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 <linux/security.h>
30	#include <linux/fs_parser.h>
31	#include "messages.h"
32	#include "delayed-inode.h"
33	#include "ctree.h"
34	#include "disk-io.h"
35	#include "transaction.h"
36	#include "btrfs_inode.h"
37	#include "props.h"
38	#include "xattr.h"
39	#include "bio.h"
40	#include "export.h"
41	#include "compression.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	#include "raid56.h"
53	#include "fs.h"
54	#include "accessors.h"
55	#include "defrag.h"
56	#include "dir-item.h"
57	#include "ioctl.h"
58	#include "scrub.h"
59	#include "verity.h"
60	#include "super.h"
61	#include "extent-tree.h"
62	#define CREATE_TRACE_POINTS
63	#include <trace/events/btrfs.h>
64
65	static const struct super_operations btrfs_super_ops;
66	static struct file_system_type btrfs_fs_type;
67
68	static void btrfs_put_super(struct super_block *sb)
69	{
70	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
71
72	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
73	close_ctree(fs_info);
74	}
75
76	/* Store the mount options related information. */
77	struct btrfs_fs_context {
78	char *subvol_name;
79	u64 subvol_objectid;
80	u64 max_inline;
81	u32 commit_interval;
82	u32 metadata_ratio;
83	u32 thread_pool_size;
84	unsigned long mount_opt;
85	unsigned long compress_type:4;
86	unsigned int compress_level;
87	refcount_t refs;
88	};
89
90	enum {
91	Opt_acl,
92	Opt_clear_cache,
93	Opt_commit_interval,
94	Opt_compress,
95	Opt_compress_force,
96	Opt_compress_force_type,
97	Opt_compress_type,
98	Opt_degraded,
99	Opt_device,
100	Opt_fatal_errors,
101	Opt_flushoncommit,
102	Opt_max_inline,
103	Opt_barrier,
104	Opt_datacow,
105	Opt_datasum,
106	Opt_defrag,
107	Opt_discard,
108	Opt_discard_mode,
109	Opt_ratio,
110	Opt_rescan_uuid_tree,
111	Opt_skip_balance,
112	Opt_space_cache,
113	Opt_space_cache_version,
114	Opt_ssd,
115	Opt_ssd_spread,
116	Opt_subvol,
117	Opt_subvol_empty,
118	Opt_subvolid,
119	Opt_thread_pool,
120	Opt_treelog,
121	Opt_user_subvol_rm_allowed,
122
123	/* Rescue options */
124	Opt_rescue,
125	Opt_usebackuproot,
126	Opt_nologreplay,
127	Opt_ignorebadroots,
128	Opt_ignoredatacsums,
129	Opt_rescue_all,
130
131	/* Debugging options */
132	Opt_enospc_debug,
133	#ifdef CONFIG_BTRFS_DEBUG
134	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
135	#endif
136	#ifdef CONFIG_BTRFS_FS_REF_VERIFY
137	Opt_ref_verify,
138	#endif
139	Opt_err,
140	};
141
142	enum {
143	Opt_fatal_errors_panic,
144	Opt_fatal_errors_bug,
145	};
146
147	static const struct constant_table btrfs_parameter_fatal_errors[] = {
148	{ "panic", Opt_fatal_errors_panic },
149	{ "bug", Opt_fatal_errors_bug },
150	{}
151	};
152
153	enum {
154	Opt_discard_sync,
155	Opt_discard_async,
156	};
157
158	static const struct constant_table btrfs_parameter_discard[] = {
159	{ "sync", Opt_discard_sync },
160	{ "async", Opt_discard_async },
161	{}
162	};
163
164	enum {
165	Opt_space_cache_v1,
166	Opt_space_cache_v2,
167	};
168
169	static const struct constant_table btrfs_parameter_space_cache[] = {
170	{ "v1", Opt_space_cache_v1 },
171	{ "v2", Opt_space_cache_v2 },
172	{}
173	};
174
175	enum {
176	Opt_rescue_usebackuproot,
177	Opt_rescue_nologreplay,
178	Opt_rescue_ignorebadroots,
179	Opt_rescue_ignoredatacsums,
180	Opt_rescue_parameter_all,
181	};
182
183	static const struct constant_table btrfs_parameter_rescue[] = {
184	{ "usebackuproot", Opt_rescue_usebackuproot },
185	{ "nologreplay", Opt_rescue_nologreplay },
186	{ "ignorebadroots", Opt_rescue_ignorebadroots },
187	{ "ibadroots", Opt_rescue_ignorebadroots },
188	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
189	{ "idatacsums", Opt_rescue_ignoredatacsums },
190	{ "all", Opt_rescue_parameter_all },
191	{}
192	};
193
194	#ifdef CONFIG_BTRFS_DEBUG
195	enum {
196	Opt_fragment_parameter_data,
197	Opt_fragment_parameter_metadata,
198	Opt_fragment_parameter_all,
199	};
200
201	static const struct constant_table btrfs_parameter_fragment[] = {
202	{ "data", Opt_fragment_parameter_data },
203	{ "metadata", Opt_fragment_parameter_metadata },
204	{ "all", Opt_fragment_parameter_all },
205	{}
206	};
207	#endif
208
209	static const struct fs_parameter_spec btrfs_fs_parameters[] = {
210	fsparam_flag_no("acl", Opt_acl),
211	fsparam_flag_no("autodefrag", Opt_defrag),
212	fsparam_flag_no("barrier", Opt_barrier),
213	fsparam_flag("clear_cache", Opt_clear_cache),
214	fsparam_u32("commit", Opt_commit_interval),
215	fsparam_flag("compress", Opt_compress),
216	fsparam_string("compress", Opt_compress_type),
217	fsparam_flag("compress-force", Opt_compress_force),
218	fsparam_string("compress-force", Opt_compress_force_type),
219	fsparam_flag_no("datacow", Opt_datacow),
220	fsparam_flag_no("datasum", Opt_datasum),
221	fsparam_flag("degraded", Opt_degraded),
222	fsparam_string("device", Opt_device),
223	fsparam_flag_no("discard", Opt_discard),
224	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
225	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
226	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
227	fsparam_string("max_inline", Opt_max_inline),
228	fsparam_u32("metadata_ratio", Opt_ratio),
229	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
230	fsparam_flag("skip_balance", Opt_skip_balance),
231	fsparam_flag_no("space_cache", Opt_space_cache),
232	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
233	fsparam_flag_no("ssd", Opt_ssd),
234	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
235	fsparam_string("subvol", Opt_subvol),
236	fsparam_flag("subvol=", Opt_subvol_empty),
237	fsparam_u64("subvolid", Opt_subvolid),
238	fsparam_u32("thread_pool", Opt_thread_pool),
239	fsparam_flag_no("treelog", Opt_treelog),
240	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
241
242	/* Rescue options. */
243	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
244	/* Deprecated, with alias rescue=nologreplay */
245	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
246	/* Deprecated, with alias rescue=usebackuproot */
247	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
248
249	/* Debugging options. */
250	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
251	#ifdef CONFIG_BTRFS_DEBUG
252	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
253	#endif
254	#ifdef CONFIG_BTRFS_FS_REF_VERIFY
255	fsparam_flag("ref_verify", Opt_ref_verify),
256	#endif
257	{}
258	};
259
260	/* No support for restricting writes to btrfs devices yet... */
261	static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
262	{
263	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
264	}
265
266	static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
267	{
268	struct btrfs_fs_context *ctx = fc->fs_private;
269	struct fs_parse_result result;
270	int opt;
271
272	opt = fs_parse(fc, desc: btrfs_fs_parameters, param, result: &result);
273	if (opt < 0)
274	return opt;
275
276	switch (opt) {
277	case Opt_degraded:
278	btrfs_set_opt(ctx->mount_opt, DEGRADED);
279	break;
280	case Opt_subvol_empty:
281	/*
282	* This exists because we used to allow it on accident, so we're
283	* keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow
284	* empty subvol= again").
285	*/
286	break;
287	case Opt_subvol:
288	kfree(objp: ctx->subvol_name);
289	ctx->subvol_name = kstrdup(s: param->string, GFP_KERNEL);
290	if (!ctx->subvol_name)
291	return -ENOMEM;
292	break;
293	case Opt_subvolid:
294	ctx->subvol_objectid = result.uint_64;
295
296	/* subvolid=0 means give me the original fs_tree. */
297	if (!ctx->subvol_objectid)
298	ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
299	break;
300	case Opt_device: {
301	struct btrfs_device *device;
302	blk_mode_t mode = btrfs_open_mode(fc);
303
304	mutex_lock(&uuid_mutex);
305	device = btrfs_scan_one_device(path: param->string, flags: mode, mount_arg_dev: false);
306	mutex_unlock(lock: &uuid_mutex);
307	if (IS_ERR(ptr: device))
308	return PTR_ERR(ptr: device);
309	break;
310	}
311	case Opt_datasum:
312	if (result.negated) {
313	btrfs_set_opt(ctx->mount_opt, NODATASUM);
314	} else {
315	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
316	btrfs_clear_opt(ctx->mount_opt, NODATASUM);
317	}
318	break;
319	case Opt_datacow:
320	if (result.negated) {
321	btrfs_clear_opt(ctx->mount_opt, COMPRESS);
322	btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
323	btrfs_set_opt(ctx->mount_opt, NODATACOW);
324	btrfs_set_opt(ctx->mount_opt, NODATASUM);
325	} else {
326	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
327	}
328	break;
329	case Opt_compress_force:
330	case Opt_compress_force_type:
331	btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
332	fallthrough;
333	case Opt_compress:
334	case Opt_compress_type:
335	if (opt == Opt_compress || opt == Opt_compress_force) {
336	ctx->compress_type = BTRFS_COMPRESS_ZLIB;
337	ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
338	btrfs_set_opt(ctx->mount_opt, COMPRESS);
339	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
340	btrfs_clear_opt(ctx->mount_opt, NODATASUM);
341	} else if (strncmp(param->string, "zlib", 4) == 0) {
342	ctx->compress_type = BTRFS_COMPRESS_ZLIB;
343	ctx->compress_level =
344	btrfs_compress_str2level(type: BTRFS_COMPRESS_ZLIB,
345	str: param->string + 4);
346	btrfs_set_opt(ctx->mount_opt, COMPRESS);
347	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
348	btrfs_clear_opt(ctx->mount_opt, NODATASUM);
349	} else if (strncmp(param->string, "lzo", 3) == 0) {
350	ctx->compress_type = BTRFS_COMPRESS_LZO;
351	ctx->compress_level = 0;
352	btrfs_set_opt(ctx->mount_opt, COMPRESS);
353	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
354	btrfs_clear_opt(ctx->mount_opt, NODATASUM);
355	} else if (strncmp(param->string, "zstd", 4) == 0) {
356	ctx->compress_type = BTRFS_COMPRESS_ZSTD;
357	ctx->compress_level =
358	btrfs_compress_str2level(type: BTRFS_COMPRESS_ZSTD,
359	str: param->string + 4);
360	btrfs_set_opt(ctx->mount_opt, COMPRESS);
361	btrfs_clear_opt(ctx->mount_opt, NODATACOW);
362	btrfs_clear_opt(ctx->mount_opt, NODATASUM);
363	} else if (strncmp(param->string, "no", 2) == 0) {
364	ctx->compress_level = 0;
365	ctx->compress_type = 0;
366	btrfs_clear_opt(ctx->mount_opt, COMPRESS);
367	btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
368	} else {
369	btrfs_err(NULL, "unrecognized compression value %s",
370	param->string);
371	return -EINVAL;
372	}
373	break;
374	case Opt_ssd:
375	if (result.negated) {
376	btrfs_set_opt(ctx->mount_opt, NOSSD);
377	btrfs_clear_opt(ctx->mount_opt, SSD);
378	btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
379	} else {
380	btrfs_set_opt(ctx->mount_opt, SSD);
381	btrfs_clear_opt(ctx->mount_opt, NOSSD);
382	}
383	break;
384	case Opt_ssd_spread:
385	if (result.negated) {
386	btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
387	} else {
388	btrfs_set_opt(ctx->mount_opt, SSD);
389	btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
390	btrfs_clear_opt(ctx->mount_opt, NOSSD);
391	}
392	break;
393	case Opt_barrier:
394	if (result.negated)
395	btrfs_set_opt(ctx->mount_opt, NOBARRIER);
396	else
397	btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
398	break;
399	case Opt_thread_pool:
400	if (result.uint_32 == 0) {
401	btrfs_err(NULL, "invalid value 0 for thread_pool");
402	return -EINVAL;
403	}
404	ctx->thread_pool_size = result.uint_32;
405	break;
406	case Opt_max_inline:
407	ctx->max_inline = memparse(ptr: param->string, NULL);
408	break;
409	case Opt_acl:
410	if (result.negated) {
411	fc->sb_flags &= ~SB_POSIXACL;
412	} else {
413	#ifdef CONFIG_BTRFS_FS_POSIX_ACL
414	fc->sb_flags |= SB_POSIXACL;
415	#else
416	btrfs_err(NULL, "support for ACL not compiled in");
417	return -EINVAL;
418	#endif
419	}
420	/*
421	* VFS limits the ability to toggle ACL on and off via remount,
422	* despite every file system allowing this. This seems to be
423	* an oversight since we all do, but it'll fail if we're
424	* remounting. So don't set the mask here, we'll check it in
425	* btrfs_reconfigure and do the toggling ourselves.
