1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #include <linux/blkdev.h> 7 #include <linux/module.h> 8 #include <linux/fs.h> 9 #include <linux/pagemap.h> 10 #include <linux/highmem.h> 11 #include <linux/time.h> 12 #include <linux/init.h> 13 #include <linux/seq_file.h> 14 #include <linux/string.h> 15 #include <linux/backing-dev.h> 16 #include <linux/mount.h> 17 #include <linux/writeback.h> 18 #include <linux/statfs.h> 19 #include <linux/compat.h> 20 #include <linux/parser.h> 21 #include <linux/ctype.h> 22 #include <linux/namei.h> 23 #include <linux/miscdevice.h> 24 #include <linux/magic.h> 25 #include <linux/slab.h> 26 #include <linux/ratelimit.h> 27 #include <linux/crc32c.h> 28 #include <linux/btrfs.h> 29 #include <linux/security.h> 30 #include <linux/fs_parser.h> 31 #include "messages.h" 32 #include "delayed-inode.h" 33 #include "ctree.h" 34 #include "disk-io.h" 35 #include "transaction.h" 36 #include "btrfs_inode.h" 37 #include "direct-io.h" 38 #include "props.h" 39 #include "xattr.h" 40 #include "bio.h" 41 #include "export.h" 42 #include "compression.h" 43 #include "dev-replace.h" 44 #include "free-space-cache.h" 45 #include "backref.h" 46 #include "space-info.h" 47 #include "sysfs.h" 48 #include "zoned.h" 49 #include "tests/btrfs-tests.h" 50 #include "block-group.h" 51 #include "discard.h" 52 #include "qgroup.h" 53 #include "raid56.h" 54 #include "fs.h" 55 #include "accessors.h" 56 #include "defrag.h" 57 #include "dir-item.h" 58 #include "ioctl.h" 59 #include "scrub.h" 60 #include "verity.h" 61 #include "super.h" 62 #include "extent-tree.h" 63 #define CREATE_TRACE_POINTS 64 #include <trace/events/btrfs.h> 65 66 static const struct super_operations btrfs_super_ops; 67 static struct file_system_type btrfs_fs_type; 68 69 static void btrfs_put_super(struct super_block *sb) 70 { 71 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 72 73 btrfs_info(fs_info, "last unmount of filesystem %pU", fs_info->fs_devices->fsid); 74 close_ctree(fs_info); 75 } 76 77 /* Store the mount options related information. */ 78 struct btrfs_fs_context { 79 char *subvol_name; 80 u64 subvol_objectid; 81 u64 max_inline; 82 u32 commit_interval; 83 u32 metadata_ratio; 84 u32 thread_pool_size; 85 unsigned long long mount_opt; 86 unsigned long compress_type:4; 87 int compress_level; 88 refcount_t refs; 89 }; 90 91 static void btrfs_emit_options(struct btrfs_fs_info *info, 92 struct btrfs_fs_context *old); 93 94 enum { 95 Opt_acl, 96 Opt_clear_cache, 97 Opt_commit_interval, 98 Opt_compress, 99 Opt_compress_force, 100 Opt_compress_force_type, 101 Opt_compress_type, 102 Opt_degraded, 103 Opt_device, 104 Opt_fatal_errors, 105 Opt_flushoncommit, 106 Opt_max_inline, 107 Opt_barrier, 108 Opt_datacow, 109 Opt_datasum, 110 Opt_defrag, 111 Opt_discard, 112 Opt_discard_mode, 113 Opt_ratio, 114 Opt_rescan_uuid_tree, 115 Opt_skip_balance, 116 Opt_space_cache, 117 Opt_space_cache_version, 118 Opt_ssd, 119 Opt_ssd_spread, 120 Opt_subvol, 121 Opt_subvol_empty, 122 Opt_subvolid, 123 Opt_thread_pool, 124 Opt_treelog, 125 Opt_user_subvol_rm_allowed, 126 Opt_norecovery, 127 128 /* Rescue options */ 129 Opt_rescue, 130 Opt_usebackuproot, 131 132 /* Debugging options */ 133 Opt_enospc_debug, 134 #ifdef CONFIG_BTRFS_DEBUG 135 Opt_fragment, Opt_fragment_data, Opt_fragment_metadata, Opt_fragment_all, 136 #endif 137 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 138 Opt_ref_verify, 139 #endif 140 Opt_err, 141 }; 142 143 enum { 144 Opt_fatal_errors_panic, 145 Opt_fatal_errors_bug, 146 }; 147 148 static const struct constant_table btrfs_parameter_fatal_errors[] = { 149 { "panic", Opt_fatal_errors_panic }, 150 { "bug", Opt_fatal_errors_bug }, 151 {} 152 }; 153 154 enum { 155 Opt_discard_sync, 156 Opt_discard_async, 157 }; 158 159 static const struct constant_table btrfs_parameter_discard[] = { 160 { "sync", Opt_discard_sync }, 161 { "async", Opt_discard_async }, 162 {} 163 }; 164 165 enum { 166 Opt_space_cache_v1, 167 Opt_space_cache_v2, 168 }; 169 170 static const struct constant_table btrfs_parameter_space_cache[] = { 171 { "v1", Opt_space_cache_v1 }, 172 { "v2", Opt_space_cache_v2 }, 173 {} 174 }; 175 176 enum { 177 Opt_rescue_usebackuproot, 178 Opt_rescue_nologreplay, 179 Opt_rescue_ignorebadroots, 180 Opt_rescue_ignoredatacsums, 181 Opt_rescue_ignoremetacsums, 182 Opt_rescue_ignoresuperflags, 183 Opt_rescue_parameter_all, 184 }; 185 186 static const struct constant_table btrfs_parameter_rescue[] = { 187 { "usebackuproot", Opt_rescue_usebackuproot }, 188 { "nologreplay", Opt_rescue_nologreplay }, 189 { "ignorebadroots", Opt_rescue_ignorebadroots }, 190 { "ibadroots", Opt_rescue_ignorebadroots }, 191 { "ignoredatacsums", Opt_rescue_ignoredatacsums }, 192 { "ignoremetacsums", Opt_rescue_ignoremetacsums}, 193 { "ignoresuperflags", Opt_rescue_ignoresuperflags}, 194 { "idatacsums", Opt_rescue_ignoredatacsums }, 195 { "imetacsums", Opt_rescue_ignoremetacsums}, 196 { "isuperflags", Opt_rescue_ignoresuperflags}, 197 { "all", Opt_rescue_parameter_all }, 198 {} 199 }; 200 201 #ifdef CONFIG_BTRFS_DEBUG 202 enum { 203 Opt_fragment_parameter_data, 204 Opt_fragment_parameter_metadata, 205 Opt_fragment_parameter_all, 206 }; 207 208 static const struct constant_table btrfs_parameter_fragment[] = { 209 { "data", Opt_fragment_parameter_data }, 210 { "metadata", Opt_fragment_parameter_metadata }, 211 { "all", Opt_fragment_parameter_all }, 212 {} 213 }; 214 #endif 215 216 static const struct fs_parameter_spec btrfs_fs_parameters[] = { 217 fsparam_flag_no("acl", Opt_acl), 218 fsparam_flag_no("autodefrag", Opt_defrag), 219 fsparam_flag_no("barrier", Opt_barrier), 220 fsparam_flag("clear_cache", Opt_clear_cache), 221 fsparam_u32("commit", Opt_commit_interval), 222 fsparam_flag("compress", Opt_compress), 223 fsparam_string("compress", Opt_compress_type), 224 fsparam_flag("compress-force", Opt_compress_force), 225 fsparam_string("compress-force", Opt_compress_force_type), 226 fsparam_flag_no("datacow", Opt_datacow), 227 fsparam_flag_no("datasum", Opt_datasum), 228 fsparam_flag("degraded", Opt_degraded), 229 fsparam_string("device", Opt_device), 230 fsparam_flag_no("discard", Opt_discard), 231 fsparam_enum("discard", Opt_discard_mode, btrfs_parameter_discard), 232 fsparam_enum("fatal_errors", Opt_fatal_errors, btrfs_parameter_fatal_errors), 233 fsparam_flag_no("flushoncommit", Opt_flushoncommit), 234 fsparam_string("max_inline", Opt_max_inline), 235 fsparam_u32("metadata_ratio", Opt_ratio), 236 fsparam_flag("rescan_uuid_tree", Opt_rescan_uuid_tree), 237 fsparam_flag("skip_balance", Opt_skip_balance), 238 fsparam_flag_no("space_cache", Opt_space_cache), 239 fsparam_enum("space_cache", Opt_space_cache_version, btrfs_parameter_space_cache), 240 fsparam_flag_no("ssd", Opt_ssd), 241 fsparam_flag_no("ssd_spread", Opt_ssd_spread), 242 fsparam_string("subvol", Opt_subvol), 243 fsparam_flag("subvol=", Opt_subvol_empty), 244 fsparam_u64("subvolid", Opt_subvolid), 245 fsparam_u32("thread_pool", Opt_thread_pool), 246 fsparam_flag_no("treelog", Opt_treelog), 247 fsparam_flag("user_subvol_rm_allowed", Opt_user_subvol_rm_allowed), 248 249 /* Rescue options. */ 250 fsparam_enum("rescue", Opt_rescue, btrfs_parameter_rescue), 251 /* Deprecated, with alias rescue=usebackuproot */ 252 __fsparam(NULL, "usebackuproot", Opt_usebackuproot, fs_param_deprecated, NULL), 253 /* For compatibility only, alias for "rescue=nologreplay". */ 254 fsparam_flag("norecovery", Opt_norecovery), 255 256 /* Debugging options. */ 257 fsparam_flag_no("enospc_debug", Opt_enospc_debug), 258 #ifdef CONFIG_BTRFS_DEBUG 259 fsparam_enum("fragment", Opt_fragment, btrfs_parameter_fragment), 260 #endif 261 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 262 fsparam_flag("ref_verify", Opt_ref_verify), 263 #endif 264 {} 265 }; 266 267 static bool btrfs_match_compress_type(const char *string, const char *type, bool may_have_level) 268 { 269 const int len = strlen(type); 270 271 return (strncmp(string, type, len) == 0) && 272 ((may_have_level && string[len] == ':') || string[len] == '\0'); 273 } 274 275 static int btrfs_parse_compress(struct btrfs_fs_context *ctx, 276 const struct fs_parameter *param, int opt) 277 { 278 const char *string = param->string; 279 280 /* 281 * Provide the same semantics as older kernels that don't use fs 282 * context, specifying the "compress" option clears "force-compress" 283 * without the need to pass "compress-force=[no|none]" before 284 * specifying "compress". 285 */ 286 if (opt != Opt_compress_force && opt != Opt_compress_force_type) 287 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 288 289 if (opt == Opt_compress || opt == Opt_compress_force) { 290 ctx->compress_type = BTRFS_COMPRESS_ZLIB; 291 ctx->compress_level = BTRFS_ZLIB_DEFAULT_LEVEL; 292 btrfs_set_opt(ctx->mount_opt, COMPRESS); 293 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 294 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 295 } else if (btrfs_match_compress_type(string, "zlib", true)) { 296 ctx->compress_type = BTRFS_COMPRESS_ZLIB; 297 ctx->compress_level = btrfs_compress_str2level(BTRFS_COMPRESS_ZLIB, 298 string + 4); 299 btrfs_set_opt(ctx->mount_opt, COMPRESS); 300 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 301 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 302 } else if (btrfs_match_compress_type(string, "lzo", false)) { 303 ctx->compress_type = BTRFS_COMPRESS_LZO; 304 ctx->compress_level = 0; 305 btrfs_set_opt(ctx->mount_opt, COMPRESS); 306 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 307 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 308 } else if (btrfs_match_compress_type(string, "zstd", true)) { 309 ctx->compress_type = BTRFS_COMPRESS_ZSTD; 310 ctx->compress_level = btrfs_compress_str2level(BTRFS_COMPRESS_ZSTD, 311 string + 4); 312 btrfs_set_opt(ctx->mount_opt, COMPRESS); 313 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 314 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 315 } else if (btrfs_match_compress_type(string, "no", false) || 316 btrfs_match_compress_type(string, "none", false)) { 317 ctx->compress_level = 0; 318 ctx->compress_type = 0; 319 btrfs_clear_opt(ctx->mount_opt, COMPRESS); 320 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 321 } else { 322 btrfs_err(NULL, "unrecognized compression value %s", string); 323 return -EINVAL; 324 } 325 return 0; 326 } 327 328 static int btrfs_parse_param(struct fs_context *fc, struct fs_parameter *param) 329 { 330 struct btrfs_fs_context *ctx = fc->fs_private; 331 struct fs_parse_result result; 332 int opt; 333 334 opt = fs_parse(fc, btrfs_fs_parameters, param, &result); 335 if (opt < 0) 336 return opt; 337 338 switch (opt) { 339 case Opt_degraded: 340 btrfs_set_opt(ctx->mount_opt, DEGRADED); 341 break; 342 case Opt_subvol_empty: 343 /* 344 * This exists because we used to allow it on accident, so we're 345 * keeping it to maintain ABI. See 37becec95ac3 ("Btrfs: allow 346 * empty subvol= again"). 