cgroup.c 143.5 KB
Newer Older
1 2 3 4 5 6
/*
 *  Generic process-grouping system.
 *
 *  Based originally on the cpuset system, extracted by Paul Menage
 *  Copyright (C) 2006 Google, Inc
 *
7 8 9 10
 *  Notifications support
 *  Copyright (C) 2009 Nokia Corporation
 *  Author: Kirill A. Shutemov
 *
11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
 *  Copyright notices from the original cpuset code:
 *  --------------------------------------------------
 *  Copyright (C) 2003 BULL SA.
 *  Copyright (C) 2004-2006 Silicon Graphics, Inc.
 *
 *  Portions derived from Patrick Mochel's sysfs code.
 *  sysfs is Copyright (c) 2001-3 Patrick Mochel
 *
 *  2003-10-10 Written by Simon Derr.
 *  2003-10-22 Updates by Stephen Hemminger.
 *  2004 May-July Rework by Paul Jackson.
 *  ---------------------------------------------------
 *
 *  This file is subject to the terms and conditions of the GNU General Public
 *  License.  See the file COPYING in the main directory of the Linux
 *  distribution for more details.
 */

#include <linux/cgroup.h>
30
#include <linux/cred.h>
31
#include <linux/ctype.h>
32 33
#include <linux/errno.h>
#include <linux/fs.h>
34
#include <linux/init_task.h>
35 36 37 38 39 40
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/mm.h>
#include <linux/mutex.h>
#include <linux/mount.h>
#include <linux/pagemap.h>
41
#include <linux/proc_fs.h>
42 43
#include <linux/rcupdate.h>
#include <linux/sched.h>
44
#include <linux/backing-dev.h>
45 46 47 48 49
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/magic.h>
#include <linux/spinlock.h>
#include <linux/string.h>
50
#include <linux/sort.h>
51
#include <linux/kmod.h>
52
#include <linux/module.h>
B
Balbir Singh 已提交
53 54
#include <linux/delayacct.h>
#include <linux/cgroupstats.h>
55
#include <linux/hashtable.h>
56
#include <linux/namei.h>
L
Li Zefan 已提交
57
#include <linux/pid_namespace.h>
58
#include <linux/idr.h>
59
#include <linux/vmalloc.h> /* TODO: replace with more sophisticated array */
60 61
#include <linux/eventfd.h>
#include <linux/poll.h>
62
#include <linux/flex_array.h> /* used in cgroup_attach_task */
63
#include <linux/kthread.h>
B
Balbir Singh 已提交
64

A
Arun Sharma 已提交
65
#include <linux/atomic.h>
66

67 68 69
/* css deactivation bias, makes css->refcnt negative to deny new trygets */
#define CSS_DEACT_BIAS		INT_MIN

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

B
Ben Blum 已提交
89 90
/*
 * Generate an array of cgroup subsystem pointers. At boot time, this is
91
 * populated with the built in subsystems, and modular subsystems are
B
Ben Blum 已提交
92 93 94
 * registered after that. The mutable section of this array is protected by
 * cgroup_mutex.
 */
95
#define SUBSYS(_x) [_x ## _subsys_id] = &_x ## _subsys,
96
#define IS_SUBSYS_ENABLED(option) IS_BUILTIN(option)
B
Ben Blum 已提交
97
static struct cgroup_subsys *subsys[CGROUP_SUBSYS_COUNT] = {
98 99 100
#include <linux/cgroup_subsys.h>
};

101 102
#define MAX_CGROUP_ROOT_NAMELEN 64

103 104 105 106 107 108 109 110 111 112 113 114
/*
 * A cgroupfs_root represents the root of a cgroup hierarchy,
 * and may be associated with a superblock to form an active
 * hierarchy
 */
struct cgroupfs_root {
	struct super_block *sb;

	/*
	 * The bitmask of subsystems intended to be attached to this
	 * hierarchy
	 */
115
	unsigned long subsys_mask;
116

117 118 119
	/* Unique id for this hierarchy. */
	int hierarchy_id;

120
	/* The bitmask of subsystems currently attached to this hierarchy */
121
	unsigned long actual_subsys_mask;
122 123 124 125 126 127 128 129 130 131

	/* A list running through the attached subsystems */
	struct list_head subsys_list;

	/* The root cgroup for this hierarchy */
	struct cgroup top_cgroup;

	/* Tracks how many cgroups are currently defined in hierarchy.*/
	int number_of_cgroups;

132
	/* A list running through the active hierarchies */
133 134
	struct list_head root_list;

135 136 137
	/* All cgroups on this root, cgroup_mutex protected */
	struct list_head allcg_list;

138 139
	/* Hierarchy-specific flags */
	unsigned long flags;
140

T
Tejun Heo 已提交
141 142 143
	/* IDs for cgroups in this hierarchy */
	struct ida cgroup_ida;

144
	/* The path to use for release notifications. */
145
	char release_agent_path[PATH_MAX];
146 147 148

	/* The name for this hierarchy - may be empty */
	char name[MAX_CGROUP_ROOT_NAMELEN];
149 150 151 152 153 154 155 156 157
};

/*
 * The "rootnode" hierarchy is the "dummy hierarchy", reserved for the
 * subsystems that are otherwise unattached - it never has more than a
 * single cgroup, and all tasks are part of that cgroup.
 */
static struct cgroupfs_root rootnode;

T
Tejun Heo 已提交
158 159 160 161 162 163 164 165 166
/*
 * cgroupfs file entry, pointed to from leaf dentry->d_fsdata.
 */
struct cfent {
	struct list_head		node;
	struct dentry			*dentry;
	struct cftype			*type;
};

K
KAMEZAWA Hiroyuki 已提交
167 168 169 170 171 172 173 174 175 176
/*
 * CSS ID -- ID per subsys's Cgroup Subsys State(CSS). used only when
 * cgroup_subsys->use_id != 0.
 */
#define CSS_ID_MAX	(65535)
struct css_id {
	/*
	 * The css to which this ID points. This pointer is set to valid value
	 * after cgroup is populated. If cgroup is removed, this will be NULL.
	 * This pointer is expected to be RCU-safe because destroy()
T
Tejun Heo 已提交
177 178
	 * is called after synchronize_rcu(). But for safe use, css_tryget()
	 * should be used for avoiding race.
K
KAMEZAWA Hiroyuki 已提交
179
	 */
A
Arnd Bergmann 已提交
180
	struct cgroup_subsys_state __rcu *css;
K
KAMEZAWA Hiroyuki 已提交
181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
	/*
	 * ID of this css.
	 */
	unsigned short id;
	/*
	 * Depth in hierarchy which this ID belongs to.
	 */
	unsigned short depth;
	/*
	 * ID is freed by RCU. (and lookup routine is RCU safe.)
	 */
	struct rcu_head rcu_head;
	/*
	 * Hierarchy of CSS ID belongs to.
	 */
	unsigned short stack[0]; /* Array of Length (depth+1) */
};

199
/*
L
Lucas De Marchi 已提交
200
 * cgroup_event represents events which userspace want to receive.
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
 */
struct cgroup_event {
	/*
	 * Cgroup which the event belongs to.
	 */
	struct cgroup *cgrp;
	/*
	 * Control file which the event associated.
	 */
	struct cftype *cft;
	/*
	 * eventfd to signal userspace about the event.
	 */
	struct eventfd_ctx *eventfd;
	/*
	 * Each of these stored in a list by the cgroup.
	 */
	struct list_head list;
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
	wait_queue_t wait;
	struct work_struct remove;
};
K
KAMEZAWA Hiroyuki 已提交
228

229 230 231
/* The list of hierarchy roots */

static LIST_HEAD(roots);
232
static int root_count;
233

234 235 236 237
static DEFINE_IDA(hierarchy_ida);
static int next_hierarchy_id;
static DEFINE_SPINLOCK(hierarchy_id_lock);

238 239 240
/* dummytop is a shorthand for the dummy hierarchy's top cgroup */
#define dummytop (&rootnode.top_cgroup)

241 242
static struct cgroup_name root_cgroup_name = { .name = "/" };

243
/* This flag indicates whether tasks in the fork and exit paths should
L
Li Zefan 已提交
244 245 246
 * 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.
247
 */
248
static int need_forkexit_callback __read_mostly;
249

250
static int cgroup_destroy_locked(struct cgroup *cgrp);
251 252
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
			      struct cftype cfts[], bool is_add);
253

254 255 256 257 258 259 260 261 262 263 264 265 266 267
#ifdef CONFIG_PROVE_LOCKING
int cgroup_lock_is_held(void)
{
	return lockdep_is_held(&cgroup_mutex);
}
#else /* #ifdef CONFIG_PROVE_LOCKING */
int cgroup_lock_is_held(void)
{
	return mutex_is_locked(&cgroup_mutex);
}
#endif /* #else #ifdef CONFIG_PROVE_LOCKING */

EXPORT_SYMBOL_GPL(cgroup_lock_is_held);

268 269 270 271 272
static int css_unbias_refcnt(int refcnt)
{
	return refcnt >= 0 ? refcnt : refcnt - CSS_DEACT_BIAS;
}

273 274 275 276 277
/* the current nr of refs, always >= 0 whether @css is deactivated or not */
static int css_refcnt(struct cgroup_subsys_state *css)
{
	int v = atomic_read(&css->refcnt);

278
	return css_unbias_refcnt(v);
279 280
}

281
/* convenient tests for these bits */
282
inline int cgroup_is_removed(const struct cgroup *cgrp)
283
{
284
	return test_bit(CGRP_REMOVED, &cgrp->flags);
285 286 287 288
}

/* bits in struct cgroupfs_root flags field */
enum {
A
Aristeu Rozanski 已提交
289 290
	ROOT_NOPREFIX,	/* mounted subsystems have no named prefix */
	ROOT_XATTR,	/* supports extended attributes */
291 292
};

293
static int cgroup_is_releasable(const struct cgroup *cgrp)
294 295
{
	const int bits =
296 297 298
		(1 << CGRP_RELEASABLE) |
		(1 << CGRP_NOTIFY_ON_RELEASE);
	return (cgrp->flags & bits) == bits;
299 300
}

301
static int notify_on_release(const struct cgroup *cgrp)
302
{
303
	return test_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
304 305
}

306 307 308 309 310 311 312
/*
 * for_each_subsys() allows you to iterate on each subsystem attached to
 * an active hierarchy
 */
#define for_each_subsys(_root, _ss) \
list_for_each_entry(_ss, &_root->subsys_list, sibling)

313 314
/* for_each_active_root() allows you to iterate across the active hierarchies */
#define for_each_active_root(_root) \
315 316
list_for_each_entry(_root, &roots, root_list)

317 318 319 320 321
static inline struct cgroup *__d_cgrp(struct dentry *dentry)
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
322
static inline struct cfent *__d_cfe(struct dentry *dentry)
323 324 325 326
{
	return dentry->d_fsdata;
}

T
Tejun Heo 已提交
327 328 329 330 331
static inline struct cftype *__d_cft(struct dentry *dentry)
{
	return __d_cfe(dentry)->type;
}

332 333 334
/* the list of cgroups eligible for automatic release. Protected by
 * release_list_lock */
static LIST_HEAD(release_list);
335
static DEFINE_RAW_SPINLOCK(release_list_lock);
336 337
static void cgroup_release_agent(struct work_struct *work);
static DECLARE_WORK(release_agent_work, cgroup_release_agent);
338
static void check_for_release(struct cgroup *cgrp);
339

340 341 342 343 344 345
/* Link structure for associating css_set objects with cgroups */
struct cg_cgroup_link {
	/*
	 * List running through cg_cgroup_links associated with a
	 * cgroup, anchored on cgroup->css_sets
	 */
346
	struct list_head cgrp_link_list;
347
	struct cgroup *cgrp;
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
	/*
	 * List running through cg_cgroup_links pointing at a
	 * single css_set object, anchored on css_set->cg_links
	 */
	struct list_head cg_link_list;
	struct css_set *cg;
};

/* The default css_set - used by init and its children prior to any
 * hierarchies being mounted. It contains a pointer to the root state
 * for each subsystem. Also used to anchor the list of css_sets. Not
 * reference-counted, to improve performance when child cgroups
 * haven't been created.
 */

static struct css_set init_css_set;
static struct cg_cgroup_link init_css_set_link;

366 367
static int cgroup_init_idr(struct cgroup_subsys *ss,
			   struct cgroup_subsys_state *css);
K
KAMEZAWA Hiroyuki 已提交
368

369 370 371 372 373 374
/* css_set_lock protects the list of css_set objects, and the
 * chain of tasks off each css_set.  Nests outside task->alloc_lock
 * due to cgroup_iter_start() */
static DEFINE_RWLOCK(css_set_lock);
static int css_set_count;

375 376 377 378 379
/*
 * 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.
 */
380
#define CSS_SET_HASH_BITS	7
381
static DEFINE_HASHTABLE(css_set_table, CSS_SET_HASH_BITS);
382

383
static unsigned long css_set_hash(struct cgroup_subsys_state *css[])
384 385
{
	int i;
386
	unsigned long key = 0UL;
387 388

	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++)
389 390
		key += (unsigned long)css[i];
	key = (key >> 16) ^ key;
391

392
	return key;
393 394
}

395 396 397 398
/* We don't maintain the lists running through each css_set to its
 * task until after the first call to cgroup_iter_start(). This
 * reduces the fork()/exit() overhead for people who have cgroups
 * compiled into their kernel but not actually in use */
399
static int use_task_css_set_links __read_mostly;
400

401
static void __put_css_set(struct css_set *cg, int taskexit)
402
{
K
KOSAKI Motohiro 已提交
403 404
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;
405 406 407 408 409 410 411 412 413 414 415 416
	/*
	 * 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(&cg->refcount, -1, 1))
		return;
	write_lock(&css_set_lock);
	if (!atomic_dec_and_test(&cg->refcount)) {
		write_unlock(&css_set_lock);
		return;
	}
417

418
	/* This css_set is dead. unlink it and release cgroup refcounts */
419
	hash_del(&cg->hlist);
420 421 422 423 424 425 426
	css_set_count--;

	list_for_each_entry_safe(link, saved_link, &cg->cg_links,
				 cg_link_list) {
		struct cgroup *cgrp = link->cgrp;
		list_del(&link->cg_link_list);
		list_del(&link->cgrp_link_list);
L
Li Zefan 已提交
427 428 429 430 431 432 433

		/*
		 * We may not be holding cgroup_mutex, and if cgrp->count is
		 * dropped to 0 the cgroup can be destroyed at any time, hence
		 * rcu_read_lock is used to keep it alive.
		 */
		rcu_read_lock();
434 435
		if (atomic_dec_and_test(&cgrp->count) &&
		    notify_on_release(cgrp)) {
436
			if (taskexit)
437 438
				set_bit(CGRP_RELEASABLE, &cgrp->flags);
			check_for_release(cgrp);
439
		}
L
Li Zefan 已提交
440
		rcu_read_unlock();
441 442

		kfree(link);
443
	}
444 445

	write_unlock(&css_set_lock);
446
	kfree_rcu(cg, rcu_head);
447 448
}

449 450 451 452 453
/*
 * refcounted get/put for css_set objects
 */
static inline void get_css_set(struct css_set *cg)
{
454
	atomic_inc(&cg->refcount);
455 456 457 458
}

static inline void put_css_set(struct css_set *cg)
{
459
	__put_css_set(cg, 0);
460 461
}

462 463
static inline void put_css_set_taskexit(struct css_set *cg)
{
464
	__put_css_set(cg, 1);
465 466
}

467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538
/*
 * compare_css_sets - helper function for find_existing_css_set().
 * @cg: candidate css_set being tested
 * @old_cg: existing css_set for a task
 * @new_cgrp: cgroup that's being entered by the task
 * @template: desired set of css pointers in css_set (pre-calculated)
 *
 * Returns true if "cg" matches "old_cg" except for the hierarchy
 * which "new_cgrp" belongs to, for which it should match "new_cgrp".
 */
static bool compare_css_sets(struct css_set *cg,
			     struct css_set *old_cg,
			     struct cgroup *new_cgrp,
			     struct cgroup_subsys_state *template[])
{
	struct list_head *l1, *l2;

	if (memcmp(template, cg->subsys, sizeof(cg->subsys))) {
		/* Not all subsystems matched */
		return false;
	}

	/*
	 * Compare cgroup pointers in order to distinguish between
	 * different cgroups in heirarchies with no subsystems. We
	 * could get by with just this check alone (and skip the
	 * memcmp above) but on most setups the memcmp check will
	 * avoid the need for this more expensive check on almost all
	 * candidates.
	 */

	l1 = &cg->cg_links;
	l2 = &old_cg->cg_links;
	while (1) {
		struct cg_cgroup_link *cgl1, *cgl2;
		struct cgroup *cg1, *cg2;

		l1 = l1->next;
		l2 = l2->next;
		/* See if we reached the end - both lists are equal length. */
		if (l1 == &cg->cg_links) {
			BUG_ON(l2 != &old_cg->cg_links);
			break;
		} else {
			BUG_ON(l2 == &old_cg->cg_links);
		}
		/* Locate the cgroups associated with these links. */
		cgl1 = list_entry(l1, struct cg_cgroup_link, cg_link_list);
		cgl2 = list_entry(l2, struct cg_cgroup_link, cg_link_list);
		cg1 = cgl1->cgrp;
		cg2 = cgl2->cgrp;
		/* Hierarchies should be linked in the same order. */
		BUG_ON(cg1->root != cg2->root);

		/*
		 * 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.
		 */
		if (cg1->root == new_cgrp->root) {
			if (cg1 != new_cgrp)
				return false;
		} else {
			if (cg1 != cg2)
				return false;
		}
	}
	return true;
}

539 540 541
/*
 * find_existing_css_set() is a helper for
 * find_css_set(), and checks to see whether an existing
542
 * css_set is suitable.
543 544 545 546
 *
 * oldcg: the cgroup group that we're using before the cgroup
 * transition
 *
547
 * cgrp: the cgroup that we're moving into
548 549 550 551 552 553
 *
 * template: location in which to build the desired set of subsystem
 * state objects for the new cgroup group
 */
static struct css_set *find_existing_css_set(
	struct css_set *oldcg,
554
	struct cgroup *cgrp,
555
	struct cgroup_subsys_state *template[])
556 557
{
	int i;
558
	struct cgroupfs_root *root = cgrp->root;
559
	struct css_set *cg;
560
	unsigned long key;
561

B
Ben Blum 已提交
562 563 564 565 566
	/*
	 * 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.
	 */
567
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
568
		if (root->subsys_mask & (1UL << i)) {
569 570 571
			/* Subsystem is in this hierarchy. So we want
			 * the subsystem state from the new
			 * cgroup */
572
			template[i] = cgrp->subsys[i];
573 574 575 576 577 578 579
		} else {
			/* Subsystem is not in this hierarchy, so we
			 * don't want to change the subsystem state */
			template[i] = oldcg->subsys[i];
		}
	}

580
	key = css_set_hash(template);
581
	hash_for_each_possible(css_set_table, cg, hlist, key) {
582 583 584 585 586
		if (!compare_css_sets(cg, oldcg, cgrp, template))
			continue;

		/* This css_set matches what we need */
		return cg;
587
	}
588 589 590 591 592

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

593 594 595 596 597 598 599 600 601 602 603
static void free_cg_links(struct list_head *tmp)
{
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;

	list_for_each_entry_safe(link, saved_link, tmp, cgrp_link_list) {
		list_del(&link->cgrp_link_list);
		kfree(link);
	}
}

604 605
/*
 * allocate_cg_links() allocates "count" cg_cgroup_link structures
606
 * and chains them on tmp through their cgrp_link_list fields. Returns 0 on
607 608 609 610 611 612 613 614 615 616
 * success or a negative error
 */
static int allocate_cg_links(int count, struct list_head *tmp)
{
	struct cg_cgroup_link *link;
	int i;
	INIT_LIST_HEAD(tmp);
	for (i = 0; i < count; i++) {
		link = kmalloc(sizeof(*link), GFP_KERNEL);
		if (!link) {
617
			free_cg_links(tmp);
618 619
			return -ENOMEM;
		}
620
		list_add(&link->cgrp_link_list, tmp);
621 622 623 624
	}
	return 0;
}

625 626 627 628 629 630 631 632 633 634 635 636 637 638 639
/**
 * link_css_set - a helper function to link a css_set to a cgroup
 * @tmp_cg_links: cg_cgroup_link objects allocated by allocate_cg_links()
 * @cg: the css_set to be linked
 * @cgrp: the destination cgroup
 */
static void link_css_set(struct list_head *tmp_cg_links,
			 struct css_set *cg, struct cgroup *cgrp)
{
	struct cg_cgroup_link *link;

	BUG_ON(list_empty(tmp_cg_links));
	link = list_first_entry(tmp_cg_links, struct cg_cgroup_link,
				cgrp_link_list);
	link->cg = cg;
640
	link->cgrp = cgrp;
641
	atomic_inc(&cgrp->count);
642
	list_move(&link->cgrp_link_list, &cgrp->css_sets);
643 644 645 646 647
	/*
	 * Always add links to the tail of the list so that the list
	 * is sorted by order of hierarchy creation
	 */
	list_add_tail(&link->cg_link_list, &cg->cg_links);
648 649
}

650 651 652 653 654 655 656 657
/*
 * find_css_set() takes an existing cgroup group and a
 * cgroup object, and returns a css_set object that's
 * equivalent to the old group, but with the given cgroup
 * substituted into the appropriate hierarchy. Must be called with
 * cgroup_mutex held
 */
static struct css_set *find_css_set(
658
	struct css_set *oldcg, struct cgroup *cgrp)
659 660 661 662 663 664
{
	struct css_set *res;
	struct cgroup_subsys_state *template[CGROUP_SUBSYS_COUNT];

	struct list_head tmp_cg_links;

665
	struct cg_cgroup_link *link;
666
	unsigned long key;
667

668 669
	/* First see if we already have a cgroup group that matches
	 * the desired set */
670
	read_lock(&css_set_lock);
671
	res = find_existing_css_set(oldcg, cgrp, template);
672 673
	if (res)
		get_css_set(res);
674
	read_unlock(&css_set_lock);
675 676 677 678 679 680 681 682 683 684 685 686 687 688

	if (res)
		return res;

	res = kmalloc(sizeof(*res), GFP_KERNEL);
	if (!res)
		return NULL;

	/* Allocate all the cg_cgroup_link objects that we'll need */
	if (allocate_cg_links(root_count, &tmp_cg_links) < 0) {
		kfree(res);
		return NULL;
	}

689
	atomic_set(&res->refcount, 1);
690 691
	INIT_LIST_HEAD(&res->cg_links);
	INIT_LIST_HEAD(&res->tasks);
692
	INIT_HLIST_NODE(&res->hlist);
693 694 695 696 697 698 699