426	*/
427	if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
428	fc->sb_flags_mask |= SB_POSIXACL;
429	break;
430	case Opt_treelog:
431	if (result.negated)
432	btrfs_set_opt(ctx->mount_opt, NOTREELOG);
433	else
434	btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
435	break;
436	case Opt_nologreplay:
437	btrfs_warn(NULL,
438	"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
439	btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
440	break;
441	case Opt_flushoncommit:
442	if (result.negated)
443	btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
444	else
445	btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
446	break;
447	case Opt_ratio:
448	ctx->metadata_ratio = result.uint_32;
449	break;
450	case Opt_discard:
451	if (result.negated) {
452	btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
453	btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
454	btrfs_set_opt(ctx->mount_opt, NODISCARD);
455	} else {
456	btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
457	btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
458	}
459	break;
460	case Opt_discard_mode:
461	switch (result.uint_32) {
462	case Opt_discard_sync:
463	btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
464	btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
465	break;
466	case Opt_discard_async:
467	btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
468	btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
469	break;
470	default:
471	btrfs_err(NULL, "unrecognized discard mode value %s",
472	param->key);
473	return -EINVAL;
474	}
475	btrfs_clear_opt(ctx->mount_opt, NODISCARD);
476	break;
477	case Opt_space_cache:
478	if (result.negated) {
479	btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
480	btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
481	btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
482	} else {
483	btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
484	btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
485	}
486	break;
487	case Opt_space_cache_version:
488	switch (result.uint_32) {
489	case Opt_space_cache_v1:
490	btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
491	btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
492	break;
493	case Opt_space_cache_v2:
494	btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
495	btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
496	break;
497	default:
498	btrfs_err(NULL, "unrecognized space_cache value %s",
499	param->key);
500	return -EINVAL;
501	}
502	break;
503	case Opt_rescan_uuid_tree:
504	btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
505	break;
506	case Opt_clear_cache:
507	btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
508	break;
509	case Opt_user_subvol_rm_allowed:
510	btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
511	break;
512	case Opt_enospc_debug:
513	if (result.negated)
514	btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
515	else
516	btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
517	break;
518	case Opt_defrag:
519	if (result.negated)
520	btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
521	else
522	btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
523	break;
524	case Opt_usebackuproot:
525	btrfs_warn(NULL,
526	"'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
527	btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
528
529	/* If we're loading the backup roots we can't trust the space cache. */
530	btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
531	break;
532	case Opt_skip_balance:
533	btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
534	break;
535	case Opt_fatal_errors:
536	switch (result.uint_32) {
537	case Opt_fatal_errors_panic:
538	btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
539	break;
540	case Opt_fatal_errors_bug:
541	btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
542	break;
543	default:
544	btrfs_err(NULL, "unrecognized fatal_errors value %s",
545	param->key);
546	return -EINVAL;
547	}
548	break;
549	case Opt_commit_interval:
550	ctx->commit_interval = result.uint_32;
551	if (ctx->commit_interval == 0)
552	ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
553	break;
554	case Opt_rescue:
555	switch (result.uint_32) {
556	case Opt_rescue_usebackuproot:
557	btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
558	break;
559	case Opt_rescue_nologreplay:
560	btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
561	break;
562	case Opt_rescue_ignorebadroots:
563	btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
564	break;
565	case Opt_rescue_ignoredatacsums:
566	btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
567	break;
568	case Opt_rescue_parameter_all:
569	btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
570	btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
571	btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
572	break;
573	default:
574	btrfs_info(NULL, "unrecognized rescue option '%s'",
575	param->key);
576	return -EINVAL;
577	}
578	break;
579	#ifdef CONFIG_BTRFS_DEBUG
580	case Opt_fragment:
581	switch (result.uint_32) {
582	case Opt_fragment_parameter_all:
583	btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
584	btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
585	break;
586	case Opt_fragment_parameter_metadata:
587	btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
588	break;
589	case Opt_fragment_parameter_data:
590	btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
591	break;
592	default:
593	btrfs_info(NULL, "unrecognized fragment option '%s'",
594	param->key);
595	return -EINVAL;
596	}
597	break;
598	#endif
599	#ifdef CONFIG_BTRFS_FS_REF_VERIFY
600	case Opt_ref_verify:
601	btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
602	break;
603	#endif
604	default:
605	btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
606	return -EINVAL;
607	}
608
609	return 0;
610	}
611
612	/*
613	* Some options only have meaning at mount time and shouldn't persist across
614	* remounts, or be displayed. Clear these at the end of mount and remount code
615	* paths.
616	*/
617	static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
618	{
619	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
620	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
621	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
622	}
623
624	static bool check_ro_option(struct btrfs_fs_info *fs_info,
625	unsigned long mount_opt, unsigned long opt,
626	const char *opt_name)
627	{
628	if (mount_opt & opt) {
629	btrfs_err(fs_info, "%s must be used with ro mount option",
630	opt_name);
631	return true;
632	}
633	return false;
634	}
635
636	bool btrfs_check_options(struct btrfs_fs_info *info, unsigned long *mount_opt,
637	unsigned long flags)
638	{
639	bool ret = true;
640
641	if (!(flags & SB_RDONLY) &&
642	(check_ro_option(fs_info: info, mount_opt: *mount_opt, opt: BTRFS_MOUNT_NOLOGREPLAY, opt_name: "nologreplay") ||
643	check_ro_option(fs_info: info, mount_opt: *mount_opt, opt: BTRFS_MOUNT_IGNOREBADROOTS, opt_name: "ignorebadroots") ||
644	check_ro_option(fs_info: info, mount_opt: *mount_opt, opt: BTRFS_MOUNT_IGNOREDATACSUMS, opt_name: "ignoredatacsums")))
645	ret = false;
646
647	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
648	!btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
649	!btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
650	btrfs_err(info, "cannot disable free-space-tree");
651	ret = false;
652	}
653	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
654	!btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
655	btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
656	ret = false;
657	}
658
659	if (btrfs_check_mountopts_zoned(info, mount_opt))
660	ret = false;
661
662	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
663	if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE))
664	btrfs_info(info, "disk space caching is enabled");
665	if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
666	btrfs_info(info, "using free-space-tree");
667	}
668
669	return ret;
670	}
671
672	/*
673	* This is subtle, we only call this during open_ctree(). We need to pre-load
674	* the mount options with the on-disk settings. Before the new mount API took
675	* effect we would do this on mount and remount. With the new mount API we'll
676	* only do this on the initial mount.
677	*
678	* This isn't a change in behavior, because we're using the current state of the
679	* file system to set the current mount options. If you mounted with special
680	* options to disable these features and then remounted we wouldn't revert the
681	* settings, because mounting without these features cleared the on-disk
682	* settings, so this being called on re-mount is not needed.
683	*/
684	void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
685	{
686	if (fs_info->sectorsize < PAGE_SIZE) {
687	btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
688	if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
689	btrfs_info(fs_info,
690	"forcing free space tree for sector size %u with page size %lu",
691	fs_info->sectorsize, PAGE_SIZE);
692	btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
693	}
694	}
695
696	/*
697	* At this point our mount options are populated, so we only mess with
698	* these settings if we don't have any settings already.
699	*/
700	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
701	return;
702
703	if (btrfs_is_zoned(fs_info) &&
704	btrfs_free_space_cache_v1_active(fs_info)) {
705	btrfs_info(fs_info, "zoned: clearing existing space cache");
706	btrfs_set_super_cache_generation(s: fs_info->super_copy, val: 0);
707	return;
708	}
709
710	if (btrfs_test_opt(fs_info, SPACE_CACHE))
711	return;
712
713	if (btrfs_test_opt(fs_info, NOSPACECACHE))
714	return;
715
716	/*
717	* At this point we don't have explicit options set by the user, set
718	* them ourselves based on the state of the file system.
719	*/
720	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
721	btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
722	else if (btrfs_free_space_cache_v1_active(fs_info))
723	btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
724	}
725
726	static void set_device_specific_options(struct btrfs_fs_info *fs_info)
727	{
728	if (!btrfs_test_opt(fs_info, NOSSD) &&
729	!fs_info->fs_devices->rotating)
730	btrfs_set_opt(fs_info->mount_opt, SSD);
731
732	/*
733	* For devices supporting discard turn on discard=async automatically,
734	* unless it's already set or disabled. This could be turned off by
735	* nodiscard for the same mount.
736	*
737	* The zoned mode piggy backs on the discard functionality for
738	* resetting a zone. There is no reason to delay the zone reset as it is
739	* fast enough. So, do not enable async discard for zoned mode.
