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 "compat.h" 43 #include "ctree.h" 44 #include "disk-io.h" 45 #include "transaction.h" 46 #include "btrfs_inode.h" 47 #include "ioctl.h" 48 #include "print-tree.h" 49 #include "xattr.h" 50 #include "volumes.h" 51 #include "version.h" 52 #include "export.h" 53 #include "compression.h" 54 55 static const struct super_operations btrfs_super_ops; 56 57 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno, 58 char nbuf[16]) 59 { 60 char *errstr = NULL; 61 62 switch (errno) { 63 case -EIO: 64 errstr = "IO failure"; 65 break; 66 case -ENOMEM: 67 errstr = "Out of memory"; 68 break; 69 case -EROFS: 70 errstr = "Readonly filesystem"; 71 break; 72 default: 73 if (nbuf) { 74 if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 75 errstr = nbuf; 76 } 77 break; 78 } 79 80 return errstr; 81 } 82 83 static void __save_error_info(struct btrfs_fs_info *fs_info) 84 { 85 /* 86 * today we only save the error info into ram. Long term we'll 87 * also send it down to the disk 88 */ 89 fs_info->fs_state = BTRFS_SUPER_FLAG_ERROR; 90 } 91 92 /* NOTE: 93 * We move write_super stuff at umount in order to avoid deadlock 94 * for umount hold all lock. 95 */ 96 static void save_error_info(struct btrfs_fs_info *fs_info) 97 { 98 __save_error_info(fs_info); 99 } 100 101 /* btrfs handle error by forcing the filesystem readonly */ 102 static void btrfs_handle_error(struct btrfs_fs_info *fs_info) 103 { 104 struct super_block *sb = fs_info->sb; 105 106 if (sb->s_flags & MS_RDONLY) 107 return; 108 109 if (fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) { 110 sb->s_flags |= MS_RDONLY; 111 printk(KERN_INFO "btrfs is forced readonly\n"); 112 } 113 } 114 115 /* 116 * __btrfs_std_error decodes expected errors from the caller and 117 * invokes the approciate error response. 118 */ 119 void __btrfs_std_error(struct btrfs_fs_info *fs_info, const char *function, 120 unsigned int line, int errno) 121 { 122 struct super_block *sb = fs_info->sb; 123 char nbuf[16]; 124 const char *errstr; 125 126 /* 127 * Special case: if the error is EROFS, and we're already 128 * under MS_RDONLY, then it is safe here. 129 */ 130 if (errno == -EROFS && (sb->s_flags & MS_RDONLY)) 131 return; 132 133 errstr = btrfs_decode_error(fs_info, errno, nbuf); 134 printk(KERN_CRIT "BTRFS error (device %s) in %s:%d: %s\n", 135 sb->s_id, function, line, errstr); 136 save_error_info(fs_info); 137 138 btrfs_handle_error(fs_info); 139 } 140 141 static void btrfs_put_super(struct super_block *sb) 142 { 143 struct btrfs_root *root = btrfs_sb(sb); 144 int ret; 145 146 ret = close_ctree(root); 147 sb->s_fs_info = NULL; 148 149 (void)ret; /* FIXME: need to fix VFS to return error? */ 150 } 151 152 enum { 153 Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum, 154 Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd, 155 Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress, 156 Opt_compress_type, Opt_compress_force, Opt_compress_force_type, 157 Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard, 158 Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed, Opt_err, 159 }; 160 161 static match_table_t tokens = { 162 {Opt_degraded, "degraded"}, 163 {Opt_subvol, "subvol=%s"}, 164 {Opt_subvolid, "subvolid=%d"}, 165 {Opt_device, "device=%s"}, 166 {Opt_nodatasum, "nodatasum"}, 167 {Opt_nodatacow, "nodatacow"}, 168 {Opt_nobarrier, "nobarrier"}, 169 {Opt_max_inline, "max_inline=%s"}, 170 {Opt_alloc_start, "alloc_start=%s"}, 171 {Opt_thread_pool, "thread_pool=%d"}, 172 {Opt_compress, "compress"}, 173 {Opt_compress_type, "compress=%s"}, 174 {Opt_compress_force, "compress-force"}, 175 {Opt_compress_force_type, "compress-force=%s"}, 176 {Opt_ssd, "ssd"}, 177 {Opt_ssd_spread, "ssd_spread"}, 178 {Opt_nossd, "nossd"}, 179 {Opt_noacl, "noacl"}, 180 {Opt_notreelog, "notreelog"}, 181 {Opt_flushoncommit, "flushoncommit"}, 182 {Opt_ratio, "metadata_ratio=%d"}, 183 {Opt_discard, "discard"}, 184 {Opt_space_cache, "space_cache"}, 