1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #include <linux/blkdev.h> 7 #include <linux/module.h> 8 #include <linux/fs.h> 9 #include <linux/pagemap.h> 10 #include <linux/highmem.h> 11 #include <linux/time.h> 12 #include <linux/init.h> 13 #include <linux/seq_file.h> 14 #include <linux/string.h> 15 #include <linux/backing-dev.h> 16 #include <linux/mount.h> 17 #include <linux/writeback.h> 18 #include <linux/statfs.h> 19 #include <linux/compat.h> 20 #include <linux/parser.h> 21 #include <linux/ctype.h> 22 #include <linux/namei.h> 23 #include <linux/miscdevice.h> 24 #include <linux/magic.h> 25 #include <linux/slab.h> 26 #include <linux/ratelimit.h> 27 #include <linux/crc32c.h> 28 #include <linux/btrfs.h> 29 #include <linux/security.h> 30 #include <linux/fs_parser.h> 31 #include "messages.h" 32 #include "delayed-inode.h" 33 #include "ctree.h" 34 #include "disk-io.h" 35 #include "transaction.h" 36 #include "btrfs_inode.h" 37 #include "direct-io.h" 38 #include "props.h" 39 #include "xattr.h" 40 #include "bio.h" 41 #include "export.h" 42 #include "compression.h" 43 #include "dev-replace.h" 44 #include "free-space-cache.h" 45 #include "backref.h" 46 #include "space-info.h" 47 #include "sysfs.h" 48 #include "zoned.h" 49 #include "tests/btrfs-tests.h" 50 #include "block-group.h" 51 #include "discard.h" 52 #include "qgroup.h" 53 #include "raid56.h" 54 #include "fs.h" 55 #include "accessors.h" 56 #include "defrag.h" 57 #include "dir-item.h" 58 #include "ioctl.h" 59 #include "scrub.h" 60 #include "verity.h" 61 #include "super.h" 62 #include "extent-tree.h" 63 #define CREATE_TRACE_POINTS 64 #include <trace/events/btrfs.h> 65 66 static const struct super_operations btrfs_super_ops; 67 static struct file_system_type btrfs_fs_type; 68 69 static void btrfs_put_super(struct super_block *sb) 70 { 71 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 72 73 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid); 74 close_ctree(fs_info); 75 } 76 77 /* Store the mount options related information. */ 78 struct btrfs_fs_context { 79 char *subvol_name; 80 u64 subvol_objectid; 81 u64 max_inline; 82 u32 commit_interval; 83 u32 metadata_ratio; 84 u32 thread_pool_size; 85 unsigned long long mount_opt; 86 unsigned long compress_type:4; 87 int compress_level; 88 refcount_t refs; 89 }; 90 91 enum { 92 Opt_acl, 93 Opt_clear_cache, 94 Opt_commit_interval, 95 Opt_compress, 96 Opt_compress_force, 97 Opt_compress_force_type, 98 Opt_compress_type, 99 Opt_degraded, 100 Opt_device, 101 Opt_fatal_errors, 102 Opt_flushoncommit, 103 Opt_max_inline, 104 Opt_barrier, 105 Opt_datacow, 106 Opt_datasum, 107 Opt_defrag, 108 Opt_discard, 109 Opt_discard_mode, 110 Opt_ratio, 111 Opt_rescan_uuid_tree, 112 Opt_skip_balance, 113 Opt_space_cache, 114 Opt_space_cache_version, 115 Opt_ssd, 116 Opt_ssd_spread, 117 Opt_subvol, 118 Opt_subvol_empty, 119 Opt_subvolid, 120 Opt_thread_pool, 121 Opt_treelog, 122 Opt_user_subvol_rm_allowed, 123 Opt_norecovery, 124 125 /* Rescue options */ 126 Opt_rescue, 127 Opt_usebackuproot, 128 129 /* Debugging options */ 130 Opt_enospc_debug, 131 #ifdef CONFIG_BTRFS_DEBUG 132 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all, 133 #endif 134 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 135 Opt_ref_verify, 136 #endif 137 Opt_err, 138 }; 139 140 enum { 141 Opt_fatal_errors_panic, 142 Opt_fatal_errors_bug, 143 }; 144 145 static const struct constant_table btrfs_parameter_fatal_errors[] = { 146 { "panic", Opt_fatal_errors_panic }, 147 { "bug", Opt_fatal_errors_bug }, 148 {} 149 }; 150 151 enum { 152 Opt_discard_sync, 153 Opt_discard_async, 154 }; 155 156 static const struct constant_table btrfs_parameter_discard[] = { 157 { "sync", Opt_discard_sync }, 158 { "async", Opt_discard_async }, 159 {} 160 }; 161 162 enum { 163 Opt_space_cache_v1, 164 Opt_space_cache_v2, 165 }; 166 167 static const struct constant_table btrfs_parameter_space_cache[] = { 168 { "v1", Opt_space_cache_v1 }, 169 { "v2", Opt_space_cache_v2 }, 170 {} 171 }; 172 173 enum { 174 Opt_rescue_usebackuproot, 175 Opt_rescue_nologreplay, 176 Opt_rescue_ignorebadroots, 177 Opt_rescue_ignoredatacsums, 178 Opt_rescue_ignoremetacsums, 179 Opt_rescue_ignoresuperflags, 180 Opt_rescue_parameter_all, 181 }; 182 183 static const struct constant_table btrfs_parameter_rescue[] = { 184 { "usebackuproot", Opt_rescue_usebackuproot }, 185 { "nologreplay", Opt_rescue_nologreplay }, 186 { "ignorebadroots", Opt_rescue_ignorebadroots }, 187 { "ibadroots", Opt_rescue_ignorebadroots }, 188 { "ignoredatacsums", Opt_rescue_ignoredatacsums }, 189 { "ignoremetacsums", Opt_rescue_ignoremetacsums}, 190 { "ignoresuperflags", Opt_rescue_ignoresuperflags}, 191 { "idatacsums", Opt_rescue_ignoredatacsums }, 192 { "imetacsums", Opt_rescue_ignoremetacsums}, 193 { "isuperflags", Opt_rescue_ignoresuperflags}, 194 { "all", Opt_rescue_parameter_all }, 195 {} 196 }; 197 198 #ifdef CONFIG_BTRFS_DEBUG 199 enum { 200 Opt_fragment_parameter_data, 201 Opt_fragment_parameter_metadata, 202 Opt_fragment_parameter_all, 203 }; 204 205 static const struct constant_table btrfs_parameter_fragment[] = { 206 { "data", Opt_fragment_parameter_data }, 207 { "metadata", Opt_fragment_parameter_metadata }, 208 { "all", Opt_fragment_parameter_all }, 209 {} 210 }; 211 #endif 212 213 static const struct fs_parameter_spec btrfs_fs_parameters[] = { 214 fsparam_flag_no("acl", Opt_acl), 215 fsparam_flag_no("autodefrag", Opt_defrag), 216 fsparam_flag_no("barrier", Opt_barrier), 217 fsparam_flag("clear_cache", Opt_clear_cache), 218 fsparam_u32("commit", Opt_commit_interval), 219 fsparam_flag("compress", Opt_compress), 220 fsparam_string("compress", Opt_compress_type), 221 fsparam_flag("compress-force", Opt_compress_force), 222 fsparam_string("compress-force", Opt_compress_force_type), 223 fsparam_flag_no("datacow", Opt_datacow), 224 fsparam_flag_no("datasum", Opt_datasum), 225 fsparam_flag("degraded", Opt_degraded), 226 fsparam_string("device", Opt_device), 227 fsparam_flag_no("discard", Opt_discard), 228 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard), 229 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors), 230 fsparam_flag_no("flushoncommit", Opt_flushoncommit), 231 fsparam_string("max_inline", Opt_max_inline), 232 fsparam_u32("metadata_ratio", Opt_ratio), 233 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree), 234 fsparam_flag("skip_balance", Opt_skip_balance), 235 fsparam_flag_no("space_cache", Opt_space_cache), 236 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache), 237 fsparam_flag_no("ssd", Opt_ssd), 238 fsparam_flag_no("ssd_spread", Opt_ssd_spread), 239 fsparam_string("subvol", Opt_subvol), 240 fsparam_flag("subvol=", Opt_subvol_empty), 241 fsparam_u64("subvolid", Opt_subvolid), 242 fsparam_u32("thread_pool", Opt_thread_pool), 243 fsparam_flag_no("treelog", Opt_treelog), 244 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed), 245 246 /* Rescue options. */ 247 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue), 248 /* Deprecated, with alias rescue=usebackuproot */ 249 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL), 250 /* For compatibility only, alias for "rescue=nologreplay". */ 251 fsparam_flag("norecovery", Opt_norecovery), 252 253 /* Debugging options. */ 254 fsparam_flag_no("enospc_debug", Opt_enospc_debug), 255 #ifdef CONFIG_BTRFS_DEBUG 256 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment), 257 #endif 258 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 259 fsparam_flag("ref_verify", Opt_ref_verify), 260 #endif 261 {} 262 }; 263 264 static bool btrfs_match_compress_type(const char *string, const char *type, bool may_have_level) 265 { 266 const int len = strlen(type); 267 268 return (strncmp(string, type, len) == 0) && 269 ((may_have_level && string[len] == ':') || string[len] == '\0'); 270 } 271 272 static int btrfs_parse_compress(struct btrfs_fs_context *ctx, 273 const struct fs_parameter *param, int opt) 274 { 275 const char *string = param->string; 276 277 /* 278 * Provide the same semantics as older kernels that don't use fs 279 * context, specifying the "compress" option clears "force-compress" 280 * without the need to pass "compress-force=[no|none]" before 281 * specifying "compress". 282 */ 283 if (opt != Opt_compress_force && opt != Opt_compress_force_type) 284 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 285 286 if (opt == Opt_compress || opt == Opt_compress_force) { 287 ctx->compress_type = BTRFS_COMPRESS_ZLIB; 288 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL; 289 btrfs_set_opt(ctx->mount_opt, COMPRESS); 290 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 291 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 292 } else if (btrfs_match_compress_type(string, "zlib", true)) { 293 ctx->compress_type = BTRFS_COMPRESS_ZLIB; 294 ctx->compress_level = btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, 295 string + 4); 296 btrfs_set_opt(ctx->mount_opt, COMPRESS); 297 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 298 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 299 } else if (btrfs_match_compress_type(string, "lzo", false)) { 300 ctx->compress_type = BTRFS_COMPRESS_LZO; 301 ctx->compress_level = 0; 302 btrfs_set_opt(ctx->mount_opt, COMPRESS); 303 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 304 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 305 } else if (btrfs_match_compress_type(string, "zstd", true)) { 306 ctx->compress_type = BTRFS_COMPRESS_ZSTD; 307 ctx->compress_level = btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, 308 string + 4); 309 btrfs_set_opt(ctx->mount_opt, COMPRESS); 310 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 311 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 312 } else if (btrfs_match_compress_type(string, "no", false) || 313 btrfs_match_compress_type(string, "none", false)) { 314 ctx->compress_level = 0; 315 ctx->compress_type = 0; 316 btrfs_clear_opt(ctx->mount_opt, COMPRESS); 317 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 318 } else { 319 btrfs_err(NULL, "unrecognized compression value %s", string); 320 return -EINVAL; 321 } 322 return 0; 323 } 324 325 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param) 326 { 327 struct btrfs_fs_context *ctx = fc->fs_private; 328 struct fs_parse_result result; 329 int opt; 330 331 opt = fs_parse(fc, btrfs_fs_parameters, param, &result); 332 if (opt < 0) 333 return opt; 334 335 switch (opt) { 336 case Opt_degraded: 337 btrfs_set_opt(ctx->mount_opt, DEGRADED); 338 break; 339 case Opt_subvol_empty: 340 /* 341 * This exists because we used to allow it on accident, so we're 342 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow 343 * empty subvol= again"). 