensures that the dst_cset of the leader is the first on the
	 * tset's dst_csets list.
	 */
	list_move_tail(&tsk->cg_list, &new_cset->mg_tasks);
B
Ben Blum 已提交
1924 1925

	/*
1926 1927 1928
	 * 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
1929
	 */
1930
	set_bit(CGRP_RELEASABLE, &old_cgrp->flags);
1931
	put_css_set_locked(old_cset, false);
B
Ben Blum 已提交
1932 1933
}

L
Li Zefan 已提交
1934
/**
1935 1936
 * cgroup_migrate_finish - cleanup after attach
 * @preloaded_csets: list of preloaded css_sets
B
Ben Blum 已提交
1937
 *
1938 1939
 * Undo cgroup_migrate_add_src() and cgroup_migrate_prepare_dst().  See
 * those functions for details.
B
Ben Blum 已提交
1940
 */
1941
static void cgroup_migrate_finish(struct list_head *preloaded_csets)
B
Ben Blum 已提交
1942
{
1943
	struct css_set *cset, *tmp_cset;
B
Ben Blum 已提交
1944

1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
	lockdep_assert_held(&cgroup_mutex);

	down_write(&css_set_rwsem);
	list_for_each_entry_safe(cset, tmp_cset, preloaded_csets, mg_preload_node) {
		cset->mg_src_cgrp = NULL;
		cset->mg_dst_cset = NULL;
		list_del_init(&cset->mg_preload_node);
		put_css_set_locked(cset, false);
	}
	up_write(&css_set_rwsem);
}

/**
 * cgroup_migrate_add_src - add a migration source css_set
 * @src_cset: the source css_set to add
 * @dst_cgrp: the destination cgroup
 * @preloaded_csets: list of preloaded css_sets
 *
 * Tasks belonging to @src_cset are about to be migrated to @dst_cgrp.  Pin
 * @src_cset and add it to @preloaded_csets, which should later be cleaned
 * up by cgroup_migrate_finish().
 *
 * This function may be called without holding threadgroup_lock even if the
 * target is a process.  Threads may be created and destroyed but as long
 * as cgroup_mutex is not dropped, no new css_set can be put into play and
 * the preloaded css_sets are guaranteed to cover all migrations.
 */
static void cgroup_migrate_add_src(struct css_set *src_cset,
				   struct cgroup *dst_cgrp,
				   struct list_head *preloaded_csets)
{
	struct cgroup *src_cgrp;

	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&css_set_rwsem);

	src_cgrp = cset_cgroup_from_root(src_cset, dst_cgrp->root);

	if (!list_empty(&src_cset->mg_preload_node))
		return;

	WARN_ON(src_cset->mg_src_cgrp);
	WARN_ON(!list_empty(&src_cset->mg_tasks));
	WARN_ON(!list_empty(&src_cset->mg_node));

	src_cset->mg_src_cgrp = src_cgrp;
	get_css_set(src_cset);
	list_add(&src_cset->mg_preload_node, preloaded_csets);
}

/**
 * cgroup_migrate_prepare_dst - prepare destination css_sets for migration
1997
 * @dst_cgrp: the destination cgroup (may be %NULL)
1998 1999 2000 2001
 * @preloaded_csets: list of preloaded source css_sets
 *
 * Tasks are about to be moved to @dst_cgrp and all the source css_sets
 * have been preloaded to @preloaded_csets.  This function looks up and
2002 2003 2004
 * pins all destination css_sets, links each to its source, and append them
 * to @preloaded_csets.  If @dst_cgrp is %NULL, the destination of each
 * source css_set is assumed to be its cgroup on the default hierarchy.
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014
 *
 * This function must be called after cgroup_migrate_add_src() has been
 * called on each migration source css_set.  After migration is performed
 * using cgroup_migrate(), cgroup_migrate_finish() must be called on
 * @preloaded_csets.
 */
static int cgroup_migrate_prepare_dst(struct cgroup *dst_cgrp,
				      struct list_head *preloaded_csets)
{
	LIST_HEAD(csets);
2015
	struct css_set *src_cset, *tmp_cset;
2016 2017 2018

	lockdep_assert_held(&cgroup_mutex);

2019 2020 2021 2022 2023 2024 2025 2026
	/*
	 * Except for the root, child_subsys_mask must be zero for a cgroup
	 * with tasks so that child cgroups don't compete against tasks.
	 */
	if (dst_cgrp && cgroup_on_dfl(dst_cgrp) && dst_cgrp->parent &&
	    dst_cgrp->child_subsys_mask)
		return -EBUSY;

2027
	/* look up the dst cset for each src cset and link it to src */
2028
	list_for_each_entry_safe(src_cset, tmp_cset, preloaded_csets, mg_preload_node) {
2029 2030
		struct css_set *dst_cset;

2031 2032
		dst_cset = find_css_set(src_cset,
					dst_cgrp ?: src_cset->dfl_cgrp);
2033 2034 2035 2036
		if (!dst_cset)
			goto err;

		WARN_ON_ONCE(src_cset->mg_dst_cset || dst_cset->mg_dst_cset);
2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050

		/*
		 * If src cset equals dst, it's noop.  Drop the src.
		 * cgroup_migrate() will skip the cset too.  Note that we
		 * can't handle src == dst as some nodes are used by both.
		 */
		if (src_cset == dst_cset) {
			src_cset->mg_src_cgrp = NULL;
			list_del_init(&src_cset->mg_preload_node);
			put_css_set(src_cset, false);
			put_css_set(dst_cset, false);
			continue;
		}

2051 2052 2053 2054 2055 2056 2057 2058
		src_cset->mg_dst_cset = dst_cset;

		if (list_empty(&dst_cset->mg_preload_node))
			list_add(&dst_cset->mg_preload_node, &csets);
		else
			put_css_set(dst_cset, false);
	}

2059
	list_splice_tail(&csets, preloaded_csets);
2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085
	return 0;
err:
	cgroup_migrate_finish(&csets);
	return -ENOMEM;
}

/**
 * cgroup_migrate - migrate a process or task to a cgroup
 * @cgrp: the destination cgroup
 * @leader: the leader of the process or the task to migrate
 * @threadgroup: whether @leader points to the whole process or a single task
 *
 * Migrate a process or task denoted by @leader to @cgrp.  If migrating a
 * process, the caller must be holding threadgroup_lock of @leader.  The
 * caller is also responsible for invoking cgroup_migrate_add_src() and
 * cgroup_migrate_prepare_dst() on the targets before invoking this
 * function and following up with cgroup_migrate_finish().
 *
 * As long as a controller's ->can_attach() doesn't fail, this function is
 * guaranteed to succeed.  This means that, excluding ->can_attach()
 * failure, when migrating multiple targets, the success or failure can be
 * decided for all targets by invoking group_migrate_prepare_dst() before
 * actually starting migrating.
 */
static int cgroup_migrate(struct cgroup *cgrp, struct task_struct *leader,
			  bool threadgroup)
B
Ben Blum 已提交
2086
{
2087 2088 2089 2090 2091
	struct cgroup_taskset tset = {
		.src_csets	= LIST_HEAD_INIT(tset.src_csets),
		.dst_csets	= LIST_HEAD_INIT(tset.dst_csets),
		.csets		= &tset.src_csets,
	};
T
Tejun Heo 已提交
2092
	struct cgroup_subsys_state *css, *failed_css = NULL;
2093 2094 2095
	struct css_set *cset, *tmp_cset;
	struct task_struct *task, *tmp_task;
	int i, ret;
B
Ben Blum 已提交
2096

2097 2098 2099 2100 2101
	/*
	 * 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.
	 */
2102
	down_write(&css_set_rwsem);
2103
	rcu_read_lock();
2104
	task = leader;
B
Ben Blum 已提交
2105
	do {
2106 2107
		/* @task either already exited or can't exit until the end */
		if (task->flags & PF_EXITING)
2108
			goto next;
2109

2110 2111
		/* leave @task alone if post_fork() hasn't linked it yet */
		if (list_empty(&task->cg_list))
2112
			goto next;
2113

2114
		cset = task_css_set(task);
2115
		if (!cset->mg_src_cgrp)
2116
			goto next;
2117

2118
		/*
2119 2120
		 * cgroup_taskset_first() must always return the leader.
		 * Take care to avoid disturbing the ordering.
2121
		 */
2122 2123 2124 2125 2126 2127
		list_move_tail(&task->cg_list, &cset->mg_tasks);
		if (list_empty(&cset->mg_node))
			list_add_tail(&cset->mg_node, &tset.src_csets);
		if (list_empty(&cset->mg_dst_cset->mg_node))
			list_move_tail(&cset->mg_dst_cset->mg_node,
				       &tset.dst_csets);
2128
	next:
2129 2130
		if (!threadgroup)
			break;
2131
	} while_each_thread(leader, task);
2132
	rcu_read_unlock();
2133
	up_write(&css_set_rwsem);
B
Ben Blum 已提交
2134

2135
	/* methods shouldn't be called if no task is actually migrating */
2136 2137
	if (list_empty(&tset.src_csets))
		return 0;
2138

2139
	/* check that we can legitimately attach to the cgroup */
2140
	for_each_e_css(css, i, cgrp) {
T
Tejun Heo 已提交
2141
		if (css->ss->can_attach) {
2142 2143
			ret = css->ss->can_attach(css, &tset);
			if (ret) {
T
Tejun Heo 已提交
2144
				failed_css = css;
B
Ben Blum 已提交
2145 2146 2147 2148 2149 2150
				goto out_cancel_attach;
			}
		}
	}

	/*
2151 2152 2153
	 * Now that we're guaranteed success, 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
2154
	 */
2155
	down_write(&css_set_rwsem);
2156 2157 2158 2159
	list_for_each_entry(cset, &tset.src_csets, mg_node) {
		list_for_each_entry_safe(task, tmp_task, &cset->mg_tasks, cg_list)
			cgroup_task_migrate(cset->mg_src_cgrp, task,
					    cset->mg_dst_cset);
B
Ben Blum 已提交
2160
	}
2161
	up_write(&css_set_rwsem);
B
Ben Blum 已提交
2162 2163

	/*
2164 2165 2166
	 * Migration is committed, all target tasks are now on dst_csets.
	 * Nothing is sensitive to fork() after this point.  Notify
	 * controllers that migration is complete.
B
Ben Blum 已提交
2167
	 */
2168
	tset.csets = &tset.dst_csets;
B
Ben Blum 已提交
2169

2170
	for_each_e_css(css, i, cgrp)
T
Tejun Heo 已提交
2171 2172
		if (css->ss->attach)
			css->ss->attach(css, &tset);
B
Ben Blum 已提交
2173

2174
	ret = 0;
2175 2176
	goto out_release_tset;

B
Ben Blum 已提交
2177
out_cancel_attach:
2178
	for_each_e_css(css, i, cgrp) {
2179 2180 2181 2182
		if (css == failed_css)
			break;
		if (css->ss->cancel_attach)
			css->ss->cancel_attach(css, &tset);
B
Ben Blum 已提交
2183
	}
2184 2185 2186 2187
out_release_tset:
	down_write(&css_set_rwsem);
	list_splice_init(&tset.dst_csets, &tset.src_csets);
	list_for_each_entry_safe(cset, tmp_cset, &tset.src_csets, mg_node) {
2188
		list_splice_tail_init(&cset->mg_tasks, &cset->tasks);
2189 2190 2191
		list_del_init(&cset->mg_node);
	}
	up_write(&css_set_rwsem);
2192
	return ret;
B
Ben Blum 已提交
2193 2194
}

2195 2196 2197 2198 2199 2200
/**
 * cgroup_attach_task - attach a task or a whole threadgroup to a cgroup
 * @dst_cgrp: the cgroup to attach to
 * @leader: the task or the leader of the threadgroup to be attached
 * @threadgroup: attach the whole threadgroup?
 *
2201
 * Call holding cgroup_mutex and threadgroup_lock of @leader.
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
 */
static int cgroup_attach_task(struct cgroup *dst_cgrp,
			      struct task_struct *leader, bool threadgroup)
{
	LIST_HEAD(preloaded_csets);
	struct task_struct *task;
	int ret;

	/* look up all src csets */
	down_read(&css_set_rwsem);
	rcu_read_lock();
	task = leader;
	do {
		cgroup_migrate_add_src(task_css_set(task), dst_cgrp,
				       &preloaded_csets);
		if (!threadgroup)
			break;
	} while_each_thread(leader, task);
	rcu_read_unlock();
	up_read(&css_set_rwsem);

	/* prepare dst csets and commit */
	ret = cgroup_migrate_prepare_dst(dst_cgrp, &preloaded_csets);
	if (!ret)
		ret = cgroup_migrate(dst_cgrp, leader, threadgroup);

	cgroup_migrate_finish(&preloaded_csets);
	return ret;
B
Ben Blum 已提交
2230 2231 2232 2233
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2234
 * function to attach either it or all tasks in its threadgroup. Will lock
2235
 * cgroup_mutex and threadgroup.
2236
 */
B
Ben Blum 已提交
2237
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2238 2239
{
	struct task_struct *tsk;
2240
	const struct cred *cred = current_cred(), *tcred;
2241 2242
	int ret;

B
Ben Blum 已提交
2243 2244 2245
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2246 2247
retry_find_task:
	rcu_read_lock();
2248
	if (pid) {
2249
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2250 2251
		if (!tsk) {
			rcu_read_unlock();
S
SeongJae Park 已提交
2252
			ret = -ESRCH;
2253
			goto out_unlock_cgroup;
2254
		}
B
Ben Blum 已提交
2255 2256 2257 2258
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2259
		tcred = __task_cred(tsk);
2260 2261 2262
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2263
			rcu_read_unlock();
2264 2265
			ret = -EACCES;
			goto out_unlock_cgroup;
2266
		}
2267 2268
	} else
		tsk = current;
2269 2270

	if (threadgroup)
2271
		tsk = tsk->group_leader;
2272 2273

	/*
2274
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2275 2276 2277
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2278
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2279 2280 2281 2282 2283
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300
	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;
		}
2301 2302 2303 2304
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2305 2306
	threadgroup_unlock(tsk);

2307
	put_task_struct(tsk);
2308
out_unlock_cgroup:
T
Tejun Heo 已提交
2309
	mutex_unlock(&cgroup_mutex);
2310 2311 2312
	return ret;
}