347 */ 348 break; 349 case Opt_subvol: 350 kfree(ctx->subvol_name); 351 ctx->subvol_name = kstrdup(param->string, GFP_KERNEL); 352 if (!ctx->subvol_name) 353 return -ENOMEM; 354 break; 355 case Opt_subvolid: 356 ctx->subvol_objectid = result.uint_64; 357 358 /* subvolid=0 means give me the original fs_tree. */ 359 if (!ctx->subvol_objectid) 360 ctx->subvol_objectid = BTRFS_FS_TREE_OBJECTID; 361 break; 362 case Opt_device: { 363 struct btrfs_device *device; 364 365 mutex_lock(&uuid_mutex); 366 device = btrfs_scan_one_device(param->string, false); 367 mutex_unlock(&uuid_mutex); 368 if (IS_ERR(device)) 369 return PTR_ERR(device); 370 break; 371 } 372 case Opt_datasum: 373 if (result.negated) { 374 btrfs_set_opt(ctx->mount_opt, NODATASUM); 375 } else { 376 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 377 btrfs_clear_opt(ctx->mount_opt, NODATASUM); 378 } 379 break; 380 case Opt_datacow: 381 if (result.negated) { 382 btrfs_clear_opt(ctx->mount_opt, COMPRESS); 383 btrfs_clear_opt(ctx->mount_opt, FORCE_COMPRESS); 384 btrfs_set_opt(ctx->mount_opt, NODATACOW); 385 btrfs_set_opt(ctx->mount_opt, NODATASUM); 386 } else { 387 btrfs_clear_opt(ctx->mount_opt, NODATACOW); 388 } 389 break; 390 case Opt_compress_force: 391 case Opt_compress_force_type: 392 btrfs_set_opt(ctx->mount_opt, FORCE_COMPRESS); 393 fallthrough; 394 case Opt_compress: 395 case Opt_compress_type: 396 if (btrfs_parse_compress(ctx, param, opt)) 397 return -EINVAL; 398 break; 399 case Opt_ssd: 400 if (result.negated) { 401 btrfs_set_opt(ctx->mount_opt, NOSSD); 402 btrfs_clear_opt(ctx->mount_opt, SSD); 403 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); 404 } else { 405 btrfs_set_opt(ctx->mount_opt, SSD); 406 btrfs_clear_opt(ctx->mount_opt, NOSSD); 407 } 408 break; 409 case Opt_ssd_spread: 410 if (result.negated) { 411 btrfs_clear_opt(ctx->mount_opt, SSD_SPREAD); 412 } else { 413 btrfs_set_opt(ctx->mount_opt, SSD); 414 btrfs_set_opt(ctx->mount_opt, SSD_SPREAD); 415 btrfs_clear_opt(ctx->mount_opt, NOSSD); 416 } 417 break; 418 case Opt_barrier: 419 if (result.negated) 420 btrfs_set_opt(ctx->mount_opt, NOBARRIER); 421 else 422 btrfs_clear_opt(ctx->mount_opt, NOBARRIER); 423 break; 424 case Opt_thread_pool: 425 if (result.uint_32 == 0) { 426 btrfs_err(NULL, "invalid value 0 for thread_pool"); 427 return -EINVAL; 428 } 429 ctx->thread_pool_size = result.uint_32; 430 break; 431 case Opt_max_inline: 432 ctx->max_inline = memparse(param->string, NULL); 433 break; 434 case Opt_acl: 435 if (result.negated) { 436 fc->sb_flags &= ~SB_POSIXACL; 437 } else { 438 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 439 fc->sb_flags |= SB_POSIXACL; 440 #else 441 btrfs_err(NULL, "support for ACL not compiled in"); 442 return -EINVAL; 443 #endif 444 } 445 /* 446 * VFS limits the ability to toggle ACL on and off via remount, 447 * despite every file system allowing this. This seems to be 448 * an oversight since we all do, but it'll fail if we're 449 * remounting. So don't set the mask here, we'll check it in 450 * btrfs_reconfigure and do the toggling ourselves. 451 */ 452 if (fc->purpose != FS_CONTEXT_FOR_RECONFIGURE) 453 fc->sb_flags_mask |= SB_POSIXACL; 454 break; 455 case Opt_treelog: 456 if (result.negated) 457 btrfs_set_opt(ctx->mount_opt, NOTREELOG); 458 else 459 btrfs_clear_opt(ctx->mount_opt, NOTREELOG); 460 break; 461 case Opt_norecovery: 462 btrfs_info(NULL, 463 "'norecovery' is for compatibility only, recommended to use 'rescue=nologreplay'"); 464 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); 465 break; 466 case Opt_flushoncommit: 467 if (result.negated) 468 btrfs_clear_opt(ctx->mount_opt, FLUSHONCOMMIT); 469 else 470 btrfs_set_opt(ctx->mount_opt, FLUSHONCOMMIT); 471 break; 472 case Opt_ratio: 473 ctx->metadata_ratio = result.uint_32; 474 break; 475 case Opt_discard: 476 if (result.negated) { 477 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC); 478 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); 479 btrfs_set_opt(ctx->mount_opt, NODISCARD); 480 } else { 481 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC); 482 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); 483 } 484 break; 485 case Opt_discard_mode: 486 switch (result.uint_32) { 487 case Opt_discard_sync: 488 btrfs_clear_opt(ctx->mount_opt, DISCARD_ASYNC); 489 btrfs_set_opt(ctx->mount_opt, DISCARD_SYNC); 490 break; 491 case Opt_discard_async: 492 btrfs_clear_opt(ctx->mount_opt, DISCARD_SYNC); 493 btrfs_set_opt(ctx->mount_opt, DISCARD_ASYNC); 494 break; 495 default: 496 btrfs_err(NULL, "unrecognized discard mode value %s", 497 param->key); 498 return -EINVAL; 499 } 500 btrfs_clear_opt(ctx->mount_opt, NODISCARD); 501 break; 502 case Opt_space_cache: 503 if (result.negated) { 504 btrfs_set_opt(ctx->mount_opt, NOSPACECACHE); 505 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE); 506 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); 507 } else { 508 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); 509 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE); 510 } 511 break; 512 case Opt_space_cache_version: 513 switch (result.uint_32) { 514 case Opt_space_cache_v1: 515 btrfs_set_opt(ctx->mount_opt, SPACE_CACHE); 516 btrfs_clear_opt(ctx->mount_opt, FREE_SPACE_TREE); 517 break; 518 case Opt_space_cache_v2: 519 btrfs_clear_opt(ctx->mount_opt, SPACE_CACHE); 520 btrfs_set_opt(ctx->mount_opt, FREE_SPACE_TREE); 521 break; 522 default: 523 btrfs_err(NULL, "unrecognized space_cache value %s", 524 param->key); 525 return -EINVAL; 526 } 527 break; 528 case Opt_rescan_uuid_tree: 529 btrfs_set_opt(ctx->mount_opt, RESCAN_UUID_TREE); 530 break; 531 case Opt_clear_cache: 532 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE); 533 break; 534 case Opt_user_subvol_rm_allowed: 535 btrfs_set_opt(ctx->mount_opt, USER_SUBVOL_RM_ALLOWED); 536 break; 537 case Opt_enospc_debug: 538 if (result.negated) 539 btrfs_clear_opt(ctx->mount_opt, ENOSPC_DEBUG); 540 else 541 btrfs_set_opt(ctx->mount_opt, ENOSPC_DEBUG); 542 break; 543 case Opt_defrag: 544 if (result.negated) 545 btrfs_clear_opt(ctx->mount_opt, AUTO_DEFRAG); 546 else 547 btrfs_set_opt(ctx->mount_opt, AUTO_DEFRAG); 548 break; 549 case Opt_usebackuproot: 550 btrfs_warn(NULL, 551 "'usebackuproot' is deprecated, use 'rescue=usebackuproot' instead"); 552 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT); 553 554 /* If we're loading the backup roots we can't trust the space cache. */ 555 btrfs_set_opt(ctx->mount_opt, CLEAR_CACHE); 556 break; 557 case Opt_skip_balance: 558 btrfs_set_opt(ctx->mount_opt, SKIP_BALANCE); 559 break; 560 case Opt_fatal_errors: 561 switch (result.uint_32) { 562 case Opt_fatal_errors_panic: 563 btrfs_set_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR); 564 break; 565 case Opt_fatal_errors_bug: 566 btrfs_clear_opt(ctx->mount_opt, PANIC_ON_FATAL_ERROR); 567 break; 568 default: 569 btrfs_err(NULL, "unrecognized fatal_errors value %s", 570 param->key); 571 return -EINVAL; 572 } 573 break; 574 case Opt_commit_interval: 575 ctx->commit_interval = result.uint_32; 576 if (ctx->commit_interval > BTRFS_WARNING_COMMIT_INTERVAL) { 577 btrfs_warn(NULL, "excessive commit interval %u, use with care", 578 ctx->commit_interval); 579 } 580 if (ctx->commit_interval == 0) 581 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 582 break; 583 case Opt_rescue: 584 switch (result.uint_32) { 585 case Opt_rescue_usebackuproot: 586 btrfs_set_opt(ctx->mount_opt, USEBACKUPROOT); 587 break; 588 case Opt_rescue_nologreplay: 589 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); 590 break; 591 case Opt_rescue_ignorebadroots: 592 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); 593 break; 594 case Opt_rescue_ignoredatacsums: 595 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); 596 break; 597 case Opt_rescue_ignoremetacsums: 598 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS); 599 break; 600 case Opt_rescue_ignoresuperflags: 601 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS); 602 break; 603 case Opt_rescue_parameter_all: 604 btrfs_set_opt(ctx->mount_opt, IGNOREDATACSUMS); 605 btrfs_set_opt(ctx->mount_opt, IGNOREMETACSUMS); 606 btrfs_set_opt(ctx->mount_opt, IGNORESUPERFLAGS); 607 btrfs_set_opt(ctx->mount_opt, IGNOREBADROOTS); 608 btrfs_set_opt(ctx->mount_opt, NOLOGREPLAY); 609 break; 610 default: 611 btrfs_info(NULL, "unrecognized rescue option '%s'", 612 param->key); 613 return -EINVAL; 614 } 615 break; 616 #ifdef CONFIG_BTRFS_DEBUG 617 case Opt_fragment: 618 switch (result.uint_32) { 619 case Opt_fragment_parameter_all: 620 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); 621 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); 622 break; 623 case Opt_fragment_parameter_metadata: 624 btrfs_set_opt(ctx->mount_opt, FRAGMENT_METADATA); 625 break; 626 case Opt_fragment_parameter_data: 627 btrfs_set_opt(ctx->mount_opt, FRAGMENT_DATA); 628 break; 629 default: 630 btrfs_info(NULL, "unrecognized fragment option '%s'", 631 param->key); 632 return -EINVAL; 633 } 634 break; 635 #endif 636 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 637 case Opt_ref_verify: 638 btrfs_set_opt(ctx->mount_opt, REF_VERIFY); 639 break; 640 #endif 641 default: 642 btrfs_err(NULL, "unrecognized mount option '%s'", param->key); 643 return -EINVAL; 644 } 645 646 return 0; 647 } 648 649 /* 650 * Some options only have meaning at mount time and shouldn't persist across 651 * remounts, or be displayed. Clear these at the end of mount and remount code 652 * paths. 