344 */ 345 break; 346 case Opt_subvol: 347 kfree(ctx->subvol_name); 348 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL); 349 if (!ctx->subvol_name) 350 return -ENOMEM; 351 break; 352 case Opt_subvolid: 353 ctx->subvol_objectid = result.uint_64; 354 355 /* subvolid=0 means give me the original fs_tree. */ 356 if (!ctx->subvol_objectid) 357 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID; 358 break; 359 case Opt_device: { 360 struct btrfs_device *device; 361 362 mutex_lock(&uuid_mutex); 363 device = btrfs_scan_one_device(param->string, false); 364 mutex_unlock(&uuid_mutex); 365 if (IS_ERR(device)) 366 return PTR_ERR(device); 367 break; 368 } 369 case Opt_datasum: 370 if (result.negated) { 371 btrfs_set_opt(ctx->mount_opt, NODATASUM); 372 } else { 373 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 374 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 375 } 376 break; 377 case Opt_datacow: 378 if (result.negated) { 379 btrfs_clear_opt(ctx->mount_opt, COMPRESS); 380 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 381 btrfs_set_opt(ctx->mount_opt, NODATACOW); 382 btrfs_set_opt(ctx->mount_opt, NODATASUM); 383 } else { 384 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 385 } 386 break; 387 case Opt_compress_force: 388 case Opt_compress_force_type: 389 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS); 390 fallthrough; 391 case Opt_compress: 392 case Opt_compress_type: 393 if (btrfs_parse_compress(ctx, param, opt)) 394 return -EINVAL; 395 break; 396 case Opt_ssd: 397 if (result.negated) { 398 btrfs_set_opt(ctx->mount_opt, NOSSD); 399 btrfs_clear_opt(ctx->mount_opt, SSD); 400 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); 401 } else { 402 btrfs_set_opt(ctx->mount_opt, SSD); 403 btrfs_clear_opt(ctx->mount_opt, NOSSD); 404 } 405 break; 406 case Opt_ssd_spread: 407 if (result.negated) { 408 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); 409 } else { 410 btrfs_set_opt(ctx->mount_opt, SSD); 411 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD); 412 btrfs_clear_opt(ctx->mount_opt, NOSSD); 413 } 414 break; 415 case Opt_barrier: 416 if (result.negated) 417 btrfs_set_opt(ctx->mount_opt, NOBARRIER); 418 else 419 btrfs_clear_opt(ctx->mount_opt, NOBARRIER); 420 break; 421 case Opt_thread_pool: 422 if (result.uint_32 == 0) { 423 btrfs_err(NULL, "invalid value 0 for thread_pool"); 424 return -EINVAL; 425 } 426 ctx->thread_pool_size = result.uint_32; 427 break; 428 case Opt_max_inline: 429 ctx->max_inline = memparse(param->string, NULL); 430 break; 431 case Opt_acl: 432 if (result.negated) { 433 fc->sb_flags &= ~SB_POSIXACL; 434 } else { 435 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 436 fc->sb_flags |= SB_POSIXACL; 437 #else 438 btrfs_err(NULL, "support for ACL not compiled in"); 439 return -EINVAL; 440 #endif 441 } 442 /* 443 * VFS limits the ability to toggle ACL on and off via remount, 444 * despite every file system allowing this. This seems to be 445 * an oversight since we all do, but it'll fail if we're 446 * remounting. So don't set the mask here, we'll check it in 447 * btrfs_reconfigure and do the toggling ourselves. 448 */ 449 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE) 450 fc->sb_flags_mask |= SB_POSIXACL; 451 break; 452 case Opt_treelog: 453 if (result.negated) 454 btrfs_set_opt(ctx->mount_opt, NOTREELOG); 455 else 456 btrfs_clear_opt(ctx->mount_opt, NOTREELOG); 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 > BTRFS_WARNING_COMMIT_INTERVAL) { 574 btrfs_warn(NULL, "excessive commit interval %u, use with care", 575 ctx->commit_interval); 576 } 577 if (ctx->commit_interval == 0) 578 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 579 break; 580 case Opt_rescue: 581 switch (result.uint_32) { 582 case Opt_rescue_usebackuproot: 583 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT); 584 break; 585 case Opt_rescue_nologreplay: 586 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); 587 break; 588 case Opt_rescue_ignorebadroots: 589 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); 590 break; 591 case Opt_rescue_ignoredatacsums: 592 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); 593 break; 594 case Opt_rescue_ignoremetacsums: 595 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS); 596 break; 597 case Opt_rescue_ignoresuperflags: 598 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS); 599 break; 600 case Opt_rescue_parameter_all: 601 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); 602 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS); 603 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS); 604 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); 605 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); 606 break; 607 default: 608 btrfs_info(NULL, "unrecognized rescue option '%s'", 609 param->key); 610 return -EINVAL; 611 } 612 break; 613 #ifdef CONFIG_BTRFS_DEBUG 614 case Opt_fragment: 615 switch (result.uint_32) { 616 case Opt_fragment_parameter_all: 617 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); 618 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); 619 break; 620 case Opt_fragment_parameter_metadata: 621 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); 622 break; 623 case Opt_fragment_parameter_data: 624 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); 625 break; 626 default: 627 btrfs_info(NULL, "unrecognized fragment option '%s'", 628 param->key); 629 return -EINVAL; 630 } 631 break; 632 #endif 633 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 634 case Opt_ref_verify: 635 btrfs_set_opt(ctx->mount_opt, REF_VERIFY); 636 break; 637 #endif 638 default: 639 btrfs_err(NULL, "unrecognized mount option '%s'", param->key); 640 return -EINVAL; 641 } 642 643 return 0; 644 } 645 646 /* 647 * Some options only have meaning at mount time and shouldn't persist across 648 * remounts, or be displayed. Clear these at the end of mount and remount code 649 * paths. 650 */ 651 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 652 { 653 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 654 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 655 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE); 656 } 657 658 static bool check_ro_option(const struct btrfs_fs_info *fs_info, 659 unsigned long long mount_opt, unsigned long long opt, 660 const char *opt_name) 661 { 662 if (mount_opt & opt) { 663 btrfs_err(fs_info, "%s must be used with ro mount option", 664 opt_name); 665 return true; 666 } 667 return false; 668 } 669 670 bool btrfs_check_options(const struct btrfs_fs_info *info, 671 unsigned long long *mount_opt, 672 unsigned long flags) 673 { 674 bool ret = true; 675 676 if (!(flags & SB_RDONLY) && 677 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") || 678 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") || 679 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") || 680 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") || 681 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags"))) 682 ret = false; 683 684 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) && 685 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) && 686 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) { 687 btrfs_err(info, "cannot disable free-space-tree"); 688 ret = false; 689 } 690 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) && 691 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) { 692 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature"); 693 ret = false; 694 } 695 696 if (btrfs_check_mountopts_zoned(info, mount_opt)) 697 ret = false; 698 699 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) { 700 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) { 701 btrfs_info(info, "disk space caching is enabled"); 702 btrfs_warn(info, 703 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2"); 704 } 705 if (btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) 706 btrfs_info(info, "using free-space-tree"); 707 } 708 709 return ret; 710 } 711 712 /* 713 * This is subtle, we only call this during open_ctree(). We need to pre-load 714 * the mount options with the on-disk settings. Before the new mount API took 715 * effect we would do this on mount and remount. With the new mount API we'll 716 * only do this on the initial mount. 717 * 718 * This isn't a change in behavior, because we're using the current state of the 719 * file system to set the current mount options. If you mounted with special 720 * options to disable these features and then remounted we wouldn't revert the 721 * settings, because mounting without these features cleared the on-disk 722 * settings, so this being called on re-mount is not needed. 723 */ 724 void btrfs_set_free_space_cache_settings(struct btrfs_fs_info *fs_info) 725 { 726 if (fs_info->sectorsize < PAGE_SIZE) { 727 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 728 if (!btrfs_test_opt(fs_info, FREE_SPACE_TREE)) { 729 btrfs_info(fs_info, 730 "forcing free space tree for sector size %u with page size %lu", 731 fs_info->sectorsize, PAGE_SIZE); 732 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); 733 } 734 } 735 736 /* 737 * At this point our mount options are populated, so we only mess with 738 * these settings if we don't have any settings already. 739 */ 740 if (btrfs_test_opt(fs_info, FREE_SPACE_TREE)) 741 return; 742 743 if (btrfs_is_zoned(fs_info) && 744 btrfs_free_space_cache_v1_active(fs_info)) { 745 btrfs_info(fs_info, "zoned: clearing existing space cache"); 746 btrfs_set_super_cache_generation(fs_info->super_copy, 0); 747 return; 748 } 749 750 if (btrfs_test_opt(fs_info, SPACE_CACHE)) 751 return; 752 753 if (btrfs_test_opt(fs_info, NOSPACECACHE)) 754 return; 755 756 /* 757 * At this point we don't have explicit options set by the user, set 758 * them ourselves based on the state of the file system. 