2313 2314 2315 2316 2317 2318 2319
/**
 * 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)
{
2320
	struct cgroup_root *root;
2321 2322
	int retval = 0;

T
Tejun Heo 已提交
2323
	mutex_lock(&cgroup_mutex);
2324
	for_each_root(root) {
2325 2326
		struct cgroup *from_cgrp;

2327
		if (root == &cgrp_dfl_root)
2328 2329
			continue;

2330 2331 2332
		down_read(&css_set_rwsem);
		from_cgrp = task_cgroup_from_root(from, root);
		up_read(&css_set_rwsem);
2333

L
Li Zefan 已提交
2334
		retval = cgroup_attach_task(from_cgrp, tsk, false);
2335 2336 2337
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2338
	mutex_unlock(&cgroup_mutex);
2339 2340 2341 2342 2343

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2344 2345
static int cgroup_tasks_write(struct cgroup_subsys_state *css,
			      struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2346
{
2347
	return attach_task_by_pid(css->cgroup, pid, false);
B
Ben Blum 已提交
2348 2349
}

2350 2351
static int cgroup_procs_write(struct cgroup_subsys_state *css,
			      struct cftype *cft, u64 tgid)
2352
{
2353
	return attach_task_by_pid(css->cgroup, tgid, true);
2354 2355
}

2356
static int cgroup_release_agent_write(struct cgroup_subsys_state *css,
2357
				      struct cftype *cft, char *buffer)
2358
{
2359
	struct cgroup_root *root = css->cgroup->root;
2360 2361

	BUILD_BUG_ON(sizeof(root->release_agent_path) < PATH_MAX);
2362
	if (!cgroup_lock_live_group(css->cgroup))
2363
		return -ENODEV;
2364
	spin_lock(&release_agent_path_lock);
2365 2366
	strlcpy(root->release_agent_path, buffer,
		sizeof(root->release_agent_path));
2367
	spin_unlock(&release_agent_path_lock);
T
Tejun Heo 已提交
2368
	mutex_unlock(&cgroup_mutex);
2369 2370 2371
	return 0;
}

2372
static int cgroup_release_agent_show(struct seq_file *seq, void *v)
2373
{
2374
	struct cgroup *cgrp = seq_css(seq)->cgroup;
2375

2376 2377 2378 2379
	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
2380
	mutex_unlock(&cgroup_mutex);
2381 2382 2383
	return 0;
}

2384
static int cgroup_sane_behavior_show(struct seq_file *seq, void *v)
2385
{
2386 2387 2388
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	seq_printf(seq, "%d\n", cgroup_sane_behavior(cgrp));
2389 2390 2391
	return 0;
}

2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537
static void cgroup_print_ss_mask(struct seq_file *seq, unsigned int ss_mask)
{
	struct cgroup_subsys *ss;
	bool printed = false;
	int ssid;

	for_each_subsys(ss, ssid) {
		if (ss_mask & (1 << ssid)) {
			if (printed)
				seq_putc(seq, ' ');
			seq_printf(seq, "%s", ss->name);
			printed = true;
		}
	}
	if (printed)
		seq_putc(seq, '\n');
}

/* show controllers which are currently attached to the default hierarchy */
static int cgroup_root_controllers_show(struct seq_file *seq, void *v)
{
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	cgroup_print_ss_mask(seq, cgrp->root->subsys_mask);
	return 0;
}

/* show controllers which are enabled from the parent */
static int cgroup_controllers_show(struct seq_file *seq, void *v)
{
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	cgroup_print_ss_mask(seq, cgrp->parent->child_subsys_mask);
	return 0;
}

/* show controllers which are enabled for a given cgroup's children */
static int cgroup_subtree_control_show(struct seq_file *seq, void *v)
{
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	cgroup_print_ss_mask(seq, cgrp->child_subsys_mask);
	return 0;
}

/**
 * cgroup_update_dfl_csses - update css assoc of a subtree in default hierarchy
 * @cgrp: root of the subtree to update csses for
 *
 * @cgrp's child_subsys_mask has changed and its subtree's (self excluded)
 * css associations need to be updated accordingly.  This function looks up
 * all css_sets which are attached to the subtree, creates the matching
 * updated css_sets and migrates the tasks to the new ones.
 */
static int cgroup_update_dfl_csses(struct cgroup *cgrp)
{
	LIST_HEAD(preloaded_csets);
	struct cgroup_subsys_state *css;
	struct css_set *src_cset;
	int ret;

	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);

	/* look up all csses currently attached to @cgrp's subtree */
	down_read(&css_set_rwsem);
	css_for_each_descendant_pre(css, cgroup_css(cgrp, NULL)) {
		struct cgrp_cset_link *link;

		/* self is not affected by child_subsys_mask change */
		if (css->cgroup == cgrp)
			continue;

		list_for_each_entry(link, &css->cgroup->cset_links, cset_link)
			cgroup_migrate_add_src(link->cset, cgrp,
					       &preloaded_csets);
	}
	up_read(&css_set_rwsem);

	/* NULL dst indicates self on default hierarchy */
	ret = cgroup_migrate_prepare_dst(NULL, &preloaded_csets);
	if (ret)
		goto out_finish;

	list_for_each_entry(src_cset, &preloaded_csets, mg_preload_node) {
		struct task_struct *last_task = NULL, *task;

		/* src_csets precede dst_csets, break on the first dst_cset */
		if (!src_cset->mg_src_cgrp)
			break;

		/*
		 * All tasks in src_cset need to be migrated to the
		 * matching dst_cset.  Empty it process by process.  We
		 * walk tasks but migrate processes.  The leader might even
		 * belong to a different cset but such src_cset would also
		 * be among the target src_csets because the default
		 * hierarchy enforces per-process membership.
		 */
		while (true) {
			down_read(&css_set_rwsem);
			task = list_first_entry_or_null(&src_cset->tasks,
						struct task_struct, cg_list);
			if (task) {
				task = task->group_leader;
				WARN_ON_ONCE(!task_css_set(task)->mg_src_cgrp);
				get_task_struct(task);
			}
			up_read(&css_set_rwsem);

			if (!task)
				break;

			/* guard against possible infinite loop */
			if (WARN(last_task == task,
				 "cgroup: update_dfl_csses failed to make progress, aborting in inconsistent state\n"))
				goto out_finish;
			last_task = task;

			threadgroup_lock(task);
			/* raced against de_thread() from another thread? */
			if (!thread_group_leader(task)) {
				threadgroup_unlock(task);
				put_task_struct(task);
				continue;
			}

			ret = cgroup_migrate(src_cset->dfl_cgrp, task, true);

			threadgroup_unlock(task);
			put_task_struct(task);

			if (WARN(ret, "cgroup: failed to update controllers for the default hierarchy (%d), further operations may crash or hang\n", ret))
				goto out_finish;
		}
	}

out_finish:
	cgroup_migrate_finish(&preloaded_csets);
	return ret;
}

/* change the enabled child controllers for a cgroup in the default hierarchy */
static int cgroup_subtree_control_write(struct cgroup_subsys_state *dummy_css,
					struct cftype *cft, char *buffer)
{
2538
	unsigned int enable_req = 0, disable_req = 0, enable, disable;
2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
	struct cgroup *cgrp = dummy_css->cgroup, *child;
	struct cgroup_subsys *ss;
	char *tok, *p;
	int ssid, ret;

	/*
	 * Parse input - white space separated list of subsystem names
	 * prefixed with either + or -.
	 */
	p = buffer;
	while ((tok = strsep(&p, " \t\n"))) {
		for_each_subsys(ss, ssid) {
			if (ss->disabled || strcmp(tok + 1, ss->name))
				continue;

			if (*tok == '+') {
				enable_req |= 1 << ssid;
				disable_req &= ~(1 << ssid);
			} else if (*tok == '-') {
				disable_req |= 1 << ssid;
				enable_req &= ~(1 << ssid);
			} else {
				return -EINVAL;
			}
			break;
		}
		if (ssid == CGROUP_SUBSYS_COUNT)
			return -EINVAL;
	}

	/*
	 * We're gonna grab cgroup_tree_mutex which nests outside kernfs
	 * active_ref.  cgroup_lock_live_group() already provides enough
	 * protection.  Ensure @cgrp stays accessible and break the
	 * active_ref protection.
	 */
	cgroup_get(cgrp);
	kernfs_break_active_protection(cgrp->control_kn);
retry:
	enable = enable_req;
	disable = disable_req;

	mutex_lock(&cgroup_tree_mutex);

	for_each_subsys(ss, ssid) {
		if (enable & (1 << ssid)) {
			if (cgrp->child_subsys_mask & (1 << ssid)) {
				enable &= ~(1 << ssid);
				continue;
			}

			/*
			 * Because css offlining is asynchronous, userland
			 * might try to re-enable the same controller while
			 * the previous instance is still around.  In such
			 * cases, wait till it's gone using offline_waitq.
			 */
			cgroup_for_each_live_child(child, cgrp) {
2597
				DEFINE_WAIT(wait);
2598 2599 2600 2601

				if (!cgroup_css(child, ss))
					continue;

2602
				cgroup_get(child);
2603 2604 2605 2606 2607
				prepare_to_wait(&child->offline_waitq, &wait,
						TASK_UNINTERRUPTIBLE);
				mutex_unlock(&cgroup_tree_mutex);
				schedule();
				finish_wait(&child->offline_waitq, &wait);
2608
				cgroup_put(child);
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
				goto retry;
			}

			/* unavailable or not enabled on the parent? */
			if (!(cgrp_dfl_root.subsys_mask & (1 << ssid)) ||
			    (cgrp->parent &&
			     !(cgrp->parent->child_subsys_mask & (1 << ssid)))) {
				ret = -ENOENT;
				goto out_unlock_tree;
			}
		} else if (disable & (1 << ssid)) {
			if (!(cgrp->child_subsys_mask & (1 << ssid))) {
				disable &= ~(1 << ssid);
				continue;
			}

			/* a child has it enabled? */
			cgroup_for_each_live_child(child, cgrp) {
				if (child->child_subsys_mask & (1 << ssid)) {
					ret = -EBUSY;
					goto out_unlock_tree;
				}
			}
		}
	}

	if (!enable && !disable) {
		ret = 0;
		goto out_unlock_tree;
	}

	if (!cgroup_lock_live_group(cgrp)) {
		ret = -ENODEV;
		goto out_unlock_tree;
	}

	/*
	 * Except for the root, child_subsys_mask must be zero for a cgroup
	 * with tasks so that child cgroups don't compete against tasks.
	 */
	if (enable && cgrp->parent && !list_empty(&cgrp->cset_links)) {
		ret = -EBUSY;
		goto out_unlock;
	}

	/*
	 * Create csses for enables and update child_subsys_mask.  This
	 * changes cgroup_e_css() results which in turn makes the
	 * subsequent cgroup_update_dfl_csses() associate all tasks in the
	 * subtree to the updated csses.
	 */
	for_each_subsys(ss, ssid) {
		if (!(enable & (1 << ssid)))
			continue;

		cgroup_for_each_live_child(child, cgrp) {
			ret = create_css(child, ss);
			if (ret)
				goto err_undo_css;
		}
	}

	cgrp->child_subsys_mask |= enable;
	cgrp->child_subsys_mask &= ~disable;

	ret = cgroup_update_dfl_csses(cgrp);
	if (ret)
		goto err_undo_css;

	/* all tasks are now migrated away from the old csses, kill them */
	for_each_subsys(ss, ssid) {
		if (!(disable & (1 << ssid)))
			continue;

		cgroup_for_each_live_child(child, cgrp)
			kill_css(cgroup_css(child, ss));
	}

	kernfs_activate(cgrp->kn);
	ret = 0;
out_unlock:
	mutex_unlock(&cgroup_mutex);
out_unlock_tree:
	mutex_unlock(&cgroup_tree_mutex);
	kernfs_unbreak_active_protection(cgrp->control_kn);
	cgroup_put(cgrp);
	return ret;

err_undo_css:
	cgrp->child_subsys_mask &= ~enable;
	cgrp->child_subsys_mask |= disable;

	for_each_subsys(ss, ssid) {
		if (!(enable & (1 << ssid)))
			continue;

		cgroup_for_each_live_child(child, cgrp) {
			struct cgroup_subsys_state *css = cgroup_css(child, ss);
			if (css)
				kill_css(css);
		}
	}
	goto out_unlock;
}

2714 2715 2716 2717 2718 2719
static int cgroup_populated_show(struct seq_file *seq, void *v)
{
	seq_printf(seq, "%d\n", (bool)seq_css(seq)->cgroup->populated_cnt);
	return 0;
}

T
Tejun Heo 已提交
2720 2721
static ssize_t cgroup_file_write(struct kernfs_open_file *of, char *buf,
				 size_t nbytes, loff_t off)
2722
{
T
Tejun Heo 已提交
2723 2724 2725
	struct cgroup *cgrp = of->kn->parent->priv;
	struct cftype *cft = of->kn->priv;
	struct cgroup_subsys_state *css;
2726
	int ret;
2727

T
Tejun Heo 已提交
2728 2729 2730 2731 2732 2733 2734 2735 2736
	/*
	 * kernfs guarantees that a file isn't deleted with operations in
	 * flight, which means that the matching css is and stays alive and
	 * doesn't need to be pinned.  The RCU locking is not necessary
	 * either.  It's just for the convenience of using cgroup_css().
	 */
	rcu_read_lock();
	css = cgroup_css(cgrp, cft->ss);
	rcu_read_unlock();
2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751

	if (cft->write_string) {
		ret = cft->write_string(css, cft, strstrip(buf));
	} else if (cft->write_u64) {
		unsigned long long v;
		ret = kstrtoull(buf, 0, &v);
		if (!ret)
			ret = cft->write_u64(css, cft, v);
	} else if (cft->write_s64) {
		long long v;
		ret = kstrtoll(buf, 0, &v);
		if (!ret)
			ret = cft->write_s64(css, cft, v);
	} else if (cft->trigger) {
		ret = cft->trigger(css, (unsigned int)cft->private);
2752
	} else {
2753
		ret = -EINVAL;
2754
	}
T
Tejun Heo 已提交
2755