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

	write_lock(&css_set_lock);
	/* Add reference counts and links from the new css_set. */
700 701 702 703 704 705
	list_for_each_entry(link, &oldcg->cg_links, cg_link_list) {
		struct cgroup *c = link->cgrp;
		if (c->root == cgrp->root)
			c = cgrp;
		link_css_set(&tmp_cg_links, res, c);
	}
706 707 708 709

	BUG_ON(!list_empty(&tmp_cg_links));

	css_set_count++;
710 711

	/* Add this cgroup group to the hash table */
712 713
	key = css_set_hash(res->subsys);
	hash_add(css_set_table, &res->hlist, key);
714

715 716 717
	write_unlock(&css_set_lock);

	return res;
718 719
}

720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754
/*
 * Return the cgroup for "task" from the given hierarchy. Must be
 * called with cgroup_mutex held.
 */
static struct cgroup *task_cgroup_from_root(struct task_struct *task,
					    struct cgroupfs_root *root)
{
	struct css_set *css;
	struct cgroup *res = NULL;

	BUG_ON(!mutex_is_locked(&cgroup_mutex));
	read_lock(&css_set_lock);
	/*
	 * No need to lock the task - since we hold cgroup_mutex the
	 * task can't change groups, so the only thing that can happen
	 * is that it exits and its css is set back to init_css_set.
	 */
	css = task->cgroups;
	if (css == &init_css_set) {
		res = &root->top_cgroup;
	} else {
		struct cg_cgroup_link *link;
		list_for_each_entry(link, &css->cg_links, cg_link_list) {
			struct cgroup *c = link->cgrp;
			if (c->root == root) {
				res = c;
				break;
			}
		}
	}
	read_unlock(&css_set_lock);
	BUG_ON(!res);
	return res;
}

755 756 757 758 759 760 761 762 763 764
/*
 * There is one global cgroup mutex. We also require taking
 * task_lock() when dereferencing a task's cgroup subsys pointers.
 * See "The task_lock() exception", at the end of this comment.
 *
 * A task must hold cgroup_mutex to modify cgroups.
 *
 * Any task can increment and decrement the count field without lock.
 * So in general, code holding cgroup_mutex can't rely on the count
 * field not changing.  However, if the count goes to zero, then only
765
 * cgroup_attach_task() can increment it again.  Because a count of zero
766 767 768 769 770 771 772 773 774 775 776 777 778
 * 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
779 780
 * to the release agent with the name of the cgroup (path relative to
 * the root of cgroup file system) as the argument.
781 782 783 784 785 786 787 788 789 790 791
 *
 * A cgroup can only be deleted if both its 'count' of using tasks
 * is zero, and its list of 'children' cgroups is empty.  Since all
 * tasks in the system use _some_ cgroup, and since there is always at
 * least one task in the system (init, pid == 1), therefore, top_cgroup
 * always has either children cgroups and/or using tasks.  So we don't
 * need a special hack to ensure that top_cgroup cannot be deleted.
 *
 *	The task_lock() exception
 *
 * The need for this exception arises from the action of
792
 * cgroup_attach_task(), which overwrites one task's cgroup pointer with
L
Li Zefan 已提交
793
 * another.  It does so using cgroup_mutex, however there are
794 795 796
 * several performance critical places that need to reference
 * task->cgroup without the expense of grabbing a system global
 * mutex.  Therefore except as noted below, when dereferencing or, as
797
 * in cgroup_attach_task(), modifying a task's cgroup pointer we use
798 799 800 801
 * task_lock(), which acts on a spinlock (task->alloc_lock) already in
 * the task_struct routinely used for such matters.
 *
 * P.S.  One more locking exception.  RCU is used to guard the
802
 * update of a tasks cgroup pointer by cgroup_attach_task()
803 804 805 806 807 808 809 810 811 812
 */

/**
 * cgroup_lock - lock out any changes to cgroup structures
 *
 */
void cgroup_lock(void)
{
	mutex_lock(&cgroup_mutex);
}
B
Ben Blum 已提交
813
EXPORT_SYMBOL_GPL(cgroup_lock);
814 815 816 817 818 819 820 821 822 823

/**
 * cgroup_unlock - release lock on cgroup changes
 *
 * Undo the lock taken in a previous cgroup_lock() call.
 */
void cgroup_unlock(void)
{
	mutex_unlock(&cgroup_mutex);
}
B
Ben Blum 已提交
824
EXPORT_SYMBOL_GPL(cgroup_unlock);
825 826 827 828 829 830 831 832

/*
 * A couple of forward declarations required, due to cyclic reference loop:
 * cgroup_mkdir -> cgroup_create -> cgroup_populate_dir ->
 * cgroup_add_file -> cgroup_create_file -> cgroup_dir_inode_operations
 * -> cgroup_mkdir.
 */

833
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode);
A
Al Viro 已提交
834
static struct dentry *cgroup_lookup(struct inode *, struct dentry *, unsigned int);
835
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry);
836 837
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       unsigned long subsys_mask);
838
static const struct inode_operations cgroup_dir_inode_operations;
839
static const struct file_operations proc_cgroupstats_operations;
840 841

static struct backing_dev_info cgroup_backing_dev_info = {
842
	.name		= "cgroup",
843
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
844
};
845

K
KAMEZAWA Hiroyuki 已提交
846 847 848
static int alloc_css_id(struct cgroup_subsys *ss,
			struct cgroup *parent, struct cgroup *child);

A
Al Viro 已提交
849
static struct inode *cgroup_new_inode(umode_t mode, struct super_block *sb)
850 851 852 853
{
	struct inode *inode = new_inode(sb);

	if (inode) {
854
		inode->i_ino = get_next_ino();
855
		inode->i_mode = mode;
856 857
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
858 859 860 861 862 863
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inode->i_mapping->backing_dev_info = &cgroup_backing_dev_info;
	}
	return inode;
}

864 865 866 867 868 869 870 871 872 873 874
static struct cgroup_name *cgroup_alloc_name(struct dentry *dentry)
{
	struct cgroup_name *name;

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

875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904
static void cgroup_free_fn(struct work_struct *work)
{
	struct cgroup *cgrp = container_of(work, struct cgroup, free_work);
	struct cgroup_subsys *ss;

	mutex_lock(&cgroup_mutex);
	/*
	 * Release the subsystem state objects.
	 */
	for_each_subsys(cgrp->root, ss)
		ss->css_free(cgrp);

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

	/*
	 * Drop the active superblock reference that we took when we
	 * created the cgroup
	 */
	deactivate_super(cgrp->root->sb);

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

	simple_xattrs_free(&cgrp->xattrs);

	ida_simple_remove(&cgrp->root->cgroup_ida, cgrp->id);
905
	kfree(rcu_dereference_raw(cgrp->name));
906 907 908 909 910 911 912 913 914 915
	kfree(cgrp);
}

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

	schedule_work(&cgrp->free_work);
}

916 917 918 919
static void cgroup_diput(struct dentry *dentry, struct inode *inode)
{
	/* is dentry a directory ? if so, kfree() associated cgroup */
	if (S_ISDIR(inode->i_mode)) {
920
		struct cgroup *cgrp = dentry->d_fsdata;
921

922
		BUG_ON(!(cgroup_is_removed(cgrp)));
923
		call_rcu(&cgrp->rcu_head, cgroup_free_rcu);
T
Tejun Heo 已提交
924 925 926
	} else {
		struct cfent *cfe = __d_cfe(dentry);
		struct cgroup *cgrp = dentry->d_parent->d_fsdata;
A
Aristeu Rozanski 已提交
927
		struct cftype *cft = cfe->type;
T
Tejun Heo 已提交
928 929 930 931 932

		WARN_ONCE(!list_empty(&cfe->node) &&
			  cgrp != &cgrp->root->top_cgroup,
			  "cfe still linked for %s\n", cfe->type->name);
		kfree(cfe);
A
Aristeu Rozanski 已提交
933
		simple_xattrs_free(&cft->xattrs);
934 935 936 937
	}
	iput(inode);
}

938 939 940 941 942
static int cgroup_delete(const struct dentry *d)
{
	return 1;
}

943 944 945 946 947 948 949 950 951
static void remove_dir(struct dentry *d)
{
	struct dentry *parent = dget(d->d_parent);

	d_delete(d);
	simple_rmdir(parent->d_inode, d);
	dput(parent);
}

952
static void cgroup_rm_file(struct cgroup *cgrp, const struct cftype *cft)
T
Tejun Heo 已提交
953 954 955 956 957 958
{
	struct cfent *cfe;

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

959 960 961 962
	/*
	 * If we're doing cleanup due to failure of cgroup_create(),
	 * the corresponding @cfe may not exist.
	 */
T
Tejun Heo 已提交
963 964 965 966 967 968 969 970
	list_for_each_entry(cfe, &cgrp->files, node) {
		struct dentry *d = cfe->dentry;

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

		dget(d);
		d_delete(d);
971
		simple_unlink(cgrp->dentry->d_inode, d);
T
Tejun Heo 已提交
972 973 974
		list_del_init(&cfe->node);
		dput(d);

975
		break;
976
	}
T
Tejun Heo 已提交
977 978
}

979 980 981 982 983 984 985 986
/**
 * cgroup_clear_directory - selective removal of base and subsystem files
 * @dir: directory containing the files
 * @base_files: true if the base files should be removed
 * @subsys_mask: mask of the subsystem ids whose files should be removed
 */
static void cgroup_clear_directory(struct dentry *dir, bool base_files,
				   unsigned long subsys_mask)
T
Tejun Heo 已提交
987 988
{
	struct cgroup *cgrp = __d_cgrp(dir);
989
	struct cgroup_subsys *ss;
T
Tejun Heo 已提交
990

991 992 993 994 995
	for_each_subsys(cgrp->root, ss) {
		struct cftype_set *set;
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
		list_for_each_entry(set, &ss->cftsets, node)
996
			cgroup_addrm_files(cgrp, NULL, set->cfts, false);
997 998 999 1000 1001
	}
	if (base_files) {
		while (!list_empty(&cgrp->files))
			cgroup_rm_file(cgrp, NULL);
	}
1002 1003 1004 1005 1006 1007 1008
}

/*
 * NOTE : the dentry must have been dget()'ed
 */
static void cgroup_d_remove_dir(struct dentry *dentry)
{
N
Nick Piggin 已提交
1009
	struct dentry *parent;
1010
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
N
Nick Piggin 已提交
1011

1012
	cgroup_clear_directory(dentry, true, root->subsys_mask);
1013

N
Nick Piggin 已提交
1014 1015
	parent = dentry->d_parent;
	spin_lock(&parent->d_lock);
1016
	spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
1017
	list_del_init(&dentry->d_u.d_child);
N
Nick Piggin 已提交
1018 1019
	spin_unlock(&dentry->d_lock);
	spin_unlock(&parent->d_lock);
1020 1021 1022
	remove_dir(dentry);
}

B
Ben Blum 已提交
1023
/*
B
Ben Blum 已提交
1024 1025 1026
 * Call with cgroup_mutex held. Drops reference counts on modules, including
 * any duplicate ones that parse_cgroupfs_options took. If this function
 * returns an error, no reference counts are touched.
B
Ben Blum 已提交
1027
 */
1028
static int rebind_subsystems(struct cgroupfs_root *root,
1029
			      unsigned long final_subsys_mask)
1030
{
1031
	unsigned long added_mask, removed_mask;
1032
	struct cgroup *cgrp = &root->top_cgroup;
1033 1034
	int i;

B
Ben Blum 已提交
1035
	BUG_ON(!mutex_is_locked(&cgroup_mutex));
T
Tejun Heo 已提交
1036
	BUG_ON(!mutex_is_locked(&cgroup_root_mutex));
B
Ben Blum 已提交
1037

1038 1039
	removed_mask = root->actual_subsys_mask & ~final_subsys_mask;
	added_mask = final_subsys_mask & ~root->actual_subsys_mask;
1040 1041
	/* Check that any added subsystems are currently free */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
L
Li Zefan 已提交
1042
		unsigned long bit = 1UL << i;
1043
		struct cgroup_subsys *ss = subsys[i];
1044
		if (!(bit & added_mask))
1045
			continue;
B
Ben Blum 已提交
1046 1047 1048 1049 1050 1051
		/*
		 * Nobody should tell us to do a subsys that doesn't exist:
		 * parse_cgroupfs_options should catch that case and refcounts
		 * ensure that subsystems won't disappear once selected.
		 */
		BUG_ON(ss == NULL);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
		if (ss->root != &rootnode) {
			/* Subsystem isn't free */
			return -EBUSY;
		}
	}

	/* Currently we don't handle adding/removing subsystems when
	 * any child cgroups exist. This is theoretically supportable
	 * but involves complex error handling, so it's being left until
	 * later */
1062
	if (root->number_of_cgroups > 1)
1063 1064 1065 1066 1067 1068
		return -EBUSY;

	/* Process each subsystem */
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
		unsigned long bit = 1UL << i;
1069
		if (bit & added_mask) {
1070
			/* We're binding this subsystem to this hierarchy */
B
Ben Blum 已提交
1071
			BUG_ON(ss == NULL);
1072
			BUG_ON(cgrp->subsys[i]);
1073 1074
			BUG_ON(!dummytop->subsys[i]);
			BUG_ON(dummytop->subsys[i]->cgroup != dummytop);
1075 1076
			cgrp->subsys[i] = dummytop->subsys[i];
			cgrp->subsys[i]->cgroup = cgrp;
1077
			list_move(&ss->sibling, &root->subsys_list);
1078
			ss->root = root;
1079
			if (ss->bind)
1080
				ss->bind(cgrp);
B
Ben Blum 已提交
1081
			/* refcount was already taken, and we're keeping it */
1082
		} else if (bit & removed_mask) {
1083
			/* We're removing this subsystem */
B
Ben Blum 已提交
1084
			BUG_ON(ss == NULL);
1085 1086
			BUG_ON(cgrp->subsys[i] != dummytop->subsys[i]);
			BUG_ON(cgrp->subsys[i]->cgroup != cgrp);
1087
			if (ss->bind)
1088
				ss->bind(dummytop);
1089
			dummytop->subsys[i]->cgroup = dummytop;
1090
			cgrp->subsys[i] = NULL;
1091
			subsys[i]->root = &rootnode;
1092
			list_move(&ss->sibling, &rootnode.subsys_list);
B
Ben Blum 已提交
1093 1094
			/* subsystem is now free - drop reference on module */
			module_put(ss->module);
1095
		} else if (bit & final_subsys_mask) {
1096
			/* Subsystem state should already exist */
B
Ben Blum 已提交
1097
			BUG_ON(ss == NULL);
1098
			BUG_ON(!cgrp->subsys[i]);
B
Ben Blum 已提交
1099 1100 1101 1102 1103 1104 1105 1106
			/*
			 * a refcount was taken, but we already had one, so
			 * drop the extra reference.
			 */
			module_put(ss->module);
#ifdef CONFIG_MODULE_UNLOAD
			BUG_ON(ss->module && !module_refcount(ss->module));
#endif
1107 1108
		} else {
			/* Subsystem state shouldn't exist */
1109
			BUG_ON(cgrp->subsys[i]);
1110 1111
		}
	}
1112
	root->subsys_mask = root->actual_subsys_mask = final_subsys_mask;
1113 1114 1115 1116

	return 0;
}

1117
static int cgroup_show_options(struct seq_file *seq, struct dentry *dentry)
1118
{
1119
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
1120 1121
	struct cgroup_subsys *ss;

T
Tejun Heo 已提交
1122
	mutex_lock(&cgroup_root_mutex);
1123 1124 1125 1126
	for_each_subsys(root, ss)
		seq_printf(seq, ",%s", ss->name);
	if (test_bit(ROOT_NOPREFIX, &root->flags))
		seq_puts(seq, ",noprefix");
A
Aristeu Rozanski 已提交
1127 1128
	if (test_bit(ROOT_XATTR, &root->flags))
		seq_puts(seq, ",xattr");
1129 1130
	if (strlen(root->release_agent_path))
		seq_printf(seq, ",release_agent=%s", root->release_agent_path);
1131
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags))
1132
		seq_puts(seq, ",clone_children");
1133 1134
	if (strlen(root->name))
		seq_printf(seq, ",name=%s", root->name);
T
Tejun Heo 已提交
1135
	mutex_unlock(&cgroup_root_mutex);
1136 1137 1138 1139
	return 0;
}

struct cgroup_sb_opts {
1140
	unsigned long subsys_mask;
1141
	unsigned long flags;
1142
	char *release_agent;
1143
	bool cpuset_clone_children;
1144
	char *name;
1145 1146
	/* User explicitly requested empty subsystem */
	bool none;
1147 1148

	struct cgroupfs_root *new_root;
1149

1150 1151
};

B
Ben Blum 已提交
1152 1153
/*
 * Convert a hierarchy specifier into a bitmask of subsystems and flags. Call
B
Ben Blum 已提交
1154 1155 1156
 * with cgroup_mutex held to protect the subsys[] array. This function takes
 * refcounts on subsystems to be used, unless it returns error, in which case
 * no refcounts are taken.
B
Ben Blum 已提交
1157
 */
B
Ben Blum 已提交
1158
static int parse_cgroupfs_options(char *data, struct cgroup_sb_opts *opts)
1159
{
1160 1161
	char *token, *o = data;
	bool all_ss = false, one_ss = false;
1162
	unsigned long mask = (unsigned long)-1;
B
Ben Blum 已提交
1163 1164
	int i;
	bool module_pin_failed = false;
1165

B
Ben Blum 已提交
1166 1167
	BUG_ON(!mutex_is_locked(&cgroup_mutex));

1168 1169 1170
#ifdef CONFIG_CPUSETS
	mask = ~(1UL << cpuset_subsys_id);
#endif
1171

1172
	memset(opts, 0, sizeof(*opts));
1173 1174 1175 1176

	while ((token = strsep(&o, ",")) != NULL) {
		if (!*token)
			return -EINVAL;
1177
		if (!strcmp(token, "none")) {
1178 1179
			/* Explicitly have no subsystems */
			opts->none = true;
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189
			continue;
		}
		if (!strcmp(token, "all")) {
			/* Mutually exclusive option 'all' + subsystem name */
			if (one_ss)
				return -EINVAL;
			all_ss = true;
			continue;
		}
		if (!strcmp(token, "noprefix")) {
1190
			set_bit(ROOT_NOPREFIX, &opts->flags);
1191 1192 1193
			continue;
		}
		if (!strcmp(token, "clone_children")) {
1194
			opts->cpuset_clone_children = true;
1195 1196
			continue;
		}
A
Aristeu Rozanski 已提交
1197 1198 1199 1200
		if (!strcmp(token, "xattr")) {
			set_bit(ROOT_XATTR, &opts->flags);
			continue;
		}
1201
		if (!strncmp(token, "release_agent=", 14)) {
1202 1203 1204
			/* Specifying two release agents is forbidden */
			if (opts->release_agent)
				return -EINVAL;
1205
			opts->release_agent =
1206
				kstrndup(token + 14, PATH_MAX - 1, GFP_KERNEL);
1207 1208
			if (!opts->release_agent)
				return -ENOMEM;
1209 1210 1211
			continue;
		}
		if (!strncmp(token, "name=", 5)) {
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
			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,
1229
					      MAX_CGROUP_ROOT_NAMELEN - 1,
1230 1231 1232
					      GFP_KERNEL);
			if (!opts->name)
				return -ENOMEM;
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248

			continue;
		}

		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
			struct cgroup_subsys *ss = subsys[i];
			if (ss == NULL)
				continue;
			if (strcmp(token, ss->name))
				continue;
			if (ss->disabled)
				continue;

			/* Mutually exclusive option 'all' + subsystem name */
			if (all_ss)
				return -EINVAL;
1249
			set_bit(i, &opts->subsys_mask);
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
			one_ss = true;

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

	/*
	 * If the 'all' option was specified select all the subsystems,
1260 1261
	 * otherwise if 'none', 'name=' and a subsystem name options
	 * were not specified, let's default to 'all'
1262
	 */
1263
	if (all_ss || (!one_ss && !opts->none && !opts->name)) {
1264 1265 1266 1267 1268 1269
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
			struct cgroup_subsys *ss = subsys[i];
			if (ss == NULL)
				continue;
			if (ss->disabled)
				continue;
1270
			set_bit(i, &opts->subsys_mask);
1271 1272 1273
		}
	}

1274 1275
	/* Consistency checks */

1276 1277 1278 1279 1280 1281
	/*
	 * Option noprefix was introduced just for backward compatibility
	 * with the old cpuset, so we allow noprefix only if mounting just
	 * the cpuset subsystem.
	 */
	if (test_bit(ROOT_NOPREFIX, &opts->flags) &&
1282
	    (opts->subsys_mask & mask))
1283 1284
		return -EINVAL;

1285 1286

	/* Can't specify "none" and some subsystems */
1287
	if (opts->subsys_mask && opts->none)
1288 1289 1290 1291 1292 1293
		return -EINVAL;

	/*
	 * We either have to specify by name or by subsystems. (So all
	 * empty hierarchies must have a name).
	 */
1294
	if (!opts->subsys_mask && !opts->name)
1295 1296
		return -EINVAL;

B
Ben Blum 已提交
1297 1298 1299 1300 1301 1302
	/*
	 * Grab references on all the modules we'll need, so the subsystems
	 * don't dance around before rebind_subsystems attaches them. This may
	 * take duplicate reference counts on a subsystem that's already used,
	 * but rebind_subsystems handles this case.
	 */
1303
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
B
Ben Blum 已提交
1304 1305
		unsigned long bit = 1UL << i;

1306
		if (!(bit & opts->subsys_mask))
B
Ben Blum 已提交
1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
			continue;
		if (!try_module_get(subsys[i]->module)) {
			module_pin_failed = true;
			break;
		}
	}
	if (module_pin_failed) {
		/*
		 * oops, one of the modules was going away. this means that we
		 * raced with a module_delete call, and to the user this is
		 * essentially a "subsystem doesn't exist" case.
		 */
1319
		for (i--; i >= 0; i--) {
B
Ben Blum 已提交
1320 1321 1322
			/* drop refcounts only on the ones we took */
			unsigned long bit = 1UL << i;

1323
			if (!(bit & opts->subsys_mask))
B
Ben Blum 已提交
1324 1325 1326 1327 1328 1329
				continue;
			module_put(subsys[i]->module);
		}
		return -ENOENT;
	}

1330 1331 1332
	return 0;
}

1333
static void drop_parsed_module_refcounts(unsigned long subsys_mask)
B
Ben Blum 已提交
1334 1335
{
	int i;
1336
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
B
Ben Blum 已提交
1337 1338
		unsigned long bit = 1UL << i;

