xref: /linux/fs/btrfs/super.c (revision 4e46774408d942efe4eb35dc62e5af3af71b9a30)
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 <linux/swap.h>
32 #include "messages.h"
33 #include "delayed-inode.h"
34 #include "ctree.h"
35 #include "disk-io.h"
36 #include "transaction.h"
37 #include "btrfs_inode.h"
38 #include "direct-io.h"
39 #include "props.h"
40 #include "xattr.h"
41 #include "bio.h"
42 #include "export.h"
43 #include "compression.h"
44 #include "dev-replace.h"
45 #include "free-space-cache.h"
46 #include "backref.h"
47 #include "space-info.h"
48 #include "sysfs.h"
49 #include "zoned.h"
50 #include "tests/btrfs-tests.h"
51 #include "block-group.h"
52 #include "discard.h"
53 #include "qgroup.h"
54 #include "raid56.h"
55 #include "fs.h"
56 #include "accessors.h"
57 #include "defrag.h"
58 #include "dir-item.h"
59 #include "ioctl.h"
60 #include "scrub.h"
61 #include "verity.h"
62 #include "super.h"
63 #include "extent-tree.h"
64 #define CREATE_TRACE_POINTS
65 #include <trace/events/btrfs.h>
66 
67 static const struct super_operations btrfs_super_ops;
68 static struct file_system_type btrfs_fs_type;
69 
btrfs_put_super(struct super_block * sb)70 static void btrfs_put_super(struct super_block *sb)
71 {
72 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
73 
74 	btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid);
75 	close_ctree(fs_info);
76 }
77 
78 /* Store the mount options related information. */
79 struct btrfs_fs_context {
80 	char *subvol_name;
81 	u64 subvol_objectid;
82 	u64 max_inline;
83 	u32 commit_interval;
84 	u32 metadata_ratio;
85 	u32 thread_pool_size;
86 	unsigned long long mount_opt;
87 	unsigned long compress_type:4;
88 	unsigned int compress_level;
89 	refcount_t refs;
90 };
91 
92 enum {
93 	Opt_acl,
94 	Opt_clear_cache,
95 	Opt_commit_interval,
96 	Opt_compress,
97 	Opt_compress_force,
98 	Opt_compress_force_type,
99 	Opt_compress_type,
100 	Opt_degraded,
101 	Opt_device,
102 	Opt_fatal_errors,
103 	Opt_flushoncommit,
104 	Opt_max_inline,
105 	Opt_barrier,
106 	Opt_datacow,
107 	Opt_datasum,
108 	Opt_defrag,
109 	Opt_discard,
110 	Opt_discard_mode,
111 	Opt_ratio,
112 	Opt_rescan_uuid_tree,
113 	Opt_skip_balance,
114 	Opt_space_cache,
115 	Opt_space_cache_version,
116 	Opt_ssd,
117 	Opt_ssd_spread,
118 	Opt_subvol,
119 	Opt_subvol_empty,
120 	Opt_subvolid,
121 	Opt_thread_pool,
122 	Opt_treelog,
123 	Opt_user_subvol_rm_allowed,
124 	Opt_norecovery,
125 
126 	/* Rescue options */
127 	Opt_rescue,
128 	Opt_usebackuproot,
129 	Opt_nologreplay,
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_ignoremetacsums,
181 	Opt_rescue_ignoresuperflags,
182 	Opt_rescue_parameter_all,
183 };
184 
185 static const struct constant_table btrfs_parameter_rescue[] = {
186 	{ "usebackuproot", Opt_rescue_usebackuproot },
187 	{ "nologreplay", Opt_rescue_nologreplay },
188 	{ "ignorebadroots", Opt_rescue_ignorebadroots },
189 	{ "ibadroots", Opt_rescue_ignorebadroots },
190 	{ "ignoredatacsums", Opt_rescue_ignoredatacsums },
191 	{ "ignoremetacsums", Opt_rescue_ignoremetacsums},
192 	{ "ignoresuperflags", Opt_rescue_ignoresuperflags},
193 	{ "idatacsums", Opt_rescue_ignoredatacsums },
194 	{ "imetacsums", Opt_rescue_ignoremetacsums},
195 	{ "isuperflags", Opt_rescue_ignoresuperflags},
196 	{ "all", Opt_rescue_parameter_all },
197 	{}
198 };
199 
200 #ifdef CONFIG_BTRFS_DEBUG
201 enum {
202 	Opt_fragment_parameter_data,
203 	Opt_fragment_parameter_metadata,
204 	Opt_fragment_parameter_all,
205 };
206 
207 static const struct constant_table btrfs_parameter_fragment[] = {
208 	{ "data", Opt_fragment_parameter_data },
209 	{ "metadata", Opt_fragment_parameter_metadata },
210 	{ "all", Opt_fragment_parameter_all },
211 	{}
212 };
213 #endif
214 
215 static const struct fs_parameter_spec btrfs_fs_parameters[] = {
216 	fsparam_flag_no("acl", Opt_acl),
217 	fsparam_flag_no("autodefrag", Opt_defrag),
218 	fsparam_flag_no("barrier", Opt_barrier),
219 	fsparam_flag("clear_cache", Opt_clear_cache),
220 	fsparam_u32("commit", Opt_commit_interval),
221 	fsparam_flag("compress", Opt_compress),
222 	fsparam_string("compress", Opt_compress_type),
223 	fsparam_flag("compress-force", Opt_compress_force),
224 	fsparam_string("compress-force", Opt_compress_force_type),
225 	fsparam_flag_no("datacow", Opt_datacow),
226 	fsparam_flag_no("datasum", Opt_datasum),
227 	fsparam_flag("degraded", Opt_degraded),
228 	fsparam_string("device", Opt_device),
229 	fsparam_flag_no("discard", Opt_discard),
230 	fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard),
231 	fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors),
232 	fsparam_flag_no("flushoncommit", Opt_flushoncommit),
233 	fsparam_string("max_inline", Opt_max_inline),
234 	fsparam_u32("metadata_ratio", Opt_ratio),
235 	fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree),
236 	fsparam_flag("skip_balance", Opt_skip_balance),
237 	fsparam_flag_no("space_cache", Opt_space_cache),
238 	fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache),
239 	fsparam_flag_no("ssd", Opt_ssd),
240 	fsparam_flag_no("ssd_spread", Opt_ssd_spread),
241 	fsparam_string("subvol", Opt_subvol),
242 	fsparam_flag("subvol=", Opt_subvol_empty),
243 	fsparam_u64("subvolid", Opt_subvolid),
244 	fsparam_u32("thread_pool", Opt_thread_pool),
245 	fsparam_flag_no("treelog", Opt_treelog),
246 	fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed),
247 
248 	/* Rescue options. */
249 	fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue),
250 	/* Deprecated, with alias rescue=nologreplay */
251 	__fsparam(NULL, "nologreplay", Opt_nologreplay, fs_param_deprecated, NULL),
252 	/* Deprecated, with alias rescue=usebackuproot */
253 	__fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL),
254 	/* For compatibility only, alias for "rescue=nologreplay". */
255 	fsparam_flag("norecovery", Opt_norecovery),
256 
257 	/* Debugging options. */
258 	fsparam_flag_no("enospc_debug", Opt_enospc_debug),
259 #ifdef CONFIG_BTRFS_DEBUG
260 	fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment),
261 #endif
262 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
263 	fsparam_flag("ref_verify", Opt_ref_verify),
264 #endif
265 	{}
266 };
267 
268 /* No support for restricting writes to btrfs devices yet... */
btrfs_open_mode(struct fs_context * fc)269 static inline blk_mode_t btrfs_open_mode(struct fs_context *fc)
270 {
271 	return sb_open_mode(fc->sb_flags) & ~BLK_OPEN_RESTRICT_WRITES;
272 }
273 
btrfs_parse_param(struct fs_context * fc,struct fs_parameter * param)274 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param)
275 {
276 	struct btrfs_fs_context *ctx = fc->fs_private;
277 	struct fs_parse_result result;
278 	int opt;
279 
280 	opt = fs_parse(fc, btrfs_fs_parameters, param, &result);
281 	if (opt < 0)
282 		return opt;
283 
284 	switch (opt) {
285 	case Opt_degraded:
286 		btrfs_set_opt(ctx->mount_opt, DEGRADED);
287 		break;
288 	case Opt_subvol_empty:
289 		/*
290 		 * This exists because we used to allow it on accident, so we're
291 		 * keeping it to maintain ABI.  See 37becec95ac3 ("Btrfs: allow
292 		 * empty subvol= again").
293 		 */
294 		break;
295 	case Opt_subvol:
296 		kfree(ctx->subvol_name);
297 		ctx->subvol_name = kstrdup(param->string, GFP_KERNEL);
298 		if (!ctx->subvol_name)
299 			return -ENOMEM;
300 		break;
301 	case Opt_subvolid:
302 		ctx->subvol_objectid = result.uint_64;
303 
304 		/* subvolid=0 means give me the original fs_tree. */
305 		if (!ctx->subvol_objectid)
306 			ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID;
307 		break;
308 	case Opt_device: {
309 		struct btrfs_device *device;
310 		blk_mode_t mode = btrfs_open_mode(fc);
311 
312 		mutex_lock(&uuid_mutex);
313 		device = btrfs_scan_one_device(param->string, mode, false);
314 		mutex_unlock(&uuid_mutex);
315 		if (IS_ERR(device))
316 			return PTR_ERR(device);
317 		break;
318 	}
319 	case Opt_datasum:
320 		if (result.negated) {
321 			btrfs_set_opt(ctx->mount_opt, NODATASUM);
322 		} else {
323 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
324 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
325 		}
326 		break;
327 	case Opt_datacow:
328 		if (result.negated) {
329 			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
330 			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
331 			btrfs_set_opt(ctx->mount_opt, NODATACOW);
332 			btrfs_set_opt(ctx->mount_opt, NODATASUM);
333 		} else {
334 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
335 		}
336 		break;
337 	case Opt_compress_force:
338 	case Opt_compress_force_type:
339 		btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS);
340 		fallthrough;
341 	case Opt_compress:
342 	case Opt_compress_type:
343 		/*
344 		 * Provide the same semantics as older kernels that don't use fs
345 		 * context, specifying the "compress" option clears
346 		 * "force-compress" without the need to pass
347 		 * "compress-force=[no|none]" before specifying "compress".