185 {Opt_clear_cache, "clear_cache"}, 186 {Opt_user_subvol_rm_allowed, "user_subvol_rm_allowed"}, 187 {Opt_err, NULL}, 188 }; 189 190 /* 191 * Regular mount options parser. Everything that is needed only when 192 * reading in a new superblock is parsed here. 193 */ 194 int btrfs_parse_options(struct btrfs_root *root, char *options) 195 { 196 struct btrfs_fs_info *info = root->fs_info; 197 substring_t args[MAX_OPT_ARGS]; 198 char *p, *num, *orig; 199 int intarg; 200 int ret = 0; 201 char *compress_type; 202 bool compress_force = false; 203 204 if (!options) 205 return 0; 206 207 /* 208 * strsep changes the string, duplicate it because parse_options 209 * gets called twice 210 */ 211 options = kstrdup(options, GFP_NOFS); 212 if (!options) 213 return -ENOMEM; 214 215 orig = options; 216 217 while ((p = strsep(&options, ",")) != NULL) { 218 int token; 219 if (!*p) 220 continue; 221 222 token = match_token(p, tokens, args); 223 switch (token) { 224 case Opt_degraded: 225 printk(KERN_INFO "btrfs: allowing degraded mounts\n"); 226 btrfs_set_opt(info->mount_opt, DEGRADED); 227 break; 228 case Opt_subvol: 229 case Opt_subvolid: 230 case Opt_device: 231 /* 232 * These are parsed by btrfs_parse_early_options 233 * and can be happily ignored here. 234 */ 235 break; 236 case Opt_nodatasum: 237 printk(KERN_INFO "btrfs: setting nodatasum\n"); 238 btrfs_set_opt(info->mount_opt, NODATASUM); 239 break; 240 case Opt_nodatacow: 241 printk(KERN_INFO "btrfs: setting nodatacow\n"); 242 btrfs_set_opt(info->mount_opt, NODATACOW); 243 btrfs_set_opt(info->mount_opt, NODATASUM); 244 break; 245 case Opt_compress_force: 246 case Opt_compress_force_type: 247 compress_force = true; 248 case Opt_compress: 249 case Opt_compress_type: 250 if (token == Opt_compress || 251 token == Opt_compress_force || 252 strcmp(args[0].from, "zlib") == 0) { 253 compress_type = "zlib"; 254 info->compress_type = BTRFS_COMPRESS_ZLIB; 255 } else if (strcmp(args[0].from, "lzo") == 0) { 256 compress_type = "lzo"; 257 info->compress_type = BTRFS_COMPRESS_LZO; 258 } else { 259 ret = -EINVAL; 260 goto out; 261 } 262 263 btrfs_set_opt(info->mount_opt, COMPRESS); 264 if (compress_force) { 265 btrfs_set_opt(info->mount_opt, FORCE_COMPRESS); 266 pr_info("btrfs: force %s compression\n", 267 compress_type); 268 } else 269 pr_info("btrfs: use %s compression\n", 270 compress_type); 271 break; 272 case Opt_ssd: 273 printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); 274 btrfs_set_opt(info->mount_opt, SSD); 275 break; 276 case Opt_ssd_spread: 277 printk(KERN_INFO "btrfs: use spread ssd " 278 "allocation scheme\n"); 279 btrfs_set_opt(info->mount_opt, SSD); 280 btrfs_set_opt(info->mount_opt, SSD_SPREAD); 281 break; 282 case Opt_nossd: 283 printk(KERN_INFO "btrfs: not using ssd allocation " 284 "scheme\n"); 285 btrfs_set_opt(info->mount_opt, NOSSD); 286 btrfs_clear_opt(info->mount_opt, SSD); 287 btrfs_clear_opt(info->mount_opt, SSD_SPREAD); 288 break; 289 case Opt_nobarrier: 290 printk(KERN_INFO "btrfs: turning off barriers\n"); 291 btrfs_set_opt(info->mount_opt, NOBARRIER); 292 break; 293 case Opt_thread_pool: 294 intarg = 0; 295 match_int(&args[0], &intarg); 296 if (intarg) { 297 info->thread_pool_size = intarg; 298 printk(KERN_INFO "btrfs: thread pool %d\n", 299 info->thread_pool_size); 300 } 301 break; 302 case Opt_max_inline: 303 num = match_strdup(&args[0]); 304 if (num) { 305 info->max_inline = memparse(num, NULL); 306 kfree(num); 307 308 if (info->max_inline) { 309 info->max_inline = max_t(u64, 310 info->max_inline, 311 root->sectorsize); 312 } 313 printk(KERN_INFO "btrfs: max_inline at %llu\n", 314 (unsigned long long)info->max_inline); 315 } 316 break; 317 case Opt_alloc_start: 318 num = match_strdup(&args[0]); 319 if (num) { 320 info->alloc_start = memparse(num, NULL); 321 kfree(num); 322 printk(KERN_INFO 323 "btrfs: allocations start at %llu\n", 324 (unsigned long long)info->alloc_start); 325 } 326 break; 327 case Opt_noacl: 