1339
		if (!(bit & subsys_mask))
B
Ben Blum 已提交
1340 1341 1342 1343 1344
			continue;
		module_put(subsys[i]->module);
	}
}

1345 1346 1347 1348
static int cgroup_remount(struct super_block *sb, int *flags, char *data)
{
	int ret = 0;
	struct cgroupfs_root *root = sb->s_fs_info;
1349
	struct cgroup *cgrp = &root->top_cgroup;
1350
	struct cgroup_sb_opts opts;
1351
	unsigned long added_mask, removed_mask;
1352

1353
	mutex_lock(&cgrp->dentry->d_inode->i_mutex);
1354
	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1355
	mutex_lock(&cgroup_root_mutex);
1356 1357 1358 1359 1360 1361

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

1362
	if (opts.subsys_mask != root->actual_subsys_mask || opts.release_agent)
1363 1364 1365
		pr_warning("cgroup: option changes via remount are deprecated (pid=%d comm=%s)\n",
			   task_tgid_nr(current), current->comm);

1366 1367
	added_mask = opts.subsys_mask & ~root->subsys_mask;
	removed_mask = root->subsys_mask & ~opts.subsys_mask;
1368

B
Ben Blum 已提交
1369 1370 1371
	/* Don't allow flags or name to change at remount */
	if (opts.flags != root->flags ||
	    (opts.name && strcmp(opts.name, root->name))) {
1372
		ret = -EINVAL;
1373
		drop_parsed_module_refcounts(opts.subsys_mask);
1374 1375 1376
		goto out_unlock;
	}

1377 1378 1379 1380 1381 1382 1383
	/*
	 * Clear out the files of subsystems that should be removed, do
	 * this before rebind_subsystems, since rebind_subsystems may
	 * change this hierarchy's subsys_list.
	 */
	cgroup_clear_directory(cgrp->dentry, false, removed_mask);

1384
	ret = rebind_subsystems(root, opts.subsys_mask);
B
Ben Blum 已提交
1385
	if (ret) {
1386 1387
		/* rebind_subsystems failed, re-populate the removed files */
		cgroup_populate_dir(cgrp, false, removed_mask);
1388
		drop_parsed_module_refcounts(opts.subsys_mask);
1389
		goto out_unlock;
B
Ben Blum 已提交
1390
	}
1391

1392
	/* re-populate subsystem files */
1393
	cgroup_populate_dir(cgrp, false, added_mask);
1394

1395 1396
	if (opts.release_agent)
		strcpy(root->release_agent_path, opts.release_agent);
1397
 out_unlock:
1398
	kfree(opts.release_agent);
1399
	kfree(opts.name);
T
Tejun Heo 已提交
1400
	mutex_unlock(&cgroup_root_mutex);
1401
	mutex_unlock(&cgroup_mutex);
1402
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
1403 1404 1405
	return ret;
}

1406
static const struct super_operations cgroup_ops = {
1407 1408 1409 1410 1411 1412
	.statfs = simple_statfs,
	.drop_inode = generic_delete_inode,
	.show_options = cgroup_show_options,
	.remount_fs = cgroup_remount,
};

1413 1414 1415 1416
static void init_cgroup_housekeeping(struct cgroup *cgrp)
{
	INIT_LIST_HEAD(&cgrp->sibling);
	INIT_LIST_HEAD(&cgrp->children);
T
Tejun Heo 已提交
1417
	INIT_LIST_HEAD(&cgrp->files);
1418
	INIT_LIST_HEAD(&cgrp->css_sets);
1419
	INIT_LIST_HEAD(&cgrp->allcg_node);
1420
	INIT_LIST_HEAD(&cgrp->release_list);
1421
	INIT_LIST_HEAD(&cgrp->pidlists);
1422
	INIT_WORK(&cgrp->free_work, cgroup_free_fn);
1423
	mutex_init(&cgrp->pidlist_mutex);
1424 1425
	INIT_LIST_HEAD(&cgrp->event_list);
	spin_lock_init(&cgrp->event_list_lock);
A
Aristeu Rozanski 已提交
1426
	simple_xattrs_init(&cgrp->xattrs);
1427
}
1428

1429 1430
static void init_cgroup_root(struct cgroupfs_root *root)
{
1431
	struct cgroup *cgrp = &root->top_cgroup;
1432

1433 1434
	INIT_LIST_HEAD(&root->subsys_list);
	INIT_LIST_HEAD(&root->root_list);
1435
	INIT_LIST_HEAD(&root->allcg_list);
1436
	root->number_of_cgroups = 1;
1437
	cgrp->root = root;
1438
	cgrp->name = &root_cgroup_name;
1439
	cgrp->top_cgroup = cgrp;
1440
	init_cgroup_housekeeping(cgrp);
1441
	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
1442 1443
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
static bool init_root_id(struct cgroupfs_root *root)
{
	int ret = 0;

	do {
		if (!ida_pre_get(&hierarchy_ida, GFP_KERNEL))
			return false;
		spin_lock(&hierarchy_id_lock);
		/* Try to allocate the next unused ID */
		ret = ida_get_new_above(&hierarchy_ida, next_hierarchy_id,
					&root->hierarchy_id);
		if (ret == -ENOSPC)
			/* Try again starting from 0 */
			ret = ida_get_new(&hierarchy_ida, &root->hierarchy_id);
		if (!ret) {
			next_hierarchy_id = root->hierarchy_id + 1;
		} else if (ret != -EAGAIN) {
			/* Can only get here if the 31-bit IDR is full ... */
			BUG_ON(ret);
		}
		spin_unlock(&hierarchy_id_lock);
	} while (ret);
	return true;
}

1469 1470
static int cgroup_test_super(struct super_block *sb, void *data)
{
1471
	struct cgroup_sb_opts *opts = data;
1472 1473
	struct cgroupfs_root *root = sb->s_fs_info;

1474 1475 1476
	/* If we asked for a name then it must match */
	if (opts->name && strcmp(opts->name, root->name))
		return 0;
1477

1478 1479 1480 1481
	/*
	 * If we asked for subsystems (or explicitly for no
	 * subsystems) then they must match
	 */
1482 1483
	if ((opts->subsys_mask || opts->none)
	    && (opts->subsys_mask != root->subsys_mask))
1484 1485 1486 1487 1488
		return 0;

	return 1;
}

1489 1490 1491 1492
static struct cgroupfs_root *cgroup_root_from_opts(struct cgroup_sb_opts *opts)
{
	struct cgroupfs_root *root;

1493
	if (!opts->subsys_mask && !opts->none)
1494 1495 1496 1497 1498 1499
		return NULL;

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

1500 1501 1502 1503
	if (!init_root_id(root)) {
		kfree(root);
		return ERR_PTR(-ENOMEM);
	}
1504
	init_cgroup_root(root);
1505

1506
	root->subsys_mask = opts->subsys_mask;
1507
	root->flags = opts->flags;
T
Tejun Heo 已提交
1508
	ida_init(&root->cgroup_ida);
1509 1510 1511 1512
	if (opts->release_agent)
		strcpy(root->release_agent_path, opts->release_agent);
	if (opts->name)
		strcpy(root->name, opts->name);
1513 1514
	if (opts->cpuset_clone_children)
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &root->top_cgroup.flags);
1515 1516 1517
	return root;
}

1518 1519 1520 1521 1522 1523 1524 1525 1526
static void cgroup_drop_root(struct cgroupfs_root *root)
{
	if (!root)
		return;

	BUG_ON(!root->hierarchy_id);
	spin_lock(&hierarchy_id_lock);
	ida_remove(&hierarchy_ida, root->hierarchy_id);
	spin_unlock(&hierarchy_id_lock);
T
Tejun Heo 已提交
1527
	ida_destroy(&root->cgroup_ida);
1528 1529 1530
	kfree(root);
}

1531 1532 1533
static int cgroup_set_super(struct super_block *sb, void *data)
{
	int ret;
1534 1535 1536 1537 1538 1539
	struct cgroup_sb_opts *opts = data;

	/* If we don't have a new root, we can't set up a new sb */
	if (!opts->new_root)
		return -EINVAL;

1540
	BUG_ON(!opts->subsys_mask && !opts->none);
1541 1542 1543 1544 1545

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

1546 1547
	sb->s_fs_info = opts->new_root;
	opts->new_root->sb = sb;
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558

	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = CGROUP_SUPER_MAGIC;
	sb->s_op = &cgroup_ops;

	return 0;
}

static int cgroup_get_rootdir(struct super_block *sb)
{
A
Al Viro 已提交
1559 1560
	static const struct dentry_operations cgroup_dops = {
		.d_iput = cgroup_diput,
1561
		.d_delete = cgroup_delete,
A
Al Viro 已提交
1562 1563
	};

1564 1565 1566 1567 1568 1569 1570 1571 1572 1573
	struct inode *inode =
		cgroup_new_inode(S_IFDIR | S_IRUGO | S_IXUGO | S_IWUSR, sb);

	if (!inode)
		return -ENOMEM;

	inode->i_fop = &simple_dir_operations;
	inode->i_op = &cgroup_dir_inode_operations;
	/* directories start off with i_nlink == 2 (for "." entry) */
	inc_nlink(inode);
1574 1575
	sb->s_root = d_make_root(inode);
	if (!sb->s_root)
1576
		return -ENOMEM;
A
Al Viro 已提交
1577 1578
	/* for everything else we want ->d_op set */
	sb->s_d_op = &cgroup_dops;
1579 1580 1581
	return 0;
}

A
Al Viro 已提交
1582
static struct dentry *cgroup_mount(struct file_system_type *fs_type,
1583
			 int flags, const char *unused_dev_name,
A
Al Viro 已提交
1584
			 void *data)
1585 1586
{
	struct cgroup_sb_opts opts;
1587
	struct cgroupfs_root *root;
1588 1589
	int ret = 0;
	struct super_block *sb;
1590
	struct cgroupfs_root *new_root;
T
Tejun Heo 已提交
1591
	struct inode *inode;
1592 1593

	/* First find the desired set of subsystems */
B
Ben Blum 已提交
1594
	mutex_lock(&cgroup_mutex);
1595
	ret = parse_cgroupfs_options(data, &opts);
B
Ben Blum 已提交
1596
	mutex_unlock(&cgroup_mutex);
1597 1598
	if (ret)
		goto out_err;
1599

1600 1601 1602 1603 1604 1605 1606
	/*
	 * Allocate a new cgroup root. We may not need it if we're
	 * reusing an existing hierarchy.
	 */
	new_root = cgroup_root_from_opts(&opts);
	if (IS_ERR(new_root)) {
		ret = PTR_ERR(new_root);
B
Ben Blum 已提交
1607
		goto drop_modules;
1608
	}
1609
	opts.new_root = new_root;
1610

1611
	/* Locate an existing or new sb for this hierarchy */
D
David Howells 已提交
1612
	sb = sget(fs_type, cgroup_test_super, cgroup_set_super, 0, &opts);
1613
	if (IS_ERR(sb)) {
1614
		ret = PTR_ERR(sb);
1615
		cgroup_drop_root(opts.new_root);
B
Ben Blum 已提交
1616
		goto drop_modules;
1617 1618
	}

1619 1620 1621 1622 1623
	root = sb->s_fs_info;
	BUG_ON(!root);
	if (root == opts.new_root) {
		/* We used the new root structure, so this is a new hierarchy */
		struct list_head tmp_cg_links;
1624
		struct cgroup *root_cgrp = &root->top_cgroup;
1625
		struct cgroupfs_root *existing_root;
1626
		const struct cred *cred;
1627
		int i;
1628
		struct css_set *cg;
1629 1630 1631 1632 1633 1634

		BUG_ON(sb->s_root != NULL);

		ret = cgroup_get_rootdir(sb);
		if (ret)
			goto drop_new_super;
1635
		inode = sb->s_root->d_inode;
1636

1637
		mutex_lock(&inode->i_mutex);
1638
		mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1639
		mutex_lock(&cgroup_root_mutex);
1640

T
Tejun Heo 已提交
1641 1642 1643 1644 1645 1646
		/* Check for name clashes with existing mounts */
		ret = -EBUSY;
		if (strlen(root->name))
			for_each_active_root(existing_root)
				if (!strcmp(existing_root->name, root->name))
					goto unlock_drop;
1647

1648 1649 1650 1651 1652 1653 1654 1655
		/*
		 * We're accessing css_set_count without locking
		 * css_set_lock here, but that's OK - it can only be
		 * increased by someone holding cgroup_lock, and
		 * that's us. The worst that can happen is that we
		 * have some link structures left over
		 */
		ret = allocate_cg_links(css_set_count, &tmp_cg_links);
T
Tejun Heo 已提交
1656 1657
		if (ret)
			goto unlock_drop;
1658

1659
		ret = rebind_subsystems(root, root->subsys_mask);
1660
		if (ret == -EBUSY) {
1661
			free_cg_links(&tmp_cg_links);
T
Tejun Heo 已提交
1662
			goto unlock_drop;
1663
		}
B
Ben Blum 已提交
1664 1665 1666 1667 1668
		/*
		 * There must be no failure case after here, since rebinding
		 * takes care of subsystems' refcounts, which are explicitly
		 * dropped in the failure exit path.
		 */
1669 1670 1671 1672 1673

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

		list_add(&root->root_list, &roots);
1674
		root_count++;
1675

1676
		sb->s_root->d_fsdata = root_cgrp;
1677 1678
		root->top_cgroup.dentry = sb->s_root;

1679 1680 1681
		/* Link the top cgroup in this hierarchy into all
		 * the css_set objects */
		write_lock(&css_set_lock);
1682
		hash_for_each(css_set_table, i, cg, hlist)
1683
			link_css_set(&tmp_cg_links, cg, root_cgrp);
1684 1685 1686 1687
		write_unlock(&css_set_lock);

		free_cg_links(&tmp_cg_links);

1688
		BUG_ON(!list_empty(&root_cgrp->children));
1689 1690
		BUG_ON(root->number_of_cgroups != 1);

1691
		cred = override_creds(&init_cred);
1692
		cgroup_populate_dir(root_cgrp, true, root->subsys_mask);
1693
		revert_creds(cred);
T
Tejun Heo 已提交
1694
		mutex_unlock(&cgroup_root_mutex);
1695
		mutex_unlock(&cgroup_mutex);
1696
		mutex_unlock(&inode->i_mutex);
1697 1698 1699 1700 1701
	} else {
		/*
		 * We re-used an existing hierarchy - the new root (if
		 * any) is not needed
		 */
1702
		cgroup_drop_root(opts.new_root);
B
Ben Blum 已提交
1703
		/* no subsys rebinding, so refcounts don't change */
1704
		drop_parsed_module_refcounts(opts.subsys_mask);
1705 1706
	}

1707 1708
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1709
	return dget(sb->s_root);
1710

T
Tejun Heo 已提交
1711 1712 1713 1714
 unlock_drop:
	mutex_unlock(&cgroup_root_mutex);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&inode->i_mutex);
1715
 drop_new_super:
1716
	deactivate_locked_super(sb);
B
Ben Blum 已提交
1717
 drop_modules:
1718
	drop_parsed_module_refcounts(opts.subsys_mask);
1719 1720 1721
 out_err:
	kfree(opts.release_agent);
	kfree(opts.name);
A
Al Viro 已提交
1722
	return ERR_PTR(ret);
1723 1724 1725 1726
}

static void cgroup_kill_sb(struct super_block *sb) {
	struct cgroupfs_root *root = sb->s_fs_info;
1727
	struct cgroup *cgrp = &root->top_cgroup;
1728
	int ret;
K
KOSAKI Motohiro 已提交
1729 1730
	struct cg_cgroup_link *link;
	struct cg_cgroup_link *saved_link;
1731 1732 1733 1734

	BUG_ON(!root);

	BUG_ON(root->number_of_cgroups != 1);
1735
	BUG_ON(!list_empty(&cgrp->children));
1736 1737

	mutex_lock(&cgroup_mutex);
T
Tejun Heo 已提交
1738
	mutex_lock(&cgroup_root_mutex);
1739 1740 1741 1742 1743 1744

	/* Rebind all subsystems back to the default hierarchy */
	ret = rebind_subsystems(root, 0);
	/* Shouldn't be able to fail ... */
	BUG_ON(ret);

1745 1746 1747 1748 1749
	/*
	 * Release all the links from css_sets to this hierarchy's
	 * root cgroup
	 */
	write_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
1750 1751 1752

	list_for_each_entry_safe(link, saved_link, &cgrp->css_sets,
				 cgrp_link_list) {
1753
		list_del(&link->cg_link_list);
1754
		list_del(&link->cgrp_link_list);
1755 1756 1757 1758
		kfree(link);
	}
	write_unlock(&css_set_lock);

1759 1760 1761 1762
	if (!list_empty(&root->root_list)) {
		list_del(&root->root_list);
		root_count--;
	}
1763

T
Tejun Heo 已提交
1764
	mutex_unlock(&cgroup_root_mutex);
1765 1766
	mutex_unlock(&cgroup_mutex);

A
Aristeu Rozanski 已提交
1767 1768
	simple_xattrs_free(&cgrp->xattrs);

1769
	kill_litter_super(sb);
1770
	cgroup_drop_root(root);
1771 1772 1773 1774
}

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

1779 1780
static struct kobject *cgroup_kobj;

L
Li Zefan 已提交
1781 1782 1783 1784 1785 1786
/**
 * cgroup_path - generate the path of a cgroup
 * @cgrp: the cgroup in question
 * @buf: the buffer to write the path into
 * @buflen: the length of the buffer
 *
1787 1788 1789 1790 1791 1792
 * Writes path of cgroup into buf.  Returns 0 on success, -errno on error.
 *
 * We can't generate cgroup path using dentry->d_name, as accessing
 * dentry->name must be protected by irq-unsafe dentry->d_lock or parent
 * inode's i_mutex, while on the other hand cgroup_path() can be called
 * with some irq-safe spinlocks held.
1793
 */
1794
int cgroup_path(const struct cgroup *cgrp, char *buf, int buflen)
1795
{
1796
	int ret = -ENAMETOOLONG;
1797
	char *start;
1798

1799 1800
	start = buf + buflen - 1;
	*start = '\0';
1801

1802 1803 1804 1805 1806 1807
	rcu_read_lock();
	while (cgrp) {
		const char *name = cgroup_name(cgrp);
		int len;

		len = strlen(name);
1808
		if ((start -= len) < buf)
1809 1810
			goto out;
		memcpy(start, name, len);
1811

1812
		if (!cgrp->parent)
1813 1814
			break;

1815
		if (--start < buf)
1816
			goto out;
1817
		*start = '/';
1818 1819

		cgrp = cgrp->parent;
1820
	}
1821
	ret = 0;
1822
	memmove(buf, start, buf + buflen - start);
1823 1824 1825
out:
	rcu_read_unlock();
	return ret;
1826
}
B
Ben Blum 已提交
1827
EXPORT_SYMBOL_GPL(cgroup_path);
1828

1829 1830 1831
/*
 * Control Group taskset
 */
1832 1833 1834
struct task_and_cgroup {
	struct task_struct	*task;
	struct cgroup		*cgrp;
1835
	struct css_set		*cg;
1836 1837
};

1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908
struct cgroup_taskset {
	struct task_and_cgroup	single;
	struct flex_array	*tc_array;
	int			tc_array_len;
	int			idx;
	struct cgroup		*cur_cgrp;
};

/**
 * cgroup_taskset_first - reset taskset and return the first task
 * @tset: taskset of interest
 *
 * @tset iteration is initialized and the first task is returned.
 */
struct task_struct *cgroup_taskset_first(struct cgroup_taskset *tset)
{
	if (tset->tc_array) {
		tset->idx = 0;
		return cgroup_taskset_next(tset);
	} else {
		tset->cur_cgrp = tset->single.cgrp;
		return tset->single.task;
	}
}
EXPORT_SYMBOL_GPL(cgroup_taskset_first);

/**
 * cgroup_taskset_next - iterate to the next task in taskset
 * @tset: taskset of interest
 *
 * Return the next task in @tset.  Iteration must have been initialized
 * with cgroup_taskset_first().
 */
struct task_struct *cgroup_taskset_next(struct cgroup_taskset *tset)
{
	struct task_and_cgroup *tc;

	if (!tset->tc_array || tset->idx >= tset->tc_array_len)
		return NULL;

	tc = flex_array_get(tset->tc_array, tset->idx++);
	tset->cur_cgrp = tc->cgrp;
	return tc->task;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_next);

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

/**
 * cgroup_taskset_size - return the number of tasks in taskset
 * @tset: taskset of interest
 */
int cgroup_taskset_size(struct cgroup_taskset *tset)
{
	return tset->tc_array ? tset->tc_array_len : 1;
}
EXPORT_SYMBOL_GPL(cgroup_taskset_size);


B
Ben Blum 已提交
1909 1910 1911
/*
 * cgroup_task_migrate - move a task from one cgroup to another.
 *
1912
 * Must be called with cgroup_mutex and threadgroup locked.
B
Ben Blum 已提交
1913
 */
1914
static void cgroup_task_migrate(struct cgroup *oldcgrp,
1915
				struct task_struct *tsk, struct css_set *newcg)
B
Ben Blum 已提交
1916 1917 1918 1919
{
	struct css_set *oldcg;

	/*
1920 1921 1922
	 * 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
1923
	 */
1924
	WARN_ON_ONCE(tsk->flags & PF_EXITING);
B
Ben Blum 已提交
1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942
	oldcg = tsk->cgroups;

	task_lock(tsk);
	rcu_assign_pointer(tsk->cgroups, newcg);
	task_unlock(tsk);

	/* Update the css_set linked lists if we're using them */
	write_lock(&css_set_lock);
	if (!list_empty(&tsk->cg_list))
		list_move(&tsk->cg_list, &newcg->tasks);
	write_unlock(&css_set_lock);

	/*
	 * We just gained a reference on oldcg by taking it from the task. As
	 * trading it for newcg is protected by cgroup_mutex, we're safe to drop
	 * it here; it will be freed under RCU.
	 */
	set_bit(CGRP_RELEASABLE, &oldcgrp->flags);
1943
	put_css_set(oldcg);
B
Ben Blum 已提交
1944 1945
}

1946
/**
M
Michael S. Tsirkin 已提交
1947 1948
 * cgroup_attach_task_all - attach task 'tsk' to all cgroups of task 'from'
 * @from: attach to all cgroups of a given task
1949 1950
 * @tsk: the task to be attached
 */
M
Michael S. Tsirkin 已提交
1951
int cgroup_attach_task_all(struct task_struct *from, struct task_struct *tsk)
1952 1953 1954 1955 1956 1957
{
	struct cgroupfs_root *root;
	int retval = 0;

	cgroup_lock();
	for_each_active_root(root) {
M
Michael S. Tsirkin 已提交
1958 1959
		struct cgroup *from_cg = task_cgroup_from_root(from, root);

