xref: /linux/fs/btrfs/super.c (revision f975f08c2e899ae2484407d7bba6bb7f8b6d9d40)
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 "direct-io.h"
38 #include "props.h"
39 #include "xattr.h"
40 #include "bio.h"
41 #include "export.h"
42 #include "compression.h"
43 #include "dev-replace.h"
44 #include "free-space-cache.h"
45 #include "backref.h"
46 #include "space-info.h"
47 #include "sysfs.h"
48 #include "zoned.h"
49 #include "tests/btrfs-tests.h"
50 #include "block-group.h"
51 #include "discard.h"
52 #include "qgroup.h"
53 #include "raid56.h"
54 #include "fs.h"
55 #include "accessors.h"
56 #include "defrag.h"
57 #include "dir-item.h"
58 #include "ioctl.h"
59 #include "scrub.h"
60 #include "verity.h"
61 #include "super.h"
62 #include "extent-tree.h"
63 #define CREATE_TRACE_POINTS
64 #include <trace/events/btrfs.h>
65 
66 static const struct super_operations btrfs_super_ops;
67 static struct file_system_type btrfs_fs_type;
68 
69 static void btrfs_put_super(struct super_block *sb)
70 {
71 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
72 
73 	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
74 	close_ctree(fs_info);
75 }
76 
77 /* Store the mount options related information. */
78 struct btrfs_fs_context {
79 	char *subvol_name;
80 	u64 subvol_objectid;
81 	u64 max_inline;
82 	u32 commit_interval;
83 	u32 metadata_ratio;
84 	u32 thread_pool_size;
85 	unsigned long long mount_opt;
86 	unsigned long compress_type:4;
87 	int compress_level;
88 	refcount_t refs;
89 };
90 
91 static void btrfs_emit_options(struct btrfs_fs_info *info,
92 			       struct btrfs_fs_context *old);
93 
94 enum {
95 	Opt_acl,
96 	Opt_clear_cache,
97 	Opt_commit_interval,
98 	Opt_compress,
99 	Opt_compress_force,
100 	Opt_compress_force_type,
101 	Opt_compress_type,
102 	Opt_degraded,
103 	Opt_device,
104 	Opt_fatal_errors,
105 	Opt_flushoncommit,
106 	Opt_max_inline,
107 	Opt_barrier,
108 	Opt_datacow,
109 	Opt_datasum,
110 	Opt_defrag,
111 	Opt_discard,
112 	Opt_discard_mode,
113 	Opt_ratio,
114 	Opt_rescan_uuid_tree,
115 	Opt_skip_balance,
116 	Opt_space_cache,
117 	Opt_space_cache_version,
118 	Opt_ssd,
119 	Opt_ssd_spread,
120 	Opt_subvol,
121 	Opt_subvol_empty,
122 	Opt_subvolid,
123 	Opt_thread_pool,
124 	Opt_treelog,
125 	Opt_user_subvol_rm_allowed,
126 	Opt_norecovery,
127 
128 	/* Rescue options */
129 	Opt_rescue,
130 	Opt_usebackuproot,
131 
132 	/* Debugging options */
133 	Opt_enospc_debug,
134 #ifdef CONFIG_BTRFS_DEBUG
135 	Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all,
136 #endif
137 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
138 	Opt_ref_verify,
139 #endif
140 	Opt_err,
141 };
142 
143 enum {
144 	Opt_fatal_errors_panic,
145 	Opt_fatal_errors_bug,
146 };
147 
148 static const struct constant_table btrfs_parameter_fatal_errors[] = {
149 	{ "panic", Opt_fatal_errors_panic },
150 	{ "bug", Opt_fatal_errors_bug },
151 	{}
152 };
153 
154 enum {
155 	Opt_discard_sync,
156 	Opt_discard_async,
157 };
158 
159 static const struct constant_table btrfs_parameter_discard[] = {
160 	{ "sync", Opt_discard_sync },
161 	{ "async", Opt_discard_async },
162 	{}
163 };
164 
165 enum {
166 	Opt_space_cache_v1,
167 	Opt_space_cache_v2,
168 };
169 
170 static const struct constant_table btrfs_parameter_space_cache[] = {
171 	{ "v1", Opt_space_cache_v1 },
172 	{ "v2", Opt_space_cache_v2 },
173 	{}
174 };
175 
176 enum {
177 	Opt_rescue_usebackuproot,
178 	Opt_rescue_nologreplay,
179 	Opt_rescue_ignorebadroots,
180 	Opt_rescue_ignoredatacsums,
181 	Opt_rescue_ignoremetacsums,
182 	Opt_rescue_ignoresuperflags,
183 	Opt_rescue_parameter_all,
184 };
185 
186 static const struct constant_table btrfs_parameter_rescue[] = {
187 	{ "usebackuproot", Opt_rescue_usebackuproot },
188 	{ "nologreplay", Opt_rescue_nologreplay },
189 	{ "ignorebadroots", Opt_rescue_ignorebadroots },
190 	{ "ibadroots", Opt_rescue_ignorebadroots },
191 	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
192 	{ "ignoremetacsums", Opt_rescue_ignoremetacsums},
193 	{ "ignoresuperflags", Opt_rescue_ignoresuperflags},
194 	{ "idatacsums", Opt_rescue_ignoredatacsums },
195 	{ "imetacsums", Opt_rescue_ignoremetacsums},
196 	{ "isuperflags", Opt_rescue_ignoresuperflags},
197 	{ "all", Opt_rescue_parameter_all },
198 	{}
199 };
200 
201 #ifdef CONFIG_BTRFS_DEBUG
202 enum {
203 	Opt_fragment_parameter_data,
204 	Opt_fragment_parameter_metadata,
205 	Opt_fragment_parameter_all,
206 };
207 
208 static const struct constant_table btrfs_parameter_fragment[] = {
209 	{ "data", Opt_fragment_parameter_data },
210 	{ "metadata", Opt_fragment_parameter_metadata },
211 	{ "all", Opt_fragment_parameter_all },
212 	{}
213 };
214 #endif
215 
216 static const struct fs_parameter_spec btrfs_fs_parameters[] = {
217 	fsparam_flag_no("acl", Opt_acl),
218 	fsparam_flag_no("autodefrag", Opt_defrag),
219 	fsparam_flag_no("barrier", Opt_barrier),
220 	fsparam_flag("clear_cache", Opt_clear_cache),
221 	fsparam_u32("commit", Opt_commit_interval),
222 	fsparam_flag("compress", Opt_compress),
223 	fsparam_string("compress", Opt_compress_type),
224 	fsparam_flag("compress-force", Opt_compress_force),
225 	fsparam_string("compress-force", Opt_compress_force_type),
226 	fsparam_flag_no("datacow", Opt_datacow),
227 	fsparam_flag_no("datasum", Opt_datasum),
228 	fsparam_flag("degraded", Opt_degraded),
229 	fsparam_string("device", Opt_device),
230 	fsparam_flag_no("discard", Opt_discard),
231 	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
232 	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
233 	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
234 	fsparam_string("max_inline", Opt_max_inline),
235 	fsparam_u32("metadata_ratio", Opt_ratio),
236 	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
237 	fsparam_flag("skip_balance", Opt_skip_balance),
238 	fsparam_flag_no("space_cache", Opt_space_cache),
239 	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
240 	fsparam_flag_no("ssd", Opt_ssd),
241 	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
242 	fsparam_string("subvol", Opt_subvol),
243 	fsparam_flag("subvol=", Opt_subvol_empty),
244 	fsparam_u64("subvolid", Opt_subvolid),
245 	fsparam_u32("thread_pool", Opt_thread_pool),
246 	fsparam_flag_no("treelog", Opt_treelog),
247 	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
248 
249 	/* Rescue options. */
250 	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
251 	/* Deprecated, with alias rescue=usebackuproot */
252 	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
253 	/* For compatibility only, alias for "rescue=nologreplay". */
254 	fsparam_flag("norecovery", Opt_norecovery),
255 
256 	/* Debugging options. */
257 	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
258 #ifdef CONFIG_BTRFS_DEBUG
259 	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
260 #endif
261 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
262 	fsparam_flag("ref_verify", Opt_ref_verify),
263 #endif
264 	{}
265 };
266 
267 static bool btrfs_match_compress_type(const char *string, const char *type, bool may_have_level)
268 {
269 	const int len = strlen(type);
270 
271 	return (strncmp(string, type, len) == 0) &&
272 		((may_have_level && string[len] == ':') || string[len] == '\0');
273 }
274 
275 static int btrfs_parse_compress(struct btrfs_fs_context *ctx,
276 				const struct fs_parameter *param, int opt)
277 {
278 	const char *string = param->string;
279 	int ret;
280 
281 	/*
282 	 * Provide the same semantics as older kernels that don't use fs
283 	 * context, specifying the "compress" option clears "force-compress"
284 	 * without the need to pass "compress-force=[no|none]" before
285 	 * specifying "compress".
286 	 */
287 	if (opt != Opt_compress_force && opt != Opt_compress_force_type)
288 		btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
289 
290 	if (opt == Opt_compress || opt == Opt_compress_force) {
291 		ctx->compress_type = BTRFS_COMPRESS_ZLIB;
292 		ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
293 		btrfs_set_opt(ctx->mount_opt, COMPRESS);
294 		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
295 		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
296 	} else if (btrfs_match_compress_type(string, "zlib", true)) {
297 		ctx->compress_type = BTRFS_COMPRESS_ZLIB;
298 		ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, string + 4,
299 					       &ctx->compress_level);
300 		if (ret < 0)
301 			goto error;
302 		btrfs_set_opt(ctx->mount_opt, COMPRESS);
303 		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
304 		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
305 	} else if (btrfs_match_compress_type(string, "lzo", true)) {
306 		ctx->compress_type = BTRFS_COMPRESS_LZO;
307 		ret = btrfs_compress_str2level(BTRFS_COMPRESS_LZO, string + 3,
308 					       &ctx->compress_level);
309 		if (ret < 0)
310 			goto error;
311 		if (string[3] == ':' && string[4])
312 			btrfs_warn(NULL, "Compression level ignored for LZO");
313 		btrfs_set_opt(ctx->mount_opt, COMPRESS);
314 		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
315 		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
316 	} else if (btrfs_match_compress_type(string, "zstd", true)) {
317 		ctx->compress_type = BTRFS_COMPRESS_ZSTD;
318 		ret = btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, string + 4,
319 					       &ctx->compress_level);
320 		if (ret < 0)
321 			goto error;
322 		btrfs_set_opt(ctx->mount_opt, COMPRESS);
323 		btrfs_clear_opt(ctx->mount_opt, NODATACOW);
324 		btrfs_clear_opt(ctx->mount_opt, NODATASUM);
325 	} else if (btrfs_match_compress_type(string, "no", false) ||
326 		   btrfs_match_compress_type(string, "none", false)) {
327 		ctx->compress_level = 0;
328 		ctx->compress_type = 0;
329 		btrfs_clear_opt(ctx->mount_opt, COMPRESS);
330 		btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
331 	} else {
332 		ret = -EINVAL;
333 		goto error;
334 	}
335 	return 0;
336 error:
337 	btrfs_err(NULL, "failed to parse compression option '%s'", string);
338 	return ret;
339 
340 }
341 
342 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
343 {
344 	struct btrfs_fs_context *ctx = fc->fs_private;
345 	struct fs_parse_result result;
346 	int opt;
347 
348 	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
349 	if (opt < 0)
350 		return opt;
351 
352 	switch (opt) {
353 	case Opt_degraded:
354 		btrfs_set_opt(ctx->mount_opt, DEGRADED);
355 		break;
356 	case Opt_subvol_empty:
357 		/*
358 		 * This exists because we used to allow it on accident, so we're
359 		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
360 		 * empty subvol= again").