653 */ 654 static void btrfs_clear_oneshot_options(struct btrfs_fs_info *fs_info) 655 { 656 btrfs_clear_opt(fs_info->mount_opt, USEBACKUPROOT); 657 btrfs_clear_opt(fs_info->mount_opt, CLEAR_CACHE); 658 btrfs_clear_opt(fs_info->mount_opt, NOSPACECACHE); 659 } 660 661 static bool check_ro_option(const struct btrfs_fs_info *fs_info, 662 unsigned long long mount_opt, unsigned long long opt, 663 const char *opt_name) 664 { 665 if (mount_opt & opt) { 666 btrfs_err(fs_info, "%s must be used with ro mount option", 667 opt_name); 668 return true; 669 } 670 return false; 671 } 672 673 bool btrfs_check_options(const struct btrfs_fs_info *info, 674 unsigned long long *mount_opt, 675 unsigned long flags) 676 { 677 bool ret = true; 678 679 if (!(flags & SB_RDONLY) && 680 (check_ro_option(info, *mount_opt, BTRFS_MOUNT_NOLOGREPLAY, "nologreplay") || 681 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREBADROOTS, "ignorebadroots") || 682 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREDATACSUMS, "ignoredatacsums") || 683 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNOREMETACSUMS, "ignoremetacsums") || 684 check_ro_option(info, *mount_opt, BTRFS_MOUNT_IGNORESUPERFLAGS, "ignoresuperflags"))) 685 ret = false; 686 687 if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE) && 688 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE) && 689 !btrfs_raw_test_opt(*mount_opt, CLEAR_CACHE)) { 690 btrfs_err(info, "cannot disable free-space-tree"); 691 ret = false; 692 } 693 if (btrfs_fs_compat_ro(info, BLOCK_GROUP_TREE) && 694 !btrfs_raw_test_opt(*mount_opt, FREE_SPACE_TREE)) { 695 btrfs_err(info, "cannot disable free-space-tree with block-group-tree feature"); 696 ret = false; 697 } 698 699 if (btrfs_check_mountopts_zoned(info, mount_opt)) 700 ret = false; 701 702 if (!test_bit(BTRFS_FS_STATE_REMOUNTING, &info->fs_state)) { 703 if (btrfs_raw_test_opt(*mount_opt, SPACE_CACHE)) { 704 btrfs_warn(info, 705 "space cache v1 is being deprecated and will be removed in a future release, please use -o space_cache=v2"); 706 } 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 btrfs_emit_options(fs_info, NULL); 984 985 inode = btrfs_iget(BTRFS_FIRST_FREE_OBJECTID, fs_info->fs_root); 986 if (IS_ERR(inode)) { 987 ret = PTR_ERR(inode); 988 btrfs_handle_fs_error(fs_info, ret, NULL); 989 goto fail_close; 990 } 991 992 sb->s_root = d_make_root(&inode->vfs_inode); 993 if (!sb->s_root) { 994 ret = -ENOMEM; 995 goto fail_close; 996 } 997 998 sb->s_flags |= SB_ACTIVE; 999 return 0; 1000 1001 fail_close: 1002 close_ctree(fs_info); 1003 return ret; 1004 } 1005 1006 int btrfs_sync_fs(struct super_block *sb, int wait) 1007 { 1008 struct btrfs_trans_handle *trans; 1009 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1010 struct btrfs_root *root = fs_info->tree_root; 1011 1012 trace_btrfs_sync_fs(fs_info, wait); 1013 1014 if (!wait) { 1015 filemap_flush(fs_info->btree_inode->i_mapping); 1016 return 0; 1017 } 1018 1019 btrfs_wait_ordered_roots(fs_info, U64_MAX, NULL); 1020 1021 trans = btrfs_attach_transaction_barrier(root); 1022 if (IS_ERR(trans)) { 1023 /* no transaction, don't bother */ 1024 if (PTR_ERR(trans) == -ENOENT) { 1025 /* 1026 * Exit unless we have some pending changes 1027 * that need to go through commit 1028 */ 1029 if (!test_bit(BTRFS_FS_NEED_TRANS_COMMIT, 1030 &fs_info->flags)) 1031 return 0; 1032 /* 1033 * A non-blocking test if the fs is frozen. We must not 1034 * start a new transaction here otherwise a deadlock 1035 * happens. The pending operations are delayed to the 1036 * next commit after thawing. 1037 */ 1038 if (sb_start_write_trylock(sb)) 1039 sb_end_write(sb); 1040 else 1041 return 0; 1042 trans = btrfs_start_transaction(root, 0); 1043 } 1044 if (IS_ERR(trans)) 1045 return PTR_ERR(trans); 1046 } 1047 return btrfs_commit_transaction(trans); 1048 } 1049 1050 static void print_rescue_option(struct seq_file *seq, const char *s, bool *printed) 1051 { 1052 seq_printf(seq, "%s%s", (*printed) ? ":" : ",rescue=", s); 1053 *printed = true; 1054 } 1055 1056 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 1057 { 1058 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 1059 const char *compress_type; 1060 const char *subvol_name; 1061 bool printed = false; 1062 1063 if (btrfs_test_opt(info, DEGRADED)) 1064 seq_puts(seq, ",degraded"); 1065 if (btrfs_test_opt(info, NODATASUM)) 1066 seq_puts(seq, ",nodatasum"); 1067 if (btrfs_test_opt(info, NODATACOW)) 1068 seq_puts(seq, ",nodatacow"); 1069 if (btrfs_test_opt(info, NOBARRIER)) 1070 seq_puts(seq, ",nobarrier"); 1071 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) 1072 seq_printf(seq, ",max_inline=%llu", info->max_inline); 1073 if (info->thread_pool_size != min_t(unsigned long, 1074 num_online_cpus() + 2, 8)) 1075 seq_printf(seq, ",thread_pool=%u", info->thread_pool_size); 1076 if (btrfs_test_opt(info, COMPRESS)) { 1077 compress_type = btrfs_compress_type2str(info->compress_type); 1078 if (btrfs_test_opt(info, FORCE_COMPRESS)) 1079 seq_printf(seq, ",compress-force=%s", compress_type); 1080 else 1081 seq_printf(seq, ",compress=%s", compress_type); 1082 if (info->compress_level) 1083 seq_printf(seq, ":%d", info->compress_level); 1084 } 1085 if (btrfs_test_opt(info, NOSSD)) 1086 seq_puts(seq, ",nossd"); 1087 if (btrfs_test_opt(info, SSD_SPREAD)) 1088 seq_puts(seq, ",ssd_spread"); 1089 else if (btrfs_test_opt(info, SSD)) 1090 seq_puts(seq, ",ssd"); 1091 if (btrfs_test_opt(info, NOTREELOG)) 1092 seq_puts(seq, ",notreelog"); 1093 if (btrfs_test_opt(info, NOLOGREPLAY)) 1094 print_rescue_option(seq, "nologreplay", &printed); 1095 if (btrfs_test_opt(info, USEBACKUPROOT)) 1096 print_rescue_option(seq, "usebackuproot", &printed); 1097 if (btrfs_test_opt(info, IGNOREBADROOTS)) 1098 print_rescue_option(seq, "ignorebadroots", &printed); 1099 if (btrfs_test_opt(info, IGNOREDATACSUMS)) 1100 print_rescue_option(seq, "ignoredatacsums", &printed); 1101 if (btrfs_test_opt(info, IGNOREMETACSUMS)) 1102 print_rescue_option(seq, "ignoremetacsums", &printed); 1103 if (btrfs_test_opt(info, IGNORESUPERFLAGS)) 1104 print_rescue_option(seq, "ignoresuperflags", &printed); 1105 if (btrfs_test_opt(info, FLUSHONCOMMIT)) 1106 seq_puts(seq, ",flushoncommit"); 1107 if (btrfs_test_opt(info, DISCARD_SYNC)) 1108 seq_puts(seq, ",discard"); 1109 if (btrfs_test_opt(info, DISCARD_ASYNC)) 1110 seq_puts(seq, ",discard=async"); 1111 if (!(info->sb->s_flags & SB_POSIXACL)) 1112 seq_puts(seq, ",noacl"); 1113 if (btrfs_free_space_cache_v1_active(info)) 1114 seq_puts(seq, ",space_cache"); 1115 else if (btrfs_fs_compat_ro(info, FREE_SPACE_TREE)) 1116 seq_puts(seq, ",space_cache=v2"); 1117 else 1118 seq_puts(seq, ",nospace_cache"); 1119 if (btrfs_test_opt(info, RESCAN_UUID_TREE)) 1120 seq_puts(seq, ",rescan_uuid_tree"); 1121 if (btrfs_test_opt(info, CLEAR_CACHE)) 1122 seq_puts(seq, ",clear_cache"); 1123 if (btrfs_test_opt(info, USER_SUBVOL_RM_ALLOWED)) 1124 seq_puts(seq, ",user_subvol_rm_allowed"); 1125 if (btrfs_test_opt(info, ENOSPC_DEBUG)) 1126 seq_puts(seq, ",enospc_debug"); 1127 if (btrfs_test_opt(info, AUTO_DEFRAG)) 1128 seq_puts(seq, ",autodefrag"); 1129 if (btrfs_test_opt(info, SKIP_BALANCE)) 1130 seq_puts(seq, ",skip_balance"); 1131 if (info->metadata_ratio) 1132 seq_printf(seq, ",metadata_ratio=%u", info->metadata_ratio); 1133 if (btrfs_test_opt(info, PANIC_ON_FATAL_ERROR)) 1134 seq_puts(seq, ",fatal_errors=panic"); 1135 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1136 seq_printf(seq, ",commit=%u", info->commit_interval); 1137 #ifdef CONFIG_BTRFS_DEBUG 1138 if (btrfs_test_opt(info, FRAGMENT_DATA)) 1139 seq_puts(seq, ",fragment=data"); 1140 if (btrfs_test_opt(info, FRAGMENT_METADATA)) 1141 seq_puts(seq, ",fragment=metadata"); 1142 #endif 1143 if (btrfs_test_opt(info, REF_VERIFY)) 1144 seq_puts(seq, ",ref_verify"); 1145 seq_printf(seq, ",subvolid=%llu", btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); 1146 subvol_name = btrfs_get_subvol_name_from_objectid(info, 1147 btrfs_root_id(BTRFS_I(d_inode(dentry))->root)); 1148 if (!IS_ERR(subvol_name)) { 1149 seq_show_option(seq, "subvol", subvol_name); 1150 kfree(subvol_name); 1151 } 1152 return 0; 1153 } 1154 1155 /* 1156 * subvolumes are identified by ino 256 1157 */ 1158 static inline bool is_subvolume_inode(struct inode *inode) 1159 { 1160 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1161 return true; 1162 return false; 1163 } 1164 1165 static struct dentry *mount_subvol(const char *subvol_name, u64 subvol_objectid, 1166 struct vfsmount *mnt) 1167 { 1168 struct dentry *root; 1169 int ret; 1170 1171 if (!subvol_name) { 1172 if (!subvol_objectid) { 1173 ret = get_default_subvol_objectid(btrfs_sb(mnt->mnt_sb), 1174 &subvol_objectid); 1175 if (ret) { 1176 root = ERR_PTR(ret); 1177 goto out; 1178 } 1179 } 1180 subvol_name = btrfs_get_subvol_name_from_objectid( 1181 btrfs_sb(mnt->mnt_sb), subvol_objectid); 1182 if (IS_ERR(subvol_name)) { 1183 root = ERR_CAST(subvol_name); 1184 subvol_name = NULL; 1185 goto out; 1186 } 1187 1188 } 1189 1190 root = mount_subtree(mnt, subvol_name); 1191 /* mount_subtree() drops our reference on the vfsmount. */ 1192 mnt = NULL; 1193 1194 if (!IS_ERR(root)) { 1195 struct super_block *s = root->d_sb; 1196 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1197 struct inode *root_inode = d_inode(root); 1198 u64 root_objectid = btrfs_root_id(BTRFS_I(root_inode)->root); 1199 1200 ret = 0; 1201 if (!is_subvolume_inode(root_inode)) { 1202 btrfs_err(fs_info, "'%s' is not a valid subvolume", 1203 subvol_name); 1204 ret = -EINVAL; 1205 } 1206 if (subvol_objectid && root_objectid != subvol_objectid) { 1207 /* 1208 * This will also catch a race condition where a 1209 * subvolume which was passed by ID is renamed and 1210 * another subvolume is renamed over the old location. 