1960
		retval = cgroup_attach_task(from_cg, tsk, false);
1961 1962 1963 1964 1965 1966 1967
		if (retval)
			break;
	}
	cgroup_unlock();

	return retval;
}
M
Michael S. Tsirkin 已提交
1968
EXPORT_SYMBOL_GPL(cgroup_attach_task_all);
1969

B
Ben Blum 已提交
1970
/**
1971
 * cgroup_attach_task - attach a task or a whole threadgroup to a cgroup
B
Ben Blum 已提交
1972
 * @cgrp: the cgroup to attach to
1973 1974
 * @tsk: the task or the leader of the threadgroup to be attached
 * @threadgroup: attach the whole threadgroup?
B
Ben Blum 已提交
1975
 *
1976
 * Call holding cgroup_mutex and the group_rwsem of the leader. Will take
1977
 * task_lock of @tsk or each thread in the threadgroup individually in turn.
B
Ben Blum 已提交
1978
 */
1979 1980
int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk,
		       bool threadgroup)
B
Ben Blum 已提交
1981 1982 1983 1984 1985
{
	int retval, i, group_size;
	struct cgroup_subsys *ss, *failed_ss = NULL;
	struct cgroupfs_root *root = cgrp->root;
	/* threadgroup list cursor and array */
1986
	struct task_struct *leader = tsk;
1987
	struct task_and_cgroup *tc;
1988
	struct flex_array *group;
1989
	struct cgroup_taskset tset = { };
B
Ben Blum 已提交
1990 1991 1992 1993 1994

	/*
	 * step 0: in order to do expensive, possibly blocking operations for
	 * every thread, we cannot iterate the thread group list, since it needs
	 * rcu or tasklist locked. instead, build an array of all threads in the
1995 1996
	 * group - group_rwsem prevents new threads from appearing, and if
	 * threads exit, this will just be an over-estimate.
B
Ben Blum 已提交
1997
	 */
1998 1999 2000 2001
	if (threadgroup)
		group_size = get_nr_threads(tsk);
	else
		group_size = 1;
2002
	/* flex_array supports very large thread-groups better than kmalloc. */
2003
	group = flex_array_alloc(sizeof(*tc), group_size, GFP_KERNEL);
B
Ben Blum 已提交
2004 2005
	if (!group)
		return -ENOMEM;
2006
	/* pre-allocate to guarantee space while iterating in rcu read-side. */
2007
	retval = flex_array_prealloc(group, 0, group_size, GFP_KERNEL);
2008 2009
	if (retval)
		goto out_free_group_list;
B
Ben Blum 已提交
2010 2011

	i = 0;
2012 2013 2014 2015 2016 2017
	/*
	 * Prevent freeing of tasks while we take a snapshot. Tasks that are
	 * already PF_EXITING could be freed from underneath us unless we
	 * take an rcu_read_lock.
	 */
	rcu_read_lock();
B
Ben Blum 已提交
2018
	do {
2019 2020
		struct task_and_cgroup ent;

2021 2022 2023 2024
		/* @tsk either already exited or can't exit until the end */
		if (tsk->flags & PF_EXITING)
			continue;

B
Ben Blum 已提交
2025 2026
		/* as per above, nr_threads may decrease, but not increase. */
		BUG_ON(i >= group_size);
2027 2028
		ent.task = tsk;
		ent.cgrp = task_cgroup_from_root(tsk, root);
2029 2030 2031
		/* nothing to do if this task is already in the cgroup */
		if (ent.cgrp == cgrp)
			continue;
2032 2033 2034 2035
		/*
		 * saying GFP_ATOMIC has no effect here because we did prealloc
		 * earlier, but it's good form to communicate our expectations.
		 */
2036
		retval = flex_array_put(group, i, &ent, GFP_ATOMIC);
2037
		BUG_ON(retval != 0);
B
Ben Blum 已提交
2038
		i++;
2039 2040 2041

		if (!threadgroup)
			break;
B
Ben Blum 已提交
2042
	} while_each_thread(leader, tsk);
2043
	rcu_read_unlock();
B
Ben Blum 已提交
2044 2045
	/* remember the number of threads in the array for later. */
	group_size = i;
2046 2047
	tset.tc_array = group;
	tset.tc_array_len = group_size;
B
Ben Blum 已提交
2048

2049 2050
	/* methods shouldn't be called if no task is actually migrating */
	retval = 0;
2051
	if (!group_size)
2052
		goto out_free_group_list;
2053

B
Ben Blum 已提交
2054 2055 2056 2057 2058
	/*
	 * step 1: check that we can legitimately attach to the cgroup.
	 */
	for_each_subsys(root, ss) {
		if (ss->can_attach) {
2059
			retval = ss->can_attach(cgrp, &tset);
B
Ben Blum 已提交
2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
			if (retval) {
				failed_ss = ss;
				goto out_cancel_attach;
			}
		}
	}

	/*
	 * step 2: make sure css_sets exist for all threads to be migrated.
	 * we use find_css_set, which allocates a new one if necessary.
	 */
	for (i = 0; i < group_size; i++) {
2072
		tc = flex_array_get(group, i);
2073 2074 2075 2076
		tc->cg = find_css_set(tc->task->cgroups, cgrp);
		if (!tc->cg) {
			retval = -ENOMEM;
			goto out_put_css_set_refs;
B
Ben Blum 已提交
2077 2078 2079 2080
		}
	}

	/*
2081 2082 2083
	 * step 3: now that we're guaranteed success wrt the css_sets,
	 * proceed to move all tasks to the new cgroup.  There are no
	 * failure cases after here, so this is the commit point.
B
Ben Blum 已提交
2084 2085
	 */
	for (i = 0; i < group_size; i++) {
2086
		tc = flex_array_get(group, i);
2087
		cgroup_task_migrate(tc->cgrp, tc->task, tc->cg);
B
Ben Blum 已提交
2088 2089 2090 2091
	}
	/* nothing is sensitive to fork() after this point. */

	/*
2092
	 * step 4: do subsystem attach callbacks.
B
Ben Blum 已提交
2093 2094 2095
	 */
	for_each_subsys(root, ss) {
		if (ss->attach)
2096
			ss->attach(cgrp, &tset);
B
Ben Blum 已提交
2097 2098 2099 2100 2101 2102
	}

	/*
	 * step 5: success! and cleanup
	 */
	retval = 0;
2103 2104 2105 2106 2107 2108 2109 2110
out_put_css_set_refs:
	if (retval) {
		for (i = 0; i < group_size; i++) {
			tc = flex_array_get(group, i);
			if (!tc->cg)
				break;
			put_css_set(tc->cg);
		}
B
Ben Blum 已提交
2111 2112 2113 2114
	}
out_cancel_attach:
	if (retval) {
		for_each_subsys(root, ss) {
2115
			if (ss == failed_ss)
B
Ben Blum 已提交
2116 2117
				break;
			if (ss->cancel_attach)
2118
				ss->cancel_attach(cgrp, &tset);
B
Ben Blum 已提交
2119 2120 2121
		}
	}
out_free_group_list:
2122
	flex_array_free(group);
B
Ben Blum 已提交
2123 2124 2125 2126 2127
	return retval;
}

/*
 * Find the task_struct of the task to attach by vpid and pass it along to the
2128 2129
 * function to attach either it or all tasks in its threadgroup. Will lock
 * cgroup_mutex and threadgroup; may take task_lock of task.
2130
 */
B
Ben Blum 已提交
2131
static int attach_task_by_pid(struct cgroup *cgrp, u64 pid, bool threadgroup)
2132 2133
{
	struct task_struct *tsk;
2134
	const struct cred *cred = current_cred(), *tcred;
2135 2136
	int ret;

B
Ben Blum 已提交
2137 2138 2139
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;

2140 2141
retry_find_task:
	rcu_read_lock();
2142
	if (pid) {
2143
		tsk = find_task_by_vpid(pid);
B
Ben Blum 已提交
2144 2145
		if (!tsk) {
			rcu_read_unlock();
2146 2147
			ret= -ESRCH;
			goto out_unlock_cgroup;
2148
		}
B
Ben Blum 已提交
2149 2150 2151 2152
		/*
		 * even if we're attaching all tasks in the thread group, we
		 * only need to check permissions on one of them.
		 */
2153
		tcred = __task_cred(tsk);
2154 2155 2156
		if (!uid_eq(cred->euid, GLOBAL_ROOT_UID) &&
		    !uid_eq(cred->euid, tcred->uid) &&
		    !uid_eq(cred->euid, tcred->suid)) {
2157
			rcu_read_unlock();
2158 2159
			ret = -EACCES;
			goto out_unlock_cgroup;
2160
		}
2161 2162
	} else
		tsk = current;
2163 2164

	if (threadgroup)
2165
		tsk = tsk->group_leader;
2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177

	/*
	 * Workqueue threads may acquire PF_THREAD_BOUND and become
	 * trapped in a cpuset, or RT worker may be born in a cgroup
	 * with no rt_runtime allocated.  Just say no.
	 */
	if (tsk == kthreadd_task || (tsk->flags & PF_THREAD_BOUND)) {
		ret = -EINVAL;
		rcu_read_unlock();
		goto out_unlock_cgroup;
	}

2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194
	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;
		}
2195 2196 2197 2198
	}

	ret = cgroup_attach_task(cgrp, tsk, threadgroup);

2199 2200
	threadgroup_unlock(tsk);

2201
	put_task_struct(tsk);
2202
out_unlock_cgroup:
B
Ben Blum 已提交
2203
	cgroup_unlock();
2204 2205 2206
	return ret;
}

2207
static int cgroup_tasks_write(struct cgroup *cgrp, struct cftype *cft, u64 pid)
B
Ben Blum 已提交
2208 2209 2210 2211 2212
{
	return attach_task_by_pid(cgrp, pid, false);
}

static int cgroup_procs_write(struct cgroup *cgrp, struct cftype *cft, u64 tgid)
2213
{
2214
	return attach_task_by_pid(cgrp, tgid, true);
2215 2216
}

2217 2218 2219 2220
/**
 * cgroup_lock_live_group - take cgroup_mutex and check that cgrp is alive.
 * @cgrp: the cgroup to be checked for liveness
 *
2221 2222
 * On success, returns true; the lock should be later released with
 * cgroup_unlock(). On failure returns false with no lock held.
2223
 */
2224
bool cgroup_lock_live_group(struct cgroup *cgrp)
2225 2226 2227 2228 2229 2230 2231 2232
{
	mutex_lock(&cgroup_mutex);
	if (cgroup_is_removed(cgrp)) {
		mutex_unlock(&cgroup_mutex);
		return false;
	}
	return true;
}
B
Ben Blum 已提交
2233
EXPORT_SYMBOL_GPL(cgroup_lock_live_group);
2234 2235 2236 2237 2238

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

static int cgroup_release_agent_show(struct cgroup *cgrp, struct cftype *cft,
				     struct seq_file *seq)
{
	if (!cgroup_lock_live_group(cgrp))
		return -ENODEV;
	seq_puts(seq, cgrp->root->release_agent_path);
	seq_putc(seq, '\n');
2257
	cgroup_unlock();
2258 2259 2260
	return 0;
}

2261 2262 2263
/* A buffer size big enough for numbers or short strings */
#define CGROUP_LOCAL_BUFFER_SIZE 64

2264
static ssize_t cgroup_write_X64(struct cgroup *cgrp, struct cftype *cft,
2265 2266 2267
				struct file *file,
				const char __user *userbuf,
				size_t nbytes, loff_t *unused_ppos)
2268
{
2269
	char buffer[CGROUP_LOCAL_BUFFER_SIZE];
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
	int retval = 0;
	char *end;

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

	buffer[nbytes] = 0;     /* nul-terminate */
2281
	if (cft->write_u64) {
K
KOSAKI Motohiro 已提交
2282
		u64 val = simple_strtoull(strstrip(buffer), &end, 0);
2283 2284 2285 2286
		if (*end)
			return -EINVAL;
		retval = cft->write_u64(cgrp, cft, val);
	} else {
K
KOSAKI Motohiro 已提交
2287
		s64 val = simple_strtoll(strstrip(buffer), &end, 0);
2288 2289 2290 2291
		if (*end)
			return -EINVAL;
		retval = cft->write_s64(cgrp, cft, val);
	}
2292 2293 2294 2295 2296
	if (!retval)
		retval = nbytes;
	return retval;
}

2297 2298 2299 2300 2301
static ssize_t cgroup_write_string(struct cgroup *cgrp, struct cftype *cft,
				   struct file *file,
				   const char __user *userbuf,
				   size_t nbytes, loff_t *unused_ppos)
{
2302
	char local_buffer[CGROUP_LOCAL_BUFFER_SIZE];
2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
	int retval = 0;
	size_t max_bytes = cft->max_write_len;
	char *buffer = local_buffer;

	if (!max_bytes)
		max_bytes = sizeof(local_buffer) - 1;
	if (nbytes >= max_bytes)
		return -E2BIG;
	/* Allocate a dynamic buffer if we need one */
	if (nbytes >= sizeof(local_buffer)) {
		buffer = kmalloc(nbytes + 1, GFP_KERNEL);
		if (buffer == NULL)
			return -ENOMEM;
	}
L
Li Zefan 已提交
2317 2318 2319 2320
	if (nbytes && copy_from_user(buffer, userbuf, nbytes)) {
		retval = -EFAULT;
		goto out;
	}
2321 2322

	buffer[nbytes] = 0;     /* nul-terminate */
K
KOSAKI Motohiro 已提交
2323
	retval = cft->write_string(cgrp, cft, strstrip(buffer));
2324 2325
	if (!retval)
		retval = nbytes;
L
Li Zefan 已提交
2326
out:
2327 2328 2329 2330 2331
	if (buffer != local_buffer)
		kfree(buffer);
	return retval;
}

2332 2333 2334 2335
static ssize_t cgroup_file_write(struct file *file, const char __user *buf,
						size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2336
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2337

2338
	if (cgroup_is_removed(cgrp))
2339
		return -ENODEV;
2340
	if (cft->write)
2341
		return cft->write(cgrp, cft, file, buf, nbytes, ppos);
2342 2343
	if (cft->write_u64 || cft->write_s64)
		return cgroup_write_X64(cgrp, cft, file, buf, nbytes, ppos);
2344 2345
	if (cft->write_string)
		return cgroup_write_string(cgrp, cft, file, buf, nbytes, ppos);
2346 2347 2348 2349
	if (cft->trigger) {
		int ret = cft->trigger(cgrp, (unsigned int)cft->private);
		return ret ? ret : nbytes;
	}
2350
	return -EINVAL;
2351 2352
}

2353 2354 2355 2356
static ssize_t cgroup_read_u64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
2357
{
2358
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2359
	u64 val = cft->read_u64(cgrp, cft);
2360 2361 2362 2363 2364
	int len = sprintf(tmp, "%llu\n", (unsigned long long) val);

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

2365 2366 2367 2368 2369
static ssize_t cgroup_read_s64(struct cgroup *cgrp, struct cftype *cft,
			       struct file *file,
			       char __user *buf, size_t nbytes,
			       loff_t *ppos)
{
2370
	char tmp[CGROUP_LOCAL_BUFFER_SIZE];
2371 2372 2373 2374 2375 2376
	s64 val = cft->read_s64(cgrp, cft);
	int len = sprintf(tmp, "%lld\n", (long long) val);

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

2377 2378 2379 2380
static ssize_t cgroup_file_read(struct file *file, char __user *buf,
				   size_t nbytes, loff_t *ppos)
{
	struct cftype *cft = __d_cft(file->f_dentry);
2381
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
2382

2383
	if (cgroup_is_removed(cgrp))
2384 2385 2386
		return -ENODEV;

	if (cft->read)
2387
		return cft->read(cgrp, cft, file, buf, nbytes, ppos);
2388 2389
	if (cft->read_u64)
		return cgroup_read_u64(cgrp, cft, file, buf, nbytes, ppos);
2390 2391
	if (cft->read_s64)
		return cgroup_read_s64(cgrp, cft, file, buf, nbytes, ppos);
2392 2393 2394
	return -EINVAL;
}

2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414
/*
 * seqfile ops/methods for returning structured data. Currently just
 * supports string->u64 maps, but can be extended in future.
 */

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

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

static int cgroup_seqfile_show(struct seq_file *m, void *arg)
{
	struct cgroup_seqfile_state *state = m->private;
	struct cftype *cft = state->cft;
2415 2416 2417 2418 2419 2420 2421 2422
	if (cft->read_map) {
		struct cgroup_map_cb cb = {
			.fill = cgroup_map_add,
			.state = m,
		};
		return cft->read_map(state->cgroup, cft, &cb);
	}
	return cft->read_seq_string(state->cgroup, cft, m);
2423 2424
}

2425
static int cgroup_seqfile_release(struct inode *inode, struct file *file)
2426 2427 2428 2429 2430 2431
{
	struct seq_file *seq = file->private_data;
	kfree(seq->private);
	return single_release(inode, file);
}

2432
static const struct file_operations cgroup_seqfile_operations = {
2433
	.read = seq_read,
2434
	.write = cgroup_file_write,
2435 2436 2437 2438
	.llseek = seq_lseek,
	.release = cgroup_seqfile_release,
};

2439 2440 2441 2442 2443 2444 2445 2446 2447
static int cgroup_file_open(struct inode *inode, struct file *file)
{
	int err;
	struct cftype *cft;

	err = generic_file_open(inode, file);
	if (err)
		return err;
	cft = __d_cft(file->f_dentry);
2448

2449
	if (cft->read_map || cft->read_seq_string) {
2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460
		struct cgroup_seqfile_state *state =
			kzalloc(sizeof(*state), GFP_USER);
		if (!state)
			return -ENOMEM;
		state->cft = cft;
		state->cgroup = __d_cgrp(file->f_dentry->d_parent);
		file->f_op = &cgroup_seqfile_operations;
		err = single_open(file, cgroup_seqfile_show, state);
		if (err < 0)
			kfree(state);
	} else if (cft->open)
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481
		err = cft->open(inode, file);
	else
		err = 0;

	return err;
}

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

/*
 * cgroup_rename - Only allow simple rename of directories in place.
 */
static int cgroup_rename(struct inode *old_dir, struct dentry *old_dentry,
			    struct inode *new_dir, struct dentry *new_dentry)
{
2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
	int ret;
	struct cgroup_name *name, *old_name;
	struct cgroup *cgrp;

	/*
	 * It's convinient to use parent dir's i_mutex to protected
	 * cgrp->name.
	 */
	lockdep_assert_held(&old_dir->i_mutex);

2492 2493 2494 2495 2496 2497
	if (!S_ISDIR(old_dentry->d_inode->i_mode))
		return -ENOTDIR;
	if (new_dentry->d_inode)
		return -EEXIST;
	if (old_dir != new_dir)
		return -EIO;
2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515

	cgrp = __d_cgrp(old_dentry);

	name = cgroup_alloc_name(new_dentry);
	if (!name)
		return -ENOMEM;

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

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

	kfree_rcu(old_name, rcu_head);
	return 0;
2516 2517
}

A
Aristeu Rozanski 已提交
2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575
static struct simple_xattrs *__d_xattrs(struct dentry *dentry)
{
	if (S_ISDIR(dentry->d_inode->i_mode))
		return &__d_cgrp(dentry)->xattrs;
	else
		return &__d_cft(dentry)->xattrs;
}

static inline int xattr_enabled(struct dentry *dentry)
{
	struct cgroupfs_root *root = dentry->d_sb->s_fs_info;
	return test_bit(ROOT_XATTR, &root->flags);
}

static bool is_valid_xattr(const char *name)
{
	if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN) ||
	    !strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN))
		return true;
	return false;
}

static int cgroup_setxattr(struct dentry *dentry, const char *name,
			   const void *val, size_t size, int flags)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_set(__d_xattrs(dentry), name, val, size, flags);
}

static int cgroup_removexattr(struct dentry *dentry, const char *name)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_remove(__d_xattrs(dentry), name);
}

static ssize_t cgroup_getxattr(struct dentry *dentry, const char *name,
			       void *buf, size_t size)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	if (!is_valid_xattr(name))
		return -EINVAL;
	return simple_xattr_get(__d_xattrs(dentry), name, buf, size);
}

static ssize_t cgroup_listxattr(struct dentry *dentry, char *buf, size_t size)
{
	if (!xattr_enabled(dentry))
		return -EOPNOTSUPP;
	return simple_xattr_list(__d_xattrs(dentry), buf, size);
}

2576
static const struct file_operations cgroup_file_operations = {
2577 2578 2579 2580 2581 2582 2583
	.read = cgroup_file_read,
	.write = cgroup_file_write,
	.llseek = generic_file_llseek,
	.open = cgroup_file_open,
	.release = cgroup_file_release,
};

A
Aristeu Rozanski 已提交
2584 2585 2586 2587 2588 2589 2590
static const struct inode_operations cgroup_file_inode_operations = {
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
};

2591
static const struct inode_operations cgroup_dir_inode_operations = {
2592
	.lookup = cgroup_lookup,
2593 2594 2595
	.mkdir = cgroup_mkdir,
	.rmdir = cgroup_rmdir,
	.rename = cgroup_rename,
A
Aristeu Rozanski 已提交
2596 2597 2598 2599
	.setxattr = cgroup_setxattr,
	.getxattr = cgroup_getxattr,
	.listxattr = cgroup_listxattr,
	.removexattr = cgroup_removexattr,
2600 2601
};

A
Al Viro 已提交
2602
static struct dentry *cgroup_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
2603 2604 2605 2606 2607 2608 2609
{
	if (dentry->d_name.len > NAME_MAX)
		return ERR_PTR(-ENAMETOOLONG);
	d_add(dentry, NULL);
	return NULL;
}

2610 2611 2612 2613 2614
/*
 * Check if a file is a control file
 */
static inline struct cftype *__file_cft(struct file *file)
{
A
Al Viro 已提交
2615
	if (file_inode(file)->i_fop != &cgroup_file_operations)
2616 2617 2618 2619
		return ERR_PTR(-EINVAL);
	return __d_cft(file->f_dentry);
}

A
Al Viro 已提交
2620
static int cgroup_create_file(struct dentry *dentry, umode_t mode,
2621 2622
				struct super_block *sb)
{
2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
	struct inode *inode;

	if (!dentry)
		return -ENOENT;
	if (dentry->d_inode)
		return -EEXIST;

	inode = cgroup_new_inode(mode, sb);
	if (!inode)
		return -ENOMEM;

	if (S_ISDIR(mode)) {
		inode->i_op = &cgroup_dir_inode_operations;
		inode->i_fop = &simple_dir_operations;

		/* start off with i_nlink == 2 (for "." entry) */
		inc_nlink(inode);
T
Tejun Heo 已提交
2640
		inc_nlink(dentry->d_parent->d_inode);
2641

