xref: /openbmc/linux/fs/btrfs/super.c (revision 63a7cb13071842966c1ce931edacbc23573aada5)
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/ratelimit.h>
27 #include <linux/crc32c.h>
28 #include <linux/btrfs.h>
29 #include "delayed-inode.h"
30 #include "ctree.h"
31 #include "disk-io.h"
32 #include "transaction.h"
33 #include "btrfs_inode.h"
34 #include "print-tree.h"
35 #include "props.h"
36 #include "xattr.h"
37 #include "volumes.h"
38 #include "export.h"
39 #include "compression.h"
40 #include "rcu-string.h"
41 #include "dev-replace.h"
42 #include "free-space-cache.h"
43 #include "backref.h"
44 #include "space-info.h"
45 #include "sysfs.h"
46 #include "zoned.h"
47 #include "tests/btrfs-tests.h"
48 #include "block-group.h"
49 #include "discard.h"
50 #include "qgroup.h"
51 #include "raid56.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 #ifdef CONFIG_PRINTK
71 
72 #define STATE_STRING_PREFACE	": state "
73 #define STATE_STRING_BUF_LEN	(sizeof(STATE_STRING_PREFACE) + BTRFS_FS_STATE_COUNT)
74 
75 /*
76  * Characters to print to indicate error conditions or uncommon filesystem state.
77  * RO is not an error.
78  */
79 static const char fs_state_chars[] = {
80 	[BTRFS_FS_STATE_ERROR]			= 'E',
81 	[BTRFS_FS_STATE_REMOUNTING]		= 'M',
82 	[BTRFS_FS_STATE_RO]			= 0,
83 	[BTRFS_FS_STATE_TRANS_ABORTED]		= 'A',
84 	[BTRFS_FS_STATE_DEV_REPLACING]		= 'R',
85 	[BTRFS_FS_STATE_DUMMY_FS_INFO]		= 0,
86 	[BTRFS_FS_STATE_NO_CSUMS]		= 'C',
87 	[BTRFS_FS_STATE_LOG_CLEANUP_ERROR]	= 'L',
88 };
89 
90 static void btrfs_state_to_string(const struct btrfs_fs_info *info, char *buf)
91 {
92 	unsigned int bit;
93 	bool states_printed = false;
94 	unsigned long fs_state = READ_ONCE(info->fs_state);
95 	char *curr = buf;
96 
97 	memcpy(curr, STATE_STRING_PREFACE, sizeof(STATE_STRING_PREFACE));
98 	curr += sizeof(STATE_STRING_PREFACE) - 1;
99 
100 	for_each_set_bit(bit, &fs_state, sizeof(fs_state)) {
101 		WARN_ON_ONCE(bit >= BTRFS_FS_STATE_COUNT);
102 		if ((bit < BTRFS_FS_STATE_COUNT) && fs_state_chars[bit]) {
103 			*curr++ = fs_state_chars[bit];
104 			states_printed = true;
105 		}
106 	}
107 
108 	/* If no states were printed, reset the buffer */
109 	if (!states_printed)
110 		curr = buf;
111 
112 	*curr++ = 0;
113 }
114 #endif
115 
116 /*
117  * Generally the error codes correspond to their respective errors, but there
118  * are a few special cases.
119  *
120  * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
121  *          instance will return EUCLEAN if any of the blocks are corrupted in
122  *          a way that is problematic.  We want to reserve EUCLEAN for these
123  *          sort of corruptions.
124  *
125  * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
126  *        need to use EROFS for this case.  We will have no idea of the
127  *        original failure, that will have been reported at the time we tripped
128  *        over the error.  Each subsequent error that doesn't have any context
129  *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
130  */
131 const char * __attribute_const__ btrfs_decode_error(int errno)
132 {
133 	char *errstr = "unknown";
134 
135 	switch (errno) {
136 	case -ENOENT:		/* -2 */
137 		errstr = "No such entry";
138 		break;
139 	case -EIO:		/* -5 */
140 		errstr = "IO failure";
141 		break;
142 	case -ENOMEM:		/* -12*/
143 		errstr = "Out of memory";
144 		break;
145 	case -EEXIST:		/* -17 */
146 		errstr = "Object already exists";
147 		break;
148 	case -ENOSPC:		/* -28 */
149 		errstr = "No space left";
150 		break;
151 	case -EROFS:		/* -30 */
152 		errstr = "Readonly filesystem";
153 		break;
154 	case -EOPNOTSUPP:	/* -95 */
155 		errstr = "Operation not supported";
156 		break;
157 	case -EUCLEAN:		/* -117 */
158 		errstr = "Filesystem corrupted";
159 		break;
160 	case -EDQUOT:		/* -122 */
161 		errstr = "Quota exceeded";
162 		break;
163 	}
164 
165 	return errstr;
166 }
167 
168 /*
169  * __btrfs_handle_fs_error decodes expected errors from the caller and
170  * invokes the appropriate error response.
171  */
172 __cold
173 void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
174 		       unsigned int line, int errno, const char *fmt, ...)
175 {
176 	struct super_block *sb = fs_info->sb;
177 #ifdef CONFIG_PRINTK
178 	char statestr[STATE_STRING_BUF_LEN];
179 	const char *errstr;
180 #endif
181 
182 	/*
183 	 * Special case: if the error is EROFS, and we're already
184 	 * under SB_RDONLY, then it is safe here.
185 	 */
186 	if (errno == -EROFS && sb_rdonly(sb))
187   		return;
188 
189 #ifdef CONFIG_PRINTK
190 	errstr = btrfs_decode_error(errno);
191 	btrfs_state_to_string(fs_info, statestr);
192 	if (fmt) {
193 		struct va_format vaf;
194 		va_list args;
195 
196 		va_start(args, fmt);
197 		vaf.fmt = fmt;
198 		vaf.va = &args;
199 
200 		pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s (%pV)\n",
201 			sb->s_id, statestr, function, line, errno, errstr, &vaf);
202 		va_end(args);
203 	} else {
204 		pr_crit("BTRFS: error (device %s%s) in %s:%d: errno=%d %s\n",
205 			sb->s_id, statestr, function, line, errno, errstr);
206 	}
207 #endif
208 
209 	/*
210 	 * Today we only save the error info to memory.  Long term we'll
211 	 * also send it down to the disk
212 	 */
213 	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
214 
215 	/* Don't go through full error handling during mount */
216 	if (!(sb->s_flags & SB_BORN))
217 		return;
218 
219 	if (sb_rdonly(sb))
220 		return;
221 
222 	btrfs_discard_stop(fs_info);
223 
224 	/* btrfs handle error by forcing the filesystem readonly */
225 	btrfs_set_sb_rdonly(sb);
226 	btrfs_info(fs_info, "forced readonly");
227 	/*
228 	 * Note that a running device replace operation is not canceled here
229 	 * although there is no way to update the progress. It would add the
230 	 * risk of a deadlock, therefore the canceling is omitted. The only
231 	 * penalty is that some I/O remains active until the procedure
232 	 * completes. The next time when the filesystem is mounted writable
233 	 * again, the device replace operation continues.
234 	 */
235 }
236 
237 #ifdef CONFIG_PRINTK
238 static const char * const logtypes[] = {
239 	"emergency",
240 	"alert",
241 	"critical",
242 	"error",
243 	"warning",
244 	"notice",
245 	"info",
246 	"debug",
247 };
248 
249 
250 /*
251  * Use one ratelimit state per log level so that a flood of less important
252  * messages doesn't cause more important ones to be dropped.
253  */
254 static struct ratelimit_state printk_limits[] = {
255 	RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
256 	RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
257 	RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
258 	RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
259 	RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
260 	RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
261 	RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
262 	RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
263 };
264 
265 void __cold _btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
266 {
267 	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
268 	struct va_format vaf;
269 	va_list args;
270 	int kern_level;
271 	const char *type = logtypes[4];
272 	struct ratelimit_state *ratelimit = &printk_limits[4];
273 
274 	va_start(args, fmt);
275 
276 	while ((kern_level = printk_get_level(fmt)) != 0) {
277 		size_t size = printk_skip_level(fmt) - fmt;
278 
279 		if (kern_level >= '0' && kern_level <= '7') {
280 			memcpy(lvl, fmt,  size);
281 			lvl[size] = '\0';
282 			type = logtypes[kern_level - '0'];
283 			ratelimit = &printk_limits[kern_level - '0'];
284 		}
285 		fmt += size;
286 	}
287 
288 	vaf.fmt = fmt;
289 	vaf.va = &args;
290 
291 	if (__ratelimit(ratelimit)) {
292 		if (fs_info) {
293 			char statestr[STATE_STRING_BUF_LEN];
294 
295 			btrfs_state_to_string(fs_info, statestr);
296 			_printk("%sBTRFS %s (device %s%s): %pV\n", lvl, type,
297 				fs_info->sb->s_id, statestr, &vaf);
298 		} else {
299 			_printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
300 		}
301 	}
302 
303 	va_end(args);
304 }
305 #endif
306 
307 #if BITS_PER_LONG == 32
308 void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
309 {
310 	if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
311 		btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
312 		btrfs_warn(fs_info,
313 "due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
314 			   BTRFS_32BIT_MAX_FILE_SIZE >> 40);
315 		btrfs_warn(fs_info,
316 			   "please consider upgrading to 64bit kernel/hardware");
317 	}
318 }
319 
320 void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
321 {
322 	if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
323 		btrfs_err(fs_info, "reached 32bit limit for logical addresses");
324 		btrfs_err(fs_info,
325 "due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
326 			  BTRFS_32BIT_MAX_FILE_SIZE >> 40);
327 		btrfs_err(fs_info,
328 			   "please consider upgrading to 64bit kernel/hardware");
329 	}
330 }
331 #endif
332 
333 /*
334  * We only mark the transaction aborted and then set the file system read-only.
335  * This will prevent new transactions from starting or trying to join this
336  * one.
337  *
338  * This means that error recovery at the call site is limited to freeing
339  * any local memory allocations and passing the error code up without
340  * further cleanup. The transaction should complete as it normally would
341  * in the call path but will return -EIO.
342  *
343  * We'll complete the cleanup in btrfs_end_transaction and
344  * btrfs_commit_transaction.
345  */
346 __cold
347 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
348 			       const char *function,
349 			       unsigned int line, int errno, bool first_hit)
350 {
351 	struct btrfs_fs_info *fs_info = trans->fs_info;
352 
353 	WRITE_ONCE(trans->aborted, errno);
354 	WRITE_ONCE(trans->transaction->aborted, errno);
355 	if (first_hit && errno == -ENOSPC)
356 		btrfs_dump_space_info_for_trans_abort(fs_info);
357 	/* Wake up anybody who may be waiting on this transaction */
358 	wake_up(&fs_info->transaction_wait);
359 	wake_up(&fs_info->transaction_blocked_wait);
360 	__btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
361 }
362 /*
363  * __btrfs_panic decodes unexpected, fatal errors from the caller,
364  * issues an alert, and either panics or BUGs, depending on mount options.
365  */
366 __cold
367 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
368 		   unsigned int line, int errno, const char *fmt, ...)