1211 */ 1212 btrfs_err(fs_info, 1213 "subvol '%s' does not match subvolid %llu", 1214 subvol_name, subvol_objectid); 1215 ret = -EINVAL; 1216 } 1217 if (ret) { 1218 dput(root); 1219 root = ERR_PTR(ret); 1220 deactivate_locked_super(s); 1221 } 1222 } 1223 1224 out: 1225 mntput(mnt); 1226 kfree(subvol_name); 1227 return root; 1228 } 1229 1230 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 1231 u32 new_pool_size, u32 old_pool_size) 1232 { 1233 if (new_pool_size == old_pool_size) 1234 return; 1235 1236 fs_info->thread_pool_size = new_pool_size; 1237 1238 btrfs_info(fs_info, "resize thread pool %d -> %d", 1239 old_pool_size, new_pool_size); 1240 1241 btrfs_workqueue_set_max(fs_info->workers, new_pool_size); 1242 btrfs_workqueue_set_max(fs_info->delalloc_workers, new_pool_size); 1243 btrfs_workqueue_set_max(fs_info->caching_workers, new_pool_size); 1244 workqueue_set_max_active(fs_info->endio_workers, new_pool_size); 1245 workqueue_set_max_active(fs_info->endio_meta_workers, new_pool_size); 1246 btrfs_workqueue_set_max(fs_info->endio_write_workers, new_pool_size); 1247 btrfs_workqueue_set_max(fs_info->endio_freespace_worker, new_pool_size); 1248 btrfs_workqueue_set_max(fs_info->delayed_workers, new_pool_size); 1249 } 1250 1251 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1252 unsigned long long old_opts, int flags) 1253 { 1254 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1255 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1256 (flags & SB_RDONLY))) { 1257 /* wait for any defraggers to finish */ 1258 wait_event(fs_info->transaction_wait, 1259 (atomic_read(&fs_info->defrag_running) == 0)); 1260 if (flags & SB_RDONLY) 1261 sync_filesystem(fs_info->sb); 1262 } 1263 } 1264 1265 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1266 unsigned long long old_opts) 1267 { 1268 const bool cache_opt = btrfs_test_opt(fs_info, SPACE_CACHE); 1269 1270 /* 1271 * We need to cleanup all defragable inodes if the autodefragment is 1272 * close or the filesystem is read only. 1273 */ 1274 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1275 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || sb_rdonly(fs_info->sb))) { 1276 btrfs_cleanup_defrag_inodes(fs_info); 1277 } 1278 1279 /* If we toggled discard async */ 1280 if (!btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1281 btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1282 btrfs_discard_resume(fs_info); 1283 else if (btrfs_raw_test_opt(old_opts, DISCARD_ASYNC) && 1284 !btrfs_test_opt(fs_info, DISCARD_ASYNC)) 1285 btrfs_discard_cleanup(fs_info); 1286 1287 /* If we toggled space cache */ 1288 if (cache_opt != btrfs_free_space_cache_v1_active(fs_info)) 1289 btrfs_set_free_space_cache_v1_active(fs_info, cache_opt); 1290 } 1291 1292 static int btrfs_remount_rw(struct btrfs_fs_info *fs_info) 1293 { 1294 int ret; 1295 1296 if (BTRFS_FS_ERROR(fs_info)) { 1297 btrfs_err(fs_info, 1298 "remounting read-write after error is not allowed"); 1299 return -EINVAL; 1300 } 1301 1302 if (fs_info->fs_devices->rw_devices == 0) 1303 return -EACCES; 1304 1305 if (!btrfs_check_rw_degradable(fs_info, NULL)) { 1306 btrfs_warn(fs_info, 1307 "too many missing devices, writable remount is not allowed"); 1308 return -EACCES; 1309 } 1310 1311 if (btrfs_super_log_root(fs_info->super_copy) != 0) { 1312 btrfs_warn(fs_info, 1313 "mount required to replay tree-log, cannot remount read-write"); 1314 return -EINVAL; 1315 } 1316 1317 /* 1318 * NOTE: when remounting with a change that does writes, don't put it 1319 * anywhere above this point, as we are not sure to be safe to write 1320 * until we pass the above checks. 1321 */ 1322 ret = btrfs_start_pre_rw_mount(fs_info); 1323 if (ret) 1324 return ret; 1325 1326 btrfs_clear_sb_rdonly(fs_info->sb); 1327 1328 set_bit(BTRFS_FS_OPEN, &fs_info->flags); 1329 1330 /* 1331 * If we've gone from readonly -> read-write, we need to get our 1332 * sync/async discard lists in the right state. 1333 */ 1334 btrfs_discard_resume(fs_info); 1335 1336 return 0; 1337 } 1338 1339 static int btrfs_remount_ro(struct btrfs_fs_info *fs_info) 1340 { 1341 /* 1342 * This also happens on 'umount -rf' or on shutdown, when the 1343 * filesystem is busy. 1344 */ 1345 cancel_work_sync(&fs_info->async_reclaim_work); 1346 cancel_work_sync(&fs_info->async_data_reclaim_work); 1347 1348 btrfs_discard_cleanup(fs_info); 1349 1350 /* Wait for the uuid_scan task to finish */ 1351 down(&fs_info->uuid_tree_rescan_sem); 1352 /* Avoid complains from lockdep et al. */ 1353 up(&fs_info->uuid_tree_rescan_sem); 1354 1355 btrfs_set_sb_rdonly(fs_info->sb); 1356 1357 /* 1358 * Setting SB_RDONLY will put the cleaner thread to sleep at the next 1359 * loop if it's already active. If it's already asleep, we'll leave 1360 * unused block groups on disk until we're mounted read-write again 1361 * unless we clean them up here. 1362 */ 1363 btrfs_delete_unused_bgs(fs_info); 1364 1365 /* 1366 * The cleaner task could be already running before we set the flag 1367 * BTRFS_FS_STATE_RO (and SB_RDONLY in the superblock). We must make 1368 * sure that after we finish the remount, i.e. after we call 1369 * btrfs_commit_super(), the cleaner can no longer start a transaction 1370 * - either because it was dropping a dead root, running delayed iputs 1371 * or deleting an unused block group (the cleaner picked a block 1372 * group from the list of unused block groups before we were able to 1373 * in the previous call to btrfs_delete_unused_bgs()). 1374 */ 1375 wait_on_bit(&fs_info->flags, BTRFS_FS_CLEANER_RUNNING, TASK_UNINTERRUPTIBLE); 1376 1377 /* 1378 * We've set the superblock to RO mode, so we might have made the 1379 * cleaner task sleep without running all pending delayed iputs. Go 1380 * through all the delayed iputs here, so that if an unmount happens 1381 * without remounting RW we don't end up at finishing close_ctree() 1382 * with a non-empty list of delayed iputs. 1383 */ 1384 btrfs_run_delayed_iputs(fs_info); 1385 1386 btrfs_dev_replace_suspend_for_unmount(fs_info); 1387 btrfs_scrub_cancel(fs_info); 1388 btrfs_pause_balance(fs_info); 1389 1390 /* 1391 * Pause the qgroup rescan worker if it is running. We don't want it to 1392 * be still running after we are in RO mode, as after that, by the time 1393 * we unmount, it might have left a transaction open, so we would leak 1394 * the transaction and/or crash. 1395 */ 1396 btrfs_qgroup_wait_for_completion(fs_info, false); 1397 1398 return btrfs_commit_super(fs_info); 1399 } 1400 1401 static void btrfs_ctx_to_info(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) 1402 { 1403 fs_info->max_inline = ctx->max_inline; 1404 fs_info->commit_interval = ctx->commit_interval; 1405 fs_info->metadata_ratio = ctx->metadata_ratio; 1406 fs_info->thread_pool_size = ctx->thread_pool_size; 1407 fs_info->mount_opt = ctx->mount_opt; 1408 fs_info->compress_type = ctx->compress_type; 1409 fs_info->compress_level = ctx->compress_level; 1410 } 1411 1412 static void btrfs_info_to_ctx(struct btrfs_fs_info *fs_info, struct btrfs_fs_context *ctx) 1413 { 1414 ctx->max_inline = fs_info->max_inline; 1415 ctx->commit_interval = fs_info->commit_interval; 1416 ctx->metadata_ratio = fs_info->metadata_ratio; 1417 ctx->thread_pool_size = fs_info->thread_pool_size; 1418 ctx->mount_opt = fs_info->mount_opt; 1419 ctx->compress_type = fs_info->compress_type; 1420 ctx->compress_level = fs_info->compress_level; 1421 } 1422 1423 #define btrfs_info_if_set(fs_info, old_ctx, opt, fmt, args...) \ 1424 do { \ 1425 if ((!old_ctx || !btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ 1426 btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ 1427 btrfs_info(fs_info, fmt, ##args); \ 1428 } while (0) 1429 1430 #define btrfs_info_if_unset(fs_info, old_ctx, opt, fmt, args...) \ 1431 do { \ 1432 if ((old_ctx && btrfs_raw_test_opt(old_ctx->mount_opt, opt)) && \ 1433 !btrfs_raw_test_opt(fs_info->mount_opt, opt)) \ 1434 btrfs_info(fs_info, fmt, ##args); \ 1435 } while (0) 1436 1437 static void btrfs_emit_options(struct btrfs_fs_info *info, 1438 struct btrfs_fs_context *old) 1439 { 1440 btrfs_info_if_set(info, old, NODATASUM, "setting nodatasum"); 1441 btrfs_info_if_set(info, old, DEGRADED, "allowing degraded mounts"); 1442 btrfs_info_if_set(info, old, NODATACOW, "setting nodatacow"); 1443 btrfs_info_if_set(info, old, SSD, "enabling ssd optimizations"); 1444 btrfs_info_if_set(info, old, SSD_SPREAD, "using spread ssd allocation scheme"); 1445 btrfs_info_if_set(info, old, NOBARRIER, "turning off barriers"); 1446 btrfs_info_if_set(info, old, NOTREELOG, "disabling tree log"); 1447 btrfs_info_if_set(info, old, NOLOGREPLAY, "disabling log replay at mount time"); 1448 btrfs_info_if_set(info, old, FLUSHONCOMMIT, "turning on flush-on-commit"); 1449 btrfs_info_if_set(info, old, DISCARD_SYNC, "turning on sync discard"); 1450 btrfs_info_if_set(info, old, DISCARD_ASYNC, "turning on async discard"); 1451 