759 */ 760 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) 761 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); 762 else if (btrfs_free_space_cache_v1_active(fs_info)) 763 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); 764 } 765 766 static void set_device_specific_options(struct btrfs_fs_info *fs_info) 767 { 768 if (!btrfs_test_opt(fs_info, NOSSD) && 769 !fs_info->fs_devices->rotating) 770 btrfs_set_opt(fs_info->mount_opt, SSD); 771 772 /* 773 * For devices supporting discard turn on discard=async automatically, 774 * unless it's already set or disabled. This could be turned off by 775 * nodiscard for the same mount. 776 * 777 * The zoned mode piggy backs on the discard functionality for 778 * resetting a zone. There is no reason to delay the zone reset as it is 779 * fast enough. So, do not enable async discard for zoned mode. 780 */ 781 if (!(btrfs_test_opt(fs_info, DISCARD_SYNC) || 782 btrfs_test_opt(fs_info, DISCARD_ASYNC) || 783 btrfs_test_opt(fs_info, NODISCARD)) && 784 fs_info->fs_devices->discardable && 785 !btrfs_is_zoned(fs_info)) 786 btrfs_set_opt(fs_info->mount_opt, DISCARD_ASYNC); 787 } 788 789 char *btrfs_get_subvol_name_from_objectid(struct btrfs_fs_info *fs_info, 790 u64 subvol_objectid) 791 { 792 struct btrfs_root *root = fs_info->tree_root; 793 struct btrfs_root *fs_root = NULL; 794 struct btrfs_root_ref *root_ref; 795 struct btrfs_inode_ref *inode_ref; 796 struct btrfs_key key; 797 struct btrfs_path *path = NULL; 798 char *name = NULL, *ptr; 799 u64 dirid; 800 int len; 801 int ret; 802 803 path = btrfs_alloc_path(); 804 if (!path) { 805 ret = -ENOMEM; 806 goto err; 807 } 808 809 name = kmalloc(PATH_MAX, GFP_KERNEL); 810 if (!name) { 811 ret = -ENOMEM; 812 goto err; 813 } 814 ptr = name + PATH_MAX - 1; 815 ptr[0] = '\0'; 816 817 /* 818 * Walk up the subvolume trees in the tree of tree roots by root 819 * backrefs until we hit the top-level subvolume. 820 */ 821 while (subvol_objectid != BTRFS_FS_TREE_OBJECTID) { 822 key.objectid = subvol_objectid; 823 key.type = BTRFS_ROOT_BACKREF_KEY; 824 key.offset = (u64)-1; 825 826 ret = btrfs_search_backwards(root, &key, path); 827 if (ret < 0) { 828 goto err; 829 } else if (ret > 0) { 830 ret = -ENOENT; 831 goto err; 832 } 833 834 subvol_objectid = key.offset; 835 836 root_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 837 struct btrfs_root_ref); 838 len = btrfs_root_ref_name_len(path->nodes[0], root_ref); 839 ptr -= len + 1; 840 if (ptr < name) { 841 ret = -ENAMETOOLONG; 842 goto err; 843 } 844 read_extent_buffer(path->nodes[0], ptr + 1, 845 (unsigned long)(root_ref + 1), len); 846 ptr[0] = '/'; 847 dirid = btrfs_root_ref_dirid(path->nodes[0], root_ref); 848 btrfs_release_path(path); 849 850 fs_root = btrfs_get_fs_root(fs_info, subvol_objectid, true); 851 if (IS_ERR(fs_root)) { 852 ret = PTR_ERR(fs_root); 853 fs_root = NULL; 854 goto err; 855 } 856 857 /* 858 * Walk up the filesystem tree by inode refs until we hit the 859 * root directory. 860 */ 861 while (dirid != BTRFS_FIRST_FREE_OBJECTID) { 862 key.objectid = dirid; 863 key.type = BTRFS_INODE_REF_KEY; 864 key.offset = (u64)-1; 865 866 ret = btrfs_search_backwards(fs_root, &key, path); 867 if (ret < 0) { 868 goto err; 869 } else if (ret > 0) { 870 ret = -ENOENT; 871 goto err; 872 } 873 874 dirid = key.offset; 875 876 inode_ref = btrfs_item_ptr(path->nodes[0], 877 path->slots[0], 878 struct btrfs_inode_ref); 879 len = btrfs_inode_ref_name_len(path->nodes[0], 880 inode_ref); 881 ptr -= len + 1; 882 if (ptr < name) { 883 ret = -ENAMETOOLONG; 884 goto err; 885 } 886 read_extent_buffer(path->nodes[0], ptr + 1, 887 (unsigned long)(inode_ref + 1), len); 888 ptr[0] = '/'; 889 btrfs_release_path(path); 890 } 891 btrfs_put_root(fs_root); 892 fs_root = NULL; 893 } 894 895 btrfs_free_path(path); 896 if (ptr == name + PATH_MAX - 1) { 897 name[0] = '/'; 898 name[1] = '\0'; 899 } else { 900 memmove(name, ptr, name + PATH_MAX - ptr); 901 } 902 return name; 903 904 err: 905 btrfs_put_root(fs_root); 906 btrfs_free_path(path); 907 kfree(name); 908 return ERR_PTR(ret); 909 } 910 911 static int get_default_subvol_objectid(struct btrfs_fs_info *fs_info, u64 *objectid) 912 { 913 struct btrfs_root *root = fs_info->tree_root; 914 struct btrfs_dir_item *di; 915 struct btrfs_path *path; 916 struct btrfs_key location; 917 struct fscrypt_str name = FSTR_INIT("default", 7); 918 u64 dir_id; 919 920 path = btrfs_alloc_path(); 921 if (!path) 922 return -ENOMEM; 923 924 /* 925 * Find the "default" dir item which points to the root item that we 926 * will mount by default if we haven't been given a specific subvolume 927 * to mount. 928 */ 929 dir_id = btrfs_super_root_dir(fs_info->super_copy); 930 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, &name, 0); 931 if (IS_ERR(di)) { 932 btrfs_free_path(path); 933 return PTR_ERR(di); 934 } 935 if (!di) { 936 /* 937 * Ok the default dir item isn't there. This is weird since 938 * it's always been there, but don't freak out, just try and 939 * mount the top-level subvolume. 940 */ 941 btrfs_free_path(path); 942 *objectid = BTRFS_FS_TREE_OBJECTID; 943 return 0; 944 } 945 946 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 947 btrfs_free_path(path); 948 *objectid = location.objectid; 949 return 0; 950 } 951 952 static int btrfs_fill_super(struct super_block *sb, 953 struct btrfs_fs_devices *fs_devices) 954 { 955 struct btrfs_inode *inode; 956 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 957 int ret; 958 959 sb->s_maxbytes = MAX_LFS_FILESIZE; 960 sb->s_magic = BTRFS_SUPER_MAGIC; 961 sb->s_op = &btrfs_super_ops; 962 set_default_d_op(sb, &btrfs_dentry_operations); 963 sb->s_export_op = &btrfs_export_ops; 964 #ifdef CONFIG_FS_VERITY 965 sb->s_vop = &btrfs_verityops; 966 #endif 967 sb->s_xattr = btrfs_xattr_handlers; 968 sb->s_time_gran = 1; 969 sb->s_iflags |= SB_I_CGROUPWB | SB_I_ALLOW_HSM; 970 971 ret = super_setup_bdi(sb); 972 if (ret) { 973 btrfs_err(fs_info, "super_setup_bdi failed"); 974 return ret; 975 } 976 977 ret = open_ctree(sb, fs_devices); 978 if (ret) { 979 btrfs_err(fs_info, "open_ctree failed: %d", ret); 980 return ret; 981 } 982 983 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root); 984 if (IS_ERR(inode)) { 985 ret = PTR_ERR(inode); 986 btrfs_handle_fs_error(fs_info, ret, NULL); 987 goto fail_close; 988 } 989 990 sb->s_root = d_make_root(&inode->vfs_inode); 991 if (!sb->s_root) { 992 ret = -ENOMEM; 993 goto fail_close; 994 } 995 996 sb->s_flags |= SB_ACTIVE; 997 return 0; 998 999 fail_close: 1000 close_ctree(fs_info); 1001 return ret; 1002 } 1003 1004 int btrfs_sync_fs(struct super_block *sb, int wait) 1005 { 1006 struct btrfs_trans_handle *trans; 1007 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1008 struct btrfs_root *root = fs_info->tree_root; 1009 1010 trace_btrfs_sync_fs(fs_info, wait); 1011 1012 if (!wait) { 1013 filemap_flush(fs_info->btree_inode->i_mapping); 1014 return 0; 1015 } 1016 1017 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL); 1018 1019 trans = btrfs_attach_transaction_barrier(root); 1020 if (IS_ERR(trans)) { 1021 /* no transaction, don't bother */ 1022 if (PTR_ERR(trans) == -ENOENT) { 1023 /* 1024 * Exit unless we have some pending changes 1025 * that need to go through commit 1026 */ 1027 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT, 1028 &fs_info->flags)) 1029 return 0; 1030 /* 1031 * A non-blocking test if the fs is frozen. We must not 1032 * start a new transaction here otherwise a deadlock 1033 * happens. The pending operations are delayed to the 1034 * next commit after thawing. 1035 */ 1036 if (sb_start_write_trylock(sb)) 1037 sb_end_write(sb); 1038 else 1039 return 0; 1040 trans = btrfs_start_transaction(root, 0); 1041 } 1042 if (IS_ERR(trans)) 1043 return PTR_ERR(trans); 1044 } 1045 return btrfs_commit_transaction(trans); 1046 } 1047 1048 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed) 1049 { 1050 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s); 1051 *printed = true; 1052 } 1053 1054 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 1055 { 1056 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 1057 const char *compress_type; 1058 const char *subvol_name; 1059 bool printed = false; 1060 1061 if (btrfs_test_opt(info, DEGRADED)) 1062 seq_puts(seq, ",degraded"); 1063 if (btrfs_test_opt(info, NODATASUM)) 1064 seq_puts(seq, ",nodatasum"); 1065 if (btrfs_test_opt(info, NODATACOW)) 1066 seq_puts(seq, ",nodatacow"); 1067 if (btrfs_test_opt(info, NOBARRIER)) 1068 seq_puts(seq, ",nobarrier"); 1069 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) 1070 seq_printf(seq, ",max_inline=%llu", info->max_inline); 1071 if (info->thread_pool_size != min_t(unsigned long, 1072 num_online_cpus() + 2, 8)) 1073 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size); 1074 if (btrfs_test_opt(info, COMPRESS)) { 1075 compress_type = btrfs_compress_type2str(info->compress_type); 1076 if (btrfs_test_opt(info, FORCE_COMPRESS)) 1077 seq_printf(seq, ",compress-force=%s", compress_type); 1078 else 1079 seq_printf(seq, ",compress=%s", compress_type); 1080 if (info->compress_level) 1081 seq_printf(seq, ":%d", info->compress_level); 1082 } 1083 if (btrfs_test_opt(info, NOSSD)) 1084 seq_puts(seq, ",nossd"); 1085 if (btrfs_test_opt(info, SSD_SPREAD)) 1086 seq_puts(seq, ",ssd_spread"); 1087 else if (btrfs_test_opt(info, SSD)) 1088 seq_puts(seq, ",ssd"); 1089 if (btrfs_test_opt(info, NOTREELOG)) 1090 seq_puts(seq, ",notreelog"); 1091 if (btrfs_test_opt(info, NOLOGREPLAY)) 1092 print_rescue_option(seq, "nologreplay", &printed); 1093 if (btrfs_test_opt(info, USEBACKUPROOT)) 1094 print_rescue_option(seq, "usebackuproot", &printed); 1095 if (btrfs_test_opt(info, IGNOREBADROOTS)) 1096 print_rescue_option(seq, "ignorebadroots", &printed); 1097 if (btrfs_test_opt(info, IGNOREDATACSUMS)) 1098 print_rescue_option(seq, "ignoredatacsums", &printed); 1099 if (btrfs_test_opt(info, IGNOREMETACSUMS)) 1100 print_rescue_option(seq, "ignoremetacsums", &printed); 1101 if (btrfs_test_opt(info, IGNORESUPERFLAGS)) 1102 print_rescue_option(seq, "ignoresuperflags", &printed); 1103 if (btrfs_test_opt(info, FLUSHONCOMMIT)) 1104 seq_puts(seq, ",flushoncommit"); 1105 if (btrfs_test_opt(info, DISCARD_SYNC)) 1106 seq_puts(seq, ",discard"); 1107 if (btrfs_test_opt(info, DISCARD_ASYNC)) 1108 seq_puts(seq, ",discard=async"); 1109 if (!