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 #include <linux/sched.h> 19 #include <linux/pagemap.h> 20 #include <linux/writeback.h> 21 #include "hash.h" 22 #include "crc32c.h" 23 #include "ctree.h" 24 #include "disk-io.h" 25 #include "print-tree.h" 26 #include "transaction.h" 27 #include "volumes.h" 28 #include "locking.h" 29 30 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK 31 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE 32 #define BLOCK_GROUP_SYSTEM EXTENT_NEW 33 34 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY 35 36 static int finish_current_insert(struct btrfs_trans_handle *trans, struct 37 btrfs_root *extent_root); 38 static int del_pending_extents(struct btrfs_trans_handle *trans, struct 39 btrfs_root *extent_root); 40 static struct btrfs_block_group_cache * 41 __btrfs_find_block_group(struct btrfs_root *root, 42 struct btrfs_block_group_cache *hint, 43 u64 search_start, int data, int owner); 44 45 void maybe_lock_mutex(struct btrfs_root *root) 46 { 47 if (root != root->fs_info->extent_root && 48 root != root->fs_info->chunk_root && 49 root != root->fs_info->dev_root) { 50 mutex_lock(&root->fs_info->alloc_mutex); 51 } 52 } 53 54 void maybe_unlock_mutex(struct btrfs_root *root) 55 { 56 if (root != root->fs_info->extent_root && 57 root != root->fs_info->chunk_root && 58 root != root->fs_info->dev_root) { 59 mutex_unlock(&root->fs_info->alloc_mutex); 60 } 61 } 62 63 static int cache_block_group(struct btrfs_root *root, 64 struct btrfs_block_group_cache *block_group) 65 { 66 struct btrfs_path *path; 67 int ret; 68 struct btrfs_key key; 69 struct extent_buffer *leaf; 70 struct extent_io_tree *free_space_cache; 71 int slot; 72 u64 last = 0; 73 u64 hole_size; 74 u64 first_free; 75 int found = 0; 76 77 if (!block_group) 78 return 0; 79 80 root = root->fs_info->extent_root; 81 free_space_cache = &root->fs_info->free_space_cache; 82 83 if (block_group->cached) 84 return 0; 85 86 path = btrfs_alloc_path(); 87 if (!path) 88 return -ENOMEM; 89 90 path->reada = 2; 91 path->skip_locking = 1; 92 first_free = block_group->key.objectid; 93 key.objectid = block_group->key.objectid; 94 key.offset = 0; 95 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); 96 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 97 if (ret < 0) 98 return ret; 99 ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); 100 if (ret < 0) 101 return ret; 102 if (ret == 0) { 103 leaf = path->nodes[0]; 104 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 105 if (key.objectid + key.offset > first_free) 106 first_free = key.objectid + key.offset; 107 } 108 while(1) { 109 leaf = path->nodes[0]; 110 slot = path->slots[0]; 111 if (slot >= btrfs_header_nritems(leaf)) { 112 ret = btrfs_next_leaf(root, path); 113 if (ret < 0) 114 goto err; 115 if (ret == 0) { 116 continue; 117 } else { 118 break; 119 } 120 } 121 btrfs_item_key_to_cpu(leaf, &key, slot); 122 if (key.objectid < block_group->key.objectid) { 123 goto next; 124 } 125 if (key.objectid >= block_group->key.objectid + 126 block_group->key.offset) { 127 break; 128 } 129 130 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) { 131 if (!found) { 132 last = first_free; 133 found = 1; 134 } 135 if (key.objectid > last) { 136 hole_size = key.objectid - last; 137 set_extent_dirty(free_space_cache, last, 138 last + hole_size - 1, 139 GFP_NOFS); 140 } 141 last = key.objectid + key.offset; 142 } 143 next: 144 path->slots[0]++; 145 } 146 147 if (!found) 148 last = first_free; 149 if (block_group->key.objectid + 150 block_group->key.offset > last) { 151 hole_size = block_group->key.objectid + 152 block_group->key.offset - last; 153 set_extent_dirty(free_space_cache, last, 154 last + hole_size - 1, GFP_NOFS); 155 } 156 block_group->cached = 1; 157 err: 158 btrfs_free_path(path); 159 return 0; 160 } 161 162 struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct 163 btrfs_fs_info *info, 164 u64 bytenr) 165 { 166 struct extent_io_tree *block_group_cache; 167 struct btrfs_block_group_cache *block_group = NULL; 168 u64 ptr; 169 u64 start; 170 u64 end; 171 int ret; 172 173 bytenr = max_t(u64, bytenr, 174 BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); 175 block_group_cache = &info->block_group_cache; 176 ret = find_first_extent_bit(block_group_cache, 177 bytenr, &start, &end, 178 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | 179 BLOCK_GROUP_SYSTEM); 180 if (ret) { 181 return NULL; 182 } 183 ret = get_state_private(block_group_cache, start, &ptr); 184 if (ret) 185 return NULL; 186 187 block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; 188 return block_group; 189 } 190 191 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct 192 btrfs_fs_info *info, 193 u64 bytenr) 194 { 195 struct extent_io_tree *block_group_cache; 196 struct btrfs_block_group_cache *block_group = NULL; 197 u64 ptr; 198 u64 start; 199 u64 end; 200 int ret; 201 202 bytenr = max_t(u64, bytenr, 203 BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE); 204 block_group_cache = &info->block_group_cache; 205 ret = find_first_extent_bit(block_group_cache, 206 bytenr, &start, &end, 207 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | 208 BLOCK_GROUP_SYSTEM); 209 if (ret) { 210 return NULL; 211 } 212 ret = get_state_private(block_group_cache, start, &ptr); 213 if (ret) 214 return NULL; 215 216 block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr; 217 if (block_group->key.objectid <= bytenr && bytenr < 218 block_group->key.objectid + block_group->key.offset) 219 return block_group; 220 return NULL; 221 } 222 223 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits) 224 { 225 return (cache->flags & bits) == bits; 226 } 227 228 static int noinline find_search_start(struct btrfs_root *root, 229 struct btrfs_block_group_cache **cache_ret, 230 u64 *start_ret, u64 num, int data) 231 { 232 int ret; 233 struct btrfs_block_group_cache *cache = *cache_ret; 234 struct extent_io_tree *free_space_cache; 235 struct extent_state *state; 236 u64 last; 237 u64 start = 0; 238 u64 cache_miss = 0; 239 u64 total_fs_bytes; 240 u64 search_start = *start_ret; 241 int wrapped = 0; 242 243 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); 244 free_space_cache = &root->fs_info->free_space_cache; 245 246 if (!cache) 247 goto out; 248 249 again: 250 ret = cache_block_group(root, cache); 251 if (ret) { 252 goto out; 253 } 254 255 last = max(search_start, cache->key.objectid); 256 if (!block_group_bits(cache, data) || cache->ro) 257 goto new_group; 258 259 spin_lock_irq(&free_space_cache->lock); 260 state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY); 261 while(1) { 262 if (!state) { 263 if (!cache_miss) 264 cache_miss = last; 265 spin_unlock_irq(&free_space_cache->lock); 266 goto new_group; 267 } 268 269 start = max(last, state->start); 270 last = state->end + 1; 271 if (last - start < num) { 272 do { 273 state = extent_state_next(state); 274 } while(state && !(state->state & EXTENT_DIRTY)); 275 continue; 276 } 277 spin_unlock_irq(&free_space_cache->lock); 278 if (cache->ro) { 279 goto new_group; 280 } 281 if (start + num > cache->key.objectid + cache->key.offset) 282 goto new_group; 283 if (!block_group_bits(cache, data)) { 284 printk("block group bits don't match %Lu %d\n", cache->flags, data); 285 } 286 *start_ret = start; 287 return 0; 288 } 289 out: 290 cache = btrfs_lookup_block_group(root->fs_info, search_start); 291 if (!cache) { 292 printk("Unable to find block group for %Lu\n", search_start); 293 WARN_ON(1); 294 } 295 return -ENOSPC; 296 297 new_group: 298 last = cache->key.objectid + cache->key.offset; 299 wrapped: 300 cache = btrfs_lookup_first_block_group(root->fs_info, last); 301 if (!cache || cache->key.objectid >= total_fs_bytes) { 302 no_cache: 303 if (!wrapped) { 304 wrapped = 1; 305 last = search_start; 306 goto wrapped; 307 } 308 goto out; 309 } 310 if (cache_miss && !cache->cached) { 311 cache_block_group(root, cache); 312 last = cache_miss; 313 cache = btrfs_lookup_first_block_group(root->fs_info, last); 314 } 315 cache_miss = 0; 316 cache = __btrfs_find_block_group(root, cache, last, data, 0); 317 if (!cache) 318 goto no_cache; 319 *cache_ret = cache; 320 goto again; 321 } 322 323 static u64 div_factor(u64 num, int factor) 324 { 325 if (factor == 10) 326 return num; 327 num *= factor; 328 do_div(num, 10); 329 return num; 330 } 331 332 static int block_group_state_bits(u64 flags) 333 { 334 int bits = 0; 335 if (flags & BTRFS_BLOCK_GROUP_DATA) 336 bits |= BLOCK_GROUP_DATA; 337 if (flags & BTRFS_BLOCK_GROUP_METADATA) 338 bits |= BLOCK_GROUP_METADATA; 339 if (flags & BTRFS_BLOCK_GROUP_SYSTEM) 340 bits |= BLOCK_GROUP_SYSTEM; 341 return bits; 342 } 343 344 static struct btrfs_block_group_cache * 345 __btrfs_find_block_group(struct btrfs_root *root, 346 struct btrfs_block_group_cache *hint, 347 u64 search_start, int data, int owner) 348 { 349 struct btrfs_block_group_cache *cache; 350 struct extent_io_tree *block_group_cache; 351 struct btrfs_block_group_cache *found_group = NULL; 352 struct btrfs_fs_info *info = root->fs_info; 353 u64 used; 354 u64 last = 0; 355 u64 start; 356 u64 end; 357 u64 free_check; 358 u64 ptr; 359 int bit; 360 int ret; 361 int full_search = 0; 362 int factor = 10; 363 int wrapped = 0; 364 365 block_group_cache = &info->block_group_cache; 366 367 if (data & BTRFS_BLOCK_GROUP_METADATA) 368 factor = 9; 369 370 bit = block_group_state_bits(data); 371 372 if (search_start) { 373 struct btrfs_block_group_cache *shint; 374 shint = btrfs_lookup_first_block_group(info, search_start); 375 if (shint && block_group_bits(shint, data) && !shint->ro) { 376 used = btrfs_block_group_used(&shint->item); 377 if (used + shint->pinned < 378 div_factor(shint->key.offset, factor)) { 379 return shint; 380 } 381 } 382 } 383 if (hint && !hint->ro && block_group_bits(hint, data)) { 384 used = btrfs_block_group_used(&hint->item); 385 if (used + hint->pinned < 386 div_factor(hint->key.offset, factor)) { 387 return hint; 388 } 389 last = hint->key.objectid + hint->key.offset; 390 } else { 391 if (hint) 392 last = max(hint->key.objectid, search_start); 393 else 394 last = search_start; 395 } 396 again: 397 while(1) { 398 ret = find_first_extent_bit(block_group_cache, last, 399 &start, &end, bit); 400 if (ret) 401 break; 402 403 ret = get_state_private(block_group_cache, start, &ptr); 404 if (ret) { 405 last = end + 1; 406 continue; 407 } 408 409 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; 410 last = cache->key.objectid + cache->key.offset; 411 used = btrfs_block_group_used(&cache->item); 412 413 if (!cache->ro && block_group_bits(cache, data)) { 414 free_check = div_factor(cache->key.offset, factor); 415 if (used + cache->pinned < free_check) { 416 found_group = cache; 417 goto found; 418 } 419 } 420 cond_resched(); 421 } 422 if (!wrapped) { 423 last = search_start; 424 wrapped = 1; 425 goto again; 426 } 427 if (!full_search && factor < 10) { 428 last = search_start; 429 full_search = 1; 430 factor = 10; 431 goto again; 432 } 433 found: 434 return found_group; 435 } 436 437 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root, 438 struct btrfs_block_group_cache 439 *hint, u64 search_start, 440 int data, int owner) 441 { 442 443 struct btrfs_block_group_cache *ret; 444 mutex_lock(&root->fs_info->alloc_mutex); 445 ret = __btrfs_find_block_group(root, hint, search_start, data, owner); 446 mutex_unlock(&root->fs_info->alloc_mutex); 447 return ret; 448 } 449 static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation, 450 u64 owner, u64 owner_offset) 451 { 452 u32 high_crc = ~(u32)0; 453 u32 low_crc = ~(u32)0; 454 __le64 lenum; 455 lenum = cpu_to_le64(root_objectid); 456 high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum)); 457 lenum = cpu_to_le64(ref_generation); 458 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); 459 if (owner >= BTRFS_FIRST_FREE_OBJECTID) { 460 lenum = cpu_to_le64(owner); 461 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); 462 lenum = cpu_to_le64(owner_offset); 463 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); 464 } 465 return ((u64)high_crc << 32) | (u64)low_crc; 466 } 467 468 static int match_extent_ref(struct extent_buffer *leaf, 469 struct btrfs_extent_ref *disk_ref, 470 struct btrfs_extent_ref *cpu_ref) 471 { 472 int ret; 473 int len; 474 475 if (cpu_ref->objectid) 476 len = sizeof(*cpu_ref); 477 else 478 len = 2 * sizeof(u64); 479 ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref, 480 len); 481 return ret == 0; 482 } 483 484 static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans, 485 struct btrfs_root *root, 486 struct btrfs_path *path, u64 bytenr, 487 u64 root_objectid, 488 u64 ref_generation, u64 owner, 489 u64 owner_offset, int del) 490 { 491 u64 hash; 492 struct btrfs_key key; 493 struct btrfs_key found_key; 494 struct btrfs_extent_ref ref; 495 struct extent_buffer *leaf; 496 struct btrfs_extent_ref *disk_ref; 497 int ret; 498 int ret2; 499 500 btrfs_set_stack_ref_root(&ref, root_objectid); 501 btrfs_set_stack_ref_generation(&ref, ref_generation); 502 btrfs_set_stack_ref_objectid(&ref, owner); 503 btrfs_set_stack_ref_offset(&ref, owner_offset); 504 505 hash = hash_extent_ref(root_objectid, ref_generation, owner, 506 owner_offset); 507 key.offset = hash; 508 key.objectid = bytenr; 509 key.type = BTRFS_EXTENT_REF_KEY; 510 511 while (1) { 512 ret = btrfs_search_slot(trans, root, &key, path, 513 del ? -1 : 0, del); 514 if (ret < 0) 515 goto out; 516 leaf = path->nodes[0]; 517 if (ret != 0) { 518 u32 nritems = btrfs_header_nritems(leaf); 519 if (path->slots[0] >= nritems) { 520 ret2 = btrfs_next_leaf(root, path); 521 if (ret2) 522 goto out; 523 leaf = path->nodes[0]; 524 } 525 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 526 if (found_key.objectid != bytenr || 527 found_key.type != BTRFS_EXTENT_REF_KEY) 528 goto out; 529 key.offset = found_key.offset; 530 if (del) { 531 btrfs_release_path(root, path); 532 continue; 533 } 534 } 535 disk_ref = btrfs_item_ptr(path->nodes[0], 536 path->slots[0], 537 struct btrfs_extent_ref); 538 if (match_extent_ref(path->nodes[0], disk_ref, &ref)) { 539 ret = 0; 540 goto out; 541 } 542 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 543 key.offset = found_key.offset + 1; 544 btrfs_release_path(root, path); 545 } 546 out: 547 return ret; 548 } 549 550 /* 551 * Back reference rules. Back refs have three main goals: 552 * 553 * 1) differentiate between all holders of references to an extent so that 554 * when a reference is dropped we can make sure it was a valid reference 555 * before freeing the extent. 556 * 557 * 2) Provide enough information to quickly find the holders of an extent 558 * if we notice a given block is corrupted or bad. 559 * 560 * 3) Make it easy to migrate blocks for FS shrinking or storage pool 561 * maintenance. This is actually the same as #2, but with a slightly 562 * different use case. 563 * 564 * File extents can be referenced by: 565 * 566 * - multiple snapshots, subvolumes, or different generations in one subvol 567 * - different files inside a single subvolume (in theory, not implemented yet) 568 * - different offsets inside a file (bookend extents in file.c) 569 * 570 * The extent ref structure has fields for: 571 * 572 * - Objectid of the subvolume root 573 * - Generation number of the tree holding the reference 574 * - objectid of the file holding the reference 575 * - offset in the file corresponding to the key holding the reference 576 * 577 * When a file extent is allocated the fields are filled in: 578 * (root_key.objectid, trans->transid, inode objectid, offset in file) 579 * 580 * When a leaf is cow'd new references are added for every file extent found 581 * in the leaf. It looks the same as the create case, but trans->transid 582 * will be different when the block is cow'd. 583 * 584 * (root_key.objectid, trans->transid, inode objectid, offset in file) 585 * 586 * When a file extent is removed either during snapshot deletion or file 587 * truncation, the corresponding back reference is found 588 * by searching for: 589 * 590 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf), 591 * inode objectid, offset in file) 592 * 593 * Btree extents can be referenced by: 594 * 595 * - Different subvolumes 596 * - Different generations of the same subvolume 597 * 598 * Storing sufficient information for a full reverse mapping of a btree 599 * block would require storing the lowest key of the block in the backref, 600 * and it would require updating that lowest key either before write out or 601 * every time it changed. Instead, the objectid of the lowest key is stored 602 * along with the level of the tree block. This provides a hint 603 * about where in the btree the block can be found. Searches through the 604 * btree only need to look for a pointer to that block, so they stop one 605 * level higher than the level recorded in the backref. 606 * 607 * Some btrees do not do reference counting on their extents. These 608 * include the extent tree and the tree of tree roots. Backrefs for these 609 * trees always have a generation of zero. 610 * 611 * When a tree block is created, back references are inserted: 612 * 613 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid) 614 * 615 * When a tree block is cow'd in a reference counted root, 616 * new back references are added for all the blocks it points to. 617 * These are of the form (trans->transid will have increased since creation): 618 * 619 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid) 620 * 621 * Because the lowest_key_objectid and the level are just hints 622 * they are not used when backrefs are deleted. When a backref is deleted: 623 * 624 * if backref was for a tree root: 625 * root_objectid = root->root_key.objectid 626 * else 627 * root_objectid = btrfs_header_owner(parent) 628 * 629 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0) 630 * 631 * Back Reference Key hashing: 632 * 633 * Back references have four fields, each 64 bits long. Unfortunately, 634 * This is hashed into a single 64 bit number and placed into the key offset. 635 * The key objectid corresponds to the first byte in the extent, and the 636 * key type is set to BTRFS_EXTENT_REF_KEY 637 */ 638 int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans, 639 struct btrfs_root *root, 640 struct btrfs_path *path, u64 bytenr, 641 u64 root_objectid, u64 ref_generation, 642 u64 owner, u64 owner_offset) 643 { 644 u64 hash; 645 struct btrfs_key key; 646 struct btrfs_extent_ref ref; 647 struct btrfs_extent_ref *disk_ref; 648 int ret; 649 650 btrfs_set_stack_ref_root(&ref, root_objectid); 651 btrfs_set_stack_ref_generation(&ref, ref_generation); 652 btrfs_set_stack_ref_objectid(&ref, owner); 653 btrfs_set_stack_ref_offset(&ref, owner_offset); 654 655 hash = hash_extent_ref(root_objectid, ref_generation, owner, 656 owner_offset); 657 key.offset = hash; 658 key.objectid = bytenr; 659 key.type = BTRFS_EXTENT_REF_KEY; 660 661 ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref)); 662 while (ret == -EEXIST) { 663 disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 664 struct btrfs_extent_ref); 665 if (match_extent_ref(path->nodes[0], disk_ref, &ref)) 666 goto out; 667 key.offset++; 668 btrfs_release_path(root, path); 669 ret = btrfs_insert_empty_item(trans, root, path, &key, 670 sizeof(ref)); 671 } 672 if (ret) 673 goto out; 674 disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 675 struct btrfs_extent_ref); 676 write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref, 677 sizeof(ref)); 678 btrfs_mark_buffer_dirty(path->nodes[0]); 679 out: 680 btrfs_release_path(root, path); 681 return ret; 682 } 683 684 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, 685 struct btrfs_root *root, 686 u64 bytenr, u64 num_bytes, 687 u64 root_objectid, u64 ref_generation, 688 u64 owner, u64 owner_offset) 689 { 690 struct btrfs_path *path; 691 int ret; 692 struct btrfs_key key; 693 struct extent_buffer *l; 694 struct btrfs_extent_item *item; 695 u32 refs; 696 697 WARN_ON(num_bytes < root->sectorsize); 698 path = btrfs_alloc_path(); 699 if (!path) 700 return -ENOMEM; 701 702 path->reada = 1; 703 key.objectid = bytenr; 704 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); 705 key.offset = num_bytes; 706 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 707 0, 1); 708 if (ret < 0) 709 return ret; 710 if (ret != 0) { 711 BUG(); 712 } 713 BUG_ON(ret != 0); 714 l = path->nodes[0]; 715 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); 716 refs = btrfs_extent_refs(l, item); 717 btrfs_set_extent_refs(l, item, refs + 1); 718 btrfs_mark_buffer_dirty(path->nodes[0]); 719 720 btrfs_release_path(root->fs_info->extent_root, path); 721 722 path->reada = 1; 723 ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root, 724 path, bytenr, root_objectid, 725 ref_generation, owner, owner_offset); 726 BUG_ON(ret); 727 finish_current_insert(trans, root->fs_info->extent_root); 728 del_pending_extents(trans, root->fs_info->extent_root); 729 730 btrfs_free_path(path); 731 return 0; 732 } 733 734 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, 735 struct btrfs_root *root, 736 u64 bytenr, u64 num_bytes, 737 u64 root_objectid, u64 ref_generation, 738 u64 owner, u64 owner_offset) 739 { 740 int ret; 741 742 mutex_lock(&root->fs_info->alloc_mutex); 743 ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 744 root_objectid, ref_generation, 745 owner, owner_offset); 746 mutex_unlock(&root->fs_info->alloc_mutex); 747 return ret; 748 } 749 750 int btrfs_extent_post_op(struct btrfs_trans_handle *trans, 751 struct btrfs_root *root) 752 { 753 finish_current_insert(trans, root->fs_info->extent_root); 754 del_pending_extents(trans, root->fs_info->extent_root); 755 return 0; 756 } 757 758 static int lookup_extent_ref(struct btrfs_trans_handle *trans, 759 struct btrfs_root *root, u64 bytenr, 760 u64 num_bytes, u32 *refs) 761 { 762 struct btrfs_path *path; 763 int ret; 764 struct btrfs_key key; 765 struct extent_buffer *l; 766 struct btrfs_extent_item *item; 767 768 WARN_ON(num_bytes < root->sectorsize); 769 path = btrfs_alloc_path(); 770 path->reada = 1; 771 key.objectid = bytenr; 772 key.offset = num_bytes; 773 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); 774 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path, 775 0, 0); 776 if (ret < 0) 777 goto out; 778 if (ret != 0) { 779 btrfs_print_leaf(root, path->nodes[0]); 780 printk("failed to find block number %Lu\n", bytenr); 781 BUG(); 782 } 783 l = path->nodes[0]; 784 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); 785 *refs = btrfs_extent_refs(l, item); 786 out: 787 btrfs_free_path(path); 788 return 0; 789 } 790 791 u32 btrfs_count_snapshots_in_path(struct btrfs_root *root, 792 struct btrfs_path *count_path, 793 u64 expected_owner, 794 u64 first_extent) 795 { 796 struct btrfs_root *extent_root = root->fs_info->extent_root; 797 struct btrfs_path *path; 798 u64 bytenr; 799 u64 found_objectid; 800 u64 found_owner; 801 u64 root_objectid = root->root_key.objectid; 802 u32 total_count = 0; 803 u32 extent_refs; 804 u32 cur_count; 805 u32 nritems; 806 int ret; 807 struct btrfs_key key; 808 struct btrfs_key found_key; 809 struct extent_buffer *l; 810 struct btrfs_extent_item *item; 811 struct btrfs_extent_ref *ref_item; 812 int level = -1; 813 814 /* FIXME, needs locking */ 815 BUG(); 816 817 mutex_lock(&root->fs_info->alloc_mutex); 818 path = btrfs_alloc_path(); 819 again: 820 if (level == -1) 821 bytenr = first_extent; 