inotify_user.c 20.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10
/*
 * fs/inotify_user.c - inotify support for userspace
 *
 * Authors:
 *	John McCutchan	<ttb@tentacle.dhs.org>
 *	Robert Love	<rml@novell.com>
 *
 * Copyright (C) 2005 John McCutchan
 * Copyright 2006 Hewlett-Packard Development Company, L.P.
 *
11 12 13
 * Copyright (C) 2009 Eric Paris <Red Hat Inc>
 * inotify was largely rewriten to make use of the fsnotify infrastructure
 *
14 15 16 17 18 19 20 21 22 23 24 25
 * 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, or (at your option) any
 * later version.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 */

#include <linux/file.h>
26 27 28 29
#include <linux/fs.h> /* struct inode */
#include <linux/fsnotify_backend.h>
#include <linux/idr.h>
#include <linux/init.h> /* module_init */
30
#include <linux/inotify.h>
31 32 33 34
#include <linux/kernel.h> /* roundup() */
#include <linux/namei.h> /* LOOKUP_FOLLOW */
#include <linux/sched.h> /* struct user */
#include <linux/slab.h> /* struct kmem_cache */
35
#include <linux/syscalls.h>
36
#include <linux/types.h>
A
Al Viro 已提交
37
#include <linux/anon_inodes.h>
38 39 40
#include <linux/uaccess.h>
#include <linux/poll.h>
#include <linux/wait.h>
41

42
#include "inotify.h"
43
#include "../fdinfo.h"
44

45
#include <asm/ioctls.h>
46 47

/* these are configurable via /proc/sys/fs/inotify/ */
48 49
static int inotify_max_user_instances __read_mostly;
static int inotify_max_queued_events __read_mostly;
50
static int inotify_max_user_watches __read_mostly;
51

52
static struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
53 54 55 56 57 58 59

#ifdef CONFIG_SYSCTL

#include <linux/sysctl.h>

static int zero;

60
struct ctl_table inotify_table[] = {
61 62 63 64 65
	{
		.procname	= "max_user_instances",
		.data		= &inotify_max_user_instances,
		.maxlen		= sizeof(int),
		.mode		= 0644,
66
		.proc_handler	= proc_dointvec_minmax,
67 68 69 70 71 72 73
		.extra1		= &zero,
	},
	{
		.procname	= "max_user_watches",
		.data		= &inotify_max_user_watches,
		.maxlen		= sizeof(int),
		.mode		= 0644,
74
		.proc_handler	= proc_dointvec_minmax,
75 76 77 78 79 80 81
		.extra1		= &zero,
	},
	{
		.procname	= "max_queued_events",
		.data		= &inotify_max_queued_events,
		.maxlen		= sizeof(int),
		.mode		= 0644,
82
		.proc_handler	= proc_dointvec_minmax,
83 84
		.extra1		= &zero
	},
85
	{ }
86 87 88
};
#endif /* CONFIG_SYSCTL */

89
static inline __u32 inotify_arg_to_mask(u32 arg)
Y
Yan Zheng 已提交
90
{
91
	__u32 mask;
Y
Yan Zheng 已提交
92

E
Eric Paris 已提交
93 94 95 96 97
	/*
	 * everything should accept their own ignored, cares about children,
	 * and should receive events when the inode is unmounted
	 */
	mask = (FS_IN_IGNORED | FS_EVENT_ON_CHILD | FS_UNMOUNT);
98

99
	/* mask off the flags used to open the fd */
100
	mask |= (arg & (IN_ALL_EVENTS | IN_ONESHOT | IN_EXCL_UNLINK));
101

102
	return mask;
103 104
}

105
static inline u32 inotify_mask_to_arg(__u32 mask)
106
{
107 108
	return mask & (IN_ALL_EVENTS | IN_ISDIR | IN_UNMOUNT | IN_IGNORED |
		       IN_Q_OVERFLOW);
109 110
}

111
/* intofiy userspace file descriptor functions */
112 113
static unsigned int inotify_poll(struct file *file, poll_table *wait)
{
114
	struct fsnotify_group *group = file->private_data;
115 116
	int ret = 0;