btrfs_info_if_set(info, old, FREE_SPACE_TREE, "enabling free space tree"); 1452 btrfs_info_if_set(info, old, SPACE_CACHE, "enabling disk space caching"); 1453 btrfs_info_if_set(info, old, CLEAR_CACHE, "force clearing of disk cache"); 1454 btrfs_info_if_set(info, old, AUTO_DEFRAG, "enabling auto defrag"); 1455 btrfs_info_if_set(info, old, FRAGMENT_DATA, "fragmenting data"); 1456 btrfs_info_if_set(info, old, FRAGMENT_METADATA, "fragmenting metadata"); 1457 btrfs_info_if_set(info, old, REF_VERIFY, "doing ref verification"); 1458 btrfs_info_if_set(info, old, USEBACKUPROOT, "trying to use backup root at mount time"); 1459 btrfs_info_if_set(info, old, IGNOREBADROOTS, "ignoring bad roots"); 1460 btrfs_info_if_set(info, old, IGNOREDATACSUMS, "ignoring data csums"); 1461 btrfs_info_if_set(info, old, IGNOREMETACSUMS, "ignoring meta csums"); 1462 btrfs_info_if_set(info, old, IGNORESUPERFLAGS, "ignoring unknown super block flags"); 1463 1464 btrfs_info_if_unset(info, old, NODATASUM, "setting datasum"); 1465 btrfs_info_if_unset(info, old, NODATACOW, "setting datacow"); 1466 btrfs_info_if_unset(info, old, SSD, "not using ssd optimizations"); 1467 btrfs_info_if_unset(info, old, SSD_SPREAD, "not using spread ssd allocation scheme"); 1468 btrfs_info_if_unset(info, old, NOBARRIER, "turning on barriers"); 1469 btrfs_info_if_unset(info, old, NOTREELOG, "enabling tree log"); 1470 btrfs_info_if_unset(info, old, SPACE_CACHE, "disabling disk space caching"); 1471 btrfs_info_if_unset(info, old, FREE_SPACE_TREE, "disabling free space tree"); 1472 btrfs_info_if_unset(info, old, AUTO_DEFRAG, "disabling auto defrag"); 1473 btrfs_info_if_unset(info, old, COMPRESS, "use no compression"); 1474 1475 /* Did the compression settings change? */ 1476 if (btrfs_test_opt(info, COMPRESS) && 1477 (!old || 1478 old->compress_type != info->compress_type || 1479 old->compress_level != info->compress_level || 1480 (!btrfs_raw_test_opt(old->mount_opt, FORCE_COMPRESS) && 1481 btrfs_raw_test_opt(info->mount_opt, FORCE_COMPRESS)))) { 1482 const char *compress_type = btrfs_compress_type2str(info->compress_type); 1483 1484 btrfs_info(info, "%s %s compression, level %d", 1485 btrfs_test_opt(info, FORCE_COMPRESS) ? "force" : "use", 1486 compress_type, info->compress_level); 1487 } 1488 1489 if (info->max_inline != BTRFS_DEFAULT_MAX_INLINE) 1490 btrfs_info(info, "max_inline set to %llu", info->max_inline); 1491 } 1492 1493 static int btrfs_reconfigure(struct fs_context *fc) 1494 { 1495 struct super_block *sb = fc->root->d_sb; 1496 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1497 struct btrfs_fs_context *ctx = fc->fs_private; 1498 struct btrfs_fs_context old_ctx; 1499 int ret = 0; 1500 bool mount_reconfigure = (fc->s_fs_info != NULL); 1501 1502 btrfs_info_to_ctx(fs_info, &old_ctx); 1503 1504 /* 1505 * This is our "bind mount" trick, we don't want to allow the user to do 1506 * anything other than mount a different ro/rw and a different subvol, 1507 * all of the mount options should be maintained. 1508 */ 1509 if (mount_reconfigure) 1510 ctx->mount_opt = old_ctx.mount_opt; 1511 1512 sync_filesystem(sb); 1513 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1514 1515 if (!btrfs_check_options(fs_info, &ctx->mount_opt, fc->sb_flags)) 1516 return -EINVAL; 1517 1518 ret = btrfs_check_features(fs_info, !(fc->sb_flags & SB_RDONLY)); 1519 if (ret < 0) 1520 return ret; 1521 1522 btrfs_ctx_to_info(fs_info, ctx); 1523 btrfs_remount_begin(fs_info, old_ctx.mount_opt, fc->sb_flags); 1524 btrfs_resize_thread_pool(fs_info, fs_info->thread_pool_size, 1525 old_ctx.thread_pool_size); 1526 1527 if ((bool)btrfs_test_opt(fs_info, FREE_SPACE_TREE) != 1528 (bool)btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE) && 1529 (!sb_rdonly(sb) || (fc->sb_flags & SB_RDONLY))) { 1530 btrfs_warn(fs_info, 1531 "remount supports changing free space tree only from RO to RW"); 1532 /* Make sure free space cache options match the state on disk. */ 1533 if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE)) { 1534 btrfs_set_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1535 btrfs_clear_opt(fs_info->mount_opt, SPACE_CACHE); 1536 } 1537 if (btrfs_free_space_cache_v1_active(fs_info)) { 1538 btrfs_clear_opt(fs_info->mount_opt, FREE_SPACE_TREE); 1539 btrfs_set_opt(fs_info->mount_opt, SPACE_CACHE); 1540 } 1541 } 1542 1543 ret = 0; 1544 if (!sb_rdonly(sb) && (fc->sb_flags & SB_RDONLY)) 1545 ret = btrfs_remount_ro(fs_info); 1546 else if (sb_rdonly(sb) && !(fc->sb_flags & SB_RDONLY)) 1547 ret = btrfs_remount_rw(fs_info); 1548 if (ret) 1549 goto restore; 1550 1551 /* 1552 * If we set the mask during the parameter parsing VFS would reject the 1553 * remount. Here we can set the mask and the value will be updated 1554 * appropriately. 1555 */ 1556 if ((fc->sb_flags & SB_POSIXACL) != (sb->s_flags & SB_POSIXACL)) 1557 fc->sb_flags_mask |= SB_POSIXACL; 1558 1559 btrfs_emit_options(fs_info, &old_ctx); 1560 wake_up_process(fs_info->transaction_kthread); 1561 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); 1562 btrfs_clear_oneshot_options(fs_info); 1563 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1564 1565 return 0; 1566 restore: 1567 btrfs_ctx_to_info(fs_info, &old_ctx); 1568 btrfs_remount_cleanup(fs_info, old_ctx.mount_opt); 1569 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1570 return ret; 1571 } 1572 1573 /* Used to sort the devices by max_avail(descending sort) */ 1574 static int btrfs_cmp_device_free_bytes(const void *a, const void *b) 1575 { 1576 const struct btrfs_device_info *dev_info1 = a; 1577 const struct btrfs_device_info *dev_info2 = b; 1578 1579 if (dev_info1->max_avail > dev_info2->max_avail) 1580 return -1; 1581 else if (dev_info1->max_avail < dev_info2->max_avail) 1582 return 1; 1583 return 0; 1584 } 1585 1586 /* 1587 * sort the devices by max_avail, in which max free extent size of each device 1588 * is stored.(Descending Sort) 1589 */ 1590 static inline void btrfs_descending_sort_devices( 1591 struct btrfs_device_info *devices, 1592 size_t nr_devices) 1593 { 1594 sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1595 btrfs_cmp_device_free_bytes, NULL); 1596 } 1597 1598 /* 1599 * The helper to calc the free space on the devices that can be used to store 1600 * file data. 1601 */ 1602 static inline int btrfs_calc_avail_data_space(struct btrfs_fs_info *fs_info, 1603 u64 *free_bytes) 1604 { 1605 struct btrfs_device_info *devices_info; 1606 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 1607 struct btrfs_device *device; 1608 u64 type; 1609 u64 avail_space; 1610 u64 min_stripe_size; 1611 int num_stripes = 1; 1612 int i = 0, nr_devices; 1613 const struct btrfs_raid_attr *rattr; 1614 1615 /* 1616 * We aren't under the device list lock, so this is racy-ish, but good 1617 * enough for our purposes. 1618 */ 1619 nr_devices = fs_info->fs_devices->open_devices; 1620 if (!nr_devices) { 1621 smp_mb(); 1622 nr_devices = fs_info->fs_devices->open_devices; 1623 ASSERT(nr_devices); 1624 if (!nr_devices) { 1625 *free_bytes = 0; 1626 return 0; 1627 } 1628 } 1629 1630 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), 1631 GFP_KERNEL); 1632 if (!devices_info) 1633 return -ENOMEM; 1634 1635 /* calc min stripe number for data space allocation */ 1636 type = btrfs_data_alloc_profile(fs_info); 1637 rattr = &btrfs_raid_array[btrfs_bg_flags_to_raid_index(type)]; 1638 1639 if (type & BTRFS_BLOCK_GROUP_RAID0) 1640 num_stripes = nr_devices; 1641 else if (type & BTRFS_BLOCK_GROUP_RAID1_MASK) 1642 num_stripes = rattr->ncopies; 1643 else if (type & BTRFS_BLOCK_GROUP_RAID10) 1644 num_stripes = 4; 1645 1646 /* Adjust for more than 1 stripe per device */ 1647 min_stripe_size = rattr->dev_stripes * BTRFS_STRIPE_LEN; 1648 1649 rcu_read_lock(); 1650 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) { 1651 if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, 1652 &device->dev_state) || 1653 !device->bdev || 1654 test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) 1655 continue; 1656 1657 if (i >= nr_devices) 1658 break; 1659 1660 avail_space = device->total_bytes - device->bytes_used; 1661 1662 /* align with stripe_len */ 1663 avail_space = rounddown(avail_space, BTRFS_STRIPE_LEN); 1664 1665 /* 1666 * Ensure we have at least min_stripe_size on top of the 1667 * reserved space on the device. 1668 */ 1669 if (avail_space <= BTRFS_DEVICE_RANGE_RESERVED + min_stripe_size) 1670 continue; 1671 1672 avail_space -= BTRFS_DEVICE_RANGE_RESERVED; 1673 1674 devices_info[i].dev = device; 1675 devices_info[i].max_avail = avail_space; 1676 1677 i++; 1678 } 1679 rcu_read_unlock(); 1680 1681 nr_devices = i; 1682 1683 btrfs_descending_sort_devices(devices_info, nr_devices); 1684 1685 i = nr_devices - 1; 1686 avail_space = 0; 1687 while (nr_devices >= rattr->devs_min) { 1688 num_stripes = min(num_stripes, nr_devices); 1689 1690 if (devices_info[i].max_avail >= min_stripe_size) { 1691 int j; 1692 u64 alloc_size; 1693 1694 avail_space += devices_info[i].max_avail * num_stripes; 1695 alloc_size = devices_info[i].max_avail; 1696 for (j = i + 1 - num_stripes; j <= i; j++) 1697 devices_info[j].max_avail -= alloc_size; 1698 } 1699 i--; 1700 nr_devices--; 1701 } 1702 1703 kfree(devices_info); 1704 *free_bytes = avail_space; 1705 return 0; 1706 } 1707 1708 /* 1709 * Calculate numbers for 'df', pessimistic in case of mixed raid profiles. 1710 * 1711 * If there's a redundant raid level at DATA block groups, use the respective 1712 * multiplier to scale the sizes. 1713 * 1714 * Unused device space usage is based on simulating the chunk allocator 1715 * algorithm that respects the device sizes and order of allocations. This is 1716 * a close approximation of the actual use but there are other factors that may 1717 * change the result (like a new metadata chunk). 