cgroup.c 144.8 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
}

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

	/*
	 * 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;
303
}
T
Tejun Heo 已提交
304
EXPORT_SYMBOL_GPL(of_css);
305

306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
/**
 * 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;
}
324

325
static int cgroup_is_releasable(const struct cgroup *cgrp)
326 327
{
	const int bits =
328 329 330
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
331 332
}

333
static int notify_on_release(const struct cgroup *cgrp)
334
{
335
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
336 337
}

T
Tejun Heo 已提交
338 339 340 341 342 343
/**
 * 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
 *
344
 * Should be called under cgroup_[tree_]mutex.
T
Tejun Heo 已提交
345 346 347 348 349
 */
#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
350
				lockdep_is_held(&cgroup_tree_mutex) ||	\
T
Tejun Heo 已提交
351 352 353
				lockdep_is_held(&cgroup_mutex)))) { }	\
		else

354 355 356 357 358 359 360 361 362 363 364 365 366 367
/**
 * 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

368
/**
T
Tejun Heo 已提交
369
 * for_each_subsys - iterate all enabled cgroup subsystems
370
 * @ss: the iteration cursor
371
 * @ssid: the index of @ss, CGROUP_SUBSYS_COUNT after reaching the end
372
 */
373
#define for_each_subsys(ss, ssid)					\
T
Tejun Heo 已提交
374 375
	for ((ssid) = 0; (ssid) < CGROUP_SUBSYS_COUNT &&		\
	     (((ss) = cgroup_subsys[ssid]) || true); (ssid)++)
376

377 378
/* iterate across the hierarchies */
#define for_each_root(root)						\
379
	list_for_each_entry((root), &cgroup_roots, root_list)
380

381 382 383 384 385 386 387 388
/* 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

389 390 391 392
/**
 * 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
393 394
 * On success, returns true; the mutex should be later unlocked.  On
 * failure returns false with no lock held.
395
 */
396
static bool cgroup_lock_live_group(struct cgroup *cgrp)
397 398
{
	mutex_lock(&cgroup_mutex);
399
	if (cgroup_is_dead(cgrp)) {
400 401 402 403 404 405
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}

406 407 408
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
409
static DEFINE_RAW_SPINLOCK(release_list_lock);
410 411
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
412
static void check_for_release(struct cgroup *cgrp);
413

414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
/*
 * 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;
432 433
};

434 435
/*
 * The default css_set - used by init and its children prior to any
436 437 438 439 440
 * 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
441
struct css_set init_css_set = {
442 443 444 445 446 447 448
	.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),
};
449

450
static int css_set_count	= 1;	/* 1 for init_css_set */
451

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
/**
 * 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);
}

489 490 491 492 493
/*
 * 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.
 */
494
#define CSS_SET_HASH_BITS	7
495
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
496

497
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
498
{
499
	unsigned long key = 0UL;
500 501
	struct cgroup_subsys *ss;
	int i;
502

503
	for_each_subsys(ss, i)
504 505
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
506

507
	return key;
508 509
}

510
static void put_css_set_locked(struct css_set *cset, bool taskexit)
511
{
512
	struct cgrp_cset_link *link, *tmp_link;
T
Tejun Heo 已提交
513 514
	struct cgroup_subsys *ss;
	int ssid;
515

516 517 518
	lockdep_assert_held(&css_set_rwsem);

	if (!atomic_dec_and_test(&cset->refcount))
519
		return;
520

521
	/* This css_set is dead. unlink it and release cgroup refcounts */
T
Tejun Heo 已提交
522 523
	for_each_subsys(ss, ssid)
		list_del(&cset->e_cset_node[ssid]);
524
	hash_del(&cset->hlist);
525 526
	css_set_count--;

527
	list_for_each_entry_safe(link, tmp_link, &cset->cgrp_links, cgrp_link) {
528
		struct cgroup *cgrp = link->cgrp;
529

530 531
		list_del(&link->cset_link);
		list_del(&link->cgrp_link);
L
Li Zefan 已提交
532

533
		/* @cgrp can't go away while we're holding css_set_rwsem */
534 535 536 537 538 539 540
		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);
			}
541
		}
542 543

		kfree(link);
544
	}
545

546
	kfree_rcu(cset, rcu_head);
547 548
}

549 550 551 552 553 554 555 556 557 558 559 560 561 562 563
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);
}

564 565 566
/*
 * refcounted get/put for css_set objects
 */
567
static inline void get_css_set(struct css_set *cset)
568
{
569
	atomic_inc(&cset->refcount);
570 571
}

572
/**
573
 * compare_css_sets - helper function for find_existing_css_set().
574 575
 * @cset: candidate css_set being tested
 * @old_cset: existing css_set for a task
576 577 578
 * @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
579
 * Returns true if "cset" matches "old_cset" except for the hierarchy
580 581
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
582 583
static bool compare_css_sets(struct css_set *cset,
			     struct css_set *old_cset,
584 585 586 587 588
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

589 590 591 592 593 594
	/*
	 * 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)))
595 596 597 598
		return false;

	/*
	 * Compare cgroup pointers in order to distinguish between
599 600 601
	 * different cgroups in hierarchies.  As different cgroups may
	 * share the same effective css, this comparison is always
	 * necessary.
602
	 */
603 604
	l1 = &cset->cgrp_links;
	l2 = &old_cset->cgrp_links;
605
	while (1) {
606
		struct cgrp_cset_link *link1, *link2;
607
		struct cgroup *cgrp1, *cgrp2;
608 609 610 611

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
612 613
		if (l1 == &cset->cgrp_links) {
			BUG_ON(l2 != &old_cset->cgrp_links);
614 615
			break;
		} else {
616
			BUG_ON(l2 == &old_cset->cgrp_links);
617 618
		}
		/* Locate the cgroups associated with these links. */
619 620 621 622
		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;
623
		/* Hierarchies should be linked in the same order. */
624
		BUG_ON(cgrp1->root != cgrp2->root);
625 626 627 628 629 630 631 632

		/*
		 * 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.
		 */
633 634
		if (cgrp1->root == new_cgrp->root) {
			if (cgrp1 != new_cgrp)
635 636
				return false;
		} else {
637
			if (cgrp1 != cgrp2)
638 639 640 641 642 643
				return false;
		}
	}
	return true;
}

644 645 646 647 648
/**
 * 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
649
 */
650 651 652
static struct css_set *find_existing_css_set(struct css_set *old_cset,
					struct cgroup *cgrp,
					struct cgroup_subsys_state *template[])
653
{
654
	struct cgroup_root *root = cgrp->root;
655
	struct cgroup_subsys *ss;
656
	struct css_set *cset;
657
	unsigned long key;
658
	int i;
659

B
Ben Blum 已提交
660 661 662 663 664
	/*
	 * 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.
	 */
665
	for_each_subsys(ss, i) {
666
		if (root->subsys_mask & (1UL << i)) {
667 668 669 670 671
			/*
			 * @ss is in this hierarchy, so we want the
			 * effective css from @cgrp.
			 */
			template[i] = cgroup_e_css(cgrp, ss);
672
		} else {
673 674 675 676
			/*
			 * @ss is not in this hierarchy, so we don't want
			 * to change the css.
			 */
677
			template[i] = old_cset->subsys[i];
678 679 680
		}
	}

681
	key = css_set_hash(template);
682 683
	hash_for_each_possible(css_set_table, cset, hlist, key) {
		if (!compare_css_sets(cset, old_cset, cgrp, template))
684 685 686
			continue;

		/* This css_set matches what we need */
687
		return cset;
688
	}
689 690 691 692 693

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

694
static void free_cgrp_cset_links(struct list_head *links_to_free)
695
{
696
	struct cgrp_cset_link *link, *tmp_link;
697

698 699
	list_for_each_entry_safe(link, tmp_link, links_to_free, cset_link) {
		list_del(&link->cset_link);
700 701 702 703
		kfree(link);
	}
}

704 705 706 707 708 709 710
/**
 * 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.
711
 */
712
static int allocate_cgrp_cset_links(int count, struct list_head *tmp_links)
713
{
714
	struct cgrp_cset_link *link;
715
	int i;
716 717 718

	INIT_LIST_HEAD(tmp_links);

719
	for (i = 0; i < count; i++) {
720
		link = kzalloc(sizeof(*link), GFP_KERNEL);
721
		if (!link) {
722
			free_cgrp_cset_links(tmp_links);
723 724
			return -ENOMEM;
		}
725
		list_add(&link->cset_link, tmp_links);
726 727 728 729
	}
	return 0;
}

730 731
/**
 * link_css_set - a helper function to link a css_set to a cgroup
732
 * @tmp_links: cgrp_cset_link objects allocated by allocate_cgrp_cset_links()
733
 * @cset: the css_set to be linked
734 735
 * @cgrp: the destination cgroup
 */
736 737
static void link_css_set(struct list_head *tmp_links, struct css_set *cset,
			 struct cgroup *cgrp)
738
{
739
	struct cgrp_cset_link *link;
740

741
	BUG_ON(list_empty(tmp_links));
T
Tejun Heo 已提交
742 743 744 745

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

746 747
	link = list_first_entry(tmp_links, struct cgrp_cset_link, cset_link);
	link->cset = cset;
748
	link->cgrp = cgrp;
749 750 751

	if (list_empty(&cgrp->cset_links))
		cgroup_update_populated(cgrp, true);
752
	list_move(&link->cset_link, &cgrp->cset_links);
753

754 755 756 757
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
758
	list_add_tail(&link->cgrp_link, &cset->cgrp_links);
759 760
}

761 762 763 764 765 766 767
/**
 * 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.
768
 */
769 770
static struct css_set *find_css_set(struct css_set *old_cset,
				    struct cgroup *cgrp)
771
{
772
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT] = { };
773
	struct css_set *cset;
774 775
	struct list_head tmp_links;
	struct cgrp_cset_link *link;
T
Tejun Heo 已提交
776
	struct cgroup_subsys *ss;
777
	unsigned long key;
T
Tejun Heo 已提交
778
	int ssid;
779

780 781
	lockdep_assert_held(&cgroup_mutex);

782 783
	/* First see if we already have a cgroup group that matches
	 * the desired set */
784
	down_read(&css_set_rwsem);
785 786 787
	cset = find_existing_css_set(old_cset, cgrp, template);
	if (cset)
		get_css_set(cset);
788
	up_read(&css_set_rwsem);
789

790 791
	if (cset)
		return cset;
792

793
	cset = kzalloc(sizeof(*cset), GFP_KERNEL);
794
	if (!cset)
795 796
		return NULL;

797
	/* Allocate all the cgrp_cset_link objects that we'll need */
798
	if (allocate_cgrp_cset_links(cgroup_root_count, &tmp_links) < 0) {
799
		kfree(cset);
800 801 802
		return NULL;
	}

803
	atomic_set(&cset->refcount, 1);
804
	INIT_LIST_HEAD(&cset->cgrp_links);
805
	INIT_LIST_HEAD(&cset->tasks);
T
Tejun Heo 已提交
806
	INIT_LIST_HEAD(&cset->mg_tasks);
807
	INIT_LIST_HEAD(&cset->mg_preload_node);
808
	INIT_LIST_HEAD(&cset->mg_node);
809
	INIT_HLIST_NODE(&cset->hlist);
810 811 812

	/* Copy the set of subsystem state objects generated in
	 * find_existing_css_set() */
813
	memcpy(cset->subsys, template, sizeof(cset->subsys));
814

815
	down_write(&css_set_rwsem);
816
	/* Add reference counts and links from the new css_set. */
817
	list_for_each_entry(link, &old_cset->cgrp_links, cgrp_link) {
818
		struct cgroup *c = link->cgrp;
819

820 821
		if (c->root == cgrp->root)
			c = cgrp;
822
		link_css_set(&tmp_links, cset, c);
823
	}
824

825
	BUG_ON(!list_empty(&tmp_links));
826 827

	css_set_count++;
828

T
Tejun Heo 已提交
829
	/* Add @cset to the hash table */
830 831
	key = css_set_hash(cset->subsys);
	hash_add(css_set_table, &cset->hlist, key);
832

T
Tejun Heo 已提交
833 834 835 836
	for_each_subsys(ss, ssid)
		list_add_tail(&cset->e_cset_node[ssid],
			      &cset->subsys[ssid]->cgroup->e_csets[ssid]);

837
	up_write(&css_set_rwsem);
838

839
	return cset;
840 841
}

842
static struct cgroup_root *cgroup_root_from_kf(struct kernfs_root *kf_root)
843
{
844
	struct cgroup *root_cgrp = kf_root->kn->priv;
T
Tejun Heo 已提交
845

846
	return root_cgrp->root;
T
Tejun Heo 已提交
847 848
}

849
static int cgroup_init_root_id(struct cgroup_root *root)
850 851 852 853 854
{
	int id;

	lockdep_assert_held(&cgroup_mutex);

855
	id = idr_alloc_cyclic(&cgroup_hierarchy_idr, root, 0, 0, GFP_KERNEL);
856 857 858 859 860 861 862
	if (id < 0)
		return id;

	root->hierarchy_id = id;
	return 0;
}

863
static void cgroup_exit_root_id(struct cgroup_root *root)
864 865 866 867 868 869 870 871 872
{
	lockdep_assert_held(&cgroup_mutex);

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

873
static void cgroup_free_root(struct cgroup_root *root)
874 875 876 877 878 879 880 881 882 883
{
	if (root) {
		/* hierarhcy ID shoulid already have been released */
		WARN_ON_ONCE(root->hierarchy_id);

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

884
static void cgroup_destroy_root(struct cgroup_root *root)
885
{
886
	struct cgroup *cgrp = &root->cgrp;
887 888
	struct cgrp_cset_link *link, *tmp_link;

T
Tejun Heo 已提交
889 890
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);
891

T
Tejun Heo 已提交
892
	BUG_ON(atomic_read(&root->nr_cgrps));
893 894 895
	BUG_ON(!list_empty(&cgrp->children));

	/* Rebind all subsystems back to the default hierarchy */
896
	rebind_subsystems(&cgrp_dfl_root, root->subsys_mask);
897 898

	/*
899 900
	 * Release all the links from cset_links to this hierarchy's
	 * root cgroup
901
	 */
902
	down_write(&css_set_rwsem);
903 904 905 906 907 908

	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);
	}
909
	up_write(&css_set_rwsem);
910 911 912 913 914 915 916 917 918 919 920

	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
921
	kernfs_destroy_root(root->kf_root);
922 923 924
	cgroup_free_root(root);
}

925 926
/* look up cgroup associated with given css_set on the specified hierarchy */
static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
927
					    struct cgroup_root *root)
928 929 930
{
	struct cgroup *res = NULL;

931 932 933
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&css_set_rwsem);

934
	if (cset == &init_css_set) {
935
		res = &root->cgrp;
936
	} else {
937 938 939
		struct cgrp_cset_link *link;

		list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
940
			struct cgroup *c = link->cgrp;
941

942 943 944 945 946 947
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
948

949 950 951 952
	BUG_ON(!res);
	return res;
}

953
/*
954 955 956 957
 * 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,
958
					    struct cgroup_root *root)
959 960 961 962 963 964 965 966 967
{
	/*
	 * 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);
}

968 969 970 971 972 973
/*
 * 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
974
 * cgroup_attach_task() can increment it again.  Because a count of zero
975 976 977 978 979 980 981 982 983 984 985 986 987
 * 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
988 989
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
990 991 992 993
 *
 * 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
994
 * least one task in the system (init, pid == 1), therefore, root cgroup
995
 * always has either children cgroups and/or using tasks.  So we don't
996
 * need a special hack to ensure that root cgroup cannot be deleted.
997 998
 *
 * P.S.  One more locking exception.  RCU is used to guard the
999
 * update of a tasks cgroup pointer by cgroup_attach_task()
1000 1001
 */

1002
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned int subsys_mask);
T
Tejun Heo 已提交
1003
static struct kernfs_syscall_ops cgroup_kf_syscall_ops;
1004
static const struct file_operations proc_cgroupstats_operations;
1005

