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 "messages.h" 31 #include "delayed-inode.h" 32 #include "ctree.h" 33 #include "disk-io.h" 34 #include "transaction.h" 35 #include "btrfs_inode.h" 36 #include "print-tree.h" 37 #include "props.h" 38 #include "xattr.h" 39 #include "bio.h" 40 #include "export.h" 41 #include "compression.h" 42 #include "rcu-string.h" 43 #include "dev-replace.h" 44 #include "free-space-cache.h" 45 #include "backref.h" 46 #include "space-info.h" 47 #include "sysfs.h" 48 #include "zoned.h" 49 #include "tests/btrfs-tests.h" 50 #include "block-group.h" 51 #include "discard.h" 52 #include "qgroup.h" 53 #include "raid56.h" 54 #include "fs.h" 55 #include "accessors.h" 56 #include "defrag.h" 57 #include "dir-item.h" 58 #include "ioctl.h" 59 #include "scrub.h" 60 #include "verity.h" 61 #include "super.h" 62 #include "extent-tree.h" 63 #define CREATE_TRACE_POINTS 64 #include <trace/events/btrfs.h> 65 66 static const struct super_operations btrfs_super_ops; 67 68 /* 69 * Types for mounting the default subvolume and a subvolume explicitly 70 * requested by subvol=/path. That way the callchain is straightforward and we 71 * don't have to play tricks with the mount options and recursive calls to 72 * btrfs_mount. 73 * 74 * The new btrfs_root_fs_type also servers as a tag for the bdev_holder. 75 */ 76 static struct file_system_type btrfs_fs_type; 77 static struct file_system_type btrfs_root_fs_type; 78 79 static int btrfs_remount(struct super_block *sb, int *flags, char *data); 80 81 static void btrfs_put_super(struct super_block *sb) 82 { 83 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 84 85 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid); 86 close_ctree(fs_info); 87 } 88 89 enum { 90 Opt_acl, Opt_noacl, 91 Opt_clear_cache, 92 Opt_commit_interval, 93 Opt_compress, 94 Opt_compress_force, 95 Opt_compress_force_type, 96 Opt_compress_type, 97 Opt_degraded, 98 Opt_device, 99 Opt_fatal_errors, 100 Opt_flushoncommit, Opt_noflushoncommit, 101 Opt_max_inline, 102 Opt_barrier, Opt_nobarrier, 103 Opt_datacow, Opt_nodatacow, 104 Opt_datasum, Opt_nodatasum, 105 Opt_defrag, Opt_nodefrag, 106 Opt_discard, Opt_nodiscard, 107 Opt_discard_mode, 108 Opt_norecovery, 109 Opt_ratio, 110 Opt_rescan_uuid_tree, 111 Opt_skip_balance, 112 Opt_space_cache, Opt_no_space_cache, 113 Opt_space_cache_version, 114 Opt_ssd, Opt_nossd, 115 Opt_ssd_spread, Opt_nossd_spread, 116 Opt_subvol, 117 Opt_subvol_empty, 118 Opt_subvolid, 119 Opt_thread_pool, 120 Opt_treelog, Opt_notreelog, 121 Opt_user_subvol_rm_allowed, 122 123 /* Rescue options */ 124 Opt_rescue, 125 Opt_usebackuproot, 126 Opt_nologreplay, 127 Opt_ignorebadroots, 128 Opt_ignoredatacsums, 129 Opt_rescue_all, 130 131 /* Deprecated options */ 132 Opt_recovery, 133 Opt_inode_cache, Opt_noinode_cache, 134 135 /* Debugging options */ 136 Opt_enospc_debug, Opt_noenospc_debug, 137 #ifdef CONFIG_BTRFS_DEBUG 138 Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all, 139 #endif 140 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 141 Opt_ref_verify, 142 #endif 143 Opt_err, 144 }; 145 146 static const match_table_t tokens = { 147 {Opt_acl, "acl"}, 148 {Opt_noacl, "noacl"}, 149 {Opt_clear_cache, "clear_cache"}, 150 {Opt_commit_interval, "commit=%u"}, 151 {Opt_compress, "compress"}, 152 {Opt_compress_type, "compress=%s"}, 153 {Opt_compress_force, "compress-force"}, 154 {Opt_compress_force_type, "compress-force=%s"}, 155 {Opt_degraded, "degraded"}, 156 {Opt_device, "device=%s"}, 157 {Opt_fatal_errors, "fatal_errors=%s"}, 158 {Opt_flushoncommit, "flushoncommit"}, 159 {Opt_noflushoncommit, "noflushoncommit"}, 160 {Opt_inode_cache, "inode_cache"}, 161 {Opt_noinode_cache, "noinode_cache"}, 162 {Opt_max_inline, "max_inline=%s"}, 163 {Opt_barrier, "barrier"}, 164 {Opt_nobarrier, "nobarrier"}, 165 {Opt_datacow, "datacow"}, 166 {Opt_nodatacow, "nodatacow"}, 167 {Opt_datasum, "datasum"}, 168 {Opt_nodatasum, "nodatasum"}, 169 {Opt_defrag, "autodefrag"}, 170 {Opt_nodefrag, "noautodefrag"}, 171 {Opt_discard, "discard"}, 172 {Opt_discard_mode, "discard=%s"}, 173 {Opt_nodiscard, "nodiscard"}, 174 {Opt_norecovery, "norecovery"}, 175 {Opt_ratio, "metadata_ratio=%u"}, 176 {Opt_rescan_uuid_tree, "rescan_uuid_tree"}, 177 {Opt_skip_balance, "skip_balance"}, 178 {Opt_space_cache, "space_cache"}, 179 {Opt_no_space_cache, "nospace_cache"}, 180 {Opt_space_cache_version, "space_cache=%s"}, 181 {Opt_ssd, "ssd"}, 182 {Opt_nossd, "nossd"}, 183 {Opt_ssd_spread, "ssd_spread"}, 184 {Opt_nossd_spread, "nossd_spread"}, 185 {Opt_subvol, "subvol=%s"}, 186 {Opt_subvol_empty, "subvol="}, 187 {Opt_subvolid, "subvolid=%s"}, 188 {Opt_thread_pool, "thread_pool=%u"}, 189 {Opt_treelog, "treelog"}, 190 {Opt_notreelog, "notreelog"}, 191 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 192 193 /* Rescue options */ 194 {Opt_rescue, "rescue=%s"}, 195 /* Deprecated, with alias rescue=nologreplay */ 196 {Opt_nologreplay, "nologreplay"}, 197 /* Deprecated, with alias rescue=usebackuproot */ 198 {Opt_usebackuproot, "usebackuproot"}, 199 200 /* Deprecated options */ 201 {Opt_recovery, "recovery"}, 202 203 /* Debugging options */ 204 {Opt_enospc_debug, "enospc_debug"}, 205 {Opt_noenospc_debug, "noenospc_debug"}, 206 #ifdef CONFIG_BTRFS_DEBUG 207 {Opt_fragment_data, "fragment=data"}, 208 {Opt_fragment_metadata, "fragment=metadata"}, 209 {Opt_fragment_all, "fragment=all"}, 210 #endif 211 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 212 {Opt_ref_verify, "ref_verify"}, 213 #endif 214 {Opt_err, NULL}, 215 }; 216 217 static const match_table_t rescue_tokens = { 218 {Opt_usebackuproot, "usebackuproot"}, 219 {Opt_nologreplay, "nologreplay"}, 220 {Opt_ignorebadroots, "ignorebadroots"}, 221 {Opt_ignorebadroots, "ibadroots"}, 222 {Opt_ignoredatacsums, "ignoredatacsums"}, 223 {Opt_ignoredatacsums, "idatacsums"}, 224 {Opt_rescue_all, "all"}, 225 {Opt_err, NULL}, 226 }; 227 228 static bool check_ro_option(struct btrfs_fs_info *fs_info, unsigned long opt, 229 const char *opt_name) 230 { 231 if (fs_info->mount_opt & opt) { 232 btrfs_err(fs_info, "%s must be used with ro mount option", 233 opt_name); 234 return true; 235 } 236 return false; 237 } 238 239 static int parse_rescue_options(struct btrfs_fs_info *info, const char *options) 240 { 241 char *opts; 242 char *orig; 243 char *p; 244 substring_t args[MAX_OPT_ARGS]; 245 int ret = 0; 246 247 opts = kstrdup(options, GFP_KERNEL); 248 if (!opts) 249 return -ENOMEM; 250 orig = opts; 251 252 while ((p = strsep(&opts, ":")) != NULL) { 253 int token; 254 255 if (!*p) 256 continue; 257 token = match_token(p, rescue_tokens, args); 258 switch (token){ 259 case Opt_usebackuproot: 260 btrfs_info(info, 261 "trying to use backup root at mount time"); 262 btrfs_set_opt(info->mount_opt, USEBACKUPROOT); 263 break; 264 case Opt_nologreplay: 265 btrfs_set_and_info(info, NOLOGREPLAY, 266 "disabling log replay at mount time"); 267 break; 268 case Opt_ignorebadroots: 269 btrfs_set_and_info(info, IGNOREBADROOTS, 270 "ignoring bad roots"); 271 break; 272 case Opt_ignoredatacsums: 273 btrfs_set_and_info(info, IGNOREDATACSUMS, 274 "ignoring data csums"); 275 break; 276 case Opt_rescue_all: 277 btrfs_info(info, "enabling all of the rescue options"); 278 btrfs_set_and_info(info, IGNOREDATACSUMS, 279 "ignoring data csums"); 280 btrfs_set_and_info(info, IGNOREBADROOTS, 281 "ignoring bad roots"); 282 btrfs_set_and_info(info, NOLOGREPLAY, 283 "disabling log replay at mount time"); 284 break; 285 case Opt_err: 286 btrfs_info(info, "unrecognized rescue option '%s'", p); 287 ret = -EINVAL; 288 goto out; 289 default: 290 break; 291 } 292 293 } 294 out: 295 kfree(orig); 296 return ret; 297 } 298 299 /* 300 * Regular mount options parser. Everything that is needed only when 301 * reading in a new superblock is parsed here. 302 * XXX JDM: This needs to be cleaned up for remount. 303 */ 304 int btrfs_parse_options(struct btrfs_fs_info *info, char *options, 305 unsigned long new_flags) 306 { 307 substring_t args[MAX_OPT_ARGS]; 308 char *p, *num; 309 int intarg; 310 int ret = 0; 311 char *compress_type; 312 bool compress_force = false; 313 enum btrfs_compression_type saved_compress_type; 314 int saved_compress_level; 315 bool saved_compress_force; 316 int no_compress = 0; 317 const bool remounting = test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state); 318 319 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE)) 320 btrfs_set_opt(info->mount_opt, FREE_SPACE_TREE); 321 else if (btrfs_free_space_cache_v1_active(info)) { 322 if (btrfs_is_zoned(info)) { 323 btrfs_info(info, 324 "zoned: clearing existing space cache"); 325 btrfs_set_super_cache_generation(info->super_copy, 0); 326 } else { 327 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 328 } 329 } 330 331 /* 332 * Even the options are empty, we still need to do extra check 333 * against new flags 334 */ 335 if (!options) 336 goto check; 337 338 while ((p = strsep(&options, ",")) != NULL) { 339 int token; 340 if (!