1718 * 1719 * If metadata is exhausted, f_bavail will be 0. 1720 */ 1721 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 1722 { 1723 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1724 struct btrfs_super_block *disk_super = fs_info->super_copy; 1725 struct btrfs_space_info *found; 1726 u64 total_used = 0; 1727 u64 total_free_data = 0; 1728 u64 total_free_meta = 0; 1729 u32 bits = fs_info->sectorsize_bits; 1730 __be32 *fsid = (__be32 *)fs_info->fs_devices->fsid; 1731 unsigned factor = 1; 1732 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv; 1733 int ret; 1734 u64 thresh = 0; 1735 int mixed = 0; 1736 1737 list_for_each_entry(found, &fs_info->space_info, list) { 1738 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 1739 int i; 1740 1741 total_free_data += found->disk_total - found->disk_used; 1742 total_free_data -= 1743 btrfs_account_ro_block_groups_free_space(found); 1744 1745 for (i = 0; i < BTRFS_NR_RAID_TYPES; i++) { 1746 if (!list_empty(&found->block_groups[i])) 1747 factor = btrfs_bg_type_to_factor( 1748 btrfs_raid_array[i].bg_flag); 1749 } 1750 } 1751 1752 /* 1753 * Metadata in mixed block group profiles are accounted in data 1754 */ 1755 if (!mixed && found->flags & BTRFS_BLOCK_GROUP_METADATA) { 1756 if (found->flags & BTRFS_BLOCK_GROUP_DATA) 1757 mixed = 1; 1758 else 1759 total_free_meta += found->disk_total - 1760 found->disk_used; 1761 } 1762 1763 total_used += found->disk_used; 1764 } 1765 1766 buf->f_blocks = div_u64(btrfs_super_total_bytes(disk_super), factor); 1767 buf->f_blocks >>= bits; 1768 buf->f_bfree = buf->f_blocks - (div_u64(total_used, factor) >> bits); 1769 1770 /* Account global block reserve as used, it's in logical size already */ 1771 spin_lock(&block_rsv->lock); 1772 /* Mixed block groups accounting is not byte-accurate, avoid overflow */ 1773 if (buf->f_bfree >= block_rsv->size >> bits) 1774 buf->f_bfree -= block_rsv->size >> bits; 1775 else 1776 buf->f_bfree = 0; 1777 spin_unlock(&block_rsv->lock); 1778 1779 buf->f_bavail = div_u64(total_free_data, factor); 1780 ret = btrfs_calc_avail_data_space(fs_info, &total_free_data); 1781 if (ret) 1782 return ret; 1783 buf->f_bavail += div_u64(total_free_data, factor); 1784 buf->f_bavail = buf->f_bavail >> bits; 1785 1786 /* 1787 * We calculate the remaining metadata space minus global reserve. If 1788 * this is (supposedly) smaller than zero, there's no space. But this 1789 * does not hold in practice, the exhausted state happens where's still 1790 * some positive delta. So we apply some guesswork and compare the 1791 * delta to a 4M threshold. (Practically observed delta was ~2M.) 1792 * 1793 * We probably cannot calculate the exact threshold value because this 1794 * depends on the internal reservations requested by various 1795 * operations, so some operations that consume a few metadata will 1796 * succeed even if the Avail is zero. But this is better than the other 1797 * way around. 1798 */ 1799 thresh = SZ_4M; 1800 1801 /* 1802 * We only want to claim there's no available space if we can no longer 1803 * allocate chunks for our metadata profile and our global reserve will 1804 * not fit in the free metadata space. If we aren't ->full then we 1805 * still can allocate chunks and thus are fine using the currently 1806 * calculated f_bavail. 1807 */ 1808 if (!mixed && block_rsv->space_info->full && 1809 (total_free_meta < thresh || total_free_meta - thresh < block_rsv->size)) 1810 buf->f_bavail = 0; 1811 1812 buf->f_type = BTRFS_SUPER_MAGIC; 1813 buf->f_bsize = fs_info->sectorsize; 1814 buf->f_namelen = BTRFS_NAME_LEN; 1815 1816 /* We treat it as constant endianness (it doesn't matter _which_) 1817 because we want the fsid to come out the same whether mounted 1818 on a big-endian or little-endian host */ 1819 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 1820 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 1821 /* Mask in the root object ID too, to disambiguate subvols */ 1822 buf->f_fsid.val[0] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root) >> 32; 1823 buf->f_fsid.val[1] ^= btrfs_root_id(BTRFS_I(d_inode(dentry))->root); 1824 1825 return 0; 1826 } 1827 1828 static int btrfs_fc_test_super(struct super_block *sb, struct fs_context *fc) 1829 { 1830 struct btrfs_fs_info *p = fc->s_fs_info; 1831 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1832 1833 return fs_info->fs_devices == p->fs_devices; 1834 } 1835 1836 static int btrfs_get_tree_super(struct fs_context *fc) 1837 { 1838 struct btrfs_fs_info *fs_info = fc->s_fs_info; 1839 struct btrfs_fs_context *ctx = fc->fs_private; 1840 struct btrfs_fs_devices *fs_devices = NULL; 1841 struct btrfs_device *device; 1842 struct super_block *sb; 1843 blk_mode_t mode = sb_open_mode(fc->sb_flags); 1844 int ret; 1845 1846 btrfs_ctx_to_info(fs_info, ctx); 1847 mutex_lock(&uuid_mutex); 1848 1849 /* 1850 * With 'true' passed to btrfs_scan_one_device() (mount time) we expect 1851 * either a valid device or an error. 1852 */ 1853 device = btrfs_scan_one_device(fc->source, true); 1854 ASSERT(device != NULL); 1855 if (IS_ERR(device)) { 1856 mutex_unlock(&uuid_mutex); 1857 return PTR_ERR(device); 1858 } 1859 fs_devices = device->fs_devices; 1860 /* 1861 * We cannot hold uuid_mutex calling sget_fc(), it will lead to a 1862 * locking order reversal with s_umount. 1863 * 1864 * So here we increase the holding number of fs_devices, this will ensure 1865 * the fs_devices itself won't be freed. 1866 */ 1867 btrfs_fs_devices_inc_holding(fs_devices); 1868 fs_info->fs_devices = fs_devices; 1869 mutex_unlock(&uuid_mutex); 1870 1871 1872 sb = sget_fc(fc, btrfs_fc_test_super, set_anon_super_fc); 1873 if (IS_ERR(sb)) { 1874 mutex_lock(&uuid_mutex); 1875 btrfs_fs_devices_dec_holding(fs_devices); 1876 /* 1877 * Since the fs_devices is not opened, it can be freed at any 1878 * time after unlocking uuid_mutex. We need to avoid double 1879 * free through put_fs_context()->btrfs_free_fs_info(). 1880 * So here we reset fs_info->fs_devices to NULL, and let the 1881 * regular fs_devices reclaim path to handle it. 1882 * 1883 * This applies to all later branches where no fs_devices is 1884 * opened. 1885 */ 1886 fs_info->fs_devices = NULL; 1887 mutex_unlock(&uuid_mutex); 1888 return PTR_ERR(sb); 1889 } 1890 1891 set_device_specific_options(fs_info); 1892 1893 if (sb->s_root) { 1894 /* 1895 * Not the first mount of the fs thus got an existing super block. 1896 * Will reuse the returned super block, fs_info and fs_devices. 1897 * 1898 * fc->s_fs_info is not touched and will be later freed by 1899 * put_fs_context() through btrfs_free_fs_context(). 1900 */ 1901 ASSERT(fc->s_fs_info == fs_info); 1902 1903 mutex_lock(&uuid_mutex); 1904 btrfs_fs_devices_dec_holding(fs_devices); 1905 fs_info->fs_devices = NULL; 1906 mutex_unlock(&uuid_mutex); 1907 /* 1908 * At this stage we may have RO flag mismatch between 1909 * fc->sb_flags and sb->s_flags. Caller should detect such 1910 * mismatch and reconfigure with sb->s_umount rwsem held if 1911 * needed. 1912 */ 1913 } else { 1914 struct block_device *bdev; 1915 1916 /* 1917 * The first mount of the fs thus a new superblock, fc->s_fs_info 1918 * must be NULL, and the ownership of our fs_info and fs_devices is 1919 * transferred to the super block. 1920 */ 1921 ASSERT(fc->s_fs_info == NULL); 1922 1923 mutex_lock(&uuid_mutex); 1924 btrfs_fs_devices_dec_holding(fs_devices); 1925 ret = btrfs_open_devices(fs_devices, mode, sb); 1926 if (ret < 0) 1927 fs_info->fs_devices = NULL; 1928 mutex_unlock(&uuid_mutex); 1929 if (ret < 0) { 1930 deactivate_locked_super(sb); 1931 return ret; 1932 } 1933 if (!(fc->sb_flags & SB_RDONLY) && fs_devices->rw_devices == 0) { 1934 deactivate_locked_super(sb); 1935 return -EACCES; 1936 } 1937 bdev = fs_devices->latest_dev->bdev; 1938 snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev); 1939 shrinker_debugfs_rename(sb->s_shrink, "sb-btrfs:%s", sb->s_id); 1940 ret = btrfs_fill_super(sb, fs_devices); 1941 if (ret) { 1942 deactivate_locked_super(sb); 1943 return ret; 1944 } 1945 } 1946 1947 btrfs_clear_oneshot_options(fs_info); 1948 1949 fc->root = dget(sb->s_root); 1950 return 0; 1951 } 1952 1953 /* 1954 * Ever since commit 0723a0473fb4 ("btrfs: allow mounting btrfs subvolumes 1955 * with different ro/rw options") the following works: 1956 * 1957 * (i) mount /dev/sda3 -o subvol=foo,ro /mnt/foo 1958 * (ii) mount /dev/sda3 -o subvol=bar,rw /mnt/bar 1959 * 1960 * which looks nice and innocent but is actually pretty intricate and deserves 1961 * a long comment. 1962 * 1963 * On another filesystem a subvolume mount is close to something like: 1964 * 1965 * (iii) # create rw superblock + initial mount 1966 * mount -t xfs /dev/sdb /opt/ 1967 * 1968 * # create ro bind mount 1969 * mount --bind -o ro /opt/foo /mnt/foo 1970 * 1971 * # unmount initial mount 1972 * umount /opt 1973 * 1974 * Of course, there's some special subvolume sauce and there's the fact that the 1975 * sb->s_root dentry is really swapped after mount_subtree(). But conceptually 1976 * it's very close and will help us understand the issue. 1977 * 1978 * The old mount API didn't cleanly distinguish between a mount being made ro 1979 * and a superblock being made ro. The only way to change the ro state of 1980 * either object was by passing ms_rdonly. If a new mount was created via 1981 * mount(2) such as: 1982 * 1983 * mount("/dev/sdb", "/mnt", "xfs", ms_rdonly, null); 1984 * 1985 * the MS_RDONLY flag being specified had two effects: 1986 * 1987 * (1) MNT_READONLY was raised -> the resulting mount got 1988 * @mnt->mnt_flags |= MNT_READONLY raised. 1989 * 1990 * (2) MS_RDONLY was passed to the filesystem's mount method and the filesystems 1991 * made the superblock ro. Note, how SB_RDONLY has the same value as 1992 * ms_rdonly and is raised whenever MS_RDONLY is passed through mount(2). 