eventpoll.c 57.6 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
/*
2 3
 *  fs/eventpoll.c (Efficient event retrieval implementation)
 *  Copyright (C) 2001,...,2009	 Davide Libenzi
Linus Torvalds's avatar
Linus Torvalds committed
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  Davide Libenzi <davidel@xmailserver.org>
 *
 */

#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/signal.h>
#include <linux/errno.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/poll.h>
#include <linux/string.h>
#include <linux/list.h>
#include <linux/hash.h>
#include <linux/spinlock.h>
#include <linux/syscalls.h>
#include <linux/rbtree.h>
#include <linux/wait.h>
#include <linux/eventpoll.h>
#include <linux/mount.h>
#include <linux/bitops.h>
34
#include <linux/mutex.h>
35
#include <linux/anon_inodes.h>
36
#include <linux/device.h>
37
#include <linux/freezer.h>
Linus Torvalds's avatar
Linus Torvalds committed
38 39 40
#include <asm/uaccess.h>
#include <asm/io.h>
#include <asm/mman.h>
41
#include <linux/atomic.h>
42 43
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
44
#include <linux/compat.h>
Linus Torvalds's avatar
Linus Torvalds committed
45 46 47 48 49

/*
 * LOCKING:
 * There are three level of locking required by epoll :
 *
50
 * 1) epmutex (mutex)
51 52
 * 2) ep->mtx (mutex)
 * 3) ep->lock (spinlock)
Linus Torvalds's avatar
Linus Torvalds committed
53 54 55 56 57 58 59 60 61
 *
 * The acquire order is the one listed above, from 1 to 3.
 * We need a spinlock (ep->lock) because we manipulate objects
 * from inside the poll callback, that might be triggered from
 * a wake_up() that in turn might be called from IRQ context.
 * So we can't sleep inside the poll callback and hence we need
 * a spinlock. During the event transfer loop (from kernel to
 * user space) we could end up sleeping due a copy_to_user(), so
 * we need a lock that will allow us to sleep. This lock is a
62 63 64 65 66
 * mutex (ep->mtx). It is acquired during the event transfer loop,
 * during epoll_ctl(EPOLL_CTL_DEL) and during eventpoll_release_file().
 * Then we also need a global mutex to serialize eventpoll_release_file()
 * and ep_free().
 * This mutex is acquired by ep_free() during the epoll file
Linus Torvalds's avatar
Linus Torvalds committed
67 68
 * cleanup path and it is also acquired by eventpoll_release_file()
 * if a file has been pushed inside an epoll set and it is then
69
 * close()d without a previous call to epoll_ctl(EPOLL_CTL_DEL).
70 71 72 73 74 75 76
 * It is also acquired when inserting an epoll fd onto another epoll
 * fd. We do this so that we walk the epoll tree and ensure that this
 * insertion does not create a cycle of epoll file descriptors, which
 * could lead to deadlock. We need a global mutex to prevent two
 * simultaneous inserts (A into B and B into A) from racing and
 * constructing a cycle without either insert observing that it is
 * going to.
77 78 79 80 81 82 83 84 85
 * It is necessary to acquire multiple "ep->mtx"es at once in the
 * case when one epoll fd is added to another. In this case, we
 * always acquire the locks in the order of nesting (i.e. after
 * epoll_ctl(e1, EPOLL_CTL_ADD, e2), e1->mtx will always be acquired
 * before e2->mtx). Since we disallow cycles of epoll file
 * descriptors, this ensures that the mutexes are well-ordered. In
 * order to communicate this nesting to lockdep, when walking a tree
 * of epoll file descriptors, we use the current recursion depth as
 * the lockdep subkey.
86 87 88
 * It is possible to drop the "ep->mtx" and to use the global
 * mutex "epmutex" (together with "ep->lock") to have it working,
 * but having "ep->mtx" will make the interface more scalable.
89
 * Events that require holding "epmutex" are very rare, while for
90 91
 * normal operations the epoll private "ep->mtx" will guarantee
 * a better scalability.
Linus Torvalds's avatar
Linus Torvalds committed
92 93 94
 */

/* Epoll private bits inside the event mask */
95
#define EP_PRIVATE_BITS (EPOLLWAKEUP | EPOLLONESHOT | EPOLLET)
Linus Torvalds's avatar
Linus Torvalds committed
96

97 98
/* Maximum number of nesting allowed inside epoll sets */
#define EP_MAX_NESTS 4
Linus Torvalds's avatar
Linus Torvalds committed
99

Davide Libenzi's avatar
Davide Libenzi committed
100 101
#define EP_MAX_EVENTS (INT_MAX / sizeof(struct epoll_event))

102 103
#define EP_UNACTIVE_PTR ((void *) -1L)

104 105
#define EP_ITEM_COST (sizeof(struct epitem) + sizeof(struct eppoll_entry))

Linus Torvalds's avatar
Linus Torvalds committed
106 107 108
struct epoll_filefd {
	struct file *file;
	int fd;
109
} __packed;
Linus Torvalds's avatar
Linus Torvalds committed
110 111

/*
112 113
 * Structure used to track possible nested calls, for too deep recursions
 * and loop cycles.
Linus Torvalds's avatar
Linus Torvalds committed
114
 */
115
struct nested_call_node {
Linus Torvalds's avatar
Linus Torvalds committed
116
	struct list_head llink;
117
	void *cookie;
118
	void *ctx;
Linus Torvalds's avatar
Linus Torvalds committed
119 120 121
};

/*
122 123
 * This structure is used as collector for nested calls, to check for
 * maximum recursion dept and loop cycles.
Linus Torvalds's avatar
Linus Torvalds committed
124
 */
125 126
struct nested_calls {
	struct list_head tasks_call_list;
Linus Torvalds's avatar
Linus Torvalds committed
127 128 129
	spinlock_t lock;
};

130 131 132
/*
 * Each file descriptor added to the eventpoll interface will
 * have an entry of this type linked to the "rbr" RB tree.
133 134
 * Avoid increasing the size of this struct, there can be many thousands
 * of these on a server and we do not want this to take another cache line.
135 136
 */
struct epitem {
137
	/* RB tree node used to link this structure to the eventpoll RB tree */
138 139 140 141 142
	struct rb_node rbn;

	/* List header used to link this structure to the eventpoll ready list */
	struct list_head rdllink;

143 144 145 146 147 148
	/*
	 * Works together "struct eventpoll"->ovflist in keeping the
	 * single linked chain of items.
	 */
	struct epitem *next;

149 150 151 152 153 154 155 156 157 158 159 160 161 162 163
	/* The file descriptor information this item refers to */
	struct epoll_filefd ffd;

	/* Number of active wait queue attached to poll operations */
	int nwait;

	/* List containing poll wait queues */
	struct list_head pwqlist;

	/* The "container" of this item */
	struct eventpoll *ep;

	/* List header used to link this item to the "struct file" items list */
	struct list_head fllink;

164
	/* wakeup_source used when EPOLLWAKEUP is set */
165
	struct wakeup_source __rcu *ws;
166

167 168
	/* The structure that describe the interested events and the source fd */
	struct epoll_event event;
169 170
};

Linus Torvalds's avatar
Linus Torvalds committed
171 172
/*
 * This structure is stored inside the "private_data" member of the file
173
 * structure and represents the main data structure for the eventpoll
Linus Torvalds's avatar
Linus Torvalds committed
174 175 176
 * interface.
 */
struct eventpoll {
177
	/* Protect the access to this structure */
178
	spinlock_t lock;
Linus Torvalds's avatar
Linus Torvalds committed
179 180

