xref: /openbmc/linux/fs/btrfs/super.c (revision 6774def6)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/blkdev.h>
20 #include <linux/module.h>
21 #include <linux/buffer_head.h>
22 #include <linux/fs.h>
23 #include <linux/pagemap.h>
24 #include <linux/highmem.h>
25 #include <linux/time.h>
26 #include <linux/init.h>
27 #include <linux/seq_file.h>
28 #include <linux/string.h>
29 #include <linux/backing-dev.h>
30 #include <linux/mount.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/parser.h>
37 #include <linux/ctype.h>
38 #include <linux/namei.h>
39 #include <linux/miscdevice.h>
40 #include <linux/magic.h>
41 #include <linux/slab.h>
42 #include <linux/cleancache.h>
43 #include <linux/ratelimit.h>
44 #include <linux/btrfs.h>
45 #include "delayed-inode.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
51 #include "hash.h"
52 #include "props.h"
53 #include "xattr.h"
54 #include "volumes.h"
55 #include "export.h"
56 #include "compression.h"
57 #include "rcu-string.h"
58 #include "dev-replace.h"
59 #include "free-space-cache.h"
60 #include "backref.h"
61 #include "tests/btrfs-tests.h"
62 
63 #include "qgroup.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66 
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69 
70 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
71 
72 static const char *btrfs_decode_error(int errno)
73 {
74 	char *errstr = "unknown";
75 
76 	switch (errno) {
77 	case -EIO:
78 		errstr = "IO failure";
79 		break;
80 	case -ENOMEM:
81 		errstr = "Out of memory";
82 		break;
83 	case -EROFS:
84 		errstr = "Readonly filesystem";
85 		break;
86 	case -EEXIST:
87 		errstr = "Object already exists";
88 		break;
89 	case -ENOSPC:
90 		errstr = "No space left";
91 		break;
92 	case -ENOENT:
93 		errstr = "No such entry";
94 		break;
95 	}
96 
97 	return errstr;
98 }
99 
100 static void save_error_info(struct btrfs_fs_info *fs_info)
101 {
102 	/*
103 	 * today we only save the error info into ram.  Long term we'll
104 	 * also send it down to the disk
105 	 */
106 	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
107 }
108 
109 /* btrfs handle error by forcing the filesystem readonly */
110 static void btrfs_handle_error(struct btrfs_fs_info *fs_info)
111 {
112 	struct super_block *sb = fs_info->sb;
113 
114 	if (sb->s_flags & MS_RDONLY)
115 		return;
116 
117 	if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
118 		sb->s_flags |= MS_RDONLY;
119 		btrfs_info(fs_info, "forced readonly");
120 		/*
121 		 * Note that a running device replace operation is not
122 		 * canceled here although there is no way to update
123 		 * the progress. It would add the risk of a deadlock,
124 		 * therefore the canceling is ommited. The only penalty
125 		 * is that some I/O remains active until the procedure
126 		 * completes. The next time when the filesystem is
127 		 * mounted writeable again, the device replace
128 		 * operation continues.
129 		 */
130 	}
131 }
132 
133 #ifdef CONFIG_PRINTK
134 /*
135  * __btrfs_std_error decodes expected errors from the caller and
136  * invokes the approciate error response.
137  */
138 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
139 		       unsigned int line, int errno, const char *fmt, ...)
140 {
141 	struct super_block *sb = fs_info->sb;
142 	const char *errstr;
143 
144 	/*
145 	 * Special case: if the error is EROFS, and we're already
146 	 * under MS_RDONLY, then it is safe here.
147 	 */
148 	if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
149   		return;
150 
151 	errstr = btrfs_decode_error(errno);
152 	if (fmt) {
153 		struct va_format vaf;
154 		va_list args;
155 
156 		va_start(args, fmt);
157 		vaf.fmt = fmt;
158 		vaf.va = &args;
159 
160 		printk(KERN_CRIT
161 			"BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
162 			sb->s_id, function, line, errno, errstr, &vaf);
163 		va_end(args);
164 	} else {
165 		printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
166 			sb->s_id, function, line, errno, errstr);
167 	}
168 
169 	/* Don't go through full error handling during mount */
170 	save_error_info(fs_info);
171 	if (sb->s_flags & MS_BORN)
172 		btrfs_handle_error(fs_info);
173 }
174 
175 static const char * const logtypes[] = {
176 	"emergency",
177 	"alert",
178 	"critical",
179 	"error",
180 	"warning",
181 	"notice",
182 	"info",
183 	"debug",
184 };
185 
186 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
187 {
188 	struct super_block *sb = fs_info->sb;
189 	char lvl[4];
190 	struct va_format vaf;
191 	va_list args;
192 	const char *type = logtypes[4];
193 	int kern_level;
194 
195 	va_start(args, fmt);
196 
197 	kern_level = printk_get_level(fmt);
198 	if (kern_level) {
199 		size_t size = printk_skip_level(fmt) - fmt;
200 		memcpy(lvl, fmt,  size);
201 		lvl[size] = '\0';
202 		fmt += size;
203 		type = logtypes[kern_level - '0'];
204 	} else
205 		*lvl = '\0';
206 
207 	vaf.fmt = fmt;
208 	vaf.va = &args;
209 
210 	printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf);
211 
212 	va_end(args);
213 }
214 
215 #else
216 
217 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function,
218 		       unsigned int line, int errno, const char *fmt, ...)
219 {
220 	struct super_block *sb = fs_info->sb;
221 
222 	/*
223 	 * Special case: if the error is EROFS, and we're already
224 	 * under MS_RDONLY, then it is safe here.
225 	 */
226 	if (errno == -EROFS && (sb->s_flags & MS_RDONLY))
227 		return;
228 
229 	/* Don't go through full error handling during mount */
230 	if (sb->s_flags & MS_BORN) {
231 		save_error_info(fs_info);
232 		btrfs_handle_error(fs_info);
233 	}
234 }
235 #endif
236 
237 /*
238  * We only mark the transaction aborted and then set the file system read-only.
239  * This will prevent new transactions from starting or trying to join this
240  * one.
241  *
242  * This means that error recovery at the call site is limited to freeing
243  * any local memory allocations and passing the error code up without
244  * further cleanup. The transaction should complete as it normally would
245  * in the call path but will return -EIO.
246  *
247  * We'll complete the cleanup in btrfs_end_transaction and
248  * btrfs_commit_transaction.
249  */
250 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
251 			       struct btrfs_root *root, const char *function,
252 			       unsigned int line, int errno)
253 {
254 	/*
255 	 * Report first abort since mount
256 	 */
257 	if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED,
258 				&root->fs_info->fs_state)) {
259 		WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n",
260 				errno);
261 	}
262 	trans->aborted = errno;
263 	/* Nothing used. The other threads that have joined this
264 	 * transaction may be able to continue. */
265 	if (!trans->blocks_used) {
266 		const char *errstr;
267 
268 		errstr = btrfs_decode_error(errno);
269 		btrfs_warn(root->fs_info,
270 		           "%s:%d: Aborting unused transaction(%s).",
271 		           function, line, errstr);
272 		return;
273 	}
274 	ACCESS_ONCE(trans->transaction->aborted) = errno;
275 	/* Wake up anybody who may be waiting on this transaction */
276 	wake_up(&root->fs_info->transaction_wait);
277 	wake_up(&root->fs_info->transaction_blocked_wait);
278 	__btrfs_std_error(root->fs_info, function, line, errno, NULL);
279 }
280 /*
281  * __btrfs_panic decodes unexpected, fatal errors from the caller,
282  * issues an alert, and either panics or BUGs, depending on mount options.
283  */
284 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
285 		   unsigned int line, int errno, const char *fmt, ...)