369 {
370 	char *s_id = "<unknown>";
371 	const char *errstr;
372 	struct va_format vaf = { .fmt = fmt };
373 	va_list args;
374 
375 	if (fs_info)
376 		s_id = fs_info->sb->s_id;
377 
378 	va_start(args, fmt);
379 	vaf.va = &args;
380 
381 	errstr = btrfs_decode_error(errno);
382 	if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
383 		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
384 			s_id, function, line, &vaf, errno, errstr);
385 
386 	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
387 		   function, line, &vaf, errno, errstr);
388 	va_end(args);
389 	/* Caller calls BUG() */
390 }
391 
392 static void btrfs_put_super(struct super_block *sb)
393 {
394 	close_ctree(btrfs_sb(sb));
395 }
396 
397 enum {
398 	Opt_acl, Opt_noacl,
399 	Opt_clear_cache,
400 	Opt_commit_interval,
401 	Opt_compress,
402 	Opt_compress_force,
403 	Opt_compress_force_type,
404 	Opt_compress_type,
405 	Opt_degraded,
406 	Opt_device,
407 	Opt_fatal_errors,
408 	Opt_flushoncommit, Opt_noflushoncommit,
409 	Opt_max_inline,
410 	Opt_barrier, Opt_nobarrier,
411 	Opt_datacow, Opt_nodatacow,
412 	Opt_datasum, Opt_nodatasum,
413 	Opt_defrag, Opt_nodefrag,
414 	Opt_discard, Opt_nodiscard,
415 	Opt_discard_mode,
416 	Opt_norecovery,
417 	Opt_ratio,
418 	Opt_rescan_uuid_tree,
419 	Opt_skip_balance,
420 	Opt_space_cache, Opt_no_space_cache,
421 	Opt_space_cache_version,
422 	Opt_ssd, Opt_nossd,
423 	Opt_ssd_spread, Opt_nossd_spread,
424 	Opt_subvol,
425 	Opt_subvol_empty,
426 	Opt_subvolid,
427 	Opt_thread_pool,
428 	Opt_treelog, Opt_notreelog,
429 	Opt_user_subvol_rm_allowed,
430 
431 	/* Rescue options */
432 	Opt_rescue,
433 	Opt_usebackuproot,
434 	Opt_nologreplay,
435 	Opt_ignorebadroots,
436 	Opt_ignoredatacsums,
437 	Opt_rescue_all,
438 
439 	/* Deprecated options */
440 	Opt_recovery,
441 	Opt_inode_cache, Opt_noinode_cache,
442 
443 	/* Debugging options */
444 	Opt_check_integrity,
445 	Opt_check_integrity_including_extent_data,
446 	Opt_check_integrity_print_mask,
447 	Opt_enospc_debug, Opt_noenospc_debug,
448 #ifdef CONFIG_BTRFS_DEBUG
449 	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
450 #endif
451 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
452 	Opt_ref_verify,
453 #endif
454 	Opt_err,
455 };
456 
457 static const match_table_t tokens = {
458 	{Opt_acl, "acl"},
459 	{Opt_noacl, "noacl"},
460 	{Opt_clear_cache, "clear_cache"},
461 	{Opt_commit_interval, "commit=%u"},
462 	{Opt_compress, "compress"},
463 	{Opt_compress_type, "compress=%s"},
464 	{Opt_compress_force, "compress-force"},
465 	{Opt_compress_force_type, "compress-force=%s"},
466 	{Opt_degraded, "degraded"},
467 	{Opt_device, "device=%s"},
468 	{Opt_fatal_errors, "fatal_errors=%s"},
469 	{Opt_flushoncommit, "flushoncommit"},
470 	{Opt_noflushoncommit, "noflushoncommit"},
471 	{Opt_inode_cache, "inode_cache"},
472 	{Opt_noinode_cache, "noinode_cache"},
473 	{Opt_max_inline, "max_inline=%s"},
474 	{Opt_barrier, "barrier"},
475 	{Opt_nobarrier, "nobarrier"},
476 	{Opt_datacow, "datacow"},
477 	{Opt_nodatacow, "nodatacow"},
478 	{Opt_datasum, "datasum"},
479 	{Opt_nodatasum, "nodatasum"},
480 	{Opt_defrag, "autodefrag"},
481 	{Opt_nodefrag, "noautodefrag"},
482 	{Opt_discard, "discard"},
483 	{Opt_discard_mode, "discard=%s"},
484 	{Opt_nodiscard, "nodiscard"},
485 	{Opt_norecovery, "norecovery"},
486 	{Opt_ratio, "metadata_ratio=%u"},
487 	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
488 	{Opt_skip_balance, "skip_balance"},
489 	{Opt_space_cache, "space_cache"},
490 	{Opt_no_space_cache, "nospace_cache"},
491 	{Opt_space_cache_version, "space_cache=%s"},
492 	{Opt_ssd, "ssd"},
493 	{Opt_nossd, "nossd"},
494 	{Opt_ssd_spread, "ssd_spread"},
495 	{Opt_nossd_spread, "nossd_spread"},
496 	{Opt_subvol, "subvol=%s"},
497 	{Opt_subvol_empty, "subvol="},
498 	{Opt_subvolid, "subvolid=%s"},
499 	{Opt_thread_pool, "thread_pool=%u"},
500 	{Opt_treelog, "treelog"},
501 	{Opt_notreelog, "notreelog"},
502 	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
503 
504 	/* Rescue options */
505 	{Opt_rescue, "rescue=%s"},
506 	/* Deprecated, with alias rescue=nologreplay */
507 	{Opt_nologreplay, "nologreplay"},
508 	/* Deprecated, with alias rescue=usebackuproot */
509 	{Opt_usebackuproot, "usebackuproot"},
510 
511 	/* Deprecated options */
512 	{Opt_recovery, "recovery"},
513 
514 	/* Debugging options */
515 	{Opt_check_integrity, "check_int"},
516 	{Opt_check_integrity_including_extent_data, "check_int_data"},
517 	{Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
518 	{Opt_enospc_debug, "enospc_debug"},
519 	{Opt_noenospc_debug, "noenospc_debug"},
520 #ifdef CONFIG_BTRFS_DEBUG
521 	{Opt_fragment_data, "fragment=data"},
522 	{Opt_fragment_metadata, "fragment=metadata"},
523 	{Opt_fragment_all, "fragment=all"},
524 #endif
525 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
526 	{Opt_ref_verify, "ref_verify"},
527 #endif
528 	{Opt_err, NULL},
529 };
530 
531 static const match_table_t rescue_tokens = {
532 	{Opt_usebackuproot, "usebackuproot"},
533 	{Opt_nologreplay, "nologreplay"},
534 	{Opt_ignorebadroots, "ignorebadroots"},
535 	{Opt_ignorebadroots, "ibadroots"},
536 	{Opt_ignoredatacsums, "ignoredatacsums"},
537 	{Opt_ignoredatacsums, "idatacsums"},
538 	{Opt_rescue_all, "all"},
539 	{Opt_err, NULL},
540 };
541 
542 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
543 			    const char *opt_name)
544 {
545 	if (fs_info->mount_opt & opt) {
546 		btrfs_err(fs_info, "%s must be used with ro mount option",
547 			  opt_name);
548 		return true;
549 	}
550 	return false;
551 }
552 
553 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
554 {
555 	char *opts;
556 	char *orig;
557 	char *p;
558 	substring_t args[MAX_OPT_ARGS];
559 	int ret = 0;
560 
561 	opts = kstrdup(options, GFP_KERNEL);
562 	if (!opts)
563 		return -ENOMEM;
564 	orig = opts;
565 
566 	while ((p = strsep(&opts, ":")) != NULL) {
567 		int token;
568 
569 		if (!*p)
570 			continue;
571 		token = match_token(p, rescue_tokens, args);
572 		switch (token){
573 		case Opt_usebackuproot:
574 			btrfs_info(info,
575 				   "trying to use backup root at mount time");
576 			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
577 			break;
578 		case Opt_nologreplay:
579 			btrfs_set_and_info(info, NOLOGREPLAY,
580 					   "disabling log replay at mount time");
581 			break;
582 		case Opt_ignorebadroots:
583 			btrfs_set_and_info(info, IGNOREBADROOTS,
584 					   "ignoring bad roots");
585 			break;
586 		case Opt_ignoredatacsums:
587 			btrfs_set_and_info(info, IGNOREDATACSUMS,
588 					   "ignoring data csums");
589 			break;
590 		case Opt_rescue_all:
591 			btrfs_info(info, "enabling all of the rescue options");
592 			btrfs_set_and_info(info, IGNOREDATACSUMS,
593 					   "ignoring data csums");
594 			btrfs_set_and_info(info, IGNOREBADROOTS,
595 					   "ignoring bad roots");
596 			btrfs_set_and_info(info, NOLOGREPLAY,
597 					   "disabling log replay at mount time");
598 			break;
599 		case Opt_err:
600 			btrfs_info(info, "unrecognized rescue option '%s'", p);
601 			ret = -EINVAL;
602 			goto out;
603 		default:
604 			break;
605 		}
606 
607 	}
608 out:
609 	kfree(orig);
610 	return ret;
611 }
612 
613 /*
614  * Regular mount options parser.  Everything that is needed only when
615  * reading in a new superblock is parsed here.
616  * XXX JDM: This needs to be cleaned up for remount.
617  */
618 int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
619 			unsigned long new_flags)
620 {
621 	substring_t args[MAX_OPT_ARGS];
622 	char *p, *num;
623 	int intarg;
624 	int ret = 0;
625 	char *compress_type;
626 	bool compress_force = false;
627 	enum btrfs_compression_type saved_compress_type;
628 	int saved_compress_level;
629 	bool saved_compress_force;
630 	int no_compress = 0;
631 	const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state);
632 
633 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
634 		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
635 	else if (btrfs_free_space_cache_v1_active(info)) {
636 		if (btrfs_is_zoned(info)) {
637 			btrfs_info(info,
638 			"zoned: clearing existing space cache");
639 			btrfs_set_super_cache_generation(info->super_copy, 0);
640 		} else {
641 			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
642 		}
643 	}
644 
645 	/*
646 	 * Even the options are empty, we still need to do extra check
647 	 * against new flags
648 	 */
649 	if (!options)
650 		goto check;
651 
652 	while ((p = strsep(&options, ",")) != NULL) {
653 		int token;
654 		if (!*p)
655 			continue;
656 
657 		token = match_token(p, tokens, args);
658 		switch (token) {
659 		case Opt_degraded:
660 			btrfs_info(info, "allowing degraded mounts");
661 			btrfs_set_opt(info->mount_opt, DEGRADED);
662 			break;
663 		case Opt_subvol:
664 		case Opt_subvol_empty:
665 		case Opt_subvolid:
666 		case Opt_device:
667 			/*
668 			 * These are parsed by btrfs_parse_subvol_options or
669 			 * btrfs_parse_device_options and can be ignored here.
670 			 */
671 			break;
672 		case Opt_nodatasum:
673 			btrfs_set_and_info(info, NODATASUM,
674 					   "setting nodatasum");
675 			break;
676 		case Opt_datasum:
677 			if (btrfs_test_opt(info, NODATASUM)) {
678 				if (btrfs_test_opt(info, NODATACOW))
679 					btrfs_info(info,
680 						   "setting datasum, datacow enabled");
681 				else
682 					btrfs_info(info, "setting datasum");
683 			}
684 			btrfs_clear_opt(info->mount_opt, NODATACOW);
685 			btrfs_clear_opt(info->mount_opt, NODATASUM);
686 			break;
687 		case Opt_nodatacow:
688 			if (!btrfs_test_opt(info, NODATACOW)) {
689 				if (!btrfs_test_opt(info, COMPRESS) ||
690 				    !btrfs_test_opt(info, FORCE_COMPRESS)) {
691 					btrfs_info(info,
692 						   "setting nodatacow, compression disabled");
693 				} else {
694 					btrfs_info(info, "setting nodatacow");
695 				}
696 			}
697 			btrfs_clear_opt(info->mount_opt, COMPRESS);
698 			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
699 			btrfs_set_opt(info->mount_opt, NODATACOW);
700 			btrfs_set_opt(info->mount_opt, NODATASUM);
701 			break;
702 		case Opt_datacow:
703 			btrfs_clear_and_info(info, NODATACOW,
704 					     "setting datacow");
705 			break;
706 		case Opt_compress_force:
707 		case Opt_compress_force_type:
708 			compress_force = true;
709 			fallthrough;
710 		case Opt_compress:
711 		case Opt_compress_type:
712 			saved_compress_type = btrfs_test_opt(info,
713 							     COMPRESS) ?
714 				info->compress_type : BTRFS_COMPRESS_NONE;
715 			saved_compress_force =
716 				btrfs_test_opt(info, FORCE_COMPRESS);
717 			saved_compress_level = info->compress_level;
718 			if (token == Opt_compress ||
719 			    token == Opt_compress_force ||
720 			    strncmp(args[0].from, "zlib", 4) == 0) {
721 				compress_type = "zlib";
722 
723 				info->compress_type = BTRFS_COMPRESS_ZLIB;
724 				info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
725 				/*
726 				 * args[0] contains uninitialized data since
727 				 * for these tokens we don't expect any
728 				 * parameter.
