super.c revision 14605409
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2007 Oracle.  All rights reserved.
4 */
5
6#include <linux/blkdev.h>
7#include <linux/module.h>
8#include <linux/fs.h>
9#include <linux/pagemap.h>
10#include <linux/highmem.h>
11#include <linux/time.h>
12#include <linux/init.h>
13#include <linux/seq_file.h>
14#include <linux/string.h>
15#include <linux/backing-dev.h>
16#include <linux/mount.h>
17#include <linux/writeback.h>
18#include <linux/statfs.h>
19#include <linux/compat.h>
20#include <linux/parser.h>
21#include <linux/ctype.h>
22#include <linux/namei.h>
23#include <linux/miscdevice.h>
24#include <linux/magic.h>
25#include <linux/slab.h>
26#include <linux/cleancache.h>
27#include <linux/ratelimit.h>
28#include <linux/crc32c.h>
29#include <linux/btrfs.h>
30#include "delayed-inode.h"
31#include "ctree.h"
32#include "disk-io.h"
33#include "transaction.h"
34#include "btrfs_inode.h"
35#include "print-tree.h"
36#include "props.h"
37#include "xattr.h"
38#include "volumes.h"
39#include "export.h"
40#include "compression.h"
41#include "rcu-string.h"
42#include "dev-replace.h"
43#include "free-space-cache.h"
44#include "backref.h"
45#include "space-info.h"
46#include "sysfs.h"
47#include "zoned.h"
48#include "tests/btrfs-tests.h"
49#include "block-group.h"
50#include "discard.h"
51#include "qgroup.h"
52#define CREATE_TRACE_POINTS
53#include <trace/events/btrfs.h>
54
55static 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 */
65static struct file_system_type btrfs_fs_type;
66static struct file_system_type btrfs_root_fs_type;
67
68static int btrfs_remount(struct super_block *sb, int *flags, char *data);
69
70/*
71 * Generally the error codes correspond to their respective errors, but there
72 * are a few special cases.
73 *
74 * EUCLEAN: Any sort of corruption that we encounter.  The tree-checker for
75 *          instance will return EUCLEAN if any of the blocks are corrupted in
76 *          a way that is problematic.  We want to reserve EUCLEAN for these
77 *          sort of corruptions.
78 *
79 * EROFS: If we check BTRFS_FS_STATE_ERROR and fail out with a return error, we
80 *        need to use EROFS for this case.  We will have no idea of the
81 *        original failure, that will have been reported at the time we tripped
82 *        over the error.  Each subsequent error that doesn't have any context
83 *        of the original error should use EROFS when handling BTRFS_FS_STATE_ERROR.
84 */
85const char * __attribute_const__ btrfs_decode_error(int errno)
86{
87	char *errstr = "unknown";
88
89	switch (errno) {
90	case -ENOENT:		/* -2 */
91		errstr = "No such entry";
92		break;
93	case -EIO:		/* -5 */
94		errstr = "IO failure";
95		break;
96	case -ENOMEM:		/* -12*/
97		errstr = "Out of memory";
98		break;
99	case -EEXIST:		/* -17 */
100		errstr = "Object already exists";
101		break;
102	case -ENOSPC:		/* -28 */
103		errstr = "No space left";
104		break;
105	case -EROFS:		/* -30 */
106		errstr = "Readonly filesystem";
107		break;
108	case -EOPNOTSUPP:	/* -95 */
109		errstr = "Operation not supported";
110		break;
111	case -EUCLEAN:		/* -117 */
112		errstr = "Filesystem corrupted";
113		break;
114	case -EDQUOT:		/* -122 */
115		errstr = "Quota exceeded";
116		break;
117	}
118
119	return errstr;
120}
121
122/*
123 * __btrfs_handle_fs_error decodes expected errors from the caller and
124 * invokes the appropriate error response.
125 */
126__cold
127void __btrfs_handle_fs_error(struct btrfs_fs_info *fs_info, const char *function,
128		       unsigned int line, int errno, const char *fmt, ...)
129{
130	struct super_block *sb = fs_info->sb;
131#ifdef CONFIG_PRINTK
132	const char *errstr;
133#endif
134
135	/*
136	 * Special case: if the error is EROFS, and we're already
137	 * under SB_RDONLY, then it is safe here.
138	 */
139	if (errno == -EROFS && sb_rdonly(sb))
140  		return;
141
142#ifdef CONFIG_PRINTK
143	errstr = btrfs_decode_error(errno);
144	if (fmt) {
145		struct va_format vaf;
146		va_list args;
147
148		va_start(args, fmt);
149		vaf.fmt = fmt;
150		vaf.va = &args;
151
152		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n",
153			sb->s_id, function, line, errno, errstr, &vaf);
154		va_end(args);
155	} else {
156		pr_crit("BTRFS: error (device %s) in %s:%d: errno=%d %s\n",
157			sb->s_id, function, line, errno, errstr);
158	}
159#endif
160
161	/*
162	 * Today we only save the error info to memory.  Long term we'll
163	 * also send it down to the disk
164	 */
165	set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state);
166
167	/* Don't go through full error handling during mount */
168	if (!(sb->s_flags & SB_BORN))
169		return;
170
171	if (sb_rdonly(sb))
172		return;
173
174	btrfs_discard_stop(fs_info);
175
176	/* btrfs handle error by forcing the filesystem readonly */
177	btrfs_set_sb_rdonly(sb);
178	btrfs_info(fs_info, "forced readonly");
179	/*
180	 * Note that a running device replace operation is not canceled here
181	 * although there is no way to update the progress. It would add the
182	 * risk of a deadlock, therefore the canceling is omitted. The only
183	 * penalty is that some I/O remains active until the procedure
184	 * completes. The next time when the filesystem is mounted writable
185	 * again, the device replace operation continues.
186	 */
187}
188
189#ifdef CONFIG_PRINTK
190static const char * const logtypes[] = {
191	"emergency",
192	"alert",
193	"critical",
194	"error",
195	"warning",
196	"notice",
197	"info",
198	"debug",
199};
200
201
202/*
203 * Use one ratelimit state per log level so that a flood of less important
204 * messages doesn't cause more important ones to be dropped.
205 */
206static struct ratelimit_state printk_limits[] = {
207	RATELIMIT_STATE_INIT(printk_limits[0], DEFAULT_RATELIMIT_INTERVAL, 100),
208	RATELIMIT_STATE_INIT(printk_limits[1], DEFAULT_RATELIMIT_INTERVAL, 100),
209	RATELIMIT_STATE_INIT(printk_limits[2], DEFAULT_RATELIMIT_INTERVAL, 100),
210	RATELIMIT_STATE_INIT(printk_limits[3], DEFAULT_RATELIMIT_INTERVAL, 100),
211	RATELIMIT_STATE_INIT(printk_limits[4], DEFAULT_RATELIMIT_INTERVAL, 100),
212	RATELIMIT_STATE_INIT(printk_limits[5], DEFAULT_RATELIMIT_INTERVAL, 100),
213	RATELIMIT_STATE_INIT(printk_limits[6], DEFAULT_RATELIMIT_INTERVAL, 100),
214	RATELIMIT_STATE_INIT(printk_limits[7], DEFAULT_RATELIMIT_INTERVAL, 100),
215};
216
217void __cold btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...)
218{
219	char lvl[PRINTK_MAX_SINGLE_HEADER_LEN + 1] = "\0";
220	struct va_format vaf;
221	va_list args;
222	int kern_level;
223	const char *type = logtypes[4];
224	struct ratelimit_state *ratelimit = &printk_limits[4];
225
226	va_start(args, fmt);
227
228	while ((kern_level = printk_get_level(fmt)) != 0) {
229		size_t size = printk_skip_level(fmt) - fmt;
230
231		if (kern_level >= '0' && kern_level <= '7') {
232			memcpy(lvl, fmt,  size);
233			lvl[size] = '\0';
234			type = logtypes[kern_level - '0'];
235			ratelimit = &printk_limits[kern_level - '0'];
236		}
237		fmt += size;
238	}
239
240	vaf.fmt = fmt;
241	vaf.va = &args;
242
243	if (__ratelimit(ratelimit)) {
244		if (fs_info)
245			printk("%sBTRFS %s (device %s): %pV\n", lvl, type,
246				fs_info->sb->s_id, &vaf);
247		else
248			printk("%sBTRFS %s: %pV\n", lvl, type, &vaf);
249	}
250
251	va_end(args);
252}
253#endif
254
255#if BITS_PER_LONG == 32
256void __cold btrfs_warn_32bit_limit(struct btrfs_fs_info *fs_info)
257{
258	if (!test_and_set_bit(BTRFS_FS_32BIT_WARN, &fs_info->flags)) {
259		btrfs_warn(fs_info, "reaching 32bit limit for logical addresses");
260		btrfs_warn(fs_info,
261"due to page cache limit on 32bit systems, btrfs can't access metadata at or beyond %lluT",
262			   BTRFS_32BIT_MAX_FILE_SIZE >> 40);
263		btrfs_warn(fs_info,
264			   "please consider upgrading to 64bit kernel/hardware");
265	}
266}
267
268void __cold btrfs_err_32bit_limit(struct btrfs_fs_info *fs_info)
269{
270	if (!test_and_set_bit(BTRFS_FS_32BIT_ERROR, &fs_info->flags)) {
271		btrfs_err(fs_info, "reached 32bit limit for logical addresses");
272		btrfs_err(fs_info,
273"due to page cache limit on 32bit systems, metadata beyond %lluT can't be accessed",
274			  BTRFS_32BIT_MAX_FILE_SIZE >> 40);
275		btrfs_err(fs_info,
276			   "please consider upgrading to 64bit kernel/hardware");
277	}
278}
279#endif
280
281/*
282 * We only mark the transaction aborted and then set the file system read-only.
283 * This will prevent new transactions from starting or trying to join this
284 * one.
285 *
286 * This means that error recovery at the call site is limited to freeing
287 * any local memory allocations and passing the error code up without
288 * further cleanup. The transaction should complete as it normally would
289 * in the call path but will return -EIO.
290 *
291 * We'll complete the cleanup in btrfs_end_transaction and
292 * btrfs_commit_transaction.
293 */
294__cold
295void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
296			       const char *function,
297			       unsigned int line, int errno)
298{
299	struct btrfs_fs_info *fs_info = trans->fs_info;
300
301	WRITE_ONCE(trans->aborted, errno);
302	WRITE_ONCE(trans->transaction->aborted, errno);
303	/* Wake up anybody who may be waiting on this transaction */
304	wake_up(&fs_info->transaction_wait);
305	wake_up(&fs_info->transaction_blocked_wait);
306	__btrfs_handle_fs_error(fs_info, function, line, errno, NULL);
307}
308/*
309 * __btrfs_panic decodes unexpected, fatal errors from the caller,
310 * issues an alert, and either panics or BUGs, depending on mount options.
311 */
312__cold
313void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function,
314		   unsigned int line, int errno, const char *fmt, ...)