740	*/
741	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
742	btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
743	btrfs_test_opt(fs_info, NODISCARD)) &&
744	fs_info->fs_devices->discardable &&
745	!btrfs_is_zoned(fs_info))
746	btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
747	}
748
749	char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
750	u64 subvol_objectid)
751	{
752	struct btrfs_root *root = fs_info->tree_root;
753	struct btrfs_root *fs_root = NULL;
754	struct btrfs_root_ref *root_ref;
755	struct btrfs_inode_ref *inode_ref;
756	struct btrfs_key key;
757	struct btrfs_path *path = NULL;
758	char *name = NULL, *ptr;
759	u64 dirid;
760	int len;
761	int ret;
762
763	path = btrfs_alloc_path();
764	if (!path) {
765	ret = -ENOMEM;
766	goto err;
767	}
768
769	name = kmalloc(PATH_MAX, GFP_KERNEL);
770	if (!name) {
771	ret = -ENOMEM;
772	goto err;
773	}
774	ptr = name + PATH_MAX - 1;
775	ptr[0] = '\0';
776
777	/*
778	* Walk up the subvolume trees in the tree of tree roots by root
779	* backrefs until we hit the top-level subvolume.
780	*/
781	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
782	key.objectid = subvol_objectid;
783	key.type = BTRFS_ROOT_BACKREF_KEY;
784	key.offset = (u64)-1;
785
786	ret = btrfs_search_backwards(root, key: &key, path);
787	if (ret < 0) {
788	goto err;
789	} else if (ret > 0) {
790	ret = -ENOENT;
791	goto err;
792	}
793
794	subvol_objectid = key.offset;
795
796	root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
797	struct btrfs_root_ref);
798	len = btrfs_root_ref_name_len(eb: path->nodes[0], s: root_ref);
799	ptr -= len + 1;
800	if (ptr < name) {
801	ret = -ENAMETOOLONG;
802	goto err;
803	}
804	read_extent_buffer(eb: path->nodes[0], dst: ptr + 1,
805	start: (unsigned long)(root_ref + 1), len);
806	ptr[0] = '/';
807	dirid = btrfs_root_ref_dirid(eb: path->nodes[0], s: root_ref);
808	btrfs_release_path(p: path);
809
810	fs_root = btrfs_get_fs_root(fs_info, objectid: subvol_objectid, check_ref: true);
811	if (IS_ERR(ptr: fs_root)) {
812	ret = PTR_ERR(ptr: fs_root);
813	fs_root = NULL;
814	goto err;
815	}
816
817	/*
818	* Walk up the filesystem tree by inode refs until we hit the
819	* root directory.
820	*/
821	while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
822	key.objectid = dirid;
823	key.type = BTRFS_INODE_REF_KEY;
824	key.offset = (u64)-1;
825
826	ret = btrfs_search_backwards(root: fs_root, key: &key, path);
827	if (ret < 0) {
828	goto err;
829	} else if (ret > 0) {
830	ret = -ENOENT;
831	goto err;
832	}
833
834	dirid = key.offset;
835
836	inode_ref = btrfs_item_ptr(path->nodes[0],
837	path->slots[0],
838	struct btrfs_inode_ref);
839	len = btrfs_inode_ref_name_len(eb: path->nodes[0],
840	s: inode_ref);
841	ptr -= len + 1;
842	if (ptr < name) {
843	ret = -ENAMETOOLONG;
844	goto err;
845	}
846	read_extent_buffer(eb: path->nodes[0], dst: ptr + 1,
847	start: (unsigned long)(inode_ref + 1), len);
848	ptr[0] = '/';
849	btrfs_release_path(p: path);
850	}
851	btrfs_put_root(root: fs_root);
852	fs_root = NULL;
853	}
854
855	btrfs_free_path(p: path);
856	if (ptr == name + PATH_MAX - 1) {
857	name[0] = '/';
858	name[1] = '\0';
859	} else {
860	memmove(name, ptr, name + PATH_MAX - ptr);
861	}
862	return name;
863
864	err:
865	btrfs_put_root(root: fs_root);
866	btrfs_free_path(p: path);
867	kfree(objp: name);
868	return ERR_PTR(error: ret);
869	}
870
871	static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
872	{
873	struct btrfs_root *root = fs_info->tree_root;
874	struct btrfs_dir_item *di;
875	struct btrfs_path *path;
876	struct btrfs_key location;
877	struct fscrypt_str name = FSTR_INIT("default", 7);
878	u64 dir_id;
879
880	path = btrfs_alloc_path();
881	if (!path)
882	return -ENOMEM;
883
884	/*
885	* Find the "default" dir item which points to the root item that we
886	* will mount by default if we haven't been given a specific subvolume
887	* to mount.
888	*/
889	dir_id = btrfs_super_root_dir(s: fs_info->super_copy);
890	di = btrfs_lookup_dir_item(NULL, root, path, dir: dir_id, name: &name, mod: 0);
891	if (IS_ERR(ptr: di)) {
892	btrfs_free_path(p: path);
893	return PTR_ERR(ptr: di);
894	}
895	if (!di) {
896	/*
897	* Ok the default dir item isn't there. This is weird since
898	* it's always been there, but don't freak out, just try and
899	* mount the top-level subvolume.
900	*/
901	btrfs_free_path(p: path);
902	*objectid = BTRFS_FS_TREE_OBJECTID;
903	return 0;
904	}
905
906	btrfs_dir_item_key_to_cpu(eb: path->nodes[0], item: di, cpu_key: &location);
907	btrfs_free_path(p: path);
908	*objectid = location.objectid;
909	return 0;
910	}
911
912	static int btrfs_fill_super(struct super_block *sb,
913	struct btrfs_fs_devices *fs_devices,
914	void *data)
915	{
916	struct inode *inode;
917	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
918	int err;
919
920	sb->s_maxbytes = MAX_LFS_FILESIZE;
921	sb->s_magic = BTRFS_SUPER_MAGIC;
922	sb->s_op = &btrfs_super_ops;
923	sb->s_d_op = &btrfs_dentry_operations;
924	sb->s_export_op = &btrfs_export_ops;
925	#ifdef CONFIG_FS_VERITY
926	sb->s_vop = &btrfs_verityops;
927	#endif
928	sb->s_xattr = btrfs_xattr_handlers;
929	sb->s_time_gran = 1;
930	sb->s_iflags |= SB_I_CGROUPWB;
931
932	err = super_setup_bdi(sb);
933	if (err) {
934	btrfs_err(fs_info, "super_setup_bdi failed");
935	return err;
936	}
937
938	err = open_ctree(sb, fs_devices, options: (char *)data);
939	if (err) {
940	btrfs_err(fs_info, "open_ctree failed");
941	return err;
942	}
943
944	inode = btrfs_iget(s: sb, BTRFS_FIRST_FREE_OBJECTID, root: fs_info->fs_root);
945	if (IS_ERR(ptr: inode)) {
946	err = PTR_ERR(ptr: inode);
947	btrfs_handle_fs_error(fs_info, err, NULL);
948	goto fail_close;
949	}
950
951	sb->s_root = d_make_root(inode);
952	if (!sb->s_root) {
953	err = -ENOMEM;
954	goto fail_close;
955	}
956
957	sb->s_flags |= SB_ACTIVE;
958	return 0;
959
960	fail_close:
961	close_ctree(fs_info);
962	return err;
963	}
964
965	int btrfs_sync_fs(struct super_block *sb, int wait)
966	{
967	struct btrfs_trans_handle *trans;
968	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
969	struct btrfs_root *root = fs_info->tree_root;
970
971	trace_btrfs_sync_fs(fs_info, wait);
972
973	if (!wait) {
974	filemap_flush(fs_info->btree_inode->i_mapping);
975	return 0;
976	}
977
978	btrfs_wait_ordered_roots(fs_info, U64_MAX, range_start: 0, range_len: (u64)-1);
979
980	trans = btrfs_attach_transaction_barrier(root);
981	if (IS_ERR(ptr: trans)) {
982	/* no transaction, don't bother */
983	if (PTR_ERR(ptr: trans) == -ENOENT) {
984	/*
985	* Exit unless we have some pending changes
986	* that need to go through commit
987	*/
988	if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
989	&fs_info->flags))
990	return 0;
991	/*
992	* A non-blocking test if the fs is frozen. We must not
993	* start a new transaction here otherwise a deadlock
994	* happens. The pending operations are delayed to the
995	* next commit after thawing.
