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, Opt_noflushoncommit, 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_noflushoncommit, "noflushoncommit"}, 354 {Opt_ratio, "metadata_ratio=%d"}, 355 {Opt_discard, "discard"}, 356 {Opt_nodiscard, "nodiscard"}, 357 {Opt_space_cache, "space_cache"}, 358 {Opt_clear_cache, "clear_cache"}, 359 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 360 {Opt_enospc_debug, "enospc_debug"}, 361 {Opt_noenospc_debug, "noenospc_debug"}, 362 {Opt_subvolrootid, "subvolrootid=%d"}, 363 {Opt_defrag, "autodefrag"}, 364 {Opt_nodefrag, "noautodefrag"}, 365 {Opt_inode_cache, "inode_cache"}, 366 {Opt_no_space_cache, "nospace_cache"}, 367 {Opt_recovery, "recovery"}, 368 {Opt_skip_balance, "skip_balance"}, 369 {Opt_check_integrity, "check_int"}, 370 {Opt_check_integrity_including_extent_data, "check_int_data"}, 371 {Opt_check_integrity_print_mask, "check_int_print_mask=%d"}, 372 {Opt_rescan_uuid_tree, "rescan_uuid_tree"}, 373 {Opt_fatal_errors, "fatal_errors=%s"}, 374 {Opt_commit_interval, "commit=%d"}, 375 {Opt_err, NULL}, 376 }; 377 378 /* 379 * Regular mount options parser. Everything that is needed only when 380 * reading in a new superblock is parsed here. 381 * XXX JDM: This needs to be cleaned up for remount. 382 */ 383 int btrfs_parse_options(struct btrfs_root *root, char *options) 384 { 385 struct btrfs_fs_info *info = root->fs_info; 386 substring_t args[MAX_OPT_ARGS]; 387 char *p, *num, *orig = NULL; 388 u64 cache_gen; 389 int intarg; 390 int ret = 0; 391 char *compress_type; 392 bool compress_force = false; 393 394 cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy); 395 if (cache_gen) 396 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 397 398 if (!options) 399 goto out; 400 401 /* 402 * strsep changes the string, duplicate it because parse_options 403 * gets called twice 404 */ 405 options = kstrdup(options, GFP_NOFS); 406 if (!options) 407 return -ENOMEM; 408 409 orig = options; 410 411 while ((p = strsep(&options, ",")) != NULL) { 412 int token; 413 if (!*p) 414 continue; 415 416 token = match_token(p, tokens, args); 417 switch (token) { 418 case Opt_degraded: 419 btrfs_info(root->fs_info, "allowing degraded mounts"); 420 btrfs_set_opt(info->mount_opt, DEGRADED); 421 break; 422 case Opt_subvol: 423 case Opt_subvolid: 424 case Opt_subvolrootid: 425 case Opt_device: 426 /* 427 * These are parsed by btrfs_parse_early_options 428 * and can be happily ignored here. 429 */ 430 break; 431 case Opt_nodatasum: 432 btrfs_info(root->fs_info, "setting nodatasum"); 433 btrfs_set_opt(info->mount_opt, NODATASUM); 434 break; 435 case Opt_nodatacow: 436 if (!btrfs_test_opt(root, COMPRESS) || 437 !btrfs_test_opt(root, FORCE_COMPRESS)) { 438 btrfs_info(root->fs_info, 439 "setting nodatacow, compression disabled"); 440 } else { 441 btrfs_info(root->fs_info, "setting nodatacow"); 442 } 443 btrfs_clear_opt(info->mount_opt, COMPRESS); 444 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 445 btrfs_set_opt(info->mount_opt, NODATACOW); 446 btrfs_set_opt(info->mount_opt, NODATASUM); 447 break; 448 case Opt_compress_force: 449 case Opt_compress_force_type: 450 compress_force = true; 451 /* Fallthrough */ 452 case Opt_compress: 453 case Opt_compress_type: 454 if (token == Opt_compress || 455 token == Opt_compress_force || 456 strcmp(args[0].from, "zlib") == 0) { 457 compress_type = "zlib"; 458 info->compress_type = BTRFS_COMPRESS_ZLIB; 459 btrfs_set_opt(info->mount_opt, COMPRESS); 460 btrfs_clear_opt(info->mount_opt, NODATACOW); 461 btrfs_clear_opt(info->mount_opt, NODATASUM); 462 } else if (strcmp(args[0].from, "lzo") == 0) { 463 compress_type = "lzo"; 464 info->compress_type = BTRFS_COMPRESS_LZO; 465 btrfs_set_opt(info->mount_opt, COMPRESS); 466 btrfs_clear_opt(info->mount_opt, NODATACOW); 467 btrfs_clear_opt(info->mount_opt, NODATASUM); 468 btrfs_set_fs_incompat(info, COMPRESS_LZO); 469 } else if (strncmp(args[0].from, "no", 2) == 0) { 470 compress_type = "no"; 471 btrfs_clear_opt(info->mount_opt, COMPRESS); 472 btrfs_clear_opt(info->mount_opt, FORCE_COMPRESS); 473 compress_force = false; 474 } else { 475 ret = -EINVAL; 476 goto out; 477 } 478 479 if (compress_force) { 480 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 481 btrfs_info(root->fs_info, "force %s compression", 482 compress_type); 483 } else if (btrfs_test_opt(root, COMPRESS)) { 484 pr_info("btrfs: use %s compression\n", 485 compress_type); 486 } 487 break; 488 case Opt_ssd: 489 btrfs_info(root->fs_info, "use ssd allocation scheme"); 490 btrfs_set_opt(info->mount_opt, SSD); 491 break; 492 case Opt_ssd_spread: 493 btrfs_info(root->fs_info, "use spread ssd allocation scheme"); 494 btrfs_set_opt(info->mount_opt, SSD); 495 btrfs_set_opt(info->mount_opt, SSD_SPREAD); 496 break; 497 case Opt_nossd: 498 btrfs_info(root->fs_info, "not using ssd allocation scheme"); 499 btrfs_set_opt(info->mount_opt, NOSSD); 500 btrfs_clear_opt(info->mount_opt, SSD); 501 btrfs_clear_opt(info->mount_opt, SSD_SPREAD); 502 break; 503 case Opt_barrier: 504 if (btrfs_test_opt(root, NOBARRIER)) 505 btrfs_info(root->fs_info, "turning on barriers"); 506 btrfs_clear_opt(info->mount_opt, NOBARRIER); 507 break; 508 case Opt_nobarrier: 509 btrfs_info(root->fs_info, "turning off barriers"); 510 btrfs_set_opt(info->mount_opt, NOBARRIER); 511 break; 512 case Opt_thread_pool: 513 ret = match_int(&args[0], &intarg); 514 if (ret) { 515 goto out; 516 } else if (intarg > 0) { 517 info->thread_pool_size = intarg; 518 } else { 519 ret = -EINVAL; 520 goto out; 521 } 522 break; 523 case Opt_max_inline: 524 num = match_strdup(&args[0]); 525 if (num) { 526 info->max_inline = memparse(num, NULL); 527 kfree(num); 528 529 if (info->max_inline) { 530 info->max_inline = max_t(u64, 531 info->max_inline, 532 root->sectorsize); 533 } 534 