117 118 119
	poll_wait(file, &group->notification_waitq, wait);
	mutex_lock(&group->notification_mutex);
	if (!fsnotify_notify_queue_is_empty(group))
120
		ret = POLLIN | POLLRDNORM;
121
	mutex_unlock(&group->notification_mutex);
122 123 124 125

	return ret;
}

126
static int round_event_name_len(struct fsnotify_event *fsn_event)
127
{
128 129 130
	struct inotify_event_info *event;

	event = INOTIFY_E(fsn_event);
131 132 133 134 135
	if (!event->name_len)
		return 0;
	return roundup(event->name_len + 1, sizeof(struct inotify_event));
}

136 137 138 139 140
/*
 * Get an inotify_kernel_event if one exists and is small
 * enough to fit in "count". Return an error pointer if
 * not large enough.
 *
141
 * Called with the group->notification_mutex held.
142
 */
143 144
static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
					    size_t count)
145 146
{
	size_t event_size = sizeof(struct inotify_event);
147
	struct fsnotify_event *event;
148

149
	if (fsnotify_notify_queue_is_empty(group))
150 151
		return NULL;

152
	event = fsnotify_peek_first_event(group);
153

E
Eric Paris 已提交
154 155
	pr_debug("%s: group=%p event=%p\n", __func__, group, event);

156
	event_size += round_event_name_len(event);
157 158 159
	if (event_size > count)
		return ERR_PTR(-EINVAL);

160 161
	/* held the notification_mutex the whole time, so this is the
	 * same event we peeked above */
162
	fsnotify_remove_first_event(group);
163 164

	return event;
165 166 167 168 169 170 171 172
}

/*
 * Copy an event to user space, returning how much we copied.
 *
 * We already checked that the event size is smaller than the
 * buffer we had in "get_one_event()" above.
 */
173
static ssize_t copy_event_to_user(struct fsnotify_group *group,
174
				  struct fsnotify_event *fsn_event,
175 176
				  char __user *buf)
{
177
	struct inotify_event inotify_event;
178
	struct inotify_event_info *event;
179
	size_t event_size = sizeof(struct inotify_event);
180 181
	size_t name_len;
	size_t pad_name_len;
182

183
	pr_debug("%s: group=%p event=%p\n", __func__, group, fsn_event);
184

185
	event = INOTIFY_E(fsn_event);
186
	name_len = event->name_len;
187
	/*
188
	 * round up name length so it is a multiple of event_size
189 190
	 * plus an extra byte for the terminating '\0'.
	 */
191
	pad_name_len = round_event_name_len(fsn_event);
192
	inotify_event.len = pad_name_len;
193 194
	inotify_event.mask = inotify_mask_to_arg(fsn_event->mask);
	inotify_event.wd = event->wd;
195
	inotify_event.cookie = event->sync_cookie;
196

197 198
	/* send the main event */
	if (copy_to_user(buf, &inotify_event, event_size))
199 200
		return -EFAULT;

201
	buf += event_size;
202

203 204 205
	/*
	 * fsnotify only stores the pathname, so here we have to send the pathname
	 * and then pad that pathname out to a multiple of sizeof(inotify_event)
206
	 * with zeros.
207
	 */
208
	if (pad_name_len) {
209
		/* copy the path name */
210
		if (copy_to_user(buf, event->name, name_len))
211
			return -EFAULT;
212
		buf += name_len;
213

214
		/* fill userspace with 0's */
215
		if (clear_user(buf, pad_name_len - name_len))
216
			return -EFAULT;
217
		event_size += pad_name_len;
218
	}
219

220 221 222
	return event_size;
}

223 224 225
static ssize_t inotify_read(struct file *file, char __user *buf,
			    size_t count, loff_t *pos)
{
226 227
	struct fsnotify_group *group;
	struct fsnotify_event *kevent;
228 229 230 231 232
	char __user *start;
	int ret;
	DEFINE_WAIT(wait);

	start = buf;
233
	group = file->private_data;
234 235

	while (1) {
236
		prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);
237

238 239 240
		mutex_lock(&group->notification_mutex);
		kevent = get_one_event(group, count);
		mutex_unlock(&group->notification_mutex);
241

E
Eric Paris 已提交
242 243
		pr_debug("%s: group=%p kevent=%p\n", __func__, group, kevent);