	/*
181 182 183 184
	 * This mutex is used to ensure that files are not removed
	 * while epoll is using them. This is held during the event
	 * collection loop, the file cleanup path, the epoll file exit
	 * code and the ctl operations.
Linus Torvalds's avatar
Linus Torvalds committed
185
	 */
186
	struct mutex mtx;
Linus Torvalds's avatar
Linus Torvalds committed
187 188 189 190 191 192 193 194 195 196

	/* Wait queue used by sys_epoll_wait() */
	wait_queue_head_t wq;

	/* Wait queue used by file->poll() */
	wait_queue_head_t poll_wait;

	/* List of ready file descriptors */
	struct list_head rdllist;

197
	/* RB tree root used to store monitored fd structs */
Linus Torvalds's avatar
Linus Torvalds committed
198
	struct rb_root rbr;
199 200 201

	/*
	 * This is a single linked list that chains all the "struct epitem" that
202
	 * happened while transferring ready events to userspace w/out
203 204 205
	 * holding ->lock.
	 */
	struct epitem *ovflist;
206

207 208 209
	/* wakeup_source used when ep_scan_ready_list is running */
	struct wakeup_source *ws;

210 211
	/* The user that created the eventpoll descriptor */
	struct user_struct *user;
Jason Baron's avatar
Jason Baron committed
212 213 214 215 216 217

	struct file *file;

	/* used to optimize loop detection check */
	int visited;
	struct list_head visited_list_link;
Linus Torvalds's avatar
Linus Torvalds committed
218 219 220 221 222 223 224 225
};

/* Wait structure used by the poll hooks */
struct eppoll_entry {
	/* List header used to link this structure to the "struct epitem" */
	struct list_head llink;

	/* The "base" pointer is set to the container "struct epitem" */
226
	struct epitem *base;
Linus Torvalds's avatar
Linus Torvalds committed
227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243

	/*
	 * Wait queue item that will be linked to the target file wait
	 * queue head.
	 */
	wait_queue_t wait;

	/* The wait queue head that linked the "wait" wait queue item */
	wait_queue_head_t *whead;
};

/* Wrapper struct used by poll queueing */
struct ep_pqueue {
	poll_table pt;
	struct epitem *epi;
};

244 245 246 247 248 249
/* Used by the ep_send_events() function as callback private data */
struct ep_send_events_data {
	int maxevents;
	struct epoll_event __user *events;
};

250 251 252 253
/*
 * Configuration options available inside /proc/sys/fs/epoll/
 */
/* Maximum number of epoll watched descriptors, per user */
254
static long max_user_watches __read_mostly;
255

Linus Torvalds's avatar
Linus Torvalds committed
256
/*
257
 * This mutex is used to serialize ep_free() and eventpoll_release_file().
Linus Torvalds's avatar
Linus Torvalds committed
258
 */
259
static DEFINE_MUTEX(epmutex);
Linus Torvalds's avatar
Linus Torvalds committed
260

261 262 263
/* Used to check for epoll file descriptor inclusion loops */
static struct nested_calls poll_loop_ncalls;

264 265 266 267 268
/* Used for safe wake up implementation */
static struct nested_calls poll_safewake_ncalls;

/* Used to call file's f_op->poll() under the nested calls boundaries */
static struct nested_calls poll_readywalk_ncalls;
Linus Torvalds's avatar
Linus Torvalds committed
269 270

/* Slab cache used to allocate "struct epitem" */
271
static struct kmem_cache *epi_cache __read_mostly;
Linus Torvalds's avatar
Linus Torvalds committed
272 273

/* Slab cache used to allocate "struct eppoll_entry" */
274
static struct kmem_cache *pwq_cache __read_mostly;
Linus Torvalds's avatar
Linus Torvalds committed
275

Jason Baron's avatar
Jason Baron committed
276 277 278 279 280 281 282 283 284
/* Visited nodes during ep_loop_check(), so we can unset them when we finish */
static LIST_HEAD(visited_list);

/*
 * List of files with newly added links, where we may need to limit the number
 * of emanating paths. Protected by the epmutex.
 */
static LIST_HEAD(tfile_check_list);

285 286 287 288
#ifdef CONFIG_SYSCTL

#include <linux/sysctl.h>

289 290
static long zero;
static long long_max = LONG_MAX;
291 292 293 294 295

ctl_table epoll_table[] = {
	{
		.procname	= "max_user_watches",
		.data		= &max_user_watches,
296
		.maxlen		= sizeof(max_user_watches),
297
		.mode		= 0644,
298
		.proc_handler	= proc_doulongvec_minmax,
299
		.extra1		= &zero,
300
		.extra2		= &long_max,
301
	},
302
	{ }
303 304 305
};
#endif /* CONFIG_SYSCTL */

Jason Baron's avatar
Jason Baron committed
306 307 308 309 310 311
static const struct file_operations eventpoll_fops;

static inline int is_file_epoll(struct file *f)
{
	return f->f_op == &eventpoll_fops;
}
312

313
/* Setup the structure that is used as key for the RB tree */
314 315 316 317 318 319 320
static inline void ep_set_ffd(struct epoll_filefd *ffd,
			      struct file *file, int fd)
{
	ffd->file = file;
	ffd->fd = fd;
}

321
/* Compare RB tree keys */
322 323 324 325 326 327 328 329 330 331 332 333 334
static inline int ep_cmp_ffd(struct epoll_filefd *p1,
			     struct epoll_filefd *p2)
{
	return (p1->file > p2->file ? +1:
	        (p1->file < p2->file ? -1 : p1->fd - p2->fd));
}

/* Tells us if the item is currently linked */
static inline int ep_is_linked(struct list_head *p)
{
	return !list_empty(p);
}

335 336 337 338 339
static inline struct eppoll_entry *ep_pwq_from_wait(wait_queue_t *p)
{
	return container_of(p, struct eppoll_entry, wait);
}

340
/* Get the "struct epitem" from a wait queue pointer */
341
static inline struct epitem *ep_item_from_wait(wait_queue_t *p)
342 343 344 345 346
{
	return container_of(p, struct eppoll_entry, wait)->base;
}

/* Get the "struct epitem" from an epoll queue wrapper */
347
static inline struct epitem *ep_item_from_epqueue(poll_table *p)
348 349 350 351 352
{
	return container_of(p, struct ep_pqueue, pt)->epi;
}

/* Tells if the epoll_ctl(2) operation needs an event copy from userspace */
353
static inline int ep_op_has_event(int op)
354
{
355
	return op != EPOLL_CTL_DEL;
356 357
}

Linus Torvalds's avatar
Linus Torvalds committed
358
/* Initialize the poll safe wake up structure */
359
static void ep_nested_calls_init(struct nested_calls *ncalls)
Linus Torvalds's avatar
Linus Torvalds committed
360
{
361 362
	INIT_LIST_HEAD(&ncalls->tasks_call_list);
	spin_lock_init(&ncalls->lock);
Linus Torvalds's avatar
Linus Torvalds committed
363 364
}

365 366 367 368 369 370 371 372 373 374 375 376 377
/**
 * ep_events_available - Checks if ready events might be available.
 *
 * @ep: Pointer to the eventpoll context.
 *
 * Returns: Returns a value different than zero if ready events are available,
 *          or zero otherwise.
 */
static inline int ep_events_available(struct eventpoll *ep)
{
	return !list_empty(&ep->rdllist) || ep->ovflist != EP_UNACTIVE_PTR;
}

