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/err.h> 20 #include <linux/uuid.h> 21 #include "ctree.h" 22 #include "transaction.h" 23 #include "disk-io.h" 24 #include "print-tree.h" 25 26 /* 27 * Read a root item from the tree. In case we detect a root item smaller then 28 * sizeof(root_item), we know it's an old version of the root structure and 29 * initialize all new fields to zero. The same happens if we detect mismatching 30 * generation numbers as then we know the root was once mounted with an older 31 * kernel that was not aware of the root item structure change. 32 */ 33 static void btrfs_read_root_item(struct extent_buffer *eb, int slot, 34 struct btrfs_root_item *item) 35 { 36 uuid_le uuid; 37 int len; 38 int need_reset = 0; 39 40 len = btrfs_item_size_nr(eb, slot); 41 read_extent_buffer(eb, item, btrfs_item_ptr_offset(eb, slot), 42 min_t(int, len, (int)sizeof(*item))); 43 if (len < sizeof(*item)) 44 need_reset = 1; 45 if (!need_reset && btrfs_root_generation(item) 46 != btrfs_root_generation_v2(item)) { 47 if (btrfs_root_generation_v2(item) != 0) { 48 btrfs_warn(eb->fs_info, 49 "mismatching generation and generation_v2 found in root item. This root was probably mounted with an older kernel. Resetting all new fields."); 50 } 51 need_reset = 1; 52 } 53 if (need_reset) { 54 memset(&item->generation_v2, 0, 55 sizeof(*item) - offsetof(struct btrfs_root_item, 56 generation_v2)); 57 58 uuid_le_gen(&uuid); 59 memcpy(item->uuid, uuid.b, BTRFS_UUID_SIZE); 60 } 61 } 62 63 /* 64 * btrfs_find_root - lookup the root by the key. 65 * root: the root of the root tree 66 * search_key: the key to search 67 * path: the path we search 68 * root_item: the root item of the tree we look for 69 * root_key: the root key of the tree we look for 70 * 71 * If ->offset of 'search_key' is -1ULL, it means we are not sure the offset 72 * of the search key, just lookup the root with the highest offset for a 73 * given objectid. 74 * 75 * If we find something return 0, otherwise > 0, < 0 on error. 76 */ 77 int btrfs_find_root(struct btrfs_root *root, const struct btrfs_key *search_key, 78 struct btrfs_path *path, struct btrfs_root_item *root_item, 79 struct btrfs_key *root_key) 80 { 81 struct btrfs_key found_key; 82 struct extent_buffer *l; 83 int ret; 84 int slot; 85 86 ret = btrfs_search_slot(NULL, root, search_key, path, 0, 0); 87 if (ret < 0) 88 return ret; 89 90 if (search_key->offset != -1ULL) { /* the search key is exact */ 91 if (ret > 0) 92 goto out; 93 } else { 94 BUG_ON(ret == 0); /* Logical error */ 95 if (path->slots[0] == 0) 96 goto out; 97 path->slots[0]--; 98 ret = 0; 99 } 100 101 l = path->nodes[0]; 102 slot = path->slots[0]; 103 104 btrfs_item_key_to_cpu(l, &found_key, slot); 105 if (found_key.objectid != search_key->objectid || 106 found_key.type != BTRFS_ROOT_ITEM_KEY) { 107 ret = 1; 108 goto out; 109 } 110 111 if (root_item) 112 btrfs_read_root_item(l, slot, root_item); 113 if (root_key) 114 memcpy(root_key, &found_key, sizeof(found_key)); 115 out: 116 btrfs_release_path(path); 117 return ret; 118 } 119 120 void btrfs_set_root_node(struct btrfs_root_item *item, 121 struct extent_buffer *node) 122 { 123 btrfs_set_root_bytenr(item, node->start); 124 btrfs_set_root_level(item, btrfs_header_level(node)); 125 btrfs_set_root_generation(item, btrfs_header_generation(node)); 126 } 127 128 /* 129 * copy the data in 'item' into the btree 130 */ 131 int btrfs_update_root(struct btrfs_trans_handle *trans, struct btrfs_root 