315{
316	char *s_id = "<unknown>";
317	const char *errstr;
318	struct va_format vaf = { .fmt = fmt };
319	va_list args;
320
321	if (fs_info)
322		s_id = fs_info->sb->s_id;
323
324	va_start(args, fmt);
325	vaf.va = &args;
326
327	errstr = btrfs_decode_error(errno);
328	if (fs_info && (btrfs_test_opt(fs_info, PANIC_ON_FATAL_ERROR)))
329		panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n",
330			s_id, function, line, &vaf, errno, errstr);
331
332	btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)",
333		   function, line, &vaf, errno, errstr);
334	va_end(args);
335	/* Caller calls BUG() */
336}
337
338static void btrfs_put_super(struct super_block *sb)
339{
340	close_ctree(btrfs_sb(sb));
341}
342
343enum {
344	Opt_acl, Opt_noacl,
345	Opt_clear_cache,
346	Opt_commit_interval,
347	Opt_compress,
348	Opt_compress_force,
349	Opt_compress_force_type,
350	Opt_compress_type,
351	Opt_degraded,
352	Opt_device,
353	Opt_fatal_errors,
354	Opt_flushoncommit, Opt_noflushoncommit,
355	Opt_max_inline,
356	Opt_barrier, Opt_nobarrier,
357	Opt_datacow, Opt_nodatacow,
358	Opt_datasum, Opt_nodatasum,
359	Opt_defrag, Opt_nodefrag,
360	Opt_discard, Opt_nodiscard,
361	Opt_discard_mode,
362	Opt_norecovery,
363	Opt_ratio,
364	Opt_rescan_uuid_tree,
365	Opt_skip_balance,
366	Opt_space_cache, Opt_no_space_cache,
367	Opt_space_cache_version,
368	Opt_ssd, Opt_nossd,
369	Opt_ssd_spread, Opt_nossd_spread,
370	Opt_subvol,
371	Opt_subvol_empty,
372	Opt_subvolid,
373	Opt_thread_pool,
374	Opt_treelog, Opt_notreelog,
375	Opt_user_subvol_rm_allowed,
376
377	/* Rescue options */
378	Opt_rescue,
379	Opt_usebackuproot,
380	Opt_nologreplay,
381	Opt_ignorebadroots,
382	Opt_ignoredatacsums,
383	Opt_rescue_all,
384
385	/* Deprecated options */
386	Opt_recovery,
387	Opt_inode_cache, Opt_noinode_cache,
388
389	/* Debugging options */
390	Opt_check_integrity,
391	Opt_check_integrity_including_extent_data,
392	Opt_check_integrity_print_mask,
393	Opt_enospc_debug, Opt_noenospc_debug,
394#ifdef CONFIG_BTRFS_DEBUG
395	Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
396#endif
397#ifdef CONFIG_BTRFS_FS_REF_VERIFY
398	Opt_ref_verify,
399#endif
400	Opt_err,
401};
402
403static const match_table_t tokens = {
404	{Opt_acl, "acl"},
405	{Opt_noacl, "noacl"},
406	{Opt_clear_cache, "clear_cache"},
407	{Opt_commit_interval, "commit=%u"},
408	{Opt_compress, "compress"},
409	{Opt_compress_type, "compress=%s"},
410	{Opt_compress_force, "compress-force"},
411	{Opt_compress_force_type, "compress-force=%s"},
412	{Opt_degraded, "degraded"},
413	{Opt_device, "device=%s"},
414	{Opt_fatal_errors, "fatal_errors=%s"},
415	{Opt_flushoncommit, "flushoncommit"},
416	{Opt_noflushoncommit, "noflushoncommit"},
417	{Opt_inode_cache, "inode_cache"},
418	{Opt_noinode_cache, "noinode_cache"},
419	{Opt_max_inline, "max_inline=%s"},
420	{Opt_barrier, "barrier"},
421	{Opt_nobarrier, "nobarrier"},
422	{Opt_datacow, "datacow"},
423	{Opt_nodatacow, "nodatacow"},
424	{Opt_datasum, "datasum"},
425	{Opt_nodatasum, "nodatasum"},
426	{Opt_defrag, "autodefrag"},
427	{Opt_nodefrag, "noautodefrag"},
428	{Opt_discard, "discard"},
429	{Opt_discard_mode, "discard=%s"},
430	{Opt_nodiscard, "nodiscard"},
431	{Opt_norecovery, "norecovery"},
432	{Opt_ratio, "metadata_ratio=%u"},
433	{Opt_rescan_uuid_tree, "rescan_uuid_tree"},
434	{Opt_skip_balance, "skip_balance"},
435	{Opt_space_cache, "space_cache"},
436	{Opt_no_space_cache, "nospace_cache"},
437	{Opt_space_cache_version, "space_cache=%s"},
438	{Opt_ssd, "ssd"},
439	{Opt_nossd, "nossd"},
440	{Opt_ssd_spread, "ssd_spread"},
441	{Opt_nossd_spread, "nossd_spread"},
442	{Opt_subvol, "subvol=%s"},
443	{Opt_subvol_empty, "subvol="},
444	{Opt_subvolid, "subvolid=%s"},
445	{Opt_thread_pool, "thread_pool=%u"},
446	{Opt_treelog, "treelog"},
447	{Opt_notreelog, "notreelog"},
448	{Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"},
449
450	/* Rescue options */
451	{Opt_rescue, "rescue=%s"},
452	/* Deprecated, with alias rescue=nologreplay */
453	{Opt_nologreplay, "nologreplay"},
454	/* Deprecated, with alias rescue=usebackuproot */
455	{Opt_usebackuproot, "usebackuproot"},
456
457	/* Deprecated options */
458	{Opt_recovery, "recovery"},
459
460	/* Debugging options */
461	{Opt_check_integrity, "check_int"},
462	{Opt_check_integrity_including_extent_data, "check_int_data"},
463	{Opt_check_integrity_print_mask, "check_int_print_mask=%u"},
464	{Opt_enospc_debug, "enospc_debug"},
465	{Opt_noenospc_debug, "noenospc_debug"},
466#ifdef CONFIG_BTRFS_DEBUG
467	{Opt_fragment_data, "fragment=data"},
468	{Opt_fragment_metadata, "fragment=metadata"},
469	{Opt_fragment_all, "fragment=all"},
470#endif
471#ifdef CONFIG_BTRFS_FS_REF_VERIFY
472	{Opt_ref_verify, "ref_verify"},
473#endif
474	{Opt_err, NULL},
475};
476
477static const match_table_t rescue_tokens = {
478	{Opt_usebackuproot, "usebackuproot"},
479	{Opt_nologreplay, "nologreplay"},
480	{Opt_ignorebadroots, "ignorebadroots"},
481	{Opt_ignorebadroots, "ibadroots"},
482	{Opt_ignoredatacsums, "ignoredatacsums"},
483	{Opt_ignoredatacsums, "idatacsums"},
484	{Opt_rescue_all, "all"},
485	{Opt_err, NULL},
486};
487
488static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt,
489			    const char *opt_name)
490{
491	if (fs_info->mount_opt & opt) {
492		btrfs_err(fs_info, "%s must be used with ro mount option",
493			  opt_name);
494		return true;
495	}
496	return false;
497}
498
499static int parse_rescue_options(struct btrfs_fs_info *info, const char *options)
500{
501	char *opts;
502	char *orig;
503	char *p;
504	substring_t args[MAX_OPT_ARGS];
505	int ret = 0;
506
507	opts = kstrdup(options, GFP_KERNEL);
508	if (!opts)
509		return -ENOMEM;
510	orig = opts;
511
512	while ((p = strsep(&opts, ":")) != NULL) {
513		int token;
514
515		if (!*p)
516			continue;
517		token = match_token(p, rescue_tokens, args);
518		switch (token){
519		case Opt_usebackuproot:
520			btrfs_info(info,
521				   "trying to use backup root at mount time");
522			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
523			break;
524		case Opt_nologreplay:
525			btrfs_set_and_info(info, NOLOGREPLAY,
526					   "disabling log replay at mount time");
527			break;
528		case Opt_ignorebadroots:
529			btrfs_set_and_info(info, IGNOREBADROOTS,
530					   "ignoring bad roots");
531			break;
532		case Opt_ignoredatacsums:
533			btrfs_set_and_info(info, IGNOREDATACSUMS,
534					   "ignoring data csums");
535			break;
536		case Opt_rescue_all:
537			btrfs_info(info, "enabling all of the rescue options");
538			btrfs_set_and_info(info, IGNOREDATACSUMS,
539					   "ignoring data csums");
540			btrfs_set_and_info(info, IGNOREBADROOTS,
541					   "ignoring bad roots");
542			btrfs_set_and_info(info, NOLOGREPLAY,
543					   "disabling log replay at mount time");
544			break;
545		case Opt_err:
546			btrfs_info(info, "unrecognized rescue option '%s'", p);
547			ret = -EINVAL;
548			goto out;
549		default:
550			break;
551		}
552
553	}
554out:
555	kfree(orig);
556	return ret;
557}
558
559/*
560 * Regular mount options parser.  Everything that is needed only when
561 * reading in a new superblock is parsed here.
562 * XXX JDM: This needs to be cleaned up for remount.
563 */
564int btrfs_parse_options(struct btrfs_fs_info *info, char *options,
565			unsigned long new_flags)
566{
567	substring_t args[MAX_OPT_ARGS];
568	char *p, *num;
569	int intarg;
570	int ret = 0;
571	char *compress_type;
572	bool compress_force = false;
573	enum btrfs_compression_type saved_compress_type;
574	int saved_compress_level;
575	bool saved_compress_force;
576	int no_compress = 0;
577
578	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
579		btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE);
580	else if (btrfs_free_space_cache_v1_active(info)) {
581		if (btrfs_is_zoned(info)) {
582			btrfs_info(info,
583			"zoned: clearing existing space cache");
584			btrfs_set_super_cache_generation(info->super_copy, 0);
585		} else {
586			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
587		}
588	}
589
590	/*
591	 * Even the options are empty, we still need to do extra check
592	 * against new flags
593	 */
594	if (!options)
595		goto check;
596
597	while ((p = strsep(&options, ",")) != NULL) {
598		int token;
599		if (!*p)
600			continue;
601
602		token = match_token(p, tokens, args);
603		switch (token) {
604		case Opt_degraded:
605			btrfs_info(info, "allowing degraded mounts");
606			btrfs_set_opt(info->mount_opt, DEGRADED);
607			break;
608		case Opt_subvol:
609		case Opt_subvol_empty:
610		case Opt_subvolid:
611		case Opt_device:
612			/*
613			 * These are parsed by btrfs_parse_subvol_options or
614			 * btrfs_parse_device_options and can be ignored here.
615			 */
616			break;
617		case Opt_nodatasum:
618			btrfs_set_and_info(info, NODATASUM,
619					   "setting nodatasum");
620			break;
621		case Opt_datasum:
622			if (btrfs_test_opt(info, NODATASUM)) {
623				if (btrfs_test_opt(info, NODATACOW))
624					btrfs_info(info,
625						   "setting datasum, datacow enabled");
626				else
627					btrfs_info(info, "setting datasum");
628			}
629			btrfs_clear_opt(info->mount_opt, NODATACOW);
630			btrfs_clear_opt(info->mount_opt, NODATASUM);
631			break;
632		case Opt_nodatacow:
633			if (!btrfs_test_opt(info, NODATACOW)) {
634				if (!btrfs_test_opt(info, COMPRESS) ||
635				    !btrfs_test_opt(info, FORCE_COMPRESS)) {
636					btrfs_info(info,
637						   "setting nodatacow, compression disabled");
638				} else {
639					btrfs_info(info, "setting nodatacow");
640				}
641			}
642			btrfs_clear_opt(info->mount_opt, COMPRESS);
643			btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
644			btrfs_set_opt(info->mount_opt, NODATACOW);
645			btrfs_set_opt(info->mount_opt, NODATASUM);
646			break;
647		case Opt_datacow:
648			btrfs_clear_and_info(info, NODATACOW,
649					     "setting datacow");
650			break;
651		case Opt_compress_force:
652		case Opt_compress_force_type:
653			compress_force = true;
654			fallthrough;
655		case Opt_compress:
656		case Opt_compress_type:
657			saved_compress_type = btrfs_test_opt(info,
658							     COMPRESS) ?
659				info->compress_type : BTRFS_COMPRESS_NONE;
660			saved_compress_force =
661				btrfs_test_opt(info, FORCE_COMPRESS);
662			saved_compress_level = info->compress_level;
663			if (token == Opt_compress ||
664			    token == Opt_compress_force ||
665			    strncmp(args[0].from, "zlib", 4) == 0) {
666				compress_type = "zlib";
667
668				info->compress_type = BTRFS_COMPRESS_ZLIB;
669				info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
670				/*
671				 * args[0] contains uninitialized data since
672				 * for these tokens we don't expect any
673				 * parameter.
