1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * Copyright (C) 2007 Oracle. All rights reserved. 4 */ 5 6 #include <linux/sched.h> 7 #include <linux/sched/signal.h> 8 #include <linux/pagemap.h> 9 #include <linux/writeback.h> 10 #include <linux/blkdev.h> 11 #include <linux/sort.h> 12 #include <linux/rcupdate.h> 13 #include <linux/kthread.h> 14 #include <linux/slab.h> 15 #include <linux/ratelimit.h> 16 #include <linux/percpu_counter.h> 17 #include <linux/lockdep.h> 18 #include <linux/crc32c.h> 19 #include "ctree.h" 20 #include "extent-tree.h" 21 #include "tree-log.h" 22 #include "disk-io.h" 23 #include "print-tree.h" 24 #include "volumes.h" 25 #include "raid56.h" 26 #include "locking.h" 27 #include "free-space-cache.h" 28 #include "free-space-tree.h" 29 #include "sysfs.h" 30 #include "qgroup.h" 31 #include "ref-verify.h" 32 #include "space-info.h" 33 #include "block-rsv.h" 34 #include "delalloc-space.h" 35 #include "discard.h" 36 #include "rcu-string.h" 37 #include "zoned.h" 38 #include "dev-replace.h" 39 #include "fs.h" 40 #include "accessors.h" 41 #include "root-tree.h" 42 #include "file-item.h" 43 #include "orphan.h" 44 #include "tree-checker.h" 45 46 #undef SCRAMBLE_DELAYED_REFS 47 48 49 static int __btrfs_free_extent(struct btrfs_trans_handle *trans, 50 struct btrfs_delayed_ref_node *node, u64 parent, 51 u64 root_objectid, u64 owner_objectid, 52 u64 owner_offset, int refs_to_drop, 53 struct btrfs_delayed_extent_op *extra_op); 54 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op, 55 struct extent_buffer *leaf, 56 struct btrfs_extent_item *ei); 57 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans, 58 u64 parent, u64 root_objectid, 59 u64 flags, u64 owner, u64 offset, 60 struct btrfs_key *ins, int ref_mod); 61 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans, 62 struct btrfs_delayed_ref_node *node, 63 struct btrfs_delayed_extent_op *extent_op); 64 static int find_next_key(struct btrfs_path *path, int level, 65 struct btrfs_key *key); 66 67 static int block_group_bits(struct btrfs_block_group *cache, u64 bits) 68 { 69 return (cache->flags & bits) == bits; 70 } 71 72 /* simple helper to search for an existing data extent at a given offset */ 73 int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len) 74 { 75 struct btrfs_root *root = btrfs_extent_root(fs_info, start); 76 int ret; 77 struct btrfs_key key; 78 struct btrfs_path *path; 79 80 path = btrfs_alloc_path(); 81 if (!path) 82 return -ENOMEM; 83 84 key.objectid = start; 85 key.offset = len; 86 key.type = BTRFS_EXTENT_ITEM_KEY; 87 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 88 btrfs_free_path(path); 89 return ret; 90 } 91 92 /* 93 * helper function to lookup reference count and flags of a tree block. 94 * 95 * the head node for delayed ref is used to store the sum of all the 96 * reference count modifications queued up in the rbtree. the head 97 * node may also store the extent flags to set. This way you can check 98 * to see what the reference count and extent flags would be if all of 99 * the delayed refs are not processed. 100 */ 101 int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans, 102 struct btrfs_fs_info *fs_info, u64 bytenr, 103 u64 offset, int metadata, u64 *refs, u64 *flags) 104 { 105 struct btrfs_root *extent_root; 106 struct btrfs_delayed_ref_head *head; 107 struct btrfs_delayed_ref_root *delayed_refs; 108 struct btrfs_path *path; 109 struct btrfs_extent_item *ei; 110 struct extent_buffer *leaf; 111 struct btrfs_key key; 112 u32 item_size; 113 u64 num_refs; 114 u64 extent_flags; 115 int ret; 116 117 /* 118 * If we don't have skinny metadata, don't bother doing anything 119 * different 120 */ 121 if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) { 122 offset = fs_info->nodesize; 123 metadata = 0; 124 } 125 126 path = btrfs_alloc_path(); 127 if (!path) 128 return -ENOMEM; 129 130 if (!trans) { 131 path->skip_locking = 1; 132 path->search_commit_root = 1; 133 } 134 135 search_again: 136 key.objectid = bytenr; 137 key.offset = offset; 138 if (metadata) 139 key.type = BTRFS_METADATA_ITEM_KEY; 140 else 141 key.type = BTRFS_EXTENT_ITEM_KEY; 142 143 extent_root = btrfs_extent_root(fs_info, bytenr); 144 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); 145 if (ret < 0) 146 goto out_free; 147 148 if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) { 149 if (path->slots[0]) { 150 path->slots[0]--; 151 btrfs_item_key_to_cpu(path->nodes[0], &key, 152 path->slots[0]); 153 if (key.objectid == bytenr && 154 key.type == BTRFS_EXTENT_ITEM_KEY && 155 key.offset == fs_info->nodesize) 156 ret = 0; 157 } 158 } 159 160 if (ret == 0) { 161 leaf = path->nodes[0]; 162 item_size = btrfs_item_size(leaf, path->slots[0]); 163 if (item_size >= sizeof(*ei)) { 164 ei = btrfs_item_ptr(leaf, path->slots[0], 165 struct btrfs_extent_item); 166 num_refs = btrfs_extent_refs(leaf, ei); 167 extent_flags = btrfs_extent_flags(leaf, ei); 168 } else { 169 ret = -EUCLEAN; 170 btrfs_err(fs_info, 171 "unexpected extent item size, has %u expect >= %zu", 172 item_size, sizeof(*ei)); 173 if (trans) 174 btrfs_abort_transaction(trans, ret); 175 else 176 btrfs_handle_fs_error(fs_info, ret, NULL); 177 178 goto out_free; 179 } 180 181 BUG_ON(num_refs == 0); 182 } else { 183 num_refs = 0; 184 extent_flags = 0; 185 ret = 0; 186 } 187 188 if (!trans) 189 goto out; 190 191 delayed_refs = &trans->transaction->delayed_refs; 192 spin_lock(&delayed_refs->lock); 193 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr); 194 if (head) { 195 if (!mutex_trylock(&head->mutex)) { 196 refcount_inc(&head->refs); 197 spin_unlock(&delayed_refs->lock); 198 199 btrfs_release_path(path); 200 201 /* 202 * Mutex was contended, block until it's released and try 203 * again 204 */ 205 mutex_lock(&head->mutex); 206 mutex_unlock(&head->mutex); 207 btrfs_put_delayed_ref_head(head); 208 goto search_again; 209 } 210 spin_lock(&head->lock); 211 if (head->extent_op && head->extent_op->update_flags) 212 extent_flags |= head->extent_op->flags_to_set; 213 else 214 BUG_ON(num_refs == 0); 215 216 num_refs += head->ref_mod; 217 spin_unlock(&head->lock); 218 mutex_unlock(&head->mutex); 219 } 220 spin_unlock(&delayed_refs->lock); 221 out: 222 WARN_ON(num_refs == 0); 223 if (refs) 224 *refs = num_refs; 225 if (flags) 226 *flags = extent_flags; 227 out_free: 228 btrfs_free_path(path); 229 return ret; 230 } 231 232 /* 233 * Back reference rules. Back refs have three main goals: 234 * 235 * 1) differentiate between all holders of references to an extent so that 236 * when a reference is dropped we can make sure it was a valid reference 237 * before freeing the extent. 