348 		 */
349 		if (opt != Opt_compress_force && opt != Opt_compress_force_type)
350 			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
351 
352 		if (opt == Opt_compress || opt == Opt_compress_force) {
353 			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
354 			ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL;
355 			btrfs_set_opt(ctx->mount_opt, COMPRESS);
356 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
357 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
358 		} else if (strncmp(param->string, "zlib", 4) == 0) {
359 			ctx->compress_type = BTRFS_COMPRESS_ZLIB;
360 			ctx->compress_level =
361 				btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB,
362 							 param->string + 4);
363 			btrfs_set_opt(ctx->mount_opt, COMPRESS);
364 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
365 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
366 		} else if (strncmp(param->string, "lzo", 3) == 0) {
367 			ctx->compress_type = BTRFS_COMPRESS_LZO;
368 			ctx->compress_level = 0;
369 			btrfs_set_opt(ctx->mount_opt, COMPRESS);
370 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
371 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
372 		} else if (strncmp(param->string, "zstd", 4) == 0) {
373 			ctx->compress_type = BTRFS_COMPRESS_ZSTD;
374 			ctx->compress_level =
375 				btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD,
376 							 param->string + 4);
377 			btrfs_set_opt(ctx->mount_opt, COMPRESS);
378 			btrfs_clear_opt(ctx->mount_opt, NODATACOW);
379 			btrfs_clear_opt(ctx->mount_opt, NODATASUM);
380 		} else if (strncmp(param->string, "no", 2) == 0) {
381 			ctx->compress_level = 0;
382 			ctx->compress_type = 0;
383 			btrfs_clear_opt(ctx->mount_opt, COMPRESS);
384 			btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS);
385 		} else {
386 			btrfs_err(NULL, "unrecognized compression value %s",
387 				  param->string);
388 			return -EINVAL;
389 		}
390 		break;
391 	case Opt_ssd:
392 		if (result.negated) {
393 			btrfs_set_opt(ctx->mount_opt, NOSSD);
394 			btrfs_clear_opt(ctx->mount_opt, SSD);
395 			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
396 		} else {
397 			btrfs_set_opt(ctx->mount_opt, SSD);
398 			btrfs_clear_opt(ctx->mount_opt, NOSSD);
399 		}
400 		break;
401 	case Opt_ssd_spread:
402 		if (result.negated) {
403 			btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD);
404 		} else {
405 			btrfs_set_opt(ctx->mount_opt, SSD);
406 			btrfs_set_opt(ctx->mount_opt, SSD_SPREAD);
407 			btrfs_clear_opt(ctx->mount_opt, NOSSD);
408 		}
409 		break;
410 	case Opt_barrier:
411 		if (result.negated)
412 			btrfs_set_opt(ctx->mount_opt, NOBARRIER);
413 		else
414 			btrfs_clear_opt(ctx->mount_opt, NOBARRIER);
415 		break;
416 	case Opt_thread_pool:
417 		if (result.uint_32 == 0) {
418 			btrfs_err(NULL, "invalid value 0 for thread_pool");
419 			return -EINVAL;
420 		}
421 		ctx->thread_pool_size = result.uint_32;
422 		break;
423 	case Opt_max_inline:
424 		ctx->max_inline = memparse(param->string, NULL);
425 		break;
426 	case Opt_acl:
427 		if (result.negated) {
428 			fc->sb_flags &= ~SB_POSIXACL;
429 		} else {
430 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
431 			fc->sb_flags |= SB_POSIXACL;
432 #else
433 			btrfs_err(NULL, "support for ACL not compiled in");
434 			return -EINVAL;
435 #endif
436 		}
437 		/*
438 		 * VFS limits the ability to toggle ACL on and off via remount,
439 		 * despite every file system allowing this.  This seems to be
440 		 * an oversight since we all do, but it'll fail if we're
441 		 * remounting.  So don't set the mask here, we'll check it in
442 		 * btrfs_reconfigure and do the toggling ourselves.
443 		 */
444 		if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE)
445 			fc->sb_flags_mask |= SB_POSIXACL;
446 		break;
447 	case Opt_treelog:
448 		if (result.negated)
449 			btrfs_set_opt(ctx->mount_opt, NOTREELOG);
450 		else
451 			btrfs_clear_opt(ctx->mount_opt, NOTREELOG);
452 		break;
453 	case Opt_nologreplay:
454 		btrfs_warn(NULL,
455 		"'nologreplay' is deprecated, use 'rescue=nologreplay' instead");
456 		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
457 		break;
458 	case Opt_norecovery:
459 		btrfs_info(NULL,
460 "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'");
461 		btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
462 		break;
463 	case Opt_flushoncommit:
464 		if (result.negated)
465 			btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT);
466 		else
467 			btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT);
468 		break;
469 	case Opt_ratio:
470 		ctx->metadata_ratio = result.uint_32;
471 		break;
472 	case Opt_discard:
473 		if (result.negated) {
474 			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
475 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
476 			btrfs_set_opt(ctx->mount_opt, NODISCARD);
477 		} else {
478 			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
479 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
480 		}
481 		break;
482 	case Opt_discard_mode:
483 		switch (result.uint_32) {
484 		case Opt_discard_sync:
485 			btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC);
486 			btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC);
487 			break;
488 		case Opt_discard_async:
489 			btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC);
490 			btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC);
491 			break;
492 		default:
493 			btrfs_err(NULL, "unrecognized discard mode value %s",
494 				  param->key);
495 			return -EINVAL;
496 		}
497 		btrfs_clear_opt(ctx->mount_opt, NODISCARD);
498 		break;
499 	case Opt_space_cache:
500 		if (result.negated) {
501 			btrfs_set_opt(ctx->mount_opt, NOSPACECACHE);
502 			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
503 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
504 		} else {
505 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
506 			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
507 		}
508 		break;
509 	case Opt_space_cache_version:
510 		switch (result.uint_32) {
511 		case Opt_space_cache_v1:
512 			btrfs_set_opt(ctx->mount_opt, SPACE_CACHE);
513 			btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE);
514 			break;
515 		case Opt_space_cache_v2:
516 			btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE);
517 			btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE);
518 			break;
519 		default:
520 			btrfs_err(NULL, "unrecognized space_cache value %s",
521 				  param->key);
522 			return -EINVAL;
523 		}
524 		break;
525 	case Opt_rescan_uuid_tree:
526 		btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE);
527 		break;
528 	case Opt_clear_cache:
529 		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
530 		break;
531 	case Opt_user_subvol_rm_allowed:
532 		btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED);
533 		break;
534 	case Opt_enospc_debug:
535 		if (result.negated)
536 			btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG);
537 		else
538 			btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG);
539 		break;
540 	case Opt_defrag:
541 		if (result.negated)
542 			btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG);
543 		else
544 			btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG);
545 		break;
546 	case Opt_usebackuproot:
547 		btrfs_warn(NULL,
548 			   "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead");
549 		btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
550 
551 		/* If we're loading the backup roots we can't trust the space cache. */
552 		btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE);
553 		break;
554 	case Opt_skip_balance:
555 		btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE);
556 		break;
557 	case Opt_fatal_errors:
558 		switch (result.uint_32) {
559 		case Opt_fatal_errors_panic:
560 			btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
561 			break;
562 		case Opt_fatal_errors_bug:
563 			btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR);
564 			break;
565 		default:
566 			btrfs_err(NULL, "unrecognized fatal_errors value %s",
567 				  param->key);
568 			return -EINVAL;
569 		}
570 		break;
571 	case Opt_commit_interval:
572 		ctx->commit_interval = result.uint_32;
573 		if (ctx->commit_interval == 0)
574 			ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
575 		break;
576 	case Opt_rescue:
577 		switch (result.uint_32) {
578 		case Opt_rescue_usebackuproot:
579 			btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT);
580 			break;
581 		case Opt_rescue_nologreplay:
582 			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
583 			break;
584 		case Opt_rescue_ignorebadroots:
585 			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
586 			break;
587 		case Opt_rescue_ignoredatacsums:
588 			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
589 			break;
590 		case Opt_rescue_ignoremetacsums:
591 			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
592 			break;
593 		case Opt_rescue_ignoresuperflags:
594 			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
595 			break;
596 		case Opt_rescue_parameter_all:
597 			btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS);
598 			btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS);
599 			btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS);
600 			btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS);
601 			btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY);
602 			break;
603 		default:
604 			btrfs_info(NULL, "unrecognized rescue option '%s'",
605 				   param->key);
606 			return -EINVAL;
607 		}
608 		break;
609 #ifdef CONFIG_BTRFS_DEBUG
610 	case Opt_fragment:
611 		switch (result.uint_32) {
612 		case Opt_fragment_parameter_all:
613 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
614 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
615 			break;
616 		case Opt_fragment_parameter_metadata:
617 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA);
618 			break;
619 		case Opt_fragment_parameter_data:
620 			btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA);
621 			break;
622 		default:
623 			btrfs_info(NULL, "unrecognized fragment option '%s'",
624 				   param->key);
625 			return -EINVAL;
626 		}
627 		break;
628 #endif
629 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
630 	case Opt_ref_verify:
631 		btrfs_set_opt(ctx->mount_opt, REF_VERIFY);
632 		break;
633 #endif
634 	default:
635 		btrfs_err(NULL, "unrecognized mount option '%s'", param->key);
636 		return -EINVAL;
637 	}
638 
639 	return 0;
640 }
641 
642 /*
643  * Some options only have meaning at mount time and shouldn't persist across
644  * remounts, or be displayed. Clear these at the end of mount and remount code
645  * paths.
646  */
btrfs_clear_oneshot_options(struct btrfs_fs_info * fs_info)647 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info)
648 {
649 	btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT);
650 	btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE);
651 	btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE);
652 }
653 
check_ro_option(const struct btrfs_fs_info * fs_info,unsigned long long mount_opt,unsigned long long opt,const char * opt_name)654 static bool check_ro_option(const struct btrfs_fs_info *fs_info,
655 			    unsigned long long mount_opt, unsigned long long opt,
656 			    const char *opt_name)
657 {
658 	if (mount_opt & opt) {
659 		btrfs_err(fs_info, "%s must be used with ro mount option",
660 			  opt_name);
661 		return true;
662 	}
663 	return false;
664 }
665 
btrfs_check_options(const struct btrfs_fs_info * info,unsigned long long * mount_opt,unsigned long flags)666 bool btrfs_check_options(const struct btrfs_fs_info *info,
667 			 unsigned long long *mount_opt,
668 			 unsigned long flags)
669 {
670 	bool ret = true;
671 
672 	if (!(flags & SB_RDONLY) &&
673 	    (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") ||
674 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") ||
675 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") ||
676 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") ||
677 	     check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags")))
678 		ret = false;
679 
680 	if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) &&
681 	    !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) &&
682 	    !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) {
683 		btrfs_err(info, "cannot disable free-space-tree");
684 		ret = false;
685 	}
686 	if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) &&
687 	     !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) {
688 		btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature");
689 		ret = false;
690 	}
691 
692 	if (btrfs_check_mountopts_zoned(info, mount_opt))
693 		ret = false;
694 
695 	if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) {
696 		if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) {
697 			btrfs_info(info, "disk space caching is enabled");
698 			btrfs_warn(info,
699 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2");
700 		}
701 		if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE))
702 			btrfs_info(info, "using free-space-tree");
703 	}
704 
705 	return ret;
706 }
707 
708 /*
709  * This is subtle, we only call this during open_ctree().  We need to pre-load
710  * the mount options with the on-disk settings.  Before the new mount API took
711  * effect we would do this on mount and remount.  With the new mount API we'll
712  * only do this on the initial mount.
713  *
714  * This isn't a change in behavior, because we're using the current state of the
715  * file system to set the current mount options.  If you mounted with special
716  * options to disable these features and then remounted we wouldn't revert the
717  * settings, because mounting without these features cleared the on-disk
718  * settings, so this being called on re-mount is not needed.
719  */
btrfs_set_free_space_cache_settings(struct btrfs_fs_info * fs_info)720 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info)
721 {
722 	if (fs_info->sectorsize < PAGE_SIZE) {
723 		btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
724 		if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) {
725 			btrfs_info(fs_info,
726 				   "forcing free space tree for sector size %u with page size %lu",
727 				   fs_info->sectorsize, PAGE_SIZE);
728 			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
729 		}
730 	}
731 
732 	/*
733 	 * At this point our mount options are populated, so we only mess with
734 	 * these settings if we don't have any settings already.
735 	 */
736 	if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
737 		return;
738 
739 	if (btrfs_is_zoned(fs_info) &&
740 	    btrfs_free_space_cache_v1_active(fs_info)) {
741 		btrfs_info(fs_info, "zoned: clearing existing space cache");
742 		btrfs_set_super_cache_generation(fs_info->super_copy, 0);
743 		return;
744 	}
745 
746 	if (btrfs_test_opt(fs_info, SPACE_CACHE))
747 		return;
748 
749 	if (btrfs_test_opt(fs_info, NOSPACECACHE))
750 		return;
751 
752 	/*
753 	 * At this point we don't have explicit options set by the user, set
754 	 * them ourselves based on the state of the file system.
755 	 */
756 	if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
757 		btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
758 	else if (btrfs_free_space_cache_v1_active(fs_info))
759 		btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
760 }
761 
set_device_specific_options(struct btrfs_fs_info * fs_info)762 static void set_device_specific_options(struct btrfs_fs_info *fs_info)
763 {
764 	if (!btrfs_test_opt(fs_info, NOSSD) &&
765 	    !fs_info->fs_devices->rotating)
766 		btrfs_set_opt(fs_info->mount_opt, SSD);
767 
768 	/*
769 	 * For devices supporting discard turn on discard=async automatically,
770 	 * unless it's already set or disabled. This could be turned off by
771 	 * nodiscard for the same mount.