1993 * 1994 * Creating a subtree mount via (iii) ends up leaving a rw superblock with a 1995 * subtree mounted ro. 1996 * 1997 * But consider the effect on the old mount API on btrfs subvolume mounting 1998 * which combines the distinct step in (iii) into a single step. 1999 * 2000 * By issuing (i) both the mount and the superblock are turned ro. Now when (ii) 2001 * is issued the superblock is ro and thus even if the mount created for (ii) is 2002 * rw it wouldn't help. Hence, btrfs needed to transition the superblock from ro 2003 * to rw for (ii) which it did using an internal remount call. 2004 * 2005 * IOW, subvolume mounting was inherently complicated due to the ambiguity of 2006 * MS_RDONLY in mount(2). Note, this ambiguity has mount(8) always translate 2007 * "ro" to MS_RDONLY. IOW, in both (i) and (ii) "ro" becomes MS_RDONLY when 2008 * passed by mount(8) to mount(2). 2009 * 2010 * Enter the new mount API. The new mount API disambiguates making a mount ro 2011 * and making a superblock ro. 2012 * 2013 * (3) To turn a mount ro the MOUNT_ATTR_ONLY flag can be used with either 2014 * fsmount() or mount_setattr() this is a pure VFS level change for a 2015 * specific mount or mount tree that is never seen by the filesystem itself. 2016 * 2017 * (4) To turn a superblock ro the "ro" flag must be used with 2018 * fsconfig(FSCONFIG_SET_FLAG, "ro"). This option is seen by the filesystem 2019 * in fc->sb_flags. 2020 * 2021 * But, currently the util-linux mount command already utilizes the new mount 2022 * API and is still setting fsconfig(FSCONFIG_SET_FLAG, "ro") no matter if it's 2023 * btrfs or not, setting the whole super block RO. To make per-subvolume mounting 2024 * work with different options work we need to keep backward compatibility. 2025 */ 2026 static int btrfs_reconfigure_for_mount(struct fs_context *fc) 2027 { 2028 int ret = 0; 2029 2030 if (!(fc->sb_flags & SB_RDONLY) && (fc->root->d_sb->s_flags & SB_RDONLY)) 2031 ret = btrfs_reconfigure(fc); 2032 2033 return ret; 2034 } 2035 2036 static int btrfs_get_tree_subvol(struct fs_context *fc) 2037 { 2038 struct btrfs_fs_info *fs_info = NULL; 2039 struct btrfs_fs_context *ctx = fc->fs_private; 2040 struct fs_context *dup_fc; 2041 struct dentry *dentry; 2042 struct vfsmount *mnt; 2043 int ret = 0; 2044 2045 /* 2046 * Setup a dummy root and fs_info for test/set super. This is because 2047 * we don't actually fill this stuff out until open_ctree, but we need 2048 * then open_ctree will properly initialize the file system specific 2049 * settings later. btrfs_init_fs_info initializes the static elements 2050 * of the fs_info (locks and such) to make cleanup easier if we find a 2051 * superblock with our given fs_devices later on at sget() time. 2052 */ 2053 fs_info = kvzalloc(sizeof(struct btrfs_fs_info), GFP_KERNEL); 2054 if (!fs_info) 2055 return -ENOMEM; 2056 2057 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 2058 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_KERNEL); 2059 if (!fs_info->super_copy || !fs_info->super_for_commit) { 2060 btrfs_free_fs_info(fs_info); 2061 return -ENOMEM; 2062 } 2063 btrfs_init_fs_info(fs_info); 2064 2065 dup_fc = vfs_dup_fs_context(fc); 2066 if (IS_ERR(dup_fc)) { 2067 btrfs_free_fs_info(fs_info); 2068 return PTR_ERR(dup_fc); 2069 } 2070 2071 /* 2072 * When we do the sget_fc this gets transferred to the sb, so we only 2073 * need to set it on the dup_fc as this is what creates the super block. 2074 */ 2075 dup_fc->s_fs_info = fs_info; 2076 2077 ret = btrfs_get_tree_super(dup_fc); 2078 if (ret) 2079 goto error; 2080 2081 ret = btrfs_reconfigure_for_mount(dup_fc); 2082 up_write(&dup_fc->root->d_sb->s_umount); 2083 if (ret) 2084 goto error; 2085 mnt = vfs_create_mount(dup_fc); 2086 put_fs_context(dup_fc); 2087 if (IS_ERR(mnt)) 2088 return PTR_ERR(mnt); 2089 2090 /* 2091 * This free's ->subvol_name, because if it isn't set we have to 2092 * allocate a buffer to hold the subvol_name, so we just drop our 2093 * reference to it here. 2094 */ 2095 dentry = mount_subvol(ctx->subvol_name, ctx->subvol_objectid, mnt); 2096 ctx->subvol_name = NULL; 2097 if (IS_ERR(dentry)) 2098 return PTR_ERR(dentry); 2099 2100 fc->root = dentry; 2101 return 0; 2102 error: 2103 put_fs_context(dup_fc); 2104 return ret; 2105 } 2106 2107 static int btrfs_get_tree(struct fs_context *fc) 2108 { 2109 ASSERT(fc->s_fs_info == NULL); 2110 2111 return btrfs_get_tree_subvol(fc); 2112 } 2113 2114 static void btrfs_kill_super(struct super_block *sb) 2115 { 2116 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2117 kill_anon_super(sb); 2118 btrfs_free_fs_info(fs_info); 2119 } 2120 2121 static void btrfs_free_fs_context(struct fs_context *fc) 2122 { 2123 struct btrfs_fs_context *ctx = fc->fs_private; 2124 struct btrfs_fs_info *fs_info = fc->s_fs_info; 2125 2126 if (fs_info) 2127 btrfs_free_fs_info(fs_info); 2128 2129 if (ctx && refcount_dec_and_test(&ctx->refs)) { 2130 kfree(ctx->subvol_name); 2131 kfree(ctx); 2132 } 2133 } 2134 2135 static int btrfs_dup_fs_context(struct fs_context *fc, struct fs_context *src_fc) 2136 { 2137 struct btrfs_fs_context *ctx = src_fc->fs_private; 2138 2139 /* 2140 * Give a ref to our ctx to this dup, as we want to keep it around for 2141 * our original fc so we can have the subvolume name or objectid. 2142 * 2143 * We unset ->source in the original fc because the dup needs it for 2144 * mounting, and then once we free the dup it'll free ->source, so we 2145 * need to make sure we're only pointing to it in one fc. 2146 */ 2147 refcount_inc(&ctx->refs); 2148 fc->fs_private = ctx; 2149 fc->source = src_fc->source; 2150 src_fc->source = NULL; 2151 return 0; 2152 } 2153 2154 static const struct fs_context_operations btrfs_fs_context_ops = { 2155 .parse_param = btrfs_parse_param, 2156 .reconfigure = btrfs_reconfigure, 2157 .get_tree = btrfs_get_tree, 2158 .dup = btrfs_dup_fs_context, 2159 .free = btrfs_free_fs_context, 2160 }; 2161 2162 static int btrfs_init_fs_context(struct fs_context *fc) 2163 { 2164 struct btrfs_fs_context *ctx; 2165 2166 ctx = kzalloc(sizeof(struct btrfs_fs_context), GFP_KERNEL); 2167 if (!ctx) 2168 return -ENOMEM; 2169 2170 refcount_set(&ctx->refs, 1); 2171 fc->fs_private = ctx; 2172 fc->ops = &btrfs_fs_context_ops; 2173 2174 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 2175 btrfs_info_to_ctx(btrfs_sb(fc->root->d_sb), ctx); 2176 } else { 2177 ctx->thread_pool_size = 2178 min_t(unsigned long, num_online_cpus() + 2, 8); 2179 ctx->max_inline = BTRFS_DEFAULT_MAX_INLINE; 2180 ctx->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 2181 } 2182 2183 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 2184 fc->sb_flags |= SB_POSIXACL; 2185 #endif 2186 fc->sb_flags |= SB_I_VERSION; 2187 2188 return 0; 2189 } 2190 2191 static struct file_system_type btrfs_fs_type = { 2192 .owner = THIS_MODULE, 2193 .name = "btrfs", 2194 .init_fs_context = btrfs_init_fs_context, 2195 .parameters = btrfs_fs_parameters, 2196 .kill_sb = btrfs_kill_super, 2197 .fs_flags = FS_REQUIRES_DEV | FS_BINARY_MOUNTDATA | 2198 FS_ALLOW_IDMAP | FS_MGTIME, 2199 }; 2200 2201 MODULE_ALIAS_FS("btrfs"); 2202 2203 static int btrfs_control_open(struct inode *inode, struct file *file) 2204 { 2205 /* 2206 * The control file's private_data is used to hold the 2207 * transaction when it is started and is used to keep 2208 * track of whether a transaction is already in progress. 2209 */ 2210 file->private_data = NULL; 2211 return 0; 2212 } 2213 2214 /* 2215 * Used by /dev/btrfs-control for devices ioctls. 2216 */ 2217 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 2218 unsigned long arg) 2219 { 2220 struct btrfs_ioctl_vol_args *vol; 2221 struct btrfs_device *device = NULL; 2222 dev_t devt = 0; 2223 int ret = -ENOTTY; 2224 2225 if (!capable(CAP_SYS_ADMIN)) 2226 return -EPERM; 2227 2228 vol = memdup_user((void __user *)arg, sizeof(*vol)); 2229 if (IS_ERR(vol)) 2230 return PTR_ERR(vol); 2231 ret = btrfs_check_ioctl_vol_args_path(vol); 2232 if (ret < 0) 2233 goto out; 2234 2235 switch (cmd) { 2236 case BTRFS_IOC_SCAN_DEV: 2237 mutex_lock(&uuid_mutex); 2238 /* 2239 * Scanning outside of mount can return NULL which would turn 2240 * into 0 error code. 2241 */ 2242 device = btrfs_scan_one_device(vol->name, false); 2243 ret = PTR_ERR_OR_ZERO(device); 2244 mutex_unlock(&uuid_mutex); 2245 break; 2246 case BTRFS_IOC_FORGET_DEV: 2247 if (vol->name[0] != 0) { 2248 ret = lookup_bdev(vol->name, &devt); 2249 if (ret) 2250 break; 2251 } 2252 ret = btrfs_forget_devices(devt); 2253 break; 2254 case BTRFS_IOC_DEVICES_READY: 2255 mutex_lock(&uuid_mutex); 2256 /* 2257 * Scanning outside of mount can return NULL which would turn 2258 * into 0 error code. 2259 */ 2260 device = btrfs_scan_one_device(vol->name, false); 2261 if (IS_ERR_OR_NULL(device)) { 2262 mutex_unlock(&uuid_mutex); 2263 if (IS_ERR(device)) 2264 ret = PTR_ERR(device); 2265 else 2266 ret = 0; 2267 break; 2268 } 2269 ret = !(device->fs_devices->num_devices == 2270 device->fs_devices->total_devices); 2271 mutex_unlock(&uuid_mutex); 2272 break; 2273 case BTRFS_IOC_GET_SUPPORTED_FEATURES: 2274 ret = btrfs_ioctl_get_supported_features((void __user*)arg); 2275 break; 2276 } 2277 2278 out: 2279 kfree(vol); 2280 return ret; 2281 } 2282 2283 static int btrfs_freeze(struct super_block *sb) 2284 { 2285 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2286 2287 set_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2288 /* 2289 * We don't need a barrier here, we'll wait for any transaction that 2290 * could be in progress on other threads (and do delayed iputs that 2291 * we want to avoid on a frozen filesystem), or do the commit 2292 * ourselves. 