1 /* 2 * Copyright (C) 2007 Oracle. All rights reserved. 3 * 4 * This program is free software; you can redistribute it and/or 5 * modify it under the terms of the GNU General Public 6 * License v2 as published by the Free Software Foundation. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 11 * General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public 14 * License along with this program; if not, write to the 15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 16 * Boston, MA 021110-1307, USA. 17 */ 18 19 #include <linux/blkdev.h> 20 #include <linux/module.h> 21 #include <linux/buffer_head.h> 22 #include <linux/fs.h> 23 #include <linux/pagemap.h> 24 #include <linux/highmem.h> 25 #include <linux/time.h> 26 #include <linux/init.h> 27 #include <linux/seq_file.h> 28 #include <linux/string.h> 29 #include <linux/backing-dev.h> 30 #include <linux/mount.h> 31 #include <linux/mpage.h> 32 #include <linux/swap.h> 33 #include <linux/writeback.h> 34 #include <linux/statfs.h> 35 #include <linux/compat.h> 36 #include <linux/parser.h> 37 #include <linux/ctype.h> 38 #include <linux/namei.h> 39 #include <linux/miscdevice.h> 40 #include <linux/magic.h> 41 #include <linux/slab.h> 42 #include <linux/cleancache.h> 43 #include <linux/ratelimit.h> 44 #include <linux/btrfs.h> 45 #include "delayed-inode.h" 46 #include "ctree.h" 47 #include "disk-io.h" 48 #include "transaction.h" 49 #include "btrfs_inode.h" 50 #include "print-tree.h" 51 #include "xattr.h" 52 #include "volumes.h" 53 #include "export.h" 54 #include "compression.h" 55 #include "rcu-string.h" 56 #include "dev-replace.h" 57 #include "free-space-cache.h" 58 #include "backref.h" 59 #include "tests/btrfs-tests.h" 60 61 #define CREATE_TRACE_POINTS 62 #include <trace/events/btrfs.h> 63 64 static const struct super_operations btrfs_super_ops; 65 static struct file_system_type btrfs_fs_type; 66 67 static const char *btrfs_decode_error(int errno) 68 { 69 char *errstr = "unknown"; 70 71 switch (errno) { 72 case -EIO: 73 errstr = "IO failure"; 74 break; 75 case -ENOMEM: 76 errstr = "Out of memory"; 77 break; 78 case -EROFS: 79 errstr = "Readonly filesystem"; 80 break; 81 case -EEXIST: 82 errstr = "Object already exists"; 83 break; 84 case -ENOSPC: 85 errstr = "No space left"; 86 break; 87 case -ENOENT: 88 errstr = "No such entry"; 89 break; 90 } 91 92 return errstr; 93 } 94 95 static void save_error_info(struct btrfs_fs_info *fs_info) 96 { 97 /* 98 * today we only save the error info into ram. Long term we'll 99 * also send it down to the disk 100 */ 101 set_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state); 102 } 103 104 /* btrfs handle error by forcing the filesystem readonly */ 105 static void btrfs_handle_error(struct btrfs_fs_info *fs_info) 106 { 107 struct super_block *sb = fs_info->sb; 108 109 if (sb->s_flags & MS_RDONLY) 110 return; 111 112 if (test_bit(BTRFS_FS_STATE_ERROR, &fs_info->fs_state)) { 113 sb->s_flags |= MS_RDONLY; 114 btrfs_info(fs_info, "forced readonly"); 115 /* 116 * Note that a running device replace operation is not 117 * canceled here although there is no way to update 118 * the progress. It would add the risk of a deadlock, 119 * therefore the canceling is ommited. The only penalty 120 * is that some I/O remains active until the procedure 121 * completes. The next time when the filesystem is 122 * mounted writeable again, the device replace 123 * operation continues. 124 */ 125 } 126 } 127 128 #ifdef CONFIG_PRINTK 129 /* 130 * __btrfs_std_error decodes expected errors from the caller and 131 * invokes the approciate error response. 132 */ 133 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, 134 unsigned int line, int errno, const char *fmt, ...) 135 { 136 struct super_block *sb = fs_info->sb; 137 const char *errstr; 138 139 /* 140 * Special case: if the error is EROFS, and we're already 141 * under MS_RDONLY, then it is safe here. 142 */ 143 if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) 144 return; 145 146 errstr = btrfs_decode_error(errno); 147 if (fmt) { 148 struct va_format vaf; 149 va_list args; 150 151 va_start(args, fmt); 152 vaf.fmt = fmt; 153 vaf.va = &args; 154 155 printk(KERN_CRIT 156 "BTRFS: error (device %s) in %s:%d: errno=%d %s (%pV)\n", 157 sb->s_id, function, line, errno, errstr, &vaf); 158 va_end(args); 159 } else { 160 printk(KERN_CRIT "BTRFS: error (device %s) in %s:%d: errno=%d %s\n", 161 sb->s_id, function, line, errno, errstr); 162 } 163 164 /* Don't go through full error handling during mount */ 165 save_error_info(fs_info); 166 if (sb->s_flags & MS_BORN) 167 btrfs_handle_error(fs_info); 168 } 169 170 static const char * const logtypes[] = { 171 "emergency", 172 "alert", 173 "critical", 174 "error", 175 "warning", 176 "notice", 177 "info", 178 "debug", 179 }; 180 181 void btrfs_printk(const struct btrfs_fs_info *fs_info, const char *fmt, ...) 182 { 183 struct super_block *sb = fs_info->sb; 184 char lvl[4]; 185 struct va_format vaf; 186 va_list args; 187 const char *type = logtypes[4]; 188 int kern_level; 189 190 va_start(args, fmt); 191 192 kern_level = printk_get_level(fmt); 193 if (kern_level) { 194 size_t size = printk_skip_level(fmt) - fmt; 195 memcpy(lvl, fmt, size); 196 lvl[size] = '\0'; 197 fmt += size; 198 type = logtypes[kern_level - '0']; 199 } else 200 *lvl = '\0'; 201 202 vaf.fmt = fmt; 203 vaf.va = &args; 204 205 printk("%sBTRFS %s (device %s): %pV\n", lvl, type, sb->s_id, &vaf); 206 207 va_end(args); 208 } 209 210 #else 211 212 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, 213 unsigned int line, int errno, const char *fmt, ...) 214 { 215 struct super_block *sb = fs_info->sb; 216 217 /* 218 * Special case: if the error is EROFS, and we're already 219 * under MS_RDONLY, then it is safe here. 220 */ 221 if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) 222 return; 223 224 /* Don't go through full error handling during mount */ 225 if (sb->s_flags & MS_BORN) { 226 save_error_info(fs_info); 227 btrfs_handle_error(fs_info); 228 } 229 } 230 #endif 231 232 /* 233 * We only mark the transaction aborted and then set the file system read-only. 234 * This will prevent new transactions from starting or trying to join this 235 * one. 236 * 237 * This means that error recovery at the call site is limited to freeing 238 * any local memory allocations and passing the error code up without 239 * further cleanup. The transaction should complete as it normally would 240 * in the call path but will return -EIO. 241 * 242 * We'll complete the cleanup in btrfs_end_transaction and 243 * btrfs_commit_transaction. 