729 				 */
730 				if (token != Opt_compress &&
731 				    token != Opt_compress_force)
732 					info->compress_level =
733 					  btrfs_compress_str2level(
734 							BTRFS_COMPRESS_ZLIB,
735 							args[0].from + 4);
736 				btrfs_set_opt(info->mount_opt, COMPRESS);
737 				btrfs_clear_opt(info->mount_opt, NODATACOW);
738 				btrfs_clear_opt(info->mount_opt, NODATASUM);
739 				no_compress = 0;
740 			} else if (strncmp(args[0].from, "lzo", 3) == 0) {
741 				compress_type = "lzo";
742 				info->compress_type = BTRFS_COMPRESS_LZO;
743 				info->compress_level = 0;
744 				btrfs_set_opt(info->mount_opt, COMPRESS);
745 				btrfs_clear_opt(info->mount_opt, NODATACOW);
746 				btrfs_clear_opt(info->mount_opt, NODATASUM);
747 				btrfs_set_fs_incompat(info, COMPRESS_LZO);
748 				no_compress = 0;
749 			} else if (strncmp(args[0].from, "zstd", 4) == 0) {
750 				compress_type = "zstd";
751 				info->compress_type = BTRFS_COMPRESS_ZSTD;
752 				info->compress_level =
753 					btrfs_compress_str2level(
754 							 BTRFS_COMPRESS_ZSTD,
755 							 args[0].from + 4);
756 				btrfs_set_opt(info->mount_opt, COMPRESS);
757 				btrfs_clear_opt(info->mount_opt, NODATACOW);
758 				btrfs_clear_opt(info->mount_opt, NODATASUM);
759 				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
760 				no_compress = 0;
761 			} else if (strncmp(args[0].from, "no", 2) == 0) {
762 				compress_type = "no";
763 				info->compress_level = 0;
764 				info->compress_type = 0;
765 				btrfs_clear_opt(info->mount_opt, COMPRESS);
766 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
767 				compress_force = false;
768 				no_compress++;
769 			} else {
770 				btrfs_err(info, "unrecognized compression value %s",
771 					  args[0].from);
772 				ret = -EINVAL;
773 				goto out;
774 			}
775 
776 			if (compress_force) {
777 				btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
778 			} else {
779 				/*
780 				 * If we remount from compress-force=xxx to
781 				 * compress=xxx, we need clear FORCE_COMPRESS
782 				 * flag, otherwise, there is no way for users
783 				 * to disable forcible compression separately.
784 				 */
785 				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
786 			}
787 			if (no_compress == 1) {
788 				btrfs_info(info, "use no compression");
789 			} else if ((info->compress_type != saved_compress_type) ||
790 				   (compress_force != saved_compress_force) ||
791 				   (info->compress_level != saved_compress_level)) {
792 				btrfs_info(info, "%s %s compression, level %d",
793 					   (compress_force) ? "force" : "use",
794 					   compress_type, info->compress_level);
795 			}
796 			compress_force = false;
797 			break;
798 		case Opt_ssd:
799 			btrfs_set_and_info(info, SSD,
800 					   "enabling ssd optimizations");
801 			btrfs_clear_opt(info->mount_opt, NOSSD);
802 			break;
803 		case Opt_ssd_spread:
804 			btrfs_set_and_info(info, SSD,
805 					   "enabling ssd optimizations");
806 			btrfs_set_and_info(info, SSD_SPREAD,
807 					   "using spread ssd allocation scheme");
808 			btrfs_clear_opt(info->mount_opt, NOSSD);
809 			break;
810 		case Opt_nossd:
811 			btrfs_set_opt(info->mount_opt, NOSSD);
812 			btrfs_clear_and_info(info, SSD,
813 					     "not using ssd optimizations");
814 			fallthrough;
815 		case Opt_nossd_spread:
816 			btrfs_clear_and_info(info, SSD_SPREAD,
817 					     "not using spread ssd allocation scheme");
818 			break;
819 		case Opt_barrier:
820 			btrfs_clear_and_info(info, NOBARRIER,
821 					     "turning on barriers");
822 			break;
823 		case Opt_nobarrier:
824 			btrfs_set_and_info(info, NOBARRIER,
825 					   "turning off barriers");
826 			break;
827 		case Opt_thread_pool:
828 			ret = match_int(&args[0], &intarg);
829 			if (ret) {
830 				btrfs_err(info, "unrecognized thread_pool value %s",
831 					  args[0].from);
832 				goto out;
833 			} else if (intarg == 0) {
834 				btrfs_err(info, "invalid value 0 for thread_pool");
835 				ret = -EINVAL;
836 				goto out;
837 			}
838 			info->thread_pool_size = intarg;
839 			break;
840 		case Opt_max_inline:
841 			num = match_strdup(&args[0]);
842 			if (num) {
843 				info->max_inline = memparse(num, NULL);
844 				kfree(num);
845 
846 				if (info->max_inline) {
847 					info->max_inline = min_t(u64,
848 						info->max_inline,
849 						info->sectorsize);
850 				}
851 				btrfs_info(info, "max_inline at %llu",
852 					   info->max_inline);
853 			} else {
854 				ret = -ENOMEM;
855 				goto out;
856 			}
857 			break;
858 		case Opt_acl:
859 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
860 			info->sb->s_flags |= SB_POSIXACL;
861 			break;
862 #else
863 			btrfs_err(info, "support for ACL not compiled in!");
864 			ret = -EINVAL;
865 			goto out;
866 #endif
867 		case Opt_noacl:
868 			info->sb->s_flags &= ~SB_POSIXACL;
869 			break;
870 		case Opt_notreelog:
871 			btrfs_set_and_info(info, NOTREELOG,
872 					   "disabling tree log");
873 			break;
874 		case Opt_treelog:
875 			btrfs_clear_and_info(info, NOTREELOG,
876 					     "enabling tree log");
877 			break;
878 		case Opt_norecovery:
879 		case Opt_nologreplay:
880 			btrfs_warn(info,
881 		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
882 			btrfs_set_and_info(info, NOLOGREPLAY,
883 					   "disabling log replay at mount time");
884 			break;
885 		case Opt_flushoncommit:
886 			btrfs_set_and_info(info, FLUSHONCOMMIT,
887 					   "turning on flush-on-commit");
888 			break;
889 		case Opt_noflushoncommit:
890 			btrfs_clear_and_info(info, FLUSHONCOMMIT,
891 					     "turning off flush-on-commit");
892 			break;
893 		case Opt_ratio:
894 			ret = match_int(&args[0], &intarg);
895 			if (ret) {
896 				btrfs_err(info, "unrecognized metadata_ratio value %s",
897 					  args[0].from);
898 				goto out;
899 			}
900 			info->metadata_ratio = intarg;
901 			btrfs_info(info, "metadata ratio %u",
902 				   info->metadata_ratio);
903 			break;
904 		case Opt_discard:
905 		case Opt_discard_mode:
906 			if (token == Opt_discard ||
907 			    strcmp(args[0].from, "sync") == 0) {
908 				btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
909 				btrfs_set_and_info(info, DISCARD_SYNC,
910 						   "turning on sync discard");
911 			} else if (strcmp(args[0].from, "async") == 0) {
912 				btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
913 				btrfs_set_and_info(info, DISCARD_ASYNC,
914 						   "turning on async discard");
915 			} else {
916 				btrfs_err(info, "unrecognized discard mode value %s",
917 					  args[0].from);
918 				ret = -EINVAL;
919 				goto out;
920 			}
921 			btrfs_clear_opt(info->mount_opt, NODISCARD);
922 			break;
923 		case Opt_nodiscard:
924 			btrfs_clear_and_info(info, DISCARD_SYNC,
925 					     "turning off discard");
926 			btrfs_clear_and_info(info, DISCARD_ASYNC,
927 					     "turning off async discard");
928 			btrfs_set_opt(info->mount_opt, NODISCARD);
929 			break;
930 		case Opt_space_cache:
931 		case Opt_space_cache_version:
932 			/*
933 			 * We already set FREE_SPACE_TREE above because we have
934 			 * compat_ro(FREE_SPACE_TREE) set, and we aren't going
935 			 * to allow v1 to be set for extent tree v2, simply
936 			 * ignore this setting if we're extent tree v2.
937 			 */
938 			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
939 				break;
940 			if (token == Opt_space_cache ||
941 			    strcmp(args[0].from, "v1") == 0) {
942 				btrfs_clear_opt(info->mount_opt,
943 						FREE_SPACE_TREE);
944 				btrfs_set_and_info(info, SPACE_CACHE,
945 					   "enabling disk space caching");
946 			} else if (strcmp(args[0].from, "v2") == 0) {
947 				btrfs_clear_opt(info->mount_opt,
948 						SPACE_CACHE);
949 				btrfs_set_and_info(info, FREE_SPACE_TREE,
950 						   "enabling free space tree");
951 			} else {
952 				btrfs_err(info, "unrecognized space_cache value %s",
953 					  args[0].from);
954 				ret = -EINVAL;
955 				goto out;
956 			}
957 			break;
958 		case Opt_rescan_uuid_tree:
959 			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
960 			break;
961 		case Opt_no_space_cache:
962 			/*
963 			 * We cannot operate without the free space tree with
964 			 * extent tree v2, ignore this option.
965 			 */
966 			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
967 				break;
968 			if (btrfs_test_opt(info, SPACE_CACHE)) {
969 				btrfs_clear_and_info(info, SPACE_CACHE,
970 					     "disabling disk space caching");
971 			}
972 			if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
973 				btrfs_clear_and_info(info, FREE_SPACE_TREE,
974 					     "disabling free space tree");
975 			}
976 			break;
977 		case Opt_inode_cache:
978 		case Opt_noinode_cache:
979 			btrfs_warn(info,
980 	"the 'inode_cache' option is deprecated and has no effect since 5.11");
981 			break;
982 		case Opt_clear_cache:
983 			/*
984 			 * We cannot clear the free space tree with extent tree
985 			 * v2, ignore this option.