2642 2643 2644 2645 2646 2647 2648 2649 2650
		/*
		 * Control reaches here with cgroup_mutex held.
		 * @inode->i_mutex should nest outside cgroup_mutex but we
		 * want to populate it immediately without releasing
		 * cgroup_mutex.  As @inode isn't visible to anyone else
		 * yet, trylock will always succeed without affecting
		 * lockdep checks.
		 */
		WARN_ON_ONCE(!mutex_trylock(&inode->i_mutex));
2651 2652 2653
	} else if (S_ISREG(mode)) {
		inode->i_size = 0;
		inode->i_fop = &cgroup_file_operations;
A
Aristeu Rozanski 已提交
2654
		inode->i_op = &cgroup_file_inode_operations;
2655 2656 2657 2658 2659 2660
	}
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
	return 0;
}

L
Li Zefan 已提交
2661 2662 2663 2664 2665 2666 2667 2668 2669
/**
 * cgroup_file_mode - deduce file mode of a control file
 * @cft: the control file in question
 *
 * returns cft->mode if ->mode is not 0
 * returns S_IRUGO|S_IWUSR if it has both a read and a write handler
 * returns S_IRUGO if it has only a read handler
 * returns S_IWUSR if it has only a write hander
 */
A
Al Viro 已提交
2670
static umode_t cgroup_file_mode(const struct cftype *cft)
L
Li Zefan 已提交
2671
{
A
Al Viro 已提交
2672
	umode_t mode = 0;
L
Li Zefan 已提交
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687

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

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

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

	return mode;
}

T
Tejun Heo 已提交
2688
static int cgroup_add_file(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2689
			   struct cftype *cft)
2690
{
2691
	struct dentry *dir = cgrp->dentry;
T
Tejun Heo 已提交
2692
	struct cgroup *parent = __d_cgrp(dir);
2693
	struct dentry *dentry;
T
Tejun Heo 已提交
2694
	struct cfent *cfe;
2695
	int error;
A
Al Viro 已提交
2696
	umode_t mode;
2697
	char name[MAX_CGROUP_TYPE_NAMELEN + MAX_CFTYPE_NAME + 2] = { 0 };
2698

A
Aristeu Rozanski 已提交
2699 2700
	simple_xattrs_init(&cft->xattrs);

2701
	if (subsys && !test_bit(ROOT_NOPREFIX, &cgrp->root->flags)) {
2702 2703 2704 2705
		strcpy(name, subsys->name);
		strcat(name, ".");
	}
	strcat(name, cft->name);
T
Tejun Heo 已提交
2706

2707
	BUG_ON(!mutex_is_locked(&dir->d_inode->i_mutex));
T
Tejun Heo 已提交
2708 2709 2710 2711 2712

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

2713
	dentry = lookup_one_len(name, dir, strlen(name));
T
Tejun Heo 已提交
2714
	if (IS_ERR(dentry)) {
2715
		error = PTR_ERR(dentry);
T
Tejun Heo 已提交
2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730
		goto out;
	}

	mode = cgroup_file_mode(cft);
	error = cgroup_create_file(dentry, mode | S_IFREG, cgrp->root->sb);
	if (!error) {
		cfe->type = (void *)cft;
		cfe->dentry = dentry;
		dentry->d_fsdata = cfe;
		list_add_tail(&cfe->node, &parent->files);
		cfe = NULL;
	}
	dput(dentry);
out:
	kfree(cfe);
2731 2732 2733
	return error;
}

2734
static int cgroup_addrm_files(struct cgroup *cgrp, struct cgroup_subsys *subsys,
A
Aristeu Rozanski 已提交
2735
			      struct cftype cfts[], bool is_add)
2736
{
A
Aristeu Rozanski 已提交
2737
	struct cftype *cft;
T
Tejun Heo 已提交
2738 2739 2740
	int err, ret = 0;

	for (cft = cfts; cft->name[0] != '\0'; cft++) {
2741 2742 2743 2744 2745 2746
		/* does cft->flags tell us to skip this file on @cgrp? */
		if ((cft->flags & CFTYPE_NOT_ON_ROOT) && !cgrp->parent)
			continue;
		if ((cft->flags & CFTYPE_ONLY_ON_ROOT) && cgrp->parent)
			continue;

2747
		if (is_add) {
2748
			err = cgroup_add_file(cgrp, subsys, cft);
2749 2750 2751
			if (err)
				pr_warn("cgroup_addrm_files: failed to add %s, err=%d\n",
					cft->name, err);
T
Tejun Heo 已提交
2752
			ret = err;
2753 2754
		} else {
			cgroup_rm_file(cgrp, cft);
T
Tejun Heo 已提交
2755
		}
2756
	}
T
Tejun Heo 已提交
2757
	return ret;
2758 2759
}

2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
static DEFINE_MUTEX(cgroup_cft_mutex);

static void cgroup_cfts_prepare(void)
	__acquires(&cgroup_cft_mutex) __acquires(&cgroup_mutex)
{
	/*
	 * Thanks to the entanglement with vfs inode locking, we can't walk
	 * the existing cgroups under cgroup_mutex and create files.
	 * Instead, we increment reference on all cgroups and build list of
	 * them using @cgrp->cft_q_node.  Grab cgroup_cft_mutex to ensure
	 * exclusive access to the field.
	 */
	mutex_lock(&cgroup_cft_mutex);
	mutex_lock(&cgroup_mutex);
}

static void cgroup_cfts_commit(struct cgroup_subsys *ss,
A
Aristeu Rozanski 已提交
2777
			       struct cftype *cfts, bool is_add)
2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802
	__releases(&cgroup_mutex) __releases(&cgroup_cft_mutex)
{
	LIST_HEAD(pending);
	struct cgroup *cgrp, *n;

	/* %NULL @cfts indicates abort and don't bother if @ss isn't attached */
	if (cfts && ss->root != &rootnode) {
		list_for_each_entry(cgrp, &ss->root->allcg_list, allcg_node) {
			dget(cgrp->dentry);
			list_add_tail(&cgrp->cft_q_node, &pending);
		}
	}

	mutex_unlock(&cgroup_mutex);

	/*
	 * All new cgroups will see @cfts update on @ss->cftsets.  Add/rm
	 * files for all cgroups which were created before.
	 */
	list_for_each_entry_safe(cgrp, n, &pending, cft_q_node) {
		struct inode *inode = cgrp->dentry->d_inode;

		mutex_lock(&inode->i_mutex);
		mutex_lock(&cgroup_mutex);
		if (!cgroup_is_removed(cgrp))
2803
			cgroup_addrm_files(cgrp, ss, cfts, is_add);
2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827
		mutex_unlock(&cgroup_mutex);
		mutex_unlock(&inode->i_mutex);

		list_del_init(&cgrp->cft_q_node);
		dput(cgrp->dentry);
	}

	mutex_unlock(&cgroup_cft_mutex);
}

/**
 * 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
2828
int cgroup_add_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2829 2830 2831 2832 2833 2834 2835 2836 2837 2838
{
	struct cftype_set *set;

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

	cgroup_cfts_prepare();
	set->cfts = cfts;
	list_add_tail(&set->node, &ss->cftsets);
2839
	cgroup_cfts_commit(ss, cfts, true);
2840 2841 2842 2843 2844

	return 0;
}
EXPORT_SYMBOL_GPL(cgroup_add_cftypes);

2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857
/**
 * cgroup_rm_cftypes - remove an array of cftypes from a subsystem
 * @ss: target cgroup subsystem
 * @cfts: zero-length name terminated array of cftypes
 *
 * Unregister @cfts from @ss.  Files described by @cfts are removed from
 * all existing cgroups to which @ss is attached and all future cgroups
 * won't have them either.  This function can be called anytime whether @ss
 * is attached or not.
 *
 * Returns 0 on successful unregistration, -ENOENT if @cfts is not
 * registered with @ss.
 */
A
Aristeu Rozanski 已提交
2858
int cgroup_rm_cftypes(struct cgroup_subsys *ss, struct cftype *cfts)
2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875
{
	struct cftype_set *set;

	cgroup_cfts_prepare();

	list_for_each_entry(set, &ss->cftsets, node) {
		if (set->cfts == cfts) {
			list_del_init(&set->node);
			cgroup_cfts_commit(ss, cfts, false);
			return 0;
		}
	}

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

L
Li Zefan 已提交
2876 2877 2878 2879 2880 2881
/**
 * cgroup_task_count - count the number of tasks in a cgroup.
 * @cgrp: the cgroup in question
 *
 * Return the number of tasks in the cgroup.
 */
2882
int cgroup_task_count(const struct cgroup *cgrp)
2883 2884
{
	int count = 0;
K
KOSAKI Motohiro 已提交
2885
	struct cg_cgroup_link *link;
2886 2887

	read_lock(&css_set_lock);
K
KOSAKI Motohiro 已提交
2888
	list_for_each_entry(link, &cgrp->css_sets, cgrp_link_list) {
2889
		count += atomic_read(&link->cg->refcount);
2890 2891
	}
	read_unlock(&css_set_lock);
2892 2893 2894
	return count;
}

2895 2896 2897 2898
/*
 * Advance a list_head iterator.  The iterator should be positioned at
 * the start of a css_set
 */
2899
static void cgroup_advance_iter(struct cgroup *cgrp,
2900
				struct cgroup_iter *it)
2901 2902 2903 2904 2905 2906 2907 2908
{
	struct list_head *l = it->cg_link;
	struct cg_cgroup_link *link;
	struct css_set *cg;

	/* Advance to the next non-empty css_set */
	do {
		l = l->next;
2909
		if (l == &cgrp->css_sets) {
2910 2911 2912
			it->cg_link = NULL;
			return;
		}
2913
		link = list_entry(l, struct cg_cgroup_link, cgrp_link_list);
2914 2915 2916 2917 2918 2919
		cg = link->cg;
	} while (list_empty(&cg->tasks));
	it->cg_link = l;
	it->task = cg->tasks.next;
}

2920 2921 2922 2923 2924 2925
/*
 * To reduce the fork() overhead for systems that are not actually
 * using their cgroups capability, we don't maintain the lists running
 * through each css_set to its tasks until we see the list actually
 * used - in other words after the first call to cgroup_iter_start().
 */
2926
static void cgroup_enable_task_cg_lists(void)
2927 2928 2929 2930
{
	struct task_struct *p, *g;
	write_lock(&css_set_lock);
	use_task_css_set_links = 1;
2931 2932 2933 2934 2935 2936 2937 2938
	/*
	 * 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);
2939 2940
	do_each_thread(g, p) {
		task_lock(p);
2941 2942 2943 2944 2945 2946
		/*
		 * We should check if the process is exiting, otherwise
		 * it will race with cgroup_exit() in that the list
		 * entry won't be deleted though the process has exited.
		 */
		if (!(p->flags & PF_EXITING) && list_empty(&p->cg_list))
2947 2948 2949
			list_add(&p->cg_list, &p->cgroups->tasks);
		task_unlock(p);
	} while_each_thread(g, p);
2950
	read_unlock(&tasklist_lock);
2951 2952 2953
	write_unlock(&css_set_lock);
}

2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994
/**
 * cgroup_next_descendant_pre - find the next descendant for pre-order walk
 * @pos: the current position (%NULL to initiate traversal)
 * @cgroup: cgroup whose descendants to walk
 *
 * To be used by cgroup_for_each_descendant_pre().  Find the next
 * descendant to visit for pre-order traversal of @cgroup's descendants.
 */
struct cgroup *cgroup_next_descendant_pre(struct cgroup *pos,
					  struct cgroup *cgroup)
{
	struct cgroup *next;

	WARN_ON_ONCE(!rcu_read_lock_held());

	/* if first iteration, pretend we just visited @cgroup */
	if (!pos) {
		if (list_empty(&cgroup->children))
			return NULL;
		pos = cgroup;
	}

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

	/* no child, visit my or the closest ancestor's next sibling */
	do {
		next = list_entry_rcu(pos->sibling.next, struct cgroup,
				      sibling);
		if (&next->sibling != &pos->parent->children)
			return next;

		pos = pos->parent;
	} while (pos != cgroup);

	return NULL;
}
EXPORT_SYMBOL_GPL(cgroup_next_descendant_pre);

2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020
/**
 * cgroup_rightmost_descendant - return the rightmost descendant of a cgroup
 * @pos: cgroup of interest
 *
 * Return the rightmost descendant of @pos.  If there's no descendant,
 * @pos is returned.  This can be used during pre-order traversal to skip
 * subtree of @pos.
 */
struct cgroup *cgroup_rightmost_descendant(struct cgroup *pos)
{
	struct cgroup *last, *tmp;

	WARN_ON_ONCE(!rcu_read_lock_held());

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

	return last;
}
EXPORT_SYMBOL_GPL(cgroup_rightmost_descendant);

3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065
static struct cgroup *cgroup_leftmost_descendant(struct cgroup *pos)
{
	struct cgroup *last;

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

	return last;
}

/**
 * cgroup_next_descendant_post - find the next descendant for post-order walk
 * @pos: the current position (%NULL to initiate traversal)
 * @cgroup: cgroup whose descendants to walk
 *
 * To be used by cgroup_for_each_descendant_post().  Find the next
 * descendant to visit for post-order traversal of @cgroup's descendants.
 */
struct cgroup *cgroup_next_descendant_post(struct cgroup *pos,
					   struct cgroup *cgroup)
{
	struct cgroup *next;

	WARN_ON_ONCE(!rcu_read_lock_held());

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

	/* if there's an unvisited sibling, visit its leftmost descendant */
	next = list_entry_rcu(pos->sibling.next, struct cgroup, sibling);
	if (&next->sibling != &pos->parent->children)
		return cgroup_leftmost_descendant(next);

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

3066
void cgroup_iter_start(struct cgroup *cgrp, struct cgroup_iter *it)
3067
	__acquires(css_set_lock)
3068 3069 3070 3071 3072 3073
{
	/*
	 * The first time anyone tries to iterate across a cgroup,
	 * we need to enable the list linking each css_set to its
	 * tasks, and fix up all existing tasks.
	 */
3074 3075 3076
	if (!use_task_css_set_links)
		cgroup_enable_task_cg_lists();

3077
	read_lock(&css_set_lock);
3078 3079
	it->cg_link = &cgrp->css_sets;
	cgroup_advance_iter(cgrp, it);
3080 3081
}

3082
struct task_struct *cgroup_iter_next(struct cgroup *cgrp,
3083 3084 3085 3086
					struct cgroup_iter *it)
{
	struct task_struct *res;
	struct list_head *l = it->task;
3087
	struct cg_cgroup_link *link;
3088 3089 3090 3091 3092 3093 3094

	/* If the iterator cg is NULL, we have no tasks */
	if (!it->cg_link)
		return NULL;
	res = list_entry(l, struct task_struct, cg_list);
	/* Advance iterator to find next entry */
	l = l->next;
3095 3096
	link = list_entry(it->cg_link, struct cg_cgroup_link, cgrp_link_list);
	if (l == &link->cg->tasks) {
3097 3098
		/* We reached the end of this task list - move on to
		 * the next cg_cgroup_link */
3099
		cgroup_advance_iter(cgrp, it);
3100 3101 3102 3103 3104 3105
	} else {
		it->task = l;
	}
	return res;
}

3106
void cgroup_iter_end(struct cgroup *cgrp, struct cgroup_iter *it)
3107
	__releases(css_set_lock)
3108 3109 3110 3111
{
	read_unlock(&css_set_lock);
}

3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248
static inline int started_after_time(struct task_struct *t1,
				     struct timespec *time,
				     struct task_struct *t2)
{
	int start_diff = timespec_compare(&t1->start_time, time);
	if (start_diff > 0) {
		return 1;
	} else if (start_diff < 0) {
		return 0;
	} else {
		/*
		 * Arbitrarily, if two processes started at the same
		 * time, we'll say that the lower pointer value
		 * started first. Note that t2 may have exited by now
		 * so this may not be a valid pointer any longer, but
		 * that's fine - it still serves to distinguish
		 * between two tasks started (effectively) simultaneously.
		 */
		return t1 > t2;
	}
}

/*
 * This function is a callback from heap_insert() and is used to order
 * the heap.
 * In this case we order the heap in descending task start time.
 */
static inline int started_after(void *p1, void *p2)
{
	struct task_struct *t1 = p1;
	struct task_struct *t2 = p2;
	return started_after_time(t1, &t2->start_time, t2);
}

/**
 * cgroup_scan_tasks - iterate though all the tasks in a cgroup
 * @scan: struct cgroup_scanner containing arguments for the scan
 *
 * Arguments include pointers to callback functions test_task() and
 * process_task().
 * Iterate through all the tasks in a cgroup, calling test_task() for each,
 * and if it returns true, call process_task() for it also.
 * The test_task pointer may be NULL, meaning always true (select all tasks).
 * Effectively duplicates cgroup_iter_{start,next,end}()
 * but does not lock css_set_lock for the call to process_task().
 * The struct cgroup_scanner may be embedded in any structure of the caller's
 * creation.
 * It is guaranteed that process_task() will act on every task that
 * is a member of the cgroup for the duration of this call. This
 * function may or may not call process_task() for tasks that exit
 * or move to a different cgroup during the call, or are forked or
 * move into the cgroup during the call.
 *
 * Note that test_task() may be called with locks held, and may in some
 * situations be called multiple times for the same task, so it should
 * be cheap.
 * If the heap pointer in the struct cgroup_scanner is non-NULL, a heap has been
 * pre-allocated and will be used for heap operations (and its "gt" member will
 * be overwritten), else a temporary heap will be used (allocation of which
 * may cause this function to fail).
 */
int cgroup_scan_tasks(struct cgroup_scanner *scan)
{
	int retval, i;
	struct cgroup_iter it;
	struct task_struct *p, *dropped;
	/* Never dereference latest_task, since it's not refcounted */
	struct task_struct *latest_task = NULL;
	struct ptr_heap tmp_heap;
	struct ptr_heap *heap;
	struct timespec latest_time = { 0, 0 };

	if (scan->heap) {
		/* The caller supplied our heap and pre-allocated its memory */
		heap = scan->heap;
		heap->gt = &started_after;
	} else {
		/* We need to allocate our own heap memory */
		heap = &tmp_heap;
		retval = heap_init(heap, PAGE_SIZE, GFP_KERNEL, &started_after);
		if (retval)
			/* cannot allocate the heap */
			return retval;
	}

 again:
	/*
	 * Scan tasks in the cgroup, using the scanner's "test_task" callback
	 * to determine which are of interest, and using the scanner's
	 * "process_task" callback to process any of them that need an update.
	 * Since we don't want to hold any locks during the task updates,
	 * gather tasks to be processed in a heap structure.
	 * The heap is sorted by descending task start time.
	 * If the statically-sized heap fills up, we overflow tasks that
	 * started later, and in future iterations only consider tasks that
	 * started after the latest task in the previous pass. This
	 * guarantees forward progress and that we don't miss any tasks.
	 */
	heap->size = 0;
	cgroup_iter_start(scan->cg, &it);
	while ((p = cgroup_iter_next(scan->cg, &it))) {
		/*
		 * Only affect tasks that qualify per the caller's callback,
		 * if he provided one
		 */
		if (scan->test_task && !scan->test_task(p, scan))
			continue;
		/*
		 * Only process tasks that started after the last task
		 * we processed
		 */
		if (!started_after_time(p, &latest_time, latest_task))
			continue;
		dropped = heap_insert(heap, p);
		if (dropped == NULL) {
			/*
			 * The new task was inserted; the heap wasn't
			 * previously full
			 */
			get_task_struct(p);
		} else if (dropped != p) {
			/*
			 * The new task was inserted, and pushed out a
			 * different task
			 */
			get_task_struct(p);
			put_task_struct(dropped);
		}
		/*
		 * Else the new task was newer than anything already in
		 * the heap and wasn't inserted
		 */
	}
	cgroup_iter_end(scan->cg, &it);

	if (heap->size) {
		for (i = 0; i < heap->size; i++) {
3249
			struct task_struct *q = heap->ptrs[i];
3250
			if (i == 0) {
3251 3252
				latest_time = q->start_time;
				latest_task = q;
3253 3254
			}
			/* Process the task per the caller's callback */
3255 3256
			scan->process_task(q, scan);
			put_task_struct(q);
3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271
		}
		/*
		 * If we had to process any tasks at all, scan again
		 * in case some of them were in the middle of forking
		 * children that didn't get processed.
		 * Not the most efficient way to do it, but it avoids
		 * having to take callback_mutex in the fork path
		 */
		goto again;
	}
	if (heap == &tmp_heap)
		heap_free(&tmp_heap);
	return 0;
}

3272
/*
3273
 * Stuff for reading the 'tasks'/'procs' files.
3274 3275 3276 3277 3278 3279 3280 3281
 *
 * 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.
 *
 */

3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313
/* which pidlist file are we talking about? */
enum cgroup_filetype {
	CGROUP_FILE_PROCS,
	CGROUP_FILE_TASKS,
};

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

3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334
/*
 * The following two functions "fix" the issue where there are more pids
 * than kmalloc will give memory for; in such cases, we use vmalloc/vfree.
 * TODO: replace with a kernel-wide solution to this problem
 */
#define PIDLIST_TOO_LARGE(c) ((c) * sizeof(pid_t) > (PAGE_SIZE * 2))
static void *pidlist_allocate(int count)
{
	if (PIDLIST_TOO_LARGE(count))
		return vmalloc(count * sizeof(pid_t));
	else
		return kmalloc(count * sizeof(pid_t), GFP_KERNEL);
}
static void pidlist_free(void *p)
{
	if (is_vmalloc_addr(p))
		vfree(p);
	else
		kfree(p);
}

3335
/*
3336
 * pidlist_uniq - given a kmalloc()ed list, strip out all duplicate entries
3337
 * Returns the number of unique elements.
3338
 */
3339
static int pidlist_uniq(pid_t *list, int length)
3340
{
3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369
	int src, dest = 1;

	/*
	 * we presume the 0th element is unique, so i starts at 1. trivial
	 * edge cases first; no work needs to be done for either
	 */
	if (length == 0 || length == 1)
		return length;
	/* src and dest walk down the list; dest counts unique elements */
	for (src = 1; src < length; src++) {
		/* find next unique element */
		while (list[src] == list[src-1]) {
			src++;
			if (src == length)
				goto after;
		}
		/* dest always points to where the next unique element goes */
		list[dest] = list[src];
		dest++;
	}
after:
	return dest;
}

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

3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380
/*
 * find the appropriate pidlist for our purpose (given procs vs tasks)
 * returns with the lock on that pidlist already held, and takes care
 * of the use count, or returns NULL with no locks held if we're out of
 * memory.
 */
static struct cgroup_pidlist *cgroup_pidlist_find(struct cgroup *cgrp,
						  enum cgroup_filetype type)
{
	struct cgroup_pidlist *l;
	/* don't need task_nsproxy() if we're looking at ourself */
3381
	struct pid_namespace *ns = task_active_pid_ns(current);
3382