361 		 */
362 		break;
363 	case Opt_subvol:
364 		kfree(ctx->subvol_name);
365 		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
366 		if (!ctx->subvol_name)
367 			return -ENOMEM;
368 		break;
369 	case Opt_subvolid:
370 		ctx->subvol_objectid = result.uint_64;
371 
372 		/* subvolid=0 means give me the original fs_tree. */
373 		if (!ctx->subvol_objectid)
374 			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
375 		break;
376 	case Opt_device: {
377 		struct btrfs_device *device;
378 
379 		mutex_lock(&uuid_mutex);
380 		device = btrfs_scan_one_device(param->string, false);
381 		mutex_unlock(&uuid_mutex);
382 		if (IS_ERR(device))
383 			return PTR_ERR(device);
384 		break;
385 	}
386 	case Opt_datasum:
387 		if (result.negated) {
388 			btrfs_set_opt(ctx->mount_opt, NODATASUM);
389 		} else {
390 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
391 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
392 		}
393 		break;
394 	case Opt_datacow:
395 		if (result.negated) {
396 			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
397 			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
398 			btrfs_set_opt(ctx->mount_opt, NODATACOW);
399 			btrfs_set_opt(ctx->mount_opt, NODATASUM);
400 		} else {
401 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
402 		}
403 		break;
404 	case Opt_compress_force:
405 	case Opt_compress_force_type:
406 		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
407 		fallthrough;
408 	case Opt_compress:
409 	case Opt_compress_type:
410 		if (btrfs_parse_compress(ctx, param, opt))
411 			return -EINVAL;
412 		break;
413 	case Opt_ssd:
414 		if (result.negated) {
415 			btrfs_set_opt(ctx->mount_opt, NOSSD);
416 			btrfs_clear_opt(ctx->mount_opt, SSD);
417 			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
418 		} else {
419 			btrfs_set_opt(ctx->mount_opt, SSD);
420 			btrfs_clear_opt(ctx->mount_opt, NOSSD);
421 		}
422 		break;
423 	case Opt_ssd_spread:
424 		if (result.negated) {
425 			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
426 		} else {
427 			btrfs_set_opt(ctx->mount_opt, SSD);
428 			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
429 			btrfs_clear_opt(ctx->mount_opt, NOSSD);
430 		}
431 		break;
432 	case Opt_barrier:
433 		if (result.negated)
434 			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
435 		else
436 			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
437 		break;
438 	case Opt_thread_pool:
439 		if (result.uint_32 == 0) {
440 			btrfs_err(NULL, "invalid value 0 for thread_pool");
441 			return -EINVAL;
442 		}
443 		ctx->thread_pool_size = result.uint_32;
444 		break;
445 	case Opt_max_inline:
446 		ctx->max_inline = memparse(param->string, NULL);
447 		break;
448 	case Opt_acl:
449 		if (result.negated) {
450 			fc->sb_flags &= ~SB_POSIXACL;
451 		} else {
452 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
453 			fc->sb_flags |= SB_POSIXACL;
454 #else
455 			btrfs_err(NULL, "support for ACL not compiled in");
456 			return -EINVAL;
457 #endif
458 		}
459 		/*
460 		 * VFS limits the ability to toggle ACL on and off via remount,
461 		 * despite every file system allowing this.  This seems to be
462 		 * an oversight since we all do, but it'll fail if we're
463 		 * remounting.  So don't set the mask here, we'll check it in
464 		 * btrfs_reconfigure and do the toggling ourselves.
465 		 */
466 		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
467 			fc->sb_flags_mask |= SB_POSIXACL;
468 		break;
469 	case Opt_treelog:
470 		if (result.negated)
471 			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
472 		else
473 			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
474 		break;
475 	case Opt_norecovery:
476 		btrfs_info(NULL,
477 "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
478 		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
479 		break;
480 	case Opt_flushoncommit:
481 		if (result.negated)
482 			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
483 		else
484 			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
485 		break;
486 	case Opt_ratio:
487 		ctx->metadata_ratio = result.uint_32;
488 		break;
489 	case Opt_discard:
490 		if (result.negated) {
491 			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
492 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
493 			btrfs_set_opt(ctx->mount_opt, NODISCARD);
494 		} else {
495 			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
496 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
497 		}
498 		break;
499 	case Opt_discard_mode:
500 		switch (result.uint_32) {
501 		case Opt_discard_sync:
502 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
503 			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
504 			break;
505 		case Opt_discard_async:
506 			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
507 			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
508 			break;
509 		default:
510 			btrfs_err(NULL, "unrecognized discard mode value %s",
511 				  param->key);
512 			return -EINVAL;
513 		}
514 		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
515 		break;
516 	case Opt_space_cache:
517 		if (result.negated) {
518 			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
519 			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
520 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
521 		} else {
522 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
523 			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
524 		}
525 		break;
526 	case Opt_space_cache_version:
527 		switch (result.uint_32) {
528 		case Opt_space_cache_v1:
529 			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
530 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
531 			break;
532 		case Opt_space_cache_v2:
533 			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
534 			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
535 			break;
536 		default:
537 			btrfs_err(NULL, "unrecognized space_cache value %s",
538 				  param->key);
539 			return -EINVAL;
540 		}
541 		break;
542 	case Opt_rescan_uuid_tree:
543 		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
544 		break;
545 	case Opt_clear_cache:
546 		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
547 		break;
548 	case Opt_user_subvol_rm_allowed:
549 		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
550 		break;
551 	case Opt_enospc_debug:
552 		if (result.negated)
553 			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
554 		else
555 			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
556 		break;
557 	case Opt_defrag:
558 		if (result.negated)
559 			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
560 		else
561 			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
562 		break;
563 	case Opt_usebackuproot:
564 		btrfs_warn(NULL,
565 			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
566 		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
567 
568 		/* If we're loading the backup roots we can't trust the space cache. */
569 		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
570 		break;
571 	case Opt_skip_balance:
572 		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
573 		break;
574 	case Opt_fatal_errors:
575 		switch (result.uint_32) {
576 		case Opt_fatal_errors_panic:
577 			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
578 			break;
579 		case Opt_fatal_errors_bug:
580 			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
581 			break;
582 		default:
583 			btrfs_err(NULL, "unrecognized fatal_errors value %s",
584 				  param->key);
585 			return -EINVAL;
586 		}
587 		break;
588 	case Opt_commit_interval:
589 		ctx->commit_interval = result.uint_32;
590 		if (ctx->commit_interval > BTRFS_WARNING_COMMIT_INTERVAL) {
591 			btrfs_warn(NULL, "excessive commit interval %u, use with care",
592 				   ctx->commit_interval);
593 		}
594 		if (ctx->commit_interval == 0)
595 			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
596 		break;
597 	case Opt_rescue:
598 		switch (result.uint_32) {
599 		case Opt_rescue_usebackuproot:
600 			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
601 			break;
602 		case Opt_rescue_nologreplay:
603 			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
604 			break;
605 		case Opt_rescue_ignorebadroots:
606 			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
607 			break;
608 		case Opt_rescue_ignoredatacsums:
609 			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
610 			break;
611 		case Opt_rescue_ignoremetacsums:
612 			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
613 			break;
614 		case Opt_rescue_ignoresuperflags:
615 			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
616 			break;
617 		case Opt_rescue_parameter_all:
618 			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
619 			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
620 			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
621 			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
622 			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
623 			break;
624 		default:
625 			btrfs_info(NULL, "unrecognized rescue option '%s'",
626 				   param->key);
627 			return -EINVAL;
628 		}
629 		break;
630 #ifdef CONFIG_BTRFS_DEBUG
631 	case Opt_fragment:
632 		switch (result.uint_32) {
633 		case Opt_fragment_parameter_all:
634 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
635 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
636 			break;
637 		case Opt_fragment_parameter_metadata:
638 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
639 			break;
640 		case Opt_fragment_parameter_data:
641 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
642 			break;
643 		default:
644 			btrfs_info(NULL, "unrecognized fragment option '%s'",
645 				   param->key);
646 			return -EINVAL;
647 		}
648 		break;
649 #endif
650 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
651 	case Opt_ref_verify:
652 		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
653 		break;
654 #endif
655 	default:
656 		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
657 		return -EINVAL;
658 	}
659 
660 	return 0;
661 }
662 
663 /*
664  * Some options only have meaning at mount time and shouldn't persist across
665  * remounts, or be displayed. Clear these at the end of mount and remount code
666  * paths.
667  */
668 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
669 {
670 	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
671 	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
672 	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
673 }
674 
675 static bool check_ro_option(const struct btrfs_fs_info *fs_info,
676 			    unsigned long long mount_opt, unsigned long long opt,
677 			    const char *opt_name)
678 {
679 	if (mount_opt & opt) {
680 		btrfs_err(fs_info, "%s must be used with ro mount option",
681 			  opt_name);
682 		return true;
683 	}
684 	return false;
685 }
686 
687 bool btrfs_check_options(const struct btrfs_fs_info *info,
688 			 unsigned long long *mount_opt,
689 			 unsigned long flags)
690 {
691 	bool ret = true;
692 
693 	if (!(flags & SB_RDONLY) &&
694 	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
695 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
696 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
697 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
698 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
699 		ret = false;
700 
701 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
702 	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
703 	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
704 		btrfs_err(info, "cannot disable free-space-tree");
705 		ret = false;
706 	}
707 	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
708 	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
709 		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
710 		ret = false;
711 	}
712 
713 	if (btrfs_check_mountopts_zoned(info, mount_opt))
714 		ret = false;
715 
716 	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
717 		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
718 			btrfs_warn(info,
719 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
720 		}
721 	}
722 
723 	return ret;
724 }
725 
726 /*
727  * This is subtle, we only call this during open_ctree().  We need to pre-load
728  * the mount options with the on-disk settings.  Before the new mount API took
729  * effect we would do this on mount and remount.  With the new mount API we'll
730  * only do this on the initial mount.