*p) 341 continue; 342 343 token = match_token(p, tokens, args); 344 switch (token) { 345 case Opt_degraded: 346 btrfs_info(info, "allowing degraded mounts"); 347 btrfs_set_opt(info->mount_opt, DEGRADED); 348 break; 349 case Opt_subvol: 350 case Opt_subvol_empty: 351 case Opt_subvolid: 352 case Opt_device: 353 /* 354 * These are parsed by btrfs_parse_subvol_options or 355 * btrfs_parse_device_options and can be ignored here. 356 */ 357 break; 358 case Opt_nodatasum: 359 btrfs_set_and_info(info, NODATASUM, 360 "setting nodatasum"); 361 break; 362 case Opt_datasum: 363 if (btrfs_test_opt(info, NODATASUM)) { 364 if (btrfs_test_opt(info, NODATACOW)) 365 btrfs_info(info, 366 "setting datasum, datacow enabled"); 367 else 368 btrfs_info(info, "setting datasum"); 369 } 370 btrfs_clear_opt(info->mount_opt, NODATACOW); 371 btrfs_clear_opt(info->mount_opt, NODATASUM); 372 break; 373 case Opt_nodatacow: 374 if (!btrfs_test_opt(info, NODATACOW)) { 375 if (!btrfs_test_opt(info, COMPRESS) || 376 !btrfs_test_opt(info, FORCE_COMPRESS)) { 377 btrfs_info(info, 378 "setting nodatacow, compression disabled"); 379 } else { 380 btrfs_info(info, "setting nodatacow"); 381 } 382 } 383 btrfs_clear_opt(info->mount_opt, COMPRESS); 384 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 385 btrfs_set_opt(info->mount_opt, NODATACOW); 386 btrfs_set_opt(info->mount_opt, NODATASUM); 387 break; 388 case Opt_datacow: 389 btrfs_clear_and_info(info, NODATACOW, 390 "setting datacow"); 391 break; 392 case Opt_compress_force: 393 case Opt_compress_force_type: 394 compress_force = true; 395 fallthrough; 396 case Opt_compress: 397 case Opt_compress_type: 398 saved_compress_type = btrfs_test_opt(info, 399 COMPRESS) ? 400 info->compress_type : BTRFS_COMPRESS_NONE; 401 saved_compress_force = 402 btrfs_test_opt(info, FORCE_COMPRESS); 403 saved_compress_level = info->compress_level; 404 if (token == Opt_compress || 405 token == Opt_compress_force || 406 strncmp(args[0].from, "zlib", 4) == 0) { 407 compress_type = "zlib"; 408 409 info->compress_type = BTRFS_COMPRESS_ZLIB; 410 info->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL; 411 /* 412 * args[0] contains uninitialized data since 413 * for these tokens we don't expect any 414 * parameter. 415 */ 416 if (token != Opt_compress && 417 token != Opt_compress_force) 418 info->compress_level = 419 btrfs_compress_str2level( 420 BTRFS_COMPRESS_ZLIB, 421 args[0].from + 4); 422 btrfs_set_opt(info->mount_opt, COMPRESS); 423 btrfs_clear_opt(info->mount_opt, NODATACOW); 424 btrfs_clear_opt(info->mount_opt, NODATASUM); 425 no_compress = 0; 426 } else if (strncmp(args[0].from, "lzo", 3) == 0) { 427 compress_type = "lzo"; 428 info->compress_type = BTRFS_COMPRESS_LZO; 429 info->compress_level = 0; 430 btrfs_set_opt(info->mount_opt, COMPRESS); 431 btrfs_clear_opt(info->mount_opt, NODATACOW); 432 btrfs_clear_opt(info->mount_opt, NODATASUM); 433 btrfs_set_fs_incompat(info, COMPRESS_LZO); 434 no_compress = 0; 435 } else if (strncmp(args[0].from, "zstd", 4) == 0) { 436 compress_type = "zstd"; 437 info->compress_type = BTRFS_COMPRESS_ZSTD; 438 info->compress_level = 439 btrfs_compress_str2level( 440 BTRFS_COMPRESS_ZSTD, 441 args[0].from + 4); 442 btrfs_set_opt(info->mount_opt, COMPRESS); 443 btrfs_clear_opt(info->mount_opt, NODATACOW); 444 btrfs_clear_opt(info->mount_opt, NODATASUM); 445 btrfs_set_fs_incompat(info, COMPRESS_ZSTD); 446 no_compress = 0; 447 } else if (strncmp(args[0].from, "no", 2) == 0) { 448 compress_type = "no"; 449 info->compress_level = 0; 450 info->compress_type = 0; 451 btrfs_clear_opt(info->mount_opt, COMPRESS); 452 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 453 compress_force = false; 454 no_compress++; 455 } else { 456 btrfs_err(info, "unrecognized compression value %s", 457 args[0].from); 458 ret = -EINVAL; 459 goto out; 460 } 461 462 if (compress_force) { 463 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 464 } else { 465 /* 466 * If we remount from compress-force=xxx to 467 * compress=xxx, we need clear FORCE_COMPRESS 468 * flag, otherwise, there is no way for users 469 * to disable forcible compression separately. 470 */ 471 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 472 } 473 if (no_compress == 1) { 474 btrfs_info(info, "use no compression"); 475 } else if ((info->compress_type != saved_compress_type) || 476 (compress_force != saved_compress_force) || 477 (info->compress_level != saved_compress_level)) { 478 btrfs_info(info, "%s %s compression, level %d", 479 (compress_force) ? "force" : "use", 480 compress_type, info->compress_level); 481 } 482 compress_force = false; 483 break; 484 case Opt_ssd: 485 btrfs_set_and_info(info, SSD, 486 "enabling ssd optimizations"); 487 btrfs_clear_opt(info->mount_opt, NOSSD); 488 break; 489 case Opt_ssd_spread: 490 btrfs_set_and_info(info, SSD, 491 "enabling ssd optimizations"); 492 btrfs_set_and_info(info, SSD_SPREAD, 493 "using spread ssd allocation scheme"); 494 btrfs_clear_opt(info->mount_opt, NOSSD); 495 break; 496 case Opt_nossd: 497 btrfs_set_opt(info->mount_opt, NOSSD); 498 btrfs_clear_and_info(info, SSD, 499 "not using ssd optimizations"); 500 fallthrough; 501 case Opt_nossd_spread: 502 btrfs_clear_and_info(info, SSD_SPREAD, 503 "not using spread ssd allocation scheme"); 504 break; 505 case Opt_barrier: 506 btrfs_clear_and_info(info, NOBARRIER, 507 "turning on barriers"); 508 break; 509 case Opt_nobarrier: 510 btrfs_set_and_info(info, NOBARRIER, 511 "turning off barriers"); 512 break; 513 case Opt_thread_pool: 514 ret = match_int(&args[0], &intarg); 515 if (ret) { 516 btrfs_err(info, "unrecognized thread_pool value %s", 517 args[0].from); 518 goto out; 519 } else if (intarg == 0) { 520 btrfs_err(info, "invalid value 0 for thread_pool"); 521 ret = -EINVAL; 522 goto out; 523 } 524 info->thread_pool_size = intarg; 525 break; 526 case Opt_max_inline: 527 num = match_strdup(&args[0]); 528 if (num) { 529 info->max_inline = memparse(num, NULL); 530 kfree(num); 531 532 if (info->max_inline) { 533 info->max_inline = min_t(u64, 534 info->max_inline, 535 info->sectorsize); 536 } 537 btrfs_info(info, "max_inline at %llu", 538 info->max_inline); 539 } else { 540 ret = -ENOMEM; 541 goto out; 542 } 543 break; 544 case Opt_acl: 545 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 546 info->sb->s_flags |= SB_POSIXACL; 547 break; 548 #else 549 btrfs_err(info, "support for ACL not compiled in!"); 550 ret = -EINVAL; 551 goto out; 552 #endif 553 case Opt_noacl: 554 info->sb->s_flags &= ~SB_POSIXACL; 555 break; 556 case Opt_notreelog: 557 btrfs_set_and_info(info, NOTREELOG, 558 "disabling tree log"); 559 break; 560 case Opt_treelog: 561 btrfs_clear_and_info(info, NOTREELOG, 562 "enabling tree log"); 563 break; 564 case Opt_norecovery: 565 case Opt_nologreplay: 566 btrfs_warn(info, 567 "'nologreplay' is deprecated, use 'rescue=nologreplay' instead"); 568 btrfs_set_and_info(info, NOLOGREPLAY, 569 "disabling log replay at mount time"); 570 break; 571 case Opt_flushoncommit: 572 btrfs_set_and_info(info, FLUSHONCOMMIT, 573 "turning on flush-on-commit"); 574 break; 575 case Opt_noflushoncommit: 576 btrfs_clear_and_info(info, FLUSHONCOMMIT, 577 "turning off flush-on-commit"); 578 break; 579 case Opt_ratio: 580 ret = match_int(&args[0], &intarg); 581 if (ret) { 582 btrfs_err(info, "unrecognized metadata_ratio value %s", 583 args[0].from); 584 goto out; 585 } 586 info->metadata_ratio = intarg; 587 btrfs_info(info, "metadata ratio %u", 588 info->metadata_ratio); 589 break; 590 case Opt_discard: 591 case Opt_discard_mode: 592 if (token == Opt_discard || 593 strcmp(args[0].from, "sync") == 0) { 594 btrfs_clear_opt(info->mount_opt, DISCARD_ASYNC); 595 btrfs_set_and_info(info, DISCARD_SYNC, 596 "turning on sync discard"); 597 } else if (strcmp(args[0].from, "async") == 0) { 598 btrfs_clear_opt(info->mount_opt, DISCARD_SYNC); 599 btrfs_set_and_info(info, DISCARD_ASYNC, 600 "turning on async discard"); 601 } else { 602 btrfs_err(info, "unrecognized discard mode value %s", 603 args[0].from); 604 ret = -EINVAL; 605 goto out; 606 } 607 btrfs_clear_opt(info->mount_opt, NODISCARD); 608 break; 609 case Opt_nodiscard: 610 btrfs_clear_and_info(info, DISCARD_SYNC, 611 "turning off discard"); 612 btrfs_clear_and_info(info, DISCARD_ASYNC, 613 "turning off async discard"); 614 btrfs_set_opt(info->mount_opt, NODISCARD); 615 break; 616 case Opt_space_cache: 617 case Opt_space_cache_version: 618 /* 619 * We already set FREE_SPACE_TREE above because we have 620 * compat_ro(FREE_SPACE_TREE) set, and we aren't going 621 * to allow v1 to be set for extent tree v2, simply 622 * ignore this setting if we're extent tree v2. 623 */ 624 if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) 625 break; 626 if (token == Opt_space_cache || 627 strcmp(args[0].from, "v1") == 0) { 628 btrfs_clear_opt(info->mount_opt, 629 FREE_SPACE_TREE); 630 btrfs_set_and_info(info, SPACE_CACHE, 631 "enabling disk space caching"); 632 } else if (strcmp(args[0].from, "v2") == 0) { 633 btrfs_clear_opt(info->mount_opt, 634 SPACE_CACHE); 635 btrfs_set_and_info(info, FREE_SPACE_TREE, 636 "enabling free space tree"); 637 } else { 638 btrfs_err(info, "unrecognized space_cache value %s", 639 args[0].from); 640 ret = -EINVAL; 641 goto out; 642 } 643 break; 644 case Opt_rescan_uuid_tree: 645 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE); 646 break; 647 case Opt_no_space_cache: 648 /* 649 * We cannot operate without the free space tree with 650 * extent tree v2, ignore this option. 