986 			 */
987 			if (btrfs_fs_incompat(info, EXTENT_TREE_V2))
988 				break;
989 			btrfs_set_and_info(info, CLEAR_CACHE,
990 					   "force clearing of disk cache");
991 			break;
992 		case Opt_user_subvol_rm_allowed:
993 			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
994 			break;
995 		case Opt_enospc_debug:
996 			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
997 			break;
998 		case Opt_noenospc_debug:
999 			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
1000 			break;
1001 		case Opt_defrag:
1002 			btrfs_set_and_info(info, AUTO_DEFRAG,
1003 					   "enabling auto defrag");
1004 			break;
1005 		case Opt_nodefrag:
1006 			btrfs_clear_and_info(info, AUTO_DEFRAG,
1007 					     "disabling auto defrag");
1008 			break;
1009 		case Opt_recovery:
1010 		case Opt_usebackuproot:
1011 			btrfs_warn(info,
1012 			"'%s' is deprecated, use 'rescue=usebackuproot' instead",
1013 				   token == Opt_recovery ? "recovery" :
1014 				   "usebackuproot");
1015 			btrfs_info(info,
1016 				   "trying to use backup root at mount time");
1017 			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
1018 			break;
1019 		case Opt_skip_balance:
1020 			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
1021 			break;
1022 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1023 		case Opt_check_integrity_including_extent_data:
1024 			btrfs_info(info,
1025 				   "enabling check integrity including extent data");
1026 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
1027 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
1028 			break;
1029 		case Opt_check_integrity:
1030 			btrfs_info(info, "enabling check integrity");
1031 			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
1032 			break;
1033 		case Opt_check_integrity_print_mask:
1034 			ret = match_int(&args[0], &intarg);
1035 			if (ret) {
1036 				btrfs_err(info,
1037 				"unrecognized check_integrity_print_mask value %s",
1038 					args[0].from);
1039 				goto out;
1040 			}
1041 			info->check_integrity_print_mask = intarg;
1042 			btrfs_info(info, "check_integrity_print_mask 0x%x",
1043 				   info->check_integrity_print_mask);
1044 			break;
1045 #else
1046 		case Opt_check_integrity_including_extent_data:
1047 		case Opt_check_integrity:
1048 		case Opt_check_integrity_print_mask:
1049 			btrfs_err(info,
1050 				  "support for check_integrity* not compiled in!");
1051 			ret = -EINVAL;
1052 			goto out;
1053 #endif
1054 		case Opt_fatal_errors:
1055 			if (strcmp(args[0].from, "panic") == 0) {
1056 				btrfs_set_opt(info->mount_opt,
1057 					      PANIC_ON_FATAL_ERROR);
1058 			} else if (strcmp(args[0].from, "bug") == 0) {
1059 				btrfs_clear_opt(info->mount_opt,
1060 					      PANIC_ON_FATAL_ERROR);
1061 			} else {
1062 				btrfs_err(info, "unrecognized fatal_errors value %s",
1063 					  args[0].from);
1064 				ret = -EINVAL;
1065 				goto out;
1066 			}
1067 			break;
1068 		case Opt_commit_interval:
1069 			intarg = 0;
1070 			ret = match_int(&args[0], &intarg);
1071 			if (ret) {
1072 				btrfs_err(info, "unrecognized commit_interval value %s",
1073 					  args[0].from);
1074 				ret = -EINVAL;
1075 				goto out;
1076 			}
1077 			if (intarg == 0) {
1078 				btrfs_info(info,
1079 					   "using default commit interval %us",
1080 					   BTRFS_DEFAULT_COMMIT_INTERVAL);
1081 				intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
1082 			} else if (intarg > 300) {
1083 				btrfs_warn(info, "excessive commit interval %d",
1084 					   intarg);
1085 			}
1086 			info->commit_interval = intarg;
1087 			break;
1088 		case Opt_rescue:
1089 			ret = parse_rescue_options(info, args[0].from);
1090 			if (ret < 0) {
1091 				btrfs_err(info, "unrecognized rescue value %s",
1092 					  args[0].from);
1093 				goto out;
1094 			}
1095 			break;
1096 #ifdef CONFIG_BTRFS_DEBUG
1097 		case Opt_fragment_all:
1098 			btrfs_info(info, "fragmenting all space");
1099 			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1100 			btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
1101 			break;
1102 		case Opt_fragment_metadata:
1103 			btrfs_info(info, "fragmenting metadata");
1104 			btrfs_set_opt(info->mount_opt,
1105 				      FRAGMENT_METADATA);
1106 			break;
1107 		case Opt_fragment_data:
1108 			btrfs_info(info, "fragmenting data");
1109 			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1110 			break;
1111 #endif
1112 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
1113 		case Opt_ref_verify:
1114 			btrfs_info(info, "doing ref verification");
1115 			btrfs_set_opt(info->mount_opt, REF_VERIFY);
1116 			break;
1117 #endif
1118 		case Opt_err:
1119 			btrfs_err(info, "unrecognized mount option '%s'", p);
1120 			ret = -EINVAL;
1121 			goto out;
1122 		default:
1123 			break;
1124 		}
1125 	}
1126 check:
1127 	/* We're read-only, don't have to check. */
1128 	if (new_flags & SB_RDONLY)
1129 		goto out;
1130 
1131 	if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1132 	    check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1133 	    check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1134 		ret = -EINVAL;
1135 out:
1136 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1137 	    !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1138 	    !btrfs_test_opt(info, CLEAR_CACHE)) {
1139 		btrfs_err(info, "cannot disable free space tree");
1140 		ret = -EINVAL;
1141 
1142 	}
1143 	if (!ret)
1144 		ret = btrfs_check_mountopts_zoned(info);
1145 	if (!ret && !remounting) {
1146 		if (btrfs_test_opt(info, SPACE_CACHE))
1147 			btrfs_info(info, "disk space caching is enabled");
1148 		if (btrfs_test_opt(info, FREE_SPACE_TREE))
1149 			btrfs_info(info, "using free space tree");
1150 	}
1151 	return ret;
1152 }
1153 
1154 /*
1155  * Parse mount options that are required early in the mount process.
1156  *
1157  * All other options will be parsed on much later in the mount process and
1158  * only when we need to allocate a new super block.
1159  */
1160 static int btrfs_parse_device_options(const char *options, fmode_t flags,
1161 				      void *holder)
1162 {
1163 	substring_t args[MAX_OPT_ARGS];
1164 	char *device_name, *opts, *orig, *p;
1165 	struct btrfs_device *device = NULL;
1166 	int error = 0;
1167 
1168 	lockdep_assert_held(&uuid_mutex);
1169 
1170 	if (!options)
1171 		return 0;
1172 
1173 	/*
1174 	 * strsep changes the string, duplicate it because btrfs_parse_options
1175 	 * gets called later
1176 	 */
1177 	opts = kstrdup(options, GFP_KERNEL);
1178 	if (!opts)
1179 		return -ENOMEM;
1180 	orig = opts;
1181 
1182 	while ((p = strsep(&opts, ",")) != NULL) {
1183 		int token;
1184 
1185 		if (!*p)
1186 			continue;
1187 
1188 		token = match_token(p, tokens, args);
1189 		if (token == Opt_device) {
1190 			device_name = match_strdup(&args[0]);
1191 			if (!device_name) {
1192 				error = -ENOMEM;
1193 				goto out;
1194 			}
1195 			device = btrfs_scan_one_device(device_name, flags,
1196 					holder);
1197 			kfree(device_name);
1198 			if (IS_ERR(device)) {
1199 				error = PTR_ERR(device);
1200 				goto out;
1201 			}
1202 		}
1203 	}
1204 
1205 out:
1206 	kfree(orig);
1207 	return error;
1208 }
1209 
1210 /*
1211  * Parse mount options that are related to subvolume id
1212  *
1213  * The value is later passed to mount_subvol()
1214  */
1215 static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1216 		u64 *subvol_objectid)
1217 {
1218 	substring_t args[MAX_OPT_ARGS];
1219 	char *opts, *orig, *p;
1220 	int error = 0;
1221 	u64 subvolid;
1222 
1223 	if (!options)
1224 		return 0;
1225 
1226 	/*
1227 	 * strsep changes the string, duplicate it because
1228 	 * btrfs_parse_device_options gets called later
1229 	 */
1230 	opts = kstrdup(options, GFP_KERNEL);
1231 	if (!opts)
1232 		return -ENOMEM;
1233 	orig = opts;
1234 
1235 	while ((p = strsep(&opts, ",")) != NULL) {
1236 		int token;
1237 		if (!*p)
1238 			continue;
1239 
1240 		token = match_token(p, tokens, args);
1241 		switch (token) {
1242 		case Opt_subvol:
1243 			kfree(*subvol_name);
1244 			*subvol_name = match_strdup(&args[0]);
1245 			if (!*subvol_name) {
1246 				error = -ENOMEM;
1247 				goto out;
1248 			}
1249 			break;
1250 		case Opt_subvolid:
1251 			error = match_u64(&args[0], &subvolid);
1252 			if (error)
1253 				goto out;
1254 
1255 			/* we want the original fs_tree */
1256 			if (subvolid == 0)
1257 				subvolid = BTRFS_FS_TREE_OBJECTID;
1258 
1259 			*subvol_objectid = subvolid;
1260 			break;
1261 		default:
1262 			break;
1263 		}
1264 	}
1265 
1266 out:
1267 	kfree(orig);
1268 	return error;
1269 }
1270 
1271 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1272 					  u64 subvol_objectid)
1273 {
1274 	struct btrfs_root *root = fs_info->tree_root;
1275 	struct btrfs_root *fs_root = NULL;
1276 	struct btrfs_root_ref *root_ref;
1277 	struct btrfs_inode_ref *inode_ref;
1278 	struct btrfs_key key;
1279 	struct btrfs_path *path = NULL;
1280 	char *name = NULL, *ptr;
1281 	u64 dirid;
1282 	int len;
1283 	int ret;
1284 
1285 	path = btrfs_alloc_path();
1286 	if (!path) {
1287 		ret = -ENOMEM;
1288 		goto err;
1289 	}
1290 
1291 	name = kmalloc(PATH_MAX, GFP_KERNEL);
1292 	if (!name) {
1293 		ret = -ENOMEM;
1294 		goto err;
1295 	}
1296 	ptr = name + PATH_MAX - 1;
1297 	ptr[0] = '\0';
1298 
1299 	/*
1300 	 * Walk up the subvolume trees in the tree of tree roots by root
1301 	 * backrefs until we hit the top-level subvolume.
1302 	 */
1303 	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1304 		key.objectid = subvol_objectid;
1305 		key.type = BTRFS_ROOT_BACKREF_KEY;
1306 		key.offset = (u64)-1;
1307 
1308 		ret = btrfs_search_backwards(root, &key, path);
1309 		if (ret < 0) {
1310 			goto err;
1311 		} else if (ret > 0) {
1312 			ret = -ENOENT;
1313 			goto err;
1314 		}
1315 
1316 		subvol_objectid = key.offset;
1317 
1318 		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1319 					  struct btrfs_root_ref);
1320 		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1321 		ptr -= len + 1;
1322 		if (ptr < name) {
1323 			ret = -ENAMETOOLONG;
1324 			goto err;
1325 		}
1326 		read_extent_buffer(path->nodes[0], ptr + 1,
1327 				   (unsigned long)(root_ref + 1), len);
1328 		ptr[0] = '/';
1329 		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1330 		btrfs_release_path(path);
1331 
1332 		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1333 		if (IS_ERR(fs_root)) {
1334 			ret = PTR_ERR(fs_root);
1335 			fs_root = NULL;
1336 			goto err;
1337 		}
1338 
1339 		/*
1340 		 * Walk up the filesystem tree by inode refs until we hit the
1341 		 * root directory.
1342 		 */
1343 		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1344 			key.objectid = dirid;
1345 			key.type = BTRFS_INODE_REF_KEY;
1346 			key.offset = (u64)-1;
1347 
1348 			ret = btrfs_search_backwards(fs_root, &key, path);
1349 			if (ret < 0) {
1350 				goto err;
1351 			} else if (ret > 0) {
1352 				ret = -ENOENT;
1353 				goto err;
1354 			}
1355 
1356 			dirid = key.offset;
1357 
1358 			inode_ref = btrfs_item_ptr(path->nodes[0],
1359 						   path->slots[0],
1360 						   struct btrfs_inode_ref);
1361 			len = btrfs_inode_ref_name_len(path->nodes[0],
1362 						       inode_ref);
1363 			ptr -= len + 1;
1364 			if (ptr < name) {
1365 				ret = -ENAMETOOLONG;
1366 				goto err;
1367 			}
1368 			read_extent_buffer(path->nodes[0], ptr + 1,
1369 					   (unsigned long)(inode_ref + 1), len);
1370 			ptr[0] = '/';
1371 			btrfs_release_path(path);
1372 		}
1373 		btrfs_put_root(fs_root);
1374 		fs_root = NULL;
1375 	}
1376 
1377 	btrfs_free_path(path);
1378 	if (ptr == name + PATH_MAX - 1) {
1379 		name[0] = '/';
1380 		name[1] = '\0';
1381 	} else {
1382 		memmove(name, ptr, name + PATH_MAX - ptr);
1383 	}
1384 	return name;
1385 
1386 err:
1387 	btrfs_put_root(fs_root);
1388 	btrfs_free_path(path);
1389 	kfree(name);
1390 	return ERR_PTR(ret);
1391 }
1392 
1393 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1394 {
1395 	struct btrfs_root *root = fs_info->tree_root;
1396 	struct btrfs_dir_item *di;
1397 	struct btrfs_path *path;
1398 	struct btrfs_key location;
1399 	u64 dir_id;
1400 
1401 	path = btrfs_alloc_path();
1402 	if (!path)
1403 		return -ENOMEM;
1404 
1405 	/*
1406 	 * Find the "default" dir item which points to the root item that we
1407 	 * will mount by default if we haven't been given a specific subvolume
1408 	 * to mount.
1409 	 */
1410 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
1411 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1412 	if (IS_ERR(di)) {
1413 		btrfs_free_path(path);
1414 		return PTR_ERR(di);
1415 	}
1416 	if (!di) {
1417 		/*
1418 		 * Ok the default dir item isn't there.  This is weird since
1419 		 * it's always been there, but don't freak out, just try and
1420 		 * mount the top-level subvolume.