2756
	return ret ?: nbytes;
2757 2758
}

2759
static void *cgroup_seqfile_start(struct seq_file *seq, loff_t *ppos)
2760
{
T
Tejun Heo 已提交
2761
	return seq_cft(seq)->seq_start(seq, ppos);
2762 2763
}

2764
static void *cgroup_seqfile_next(struct seq_file *seq, void *v, loff_t *ppos)
2765
{
T
Tejun Heo 已提交
2766
	return seq_cft(seq)->seq_next(seq, v, ppos);
2767 2768
}

2769
static void cgroup_seqfile_stop(struct seq_file *seq, void *v)
2770
{
T
Tejun Heo 已提交
2771
	seq_cft(seq)->seq_stop(seq, v);
2772 2773
}

2774
static int cgroup_seqfile_show(struct seq_file *m, void *arg)
2775
{
2776 2777
	struct cftype *cft = seq_cft(m);
	struct cgroup_subsys_state *css = seq_css(m);
2778

2779 2780
	if (cft->seq_show)
		return cft->seq_show(m, arg);
2781

2782
	if (cft->read_u64)
2783 2784 2785 2786 2787 2788
		seq_printf(m, "%llu\n", cft->read_u64(css, cft));
	else if (cft->read_s64)
		seq_printf(m, "%lld\n", cft->read_s64(css, cft));
	else
		return -EINVAL;
	return 0;
2789 2790
}

T
Tejun Heo 已提交
2791 2792 2793 2794
static struct kernfs_ops cgroup_kf_single_ops = {
	.atomic_write_len	= PAGE_SIZE,
	.write			= cgroup_file_write,
	.seq_show		= cgroup_seqfile_show,
2795 2796
};

T
Tejun Heo 已提交
2797 2798 2799 2800 2801 2802 2803 2804
static struct kernfs_ops cgroup_kf_ops = {
	.atomic_write_len	= PAGE_SIZE,
	.write			= cgroup_file_write,
	.seq_start		= cgroup_seqfile_start,
	.seq_next		= cgroup_seqfile_next,
	.seq_stop		= cgroup_seqfile_stop,
	.seq_show		= cgroup_seqfile_show,
};
2805 2806 2807 2808

/*
 * cgroup_rename - Only allow simple rename of directories in place.
 */
T
Tejun Heo 已提交
2809 2810
static int cgroup_rename(struct kernfs_node *kn, struct kernfs_node *new_parent,
			 const char *new_name_str)
2811
{
T
Tejun Heo 已提交
2812
	struct cgroup *cgrp = kn->priv;
2813 2814
	int ret;

T
Tejun Heo 已提交
2815
	if (kernfs_type(kn) != KERNFS_DIR)
2816
		return -ENOTDIR;
T
Tejun Heo 已提交
2817
	if (kn->parent != new_parent)
2818
		return -EIO;
2819

2820 2821 2822 2823 2824 2825
	/*
	 * This isn't a proper migration and its usefulness is very
	 * limited.  Disallow if sane_behavior.
	 */
	if (cgroup_sane_behavior(cgrp))
		return -EPERM;
L
Li Zefan 已提交
2826

2827 2828 2829 2830 2831 2832 2833
	/*
	 * We're gonna grab cgroup_tree_mutex which nests outside kernfs
	 * active_ref.  kernfs_rename() doesn't require active_ref
	 * protection.  Break them before grabbing cgroup_tree_mutex.
	 */
	kernfs_break_active_protection(new_parent);
	kernfs_break_active_protection(kn);
L
Li Zefan 已提交
2834

T
Tejun Heo 已提交
2835 2836
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);
L
Li Zefan 已提交
2837

T
Tejun Heo 已提交
2838
	ret = kernfs_rename(kn, new_parent, new_name_str);
L
Li Zefan 已提交
2839

T
Tejun Heo 已提交
2840 2841
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&cgroup_tree_mutex);
2842 2843 2844

	kernfs_unbreak_active_protection(kn);
	kernfs_unbreak_active_protection(new_parent);
T
Tejun Heo 已提交
2845
	return ret;
L
Li Zefan 已提交
2846 2847
}

2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
/* set uid and gid of cgroup dirs and files to that of the creator */
static int cgroup_kn_set_ugid(struct kernfs_node *kn)
{
	struct iattr iattr = { .ia_valid = ATTR_UID | ATTR_GID,
			       .ia_uid = current_fsuid(),
			       .ia_gid = current_fsgid(), };

	if (uid_eq(iattr.ia_uid, GLOBAL_ROOT_UID) &&
	    gid_eq(iattr.ia_gid, GLOBAL_ROOT_GID))
		return 0;

	return kernfs_setattr(kn, &iattr);
}

2862
static int cgroup_add_file(struct cgroup *cgrp, struct cftype *cft)
2863
{
T
Tejun Heo 已提交
2864
	char name[CGROUP_FILE_NAME_MAX];
T
Tejun Heo 已提交
2865 2866
	struct kernfs_node *kn;
	struct lock_class_key *key = NULL;
2867
	int ret;
T
Tejun Heo 已提交
2868

T
Tejun Heo 已提交
2869 2870 2871 2872 2873 2874
#ifdef CONFIG_DEBUG_LOCK_ALLOC
	key = &cft->lockdep_key;
#endif
	kn = __kernfs_create_file(cgrp->kn, cgroup_file_name(cgrp, cft, name),
				  cgroup_file_mode(cft), 0, cft->kf_ops, cft,
				  NULL, false, key);
2875 2876 2877 2878
	if (IS_ERR(kn))
		return PTR_ERR(kn);

	ret = cgroup_kn_set_ugid(kn);
2879
	if (ret) {
2880
		kernfs_remove(kn);
2881 2882 2883 2884 2885
		return ret;
	}

	if (cft->seq_show == cgroup_subtree_control_show)
		cgrp->control_kn = kn;
2886 2887
	else if (cft->seq_show == cgroup_populated_show)
		cgrp->populated_kn = kn;
2888
	return 0;
2889 2890
}

2891 2892 2893 2894 2895 2896 2897
/**
 * 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.
2898 2899 2900
 * For removals, this function never fails.  If addition fails, this
 * function doesn't remove files already added.  The caller is responsible
 * for cleaning up.
2901
 */
2902 2903
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add)
2904
{
A
Aristeu Rozanski 已提交
2905
	struct cftype *cft;
2906 2907
	int ret;

T
Tejun Heo 已提交
2908
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
2909 2910

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2911
		/* does cft->flags tell us to skip this file on @cgrp? */
T
Tejun Heo 已提交
2912 2913
		if ((cft->flags & CFTYPE_ONLY_ON_DFL) && !cgroup_on_dfl(cgrp))
			continue;
2914 2915
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2916 2917 2918 2919 2920
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2921
		if (is_add) {
2922
			ret = cgroup_add_file(cgrp, cft);
2923
			if (ret) {
2924 2925
				pr_warn("%s: failed to add %s, err=%d\n",
					__func__, cft->name, ret);
2926 2927
				return ret;
			}
2928 2929
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2930
		}
2931
	}
2932
	return 0;
2933 2934
}

2935
static int cgroup_apply_cftypes(struct cftype *cfts, bool is_add)
2936 2937
{
	LIST_HEAD(pending);
2938
	struct cgroup_subsys *ss = cfts[0].ss;
2939
	struct cgroup *root = &ss->root->cgrp;
2940
	struct cgroup_subsys_state *css;
2941
	int ret = 0;
2942

2943
	lockdep_assert_held(&cgroup_tree_mutex);
2944

2945
	/* add/rm files for all cgroups created before */
2946
	css_for_each_descendant_pre(css, cgroup_css(root, ss)) {
2947 2948
		struct cgroup *cgrp = css->cgroup;

2949 2950 2951
		if (cgroup_is_dead(cgrp))
			continue;

2952
		ret = cgroup_addrm_files(cgrp, cfts, is_add);
2953 2954
		if (ret)
			break;
2955
	}
2956 2957 2958

	if (is_add && !ret)
		kernfs_activate(root->kn);
2959
	return ret;
2960 2961
}

2962
static void cgroup_exit_cftypes(struct cftype *cfts)
2963
{
2964
	struct cftype *cft;
2965

T
Tejun Heo 已提交
2966 2967 2968 2969 2970
	for (cft = cfts; cft->name[0] != '\0'; cft++) {
		/* free copy for custom atomic_write_len, see init_cftypes() */
		if (cft->max_write_len && cft->max_write_len != PAGE_SIZE)
			kfree(cft->kf_ops);
		cft->kf_ops = NULL;
2971
		cft->ss = NULL;
T
Tejun Heo 已提交
2972
	}
2973 2974
}

T
Tejun Heo 已提交
2975
static int cgroup_init_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2976 2977 2978
{
	struct cftype *cft;

T
Tejun Heo 已提交
2979 2980 2981
	for (cft = cfts; cft->name[0] != '\0'; cft++) {
		struct kernfs_ops *kf_ops;

T
Tejun Heo 已提交
2982 2983
		WARN_ON(cft->ss || cft->kf_ops);

T
Tejun Heo 已提交
2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000
		if (cft->seq_start)
			kf_ops = &cgroup_kf_ops;
		else
			kf_ops = &cgroup_kf_single_ops;

		/*
		 * Ugh... if @cft wants a custom max_write_len, we need to
		 * make a copy of kf_ops to set its atomic_write_len.
		 */
		if (cft->max_write_len && cft->max_write_len != PAGE_SIZE) {
			kf_ops = kmemdup(kf_ops, sizeof(*kf_ops), GFP_KERNEL);
			if (!kf_ops) {
				cgroup_exit_cftypes(cfts);
				return -ENOMEM;
			}
			kf_ops->atomic_write_len = cft->max_write_len;
		}
3001

T
Tejun Heo 已提交
3002
		cft->kf_ops = kf_ops;
3003
		cft->ss = ss;
T
Tejun Heo 已提交
3004
	}
3005

T
Tejun Heo 已提交
3006
	return 0;
3007 3008
}

3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019
static int cgroup_rm_cftypes_locked(struct cftype *cfts)
{
	lockdep_assert_held(&cgroup_tree_mutex);

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

	list_del(&cfts->node);
	cgroup_apply_cftypes(cfts, false);
	cgroup_exit_cftypes(cfts);
	return 0;
3020 3021
}

3022 3023 3024 3025
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
3026 3027 3028
 * 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.
3029 3030
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
3031
 * registered.
3032
 */
3033
int cgroup_rm_cftypes(struct cftype *cfts)
3034
{
3035
	int ret;
3036

3037 3038 3039 3040
	mutex_lock(&cgroup_tree_mutex);
	ret = cgroup_rm_cftypes_locked(cfts);
	mutex_unlock(&cgroup_tree_mutex);
	return ret;
T
Tejun Heo 已提交
3041 3042
}

3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056
/**
 * 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
3057
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
3058
{
3059
	int ret;
3060

3061 3062
	if (!cfts || cfts[0].name[0] == '\0')
		return 0;
3063

T
Tejun Heo 已提交
3064 3065 3066
	ret = cgroup_init_cftypes(ss, cfts);
	if (ret)
		return ret;
3067

3068 3069
	mutex_lock(&cgroup_tree_mutex);

T
Tejun Heo 已提交
3070
	list_add_tail(&cfts->node, &ss->cfts);
3071
	ret = cgroup_apply_cftypes(cfts, true);
3072
	if (ret)
3073
		cgroup_rm_cftypes_locked(cfts);
3074

3075
	mutex_unlock(&cgroup_tree_mutex);
3076
	return ret;
3077 3078
}

L
Li Zefan 已提交
3079 3080 3081 3082 3083 3084
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
3085
static int cgroup_task_count(const struct cgroup *cgrp)
3086 3087
{
	int count = 0;
3088
	struct cgrp_cset_link *link;
3089

3090
	down_read(&css_set_rwsem);
3091 3092
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
3093
	up_read(&css_set_rwsem);
3094 3095 3096
	return count;
}

3097
/**
3098 3099 3100
 * css_next_child - find the next child of a given css
 * @pos_css: the current position (%NULL to initiate traversal)
 * @parent_css: css whose children to walk
3101
 *
3102
 * This function returns the next child of @parent_css and should be called
3103 3104 3105
 * under either cgroup_mutex or RCU read lock.  The only requirement is
 * that @parent_css and @pos_css are accessible.  The next sibling is
 * guaranteed to be returned regardless of their states.
3106
 */
3107 3108 3109
struct cgroup_subsys_state *
css_next_child(struct cgroup_subsys_state *pos_css,
	       struct cgroup_subsys_state *parent_css)
3110
{
3111 3112
	struct cgroup *pos = pos_css ? pos_css->cgroup : NULL;
	struct cgroup *cgrp = parent_css->cgroup;
3113 3114
	struct cgroup *next;

T
Tejun Heo 已提交
3115
	cgroup_assert_mutexes_or_rcu_locked();
3116 3117 3118 3119

	/*
	 * @pos could already have been removed.  Once a cgroup is removed,
	 * its ->sibling.next is no longer updated when its next sibling
3120 3121 3122 3123 3124 3125 3126
	 * 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.
3127 3128 3129 3130 3131 3132 3133 3134
	 *
	 * If @pos is dead, its next pointer can't be dereferenced;
	 * however, as each cgroup is given a monotonically increasing
	 * unique serial number and always appended to the sibling list,
	 * the next one can be found by walking the parent's children until
	 * we see a cgroup with higher serial number than @pos's.  While
	 * this path can be slower, it's taken only when either the current
	 * cgroup is removed or iteration and removal race.
3135
	 */
3136 3137 3138
	if (!pos) {
		next = list_entry_rcu(cgrp->children.next, struct cgroup, sibling);
	} else if (likely(!cgroup_is_dead(pos))) {
3139
		next = list_entry_rcu(pos->sibling.next, struct cgroup, sibling);
3140 3141 3142 3143
	} else {
		list_for_each_entry_rcu(next, &cgrp->children, sibling)
			if (next->serial_nr > pos->serial_nr)
				break;
3144 3145
	}

3146 3147 3148 3149 3150 3151 3152
	/*
	 * @next, if not pointing to the head, can be dereferenced and is
	 * the next sibling; however, it might have @ss disabled.  If so,
	 * fast-forward to the next enabled one.
	 */
	while (&next->sibling != &cgrp->children) {
		struct cgroup_subsys_state *next_css = cgroup_css(next, parent_css->ss);
3153