378 379 380 381 382 383 384 385 386 387 388
/**
 * ep_call_nested - Perform a bound (possibly) nested call, by checking
 *                  that the recursion limit is not exceeded, and that
 *                  the same nested call (by the meaning of same cookie) is
 *                  no re-entered.
 *
 * @ncalls: Pointer to the nested_calls structure to be used for this call.
 * @max_nests: Maximum number of allowed nesting calls.
 * @nproc: Nested call core function pointer.
 * @priv: Opaque data to be passed to the @nproc callback.
 * @cookie: Cookie to be used to identify this nested call.
389
 * @ctx: This instance context.
390 391 392
 *
 * Returns: Returns the code returned by the @nproc callback, or -1 if
 *          the maximum recursion limit has been exceeded.
Linus Torvalds's avatar
Linus Torvalds committed
393
 */
394 395
static int ep_call_nested(struct nested_calls *ncalls, int max_nests,
			  int (*nproc)(void *, void *, int), void *priv,
396
			  void *cookie, void *ctx)
Linus Torvalds's avatar
Linus Torvalds committed
397
{
398
	int error, call_nests = 0;
Linus Torvalds's avatar
Linus Torvalds committed
399
	unsigned long flags;
400 401 402
	struct list_head *lsthead = &ncalls->tasks_call_list;
	struct nested_call_node *tncur;
	struct nested_call_node tnode;
Linus Torvalds's avatar
Linus Torvalds committed
403

404
	spin_lock_irqsave(&ncalls->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
405

406 407 408 409 410
	/*
	 * Try to see if the current task is already inside this wakeup call.
	 * We use a list here, since the population inside this set is always
	 * very much limited.
	 */
411
	list_for_each_entry(tncur, lsthead, llink) {
412
		if (tncur->ctx == ctx &&
413
		    (tncur->cookie == cookie || ++call_nests > max_nests)) {
Linus Torvalds's avatar
Linus Torvalds committed
414 415 416 417
			/*
			 * Ops ... loop detected or maximum nest level reached.
			 * We abort this wake by breaking the cycle itself.
			 */
418 419
			error = -1;
			goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
420 421 422
		}
	}

423
	/* Add the current task and cookie to the list */
424
	tnode.ctx = ctx;
425
	tnode.cookie = cookie;
Linus Torvalds's avatar
Linus Torvalds committed
426 427
	list_add(&tnode.llink, lsthead);

428
	spin_unlock_irqrestore(&ncalls->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
429

430 431
	/* Call the nested function */
	error = (*nproc)(priv, cookie, call_nests);
Linus Torvalds's avatar
Linus Torvalds committed
432 433

	/* Remove the current task from the list */
434
	spin_lock_irqsave(&ncalls->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
435
	list_del(&tnode.llink);
436
out_unlock:
437 438 439 440 441
	spin_unlock_irqrestore(&ncalls->lock, flags);

	return error;
}

442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466
/*
 * As described in commit 0ccf831cb lockdep: annotate epoll
 * the use of wait queues used by epoll is done in a very controlled
 * manner. Wake ups can nest inside each other, but are never done
 * with the same locking. For example:
 *
 *   dfd = socket(...);
 *   efd1 = epoll_create();
 *   efd2 = epoll_create();
 *   epoll_ctl(efd1, EPOLL_CTL_ADD, dfd, ...);
 *   epoll_ctl(efd2, EPOLL_CTL_ADD, efd1, ...);
 *
 * When a packet arrives to the device underneath "dfd", the net code will
 * issue a wake_up() on its poll wake list. Epoll (efd1) has installed a
 * callback wakeup entry on that queue, and the wake_up() performed by the
 * "dfd" net code will end up in ep_poll_callback(). At this point epoll
 * (efd1) notices that it may have some event ready, so it needs to wake up
 * the waiters on its poll wait list (efd2). So it calls ep_poll_safewake()
 * that ends up in another wake_up(), after having checked about the
 * recursion constraints. That are, no more than EP_MAX_POLLWAKE_NESTS, to
 * avoid stack blasting.
 *
 * When CONFIG_DEBUG_LOCK_ALLOC is enabled, make sure lockdep can handle
 * this special case of epoll.
 */
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
#ifdef CONFIG_DEBUG_LOCK_ALLOC
static inline void ep_wake_up_nested(wait_queue_head_t *wqueue,
				     unsigned long events, int subclass)
{
	unsigned long flags;

	spin_lock_irqsave_nested(&wqueue->lock, flags, subclass);
	wake_up_locked_poll(wqueue, events);
	spin_unlock_irqrestore(&wqueue->lock, flags);
}
#else
static inline void ep_wake_up_nested(wait_queue_head_t *wqueue,
				     unsigned long events, int subclass)
{
	wake_up_poll(wqueue, events);
}
#endif

485 486
static int ep_poll_wakeup_proc(void *priv, void *cookie, int call_nests)
{
487 488
	ep_wake_up_nested((wait_queue_head_t *) cookie, POLLIN,
			  1 + call_nests);
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
	return 0;
}

/*
 * Perform a safe wake up of the poll wait list. The problem is that
 * with the new callback'd wake up system, it is possible that the
 * poll callback is reentered from inside the call to wake_up() done
 * on the poll wait queue head. The rule is that we cannot reenter the
 * wake up code from the same task more than EP_MAX_NESTS times,
 * and we cannot reenter the same wait queue head at all. This will
 * enable to have a hierarchy of epoll file descriptor of no more than
 * EP_MAX_NESTS deep.
 */
static void ep_poll_safewake(wait_queue_head_t *wq)
{
504 505
	int this_cpu = get_cpu();

506
	ep_call_nested(&poll_safewake_ncalls, EP_MAX_NESTS,
507 508 509
		       ep_poll_wakeup_proc, NULL, wq, (void *) (long) this_cpu);

	put_cpu();
Linus Torvalds's avatar
Linus Torvalds committed
510 511
}

512 513 514 515 516 517 518 519 520 521 522 523
static void ep_remove_wait_queue(struct eppoll_entry *pwq)
{
	wait_queue_head_t *whead;

	rcu_read_lock();
	/* If it is cleared by POLLFREE, it should be rcu-safe */
	whead = rcu_dereference(pwq->whead);
	if (whead)
		remove_wait_queue(whead, &pwq->wait);
	rcu_read_unlock();
}

Linus Torvalds's avatar
Linus Torvalds committed
524
/*
525 526 527
 * This function unregisters poll callbacks from the associated file
 * descriptor.  Must be called with "mtx" held (or "epmutex" if called from
 * ep_free).
Linus Torvalds's avatar
Linus Torvalds committed
528
 */
529
static void ep_unregister_pollwait(struct eventpoll *ep, struct epitem *epi)
Linus Torvalds's avatar
Linus Torvalds committed
530
{
531 532
	struct list_head *lsthead = &epi->pwqlist;
	struct eppoll_entry *pwq;
Linus Torvalds's avatar
Linus Torvalds committed
533

534 535
	while (!list_empty(lsthead)) {
		pwq = list_first_entry(lsthead, struct eppoll_entry, llink);
Linus Torvalds's avatar
Linus Torvalds committed
536