238 * 239 * 2) Provide enough information to quickly find the holders of an extent 240 * if we notice a given block is corrupted or bad. 241 * 242 * 3) Make it easy to migrate blocks for FS shrinking or storage pool 243 * maintenance. This is actually the same as #2, but with a slightly 244 * different use case. 245 * 246 * There are two kinds of back refs. The implicit back refs is optimized 247 * for pointers in non-shared tree blocks. For a given pointer in a block, 248 * back refs of this kind provide information about the block's owner tree 249 * and the pointer's key. These information allow us to find the block by 250 * b-tree searching. The full back refs is for pointers in tree blocks not 251 * referenced by their owner trees. The location of tree block is recorded 252 * in the back refs. Actually the full back refs is generic, and can be 253 * used in all cases the implicit back refs is used. The major shortcoming 254 * of the full back refs is its overhead. Every time a tree block gets 255 * COWed, we have to update back refs entry for all pointers in it. 256 * 257 * For a newly allocated tree block, we use implicit back refs for 258 * pointers in it. This means most tree related operations only involve 259 * implicit back refs. For a tree block created in old transaction, the 260 * only way to drop a reference to it is COW it. So we can detect the 261 * event that tree block loses its owner tree's reference and do the 262 * back refs conversion. 263 * 264 * When a tree block is COWed through a tree, there are four cases: 265 * 266 * The reference count of the block is one and the tree is the block's 267 * owner tree. Nothing to do in this case. 268 * 269 * The reference count of the block is one and the tree is not the 270 * block's owner tree. In this case, full back refs is used for pointers 271 * in the block. Remove these full back refs, add implicit back refs for 272 * every pointers in the new block. 273 * 274 * The reference count of the block is greater than one and the tree is 275 * the block's owner tree. In this case, implicit back refs is used for 276 * pointers in the block. Add full back refs for every pointers in the 277 * block, increase lower level extents' reference counts. The original 278 * implicit back refs are entailed to the new block. 279 * 280 * The reference count of the block is greater than one and the tree is 281 * not the block's owner tree. Add implicit back refs for every pointer in 282 * the new block, increase lower level extents' reference count. 283 * 284 * Back Reference Key composing: 285 * 286 * The key objectid corresponds to the first byte in the extent, 287 * The key type is used to differentiate between types of back refs. 288 * There are different meanings of the key offset for different types 289 * of back refs. 290 * 291 * File extents can be referenced by: 292 * 293 * - multiple snapshots, subvolumes, or different generations in one subvol 294 * - different files inside a single subvolume 295 * - different offsets inside a file (bookend extents in file.c) 296 * 297 * The extent ref structure for the implicit back refs has fields for: 298 * 299 * - Objectid of the subvolume root 300 * - objectid of the file holding the reference 301 * - original offset in the file 302 * - how many bookend extents 303 * 304 * The key offset for the implicit back refs is hash of the first 305 * three fields. 306 * 307 * The extent ref structure for the full back refs has field for: 308 * 309 * - number of pointers in the tree leaf 310 * 311 * The key offset for the implicit back refs is the first byte of 312 * the tree leaf 313 * 314 * When a file extent is allocated, The implicit back refs is used. 315 * the fields are filled in: 316 * 317 * (root_key.objectid, inode objectid, offset in file, 1) 318 * 319 * When a file extent is removed file truncation, we find the 320 * corresponding implicit back refs and check the following fields: 321 * 322 * (btrfs_header_owner(leaf), inode objectid, offset in file) 323 * 324 * Btree extents can be referenced by: 325 * 326 * - Different subvolumes 327 * 328 * Both the implicit back refs and the full back refs for tree blocks 329 * only consist of key. The key offset for the implicit back refs is 330 * objectid of block's owner tree. The key offset for the full back refs 331 * is the first byte of parent block. 332 * 333 * When implicit back refs is used, information about the lowest key and 334 * level of the tree block are required. These information are stored in 335 * tree block info structure. 336 */ 337 338 /* 339 * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required, 340 * is_data == BTRFS_REF_TYPE_DATA, data type is requiried, 341 * is_data == BTRFS_REF_TYPE_ANY, either type is OK. 342 */ 343 int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb, 344 struct btrfs_extent_inline_ref *iref, 345 enum btrfs_inline_ref_type is_data) 346 { 347 int type = btrfs_extent_inline_ref_type(eb, iref); 348 u64 offset = btrfs_extent_inline_ref_offset(eb, iref); 349 350 if (type == BTRFS_TREE_BLOCK_REF_KEY || 351 type == BTRFS_SHARED_BLOCK_REF_KEY || 352 type == BTRFS_SHARED_DATA_REF_KEY || 353 type == BTRFS_EXTENT_DATA_REF_KEY) { 354 if (is_data == BTRFS_REF_TYPE_BLOCK) { 355 if (type == BTRFS_TREE_BLOCK_REF_KEY) 356 return type; 357 if (type == BTRFS_SHARED_BLOCK_REF_KEY) { 358 ASSERT(eb->fs_info); 359 /* 360 * Every shared one has parent tree block, 361 * which must be aligned to sector size. 362 */ 363 if (offset && 364 IS_ALIGNED(offset, eb->fs_info->sectorsize)) 365 return type; 366 } 367 } else if (is_data == BTRFS_REF_TYPE_DATA) { 368 if (type == BTRFS_EXTENT_DATA_REF_KEY) 369 return type; 370 if (type == BTRFS_SHARED_DATA_REF_KEY) { 371 ASSERT(eb->fs_info); 372 /* 373 * Every shared one has parent tree block, 374 * which must be aligned to sector size. 375 */ 376 if (offset && 377 IS_ALIGNED(offset, eb->fs_info->sectorsize)) 378 return type; 379 } 380 } else { 381 ASSERT(is_data == BTRFS_REF_TYPE_ANY); 382 return type; 383 } 384 } 385 386 WARN_ON(1); 387 btrfs_print_leaf(eb); 388 btrfs_err(eb->fs_info, 389 "eb %llu iref 0x%lx invalid extent inline ref type %d", 390 eb->start, (unsigned long)iref, type); 391 392 return BTRFS_REF_TYPE_INVALID; 393 } 394 395 u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset) 396 { 397 u32 high_crc = ~(u32)0; 398 u32 low_crc = ~(u32)0; 399 __le64 lenum; 400 401 lenum = cpu_to_le64(root_objectid); 402 high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum)); 403 lenum = cpu_to_le64(owner); 404 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); 405 lenum = cpu_to_le64(offset); 406 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum)); 407 408 return ((u64)high_crc << 31) ^ (u64)low_crc; 409 } 410 411 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf, 412 struct btrfs_extent_data_ref *ref) 413 { 414 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref), 415 btrfs_extent_data_ref_objectid(leaf, ref), 416 btrfs_extent_data_ref_offset(leaf, ref)); 417 } 418 419 static int match_extent_data_ref(struct extent_buffer *leaf, 420 struct btrfs_extent_data_ref *ref, 421 u64 root_objectid, u64 owner, u64 offset) 422 { 423 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid || 424 btrfs_extent_data_ref_objectid(leaf, ref) != owner || 425 btrfs_extent_data_ref_offset(leaf, ref) != offset) 426 return 0; 427 return 1; 428 } 429 430 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans, 431 struct btrfs_path *path, 432 