244 245 246 247
		if (kevent) {
			ret = PTR_ERR(kevent);
			if (IS_ERR(kevent))
				break;
248
			ret = copy_event_to_user(group, kevent, buf);
249
			fsnotify_destroy_event(group, kevent);
250 251 252 253 254
			if (ret < 0)
				break;
			buf += ret;
			count -= ret;
			continue;
255 256
		}

257 258
		ret = -EAGAIN;
		if (file->f_flags & O_NONBLOCK)
259
			break;
260
		ret = -ERESTARTSYS;
261
		if (signal_pending(current))
262
			break;
263

264
		if (start != buf)
265
			break;
266

267
		schedule();
268 269
	}

270
	finish_wait(&group->notification_waitq, &wait);
271 272
	if (start != buf && ret != -EFAULT)
		ret = buf - start;
273 274 275 276 277
	return ret;
}

static int inotify_release(struct inode *ignored, struct file *file)
{
278
	struct fsnotify_group *group = file->private_data;
279

E
Eric Paris 已提交
280 281
	pr_debug("%s: group=%p\n", __func__, group);

282
	/* free this group, matching get was inotify_init->fsnotify_obtain_group */
283
	fsnotify_destroy_group(group);
284 285 286 287 288 289 290

	return 0;
}

static long inotify_ioctl(struct file *file, unsigned int cmd,
			  unsigned long arg)
{
291
	struct fsnotify_group *group;
292
	struct fsnotify_event *fsn_event;
293 294
	void __user *p;
	int ret = -ENOTTY;
295
	size_t send_len = 0;
296

297
	group = file->private_data;
298 299
	p = (void __user *) arg;

E
Eric Paris 已提交
300 301
	pr_debug("%s: group=%p cmd=%u\n", __func__, group, cmd);

302 303
	switch (cmd) {
	case FIONREAD:
304
		mutex_lock(&group->notification_mutex);
305 306
		list_for_each_entry(fsn_event, &group->notification_list,
				    list) {
307
			send_len += sizeof(struct inotify_event);
308
			send_len += round_event_name_len(fsn_event);
309 310 311
		}
		mutex_unlock(&group->notification_mutex);
		ret = put_user(send_len, (int __user *) p);
312 313 314 315 316 317 318
		break;
	}

	return ret;
}

static const struct file_operations inotify_fops = {
319
	.show_fdinfo	= inotify_show_fdinfo,
320 321
	.poll		= inotify_poll,
	.read		= inotify_read,
322
	.fasync		= fsnotify_fasync,
323 324
	.release	= inotify_release,
	.unlocked_ioctl	= inotify_ioctl,
325
	.compat_ioctl	= inotify_ioctl,
326
	.llseek		= noop_llseek,
327 328 329
};


330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
/*
 * find_inode - resolve a user-given path to a specific inode
 */
static int inotify_find_inode(const char __user *dirname, struct path *path, unsigned flags)
{
	int error;

	error = user_path_at(AT_FDCWD, dirname, flags, path);
	if (error)
		return error;
	/* you can only watch an inode if you have read permissions on it */
	error = inode_permission(path->dentry->d_inode, MAY_READ);
	if (error)
		path_put(path);
	return error;
}

347
static int inotify_add_to_idr(struct idr *idr, spinlock_t *idr_lock,
348
			      struct inotify_inode_mark *i_mark)
349 350 351
{
	int ret;

T
Tejun Heo 已提交
352 353
	idr_preload(GFP_KERNEL);
	spin_lock(idr_lock);
354

355
	ret = idr_alloc_cyclic(idr, i_mark, 1, 0, GFP_NOWAIT);
T
Tejun Heo 已提交
356
	if (ret >= 0) {
357
		/* we added the mark to the idr, take a reference */
T
Tejun Heo 已提交
358 359 360
		i_mark->wd = ret;
		fsnotify_get_mark(&i_mark->fsn_mark);
	}
361

T
Tejun Heo 已提交
362 363 364
	spin_unlock(idr_lock);
	idr_preload_end();
	return ret < 0 ? ret : 0;
365 366
}

367
static struct inotify_inode_mark *inotify_idr_find_locked(struct fsnotify_group *group,
368 369 370 371
								int wd)
{
	struct idr *idr = &group->inotify_data.idr;
	spinlock_t *idr_lock = &group->inotify_data.idr_lock;
372
	struct inotify_inode_mark *i_mark;
373 374 375

	assert_spin_locked(idr_lock);