537
		list_del(&pwq->llink);
538
		ep_remove_wait_queue(pwq);
539
		kmem_cache_free(pwq_cache, pwq);
Linus Torvalds's avatar
Linus Torvalds committed
540 541 542
	}
}

543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574
/* call only when ep->mtx is held */
static inline struct wakeup_source *ep_wakeup_source(struct epitem *epi)
{
	return rcu_dereference_check(epi->ws, lockdep_is_held(&epi->ep->mtx));
}

/* call only when ep->mtx is held */
static inline void ep_pm_stay_awake(struct epitem *epi)
{
	struct wakeup_source *ws = ep_wakeup_source(epi);

	if (ws)
		__pm_stay_awake(ws);
}

static inline bool ep_has_wakeup_source(struct epitem *epi)
{
	return rcu_access_pointer(epi->ws) ? true : false;
}

/* call when ep->mtx cannot be held (ep_poll_callback) */
static inline void ep_pm_stay_awake_rcu(struct epitem *epi)
{
	struct wakeup_source *ws;

	rcu_read_lock();
	ws = rcu_dereference(epi->ws);
	if (ws)
		__pm_stay_awake(ws);
	rcu_read_unlock();
}

575 576 577 578 579 580 581 582
/**
 * ep_scan_ready_list - Scans the ready list in a way that makes possible for
 *                      the scan code, to call f_op->poll(). Also allows for
 *                      O(NumReady) performance.
 *
 * @ep: Pointer to the epoll private data structure.
 * @sproc: Pointer to the scan callback.
 * @priv: Private opaque data passed to the @sproc callback.
583
 * @depth: The current depth of recursive f_op->poll calls.
584 585 586 587 588 589
 *
 * Returns: The same integer error code returned by the @sproc callback.
 */
static int ep_scan_ready_list(struct eventpoll *ep,
			      int (*sproc)(struct eventpoll *,
					   struct list_head *, void *),
590 591
			      void *priv,
			      int depth)
592 593 594 595
{
	int error, pwake = 0;
	unsigned long flags;
	struct epitem *epi, *nepi;
596
	LIST_HEAD(txlist);
597 598 599

	/*
	 * We need to lock this because we could be hit by
600
	 * eventpoll_release_file() and epoll_ctl().
601
	 */
602
	mutex_lock_nested(&ep->mtx, depth);
603 604 605 606 607 608 609 610 611 612

	/*
	 * Steal the ready list, and re-init the original one to the
	 * empty list. Also, set ep->ovflist to NULL so that events
	 * happening while looping w/out locks, are not lost. We cannot
	 * have the poll callback to queue directly on ep->rdllist,
	 * because we want the "sproc" callback to be able to do it
	 * in a lockless way.
	 */
	spin_lock_irqsave(&ep->lock, flags);
613
	list_splice_init(&ep->rdllist, &txlist);
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
	ep->ovflist = NULL;
	spin_unlock_irqrestore(&ep->lock, flags);

	/*
	 * Now call the callback function.
	 */
	error = (*sproc)(ep, &txlist, priv);

	spin_lock_irqsave(&ep->lock, flags);
	/*
	 * During the time we spent inside the "sproc" callback, some
	 * other events might have been queued by the poll callback.
	 * We re-insert them inside the main ready-list here.
	 */
	for (nepi = ep->ovflist; (epi = nepi) != NULL;
	     nepi = epi->next, epi->next = EP_UNACTIVE_PTR) {
		/*
		 * We need to check if the item is already in the list.
		 * During the "sproc" callback execution time, items are
		 * queued into ->ovflist but the "txlist" might already
		 * contain them, and the list_splice() below takes care of them.
		 */
636
		if (!ep_is_linked(&epi->rdllink)) {
637
			list_add_tail(&epi->rdllink, &ep->rdllist);
638
			ep_pm_stay_awake(epi);
639
		}
640 641 642 643 644 645 646 647 648 649 650 651
	}
	/*
	 * We need to set back ep->ovflist to EP_UNACTIVE_PTR, so that after
	 * releasing the lock, events will be queued in the normal way inside
	 * ep->rdllist.
	 */
	ep->ovflist = EP_UNACTIVE_PTR;

	/*
	 * Quickly re-inject items left on "txlist".
	 */
	list_splice(&txlist, &ep->rdllist);
652
	__pm_relax(ep->ws);
653 654 655

	if (!list_empty(&ep->rdllist)) {
		/*
656 657
		 * Wake up (if active) both the eventpoll wait list and
		 * the ->poll() wait list (delayed after we release the lock).
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674
		 */
		if (waitqueue_active(&ep->wq))
			wake_up_locked(&ep->wq);
		if (waitqueue_active(&ep->poll_wait))
			pwake++;
	}
	spin_unlock_irqrestore(&ep->lock, flags);

	mutex_unlock(&ep->mtx);

	/* We have to call this outside the lock */
	if (pwake)
		ep_poll_safewake(&ep->poll_wait);

	return error;
}

675 676
/*
 * Removes a "struct epitem" from the eventpoll RB tree and deallocates
677
 * all the associated resources. Must be called with "mtx" held.
678 679 680 681 682
 */
static int ep_remove(struct eventpoll *ep, struct epitem *epi)
{
	unsigned long flags;
	struct file *file = epi->ffd.file;
Linus Torvalds's avatar
Linus Torvalds committed
683 684

	/*
685 686 687 688 689 690
	 * Removes poll wait queue hooks. We _have_ to do this without holding
	 * the "ep->lock" otherwise a deadlock might occur. This because of the
	 * sequence of the lock acquisition. Here we do "ep->lock" then the wait
	 * queue head lock when unregistering the wait queue. The wakeup callback
	 * will run by holding the wait queue head lock and will call our callback
	 * that will try to get "ep->lock".
Linus Torvalds's avatar
Linus Torvalds committed
691
	 */
692
	ep_unregister_pollwait(ep, epi);
Linus Torvalds's avatar
Linus Torvalds committed
693

694
	/* Remove the current item from the list of epoll hooks */
695
	spin_lock(&file->f_lock);
696 697
	if (ep_is_linked(&epi->fllink))
		list_del_init(&epi->fllink);
698
	spin_unlock(&file->f_lock);
Linus Torvalds's avatar
Linus Torvalds committed
699

700
	rb_erase(&epi->rbn, &ep->rbr);
Linus Torvalds's avatar
Linus Torvalds committed
701

702 703 704 705
	spin_lock_irqsave(&ep->lock, flags);
	if (ep_is_linked(&epi->rdllink))
		list_del_init(&epi->rdllink);
	spin_unlock_irqrestore(&ep->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
706

707
	wakeup_source_unregister(ep_wakeup_source(epi));
708

709
	/* At this point it is safe to free the eventpoll item */
710
	kmem_cache_free(epi_cache, epi);
Linus Torvalds's avatar
Linus Torvalds committed
711

712
	atomic_long_dec(&ep->user->epoll_watches);
713

714
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
715 716
}

717
static void ep_free(struct eventpoll *ep)
Linus Torvalds's avatar
Linus Torvalds committed
718
{
719 720
	struct rb_node *rbp;
	struct epitem *epi;
Linus Torvalds's avatar
Linus Torvalds committed
721

722 723
	/* We need to release all tasks waiting for these file */
	if (waitqueue_active(&ep->poll_wait))
724
		ep_poll_safewake(&ep->poll_wait);
Linus Torvalds's avatar
Linus Torvalds committed
725

