• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #include <linux/blkdev.h>
7 #include <linux/module.h>
8 #include <linux/fs.h>
9 #include <linux/pagemap.h>
10 #include <linux/highmem.h>
11 #include <linux/time.h>
12 #include <linux/init.h>
13 #include <linux/seq_file.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/writeback.h>
18 #include <linux/statfs.h>
19 #include <linux/compat.h>
20 #include <linux/parser.h>
21 #include <linux/ctype.h>
22 #include <linux/namei.h>
23 #include <linux/miscdevice.h>
24 #include <linux/magic.h>
25 #include <linux/slab.h>
26 #include <linux/cleancache.h>
27 #include <linux/ratelimit.h>
28 #include <linux/crc32c.h>
29 #include <linux/btrfs.h>
30 #include "delayed-inode.h"
31 #include "ctree.h"
32 #include "disk-io.h"
33 #include "transaction.h"
34 #include "btrfs_inode.h"
35 #include "print-tree.h"
36 #include "props.h"
37 #include "xattr.h"
38 #include "volumes.h"
39 #include "export.h"
40 #include "compression.h"
41 #include "rcu-string.h"
42 #include "dev-replace.h"
43 #include "free-space-cache.h"
44 #include "backref.h"
45 #include "space-info.h"
46 #include "sysfs.h"
47 #include "tests/btrfs-tests.h"
48 #include "block-group.h"
49 #include "discard.h"
50 
51 #include "qgroup.h"
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/btrfs.h>
54 
55 static const struct super_operations btrfs_super_ops;
56 
57 /*
58  * Types for mounting the default subvolume and a subvolume explicitly
59  * requested by subvol=/path. That way the callchain is straightforward and we
60  * don't have to play tricks with the mount options and recursive calls to
61  * btrfs_mount.
62  *
63  * The new btrfs_root_fs_type also servers as a tag for the bdev_holder.
64  */
65 static struct file_system_type btrfs_fs_type;
66 static struct file_system_type btrfs_root_fs_type;
67 
68 static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69 
70 /*
71  * Generally the error codes correspond to their respective errors, but there
72  * are a few special cases.
73  *
74  * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
75  *          instance will return EUCLEAN if any of the blocks are corrupted in
76  *          a way that is problematic.  We want to reserve EUCLEAN for these
77  *          sort of corruptions.
78  *
79  * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80  *        need to use EROFS for this case.  We will have no idea of the
81  *        original failure, that will have been reported at the time we tripped
82  *        over the error.  Each subsequent error that doesn't have any context
83  *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84  */
btrfs_decode_error(int errno)85 const char * __attribute_const__ btrfs_decode_error(int errno)
86 {
87 	char *errstr = "unknown";
88 
89 	switch (errno) {
90 	case -ENOENT:		/* -2 */
91 		errstr = "No such entry";
92 		break;
93 	case -EIO:		/* -5 */
94 		errstr = "IO failure";
95 		break;
96 	case -ENOMEM:		/* -12*/
97 		errstr = "Out of memory";
98 		break;
99 	case -EEXIST:		/* -17 */
100 		errstr = "Object already exists";
101 		break;
102 	case -ENOSPC:		/* -28 */
103 		errstr = "No space left";
104 		break;
105 	case -EROFS:		/* -30 */
106 		errstr = "Readonly filesystem";
107 		break;
108 	case -EOPNOTSUPP:	/* -95 */
109 		errstr = "Operation not supported";
110 		break;
111 	case -EUCLEAN:		/* -117 */
112 		errstr = "Filesystem corrupted";
113 		break;
114 	case -EDQUOT:		/* -122 */
115 		errstr = "Quota exceeded";
116 		break;
117 	}
118 
119 	return errstr;
120 }
121 
122 /*
123  * __btrfs_handle_fs_error decodes expected errors from the caller and
124  * invokes the appropriate error response.
125  */
126 __cold
__btrfs_handle_fs_error(struct btrfs_fs_info * fs_info,const char * function,unsigned int line,int errno,const char * fmt,...)127 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128 		       unsigned int line, int errno, const char *fmt, ...)
129 {
130 	struct super_block *sb = fs_info->sb;
131 #ifdef CONFIG_PRINTK
132 	const char *errstr;
133 #endif
134 
135 	/*
136 	 * Special case: if the error is EROFS, and we're already
137 	 * under SB_RDONLY, then it is safe here.
138 	 */
139 	if (errno == -EROFS && sb_rdonly(sb))
140   		return;
141 
142 #ifdef CONFIG_PRINTK
143 	errstr = btrfs_decode_error(errno);
144 	if (fmt) {
145 		struct va_format vaf;
146 		va_list args;
147 
148 		va_start(args, fmt);
149 		vaf.fmt = fmt;
150 		vaf.va = &args;
151 
152 		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153 			sb->s_id, function, line, errno, errstr, &vaf);
154 		va_end(args);
155 	} else {
156 		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157 			sb->s_id, function, line, errno, errstr);
158 	}
159 #endif
160 
161 	/*
162 	 * Today we only save the error info to memory.  Long term we'll
163 	 * also send it down to the disk
164 	 */
165 	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166 
167 	/* Don't go through full error handling during mount */
168 	if (!(sb->s_flags & SB_BORN))
169 		return;
170 
171 	if (sb_rdonly(sb))
172 		return;
173 
174 	btrfs_discard_stop(fs_info);
175 
176 	/* btrfs handle error by forcing the filesystem readonly */
177 	sb->s_flags |= SB_RDONLY;
178 	btrfs_info(fs_info, "forced readonly");
179 	/*
180 	 * Note that a running device replace operation is not canceled here
181 	 * although there is no way to update the progress. It would add the
182 	 * risk of a deadlock, therefore the canceling is omitted. The only
183 	 * penalty is that some I/O remains active until the procedure
184 	 * completes. The next time when the filesystem is mounted writable
185 	 * again, the device replace operation continues.
186 	 */
187 }
188 
189 #ifdef CONFIG_PRINTK
190 static const char * const logtypes[] = {
191 	"emergency",
192 	"alert",
193 	"critical",
194 	"error",
195 	"warning",
196 	"notice",
197 	"info",
198 	"debug",
199 };
200 
201 
202 /*
203  * Use one ratelimit state per log level so that a flood of less important
204  * messages doesn't cause more important ones to be dropped.
205  */
206 static struct ratelimit_state printk_limits[] = {
207 	RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208 	RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209 	RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210 	RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211 	RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212 	RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213 	RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214 	RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
215 };
216 
btrfs_printk(const struct btrfs_fs_info * fs_info,const char * fmt,...)217 void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218 {
219 	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220 	struct va_format vaf;
221 	va_list args;
222 	int kern_level;
223 	const char *type = logtypes[4];
224 	struct ratelimit_state *ratelimit = &printk_limits[4];
225 
226 	va_start(args, fmt);
227 
228 	while ((kern_level = printk_get_level(fmt)) != 0) {
229 		size_t size = printk_skip_level(fmt) - fmt;
230 
231 		if (kern_level >= '0' && kern_level <= '7') {
232 			memcpy(lvl, fmt,  size);
233 			lvl[size] = '\0';
234 			type = logtypes[kern_level - '0'];
235 			ratelimit = &printk_limits[kern_level - '0'];
236 		}
237 		fmt += size;
238 	}
239 
240 	vaf.fmt = fmt;
241 	vaf.va = &args;
242 
243 	if (__ratelimit(ratelimit))
244 		printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
245 			fs_info ? fs_info->sb->s_id : "<unknown>", &vaf);
246 
247 	va_end(args);
248 }
249 #endif
250 
251 /*
252  * We only mark the transaction aborted and then set the file system read-only.
253  * This will prevent new transactions from starting or trying to join this
254  * one.
255  *
256  * This means that error recovery at the call site is limited to freeing
257  * any local memory allocations and passing the error code up without
258  * further cleanup. The transaction should complete as it normally would
259  * in the call path but will return -EIO.
260  *
261  * We'll complete the cleanup in btrfs_end_transaction and
262  * btrfs_commit_transaction.
263  */
264 __cold
__btrfs_abort_transaction(struct btrfs_trans_handle * trans,const char * function,unsigned int line,int errno)265 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
266 			       const char *function,
267 			       unsigned int line, int errno)
268 {
269 	struct btrfs_fs_info *fs_info = trans->fs_info;
270 
271 	WRITE_ONCE(trans->aborted, errno);
272 	/* Nothing used. The other threads that have joined this
273 	 * transaction may be able to continue. */
274 	if (!trans->dirty && list_empty(&trans->new_bgs)) {
275 		const char *errstr;
276 
277 		errstr = btrfs_decode_error(errno);
278 		btrfs_warn(fs_info,
279 		           "%s:%d: Aborting unused transaction(%s).",
280 		           function, line, errstr);
281 		return;
282 	}
283 	WRITE_ONCE(trans->transaction->aborted, errno);
284 	/* Wake up anybody who may be waiting on this transaction */
285 	wake_up(&fs_info->transaction_wait);
286 	wake_up(&fs_info->transaction_blocked_wait);
287 	__btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
288 }
289 /*
290  * __btrfs_panic decodes unexpected, fatal errors from the caller,
291  * issues an alert, and either panics or BUGs, depending on mount options.
292  */
293 __cold
__btrfs_panic(struct btrfs_fs_info * fs_info,const char * function,unsigned int line,int errno,const char * fmt,...)294 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
295 		   unsigned int line, int errno, const char *fmt, ...)