3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406
	/*
	 * We can't drop the pidlist_mutex before taking the l->mutex in case
	 * the last ref-holder is trying to remove l from the list at the same
	 * time. Holding the pidlist_mutex precludes somebody taking whichever
	 * list we find out from under us - compare release_pid_array().
	 */
	mutex_lock(&cgrp->pidlist_mutex);
	list_for_each_entry(l, &cgrp->pidlists, links) {
		if (l->key.type == type && l->key.ns == ns) {
			/* make sure l doesn't vanish out from under us */
			down_write(&l->mutex);
			mutex_unlock(&cgrp->pidlist_mutex);
			return l;
		}
	}
	/* entry not found; create a new one */
	l = kmalloc(sizeof(struct cgroup_pidlist), GFP_KERNEL);
	if (!l) {
		mutex_unlock(&cgrp->pidlist_mutex);
		return l;
	}
	init_rwsem(&l->mutex);
	down_write(&l->mutex);
	l->key.type = type;
3407
	l->key.ns = get_pid_ns(ns);
3408 3409 3410 3411 3412 3413 3414 3415
	l->use_count = 0; /* don't increment here */
	l->list = NULL;
	l->owner = cgrp;
	list_add(&l->links, &cgrp->pidlists);
	mutex_unlock(&cgrp->pidlist_mutex);
	return l;
}

3416 3417 3418
/*
 * Load a cgroup's pidarray with either procs' tgids or tasks' pids
 */
3419 3420
static int pidlist_array_load(struct cgroup *cgrp, enum cgroup_filetype type,
			      struct cgroup_pidlist **lp)
3421 3422 3423 3424
{
	pid_t *array;
	int length;
	int pid, n = 0; /* used for populating the array */
3425 3426
	struct cgroup_iter it;
	struct task_struct *tsk;
3427 3428 3429 3430 3431 3432 3433 3434 3435
	struct cgroup_pidlist *l;

	/*
	 * If cgroup gets more users after we read count, we won't have
	 * enough space - tough.  This race is indistinguishable to the
	 * caller from the case that the additional cgroup users didn't
	 * show up until sometime later on.
	 */
	length = cgroup_task_count(cgrp);
3436
	array = pidlist_allocate(length);
3437 3438 3439
	if (!array)
		return -ENOMEM;
	/* now, populate the array */
3440 3441
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
3442
		if (unlikely(n == length))
3443
			break;
3444
		/* get tgid or pid for procs or tasks file respectively */
3445 3446 3447 3448
		if (type == CGROUP_FILE_PROCS)
			pid = task_tgid_vnr(tsk);
		else
			pid = task_pid_vnr(tsk);
3449 3450
		if (pid > 0) /* make sure to only use valid results */
			array[n++] = pid;
3451
	}
3452
	cgroup_iter_end(cgrp, &it);
3453 3454 3455
	length = n;
	/* now sort & (if procs) strip out duplicates */
	sort(array, length, sizeof(pid_t), cmppid, NULL);
3456
	if (type == CGROUP_FILE_PROCS)
3457
		length = pidlist_uniq(array, length);
3458 3459
	l = cgroup_pidlist_find(cgrp, type);
	if (!l) {
3460
		pidlist_free(array);
3461
		return -ENOMEM;
3462
	}
3463
	/* store array, freeing old if necessary - lock already held */
3464
	pidlist_free(l->list);
3465 3466 3467 3468
	l->list = array;
	l->length = length;
	l->use_count++;
	up_write(&l->mutex);
3469
	*lp = l;
3470
	return 0;
3471 3472
}

B
Balbir Singh 已提交
3473
/**
L
Li Zefan 已提交
3474
 * cgroupstats_build - build and fill cgroupstats
B
Balbir Singh 已提交
3475 3476 3477
 * @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
3478 3479 3480
 *
 * Build and fill cgroupstats so that taskstats can export it to user
 * space.
B
Balbir Singh 已提交
3481 3482 3483 3484
 */
int cgroupstats_build(struct cgroupstats *stats, struct dentry *dentry)
{
	int ret = -EINVAL;
3485
	struct cgroup *cgrp;
B
Balbir Singh 已提交
3486 3487
	struct cgroup_iter it;
	struct task_struct *tsk;
3488

B
Balbir Singh 已提交
3489
	/*
3490 3491
	 * Validate dentry by checking the superblock operations,
	 * and make sure it's a directory.
B
Balbir Singh 已提交
3492
	 */
3493 3494
	if (dentry->d_sb->s_op != &cgroup_ops ||
	    !S_ISDIR(dentry->d_inode->i_mode))
B
Balbir Singh 已提交
3495 3496 3497
		 goto err;

	ret = 0;
3498
	cgrp = dentry->d_fsdata;
B
Balbir Singh 已提交
3499

3500 3501
	cgroup_iter_start(cgrp, &it);
	while ((tsk = cgroup_iter_next(cgrp, &it))) {
B
Balbir Singh 已提交
3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520
		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;
		}
	}
3521
	cgroup_iter_end(cgrp, &it);
B
Balbir Singh 已提交
3522 3523 3524 3525 3526

err:
	return ret;
}

3527

3528
/*
3529
 * seq_file methods for the tasks/procs files. The seq_file position is the
3530
 * next pid to display; the seq_file iterator is a pointer to the pid
3531
 * in the cgroup->l->list array.
3532
 */
3533

3534
static void *cgroup_pidlist_start(struct seq_file *s, loff_t *pos)
3535
{
3536 3537 3538 3539 3540 3541
	/*
	 * 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
	 */
3542
	struct cgroup_pidlist *l = s->private;
3543 3544 3545
	int index = 0, pid = *pos;
	int *iter;

3546
	down_read(&l->mutex);
3547
	if (pid) {
3548
		int end = l->length;
S
Stephen Rothwell 已提交
3549

3550 3551
		while (index < end) {
			int mid = (index + end) / 2;
3552
			if (l->list[mid] == pid) {
3553 3554
				index = mid;
				break;
3555
			} else if (l->list[mid] <= pid)
3556 3557 3558 3559 3560 3561
				index = mid + 1;
			else
				end = mid;
		}
	}
	/* If we're off the end of the array, we're done */
3562
	if (index >= l->length)
3563 3564
		return NULL;
	/* Update the abstract position to be the actual pid that we found */
3565
	iter = l->list + index;
3566 3567 3568 3569
	*pos = *iter;
	return iter;
}

3570
static void cgroup_pidlist_stop(struct seq_file *s, void *v)
3571
{
3572 3573
	struct cgroup_pidlist *l = s->private;
	up_read(&l->mutex);
3574 3575
}

3576
static void *cgroup_pidlist_next(struct seq_file *s, void *v, loff_t *pos)
3577
{
3578 3579 3580
	struct cgroup_pidlist *l = s->private;
	pid_t *p = v;
	pid_t *end = l->list + l->length;
3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593
	/*
	 * Advance to the next pid in the array. If this goes off the
	 * end, we're done
	 */
	p++;
	if (p >= end) {
		return NULL;
	} else {
		*pos = *p;
		return p;
	}
}

3594
static int cgroup_pidlist_show(struct seq_file *s, void *v)
3595 3596 3597
{
	return seq_printf(s, "%d\n", *(int *)v);
}
3598

3599 3600 3601 3602 3603 3604 3605 3606 3607
/*
 * seq_operations functions for iterating on pidlists through seq_file -
 * independent of whether it's tasks or procs
 */
static const struct seq_operations cgroup_pidlist_seq_operations = {
	.start = cgroup_pidlist_start,
	.stop = cgroup_pidlist_stop,
	.next = cgroup_pidlist_next,
	.show = cgroup_pidlist_show,
3608 3609
};

3610
static void cgroup_release_pid_array(struct cgroup_pidlist *l)
3611
{
3612 3613 3614 3615 3616 3617 3618
	/*
	 * the case where we're the last user of this particular pidlist will
	 * have us remove it from the cgroup's list, which entails taking the
	 * mutex. since in pidlist_find the pidlist->lock depends on cgroup->
	 * pidlist_mutex, we have to take pidlist_mutex first.
	 */
	mutex_lock(&l->owner->pidlist_mutex);
3619 3620 3621
	down_write(&l->mutex);
	BUG_ON(!l->use_count);
	if (!--l->use_count) {
3622 3623 3624
		/* we're the last user if refcount is 0; remove and free */
		list_del(&l->links);
		mutex_unlock(&l->owner->pidlist_mutex);
3625
		pidlist_free(l->list);
3626 3627 3628 3629
		put_pid_ns(l->key.ns);
		up_write(&l->mutex);
		kfree(l);
		return;
3630
	}
3631
	mutex_unlock(&l->owner->pidlist_mutex);
3632
	up_write(&l->mutex);
3633 3634
}

3635
static int cgroup_pidlist_release(struct inode *inode, struct file *file)
3636
{
3637
	struct cgroup_pidlist *l;
3638 3639
	if (!(file->f_mode & FMODE_READ))
		return 0;
3640 3641 3642 3643 3644 3645
	/*
	 * the seq_file will only be initialized if the file was opened for
	 * reading; hence we check if it's not null only in that case.
	 */
	l = ((struct seq_file *)file->private_data)->private;
	cgroup_release_pid_array(l);
3646 3647 3648
	return seq_release(inode, file);
}

3649
static const struct file_operations cgroup_pidlist_operations = {
3650 3651 3652
	.read = seq_read,
	.llseek = seq_lseek,
	.write = cgroup_file_write,
3653
	.release = cgroup_pidlist_release,
3654 3655
};

3656
/*
3657 3658 3659
 * The following functions handle opens on a file that displays a pidlist
 * (tasks or procs). Prepare an array of the process/thread IDs of whoever's
 * in the cgroup.
3660
 */
3661
/* helper function for the two below it */
3662
static int cgroup_pidlist_open(struct file *file, enum cgroup_filetype type)
3663
{
3664
	struct cgroup *cgrp = __d_cgrp(file->f_dentry->d_parent);
3665
	struct cgroup_pidlist *l;
3666
	int retval;
3667

3668
	/* Nothing to do for write-only files */
3669 3670 3671
	if (!(file->f_mode & FMODE_READ))
		return 0;

3672
	/* have the array populated */
3673
	retval = pidlist_array_load(cgrp, type, &l);
3674 3675 3676 3677
	if (retval)
		return retval;
	/* configure file information */
	file->f_op = &cgroup_pidlist_operations;
3678

3679
	retval = seq_open(file, &cgroup_pidlist_seq_operations);
3680
	if (retval) {
3681
		cgroup_release_pid_array(l);
3682
		return retval;
3683
	}
3684
	((struct seq_file *)file->private_data)->private = l;
3685 3686
	return 0;
}
3687 3688
static int cgroup_tasks_open(struct inode *unused, struct file *file)
{
3689
	return cgroup_pidlist_open(file, CGROUP_FILE_TASKS);
3690 3691 3692
}
static int cgroup_procs_open(struct inode *unused, struct file *file)
{
3693
	return cgroup_pidlist_open(file, CGROUP_FILE_PROCS);
3694
}
3695

3696
static u64 cgroup_read_notify_on_release(struct cgroup *cgrp,
3697 3698
					    struct cftype *cft)
{
3699
	return notify_on_release(cgrp);
3700 3701
}

3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713
static int cgroup_write_notify_on_release(struct cgroup *cgrp,
					  struct cftype *cft,
					  u64 val)
{
	clear_bit(CGRP_RELEASABLE, &cgrp->flags);
	if (val)
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	else
		clear_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);
	return 0;
}

3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
static void cgroup_event_remove(struct work_struct *work)
{
	struct cgroup_event *event = container_of(work, struct cgroup_event,
			remove);
	struct cgroup *cgrp = event->cgrp;

3725 3726
	remove_wait_queue(event->wqh, &event->wait);

3727 3728
	event->cft->unregister_event(cgrp, event->cft, event->eventfd);

3729 3730 3731
	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

3732 3733
	eventfd_ctx_put(event->eventfd);
	kfree(event);
3734
	dput(cgrp->dentry);
3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751
}

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

	if (flags & POLLHUP) {
		/*
3752 3753 3754 3755 3756 3757 3758
		 * If the event has been detached at cgroup removal, we
		 * can simply return knowing the other side will cleanup
		 * for us.
		 *
		 * We can't race against event freeing since the other
		 * side will require wqh->lock via remove_wait_queue(),
		 * which we hold.
3759
		 */
3760 3761 3762 3763 3764 3765 3766 3767 3768 3769
		spin_lock(&cgrp->event_list_lock);
		if (!list_empty(&event->list)) {
			list_del_init(&event->list);
			/*
			 * We are in atomic context, but cgroup_event_remove()
			 * may sleep, so we have to call it in workqueue.
			 */
			schedule_work(&event->remove);
		}
		spin_unlock(&cgrp->event_list_lock);
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794
	}

	return 0;
}

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

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

/*
 * Parse input and register new cgroup event handler.
 *
 * Input must be in format '<event_fd> <control_fd> <args>'.
 * Interpretation of args is defined by control file implementation.
 */
static int cgroup_write_event_control(struct cgroup *cgrp, struct cftype *cft,
				      const char *buffer)
{
	struct cgroup_event *event = NULL;
3795
	struct cgroup *cgrp_cfile;
3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
	unsigned int efd, cfd;
	struct file *efile = NULL;
	struct file *cfile = NULL;
	char *endp;
	int ret;

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

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

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

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

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

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

	/* the process need read permission on control file */
A
Al Viro 已提交
3840
	/* AV: shouldn't we check that it's been opened for read instead? */
A
Al Viro 已提交
3841
	ret = inode_permission(file_inode(cfile), MAY_READ);
3842 3843 3844 3845 3846 3847 3848 3849 3850
	if (ret < 0)
		goto fail;

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

3851 3852 3853 3854 3855 3856 3857 3858 3859 3860
	/*
	 * The file to be monitored must be in the same cgroup as
	 * cgroup.event_control is.
	 */
	cgrp_cfile = __d_cgrp(cfile->f_dentry->d_parent);
	if (cgrp_cfile != cgrp) {
		ret = -EINVAL;
		goto fail;
	}

3861 3862 3863 3864 3865 3866 3867 3868 3869 3870
	if (!event->cft->register_event || !event->cft->unregister_event) {
		ret = -EINVAL;
		goto fail;
	}

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

3871 3872 3873 3874 3875 3876 3877
	/*
	 * Events should be removed after rmdir of cgroup directory, but before
	 * destroying subsystem state objects. Let's take reference to cgroup
	 * directory dentry to do that.
	 */
	dget(cgrp->dentry);

3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901
	spin_lock(&cgrp->event_list_lock);
	list_add(&event->list, &cgrp->event_list);
	spin_unlock(&cgrp->event_list_lock);

	fput(cfile);
	fput(efile);

	return 0;

fail:
	if (cfile)
		fput(cfile);

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

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

	kfree(event);

	return ret;
}

3902 3903 3904
static u64 cgroup_clone_children_read(struct cgroup *cgrp,
				    struct cftype *cft)
{
3905
	return test_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
3906 3907 3908 3909 3910 3911 3912
}

static int cgroup_clone_children_write(struct cgroup *cgrp,
				     struct cftype *cft,
				     u64 val)
{
	if (val)
3913
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
3914
	else
3915
		clear_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
3916 3917 3918
	return 0;
}

3919 3920 3921
/*
 * for the common functions, 'private' gives the type of file
 */
3922 3923
/* for hysterical raisins, we can't put this on the older files */
#define CGROUP_FILE_GENERIC_PREFIX "cgroup."
3924 3925 3926 3927
static struct cftype files[] = {
	{
		.name = "tasks",
		.open = cgroup_tasks_open,
3928
		.write_u64 = cgroup_tasks_write,
3929
		.release = cgroup_pidlist_release,
L
Li Zefan 已提交
3930
		.mode = S_IRUGO | S_IWUSR,
3931
	},
3932 3933 3934
	{
		.name = CGROUP_FILE_GENERIC_PREFIX "procs",
		.open = cgroup_procs_open,
B
Ben Blum 已提交
3935
		.write_u64 = cgroup_procs_write,
3936
		.release = cgroup_pidlist_release,
B
Ben Blum 已提交
3937
		.mode = S_IRUGO | S_IWUSR,
3938
	},
3939 3940
	{
		.name = "notify_on_release",
3941
		.read_u64 = cgroup_read_notify_on_release,
3942
		.write_u64 = cgroup_write_notify_on_release,
3943
	},
3944 3945 3946 3947 3948
	{
		.name = CGROUP_FILE_GENERIC_PREFIX "event_control",
		.write_string = cgroup_write_event_control,
		.mode = S_IWUGO,
	},
3949 3950 3951 3952 3953
	{
		.name = "cgroup.clone_children",
		.read_u64 = cgroup_clone_children_read,
		.write_u64 = cgroup_clone_children_write,
	},
3954 3955 3956 3957 3958 3959 3960
	{
		.name = "release_agent",
		.flags = CFTYPE_ONLY_ON_ROOT,
		.read_seq_string = cgroup_release_agent_show,
		.write_string = cgroup_release_agent_write,
		.max_write_len = PATH_MAX,
	},
T
Tejun Heo 已提交
3961
	{ }	/* terminate */
3962 3963
};

3964 3965 3966 3967 3968 3969 3970 3971
/**
 * cgroup_populate_dir - selectively creation of files in a directory
 * @cgrp: target cgroup
 * @base_files: true if the base files should be added
 * @subsys_mask: mask of the subsystem ids whose files should be added
 */
static int cgroup_populate_dir(struct cgroup *cgrp, bool base_files,
			       unsigned long subsys_mask)
3972 3973 3974 3975
{
	int err;
	struct cgroup_subsys *ss;

3976 3977 3978 3979 3980
	if (base_files) {
		err = cgroup_addrm_files(cgrp, NULL, files, true);
		if (err < 0)
			return err;
	}
3981

3982
	/* process cftsets of each subsystem */
3983
	for_each_subsys(cgrp->root, ss) {
3984
		struct cftype_set *set;
3985 3986
		if (!test_bit(ss->subsys_id, &subsys_mask))
			continue;
3987

T
Tejun Heo 已提交
3988
		list_for_each_entry(set, &ss->cftsets, node)
3989
			cgroup_addrm_files(cgrp, ss, set->cfts, true);
3990
	}
3991

K
KAMEZAWA Hiroyuki 已提交
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002
	/* This cgroup is ready now */
	for_each_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
		/*
		 * Update id->css pointer and make this css visible from
		 * CSS ID functions. This pointer will be dereferened
		 * from RCU-read-side without locks.
		 */
		if (css->id)
			rcu_assign_pointer(css->id->css, css);
	}
4003 4004 4005 4006

	return 0;
}

4007 4008 4009 4010
static void css_dput_fn(struct work_struct *work)
{
	struct cgroup_subsys_state *css =
		container_of(work, struct cgroup_subsys_state, dput_work);
4011 4012
	struct dentry *dentry = css->cgroup->dentry;
	struct super_block *sb = dentry->d_sb;
4013

4014 4015 4016
	atomic_inc(&sb->s_active);
	dput(dentry);
	deactivate_super(sb);
4017 4018
}

4019 4020
static void init_cgroup_css(struct cgroup_subsys_state *css,
			       struct cgroup_subsys *ss,
4021
			       struct cgroup *cgrp)
4022
{
4023
	css->cgroup = cgrp;
P
Paul Menage 已提交
4024
	atomic_set(&css->refcnt, 1);
4025
	css->flags = 0;
K
KAMEZAWA Hiroyuki 已提交
4026
	css->id = NULL;
4027
	if (cgrp == dummytop)
4028
		css->flags |= CSS_ROOT;
4029 4030
	BUG_ON(cgrp->subsys[ss->subsys_id]);
	cgrp->subsys[ss->subsys_id] = css;
4031 4032

	/*
4033 4034 4035 4036
	 * css holds an extra ref to @cgrp->dentry which is put on the last
	 * css_put().  dput() requires process context, which css_put() may
	 * be called without.  @css->dput_work will be used to invoke
	 * dput() asynchronously from css_put().
4037 4038
	 */
	INIT_WORK(&css->dput_work, css_dput_fn);
4039 4040
}

T
Tejun Heo 已提交
4041 4042
/* invoke ->post_create() on a new CSS and mark it online if successful */
static int online_css(struct cgroup_subsys *ss, struct cgroup *cgrp)
4043
{
T
Tejun Heo 已提交
4044 4045
	int ret = 0;

4046 4047
	lockdep_assert_held(&cgroup_mutex);

4048 4049
	if (ss->css_online)
		ret = ss->css_online(cgrp);
T
Tejun Heo 已提交
4050 4051 4052
	if (!ret)
		cgrp->subsys[ss->subsys_id]->flags |= CSS_ONLINE;
	return ret;
4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065
}

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

	lockdep_assert_held(&cgroup_mutex);

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

4066
	if (ss->css_offline)
4067
		ss->css_offline(cgrp);
4068 4069 4070 4071

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

4072
/*
L
Li Zefan 已提交
4073 4074 4075 4076
 * cgroup_create - create a cgroup
 * @parent: cgroup that will be parent of the new cgroup
 * @dentry: dentry of the new cgroup
 * @mode: mode to set on new inode
4077
 *
L
Li Zefan 已提交
4078
 * Must be called with the mutex on the parent inode held
4079 4080
 */
static long cgroup_create(struct cgroup *parent, struct dentry *dentry,
A
Al Viro 已提交
4081
			     umode_t mode)
4082
{
4083
	struct cgroup *cgrp;
4084
	struct cgroup_name *name;
4085 4086 4087 4088 4089
	struct cgroupfs_root *root = parent->root;
	int err = 0;
	struct cgroup_subsys *ss;
	struct super_block *sb = root->sb;

T
Tejun Heo 已提交
4090
	/* allocate the cgroup and its ID, 0 is reserved for the root */
4091 4092
	cgrp = kzalloc(sizeof(*cgrp), GFP_KERNEL);
	if (!cgrp)
4093 4094
		return -ENOMEM;

4095 4096 4097 4098 4099
	name = cgroup_alloc_name(dentry);
	if (!name)
		goto err_free_cgrp;
	rcu_assign_pointer(cgrp->name, name);

T
Tejun Heo 已提交
4100 4101
	cgrp->id = ida_simple_get(&root->cgroup_ida, 1, 0, GFP_KERNEL);
	if (cgrp->id < 0)
4102
		goto err_free_name;
T
Tejun Heo 已提交
4103

4104 4105 4106 4107 4108 4109 4110 4111 4112
	/*
	 * 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
4113
		goto err_free_id;
4114 4115
	}