u64 bytenr, u64 parent, 433 u64 root_objectid, 434 u64 owner, u64 offset) 435 { 436 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr); 437 struct btrfs_key key; 438 struct btrfs_extent_data_ref *ref; 439 struct extent_buffer *leaf; 440 u32 nritems; 441 int ret; 442 int recow; 443 int err = -ENOENT; 444 445 key.objectid = bytenr; 446 if (parent) { 447 key.type = BTRFS_SHARED_DATA_REF_KEY; 448 key.offset = parent; 449 } else { 450 key.type = BTRFS_EXTENT_DATA_REF_KEY; 451 key.offset = hash_extent_data_ref(root_objectid, 452 owner, offset); 453 } 454 again: 455 recow = 0; 456 ret = btrfs_search_slot(trans, root, &key, path, -1, 1); 457 if (ret < 0) { 458 err = ret; 459 goto fail; 460 } 461 462 if (parent) { 463 if (!ret) 464 return 0; 465 goto fail; 466 } 467 468 leaf = path->nodes[0]; 469 nritems = btrfs_header_nritems(leaf); 470 while (1) { 471 if (path->slots[0] >= nritems) { 472 ret = btrfs_next_leaf(root, path); 473 if (ret < 0) 474 err = ret; 475 if (ret) 476 goto fail; 477 478 leaf = path->nodes[0]; 479 nritems = btrfs_header_nritems(leaf); 480 recow = 1; 481 } 482 483 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 484 if (key.objectid != bytenr || 485 key.type != BTRFS_EXTENT_DATA_REF_KEY) 486 goto fail; 487 488 ref = btrfs_item_ptr(leaf, path->slots[0], 489 struct btrfs_extent_data_ref); 490 491 if (match_extent_data_ref(leaf, ref, root_objectid, 492 owner, offset)) { 493 if (recow) { 494 btrfs_release_path(path); 495 goto again; 496 } 497 err = 0; 498 break; 499 } 500 path->slots[0]++; 501 } 502 fail: 503 return err; 504 } 505 506 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans, 507 struct btrfs_path *path, 508 u64 bytenr, u64 parent, 509 u64 root_objectid, u64 owner, 510 u64 offset, int refs_to_add) 511 { 512 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr); 513 struct btrfs_key key; 514 struct extent_buffer *leaf; 515 u32 size; 516 u32 num_refs; 517 int ret; 518 519 key.objectid = bytenr; 520 if (parent) { 521 key.type = BTRFS_SHARED_DATA_REF_KEY; 522 key.offset = parent; 523 size = sizeof(struct btrfs_shared_data_ref); 524 } else { 525 key.type = BTRFS_EXTENT_DATA_REF_KEY; 526 key.offset = hash_extent_data_ref(root_objectid, 527 owner, offset); 528 size = sizeof(struct btrfs_extent_data_ref); 529 } 530 531 ret = btrfs_insert_empty_item(trans, root, path, &key, size); 532 if (ret && ret != -EEXIST) 533 goto fail; 534 535 leaf = path->nodes[0]; 536 if (parent) { 537 struct btrfs_shared_data_ref *ref; 538 ref = btrfs_item_ptr(leaf, path->slots[0], 539 struct btrfs_shared_data_ref); 540 if (ret == 0) { 541 btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add); 542 } else { 543 num_refs = btrfs_shared_data_ref_count(leaf, ref); 544 num_refs += refs_to_add; 545 btrfs_set_shared_data_ref_count(leaf, ref, num_refs); 546 } 547 } else { 548 struct btrfs_extent_data_ref *ref; 549 while (ret == -EEXIST) { 550 ref = btrfs_item_ptr(leaf, path->slots[0], 551 struct btrfs_extent_data_ref); 552 if (match_extent_data_ref(leaf, ref, root_objectid, 553 owner, offset)) 554 break; 555 btrfs_release_path(path); 556 key.offset++; 557 ret = btrfs_insert_empty_item(trans, root, path, &key, 558 size); 559 if (ret && ret != -EEXIST) 560 goto fail; 561 562 leaf = path->nodes[0]; 563 } 564 ref = btrfs_item_ptr(leaf, path->slots[0], 565 struct btrfs_extent_data_ref); 566 if (ret == 0) { 567 btrfs_set_extent_data_ref_root(leaf, ref, 568 root_objectid); 569 btrfs_set_extent_data_ref_objectid(leaf, ref, owner); 570 btrfs_set_extent_data_ref_offset(leaf, ref, offset); 571 btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add); 572 } else { 573 num_refs = btrfs_extent_data_ref_count(leaf, ref); 574 num_refs += refs_to_add; 575 btrfs_set_extent_data_ref_count(leaf, ref, num_refs); 576 } 577 } 578 btrfs_mark_buffer_dirty(trans, leaf); 579 ret = 0; 580 fail: 581 btrfs_release_path(path); 582 return ret; 583 } 584 585 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans, 586 struct btrfs_root *root, 587 struct btrfs_path *path, 588 int refs_to_drop) 589 { 590 struct btrfs_key key; 591 struct btrfs_extent_data_ref *ref1 = NULL; 592 struct btrfs_shared_data_ref *ref2 = NULL; 593 struct extent_buffer *leaf; 594 u32 num_refs = 0; 595 int ret = 0; 596 597 leaf = path->nodes[0]; 598 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 599 600 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) { 601 ref1 = btrfs_item_ptr(leaf, path->slots[0], 602 struct btrfs_extent_data_ref); 603 num_refs = btrfs_extent_data_ref_count(leaf, ref1); 604 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) { 605 ref2 = btrfs_item_ptr(leaf, path->slots[0], 606 struct btrfs_shared_data_ref); 607 num_refs = btrfs_shared_data_ref_count(leaf, ref2); 608 } else { 609 btrfs_err(trans->fs_info, 610 "unrecognized backref key (%llu %u %llu)", 611 key.objectid, key.type, key.offset); 612 btrfs_abort_transaction(trans, -EUCLEAN); 613 return -EUCLEAN; 614 } 615 616 BUG_ON(num_refs < refs_to_drop); 617 num_refs -= refs_to_drop; 618 619 if (num_refs == 0) { 620 ret = btrfs_del_item(trans, root, path); 621 } else { 622 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) 623 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs); 624 else if (key.type == BTRFS_SHARED_DATA_REF_KEY) 625 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs); 626 btrfs_mark_buffer_dirty(trans, leaf); 627 } 628 return ret; 629 } 630 631 static noinline u32 extent_data_ref_count(struct btrfs_path *path, 632 struct btrfs_extent_inline_ref *iref) 633 { 634 struct btrfs_key key; 635 struct extent_buffer *leaf; 636 struct btrfs_extent_data_ref *ref1; 637 struct btrfs_shared_data_ref *ref2; 638 u32 num_refs = 0; 639 int type; 640 641 leaf = path->nodes[0]; 642 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 643 644 if (iref) { 645 /* 646 * If type is invalid, we should have bailed out earlier than 647 * this call. 648 */ 649 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA); 650 ASSERT(type != BTRFS_REF_TYPE_INVALID); 651 if (type == BTRFS_EXTENT_DATA_REF_KEY) { 652 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset); 653 num_refs = btrfs_extent_data_ref_count(leaf, ref1); 654 } else { 655 ref2 = (struct btrfs_shared_data_ref *)(iref + 1); 656 num_refs = btrfs_shared_data_ref_count(leaf, ref2); 657 } 658 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) { 659 ref1 = btrfs_item_ptr(leaf, path->slots[0], 660 struct btrfs_extent_data_ref); 661 num_refs = btrfs_extent_data_ref_count(leaf, ref1); 662 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) { 663 ref2 = btrfs_item_ptr(leaf, path->slots[0], 664 struct btrfs_shared_data_ref); 665 num_refs = btrfs_shared_data_ref_count(leaf, ref2); 666 } else { 667 WARN_ON(1); 668 } 669 return num_refs; 670 } 671 672 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans, 673 struct btrfs_path *path, 674 u64 bytenr, u64 parent, 675 u64 root_objectid) 676 { 677 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr); 678 struct btrfs_key key; 679 int ret; 680 681 key.objectid = bytenr; 682 if (parent) { 683 key.type = BTRFS_SHARED_BLOCK_REF_KEY; 684 key.offset = parent; 685 } else { 686 key.type = BTRFS_TREE_BLOCK_REF_KEY; 687 key.offset = root_objectid; 688 } 689 690 ret = btrfs_search_slot(trans, root, &key, path, -1, 1); 691 if (ret > 0) 692 ret = -ENOENT; 693 return ret; 694 } 695 696 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans, 697 struct btrfs_path *path, 698 u64 bytenr, u64 parent, 699 u64 root_objectid) 700 { 701 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr); 702 struct btrfs_key key; 703 int ret; 704 705 key.objectid = bytenr; 706 if (parent) { 707 key.type = BTRFS_SHARED_BLOCK_REF_KEY; 708 key.offset = parent; 709 } else { 710 key.type = BTRFS_TREE_BLOCK_REF_KEY; 711 key.offset = root_objectid; 712 } 713 714 ret = btrfs_insert_empty_item(trans, root, path, &key, 0); 715 btrfs_release_path(path); 716 return ret; 717 } 718 719 static inline int extent_ref_type(u64 parent, u64 owner) 720 { 721 int type; 722 if (owner < BTRFS_FIRST_FREE_OBJECTID) { 723 if (parent > 0) 724 type = BTRFS_SHARED_BLOCK_REF_KEY; 725 else 726 type = BTRFS_TREE_BLOCK_REF_KEY; 727 } else { 728 if (parent > 0) 729 type = BTRFS_SHARED_DATA_REF_KEY; 730 else 731 type = BTRFS_EXTENT_DATA_REF_KEY; 732 } 733 return type; 734 } 735 736 static int find_next_key(struct btrfs_path *path, int level, 737 struct btrfs_key *key) 738 739 { 740 for (; level < BTRFS_MAX_LEVEL; level++) { 741 if (!path->nodes[level]) 742 break; 743 if (path->slots[level] + 1 >= 744 btrfs_header_nritems(path->nodes[level])) 745 continue; 746 if (level == 0) 747 btrfs_item_key_to_cpu(path->nodes[level], key, 748 path->slots[level] + 1); 749 else 750 btrfs_node_key_to_cpu(path->nodes[level], key, 751 path->slots[level] + 1); 752 return 0; 753 } 754 return 1; 755 } 756 757 /* 758 * look for inline back ref. if back ref is found, *ref_ret is set 759 * to the address of inline back ref, and 0 is returned. 760 * 761 * if back ref isn't found, *ref_ret is set to the address where it 762 * should be inserted, and -ENOENT is returned. 763 * 764 * if insert is true and there are too many inline back refs, the path 765 * points to the extent item, and -EAGAIN is returned. 766 * 767 * NOTE: inline back refs are ordered in the same way that back ref 768 * items in the tree are ordered. 769 */ 770 static noinline_for_stack 771 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans, 772 struct btrfs_path *path, 773 struct btrfs_extent_inline_ref **ref_ret, 774 u64 bytenr, u64 num_bytes, 775 u64 parent, u64 root_objectid, 776 u64 owner, u64 offset, int insert) 777 { 778 struct btrfs_fs_info *fs_info = trans->fs_info; 779 struct btrfs_root *root = btrfs_extent_root(fs_info, bytenr); 780 struct btrfs_key key; 781 struct extent_buffer *leaf; 782 struct btrfs_extent_item *ei; 783 struct btrfs_extent_inline_ref *iref; 784 u64 flags; 785 u64 item_size; 786 unsigned long ptr; 787 unsigned long end; 788 int extra_size; 789 int type; 790 int want; 791 int ret; 792 int err = 0; 793 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA); 794 int needed; 795 796 key.objectid = bytenr; 797 key.type = BTRFS_EXTENT_ITEM_KEY; 798 key.offset = num_bytes; 799 800 want = extent_ref_type(parent, owner); 801 if (insert) { 802 extra_size = btrfs_extent_inline_ref_size(want); 803 path->search_for_extension = 1; 804 path->keep_locks = 1; 805 } else 806 extra_size = -1; 807 808 /* 809 * Owner is our level, so we can just add one to get the level for the 810 * block we are interested in. 811 */ 812 if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) { 813 key.type = BTRFS_METADATA_ITEM_KEY; 814 key.offset = owner; 815 } 816 817 again: 818 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1); 819 if (ret < 0) { 820 err = ret; 821 goto out; 822 } 823 824 /* 825 * We may be a newly converted file system which still has the old fat 826 * extent entries for metadata, so try and see if we have one of those. 827 */ 828 if (ret > 0 && skinny_metadata) { 829 skinny_metadata = false; 830 if (path->slots[0]) { 831 path->slots[0]--; 832 btrfs_item_key_to_cpu(path->nodes[0], &key, 833 path->slots[0]); 834 if (key.objectid == bytenr && 835 key.type == BTRFS_EXTENT_ITEM_KEY && 836 key.offset == num_bytes) 837 ret = 0; 838 } 839 if (ret) { 840 key.objectid = bytenr; 841 key.type = BTRFS_EXTENT_ITEM_KEY; 842 key.offset = num_bytes; 843 btrfs_release_path(path); 844 goto again; 845 } 846 } 847 848 if (ret && !insert) { 849 err = -ENOENT; 850 goto out; 851 } else if (WARN_ON(ret)) { 852 btrfs_print_leaf(path->nodes[0]); 853 btrfs_err(fs_info, 854 "extent item not found for insert, bytenr %llu num_bytes %llu parent %llu root_objectid %llu owner %llu offset %llu", 855 bytenr, num_bytes, parent, root_objectid, owner, 856 offset); 857 err = -EIO; 858 goto out; 859 } 860 861 leaf = path->nodes[0]; 862 item_size = btrfs_item_size(leaf, path->slots[0]); 863 if (unlikely(item_size < sizeof(*ei))) { 864 err = -EUCLEAN; 865 btrfs_err(fs_info, 866 "unexpected extent item size, has %llu expect >= %zu", 867 item_size, sizeof(*ei)); 868 btrfs_abort_transaction(trans, err); 869 goto out; 870 } 871 872 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 873 flags = btrfs_extent_flags(leaf, ei); 874 875 ptr = (unsigned long)(ei + 1); 876 end = (unsigned long)ei + item_size; 877 878 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) { 879 ptr += sizeof(struct btrfs_tree_block_info); 880 BUG_ON(ptr > end); 881 } 882 883 if (owner >= BTRFS_FIRST_FREE_OBJECTID) 884 needed = BTRFS_REF_TYPE_DATA; 885 else 886 needed = BTRFS_REF_TYPE_BLOCK; 887 888 err = -ENOENT; 889 while (1) { 890 if (ptr >= end) { 891 if (ptr > end) { 892 err = -EUCLEAN; 893 btrfs_print_leaf(path->nodes[0]); 894 btrfs_crit(fs_info, 895 "overrun extent record at slot %d while looking for inline extent for root %llu owner %llu offset %llu parent %llu", 896 path->slots[0], root_objectid, owner, offset, parent); 897 } 898 break; 899 } 900 iref = (struct btrfs_extent_inline_ref *)ptr; 901 type = btrfs_get_extent_inline_ref_type(leaf, iref, needed); 902 if (type == BTRFS_REF_TYPE_INVALID) { 903 err = -EUCLEAN; 904 goto out; 905 } 906 907 if (want < type) 908 break; 909 if (want > type) { 910 ptr += btrfs_extent_inline_ref_size(type); 911 continue; 912 } 913 914 if (type == BTRFS_EXTENT_DATA_REF_KEY) { 915 struct btrfs_extent_data_ref *dref; 916 dref = (struct btrfs_extent_data_ref *)(&iref->offset); 917 if (match_extent_data_ref(leaf, dref, root_objectid, 918 owner, offset)) { 919 err = 0; 920 break; 921 } 922 if (hash_extent_data_ref_item(leaf, dref) < 923 hash_extent_data_ref(root_objectid, owner, offset)) 924 break; 925 } else { 926 u64 ref_offset; 927 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref); 928 if (parent > 0) { 929 if (parent == ref_offset) { 930 err = 0; 931 break; 932 } 933 if (ref_offset < parent) 934 break; 935 } else { 936 if (root_objectid == ref_offset) { 937 err = 0; 938 break; 939 } 940 if (ref_offset < root_objectid) 941 break; 942 } 943 } 944 ptr += btrfs_extent_inline_ref_size(type); 945 } 946 if (err == -ENOENT && insert) { 947 if (item_size + extra_size >= 948 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) { 949 err = -EAGAIN; 950 goto out; 951 } 952 /* 953 * To add new inline back ref, we have to make sure 954 * there is no corresponding back ref item. 955 * For simplicity, we just do not add new inline back 956 * ref if there is any kind of item for this block 957 */ 958 if (find_next_key(path, 0, &key) == 0 && 959 key.objectid == bytenr && 960 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) { 961 err = -EAGAIN; 962 goto out; 963 } 964 } 965 *ref_ret = (struct btrfs_extent_inline_ref *)ptr; 966 out: 967 if (insert) { 968 path->keep_locks = 0; 969 path->search_for_extension = 0; 970 btrfs_unlock_up_safe(path, 1); 971 } 972 return err; 973 } 974 975 /* 976 * helper to add new inline back ref 977 */ 978 static noinline_for_stack 979 void setup_inline_extent_backref(struct btrfs_trans_handle *trans, 980 struct btrfs_path *path, 981 struct btrfs_extent_inline_ref *iref, 982 u64 parent, u64 root_objectid, 983 u64 owner, u64 offset, int refs_to_add, 984 struct btrfs_delayed_extent_op *extent_op) 985 { 986 struct extent_buffer *leaf; 987 struct btrfs_extent_item *ei; 988 unsigned long ptr; 989 unsigned long end; 990 unsigned long item_offset; 991 u64 refs; 992 int size; 993 int type; 994 995 leaf = path->nodes[0]; 996 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 997 item_offset = (unsigned long)iref - (unsigned long)ei; 998 999 type = extent_ref_type(parent, owner); 1000 size = btrfs_extent_inline_ref_size(type); 1001 1002 btrfs_extend_item(trans, path, size); 1003 1004 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 1005 refs = btrfs_extent_refs(leaf, ei); 1006 refs += refs_to_add; 1007 btrfs_set_extent_refs(leaf, ei, refs); 1008 if (extent_op) 1009 __run_delayed_extent_op(extent_op, leaf, ei); 1010 1011 ptr = (unsigned long)ei + item_offset; 1012 end = (unsigned long)ei + btrfs_item_size(leaf, path->slots[0]); 1013 if (ptr < end - size) 1014 memmove_extent_buffer(leaf, ptr + size, ptr, 1015 end - size - ptr); 1016 1017 iref = (struct btrfs_extent_inline_ref *)ptr; 1018 btrfs_set_extent_inline_ref_type(leaf, iref, type); 1019 if (type == BTRFS_EXTENT_DATA_REF_KEY) { 1020 struct btrfs_extent_data_ref *dref; 1021 dref = (struct btrfs_extent_data_ref *)(&iref->offset); 1022 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid); 1023 btrfs_set_extent_data_ref_objectid(leaf, dref, owner); 1024 btrfs_set_extent_data_ref_offset(leaf, dref, offset); 1025 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add); 1026 } else if (type == BTRFS_SHARED_DATA_REF_KEY) { 1027 struct btrfs_shared_data_ref *sref; 1028 sref = (struct btrfs_shared_data_ref *)(iref + 1); 1029 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add); 1030 btrfs_set_extent_inline_ref_offset(leaf, iref, parent); 1031 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) { 1032 btrfs_set_extent_inline_ref_offset(leaf, iref, parent); 1033 } else { 1034 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid); 1035 } 1036 btrfs_mark_buffer_dirty(trans, leaf); 1037 } 1038 1039 static int lookup_extent_backref(struct btrfs_trans_handle *trans, 1040 struct btrfs_path *path, 1041 struct btrfs_extent_inline_ref **ref_ret, 1042 u64 bytenr, u64 num_bytes, u64 parent, 1043 u64 root_objectid, u64 owner, u64 offset) 1044 { 1045 int ret; 1046 1047 ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr, 1048 num_bytes, parent, root_objectid, 1049 owner, offset, 0); 1050 if (ret != -ENOENT) 1051 return ret; 1052 1053 btrfs_release_path(path); 1054 *ref_ret = NULL; 1055 1056 if (owner < BTRFS_FIRST_FREE_OBJECTID) { 1057 ret = lookup_tree_block_ref(trans, path, bytenr, parent, 1058 root_objectid); 1059 } else { 1060 ret = lookup_extent_data_ref(trans, path, bytenr, parent, 1061 root_objectid, owner, offset); 1062 } 1063 return ret; 1064 } 1065 1066 /* 1067 * helper to update/remove inline back ref 1068 */ 1069 static noinline_for_stack int update_inline_extent_backref( 1070 struct btrfs_trans_handle *trans, 1071 struct btrfs_path *path, 1072 struct btrfs_extent_inline_ref *iref, 1073 int refs_to_mod, 1074 struct btrfs_delayed_extent_op *extent_op) 1075 { 1076 struct extent_buffer *leaf = path->nodes[0]; 1077 struct btrfs_fs_info *fs_info = leaf->fs_info; 1078 struct btrfs_extent_item *ei; 1079 struct btrfs_extent_data_ref *dref = NULL; 1080 struct btrfs_shared_data_ref *sref = NULL; 1081 unsigned long ptr; 1082 unsigned long end; 1083 u32 item_size; 1084 int size; 1085 int type; 1086 u64 refs; 1087 1088 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 1089 refs = btrfs_extent_refs(leaf, ei); 1090 if (unlikely(refs_to_mod < 0 && refs + refs_to_mod <= 0)) { 1091 struct btrfs_key key; 1092 u32 extent_size; 1093 1094 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 1095 if (key.type == BTRFS_METADATA_ITEM_KEY) 1096 extent_size = fs_info->nodesize; 1097 else 1098 extent_size = key.offset; 1099 btrfs_print_leaf(leaf); 1100 btrfs_err(fs_info, 1101 "invalid refs_to_mod for extent %llu num_bytes %u, has %d expect >= -%llu", 1102 key.objectid, extent_size, refs_to_mod, refs); 1103 return -EUCLEAN; 1104 } 1105 refs += refs_to_mod; 1106 btrfs_set_extent_refs(leaf, ei, refs); 1107 if (extent_op) 1108 __run_delayed_extent_op(extent_op, leaf, ei); 1109 1110 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY); 1111 /* 1112 * Function btrfs_get_extent_inline_ref_type() has already printed 1113 * error messages. 1114 */ 1115 if (unlikely(type == BTRFS_REF_TYPE_INVALID)) 1116 return -EUCLEAN; 1117 1118 if (type == BTRFS_EXTENT_DATA_REF_KEY) { 1119 dref = (struct btrfs_extent_data_ref *)(&iref->offset); 1120 refs = btrfs_extent_data_ref_count(leaf, dref); 1121 } else if (type == BTRFS_SHARED_DATA_REF_KEY) { 1122 sref = (struct btrfs_shared_data_ref *)(iref + 1); 1123 refs = btrfs_shared_data_ref_count(leaf, sref); 1124 } else { 1125 refs = 1; 1126 /* 1127 * For tree blocks we can only drop one ref for it, and tree 1128 * blocks should not have refs > 1. 1129 * 1130 * Furthermore if we're inserting a new inline backref, we 1131 * won't reach this path either. That would be 1132 * setup_inline_extent_backref(). 