726 727 728
	/*
	 * We need to lock this because we could be hit by
	 * eventpoll_release_file() while we're freeing the "struct eventpoll".
729
	 * We do not need to hold "ep->mtx" here because the epoll file
730 731
	 * is on the way to be removed and no one has references to it
	 * anymore. The only hit might come from eventpoll_release_file() but
732
	 * holding "epmutex" is sufficient here.
733 734
	 */
	mutex_lock(&epmutex);
Linus Torvalds's avatar
Linus Torvalds committed
735 736

	/*
737
	 * Walks through the whole tree by unregistering poll callbacks.
Linus Torvalds's avatar
Linus Torvalds committed
738
	 */
739 740 741 742
	for (rbp = rb_first(&ep->rbr); rbp; rbp = rb_next(rbp)) {
		epi = rb_entry(rbp, struct epitem, rbn);

		ep_unregister_pollwait(ep, epi);
743
		cond_resched();
744
	}
Linus Torvalds's avatar
Linus Torvalds committed
745 746

	/*
747 748
	 * Walks through the whole tree by freeing each "struct epitem". At this
	 * point we are sure no poll callbacks will be lingering around, and also by
749
	 * holding "epmutex" we can be sure that no file cleanup code will hit
750
	 * us during this operation. So we can avoid the lock on "ep->lock".
751 752
	 * We do not need to lock ep->mtx, either, we only do it to prevent
	 * a lockdep warning.
Linus Torvalds's avatar
Linus Torvalds committed
753
	 */
754
	mutex_lock(&ep->mtx);
755
	while ((rbp = rb_first(&ep->rbr)) != NULL) {
756 757
		epi = rb_entry(rbp, struct epitem, rbn);
		ep_remove(ep, epi);
758
		cond_resched();
759
	}
760
	mutex_unlock(&ep->mtx);
Linus Torvalds's avatar
Linus Torvalds committed
761

762
	mutex_unlock(&epmutex);
763
	mutex_destroy(&ep->mtx);
764
	free_uid(ep->user);
765
	wakeup_source_unregister(ep->ws);
766
	kfree(ep);
767
}
Linus Torvalds's avatar
Linus Torvalds committed
768

769 770 771
static int ep_eventpoll_release(struct inode *inode, struct file *file)
{
	struct eventpoll *ep = file->private_data;
Linus Torvalds's avatar
Linus Torvalds committed
772

773
	if (ep)
774 775 776
		ep_free(ep);

	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
777 778
}

779 780 781 782 783 784 785
static inline unsigned int ep_item_poll(struct epitem *epi, poll_table *pt)
{
	pt->_key = epi->event.events;

	return epi->ffd.file->f_op->poll(epi->ffd.file, pt) & epi->event.events;
}

786 787
static int ep_read_events_proc(struct eventpoll *ep, struct list_head *head,
			       void *priv)
788 789
{
	struct epitem *epi, *tmp;
790
	poll_table pt;
791

792
	init_poll_funcptr(&pt, NULL);
793

794
	list_for_each_entry_safe(epi, tmp, head, rdllink) {
795
		if (ep_item_poll(epi, &pt))
796
			return POLLIN | POLLRDNORM;
797
		else {
798 799 800 801 802
			/*
			 * Item has been dropped into the ready list by the poll
			 * callback, but it's not actually ready, as far as
			 * caller requested events goes. We can remove it here.
			 */
803
			__pm_relax(ep_wakeup_source(epi));
804
			list_del_init(&epi->rdllink);
805
		}
806 807 808 809 810 811 812
	}

	return 0;
}

static int ep_poll_readyevents_proc(void *priv, void *cookie, int call_nests)
{
813
	return ep_scan_ready_list(priv, ep_read_events_proc, NULL, call_nests + 1);
814 815
}

816 817
static unsigned int ep_eventpoll_poll(struct file *file, poll_table *wait)
{
818
	int pollflags;
819
	struct eventpoll *ep = file->private_data;
Linus Torvalds's avatar
Linus Torvalds committed
820

821 822 823
	/* Insert inside our poll wait queue */
	poll_wait(file, &ep->poll_wait, wait);

824 825 826 827 828 829 830
	/*
	 * Proceed to find out if wanted events are really available inside
	 * the ready list. This need to be done under ep_call_nested()
	 * supervision, since the call to f_op->poll() done on listed files
	 * could re-enter here.
	 */
	pollflags = ep_call_nested(&poll_readywalk_ncalls, EP_MAX_NESTS,
831
				   ep_poll_readyevents_proc, ep, ep, current);
832

833
	return pollflags != -1 ? pollflags : 0;
834 835
}

836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858
#ifdef CONFIG_PROC_FS
static int ep_show_fdinfo(struct seq_file *m, struct file *f)
{
	struct eventpoll *ep = f->private_data;
	struct rb_node *rbp;
	int ret = 0;

	mutex_lock(&ep->mtx);
	for (rbp = rb_first(&ep->rbr); rbp; rbp = rb_next(rbp)) {
		struct epitem *epi = rb_entry(rbp, struct epitem, rbn);

		ret = seq_printf(m, "tfd: %8d events: %8x data: %16llx\n",
				 epi->ffd.fd, epi->event.events,
				 (long long)epi->event.data);
		if (ret)
			break;
	}
	mutex_unlock(&ep->mtx);

	return ret;
}
#endif

859 860
/* File callbacks that implement the eventpoll file behaviour */
static const struct file_operations eventpoll_fops = {
861 862 863
#ifdef CONFIG_PROC_FS
	.show_fdinfo	= ep_show_fdinfo,
#endif
864
	.release	= ep_eventpoll_release,
865 866
	.poll		= ep_eventpoll_poll,
	.llseek		= noop_llseek,
867 868
};

Davide Libenzi's avatar
Davide Libenzi committed
869
/*
870 871 872
 * This is called from eventpoll_release() to unlink files from the eventpoll
 * interface. We need to have this facility to cleanup correctly files that are
 * closed without being removed from the eventpoll interface.
Davide Libenzi's avatar
Davide Libenzi committed
873
 */
874
void eventpoll_release_file(struct file *file)
Davide Libenzi's avatar
Davide Libenzi committed
875
{
876 877 878
	struct list_head *lsthead = &file->f_ep_links;
	struct eventpoll *ep;
	struct epitem *epi;
Davide Libenzi's avatar
Davide Libenzi committed
879 880

	/*
881
	 * We don't want to get "file->f_lock" because it is not
882
	 * necessary. It is not necessary because we're in the "struct file"
883
	 * cleanup path, and this means that no one is using this file anymore.
884
	 * So, for example, epoll_ctl() cannot hit here since if we reach this
885
	 * point, the file counter already went to zero and fget() would fail.
886
	 * The only hit might come from ep_free() but by holding the mutex
887
	 * will correctly serialize the operation. We do need to acquire
888
	 * "ep->mtx" after "epmutex" because ep_remove() requires it when called
889
	 * from anywhere but ep_free().
890 891
	 *
	 * Besides, ep_remove() acquires the lock, so we can't hold it here.
Davide Libenzi's avatar
Davide Libenzi committed
892
	 */
893
	mutex_lock(&epmutex);
Davide Libenzi's avatar
Davide Libenzi committed
894

895 896
	while (!list_empty(lsthead)) {
		epi = list_first_entry(lsthead, struct epitem, fllink);
Davide Libenzi's avatar
Davide Libenzi committed
897