328 root->fs_info->sb->s_flags &= ~MS_POSIXACL; 329 break; 330 case Opt_notreelog: 331 printk(KERN_INFO "btrfs: disabling tree log\n"); 332 btrfs_set_opt(info->mount_opt, NOTREELOG); 333 break; 334 case Opt_flushoncommit: 335 printk(KERN_INFO "btrfs: turning on flush-on-commit\n"); 336 btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT); 337 break; 338 case Opt_ratio: 339 intarg = 0; 340 match_int(&args[0], &intarg); 341 if (intarg) { 342 info->metadata_ratio = intarg; 343 printk(KERN_INFO "btrfs: metadata ratio %d\n", 344 info->metadata_ratio); 345 } 346 break; 347 case Opt_discard: 348 btrfs_set_opt(info->mount_opt, DISCARD); 349 break; 350 case Opt_space_cache: 351 printk(KERN_INFO "btrfs: enabling disk space caching\n"); 352 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 353 break; 354 case Opt_clear_cache: 355 printk(KERN_INFO "btrfs: force clearing of disk cache\n"); 356 btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 357 break; 358 case Opt_user_subvol_rm_allowed: 359 btrfs_set_opt(info->mount_opt, USER_SUBVOL_RM_ALLOWED); 360 break; 361 case Opt_err: 362 printk(KERN_INFO "btrfs: unrecognized mount option " 363 "'%s'\n", p); 364 ret = -EINVAL; 365 goto out; 366 default: 367 break; 368 } 369 } 370 out: 371 kfree(orig); 372 return ret; 373 } 374 375 /* 376 * Parse mount options that are required early in the mount process. 377 * 378 * All other options will be parsed on much later in the mount process and 379 * only when we need to allocate a new super block. 380 */ 381 static int btrfs_parse_early_options(const char *options, fmode_t flags, 382 void *holder, char **subvol_name, u64 *subvol_objectid, 383 struct btrfs_fs_devices **fs_devices) 384 { 385 substring_t args[MAX_OPT_ARGS]; 386 char *opts, *p; 387 int error = 0; 388 int intarg; 389 390 if (!options) 391 goto out; 392 393 /* 394 * strsep changes the string, duplicate it because parse_options 395 * gets called twice 396 */ 397 opts = kstrdup(options, GFP_KERNEL); 398 if (!opts) 399 return -ENOMEM; 400 401 while ((p = strsep(&opts, ",")) != NULL) { 402 int token; 403 if (!*p) 404 continue; 405 406 token = match_token(p, tokens, args); 407 switch (token) { 408 case Opt_subvol: 409 *subvol_name = match_strdup(&args[0]); 410 break; 411 case Opt_subvolid: 412 intarg = 0; 413 error = match_int(&args[0], &intarg); 414 if (!error) { 415 /* we want the original fs_tree */ 416 if (!intarg) 417 *subvol_objectid = 418 BTRFS_FS_TREE_OBJECTID; 419 else 420 *subvol_objectid = intarg; 421 } 422 break; 423 case Opt_device: 424 error = btrfs_scan_one_device(match_strdup(&args[0]), 425 flags, holder, fs_devices); 426 if (error) 427 goto out_free_opts; 428 break; 429 default: 430 break; 431 } 432 } 433 434 out_free_opts: 435 kfree(opts); 436 out: 437 /* 438 * If no subvolume name is specified we use the default one. Allocate 439 * a copy of the string "." here so that code later in the 440 * mount path doesn't care if it's the default volume or another one. 441 */ 442 if (!*subvol_name) { 443 *subvol_name = kstrdup(".", GFP_KERNEL); 444 if (!*subvol_name) 445 return -ENOMEM; 446 } 447 return error; 448 } 449 450 static struct dentry *get_default_root(struct super_block *sb, 451 u64 subvol_objectid) 452 { 453 struct btrfs_root *root = sb->s_fs_info; 454 struct btrfs_root *new_root; 455 struct btrfs_dir_item *di; 456 struct btrfs_path *path; 457 struct btrfs_key location; 458 struct inode *inode; 459 struct dentry *dentry; 460 u64 dir_id; 461 int new = 0; 462 463 /* 464 * We have a specific subvol we want to mount, just setup location and 465 * go look up the root. 466 */ 467 if (subvol_objectid) { 468 location.objectid = subvol_objectid; 469 location.type = BTRFS_ROOT_ITEM_KEY; 470 location.offset = (u64)-1; 471 goto find_root; 472 } 473 474 path = btrfs_alloc_path(); 475 if (!path) 476 return ERR_PTR(-ENOMEM); 477 path->leave_spinning = 1; 478 479 /* 480 * Find the "default" dir item which points to the root item that we 481 * will mount by default if we haven't been given a specific subvolume 482 * to mount. 