1421 		 */
1422 		btrfs_free_path(path);
1423 		*objectid = BTRFS_FS_TREE_OBJECTID;
1424 		return 0;
1425 	}
1426 
1427 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1428 	btrfs_free_path(path);
1429 	*objectid = location.objectid;
1430 	return 0;
1431 }
1432 
1433 static int btrfs_fill_super(struct super_block *sb,
1434 			    struct btrfs_fs_devices *fs_devices,
1435 			    void *data)
1436 {
1437 	struct inode *inode;
1438 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1439 	int err;
1440 
1441 	sb->s_maxbytes = MAX_LFS_FILESIZE;
1442 	sb->s_magic = BTRFS_SUPER_MAGIC;
1443 	sb->s_op = &btrfs_super_ops;
1444 	sb->s_d_op = &btrfs_dentry_operations;
1445 	sb->s_export_op = &btrfs_export_ops;
1446 #ifdef CONFIG_FS_VERITY
1447 	sb->s_vop = &btrfs_verityops;
1448 #endif
1449 	sb->s_xattr = btrfs_xattr_handlers;
1450 	sb->s_time_gran = 1;
1451 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
1452 	sb->s_flags |= SB_POSIXACL;
1453 #endif
1454 	sb->s_flags |= SB_I_VERSION;
1455 	sb->s_iflags |= SB_I_CGROUPWB;
1456 
1457 	err = super_setup_bdi(sb);
1458 	if (err) {
1459 		btrfs_err(fs_info, "super_setup_bdi failed");
1460 		return err;
1461 	}
1462 
1463 	err = open_ctree(sb, fs_devices, (char *)data);
1464 	if (err) {
1465 		btrfs_err(fs_info, "open_ctree failed");
1466 		return err;
1467 	}
1468 
1469 	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1470 	if (IS_ERR(inode)) {
1471 		err = PTR_ERR(inode);
1472 		goto fail_close;
1473 	}
1474 
1475 	sb->s_root = d_make_root(inode);
1476 	if (!sb->s_root) {
1477 		err = -ENOMEM;
1478 		goto fail_close;
1479 	}
1480 
1481 	sb->s_flags |= SB_ACTIVE;
1482 	return 0;
1483 
1484 fail_close:
1485 	close_ctree(fs_info);
1486 	return err;
1487 }
1488 
1489 int btrfs_sync_fs(struct super_block *sb, int wait)
1490 {
1491 	struct btrfs_trans_handle *trans;
1492 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1493 	struct btrfs_root *root = fs_info->tree_root;
1494 
1495 	trace_btrfs_sync_fs(fs_info, wait);
1496 
1497 	if (!wait) {
1498 		filemap_flush(fs_info->btree_inode->i_mapping);
1499 		return 0;
1500 	}
1501 
1502 	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1503 
1504 	trans = btrfs_attach_transaction_barrier(root);
1505 	if (IS_ERR(trans)) {
1506 		/* no transaction, don't bother */
1507 		if (PTR_ERR(trans) == -ENOENT) {
1508 			/*
1509 			 * Exit unless we have some pending changes
1510 			 * that need to go through commit
1511 			 */
1512 			if (fs_info->pending_changes == 0)
1513 				return 0;
1514 			/*
1515 			 * A non-blocking test if the fs is frozen. We must not
1516 			 * start a new transaction here otherwise a deadlock
1517 			 * happens. The pending operations are delayed to the
1518 			 * next commit after thawing.
1519 			 */
1520 			if (sb_start_write_trylock(sb))
1521 				sb_end_write(sb);
1522 			else
1523 				return 0;
1524 			trans = btrfs_start_transaction(root, 0);
1525 		}
1526 		if (IS_ERR(trans))
1527 			return PTR_ERR(trans);
1528 	}
1529 	return btrfs_commit_transaction(trans);
1530 }
1531 
1532 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1533 {
1534 	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1535 	*printed = true;
1536 }
1537 
1538 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1539 {
1540 	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1541 	const char *compress_type;
1542 	const char *subvol_name;
1543 	bool printed = false;
1544 
1545 	if (btrfs_test_opt(info, DEGRADED))
1546 		seq_puts(seq, ",degraded");
1547 	if (btrfs_test_opt(info, NODATASUM))
1548 		seq_puts(seq, ",nodatasum");
1549 	if (btrfs_test_opt(info, NODATACOW))
1550 		seq_puts(seq, ",nodatacow");
1551 	if (btrfs_test_opt(info, NOBARRIER))
1552 		seq_puts(seq, ",nobarrier");
1553 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1554 		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1555 	if (info->thread_pool_size !=  min_t(unsigned long,
1556 					     num_online_cpus() + 2, 8))
1557 		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1558 	if (btrfs_test_opt(info, COMPRESS)) {
1559 		compress_type = btrfs_compress_type2str(info->compress_type);
1560 		if (btrfs_test_opt(info, FORCE_COMPRESS))
1561 			seq_printf(seq, ",compress-force=%s", compress_type);
1562 		else
1563 			seq_printf(seq, ",compress=%s", compress_type);
1564 		if (info->compress_level)
1565 			seq_printf(seq, ":%d", info->compress_level);
1566 	}
1567 	if (btrfs_test_opt(info, NOSSD))
1568 		seq_puts(seq, ",nossd");
1569 	if (btrfs_test_opt(info, SSD_SPREAD))
1570 		seq_puts(seq, ",ssd_spread");
1571 	else if (btrfs_test_opt(info, SSD))
1572 		seq_puts(seq, ",ssd");
1573 	if (btrfs_test_opt(info, NOTREELOG))
1574 		seq_puts(seq, ",notreelog");
1575 	if (btrfs_test_opt(info, NOLOGREPLAY))
1576 		print_rescue_option(seq, "nologreplay", &printed);
1577 	if (btrfs_test_opt(info, USEBACKUPROOT))
1578 		print_rescue_option(seq, "usebackuproot", &printed);
1579 	if (btrfs_test_opt(info, IGNOREBADROOTS))
1580 		print_rescue_option(seq, "ignorebadroots", &printed);
1581 	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1582 		print_rescue_option(seq, "ignoredatacsums", &printed);
1583 	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1584 		seq_puts(seq, ",flushoncommit");
1585 	if (btrfs_test_opt(info, DISCARD_SYNC))
1586 		seq_puts(seq, ",discard");
1587 	if (btrfs_test_opt(info, DISCARD_ASYNC))
1588 		seq_puts(seq, ",discard=async");
1589 	if (!(info->sb->s_flags & SB_POSIXACL))
1590 		seq_puts(seq, ",noacl");
1591 	if (btrfs_free_space_cache_v1_active(info))
1592 		seq_puts(seq, ",space_cache");
1593 	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1594 		seq_puts(seq, ",space_cache=v2");
1595 	else
1596 		seq_puts(seq, ",nospace_cache");
1597 	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1598 		seq_puts(seq, ",rescan_uuid_tree");
1599 	if (btrfs_test_opt(info, CLEAR_CACHE))
1600 		seq_puts(seq, ",clear_cache");
1601 	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1602 		seq_puts(seq, ",user_subvol_rm_allowed");
1603 	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1604 		seq_puts(seq, ",enospc_debug");
1605 	if (btrfs_test_opt(info, AUTO_DEFRAG))
1606 		seq_puts(seq, ",autodefrag");
1607 	if (btrfs_test_opt(info, SKIP_BALANCE))
1608 		seq_puts(seq, ",skip_balance");
1609 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1610 	if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1611 		seq_puts(seq, ",check_int_data");
1612 	else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1613 		seq_puts(seq, ",check_int");
1614 	if (info->check_integrity_print_mask)
1615 		seq_printf(seq, ",check_int_print_mask=%d",
1616 				info->check_integrity_print_mask);
1617 #endif
1618 	if (info->metadata_ratio)
1619 		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1620 	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1621 		seq_puts(seq, ",fatal_errors=panic");
1622 	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1623 		seq_printf(seq, ",commit=%u", info->commit_interval);
1624 #ifdef CONFIG_BTRFS_DEBUG
1625 	if (btrfs_test_opt(info, FRAGMENT_DATA))
1626 		seq_puts(seq, ",fragment=data");
1627 	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1628 		seq_puts(seq, ",fragment=metadata");
1629 #endif
1630 	if (btrfs_test_opt(info, REF_VERIFY))
1631 		seq_puts(seq, ",ref_verify");
1632 	seq_printf(seq, ",subvolid=%llu",
1633 		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1634 	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1635 			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1636 	if (!IS_ERR(subvol_name)) {
1637 		seq_puts(seq, ",subvol=");
1638 		seq_escape(seq, subvol_name, " \t\n\\");
1639 		kfree(subvol_name);
1640 	}
1641 	return 0;
1642 }
1643 
1644 static int btrfs_test_super(struct super_block *s, void *data)
1645 {
1646 	struct btrfs_fs_info *p = data;
1647 	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1648 
1649 	return fs_info->fs_devices == p->fs_devices;
1650 }
1651 
1652 static int btrfs_set_super(struct super_block *s, void *data)
1653 {
1654 	int err = set_anon_super(s, data);
1655 	if (!err)
1656 		s->s_fs_info = data;
1657 	return err;
1658 }
1659 
1660 /*
1661  * subvolumes are identified by ino 256
1662  */
1663 static inline int is_subvolume_inode(struct inode *inode)
1664 {
1665 	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1666 		return 1;
1667 	return 0;
1668 }
1669 
1670 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1671 				   struct vfsmount *mnt)
1672 {
1673 	struct dentry *root;
1674 	int ret;
1675 
1676 	if (!subvol_name) {
1677 		if (!subvol_objectid) {
1678 			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1679 							  &subvol_objectid);
1680 			if (ret) {
1681 				root = ERR_PTR(ret);
1682 				goto out;
1683 			}
1684 		}
1685 		subvol_name = btrfs_get_subvol_name_from_objectid(
1686 					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1687 		if (IS_ERR(subvol_name)) {
1688 			root = ERR_CAST(subvol_name);
1689 			subvol_name = NULL;
1690 			goto out;
1691 		}
1692 
1693 	}
1694 
1695 	root = mount_subtree(mnt, subvol_name);
1696 	/* mount_subtree() drops our reference on the vfsmount. */
1697 	mnt = NULL;
1698 
1699 	if (!IS_ERR(root)) {
1700 		struct super_block *s = root->d_sb;
1701 		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1702 		struct inode *root_inode = d_inode(root);
1703 		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1704 
1705 		ret = 0;
1706 		if (!is_subvolume_inode(root_inode)) {
1707 			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1708 			       subvol_name);
1709 			ret = -EINVAL;
1710 		}
1711 		if (subvol_objectid && root_objectid != subvol_objectid) {
1712 			/*
1713 			 * This will also catch a race condition where a
1714 			 * subvolume which was passed by ID is renamed and
1715 			 * another subvolume is renamed over the old location.
1716 			 */
1717 			btrfs_err(fs_info,
1718 				  "subvol '%s' does not match subvolid %llu",
1719 				  subvol_name, subvol_objectid);
1720 			ret = -EINVAL;
1721 		}
1722 		if (ret) {
1723 			dput(root);
1724 			root = ERR_PTR(ret);
1725 			deactivate_locked_super(s);
1726 		}
1727 	}
1728 
1729 out:
1730 	mntput(mnt);
1731 	kfree(subvol_name);
1732 	return root;
1733 }
1734 
1735 /*
1736  * Find a superblock for the given device / mount point.
1737  *
1738  * Note: This is based on mount_bdev from fs/super.c with a few additions
1739  *       for multiple device setup.  Make sure to keep it in sync.
1740  */
1741 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1742 		int flags, const char *device_name, void *data)
1743 {
1744 	struct block_device *bdev = NULL;
1745 	struct super_block *s;
1746 	struct btrfs_device *device = NULL;
1747 	struct btrfs_fs_devices *fs_devices = NULL;
1748 	struct btrfs_fs_info *fs_info = NULL;
1749 	void *new_sec_opts = NULL;
1750 	fmode_t mode = FMODE_READ;
1751 	int error = 0;
1752 
1753 	if (!(flags & SB_RDONLY))
1754 		mode |= FMODE_WRITE;
1755 
1756 	if (data) {
1757 		error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1758 		if (error)
1759 			return ERR_PTR(error);
1760 	}
1761 
1762 	/*
1763 	 * Setup a dummy root and fs_info for test/set super.  This is because
1764 	 * we don't actually fill this stuff out until open_ctree, but we need
1765 	 * then open_ctree will properly initialize the file system specific
1766 	 * settings later.  btrfs_init_fs_info initializes the static elements
1767 	 * of the fs_info (locks and such) to make cleanup easier if we find a
1768 	 * superblock with our given fs_devices later on at sget() time.