3154 3155 3156 3157 3158
		if (next_css)
			return next_css;
		next = list_entry_rcu(next->sibling.next, struct cgroup, sibling);
	}
	return NULL;
3159 3160
}

3161
/**
3162
 * css_next_descendant_pre - find the next descendant for pre-order walk
3163
 * @pos: the current position (%NULL to initiate traversal)
3164
 * @root: css whose descendants to walk
3165
 *
3166
 * To be used by css_for_each_descendant_pre().  Find the next descendant
3167 3168
 * to visit for pre-order traversal of @root's descendants.  @root is
 * included in the iteration and the first node to be visited.
3169
 *
3170 3171 3172 3173
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct next descendant as long
 * as both @pos and @root are accessible and @pos is a descendant of @root.
3174
 */
3175 3176 3177
struct cgroup_subsys_state *
css_next_descendant_pre(struct cgroup_subsys_state *pos,
			struct cgroup_subsys_state *root)
3178
{
3179
	struct cgroup_subsys_state *next;
3180

T
Tejun Heo 已提交
3181
	cgroup_assert_mutexes_or_rcu_locked();
3182

3183
	/* if first iteration, visit @root */
3184
	if (!pos)
3185
		return root;
3186 3187

	/* visit the first child if exists */
3188
	next = css_next_child(NULL, pos);
3189 3190 3191 3192
	if (next)
		return next;

	/* no child, visit my or the closest ancestor's next sibling */
3193 3194
	while (pos != root) {
		next = css_next_child(pos, css_parent(pos));
3195
		if (next)
3196
			return next;
3197
		pos = css_parent(pos);
3198
	}
3199 3200 3201 3202

	return NULL;
}

3203
/**
3204 3205
 * css_rightmost_descendant - return the rightmost descendant of a css
 * @pos: css of interest
3206
 *
3207 3208
 * Return the rightmost descendant of @pos.  If there's no descendant, @pos
 * is returned.  This can be used during pre-order traversal to skip
3209
 * subtree of @pos.
3210
 *
3211 3212 3213 3214
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct rightmost descendant as
 * long as @pos is accessible.
3215
 */
3216 3217
struct cgroup_subsys_state *
css_rightmost_descendant(struct cgroup_subsys_state *pos)
3218
{
3219
	struct cgroup_subsys_state *last, *tmp;
3220

T
Tejun Heo 已提交
3221
	cgroup_assert_mutexes_or_rcu_locked();
3222 3223 3224 3225 3226

	do {
		last = pos;
		/* ->prev isn't RCU safe, walk ->next till the end */
		pos = NULL;
3227
		css_for_each_child(tmp, last)
3228 3229 3230 3231 3232 3233
			pos = tmp;
	} while (pos);

	return last;
}

3234 3235
static struct cgroup_subsys_state *
css_leftmost_descendant(struct cgroup_subsys_state *pos)
3236
{
3237
	struct cgroup_subsys_state *last;
3238 3239 3240

	do {
		last = pos;
3241
		pos = css_next_child(NULL, pos);
3242 3243 3244 3245 3246 3247
	} while (pos);

	return last;
}

/**
3248
 * css_next_descendant_post - find the next descendant for post-order walk
3249
 * @pos: the current position (%NULL to initiate traversal)
3250
 * @root: css whose descendants to walk
3251
 *
3252
 * To be used by css_for_each_descendant_post().  Find the next descendant
3253 3254
 * to visit for post-order traversal of @root's descendants.  @root is
 * included in the iteration and the last node to be visited.
3255
 *
3256 3257 3258 3259 3260
 * While this function requires cgroup_mutex or RCU read locking, it
 * doesn't require the whole traversal to be contained in a single critical
 * section.  This function will return the correct next descendant as long
 * as both @pos and @cgroup are accessible and @pos is a descendant of
 * @cgroup.
3261
 */
3262 3263 3264
struct cgroup_subsys_state *
css_next_descendant_post(struct cgroup_subsys_state *pos,
			 struct cgroup_subsys_state *root)
3265
{
3266
	struct cgroup_subsys_state *next;
3267

T
Tejun Heo 已提交
3268
	cgroup_assert_mutexes_or_rcu_locked();
3269

3270 3271 3272
	/* if first iteration, visit leftmost descendant which may be @root */
	if (!pos)
		return css_leftmost_descendant(root);
3273

3274 3275 3276 3277
	/* if we visited @root, we're done */
	if (pos == root)
		return NULL;

3278
	/* if there's an unvisited sibling, visit its leftmost descendant */
3279
	next = css_next_child(pos, css_parent(pos));
3280
	if (next)
3281
		return css_leftmost_descendant(next);
3282 3283

	/* no sibling left, visit parent */
3284
	return css_parent(pos);
3285 3286
}

3287
/**
3288
 * css_advance_task_iter - advance a task itererator to the next css_set
3289 3290 3291
 * @it: the iterator to advance
 *
 * Advance @it to the next css_set to walk.
3292
 */
3293
static void css_advance_task_iter(struct css_task_iter *it)
3294
{
T
Tejun Heo 已提交
3295
	struct list_head *l = it->cset_pos;
3296 3297 3298 3299 3300 3301
	struct cgrp_cset_link *link;
	struct css_set *cset;

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
T
Tejun Heo 已提交
3302 3303
		if (l == it->cset_head) {
			it->cset_pos = NULL;
3304 3305
			return;
		}
3306 3307 3308 3309 3310 3311 3312 3313

		if (it->ss) {
			cset = container_of(l, struct css_set,
					    e_cset_node[it->ss->id]);
		} else {
			link = list_entry(l, struct cgrp_cset_link, cset_link);
			cset = link->cset;
		}
T
Tejun Heo 已提交
3314 3315
	} while (list_empty(&cset->tasks) && list_empty(&cset->mg_tasks));

T
Tejun Heo 已提交
3316
	it->cset_pos = l;
T
Tejun Heo 已提交
3317 3318

	if (!list_empty(&cset->tasks))
T
Tejun Heo 已提交
3319
		it->task_pos = cset->tasks.next;
T
Tejun Heo 已提交
3320
	else
T
Tejun Heo 已提交
3321 3322 3323 3324
		it->task_pos = cset->mg_tasks.next;

	it->tasks_head = &cset->tasks;
	it->mg_tasks_head = &cset->mg_tasks;
3325 3326
}

3327
/**
3328 3329
 * css_task_iter_start - initiate task iteration
 * @css: the css to walk tasks of
3330 3331
 * @it: the task iterator to use
 *
3332 3333 3334 3335
 * Initiate iteration through the tasks of @css.  The caller can call
 * css_task_iter_next() to walk through the tasks until the function
 * returns NULL.  On completion of iteration, css_task_iter_end() must be
 * called.
3336 3337 3338 3339 3340
 *
 * Note that this function acquires a lock which is released when the
 * iteration finishes.  The caller can't sleep while iteration is in
 * progress.
 */
3341 3342
void css_task_iter_start(struct cgroup_subsys_state *css,
			 struct css_task_iter *it)
3343
	__acquires(css_set_rwsem)
3344
{
3345 3346
	/* no one should try to iterate before mounting cgroups */
	WARN_ON_ONCE(!use_task_css_set_links);
3347

3348
	down_read(&css_set_rwsem);
3349

3350 3351 3352 3353 3354 3355 3356
	it->ss = css->ss;

	if (it->ss)
		it->cset_pos = &css->cgroup->e_csets[css->ss->id];
	else
		it->cset_pos = &css->cgroup->cset_links;

T
Tejun Heo 已提交
3357
	it->cset_head = it->cset_pos;
3358

3359
	css_advance_task_iter(it);
3360 3361
}

3362
/**
3363
 * css_task_iter_next - return the next task for the iterator
3364 3365 3366
 * @it: the task iterator being iterated
 *
 * The "next" function for task iteration.  @it should have been
3367 3368
 * initialized via css_task_iter_start().  Returns NULL when the iteration
 * reaches the end.
3369
 */
3370
struct task_struct *css_task_iter_next(struct css_task_iter *it)
3371 3372
{
	struct task_struct *res;
T
Tejun Heo 已提交
3373
	struct list_head *l = it->task_pos;
3374 3375

	/* If the iterator cg is NULL, we have no tasks */
T
Tejun Heo 已提交
3376
	if (!it->cset_pos)
3377 3378
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
T
Tejun Heo 已提交
3379 3380 3381 3382 3383 3384

	/*
	 * Advance iterator to find next entry.  cset->tasks is consumed
	 * first and then ->mg_tasks.  After ->mg_tasks, we move onto the
	 * next cset.
	 */
3385
	l = l->next;
T
Tejun Heo 已提交
3386

T
Tejun Heo 已提交
3387 3388
	if (l == it->tasks_head)
		l = it->mg_tasks_head->next;
T
Tejun Heo 已提交
3389

T
Tejun Heo 已提交
3390
	if (l == it->mg_tasks_head)
3391
		css_advance_task_iter(it);
T
Tejun Heo 已提交
3392
	else
T
Tejun Heo 已提交
3393
		it->task_pos = l;
T
Tejun Heo 已提交
3394

3395 3396 3397
	return res;
}

3398
/**
3399
 * css_task_iter_end - finish task iteration
3400 3401
 * @it: the task iterator to finish
 *
3402
 * Finish task iteration started by css_task_iter_start().
3403
 */
3404
void css_task_iter_end(struct css_task_iter *it)
3405
	__releases(css_set_rwsem)
3406
{
3407
	up_read(&css_set_rwsem);
3408 3409 3410
}

/**
3411 3412 3413
 * 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
3414
 *
3415 3416 3417 3418 3419
 * Locking rules between cgroup_post_fork() and the migration path
 * guarantee that, if a task is forking while being migrated, the new child
 * is guaranteed to be either visible in the source cgroup after the
 * parent's migration is complete or put into the target cgroup.  No task
 * can slip out of migration through forking.
3420
 */
3421
int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from)
3422
{
3423 3424
	LIST_HEAD(preloaded_csets);
	struct cgrp_cset_link *link;
3425
	struct css_task_iter it;
3426
	struct task_struct *task;
3427
	int ret;
3428

3429
	mutex_lock(&cgroup_mutex);
3430

3431 3432 3433 3434 3435
	/* all tasks in @from are being moved, all csets are source */
	down_read(&css_set_rwsem);
	list_for_each_entry(link, &from->cset_links, cset_link)
		cgroup_migrate_add_src(link->cset, to, &preloaded_csets);
	up_read(&css_set_rwsem);
3436

3437 3438 3439
	ret = cgroup_migrate_prepare_dst(to, &preloaded_csets);
	if (ret)
		goto out_err;
3440

3441 3442 3443 3444
	/*
	 * Migrate tasks one-by-one until @form is empty.  This fails iff
	 * ->can_attach() fails.
	 */
3445 3446 3447 3448 3449 3450 3451 3452
	do {
		css_task_iter_start(&from->dummy_css, &it);
		task = css_task_iter_next(&it);
		if (task)
			get_task_struct(task);
		css_task_iter_end(&it);

		if (task) {
3453
			ret = cgroup_migrate(to, task, false);
3454 3455 3456
			put_task_struct(task);
		}
	} while (task && !ret);
3457 3458
out_err:
	cgroup_migrate_finish(&preloaded_csets);
T
Tejun Heo 已提交
3459
	mutex_unlock(&cgroup_mutex);
3460
	return ret;
3461 3462
}

3463
/*
3464
 * Stuff for reading the 'tasks'/'procs' files.
3465 3466 3467 3468 3469 3470 3471 3472
 *
 * 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.
 *
 */

3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498
/* which pidlist file are we talking about? */
enum cgroup_filetype {
	CGROUP_FILE_PROCS,
	CGROUP_FILE_TASKS,
};

/*
 * A pidlist is a list of pids that virtually represents the contents of one
 * of the cgroup files ("procs" or "tasks"). We keep a list of such pidlists,
 * a pair (one each for procs, tasks) for each pid namespace that's relevant
 * to the cgroup.
 */
struct cgroup_pidlist {
	/*
	 * used to find which pidlist is wanted. doesn't change as long as
	 * this particular list stays in the list.
	*/
	struct { enum cgroup_filetype type; struct pid_namespace *ns; } key;
	/* array of xids */
	pid_t *list;
	/* how many elements the above list has */
	int length;
	/* each of these stored in a list by its cgroup */
	struct list_head links;
	/* pointer to the cgroup we belong to, for list removal purposes */
	struct cgroup *owner;
3499 3500
	/* for delayed destruction */
	struct delayed_work destroy_dwork;
3501 3502
};

3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515
/*
 * 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);
}
3516

3517 3518 3519 3520 3521 3522 3523 3524
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}

3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551
/*
 * Used to destroy all pidlists lingering waiting for destroy timer.  None
 * should be left afterwards.
 */
static void cgroup_pidlist_destroy_all(struct cgroup *cgrp)
{
	struct cgroup_pidlist *l, *tmp_l;

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

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

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

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

	/*
3552 3553
	 * Destroy iff we didn't get queued again.  The state won't change
	 * as destroy_dwork can only be queued while locked.
3554
	 */
3555
	if (!delayed_work_pending(dwork)) {
3556 3557 3558 3559 3560 3561 3562 3563 3564 3565
		list_del(&l->links);
		pidlist_free(l->list);
		put_pid_ns(l->key.ns);
		tofree = l;
	}

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

3566
/*
3567
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3568
 * Returns the number of unique elements.
3569
 */
3570
static int pidlist_uniq(pid_t *list, int length)
3571
{
3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595
	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;
}

3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628
/*
 * The two pid files - task and cgroup.procs - guaranteed that the result
 * is sorted, which forced this whole pidlist fiasco.  As pid order is
 * different per namespace, each namespace needs differently sorted list,
 * making it impossible to use, for example, single rbtree of member tasks
 * sorted by task pointer.  As pidlists can be fairly large, allocating one
 * per open file is dangerous, so cgroup had to implement shared pool of
 * pidlists keyed by cgroup and namespace.
 *
 * All this extra complexity was caused by the original implementation
 * committing to an entirely unnecessary property.  In the long term, we
 * want to do away with it.  Explicitly scramble sort order if
 * sane_behavior so that no such expectation exists in the new interface.
 *
 * Scrambling is done by swapping every two consecutive bits, which is
 * non-identity one-to-one mapping which disturbs sort order sufficiently.
 */
static pid_t pid_fry(pid_t pid)
{
	unsigned a = pid & 0x55555555;
	unsigned b = pid & 0xAAAAAAAA;