2293 */ 2294 return btrfs_commit_current_transaction(fs_info->tree_root); 2295 } 2296 2297 static int check_dev_super(struct btrfs_device *dev) 2298 { 2299 struct btrfs_fs_info *fs_info = dev->fs_info; 2300 struct btrfs_super_block *sb; 2301 u64 last_trans; 2302 u16 csum_type; 2303 int ret = 0; 2304 2305 /* This should be called with fs still frozen. */ 2306 ASSERT(test_bit(BTRFS_FS_FROZEN, &fs_info->flags)); 2307 2308 /* Missing dev, no need to check. */ 2309 if (!dev->bdev) 2310 return 0; 2311 2312 /* Only need to check the primary super block. */ 2313 sb = btrfs_read_disk_super(dev->bdev, 0, true); 2314 if (IS_ERR(sb)) 2315 return PTR_ERR(sb); 2316 2317 /* Verify the checksum. */ 2318 csum_type = btrfs_super_csum_type(sb); 2319 if (csum_type != btrfs_super_csum_type(fs_info->super_copy)) { 2320 btrfs_err(fs_info, "csum type changed, has %u expect %u", 2321 csum_type, btrfs_super_csum_type(fs_info->super_copy)); 2322 ret = -EUCLEAN; 2323 goto out; 2324 } 2325 2326 if (btrfs_check_super_csum(fs_info, sb)) { 2327 btrfs_err(fs_info, "csum for on-disk super block no longer matches"); 2328 ret = -EUCLEAN; 2329 goto out; 2330 } 2331 2332 /* Btrfs_validate_super() includes fsid check against super->fsid. */ 2333 ret = btrfs_validate_super(fs_info, sb, 0); 2334 if (ret < 0) 2335 goto out; 2336 2337 last_trans = btrfs_get_last_trans_committed(fs_info); 2338 if (btrfs_super_generation(sb) != last_trans) { 2339 btrfs_err(fs_info, "transid mismatch, has %llu expect %llu", 2340 btrfs_super_generation(sb), last_trans); 2341 ret = -EUCLEAN; 2342 goto out; 2343 } 2344 out: 2345 btrfs_release_disk_super(sb); 2346 return ret; 2347 } 2348 2349 static int btrfs_unfreeze(struct super_block *sb) 2350 { 2351 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2352 struct btrfs_device *device; 2353 int ret = 0; 2354 2355 /* 2356 * Make sure the fs is not changed by accident (like hibernation then 2357 * modified by other OS). 2358 * If we found anything wrong, we mark the fs error immediately. 2359 * 2360 * And since the fs is frozen, no one can modify the fs yet, thus 2361 * we don't need to hold device_list_mutex. 2362 */ 2363 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) { 2364 ret = check_dev_super(device); 2365 if (ret < 0) { 2366 btrfs_handle_fs_error(fs_info, ret, 2367 "super block on devid %llu got modified unexpectedly", 2368 device->devid); 2369 break; 2370 } 2371 } 2372 clear_bit(BTRFS_FS_FROZEN, &fs_info->flags); 2373 2374 /* 2375 * We still return 0, to allow VFS layer to unfreeze the fs even the 2376 * above checks failed. Since the fs is either fine or read-only, we're 2377 * safe to continue, without causing further damage. 2378 */ 2379 return 0; 2380 } 2381 2382 static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 2383 { 2384 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 2385 2386 /* 2387 * There should be always a valid pointer in latest_dev, it may be stale 2388 * for a short moment in case it's being deleted but still valid until 2389 * the end of RCU grace period. 2390 */ 2391 rcu_read_lock(); 2392 seq_escape(m, btrfs_dev_name(fs_info->fs_devices->latest_dev), " \t\n\\"); 2393 rcu_read_unlock(); 2394 2395 return 0; 2396 } 2397 2398 static long btrfs_nr_cached_objects(struct super_block *sb, struct shrink_control *sc) 2399 { 2400 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2401 const s64 nr = percpu_counter_sum_positive(&fs_info->evictable_extent_maps); 2402 2403 trace_btrfs_extent_map_shrinker_count(fs_info, nr); 2404 2405 return nr; 2406 } 2407 2408 static long btrfs_free_cached_objects(struct super_block *sb, struct shrink_control *sc) 2409 { 2410 const long nr_to_scan = min_t(unsigned long, LONG_MAX, sc->nr_to_scan); 2411 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 2412 2413 btrfs_free_extent_maps(fs_info, nr_to_scan); 2414 2415 /* The extent map shrinker runs asynchronously, so always return 0. */ 2416 return 0; 2417 } 2418 2419 static const struct super_operations btrfs_super_ops = { 2420 .drop_inode = btrfs_drop_inode, 2421 .evict_inode = btrfs_evict_inode, 2422 .put_super = btrfs_put_super, 2423 .sync_fs = btrfs_sync_fs, 2424 .show_options = btrfs_show_options, 2425 .show_devname = btrfs_show_devname, 2426 .alloc_inode = btrfs_alloc_inode, 2427 .destroy_inode = btrfs_destroy_inode, 2428 .free_inode = btrfs_free_inode, 2429 .statfs = btrfs_statfs, 2430 .freeze_fs = btrfs_freeze, 2431 .unfreeze_fs = btrfs_unfreeze, 2432 .nr_cached_objects = btrfs_nr_cached_objects, 2433 .free_cached_objects = btrfs_free_cached_objects, 2434 }; 2435 2436 static const struct file_operations btrfs_ctl_fops = { 2437 .open = btrfs_control_open, 2438 .unlocked_ioctl = btrfs_control_ioctl, 2439 .compat_ioctl = compat_ptr_ioctl, 2440 .owner = THIS_MODULE, 2441 .llseek = noop_llseek, 2442 }; 2443 2444 static struct miscdevice btrfs_misc = { 2445 .minor = BTRFS_MINOR, 2446 .name = "btrfs-control", 2447 .fops = &btrfs_ctl_fops 2448 }; 2449 2450 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 2451 MODULE_ALIAS("devname:btrfs-control"); 2452 2453 static int __init btrfs_interface_init(void) 2454 { 2455 return misc_register(&btrfs_misc); 2456 } 2457 2458 static __cold void btrfs_interface_exit(void) 2459 { 2460 misc_deregister(&btrfs_misc); 2461 } 2462 2463 static int __init btrfs_print_mod_info(void) 2464 { 2465 static const char options[] = "" 2466 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2467 ", experimental=on" 2468 #endif 2469 #ifdef CONFIG_BTRFS_DEBUG 2470 ", debug=on" 2471 #endif 2472 #ifdef CONFIG_BTRFS_ASSERT 2473 ", assert=on" 2474 #endif 2475 #ifdef CONFIG_BTRFS_FS_REF_VERIFY 2476 ", ref-verify=on" 2477 #endif 2478 #ifdef CONFIG_BLK_DEV_ZONED 2479 ", zoned=yes" 2480 #else 2481 ", zoned=no" 2482 #endif 2483 #ifdef CONFIG_FS_VERITY 2484 ", fsverity=yes" 2485 #else 2486 ", fsverity=no" 2487 #endif 2488 ; 2489 2490 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2491 if (btrfs_get_mod_read_policy() == NULL) 2492 pr_info("Btrfs loaded%s\n", options); 2493 else 2494 pr_info("Btrfs loaded%s, read_policy=%s\n", 2495 options, btrfs_get_mod_read_policy()); 2496 #else 2497 pr_info("Btrfs loaded%s\n", options); 2498 #endif 2499 2500 return 0; 2501 } 2502 2503 static int register_btrfs(void) 2504 { 2505 return register_filesystem(&btrfs_fs_type); 2506 } 2507 2508 static void unregister_btrfs(void) 2509 { 2510 unregister_filesystem(&btrfs_fs_type); 2511 } 2512 2513 /* Helper structure for long init/exit functions. */ 2514 struct init_sequence { 2515 int (*init_func)(void); 2516 /* Can be NULL if the init_func doesn't need cleanup. */ 2517 void (*exit_func)(void); 2518 }; 2519 2520 static const struct init_sequence mod_init_seq[] = { 2521 { 2522 .init_func = btrfs_props_init, 2523 .exit_func = NULL, 2524 }, { 2525 .init_func = btrfs_init_sysfs, 2526 .exit_func = btrfs_exit_sysfs, 2527 }, { 2528 .init_func = btrfs_init_compress, 2529 .exit_func = btrfs_exit_compress, 2530 }, { 2531 .init_func = btrfs_init_cachep, 2532 .exit_func = btrfs_destroy_cachep, 2533 }, { 2534 .init_func = btrfs_init_dio, 2535 .exit_func = btrfs_destroy_dio, 2536 }, { 2537 .init_func = btrfs_transaction_init, 2538 .exit_func = btrfs_transaction_exit, 2539 }, { 2540 .init_func = btrfs_ctree_init, 2541 .exit_func = btrfs_ctree_exit, 2542 }, { 2543 .init_func = btrfs_free_space_init, 2544 .exit_func = btrfs_free_space_exit, 2545 }, { 2546 .init_func = btrfs_extent_state_init_cachep, 2547 .exit_func = btrfs_extent_state_free_cachep, 2548 }, { 2549 .init_func = extent_buffer_init_cachep, 2550 .exit_func = extent_buffer_free_cachep, 2551 }, { 2552 .init_func = btrfs_bioset_init, 2553 .exit_func = btrfs_bioset_exit, 2554 }, { 2555 .init_func = btrfs_extent_map_init, 2556 .exit_func = btrfs_extent_map_exit, 2557 #ifdef CONFIG_BTRFS_EXPERIMENTAL 2558 }, { 2559 .init_func = btrfs_read_policy_init, 2560 .exit_func = NULL, 2561 #endif 2562 }, { 2563 .init_func = ordered_data_init, 2564 .exit_func = ordered_data_exit, 2565 }, { 2566 .init_func = btrfs_delayed_inode_init, 2567 .exit_func = btrfs_delayed_inode_exit, 2568 }, { 2569 .init_func = btrfs_auto_defrag_init, 2570 .exit_func = btrfs_auto_defrag_exit, 2571 }, { 2572 .init_func = btrfs_delayed_ref_init, 2573 .exit_func = btrfs_delayed_ref_exit, 2574 }, { 2575 .init_func = btrfs_prelim_ref_init, 2576 .exit_func = btrfs_prelim_ref_exit, 2577 }, { 2578 .init_func = btrfs_interface_init, 2579 .exit_func = btrfs_interface_exit, 2580 }, { 2581 .init_func = btrfs_print_mod_info, 2582 .exit_func = NULL, 2583 }, { 2584 .init_func = btrfs_run_sanity_tests, 2585 .exit_func = NULL, 2586 }, { 2587 .init_func = register_btrfs, 2588 .exit_func = unregister_btrfs, 2589 } 2590 }; 2591 2592 static bool mod_init_result[ARRAY_SIZE(mod_init_seq)]; 2593 2594 static __always_inline void btrfs_exit_btrfs_fs(void) 2595 { 2596 int i; 2597 2598 for (i = ARRAY_SIZE(mod_init_seq) - 1; i >= 0; i--) { 2599 if (!mod_init_result[i]) 2600 continue; 2601 if (mod_init_seq[i].exit_func) 2602 mod_init_seq[i].exit_func(); 2603 mod_init_result[i] = false; 2604 } 2605 } 2606 2607 static void __exit exit_btrfs_fs(void) 2608 { 2609 btrfs_exit_btrfs_fs(); 2610 btrfs_cleanup_fs_uuids(); 2611 } 2612 2613 static int __init init_btrfs_fs(void) 2614 { 2615 int ret; 2616 int i; 2617 2618 for (i = 0; i < ARRAY_SIZE(mod_init_seq); i++) { 2619 ASSERT(!mod_init_result[i]); 2620 ret = mod_init_seq[i].init_func(); 2621 if (ret < 0) { 2622 btrfs_exit_btrfs_fs(); 2623 return ret; 2624 } 2625 mod_init_result[i] = true; 2626 } 2627 return 0; 2628 } 2629 2630 late_initcall(init_btrfs_fs); 2631 module_exit(exit_btrfs_fs) 2632 2633 MODULE_DESCRIPTION("B-Tree File System (BTRFS)"); 2634 MODULE_LICENSE("GPL"); 2635 MODULE_SOFTDEP("pre: crc32c"); 2636 MODULE_SOFTDEP("pre: xxhash64"); 2637 MODULE_SOFTDEP("pre: sha256"); 2638 MODULE_SOFTDEP("pre: blake2b-256"); 2639