483 */ 484 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy); 485 di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0); 486 if (IS_ERR(di)) 487 return ERR_CAST(di); 488 if (!di) { 489 /* 490 * Ok the default dir item isn't there. This is weird since 491 * it's always been there, but don't freak out, just try and 492 * mount to root most subvolume. 493 */ 494 btrfs_free_path(path); 495 dir_id = BTRFS_FIRST_FREE_OBJECTID; 496 new_root = root->fs_info->fs_root; 497 goto setup_root; 498 } 499 500 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location); 501 btrfs_free_path(path); 502 503 find_root: 504 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location); 505 if (IS_ERR(new_root)) 506 return ERR_CAST(new_root); 507 508 if (btrfs_root_refs(&new_root->root_item) == 0) 509 return ERR_PTR(-ENOENT); 510 511 dir_id = btrfs_root_dirid(&new_root->root_item); 512 setup_root: 513 location.objectid = dir_id; 514 location.type = BTRFS_INODE_ITEM_KEY; 515 location.offset = 0; 516 517 inode = btrfs_iget(sb, &location, new_root, &new); 518 if (IS_ERR(inode)) 519 return ERR_CAST(inode); 520 521 /* 522 * If we're just mounting the root most subvol put the inode and return 523 * a reference to the dentry. We will have already gotten a reference 524 * to the inode in btrfs_fill_super so we're good to go. 525 */ 526 if (!new && sb->s_root->d_inode == inode) { 527 iput(inode); 528 return dget(sb->s_root); 529 } 530 531 if (new) { 532 const struct qstr name = { .name = "/", .len = 1 }; 533 534 /* 535 * New inode, we need to make the dentry a sibling of s_root so 536 * everything gets cleaned up properly on unmount. 537 */ 538 dentry = d_alloc(sb->s_root, &name); 539 if (!dentry) { 540 iput(inode); 541 return ERR_PTR(-ENOMEM); 542 } 543 d_splice_alias(inode, dentry); 544 } else { 545 /* 546 * We found the inode in cache, just find a dentry for it and 547 * put the reference to the inode we just got. 548 */ 549 dentry = d_find_alias(inode); 550 iput(inode); 551 } 552 553 return dentry; 554 } 555 556 static int btrfs_fill_super(struct super_block *sb, 557 struct btrfs_fs_devices *fs_devices, 558 void *data, int silent) 559 { 560 struct inode *inode; 561 struct dentry *root_dentry; 562 struct btrfs_root *tree_root; 563 struct btrfs_key key; 564 int err; 565 566 sb->s_maxbytes = MAX_LFS_FILESIZE; 567 sb->s_magic = BTRFS_SUPER_MAGIC; 568 sb->s_op = &btrfs_super_ops; 569 sb->s_d_op = &btrfs_dentry_operations; 570 sb->s_export_op = &btrfs_export_ops; 571 sb->s_xattr = btrfs_xattr_handlers; 572 sb->s_time_gran = 1; 573 #ifdef CONFIG_BTRFS_FS_POSIX_ACL 574 sb->s_flags |= MS_POSIXACL; 575 #endif 576 577 tree_root = open_ctree(sb, fs_devices, (char *)data); 578 579 if (IS_ERR(tree_root)) { 580 printk("btrfs: open_ctree failed\n"); 581 return PTR_ERR(tree_root); 582 } 583 sb->s_fs_info = tree_root; 584 585 key.objectid = BTRFS_FIRST_FREE_OBJECTID; 586 key.type = BTRFS_INODE_ITEM_KEY; 587 key.offset = 0; 588 inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root, NULL); 589 if (IS_ERR(inode)) { 590 err = PTR_ERR(inode); 591 goto fail_close; 592 } 593 594 root_dentry = d_alloc_root(inode); 595 if (!root_dentry) { 596 iput(inode); 597 err = -ENOMEM; 598 goto fail_close; 599 } 600 601 sb->s_root = root_dentry; 602 603 save_mount_options(sb, data); 604 return 0; 605 606 fail_close: 607 close_ctree(tree_root); 608 return err; 609 } 610 611 int btrfs_sync_fs(struct super_block *sb, int wait) 612 { 613 struct btrfs_trans_handle *trans; 614 struct btrfs_root *root = btrfs_sb(sb); 615 int ret; 616 617 if (!wait) { 618 filemap_flush(root->fs_info->btree_inode->i_mapping); 619 return 0; 620 } 621 622 btrfs_start_delalloc_inodes(root, 0); 623 btrfs_wait_ordered_extents(root, 0, 0); 624 625 trans = btrfs_start_transaction(root, 0); 626 ret = btrfs_commit_transaction(trans, root); 627 return ret; 628 } 629 630 