286 {
287 	char *s_id = "<unknown>";
288 	const char *errstr;
289 	struct va_format vaf = { .fmt = fmt };
290 	va_list args;
291 
292 	if (fs_info)
293 		s_id = fs_info->sb->s_id;
294 
295 	va_start(args, fmt);
296 	vaf.va = &args;
297 
298 	errstr = btrfs_decode_error(errno);
299 	if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR))
300 		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
301 			s_id, function, line, &vaf, errno, errstr);
302 
303 	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
304 		   function, line, &vaf, errno, errstr);
305 	va_end(args);
306 	/* Caller calls BUG() */
307 }
308 
309 static void btrfs_put_super(struct super_block *sb)
310 {
311 	close_ctree(btrfs_sb(sb)->tree_root);
312 }
313 
314 enum {
315 	Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
316 	Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
317 	Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
318 	Opt_compress_type, Opt_compress_force, Opt_compress_force_type,
319 	Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
320 	Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
321 	Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache,
322 	Opt_no_space_cache, Opt_recovery, Opt_skip_balance,
323 	Opt_check_integrity, Opt_check_integrity_including_extent_data,
324 	Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree,
325 	Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard,
326 	Opt_noenospc_debug, Opt_noflushoncommit, Opt_acl, Opt_datacow,
327 	Opt_datasum, Opt_treelog, Opt_noinode_cache,
328 	Opt_err,
329 };
330 
331 static match_table_t tokens = {
332 	{Opt_degraded, "degraded"},
333 	{Opt_subvol, "subvol=%s"},
334 	{Opt_subvolid, "subvolid=%s"},
335 	{Opt_device, "device=%s"},
336 	{Opt_nodatasum, "nodatasum"},
337 	{Opt_datasum, "datasum"},
338 	{Opt_nodatacow, "nodatacow"},
339 	{Opt_datacow, "datacow"},
340 	{Opt_nobarrier, "nobarrier"},
341 	{Opt_barrier, "barrier"},
342 	{Opt_max_inline, "max_inline=%s"},
343 	{Opt_alloc_start, "alloc_start=%s"},
344 	{Opt_thread_pool, "thread_pool=%d"},
345 	{Opt_compress, "compress"},
346 	{Opt_compress_type, "compress=%s"},
347 	{Opt_compress_force, "compress-force"},
348 	{Opt_compress_force_type, "compress-force=%s"},
349 	{Opt_ssd, "ssd"},
350 	{Opt_ssd_spread, "ssd_spread"},
351 	{Opt_nossd, "nossd"},
352 	{Opt_acl, "acl"},
353 	{Opt_noacl, "noacl"},
354 	{Opt_notreelog, "notreelog"},
355 	{Opt_treelog, "treelog"},
356 	{Opt_flushoncommit, "flushoncommit"},
357 	{Opt_noflushoncommit, "noflushoncommit"},
358 	{Opt_ratio, "metadata_ratio=%d"},
359 	{Opt_discard, "discard"},
360 	{Opt_nodiscard, "nodiscard"},
361 	{Opt_space_cache, "space_cache"},
362 	{Opt_clear_cache, "clear_cache"},
363 	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
364 	{Opt_enospc_debug, "enospc_debug"},
365 	{Opt_noenospc_debug, "noenospc_debug"},
366 	{Opt_subvolrootid, "subvolrootid=%d"},
367 	{Opt_defrag, "autodefrag"},
368 	{Opt_nodefrag, "noautodefrag"},
369 	{Opt_inode_cache, "inode_cache"},
370 	{Opt_noinode_cache, "noinode_cache"},
371 	{Opt_no_space_cache, "nospace_cache"},
372 	{Opt_recovery, "recovery"},
373 	{Opt_skip_balance, "skip_balance"},
374 	{Opt_check_integrity, "check_int"},
375 	{Opt_check_integrity_including_extent_data, "check_int_data"},
376 	{Opt_check_integrity_print_mask, "check_int_print_mask=%d"},
377 	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
378 	{Opt_fatal_errors, "fatal_errors=%s"},
379 	{Opt_commit_interval, "commit=%d"},
380 	{Opt_err, NULL},
381 };
382 
383 /*
384  * Regular mount options parser.  Everything that is needed only when
385  * reading in a new superblock is parsed here.
386  * XXX JDM: This needs to be cleaned up for remount.
387  */
388 int btrfs_parse_options(struct btrfs_root *root, char *options)
389 {
390 	struct btrfs_fs_info *info = root->fs_info;
391 	substring_t args[MAX_OPT_ARGS];
392 	char *p, *num, *orig = NULL;
393 	u64 cache_gen;
394 	int intarg;
395 	int ret = 0;
396 	char *compress_type;
397 	bool compress_force = false;
398 
399 	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
400 	if (cache_gen)
401 		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
402 
403 	if (!options)
404 		goto out;
405 
406 	/*
407 	 * strsep changes the string, duplicate it because parse_options
408 	 * gets called twice
409 	 */
410 	options = kstrdup(options, GFP_NOFS);
411 	if (!options)
412 		return -ENOMEM;
413 
414 	orig = options;
415 
416 	while ((p = strsep(&options, ",")) != NULL) {
417 		int token;
418 		if (!*p)
419 			continue;
420 
421 		token = match_token(p, tokens, args);
422 		switch (token) {
423 		case Opt_degraded:
424 			btrfs_info(root->fs_info, "allowing degraded mounts");
425 			btrfs_set_opt(info->mount_opt, DEGRADED);
426 			break;
427 		case Opt_subvol:
428 		case Opt_subvolid:
429 		case Opt_subvolrootid:
430 		case Opt_device:
431 			/*
432 			 * These are parsed by btrfs_parse_early_options
433 			 * and can be happily ignored here.
434 			 */
435 			break;
436 		case Opt_nodatasum:
437 			btrfs_set_and_info(root, NODATASUM,
438 					   "setting nodatasum");
439 			break;
440 		case Opt_datasum:
441 			if (btrfs_test_opt(root, NODATASUM)) {
442 				if (btrfs_test_opt(root, NODATACOW))
443 					btrfs_info(root->fs_info, "setting datasum, datacow enabled");
444 				else
445 					btrfs_info(root->fs_info, "setting datasum");
446 			}
447 			btrfs_clear_opt(info->mount_opt, NODATACOW);
448 			btrfs_clear_opt(info->mount_opt, NODATASUM);
449 			break;
450 		case Opt_nodatacow:
451 			if (!btrfs_test_opt(root, NODATACOW)) {
452 				if (!btrfs_test_opt(root, COMPRESS) ||
453 				    !btrfs_test_opt(root, FORCE_COMPRESS)) {
454 					btrfs_info(root->fs_info,
455 						   "setting nodatacow, compression disabled");
456 				} else {
457 					btrfs_info(root->fs_info, "setting nodatacow");
458 				}
459 			}
460 			btrfs_clear_opt(info->mount_opt, COMPRESS);
461 			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
462 			btrfs_set_opt(info->mount_opt, NODATACOW);
463 			btrfs_set_opt(info->mount_opt, NODATASUM);
464 			break;
465 		case Opt_datacow:
466 			btrfs_clear_and_info(root, NODATACOW,
467 					     "setting datacow");
468 			break;
469 		case Opt_compress_force:
470 		case Opt_compress_force_type:
471 			compress_force = true;
472 			/* Fallthrough */
473 		case Opt_compress:
474 		case Opt_compress_type:
475 			if (token == Opt_compress ||
476 			    token == Opt_compress_force ||
477 			    strcmp(args[0].from, "zlib") == 0) {
478 				compress_type = "zlib";
479 				info->compress_type = BTRFS_COMPRESS_ZLIB;
480 				btrfs_set_opt(info->mount_opt, COMPRESS);
481 				btrfs_clear_opt(info->mount_opt, NODATACOW);
482 				btrfs_clear_opt(info->mount_opt, NODATASUM);
483 			} else if (strcmp(args[0].from, "lzo") == 0) {
484 				compress_type = "lzo";
485 				info->compress_type = BTRFS_COMPRESS_LZO;
486 				btrfs_set_opt(info->mount_opt, COMPRESS);
487 				btrfs_clear_opt(info->mount_opt, NODATACOW);
488 				btrfs_clear_opt(info->mount_opt, NODATASUM);
489 				btrfs_set_fs_incompat(info, COMPRESS_LZO);
490 			} else if (strncmp(args[0].from, "no", 2) == 0) {
491 				compress_type = "no";
492 				btrfs_clear_opt(info->mount_opt, COMPRESS);
493 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
494 				compress_force = false;
495 			} else {
496 				ret = -EINVAL;
497 				goto out;
498 			}
499 
500 			if (compress_force) {
501 				btrfs_set_and_info(root, FORCE_COMPRESS,
502 						   "force %s compression",
503 						   compress_type);
504 			} else {
505 				if (!btrfs_test_opt(root, COMPRESS))
506 					btrfs_info(root->fs_info,
507 						   "btrfs: use %s compression",
508 						   compress_type);
509 				/*
510 				 * If we remount from compress-force=xxx to
511 				 * compress=xxx, we need clear FORCE_COMPRESS
512 				 * flag, otherwise, there is no way for users
513 				 * to disable forcible compression separately.