651 */ 652 if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) 653 break; 654 if (btrfs_test_opt(info, SPACE_CACHE)) { 655 btrfs_clear_and_info(info, SPACE_CACHE, 656 "disabling disk space caching"); 657 } 658 if (btrfs_test_opt(info, FREE_SPACE_TREE)) { 659 btrfs_clear_and_info(info, FREE_SPACE_TREE, 660 "disabling free space tree"); 661 } 662 break; 663 case Opt_inode_cache: 664 case Opt_noinode_cache: 665 btrfs_warn(info, 666 "the 'inode_cache' option is deprecated and has no effect since 5.11"); 667 break; 668 case Opt_clear_cache: 669 /* 670 * We cannot clear the free space tree with extent tree 671 * v2, ignore this option. 672 */ 673 if (btrfs_fs_incompat(info, EXTENT_TREE_V2)) 674 break; 675 btrfs_set_and_info(info, CLEAR_CACHE, 676 "force clearing of disk cache"); 677 break; 678 case Opt_user_subvol_rm_allowed: 679 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 680 break; 681 case Opt_enospc_debug: 682 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); 683 break; 684 case Opt_noenospc_debug: 685 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG); 686 break; 687 case Opt_defrag: 688 btrfs_set_and_info(info, AUTO_DEFRAG, 689 "enabling auto defrag"); 690 break; 691 case Opt_nodefrag: 692 btrfs_clear_and_info(info, AUTO_DEFRAG, 693 "disabling auto defrag"); 694 break; 695 case Opt_recovery: 696 case Opt_usebackuproot: 697 btrfs_warn(info, 698 "'%s' is deprecated, use 'rescue=usebackuproot' instead", 699 token == Opt_recovery ? "recovery" : 700 "usebackuproot"); 701 btrfs_info(info, 702 "trying to use backup root at mount time"); 703 btrfs_set_opt(info->mount_opt, USEBACKUPROOT); 704 break; 705 case Opt_skip_balance: 706 btrfs_set_opt(info->mount_opt, SKIP_BALANCE); 707 break; 708 case Opt_fatal_errors: 709 if (strcmp(args[0].from, "panic") == 0) { 710 btrfs_set_opt(info->mount_opt, 711 PANIC_ON_FATAL_ERROR); 712 } else if (strcmp(args[0].from, "bug") == 0) { 713 btrfs_clear_opt(info->mount_opt, 714 PANIC_ON_FATAL_ERROR); 715 } else { 716 btrfs_err(info, "unrecognized fatal_errors value %s", 717 args[0].from); 718 ret = -EINVAL; 719 goto out; 720 } 721 break; 722 case Opt_commit_interval: 723 intarg = 0; 724 ret = match_int(&args[0], &intarg); 725 if (ret) { 726 btrfs_err(info, "unrecognized commit_interval value %s", 727 args[0].from); 728 ret = -EINVAL; 729 goto out; 730 } 731 if (intarg == 0) { 732 btrfs_info(info, 733 "using default commit interval %us", 734 BTRFS_DEFAULT_COMMIT_INTERVAL); 735 intarg = BTRFS_DEFAULT_COMMIT_INTERVAL; 736 } else if (intarg > 300) { 737 btrfs_warn(info, "excessive commit interval %d", 738 intarg); 739 } 740 info->commit_interval = intarg; 741 break; 742 case Opt_rescue: 743 ret = parse_rescue_options(info, args[0].from); 744 if (ret < 0) { 745 btrfs_err(info, "unrecognized rescue value %s", 746 args[0].from); 747 goto out; 748 } 749 break; 750 #ifdef CONFIG_BTRFS_DEBUG 751 case Opt_fragment_all: 752 btrfs_info(info, "fragmenting all space"); 753 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA); 754 btrfs_set_opt(info->mount_opt, FRAGMENT_METADATA); 755 break; 756 case Opt_fragment_metadata: 757 btrfs_info(info, "fragmenting metadata"); 758 btrfs_set_opt(info->mount_opt, 759 FRAGMENT_METADATA); 760 break; 761 case Opt_fragment_data: 762 btrfs_info(info, "fragmenting data"); 763 btrfs_set_opt(info->mount_opt, FRAGMENT_DATA); 764 break; 765 #endif 766 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 767 case Opt_ref_verify: 768 btrfs_info(info, "doing ref verification"); 769 btrfs_set_opt(info->mount_opt, REF_VERIFY); 770 break; 771 #endif 772 case Opt_err: 773 btrfs_err(info, "unrecognized mount option '%s'", p); 774 ret = -EINVAL; 775 goto out; 776 default: 777 break; 778 } 779 } 780 check: 781 /* We're read-only, don't have to check. */ 782 if (new_flags & SB_RDONLY) 783 goto out; 784 785 if (check_ro_option(info, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") || 786 check_ro_option(info, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") || 787 check_ro_option(info, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums")) 788 ret = -EINVAL; 789 out: 790 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) && 791 !btrfs_test_opt(info, FREE_SPACE_TREE) && 792 !btrfs_test_opt(info, CLEAR_CACHE)) { 793 btrfs_err(info, "cannot disable free space tree"); 794 ret = -EINVAL; 795 } 796 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) && 797 !btrfs_test_opt(info, FREE_SPACE_TREE)) { 798 btrfs_err(info, "cannot disable free space tree with block-group-tree feature"); 799 ret = -EINVAL; 800 } 801 if (!ret) 802 ret = btrfs_check_mountopts_zoned(info); 803 if (!ret && !remounting) { 804 if (btrfs_test_opt(info, SPACE_CACHE)) 805 btrfs_info(info, "disk space caching is enabled"); 806 if (btrfs_test_opt(info, FREE_SPACE_TREE)) 807 btrfs_info(info, "using free space tree"); 808 } 809 return ret; 810 } 811 812 /* 813 * Parse mount options that are required early in the mount process. 814 * 815 * All other options will be parsed on much later in the mount process and 816 * only when we need to allocate a new super block. 817 */ 818 static int btrfs_parse_device_options(const char *options, blk_mode_t flags) 819 { 820 substring_t args[MAX_OPT_ARGS]; 821 char *device_name, *opts, *orig, *p; 822 struct btrfs_device *device = NULL; 823 int error = 0; 824 825 lockdep_assert_held(&uuid_mutex); 826 827 if (!options) 828 return 0; 829 830 /* 831 * strsep changes the string, duplicate it because btrfs_parse_options 832 * gets called later 833 */ 834 opts = kstrdup(options, GFP_KERNEL); 835 if (!opts) 836 return -ENOMEM; 837 orig = opts; 838 839 while ((p = strsep(&opts, ",")) != NULL) { 840 int token; 841 842 if (!*p) 843 continue; 844 845 token = match_token(p, tokens, args); 846 if (token == Opt_device) { 847 device_name = match_strdup(&args[0]); 848 if (!device_name) { 849 error = -ENOMEM; 850 goto out; 851 } 852 device = btrfs_scan_one_device(device_name, flags, false); 853 kfree(device_name); 854 if (IS_ERR(device)) { 855 error = PTR_ERR(device); 856 goto out; 857 } 858 } 859 } 860 861 out: 862 kfree(orig); 863 return error; 864 } 865 866 /* 867 * Parse mount options that are related to subvolume id 868 * 869 * The value is later passed to mount_subvol() 870 */ 871 static int btrfs_parse_subvol_options(const char *options, char **subvol_name, 872 u64 *subvol_objectid) 873 { 874 substring_t args[MAX_OPT_ARGS]; 875 char *opts, *orig, *p; 876 int error = 0; 877 u64 subvolid; 878 879 if (!options) 880 return 0; 881 882 /* 883 * strsep changes the string, duplicate it because 884 * btrfs_parse_device_options gets called later 885 */ 886 opts = kstrdup(options, GFP_KERNEL); 887 if (!opts) 888 return -ENOMEM; 889 orig = opts; 890 891 while ((p = strsep(&opts, ",")) != NULL) { 892 int token; 893 if (!*p) 894 continue; 895 896 token = match_token(p, tokens, args); 897 switch (token) { 898 case Opt_subvol: 899 kfree(*subvol_name); 900 *subvol_name = match_strdup(&args[0]); 901 if (!*subvol_name) { 902 error = -ENOMEM; 903 goto out; 904 } 905 break; 906 case Opt_subvolid: 907 error = match_u64(&args[0], &subvolid); 908 if (error) 909 goto out; 910 911 /* we want the original fs_tree */ 912 if (subvolid == 0) 913 subvolid = BTRFS_FS_TREE_OBJECTID; 914 915 *subvol_objectid = subvolid; 916 break; 917 default: 918 break; 919 } 920 } 921 922 out: 923 kfree(orig); 924 return error; 925 } 926 927 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info, 928 u64 subvol_objectid) 929 { 930 struct btrfs_root *root = fs_info->tree_root; 931 struct btrfs_root *fs_root = NULL; 932 struct btrfs_root_ref *root_ref; 933 struct btrfs_inode_ref *inode_ref; 934 struct btrfs_key key; 935 struct btrfs_path *path = NULL; 936 char *name = NULL, *ptr; 937 u64 dirid; 938 int len; 939 int ret; 940 941 path = btrfs_alloc_path(); 942 if (!path) { 943 ret = -ENOMEM; 944 goto err; 945 } 946 947 name = kmalloc(PATH_MAX, GFP_KERNEL); 948 if (!name) { 949 ret = -ENOMEM; 950 goto err; 951 } 952 ptr = name + PATH_MAX - 1; 953 ptr[0] = '\0'; 954 955 /* 956 * Walk up the subvolume trees in the tree of tree roots by root 957 * backrefs until we hit the top-level subvolume. 958 */ 959 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) { 960 key.objectid = subvol_objectid; 961 key.type = BTRFS_ROOT_BACKREF_KEY; 962 key.offset = (u64)-1; 963 964 ret = btrfs_search_backwards(root, &key, path); 965 if (ret < 0) { 966 goto err; 967 } else if (ret > 0) { 968 ret = -ENOENT; 969 goto err; 970 } 971 972 subvol_objectid = key.offset; 973 974 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 975 struct btrfs_root_ref); 976 len = btrfs_root_ref_name_len(path->nodes[0], root_ref); 977 ptr -= len + 1; 978 if (ptr < name) { 979 ret = -ENAMETOOLONG; 980 goto err; 981 } 982 read_extent_buffer(path->nodes[0], ptr + 1, 983 (unsigned long)(root_ref + 1), len); 984 ptr[0] = '/'; 985 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref); 986 btrfs_release_path(path); 987 988 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true); 989 if (IS_ERR(fs_root)) { 990 ret = PTR_ERR(fs_root); 991 fs_root = NULL; 992 goto err; 993 } 994 995 /* 996 * Walk up the filesystem tree by inode refs until we hit the 997 * root directory. 