674				 */
675				if (token != Opt_compress &&
676				    token != Opt_compress_force)
677					info->compress_level =
678					  btrfs_compress_str2level(
679							BTRFS_COMPRESS_ZLIB,
680							args[0].from + 4);
681				btrfs_set_opt(info->mount_opt, COMPRESS);
682				btrfs_clear_opt(info->mount_opt, NODATACOW);
683				btrfs_clear_opt(info->mount_opt, NODATASUM);
684				no_compress = 0;
685			} else if (strncmp(args[0].from, "lzo", 3) == 0) {
686				compress_type = "lzo";
687				info->compress_type = BTRFS_COMPRESS_LZO;
688				info->compress_level = 0;
689				btrfs_set_opt(info->mount_opt, COMPRESS);
690				btrfs_clear_opt(info->mount_opt, NODATACOW);
691				btrfs_clear_opt(info->mount_opt, NODATASUM);
692				btrfs_set_fs_incompat(info, COMPRESS_LZO);
693				no_compress = 0;
694			} else if (strncmp(args[0].from, "zstd", 4) == 0) {
695				compress_type = "zstd";
696				info->compress_type = BTRFS_COMPRESS_ZSTD;
697				info->compress_level =
698					btrfs_compress_str2level(
699							 BTRFS_COMPRESS_ZSTD,
700							 args[0].from + 4);
701				btrfs_set_opt(info->mount_opt, COMPRESS);
702				btrfs_clear_opt(info->mount_opt, NODATACOW);
703				btrfs_clear_opt(info->mount_opt, NODATASUM);
704				btrfs_set_fs_incompat(info, COMPRESS_ZSTD);
705				no_compress = 0;
706			} else if (strncmp(args[0].from, "no", 2) == 0) {
707				compress_type = "no";
708				info->compress_level = 0;
709				info->compress_type = 0;
710				btrfs_clear_opt(info->mount_opt, COMPRESS);
711				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
712				compress_force = false;
713				no_compress++;
714			} else {
715				ret = -EINVAL;
716				goto out;
717			}
718
719			if (compress_force) {
720				btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
721			} else {
722				/*
723				 * If we remount from compress-force=xxx to
724				 * compress=xxx, we need clear FORCE_COMPRESS
725				 * flag, otherwise, there is no way for users
726				 * to disable forcible compression separately.
727				 */
728				btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS);
729			}
730			if (no_compress == 1) {
731				btrfs_info(info, "use no compression");
732			} else if ((info->compress_type != saved_compress_type) ||
733				   (compress_force != saved_compress_force) ||
734				   (info->compress_level != saved_compress_level)) {
735				btrfs_info(info, "%s %s compression, level %d",
736					   (compress_force) ? "force" : "use",
737					   compress_type, info->compress_level);
738			}
739			compress_force = false;
740			break;
741		case Opt_ssd:
742			btrfs_set_and_info(info, SSD,
743					   "enabling ssd optimizations");
744			btrfs_clear_opt(info->mount_opt, NOSSD);
745			break;
746		case Opt_ssd_spread:
747			btrfs_set_and_info(info, SSD,
748					   "enabling ssd optimizations");
749			btrfs_set_and_info(info, SSD_SPREAD,
750					   "using spread ssd allocation scheme");
751			btrfs_clear_opt(info->mount_opt, NOSSD);
752			break;
753		case Opt_nossd:
754			btrfs_set_opt(info->mount_opt, NOSSD);
755			btrfs_clear_and_info(info, SSD,
756					     "not using ssd optimizations");
757			fallthrough;
758		case Opt_nossd_spread:
759			btrfs_clear_and_info(info, SSD_SPREAD,
760					     "not using spread ssd allocation scheme");
761			break;
762		case Opt_barrier:
763			btrfs_clear_and_info(info, NOBARRIER,
764					     "turning on barriers");
765			break;
766		case Opt_nobarrier:
767			btrfs_set_and_info(info, NOBARRIER,
768					   "turning off barriers");
769			break;
770		case Opt_thread_pool:
771			ret = match_int(&args[0], &intarg);
772			if (ret) {
773				goto out;
774			} else if (intarg == 0) {
775				ret = -EINVAL;
776				goto out;
777			}
778			info->thread_pool_size = intarg;
779			break;
780		case Opt_max_inline:
781			num = match_strdup(&args[0]);
782			if (num) {
783				info->max_inline = memparse(num, NULL);
784				kfree(num);
785
786				if (info->max_inline) {
787					info->max_inline = min_t(u64,
788						info->max_inline,
789						info->sectorsize);
790				}
791				btrfs_info(info, "max_inline at %llu",
792					   info->max_inline);
793			} else {
794				ret = -ENOMEM;
795				goto out;
796			}
797			break;
798		case Opt_acl:
799#ifdef CONFIG_BTRFS_FS_POSIX_ACL
800			info->sb->s_flags |= SB_POSIXACL;
801			break;
802#else
803			btrfs_err(info, "support for ACL not compiled in!");
804			ret = -EINVAL;
805			goto out;
806#endif
807		case Opt_noacl:
808			info->sb->s_flags &= ~SB_POSIXACL;
809			break;
810		case Opt_notreelog:
811			btrfs_set_and_info(info, NOTREELOG,
812					   "disabling tree log");
813			break;
814		case Opt_treelog:
815			btrfs_clear_and_info(info, NOTREELOG,
816					     "enabling tree log");
817			break;
818		case Opt_norecovery:
819		case Opt_nologreplay:
820			btrfs_warn(info,
821		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
822			btrfs_set_and_info(info, NOLOGREPLAY,
823					   "disabling log replay at mount time");
824			break;
825		case Opt_flushoncommit:
826			btrfs_set_and_info(info, FLUSHONCOMMIT,
827					   "turning on flush-on-commit");
828			break;
829		case Opt_noflushoncommit:
830			btrfs_clear_and_info(info, FLUSHONCOMMIT,
831					     "turning off flush-on-commit");
832			break;
833		case Opt_ratio:
834			ret = match_int(&args[0], &intarg);
835			if (ret)
836				goto out;
837			info->metadata_ratio = intarg;
838			btrfs_info(info, "metadata ratio %u",
839				   info->metadata_ratio);
840			break;
841		case Opt_discard:
842		case Opt_discard_mode:
843			if (token == Opt_discard ||
844			    strcmp(args[0].from, "sync") == 0) {
845				btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC);
846				btrfs_set_and_info(info, DISCARD_SYNC,
847						   "turning on sync discard");
848			} else if (strcmp(args[0].from, "async") == 0) {
849				btrfs_clear_opt(info->mount_opt, DISCARD_SYNC);
850				btrfs_set_and_info(info, DISCARD_ASYNC,
851						   "turning on async discard");
852			} else {
853				ret = -EINVAL;
854				goto out;
855			}
856			break;
857		case Opt_nodiscard:
858			btrfs_clear_and_info(info, DISCARD_SYNC,
859					     "turning off discard");
860			btrfs_clear_and_info(info, DISCARD_ASYNC,
861					     "turning off async discard");
862			break;
863		case Opt_space_cache:
864		case Opt_space_cache_version:
865			if (token == Opt_space_cache ||
866			    strcmp(args[0].from, "v1") == 0) {
867				btrfs_clear_opt(info->mount_opt,
868						FREE_SPACE_TREE);
869				btrfs_set_and_info(info, SPACE_CACHE,
870					   "enabling disk space caching");
871			} else if (strcmp(args[0].from, "v2") == 0) {
872				btrfs_clear_opt(info->mount_opt,
873						SPACE_CACHE);
874				btrfs_set_and_info(info, FREE_SPACE_TREE,
875						   "enabling free space tree");
876			} else {
877				ret = -EINVAL;
878				goto out;
879			}
880			break;
881		case Opt_rescan_uuid_tree:
882			btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE);
883			break;
884		case Opt_no_space_cache:
885			if (btrfs_test_opt(info, SPACE_CACHE)) {
886				btrfs_clear_and_info(info, SPACE_CACHE,
887					     "disabling disk space caching");
888			}
889			if (btrfs_test_opt(info, FREE_SPACE_TREE)) {
890				btrfs_clear_and_info(info, FREE_SPACE_TREE,
891					     "disabling free space tree");
892			}
893			break;
894		case Opt_inode_cache:
895		case Opt_noinode_cache:
896			btrfs_warn(info,
897	"the 'inode_cache' option is deprecated and has no effect since 5.11");
898			break;
899		case Opt_clear_cache:
900			btrfs_set_and_info(info, CLEAR_CACHE,
901					   "force clearing of disk cache");
902			break;
903		case Opt_user_subvol_rm_allowed:
904			btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED);
905			break;
906		case Opt_enospc_debug:
907			btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG);
908			break;
909		case Opt_noenospc_debug:
910			btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG);
911			break;
912		case Opt_defrag:
913			btrfs_set_and_info(info, AUTO_DEFRAG,
914					   "enabling auto defrag");
915			break;
916		case Opt_nodefrag:
917			btrfs_clear_and_info(info, AUTO_DEFRAG,
918					     "disabling auto defrag");
919			break;
920		case Opt_recovery:
921		case Opt_usebackuproot:
922			btrfs_warn(info,
923			"'%s' is deprecated, use 'rescue=usebackuproot' instead",
924				   token == Opt_recovery ? "recovery" :
925				   "usebackuproot");
926			btrfs_info(info,
927				   "trying to use backup root at mount time");
928			btrfs_set_opt(info->mount_opt, USEBACKUPROOT);
929			break;
930		case Opt_skip_balance:
931			btrfs_set_opt(info->mount_opt, SKIP_BALANCE);
932			break;
933#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
934		case Opt_check_integrity_including_extent_data:
935			btrfs_info(info,
936				   "enabling check integrity including extent data");
937			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY_DATA);
938			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
939			break;
940		case Opt_check_integrity:
941			btrfs_info(info, "enabling check integrity");
942			btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY);
943			break;
944		case Opt_check_integrity_print_mask:
945			ret = match_int(&args[0], &intarg);
946			if (ret)
947				goto out;
948			info->check_integrity_print_mask = intarg;
949			btrfs_info(info, "check_integrity_print_mask 0x%x",
950				   info->check_integrity_print_mask);
951			break;
952#else
953		case Opt_check_integrity_including_extent_data:
954		case Opt_check_integrity:
955		case Opt_check_integrity_print_mask:
956			btrfs_err(info,
957				  "support for check_integrity* not compiled in!");
958			ret = -EINVAL;
959			goto out;
960#endif
961		case Opt_fatal_errors:
962			if (strcmp(args[0].from, "panic") == 0)
963				btrfs_set_opt(info->mount_opt,
964					      PANIC_ON_FATAL_ERROR);
965			else if (strcmp(args[0].from, "bug") == 0)
966				btrfs_clear_opt(info->mount_opt,
967					      PANIC_ON_FATAL_ERROR);
968			else {
969				ret = -EINVAL;
970				goto out;
971			}
972			break;
973		case Opt_commit_interval:
974			intarg = 0;
975			ret = match_int(&args[0], &intarg);
976			if (ret)
977				goto out;
978			if (intarg == 0) {
979				btrfs_info(info,
980					   "using default commit interval %us",
981					   BTRFS_DEFAULT_COMMIT_INTERVAL);
982				intarg = BTRFS_DEFAULT_COMMIT_INTERVAL;
983			} else if (intarg > 300) {
984				btrfs_warn(info, "excessive commit interval %d",
985					   intarg);
986			}
987			info->commit_interval = intarg;
988			break;
989		case Opt_rescue:
990			ret = parse_rescue_options(info, args[0].from);
991			if (ret < 0)
992				goto out;
993			break;
994#ifdef CONFIG_BTRFS_DEBUG
995		case Opt_fragment_all:
996			btrfs_info(info, "fragmenting all space");
997			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
998			btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA);
999			break;
1000		case Opt_fragment_metadata:
1001			btrfs_info(info, "fragmenting metadata");
1002			btrfs_set_opt(info->mount_opt,
1003				      FRAGMENT_METADATA);
1004			break;
1005		case Opt_fragment_data:
1006			btrfs_info(info, "fragmenting data");
1007			btrfs_set_opt(info->mount_opt, FRAGMENT_DATA);
1008			break;
1009#endif
1010#ifdef CONFIG_BTRFS_FS_REF_VERIFY
1011		case Opt_ref_verify:
1012			btrfs_info(info, "doing ref verification");
1013			btrfs_set_opt(info->mount_opt, REF_VERIFY);
1014			break;
1015#endif
1016		case Opt_err:
1017			btrfs_err(info, "unrecognized mount option '%s'", p);
1018			ret = -EINVAL;
1019			goto out;
1020		default:
1021			break;
1022		}
1023	}
1024check:
1025	/* We're read-only, don't have to check. */
1026	if (new_flags & SB_RDONLY)
1027		goto out;
1028
1029	if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
1030	    check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
1031	    check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums"))
1032		ret = -EINVAL;
1033out:
1034	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
1035	    !btrfs_test_opt(info, FREE_SPACE_TREE) &&
1036	    !btrfs_test_opt(info, CLEAR_CACHE)) {
1037		btrfs_err(info, "cannot disable free space tree");
1038		ret = -EINVAL;
1039
1040	}
1041	if (!ret)
1042		ret = btrfs_check_mountopts_zoned(info);
1043	if (!ret && btrfs_test_opt(info, SPACE_CACHE))
1044		btrfs_info(info, "disk space caching is enabled");
1045	if (!ret && btrfs_test_opt(info, FREE_SPACE_TREE))
1046		btrfs_info(info, "using free space tree");
1047	return ret;
1048}
1049
1050/*
1051 * Parse mount options that are required early in the mount process.