296 {
297 	char *s_id = "<unknown>";
298 	const char *errstr;
299 	struct va_format vaf = { .fmt = fmt };
300 	va_list args;
301 
302 	if (fs_info)
303 		s_id = fs_info->sb->s_id;
304 
305 	va_start(args, fmt);
306 	vaf.va = &args;
307 
308 	errstr = btrfs_decode_error(errno);
309 	if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
310 		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
311 			s_id, function, line, &vaf, errno, errstr);
312 
313 	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
314 		   function, line, &vaf, errno, errstr);
315 	va_end(args);
316 	/* Caller calls BUG() */
317 }
318 
btrfs_put_super(struct super_block * sb)319 static void btrfs_put_super(struct super_block *sb)
320 {
321 	close_ctree(btrfs_sb(sb));
322 }
323 
324 enum {
325 	Opt_acl, Opt_noacl,
326 	Opt_clear_cache,
327 	Opt_commit_interval,
328 	Opt_compress,
329 	Opt_compress_force,
330 	Opt_compress_force_type,
331 	Opt_compress_type,
332 	Opt_degraded,
333 	Opt_device,
334 	Opt_fatal_errors,
335 	Opt_flushoncommit, Opt_noflushoncommit,
336 	Opt_inode_cache, Opt_noinode_cache,
337 	Opt_max_inline,
338 	Opt_barrier, Opt_nobarrier,
339 	Opt_datacow, Opt_nodatacow,
340 	Opt_datasum, Opt_nodatasum,
341 	Opt_defrag, Opt_nodefrag,
342 	Opt_discard, Opt_nodiscard,
343 	Opt_discard_mode,
344 	Opt_norecovery,
345 	Opt_ratio,
346 	Opt_rescan_uuid_tree,
347 	Opt_skip_balance,
348 	Opt_space_cache, Opt_no_space_cache,
349 	Opt_space_cache_version,
350 	Opt_ssd, Opt_nossd,
351 	Opt_ssd_spread, Opt_nossd_spread,
352 	Opt_subvol,
353 	Opt_subvol_empty,
354 	Opt_subvolid,
355 	Opt_thread_pool,
356 	Opt_treelog, Opt_notreelog,
357 	Opt_user_subvol_rm_allowed,
358 
359 	/* Rescue options */
360 	Opt_rescue,
361 	Opt_usebackuproot,
362 	Opt_nologreplay,
363 
364 	/* Deprecated options */
365 	Opt_recovery,
366 
367 	/* Debugging options */
368 	Opt_check_integrity,
369 	Opt_check_integrity_including_extent_data,
370 	Opt_check_integrity_print_mask,
371 	Opt_enospc_debug, Opt_noenospc_debug,
372 #ifdef CONFIG_BTRFS_DEBUG
373 	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
374 #endif
375 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
376 	Opt_ref_verify,
377 #endif
378 	Opt_err,
379 };
380 
381 static const match_table_t tokens = {
382 	{Opt_acl, "acl"},
383 	{Opt_noacl, "noacl"},
384 	{Opt_clear_cache, "clear_cache"},
385 	{Opt_commit_interval, "commit=%u"},
386 	{Opt_compress, "compress"},
387 	{Opt_compress_type, "compress=%s"},
388 	{Opt_compress_force, "compress-force"},
389 	{Opt_compress_force_type, "compress-force=%s"},
390 	{Opt_degraded, "degraded"},
391 	{Opt_device, "device=%s"},
392 	{Opt_fatal_errors, "fatal_errors=%s"},
393 	{Opt_flushoncommit, "flushoncommit"},
394 	{Opt_noflushoncommit, "noflushoncommit"},
395 	{Opt_inode_cache, "inode_cache"},
396 	{Opt_noinode_cache, "noinode_cache"},
397 	{Opt_max_inline, "max_inline=%s"},
398 	{Opt_barrier, "barrier"},
399 	{Opt_nobarrier, "nobarrier"},
400 	{Opt_datacow, "datacow"},
401 	{Opt_nodatacow, "nodatacow"},
402 	{Opt_datasum, "datasum"},
403 	{Opt_nodatasum, "nodatasum"},
404 	{Opt_defrag, "autodefrag"},
405 	{Opt_nodefrag, "noautodefrag"},
406 	{Opt_discard, "discard"},
407 	{Opt_discard_mode, "discard=%s"},
408 	{Opt_nodiscard, "nodiscard"},
409 	{Opt_norecovery, "norecovery"},
410 	{Opt_ratio, "metadata_ratio=%u"},
411 	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
412 	{Opt_skip_balance, "skip_balance"},
413 	{Opt_space_cache, "space_cache"},
414 	{Opt_no_space_cache, "nospace_cache"},
415 	{Opt_space_cache_version, "space_cache=%s"},
416 	{Opt_ssd, "ssd"},
417 	{Opt_nossd, "nossd"},
418 	{Opt_ssd_spread, "ssd_spread"},
419 	{Opt_nossd_spread, "nossd_spread"},
420 	{Opt_subvol, "subvol=%s"},
421 	{Opt_subvol_empty, "subvol="},
422 	{Opt_subvolid, "subvolid=%s"},
423 	{Opt_thread_pool, "thread_pool=%u"},
424 	{Opt_treelog, "treelog"},
425 	{Opt_notreelog, "notreelog"},
426 	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
427 
428 	/* Rescue options */
429 	{Opt_rescue, "rescue=%s"},
430 	/* Deprecated, with alias rescue=nologreplay */
431 	{Opt_nologreplay, "nologreplay"},
432 	/* Deprecated, with alias rescue=usebackuproot */
433 	{Opt_usebackuproot, "usebackuproot"},
434 
435 	/* Deprecated options */
436 	{Opt_recovery, "recovery"},
437 
438 	/* Debugging options */
439 	{Opt_check_integrity, "check_int"},
440 	{Opt_check_integrity_including_extent_data, "check_int_data"},
441 	{Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
442 	{Opt_enospc_debug, "enospc_debug"},
443 	{Opt_noenospc_debug, "noenospc_debug"},
444 #ifdef CONFIG_BTRFS_DEBUG
445 	{Opt_fragment_data, "fragment=data"},
446 	{Opt_fragment_metadata, "fragment=metadata"},
447 	{Opt_fragment_all, "fragment=all"},
448 #endif
449 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
450 	{Opt_ref_verify, "ref_verify"},
451 #endif
452 	{Opt_err, NULL},
453 };
454 
455 static const match_table_t rescue_tokens = {
456 	{Opt_usebackuproot, "usebackuproot"},
457 	{Opt_nologreplay, "nologreplay"},
458 	{Opt_err, NULL},
459 };
460 
parse_rescue_options(struct btrfs_fs_info * info,const char * options)461 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
462 {
463 	char *opts;
464 	char *orig;
465 	char *p;
466 	substring_t args[MAX_OPT_ARGS];
467 	int ret = 0;
468 
469 	opts = kstrdup(options, GFP_KERNEL);
470 	if (!opts)
471 		return -ENOMEM;
472 	orig = opts;
473 
474 	while ((p = strsep(&opts, ":")) != NULL) {
475 		int token;
476 
477 		if (!*p)
478 			continue;
479 		token = match_token(p, rescue_tokens, args);
480 		switch (token){
481 		case Opt_usebackuproot:
482 			btrfs_info(info,
483 				   "trying to use backup root at mount time");
484 			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
485 			break;
486 		case Opt_nologreplay:
487 			btrfs_set_and_info(info, NOLOGREPLAY,
488 					   "disabling log replay at mount time");
489 			break;
490 		case Opt_err:
491 			btrfs_info(info, "unrecognized rescue option '%s'", p);
492 			ret = -EINVAL;
493 			goto out;
494 		default:
495 			break;
496 		}
497 
498 	}
499 out:
500 	kfree(orig);
501 	return ret;
502 }
503 
504 /*
505  * Regular mount options parser.  Everything that is needed only when
506  * reading in a new superblock is parsed here.
507  * XXX JDM: This needs to be cleaned up for remount.
508  */
btrfs_parse_options(struct btrfs_fs_info * info,char * options,unsigned long new_flags)509 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
510 			unsigned long new_flags)
511 {
512 	substring_t args[MAX_OPT_ARGS];
513 	char *p, *num;
514 	u64 cache_gen;
515 	int intarg;
516 	int ret = 0;
517 	char *compress_type;
518 	bool compress_force = false;
519 	enum btrfs_compression_type saved_compress_type;
520 	int saved_compress_level;
521 	bool saved_compress_force;
522 	int no_compress = 0;
523 
524 	cache_gen = btrfs_super_cache_generation(info->super_copy);
525 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
526 		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
527 	else if (cache_gen)
528 		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
529 
530 	/*
531 	 * Even the options are empty, we still need to do extra check
532 	 * against new flags
533 	 */
534 	if (!options)
535 		goto check;
536 
537 	while ((p = strsep(&options, ",")) != NULL) {
538 		int token;
539 		if (!*p)
540 			continue;
541 
542 		token = match_token(p, tokens, args);
543 		switch (token) {
544 		case Opt_degraded:
545 			btrfs_info(info, "allowing degraded mounts");
546 			btrfs_set_opt(info->mount_opt, DEGRADED);
547 			break;
548 		case Opt_subvol:
549 		case Opt_subvol_empty:
550 		case Opt_subvolid:
551 		case Opt_device:
552 			/*
553 			 * These are parsed by btrfs_parse_subvol_options or
554 			 * btrfs_parse_device_options and can be ignored here.
555 			 */
556 			break;
557 		case Opt_nodatasum:
558 			btrfs_set_and_info(info, NODATASUM,
559 					   "setting nodatasum");
560 			break;
561 		case Opt_datasum:
562 			if (btrfs_test_opt(info, NODATASUM)) {
563 				if (btrfs_test_opt(info, NODATACOW))
564 					btrfs_info(info,
565 						   "setting datasum, datacow enabled");
566 				else
567 					btrfs_info(info, "setting datasum");
568 			}
569 			btrfs_clear_opt(info->mount_opt, NODATACOW);
570 			btrfs_clear_opt(info->mount_opt, NODATASUM);
571 			break;
572 		case Opt_nodatacow:
573 			if (!btrfs_test_opt(info, NODATACOW)) {
574 				if (!btrfs_test_opt(info, COMPRESS) ||
575 				    !btrfs_test_opt(info, FORCE_COMPRESS)) {
576 					btrfs_info(info,
577 						   "setting nodatacow, compression disabled");
578 				} else {
579 					btrfs_info(info, "setting nodatacow");
580 				}
581 			}
582 			btrfs_clear_opt(info->mount_opt, COMPRESS);
583 			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
584 			btrfs_set_opt(info->mount_opt, NODATACOW);
585 			btrfs_set_opt(info->mount_opt, NODATASUM);
586 			break;
587 		case Opt_datacow:
588 			btrfs_clear_and_info(info, NODATACOW,
589 					     "setting datacow");
590 			break;
591 		case Opt_compress_force:
592 		case Opt_compress_force_type:
593 			compress_force = true;
594 			fallthrough;
595 		case Opt_compress:
596 		case Opt_compress_type:
597 			saved_compress_type = btrfs_test_opt(info,
598 							     COMPRESS) ?
599 				info->compress_type : BTRFS_COMPRESS_NONE;
600 			saved_compress_force =
601 				btrfs_test_opt(info, FORCE_COMPRESS);
602 			saved_compress_level = info->compress_level;
603 			if (token == Opt_compress ||
604 			    token == Opt_compress_force ||
605 			    strncmp(args[0].from, "zlib", 4) == 0) {
606 				compress_type = "zlib";
607 
608 				info->compress_type = BTRFS_COMPRESS_ZLIB;
609 				info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
610 				/*
611 				 * args[0] contains uninitialized data since
612 				 * for these tokens we don't expect any
613 				 * parameter.
614 				 */
615 				if (token != Opt_compress &&
616 				    token != Opt_compress_force)
617 					info->compress_level =
618 					  btrfs_compress_str2level(
619 							BTRFS_COMPRESS_ZLIB,
620 							args[0].from + 4);
621 				btrfs_set_opt(info->mount_opt, COMPRESS);
622 				btrfs_clear_opt(info->mount_opt, NODATACOW);
623 				btrfs_clear_opt(info->mount_opt, NODATASUM);
624 				no_compress = 0;
625 			} else if (strncmp(args[0].from, "lzo", 3) == 0) {
626 				compress_type = "lzo";
627 				info->compress_type = BTRFS_COMPRESS_LZO;
628 				info->compress_level = 0;
629 				btrfs_set_opt(info->mount_opt, COMPRESS);
630 				btrfs_clear_opt(info->mount_opt, NODATACOW);
631 				btrfs_clear_opt(info->mount_opt, NODATASUM);
632 				btrfs_set_fs_incompat(info, COMPRESS_LZO);
633 				no_compress = 0;
634 			} else if (strncmp(args[0].from, "zstd", 4) == 0) {
635 				compress_type = "zstd";
636 				info->compress_type = BTRFS_COMPRESS_ZSTD;
637 				info->compress_level =
638 					btrfs_compress_str2level(
639 							 BTRFS_COMPRESS_ZSTD,
640 							 args[0].from + 4);
641 				btrfs_set_opt(info->mount_opt, COMPRESS);
642 				btrfs_clear_opt(info->mount_opt, NODATACOW);
643 				btrfs_clear_opt(info->mount_opt, NODATASUM);
644 				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
645 				no_compress = 0;
646 			} else if (strncmp(args[0].from, "no", 2) == 0) {
647 				compress_type = "no";
648 				info->compress_level = 0;
649 				info->compress_type = 0;
650 				btrfs_clear_opt(info->mount_opt, COMPRESS);
651 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
652 				compress_force = false;
653 				no_compress++;
654 			} else {
655 				btrfs_err(info, "unrecognized compression value %s",
656 					  args[0].from);
657 				ret = -EINVAL;
658 				goto out;
659 			}
660 
661 			if (compress_force) {
662 				btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
663 			} else {
664 				/*
665 				 * If we remount from compress-force=xxx to
666 				 * compress=xxx, we need clear FORCE_COMPRESS
667 				 * flag, otherwise, there is no way for users
668 				 * to disable forcible compression separately.