btrfs_info(root->fs_info, "max_inline at %llu", 535 info->max_inline); 536 } else { 537 ret = -ENOMEM; 538 goto out; 539 } 540 break; 541 case Opt_alloc_start: 542 num = match_strdup(&args[0]); 543 if (num) { 544 mutex_lock(&info->chunk_mutex); 545 info->alloc_start = memparse(num, NULL); 546 mutex_unlock(&info->chunk_mutex); 547 kfree(num); 548 btrfs_info(root->fs_info, "allocations start at %llu", 549 info->alloc_start); 550 } else { 551 ret = -ENOMEM; 552 goto out; 553 } 554 break; 555 case Opt_noacl: 556 root->fs_info->sb->s_flags &= ~MS_POSIXACL; 557 break; 558 case Opt_notreelog: 559 btrfs_info(root->fs_info, "disabling tree log"); 560 btrfs_set_opt(info->mount_opt, NOTREELOG); 561 break; 562 case Opt_flushoncommit: 563 btrfs_info(root->fs_info, "turning on flush-on-commit"); 564 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); 565 break; 566 case Opt_noflushoncommit: 567 if (btrfs_test_opt(root, FLUSHONCOMMIT)) 568 btrfs_info(root->fs_info, "turning off flush-on-commit"); 569 btrfs_clear_opt(info->mount_opt, FLUSHONCOMMIT); 570 break; 571 case Opt_ratio: 572 ret = match_int(&args[0], &intarg); 573 if (ret) { 574 goto out; 575 } else if (intarg >= 0) { 576 info->metadata_ratio = intarg; 577 btrfs_info(root->fs_info, "metadata ratio %d", 578 info->metadata_ratio); 579 } else { 580 ret = -EINVAL; 581 goto out; 582 } 583 break; 584 case Opt_discard: 585 btrfs_set_opt(info->mount_opt, DISCARD); 586 break; 587 case Opt_nodiscard: 588 btrfs_clear_opt(info->mount_opt, DISCARD); 589 break; 590 case Opt_space_cache: 591 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 592 break; 593 case Opt_rescan_uuid_tree: 594 btrfs_set_opt(info->mount_opt, RESCAN_UUID_TREE); 595 break; 596 case Opt_no_space_cache: 597 btrfs_info(root->fs_info, "disabling disk space caching"); 598 btrfs_clear_opt(info->mount_opt, SPACE_CACHE); 599 break; 600 case Opt_inode_cache: 601 btrfs_info(root->fs_info, "enabling inode map caching"); 602 btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE); 603 break; 604 case Opt_clear_cache: 605 btrfs_info(root->fs_info, "force clearing of disk cache"); 606 btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 607 break; 608 case Opt_user_subvol_rm_allowed: 609 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 610 break; 611 case Opt_enospc_debug: 612 btrfs_set_opt(info->mount_opt, ENOSPC_DEBUG); 613 break; 614 case Opt_noenospc_debug: 615 btrfs_clear_opt(info->mount_opt, ENOSPC_DEBUG); 616 break; 617 case Opt_defrag: 618 btrfs_info(root->fs_info, "enabling auto defrag"); 619 btrfs_set_opt(info->mount_opt, AUTO_DEFRAG); 620 break; 621 case Opt_nodefrag: 622 if (btrfs_test_opt(root, AUTO_DEFRAG)) 623 btrfs_info(root->fs_info, "disabling auto defrag"); 624 btrfs_clear_opt(info->mount_opt, AUTO_DEFRAG); 625 break; 626 case Opt_recovery: 627 btrfs_info(root->fs_info, "enabling auto recovery"); 628 btrfs_set_opt(info->mount_opt, RECOVERY); 629 break; 630 case Opt_skip_balance: 631 btrfs_set_opt(info->mount_opt, SKIP_BALANCE); 632 break; 633 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 634 case Opt_check_integrity_including_extent_data: 635 btrfs_info(root->fs_info, 636 "enabling check integrity including extent data"); 637 btrfs_set_opt(info->mount_opt, 638 CHECK_INTEGRITY_INCLUDING_EXTENT_DATA); 639 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 640 break; 641 case Opt_check_integrity: 642 btrfs_info(root->fs_info, "enabling check integrity"); 643 btrfs_set_opt(info->mount_opt, CHECK_INTEGRITY); 644 break; 645 case Opt_check_integrity_print_mask: 646 ret = match_int(&args[0], &intarg); 647 if (ret) { 648 goto out; 649 } else if (intarg >= 0) { 650 info->check_integrity_print_mask = intarg; 651 btrfs_info(root->fs_info, "check_integrity_print_mask 0x%x", 652 info->check_integrity_print_mask); 653 } else { 654 ret = -EINVAL; 655 goto out; 656 } 657 break; 658 #else 659 case Opt_check_integrity_including_extent_data: 660 case Opt_check_integrity: 661 case Opt_check_integrity_print_mask: 662 btrfs_err(root->fs_info, 663 "support for check_integrity* not compiled in!"); 664 ret = -EINVAL; 665 goto out; 666 #endif 667 case Opt_fatal_errors: 668 if (strcmp(args[0].from, "panic") == 0) 669 btrfs_set_opt(info->mount_opt, 670 PANIC_ON_FATAL_ERROR); 671 else if (strcmp(args[0].from, "bug") == 0) 672 btrfs_clear_opt(info->mount_opt, 673 PANIC_ON_FATAL_ERROR); 674 else { 675 ret = -EINVAL; 676 goto out; 677 } 678 break; 679 case Opt_commit_interval: 680 intarg = 0; 681 ret = match_int(&args[0], &intarg); 682 if (ret < 0) { 683 btrfs_err(root->fs_info, "invalid commit interval"); 684 ret = -EINVAL; 685 goto out; 686 } 687 if (intarg > 0) { 688 if (intarg > 300) { 689 btrfs_warn(root->fs_info, "excessive commit interval %d", 690 intarg); 691 } 692 info->commit_interval = intarg; 693 } else { 694 btrfs_info(root->fs_info, "using default commit interval %ds", 695 BTRFS_DEFAULT_COMMIT_INTERVAL); 696 info->commit_interval = BTRFS_DEFAULT_COMMIT_INTERVAL; 697 } 698 break; 699 case Opt_err: 700 btrfs_info(root->fs_info, "unrecognized mount option '%s'", p); 701 ret = -EINVAL; 702 goto out; 703 default: 704 break; 705 } 706 } 707 out: 708 if (!ret && btrfs_test_opt(root, SPACE_CACHE)) 709 btrfs_info(root->fs_info, "disk space caching is enabled"); 710 kfree(orig); 711 return ret; 712 } 713 714 /* 715 * Parse mount options that are required early in the mount process. 716 * 717 * All other options will be parsed on much later in the mount process and 718 * only when we need to allocate a new super block. 