998 */ 999 while (dirid != BTRFS_FIRST_FREE_OBJECTID) { 1000 key.objectid = dirid; 1001 key.type = BTRFS_INODE_REF_KEY; 1002 key.offset = (u64)-1; 1003 1004 ret = btrfs_search_backwards(fs_root, &key, path); 1005 if (ret < 0) { 1006 goto err; 1007 } else if (ret > 0) { 1008 ret = -ENOENT; 1009 goto err; 1010 } 1011 1012 dirid = key.offset; 1013 1014 inode_ref = btrfs_item_ptr(path->nodes[0], 1015 path->slots[0], 1016 struct btrfs_inode_ref); 1017 len = btrfs_inode_ref_name_len(path->nodes[0], 1018 inode_ref); 1019 ptr -= len + 1; 1020 if (ptr < name) { 1021 ret = -ENAMETOOLONG; 1022 goto err; 1023 } 1024 read_extent_buffer(path->nodes[0], ptr + 1, 1025 (unsigned long)(inode_ref + 1), len); 1026 ptr[0] = '/'; 1027 btrfs_release_path(path); 1028 } 1029 btrfs_put_root(fs_root); 1030 fs_root = NULL; 1031 } 1032 1033 btrfs_free_path(path); 1034 if (ptr == name + PATH_MAX - 1) { 1035 name[0] = '/'; 1036 name[1] = '\0'; 1037 } else { 1038 memmove(name, ptr, name + PATH_MAX - ptr); 1039 } 1040 return name; 1041 1042 err: 1043 btrfs_put_root(fs_root); 1044 btrfs_free_path(path); 1045 kfree(name); 1046 return ERR_PTR(ret); 1047 } 1048 1049 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid) 1050 { 1051 struct btrfs_root *root = fs_info->tree_root; 1052 struct btrfs_dir_item *di; 1053 struct btrfs_path *path; 1054 struct btrfs_key location; 1055 struct fscrypt_str name = FSTR_INIT("default", 7); 1056 u64 dir_id; 1057 1058 path = btrfs_alloc_path(); 1059 if (!path) 1060 return -ENOMEM; 1061 1062 /* 1063 * Find the "default" dir item which points to the root item that we 1064 * will mount by default if we haven't been given a specific subvolume 1065 * to mount. 1066 */ 1067 dir_id = btrfs_super_root_dir(fs_info->super_copy); 1068 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0); 1069 if (IS_ERR(di)) { 1070 btrfs_free_path(path); 1071 return PTR_ERR(di); 1072 } 1073 if (!di) { 1074 /* 1075 * Ok the default dir item isn't there. This is weird since 1076 * it's always been there, but don't freak out, just try and 1077 * mount the top-level subvolume. 1078 */ 1079 btrfs_free_path(path); 1080 *objectid = BTRFS_FS_TREE_OBJECTID; 1081 return 0; 1082 } 1083 1084 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 1085 btrfs_free_path(path); 1086 *objectid = location.objectid; 1087 return 0; 1088 } 1089 1090 static int btrfs_fill_super(struct super_block *sb, 1091 struct btrfs_fs_devices *fs_devices, 1092 void *data) 1093 { 1094 struct inode *inode; 1095 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1096 int err; 1097 1098 sb->s_maxbytes = MAX_LFS_FILESIZE; 1099 sb->s_magic = BTRFS_SUPER_MAGIC; 1100 sb->s_op = &btrfs_super_ops; 1101 sb->s_d_op = &btrfs_dentry_operations; 1102 sb->s_export_op = &btrfs_export_ops; 1103 #ifdef CONFIG_FS_VERITY 1104 sb->s_vop = &btrfs_verityops; 1105 #endif 1106 sb->s_xattr = btrfs_xattr_handlers; 1107 sb->s_time_gran = 1; 1108 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 1109 sb->s_flags |= SB_POSIXACL; 1110 #endif 1111 sb->s_flags |= SB_I_VERSION; 1112 sb->s_iflags |= SB_I_CGROUPWB; 1113 1114 err = super_setup_bdi(sb); 1115 if (err) { 1116 btrfs_err(fs_info, "super_setup_bdi failed"); 1117 return err; 1118 } 1119 1120 err = open_ctree(sb, fs_devices, (char *)data); 1121 if (err) { 1122 btrfs_err(fs_info, "open_ctree failed"); 1123 return err; 1124 } 1125 1126 inode = btrfs_iget(sb, BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root); 1127 if (IS_ERR(inode)) { 1128 err = PTR_ERR(inode); 1129 btrfs_handle_fs_error(fs_info, err, NULL); 1130 goto fail_close; 1131 } 1132 1133 sb->s_root = d_make_root(inode); 1134 if (!sb->s_root) { 1135 err = -ENOMEM; 1136 goto fail_close; 1137 } 1138 1139 sb->s_flags |= SB_ACTIVE; 1140 return 0; 1141 1142 fail_close: 1143 close_ctree(fs_info); 1144 return err; 1145 } 1146 1147 int btrfs_sync_fs(struct super_block *sb, int wait) 1148 { 1149 struct btrfs_trans_handle *trans; 1150 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1151 struct btrfs_root *root = fs_info->tree_root; 1152 1153 trace_btrfs_sync_fs(fs_info, wait); 1154 1155 if (!wait) { 1156 filemap_flush(fs_info->btree_inode->i_mapping); 1157 return 0; 1158 } 1159 1160 btrfs_wait_ordered_roots(fs_info, U64_MAX, 0, (u64)-1); 1161 1162 trans = btrfs_attach_transaction_barrier(root); 1163 if (IS_ERR(trans)) { 1164 /* no transaction, don't bother */ 1165 if (PTR_ERR(trans) == -ENOENT) { 1166 /* 1167 * Exit unless we have some pending changes 1168 * that need to go through commit 1169 */ 1170 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT, 1171 &fs_info->flags)) 1172 return 0; 1173 /* 1174 * A non-blocking test if the fs is frozen. We must not 1175 * start a new transaction here otherwise a deadlock 1176 * happens. The pending operations are delayed to the 1177 * next commit after thawing. 1178 */ 1179 if (sb_start_write_trylock(sb)) 1180 sb_end_write(sb); 1181 else 1182 return 0; 1183 trans = btrfs_start_transaction(root, 0); 1184 } 1185 if (IS_ERR(trans)) 1186 return PTR_ERR(trans); 1187 } 1188 return btrfs_commit_transaction(trans); 1189 } 1190 1191 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed) 1192 { 1193 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s); 1194 *printed = true; 1195 } 1196 1197 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 1198 { 1199 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 1200 const char *compress_type; 1201 const char *subvol_name; 1202 bool printed = false; 1203 1204 if (btrfs_test_opt(info, DEGRADED)) 1205 seq_puts(seq, ",degraded"); 1206 if (btrfs_test_opt(info, NODATASUM)) 1207 seq_puts(seq, ",nodatasum"); 1208 if (btrfs_test_opt(info, NODATACOW)) 1209 seq_puts(seq, ",nodatacow"); 1210 if (btrfs_test_opt(info, NOBARRIER)) 1211 seq_puts(seq, ",nobarrier"); 1212 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) 1213 seq_printf(seq, ",max_inline=%llu", info->max_inline); 1214 if (info->thread_pool_size != min_t(unsigned long, 1215 num_online_cpus() + 2, 8)) 1216 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size); 1217 if (btrfs_test_opt(info, COMPRESS)) { 1218 compress_type = btrfs_compress_type2str(info->compress_type); 1219 if (btrfs_test_opt(info, FORCE_COMPRESS)) 1220 seq_printf(seq, ",compress-force=%s", compress_type); 1221 else 1222 seq_printf(seq, ",compress=%s", compress_type); 1223 if (info->compress_level) 1224 seq_printf(seq, ":%d", info->compress_level); 1225 } 1226 if (btrfs_test_opt(info, NOSSD)) 1227 seq_puts(seq, ",nossd"); 1228 if (btrfs_test_opt(info, SSD_SPREAD)) 1229 seq_puts(seq, ",ssd_spread"); 1230 else if (btrfs_test_opt(info, SSD)) 1231 seq_puts(seq, ",ssd"); 1232 if (btrfs_test_opt(info, NOTREELOG)) 1233 seq_puts(seq, ",notreelog"); 1234 if (btrfs_test_opt(info, NOLOGREPLAY)) 1235 print_rescue_option(seq, "nologreplay", &printed); 1236 if (btrfs_test_opt(info, USEBACKUPROOT)) 1237 print_rescue_option(seq, "usebackuproot", &printed); 1238 if (btrfs_test_opt(info, IGNOREBADROOTS)) 1239 print_rescue_option(seq, "ignorebadroots", &printed); 1240 if (btrfs_test_opt(info, IGNOREDATACSUMS)) 1241 print_rescue_option(seq, "ignoredatacsums", &printed); 1242 if (btrfs_test_opt(info, FLUSHONCOMMIT)) 1243 seq_puts(seq, ",flushoncommit"); 1244 if (btrfs_test_opt(info, DISCARD_SYNC)) 1245 seq_puts(seq, ",discard"); 1246 if (btrfs_test_opt(info, DISCARD_ASYNC)) 1247 seq_puts(seq, ",discard=async"); 1248 if (!(info->sb->s_flags & SB_POSIXACL)) 1249 seq_puts(seq, ",noacl"); 1250 if (btrfs_free_space_cache_v1_active(info)) 1251 seq_puts(seq, ",space_cache"); 1252 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE)) 1253 seq_puts(seq, ",space_cache=v2"); 1254 else 1255 seq_puts(seq, ",nospace_cache"); 1256 if (btrfs_test_opt(info, RESCAN_UUID_TREE)) 1257 seq_puts(seq, ",rescan_uuid_tree"); 1258 if (btrfs_test_opt(info, CLEAR_CACHE)) 1259 seq_puts(seq, ",clear_cache"); 1260 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED)) 1261 seq_puts(seq, ",user_subvol_rm_allowed"); 1262 if (btrfs_test_opt(info, ENOSPC_DEBUG)) 1263 seq_puts(seq, ",enospc_debug"); 1264 if (btrfs_test_opt(info, AUTO_DEFRAG)) 1265 seq_puts(seq, ",autodefrag"); 1266 if (btrfs_test_opt(info, SKIP_BALANCE)) 1267 seq_puts(seq, ",skip_balance"); 1268 if (info->metadata_ratio) 1269 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio); 1270 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR)) 1271 seq_puts(seq, ",fatal_errors=panic"); 1272 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1273 seq_printf(seq, ",commit=%u", info->commit_interval); 1274 #ifdef CONFIG_BTRFS_DEBUG 1275 if (btrfs_test_opt(info, FRAGMENT_DATA)) 1276 seq_puts(seq, ",fragment=data"); 1277 if (btrfs_test_opt(info, FRAGMENT_METADATA)) 1278 seq_puts(seq, ",fragment=metadata"); 1279 #endif 1280 if (btrfs_test_opt(info, REF_VERIFY)) 1281 seq_puts(seq, ",ref_verify"); 1282 seq_printf(seq, ",subvolid=%llu", 1283 BTRFS_I(d_inode(dentry))->root->root_key.objectid); 1284 subvol_name = btrfs_get_subvol_name_from_objectid(info, 1285 BTRFS_I(d_inode(dentry))->root->root_key.objectid); 1286 if (!IS_ERR(subvol_name)) { 1287 seq_puts(seq, ",subvol="); 1288 seq_escape(seq, subvol_name, " \t\n\\"); 1289 kfree(subvol_name); 1290 } 1291 return 0; 1292 } 1293 1294 static int btrfs_test_super(struct super_block *s, void *data) 1295 { 1296 struct btrfs_fs_info *p = data; 1297 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1298 1299 return fs_info->fs_devices == p->fs_devices; 1300 } 1301 1302 static int btrfs_set_super(struct super_block *s, void *data) 1303 { 1304 int err = set_anon_super(s, data); 1305 if (!err) 1306 s->s_fs_info = data; 1307 return err; 1308 } 1309 1310 /* 1311 * subvolumes are identified by ino 256 1312 */ 1313 static inline int is_subvolume_inode(struct inode *inode) 1314 { 1315 