669 				 */
670 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
671 			}
672 			if (no_compress == 1) {
673 				btrfs_info(info, "use no compression");
674 			} else if ((info->compress_type != saved_compress_type) ||
675 				   (compress_force != saved_compress_force) ||
676 				   (info->compress_level != saved_compress_level)) {
677 				btrfs_info(info, "%s %s compression, level %d",
678 					   (compress_force) ? "force" : "use",
679 					   compress_type, info->compress_level);
680 			}
681 			compress_force = false;
682 			break;
683 		case Opt_ssd:
684 			btrfs_set_and_info(info, SSD,
685 					   "enabling ssd optimizations");
686 			btrfs_clear_opt(info->mount_opt, NOSSD);
687 			break;
688 		case Opt_ssd_spread:
689 			btrfs_set_and_info(info, SSD,
690 					   "enabling ssd optimizations");
691 			btrfs_set_and_info(info, SSD_SPREAD,
692 					   "using spread ssd allocation scheme");
693 			btrfs_clear_opt(info->mount_opt, NOSSD);
694 			break;
695 		case Opt_nossd:
696 			btrfs_set_opt(info->mount_opt, NOSSD);
697 			btrfs_clear_and_info(info, SSD,
698 					     "not using ssd optimizations");
699 			fallthrough;
700 		case Opt_nossd_spread:
701 			btrfs_clear_and_info(info, SSD_SPREAD,
702 					     "not using spread ssd allocation scheme");
703 			break;
704 		case Opt_barrier:
705 			btrfs_clear_and_info(info, NOBARRIER,
706 					     "turning on barriers");
707 			break;
708 		case Opt_nobarrier:
709 			btrfs_set_and_info(info, NOBARRIER,
710 					   "turning off barriers");
711 			break;
712 		case Opt_thread_pool:
713 			ret = match_int(&args[0], &intarg);
714 			if (ret) {
715 				btrfs_err(info, "unrecognized thread_pool value %s",
716 					  args[0].from);
717 				goto out;
718 			} else if (intarg == 0) {
719 				btrfs_err(info, "invalid value 0 for thread_pool");
720 				ret = -EINVAL;
721 				goto out;
722 			}
723 			info->thread_pool_size = intarg;
724 			break;
725 		case Opt_max_inline:
726 			num = match_strdup(&args[0]);
727 			if (num) {
728 				info->max_inline = memparse(num, NULL);
729 				kfree(num);
730 
731 				if (info->max_inline) {
732 					info->max_inline = min_t(u64,
733 						info->max_inline,
734 						info->sectorsize);
735 				}
736 				btrfs_info(info, "max_inline at %llu",
737 					   info->max_inline);
738 			} else {
739 				ret = -ENOMEM;
740 				goto out;
741 			}
742 			break;
743 		case Opt_acl:
744 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
745 			info->sb->s_flags |= SB_POSIXACL;
746 			break;
747 #else
748 			btrfs_err(info, "support for ACL not compiled in!");
749 			ret = -EINVAL;
750 			goto out;
751 #endif
752 		case Opt_noacl:
753 			info->sb->s_flags &= ~SB_POSIXACL;
754 			break;
755 		case Opt_notreelog:
756 			btrfs_set_and_info(info, NOTREELOG,
757 					   "disabling tree log");
758 			break;
759 		case Opt_treelog:
760 			btrfs_clear_and_info(info, NOTREELOG,
761 					     "enabling tree log");
762 			break;
763 		case Opt_norecovery:
764 		case Opt_nologreplay:
765 			btrfs_warn(info,
766 		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
767 			btrfs_set_and_info(info, NOLOGREPLAY,
768 					   "disabling log replay at mount time");
769 			break;
770 		case Opt_flushoncommit:
771 			btrfs_set_and_info(info, FLUSHONCOMMIT,
772 					   "turning on flush-on-commit");
773 			break;
774 		case Opt_noflushoncommit:
775 			btrfs_clear_and_info(info, FLUSHONCOMMIT,
776 					     "turning off flush-on-commit");
777 			break;
778 		case Opt_ratio:
779 			ret = match_int(&args[0], &intarg);
780 			if (ret) {
781 				btrfs_err(info, "unrecognized metadata_ratio value %s",
782 					  args[0].from);
783 				goto out;
784 			}
785 			info->metadata_ratio = intarg;
786 			btrfs_info(info, "metadata ratio %u",
787 				   info->metadata_ratio);
788 			break;
789 		case Opt_discard:
790 		case Opt_discard_mode:
791 			if (token == Opt_discard ||
792 			    strcmp(args[0].from, "sync") == 0) {
793 				btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
794 				btrfs_set_and_info(info, DISCARD_SYNC,
795 						   "turning on sync discard");
796 			} else if (strcmp(args[0].from, "async") == 0) {
797 				btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
798 				btrfs_set_and_info(info, DISCARD_ASYNC,
799 						   "turning on async discard");
800 			} else {
801 				btrfs_err(info, "unrecognized discard mode value %s",
802 					  args[0].from);
803 				ret = -EINVAL;
804 				goto out;
805 			}
806 			break;
807 		case Opt_nodiscard:
808 			btrfs_clear_and_info(info, DISCARD_SYNC,
809 					     "turning off discard");
810 			btrfs_clear_and_info(info, DISCARD_ASYNC,
811 					     "turning off async discard");
812 			break;
813 		case Opt_space_cache:
814 		case Opt_space_cache_version:
815 			if (token == Opt_space_cache ||
816 			    strcmp(args[0].from, "v1") == 0) {
817 				btrfs_clear_opt(info->mount_opt,
818 						FREE_SPACE_TREE);
819 				btrfs_set_and_info(info, SPACE_CACHE,
820 					   "enabling disk space caching");
821 			} else if (strcmp(args[0].from, "v2") == 0) {
822 				btrfs_clear_opt(info->mount_opt,
823 						SPACE_CACHE);
824 				btrfs_set_and_info(info, FREE_SPACE_TREE,
825 						   "enabling free space tree");
826 			} else {
827 				btrfs_err(info, "unrecognized space_cache value %s",
828 					  args[0].from);
829 				ret = -EINVAL;
830 				goto out;
831 			}
832 			break;
833 		case Opt_rescan_uuid_tree:
834 			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
835 			break;
836 		case Opt_no_space_cache:
837 			if (btrfs_test_opt(info, SPACE_CACHE)) {
838 				btrfs_clear_and_info(info, SPACE_CACHE,
839 					     "disabling disk space caching");
840 			}
841 			if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
842 				btrfs_clear_and_info(info, FREE_SPACE_TREE,
843 					     "disabling free space tree");
844 			}
845 			break;
846 		case Opt_inode_cache:
847 			btrfs_warn(info,
848 	"the 'inode_cache' option is deprecated and will have no effect from 5.11");
849 			btrfs_set_pending_and_info(info, INODE_MAP_CACHE,
850 					   "enabling inode map caching");
851 			break;
852 		case Opt_noinode_cache:
853 			btrfs_clear_pending_and_info(info, INODE_MAP_CACHE,
854 					     "disabling inode map caching");
855 			break;
856 		case Opt_clear_cache:
857 			btrfs_set_and_info(info, CLEAR_CACHE,
858 					   "force clearing of disk cache");
859 			break;
860 		case Opt_user_subvol_rm_allowed:
861 			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
862 			break;
863 		case Opt_enospc_debug:
864 			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
865 			break;
866 		case Opt_noenospc_debug:
867 			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
868 			break;
869 		case Opt_defrag:
870 			btrfs_set_and_info(info, AUTO_DEFRAG,
871 					   "enabling auto defrag");
872 			break;
873 		case Opt_nodefrag:
874 			btrfs_clear_and_info(info, AUTO_DEFRAG,
875 					     "disabling auto defrag");
876 			break;
877 		case Opt_recovery:
878 		case Opt_usebackuproot:
879 			btrfs_warn(info,
880 			"'%s' is deprecated, use 'rescue=usebackuproot' instead",
881 				   token == Opt_recovery ? "recovery" :
882 				   "usebackuproot");
883 			btrfs_info(info,
884 				   "trying to use backup root at mount time");
885 			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
886 			break;
887 		case Opt_skip_balance:
888 			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
889 			break;
890 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
891 		case Opt_check_integrity_including_extent_data:
892 			btrfs_info(info,
893 				   "enabling check integrity including extent data");
894 			btrfs_set_opt(info->mount_opt,
895 				      CHECK_INTEGRITY_INCLUDING_EXTENT_DATA);
896 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
897 			break;
898 		case Opt_check_integrity:
899 			btrfs_info(info, "enabling check integrity");
900 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
901 			break;
902 		case Opt_check_integrity_print_mask:
903 			ret = match_int(&args[0], &intarg);
904 			if (ret) {
905 				btrfs_err(info,
906 				"unrecognized check_integrity_print_mask value %s",
907 					args[0].from);
908 				goto out;
909 			}
910 			info->check_integrity_print_mask = intarg;
911 			btrfs_info(info, "check_integrity_print_mask 0x%x",
912 				   info->check_integrity_print_mask);
913 			break;
914 #else
915 		case Opt_check_integrity_including_extent_data:
916 		case Opt_check_integrity:
917 		case Opt_check_integrity_print_mask:
918 			btrfs_err(info,
919 				  "support for check_integrity* not compiled in!");
920 			ret = -EINVAL;
921 			goto out;
922 #endif
923 		case Opt_fatal_errors:
924 			if (strcmp(args[0].from, "panic") == 0) {
925 				btrfs_set_opt(info->mount_opt,
926 					      PANIC_ON_FATAL_ERROR);
927 			} else if (strcmp(args[0].from, "bug") == 0) {
928 				btrfs_clear_opt(info->mount_opt,
929 					      PANIC_ON_FATAL_ERROR);
930 			} else {
931 				btrfs_err(info, "unrecognized fatal_errors value %s",
932 					  args[0].from);
933 				ret = -EINVAL;
934 				goto out;
935 			}
936 			break;
937 		case Opt_commit_interval:
938 			intarg = 0;
939 			ret = match_int(&args[0], &intarg);
940 			if (ret) {
941 				btrfs_err(info, "unrecognized commit_interval value %s",
942 					  args[0].from);
943 				ret = -EINVAL;
944 				goto out;
945 			}
946 			if (intarg == 0) {
947 				btrfs_info(info,
948 					   "using default commit interval %us",
949 					   BTRFS_DEFAULT_COMMIT_INTERVAL);
950 				intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
951 			} else if (intarg > 300) {
952 				btrfs_warn(info, "excessive commit interval %d",
953 					   intarg);
954 			}
955 			info->commit_interval = intarg;
956 			break;
957 		case Opt_rescue:
958 			ret = parse_rescue_options(info, args[0].from);
959 			if (ret < 0) {
960 				btrfs_err(info, "unrecognized rescue value %s",
961 					  args[0].from);
962 				goto out;
963 			}
964 			break;
965 #ifdef CONFIG_BTRFS_DEBUG
966 		case Opt_fragment_all:
967 			btrfs_info(info, "fragmenting all space");
968 			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
969 			btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
970 			break;
971 		case Opt_fragment_metadata:
972 			btrfs_info(info, "fragmenting metadata");
973 			btrfs_set_opt(info->mount_opt,
974 				      FRAGMENT_METADATA);
975 			break;
976 		case Opt_fragment_data:
977 			btrfs_info(info, "fragmenting data");
978 			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
979 			break;
980 #endif
981 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
982 		case Opt_ref_verify:
983 			btrfs_info(info, "doing ref verification");
984 			btrfs_set_opt(info->mount_opt, REF_VERIFY);
985 			break;
986 #endif
987 		case Opt_err:
988 			btrfs_err(info, "unrecognized mount option '%s'", p);
989 			ret = -EINVAL;
990 			goto out;
991 		default:
992 			break;
993 		}
994 	}
995 check:
996 	/*
997 	 * Extra check for current option against current flag
998 	 */
999 	if (btrfs_test_opt(info, NOLOGREPLAY) && !(new_flags & SB_RDONLY)) {
1000 		btrfs_err(info,
1001 			  "nologreplay must be used with ro mount option");
1002 		ret = -EINVAL;
1003 	}
1004 out:
1005 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1006 	    !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1007 	    !btrfs_test_opt(info, CLEAR_CACHE)) {
1008 		btrfs_err(info, "cannot disable free space tree");
1009 		ret = -EINVAL;
1010 
1011 	}
1012 	if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1013 		btrfs_info(info, "disk space caching is enabled");
1014 	if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1015 		btrfs_info(info, "using free space tree");
1016 	return ret;
1017 }
1018 
1019 /*
1020  * Parse mount options that are required early in the mount process.