1052 *
1053 * All other options will be parsed on much later in the mount process and
1054 * only when we need to allocate a new super block.
1055 */
1056static int btrfs_parse_device_options(const char *options, fmode_t flags,
1057				      void *holder)
1058{
1059	substring_t args[MAX_OPT_ARGS];
1060	char *device_name, *opts, *orig, *p;
1061	struct btrfs_device *device = NULL;
1062	int error = 0;
1063
1064	lockdep_assert_held(&uuid_mutex);
1065
1066	if (!options)
1067		return 0;
1068
1069	/*
1070	 * strsep changes the string, duplicate it because btrfs_parse_options
1071	 * gets called later
1072	 */
1073	opts = kstrdup(options, GFP_KERNEL);
1074	if (!opts)
1075		return -ENOMEM;
1076	orig = opts;
1077
1078	while ((p = strsep(&opts, ",")) != NULL) {
1079		int token;
1080
1081		if (!*p)
1082			continue;
1083
1084		token = match_token(p, tokens, args);
1085		if (token == Opt_device) {
1086			device_name = match_strdup(&args[0]);
1087			if (!device_name) {
1088				error = -ENOMEM;
1089				goto out;
1090			}
1091			device = btrfs_scan_one_device(device_name, flags,
1092					holder);
1093			kfree(device_name);
1094			if (IS_ERR(device)) {
1095				error = PTR_ERR(device);
1096				goto out;
1097			}
1098		}
1099	}
1100
1101out:
1102	kfree(orig);
1103	return error;
1104}
1105
1106/*
1107 * Parse mount options that are related to subvolume id
1108 *
1109 * The value is later passed to mount_subvol()
1110 */
1111static int btrfs_parse_subvol_options(const char *options, char **subvol_name,
1112		u64 *subvol_objectid)
1113{
1114	substring_t args[MAX_OPT_ARGS];
1115	char *opts, *orig, *p;
1116	int error = 0;
1117	u64 subvolid;
1118
1119	if (!options)
1120		return 0;
1121
1122	/*
1123	 * strsep changes the string, duplicate it because
1124	 * btrfs_parse_device_options gets called later
1125	 */
1126	opts = kstrdup(options, GFP_KERNEL);
1127	if (!opts)
1128		return -ENOMEM;
1129	orig = opts;
1130
1131	while ((p = strsep(&opts, ",")) != NULL) {
1132		int token;
1133		if (!*p)
1134			continue;
1135
1136		token = match_token(p, tokens, args);
1137		switch (token) {
1138		case Opt_subvol:
1139			kfree(*subvol_name);
1140			*subvol_name = match_strdup(&args[0]);
1141			if (!*subvol_name) {
1142				error = -ENOMEM;
1143				goto out;
1144			}
1145			break;
1146		case Opt_subvolid:
1147			error = match_u64(&args[0], &subvolid);
1148			if (error)
1149				goto out;
1150
1151			/* we want the original fs_tree */
1152			if (subvolid == 0)
1153				subvolid = BTRFS_FS_TREE_OBJECTID;
1154
1155			*subvol_objectid = subvolid;
1156			break;
1157		default:
1158			break;
1159		}
1160	}
1161
1162out:
1163	kfree(orig);
1164	return error;
1165}
1166
1167char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
1168					  u64 subvol_objectid)
1169{
1170	struct btrfs_root *root = fs_info->tree_root;
1171	struct btrfs_root *fs_root = NULL;
1172	struct btrfs_root_ref *root_ref;
1173	struct btrfs_inode_ref *inode_ref;
1174	struct btrfs_key key;
1175	struct btrfs_path *path = NULL;
1176	char *name = NULL, *ptr;
1177	u64 dirid;
1178	int len;
1179	int ret;
1180
1181	path = btrfs_alloc_path();
1182	if (!path) {
1183		ret = -ENOMEM;
1184		goto err;
1185	}
1186
1187	name = kmalloc(PATH_MAX, GFP_KERNEL);
1188	if (!name) {
1189		ret = -ENOMEM;
1190		goto err;
1191	}
1192	ptr = name + PATH_MAX - 1;
1193	ptr[0] = '\0';
1194
1195	/*
1196	 * Walk up the subvolume trees in the tree of tree roots by root
1197	 * backrefs until we hit the top-level subvolume.
1198	 */
1199	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
1200		key.objectid = subvol_objectid;
1201		key.type = BTRFS_ROOT_BACKREF_KEY;
1202		key.offset = (u64)-1;
1203
1204		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1205		if (ret < 0) {
1206			goto err;
1207		} else if (ret > 0) {
1208			ret = btrfs_previous_item(root, path, subvol_objectid,
1209						  BTRFS_ROOT_BACKREF_KEY);
1210			if (ret < 0) {
1211				goto err;
1212			} else if (ret > 0) {
1213				ret = -ENOENT;
1214				goto err;
1215			}
1216		}
1217
1218		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1219		subvol_objectid = key.offset;
1220
1221		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
1222					  struct btrfs_root_ref);
1223		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
1224		ptr -= len + 1;
1225		if (ptr < name) {
1226			ret = -ENAMETOOLONG;
1227			goto err;
1228		}
1229		read_extent_buffer(path->nodes[0], ptr + 1,
1230				   (unsigned long)(root_ref + 1), len);
1231		ptr[0] = '/';
1232		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
1233		btrfs_release_path(path);
1234
1235		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
1236		if (IS_ERR(fs_root)) {
1237			ret = PTR_ERR(fs_root);
1238			fs_root = NULL;
1239			goto err;
1240		}
1241
1242		/*
1243		 * Walk up the filesystem tree by inode refs until we hit the
1244		 * root directory.
1245		 */
1246		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
1247			key.objectid = dirid;
1248			key.type = BTRFS_INODE_REF_KEY;
1249			key.offset = (u64)-1;
1250
1251			ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1252			if (ret < 0) {
1253				goto err;
1254			} else if (ret > 0) {
1255				ret = btrfs_previous_item(fs_root, path, dirid,
1256							  BTRFS_INODE_REF_KEY);
1257				if (ret < 0) {
1258					goto err;
1259				} else if (ret > 0) {
1260					ret = -ENOENT;
1261					goto err;
1262				}
1263			}
1264
1265			btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1266			dirid = key.offset;
1267
1268			inode_ref = btrfs_item_ptr(path->nodes[0],
1269						   path->slots[0],
1270						   struct btrfs_inode_ref);
1271			len = btrfs_inode_ref_name_len(path->nodes[0],
1272						       inode_ref);
1273			ptr -= len + 1;
1274			if (ptr < name) {
1275				ret = -ENAMETOOLONG;
1276				goto err;
1277			}
1278			read_extent_buffer(path->nodes[0], ptr + 1,
1279					   (unsigned long)(inode_ref + 1), len);
1280			ptr[0] = '/';
1281			btrfs_release_path(path);
1282		}
1283		btrfs_put_root(fs_root);
1284		fs_root = NULL;
1285	}
1286
1287	btrfs_free_path(path);
1288	if (ptr == name + PATH_MAX - 1) {
1289		name[0] = '/';
1290		name[1] = '\0';
1291	} else {
1292		memmove(name, ptr, name + PATH_MAX - ptr);
1293	}
1294	return name;
1295
1296err:
1297	btrfs_put_root(fs_root);
1298	btrfs_free_path(path);
1299	kfree(name);
1300	return ERR_PTR(ret);
1301}
1302
1303static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
1304{
1305	struct btrfs_root *root = fs_info->tree_root;
1306	struct btrfs_dir_item *di;
1307	struct btrfs_path *path;
1308	struct btrfs_key location;
1309	u64 dir_id;
1310
1311	path = btrfs_alloc_path();
1312	if (!path)
1313		return -ENOMEM;
1314
1315	/*
1316	 * Find the "default" dir item which points to the root item that we
1317	 * will mount by default if we haven't been given a specific subvolume
1318	 * to mount.
1319	 */
1320	dir_id = btrfs_super_root_dir(fs_info->super_copy);
1321	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
1322	if (IS_ERR(di)) {
1323		btrfs_free_path(path);
1324		return PTR_ERR(di);
1325	}
1326	if (!di) {
1327		/*
1328		 * Ok the default dir item isn't there.  This is weird since
1329		 * it's always been there, but don't freak out, just try and
1330		 * mount the top-level subvolume.