static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) 631 { 632 struct btrfs_root *root = btrfs_sb(vfs->mnt_sb); 633 struct btrfs_fs_info *info = root->fs_info; 634 635 if (btrfs_test_opt(root, DEGRADED)) 636 seq_puts(seq, ",degraded"); 637 if (btrfs_test_opt(root, NODATASUM)) 638 seq_puts(seq, ",nodatasum"); 639 if (btrfs_test_opt(root, NODATACOW)) 640 seq_puts(seq, ",nodatacow"); 641 if (btrfs_test_opt(root, NOBARRIER)) 642 seq_puts(seq, ",nobarrier"); 643 if (info->max_inline != 8192 * 1024) 644 seq_printf(seq, ",max_inline=%llu", 645 (unsigned long long)info->max_inline); 646 if (info->alloc_start != 0) 647 seq_printf(seq, ",alloc_start=%llu", 648 (unsigned long long)info->alloc_start); 649 if (info->thread_pool_size != min_t(unsigned long, 650 num_online_cpus() + 2, 8)) 651 seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); 652 if (btrfs_test_opt(root, COMPRESS)) 653 seq_puts(seq, ",compress"); 654 if (btrfs_test_opt(root, NOSSD)) 655 seq_puts(seq, ",nossd"); 656 if (btrfs_test_opt(root, SSD_SPREAD)) 657 seq_puts(seq, ",ssd_spread"); 658 else if (btrfs_test_opt(root, SSD)) 659 seq_puts(seq, ",ssd"); 660 if (btrfs_test_opt(root, NOTREELOG)) 661 seq_puts(seq, ",notreelog"); 662 if (btrfs_test_opt(root, FLUSHONCOMMIT)) 663 seq_puts(seq, ",flushoncommit"); 664 if (btrfs_test_opt(root, DISCARD)) 665 seq_puts(seq, ",discard"); 666 if (!(root->fs_info->sb->s_flags & MS_POSIXACL)) 667 seq_puts(seq, ",noacl"); 668 return 0; 669 } 670 671 static int btrfs_test_super(struct super_block *s, void *data) 672 { 673 struct btrfs_root *test_root = data; 674 struct btrfs_root *root = btrfs_sb(s); 675 676 /* 677 * If this super block is going away, return false as it 678 * can't match as an existing super block. 679 */ 680 if (!atomic_read(&s->s_active)) 681 return 0; 682 return root->fs_info->fs_devices == test_root->fs_info->fs_devices; 683 } 684 685 static int btrfs_set_super(struct super_block *s, void *data) 686 { 687 s->s_fs_info = data; 688 689 return set_anon_super(s, data); 690 } 691 692 693 /* 694 * Find a superblock for the given device / mount point. 695 * 696 * Note: This is based on get_sb_bdev from fs/super.c with a few additions 697 * for multiple device setup. Make sure to keep it in sync. 698 */ 699 static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags, 700 const char *dev_name, void *data) 701 { 702 struct block_device *bdev = NULL; 703 struct super_block *s; 704 struct dentry *root; 705 struct btrfs_fs_devices *fs_devices = NULL; 706 struct btrfs_root *tree_root = NULL; 707 struct btrfs_fs_info *fs_info = NULL; 708 fmode_t mode = FMODE_READ; 709 char *subvol_name = NULL; 710 u64 subvol_objectid = 0; 711 int error = 0; 712 713 if (!(flags & MS_RDONLY)) 714 mode |= FMODE_WRITE; 715 716 error = btrfs_parse_early_options(data, mode, fs_type, 717 &subvol_name, &subvol_objectid, 718 &fs_devices); 719 if (error) 720 return ERR_PTR(error); 721 722 error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices); 723 if (error) 724 goto error_free_subvol_name; 725 726 error = btrfs_open_devices(fs_devices, mode, fs_type); 727 if (error) 728 goto error_free_subvol_name; 729 730 if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) { 731 error = -EACCES; 732 goto error_close_devices; 733 } 734 735 /* 736 * Setup a dummy root and fs_info for test/set super. This is because 737 * we don't actually fill this stuff out until open_ctree, but we need 738 * it for searching for existing supers, so this lets us do that and 739 * then open_ctree will properly initialize everything later. 