132 *root, struct btrfs_key *key, struct btrfs_root_item 133 *item) 134 { 135 struct btrfs_fs_info *fs_info = root->fs_info; 136 struct btrfs_path *path; 137 struct extent_buffer *l; 138 int ret; 139 int slot; 140 unsigned long ptr; 141 u32 old_len; 142 143 path = btrfs_alloc_path(); 144 if (!path) 145 return -ENOMEM; 146 147 ret = btrfs_search_slot(trans, root, key, path, 0, 1); 148 if (ret < 0) { 149 btrfs_abort_transaction(trans, ret); 150 goto out; 151 } 152 153 if (ret != 0) { 154 btrfs_print_leaf(fs_info, path->nodes[0]); 155 btrfs_crit(fs_info, "unable to update root key %llu %u %llu", 156 key->objectid, key->type, key->offset); 157 BUG_ON(1); 158 } 159 160 l = path->nodes[0]; 161 slot = path->slots[0]; 162 ptr = btrfs_item_ptr_offset(l, slot); 163 old_len = btrfs_item_size_nr(l, slot); 164 165 /* 166 * If this is the first time we update the root item which originated 167 * from an older kernel, we need to enlarge the item size to make room 168 * for the added fields. 169 */ 170 if (old_len < sizeof(*item)) { 171 btrfs_release_path(path); 172 ret = btrfs_search_slot(trans, root, key, path, 173 -1, 1); 174 if (ret < 0) { 175 btrfs_abort_transaction(trans, ret); 176 goto out; 177 } 178 179 ret = btrfs_del_item(trans, root, path); 180 if (ret < 0) { 181 btrfs_abort_transaction(trans, ret); 182 goto out; 183 } 184 btrfs_release_path(path); 185 ret = btrfs_insert_empty_item(trans, root, path, 186 key, sizeof(*item)); 187 if (ret < 0) { 188 btrfs_abort_transaction(trans, ret); 189 goto out; 190 } 191 l = path->nodes[0]; 192 slot = path->slots[0]; 193 ptr = btrfs_item_ptr_offset(l, slot); 194 } 195 196 /* 197 * Update generation_v2 so at the next mount we know the new root 198 * fields are valid. 199 */ 200 btrfs_set_root_generation_v2(item, btrfs_root_generation(item)); 201 202 write_extent_buffer(l, item, ptr, sizeof(*item)); 203 btrfs_mark_buffer_dirty(path->nodes[0]); 204 out: 205 btrfs_free_path(path); 206 return ret; 207 } 208 209 int btrfs_insert_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, 210 const struct btrfs_key *key, struct btrfs_root_item *item) 211 { 212 /* 213 * Make sure generation v1 and v2 match. See update_root for details. 214 */ 215 btrfs_set_root_generation_v2(item, btrfs_root_generation(item)); 216 return btrfs_insert_item(trans, root, key, item, sizeof(*item)); 217 } 218 219 int btrfs_find_orphan_roots(struct btrfs_fs_info *fs_info) 220 { 221 struct btrfs_root *tree_root = fs_info->tree_root; 222 struct extent_buffer *leaf; 223 struct btrfs_path *path; 224 struct btrfs_key key; 225 struct btrfs_key root_key; 226 struct btrfs_root *root; 227 int err = 0; 228 int ret; 229 bool can_recover = true; 230 231 if (fs_info->sb->s_flags & MS_RDONLY) 232 can_recover = false; 233 234 path = btrfs_alloc_path(); 235 if (!path) 236 return -ENOMEM; 237 238 key.objectid = BTRFS_ORPHAN_OBJECTID; 239 key.type = BTRFS_ORPHAN_ITEM_KEY; 240 key.offset = 0; 241 242 root_key.type = BTRFS_ROOT_ITEM_KEY; 243 root_key.offset = (u64)-1; 244 245 while (1) { 246 ret = btrfs_search_slot(NULL, tree_root, &key, path, 0, 0); 247 if (ret < 0) { 248 err = ret; 249 break; 250 } 251 252 leaf = path->nodes[0]; 253 if (path->slots[0] >= btrfs_header_nritems(leaf)) { 254 ret = btrfs_next_leaf(tree_root, path); 255 if (ret < 0) 256 err = ret; 257 if (ret != 0) 258 break; 259 leaf = path->nodes[0]; 260 } 261 262 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 263 btrfs_release_path(path); 264 265 if (key.objectid != BTRFS_ORPHAN_OBJECTID || 266 key.type != BTRFS_ORPHAN_ITEM_KEY) 267 break; 268 269 root_key.objectid = key.offset; 270 key.offset++; 271 272 /* 273 * The root might have been inserted already, as before we look 274 * for orphan roots, log replay might have happened, which 275 * triggers a transaction commit and qgroup accounting, which 276 * in turn reads and inserts fs roots while doing backref 277 * walking. 