731  *
732  * This isn't a change in behavior, because we're using the current state of the
733  * file system to set the current mount options.  If you mounted with special
734  * options to disable these features and then remounted we wouldn't revert the
735  * settings, because mounting without these features cleared the on-disk
736  * settings, so this being called on re-mount is not needed.
737  */
738 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
739 {
740 	if (fs_info->sectorsize < PAGE_SIZE) {
741 		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
742 		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
743 			btrfs_info(fs_info,
744 				   "forcing free space tree for sector size %u with page size %lu",
745 				   fs_info->sectorsize, PAGE_SIZE);
746 			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
747 		}
748 	}
749 
750 	/*
751 	 * At this point our mount options are populated, so we only mess with
752 	 * these settings if we don't have any settings already.
753 	 */
754 	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
755 		return;
756 
757 	if (btrfs_is_zoned(fs_info) &&
758 	    btrfs_free_space_cache_v1_active(fs_info)) {
759 		btrfs_info(fs_info, "zoned: clearing existing space cache");
760 		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
761 		return;
762 	}
763 
764 	if (btrfs_test_opt(fs_info, SPACE_CACHE))
765 		return;
766 
767 	if (btrfs_test_opt(fs_info, NOSPACECACHE))
768 		return;
769 
770 	/*
771 	 * At this point we don't have explicit options set by the user, set
772 	 * them ourselves based on the state of the file system.
773 	 */
774 	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
775 		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
776 	else if (btrfs_free_space_cache_v1_active(fs_info))
777 		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
778 }
779 
780 static void set_device_specific_options(struct btrfs_fs_info *fs_info)
781 {
782 	if (!btrfs_test_opt(fs_info, NOSSD) &&
783 	    !fs_info->fs_devices->rotating)
784 		btrfs_set_opt(fs_info->mount_opt, SSD);
785 
786 	/*
787 	 * For devices supporting discard turn on discard=async automatically,
788 	 * unless it's already set or disabled. This could be turned off by
789 	 * nodiscard for the same mount.
790 	 *
791 	 * The zoned mode piggy backs on the discard functionality for
792 	 * resetting a zone. There is no reason to delay the zone reset as it is
793 	 * fast enough. So, do not enable async discard for zoned mode.
794 	 */
795 	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
796 	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
797 	      btrfs_test_opt(fs_info, NODISCARD)) &&
798 	    fs_info->fs_devices->discardable &&
799 	    !btrfs_is_zoned(fs_info))
800 		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
801 }
802 
803 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
804 					  u64 subvol_objectid)
805 {
806 	struct btrfs_root *root = fs_info->tree_root;
807 	struct btrfs_root *fs_root = NULL;
808 	struct btrfs_root_ref *root_ref;
809 	struct btrfs_inode_ref *inode_ref;
810 	struct btrfs_key key;
811 	struct btrfs_path *path = NULL;
812 	char *name = NULL, *ptr;
813 	u64 dirid;
814 	int len;
815 	int ret;
816 
817 	path = btrfs_alloc_path();
818 	if (!path) {
819 		ret = -ENOMEM;
820 		goto err;
821 	}
822 
823 	name = kmalloc(PATH_MAX, GFP_KERNEL);
824 	if (!name) {
825 		ret = -ENOMEM;
826 		goto err;
827 	}
828 	ptr = name + PATH_MAX - 1;
829 	ptr[0] = '\0';
830 
831 	/*
832 	 * Walk up the subvolume trees in the tree of tree roots by root
833 	 * backrefs until we hit the top-level subvolume.
834 	 */
835 	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
836 		key.objectid = subvol_objectid;
837 		key.type = BTRFS_ROOT_BACKREF_KEY;
838 		key.offset = (u64)-1;
839 
840 		ret = btrfs_search_backwards(root, &key, path);
841 		if (ret < 0) {
842 			goto err;
843 		} else if (ret > 0) {
844 			ret = -ENOENT;
845 			goto err;
846 		}
847 
848 		subvol_objectid = key.offset;
849 
850 		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
851 					  struct btrfs_root_ref);
852 		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
853 		ptr -= len + 1;
854 		if (ptr < name) {
855 			ret = -ENAMETOOLONG;
856 			goto err;
857 		}
858 		read_extent_buffer(path->nodes[0], ptr + 1,
859 				   (unsigned long)(root_ref + 1), len);
860 		ptr[0] = '/';
861 		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
862 		btrfs_release_path(path);
863 
864 		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
865 		if (IS_ERR(fs_root)) {
866 			ret = PTR_ERR(fs_root);
867 			fs_root = NULL;
868 			goto err;
869 		}
870 
871 		/*
872 		 * Walk up the filesystem tree by inode refs until we hit the
873 		 * root directory.
874 		 */
875 		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
876 			key.objectid = dirid;
877 			key.type = BTRFS_INODE_REF_KEY;
878 			key.offset = (u64)-1;
879 
880 			ret = btrfs_search_backwards(fs_root, &key, path);
881 			if (ret < 0) {
882 				goto err;
883 			} else if (ret > 0) {
884 				ret = -ENOENT;
885 				goto err;
886 			}
887 
888 			dirid = key.offset;
889 
890 			inode_ref = btrfs_item_ptr(path->nodes[0],
891 						   path->slots[0],
892 						   struct btrfs_inode_ref);
893 			len = btrfs_inode_ref_name_len(path->nodes[0],
894 						       inode_ref);
895 			ptr -= len + 1;
896 			if (ptr < name) {
897 				ret = -ENAMETOOLONG;
898 				goto err;
899 			}
900 			read_extent_buffer(path->nodes[0], ptr + 1,
901 					   (unsigned long)(inode_ref + 1), len);
902 			ptr[0] = '/';
903 			btrfs_release_path(path);
904 		}
905 		btrfs_put_root(fs_root);
906 		fs_root = NULL;
907 	}
908 
909 	btrfs_free_path(path);
910 	if (ptr == name + PATH_MAX - 1) {
911 		name[0] = '/';
912 		name[1] = '\0';
913 	} else {
914 		memmove(name, ptr, name + PATH_MAX - ptr);
915 	}
916 	return name;
917 
918 err:
919 	btrfs_put_root(fs_root);
920 	btrfs_free_path(path);
921 	kfree(name);
922 	return ERR_PTR(ret);
923 }
924 
925 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
926 {
927 	struct btrfs_root *root = fs_info->tree_root;
928 	struct btrfs_dir_item *di;
929 	struct btrfs_path *path;
930 	struct btrfs_key location;
931 	struct fscrypt_str name = FSTR_INIT("default", 7);
932 	u64 dir_id;
933 
934 	path = btrfs_alloc_path();
935 	if (!path)
936 		return -ENOMEM;
937 
938 	/*
939 	 * Find the "default" dir item which points to the root item that we
940 	 * will mount by default if we haven't been given a specific subvolume
941 	 * to mount.
942 	 */
943 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
944 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
945 	if (IS_ERR(di)) {
946 		btrfs_free_path(path);
947 		return PTR_ERR(di);
948 	}
949 	if (!di) {
950 		/*
951 		 * Ok the default dir item isn't there.  This is weird since
952 		 * it's always been there, but don't freak out, just try and
953 		 * mount the top-level subvolume.
954 		 */
955 		btrfs_free_path(path);
956 		*objectid = BTRFS_FS_TREE_OBJECTID;
957 		return 0;
958 	}
959 
960 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
961 	btrfs_free_path(path);
962 	*objectid = location.objectid;
963 	return 0;
964 }
965 
966 static int btrfs_fill_super(struct super_block *sb,
967 			    struct btrfs_fs_devices *fs_devices)
968 {
969 	struct btrfs_inode *inode;
970 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
971 	int ret;
972 
973 	sb->s_maxbytes = MAX_LFS_FILESIZE;
974 	sb->s_magic = BTRFS_SUPER_MAGIC;
975 	sb->s_op = &btrfs_super_ops;
976 	set_default_d_op(sb, &btrfs_dentry_operations);
977 	sb->s_export_op = &btrfs_export_ops;
978 #ifdef CONFIG_FS_VERITY
979 	sb->s_vop = &btrfs_verityops;
980 #endif
981 	sb->s_xattr = btrfs_xattr_handlers;
982 	sb->s_time_gran = 1;
983 	sb->s_iflags |= SB_I_CGROUPWB | SB_I_ALLOW_HSM;
984 
985 	ret = super_setup_bdi(sb);
986 	if (ret) {
987 		btrfs_err(fs_info, "super_setup_bdi failed");
988 		return ret;
989 	}
990 
991 	ret = open_ctree(sb, fs_devices);
992 	if (ret) {
993 		btrfs_err(fs_info, "open_ctree failed: %d", ret);
994 		return ret;
995 	}
996 
997 	btrfs_emit_options(fs_info, NULL);
998 
999 	inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
1000 	if (IS_ERR(inode)) {
1001 		ret = PTR_ERR(inode);
1002 		btrfs_handle_fs_error(fs_info, ret, NULL);
1003 		goto fail_close;
1004 	}
1005 
1006 	sb->s_root = d_make_root(&inode->vfs_inode);
1007 	if (!sb->s_root) {
1008 		ret = -ENOMEM;
1009 		goto fail_close;
1010 	}
1011 
1012 	sb->s_flags |= SB_ACTIVE;
1013 	return 0;
1014 
1015 fail_close:
1016 	close_ctree(fs_info);
1017 	return ret;
1018 }
1019 
1020 int btrfs_sync_fs(struct super_block *sb, int wait)
1021 {
1022 	struct btrfs_trans_handle *trans;
1023 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1024 	struct btrfs_root *root = fs_info->tree_root;
1025 
1026 	trace_btrfs_sync_fs(fs_info, wait);
1027 
1028 	if (!wait) {
1029 		filemap_flush(fs_info->btree_inode->i_mapping);
1030 		return 0;
1031 	}
1032 
1033 	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1034 
1035 	trans = btrfs_attach_transaction_barrier(root);
1036 	if (IS_ERR(trans)) {
1037 		/* no transaction, don't bother */
1038 		if (PTR_ERR(trans) == -ENOENT) {
1039 			/*
1040 			 * Exit unless we have some pending changes
1041 			 * that need to go through commit
1042 			 */
1043 			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1044 				      &fs_info->flags))
1045 				return 0;
1046 			/*
1047 			 * A non-blocking test if the fs is frozen. We must not
1048 			 * start a new transaction here otherwise a deadlock
1049 			 * happens. The pending operations are delayed to the
1050 			 * next commit after thawing.