1331		 */
1332		btrfs_free_path(path);
1333		*objectid = BTRFS_FS_TREE_OBJECTID;
1334		return 0;
1335	}
1336
1337	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
1338	btrfs_free_path(path);
1339	*objectid = location.objectid;
1340	return 0;
1341}
1342
1343static int btrfs_fill_super(struct super_block *sb,
1344			    struct btrfs_fs_devices *fs_devices,
1345			    void *data)
1346{
1347	struct inode *inode;
1348	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1349	int err;
1350
1351	sb->s_maxbytes = MAX_LFS_FILESIZE;
1352	sb->s_magic = BTRFS_SUPER_MAGIC;
1353	sb->s_op = &btrfs_super_ops;
1354	sb->s_d_op = &btrfs_dentry_operations;
1355	sb->s_export_op = &btrfs_export_ops;
1356#ifdef CONFIG_FS_VERITY
1357	sb->s_vop = &btrfs_verityops;
1358#endif
1359	sb->s_xattr = btrfs_xattr_handlers;
1360	sb->s_time_gran = 1;
1361#ifdef CONFIG_BTRFS_FS_POSIX_ACL
1362	sb->s_flags |= SB_POSIXACL;
1363#endif
1364	sb->s_flags |= SB_I_VERSION;
1365	sb->s_iflags |= SB_I_CGROUPWB;
1366
1367	err = super_setup_bdi(sb);
1368	if (err) {
1369		btrfs_err(fs_info, "super_setup_bdi failed");
1370		return err;
1371	}
1372
1373	err = open_ctree(sb, fs_devices, (char *)data);
1374	if (err) {
1375		btrfs_err(fs_info, "open_ctree failed");
1376		return err;
1377	}
1378
1379	inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1380	if (IS_ERR(inode)) {
1381		err = PTR_ERR(inode);
1382		goto fail_close;
1383	}
1384
1385	sb->s_root = d_make_root(inode);
1386	if (!sb->s_root) {
1387		err = -ENOMEM;
1388		goto fail_close;
1389	}
1390
1391	cleancache_init_fs(sb);
1392	sb->s_flags |= SB_ACTIVE;
1393	return 0;
1394
1395fail_close:
1396	close_ctree(fs_info);
1397	return err;
1398}
1399
1400int btrfs_sync_fs(struct super_block *sb, int wait)
1401{
1402	struct btrfs_trans_handle *trans;
1403	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1404	struct btrfs_root *root = fs_info->tree_root;
1405
1406	trace_btrfs_sync_fs(fs_info, wait);
1407
1408	if (!wait) {
1409		filemap_flush(fs_info->btree_inode->i_mapping);
1410		return 0;
1411	}
1412
1413	btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1);
1414
1415	trans = btrfs_attach_transaction_barrier(root);
1416	if (IS_ERR(trans)) {
1417		/* no transaction, don't bother */
1418		if (PTR_ERR(trans) == -ENOENT) {
1419			/*
1420			 * Exit unless we have some pending changes
1421			 * that need to go through commit
1422			 */
1423			if (fs_info->pending_changes == 0)
1424				return 0;
1425			/*
1426			 * A non-blocking test if the fs is frozen. We must not
1427			 * start a new transaction here otherwise a deadlock
1428			 * happens. The pending operations are delayed to the
1429			 * next commit after thawing.
1430			 */
1431			if (sb_start_write_trylock(sb))
1432				sb_end_write(sb);
1433			else
1434				return 0;
1435			trans = btrfs_start_transaction(root, 0);
1436		}
1437		if (IS_ERR(trans))
1438			return PTR_ERR(trans);
1439	}
1440	return btrfs_commit_transaction(trans);
1441}
1442
1443static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1444{
1445	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1446	*printed = true;
1447}
1448
1449static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1450{
1451	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1452	const char *compress_type;
1453	const char *subvol_name;
1454	bool printed = false;
1455
1456	if (btrfs_test_opt(info, DEGRADED))
1457		seq_puts(seq, ",degraded");
1458	if (btrfs_test_opt(info, NODATASUM))
1459		seq_puts(seq, ",nodatasum");
1460	if (btrfs_test_opt(info, NODATACOW))
1461		seq_puts(seq, ",nodatacow");
1462	if (btrfs_test_opt(info, NOBARRIER))
1463		seq_puts(seq, ",nobarrier");
1464	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1465		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1466	if (info->thread_pool_size !=  min_t(unsigned long,
1467					     num_online_cpus() + 2, 8))
1468		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1469	if (btrfs_test_opt(info, COMPRESS)) {
1470		compress_type = btrfs_compress_type2str(info->compress_type);
1471		if (btrfs_test_opt(info, FORCE_COMPRESS))
1472			seq_printf(seq, ",compress-force=%s", compress_type);
1473		else
1474			seq_printf(seq, ",compress=%s", compress_type);
1475		if (info->compress_level)
1476			seq_printf(seq, ":%d", info->compress_level);
1477	}
1478	if (btrfs_test_opt(info, NOSSD))
1479		seq_puts(seq, ",nossd");
1480	if (btrfs_test_opt(info, SSD_SPREAD))
1481		seq_puts(seq, ",ssd_spread");
1482	else if (btrfs_test_opt(info, SSD))
1483		seq_puts(seq, ",ssd");
1484	if (btrfs_test_opt(info, NOTREELOG))
1485		seq_puts(seq, ",notreelog");
1486	if (btrfs_test_opt(info, NOLOGREPLAY))
1487		print_rescue_option(seq, "nologreplay", &printed);
1488	if (btrfs_test_opt(info, USEBACKUPROOT))
1489		print_rescue_option(seq, "usebackuproot", &printed);
1490	if (btrfs_test_opt(info, IGNOREBADROOTS))
1491		print_rescue_option(seq, "ignorebadroots", &printed);
1492	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1493		print_rescue_option(seq, "ignoredatacsums", &printed);
1494	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1495		seq_puts(seq, ",flushoncommit");
1496	if (btrfs_test_opt(info, DISCARD_SYNC))
1497		seq_puts(seq, ",discard");
1498	if (btrfs_test_opt(info, DISCARD_ASYNC))
1499		seq_puts(seq, ",discard=async");
1500	if (!(info->sb->s_flags & SB_POSIXACL))
1501		seq_puts(seq, ",noacl");
1502	if (btrfs_free_space_cache_v1_active(info))
1503		seq_puts(seq, ",space_cache");
1504	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1505		seq_puts(seq, ",space_cache=v2");
1506	else
1507		seq_puts(seq, ",nospace_cache");
1508	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1509		seq_puts(seq, ",rescan_uuid_tree");
1510	if (btrfs_test_opt(info, CLEAR_CACHE))
1511		seq_puts(seq, ",clear_cache");
1512	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1513		seq_puts(seq, ",user_subvol_rm_allowed");
1514	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1515		seq_puts(seq, ",enospc_debug");
1516	if (btrfs_test_opt(info, AUTO_DEFRAG))
1517		seq_puts(seq, ",autodefrag");
1518	if (btrfs_test_opt(info, SKIP_BALANCE))
1519		seq_puts(seq, ",skip_balance");
1520#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
1521	if (btrfs_test_opt(info, CHECK_INTEGRITY_DATA))
1522		seq_puts(seq, ",check_int_data");
1523	else if (btrfs_test_opt(info, CHECK_INTEGRITY))
1524		seq_puts(seq, ",check_int");
1525	if (info->check_integrity_print_mask)
1526		seq_printf(seq, ",check_int_print_mask=%d",
1527				info->check_integrity_print_mask);
1528#endif
1529	if (info->metadata_ratio)
1530		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1531	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1532		seq_puts(seq, ",fatal_errors=panic");
1533	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1534		seq_printf(seq, ",commit=%u", info->commit_interval);
1535#ifdef CONFIG_BTRFS_DEBUG
1536	if (btrfs_test_opt(info, FRAGMENT_DATA))
1537		seq_puts(seq, ",fragment=data");
1538	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1539		seq_puts(seq, ",fragment=metadata");
1540#endif
1541	if (btrfs_test_opt(info, REF_VERIFY))
1542		seq_puts(seq, ",ref_verify");
1543	seq_printf(seq, ",subvolid=%llu",
1544		  BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1545	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1546			BTRFS_I(d_inode(dentry))->root->root_key.objectid);
1547	if (!IS_ERR(subvol_name)) {
1548		seq_puts(seq, ",subvol=");
1549		seq_escape(seq, subvol_name, " \t\n\\");
1550		kfree(subvol_name);
1551	}
1552	return 0;
1553}
1554
1555static int btrfs_test_super(struct super_block *s, void *data)
1556{
1557	struct btrfs_fs_info *p = data;
1558	struct btrfs_fs_info *fs_info = btrfs_sb(s);
1559
1560	return fs_info->fs_devices == p->fs_devices;
1561}
1562
1563static int btrfs_set_super(struct super_block *s, void *data)
1564{
1565	int err = set_anon_super(s, data);
1566	if (!err)
1567		s->s_fs_info = data;
1568	return err;
1569}
1570
1571/*
1572 * subvolumes are identified by ino 256
1573 */
1574static inline int is_subvolume_inode(struct inode *inode)
1575{
1576	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1577		return 1;
1578	return 0;
1579}
1580
1581static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1582				   struct vfsmount *mnt)
1583{
1584	struct dentry *root;
1585	int ret;
1586
1587	if (!subvol_name) {
1588		if (!subvol_objectid) {
1589			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1590							  &subvol_objectid);
1591			if (ret) {
1592				root = ERR_PTR(ret);
1593				goto out;
1594			}
1595		}
1596		subvol_name = btrfs_get_subvol_name_from_objectid(
1597					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1598		if (IS_ERR(subvol_name)) {
1599			root = ERR_CAST(subvol_name);
1600			subvol_name = NULL;
1601			goto out;
1602		}
1603
1604	}
1605
1606	root = mount_subtree(mnt, subvol_name);
1607	/* mount_subtree() drops our reference on the vfsmount. */
1608	mnt = NULL;
1609
1610	if (!IS_ERR(root)) {
1611		struct super_block *s = root->d_sb;
1612		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1613		struct inode *root_inode = d_inode(root);
1614		u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid;
1615
1616		ret = 0;
1617		if (!is_subvolume_inode(root_inode)) {
1618			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1619			       subvol_name);
1620			ret = -EINVAL;
1621		}
1622		if (subvol_objectid && root_objectid != subvol_objectid) {
1623			/*
1624			 * This will also catch a race condition where a
1625			 * subvolume which was passed by ID is renamed and
1626			 * another subvolume is renamed over the old location.
1627			 */
1628			btrfs_err(fs_info,
1629				  "subvol '%s' does not match subvolid %llu",
1630				  subvol_name, subvol_objectid);
1631			ret = -EINVAL;
1632		}
1633		if (ret) {
1634			dput(root);
1635			root = ERR_PTR(ret);
1636			deactivate_locked_super(s);
1637		}
1638	}
1639
1640out:
1641	mntput(mnt);
1642	kfree(subvol_name);
1643	return root;
1644}
1645
1646/*
1647 * Find a superblock for the given device / mount point.
1648 *
1649 * Note: This is based on mount_bdev from fs/super.c with a few additions
1650 *       for multiple device setup.  Make sure to keep it in sync.