514 				 */
515 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
516 			}
517 			break;
518 		case Opt_ssd:
519 			btrfs_set_and_info(root, SSD,
520 					   "use ssd allocation scheme");
521 			break;
522 		case Opt_ssd_spread:
523 			btrfs_set_and_info(root, SSD_SPREAD,
524 					   "use spread ssd allocation scheme");
525 			btrfs_set_opt(info->mount_opt, SSD);
526 			break;
527 		case Opt_nossd:
528 			btrfs_set_and_info(root, NOSSD,
529 					     "not using ssd allocation scheme");
530 			btrfs_clear_opt(info->mount_opt, SSD);
531 			break;
532 		case Opt_barrier:
533 			btrfs_clear_and_info(root, NOBARRIER,
534 					     "turning on barriers");
535 			break;
536 		case Opt_nobarrier:
537 			btrfs_set_and_info(root, NOBARRIER,
538 					   "turning off barriers");
539 			break;
540 		case Opt_thread_pool:
541 			ret = match_int(&args[0], &intarg);
542 			if (ret) {
543 				goto out;
544 			} else if (intarg > 0) {
545 				info->thread_pool_size = intarg;
546 			} else {
547 				ret = -EINVAL;
548 				goto out;
549 			}
550 			break;
551 		case Opt_max_inline:
552 			num = match_strdup(&args[0]);
553 			if (num) {
554 				info->max_inline = memparse(num, NULL);
555 				kfree(num);
556 
557 				if (info->max_inline) {
558 					info->max_inline = min_t(u64,
559 						info->max_inline,
560 						root->sectorsize);
561 				}
562 				btrfs_info(root->fs_info, "max_inline at %llu",
563 					info->max_inline);
564 			} else {
565 				ret = -ENOMEM;
566 				goto out;
567 			}
568 			break;
569 		case Opt_alloc_start:
570 			num = match_strdup(&args[0]);
571 			if (num) {
572 				mutex_lock(&info->chunk_mutex);
573 				info->alloc_start = memparse(num, NULL);
574 				mutex_unlock(&info->chunk_mutex);
575 				kfree(num);
576 				btrfs_info(root->fs_info, "allocations start at %llu",
577 					info->alloc_start);
578 			} else {
579 				ret = -ENOMEM;
580 				goto out;
581 			}
582 			break;
583 		case Opt_acl:
584 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
585 			root->fs_info->sb->s_flags |= MS_POSIXACL;
586 			break;
587 #else
588 			btrfs_err(root->fs_info,
589 				"support for ACL not compiled in!");
590 			ret = -EINVAL;
591 			goto out;
592 #endif
593 		case Opt_noacl:
594 			root->fs_info->sb->s_flags &= ~MS_POSIXACL;
595 			break;
596 		case Opt_notreelog:
597 			btrfs_set_and_info(root, NOTREELOG,
598 					   "disabling tree log");
599 			break;
600 		case Opt_treelog:
601 			btrfs_clear_and_info(root, NOTREELOG,
602 					     "enabling tree log");
603 			break;
604 		case Opt_flushoncommit:
605 			btrfs_set_and_info(root, FLUSHONCOMMIT,
606 					   "turning on flush-on-commit");
607 			break;
608 		case Opt_noflushoncommit:
609 			btrfs_clear_and_info(root, FLUSHONCOMMIT,
610 					     "turning off flush-on-commit");
611 			break;
612 		case Opt_ratio:
613 			ret = match_int(&args[0], &intarg);
614 			if (ret) {
615 				goto out;
616 			} else if (intarg >= 0) {
617 				info->metadata_ratio = intarg;
618 				btrfs_info(root->fs_info, "metadata ratio %d",
619 				       info->metadata_ratio);
620 			} else {
621 				ret = -EINVAL;
622 				goto out;
623 			}
624 			break;
625 		case Opt_discard:
626 			btrfs_set_and_info(root, DISCARD,
627 					   "turning on discard");
628 			break;
629 		case Opt_nodiscard:
630 			btrfs_clear_and_info(root, DISCARD,
631 					     "turning off discard");
632 			break;
633 		case Opt_space_cache:
634 			btrfs_set_and_info(root, SPACE_CACHE,
635 					   "enabling disk space caching");
636 			break;
637 		case Opt_rescan_uuid_tree:
638 			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
639 			break;
640 		case Opt_no_space_cache:
641 			btrfs_clear_and_info(root, SPACE_CACHE,
642 					     "disabling disk space caching");
643 			break;
644 		case Opt_inode_cache:
645 			btrfs_set_and_info(root, CHANGE_INODE_CACHE,
646 					   "enabling inode map caching");
647 			break;
648 		case Opt_noinode_cache:
649 			btrfs_clear_and_info(root, CHANGE_INODE_CACHE,
650 					     "disabling inode map caching");
651 			break;
652 		case Opt_clear_cache:
653 			btrfs_set_and_info(root, CLEAR_CACHE,
654 					   "force clearing of disk cache");
655 			break;
656 		case Opt_user_subvol_rm_allowed:
657 			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
658 			break;
659 		case Opt_enospc_debug:
660 			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
661 			break;
662 		case Opt_noenospc_debug:
663 			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
664 			break;
665 		case Opt_defrag:
666 			btrfs_set_and_info(root, AUTO_DEFRAG,
667 					   "enabling auto defrag");
668 			break;
669 		case Opt_nodefrag:
670 			btrfs_clear_and_info(root, AUTO_DEFRAG,
671 					     "disabling auto defrag");
672 			break;
673 		case Opt_recovery:
674 			btrfs_info(root->fs_info, "enabling auto recovery");
675 			btrfs_set_opt(info->mount_opt, RECOVERY);
676 			break;
677 		case Opt_skip_balance:
678 			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
679 			break;
680 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
681 		case Opt_check_integrity_including_extent_data:
682 			btrfs_info(root->fs_info,
683 				   "enabling check integrity including extent data");
684 			btrfs_set_opt(info->mount_opt,
685 				      CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
686 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
687 			break;
688 		case Opt_check_integrity:
689 			btrfs_info(root->fs_info, "enabling check integrity");
690 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
691 			break;
692 		case Opt_check_integrity_print_mask:
693 			ret = match_int(&args[0], &intarg);
694 			if (ret) {
695 				goto out;
696 			} else if (intarg >= 0) {
697 				info->check_integrity_print_mask = intarg;
698 				btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x",
699 				       info->check_integrity_print_mask);
700 			} else {
701 				ret = -EINVAL;
702 				goto out;
703 			}
704 			break;
705 #else
706 		case Opt_check_integrity_including_extent_data:
707 		case Opt_check_integrity:
708 		case Opt_check_integrity_print_mask:
709 			btrfs_err(root->fs_info,
710 				"support for check_integrity* not compiled in!");
711 			ret = -EINVAL;
712 			goto out;
713 #endif
714 		case Opt_fatal_errors:
715 			if (strcmp(args[0].from, "panic") == 0)
716 				btrfs_set_opt(info->mount_opt,
717 					      PANIC_ON_FATAL_ERROR);
718 			else if (strcmp(args[0].from, "bug") == 0)
719 				btrfs_clear_opt(info->mount_opt,
720 					      PANIC_ON_FATAL_ERROR);
721 			else {
722 				ret = -EINVAL;
723 				goto out;
724 			}
725 			break;
726 		case Opt_commit_interval:
727 			intarg = 0;
728 			ret = match_int(&args[0], &intarg);
729 			if (ret < 0) {
730 				btrfs_err(root->fs_info, "invalid commit interval");
731 				ret = -EINVAL;
732 				goto out;
733 			}
734 			if (intarg > 0) {
735 				if (intarg > 300) {
736 					btrfs_warn(root->fs_info, "excessive commit interval %d",
737 							intarg);
738 				}
739 				info->commit_interval = intarg;
740 			} else {
741 				btrfs_info(root->fs_info, "using default commit interval %ds",
742 				    BTRFS_DEFAULT_COMMIT_INTERVAL);
743 				info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
744 			}
745 			break;
746 		case Opt_err:
747 			btrfs_info(root->fs_info, "unrecognized mount option '%s'", p);
748 			ret = -EINVAL;
749 			goto out;
750 		default:
751 			break;
752 		}
753 	}
754 out:
755 	if (!ret && btrfs_test_opt(root, SPACE_CACHE))
756 		btrfs_info(root->fs_info, "disk space caching is enabled");
757 	kfree(orig);
758 	return ret;
759 }
760 
761 /*
762  * Parse mount options that are required early in the mount process.
763  *
764  * All other options will be parsed on much later in the mount process and
765  * only when we need to allocate a new super block.
766  */
767 static int btrfs_parse_early_options(const char *options, fmode_t flags,
768 		void *holder, char **subvol_name, u64 *subvol_objectid,
769 		struct btrfs_fs_devices **fs_devices)
770 {
771 	substring_t args[MAX_OPT_ARGS];
772 	char *device_name, *opts, *orig, *p;
773 	char *num = NULL;
774 	int error = 0;
775 
776 	if (!options)
777 		return 0;
778 
779 	/*
780 	 * strsep changes the string, duplicate it because parse_options
781 	 * gets called twice
782 	 */
783 	opts = kstrdup(options, GFP_KERNEL);
784 	if (!opts)
785 		return -ENOMEM;
786 	orig = opts;
787 
788 	while ((p = strsep(&opts, ",")) != NULL) {
789 		int token;
790 		if (!*p)
791 			continue;
792 
793 		token = match_token(p, tokens, args);
794 		switch (token) {
795 		case Opt_subvol:
796 			kfree(*subvol_name);
797 			*subvol_name = match_strdup(&args[0]);
798 			if (!*subvol_name) {
799 				error = -ENOMEM;
800 				goto out;
801 			}
802 			break;
803 		case Opt_subvolid:
804 			num = match_strdup(&args[0]);
805 			if (num) {
806 				*subvol_objectid = memparse(num, NULL);
807 				kfree(num);
808 				/* we want the original fs_tree */
809 				if (!*subvol_objectid)
810 					*subvol_objectid =
811 						BTRFS_FS_TREE_OBJECTID;
812 			} else {
813 				error = -EINVAL;
814 				goto out;
815 			}
816 			break;
817 		case Opt_subvolrootid:
818 			printk(KERN_WARNING
819 				"BTRFS: 'subvolrootid' mount option is deprecated and has "
820 				"no effect\n");
821 			break;
822 		case Opt_device:
823 			device_name = match_strdup(&args[0]);
824 			if (!device_name) {
825 				error = -ENOMEM;
826 				goto out;
827 			}
828 			error = btrfs_scan_one_device(device_name,
829 					flags, holder, fs_devices);
830 			kfree(device_name);
831 			if (error)
832 				goto out;
833 			break;
834 		default:
835 			break;
836 		}
837 	}
838 
839 out:
840 	kfree(orig);
841 	return error;
842 }
843 
844 static struct dentry *get_default_root(struct super_block *sb,
845 				       u64 subvol_objectid)
846 {
847 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
848 	struct btrfs_root *root = fs_info->tree_root;
849 	struct btrfs_root *new_root;
850 	struct btrfs_dir_item *di;
851 	struct btrfs_path *path;
852 	struct btrfs_key location;
853 	struct inode *inode;
854 	u64 dir_id;
855 	int new = 0;
856 
857 	/*
858 	 * We have a specific subvol we want to mount, just setup location and
859 	 * go look up the root.