1051 			 */
1052 			if (sb_start_write_trylock(sb))
1053 				sb_end_write(sb);
1054 			else
1055 				return 0;
1056 			trans = btrfs_start_transaction(root, 0);
1057 		}
1058 		if (IS_ERR(trans))
1059 			return PTR_ERR(trans);
1060 	}
1061 	return btrfs_commit_transaction(trans);
1062 }
1063 
1064 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1065 {
1066 	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1067 	*printed = true;
1068 }
1069 
1070 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1071 {
1072 	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1073 	const char *compress_type;
1074 	const char *subvol_name;
1075 	bool printed = false;
1076 
1077 	if (btrfs_test_opt(info, DEGRADED))
1078 		seq_puts(seq, ",degraded");
1079 	if (btrfs_test_opt(info, NODATASUM))
1080 		seq_puts(seq, ",nodatasum");
1081 	if (btrfs_test_opt(info, NODATACOW))
1082 		seq_puts(seq, ",nodatacow");
1083 	if (btrfs_test_opt(info, NOBARRIER))
1084 		seq_puts(seq, ",nobarrier");
1085 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1086 		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1087 	if (info->thread_pool_size !=  min_t(unsigned long,
1088 					     num_online_cpus() + 2, 8))
1089 		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1090 	if (btrfs_test_opt(info, COMPRESS)) {
1091 		compress_type = btrfs_compress_type2str(info->compress_type);
1092 		if (btrfs_test_opt(info, FORCE_COMPRESS))
1093 			seq_printf(seq, ",compress-force=%s", compress_type);
1094 		else
1095 			seq_printf(seq, ",compress=%s", compress_type);
1096 		if (info->compress_level && info->compress_type != BTRFS_COMPRESS_LZO)
1097 			seq_printf(seq, ":%d", info->compress_level);
1098 	}
1099 	if (btrfs_test_opt(info, NOSSD))
1100 		seq_puts(seq, ",nossd");
1101 	if (btrfs_test_opt(info, SSD_SPREAD))
1102 		seq_puts(seq, ",ssd_spread");
1103 	else if (btrfs_test_opt(info, SSD))
1104 		seq_puts(seq, ",ssd");
1105 	if (btrfs_test_opt(info, NOTREELOG))
1106 		seq_puts(seq, ",notreelog");
1107 	if (btrfs_test_opt(info, NOLOGREPLAY))
1108 		print_rescue_option(seq, "nologreplay", &printed);
1109 	if (btrfs_test_opt(info, USEBACKUPROOT))
1110 		print_rescue_option(seq, "usebackuproot", &printed);
1111 	if (btrfs_test_opt(info, IGNOREBADROOTS))
1112 		print_rescue_option(seq, "ignorebadroots", &printed);
1113 	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1114 		print_rescue_option(seq, "ignoredatacsums", &printed);
1115 	if (btrfs_test_opt(info, IGNOREMETACSUMS))
1116 		print_rescue_option(seq, "ignoremetacsums", &printed);
1117 	if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1118 		print_rescue_option(seq, "ignoresuperflags", &printed);
1119 	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1120 		seq_puts(seq, ",flushoncommit");
1121 	if (btrfs_test_opt(info, DISCARD_SYNC))
1122 		seq_puts(seq, ",discard");
1123 	if (btrfs_test_opt(info, DISCARD_ASYNC))
1124 		seq_puts(seq, ",discard=async");
1125 	if (!(info->sb->s_flags & SB_POSIXACL))
1126 		seq_puts(seq, ",noacl");
1127 	if (btrfs_free_space_cache_v1_active(info))
1128 		seq_puts(seq, ",space_cache");
1129 	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1130 		seq_puts(seq, ",space_cache=v2");
1131 	else
1132 		seq_puts(seq, ",nospace_cache");
1133 	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1134 		seq_puts(seq, ",rescan_uuid_tree");
1135 	if (btrfs_test_opt(info, CLEAR_CACHE))
1136 		seq_puts(seq, ",clear_cache");
1137 	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1138 		seq_puts(seq, ",user_subvol_rm_allowed");
1139 	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1140 		seq_puts(seq, ",enospc_debug");
1141 	if (btrfs_test_opt(info, AUTO_DEFRAG))
1142 		seq_puts(seq, ",autodefrag");
1143 	if (btrfs_test_opt(info, SKIP_BALANCE))
1144 		seq_puts(seq, ",skip_balance");
1145 	if (info->metadata_ratio)
1146 		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1147 	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1148 		seq_puts(seq, ",fatal_errors=panic");
1149 	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1150 		seq_printf(seq, ",commit=%u", info->commit_interval);
1151 #ifdef CONFIG_BTRFS_DEBUG
1152 	if (btrfs_test_opt(info, FRAGMENT_DATA))
1153 		seq_puts(seq, ",fragment=data");
1154 	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1155 		seq_puts(seq, ",fragment=metadata");
1156 #endif
1157 	if (btrfs_test_opt(info, REF_VERIFY))
1158 		seq_puts(seq, ",ref_verify");
1159 	seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1160 	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1161 			btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1162 	if (!IS_ERR(subvol_name)) {
1163 		seq_show_option(seq, "subvol", subvol_name);
1164 		kfree(subvol_name);
1165 	}
1166 	return 0;
1167 }
1168 
1169 /*
1170  * subvolumes are identified by ino 256
1171  */
1172 static inline bool is_subvolume_inode(struct inode *inode)
1173 {
1174 	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1175 		return true;
1176 	return false;
1177 }
1178 
1179 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1180 				   struct vfsmount *mnt)
1181 {
1182 	struct dentry *root;
1183 	int ret;
1184 
1185 	if (!subvol_name) {
1186 		if (!subvol_objectid) {
1187 			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1188 							  &subvol_objectid);
1189 			if (ret) {
1190 				root = ERR_PTR(ret);
1191 				goto out;
1192 			}
1193 		}
1194 		subvol_name = btrfs_get_subvol_name_from_objectid(
1195 					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1196 		if (IS_ERR(subvol_name)) {
1197 			root = ERR_CAST(subvol_name);
1198 			subvol_name = NULL;
1199 			goto out;
1200 		}
1201 
1202 	}
1203 
1204 	root = mount_subtree(mnt, subvol_name);
1205 	/* mount_subtree() drops our reference on the vfsmount. */
1206 	mnt = NULL;
1207 
1208 	if (!IS_ERR(root)) {
1209 		struct super_block *s = root->d_sb;
1210 		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1211 		struct inode *root_inode = d_inode(root);
1212 		u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1213 
1214 		ret = 0;
1215 		if (!is_subvolume_inode(root_inode)) {
1216 			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1217 			       subvol_name);
1218 			ret = -EINVAL;
1219 		}
1220 		if (subvol_objectid && root_objectid != subvol_objectid) {
1221 			/*
1222 			 * This will also catch a race condition where a
1223 			 * subvolume which was passed by ID is renamed and
1224 			 * another subvolume is renamed over the old location.
1225 			 */
1226 			btrfs_err(fs_info,
1227 				  "subvol '%s' does not match subvolid %llu",
1228 				  subvol_name, subvol_objectid);
1229 			ret = -EINVAL;
1230 		}
1231 		if (ret) {
1232 			dput(root);
1233 			root = ERR_PTR(ret);
1234 			deactivate_locked_super(s);
1235 		}
1236 	}
1237 
1238 out:
1239 	mntput(mnt);
1240 	kfree(subvol_name);
1241 	return root;
1242 }
1243 
1244 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1245 				     u32 new_pool_size, u32 old_pool_size)
1246 {
1247 	if (new_pool_size == old_pool_size)
1248 		return;
1249 
1250 	fs_info->thread_pool_size = new_pool_size;
1251 
1252 	btrfs_info(fs_info, "resize thread pool %d -> %d",
1253 	       old_pool_size, new_pool_size);
1254 
1255 	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1256 	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1257 	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1258 	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1259 	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1260 	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1261 	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1262 	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1263 }
1264 
1265 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1266 				       unsigned long long old_opts, int flags)
1267 {
1268 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1269 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1270 	     (flags & SB_RDONLY))) {
1271 		/* wait for any defraggers to finish */
1272 		wait_event(fs_info->transaction_wait,
1273 			   (atomic_read(&fs_info->defrag_running) == 0));
1274 		if (flags & SB_RDONLY)
1275 			sync_filesystem(fs_info->sb);
1276 	}
1277 }
1278 
1279 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1280 					 unsigned long long old_opts)
1281 {
1282 	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1283 
1284 	/*
1285 	 * We need to cleanup all defragable inodes if the autodefragment is
1286 	 * close or the filesystem is read only.
1287 	 */
1288 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1289 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1290 		btrfs_cleanup_defrag_inodes(fs_info);
1291 	}
1292 
1293 	/* If we toggled discard async */
1294 	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1295 	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1296 		btrfs_discard_resume(fs_info);
1297 	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1298 		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1299 		btrfs_discard_cleanup(fs_info);
1300 
1301 	/* If we toggled space cache */
1302 	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1303 		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1304 }
1305 
1306 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1307 {
1308 	int ret;
1309 
1310 	if (BTRFS_FS_ERROR(fs_info)) {
1311 		btrfs_err(fs_info,
1312 			  "remounting read-write after error is not allowed");
1313 		return -EINVAL;
1314 	}
1315 
1316 	if (fs_info->fs_devices->rw_devices == 0)
1317 		return -EACCES;
1318 
1319 	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1320 		btrfs_warn(fs_info,
1321 			   "too many missing devices, writable remount is not allowed");
1322 		return -EACCES;
1323 	}
1324 
1325 	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1326 		btrfs_warn(fs_info,
1327 			   "mount required to replay tree-log, cannot remount read-write");
1328 		return -EINVAL;
1329 	}
1330 
1331 	/*
1332 	 * NOTE: when remounting with a change that does writes, don't put it
1333 	 * anywhere above this point, as we are not sure to be safe to write
1334 	 * until we pass the above checks.
1335 	 */
1336 	ret = btrfs_start_pre_rw_mount(fs_info);
1337 	if (ret)
1338 		return ret;
1339 
1340 	btrfs_clear_sb_rdonly(fs_info->sb);
1341 
1342 	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1343 
1344 	/*
1345 	 * If we've gone from readonly -> read-write, we need to get our
1346 	 * sync/async discard lists in the right state.
1347 	 */
1348 	btrfs_discard_resume(fs_info);
1349 
1350 	return 0;
1351 }
1352 
1353 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1354 {
1355 	/*
1356 	 * This also happens on 'umount -rf' or on shutdown, when the
1357 	 * filesystem is busy.
1358 	 */
1359 	cancel_work_sync(&fs_info->async_reclaim_work);
1360 	cancel_work_sync(&fs_info->async_data_reclaim_work);
1361 
1362 	btrfs_discard_cleanup(fs_info);
1363 
1364 	/* Wait for the uuid_scan task to finish */
1365 	down(&fs_info->uuid_tree_rescan_sem);
1366 	/* Avoid complains from lockdep et al. */
1367 	up(&fs_info->uuid_tree_rescan_sem);
1368 
1369 	btrfs_set_sb_rdonly(fs_info->sb);
1370 
1371 	/*
1372 	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1373 	 * loop if it's already active.  If it's already asleep, we'll leave
1374 	 * unused block groups on disk until we're mounted read-write again
1375 	 * unless we clean them up here.