1651 */
1652static struct dentry *btrfs_mount_root(struct file_system_type *fs_type,
1653		int flags, const char *device_name, void *data)
1654{
1655	struct block_device *bdev = NULL;
1656	struct super_block *s;
1657	struct btrfs_device *device = NULL;
1658	struct btrfs_fs_devices *fs_devices = NULL;
1659	struct btrfs_fs_info *fs_info = NULL;
1660	void *new_sec_opts = NULL;
1661	fmode_t mode = FMODE_READ;
1662	int error = 0;
1663
1664	if (!(flags & SB_RDONLY))
1665		mode |= FMODE_WRITE;
1666
1667	if (data) {
1668		error = security_sb_eat_lsm_opts(data, &new_sec_opts);
1669		if (error)
1670			return ERR_PTR(error);
1671	}
1672
1673	/*
1674	 * Setup a dummy root and fs_info for test/set super.  This is because
1675	 * we don't actually fill this stuff out until open_ctree, but we need
1676	 * then open_ctree will properly initialize the file system specific
1677	 * settings later.  btrfs_init_fs_info initializes the static elements
1678	 * of the fs_info (locks and such) to make cleanup easier if we find a
1679	 * superblock with our given fs_devices later on at sget() time.
1680	 */
1681	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
1682	if (!fs_info) {
1683		error = -ENOMEM;
1684		goto error_sec_opts;
1685	}
1686	btrfs_init_fs_info(fs_info);
1687
1688	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1689	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
1690	if (!fs_info->super_copy || !fs_info->super_for_commit) {
1691		error = -ENOMEM;
1692		goto error_fs_info;
1693	}
1694
1695	mutex_lock(&uuid_mutex);
1696	error = btrfs_parse_device_options(data, mode, fs_type);
1697	if (error) {
1698		mutex_unlock(&uuid_mutex);
1699		goto error_fs_info;
1700	}
1701
1702	device = btrfs_scan_one_device(device_name, mode, fs_type);
1703	if (IS_ERR(device)) {
1704		mutex_unlock(&uuid_mutex);
1705		error = PTR_ERR(device);
1706		goto error_fs_info;
1707	}
1708
1709	fs_devices = device->fs_devices;
1710	fs_info->fs_devices = fs_devices;
1711
1712	error = btrfs_open_devices(fs_devices, mode, fs_type);
1713	mutex_unlock(&uuid_mutex);
1714	if (error)
1715		goto error_fs_info;
1716
1717	if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1718		error = -EACCES;
1719		goto error_close_devices;
1720	}
1721
1722	bdev = fs_devices->latest_bdev;
1723	s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC,
1724		 fs_info);
1725	if (IS_ERR(s)) {
1726		error = PTR_ERR(s);
1727		goto error_close_devices;
1728	}
1729
1730	if (s->s_root) {
1731		btrfs_close_devices(fs_devices);
1732		btrfs_free_fs_info(fs_info);
1733		if ((flags ^ s->s_flags) & SB_RDONLY)
1734			error = -EBUSY;
1735	} else {
1736		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
1737		btrfs_sb(s)->bdev_holder = fs_type;
1738		if (!strstr(crc32c_impl(), "generic"))
1739			set_bit(BTRFS_FS_CSUM_IMPL_FAST, &fs_info->flags);
1740		error = btrfs_fill_super(s, fs_devices, data);
1741	}
1742	if (!error)
1743		error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL);
1744	security_free_mnt_opts(&new_sec_opts);
1745	if (error) {
1746		deactivate_locked_super(s);
1747		return ERR_PTR(error);
1748	}
1749
1750	return dget(s->s_root);
1751
1752error_close_devices:
1753	btrfs_close_devices(fs_devices);
1754error_fs_info:
1755	btrfs_free_fs_info(fs_info);
1756error_sec_opts:
1757	security_free_mnt_opts(&new_sec_opts);
1758	return ERR_PTR(error);
1759}
1760
1761/*
1762 * Mount function which is called by VFS layer.
1763 *
1764 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree()
1765 * which needs vfsmount* of device's root (/).  This means device's root has to
1766 * be mounted internally in any case.
1767 *
1768 * Operation flow:
1769 *   1. Parse subvol id related options for later use in mount_subvol().
1770 *
1771 *   2. Mount device's root (/) by calling vfs_kern_mount().
1772 *
1773 *      NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the
1774 *      first place. In order to avoid calling btrfs_mount() again, we use
1775 *      different file_system_type which is not registered to VFS by
1776 *      register_filesystem() (btrfs_root_fs_type). As a result,
1777 *      btrfs_mount_root() is called. The return value will be used by
1778 *      mount_subtree() in mount_subvol().
1779 *
1780 *   3. Call mount_subvol() to get the dentry of subvolume. Since there is
1781 *      "btrfs subvolume set-default", mount_subvol() is called always.
1782 */
1783static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
1784		const char *device_name, void *data)
1785{
1786	struct vfsmount *mnt_root;
1787	struct dentry *root;
1788	char *subvol_name = NULL;
1789	u64 subvol_objectid = 0;
1790	int error = 0;
1791
1792	error = btrfs_parse_subvol_options(data, &subvol_name,
1793					&subvol_objectid);
1794	if (error) {
1795		kfree(subvol_name);
1796		return ERR_PTR(error);
1797	}
1798
1799	/* mount device's root (/) */
1800	mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data);
1801	if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) {
1802		if (flags & SB_RDONLY) {
1803			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1804				flags & ~SB_RDONLY, device_name, data);
1805		} else {
1806			mnt_root = vfs_kern_mount(&btrfs_root_fs_type,
1807				flags | SB_RDONLY, device_name, data);
1808			if (IS_ERR(mnt_root)) {
1809				root = ERR_CAST(mnt_root);
1810				kfree(subvol_name);
1811				goto out;
1812			}
1813
1814			down_write(&mnt_root->mnt_sb->s_umount);
1815			error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL);
1816			up_write(&mnt_root->mnt_sb->s_umount);
1817			if (error < 0) {
1818				root = ERR_PTR(error);
1819				mntput(mnt_root);
1820				kfree(subvol_name);
1821				goto out;
1822			}
1823		}
1824	}
1825	if (IS_ERR(mnt_root)) {
1826		root = ERR_CAST(mnt_root);
1827		kfree(subvol_name);
1828		goto out;
1829	}
1830
1831	/* mount_subvol() will free subvol_name and mnt_root */
1832	root = mount_subvol(subvol_name, subvol_objectid, mnt_root);
1833
1834out:
1835	return root;
1836}
1837
1838static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1839				     u32 new_pool_size, u32 old_pool_size)
1840{
1841	if (new_pool_size == old_pool_size)
1842		return;
1843
1844	fs_info->thread_pool_size = new_pool_size;
1845
1846	btrfs_info(fs_info, "resize thread pool %d -> %d",
1847	       old_pool_size, new_pool_size);
1848
1849	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1850	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1851	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1852	btrfs_workqueue_set_max(fs_info->endio_workers, new_pool_size);
1853	btrfs_workqueue_set_max(fs_info->endio_meta_workers, new_pool_size);
1854	btrfs_workqueue_set_max(fs_info->endio_meta_write_workers,
1855				new_pool_size);
1856	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1857	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1858	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1859	btrfs_workqueue_set_max(fs_info->readahead_workers, new_pool_size);
1860	btrfs_workqueue_set_max(fs_info->scrub_wr_completion_workers,
1861				new_pool_size);
1862}
1863
1864static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1865				       unsigned long old_opts, int flags)
1866{
1867	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1868	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1869	     (flags & SB_RDONLY))) {
1870		/* wait for any defraggers to finish */
1871		wait_event(fs_info->transaction_wait,
1872			   (atomic_read(&fs_info->defrag_running) == 0));
1873		if (flags & SB_RDONLY)
1874			sync_filesystem(fs_info->sb);
1875	}
1876}
1877
1878static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1879					 unsigned long old_opts)
1880{
1881	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1882
1883	/*
1884	 * We need to cleanup all defragable inodes if the autodefragment is
1885	 * close or the filesystem is read only.
1886	 */
1887	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1888	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1889		btrfs_cleanup_defrag_inodes(fs_info);
1890	}
1891
1892	/* If we toggled discard async */
1893	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1894	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1895		btrfs_discard_resume(fs_info);
1896	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1897		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1898		btrfs_discard_cleanup(fs_info);
1899
1900	/* If we toggled space cache */
1901	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1902		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1903}
1904
1905static int btrfs_remount(struct super_block *sb, int *flags, char *data)
1906{
1907	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1908	unsigned old_flags = sb->s_flags;
1909	unsigned long old_opts = fs_info->mount_opt;
1910	unsigned long old_compress_type = fs_info->compress_type;
1911	u64 old_max_inline = fs_info->max_inline;
1912	u32 old_thread_pool_size = fs_info->thread_pool_size;
1913	u32 old_metadata_ratio = fs_info->metadata_ratio;
1914	int ret;
1915
1916	sync_filesystem(sb);
1917	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1918
1919	if (data) {
1920		void *new_sec_opts = NULL;
1921
1922		ret = security_sb_eat_lsm_opts(data, &new_sec_opts);
1923		if (!ret)
1924			ret = security_sb_remount(sb, new_sec_opts);
1925		security_free_mnt_opts(&new_sec_opts);
1926		if (ret)
1927			goto restore;
1928	}
1929
1930	ret = btrfs_parse_options(fs_info, data, *flags);
1931	if (ret)
1932		goto restore;
1933
1934	btrfs_remount_begin(fs_info, old_opts, *flags);
1935	btrfs_resize_thread_pool(fs_info,
1936		fs_info->thread_pool_size, old_thread_pool_size);
1937
1938	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1939	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1940	    (!sb_rdonly(sb) || (*flags & SB_RDONLY))) {
1941		btrfs_warn(fs_info,
1942		"remount supports changing free space tree only from ro to rw");
1943		/* Make sure free space cache options match the state on disk */
1944		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1945			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1946			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1947		}
1948		if (btrfs_free_space_cache_v1_active(fs_info)) {
1949			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1950			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1951		}
1952	}
1953
1954	if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb))
1955		goto out;
1956
1957	if (*flags & SB_RDONLY) {
1958		/*
1959		 * this also happens on 'umount -rf' or on shutdown, when
1960		 * the filesystem is busy.
1961		 */
1962		cancel_work_sync(&fs_info->async_reclaim_work);
1963		cancel_work_sync(&fs_info->async_data_reclaim_work);
1964
1965		btrfs_discard_cleanup(fs_info);
1966
1967		/* wait for the uuid_scan task to finish */
1968		down(&fs_info->uuid_tree_rescan_sem);
1969		/* avoid complains from lockdep et al. */
1970		up(&fs_info->uuid_tree_rescan_sem);
1971
1972		btrfs_set_sb_rdonly(sb);
1973
1974		/*
1975		 * Setting SB_RDONLY will put the cleaner thread to
1976		 * sleep at the next loop if it's already active.
1977		 * If it's already asleep, we'll leave unused block
1978		 * groups on disk until we're mounted read-write again
1979		 * unless we clean them up here.
1980		 */
1981		btrfs_delete_unused_bgs(fs_info);
1982
1983		/*
1984		 * The cleaner task could be already running before we set the
1985		 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).
1986		 * We must make sure that after we finish the remount, i.e. after
1987		 * we call btrfs_commit_super(), the cleaner can no longer start
1988		 * a transaction - either because it was dropping a dead root,
1989		 * running delayed iputs or deleting an unused block group (the
1990		 * cleaner picked a block group from the list of unused block
1991		 * groups before we were able to in the previous call to
1992		 * btrfs_delete_unused_bgs()).