719 */ 720 static int btrfs_parse_early_options(const char *options, fmode_t flags, 721 void *holder, char **subvol_name, u64 *subvol_objectid, 722 struct btrfs_fs_devices **fs_devices) 723 { 724 substring_t args[MAX_OPT_ARGS]; 725 char *device_name, *opts, *orig, *p; 726 char *num = NULL; 727 int error = 0; 728 729 if (!options) 730 return 0; 731 732 /* 733 * strsep changes the string, duplicate it because parse_options 734 * gets called twice 735 */ 736 opts = kstrdup(options, GFP_KERNEL); 737 if (!opts) 738 return -ENOMEM; 739 orig = opts; 740 741 while ((p = strsep(&opts, ",")) != NULL) { 742 int token; 743 if (!*p) 744 continue; 745 746 token = match_token(p, tokens, args); 747 switch (token) { 748 case Opt_subvol: 749 kfree(*subvol_name); 750 *subvol_name = match_strdup(&args[0]); 751 if (!*subvol_name) { 752 error = -ENOMEM; 753 goto out; 754 } 755 break; 756 case Opt_subvolid: 757 num = match_strdup(&args[0]); 758 if (num) { 759 *subvol_objectid = memparse(num, NULL); 760 kfree(num); 761 /* we want the original fs_tree */ 762 if (!*subvol_objectid) 763 *subvol_objectid = 764 BTRFS_FS_TREE_OBJECTID; 765 } else { 766 error = -EINVAL; 767 goto out; 768 } 769 break; 770 case Opt_subvolrootid: 771 printk(KERN_WARNING 772 "BTRFS: 'subvolrootid' mount option is deprecated and has " 773 "no effect\n"); 774 break; 775 case Opt_device: 776 device_name = match_strdup(&args[0]); 777 if (!device_name) { 778 error = -ENOMEM; 779 goto out; 780 } 781 error = btrfs_scan_one_device(device_name, 782 flags, holder, fs_devices); 783 kfree(device_name); 784 if (error) 785 goto out; 786 break; 787 default: 788 break; 789 } 790 } 791 792 out: 793 kfree(orig); 794 return error; 795 } 796 797 static struct dentry *get_default_root(struct super_block *sb, 798 u64 subvol_objectid) 799 { 800 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 801 struct btrfs_root *root = fs_info->tree_root; 802 struct btrfs_root *new_root; 803 struct btrfs_dir_item *di; 804 struct btrfs_path *path; 805 struct btrfs_key location; 806 struct inode *inode; 807 u64 dir_id; 808 int new = 0; 809 810 /* 811 * We have a specific subvol we want to mount, just setup location and 812 * go look up the root. 813 */ 814 if (subvol_objectid) { 815 location.objectid = subvol_objectid; 816 location.type = BTRFS_ROOT_ITEM_KEY; 817 location.offset = (u64)-1; 818 goto find_root; 819 } 820 821 path = btrfs_alloc_path(); 822 if (!path) 823 return ERR_PTR(-ENOMEM); 824 path->leave_spinning = 1; 825 826 /* 827 * Find the "default" dir item which points to the root item that we 828 * will mount by default if we haven't been given a specific subvolume 829 * to mount. 830 */ 831 dir_id = btrfs_super_root_dir(fs_info->super_copy); 832 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); 833 if (IS_ERR(di)) { 834 btrfs_free_path(path); 835 return ERR_CAST(di); 836 } 837 if (!di) { 838 /* 839 * Ok the default dir item isn't there. This is weird since 840 * it's always been there, but don't freak out, just try and 841 * mount to root most subvolume. 842 */ 843 btrfs_free_path(path); 844 dir_id = BTRFS_FIRST_FREE_OBJECTID; 845 new_root = fs_info->fs_root; 846 goto setup_root; 847 } 848 849 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 850 btrfs_free_path(path); 851 852 find_root: 853 new_root = btrfs_read_fs_root_no_name(fs_info, &location); 854 if (IS_ERR(new_root)) 855 return ERR_CAST(new_root); 856 857 dir_id = btrfs_root_dirid(&new_root->root_item); 858 setup_root: 859 location.objectid = dir_id; 860 location.type = BTRFS_INODE_ITEM_KEY; 861 location.offset = 0; 862 863 inode = btrfs_iget(sb, &location, new_root, &new); 864 if (IS_ERR(inode)) 865 return ERR_CAST(inode); 866 867 /* 868 * If we're just mounting the root most subvol put the inode and return 869 * a reference to the dentry. We will have already gotten a reference 870 * to the inode in btrfs_fill_super so we're good to go. 871 */ 872 if (!new && sb->s_root->d_inode == inode) { 873 iput(inode); 874 return dget(sb->s_root); 875 } 876 877 return d_obtain_alias(inode); 878 } 879 880 static int btrfs_fill_super(struct super_block *sb, 881 struct btrfs_fs_devices *fs_devices, 882 void *data, int silent) 883 { 884 struct inode *inode; 885 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 886 struct btrfs_key key; 887 int err; 888 889 sb->s_maxbytes = MAX_LFS_FILESIZE; 890 sb->s_magic = BTRFS_SUPER_MAGIC; 891 sb->s_op = &btrfs_super_ops; 892 sb->s_d_op = &btrfs_dentry_operations; 893 sb->s_export_op = &btrfs_export_ops; 894 sb->s_xattr = btrfs_xattr_handlers; 895 sb->s_time_gran = 1; 896 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 897 sb->s_flags |= MS_POSIXACL; 898 #endif 899 sb->s_flags |= MS_I_VERSION; 900 err = open_ctree(sb, fs_devices, (char *)data); 901 if (err) { 902 printk(KERN_ERR "BTRFS: open_ctree failed\n"); 903 return err; 904 } 905 906 key.objectid = BTRFS_FIRST_FREE_OBJECTID; 907 key.type = BTRFS_INODE_ITEM_KEY; 908 key.offset = 0; 909 inode = btrfs_iget(sb, &key, fs_info->fs_root, NULL); 910 if (IS_ERR(inode)) { 911 err = PTR_ERR(inode); 912 goto fail_close; 913 } 914 915 sb->s_root = d_make_root(inode); 916 if (!sb->s_root) { 917 err = -ENOMEM; 918 goto fail_close; 919 } 920 921 save_mount_options(sb, data); 922 cleancache_init_fs(sb); 923 sb->s_flags |= MS_ACTIVE; 924 return 0; 925 926 fail_close: 927 close_ctree(fs_info->tree_root); 928 return err; 929 } 930 931 int btrfs_sync_fs(struct super_block *sb, int wait) 932 { 933 struct btrfs_trans_handle *trans; 934 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 935 struct btrfs_root *root = fs_info->tree_root; 936 937 trace_btrfs_sync_fs(wait); 938 939 if (!wait) { 940 filemap_flush(fs_info->btree_inode->i_mapping); 941 return 0; 942 } 943 944 btrfs_wait_ordered_roots(fs_info, -1); 945 946 trans = btrfs_attach_transaction_barrier(root); 947 if (IS_ERR(trans)) { 948 /* no transaction, don't bother */ 949 if (PTR_ERR(trans) == -ENOENT) 950 return 0; 951 return PTR_ERR(trans); 952 } 953 return btrfs_commit_transaction(trans, root); 954 } 955 956 static int btrfs_show_options(struct seq_file *seq, struct dentry *dentry) 957 { 958 struct btrfs_fs_info *info = btrfs_sb(dentry->d_sb); 959 struct btrfs_root *root = info->tree_root; 960 char *compress_type; 961 962 if (btrfs_test_opt(root, DEGRADED)) 963 seq_puts(seq, ",degraded"); 964 if (btrfs_test_opt(root, NODATASUM)) 965 seq_puts(seq, ",nodatasum"); 966 if (btrfs_test_opt(root, NODATACOW)) 967 seq_puts(seq, ",nodatacow"); 968 if (btrfs_test_opt(root, NOBARRIER)) 969 seq_puts(seq, ",nobarrier"); 970 if (info->max_inline != 8192 * 1024) 971 seq_printf(seq, ",max_inline=%llu", info->max_inline); 972 if (info->alloc_start != 0) 973 seq_printf(seq, ",alloc_start=%llu", info->alloc_start); 974 if (info->thread_pool_size != min_t(unsigned long, 975 num_online_cpus() + 2, 8)) 976 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); 977 if (btrfs_test_opt(root, COMPRESS)) { 978 if (info->compress_type == BTRFS_COMPRESS_ZLIB) 979 compress_type = "zlib"; 980 else 981 compress_type = "lzo"; 982 if (btrfs_test_opt(root, FORCE_COMPRESS)) 983 seq_printf(seq, ",compress-force=%s", compress_type); 984 else 985 seq_printf(seq, ",compress=%s", compress_type); 986 } 987 if (btrfs_test_opt(root, NOSSD)) 988 seq_puts(seq, ",nossd"); 989 if (btrfs_test_opt(root, SSD_SPREAD)) 990 seq_puts(seq, ",ssd_spread"); 991 else if (btrfs_test_opt(root, SSD)) 992 seq_puts(seq, ",ssd"); 993 if (btrfs_test_opt(root, NOTREELOG)) 994 seq_puts(seq, ",notreelog"); 995 if (btrfs_test_opt(root, FLUSHONCOMMIT)) 996 seq_puts(seq, ",flushoncommit"); 997 if (btrfs_test_opt(root, DISCARD)) 998 seq_puts(seq, ",discard"); 999 if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) 1000 seq_puts(seq, ",noacl"); 1001 if (btrfs_test_opt(root, SPACE_CACHE)) 1002 seq_puts(seq, ",space_cache"); 1003 else 1004 seq_puts(seq, ",nospace_cache"); 1005 if (btrfs_test_opt(root, RESCAN_UUID_TREE)) 1006 seq_puts(seq, ",rescan_uuid_tree"); 1007 if (btrfs_test_opt(root, CLEAR_CACHE)) 1008 seq_puts(seq, ",clear_cache"); 1009 if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED)) 1010 seq_puts(seq, ",user_subvol_rm_allowed"); 1011 if (btrfs_test_opt(root, ENOSPC_DEBUG)) 1012 seq_puts(seq, ",enospc_debug"); 1013 if (btrfs_test_opt(root, AUTO_DEFRAG)) 1014 seq_puts(seq, ",autodefrag"); 1015 if (btrfs_test_opt(root, INODE_MAP_CACHE)) 1016 seq_puts(seq, ",inode_cache"); 1017 if (btrfs_test_opt(root, SKIP_BALANCE)) 1018 seq_puts(seq, ",skip_balance"); 1019 if (btrfs_test_opt(root, RECOVERY)) 1020 seq_puts(seq, ",recovery"); 1021 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 1022 if (btrfs_test_opt(root, CHECK_INTEGRITY_INCLUDING_EXTENT_DATA)) 1023 seq_puts(seq, ",check_int_data"); 1024 else if (btrfs_test_opt(root, CHECK_INTEGRITY)) 1025 seq_puts(seq, ",check_int"); 1026 if (info->check_integrity_print_mask) 1027 seq_printf(seq, ",check_int_print_mask=%d", 1028 info->check_integrity_print_mask); 1029 #endif 1030 if (info->metadata_ratio) 1031 seq_printf(seq, ",metadata_ratio=%d", 1032 info->metadata_ratio); 1033 if (btrfs_test_opt(root, PANIC_ON_FATAL_ERROR)) 1034 seq_puts(seq, ",fatal_errors=panic"); 1035 if (info->commit_interval != BTRFS_DEFAULT_COMMIT_INTERVAL) 1036 seq_printf(seq, ",commit=%d", info->commit_interval); 1037 return 0; 1038 } 1039 1040 static int btrfs_test_super(struct super_block *s, void *data) 1041 { 1042 struct btrfs_fs_info *p = data; 1043 struct btrfs_fs_info *fs_info = btrfs_sb(s); 1044 1045 return fs_info->fs_devices == p->fs_devices; 1046 } 1047 1048 static int btrfs_set_super(struct super_block *s, void *data) 1049 { 1050 int err = set_anon_super(s, data); 1051 if (!err) 1052 s->s_fs_info = data; 1053 return err; 1054 } 1055 1056 /* 1057 * subvolumes are identified by ino 256 1058 */ 1059 static inline int is_subvolume_inode(struct inode *inode) 1060 { 1061 if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) 1062 return 1; 1063 return 0; 1064 } 1065 1066 /* 1067 * This will strip out the subvol=%s argument for an argument string and add 1068 * subvolid=0 to make sure we get the actual tree root for path walking to the 1069 * subvol we want. 1070 */ 1071 static char *setup_root_args(char *args) 1072 { 1073 unsigned len = strlen(args) + 2 + 1; 1074 char *src, *dst, *buf; 1075 1076 /* 1077 * We need the same args as before, but with this substitution: 1078 * s!subvol=[^,]+!subvolid=0! 1079 * 1080 * Since the replacement string is up to 2 bytes longer than the 1081 * original, allocate strlen(args) + 2 + 1 bytes. 1082 */ 1083 1084 src = strstr(args, "subvol="); 1085 /* This shouldn't happen, but just in case.. */ 1086 if (!src) 1087 return NULL; 1088 1089 buf = dst = kmalloc(len, GFP_NOFS); 1090 if (!buf) 1091 return NULL; 1092 1093 /* 1094 * If the subvol= arg is not at the start of the string, 1095 * copy whatever precedes it into buf. 1096 */ 1097 if (src != args) { 1098 *src++ = '\0'; 1099 strcpy(buf, args); 1100 dst += strlen(args); 1101 } 1102 1103 strcpy(dst, "subvolid=0"); 1104 dst += strlen("subvolid=0"); 1105 1106 /* 1107 * If there is a "," after the original subvol=... string, 1108 * copy that suffix into our buffer. Otherwise, we're done. 1109 */ 1110 src = strchr(src, ','); 1111 if (src) 1112 strcpy(dst, src); 1113 1114 return buf; 1115 } 1116 1117 static struct dentry *mount_subvol(const char *subvol_name, int flags, 1118 const char *device_name, char *data) 1119 { 1120 struct dentry *root; 1121 struct vfsmount *mnt; 1122 char *newargs; 1123 1124 newargs = setup_root_args(data); 1125 if (!newargs) 1126 return ERR_PTR(-ENOMEM); 1127 mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name, 1128 newargs); 1129 kfree(newargs); 1130 if (IS_ERR(mnt)) 1131 return ERR_CAST(mnt); 1132 1133 root = mount_subtree(mnt, subvol_name); 1134 1135 if (!IS_ERR(root) && !is_subvolume_inode(root->d_inode)) { 1136 struct super_block *s = root->d_sb; 1137 dput(root); 1138 root = ERR_PTR(-EINVAL); 1139 deactivate_locked_super(s); 1140 printk(KERN_ERR "BTRFS: '%s' is not a valid subvolume\n", 1141 subvol_name); 1142 } 1143 1144 return root; 1145 } 1146 1147 /* 1148 * Find a superblock for the given device / mount point. 