772 	 *
773 	 * The zoned mode piggy backs on the discard functionality for
774 	 * resetting a zone. There is no reason to delay the zone reset as it is
775 	 * fast enough. So, do not enable async discard for zoned mode.
776 	 */
777 	if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) ||
778 	      btrfs_test_opt(fs_info, DISCARD_ASYNC) ||
779 	      btrfs_test_opt(fs_info, NODISCARD)) &&
780 	    fs_info->fs_devices->discardable &&
781 	    !btrfs_is_zoned(fs_info))
782 		btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC);
783 }
784 
btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info * fs_info,u64 subvol_objectid)785 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info,
786 					  u64 subvol_objectid)
787 {
788 	struct btrfs_root *root = fs_info->tree_root;
789 	struct btrfs_root *fs_root = NULL;
790 	struct btrfs_root_ref *root_ref;
791 	struct btrfs_inode_ref *inode_ref;
792 	struct btrfs_key key;
793 	struct btrfs_path *path = NULL;
794 	char *name = NULL, *ptr;
795 	u64 dirid;
796 	int len;
797 	int ret;
798 
799 	path = btrfs_alloc_path();
800 	if (!path) {
801 		ret = -ENOMEM;
802 		goto err;
803 	}
804 
805 	name = kmalloc(PATH_MAX, GFP_KERNEL);
806 	if (!name) {
807 		ret = -ENOMEM;
808 		goto err;
809 	}
810 	ptr = name + PATH_MAX - 1;
811 	ptr[0] = '\0';
812 
813 	/*
814 	 * Walk up the subvolume trees in the tree of tree roots by root
815 	 * backrefs until we hit the top-level subvolume.
816 	 */
817 	while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) {
818 		key.objectid = subvol_objectid;
819 		key.type = BTRFS_ROOT_BACKREF_KEY;
820 		key.offset = (u64)-1;
821 
822 		ret = btrfs_search_backwards(root, &key, path);
823 		if (ret < 0) {
824 			goto err;
825 		} else if (ret > 0) {
826 			ret = -ENOENT;
827 			goto err;
828 		}
829 
830 		subvol_objectid = key.offset;
831 
832 		root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
833 					  struct btrfs_root_ref);
834 		len = btrfs_root_ref_name_len(path->nodes[0], root_ref);
835 		ptr -= len + 1;
836 		if (ptr < name) {
837 			ret = -ENAMETOOLONG;
838 			goto err;
839 		}
840 		read_extent_buffer(path->nodes[0], ptr + 1,
841 				   (unsigned long)(root_ref + 1), len);
842 		ptr[0] = '/';
843 		dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref);
844 		btrfs_release_path(path);
845 
846 		fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true);
847 		if (IS_ERR(fs_root)) {
848 			ret = PTR_ERR(fs_root);
849 			fs_root = NULL;
850 			goto err;
851 		}
852 
853 		/*
854 		 * Walk up the filesystem tree by inode refs until we hit the
855 		 * root directory.
856 		 */
857 		while (dirid != BTRFS_FIRST_FREE_OBJECTID) {
858 			key.objectid = dirid;
859 			key.type = BTRFS_INODE_REF_KEY;
860 			key.offset = (u64)-1;
861 
862 			ret = btrfs_search_backwards(fs_root, &key, path);
863 			if (ret < 0) {
864 				goto err;
865 			} else if (ret > 0) {
866 				ret = -ENOENT;
867 				goto err;
868 			}
869 
870 			dirid = key.offset;
871 
872 			inode_ref = btrfs_item_ptr(path->nodes[0],
873 						   path->slots[0],
874 						   struct btrfs_inode_ref);
875 			len = btrfs_inode_ref_name_len(path->nodes[0],
876 						       inode_ref);
877 			ptr -= len + 1;
878 			if (ptr < name) {
879 				ret = -ENAMETOOLONG;
880 				goto err;
881 			}
882 			read_extent_buffer(path->nodes[0], ptr + 1,
883 					   (unsigned long)(inode_ref + 1), len);
884 			ptr[0] = '/';
885 			btrfs_release_path(path);
886 		}
887 		btrfs_put_root(fs_root);
888 		fs_root = NULL;
889 	}
890 
891 	btrfs_free_path(path);
892 	if (ptr == name + PATH_MAX - 1) {
893 		name[0] = '/';
894 		name[1] = '\0';
895 	} else {
896 		memmove(name, ptr, name + PATH_MAX - ptr);
897 	}
898 	return name;
899 
900 err:
901 	btrfs_put_root(fs_root);
902 	btrfs_free_path(path);
903 	kfree(name);
904 	return ERR_PTR(ret);
905 }
906 
get_default_subvol_objectid(struct btrfs_fs_info * fs_info,u64 * objectid)907 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid)
908 {
909 	struct btrfs_root *root = fs_info->tree_root;
910 	struct btrfs_dir_item *di;
911 	struct btrfs_path *path;
912 	struct btrfs_key location;
913 	struct fscrypt_str name = FSTR_INIT("default", 7);
914 	u64 dir_id;
915 
916 	path = btrfs_alloc_path();
917 	if (!path)
918 		return -ENOMEM;
919 
920 	/*
921 	 * Find the "default" dir item which points to the root item that we
922 	 * will mount by default if we haven't been given a specific subvolume
923 	 * to mount.
924 	 */
925 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
926 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0);
927 	if (IS_ERR(di)) {
928 		btrfs_free_path(path);
929 		return PTR_ERR(di);
930 	}
931 	if (!di) {
932 		/*
933 		 * Ok the default dir item isn't there.  This is weird since
934 		 * it's always been there, but don't freak out, just try and
935 		 * mount the top-level subvolume.
936 		 */
937 		btrfs_free_path(path);
938 		*objectid = BTRFS_FS_TREE_OBJECTID;
939 		return 0;
940 	}
941 
942 	btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
943 	btrfs_free_path(path);
944 	*objectid = location.objectid;
945 	return 0;
946 }
947 
btrfs_fill_super(struct super_block * sb,struct btrfs_fs_devices * fs_devices,void * data)948 static int btrfs_fill_super(struct super_block *sb,
949 			    struct btrfs_fs_devices *fs_devices,
950 			    void *data)
951 {
952 	struct inode *inode;
953 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
954 	int err;
955 
956 	sb->s_maxbytes = MAX_LFS_FILESIZE;
957 	sb->s_magic = BTRFS_SUPER_MAGIC;
958 	sb->s_op = &btrfs_super_ops;
959 	sb->s_d_op = &btrfs_dentry_operations;
960 	sb->s_export_op = &btrfs_export_ops;
961 #ifdef CONFIG_FS_VERITY
962 	sb->s_vop = &btrfs_verityops;
963 #endif
964 	sb->s_xattr = btrfs_xattr_handlers;
965 	sb->s_time_gran = 1;
966 	sb->s_iflags |= SB_I_CGROUPWB;
967 
968 	err = super_setup_bdi(sb);
969 	if (err) {
970 		btrfs_err(fs_info, "super_setup_bdi failed");
971 		return err;
972 	}
973 
974 	err = open_ctree(sb, fs_devices, (char *)data);
975 	if (err) {
976 		btrfs_err(fs_info, "open_ctree failed");
977 		return err;
978 	}
979 
980 	inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root);
981 	if (IS_ERR(inode)) {
982 		err = PTR_ERR(inode);
983 		btrfs_handle_fs_error(fs_info, err, NULL);
984 		goto fail_close;
985 	}
986 
987 	sb->s_root = d_make_root(inode);
988 	if (!sb->s_root) {
989 		err = -ENOMEM;
990 		goto fail_close;
991 	}
992 
993 	sb->s_flags |= SB_ACTIVE;
994 	return 0;
995 
996 fail_close:
997 	close_ctree(fs_info);
998 	return err;
999 }
1000 
btrfs_sync_fs(struct super_block * sb,int wait)1001 int btrfs_sync_fs(struct super_block *sb, int wait)
1002 {
1003 	struct btrfs_trans_handle *trans;
1004 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1005 	struct btrfs_root *root = fs_info->tree_root;
1006 
1007 	trace_btrfs_sync_fs(fs_info, wait);
1008 
1009 	if (!wait) {
1010 		filemap_flush(fs_info->btree_inode->i_mapping);
1011 		return 0;
1012 	}
1013 
1014 	btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL);
1015 
1016 	trans = btrfs_attach_transaction_barrier(root);
1017 	if (IS_ERR(trans)) {
1018 		/* no transaction, don't bother */
1019 		if (PTR_ERR(trans) == -ENOENT) {
1020 			/*
1021 			 * Exit unless we have some pending changes
1022 			 * that need to go through commit
1023 			 */
1024 			if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT,
1025 				      &fs_info->flags))
1026 				return 0;
1027 			/*
1028 			 * A non-blocking test if the fs is frozen. We must not
1029 			 * start a new transaction here otherwise a deadlock
1030 			 * happens. The pending operations are delayed to the
1031 			 * next commit after thawing.