822 else 823 bytenr = count_path->nodes[level]->start; 824 825 cur_count = 0; 826 key.objectid = bytenr; 827 key.offset = 0; 828 829 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); 830 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); 831 if (ret < 0) 832 goto out; 833 BUG_ON(ret == 0); 834 835 l = path->nodes[0]; 836 btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); 837 838 if (found_key.objectid != bytenr || 839 found_key.type != BTRFS_EXTENT_ITEM_KEY) { 840 goto out; 841 } 842 843 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item); 844 extent_refs = btrfs_extent_refs(l, item); 845 while (1) { 846 l = path->nodes[0]; 847 nritems = btrfs_header_nritems(l); 848 if (path->slots[0] >= nritems) { 849 ret = btrfs_next_leaf(extent_root, path); 850 if (ret == 0) 851 continue; 852 break; 853 } 854 btrfs_item_key_to_cpu(l, &found_key, path->slots[0]); 855 if (found_key.objectid != bytenr) 856 break; 857 858 if (found_key.type != BTRFS_EXTENT_REF_KEY) { 859 path->slots[0]++; 860 continue; 861 } 862 863 cur_count++; 864 ref_item = btrfs_item_ptr(l, path->slots[0], 865 struct btrfs_extent_ref); 866 found_objectid = btrfs_ref_root(l, ref_item); 867 868 if (found_objectid != root_objectid) { 869 total_count = 2; 870 goto out; 871 } 872 if (level == -1) { 873 found_owner = btrfs_ref_objectid(l, ref_item); 874 if (found_owner != expected_owner) { 875 total_count = 2; 876 goto out; 877 } 878 /* 879 * nasty. we don't count a reference held by 880 * the running transaction. This allows nodatacow 881 * to avoid cow most of the time 882 */ 883 if (found_owner >= BTRFS_FIRST_FREE_OBJECTID && 884 btrfs_ref_generation(l, ref_item) == 885 root->fs_info->generation) { 886 extent_refs--; 887 } 888 } 889 total_count = 1; 890 path->slots[0]++; 891 } 892 /* 893 * if there is more than one reference against a data extent, 894 * we have to assume the other ref is another snapshot 895 */ 896 if (level == -1 && extent_refs > 1) { 897 total_count = 2; 898 goto out; 899 } 900 if (cur_count == 0) { 901 total_count = 0; 902 goto out; 903 } 904 if (level >= 0 && root->node == count_path->nodes[level]) 905 goto out; 906 level++; 907 btrfs_release_path(root, path); 908 goto again; 909 910 out: 911 btrfs_free_path(path); 912 mutex_unlock(&root->fs_info->alloc_mutex); 913 return total_count; 914 } 915 916 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, 917 struct extent_buffer *buf) 918 { 919 u64 bytenr; 920 u32 nritems; 921 struct btrfs_key key; 922 struct btrfs_file_extent_item *fi; 923 int i; 924 int level; 925 int ret; 926 int faili; 927 928 if (!root->ref_cows) 929 return 0; 930 931 mutex_lock(&root->fs_info->alloc_mutex); 932 level = btrfs_header_level(buf); 933 nritems = btrfs_header_nritems(buf); 934 for (i = 0; i < nritems; i++) { 935 if (level == 0) { 936 u64 disk_bytenr; 937 btrfs_item_key_to_cpu(buf, &key, i); 938 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) 939 continue; 940 fi = btrfs_item_ptr(buf, i, 941 struct btrfs_file_extent_item); 942 if (btrfs_file_extent_type(buf, fi) == 943 BTRFS_FILE_EXTENT_INLINE) 944 continue; 945 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); 946 if (disk_bytenr == 0) 947 continue; 948 ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr, 949 btrfs_file_extent_disk_num_bytes(buf, fi), 950 root->root_key.objectid, trans->transid, 951 key.objectid, key.offset); 952 if (ret) { 953 faili = i; 954 goto fail; 955 } 956 } else { 957 bytenr = btrfs_node_blockptr(buf, i); 958 btrfs_node_key_to_cpu(buf, &key, i); 959 ret = __btrfs_inc_extent_ref(trans, root, bytenr, 960 btrfs_level_size(root, level - 1), 961 root->root_key.objectid, 962 trans->transid, 963 level - 1, key.objectid); 964 if (ret) { 965 faili = i; 966 goto fail; 967 } 968 } 969 } 970 mutex_unlock(&root->fs_info->alloc_mutex); 971 return 0; 972 fail: 973 WARN_ON(1); 974 #if 0 975 for (i =0; i < faili; i++) { 976 if (level == 0) { 977 u64 disk_bytenr; 978 btrfs_item_key_to_cpu(buf, &key, i); 979 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) 980 continue; 981 fi = btrfs_item_ptr(buf, i, 982 struct btrfs_file_extent_item); 983 if (btrfs_file_extent_type(buf, fi) == 984 BTRFS_FILE_EXTENT_INLINE) 985 continue; 986 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi); 987 if (disk_bytenr == 0) 988 continue; 989 err = btrfs_free_extent(trans, root, disk_bytenr, 990 btrfs_file_extent_disk_num_bytes(buf, 991 fi), 0); 992 BUG_ON(err); 993 } else { 994 bytenr = btrfs_node_blockptr(buf, i); 995 err = btrfs_free_extent(trans, root, bytenr, 996 btrfs_level_size(root, level - 1), 0); 997 BUG_ON(err); 998 } 999 } 1000 #endif 1001 mutex_unlock(&root->fs_info->alloc_mutex); 1002 return ret; 1003 } 1004 1005 static int write_one_cache_group(struct btrfs_trans_handle *trans, 1006 struct btrfs_root *root, 1007 struct btrfs_path *path, 1008 struct btrfs_block_group_cache *cache) 1009 { 1010 int ret; 1011 int pending_ret; 1012 struct btrfs_root *extent_root = root->fs_info->extent_root; 1013 unsigned long bi; 1014 struct extent_buffer *leaf; 1015 1016 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1); 1017 if (ret < 0) 1018 goto fail; 1019 BUG_ON(ret); 1020 1021 leaf = path->nodes[0]; 1022 bi = btrfs_item_ptr_offset(leaf, path->slots[0]); 1023 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item)); 1024 btrfs_mark_buffer_dirty(leaf); 1025 btrfs_release_path(extent_root, path); 1026 fail: 1027 finish_current_insert(trans, extent_root); 1028 pending_ret = del_pending_extents(trans, extent_root); 1029 if (ret) 1030 return ret; 1031 if (pending_ret) 1032 return pending_ret; 1033 return 0; 1034 1035 } 1036 1037 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans, 1038 struct btrfs_root *root) 1039 { 1040 struct extent_io_tree *block_group_cache; 1041 struct btrfs_block_group_cache *cache; 1042 int ret; 1043 int err = 0; 1044 int werr = 0; 1045 struct btrfs_path *path; 1046 u64 last = 0; 1047 u64 start; 1048 u64 end; 1049 u64 ptr; 1050 1051 block_group_cache = &root->fs_info->block_group_cache; 1052 path = btrfs_alloc_path(); 1053 if (!path) 1054 return -ENOMEM; 1055 1056 mutex_lock(&root->fs_info->alloc_mutex); 1057 while(1) { 1058 ret = find_first_extent_bit(block_group_cache, last, 1059 &start, &end, BLOCK_GROUP_DIRTY); 1060 if (ret) 1061 break; 1062 1063 last = end + 1; 1064 ret = get_state_private(block_group_cache, start, &ptr); 1065 if (ret) 1066 break; 1067 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr; 1068 err = write_one_cache_group(trans, root, 1069 path, cache); 1070 /* 1071 * if we fail to write the cache group, we want 1072 * to keep it marked dirty in hopes that a later 1073 * write will work 1074 */ 1075 if (err) { 1076 werr = err; 1077 continue; 1078 } 1079 clear_extent_bits(block_group_cache, start, end, 1080 BLOCK_GROUP_DIRTY, GFP_NOFS); 1081 } 1082 btrfs_free_path(path); 1083 mutex_unlock(&root->fs_info->alloc_mutex); 1084 return werr; 1085 } 1086 1087 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info, 1088 u64 flags) 1089 { 1090 struct list_head *head = &info->space_info; 1091 struct list_head *cur; 1092 struct btrfs_space_info *found; 1093 list_for_each(cur, head) { 1094 found = list_entry(cur, struct btrfs_space_info, list); 1095 if (found->flags == flags) 1096 return found; 1097 } 1098 return NULL; 1099 1100 } 1101 1102 static int update_space_info(struct btrfs_fs_info *info, u64 flags, 1103 u64 total_bytes, u64 bytes_used, 1104 struct btrfs_space_info **space_info) 1105 { 1106 struct btrfs_space_info *found; 1107 1108 found = __find_space_info(info, flags); 1109 if (found) { 1110 found->total_bytes += total_bytes; 1111 found->bytes_used += bytes_used; 1112 found->full = 0; 1113 WARN_ON(found->total_bytes < found->bytes_used); 1114 *space_info = found; 1115 return 0; 1116 } 1117 found = kmalloc(sizeof(*found), GFP_NOFS); 1118 if (!found) 1119 return -ENOMEM; 1120 1121 list_add(&found->list, &info->space_info); 1122 found->flags = flags; 1123 found->total_bytes = total_bytes; 1124 found->bytes_used = bytes_used; 1125 found->bytes_pinned = 0; 1126 found->full = 0; 1127 found->force_alloc = 0; 1128 *space_info = found; 1129 return 0; 1130 } 1131 1132 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags) 1133 { 1134 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 | 1135 BTRFS_BLOCK_GROUP_RAID1 | 1136 BTRFS_BLOCK_GROUP_RAID10 | 1137 BTRFS_BLOCK_GROUP_DUP); 1138 if (extra_flags) { 1139 if (flags & BTRFS_BLOCK_GROUP_DATA) 1140 fs_info->avail_data_alloc_bits |= extra_flags; 1141 if (flags & BTRFS_BLOCK_GROUP_METADATA) 1142 fs_info->avail_metadata_alloc_bits |= extra_flags; 1143 if (flags & BTRFS_BLOCK_GROUP_SYSTEM) 1144 fs_info->avail_system_alloc_bits |= extra_flags; 1145 } 1146 } 1147 1148 static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags) 1149 { 1150 u64 num_devices = root->fs_info->fs_devices->num_devices; 1151 1152 if (num_devices == 1) 1153 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0); 1154 if (num_devices < 4) 1155 flags &= ~BTRFS_BLOCK_GROUP_RAID10; 1156 1157 if ((flags & BTRFS_BLOCK_GROUP_DUP) && 1158 (flags & (BTRFS_BLOCK_GROUP_RAID1 | 1159 BTRFS_BLOCK_GROUP_RAID10))) { 1160 flags &= ~BTRFS_BLOCK_GROUP_DUP; 1161 } 1162 1163 if ((flags & BTRFS_BLOCK_GROUP_RAID1) && 1164 (flags & BTRFS_BLOCK_GROUP_RAID10)) { 1165 flags &= ~BTRFS_BLOCK_GROUP_RAID1; 1166 } 1167 1168 if ((flags & BTRFS_BLOCK_GROUP_RAID0) && 1169 ((flags & BTRFS_BLOCK_GROUP_RAID1) | 1170 (flags & BTRFS_BLOCK_GROUP_RAID10) | 1171 (flags & BTRFS_BLOCK_GROUP_DUP))) 1172 flags &= ~BTRFS_BLOCK_GROUP_RAID0; 1173 return flags; 1174 } 1175 1176 static int do_chunk_alloc(struct btrfs_trans_handle *trans, 1177 struct btrfs_root *extent_root, u64 alloc_bytes, 1178 u64 flags, int force) 1179 { 1180 struct btrfs_space_info *space_info; 1181 u64 thresh; 1182 u64 start; 1183 u64 num_bytes; 1184 int ret; 1185 1186 flags = reduce_alloc_profile(extent_root, flags); 1187 1188 space_info = __find_space_info(extent_root->fs_info, flags); 1189 if (!space_info) { 1190 ret = update_space_info(extent_root->fs_info, flags, 1191 0, 0, &space_info); 1192 BUG_ON(ret); 1193 } 1194 BUG_ON(!space_info); 1195 1196 if (space_info->force_alloc) { 1197 force = 1; 1198 space_info->force_alloc = 0; 1199 } 1200 if (space_info->full) 1201 goto out; 1202 1203 thresh = div_factor(space_info->total_bytes, 6); 1204 if (!force && 1205 (space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) < 1206 thresh) 1207 goto out; 1208 1209 mutex_lock(&extent_root->fs_info->chunk_mutex); 1210 ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags); 1211 if (ret == -ENOSPC) { 1212 printk("space info full %Lu\n", flags); 1213 space_info->full = 1; 1214 goto out; 1215 } 1216 BUG_ON(ret); 1217 1218 ret = btrfs_make_block_group(trans, extent_root, 0, flags, 1219 BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes); 1220 BUG_ON(ret); 1221 mutex_unlock(&extent_root->fs_info->chunk_mutex); 1222 out: 1223 return 0; 1224 } 1225 1226 static int update_block_group(struct btrfs_trans_handle *trans, 1227 struct btrfs_root *root, 1228 u64 bytenr, u64 num_bytes, int alloc, 1229 int mark_free) 1230 { 1231 struct btrfs_block_group_cache *cache; 1232 struct btrfs_fs_info *info = root->fs_info; 1233 u64 total = num_bytes; 1234 u64 old_val; 1235 u64 byte_in_group; 1236 u64 start; 1237 u64 end; 1238 1239 while(total) { 1240 cache = btrfs_lookup_block_group(info, bytenr); 1241 if (!cache) { 1242 return -1; 1243 } 1244 byte_in_group = bytenr - cache->key.objectid; 1245 WARN_ON(byte_in_group > cache->key.offset); 1246 start = cache->key.objectid; 1247 end = start + cache->key.offset - 1; 1248 set_extent_bits(&info->block_group_cache, start, end, 1249 BLOCK_GROUP_DIRTY, GFP_NOFS); 1250 1251 old_val = btrfs_block_group_used(&cache->item); 