860 	 */
861 	if (subvol_objectid) {
862 		location.objectid = subvol_objectid;
863 		location.type = BTRFS_ROOT_ITEM_KEY;
864 		location.offset = (u64)-1;
865 		goto find_root;
866 	}
867 
868 	path = btrfs_alloc_path();
869 	if (!path)
870 		return ERR_PTR(-ENOMEM);
871 	path->leave_spinning = 1;
872 
873 	/*
874 	 * Find the "default" dir item which points to the root item that we
875 	 * will mount by default if we haven't been given a specific subvolume
876 	 * to mount.
877 	 */
878 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
879 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
880 	if (IS_ERR(di)) {
881 		btrfs_free_path(path);
882 		return ERR_CAST(di);
883 	}
884 	if (!di) {
885 		/*
886 		 * Ok the default dir item isn't there.  This is weird since
887 		 * it's always been there, but don't freak out, just try and
888 		 * mount to root most subvolume.
889 		 */
890 		btrfs_free_path(path);
891 		dir_id = BTRFS_FIRST_FREE_OBJECTID;
892 		new_root = fs_info->fs_root;
893 		goto setup_root;
894 	}
895 
896 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
897 	btrfs_free_path(path);
898 
899 find_root:
900 	new_root = btrfs_read_fs_root_no_name(fs_info, &location);
901 	if (IS_ERR(new_root))
902 		return ERR_CAST(new_root);
903 
904 	dir_id = btrfs_root_dirid(&new_root->root_item);
905 setup_root:
906 	location.objectid = dir_id;
907 	location.type = BTRFS_INODE_ITEM_KEY;
908 	location.offset = 0;
909 
910 	inode = btrfs_iget(sb, &location, new_root, &new);
911 	if (IS_ERR(inode))
912 		return ERR_CAST(inode);
913 
914 	/*
915 	 * If we're just mounting the root most subvol put the inode and return
916 	 * a reference to the dentry.  We will have already gotten a reference
917 	 * to the inode in btrfs_fill_super so we're good to go.
918 	 */
919 	if (!new && sb->s_root->d_inode == inode) {
920 		iput(inode);
921 		return dget(sb->s_root);
922 	}
923 
924 	return d_obtain_root(inode);
925 }
926 
927 static int btrfs_fill_super(struct super_block *sb,
928 			    struct btrfs_fs_devices *fs_devices,
929 			    void *data, int silent)
930 {
931 	struct inode *inode;
932 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
933 	struct btrfs_key key;
934 	int err;
935 
936 	sb->s_maxbytes = MAX_LFS_FILESIZE;
937 	sb->s_magic = BTRFS_SUPER_MAGIC;
938 	sb->s_op = &btrfs_super_ops;
939 	sb->s_d_op = &btrfs_dentry_operations;
940 	sb->s_export_op = &btrfs_export_ops;
941 	sb->s_xattr = btrfs_xattr_handlers;
942 	sb->s_time_gran = 1;
943 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
944 	sb->s_flags |= MS_POSIXACL;
945 #endif
946 	sb->s_flags |= MS_I_VERSION;
947 	err = open_ctree(sb, fs_devices, (char *)data);
948 	if (err) {
949 		printk(KERN_ERR "BTRFS: open_ctree failed\n");
950 		return err;
951 	}
952 
953 	key.objectid = BTRFS_FIRST_FREE_OBJECTID;
954 	key.type = BTRFS_INODE_ITEM_KEY;
955 	key.offset = 0;
956 	inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL);
957 	if (IS_ERR(inode)) {
958 		err = PTR_ERR(inode);
959 		goto fail_close;
960 	}
961 
962 	sb->s_root = d_make_root(inode);
963 	if (!sb->s_root) {
964 		err = -ENOMEM;
965 		goto fail_close;
966 	}
967 
968 	save_mount_options(sb, data);
969 	cleancache_init_fs(sb);
970 	sb->s_flags |= MS_ACTIVE;
971 	return 0;
972 
973 fail_close:
974 	close_ctree(fs_info->tree_root);
975 	return err;
976 }
977 
978 int btrfs_sync_fs(struct super_block *sb, int wait)
979 {
980 	struct btrfs_trans_handle *trans;
981 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
982 	struct btrfs_root *root = fs_info->tree_root;
983 
984 	trace_btrfs_sync_fs(wait);
985 
986 	if (!wait) {
987 		filemap_flush(fs_info->btree_inode->i_mapping);
988 		return 0;
989 	}
990 
991 	btrfs_wait_ordered_roots(fs_info, -1);
992 
993 	trans = btrfs_attach_transaction_barrier(root);
994 	if (IS_ERR(trans)) {
995 		/* no transaction, don't bother */
996 		if (PTR_ERR(trans) == -ENOENT)
997 			return 0;
998 		return PTR_ERR(trans);
999 	}
1000 	return btrfs_commit_transaction(trans, root);
1001 }
1002 
1003 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1004 {
1005 	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1006 	struct btrfs_root *root = info->tree_root;
1007 	char *compress_type;
1008 
1009 	if (btrfs_test_opt(root, DEGRADED))
1010 		seq_puts(seq, ",degraded");
1011 	if (btrfs_test_opt(root, NODATASUM))
1012 		seq_puts(seq, ",nodatasum");
1013 	if (btrfs_test_opt(root, NODATACOW))
1014 		seq_puts(seq, ",nodatacow");
1015 	if (btrfs_test_opt(root, NOBARRIER))
1016 		seq_puts(seq, ",nobarrier");
1017 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1018 		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1019 	if (info->alloc_start != 0)
1020 		seq_printf(seq, ",alloc_start=%llu", info->alloc_start);
1021 	if (info->thread_pool_size !=  min_t(unsigned long,
1022 					     num_online_cpus() + 2, 8))
1023 		seq_printf(seq, ",thread_pool=%d", info->thread_pool_size);
1024 	if (btrfs_test_opt(root, COMPRESS)) {
1025 		if (info->compress_type == BTRFS_COMPRESS_ZLIB)
1026 			compress_type = "zlib";
1027 		else
1028 			compress_type = "lzo";
1029 		if (btrfs_test_opt(root, FORCE_COMPRESS))
1030 			seq_printf(seq, ",compress-force=%s", compress_type);
1031 		else
1032 			seq_printf(seq, ",compress=%s", compress_type);
1033 	}
1034 	if (btrfs_test_opt(root, NOSSD))
1035 		seq_puts(seq, ",nossd");
1036 	if (btrfs_test_opt(root, SSD_SPREAD))
1037 		seq_puts(seq, ",ssd_spread");
1038 	else if (btrfs_test_opt(root, SSD))
1039 		seq_puts(seq, ",ssd");
1040 	if (btrfs_test_opt(root, NOTREELOG))
1041 		seq_puts(seq, ",notreelog");
1042 	if (btrfs_test_opt(root, FLUSHONCOMMIT))
1043 		seq_puts(seq, ",flushoncommit");
1044 	if (btrfs_test_opt(root, DISCARD))
1045 		seq_puts(seq, ",discard");
1046 	if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
1047 		seq_puts(seq, ",noacl");
1048 	if (btrfs_test_opt(root, SPACE_CACHE))
1049 		seq_puts(seq, ",space_cache");
1050 	else
1051 		seq_puts(seq, ",nospace_cache");
1052 	if (btrfs_test_opt(root, RESCAN_UUID_TREE))
1053 		seq_puts(seq, ",rescan_uuid_tree");
1054 	if (btrfs_test_opt(root, CLEAR_CACHE))
1055 		seq_puts(seq, ",clear_cache");
1056 	if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1057 		seq_puts(seq, ",user_subvol_rm_allowed");
1058 	if (btrfs_test_opt(root, ENOSPC_DEBUG))
1059 		seq_puts(seq, ",enospc_debug");
1060 	if (btrfs_test_opt(root, AUTO_DEFRAG))
1061 		seq_puts(seq, ",autodefrag");
1062 	if (btrfs_test_opt(root, INODE_MAP_CACHE))
1063 		seq_puts(seq, ",inode_cache");
1064 	if (btrfs_test_opt(root, SKIP_BALANCE))
1065 		seq_puts(seq, ",skip_balance");
1066 	if (btrfs_test_opt(root, RECOVERY))
1067 		seq_puts(seq, ",recovery");
1068 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1069 	if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1070 		seq_puts(seq, ",check_int_data");
1071 	else if (btrfs_test_opt(root, CHECK_INTEGRITY))
1072 		seq_puts(seq, ",check_int");
1073 	if (info->check_integrity_print_mask)
1074 		seq_printf(seq, ",check_int_print_mask=%d",
1075 				info->check_integrity_print_mask);
1076 #endif
1077 	if (info->metadata_ratio)
1078 		seq_printf(seq, ",metadata_ratio=%d",
1079 				info->metadata_ratio);
1080 	if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR))
1081 		seq_puts(seq, ",fatal_errors=panic");
1082 	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1083 		seq_printf(seq, ",commit=%d", info->commit_interval);
1084 	return 0;
1085 }
1086 
1087 static int btrfs_test_super(struct super_block *s, void *data)
1088 {
1089 	struct btrfs_fs_info *p = data;
1090 	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1091 
1092 	return fs_info->fs_devices == p->fs_devices;
1093 }
1094 
1095 static int btrfs_set_super(struct super_block *s, void *data)
1096 {
1097 	int err = set_anon_super(s, data);
1098 	if (!err)
1099 		s->s_fs_info = data;
1100 	return err;
1101 }
1102 
1103 /*
1104  * subvolumes are identified by ino 256
1105  */
1106 static inline int is_subvolume_inode(struct inode *inode)
1107 {
1108 	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1109 		return 1;
1110 	return 0;
1111 }
1112 
1113 /*
1114  * This will strip out the subvol=%s argument for an argument string and add
1115  * subvolid=0 to make sure we get the actual tree root for path walking to the
1116  * subvol we want.