1993		 */
1994		wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING,
1995			    TASK_UNINTERRUPTIBLE);
1996
1997		/*
1998		 * We've set the superblock to RO mode, so we might have made
1999		 * the cleaner task sleep without running all pending delayed
2000		 * iputs. Go through all the delayed iputs here, so that if an
2001		 * unmount happens without remounting RW we don't end up at
2002		 * finishing close_ctree() with a non-empty list of delayed
2003		 * iputs.
2004		 */
2005		btrfs_run_delayed_iputs(fs_info);
2006
2007		btrfs_dev_replace_suspend_for_unmount(fs_info);
2008		btrfs_scrub_cancel(fs_info);
2009		btrfs_pause_balance(fs_info);
2010
2011		/*
2012		 * Pause the qgroup rescan worker if it is running. We don't want
2013		 * it to be still running after we are in RO mode, as after that,
2014		 * by the time we unmount, it might have left a transaction open,
2015		 * so we would leak the transaction and/or crash.
2016		 */
2017		btrfs_qgroup_wait_for_completion(fs_info, false);
2018
2019		ret = btrfs_commit_super(fs_info);
2020		if (ret)
2021			goto restore;
2022	} else {
2023		if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) {
2024			btrfs_err(fs_info,
2025				"Remounting read-write after error is not allowed");
2026			ret = -EINVAL;
2027			goto restore;
2028		}
2029		if (fs_info->fs_devices->rw_devices == 0) {
2030			ret = -EACCES;
2031			goto restore;
2032		}
2033
2034		if (!btrfs_check_rw_degradable(fs_info, NULL)) {
2035			btrfs_warn(fs_info,
2036		"too many missing devices, writable remount is not allowed");
2037			ret = -EACCES;
2038			goto restore;
2039		}
2040
2041		if (btrfs_super_log_root(fs_info->super_copy) != 0) {
2042			btrfs_warn(fs_info,
2043		"mount required to replay tree-log, cannot remount read-write");
2044			ret = -EINVAL;
2045			goto restore;
2046		}
2047
2048		/*
2049		 * NOTE: when remounting with a change that does writes, don't
2050		 * put it anywhere above this point, as we are not sure to be
2051		 * safe to write until we pass the above checks.
2052		 */
2053		ret = btrfs_start_pre_rw_mount(fs_info);
2054		if (ret)
2055			goto restore;
2056
2057		btrfs_clear_sb_rdonly(sb);
2058
2059		set_bit(BTRFS_FS_OPEN, &fs_info->flags);
2060	}
2061out:
2062	/*
2063	 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS,
2064	 * since the absence of the flag means it can be toggled off by remount.
2065	 */
2066	*flags |= SB_I_VERSION;
2067
2068	wake_up_process(fs_info->transaction_kthread);
2069	btrfs_remount_cleanup(fs_info, old_opts);
2070	btrfs_clear_oneshot_options(fs_info);
2071	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2072
2073	return 0;
2074
2075restore:
2076	/* We've hit an error - don't reset SB_RDONLY */
2077	if (sb_rdonly(sb))
2078		old_flags |= SB_RDONLY;
2079	if (!(old_flags & SB_RDONLY))
2080		clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state);
2081	sb->s_flags = old_flags;
2082	fs_info->mount_opt = old_opts;
2083	fs_info->compress_type = old_compress_type;
2084	fs_info->max_inline = old_max_inline;
2085	btrfs_resize_thread_pool(fs_info,
2086		old_thread_pool_size, fs_info->thread_pool_size);
2087	fs_info->metadata_ratio = old_metadata_ratio;
2088	btrfs_remount_cleanup(fs_info, old_opts);
2089	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
2090
2091	return ret;
2092}
2093
2094/* Used to sort the devices by max_avail(descending sort) */
2095static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
2096{
2097	const struct btrfs_device_info *dev_info1 = a;
2098	const struct btrfs_device_info *dev_info2 = b;
2099
2100	if (dev_info1->max_avail > dev_info2->max_avail)
2101		return -1;
2102	else if (dev_info1->max_avail < dev_info2->max_avail)
2103		return 1;
2104	return 0;
2105}
2106
2107/*
2108 * sort the devices by max_avail, in which max free extent size of each device
2109 * is stored.(Descending Sort)
2110 */
2111static inline void btrfs_descending_sort_devices(
2112					struct btrfs_device_info *devices,
2113					size_t nr_devices)
2114{
2115	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
2116	     btrfs_cmp_device_free_bytes, NULL);
2117}
2118
2119/*
2120 * The helper to calc the free space on the devices that can be used to store
2121 * file data.
2122 */
2123static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
2124					      u64 *free_bytes)
2125{
2126	struct btrfs_device_info *devices_info;
2127	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2128	struct btrfs_device *device;
2129	u64 type;
2130	u64 avail_space;
2131	u64 min_stripe_size;
2132	int num_stripes = 1;
2133	int i = 0, nr_devices;
2134	const struct btrfs_raid_attr *rattr;
2135
2136	/*
2137	 * We aren't under the device list lock, so this is racy-ish, but good
2138	 * enough for our purposes.
2139	 */
2140	nr_devices = fs_info->fs_devices->open_devices;
2141	if (!nr_devices) {
2142		smp_mb();
2143		nr_devices = fs_info->fs_devices->open_devices;
2144		ASSERT(nr_devices);
2145		if (!nr_devices) {
2146			*free_bytes = 0;
2147			return 0;
2148		}
2149	}
2150
2151	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
2152			       GFP_KERNEL);
2153	if (!devices_info)
2154		return -ENOMEM;
2155
2156	/* calc min stripe number for data space allocation */
2157	type = btrfs_data_alloc_profile(fs_info);
2158	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
2159
2160	if (type & BTRFS_BLOCK_GROUP_RAID0)
2161		num_stripes = nr_devices;
2162	else if (type & BTRFS_BLOCK_GROUP_RAID1)
2163		num_stripes = 2;
2164	else if (type & BTRFS_BLOCK_GROUP_RAID1C3)
2165		num_stripes = 3;
2166	else if (type & BTRFS_BLOCK_GROUP_RAID1C4)
2167		num_stripes = 4;
2168	else if (type & BTRFS_BLOCK_GROUP_RAID10)
2169		num_stripes = 4;
2170
2171	/* Adjust for more than 1 stripe per device */
2172	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
2173
2174	rcu_read_lock();
2175	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2176		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
2177						&device->dev_state) ||
2178		    !device->bdev ||
2179		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
2180			continue;
2181
2182		if (i >= nr_devices)
2183			break;
2184
2185		avail_space = device->total_bytes - device->bytes_used;
2186
2187		/* align with stripe_len */
2188		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
2189
2190		/*
2191		 * In order to avoid overwriting the superblock on the drive,
2192		 * btrfs starts at an offset of at least 1MB when doing chunk
2193		 * allocation.
2194		 *
2195		 * This ensures we have at least min_stripe_size free space
2196		 * after excluding 1MB.
2197		 */
2198		if (avail_space <= SZ_1M + min_stripe_size)
2199			continue;
2200
2201		avail_space -= SZ_1M;
2202
2203		devices_info[i].dev = device;
2204		devices_info[i].max_avail = avail_space;
2205
2206		i++;
2207	}
2208	rcu_read_unlock();
2209
2210	nr_devices = i;
2211
2212	btrfs_descending_sort_devices(devices_info, nr_devices);
2213
2214	i = nr_devices - 1;
2215	avail_space = 0;
2216	while (nr_devices >= rattr->devs_min) {
2217		num_stripes = min(num_stripes, nr_devices);
2218
2219		if (devices_info[i].max_avail >= min_stripe_size) {
2220			int j;
2221			u64 alloc_size;
2222
2223			avail_space += devices_info[i].max_avail * num_stripes;
2224			alloc_size = devices_info[i].max_avail;
2225			for (j = i + 1 - num_stripes; j <= i; j++)
2226				devices_info[j].max_avail -= alloc_size;
2227		}
2228		i--;
2229		nr_devices--;
2230	}
2231
2232	kfree(devices_info);
2233	*free_bytes = avail_space;
2234	return 0;
2235}
2236
2237/*
2238 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
2239 *
2240 * If there's a redundant raid level at DATA block groups, use the respective
2241 * multiplier to scale the sizes.
2242 *
2243 * Unused device space usage is based on simulating the chunk allocator
2244 * algorithm that respects the device sizes and order of allocations.  This is
2245 * a close approximation of the actual use but there are other factors that may
2246 * change the result (like a new metadata chunk).
2247 *
2248 * If metadata is exhausted, f_bavail will be 0.
2249 */
2250static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
2251{
2252	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
2253	struct btrfs_super_block *disk_super = fs_info->super_copy;
2254	struct btrfs_space_info *found;
2255	u64 total_used = 0;
2256	u64 total_free_data = 0;
2257	u64 total_free_meta = 0;
2258	u32 bits = fs_info->sectorsize_bits;
2259	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
2260	unsigned factor = 1;
2261	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
2262	int ret;
2263	u64 thresh = 0;
2264	int mixed = 0;
2265
2266	list_for_each_entry(found, &fs_info->space_info, list) {
2267		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
2268			int i;
2269
2270			total_free_data += found->disk_total - found->disk_used;
2271			total_free_data -=
2272				btrfs_account_ro_block_groups_free_space(found);
2273
2274			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
2275				if (!list_empty(&found->block_groups[i]))
2276					factor = btrfs_bg_type_to_factor(
2277						btrfs_raid_array[i].bg_flag);
2278			}
2279		}
2280
2281		/*
2282		 * Metadata in mixed block goup profiles are accounted in data
2283		 */
2284		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
2285			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
2286				mixed = 1;
2287			else
2288				total_free_meta += found->disk_total -
2289					found->disk_used;
2290		}
2291
2292		total_used += found->disk_used;
2293	}
2294
2295	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
2296	buf->f_blocks >>= bits;
2297	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
2298
2299	/* Account global block reserve as used, it's in logical size already */
2300	spin_lock(&block_rsv->lock);
2301	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
2302	if (buf->f_bfree >= block_rsv->size >> bits)
2303		buf->f_bfree -= block_rsv->size >> bits;
2304	else
2305		buf->f_bfree = 0;
2306	spin_unlock(&block_rsv->lock);
2307
2308	buf->f_bavail = div_u64(total_free_data, factor);
2309	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
2310	if (ret)
2311		return ret;
2312	buf->f_bavail += div_u64(total_free_data, factor);
2313	buf->f_bavail = buf->f_bavail >> bits;
2314
2315	/*
2316	 * We calculate the remaining metadata space minus global reserve. If
2317	 * this is (supposedly) smaller than zero, there's no space. But this
2318	 * does not hold in practice, the exhausted state happens where's still
2319	 * some positive delta. So we apply some guesswork and compare the
2320	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
2321	 *
2322	 * We probably cannot calculate the exact threshold value because this
2323	 * depends on the internal reservations requested by various
2324	 * operations, so some operations that consume a few metadata will
2325	 * succeed even if the Avail is zero. But this is better than the other
2326	 * way around.
2327	 */
2328	thresh = SZ_4M;
2329
2330	/*
2331	 * We only want to claim there's no available space if we can no longer
2332	 * allocate chunks for our metadata profile and our global reserve will
2333	 * not fit in the free metadata space.  If we aren't ->full then we
2334	 * still can allocate chunks and thus are fine using the currently
2335	 * calculated f_bavail.