1769 	 */
1770 	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1771 	if (!fs_info) {
1772 		error = -ENOMEM;
1773 		goto error_sec_opts;
1774 	}
1775 	btrfs_init_fs_info(fs_info);
1776 
1777 	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1778 	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1779 	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1780 		error = -ENOMEM;
1781 		goto error_fs_info;
1782 	}
1783 
1784 	mutex_lock(&uuid_mutex);
1785 	error = btrfs_parse_device_options(data, mode, fs_type);
1786 	if (error) {
1787 		mutex_unlock(&uuid_mutex);
1788 		goto error_fs_info;
1789 	}
1790 
1791 	device = btrfs_scan_one_device(device_name, mode, fs_type);
1792 	if (IS_ERR(device)) {
1793 		mutex_unlock(&uuid_mutex);
1794 		error = PTR_ERR(device);
1795 		goto error_fs_info;
1796 	}
1797 
1798 	fs_devices = device->fs_devices;
1799 	fs_info->fs_devices = fs_devices;
1800 
1801 	error = btrfs_open_devices(fs_devices, mode, fs_type);
1802 	mutex_unlock(&uuid_mutex);
1803 	if (error)
1804 		goto error_fs_info;
1805 
1806 	if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1807 		error = -EACCES;
1808 		goto error_close_devices;
1809 	}
1810 
1811 	bdev = fs_devices->latest_dev->bdev;
1812 	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1813 		 fs_info);
1814 	if (IS_ERR(s)) {
1815 		error = PTR_ERR(s);
1816 		goto error_close_devices;
1817 	}
1818 
1819 	if (s->s_root) {
1820 		btrfs_close_devices(fs_devices);
1821 		btrfs_free_fs_info(fs_info);
1822 		if ((flags ^ s->s_flags) & SB_RDONLY)
1823 			error = -EBUSY;
1824 	} else {
1825 		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1826 		shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s", fs_type->name,
1827 					s->s_id);
1828 		btrfs_sb(s)->bdev_holder = fs_type;
1829 		if (!strstr(crc32c_impl(), "generic"))
1830 			set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1831 		error = btrfs_fill_super(s, fs_devices, data);
1832 	}
1833 	if (!error)
1834 		error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1835 	security_free_mnt_opts(&new_sec_opts);
1836 	if (error) {
1837 		deactivate_locked_super(s);
1838 		return ERR_PTR(error);
1839 	}
1840 
1841 	return dget(s->s_root);
1842 
1843 error_close_devices:
1844 	btrfs_close_devices(fs_devices);
1845 error_fs_info:
1846 	btrfs_free_fs_info(fs_info);
1847 error_sec_opts:
1848 	security_free_mnt_opts(&new_sec_opts);
1849 	return ERR_PTR(error);
1850 }
1851 
1852 /*
1853  * Mount function which is called by VFS layer.
1854  *
1855  * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1856  * which needs vfsmount* of device's root (/).  This means device's root has to
1857  * be mounted internally in any case.
1858  *
1859  * Operation flow:
1860  *   1. Parse subvol id related options for later use in mount_subvol().
1861  *
1862  *   2. Mount device's root (/) by calling vfs_kern_mount().
1863  *
1864  *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1865  *      first place. In order to avoid calling btrfs_mount() again, we use
1866  *      different file_system_type which is not registered to VFS by
1867  *      register_filesystem() (btrfs_root_fs_type). As a result,
1868  *      btrfs_mount_root() is called. The return value will be used by
1869  *      mount_subtree() in mount_subvol().
1870  *
1871  *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1872  *      "btrfs subvolume set-default", mount_subvol() is called always.
1873  */
1874 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1875 		const char *device_name, void *data)
1876 {
1877 	struct vfsmount *mnt_root;
1878 	struct dentry *root;
1879 	char *subvol_name = NULL;
1880 	u64 subvol_objectid = 0;
1881 	int error = 0;
1882 
1883 	error = btrfs_parse_subvol_options(data, &subvol_name,
1884 					&subvol_objectid);
1885 	if (error) {
1886 		kfree(subvol_name);
1887 		return ERR_PTR(error);
1888 	}
1889 
1890 	/* mount device's root (/) */
1891 	mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1892 	if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1893 		if (flags & SB_RDONLY) {
1894 			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1895 				flags & ~SB_RDONLY, device_name, data);
1896 		} else {
1897 			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1898 				flags | SB_RDONLY, device_name, data);
1899 			if (IS_ERR(mnt_root)) {
1900 				root = ERR_CAST(mnt_root);
1901 				kfree(subvol_name);
1902 				goto out;
1903 			}
1904 
1905 			down_write(&mnt_root->mnt_sb->s_umount);
1906 			error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1907 			up_write(&mnt_root->mnt_sb->s_umount);
1908 			if (error < 0) {
1909 				root = ERR_PTR(error);
1910 				mntput(mnt_root);
1911 				kfree(subvol_name);
1912 				goto out;
1913 			}
1914 		}
1915 	}
1916 	if (IS_ERR(mnt_root)) {
1917 		root = ERR_CAST(mnt_root);
1918 		kfree(subvol_name);
1919 		goto out;
1920 	}
1921 
1922 	/* mount_subvol() will free subvol_name and mnt_root */
1923 	root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1924 
1925 out:
1926 	return root;
1927 }
1928 
1929 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1930 				     u32 new_pool_size, u32 old_pool_size)
1931 {
1932 	if (new_pool_size == old_pool_size)
1933 		return;
1934 
1935 	fs_info->thread_pool_size = new_pool_size;
1936 
1937 	btrfs_info(fs_info, "resize thread pool %d -> %d",
1938 	       old_pool_size, new_pool_size);
1939 
1940 	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1941 	btrfs_workqueue_set_max(fs_info->hipri_workers, new_pool_size);
1942 	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1943 	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1944 	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1945 	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1946 	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1947 }
1948 
1949 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1950 				       unsigned long old_opts, int flags)
1951 {
1952 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1953 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1954 	     (flags & SB_RDONLY))) {
1955 		/* wait for any defraggers to finish */
1956 		wait_event(fs_info->transaction_wait,
1957 			   (atomic_read(&fs_info->defrag_running) == 0));
1958 		if (flags & SB_RDONLY)
1959 			sync_filesystem(fs_info->sb);
1960 	}
1961 }
1962 
1963 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1964 					 unsigned long old_opts)
1965 {
1966 	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1967 
1968 	/*
1969 	 * We need to cleanup all defragable inodes if the autodefragment is
1970 	 * close or the filesystem is read only.
1971 	 */
1972 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1973 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1974 		btrfs_cleanup_defrag_inodes(fs_info);
1975 	}
1976 
1977 	/* If we toggled discard async */
1978 	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1979 	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1980 		btrfs_discard_resume(fs_info);
1981 	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1982 		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1983 		btrfs_discard_cleanup(fs_info);
1984 
1985 	/* If we toggled space cache */
1986 	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1987 		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1988 }
1989 
1990 static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1991 {
1992 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1993 	unsigned old_flags = sb->s_flags;
1994 	unsigned long old_opts = fs_info->mount_opt;
1995 	unsigned long old_compress_type = fs_info->compress_type;
1996 	u64 old_max_inline = fs_info->max_inline;
1997 	u32 old_thread_pool_size = fs_info->thread_pool_size;
1998 	u32 old_metadata_ratio = fs_info->metadata_ratio;
1999 	int ret;
2000 
2001 	sync_filesystem(sb);
2002 	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2003 
2004 	if (data) {
2005 		void *new_sec_opts = NULL;
2006 
2007 		ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
2008 		if (!ret)
2009 			ret = security_sb_remount(sb, new_sec_opts);
2010 		security_free_mnt_opts(&new_sec_opts);
2011 		if (ret)
2012 			goto restore;
2013 	}
2014 
2015 	ret = btrfs_parse_options(fs_info, data, *flags);
2016 	if (ret)
2017 		goto restore;
2018 
2019 	ret = btrfs_check_features(fs_info, sb);
2020 	if (ret < 0)
2021 		goto restore;
2022 
2023 	btrfs_remount_begin(fs_info, old_opts, *flags);
2024 	btrfs_resize_thread_pool(fs_info,
2025 		fs_info->thread_pool_size, old_thread_pool_size);
2026 
2027 	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
2028 	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
2029 	    (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
2030 		btrfs_warn(fs_info,
2031 		"remount supports changing free space tree only from ro to rw");
2032 		/* Make sure free space cache options match the state on disk */
2033 		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
2034 			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
2035 			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
2036 		}
2037 		if (btrfs_free_space_cache_v1_active(fs_info)) {
2038 			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
2039 			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
2040 		}
2041 	}
2042 
2043 	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
2044 		goto out;
2045 
2046 	if (*flags & SB_RDONLY) {
2047 		/*
2048 		 * this also happens on 'umount -rf' or on shutdown, when
2049 		 * the filesystem is busy.
2050 		 */
2051 		cancel_work_sync(&fs_info->async_reclaim_work);
2052 		cancel_work_sync(&fs_info->async_data_reclaim_work);
2053 
2054 		btrfs_discard_cleanup(fs_info);
2055 
2056 		/* wait for the uuid_scan task to finish */
2057 		down(&fs_info->uuid_tree_rescan_sem);
2058 		/* avoid complains from lockdep et al. */
2059 		up(&fs_info->uuid_tree_rescan_sem);
2060 
2061 		btrfs_set_sb_rdonly(sb);
2062 
2063 		/*
2064 		 * Setting SB_RDONLY will put the cleaner thread to
2065 		 * sleep at the next loop if it's already active.
2066 		 * If it's already asleep, we'll leave unused block
2067 		 * groups on disk until we're mounted read-write again
2068 		 * unless we clean them up here.
2069 		 */
2070 		btrfs_delete_unused_bgs(fs_info);
2071 
2072 		/*
2073 		 * The cleaner task could be already running before we set the
2074 		 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
2075 		 * We must make sure that after we finish the remount, i.e. after
2076 		 * we call btrfs_commit_super(), the cleaner can no longer start
2077 		 * a transaction - either because it was dropping a dead root,
2078 		 * running delayed iputs or deleting an unused block group (the
2079 		 * cleaner picked a block group from the list of unused block
2080 		 * groups before we were able to in the previous call to
2081 		 * btrfs_delete_unused_bgs()).
2082 		 */
2083 		wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
2084 			    TASK_UNINTERRUPTIBLE);
2085 
2086 		/*
2087 		 * We've set the superblock to RO mode, so we might have made
2088 		 * the cleaner task sleep without running all pending delayed
2089 		 * iputs. Go through all the delayed iputs here, so that if an
2090 		 * unmount happens without remounting RW we don't end up at
2091 		 * finishing close_ctree() with a non-empty list of delayed
2092 		 * iputs.
2093 		 */
2094 		btrfs_run_delayed_iputs(fs_info);
2095 
2096 		btrfs_dev_replace_suspend_for_unmount(fs_info);
2097 		btrfs_scrub_cancel(fs_info);
2098 		btrfs_pause_balance(fs_info);
2099 
2100 		/*
2101 		 * Pause the qgroup rescan worker if it is running. We don't want
2102 		 * it to be still running after we are in RO mode, as after that,
2103 		 * by the time we unmount, it might have left a transaction open,
2104 		 * so we would leak the transaction and/or crash.
2105 		 */
2106 		btrfs_qgroup_wait_for_completion(fs_info, false);
2107 
2108 		ret = btrfs_commit_super(fs_info);
2109 		if (ret)
2110 			goto restore;
2111 	} else {
2112 		if (BTRFS_FS_ERROR(fs_info)) {
2113 			btrfs_err(fs_info,
2114 				"Remounting read-write after error is not allowed");
2115 			ret = -EINVAL;
2116 			goto restore;
2117 		}
2118 		if (fs_info->fs_devices->rw_devices == 0) {
2119 			ret = -EACCES;
2120 			goto restore;
2121 		}
2122 
2123 		if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2124 			btrfs_warn(fs_info,
2125 		"too many missing devices, writable remount is not allowed");
2126 			ret = -EACCES;
2127 			goto restore;
2128 		}
2129 
2130 		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2131 			btrfs_warn(fs_info,
2132 		"mount required to replay tree-log, cannot remount read-write");
2133 			ret = -EINVAL;
2134 			goto restore;
2135 		}
2136 
2137 		/*
2138 		 * NOTE: when remounting with a change that does writes, don't
2139 		 * put it anywhere above this point, as we are not sure to be
2140 		 * safe to write until we pass the above checks.