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

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

3629 3630 3631 3632 3633
static int cmppid(const void *a, const void *b)
{
	return *(pid_t *)a - *(pid_t *)b;
}

3634 3635 3636 3637 3638
static int fried_cmppid(const void *a, const void *b)
{
	return pid_fry(*(pid_t *)a) - pid_fry(*(pid_t *)b);
}

T
Tejun Heo 已提交
3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
						  enum cgroup_filetype type)
{
	struct cgroup_pidlist *l;
	/* don't need task_nsproxy() if we're looking at ourself */
	struct pid_namespace *ns = task_active_pid_ns(current);

	lockdep_assert_held(&cgrp->pidlist_mutex);

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

3654 3655 3656 3657 3658 3659
/*
 * find the appropriate pidlist for our purpose (given procs vs tasks)
 * returns with the lock on that pidlist already held, and takes care
 * of the use count, or returns NULL with no locks held if we're out of
 * memory.
 */
T
Tejun Heo 已提交
3660 3661
static struct cgroup_pidlist *cgroup_pidlist_find_create(struct cgroup *cgrp,
						enum cgroup_filetype type)
3662 3663
{
	struct cgroup_pidlist *l;
3664

T
Tejun Heo 已提交
3665 3666 3667 3668 3669 3670
	lockdep_assert_held(&cgrp->pidlist_mutex);

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

3671
	/* entry not found; create a new one */
3672
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
T
Tejun Heo 已提交
3673
	if (!l)
3674
		return l;
T
Tejun Heo 已提交
3675

3676
	INIT_DELAYED_WORK(&l->destroy_dwork, cgroup_pidlist_destroy_work_fn);
3677
	l->key.type = type;
T
Tejun Heo 已提交
3678 3679
	/* don't need task_nsproxy() if we're looking at ourself */
	l->key.ns = get_pid_ns(task_active_pid_ns(current));
3680 3681 3682 3683 3684
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	return l;
}

3685 3686 3687
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3688 3689
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3690 3691 3692 3693
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3694
	struct css_task_iter it;
3695
	struct task_struct *tsk;
3696 3697
	struct cgroup_pidlist *l;

3698 3699
	lockdep_assert_held(&cgrp->pidlist_mutex);

3700 3701 3702 3703 3704 3705 3706
	/*
	 * 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);
3707
	array = pidlist_allocate(length);
3708 3709 3710
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3711 3712
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
3713
		if (unlikely(n == length))
3714
			break;
3715
		/* get tgid or pid for procs or tasks file respectively */
3716 3717 3718 3719
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3720 3721
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3722
	}
3723
	css_task_iter_end(&it);
3724 3725
	length = n;
	/* now sort & (if procs) strip out duplicates */
3726 3727 3728 3729
	if (cgroup_sane_behavior(cgrp))
		sort(array, length, sizeof(pid_t), fried_cmppid, NULL);
	else
		sort(array, length, sizeof(pid_t), cmppid, NULL);
3730
	if (type == CGROUP_FILE_PROCS)
3731
		length = pidlist_uniq(array, length);
T
Tejun Heo 已提交
3732 3733

	l = cgroup_pidlist_find_create(cgrp, type);
3734
	if (!l) {
T
Tejun Heo 已提交
3735
		mutex_unlock(&cgrp->pidlist_mutex);
3736
		pidlist_free(array);
3737
		return -ENOMEM;
3738
	}
T
Tejun Heo 已提交
3739 3740

	/* store array, freeing old if necessary */
3741
	pidlist_free(l->list);
3742 3743
	l->list = array;
	l->length = length;
3744
	*lp = l;
3745
	return 0;
3746 3747
}

B
Balbir Singh 已提交
3748
/**
L
Li Zefan 已提交
3749
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3750 3751 3752
 * @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
3753 3754 3755
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3756 3757 3758
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
T
Tejun Heo 已提交
3759
	struct kernfs_node *kn = kernfs_node_from_dentry(dentry);
3760
	struct cgroup *cgrp;
3761
	struct css_task_iter it;
B
Balbir Singh 已提交
3762
	struct task_struct *tsk;
3763

T
Tejun Heo 已提交
3764 3765 3766 3767 3768
	/* it should be kernfs_node belonging to cgroupfs and is a directory */
	if (dentry->d_sb->s_type != &cgroup_fs_type || !kn ||
	    kernfs_type(kn) != KERNFS_DIR)
		return -EINVAL;

3769 3770
	mutex_lock(&cgroup_mutex);

B
Balbir Singh 已提交
3771
	/*
T
Tejun Heo 已提交
3772 3773 3774
	 * We aren't being called from kernfs and there's no guarantee on
	 * @kn->priv's validity.  For this and css_tryget_from_dir(),
	 * @kn->priv is RCU safe.  Let's do the RCU dancing.
B
Balbir Singh 已提交
3775
	 */
T
Tejun Heo 已提交
3776 3777
	rcu_read_lock();
	cgrp = rcu_dereference(kn->priv);
3778
	if (!cgrp || cgroup_is_dead(cgrp)) {
T
Tejun Heo 已提交
3779
		rcu_read_unlock();
3780
		mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
3781 3782
		return -ENOENT;
	}
3783
	rcu_read_unlock();
B
Balbir Singh 已提交
3784

3785 3786
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
B
Balbir Singh 已提交
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805
		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;
		}
	}
3806
	css_task_iter_end(&it);
B
Balbir Singh 已提交
3807

3808
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
3809
	return 0;
B
Balbir Singh 已提交
3810 3811
}

3812

3813
/*
3814
 * seq_file methods for the tasks/procs files. The seq_file position is the
3815
 * next pid to display; the seq_file iterator is a pointer to the pid
3816
 * in the cgroup->l->list array.
3817
 */
3818

3819
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3820
{
3821 3822 3823 3824 3825 3826
	/*
	 * 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
	 */
T
Tejun Heo 已提交
3827
	struct kernfs_open_file *of = s->private;
3828
	struct cgroup *cgrp = seq_css(s)->cgroup;
3829
	struct cgroup_pidlist *l;
3830
	enum cgroup_filetype type = seq_cft(s)->private;
3831
	int index = 0, pid = *pos;
3832 3833 3834 3835 3836
	int *iter, ret;

	mutex_lock(&cgrp->pidlist_mutex);

	/*
3837
	 * !NULL @of->priv indicates that this isn't the first start()
3838
	 * after open.  If the matching pidlist is around, we can use that.
3839
	 * Look for it.  Note that @of->priv can't be used directly.  It
3840 3841
	 * could already have been destroyed.
	 */
3842 3843
	if (of->priv)
		of->priv = cgroup_pidlist_find(cgrp, type);
3844 3845 3846 3847 3848

	/*
	 * Either this is the first start() after open or the matching
	 * pidlist has been destroyed inbetween.  Create a new one.
	 */
3849 3850 3851
	if (!of->priv) {
		ret = pidlist_array_load(cgrp, type,
					 (struct cgroup_pidlist **)&of->priv);
3852 3853 3854
		if (ret)
			return ERR_PTR(ret);
	}
3855
	l = of->priv;
3856 3857

	if (pid) {
3858
		int end = l->length;
S
Stephen Rothwell 已提交
3859

3860 3861
		while (index < end) {
			int mid = (index + end) / 2;
3862
			if (cgroup_pid_fry(cgrp, l->list[mid]) == pid) {
3863 3864
				index = mid;
				break;
3865
			} else if (cgroup_pid_fry(cgrp, l->list[mid]) <= pid)
3866 3867 3868 3869 3870 3871
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3872
	if (index >= l->length)
3873 3874
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3875
	iter = l->list + index;
3876
	*pos = cgroup_pid_fry(cgrp, *iter);
3877 3878 3879
	return iter;
}

3880
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3881
{
T
Tejun Heo 已提交
3882
	struct kernfs_open_file *of = s->private;
3883
	struct cgroup_pidlist *l = of->priv;
3884

3885 3886
	if (l)
		mod_delayed_work(cgroup_pidlist_destroy_wq, &l->destroy_dwork,
3887
				 CGROUP_PIDLIST_DESTROY_DELAY);
3888
	mutex_unlock(&seq_css(s)->cgroup->pidlist_mutex);
3889 3890
}

3891
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3892
{
T
Tejun Heo 已提交
3893
	struct kernfs_open_file *of = s->private;
3894
	struct cgroup_pidlist *l = of->priv;
3895 3896
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3897 3898 3899 3900 3901 3902 3903 3904
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
3905
		*pos = cgroup_pid_fry(seq_css(s)->cgroup, *p);
3906 3907 3908 3909
		return p;
	}
}

3910
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3911 3912 3913
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3914

3915 3916
static u64 cgroup_read_notify_on_release(struct cgroup_subsys_state *css,
					 struct cftype *cft)
3917
{
3918
	return notify_on_release(css->cgroup);
3919 3920
}

3921 3922
static int cgroup_write_notify_on_release(struct cgroup_subsys_state *css,
					  struct cftype *cft, u64 val)
3923
{
3924
	clear_bit(CGRP_RELEASABLE, &css->cgroup->flags);
3925
	if (val)
3926
		set_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3927
	else
3928
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3929 3930 3931
	return 0;
}

3932 3933
static u64 cgroup_clone_children_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
3934
{
3935
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3936 3937
}

3938 3939
static int cgroup_clone_children_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
3940 3941
{
	if (val)
3942
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3943
	else
3944
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3945 3946 3947
	return 0;
}

3948
static struct cftype cgroup_base_files[] = {
3949
	{
3950
		.name = "cgroup.procs",
3951 3952 3953 3954
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
3955
		.private = CGROUP_FILE_PROCS,
B
Ben Blum 已提交
3956 3957
		.write_u64 = cgroup_procs_write,
		.mode = S_IRUGO | S_IWUSR,
3958
	},
3959 3960
	{
		.name = "cgroup.clone_children",
3961
		.flags = CFTYPE_INSANE,
3962 3963 3964
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
3965 3966 3967
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
3968
		.seq_show = cgroup_sane_behavior_show,
3969
	},
3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985
	{
		.name = "cgroup.controllers",
		.flags = CFTYPE_ONLY_ON_DFL | CFTYPE_ONLY_ON_ROOT,
		.seq_show = cgroup_root_controllers_show,
	},
	{
		.name = "cgroup.controllers",
		.flags = CFTYPE_ONLY_ON_DFL | CFTYPE_NOT_ON_ROOT,
		.seq_show = cgroup_controllers_show,
	},
	{
		.name = "cgroup.subtree_control",
		.flags = CFTYPE_ONLY_ON_DFL,
		.seq_show = cgroup_subtree_control_show,
		.write_string = cgroup_subtree_control_write,
	},
3986 3987 3988 3989 3990
	{
		.name = "cgroup.populated",
		.flags = CFTYPE_ONLY_ON_DFL | CFTYPE_NOT_ON_ROOT,
		.seq_show = cgroup_populated_show,
	},
3991 3992 3993 3994 3995 3996 3997 3998 3999

	/*
	 * 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 */
4000 4001 4002 4003
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
4004
		.private = CGROUP_FILE_TASKS,
4005 4006 4007 4008 4009 4010 4011 4012 4013
		.write_u64 = cgroup_tasks_write,
		.mode = S_IRUGO | S_IWUSR,
	},
	{
		.name = "notify_on_release",
		.flags = CFTYPE_INSANE,
		.read_u64 = cgroup_read_notify_on_release,
		.write_u64 = cgroup_write_notify_on_release,
	},
4014 4015
	{
		.name = "release_agent",
4016
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
4017
		.seq_show = cgroup_release_agent_show,
4018
		.write_string = cgroup_release_agent_write,
4019
		.max_write_len = PATH_MAX - 1,
4020
	},
T
Tejun Heo 已提交
4021
	{ }	/* terminate */
4022 4023
};

4024
/**
4025
 * cgroup_populate_dir - create subsys files in a cgroup directory
4026 4027
 * @cgrp: target cgroup
 * @subsys_mask: mask of the subsystem ids whose files should be added
4028 4029
 *
 * On failure, no file is added.
4030
 */
4031
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned int subsys_mask)
4032 4033
{
	struct cgroup_subsys *ss;
4034
	int i, ret = 0;
4035

4036
	/* process cftsets of each subsystem */
4037
	for_each_subsys(ss, i) {
T
Tejun Heo 已提交
4038
		struct cftype *cfts;
4039

4040
		if (!(subsys_mask & (1 << i)))
4041
			continue;
4042

T
Tejun Heo 已提交
4043 4044
		list_for_each_entry(cfts, &ss->cfts, node) {
			ret = cgroup_addrm_files(cgrp, cfts, true);
4045 4046 4047
			if (ret < 0)
				goto err;
		}
4048 4049
	}
	return 0;
4050 4051 4052
err:
	cgroup_clear_dir(cgrp, subsys_mask);
	return ret;
4053 4054
}

4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076
/*
 * css destruction is four-stage process.
 *
 * 1. Destruction starts.  Killing of the percpu_ref is initiated.
 *    Implemented in kill_css().
 *
 * 2. When the percpu_ref is confirmed to be visible as killed on all CPUs
 *    and thus css_tryget() is guaranteed to fail, the css can be offlined
 *    by invoking offline_css().  After offlining, the base ref is put.
 *    Implemented in css_killed_work_fn().
 *
 * 3. When the percpu_ref reaches zero, the only possible remaining
 *    accessors are inside RCU read sections.  css_release() schedules the
 *    RCU callback.
 *
 * 4. After the grace period, the css can be freed.  Implemented in
 *    css_free_work_fn().
 *
 * It is actually hairier because both step 2 and 4 require process context
 * and thus involve punting to css->destroy_work adding two additional
 * steps to the already complex sequence.
 */
4077
static void css_free_work_fn(struct work_struct *work)
4078 4079
{
	struct cgroup_subsys_state *css =
4080
		container_of(work, struct cgroup_subsys_state, destroy_work);
4081
	struct cgroup *cgrp = css->cgroup;
4082

4083 4084 4085
	if (css->parent)
		css_put(css->parent);

4086
	css->ss->css_free(css);
T
Tejun Heo 已提交
4087
	cgroup_put(cgrp);
4088 4089
}