1021  *
1022  * All other options will be parsed on much later in the mount process and
1023  * only when we need to allocate a new super block.
1024  */
btrfs_parse_device_options(const char * options,fmode_t flags,void * holder)1025 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1026 				      void *holder)
1027 {
1028 	substring_t args[MAX_OPT_ARGS];
1029 	char *device_name, *opts, *orig, *p;
1030 	struct btrfs_device *device = NULL;
1031 	int error = 0;
1032 
1033 	lockdep_assert_held(&uuid_mutex);
1034 
1035 	if (!options)
1036 		return 0;
1037 
1038 	/*
1039 	 * strsep changes the string, duplicate it because btrfs_parse_options
1040 	 * gets called later
1041 	 */
1042 	opts = kstrdup(options, GFP_KERNEL);
1043 	if (!opts)
1044 		return -ENOMEM;
1045 	orig = opts;
1046 
1047 	while ((p = strsep(&opts, ",")) != NULL) {
1048 		int token;
1049 
1050 		if (!*p)
1051 			continue;
1052 
1053 		token = match_token(p, tokens, args);
1054 		if (token == Opt_device) {
1055 			device_name = match_strdup(&args[0]);
1056 			if (!device_name) {
1057 				error = -ENOMEM;
1058 				goto out;
1059 			}
1060 			device = btrfs_scan_one_device(device_name, flags,
1061 					holder);
1062 			kfree(device_name);
1063 			if (IS_ERR(device)) {
1064 				error = PTR_ERR(device);
1065 				goto out;
1066 			}
1067 		}
1068 	}
1069 
1070 out:
1071 	kfree(orig);
1072 	return error;
1073 }
1074 
1075 /*
1076  * Parse mount options that are related to subvolume id
1077  *
1078  * The value is later passed to mount_subvol()
1079  */
btrfs_parse_subvol_options(const char * options,char ** subvol_name,u64 * subvol_objectid)1080 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1081 		u64 *subvol_objectid)
1082 {
1083 	substring_t args[MAX_OPT_ARGS];
1084 	char *opts, *orig, *p;
1085 	int error = 0;
1086 	u64 subvolid;
1087 
1088 	if (!options)
1089 		return 0;
1090 
1091 	/*
1092 	 * strsep changes the string, duplicate it because
1093 	 * btrfs_parse_device_options gets called later
1094 	 */
1095 	opts = kstrdup(options, GFP_KERNEL);
1096 	if (!opts)
1097 		return -ENOMEM;
1098 	orig = opts;
1099 
1100 	while ((p = strsep(&opts, ",")) != NULL) {
1101 		int token;
1102 		if (!*p)
1103 			continue;
1104 
1105 		token = match_token(p, tokens, args);
1106 		switch (token) {
1107 		case Opt_subvol:
1108 			kfree(*subvol_name);
1109 			*subvol_name = match_strdup(&args[0]);
1110 			if (!*subvol_name) {
1111 				error = -ENOMEM;
1112 				goto out;
1113 			}
1114 			break;
1115 		case Opt_subvolid:
1116 			error = match_u64(&args[0], &subvolid);
1117 			if (error)
1118 				goto out;
1119 
1120 			/* we want the original fs_tree */
1121 			if (subvolid == 0)
1122 				subvolid = BTRFS_FS_TREE_OBJECTID;
1123 
1124 			*subvol_objectid = subvolid;
1125 			break;
1126 		default:
1127 			break;
1128 		}
1129 	}
1130 
1131 out:
1132 	kfree(orig);
1133 	return error;
1134 }
1135 
btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info * fs_info,u64 subvol_objectid)1136 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1137 					  u64 subvol_objectid)
1138 {
1139 	struct btrfs_root *root = fs_info->tree_root;
1140 	struct btrfs_root *fs_root = NULL;
1141 	struct btrfs_root_ref *root_ref;
1142 	struct btrfs_inode_ref *inode_ref;
1143 	struct btrfs_key key;
1144 	struct btrfs_path *path = NULL;
1145 	char *name = NULL, *ptr;
1146 	u64 dirid;
1147 	int len;
1148 	int ret;
1149 
1150 	path = btrfs_alloc_path();
1151 	if (!path) {
1152 		ret = -ENOMEM;
1153 		goto err;
1154 	}
1155 	path->leave_spinning = 1;
1156 
1157 	name = kmalloc(PATH_MAX, GFP_KERNEL);
1158 	if (!name) {
1159 		ret = -ENOMEM;
1160 		goto err;
1161 	}
1162 	ptr = name + PATH_MAX - 1;
1163 	ptr[0] = '\0';
1164 
1165 	/*
1166 	 * Walk up the subvolume trees in the tree of tree roots by root
1167 	 * backrefs until we hit the top-level subvolume.
1168 	 */
1169 	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1170 		key.objectid = subvol_objectid;
1171 		key.type = BTRFS_ROOT_BACKREF_KEY;
1172 		key.offset = (u64)-1;
1173 
1174 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1175 		if (ret < 0) {
1176 			goto err;
1177 		} else if (ret > 0) {
1178 			ret = btrfs_previous_item(root, path, subvol_objectid,
1179 						  BTRFS_ROOT_BACKREF_KEY);
1180 			if (ret < 0) {
1181 				goto err;
1182 			} else if (ret > 0) {
1183 				ret = -ENOENT;
1184 				goto err;
1185 			}
1186 		}
1187 
1188 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1189 		subvol_objectid = key.offset;
1190 
1191 		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1192 					  struct btrfs_root_ref);
1193 		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1194 		ptr -= len + 1;
1195 		if (ptr < name) {
1196 			ret = -ENAMETOOLONG;
1197 			goto err;
1198 		}
1199 		read_extent_buffer(path->nodes[0], ptr + 1,
1200 				   (unsigned long)(root_ref + 1), len);
1201 		ptr[0] = '/';
1202 		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1203 		btrfs_release_path(path);
1204 
1205 		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1206 		if (IS_ERR(fs_root)) {
1207 			ret = PTR_ERR(fs_root);
1208 			fs_root = NULL;
1209 			goto err;
1210 		}
1211 
1212 		/*
1213 		 * Walk up the filesystem tree by inode refs until we hit the
1214 		 * root directory.
1215 		 */
1216 		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1217 			key.objectid = dirid;
1218 			key.type = BTRFS_INODE_REF_KEY;
1219 			key.offset = (u64)-1;
1220 
1221 			ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1222 			if (ret < 0) {
1223 				goto err;
1224 			} else if (ret > 0) {
1225 				ret = btrfs_previous_item(fs_root, path, dirid,
1226 							  BTRFS_INODE_REF_KEY);
1227 				if (ret < 0) {
1228 					goto err;
1229 				} else if (ret > 0) {
1230 					ret = -ENOENT;
1231 					goto err;
1232 				}
1233 			}
1234 
1235 			btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1236 			dirid = key.offset;
1237 
1238 			inode_ref = btrfs_item_ptr(path->nodes[0],
1239 						   path->slots[0],
1240 						   struct btrfs_inode_ref);
1241 			len = btrfs_inode_ref_name_len(path->nodes[0],
1242 						       inode_ref);
1243 			ptr -= len + 1;
1244 			if (ptr < name) {
1245 				ret = -ENAMETOOLONG;
1246 				goto err;
1247 			}
1248 			read_extent_buffer(path->nodes[0], ptr + 1,
1249 					   (unsigned long)(inode_ref + 1), len);
1250 			ptr[0] = '/';
1251 			btrfs_release_path(path);
1252 		}
1253 		btrfs_put_root(fs_root);
1254 		fs_root = NULL;
1255 	}
1256 
1257 	btrfs_free_path(path);
1258 	if (ptr == name + PATH_MAX - 1) {
1259 		name[0] = '/';
1260 		name[1] = '\0';
1261 	} else {
1262 		memmove(name, ptr, name + PATH_MAX - ptr);
1263 	}
1264 	return name;
1265 
1266 err:
1267 	btrfs_put_root(fs_root);
1268 	btrfs_free_path(path);
1269 	kfree(name);
1270 	return ERR_PTR(ret);
1271 }
1272 
get_default_subvol_objectid(struct btrfs_fs_info * fs_info,u64 * objectid)1273 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1274 {
1275 	struct btrfs_root *root = fs_info->tree_root;
1276 	struct btrfs_dir_item *di;
1277 	struct btrfs_path *path;
1278 	struct btrfs_key location;
1279 	u64 dir_id;
1280 
1281 	path = btrfs_alloc_path();
1282 	if (!path)
1283 		return -ENOMEM;
1284 	path->leave_spinning = 1;
1285 
1286 	/*
1287 	 * Find the "default" dir item which points to the root item that we
1288 	 * will mount by default if we haven't been given a specific subvolume
1289 	 * to mount.
1290 	 */
1291 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
1292 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1293 	if (IS_ERR(di)) {
1294 		btrfs_free_path(path);
1295 		return PTR_ERR(di);
1296 	}
1297 	if (!di) {
1298 		/*
1299 		 * Ok the default dir item isn't there.  This is weird since
1300 		 * it's always been there, but don't freak out, just try and
1301 		 * mount the top-level subvolume.