996	*/
997	if (sb_start_write_trylock(sb))
998	sb_end_write(sb);
999	else
1000	return 0;
1001	trans = btrfs_start_transaction(root, num_items: 0);
1002	}
1003	if (IS_ERR(ptr: trans))
1004	return PTR_ERR(ptr: trans);
1005	}
1006	return btrfs_commit_transaction(trans);
1007	}
1008
1009	static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1010	{
1011	seq_printf(m: seq, fmt: "%s%s", (*printed) ? ":" : ",rescue=", s);
1012	*printed = true;
1013	}
1014
1015	static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1016	{
1017	struct btrfs_fs_info *info = btrfs_sb(sb: dentry->d_sb);
1018	const char *compress_type;
1019	const char *subvol_name;
1020	bool printed = false;
1021
1022	if (btrfs_test_opt(info, DEGRADED))
1023	seq_puts(m: seq, s: ",degraded");
1024	if (btrfs_test_opt(info, NODATASUM))
1025	seq_puts(m: seq, s: ",nodatasum");
1026	if (btrfs_test_opt(info, NODATACOW))
1027	seq_puts(m: seq, s: ",nodatacow");
1028	if (btrfs_test_opt(info, NOBARRIER))
1029	seq_puts(m: seq, s: ",nobarrier");
1030	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1031	seq_printf(m: seq, fmt: ",max_inline=%llu", info->max_inline);
1032	if (info->thread_pool_size != min_t(unsigned long,
1033	num_online_cpus() + 2, 8))
1034	seq_printf(m: seq, fmt: ",thread_pool=%u", info->thread_pool_size);
1035	if (btrfs_test_opt(info, COMPRESS)) {
1036	compress_type = btrfs_compress_type2str(type: info->compress_type);
1037	if (btrfs_test_opt(info, FORCE_COMPRESS))
1038	seq_printf(m: seq, fmt: ",compress-force=%s", compress_type);
1039	else
1040	seq_printf(m: seq, fmt: ",compress=%s", compress_type);
1041	if (info->compress_level)
1042	seq_printf(m: seq, fmt: ":%d", info->compress_level);
1043	}
1044	if (btrfs_test_opt(info, NOSSD))
1045	seq_puts(m: seq, s: ",nossd");
1046	if (btrfs_test_opt(info, SSD_SPREAD))
1047	seq_puts(m: seq, s: ",ssd_spread");
1048	else if (btrfs_test_opt(info, SSD))
1049	seq_puts(m: seq, s: ",ssd");
1050	if (btrfs_test_opt(info, NOTREELOG))
1051	seq_puts(m: seq, s: ",notreelog");
1052	if (btrfs_test_opt(info, NOLOGREPLAY))
1053	print_rescue_option(seq, s: "nologreplay", printed: &printed);
1054	if (btrfs_test_opt(info, USEBACKUPROOT))
1055	print_rescue_option(seq, s: "usebackuproot", printed: &printed);
1056	if (btrfs_test_opt(info, IGNOREBADROOTS))
1057	print_rescue_option(seq, s: "ignorebadroots", printed: &printed);
1058	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1059	print_rescue_option(seq, s: "ignoredatacsums", printed: &printed);
1060	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1061	seq_puts(m: seq, s: ",flushoncommit");
1062	if (btrfs_test_opt(info, DISCARD_SYNC))
1063	seq_puts(m: seq, s: ",discard");
1064	if (btrfs_test_opt(info, DISCARD_ASYNC))
1065	seq_puts(m: seq, s: ",discard=async");
1066	if (!(info->sb->s_flags & SB_POSIXACL))
1067	seq_puts(m: seq, s: ",noacl");
1068	if (btrfs_free_space_cache_v1_active(fs_info: info))
1069	seq_puts(m: seq, s: ",space_cache");
1070	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1071	seq_puts(m: seq, s: ",space_cache=v2");
1072	else
1073	seq_puts(m: seq, s: ",nospace_cache");
1074	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1075	seq_puts(m: seq, s: ",rescan_uuid_tree");
1076	if (btrfs_test_opt(info, CLEAR_CACHE))
1077	seq_puts(m: seq, s: ",clear_cache");
1078	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1079	seq_puts(m: seq, s: ",user_subvol_rm_allowed");
1080	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1081	seq_puts(m: seq, s: ",enospc_debug");
1082	if (btrfs_test_opt(info, AUTO_DEFRAG))
1083	seq_puts(m: seq, s: ",autodefrag");
1084	if (btrfs_test_opt(info, SKIP_BALANCE))
1085	seq_puts(m: seq, s: ",skip_balance");
1086	if (info->metadata_ratio)
1087	seq_printf(m: seq, fmt: ",metadata_ratio=%u", info->metadata_ratio);
1088	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1089	seq_puts(m: seq, s: ",fatal_errors=panic");
1090	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1091	seq_printf(m: seq, fmt: ",commit=%u", info->commit_interval);
1092	#ifdef CONFIG_BTRFS_DEBUG
1093	if (btrfs_test_opt(info, FRAGMENT_DATA))
1094	seq_puts(m: seq, s: ",fragment=data");
1095	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1096	seq_puts(m: seq, s: ",fragment=metadata");
1097	#endif
1098	if (btrfs_test_opt(info, REF_VERIFY))
1099	seq_puts(m: seq, s: ",ref_verify");
1100	seq_printf(m: seq, fmt: ",subvolid=%llu",
1101	BTRFS_I(inode: d_inode(dentry))->root->root_key.objectid);
1102	subvol_name = btrfs_get_subvol_name_from_objectid(fs_info: info,
1103	subvol_objectid: BTRFS_I(inode: d_inode(dentry))->root->root_key.objectid);
1104	if (!IS_ERR(ptr: subvol_name)) {
1105	seq_puts(m: seq, s: ",subvol=");
1106	seq_escape(m: seq, s: subvol_name, esc: " \t\n\\");
1107	kfree(objp: subvol_name);
1108	}
1109	return 0;
1110	}
1111
1112	/*
1113	* subvolumes are identified by ino 256
1114	*/
1115	static inline int is_subvolume_inode(struct inode *inode)
1116	{
1117	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1118	return 1;
1119	return 0;
1120	}
1121
1122	static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1123	struct vfsmount *mnt)
1124	{
1125	struct dentry *root;
1126	int ret;
1127
1128	if (!subvol_name) {
1129	if (!subvol_objectid) {
1130	ret = get_default_subvol_objectid(fs_info: btrfs_sb(sb: mnt->mnt_sb),
1131	objectid: &subvol_objectid);
1132	if (ret) {
1133	root = ERR_PTR(error: ret);
1134	goto out;
1135	}
1136	}
1137	subvol_name = btrfs_get_subvol_name_from_objectid(
1138	fs_info: btrfs_sb(sb: mnt->mnt_sb), subvol_objectid);
1139	if (IS_ERR(ptr: subvol_name)) {
1140	root = ERR_CAST(ptr: subvol_name);
1141	subvol_name = NULL;
1142	goto out;
1143	}
1144
1145	}
1146
1147	root = mount_subtree(mnt, path: subvol_name);
1148	/* mount_subtree() drops our reference on the vfsmount. */
1149	mnt = NULL;
1150
1151	if (!IS_ERR(ptr: root)) {
1152	struct super_block *s = root->d_sb;
1153	struct btrfs_fs_info *fs_info = btrfs_sb(sb: s);
1154	struct inode *root_inode = d_inode(dentry: root);
1155	u64 root_objectid = BTRFS_I(inode: root_inode)->root->root_key.objectid;
1156
1157	ret = 0;
1158	if (!is_subvolume_inode(inode: root_inode)) {
1159	btrfs_err(fs_info, "'%s' is not a valid subvolume",
1160	subvol_name);
1161	ret = -EINVAL;
1162	}
1163	if (subvol_objectid && root_objectid != subvol_objectid) {
1164	/*
1165	* This will also catch a race condition where a
1166	* subvolume which was passed by ID is renamed and
1167	* another subvolume is renamed over the old location.
1168	*/
1169	btrfs_err(fs_info,
1170	"subvol '%s' does not match subvolid %llu",
1171	subvol_name, subvol_objectid);
1172	ret = -EINVAL;
1173	}
1174	if (ret) {
1175	dput(root);
1176	root = ERR_PTR(error: ret);
1177	deactivate_locked_super(sb: s);
1178	}
1179	}
1180
1181	out:
1182	mntput(mnt);
1183	kfree(objp: subvol_name);
1184	return root;
1185	}
1186
1187	static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1188	u32 new_pool_size, u32 old_pool_size)
1189	{
1190	if (new_pool_size == old_pool_size)
1191	return;
1192
1193	fs_info->thread_pool_size = new_pool_size;
1194
1195	btrfs_info(fs_info, "resize thread pool %d -> %d",
1196	old_pool_size, new_pool_size);
1197
1198	btrfs_workqueue_set_max(wq: fs_info->workers, max: new_pool_size);
1199	btrfs_workqueue_set_max(wq: fs_info->delalloc_workers, max: new_pool_size);
1200	btrfs_workqueue_set_max(wq: fs_info->caching_workers, max: new_pool_size);
1201	workqueue_set_max_active(wq: fs_info->endio_workers, max_active: new_pool_size);
1202	workqueue_set_max_active(wq: fs_info->endio_meta_workers, max_active: new_pool_size);
1203	btrfs_workqueue_set_max(wq: fs_info->endio_write_workers, max: new_pool_size);
1204	btrfs_workqueue_set_max(wq: fs_info->endio_freespace_worker, max: new_pool_size);
1205	btrfs_workqueue_set_max(wq: fs_info->delayed_workers, max: new_pool_size);
1206	}
1207
1208	static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1209	unsigned long old_opts, int flags)
1210	{
1211	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1212	(!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1213	(flags & SB_RDONLY))) {
1214	/* wait for any defraggers to finish */
1215	wait_event(fs_info->transaction_wait,
1216	(atomic_read(&fs_info->defrag_running) == 0));
1217	if (flags & SB_RDONLY)
1218	sync_filesystem(fs_info->sb);
1219	}
1220	}
1221
1222	static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1223	unsigned long old_opts)
1224	{
1225	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1226
1227	/*
1228	* We need to cleanup all defragable inodes if the autodefragment is
1229	* close or the filesystem is read only.
1230	*/
1231	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1232	(!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(sb: fs_info->sb))) {
1233	btrfs_cleanup_defrag_inodes(fs_info);
1234	}
1235
1236	/* If we toggled discard async */
1237	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1238	btrfs_test_opt(fs_info, DISCARD_ASYNC))
1239	btrfs_discard_resume(fs_info);
1240	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1241	!btrfs_test_opt(fs_info, DISCARD_ASYNC))
1242	btrfs_discard_cleanup(fs_info);
1243
1244	/* If we toggled space cache */
1245	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1246	btrfs_set_free_space_cache_v1_active(fs_info, active: cache_opt);
1247	}
1248
1249	static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1250	{
1251	int ret;
1252
1253	if (BTRFS_FS_ERROR(fs_info)) {
1254	btrfs_err(fs_info,
1255	"remounting read-write after error is not allowed");
1256	return -EINVAL;
1257	}
1258
1259	if (fs_info->fs_devices->rw_devices == 0)
1260	return -EACCES;
1261
1262	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1263	btrfs_warn(fs_info,
1264	"too many missing devices, writable remount is not allowed");
1265	return -EACCES;
1266	}
1267
1268	if (btrfs_super_log_root(s: fs_info->super_copy) != 0) {
1269	btrfs_warn(fs_info,
1270	"mount required to replay tree-log, cannot remount read-write");
1271	return -EINVAL;
1272	}
1273
1274	/*
1275	* NOTE: when remounting with a change that does writes, don't put it
1276	* anywhere above this point, as we are not sure to be safe to write
1277	* until we pass the above checks.
1278	*/
1279	ret = btrfs_start_pre_rw_mount(fs_info);
1280	if (ret)
1281	return ret;
1282
1283	btrfs_clear_sb_rdonly(sb: fs_info->sb);
1284
1285	set_bit(nr: BTRFS_FS_OPEN, addr: &fs_info->flags);
1286
1287	/*
1288	* If we've gone from readonly -> read-write, we need to get our
1289	* sync/async discard lists in the right state.
1290	*/
1291	btrfs_discard_resume(fs_info);
1292
1293	return 0;
1294	}
1295
1296	static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1297	{
1298	/*
1299	* This also happens on 'umount -rf' or on shutdown, when the
1300	* filesystem is busy.
1301	*/
1302	cancel_work_sync(work: &fs_info->async_reclaim_work);
1303	cancel_work_sync(work: &fs_info->async_data_reclaim_work);
1304
1305	btrfs_discard_cleanup(fs_info);
1306
1307	/* Wait for the uuid_scan task to finish */
1308	down(sem: &fs_info->uuid_tree_rescan_sem);
1309	/* Avoid complains from lockdep et al. */
1310	up(sem: &fs_info->uuid_tree_rescan_sem);
1311
1312	btrfs_set_sb_rdonly(sb: fs_info->sb);
1313
1314	/*
1315	* Setting SB_RDONLY will put the cleaner thread to sleep at the next
1316	* loop if it's already active. If it's already asleep, we'll leave
1317	* unused block groups on disk until we're mounted read-write again
1318	* unless we clean them up here.
1319	*/
1320	btrfs_delete_unused_bgs(fs_info);
1321
1322	/*
1323	* The cleaner task could be already running before we set the flag
1324	* BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make
1325	* sure that after we finish the remount, i.e. after we call
1326	* btrfs_commit_super(), the cleaner can no longer start a transaction
1327	* - either because it was dropping a dead root, running delayed iputs
1328	* or deleting an unused block group (the cleaner picked a block
1329	* group from the list of unused block groups before we were able to
1330	* in the previous call to btrfs_delete_unused_bgs()).