898 899
		ep = epi->ep;
		list_del_init(&epi->fllink);
900
		mutex_lock_nested(&ep->mtx, 0);
901
		ep_remove(ep, epi);
902
		mutex_unlock(&ep->mtx);
Davide Libenzi's avatar
Davide Libenzi committed
903 904
	}

905
	mutex_unlock(&epmutex);
Davide Libenzi's avatar
Davide Libenzi committed
906 907
}

908
static int ep_alloc(struct eventpoll **pep)
Linus Torvalds's avatar
Linus Torvalds committed
909
{
910 911 912
	int error;
	struct user_struct *user;
	struct eventpoll *ep;
Linus Torvalds's avatar
Linus Torvalds committed
913

914 915 916 917 918
	user = get_current_user();
	error = -ENOMEM;
	ep = kzalloc(sizeof(*ep), GFP_KERNEL);
	if (unlikely(!ep))
		goto free_uid;
Linus Torvalds's avatar
Linus Torvalds committed
919

920
	spin_lock_init(&ep->lock);
921
	mutex_init(&ep->mtx);
Linus Torvalds's avatar
Linus Torvalds committed
922 923 924 925
	init_waitqueue_head(&ep->wq);
	init_waitqueue_head(&ep->poll_wait);
	INIT_LIST_HEAD(&ep->rdllist);
	ep->rbr = RB_ROOT;
926
	ep->ovflist = EP_UNACTIVE_PTR;
927
	ep->user = user;
Linus Torvalds's avatar
Linus Torvalds committed
928

929
	*pep = ep;
Linus Torvalds's avatar
Linus Torvalds committed
930 931

	return 0;
932 933 934 935

free_uid:
	free_uid(user);
	return error;
Linus Torvalds's avatar
Linus Torvalds committed
936 937 938
}

/*
939 940 941
 * Search the file inside the eventpoll tree. The RB tree operations
 * are protected by the "mtx" mutex, and ep_find() must be called with
 * "mtx" held.
Linus Torvalds's avatar
Linus Torvalds committed
942 943 944 945 946 947 948 949
 */
static struct epitem *ep_find(struct eventpoll *ep, struct file *file, int fd)
{
	int kcmp;
	struct rb_node *rbp;
	struct epitem *epi, *epir = NULL;
	struct epoll_filefd ffd;

950
	ep_set_ffd(&ffd, file, fd);
Linus Torvalds's avatar
Linus Torvalds committed
951 952
	for (rbp = ep->rbr.rb_node; rbp; ) {
		epi = rb_entry(rbp, struct epitem, rbn);
953
		kcmp = ep_cmp_ffd(&ffd, &epi->ffd);
Linus Torvalds's avatar
Linus Torvalds committed
954 955 956 957 958 959 960 961 962 963 964 965 966 967
		if (kcmp > 0)
			rbp = rbp->rb_right;
		else if (kcmp < 0)
			rbp = rbp->rb_left;
		else {
			epir = epi;
			break;
		}
	}

	return epir;
}

/*
968
 * This is the callback that is passed to the wait queue wakeup
969
 * mechanism. It is called by the stored file descriptors when they
970
 * have events to report.
Linus Torvalds's avatar
Linus Torvalds committed
971
 */
972
static int ep_poll_callback(wait_queue_t *wait, unsigned mode, int sync, void *key)
Linus Torvalds's avatar
Linus Torvalds committed
973
{
974 975 976 977
	int pwake = 0;
	unsigned long flags;
	struct epitem *epi = ep_item_from_wait(wait);
	struct eventpoll *ep = epi->ep;
Linus Torvalds's avatar
Linus Torvalds committed
978

979 980 981 982 983 984 985 986
	if ((unsigned long)key & POLLFREE) {
		ep_pwq_from_wait(wait)->whead = NULL;
		/*
		 * whead = NULL above can race with ep_remove_wait_queue()
		 * which can do another remove_wait_queue() after us, so we
		 * can't use __remove_wait_queue(). whead->lock is held by
		 * the caller.
		 */
987
		list_del_init(&wait->task_list);
988
	}
989

990
	spin_lock_irqsave(&ep->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
991

992 993 994 995 996 997 998
	/*
	 * If the event mask does not contain any poll(2) event, we consider the
	 * descriptor to be disabled. This condition is likely the effect of the
	 * EPOLLONESHOT bit that disables the descriptor when an event is received,
	 * until the next EPOLL_CTL_MOD will be issued.
	 */
	if (!(epi->event.events & ~EP_PRIVATE_BITS))
999 1000
		goto out_unlock;

1001 1002 1003 1004 1005 1006 1007 1008 1009
	/*
	 * Check the events coming with the callback. At this stage, not
	 * every device reports the events in the "key" parameter of the
	 * callback. We need to be able to handle both cases here, hence the
	 * test for "key" != NULL before the event match test.
	 */
	if (key && !((unsigned long) key & epi->event.events))
		goto out_unlock;

1010
	/*
1011
	 * If we are transferring events to userspace, we can hold no locks
1012
	 * (because we're accessing user memory, and because of linux f_op->poll()
1013
	 * semantics). All the events that happen during that period of time are
1014 1015 1016 1017 1018 1019
	 * chained in ep->ovflist and requeued later on.
	 */
	if (unlikely(ep->ovflist != EP_UNACTIVE_PTR)) {
		if (epi->next == EP_UNACTIVE_PTR) {
			epi->next = ep->ovflist;
			ep->ovflist = epi;
1020 1021 1022 1023 1024 1025 1026 1027
			if (epi->ws) {
				/*
				 * Activate ep->ws since epi->ws may get
				 * deactivated at any time.
				 */
				__pm_stay_awake(ep->ws);
			}

1028 1029 1030
		}
		goto out_unlock;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1031

1032
	/* If this file is already in the ready list we exit soon */
1033
	if (!ep_is_linked(&epi->rdllink)) {
1034
		list_add_tail(&epi->rdllink, &ep->rdllist);
1035
		ep_pm_stay_awake_rcu(epi);
1036
	}
1037 1038 1039 1040 1041 1042

	/*
	 * Wake up ( if active ) both the eventpoll wait list and the ->poll()
	 * wait list.
	 */
	if (waitqueue_active(&ep->wq))
1043
		wake_up_locked(&ep->wq);
1044 1045 1046
	if (waitqueue_active(&ep->poll_wait))
		pwake++;

1047
out_unlock:
1048
	spin_unlock_irqrestore(&ep->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1049

1050 1051
	/* We have to call this outside the lock */
	if (pwake)
1052
		ep_poll_safewake(&ep->poll_wait);
1053 1054 1055

	return 1;
}
Linus Torvalds's avatar
Linus Torvalds committed
1056 1057 1058 1059 1060 1061 1062 1063

/*
 * This is the callback that is used to add our wait queue to the
 * target file wakeup lists.
 */
static void ep_ptable_queue_proc(struct file *file, wait_queue_head_t *whead,
				 poll_table *pt)
{
1064
	struct epitem *epi = ep_item_from_epqueue(pt);
Linus Torvalds's avatar
Linus Torvalds committed
1065 1066
	struct eppoll_entry *pwq;