1032 			 */
1033 			if (sb_start_write_trylock(sb))
1034 				sb_end_write(sb);
1035 			else
1036 				return 0;
1037 			trans = btrfs_start_transaction(root, 0);
1038 		}
1039 		if (IS_ERR(trans))
1040 			return PTR_ERR(trans);
1041 	}
1042 	return btrfs_commit_transaction(trans);
1043 }
1044 
print_rescue_option(struct seq_file * seq,const char * s,bool * printed)1045 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed)
1046 {
1047 	seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s);
1048 	*printed = true;
1049 }
1050 
btrfs_show_options(struct seq_file * seq,struct dentry * dentry)1051 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry)
1052 {
1053 	struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb);
1054 	const char *compress_type;
1055 	const char *subvol_name;
1056 	bool printed = false;
1057 
1058 	if (btrfs_test_opt(info, DEGRADED))
1059 		seq_puts(seq, ",degraded");
1060 	if (btrfs_test_opt(info, NODATASUM))
1061 		seq_puts(seq, ",nodatasum");
1062 	if (btrfs_test_opt(info, NODATACOW))
1063 		seq_puts(seq, ",nodatacow");
1064 	if (btrfs_test_opt(info, NOBARRIER))
1065 		seq_puts(seq, ",nobarrier");
1066 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1067 		seq_printf(seq, ",max_inline=%llu", info->max_inline);
1068 	if (info->thread_pool_size !=  min_t(unsigned long,
1069 					     num_online_cpus() + 2, 8))
1070 		seq_printf(seq, ",thread_pool=%u", info->thread_pool_size);
1071 	if (btrfs_test_opt(info, COMPRESS)) {
1072 		compress_type = btrfs_compress_type2str(info->compress_type);
1073 		if (btrfs_test_opt(info, FORCE_COMPRESS))
1074 			seq_printf(seq, ",compress-force=%s", compress_type);
1075 		else
1076 			seq_printf(seq, ",compress=%s", compress_type);
1077 		if (info->compress_level)
1078 			seq_printf(seq, ":%d", info->compress_level);
1079 	}
1080 	if (btrfs_test_opt(info, NOSSD))
1081 		seq_puts(seq, ",nossd");
1082 	if (btrfs_test_opt(info, SSD_SPREAD))
1083 		seq_puts(seq, ",ssd_spread");
1084 	else if (btrfs_test_opt(info, SSD))
1085 		seq_puts(seq, ",ssd");
1086 	if (btrfs_test_opt(info, NOTREELOG))
1087 		seq_puts(seq, ",notreelog");
1088 	if (btrfs_test_opt(info, NOLOGREPLAY))
1089 		print_rescue_option(seq, "nologreplay", &printed);
1090 	if (btrfs_test_opt(info, USEBACKUPROOT))
1091 		print_rescue_option(seq, "usebackuproot", &printed);
1092 	if (btrfs_test_opt(info, IGNOREBADROOTS))
1093 		print_rescue_option(seq, "ignorebadroots", &printed);
1094 	if (btrfs_test_opt(info, IGNOREDATACSUMS))
1095 		print_rescue_option(seq, "ignoredatacsums", &printed);
1096 	if (btrfs_test_opt(info, IGNOREMETACSUMS))
1097 		print_rescue_option(seq, "ignoremetacsums", &printed);
1098 	if (btrfs_test_opt(info, IGNORESUPERFLAGS))
1099 		print_rescue_option(seq, "ignoresuperflags", &printed);
1100 	if (btrfs_test_opt(info, FLUSHONCOMMIT))
1101 		seq_puts(seq, ",flushoncommit");
1102 	if (btrfs_test_opt(info, DISCARD_SYNC))
1103 		seq_puts(seq, ",discard");
1104 	if (btrfs_test_opt(info, DISCARD_ASYNC))
1105 		seq_puts(seq, ",discard=async");
1106 	if (!(info->sb->s_flags & SB_POSIXACL))
1107 		seq_puts(seq, ",noacl");
1108 	if (btrfs_free_space_cache_v1_active(info))
1109 		seq_puts(seq, ",space_cache");
1110 	else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE))
1111 		seq_puts(seq, ",space_cache=v2");
1112 	else
1113 		seq_puts(seq, ",nospace_cache");
1114 	if (btrfs_test_opt(info, RESCAN_UUID_TREE))
1115 		seq_puts(seq, ",rescan_uuid_tree");
1116 	if (btrfs_test_opt(info, CLEAR_CACHE))
1117 		seq_puts(seq, ",clear_cache");
1118 	if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED))
1119 		seq_puts(seq, ",user_subvol_rm_allowed");
1120 	if (btrfs_test_opt(info, ENOSPC_DEBUG))
1121 		seq_puts(seq, ",enospc_debug");
1122 	if (btrfs_test_opt(info, AUTO_DEFRAG))
1123 		seq_puts(seq, ",autodefrag");
1124 	if (btrfs_test_opt(info, SKIP_BALANCE))
1125 		seq_puts(seq, ",skip_balance");
1126 	if (info->metadata_ratio)
1127 		seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio);
1128 	if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR))
1129 		seq_puts(seq, ",fatal_errors=panic");
1130 	if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL)
1131 		seq_printf(seq, ",commit=%u", info->commit_interval);
1132 #ifdef CONFIG_BTRFS_DEBUG
1133 	if (btrfs_test_opt(info, FRAGMENT_DATA))
1134 		seq_puts(seq, ",fragment=data");
1135 	if (btrfs_test_opt(info, FRAGMENT_METADATA))
1136 		seq_puts(seq, ",fragment=metadata");
1137 #endif
1138 	if (btrfs_test_opt(info, REF_VERIFY))
1139 		seq_puts(seq, ",ref_verify");
1140 	seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1141 	subvol_name = btrfs_get_subvol_name_from_objectid(info,
1142 			btrfs_root_id(BTRFS_I(d_inode(dentry))->root));
1143 	if (!IS_ERR(subvol_name)) {
1144 		seq_puts(seq, ",subvol=");
1145 		seq_escape(seq, subvol_name, " \t\n\\");
1146 		kfree(subvol_name);
1147 	}
1148 	return 0;
1149 }
1150 
1151 /*
1152  * subvolumes are identified by ino 256
1153  */
is_subvolume_inode(struct inode * inode)1154 static inline int is_subvolume_inode(struct inode *inode)
1155 {
1156 	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
1157 		return 1;
1158 	return 0;
1159 }
1160 
mount_subvol(const char * subvol_name,u64 subvol_objectid,struct vfsmount * mnt)1161 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid,
1162 				   struct vfsmount *mnt)
1163 {
1164 	struct dentry *root;
1165 	int ret;
1166 
1167 	if (!subvol_name) {
1168 		if (!subvol_objectid) {
1169 			ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb),
1170 							  &subvol_objectid);
1171 			if (ret) {
1172 				root = ERR_PTR(ret);
1173 				goto out;
1174 			}
1175 		}
1176 		subvol_name = btrfs_get_subvol_name_from_objectid(
1177 					btrfs_sb(mnt->mnt_sb), subvol_objectid);
1178 		if (IS_ERR(subvol_name)) {
1179 			root = ERR_CAST(subvol_name);
1180 			subvol_name = NULL;
1181 			goto out;
1182 		}
1183 
1184 	}
1185 
1186 	root = mount_subtree(mnt, subvol_name);
1187 	/* mount_subtree() drops our reference on the vfsmount. */
1188 	mnt = NULL;
1189 
1190 	if (!IS_ERR(root)) {
1191 		struct super_block *s = root->d_sb;
1192 		struct btrfs_fs_info *fs_info = btrfs_sb(s);
1193 		struct inode *root_inode = d_inode(root);
1194 		u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root);
1195 
1196 		ret = 0;
1197 		if (!is_subvolume_inode(root_inode)) {
1198 			btrfs_err(fs_info, "'%s' is not a valid subvolume",
1199 			       subvol_name);
1200 			ret = -EINVAL;
1201 		}
1202 		if (subvol_objectid && root_objectid != subvol_objectid) {
1203 			/*
1204 			 * This will also catch a race condition where a
1205 			 * subvolume which was passed by ID is renamed and
1206 			 * another subvolume is renamed over the old location.
1207 			 */
1208 			btrfs_err(fs_info,
1209 				  "subvol '%s' does not match subvolid %llu",
1210 				  subvol_name, subvol_objectid);
1211 			ret = -EINVAL;
1212 		}
1213 		if (ret) {
1214 			dput(root);
1215 			root = ERR_PTR(ret);
1216 			deactivate_locked_super(s);
1217 		}
1218 	}
1219 
1220 out:
1221 	mntput(mnt);
1222 	kfree(subvol_name);
1223 	return root;
1224 }
1225 
btrfs_resize_thread_pool(struct btrfs_fs_info * fs_info,u32 new_pool_size,u32 old_pool_size)1226 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info,
1227 				     u32 new_pool_size, u32 old_pool_size)
1228 {
1229 	if (new_pool_size == old_pool_size)
1230 		return;
1231 
1232 	fs_info->thread_pool_size = new_pool_size;
1233 
1234 	btrfs_info(fs_info, "resize thread pool %d -> %d",
1235 	       old_pool_size, new_pool_size);
1236 
1237 	btrfs_workqueue_set_max(fs_info->workers, new_pool_size);
1238 	btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size);
1239 	btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size);
1240 	workqueue_set_max_active(fs_info->endio_workers, new_pool_size);
1241 	workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size);
1242 	btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size);
1243 	btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size);
1244 	btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size);
1245 }
1246 
btrfs_remount_begin(struct btrfs_fs_info * fs_info,unsigned long long old_opts,int flags)1247 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info,
1248 				       unsigned long long old_opts, int flags)
1249 {
1250 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1251 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) ||
1252 	     (flags & SB_RDONLY))) {
1253 		/* wait for any defraggers to finish */
1254 		wait_event(fs_info->transaction_wait,
1255 			   (atomic_read(&fs_info->defrag_running) == 0));
1256 		if (flags & SB_RDONLY)
1257 			sync_filesystem(fs_info->sb);
1258 	}
1259 }
1260 
btrfs_remount_cleanup(struct btrfs_fs_info * fs_info,unsigned long long old_opts)1261 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info,
1262 					 unsigned long long old_opts)
1263 {
1264 	const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE);
1265 
1266 	/*
1267 	 * We need to cleanup all defragable inodes if the autodefragment is
1268 	 * close or the filesystem is read only.
1269 	 */
1270 	if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) &&
1271 	    (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) {
1272 		btrfs_cleanup_defrag_inodes(fs_info);
1273 	}
1274 
1275 	/* If we toggled discard async */
1276 	if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1277 	    btrfs_test_opt(fs_info, DISCARD_ASYNC))
1278 		btrfs_discard_resume(fs_info);
1279 	else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) &&
1280 		 !btrfs_test_opt(fs_info, DISCARD_ASYNC))
1281 		btrfs_discard_cleanup(fs_info);
1282 
1283 	/* If we toggled space cache */
1284 	if (cache_opt != btrfs_free_space_cache_v1_active(fs_info))
1285 		btrfs_set_free_space_cache_v1_active(fs_info, cache_opt);
1286 }
1287 
btrfs_remount_rw(struct btrfs_fs_info * fs_info)1288 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info)
1289 {
1290 	int ret;
1291 
1292 	if (BTRFS_FS_ERROR(fs_info)) {
1293 		btrfs_err(fs_info,
1294 			  "remounting read-write after error is not allowed");
1295 		return -EINVAL;
1296 	}
1297 
1298 	if (fs_info->fs_devices->rw_devices == 0)
1299 		return -EACCES;
1300 
1301 	if (!btrfs_check_rw_degradable(fs_info, NULL)) {
1302 		btrfs_warn(fs_info,
1303 			   "too many missing devices, writable remount is not allowed");
1304 		return -EACCES;
1305 	}
1306 
1307 	if (btrfs_super_log_root(fs_info->super_copy) != 0) {
1308 		btrfs_warn(fs_info,
1309 			   "mount required to replay tree-log, cannot remount read-write");
1310 		return -EINVAL;
1311 	}
1312 
1313 	/*
1314 	 * NOTE: when remounting with a change that does writes, don't put it
1315 	 * anywhere above this point, as we are not sure to be safe to write
1316 	 * until we pass the above checks.
1317 	 */
1318 	ret = btrfs_start_pre_rw_mount(fs_info);
1319 	if (ret)
1320 		return ret;
1321 
1322 	btrfs_clear_sb_rdonly(fs_info->sb);
1323 
1324 	set_bit(BTRFS_FS_OPEN, &fs_info->flags);
1325 
1326 	/*
1327 	 * If we've gone from readonly -> read-write, we need to get our
1328 	 * sync/async discard lists in the right state.
1329 	 */
1330 	btrfs_discard_resume(fs_info);
1331 
1332 	return 0;
1333 }
1334 
btrfs_remount_ro(struct btrfs_fs_info * fs_info)1335 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info)
1336 {
1337 	/*
1338 	 * This also happens on 'umount -rf' or on shutdown, when the
1339 	 * filesystem is busy.
1340 	 */
1341 	cancel_work_sync(&fs_info->async_reclaim_work);
1342 	cancel_work_sync(&fs_info->async_data_reclaim_work);
1343 
1344 	btrfs_discard_cleanup(fs_info);
1345 
1346 	/* Wait for the uuid_scan task to finish */
1347 	down(&fs_info->uuid_tree_rescan_sem);
1348 	/* Avoid complains from lockdep et al. */
1349 	up(&fs_info->uuid_tree_rescan_sem);
1350 
1351 	btrfs_set_sb_rdonly(fs_info->sb);
1352 
1353 	/*
1354 	 * Setting SB_RDONLY will put the cleaner thread to sleep at the next
1355 	 * loop if it's already active.  If it's already asleep, we'll leave
1356 	 * unused block groups on disk until we're mounted read-write again
1357 	 * unless we clean them up here.
1358 	 */
1359 	btrfs_delete_unused_bgs(fs_info);
1360 
1361 	/*
1362 	 * The cleaner task could be already running before we set the flag
1363 	 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock).  We must make
1364 	 * sure that after we finish the remount, i.e. after we call
1365 	 * btrfs_commit_super(), the cleaner can no longer start a transaction
1366 	 * - either because it was dropping a dead root, running delayed iputs
1367 	 *   or deleting an unused block group (the cleaner picked a block
1368 	 *   group from the list of unused block groups before we were able to
1369 	 *   in the previous call to btrfs_delete_unused_bgs()).
1370 	 */
1371 	wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE);
1372 
1373 	/*
1374 	 * We've set the superblock to RO mode, so we might have made the
1375 	 * cleaner task sleep without running all pending delayed iputs. Go
1376 	 * through all the delayed iputs here, so that if an unmount happens
1377 	 * without remounting RW we don't end up at finishing close_ctree()
1378 	 * with a non-empty list of delayed iputs.