1302 		 */
1303 		btrfs_free_path(path);
1304 		*objectid = BTRFS_FS_TREE_OBJECTID;
1305 		return 0;
1306 	}
1307 
1308 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1309 	btrfs_free_path(path);
1310 	*objectid = location.objectid;
1311 	return 0;
1312 }
1313 
btrfs_fill_super(struct super_block * sb,struct btrfs_fs_devices * fs_devices,void * data)1314 static int btrfs_fill_super(struct super_block *sb,
1315 			    struct btrfs_fs_devices *fs_devices,
1316 			    void *data)
1317 {
1318 	struct inode *inode;
1319 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1320 	int err;
1321 
1322 	sb->s_maxbytes = MAX_LFS_FILESIZE;
1323 	sb->s_magic = BTRFS_SUPER_MAGIC;
1324 	sb->s_op = &btrfs_super_ops;
1325 	sb->s_d_op = &btrfs_dentry_operations;
1326 	sb->s_export_op = &btrfs_export_ops;
1327 	sb->s_xattr = btrfs_xattr_handlers;
1328 	sb->s_time_gran = 1;
1329 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1330 	sb->s_flags |= SB_POSIXACL;
1331 #endif
1332 	sb->s_flags |= SB_I_VERSION;
1333 	sb->s_iflags |= SB_I_CGROUPWB;
1334 
1335 	err = super_setup_bdi(sb);
1336 	if (err) {
1337 		btrfs_err(fs_info, "super_setup_bdi failed");
1338 		return err;
1339 	}
1340 
1341 	err = open_ctree(sb, fs_devices, (char *)data);
1342 	if (err) {
1343 		btrfs_err(fs_info, "open_ctree failed");
1344 		return err;
1345 	}
1346 
1347 	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1348 	if (IS_ERR(inode)) {
1349 		err = PTR_ERR(inode);
1350 		goto fail_close;
1351 	}
1352 
1353 	sb->s_root = d_make_root(inode);
1354 	if (!sb->s_root) {
1355 		err = -ENOMEM;
1356 		goto fail_close;
1357 	}
1358 
1359 	cleancache_init_fs(sb);
1360 	sb->s_flags |= SB_ACTIVE;
1361 	return 0;
1362 
1363 fail_close:
1364 	close_ctree(fs_info);
1365 	return err;
1366 }
1367 
btrfs_sync_fs(struct super_block * sb,int wait)1368 int btrfs_sync_fs(struct super_block *sb, int wait)
1369 {
1370 	struct btrfs_trans_handle *trans;
1371 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1372 	struct btrfs_root *root = fs_info->tree_root;
1373 
1374 	trace_btrfs_sync_fs(fs_info, wait);
1375 
1376 	if (!wait) {
1377 		filemap_flush(fs_info->btree_inode->i_mapping);
1378 		return 0;
1379 	}
1380 
1381 	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1382 
1383 	trans = btrfs_attach_transaction_barrier(root);
1384 	if (IS_ERR(trans)) {
1385 		/* no transaction, don't bother */
1386 		if (PTR_ERR(trans) == -ENOENT) {
1387 			/*
1388 			 * Exit unless we have some pending changes
1389 			 * that need to go through commit
1390 			 */
1391 			if (fs_info->pending_changes == 0)
1392 				return 0;
1393 			/*
1394 			 * A non-blocking test if the fs is frozen. We must not
1395 			 * start a new transaction here otherwise a deadlock
1396 			 * happens. The pending operations are delayed to the
1397 			 * next commit after thawing.
1398 			 */
1399 			if (sb_start_write_trylock(sb))
1400 				sb_end_write(sb);
1401 			else
1402 				return 0;
1403 			trans = btrfs_start_transaction(root, 0);
1404 		}
1405 		if (IS_ERR(trans))
1406 			return PTR_ERR(trans);
1407 	}
1408 	return btrfs_commit_transaction(trans);
1409 }
1410 
btrfs_show_options(struct seq_file * seq,struct dentry * dentry)1411 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1412 {
1413 	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1414 	const char *compress_type;
1415 	const char *subvol_name;
1416 
1417 	if (btrfs_test_opt(info, DEGRADED))
1418 		seq_puts(seq, ",degraded");
1419 	if (btrfs_test_opt(info, NODATASUM))
1420 		seq_puts(seq, ",nodatasum");
1421 	if (btrfs_test_opt(info, NODATACOW))
1422 		seq_puts(seq, ",nodatacow");
1423 	if (btrfs_test_opt(info, NOBARRIER))
1424 		seq_puts(seq, ",nobarrier");
1425 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1426 		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1427 	if (info->thread_pool_size !=  min_t(unsigned long,
1428 					     num_online_cpus() + 2, 8))
1429 		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1430 	if (btrfs_test_opt(info, COMPRESS)) {
1431 		compress_type = btrfs_compress_type2str(info->compress_type);
1432 		if (btrfs_test_opt(info, FORCE_COMPRESS))
1433 			seq_printf(seq, ",compress-force=%s", compress_type);
1434 		else
1435 			seq_printf(seq, ",compress=%s", compress_type);
1436 		if (info->compress_level)
1437 			seq_printf(seq, ":%d", info->compress_level);
1438 	}
1439 	if (btrfs_test_opt(info, NOSSD))
1440 		seq_puts(seq, ",nossd");
1441 	if (btrfs_test_opt(info, SSD_SPREAD))
1442 		seq_puts(seq, ",ssd_spread");
1443 	else if (btrfs_test_opt(info, SSD))
1444 		seq_puts(seq, ",ssd");
1445 	if (btrfs_test_opt(info, NOTREELOG))
1446 		seq_puts(seq, ",notreelog");
1447 	if (btrfs_test_opt(info, NOLOGREPLAY))
1448 		seq_puts(seq, ",rescue=nologreplay");
1449 	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1450 		seq_puts(seq, ",flushoncommit");
1451 	if (btrfs_test_opt(info, DISCARD_SYNC))
1452 		seq_puts(seq, ",discard");
1453 	if (btrfs_test_opt(info, DISCARD_ASYNC))
1454 		seq_puts(seq, ",discard=async");
1455 	if (!(info->sb->s_flags & SB_POSIXACL))
1456 		seq_puts(seq, ",noacl");
1457 	if (btrfs_test_opt(info, SPACE_CACHE))
1458 		seq_puts(seq, ",space_cache");
1459 	else if (btrfs_test_opt(info, FREE_SPACE_TREE))
1460 		seq_puts(seq, ",space_cache=v2");
1461 	else
1462 		seq_puts(seq, ",nospace_cache");
1463 	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1464 		seq_puts(seq, ",rescan_uuid_tree");
1465 	if (btrfs_test_opt(info, CLEAR_CACHE))
1466 		seq_puts(seq, ",clear_cache");
1467 	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1468 		seq_puts(seq, ",user_subvol_rm_allowed");
1469 	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1470 		seq_puts(seq, ",enospc_debug");
1471 	if (btrfs_test_opt(info, AUTO_DEFRAG))
1472 		seq_puts(seq, ",autodefrag");
1473 	if (btrfs_test_opt(info, INODE_MAP_CACHE))
1474 		seq_puts(seq, ",inode_cache");
1475 	if (btrfs_test_opt(info, SKIP_BALANCE))
1476 		seq_puts(seq, ",skip_balance");
1477 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1478 	if (btrfs_test_opt(info, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA))
1479 		seq_puts(seq, ",check_int_data");
1480 	else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1481 		seq_puts(seq, ",check_int");
1482 	if (info->check_integrity_print_mask)
1483 		seq_printf(seq, ",check_int_print_mask=%d",
1484 				info->check_integrity_print_mask);
1485 #endif
1486 	if (info->metadata_ratio)
1487 		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1488 	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1489 		seq_puts(seq, ",fatal_errors=panic");
1490 	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1491 		seq_printf(seq, ",commit=%u", info->commit_interval);
1492 #ifdef CONFIG_BTRFS_DEBUG
1493 	if (btrfs_test_opt(info, FRAGMENT_DATA))
1494 		seq_puts(seq, ",fragment=data");
1495 	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1496 		seq_puts(seq, ",fragment=metadata");
1497 #endif
1498 	if (btrfs_test_opt(info, REF_VERIFY))
1499 		seq_puts(seq, ",ref_verify");
1500 	seq_printf(seq, ",subvolid=%llu",
1501 		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1502 	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1503 			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1504 	if (!IS_ERR(subvol_name)) {
1505 		seq_puts(seq, ",subvol=");
1506 		seq_escape(seq, subvol_name, " \t\n\\");
1507 		kfree(subvol_name);
1508 	}
1509 	return 0;
1510 }
1511 
btrfs_test_super(struct super_block * s,void * data)1512 static int btrfs_test_super(struct super_block *s, void *data)
1513 {
1514 	struct btrfs_fs_info *p = data;
1515 	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1516 
1517 	return fs_info->fs_devices == p->fs_devices;
1518 }
1519 
btrfs_set_super(struct super_block * s,void * data)1520 static int btrfs_set_super(struct super_block *s, void *data)
1521 {
1522 	int err = set_anon_super(s, data);
1523 	if (!err)
1524 		s->s_fs_info = data;
1525 	return err;
1526 }
1527 
1528 /*
1529  * subvolumes are identified by ino 256
1530  */
is_subvolume_inode(struct inode * inode)1531 static inline int is_subvolume_inode(struct inode *inode)
1532 {
1533 	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1534 		return 1;
1535 	return 0;
1536 }
1537 
mount_subvol(const char * subvol_name,u64 subvol_objectid,struct vfsmount * mnt)1538 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1539 				   struct vfsmount *mnt)
1540 {
1541 	struct dentry *root;
1542 	int ret;
1543 
1544 	if (!subvol_name) {
1545 		if (!subvol_objectid) {
1546 			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1547 							  &subvol_objectid);
1548 			if (ret) {
1549 				root = ERR_PTR(ret);
1550 				goto out;
1551 			}
1552 		}
1553 		subvol_name = btrfs_get_subvol_name_from_objectid(
1554 					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1555 		if (IS_ERR(subvol_name)) {
1556 			root = ERR_CAST(subvol_name);
1557 			subvol_name = NULL;
1558 			goto out;
1559 		}
1560 
1561 	}
1562 
1563 	root = mount_subtree(mnt, subvol_name);
1564 	/* mount_subtree() drops our reference on the vfsmount. */
1565 	mnt = NULL;
1566 
1567 	if (!IS_ERR(root)) {
1568 		struct super_block *s = root->d_sb;
1569 		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1570 		struct inode *root_inode = d_inode(root);
1571 		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1572 
1573 		ret = 0;
1574 		if (!is_subvolume_inode(root_inode)) {
1575 			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1576 			       subvol_name);
1577 			ret = -EINVAL;
1578 		}
1579 		if (subvol_objectid && root_objectid != subvol_objectid) {
1580 			/*
1581 			 * This will also catch a race condition where a
1582 			 * subvolume which was passed by ID is renamed and
1583 			 * another subvolume is renamed over the old location.
1584 			 */
1585 			btrfs_err(fs_info,
1586 				  "subvol '%s' does not match subvolid %llu",
1587 				  subvol_name, subvol_objectid);
1588 			ret = -EINVAL;
1589 		}
1590 		if (ret) {
1591 			dput(root);
1592 			root = ERR_PTR(ret);
1593 			deactivate_locked_super(s);
1594 		}
1595 	}
1596 
1597 out:
1598 	mntput(mnt);
1599 	kfree(subvol_name);
1600 	return root;
1601 }
1602 
1603 /*
1604  * Find a superblock for the given device / mount point.
1605  *
1606  * Note: This is based on mount_bdev from fs/super.c with a few additions
1607  *       for multiple device setup.  Make sure to keep it in sync.
1608  */
btrfs_mount_root(struct file_system_type * fs_type,int flags,const char * device_name,void * data)1609 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1610 		int flags, const char *device_name, void *data)
1611 {
1612 	struct block_device *bdev = NULL;
1613 	struct super_block *s;
1614 	struct btrfs_device *device = NULL;
1615 	struct btrfs_fs_devices *fs_devices = NULL;
1616 	struct btrfs_fs_info *fs_info = NULL;
1617 	void *new_sec_opts = NULL;
1618 	fmode_t mode = FMODE_READ;
1619 	int error = 0;
1620 
1621 	if (!(flags & SB_RDONLY))
1622 		mode |= FMODE_WRITE;
1623 
1624 	if (data) {
1625 		error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1626 		if (error)
1627 			return ERR_PTR(error);
1628 	}
1629 
1630 	/*
1631 	 * Setup a dummy root and fs_info for test/set super.  This is because
1632 	 * we don't actually fill this stuff out until open_ctree, but we need
1633 	 * then open_ctree will properly initialize the file system specific
1634 	 * settings later.  btrfs_init_fs_info initializes the static elements
1635 	 * of the fs_info (locks and such) to make cleanup easier if we find a
1636 	 * superblock with our given fs_devices later on at sget() time.