1252 num_bytes = min(total, cache->key.offset - byte_in_group); 1253 if (alloc) { 1254 old_val += num_bytes; 1255 cache->space_info->bytes_used += num_bytes; 1256 } else { 1257 old_val -= num_bytes; 1258 cache->space_info->bytes_used -= num_bytes; 1259 if (mark_free) { 1260 set_extent_dirty(&info->free_space_cache, 1261 bytenr, bytenr + num_bytes - 1, 1262 GFP_NOFS); 1263 } 1264 } 1265 btrfs_set_block_group_used(&cache->item, old_val); 1266 total -= num_bytes; 1267 bytenr += num_bytes; 1268 } 1269 return 0; 1270 } 1271 1272 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start) 1273 { 1274 u64 start; 1275 u64 end; 1276 int ret; 1277 ret = find_first_extent_bit(&root->fs_info->block_group_cache, 1278 search_start, &start, &end, 1279 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA | 1280 BLOCK_GROUP_SYSTEM); 1281 if (ret) 1282 return 0; 1283 return start; 1284 } 1285 1286 1287 static int update_pinned_extents(struct btrfs_root *root, 1288 u64 bytenr, u64 num, int pin) 1289 { 1290 u64 len; 1291 struct btrfs_block_group_cache *cache; 1292 struct btrfs_fs_info *fs_info = root->fs_info; 1293 1294 if (pin) { 1295 set_extent_dirty(&fs_info->pinned_extents, 1296 bytenr, bytenr + num - 1, GFP_NOFS); 1297 } else { 1298 clear_extent_dirty(&fs_info->pinned_extents, 1299 bytenr, bytenr + num - 1, GFP_NOFS); 1300 } 1301 while (num > 0) { 1302 cache = btrfs_lookup_block_group(fs_info, bytenr); 1303 if (!cache) { 1304 u64 first = first_logical_byte(root, bytenr); 1305 WARN_ON(first < bytenr); 1306 len = min(first - bytenr, num); 1307 } else { 1308 len = min(num, cache->key.offset - 1309 (bytenr - cache->key.objectid)); 1310 } 1311 if (pin) { 1312 if (cache) { 1313 cache->pinned += len; 1314 cache->space_info->bytes_pinned += len; 1315 } 1316 fs_info->total_pinned += len; 1317 } else { 1318 if (cache) { 1319 cache->pinned -= len; 1320 cache->space_info->bytes_pinned -= len; 1321 } 1322 fs_info->total_pinned -= len; 1323 } 1324 bytenr += len; 1325 num -= len; 1326 } 1327 return 0; 1328 } 1329 1330 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy) 1331 { 1332 u64 last = 0; 1333 u64 start; 1334 u64 end; 1335 struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents; 1336 int ret; 1337 1338 while(1) { 1339 ret = find_first_extent_bit(pinned_extents, last, 1340 &start, &end, EXTENT_DIRTY); 1341 if (ret) 1342 break; 1343 set_extent_dirty(copy, start, end, GFP_NOFS); 1344 last = end + 1; 1345 } 1346 return 0; 1347 } 1348 1349 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans, 1350 struct btrfs_root *root, 1351 struct extent_io_tree *unpin) 1352 { 1353 u64 start; 1354 u64 end; 1355 int ret; 1356 struct extent_io_tree *free_space_cache; 1357 free_space_cache = &root->fs_info->free_space_cache; 1358 1359 mutex_lock(&root->fs_info->alloc_mutex); 1360 while(1) { 1361 ret = find_first_extent_bit(unpin, 0, &start, &end, 1362 EXTENT_DIRTY); 1363 if (ret) 1364 break; 1365 update_pinned_extents(root, start, end + 1 - start, 0); 1366 clear_extent_dirty(unpin, start, end, GFP_NOFS); 1367 set_extent_dirty(free_space_cache, start, end, GFP_NOFS); 1368 } 1369 mutex_unlock(&root->fs_info->alloc_mutex); 1370 return 0; 1371 } 1372 1373 static int finish_current_insert(struct btrfs_trans_handle *trans, 1374 struct btrfs_root *extent_root) 1375 { 1376 u64 start; 1377 u64 end; 1378 struct btrfs_fs_info *info = extent_root->fs_info; 1379 struct extent_buffer *eb; 1380 struct btrfs_path *path; 1381 struct btrfs_key ins; 1382 struct btrfs_disk_key first; 1383 struct btrfs_extent_item extent_item; 1384 int ret; 1385 int level; 1386 int err = 0; 1387 1388 btrfs_set_stack_extent_refs(&extent_item, 1); 1389 btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY); 1390 path = btrfs_alloc_path(); 1391 1392 while(1) { 1393 ret = find_first_extent_bit(&info->extent_ins, 0, &start, 1394 &end, EXTENT_LOCKED); 1395 if (ret) 1396 break; 1397 1398 ins.objectid = start; 1399 ins.offset = end + 1 - start; 1400 err = btrfs_insert_item(trans, extent_root, &ins, 1401 &extent_item, sizeof(extent_item)); 1402 clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED, 1403 GFP_NOFS); 1404 eb = read_tree_block(extent_root, ins.objectid, ins.offset, 1405 trans->transid); 1406 btrfs_tree_lock(eb); 1407 level = btrfs_header_level(eb); 1408 if (level == 0) { 1409 btrfs_item_key(eb, &first, 0); 1410 } else { 1411 btrfs_node_key(eb, &first, 0); 1412 } 1413 btrfs_tree_unlock(eb); 1414 free_extent_buffer(eb); 1415 /* 1416 * the first key is just a hint, so the race we've created 1417 * against reading it is fine 1418 */ 1419 err = btrfs_insert_extent_backref(trans, extent_root, path, 1420 start, extent_root->root_key.objectid, 1421 0, level, 1422 btrfs_disk_key_objectid(&first)); 1423 BUG_ON(err); 1424 } 1425 btrfs_free_path(path); 1426 return 0; 1427 } 1428 1429 static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes, 1430 int pending) 1431 { 1432 int err = 0; 1433 1434 if (!pending) { 1435 #if 0 1436 struct extent_buffer *buf; 1437 buf = btrfs_find_tree_block(root, bytenr, num_bytes); 1438 if (buf) { 1439 if (!btrfs_try_tree_lock(buf) && 1440 btrfs_buffer_uptodate(buf, 0)) { 1441 u64 transid = 1442 root->fs_info->running_transaction->transid; 1443 u64 header_transid = 1444 btrfs_header_generation(buf); 1445 if (header_transid == transid && 1446 !btrfs_header_flag(buf, 1447 BTRFS_HEADER_FLAG_WRITTEN)) { 1448 clean_tree_block(NULL, root, buf); 1449 btrfs_tree_unlock(buf); 1450 free_extent_buffer(buf); 1451 return 1; 1452 } 1453 btrfs_tree_unlock(buf); 1454 } 1455 free_extent_buffer(buf); 1456 } 1457 #endif 1458 update_pinned_extents(root, bytenr, num_bytes, 1); 1459 } else { 1460 set_extent_bits(&root->fs_info->pending_del, 1461 bytenr, bytenr + num_bytes - 1, 1462 EXTENT_LOCKED, GFP_NOFS); 1463 } 1464 BUG_ON(err < 0); 1465 return 0; 1466 } 1467 1468 /* 1469 * remove an extent from the root, returns 0 on success 1470 */ 1471 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root 1472 *root, u64 bytenr, u64 num_bytes, 1473 u64 root_objectid, u64 ref_generation, 1474 u64 owner_objectid, u64 owner_offset, int pin, 1475 int mark_free) 1476 { 1477 struct btrfs_path *path; 1478 struct btrfs_key key; 1479 struct btrfs_fs_info *info = root->fs_info; 1480 struct btrfs_root *extent_root = info->extent_root; 1481 struct extent_buffer *leaf; 1482 int ret; 1483 int extent_slot = 0; 1484 int found_extent = 0; 1485 int num_to_del = 1; 1486 struct btrfs_extent_item *ei; 1487 u32 refs; 1488 1489 key.objectid = bytenr; 1490 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY); 1491 key.offset = num_bytes; 1492 path = btrfs_alloc_path(); 1493 if (!path) 1494 return -ENOMEM; 1495 1496 path->reada = 1; 1497 ret = lookup_extent_backref(trans, extent_root, path, 1498 bytenr, root_objectid, 1499 ref_generation, 1500 owner_objectid, owner_offset, 1); 1501 if (ret == 0) { 1502 struct btrfs_key found_key; 1503 extent_slot = path->slots[0]; 1504 while(extent_slot > 0) { 1505 extent_slot--; 1506 btrfs_item_key_to_cpu(path->nodes[0], &found_key, 1507 extent_slot); 1508 if (found_key.objectid != bytenr) 1509 break; 1510 if (found_key.type == BTRFS_EXTENT_ITEM_KEY && 1511 found_key.offset == num_bytes) { 1512 found_extent = 1; 1513 break; 1514 } 1515 if (path->slots[0] - extent_slot > 5) 1516 break; 1517 } 1518 if (!found_extent) 1519 ret = btrfs_del_item(trans, extent_root, path); 1520 } else { 1521 btrfs_print_leaf(extent_root, path->nodes[0]); 1522 WARN_ON(1); 1523 printk("Unable to find ref byte nr %Lu root %Lu " 1524 " gen %Lu owner %Lu offset %Lu\n", bytenr, 1525 root_objectid, ref_generation, owner_objectid, 1526 owner_offset); 1527 } 1528 if (!found_extent) { 1529 btrfs_release_path(extent_root, path); 1530 ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1); 1531 if (ret < 0) 1532 return ret; 1533 BUG_ON(ret); 1534 extent_slot = path->slots[0]; 1535 } 1536 1537 leaf = path->nodes[0]; 1538 ei = btrfs_item_ptr(leaf, extent_slot, 1539 struct btrfs_extent_item); 1540 refs = btrfs_extent_refs(leaf, ei); 1541 BUG_ON(refs == 0); 1542 refs -= 1; 1543 btrfs_set_extent_refs(leaf, ei, refs); 1544 1545 btrfs_mark_buffer_dirty(leaf); 1546 1547 if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) { 1548 /* if the back ref and the extent are next to each other 1549 * they get deleted below in one shot 1550 */ 1551 path->slots[0] = extent_slot; 1552 num_to_del = 2; 1553 } else if (found_extent) { 1554 /* otherwise delete the extent back ref */ 1555 ret = btrfs_del_item(trans, extent_root, path); 1556 BUG_ON(ret); 1557 /* if refs are 0, we need to setup the path for deletion */ 1558 if (refs == 0) { 1559 btrfs_release_path(extent_root, path); 1560 ret = btrfs_search_slot(trans, extent_root, &key, path, 1561 -1, 1); 1562 if (ret < 0) 1563 return ret; 1564 BUG_ON(ret); 1565 } 1566 } 1567 1568 if (refs == 0) { 1569 u64 super_used; 1570 u64 root_used; 1571 1572 if (pin) { 1573 ret = pin_down_bytes(root, bytenr, num_bytes, 0); 1574 if (ret > 0) 1575 mark_free = 1; 1576 BUG_ON(ret < 0); 1577 } 1578 1579 /* block accounting for super block */ 1580 spin_lock_irq(&info->delalloc_lock); 1581 super_used = btrfs_super_bytes_used(&info->super_copy); 1582 btrfs_set_super_bytes_used(&info->super_copy, 1583 super_used - num_bytes); 1584 spin_unlock_irq(&info->delalloc_lock); 1585 1586 /* block accounting for root item */ 1587 root_used = btrfs_root_used(&root->root_item); 1588 btrfs_set_root_used(&root->root_item, 1589 root_used - num_bytes); 1590 ret = btrfs_del_items(trans, extent_root, path, path->slots[0], 1591 num_to_del); 1592 if (ret) { 1593 return ret; 1594 } 1595 ret = update_block_group(trans, root, bytenr, num_bytes, 0, 1596 mark_free); 1597 BUG_ON(ret); 1598 } 1599 btrfs_free_path(path); 1600 finish_current_insert(trans, extent_root); 1601 return ret; 1602 } 1603 1604 /* 1605 * find all the blocks marked as pending in the radix tree and remove 1606 * them from the extent map 1607 */ 1608 static int del_pending_extents(struct btrfs_trans_handle *trans, struct 1609 btrfs_root *extent_root) 1610 { 1611 int ret; 1612 int err = 0; 1613 u64 start; 1614 u64 end; 1615 struct extent_io_tree *pending_del; 1616 struct extent_io_tree *pinned_extents; 1617 1618 pending_del = &extent_root->fs_info->pending_del; 1619 pinned_extents = &extent_root->fs_info->pinned_extents; 1620 1621 while(1) { 1622 ret = find_first_extent_bit(pending_del, 0, &start, &end, 1623 EXTENT_LOCKED); 1624 if (ret) 1625 break; 1626 update_pinned_extents(extent_root, start, end + 1 - start, 1); 1627 clear_extent_bits(pending_del, start, end, EXTENT_LOCKED, 1628 GFP_NOFS); 1629 ret = __free_extent(trans, extent_root, 1630 start, end + 1 - start, 1631 extent_root->root_key.objectid, 1632 0, 0, 0, 0, 0); 1633 if (ret) 1634 err = ret; 1635 } 1636 return err; 1637 } 1638 1639 /* 1640 * remove an extent from the root, returns 0 on success 1641 */ 1642 static int __btrfs_free_extent(struct btrfs_trans_handle *trans, 1643 struct btrfs_root *root, u64 bytenr, 1644 u64 num_bytes, u64 root_objectid, 1645 u64 ref_generation, u64 owner_objectid, 1646 u64 owner_offset, int pin) 1647 { 1648 struct btrfs_root *extent_root = root->fs_info->extent_root; 1649 int pending_ret; 1650 int ret; 1651 1652 WARN_ON(num_bytes < root->sectorsize); 1653 if (!root->ref_cows) 1654 ref_generation = 0; 1655 1656 if (root == extent_root) { 1657 pin_down_bytes(root, bytenr, num_bytes, 1); 1658 return 0; 1659 } 1660 ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid, 1661 ref_generation, owner_objectid, owner_offset, 1662 pin, pin == 0); 1663 pending_ret = del_pending_extents(trans, root->fs_info->extent_root); 1664 return ret ? ret : pending_ret; 1665 } 1666 1667 int btrfs_free_extent(struct btrfs_trans_handle *trans, 1668 struct btrfs_root *root, u64 bytenr, 1669 u64 num_bytes, u64 root_objectid, 1670 u64 ref_generation, u64 owner_objectid, 1671 u64 owner_offset, int pin) 1672 { 1673 int ret; 1674 1675 maybe_lock_mutex(root); 1676 ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, 1677 root_objectid, ref_generation, 1678 owner_objectid, owner_offset, pin); 1679 maybe_unlock_mutex(root); 1680 return ret; 1681 } 1682 1683 static u64 stripe_align(struct btrfs_root *root, u64 val) 1684 { 1685 u64 mask = ((u64)root->stripesize - 1); 1686 u64 ret = (val + mask) & ~mask; 1687 return ret; 1688 } 1689 1690 /* 1691 * walks the btree of allocated extents and find a hole of a given size. 1692 * The key ins is changed to record the hole: 1693 * ins->objectid == block start 1694 * ins->flags = BTRFS_EXTENT_ITEM_KEY 1695 * ins->offset == number of blocks 1696 * Any available blocks before search_start are skipped. 