1379 	 */
1380 	btrfs_run_delayed_iputs(fs_info);
1381 
1382 	btrfs_dev_replace_suspend_for_unmount(fs_info);
1383 	btrfs_scrub_cancel(fs_info);
1384 	btrfs_pause_balance(fs_info);
1385 
1386 	/*
1387 	 * Pause the qgroup rescan worker if it is running. We don't want it to
1388 	 * be still running after we are in RO mode, as after that, by the time
1389 	 * we unmount, it might have left a transaction open, so we would leak
1390 	 * the transaction and/or crash.
1391 	 */
1392 	btrfs_qgroup_wait_for_completion(fs_info, false);
1393 
1394 	return btrfs_commit_super(fs_info);
1395 }
1396 
btrfs_ctx_to_info(struct btrfs_fs_info * fs_info,struct btrfs_fs_context * ctx)1397 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1398 {
1399 	fs_info->max_inline = ctx->max_inline;
1400 	fs_info->commit_interval = ctx->commit_interval;
1401 	fs_info->metadata_ratio = ctx->metadata_ratio;
1402 	fs_info->thread_pool_size = ctx->thread_pool_size;
1403 	fs_info->mount_opt = ctx->mount_opt;
1404 	fs_info->compress_type = ctx->compress_type;
1405 	fs_info->compress_level = ctx->compress_level;
1406 }
1407 
btrfs_info_to_ctx(struct btrfs_fs_info * fs_info,struct btrfs_fs_context * ctx)1408 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx)
1409 {
1410 	ctx->max_inline = fs_info->max_inline;
1411 	ctx->commit_interval = fs_info->commit_interval;
1412 	ctx->metadata_ratio = fs_info->metadata_ratio;
1413 	ctx->thread_pool_size = fs_info->thread_pool_size;
1414 	ctx->mount_opt = fs_info->mount_opt;
1415 	ctx->compress_type = fs_info->compress_type;
1416 	ctx->compress_level = fs_info->compress_level;
1417 }
1418 
1419 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...)			\
1420 do {										\
1421 	if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1422 	    btrfs_raw_test_opt(fs_info->mount_opt, opt))			\
1423 		btrfs_info(fs_info, fmt, ##args);				\
1424 } while (0)
1425 
1426 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...)	\
1427 do {									\
1428 	if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) &&	\
1429 	    !btrfs_raw_test_opt(fs_info->mount_opt, opt))		\
1430 		btrfs_info(fs_info, fmt, ##args);			\
1431 } while (0)
1432 
btrfs_emit_options(struct btrfs_fs_info * info,struct btrfs_fs_context * old)1433 static void btrfs_emit_options(struct btrfs_fs_info *info,
1434 			       struct btrfs_fs_context *old)
1435 {
1436 	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1437 	btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts");
1438 	btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum");
1439 	btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations");
1440 	btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme");
1441 	btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers");
1442 	btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log");
1443 	btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time");
1444 	btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit");
1445 	btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard");
1446 	btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard");
1447 	btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree");
1448 	btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching");
1449 	btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache");
1450 	btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag");
1451 	btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data");
1452 	btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata");
1453 	btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification");
1454 	btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time");
1455 	btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots");
1456 	btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums");
1457 	btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums");
1458 	btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags");
1459 
1460 	btrfs_info_if_unset(info, old, NODATACOW, "setting datacow");
1461 	btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations");
1462 	btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme");
1463 	btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers");
1464 	btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log");
1465 	btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching");
1466 	btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree");
1467 	btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag");
1468 	btrfs_info_if_unset(info, old, COMPRESS, "use no compression");
1469 
1470 	/* Did the compression settings change? */
1471 	if (btrfs_test_opt(info, COMPRESS) &&
1472 	    (!old ||
1473 	     old->compress_type != info->compress_type ||
1474 	     old->compress_level != info->compress_level ||
1475 	     (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) &&
1476 	      btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) {
1477 		const char *compress_type = btrfs_compress_type2str(info->compress_type);
1478 
1479 		btrfs_info(info, "%s %s compression, level %d",
1480 			   btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use",
1481 			   compress_type, info->compress_level);
1482 	}
1483 
1484 	if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE)
1485 		btrfs_info(info, "max_inline set to %llu", info->max_inline);
1486 }
1487 
btrfs_reconfigure(struct fs_context * fc)1488 static int btrfs_reconfigure(struct fs_context *fc)
1489 {
1490 	struct super_block *sb = fc->root->d_sb;
1491 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1492 	struct btrfs_fs_context *ctx = fc->fs_private;
1493 	struct btrfs_fs_context old_ctx;
1494 	int ret = 0;
1495 	bool mount_reconfigure = (fc->s_fs_info != NULL);
1496 
1497 	btrfs_info_to_ctx(fs_info, &old_ctx);
1498 
1499 	/*
1500 	 * This is our "bind mount" trick, we don't want to allow the user to do
1501 	 * anything other than mount a different ro/rw and a different subvol,
1502 	 * all of the mount options should be maintained.
1503 	 */
1504 	if (mount_reconfigure)
1505 		ctx->mount_opt = old_ctx.mount_opt;
1506 
1507 	sync_filesystem(sb);
1508 	set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1509 
1510 	if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags))
1511 		return -EINVAL;
1512 
1513 	ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY));
1514 	if (ret < 0)
1515 		return ret;
1516 
1517 	btrfs_ctx_to_info(fs_info, ctx);
1518 	btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags);
1519 	btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size,
1520 				 old_ctx.thread_pool_size);
1521 
1522 	if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) !=
1523 	    (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) &&
1524 	    (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) {
1525 		btrfs_warn(fs_info,
1526 		"remount supports changing free space tree only from RO to RW");
1527 		/* Make sure free space cache options match the state on disk. */
1528 		if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) {
1529 			btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1530 			btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE);
1531 		}
1532 		if (btrfs_free_space_cache_v1_active(fs_info)) {
1533 			btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE);
1534 			btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE);
1535 		}
1536 	}
1537 
1538 	ret = 0;
1539 	if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY))
1540 		ret = btrfs_remount_ro(fs_info);
1541 	else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY))
1542 		ret = btrfs_remount_rw(fs_info);
1543 	if (ret)
1544 		goto restore;
1545 
1546 	/*
1547 	 * If we set the mask during the parameter parsing VFS would reject the
1548 	 * remount.  Here we can set the mask and the value will be updated
1549 	 * appropriately.
1550 	 */
1551 	if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL))
1552 		fc->sb_flags_mask |= SB_POSIXACL;
1553 
1554 	btrfs_emit_options(fs_info, &old_ctx);
1555 	wake_up_process(fs_info->transaction_kthread);
1556 	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1557 	btrfs_clear_oneshot_options(fs_info);
1558 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1559 
1560 	return 0;
1561 restore:
1562 	btrfs_ctx_to_info(fs_info, &old_ctx);
1563 	btrfs_remount_cleanup(fs_info, old_ctx.mount_opt);
1564 	clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state);
1565 	return ret;
1566 }
1567 
1568 /* Used to sort the devices by max_avail(descending sort) */
btrfs_cmp_device_free_bytes(const void * a,const void * b)1569 static int btrfs_cmp_device_free_bytes(const void *a, const void *b)
1570 {
1571 	const struct btrfs_device_info *dev_info1 = a;
1572 	const struct btrfs_device_info *dev_info2 = b;
1573 
1574 	if (dev_info1->max_avail > dev_info2->max_avail)
1575 		return -1;
1576 	else if (dev_info1->max_avail < dev_info2->max_avail)
1577 		return 1;
1578 	return 0;
1579 }
1580 
1581 /*
1582  * sort the devices by max_avail, in which max free extent size of each device
1583  * is stored.(Descending Sort)
1584  */
btrfs_descending_sort_devices(struct btrfs_device_info * devices,size_t nr_devices)1585 static inline void btrfs_descending_sort_devices(
1586 					struct btrfs_device_info *devices,
1587 					size_t nr_devices)
1588 {
1589 	sort(devices, nr_devices, sizeof(struct btrfs_device_info),
1590 	     btrfs_cmp_device_free_bytes, NULL);
1591 }
1592 
1593 /*
1594  * The helper to calc the free space on the devices that can be used to store
1595  * file data.
1596  */
btrfs_calc_avail_data_space(struct btrfs_fs_info * fs_info,u64 * free_bytes)1597 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info,
1598 					      u64 *free_bytes)
1599 {
1600 	struct btrfs_device_info *devices_info;
1601 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
1602 	struct btrfs_device *device;
1603 	u64 type;
1604 	u64 avail_space;
1605 	u64 min_stripe_size;
1606 	int num_stripes = 1;
1607 	int i = 0, nr_devices;
1608 	const struct btrfs_raid_attr *rattr;
1609 
1610 	/*
1611 	 * We aren't under the device list lock, so this is racy-ish, but good
1612 	 * enough for our purposes.
1613 	 */
1614 	nr_devices = fs_info->fs_devices->open_devices;
1615 	if (!nr_devices) {
1616 		smp_mb();
1617 		nr_devices = fs_info->fs_devices->open_devices;
1618 		ASSERT(nr_devices);
1619 		if (!nr_devices) {
1620 			*free_bytes = 0;
1621 			return 0;
1622 		}
1623 	}
1624 
1625 	devices_info = kmalloc_array(nr_devices, sizeof(*devices_info),
1626 			       GFP_KERNEL);
1627 	if (!devices_info)
1628 		return -ENOMEM;
1629 
1630 	/* calc min stripe number for data space allocation */
1631 	type = btrfs_data_alloc_profile(fs_info);
1632 	rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)];
1633 
1634 	if (type & BTRFS_BLOCK_GROUP_RAID0)
1635 		num_stripes = nr_devices;
1636 	else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK)
1637 		num_stripes = rattr->ncopies;
1638 	else if (type & BTRFS_BLOCK_GROUP_RAID10)
1639 		num_stripes = 4;
1640 
1641 	/* Adjust for more than 1 stripe per device */
1642 	min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN;
1643 
1644 	rcu_read_lock();
1645 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
1646 		if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA,
1647 						&device->dev_state) ||
1648 		    !device->bdev ||
1649 		    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state))
1650 			continue;
1651 
1652 		if (i >= nr_devices)
1653 			break;
1654 
1655 		avail_space = device->total_bytes - device->bytes_used;
1656 
1657 		/* align with stripe_len */
1658 		avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN);
1659 
1660 		/*
1661 		 * Ensure we have at least min_stripe_size on top of the
1662 		 * reserved space on the device.
1663 		 */
1664 		if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size)
1665 			continue;
1666 
1667 		avail_space -= BTRFS_DEVICE_RANGE_RESERVED;
1668 
1669 		devices_info[i].dev = device;
1670 		devices_info[i].max_avail = avail_space;
1671 
1672 		i++;
1673 	}
1674 	rcu_read_unlock();
1675 
1676 	nr_devices = i;
1677 
1678 	btrfs_descending_sort_devices(devices_info, nr_devices);
1679 
1680 	i = nr_devices - 1;
1681 	avail_space = 0;
1682 	while (nr_devices >= rattr->devs_min) {
1683 		num_stripes = min(num_stripes, nr_devices);
1684 
1685 		if (devices_info[i].max_avail >= min_stripe_size) {
1686 			int j;
1687 			u64 alloc_size;
1688 
1689 			avail_space += devices_info[i].max_avail * num_stripes;
1690 			alloc_size = devices_info[i].max_avail;
1691 			for (j = i + 1 - num_stripes; j <= i; j++)
1692 				devices_info[j].max_avail -= alloc_size;
1693 		}
1694 		i--;
1695 		nr_devices--;
1696 	}
1697 
1698 	kfree(devices_info);
1699 	*free_bytes = avail_space;
1700 	return 0;
1701 }
1702 
1703 /*
1704  * Calculate numbers for 'df', pessimistic in case of mixed raid profiles.
1705  *
1706  * If there's a redundant raid level at DATA block groups, use the respective
1707  * multiplier to scale the sizes.
1708  *
1709  * Unused device space usage is based on simulating the chunk allocator
1710  * algorithm that respects the device sizes and order of allocations.  This is
1711  * a close approximation of the actual use but there are other factors that may
1712  * change the result (like a new metadata chunk).