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1316 return 1; 1317 return 0; 1318 } 1319 1320 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid, 1321 struct vfsmount *mnt) 1322 { 1323 struct dentry *root; 1324 int ret; 1325 1326 if (!subvol_name) { 1327 if (!subvol_objectid) { 1328 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb), 1329 &subvol_objectid); 1330 if (ret) { 1331 root = ERR_PTR(ret); 1332 goto out; 1333 } 1334 } 1335 subvol_name = btrfs_get_subvol_name_from_objectid( 1336 btrfs_sb(mnt->mnt_sb), subvol_objectid); 1337 if (IS_ERR(subvol_name)) { 1338 root = ERR_CAST(subvol_name); 1339 subvol_name = NULL; 1340 goto out; 1341 } 1342 1343 } 1344 1345 root = mount_subtree(mnt, subvol_name); 1346 /* mount_subtree() drops our reference on the vfsmount. */ 1347 mnt = NULL; 1348 1349 if (!IS_ERR(root)) { 1350 struct super_block *s = root->d_sb; 1351 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1352 struct inode *root_inode = d_inode(root); 1353 u64 root_objectid = BTRFS_I(root_inode)->root->root_key.objectid; 1354 1355 ret = 0; 1356 if (!is_subvolume_inode(root_inode)) { 1357 btrfs_err(fs_info, "'%s' is not a valid subvolume", 1358 subvol_name); 1359 ret = -EINVAL; 1360 } 1361 if (subvol_objectid && root_objectid != subvol_objectid) { 1362 /* 1363 * This will also catch a race condition where a 1364 * subvolume which was passed by ID is renamed and 1365 * another subvolume is renamed over the old location. 1366 */ 1367 btrfs_err(fs_info, 1368 "subvol '%s' does not match subvolid %llu", 1369 subvol_name, subvol_objectid); 1370 ret = -EINVAL; 1371 } 1372 if (ret) { 1373 dput(root); 1374 root = ERR_PTR(ret); 1375 deactivate_locked_super(s); 1376 } 1377 } 1378 1379 out: 1380 mntput(mnt); 1381 kfree(subvol_name); 1382 return root; 1383 } 1384 1385 /* 1386 * Find a superblock for the given device / mount point. 1387 * 1388 * Note: This is based on mount_bdev from fs/super.c with a few additions 1389 * for multiple device setup. Make sure to keep it in sync. 1390 */ 1391 static struct dentry *btrfs_mount_root(struct file_system_type *fs_type, 1392 int flags, const char *device_name, void *data) 1393 { 1394 struct block_device *bdev = NULL; 1395 struct super_block *s; 1396 struct btrfs_device *device = NULL; 1397 struct btrfs_fs_devices *fs_devices = NULL; 1398 struct btrfs_fs_info *fs_info = NULL; 1399 void *new_sec_opts = NULL; 1400 blk_mode_t mode = sb_open_mode(flags); 1401 int error = 0; 1402 1403 if (data) { 1404 error = security_sb_eat_lsm_opts(data, &new_sec_opts); 1405 if (error) 1406 return ERR_PTR(error); 1407 } 1408 1409 /* 1410 * Setup a dummy root and fs_info for test/set super. This is because 1411 * we don't actually fill this stuff out until open_ctree, but we need 1412 * then open_ctree will properly initialize the file system specific 1413 * settings later. btrfs_init_fs_info initializes the static elements 1414 * of the fs_info (locks and such) to make cleanup easier if we find a 1415 * superblock with our given fs_devices later on at sget() time. 1416 */ 1417 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL); 1418 if (!fs_info) { 1419 error = -ENOMEM; 1420 goto error_sec_opts; 1421 } 1422 btrfs_init_fs_info(fs_info); 1423 1424 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 1425 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 1426 if (!fs_info->super_copy || !fs_info->super_for_commit) { 1427 error = -ENOMEM; 1428 goto error_fs_info; 1429 } 1430 1431 mutex_lock(&uuid_mutex); 1432 error = btrfs_parse_device_options(data, mode); 1433 if (error) { 1434 mutex_unlock(&uuid_mutex); 1435 goto error_fs_info; 1436 } 1437 1438 /* 1439 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect 1440 * either a valid device or an error. 1441 */ 1442 device = btrfs_scan_one_device(device_name, mode, true); 1443 ASSERT(device != NULL); 1444 if (IS_ERR(device)) { 1445 mutex_unlock(&uuid_mutex); 1446 error = PTR_ERR(device); 1447 goto error_fs_info; 1448 } 1449 1450 fs_devices = device->fs_devices; 1451 fs_info->fs_devices = fs_devices; 1452 1453 error = btrfs_open_devices(fs_devices, mode, fs_type); 1454 mutex_unlock(&uuid_mutex); 1455 if (error) 1456 goto error_fs_info; 1457 1458 if (!(flags & SB_RDONLY) && fs_devices->rw_devices == 0) { 1459 error = -EACCES; 1460 goto error_close_devices; 1461 } 1462 1463 bdev = fs_devices->latest_dev->bdev; 1464 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | SB_NOSEC, 1465 fs_info); 1466 if (IS_ERR(s)) { 1467 error = PTR_ERR(s); 1468 goto error_close_devices; 1469 } 1470 1471 if (s->s_root) { 1472 btrfs_close_devices(fs_devices); 1473 btrfs_free_fs_info(fs_info); 1474 if ((flags ^ s->s_flags) & SB_RDONLY) 1475 error = -EBUSY; 1476 } else { 1477 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev); 1478 shrinker_debugfs_rename(s->s_shrink, "sb-%s:%s", fs_type->name, 1479 s->s_id); 1480 btrfs_sb(s)->bdev_holder = fs_type; 1481 error = btrfs_fill_super(s, fs_devices, data); 1482 } 1483 if (!error) 1484 error = security_sb_set_mnt_opts(s, new_sec_opts, 0, NULL); 1485 security_free_mnt_opts(&new_sec_opts); 1486 if (error) { 1487 deactivate_locked_super(s); 1488 return ERR_PTR(error); 1489 } 1490 1491 return dget(s->s_root); 1492 1493 error_close_devices: 1494 btrfs_close_devices(fs_devices); 1495 error_fs_info: 1496 btrfs_free_fs_info(fs_info); 1497 error_sec_opts: 1498 security_free_mnt_opts(&new_sec_opts); 1499 return ERR_PTR(error); 1500 } 1501 1502 /* 1503 * Mount function which is called by VFS layer. 1504 * 1505 * In order to allow mounting a subvolume directly, btrfs uses mount_subtree() 1506 * which needs vfsmount* of device's root (/). This means device's root has to 1507 * be mounted internally in any case. 1508 * 1509 * Operation flow: 1510 * 1. Parse subvol id related options for later use in mount_subvol(). 1511 * 1512 * 2. Mount device's root (/) by calling vfs_kern_mount(). 1513 * 1514 * NOTE: vfs_kern_mount() is used by VFS to call btrfs_mount() in the 1515 * first place. In order to avoid calling btrfs_mount() again, we use 1516 * different file_system_type which is not registered to VFS by 1517 * register_filesystem() (btrfs_root_fs_type). As a result, 1518 * btrfs_mount_root() is called. The return value will be used by 1519 * mount_subtree() in mount_subvol(). 1520 * 1521 * 3. Call mount_subvol() to get the dentry of subvolume. Since there is 1522 * "btrfs subvolume set-default", mount_subvol() is called always. 1523 */ 1524 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, 1525 const char *device_name, void *data) 1526 { 1527 struct vfsmount *mnt_root; 1528 struct dentry *root; 1529 char *subvol_name = NULL; 1530 u64 subvol_objectid = 0; 1531 int error = 0; 1532 1533 error = btrfs_parse_subvol_options(data, &subvol_name, 1534 &subvol_objectid); 1535 if (error) { 1536 kfree(subvol_name); 1537 return ERR_PTR(error); 1538 } 1539 1540 /* mount device's root (/) */ 1541 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, flags, device_name, data); 1542 if (PTR_ERR_OR_ZERO(mnt_root) == -EBUSY) { 1543 if (flags & SB_RDONLY) { 1544 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, 1545 flags & ~SB_RDONLY, device_name, data); 1546 } else { 1547 mnt_root = vfs_kern_mount(&btrfs_root_fs_type, 1548 flags | SB_RDONLY, device_name, data); 1549 if (IS_ERR(mnt_root)) { 1550 root = ERR_CAST(mnt_root); 1551 kfree(subvol_name); 1552 goto out; 1553 } 1554 1555 down_write(&mnt_root->mnt_sb->s_umount); 1556 error = btrfs_remount(mnt_root->mnt_sb, &flags, NULL); 1557 up_write(&mnt_root->mnt_sb->s_umount); 1558 if (error < 0) { 1559 root = ERR_PTR(error); 1560 mntput(mnt_root); 1561 kfree(subvol_name); 1562 goto out; 1563 } 1564 } 1565 } 1566 if (IS_ERR(mnt_root)) { 1567 root = ERR_CAST(mnt_root); 1568 kfree(subvol_name); 1569 goto out; 1570 } 1571 1572 /* mount_subvol() will free subvol_name and mnt_root */ 1573 root = mount_subvol(subvol_name, subvol_objectid, mnt_root); 1574 1575 out: 1576 return root; 1577 } 1578 1579 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 1580 u32 new_pool_size, u32 old_pool_size) 1581 { 1582 if (new_pool_size == old_pool_size) 1583 return; 1584 1585 fs_info->thread_pool_size = new_pool_size; 1586 1587 btrfs_info(fs_info, "resize thread pool %d -> %d", 1588 old_pool_size, new_pool_size); 1589 1590 btrfs_workqueue_set_max(fs_info->workers, new_pool_size); 1591 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size); 1592 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size); 1593 workqueue_set_max_active(fs_info->endio_workers, new_pool_size); 1594 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size); 1595 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size); 1596 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size); 1597 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size); 1598 } 1599 1600 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1601 unsigned long old_opts, int flags) 1602 { 1603 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1604 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1605 (flags & SB_RDONLY))) { 1606 /* wait for any defraggers to finish */ 1607 wait_event(fs_info->transaction_wait, 1608 (atomic_read(&fs_info->defrag_running) == 0)); 1609 if (flags & SB_RDONLY) 1610 sync_filesystem(fs_info->sb); 1611 } 1612 } 1613 1614 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1615 unsigned