376 377 378
	i_mark = idr_find(idr, wd);
	if (i_mark) {
		struct fsnotify_mark *fsn_mark = &i_mark->fsn_mark;
379

380
		fsnotify_get_mark(fsn_mark);
381
		/* One ref for being in the idr, one ref we just took */
382
		BUG_ON(atomic_read(&fsn_mark->refcnt) < 2);
383 384
	}

385
	return i_mark;
386 387
}

388
static struct inotify_inode_mark *inotify_idr_find(struct fsnotify_group *group,
389 390
							 int wd)
{
391
	struct inotify_inode_mark *i_mark;
392 393 394
	spinlock_t *idr_lock = &group->inotify_data.idr_lock;

	spin_lock(idr_lock);
395
	i_mark = inotify_idr_find_locked(group, wd);
396 397
	spin_unlock(idr_lock);

398
	return i_mark;
399 400 401
}

static void do_inotify_remove_from_idr(struct fsnotify_group *group,
402
				       struct inotify_inode_mark *i_mark)
403 404 405
{
	struct idr *idr = &group->inotify_data.idr;
	spinlock_t *idr_lock = &group->inotify_data.idr_lock;
406
	int wd = i_mark->wd;
407 408 409 410 411 412

	assert_spin_locked(idr_lock);

	idr_remove(idr, wd);

	/* removed from the idr, drop that ref */
413
	fsnotify_put_mark(&i_mark->fsn_mark);
414 415
}

416 417 418 419
/*
 * Remove the mark from the idr (if present) and drop the reference
 * on the mark because it was in the idr.
 */
420
static void inotify_remove_from_idr(struct fsnotify_group *group,
421
				    struct inotify_inode_mark *i_mark)
422
{
423
	spinlock_t *idr_lock = &group->inotify_data.idr_lock;
424
	struct inotify_inode_mark *found_i_mark = NULL;
425
	int wd;
426

427
	spin_lock(idr_lock);
428
	wd = i_mark->wd;
429

430
	/*
431
	 * does this i_mark think it is in the idr?  we shouldn't get called
432 433 434
	 * if it wasn't....
	 */
	if (wd == -1) {
435 436 437
		WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
438
		goto out;
439
	}
440

441
	/* Lets look in the idr to see if we find it */
442 443 444 445 446
	found_i_mark = inotify_idr_find_locked(group, wd);
	if (unlikely(!found_i_mark)) {
		WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
447
		goto out;
448
	}
449

450
	/*
451 452
	 * We found an mark in the idr at the right wd, but it's
	 * not the mark we were told to remove.  eparis seriously
453 454
	 * fucked up somewhere.
	 */
455 456 457 458 459 460 461 462
	if (unlikely(found_i_mark != i_mark)) {
		WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p "
			"mark->inode=%p found_i_mark=%p found_i_mark->wd=%d "
			"found_i_mark->group=%p found_i_mark->inode=%p\n",
			__func__, i_mark, i_mark->wd, i_mark->fsn_mark.group,
			i_mark->fsn_mark.i.inode, found_i_mark, found_i_mark->wd,
			found_i_mark->fsn_mark.group,
			found_i_mark->fsn_mark.i.inode);
463 464 465
		goto out;
	}

466 467 468 469 470
	/*
	 * One ref for being in the idr
	 * one ref held by the caller trying to kill us
	 * one ref grabbed by inotify_idr_find
	 */
471 472 473 474
	if (unlikely(atomic_read(&i_mark->fsn_mark.refcnt) < 3)) {
		printk(KERN_ERR "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
475 476 477
		/* we can't really recover with bad ref cnting.. */
		BUG();
	}
478

479
	do_inotify_remove_from_idr(group, i_mark);
480
out:
481
	/* match the ref taken by inotify_idr_find_locked() */
482 483 484
	if (found_i_mark)
		fsnotify_put_mark(&found_i_mark->fsn_mark);
	i_mark->wd = -1;
485
	spin_unlock(idr_lock);
486
}
487