1713  *
1714  * If metadata is exhausted, f_bavail will be 0.
1715  */
btrfs_statfs(struct dentry * dentry,struct kstatfs * buf)1716 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
1717 {
1718 	struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb);
1719 	struct btrfs_super_block *disk_super = fs_info->super_copy;
1720 	struct btrfs_space_info *found;
1721 	u64 total_used = 0;
1722 	u64 total_free_data = 0;
1723 	u64 total_free_meta = 0;
1724 	u32 bits = fs_info->sectorsize_bits;
1725 	__be32 *fsid = (__be32 *)fs_info->fs_devices->fsid;
1726 	unsigned factor = 1;
1727 	struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
1728 	int ret;
1729 	u64 thresh = 0;
1730 	int mixed = 0;
1731 
1732 	list_for_each_entry(found, &fs_info->space_info, list) {
1733 		if (found->flags & BTRFS_BLOCK_GROUP_DATA) {
1734 			int i;
1735 
1736 			total_free_data += found->disk_total - found->disk_used;
1737 			total_free_data -=
1738 				btrfs_account_ro_block_groups_free_space(found);
1739 
1740 			for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) {
1741 				if (!list_empty(&found->block_groups[i]))
1742 					factor = btrfs_bg_type_to_factor(
1743 						btrfs_raid_array[i].bg_flag);
1744 			}
1745 		}
1746 
1747 		/*
1748 		 * Metadata in mixed block group profiles are accounted in data
1749 		 */
1750 		if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) {
1751 			if (found->flags & BTRFS_BLOCK_GROUP_DATA)
1752 				mixed = 1;
1753 			else
1754 				total_free_meta += found->disk_total -
1755 					found->disk_used;
1756 		}
1757 
1758 		total_used += found->disk_used;
1759 	}
1760 
1761 	buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor);
1762 	buf->f_blocks >>= bits;
1763 	buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits);
1764 
1765 	/* Account global block reserve as used, it's in logical size already */
1766 	spin_lock(&block_rsv->lock);
1767 	/* Mixed block groups accounting is not byte-accurate, avoid overflow */
1768 	if (buf->f_bfree >= block_rsv->size >> bits)
1769 		buf->f_bfree -= block_rsv->size >> bits;
1770 	else
1771 		buf->f_bfree = 0;
1772 	spin_unlock(&block_rsv->lock);
1773 
1774 	buf->f_bavail = div_u64(total_free_data, factor);
1775 	ret = btrfs_calc_avail_data_space(fs_info, &total_free_data);
1776 	if (ret)
1777 		return ret;
1778 	buf->f_bavail += div_u64(total_free_data, factor);
1779 	buf->f_bavail = buf->f_bavail >> bits;
1780 
1781 	/*
1782 	 * We calculate the remaining metadata space minus global reserve. If
1783 	 * this is (supposedly) smaller than zero, there's no space. But this
1784 	 * does not hold in practice, the exhausted state happens where's still
1785 	 * some positive delta. So we apply some guesswork and compare the
1786 	 * delta to a 4M threshold.  (Practically observed delta was ~2M.)
1787 	 *
1788 	 * We probably cannot calculate the exact threshold value because this
1789 	 * depends on the internal reservations requested by various
1790 	 * operations, so some operations that consume a few metadata will
1791 	 * succeed even if the Avail is zero. But this is better than the other
1792 	 * way around.
1793 	 */
1794 	thresh = SZ_4M;
1795 
1796 	/*
1797 	 * We only want to claim there's no available space if we can no longer
1798 	 * allocate chunks for our metadata profile and our global reserve will
1799 	 * not fit in the free metadata space.  If we aren't ->full then we
1800 	 * still can allocate chunks and thus are fine using the currently
1801 	 * calculated f_bavail.
1802 	 */
1803 	if (!mixed && block_rsv->space_info->full &&
1804 	    (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size))
1805 		buf->f_bavail = 0;
1806 
1807 	buf->f_type = BTRFS_SUPER_MAGIC;
1808 	buf->f_bsize = fs_info->sectorsize;
1809 	buf->f_namelen = BTRFS_NAME_LEN;
1810 
1811 	/* We treat it as constant endianness (it doesn't matter _which_)
1812 	   because we want the fsid to come out the same whether mounted
1813 	   on a big-endian or little-endian host */
1814 	buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
1815 	buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
1816 	/* Mask in the root object ID too, to disambiguate subvols */
1817 	buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32;
1818 	buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root);
1819 
1820 	return 0;
1821 }
1822 
btrfs_fc_test_super(struct super_block * sb,struct fs_context * fc)1823 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc)
1824 {
1825 	struct btrfs_fs_info *p = fc->s_fs_info;
1826 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
1827 
1828 	return fs_info->fs_devices == p->fs_devices;
1829 }
1830 
btrfs_get_tree_super(struct fs_context * fc)1831 static int btrfs_get_tree_super(struct fs_context *fc)
1832 {
1833 	struct btrfs_fs_info *fs_info = fc->s_fs_info;
1834 	struct btrfs_fs_context *ctx = fc->fs_private;
1835 	struct btrfs_fs_devices *fs_devices = NULL;
1836 	struct block_device *bdev;
1837 	struct btrfs_device *device;
1838 	struct super_block *sb;
1839 	blk_mode_t mode = btrfs_open_mode(fc);
1840 	int ret;
1841 
1842 	btrfs_ctx_to_info(fs_info, ctx);
1843 	mutex_lock(&uuid_mutex);
1844 
1845 	/*
1846 	 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect
1847 	 * either a valid device or an error.
1848 	 */
1849 	device = btrfs_scan_one_device(fc->source, mode, true);
1850 	ASSERT(device != NULL);
1851 	if (IS_ERR(device)) {
1852 		mutex_unlock(&uuid_mutex);
1853 		return PTR_ERR(device);
1854 	}
1855 
1856 	fs_devices = device->fs_devices;
1857 	fs_info->fs_devices = fs_devices;
1858 
1859 	ret = btrfs_open_devices(fs_devices, mode, &btrfs_fs_type);
1860 	mutex_unlock(&uuid_mutex);
1861 	if (ret)
1862 		return ret;
1863 
1864 	if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) {
1865 		ret = -EACCES;
1866 		goto error;
1867 	}
1868 
1869 	bdev = fs_devices->latest_dev->bdev;
1870 
1871 	/*
1872 	 * From now on the error handling is not straightforward.
1873 	 *
1874 	 * If successful, this will transfer the fs_info into the super block,
1875 	 * and fc->s_fs_info will be NULL.  However if there's an existing
1876 	 * super, we'll still have fc->s_fs_info populated.  If we error
1877 	 * completely out it'll be cleaned up when we drop the fs_context,
1878 	 * otherwise it's tied to the lifetime of the super_block.
1879 	 */
1880 	sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc);
1881 	if (IS_ERR(sb)) {
1882 		ret = PTR_ERR(sb);
1883 		goto error;
1884 	}
1885 
1886 	set_device_specific_options(fs_info);
1887 
1888 	if (sb->s_root) {
1889 		btrfs_close_devices(fs_devices);
1890 		if ((fc->sb_flags ^ sb->s_flags) & SB_RDONLY)
1891 			ret = -EBUSY;
1892 	} else {
1893 		snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
1894 		shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id);
1895 		btrfs_sb(sb)->bdev_holder = &btrfs_fs_type;
1896 		ret = btrfs_fill_super(sb, fs_devices, NULL);
1897 	}
1898 
1899 	if (ret) {
1900 		deactivate_locked_super(sb);
1901 		return ret;
1902 	}
1903 
1904 	btrfs_clear_oneshot_options(fs_info);
1905 
1906 	fc->root = dget(sb->s_root);
1907 	return 0;
1908 
1909 error:
1910 	btrfs_close_devices(fs_devices);
1911 	return ret;
1912 }
1913 
1914 /*
1915  * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes
1916  * with different ro/rw options") the following works:
1917  *
1918  *        (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo
1919  *       (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar
1920  *
1921  * which looks nice and innocent but is actually pretty intricate and deserves
1922  * a long comment.
1923  *
1924  * On another filesystem a subvolume mount is close to something like:
1925  *
1926  *	(iii) # create rw superblock + initial mount
1927  *	      mount -t xfs /dev/sdb /opt/
1928  *
1929  *	      # create ro bind mount
1930  *	      mount --bind -o ro /opt/foo /mnt/foo
1931  *
1932  *	      # unmount initial mount
1933  *	      umount /opt
1934  *
1935  * Of course, there's some special subvolume sauce and there's the fact that the
1936  * sb->s_root dentry is really swapped after mount_subtree(). But conceptually
1937  * it's very close and will help us understand the issue.
1938  *
1939  * The old mount API didn't cleanly distinguish between a mount being made ro
1940  * and a superblock being made ro.  The only way to change the ro state of
1941  * either object was by passing ms_rdonly. If a new mount was created via
1942  * mount(2) such as:
1943  *
1944  *      mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null);
1945  *
1946  * the MS_RDONLY flag being specified had two effects:
1947  *
1948  * (1) MNT_READONLY was raised -> the resulting mount got
1949  *     @mnt->mnt_flags |= MNT_READONLY raised.
1950  *
1951  * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems
1952  *     made the superblock ro. Note, how SB_RDONLY has the same value as
1953  *     ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2).
1954  *
1955  * Creating a subtree mount via (iii) ends up leaving a rw superblock with a
1956  * subtree mounted ro.
1957  *
1958  * But consider the effect on the old mount API on btrfs subvolume mounting
1959  * which combines the distinct step in (iii) into a single step.
1960  *
1961  * By issuing (i) both the mount and the superblock are turned ro. Now when (ii)
1962  * is issued the superblock is ro and thus even if the mount created for (ii) is
1963  * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro
1964  * to rw for (ii) which it did using an internal remount call.
1965  *
1966  * IOW, subvolume mounting was inherently complicated due to the ambiguity of
1967  * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate
1968  * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when
1969  * passed by mount(8) to mount(2).
1970  *
1971  * Enter the new mount API. The new mount API disambiguates making a mount ro
1972  * and making a superblock ro.
1973  *
1974  * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either
1975  *     fsmount() or mount_setattr() this is a pure VFS level change for a
1976  *     specific mount or mount tree that is never seen by the filesystem itself.
1977  *
1978  * (4) To turn a superblock ro the "ro" flag must be used with
1979  *     fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem
1980  *     in fc->sb_flags.
1981  *
1982  * This disambiguation has rather positive consequences.  Mounting a subvolume
1983  * ro will not also turn the superblock ro. Only the mount for the subvolume
1984  * will become ro.
1985  *
1986  * So, if the superblock creation request comes from the new mount API the
1987  * caller must have explicitly done:
1988  *
1989  *      fsconfig(FSCONFIG_SET_FLAG, "ro")
1990  *      fsmount/mount_setattr(MOUNT_ATTR_RDONLY)
1991  *
1992  * IOW, at some point the caller must have explicitly turned the whole
1993  * superblock ro and we shouldn't just undo it like we did for the old mount
1994  * API. In any case, it lets us avoid the hack in the new mount API.
1995  *
1996  * Consequently, the remounting hack must only be used for requests originating
1997  * from the old mount API and should be marked for full deprecation so it can be
1998  * turned off in a couple of years.
1999  *
2000  * The new mount API has no reason to support this hack.
2001  */
btrfs_reconfigure_for_mount(struct fs_context * fc)2002 static struct vfsmount *btrfs_reconfigure_for_mount(struct fs_context *fc)
2003 {
2004 	struct vfsmount *mnt;
2005 	int ret;
2006 	const bool ro2rw = !(fc->sb_flags & SB_RDONLY);
2007 
2008 	/*
2009 	 * We got an EBUSY because our SB_RDONLY flag didn't match the existing
2010 	 * super block, so invert our setting here and retry the mount so we
2011 	 * can get our vfsmount.