(info->sb->s_flags & SB_POSIXACL)) 1110 seq_puts(seq, ",noacl"); 1111 if (btrfs_free_space_cache_v1_active(info)) 1112 seq_puts(seq, ",space_cache"); 1113 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE)) 1114 seq_puts(seq, ",space_cache=v2"); 1115 else 1116 seq_puts(seq, ",nospace_cache"); 1117 if (btrfs_test_opt(info, RESCAN_UUID_TREE)) 1118 seq_puts(seq, ",rescan_uuid_tree"); 1119 if (btrfs_test_opt(info, CLEAR_CACHE)) 1120 seq_puts(seq, ",clear_cache"); 1121 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED)) 1122 seq_puts(seq, ",user_subvol_rm_allowed"); 1123 if (btrfs_test_opt(info, ENOSPC_DEBUG)) 1124 seq_puts(seq, ",enospc_debug"); 1125 if (btrfs_test_opt(info, AUTO_DEFRAG)) 1126 seq_puts(seq, ",autodefrag"); 1127 if (btrfs_test_opt(info, SKIP_BALANCE)) 1128 seq_puts(seq, ",skip_balance"); 1129 if (info->metadata_ratio) 1130 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio); 1131 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR)) 1132 seq_puts(seq, ",fatal_errors=panic"); 1133 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1134 seq_printf(seq, ",commit=%u", info->commit_interval); 1135 #ifdef CONFIG_BTRFS_DEBUG 1136 if (btrfs_test_opt(info, FRAGMENT_DATA)) 1137 seq_puts(seq, ",fragment=data"); 1138 if (btrfs_test_opt(info, FRAGMENT_METADATA)) 1139 seq_puts(seq, ",fragment=metadata"); 1140 #endif 1141 if (btrfs_test_opt(info, REF_VERIFY)) 1142 seq_puts(seq, ",ref_verify"); 1143 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); 1144 subvol_name = btrfs_get_subvol_name_from_objectid(info, 1145 btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); 1146 if (!IS_ERR(subvol_name)) { 1147 seq_show_option(seq, "subvol", subvol_name); 1148 kfree(subvol_name); 1149 } 1150 return 0; 1151 } 1152 1153 /* 1154 * subvolumes are identified by ino 256 1155 */ 1156 static inline bool is_subvolume_inode(struct inode *inode) 1157 { 1158 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1159 return true; 1160 return false; 1161 } 1162 1163 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid, 1164 struct vfsmount *mnt) 1165 { 1166 struct dentry *root; 1167 int ret; 1168 1169 if (!subvol_name) { 1170 if (!subvol_objectid) { 1171 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb), 1172 &subvol_objectid); 1173 if (ret) { 1174 root = ERR_PTR(ret); 1175 goto out; 1176 } 1177 } 1178 subvol_name = btrfs_get_subvol_name_from_objectid( 1179 btrfs_sb(mnt->mnt_sb), subvol_objectid); 1180 if (IS_ERR(subvol_name)) { 1181 root = ERR_CAST(subvol_name); 1182 subvol_name = NULL; 1183 goto out; 1184 } 1185 1186 } 1187 1188 root = mount_subtree(mnt, subvol_name); 1189 /* mount_subtree() drops our reference on the vfsmount. */ 1190 mnt = NULL; 1191 1192 if (!IS_ERR(root)) { 1193 struct super_block *s = root->d_sb; 1194 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1195 struct inode *root_inode = d_inode(root); 1196 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root); 1197 1198 ret = 0; 1199 if (!is_subvolume_inode(root_inode)) { 1200 btrfs_err(fs_info, "'%s' is not a valid subvolume", 1201 subvol_name); 1202 ret = -EINVAL; 1203 } 1204 if (subvol_objectid && root_objectid != subvol_objectid) { 1205 /* 1206 * This will also catch a race condition where a 1207 * subvolume which was passed by ID is renamed and 1208 * another subvolume is renamed over the old location. 1209 */ 1210 btrfs_err(fs_info, 1211 "subvol '%s' does not match subvolid %llu", 1212 subvol_name, subvol_objectid); 1213 ret = -EINVAL; 1214 } 1215 if (ret) { 1216 dput(root); 1217 root = ERR_PTR(ret); 1218 deactivate_locked_super(s); 1219 } 1220 } 1221 1222 out: 1223 mntput(mnt); 1224 kfree(subvol_name); 1225 return root; 1226 } 1227 1228 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 1229 u32 new_pool_size, u32 old_pool_size) 1230 { 1231 if (new_pool_size == old_pool_size) 1232 return; 1233 1234 fs_info->thread_pool_size = new_pool_size; 1235 1236 btrfs_info(fs_info, "resize thread pool %d -> %d", 1237 old_pool_size, new_pool_size); 1238 1239 btrfs_workqueue_set_max(fs_info->workers, new_pool_size); 1240 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size); 1241 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size); 1242 workqueue_set_max_active(fs_info->endio_workers, new_pool_size); 1243 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size); 1244 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size); 1245 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size); 1246 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size); 1247 } 1248 1249 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1250 unsigned long long old_opts, int flags) 1251 { 1252 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1253 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1254 (flags & SB_RDONLY))) { 1255 /* wait for any defraggers to finish */ 1256 wait_event(fs_info->transaction_wait, 1257 (atomic_read(&fs_info->defrag_running) == 0)); 1258 if (flags & SB_RDONLY) 1259 sync_filesystem(fs_info->sb); 1260 } 1261 } 1262 1263 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1264 unsigned long long old_opts) 1265 { 1266 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 1267 1268 /* 1269 * We need to cleanup all defragable inodes if the autodefragment is 1270 * close or the filesystem is read only. 1271 */ 1272 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1273 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) { 1274 btrfs_cleanup_defrag_inodes(fs_info); 1275 } 1276 1277 /* If we toggled discard async */ 1278 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1279 btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1280 btrfs_discard_resume(fs_info); 1281 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1282 !btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1283 btrfs_discard_cleanup(fs_info); 1284 1285 /* If we toggled space cache */ 1286 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) 1287 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 1288 } 1289 1290 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info) 1291 { 1292 int ret; 1293 1294 if (BTRFS_FS_ERROR(fs_info)) { 1295 btrfs_err(fs_info, 1296 "remounting read-write after error is not allowed"); 1297 return -EINVAL; 1298 } 1299 1300 if (fs_info->fs_devices->rw_devices == 0) 1301 return -EACCES; 1302 1303 if (!btrfs_check_rw_degradable(fs_info, NULL)) { 1304 btrfs_warn(fs_info, 1305 "too many missing devices, writable remount is not allowed"); 1306 return -EACCES; 1307 } 1308 1309 if (btrfs_super_log_root(fs_info->super_copy) != 0) { 1310 btrfs_warn(fs_info, 1311 "mount required to replay tree-log, cannot remount read-write"); 1312 return -EINVAL; 1313 } 1314 1315 /* 1316 * NOTE: when remounting with a change that does writes, don't put it 1317 * anywhere above this point, as we are not sure to be safe to write 1318 * until we pass the above checks. 1319 */ 1320 ret = btrfs_start_pre_rw_mount(fs_info); 1321 if (ret) 1322 return ret; 1323 1324 btrfs_clear_sb_rdonly(fs_info->sb); 1325 1326 set_bit(BTRFS_FS_OPEN, &fs_info->flags); 1327 1328 /* 1329 * If we've gone from readonly -> read-write, we need to get our 1330 * sync/async discard lists in the right state. 1331 */ 1332 btrfs_discard_resume(fs_info); 1333 1334 return 0; 1335 } 1336 1337 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info) 1338 { 1339 /* 1340 * This also happens on 'umount -rf' or on shutdown, when the 1341 * filesystem is busy. 1342 */ 1343 cancel_work_sync(&fs_info->async_reclaim_work); 1344 cancel_work_sync(&fs_info->async_data_reclaim_work); 1345 1346 btrfs_discard_cleanup(fs_info); 1347 1348 /* Wait for the uuid_scan task to finish */ 1349 down(&fs_info->uuid_tree_rescan_sem); 1350 /* Avoid complains from lockdep et al. */ 1351 up(&fs_info->uuid_tree_rescan_sem); 1352 1353 btrfs_set_sb_rdonly(fs_info->sb); 1354 1355 /* 1356 * Setting SB_RDONLY will put the cleaner thread to sleep at the next 1357 * loop if it's already active. If it's already asleep, we'll leave 1358 * unused block groups on disk until we're mounted read-write again 1359 * unless we clean them up here. 1360 */ 1361 btrfs_delete_unused_bgs(fs_info); 1362 1363 /* 1364 * The cleaner task could be already running before we set the flag 1365 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make 1366 * sure that after we finish the remount, i.e. after we call 1367 * btrfs_commit_super(), the cleaner can no longer start a transaction 1368 * - either because it was dropping a dead root, running delayed iputs 1369 * or deleting an unused block group (the cleaner picked a block 1370 * group from the list of unused block groups before we were able to 1371 * in the previous call to btrfs_delete_unused_bgs()). 1372 */ 1373 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE); 1374 1375 /* 1376 * We've set the superblock to RO mode, so we might have made the 1377 * cleaner task sleep without running all pending delayed iputs. Go 1378 * through all the delayed iputs here, so that if an unmount happens 1379 * without remounting RW we don't end up at finishing close_ctree() 1380 * with a non-empty list of delayed iputs. 1381 */ 1382 btrfs_run_delayed_iputs(fs_info); 1383 1384 btrfs_dev_replace_suspend_for_unmount(fs_info); 1385 btrfs_scrub_cancel(fs_info); 1386 btrfs_pause_balance(fs_info); 1387 1388 /* 1389 * Pause the qgroup rescan worker if it is running. We don't want it to 1390 * be still running after we are in RO mode, as after that, by the time 1391 * we unmount, it might have left a transaction open, so we would leak 1392 * the transaction and/or crash. 