long old_opts) 1616 { 1617 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 1618 1619 /* 1620 * We need to cleanup all defragable inodes if the autodefragment is 1621 * close or the filesystem is read only. 1622 */ 1623 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1624 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) { 1625 btrfs_cleanup_defrag_inodes(fs_info); 1626 } 1627 1628 /* If we toggled discard async */ 1629 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1630 btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1631 btrfs_discard_resume(fs_info); 1632 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1633 !btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1634 btrfs_discard_cleanup(fs_info); 1635 1636 /* If we toggled space cache */ 1637 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) 1638 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 1639 } 1640 1641 static int btrfs_remount(struct super_block *sb, int *flags, char *data) 1642 { 1643 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1644 unsigned old_flags = sb->s_flags; 1645 unsigned long old_opts = fs_info->mount_opt; 1646 unsigned long old_compress_type = fs_info->compress_type; 1647 u64 old_max_inline = fs_info->max_inline; 1648 u32 old_thread_pool_size = fs_info->thread_pool_size; 1649 u32 old_metadata_ratio = fs_info->metadata_ratio; 1650 int ret; 1651 1652 sync_filesystem(sb); 1653 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1654 1655 if (data) { 1656 void *new_sec_opts = NULL; 1657 1658 ret = security_sb_eat_lsm_opts(data, &new_sec_opts); 1659 if (!ret) 1660 ret = security_sb_remount(sb, new_sec_opts); 1661 security_free_mnt_opts(&new_sec_opts); 1662 if (ret) 1663 goto restore; 1664 } 1665 1666 ret = btrfs_parse_options(fs_info, data, *flags); 1667 if (ret) 1668 goto restore; 1669 1670 ret = btrfs_check_features(fs_info, !(*flags & SB_RDONLY)); 1671 if (ret < 0) 1672 goto restore; 1673 1674 btrfs_remount_begin(fs_info, old_opts, *flags); 1675 btrfs_resize_thread_pool(fs_info, 1676 fs_info->thread_pool_size, old_thread_pool_size); 1677 1678 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) != 1679 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 1680 (!sb_rdonly(sb) || (*flags & SB_RDONLY))) { 1681 btrfs_warn(fs_info, 1682 "remount supports changing free space tree only from ro to rw"); 1683 /* Make sure free space cache options match the state on disk */ 1684 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 1685 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1686 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 1687 } 1688 if (btrfs_free_space_cache_v1_active(fs_info)) { 1689 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1690 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); 1691 } 1692 } 1693 1694 if ((bool)(*flags & SB_RDONLY) == sb_rdonly(sb)) 1695 goto out; 1696 1697 if (*flags & SB_RDONLY) { 1698 /* 1699 * this also happens on 'umount -rf' or on shutdown, when 1700 * the filesystem is busy. 1701 */ 1702 cancel_work_sync(&fs_info->async_reclaim_work); 1703 cancel_work_sync(&fs_info->async_data_reclaim_work); 1704 1705 btrfs_discard_cleanup(fs_info); 1706 1707 /* wait for the uuid_scan task to finish */ 1708 down(&fs_info->uuid_tree_rescan_sem); 1709 /* avoid complains from lockdep et al. */ 1710 up(&fs_info->uuid_tree_rescan_sem); 1711 1712 btrfs_set_sb_rdonly(sb); 1713 1714 /* 1715 * Setting SB_RDONLY will put the cleaner thread to 1716 * sleep at the next loop if it's already active. 1717 * If it's already asleep, we'll leave unused block 1718 * groups on disk until we're mounted read-write again 1719 * unless we clean them up here. 1720 */ 1721 btrfs_delete_unused_bgs(fs_info); 1722 1723 /* 1724 * The cleaner task could be already running before we set the 1725 * flag BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). 1726 * We must make sure that after we finish the remount, i.e. after 1727 * we call btrfs_commit_super(), the cleaner can no longer start 1728 * a transaction - either because it was dropping a dead root, 1729 * running delayed iputs or deleting an unused block group (the 1730 * cleaner picked a block group from the list of unused block 1731 * groups before we were able to in the previous call to 1732 * btrfs_delete_unused_bgs()). 1733 */ 1734 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, 1735 TASK_UNINTERRUPTIBLE); 1736 1737 /* 1738 * We've set the superblock to RO mode, so we might have made 1739 * the cleaner task sleep without running all pending delayed 1740 * iputs. Go through all the delayed iputs here, so that if an 1741 * unmount happens without remounting RW we don't end up at 1742 * finishing close_ctree() with a non-empty list of delayed 1743 * iputs. 1744 */ 1745 btrfs_run_delayed_iputs(fs_info); 1746 1747 btrfs_dev_replace_suspend_for_unmount(fs_info); 1748 btrfs_scrub_cancel(fs_info); 1749 btrfs_pause_balance(fs_info); 1750 1751 /* 1752 * Pause the qgroup rescan worker if it is running. We don't want 1753 * it to be still running after we are in RO mode, as after that, 1754 * by the time we unmount, it might have left a transaction open, 1755 * so we would leak the transaction and/or crash. 1756 */ 1757 btrfs_qgroup_wait_for_completion(fs_info, false); 1758 1759 ret = btrfs_commit_super(fs_info); 1760 if (ret) 1761 goto restore; 1762 } else { 1763 if (BTRFS_FS_ERROR(fs_info)) { 1764 btrfs_err(fs_info, 1765 "Remounting read-write after error is not allowed"); 1766 ret = -EINVAL; 1767 goto restore; 1768 } 1769 if (fs_info->fs_devices->rw_devices == 0) { 1770 ret = -EACCES; 1771 goto restore; 1772 } 1773 1774 if (!btrfs_check_rw_degradable(fs_info, NULL)) { 1775 btrfs_warn(fs_info, 1776 "too many missing devices, writable remount is not allowed"); 1777 ret = -EACCES; 1778 goto restore; 1779 } 1780 1781 if (btrfs_super_log_root(fs_info->super_copy) != 0) { 1782 btrfs_warn(fs_info, 1783 "mount required to replay tree-log, cannot remount read-write"); 1784 ret = -EINVAL; 1785 goto restore; 1786 } 1787 1788 /* 1789 * NOTE: when remounting with a change that does writes, don't 1790 * put it anywhere above this point, as we are not sure to be 1791 * safe to write until we pass the above checks. 1792 */ 1793 ret = btrfs_start_pre_rw_mount(fs_info); 1794 if (ret) 1795 goto restore; 1796 1797 btrfs_clear_sb_rdonly(sb); 1798 1799 set_bit(BTRFS_FS_OPEN, &fs_info->flags); 1800 1801 /* 1802 * If we've gone from readonly -> read/write, we need to get 1803 * our sync/async discard lists in the right state. 1804 */ 1805 btrfs_discard_resume(fs_info); 1806 } 1807 out: 1808 /* 1809 * We need to set SB_I_VERSION here otherwise it'll get cleared by VFS, 1810 * since the absence of the flag means it can be toggled off by remount. 1811 */ 1812 *flags |= SB_I_VERSION; 1813 1814 wake_up_process(fs_info->transaction_kthread); 1815 btrfs_remount_cleanup(fs_info, old_opts); 1816 btrfs_clear_oneshot_options(fs_info); 1817 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1818 1819 return 0; 1820 1821 restore: 1822 /* We've hit an error - don't reset SB_RDONLY */ 1823 if (sb_rdonly(sb)) 1824 old_flags |= SB_RDONLY; 1825 if (!(old_flags & SB_RDONLY)) 1826 clear_bit(BTRFS_FS_STATE_RO, &fs_info->fs_state); 1827 sb->s_flags = old_flags; 1828 fs_info->mount_opt = old_opts; 1829 fs_info->compress_type = old_compress_type; 1830 fs_info->max_inline = old_max_inline; 1831 btrfs_resize_thread_pool(fs_info, 1832 old_thread_pool_size, fs_info->thread_pool_size); 1833 fs_info->metadata_ratio = old_metadata_ratio; 1834 btrfs_remount_cleanup(fs_info, old_opts); 1835 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1836 1837 return ret; 1838 } 1839 1840 /* Used to sort the devices by max_avail(descending sort) */ 1841 static int btrfs_cmp_device_free_bytes(const void *a, const void *b) 1842 { 1843 const struct btrfs_device_info *dev_info1 = a; 1844 const struct btrfs_device_info *dev_info2 = b; 1845 1846 if (dev_info1->max_avail > dev_info2->max_avail) 1847 return -1; 1848 else if (dev_info1->max_avail < dev_info2->max_avail) 1849 return 1; 1850 return 0; 1851 } 1852 1853 /* 1854 * sort the devices by max_avail, in which max free extent size of each device 1855 * is stored.(Descending Sort) 1856 */ 1857 static inline void btrfs_descending_sort_devices( 1858 struct btrfs_device_info *devices, 1859 size_t nr_devices) 1860 { 1861 sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1862 btrfs_cmp_device_free_bytes, NULL); 1863 } 1864 1865 /* 1866 * The helper to calc the free space on the devices that can be used to store 1867 * file data. 1868 */ 1869 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info, 1870 u64 *free_bytes) 1871 { 1872 struct btrfs_device_info *devices_info; 1873 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 1874 struct btrfs_device *device; 1875 u64 type; 1876 u64 avail_space; 1877 u64 min_stripe_size; 1878 int num_stripes = 1; 1879 int i = 0, nr_devices; 1880 const struct btrfs_raid_attr *rattr; 1881 1882 /* 1883 * We aren't under the device list lock, so this is racy-ish, but good 1884 * enough for our purposes. 