T
Tejun Heo 已提交
1006 1007
static char *cgroup_file_name(struct cgroup *cgrp, const struct cftype *cft,
			      char *buf)
1008
{
T
Tejun Heo 已提交
1009 1010 1011 1012 1013 1014 1015
	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;
1016 1017
}

1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
/**
 * 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)
1028
{
1029
	umode_t mode = 0;
1030

1031 1032 1033 1034 1035 1036
	if (cft->mode)
		return cft->mode;

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

1037
	if (cft->write_u64 || cft->write_s64 || cft->write)
1038 1039 1040
		mode |= S_IWUSR;

	return mode;
1041 1042
}

1043 1044
static void cgroup_free_fn(struct work_struct *work)
{
1045
	struct cgroup *cgrp = container_of(work, struct cgroup, destroy_work);
1046

1047
	atomic_dec(&cgrp->root->nr_cgrps);
1048
	cgroup_pidlist_destroy_all(cgrp);
1049

T
Tejun Heo 已提交
1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
	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 {
		/*
1061
		 * This is root cgroup's refcnt reaching zero, which
T
Tejun Heo 已提交
1062 1063 1064 1065
		 * indicates that the root should be released.
		 */
		cgroup_destroy_root(cgrp->root);
	}
1066 1067 1068 1069 1070 1071
}

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

1072
	INIT_WORK(&cgrp->destroy_work, cgroup_free_fn);
1073
	queue_work(cgroup_destroy_wq, &cgrp->destroy_work);
1074 1075
}

1076
static void cgroup_get(struct cgroup *cgrp)
1077
{
T
Tejun Heo 已提交
1078 1079 1080
	WARN_ON_ONCE(cgroup_is_dead(cgrp));
	WARN_ON_ONCE(atomic_read(&cgrp->refcnt) <= 0);
	atomic_inc(&cgrp->refcnt);
1081 1082
}

1083
static void cgroup_put(struct cgroup *cgrp)
T
Tejun Heo 已提交
1084
{
T
Tejun Heo 已提交
1085 1086
	if (!atomic_dec_and_test(&cgrp->refcnt))
		return;
T
Tejun Heo 已提交
1087
	if (WARN_ON_ONCE(cgrp->parent && !cgroup_is_dead(cgrp)))
T
Tejun Heo 已提交
1088
		return;
T
Tejun Heo 已提交
1089

1090
	cgroup_idr_remove(&cgrp->root->cgroup_idr, cgrp->id);
T
Tejun Heo 已提交
1091
	cgrp->id = -1;
T
Tejun Heo 已提交
1092

T
Tejun Heo 已提交
1093
	call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
1094
}
T
Tejun Heo 已提交
1095

1096
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
1097
{
T
Tejun Heo 已提交
1098
	char name[CGROUP_FILE_NAME_MAX];
T
Tejun Heo 已提交
1099

T
Tejun Heo 已提交
1100
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
1101
	kernfs_remove_by_name(cgrp->kn, cgroup_file_name(cgrp, cft, name));
T
Tejun Heo 已提交
1102 1103
}

1104
/**
1105
 * cgroup_clear_dir - remove subsys files in a cgroup directory
1106
 * @cgrp: target cgroup
1107 1108
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
1109
static void cgroup_clear_dir(struct cgroup *cgrp, unsigned int subsys_mask)
T
Tejun Heo 已提交
1110
{
1111
	struct cgroup_subsys *ss;
1112
	int i;
T
Tejun Heo 已提交
1113

1114
	for_each_subsys(ss, i) {
T
Tejun Heo 已提交
1115
		struct cftype *cfts;
1116

1117
		if (!(subsys_mask & (1 << i)))
1118
			continue;
T
Tejun Heo 已提交
1119 1120
		list_for_each_entry(cfts, &ss->cfts, node)
			cgroup_addrm_files(cgrp, cfts, false);
1121
	}
1122 1123
}

1124
static int rebind_subsystems(struct cgroup_root *dst_root, unsigned int ss_mask)
1125
{
1126
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
1127
	int ssid, i, ret;
1128

T
Tejun Heo 已提交
1129 1130
	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
1131

1132 1133 1134
	for_each_subsys(ss, ssid) {
		if (!(ss_mask & (1 << ssid)))
			continue;
B
Ben Blum 已提交
1135

1136 1137
		/* 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
1138
			return -EBUSY;
1139

1140
		/* can't move between two non-dummy roots either */
1141
		if (ss->root != &cgrp_dfl_root && dst_root != &cgrp_dfl_root)
1142
			return -EBUSY;
1143 1144
	}

T
Tejun Heo 已提交
1145 1146 1147
	ret = cgroup_populate_dir(&dst_root->cgrp, ss_mask);
	if (ret) {
		if (dst_root != &cgrp_dfl_root)
1148
			return ret;
1149

T
Tejun Heo 已提交
1150 1151 1152 1153 1154 1155 1156
		/*
		 * 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) {
1157
			pr_warn("failed to create files (%d) while rebinding 0x%x to default root\n",
1158
				ret, ss_mask);
1159
			pr_warn("you may retry by moving them to a different hierarchy and unbinding\n");
T
Tejun Heo 已提交
1160
		}
1161
	}
1162 1163 1164 1165 1166

	/*
	 * Nothing can fail from this point on.  Remove files for the
	 * removed subsystems and rebind each subsystem.
	 */
1167
	mutex_unlock(&cgroup_mutex);
1168
	for_each_subsys(ss, ssid)
T
Tejun Heo 已提交
1169
		if (ss_mask & (1 << ssid))
1170
			cgroup_clear_dir(&ss->root->cgrp, 1 << ssid);
1171
	mutex_lock(&cgroup_mutex);
1172

1173
	for_each_subsys(ss, ssid) {
1174
		struct cgroup_root *src_root;
1175
		struct cgroup_subsys_state *css;
T
Tejun Heo 已提交
1176
		struct css_set *cset;
1177

1178 1179
		if (!(ss_mask & (1 << ssid)))
			continue;
1180

1181
		src_root = ss->root;
1182
		css = cgroup_css(&src_root->cgrp, ss);
1183

1184
		WARN_ON(!css || cgroup_css(&dst_root->cgrp, ss));
1185

1186 1187
		RCU_INIT_POINTER(src_root->cgrp.subsys[ssid], NULL);
		rcu_assign_pointer(dst_root->cgrp.subsys[ssid], css);
1188
		ss->root = dst_root;
1189
		css->cgroup = &dst_root->cgrp;
1190

T
Tejun Heo 已提交
1191 1192 1193 1194 1195 1196
		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);

1197 1198 1199
		src_root->subsys_mask &= ~(1 << ssid);
		src_root->cgrp.child_subsys_mask &= ~(1 << ssid);

1200
		/* default hierarchy doesn't enable controllers by default */
1201
		dst_root->subsys_mask |= 1 << ssid;
1202 1203
		if (dst_root != &cgrp_dfl_root)
			dst_root->cgrp.child_subsys_mask |= 1 << ssid;
1204

1205 1206
		if (ss->bind)
			ss->bind(css);
1207 1208
	}

T
Tejun Heo 已提交
1209
	kernfs_activate(dst_root->cgrp.kn);
1210 1211 1212
	return 0;
}

T
Tejun Heo 已提交
1213 1214
static int cgroup_show_options(struct seq_file *seq,
			       struct kernfs_root *kf_root)
1215
{
1216
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
1217
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
1218
	int ssid;
1219

T
Tejun Heo 已提交
1220
	for_each_subsys(ss, ssid)
1221
		if (root->subsys_mask & (1 << ssid))
T
Tejun Heo 已提交
1222
			seq_printf(seq, ",%s", ss->name);
1223 1224
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR)
		seq_puts(seq, ",sane_behavior");
1225
	if (root->flags & CGRP_ROOT_NOPREFIX)
1226
		seq_puts(seq, ",noprefix");
1227
	if (root->flags & CGRP_ROOT_XATTR)
A
Aristeu Rozanski 已提交
1228
		seq_puts(seq, ",xattr");
1229 1230

	spin_lock(&release_agent_path_lock);
1231 1232
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1233 1234
	spin_unlock(&release_agent_path_lock);

1235
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->cgrp.flags))
1236
		seq_puts(seq, ",clone_children");
1237 1238
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
1239 1240 1241 1242
	return 0;
}

struct cgroup_sb_opts {
1243 1244
	unsigned int subsys_mask;
	unsigned int flags;
1245
	char *release_agent;
1246
	bool cpuset_clone_children;
1247
	char *name;
1248 1249
	/* User explicitly requested empty subsystem */
	bool none;
1250 1251
};

B
Ben Blum 已提交
1252
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1253
{
1254 1255
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1256
	unsigned int mask = -1U;
1257 1258
	struct cgroup_subsys *ss;
	int i;
1259 1260

#ifdef CONFIG_CPUSETS
1261
	mask = ~(1U << cpuset_cgrp_id);
1262
#endif
1263

1264
	memset(opts, 0, sizeof(*opts));
1265 1266 1267 1268

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1269
		if (!strcmp(token, "none")) {
1270 1271
			/* Explicitly have no subsystems */
			opts->none = true;
1272 1273 1274 1275 1276 1277 1278 1279 1280
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
1281 1282 1283 1284
		if (!strcmp(token, "__DEVEL__sane_behavior")) {
			opts->flags |= CGRP_ROOT_SANE_BEHAVIOR;
			continue;
		}
1285
		if (!strcmp(token, "noprefix")) {
1286
			opts->flags |= CGRP_ROOT_NOPREFIX;
1287 1288 1289
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1290
			opts->cpuset_clone_children = true;
1291 1292
			continue;
		}
A
Aristeu Rozanski 已提交
1293
		if (!strcmp(token, "xattr")) {
1294
			opts->flags |= CGRP_ROOT_XATTR;
A
Aristeu Rozanski 已提交
1295 1296
			continue;
		}
1297
		if (!strncmp(token, "release_agent=", 14)) {
1298 1299 1300
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1301
			opts->release_agent =
1302
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1303 1304
			if (!opts->release_agent)
				return -ENOMEM;
1305 1306 1307
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
			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,
1325
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1326 1327 1328
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1329 1330 1331 1332

			continue;
		}

1333
		for_each_subsys(ss, i) {
1334 1335 1336 1337 1338 1339 1340 1341
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1342
			opts->subsys_mask |= (1 << i);
1343 1344 1345 1346 1347 1348 1349 1350
			one_ss = true;

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

1351 1352
	/* Consistency checks */

1353
	if (opts->flags & CGRP_ROOT_SANE_BEHAVIOR) {
1354
		pr_warn("sane_behavior: this is still under development and its behaviors will change, proceed at your own risk\n");
1355

1356 1357 1358
		if ((opts->flags & (CGRP_ROOT_NOPREFIX | CGRP_ROOT_XATTR)) ||
		    opts->cpuset_clone_children || opts->release_agent ||
		    opts->name) {
1359
			pr_err("sane_behavior: noprefix, xattr, clone_children, release_agent and name are not allowed\n");
1360 1361
			return -EINVAL;
		}
T
Tejun Heo 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370
	} 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)
1371
					opts->subsys_mask |= (1 << i);
1372

T
Tejun Heo 已提交
1373 1374 1375 1376 1377
		/*
		 * We either have to specify by name or by subsystems. (So
		 * all empty hierarchies must have a name).
		 */
		if (!opts->subsys_mask && !opts->name)
1378 1379 1380
			return -EINVAL;
	}

1381 1382 1383 1384 1385
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
1386
	if ((opts->flags & CGRP_ROOT_NOPREFIX) && (opts->subsys_mask & mask))
1387 1388
		return -EINVAL;

1389 1390

	/* Can't specify "none" and some subsystems */
1391
	if (opts->subsys_mask && opts->none)
1392 1393
		return -EINVAL;

1394 1395 1396
	return 0;
}

T
Tejun Heo 已提交
1397
static int cgroup_remount(struct kernfs_root *kf_root, int *flags, char *data)
1398 1399
{
	int ret = 0;
1400
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
1401
	struct cgroup_sb_opts opts;
1402
	unsigned int added_mask, removed_mask;
1403

1404
	if (root->flags & CGRP_ROOT_SANE_BEHAVIOR) {
1405
		pr_err("sane_behavior: remount is not allowed\n");
1406 1407 1408
		return -EINVAL;
	}

T
Tejun Heo 已提交
1409
	mutex_lock(&cgroup_tree_mutex);
1410 1411 1412 1413 1414 1415 1416
	mutex_lock(&cgroup_mutex);

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

1417
	if (opts.subsys_mask != root->subsys_mask || opts.release_agent)
1418
		pr_warn("option changes via remount are deprecated (pid=%d comm=%s)\n",
1419
			task_tgid_nr(current), current->comm);
1420

1421 1422
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1423

B
Ben Blum 已提交
1424
	/* Don't allow flags or name to change at remount */
1425
	if (((opts.flags ^ root->flags) & CGRP_ROOT_OPTION_MASK) ||
B
Ben Blum 已提交
1426
	    (opts.name && strcmp(opts.name, root->name))) {
1427
		pr_err("option or name mismatch, new: 0x%x \"%s\", old: 0x%x \"%s\"\n",
1428 1429
		       opts.flags & CGRP_ROOT_OPTION_MASK, opts.name ?: "",
		       root->flags & CGRP_ROOT_OPTION_MASK, root->name);
1430 1431 1432 1433
		ret = -EINVAL;
		goto out_unlock;
	}

1434
	/* remounting is not allowed for populated hierarchies */
1435
	if (!list_empty(&root->cgrp.children)) {
1436
		ret = -EBUSY;
1437
		goto out_unlock;
B
Ben Blum 已提交
1438
	}
1439

1440
	ret = rebind_subsystems(root, added_mask);
1441
	if (ret)
1442
		goto out_unlock;
1443

1444
	rebind_subsystems(&cgrp_dfl_root, removed_mask);
1445

1446 1447
	if (opts.release_agent) {
		spin_lock(&release_agent_path_lock);
1448
		strcpy(root->release_agent_path, opts.release_agent);
1449 1450
		spin_unlock(&release_agent_path_lock);
	}
1451
 out_unlock:
1452
	kfree(opts.release_agent);
1453
	kfree(opts.name);
1454
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1455
	mutex_unlock(&cgroup_tree_mutex);
1456 1457 1458
	return ret;
}

1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
/*
 * 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;

1471
	down_write(&css_set_rwsem);
1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493

	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.
1494 1495
		 * Do it while holding siglock so that we don't end up
		 * racing against cgroup_exit().
1496
		 */
1497
		spin_lock_irq(&p->sighand->siglock);
1498 1499 1500 1501 1502 1503
		if (!(p->flags & PF_EXITING)) {
			struct css_set *cset = task_css_set(p);

			list_add(&p->cg_list, &cset->tasks);
			get_css_set(cset);
		}
1504
		spin_unlock_irq(&p->sighand->siglock);
1505 1506 1507
	} while_each_thread(g, p);
	read_unlock(&tasklist_lock);
out_unlock:
1508
	up_write(&css_set_rwsem);
1509
}
1510

1511 1512
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
T
Tejun Heo 已提交
1513 1514 1515
	struct cgroup_subsys *ss;
	int ssid;

T
Tejun Heo 已提交
1516
	atomic_set(&cgrp->refcnt, 1);
1517 1518
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
1519
	INIT_LIST_HEAD(&cgrp->cset_links);
1520
	INIT_LIST_HEAD(&cgrp->release_list);
1521 1522
	INIT_LIST_HEAD(&cgrp->pidlists);
	mutex_init(&cgrp->pidlist_mutex);
T
Tejun Heo 已提交
1523
	cgrp->dummy_css.cgroup = cgrp;
T
Tejun Heo 已提交
1524 1525 1526

	for_each_subsys(ss, ssid)
		INIT_LIST_HEAD(&cgrp->e_csets[ssid]);
1527 1528

	init_waitqueue_head(&cgrp->offline_waitq);
1529
}
1530

1531
static void init_cgroup_root(struct cgroup_root *root,
1532
			     struct cgroup_sb_opts *opts)
1533
{
1534
	struct cgroup *cgrp = &root->cgrp;
1535