1393 */ 1394 btrfs_qgroup_wait_for_completion(fs_info, false); 1395 1396 return btrfs_commit_super(fs_info); 1397 } 1398 1399 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) 1400 { 1401 fs_info->max_inline = ctx->max_inline; 1402 fs_info->commit_interval = ctx->commit_interval; 1403 fs_info->metadata_ratio = ctx->metadata_ratio; 1404 fs_info->thread_pool_size = ctx->thread_pool_size; 1405 fs_info->mount_opt = ctx->mount_opt; 1406 fs_info->compress_type = ctx->compress_type; 1407 fs_info->compress_level = ctx->compress_level; 1408 } 1409 1410 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) 1411 { 1412 ctx->max_inline = fs_info->max_inline; 1413 ctx->commit_interval = fs_info->commit_interval; 1414 ctx->metadata_ratio = fs_info->metadata_ratio; 1415 ctx->thread_pool_size = fs_info->thread_pool_size; 1416 ctx->mount_opt = fs_info->mount_opt; 1417 ctx->compress_type = fs_info->compress_type; 1418 ctx->compress_level = fs_info->compress_level; 1419 } 1420 1421 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \ 1422 do { \ 1423 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ 1424 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ 1425 btrfs_info(fs_info, fmt, ##args); \ 1426 } while (0) 1427 1428 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \ 1429 do { \ 1430 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ 1431 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ 1432 btrfs_info(fs_info, fmt, ##args); \ 1433 } while (0) 1434 1435 static void btrfs_emit_options(struct btrfs_fs_info *info, 1436 struct btrfs_fs_context *old) 1437 { 1438 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum"); 1439 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts"); 1440 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum"); 1441 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations"); 1442 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme"); 1443 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers"); 1444 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log"); 1445 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time"); 1446 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit"); 1447 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard"); 1448 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard"); 1449 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree"); 1450 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching"); 1451 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache"); 1452 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag"); 1453 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data"); 1454 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata"); 1455 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification"); 1456 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time"); 1457 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots"); 1458 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums"); 1459 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums"); 1460 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags"); 1461 1462 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow"); 1463 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations"); 1464 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme"); 1465 btrfs_info_if_unset(info, old, NOBARRIER, "turning off barriers"); 1466 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log"); 1467 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching"); 1468 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree"); 1469 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag"); 1470 btrfs_info_if_unset(info, old, COMPRESS, "use no compression"); 1471 1472 /* Did the compression settings change? */ 1473 if (btrfs_test_opt(info, COMPRESS) && 1474 (!old || 1475 old->compress_type != info->compress_type || 1476 old->compress_level != info->compress_level || 1477 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) && 1478 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) { 1479 const char *compress_type = btrfs_compress_type2str(info->compress_type); 1480 1481 btrfs_info(info, "%s %s compression, level %d", 1482 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use", 1483 compress_type, info->compress_level); 1484 } 1485 1486 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) 1487 btrfs_info(info, "max_inline set to %llu", info->max_inline); 1488 } 1489 1490 static int btrfs_reconfigure(struct fs_context *fc) 1491 { 1492 struct super_block *sb = fc->root->d_sb; 1493 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1494 struct btrfs_fs_context *ctx = fc->fs_private; 1495 struct btrfs_fs_context old_ctx; 1496 int ret = 0; 1497 bool mount_reconfigure = (fc->s_fs_info != NULL); 1498 1499 btrfs_info_to_ctx(fs_info, &old_ctx); 1500 1501 /* 1502 * This is our "bind mount" trick, we don't want to allow the user to do 1503 * anything other than mount a different ro/rw and a different subvol, 1504 * all of the mount options should be maintained. 1505 */ 1506 if (mount_reconfigure) 1507 ctx->mount_opt = old_ctx.mount_opt; 1508 1509 sync_filesystem(sb); 1510 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1511 1512 if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags)) 1513 return -EINVAL; 1514 1515 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY)); 1516 if (ret < 0) 1517 return ret; 1518 1519 btrfs_ctx_to_info(fs_info, ctx); 1520 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags); 1521 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size, 1522 old_ctx.thread_pool_size); 1523 1524 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) != 1525 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 1526 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) { 1527 btrfs_warn(fs_info, 1528 "remount supports changing free space tree only from RO to RW"); 1529 /* Make sure free space cache options match the state on disk. */ 1530 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 1531 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1532 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 1533 } 1534 if (btrfs_free_space_cache_v1_active(fs_info)) { 1535 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1536 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); 1537 } 1538 } 1539 1540 ret = 0; 1541 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY)) 1542 ret = btrfs_remount_ro(fs_info); 1543 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY)) 1544 ret = btrfs_remount_rw(fs_info); 1545 if (ret) 1546 goto restore; 1547 1548 /* 1549 * If we set the mask during the parameter parsing VFS would reject the 1550 * remount. Here we can set the mask and the value will be updated 1551 * appropriately. 1552 */ 1553 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL)) 1554 fc->sb_flags_mask |= SB_POSIXACL; 1555 1556 btrfs_emit_options(fs_info, &old_ctx); 1557 wake_up_process(fs_info->transaction_kthread); 1558 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); 1559 btrfs_clear_oneshot_options(fs_info); 1560 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1561 1562 return 0; 1563 restore: 1564 btrfs_ctx_to_info(fs_info, &old_ctx); 1565 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); 1566 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1567 return ret; 1568 } 1569 1570 /* Used to sort the devices by max_avail(descending sort) */ 1571 static int btrfs_cmp_device_free_bytes(const void *a, const void *b) 1572 { 1573 const struct btrfs_device_info *dev_info1 = a; 1574 const struct btrfs_device_info *dev_info2 = b; 1575 1576 if (dev_info1->max_avail > dev_info2->max_avail) 1577 return -1; 1578 else if (dev_info1->max_avail < dev_info2->max_avail) 1579 return 1; 1580 return 0; 1581 } 1582 1583 /* 1584 * sort the devices by max_avail, in which max free extent size of each device 1585 * is stored.(Descending Sort) 1586 */ 1587 static inline void btrfs_descending_sort_devices( 1588 struct btrfs_device_info *devices, 1589 size_t nr_devices) 1590 { 1591 sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1592 btrfs_cmp_device_free_bytes, NULL); 1593 } 1594 1595 /* 1596 * The helper to calc the free space on the devices that can be used to store 1597 * file data. 1598 */ 1599 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info, 1600 u64 *free_bytes) 1601 { 1602 struct btrfs_device_info *devices_info; 1603 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 1604 struct btrfs_device *device; 1605 u64 type; 1606 u64 avail_space; 1607 u64 min_stripe_size; 1608 int num_stripes = 1; 1609 int i = 0, nr_devices; 1610 const struct btrfs_raid_attr *rattr; 1611 1612 /* 1613 * We aren't under the device list lock, so this is racy-ish, but good 1614 * enough for our purposes. 1615 */ 1616 nr_devices = fs_info->fs_devices->open_devices; 1617 if (!nr_devices) { 1618 smp_mb(); 1619 nr_devices = fs_info->fs_devices->open_devices; 1620 ASSERT(nr_devices); 1621 if (!nr_devices) { 1622 *free_bytes = 0; 1623 return 0; 1624 } 1625 } 1626 1627 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), 1628 GFP_KERNEL); 1629 if (!devices_info) 1630 return -ENOMEM; 1631 1632 /* calc min stripe number for data space allocation */ 1633 type = btrfs_data_alloc_profile(fs_info); 1634 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)]; 1635 1636 if (type & BTRFS_BLOCK_GROUP_RAID0) 1637 num_stripes = nr_devices; 1638 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK) 1639 num_stripes = rattr->ncopies; 1640 else if (type & BTRFS_BLOCK_GROUP_RAID10) 1641 num_stripes = 4; 1642 1643 /* Adjust for more than 1 stripe per device */ 1644 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN; 1645 1646 rcu_read_lock(); 1647 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 1648 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, 1649 &device->dev_state) || 1650 !device->bdev || 1651 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) 1652 continue; 1653 1654 if (i >= nr_devices) 1655 break; 1656 1657 avail_space = device->total_bytes - device->bytes_used; 1658 1659 /* align with stripe_len */ 1660 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN); 1661 1662 /* 1663 * Ensure we have at least min_stripe_size on top of the 1664 * reserved space on the device. 