1331	*/
1332	wait_on_bit(word: &fs_info->flags, bit: BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1333
1334	/*
1335	* We've set the superblock to RO mode, so we might have made the
1336	* cleaner task sleep without running all pending delayed iputs. Go
1337	* through all the delayed iputs here, so that if an unmount happens
1338	* without remounting RW we don't end up at finishing close_ctree()
1339	* with a non-empty list of delayed iputs.
1340	*/
1341	btrfs_run_delayed_iputs(fs_info);
1342
1343	btrfs_dev_replace_suspend_for_unmount(fs_info);
1344	btrfs_scrub_cancel(info: fs_info);
1345	btrfs_pause_balance(fs_info);
1346
1347	/*
1348	* Pause the qgroup rescan worker if it is running. We don't want it to
1349	* be still running after we are in RO mode, as after that, by the time
1350	* we unmount, it might have left a transaction open, so we would leak
1351	* the transaction and/or crash.
1352	*/
1353	btrfs_qgroup_wait_for_completion(fs_info, interruptible: false);
1354
1355	return btrfs_commit_super(fs_info);
1356	}
1357
1358	static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1359	{
1360	fs_info->max_inline = ctx->max_inline;
1361	fs_info->commit_interval = ctx->commit_interval;
1362	fs_info->metadata_ratio = ctx->metadata_ratio;
1363	fs_info->thread_pool_size = ctx->thread_pool_size;
1364	fs_info->mount_opt = ctx->mount_opt;
1365	fs_info->compress_type = ctx->compress_type;
1366	fs_info->compress_level = ctx->compress_level;
1367	}
1368
1369	static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1370	{
1371	ctx->max_inline = fs_info->max_inline;
1372	ctx->commit_interval = fs_info->commit_interval;
1373	ctx->metadata_ratio = fs_info->metadata_ratio;
1374	ctx->thread_pool_size = fs_info->thread_pool_size;
1375	ctx->mount_opt = fs_info->mount_opt;
1376	ctx->compress_type = fs_info->compress_type;
1377	ctx->compress_level = fs_info->compress_level;
1378	}
1379
1380	#define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \
1381	do { \
1382	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1383	btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1384	btrfs_info(fs_info, fmt, ##args); \
1385	} while (0)
1386
1387	#define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \
1388	do { \
1389	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \
1390	!btrfs_raw_test_opt(fs_info->mount_opt, opt)) \
1391	btrfs_info(fs_info, fmt, ##args); \
1392	} while (0)
1393
1394	static void btrfs_emit_options(struct btrfs_fs_info *info,
1395	struct btrfs_fs_context *old)
1396	{
1397	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1398	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1399	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1400	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1401	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1402	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1403	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1404	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1405	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1406	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1407	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1408	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1409	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1410	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1411	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1412	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1413	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1414	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1415	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1416	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1417	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1418
1419	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1420	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1421	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1422	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1423	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1424	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1425	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1426	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1427	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1428
1429	/* Did the compression settings change? */
1430	if (btrfs_test_opt(info, COMPRESS) &&
1431	(!old ||
1432	old->compress_type != info->compress_type ||
1433	old->compress_level != info->compress_level ||
1434	(!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1435	btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1436	const char *compress_type = btrfs_compress_type2str(type: info->compress_type);
1437
1438	btrfs_info(info, "%s %s compression, level %d",
1439	btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1440	compress_type, info->compress_level);
1441	}
1442
1443	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1444	btrfs_info(info, "max_inline set to %llu", info->max_inline);
1445	}
1446
1447	static int btrfs_reconfigure(struct fs_context *fc)
1448	{
1449	struct super_block *sb = fc->root->d_sb;
1450	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1451	struct btrfs_fs_context *ctx = fc->fs_private;
1452	struct btrfs_fs_context old_ctx;
1453	int ret = 0;
1454	bool mount_reconfigure = (fc->s_fs_info != NULL);
1455
1456	btrfs_info_to_ctx(fs_info, ctx: &old_ctx);
1457
1458	/*
1459	* This is our "bind mount" trick, we don't want to allow the user to do
1460	* anything other than mount a different ro/rw and a different subvol,
1461	* all of the mount options should be maintained.
1462	*/
1463	if (mount_reconfigure)
1464	ctx->mount_opt = old_ctx.mount_opt;
1465
1466	sync_filesystem(sb);
1467	set_bit(nr: BTRFS_FS_STATE_REMOUNTING, addr: &fs_info->fs_state);
1468
1469	if (!mount_reconfigure &&
1470	!btrfs_check_options(info: fs_info, mount_opt: &ctx->mount_opt, flags: fc->sb_flags))
1471	return -EINVAL;
1472
1473	ret = btrfs_check_features(fs_info, is_rw_mount: !(fc->sb_flags & SB_RDONLY));
1474	if (ret < 0)
1475	return ret;
1476
1477	btrfs_ctx_to_info(fs_info, ctx);
1478	btrfs_remount_begin(fs_info, old_opts: old_ctx.mount_opt, flags: fc->sb_flags);
1479	btrfs_resize_thread_pool(fs_info, new_pool_size: fs_info->thread_pool_size,
1480	old_pool_size: old_ctx.thread_pool_size);
1481
1482	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1483	(bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1484	(!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1485	btrfs_warn(fs_info,
1486	"remount supports changing free space tree only from RO to RW");
1487	/* Make sure free space cache options match the state on disk. */
1488	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1489	btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1490	btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1491	}
1492	if (btrfs_free_space_cache_v1_active(fs_info)) {
1493	btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1494	btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1495	}
1496	}
1497
1498	ret = 0;
1499	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1500	ret = btrfs_remount_ro(fs_info);
1501	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1502	ret = btrfs_remount_rw(fs_info);
1503	if (ret)
1504	goto restore;
1505
1506	/*
1507	* If we set the mask during the parameter parsing VFS would reject the
1508	* remount. Here we can set the mask and the value will be updated
1509	* appropriately.
1510	*/
1511	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1512	fc->sb_flags_mask |= SB_POSIXACL;
1513
1514	btrfs_emit_options(info: fs_info, old: &old_ctx);
1515	wake_up_process(tsk: fs_info->transaction_kthread);
1516	btrfs_remount_cleanup(fs_info, old_opts: old_ctx.mount_opt);
1517	btrfs_clear_oneshot_options(fs_info);
1518	clear_bit(nr: BTRFS_FS_STATE_REMOUNTING, addr: &fs_info->fs_state);
1519
1520	return 0;
1521	restore:
1522	btrfs_ctx_to_info(fs_info, ctx: &old_ctx);
1523	btrfs_remount_cleanup(fs_info, old_opts: old_ctx.mount_opt);
1524	clear_bit(nr: BTRFS_FS_STATE_REMOUNTING, addr: &fs_info->fs_state);
1525	return ret;
1526	}
1527
1528	/* Used to sort the devices by max_avail(descending sort) */
1529	static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1530	{
1531	const struct btrfs_device_info *dev_info1 = a;
1532	const struct btrfs_device_info *dev_info2 = b;
1533
1534	if (dev_info1->max_avail > dev_info2->max_avail)
1535	return -1;
1536	else if (dev_info1->max_avail < dev_info2->max_avail)
1537	return 1;
1538	return 0;
1539	}
1540
1541	/*
1542	* sort the devices by max_avail, in which max free extent size of each device
1543	* is stored.(Descending Sort)
1544	*/
1545	static inline void btrfs_descending_sort_devices(
1546	struct btrfs_device_info *devices,
1547	size_t nr_devices)
1548	{
1549	sort(base: devices, num: nr_devices, size: sizeof(struct btrfs_device_info),
1550	cmp_func: btrfs_cmp_device_free_bytes, NULL);
1551	}
1552
1553	/*
1554	* The helper to calc the free space on the devices that can be used to store
1555	* file data.
1556	*/
1557	static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1558	u64 *free_bytes)
1559	{
1560	struct btrfs_device_info *devices_info;
1561	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1562	struct btrfs_device *device;
1563	u64 type;
1564	u64 avail_space;
1565	u64 min_stripe_size;
1566	int num_stripes = 1;
1567	int i = 0, nr_devices;
1568	const struct btrfs_raid_attr *rattr;
1569
1570	/*
1571	* We aren't under the device list lock, so this is racy-ish, but good
1572	* enough for our purposes.
1573	*/
1574	nr_devices = fs_info->fs_devices->open_devices;
1575	if (!nr_devices) {
1576	smp_mb();
1577	nr_devices = fs_info->fs_devices->open_devices;
1578	ASSERT(nr_devices);
1579	if (!nr_devices) {
1580	*free_bytes = 0;
1581	return 0;
1582	}
1583	}
1584
1585	devices_info = kmalloc_array(n: nr_devices, size: sizeof(*devices_info),
1586	GFP_KERNEL);
1587	if (!devices_info)
1588	return -ENOMEM;
1589
1590	/* calc min stripe number for data space allocation */
1591	type = btrfs_data_alloc_profile(fs_info);
1592	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(flags: type)];
1593
1594	if (type & BTRFS_BLOCK_GROUP_RAID0)
1595	num_stripes = nr_devices;
1596	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1597	num_stripes = rattr->ncopies;
1598	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1599	num_stripes = 4;
1600
1601	/* Adjust for more than 1 stripe per device */
1602	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1603
1604	rcu_read_lock();
1605	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1606	if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1607	&device->dev_state) ||
1608	!device->bdev ||
1609	test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1610	continue;
1611
1612	if (i >= nr_devices)
1613	break;
1614
1615	avail_space = device->total_bytes - device->bytes_used;
1616
1617	/* align with stripe_len */
1618	avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1619
1620	/*
1621	* Ensure we have at least min_stripe_size on top of the
1622	* reserved space on the device.
1623	*/
1624	if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1625	continue;
1626
1627	avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1628
1629	devices_info[i].dev = device;
1630	devices_info[i].max_avail = avail_space;
1631
1632	i++;
1633	}
1634	rcu_read_unlock();
1635
1636	nr_devices = i;
1637
1638	btrfs_descending_sort_devices(devices: devices_info, nr_devices);
1639
1640	i = nr_devices - 1;
1641	avail_space = 0;
1642	while (nr_devices >= rattr->devs_min) {
1643	num_stripes = min(num_stripes, nr_devices);
1644
1645	if (devices_info[i].max_avail >= min_stripe_size) {
1646	int j;
1647	u64 alloc_size;
1648
1649	avail_space += devices_info[i].max_avail * num_stripes;
1650	alloc_size = devices_info[i].max_avail;
1651	for (j = i + 1 - num_stripes; j <= i; j++)
1652	devices_info[j].max_avail -= alloc_size;
1653	}
1654	i--;
1655	nr_devices--;
1656	}
1657
1658	kfree(objp: devices_info);
1659	*free_bytes = avail_space;
1660	return 0;
1661	}
1662
1663	/*
1664	* Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1665	*
1666	* If there's a redundant raid level at DATA block groups, use the respective
1667	* multiplier to scale the sizes.