244 */ 245 void __btrfs_abort_transaction(struct btrfs_trans_handle *trans, 246 struct btrfs_root *root, const char *function, 247 unsigned int line, int errno) 248 { 249 /* 250 * Report first abort since mount 251 */ 252 if (!test_and_set_bit(BTRFS_FS_STATE_TRANS_ABORTED, 253 &root->fs_info->fs_state)) { 254 WARN(1, KERN_DEBUG "BTRFS: Transaction aborted (error %d)\n", 255 errno); 256 } 257 trans->aborted = errno; 258 /* Nothing used. The other threads that have joined this 259 * transaction may be able to continue. */ 260 if (!trans->blocks_used) { 261 const char *errstr; 262 263 errstr = btrfs_decode_error(errno); 264 btrfs_warn(root->fs_info, 265 "%s:%d: Aborting unused transaction(%s).", 266 function, line, errstr); 267 return; 268 } 269 ACCESS_ONCE(trans->transaction->aborted) = errno; 270 /* Wake up anybody who may be waiting on this transaction */ 271 wake_up(&root->fs_info->transaction_wait); 272 wake_up(&root->fs_info->transaction_blocked_wait); 273 __btrfs_std_error(root->fs_info, function, line, errno, NULL); 274 } 275 /* 276 * __btrfs_panic decodes unexpected, fatal errors from the caller, 277 * issues an alert, and either panics or BUGs, depending on mount options. 278 */ 279 void __btrfs_panic(struct btrfs_fs_info *fs_info, const char *function, 280 unsigned int line, int errno, const char *fmt, ...) 281 { 282 char *s_id = "<unknown>"; 283 const char *errstr; 284 struct va_format vaf = { .fmt = fmt }; 285 va_list args; 286 287 if (fs_info) 288 s_id = fs_info->sb->s_id; 289 290 va_start(args, fmt); 291 vaf.va = &args; 292 293 errstr = btrfs_decode_error(errno); 294 if (fs_info && (fs_info->mount_opt & BTRFS_MOUNT_PANIC_ON_FATAL_ERROR)) 295 panic(KERN_CRIT "BTRFS panic (device %s) in %s:%d: %pV (errno=%d %s)\n", 296 s_id, function, line, &vaf, errno, errstr); 297 298 btrfs_crit(fs_info, "panic in %s:%d: %pV (errno=%d %s)", 299 function, line, &vaf, errno, errstr); 300 va_end(args); 301 /* Caller calls BUG() */ 302 } 303 304 static void btrfs_put_super(struct super_block *sb) 305 { 306 (void)close_ctree(btrfs_sb(sb)->tree_root); 307 /* FIXME: need to fix VFS to return error? */ 308 /* AV: return it _where_? ->put_super() can be triggered by any number 309 * of async events, up to and including delivery of SIGKILL to the 310 * last process that kept it busy. Or segfault in the aforementioned 311 * process... Whom would you report that to? 312 */ 313 } 314 315 enum { 316 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, 317 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, 318 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, 319 Opt_compress_type, Opt_compress_force, Opt_compress_force_type, 320 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, 321 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, 322 Opt_enospc_debug, Opt_subvolrootid, Opt_defrag, Opt_inode_cache, 323 Opt_no_space_cache, Opt_recovery, Opt_skip_balance, 324 Opt_check_integrity, Opt_check_integrity_including_extent_data, 325 Opt_check_integrity_print_mask, Opt_fatal_errors, Opt_rescan_uuid_tree, 326 Opt_commit_interval, Opt_barrier, Opt_nodefrag, Opt_nodiscard, 327 Opt_noenospc_debug, 328 Opt_err, 329 }; 330 331 static match_table_t tokens = { 332 {Opt_degraded, "degraded"}, 333 {Opt_subvol, "subvol=%s"}, 334 {Opt_subvolid, "subvolid=%s"}, 335 {Opt_device, "device=%s"}, 336 {Opt_nodatasum, "nodatasum"}, 337 {Opt_nodatacow, "nodatacow"}, 338 {Opt_nobarrier, "nobarrier"}, 339 {Opt_barrier, "barrier"}, 340 {Opt_max_inline, "max_inline=%s"}, 341 {Opt_alloc_start, "alloc_start=%s"}, 342 {Opt_thread_pool, "thread_pool=%d"}, 343 {Opt_compress, "compress"}, 344 {Opt_compress_type, "compress=%s"}, 345 {Opt_compress_force, "compress-force"}, 346 {Opt_compress_force_type, "compress-force=%s"}, 347 {Opt_ssd, "ssd"}, 348 {Opt_ssd_spread, "ssd_spread"}, 349 {Opt_nossd, "nossd"}, 350 {Opt_noacl, "noacl"}, 351 {Opt_notreelog, "notreelog"}, 352 {Opt_flushoncommit, "flushoncommit"}, 353 {Opt_ratio, "metadata_ratio=%d"}, 354 {Opt_discard, "discard"}, 355 {Opt_nodiscard, "nodiscard"}, 356 {Opt_space_cache, "space_cache"}, 357 {Opt_clear_cache, "clear_cache"}, 358 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 359 {Opt_enospc_debug, "enospc_debug"}, 360 {Opt_noenospc_debug, "noenospc_debug"}, 361 {Opt_subvolrootid, "subvolrootid=%d"}, 362 {Opt_defrag, "autodefrag"}, 363 {Opt_nodefrag, "noautodefrag"}, 364 {Opt_inode_cache, "inode_cache"}, 365 {Opt_no_space_cache, "nospace_cache"}, 366 {Opt_recovery, "recovery"}, 367 {Opt_skip_balance, "skip_balance"}, 368 {Opt_check_integrity, "check_int"}, 369 {Opt_check_integrity_including_extent_data, "check_int_data"}, 370 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"}, 371 {Opt_rescan_uuid_tree, "rescan_uuid_tree"}, 372 {Opt_fatal_errors, "fatal_errors=%s"}, 373 {Opt_commit_interval, "commit=%d"}, 374 {Opt_err, NULL}, 375 }; 376 377 /* 378 * Regular mount options parser. Everything that is needed only when 379 * reading in a new superblock is parsed here. 380 * XXX JDM: This needs to be cleaned up for remount. 381 */ 382 int btrfs_parse_options(struct btrfs_root *root, char *options) 383 { 384 struct btrfs_fs_info *info = root->fs_info; 385 substring_t args[MAX_OPT_ARGS]; 386 char *p, *num, *orig = NULL; 387 u64 cache_gen; 388 int intarg; 389 int ret = 0; 390 char *compress_type; 391 bool compress_force = false; 392 393 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy); 394 if (cache_gen) 395 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 396 397 if (!options) 398 goto out; 399 400 /* 401 * strsep changes the string, duplicate it because parse_options 402 * gets called twice 403 */ 404 options = kstrdup(options, GFP_NOFS); 405 if (!options) 406 return -ENOMEM; 407 408 orig = options; 409 410 while ((p = strsep(&options, ",")) != NULL) { 411 int token; 412 if (!*p) 413 continue; 414 415 token = match_token(p, tokens, args); 416 switch (token) { 417 case Opt_degraded: 418 btrfs_info(root->fs_info, "allowing degraded mounts"); 419 btrfs_set_opt(info->mount_opt, DEGRADED); 420 break; 421 case Opt_subvol: 422 case Opt_subvolid: 423 case Opt_subvolrootid: 424 case Opt_device: 425 /* 426 * These are parsed by btrfs_parse_early_options 427 * and can be happily ignored here. 428 */ 429 break; 430 case Opt_nodatasum: 431 btrfs_info(root->fs_info, "setting nodatasum"); 432 btrfs_set_opt(info->mount_opt, NODATASUM); 433 break; 434 case Opt_nodatacow: 435 if (!btrfs_test_opt(root, COMPRESS) || 436 !btrfs_test_opt(root, FORCE_COMPRESS)) { 437 btrfs_info(root->fs_info, 438 "setting nodatacow, compression disabled"); 439 } else { 440 btrfs_info(root->fs_info, "setting nodatacow"); 441 } 442 btrfs_clear_opt(info->mount_opt, COMPRESS); 443 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 444 btrfs_set_opt(info->mount_opt, NODATACOW); 445 btrfs_set_opt(info->mount_opt, NODATASUM); 446 break; 447 case Opt_compress_force: 448 case Opt_compress_force_type: 449 compress_force = true; 450 /* Fallthrough */ 451 case Opt_compress: 452 case Opt_compress_type: 453 if (token == Opt_compress || 454 token == Opt_compress_force || 455 strcmp(args[0].from, "zlib") == 0) { 456 compress_type = "zlib"; 457 info->compress_type = BTRFS_COMPRESS_ZLIB; 458 btrfs_set_opt(info->mount_opt, COMPRESS); 459 btrfs_clear_opt(info->mount_opt, NODATACOW); 460 btrfs_clear_opt(info->mount_opt, NODATASUM); 461 } else if (strcmp(args[0].from, "lzo") == 0) { 462 compress_type = "lzo"; 463 info->compress_type = BTRFS_COMPRESS_LZO; 464 