1376 	 */
1377 	btrfs_delete_unused_bgs(fs_info);
1378 
1379 	/*
1380 	 * The cleaner task could be already running before we set the flag
1381 	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1382 	 * sure that after we finish the remount, i.e. after we call
1383 	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1384 	 * - either because it was dropping a dead root, running delayed iputs
1385 	 *   or deleting an unused block group (the cleaner picked a block
1386 	 *   group from the list of unused block groups before we were able to
1387 	 *   in the previous call to btrfs_delete_unused_bgs()).
1388 	 */
1389 	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1390 
1391 	/*
1392 	 * We've set the superblock to RO mode, so we might have made the
1393 	 * cleaner task sleep without running all pending delayed iputs. Go
1394 	 * through all the delayed iputs here, so that if an unmount happens
1395 	 * without remounting RW we don't end up at finishing close_ctree()
1396 	 * with a non-empty list of delayed iputs.
1397 	 */
1398 	btrfs_run_delayed_iputs(fs_info);
1399 
1400 	btrfs_dev_replace_suspend_for_unmount(fs_info);
1401 	btrfs_scrub_cancel(fs_info);
1402 	btrfs_pause_balance(fs_info);
1403 
1404 	/*
1405 	 * Pause the qgroup rescan worker if it is running. We don't want it to
1406 	 * be still running after we are in RO mode, as after that, by the time
1407 	 * we unmount, it might have left a transaction open, so we would leak
1408 	 * the transaction and/or crash.
1409 	 */
1410 	btrfs_qgroup_wait_for_completion(fs_info, false);
1411 
1412 	return btrfs_commit_super(fs_info);
1413 }
1414 
1415 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1416 {
1417 	fs_info->max_inline = ctx->max_inline;
1418 	fs_info->commit_interval = ctx->commit_interval;
1419 	fs_info->metadata_ratio = ctx->metadata_ratio;
1420 	fs_info->thread_pool_size = ctx->thread_pool_size;
1421 	fs_info->mount_opt = ctx->mount_opt;
1422 	fs_info->compress_type = ctx->compress_type;
1423 	fs_info->compress_level = ctx->compress_level;
1424 }
1425 
1426 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1427 {
1428 	ctx->max_inline = fs_info->max_inline;
1429 	ctx->commit_interval = fs_info->commit_interval;
1430 	ctx->metadata_ratio = fs_info->metadata_ratio;
1431 	ctx->thread_pool_size = fs_info->thread_pool_size;
1432 	ctx->mount_opt = fs_info->mount_opt;
1433 	ctx->compress_type = fs_info->compress_type;
1434 	ctx->compress_level = fs_info->compress_level;
1435 }
1436 
1437 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1438 do {										\
1439 	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1440 	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1441 		btrfs_info(fs_info, fmt, ##args);				\
1442 } while (0)
1443 
1444 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1445 do {									\
1446 	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1447 	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1448 		btrfs_info(fs_info, fmt, ##args);			\
1449 } while (0)
1450 
1451 static void btrfs_emit_options(struct btrfs_fs_info *info,
1452 			       struct btrfs_fs_context *old)
1453 {
1454 	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1455 	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1456 	btrfs_info_if_set(info, old, NODATACOW, "setting nodatacow");
1457 	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1458 	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1459 	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1460 	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1461 	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1462 	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1463 	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1464 	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1465 	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1466 	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1467 	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1468 	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1469 	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1470 	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1471 	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1472 	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1473 	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1474 	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1475 	btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1476 	btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1477 
1478 	btrfs_info_if_unset(info, old, NODATASUM, "setting datasum");
1479 	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1480 	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1481 	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1482 	btrfs_info_if_unset(info, old, NOBARRIER, "turning on barriers");
1483 	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1484 	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1485 	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1486 	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1487 	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1488 
1489 	/* Did the compression settings change? */
1490 	if (btrfs_test_opt(info, COMPRESS) &&
1491 	    (!old ||
1492 	     old->compress_type != info->compress_type ||
1493 	     old->compress_level != info->compress_level ||
1494 	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1495 	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1496 		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1497 
1498 		btrfs_info(info, "%s %s compression, level %d",
1499 			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1500 			   compress_type, info->compress_level);
1501 	}
1502 
1503 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1504 		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1505 }
1506 
1507 static int btrfs_reconfigure(struct fs_context *fc)
1508 {
1509 	struct super_block *sb = fc->root->d_sb;
1510 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1511 	struct btrfs_fs_context *ctx = fc->fs_private;
1512 	struct btrfs_fs_context old_ctx;
1513 	int ret = 0;
1514 	bool mount_reconfigure = (fc->s_fs_info != NULL);
1515 
1516 	btrfs_info_to_ctx(fs_info, &old_ctx);
1517 
1518 	/*
1519 	 * This is our "bind mount" trick, we don't want to allow the user to do
1520 	 * anything other than mount a different ro/rw and a different subvol,
1521 	 * all of the mount options should be maintained.
1522 	 */
1523 	if (mount_reconfigure)
1524 		ctx->mount_opt = old_ctx.mount_opt;
1525 
1526 	sync_filesystem(sb);
1527 	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1528 
1529 	if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1530 		return -EINVAL;
1531 
1532 	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1533 	if (ret < 0)
1534 		return ret;
1535 
1536 	btrfs_ctx_to_info(fs_info, ctx);
1537 	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1538 	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1539 				 old_ctx.thread_pool_size);
1540 
1541 	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1542 	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1543 	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1544 		btrfs_warn(fs_info,
1545 		"remount supports changing free space tree only from RO to RW");
1546 		/* Make sure free space cache options match the state on disk. */
1547 		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1548 			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1549 			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1550 		}
1551 		if (btrfs_free_space_cache_v1_active(fs_info)) {
1552 			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1553 			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1554 		}
1555 	}
1556 
1557 	ret = 0;
1558 	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1559 		ret = btrfs_remount_ro(fs_info);
1560 	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1561 		ret = btrfs_remount_rw(fs_info);
1562 	if (ret)
1563 		goto restore;
1564 
1565 	/*
1566 	 * If we set the mask during the parameter parsing VFS would reject the
1567 	 * remount.  Here we can set the mask and the value will be updated
1568 	 * appropriately.
1569 	 */
1570 	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1571 		fc->sb_flags_mask |= SB_POSIXACL;
1572 
1573 	btrfs_emit_options(fs_info, &old_ctx);
1574 	wake_up_process(fs_info->transaction_kthread);
1575 	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1576 	btrfs_clear_oneshot_options(fs_info);
1577 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1578 
1579 	return 0;
1580 restore:
1581 	btrfs_ctx_to_info(fs_info, &old_ctx);
1582 	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1583 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1584 	return ret;
1585 }
1586 
1587 /* Used to sort the devices by max_avail(descending sort) */
1588 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1589 {
1590 	const struct btrfs_device_info *dev_info1 = a;
1591 	const struct btrfs_device_info *dev_info2 = b;
1592 
1593 	if (dev_info1->max_avail > dev_info2->max_avail)
1594 		return -1;
1595 	else if (dev_info1->max_avail < dev_info2->max_avail)
1596 		return 1;
1597 	return 0;
1598 }
1599 
1600 /*
1601  * sort the devices by max_avail, in which max free extent size of each device
1602  * is stored.(Descending Sort)
1603  */
1604 static inline void btrfs_descending_sort_devices(
1605 					struct btrfs_device_info *devices,
1606 					size_t nr_devices)
1607 {
1608 	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1609 	     btrfs_cmp_device_free_bytes, NULL);
1610 }
1611 
1612 /*
1613  * The helper to calc the free space on the devices that can be used to store
1614  * file data.
1615  */
1616 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1617 					      u64 *free_bytes)
1618 {
1619 	struct btrfs_device_info *devices_info;
1620 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1621 	struct btrfs_device *device;
1622 	u64 type;
1623 	u64 avail_space;
1624 	u64 min_stripe_size;
1625 	int num_stripes = 1;
1626 	int i = 0, nr_devices;
1627 	const struct btrfs_raid_attr *rattr;
1628 
1629 	/*
1630 	 * We aren't under the device list lock, so this is racy-ish, but good
1631 	 * enough for our purposes.
1632 	 */
1633 	nr_devices = fs_info->fs_devices->open_devices;
1634 	if (!nr_devices) {
1635 		smp_mb();
1636 		nr_devices = fs_info->fs_devices->open_devices;
1637 		ASSERT(nr_devices);
1638 		if (!nr_devices) {
1639 			*free_bytes = 0;
1640 			return 0;
1641 		}
1642 	}
1643 
1644 	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1645 			       GFP_KERNEL);
1646 	if (!devices_info)
1647 		return -ENOMEM;
1648 
1649 	/* calc min stripe number for data space allocation */
1650 	type = btrfs_data_alloc_profile(fs_info);
1651 	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1652 
1653 	if (type & BTRFS_BLOCK_GROUP_RAID0)
1654 		num_stripes = nr_devices;
1655 	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1656 		num_stripes = rattr->ncopies;
1657 	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1658 		num_stripes = 4;
1659 
1660 	/* Adjust for more than 1 stripe per device */
1661 	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1662 
1663 	rcu_read_lock();
1664 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1665 		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1666 						&device->dev_state) ||
1667 		    !device->bdev ||
1668 		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1669 			continue;
1670 
1671 		if (i >= nr_devices)
1672 			break;
1673 
1674 		avail_space = device->total_bytes - device->bytes_used;
1675 
1676 		/* align with stripe_len */
1677 		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1678 
1679 		/*
1680 		 * Ensure we have at least min_stripe_size on top of the
1681 		 * reserved space on the device.
1682 		 */
1683 		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1684 			continue;
1685 
1686 		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1687 
1688 		devices_info[i].dev = device;
1689 		devices_info[i].max_avail = avail_space;
1690 
1691 		i++;
1692 	}
1693 	rcu_read_unlock();
1694 
1695 	nr_devices = i;
1696 
1697 	btrfs_descending_sort_devices(devices_info, nr_devices);
1698 
1699 	i = nr_devices - 1;
1700 	avail_space = 0;
1701 	while (nr_devices >= rattr->devs_min) {
1702 		num_stripes = min(num_stripes, nr_devices);
1703 
1704 		if (devices_info[i].max_avail >= min_stripe_size) {
1705 			int j;
1706 			u64 alloc_size;
1707 
1708 			avail_space += devices_info[i].max_avail * num_stripes;
1709 			alloc_size = devices_info[i].max_avail;
1710 			for (j = i + 1 - num_stripes; j <= i; j++)
1711 				devices_info[j].max_avail -= alloc_size;
1712 		}
1713 		i--;
1714 		nr_devices--;
1715 	}
1716 
1717 	kfree(devices_info);
1718 	*free_bytes = avail_space;
1719 	return 0;
1720 }
1721 
1722 /*
1723  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1724  *
1725  * If there's a redundant raid level at DATA block groups, use the respective
1726  * multiplier to scale the sizes.