1067
	if (epi->nwait >= 0 && (pwq = kmem_cache_alloc(pwq_cache, GFP_KERNEL))) {
Linus Torvalds's avatar
Linus Torvalds committed
1068 1069 1070 1071 1072 1073
		init_waitqueue_func_entry(&pwq->wait, ep_poll_callback);
		pwq->whead = whead;
		pwq->base = epi;
		add_wait_queue(whead, &pwq->wait);
		list_add_tail(&pwq->llink, &epi->pwqlist);
		epi->nwait++;
1074
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
1075 1076
		/* We have to signal that an error occurred */
		epi->nwait = -1;
1077
	}
Linus Torvalds's avatar
Linus Torvalds committed
1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
}

static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi)
{
	int kcmp;
	struct rb_node **p = &ep->rbr.rb_node, *parent = NULL;
	struct epitem *epic;

	while (*p) {
		parent = *p;
		epic = rb_entry(parent, struct epitem, rbn);
1089
		kcmp = ep_cmp_ffd(&epi->ffd, &epic->ffd);
Linus Torvalds's avatar
Linus Torvalds committed
1090 1091 1092 1093 1094 1095 1096 1097 1098
		if (kcmp > 0)
			p = &parent->rb_right;
		else
			p = &parent->rb_left;
	}
	rb_link_node(&epi->rbn, parent, p);
	rb_insert_color(&epi->rbn, &ep->rbr);
}

1099 1100


Jason Baron's avatar
Jason Baron committed
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
#define PATH_ARR_SIZE 5
/*
 * These are the number paths of length 1 to 5, that we are allowing to emanate
 * from a single file of interest. For example, we allow 1000 paths of length
 * 1, to emanate from each file of interest. This essentially represents the
 * potential wakeup paths, which need to be limited in order to avoid massive
 * uncontrolled wakeup storms. The common use case should be a single ep which
 * is connected to n file sources. In this case each file source has 1 path
 * of length 1. Thus, the numbers below should be more than sufficient. These
 * path limits are enforced during an EPOLL_CTL_ADD operation, since a modify
 * and delete can't add additional paths. Protected by the epmutex.
 */
static const int path_limits[PATH_ARR_SIZE] = { 1000, 500, 100, 50, 10 };
static int path_count[PATH_ARR_SIZE];

static int path_count_inc(int nests)
{
1118 1119 1120 1121
	/* Allow an arbitrary number of depth 1 paths */
	if (nests == 0)
		return 0;

Jason Baron's avatar
Jason Baron committed
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
	if (++path_count[nests] > path_limits[nests])
		return -1;
	return 0;
}

static void path_count_init(void)
{
	int i;

	for (i = 0; i < PATH_ARR_SIZE; i++)
		path_count[i] = 0;
}

static int reverse_path_check_proc(void *priv, void *cookie, int call_nests)
{
	int error = 0;
	struct file *file = priv;
	struct file *child_file;
	struct epitem *epi;

	list_for_each_entry(epi, &file->f_ep_links, fllink) {
		child_file = epi->ep->file;
		if (is_file_epoll(child_file)) {
			if (list_empty(&child_file->f_ep_links)) {
				if (path_count_inc(call_nests)) {
					error = -1;
					break;
				}
			} else {
				error = ep_call_nested(&poll_loop_ncalls,
							EP_MAX_NESTS,
							reverse_path_check_proc,
							child_file, child_file,
							current);
			}
			if (error != 0)
				break;
		} else {
			printk(KERN_ERR "reverse_path_check_proc: "
				"file is not an ep!\n");
		}
	}
	return error;
}

/**
 * reverse_path_check - The tfile_check_list is list of file *, which have
 *                      links that are proposed to be newly added. We need to
 *                      make sure that those added links don't add too many
 *                      paths such that we will spend all our time waking up
 *                      eventpoll objects.
 *
 * Returns: Returns zero if the proposed links don't create too many paths,
 *	    -1 otherwise.
 */
static int reverse_path_check(void)
{
	int error = 0;
	struct file *current_file;

	/* let's call this for all tfiles */
	list_for_each_entry(current_file, &tfile_check_list, f_tfile_llink) {
		path_count_init();
		error = ep_call_nested(&poll_loop_ncalls, EP_MAX_NESTS,
					reverse_path_check_proc, current_file,
					current_file, current);
		if (error)
			break;
	}
	return error;
}

1194 1195 1196
static int ep_create_wakeup_source(struct epitem *epi)
{
	const char *name;
1197
	struct wakeup_source *ws;
1198 1199 1200 1201 1202 1203 1204 1205

	if (!epi->ep->ws) {
		epi->ep->ws = wakeup_source_register("eventpoll");
		if (!epi->ep->ws)
			return -ENOMEM;
	}

	name = epi->ffd.file->f_path.dentry->d_name.name;
1206 1207 1208
	ws = wakeup_source_register(name);

	if (!ws)
1209
		return -ENOMEM;
1210
	rcu_assign_pointer(epi->ws, ws);
1211 1212 1213 1214

	return 0;
}

1215 1216
/* rare code path, only used when EPOLL_CTL_MOD removes a wakeup source */
static noinline void ep_destroy_wakeup_source(struct epitem *epi)
1217
{
1218 1219
	struct wakeup_source *ws = ep_wakeup_source(epi);

1220
	RCU_INIT_POINTER(epi->ws, NULL);
1221 1222 1223 1224 1225 1226 1227 1228

	/*
	 * wait for ep_pm_stay_awake_rcu to finish, synchronize_rcu is
	 * used internally by wakeup_source_remove, too (called by
	 * wakeup_source_unregister), so we cannot use call_rcu
	 */
	synchronize_rcu();
	wakeup_source_unregister(ws);
1229 1230
}

1231 1232 1233
/*
 * Must be called with "mtx" held.
 */
Linus Torvalds's avatar
Linus Torvalds committed
1234 1235 1236 1237 1238
static int ep_insert(struct eventpoll *ep, struct epoll_event *event,
		     struct file *tfile, int fd)
{
	int error, revents, pwake = 0;
	unsigned long flags;
1239
	long user_watches;
Linus Torvalds's avatar
Linus Torvalds committed
1240 1241 1242
	struct epitem *epi;
	struct ep_pqueue epq;

1243 1244
	user_watches = atomic_long_read(&ep->user->epoll_watches);
	if (unlikely(user_watches >= max_user_watches))
1245
		return -ENOSPC;
1246
	if (!(epi = kmem_cache_alloc(epi_cache, GFP_KERNEL)))
1247
		return -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
1248 1249 1250 1251 1252 1253

	/* Item initialization follow here ... */
	INIT_LIST_HEAD(&epi->rdllink);
	INIT_LIST_HEAD(&epi->fllink);
	INIT_LIST_HEAD(&epi->pwqlist);
	epi->ep = ep;
1254
	ep_set_ffd(&epi->ffd, tfile, fd);
Linus Torvalds's avatar
Linus Torvalds committed
1255 1256
	epi->event = *event;
	epi->nwait = 0;
1257
	epi->next = EP_UNACTIVE_PTR;
1258 1259 1260 1261 1262
	if (epi->event.events & EPOLLWAKEUP) {
		error = ep_create_wakeup_source(epi);
		if (error)
			goto error_create_wakeup_source;
	} else {
1263
		RCU_INIT_POINTER(epi->ws, NULL);
1264
	}
Linus Torvalds's avatar
Linus Torvalds committed
1265 1266 1267 1268 1269 1270 1271 1272

	/* Initialize the poll table using the queue callback */
	epq.epi = epi;
	init_poll_funcptr(&epq.pt, ep_ptable_queue_proc);