btrfs_set_opt(info->mount_opt, COMPRESS); 465 btrfs_clear_opt(info->mount_opt, NODATACOW); 466 btrfs_clear_opt(info->mount_opt, NODATASUM); 467 btrfs_set_fs_incompat(info, COMPRESS_LZO); 468 } else if (strncmp(args[0].from, "no", 2) == 0) { 469 compress_type = "no"; 470 btrfs_clear_opt(info->mount_opt, COMPRESS); 471 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 472 compress_force = false; 473 } else { 474 ret = -EINVAL; 475 goto out; 476 } 477 478 if (compress_force) { 479 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 480 btrfs_info(root->fs_info, "force %s compression", 481 compress_type); 482 } else if (btrfs_test_opt(root, COMPRESS)) { 483 pr_info("btrfs: use %s compression\n", 484 compress_type); 485 } 486 break; 487 case Opt_ssd: 488 btrfs_info(root->fs_info, "use ssd allocation scheme"); 489 btrfs_set_opt(info->mount_opt, SSD); 490 break; 491 case Opt_ssd_spread: 492 btrfs_info(root->fs_info, "use spread ssd allocation scheme"); 493 btrfs_set_opt(info->mount_opt, SSD); 494 btrfs_set_opt(info->mount_opt, SSD_SPREAD); 495 break; 496 case Opt_nossd: 497 btrfs_info(root->fs_info, "not using ssd allocation scheme"); 498 btrfs_set_opt(info->mount_opt, NOSSD); 499 btrfs_clear_opt(info->mount_opt, SSD); 500 btrfs_clear_opt(info->mount_opt, SSD_SPREAD); 501 break; 502 case Opt_barrier: 503 if (btrfs_test_opt(root, NOBARRIER)) 504 btrfs_info(root->fs_info, "turning on barriers"); 505 btrfs_clear_opt(info->mount_opt, NOBARRIER); 506 break; 507 case Opt_nobarrier: 508 btrfs_info(root->fs_info, "turning off barriers"); 509 btrfs_set_opt(info->mount_opt, NOBARRIER); 510 break; 511 case Opt_thread_pool: 512 ret = match_int(&args[0], &intarg); 513 if (ret) { 514 goto out; 515 } else if (intarg > 0) { 516 info->thread_pool_size = intarg; 517 } else { 518 ret = -EINVAL; 519 goto out; 520 } 521 break; 522 case Opt_max_inline: 523 num = match_strdup(&args[0]); 524 if (num) { 525 info->max_inline = memparse(num, NULL); 526 kfree(num); 527 528 if (info->max_inline) { 529 info->max_inline = max_t(u64, 530 info->max_inline, 531 root->sectorsize); 532 } 533 btrfs_info(root->fs_info, "max_inline at %llu", 534 info->max_inline); 535 } else { 536 ret = -ENOMEM; 537 goto out; 538 } 539 break; 540 case Opt_alloc_start: 541 num = match_strdup(&args[0]); 542 if (num) { 543 mutex_lock(&info->chunk_mutex); 544 info->alloc_start = memparse(num, NULL); 545 mutex_unlock(&info->chunk_mutex); 546 kfree(num); 547 btrfs_info(root->fs_info, "allocations start at %llu", 548 info->alloc_start); 549 } else { 550 ret = -ENOMEM; 551 goto out; 552 } 553 break; 554 case Opt_noacl: 555 root->fs_info->sb->s_flags &= ~MS_POSIXACL; 556 break; 557 case Opt_notreelog: 558 btrfs_info(root->fs_info, "disabling tree log"); 559 btrfs_set_opt(info->mount_opt, NOTREELOG); 560 break; 561 case Opt_flushoncommit: 562 btrfs_info(root->fs_info, "turning on flush-on-commit"); 563 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); 564 break; 565 case Opt_ratio: 566 ret = match_int(&args[0], &intarg); 567 if (ret) { 568 goto out; 569 } else if (intarg >= 0) { 570 info->metadata_ratio = intarg; 571 btrfs_info(root->fs_info, "metadata ratio %d", 572 info->metadata_ratio); 573 } else { 574 ret = -EINVAL; 575 goto out; 576 } 577 break; 578 case Opt_discard: 579 btrfs_set_opt(info->mount_opt, DISCARD); 580 break; 581 case Opt_nodiscard: 582 btrfs_clear_opt(info->mount_opt, DISCARD); 583 break; 584 case Opt_space_cache: 585 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 586 break; 587 case Opt_rescan_uuid_tree: 588 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE); 589 break; 590 case Opt_no_space_cache: 591 btrfs_info(root->fs_info, "disabling disk space caching"); 592 btrfs_clear_opt(info->mount_opt, SPACE_CACHE); 593 break; 594 case Opt_inode_cache: 595 btrfs_info(root->fs_info, "enabling inode map caching"); 596 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); 597 break; 598 case Opt_clear_cache: 599 btrfs_info(root->fs_info, "force clearing of disk cache"); 600 btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 601 break; 602 case Opt_user_subvol_rm_allowed: 603 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 604 break; 605 case Opt_enospc_debug: 606 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); 607 break; 608 case Opt_noenospc_debug: 609 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG); 610 break; 611 case Opt_defrag: 612 btrfs_info(root->fs_info, "enabling auto defrag"); 613 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG); 614 break; 615 case Opt_nodefrag: 616 if (btrfs_test_opt(root, AUTO_DEFRAG)) 617 btrfs_info(root->fs_info, "disabling auto defrag"); 618 btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG); 619 break; 620 case Opt_recovery: 621 btrfs_info(root->fs_info, "enabling auto recovery"); 622 btrfs_set_opt(info->mount_opt, RECOVERY); 623 break; 624 case Opt_skip_balance: 625 btrfs_set_opt(info->mount_opt, SKIP_BALANCE); 626 break; 627 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 628 case Opt_check_integrity_including_extent_data: 629 btrfs_info(root->fs_info, 630 "enabling check integrity including extent data"); 631 btrfs_set_opt(info->mount_opt, 632 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA); 633 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 634 break; 635 case Opt_check_integrity: 636 btrfs_info(root->fs_info, "enabling check integrity"); 637 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 638 break; 639 case Opt_check_integrity_print_mask: 640 ret = match_int(&args[0], &intarg); 641 if (ret) { 642 goto out; 643 } else if (intarg >= 0) { 644 info->check_integrity_print_mask = intarg; 645 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x", 646 info->check_integrity_print_mask); 647 } else { 648 ret = -EINVAL; 649 goto out; 650 } 651 break; 652 #else 653 case Opt_check_integrity_including_extent_data: 654 case Opt_check_integrity: 655 case Opt_check_integrity_print_mask: 656 btrfs_err(root->fs_info, 657 "support for check_integrity* not compiled in!"); 658 ret = -EINVAL; 659 goto out; 660 #endif 661 case Opt_fatal_errors: 662 if (strcmp(args[0].from, "panic") == 0) 663 btrfs_set_opt(info->mount_opt, 664 PANIC_ON_FATAL_ERROR); 665 else if (strcmp(args[0].from, "bug") == 0) 666 btrfs_clear_opt(info->mount_opt, 667 PANIC_ON_FATAL_ERROR); 668 else { 669 ret = -EINVAL; 670 goto out; 671 } 672 break; 673 case Opt_commit_interval: 674 intarg = 0; 675 ret = match_int(&args[0], &intarg); 676 if (ret < 0) { 677 btrfs_err(root->fs_info, "invalid commit interval"); 678 ret = -EINVAL; 679 goto out; 680 } 681 if (intarg > 0) { 682 if (intarg > 300) { 683 btrfs_warn(root->fs_info, "excessive commit interval %d", 684 intarg); 685 } 686 info->commit_interval = intarg; 687 } else { 688 btrfs_info(root->fs_info, "using default commit interval %ds", 689 BTRFS_DEFAULT_COMMIT_INTERVAL); 690 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 691 } 692 break; 693 case Opt_err: 694 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p); 695 ret = -EINVAL; 696 goto out; 697 default: 698 break; 699 } 700 } 701 out: 702 if (!ret && btrfs_test_opt(root, SPACE_CACHE)) 703 btrfs_info(root->fs_info, "disk space caching is enabled"); 704 kfree(orig); 705 return ret; 706 } 707 708 /* 709 * Parse mount options that are required early in the mount process. 