1697 */ 1698 static int noinline find_free_extent(struct btrfs_trans_handle *trans, 1699 struct btrfs_root *orig_root, 1700 u64 num_bytes, u64 empty_size, 1701 u64 search_start, u64 search_end, 1702 u64 hint_byte, struct btrfs_key *ins, 1703 u64 exclude_start, u64 exclude_nr, 1704 int data) 1705 { 1706 int ret; 1707 u64 orig_search_start; 1708 struct btrfs_root * root = orig_root->fs_info->extent_root; 1709 struct btrfs_fs_info *info = root->fs_info; 1710 u64 total_needed = num_bytes; 1711 u64 *last_ptr = NULL; 1712 struct btrfs_block_group_cache *block_group; 1713 int full_scan = 0; 1714 int wrapped = 0; 1715 int chunk_alloc_done = 0; 1716 int empty_cluster = 2 * 1024 * 1024; 1717 int allowed_chunk_alloc = 0; 1718 1719 WARN_ON(num_bytes < root->sectorsize); 1720 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY); 1721 1722 if (orig_root->ref_cows || empty_size) 1723 allowed_chunk_alloc = 1; 1724 1725 if (data & BTRFS_BLOCK_GROUP_METADATA) { 1726 last_ptr = &root->fs_info->last_alloc; 1727 empty_cluster = 256 * 1024; 1728 } 1729 1730 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) { 1731 last_ptr = &root->fs_info->last_data_alloc; 1732 } 1733 1734 if (last_ptr) { 1735 if (*last_ptr) 1736 hint_byte = *last_ptr; 1737 else { 1738 empty_size += empty_cluster; 1739 } 1740 } 1741 1742 search_start = max(search_start, first_logical_byte(root, 0)); 1743 orig_search_start = search_start; 1744 1745 if (search_end == (u64)-1) 1746 search_end = btrfs_super_total_bytes(&info->super_copy); 1747 1748 if (hint_byte) { 1749 block_group = btrfs_lookup_first_block_group(info, hint_byte); 1750 if (!block_group) 1751 hint_byte = search_start; 1752 block_group = __btrfs_find_block_group(root, block_group, 1753 hint_byte, data, 1); 1754 if (last_ptr && *last_ptr == 0 && block_group) 1755 hint_byte = block_group->key.objectid; 1756 } else { 1757 block_group = __btrfs_find_block_group(root, 1758 trans->block_group, 1759 search_start, data, 1); 1760 } 1761 search_start = max(search_start, hint_byte); 1762 1763 total_needed += empty_size; 1764 1765 check_failed: 1766 if (!block_group) { 1767 block_group = btrfs_lookup_first_block_group(info, 1768 search_start); 1769 if (!block_group) 1770 block_group = btrfs_lookup_first_block_group(info, 1771 orig_search_start); 1772 } 1773 if (full_scan && !chunk_alloc_done) { 1774 if (allowed_chunk_alloc) { 1775 do_chunk_alloc(trans, root, 1776 num_bytes + 2 * 1024 * 1024, data, 1); 1777 allowed_chunk_alloc = 0; 1778 } else if (block_group && block_group_bits(block_group, data)) { 1779 block_group->space_info->force_alloc = 1; 1780 } 1781 chunk_alloc_done = 1; 1782 } 1783 ret = find_search_start(root, &block_group, &search_start, 1784 total_needed, data); 1785 if (ret == -ENOSPC && last_ptr && *last_ptr) { 1786 *last_ptr = 0; 1787 block_group = btrfs_lookup_first_block_group(info, 1788 orig_search_start); 1789 search_start = orig_search_start; 1790 ret = find_search_start(root, &block_group, &search_start, 1791 total_needed, data); 1792 } 1793 if (ret == -ENOSPC) 1794 goto enospc; 1795 if (ret) 1796 goto error; 1797 1798 if (last_ptr && *last_ptr && search_start != *last_ptr) { 1799 *last_ptr = 0; 1800 if (!empty_size) { 1801 empty_size += empty_cluster; 1802 total_needed += empty_size; 1803 } 1804 block_group = btrfs_lookup_first_block_group(info, 1805 orig_search_start); 1806 search_start = orig_search_start; 1807 ret = find_search_start(root, &block_group, 1808 &search_start, total_needed, data); 1809 if (ret == -ENOSPC) 1810 goto enospc; 1811 if (ret) 1812 goto error; 1813 } 1814 1815 search_start = stripe_align(root, search_start); 1816 ins->objectid = search_start; 1817 ins->offset = num_bytes; 1818 1819 if (ins->objectid + num_bytes >= search_end) 1820 goto enospc; 1821 1822 if (ins->objectid + num_bytes > 1823 block_group->key.objectid + block_group->key.offset) { 1824 search_start = block_group->key.objectid + 1825 block_group->key.offset; 1826 goto new_group; 1827 } 1828 1829 if (test_range_bit(&info->extent_ins, ins->objectid, 1830 ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) { 1831 search_start = ins->objectid + num_bytes; 1832 goto new_group; 1833 } 1834 1835 if (test_range_bit(&info->pinned_extents, ins->objectid, 1836 ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) { 1837 search_start = ins->objectid + num_bytes; 1838 goto new_group; 1839 } 1840 1841 if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start && 1842 ins->objectid < exclude_start + exclude_nr)) { 1843 search_start = exclude_start + exclude_nr; 1844 goto new_group; 1845 } 1846 1847 if (!(data & BTRFS_BLOCK_GROUP_DATA)) { 1848 block_group = btrfs_lookup_block_group(info, ins->objectid); 1849 if (block_group) 1850 trans->block_group = block_group; 1851 } 1852 ins->offset = num_bytes; 1853 if (last_ptr) { 1854 *last_ptr = ins->objectid + ins->offset; 1855 if (*last_ptr == 1856 btrfs_super_total_bytes(&root->fs_info->super_copy)) { 1857 *last_ptr = 0; 1858 } 1859 } 1860 return 0; 1861 1862 new_group: 1863 if (search_start + num_bytes >= search_end) { 1864 enospc: 1865 search_start = orig_search_start; 1866 if (full_scan) { 1867 ret = -ENOSPC; 1868 goto error; 1869 } 1870 if (wrapped) { 1871 if (!full_scan) 1872 total_needed -= empty_size; 1873 full_scan = 1; 1874 } else 1875 wrapped = 1; 1876 } 1877 block_group = btrfs_lookup_first_block_group(info, search_start); 1878 cond_resched(); 1879 block_group = __btrfs_find_block_group(root, block_group, 1880 search_start, data, 0); 1881 goto check_failed; 1882 1883 error: 1884 return ret; 1885 } 1886 1887 /* 1888 * finds a free extent and does all the dirty work required for allocation 1889 * returns the key for the extent through ins, and a tree buffer for 1890 * the first block of the extent through buf. 1891 * 1892 * returns 0 if everything worked, non-zero otherwise. 1893 */ 1894 int btrfs_alloc_extent(struct btrfs_trans_handle *trans, 1895 struct btrfs_root *root, 1896 u64 num_bytes, u64 min_alloc_size, 1897 u64 root_objectid, u64 ref_generation, 1898 u64 owner, u64 owner_offset, 1899 u64 empty_size, u64 hint_byte, 1900 u64 search_end, struct btrfs_key *ins, u64 data) 1901 { 1902 int ret; 1903 int pending_ret; 1904 u64 super_used; 1905 u64 root_used; 1906 u64 search_start = 0; 1907 u64 alloc_profile; 1908 u32 sizes[2]; 1909 struct btrfs_fs_info *info = root->fs_info; 1910 struct btrfs_root *extent_root = info->extent_root; 1911 struct btrfs_extent_item *extent_item; 1912 struct btrfs_extent_ref *ref; 1913 struct btrfs_path *path; 1914 struct btrfs_key keys[2]; 1915 1916 maybe_lock_mutex(root); 1917 1918 if (data) { 1919 alloc_profile = info->avail_data_alloc_bits & 1920 info->data_alloc_profile; 1921 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile; 1922 } else if (root == root->fs_info->chunk_root) { 1923 alloc_profile = info->avail_system_alloc_bits & 1924 info->system_alloc_profile; 1925 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile; 1926 } else { 1927 alloc_profile = info->avail_metadata_alloc_bits & 1928 info->metadata_alloc_profile; 1929 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile; 1930 } 1931 again: 1932 data = reduce_alloc_profile(root, data); 1933 /* 1934 * the only place that sets empty_size is btrfs_realloc_node, which 1935 * is not called recursively on allocations 1936 */ 1937 if (empty_size || root->ref_cows) { 1938 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) { 1939 ret = do_chunk_alloc(trans, root->fs_info->extent_root, 1940 2 * 1024 * 1024, 1941 BTRFS_BLOCK_GROUP_METADATA | 1942 (info->metadata_alloc_profile & 1943 info->avail_metadata_alloc_bits), 0); 1944 BUG_ON(ret); 1945 } 1946 ret = do_chunk_alloc(trans, root->fs_info->extent_root, 1947 num_bytes + 2 * 1024 * 1024, data, 0); 1948 BUG_ON(ret); 1949 } 1950 1951 WARN_ON(num_bytes < root->sectorsize); 1952 ret = find_free_extent(trans, root, num_bytes, empty_size, 1953 search_start, search_end, hint_byte, ins, 1954 trans->alloc_exclude_start, 1955 trans->alloc_exclude_nr, data); 1956 1957 if (ret == -ENOSPC && num_bytes > min_alloc_size) { 1958 num_bytes = num_bytes >> 1; 1959 num_bytes = max(num_bytes, min_alloc_size); 1960 do_chunk_alloc(trans, root->fs_info->extent_root, 1961 num_bytes, data, 1); 1962 goto again; 1963 } 1964 if (ret) { 1965 printk("allocation failed flags %Lu\n", data); 1966 } 1967 if (ret) { 1968 BUG(); 1969 goto out; 1970 } 1971 1972 /* block accounting for super block */ 1973 spin_lock_irq(&info->delalloc_lock); 1974 super_used = btrfs_super_bytes_used(&info->super_copy); 1975 btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes); 1976 spin_unlock_irq(&info->delalloc_lock); 1977 1978 /* block accounting for root item */ 1979 root_used = btrfs_root_used(&root->root_item); 1980 btrfs_set_root_used(&root->root_item, root_used + num_bytes); 1981 1982 clear_extent_dirty(&root->fs_info->free_space_cache, 1983 ins->objectid, ins->objectid + ins->offset - 1, 1984 GFP_NOFS); 1985 1986 if (root == extent_root) { 1987 set_extent_bits(&root->fs_info->extent_ins, ins->objectid, 1988 ins->objectid + ins->offset - 1, 1989 EXTENT_LOCKED, GFP_NOFS); 1990 goto update_block; 1991 } 1992 1993 WARN_ON(trans->alloc_exclude_nr); 1994 trans->alloc_exclude_start = ins->objectid; 1995 trans->alloc_exclude_nr = ins->offset; 1996 1997 memcpy(&keys[0], ins, sizeof(*ins)); 1998 keys[1].offset = hash_extent_ref(root_objectid, ref_generation, 1999 owner, owner_offset); 2000 keys[1].objectid = ins->objectid; 2001 keys[1].type = BTRFS_EXTENT_REF_KEY; 2002 sizes[0] = sizeof(*extent_item); 2003 sizes[1] = sizeof(*ref); 2004 2005 path = btrfs_alloc_path(); 2006 BUG_ON(!path); 2007 2008 ret = btrfs_insert_empty_items(trans, extent_root, path, keys, 2009 sizes, 2); 2010 2011 BUG_ON(ret); 2012 extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0], 2013 struct btrfs_extent_item); 2014 btrfs_set_extent_refs(path->nodes[0], extent_item, 1); 2015 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1, 2016 struct btrfs_extent_ref); 2017 2018 btrfs_set_ref_root(path->nodes[0], ref, root_objectid); 2019 btrfs_set_ref_generation(path->nodes[0], ref, ref_generation); 2020 btrfs_set_ref_objectid(path->nodes[0], ref, owner); 2021 btrfs_set_ref_offset(path->nodes[0], ref, owner_offset); 2022 2023 btrfs_mark_buffer_dirty(path->nodes[0]); 2024 2025 trans->alloc_exclude_start = 0; 2026 trans->alloc_exclude_nr = 0; 2027 btrfs_free_path(path); 2028 finish_current_insert(trans, extent_root); 2029 pending_ret = del_pending_extents(trans, extent_root); 2030 2031 if (ret) 2032 goto out; 2033 if (pending_ret) { 2034 ret = pending_ret; 2035 goto out; 2036 } 2037 2038 update_block: 2039 ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0); 2040 if (ret) { 2041 printk("update block group failed for %Lu %Lu\n", 2042 ins->objectid, ins->offset); 2043 BUG(); 2044 } 2045 out: 2046 maybe_unlock_mutex(root); 2047 return