1727  *
1728  * Unused device space usage is based on simulating the chunk allocator
1729  * algorithm that respects the device sizes and order of allocations.  This is
1730  * a close approximation of the actual use but there are other factors that may
1731  * change the result (like a new metadata chunk).
1732  *
1733  * If metadata is exhausted, f_bavail will be 0.
1734  */
1735 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1736 {
1737 	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1738 	struct btrfs_super_block *disk_super = fs_info->super_copy;
1739 	struct btrfs_space_info *found;
1740 	u64 total_used = 0;
1741 	u64 total_free_data = 0;
1742 	u64 total_free_meta = 0;
1743 	u32 bits = fs_info->sectorsize_bits;
1744 	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1745 	unsigned factor = 1;
1746 	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1747 	int ret;
1748 	u64 thresh = 0;
1749 	int mixed = 0;
1750 
1751 	list_for_each_entry(found, &fs_info->space_info, list) {
1752 		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1753 			int i;
1754 
1755 			total_free_data += found->disk_total - found->disk_used;
1756 			total_free_data -=
1757 				btrfs_account_ro_block_groups_free_space(found);
1758 
1759 			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1760 				if (!list_empty(&found->block_groups[i]))
1761 					factor = btrfs_bg_type_to_factor(
1762 						btrfs_raid_array[i].bg_flag);
1763 			}
1764 		}
1765 
1766 		/*
1767 		 * Metadata in mixed block group profiles are accounted in data
1768 		 */
1769 		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1770 			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1771 				mixed = 1;
1772 			else
1773 				total_free_meta += found->disk_total -
1774 					found->disk_used;
1775 		}
1776 
1777 		total_used += found->disk_used;
1778 	}
1779 
1780 	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1781 	buf->f_blocks >>= bits;
1782 	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1783 
1784 	/* Account global block reserve as used, it's in logical size already */
1785 	spin_lock(&block_rsv->lock);
1786 	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1787 	if (buf->f_bfree >= block_rsv->size >> bits)
1788 		buf->f_bfree -= block_rsv->size >> bits;
1789 	else
1790 		buf->f_bfree = 0;
1791 	spin_unlock(&block_rsv->lock);
1792 
1793 	buf->f_bavail = div_u64(total_free_data, factor);
1794 	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1795 	if (ret)
1796 		return ret;
1797 	buf->f_bavail += div_u64(total_free_data, factor);
1798 	buf->f_bavail = buf->f_bavail >> bits;
1799 
1800 	/*
1801 	 * We calculate the remaining metadata space minus global reserve. If
1802 	 * this is (supposedly) smaller than zero, there's no space. But this
1803 	 * does not hold in practice, the exhausted state happens where's still
1804 	 * some positive delta. So we apply some guesswork and compare the
1805 	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1806 	 *
1807 	 * We probably cannot calculate the exact threshold value because this
1808 	 * depends on the internal reservations requested by various
1809 	 * operations, so some operations that consume a few metadata will
1810 	 * succeed even if the Avail is zero. But this is better than the other
1811 	 * way around.
1812 	 */
1813 	thresh = SZ_4M;
1814 
1815 	/*
1816 	 * We only want to claim there's no available space if we can no longer
1817 	 * allocate chunks for our metadata profile and our global reserve will
1818 	 * not fit in the free metadata space.  If we aren't ->full then we
1819 	 * still can allocate chunks and thus are fine using the currently
1820 	 * calculated f_bavail.
1821 	 */
1822 	if (!mixed && block_rsv->space_info->full &&
1823 	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1824 		buf->f_bavail = 0;
1825 
1826 	buf->f_type = BTRFS_SUPER_MAGIC;
1827 	buf->f_bsize = fs_info->sectorsize;
1828 	buf->f_namelen = BTRFS_NAME_LEN;
1829 
1830 	/* We treat it as constant endianness (it doesn't matter _which_)
1831 	   because we want the fsid to come out the same whether mounted
1832 	   on a big-endian or little-endian host */
1833 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1834 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1835 	/* Mask in the root object ID too, to disambiguate subvols */
1836 	buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1837 	buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1838 
1839 	return 0;
1840 }
1841 
1842 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1843 {
1844 	struct btrfs_fs_info *p = fc->s_fs_info;
1845 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1846 
1847 	return fs_info->fs_devices == p->fs_devices;
1848 }
1849 
1850 static int btrfs_get_tree_super(struct fs_context *fc)
1851 {
1852 	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1853 	struct btrfs_fs_context *ctx = fc->fs_private;
1854 	struct btrfs_fs_devices *fs_devices = NULL;
1855 	struct btrfs_device *device;
1856 	struct super_block *sb;
1857 	blk_mode_t mode = sb_open_mode(fc->sb_flags);
1858 	int ret;
1859 
1860 	btrfs_ctx_to_info(fs_info, ctx);
1861 	mutex_lock(&uuid_mutex);
1862 
1863 	/*
1864 	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1865 	 * either a valid device or an error.
1866 	 */
1867 	device = btrfs_scan_one_device(fc->source, true);
1868 	ASSERT(device != NULL);
1869 	if (IS_ERR(device)) {
1870 		mutex_unlock(&uuid_mutex);
1871 		return PTR_ERR(device);
1872 	}
1873 	fs_devices = device->fs_devices;
1874 	/*
1875 	 * We cannot hold uuid_mutex calling sget_fc(), it will lead to a
1876 	 * locking order reversal with s_umount.
1877 	 *
1878 	 * So here we increase the holding number of fs_devices, this will ensure
1879 	 * the fs_devices itself won't be freed.
1880 	 */
1881 	btrfs_fs_devices_inc_holding(fs_devices);
1882 	fs_info->fs_devices = fs_devices;
1883 	mutex_unlock(&uuid_mutex);
1884 
1885 
1886 	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1887 	if (IS_ERR(sb)) {
1888 		mutex_lock(&uuid_mutex);
1889 		btrfs_fs_devices_dec_holding(fs_devices);
1890 		/*
1891 		 * Since the fs_devices is not opened, it can be freed at any
1892 		 * time after unlocking uuid_mutex.  We need to avoid double
1893 		 * free through put_fs_context()->btrfs_free_fs_info().
1894 		 * So here we reset fs_info->fs_devices to NULL, and let the
1895 		 * regular fs_devices reclaim path to handle it.
1896 		 *
1897 		 * This applies to all later branches where no fs_devices is
1898 		 * opened.
1899 		 */
1900 		fs_info->fs_devices = NULL;
1901 		mutex_unlock(&uuid_mutex);
1902 		return PTR_ERR(sb);
1903 	}
1904 
1905 	set_device_specific_options(fs_info);
1906 
1907 	if (sb->s_root) {
1908 		/*
1909 		 * Not the first mount of the fs thus got an existing super block.
1910 		 * Will reuse the returned super block, fs_info and fs_devices.
1911 		 *
1912 		 * fc->s_fs_info is not touched and will be later freed by
1913 		 * put_fs_context() through btrfs_free_fs_context().
1914 		 */
1915 		ASSERT(fc->s_fs_info == fs_info);
1916 
1917 		mutex_lock(&uuid_mutex);
1918 		btrfs_fs_devices_dec_holding(fs_devices);
1919 		fs_info->fs_devices = NULL;
1920 		mutex_unlock(&uuid_mutex);
1921 		/*
1922 		 * At this stage we may have RO flag mismatch between
1923 		 * fc->sb_flags and sb->s_flags.  Caller should detect such
1924 		 * mismatch and reconfigure with sb->s_umount rwsem held if
1925 		 * needed.
1926 		 */
1927 	} else {
1928 		struct block_device *bdev;
1929 
1930 		/*
1931 		 * The first mount of the fs thus a new superblock, fc->s_fs_info
1932 		 * must be NULL, and the ownership of our fs_info and fs_devices is
1933 		 * transferred to the super block.
1934 		 */
1935 		ASSERT(fc->s_fs_info == NULL);
1936 
1937 		mutex_lock(&uuid_mutex);
1938 		btrfs_fs_devices_dec_holding(fs_devices);
1939 		ret = btrfs_open_devices(fs_devices, mode, sb);
1940 		if (ret < 0)
1941 			fs_info->fs_devices = NULL;
1942 		mutex_unlock(&uuid_mutex);
1943 		if (ret < 0) {
1944 			deactivate_locked_super(sb);
1945 			return ret;
1946 		}
1947 		if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1948 			deactivate_locked_super(sb);
1949 			return -EACCES;
1950 		}
1951 		bdev = fs_devices->latest_dev->bdev;
1952 		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1953 		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1954 		ret = btrfs_fill_super(sb, fs_devices);
1955 		if (ret) {
1956 			deactivate_locked_super(sb);
1957 			return ret;
1958 		}
1959 	}
1960 
1961 	btrfs_clear_oneshot_options(fs_info);
1962 
1963 	fc->root = dget(sb->s_root);
1964 	return 0;
1965 }
1966 
1967 /*
1968  * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1969  * with different ro/rw options") the following works:
1970  *
1971  *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1972  *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1973  *
1974  * which looks nice and innocent but is actually pretty intricate and deserves
1975  * a long comment.
1976  *
1977  * On another filesystem a subvolume mount is close to something like:
1978  *
1979  *	(iii) # create rw superblock + initial mount
1980  *	      mount -t xfs /dev/sdb /opt/
1981  *
1982  *	      # create ro bind mount
1983  *	      mount --bind -o ro /opt/foo /mnt/foo
1984  *
1985  *	      # unmount initial mount
1986  *	      umount /opt
1987  *
1988  * Of course, there's some special subvolume sauce and there's the fact that the
1989  * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1990  * it's very close and will help us understand the issue.
1991  *
1992  * The old mount API didn't cleanly distinguish between a mount being made ro
1993  * and a superblock being made ro.  The only way to change the ro state of
1994  * either object was by passing ms_rdonly. If a new mount was created via
1995  * mount(2) such as:
1996  *
1997  *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1998  *
1999  * the MS_RDONLY flag being specified had two effects:
2000  *
2001  * (1) MNT_READONLY was raised -> the resulting mount got
2002  *     @mnt->mnt_flags |= MNT_READONLY raised.
2003  *
2004  * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
2005  *     made the superblock ro. Note, how SB_RDONLY has the same value as
2006  *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
2007  *
2008  * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
2009  * subtree mounted ro.
2010  *
2011  * But consider the effect on the old mount API on btrfs subvolume mounting
2012  * which combines the distinct step in (iii) into a single step.
2013  *
2014  * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
2015  * is issued the superblock is ro and thus even if the mount created for (ii) is
2016  * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
2017  * to rw for (ii) which it did using an internal remount call.
2018  *
2019  * IOW, subvolume mounting was inherently complicated due to the ambiguity of
2020  * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
2021  * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
2022  * passed by mount(8) to mount(2).