4090
static void css_free_rcu_fn(struct rcu_head *rcu_head)
4091 4092
{
	struct cgroup_subsys_state *css =
4093
		container_of(rcu_head, struct cgroup_subsys_state, rcu_head);
4094

4095
	INIT_WORK(&css->destroy_work, css_free_work_fn);
4096
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4097 4098
}

4099 4100 4101 4102
static void css_release(struct percpu_ref *ref)
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);
4103 4104 4105
	struct cgroup_subsys *ss = css->ss;

	cgroup_idr_remove(&ss->css_idr, css->id);
4106

4107
	call_rcu(&css->rcu_head, css_free_rcu_fn);
4108 4109
}

4110 4111
static void init_and_link_css(struct cgroup_subsys_state *css,
			      struct cgroup_subsys *ss, struct cgroup *cgrp)
4112
{
4113 4114
	cgroup_get(cgrp);

4115
	css->cgroup = cgrp;
4116
	css->ss = ss;
4117
	css->flags = 0;
4118

4119
	if (cgrp->parent) {
4120
		css->parent = cgroup_css(cgrp->parent, ss);
4121 4122
		css_get(css->parent);
	} else {
4123
		css->flags |= CSS_ROOT;
4124
	}
4125

4126
	BUG_ON(cgroup_css(cgrp, ss));
4127 4128
}

4129
/* invoke ->css_online() on a new CSS and mark it online if successful */
4130
static int online_css(struct cgroup_subsys_state *css)
4131
{
4132
	struct cgroup_subsys *ss = css->ss;
T
Tejun Heo 已提交
4133 4134
	int ret = 0;

T
Tejun Heo 已提交
4135
	lockdep_assert_held(&cgroup_tree_mutex);
4136 4137
	lockdep_assert_held(&cgroup_mutex);

4138
	if (ss->css_online)
4139
		ret = ss->css_online(css);
4140
	if (!ret) {
4141
		css->flags |= CSS_ONLINE;
4142
		css->cgroup->nr_css++;
4143
		rcu_assign_pointer(css->cgroup->subsys[ss->id], css);
4144
	}
T
Tejun Heo 已提交
4145
	return ret;
4146 4147
}

4148
/* if the CSS is online, invoke ->css_offline() on it and mark it offline */
4149
static void offline_css(struct cgroup_subsys_state *css)
4150
{
4151
	struct cgroup_subsys *ss = css->ss;
4152

T
Tejun Heo 已提交
4153
	lockdep_assert_held(&cgroup_tree_mutex);
4154 4155 4156 4157 4158
	lockdep_assert_held(&cgroup_mutex);

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

4159
	if (ss->css_offline)
4160
		ss->css_offline(css);
4161

4162
	css->flags &= ~CSS_ONLINE;
4163
	css->cgroup->nr_css--;
4164
	RCU_INIT_POINTER(css->cgroup->subsys[ss->id], NULL);
4165 4166

	wake_up_all(&css->cgroup->offline_waitq);
4167 4168
}

4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189
/**
 * create_css - create a cgroup_subsys_state
 * @cgrp: the cgroup new css will be associated with
 * @ss: the subsys of new css
 *
 * Create a new css associated with @cgrp - @ss pair.  On success, the new
 * css is online and installed in @cgrp with all interface files created.
 * Returns 0 on success, -errno on failure.
 */
static int create_css(struct cgroup *cgrp, struct cgroup_subsys *ss)
{
	struct cgroup *parent = cgrp->parent;
	struct cgroup_subsys_state *css;
	int err;

	lockdep_assert_held(&cgroup_mutex);

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

4190
	init_and_link_css(css, ss, cgrp);
4191

4192 4193
	err = percpu_ref_init(&css->refcnt, css_release);
	if (err)
4194
		goto err_free_css;
4195

4196 4197 4198 4199 4200
	err = cgroup_idr_alloc(&ss->css_idr, NULL, 2, 0, GFP_NOWAIT);
	if (err < 0)
		goto err_free_percpu_ref;
	css->id = err;

4201
	err = cgroup_populate_dir(cgrp, 1 << ss->id);
4202
	if (err)
4203 4204 4205 4206
		goto err_free_id;

	/* @css is ready to be brought online now, make it visible */
	cgroup_idr_replace(&ss->css_idr, css, css->id);
4207 4208 4209

	err = online_css(css);
	if (err)
4210
		goto err_clear_dir;
4211 4212 4213

	if (ss->broken_hierarchy && !ss->warned_broken_hierarchy &&
	    parent->parent) {
4214
		pr_warn("%s (%d) created nested cgroup for controller \"%s\" which has incomplete hierarchy support. Nested cgroups may change behavior in the future.\n",
4215
			current->comm, current->pid, ss->name);
4216
		if (!strcmp(ss->name, "memory"))
4217
			pr_warn("\"memory\" requires setting use_hierarchy to 1 on the root\n");
4218 4219 4220 4221 4222
		ss->warned_broken_hierarchy = true;
	}

	return 0;

4223
err_clear_dir:
4224
	cgroup_clear_dir(css->cgroup, 1 << css->ss->id);
4225 4226
err_free_id:
	cgroup_idr_remove(&ss->css_idr, css->id);
4227
err_free_percpu_ref:
4228
	percpu_ref_cancel_init(&css->refcnt);
4229
err_free_css:
4230
	call_rcu(&css->rcu_head, css_free_rcu_fn);
4231 4232 4233
	return err;
}

T
Tejun Heo 已提交
4234
/**
L
Li Zefan 已提交
4235 4236
 * cgroup_create - create a cgroup
 * @parent: cgroup that will be parent of the new cgroup
T
Tejun Heo 已提交
4237
 * @name: name of the new cgroup
T
Tejun Heo 已提交
4238
 * @mode: mode to set on new cgroup
4239
 */
T
Tejun Heo 已提交
4240
static long cgroup_create(struct cgroup *parent, const char *name,
T
Tejun Heo 已提交
4241
			  umode_t mode)
4242
{
4243
	struct cgroup *cgrp;
4244
	struct cgroup_root *root = parent->root;
4245
	int ssid, err;
4246
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
4247
	struct kernfs_node *kn;
4248

T
Tejun Heo 已提交
4249
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4250 4251
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4252 4253
		return -ENOMEM;

T
Tejun Heo 已提交
4254
	mutex_lock(&cgroup_tree_mutex);
4255

4256 4257 4258 4259 4260 4261 4262 4263 4264
	/*
	 * 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
4265
		goto err_unlock_tree;
4266 4267 4268 4269 4270 4271
	}

	/*
	 * Temporarily set the pointer to NULL, so idr_find() won't return
	 * a half-baked cgroup.
	 */
4272
	cgrp->id = cgroup_idr_alloc(&root->cgroup_idr, NULL, 2, 0, GFP_NOWAIT);
4273 4274 4275
	if (cgrp->id < 0) {
		err = -ENOMEM;
		goto err_unlock;
4276 4277
	}

4278
	init_cgroup_housekeeping(cgrp);
4279

4280
	cgrp->parent = parent;
4281
	cgrp->dummy_css.parent = &parent->dummy_css;
4282
	cgrp->root = parent->root;
4283

4284 4285 4286
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4287 4288
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4289

T
Tejun Heo 已提交
4290
	/* create the directory */
T
Tejun Heo 已提交
4291
	kn = kernfs_create_dir(parent->kn, name, mode, cgrp);
T
Tejun Heo 已提交
4292 4293
	if (IS_ERR(kn)) {
		err = PTR_ERR(kn);
4294
		goto err_free_id;
T
Tejun Heo 已提交
4295 4296
	}
	cgrp->kn = kn;
4297

4298
	/*
4299 4300
	 * This extra ref will be put in cgroup_free_fn() and guarantees
	 * that @cgrp->kn is always accessible.
4301
	 */
4302
	kernfs_get(kn);
4303

4304
	cgrp->serial_nr = cgroup_serial_nr_next++;
4305

4306 4307
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
4308
	atomic_inc(&root->nr_cgrps);
4309
	cgroup_get(parent);
4310

4311 4312 4313 4314
	/*
	 * @cgrp is now fully operational.  If something fails after this
	 * point, it'll be released via the normal destruction path.
	 */
4315
	cgroup_idr_replace(&root->cgroup_idr, cgrp, cgrp->id);
4316

4317 4318 4319 4320
	err = cgroup_kn_set_ugid(kn);
	if (err)
		goto err_destroy;

4321
	err = cgroup_addrm_files(cgrp, cgroup_base_files, true);
4322 4323 4324
	if (err)
		goto err_destroy;

4325
	/* let's create and online css's */
T
Tejun Heo 已提交
4326
	for_each_subsys(ss, ssid) {
4327
		if (parent->child_subsys_mask & (1 << ssid)) {
T
Tejun Heo 已提交
4328 4329 4330 4331
			err = create_css(cgrp, ss);
			if (err)
				goto err_destroy;
		}
4332
	}
4333

4334 4335 4336 4337 4338 4339
	/*
	 * On the default hierarchy, a child doesn't automatically inherit
	 * child_subsys_mask from the parent.  Each is configured manually.
	 */
	if (!cgroup_on_dfl(cgrp))
		cgrp->child_subsys_mask = parent->child_subsys_mask;
4340

T
Tejun Heo 已提交
4341 4342
	kernfs_activate(kn);

4343
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4344
	mutex_unlock(&cgroup_tree_mutex);
4345 4346 4347

	return 0;

T
Tejun Heo 已提交
4348
err_free_id:
4349
	cgroup_idr_remove(&root->cgroup_idr, cgrp->id);
4350 4351
err_unlock:
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4352 4353
err_unlock_tree:
	mutex_unlock(&cgroup_tree_mutex);
4354
	kfree(cgrp);
4355
	return err;
4356 4357 4358 4359

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4360
	mutex_unlock(&cgroup_tree_mutex);
4361
	return err;
4362 4363
}

T
Tejun Heo 已提交
4364 4365
static int cgroup_mkdir(struct kernfs_node *parent_kn, const char *name,
			umode_t mode)
4366
{
T
Tejun Heo 已提交
4367
	struct cgroup *parent = parent_kn->priv;
4368
	int ret;
4369

4370 4371 4372 4373 4374 4375 4376 4377
	/*
	 * cgroup_create() grabs cgroup_tree_mutex which nests outside
	 * kernfs active_ref and cgroup_create() already synchronizes
	 * properly against removal through cgroup_lock_live_group().
	 * Break it before calling cgroup_create().
	 */
	cgroup_get(parent);
	kernfs_break_active_protection(parent_kn);
4378

4379 4380 4381 4382 4383
	ret = cgroup_create(parent, name, mode);

	kernfs_unbreak_active_protection(parent_kn);
	cgroup_put(parent);
	return ret;
4384 4385
}

4386 4387 4388 4389 4390
/*
 * This is called when the refcnt of a css is confirmed to be killed.
 * css_tryget() is now guaranteed to fail.
 */
static void css_killed_work_fn(struct work_struct *work)
4391
{
4392 4393 4394
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, destroy_work);
	struct cgroup *cgrp = css->cgroup;
4395

T
Tejun Heo 已提交
4396
	mutex_lock(&cgroup_tree_mutex);
4397 4398
	mutex_lock(&cgroup_mutex);

4399 4400 4401 4402 4403 4404
	/*
	 * css_tryget() is guaranteed to fail now.  Tell subsystems to
	 * initate destruction.
	 */
	offline_css(css);

4405 4406 4407 4408 4409
	/*
	 * If @cgrp is marked dead, it's waiting for refs of all css's to
	 * be disabled before proceeding to the second phase of cgroup
	 * destruction.  If we are the last one, kick it off.
	 */
4410
	if (!cgrp->nr_css && cgroup_is_dead(cgrp))
4411 4412 4413
		cgroup_destroy_css_killed(cgrp);

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4414
	mutex_unlock(&cgroup_tree_mutex);
4415 4416 4417 4418 4419 4420 4421 4422 4423

	/*
	 * Put the css refs from kill_css().  Each css holds an extra
	 * reference to the cgroup's dentry and cgroup removal proceeds
	 * regardless of css refs.  On the last put of each css, whenever
	 * that may be, the extra dentry ref is put so that dentry
	 * destruction happens only after all css's are released.
	 */
	css_put(css);
4424 4425
}

4426 4427
/* css kill confirmation processing requires process context, bounce */
static void css_killed_ref_fn(struct percpu_ref *ref)
4428 4429 4430 4431
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

4432
	INIT_WORK(&css->destroy_work, css_killed_work_fn);
4433
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4434 4435
}

4436 4437 4438 4439 4440 4441 4442 4443 4444 4445
/**
 * kill_css - destroy a css
 * @css: css to destroy
 *
 * This function initiates destruction of @css by removing cgroup interface
 * files and putting its base reference.  ->css_offline() will be invoked
 * asynchronously once css_tryget() is guaranteed to fail and when the
 * reference count reaches zero, @css will be released.
 */
static void kill_css(struct cgroup_subsys_state *css)
T
Tejun Heo 已提交
4446
{
4447 4448
	lockdep_assert_held(&cgroup_tree_mutex);

T
Tejun Heo 已提交
4449 4450 4451 4452
	/*
	 * This must happen before css is disassociated with its cgroup.
	 * See seq_css() for details.
	 */
4453
	cgroup_clear_dir(css->cgroup, 1 << css->ss->id);
4454

T
Tejun Heo 已提交
4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469 4470 4471
	/*
	 * Killing would put the base ref, but we need to keep it alive
	 * until after ->css_offline().
	 */
	css_get(css);

	/*
	 * cgroup core guarantees that, by the time ->css_offline() is
	 * invoked, no new css reference will be given out via
	 * css_tryget().  We can't simply call percpu_ref_kill() and
	 * proceed to offlining css's because percpu_ref_kill() doesn't
	 * guarantee that the ref is seen as killed on all CPUs on return.
	 *
	 * Use percpu_ref_kill_and_confirm() to get notifications as each
	 * css is confirmed to be seen as killed on all CPUs.
	 */
	percpu_ref_kill_and_confirm(&css->refcnt, css_killed_ref_fn);
4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497
}

/**
 * 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.
 */
4498 4499
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4500
{
4501
	struct cgroup *child;
T
Tejun Heo 已提交
4502
	struct cgroup_subsys_state *css;
4503
	bool empty;
T
Tejun Heo 已提交
4504
	int ssid;
4505