ret; 2048 } 2049 /* 2050 * helper function to allocate a block for a given tree 2051 * returns the tree buffer or NULL. 2052 */ 2053 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans, 2054 struct btrfs_root *root, 2055 u32 blocksize, 2056 u64 root_objectid, 2057 u64 ref_generation, 2058 u64 first_objectid, 2059 int level, 2060 u64 hint, 2061 u64 empty_size) 2062 { 2063 struct btrfs_key ins; 2064 int ret; 2065 struct extent_buffer *buf; 2066 2067 ret = btrfs_alloc_extent(trans, root, blocksize, blocksize, 2068 root_objectid, ref_generation, 2069 level, first_objectid, empty_size, hint, 2070 (u64)-1, &ins, 0); 2071 if (ret) { 2072 BUG_ON(ret > 0); 2073 return ERR_PTR(ret); 2074 } 2075 buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize); 2076 if (!buf) { 2077 btrfs_free_extent(trans, root, ins.objectid, blocksize, 2078 root->root_key.objectid, ref_generation, 2079 0, 0, 0); 2080 return ERR_PTR(-ENOMEM); 2081 } 2082 btrfs_set_header_generation(buf, trans->transid); 2083 btrfs_tree_lock(buf); 2084 clean_tree_block(trans, root, buf); 2085 btrfs_set_buffer_uptodate(buf); 2086 2087 if (PageDirty(buf->first_page)) { 2088 printk("page %lu dirty\n", buf->first_page->index); 2089 WARN_ON(1); 2090 } 2091 2092 set_extent_dirty(&trans->transaction->dirty_pages, buf->start, 2093 buf->start + buf->len - 1, GFP_NOFS); 2094 if (!btrfs_test_opt(root, SSD)) 2095 btrfs_set_buffer_defrag(buf); 2096 trans->blocks_used++; 2097 return buf; 2098 } 2099 2100 static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans, 2101 struct btrfs_root *root, 2102 struct extent_buffer *leaf) 2103 { 2104 u64 leaf_owner; 2105 u64 leaf_generation; 2106 struct btrfs_key key; 2107 struct btrfs_file_extent_item *fi; 2108 int i; 2109 int nritems; 2110 int ret; 2111 2112 BUG_ON(!btrfs_is_leaf(leaf)); 2113 nritems = btrfs_header_nritems(leaf); 2114 leaf_owner = btrfs_header_owner(leaf); 2115 leaf_generation = btrfs_header_generation(leaf); 2116 2117 for (i = 0; i < nritems; i++) { 2118 u64 disk_bytenr; 2119 2120 btrfs_item_key_to_cpu(leaf, &key, i); 2121 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY) 2122 continue; 2123 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item); 2124 if (btrfs_file_extent_type(leaf, fi) == 2125 BTRFS_FILE_EXTENT_INLINE) 2126 continue; 2127 /* 2128 * FIXME make sure to insert a trans record that 2129 * repeats the snapshot del on crash 2130 */ 2131 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi); 2132 if (disk_bytenr == 0) 2133 continue; 2134 ret = __btrfs_free_extent(trans, root, disk_bytenr, 2135 btrfs_file_extent_disk_num_bytes(leaf, fi), 2136 leaf_owner, leaf_generation, 2137 key.objectid, key.offset, 0); 2138 BUG_ON(ret); 2139 } 2140 return 0; 2141 } 2142 2143 static void noinline reada_walk_down(struct btrfs_root *root, 2144 struct extent_buffer *node, 2145 int slot) 2146 { 2147 u64 bytenr; 2148 u64 last = 0; 2149 u32 nritems; 2150 u32 refs; 2151 u32 blocksize; 2152 int ret; 2153 int i; 2154 int level; 2155 int skipped = 0; 2156 2157 nritems = btrfs_header_nritems(node); 2158 level = btrfs_header_level(node); 2159 if (level) 2160 return; 2161 2162 for (i = slot; i < nritems && skipped < 32; i++) { 2163 bytenr = btrfs_node_blockptr(node, i); 2164 if (last && ((bytenr > last && bytenr - last > 32 * 1024) || 2165 (last > bytenr && last - bytenr > 32 * 1024))) { 2166 skipped++; 2167 continue; 2168 } 2169 blocksize = btrfs_level_size(root, level - 1); 2170 if (i != slot) { 2171 ret = lookup_extent_ref(NULL, root, bytenr, 2172 blocksize, &refs); 2173 BUG_ON(ret); 2174 if (refs != 1) { 2175 skipped++; 2176 continue; 2177 } 2178 } 2179 mutex_unlock(&root->fs_info->alloc_mutex); 2180 ret = readahead_tree_block(root, bytenr, blocksize, 2181 btrfs_node_ptr_generation(node, i)); 2182 last = bytenr + blocksize; 2183 cond_resched(); 2184 mutex_lock(&root->fs_info->alloc_mutex); 2185 if (ret) 2186 break; 2187 } 2188 } 2189 2190 /* 2191 * helper function for drop_snapshot, this walks down the tree dropping ref 2192 * counts as it goes. 2193 */ 2194 static int noinline walk_down_tree(struct btrfs_trans_handle *trans, 2195 struct btrfs_root *root, 2196 struct btrfs_path *path, int *level) 2197 { 2198 u64 root_owner; 2199 u64 root_gen; 2200 u64 bytenr; 2201 u64 ptr_gen; 2202 struct extent_buffer *next; 2203 struct extent_buffer *cur; 2204 struct extent_buffer *parent; 2205 u32 blocksize; 2206 int ret; 2207 u32 refs; 2208 2209 mutex_lock(&root->fs_info->alloc_mutex); 2210 2211 WARN_ON(*level < 0); 2212 WARN_ON(*level >= BTRFS_MAX_LEVEL); 2213 ret = lookup_extent_ref(trans, root, 2214 path->nodes[*level]->start, 2215 path->nodes[*level]->len, &refs); 2216 BUG_ON(ret); 2217 if (refs > 1) 2218 goto out; 2219 2220 /* 2221 * walk down to the last node level and free all the leaves 2222 */ 2223 while(*level >= 0) { 2224 WARN_ON(*level < 0); 2225 WARN_ON(*level >= BTRFS_MAX_LEVEL); 2226 cur = path->nodes[*level]; 2227 2228 if (btrfs_header_level(cur) != *level) 2229 WARN_ON(1); 2230 2231 if (path->slots[*level] >= 2232 btrfs_header_nritems(cur)) 2233 break; 2234 if (*level == 0) { 2235 ret = drop_leaf_ref(trans, root, cur); 2236 BUG_ON(ret); 2237 break; 2238 } 2239 bytenr = btrfs_node_blockptr(cur, path->slots[*level]); 2240 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]); 2241 blocksize = btrfs_level_size(root, *level - 1); 2242 2243 ret = lookup_extent_ref(trans, root, bytenr, blocksize, &refs); 2244 BUG_ON(ret); 2245 if (refs != 1) { 2246 parent = path->nodes[*level]; 2247 root_owner = btrfs_header_owner(parent); 2248 root_gen = btrfs_header_generation(parent); 2249 path->slots[*level]++; 2250 ret = __btrfs_free_extent(trans, root, bytenr, 2251 blocksize, root_owner, 2252 root_gen, 0, 0, 1); 2253 BUG_ON(ret); 2254 continue; 2255 } 2256 next = btrfs_find_tree_block(root, bytenr, blocksize); 2257 if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) { 2258 free_extent_buffer(next); 2259 reada_walk_down(root, cur, path->slots[*level]); 2260 2261 mutex_unlock(&root->fs_info->alloc_mutex); 2262 next = read_tree_block(root, bytenr, blocksize, 2263 ptr_gen); 2264 mutex_lock(&root->fs_info->alloc_mutex); 2265 2266 /* we've dropped the lock, double check */ 2267 ret = lookup_extent_ref(trans, root, bytenr, 2268 blocksize, &refs); 2269 BUG_ON(ret); 2270 if (refs != 1) { 2271 parent = path->nodes[*level]; 2272 root_owner = btrfs_header_owner(parent); 2273 root_gen = btrfs_header_generation(parent); 2274 2275 path->slots[*level]++; 2276 free_extent_buffer(next); 2277 ret = __btrfs_free_extent(trans, root, bytenr, 2278 blocksize, 2279 root_owner, 2280 root_gen, 0, 0, 1); 2281 BUG_ON(ret); 2282 continue; 2283 } 2284 } 2285 WARN_ON(*level <= 0); 2286 if (path->nodes[*level-1]) 2287 free_extent_buffer(path->nodes[*level-1]); 2288 path->nodes[*level-1] = next; 2289 *level = btrfs_header_level(next); 2290 path->slots[*level] = 0; 2291 } 2292 out: 2293 WARN_ON(*level < 0); 2294 WARN_ON(*level >= BTRFS_MAX_LEVEL); 2295 2296 if (path->nodes[*level] == root->node) { 2297 root_owner = root->root_key.objectid; 2298 parent = path->nodes[*level]; 2299 } else { 2300 parent = path->nodes[*level + 1]; 2301 root_owner = btrfs_header_owner(parent); 2302 } 2303 2304 root_gen = btrfs_header_generation(parent); 2305 ret = __btrfs_free_extent(trans, root, path->nodes[*level]->start, 2306 path->nodes[*level]->len, 2307 root_owner, root_gen, 0, 0, 1); 2308 free_extent_buffer(path->nodes[*level]); 2309 path->nodes[*level] = NULL; 2310 *level += 1; 2311 BUG_ON(ret); 2312 mutex_unlock(&root->fs_info->alloc_mutex); 2313 return 0; 2314 } 2315 2316 /* 2317 * helper for dropping snapshots. This walks back up the tree in the path 2318 * to find the first node higher up where we haven't yet gone through 2319 * all the slots 2320 */ 2321 static int noinline walk_up_tree(struct btrfs_trans_handle *trans, 2322 struct btrfs_root *root, 2323 struct btrfs_path *path, int *level) 2324 { 2325 u64 root_owner; 2326 u64 root_gen; 2327 struct btrfs_root_item *root_item = &root->root_item; 2328 int i; 2329 int slot; 2330 int ret; 2331 2332 for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) { 2333 slot = path->slots[i]; 2334 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) { 2335 struct extent_buffer *node; 2336 struct btrfs_disk_key disk_key; 2337 node = path->nodes[i]; 2338 path->slots[i]++; 2339 *level = i; 2340 WARN_ON(*level == 0); 2341 btrfs_node_key(node, &disk_key, path->slots[i]); 2342 memcpy(&root_item->drop_progress, 2343 &disk_key, sizeof(disk_key)); 2344 root_item->drop_level = i; 2345 return 0; 2346 } else { 2347 if (path->nodes[*level] == root->node) { 2348 root_owner = root->root_key.objectid; 2349 root_gen = 2350 btrfs_header_generation(path->nodes[*level]); 2351 } else { 2352 struct extent_buffer *node; 2353 node = path->nodes[*level + 1]; 2354 root_owner = btrfs_header_owner(node); 2355 root_gen = btrfs_header_generation(node); 2356 } 2357 ret = btrfs_free_extent(trans, root, 2358 path->nodes[*level]->start, 2359 path->nodes[*level]->len, 2360 root_owner, root_gen, 0, 0, 1); 2361 BUG_ON(ret); 2362 free_extent_buffer(path->nodes[*level]); 2363 path->nodes[*level] = NULL; 2364 *level = i + 1; 2365 } 2366 } 2367 return 1; 2368 } 2369 2370 /* 2371 * drop the reference count on the tree rooted at 'snap'. This traverses 2372 * the tree freeing any blocks that have a ref count of zero after being 2373 * decremented. 2374 */ 2375 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root 2376 *root) 2377 { 2378 int ret = 0; 2379 int wret; 2380 int level; 2381 struct btrfs_path *path; 2382 int i; 2383 int orig_level; 2384 struct btrfs_root_item *root_item = &root->root_item; 2385 2386 WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex)); 2387 path = btrfs_alloc_path(); 2388 BUG_ON(!path); 2389 2390 level = btrfs_header_level(root->node); 2391 orig_level = level; 2392 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { 2393 path->nodes[level] = root->node; 2394 extent_buffer_get(root->node); 2395 path->slots[level] = 0; 2396 } else { 2397 struct btrfs_key key; 2398 struct btrfs_disk_key found_key; 2399 struct extent_buffer *node; 2400 2401 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); 2402 level = root_item->drop_level; 2403 path->lowest_level = level; 2404 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2405 if (wret < 0) { 2406 ret = wret; 2407 goto out; 2408 } 2409 node = path->nodes[level]; 2410 btrfs_node_key(node, &found_key, path->slots[level]); 2411 WARN_ON(memcmp(&found_key, &root_item->drop_progress, 2412 sizeof(found_key))); 2413 for (i = 0; i < BTRFS_MAX_LEVEL; i++) { 2414 if (path->nodes[i] && path->locks[i]) { 2415 path->locks[i] = 0; 2416 btrfs_tree_unlock(path->nodes[i]); 2417 } 2418 } 2419 } 2420 while(1) { 2421 wret = walk_down_tree(trans, root, path, &level); 2422 if (wret > 0) 2423 break; 2424 if (wret < 0) 2425 ret = wret; 2426 2427 wret = walk_up_tree(trans, root, path, &level); 2428 if (wret > 0) 2429 break; 2430 if (wret < 0) 2431 ret = wret; 2432 ret = -EAGAIN; 2433 break; 2434 } 2435 for (i = 0; i <= orig_level; i++) { 2436 if (path->nodes[i]) { 2437 free_extent_buffer(path->nodes[i]); 2438 path->nodes[i] = NULL; 2439 } 2440 } 2441 out: 2442 btrfs_free_path(path); 2443 return ret; 2444 } 2445 2446 int btrfs_free_block_groups(struct btrfs_fs_info *info) 2447 { 2448 u64 start; 2449 u64 end; 2450 u64 ptr; 2451 int ret; 2452 2453 mutex_lock(&info->alloc_mutex); 2454 while(1) { 2455 ret = find_first_extent_bit(&info->block_group_cache, 0, 2456 &start, &end, (unsigned int)-1); 2457 if (ret) 2458 break; 2459 ret = get_state_private(&info->block_group_cache, start, &ptr); 2460 if (!ret) 2461 kfree((void *)(unsigned long)ptr); 2462 clear_extent_bits(&info->block_group_cache, start, 2463 end, (unsigned