1668	*
1669	* Unused device space usage is based on simulating the chunk allocator
1670	* algorithm that respects the device sizes and order of allocations. This is
1671	* a close approximation of the actual use but there are other factors that may
1672	* change the result (like a new metadata chunk).
1673	*
1674	* If metadata is exhausted, f_bavail will be 0.
1675	*/
1676	static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1677	{
1678	struct btrfs_fs_info *fs_info = btrfs_sb(sb: dentry->d_sb);
1679	struct btrfs_super_block *disk_super = fs_info->super_copy;
1680	struct btrfs_space_info *found;
1681	u64 total_used = 0;
1682	u64 total_free_data = 0;
1683	u64 total_free_meta = 0;
1684	u32 bits = fs_info->sectorsize_bits;
1685	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1686	unsigned factor = 1;
1687	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1688	int ret;
1689	u64 thresh = 0;
1690	int mixed = 0;
1691
1692	list_for_each_entry(found, &fs_info->space_info, list) {
1693	if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1694	int i;
1695
1696	total_free_data += found->disk_total - found->disk_used;
1697	total_free_data -=
1698	btrfs_account_ro_block_groups_free_space(sinfo: found);
1699
1700	for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1701	if (!list_empty(head: &found->block_groups[i]))
1702	factor = btrfs_bg_type_to_factor(
1703	flags: btrfs_raid_array[i].bg_flag);
1704	}
1705	}
1706
1707	/*
1708	* Metadata in mixed block group profiles are accounted in data
1709	*/
1710	if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1711	if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1712	mixed = 1;
1713	else
1714	total_free_meta += found->disk_total -
1715	found->disk_used;
1716	}
1717
1718	total_used += found->disk_used;
1719	}
1720
1721	buf->f_blocks = div_u64(dividend: btrfs_super_total_bytes(s: disk_super), divisor: factor);
1722	buf->f_blocks >>= bits;
1723	buf->f_bfree = buf->f_blocks - (div_u64(dividend: total_used, divisor: factor) >> bits);
1724
1725	/* Account global block reserve as used, it's in logical size already */
1726	spin_lock(lock: &block_rsv->lock);
1727	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1728	if (buf->f_bfree >= block_rsv->size >> bits)
1729	buf->f_bfree -= block_rsv->size >> bits;
1730	else
1731	buf->f_bfree = 0;
1732	spin_unlock(lock: &block_rsv->lock);
1733
1734	buf->f_bavail = div_u64(dividend: total_free_data, divisor: factor);
1735	ret = btrfs_calc_avail_data_space(fs_info, free_bytes: &total_free_data);
1736	if (ret)
1737	return ret;
1738	buf->f_bavail += div_u64(dividend: total_free_data, divisor: factor);
1739	buf->f_bavail = buf->f_bavail >> bits;
1740
1741	/*
1742	* We calculate the remaining metadata space minus global reserve. If
1743	* this is (supposedly) smaller than zero, there's no space. But this
1744	* does not hold in practice, the exhausted state happens where's still
1745	* some positive delta. So we apply some guesswork and compare the
1746	* delta to a 4M threshold. (Practically observed delta was ~2M.)
1747	*
1748	* We probably cannot calculate the exact threshold value because this
1749	* depends on the internal reservations requested by various
1750	* operations, so some operations that consume a few metadata will
1751	* succeed even if the Avail is zero. But this is better than the other
1752	* way around.
1753	*/
1754	thresh = SZ_4M;
1755
1756	/*
1757	* We only want to claim there's no available space if we can no longer
1758	* allocate chunks for our metadata profile and our global reserve will
1759	* not fit in the free metadata space. If we aren't ->full then we
1760	* still can allocate chunks and thus are fine using the currently
1761	* calculated f_bavail.
1762	*/
1763	if (!mixed && block_rsv->space_info->full &&
1764	(total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1765	buf->f_bavail = 0;
1766
1767	buf->f_type = BTRFS_SUPER_MAGIC;
1768	buf->f_bsize = fs_info->sectorsize;
1769	buf->f_namelen = BTRFS_NAME_LEN;
1770
1771	/* We treat it as constant endianness (it doesn't matter _which_)
1772	because we want the fsid to come out the same whether mounted
1773	on a big-endian or little-endian host */
1774	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1775	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1776	/* Mask in the root object ID too, to disambiguate subvols */
1777	buf->f_fsid.val[0] ^=
1778	BTRFS_I(inode: d_inode(dentry))->root->root_key.objectid >> 32;
1779	buf->f_fsid.val[1] ^=
1780	BTRFS_I(inode: d_inode(dentry))->root->root_key.objectid;
1781
1782	return 0;
1783	}
1784
1785	static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1786	{
1787	struct btrfs_fs_info *p = fc->s_fs_info;
1788	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1789
1790	return fs_info->fs_devices == p->fs_devices;
1791	}
1792
1793	static int btrfs_get_tree_super(struct fs_context *fc)
1794	{
1795	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1796	struct btrfs_fs_context *ctx = fc->fs_private;
1797	struct btrfs_fs_devices *fs_devices = NULL;
1798	struct block_device *bdev;
1799	struct btrfs_device *device;
1800	struct super_block *sb;
1801	blk_mode_t mode = btrfs_open_mode(fc);
1802	int ret;
1803
1804	btrfs_ctx_to_info(fs_info, ctx);
1805	mutex_lock(&uuid_mutex);
1806
1807	/*
1808	* With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1809	* either a valid device or an error.
1810	*/
1811	device = btrfs_scan_one_device(path: fc->source, flags: mode, mount_arg_dev: true);
1812	ASSERT(device != NULL);
1813	if (IS_ERR(ptr: device)) {
1814	mutex_unlock(lock: &uuid_mutex);
1815	return PTR_ERR(ptr: device);
1816	}
1817
1818	fs_devices = device->fs_devices;
1819	fs_info->fs_devices = fs_devices;
1820
1821	ret = btrfs_open_devices(fs_devices, flags: mode, holder: &btrfs_fs_type);
1822	mutex_unlock(lock: &uuid_mutex);
1823	if (ret)
1824	return ret;
1825
1826	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1827	ret = -EACCES;
1828	goto error;
1829	}
1830
1831	bdev = fs_devices->latest_dev->bdev;
1832
1833	/*
1834	* From now on the error handling is not straightforward.
1835	*
1836	* If successful, this will transfer the fs_info into the super block,
1837	* and fc->s_fs_info will be NULL. However if there's an existing
1838	* super, we'll still have fc->s_fs_info populated. If we error
1839	* completely out it'll be cleaned up when we drop the fs_context,
1840	* otherwise it's tied to the lifetime of the super_block.
1841	*/
1842	sb = sget_fc(fc, test: btrfs_fc_test_super, set: set_anon_super_fc);
1843	if (IS_ERR(ptr: sb)) {
1844	ret = PTR_ERR(ptr: sb);
1845	goto error;
1846	}
1847
1848	set_device_specific_options(fs_info);
1849
1850	if (sb->s_root) {
1851	btrfs_close_devices(fs_devices);
1852	if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1853	ret = -EBUSY;
1854	} else {
1855	snprintf(buf: sb->s_id, size: sizeof(sb->s_id), fmt: "%pg", bdev);
1856	shrinker_debugfs_rename(shrinker: sb->s_shrink, fmt: "sb-btrfs:%s", sb->s_id);
1857	btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1858	ret = btrfs_fill_super(sb, fs_devices, NULL);
1859	}
1860
1861	if (ret) {
1862	deactivate_locked_super(sb);
1863	return ret;
1864	}
1865
1866	btrfs_clear_oneshot_options(fs_info);
1867
1868	fc->root = dget(dentry: sb->s_root);
1869	return 0;
1870
1871	error:
1872	btrfs_close_devices(fs_devices);
1873	return ret;
1874	}
1875
1876	/*
1877	* Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1878	* with different ro/rw options") the following works:
1879	*
1880	* (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1881	* (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1882	*
1883	* which looks nice and innocent but is actually pretty intricate and deserves
1884	* a long comment.
1885	*
1886	* On another filesystem a subvolume mount is close to something like:
1887	*
1888	* (iii) # create rw superblock + initial mount
1889	* mount -t xfs /dev/sdb /opt/
1890	*
1891	* # create ro bind mount
1892	* mount --bind -o ro /opt/foo /mnt/foo
1893	*
1894	* # unmount initial mount
1895	* umount /opt
1896	*
1897	* Of course, there's some special subvolume sauce and there's the fact that the
1898	* sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1899	* it's very close and will help us understand the issue.
1900	*
1901	* The old mount API didn't cleanly distinguish between a mount being made ro
1902	* and a superblock being made ro. The only way to change the ro state of
1903	* either object was by passing ms_rdonly. If a new mount was created via
1904	* mount(2) such as:
1905	*
1906	* mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1907	*
1908	* the MS_RDONLY flag being specified had two effects:
1909	*
1910	* (1) MNT_READONLY was raised -> the resulting mount got
1911	* @mnt->mnt_flags |= MNT_READONLY raised.
1912	*
1913	* (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1914	* made the superblock ro. Note, how SB_RDONLY has the same value as
1915	* ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1916	*
1917	* Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1918	* subtree mounted ro.
1919	*
1920	* But consider the effect on the old mount API on btrfs subvolume mounting
1921	* which combines the distinct step in (iii) into a single step.
1922	*
1923	* By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1924	* is issued the superblock is ro and thus even if the mount created for (ii) is
1925	* rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1926	* to rw for (ii) which it did using an internal remount call.
1927	*
1928	* IOW, subvolume mounting was inherently complicated due to the ambiguity of
1929	* MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1930	* "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1931	* passed by mount(8) to mount(2).
1932	*
1933	* Enter the new mount API. The new mount API disambiguates making a mount ro
1934	* and making a superblock ro.
1935	*
1936	* (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1937	* fsmount() or mount_setattr() this is a pure VFS level change for a
1938	* specific mount or mount tree that is never seen by the filesystem itself.
1939	*
1940	* (4) To turn a superblock ro the "ro" flag must be used with
1941	* fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1942	* in fc->sb_flags.
1943	*
1944	* This disambiguation has rather positive consequences. Mounting a subvolume
1945	* ro will not also turn the superblock ro. Only the mount for the subvolume
1946	* will become ro.
1947	*
1948	* So, if the superblock creation request comes from the new mount API the
1949	* caller must have explicitly done:
1950	*
1951	* fsconfig(FSCONFIG_SET_FLAG, "ro")
1952	* fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1953	*
1954	* IOW, at some point the caller must have explicitly turned the whole
1955	* superblock ro and we shouldn't just undo it like we did for the old mount
1956	* API. In any case, it lets us avoid the hack in the new mount API.
1957	*
1958	* Consequently, the remounting hack must only be used for requests originating
1959	* from the old mount API and should be marked for full deprecation so it can be
1960	* turned off in a couple of years.