1133 */ 1134 if (unlikely(refs_to_mod != -1)) { 1135 struct btrfs_key key; 1136 1137 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 1138 1139 btrfs_print_leaf(leaf); 1140 btrfs_err(fs_info, 1141 "invalid refs_to_mod for tree block %llu, has %d expect -1", 1142 key.objectid, refs_to_mod); 1143 return -EUCLEAN; 1144 } 1145 } 1146 1147 if (unlikely(refs_to_mod < 0 && refs < -refs_to_mod)) { 1148 struct btrfs_key key; 1149 u32 extent_size; 1150 1151 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 1152 if (key.type == BTRFS_METADATA_ITEM_KEY) 1153 extent_size = fs_info->nodesize; 1154 else 1155 extent_size = key.offset; 1156 btrfs_print_leaf(leaf); 1157 btrfs_err(fs_info, 1158 "invalid refs_to_mod for backref entry, iref %lu extent %llu num_bytes %u, has %d expect >= -%llu", 1159 (unsigned long)iref, key.objectid, extent_size, 1160 refs_to_mod, refs); 1161 return -EUCLEAN; 1162 } 1163 refs += refs_to_mod; 1164 1165 if (refs > 0) { 1166 if (type == BTRFS_EXTENT_DATA_REF_KEY) 1167 btrfs_set_extent_data_ref_count(leaf, dref, refs); 1168 else 1169 btrfs_set_shared_data_ref_count(leaf, sref, refs); 1170 } else { 1171 size = btrfs_extent_inline_ref_size(type); 1172 item_size = btrfs_item_size(leaf, path->slots[0]); 1173 ptr = (unsigned long)iref; 1174 end = (unsigned long)ei + item_size; 1175 if (ptr + size < end) 1176 memmove_extent_buffer(leaf, ptr, ptr + size, 1177 end - ptr - size); 1178 item_size -= size; 1179 btrfs_truncate_item(trans, path, item_size, 1); 1180 } 1181 btrfs_mark_buffer_dirty(trans, leaf); 1182 return 0; 1183 } 1184 1185 static noinline_for_stack 1186 int insert_inline_extent_backref(struct btrfs_trans_handle *trans, 1187 struct btrfs_path *path, 1188 u64 bytenr, u64 num_bytes, u64 parent, 1189 u64 root_objectid, u64 owner, 1190 u64 offset, int refs_to_add, 1191 struct btrfs_delayed_extent_op *extent_op) 1192 { 1193 struct btrfs_extent_inline_ref *iref; 1194 int ret; 1195 1196 ret = lookup_inline_extent_backref(trans, path, &iref, bytenr, 1197 num_bytes, parent, root_objectid, 1198 owner, offset, 1); 1199 if (ret == 0) { 1200 /* 1201 * We're adding refs to a tree block we already own, this 1202 * should not happen at all. 1203 */ 1204 if (owner < BTRFS_FIRST_FREE_OBJECTID) { 1205 btrfs_print_leaf(path->nodes[0]); 1206 btrfs_crit(trans->fs_info, 1207 "adding refs to an existing tree ref, bytenr %llu num_bytes %llu root_objectid %llu slot %u", 1208 bytenr, num_bytes, root_objectid, path->slots[0]); 1209 return -EUCLEAN; 1210 } 1211 ret = update_inline_extent_backref(trans, path, iref, 1212 refs_to_add, extent_op); 1213 } else if (ret == -ENOENT) { 1214 setup_inline_extent_backref(trans, path, iref, parent, 1215 root_objectid, owner, offset, 1216 refs_to_add, extent_op); 1217 ret = 0; 1218 } 1219 return ret; 1220 } 1221 1222 static int remove_extent_backref(struct btrfs_trans_handle *trans, 1223 struct btrfs_root *root, 1224 struct btrfs_path *path, 1225 struct btrfs_extent_inline_ref *iref, 1226 int refs_to_drop, int is_data) 1227 { 1228 int ret = 0; 1229 1230 BUG_ON(!is_data && refs_to_drop != 1); 1231 if (iref) 1232 ret = update_inline_extent_backref(trans, path, iref, 1233 -refs_to_drop, NULL); 1234 else if (is_data) 1235 ret = remove_extent_data_ref(trans, root, path, refs_to_drop); 1236 else 1237 ret = btrfs_del_item(trans, root, path); 1238 return ret; 1239 } 1240 1241 static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len, 1242 u64 *discarded_bytes) 1243 { 1244 int j, ret = 0; 1245 u64 bytes_left, end; 1246 u64 aligned_start = ALIGN(start, 1 << SECTOR_SHIFT); 1247 1248 /* Adjust the range to be aligned to 512B sectors if necessary. */ 1249 if (start != aligned_start) { 1250 len -= aligned_start - start; 1251 len = round_down(len, 1 << SECTOR_SHIFT); 1252 start = aligned_start; 1253 } 1254 1255 *discarded_bytes = 0; 1256 1257 if (!len) 1258 return 0; 1259 1260 end = start + len; 1261 bytes_left = len; 1262 1263 /* Skip any superblocks on this device. */ 1264 for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) { 1265 u64 sb_start = btrfs_sb_offset(j); 1266 u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE; 1267 u64 size = sb_start - start; 1268 1269 if (!in_range(sb_start, start, bytes_left) && 1270 !in_range(sb_end, start, bytes_left) && 1271 !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE)) 1272 continue; 1273 1274 /* 1275 * Superblock spans beginning of range. Adjust start and 1276 * try again. 1277 */ 1278 if (sb_start <= start) { 1279 start += sb_end - start; 1280 if (start > end) { 1281 bytes_left = 0; 1282 break; 1283 } 1284 bytes_left = end - start; 1285 continue; 1286 } 1287 1288 if (size) { 1289 ret = blkdev_issue_discard(bdev, start >> SECTOR_SHIFT, 1290 size >> SECTOR_SHIFT, 1291 GFP_NOFS); 1292 if (!ret) 1293 *discarded_bytes += size; 1294 else if (ret != -EOPNOTSUPP) 1295 return ret; 1296 } 1297 1298 start = sb_end; 1299 if (start > end) { 1300 bytes_left = 0; 1301 break; 1302 } 1303 bytes_left = end - start; 1304 } 1305 1306 while (bytes_left) { 1307 u64 bytes_to_discard = min(BTRFS_MAX_DISCARD_CHUNK_SIZE, bytes_left); 1308 1309 ret = blkdev_issue_discard(bdev, start >> SECTOR_SHIFT, 1310 bytes_to_discard >> SECTOR_SHIFT, 1311 GFP_NOFS); 1312 1313 if (ret) { 1314 if (ret != -EOPNOTSUPP) 1315 break; 1316 continue; 1317 } 1318 1319 start += bytes_to_discard; 1320 bytes_left -= bytes_to_discard; 1321 *discarded_bytes += bytes_to_discard; 1322 1323 if (btrfs_trim_interrupted()) { 1324 ret = -ERESTARTSYS; 1325 break; 1326 } 1327 } 1328 1329 return ret; 1330 } 1331 1332 static int do_discard_extent(struct btrfs_discard_stripe *stripe, u64 *bytes) 1333 { 1334 struct btrfs_device *dev = stripe->dev; 1335 struct btrfs_fs_info *fs_info = dev->fs_info; 1336 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace; 1337 u64 phys = stripe->physical; 1338 u64 len = stripe->length; 1339 u64 discarded = 0; 1340 int ret = 0; 1341 1342 /* Zone reset on a zoned filesystem */ 1343 if (btrfs_can_zone_reset(dev, phys, len)) { 1344 u64 src_disc; 1345 1346 ret = btrfs_reset_device_zone(dev, phys, len, &discarded); 1347 if (ret) 1348 goto out; 1349 1350 if (!btrfs_dev_replace_is_ongoing(dev_replace) || 1351 dev != dev_replace->srcdev) 1352 goto out; 1353 1354 src_disc = discarded; 1355 1356 /* Send to replace target as well */ 1357 ret = btrfs_reset_device_zone(dev_replace->tgtdev, phys, len, 1358 &discarded); 1359 discarded += src_disc; 1360 } else if (bdev_max_discard_sectors(stripe->dev->bdev)) { 1361 ret = btrfs_issue_discard(dev->bdev, phys, len, &discarded); 1362 } else { 1363 ret = 0; 1364 *bytes = 0; 1365 } 1366 1367 out: 1368 *bytes = discarded; 1369 return ret; 1370 } 1371 1372 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr, 1373 u64 num_bytes, u64 *actual_bytes) 1374 { 1375 int ret = 0; 1376 u64 discarded_bytes = 0; 1377 u64 end = bytenr + num_bytes; 1378 u64 cur = bytenr; 1379 1380 /* 1381 * Avoid races with device replace and make sure the devices in the 1382 * stripes don't go away while we are discarding. 1383 */ 1384 btrfs_bio_counter_inc_blocked(fs_info); 1385 while (cur < end) { 1386 struct btrfs_discard_stripe *stripes; 1387 unsigned int num_stripes; 1388 int i; 1389 1390 num_bytes = end - cur; 1391 stripes = btrfs_map_discard(fs_info, cur, &num_bytes, &num_stripes); 1392 if (IS_ERR(stripes)) { 1393 ret = PTR_ERR(stripes); 1394 if (ret == -EOPNOTSUPP) 1395 ret = 0; 1396 break; 1397 } 1398 1399 for (i = 0; i < num_stripes; i++) { 1400 struct btrfs_discard_stripe *stripe = stripes + i; 1401 u64 bytes; 1402 1403 if (!stripe->dev->bdev) { 1404 ASSERT(btrfs_test_opt(fs_info, DEGRADED)); 1405 continue; 1406 } 1407 1408 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, 1409 &stripe->dev->dev_state)) 1410 continue; 1411 1412 ret = do_discard_extent(stripe, &bytes); 1413 if (ret) { 1414 /* 1415 * Keep going if discard is not supported by the 1416 * device. 