1536
	INIT_LIST_HEAD(&root->root_list);
1537
	atomic_set(&root->nr_cgrps, 1);
1538
	cgrp->root = root;
1539
	init_cgroup_housekeeping(cgrp);
1540
	idr_init(&root->cgroup_idr);
1541 1542 1543 1544 1545 1546

	root->flags = opts->flags;
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1547
	if (opts->cpuset_clone_children)
1548
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->cgrp.flags);
1549 1550
}

1551
static int cgroup_setup_root(struct cgroup_root *root, unsigned int ss_mask)
1552
{
1553
	LIST_HEAD(tmp_links);
1554
	struct cgroup *root_cgrp = &root->cgrp;
1555 1556
	struct css_set *cset;
	int i, ret;
1557

1558 1559
	lockdep_assert_held(&cgroup_tree_mutex);
	lockdep_assert_held(&cgroup_mutex);
1560

1561
	ret = cgroup_idr_alloc(&root->cgroup_idr, root_cgrp, 1, 2, GFP_NOWAIT);
1562
	if (ret < 0)
T
Tejun Heo 已提交
1563
		goto out;
1564
	root_cgrp->id = ret;
1565

1566
	/*
1567
	 * We're accessing css_set_count without locking css_set_rwsem here,
1568 1569 1570 1571 1572 1573
	 * 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
1574
		goto out;
1575

1576
	ret = cgroup_init_root_id(root);
1577
	if (ret)
T
Tejun Heo 已提交
1578
		goto out;
1579

T
Tejun Heo 已提交
1580 1581 1582 1583 1584 1585 1586 1587
	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;
1588

1589 1590
	ret = cgroup_addrm_files(root_cgrp, cgroup_base_files, true);
	if (ret)
T
Tejun Heo 已提交
1591
		goto destroy_root;
1592

1593
	ret = rebind_subsystems(root, ss_mask);
1594
	if (ret)
T
Tejun Heo 已提交
1595
		goto destroy_root;
1596

1597 1598 1599 1600 1601 1602 1603
	/*
	 * 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
1604

1605
	/*
1606
	 * Link the root cgroup in this hierarchy into all the css_set
1607 1608
	 * objects.
	 */
1609
	down_write(&css_set_rwsem);
1610 1611
	hash_for_each(css_set_table, i, cset, hlist)
		link_css_set(&tmp_links, cset, root_cgrp);
1612
	up_write(&css_set_rwsem);
1613

1614
	BUG_ON(!list_empty(&root_cgrp->children));
1615
	BUG_ON(atomic_read(&root->nr_cgrps) != 1);
1616

T
Tejun Heo 已提交
1617
	kernfs_activate(root_cgrp->kn);
1618
	ret = 0;
T
Tejun Heo 已提交
1619
	goto out;
1620

T
Tejun Heo 已提交
1621 1622 1623 1624
destroy_root:
	kernfs_destroy_root(root->kf_root);
	root->kf_root = NULL;
exit_root_id:
1625
	cgroup_exit_root_id(root);
T
Tejun Heo 已提交
1626
out:
1627 1628
	free_cgrp_cset_links(&tmp_links);
	return ret;
1629 1630
}

A
Al Viro 已提交
1631
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1632
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1633
			 void *data)
1634
{
1635
	struct cgroup_root *root;
1636
	struct cgroup_sb_opts opts;
T
Tejun Heo 已提交
1637
	struct dentry *dentry;
1638
	int ret;
L
Li Zefan 已提交
1639
	bool new_sb;
1640

1641 1642 1643 1644 1645 1646
	/*
	 * 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();
1647

1648
	mutex_lock(&cgroup_tree_mutex);
B
Ben Blum 已提交
1649
	mutex_lock(&cgroup_mutex);
1650 1651

	/* First find the desired set of subsystems */
1652
	ret = parse_cgroupfs_options(data, &opts);
1653
	if (ret)
1654
		goto out_unlock;
1655
retry:
T
Tejun Heo 已提交
1656
	/* look for a matching existing root */
T
Tejun Heo 已提交
1657 1658 1659 1660 1661 1662
	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;
1663 1664
	}

1665
	for_each_root(root) {
T
Tejun Heo 已提交
1666
		bool name_match = false;
1667

1668
		if (root == &cgrp_dfl_root)
1669
			continue;
1670

B
Ben Blum 已提交
1671
		/*
T
Tejun Heo 已提交
1672 1673 1674
		 * 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
1675
		 */
T
Tejun Heo 已提交
1676 1677 1678 1679 1680
		if (opts.name) {
			if (strcmp(opts.name, root->name))
				continue;
			name_match = true;
		}
1681

1682
		/*
T
Tejun Heo 已提交
1683 1684
		 * If we asked for subsystems (or explicitly for no
		 * subsystems) then they must match.
1685
		 */
T
Tejun Heo 已提交
1686
		if ((opts.subsys_mask || opts.none) &&
1687
		    (opts.subsys_mask != root->subsys_mask)) {
T
Tejun Heo 已提交
1688 1689 1690 1691 1692
			if (!name_match)
				continue;
			ret = -EBUSY;
			goto out_unlock;
		}
1693

1694
		if ((root->flags ^ opts.flags) & CGRP_ROOT_OPTION_MASK) {
1695
			if ((root->flags | opts.flags) & CGRP_ROOT_SANE_BEHAVIOR) {
1696
				pr_err("sane_behavior: new mount options should match the existing superblock\n");
1697
				ret = -EINVAL;
1698
				goto out_unlock;
1699
			} else {
1700
				pr_warn("new mount options do not match the existing superblock, will be ignored\n");
1701
			}
1702
		}
1703

T
Tejun Heo 已提交
1704
		/*
1705
		 * A root's lifetime is governed by its root cgroup.  Zero
T
Tejun Heo 已提交
1706 1707 1708 1709 1710
		 * 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.
		 */
1711
		if (!atomic_inc_not_zero(&root->cgrp.refcnt)) {
T
Tejun Heo 已提交
1712 1713 1714
			mutex_unlock(&cgroup_mutex);
			mutex_unlock(&cgroup_tree_mutex);
			msleep(10);
1715 1716
			mutex_lock(&cgroup_tree_mutex);
			mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1717 1718
			goto retry;
		}
1719

T
Tejun Heo 已提交
1720
		ret = 0;
T
Tejun Heo 已提交
1721
		goto out_unlock;
1722 1723
	}

1724
	/*
1725 1726 1727
	 * 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.
1728
	 */
1729 1730 1731
	if (!opts.subsys_mask && !opts.none) {
		ret = -EINVAL;
		goto out_unlock;
1732 1733
	}

1734 1735 1736
	root = kzalloc(sizeof(*root), GFP_KERNEL);
	if (!root) {
		ret = -ENOMEM;
T
Tejun Heo 已提交
1737
		goto out_unlock;
1738
	}
1739

1740 1741
	init_cgroup_root(root, &opts);

T
Tejun Heo 已提交
1742
	ret = cgroup_setup_root(root, opts.subsys_mask);
T
Tejun Heo 已提交
1743 1744
	if (ret)
		cgroup_free_root(root);
1745

1746
out_unlock:
1747
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1748
	mutex_unlock(&cgroup_tree_mutex);
1749

1750 1751
	kfree(opts.release_agent);
	kfree(opts.name);
A
Aristeu Rozanski 已提交
1752

T
Tejun Heo 已提交
1753
	if (ret)
1754
		return ERR_PTR(ret);
T
Tejun Heo 已提交
1755

L
Li Zefan 已提交
1756 1757
	dentry = kernfs_mount(fs_type, flags, root->kf_root, &new_sb);
	if (IS_ERR(dentry) || !new_sb)
1758
		cgroup_put(&root->cgrp);
T
Tejun Heo 已提交
1759 1760 1761 1762 1763 1764
	return dentry;
}

static void cgroup_kill_sb(struct super_block *sb)
{
	struct kernfs_root *kf_root = kernfs_root_from_sb(sb);
1765
	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
T
Tejun Heo 已提交
1766

1767
	cgroup_put(&root->cgrp);
T
Tejun Heo 已提交
1768
	kernfs_kill_sb(sb);
1769 1770 1771 1772
}

static struct file_system_type cgroup_fs_type = {
	.name = "cgroup",
A
Al Viro 已提交
1773
	.mount = cgroup_mount,
1774 1775 1776
	.kill_sb = cgroup_kill_sb,
};

1777 1778
static struct kobject *cgroup_kobj;

1779
/**
1780
 * task_cgroup_path - cgroup path of a task in the first cgroup hierarchy
1781 1782 1783 1784
 * @task: target task
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1785 1786 1787 1788 1789
 * 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
1790
 * Return value is the same as kernfs_path().
1791
 */
T
Tejun Heo 已提交
1792
char *task_cgroup_path(struct task_struct *task, char *buf, size_t buflen)
1793
{
1794
	struct cgroup_root *root;
1795
	struct cgroup *cgrp;
T
Tejun Heo 已提交
1796 1797
	int hierarchy_id = 1;
	char *path = NULL;
1798 1799

	mutex_lock(&cgroup_mutex);
1800
	down_read(&css_set_rwsem);
1801

1802 1803
	root = idr_get_next(&cgroup_hierarchy_idr, &hierarchy_id);

1804 1805
	if (root) {
		cgrp = task_cgroup_from_root(task, root);
T
Tejun Heo 已提交
1806
		path = cgroup_path(cgrp, buf, buflen);
1807 1808
	} else {
		/* if no hierarchy exists, everyone is in "/" */
T
Tejun Heo 已提交
1809 1810
		if (strlcpy(buf, "/", buflen) < buflen)
			path = buf;
1811 1812
	}

1813
	up_read(&css_set_rwsem);
1814
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
1815
	return path;
1816
}
1817
EXPORT_SYMBOL_GPL(task_cgroup_path);
1818

1819
/* used to track tasks and other necessary states during migration */
1820
struct cgroup_taskset {
1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838
	/* 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;
1839 1840 1841 1842 1843 1844 1845 1846 1847 1848
};

/**
 * 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)
{
1849 1850 1851 1852
	tset->cur_cset = list_first_entry(tset->csets, struct css_set, mg_node);
	tset->cur_task = NULL;

	return cgroup_taskset_next(tset);
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863
}

/**
 * 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)
{
1864 1865
	struct css_set *cset = tset->cur_cset;
	struct task_struct *task = tset->cur_task;
1866

1867 1868 1869 1870 1871 1872
	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);
1873

1874 1875 1876 1877 1878
		if (&task->cg_list != &cset->mg_tasks) {
			tset->cur_cset = cset;
			tset->cur_task = task;
			return task;
		}
1879

1880 1881 1882
		cset = list_next_entry(cset, mg_node);
		task = NULL;
	}
1883

1884
	return NULL;
1885 1886
}

1887
/**
B
Ben Blum 已提交
1888
 * cgroup_task_migrate - move a task from one cgroup to another.
1889
 * @old_cgrp: the cgroup @tsk is being migrated from
1890 1891
 * @tsk: the task being migrated
 * @new_cset: the new css_set @tsk is being attached to
B
Ben Blum 已提交
1892
 *
1893
 * Must be called with cgroup_mutex, threadgroup and css_set_rwsem locked.
B
Ben Blum 已提交
1894
 */
1895 1896 1897
static void cgroup_task_migrate(struct cgroup *old_cgrp,
				struct task_struct *tsk,
				struct css_set *new_cset)
B
Ben Blum 已提交
1898
{
1899
	struct css_set *old_cset;
B
Ben Blum 已提交
1900

1901 1902 1903
	lockdep_assert_held(&cgroup_mutex);
	lockdep_assert_held(&css_set_rwsem);

B
Ben Blum 已提交
1904
	/*
1905 1906 1907
	 * 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
1908
	 */
1909
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
1910
	old_cset = task_css_set(tsk);
B
Ben Blum 已提交
1911

1912
	get_css_set(new_cset);
1913
	rcu_assign_pointer(tsk->cgroups, new_cset);
B
Ben Blum 已提交
1914

1915 1916 1917 1918 1919 1920 1921
	/*
	 * 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
1922 1923

	/*
1924 1925 1926
	 * 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
1927
	 */
1928
	set_bit(CGRP_RELEASABLE, &old_cgrp->flags);
1929
	put_css_set_locked(old_cset, false);
B
Ben Blum 已提交
1930 1931
}

L
Li Zefan 已提交
1932
/**
1933 1934
 * cgroup_migrate_finish - cleanup after attach
 * @preloaded_csets: list of preloaded css_sets
B
Ben Blum 已提交
1935
 *
1936 1937
 * Undo cgroup_migrate_add_src() and cgroup_migrate_prepare_dst().  See
 * those functions for details.
B
Ben Blum 已提交
1938
 */
1939
static void cgroup_migrate_finish(struct list_head *preloaded_csets)
B
Ben Blum 已提交
1940
{
1941
	struct css_set *cset, *tmp_cset;
B
Ben Blum 已提交
1942

1943 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
	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
1995
 * @dst_cgrp: the destination cgroup (may be %NULL)
1996 1997 1998 1999
 * @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
2000 2001 2002
 * 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.
2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
 *
 * 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);
2013
	struct css_set *src_cset, *tmp_cset;
2014 2015 2016

	lockdep_assert_held(&cgroup_mutex);

2017 2018 2019 2020 2021 2022 2023 2024
	/*
	 * 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;

2025
	/* look up the dst cset for each src cset and link it to src */
2026
	list_for_each_entry_safe(src_cset, tmp_cset, preloaded_csets, mg_preload_node) {
2027 2028
		struct css_set *dst_cset;

2029 2030
		dst_cset = find_css_set(src_cset,
					dst_cgrp ?: src_cset->dfl_cgrp);
2031 2032 2033 2034
		if (!dst_cset)
			goto err;

		WARN_ON_ONCE(src_cset->mg_dst_cset || dst_cset->mg_dst_cset);
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048

		/*
		 * 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;
		}

2049 2050 2051 2052 2053 2054 2055 2056
		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);
	}

2057
	list_splice_tail(&csets, preloaded_csets);
2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083
	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
2084
{
2085 2086 2087 2088 2089
	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
2090
	struct cgroup_subsys_state *css, *failed_css = NULL;
2091 2092 2093
	struct css_set *cset, *tmp_cset;
	struct task_struct *task, *tmp_task;
	int i, ret;
B
Ben Blum 已提交
2094

2095 2096 2097 2098 2099
	/*
	 * 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.
	 */
2100
	down_write(&css_set_rwsem);
2101
	rcu_read_lock();
2102
	task = leader;
B
Ben Blum 已提交
2103
	do {
2104 2105
		/* @task either already exited or can't exit until the end */
		if (task->flags & PF_EXITING)
2106
			goto next;
2107

2108 2109
		/* leave @task alone if post_fork() hasn't linked it yet */
		if (list_empty(&task->cg_list))
2110
			goto next;
2111

2112
		cset = task_css_set(task);
2113
		if (!cset->mg_src_cgrp)
2114
			goto next;
2115

2116
		/*
2117 2118
		 * cgroup_taskset_first() must always return the leader.
		 * Take care to avoid disturbing the ordering.
2119
		 */
2120 2121 2122 2123 2124 2125
		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);
2126
	next:
2127 2128
		if (!threadgroup)
			break;
2129
	} while_each_thread(leader, task);
2130
	rcu_read_unlock();
2131
	up_write(&css_set_rwsem);
B
Ben Blum 已提交
2132

2133
	/* methods shouldn't be called if no task is actually migrating */
2134 2135
	if (list_empty(&tset.src_csets))
		return 0;
2136