1117  */
1118 static char *setup_root_args(char *args)
1119 {
1120 	unsigned len = strlen(args) + 2 + 1;
1121 	char *src, *dst, *buf;
1122 
1123 	/*
1124 	 * We need the same args as before, but with this substitution:
1125 	 * s!subvol=[^,]+!subvolid=0!
1126 	 *
1127 	 * Since the replacement string is up to 2 bytes longer than the
1128 	 * original, allocate strlen(args) + 2 + 1 bytes.
1129 	 */
1130 
1131 	src = strstr(args, "subvol=");
1132 	/* This shouldn't happen, but just in case.. */
1133 	if (!src)
1134 		return NULL;
1135 
1136 	buf = dst = kmalloc(len, GFP_NOFS);
1137 	if (!buf)
1138 		return NULL;
1139 
1140 	/*
1141 	 * If the subvol= arg is not at the start of the string,
1142 	 * copy whatever precedes it into buf.
1143 	 */
1144 	if (src != args) {
1145 		*src++ = '\0';
1146 		strcpy(buf, args);
1147 		dst += strlen(args);
1148 	}
1149 
1150 	strcpy(dst, "subvolid=0");
1151 	dst += strlen("subvolid=0");
1152 
1153 	/*
1154 	 * If there is a "," after the original subvol=... string,
1155 	 * copy that suffix into our buffer.  Otherwise, we're done.
1156 	 */
1157 	src = strchr(src, ',');
1158 	if (src)
1159 		strcpy(dst, src);
1160 
1161 	return buf;
1162 }
1163 
1164 static struct dentry *mount_subvol(const char *subvol_name, int flags,
1165 				   const char *device_name, char *data)
1166 {
1167 	struct dentry *root;
1168 	struct vfsmount *mnt;
1169 	char *newargs;
1170 
1171 	newargs = setup_root_args(data);
1172 	if (!newargs)
1173 		return ERR_PTR(-ENOMEM);
1174 	mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
1175 			     newargs);
1176 
1177 	if (PTR_RET(mnt) == -EBUSY) {
1178 		if (flags & MS_RDONLY) {
1179 			mnt = vfs_kern_mount(&btrfs_fs_type, flags & ~MS_RDONLY, device_name,
1180 					     newargs);
1181 		} else {
1182 			int r;
1183 			mnt = vfs_kern_mount(&btrfs_fs_type, flags | MS_RDONLY, device_name,
1184 					     newargs);
1185 			if (IS_ERR(mnt)) {
1186 				kfree(newargs);
1187 				return ERR_CAST(mnt);
1188 			}
1189 
1190 			r = btrfs_remount(mnt->mnt_sb, &flags, NULL);
1191 			if (r < 0) {
1192 				/* FIXME: release vfsmount mnt ??*/
1193 				kfree(newargs);
1194 				return ERR_PTR(r);
1195 			}
1196 		}
1197 	}
1198 
1199 	kfree(newargs);
1200 
1201 	if (IS_ERR(mnt))
1202 		return ERR_CAST(mnt);
1203 
1204 	root = mount_subtree(mnt, subvol_name);
1205 
1206 	if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) {
1207 		struct super_block *s = root->d_sb;
1208 		dput(root);
1209 		root = ERR_PTR(-EINVAL);
1210 		deactivate_locked_super(s);
1211 		printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n",
1212 				subvol_name);
1213 	}
1214 
1215 	return root;
1216 }
1217 
1218 static int parse_security_options(char *orig_opts,
1219 				  struct security_mnt_opts *sec_opts)
1220 {
1221 	char *secdata = NULL;
1222 	int ret = 0;
1223 
1224 	secdata = alloc_secdata();
1225 	if (!secdata)
1226 		return -ENOMEM;
1227 	ret = security_sb_copy_data(orig_opts, secdata);
1228 	if (ret) {
1229 		free_secdata(secdata);
1230 		return ret;
1231 	}
1232 	ret = security_sb_parse_opts_str(secdata, sec_opts);
1233 	free_secdata(secdata);
1234 	return ret;
1235 }
1236 
1237 static int setup_security_options(struct btrfs_fs_info *fs_info,
1238 				  struct super_block *sb,
1239 				  struct security_mnt_opts *sec_opts)
1240 {
1241 	int ret = 0;
1242 
1243 	/*
1244 	 * Call security_sb_set_mnt_opts() to check whether new sec_opts
1245 	 * is valid.
1246 	 */
1247 	ret = security_sb_set_mnt_opts(sb, sec_opts, 0, NULL);
1248 	if (ret)
1249 		return ret;
1250 
1251 #ifdef CONFIG_SECURITY
1252 	if (!fs_info->security_opts.num_mnt_opts) {
1253 		/* first time security setup, copy sec_opts to fs_info */
1254 		memcpy(&fs_info->security_opts, sec_opts, sizeof(*sec_opts));
1255 	} else {
1256 		/*
1257 		 * Since SELinux(the only one supports security_mnt_opts) does
1258 		 * NOT support changing context during remount/mount same sb,
1259 		 * This must be the same or part of the same security options,
1260 		 * just free it.
1261 		 */
1262 		security_free_mnt_opts(sec_opts);
1263 	}
1264 #endif
1265 	return ret;
1266 }
1267 
1268 /*
1269  * Find a superblock for the given device / mount point.
1270  *
1271  * Note:  This is based on get_sb_bdev from fs/super.c with a few additions
1272  *	  for multiple device setup.  Make sure to keep it in sync.
1273  */
1274 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1275 		const char *device_name, void *data)
1276 {
1277 	struct block_device *bdev = NULL;
1278 	struct super_block *s;
1279 	struct dentry *root;
1280 	struct btrfs_fs_devices *fs_devices = NULL;
1281 	struct btrfs_fs_info *fs_info = NULL;
1282 	struct security_mnt_opts new_sec_opts;
1283 	fmode_t mode = FMODE_READ;
1284 	char *subvol_name = NULL;
1285 	u64 subvol_objectid = 0;
1286 	int error = 0;
1287 
1288 	if (!(flags & MS_RDONLY))
1289 		mode |= FMODE_WRITE;
1290 
1291 	error = btrfs_parse_early_options(data, mode, fs_type,
1292 					  &subvol_name, &subvol_objectid,
1293 					  &fs_devices);
1294 	if (error) {
1295 		kfree(subvol_name);
1296 		return ERR_PTR(error);
1297 	}
1298 
1299 	if (subvol_name) {
1300 		root = mount_subvol(subvol_name, flags, device_name, data);
1301 		kfree(subvol_name);
1302 		return root;
1303 	}
1304 
1305 	security_init_mnt_opts(&new_sec_opts);
1306 	if (data) {
1307 		error = parse_security_options(data, &new_sec_opts);
1308 		if (error)
1309 			return ERR_PTR(error);
1310 	}
1311 
1312 	error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
1313 	if (error)
1314 		goto error_sec_opts;
1315 
1316 	/*
1317 	 * Setup a dummy root and fs_info for test/set super.  This is because
1318 	 * we don't actually fill this stuff out until open_ctree, but we need
1319 	 * it for searching for existing supers, so this lets us do that and
1320 	 * then open_ctree will properly initialize everything later.