1149 * 1150 * Note: This is based on get_sb_bdev from fs/super.c with a few additions 1151 * for multiple device setup. Make sure to keep it in sync. 1152 */ 1153 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, 1154 const char *device_name, void *data) 1155 { 1156 struct block_device *bdev = NULL; 1157 struct super_block *s; 1158 struct dentry *root; 1159 struct btrfs_fs_devices *fs_devices = NULL; 1160 struct btrfs_fs_info *fs_info = NULL; 1161 fmode_t mode = FMODE_READ; 1162 char *subvol_name = NULL; 1163 u64 subvol_objectid = 0; 1164 int error = 0; 1165 1166 if (!(flags & MS_RDONLY)) 1167 mode |= FMODE_WRITE; 1168 1169 error = btrfs_parse_early_options(data, mode, fs_type, 1170 &subvol_name, &subvol_objectid, 1171 &fs_devices); 1172 if (error) { 1173 kfree(subvol_name); 1174 return ERR_PTR(error); 1175 } 1176 1177 if (subvol_name) { 1178 root = mount_subvol(subvol_name, flags, device_name, data); 1179 kfree(subvol_name); 1180 return root; 1181 } 1182 1183 error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices); 1184 if (error) 1185 return ERR_PTR(error); 1186 1187 /* 1188 * Setup a dummy root and fs_info for test/set super. This is because 1189 * we don't actually fill this stuff out until open_ctree, but we need 1190 * it for searching for existing supers, so this lets us do that and 1191 * then open_ctree will properly initialize everything later. 1192 */ 1193 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); 1194 if (!fs_info) 1195 return ERR_PTR(-ENOMEM); 1196 1197 fs_info->fs_devices = fs_devices; 1198 1199 fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 1200 fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS); 1201 if (!fs_info->super_copy || !fs_info->super_for_commit) { 1202 error = -ENOMEM; 1203 goto error_fs_info; 1204 } 1205 1206 error = btrfs_open_devices(fs_devices, mode, fs_type); 1207 if (error) 1208 goto error_fs_info; 1209 1210 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { 1211 error = -EACCES; 1212 goto error_close_devices; 1213 } 1214 1215 bdev = fs_devices->latest_bdev; 1216 s = sget(fs_type, btrfs_test_super, btrfs_set_super, flags | MS_NOSEC, 1217 fs_info); 1218 if (IS_ERR(s)) { 1219 error = PTR_ERR(s); 1220 goto error_close_devices; 1221 } 1222 1223 if (s->s_root) { 1224 btrfs_close_devices(fs_devices); 1225 free_fs_info(fs_info); 1226 if ((flags ^ s->s_flags) & MS_RDONLY) 1227 error = -EBUSY; 1228 } else { 1229 char b[BDEVNAME_SIZE]; 1230 1231 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); 1232 btrfs_sb(s)->bdev_holder = fs_type; 1233 error = btrfs_fill_super(s, fs_devices, data, 1234 flags & MS_SILENT ? 1 : 0); 1235 } 1236 1237 root = !error ? get_default_root(s, subvol_objectid) : ERR_PTR(error); 1238 if (IS_ERR(root)) 1239 deactivate_locked_super(s); 1240 1241 return root; 1242 1243 error_close_devices: 1244 btrfs_close_devices(fs_devices); 1245 error_fs_info: 1246 free_fs_info(fs_info); 1247 return ERR_PTR(error); 1248 } 1249 1250 static void btrfs_set_max_workers(struct btrfs_workers *workers, int new_limit) 1251 { 1252 spin_lock_irq(&workers->lock); 1253 workers->max_workers = new_limit; 1254 spin_unlock_irq(&workers->lock); 1255 } 1256 1257 static void btrfs_resize_thread_pool(struct btrfs_fs_info *fs_info, 1258 int new_pool_size, int old_pool_size) 1259 { 1260 if (new_pool_size == old_pool_size) 1261 return; 1262 1263 fs_info->thread_pool_size = new_pool_size; 1264 1265 btrfs_info(fs_info, "resize thread pool %d -> %d", 1266 old_pool_size, new_pool_size); 1267 1268 btrfs_set_max_workers(&fs_info->generic_worker, new_pool_size); 1269 btrfs_set_max_workers(&fs_info->workers, new_pool_size); 1270 btrfs_set_max_workers(&fs_info->delalloc_workers, new_pool_size); 1271 btrfs_set_max_workers(&fs_info->submit_workers, new_pool_size); 1272 btrfs_set_max_workers(&fs_info->caching_workers, new_pool_size); 1273 btrfs_set_max_workers(&fs_info->fixup_workers, new_pool_size); 1274 btrfs_set_max_workers(&fs_info->endio_workers, new_pool_size); 1275 btrfs_set_max_workers(&fs_info->endio_meta_workers, new_pool_size); 1276 btrfs_set_max_workers(&fs_info->endio_meta_write_workers, new_pool_size); 1277 btrfs_set_max_workers(&fs_info->endio_write_workers, new_pool_size); 1278 btrfs_set_max_workers(&fs_info->endio_freespace_worker, new_pool_size); 1279 btrfs_set_max_workers(&fs_info->delayed_workers, new_pool_size); 1280 btrfs_set_max_workers(&fs_info->readahead_workers, new_pool_size); 1281 btrfs_set_max_workers(&fs_info->scrub_wr_completion_workers, 1282 new_pool_size); 1283 } 1284 1285 static inline void btrfs_remount_prepare(struct btrfs_fs_info *fs_info) 1286 { 1287 set_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1288 } 1289 1290 static inline void btrfs_remount_begin(struct btrfs_fs_info *fs_info, 1291 unsigned long old_opts, int flags) 1292 { 1293 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1294 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1295 (flags & MS_RDONLY))) { 1296 /* wait for any defraggers to finish */ 1297 wait_event(fs_info->transaction_wait, 1298 (atomic_read(&fs_info->defrag_running) == 0)); 1299 if (flags & MS_RDONLY) 1300 sync_filesystem(fs_info->sb); 1301 } 1302 } 1303 1304 static inline void btrfs_remount_cleanup(struct btrfs_fs_info *fs_info, 1305 unsigned long old_opts) 1306 { 1307 /* 1308 * We need cleanup all defragable inodes if the autodefragment is 1309 * close or the fs is R/O. 1310 */ 1311 if (btrfs_raw_test_opt(old_opts, AUTO_DEFRAG) && 1312 (!btrfs_raw_test_opt(fs_info->mount_opt, AUTO_DEFRAG) || 1313 (fs_info->sb->s_flags & MS_RDONLY))) { 1314 btrfs_cleanup_defrag_inodes(fs_info); 1315 } 1316 1317 clear_bit(BTRFS_FS_STATE_REMOUNTING, &fs_info->fs_state); 1318 } 1319 1320 static int btrfs_remount(struct super_block *sb, int *flags, char *data) 1321 { 1322 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1323 struct btrfs_root *root = fs_info->tree_root; 1324 unsigned old_flags = sb->s_flags; 1325 unsigned long old_opts = fs_info->mount_opt; 1326 unsigned long old_compress_type = fs_info->compress_type; 1327 u64 old_max_inline = fs_info->max_inline; 1328 u64 old_alloc_start = fs_info->alloc_start; 1329 int old_thread_pool_size = fs_info->thread_pool_size; 1330 unsigned int old_metadata_ratio = fs_info->metadata_ratio; 1331 int ret; 1332 1333 btrfs_remount_prepare(fs_info); 1334 1335 ret = btrfs_parse_options(root, data); 1336 if (ret) { 1337 ret = -EINVAL; 1338 goto restore; 1339 } 1340 1341 btrfs_remount_begin(fs_info, old_opts, *flags); 1342 btrfs_resize_thread_pool(fs_info, 1343 fs_info->thread_pool_size, old_thread_pool_size); 1344 1345 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) 1346 goto out; 1347 1348 if (*flags & MS_RDONLY) { 1349 /* 1350 * this also happens on 'umount -rf' or on shutdown, when 1351 * the filesystem is busy. 1352 */ 1353 1354 /* wait for the uuid_scan task to finish */ 1355 down(&fs_info->uuid_tree_rescan_sem); 1356 /* avoid complains from lockdep et al. */ 1357 up(&fs_info->uuid_tree_rescan_sem); 1358 1359 sb->s_flags |= MS_RDONLY; 1360 1361 btrfs_dev_replace_suspend_for_unmount(fs_info); 1362 btrfs_scrub_cancel(fs_info); 1363 btrfs_pause_balance(fs_info); 1364 1365 ret = btrfs_commit_super(root); 1366 if (ret) 1367 goto restore; 1368 } else { 1369 if (test_bit(BTRFS_FS_STATE_ERROR, &root->fs_info->fs_state)) { 1370 btrfs_err(fs_info, 1371 "Remounting read-write after error is not allowed"); 1372 ret = -EINVAL; 1373 goto restore; 1374 } 1375 if (fs_info->fs_devices->rw_devices == 0) { 1376 ret = -EACCES; 1377 goto restore; 1378 } 1379 1380 if (fs_info->fs_devices->missing_devices > 1381 fs_info->num_tolerated_disk_barrier_failures && 1382 !(*flags & MS_RDONLY)) { 1383 btrfs_warn(fs_info, 1384 "too many missing devices, writeable remount is not allowed"); 1385 ret = -EACCES; 1386 goto restore; 1387 } 1388 1389 if (btrfs_super_log_root(fs_info->super_copy) != 0) { 1390 ret = -EINVAL; 1391 goto restore; 1392 } 1393 1394 ret = btrfs_cleanup_fs_roots(fs_info); 1395 if (ret) 1396 goto restore; 1397 1398 /* recover relocation */ 1399 ret = btrfs_recover_relocation(root); 1400 if (ret) 1401 goto restore; 1402 1403 ret = btrfs_resume_balance_async(fs_info); 1404 if (ret) 1405 goto restore; 1406 1407 ret = btrfs_resume_dev_replace_async(fs_info); 1408 if (ret) { 1409 btrfs_warn(fs_info, "failed to resume dev_replace"); 1410 goto restore; 1411 } 1412 1413 if (!fs_info->uuid_root) { 1414 btrfs_info(fs_info, "creating UUID tree"); 1415 ret = btrfs_create_uuid_tree(fs_info); 1416 if (ret) { 1417 btrfs_warn(fs_info, "failed to create the UUID tree %d", ret); 1418 goto restore; 1419 } 1420 } 1421 sb->s_flags &= ~MS_RDONLY; 1422 } 1423 out: 1424 btrfs_remount_cleanup(fs_info, old_opts); 1425 return 0; 1426 1427 restore: 1428 /* We've hit an error - don't reset MS_RDONLY */ 1429 if (sb->s_flags & MS_RDONLY) 1430 old_flags |= MS_RDONLY; 1431 sb->s_flags = old_flags; 1432 fs_info->mount_opt = old_opts; 1433 fs_info->compress_type = old_compress_type; 1434 fs_info->max_inline = old_max_inline; 1435 mutex_lock(&fs_info->chunk_mutex); 1436 fs_info->alloc_start = old_alloc_start; 1437 mutex_unlock(&fs_info->chunk_mutex); 1438 btrfs_resize_thread_pool(fs_info, 1439 old_thread_pool_size, fs_info->thread_pool_size); 1440 fs_info->metadata_ratio = old_metadata_ratio; 1441 btrfs_remount_cleanup(fs_info, old_opts); 1442 return ret; 1443 } 1444 1445 /* Used to sort the devices by max_avail(descending sort) */ 1446 static int btrfs_cmp_device_free_bytes(const void *dev_info1, 1447 const void *dev_info2) 1448 { 1449 if (((struct btrfs_device_info *)dev_info1)->max_avail > 1450 ((struct btrfs_device_info *)dev_info2)->max_avail) 1451 return -1; 1452 else if (((struct btrfs_device_info *)dev_info1)->max_avail < 1453 ((struct btrfs_device_info *)dev_info2)->max_avail) 1454 return 1; 1455 else 1456 return 0; 1457 } 1458 1459 /* 1460 * sort the devices by max_avail, in which max free extent size of each device 1461 * is stored.(Descending Sort) 1462 */ 1463 static inline void btrfs_descending_sort_devices( 1464 struct btrfs_device_info *devices, 1465 size_t nr_devices) 1466 { 1467 sort(devices, nr_devices, sizeof(struct btrfs_device_info), 1468 btrfs_cmp_device_free_bytes, NULL); 1469 } 1470 1471 /* 1472 * The helper to calc the free space on the devices that can be used to store 1473 * file data. 1474 */ 1475 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) 1476 { 1477 struct btrfs_fs_info *fs_info = root->fs_info; 1478 struct btrfs_device_info *devices_info; 1479 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 1480 struct btrfs_device *device; 1481 u64 skip_space; 1482 u64 type; 1483 u64 avail_space; 1484 u64 used_space; 1485 u64 min_stripe_size; 1486 int min_stripes = 1, num_stripes = 1; 1487 int i = 0, nr_devices; 1488 int ret; 1489 1490 nr_devices = fs_info->fs_devices->open_devices; 1491 BUG_ON(!nr_devices); 1492 1493 devices_info = kmalloc_array(nr_devices, sizeof(*devices_info), 1494 GFP_NOFS); 1495 if (!devices_info) 1496 return -ENOMEM; 1497 1498 /* calc min stripe number for data space alloction */ 1499 type = btrfs_get_alloc_profile(root, 1); 1500 if (type & BTRFS_BLOCK_GROUP_RAID0) { 1501 min_stripes = 2; 1502 num_stripes = nr_devices; 1503 } else if (type & BTRFS_BLOCK_GROUP_RAID1) { 1504 min_stripes = 2; 1505 num_stripes = 2; 1506 } else if (type & BTRFS_BLOCK_GROUP_RAID10) { 1507 min_stripes = 4; 1508 num_stripes = 4; 1509 } 1510 1511 if (type & BTRFS_BLOCK_GROUP_DUP) 1512 min_stripe_size = 2 * BTRFS_STRIPE_LEN; 1513 else 1514 min_stripe_size = BTRFS_STRIPE_LEN; 1515 1516 list_for_each_entry(device, &fs_devices->devices, dev_list) { 1517 if (!device->in_fs_metadata || !device->bdev || 1518 device->is_tgtdev_for_dev_replace) 1519 continue; 1520 1521 avail_space = device->total_bytes - device->bytes_used; 1522 1523 /* align with stripe_len */ 1524 do_div(avail_space, BTRFS_STRIPE_LEN); 1525 avail_space *= BTRFS_STRIPE_LEN; 1526 1527 /* 1528 * In order to avoid overwritting the superblock on the drive, 1529 * btrfs starts at an offset of at least 1MB when doing chunk 1530 * allocation. 