T
Tejun Heo 已提交
4506
	lockdep_assert_held(&cgroup_tree_mutex);
4507 4508
	lockdep_assert_held(&cgroup_mutex);

4509
	/*
4510
	 * css_set_rwsem synchronizes access to ->cset_links and prevents
4511
	 * @cgrp from being removed while put_css_set() is in progress.
4512
	 */
4513
	down_read(&css_set_rwsem);
4514
	empty = list_empty(&cgrp->cset_links);
4515
	up_read(&css_set_rwsem);
4516
	if (!empty)
4517
		return -EBUSY;
L
Li Zefan 已提交
4518

4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534
	/*
	 * Make sure there's no live children.  We can't test ->children
	 * emptiness as dead children linger on it while being destroyed;
	 * otherwise, "rmdir parent/child parent" may fail with -EBUSY.
	 */
	empty = true;
	rcu_read_lock();
	list_for_each_entry_rcu(child, &cgrp->children, sibling) {
		empty = cgroup_is_dead(child);
		if (!empty)
			break;
	}
	rcu_read_unlock();
	if (!empty)
		return -EBUSY;

4535 4536 4537
	/*
	 * Mark @cgrp dead.  This prevents further task migration and child
	 * creation by disabling cgroup_lock_live_group().  Note that
4538
	 * CGRP_DEAD assertion is depended upon by css_next_child() to
4539
	 * resume iteration after dropping RCU read lock.  See
4540
	 * css_next_child() for details.
4541
	 */
4542
	set_bit(CGRP_DEAD, &cgrp->flags);
4543

4544
	/*
T
Tejun Heo 已提交
4545 4546
	 * Initiate massacre of all css's.  cgroup_destroy_css_killed()
	 * will be invoked to perform the rest of destruction once the
4547 4548
	 * percpu refs of all css's are confirmed to be killed.  This
	 * involves removing the subsystem's files, drop cgroup_mutex.
4549
	 */
4550
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4551 4552
	for_each_css(css, ssid, cgrp)
		kill_css(css);
4553
	mutex_lock(&cgroup_mutex);
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 4564 4565 4566 4567 4568 4569
	 * If @cgrp has css's attached, the second stage of cgroup
	 * destruction is kicked off from css_killed_work_fn() after the
	 * refs of all attached css's are killed.  If @cgrp doesn't have
	 * any css, we kick it off here.
	 */
	if (!cgrp->nr_css)
		cgroup_destroy_css_killed(cgrp);

T
Tejun Heo 已提交
4570 4571 4572
	/* remove @cgrp directory along with the base files */
	mutex_unlock(&cgroup_mutex);

4573
	/*
T
Tejun Heo 已提交
4574 4575 4576 4577 4578
	 * There are two control paths which try to determine cgroup from
	 * dentry without going through kernfs - cgroupstats_build() and
	 * css_tryget_from_dir().  Those are supported by RCU protecting
	 * clearing of cgrp->kn->priv backpointer, which should happen
	 * after all files under it have been removed.
4579
	 */
4580
	kernfs_remove(cgrp->kn);	/* @cgrp has an extra ref on its kn */
T
Tejun Heo 已提交
4581 4582
	RCU_INIT_POINTER(*(void __rcu __force **)&cgrp->kn->priv, NULL);

4583
	mutex_lock(&cgroup_mutex);
4584

4585 4586 4587
	return 0;
};

4588
/**
4589
 * cgroup_destroy_css_killed - the second step of cgroup destruction
4590
 * @cgrp: the cgroup whose csses have just finished offlining
4591 4592
 *
 * This function is invoked from a work item for a cgroup which is being
4593 4594 4595
 * destroyed after all css's are offlined and performs the rest of
 * destruction.  This is the second step of destruction described in the
 * comment above cgroup_destroy_locked().
4596
 */
4597
static void cgroup_destroy_css_killed(struct cgroup *cgrp)
4598 4599 4600
{
	struct cgroup *parent = cgrp->parent;

T
Tejun Heo 已提交
4601
	lockdep_assert_held(&cgroup_tree_mutex);
4602
	lockdep_assert_held(&cgroup_mutex);
4603

4604
	/* delete this cgroup from parent->children */
4605
	list_del_rcu(&cgrp->sibling);
4606

4607
	cgroup_put(cgrp);
4608

4609
	set_bit(CGRP_RELEASABLE, &parent->flags);
4610
	check_for_release(parent);
4611 4612
}

T
Tejun Heo 已提交
4613
static int cgroup_rmdir(struct kernfs_node *kn)
4614
{
T
Tejun Heo 已提交
4615 4616
	struct cgroup *cgrp = kn->priv;
	int ret = 0;
4617

T
Tejun Heo 已提交
4618 4619 4620 4621 4622 4623 4624 4625
	/*
	 * This is self-destruction but @kn can't be removed while this
	 * callback is in progress.  Let's break active protection.  Once
	 * the protection is broken, @cgrp can be destroyed at any point.
	 * Pin it so that it stays accessible.
	 */
	cgroup_get(cgrp);
	kernfs_break_active_protection(kn);
4626

T
Tejun Heo 已提交
4627
	mutex_lock(&cgroup_tree_mutex);
4628
	mutex_lock(&cgroup_mutex);
4629 4630

	/*
T
Tejun Heo 已提交
4631 4632
	 * @cgrp might already have been destroyed while we're trying to
	 * grab the mutexes.
4633
	 */
T
Tejun Heo 已提交
4634 4635
	if (!cgroup_is_dead(cgrp))
		ret = cgroup_destroy_locked(cgrp);
4636

4637
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4638
	mutex_unlock(&cgroup_tree_mutex);
4639

T
Tejun Heo 已提交
4640 4641
	kernfs_unbreak_active_protection(kn);
	cgroup_put(cgrp);
4642
	return ret;
4643 4644
}

T
Tejun Heo 已提交
4645 4646 4647 4648 4649 4650 4651 4652
static struct kernfs_syscall_ops cgroup_kf_syscall_ops = {
	.remount_fs		= cgroup_remount,
	.show_options		= cgroup_show_options,
	.mkdir			= cgroup_mkdir,
	.rmdir			= cgroup_rmdir,
	.rename			= cgroup_rename,
};

4653
static void __init cgroup_init_subsys(struct cgroup_subsys *ss, bool early)
4654 4655
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4656 4657

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4658

T
Tejun Heo 已提交
4659
	mutex_lock(&cgroup_tree_mutex);
4660 4661
	mutex_lock(&cgroup_mutex);

4662
	idr_init(&ss->css_idr);
T
Tejun Heo 已提交
4663
	INIT_LIST_HEAD(&ss->cfts);
4664

4665 4666 4667
	/* Create the root cgroup state for this subsystem */
	ss->root = &cgrp_dfl_root;
	css = ss->css_alloc(cgroup_css(&cgrp_dfl_root.cgrp, ss));
4668 4669
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4670
	init_and_link_css(css, ss, &cgrp_dfl_root.cgrp);
4671 4672 4673 4674 4675 4676 4677
	if (early) {
		/* idr_alloc() can't be called safely during early init */
		css->id = 1;
	} else {
		css->id = cgroup_idr_alloc(&ss->css_idr, css, 1, 2, GFP_KERNEL);
		BUG_ON(css->id < 0);
	}
4678

L
Li Zefan 已提交
4679
	/* Update the init_css_set to contain a subsys
4680
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4681
	 * newly registered, all tasks and hence the
4682
	 * init_css_set is in the subsystem's root cgroup. */
4683
	init_css_set.subsys[ss->id] = css;
4684 4685 4686

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

L
Li Zefan 已提交
4687 4688 4689 4690 4691
	/* 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));

4692
	BUG_ON(online_css(css));
4693

4694
	cgrp_dfl_root.subsys_mask |= 1 << ss->id;
B
Ben Blum 已提交
4695 4696

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4697
	mutex_unlock(&cgroup_tree_mutex);
B
Ben Blum 已提交
4698 4699
}

4700
/**
L
Li Zefan 已提交
4701 4702 4703 4704
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4705 4706 4707
 */
int __init cgroup_init_early(void)
{
T
Tejun Heo 已提交
4708 4709
	static struct cgroup_sb_opts __initdata opts =
		{ .flags = CGRP_ROOT_SANE_BEHAVIOR };
4710
	struct cgroup_subsys *ss;
4711
	int i;
4712

4713
	init_cgroup_root(&cgrp_dfl_root, &opts);
4714
	RCU_INIT_POINTER(init_task.cgroups, &init_css_set);
4715

T
Tejun Heo 已提交
4716
	for_each_subsys(ss, i) {
4717
		WARN(!ss->css_alloc || !ss->css_free || ss->name || ss->id,
4718 4719
		     "invalid cgroup_subsys %d:%s css_alloc=%p css_free=%p name:id=%d:%s\n",
		     i, cgroup_subsys_name[i], ss->css_alloc, ss->css_free,
4720
		     ss->id, ss->name);
4721 4722 4723
		WARN(strlen(cgroup_subsys_name[i]) > MAX_CGROUP_TYPE_NAMELEN,
		     "cgroup_subsys_name %s too long\n", cgroup_subsys_name[i]);

4724
		ss->id = i;
4725
		ss->name = cgroup_subsys_name[i];
4726 4727

		if (ss->early_init)
4728
			cgroup_init_subsys(ss, true);
4729 4730 4731 4732 4733
	}
	return 0;
}

/**
L
Li Zefan 已提交
4734 4735 4736 4737
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4738 4739 4740
 */
int __init cgroup_init(void)
{
4741
	struct cgroup_subsys *ss;
4742
	unsigned long key;
4743
	int ssid, err;
4744

T
Tejun Heo 已提交
4745
	BUG_ON(cgroup_init_cftypes(NULL, cgroup_base_files));
4746

4747
	mutex_lock(&cgroup_tree_mutex);
T
Tejun Heo 已提交
4748 4749
	mutex_lock(&cgroup_mutex);

4750 4751 4752 4753
	/* 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);

4754
	BUG_ON(cgroup_setup_root(&cgrp_dfl_root, 0));
4755

T
Tejun Heo 已提交
4756
	mutex_unlock(&cgroup_mutex);
4757
	mutex_unlock(&cgroup_tree_mutex);
T
Tejun Heo 已提交
4758

4759
	for_each_subsys(ss, ssid) {
4760 4761 4762 4763 4764 4765 4766 4767 4768 4769
		if (ss->early_init) {
			struct cgroup_subsys_state *css =
				init_css_set.subsys[ss->id];

			css->id = cgroup_idr_alloc(&ss->css_idr, css, 1, 2,
						   GFP_KERNEL);
			BUG_ON(css->id < 0);
		} else {
			cgroup_init_subsys(ss, false);
		}
4770

T
Tejun Heo 已提交
4771 4772 4773
		list_add_tail(&init_css_set.e_cset_node[ssid],
			      &cgrp_dfl_root.cgrp.e_csets[ssid]);

4774 4775 4776 4777 4778 4779
		/*
		 * cftype registration needs kmalloc and can't be done
		 * during early_init.  Register base cftypes separately.
		 */
		if (ss->base_cftypes)
			WARN_ON(cgroup_add_cftypes(ss, ss->base_cftypes));
4780 4781 4782
	}

	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
T
Tejun Heo 已提交
4783 4784
	if (!cgroup_kobj)
		return -ENOMEM;
4785

4786
	err = register_filesystem(&cgroup_fs_type);
4787 4788
	if (err < 0) {
		kobject_put(cgroup_kobj);
T
Tejun Heo 已提交
4789
		return err;
4790
	}
4791

L
Li Zefan 已提交
4792
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
T
Tejun Heo 已提交
4793
	return 0;
4794
}
4795

4796 4797 4798 4799 4800
static int __init cgroup_wq_init(void)
{
	/*
	 * There isn't much point in executing destruction path in
	 * parallel.  Good chunk is serialized with cgroup_mutex anyway.
4801
	 * Use 1 for @max_active.
4802 4803 4804 4805
	 *
	 * We would prefer to do this in cgroup_init() above, but that
	 * is called before init_workqueues(): so leave this until after.
	 */
4806
	cgroup_destroy_wq = alloc_workqueue("cgroup_destroy", 0, 1);
4807
	BUG_ON(!cgroup_destroy_wq);
4808 4809 4810 4811 4812 4813 4814 4815 4816

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

4817 4818 4819 4820
	return 0;
}
core_initcall(cgroup_wq_init);

4821 4822 4823 4824 4825 4826 4827
/*
 * proc_cgroup_show()
 *  - Print task's cgroup paths into seq_file, one line for each hierarchy
 *  - Used for /proc/<pid>/cgroup.
 */

/* TODO: Use a proper seq_file iterator */
4828
int proc_cgroup_show(struct seq_file *m, void *v)
4829 4830 4831
{
	struct pid *pid;
	struct task_struct *tsk;
T
Tejun Heo 已提交
4832
	char *buf, *path;
4833
	int retval;
4834
	struct cgroup_root *root;
4835 4836

	retval = -ENOMEM;
T
Tejun Heo 已提交
4837
	buf = kmalloc(PATH_MAX, GFP_KERNEL);
4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849
	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);
4850
	down_read(&css_set_rwsem);
4851

4852
	for_each_root(root) {
4853
		struct cgroup_subsys *ss;
4854
		struct cgroup *cgrp;
T
Tejun Heo 已提交
4855
		int ssid, count = 0;
4856

T
Tejun Heo 已提交
4857
		if (root == &cgrp_dfl_root && !cgrp_dfl_root_visible)
4858 4859
			continue;

4860
		seq_printf(m, "%d:", root->hierarchy_id);
T
Tejun Heo 已提交
4861
		for_each_subsys(ss, ssid)
4862
			if (root->subsys_mask & (1 << ssid))
T
Tejun Heo 已提交
4863
				seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4864 4865 4866
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4867
		seq_putc(m, ':');
4868
		cgrp = task_cgroup_from_root(tsk, root);
T
Tejun Heo 已提交
4869 4870 4871
		path = cgroup_path(cgrp, buf, PATH_MAX);
		if (!path) {
			retval = -ENAMETOOLONG;
4872
			goto out_unlock;
T
Tejun Heo 已提交
4873 4874
		}
		seq_puts(m, path);
4875 4876 4877 4878
		seq_putc(m, '\n');
	}

out_unlock:
4879
	up_read(&css_set_rwsem);
4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890
	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)
{
4891
	struct cgroup_subsys *ss;
4892 4893
	int i;