488
/*
489
 * Send IN_IGNORED for this wd, remove this wd from the idr.
490
 */
491
void inotify_ignored_and_remove_idr(struct fsnotify_mark *fsn_mark,
492
				    struct fsnotify_group *group)
493
{
494
	struct inotify_inode_mark *i_mark;
495

496 497
	/* Queue ignore event for the watch */
	inotify_handle_event(group, NULL, fsn_mark, NULL, FS_IN_IGNORED,
498
			     NULL, FSNOTIFY_EVENT_NONE, NULL, 0);
499

500
	i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
501 502
	/* remove this mark from the idr */
	inotify_remove_from_idr(group, i_mark);
503

504
	atomic_dec(&group->inotify_data.user->inotify_watches);
505 506 507
}

/* ding dong the mark is dead */
508
static void inotify_free_mark(struct fsnotify_mark *fsn_mark)
509
{
510
	struct inotify_inode_mark *i_mark;
511

512
	i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
513

514
	kmem_cache_free(inotify_inode_mark_cachep, i_mark);
515 516
}

517 518 519
static int inotify_update_existing_watch(struct fsnotify_group *group,
					 struct inode *inode,
					 u32 arg)
520
{
521 522
	struct fsnotify_mark *fsn_mark;
	struct inotify_inode_mark *i_mark;
523
	__u32 old_mask, new_mask;
524 525 526
	__u32 mask;
	int add = (arg & IN_MASK_ADD);
	int ret;
527 528 529

	mask = inotify_arg_to_mask(arg);

530
	fsn_mark = fsnotify_find_inode_mark(group, inode);
531
	if (!fsn_mark)
532
		return -ENOENT;
533

534
	i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
535

536
	spin_lock(&fsn_mark->lock);
537

538
	old_mask = fsn_mark->mask;
539 540 541 542 543
	if (add)
		fsnotify_set_mark_mask_locked(fsn_mark, (fsn_mark->mask | mask));
	else
		fsnotify_set_mark_mask_locked(fsn_mark, mask);
	new_mask = fsn_mark->mask;
544

545
	spin_unlock(&fsn_mark->lock);
546 547 548 549

	if (old_mask != new_mask) {
		/* more bits in old than in new? */
		int dropped = (old_mask & ~new_mask);
550
		/* more bits in this fsn_mark than the inode's mask? */
551 552
		int do_inode = (new_mask & ~inode->i_fsnotify_mask);

553
		/* update the inode with this new fsn_mark */
554 555 556 557 558
		if (dropped || do_inode)
			fsnotify_recalc_inode_mask(inode);

	}

559
	/* return the wd */
560
	ret = i_mark->wd;
561

562
	/* match the get from fsnotify_find_mark() */
563
	fsnotify_put_mark(fsn_mark);
564

565 566 567 568 569 570 571
	return ret;
}

static int inotify_new_watch(struct fsnotify_group *group,
			     struct inode *inode,
			     u32 arg)
{
572
	struct inotify_inode_mark *tmp_i_mark;
573 574
	__u32 mask;
	int ret;
575 576
	struct idr *idr = &group->inotify_data.idr;
	spinlock_t *idr_lock = &group->inotify_data.idr_lock;
577 578 579

	mask = inotify_arg_to_mask(arg);

580 581
	tmp_i_mark = kmem_cache_alloc(inotify_inode_mark_cachep, GFP_KERNEL);
	if (unlikely(!tmp_i_mark))
582 583
		return -ENOMEM;

584 585 586
	fsnotify_init_mark(&tmp_i_mark->fsn_mark, inotify_free_mark);
	tmp_i_mark->fsn_mark.mask = mask;
	tmp_i_mark->wd = -1;
587 588 589 590

	ret = -ENOSPC;
	if (atomic_read(&group->inotify_data.user->inotify_watches) >= inotify_max_user_watches)
		goto out_err;
591

592
	ret = inotify_add_to_idr(idr, idr_lock, tmp_i_mark);
593
	if (ret)
594 595 596
		goto out_err;

	/* we are on the idr, now get on the inode */
597 598
	ret = fsnotify_add_mark_locked(&tmp_i_mark->fsn_mark, group, inode,
				       NULL, 0);
599 600
	if (ret) {
		/* we failed to get on the inode, get off the idr */
601
		inotify_remove_from_idr(group, tmp_i_mark);
602 603 604 605 606 607
		goto out_err;
	}