1961	*
1962	* The new mount API has no reason to support this hack.
1963	*/
1964	static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
1965	{
1966	struct vfsmount *mnt;
1967	int ret;
1968	const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
1969
1970	/*
1971	* We got an EBUSY because our SB_RDONLY flag didn't match the existing
1972	* super block, so invert our setting here and retry the mount so we
1973	* can get our vfsmount.
1974	*/
1975	if (ro2rw)
1976	fc->sb_flags |= SB_RDONLY;
1977	else
1978	fc->sb_flags &= ~SB_RDONLY;
1979
1980	mnt = fc_mount(fc);
1981	if (IS_ERR(ptr: mnt))
1982	return mnt;
1983
1984	if (!fc->oldapi || !ro2rw)
1985	return mnt;
1986
1987	/* We need to convert to rw, call reconfigure. */
1988	fc->sb_flags &= ~SB_RDONLY;
1989	down_write(sem: &mnt->mnt_sb->s_umount);
1990	ret = btrfs_reconfigure(fc);
1991	up_write(sem: &mnt->mnt_sb->s_umount);
1992	if (ret) {
1993	mntput(mnt);
1994	return ERR_PTR(error: ret);
1995	}
1996	return mnt;
1997	}
1998
1999	static int btrfs_get_tree_subvol(struct fs_context *fc)
2000	{
2001	struct btrfs_fs_info *fs_info = NULL;
2002	struct btrfs_fs_context *ctx = fc->fs_private;
2003	struct fs_context *dup_fc;
2004	struct dentry *dentry;
2005	struct vfsmount *mnt;
2006
2007	/*
2008	* Setup a dummy root and fs_info for test/set super. This is because
2009	* we don't actually fill this stuff out until open_ctree, but we need
2010	* then open_ctree will properly initialize the file system specific
2011	* settings later. btrfs_init_fs_info initializes the static elements
2012	* of the fs_info (locks and such) to make cleanup easier if we find a
2013	* superblock with our given fs_devices later on at sget() time.
2014	*/
2015	fs_info = kvzalloc(size: sizeof(struct btrfs_fs_info), GFP_KERNEL);
2016	if (!fs_info)
2017	return -ENOMEM;
2018
2019	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2020	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2021	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2022	btrfs_free_fs_info(fs_info);
2023	return -ENOMEM;
2024	}
2025	btrfs_init_fs_info(fs_info);
2026
2027	dup_fc = vfs_dup_fs_context(fc);
2028	if (IS_ERR(ptr: dup_fc)) {
2029	btrfs_free_fs_info(fs_info);
2030	return PTR_ERR(ptr: dup_fc);
2031	}
2032
2033	/*
2034	* When we do the sget_fc this gets transferred to the sb, so we only
2035	* need to set it on the dup_fc as this is what creates the super block.
2036	*/
2037	dup_fc->s_fs_info = fs_info;
2038
2039	/*
2040	* We'll do the security settings in our btrfs_get_tree_super() mount
2041	* loop, they were duplicated into dup_fc, we can drop the originals
2042	* here.
2043	*/
2044	security_free_mnt_opts(mnt_opts: &fc->security);
2045	fc->security = NULL;
2046
2047	mnt = fc_mount(fc: dup_fc);
2048	if (PTR_ERR_OR_ZERO(ptr: mnt) == -EBUSY)
2049	mnt = btrfs_reconfigure_for_mount(fc: dup_fc);
2050	put_fs_context(fc: dup_fc);
2051	if (IS_ERR(ptr: mnt))
2052	return PTR_ERR(ptr: mnt);
2053
2054	/*
2055	* This free's ->subvol_name, because if it isn't set we have to
2056	* allocate a buffer to hold the subvol_name, so we just drop our
2057	* reference to it here.
2058	*/
2059	dentry = mount_subvol(subvol_name: ctx->subvol_name, subvol_objectid: ctx->subvol_objectid, mnt);
2060	ctx->subvol_name = NULL;
2061	if (IS_ERR(ptr: dentry))
2062	return PTR_ERR(ptr: dentry);
2063
2064	fc->root = dentry;
2065	return 0;
2066	}
2067
2068	static int btrfs_get_tree(struct fs_context *fc)
2069	{
2070	/*
2071	* Since we use mount_subtree to mount the default/specified subvol, we
2072	* have to do mounts in two steps.
2073	*
2074	* First pass through we call btrfs_get_tree_subvol(), this is just a
2075	* wrapper around fc_mount() to call back into here again, and this time
2076	* we'll call btrfs_get_tree_super(). This will do the open_ctree() and
2077	* everything to open the devices and file system. Then we return back
2078	* with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2079	* from there we can do our mount_subvol() call, which will lookup
2080	* whichever subvol we're mounting and setup this fc with the
2081	* appropriate dentry for the subvol.
2082	*/
2083	if (fc->s_fs_info)
2084	return btrfs_get_tree_super(fc);
2085	return btrfs_get_tree_subvol(fc);
2086	}
2087
2088	static void btrfs_kill_super(struct super_block *sb)
2089	{
2090	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2091	kill_anon_super(sb);
2092	btrfs_free_fs_info(fs_info);
2093	}
2094
2095	static void btrfs_free_fs_context(struct fs_context *fc)
2096	{
2097	struct btrfs_fs_context *ctx = fc->fs_private;
2098	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2099
2100	if (fs_info)
2101	btrfs_free_fs_info(fs_info);
2102
2103	if (ctx && refcount_dec_and_test(r: &ctx->refs)) {
2104	kfree(objp: ctx->subvol_name);
2105	kfree(objp: ctx);
2106	}
2107	}
2108
2109	static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2110	{
2111	struct btrfs_fs_context *ctx = src_fc->fs_private;
2112
2113	/*
2114	* Give a ref to our ctx to this dup, as we want to keep it around for
2115	* our original fc so we can have the subvolume name or objectid.
2116	*
2117	* We unset ->source in the original fc because the dup needs it for
2118	* mounting, and then once we free the dup it'll free ->source, so we
2119	* need to make sure we're only pointing to it in one fc.
2120	*/
2121	refcount_inc(r: &ctx->refs);
2122	fc->fs_private = ctx;
2123	fc->source = src_fc->source;
2124	src_fc->source = NULL;
2125	return 0;
2126	}
2127
2128	static const struct fs_context_operations btrfs_fs_context_ops = {
2129	.parse_param = btrfs_parse_param,
2130	.reconfigure = btrfs_reconfigure,
2131	.get_tree = btrfs_get_tree,
2132	.dup = btrfs_dup_fs_context,
2133	.free = btrfs_free_fs_context,
2134	};
2135
2136	static int btrfs_init_fs_context(struct fs_context *fc)
2137	{
2138	struct btrfs_fs_context *ctx;
2139
2140	ctx = kzalloc(size: sizeof(struct btrfs_fs_context), GFP_KERNEL);
2141	if (!ctx)
2142	return -ENOMEM;
2143
2144	refcount_set(r: &ctx->refs, n: 1);
2145	fc->fs_private = ctx;
2146	fc->ops = &btrfs_fs_context_ops;
2147
2148	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2149	btrfs_info_to_ctx(fs_info: btrfs_sb(sb: fc->root->d_sb), ctx);
2150	} else {
2151	ctx->thread_pool_size =
2152	min_t(unsigned long, num_online_cpus() + 2, 8);
2153	ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2154	ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2155	}
2156
2157	#ifdef CONFIG_BTRFS_FS_POSIX_ACL
2158	fc->sb_flags |= SB_POSIXACL;
2159	#endif
2160	fc->sb_flags |= SB_I_VERSION;
2161
2162	return 0;
2163	}
2164
2165	static struct file_system_type btrfs_fs_type = {
2166	.owner = THIS_MODULE,
2167	.name = "btrfs",
2168	.init_fs_context = btrfs_init_fs_context,
2169	.parameters = btrfs_fs_parameters,
2170	.kill_sb = btrfs_kill_super,
2171	.fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2172	};
2173
2174	MODULE_ALIAS_FS("btrfs");
2175
2176	static int btrfs_control_open(struct inode *inode, struct file *file)
2177	{
2178	/*
2179	* The control file's private_data is used to hold the
2180	* transaction when it is started and is used to keep
2181	* track of whether a transaction is already in progress.
2182	*/
2183	file->private_data = NULL;
2184	return 0;
2185	}
2186
2187	/*
2188	* Used by /dev/btrfs-control for devices ioctls.
2189	*/
2190	static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2191	unsigned long arg)
2192	{
2193	struct btrfs_ioctl_vol_args *vol;
2194	struct btrfs_device *device = NULL;
2195	dev_t devt = 0;
2196	int ret = -ENOTTY;
2197
2198	if (!capable(CAP_SYS_ADMIN))
2199	return -EPERM;
2200
2201	vol = memdup_user((void __user *)arg, sizeof(*vol));
2202	if (IS_ERR(ptr: vol))
2203	return PTR_ERR(ptr: vol);
2204	ret = btrfs_check_ioctl_vol_args_path(vol_args: vol);
2205	if (ret < 0)
2206	goto out;
2207
2208	switch (cmd) {
2209	case BTRFS_IOC_SCAN_DEV:
2210	mutex_lock(&uuid_mutex);
2211	/*
2212	* Scanning outside of mount can return NULL which would turn
2213	* into 0 error code.
2214	*/
2215	device = btrfs_scan_one_device(path: vol->name, BLK_OPEN_READ, mount_arg_dev: false);
2216	ret = PTR_ERR_OR_ZERO(ptr: device);
2217	mutex_unlock(lock: &uuid_mutex);
2218	break;
2219	case BTRFS_IOC_FORGET_DEV:
2220	if (vol->name[0] != 0) {
2221	ret = lookup_bdev(pathname: vol->name, dev: &devt);
2222	if (ret)
2223	break;
2224	}
2225	ret = btrfs_forget_devices(devt);
2226	break;
2227	case BTRFS_IOC_DEVICES_READY:
2228	mutex_lock(&uuid_mutex);
2229	/*
2230	* Scanning outside of mount can return NULL which would turn
2231	* into 0 error code.
2232	*/
2233	device = btrfs_scan_one_device(path: vol->name, BLK_OPEN_READ, mount_arg_dev: false);
2234	if (IS_ERR_OR_NULL(ptr: device)) {
2235	mutex_unlock(lock: &uuid_mutex);
2236	ret = PTR_ERR(ptr: device);
2237	break;
2238	}
2239	ret = !(device->fs_devices->num_devices ==
2240	device->fs_devices->total_devices);
2241	mutex_unlock(lock: &uuid_mutex);
2242	break;
2243	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2244	ret = btrfs_ioctl_get_supported_features(arg: (void __user*)arg);
2245	break;
2246	}
2247
2248	out:
2249	kfree(objp: vol);
2250	return ret;
2251	}
2252
2253	static int btrfs_freeze(struct super_block *sb)
2254	{
2255	struct btrfs_trans_handle *trans;
2256	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2257	struct btrfs_root *root = fs_info->tree_root;
2258
2259	set_bit(nr: BTRFS_FS_FROZEN, addr: &fs_info->flags);
2260	/*
2261	* We don't need a barrier here, we'll wait for any transaction that
2262	* could be in progress on other threads (and do delayed iputs that
2263	* we want to avoid on a frozen filesystem), or do the commit
2264	* ourselves.