4116 4117 4118 4119 4120 4121 4122
	/* Grab a reference on the superblock so the hierarchy doesn't
	 * get deleted on unmount if there are child cgroups.  This
	 * can be done outside cgroup_mutex, since the sb can't
	 * disappear while someone has an open control file on the
	 * fs */
	atomic_inc(&sb->s_active);

4123
	init_cgroup_housekeeping(cgrp);
4124

4125 4126 4127
	dentry->d_fsdata = cgrp;
	cgrp->dentry = dentry;

4128 4129 4130
	cgrp->parent = parent;
	cgrp->root = parent->root;
	cgrp->top_cgroup = parent->top_cgroup;
4131

4132 4133 4134
	if (notify_on_release(parent))
		set_bit(CGRP_NOTIFY_ON_RELEASE, &cgrp->flags);

4135 4136
	if (test_bit(CGRP_CPUSET_CLONE_CHILDREN, &parent->flags))
		set_bit(CGRP_CPUSET_CLONE_CHILDREN, &cgrp->flags);
4137

4138
	for_each_subsys(root, ss) {
4139
		struct cgroup_subsys_state *css;
4140

4141
		css = ss->css_alloc(cgrp);
4142 4143
		if (IS_ERR(css)) {
			err = PTR_ERR(css);
4144
			goto err_free_all;
4145
		}
4146
		init_cgroup_css(css, ss, cgrp);
4147 4148 4149
		if (ss->use_id) {
			err = alloc_css_id(ss, parent, cgrp);
			if (err)
4150
				goto err_free_all;
4151
		}
4152 4153
	}

4154 4155 4156 4157 4158
	/*
	 * Create directory.  cgroup_create_file() returns with the new
	 * directory locked on success so that it can be populated without
	 * dropping cgroup_mutex.
	 */
T
Tejun Heo 已提交
4159
	err = cgroup_create_file(dentry, S_IFDIR | mode, sb);
4160
	if (err < 0)
4161
		goto err_free_all;
4162
	lockdep_assert_held(&dentry->d_inode->i_mutex);
4163

4164 4165 4166 4167
	/* allocation complete, commit to creation */
	list_add_tail(&cgrp->allcg_node, &root->allcg_list);
	list_add_tail_rcu(&cgrp->sibling, &cgrp->parent->children);
	root->number_of_cgroups++;
T
Tejun Heo 已提交
4168

T
Tejun Heo 已提交
4169 4170
	/* each css holds a ref to the cgroup's dentry */
	for_each_subsys(root, ss)
4171
		dget(dentry);
4172

T
Tejun Heo 已提交
4173 4174 4175 4176 4177
	/* creation succeeded, notify subsystems */
	for_each_subsys(root, ss) {
		err = online_css(ss, cgrp);
		if (err)
			goto err_destroy;
4178 4179 4180 4181 4182 4183 4184 4185 4186

		if (ss->broken_hierarchy && !ss->warned_broken_hierarchy &&
		    parent->parent) {
			pr_warning("cgroup: %s (%d) created nested cgroup for controller \"%s\" which has incomplete hierarchy support. Nested cgroups may change behavior in the future.\n",
				   current->comm, current->pid, ss->name);
			if (!strcmp(ss->name, "memory"))
				pr_warning("cgroup: \"memory\" requires setting use_hierarchy to 1 on the root.\n");
			ss->warned_broken_hierarchy = true;
		}
4187 4188
	}

4189
	err = cgroup_populate_dir(cgrp, true, root->subsys_mask);
4190 4191
	if (err)
		goto err_destroy;
4192 4193

	mutex_unlock(&cgroup_mutex);
4194
	mutex_unlock(&cgrp->dentry->d_inode->i_mutex);
4195 4196 4197

	return 0;

4198
err_free_all:
4199
	for_each_subsys(root, ss) {
4200
		if (cgrp->subsys[ss->subsys_id])
4201
			ss->css_free(cgrp);
4202 4203 4204 4205
	}
	mutex_unlock(&cgroup_mutex);
	/* Release the reference count that we took on the superblock */
	deactivate_super(sb);
T
Tejun Heo 已提交
4206 4207
err_free_id:
	ida_simple_remove(&root->cgroup_ida, cgrp->id);
4208 4209
err_free_name:
	kfree(rcu_dereference_raw(cgrp->name));
4210
err_free_cgrp:
4211
	kfree(cgrp);
4212
	return err;
4213 4214 4215 4216 4217 4218

err_destroy:
	cgroup_destroy_locked(cgrp);
	mutex_unlock(&cgroup_mutex);
	mutex_unlock(&dentry->d_inode->i_mutex);
	return err;
4219 4220
}

4221
static int cgroup_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
4222 4223 4224 4225 4226 4227 4228
{
	struct cgroup *c_parent = dentry->d_parent->d_fsdata;

	/* the vfs holds inode->i_mutex already */
	return cgroup_create(c_parent, dentry, mode | S_IFDIR);
}

4229 4230
static int cgroup_destroy_locked(struct cgroup *cgrp)
	__releases(&cgroup_mutex) __acquires(&cgroup_mutex)
4231
{
4232 4233
	struct dentry *d = cgrp->dentry;
	struct cgroup *parent = cgrp->parent;
4234
	struct cgroup_event *event, *tmp;
4235
	struct cgroup_subsys *ss;
4236

4237 4238 4239 4240
	lockdep_assert_held(&d->d_inode->i_mutex);
	lockdep_assert_held(&cgroup_mutex);

	if (atomic_read(&cgrp->count) || !list_empty(&cgrp->children))
4241
		return -EBUSY;
L
Li Zefan 已提交
4242

4243
	/*
4244 4245 4246 4247
	 * Block new css_tryget() by deactivating refcnt and mark @cgrp
	 * removed.  This makes future css_tryget() and child creation
	 * attempts fail thus maintaining the removal conditions verified
	 * above.
4248
	 */
4249 4250
	for_each_subsys(cgrp->root, ss) {
		struct cgroup_subsys_state *css = cgrp->subsys[ss->subsys_id];
4251

4252 4253
		WARN_ON(atomic_read(&css->refcnt) < 0);
		atomic_add(CSS_DEACT_BIAS, &css->refcnt);
4254
	}
4255
	set_bit(CGRP_REMOVED, &cgrp->flags);
4256

4257
	/* tell subsystems to initate destruction */
4258
	for_each_subsys(cgrp->root, ss)
4259
		offline_css(ss, cgrp);
4260 4261 4262 4263 4264 4265 4266 4267

	/*
	 * Put all the base refs.  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.
	 */
T
Tejun Heo 已提交
4268 4269
	for_each_subsys(cgrp->root, ss)
		css_put(cgrp->subsys[ss->subsys_id]);
4270

4271
	raw_spin_lock(&release_list_lock);
4272
	if (!list_empty(&cgrp->release_list))
4273
		list_del_init(&cgrp->release_list);
4274
	raw_spin_unlock(&release_list_lock);
4275 4276

	/* delete this cgroup from parent->children */
4277
	list_del_rcu(&cgrp->sibling);
4278 4279
	list_del_init(&cgrp->allcg_node);

4280
	dget(d);
4281 4282 4283
	cgroup_d_remove_dir(d);
	dput(d);

4284
	set_bit(CGRP_RELEASABLE, &parent->flags);
4285 4286
	check_for_release(parent);

4287 4288 4289
	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
4290
	 * directory to avoid race between userspace and kernelspace.
4291 4292
	 */
	spin_lock(&cgrp->event_list_lock);
4293
	list_for_each_entry_safe(event, tmp, &cgrp->event_list, list) {
4294
		list_del_init(&event->list);
4295 4296
		schedule_work(&event->remove);
	}
4297
	spin_unlock(&cgrp->event_list_lock);
4298

4299 4300 4301
	return 0;
}

4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312
static int cgroup_rmdir(struct inode *unused_dir, struct dentry *dentry)
{
	int ret;

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

	return ret;
}

4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326
static void __init_or_module cgroup_init_cftsets(struct cgroup_subsys *ss)
{
	INIT_LIST_HEAD(&ss->cftsets);

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

4327
static void __init cgroup_init_subsys(struct cgroup_subsys *ss)
4328 4329
{
	struct cgroup_subsys_state *css;
D
Diego Calleja 已提交
4330 4331

	printk(KERN_INFO "Initializing cgroup subsys %s\n", ss->name);
4332

4333 4334
	mutex_lock(&cgroup_mutex);

4335 4336 4337
	/* init base cftset */
	cgroup_init_cftsets(ss);

4338
	/* Create the top cgroup state for this subsystem */
4339
	list_add(&ss->sibling, &rootnode.subsys_list);
4340
	ss->root = &rootnode;
4341
	css = ss->css_alloc(dummytop);
4342 4343 4344 4345
	/* We don't handle early failures gracefully */
	BUG_ON(IS_ERR(css));
	init_cgroup_css(css, ss, dummytop);

L
Li Zefan 已提交
4346
	/* Update the init_css_set to contain a subsys
4347
	 * pointer to this state - since the subsystem is
L
Li Zefan 已提交
4348 4349
	 * newly registered, all tasks and hence the
	 * init_css_set is in the subsystem's top cgroup. */
4350
	init_css_set.subsys[ss->subsys_id] = css;
4351 4352 4353

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

L
Li Zefan 已提交
4354 4355 4356 4357 4358
	/* 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));

4359
	ss->active = 1;
T
Tejun Heo 已提交
4360
	BUG_ON(online_css(ss, dummytop));
4361

4362 4363
	mutex_unlock(&cgroup_mutex);

4364 4365 4366 4367 4368 4369 4370 4371 4372 4373
	/* this function shouldn't be used with modular subsystems, since they
	 * need to register a subsys_id, among other things */
	BUG_ON(ss->module);
}

/**
 * cgroup_load_subsys: load and register a modular subsystem at runtime
 * @ss: the subsystem to load
 *
 * This function should be called in a modular subsystem's initcall. If the
T
Thomas Weber 已提交
4374
 * subsystem is built as a module, it will be assigned a new subsys_id and set
4375 4376 4377 4378 4379 4380
 * up for use. If the subsystem is built-in anyway, work is delegated to the
 * simpler cgroup_init_subsys.
 */
int __init_or_module cgroup_load_subsys(struct cgroup_subsys *ss)
{
	struct cgroup_subsys_state *css;
4381
	int i, ret;
4382
	struct hlist_node *tmp;
4383 4384
	struct css_set *cg;
	unsigned long key;
4385 4386 4387

	/* check name and function validity */
	if (ss->name == NULL || strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN ||
4388
	    ss->css_alloc == NULL || ss->css_free == NULL)
4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404
		return -EINVAL;

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

	/*
	 * an optionally modular subsystem is built-in: we want to do nothing,
	 * since cgroup_init_subsys will have already taken care of it.
	 */
	if (ss->module == NULL) {
4405
		/* a sanity check */
4406 4407 4408 4409
		BUG_ON(subsys[ss->subsys_id] != ss);
		return 0;
	}

4410 4411 4412
	/* init base cftset */
	cgroup_init_cftsets(ss);

4413
	mutex_lock(&cgroup_mutex);
4414
	subsys[ss->subsys_id] = ss;
4415 4416

	/*
4417 4418 4419
	 * no ss->css_alloc seems to need anything important in the ss
	 * struct, so this can happen first (i.e. before the rootnode
	 * attachment).
4420
	 */
4421
	css = ss->css_alloc(dummytop);
4422 4423
	if (IS_ERR(css)) {
		/* failure case - need to deassign the subsys[] slot. */
4424
		subsys[ss->subsys_id] = NULL;
4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435
		mutex_unlock(&cgroup_mutex);
		return PTR_ERR(css);
	}

	list_add(&ss->sibling, &rootnode.subsys_list);
	ss->root = &rootnode;

	/* our new subsystem will be attached to the dummy hierarchy. */
	init_cgroup_css(css, ss, dummytop);
	/* init_idr must be after init_cgroup_css because it sets css->id. */
	if (ss->use_id) {
4436 4437 4438
		ret = cgroup_init_idr(ss, css);
		if (ret)
			goto err_unload;
4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449
	}

	/*
	 * Now we need to entangle the css into the existing css_sets. unlike
	 * in cgroup_init_subsys, there are now multiple css_sets, so each one
	 * will need a new pointer to it; done by iterating the css_set_table.
	 * furthermore, modifying the existing css_sets will corrupt the hash
	 * table state, so each changed css_set will need its hash recomputed.
	 * this is all done under the css_set_lock.
	 */
	write_lock(&css_set_lock);
4450
	hash_for_each_safe(css_set_table, i, tmp, cg, hlist) {
4451 4452 4453 4454 4455 4456 4457 4458 4459
		/* skip entries that we already rehashed */
		if (cg->subsys[ss->subsys_id])
			continue;
		/* remove existing entry */
		hash_del(&cg->hlist);
		/* set new value */
		cg->subsys[ss->subsys_id] = css;
		/* recompute hash and restore entry */
		key = css_set_hash(cg->subsys);
4460
		hash_add(css_set_table, &cg->hlist, key);
4461 4462 4463 4464
	}
	write_unlock(&css_set_lock);

	ss->active = 1;
T
Tejun Heo 已提交
4465 4466 4467
	ret = online_css(ss, dummytop);
	if (ret)
		goto err_unload;
4468

4469 4470 4471
	/* success! */
	mutex_unlock(&cgroup_mutex);
	return 0;
4472 4473 4474 4475 4476 4477

err_unload:
	mutex_unlock(&cgroup_mutex);
	/* @ss can't be mounted here as try_module_get() would fail */
	cgroup_unload_subsys(ss);
	return ret;
4478
}
4479
EXPORT_SYMBOL_GPL(cgroup_load_subsys);
4480

B
Ben Blum 已提交
4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502
/**
 * cgroup_unload_subsys: unload a modular subsystem
 * @ss: the subsystem to unload
 *
 * This function should be called in a modular subsystem's exitcall. When this
 * function is invoked, the refcount on the subsystem's module will be 0, so
 * the subsystem will not be attached to any hierarchy.
 */
void cgroup_unload_subsys(struct cgroup_subsys *ss)
{
	struct cg_cgroup_link *link;

	BUG_ON(ss->module == NULL);

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

	mutex_lock(&cgroup_mutex);
4503

4504
	offline_css(ss, dummytop);
4505 4506
	ss->active = 0;

T
Tejun Heo 已提交
4507
	if (ss->use_id)
4508 4509
		idr_destroy(&ss->idr);

B
Ben Blum 已提交
4510 4511 4512 4513
	/* deassign the subsys_id */
	subsys[ss->subsys_id] = NULL;

	/* remove subsystem from rootnode's list of subsystems */
4514
	list_del_init(&ss->sibling);
B
Ben Blum 已提交
4515 4516 4517 4518 4519 4520 4521 4522

	/*
	 * disentangle the css from all css_sets attached to the dummytop. as
	 * in loading, we need to pay our respects to the hashtable gods.
	 */
	write_lock(&css_set_lock);
	list_for_each_entry(link, &dummytop->css_sets, cgrp_link_list) {
		struct css_set *cg = link->cg;
4523
		unsigned long key;
B
Ben Blum 已提交
4524

4525
		hash_del(&cg->hlist);
B
Ben Blum 已提交
4526
		cg->subsys[ss->subsys_id] = NULL;
4527 4528
		key = css_set_hash(cg->subsys);
		hash_add(css_set_table, &cg->hlist, key);
B
Ben Blum 已提交
4529 4530 4531 4532
	}
	write_unlock(&css_set_lock);

	/*
4533 4534 4535 4536
	 * remove subsystem's css from the dummytop and free it - need to
	 * free before marking as null because ss->css_free needs the
	 * cgrp->subsys pointer to find their state. note that this also
	 * takes care of freeing the css_id.
B
Ben Blum 已提交
4537
	 */
4538
	ss->css_free(dummytop);
B
Ben Blum 已提交
4539 4540 4541 4542 4543 4544
	dummytop->subsys[ss->subsys_id] = NULL;

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

4545
/**
L
Li Zefan 已提交
4546 4547 4548 4549
 * cgroup_init_early - cgroup initialization at system boot
 *
 * Initialize cgroups at system boot, and initialize any
 * subsystems that request early init.
4550 4551 4552 4553
 */
int __init cgroup_init_early(void)
{
	int i;
4554
	atomic_set(&init_css_set.refcount, 1);
4555 4556
	INIT_LIST_HEAD(&init_css_set.cg_links);
	INIT_LIST_HEAD(&init_css_set.tasks);
4557
	INIT_HLIST_NODE(&init_css_set.hlist);
4558
	css_set_count = 1;
4559
	init_cgroup_root(&rootnode);
4560 4561 4562 4563
	root_count = 1;
	init_task.cgroups = &init_css_set;

	init_css_set_link.cg = &init_css_set;
4564
	init_css_set_link.cgrp = dummytop;
4565
	list_add(&init_css_set_link.cgrp_link_list,
4566 4567 4568
		 &rootnode.top_cgroup.css_sets);
	list_add(&init_css_set_link.cg_link_list,
		 &init_css_set.cg_links);
4569

4570
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
4571 4572
		struct cgroup_subsys *ss = subsys[i];

4573 4574 4575 4576
		/* at bootup time, we don't worry about modular subsystems */
		if (!ss || ss->module)
			continue;

4577 4578
		BUG_ON(!ss->name);
		BUG_ON(strlen(ss->name) > MAX_CGROUP_TYPE_NAMELEN);
4579 4580
		BUG_ON(!ss->css_alloc);
		BUG_ON(!ss->css_free);
4581
		if (ss->subsys_id != i) {
D
Diego Calleja 已提交
4582
			printk(KERN_ERR "cgroup: Subsys %s id == %d\n",
4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593
			       ss->name, ss->subsys_id);
			BUG();
		}

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

/**
L
Li Zefan 已提交
4594 4595 4596 4597
 * cgroup_init - cgroup initialization
 *
 * Register cgroup filesystem and /proc file, and initialize
 * any subsystems that didn't request early init.
4598 4599 4600 4601 4602
 */
int __init cgroup_init(void)
{
	int err;
	int i;
4603
	unsigned long key;
4604 4605 4606 4607

	err = bdi_init(&cgroup_backing_dev_info);
	if (err)
		return err;
4608

4609
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
4610
		struct cgroup_subsys *ss = subsys[i];
4611 4612 4613 4614

		/* at bootup time, we don't worry about modular subsystems */
		if (!ss || ss->module)
			continue;
4615 4616
		if (!ss->early_init)
			cgroup_init_subsys(ss);
K
KAMEZAWA Hiroyuki 已提交
4617
		if (ss->use_id)
4618
			cgroup_init_idr(ss, init_css_set.subsys[ss->subsys_id]);
4619 4620
	}

4621
	/* Add init_css_set to the hash table */
4622 4623
	key = css_set_hash(init_css_set.subsys);
	hash_add(css_set_table, &init_css_set.hlist, key);
4624
	BUG_ON(!init_root_id(&rootnode));
4625 4626 4627 4628 4629 4630 4631

	cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
	if (!cgroup_kobj) {
		err = -ENOMEM;
		goto out;
	}

4632
	err = register_filesystem(&cgroup_fs_type);
4633 4634
	if (err < 0) {
		kobject_put(cgroup_kobj);
4635
		goto out;
4636
	}
4637

L
Li Zefan 已提交
4638
	proc_create("cgroups", 0, NULL, &proc_cgroupstats_operations);
4639

4640
out:
4641 4642 4643
	if (err)
		bdi_destroy(&cgroup_backing_dev_info);

4644 4645
	return err;
}
4646

4647 4648 4649 4650 4651 4652
/*
 * proc_cgroup_show()
 *  - Print task's cgroup paths into seq_file, one line for each hierarchy
 *  - Used for /proc/<pid>/cgroup.
 *  - No need to task_lock(tsk) on this tsk->cgroup reference, as it
 *    doesn't really matter if tsk->cgroup changes after we read it,
4653
 *    and we take cgroup_mutex, keeping cgroup_attach_task() from changing it
4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682
 *    anyway.  No need to check that tsk->cgroup != NULL, thanks to
 *    the_top_cgroup_hack in cgroup_exit(), which sets an exiting tasks
 *    cgroup to top_cgroup.
 */

/* TODO: Use a proper seq_file iterator */
static int proc_cgroup_show(struct seq_file *m, void *v)
{
	struct pid *pid;
	struct task_struct *tsk;
	char *buf;
	int retval;
	struct cgroupfs_root *root;

	retval = -ENOMEM;
	buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!buf)
		goto out;

	retval = -ESRCH;
	pid = m->private;
	tsk = get_pid_task(pid, PIDTYPE_PID);
	if (!tsk)
		goto out_free;

	retval = 0;

	mutex_lock(&cgroup_mutex);

4683
	for_each_active_root(root) {
4684
		struct cgroup_subsys *ss;
4685
		struct cgroup *cgrp;
4686 4687
		int count = 0;

4688
		seq_printf(m, "%d:", root->hierarchy_id);
4689 4690
		for_each_subsys(root, ss)
			seq_printf(m, "%s%s", count++ ? "," : "", ss->name);
4691 4692 4693
		if (strlen(root->name))
			seq_printf(m, "%sname=%s", count ? "," : "",
				   root->name);
4694
		seq_putc(m, ':');
4695
		cgrp = task_cgroup_from_root(tsk, root);
4696
		retval = cgroup_path(cgrp, buf, PAGE_SIZE);
4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717
		if (retval < 0)
			goto out_unlock;
		seq_puts(m, buf);
		seq_putc(m, '\n');
	}

out_unlock:
	mutex_unlock(&cgroup_mutex);
	put_task_struct(tsk);
out_free:
	kfree(buf);
out:
	return retval;
}

static int cgroup_open(struct inode *inode, struct file *file)
{
	struct pid *pid = PROC_I(inode)->pid;
	return single_open(file, proc_cgroup_show, pid);
}

4718
const struct file_operations proc_cgroup_operations = {
4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729
	.open		= cgroup_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};

/* Display information about each subsystem and each hierarchy */
static int proc_cgroupstats_show(struct seq_file *m, void *v)
{
	int i;

4730
	seq_puts(m, "#subsys_name\thierarchy\tnum_cgroups\tenabled\n");
B
Ben Blum 已提交
4731 4732 4733 4734 4735
	/*
	 * 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.
	 */
4736 4737 4738
	mutex_lock(&cgroup_mutex);
	for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
		struct cgroup_subsys *ss = subsys[i];
B
Ben Blum 已提交
4739 4740
		if (ss == NULL)
			continue;
4741 4742
		seq_printf(m, "%s\t%d\t%d\t%d\n",
			   ss->name, ss->root->hierarchy_id,
4743
			   ss->root->number_of_cgroups, !ss->disabled);
4744 4745 4746 4747 4748 4749 4750
	}
	mutex_unlock(&cgroup_mutex);
	return 0;
}

static int cgroupstats_open(struct inode *inode, struct file *file)
{
A
Al Viro 已提交
4751
	return single_open(file, proc_cgroupstats_show, NULL);
4752 4753
}