710 * 711 * All other options will be parsed on much later in the mount process and 712 * only when we need to allocate a new super block. 713 */ 714 static int btrfs_parse_early_options(const char *options, fmode_t flags, 715 void *holder, char **subvol_name, u64 *subvol_objectid, 716 struct btrfs_fs_devices **fs_devices) 717 { 718 substring_t args[MAX_OPT_ARGS]; 719 char *device_name, *opts, *orig, *p; 720 char *num = NULL; 721 int error = 0; 722 723 if (!options) 724 return 0; 725 726 /* 727 * strsep changes the string, duplicate it because parse_options 728 * gets called twice 729 */ 730 opts = kstrdup(options, GFP_KERNEL); 731 if (!opts) 732 return -ENOMEM; 733 orig = opts; 734 735 while ((p = strsep(&opts, ",")) != NULL) { 736 int token; 737 if (!*p) 738 continue; 739 740 token = match_token(p, tokens, args); 741 switch (token) { 742 case Opt_subvol: 743 kfree(*subvol_name); 744 *subvol_name = match_strdup(&args[0]); 745 if (!*subvol_name) { 746 error = -ENOMEM; 747 goto out; 748 } 749 break; 750 case Opt_subvolid: 751 num = match_strdup(&args[0]); 752 if (num) { 753 *subvol_objectid = memparse(num, NULL); 754 kfree(num); 755 /* we want the original fs_tree */ 756 if (!*subvol_objectid) 757 *subvol_objectid = 758 BTRFS_FS_TREE_OBJECTID; 759 } else { 760 error = -EINVAL; 761 goto out; 762 } 763 break; 764 case Opt_subvolrootid: 765 printk(KERN_WARNING 766 "BTRFS: 'subvolrootid' mount option is deprecated and has " 767 "no effect\n"); 768 break; 769 case Opt_device: 770 device_name = match_strdup(&args[0]); 771 if (!device_name) { 772 error = -ENOMEM; 773 goto out; 774 } 775 error = btrfs_scan_one_device(device_name, 776 flags, holder, fs_devices); 777 kfree(device_name); 778 if (error) 779 goto out; 780 break; 781 default: 782 break; 783 } 784 } 785 786 out: 787 kfree(orig); 788 return error; 789 } 790 791 static struct dentry *get_default_root(struct super_block *sb, 792 u64 subvol_objectid) 793 { 794 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 795 struct btrfs_root *root = fs_info->tree_root; 796 struct btrfs_root *new_root; 797 struct btrfs_dir_item *di; 798 struct btrfs_path *path; 799 struct btrfs_key location; 800 struct inode *inode; 801 u64 dir_id; 802 int new = 0; 803 804 /* 805 * We have a specific subvol we want to mount, just setup location and 806 * go look up the root. 807 */ 808 if (subvol_objectid) { 809 location.objectid = subvol_objectid; 810 location.type = BTRFS_ROOT_ITEM_KEY; 811 location.offset = (u64)-1; 812 goto find_root; 813 } 814 815 path = btrfs_alloc_path(); 816 if (!path) 817 return ERR_PTR(-ENOMEM); 818 path->leave_spinning = 1; 819 820 /* 821 * Find the "default" dir item which points to the root item that we 822 * will mount by default if we haven't been given a specific subvolume 823 * to mount. 824 */ 825 dir_id = btrfs_super_root_dir(fs_info->super_copy); 826 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); 827 if (IS_ERR(di)) { 828 btrfs_free_path(path); 829 return ERR_CAST(di); 830 } 831 if (!di) { 832 /* 833 * Ok the default dir item isn't there. This is weird since 834 * it's always been there, but don't freak out, just try and 835 * mount to root most subvolume. 836 */ 837 btrfs_free_path(path); 838 dir_id = BTRFS_FIRST_FREE_OBJECTID; 839 new_root = fs_info->fs_root; 840 goto setup_root; 841 } 842 843 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 844 btrfs_free_path(path); 845 846 find_root: 847 new_root = btrfs_read_fs_root_no_name(fs_info, &location); 848 if (IS_ERR(new_root)) 849 return ERR_CAST(new_root); 850 851 dir_id = btrfs_root_dirid(&new_root->root_item); 852 setup_root: 853 location.objectid = dir_id; 854 location.type = BTRFS_INODE_ITEM_KEY; 855 location.offset = 0; 856 857 inode = btrfs_iget(sb, &location, new_root, &new); 858 if (IS_ERR(inode)) 859 return ERR_CAST(inode); 860 861 /* 862 * If we're just mounting the root most subvol put the inode and return 863 * a reference to the dentry. We will have already gotten a reference 864 * to the inode in btrfs_fill_super so we're good to go. 865 */ 866 if (!new && sb->s_root->d_inode == inode) { 867 iput(inode); 868 return dget(sb->s_root); 869 } 870 871 return d_obtain_alias(inode); 872 } 873 874 static int btrfs_fill_super(struct super_block *sb, 875 struct btrfs_fs_devices *fs_devices, 876 void *data, int silent) 877 { 878 struct inode *inode; 879 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 880 struct btrfs_key key; 881 int err; 882 883 sb->s_maxbytes = MAX_LFS_FILESIZE; 884 sb->s_magic = BTRFS_SUPER_MAGIC; 885 sb->s_op = &btrfs_super_ops; 886 sb->s_d_op = &btrfs_dentry_operations; 887 sb->s_export_op = &btrfs_export_ops; 888 sb->s_xattr = btrfs_xattr_handlers; 889 sb->s_time_gran = 1; 890 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 891 sb->s_flags |= MS_POSIXACL; 892 #endif 893 sb->s_flags |= MS_I_VERSION; 894 err = open_ctree(sb, fs_devices, (char *)data); 895 if (err) { 896 printk(KERN_ERR "BTRFS: open_ctree failed\n"); 897 return err; 898 } 899 900 key.objectid = BTRFS_FIRST_FREE_OBJECTID; 901 key.type = BTRFS_INODE_ITEM_KEY; 902 key.offset = 0; 903 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL); 904 if (IS_ERR(inode)) { 905 err = PTR_ERR(inode); 906 goto fail_close; 907 } 908 909 sb->s_root = d_make_root(inode); 910 if (!sb->s_root) { 911 err = -ENOMEM; 912 goto fail_close; 913 } 914 915 save_mount_options(sb, data); 916 cleancache_init_fs(sb); 917 sb->s_flags |= MS_ACTIVE; 918 return 0; 919 920 fail_close: 921 close_ctree(fs_info->tree_root); 922 return err; 923 } 924 925 int btrfs_sync_fs(struct super_block *sb, int wait) 926 { 927 struct btrfs_trans_handle *trans; 928 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 929 struct btrfs_root *root = fs_info->tree_root; 930 931 trace_btrfs_sync_fs(wait); 932 933 if (!wait) { 934 filemap_flush(fs_info->btree_inode->i_mapping); 935 return 0; 936 } 937 938 btrfs_wait_ordered_roots(fs_info, -1); 939 940 trans = btrfs_attach_transaction_barrier(root); 941 if (IS_ERR(trans)) { 942 /* no transaction, don't bother */ 943 if (PTR_ERR(trans) == -ENOENT) 944 return 0; 945 return PTR_ERR(trans); 946 } 947 return btrfs_commit_transaction(trans, root); 948 } 949 950 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 951 { 952 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 953 struct btrfs_root *root = info->tree_root; 954 char *compress_type; 955 956 if (btrfs_test_opt(root, DEGRADED)) 957 seq_puts(seq, ",degraded"); 958 if (btrfs_test_opt(root, NODATASUM)) 959 seq_puts(seq, ",nodatasum"); 960 if (btrfs_test_opt(root, NODATACOW)) 961 seq_puts(seq, ",nodatacow"); 962 if (btrfs_test_opt(root, NOBARRIER)) 963 seq_puts(seq, ",nobarrier"); 964 if (info->max_inline != 8192 * 1024) 965 