1417 */ 1418 if (ret != -EOPNOTSUPP) 1419 break; 1420 ret = 0; 1421 } else { 1422 discarded_bytes += bytes; 1423 } 1424 } 1425 kfree(stripes); 1426 if (ret) 1427 break; 1428 cur += num_bytes; 1429 } 1430 btrfs_bio_counter_dec(fs_info); 1431 if (actual_bytes) 1432 *actual_bytes = discarded_bytes; 1433 return ret; 1434 } 1435 1436 /* Can return -ENOMEM */ 1437 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, 1438 struct btrfs_ref *generic_ref) 1439 { 1440 struct btrfs_fs_info *fs_info = trans->fs_info; 1441 int ret; 1442 1443 ASSERT(generic_ref->type != BTRFS_REF_NOT_SET && 1444 generic_ref->action); 1445 BUG_ON(generic_ref->type == BTRFS_REF_METADATA && 1446 generic_ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID); 1447 1448 if (generic_ref->type == BTRFS_REF_METADATA) 1449 ret = btrfs_add_delayed_tree_ref(trans, generic_ref, NULL); 1450 else 1451 ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0); 1452 1453 btrfs_ref_tree_mod(fs_info, generic_ref); 1454 1455 return ret; 1456 } 1457 1458 /* 1459 * __btrfs_inc_extent_ref - insert backreference for a given extent 1460 * 1461 * The counterpart is in __btrfs_free_extent(), with examples and more details 1462 * how it works. 1463 * 1464 * @trans: Handle of transaction 1465 * 1466 * @node: The delayed ref node used to get the bytenr/length for 1467 * extent whose references are incremented. 1468 * 1469 * @parent: If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/ 1470 * BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical 1471 * bytenr of the parent block. Since new extents are always 1472 * created with indirect references, this will only be the case 1473 * when relocating a shared extent. In that case, root_objectid 1474 * will be BTRFS_TREE_RELOC_OBJECTID. Otherwise, parent must 1475 * be 0 1476 * 1477 * @root_objectid: The id of the root where this modification has originated, 1478 * this can be either one of the well-known metadata trees or 1479 * the subvolume id which references this extent. 1480 * 1481 * @owner: For data extents it is the inode number of the owning file. 1482 * For metadata extents this parameter holds the level in the 1483 * tree of the extent. 1484 * 1485 * @offset: For metadata extents the offset is ignored and is currently 1486 * always passed as 0. For data extents it is the fileoffset 1487 * this extent belongs to. 1488 * 1489 * @refs_to_add Number of references to add 1490 * 1491 * @extent_op Pointer to a structure, holding information necessary when 1492 * updating a tree block's flags 1493 * 1494 */ 1495 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans, 1496 struct btrfs_delayed_ref_node *node, 1497 u64 parent, u64 root_objectid, 1498 u64 owner, u64 offset, int refs_to_add, 1499 struct btrfs_delayed_extent_op *extent_op) 1500 { 1501 struct btrfs_path *path; 1502 struct extent_buffer *leaf; 1503 struct btrfs_extent_item *item; 1504 struct btrfs_key key; 1505 u64 bytenr = node->bytenr; 1506 u64 num_bytes = node->num_bytes; 1507 u64 refs; 1508 int ret; 1509 1510 path = btrfs_alloc_path(); 1511 if (!path) 1512 return -ENOMEM; 1513 1514 /* this will setup the path even if it fails to insert the back ref */ 1515 ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes, 1516 parent, root_objectid, owner, 1517 offset, refs_to_add, extent_op); 1518 if ((ret < 0 && ret != -EAGAIN) || !ret) 1519 goto out; 1520 1521 /* 1522 * Ok we had -EAGAIN which means we didn't have space to insert and 1523 * inline extent ref, so just update the reference count and add a 1524 * normal backref. 1525 */ 1526 leaf = path->nodes[0]; 1527 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 1528 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 1529 refs = btrfs_extent_refs(leaf, item); 1530 btrfs_set_extent_refs(leaf, item, refs + refs_to_add); 1531 if (extent_op) 1532 __run_delayed_extent_op(extent_op, leaf, item); 1533 1534 btrfs_mark_buffer_dirty(trans, leaf); 1535 btrfs_release_path(path); 1536 1537 /* now insert the actual backref */ 1538 if (owner < BTRFS_FIRST_FREE_OBJECTID) 1539 ret = insert_tree_block_ref(trans, path, bytenr, parent, 1540 root_objectid); 1541 else 1542 ret = insert_extent_data_ref(trans, path, bytenr, parent, 1543 root_objectid, owner, offset, 1544 refs_to_add); 1545 1546 if (ret) 1547 btrfs_abort_transaction(trans, ret); 1548 out: 1549 btrfs_free_path(path); 1550 return ret; 1551 } 1552 1553 static int run_delayed_data_ref(struct btrfs_trans_handle *trans, 1554 struct btrfs_delayed_ref_node *node, 1555 struct btrfs_delayed_extent_op *extent_op, 1556 bool insert_reserved) 1557 { 1558 int ret = 0; 1559 struct btrfs_delayed_data_ref *ref; 1560 struct btrfs_key ins; 1561 u64 parent = 0; 1562 u64 ref_root = 0; 1563 u64 flags = 0; 1564 1565 ins.objectid = node->bytenr; 1566 ins.offset = node->num_bytes; 1567 ins.type = BTRFS_EXTENT_ITEM_KEY; 1568 1569 ref = btrfs_delayed_node_to_data_ref(node); 1570 trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action); 1571 1572 if (node->type == BTRFS_SHARED_DATA_REF_KEY) 1573 parent = ref->parent; 1574 ref_root = ref->root; 1575 1576 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) { 1577 if (extent_op) 1578 flags |= extent_op->flags_to_set; 1579 ret = alloc_reserved_file_extent(trans, parent, ref_root, 1580 flags, ref->objectid, 1581 ref->offset, &ins, 1582 node->ref_mod); 1583 } else if (node->action == BTRFS_ADD_DELAYED_REF) { 1584 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root, 1585 ref->objectid, ref->offset, 1586 node->ref_mod, extent_op); 1587 } else if (node->action == BTRFS_DROP_DELAYED_REF) { 1588 ret = __btrfs_free_extent(trans, node, parent, 1589 ref_root, ref->objectid, 1590 ref->offset, node->ref_mod, 1591 extent_op); 1592 } else { 1593 BUG(); 1594 } 1595 return ret; 1596 } 1597 1598 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op, 1599 struct extent_buffer *leaf, 1600 struct btrfs_extent_item *ei) 1601 { 1602 u64 flags = btrfs_extent_flags(leaf, ei); 1603 if (extent_op->update_flags) { 1604 flags |= extent_op->flags_to_set; 1605 btrfs_set_extent_flags(leaf, ei, flags); 1606 } 1607 1608 if (extent_op->update_key) { 1609 struct btrfs_tree_block_info *bi; 1610 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)); 1611 bi = (struct btrfs_tree_block_info *)(ei + 1); 1612 btrfs_set_tree_block_key(leaf, bi, &extent_op->key); 1613 } 1614 } 1615 1616 static int run_delayed_extent_op(struct btrfs_trans_handle *trans, 1617 struct btrfs_delayed_ref_head *head, 1618 struct btrfs_delayed_extent_op *extent_op) 1619 { 1620 struct btrfs_fs_info *fs_info = trans->fs_info; 1621 struct btrfs_root *root; 1622 struct btrfs_key key; 1623 struct btrfs_path *path; 1624 struct btrfs_extent_item *ei; 1625 struct extent_buffer *leaf; 1626 u32 item_size; 1627 int ret; 1628 int err = 0; 1629 int metadata = 1; 1630 1631 if (TRANS_ABORTED(trans)) 1632 return 0; 1633 1634 if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA)) 1635 metadata = 0; 1636 1637 path = btrfs_alloc_path(); 1638 if (!path) 1639 return -ENOMEM; 1640 1641 key.objectid = head->bytenr; 1642 1643 if (metadata) { 1644 key.type = BTRFS_METADATA_ITEM_KEY; 1645 key.offset = extent_op->level; 