1637 	 */
1638 	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1639 	if (!fs_info) {
1640 		error = -ENOMEM;
1641 		goto error_sec_opts;
1642 	}
1643 	btrfs_init_fs_info(fs_info);
1644 
1645 	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1646 	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1647 	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1648 		error = -ENOMEM;
1649 		goto error_fs_info;
1650 	}
1651 
1652 	mutex_lock(&uuid_mutex);
1653 	error = btrfs_parse_device_options(data, mode, fs_type);
1654 	if (error) {
1655 		mutex_unlock(&uuid_mutex);
1656 		goto error_fs_info;
1657 	}
1658 
1659 	device = btrfs_scan_one_device(device_name, mode, fs_type);
1660 	if (IS_ERR(device)) {
1661 		mutex_unlock(&uuid_mutex);
1662 		error = PTR_ERR(device);
1663 		goto error_fs_info;
1664 	}
1665 
1666 	fs_devices = device->fs_devices;
1667 	fs_info->fs_devices = fs_devices;
1668 
1669 	error = btrfs_open_devices(fs_devices, mode, fs_type);
1670 	mutex_unlock(&uuid_mutex);
1671 	if (error)
1672 		goto error_fs_info;
1673 
1674 	if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1675 		error = -EACCES;
1676 		goto error_close_devices;
1677 	}
1678 
1679 	bdev = fs_devices->latest_bdev;
1680 	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1681 		 fs_info);
1682 	if (IS_ERR(s)) {
1683 		error = PTR_ERR(s);
1684 		goto error_close_devices;
1685 	}
1686 
1687 	if (s->s_root) {
1688 		btrfs_close_devices(fs_devices);
1689 		btrfs_free_fs_info(fs_info);
1690 		if ((flags ^ s->s_flags) & SB_RDONLY)
1691 			error = -EBUSY;
1692 	} else {
1693 		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1694 		btrfs_sb(s)->bdev_holder = fs_type;
1695 		if (!strstr(crc32c_impl(), "generic"))
1696 			set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1697 		error = btrfs_fill_super(s, fs_devices, data);
1698 	}
1699 	if (!error)
1700 		error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1701 	security_free_mnt_opts(&new_sec_opts);
1702 	if (error) {
1703 		deactivate_locked_super(s);
1704 		return ERR_PTR(error);
1705 	}
1706 
1707 	return dget(s->s_root);
1708 
1709 error_close_devices:
1710 	btrfs_close_devices(fs_devices);
1711 error_fs_info:
1712 	btrfs_free_fs_info(fs_info);
1713 error_sec_opts:
1714 	security_free_mnt_opts(&new_sec_opts);
1715 	return ERR_PTR(error);
1716 }
1717 
1718 /*
1719  * Mount function which is called by VFS layer.
1720  *
1721  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1722  * which needs vfsmount* of device's root (/).  This means device's root has to
1723  * be mounted internally in any case.
1724  *
1725  * Operation flow:
1726  *   1. Parse subvol id related options for later use in mount_subvol().
1727  *
1728  *   2. Mount device's root (/) by calling vfs_kern_mount().
1729  *
1730  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1731  *      first place. In order to avoid calling btrfs_mount() again, we use
1732  *      different file_system_type which is not registered to VFS by
1733  *      register_filesystem() (btrfs_root_fs_type). As a result,
1734  *      btrfs_mount_root() is called. The return value will be used by
1735  *      mount_subtree() in mount_subvol().
1736  *
1737  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1738  *      "btrfs subvolume set-default", mount_subvol() is called always.
1739  */
btrfs_mount(struct file_system_type * fs_type,int flags,const char * device_name,void * data)1740 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1741 		const char *device_name, void *data)
1742 {
1743 	struct vfsmount *mnt_root;
1744 	struct dentry *root;
1745 	char *subvol_name = NULL;
1746 	u64 subvol_objectid = 0;
1747 	int error = 0;
1748 
1749 	error = btrfs_parse_subvol_options(data, &subvol_name,
1750 					&subvol_objectid);
1751 	if (error) {
1752 		kfree(subvol_name);
1753 		return ERR_PTR(error);
1754 	}
1755 
1756 	/* mount device's root (/) */
1757 	mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1758 	if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1759 		if (flags & SB_RDONLY) {
1760 			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1761 				flags & ~SB_RDONLY, device_name, data);
1762 		} else {
1763 			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1764 				flags | SB_RDONLY, device_name, data);
1765 			if (IS_ERR(mnt_root)) {
1766 				root = ERR_CAST(mnt_root);
1767 				kfree(subvol_name);
1768 				goto out;
1769 			}
1770 
1771 			down_write(&mnt_root->mnt_sb->s_umount);
1772 			error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1773 			up_write(&mnt_root->mnt_sb->s_umount);
1774 			if (error < 0) {
1775 				root = ERR_PTR(error);
1776 				mntput(mnt_root);
1777 				kfree(subvol_name);
1778 				goto out;
1779 			}
1780 		}
1781 	}
1782 	if (IS_ERR(mnt_root)) {
1783 		root = ERR_CAST(mnt_root);
1784 		kfree(subvol_name);
1785 		goto out;
1786 	}
1787 
1788 	/* mount_subvol() will free subvol_name and mnt_root */
1789 	root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1790 
1791 out:
1792 	return root;
1793 }
1794 
btrfs_resize_thread_pool(struct btrfs_fs_info * fs_info,u32 new_pool_size,u32 old_pool_size)1795 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1796 				     u32 new_pool_size, u32 old_pool_size)
1797 {
1798 	if (new_pool_size == old_pool_size)
1799 		return;
1800 
1801 	fs_info->thread_pool_size = new_pool_size;
1802 
1803 	btrfs_info(fs_info, "resize thread pool %d -> %d",
1804 	       old_pool_size, new_pool_size);
1805 
1806 	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1807 	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1808 	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1809 	btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1810 	btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1811 	btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1812 				new_pool_size);
1813 	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1814 	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1815 	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1816 	btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1817 	btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1818 				new_pool_size);
1819 }
1820 
btrfs_remount_begin(struct btrfs_fs_info * fs_info,unsigned long old_opts,int flags)1821 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1822 				       unsigned long old_opts, int flags)
1823 {
1824 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1825 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1826 	     (flags & SB_RDONLY))) {
1827 		/* wait for any defraggers to finish */
1828 		wait_event(fs_info->transaction_wait,
1829 			   (atomic_read(&fs_info->defrag_running) == 0));
1830 		if (flags & SB_RDONLY)
1831 			sync_filesystem(fs_info->sb);
1832 	}
1833 }
1834 
btrfs_remount_cleanup(struct btrfs_fs_info * fs_info,unsigned long old_opts)1835 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1836 					 unsigned long old_opts)
1837 {
1838 	/*
1839 	 * We need to cleanup all defragable inodes if the autodefragment is
1840 	 * close or the filesystem is read only.
1841 	 */
1842 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1843 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1844 		btrfs_cleanup_defrag_inodes(fs_info);
1845 	}
1846 
1847 	/* If we toggled discard async */
1848 	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1849 	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1850 		btrfs_discard_resume(fs_info);
1851 	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1852 		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1853 		btrfs_discard_cleanup(fs_info);
1854 }
1855 
btrfs_remount(struct super_block * sb,int * flags,char * data)1856 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1857 {
1858 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1859 	struct btrfs_root *root = fs_info->tree_root;
1860 	unsigned old_flags = sb->s_flags;
1861 	unsigned long old_opts = fs_info->mount_opt;
1862 	unsigned long old_compress_type = fs_info->compress_type;
1863 	u64 old_max_inline = fs_info->max_inline;
1864 	u32 old_thread_pool_size = fs_info->thread_pool_size;
1865 	u32 old_metadata_ratio = fs_info->metadata_ratio;
1866 	int ret;
1867 
1868 	sync_filesystem(sb);
1869 	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1870 
1871 	if (data) {
1872 		void *new_sec_opts = NULL;
1873 
1874 		ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1875 		if (!ret)
1876 			ret = security_sb_remount(sb, new_sec_opts);
1877 		security_free_mnt_opts(&new_sec_opts);
1878 		if (ret)
1879 			goto restore;
1880 	}
1881 
1882 	ret = btrfs_parse_options(fs_info, data, *flags);
1883 	if (ret)
1884 		goto restore;
1885 
1886 	btrfs_remount_begin(fs_info, old_opts, *flags);
1887 	btrfs_resize_thread_pool(fs_info,
1888 		fs_info->thread_pool_size, old_thread_pool_size);
1889 
1890 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1891 		goto out;
1892 
1893 	if (*flags & SB_RDONLY) {
1894 		/*
1895 		 * this also happens on 'umount -rf' or on shutdown, when
1896 		 * the filesystem is busy.
1897 		 */
1898 		cancel_work_sync(&fs_info->async_reclaim_work);
1899 		cancel_work_sync(&fs_info->async_data_reclaim_work);
1900 
1901 		btrfs_discard_cleanup(fs_info);
1902 
1903 		/* wait for the uuid_scan task to finish */
1904 		down(&fs_info->uuid_tree_rescan_sem);
1905 		/* avoid complains from lockdep et al. */
1906 		up(&fs_info->uuid_tree_rescan_sem);
1907 
1908 		sb->s_flags |= SB_RDONLY;
1909 
1910 		/*
1911 		 * Setting SB_RDONLY will put the cleaner thread to
1912 		 * sleep at the next loop if it's already active.
1913 		 * If it's already asleep, we'll leave unused block
1914 		 * groups on disk until we're mounted read-write again
1915 		 * unless we clean them up here.
1916 		 */
1917 		btrfs_delete_unused_bgs(fs_info);
1918 
1919 		btrfs_dev_replace_suspend_for_unmount(fs_info);
1920 		btrfs_scrub_cancel(fs_info);
1921 		btrfs_pause_balance(fs_info);
1922 
1923 		/*
1924 		 * Pause the qgroup rescan worker if it is running. We don't want
1925 		 * it to be still running after we are in RO mode, as after that,
1926 		 * by the time we unmount, it might have left a transaction open,
1927 		 * so we would leak the transaction and/or crash.
1928 		 */
1929 		btrfs_qgroup_wait_for_completion(fs_info, false);
1930 
1931 		ret = btrfs_commit_super(fs_info);
1932 		if (ret)
1933 			goto restore;
1934 	} else {
1935 		if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
1936 			btrfs_err(fs_info,
1937 				"Remounting read-write after error is not allowed");
1938 			ret = -EINVAL;
1939 			goto restore;
1940 		}
1941 		if (fs_info->fs_devices->rw_devices == 0) {
1942 			ret = -EACCES;
1943 			goto restore;
1944 		}
1945 
1946 		if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1947 			btrfs_warn(fs_info,
1948 		"too many missing devices, writable remount is not allowed");
1949 			ret = -EACCES;
1950 			goto restore;
1951 		}
1952 
1953 		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1954 			btrfs_warn(fs_info,
1955 		"mount required to replay tree-log, cannot remount read-write");
1956 			ret = -EINVAL;
1957 			goto restore;
1958 		}
1959 
1960 		ret = btrfs_cleanup_fs_roots(fs_info);
1961 		if (ret)
1962 			goto restore;
1963 
1964 		/* recover relocation */
1965 		mutex_lock(&fs_info->cleaner_mutex);
1966 		ret = btrfs_recover_relocation(root);
1967 		mutex_unlock(&fs_info->cleaner_mutex);
1968 		if (ret)
1969 			goto restore;
1970 
1971 		ret = btrfs_resume_balance_async(fs_info);
1972 		if (ret)
1973 			goto restore;
1974 
1975 		ret = btrfs_resume_dev_replace_async(fs_info);
1976 		if (ret) {
1977 			btrfs_warn(fs_info, "failed to resume dev_replace");
1978 			goto restore;
1979 		}
1980 
1981 		btrfs_qgroup_rescan_resume(fs_info);
1982 
1983 		if (!fs_info->uuid_root) {
1984 			btrfs_info(fs_info, "creating UUID tree");
1985 			ret = btrfs_create_uuid_tree(fs_info);
1986 			if (ret) {
1987 				btrfs_warn(fs_info,
1988 					   "failed to create the UUID tree %d",
1989 					   ret);
1990 				goto restore;
1991 			}
1992 		}
1993 		sb->s_flags &= ~SB_RDONLY;
1994 
1995 		set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1996 	}
1997 out:
1998 	/*
1999 	 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2000 	 * since the absence of the flag means it can be toggled off by remount.