int)-1, GFP_NOFS); 2464 } 2465 while(1) { 2466 ret = find_first_extent_bit(&info->free_space_cache, 0, 2467 &start, &end, EXTENT_DIRTY); 2468 if (ret) 2469 break; 2470 clear_extent_dirty(&info->free_space_cache, start, 2471 end, GFP_NOFS); 2472 } 2473 mutex_unlock(&info->alloc_mutex); 2474 return 0; 2475 } 2476 2477 static unsigned long calc_ra(unsigned long start, unsigned long last, 2478 unsigned long nr) 2479 { 2480 return min(last, start + nr - 1); 2481 } 2482 2483 static int noinline relocate_inode_pages(struct inode *inode, u64 start, 2484 u64 len) 2485 { 2486 u64 page_start; 2487 u64 page_end; 2488 unsigned long last_index; 2489 unsigned long i; 2490 struct page *page; 2491 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; 2492 struct file_ra_state *ra; 2493 unsigned long total_read = 0; 2494 unsigned long ra_pages; 2495 struct btrfs_trans_handle *trans; 2496 2497 ra = kzalloc(sizeof(*ra), GFP_NOFS); 2498 2499 mutex_lock(&inode->i_mutex); 2500 i = start >> PAGE_CACHE_SHIFT; 2501 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT; 2502 2503 ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages; 2504 2505 file_ra_state_init(ra, inode->i_mapping); 2506 2507 for (; i <= last_index; i++) { 2508 if (total_read % ra_pages == 0) { 2509 btrfs_force_ra(inode->i_mapping, ra, NULL, i, 2510 calc_ra(i, last_index, ra_pages)); 2511 } 2512 total_read++; 2513 if (((u64)i << PAGE_CACHE_SHIFT) > inode->i_size) 2514 goto truncate_racing; 2515 2516 page = grab_cache_page(inode->i_mapping, i); 2517 if (!page) { 2518 goto out_unlock; 2519 } 2520 if (!PageUptodate(page)) { 2521 btrfs_readpage(NULL, page); 2522 lock_page(page); 2523 if (!PageUptodate(page)) { 2524 unlock_page(page); 2525 page_cache_release(page); 2526 goto out_unlock; 2527 } 2528 } 2529 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18) 2530 ClearPageDirty(page); 2531 #else 2532 cancel_dirty_page(page, PAGE_CACHE_SIZE); 2533 #endif 2534 wait_on_page_writeback(page); 2535 set_page_extent_mapped(page); 2536 page_start = (u64)page->index << PAGE_CACHE_SHIFT; 2537 page_end = page_start + PAGE_CACHE_SIZE - 1; 2538 2539 lock_extent(io_tree, page_start, page_end, GFP_NOFS); 2540 2541 set_extent_delalloc(io_tree, page_start, 2542 page_end, GFP_NOFS); 2543 set_page_dirty(page); 2544 2545 unlock_extent(io_tree, page_start, page_end, GFP_NOFS); 2546 unlock_page(page); 2547 page_cache_release(page); 2548 } 2549 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 2550 total_read); 2551 2552 out_unlock: 2553 kfree(ra); 2554 trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1); 2555 if (trans) { 2556 btrfs_add_ordered_inode(inode); 2557 btrfs_end_transaction(trans, BTRFS_I(inode)->root); 2558 mark_inode_dirty(inode); 2559 } 2560 mutex_unlock(&inode->i_mutex); 2561 return 0; 2562 2563 truncate_racing: 2564 vmtruncate(inode, inode->i_size); 2565 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 2566 total_read); 2567 goto out_unlock; 2568 } 2569 2570 /* 2571 * The back references tell us which tree holds a ref on a block, 2572 * but it is possible for the tree root field in the reference to 2573 * reflect the original root before a snapshot was made. In this 2574 * case we should search through all the children of a given root 2575 * to find potential holders of references on a block. 2576 * 2577 * Instead, we do something a little less fancy and just search 2578 * all the roots for a given key/block combination. 2579 */ 2580 static int find_root_for_ref(struct btrfs_root *root, 2581 struct btrfs_path *path, 2582 struct btrfs_key *key0, 2583 int level, 2584 int file_key, 2585 struct btrfs_root **found_root, 2586 u64 bytenr) 2587 { 2588 struct btrfs_key root_location; 2589 struct btrfs_root *cur_root = *found_root; 2590 struct btrfs_file_extent_item *file_extent; 2591 u64 root_search_start = BTRFS_FS_TREE_OBJECTID; 2592 u64 found_bytenr; 2593 int ret; 2594 int i; 2595 2596 root_location.offset = (u64)-1; 2597 root_location.type = BTRFS_ROOT_ITEM_KEY; 2598 path->lowest_level = level; 2599 path->reada = 0; 2600 while(1) { 2601 ret = btrfs_search_slot(NULL, cur_root, key0, path, 0, 0); 2602 found_bytenr = 0; 2603 if (ret == 0 && file_key) { 2604 struct extent_buffer *leaf = path->nodes[0]; 2605 file_extent = btrfs_item_ptr(leaf, path->slots[0], 2606 struct btrfs_file_extent_item); 2607 if (btrfs_file_extent_type(leaf, file_extent) == 2608 BTRFS_FILE_EXTENT_REG) { 2609 found_bytenr = 2610 btrfs_file_extent_disk_bytenr(leaf, 2611 file_extent); 2612 } 2613 } else if (!file_key) { 2614 if (path->nodes[level]) 2615 found_bytenr = path->nodes[level]->start; 2616 } 2617 2618 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 2619 if (!path->nodes[i]) 2620 break; 2621 free_extent_buffer(path->nodes[i]); 2622 path->nodes[i] = NULL; 2623 } 2624 btrfs_release_path(cur_root, path); 2625 2626 if (found_bytenr == bytenr) { 2627 *found_root = cur_root; 2628 ret = 0; 2629 goto out; 2630 } 2631 ret = btrfs_search_root(root->fs_info->tree_root, 2632 root_search_start, &root_search_start); 2633 if (ret) 2634 break; 2635 2636 root_location.objectid = root_search_start; 2637 cur_root = btrfs_read_fs_root_no_name(root->fs_info, 2638 &root_location); 2639 if (!cur_root) { 2640 ret = 1; 2641 break; 2642 } 2643 } 2644 out: 2645 path->lowest_level = 0; 2646 return ret; 2647 } 2648 2649 /* 2650 * note, this releases the path 2651 */ 2652 static int noinline relocate_one_reference(struct btrfs_root *extent_root, 2653 struct btrfs_path *path, 2654 struct btrfs_key *extent_key, 2655 u64 *last_file_objectid, 2656 u64 *last_file_offset, 2657 u64 *last_file_root, 2658 u64 last_extent) 2659 { 2660 struct inode *inode; 2661 struct btrfs_root *found_root; 2662 struct btrfs_key root_location; 2663 struct btrfs_key found_key; 2664 struct btrfs_extent_ref *ref; 2665 u64 ref_root; 2666 u64 ref_gen; 2667 u64 ref_objectid; 2668 u64 ref_offset; 2669 int ret; 2670 int level; 2671 2672 ref = btrfs_item_ptr(path->nodes[0], path->slots[0], 2673 struct btrfs_extent_ref); 2674 ref_root = btrfs_ref_root(path->nodes[0], ref); 2675 ref_gen = btrfs_ref_generation(path->nodes[0], ref); 2676 ref_objectid = btrfs_ref_objectid(path->nodes[0], ref); 2677 ref_offset = btrfs_ref_offset(path->nodes[0], ref); 2678 btrfs_release_path(extent_root, path); 2679 2680 root_location.objectid = ref_root; 2681 if (ref_gen == 0) 2682 root_location.offset = 0; 2683 else 2684 root_location.offset = (u64)-1; 2685 root_location.type = BTRFS_ROOT_ITEM_KEY; 2686 2687 found_root = btrfs_read_fs_root_no_name(extent_root->fs_info, 2688 &root_location); 2689 BUG_ON(!found_root); 2690 2691 if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) { 2692 found_key.objectid = ref_objectid; 2693 found_key.type = BTRFS_EXTENT_DATA_KEY; 2694 found_key.offset = ref_offset; 2695 level = 0; 2696 2697 if (last_extent == extent_key->objectid && 2698 *last_file_objectid == ref_objectid && 2699 *last_file_offset == ref_offset && 2700 *last_file_root == ref_root) 2701 goto out; 2702 2703 ret = find_root_for_ref(extent_root, path, &found_key, 2704 level, 1, &found_root, 2705 extent_key->objectid); 2706 2707 if (ret) 2708 goto out; 2709 2710 if (last_extent == extent_key->objectid && 2711 *last_file_objectid == ref_objectid && 2712 *last_file_offset == ref_offset && 2713 *last_file_root == ref_root) 2714 goto out; 2715 2716 inode = btrfs_iget_locked(extent_root->fs_info->sb, 2717 ref_objectid, found_root); 2718 if (inode->i_state & I_NEW) { 2719 /* the inode and parent dir are two different roots */ 2720 BTRFS_I(inode)->root = found_root; 2721 BTRFS_I(inode)->location.objectid = ref_objectid; 2722 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY; 2723 BTRFS_I(inode)->location.offset = 0; 2724 btrfs_read_locked_inode(inode); 2725 unlock_new_inode(inode); 2726 2727 } 2728 /* this can happen if the reference is not against 2729 * the latest version of the tree root 2730 */ 2731 if (is_bad_inode(inode)) { 2732 goto out; 2733 } 2734 *last_file_objectid = inode->i_ino; 2735 *last_file_root = found_root->root_key.objectid; 2736 *last_file_offset = ref_offset; 2737 2738 relocate_inode_pages(inode, ref_offset, extent_key->offset); 2739 iput(inode); 2740 } else { 2741 struct btrfs_trans_handle *trans; 2742 struct extent_buffer *eb; 2743 int i; 2744 2745 eb = read_tree_block(found_root, extent_key->objectid, 2746 extent_key->offset, 0); 2747 btrfs_tree_lock(eb); 2748 level = btrfs_header_level(eb); 2749 2750 if (level == 0) 2751 btrfs_item_key_to_cpu(eb, &found_key, 0); 2752 else 2753 btrfs_node_key_to_cpu(eb, &found_key, 0); 2754 2755 btrfs_tree_unlock(eb); 2756 free_extent_buffer(eb); 2757 2758 ret = find_root_for_ref(extent_root, path, &found_key, 2759 level, 0, &found_root, 2760 extent_key->objectid); 2761 2762 if (ret) 2763 goto out; 2764 2765 trans = btrfs_start_transaction(found_root, 1); 2766 2767 path->lowest_level = level; 2768 path->reada = 2; 2769 ret = btrfs_search_slot(trans, found_root, &found_key, path, 2770 0, 1); 2771 path->lowest_level = 0; 2772 for (i = level; i < BTRFS_MAX_LEVEL; i++) { 2773 if (!path->nodes[i]) 2774 break; 2775 free_extent_buffer(path->nodes[i]); 2776 path->nodes[i] = NULL; 2777 } 2778 btrfs_release_path(found_root, path); 2779 if (found_root == found_root->fs_info->extent_root) 2780 btrfs_extent_post_op(trans, found_root); 2781 btrfs_end_transaction(trans, found_root); 2782 } 2783 2784 out: 2785 return 0; 2786 } 2787 2788 static int noinline del_extent_zero(struct btrfs_root *extent_root, 2789 struct btrfs_path *path, 2790 struct btrfs_key *extent_key) 2791 { 2792 int ret; 2793 struct btrfs_trans_handle *trans; 2794 2795 trans = btrfs_start_transaction(extent_root, 1); 2796 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1); 2797 if (ret > 0) { 2798 ret = -EIO; 2799 goto out; 2800 } 2801 if (ret < 0) 2802 goto out; 2803 ret = btrfs_del_item(trans, extent_root, path); 2804 out: 2805 btrfs_end_transaction(trans, extent_root); 2806 return ret; 2807 } 2808 2809 static int noinline relocate_one_extent(struct btrfs_root *extent_root, 2810 struct btrfs_path *path, 2811 struct btrfs_key *extent_key) 2812 { 2813 struct btrfs_key key; 2814 struct btrfs_key found_key; 2815 struct extent_buffer *leaf; 2816 u64 last_file_objectid = 0; 2817 u64 last_file_root = 0; 2818 u64 last_file_offset = (u64)-1; 2819 u64 last_extent = 0; 2820 u32 nritems; 2821 u32 item_size; 2822 int ret = 0; 2823 2824 if (extent_key->objectid == 0) { 2825 ret = del_extent_zero(extent_root, path, extent_key); 2826 goto out; 2827 } 2828 key.objectid = extent_key->objectid; 2829 key.type = BTRFS_EXTENT_REF_KEY; 2830 key.offset = 0; 2831 2832 while(1) { 2833 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); 2834 2835 if (ret < 0) 2836 goto out; 2837 2838 ret = 0; 2839 leaf = path->nodes[0]; 2840 nritems = btrfs_header_nritems(leaf); 2841 if (path->slots[0] == nritems) { 2842 ret = btrfs_next_leaf(extent_root, path); 2843 if (ret > 0) { 2844 ret = 0; 2845 goto out; 2846 } 2847 if (ret < 0) 2848 goto out; 2849 leaf = path->nodes[0]; 2850 } 2851 2852 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 2853 if (found_key.objectid != extent_key->objectid) { 2854 break; 2855 } 2856 2857 if (found_key.type != BTRFS_EXTENT_REF_KEY) { 2858 break; 2859 } 2860 2861 key.offset = found_key.offset + 1; 2862 item_size = btrfs_item_size_nr(leaf, path->slots[0]); 2863 2864 ret = relocate_one_reference(extent_root, path, extent_key, 2865 &last_file_objectid, 2866 &last_file_offset, 2867 &last_file_root, last_extent); 2868 if (ret) 2869 goto out; 2870 last_extent = extent_key->objectid; 2871 } 2872 ret = 0; 2873 out: 2874 btrfs_release_path(extent_root, path); 2875 return ret; 2876 } 2877 2878 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags) 2879 { 