2265	*/
2266	trans = btrfs_attach_transaction_barrier(root);
2267	if (IS_ERR(ptr: trans)) {
2268	/* no transaction, don't bother */
2269	if (PTR_ERR(ptr: trans) == -ENOENT)
2270	return 0;
2271	return PTR_ERR(ptr: trans);
2272	}
2273	return btrfs_commit_transaction(trans);
2274	}
2275
2276	static int check_dev_super(struct btrfs_device *dev)
2277	{
2278	struct btrfs_fs_info *fs_info = dev->fs_info;
2279	struct btrfs_super_block *sb;
2280	u64 last_trans;
2281	u16 csum_type;
2282	int ret = 0;
2283
2284	/* This should be called with fs still frozen. */
2285	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2286
2287	/* Missing dev, no need to check. */
2288	if (!dev->bdev)
2289	return 0;
2290
2291	/* Only need to check the primary super block. */
2292	sb = btrfs_read_dev_one_super(bdev: dev->bdev, copy_num: 0, drop_cache: true);
2293	if (IS_ERR(ptr: sb))
2294	return PTR_ERR(ptr: sb);
2295
2296	/* Verify the checksum. */
2297	csum_type = btrfs_super_csum_type(s: sb);
2298	if (csum_type != btrfs_super_csum_type(s: fs_info->super_copy)) {
2299	btrfs_err(fs_info, "csum type changed, has %u expect %u",
2300	csum_type, btrfs_super_csum_type(fs_info->super_copy));
2301	ret = -EUCLEAN;
2302	goto out;
2303	}
2304
2305	if (btrfs_check_super_csum(fs_info, disk_sb: sb)) {
2306	btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2307	ret = -EUCLEAN;
2308	goto out;
2309	}
2310
2311	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2312	ret = btrfs_validate_super(fs_info, sb, mirror_num: 0);
2313	if (ret < 0)
2314	goto out;
2315
2316	last_trans = btrfs_get_last_trans_committed(fs_info);
2317	if (btrfs_super_generation(s: sb) != last_trans) {
2318	btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2319	btrfs_super_generation(sb), last_trans);
2320	ret = -EUCLEAN;
2321	goto out;
2322	}
2323	out:
2324	btrfs_release_disk_super(super: sb);
2325	return ret;
2326	}
2327
2328	static int btrfs_unfreeze(struct super_block *sb)
2329	{
2330	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2331	struct btrfs_device *device;
2332	int ret = 0;
2333
2334	/*
2335	* Make sure the fs is not changed by accident (like hibernation then
2336	* modified by other OS).
2337	* If we found anything wrong, we mark the fs error immediately.
2338	*
2339	* And since the fs is frozen, no one can modify the fs yet, thus
2340	* we don't need to hold device_list_mutex.
2341	*/
2342	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2343	ret = check_dev_super(dev: device);
2344	if (ret < 0) {
2345	btrfs_handle_fs_error(fs_info, ret,
2346	"super block on devid %llu got modified unexpectedly",
2347	device->devid);
2348	break;
2349	}
2350	}
2351	clear_bit(nr: BTRFS_FS_FROZEN, addr: &fs_info->flags);
2352
2353	/*
2354	* We still return 0, to allow VFS layer to unfreeze the fs even the
2355	* above checks failed. Since the fs is either fine or read-only, we're
2356	* safe to continue, without causing further damage.
2357	*/
2358	return 0;
2359	}
2360
2361	static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2362	{
2363	struct btrfs_fs_info *fs_info = btrfs_sb(sb: root->d_sb);
2364
2365	/*
2366	* There should be always a valid pointer in latest_dev, it may be stale
2367	* for a short moment in case it's being deleted but still valid until
2368	* the end of RCU grace period.
2369	*/
2370	rcu_read_lock();
2371	seq_escape(m, s: btrfs_dev_name(device: fs_info->fs_devices->latest_dev), esc: " \t\n\\");
2372	rcu_read_unlock();
2373
2374	return 0;
2375	}
2376
2377	static const struct super_operations btrfs_super_ops = {
2378	.drop_inode = btrfs_drop_inode,
2379	.evict_inode = btrfs_evict_inode,
2380	.put_super = btrfs_put_super,
2381	.sync_fs = btrfs_sync_fs,
2382	.show_options = btrfs_show_options,
2383	.show_devname = btrfs_show_devname,
2384	.alloc_inode = btrfs_alloc_inode,
2385	.destroy_inode = btrfs_destroy_inode,
2386	.free_inode = btrfs_free_inode,
2387	.statfs = btrfs_statfs,
2388	.freeze_fs = btrfs_freeze,
2389	.unfreeze_fs = btrfs_unfreeze,
2390	};
2391
2392	static const struct file_operations btrfs_ctl_fops = {
2393	.open = btrfs_control_open,
2394	.unlocked_ioctl = btrfs_control_ioctl,
2395	.compat_ioctl = compat_ptr_ioctl,
2396	.owner = THIS_MODULE,
2397	.llseek = noop_llseek,
2398	};
2399
2400	static struct miscdevice btrfs_misc = {
2401	.minor = BTRFS_MINOR,
2402	.name = "btrfs-control",
2403	.fops = &btrfs_ctl_fops
2404	};
2405
2406	MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2407	MODULE_ALIAS("devname:btrfs-control");
2408
2409	static int __init btrfs_interface_init(void)
2410	{
2411	return misc_register(misc: &btrfs_misc);
2412	}
2413
2414	static __cold void btrfs_interface_exit(void)
2415	{
2416	misc_deregister(misc: &btrfs_misc);
2417	}
2418
2419	static int __init btrfs_print_mod_info(void)
2420	{
2421	static const char options[] = ""
2422	#ifdef CONFIG_BTRFS_DEBUG
2423	", debug=on"
2424	#endif
2425	#ifdef CONFIG_BTRFS_ASSERT
2426	", assert=on"
2427	#endif
2428	#ifdef CONFIG_BTRFS_FS_REF_VERIFY
2429	", ref-verify=on"
2430	#endif
2431	#ifdef CONFIG_BLK_DEV_ZONED
2432	", zoned=yes"
2433	#else
2434	", zoned=no"
2435	#endif
2436	#ifdef CONFIG_FS_VERITY
2437	", fsverity=yes"
2438	#else
2439	", fsverity=no"
2440	#endif
2441	;
2442	pr_info("Btrfs loaded%s\n", options);
2443	return 0;
2444	}
2445
2446	static int register_btrfs(void)
2447	{
2448	return register_filesystem(&btrfs_fs_type);
2449	}
2450
2451	static void unregister_btrfs(void)
2452	{
2453	unregister_filesystem(&btrfs_fs_type);
2454	}
2455
2456	/* Helper structure for long init/exit functions. */
2457	struct init_sequence {
2458	int (*init_func)(void);
2459	/* Can be NULL if the init_func doesn't need cleanup. */
2460	void (*exit_func)(void);
2461	};
2462
2463	static const struct init_sequence mod_init_seq[] = {
2464	{
2465	.init_func = btrfs_props_init,
2466	.exit_func = NULL,
2467	}, {
2468	.init_func = btrfs_init_sysfs,
2469	.exit_func = btrfs_exit_sysfs,
2470	}, {
2471	.init_func = btrfs_init_compress,
2472	.exit_func = btrfs_exit_compress,
2473	}, {
2474	.init_func = btrfs_init_cachep,
2475	.exit_func = btrfs_destroy_cachep,
2476	}, {
2477	.init_func = btrfs_transaction_init,
2478	.exit_func = btrfs_transaction_exit,
2479	}, {
2480	.init_func = btrfs_ctree_init,
2481	.exit_func = btrfs_ctree_exit,
2482	}, {
2483	.init_func = btrfs_free_space_init,
2484	.exit_func = btrfs_free_space_exit,
2485	}, {
2486	.init_func = extent_state_init_cachep,
2487	.exit_func = extent_state_free_cachep,
2488	}, {
2489	.init_func = extent_buffer_init_cachep,
2490	.exit_func = extent_buffer_free_cachep,
2491	}, {
2492	.init_func = btrfs_bioset_init,
2493	.exit_func = btrfs_bioset_exit,
2494	}, {
2495	.init_func = extent_map_init,
2496	.exit_func = extent_map_exit,
2497	}, {
2498	.init_func = ordered_data_init,
2499	.exit_func = ordered_data_exit,
2500	}, {
2501	.init_func = btrfs_delayed_inode_init,
2502	.exit_func = btrfs_delayed_inode_exit,
2503	}, {
2504	.init_func = btrfs_auto_defrag_init,
2505	.exit_func = btrfs_auto_defrag_exit,
2506	}, {
2507	.init_func = btrfs_delayed_ref_init,
2508	.exit_func = btrfs_delayed_ref_exit,
2509	}, {
2510	.init_func = btrfs_prelim_ref_init,
2511	.exit_func = btrfs_prelim_ref_exit,
2512	}, {
2513	.init_func = btrfs_interface_init,
2514	.exit_func = btrfs_interface_exit,
2515	}, {
2516	.init_func = btrfs_print_mod_info,
2517	.exit_func = NULL,
2518	}, {
2519	.init_func = btrfs_run_sanity_tests,
2520	.exit_func = NULL,
2521	}, {
2522	.init_func = register_btrfs,
2523	.exit_func = unregister_btrfs,
2524	}
2525	};
2526
2527	static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2528
2529	static __always_inline void btrfs_exit_btrfs_fs(void)
2530	{
2531	int i;
2532
2533	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2534	if (!mod_init_result[i])
2535	continue;
2536	if (mod_init_seq[i].exit_func)
2537	mod_init_seq[i].exit_func();
2538	mod_init_result[i] = false;
2539	}
2540	}
2541
2542	static void __exit exit_btrfs_fs(void)
2543	{
2544	btrfs_exit_btrfs_fs();
2545	btrfs_cleanup_fs_uuids();
2546	}
2547
2548	static int __init init_btrfs_fs(void)
2549	{
2550	int ret;
2551	int i;
2552
2553	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2554	ASSERT(!mod_init_result[i]);
2555	ret = mod_init_seq[i].init_func();
2556	if (ret < 0) {
2557	btrfs_exit_btrfs_fs();
2558	return ret;
2559	}
2560	mod_init_result[i] = true;
2561	}
2562	return 0;
2563	}
2564
2565	late_initcall(init_btrfs_fs);
2566	module_exit(exit_btrfs_fs)
2567
2568	MODULE_LICENSE("GPL");
2569	MODULE_SOFTDEP("pre: crc32c");
2570	MODULE_SOFTDEP("pre: xxhash64");
2571	MODULE_SOFTDEP("pre: sha256");
2572	MODULE_SOFTDEP("pre: blake2b-256");
2573