2137
	/* check that we can legitimately attach to the cgroup */
2138
	for_each_e_css(css, i, cgrp) {
T
Tejun Heo 已提交
2139
		if (css->ss->can_attach) {
2140 2141
			ret = css->ss->can_attach(css, &tset);
			if (ret) {
T
Tejun Heo 已提交
2142
				failed_css = css;
B
Ben Blum 已提交
2143 2144 2145 2146 2147 2148
				goto out_cancel_attach;
			}
		}
	}

	/*
2149 2150 2151
	 * 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
2152
	 */
2153
	down_write(&css_set_rwsem);
2154 2155 2156 2157
	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
2158
	}
2159
	up_write(&css_set_rwsem);
B
Ben Blum 已提交
2160 2161

	/*
2162 2163 2164
	 * 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
2165
	 */
2166
	tset.csets = &tset.dst_csets;
B
Ben Blum 已提交
2167

2168
	for_each_e_css(css, i, cgrp)
T
Tejun Heo 已提交
2169 2170
		if (css->ss->attach)
			css->ss->attach(css, &tset);
B
Ben Blum 已提交
2171

2172
	ret = 0;
2173 2174
	goto out_release_tset;

B
Ben Blum 已提交
2175
out_cancel_attach:
2176
	for_each_e_css(css, i, cgrp) {
2177 2178 2179 2180
		if (css == failed_css)
			break;
		if (css->ss->cancel_attach)
			css->ss->cancel_attach(css, &tset);
B
Ben Blum 已提交
2181
	}
2182 2183 2184 2185
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) {
2186
		list_splice_tail_init(&cset->mg_tasks, &cset->tasks);
2187 2188 2189
		list_del_init(&cset->mg_node);
	}
	up_write(&css_set_rwsem);
2190
	return ret;
B
Ben Blum 已提交
2191 2192
}

2193 2194 2195 2196 2197 2198
/**
 * 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?
 *
2199
 * Call holding cgroup_mutex and threadgroup_lock of @leader.
2200 2201 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
 */
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
2228 2229 2230 2231
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2232
 * function to attach either it or all tasks in its threadgroup. Will lock
2233
 * cgroup_mutex and threadgroup.
2234
 */
2235 2236
static ssize_t __cgroup_procs_write(struct kernfs_open_file *of, char *buf,
				    size_t nbytes, loff_t off, bool threadgroup)
2237 2238
{
	struct task_struct *tsk;
2239
	const struct cred *cred = current_cred(), *tcred;
2240 2241
	struct cgroup *cgrp = of_css(of)->cgroup;
	pid_t pid;
2242 2243
	int ret;

2244 2245 2246
	if (kstrtoint(strstrip(buf), 0, &pid) || pid < 0)
		return -EINVAL;

B
Ben Blum 已提交
2247 2248 2249
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2250 2251
retry_find_task:
	rcu_read_lock();
2252
	if (pid) {
2253
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2254 2255
		if (!tsk) {
			rcu_read_unlock();
S
SeongJae Park 已提交
2256
			ret = -ESRCH;
2257
			goto out_unlock_cgroup;
2258
		}
B
Ben Blum 已提交
2259 2260 2261 2262
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2263
		tcred = __task_cred(tsk);
2264 2265 2266
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2267
			rcu_read_unlock();
2268 2269
			ret = -EACCES;
			goto out_unlock_cgroup;
2270
		}
2271 2272
	} else
		tsk = current;
2273 2274

	if (threadgroup)
2275
		tsk = tsk->group_leader;
2276 2277

	/*
2278
	 * Workqueue threads may acquire PF_NO_SETAFFINITY and become
2279 2280 2281
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
2282
	if (tsk == kthreadd_task || (tsk->flags & PF_NO_SETAFFINITY)) {
2283 2284 2285 2286 2287
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304
	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;
		}
2305 2306 2307 2308
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2309 2310
	threadgroup_unlock(tsk);

2311
	put_task_struct(tsk);
2312
out_unlock_cgroup:
T
Tejun Heo 已提交
2313
	mutex_unlock(&cgroup_mutex);
2314
	return ret ?: nbytes;
2315 2316
}

2317 2318 2319 2320 2321 2322 2323
/**
 * 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)
{
2324
	struct cgroup_root *root;
2325 2326
	int retval = 0;

T
Tejun Heo 已提交
2327
	mutex_lock(&cgroup_mutex);
2328
	for_each_root(root) {
2329 2330
		struct cgroup *from_cgrp;

2331
		if (root == &cgrp_dfl_root)
2332 2333
			continue;

2334 2335 2336
		down_read(&css_set_rwsem);
		from_cgrp = task_cgroup_from_root(from, root);
		up_read(&css_set_rwsem);
2337

L
Li Zefan 已提交
2338
		retval = cgroup_attach_task(from_cgrp, tsk, false);
2339 2340 2341
		if (retval)
			break;
	}
T
Tejun Heo 已提交
2342
	mutex_unlock(&cgroup_mutex);
2343 2344 2345 2346 2347

	return retval;
}
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);

2348 2349
static ssize_t cgroup_tasks_write(struct kernfs_open_file *of,
				  char *buf, size_t nbytes, loff_t off)
B
Ben Blum 已提交
2350
{
2351
	return __cgroup_procs_write(of, buf, nbytes, off, false);
B
Ben Blum 已提交
2352 2353
}

2354 2355
static ssize_t cgroup_procs_write(struct kernfs_open_file *of,
				  char *buf, size_t nbytes, loff_t off)
2356
{
2357
	return __cgroup_procs_write(of, buf, nbytes, off, true);
2358 2359
}

2360 2361
static ssize_t cgroup_release_agent_write(struct kernfs_open_file *of,
					  char *buf, size_t nbytes, loff_t off)
2362
{
2363 2364
	struct cgroup *cgrp = of_css(of)->cgroup;
	struct cgroup_root *root = cgrp->root;
2365 2366

	BUILD_BUG_ON(sizeof(root->release_agent_path) < PATH_MAX);
2367
	if (!cgroup_lock_live_group(cgrp))
2368
		return -ENODEV;
2369
	spin_lock(&release_agent_path_lock);
2370
	strlcpy(root->release_agent_path, strstrip(buf),
2371
		sizeof(root->release_agent_path));
2372
	spin_unlock(&release_agent_path_lock);
T
Tejun Heo 已提交
2373
	mutex_unlock(&cgroup_mutex);
2374
	return nbytes;
2375 2376
}

2377
static int cgroup_release_agent_show(struct seq_file *seq, void *v)
2378
{
2379
	struct cgroup *cgrp = seq_css(seq)->cgroup;
2380

2381
	spin_lock(&release_agent_path_lock);
2382
	seq_puts(seq, cgrp->root->release_agent_path);
2383
	spin_unlock(&release_agent_path_lock);
2384 2385 2386 2387
	seq_putc(seq, '\n');
	return 0;
}

2388
static int cgroup_sane_behavior_show(struct seq_file *seq, void *v)
2389
{
2390 2391 2392
	struct cgroup *cgrp = seq_css(seq)->cgroup;

	seq_printf(seq, "%d\n", cgroup_sane_behavior(cgrp));
2393 2394 2395
	return 0;
}

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 2538
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 */
2539 2540 2541
static ssize_t cgroup_subtree_control_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
2542
{
2543
	unsigned int enable = 0, disable = 0;
2544
	struct cgroup *cgrp = of_css(of)->cgroup, *child;
2545
	struct cgroup_subsys *ss;
2546
	char *tok;
2547 2548 2549
	int ssid, ret;

	/*
2550 2551
	 * Parse input - space separated list of subsystem names prefixed
	 * with either + or -.
2552
	 */
2553 2554
	buf = strstrip(buf);
	while ((tok = strsep(&buf, " "))) {
2555 2556
		if (tok[0] == '\0')
			continue;
2557 2558 2559 2560 2561
		for_each_subsys(ss, ssid) {
			if (ss->disabled || strcmp(tok + 1, ss->name))
				continue;

			if (*tok == '+') {
2562 2563
				enable |= 1 << ssid;
				disable &= ~(1 << ssid);
2564
			} else if (*tok == '-') {
2565 2566
				disable |= 1 << ssid;
				enable &= ~(1 << ssid);
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600
			} 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);

	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) {
2601
				DEFINE_WAIT(wait);
2602 2603 2604 2605

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

2606
				cgroup_get(child);
2607 2608 2609 2610 2611
				prepare_to_wait(&child->offline_waitq, &wait,
						TASK_UNINTERRUPTIBLE);
				mutex_unlock(&cgroup_tree_mutex);
				schedule();
				finish_wait(&child->offline_waitq, &wait);
2612
				cgroup_put(child);
2613 2614 2615

				ret = restart_syscall();
				goto out_unbreak;
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
			}

			/* 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);
2699
out_unbreak:
2700 2701
	kernfs_unbreak_active_protection(cgrp->control_kn);
	cgroup_put(cgrp);
2702
	return ret ?: nbytes;
2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720

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;
}

2721 2722 2723 2724 2725 2726
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
2727 2728
static ssize_t cgroup_file_write(struct kernfs_open_file *of, char *buf,
				 size_t nbytes, loff_t off)
2729
{
T
Tejun Heo 已提交
2730 2731 2732
	struct cgroup *cgrp = of->kn->parent->priv;
	struct cftype *cft = of->kn->priv;
	struct cgroup_subsys_state *css;
2733
	int ret;
2734

T
Tejun Heo 已提交
2735 2736 2737
	if (cft->write)
		return cft->write(of, buf, nbytes, off);

T
Tejun Heo 已提交
2738 2739 2740 2741 2742 2743 2744 2745 2746
	/*
	 * 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();
2747

2748
	if (cft->write_u64) {
2749 2750 2751 2752 2753 2754 2755 2756 2757
		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);
2758
	} else {
2759
		ret = -EINVAL;
2760
	}
T
Tejun Heo 已提交
2761

2762
	return ret ?: nbytes;
2763 2764
}

2765
static void *cgroup_seqfile_start(struct seq_file *seq, loff_t *ppos)
2766
{
T
Tejun Heo 已提交
2767
	return seq_cft(seq)->seq_start(seq, ppos);
2768 2769
}

2770
static void *cgroup_seqfile_next(struct seq_file *seq, void *v, loff_t *ppos)
2771
{
T
Tejun Heo 已提交
2772
	return seq_cft(seq)->seq_next(seq, v, ppos);
2773 2774
}

2775
static void cgroup_seqfile_stop(struct seq_file *seq, void *v)
2776
{
T
Tejun Heo 已提交
2777
	seq_cft(seq)->seq_stop(seq, v);
2778 2779
}

2780
static int cgroup_seqfile_show(struct seq_file *m, void *arg)
2781
{
2782 2783
	struct cftype *cft = seq_cft(m);
	struct cgroup_subsys_state *css = seq_css(m);
2784

2785 2786
	if (cft->seq_show)
		return cft->seq_show(m, arg);
2787

2788
	if (cft->read_u64)
2789 2790 2791 2792 2793 2794
		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;
2795 2796
}

T
Tejun Heo 已提交
2797 2798 2799 2800
static struct kernfs_ops cgroup_kf_single_ops = {
	.atomic_write_len	= PAGE_SIZE,
	.write			= cgroup_file_write,
	.seq_show		= cgroup_seqfile_show,
2801 2802
};

T
Tejun Heo 已提交
2803 2804 2805 2806 2807 2808 2809 2810
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,
};
2811 2812 2813 2814

/*
 * cgroup_rename - Only allow simple rename of directories in place.
 */
T
Tejun Heo 已提交
2815 2816
static int cgroup_rename(struct kernfs_node *kn, struct kernfs_node *new_parent,
			 const char *new_name_str)
2817
{
T
Tejun Heo 已提交
2818
	struct cgroup *cgrp = kn->priv;
2819 2820
	int ret;

T
Tejun Heo 已提交
2821
	if (kernfs_type(kn) != KERNFS_DIR)
2822
		return -ENOTDIR;
T
Tejun Heo 已提交
2823
	if (kn->parent != new_parent)
2824
		return -EIO;
2825

2826 2827 2828 2829 2830 2831
	/*
	 * 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
2832

2833 2834 2835 2836 2837 2838 2839
	/*
	 * 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
2840

T
Tejun Heo 已提交
2841 2842
	mutex_lock(&cgroup_tree_mutex);
	mutex_lock(&cgroup_mutex);
L
Li Zefan 已提交
2843

T
Tejun Heo 已提交
2844
	ret = kernfs_rename(kn, new_parent, new_name_str);
L
Li Zefan 已提交
2845

T
Tejun Heo 已提交
2846 2847
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&cgroup_tree_mutex);
2848 2849 2850

	kernfs_unbreak_active_protection(kn);
	kernfs_unbreak_active_protection(new_parent);
T
Tejun Heo 已提交
2851
	return ret;
L
Li Zefan 已提交
2852 2853
}

2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
/* 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);
}

2868
static int cgroup_add_file(struct cgroup *cgrp, struct cftype *cft)
2869
{
T
Tejun Heo 已提交
2870
	char name[CGROUP_FILE_NAME_MAX];
T
Tejun Heo 已提交
2871 2872
	struct kernfs_node *kn;
	struct lock_class_key *key = NULL;
2873
	int ret;
T
Tejun Heo 已提交
2874

T
Tejun Heo 已提交
2875 2876 2877 2878 2879 2880
#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);
2881 2882 2883 2884
	if (IS_ERR(kn))
		return PTR_ERR(kn);

	ret = cgroup_kn_set_ugid(kn);
2885
	if (ret) {
2886
		kernfs_remove(kn);
2887 2888 2889 2890 2891
		return ret;
	}

	if (cft->seq_show == cgroup_subtree_control_show)
		cgrp->control_kn = kn;
2892 2893
	else if (cft->seq_show == cgroup_populated_show)
		cgrp->populated_kn = kn;
2894
	return 0;
2895 2896
}

2897 2898 2899 2900 2901 2902 2903
/**
 * 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.
2904 2905 2906
 * For removals, this function never fails.  If addition fails, this
 * function doesn't remove files already added.  The caller is responsible
 * for cleaning up.
2907
 */
2908 2909
static int cgroup_addrm_files(struct cgroup *cgrp, struct cftype cfts[],
			      bool is_add)
2910
{
A
Aristeu Rozanski 已提交
2911
	struct cftype *cft;
2912 2913
	int ret;

T
Tejun Heo 已提交
2914
	lockdep_assert_held(&cgroup_tree_mutex);
T
Tejun Heo 已提交
2915 2916

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2917
		/* does cft->flags tell us to skip this file on @cgrp? */
T
Tejun Heo 已提交
2918 2919
		if ((cft->flags & CFTYPE_ONLY_ON_DFL) && !cgroup_on_dfl(cgrp))
			continue;
2920 2921
		if ((cft->flags & CFTYPE_INSANE) && cgroup_sane_behavior(cgrp))
			continue;
2922 2923 2924 2925 2926
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2927
		if (is_add) {
2928
			ret = cgroup_add_file(cgrp, cft);
2929
			if (ret) {
2930 2931
				pr_warn("%s: failed to add %s, err=%d\n",
					__func__, cft->name, ret);
2932 2933
				return ret;
			}
2934 2935
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2936
		}
2937
	}
2938
	return 0;
2939 2940
}

2941
static int cgroup_apply_cftypes(struct cftype *cfts, bool is_add)
2942 2943
{
	LIST_HEAD(pending);
2944
	struct cgroup_subsys *ss = cfts[0].ss;
2945
	struct cgroup *root = &ss->root->cgrp;
2946
	struct cgroup_subsys_state *css;
2947
	int ret = 0;
2948

2949
	lockdep_assert_held(&cgroup_tree_mutex);
2950

2951
	/* add/rm files for all cgroups created before */
2952
	css_for_each_descendant_pre(css, cgroup_css(root, ss)) {
2953 2954
		struct cgroup *cgrp = css->cgroup;

2955 2956 2957
		if (cgroup_is_dead(cgrp))
			continue;