2001 	 */
2002 	*flags |= SB_I_VERSION;
2003 
2004 	wake_up_process(fs_info->transaction_kthread);
2005 	btrfs_remount_cleanup(fs_info, old_opts);
2006 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2007 
2008 	return 0;
2009 
2010 restore:
2011 	/* We've hit an error - don't reset SB_RDONLY */
2012 	if (sb_rdonly(sb))
2013 		old_flags |= SB_RDONLY;
2014 	sb->s_flags = old_flags;
2015 	fs_info->mount_opt = old_opts;
2016 	fs_info->compress_type = old_compress_type;
2017 	fs_info->max_inline = old_max_inline;
2018 	btrfs_resize_thread_pool(fs_info,
2019 		old_thread_pool_size, fs_info->thread_pool_size);
2020 	fs_info->metadata_ratio = old_metadata_ratio;
2021 	btrfs_remount_cleanup(fs_info, old_opts);
2022 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2023 
2024 	return ret;
2025 }
2026 
2027 /* Used to sort the devices by max_avail(descending sort) */
btrfs_cmp_device_free_bytes(const void * dev_info1,const void * dev_info2)2028 static inline int btrfs_cmp_device_free_bytes(const void *dev_info1,
2029 				       const void *dev_info2)
2030 {
2031 	if (((struct btrfs_device_info *)dev_info1)->max_avail >
2032 	    ((struct btrfs_device_info *)dev_info2)->max_avail)
2033 		return -1;
2034 	else if (((struct btrfs_device_info *)dev_info1)->max_avail <
2035 		 ((struct btrfs_device_info *)dev_info2)->max_avail)
2036 		return 1;
2037 	else
2038 	return 0;
2039 }
2040 
2041 /*
2042  * sort the devices by max_avail, in which max free extent size of each device
2043  * is stored.(Descending Sort)
2044  */
btrfs_descending_sort_devices(struct btrfs_device_info * devices,size_t nr_devices)2045 static inline void btrfs_descending_sort_devices(
2046 					struct btrfs_device_info *devices,
2047 					size_t nr_devices)
2048 {
2049 	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2050 	     btrfs_cmp_device_free_bytes, NULL);
2051 }
2052 
2053 /*
2054  * The helper to calc the free space on the devices that can be used to store
2055  * file data.
2056  */
btrfs_calc_avail_data_space(struct btrfs_fs_info * fs_info,u64 * free_bytes)2057 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2058 					      u64 *free_bytes)
2059 {
2060 	struct btrfs_device_info *devices_info;
2061 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2062 	struct btrfs_device *device;
2063 	u64 type;
2064 	u64 avail_space;
2065 	u64 min_stripe_size;
2066 	int num_stripes = 1;
2067 	int i = 0, nr_devices;
2068 	const struct btrfs_raid_attr *rattr;
2069 
2070 	/*
2071 	 * We aren't under the device list lock, so this is racy-ish, but good
2072 	 * enough for our purposes.
2073 	 */
2074 	nr_devices = fs_info->fs_devices->open_devices;
2075 	if (!nr_devices) {
2076 		smp_mb();
2077 		nr_devices = fs_info->fs_devices->open_devices;
2078 		ASSERT(nr_devices);
2079 		if (!nr_devices) {
2080 			*free_bytes = 0;
2081 			return 0;
2082 		}
2083 	}
2084 
2085 	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2086 			       GFP_KERNEL);
2087 	if (!devices_info)
2088 		return -ENOMEM;
2089 
2090 	/* calc min stripe number for data space allocation */
2091 	type = btrfs_data_alloc_profile(fs_info);
2092 	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2093 
2094 	if (type & BTRFS_BLOCK_GROUP_RAID0)
2095 		num_stripes = nr_devices;
2096 	else if (type & BTRFS_BLOCK_GROUP_RAID1)
2097 		num_stripes = 2;
2098 	else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2099 		num_stripes = 3;
2100 	else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2101 		num_stripes = 4;
2102 	else if (type & BTRFS_BLOCK_GROUP_RAID10)
2103 		num_stripes = 4;
2104 
2105 	/* Adjust for more than 1 stripe per device */
2106 	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2107 
2108 	rcu_read_lock();
2109 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2110 		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2111 						&device->dev_state) ||
2112 		    !device->bdev ||
2113 		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2114 			continue;
2115 
2116 		if (i >= nr_devices)
2117 			break;
2118 
2119 		avail_space = device->total_bytes - device->bytes_used;
2120 
2121 		/* align with stripe_len */
2122 		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2123 
2124 		/*
2125 		 * In order to avoid overwriting the superblock on the drive,
2126 		 * btrfs starts at an offset of at least 1MB when doing chunk
2127 		 * allocation.
2128 		 *
2129 		 * This ensures we have at least min_stripe_size free space
2130 		 * after excluding 1MB.
2131 		 */
2132 		if (avail_space <= SZ_1M + min_stripe_size)
2133 			continue;
2134 
2135 		avail_space -= SZ_1M;
2136 
2137 		devices_info[i].dev = device;
2138 		devices_info[i].max_avail = avail_space;
2139 
2140 		i++;
2141 	}
2142 	rcu_read_unlock();
2143 
2144 	nr_devices = i;
2145 
2146 	btrfs_descending_sort_devices(devices_info, nr_devices);
2147 
2148 	i = nr_devices - 1;
2149 	avail_space = 0;
2150 	while (nr_devices >= rattr->devs_min) {
2151 		num_stripes = min(num_stripes, nr_devices);
2152 
2153 		if (devices_info[i].max_avail >= min_stripe_size) {
2154 			int j;
2155 			u64 alloc_size;
2156 
2157 			avail_space += devices_info[i].max_avail * num_stripes;
2158 			alloc_size = devices_info[i].max_avail;
2159 			for (j = i + 1 - num_stripes; j <= i; j++)
2160 				devices_info[j].max_avail -= alloc_size;
2161 		}
2162 		i--;
2163 		nr_devices--;
2164 	}
2165 
2166 	kfree(devices_info);
2167 	*free_bytes = avail_space;
2168 	return 0;
2169 }
2170 
2171 /*
2172  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2173  *
2174  * If there's a redundant raid level at DATA block groups, use the respective
2175  * multiplier to scale the sizes.
2176  *
2177  * Unused device space usage is based on simulating the chunk allocator
2178  * algorithm that respects the device sizes and order of allocations.  This is
2179  * a close approximation of the actual use but there are other factors that may
2180  * change the result (like a new metadata chunk).
2181  *
2182  * If metadata is exhausted, f_bavail will be 0.
2183  */
btrfs_statfs(struct dentry * dentry,struct kstatfs * buf)2184 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2185 {
2186 	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2187 	struct btrfs_super_block *disk_super = fs_info->super_copy;
2188 	struct btrfs_space_info *found;
2189 	u64 total_used = 0;
2190 	u64 total_free_data = 0;
2191 	u64 total_free_meta = 0;
2192 	int bits = dentry->d_sb->s_blocksize_bits;
2193 	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2194 	unsigned factor = 1;
2195 	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2196 	int ret;
2197 	u64 thresh = 0;
2198 	int mixed = 0;
2199 
2200 	list_for_each_entry(found, &fs_info->space_info, list) {
2201 		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2202 			int i;
2203 
2204 			total_free_data += found->disk_total - found->disk_used;
2205 			total_free_data -=
2206 				btrfs_account_ro_block_groups_free_space(found);
2207 
2208 			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2209 				if (!list_empty(&found->block_groups[i]))
2210 					factor = btrfs_bg_type_to_factor(
2211 						btrfs_raid_array[i].bg_flag);
2212 			}
2213 		}
2214 
2215 		/*
2216 		 * Metadata in mixed block goup profiles are accounted in data
2217 		 */
2218 		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2219 			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2220 				mixed = 1;
2221 			else
2222 				total_free_meta += found->disk_total -
2223 					found->disk_used;
2224 		}
2225 
2226 		total_used += found->disk_used;
2227 	}
2228 
2229 	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2230 	buf->f_blocks >>= bits;
2231 	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2232 
2233 	/* Account global block reserve as used, it's in logical size already */
2234 	spin_lock(&block_rsv->lock);
2235 	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
2236 	if (buf->f_bfree >= block_rsv->size >> bits)
2237 		buf->f_bfree -= block_rsv->size >> bits;
2238 	else
2239 		buf->f_bfree = 0;
2240 	spin_unlock(&block_rsv->lock);
2241 
2242 	buf->f_bavail = div_u64(total_free_data, factor);
2243 	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2244 	if (ret)
2245 		return ret;
2246 	buf->f_bavail += div_u64(total_free_data, factor);
2247 	buf->f_bavail = buf->f_bavail >> bits;
2248 
2249 	/*
2250 	 * We calculate the remaining metadata space minus global reserve. If
2251 	 * this is (supposedly) smaller than zero, there's no space. But this
2252 	 * does not hold in practice, the exhausted state happens where's still
2253 	 * some positive delta. So we apply some guesswork and compare the
2254 	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2255 	 *
2256 	 * We probably cannot calculate the exact threshold value because this
2257 	 * depends on the internal reservations requested by various
2258 	 * operations, so some operations that consume a few metadata will
2259 	 * succeed even if the Avail is zero. But this is better than the other
2260 	 * way around.
2261 	 */
2262 	thresh = SZ_4M;
2263 
2264 	/*
2265 	 * We only want to claim there's no available space if we can no longer
2266 	 * allocate chunks for our metadata profile and our global reserve will
2267 	 * not fit in the free metadata space.  If we aren't ->full then we
2268 	 * still can allocate chunks and thus are fine using the currently
2269 	 * calculated f_bavail.