1665 */ 1666 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size) 1667 continue; 1668 1669 avail_space -= BTRFS_DEVICE_RANGE_RESERVED; 1670 1671 devices_info[i].dev = device; 1672 devices_info[i].max_avail = avail_space; 1673 1674 i++; 1675 } 1676 rcu_read_unlock(); 1677 1678 nr_devices = i; 1679 1680 btrfs_descending_sort_devices(devices_info, nr_devices); 1681 1682 i = nr_devices - 1; 1683 avail_space = 0; 1684 while (nr_devices >= rattr->devs_min) { 1685 num_stripes = min(num_stripes, nr_devices); 1686 1687 if (devices_info[i].max_avail >= min_stripe_size) { 1688 int j; 1689 u64 alloc_size; 1690 1691 avail_space += devices_info[i].max_avail * num_stripes; 1692 alloc_size = devices_info[i].max_avail; 1693 for (j = i + 1 - num_stripes; j <= i; j++) 1694 devices_info[j].max_avail -= alloc_size; 1695 } 1696 i--; 1697 nr_devices--; 1698 } 1699 1700 kfree(devices_info); 1701 *free_bytes = avail_space; 1702 return 0; 1703 } 1704 1705 /* 1706 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles. 1707 * 1708 * If there's a redundant raid level at DATA block groups, use the respective 1709 * multiplier to scale the sizes. 1710 * 1711 * Unused device space usage is based on simulating the chunk allocator 1712 * algorithm that respects the device sizes and order of allocations. This is 1713 * a close approximation of the actual use but there are other factors that may 1714 * change the result (like a new metadata chunk). 1715 * 1716 * If metadata is exhausted, f_bavail will be 0. 1717 */ 1718 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 1719 { 1720 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1721 struct btrfs_super_block *disk_super = fs_info->super_copy; 1722 struct btrfs_space_info *found; 1723 u64 total_used = 0; 1724 u64 total_free_data = 0; 1725 u64 total_free_meta = 0; 1726 u32 bits = fs_info->sectorsize_bits; 1727 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid; 1728 unsigned factor = 1; 1729 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 1730 int ret; 1731 u64 thresh = 0; 1732 int mixed = 0; 1733 1734 list_for_each_entry(found, &fs_info->space_info, list) { 1735 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 1736 int i; 1737 1738 total_free_data += found->disk_total - found->disk_used; 1739 total_free_data -= 1740 btrfs_account_ro_block_groups_free_space(found); 1741 1742 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { 1743 if (!list_empty(&found->block_groups[i])) 1744 factor = btrfs_bg_type_to_factor( 1745 btrfs_raid_array[i].bg_flag); 1746 } 1747 } 1748 1749 /* 1750 * Metadata in mixed block group profiles are accounted in data 1751 */ 1752 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) { 1753 if (found->flags & BTRFS_BLOCK_GROUP_DATA) 1754 mixed = 1; 1755 else 1756 total_free_meta += found->disk_total - 1757 found->disk_used; 1758 } 1759 1760 total_used += found->disk_used; 1761 } 1762 1763 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor); 1764 buf->f_blocks >>= bits; 1765 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits); 1766 1767 /* Account global block reserve as used, it's in logical size already */ 1768 spin_lock(&block_rsv->lock); 1769 /* Mixed block groups accounting is not byte-accurate, avoid overflow */ 1770 if (buf->f_bfree >= block_rsv->size >> bits) 1771 buf->f_bfree -= block_rsv->size >> bits; 1772 else 1773 buf->f_bfree = 0; 1774 spin_unlock(&block_rsv->lock); 1775 1776 buf->f_bavail = div_u64(total_free_data, factor); 1777 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data); 1778 if (ret) 1779 return ret; 1780 buf->f_bavail += div_u64(total_free_data, factor); 1781 buf->f_bavail = buf->f_bavail >> bits; 1782 1783 /* 1784 * We calculate the remaining metadata space minus global reserve. If 1785 * this is (supposedly) smaller than zero, there's no space. But this 1786 * does not hold in practice, the exhausted state happens where's still 1787 * some positive delta. So we apply some guesswork and compare the 1788 * delta to a 4M threshold. (Practically observed delta was ~2M.) 1789 * 1790 * We probably cannot calculate the exact threshold value because this 1791 * depends on the internal reservations requested by various 1792 * operations, so some operations that consume a few metadata will 1793 * succeed even if the Avail is zero. But this is better than the other 1794 * way around. 1795 */ 1796 thresh = SZ_4M; 1797 1798 /* 1799 * We only want to claim there's no available space if we can no longer 1800 * allocate chunks for our metadata profile and our global reserve will 1801 * not fit in the free metadata space. If we aren't ->full then we 1802 * still can allocate chunks and thus are fine using the currently 1803 * calculated f_bavail. 1804 */ 1805 if (!mixed && block_rsv->space_info->full && 1806 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size)) 1807 buf->f_bavail = 0; 1808 1809 buf->f_type = BTRFS_SUPER_MAGIC; 1810 buf->f_bsize = fs_info->sectorsize; 1811 buf->f_namelen = BTRFS_NAME_LEN; 1812 1813 /* We treat it as constant endianness (it doesn't matter _which_) 1814 because we want the fsid to come out the same whether mounted 1815 on a big-endian or little-endian host */ 1816 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 1817 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 1818 /* Mask in the root object ID too, to disambiguate subvols */ 1819 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32; 1820 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root); 1821 1822 return 0; 1823 } 1824 1825 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc) 1826 { 1827 struct btrfs_fs_info *p = fc->s_fs_info; 1828 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1829 1830 return fs_info->fs_devices == p->fs_devices; 1831 } 1832 1833 static int btrfs_get_tree_super(struct fs_context *fc) 1834 { 1835 struct btrfs_fs_info *fs_info = fc->s_fs_info; 1836 struct btrfs_fs_context *ctx = fc->fs_private; 1837 struct btrfs_fs_devices *fs_devices = NULL; 1838 struct btrfs_device *device; 1839 struct super_block *sb; 1840 blk_mode_t mode = sb_open_mode(fc->sb_flags); 1841 int ret; 1842 1843 btrfs_ctx_to_info(fs_info, ctx); 1844 mutex_lock(&uuid_mutex); 1845 1846 /* 1847 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect 1848 * either a valid device or an error. 1849 */ 1850 device = btrfs_scan_one_device(fc->source, true); 1851 ASSERT(device != NULL); 1852 if (IS_ERR(device)) { 1853 mutex_unlock(&uuid_mutex); 1854 return PTR_ERR(device); 1855 } 1856 fs_devices = device->fs_devices; 1857 /* 1858 * We cannot hold uuid_mutex calling sget_fc(), it will lead to a 1859 * locking order reversal with s_umount. 1860 * 1861 * So here we increase the holding number of fs_devices, this will ensure 1862 * the fs_devices itself won't be freed. 1863 */ 1864 btrfs_fs_devices_inc_holding(fs_devices); 1865 fs_info->fs_devices = fs_devices; 1866 mutex_unlock(&uuid_mutex); 1867 1868 1869 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc); 1870 if (IS_ERR(sb)) { 1871 mutex_lock(&uuid_mutex); 1872 btrfs_fs_devices_dec_holding(fs_devices); 1873 /* 1874 * Since the fs_devices is not opened, it can be freed at any 1875 * time after unlocking uuid_mutex. We need to avoid double 1876 * free through put_fs_context()->btrfs_free_fs_info(). 1877 * So here we reset fs_info->fs_devices to NULL, and let the 1878 * regular fs_devices reclaim path to handle it. 1879 * 1880 * This applies to all later branches where no fs_devices is 1881 * opened. 1882 */ 1883 fs_info->fs_devices = NULL; 1884 mutex_unlock(&uuid_mutex); 1885 return PTR_ERR(sb); 1886 } 1887 1888 set_device_specific_options(fs_info); 1889 1890 if (sb->s_root) { 1891 /* 1892 * Not the first mount of the fs thus got an existing super block. 1893 * Will reuse the returned super block, fs_info and fs_devices. 1894 * 1895 * fc->s_fs_info is not touched and will be later freed by 1896 * put_fs_context() through btrfs_free_fs_context(). 1897 */ 1898 ASSERT(fc->s_fs_info == fs_info); 1899 1900 mutex_lock(&uuid_mutex); 1901 btrfs_fs_devices_dec_holding(fs_devices); 1902 fs_info->fs_devices = NULL; 1903 mutex_unlock(&uuid_mutex); 1904 /* 1905 * At this stage we may have RO flag mismatch between 1906 * fc->sb_flags and sb->s_flags. Caller should detect such 1907 * mismatch and reconfigure with sb->s_umount rwsem held if 1908 * needed. 1909 */ 1910 } else { 1911 struct block_device *bdev; 1912 1913 /* 1914 * The first mount of the fs thus a new superblock, fc->s_fs_info 1915 * must be NULL, and the ownership of our fs_info and fs_devices is 1916 * transferred to the super block. 1917 */ 1918 ASSERT(fc->s_fs_info == NULL); 1919 1920 mutex_lock(&uuid_mutex); 1921 btrfs_fs_devices_dec_holding(fs_devices); 1922 ret = btrfs_open_devices(fs_devices, mode, sb); 1923 if (ret < 0) 1924 fs_info->fs_devices = NULL; 1925 mutex_unlock(&uuid_mutex); 1926 if (ret < 0) { 1927 deactivate_locked_super(sb); 1928 return ret; 1929 } 1930 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) { 1931 deactivate_locked_super(sb); 1932 return -EACCES; 1933 } 1934 bdev = fs_devices->latest_dev->bdev; 1935 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev); 1936 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id); 1937 ret = btrfs_fill_super(sb, fs_devices); 1938 if (ret) { 1939 deactivate_locked_super(sb); 1940 return ret; 1941 } 1942 } 1943 1944 btrfs_clear_oneshot_options(fs_info); 1945 1946 fc->root = dget(sb->s_root); 1947 return 0; 1948 } 1949 1950 /* 1951 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes 1952 * with different ro/rw options") the following works: 1953 * 1954 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo 1955 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar 1956 * 1957 * which looks nice and innocent but is actually pretty intricate and deserves 1958 * a long comment. 