740 */ 741 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS); 742 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS); 743 if (!fs_info || !tree_root) { 744 error = -ENOMEM; 745 goto error_close_devices; 746 } 747 fs_info->tree_root = tree_root; 748 fs_info->fs_devices = fs_devices; 749 tree_root->fs_info = fs_info; 750 751 bdev = fs_devices->latest_bdev; 752 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root); 753 if (IS_ERR(s)) 754 goto error_s; 755 756 if (s->s_root) { 757 if ((flags ^ s->s_flags) & MS_RDONLY) { 758 deactivate_locked_super(s); 759 error = -EBUSY; 760 goto error_close_devices; 761 } 762 763 btrfs_close_devices(fs_devices); 764 } else { 765 char b[BDEVNAME_SIZE]; 766 767 s->s_flags = flags; 768 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); 769 error = btrfs_fill_super(s, fs_devices, data, 770 flags & MS_SILENT ? 1 : 0); 771 if (error) { 772 deactivate_locked_super(s); 773 goto error_free_subvol_name; 774 } 775 776 btrfs_sb(s)->fs_info->bdev_holder = fs_type; 777 s->s_flags |= MS_ACTIVE; 778 } 779 780 root = get_default_root(s, subvol_objectid); 781 if (IS_ERR(root)) { 782 error = PTR_ERR(root); 783 deactivate_locked_super(s); 784 goto error_free_subvol_name; 785 } 786 /* if they gave us a subvolume name bind mount into that */ 787 if (strcmp(subvol_name, ".")) { 788 struct dentry *new_root; 789 mutex_lock(&root->d_inode->i_mutex); 790 new_root = lookup_one_len(subvol_name, root, 791 strlen(subvol_name)); 792 mutex_unlock(&root->d_inode->i_mutex); 793 794 if (IS_ERR(new_root)) { 795 dput(root); 796 deactivate_locked_super(s); 797 error = PTR_ERR(new_root); 798 goto error_free_subvol_name; 799 } 800 if (!new_root->d_inode) { 801 dput(root); 802 dput(new_root); 803 deactivate_locked_super(s); 804 error = -ENXIO; 805 goto error_free_subvol_name; 806 } 807 dput(root); 808 root = new_root; 809 } 810 811 kfree(subvol_name); 812 return root; 813 814 error_s: 815 error = PTR_ERR(s); 816 error_close_devices: 817 btrfs_close_devices(fs_devices); 818 kfree(fs_info); 819 kfree(tree_root); 820 error_free_subvol_name: 821 kfree(subvol_name); 822 return ERR_PTR(error); 823 } 824 825 static int btrfs_remount(struct super_block *sb, int *flags, char *data) 826 { 827 struct btrfs_root *root = btrfs_sb(sb); 828 int ret; 829 830 ret = btrfs_parse_options(root, data); 831 if (ret) 832 return -EINVAL; 833 834 if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) 835 return 0; 836 837 if (*flags & MS_RDONLY) { 838 sb->s_flags |= MS_RDONLY; 839 840 ret = btrfs_commit_super(root); 841 WARN_ON(ret); 842 } else { 843 if (root->fs_info->fs_devices->rw_devices == 0) 844 return -EACCES; 845 846 if (btrfs_super_log_root(&root->fs_info->super_copy) != 0) 847 return -EINVAL; 848 849 ret = btrfs_cleanup_fs_roots(root->fs_info); 850 WARN_ON(ret); 851 852 /* recover relocation */ 853 ret = btrfs_recover_relocation(root); 854 WARN_ON(ret); 855 856 sb->s_flags &= ~MS_RDONLY; 857 } 858 859 return 0; 860 } 861 862 /* 863 * The helper to calc the free space on the devices that can be used to store 864 * file data. 865 */ 866 static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes) 867 { 868 struct btrfs_fs_info *fs_info = root->fs_info; 869 struct btrfs_device_info *devices_info; 870 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 871 struct btrfs_device *device; 872 u64 skip_space; 873 u64 type; 874 u64 avail_space; 875 u64 used_space; 876 u64 min_stripe_size; 877 int min_stripes = 1; 878 int i = 0, nr_devices; 879 int ret; 880 881 nr_devices = fs_info->fs_devices->rw_devices; 882 BUG_ON(!nr_devices); 883 884 devices_info = kmalloc(sizeof(*devices_info) * nr_devices, 885 GFP_NOFS); 886 if (!devices_info) 887 return -ENOMEM; 888 889 /* calc min stripe number for data space alloction */ 890 type = btrfs_get_alloc_profile(root, 1); 891 if (type & BTRFS_BLOCK_GROUP_RAID0) 892 min_stripes = 2; 893 else if (type & BTRFS_BLOCK_GROUP_RAID1) 894 min_stripes = 2; 895 else if (type & BTRFS_BLOCK_GROUP_RAID10) 896 min_stripes = 4; 897 898 if (type & BTRFS_BLOCK_GROUP_DUP) 899 min_stripe_size = 2 * BTRFS_STRIPE_LEN; 900 else 901 min_stripe_size = BTRFS_STRIPE_LEN; 902 903 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) { 904 if (!device->in_fs_metadata) 905 continue; 906 907 avail_space = device->total_bytes - device->bytes_used; 908 909 /* align with stripe_len */ 910 do_div(avail_space, BTRFS_STRIPE_LEN); 911 avail_space *= BTRFS_STRIPE_LEN; 912 913 /* 914 * In order to avoid overwritting the superblock on the drive, 915 * btrfs starts at an offset of at least 1MB when doing chunk 916 * allocation. 