2012 	 */
2013 	if (ro2rw)
2014 		fc->sb_flags |= SB_RDONLY;
2015 	else
2016 		fc->sb_flags &= ~SB_RDONLY;
2017 
2018 	mnt = fc_mount(fc);
2019 	if (IS_ERR(mnt))
2020 		return mnt;
2021 
2022 	if (!fc->oldapi || !ro2rw)
2023 		return mnt;
2024 
2025 	/* We need to convert to rw, call reconfigure. */
2026 	fc->sb_flags &= ~SB_RDONLY;
2027 	down_write(&mnt->mnt_sb->s_umount);
2028 	ret = btrfs_reconfigure(fc);
2029 	up_write(&mnt->mnt_sb->s_umount);
2030 	if (ret) {
2031 		mntput(mnt);
2032 		return ERR_PTR(ret);
2033 	}
2034 	return mnt;
2035 }
2036 
btrfs_get_tree_subvol(struct fs_context * fc)2037 static int btrfs_get_tree_subvol(struct fs_context *fc)
2038 {
2039 	struct btrfs_fs_info *fs_info = NULL;
2040 	struct btrfs_fs_context *ctx = fc->fs_private;
2041 	struct fs_context *dup_fc;
2042 	struct dentry *dentry;
2043 	struct vfsmount *mnt;
2044 
2045 	/*
2046 	 * Setup a dummy root and fs_info for test/set super.  This is because
2047 	 * we don't actually fill this stuff out until open_ctree, but we need
2048 	 * then open_ctree will properly initialize the file system specific
2049 	 * settings later.  btrfs_init_fs_info initializes the static elements
2050 	 * of the fs_info (locks and such) to make cleanup easier if we find a
2051 	 * superblock with our given fs_devices later on at sget() time.
2052 	 */
2053 	fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL);
2054 	if (!fs_info)
2055 		return -ENOMEM;
2056 
2057 	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2058 	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL);
2059 	if (!fs_info->super_copy || !fs_info->super_for_commit) {
2060 		btrfs_free_fs_info(fs_info);
2061 		return -ENOMEM;
2062 	}
2063 	btrfs_init_fs_info(fs_info);
2064 
2065 	dup_fc = vfs_dup_fs_context(fc);
2066 	if (IS_ERR(dup_fc)) {
2067 		btrfs_free_fs_info(fs_info);
2068 		return PTR_ERR(dup_fc);
2069 	}
2070 
2071 	/*
2072 	 * When we do the sget_fc this gets transferred to the sb, so we only
2073 	 * need to set it on the dup_fc as this is what creates the super block.
2074 	 */
2075 	dup_fc->s_fs_info = fs_info;
2076 
2077 	/*
2078 	 * We'll do the security settings in our btrfs_get_tree_super() mount
2079 	 * loop, they were duplicated into dup_fc, we can drop the originals
2080 	 * here.
2081 	 */
2082 	security_free_mnt_opts(&fc->security);
2083 	fc->security = NULL;
2084 
2085 	mnt = fc_mount(dup_fc);
2086 	if (PTR_ERR_OR_ZERO(mnt) == -EBUSY)
2087 		mnt = btrfs_reconfigure_for_mount(dup_fc);
2088 	put_fs_context(dup_fc);
2089 	if (IS_ERR(mnt))
2090 		return PTR_ERR(mnt);
2091 
2092 	/*
2093 	 * This free's ->subvol_name, because if it isn't set we have to
2094 	 * allocate a buffer to hold the subvol_name, so we just drop our
2095 	 * reference to it here.
2096 	 */
2097 	dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt);
2098 	ctx->subvol_name = NULL;
2099 	if (IS_ERR(dentry))
2100 		return PTR_ERR(dentry);
2101 
2102 	fc->root = dentry;
2103 	return 0;
2104 }
2105 
btrfs_get_tree(struct fs_context * fc)2106 static int btrfs_get_tree(struct fs_context *fc)
2107 {
2108 	/*
2109 	 * Since we use mount_subtree to mount the default/specified subvol, we
2110 	 * have to do mounts in two steps.
2111 	 *
2112 	 * First pass through we call btrfs_get_tree_subvol(), this is just a
2113 	 * wrapper around fc_mount() to call back into here again, and this time
2114 	 * we'll call btrfs_get_tree_super().  This will do the open_ctree() and
2115 	 * everything to open the devices and file system.  Then we return back
2116 	 * with a fully constructed vfsmount in btrfs_get_tree_subvol(), and
2117 	 * from there we can do our mount_subvol() call, which will lookup
2118 	 * whichever subvol we're mounting and setup this fc with the
2119 	 * appropriate dentry for the subvol.
2120 	 */
2121 	if (fc->s_fs_info)
2122 		return btrfs_get_tree_super(fc);
2123 	return btrfs_get_tree_subvol(fc);
2124 }
2125 
btrfs_kill_super(struct super_block * sb)2126 static void btrfs_kill_super(struct super_block *sb)
2127 {
2128 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2129 	kill_anon_super(sb);
2130 	btrfs_free_fs_info(fs_info);
2131 }
2132 
btrfs_free_fs_context(struct fs_context * fc)2133 static void btrfs_free_fs_context(struct fs_context *fc)
2134 {
2135 	struct btrfs_fs_context *ctx = fc->fs_private;
2136 	struct btrfs_fs_info *fs_info = fc->s_fs_info;
2137 
2138 	if (fs_info)
2139 		btrfs_free_fs_info(fs_info);
2140 
2141 	if (ctx && refcount_dec_and_test(&ctx->refs)) {
2142 		kfree(ctx->subvol_name);
2143 		kfree(ctx);
2144 	}
2145 }
2146 
btrfs_dup_fs_context(struct fs_context * fc,struct fs_context * src_fc)2147 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc)
2148 {
2149 	struct btrfs_fs_context *ctx = src_fc->fs_private;
2150 
2151 	/*
2152 	 * Give a ref to our ctx to this dup, as we want to keep it around for
2153 	 * our original fc so we can have the subvolume name or objectid.
2154 	 *
2155 	 * We unset ->source in the original fc because the dup needs it for
2156 	 * mounting, and then once we free the dup it'll free ->source, so we
2157 	 * need to make sure we're only pointing to it in one fc.
2158 	 */
2159 	refcount_inc(&ctx->refs);
2160 	fc->fs_private = ctx;
2161 	fc->source = src_fc->source;
2162 	src_fc->source = NULL;
2163 	return 0;
2164 }
2165 
2166 static const struct fs_context_operations btrfs_fs_context_ops = {
2167 	.parse_param	= btrfs_parse_param,
2168 	.reconfigure	= btrfs_reconfigure,
2169 	.get_tree	= btrfs_get_tree,
2170 	.dup		= btrfs_dup_fs_context,
2171 	.free		= btrfs_free_fs_context,
2172 };
2173 
btrfs_init_fs_context(struct fs_context * fc)2174 static int btrfs_init_fs_context(struct fs_context *fc)
2175 {
2176 	struct btrfs_fs_context *ctx;
2177 
2178 	ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL);
2179 	if (!ctx)
2180 		return -ENOMEM;
2181 
2182 	refcount_set(&ctx->refs, 1);
2183 	fc->fs_private = ctx;
2184 	fc->ops = &btrfs_fs_context_ops;
2185 
2186 	if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2187 		btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx);
2188 	} else {
2189 		ctx->thread_pool_size =
2190 			min_t(unsigned long, num_online_cpus() + 2, 8);
2191 		ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE;
2192 		ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL;
2193 	}
2194 
2195 #ifdef CONFIG_BTRFS_FS_POSIX_ACL
2196 	fc->sb_flags |= SB_POSIXACL;
2197 #endif
2198 	fc->sb_flags |= SB_I_VERSION;
2199 
2200 	return 0;
2201 }
2202 
2203 static struct file_system_type btrfs_fs_type = {
2204 	.owner			= THIS_MODULE,
2205 	.name			= "btrfs",
2206 	.init_fs_context	= btrfs_init_fs_context,
2207 	.parameters		= btrfs_fs_parameters,
2208 	.kill_sb		= btrfs_kill_super,
2209 	.fs_flags		= FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP,
2210  };
2211 
2212 MODULE_ALIAS_FS("btrfs");
2213 
btrfs_control_open(struct inode * inode,struct file * file)2214 static int btrfs_control_open(struct inode *inode, struct file *file)
2215 {
2216 	/*
2217 	 * The control file's private_data is used to hold the
2218 	 * transaction when it is started and is used to keep
2219 	 * track of whether a transaction is already in progress.
2220 	 */
2221 	file->private_data = NULL;
2222 	return 0;
2223 }
2224 
2225 /*
2226  * Used by /dev/btrfs-control for devices ioctls.
2227  */
btrfs_control_ioctl(struct file * file,unsigned int cmd,unsigned long arg)2228 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
2229 				unsigned long arg)
2230 {
2231 	struct btrfs_ioctl_vol_args *vol;
2232 	struct btrfs_device *device = NULL;
2233 	dev_t devt = 0;
2234 	int ret = -ENOTTY;
2235 
2236 	if (!capable(CAP_SYS_ADMIN))
2237 		return -EPERM;
2238 
2239 	vol = memdup_user((void __user *)arg, sizeof(*vol));
2240 	if (IS_ERR(vol))
2241 		return PTR_ERR(vol);
2242 	ret = btrfs_check_ioctl_vol_args_path(vol);
2243 	if (ret < 0)
2244 		goto out;
2245 
2246 	switch (cmd) {
2247 	case BTRFS_IOC_SCAN_DEV:
2248 		mutex_lock(&uuid_mutex);
2249 		/*
2250 		 * Scanning outside of mount can return NULL which would turn
2251 		 * into 0 error code.
2252 		 */
2253 		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2254 		ret = PTR_ERR_OR_ZERO(device);
2255 		mutex_unlock(&uuid_mutex);
2256 		break;
2257 	case BTRFS_IOC_FORGET_DEV:
2258 		if (vol->name[0] != 0) {
2259 			ret = lookup_bdev(vol->name, &devt);
2260 			if (ret)
2261 				break;
2262 		}
2263 		ret = btrfs_forget_devices(devt);
2264 		break;
2265 	case BTRFS_IOC_DEVICES_READY:
2266 		mutex_lock(&uuid_mutex);
2267 		/*
2268 		 * Scanning outside of mount can return NULL which would turn
2269 		 * into 0 error code.
2270 		 */
2271 		device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false);
2272 		if (IS_ERR_OR_NULL(device)) {
2273 			mutex_unlock(&uuid_mutex);
2274 			ret = PTR_ERR(device);
2275 			break;
2276 		}
2277 		ret = !(device->fs_devices->num_devices ==
2278 			device->fs_devices->total_devices);
2279 		mutex_unlock(&uuid_mutex);
2280 		break;
2281 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
2282 		ret = btrfs_ioctl_get_supported_features((void __user*)arg);
2283 		break;
2284 	}
2285 
2286 out:
2287 	kfree(vol);
2288 	return ret;
2289 }
2290 
btrfs_freeze(struct super_block * sb)2291 static int btrfs_freeze(struct super_block *sb)
2292 {
2293 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2294 
2295 	set_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2296 	/*
2297 	 * We don't need a barrier here, we'll wait for any transaction that
2298 	 * could be in progress on other threads (and do delayed iputs that
2299 	 * we want to avoid on a frozen filesystem), or do the commit
2300 	 * ourselves.