seq_printf(seq, ",max_inline=%llu", info->max_inline); 966 if (info->alloc_start != 0) 967 seq_printf(seq, ",alloc_start=%llu", info->alloc_start); 968 if (info->thread_pool_size != min_t(unsigned long, 969 num_online_cpus() + 2, 8)) 970 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); 971 if (btrfs_test_opt(root, COMPRESS)) { 972 if (info->compress_type == BTRFS_COMPRESS_ZLIB) 973 compress_type = "zlib"; 974 else 975 compress_type = "lzo"; 976 if (btrfs_test_opt(root, FORCE_COMPRESS)) 977 seq_printf(seq, ",compress-force=%s", compress_type); 978 else 979 seq_printf(seq, ",compress=%s", compress_type); 980 } 981 if (btrfs_test_opt(root, NOSSD)) 982 seq_puts(seq, ",nossd"); 983 if (btrfs_test_opt(root, SSD_SPREAD)) 984 seq_puts(seq, ",ssd_spread"); 985 else if (btrfs_test_opt(root, SSD)) 986 seq_puts(seq, ",ssd"); 987 if (btrfs_test_opt(root, NOTREELOG)) 988 seq_puts(seq, ",notreelog"); 989 if (btrfs_test_opt(root, FLUSHONCOMMIT)) 990 seq_puts(seq, ",flushoncommit"); 991 if (btrfs_test_opt(root, DISCARD)) 992 seq_puts(seq, ",discard"); 993 if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) 994 seq_puts(seq, ",noacl"); 995 if (btrfs_test_opt(root, SPACE_CACHE)) 996 seq_puts(seq, ",space_cache"); 997 else 998 seq_puts(seq, ",nospace_cache"); 999 if (btrfs_test_opt(root, RESCAN_UUID_TREE)) 1000 seq_puts(seq, ",rescan_uuid_tree"); 1001 if (btrfs_test_opt(root, CLEAR_CACHE)) 1002 seq_puts(seq, ",clear_cache"); 1003 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 1004 seq_puts(seq, ",user_subvol_rm_allowed"); 1005 if (btrfs_test_opt(root, ENOSPC_DEBUG)) 1006 seq_puts(seq, ",enospc_debug"); 1007 if (btrfs_test_opt(root, AUTO_DEFRAG)) 1008 seq_puts(seq, ",autodefrag"); 1009 if (btrfs_test_opt(root, INODE_MAP_CACHE)) 1010 seq_puts(seq, ",inode_cache"); 1011 if (btrfs_test_opt(root, SKIP_BALANCE)) 1012 seq_puts(seq, ",skip_balance"); 1013 if (btrfs_test_opt(root, RECOVERY)) 1014 seq_puts(seq, ",recovery"); 1015 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 1016 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA)) 1017 seq_puts(seq, ",check_int_data"); 1018 else if (btrfs_test_opt(root, CHECK_INTEGRITY)) 1019 seq_puts(seq, ",check_int"); 1020 if (info->check_integrity_print_mask) 1021 seq_printf(seq, ",check_int_print_mask=%d", 1022 info->check_integrity_print_mask); 1023 #endif 1024 if (info->metadata_ratio) 1025 seq_printf(seq, ",metadata_ratio=%d", 1026 info->metadata_ratio); 1027 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR)) 1028 seq_puts(seq, ",fatal_errors=panic"); 1029 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1030 seq_printf(seq, ",commit=%d", info->commit_interval); 1031 return 0; 1032 } 1033 1034 static int btrfs_test_super(struct super_block *s, void *data) 1035 { 1036 struct btrfs_fs_info *p = data; 1037 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1038 1039 return fs_info->fs_devices == p->fs_devices; 1040 } 1041 1042 static int btrfs_set_super(struct super_block *s, void *data) 1043 { 1044 int err = set_anon_super(s, data); 1045 if (!err) 1046 s->s_fs_info = data; 1047 return err; 1048 } 1049 1050 /* 1051 * subvolumes are identified by ino 256 1052 */ 1053 static inline int is_subvolume_inode(struct inode *inode) 1054 { 1055 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1056 return 1; 1057 return 0; 1058 } 1059 1060 /* 1061 * This will strip out the subvol=%s argument for an argument string and add 1062 * subvolid=0 to make sure we get the actual tree root for path walking to the 1063 * subvol we want. 1064 */ 1065 static char *setup_root_args(char *args) 1066 { 1067 unsigned len = strlen(args) + 2 + 1; 1068 char *src, *dst, *buf; 1069 1070 /* 1071 * We need the same args as before, but with this substitution: 1072 * s!subvol=[^,]+!subvolid=0! 1073 * 1074 * Since the replacement string is up to 2 bytes longer than the 1075 * original, allocate strlen(args) + 2 + 1 bytes. 1076 */ 1077 1078 src = strstr(args, "subvol="); 1079 /* This shouldn't happen, but just in case.. */ 1080 if (!src) 1081 return NULL; 1082 1083 buf = dst = kmalloc(len, GFP_NOFS); 1084 if (!buf) 1085 return NULL; 1086 1087 /* 1088 * If the subvol= arg is not at the start of the string, 1089 * copy whatever precedes it into buf. 1090 */ 1091 if (src != args) { 1092 *src++ = '\0'; 1093 strcpy(buf, args); 1094 dst += strlen(args); 1095 } 1096 1097 strcpy(dst, "subvolid=0"); 1098 dst += strlen("subvolid=0"); 1099 1100 /* 1101 * If there is a "," after the original subvol=... string, 1102 * copy that suffix into our buffer. Otherwise, we're done. 1103 */ 1104 src = strchr(src, ','); 1105 if (src) 1106 strcpy(dst, src); 1107 1108 return buf; 1109 } 1110 1111 static struct dentry *mount_subvol(const char *subvol_name, int flags, 1112 const char *device_name, char *data) 1113 { 1114 struct dentry *root; 1115 struct vfsmount *mnt; 1116 char *newargs; 1117 1118 newargs = setup_root_args(data); 1119 if (!newargs) 1120 return ERR_PTR(-ENOMEM); 1121 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, 1122 newargs); 1123 kfree(newargs); 1124 if (IS_ERR(mnt)) 1125 return ERR_CAST(mnt); 1126 1127 root = mount_subtree(mnt, subvol_name); 1128 1129 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) { 1130 struct super_block *s = root->d_sb; 1131 dput(root); 1132 root = ERR_PTR(-EINVAL); 1133 deactivate_locked_super(s); 1134 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n", 1135 subvol_name); 1136 } 1137 1138 return root; 1139 } 1140 1141 /* 1142 * Find a superblock for the given device / mount point. 1143 * 1144 * Note: This is based on get_sb_bdev from fs/super.c with a few additions 1145 * for multiple device setup. Make sure to keep it in sync. 1146 */ 1147 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, 1148 const char *device_name, void *data) 1149 { 1150 struct block_device *bdev = NULL; 1151 struct super_block *s; 1152 struct dentry *root; 1153 struct btrfs_fs_devices *fs_devices = NULL; 1154 struct btrfs_fs_info *fs_info = NULL; 1155 fmode_t mode = FMODE_READ; 1156 char *subvol_name = NULL; 1157 u64 subvol_objectid = 0; 1158 int error = 0; 1159 1160 if (!(flags & MS_RDONLY)) 1161 mode |= FMODE_WRITE; 1162 1163 error = btrfs_parse_early_options(data, mode, fs_type, 1164 &subvol_name, &subvol_objectid, 1165 &fs_devices); 1166 if (error) { 1167 kfree(subvol_name); 1168 return ERR_PTR(error); 1169 } 1170 1171 if (subvol_name) { 1172 root = mount_subvol(subvol_name, flags, device_name, data); 1173 kfree(subvol_name); 1174 return root; 1175 } 1176 1177 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices); 1178 if (error) 1179 return ERR_PTR(error); 1180 1181 /* 1182 * Setup a dummy root and fs_info for test/set super. This is because 1183 * we don't actually fill this stuff out until open_ctree, but we need 1184 * it for searching for existing supers, so this lets us do that and 1185 * then open_ctree will properly initialize everything later. 