1321 	 */
1322 	fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
1323 	if (!fs_info) {
1324 		error = -ENOMEM;
1325 		goto error_sec_opts;
1326 	}
1327 
1328 	fs_info->fs_devices = fs_devices;
1329 
1330 	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1331 	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
1332 	security_init_mnt_opts(&fs_info->security_opts);
1333 	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1334 		error = -ENOMEM;
1335 		goto error_fs_info;
1336 	}
1337 
1338 	error = btrfs_open_devices(fs_devices, mode, fs_type);
1339 	if (error)
1340 		goto error_fs_info;
1341 
1342 	if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
1343 		error = -EACCES;
1344 		goto error_close_devices;
1345 	}
1346 
1347 	bdev = fs_devices->latest_bdev;
1348 	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC,
1349 		 fs_info);
1350 	if (IS_ERR(s)) {
1351 		error = PTR_ERR(s);
1352 		goto error_close_devices;
1353 	}
1354 
1355 	if (s->s_root) {
1356 		btrfs_close_devices(fs_devices);
1357 		free_fs_info(fs_info);
1358 		if ((flags ^ s->s_flags) & MS_RDONLY)
1359 			error = -EBUSY;
1360 	} else {
1361 		char b[BDEVNAME_SIZE];
1362 
1363 		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1364 		btrfs_sb(s)->bdev_holder = fs_type;
1365 		error = btrfs_fill_super(s, fs_devices, data,
1366 					 flags & MS_SILENT ? 1 : 0);
1367 	}
1368 
1369 	root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error);
1370 	if (IS_ERR(root)) {
1371 		deactivate_locked_super(s);
1372 		error = PTR_ERR(root);
1373 		goto error_sec_opts;
1374 	}
1375 
1376 	fs_info = btrfs_sb(s);
1377 	error = setup_security_options(fs_info, s, &new_sec_opts);
1378 	if (error) {
1379 		dput(root);
1380 		deactivate_locked_super(s);
1381 		goto error_sec_opts;
1382 	}
1383 
1384 	return root;
1385 
1386 error_close_devices:
1387 	btrfs_close_devices(fs_devices);
1388 error_fs_info:
1389 	free_fs_info(fs_info);
1390 error_sec_opts:
1391 	security_free_mnt_opts(&new_sec_opts);
1392 	return ERR_PTR(error);
1393 }
1394 
1395 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1396 				     int new_pool_size, int old_pool_size)
1397 {
1398 	if (new_pool_size == old_pool_size)
1399 		return;
1400 
1401 	fs_info->thread_pool_size = new_pool_size;
1402 
1403 	btrfs_info(fs_info, "resize thread pool %d -> %d",
1404 	       old_pool_size, new_pool_size);
1405 
1406 	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1407 	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1408 	btrfs_workqueue_set_max(fs_info->submit_workers, new_pool_size);
1409 	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1410 	btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1411 	btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1412 	btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1413 				new_pool_size);
1414 	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1415 	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1416 	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1417 	btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1418 	btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1419 				new_pool_size);
1420 }
1421 
1422 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info)
1423 {
1424 	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1425 }
1426 
1427 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1428 				       unsigned long old_opts, int flags)
1429 {
1430 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1431 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1432 	     (flags & MS_RDONLY))) {
1433 		/* wait for any defraggers to finish */
1434 		wait_event(fs_info->transaction_wait,
1435 			   (atomic_read(&fs_info->defrag_running) == 0));
1436 		if (flags & MS_RDONLY)
1437 			sync_filesystem(fs_info->sb);
1438 	}
1439 }
1440 
1441 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1442 					 unsigned long old_opts)
1443 {
1444 	/*
1445 	 * We need cleanup all defragable inodes if the autodefragment is
1446 	 * close or the fs is R/O.
1447 	 */
1448 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1449 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1450 	     (fs_info->sb->s_flags & MS_RDONLY))) {
1451 		btrfs_cleanup_defrag_inodes(fs_info);
1452 	}
1453 
1454 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1455 }
1456 
1457 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1458 {
1459 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1460 	struct btrfs_root *root = fs_info->tree_root;
1461 	unsigned old_flags = sb->s_flags;
1462 	unsigned long old_opts = fs_info->mount_opt;
1463 	unsigned long old_compress_type = fs_info->compress_type;
1464 	u64 old_max_inline = fs_info->max_inline;
1465 	u64 old_alloc_start = fs_info->alloc_start;
1466 	int old_thread_pool_size = fs_info->thread_pool_size;
1467 	unsigned int old_metadata_ratio = fs_info->metadata_ratio;
1468 	int ret;
1469 
1470 	sync_filesystem(sb);
1471 	btrfs_remount_prepare(fs_info);
1472 
1473 	if (data) {
1474 		struct security_mnt_opts new_sec_opts;
1475 
1476 		security_init_mnt_opts(&new_sec_opts);
1477 		ret = parse_security_options(data, &new_sec_opts);
1478 		if (ret)
1479 			goto restore;
1480 		ret = setup_security_options(fs_info, sb,
1481 					     &new_sec_opts);
1482 		if (ret) {
1483 			security_free_mnt_opts(&new_sec_opts);
1484 			goto restore;
1485 		}
1486 	}
1487 
1488 	ret = btrfs_parse_options(root, data);
1489 	if (ret) {
1490 		ret = -EINVAL;
1491 		goto restore;
1492 	}
1493 
1494 	btrfs_remount_begin(fs_info, old_opts, *flags);
1495 	btrfs_resize_thread_pool(fs_info,
1496 		fs_info->thread_pool_size, old_thread_pool_size);
1497 
1498 	if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
1499 		goto out;
1500 
1501 	if (*flags & MS_RDONLY) {
1502 		/*
1503 		 * this also happens on 'umount -rf' or on shutdown, when
1504 		 * the filesystem is busy.
1505 		 */
1506 		cancel_work_sync(&fs_info->async_reclaim_work);
1507 
1508 		/* wait for the uuid_scan task to finish */
1509 		down(&fs_info->uuid_tree_rescan_sem);
1510 		/* avoid complains from lockdep et al. */
1511 		up(&fs_info->uuid_tree_rescan_sem);
1512 
1513 		sb->s_flags |= MS_RDONLY;
1514 
1515 		btrfs_dev_replace_suspend_for_unmount(fs_info);
1516 		btrfs_scrub_cancel(fs_info);
1517 		btrfs_pause_balance(fs_info);
1518 
1519 		ret = btrfs_commit_super(root);
1520 		if (ret)
1521 			goto restore;
1522 	} else {
1523 		if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) {
1524 			btrfs_err(fs_info,
1525 				"Remounting read-write after error is not allowed");
1526 			ret = -EINVAL;
1527 			goto restore;
1528 		}
1529 		if (fs_info->fs_devices->rw_devices == 0) {
1530 			ret = -EACCES;
1531 			goto restore;
1532 		}
1533 
1534 		if (fs_info->fs_devices->missing_devices >
1535 		     fs_info->num_tolerated_disk_barrier_failures &&
1536 		    !(*flags & MS_RDONLY)) {
1537 			btrfs_warn(fs_info,
1538 				"too many missing devices, writeable remount is not allowed");
1539 			ret = -EACCES;
1540 			goto restore;
1541 		}
1542 
1543 		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1544 			ret = -EINVAL;
1545 			goto restore;
1546 		}
1547 
1548 		ret = btrfs_cleanup_fs_roots(fs_info);
1549 		if (ret)
1550 			goto restore;
1551 
1552 		/* recover relocation */
1553 		mutex_lock(&fs_info->cleaner_mutex);
1554 		ret = btrfs_recover_relocation(root);
1555 		mutex_unlock(&fs_info->cleaner_mutex);
1556 		if (ret)
1557 			goto restore;
1558 
1559 		ret = btrfs_resume_balance_async(fs_info);
1560 		if (ret)
1561 			goto restore;
1562 
1563 		ret = btrfs_resume_dev_replace_async(fs_info);
1564 		if (ret) {
1565 			btrfs_warn(fs_info, "failed to resume dev_replace");
1566 			goto restore;
1567 		}
1568 
1569 		if (!fs_info->uuid_root) {
1570 			btrfs_info(fs_info, "creating UUID tree");
1571 			ret = btrfs_create_uuid_tree(fs_info);
1572 			if (ret) {
1573 				btrfs_warn(fs_info, "failed to create the UUID tree %d", ret);
1574 				goto restore;
1575 			}
1576 		}
1577 		sb->s_flags &= ~MS_RDONLY;
1578 	}
1579 out:
1580 	wake_up_process(fs_info->transaction_kthread);
1581 	btrfs_remount_cleanup(fs_info, old_opts);
1582 	return 0;
1583 
1584 restore:
1585 	/* We've hit an error - don't reset MS_RDONLY */
1586 	if (sb->s_flags & MS_RDONLY)
1587 		old_flags |= MS_RDONLY;
1588 	sb->s_flags = old_flags;
1589 	fs_info->mount_opt = old_opts;
1590 	fs_info->compress_type = old_compress_type;
1591 	fs_info->max_inline = old_max_inline;
1592 	mutex_lock(&fs_info->chunk_mutex);
1593 	fs_info->alloc_start = old_alloc_start;
1594 	mutex_unlock(&fs_info->chunk_mutex);
1595 	btrfs_resize_thread_pool(fs_info,
1596 		old_thread_pool_size, fs_info->thread_pool_size);
1597 	fs_info->metadata_ratio = old_metadata_ratio;
1598 	btrfs_remount_cleanup(fs_info, old_opts);
1599 	return ret;
1600 }
1601 
1602 /* Used to sort the devices by max_avail(descending sort) */
1603 static int btrfs_cmp_device_free_bytes(const void *dev_info1,
1604 				       const void *dev_info2)
1605 {
1606 	if (((struct btrfs_device_info *)dev_info1)->max_avail >
1607 	    ((struct btrfs_device_info *)dev_info2)->max_avail)
1608 		return -1;
1609 	else if (((struct btrfs_device_info *)dev_info1)->max_avail <
1610 		 ((struct btrfs_device_info *)dev_info2)->max_avail)
1611 		return 1;
1612 	else
1613 	return 0;
1614 }
1615 
1616 /*
1617  * sort the devices by max_avail, in which max free extent size of each device
1618  * is stored.(Descending Sort)
1619  */
1620 static inline void btrfs_descending_sort_devices(
1621 					struct btrfs_device_info *devices,
1622 					size_t nr_devices)
1623 {
1624 	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1625 	     btrfs_cmp_device_free_bytes, NULL);
1626 }
1627 
1628 /*
1629  * The helper to calc the free space on the devices that can be used to store
1630  * file data.