1531 */ 1532 skip_space = 1024 * 1024; 1533 1534 /* user can set the offset in fs_info->alloc_start. */ 1535 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= 1536 device->total_bytes) 1537 skip_space = max(fs_info->alloc_start, skip_space); 1538 1539 /* 1540 * btrfs can not use the free space in [0, skip_space - 1], 1541 * we must subtract it from the total. In order to implement 1542 * it, we account the used space in this range first. 1543 */ 1544 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, 1545 &used_space); 1546 if (ret) { 1547 kfree(devices_info); 1548 return ret; 1549 } 1550 1551 /* calc the free space in [0, skip_space - 1] */ 1552 skip_space -= used_space; 1553 1554 /* 1555 * we can use the free space in [0, skip_space - 1], subtract 1556 * it from the total. 1557 */ 1558 if (avail_space && avail_space >= skip_space) 1559 avail_space -= skip_space; 1560 else 1561 avail_space = 0; 1562 1563 if (avail_space < min_stripe_size) 1564 continue; 1565 1566 devices_info[i].dev = device; 1567 devices_info[i].max_avail = avail_space; 1568 1569 i++; 1570 } 1571 1572 nr_devices = i; 1573 1574 btrfs_descending_sort_devices(devices_info, nr_devices); 1575 1576 i = nr_devices - 1; 1577 avail_space = 0; 1578 while (nr_devices >= min_stripes) { 1579 if (num_stripes > nr_devices) 1580 num_stripes = nr_devices; 1581 1582 if (devices_info[i].max_avail >= min_stripe_size) { 1583 int j; 1584 u64 alloc_size; 1585 1586 avail_space += devices_info[i].max_avail * num_stripes; 1587 alloc_size = devices_info[i].max_avail; 1588 for (j = i + 1 - num_stripes; j <= i; j++) 1589 devices_info[j].max_avail -= alloc_size; 1590 } 1591 i--; 1592 nr_devices--; 1593 } 1594 1595 kfree(devices_info); 1596 *free_bytes = avail_space; 1597 return 0; 1598 } 1599 1600 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 1601 { 1602 struct btrfs_fs_info *fs_info = btrfs_sb(dentry->d_sb); 1603 struct btrfs_super_block *disk_super = fs_info->super_copy; 1604 struct list_head *head = &fs_info->space_info; 1605 struct btrfs_space_info *found; 1606 u64 total_used = 0; 1607 u64 total_free_data = 0; 1608 int bits = dentry->d_sb->s_blocksize_bits; 1609 __be32 *fsid = (__be32 *)fs_info->fsid; 1610 int ret; 1611 1612 /* holding chunk_muext to avoid allocating new chunks */ 1613 mutex_lock(&fs_info->chunk_mutex); 1614 rcu_read_lock(); 1615 list_for_each_entry_rcu(found, head, list) { 1616 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 1617 total_free_data += found->disk_total - found->disk_used; 1618 total_free_data -= 1619 btrfs_account_ro_block_groups_free_space(found); 1620 } 1621 1622 total_used += found->disk_used; 1623 } 1624 rcu_read_unlock(); 1625 1626 buf->f_namelen = BTRFS_NAME_LEN; 1627 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; 1628 buf->f_bfree = buf->f_blocks - (total_used >> bits); 1629 buf->f_bsize = dentry->d_sb->s_blocksize; 1630 buf->f_type = BTRFS_SUPER_MAGIC; 1631 buf->f_bavail = total_free_data; 1632 ret = btrfs_calc_avail_data_space(fs_info->tree_root, &total_free_data); 1633 if (ret) { 1634 mutex_unlock(&fs_info->chunk_mutex); 1635 return ret; 1636 } 1637 buf->f_bavail += total_free_data; 1638 buf->f_bavail = buf->f_bavail >> bits; 1639 mutex_unlock(&fs_info->chunk_mutex); 1640 1641 /* We treat it as constant endianness (it doesn't matter _which_) 1642 because we want the fsid to come out the same whether mounted 1643 on a big-endian or little-endian host */ 1644 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 1645 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 1646 /* Mask in the root object ID too, to disambiguate subvols */ 1647 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; 1648 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; 1649 1650 return 0; 1651 } 1652 1653 static void btrfs_kill_super(struct super_block *sb) 1654 { 1655 struct btrfs_fs_info *fs_info = btrfs_sb(sb); 1656 kill_anon_super(sb); 1657 free_fs_info(fs_info); 1658 } 1659 1660 static struct file_system_type btrfs_fs_type = { 1661 .owner = THIS_MODULE, 1662 .name = "btrfs", 1663 .mount = btrfs_mount, 1664 .kill_sb = btrfs_kill_super, 1665 .fs_flags = FS_REQUIRES_DEV, 1666 }; 1667 MODULE_ALIAS_FS("btrfs"); 1668 1669 /* 1670 * used by btrfsctl to scan devices when no FS is mounted 1671 */ 1672 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 1673 unsigned long arg) 1674 { 1675 struct btrfs_ioctl_vol_args *vol; 1676 struct btrfs_fs_devices *fs_devices; 1677 int ret = -ENOTTY; 1678 1679 if (!capable(CAP_SYS_ADMIN)) 1680 return -EPERM; 1681 1682 vol = memdup_user((void __user *)arg, sizeof(*vol)); 1683 if (IS_ERR(vol)) 1684 return PTR_ERR(vol); 1685 1686 switch (cmd) { 1687 case BTRFS_IOC_SCAN_DEV: 1688 ret = btrfs_scan_one_device(vol->name, FMODE_READ, 1689 &btrfs_fs_type, &fs_devices); 1690 break; 1691 case BTRFS_IOC_DEVICES_READY: 1692 ret = btrfs_scan_one_device(vol->name, FMODE_READ, 1693 &btrfs_fs_type, &fs_devices); 1694 if (ret) 1695 break; 1696 ret = !