1186 */ 1187 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); 1188 if (!fs_info) 1189 return ERR_PTR(-ENOMEM); 1190 1191 fs_info->fs_devices = fs_devices; 1192 1193 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 1194 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 1195 if (!fs_info->super_copy || !fs_info->super_for_commit) { 1196 error = -ENOMEM; 1197 goto error_fs_info; 1198 } 1199 1200 error = btrfs_open_devices(fs_devices, mode, fs_type); 1201 if (error) 1202 goto error_fs_info; 1203 1204 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { 1205 error = -EACCES; 1206 goto error_close_devices; 1207 } 1208 1209 bdev = fs_devices->latest_bdev; 1210 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC, 1211 fs_info); 1212 if (IS_ERR(s)) { 1213 error = PTR_ERR(s); 1214 goto error_close_devices; 1215 } 1216 1217 if (s->s_root) { 1218 btrfs_close_devices(fs_devices); 1219 free_fs_info(fs_info); 1220 if ((flags ^ s->s_flags) & MS_RDONLY) 1221 error = -EBUSY; 1222 } else { 1223 char b[BDEVNAME_SIZE]; 1224 1225 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); 1226 btrfs_sb(s)->bdev_holder = fs_type; 1227 error = btrfs_fill_super(s, fs_devices, data, 1228 flags & MS_SILENT ? 1 : 0); 1229 } 1230 1231 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error); 1232 if (IS_ERR(root)) 1233 deactivate_locked_super(s); 1234 1235 return root; 1236 1237 error_close_devices: 1238 btrfs_close_devices(fs_devices); 1239 error_fs_info: 1240 free_fs_info(fs_info); 1241 return ERR_PTR(error); 1242 } 1243 1244 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit) 1245 { 1246 spin_lock_irq(&workers->lock); 1247 workers->max_workers = new_limit; 1248 spin_unlock_irq(&workers->lock); 1249 } 1250 1251 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 1252 int new_pool_size, int old_pool_size) 1253 { 1254 if (new_pool_size == old_pool_size) 1255 return; 1256 1257 fs_info->thread_pool_size = new_pool_size; 1258 1259 btrfs_info(fs_info, "resize thread pool %d -> %d", 1260 old_pool_size, new_pool_size); 1261 1262 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size); 1263 btrfs_set_max_workers(&fs_info->workers, new_pool_size); 1264 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size); 1265 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size); 1266 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size); 1267 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size); 1268 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size); 1269 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size); 1270 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size); 1271 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size); 1272 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size); 1273 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size); 1274 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size); 1275 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers, 1276 new_pool_size); 1277 } 1278 1279 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info) 1280 { 1281 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1282 } 1283 1284 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1285 unsigned long old_opts, int flags) 1286 { 1287 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1288 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1289 (flags & MS_RDONLY))) { 1290 /* wait for any defraggers to finish */ 1291 wait_event(fs_info->transaction_wait, 1292 (atomic_read(&fs_info->defrag_running) == 0)); 1293 if (flags & MS_RDONLY) 1294 sync_filesystem(fs_info->sb); 1295 } 1296 } 1297 1298 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1299 unsigned long old_opts) 1300 { 1301 /* 1302 * We need cleanup all defragable inodes if the autodefragment is 1303 * close or the fs is R/O. 1304 */ 1305 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1306 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1307 (fs_info->sb->s_flags & MS_RDONLY))) { 1308 btrfs_cleanup_defrag_inodes(fs_info); 1309 } 1310 1311 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1312 } 1313 1314 static int btrfs_remount(struct super_block *sb, int *flags, char *data) 1315 { 1316 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1317 struct btrfs_root *root = fs_info->tree_root; 1318 unsigned old_flags = sb->s_flags; 1319 unsigned long old_opts = fs_info->mount_opt; 1320 unsigned long old_compress_type = fs_info->compress_type; 1321 u64 old_max_inline = fs_info->max_inline; 1322 u64 old_alloc_start = fs_info->alloc_start; 1323 int old_thread_pool_size = fs_info->thread_pool_size; 1324 unsigned int old_metadata_ratio = fs_info->metadata_ratio; 1325 int ret; 1326 1327 btrfs_remount_prepare(fs_info); 1328 1329 ret = btrfs_parse_options(root, data); 1330 if (ret) { 1331 ret = -EINVAL; 1332 goto restore; 1333 } 1334 1335 btrfs_remount_begin(fs_info, old_opts, *flags); 1336 btrfs_resize_thread_pool(fs_info, 1337 fs_info->thread_pool_size, old_thread_pool_size); 1338 1339 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) 1340 goto out; 1341 1342 if (*flags & MS_RDONLY) { 1343 /* 1344 * this also happens on 'umount -rf' or on shutdown, when 1345 * the filesystem is busy. 1346 */ 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 sb->s_flags |= MS_RDONLY; 1354 1355 btrfs_dev_replace_suspend_for_unmount(fs_info); 1356 btrfs_scrub_cancel(fs_info); 1357 btrfs_pause_balance(fs_info); 1358 1359 ret = btrfs_commit_super(root); 1360 if (ret) 1361 goto restore; 1362 } else { 1363 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) { 1364 btrfs_err(fs_info, 1365 "Remounting read-write after error is not allowed"); 1366 ret = -EINVAL; 1367 goto restore; 1368 } 1369 if (fs_info->fs_devices->rw_devices == 0) { 1370 ret = -EACCES; 1371 goto restore; 1372 } 1373 1374 if (fs_info->fs_devices->missing_devices > 1375 fs_info->num_tolerated_disk_barrier_failures && 1376 !(*flags & MS_RDONLY)) { 1377 btrfs_warn(fs_info, 1378 "too many missing devices, writeable remount is not allowed"); 1379 ret = -EACCES; 1380 goto restore; 1381 } 1382 1383 if (btrfs_super_log_root(fs_info->super_copy) != 0) { 1384 ret = -EINVAL; 1385 goto restore; 1386 } 1387 1388 ret = btrfs_cleanup_fs_roots(fs_info); 1389 if (ret) 1390 goto restore; 1391 1392 /* recover relocation */ 1393 ret = btrfs_recover_relocation(root); 1394 if (ret) 1395 goto restore; 1396 1397 ret = btrfs_resume_balance_async(fs_info); 1398 if (ret) 1399 goto restore; 1400 1401 ret = btrfs_resume_dev_replace_async(fs_info); 1402 if (ret) { 1403 btrfs_warn(fs_info, "failed to resume dev_replace"); 1404 goto restore; 1405 } 1406 1407 if (!fs_info->uuid_root) { 1408 btrfs_info(fs_info, "creating UUID tree"); 1409 ret = btrfs_create_uuid_tree(fs_info); 1410 if (ret) { 1411 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret); 1412 goto restore; 1413 } 1414 } 1415 sb->s_flags &= ~MS_RDONLY; 1416 } 1417 out: 1418 btrfs_remount_cleanup(fs_info, old_opts); 1419 return 0; 1420 1421 restore: 1422 /* We've hit an error - don't reset MS_RDONLY */ 1423 if (sb->s_flags & MS_RDONLY) 1424 old_flags |= MS_RDONLY; 1425 sb->s_flags = old_flags; 