2141 		 */
2142 		ret = btrfs_start_pre_rw_mount(fs_info);
2143 		if (ret)
2144 			goto restore;
2145 
2146 		btrfs_clear_sb_rdonly(sb);
2147 
2148 		set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2149 	}
2150 out:
2151 	/*
2152 	 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2153 	 * since the absence of the flag means it can be toggled off by remount.
2154 	 */
2155 	*flags |= SB_I_VERSION;
2156 
2157 	wake_up_process(fs_info->transaction_kthread);
2158 	btrfs_remount_cleanup(fs_info, old_opts);
2159 	btrfs_clear_oneshot_options(fs_info);
2160 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2161 
2162 	return 0;
2163 
2164 restore:
2165 	/* We've hit an error - don't reset SB_RDONLY */
2166 	if (sb_rdonly(sb))
2167 		old_flags |= SB_RDONLY;
2168 	if (!(old_flags & SB_RDONLY))
2169 		clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2170 	sb->s_flags = old_flags;
2171 	fs_info->mount_opt = old_opts;
2172 	fs_info->compress_type = old_compress_type;
2173 	fs_info->max_inline = old_max_inline;
2174 	btrfs_resize_thread_pool(fs_info,
2175 		old_thread_pool_size, fs_info->thread_pool_size);
2176 	fs_info->metadata_ratio = old_metadata_ratio;
2177 	btrfs_remount_cleanup(fs_info, old_opts);
2178 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2179 
2180 	return ret;
2181 }
2182 
2183 /* Used to sort the devices by max_avail(descending sort) */
2184 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2185 {
2186 	const struct btrfs_device_info *dev_info1 = a;
2187 	const struct btrfs_device_info *dev_info2 = b;
2188 
2189 	if (dev_info1->max_avail > dev_info2->max_avail)
2190 		return -1;
2191 	else if (dev_info1->max_avail < dev_info2->max_avail)
2192 		return 1;
2193 	return 0;
2194 }
2195 
2196 /*
2197  * sort the devices by max_avail, in which max free extent size of each device
2198  * is stored.(Descending Sort)
2199  */
2200 static inline void btrfs_descending_sort_devices(
2201 					struct btrfs_device_info *devices,
2202 					size_t nr_devices)
2203 {
2204 	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2205 	     btrfs_cmp_device_free_bytes, NULL);
2206 }
2207 
2208 /*
2209  * The helper to calc the free space on the devices that can be used to store
2210  * file data.
2211  */
2212 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2213 					      u64 *free_bytes)
2214 {
2215 	struct btrfs_device_info *devices_info;
2216 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2217 	struct btrfs_device *device;
2218 	u64 type;
2219 	u64 avail_space;
2220 	u64 min_stripe_size;
2221 	int num_stripes = 1;
2222 	int i = 0, nr_devices;
2223 	const struct btrfs_raid_attr *rattr;
2224 
2225 	/*
2226 	 * We aren't under the device list lock, so this is racy-ish, but good
2227 	 * enough for our purposes.
2228 	 */
2229 	nr_devices = fs_info->fs_devices->open_devices;
2230 	if (!nr_devices) {
2231 		smp_mb();
2232 		nr_devices = fs_info->fs_devices->open_devices;
2233 		ASSERT(nr_devices);
2234 		if (!nr_devices) {
2235 			*free_bytes = 0;
2236 			return 0;
2237 		}
2238 	}
2239 
2240 	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2241 			       GFP_KERNEL);
2242 	if (!devices_info)
2243 		return -ENOMEM;
2244 
2245 	/* calc min stripe number for data space allocation */
2246 	type = btrfs_data_alloc_profile(fs_info);
2247 	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2248 
2249 	if (type & BTRFS_BLOCK_GROUP_RAID0)
2250 		num_stripes = nr_devices;
2251 	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
2252 		num_stripes = rattr->ncopies;
2253 	else if (type & BTRFS_BLOCK_GROUP_RAID10)
2254 		num_stripes = 4;
2255 
2256 	/* Adjust for more than 1 stripe per device */
2257 	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2258 
2259 	rcu_read_lock();
2260 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2261 		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2262 						&device->dev_state) ||
2263 		    !device->bdev ||
2264 		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2265 			continue;
2266 
2267 		if (i >= nr_devices)
2268 			break;
2269 
2270 		avail_space = device->total_bytes - device->bytes_used;
2271 
2272 		/* align with stripe_len */
2273 		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2274 
2275 		/*
2276 		 * Ensure we have at least min_stripe_size on top of the
2277 		 * reserved space on the device.
2278 		 */
2279 		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
2280 			continue;
2281 
2282 		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
2283 
2284 		devices_info[i].dev = device;
2285 		devices_info[i].max_avail = avail_space;
2286 
2287 		i++;
2288 	}
2289 	rcu_read_unlock();
2290 
2291 	nr_devices = i;
2292 
2293 	btrfs_descending_sort_devices(devices_info, nr_devices);
2294 
2295 	i = nr_devices - 1;
2296 	avail_space = 0;
2297 	while (nr_devices >= rattr->devs_min) {
2298 		num_stripes = min(num_stripes, nr_devices);
2299 
2300 		if (devices_info[i].max_avail >= min_stripe_size) {
2301 			int j;
2302 			u64 alloc_size;
2303 
2304 			avail_space += devices_info[i].max_avail * num_stripes;
2305 			alloc_size = devices_info[i].max_avail;
2306 			for (j = i + 1 - num_stripes; j <= i; j++)
2307 				devices_info[j].max_avail -= alloc_size;
2308 		}
2309 		i--;
2310 		nr_devices--;
2311 	}
2312 
2313 	kfree(devices_info);
2314 	*free_bytes = avail_space;
2315 	return 0;
2316 }
2317 
2318 /*
2319  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2320  *
2321  * If there's a redundant raid level at DATA block groups, use the respective
2322  * multiplier to scale the sizes.
2323  *
2324  * Unused device space usage is based on simulating the chunk allocator
2325  * algorithm that respects the device sizes and order of allocations.  This is
2326  * a close approximation of the actual use but there are other factors that may
2327  * change the result (like a new metadata chunk).
2328  *
2329  * If metadata is exhausted, f_bavail will be 0.
2330  */
2331 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2332 {
2333 	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2334 	struct btrfs_super_block *disk_super = fs_info->super_copy;
2335 	struct btrfs_space_info *found;
2336 	u64 total_used = 0;
2337 	u64 total_free_data = 0;
2338 	u64 total_free_meta = 0;
2339 	u32 bits = fs_info->sectorsize_bits;
2340 	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2341 	unsigned factor = 1;
2342 	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2343 	int ret;
2344 	u64 thresh = 0;
2345 	int mixed = 0;
2346 
2347 	list_for_each_entry(found, &fs_info->space_info, list) {
2348 		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2349 			int i;
2350 
2351 			total_free_data += found->disk_total - found->disk_used;
2352 			total_free_data -=
2353 				btrfs_account_ro_block_groups_free_space(found);
2354 
2355 			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2356 				if (!list_empty(&found->block_groups[i]))
2357 					factor = btrfs_bg_type_to_factor(
2358 						btrfs_raid_array[i].bg_flag);
2359 			}
2360 		}
2361 
2362 		/*
2363 		 * Metadata in mixed block goup profiles are accounted in data
2364 		 */
2365 		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2366 			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2367 				mixed = 1;
2368 			else
2369 				total_free_meta += found->disk_total -
2370 					found->disk_used;
2371 		}
2372 
2373 		total_used += found->disk_used;
2374 	}
2375 
2376 	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2377 	buf->f_blocks >>= bits;
2378 	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2379 
2380 	/* Account global block reserve as used, it's in logical size already */
2381 	spin_lock(&block_rsv->lock);
2382 	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
2383 	if (buf->f_bfree >= block_rsv->size >> bits)
2384 		buf->f_bfree -= block_rsv->size >> bits;
2385 	else
2386 		buf->f_bfree = 0;
2387 	spin_unlock(&block_rsv->lock);
2388 
2389 	buf->f_bavail = div_u64(total_free_data, factor);
2390 	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2391 	if (ret)
2392 		return ret;
2393 	buf->f_bavail += div_u64(total_free_data, factor);
2394 	buf->f_bavail = buf->f_bavail >> bits;
2395 
2396 	/*
2397 	 * We calculate the remaining metadata space minus global reserve. If
2398 	 * this is (supposedly) smaller than zero, there's no space. But this
2399 	 * does not hold in practice, the exhausted state happens where's still
2400 	 * some positive delta. So we apply some guesswork and compare the
2401 	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2402 	 *
2403 	 * We probably cannot calculate the exact threshold value because this
2404 	 * depends on the internal reservations requested by various
2405 	 * operations, so some operations that consume a few metadata will
2406 	 * succeed even if the Avail is zero. But this is better than the other
2407 	 * way around.
2408 	 */
2409 	thresh = SZ_4M;
2410 
2411 	/*
2412 	 * We only want to claim there's no available space if we can no longer
2413 	 * allocate chunks for our metadata profile and our global reserve will
2414 	 * not fit in the free metadata space.  If we aren't ->full then we
2415 	 * still can allocate chunks and thus are fine using the currently
2416 	 * calculated f_bavail.
2417 	 */
2418 	if (!mixed && block_rsv->space_info->full &&
2419 	    total_free_meta - thresh < block_rsv->size)
2420 		buf->f_bavail = 0;
2421 
2422 	buf->f_type = BTRFS_SUPER_MAGIC;
2423 	buf->f_bsize = dentry->d_sb->s_blocksize;
2424 	buf->f_namelen = BTRFS_NAME_LEN;
2425 
2426 	/* We treat it as constant endianness (it doesn't matter _which_)
2427 	   because we want the fsid to come out the same whether mounted
2428 	   on a big-endian or little-endian host */
2429 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2430 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2431 	/* Mask in the root object ID too, to disambiguate subvols */
2432 	buf->f_fsid.val[0] ^=
2433 		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2434 	buf->f_fsid.val[1] ^=
2435 		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2436 
2437 	return 0;
2438 }
2439 
2440 static void btrfs_kill_super(struct super_block *sb)
2441 {
2442 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2443 	kill_anon_super(sb);
2444 	btrfs_free_fs_info(fs_info);
2445 }
2446 
2447 static struct file_system_type btrfs_fs_type = {
2448 	.owner		= THIS_MODULE,
2449 	.name		= "btrfs",
2450 	.mount		= btrfs_mount,
2451 	.kill_sb	= btrfs_kill_super,
2452 	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2453 };
2454 
2455 static struct file_system_type btrfs_root_fs_type = {
2456 	.owner		= THIS_MODULE,
2457 	.name		= "btrfs",
2458 	.mount		= btrfs_mount_root,
2459 	.kill_sb	= btrfs_kill_super,
2460 	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2461 };
2462 
2463 MODULE_ALIAS_FS("btrfs");
2464 
2465 static int btrfs_control_open(struct inode *inode, struct file *file)
2466 {
2467 	/*
2468 	 * The control file's private_data is used to hold the
2469 	 * transaction when it is started and is used to keep
2470 	 * track of whether a transaction is already in progress.
2471 	 */
2472 	file->private_data = NULL;
2473 	return 0;
2474 }
2475 
2476 /*
2477  * Used by /dev/btrfs-control for devices ioctls.