1885 */ 1886 nr_devices = fs_info->fs_devices->open_devices; 1887 if (!nr_devices) { 1888 smp_mb(); 1889 nr_devices = fs_info->fs_devices->open_devices; 1890 ASSERT(nr_devices); 1891 if (!nr_devices) { 1892 *free_bytes = 0; 1893 return 0; 1894 } 1895 } 1896 1897 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), 1898 GFP_KERNEL); 1899 if (!devices_info) 1900 return -ENOMEM; 1901 1902 /* calc min stripe number for data space allocation */ 1903 type = btrfs_data_alloc_profile(fs_info); 1904 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)]; 1905 1906 if (type & BTRFS_BLOCK_GROUP_RAID0) 1907 num_stripes = nr_devices; 1908 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK) 1909 num_stripes = rattr->ncopies; 1910 else if (type & BTRFS_BLOCK_GROUP_RAID10) 1911 num_stripes = 4; 1912 1913 /* Adjust for more than 1 stripe per device */ 1914 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN; 1915 1916 rcu_read_lock(); 1917 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 1918 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, 1919 &device->dev_state) || 1920 !device->bdev || 1921 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) 1922 continue; 1923 1924 if (i >= nr_devices) 1925 break; 1926 1927 avail_space = device->total_bytes - device->bytes_used; 1928 1929 /* align with stripe_len */ 1930 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN); 1931 1932 /* 1933 * Ensure we have at least min_stripe_size on top of the 1934 * reserved space on the device. 1935 */ 1936 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size) 1937 continue; 1938 1939 avail_space -= BTRFS_DEVICE_RANGE_RESERVED; 1940 1941 devices_info[i].dev = device; 1942 devices_info[i].max_avail = avail_space; 1943 1944 i++; 1945 } 1946 rcu_read_unlock(); 1947 1948 nr_devices = i; 1949 1950 btrfs_descending_sort_devices(devices_info, nr_devices); 1951 1952 i = nr_devices - 1; 1953 avail_space = 0; 1954 while (nr_devices >= rattr->devs_min) { 1955 num_stripes = min(num_stripes, nr_devices); 1956 1957 if (devices_info[i].max_avail >= min_stripe_size) { 1958 int j; 1959 u64 alloc_size; 1960 1961 avail_space += devices_info[i].max_avail * num_stripes; 1962 alloc_size = devices_info[i].max_avail; 1963 for (j = i + 1 - num_stripes; j <= i; j++) 1964 devices_info[j].max_avail -= alloc_size; 1965 } 1966 i--; 1967 nr_devices--; 1968 } 1969 1970 kfree(devices_info); 1971 *free_bytes = avail_space; 1972 return 0; 1973 } 1974 1975 /* 1976 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles. 1977 * 1978 * If there's a redundant raid level at DATA block groups, use the respective 1979 * multiplier to scale the sizes. 1980 * 1981 * Unused device space usage is based on simulating the chunk allocator 1982 * algorithm that respects the device sizes and order of allocations. This is 1983 * a close approximation of the actual use but there are other factors that may 1984 * change the result (like a new metadata chunk). 1985 * 1986 * If metadata is exhausted, f_bavail will be 0. 1987 */ 1988 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 1989 { 1990 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1991 struct btrfs_super_block *disk_super = fs_info->super_copy; 1992 struct btrfs_space_info *found; 1993 u64 total_used = 0; 1994 u64 total_free_data = 0; 1995 u64 total_free_meta = 0; 1996 u32 bits = fs_info->sectorsize_bits; 1997 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid; 1998 unsigned factor = 1; 1999 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 2000 int ret; 2001 u64 thresh = 0; 2002 int mixed = 0; 2003 2004 list_for_each_entry(found, &fs_info->space_info, list) { 2005 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 2006 int i; 2007 2008 total_free_data += found->disk_total - found->disk_used; 2009 total_free_data -= 2010 btrfs_account_ro_block_groups_free_space(found); 2011 2012 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { 2013 if (!list_empty(&found->block_groups[i])) 2014 factor = btrfs_bg_type_to_factor( 2015 btrfs_raid_array[i].bg_flag); 2016 } 2017 } 2018 2019 /* 2020 * Metadata in mixed block group profiles are accounted in data 2021 */ 2022 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) { 2023 if (found->flags & BTRFS_BLOCK_GROUP_DATA) 2024 mixed = 1; 2025 else 2026 total_free_meta += found->disk_total - 2027 found->disk_used; 2028 } 2029 2030 total_used += found->disk_used; 2031 } 2032 2033 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor); 2034 buf->f_blocks >>= bits; 2035 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits); 2036 2037 /* Account global block reserve as used, it's in logical size already */ 2038 spin_lock(&block_rsv->lock); 2039 /* Mixed block groups accounting is not byte-accurate, avoid overflow */ 2040 if (buf->f_bfree >= block_rsv->size >> bits) 2041 buf->f_bfree -= block_rsv->size >> bits; 2042 else 2043 buf->f_bfree = 0; 2044 spin_unlock(&block_rsv->lock); 2045 2046 buf->f_bavail = div_u64(total_free_data, factor); 2047 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data); 2048 if (ret) 2049 return ret; 2050 buf->f_bavail += div_u64(total_free_data, factor); 2051 buf->f_bavail = buf->f_bavail >> bits; 2052 2053 /* 2054 * We calculate the remaining metadata space minus global reserve. If 2055 * this is (supposedly) smaller than zero, there's no space. But this 2056 * does not hold in practice, the exhausted state happens where's still 2057 * some positive delta. So we apply some guesswork and compare the 2058 * delta to a 4M threshold. (Practically observed delta was ~2M.) 2059 * 2060 * We probably cannot calculate the exact threshold value because this 2061 * depends on the internal reservations requested by various 2062 * operations, so some operations that consume a few metadata will 2063 * succeed even if the Avail is zero. But this is better than the other 2064 * way around. 2065 */ 2066 thresh = SZ_4M; 2067 2068 /* 2069 * We only want to claim there's no available space if we can no longer 2070 * allocate chunks for our metadata profile and our global reserve will 2071 * not fit in the free metadata space. If we aren't ->full then we 2072 * still can allocate chunks and thus are fine using the currently 2073 * calculated f_bavail. 2074 */ 2075 if (!mixed && block_rsv->space_info->full && 2076 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size)) 2077 buf->f_bavail = 0; 2078 2079 buf->f_type = BTRFS_SUPER_MAGIC; 2080 buf->f_bsize = dentry->d_sb->s_blocksize; 2081 buf->f_namelen = BTRFS_NAME_LEN; 2082 2083 /* We treat it as constant endianness (it doesn't matter _which_) 2084 because we want the fsid to come out the same whether mounted 2085 on a big-endian or little-endian host */ 2086 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 2087 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 2088 /* Mask in the root object ID too, to disambiguate subvols */ 2089 buf->f_fsid.val[0] ^= 2090 BTRFS_I(d_inode(dentry))->root->root_key.objectid >> 32; 2091 buf->f_fsid.val[1] ^= 2092 BTRFS_I(d_inode(dentry))->root->root_key.objectid; 2093 2094 return 0; 2095 } 2096 2097 static void btrfs_kill_super(struct super_block *sb) 2098 { 2099 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2100 kill_anon_super(sb); 2101 btrfs_free_fs_info(fs_info); 2102 } 2103 2104 static struct file_system_type btrfs_fs_type = { 2105 .owner = THIS_MODULE, 2106 .name = "btrfs", 2107 .mount = btrfs_mount, 2108 .kill_sb = btrfs_kill_super, 2109 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA, 2110 }; 2111 2112 static struct file_system_type btrfs_root_fs_type = { 2113 .owner = THIS_MODULE, 2114 .name = "btrfs", 2115 .mount = btrfs_mount_root, 2116 .kill_sb = btrfs_kill_super, 2117 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | FS_ALLOW_IDMAP, 2118 }; 2119 2120 MODULE_ALIAS_FS("btrfs"); 2121 2122 static int btrfs_control_open(struct inode *inode, struct file *file) 2123 { 2124 /* 2125 * The control file's private_data is used to hold the 2126 * transaction when it is started and is used to keep 2127 * track of whether a transaction is already in progress. 2128 */ 2129 file->private_data = NULL; 2130 return 0; 2131 } 2132 2133 /* 2134 * Used by /dev/btrfs-control for devices ioctls. 2135 */ 2136 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 2137 unsigned long arg) 2138 { 2139 struct btrfs_ioctl_vol_args *vol; 2140 struct btrfs_device *device = NULL; 2141 dev_t devt = 0; 2142 int ret = -ENOTTY; 2143 2144 if (!capable(CAP_SYS_ADMIN)) 2145 return -EPERM; 2146 2147 vol = memdup_user((void __user *)arg, sizeof(*vol)); 2148 if (IS_ERR(vol)) 2149 return PTR_ERR(vol); 2150 vol->name[BTRFS_PATH_NAME_MAX] = '\0'; 2151 2152 switch (cmd) { 2153 case BTRFS_IOC_SCAN_DEV: 2154 mutex_lock(&uuid_mutex); 2155 /* 2156 * Scanning outside of mount can return NULL which would turn 2157 * into 0 error code. 2158 */ 2159 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false); 2160 ret = PTR_ERR_OR_ZERO(device); 2161 mutex_unlock(&uuid_mutex); 2162 break; 2163 case BTRFS_IOC_FORGET_DEV: 2164 if (vol->name[0] != 0) { 2165 ret = lookup_bdev(vol->name, &devt); 2166 if (ret) 2167 break; 2168 } 2169 ret = btrfs_forget_devices(devt); 2170 break; 2171 case BTRFS_IOC_DEVICES_READY: 2172 mutex_lock(&uuid_mutex); 2173 /* 2174 * Scanning outside of mount can return NULL which would turn 2175 * into 0 error code. 2176 */ 2177 device = btrfs_scan_one_device(vol->name, BLK_OPEN_READ, false); 2178 if (IS_ERR_OR_NULL(device)) { 2179 mutex_unlock(&uuid_mutex); 2180 ret = PTR_ERR(device); 2181 break; 2182 } 2183 ret = !