2880 u64 num_devices; 2881 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 | 2882 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10; 2883 2884 num_devices = root->fs_info->fs_devices->num_devices; 2885 if (num_devices == 1) { 2886 stripped |= BTRFS_BLOCK_GROUP_DUP; 2887 stripped = flags & ~stripped; 2888 2889 /* turn raid0 into single device chunks */ 2890 if (flags & BTRFS_BLOCK_GROUP_RAID0) 2891 return stripped; 2892 2893 /* turn mirroring into duplication */ 2894 if (flags & (BTRFS_BLOCK_GROUP_RAID1 | 2895 BTRFS_BLOCK_GROUP_RAID10)) 2896 return stripped | BTRFS_BLOCK_GROUP_DUP; 2897 return flags; 2898 } else { 2899 /* they already had raid on here, just return */ 2900 if (flags & stripped) 2901 return flags; 2902 2903 stripped |= BTRFS_BLOCK_GROUP_DUP; 2904 stripped = flags & ~stripped; 2905 2906 /* switch duplicated blocks with raid1 */ 2907 if (flags & BTRFS_BLOCK_GROUP_DUP) 2908 return stripped | BTRFS_BLOCK_GROUP_RAID1; 2909 2910 /* turn single device chunks into raid0 */ 2911 return stripped | BTRFS_BLOCK_GROUP_RAID0; 2912 } 2913 return flags; 2914 } 2915 2916 int __alloc_chunk_for_shrink(struct btrfs_root *root, 2917 struct btrfs_block_group_cache *shrink_block_group, 2918 int force) 2919 { 2920 struct btrfs_trans_handle *trans; 2921 u64 new_alloc_flags; 2922 u64 calc; 2923 2924 if (btrfs_block_group_used(&shrink_block_group->item) > 0) { 2925 2926 trans = btrfs_start_transaction(root, 1); 2927 new_alloc_flags = update_block_group_flags(root, 2928 shrink_block_group->flags); 2929 if (new_alloc_flags != shrink_block_group->flags) { 2930 calc = 2931 btrfs_block_group_used(&shrink_block_group->item); 2932 } else { 2933 calc = shrink_block_group->key.offset; 2934 } 2935 do_chunk_alloc(trans, root->fs_info->extent_root, 2936 calc + 2 * 1024 * 1024, new_alloc_flags, force); 2937 btrfs_end_transaction(trans, root); 2938 } 2939 return 0; 2940 } 2941 2942 int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start) 2943 { 2944 struct btrfs_trans_handle *trans; 2945 struct btrfs_root *tree_root = root->fs_info->tree_root; 2946 struct btrfs_path *path; 2947 u64 cur_byte; 2948 u64 total_found; 2949 u64 shrink_last_byte; 2950 struct btrfs_block_group_cache *shrink_block_group; 2951 struct btrfs_fs_info *info = root->fs_info; 2952 struct btrfs_key key; 2953 struct btrfs_key found_key; 2954 struct extent_buffer *leaf; 2955 u32 nritems; 2956 int ret; 2957 int progress; 2958 2959 mutex_lock(&root->fs_info->alloc_mutex); 2960 shrink_block_group = btrfs_lookup_block_group(root->fs_info, 2961 shrink_start); 2962 BUG_ON(!shrink_block_group); 2963 2964 shrink_last_byte = shrink_block_group->key.objectid + 2965 shrink_block_group->key.offset; 2966 2967 shrink_block_group->space_info->total_bytes -= 2968 shrink_block_group->key.offset; 2969 path = btrfs_alloc_path(); 2970 root = root->fs_info->extent_root; 2971 path->reada = 2; 2972 2973 printk("btrfs relocating block group %llu flags %llu\n", 2974 (unsigned long long)shrink_start, 2975 (unsigned long long)shrink_block_group->flags); 2976 2977 __alloc_chunk_for_shrink(root, shrink_block_group, 1); 2978 2979 again: 2980 2981 shrink_block_group->ro = 1; 2982 2983 total_found = 0; 2984 progress = 0; 2985 key.objectid = shrink_start; 2986 key.offset = 0; 2987 key.type = 0; 2988 cur_byte = key.objectid; 2989 2990 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 2991 if (ret < 0) 2992 goto out; 2993 2994 ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY); 2995 if (ret < 0) 2996 goto out; 2997 2998 if (ret == 0) { 2999 leaf = path->nodes[0]; 3000 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 3001 if (found_key.objectid + found_key.offset > shrink_start && 3002 found_key.objectid < shrink_last_byte) { 3003 cur_byte = found_key.objectid; 3004 key.objectid = cur_byte; 3005 } 3006 } 3007 btrfs_release_path(root, path); 3008 3009 while(1) { 3010 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 3011 if (ret < 0) 3012 goto out; 3013 3014 leaf = path->nodes[0]; 3015 nritems = btrfs_header_nritems(leaf); 3016 next: 3017 if (path->slots[0] >= nritems) { 3018 ret = btrfs_next_leaf(root, path); 3019 if (ret < 0) 3020 goto out; 3021 if (ret == 1) { 3022 ret = 0; 3023 break; 3024 } 3025 leaf = path->nodes[0]; 3026 nritems = btrfs_header_nritems(leaf); 3027 } 3028 3029 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 3030 3031 if (found_key.objectid >= shrink_last_byte) 3032 break; 3033 3034 if (progress && need_resched()) { 3035 memcpy(&key, &found_key, sizeof(key)); 3036 cond_resched(); 3037 btrfs_release_path(root, path); 3038 btrfs_search_slot(NULL, root, &key, path, 0, 0); 3039 progress = 0; 3040 goto next; 3041 } 3042 progress = 1; 3043 3044 if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY || 3045 found_key.objectid + found_key.offset <= cur_byte) { 3046 memcpy(&key, &found_key, sizeof(key)); 3047 key.offset++; 3048 path->slots[0]++; 3049 goto next; 3050 } 3051 3052 total_found++; 3053 cur_byte = found_key.objectid + found_key.offset; 3054 key.objectid = cur_byte; 3055 btrfs_release_path(root, path); 3056 ret = relocate_one_extent(root, path, &found_key); 3057 __alloc_chunk_for_shrink(root, shrink_block_group, 0); 3058 } 3059 3060 btrfs_release_path(root, path); 3061 3062 if (total_found > 0) { 3063 printk("btrfs relocate found %llu last extent was %llu\n", 3064 (unsigned long long)total_found, 3065 (unsigned long long)found_key.objectid); 3066 trans = btrfs_start_transaction(tree_root, 1); 3067 btrfs_commit_transaction(trans, tree_root); 3068 3069 btrfs_clean_old_snapshots(tree_root); 3070 3071 trans = btrfs_start_transaction(tree_root, 1); 3072 btrfs_commit_transaction(trans, tree_root); 3073 goto again; 3074 } 3075 3076 /* 3077 * we've freed all the extents, now remove the block 3078 * group item from the tree 3079 */ 3080 trans = btrfs_start_transaction(root, 1); 3081 memcpy(&key, &shrink_block_group->key, sizeof(key)); 3082 3083 ret = btrfs_search_slot(trans, root, &key, path, -1, 1); 3084 if (ret > 0) 3085 ret = -EIO; 3086 if (ret < 0) 3087 goto out; 3088 3089 clear_extent_bits(&info->block_group_cache, key.objectid, 3090 key.objectid + key.offset - 1, 3091 (unsigned int)-1, GFP_NOFS); 3092 3093 3094 clear_extent_bits(&info->free_space_cache, 3095 key.objectid, key.objectid + key.offset - 1, 3096 (unsigned int)-1, GFP_NOFS); 3097 3098 memset(shrink_block_group, 0, sizeof(*shrink_block_group)); 3099 kfree(shrink_block_group); 3100 3101 btrfs_del_item(trans, root, path); 3102 btrfs_commit_transaction(trans, root); 3103 3104 /* the code to unpin extents might set a few bits in the free 3105 * space cache for this range again 3106 */ 3107 clear_extent_bits(&info->free_space_cache, 3108 key.objectid, key.objectid + key.offset - 1, 3109 (unsigned int)-1, GFP_NOFS); 3110 out: 3111 btrfs_free_path(path); 3112 mutex_unlock(&root->fs_info->alloc_mutex); 3113 return ret; 3114 } 3115 3116 int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path, 3117 struct btrfs_key *key) 3118 { 3119 int ret = 0; 3120 struct btrfs_key found_key; 3121 struct extent_buffer *leaf; 3122 int slot; 3123 3124 ret = btrfs_search_slot(NULL, root, key, path, 0, 0); 3125 if (ret < 0) 3126 goto out; 3127 3128 while(1) { 3129 slot = path->slots[0]; 3130 leaf = path->nodes[0]; 3131 if (slot >= btrfs_header_nritems(leaf)) { 3132 ret = btrfs_next_leaf(root, path); 3133 if (ret == 0) 3134 continue; 3135 if (ret < 0) 3136 goto out; 3137 break; 3138 } 3139 btrfs_item_key_to_cpu(leaf, &found_key, slot); 3140 3141 if (found_key.objectid >= key->objectid && 3142 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) { 3143 ret = 0; 3144 goto out; 3145 } 3146 path->slots[0]++; 3147 } 3148 ret = -ENOENT; 3149 out: 3150 return ret; 3151 } 3152 3153 int btrfs_read_block_groups(struct btrfs_root *root) 3154 { 3155 struct btrfs_path *path; 3156 int ret; 3157 int bit; 3158 struct btrfs_block_group_cache *cache; 3159 struct btrfs_fs_info *info = root->fs_info; 3160 struct btrfs_space_info *space_info; 3161 struct extent_io_tree *block_group_cache; 3162 struct btrfs_key key; 3163 struct btrfs_key found_key; 3164 struct extent_buffer *leaf; 3165 3166 block_group_cache = &info->block_group_cache; 3167 root = info->extent_root; 3168 key.objectid = 0; 3169 key.offset = 0; 3170 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY); 3171 path = btrfs_alloc_path(); 3172 if (!path) 3173 return -ENOMEM; 3174 3175 mutex_lock(&root->fs_info->alloc_mutex); 3176 while(1) { 3177 ret = find_first_block_group(root, path, &key); 3178 if (ret > 0) { 3179 ret = 0; 3180 goto error; 3181 } 3182 if (ret != 0) 3183 goto error; 3184 3185 leaf = path->nodes[0]; 3186 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 3187 cache = kzalloc(sizeof(*cache), GFP_NOFS); 3188 if (!cache) { 3189 ret = -ENOMEM; 3190 break; 3191 } 3192 3193 read_extent_buffer(leaf, &cache->item, 3194 btrfs_item_ptr_offset(leaf, path->slots[0]), 3195 sizeof(cache->item)); 3196 memcpy(&cache->key, &found_key, sizeof(found_key)); 3197 3198 key.objectid = found_key.objectid + found_key.offset; 3199 btrfs_release_path(root, path); 3200 cache->flags = btrfs_block_group_flags(&cache->item); 3201 bit = 0; 3202 if (cache->flags & BTRFS_BLOCK_GROUP_DATA) { 3203 bit = BLOCK_GROUP_DATA; 3204 } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) { 3205 bit = BLOCK_GROUP_SYSTEM; 3206 } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) { 3207 bit = BLOCK_GROUP_METADATA; 3208 } 3209 set_avail_alloc_bits(info, cache->flags); 3210 3211 ret = update_space_info(info, cache->flags, found_key.offset, 3212 btrfs_block_group_used(&cache->item), 3213 &space_info); 3214 BUG_ON(ret); 3215 cache->space_info = space_info; 3216 3217 /* use EXTENT_LOCKED to prevent merging */ 3218 set_extent_bits(block_group_cache, found_key.objectid, 3219 found_key.objectid + found_key.offset - 1, 3220 bit | EXTENT_LOCKED, GFP_NOFS); 3221 set_state_private(block_group_cache, found_key.objectid, 3222 (unsigned long)cache); 3223 3224 if (key.objectid >= 3225 btrfs_super_total_bytes(&info->super_copy)) 3226 break; 3227 } 3228 ret = 0; 3229 error: 3230 btrfs_free_path(path); 3231 mutex_unlock(&root->fs_info->alloc_mutex); 3232 return ret; 3233 } 3234 3235 int btrfs_make_block_group(struct btrfs_trans_handle *trans, 3236 struct btrfs_root *root, u64 bytes_used, 3237 u64 type, u64 chunk_objectid, u64 chunk_offset, 3238 u64 size) 3239 { 3240 int ret; 3241 int bit = 0; 3242 struct btrfs_root *extent_root; 3243 struct btrfs_block_group_cache *cache; 3244 struct extent_io_tree *block_group_cache; 3245 3246 extent_root = root->fs_info->extent_root; 3247 block_group_cache = &root->fs_info->block_group_cache; 3248 3249 cache = kzalloc(sizeof(*cache), GFP_NOFS); 3250 BUG_ON(!cache); 3251 cache->key.objectid = chunk_offset; 3252 cache->key.offset = size; 3253 btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY); 3254 3255 btrfs_set_block_group_used(&cache->item, bytes_used); 3256 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid); 3257 cache->flags = type; 3258 btrfs_set_block_group_flags(&cache->item, type); 3259 3260 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used, 3261 &cache->space_info); 3262 BUG_ON(ret); 3263 3264 bit = block_group_state_bits(type); 3265 set_extent_bits(block_group_cache, chunk_offset, 3266 chunk_offset + size - 1, 3267 bit | EXTENT_LOCKED, GFP_NOFS); 3268 3269 set_state_private(block_group_cache, chunk_offset, 3270 (unsigned long)cache); 3271 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item, 3272 sizeof(cache->item)); 3273 BUG_ON(ret); 3274 3275 finish_current_insert(trans, extent_root); 3276 ret = del_pending_extents(trans, extent_root); 3277 BUG_ON(ret); 3278 set_avail_alloc_bits(extent_root->fs_info, type); 3279 3280 return 0; 3281 } 3282