278 */ 279 root = btrfs_lookup_fs_root(fs_info, root_key.objectid); 280 if (root) { 281 WARN_ON(!test_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, 282 &root->state)); 283 if (btrfs_root_refs(&root->root_item) == 0) 284 btrfs_add_dead_root(root); 285 continue; 286 } 287 288 root = btrfs_read_fs_root(tree_root, &root_key); 289 err = PTR_ERR_OR_ZERO(root); 290 if (err && err != -ENOENT) { 291 break; 292 } else if (err == -ENOENT) { 293 struct btrfs_trans_handle *trans; 294 295 btrfs_release_path(path); 296 297 trans = btrfs_join_transaction(tree_root); 298 if (IS_ERR(trans)) { 299 err = PTR_ERR(trans); 300 btrfs_handle_fs_error(fs_info, err, 301 "Failed to start trans to delete orphan item"); 302 break; 303 } 304 err = btrfs_del_orphan_item(trans, tree_root, 305 root_key.objectid); 306 btrfs_end_transaction(trans); 307 if (err) { 308 btrfs_handle_fs_error(fs_info, err, 309 "Failed to delete root orphan item"); 310 break; 311 } 312 continue; 313 } 314 315 err = btrfs_init_fs_root(root); 316 if (err) { 317 btrfs_free_fs_root(root); 318 break; 319 } 320 321 set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &root->state); 322 323 err = btrfs_insert_fs_root(fs_info, root); 324 if (err) { 325 BUG_ON(err == -EEXIST); 326 btrfs_free_fs_root(root); 327 break; 328 } 329 330 if (btrfs_root_refs(&root->root_item) == 0) 331 btrfs_add_dead_root(root); 332 } 333 334 btrfs_free_path(path); 335 return err; 336 } 337 338 /* drop the root item for 'key' from 'root' */ 339 int btrfs_del_root(struct btrfs_trans_handle *trans, struct btrfs_root *root, 340 const struct btrfs_key *key) 341 { 342 struct btrfs_path *path; 343 int ret; 344 345 path = btrfs_alloc_path(); 346 if (!path) 347 return -ENOMEM; 348 ret = btrfs_search_slot(trans, root, key, path, -1, 1); 349 if (ret < 0) 350 goto out; 351 352 BUG_ON(ret != 0); 353 354 ret = btrfs_del_item(trans, root, path); 355 out: 356 btrfs_free_path(path); 357 return ret; 358 } 359 360 int btrfs_del_root_ref(struct btrfs_trans_handle *trans, 361 struct btrfs_fs_info *fs_info, 362 u64 root_id, u64 ref_id, u64 dirid, u64 *sequence, 363 const char *name, int name_len) 364 365 { 366 struct btrfs_root *tree_root = fs_info->tree_root; 367 struct btrfs_path *path; 368 struct btrfs_root_ref *ref; 369 struct extent_buffer *leaf; 370 struct btrfs_key key; 371 unsigned long ptr; 372 int err = 0; 373 int ret; 374 375 path = btrfs_alloc_path(); 376 if (!path) 377 return -ENOMEM; 378 379 key.objectid = root_id; 380 key.type = BTRFS_ROOT_BACKREF_KEY; 381 key.offset = ref_id; 382 again: 383 ret = btrfs_search_slot(trans, tree_root, &key, path, -1, 1); 384 BUG_ON(ret < 0); 385 if (ret == 0) { 386 leaf = path->nodes[0]; 387 ref = btrfs_item_ptr(leaf, path->slots[0], 388 struct btrfs_root_ref); 389 390 WARN_ON(btrfs_root_ref_dirid(leaf, ref) != dirid); 391 WARN_ON(btrfs_root_ref_name_len(leaf, ref) != name_len); 392 ptr = (unsigned long)(ref + 1); 393 WARN_ON(memcmp_extent_buffer(leaf, name, ptr, name_len)); 394 *sequence = btrfs_root_ref_sequence(leaf, ref); 395 396 ret = btrfs_del_item(trans, tree_root, path); 