1631  */
1632 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
1633 {
1634 	struct btrfs_fs_info *fs_info = root->fs_info;
1635 	struct btrfs_device_info *devices_info;
1636 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1637 	struct btrfs_device *device;
1638 	u64 skip_space;
1639 	u64 type;
1640 	u64 avail_space;
1641 	u64 used_space;
1642 	u64 min_stripe_size;
1643 	int min_stripes = 1, num_stripes = 1;
1644 	int i = 0, nr_devices;
1645 	int ret;
1646 
1647 	nr_devices = fs_info->fs_devices->open_devices;
1648 	BUG_ON(!nr_devices);
1649 
1650 	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1651 			       GFP_NOFS);
1652 	if (!devices_info)
1653 		return -ENOMEM;
1654 
1655 	/* calc min stripe number for data space alloction */
1656 	type = btrfs_get_alloc_profile(root, 1);
1657 	if (type & BTRFS_BLOCK_GROUP_RAID0) {
1658 		min_stripes = 2;
1659 		num_stripes = nr_devices;
1660 	} else if (type & BTRFS_BLOCK_GROUP_RAID1) {
1661 		min_stripes = 2;
1662 		num_stripes = 2;
1663 	} else if (type & BTRFS_BLOCK_GROUP_RAID10) {
1664 		min_stripes = 4;
1665 		num_stripes = 4;
1666 	}
1667 
1668 	if (type & BTRFS_BLOCK_GROUP_DUP)
1669 		min_stripe_size = 2 * BTRFS_STRIPE_LEN;
1670 	else
1671 		min_stripe_size = BTRFS_STRIPE_LEN;
1672 
1673 	list_for_each_entry(device, &fs_devices->devices, dev_list) {
1674 		if (!device->in_fs_metadata || !device->bdev ||
1675 		    device->is_tgtdev_for_dev_replace)
1676 			continue;
1677 
1678 		avail_space = device->total_bytes - device->bytes_used;
1679 
1680 		/* align with stripe_len */
1681 		do_div(avail_space, BTRFS_STRIPE_LEN);
1682 		avail_space *= BTRFS_STRIPE_LEN;
1683 
1684 		/*
1685 		 * In order to avoid overwritting the superblock on the drive,
1686 		 * btrfs starts at an offset of at least 1MB when doing chunk
1687 		 * allocation.
1688 		 */
1689 		skip_space = 1024 * 1024;
1690 
1691 		/* user can set the offset in fs_info->alloc_start. */
1692 		if (fs_info->alloc_start + BTRFS_STRIPE_LEN <=
1693 		    device->total_bytes)
1694 			skip_space = max(fs_info->alloc_start, skip_space);
1695 
1696 		/*
1697 		 * btrfs can not use the free space in [0, skip_space - 1],
1698 		 * we must subtract it from the total. In order to implement
1699 		 * it, we account the used space in this range first.
1700 		 */
1701 		ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1,
1702 						     &used_space);
1703 		if (ret) {
1704 			kfree(devices_info);
1705 			return ret;
1706 		}
1707 
1708 		/* calc the free space in [0, skip_space - 1] */
1709 		skip_space -= used_space;
1710 
1711 		/*
1712 		 * we can use the free space in [0, skip_space - 1], subtract
1713 		 * it from the total.
1714 		 */
1715 		if (avail_space && avail_space >= skip_space)
1716 			avail_space -= skip_space;
1717 		else
1718 			avail_space = 0;
1719 
1720 		if (avail_space < min_stripe_size)
1721 			continue;
1722 
1723 		devices_info[i].dev = device;
1724 		devices_info[i].max_avail = avail_space;
1725 
1726 		i++;
1727 	}
1728 
1729 	nr_devices = i;
1730 
1731 	btrfs_descending_sort_devices(devices_info, nr_devices);
1732 
1733 	i = nr_devices - 1;
1734 	avail_space = 0;
1735 	while (nr_devices >= min_stripes) {
1736 		if (num_stripes > nr_devices)
1737 			num_stripes = nr_devices;
1738 
1739 		if (devices_info[i].max_avail >= min_stripe_size) {
1740 			int j;
1741 			u64 alloc_size;
1742 
1743 			avail_space += devices_info[i].max_avail * num_stripes;
1744 			alloc_size = devices_info[i].max_avail;
1745 			for (j = i + 1 - num_stripes; j <= i; j++)
1746 				devices_info[j].max_avail -= alloc_size;
1747 		}
1748 		i--;
1749 		nr_devices--;
1750 	}
1751 
1752 	kfree(devices_info);
1753 	*free_bytes = avail_space;
1754 	return 0;
1755 }
1756 
1757 /*
1758  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1759  *
1760  * If there's a redundant raid level at DATA block groups, use the respective
1761  * multiplier to scale the sizes.
1762  *
1763  * Unused device space usage is based on simulating the chunk allocator
1764  * algorithm that respects the device sizes, order of allocations and the
1765  * 'alloc_start' value, this is a close approximation of the actual use but
1766  * there are other factors that may change the result (like a new metadata
1767  * chunk).
1768  *
1769  * FIXME: not accurate for mixed block groups, total and free/used are ok,
1770  * available appears slightly larger.
1771  */
1772 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1773 {
1774 	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1775 	struct btrfs_super_block *disk_super = fs_info->super_copy;
1776 	struct list_head *head = &fs_info->space_info;
1777 	struct btrfs_space_info *found;
1778 	u64 total_used = 0;
1779 	u64 total_free_data = 0;
1780 	int bits = dentry->d_sb->s_blocksize_bits;
1781 	__be32 *fsid = (__be32 *)fs_info->fsid;
1782 	unsigned factor = 1;
1783 	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1784 	int ret;
1785 
1786 	/*
1787 	 * holding chunk_muext to avoid allocating new chunks, holding
1788 	 * device_list_mutex to avoid the device being removed
1789 	 */
1790 	mutex_lock(&fs_info->fs_devices->device_list_mutex);
1791 	mutex_lock(&fs_info->chunk_mutex);
1792 	rcu_read_lock();
1793 	list_for_each_entry_rcu(found, head, list) {
1794 		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1795 			int i;
1796 
1797 			total_free_data += found->disk_total - found->disk_used;
1798 			total_free_data -=
1799 				btrfs_account_ro_block_groups_free_space(found);
1800 
1801 			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1802 				if (!list_empty(&found->block_groups[i])) {
1803 					switch (i) {
1804 					case BTRFS_RAID_DUP:
1805 					case BTRFS_RAID_RAID1:
1806 					case BTRFS_RAID_RAID10:
1807 						factor = 2;
1808 					}
1809 				}
1810 			}
1811 		}
1812 
1813 		total_used += found->disk_used;
1814 	}
1815 
1816 	rcu_read_unlock();
1817 
1818 	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1819 	buf->f_blocks >>= bits;
1820 	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1821 
1822 	/* Account global block reserve as used, it's in logical size already */
1823 	spin_lock(&block_rsv->lock);
1824 	buf->f_bfree -= block_rsv->size >> bits;
1825 	spin_unlock(&block_rsv->lock);
1826 
1827 	buf->f_bavail = total_free_data;
1828 	ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data);
1829 	if (ret) {
1830 		mutex_unlock(&fs_info->chunk_mutex);
1831 		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1832 		return ret;
1833 	}
1834 	buf->f_bavail += div_u64(total_free_data, factor);
1835 	buf->f_bavail = buf->f_bavail >> bits;
1836 	mutex_unlock(&fs_info->chunk_mutex);
1837 	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1838 
1839 	buf->f_type = BTRFS_SUPER_MAGIC;
1840 	buf->f_bsize = dentry->d_sb->s_blocksize;
1841 	buf->f_namelen = BTRFS_NAME_LEN;
1842 
1843 	/* We treat it as constant endianness (it doesn't matter _which_)
1844 	   because we want the fsid to come out the same whether mounted
1845 	   on a big-endian or little-endian host */
1846 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1847 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1848 	/* Mask in the root object ID too, to disambiguate subvols */
1849 	buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
1850 	buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
1851 
1852 	return 0;
1853 }
1854 
1855 static void btrfs_kill_super(struct super_block *sb)
1856 {
1857 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1858 	kill_anon_super(sb);
1859 	free_fs_info(fs_info);
1860 }
1861 
1862 static struct file_system_type btrfs_fs_type = {