4894
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4895 4896 4897 4898 4899
	/*
	 * 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.
	 */
4900
	mutex_lock(&cgroup_mutex);
4901 4902

	for_each_subsys(ss, i)
4903 4904
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
4905
			   atomic_read(&ss->root->nr_cgrps), !ss->disabled);
4906

4907 4908 4909 4910 4911 4912
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
4913
	return single_open(file, proc_cgroupstats_show, NULL);
4914 4915
}

4916
static const struct file_operations proc_cgroupstats_operations = {
4917 4918 4919 4920 4921 4922
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

4923
/**
4924
 * cgroup_fork - initialize cgroup related fields during copy_process()
L
Li Zefan 已提交
4925
 * @child: pointer to task_struct of forking parent process.
4926
 *
4927 4928 4929
 * A task is associated with the init_css_set until cgroup_post_fork()
 * attaches it to the parent's css_set.  Empty cg_list indicates that
 * @child isn't holding reference to its css_set.
4930 4931 4932
 */
void cgroup_fork(struct task_struct *child)
{
4933
	RCU_INIT_POINTER(child->cgroups, &init_css_set);
4934
	INIT_LIST_HEAD(&child->cg_list);
4935 4936
}

4937
/**
L
Li Zefan 已提交
4938 4939 4940
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
4941 4942 4943
 * 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
4944
 * cgroup_task_iter_start() - to guarantee that the new task ends up on its
4945
 * list.
L
Li Zefan 已提交
4946
 */
4947 4948
void cgroup_post_fork(struct task_struct *child)
{
4949
	struct cgroup_subsys *ss;
4950 4951
	int i;

4952
	/*
4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971
	 * This may race against cgroup_enable_task_cg_links().  As that
	 * function sets use_task_css_set_links before grabbing
	 * tasklist_lock and we just went through tasklist_lock to add
	 * @child, it's guaranteed that either we see the set
	 * use_task_css_set_links or cgroup_enable_task_cg_lists() sees
	 * @child during its iteration.
	 *
	 * If we won the race, @child is associated with %current's
	 * css_set.  Grabbing css_set_rwsem guarantees both that the
	 * association is stable, and, on completion of the parent's
	 * migration, @child is visible in the source of migration or
	 * already in the destination cgroup.  This guarantee is necessary
	 * when implementing operations which need to migrate all tasks of
	 * a cgroup to another.
	 *
	 * Note that if we lose to cgroup_enable_task_cg_links(), @child
	 * will remain in init_css_set.  This is safe because all tasks are
	 * in the init_css_set before cg_links is enabled and there's no
	 * operation which transfers all tasks out of init_css_set.
4972
	 */
4973
	if (use_task_css_set_links) {
4974 4975
		struct css_set *cset;

4976
		down_write(&css_set_rwsem);
4977
		cset = task_css_set(current);
4978 4979 4980 4981 4982
		if (list_empty(&child->cg_list)) {
			rcu_assign_pointer(child->cgroups, cset);
			list_add(&child->cg_list, &cset->tasks);
			get_css_set(cset);
		}
4983
		up_write(&css_set_rwsem);
4984
	}
4985 4986 4987 4988 4989 4990 4991

	/*
	 * Call ss->fork().  This must happen after @child is linked on
	 * css_set; otherwise, @child might change state between ->fork()
	 * and addition to css_set.
	 */
	if (need_forkexit_callback) {
T
Tejun Heo 已提交
4992
		for_each_subsys(ss, i)
4993 4994 4995
			if (ss->fork)
				ss->fork(child);
	}
4996
}
4997

4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
 *
 * 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.
 *
5010 5011 5012 5013 5014
 * We 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.  No need to bother with
 * init_css_set refcnting.  init_css_set never goes away and we can't race
5015
 * with migration path - PF_EXITING is visible to migration path.
5016
 */
5017
void cgroup_exit(struct task_struct *tsk)
5018
{
5019
	struct cgroup_subsys *ss;
5020
	struct css_set *cset;
5021
	bool put_cset = false;
5022
	int i;
5023 5024

	/*
5025 5026
	 * Unlink from @tsk from its css_set.  As migration path can't race
	 * with us, we can check cg_list without grabbing css_set_rwsem.
5027 5028
	 */
	if (!list_empty(&tsk->cg_list)) {
5029
		down_write(&css_set_rwsem);
5030
		list_del_init(&tsk->cg_list);
5031
		up_write(&css_set_rwsem);
5032
		put_cset = true;
5033 5034
	}

5035
	/* Reassign the task to the init_css_set. */
5036 5037
	cset = task_css_set(tsk);
	RCU_INIT_POINTER(tsk->cgroups, &init_css_set);
5038

5039
	if (need_forkexit_callback) {
T
Tejun Heo 已提交
5040 5041
		/* see cgroup_post_fork() for details */
		for_each_subsys(ss, i) {
5042
			if (ss->exit) {
5043 5044
				struct cgroup_subsys_state *old_css = cset->subsys[i];
				struct cgroup_subsys_state *css = task_css(tsk, i);
5045

5046
				ss->exit(css, old_css, tsk);
5047 5048 5049 5050
			}
		}
	}

5051 5052
	if (put_cset)
		put_css_set(cset, true);
5053
}
5054

5055
static void check_for_release(struct cgroup *cgrp)
5056
{
5057
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
5058
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
5059 5060
		/*
		 * Control Group is currently removeable. If it's not
5061
		 * already queued for a userspace notification, queue
5062 5063
		 * it now
		 */
5064
		int need_schedule_work = 0;
5065

5066
		raw_spin_lock(&release_list_lock);
5067
		if (!cgroup_is_dead(cgrp) &&
5068 5069
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
5070 5071
			need_schedule_work = 1;
		}
5072
		raw_spin_unlock(&release_list_lock);
5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104
		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);
5105
	raw_spin_lock(&release_list_lock);
5106 5107 5108
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
T
Tejun Heo 已提交
5109
		char *pathbuf = NULL, *agentbuf = NULL, *path;
5110
		struct cgroup *cgrp = list_entry(release_list.next,
5111 5112
						    struct cgroup,
						    release_list);
5113
		list_del_init(&cgrp->release_list);
5114
		raw_spin_unlock(&release_list_lock);
T
Tejun Heo 已提交
5115
		pathbuf = kmalloc(PATH_MAX, GFP_KERNEL);
5116 5117
		if (!pathbuf)
			goto continue_free;
T
Tejun Heo 已提交
5118 5119
		path = cgroup_path(cgrp, pathbuf, PATH_MAX);
		if (!path)
5120 5121 5122 5123
			goto continue_free;
		agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
		if (!agentbuf)
			goto continue_free;
5124 5125

		i = 0;
5126
		argv[i++] = agentbuf;
T
Tejun Heo 已提交
5127
		argv[i++] = path;
5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141
		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);
5142 5143 5144
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5145
		raw_spin_lock(&release_list_lock);
5146
	}
5147
	raw_spin_unlock(&release_list_lock);
5148 5149
	mutex_unlock(&cgroup_mutex);
}
5150 5151 5152

static int __init cgroup_disable(char *str)
{
5153
	struct cgroup_subsys *ss;
5154
	char *token;
5155
	int i;
5156 5157 5158 5159

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5160

T
Tejun Heo 已提交
5161
		for_each_subsys(ss, i) {
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172
			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
5173

5174
/**
5175
 * css_tryget_from_dir - get corresponding css from the dentry of a cgroup dir
5176 5177
 * @dentry: directory dentry of interest
 * @ss: subsystem of interest
5178
 *
5179 5180 5181
 * If @dentry is a directory for a cgroup which has @ss enabled on it, try
 * to get the corresponding css and return it.  If such css doesn't exist
 * or can't be pinned, an ERR_PTR value is returned.
S
Stephane Eranian 已提交
5182
 */
5183 5184
struct cgroup_subsys_state *css_tryget_from_dir(struct dentry *dentry,
						struct cgroup_subsys *ss)
S
Stephane Eranian 已提交
5185
{
T
Tejun Heo 已提交
5186 5187
	struct kernfs_node *kn = kernfs_node_from_dentry(dentry);
	struct cgroup_subsys_state *css = NULL;
S
Stephane Eranian 已提交
5188 5189
	struct cgroup *cgrp;

5190
	/* is @dentry a cgroup dir? */
T
Tejun Heo 已提交
5191 5192
	if (dentry->d_sb->s_type != &cgroup_fs_type || !kn ||
	    kernfs_type(kn) != KERNFS_DIR)
S
Stephane Eranian 已提交
5193 5194
		return ERR_PTR(-EBADF);

5195 5196
	rcu_read_lock();

T
Tejun Heo 已提交
5197 5198 5199 5200 5201 5202 5203 5204
	/*
	 * This path doesn't originate from kernfs and @kn could already
	 * have been or be removed at any point.  @kn->priv is RCU
	 * protected for this access.  See destroy_locked() for details.
	 */
	cgrp = rcu_dereference(kn->priv);
	if (cgrp)
		css = cgroup_css(cgrp, ss);
5205 5206 5207 5208 5209 5210

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

	rcu_read_unlock();
	return css;
S
Stephane Eranian 已提交
5211 5212
}

5213 5214 5215 5216 5217 5218 5219 5220 5221 5222
/**
 * css_from_id - lookup css by id
 * @id: the cgroup id
 * @ss: cgroup subsys to be looked into
 *
 * Returns the css if there's valid one with @id, otherwise returns NULL.
 * Should be called under rcu_read_lock().
 */
struct cgroup_subsys_state *css_from_id(int id, struct cgroup_subsys *ss)
{
5223
	WARN_ON_ONCE(!rcu_read_lock_held());
5224
	return idr_find(&ss->css_idr, id);
S
Stephane Eranian 已提交
5225 5226
}

5227
#ifdef CONFIG_CGROUP_DEBUG
5228 5229
static struct cgroup_subsys_state *
debug_css_alloc(struct cgroup_subsys_state *parent_css)
5230 5231 5232 5233 5234 5235 5236 5237 5238
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

5239
static void debug_css_free(struct cgroup_subsys_state *css)
5240
{
5241
	kfree(css);
5242 5243
}

5244 5245
static u64 debug_taskcount_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5246
{
5247
	return cgroup_task_count(css->cgroup);
5248 5249
}

5250 5251
static u64 current_css_set_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5252 5253 5254 5255
{
	return (u64)(unsigned long)current->cgroups;
}

5256
static u64 current_css_set_refcount_read(struct cgroup_subsys_state *css,
L
Li Zefan 已提交
5257
					 struct cftype *cft)
5258 5259 5260 5261
{
	u64 count;

	rcu_read_lock();
5262
	count = atomic_read(&task_css_set(current)->refcount);
5263 5264 5265 5266
	rcu_read_unlock();
	return count;
}

5267
static int current_css_set_cg_links_read(struct seq_file *seq, void *v)
5268
{
5269
	struct cgrp_cset_link *link;
5270
	struct css_set *cset;
T
Tejun Heo 已提交
5271 5272 5273 5274 5275
	char *name_buf;

	name_buf = kmalloc(NAME_MAX + 1, GFP_KERNEL);
	if (!name_buf)
		return -ENOMEM;
5276

5277
	down_read(&css_set_rwsem);
5278
	rcu_read_lock();
5279
	cset = rcu_dereference(current->cgroups);
5280
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5281 5282
		struct cgroup *c = link->cgrp;

T
Tejun Heo 已提交
5283
		cgroup_name(c, name_buf, NAME_MAX + 1);
5284
		seq_printf(seq, "Root %d group %s\n",
T
Tejun Heo 已提交
5285
			   c->root->hierarchy_id, name_buf);
5286 5287
	}
	rcu_read_unlock();
5288
	up_read(&css_set_rwsem);
T
Tejun Heo 已提交
5289
	kfree(name_buf);
5290 5291 5292 5293
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
5294
static int cgroup_css_links_read(struct seq_file *seq, void *v)
5295
{
5296
	struct cgroup_subsys_state *css = seq_css(seq);
5297
	struct cgrp_cset_link *link;
5298

5299
	down_read(&css_set_rwsem);
5300
	list_for_each_entry(link, &css->cgroup->cset_links, cset_link) {
5301
		struct css_set *cset = link->cset;
5302 5303
		struct task_struct *task;
		int count = 0;
T
Tejun Heo 已提交
5304

5305
		seq_printf(seq, "css_set %p\n", cset);
T
Tejun Heo 已提交
5306

5307
		list_for_each_entry(task, &cset->tasks, cg_list) {
T
Tejun Heo 已提交
5308 5309 5310 5311 5312 5313 5314 5315 5316
			if (count++ > MAX_TASKS_SHOWN_PER_CSS)
				goto overflow;
			seq_printf(seq, "  task %d\n", task_pid_vnr(task));
		}

		list_for_each_entry(task, &cset->mg_tasks, cg_list) {
			if (count++ > MAX_TASKS_SHOWN_PER_CSS)
				goto overflow;
			seq_printf(seq, "  task %d\n", task_pid_vnr(task));
5317
		}
T
Tejun Heo 已提交
5318 5319 5320
		continue;
	overflow:
		seq_puts(seq, "  ...\n");
5321
	}
5322
	up_read(&css_set_rwsem);
5323 5324 5325
	return 0;
}

5326
static u64 releasable_read(struct cgroup_subsys_state *css, struct cftype *cft)
5327
{
5328
	return test_bit(CGRP_RELEASABLE, &css->cgroup->flags);
5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346
}

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,
	},

5347 5348
	{
		.name = "current_css_set_cg_links",
5349
		.seq_show = current_css_set_cg_links_read,
5350 5351 5352 5353
	},

	{
		.name = "cgroup_css_links",
5354
		.seq_show = cgroup_css_links_read,
5355 5356
	},

5357 5358 5359 5360 5361
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5362 5363
	{ }	/* terminate */
};
5364

5365
struct cgroup_subsys debug_cgrp_subsys = {
5366 5367
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5368
	.base_cftypes = debug_files,
5369 5370
};
#endif /* CONFIG_CGROUP_DEBUG */