2270 	 */
2271 	if (!mixed && block_rsv->space_info->full &&
2272 	    total_free_meta - thresh < block_rsv->size)
2273 		buf->f_bavail = 0;
2274 
2275 	buf->f_type = BTRFS_SUPER_MAGIC;
2276 	buf->f_bsize = dentry->d_sb->s_blocksize;
2277 	buf->f_namelen = BTRFS_NAME_LEN;
2278 
2279 	/* We treat it as constant endianness (it doesn't matter _which_)
2280 	   because we want the fsid to come out the same whether mounted
2281 	   on a big-endian or little-endian host */
2282 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2283 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2284 	/* Mask in the root object ID too, to disambiguate subvols */
2285 	buf->f_fsid.val[0] ^=
2286 		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2287 	buf->f_fsid.val[1] ^=
2288 		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2289 
2290 	return 0;
2291 }
2292 
btrfs_kill_super(struct super_block * sb)2293 static void btrfs_kill_super(struct super_block *sb)
2294 {
2295 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2296 	kill_anon_super(sb);
2297 	btrfs_free_fs_info(fs_info);
2298 }
2299 
2300 static struct file_system_type btrfs_fs_type = {
2301 	.owner		= THIS_MODULE,
2302 	.name		= "btrfs",
2303 	.mount		= btrfs_mount,
2304 	.kill_sb	= btrfs_kill_super,
2305 	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2306 };
2307 
2308 static struct file_system_type btrfs_root_fs_type = {
2309 	.owner		= THIS_MODULE,
2310 	.name		= "btrfs",
2311 	.mount		= btrfs_mount_root,
2312 	.kill_sb	= btrfs_kill_super,
2313 	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2314 };
2315 
2316 MODULE_ALIAS_FS("btrfs");
2317 
btrfs_control_open(struct inode * inode,struct file * file)2318 static int btrfs_control_open(struct inode *inode, struct file *file)
2319 {
2320 	/*
2321 	 * The control file's private_data is used to hold the
2322 	 * transaction when it is started and is used to keep
2323 	 * track of whether a transaction is already in progress.
2324 	 */
2325 	file->private_data = NULL;
2326 	return 0;
2327 }
2328 
2329 /*
2330  * Used by /dev/btrfs-control for devices ioctls.
2331  */
btrfs_control_ioctl(struct file * file,unsigned int cmd,unsigned long arg)2332 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2333 				unsigned long arg)
2334 {
2335 	struct btrfs_ioctl_vol_args *vol;
2336 	struct btrfs_device *device = NULL;
2337 	int ret = -ENOTTY;
2338 
2339 	if (!capable(CAP_SYS_ADMIN))
2340 		return -EPERM;
2341 
2342 	vol = memdup_user((void __user *)arg, sizeof(*vol));
2343 	if (IS_ERR(vol))
2344 		return PTR_ERR(vol);
2345 	vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2346 
2347 	switch (cmd) {
2348 	case BTRFS_IOC_SCAN_DEV:
2349 		mutex_lock(&uuid_mutex);
2350 		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2351 					       &btrfs_root_fs_type);
2352 		ret = PTR_ERR_OR_ZERO(device);
2353 		mutex_unlock(&uuid_mutex);
2354 		break;
2355 	case BTRFS_IOC_FORGET_DEV:
2356 		ret = btrfs_forget_devices(vol->name);
2357 		break;
2358 	case BTRFS_IOC_DEVICES_READY:
2359 		mutex_lock(&uuid_mutex);
2360 		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2361 					       &btrfs_root_fs_type);
2362 		if (IS_ERR(device)) {
2363 			mutex_unlock(&uuid_mutex);
2364 			ret = PTR_ERR(device);
2365 			break;
2366 		}
2367 		ret = !(device->fs_devices->num_devices ==
2368 			device->fs_devices->total_devices);
2369 		mutex_unlock(&uuid_mutex);
2370 		break;
2371 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2372 		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2373 		break;
2374 	}
2375 
2376 	kfree(vol);
2377 	return ret;
2378 }
2379 
btrfs_freeze(struct super_block * sb)2380 static int btrfs_freeze(struct super_block *sb)
2381 {
2382 	struct btrfs_trans_handle *trans;
2383 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2384 	struct btrfs_root *root = fs_info->tree_root;
2385 
2386 	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2387 	/*
2388 	 * We don't need a barrier here, we'll wait for any transaction that
2389 	 * could be in progress on other threads (and do delayed iputs that
2390 	 * we want to avoid on a frozen filesystem), or do the commit
2391 	 * ourselves.
2392 	 */
2393 	trans = btrfs_attach_transaction_barrier(root);
2394 	if (IS_ERR(trans)) {
2395 		/* no transaction, don't bother */
2396 		if (PTR_ERR(trans) == -ENOENT)
2397 			return 0;
2398 		return PTR_ERR(trans);
2399 	}
2400 	return btrfs_commit_transaction(trans);
2401 }
2402 
btrfs_unfreeze(struct super_block * sb)2403 static int btrfs_unfreeze(struct super_block *sb)
2404 {
2405 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2406 
2407 	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2408 	return 0;
2409 }
2410 
btrfs_show_devname(struct seq_file * m,struct dentry * root)2411 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2412 {
2413 	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2414 	struct btrfs_device *dev, *first_dev = NULL;
2415 
2416 	/*
2417 	 * Lightweight locking of the devices. We should not need
2418 	 * device_list_mutex here as we only read the device data and the list
2419 	 * is protected by RCU.  Even if a device is deleted during the list
2420 	 * traversals, we'll get valid data, the freeing callback will wait at
2421 	 * least until the rcu_read_unlock.
2422 	 */
2423 	rcu_read_lock();
2424 	list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2425 		if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2426 			continue;
2427 		if (!dev->name)
2428 			continue;
2429 		if (!first_dev || dev->devid < first_dev->devid)
2430 			first_dev = dev;
2431 	}
2432 
2433 	if (first_dev)
2434 		seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2435 	else
2436 		WARN_ON(1);
2437 	rcu_read_unlock();
2438 	return 0;
2439 }
2440 
2441 static const struct super_operations btrfs_super_ops = {
2442 	.drop_inode	= btrfs_drop_inode,
2443 	.evict_inode	= btrfs_evict_inode,
2444 	.put_super	= btrfs_put_super,
2445 	.sync_fs	= btrfs_sync_fs,
2446 	.show_options	= btrfs_show_options,
2447 	.show_devname	= btrfs_show_devname,
2448 	.alloc_inode	= btrfs_alloc_inode,
2449 	.destroy_inode	= btrfs_destroy_inode,
2450 	.free_inode	= btrfs_free_inode,
2451 	.statfs		= btrfs_statfs,
2452 	.remount_fs	= btrfs_remount,
2453 	.freeze_fs	= btrfs_freeze,
2454 	.unfreeze_fs	= btrfs_unfreeze,
2455 };
2456 
2457 static const struct file_operations btrfs_ctl_fops = {
2458 	.open = btrfs_control_open,
2459 	.unlocked_ioctl	 = btrfs_control_ioctl,
2460 	.compat_ioctl = compat_ptr_ioctl,
2461 	.owner	 = THIS_MODULE,
2462 	.llseek = noop_llseek,
2463 };
2464 
2465 static struct miscdevice btrfs_misc = {
2466 	.minor		= BTRFS_MINOR,
2467 	.name		= "btrfs-control",
2468 	.fops		= &btrfs_ctl_fops
2469 };
2470 
2471 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2472 MODULE_ALIAS("devname:btrfs-control");
2473 
btrfs_interface_init(void)2474 static int __init btrfs_interface_init(void)
2475 {
2476 	return misc_register(&btrfs_misc);
2477 }
2478 
btrfs_interface_exit(void)2479 static __cold void btrfs_interface_exit(void)
2480 {
2481 	misc_deregister(&btrfs_misc);
2482 }
2483 
btrfs_print_mod_info(void)2484 static void __init btrfs_print_mod_info(void)
2485 {
2486 	static const char options[] = ""
2487 #ifdef CONFIG_BTRFS_DEBUG
2488 			", debug=on"
2489 #endif
2490 #ifdef CONFIG_BTRFS_ASSERT
2491 			", assert=on"
2492 #endif
2493 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2494 			", integrity-checker=on"
2495 #endif
2496 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2497 			", ref-verify=on"
2498 #endif
2499 			;
2500 	pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2501 }
2502 
init_btrfs_fs(void)2503 static int __init init_btrfs_fs(void)
2504 {
2505 	int err;
2506 
2507 	btrfs_props_init();
2508 
2509 	err = btrfs_init_sysfs();
2510 	if (err)
2511 		return err;
2512 
2513 	btrfs_init_compress();
2514 
2515 	err = btrfs_init_cachep();
2516 	if (err)
2517 		goto free_compress;
2518 
2519 	err = extent_io_init();
2520 	if (err)
2521 		goto free_cachep;
2522 
2523 	err = extent_state_cache_init();
2524 	if (err)
2525 		goto free_extent_io;
2526 
2527 	err = extent_map_init();
2528 	if (err)
2529 		goto free_extent_state_cache;
2530 
2531 	err = ordered_data_init();
2532 	if (err)
2533 		goto free_extent_map;
2534 
2535 	err = btrfs_delayed_inode_init();
2536 	if (err)
2537 		goto free_ordered_data;
2538 
2539 	err = btrfs_auto_defrag_init();
2540 	if (err)
2541 		goto free_delayed_inode;
2542 
2543 	err = btrfs_delayed_ref_init();
2544 	if (err)
2545 		goto free_auto_defrag;
2546 
2547 	err = btrfs_prelim_ref_init();
2548 	if (err)
2549 		goto free_delayed_ref;
2550 
2551 	err = btrfs_end_io_wq_init();
2552 	if (err)
2553 		goto free_prelim_ref;
2554 
2555 	err = btrfs_interface_init();
2556 	if (err)
2557 		goto free_end_io_wq;
2558 
2559 	btrfs_init_lockdep();
2560 
2561 	btrfs_print_mod_info();
2562 
2563 	err = btrfs_run_sanity_tests();
2564 	if (err)
2565 		goto unregister_ioctl;
2566 
2567 	err = register_filesystem(&btrfs_fs_type);
2568 	if (err)
2569 		goto unregister_ioctl;
2570 
2571 	return 0;
2572 
2573 unregister_ioctl:
2574 	btrfs_interface_exit();
2575 free_end_io_wq:
2576 	btrfs_end_io_wq_exit();
2577 free_prelim_ref:
2578 	btrfs_prelim_ref_exit();
2579 free_delayed_ref:
2580 	btrfs_delayed_ref_exit();
2581 free_auto_defrag:
2582 	btrfs_auto_defrag_exit();
2583 free_delayed_inode:
2584 	btrfs_delayed_inode_exit();
2585 free_ordered_data:
2586 	ordered_data_exit();
2587 free_extent_map:
2588 	extent_map_exit();
2589 free_extent_state_cache:
2590 	extent_state_cache_exit();
2591 free_extent_io:
2592 	extent_io_exit();
2593 free_cachep:
2594 	btrfs_destroy_cachep();
2595 free_compress:
2596 	btrfs_exit_compress();
2597 	btrfs_exit_sysfs();
2598 
2599 	return err;
2600 }
2601 
exit_btrfs_fs(void)2602 static void __exit exit_btrfs_fs(void)
2603 {
2604 	btrfs_destroy_cachep();
2605 	btrfs_delayed_ref_exit();
2606 	btrfs_auto_defrag_exit();
2607 	btrfs_delayed_inode_exit();
2608 	btrfs_prelim_ref_exit();
2609 	ordered_data_exit();
2610 	extent_map_exit();
2611 	extent_state_cache_exit();
2612 	extent_io_exit();
2613 	btrfs_interface_exit();
2614 	btrfs_end_io_wq_exit();
2615 	unregister_filesystem(&btrfs_fs_type);
2616 	btrfs_exit_sysfs();
2617 	btrfs_cleanup_fs_uuids();
2618 	btrfs_exit_compress();
2619 }
2620 
2621 late_initcall(init_btrfs_fs);
2622 module_exit(exit_btrfs_fs)
2623 
2624 MODULE_LICENSE("GPL");
2625 MODULE_SOFTDEP("pre: crc32c");
2626 MODULE_SOFTDEP("pre: xxhash64");
2627 MODULE_SOFTDEP("pre: sha256");
2628 MODULE_SOFTDEP("pre: blake2b-256");
2629