1863 	.owner		= THIS_MODULE,
1864 	.name		= "btrfs",
1865 	.mount		= btrfs_mount,
1866 	.kill_sb	= btrfs_kill_super,
1867 	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
1868 };
1869 MODULE_ALIAS_FS("btrfs");
1870 
1871 /*
1872  * used by btrfsctl to scan devices when no FS is mounted
1873  */
1874 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
1875 				unsigned long arg)
1876 {
1877 	struct btrfs_ioctl_vol_args *vol;
1878 	struct btrfs_fs_devices *fs_devices;
1879 	int ret = -ENOTTY;
1880 
1881 	if (!capable(CAP_SYS_ADMIN))
1882 		return -EPERM;
1883 
1884 	vol = memdup_user((void __user *)arg, sizeof(*vol));
1885 	if (IS_ERR(vol))
1886 		return PTR_ERR(vol);
1887 
1888 	switch (cmd) {
1889 	case BTRFS_IOC_SCAN_DEV:
1890 		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1891 					    &btrfs_fs_type, &fs_devices);
1892 		break;
1893 	case BTRFS_IOC_DEVICES_READY:
1894 		ret = btrfs_scan_one_device(vol->name, FMODE_READ,
1895 					    &btrfs_fs_type, &fs_devices);
1896 		if (ret)
1897 			break;
1898 		ret = !(fs_devices->num_devices == fs_devices->total_devices);
1899 		break;
1900 	}
1901 
1902 	kfree(vol);
1903 	return ret;
1904 }
1905 
1906 static int btrfs_freeze(struct super_block *sb)
1907 {
1908 	struct btrfs_trans_handle *trans;
1909 	struct btrfs_root *root = btrfs_sb(sb)->tree_root;
1910 
1911 	trans = btrfs_attach_transaction_barrier(root);
1912 	if (IS_ERR(trans)) {
1913 		/* no transaction, don't bother */
1914 		if (PTR_ERR(trans) == -ENOENT)
1915 			return 0;
1916 		return PTR_ERR(trans);
1917 	}
1918 	return btrfs_commit_transaction(trans, root);
1919 }
1920 
1921 static int btrfs_unfreeze(struct super_block *sb)
1922 {
1923 	return 0;
1924 }
1925 
1926 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
1927 {
1928 	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
1929 	struct btrfs_fs_devices *cur_devices;
1930 	struct btrfs_device *dev, *first_dev = NULL;
1931 	struct list_head *head;
1932 	struct rcu_string *name;
1933 
1934 	mutex_lock(&fs_info->fs_devices->device_list_mutex);
1935 	cur_devices = fs_info->fs_devices;
1936 	while (cur_devices) {
1937 		head = &cur_devices->devices;
1938 		list_for_each_entry(dev, head, dev_list) {
1939 			if (dev->missing)
1940 				continue;
1941 			if (!dev->name)
1942 				continue;
1943 			if (!first_dev || dev->devid < first_dev->devid)
1944 				first_dev = dev;
1945 		}
1946 		cur_devices = cur_devices->seed;
1947 	}
1948 
1949 	if (first_dev) {
1950 		rcu_read_lock();
1951 		name = rcu_dereference(first_dev->name);
1952 		seq_escape(m, name->str, " \t\n\\");
1953 		rcu_read_unlock();
1954 	} else {
1955 		WARN_ON(1);
1956 	}
1957 	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
1958 	return 0;
1959 }
1960 
1961 static const struct super_operations btrfs_super_ops = {
1962 	.drop_inode	= btrfs_drop_inode,
1963 	.evict_inode	= btrfs_evict_inode,
1964 	.put_super	= btrfs_put_super,
1965 	.sync_fs	= btrfs_sync_fs,
1966 	.show_options	= btrfs_show_options,
1967 	.show_devname	= btrfs_show_devname,
1968 	.write_inode	= btrfs_write_inode,
1969 	.alloc_inode	= btrfs_alloc_inode,
1970 	.destroy_inode	= btrfs_destroy_inode,
1971 	.statfs		= btrfs_statfs,
1972 	.remount_fs	= btrfs_remount,
1973 	.freeze_fs	= btrfs_freeze,
1974 	.unfreeze_fs	= btrfs_unfreeze,
1975 };
1976 
1977 static const struct file_operations btrfs_ctl_fops = {
1978 	.unlocked_ioctl	 = btrfs_control_ioctl,
1979 	.compat_ioctl = btrfs_control_ioctl,
1980 	.owner	 = THIS_MODULE,
1981 	.llseek = noop_llseek,
1982 };
1983 
1984 static struct miscdevice btrfs_misc = {
1985 	.minor		= BTRFS_MINOR,
1986 	.name		= "btrfs-control",
1987 	.fops		= &btrfs_ctl_fops
1988 };
1989 
1990 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
1991 MODULE_ALIAS("devname:btrfs-control");
1992 
1993 static int btrfs_interface_init(void)
1994 {
1995 	return misc_register(&btrfs_misc);
1996 }
1997 
1998 static void btrfs_interface_exit(void)
1999 {
2000 	if (misc_deregister(&btrfs_misc) < 0)
2001 		printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n");
2002 }
2003 
2004 static void btrfs_print_info(void)
2005 {
2006 	printk(KERN_INFO "Btrfs loaded"
2007 #ifdef CONFIG_BTRFS_DEBUG
2008 			", debug=on"
2009 #endif
2010 #ifdef CONFIG_BTRFS_ASSERT
2011 			", assert=on"
2012 #endif
2013 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2014 			", integrity-checker=on"
2015 #endif
2016 			"\n");
2017 }
2018 
2019 static int btrfs_run_sanity_tests(void)
2020 {
2021 	int ret;
2022 
2023 	ret = btrfs_init_test_fs();
2024 	if (ret)
2025 		return ret;
2026 
2027 	ret = btrfs_test_free_space_cache();
2028 	if (ret)
2029 		goto out;
2030 	ret = btrfs_test_extent_buffer_operations();
2031 	if (ret)
2032 		goto out;
2033 	ret = btrfs_test_extent_io();
2034 	if (ret)
2035 		goto out;
2036 	ret = btrfs_test_inodes();
2037 	if (ret)
2038 		goto out;
2039 	ret = btrfs_test_qgroups();
2040 out:
2041 	btrfs_destroy_test_fs();
2042 	return ret;
2043 }
2044 
2045 static int __init init_btrfs_fs(void)
2046 {
2047 	int err;
2048 
2049 	err = btrfs_hash_init();
2050 	if (err)
2051 		return err;
2052 
2053 	btrfs_props_init();
2054 
2055 	err = btrfs_init_sysfs();
2056 	if (err)
2057 		goto free_hash;
2058 
2059 	btrfs_init_compress();
2060 
2061 	err = btrfs_init_cachep();
2062 	if (err)
2063 		goto free_compress;
2064 
2065 	err = extent_io_init();
2066 	if (err)
2067 		goto free_cachep;
2068 
2069 	err = extent_map_init();
2070 	if (err)
2071 		goto free_extent_io;
2072 
2073 	err = ordered_data_init();
2074 	if (err)
2075 		goto free_extent_map;
2076 
2077 	err = btrfs_delayed_inode_init();
2078 	if (err)
2079 		goto free_ordered_data;
2080 
2081 	err = btrfs_auto_defrag_init();
2082 	if (err)
2083 		goto free_delayed_inode;
2084 
2085 	err = btrfs_delayed_ref_init();
2086 	if (err)
2087 		goto free_auto_defrag;
2088 
2089 	err = btrfs_prelim_ref_init();
2090 	if (err)
2091 		goto free_delayed_ref;
2092 
2093 	err = btrfs_end_io_wq_init();
2094 	if (err)
2095 		goto free_prelim_ref;
2096 
2097 	err = btrfs_interface_init();
2098 	if (err)
2099 		goto free_end_io_wq;
2100 
2101 	btrfs_init_lockdep();
2102 
2103 	btrfs_print_info();
2104 
2105 	err = btrfs_run_sanity_tests();
2106 	if (err)
2107 		goto unregister_ioctl;
2108 
2109 	err = register_filesystem(&btrfs_fs_type);
2110 	if (err)
2111 		goto unregister_ioctl;
2112 
2113 	return 0;
2114 
2115 unregister_ioctl:
2116 	btrfs_interface_exit();
2117 free_end_io_wq:
2118 	btrfs_end_io_wq_exit();
2119 free_prelim_ref:
2120 	btrfs_prelim_ref_exit();
2121 free_delayed_ref:
2122 	btrfs_delayed_ref_exit();
2123 free_auto_defrag:
2124 	btrfs_auto_defrag_exit();
2125 free_delayed_inode:
2126 	btrfs_delayed_inode_exit();
2127 free_ordered_data:
2128 	ordered_data_exit();
2129 free_extent_map:
2130 	extent_map_exit();
2131 free_extent_io:
2132 	extent_io_exit();
2133 free_cachep:
2134 	btrfs_destroy_cachep();
2135 free_compress:
2136 	btrfs_exit_compress();
2137 	btrfs_exit_sysfs();
2138 free_hash:
2139 	btrfs_hash_exit();
2140 	return err;
2141 }
2142 
2143 static void __exit exit_btrfs_fs(void)
2144 {
2145 	btrfs_destroy_cachep();
2146 	btrfs_delayed_ref_exit();
2147 	btrfs_auto_defrag_exit();
2148 	btrfs_delayed_inode_exit();
2149 	btrfs_prelim_ref_exit();
2150 	ordered_data_exit();
2151 	extent_map_exit();
2152 	extent_io_exit();
2153 	btrfs_interface_exit();
2154 	btrfs_end_io_wq_exit();
2155 	unregister_filesystem(&btrfs_fs_type);
2156 	btrfs_exit_sysfs();
2157 	btrfs_cleanup_fs_uuids();
2158 	btrfs_exit_compress();
2159 	btrfs_hash_exit();
2160 }
2161 
2162 late_initcall(init_btrfs_fs);
2163 module_exit(exit_btrfs_fs)
2164 
2165 MODULE_LICENSE("GPL");
2166