1426 fs_info->mount_opt = old_opts; 1427 fs_info->compress_type = old_compress_type; 1428 fs_info->max_inline = old_max_inline; 1429 mutex_lock(&fs_info->chunk_mutex); 1430 fs_info->alloc_start = old_alloc_start; 1431 mutex_unlock(&fs_info->chunk_mutex); 1432 btrfs_resize_thread_pool(fs_info, 1433 old_thread_pool_size, fs_info->thread_pool_size); 1434 fs_info->metadata_ratio = old_metadata_ratio; 1435 btrfs_remount_cleanup(fs_info, old_opts); 1436 return ret; 1437 } 1438 1439 /* Used to sort the devices by max_avail(descending sort) */ 1440 static int btrfs_cmp_device_free_bytes(const void *dev_info1, 1441 const void *dev_info2) 1442 { 1443 if (((struct btrfs_device_info *)dev_info1)->max_avail > 1444 ((struct btrfs_device_info *)dev_info2)->max_avail) 1445 return -1; 1446 else if (((struct btrfs_device_info *)dev_info1)->max_avail < 1447 ((struct btrfs_device_info *)dev_info2)->max_avail) 1448 return 1; 1449 else 1450 return 0; 1451 } 1452 1453 /* 1454 * sort the devices by max_avail, in which max free extent size of each device 1455 * is stored.(Descending Sort) 1456 */ 1457 static inline void btrfs_descending_sort_devices( 1458 struct btrfs_device_info *devices, 1459 size_t nr_devices) 1460 { 1461 sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1462 btrfs_cmp_device_free_bytes, NULL); 1463 } 1464 1465 /* 1466 * The helper to calc the free space on the devices that can be used to store 1467 * file data. 1468 */ 1469 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) 1470 { 1471 struct btrfs_fs_info *fs_info = root->fs_info; 1472 struct btrfs_device_info *devices_info; 1473 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 1474 struct btrfs_device *device; 1475 u64 skip_space; 1476 u64 type; 1477 u64 avail_space; 1478 u64 used_space; 1479 u64 min_stripe_size; 1480 int min_stripes = 1, num_stripes = 1; 1481 int i = 0, nr_devices; 1482 int ret; 1483 1484 nr_devices = fs_info->fs_devices->open_devices; 1485 BUG_ON(!nr_devices); 1486 1487 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), 1488 GFP_NOFS); 1489 if (!devices_info) 1490 return -ENOMEM; 1491 1492 /* calc min stripe number for data space alloction */ 1493 type = btrfs_get_alloc_profile(root, 1); 1494 if (type & BTRFS_BLOCK_GROUP_RAID0) { 1495 min_stripes = 2; 1496 num_stripes = nr_devices; 1497 } else if (type & BTRFS_BLOCK_GROUP_RAID1) { 1498 min_stripes = 2; 1499 num_stripes = 2; 1500 } else if (type & BTRFS_BLOCK_GROUP_RAID10) { 1501 min_stripes = 4; 1502 num_stripes = 4; 1503 } 1504 1505 if (type & BTRFS_BLOCK_GROUP_DUP) 1506 min_stripe_size = 2 * BTRFS_STRIPE_LEN; 1507 else 1508 min_stripe_size = BTRFS_STRIPE_LEN; 1509 1510 list_for_each_entry(device, &fs_devices->devices, dev_list) { 1511 if (!device->in_fs_metadata || !device->bdev || 1512 device->is_tgtdev_for_dev_replace) 1513 continue; 1514 1515 avail_space = device->total_bytes - device->bytes_used; 1516 1517 /* align with stripe_len */ 1518 do_div(avail_space, BTRFS_STRIPE_LEN); 1519 avail_space *= BTRFS_STRIPE_LEN; 1520 1521 /* 1522 * In order to avoid overwritting the superblock on the drive, 1523 * btrfs starts at an offset of at least 1MB when doing chunk 1524 * allocation. 1525 */ 1526 skip_space = 1024 * 1024; 1527 1528 /* user can set the offset in fs_info->alloc_start. */ 1529 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= 1530 device->total_bytes) 1531 skip_space = max(fs_info->alloc_start, skip_space); 1532 1533 /* 1534 * btrfs can not use the free space in [0, skip_space - 1], 1535 * we must subtract it from the total. In order to implement 1536 * it, we account the used space in this range first. 1537 */ 1538 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, 1539 &used_space); 1540 if (ret) { 1541 kfree(devices_info); 1542 return ret; 1543 } 1544 1545 /* calc the free space in [0, skip_space - 1] */ 1546 skip_space -= used_space; 1547 1548 /* 1549 * we can use the free space in [0, skip_space - 1], subtract 1550 * it from the total. 1551 */ 1552 if (avail_space && avail_space >= skip_space) 1553 avail_space -= skip_space; 1554 else 1555 avail_space = 0; 1556 1557 if (avail_space < min_stripe_size) 1558 continue; 1559 1560 devices_info[i].dev = device; 1561 devices_info[i].max_avail = avail_space; 1562 1563 i++; 1564 } 1565 1566 nr_devices = i; 1567 1568 btrfs_descending_sort_devices(devices_info, nr_devices); 1569 1570 i = nr_devices - 1; 1571 avail_space = 0; 1572 while (nr_devices >= min_stripes) { 1573 if (num_stripes > nr_devices) 1574 num_stripes = nr_devices; 1575 1576 if (devices_info[i].max_avail >= min_stripe_size) { 1577 int j; 1578 u64 alloc_size; 1579 1580 avail_space += devices_info[i].max_avail * num_stripes; 1581 alloc_size = devices_info[i].max_avail; 1582 for (j = i + 1 - num_stripes; j <= i; j++) 1583 devices_info[j].max_avail -= alloc_size; 1584 } 1585 i--; 1586 nr_devices--; 1587 } 1588 1589 kfree(devices_info); 1590 *free_bytes = avail_space; 1591 return 0; 1592 } 1593 1594 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 1595 { 1596 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1597 struct btrfs_super_block *disk_super = fs_info->super_copy; 1598 struct list_head *head = &fs_info->space_info; 1599 struct btrfs_space_info *found; 1600 u64 total_used = 0; 1601 u64 total_free_data = 0; 1602 int bits = dentry->d_sb->s_blocksize_bits; 1603 __be32 *fsid = (__be32 *)fs_info->fsid; 1604 int ret; 1605 1606 /* holding chunk_muext to avoid allocating new chunks */ 1607 mutex_lock(&fs_info->chunk_mutex); 1608 rcu_read_lock(); 1609 list_for_each_entry_rcu(found, head, list) { 1610 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 1611 total_free_data += found->disk_total - found->disk_used; 1612 total_free_data -= 1613 btrfs_account_ro_block_groups_free_space(found); 1614 } 1615 1616 total_used += found->disk_used; 1617 } 1618 rcu_read_unlock(); 1619 1620 buf->f_namelen = BTRFS_NAME_LEN; 1621 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; 1622 buf->f_bfree = buf->f_blocks - (total_used >> bits); 1623 buf->f_bsize = dentry->d_sb->s_blocksize; 1624 buf->f_type = BTRFS_SUPER_MAGIC; 1625 buf->f_bavail = total_free_data; 1626 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data); 1627 if (ret) { 1628 mutex_unlock(&fs_info->chunk_mutex); 1629 return ret; 1630 } 1631 buf->f_bavail += total_free_data; 1632 buf->f_bavail = buf->f_bavail >> bits; 1633 mutex_unlock(&fs_info->chunk_mutex); 1634 1635 /* We treat it as constant endianness (it doesn't matter _which_) 1636 because we want the fsid to come out the same whether mounted 1637 on a big-endian or little-endian host */ 1638 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 1639 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 1640 /* Mask in the root object ID too, to disambiguate subvols */ 1641 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; 1642 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; 1643 1644 return 0; 1645 } 1646 1647 static void btrfs_kill_super(struct super_block *sb) 1648 { 1649 