2023  *
2024  * Enter the new mount API. The new mount API disambiguates making a mount ro
2025  * and making a superblock ro.
2026  *
2027  * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
2028  *     fsmount() or mount_setattr() this is a pure VFS level change for a
2029  *     specific mount or mount tree that is never seen by the filesystem itself.
2030  *
2031  * (4) To turn a superblock ro the "ro" flag must be used with
2032  *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
2033  *     in fc->sb_flags.
2034  *
2035  * But, currently the util-linux mount command already utilizes the new mount
2036  * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's
2037  * btrfs or not, setting the whole super block RO.  To make per-subvolume mounting
2038  * work with different options work we need to keep backward compatibility.
2039  */
2040 static int btrfs_reconfigure_for_mount(struct fs_context *fc)
2041 {
2042 	int ret = 0;
2043 
2044 	if (!(fc->sb_flags & SB_RDONLY) && (fc->root->d_sb->s_flags & SB_RDONLY))
2045 		ret = btrfs_reconfigure(fc);
2046 
2047 	return ret;
2048 }
2049 
2050 static int btrfs_get_tree_subvol(struct fs_context *fc)
2051 {
2052 	struct btrfs_fs_info *fs_info = NULL;
2053 	struct btrfs_fs_context *ctx = fc->fs_private;
2054 	struct fs_context *dup_fc;
2055 	struct dentry *dentry;
2056 	struct vfsmount *mnt;
2057 	int ret = 0;
2058 
2059 	/*
2060 	 * Setup a dummy root and fs_info for test/set super.  This is because
2061 	 * we don't actually fill this stuff out until open_ctree, but we need
2062 	 * then open_ctree will properly initialize the file system specific
2063 	 * settings later.  btrfs_init_fs_info initializes the static elements
2064 	 * of the fs_info (locks and such) to make cleanup easier if we find a
2065 	 * superblock with our given fs_devices later on at sget() time.
2066 	 */
2067 	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2068 	if (!fs_info)
2069 		return -ENOMEM;
2070 
2071 	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2072 	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2073 	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2074 		btrfs_free_fs_info(fs_info);
2075 		return -ENOMEM;
2076 	}
2077 	btrfs_init_fs_info(fs_info);
2078 
2079 	dup_fc = vfs_dup_fs_context(fc);
2080 	if (IS_ERR(dup_fc)) {
2081 		btrfs_free_fs_info(fs_info);
2082 		return PTR_ERR(dup_fc);
2083 	}
2084 
2085 	/*
2086 	 * When we do the sget_fc this gets transferred to the sb, so we only
2087 	 * need to set it on the dup_fc as this is what creates the super block.
2088 	 */
2089 	dup_fc->s_fs_info = fs_info;
2090 
2091 	ret = btrfs_get_tree_super(dup_fc);
2092 	if (ret)
2093 		goto error;
2094 
2095 	ret = btrfs_reconfigure_for_mount(dup_fc);
2096 	up_write(&dup_fc->root->d_sb->s_umount);
2097 	if (ret)
2098 		goto error;
2099 	mnt = vfs_create_mount(dup_fc);
2100 	put_fs_context(dup_fc);
2101 	if (IS_ERR(mnt))
2102 		return PTR_ERR(mnt);
2103 
2104 	/*
2105 	 * This free's ->subvol_name, because if it isn't set we have to
2106 	 * allocate a buffer to hold the subvol_name, so we just drop our
2107 	 * reference to it here.
2108 	 */
2109 	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2110 	ctx->subvol_name = NULL;
2111 	if (IS_ERR(dentry))
2112 		return PTR_ERR(dentry);
2113 
2114 	fc->root = dentry;
2115 	return 0;
2116 error:
2117 	put_fs_context(dup_fc);
2118 	return ret;
2119 }
2120 
2121 static int btrfs_get_tree(struct fs_context *fc)
2122 {
2123 	ASSERT(fc->s_fs_info == NULL);
2124 
2125 	return btrfs_get_tree_subvol(fc);
2126 }
2127 
2128 static void btrfs_kill_super(struct super_block *sb)
2129 {
2130 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2131 	kill_anon_super(sb);
2132 	btrfs_free_fs_info(fs_info);
2133 }
2134 
2135 static void btrfs_free_fs_context(struct fs_context *fc)
2136 {
2137 	struct btrfs_fs_context *ctx = fc->fs_private;
2138 	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2139 
2140 	if (fs_info)
2141 		btrfs_free_fs_info(fs_info);
2142 
2143 	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2144 		kfree(ctx->subvol_name);
2145 		kfree(ctx);
2146 	}
2147 }
2148 
2149 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2150 {
2151 	struct btrfs_fs_context *ctx = src_fc->fs_private;
2152 
2153 	/*
2154 	 * Give a ref to our ctx to this dup, as we want to keep it around for
2155 	 * our original fc so we can have the subvolume name or objectid.
2156 	 *
2157 	 * We unset ->source in the original fc because the dup needs it for
2158 	 * mounting, and then once we free the dup it'll free ->source, so we
2159 	 * need to make sure we're only pointing to it in one fc.
2160 	 */
2161 	refcount_inc(&ctx->refs);
2162 	fc->fs_private = ctx;
2163 	fc->source = src_fc->source;
2164 	src_fc->source = NULL;
2165 	return 0;
2166 }
2167 
2168 static const struct fs_context_operations btrfs_fs_context_ops = {
2169 	.parse_param	= btrfs_parse_param,
2170 	.reconfigure	= btrfs_reconfigure,
2171 	.get_tree	= btrfs_get_tree,
2172 	.dup		= btrfs_dup_fs_context,
2173 	.free		= btrfs_free_fs_context,
2174 };
2175 
2176 static int btrfs_init_fs_context(struct fs_context *fc)
2177 {
2178 	struct btrfs_fs_context *ctx;
2179 
2180 	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2181 	if (!ctx)
2182 		return -ENOMEM;
2183 
2184 	refcount_set(&ctx->refs, 1);
2185 	fc->fs_private = ctx;
2186 	fc->ops = &btrfs_fs_context_ops;
2187 
2188 	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2189 		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2190 	} else {
2191 		ctx->thread_pool_size =
2192 			min_t(unsigned long, num_online_cpus() + 2, 8);
2193 		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2194 		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2195 	}
2196 
2197 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
2198 	fc->sb_flags |= SB_POSIXACL;
2199 #endif
2200 	fc->sb_flags |= SB_I_VERSION;
2201 
2202 	return 0;
2203 }
2204 
2205 static struct file_system_type btrfs_fs_type = {
2206 	.owner			= THIS_MODULE,
2207 	.name			= "btrfs",
2208 	.init_fs_context	= btrfs_init_fs_context,
2209 	.parameters		= btrfs_fs_parameters,
2210 	.kill_sb		= btrfs_kill_super,
2211 	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA |
2212 				  FS_ALLOW_IDMAP | FS_MGTIME,
2213  };
2214 
2215 MODULE_ALIAS_FS("btrfs");
2216 
2217 static int btrfs_control_open(struct inode *inode, struct file *file)
2218 {
2219 	/*
2220 	 * The control file's private_data is used to hold the
2221 	 * transaction when it is started and is used to keep
2222 	 * track of whether a transaction is already in progress.
2223 	 */
2224 	file->private_data = NULL;
2225 	return 0;
2226 }
2227 
2228 /*
2229  * Used by /dev/btrfs-control for devices ioctls.
2230  */
2231 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2232 				unsigned long arg)
2233 {
2234 	struct btrfs_ioctl_vol_args *vol;
2235 	struct btrfs_device *device = NULL;
2236 	dev_t devt = 0;
2237 	int ret = -ENOTTY;
2238 
2239 	if (!capable(CAP_SYS_ADMIN))
2240 		return -EPERM;
2241 
2242 	vol = memdup_user((void __user *)arg, sizeof(*vol));
2243 	if (IS_ERR(vol))
2244 		return PTR_ERR(vol);
2245 	ret = btrfs_check_ioctl_vol_args_path(vol);
2246 	if (ret < 0)
2247 		goto out;
2248 
2249 	switch (cmd) {
2250 	case BTRFS_IOC_SCAN_DEV:
2251 		mutex_lock(&uuid_mutex);
2252 		/*
2253 		 * Scanning outside of mount can return NULL which would turn
2254 		 * into 0 error code.
2255 		 */
2256 		device = btrfs_scan_one_device(vol->name, false);
2257 		ret = PTR_ERR_OR_ZERO(device);
2258 		mutex_unlock(&uuid_mutex);
2259 		break;
2260 	case BTRFS_IOC_FORGET_DEV:
2261 		if (vol->name[0] != 0) {
2262 			ret = lookup_bdev(vol->name, &devt);
2263 			if (ret)
2264 				break;
2265 		}
2266 		ret = btrfs_forget_devices(devt);
2267 		break;
2268 	case BTRFS_IOC_DEVICES_READY:
2269 		mutex_lock(&uuid_mutex);
2270 		/*
2271 		 * Scanning outside of mount can return NULL which would turn
2272 		 * into 0 error code.
2273 		 */
2274 		device = btrfs_scan_one_device(vol->name, false);
2275 		if (IS_ERR_OR_NULL(device)) {
2276 			mutex_unlock(&uuid_mutex);
2277 			if (IS_ERR(device))
2278 				ret = PTR_ERR(device);
2279 			else
2280 				ret = 0;
2281 			break;
2282 		}
2283 		ret = !(device->fs_devices->num_devices ==
2284 			device->fs_devices->total_devices);
2285 		mutex_unlock(&uuid_mutex);
2286 		break;
2287 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2288 		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2289 		break;
2290 	}
2291 
2292 out:
2293 	kfree(vol);
2294 	return ret;
2295 }
2296 
2297 static int btrfs_freeze(struct super_block *sb)
2298 {
2299 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2300 
2301 	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2302 	/*
2303 	 * We don't need a barrier here, we'll wait for any transaction that
2304 	 * could be in progress on other threads (and do delayed iputs that
2305 	 * we want to avoid on a frozen filesystem), or do the commit
2306 	 * ourselves.