	/* increment the number of watches the user has */
	atomic_inc(&group->inotify_data.user->inotify_watches);

608 609
	/* return the watch descriptor for this new mark */
	ret = tmp_i_mark->wd;
610

611
out_err:
612 613
	/* match the ref from fsnotify_init_mark() */
	fsnotify_put_mark(&tmp_i_mark->fsn_mark);
614 615 616 617 618 619 620 621

	return ret;
}

static int inotify_update_watch(struct fsnotify_group *group, struct inode *inode, u32 arg)
{
	int ret = 0;

622
	mutex_lock(&group->mark_mutex);
623 624 625 626 627
	/* try to update and existing watch with the new arg */
	ret = inotify_update_existing_watch(group, inode, arg);
	/* no mark present, try to add a new one */
	if (ret == -ENOENT)
		ret = inotify_new_watch(group, inode, arg);
628
	mutex_unlock(&group->mark_mutex);
629

630 631 632
	return ret;
}

633
static struct fsnotify_group *inotify_new_group(unsigned int max_events)
634 635
{
	struct fsnotify_group *group;
636
	struct inotify_event_info *oevent;
637

638
	group = fsnotify_alloc_group(&inotify_fsnotify_ops);
639 640 641
	if (IS_ERR(group))
		return group;

642 643 644 645 646 647 648 649 650 651 652
	oevent = kmalloc(sizeof(struct inotify_event_info), GFP_KERNEL);
	if (unlikely(!oevent)) {
		fsnotify_destroy_group(group);
		return ERR_PTR(-ENOMEM);
	}
	group->overflow_event = &oevent->fse;
	fsnotify_init_event(group->overflow_event, NULL, FS_Q_OVERFLOW);
	oevent->wd = -1;
	oevent->sync_cookie = 0;
	oevent->name_len = 0;

653 654 655 656
	group->max_events = max_events;

	spin_lock_init(&group->inotify_data.idr_lock);
	idr_init(&group->inotify_data.idr);
657 658 659 660
	group->inotify_data.user = get_current_user();

	if (atomic_inc_return(&group->inotify_data.user->inotify_devs) >
	    inotify_max_user_instances) {
661
		fsnotify_destroy_group(group);
662 663
		return ERR_PTR(-EMFILE);
	}
664 665 666 667 668 669

	return group;
}


/* inotify syscalls */
670
SYSCALL_DEFINE1(inotify_init1, int, flags)
671
{
672
	struct fsnotify_group *group;
A
Al Viro 已提交
673
	int ret;
674

675 676 677 678
	/* Check the IN_* constants for consistency.  */
	BUILD_BUG_ON(IN_CLOEXEC != O_CLOEXEC);
	BUILD_BUG_ON(IN_NONBLOCK != O_NONBLOCK);

679
	if (flags & ~(IN_CLOEXEC | IN_NONBLOCK))
U
Ulrich Drepper 已提交
680 681
		return -EINVAL;

682
	/* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */
683 684 685
	group = inotify_new_group(inotify_max_queued_events);
	if (IS_ERR(group))
		return PTR_ERR(group);
A
Al Viro 已提交
686

A
Al Viro 已提交
687 688
	ret = anon_inode_getfd("inotify", &inotify_fops, group,
				  O_RDONLY | flags);
689
	if (ret < 0)
690
		fsnotify_destroy_group(group);
A
Al Viro 已提交
691

692 693 694
	return ret;
}

695
SYSCALL_DEFINE0(inotify_init)
U
Ulrich Drepper 已提交
696 697 698 699
{
	return sys_inotify_init1(0);
}

700 701
SYSCALL_DEFINE3(inotify_add_watch, int, fd, const char __user *, pathname,
		u32, mask)
702
{
703
	struct fsnotify_group *group;
704
	struct inode *inode;
705
	struct path path;
706 707
	struct fd f;
	int ret;
708 709
	unsigned flags = 0;

710 711 712 713
	/* don't allow invalid bits: we don't want flags set */
	if (unlikely(!(mask & ALL_INOTIFY_BITS)))
		return -EINVAL;

714 715
	f = fdget(fd);
	if (unlikely(!f.file))
716 717 718
		return -EBADF;