	/*
	 * Attach the item to the poll hooks and get current event bits.
	 * We can safely use the file* here because its usage count has
1273 1274 1275
	 * been increased by the caller of this function. Note that after
	 * this operation completes, the poll callback can start hitting
	 * the new item.
Linus Torvalds's avatar
Linus Torvalds committed
1276
	 */
1277
	revents = ep_item_poll(epi, &epq.pt);
Linus Torvalds's avatar
Linus Torvalds committed
1278 1279 1280 1281 1282 1283

	/*
	 * We have to check if something went wrong during the poll wait queue
	 * install process. Namely an allocation for a wait queue failed due
	 * high memory pressure.
	 */
1284
	error = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
1285
	if (epi->nwait < 0)
1286
		goto error_unregister;
Linus Torvalds's avatar
Linus Torvalds committed
1287 1288

	/* Add the current item to the list of active epoll hook for this file */
1289
	spin_lock(&tfile->f_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1290
	list_add_tail(&epi->fllink, &tfile->f_ep_links);
1291
	spin_unlock(&tfile->f_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1292

1293 1294 1295 1296
	/*
	 * Add the current item to the RB tree. All RB tree operations are
	 * protected by "mtx", and ep_insert() is called with "mtx" held.
	 */
Linus Torvalds's avatar
Linus Torvalds committed
1297 1298
	ep_rbtree_insert(ep, epi);

Jason Baron's avatar
Jason Baron committed
1299 1300 1301 1302 1303
	/* now check if we've created too many backpaths */
	error = -EINVAL;
	if (reverse_path_check())
		goto error_remove_epi;

1304 1305 1306
	/* We have to drop the new item inside our item list to keep track of it */
	spin_lock_irqsave(&ep->lock, flags);

Linus Torvalds's avatar
Linus Torvalds committed
1307
	/* If the file is already "ready" we drop it inside the ready list */
1308
	if ((revents & event->events) && !ep_is_linked(&epi->rdllink)) {
Linus Torvalds's avatar
Linus Torvalds committed
1309
		list_add_tail(&epi->rdllink, &ep->rdllist);
1310
		ep_pm_stay_awake(epi);
Linus Torvalds's avatar
Linus Torvalds committed
1311 1312 1313

		/* Notify waiting tasks that events are available */
		if (waitqueue_active(&ep->wq))
1314
			wake_up_locked(&ep->wq);
Linus Torvalds's avatar
Linus Torvalds committed
1315 1316 1317 1318
		if (waitqueue_active(&ep->poll_wait))
			pwake++;
	}

1319
	spin_unlock_irqrestore(&ep->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1320

1321
	atomic_long_inc(&ep->user->epoll_watches);
1322

Linus Torvalds's avatar
Linus Torvalds committed
1323 1324
	/* We have to call this outside the lock */
	if (pwake)
1325
		ep_poll_safewake(&ep->poll_wait);
Linus Torvalds's avatar
Linus Torvalds committed
1326 1327 1328

	return 0;

Jason Baron's avatar
Jason Baron committed
1329 1330 1331 1332 1333 1334 1335 1336
error_remove_epi:
	spin_lock(&tfile->f_lock);
	if (ep_is_linked(&epi->fllink))
		list_del_init(&epi->fllink);
	spin_unlock(&tfile->f_lock);

	rb_erase(&epi->rbn, &ep->rbr);

1337
error_unregister:
Linus Torvalds's avatar
Linus Torvalds committed
1338 1339 1340 1341
	ep_unregister_pollwait(ep, epi);

	/*
	 * We need to do this because an event could have been arrived on some
1342 1343 1344
	 * allocated wait queue. Note that we don't care about the ep->ovflist
	 * list, since that is used/cleaned only inside a section bound by "mtx".
	 * And ep_insert() is called with "mtx" held.
Linus Torvalds's avatar
Linus Torvalds committed
1345
	 */
1346
	spin_lock_irqsave(&ep->lock, flags);
1347
	if (ep_is_linked(&epi->rdllink))
1348
		list_del_init(&epi->rdllink);
1349
	spin_unlock_irqrestore(&ep->lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1350

1351
	wakeup_source_unregister(ep_wakeup_source(epi));
1352 1353

error_create_wakeup_source:
1354
	kmem_cache_free(epi_cache, epi);
1355

Linus Torvalds's avatar
Linus Torvalds committed
1356 1357 1358 1359 1360
	return error;
}

/*
 * Modify the interest event mask by dropping an event if the new mask
1361
 * has a match in the current file status. Must be called with "mtx" held.
Linus Torvalds's avatar
Linus Torvalds committed
1362 1363 1364 1365 1366
 */
static int ep_modify(struct eventpoll *ep, struct epitem *epi, struct epoll_event *event)
{
	int pwake = 0;
	unsigned int revents;
1367 1368 1369
	poll_table pt;

	init_poll_funcptr(&pt, NULL);
Linus Torvalds's avatar
Linus Torvalds committed
1370 1371

	/*
1372 1373 1374
	 * Set the new event interest mask before calling f_op->poll();
	 * otherwise we might miss an event that happens between the
	 * f_op->poll() call and the new event set registering.
Linus Torvalds's avatar
Linus Torvalds committed
1375
	 */
1376
	epi->event.events = event->events; /* need barrier below */
1377
	epi->event.data = event->data; /* protected by mtx */
1378
	if (epi->event.events & EPOLLWAKEUP) {
1379
		if (!ep_has_wakeup_source(epi))
1380
			ep_create_wakeup_source(epi);
1381
	} else if (ep_has_wakeup_source(epi)) {
1382 1383
		ep_destroy_wakeup_source(epi);
	}
Linus Torvalds's avatar
Linus Torvalds committed
1384

1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404
	/*
	 * The following barrier has two effects:
	 *
	 * 1) Flush epi changes above to other CPUs.  This ensures
	 *    we do not miss events from ep_poll_callback if an
	 *    event occurs immediately after we call f_op->poll().
	 *    We need this because we did not take ep->lock while
	 *    changing epi above (but ep_poll_callback does take
	 *    ep->lock).
	 *
	 * 2) We also need to ensure we do not miss _past_ events
	 *    when calling f_op->poll().  This barrier also
	 *    pairs with the barrier in wq_has_sleeper (see
	 *    comments for wq_has_sleeper).
	 *
	 * This barrier will now guarantee ep_poll_callback or f_op->poll
	 * (or both) will notice the readiness of an item.
	 */
	smp_mb();

Linus Torvalds's avatar
Linus Torvalds committed
1405 1406 1407 1408
	/*
	 * Get current event bits. We can safely use the file* here because
	 * its usage count has been increased by the caller of this function.
	 */
1409
	revents = ep_item_poll(epi, &pt);
Linus Torvalds's avatar
Linus Torvalds committed
1410 1411

	/*
1412
	 * If the item is "hot" and it is not registered inside the ready
1413
	 * list, push it inside.
Linus Torvalds's avatar
Linus Torvalds committed
1414
	 */
1415
	if (revents & event->events) {
1416
		spin_lock_irq(&ep->lock);
1417 1418
		if (!ep_is_linked(&epi->rdllink)) {
			list_add_tail(&epi->rdllink, &ep->rdllist);
1419
			ep_pm_stay_awake(epi);
1420 1421 1422

			/* Notify waiting tasks that events are available */
			if (waitqueue_active(&ep->wq))
1423