2958
		ret = cgroup_addrm_files(cgrp, cfts, is_add);
2959 2960
		if (ret)
			break;
2961
	}
2962 2963 2964

	if (is_add && !ret)
		kernfs_activate(root->kn);
2965
	return ret;
2966 2967
}

2968
static void cgroup_exit_cftypes(struct cftype *cfts)
2969
{
2970
	struct cftype *cft;
2971

T
Tejun Heo 已提交
2972 2973 2974 2975 2976
	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;
2977
		cft->ss = NULL;
T
Tejun Heo 已提交
2978
	}
2979 2980
}

T
Tejun Heo 已提交
2981
static int cgroup_init_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2982 2983 2984
{
	struct cftype *cft;

T
Tejun Heo 已提交
2985 2986 2987
	for (cft = cfts; cft->name[0] != '\0'; cft++) {
		struct kernfs_ops *kf_ops;

T
Tejun Heo 已提交
2988 2989
		WARN_ON(cft->ss || cft->kf_ops);

T
Tejun Heo 已提交
2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006
		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;
		}
3007

T
Tejun Heo 已提交
3008
		cft->kf_ops = kf_ops;
3009
		cft->ss = ss;
T
Tejun Heo 已提交
3010
	}
3011

T
Tejun Heo 已提交
3012
	return 0;
3013 3014
}

3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025
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;
3026 3027
}

3028 3029 3030 3031
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
3032 3033 3034
 * 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.
3035 3036
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
3037
 * registered.
3038
 */
3039
int cgroup_rm_cftypes(struct cftype *cfts)
3040
{
3041
	int ret;
3042

3043 3044 3045 3046
	mutex_lock(&cgroup_tree_mutex);
	ret = cgroup_rm_cftypes_locked(cfts);
	mutex_unlock(&cgroup_tree_mutex);
	return ret;
T
Tejun Heo 已提交
3047 3048
}

3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062
/**
 * 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
3063
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
3064
{
3065
	int ret;
3066

3067 3068
	if (!cfts || cfts[0].name[0] == '\0')
		return 0;
3069

T
Tejun Heo 已提交
3070 3071 3072
	ret = cgroup_init_cftypes(ss, cfts);
	if (ret)
		return ret;
3073

3074 3075
	mutex_lock(&cgroup_tree_mutex);

T
Tejun Heo 已提交
3076
	list_add_tail(&cfts->node, &ss->cfts);
3077
	ret = cgroup_apply_cftypes(cfts, true);
3078
	if (ret)
3079
		cgroup_rm_cftypes_locked(cfts);
3080

3081
	mutex_unlock(&cgroup_tree_mutex);
3082
	return ret;
3083 3084
}

L
Li Zefan 已提交
3085 3086 3087 3088 3089 3090
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
3091
static int cgroup_task_count(const struct cgroup *cgrp)
3092 3093
{
	int count = 0;
3094
	struct cgrp_cset_link *link;
3095

3096
	down_read(&css_set_rwsem);
3097 3098
	list_for_each_entry(link, &cgrp->cset_links, cset_link)
		count += atomic_read(&link->cset->refcount);
3099
	up_read(&css_set_rwsem);
3100 3101 3102
	return count;
}

3103
/**
3104 3105 3106
 * 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
3107
 *
3108
 * This function returns the next child of @parent_css and should be called
3109 3110 3111
 * 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.
3112
 */
3113 3114 3115
struct cgroup_subsys_state *
css_next_child(struct cgroup_subsys_state *pos_css,
	       struct cgroup_subsys_state *parent_css)
3116
{
3117 3118
	struct cgroup *pos = pos_css ? pos_css->cgroup : NULL;
	struct cgroup *cgrp = parent_css->cgroup;
3119 3120
	struct cgroup *next;

T
Tejun Heo 已提交
3121
	cgroup_assert_mutexes_or_rcu_locked();
3122 3123 3124 3125

	/*
	 * @pos could already have been removed.  Once a cgroup is removed,
	 * its ->sibling.next is no longer updated when its next sibling
3126 3127 3128 3129 3130 3131 3132
	 * 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.
3133 3134 3135 3136 3137 3138 3139 3140
	 *
	 * 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.
3141
	 */
3142 3143 3144
	if (!pos) {
		next = list_entry_rcu(cgrp->children.next, struct cgroup, sibling);
	} else if (likely(!cgroup_is_dead(pos))) {
3145
		next = list_entry_rcu(pos->sibling.next, struct cgroup, sibling);
3146 3147 3148 3149
	} else {
		list_for_each_entry_rcu(next, &cgrp->children, sibling)
			if (next->serial_nr > pos->serial_nr)
				break;
3150 3151
	}

3152 3153 3154 3155 3156 3157 3158
	/*
	 * @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);
3159

3160 3161 3162 3163 3164
		if (next_css)
			return next_css;
		next = list_entry_rcu(next->sibling.next, struct cgroup, sibling);
	}
	return NULL;
3165 3166
}

3167
/**
3168
 * css_next_descendant_pre - find the next descendant for pre-order walk
3169
 * @pos: the current position (%NULL to initiate traversal)
3170
 * @root: css whose descendants to walk
3171
 *
3172
 * To be used by css_for_each_descendant_pre().  Find the next descendant
3173 3174
 * to visit for pre-order traversal of @root's descendants.  @root is
 * included in the iteration and the first node to be visited.
3175
 *
3176 3177 3178 3179
 * 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.
3180
 */
3181 3182 3183
struct cgroup_subsys_state *
css_next_descendant_pre(struct cgroup_subsys_state *pos,
			struct cgroup_subsys_state *root)
3184
{
3185
	struct cgroup_subsys_state *next;
3186

T
Tejun Heo 已提交
3187
	cgroup_assert_mutexes_or_rcu_locked();
3188

3189
	/* if first iteration, visit @root */
3190
	if (!pos)
3191
		return root;
3192 3193

	/* visit the first child if exists */
3194
	next = css_next_child(NULL, pos);
3195 3196 3197 3198
	if (next)
		return next;

	/* no child, visit my or the closest ancestor's next sibling */
3199 3200
	while (pos != root) {
		next = css_next_child(pos, css_parent(pos));
3201
		if (next)
3202
			return next;
3203
		pos = css_parent(pos);
3204
	}
3205 3206 3207 3208

	return NULL;
}

3209
/**
3210 3211
 * css_rightmost_descendant - return the rightmost descendant of a css
 * @pos: css of interest
3212
 *
3213 3214
 * Return the rightmost descendant of @pos.  If there's no descendant, @pos
 * is returned.  This can be used during pre-order traversal to skip
3215
 * subtree of @pos.
3216
 *
3217 3218 3219 3220
 * 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.
3221
 */
3222 3223
struct cgroup_subsys_state *
css_rightmost_descendant(struct cgroup_subsys_state *pos)
3224
{
3225
	struct cgroup_subsys_state *last, *tmp;
3226

T
Tejun Heo 已提交
3227
	cgroup_assert_mutexes_or_rcu_locked();
3228 3229 3230 3231 3232

	do {
		last = pos;
		/* ->prev isn't RCU safe, walk ->next till the end */
		pos = NULL;
3233
		css_for_each_child(tmp, last)
3234 3235 3236 3237 3238 3239
			pos = tmp;
	} while (pos);

	return last;
}

3240 3241
static struct cgroup_subsys_state *
css_leftmost_descendant(struct cgroup_subsys_state *pos)
3242
{
3243
	struct cgroup_subsys_state *last;
3244 3245 3246

	do {
		last = pos;
3247
		pos = css_next_child(NULL, pos);
3248 3249 3250 3251 3252 3253
	} while (pos);

	return last;
}

/**
3254
 * css_next_descendant_post - find the next descendant for post-order walk
3255
 * @pos: the current position (%NULL to initiate traversal)
3256
 * @root: css whose descendants to walk
3257
 *
3258
 * To be used by css_for_each_descendant_post().  Find the next descendant
3259 3260
 * to visit for post-order traversal of @root's descendants.  @root is
 * included in the iteration and the last node to be visited.
3261
 *
3262 3263 3264 3265 3266
 * 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.
3267
 */
3268 3269 3270
struct cgroup_subsys_state *
css_next_descendant_post(struct cgroup_subsys_state *pos,
			 struct cgroup_subsys_state *root)
3271
{
3272
	struct cgroup_subsys_state *next;
3273

T
Tejun Heo 已提交
3274
	cgroup_assert_mutexes_or_rcu_locked();
3275

3276 3277 3278
	/* if first iteration, visit leftmost descendant which may be @root */
	if (!pos)
		return css_leftmost_descendant(root);
3279

3280 3281 3282 3283
	/* if we visited @root, we're done */
	if (pos == root)
		return NULL;

3284
	/* if there's an unvisited sibling, visit its leftmost descendant */
3285
	next = css_next_child(pos, css_parent(pos));
3286
	if (next)
3287
		return css_leftmost_descendant(next);
3288 3289

	/* no sibling left, visit parent */
3290
	return css_parent(pos);
3291 3292
}

3293
/**
3294
 * css_advance_task_iter - advance a task itererator to the next css_set
3295 3296 3297
 * @it: the iterator to advance
 *
 * Advance @it to the next css_set to walk.
3298
 */
3299
static void css_advance_task_iter(struct css_task_iter *it)
3300
{
T
Tejun Heo 已提交
3301
	struct list_head *l = it->cset_pos;
3302 3303 3304 3305 3306 3307
	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
3308 3309
		if (l == it->cset_head) {
			it->cset_pos = NULL;
3310 3311
			return;
		}
3312 3313 3314 3315 3316 3317 3318 3319

		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
3320 3321
	} while (list_empty(&cset->tasks) && list_empty(&cset->mg_tasks));

T
Tejun Heo 已提交
3322
	it->cset_pos = l;
T
Tejun Heo 已提交
3323 3324

	if (!list_empty(&cset->tasks))
T
Tejun Heo 已提交
3325
		it->task_pos = cset->tasks.next;
T
Tejun Heo 已提交
3326
	else
T
Tejun Heo 已提交
3327 3328 3329 3330
		it->task_pos = cset->mg_tasks.next;

	it->tasks_head = &cset->tasks;
	it->mg_tasks_head = &cset->mg_tasks;
3331 3332
}

3333
/**
3334 3335
 * css_task_iter_start - initiate task iteration
 * @css: the css to walk tasks of
3336 3337
 * @it: the task iterator to use
 *
3338 3339 3340 3341
 * 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.
3342 3343 3344 3345 3346
 *
 * Note that this function acquires a lock which is released when the
 * iteration finishes.  The caller can't sleep while iteration is in
 * progress.
 */
3347 3348
void css_task_iter_start(struct cgroup_subsys_state *css,
			 struct css_task_iter *it)
3349
	__acquires(css_set_rwsem)
3350
{
3351 3352
	/* no one should try to iterate before mounting cgroups */
	WARN_ON_ONCE(!use_task_css_set_links);
3353

3354
	down_read(&css_set_rwsem);
3355

3356 3357 3358 3359 3360 3361 3362
	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
3363
	it->cset_head = it->cset_pos;
3364

3365
	css_advance_task_iter(it);
3366 3367
}

3368
/**
3369
 * css_task_iter_next - return the next task for the iterator
3370 3371 3372
 * @it: the task iterator being iterated
 *
 * The "next" function for task iteration.  @it should have been
3373 3374
 * initialized via css_task_iter_start().  Returns NULL when the iteration
 * reaches the end.
3375
 */
3376
struct task_struct *css_task_iter_next(struct css_task_iter *it)
3377 3378
{
	struct task_struct *res;
T
Tejun Heo 已提交
3379
	struct list_head *l = it->task_pos;
3380 3381

	/* If the iterator cg is NULL, we have no tasks */
T
Tejun Heo 已提交
3382
	if (!it->cset_pos)
3383 3384
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
T
Tejun Heo 已提交
3385 3386 3387 3388 3389 3390

	/*
	 * Advance iterator to find next entry.  cset->tasks is consumed
	 * first and then ->mg_tasks.  After ->mg_tasks, we move onto the
	 * next cset.
	 */
3391
	l = l->next;
T
Tejun Heo 已提交
3392

T
Tejun Heo 已提交
3393 3394
	if (l == it->tasks_head)
		l = it->mg_tasks_head->next;
T
Tejun Heo 已提交
3395

T
Tejun Heo 已提交
3396
	if (l == it->mg_tasks_head)
3397
		css_advance_task_iter(it);
T
Tejun Heo 已提交
3398
	else
T
Tejun Heo 已提交
3399
		it->task_pos = l;
T
Tejun Heo 已提交
3400

3401 3402 3403
	return res;
}

3404
/**
3405
 * css_task_iter_end - finish task iteration
3406 3407
 * @it: the task iterator to finish
 *
3408
 * Finish task iteration started by css_task_iter_start().
3409
 */
3410
void css_task_iter_end(struct css_task_iter *it)
3411
	__releases(css_set_rwsem)
3412
{
3413
	up_read(&css_set_rwsem);
3414 3415 3416
}

/**
3417 3418 3419
 * 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
3420
 *
3421 3422 3423 3424 3425
 * 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.
3426
 */
3427
int cgroup_transfer_tasks(struct cgroup *to, struct cgroup *from)
3428
{
3429 3430
	LIST_HEAD(preloaded_csets);
	struct cgrp_cset_link *link;
3431
	struct css_task_iter it;
3432
	struct task_struct *task;
3433
	int ret;
3434

3435
	mutex_lock(&cgroup_mutex);
3436

3437 3438 3439 3440 3441
	/* 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);
3442

3443 3444 3445
	ret = cgroup_migrate_prepare_dst(to, &preloaded_csets);
	if (ret)
		goto out_err;
3446

3447 3448 3449 3450
	/*
	 * Migrate tasks one-by-one until @form is empty.  This fails iff
	 * ->can_attach() fails.
	 */
3451 3452 3453 3454 3455 3456 3457 3458
	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) {
3459
			ret = cgroup_migrate(to, task, false);
3460 3461 3462
			put_task_struct(task);
		}
	} while (task && !ret);
3463 3464
out_err:
	cgroup_migrate_finish(&preloaded_csets);
T
Tejun Heo 已提交
3465
	mutex_unlock(&cgroup_mutex);
3466
	return ret;
3467 3468
}

3469
/*
3470
 * Stuff for reading the 'tasks'/'procs' files.
3471 3472 3473 3474 3475 3476 3477 3478
 *
 * 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.
 *
 */

3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504
/* 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;
3505 3506
	/* for delayed destruction */
	struct delayed_work destroy_dwork;
3507 3508
};

3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521
/*
 * 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);
}
3522

3523 3524 3525 3526 3527 3528 3529 3530
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}

3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557
/*
 * 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);

	/*
3558 3559
	 * Destroy iff we didn't get queued again.  The state won't change
	 * as destroy_dwork can only be queued while locked.
3560
	 */
3561
	if (!delayed_work_pending(dwork)) {
3562 3563 3564 3565 3566 3567 3568 3569 3570 3571
		list_del(&l->links);
		pidlist_free(l->list);
		put_pid_ns(l->key.ns);
		tofree = l;
	}

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

3572
/*
3573
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3574
 * Returns the number of unique elements.
3575
 */
3576
static int pidlist_uniq(pid_t *list, int length)
3577
{
3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601
	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;
}

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 3629 3630 3631 3632 3633 3634
/*
 * 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;
}

3635 3636 3637 3638 3639
static int cmppid(const void *a, const void *b)
{
	return *(pid_t *)a - *(pid_t *)b;
}

3640 3641 3642 3643 3644
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
3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659
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;
}

3660 3661 3662 3663 3664 3665
/*
 * 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
3666 3667
static struct cgroup_pidlist *cgroup_pidlist_find_create(struct cgroup *cgrp,
						enum cgroup_filetype type)
3668 3669
{
	struct cgroup_pidlist *l;
3670

T
Tejun Heo 已提交
3671 3672 3673 3674 3675 3676
	lockdep_assert_held(&cgrp->pidlist_mutex);