917 */ 918 skip_space = 1024 * 1024; 919 920 /* user can set the offset in fs_info->alloc_start. */ 921 if (fs_info->alloc_start + BTRFS_STRIPE_LEN <= 922 device->total_bytes) 923 skip_space = max(fs_info->alloc_start, skip_space); 924 925 /* 926 * btrfs can not use the free space in [0, skip_space - 1], 927 * we must subtract it from the total. In order to implement 928 * it, we account the used space in this range first. 929 */ 930 ret = btrfs_account_dev_extents_size(device, 0, skip_space - 1, 931 &used_space); 932 if (ret) { 933 kfree(devices_info); 934 return ret; 935 } 936 937 /* calc the free space in [0, skip_space - 1] */ 938 skip_space -= used_space; 939 940 /* 941 * we can use the free space in [0, skip_space - 1], subtract 942 * it from the total. 943 */ 944 if (avail_space && avail_space >= skip_space) 945 avail_space -= skip_space; 946 else 947 avail_space = 0; 948 949 if (avail_space < min_stripe_size) 950 continue; 951 952 devices_info[i].dev = device; 953 devices_info[i].max_avail = avail_space; 954 955 i++; 956 } 957 958 nr_devices = i; 959 960 btrfs_descending_sort_devices(devices_info, nr_devices); 961 962 i = nr_devices - 1; 963 avail_space = 0; 964 while (nr_devices >= min_stripes) { 965 if (devices_info[i].max_avail >= min_stripe_size) { 966 int j; 967 u64 alloc_size; 968 969 avail_space += devices_info[i].max_avail * min_stripes; 970 alloc_size = devices_info[i].max_avail; 971 for (j = i + 1 - min_stripes; j <= i; j++) 972 devices_info[j].max_avail -= alloc_size; 973 } 974 i--; 975 nr_devices--; 976 } 977 978 kfree(devices_info); 979 *free_bytes = avail_space; 980 return 0; 981 } 982 983 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf) 984 { 985 struct btrfs_root *root = btrfs_sb(dentry->d_sb); 986 struct btrfs_super_block *disk_super = &root->fs_info->super_copy; 987 struct list_head *head = &root->fs_info->space_info; 988 struct btrfs_space_info *found; 989 u64 total_used = 0; 990 u64 total_free_data = 0; 991 int bits = dentry->d_sb->s_blocksize_bits; 992 __be32 *fsid = (__be32 *)root->fs_info->fsid; 993 int ret; 994 995 /* holding chunk_muext to avoid allocating new chunks */ 996 mutex_lock(&root->fs_info->chunk_mutex); 997 rcu_read_lock(); 998 list_for_each_entry_rcu(found, head, list) { 999 if (found->flags & BTRFS_BLOCK_GROUP_DATA) { 1000 total_free_data += found->disk_total - found->disk_used; 1001 total_free_data -= 1002 btrfs_account_ro_block_groups_free_space(found); 1003 } 1004 1005 total_used += found->disk_used; 1006 } 1007 rcu_read_unlock(); 1008 1009 buf->f_namelen = BTRFS_NAME_LEN; 1010 buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits; 1011 buf->f_bfree = buf->f_blocks - (total_used >> bits); 1012 buf->f_bsize = dentry->d_sb->s_blocksize; 1013 buf->f_type = BTRFS_SUPER_MAGIC; 1014 buf->f_bavail = total_free_data; 1015 ret = btrfs_calc_avail_data_space(root, &total_free_data); 1016 if (ret) { 1017 mutex_unlock(&root->fs_info->chunk_mutex); 1018 return ret; 1019 } 1020 buf->f_bavail += total_free_data; 1021 buf->f_bavail = buf->f_bavail >> bits; 1022 mutex_unlock(&root->fs_info->chunk_mutex); 1023 1024 /* We treat it as constant endianness (it doesn't matter _which_) 1025 because we want the fsid to come out the same whether mounted 1026 on a big-endian or little-endian host */ 1027 buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]); 1028 buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]); 1029 /* Mask in the root object ID too, to disambiguate subvols */ 1030 buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32; 1031 buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid; 1032 1033 return 0; 1034 } 1035 1036 static struct file_system_type btrfs_fs_type = { 1037 .owner = THIS_MODULE, 1038 .name = "btrfs", 1039 .mount = btrfs_mount, 1040 .kill_sb = kill_anon_super, 1041 .fs_flags = FS_REQUIRES_DEV, 1042 }; 1043 1044 /* 1045 * used by btrfsctl to scan devices when no FS is mounted 1046 */ 1047 static long btrfs_control_ioctl(struct file *file, unsigned int cmd, 1048 unsigned long arg) 1049 { 1050 struct btrfs_ioctl_vol_args *vol; 1051 struct btrfs_fs_devices *fs_devices; 1052 int ret = -ENOTTY; 1053 1054 if (!capable(CAP_SYS_ADMIN)) 1055 return -EPERM; 1056 1057 vol = memdup_user((void __user *)arg, sizeof(*vol)); 1058 if (IS_ERR(vol)) 1059 return PTR_ERR(vol); 1060 1061 switch (cmd) { 1062 case BTRFS_IOC_SCAN_DEV: 1063 ret = btrfs_scan_one_device(vol->name, FMODE_READ, 1064 &btrfs_fs_type, &fs_devices); 1065 break; 1066 } 1067 1068 kfree(vol); 1069 return ret; 1070 } 1071 1072 static int btrfs_freeze(struct super_block *sb) 1073 { 1074 struct btrfs_root *root = btrfs_sb(sb); 1075 mutex_lock(&root->fs_info->transaction_kthread_mutex); 1076 mutex_lock(&root->fs_info->cleaner_mutex); 1077 return 0; 1078 } 1079 1080 static int btrfs_unfreeze(struct super_block *sb) 1081 { 1082 struct btrfs_root *root = btrfs_sb(sb); 1083 mutex_unlock(&root->fs_info->cleaner_mutex); 1084 mutex_unlock(&root->fs_info->transaction_kthread_mutex); 1085 return 0; 1086 } 1087 1088 static const struct super_operations btrfs_super_ops = { 1089 .drop_inode = btrfs_drop_inode, 1090 .evict_inode = btrfs_evict_inode, 1091 .put_super = btrfs_put_super, 1092 .sync_fs = btrfs_sync_fs, 1093 .show_options = btrfs_show_options, 1094 .write_inode = btrfs_write_inode, 1095 .dirty_inode = btrfs_dirty_inode, 1096 .alloc_inode = btrfs_alloc_inode, 1097 .destroy_inode = btrfs_destroy_inode, 1098 .statfs = btrfs_statfs, 1099 .remount_fs = btrfs_remount, 1100 .freeze_fs = btrfs_freeze, 1101 .unfreeze_fs = btrfs_unfreeze, 1102 }; 1103 1104 static const struct file_operations btrfs_ctl_fops = { 1105 .unlocked_ioctl = btrfs_control_ioctl, 1106 .compat_ioctl = btrfs_control_ioctl, 1107 .owner = THIS_MODULE, 1108 .llseek = noop_llseek, 1109 }; 1110 1111 static struct miscdevice btrfs_misc = { 1112 .minor = BTRFS_MINOR, 1113 .name = "btrfs-control", 1114 .fops = &btrfs_ctl_fops 1115 }; 1116 1117 MODULE_ALIAS_MISCDEV(BTRFS_MINOR); 1118 MODULE_ALIAS("devname:btrfs-control"); 1119 1120 static int btrfs_interface_init(void) 1121 { 1122 return misc_register(&btrfs_misc); 1123 } 1124 1125 static void btrfs_interface_exit(void) 1126 { 1127 if (misc_deregister(&btrfs_misc) < 0) 1128 printk(KERN_INFO "misc_deregister failed for control device"); 1129 } 1130 1131 static int __init init_btrfs_fs(void) 1132 { 1133 int err; 1134 1135 err = btrfs_init_sysfs(); 1136 if (err) 1137 return err; 1138 1139 err = btrfs_init_compress(); 1140 if (err) 1141 goto free_sysfs; 1142 1143 err = btrfs_init_cachep(); 1144 if (err) 1145 goto free_compress; 1146 1147 err = extent_io_init(); 1148 if (err) 1149 goto free_cachep; 1150 1151 err = extent_map_init(); 1152 if (err) 1153 goto free_extent_io; 1154 1155 err = btrfs_interface_init(); 1156 if (err) 1157 goto free_extent_map; 1158 1159 err = register_filesystem(&btrfs_fs_type); 1160 if (err) 1161 goto unregister_ioctl; 1162 1163 printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION); 1164 return 0; 1165 1166 unregister_ioctl: 1167 btrfs_interface_exit(); 1168 free_extent_map: 1169 extent_map_exit(); 1170 free_extent_io: 1171 extent_io_exit(); 1172 free_cachep: 1173 btrfs_destroy_cachep(); 1174 free_compress: 1175 btrfs_exit_compress(); 1176 free_sysfs: 1177 btrfs_exit_sysfs(); 1178 return err; 1179 } 1180 1181 static void __exit exit_btrfs_fs(void) 1182 { 1183 btrfs_destroy_cachep(); 1184 extent_map_exit(); 1185 extent_io_exit(); 1186 btrfs_interface_exit(); 1187 unregister_filesystem(&btrfs_fs_type); 1188 btrfs_exit_sysfs(); 1189 btrfs_cleanup_fs_uuids(); 1190 btrfs_exit_compress(); 1191 } 1192 1193 module_init(init_btrfs_fs) 1194 module_exit(exit_btrfs_fs) 1195 1196 MODULE_LICENSE("GPL"); 1197