2307 	 */
2308 	return btrfs_commit_current_transaction(fs_info->tree_root);
2309 }
2310 
2311 static int check_dev_super(struct btrfs_device *dev)
2312 {
2313 	struct btrfs_fs_info *fs_info = dev->fs_info;
2314 	struct btrfs_super_block *sb;
2315 	u64 last_trans;
2316 	u16 csum_type;
2317 	int ret = 0;
2318 
2319 	/* This should be called with fs still frozen. */
2320 	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2321 
2322 	/* Missing dev, no need to check. */
2323 	if (!dev->bdev)
2324 		return 0;
2325 
2326 	/* Only need to check the primary super block. */
2327 	sb = btrfs_read_disk_super(dev->bdev, 0, true);
2328 	if (IS_ERR(sb))
2329 		return PTR_ERR(sb);
2330 
2331 	/* Verify the checksum. */
2332 	csum_type = btrfs_super_csum_type(sb);
2333 	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2334 		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2335 			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2336 		ret = -EUCLEAN;
2337 		goto out;
2338 	}
2339 
2340 	if (btrfs_check_super_csum(fs_info, sb)) {
2341 		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2342 		ret = -EUCLEAN;
2343 		goto out;
2344 	}
2345 
2346 	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2347 	ret = btrfs_validate_super(fs_info, sb, 0);
2348 	if (ret < 0)
2349 		goto out;
2350 
2351 	last_trans = btrfs_get_last_trans_committed(fs_info);
2352 	if (btrfs_super_generation(sb) != last_trans) {
2353 		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2354 			  btrfs_super_generation(sb), last_trans);
2355 		ret = -EUCLEAN;
2356 		goto out;
2357 	}
2358 out:
2359 	btrfs_release_disk_super(sb);
2360 	return ret;
2361 }
2362 
2363 static int btrfs_unfreeze(struct super_block *sb)
2364 {
2365 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2366 	struct btrfs_device *device;
2367 	int ret = 0;
2368 
2369 	/*
2370 	 * Make sure the fs is not changed by accident (like hibernation then
2371 	 * modified by other OS).
2372 	 * If we found anything wrong, we mark the fs error immediately.
2373 	 *
2374 	 * And since the fs is frozen, no one can modify the fs yet, thus
2375 	 * we don't need to hold device_list_mutex.
2376 	 */
2377 	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2378 		ret = check_dev_super(device);
2379 		if (ret < 0) {
2380 			btrfs_handle_fs_error(fs_info, ret,
2381 				"super block on devid %llu got modified unexpectedly",
2382 				device->devid);
2383 			break;
2384 		}
2385 	}
2386 	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2387 
2388 	/*
2389 	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2390 	 * above checks failed. Since the fs is either fine or read-only, we're
2391 	 * safe to continue, without causing further damage.
2392 	 */
2393 	return 0;
2394 }
2395 
2396 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2397 {
2398 	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2399 
2400 	/*
2401 	 * There should be always a valid pointer in latest_dev, it may be stale
2402 	 * for a short moment in case it's being deleted but still valid until
2403 	 * the end of RCU grace period.
2404 	 */
2405 	rcu_read_lock();
2406 	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2407 	rcu_read_unlock();
2408 
2409 	return 0;
2410 }
2411 
2412 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2413 {
2414 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2415 	const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2416 
2417 	trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2418 
2419 	return nr;
2420 }
2421 
2422 static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc)
2423 {
2424 	const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan);
2425 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2426 
2427 	btrfs_free_extent_maps(fs_info, nr_to_scan);
2428 
2429 	/* The extent map shrinker runs asynchronously, so always return 0. */
2430 	return 0;
2431 }
2432 
2433 static const struct super_operations btrfs_super_ops = {
2434 	.drop_inode	= btrfs_drop_inode,
2435 	.evict_inode	= btrfs_evict_inode,
2436 	.put_super	= btrfs_put_super,
2437 	.sync_fs	= btrfs_sync_fs,
2438 	.show_options	= btrfs_show_options,
2439 	.show_devname	= btrfs_show_devname,
2440 	.alloc_inode	= btrfs_alloc_inode,
2441 	.destroy_inode	= btrfs_destroy_inode,
2442 	.free_inode	= btrfs_free_inode,
2443 	.statfs		= btrfs_statfs,
2444 	.freeze_fs	= btrfs_freeze,
2445 	.unfreeze_fs	= btrfs_unfreeze,
2446 	.nr_cached_objects = btrfs_nr_cached_objects,
2447 	.free_cached_objects = btrfs_free_cached_objects,
2448 };
2449 
2450 static const struct file_operations btrfs_ctl_fops = {
2451 	.open = btrfs_control_open,
2452 	.unlocked_ioctl	 = btrfs_control_ioctl,
2453 	.compat_ioctl = compat_ptr_ioctl,
2454 	.owner	 = THIS_MODULE,
2455 	.llseek = noop_llseek,
2456 };
2457 
2458 static struct miscdevice btrfs_misc = {
2459 	.minor		= BTRFS_MINOR,
2460 	.name		= "btrfs-control",
2461 	.fops		= &btrfs_ctl_fops
2462 };
2463 
2464 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2465 MODULE_ALIAS("devname:btrfs-control");
2466 
2467 static int __init btrfs_interface_init(void)
2468 {
2469 	return misc_register(&btrfs_misc);
2470 }
2471 
2472 static __cold void btrfs_interface_exit(void)
2473 {
2474 	misc_deregister(&btrfs_misc);
2475 }
2476 
2477 static int __init btrfs_print_mod_info(void)
2478 {
2479 	static const char options[] = ""
2480 #ifdef CONFIG_BTRFS_EXPERIMENTAL
2481 			", experimental=on"
2482 #endif
2483 #ifdef CONFIG_BTRFS_DEBUG
2484 			", debug=on"
2485 #endif
2486 #ifdef CONFIG_BTRFS_ASSERT
2487 			", assert=on"
2488 #endif
2489 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2490 			", ref-verify=on"
2491 #endif
2492 #ifdef CONFIG_BLK_DEV_ZONED
2493 			", zoned=yes"
2494 #else
2495 			", zoned=no"
2496 #endif
2497 #ifdef CONFIG_FS_VERITY
2498 			", fsverity=yes"
2499 #else
2500 			", fsverity=no"
2501 #endif
2502 			;
2503 
2504 #ifdef CONFIG_BTRFS_EXPERIMENTAL
2505 	if (btrfs_get_mod_read_policy() == NULL)
2506 		pr_info("Btrfs loaded%s\n", options);
2507 	else
2508 		pr_info("Btrfs loaded%s, read_policy=%s\n",
2509 			 options, btrfs_get_mod_read_policy());
2510 #else
2511 	pr_info("Btrfs loaded%s\n", options);
2512 #endif
2513 
2514 	return 0;
2515 }
2516 
2517 static int register_btrfs(void)
2518 {
2519 	return register_filesystem(&btrfs_fs_type);
2520 }
2521 
2522 static void unregister_btrfs(void)
2523 {
2524 	unregister_filesystem(&btrfs_fs_type);
2525 }
2526 
2527 /* Helper structure for long init/exit functions. */
2528 struct init_sequence {
2529 	int (*init_func)(void);
2530 	/* Can be NULL if the init_func doesn't need cleanup. */
2531 	void (*exit_func)(void);
2532 };
2533 
2534 static const struct init_sequence mod_init_seq[] = {
2535 	{
2536 		.init_func = btrfs_props_init,
2537 		.exit_func = NULL,
2538 	}, {
2539 		.init_func = btrfs_init_sysfs,
2540 		.exit_func = btrfs_exit_sysfs,
2541 	}, {
2542 		.init_func = btrfs_init_compress,
2543 		.exit_func = btrfs_exit_compress,
2544 	}, {
2545 		.init_func = btrfs_init_cachep,
2546 		.exit_func = btrfs_destroy_cachep,
2547 	}, {
2548 		.init_func = btrfs_init_dio,
2549 		.exit_func = btrfs_destroy_dio,
2550 	}, {
2551 		.init_func = btrfs_transaction_init,
2552 		.exit_func = btrfs_transaction_exit,
2553 	}, {
2554 		.init_func = btrfs_ctree_init,
2555 		.exit_func = btrfs_ctree_exit,
2556 	}, {
2557 		.init_func = btrfs_free_space_init,
2558 		.exit_func = btrfs_free_space_exit,
2559 	}, {
2560 		.init_func = btrfs_extent_state_init_cachep,
2561 		.exit_func = btrfs_extent_state_free_cachep,
2562 	}, {
2563 		.init_func = extent_buffer_init_cachep,
2564 		.exit_func = extent_buffer_free_cachep,
2565 	}, {
2566 		.init_func = btrfs_bioset_init,
2567 		.exit_func = btrfs_bioset_exit,
2568 	}, {
2569 		.init_func = btrfs_extent_map_init,
2570 		.exit_func = btrfs_extent_map_exit,
2571 #ifdef CONFIG_BTRFS_EXPERIMENTAL
2572 	}, {
2573 		.init_func = btrfs_read_policy_init,
2574 		.exit_func = NULL,
2575 #endif
2576 	}, {
2577 		.init_func = ordered_data_init,
2578 		.exit_func = ordered_data_exit,
2579 	}, {
2580 		.init_func = btrfs_delayed_inode_init,
2581 		.exit_func = btrfs_delayed_inode_exit,
2582 	}, {
2583 		.init_func = btrfs_auto_defrag_init,
2584 		.exit_func = btrfs_auto_defrag_exit,
2585 	}, {
2586 		.init_func = btrfs_delayed_ref_init,
2587 		.exit_func = btrfs_delayed_ref_exit,
2588 	}, {
2589 		.init_func = btrfs_prelim_ref_init,
2590 		.exit_func = btrfs_prelim_ref_exit,
2591 	}, {
2592 		.init_func = btrfs_interface_init,
2593 		.exit_func = btrfs_interface_exit,
2594 	}, {
2595 		.init_func = btrfs_print_mod_info,
2596 		.exit_func = NULL,
2597 	}, {
2598 		.init_func = btrfs_run_sanity_tests,
2599 		.exit_func = NULL,
2600 	}, {
2601 		.init_func = register_btrfs,
2602 		.exit_func = unregister_btrfs,
2603 	}
2604 };
2605 
2606 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2607 
2608 static __always_inline void btrfs_exit_btrfs_fs(void)
2609 {
2610 	int i;
2611 
2612 	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2613 		if (!mod_init_result[i])
2614 			continue;
2615 		if (mod_init_seq[i].exit_func)
2616 			mod_init_seq[i].exit_func();
2617 		mod_init_result[i] = false;
2618 	}
2619 }
2620 
2621 static void __exit exit_btrfs_fs(void)
2622 {
2623 	btrfs_exit_btrfs_fs();
2624 	btrfs_cleanup_fs_uuids();
2625 }
2626 
2627 static int __init init_btrfs_fs(void)
2628 {
2629 	int ret;
2630 	int i;
2631 
2632 	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2633 		ASSERT(!mod_init_result[i]);
2634 		ret = mod_init_seq[i].init_func();
2635 		if (ret < 0) {
2636 			btrfs_exit_btrfs_fs();
2637 			return ret;
2638 		}
2639 		mod_init_result[i] = true;
2640 	}
2641 	return 0;
2642 }
2643 
2644 late_initcall(init_btrfs_fs);
2645 module_exit(exit_btrfs_fs)
2646 
2647 MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2648 MODULE_LICENSE("GPL");
2649 MODULE_SOFTDEP("pre: crc32c");
2650 MODULE_SOFTDEP("pre: xxhash64");
2651 MODULE_SOFTDEP("pre: sha256");
2652 MODULE_SOFTDEP("pre: blake2b-256");
2653