2301 	 */
2302 	return btrfs_commit_current_transaction(fs_info->tree_root);
2303 }
2304 
check_dev_super(struct btrfs_device * dev)2305 static int check_dev_super(struct btrfs_device *dev)
2306 {
2307 	struct btrfs_fs_info *fs_info = dev->fs_info;
2308 	struct btrfs_super_block *sb;
2309 	u64 last_trans;
2310 	u16 csum_type;
2311 	int ret = 0;
2312 
2313 	/* This should be called with fs still frozen. */
2314 	ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags));
2315 
2316 	/* Missing dev, no need to check. */
2317 	if (!dev->bdev)
2318 		return 0;
2319 
2320 	/* Only need to check the primary super block. */
2321 	sb = btrfs_read_dev_one_super(dev->bdev, 0, true);
2322 	if (IS_ERR(sb))
2323 		return PTR_ERR(sb);
2324 
2325 	/* Verify the checksum. */
2326 	csum_type = btrfs_super_csum_type(sb);
2327 	if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) {
2328 		btrfs_err(fs_info, "csum type changed, has %u expect %u",
2329 			  csum_type, btrfs_super_csum_type(fs_info->super_copy));
2330 		ret = -EUCLEAN;
2331 		goto out;
2332 	}
2333 
2334 	if (btrfs_check_super_csum(fs_info, sb)) {
2335 		btrfs_err(fs_info, "csum for on-disk super block no longer matches");
2336 		ret = -EUCLEAN;
2337 		goto out;
2338 	}
2339 
2340 	/* Btrfs_validate_super() includes fsid check against super->fsid. */
2341 	ret = btrfs_validate_super(fs_info, sb, 0);
2342 	if (ret < 0)
2343 		goto out;
2344 
2345 	last_trans = btrfs_get_last_trans_committed(fs_info);
2346 	if (btrfs_super_generation(sb) != last_trans) {
2347 		btrfs_err(fs_info, "transid mismatch, has %llu expect %llu",
2348 			  btrfs_super_generation(sb), last_trans);
2349 		ret = -EUCLEAN;
2350 		goto out;
2351 	}
2352 out:
2353 	btrfs_release_disk_super(sb);
2354 	return ret;
2355 }
2356 
btrfs_unfreeze(struct super_block * sb)2357 static int btrfs_unfreeze(struct super_block *sb)
2358 {
2359 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2360 	struct btrfs_device *device;
2361 	int ret = 0;
2362 
2363 	/*
2364 	 * Make sure the fs is not changed by accident (like hibernation then
2365 	 * modified by other OS).
2366 	 * If we found anything wrong, we mark the fs error immediately.
2367 	 *
2368 	 * And since the fs is frozen, no one can modify the fs yet, thus
2369 	 * we don't need to hold device_list_mutex.
2370 	 */
2371 	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
2372 		ret = check_dev_super(device);
2373 		if (ret < 0) {
2374 			btrfs_handle_fs_error(fs_info, ret,
2375 				"super block on devid %llu got modified unexpectedly",
2376 				device->devid);
2377 			break;
2378 		}
2379 	}
2380 	clear_bit(BTRFS_FS_FROZEN, &fs_info->flags);
2381 
2382 	/*
2383 	 * We still return 0, to allow VFS layer to unfreeze the fs even the
2384 	 * above checks failed. Since the fs is either fine or read-only, we're
2385 	 * safe to continue, without causing further damage.
2386 	 */
2387 	return 0;
2388 }
2389 
btrfs_show_devname(struct seq_file * m,struct dentry * root)2390 static int btrfs_show_devname(struct seq_file *m, struct dentry *root)
2391 {
2392 	struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb);
2393 
2394 	/*
2395 	 * There should be always a valid pointer in latest_dev, it may be stale
2396 	 * for a short moment in case it's being deleted but still valid until
2397 	 * the end of RCU grace period.
2398 	 */
2399 	rcu_read_lock();
2400 	seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\");
2401 	rcu_read_unlock();
2402 
2403 	return 0;
2404 }
2405 
btrfs_nr_cached_objects(struct super_block * sb,struct shrink_control * sc)2406 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc)
2407 {
2408 	struct btrfs_fs_info *fs_info = btrfs_sb(sb);
2409 	const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps);
2410 
2411 	trace_btrfs_extent_map_shrinker_count(fs_info, nr);
2412 
2413 	/*
2414 	 * Only report the real number for DEBUG builds, as there are reports of
2415 	 * serious performance degradation caused by too frequent shrinks.
2416 	 */
2417 	if (IS_ENABLED(CONFIG_BTRFS_DEBUG))
2418 		return nr;
2419 	return 0;
2420 }
2421 
btrfs_free_cached_objects(struct super_block * sb,struct shrink_control * sc)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 	/*
2428 	 * We may be called from any task trying to allocate memory and we don't
2429 	 * want to slow it down with scanning and dropping extent maps. It would
2430 	 * also cause heavy lock contention if many tasks concurrently enter
2431 	 * here. Therefore only allow kswapd tasks to scan and drop extent maps.
2432 	 */
2433 	if (!current_is_kswapd())
2434 		return 0;
2435 
2436 	return btrfs_free_extent_maps(fs_info, nr_to_scan);
2437 }
2438 
2439 static const struct super_operations btrfs_super_ops = {
2440 	.drop_inode	= btrfs_drop_inode,
2441 	.evict_inode	= btrfs_evict_inode,
2442 	.put_super	= btrfs_put_super,
2443 	.sync_fs	= btrfs_sync_fs,
2444 	.show_options	= btrfs_show_options,
2445 	.show_devname	= btrfs_show_devname,
2446 	.alloc_inode	= btrfs_alloc_inode,
2447 	.destroy_inode	= btrfs_destroy_inode,
2448 	.free_inode	= btrfs_free_inode,
2449 	.statfs		= btrfs_statfs,
2450 	.freeze_fs	= btrfs_freeze,
2451 	.unfreeze_fs	= btrfs_unfreeze,
2452 	.nr_cached_objects = btrfs_nr_cached_objects,
2453 	.free_cached_objects = btrfs_free_cached_objects,
2454 };
2455 
2456 static const struct file_operations btrfs_ctl_fops = {
2457 	.open = btrfs_control_open,
2458 	.unlocked_ioctl	 = btrfs_control_ioctl,
2459 	.compat_ioctl = compat_ptr_ioctl,
2460 	.owner	 = THIS_MODULE,
2461 	.llseek = noop_llseek,
2462 };
2463 
2464 static struct miscdevice btrfs_misc = {
2465 	.minor		= BTRFS_MINOR,
2466 	.name		= "btrfs-control",
2467 	.fops		= &btrfs_ctl_fops
2468 };
2469 
2470 MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
2471 MODULE_ALIAS("devname:btrfs-control");
2472 
btrfs_interface_init(void)2473 static int __init btrfs_interface_init(void)
2474 {
2475 	return misc_register(&btrfs_misc);
2476 }
2477 
btrfs_interface_exit(void)2478 static __cold void btrfs_interface_exit(void)
2479 {
2480 	misc_deregister(&btrfs_misc);
2481 }
2482 
btrfs_print_mod_info(void)2483 static int __init btrfs_print_mod_info(void)
2484 {
2485 	static const char options[] = ""
2486 #ifdef CONFIG_BTRFS_DEBUG
2487 			", debug=on"
2488 #endif
2489 #ifdef CONFIG_BTRFS_ASSERT
2490 			", assert=on"
2491 #endif
2492 #ifdef CONFIG_BTRFS_FS_REF_VERIFY
2493 			", ref-verify=on"
2494 #endif
2495 #ifdef CONFIG_BLK_DEV_ZONED
2496 			", zoned=yes"
2497 #else
2498 			", zoned=no"
2499 #endif
2500 #ifdef CONFIG_FS_VERITY
2501 			", fsverity=yes"
2502 #else
2503 			", fsverity=no"
2504 #endif
2505 			;
2506 	pr_info("Btrfs loaded%s\n", options);
2507 	return 0;
2508 }
2509 
register_btrfs(void)2510 static int register_btrfs(void)
2511 {
2512 	return register_filesystem(&btrfs_fs_type);
2513 }
2514 
unregister_btrfs(void)2515 static void unregister_btrfs(void)
2516 {
2517 	unregister_filesystem(&btrfs_fs_type);
2518 }
2519 
2520 /* Helper structure for long init/exit functions. */
2521 struct init_sequence {
2522 	int (*init_func)(void);
2523 	/* Can be NULL if the init_func doesn't need cleanup. */
2524 	void (*exit_func)(void);
2525 };
2526 
2527 static const struct init_sequence mod_init_seq[] = {
2528 	{
2529 		.init_func = btrfs_props_init,
2530 		.exit_func = NULL,
2531 	}, {
2532 		.init_func = btrfs_init_sysfs,
2533 		.exit_func = btrfs_exit_sysfs,
2534 	}, {
2535 		.init_func = btrfs_init_compress,
2536 		.exit_func = btrfs_exit_compress,
2537 	}, {
2538 		.init_func = btrfs_init_cachep,
2539 		.exit_func = btrfs_destroy_cachep,
2540 	}, {
2541 		.init_func = btrfs_init_dio,
2542 		.exit_func = btrfs_destroy_dio,
2543 	}, {
2544 		.init_func = btrfs_transaction_init,
2545 		.exit_func = btrfs_transaction_exit,
2546 	}, {
2547 		.init_func = btrfs_ctree_init,
2548 		.exit_func = btrfs_ctree_exit,
2549 	}, {
2550 		.init_func = btrfs_free_space_init,
2551 		.exit_func = btrfs_free_space_exit,
2552 	}, {
2553 		.init_func = extent_state_init_cachep,
2554 		.exit_func = extent_state_free_cachep,
2555 	}, {
2556 		.init_func = extent_buffer_init_cachep,
2557 		.exit_func = extent_buffer_free_cachep,
2558 	}, {
2559 		.init_func = btrfs_bioset_init,
2560 		.exit_func = btrfs_bioset_exit,
2561 	}, {
2562 		.init_func = extent_map_init,
2563 		.exit_func = extent_map_exit,
2564 	}, {
2565 		.init_func = ordered_data_init,
2566 		.exit_func = ordered_data_exit,
2567 	}, {
2568 		.init_func = btrfs_delayed_inode_init,
2569 		.exit_func = btrfs_delayed_inode_exit,
2570 	}, {
2571 		.init_func = btrfs_auto_defrag_init,
2572 		.exit_func = btrfs_auto_defrag_exit,
2573 	}, {
2574 		.init_func = btrfs_delayed_ref_init,
2575 		.exit_func = btrfs_delayed_ref_exit,
2576 	}, {
2577 		.init_func = btrfs_prelim_ref_init,
2578 		.exit_func = btrfs_prelim_ref_exit,
2579 	}, {
2580 		.init_func = btrfs_interface_init,
2581 		.exit_func = btrfs_interface_exit,
2582 	}, {
2583 		.init_func = btrfs_print_mod_info,
2584 		.exit_func = NULL,
2585 	}, {
2586 		.init_func = btrfs_run_sanity_tests,
2587 		.exit_func = NULL,
2588 	}, {
2589 		.init_func = register_btrfs,
2590 		.exit_func = unregister_btrfs,
2591 	}
2592 };
2593 
2594 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)];
2595 
btrfs_exit_btrfs_fs(void)2596 static __always_inline void btrfs_exit_btrfs_fs(void)
2597 {
2598 	int i;
2599 
2600 	for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) {
2601 		if (!mod_init_result[i])
2602 			continue;
2603 		if (mod_init_seq[i].exit_func)
2604 			mod_init_seq[i].exit_func();
2605 		mod_init_result[i] = false;
2606 	}
2607 }
2608 
exit_btrfs_fs(void)2609 static void __exit exit_btrfs_fs(void)
2610 {
2611 	btrfs_exit_btrfs_fs();
2612 	btrfs_cleanup_fs_uuids();
2613 }
2614 
init_btrfs_fs(void)2615 static int __init init_btrfs_fs(void)
2616 {
2617 	int ret;
2618 	int i;
2619 
2620 	for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) {
2621 		ASSERT(!mod_init_result[i]);
2622 		ret = mod_init_seq[i].init_func();
2623 		if (ret < 0) {
2624 			btrfs_exit_btrfs_fs();
2625 			return ret;
2626 		}
2627 		mod_init_result[i] = true;
2628 	}
2629 	return 0;
2630 }
2631 
2632 late_initcall(init_btrfs_fs);
2633 module_exit(exit_btrfs_fs)
2634 
2635 MODULE_DESCRIPTION("B-Tree File System (BTRFS)");
2636 MODULE_LICENSE("GPL");
2637 MODULE_SOFTDEP("pre: crc32c");
2638 MODULE_SOFTDEP("pre: xxhash64");
2639 MODULE_SOFTDEP("pre: sha256");
2640 MODULE_SOFTDEP("pre: blake2b-256");
2641