2336	 */
2337	if (!mixed && block_rsv->space_info->full &&
2338	    total_free_meta - thresh < block_rsv->size)
2339		buf->f_bavail = 0;
2340
2341	buf->f_type = BTRFS_SUPER_MAGIC;
2342	buf->f_bsize = dentry->d_sb->s_blocksize;
2343	buf->f_namelen = BTRFS_NAME_LEN;
2344
2345	/* We treat it as constant endianness (it doesn't matter _which_)
2346	   because we want the fsid to come out the same whether mounted
2347	   on a big-endian or little-endian host */
2348	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
2349	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
2350	/* Mask in the root object ID too, to disambiguate subvols */
2351	buf->f_fsid.val[0] ^=
2352		BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32;
2353	buf->f_fsid.val[1] ^=
2354		BTRFS_I(d_inode(dentry))->root->root_key.objectid;
2355
2356	return 0;
2357}
2358
2359static void btrfs_kill_super(struct super_block *sb)
2360{
2361	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2362	kill_anon_super(sb);
2363	btrfs_free_fs_info(fs_info);
2364}
2365
2366static struct file_system_type btrfs_fs_type = {
2367	.owner		= THIS_MODULE,
2368	.name		= "btrfs",
2369	.mount		= btrfs_mount,
2370	.kill_sb	= btrfs_kill_super,
2371	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2372};
2373
2374static struct file_system_type btrfs_root_fs_type = {
2375	.owner		= THIS_MODULE,
2376	.name		= "btrfs",
2377	.mount		= btrfs_mount_root,
2378	.kill_sb	= btrfs_kill_super,
2379	.fs_flags	= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA,
2380};
2381
2382MODULE_ALIAS_FS("btrfs");
2383
2384static int btrfs_control_open(struct inode *inode, struct file *file)
2385{
2386	/*
2387	 * The control file's private_data is used to hold the
2388	 * transaction when it is started and is used to keep
2389	 * track of whether a transaction is already in progress.
2390	 */
2391	file->private_data = NULL;
2392	return 0;
2393}
2394
2395/*
2396 * Used by /dev/btrfs-control for devices ioctls.
2397 */
2398static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2399				unsigned long arg)
2400{
2401	struct btrfs_ioctl_vol_args *vol;
2402	struct btrfs_device *device = NULL;
2403	int ret = -ENOTTY;
2404
2405	if (!capable(CAP_SYS_ADMIN))
2406		return -EPERM;
2407
2408	vol = memdup_user((void __user *)arg, sizeof(*vol));
2409	if (IS_ERR(vol))
2410		return PTR_ERR(vol);
2411	vol->name[BTRFS_PATH_NAME_MAX] = '\0';
2412
2413	switch (cmd) {
2414	case BTRFS_IOC_SCAN_DEV:
2415		mutex_lock(&uuid_mutex);
2416		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2417					       &btrfs_root_fs_type);
2418		ret = PTR_ERR_OR_ZERO(device);
2419		mutex_unlock(&uuid_mutex);
2420		break;
2421	case BTRFS_IOC_FORGET_DEV:
2422		ret = btrfs_forget_devices(vol->name);
2423		break;
2424	case BTRFS_IOC_DEVICES_READY:
2425		mutex_lock(&uuid_mutex);
2426		device = btrfs_scan_one_device(vol->name, FMODE_READ,
2427					       &btrfs_root_fs_type);
2428		if (IS_ERR(device)) {
2429			mutex_unlock(&uuid_mutex);
2430			ret = PTR_ERR(device);
2431			break;
2432		}
2433		ret = !(device->fs_devices->num_devices ==
2434			device->fs_devices->total_devices);
2435		mutex_unlock(&uuid_mutex);
2436		break;
2437	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2438		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2439		break;
2440	}
2441
2442	kfree(vol);
2443	return ret;
2444}
2445
2446static int btrfs_freeze(struct super_block *sb)
2447{
2448	struct btrfs_trans_handle *trans;
2449	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2450	struct btrfs_root *root = fs_info->tree_root;
2451
2452	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2453	/*
2454	 * We don't need a barrier here, we'll wait for any transaction that
2455	 * could be in progress on other threads (and do delayed iputs that
2456	 * we want to avoid on a frozen filesystem), or do the commit
2457	 * ourselves.
2458	 */
2459	trans = btrfs_attach_transaction_barrier(root);
2460	if (IS_ERR(trans)) {
2461		/* no transaction, don't bother */
2462		if (PTR_ERR(trans) == -ENOENT)
2463			return 0;
2464		return PTR_ERR(trans);
2465	}
2466	return btrfs_commit_transaction(trans);
2467}
2468
2469static int btrfs_unfreeze(struct super_block *sb)
2470{
2471	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2472
2473	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2474	return 0;
2475}
2476
2477static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2478{
2479	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2480	struct btrfs_device *dev, *first_dev = NULL;
2481
2482	/*
2483	 * Lightweight locking of the devices. We should not need
2484	 * device_list_mutex here as we only read the device data and the list
2485	 * is protected by RCU.  Even if a device is deleted during the list
2486	 * traversals, we'll get valid data, the freeing callback will wait at
2487	 * least until the rcu_read_unlock.
2488	 */
2489	rcu_read_lock();
2490	list_for_each_entry_rcu(dev, &fs_info->fs_devices->devices, dev_list) {
2491		if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state))
2492			continue;
2493		if (!dev->name)
2494			continue;
2495		if (!first_dev || dev->devid < first_dev->devid)
2496			first_dev = dev;
2497	}
2498
2499	if (first_dev)
2500		seq_escape(m, rcu_str_deref(first_dev->name), " \t\n\\");
2501	else
2502		WARN_ON(1);
2503	rcu_read_unlock();
2504	return 0;
2505}
2506
2507static const struct super_operations btrfs_super_ops = {
2508	.drop_inode	= btrfs_drop_inode,
2509	.evict_inode	= btrfs_evict_inode,
2510	.put_super	= btrfs_put_super,
2511	.sync_fs	= btrfs_sync_fs,
2512	.show_options	= btrfs_show_options,
2513	.show_devname	= btrfs_show_devname,
2514	.alloc_inode	= btrfs_alloc_inode,
2515	.destroy_inode	= btrfs_destroy_inode,
2516	.free_inode	= btrfs_free_inode,
2517	.statfs		= btrfs_statfs,
2518	.remount_fs	= btrfs_remount,
2519	.freeze_fs	= btrfs_freeze,
2520	.unfreeze_fs	= btrfs_unfreeze,
2521};
2522
2523static const struct file_operations btrfs_ctl_fops = {
2524	.open = btrfs_control_open,
2525	.unlocked_ioctl	 = btrfs_control_ioctl,
2526	.compat_ioctl = compat_ptr_ioctl,
2527	.owner	 = THIS_MODULE,
2528	.llseek = noop_llseek,
2529};
2530
2531static struct miscdevice btrfs_misc = {
2532	.minor		= BTRFS_MINOR,
2533	.name		= "btrfs-control",
2534	.fops		= &btrfs_ctl_fops
2535};
2536
2537MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2538MODULE_ALIAS("devname:btrfs-control");
2539
2540static int __init btrfs_interface_init(void)
2541{
2542	return misc_register(&btrfs_misc);
2543}
2544
2545static __cold void btrfs_interface_exit(void)
2546{
2547	misc_deregister(&btrfs_misc);
2548}
2549
2550static void __init btrfs_print_mod_info(void)
2551{
2552	static const char options[] = ""
2553#ifdef CONFIG_BTRFS_DEBUG
2554			", debug=on"
2555#endif
2556#ifdef CONFIG_BTRFS_ASSERT
2557			", assert=on"
2558#endif
2559#ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY
2560			", integrity-checker=on"
2561#endif
2562#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2563			", ref-verify=on"
2564#endif
2565#ifdef CONFIG_BLK_DEV_ZONED
2566			", zoned=yes"
2567#else
2568			", zoned=no"
2569#endif
2570			;
2571	pr_info("Btrfs loaded, crc32c=%s%s\n", crc32c_impl(), options);
2572}
2573
2574static int __init init_btrfs_fs(void)
2575{
2576	int err;
2577
2578	btrfs_props_init();
2579
2580	err = btrfs_init_sysfs();
2581	if (err)
2582		return err;
2583
2584	btrfs_init_compress();
2585
2586	err = btrfs_init_cachep();
2587	if (err)
2588		goto free_compress;
2589
2590	err = extent_io_init();
2591	if (err)
2592		goto free_cachep;
2593
2594	err = extent_state_cache_init();
2595	if (err)
2596		goto free_extent_io;
2597
2598	err = extent_map_init();
2599	if (err)
2600		goto free_extent_state_cache;
2601
2602	err = ordered_data_init();
2603	if (err)
2604		goto free_extent_map;
2605
2606	err = btrfs_delayed_inode_init();
2607	if (err)
2608		goto free_ordered_data;
2609
2610	err = btrfs_auto_defrag_init();
2611	if (err)
2612		goto free_delayed_inode;
2613
2614	err = btrfs_delayed_ref_init();
2615	if (err)
2616		goto free_auto_defrag;
2617
2618	err = btrfs_prelim_ref_init();
2619	if (err)
2620		goto free_delayed_ref;
2621
2622	err = btrfs_end_io_wq_init();
2623	if (err)
2624		goto free_prelim_ref;
2625
2626	err = btrfs_interface_init();
2627	if (err)
2628		goto free_end_io_wq;
2629
2630	btrfs_print_mod_info();
2631
2632	err = btrfs_run_sanity_tests();
2633	if (err)
2634		goto unregister_ioctl;
2635
2636	err = register_filesystem(&btrfs_fs_type);
2637	if (err)
2638		goto unregister_ioctl;
2639
2640	return 0;
2641
2642unregister_ioctl:
2643	btrfs_interface_exit();
2644free_end_io_wq:
2645	btrfs_end_io_wq_exit();
2646free_prelim_ref:
2647	btrfs_prelim_ref_exit();
2648free_delayed_ref:
2649	btrfs_delayed_ref_exit();
2650free_auto_defrag:
2651	btrfs_auto_defrag_exit();
2652free_delayed_inode:
2653	btrfs_delayed_inode_exit();
2654free_ordered_data:
2655	ordered_data_exit();
2656free_extent_map:
2657	extent_map_exit();
2658free_extent_state_cache:
2659	extent_state_cache_exit();
2660free_extent_io:
2661	extent_io_exit();
2662free_cachep:
2663	btrfs_destroy_cachep();
2664free_compress:
2665	btrfs_exit_compress();
2666	btrfs_exit_sysfs();
2667
2668	return err;
2669}
2670
2671static void __exit exit_btrfs_fs(void)
2672{
2673	btrfs_destroy_cachep();
2674	btrfs_delayed_ref_exit();
2675	btrfs_auto_defrag_exit();
2676	btrfs_delayed_inode_exit();
2677	btrfs_prelim_ref_exit();
2678	ordered_data_exit();
2679	extent_map_exit();
2680	extent_state_cache_exit();
2681	extent_io_exit();
2682	btrfs_interface_exit();
2683	btrfs_end_io_wq_exit();
2684	unregister_filesystem(&btrfs_fs_type);
2685	btrfs_exit_sysfs();
2686	btrfs_cleanup_fs_uuids();
2687	btrfs_exit_compress();
2688}
2689
2690late_initcall(init_btrfs_fs);
2691module_exit(exit_btrfs_fs)
2692
2693MODULE_LICENSE("GPL");
2694MODULE_SOFTDEP("pre: crc32c");
2695MODULE_SOFTDEP("pre: xxhash64");
2696MODULE_SOFTDEP("pre: sha256");
2697MODULE_SOFTDEP("pre: blake2b-256");
2698