2478  */
2479 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2480 				unsigned long arg)
2481 {
2482 	struct btrfs_ioctl_vol_args *vol;
2483 	struct btrfs_device *device = NULL;
2484 	dev_t devt = 0;
2485 	int ret = -ENOTTY;
2486 
2487 	if (!capable(CAP_SYS_ADMIN))
2488 		return -EPERM;
2489 
2490 	vol = memdup_user((void __user *)arg, sizeof(*vol));
2491 	if (IS_ERR(vol))
2492 		return PTR_ERR(vol);
2493 	vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2494 
2495 	switch (cmd) {
2496 	case BTRFS_IOC_SCAN_DEV:
2497 		mutex_lock(&uuid_mutex);
2498 		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2499 					       &btrfs_root_fs_type);
2500 		ret = PTR_ERR_OR_ZERO(device);
2501 		mutex_unlock(&uuid_mutex);
2502 		break;
2503 	case BTRFS_IOC_FORGET_DEV:
2504 		if (vol->name[0] != 0) {
2505 			ret = lookup_bdev(vol->name, &devt);
2506 			if (ret)
2507 				break;
2508 		}
2509 		ret = btrfs_forget_devices(devt);
2510 		break;
2511 	case BTRFS_IOC_DEVICES_READY:
2512 		mutex_lock(&uuid_mutex);
2513 		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2514 					       &btrfs_root_fs_type);
2515 		if (IS_ERR(device)) {
2516 			mutex_unlock(&uuid_mutex);
2517 			ret = PTR_ERR(device);
2518 			break;
2519 		}
2520 		ret = !(device->fs_devices->num_devices ==
2521 			device->fs_devices->total_devices);
2522 		mutex_unlock(&uuid_mutex);
2523 		break;
2524 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2525 		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2526 		break;
2527 	}
2528 
2529 	kfree(vol);
2530 	return ret;
2531 }
2532 
2533 static int btrfs_freeze(struct super_block *sb)
2534 {
2535 	struct btrfs_trans_handle *trans;
2536 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2537 	struct btrfs_root *root = fs_info->tree_root;
2538 
2539 	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2540 	/*
2541 	 * We don't need a barrier here, we'll wait for any transaction that
2542 	 * could be in progress on other threads (and do delayed iputs that
2543 	 * we want to avoid on a frozen filesystem), or do the commit
2544 	 * ourselves.
2545 	 */
2546 	trans = btrfs_attach_transaction_barrier(root);
2547 	if (IS_ERR(trans)) {
2548 		/* no transaction, don't bother */
2549 		if (PTR_ERR(trans) == -ENOENT)
2550 			return 0;
2551 		return PTR_ERR(trans);
2552 	}
2553 	return btrfs_commit_transaction(trans);
2554 }
2555 
2556 static int check_dev_super(struct btrfs_device *dev)
2557 {
2558 	struct btrfs_fs_info *fs_info = dev->fs_info;
2559 	struct btrfs_super_block *sb;
2560 	u16 csum_type;
2561 	int ret = 0;
2562 
2563 	/* This should be called with fs still frozen. */
2564 	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2565 
2566 	/* Missing dev, no need to check. */
2567 	if (!dev->bdev)
2568 		return 0;
2569 
2570 	/* Only need to check the primary super block. */
2571 	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2572 	if (IS_ERR(sb))
2573 		return PTR_ERR(sb);
2574 
2575 	/* Verify the checksum. */
2576 	csum_type = btrfs_super_csum_type(sb);
2577 	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2578 		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2579 			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2580 		ret = -EUCLEAN;
2581 		goto out;
2582 	}
2583 
2584 	if (btrfs_check_super_csum(fs_info, sb)) {
2585 		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2586 		ret = -EUCLEAN;
2587 		goto out;
2588 	}
2589 
2590 	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2591 	ret = btrfs_validate_super(fs_info, sb, 0);
2592 	if (ret < 0)
2593 		goto out;
2594 
2595 	if (btrfs_super_generation(sb) != fs_info->last_trans_committed) {
2596 		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2597 			btrfs_super_generation(sb),
2598 			fs_info->last_trans_committed);
2599 		ret = -EUCLEAN;
2600 		goto out;
2601 	}
2602 out:
2603 	btrfs_release_disk_super(sb);
2604 	return ret;
2605 }
2606 
2607 static int btrfs_unfreeze(struct super_block *sb)
2608 {
2609 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2610 	struct btrfs_device *device;
2611 	int ret = 0;
2612 
2613 	/*
2614 	 * Make sure the fs is not changed by accident (like hibernation then
2615 	 * modified by other OS).
2616 	 * If we found anything wrong, we mark the fs error immediately.
2617 	 *
2618 	 * And since the fs is frozen, no one can modify the fs yet, thus
2619 	 * we don't need to hold device_list_mutex.
2620 	 */
2621 	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2622 		ret = check_dev_super(device);
2623 		if (ret < 0) {
2624 			btrfs_handle_fs_error(fs_info, ret,
2625 				"super block on devid %llu got modified unexpectedly",
2626 				device->devid);
2627 			break;
2628 		}
2629 	}
2630 	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2631 
2632 	/*
2633 	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2634 	 * above checks failed. Since the fs is either fine or read-only, we're
2635 	 * safe to continue, without causing further damage.
2636 	 */
2637 	return 0;
2638 }
2639 
2640 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2641 {
2642 	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2643 
2644 	/*
2645 	 * There should be always a valid pointer in latest_dev, it may be stale
2646 	 * for a short moment in case it's being deleted but still valid until
2647 	 * the end of RCU grace period.
2648 	 */
2649 	rcu_read_lock();
2650 	seq_escape(m, rcu_str_deref(fs_info->fs_devices->latest_dev->name), " \t\n\\");
2651 	rcu_read_unlock();
2652 
2653 	return 0;
2654 }
2655 
2656 static const struct super_operations btrfs_super_ops = {
2657 	.drop_inode	= btrfs_drop_inode,
2658 	.evict_inode	= btrfs_evict_inode,
2659 	.put_super	= btrfs_put_super,
2660 	.sync_fs	= btrfs_sync_fs,
2661 	.show_options	= btrfs_show_options,
2662 	.show_devname	= btrfs_show_devname,
2663 	.alloc_inode	= btrfs_alloc_inode,
2664 	.destroy_inode	= btrfs_destroy_inode,
2665 	.free_inode	= btrfs_free_inode,
2666 	.statfs		= btrfs_statfs,
2667 	.remount_fs	= btrfs_remount,
2668 	.freeze_fs	= btrfs_freeze,
2669 	.unfreeze_fs	= btrfs_unfreeze,
2670 };
2671 
2672 static const struct file_operations btrfs_ctl_fops = {
2673 	.open = btrfs_control_open,
2674 	.unlocked_ioctl	 = btrfs_control_ioctl,
2675 	.compat_ioctl = compat_ptr_ioctl,
2676 	.owner	 = THIS_MODULE,
2677 	.llseek = noop_llseek,
2678 };
2679 
2680 static struct miscdevice btrfs_misc = {
2681 	.minor		= BTRFS_MINOR,
2682 	.name		= "btrfs-control",
2683 	.fops		= &btrfs_ctl_fops
2684 };
2685 
2686 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2687 MODULE_ALIAS("devname:btrfs-control");
2688 
2689 static int __init btrfs_interface_init(void)
2690 {
2691 	return misc_register(&btrfs_misc);
2692 }
2693 
2694 static __cold void btrfs_interface_exit(void)
2695 {
2696 	misc_deregister(&btrfs_misc);
2697 }
2698 
2699 static int __init btrfs_print_mod_info(void)
2700 {
2701 	static const char options[] = ""
2702 #ifdef CONFIG_BTRFS_DEBUG
2703 			", debug=on"
2704 #endif
2705 #ifdef CONFIG_BTRFS_ASSERT
2706 			", assert=on"
2707 #endif
2708 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2709 			", integrity-checker=on"
2710 #endif
2711 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2712 			", ref-verify=on"
2713 #endif
2714 #ifdef CONFIG_BLK_DEV_ZONED
2715 			", zoned=yes"
2716 #else
2717 			", zoned=no"
2718 #endif
2719 #ifdef CONFIG_FS_VERITY
2720 			", fsverity=yes"
2721 #else
2722 			", fsverity=no"
2723 #endif
2724 			;
2725 	pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2726 	return 0;
2727 }
2728 
2729 static int register_btrfs(void)
2730 {
2731 	return register_filesystem(&btrfs_fs_type);
2732 }
2733 
2734 static void unregister_btrfs(void)
2735 {
2736 	unregister_filesystem(&btrfs_fs_type);
2737 }
2738 
2739 /* Helper structure for long init/exit functions. */
2740 struct init_sequence {
2741 	int (*init_func)(void);
2742 	/* Can be NULL if the init_func doesn't need cleanup. */
2743 	void (*exit_func)(void);
2744 };
2745 
2746 static const struct init_sequence mod_init_seq[] = {
2747 	{
2748 		.init_func = btrfs_props_init,
2749 		.exit_func = NULL,
2750 	}, {
2751 		.init_func = btrfs_init_sysfs,
2752 		.exit_func = btrfs_exit_sysfs,
2753 	}, {
2754 		.init_func = btrfs_init_compress,
2755 		.exit_func = btrfs_exit_compress,
2756 	}, {
2757 		.init_func = btrfs_init_cachep,
2758 		.exit_func = btrfs_destroy_cachep,
2759 	}, {
2760 		.init_func = btrfs_transaction_init,
2761 		.exit_func = btrfs_transaction_exit,
2762 	}, {
2763 		.init_func = btrfs_ctree_init,
2764 		.exit_func = btrfs_ctree_exit,
2765 	}, {
2766 		.init_func = btrfs_free_space_init,
2767 		.exit_func = btrfs_free_space_exit,
2768 	}, {
2769 		.init_func = extent_state_init_cachep,
2770 		.exit_func = extent_state_free_cachep,
2771 	}, {
2772 		.init_func = extent_buffer_init_cachep,
2773 		.exit_func = extent_buffer_free_cachep,
2774 	}, {
2775 		.init_func = btrfs_bioset_init,
2776 		.exit_func = btrfs_bioset_exit,
2777 	}, {
2778 		.init_func = extent_map_init,
2779 		.exit_func = extent_map_exit,
2780 	}, {
2781 		.init_func = ordered_data_init,
2782 		.exit_func = ordered_data_exit,
2783 	}, {
2784 		.init_func = btrfs_delayed_inode_init,
2785 		.exit_func = btrfs_delayed_inode_exit,
2786 	}, {
2787 		.init_func = btrfs_auto_defrag_init,
2788 		.exit_func = btrfs_auto_defrag_exit,
2789 	}, {
2790 		.init_func = btrfs_delayed_ref_init,
2791 		.exit_func = btrfs_delayed_ref_exit,
2792 	}, {
2793 		.init_func = btrfs_prelim_ref_init,
2794 		.exit_func = btrfs_prelim_ref_exit,
2795 	}, {
2796 		.init_func = btrfs_interface_init,
2797 		.exit_func = btrfs_interface_exit,
2798 	}, {
2799 		.init_func = btrfs_print_mod_info,
2800 		.exit_func = NULL,
2801 	}, {
2802 		.init_func = btrfs_run_sanity_tests,
2803 		.exit_func = NULL,
2804 	}, {
2805 		.init_func = register_btrfs,
2806 		.exit_func = unregister_btrfs,
2807 	}
2808 };
2809 
2810 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2811 
2812 static void __exit exit_btrfs_fs(void)
2813 {
2814 	int i;
2815 
2816 	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2817 		if (!mod_init_result[i])
2818 			continue;
2819 		if (mod_init_seq[i].exit_func)
2820 			mod_init_seq[i].exit_func();
2821 		mod_init_result[i] = false;
2822 	}
2823 }
2824 
2825 static int __init init_btrfs_fs(void)
2826 {
2827 	int ret;
2828 	int i;
2829 
2830 	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2831 		ASSERT(!mod_init_result[i]);
2832 		ret = mod_init_seq[i].init_func();
2833 		if (ret < 0)
2834 			goto error;
2835 		mod_init_result[i] = true;
2836 	}
2837 	return 0;
2838 
2839 error:
2840 	/*
2841 	 * If we call exit_btrfs_fs() it would cause section mismatch.
2842 	 * As init_btrfs_fs() belongs to .init.text, while exit_btrfs_fs()
2843 	 * belongs to .exit.text.
2844 	 */
2845 	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2846 		if (!mod_init_result[i])
2847 			continue;
2848 		if (mod_init_seq[i].exit_func)
2849 			mod_init_seq[i].exit_func();
2850 		mod_init_result[i] = false;
2851 	}
2852 	return ret;
2853 }
2854 
2855 late_initcall(init_btrfs_fs);
2856 module_exit(exit_btrfs_fs)
2857 
2858 MODULE_LICENSE("GPL");
2859 MODULE_SOFTDEP("pre: crc32c");
2860 MODULE_SOFTDEP("pre: xxhash64");
2861 MODULE_SOFTDEP("pre: sha256");
2862 MODULE_SOFTDEP("pre: blake2b-256");
2863