397 if (ret) { 398 err = ret; 399 goto out; 400 } 401 } else 402 err = -ENOENT; 403 404 if (key.type == BTRFS_ROOT_BACKREF_KEY) { 405 btrfs_release_path(path); 406 key.objectid = ref_id; 407 key.type = BTRFS_ROOT_REF_KEY; 408 key.offset = root_id; 409 goto again; 410 } 411 412 out: 413 btrfs_free_path(path); 414 return err; 415 } 416 417 /* 418 * add a btrfs_root_ref item. type is either BTRFS_ROOT_REF_KEY 419 * or BTRFS_ROOT_BACKREF_KEY. 420 * 421 * The dirid, sequence, name and name_len refer to the directory entry 422 * that is referencing the root. 423 * 424 * For a forward ref, the root_id is the id of the tree referencing 425 * the root and ref_id is the id of the subvol or snapshot. 426 * 427 * For a back ref the root_id is the id of the subvol or snapshot and 428 * ref_id is the id of the tree referencing it. 429 * 430 * Will return 0, -ENOMEM, or anything from the CoW path 431 */ 432 int btrfs_add_root_ref(struct btrfs_trans_handle *trans, 433 struct btrfs_fs_info *fs_info, 434 u64 root_id, u64 ref_id, u64 dirid, u64 sequence, 435 const char *name, int name_len) 436 { 437 struct btrfs_root *tree_root = fs_info->tree_root; 438 struct btrfs_key key; 439 int ret; 440 struct btrfs_path *path; 441 struct btrfs_root_ref *ref; 442 struct extent_buffer *leaf; 443 unsigned long ptr; 444 445 path = btrfs_alloc_path(); 446 if (!path) 447 return -ENOMEM; 448 449 key.objectid = root_id; 450 key.type = BTRFS_ROOT_BACKREF_KEY; 451 key.offset = ref_id; 452 again: 453 ret = btrfs_insert_empty_item(trans, tree_root, path, &key, 454 sizeof(*ref) + name_len); 455 if (ret) { 456 btrfs_abort_transaction(trans, ret); 457 btrfs_free_path(path); 458 return ret; 459 } 460 461 leaf = path->nodes[0]; 462 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref); 463 btrfs_set_root_ref_dirid(leaf, ref, dirid); 464 btrfs_set_root_ref_sequence(leaf, ref, sequence); 465 btrfs_set_root_ref_name_len(leaf, ref, name_len); 466 ptr = (unsigned long)(ref + 1); 467 write_extent_buffer(leaf, name, ptr, name_len); 468 btrfs_mark_buffer_dirty(leaf); 469 470 if (key.type == BTRFS_ROOT_BACKREF_KEY) { 471 btrfs_release_path(path); 472 key.objectid = ref_id; 473 key.type = BTRFS_ROOT_REF_KEY; 474 key.offset = root_id; 475 goto again; 476 } 477 478 btrfs_free_path(path); 479 return 0; 480 } 481 482 /* 483 * Old btrfs forgets to init root_item->flags and root_item->byte_limit 484 * for subvolumes. To work around this problem, we steal a bit from 485 * root_item->inode_item->flags, and use it to indicate if those fields 486 * have been properly initialized. 487 */ 488 void btrfs_check_and_init_root_item(struct btrfs_root_item *root_item) 489 { 490 u64 inode_flags = btrfs_stack_inode_flags(&root_item->inode); 491 492 if (!(inode_flags & BTRFS_INODE_ROOT_ITEM_INIT)) { 493 inode_flags |= BTRFS_INODE_ROOT_ITEM_INIT; 494 btrfs_set_stack_inode_flags(&root_item->inode, inode_flags); 495 btrfs_set_root_flags(root_item, 0); 496 btrfs_set_root_limit(root_item, 0); 497 } 498 } 499 500 void btrfs_update_root_times(struct btrfs_trans_handle *trans, 501 struct btrfs_root *root) 502 { 503 struct btrfs_root_item *item = &root->root_item; 504 struct timespec ct = current_fs_time(root->fs_info->sb); 505 506 spin_lock(&root->root_item_lock); 507 btrfs_set_root_ctransid(item, trans->transid); 508 btrfs_set_stack_timespec_sec(&item->ctime, ct.tv_sec); 509 btrfs_set_stack_timespec_nsec(&item->ctime, ct.tv_nsec); 510 spin_unlock(&root->root_item_lock); 511 } 512