(device->fs_devices->num_devices == 2184 device->fs_devices->total_devices); 2185 mutex_unlock(&uuid_mutex); 2186 break; 2187 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 2188 ret = btrfs_ioctl_get_supported_features((void __user*)arg); 2189 break; 2190 } 2191 2192 kfree(vol); 2193 return ret; 2194 } 2195 2196 static int btrfs_freeze(struct super_block *sb) 2197 { 2198 struct btrfs_trans_handle *trans; 2199 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2200 struct btrfs_root *root = fs_info->tree_root; 2201 2202 set_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2203 /* 2204 * We don't need a barrier here, we'll wait for any transaction that 2205 * could be in progress on other threads (and do delayed iputs that 2206 * we want to avoid on a frozen filesystem), or do the commit 2207 * ourselves. 2208 */ 2209 trans = btrfs_attach_transaction_barrier(root); 2210 if (IS_ERR(trans)) { 2211 /* no transaction, don't bother */ 2212 if (PTR_ERR(trans) == -ENOENT) 2213 return 0; 2214 return PTR_ERR(trans); 2215 } 2216 return btrfs_commit_transaction(trans); 2217 } 2218 2219 static int check_dev_super(struct btrfs_device *dev) 2220 { 2221 struct btrfs_fs_info *fs_info = dev->fs_info; 2222 struct btrfs_super_block *sb; 2223 u64 last_trans; 2224 u16 csum_type; 2225 int ret = 0; 2226 2227 /* This should be called with fs still frozen. */ 2228 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags)); 2229 2230 /* Missing dev, no need to check. */ 2231 if (!dev->bdev) 2232 return 0; 2233 2234 /* Only need to check the primary super block. */ 2235 sb = btrfs_read_dev_one_super(dev->bdev, 0, true); 2236 if (IS_ERR(sb)) 2237 return PTR_ERR(sb); 2238 2239 /* Verify the checksum. */ 2240 csum_type = btrfs_super_csum_type(sb); 2241 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) { 2242 btrfs_err(fs_info, "csum type changed, has %u expect %u", 2243 csum_type, btrfs_super_csum_type(fs_info->super_copy)); 2244 ret = -EUCLEAN; 2245 goto out; 2246 } 2247 2248 if (btrfs_check_super_csum(fs_info, sb)) { 2249 btrfs_err(fs_info, "csum for on-disk super block no longer matches"); 2250 ret = -EUCLEAN; 2251 goto out; 2252 } 2253 2254 /* Btrfs_validate_super() includes fsid check against super->fsid. */ 2255 ret = btrfs_validate_super(fs_info, sb, 0); 2256 if (ret < 0) 2257 goto out; 2258 2259 last_trans = btrfs_get_last_trans_committed(fs_info); 2260 if (btrfs_super_generation(sb) != last_trans) { 2261 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu", 2262 btrfs_super_generation(sb), last_trans); 2263 ret = -EUCLEAN; 2264 goto out; 2265 } 2266 out: 2267 btrfs_release_disk_super(sb); 2268 return ret; 2269 } 2270 2271 static int btrfs_unfreeze(struct super_block *sb) 2272 { 2273 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2274 struct btrfs_device *device; 2275 int ret = 0; 2276 2277 /* 2278 * Make sure the fs is not changed by accident (like hibernation then 2279 * modified by other OS). 2280 * If we found anything wrong, we mark the fs error immediately. 2281 * 2282 * And since the fs is frozen, no one can modify the fs yet, thus 2283 * we don't need to hold device_list_mutex. 2284 */ 2285 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) { 2286 ret = check_dev_super(device); 2287 if (ret < 0) { 2288 btrfs_handle_fs_error(fs_info, ret, 2289 "super block on devid %llu got modified unexpectedly", 2290 device->devid); 2291 break; 2292 } 2293 } 2294 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2295 2296 /* 2297 * We still return 0, to allow VFS layer to unfreeze the fs even the 2298 * above checks failed. Since the fs is either fine or read-only, we're 2299 * safe to continue, without causing further damage. 2300 */ 2301 return 0; 2302 } 2303 2304 static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 2305 { 2306 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 2307 2308 /* 2309 * There should be always a valid pointer in latest_dev, it may be stale 2310 * for a short moment in case it's being deleted but still valid until 2311 * the end of RCU grace period. 2312 */ 2313 rcu_read_lock(); 2314 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\"); 2315 rcu_read_unlock(); 2316 2317 return 0; 2318 } 2319 2320 static const struct super_operations btrfs_super_ops = { 2321 .drop_inode = btrfs_drop_inode, 2322 .evict_inode = btrfs_evict_inode, 2323 .put_super = btrfs_put_super, 2324 .sync_fs = btrfs_sync_fs, 2325 .show_options = btrfs_show_options, 2326 .show_devname = btrfs_show_devname, 2327 .alloc_inode = btrfs_alloc_inode, 2328 .destroy_inode = btrfs_destroy_inode, 2329 .free_inode = btrfs_free_inode, 2330 .statfs = btrfs_statfs, 2331 .remount_fs = btrfs_remount, 2332 .freeze_fs = btrfs_freeze, 2333 .unfreeze_fs = btrfs_unfreeze, 2334 }; 2335 2336 static const struct file_operations btrfs_ctl_fops = { 2337 .open = btrfs_control_open, 2338 .unlocked_ioctl = btrfs_control_ioctl, 2339 .compat_ioctl = compat_ptr_ioctl, 2340 .owner = THIS_MODULE, 2341 .llseek = noop_llseek, 2342 }; 2343 2344 static struct miscdevice btrfs_misc = { 2345 .minor = BTRFS_MINOR, 2346 .name = "btrfs-control", 2347 .fops = &btrfs_ctl_fops 2348 }; 2349 2350 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2351 MODULE_ALIAS("devname:btrfs-control"); 2352 2353 static int __init btrfs_interface_init(void) 2354 { 2355 return misc_register(&btrfs_misc); 2356 } 2357 2358 static __cold void btrfs_interface_exit(void) 2359 { 2360 misc_deregister(&btrfs_misc); 2361 } 2362 2363 static int __init btrfs_print_mod_info(void) 2364 { 2365 static const char options[] = "" 2366 #ifdef CONFIG_BTRFS_DEBUG 2367 ", debug=on" 2368 #endif 2369 #ifdef CONFIG_BTRFS_ASSERT 2370 ", assert=on" 2371 #endif 2372 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 2373 ", ref-verify=on" 2374 #endif 2375 #ifdef CONFIG_BLK_DEV_ZONED 2376 ", zoned=yes" 2377 #else 2378 ", zoned=no" 2379 #endif 2380 #ifdef CONFIG_FS_VERITY 2381 ", fsverity=yes" 2382 #else 2383 ", fsverity=no" 2384 #endif 2385 ; 2386 pr_info("Btrfs loaded%s\n", options); 2387 return 0; 2388 } 2389 2390 static int register_btrfs(void) 2391 { 2392 return register_filesystem(&btrfs_fs_type); 2393 } 2394 2395 static void unregister_btrfs(void) 2396 { 2397 unregister_filesystem(&btrfs_fs_type); 2398 } 2399 2400 /* Helper structure for long init/exit functions. */ 2401 struct init_sequence { 2402 int (*init_func)(void); 2403 /* Can be NULL if the init_func doesn't need cleanup. */ 2404 void (*exit_func)(void); 2405 }; 2406 2407 static const struct init_sequence mod_init_seq[] = { 2408 { 2409 .init_func = btrfs_props_init, 2410 .exit_func = NULL, 2411 }, { 2412 .init_func = btrfs_init_sysfs, 2413 .exit_func = btrfs_exit_sysfs, 2414 }, { 2415 .init_func = btrfs_init_compress, 2416 .exit_func = btrfs_exit_compress, 2417 }, { 2418 .init_func = btrfs_init_cachep, 2419 .exit_func = btrfs_destroy_cachep, 2420 }, { 2421 .init_func = btrfs_transaction_init, 2422 .exit_func = btrfs_transaction_exit, 2423 }, { 2424 .init_func = btrfs_ctree_init, 2425 .exit_func = btrfs_ctree_exit, 2426 }, { 2427 .init_func = btrfs_free_space_init, 2428 .exit_func = btrfs_free_space_exit, 2429 }, { 2430 .init_func = extent_state_init_cachep, 2431 .exit_func = extent_state_free_cachep, 2432 }, { 2433 .init_func = extent_buffer_init_cachep, 2434 .exit_func = extent_buffer_free_cachep, 2435 }, { 2436 .init_func = btrfs_bioset_init, 2437 .exit_func = btrfs_bioset_exit, 2438 }, { 2439 .init_func = extent_map_init, 2440 .exit_func = extent_map_exit, 2441 }, { 2442 .init_func = ordered_data_init, 2443 .exit_func = ordered_data_exit, 2444 }, { 2445 .init_func = btrfs_delayed_inode_init, 2446 .exit_func = btrfs_delayed_inode_exit, 2447 }, { 2448 .init_func = btrfs_auto_defrag_init, 2449 .exit_func = btrfs_auto_defrag_exit, 2450 }, { 2451 .init_func = btrfs_delayed_ref_init, 2452 .exit_func = btrfs_delayed_ref_exit, 2453 }, { 2454 .init_func = btrfs_prelim_ref_init, 2455 .exit_func = btrfs_prelim_ref_exit, 2456 }, { 2457 .init_func = btrfs_interface_init, 2458 .exit_func = btrfs_interface_exit, 2459 }, { 2460 .init_func = btrfs_print_mod_info, 2461 .exit_func = NULL, 2462 }, { 2463 .init_func = btrfs_run_sanity_tests, 2464 .exit_func = NULL, 2465 }, { 2466 .init_func = register_btrfs, 2467 .exit_func = unregister_btrfs, 2468 } 2469 }; 2470 2471 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)]; 2472 2473 static __always_inline void btrfs_exit_btrfs_fs(void) 2474 { 2475 int i; 2476 2477 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) { 2478 if (!mod_init_result[i]) 2479 continue; 2480 if (mod_init_seq[i].exit_func) 2481 mod_init_seq[i].exit_func(); 2482 mod_init_result[i] = false; 2483 } 2484 } 2485 2486 static void __exit exit_btrfs_fs(void) 2487 { 2488 btrfs_exit_btrfs_fs(); 2489 btrfs_cleanup_fs_uuids(); 2490 } 2491 2492 static int __init init_btrfs_fs(void) 2493 { 2494 int ret; 2495 int i; 2496 2497 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) { 2498 ASSERT(!mod_init_result[i]); 2499 ret = mod_init_seq[i].init_func(); 2500 if (ret < 0) { 2501 btrfs_exit_btrfs_fs(); 2502 return ret; 2503 } 2504 mod_init_result[i] = true; 2505 } 2506 return 0; 2507 } 2508 2509 late_initcall(init_btrfs_fs); 2510 module_exit(exit_btrfs_fs) 2511 2512 MODULE_LICENSE("GPL"); 2513 MODULE_SOFTDEP("pre: crc32c"); 2514 MODULE_SOFTDEP("pre: xxhash64"); 2515 MODULE_SOFTDEP("pre: sha256"); 2516 MODULE_SOFTDEP("pre: blake2b-256"); 2517