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

3677
	/* entry not found; create a new one */
3678
	l = kzalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
T
Tejun Heo 已提交
3679
	if (!l)
3680
		return l;
T
Tejun Heo 已提交
3681

3682
	INIT_DELAYED_WORK(&l->destroy_dwork, cgroup_pidlist_destroy_work_fn);
3683
	l->key.type = type;
T
Tejun Heo 已提交
3684 3685
	/* don't need task_nsproxy() if we're looking at ourself */
	l->key.ns = get_pid_ns(task_active_pid_ns(current));
3686 3687 3688 3689 3690
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	return l;
}

3691 3692 3693
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3694 3695
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3696 3697 3698 3699
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3700
	struct css_task_iter it;
3701
	struct task_struct *tsk;
3702 3703
	struct cgroup_pidlist *l;

3704 3705
	lockdep_assert_held(&cgrp->pidlist_mutex);

3706 3707 3708 3709 3710 3711 3712
	/*
	 * 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);
3713
	array = pidlist_allocate(length);
3714 3715 3716
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3717 3718
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
3719
		if (unlikely(n == length))
3720
			break;
3721
		/* get tgid or pid for procs or tasks file respectively */
3722 3723 3724 3725
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3726 3727
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3728
	}
3729
	css_task_iter_end(&it);
3730 3731
	length = n;
	/* now sort & (if procs) strip out duplicates */
3732 3733 3734 3735
	if (cgroup_sane_behavior(cgrp))
		sort(array, length, sizeof(pid_t), fried_cmppid, NULL);
	else
		sort(array, length, sizeof(pid_t), cmppid, NULL);
3736
	if (type == CGROUP_FILE_PROCS)
3737
		length = pidlist_uniq(array, length);
T
Tejun Heo 已提交
3738 3739

	l = cgroup_pidlist_find_create(cgrp, type);
3740
	if (!l) {
T
Tejun Heo 已提交
3741
		mutex_unlock(&cgrp->pidlist_mutex);
3742
		pidlist_free(array);
3743
		return -ENOMEM;
3744
	}
T
Tejun Heo 已提交
3745 3746

	/* store array, freeing old if necessary */
3747
	pidlist_free(l->list);
3748 3749
	l->list = array;
	l->length = length;
3750
	*lp = l;
3751
	return 0;
3752 3753
}

B
Balbir Singh 已提交
3754
/**
L
Li Zefan 已提交
3755
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3756 3757 3758
 * @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
3759 3760 3761
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3762 3763 3764
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
T
Tejun Heo 已提交
3765
	struct kernfs_node *kn = kernfs_node_from_dentry(dentry);
3766
	struct cgroup *cgrp;
3767
	struct css_task_iter it;
B
Balbir Singh 已提交
3768
	struct task_struct *tsk;
3769

T
Tejun Heo 已提交
3770 3771 3772 3773 3774
	/* 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;

3775 3776
	mutex_lock(&cgroup_mutex);

B
Balbir Singh 已提交
3777
	/*
T
Tejun Heo 已提交
3778
	 * We aren't being called from kernfs and there's no guarantee on
3779
	 * @kn->priv's validity.  For this and css_tryget_online_from_dir(),
T
Tejun Heo 已提交
3780
	 * @kn->priv is RCU safe.  Let's do the RCU dancing.
B
Balbir Singh 已提交
3781
	 */
T
Tejun Heo 已提交
3782 3783
	rcu_read_lock();
	cgrp = rcu_dereference(kn->priv);
3784
	if (!cgrp || cgroup_is_dead(cgrp)) {
T
Tejun Heo 已提交
3785
		rcu_read_unlock();
3786
		mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
3787 3788
		return -ENOENT;
	}
3789
	rcu_read_unlock();
B
Balbir Singh 已提交
3790

3791 3792
	css_task_iter_start(&cgrp->dummy_css, &it);
	while ((tsk = css_task_iter_next(&it))) {
B
Balbir Singh 已提交
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811
		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;
		}
	}
3812
	css_task_iter_end(&it);
B
Balbir Singh 已提交
3813

3814
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
3815
	return 0;
B
Balbir Singh 已提交
3816 3817
}

3818

3819
/*
3820
 * seq_file methods for the tasks/procs files. The seq_file position is the
3821
 * next pid to display; the seq_file iterator is a pointer to the pid
3822
 * in the cgroup->l->list array.
3823
 */
3824

3825
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3826
{
3827 3828 3829 3830 3831 3832
	/*
	 * 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
3833
	struct kernfs_open_file *of = s->private;
3834
	struct cgroup *cgrp = seq_css(s)->cgroup;
3835
	struct cgroup_pidlist *l;
3836
	enum cgroup_filetype type = seq_cft(s)->private;
3837
	int index = 0, pid = *pos;
3838 3839 3840 3841 3842
	int *iter, ret;

	mutex_lock(&cgrp->pidlist_mutex);

	/*
3843
	 * !NULL @of->priv indicates that this isn't the first start()
3844
	 * after open.  If the matching pidlist is around, we can use that.
3845
	 * Look for it.  Note that @of->priv can't be used directly.  It
3846 3847
	 * could already have been destroyed.
	 */
3848 3849
	if (of->priv)
		of->priv = cgroup_pidlist_find(cgrp, type);
3850 3851 3852 3853 3854

	/*
	 * Either this is the first start() after open or the matching
	 * pidlist has been destroyed inbetween.  Create a new one.
	 */
3855 3856 3857
	if (!of->priv) {
		ret = pidlist_array_load(cgrp, type,
					 (struct cgroup_pidlist **)&of->priv);
3858 3859 3860
		if (ret)
			return ERR_PTR(ret);
	}
3861
	l = of->priv;
3862 3863

	if (pid) {
3864
		int end = l->length;
S
Stephen Rothwell 已提交
3865

3866 3867
		while (index < end) {
			int mid = (index + end) / 2;
3868
			if (cgroup_pid_fry(cgrp, l->list[mid]) == pid) {
3869 3870
				index = mid;
				break;
3871
			} else if (cgroup_pid_fry(cgrp, l->list[mid]) <= pid)
3872 3873 3874 3875 3876 3877
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3878
	if (index >= l->length)
3879 3880
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3881
	iter = l->list + index;
3882
	*pos = cgroup_pid_fry(cgrp, *iter);
3883 3884 3885
	return iter;
}

3886
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3887
{
T
Tejun Heo 已提交
3888
	struct kernfs_open_file *of = s->private;
3889
	struct cgroup_pidlist *l = of->priv;
3890

3891 3892
	if (l)
		mod_delayed_work(cgroup_pidlist_destroy_wq, &l->destroy_dwork,
3893
				 CGROUP_PIDLIST_DESTROY_DELAY);
3894
	mutex_unlock(&seq_css(s)->cgroup->pidlist_mutex);
3895 3896
}

3897
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3898
{
T
Tejun Heo 已提交
3899
	struct kernfs_open_file *of = s->private;
3900
	struct cgroup_pidlist *l = of->priv;
3901 3902
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3903 3904 3905 3906 3907 3908 3909 3910
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
3911
		*pos = cgroup_pid_fry(seq_css(s)->cgroup, *p);
3912 3913 3914 3915
		return p;
	}
}

3916
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3917 3918 3919
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3920

3921 3922
static u64 cgroup_read_notify_on_release(struct cgroup_subsys_state *css,
					 struct cftype *cft)
3923
{
3924
	return notify_on_release(css->cgroup);
3925 3926
}

3927 3928
static int cgroup_write_notify_on_release(struct cgroup_subsys_state *css,
					  struct cftype *cft, u64 val)
3929
{
3930
	clear_bit(CGRP_RELEASABLE, &css->cgroup->flags);
3931
	if (val)
3932
		set_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3933
	else
3934
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &css->cgroup->flags);
3935 3936 3937
	return 0;
}

3938 3939
static u64 cgroup_clone_children_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
3940
{
3941
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3942 3943
}

3944 3945
static int cgroup_clone_children_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
3946 3947
{
	if (val)
3948
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3949
	else
3950
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &css->cgroup->flags);
3951 3952 3953
	return 0;
}

3954
static struct cftype cgroup_base_files[] = {
3955
	{
3956
		.name = "cgroup.procs",
3957 3958 3959 3960
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
3961
		.private = CGROUP_FILE_PROCS,
3962
		.write = cgroup_procs_write,
B
Ben Blum 已提交
3963
		.mode = S_IRUGO | S_IWUSR,
3964
	},
3965 3966
	{
		.name = "cgroup.clone_children",
3967
		.flags = CFTYPE_INSANE,
3968 3969 3970
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
3971 3972 3973
	{
		.name = "cgroup.sane_behavior",
		.flags = CFTYPE_ONLY_ON_ROOT,
3974
		.seq_show = cgroup_sane_behavior_show,
3975
	},
3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989
	{
		.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,
3990
		.write = cgroup_subtree_control_write,
3991
	},
3992 3993 3994 3995 3996
	{
		.name = "cgroup.populated",
		.flags = CFTYPE_ONLY_ON_DFL | CFTYPE_NOT_ON_ROOT,
		.seq_show = cgroup_populated_show,
	},
3997 3998 3999 4000 4001 4002 4003 4004 4005

	/*
	 * 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 */
4006 4007 4008 4009
		.seq_start = cgroup_pidlist_start,
		.seq_next = cgroup_pidlist_next,
		.seq_stop = cgroup_pidlist_stop,
		.seq_show = cgroup_pidlist_show,
4010
		.private = CGROUP_FILE_TASKS,
4011
		.write = cgroup_tasks_write,
4012 4013 4014 4015 4016 4017 4018 4019
		.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,
	},
4020 4021
	{
		.name = "release_agent",
4022
		.flags = CFTYPE_INSANE | CFTYPE_ONLY_ON_ROOT,
4023
		.seq_show = cgroup_release_agent_show,
4024
		.write = cgroup_release_agent_write,
4025
		.max_write_len = PATH_MAX - 1,
4026
	},
T
Tejun Heo 已提交
4027
	{ }	/* terminate */
4028 4029
};

4030
/**
4031
 * cgroup_populate_dir - create subsys files in a cgroup directory
4032 4033
 * @cgrp: target cgroup
 * @subsys_mask: mask of the subsystem ids whose files should be added
4034 4035
 *
 * On failure, no file is added.
4036
 */
4037
static int cgroup_populate_dir(struct cgroup *cgrp, unsigned int subsys_mask)
4038 4039
{
	struct cgroup_subsys *ss;
4040
	int i, ret = 0;
4041

4042
	/* process cftsets of each subsystem */
4043
	for_each_subsys(ss, i) {
T
Tejun Heo 已提交
4044
		struct cftype *cfts;
4045

4046
		if (!(subsys_mask & (1 << i)))
4047
			continue;
4048

T
Tejun Heo 已提交
4049 4050
		list_for_each_entry(cfts, &ss->cfts, node) {
			ret = cgroup_addrm_files(cgrp, cfts, true);
4051 4052 4053
			if (ret < 0)
				goto err;
		}
4054 4055
	}
	return 0;
4056 4057 4058
err:
	cgroup_clear_dir(cgrp, subsys_mask);
	return ret;
4059 4060
}

4061 4062 4063 4064 4065 4066 4067
/*
 * 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
4068 4069 4070
 *    and thus css_tryget_online() 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().
4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082
 *
 * 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.
 */
4083
static void css_free_work_fn(struct work_struct *work)
4084 4085
{
	struct cgroup_subsys_state *css =
4086
		container_of(work, struct cgroup_subsys_state, destroy_work);
4087
	struct cgroup *cgrp = css->cgroup;
4088

4089 4090 4091
	if (css->parent)
		css_put(css->parent);

4092
	css->ss->css_free(css);
T
Tejun Heo 已提交
4093
	cgroup_put(cgrp);
4094 4095
}

4096
static void css_free_rcu_fn(struct rcu_head *rcu_head)
4097 4098
{
	struct cgroup_subsys_state *css =
4099
		container_of(rcu_head, struct cgroup_subsys_state, rcu_head);
4100

4101
	INIT_WORK(&css->destroy_work, css_free_work_fn);
4102
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4103 4104
}

4105 4106 4107 4108
static void css_release(struct percpu_ref *ref)
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);
4109 4110 4111
	struct cgroup_subsys *ss = css->ss;

	cgroup_idr_remove(&ss->css_idr, css->id);
4112

4113
	call_rcu(&css->rcu_head, css_free_rcu_fn);
4114 4115
}

4116 4117
static void init_and_link_css(struct cgroup_subsys_state *css,
			      struct cgroup_subsys *ss, struct cgroup *cgrp)
4118
{
4119 4120
	cgroup_get(cgrp);

4121
	css->cgroup = cgrp;
4122
	css->ss = ss;
4123
	css->flags = 0;
4124

4125
	if (cgrp->parent) {
4126
		css->parent = cgroup_css(cgrp->parent, ss);
4127 4128
		css_get(css->parent);
	} else {
4129
		css->flags |= CSS_ROOT;
4130
	}
4131

4132
	BUG_ON(cgroup_css(cgrp, ss));
4133 4134
}

4135
/* invoke ->css_online() on a new CSS and mark it online if successful */
4136
static int online_css(struct cgroup_subsys_state *css)
4137
{
4138
	struct cgroup_subsys *ss = css->ss;
T
Tejun Heo 已提交
4139 4140
	int ret = 0;

T
Tejun Heo 已提交
4141
	lockdep_assert_held(&cgroup_tree_mutex);
4142 4143
	lockdep_assert_held(&cgroup_mutex);

4144
	if (ss->css_online)
4145
		ret = ss->css_online(css);
4146
	if (!ret) {
4147
		css->flags |= CSS_ONLINE;
4148
		css->cgroup->nr_css++;
4149
		rcu_assign_pointer(css->cgroup->subsys[ss->id], css);
4150
	}
T
Tejun Heo 已提交
4151
	return ret;
4152 4153
}

4154
/* if the CSS is online, invoke ->css_offline() on it and mark it offline */
4155
static void offline_css(struct cgroup_subsys_state *css)
4156
{
4157
	struct cgroup_subsys *ss = css->ss;
4158

T
Tejun Heo 已提交
4159
	lockdep_assert_held(&cgroup_tree_mutex);
4160 4161 4162 4163 4164
	lockdep_assert_held(&cgroup_mutex);

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

4165
	if (ss->css_offline)
4166
		ss->css_offline(css);
4167

4168
	css->flags &= ~CSS_ONLINE;
4169
	css->cgroup->nr_css--;
4170
	RCU_INIT_POINTER(css->cgroup->subsys[ss->id], NULL);
4171 4172

	wake_up_all(&css->cgroup->offline_waitq);
4173 4174
}

4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195
/**
 * 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);

4196
	init_and_link_css(css, ss, cgrp);
4197

4198 4199
	err = percpu_ref_init(&css->refcnt, css_release);
	if (err)
4200
		goto err_free_css;
4201

4202 4203 4204 4205 4206
	err = cgroup_idr_alloc(&ss->css_idr, NULL, 2, 0, GFP_NOWAIT);
	if (err < 0)
		goto err_free_percpu_ref;
	css->id = err;

4207
	err = cgroup_populate_dir(cgrp, 1 << ss->id);
4208
	if (err)
4209 4210 4211 4212
		goto err_free_id;

	/* @css is ready to be brought online now, make it visible */
	cgroup_idr_replace(&ss->css_idr, css, css->id);
4213 4214 4215

	err = online_css(css);
	if (err)
4216
		goto err_clear_dir;
4217 4218 4219

	if (ss->broken_hierarchy && !ss->warned_broken_hierarchy &&
	    parent->parent) {
4220
		pr_warn("%s (%d) created nested cgroup for controller \"%s\" which has incomplete hierarchy support. Nested cgroups may change behavior in the future.\n",
4221
			current->comm, current->pid, ss->name);
4222
		if (!strcmp(ss->name, "memory"))
4223
			pr_warn("\"memory\" requires setting use_hierarchy to 1 on the root\n");
4224 4225 4226 4227 4228
		ss->warned_broken_hierarchy = true;
	}

	return 0;