1959 * 1960 * On another filesystem a subvolume mount is close to something like: 1961 * 1962 * (iii) # create rw superblock + initial mount 1963 * mount -t xfs /dev/sdb /opt/ 1964 * 1965 * # create ro bind mount 1966 * mount --bind -o ro /opt/foo /mnt/foo 1967 * 1968 * # unmount initial mount 1969 * umount /opt 1970 * 1971 * Of course, there's some special subvolume sauce and there's the fact that the 1972 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually 1973 * it's very close and will help us understand the issue. 1974 * 1975 * The old mount API didn't cleanly distinguish between a mount being made ro 1976 * and a superblock being made ro. The only way to change the ro state of 1977 * either object was by passing ms_rdonly. If a new mount was created via 1978 * mount(2) such as: 1979 * 1980 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null); 1981 * 1982 * the MS_RDONLY flag being specified had two effects: 1983 * 1984 * (1) MNT_READONLY was raised -> the resulting mount got 1985 * @mnt->mnt_flags |= MNT_READONLY raised. 1986 * 1987 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems 1988 * made the superblock ro. Note, how SB_RDONLY has the same value as 1989 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2). 1990 * 1991 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a 1992 * subtree mounted ro. 1993 * 1994 * But consider the effect on the old mount API on btrfs subvolume mounting 1995 * which combines the distinct step in (iii) into a single step. 1996 * 1997 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii) 1998 * is issued the superblock is ro and thus even if the mount created for (ii) is 1999 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro 2000 * to rw for (ii) which it did using an internal remount call. 2001 * 2002 * IOW, subvolume mounting was inherently complicated due to the ambiguity of 2003 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate 2004 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when 2005 * passed by mount(8) to mount(2). 2006 * 2007 * Enter the new mount API. The new mount API disambiguates making a mount ro 2008 * and making a superblock ro. 2009 * 2010 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either 2011 * fsmount() or mount_setattr() this is a pure VFS level change for a 2012 * specific mount or mount tree that is never seen by the filesystem itself. 2013 * 2014 * (4) To turn a superblock ro the "ro" flag must be used with 2015 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem 2016 * in fc->sb_flags. 2017 * 2018 * But, currently the util-linux mount command already utilizes the new mount 2019 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's 2020 * btrfs or not, setting the whole super block RO. To make per-subvolume mounting 2021 * work with different options work we need to keep backward compatibility. 2022 */ 2023 static int btrfs_reconfigure_for_mount(struct fs_context *fc) 2024 { 2025 int ret = 0; 2026 2027 if (!(fc->sb_flags & SB_RDONLY) && (fc->root->d_sb->s_flags & SB_RDONLY)) 2028 ret = btrfs_reconfigure(fc); 2029 2030 return ret; 2031 } 2032 2033 static int btrfs_get_tree_subvol(struct fs_context *fc) 2034 { 2035 struct btrfs_fs_info *fs_info = NULL; 2036 struct btrfs_fs_context *ctx = fc->fs_private; 2037 struct fs_context *dup_fc; 2038 struct dentry *dentry; 2039 struct vfsmount *mnt; 2040 int ret = 0; 2041 2042 /* 2043 * Setup a dummy root and fs_info for test/set super. This is because 2044 * we don't actually fill this stuff out until open_ctree, but we need 2045 * then open_ctree will properly initialize the file system specific 2046 * settings later. btrfs_init_fs_info initializes the static elements 2047 * of the fs_info (locks and such) to make cleanup easier if we find a 2048 * superblock with our given fs_devices later on at sget() time. 2049 */ 2050 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL); 2051 if (!fs_info) 2052 return -ENOMEM; 2053 2054 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 2055 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 2056 if (!fs_info->super_copy || !fs_info->super_for_commit) { 2057 btrfs_free_fs_info(fs_info); 2058 return -ENOMEM; 2059 } 2060 btrfs_init_fs_info(fs_info); 2061 2062 dup_fc = vfs_dup_fs_context(fc); 2063 if (IS_ERR(dup_fc)) { 2064 btrfs_free_fs_info(fs_info); 2065 return PTR_ERR(dup_fc); 2066 } 2067 2068 /* 2069 * When we do the sget_fc this gets transferred to the sb, so we only 2070 * need to set it on the dup_fc as this is what creates the super block. 2071 */ 2072 dup_fc->s_fs_info = fs_info; 2073 2074 ret = btrfs_get_tree_super(dup_fc); 2075 if (ret) 2076 goto error; 2077 2078 ret = btrfs_reconfigure_for_mount(dup_fc); 2079 up_write(&dup_fc->root->d_sb->s_umount); 2080 if (ret) 2081 goto error; 2082 mnt = vfs_create_mount(dup_fc); 2083 put_fs_context(dup_fc); 2084 if (IS_ERR(mnt)) 2085 return PTR_ERR(mnt); 2086 2087 /* 2088 * This free's ->subvol_name, because if it isn't set we have to 2089 * allocate a buffer to hold the subvol_name, so we just drop our 2090 * reference to it here. 2091 */ 2092 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt); 2093 ctx->subvol_name = NULL; 2094 if (IS_ERR(dentry)) 2095 return PTR_ERR(dentry); 2096 2097 fc->root = dentry; 2098 return 0; 2099 error: 2100 put_fs_context(dup_fc); 2101 return ret; 2102 } 2103 2104 static int btrfs_get_tree(struct fs_context *fc) 2105 { 2106 ASSERT(fc->s_fs_info == NULL); 2107 2108 return btrfs_get_tree_subvol(fc); 2109 } 2110 2111 static void btrfs_kill_super(struct super_block *sb) 2112 { 2113 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2114 kill_anon_super(sb); 2115 btrfs_free_fs_info(fs_info); 2116 } 2117 2118 static void btrfs_free_fs_context(struct fs_context *fc) 2119 { 2120 struct btrfs_fs_context *ctx = fc->fs_private; 2121 struct btrfs_fs_info *fs_info = fc->s_fs_info; 2122 2123 if (fs_info) 2124 btrfs_free_fs_info(fs_info); 2125 2126 if (ctx && refcount_dec_and_test(&ctx->refs)) { 2127 kfree(ctx->subvol_name); 2128 kfree(ctx); 2129 } 2130 } 2131 2132 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc) 2133 { 2134 struct btrfs_fs_context *ctx = src_fc->fs_private; 2135 2136 /* 2137 * Give a ref to our ctx to this dup, as we want to keep it around for 2138 * our original fc so we can have the subvolume name or objectid. 2139 * 2140 * We unset ->source in the original fc because the dup needs it for 2141 * mounting, and then once we free the dup it'll free ->source, so we 2142 * need to make sure we're only pointing to it in one fc. 2143 */ 2144 refcount_inc(&ctx->refs); 2145 fc->fs_private = ctx; 2146 fc->source = src_fc->source; 2147 src_fc->source = NULL; 2148 return 0; 2149 } 2150 2151 static const struct fs_context_operations btrfs_fs_context_ops = { 2152 .parse_param = btrfs_parse_param, 2153 .reconfigure = btrfs_reconfigure, 2154 .get_tree = btrfs_get_tree, 2155 .dup = btrfs_dup_fs_context, 2156 .free = btrfs_free_fs_context, 2157 }; 2158 2159 static int btrfs_init_fs_context(struct fs_context *fc) 2160 { 2161 struct btrfs_fs_context *ctx; 2162 2163 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL); 2164 if (!ctx) 2165 return -ENOMEM; 2166 2167 refcount_set(&ctx->refs, 1); 2168 fc->fs_private = ctx; 2169 fc->ops = &btrfs_fs_context_ops; 2170 2171 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2172 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx); 2173 } else { 2174 ctx->thread_pool_size = 2175 min_t(unsigned long, num_online_cpus() + 2, 8); 2176 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE; 2177 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2178 } 2179 2180 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 2181 fc->sb_flags |= SB_POSIXACL; 2182 #endif 2183 fc->sb_flags |= SB_I_VERSION; 2184 2185 return 0; 2186 } 2187 2188 static struct file_system_type btrfs_fs_type = { 2189 .owner = THIS_MODULE, 2190 .name = "btrfs", 2191 .init_fs_context = btrfs_init_fs_context, 2192 .parameters = btrfs_fs_parameters, 2193 .kill_sb = btrfs_kill_super, 2194 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | 2195 FS_ALLOW_IDMAP | FS_MGTIME, 2196 }; 2197 2198 MODULE_ALIAS_FS("btrfs"); 2199 2200 static int btrfs_control_open(struct inode *inode, struct file *file) 2201 { 2202 /* 2203 * The control file's private_data is used to hold the 2204 * transaction when it is started and is used to keep 2205 * track of whether a transaction is already in progress. 2206 */ 2207 file->private_data = NULL; 2208 return 0; 2209 } 2210 2211 /* 2212 * Used by /dev/btrfs-control for devices ioctls. 2213 */ 2214 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 2215 unsigned long arg) 2216 { 2217 struct btrfs_ioctl_vol_args *vol; 2218 struct btrfs_device *device = NULL; 2219 dev_t devt = 0; 2220 int ret = -ENOTTY; 2221 2222 if (!capable(CAP_SYS_ADMIN)) 2223 return -EPERM; 2224 2225 vol = memdup_user((void __user *)arg, sizeof(*vol)); 2226 if (IS_ERR(vol)) 2227 return PTR_ERR(vol); 2228 ret = btrfs_check_ioctl_vol_args_path(vol); 2229 if (ret < 0) 2230 goto out; 2231 2232 switch (cmd) { 2233 case BTRFS_IOC_SCAN_DEV: 2234 mutex_lock(&uuid_mutex); 2235 /* 2236 * Scanning outside of mount can return NULL which would turn 2237 * into 0 error code. 2238 */ 2239 device = btrfs_scan_one_device(vol->name, false); 2240 ret = PTR_ERR_OR_ZERO(device); 2241 mutex_unlock(&uuid_mutex); 2242 break; 2243 case BTRFS_IOC_FORGET_DEV: 2244 if (vol->name[0] != 0) { 2245 ret = lookup_bdev(vol->name, &devt); 2246 if (ret) 2247 break; 2248 } 2249 ret = btrfs_forget_devices(devt); 2250 break; 2251 case BTRFS_IOC_DEVICES_READY: 2252 mutex_lock(&uuid_mutex); 2253 /* 2254 * Scanning outside of mount can return NULL which would turn 2255 * into 0 error code. 2256 */ 2257 device = btrfs_scan_one_device(vol->name, false); 2258 if (IS_ERR_OR_NULL(device)) { 2259 mutex_unlock(&uuid_mutex); 2260 if (IS_ERR(device)) 2261 ret = PTR_ERR(device); 2262 else 2263 ret = 0; 2264 break; 2265 } 2266 ret = !(device->fs_devices->num_devices == 2267 device->fs_devices->total_devices); 2268 mutex_unlock(&uuid_mutex); 2269 break; 2270 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 2271 ret = btrfs_ioctl_get_supported_features((void __user*)arg); 2272 break; 2273 } 2274 2275 out: 2276 kfree(vol); 2277 return ret; 2278 } 2279 2280 static int btrfs_freeze(struct super_block *sb) 2281 { 2282 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2283 2284 set_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2285 /* 2286 * We don't need a barrier here, we'll wait for any transaction that 2287 * could be in progress on other threads (and do delayed iputs that 2288 * we want to avoid on a frozen filesystem), or do the commit 2289 * ourselves. 