(fs_devices->num_devices == fs_devices->total_devices); 1697 break; 1698 } 1699 1700 kfree(vol); 1701 return ret; 1702 } 1703 1704 static int btrfs_freeze(struct super_block *sb) 1705 { 1706 struct btrfs_trans_handle *trans; 1707 struct btrfs_root *root = btrfs_sb(sb)->tree_root; 1708 1709 trans = btrfs_attach_transaction_barrier(root); 1710 if (IS_ERR(trans)) { 1711 /* no transaction, don't bother */ 1712 if (PTR_ERR(trans) == -ENOENT) 1713 return 0; 1714 return PTR_ERR(trans); 1715 } 1716 return btrfs_commit_transaction(trans, root); 1717 } 1718 1719 static int btrfs_unfreeze(struct super_block *sb) 1720 { 1721 return 0; 1722 } 1723 1724 static int btrfs_show_devname(struct seq_file *m, struct dentry *root) 1725 { 1726 struct btrfs_fs_info *fs_info = btrfs_sb(root->d_sb); 1727 struct btrfs_fs_devices *cur_devices; 1728 struct btrfs_device *dev, *first_dev = NULL; 1729 struct list_head *head; 1730 struct rcu_string *name; 1731 1732 mutex_lock(&fs_info->fs_devices->device_list_mutex); 1733 cur_devices = fs_info->fs_devices; 1734 while (cur_devices) { 1735 head = &cur_devices->devices; 1736 list_for_each_entry(dev, head, dev_list) { 1737 if (dev->missing) 1738 continue; 1739 if (!first_dev || dev->devid < first_dev->devid) 1740 first_dev = dev; 1741 } 1742 cur_devices = cur_devices->seed; 1743 } 1744 1745 if (first_dev) { 1746 rcu_read_lock(); 1747 name = rcu_dereference(first_dev->name); 1748 seq_escape(m, name->str, " \t\n\\"); 1749 rcu_read_unlock(); 1750 } else { 1751 WARN_ON(1); 1752 } 1753 mutex_unlock(&fs_info->fs_devices->device_list_mutex); 1754 return 0; 1755 } 1756 1757 static const struct super_operations btrfs_super_ops = { 1758 .drop_inode = btrfs_drop_inode, 1759 .evict_inode = btrfs_evict_inode, 1760 .put_super = btrfs_put_super, 1761 .sync_fs = btrfs_sync_fs, 1762 .show_options = btrfs_show_options, 1763 .show_devname = btrfs_show_devname, 1764 .write_inode = btrfs_write_inode, 1765 .alloc_inode = btrfs_alloc_inode, 1766 .destroy_inode = btrfs_destroy_inode, 1767 .statfs = btrfs_statfs, 1768 .remount_fs = btrfs_remount, 1769 .freeze_fs = btrfs_freeze, 1770 .unfreeze_fs = btrfs_unfreeze, 1771 }; 1772 1773 static const struct file_operations btrfs_ctl_fops = { 1774 .unlocked_ioctl = btrfs_control_ioctl, 1775 .compat_ioctl = btrfs_control_ioctl, 1776 .owner = THIS_MODULE, 1777 .llseek = noop_llseek, 1778 }; 1779 1780 static struct miscdevice btrfs_misc = { 1781 .minor = BTRFS_MINOR, 1782 .name = "btrfs-control", 1783 .fops = &btrfs_ctl_fops 1784 }; 1785 1786 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 1787 MODULE_ALIAS("devname:btrfs-control"); 1788 1789 static int btrfs_interface_init(void) 1790 { 1791 return misc_register(&btrfs_misc); 1792 } 1793 1794 static void btrfs_interface_exit(void) 1795 { 1796 if (misc_deregister(&btrfs_misc) < 0) 1797 printk(KERN_INFO "BTRFS: misc_deregister failed for control device\n"); 1798 } 1799 1800 static void btrfs_print_info(void) 1801 { 1802 printk(KERN_INFO "Btrfs loaded" 1803 #ifdef CONFIG_BTRFS_DEBUG 1804 ", debug=on" 1805 #endif 1806 #ifdef CONFIG_BTRFS_ASSERT 1807 ", assert=on" 1808 #endif 1809 #ifdef CONFIG_BTRFS_FS_CHECK_INTEGRITY 1810 ", integrity-checker=on" 1811 #endif 1812 "\n"); 1813 } 1814 1815 static int btrfs_run_sanity_tests(void) 1816 { 1817 int ret; 1818 1819 ret = btrfs_init_test_fs(); 1820 if (ret) 1821 return ret; 1822 1823 ret = btrfs_test_free_space_cache(); 1824 if (ret) 1825 goto out; 1826 ret = btrfs_test_extent_buffer_operations(); 1827 if (ret) 1828 goto out; 1829 ret = btrfs_test_extent_io(); 1830 if (ret) 1831 goto out; 1832 ret = btrfs_test_inodes(); 1833 out: 1834 btrfs_destroy_test_fs(); 1835 return ret; 1836 } 1837 1838 static int __init init_btrfs_fs(void) 1839 { 1840 int err; 1841 1842 err = btrfs_init_sysfs(); 1843 if (err) 1844 return err; 1845 1846 btrfs_init_compress(); 1847 1848 err = btrfs_init_cachep(); 1849 if (err) 1850 goto free_compress; 1851 1852 err = extent_io_init(); 1853 if (err) 1854 goto free_cachep; 1855 1856 err = extent_map_init(); 1857 if (err) 1858 goto free_extent_io; 1859 1860 err = ordered_data_init(); 1861 if (err) 1862 goto free_extent_map; 1863 1864 err = btrfs_delayed_inode_init(); 1865 if (err) 1866 goto free_ordered_data; 1867 1868 err = btrfs_auto_defrag_init(); 1869 if (err) 1870 goto free_delayed_inode; 1871 1872 err = btrfs_delayed_ref_init(); 1873 if (err) 1874 goto free_auto_defrag; 1875 1876 err = btrfs_prelim_ref_init(); 1877 if (err) 1878 goto free_prelim_ref; 1879 1880 err = btrfs_interface_init(); 1881 if (err) 1882 goto free_delayed_ref; 1883 1884 btrfs_init_lockdep(); 1885 1886 btrfs_print_info(); 1887 1888 err = btrfs_run_sanity_tests(); 1889 if (err) 1890 goto unregister_ioctl; 1891 1892 err = register_filesystem(&btrfs_fs_type); 1893 if (err) 1894 goto unregister_ioctl; 1895 1896 return 0; 1897 1898 unregister_ioctl: 1899 btrfs_interface_exit(); 1900 free_prelim_ref: 1901 btrfs_prelim_ref_exit(); 1902 free_delayed_ref: 1903 btrfs_delayed_ref_exit(); 1904 free_auto_defrag: 1905 btrfs_auto_defrag_exit(); 1906 free_delayed_inode: 1907 btrfs_delayed_inode_exit(); 1908 free_ordered_data: 1909 ordered_data_exit(); 1910 free_extent_map: 1911 extent_map_exit(); 1912 free_extent_io: 1913 extent_io_exit(); 1914 free_cachep: 1915 btrfs_destroy_cachep(); 1916 free_compress: 1917 btrfs_exit_compress(); 1918 btrfs_exit_sysfs(); 1919 return err; 1920 } 1921 1922 static void __exit exit_btrfs_fs(void) 1923 { 1924 btrfs_destroy_cachep(); 1925 btrfs_delayed_ref_exit(); 1926 btrfs_auto_defrag_exit(); 1927 btrfs_delayed_inode_exit(); 1928 btrfs_prelim_ref_exit(); 1929 ordered_data_exit(); 1930 extent_map_exit(); 1931 extent_io_exit(); 1932 btrfs_interface_exit(); 1933 unregister_filesystem(&btrfs_fs_type); 1934 btrfs_exit_sysfs(); 1935 btrfs_cleanup_fs_uuids(); 1936 btrfs_exit_compress(); 1937 } 1938 1939 module_init(init_btrfs_fs) 1940 module_exit(exit_btrfs_fs) 1941 1942 MODULE_LICENSE("GPL"); 1943