4229
err_clear_dir:
4230
	cgroup_clear_dir(css->cgroup, 1 << css->ss->id);
4231 4232
err_free_id:
	cgroup_idr_remove(&ss->css_idr, css->id);
4233
err_free_percpu_ref:
4234
	percpu_ref_cancel_init(&css->refcnt);
4235
err_free_css:
4236
	call_rcu(&css->rcu_head, css_free_rcu_fn);
4237 4238 4239
	return err;
}

T
Tejun Heo 已提交
4240
/**
L
Li Zefan 已提交
4241 4242
 * cgroup_create - create a cgroup
 * @parent: cgroup that will be parent of the new cgroup
T
Tejun Heo 已提交
4243
 * @name: name of the new cgroup
T
Tejun Heo 已提交
4244
 * @mode: mode to set on new cgroup
4245
 */
T
Tejun Heo 已提交
4246
static long cgroup_create(struct cgroup *parent, const char *name,
T
Tejun Heo 已提交
4247
			  umode_t mode)
4248
{
4249
	struct cgroup *cgrp;
4250
	struct cgroup_root *root = parent->root;
4251
	int ssid, err;
4252
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
4253
	struct kernfs_node *kn;
4254

T
Tejun Heo 已提交
4255
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4256 4257
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4258 4259
		return -ENOMEM;

T
Tejun Heo 已提交
4260
	mutex_lock(&cgroup_tree_mutex);
4261

4262 4263 4264 4265 4266 4267 4268 4269 4270
	/*
	 * 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
4271
		goto err_unlock_tree;
4272 4273 4274 4275 4276 4277
	}

	/*
	 * Temporarily set the pointer to NULL, so idr_find() won't return
	 * a half-baked cgroup.
	 */
4278
	cgrp->id = cgroup_idr_alloc(&root->cgroup_idr, NULL, 2, 0, GFP_NOWAIT);
4279 4280 4281
	if (cgrp->id < 0) {
		err = -ENOMEM;
		goto err_unlock;
4282 4283
	}

4284
	init_cgroup_housekeeping(cgrp);
4285

4286
	cgrp->parent = parent;
4287
	cgrp->dummy_css.parent = &parent->dummy_css;
4288
	cgrp->root = parent->root;
4289

4290 4291 4292
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4293 4294
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4295

T
Tejun Heo 已提交
4296
	/* create the directory */
T
Tejun Heo 已提交
4297
	kn = kernfs_create_dir(parent->kn, name, mode, cgrp);
T
Tejun Heo 已提交
4298 4299
	if (IS_ERR(kn)) {
		err = PTR_ERR(kn);
4300
		goto err_free_id;
T
Tejun Heo 已提交
4301 4302
	}
	cgrp->kn = kn;
4303

4304
	/*
4305 4306
	 * This extra ref will be put in cgroup_free_fn() and guarantees
	 * that @cgrp->kn is always accessible.
4307
	 */
4308
	kernfs_get(kn);
4309

4310
	cgrp->serial_nr = cgroup_serial_nr_next++;
4311

4312 4313
	/* allocation complete, commit to creation */
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
4314
	atomic_inc(&root->nr_cgrps);
4315
	cgroup_get(parent);
4316

4317 4318 4319 4320
	/*
	 * @cgrp is now fully operational.  If something fails after this
	 * point, it'll be released via the normal destruction path.
	 */
4321
	cgroup_idr_replace(&root->cgroup_idr, cgrp, cgrp->id);
4322

4323 4324 4325 4326
	err = cgroup_kn_set_ugid(kn);
	if (err)
		goto err_destroy;

4327
	err = cgroup_addrm_files(cgrp, cgroup_base_files, true);
4328 4329 4330
	if (err)
		goto err_destroy;

4331
	/* let's create and online css's */
T
Tejun Heo 已提交
4332
	for_each_subsys(ss, ssid) {
4333
		if (parent->child_subsys_mask & (1 << ssid)) {
T
Tejun Heo 已提交
4334 4335 4336 4337
			err = create_css(cgrp, ss);
			if (err)
				goto err_destroy;
		}
4338
	}
4339

4340 4341 4342 4343 4344 4345
	/*
	 * 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;
4346

T
Tejun Heo 已提交
4347 4348
	kernfs_activate(kn);

4349
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4350
	mutex_unlock(&cgroup_tree_mutex);
4351 4352 4353

	return 0;

T
Tejun Heo 已提交
4354
err_free_id:
4355
	cgroup_idr_remove(&root->cgroup_idr, cgrp->id);
4356 4357
err_unlock:
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4358 4359
err_unlock_tree:
	mutex_unlock(&cgroup_tree_mutex);
4360
	kfree(cgrp);
4361
	return err;
4362 4363 4364 4365

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4366
	mutex_unlock(&cgroup_tree_mutex);
4367
	return err;
4368 4369
}

T
Tejun Heo 已提交
4370 4371
static int cgroup_mkdir(struct kernfs_node *parent_kn, const char *name,
			umode_t mode)
4372
{
T
Tejun Heo 已提交
4373
	struct cgroup *parent = parent_kn->priv;
4374
	int ret;
4375

4376 4377 4378 4379 4380 4381 4382 4383
	/*
	 * 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);
4384

4385 4386 4387 4388 4389
	ret = cgroup_create(parent, name, mode);

	kernfs_unbreak_active_protection(parent_kn);
	cgroup_put(parent);
	return ret;
4390 4391
}

4392 4393
/*
 * This is called when the refcnt of a css is confirmed to be killed.
4394
 * css_tryget_online() is now guaranteed to fail.
4395 4396
 */
static void css_killed_work_fn(struct work_struct *work)
4397
{
4398 4399 4400
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, destroy_work);
	struct cgroup *cgrp = css->cgroup;
4401

T
Tejun Heo 已提交
4402
	mutex_lock(&cgroup_tree_mutex);
4403 4404
	mutex_lock(&cgroup_mutex);

4405
	/*
4406 4407
	 * css_tryget_online() is guaranteed to fail now.  Tell subsystems
	 * to initate destruction.
4408 4409 4410
	 */
	offline_css(css);

4411 4412 4413 4414 4415
	/*
	 * 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.
	 */
4416
	if (!cgrp->nr_css && cgroup_is_dead(cgrp))
4417 4418 4419
		cgroup_destroy_css_killed(cgrp);

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4420
	mutex_unlock(&cgroup_tree_mutex);
4421 4422 4423 4424 4425 4426 4427 4428 4429

	/*
	 * 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);
4430 4431
}

4432 4433
/* css kill confirmation processing requires process context, bounce */
static void css_killed_ref_fn(struct percpu_ref *ref)
4434 4435 4436 4437
{
	struct cgroup_subsys_state *css =
		container_of(ref, struct cgroup_subsys_state, refcnt);

4438
	INIT_WORK(&css->destroy_work, css_killed_work_fn);
4439
	queue_work(cgroup_destroy_wq, &css->destroy_work);
4440 4441
}

4442 4443 4444 4445 4446 4447
/**
 * 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
4448 4449
 * asynchronously once css_tryget_online() is guaranteed to fail and when
 * the reference count reaches zero, @css will be released.
4450 4451
 */
static void kill_css(struct cgroup_subsys_state *css)
T
Tejun Heo 已提交
4452
{
4453 4454
	lockdep_assert_held(&cgroup_tree_mutex);

T
Tejun Heo 已提交
4455 4456 4457 4458
	/*
	 * This must happen before css is disassociated with its cgroup.
	 * See seq_css() for details.
	 */
4459
	cgroup_clear_dir(css->cgroup, 1 << css->ss->id);
4460

T
Tejun Heo 已提交
4461 4462 4463 4464 4465 4466 4467 4468 4469
	/*
	 * 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
4470
	 * css_tryget_online().  We can't simply call percpu_ref_kill() and
T
Tejun Heo 已提交
4471 4472 4473 4474 4475 4476 4477
	 * 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);
4478 4479 4480 4481 4482 4483 4484 4485
}

/**
 * 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
4486 4487 4488
 * guarantee that css_tryget_online() won't succeed by the time
 * ->css_offline() is invoked.  To satisfy all the requirements,
 * destruction is implemented in the following two steps.
4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503
 *
 * 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.
 */
4504 4505
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4506
{
4507
	struct cgroup *child;
T
Tejun Heo 已提交
4508
	struct cgroup_subsys_state *css;
4509
	bool empty;
T
Tejun Heo 已提交
4510
	int ssid;
4511

T
Tejun Heo 已提交
4512
	lockdep_assert_held(&cgroup_tree_mutex);
4513 4514
	lockdep_assert_held(&cgroup_mutex);

4515
	/*
4516
	 * css_set_rwsem synchronizes access to ->cset_links and prevents
4517
	 * @cgrp from being removed while put_css_set() is in progress.
4518
	 */
4519
	down_read(&css_set_rwsem);
4520
	empty = list_empty(&cgrp->cset_links);
4521
	up_read(&css_set_rwsem);
4522
	if (!empty)
4523
		return -EBUSY;
L
Li Zefan 已提交
4524

4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540
	/*
	 * 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;

4541 4542 4543
	/*
	 * Mark @cgrp dead.  This prevents further task migration and child
	 * creation by disabling cgroup_lock_live_group().  Note that
4544
	 * CGRP_DEAD assertion is depended upon by css_next_child() to
4545
	 * resume iteration after dropping RCU read lock.  See
4546
	 * css_next_child() for details.
4547
	 */
4548
	set_bit(CGRP_DEAD, &cgrp->flags);
4549

4550
	/*
T
Tejun Heo 已提交
4551 4552
	 * Initiate massacre of all css's.  cgroup_destroy_css_killed()
	 * will be invoked to perform the rest of destruction once the
4553 4554
	 * percpu refs of all css's are confirmed to be killed.  This
	 * involves removing the subsystem's files, drop cgroup_mutex.
4555
	 */
4556
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4557 4558
	for_each_css(css, ssid, cgrp)
		kill_css(css);
4559
	mutex_lock(&cgroup_mutex);
4560 4561 4562 4563 4564 4565 4566 4567

	/* 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);

	/*
4568 4569 4570 4571 4572 4573 4574 4575
	 * 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
4576 4577 4578
	/* remove @cgrp directory along with the base files */
	mutex_unlock(&cgroup_mutex);

4579
	/*
T
Tejun Heo 已提交
4580 4581
	 * There are two control paths which try to determine cgroup from
	 * dentry without going through kernfs - cgroupstats_build() and
4582 4583 4584
	 * css_tryget_online_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.
4585
	 */
4586
	kernfs_remove(cgrp->kn);	/* @cgrp has an extra ref on its kn */
T
Tejun Heo 已提交
4587 4588
	RCU_INIT_POINTER(*(void __rcu __force **)&cgrp->kn->priv, NULL);

4589
	mutex_lock(&cgroup_mutex);
4590

4591 4592 4593
	return 0;
};

4594
/**
4595
 * cgroup_destroy_css_killed - the second step of cgroup destruction
4596
 * @cgrp: the cgroup whose csses have just finished offlining
4597 4598
 *
 * This function is invoked from a work item for a cgroup which is being
4599 4600 4601
 * 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().
4602
 */
4603
static void cgroup_destroy_css_killed(struct cgroup *cgrp)
4604 4605 4606
{
	struct cgroup *parent = cgrp->parent;

T
Tejun Heo 已提交
4607
	lockdep_assert_held(&cgroup_tree_mutex);
4608
	lockdep_assert_held(&cgroup_mutex);
4609

4610
	/* delete this cgroup from parent->children */
4611
	list_del_rcu(&cgrp->sibling);
4612

4613
	cgroup_put(cgrp);
4614

4615
	set_bit(CGRP_RELEASABLE, &parent->flags);
4616
	check_for_release(parent);
4617 4618
}

T
Tejun Heo 已提交
4619
static int cgroup_rmdir(struct kernfs_node *kn)
4620
{
T
Tejun Heo 已提交
4621 4622
	struct cgroup *cgrp = kn->priv;
	int ret = 0;
4623

T
Tejun Heo 已提交
4624 4625 4626 4627 4628 4629 4630 4631
	/*
	 * 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);
4632

T
Tejun Heo 已提交
4633
	mutex_lock(&cgroup_tree_mutex);
4634
	mutex_lock(&cgroup_mutex);
4635 4636

	/*
T
Tejun Heo 已提交
4637 4638
	 * @cgrp might already have been destroyed while we're trying to
	 * grab the mutexes.
4639
	 */
T
Tejun Heo 已提交
4640 4641
	if (!cgroup_is_dead(cgrp))
		ret = cgroup_destroy_locked(cgrp);
4642

4643
	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4644
	mutex_unlock(&cgroup_tree_mutex);
4645

T
Tejun Heo 已提交
4646 4647
	kernfs_unbreak_active_protection(kn);
	cgroup_put(cgrp);
4648
	return ret;
4649 4650
}

T
Tejun Heo 已提交
4651 4652 4653 4654 4655 4656 4657 4658
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,
};

4659
static void __init cgroup_init_subsys(struct cgroup_subsys *ss, bool early)
4660 4661
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4662 4663

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4664

T
Tejun Heo 已提交
4665
	mutex_lock(&cgroup_tree_mutex);
4666 4667
	mutex_lock(&cgroup_mutex);

4668
	idr_init(&ss->css_idr);
T
Tejun Heo 已提交
4669
	INIT_LIST_HEAD(&ss->cfts);
4670

4671 4672 4673
	/* Create the root cgroup state for this subsystem */
	ss->root = &cgrp_dfl_root;
	css = ss->css_alloc(cgroup_css(&cgrp_dfl_root.cgrp, ss));
4674 4675
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
4676
	init_and_link_css(css, ss, &cgrp_dfl_root.cgrp);
4677 4678 4679 4680 4681 4682 4683
	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);
	}
4684

L
Li Zefan 已提交
4685
	/* Update the init_css_set to contain a subsys
4686
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4687
	 * newly registered, all tasks and hence the
4688
	 * init_css_set is in the subsystem's root cgroup. */
4689
	init_css_set.subsys[ss->id] = css;
4690 4691 4692

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

L
Li Zefan 已提交
4693 4694 4695 4696 4697
	/* 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));

4698
	BUG_ON(online_css(css));
4699

4700
	cgrp_dfl_root.subsys_mask |= 1 << ss->id;
B
Ben Blum 已提交
4701 4702

	mutex_unlock(&cgroup_mutex);
T
Tejun Heo 已提交
4703
	mutex_unlock(&cgroup_tree_mutex);
B
Ben Blum 已提交
4704 4705
}

4706
/**
L
Li Zefan 已提交
4707 4708 4709 4710
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4711 4712 4713
 */
int __init cgroup_init_early(void)
{
T
Tejun Heo 已提交
4714 4715
	static struct cgroup_sb_opts __initdata opts =
		{ .flags = CGRP_ROOT_SANE_BEHAVIOR };
4716
	struct cgroup_subsys *ss;
4717
	int i;
4718

4719
	init_cgroup_root(&cgrp_dfl_root, &opts);
4720
	RCU_INIT_POINTER(init_task.cgroups, &init_css_set);
4721

T
Tejun Heo 已提交
4722
	for_each_subsys(ss, i) {
4723
		WARN(!ss->css_alloc || !ss->css_free || ss->name || ss->id,
4724 4725
		     "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,
4726
		     ss->id, ss->name);
4727 4728 4729
		WARN(strlen(cgroup_subsys_name[i]) > MAX_CGROUP_TYPE_NAMELEN,
		     "cgroup_subsys_name %s too long\n", cgroup_subsys_name[i]);

4730
		ss->id = i;
4731
		ss->name = cgroup_subsys_name[i];
4732 4733

		if (ss->early_init)
4734
			cgroup_init_subsys(ss, true);
4735 4736 4737 4738 4739
	}
	return 0;
}