2290 */ 2291 return btrfs_commit_current_transaction(fs_info->tree_root); 2292 } 2293 2294 static int check_dev_super(struct btrfs_device *dev) 2295 { 2296 struct btrfs_fs_info *fs_info = dev->fs_info; 2297 struct btrfs_super_block *sb; 2298 u64 last_trans; 2299 u16 csum_type; 2300 int ret = 0; 2301 2302 /* This should be called with fs still frozen. */ 2303 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags)); 2304 2305 /* Missing dev, no need to check. */ 2306 if (!dev->bdev) 2307 return 0; 2308 2309 /* Only need to check the primary super block. */ 2310 sb = btrfs_read_disk_super(dev->bdev, 0, true); 2311 if (IS_ERR(sb)) 2312 return PTR_ERR(sb); 2313 2314 /* Verify the checksum. */ 2315 csum_type = btrfs_super_csum_type(sb); 2316 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) { 2317 btrfs_err(fs_info, "csum type changed, has %u expect %u", 2318 csum_type, btrfs_super_csum_type(fs_info->super_copy)); 2319 ret = -EUCLEAN; 2320 goto out; 2321 } 2322 2323 if (btrfs_check_super_csum(fs_info, sb)) { 2324 btrfs_err(fs_info, "csum for on-disk super block no longer matches"); 2325 ret = -EUCLEAN; 2326 goto out; 2327 } 2328 2329 /* Btrfs_validate_super() includes fsid check against super->fsid. */ 2330 ret = btrfs_validate_super(fs_info, sb, 0); 2331 if (ret < 0) 2332 goto out; 2333 2334 last_trans = btrfs_get_last_trans_committed(fs_info); 2335 if (btrfs_super_generation(sb) != last_trans) { 2336 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu", 2337 btrfs_super_generation(sb), last_trans); 2338 ret = -EUCLEAN; 2339 goto out; 2340 } 2341 out: 2342 btrfs_release_disk_super(sb); 2343 return ret; 2344 } 2345 2346 static int btrfs_unfreeze(struct super_block *sb) 2347 { 2348 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2349 struct btrfs_device *device; 2350 int ret = 0; 2351 2352 /* 2353 * Make sure the fs is not changed by accident (like hibernation then 2354 * modified by other OS). 2355 * If we found anything wrong, we mark the fs error immediately. 2356 * 2357 * And since the fs is frozen, no one can modify the fs yet, thus 2358 * we don't need to hold device_list_mutex. 2359 */ 2360 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) { 2361 ret = check_dev_super(device); 2362 if (ret < 0) { 2363 btrfs_handle_fs_error(fs_info, ret, 2364 "super block on devid %llu got modified unexpectedly", 2365 device->devid); 2366 break; 2367 } 2368 } 2369 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2370 2371 /* 2372 * We still return 0, to allow VFS layer to unfreeze the fs even the 2373 * above checks failed. Since the fs is either fine or read-only, we're 2374 * safe to continue, without causing further damage. 2375 */ 2376 return 0; 2377 } 2378 2379 static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 2380 { 2381 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 2382 2383 /* 2384 * There should be always a valid pointer in latest_dev, it may be stale 2385 * for a short moment in case it's being deleted but still valid until 2386 * the end of RCU grace period. 2387 */ 2388 rcu_read_lock(); 2389 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\"); 2390 rcu_read_unlock(); 2391 2392 return 0; 2393 } 2394 2395 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc) 2396 { 2397 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2398 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps); 2399 2400 trace_btrfs_extent_map_shrinker_count(fs_info, nr); 2401 2402 return nr; 2403 } 2404 2405 static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc) 2406 { 2407 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan); 2408 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2409 2410 btrfs_free_extent_maps(fs_info, nr_to_scan); 2411 2412 /* The extent map shrinker runs asynchronously, so always return 0. */ 2413 return 0; 2414 } 2415 2416 static const struct super_operations btrfs_super_ops = { 2417 .drop_inode = btrfs_drop_inode, 2418 .evict_inode = btrfs_evict_inode, 2419 .put_super = btrfs_put_super, 2420 .sync_fs = btrfs_sync_fs, 2421 .show_options = btrfs_show_options, 2422 .show_devname = btrfs_show_devname, 2423 .alloc_inode = btrfs_alloc_inode, 2424 .destroy_inode = btrfs_destroy_inode, 2425 .free_inode = btrfs_free_inode, 2426 .statfs = btrfs_statfs, 2427 .freeze_fs = btrfs_freeze, 2428 .unfreeze_fs = btrfs_unfreeze, 2429 .nr_cached_objects = btrfs_nr_cached_objects, 2430 .free_cached_objects = btrfs_free_cached_objects, 2431 }; 2432 2433 static const struct file_operations btrfs_ctl_fops = { 2434 .open = btrfs_control_open, 2435 .unlocked_ioctl = btrfs_control_ioctl, 2436 .compat_ioctl = compat_ptr_ioctl, 2437 .owner = THIS_MODULE, 2438 .llseek = noop_llseek, 2439 }; 2440 2441 static struct miscdevice btrfs_misc = { 2442 .minor = BTRFS_MINOR, 2443 .name = "btrfs-control", 2444 .fops = &btrfs_ctl_fops 2445 }; 2446 2447 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2448 MODULE_ALIAS("devname:btrfs-control"); 2449 2450 static int __init btrfs_interface_init(void) 2451 { 2452 return misc_register(&btrfs_misc); 2453 } 2454 2455 static __cold void btrfs_interface_exit(void) 2456 { 2457 misc_deregister(&btrfs_misc); 2458 } 2459 2460 static int __init btrfs_print_mod_info(void) 2461 { 2462 static const char options[] = "" 2463 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2464 ", experimental=on" 2465 #endif 2466 #ifdef CONFIG_BTRFS_DEBUG 2467 ", debug=on" 2468 #endif 2469 #ifdef CONFIG_BTRFS_ASSERT 2470 ", assert=on" 2471 #endif 2472 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 2473 ", ref-verify=on" 2474 #endif 2475 #ifdef CONFIG_BLK_DEV_ZONED 2476 ", zoned=yes" 2477 #else 2478 ", zoned=no" 2479 #endif 2480 #ifdef CONFIG_FS_VERITY 2481 ", fsverity=yes" 2482 #else 2483 ", fsverity=no" 2484 #endif 2485 ; 2486 2487 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2488 if (btrfs_get_mod_read_policy() == NULL) 2489 pr_info("Btrfs loaded%s\n", options); 2490 else 2491 pr_info("Btrfs loaded%s, read_policy=%s\n", 2492 options, btrfs_get_mod_read_policy()); 2493 #else 2494 pr_info("Btrfs loaded%s\n", options); 2495 #endif 2496 2497 return 0; 2498 } 2499 2500 static int register_btrfs(void) 2501 { 2502 return register_filesystem(&btrfs_fs_type); 2503 } 2504 2505 static void unregister_btrfs(void) 2506 { 2507 unregister_filesystem(&btrfs_fs_type); 2508 } 2509 2510 /* Helper structure for long init/exit functions. */ 2511 struct init_sequence { 2512 int (*init_func)(void); 2513 /* Can be NULL if the init_func doesn't need cleanup. */ 2514 void (*exit_func)(void); 2515 }; 2516 2517 static const struct init_sequence mod_init_seq[] = { 2518 { 2519 .init_func = btrfs_props_init, 2520 .exit_func = NULL, 2521 }, { 2522 .init_func = btrfs_init_sysfs, 2523 .exit_func = btrfs_exit_sysfs, 2524 }, { 2525 .init_func = btrfs_init_compress, 2526 .exit_func = btrfs_exit_compress, 2527 }, { 2528 .init_func = btrfs_init_cachep, 2529 .exit_func = btrfs_destroy_cachep, 2530 }, { 2531 .init_func = btrfs_init_dio, 2532 .exit_func = btrfs_destroy_dio, 2533 }, { 2534 .init_func = btrfs_transaction_init, 2535 .exit_func = btrfs_transaction_exit, 2536 }, { 2537 .init_func = btrfs_ctree_init, 2538 .exit_func = btrfs_ctree_exit, 2539 }, { 2540 .init_func = btrfs_free_space_init, 2541 .exit_func = btrfs_free_space_exit, 2542 }, { 2543 .init_func = btrfs_extent_state_init_cachep, 2544 .exit_func = btrfs_extent_state_free_cachep, 2545 }, { 2546 .init_func = extent_buffer_init_cachep, 2547 .exit_func = extent_buffer_free_cachep, 2548 }, { 2549 .init_func = btrfs_bioset_init, 2550 .exit_func = btrfs_bioset_exit, 2551 }, { 2552 .init_func = btrfs_extent_map_init, 2553 .exit_func = btrfs_extent_map_exit, 2554 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2555 }, { 2556 .init_func = btrfs_read_policy_init, 2557 .exit_func = NULL, 2558 #endif 2559 }, { 2560 .init_func = ordered_data_init, 2561 .exit_func = ordered_data_exit, 2562 }, { 2563 .init_func = btrfs_delayed_inode_init, 2564 .exit_func = btrfs_delayed_inode_exit, 2565 }, { 2566 .init_func = btrfs_auto_defrag_init, 2567 .exit_func = btrfs_auto_defrag_exit, 2568 }, { 2569 .init_func = btrfs_delayed_ref_init, 2570 .exit_func = btrfs_delayed_ref_exit, 2571 }, { 2572 .init_func = btrfs_prelim_ref_init, 2573 .exit_func = btrfs_prelim_ref_exit, 2574 }, { 2575 .init_func = btrfs_interface_init, 2576 .exit_func = btrfs_interface_exit, 2577 }, { 2578 .init_func = btrfs_print_mod_info, 2579 .exit_func = NULL, 2580 }, { 2581 .init_func = btrfs_run_sanity_tests, 2582 .exit_func = NULL, 2583 }, { 2584 .init_func = register_btrfs, 2585 .exit_func = unregister_btrfs, 2586 } 2587 }; 2588 2589 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)]; 2590 2591 static __always_inline void btrfs_exit_btrfs_fs(void) 2592 { 2593 int i; 2594 2595 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) { 2596 if (!mod_init_result[i]) 2597 continue; 2598 if (mod_init_seq[i].exit_func) 2599 mod_init_seq[i].exit_func(); 2600 mod_init_result[i] = false; 2601 } 2602 } 2603 2604 static void __exit exit_btrfs_fs(void) 2605 { 2606 btrfs_exit_btrfs_fs(); 2607 btrfs_cleanup_fs_uuids(); 2608 } 2609 2610 static int __init init_btrfs_fs(void) 2611 { 2612 int ret; 2613 int i; 2614 2615 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) { 2616 ASSERT(!mod_init_result[i]); 2617 ret = mod_init_seq[i].init_func(); 2618 if (ret < 0) { 2619 btrfs_exit_btrfs_fs(); 2620 return ret; 2621 } 2622 mod_init_result[i] = true; 2623 } 2624 return 0; 2625 } 2626 2627 late_initcall(init_btrfs_fs); 2628 module_exit(exit_btrfs_fs) 2629 2630 MODULE_DESCRIPTION("B-Tree File System (BTRFS)"); 2631 MODULE_LICENSE("GPL"); 2632 MODULE_SOFTDEP("pre: crc32c"); 2633 MODULE_SOFTDEP("pre: xxhash64"); 2634 MODULE_SOFTDEP("pre: sha256"); 2635 MODULE_SOFTDEP("pre: blake2b-256"); 2636