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1650 kill_anon_super(sb); 1651 free_fs_info(fs_info); 1652 } 1653 1654 static struct file_system_type btrfs_fs_type = { 1655 .owner = THIS_MODULE, 1656 .name = "btrfs", 1657 .mount = btrfs_mount, 1658 .kill_sb = btrfs_kill_super, 1659 .fs_flags = FS_REQUIRES_DEV, 1660 }; 1661 MODULE_ALIAS_FS("btrfs"); 1662 1663 /* 1664 * used by btrfsctl to scan devices when no FS is mounted 1665 */ 1666 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 1667 unsigned long arg) 1668 { 1669 struct btrfs_ioctl_vol_args *vol; 1670 struct btrfs_fs_devices *fs_devices; 1671 int ret = -ENOTTY; 1672 1673 if (!capable(CAP_SYS_ADMIN)) 1674 return -EPERM; 1675 1676 vol = memdup_user((void __user *)arg, sizeof(*vol)); 1677 if (IS_ERR(vol)) 1678 return PTR_ERR(vol); 1679 1680 switch (cmd) { 1681 case BTRFS_IOC_SCAN_DEV: 1682 ret = btrfs_scan_one_device(vol->name, FMODE_READ, 1683 &btrfs_fs_type, &fs_devices); 1684 break; 1685 case BTRFS_IOC_DEVICES_READY: 1686 ret = btrfs_scan_one_device(vol->name, FMODE_READ, 1687 &btrfs_fs_type, &fs_devices); 1688 if (ret) 1689 break; 1690 ret = !(fs_devices->num_devices == fs_devices->total_devices); 1691 break; 1692 } 1693 1694 kfree(vol); 1695 return ret; 1696 } 1697 1698 static int btrfs_freeze(struct super_block *sb) 1699 { 1700 struct btrfs_trans_handle *trans; 1701 struct btrfs_root *root = btrfs_sb(sb)->tree_root; 1702 1703 trans = btrfs_attach_transaction_barrier(root); 1704 if (IS_ERR(trans)) { 1705 /* no transaction, don't bother */ 1706 if (PTR_ERR(trans) == -ENOENT) 1707 return 0; 1708 return PTR_ERR(trans); 1709 } 1710 return btrfs_commit_transaction(trans, root); 1711 } 1712 1713 static int btrfs_unfreeze(struct super_block *sb) 1714 { 1715 return 0; 1716 } 1717 1718 static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 1719 { 1720 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 1721 struct btrfs_fs_devices *cur_devices; 1722 struct btrfs_device *dev, *first_dev = NULL; 1723 struct list_head *head; 1724 struct rcu_string *name; 1725 1726 mutex_lock(&fs_info->fs_devices->device_list_mutex); 1727 cur_devices = fs_info->fs_devices; 1728 while (cur_devices) { 1729 head = &cur_devices->devices; 1730 list_for_each_entry(dev, head, dev_list) { 1731 if (dev->missing) 1732 continue; 1733 if (!first_dev || dev->devid < first_dev->devid) 1734 first_dev = dev; 1735 } 1736 cur_devices = cur_devices->seed; 1737 } 1738 1739 if (first_dev) { 1740 rcu_read_lock(); 1741 name = rcu_dereference(first_dev->name); 1742 seq_escape(m, name->str, " \t\n\\"); 1743 rcu_read_unlock(); 1744 } else { 1745 WARN_ON(1); 1746 } 1747 mutex_unlock(&fs_info->fs_devices->device_list_mutex); 1748 return 0; 1749 } 1750 1751 static const struct super_operations btrfs_super_ops = { 1752 .drop_inode = btrfs_drop_inode, 1753 .evict_inode = btrfs_evict_inode, 1754 .put_super = btrfs_put_super, 1755 .sync_fs = btrfs_sync_fs, 1756 .show_options = btrfs_show_options, 1757 .show_devname = btrfs_show_devname, 1758 .write_inode = btrfs_write_inode, 1759 .alloc_inode = btrfs_alloc_inode, 1760 .destroy_inode = btrfs_destroy_inode, 1761 .statfs = btrfs_statfs, 1762 .remount_fs = btrfs_remount, 1763 .freeze_fs = btrfs_freeze, 1764 .unfreeze_fs = btrfs_unfreeze, 1765 }; 1766 1767 static const struct file_operations btrfs_ctl_fops = { 1768 .unlocked_ioctl = btrfs_control_ioctl, 1769 .compat_ioctl = btrfs_control_ioctl, 1770 .owner = THIS_MODULE, 1771 .llseek = noop_llseek, 1772 }; 1773 1774 static struct miscdevice btrfs_misc = { 1775 .minor = BTRFS_MINOR, 1776 .name = "btrfs-control", 1777 .fops = &btrfs_ctl_fops 1778 }; 1779 1780 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 1781 MODULE_ALIAS("devname:btrfs-control"); 1782 1783 static int btrfs_interface_init(void) 1784 { 1785 return misc_register(&btrfs_misc); 1786 } 1787 1788 static void btrfs_interface_exit(void) 1789 { 1790 if (misc_deregister(&btrfs_misc) < 0) 1791 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n"); 1792 } 1793 1794 static void btrfs_print_info(void) 1795 { 1796 printk(KERN_INFO "Btrfs loaded" 1797 #ifdef CONFIG_BTRFS_DEBUG 1798 ", debug=on" 1799 #endif 1800 #ifdef CONFIG_BTRFS_ASSERT 1801 ", assert=on" 1802 #endif 1803 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 1804 ", integrity-checker=on" 1805 #endif 1806 "\n"); 1807 } 1808 1809 static int btrfs_run_sanity_tests(void) 1810 { 1811 int ret; 1812 1813 ret = btrfs_init_test_fs(); 1814 if (ret) 1815 return ret; 1816 1817 ret = btrfs_test_free_space_cache(); 1818 if (ret) 1819 goto out; 1820 ret = btrfs_test_extent_buffer_operations(); 1821 if (ret) 1822 goto out; 1823 ret = btrfs_test_extent_io(); 1824 if (ret) 1825 goto out; 1826 ret = btrfs_test_inodes(); 1827 out: 1828 btrfs_destroy_test_fs(); 1829 return ret; 1830 } 1831 1832 static int __init init_btrfs_fs(void) 1833 { 1834 int err; 1835 1836 err = btrfs_init_sysfs(); 1837 if (err) 1838 return err; 1839 1840 btrfs_init_compress(); 1841 1842 err = btrfs_init_cachep(); 1843 if (err) 1844 goto free_compress; 1845 1846 err = extent_io_init(); 1847 if (err) 1848 goto free_cachep; 1849 1850 err = extent_map_init(); 1851 if (err) 1852 goto free_extent_io; 1853 1854 err = ordered_data_init(); 1855 if (err) 1856 goto free_extent_map; 1857 1858 err = btrfs_delayed_inode_init(); 1859 if (err) 1860 goto free_ordered_data; 1861 1862 err = btrfs_auto_defrag_init(); 1863 if (err) 1864 goto free_delayed_inode; 1865 1866 err = btrfs_delayed_ref_init(); 1867 if (err) 1868 goto free_auto_defrag; 1869 1870 err = btrfs_prelim_ref_init(); 1871 if (err) 1872 goto free_prelim_ref; 1873 1874 err = btrfs_interface_init(); 1875 if (err) 1876 goto free_delayed_ref; 1877 1878 btrfs_init_lockdep(); 1879 1880 btrfs_print_info(); 1881 1882 err = btrfs_run_sanity_tests(); 1883 if (err) 1884 goto unregister_ioctl; 1885 1886 err = register_filesystem(&btrfs_fs_type); 1887 if (err) 1888 goto unregister_ioctl; 1889 1890 return 0; 1891 1892 unregister_ioctl: 1893 btrfs_interface_exit(); 1894 free_prelim_ref: 1895 btrfs_prelim_ref_exit(); 1896 free_delayed_ref: 1897 btrfs_delayed_ref_exit(); 1898 free_auto_defrag: 1899 btrfs_auto_defrag_exit(); 1900 free_delayed_inode: 1901 btrfs_delayed_inode_exit(); 1902 free_ordered_data: 1903 ordered_data_exit(); 1904 free_extent_map: 1905 extent_map_exit(); 1906 free_extent_io: 1907 extent_io_exit(); 1908 free_cachep: 1909 btrfs_destroy_cachep(); 1910 free_compress: 1911 btrfs_exit_compress(); 1912 btrfs_exit_sysfs(); 1913 return err; 1914 } 1915 1916 static void __exit exit_btrfs_fs(void) 1917 { 1918 btrfs_destroy_cachep(); 1919 btrfs_delayed_ref_exit(); 1920 btrfs_auto_defrag_exit(); 1921 btrfs_delayed_inode_exit(); 1922 btrfs_prelim_ref_exit(); 1923 ordered_data_exit(); 1924 extent_map_exit(); 1925 extent_io_exit(); 1926 btrfs_interface_exit(); 1927 unregister_filesystem(&btrfs_fs_type); 1928 btrfs_exit_sysfs(); 1929 btrfs_cleanup_fs_uuids(); 1930 btrfs_exit_compress(); 1931 } 1932 1933 module_init(init_btrfs_fs) 1934 module_exit(exit_btrfs_fs) 1935 1936 MODULE_LICENSE("GPL"); 1937