/**
L
Li Zefan 已提交
4740 4741 4742 4743
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4744 4745 4746
 */
int __init cgroup_init(void)
{
4747
	struct cgroup_subsys *ss;
4748
	unsigned long key;
4749
	int ssid, err;
4750

T
Tejun Heo 已提交
4751
	BUG_ON(cgroup_init_cftypes(NULL, cgroup_base_files));
4752

4753
	mutex_lock(&cgroup_tree_mutex);
T
Tejun Heo 已提交
4754 4755
	mutex_lock(&cgroup_mutex);

4756 4757 4758 4759
	/* 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);

4760
	BUG_ON(cgroup_setup_root(&cgrp_dfl_root, 0));
4761

T
Tejun Heo 已提交
4762
	mutex_unlock(&cgroup_mutex);
4763
	mutex_unlock(&cgroup_tree_mutex);
T
Tejun Heo 已提交
4764

4765
	for_each_subsys(ss, ssid) {
4766 4767 4768 4769 4770 4771 4772 4773 4774 4775
		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);
		}
4776

T
Tejun Heo 已提交
4777 4778 4779
		list_add_tail(&init_css_set.e_cset_node[ssid],
			      &cgrp_dfl_root.cgrp.e_csets[ssid]);

4780 4781 4782 4783 4784 4785
		/*
		 * 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));
4786 4787 4788
	}

	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
T
Tejun Heo 已提交
4789 4790
	if (!cgroup_kobj)
		return -ENOMEM;
4791

4792
	err = register_filesystem(&cgroup_fs_type);
4793 4794
	if (err < 0) {
		kobject_put(cgroup_kobj);
T
Tejun Heo 已提交
4795
		return err;
4796
	}
4797

L
Li Zefan 已提交
4798
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
T
Tejun Heo 已提交
4799
	return 0;
4800
}
4801

4802 4803 4804 4805 4806
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.
4807
	 * Use 1 for @max_active.
4808 4809 4810 4811
	 *
	 * We would prefer to do this in cgroup_init() above, but that
	 * is called before init_workqueues(): so leave this until after.
	 */
4812
	cgroup_destroy_wq = alloc_workqueue("cgroup_destroy", 0, 1);
4813
	BUG_ON(!cgroup_destroy_wq);
4814 4815 4816 4817 4818 4819 4820 4821 4822

	/*
	 * 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);

4823 4824 4825 4826
	return 0;
}
core_initcall(cgroup_wq_init);

4827 4828 4829 4830 4831 4832 4833
/*
 * 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 */
4834
int proc_cgroup_show(struct seq_file *m, void *v)
4835 4836 4837
{
	struct pid *pid;
	struct task_struct *tsk;
T
Tejun Heo 已提交
4838
	char *buf, *path;
4839
	int retval;
4840
	struct cgroup_root *root;
4841 4842

	retval = -ENOMEM;
T
Tejun Heo 已提交
4843
	buf = kmalloc(PATH_MAX, GFP_KERNEL);
4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855
	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);
4856
	down_read(&css_set_rwsem);
4857

4858
	for_each_root(root) {
4859
		struct cgroup_subsys *ss;
4860
		struct cgroup *cgrp;
T
Tejun Heo 已提交
4861
		int ssid, count = 0;
4862

T
Tejun Heo 已提交
4863
		if (root == &cgrp_dfl_root && !cgrp_dfl_root_visible)
4864 4865
			continue;

4866
		seq_printf(m, "%d:", root->hierarchy_id);
T
Tejun Heo 已提交
4867
		for_each_subsys(ss, ssid)
4868
			if (root->subsys_mask & (1 << ssid))
T
Tejun Heo 已提交
4869
				seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4870 4871 4872
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4873
		seq_putc(m, ':');
4874
		cgrp = task_cgroup_from_root(tsk, root);
T
Tejun Heo 已提交
4875 4876 4877
		path = cgroup_path(cgrp, buf, PATH_MAX);
		if (!path) {
			retval = -ENAMETOOLONG;
4878
			goto out_unlock;
T
Tejun Heo 已提交
4879 4880
		}
		seq_puts(m, path);
4881 4882 4883 4884
		seq_putc(m, '\n');
	}

out_unlock:
4885
	up_read(&css_set_rwsem);
4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896
	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)
{
4897
	struct cgroup_subsys *ss;
4898 4899
	int i;

4900
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4901 4902 4903 4904 4905
	/*
	 * 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.
	 */
4906
	mutex_lock(&cgroup_mutex);
4907 4908

	for_each_subsys(ss, i)
4909 4910
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
4911
			   atomic_read(&ss->root->nr_cgrps), !ss->disabled);
4912

4913 4914 4915 4916 4917 4918
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
4919
	return single_open(file, proc_cgroupstats_show, NULL);
4920 4921
}

4922
static const struct file_operations proc_cgroupstats_operations = {
4923 4924 4925 4926 4927 4928
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

4929
/**
4930
 * cgroup_fork - initialize cgroup related fields during copy_process()
L
Li Zefan 已提交
4931
 * @child: pointer to task_struct of forking parent process.
4932
 *
4933 4934 4935
 * 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.
4936 4937 4938
 */
void cgroup_fork(struct task_struct *child)
{
4939
	RCU_INIT_POINTER(child->cgroups, &init_css_set);
4940
	INIT_LIST_HEAD(&child->cg_list);
4941 4942
}

4943
/**
L
Li Zefan 已提交
4944 4945 4946
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
4947 4948 4949
 * 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
4950
 * cgroup_task_iter_start() - to guarantee that the new task ends up on its
4951
 * list.
L
Li Zefan 已提交
4952
 */
4953 4954
void cgroup_post_fork(struct task_struct *child)
{
4955
	struct cgroup_subsys *ss;
4956 4957
	int i;

4958
	/*
4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977
	 * 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.
4978
	 */
4979
	if (use_task_css_set_links) {
4980 4981
		struct css_set *cset;

4982
		down_write(&css_set_rwsem);
4983
		cset = task_css_set(current);
4984 4985 4986 4987 4988
		if (list_empty(&child->cg_list)) {
			rcu_assign_pointer(child->cgroups, cset);
			list_add(&child->cg_list, &cset->tasks);
			get_css_set(cset);
		}
4989
		up_write(&css_set_rwsem);
4990
	}
4991 4992 4993 4994 4995 4996 4997

	/*
	 * 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
4998
		for_each_subsys(ss, i)
4999 5000 5001
			if (ss->fork)
				ss->fork(child);
	}
5002
}
5003

5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015
/**
 * 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.
 *
5016 5017 5018 5019 5020
 * 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
5021
 * with migration path - PF_EXITING is visible to migration path.
5022
 */
5023
void cgroup_exit(struct task_struct *tsk)
5024
{
5025
	struct cgroup_subsys *ss;
5026
	struct css_set *cset;
5027
	bool put_cset = false;
5028
	int i;
5029 5030

	/*
5031 5032
	 * 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.
5033 5034
	 */
	if (!list_empty(&tsk->cg_list)) {
5035
		down_write(&css_set_rwsem);
5036
		list_del_init(&tsk->cg_list);
5037
		up_write(&css_set_rwsem);
5038
		put_cset = true;
5039 5040
	}

5041
	/* Reassign the task to the init_css_set. */
5042 5043
	cset = task_css_set(tsk);
	RCU_INIT_POINTER(tsk->cgroups, &init_css_set);
5044

5045
	if (need_forkexit_callback) {
T
Tejun Heo 已提交
5046 5047
		/* see cgroup_post_fork() for details */
		for_each_subsys(ss, i) {
5048
			if (ss->exit) {
5049 5050
				struct cgroup_subsys_state *old_css = cset->subsys[i];
				struct cgroup_subsys_state *css = task_css(tsk, i);
5051

5052
				ss->exit(css, old_css, tsk);
5053 5054 5055 5056
			}
		}
	}

5057 5058
	if (put_cset)
		put_css_set(cset, true);
5059
}
5060

5061
static void check_for_release(struct cgroup *cgrp)
5062
{
5063
	if (cgroup_is_releasable(cgrp) &&
T
Tejun Heo 已提交
5064
	    list_empty(&cgrp->cset_links) && list_empty(&cgrp->children)) {
5065 5066
		/*
		 * Control Group is currently removeable. If it's not
5067
		 * already queued for a userspace notification, queue
5068 5069
		 * it now
		 */
5070
		int need_schedule_work = 0;
5071

5072
		raw_spin_lock(&release_list_lock);
5073
		if (!cgroup_is_dead(cgrp) &&
5074 5075
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
5076 5077
			need_schedule_work = 1;
		}
5078
		raw_spin_unlock(&release_list_lock);
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 5105 5106 5107 5108 5109 5110
		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);
5111
	raw_spin_lock(&release_list_lock);
5112 5113 5114
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
T
Tejun Heo 已提交
5115
		char *pathbuf = NULL, *agentbuf = NULL, *path;
5116
		struct cgroup *cgrp = list_entry(release_list.next,
5117 5118
						    struct cgroup,
						    release_list);
5119
		list_del_init(&cgrp->release_list);
5120
		raw_spin_unlock(&release_list_lock);
T
Tejun Heo 已提交
5121
		pathbuf = kmalloc(PATH_MAX, GFP_KERNEL);
5122 5123
		if (!pathbuf)
			goto continue_free;
T
Tejun Heo 已提交
5124 5125
		path = cgroup_path(cgrp, pathbuf, PATH_MAX);
		if (!path)
5126 5127 5128 5129
			goto continue_free;
		agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
		if (!agentbuf)
			goto continue_free;
5130 5131

		i = 0;
5132
		argv[i++] = agentbuf;
T
Tejun Heo 已提交
5133
		argv[i++] = path;
5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147
		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);
5148 5149 5150
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5151
		raw_spin_lock(&release_list_lock);
5152
	}
5153
	raw_spin_unlock(&release_list_lock);
5154 5155
	mutex_unlock(&cgroup_mutex);
}
5156 5157 5158

static int __init cgroup_disable(char *str)
{
5159
	struct cgroup_subsys *ss;
5160
	char *token;
5161
	int i;
5162 5163 5164 5165

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5166

T
Tejun Heo 已提交
5167
		for_each_subsys(ss, i) {
5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178
			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
5179

5180
/**
5181
 * css_tryget_online_from_dir - get corresponding css from a cgroup dentry
5182 5183
 * @dentry: directory dentry of interest
 * @ss: subsystem of interest
5184
 *
5185 5186 5187
 * 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
5188
 */
5189 5190
struct cgroup_subsys_state *css_tryget_online_from_dir(struct dentry *dentry,
						       struct cgroup_subsys *ss)
S
Stephane Eranian 已提交
5191
{
T
Tejun Heo 已提交
5192 5193
	struct kernfs_node *kn = kernfs_node_from_dentry(dentry);
	struct cgroup_subsys_state *css = NULL;
S
Stephane Eranian 已提交
5194 5195
	struct cgroup *cgrp;

5196
	/* is @dentry a cgroup dir? */
T
Tejun Heo 已提交
5197 5198
	if (dentry->d_sb->s_type != &cgroup_fs_type || !kn ||
	    kernfs_type(kn) != KERNFS_DIR)
S
Stephane Eranian 已提交
5199 5200
		return ERR_PTR(-EBADF);

5201 5202
	rcu_read_lock();

T
Tejun Heo 已提交
5203 5204 5205 5206 5207 5208 5209 5210
	/*
	 * 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);
5211

5212
	if (!css || !css_tryget_online(css))
5213 5214 5215 5216
		css = ERR_PTR(-ENOENT);

	rcu_read_unlock();
	return css;
S
Stephane Eranian 已提交
5217 5218
}

5219 5220 5221 5222 5223 5224 5225 5226 5227 5228
/**
 * 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)
{
5229
	WARN_ON_ONCE(!rcu_read_lock_held());
5230
	return idr_find(&ss->css_idr, id);
S
Stephane Eranian 已提交
5231 5232
}

5233
#ifdef CONFIG_CGROUP_DEBUG
5234 5235
static struct cgroup_subsys_state *
debug_css_alloc(struct cgroup_subsys_state *parent_css)
5236 5237 5238 5239 5240 5241 5242 5243 5244
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

5245
static void debug_css_free(struct cgroup_subsys_state *css)
5246
{
5247
	kfree(css);
5248 5249
}

5250 5251
static u64 debug_taskcount_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5252
{
5253
	return cgroup_task_count(css->cgroup);
5254 5255
}

5256 5257
static u64 current_css_set_read(struct cgroup_subsys_state *css,
				struct cftype *cft)
5258 5259 5260 5261
{
	return (u64)(unsigned long)current->cgroups;
}

5262
static u64 current_css_set_refcount_read(struct cgroup_subsys_state *css,
L
Li Zefan 已提交
5263
					 struct cftype *cft)
5264 5265 5266 5267
{
	u64 count;

	rcu_read_lock();
5268
	count = atomic_read(&task_css_set(current)->refcount);
5269 5270 5271 5272
	rcu_read_unlock();
	return count;
}

5273
static int current_css_set_cg_links_read(struct seq_file *seq, void *v)
5274
{
5275
	struct cgrp_cset_link *link;
5276
	struct css_set *cset;
T
Tejun Heo 已提交
5277 5278 5279 5280 5281
	char *name_buf;

	name_buf = kmalloc(NAME_MAX + 1, GFP_KERNEL);
	if (!name_buf)
		return -ENOMEM;
5282

5283
	down_read(&css_set_rwsem);
5284
	rcu_read_lock();
5285
	cset = rcu_dereference(current->cgroups);
5286
	list_for_each_entry(link, &cset->cgrp_links, cgrp_link) {
5287 5288
		struct cgroup *c = link->cgrp;

T
Tejun Heo 已提交
5289
		cgroup_name(c, name_buf, NAME_MAX + 1);
5290
		seq_printf(seq, "Root %d group %s\n",
T
Tejun Heo 已提交
5291
			   c->root->hierarchy_id, name_buf);
5292 5293
	}
	rcu_read_unlock();
5294
	up_read(&css_set_rwsem);
T
Tejun Heo 已提交
5295
	kfree(name_buf);
5296 5297 5298 5299
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
5300
static int cgroup_css_links_read(struct seq_file *seq, void *v)
5301
{
5302
	struct cgroup_subsys_state *css = seq_css(seq);
5303
	struct cgrp_cset_link *link;
5304

5305
	down_read(&css_set_rwsem);
5306
	list_for_each_entry(link, &css->cgroup->cset_links, cset_link) {
5307
		struct css_set *cset = link->cset;
5308 5309
		struct task_struct *task;
		int count = 0;
T
Tejun Heo 已提交
5310

5311
		seq_printf(seq, "css_set %p\n", cset);
T
Tejun Heo 已提交
5312

5313
		list_for_each_entry(task, &cset->tasks, cg_list) {
T
Tejun Heo 已提交
5314 5315 5316 5317 5318 5319 5320 5321 5322
			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));
5323
		}
T
Tejun Heo 已提交
5324 5325 5326
		continue;
	overflow:
		seq_puts(seq, "  ...\n");
5327
	}
5328
	up_read(&css_set_rwsem);
5329 5330 5331
	return 0;
}

5332
static u64 releasable_read(struct cgroup_subsys_state *css, struct cftype *cft)
5333
{
5334
	return test_bit(CGRP_RELEASABLE, &css->cgroup->flags);
5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352
}

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,
	},

5353 5354
	{
		.name = "current_css_set_cg_links",
5355
		.seq_show = current_css_set_cg_links_read,
5356 5357 5358 5359
	},

	{
		.name = "cgroup_css_links",
5360
		.seq_show = cgroup_css_links_read,
5361 5362
	},

5363 5364 5365 5366 5367
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5368 5369
	{ }	/* terminate */
};
5370

5371
struct cgroup_subsys debug_cgrp_subsys = {
5372 5373
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5374
	.base_cftypes = debug_files,
5375 5376
};
#endif /* CONFIG_CGROUP_DEBUG */