	/* verify that this is indeed an inotify instance */
719
	if (unlikely(f.file->f_op != &inotify_fops)) {
720 721 722 723 724 725 726 727 728
		ret = -EINVAL;
		goto fput_and_out;
	}

	if (!(mask & IN_DONT_FOLLOW))
		flags |= LOOKUP_FOLLOW;
	if (mask & IN_ONLYDIR)
		flags |= LOOKUP_DIRECTORY;

729 730
	ret = inotify_find_inode(pathname, &path, flags);
	if (ret)
731 732
		goto fput_and_out;

733
	/* inode held in place by reference to path; group by fget on fd */
734
	inode = path.dentry->d_inode;
735
	group = f.file->private_data;
736

737 738
	/* create/update an inode mark */
	ret = inotify_update_watch(group, inode, mask);
739
	path_put(&path);
740
fput_and_out:
741
	fdput(f);
742 743 744
	return ret;
}

745
SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd)
746
{
747
	struct fsnotify_group *group;
748
	struct inotify_inode_mark *i_mark;
749 750
	struct fd f;
	int ret = 0;
751

752 753
	f = fdget(fd);
	if (unlikely(!f.file))
754 755 756
		return -EBADF;

	/* verify that this is indeed an inotify instance */
757
	ret = -EINVAL;
758
	if (unlikely(f.file->f_op != &inotify_fops))
759 760
		goto out;

761
	group = f.file->private_data;
762

763
	ret = -EINVAL;
764 765
	i_mark = inotify_idr_find(group, wd);
	if (unlikely(!i_mark))
766 767
		goto out;

768 769
	ret = 0;

770
	fsnotify_destroy_mark(&i_mark->fsn_mark, group);
771 772

	/* match ref taken by inotify_idr_find */
773
	fsnotify_put_mark(&i_mark->fsn_mark);
774 775

out:
776
	fdput(f);
777 778 779 780
	return ret;
}

/*
781
 * inotify_user_setup - Our initialization function.  Note that we cannot return
782 783 784 785 786
 * error because we have compiled-in VFS hooks.  So an (unlikely) failure here
 * must result in panic().
 */
static int __init inotify_user_setup(void)
{
787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802
	BUILD_BUG_ON(IN_ACCESS != FS_ACCESS);
	BUILD_BUG_ON(IN_MODIFY != FS_MODIFY);
	BUILD_BUG_ON(IN_ATTRIB != FS_ATTRIB);
	BUILD_BUG_ON(IN_CLOSE_WRITE != FS_CLOSE_WRITE);
	BUILD_BUG_ON(IN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
	BUILD_BUG_ON(IN_OPEN != FS_OPEN);
	BUILD_BUG_ON(IN_MOVED_FROM != FS_MOVED_FROM);
	BUILD_BUG_ON(IN_MOVED_TO != FS_MOVED_TO);
	BUILD_BUG_ON(IN_CREATE != FS_CREATE);
	BUILD_BUG_ON(IN_DELETE != FS_DELETE);
	BUILD_BUG_ON(IN_DELETE_SELF != FS_DELETE_SELF);
	BUILD_BUG_ON(IN_MOVE_SELF != FS_MOVE_SELF);
	BUILD_BUG_ON(IN_UNMOUNT != FS_UNMOUNT);
	BUILD_BUG_ON(IN_Q_OVERFLOW != FS_Q_OVERFLOW);
	BUILD_BUG_ON(IN_IGNORED != FS_IN_IGNORED);
	BUILD_BUG_ON(IN_EXCL_UNLINK != FS_EXCL_UNLINK);
803
	BUILD_BUG_ON(IN_ISDIR != FS_ISDIR);
804 805 806 807
	BUILD_BUG_ON(IN_ONESHOT != FS_IN_ONESHOT);

	BUG_ON(hweight32(ALL_INOTIFY_BITS) != 21);

808
	inotify_inode_mark_cachep = KMEM_CACHE(inotify_inode_mark, SLAB_PANIC);
809

810 811 812 813 814 815 816
	inotify_max_queued_events = 16384;
	inotify_max_user_instances = 128;
	inotify_max_user_watches = 8192;

	return 0;
}
module_init(inotify_user_setup);