4754
static const struct file_operations proc_cgroupstats_operations = {
4755 4756 4757 4758 4759 4760
	.open = cgroupstats_open,
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

4761 4762
/**
 * cgroup_fork - attach newly forked task to its parents cgroup.
L
Li Zefan 已提交
4763
 * @child: pointer to task_struct of forking parent process.
4764 4765 4766 4767 4768
 *
 * Description: A task inherits its parent's cgroup at fork().
 *
 * A pointer to the shared css_set was automatically copied in
 * fork.c by dup_task_struct().  However, we ignore that copy, since
4769 4770 4771 4772
 * it was not made under the protection of RCU or cgroup_mutex, so
 * might no longer be a valid cgroup pointer.  cgroup_attach_task() might
 * have already changed current->cgroups, allowing the previously
 * referenced cgroup group to be removed and freed.
4773 4774 4775 4776 4777 4778
 *
 * At the point that cgroup_fork() is called, 'current' is the parent
 * task, and the passed argument 'child' points to the child task.
 */
void cgroup_fork(struct task_struct *child)
{
4779
	task_lock(current);
4780 4781
	child->cgroups = current->cgroups;
	get_css_set(child->cgroups);
4782
	task_unlock(current);
4783
	INIT_LIST_HEAD(&child->cg_list);
4784 4785
}

4786
/**
L
Li Zefan 已提交
4787 4788 4789
 * cgroup_post_fork - called on a new task after adding it to the task list
 * @child: the task in question
 *
4790 4791 4792 4793 4794
 * Adds the task to the list running through its css_set if necessary and
 * call the subsystem fork() callbacks.  Has to be after the task is
 * visible on the task list in case we race with the first call to
 * cgroup_iter_start() - to guarantee that the new task ends up on its
 * list.
L
Li Zefan 已提交
4795
 */
4796 4797
void cgroup_post_fork(struct task_struct *child)
{
4798 4799
	int i;

4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810
	/*
	 * use_task_css_set_links is set to 1 before we walk the tasklist
	 * under the tasklist_lock and we read it here after we added the child
	 * to the tasklist under the tasklist_lock as well. If the child wasn't
	 * yet in the tasklist when we walked through it from
	 * cgroup_enable_task_cg_lists(), then use_task_css_set_links value
	 * should be visible now due to the paired locking and barriers implied
	 * by LOCK/UNLOCK: it is written before the tasklist_lock unlock
	 * in cgroup_enable_task_cg_lists() and read here after the tasklist_lock
	 * lock on fork.
	 */
4811 4812
	if (use_task_css_set_links) {
		write_lock(&css_set_lock);
4813 4814
		task_lock(child);
		if (list_empty(&child->cg_list))
4815
			list_add(&child->cg_list, &child->cgroups->tasks);
4816
		task_unlock(child);
4817 4818
		write_unlock(&css_set_lock);
	}
4819 4820 4821 4822 4823 4824 4825

	/*
	 * 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) {
4826 4827 4828 4829 4830 4831 4832 4833 4834
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, and the builtin section of the subsys
		 * array is immutable, so we don't need to lock the
		 * subsys array here. On the other hand, modular section
		 * of the array can be freed at module unload, so we
		 * can't touch that.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
4835 4836 4837 4838 4839 4840
			struct cgroup_subsys *ss = subsys[i];

			if (ss->fork)
				ss->fork(child);
		}
	}
4841
}
4842

4843 4844 4845
/**
 * cgroup_exit - detach cgroup from exiting task
 * @tsk: pointer to task_struct of exiting process
L
Li Zefan 已提交
4846
 * @run_callback: run exit callbacks?
4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874
 *
 * Description: Detach cgroup from @tsk and release it.
 *
 * Note that cgroups marked notify_on_release force every task in
 * them to take the global cgroup_mutex mutex when exiting.
 * This could impact scaling on very large systems.  Be reluctant to
 * use notify_on_release cgroups where very high task exit scaling
 * is required on large systems.
 *
 * the_top_cgroup_hack:
 *
 *    Set the exiting tasks cgroup to the root cgroup (top_cgroup).
 *
 *    We call cgroup_exit() while the task is still competent to
 *    handle notify_on_release(), then leave the task attached to the
 *    root cgroup in each hierarchy for the remainder of its exit.
 *
 *    To do this properly, we would increment the reference count on
 *    top_cgroup, and near the very end of the kernel/exit.c do_exit()
 *    code we would add a second cgroup function call, to drop that
 *    reference.  This would just create an unnecessary hot spot on
 *    the top_cgroup reference count, to no avail.
 *
 *    Normally, holding a reference to a cgroup without bumping its
 *    count is unsafe.   The cgroup could go away, or someone could
 *    attach us to a different cgroup, decrementing the count on
 *    the first cgroup that we never incremented.  But in this case,
 *    top_cgroup isn't going away, and either task has PF_EXITING set,
4875 4876
 *    which wards off any cgroup_attach_task() attempts, or task is a failed
 *    fork, never visible to cgroup_attach_task.
4877 4878 4879
 */
void cgroup_exit(struct task_struct *tsk, int run_callbacks)
{
4880
	struct css_set *cg;
4881
	int i;
4882 4883 4884 4885 4886 4887 4888 4889 4890

	/*
	 * Unlink from the css_set task list if necessary.
	 * Optimistically check cg_list before taking
	 * css_set_lock
	 */
	if (!list_empty(&tsk->cg_list)) {
		write_lock(&css_set_lock);
		if (!list_empty(&tsk->cg_list))
4891
			list_del_init(&tsk->cg_list);
4892 4893 4894
		write_unlock(&css_set_lock);
	}

4895 4896
	/* Reassign the task to the init_css_set. */
	task_lock(tsk);
4897 4898
	cg = tsk->cgroups;
	tsk->cgroups = &init_css_set;
4899 4900

	if (run_callbacks && need_forkexit_callback) {
4901 4902 4903 4904 4905
		/*
		 * fork/exit callbacks are supported only for builtin
		 * subsystems, see cgroup_post_fork() for details.
		 */
		for (i = 0; i < CGROUP_BUILTIN_SUBSYS_COUNT; i++) {
4906
			struct cgroup_subsys *ss = subsys[i];
4907

4908 4909 4910 4911
			if (ss->exit) {
				struct cgroup *old_cgrp =
					rcu_dereference_raw(cg->subsys[i])->cgroup;
				struct cgroup *cgrp = task_cgroup(tsk, i);
4912
				ss->exit(cgrp, old_cgrp, tsk);
4913 4914 4915
			}
		}
	}
4916
	task_unlock(tsk);
4917

4918
	put_css_set_taskexit(cg);
4919
}
4920

4921
static void check_for_release(struct cgroup *cgrp)
4922 4923 4924
{
	/* All of these checks rely on RCU to keep the cgroup
	 * structure alive */
4925 4926 4927 4928
	if (cgroup_is_releasable(cgrp) &&
	    !atomic_read(&cgrp->count) && list_empty(&cgrp->children)) {
		/*
		 * Control Group is currently removeable. If it's not
4929
		 * already queued for a userspace notification, queue
4930 4931
		 * it now
		 */
4932
		int need_schedule_work = 0;
4933

4934
		raw_spin_lock(&release_list_lock);
4935 4936 4937
		if (!cgroup_is_removed(cgrp) &&
		    list_empty(&cgrp->release_list)) {
			list_add(&cgrp->release_list, &release_list);
4938 4939
			need_schedule_work = 1;
		}
4940
		raw_spin_unlock(&release_list_lock);
4941 4942 4943 4944 4945
		if (need_schedule_work)
			schedule_work(&release_agent_work);
	}
}

4946
/* Caller must verify that the css is not for root cgroup */
4947 4948
bool __css_tryget(struct cgroup_subsys_state *css)
{
T
Tejun Heo 已提交
4949 4950
	while (true) {
		int t, v;
4951

T
Tejun Heo 已提交
4952 4953 4954
		v = css_refcnt(css);
		t = atomic_cmpxchg(&css->refcnt, v, v + 1);
		if (likely(t == v))
4955
			return true;
T
Tejun Heo 已提交
4956 4957
		else if (t < 0)
			return false;
4958
		cpu_relax();
T
Tejun Heo 已提交
4959
	}
4960 4961 4962 4963 4964
}
EXPORT_SYMBOL_GPL(__css_tryget);

/* Caller must verify that the css is not for root cgroup */
void __css_put(struct cgroup_subsys_state *css)
4965
{
4966
	int v;
4967

4968
	v = css_unbias_refcnt(atomic_dec_return(&css->refcnt));
4969
	if (v == 0)
4970
		schedule_work(&css->dput_work);
4971
}
B
Ben Blum 已提交
4972
EXPORT_SYMBOL_GPL(__css_put);
4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000

/*
 * 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);
5001
	raw_spin_lock(&release_list_lock);
5002 5003 5004
	while (!list_empty(&release_list)) {
		char *argv[3], *envp[3];
		int i;
5005
		char *pathbuf = NULL, *agentbuf = NULL;
5006
		struct cgroup *cgrp = list_entry(release_list.next,
5007 5008
						    struct cgroup,
						    release_list);
5009
		list_del_init(&cgrp->release_list);
5010
		raw_spin_unlock(&release_list_lock);
5011
		pathbuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
5012 5013 5014 5015 5016 5017 5018
		if (!pathbuf)
			goto continue_free;
		if (cgroup_path(cgrp, pathbuf, PAGE_SIZE) < 0)
			goto continue_free;
		agentbuf = kstrdup(cgrp->root->release_agent_path, GFP_KERNEL);
		if (!agentbuf)
			goto continue_free;
5019 5020

		i = 0;
5021 5022
		argv[i++] = agentbuf;
		argv[i++] = pathbuf;
5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036
		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);
5037 5038 5039
 continue_free:
		kfree(pathbuf);
		kfree(agentbuf);
5040
		raw_spin_lock(&release_list_lock);
5041
	}
5042
	raw_spin_unlock(&release_list_lock);
5043 5044
	mutex_unlock(&cgroup_mutex);
}
5045 5046 5047 5048 5049 5050 5051 5052 5053

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

	while ((token = strsep(&str, ",")) != NULL) {
		if (!*token)
			continue;
5054
		for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
5055 5056
			struct cgroup_subsys *ss = subsys[i];

5057 5058 5059 5060 5061 5062 5063 5064
			/*
			 * cgroup_disable, being at boot time, can't
			 * know about module subsystems, so we don't
			 * worry about them.
			 */
			if (!ss || ss->module)
				continue;

5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075
			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
5076 5077 5078 5079 5080 5081 5082 5083 5084 5085

/*
 * Functons for CSS ID.
 */

/*
 *To get ID other than 0, this should be called when !cgroup_is_removed().
 */
unsigned short css_id(struct cgroup_subsys_state *css)
{
5086 5087 5088 5089 5090 5091 5092
	struct css_id *cssid;

	/*
	 * This css_id() can return correct value when somone has refcnt
	 * on this or this is under rcu_read_lock(). Once css->id is allocated,
	 * it's unchanged until freed.
	 */
5093
	cssid = rcu_dereference_check(css->id, css_refcnt(css));
K
KAMEZAWA Hiroyuki 已提交
5094 5095 5096 5097 5098

	if (cssid)
		return cssid->id;
	return 0;
}
B
Ben Blum 已提交
5099
EXPORT_SYMBOL_GPL(css_id);
K
KAMEZAWA Hiroyuki 已提交
5100 5101 5102

unsigned short css_depth(struct cgroup_subsys_state *css)
{
5103 5104
	struct css_id *cssid;

5105
	cssid = rcu_dereference_check(css->id, css_refcnt(css));
K
KAMEZAWA Hiroyuki 已提交
5106 5107 5108 5109 5110

	if (cssid)
		return cssid->depth;
	return 0;
}
B
Ben Blum 已提交
5111
EXPORT_SYMBOL_GPL(css_depth);
K
KAMEZAWA Hiroyuki 已提交
5112

5113 5114 5115 5116 5117 5118
/**
 *  css_is_ancestor - test "root" css is an ancestor of "child"
 * @child: the css to be tested.
 * @root: the css supporsed to be an ancestor of the child.
 *
 * Returns true if "root" is an ancestor of "child" in its hierarchy. Because
5119
 * this function reads css->id, the caller must hold rcu_read_lock().
5120 5121 5122 5123 5124 5125
 * But, considering usual usage, the csses should be valid objects after test.
 * Assuming that the caller will do some action to the child if this returns
 * returns true, the caller must take "child";s reference count.
 * If "child" is valid object and this returns true, "root" is valid, too.
 */

K
KAMEZAWA Hiroyuki 已提交
5126
bool css_is_ancestor(struct cgroup_subsys_state *child,
5127
		    const struct cgroup_subsys_state *root)
K
KAMEZAWA Hiroyuki 已提交
5128
{
5129 5130
	struct css_id *child_id;
	struct css_id *root_id;
K
KAMEZAWA Hiroyuki 已提交
5131

5132
	child_id  = rcu_dereference(child->id);
5133 5134
	if (!child_id)
		return false;
5135
	root_id = rcu_dereference(root->id);
5136 5137 5138 5139 5140 5141 5142
	if (!root_id)
		return false;
	if (child_id->depth < root_id->depth)
		return false;
	if (child_id->stack[root_id->depth] != root_id->id)
		return false;
	return true;
K
KAMEZAWA Hiroyuki 已提交
5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155
}

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

	BUG_ON(!ss->use_id);

	rcu_assign_pointer(id->css, NULL);
	rcu_assign_pointer(css->id, NULL);
5156
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5157
	idr_remove(&ss->idr, id->id);
5158
	spin_unlock(&ss->id_lock);
5159
	kfree_rcu(id, rcu_head);
K
KAMEZAWA Hiroyuki 已提交
5160
}
B
Ben Blum 已提交
5161
EXPORT_SYMBOL_GPL(free_css_id);
K
KAMEZAWA Hiroyuki 已提交
5162 5163 5164 5165 5166 5167 5168 5169 5170

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

static struct css_id *get_new_cssid(struct cgroup_subsys *ss, int depth)
{
	struct css_id *newid;
T
Tejun Heo 已提交
5171
	int ret, size;
K
KAMEZAWA Hiroyuki 已提交
5172 5173 5174 5175 5176 5177 5178

	BUG_ON(!ss->use_id);

	size = sizeof(*newid) + sizeof(unsigned short) * (depth + 1);
	newid = kzalloc(size, GFP_KERNEL);
	if (!newid)
		return ERR_PTR(-ENOMEM);
T
Tejun Heo 已提交
5179 5180

	idr_preload(GFP_KERNEL);
5181
	spin_lock(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5182
	/* Don't use 0. allocates an ID of 1-65535 */
T
Tejun Heo 已提交
5183
	ret = idr_alloc(&ss->idr, newid, 1, CSS_ID_MAX + 1, GFP_NOWAIT);
5184
	spin_unlock(&ss->id_lock);
T
Tejun Heo 已提交
5185
	idr_preload_end();
K
KAMEZAWA Hiroyuki 已提交
5186 5187

	/* Returns error when there are no free spaces for new ID.*/
T
Tejun Heo 已提交
5188
	if (ret < 0)
K
KAMEZAWA Hiroyuki 已提交
5189 5190
		goto err_out;

T
Tejun Heo 已提交
5191
	newid->id = ret;
K
KAMEZAWA Hiroyuki 已提交
5192 5193 5194 5195
	newid->depth = depth;
	return newid;
err_out:
	kfree(newid);
T
Tejun Heo 已提交
5196
	return ERR_PTR(ret);
K
KAMEZAWA Hiroyuki 已提交
5197 5198 5199

}

5200 5201
static int __init_or_module cgroup_init_idr(struct cgroup_subsys *ss,
					    struct cgroup_subsys_state *rootcss)
K
KAMEZAWA Hiroyuki 已提交
5202 5203 5204
{
	struct css_id *newid;

5205
	spin_lock_init(&ss->id_lock);
K
KAMEZAWA Hiroyuki 已提交
5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222
	idr_init(&ss->idr);

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

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

static int alloc_css_id(struct cgroup_subsys *ss, struct cgroup *parent,
			struct cgroup *child)
{
	int subsys_id, i, depth = 0;
	struct cgroup_subsys_state *parent_css, *child_css;
5223
	struct css_id *child_id, *parent_id;
K
KAMEZAWA Hiroyuki 已提交
5224 5225 5226 5227 5228

	subsys_id = ss->subsys_id;
	parent_css = parent->subsys[subsys_id];
	child_css = child->subsys[subsys_id];
	parent_id = parent_css->id;
5229
	depth = parent_id->depth + 1;
K
KAMEZAWA Hiroyuki 已提交
5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266

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

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

	return 0;
}

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

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

	if (unlikely(!cssid))
		return NULL;

	return rcu_dereference(cssid->css);
}
B
Ben Blum 已提交
5267
EXPORT_SYMBOL_GPL(css_lookup);
K
KAMEZAWA Hiroyuki 已提交
5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292

/**
 * css_get_next - lookup next cgroup under specified hierarchy.
 * @ss: pointer to subsystem
 * @id: current position of iteration.
 * @root: pointer to css. search tree under this.
 * @foundid: position of found object.
 *
 * Search next css under the specified hierarchy of rootid. Calling under
 * rcu_read_lock() is necessary. Returns NULL if it reaches the end.
 */
struct cgroup_subsys_state *
css_get_next(struct cgroup_subsys *ss, int id,
	     struct cgroup_subsys_state *root, int *foundid)
{
	struct cgroup_subsys_state *ret = NULL;
	struct css_id *tmp;
	int tmpid;
	int rootid = css_id(root);
	int depth = css_depth(root);

	if (!rootid)
		return NULL;

	BUG_ON(!ss->use_id);
5293 5294
	WARN_ON_ONCE(!rcu_read_lock_held());

K
KAMEZAWA Hiroyuki 已提交
5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317
	/* fill start point for scan */
	tmpid = id;
	while (1) {
		/*
		 * scan next entry from bitmap(tree), tmpid is updated after
		 * idr_get_next().
		 */
		tmp = idr_get_next(&ss->idr, &tmpid);
		if (!tmp)
			break;
		if (tmp->depth >= depth && tmp->stack[depth] == rootid) {
			ret = rcu_dereference(tmp->css);
			if (ret) {
				*foundid = tmpid;
				break;
			}
		}
		/* continue to scan from next id */
		tmpid = tmpid + 1;
	}
	return ret;
}

S
Stephane Eranian 已提交
5318 5319 5320 5321 5322 5323 5324 5325 5326
/*
 * get corresponding css from file open on cgroupfs directory
 */
struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id)
{
	struct cgroup *cgrp;
	struct inode *inode;
	struct cgroup_subsys_state *css;

A
Al Viro 已提交
5327
	inode = file_inode(f);
S
Stephane Eranian 已提交
5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340
	/* check in cgroup filesystem dir */
	if (inode->i_op != &cgroup_dir_inode_operations)
		return ERR_PTR(-EBADF);

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

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

5341
#ifdef CONFIG_CGROUP_DEBUG
5342
static struct cgroup_subsys_state *debug_css_alloc(struct cgroup *cont)
5343 5344 5345 5346 5347 5348 5349 5350 5351
{
	struct cgroup_subsys_state *css = kzalloc(sizeof(*css), GFP_KERNEL);

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

	return css;
}

5352
static void debug_css_free(struct cgroup *cont)
5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
{
	kfree(cont->subsys[debug_subsys_id]);
}

static u64 cgroup_refcount_read(struct cgroup *cont, struct cftype *cft)
{
	return atomic_read(&cont->count);
}

static u64 debug_taskcount_read(struct cgroup *cont, struct cftype *cft)
{
	return cgroup_task_count(cont);
}

static u64 current_css_set_read(struct cgroup *cont, struct cftype *cft)
{
	return (u64)(unsigned long)current->cgroups;
}

static u64 current_css_set_refcount_read(struct cgroup *cont,
					   struct cftype *cft)
{
	u64 count;

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

5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400
static int current_css_set_cg_links_read(struct cgroup *cont,
					 struct cftype *cft,
					 struct seq_file *seq)
{
	struct cg_cgroup_link *link;
	struct css_set *cg;

	read_lock(&css_set_lock);
	rcu_read_lock();
	cg = rcu_dereference(current->cgroups);
	list_for_each_entry(link, &cg->cg_links, cg_link_list) {
		struct cgroup *c = link->cgrp;
		const char *name;

		if (c->dentry)
			name = c->dentry->d_name.name;
		else
			name = "?";
5401 5402
		seq_printf(seq, "Root %d group %s\n",
			   c->root->hierarchy_id, name);
5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435
	}
	rcu_read_unlock();
	read_unlock(&css_set_lock);
	return 0;
}

#define MAX_TASKS_SHOWN_PER_CSS 25
static int cgroup_css_links_read(struct cgroup *cont,
				 struct cftype *cft,
				 struct seq_file *seq)
{
	struct cg_cgroup_link *link;

	read_lock(&css_set_lock);
	list_for_each_entry(link, &cont->css_sets, cgrp_link_list) {
		struct css_set *cg = link->cg;
		struct task_struct *task;
		int count = 0;
		seq_printf(seq, "css_set %p\n", cg);
		list_for_each_entry(task, &cg->tasks, cg_list) {
			if (count++ > MAX_TASKS_SHOWN_PER_CSS) {
				seq_puts(seq, "  ...\n");
				break;
			} else {
				seq_printf(seq, "  task %d\n",
					   task_pid_vnr(task));
			}
		}
	}
	read_unlock(&css_set_lock);
	return 0;
}

5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460
static u64 releasable_read(struct cgroup *cgrp, struct cftype *cft)
{
	return test_bit(CGRP_RELEASABLE, &cgrp->flags);
}

static struct cftype debug_files[] =  {
	{
		.name = "cgroup_refcount",
		.read_u64 = cgroup_refcount_read,
	},
	{
		.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,
	},

5461 5462 5463 5464 5465 5466 5467 5468 5469 5470
	{
		.name = "current_css_set_cg_links",
		.read_seq_string = current_css_set_cg_links_read,
	},

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

5471 5472 5473 5474 5475
	{
		.name = "releasable",
		.read_u64 = releasable_read,
	},

5476 5477
	{ }	/* terminate */
};
5478 5479 5480

struct cgroup_subsys debug_subsys = {
	.name = "debug",
5481 5482
	.css_alloc = debug_css_alloc,
	.css_free = debug_css_free,
5483
	.subsys_id = debug_subsys_id,
5484
	.base_cftypes = debug_files,
5485 5486
};
#endif /* CONFIG_CGROUP_DEBUG */