1646 } else { 1647 key.type = BTRFS_EXTENT_ITEM_KEY; 1648 key.offset = head->num_bytes; 1649 } 1650 1651 root = btrfs_extent_root(fs_info, key.objectid); 1652 again: 1653 ret = btrfs_search_slot(trans, root, &key, path, 0, 1); 1654 if (ret < 0) { 1655 err = ret; 1656 goto out; 1657 } 1658 if (ret > 0) { 1659 if (metadata) { 1660 if (path->slots[0] > 0) { 1661 path->slots[0]--; 1662 btrfs_item_key_to_cpu(path->nodes[0], &key, 1663 path->slots[0]); 1664 if (key.objectid == head->bytenr && 1665 key.type == BTRFS_EXTENT_ITEM_KEY && 1666 key.offset == head->num_bytes) 1667 ret = 0; 1668 } 1669 if (ret > 0) { 1670 btrfs_release_path(path); 1671 metadata = 0; 1672 1673 key.objectid = head->bytenr; 1674 key.offset = head->num_bytes; 1675 key.type = BTRFS_EXTENT_ITEM_KEY; 1676 goto again; 1677 } 1678 } else { 1679 err = -EUCLEAN; 1680 btrfs_err(fs_info, 1681 "missing extent item for extent %llu num_bytes %llu level %d", 1682 head->bytenr, head->num_bytes, extent_op->level); 1683 goto out; 1684 } 1685 } 1686 1687 leaf = path->nodes[0]; 1688 item_size = btrfs_item_size(leaf, path->slots[0]); 1689 1690 if (unlikely(item_size < sizeof(*ei))) { 1691 err = -EUCLEAN; 1692 btrfs_err(fs_info, 1693 "unexpected extent item size, has %u expect >= %zu", 1694 item_size, sizeof(*ei)); 1695 btrfs_abort_transaction(trans, err); 1696 goto out; 1697 } 1698 1699 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 1700 __run_delayed_extent_op(extent_op, leaf, ei); 1701 1702 btrfs_mark_buffer_dirty(trans, leaf); 1703 out: 1704 btrfs_free_path(path); 1705 return err; 1706 } 1707 1708 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans, 1709 struct btrfs_delayed_ref_node *node, 1710 struct btrfs_delayed_extent_op *extent_op, 1711 bool insert_reserved) 1712 { 1713 int ret = 0; 1714 struct btrfs_delayed_tree_ref *ref; 1715 u64 parent = 0; 1716 u64 ref_root = 0; 1717 1718 ref = btrfs_delayed_node_to_tree_ref(node); 1719 trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action); 1720 1721 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) 1722 parent = ref->parent; 1723 ref_root = ref->root; 1724 1725 if (unlikely(node->ref_mod != 1)) { 1726 btrfs_err(trans->fs_info, 1727 "btree block %llu has %d references rather than 1: action %d ref_root %llu parent %llu", 1728 node->bytenr, node->ref_mod, node->action, ref_root, 1729 parent); 1730 return -EUCLEAN; 1731 } 1732 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) { 1733 BUG_ON(!extent_op || !extent_op->update_flags); 1734 ret = alloc_reserved_tree_block(trans, node, extent_op); 1735 } else if (node->action == BTRFS_ADD_DELAYED_REF) { 1736 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root, 1737 ref->level, 0, 1, extent_op); 1738 } else if (node->action == BTRFS_DROP_DELAYED_REF) { 1739 ret = __btrfs_free_extent(trans, node, parent, ref_root, 1740 ref->level, 0, 1, extent_op); 1741 } else { 1742 BUG(); 1743 } 1744 return ret; 1745 } 1746 1747 /* helper function to actually process a single delayed ref entry */ 1748 static int run_one_delayed_ref(struct btrfs_trans_handle *trans, 1749 struct btrfs_delayed_ref_node *node, 1750 struct btrfs_delayed_extent_op *extent_op, 1751 bool insert_reserved) 1752 { 1753 int ret = 0; 1754 1755 if (TRANS_ABORTED(trans)) { 1756 if (insert_reserved) 1757 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1); 1758 return 0; 1759 } 1760 1761 if (node->type == BTRFS_TREE_BLOCK_REF_KEY || 1762 node->type == BTRFS_SHARED_BLOCK_REF_KEY) 1763 ret = run_delayed_tree_ref(trans, node, extent_op, 1764 insert_reserved); 1765 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY || 1766 node->type == BTRFS_SHARED_DATA_REF_KEY) 1767 ret = run_delayed_data_ref(trans, node, extent_op, 1768 insert_reserved); 1769 else 1770 BUG(); 1771 if (ret && insert_reserved) 1772 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1); 1773 if (ret < 0) 1774 btrfs_err(trans->fs_info, 1775 "failed to run delayed ref for logical %llu num_bytes %llu type %u action %u ref_mod %d: %d", 1776 node->bytenr, node->num_bytes, node->type, 1777 node->action, node->ref_mod, ret); 1778 return ret; 1779 } 1780 1781 static inline struct btrfs_delayed_ref_node * 1782 select_delayed_ref(struct btrfs_delayed_ref_head *head) 1783 { 1784 struct btrfs_delayed_ref_node *ref; 1785 1786 if (RB_EMPTY_ROOT(&head->ref_tree.rb_root)) 1787 return NULL; 1788 1789 /* 1790 * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first. 1791 * This is to prevent a ref count from going down to zero, which deletes 1792 * the extent item from the extent tree, when there still are references 1793 * to add, which would fail because they would not find the extent item. 1794 */ 1795 if (!list_empty(&head->ref_add_list)) 1796 return list_first_entry(&head->ref_add_list, 1797 struct btrfs_delayed_ref_node, add_list); 1798 1799 ref = rb_entry(rb_first_cached(&head->ref_tree), 1800 struct btrfs_delayed_ref_node, ref_node); 1801 ASSERT(list_empty(&ref->add_list)); 1802 return ref; 1803 } 1804 1805 static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs, 1806 struct btrfs_delayed_ref_head *head) 1807 { 1808 spin_lock(&delayed_refs->lock); 1809 head->processing = false; 1810 delayed_refs->num_heads_ready++; 1811 spin_unlock(&delayed_refs->lock); 1812 btrfs_delayed_ref_unlock(head); 1813 } 1814 1815 static struct btrfs_delayed_extent_op *cleanup_extent_op( 1816 struct btrfs_delayed_ref_head *head) 1817 { 1818 struct btrfs_delayed_extent_op *extent_op = head->extent_op; 1819 1820 if (!extent_op) 1821 return NULL; 1822 1823 if (head->must_insert_reserved) { 1824 head->extent_op = NULL; 1825 btrfs_free_delayed_extent_op(extent_op); 1826 return NULL; 1827 } 1828 return extent_op; 1829 } 1830 1831 static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans, 1832 struct btrfs_delayed_ref_head *head) 1833 { 1834 struct btrfs_delayed_extent_op *extent_op; 1835 int ret; 1836 1837 extent_op = cleanup_extent_op(head); 1838 if (!extent_op) 1839 return 0; 1840 head->extent_op = NULL; 1841 spin_unlock(&head->lock); 1842 ret = run_delayed_extent_op(trans, head, extent_op); 1843 btrfs_free_delayed_extent_op(extent_op); 1844 return ret ? ret : 1; 1845 } 1846 1847 void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info, 1848 struct btrfs_delayed_ref_root *delayed_refs, 1849 struct btrfs_delayed_ref_head *head) 1850 { 1851 int nr_items = 1; /* Dropping this ref head update. */ 1852 1853 /* 1854 * We had csum deletions accounted for in our delayed refs rsv, we need 1855 * to drop the csum leaves for this update from our delayed_refs_rsv. 1856 */ 1857 if (head->total_ref_mod < 0 && head->is_data) { 1858 spin_lock(&delayed_refs->lock); 1859 delayed_refs->pending_csums -= head->num_bytes; 1860 spin_unlock(&delayed_refs->lock); 1861 nr_items += btrfs_csum_bytes_to_leaves(fs_info, head->num_bytes); 1862 } 1863 1864 btrfs_delayed_refs_rsv_release(fs_info, nr_items); 1865 } 1866 1867 static int cleanup_ref_head(struct btrfs_trans_handle *trans, 1868 struct btrfs_delayed_ref_head *head) 1869 { 1870 1871 struct btrfs_fs_info *fs_info = trans->fs_info; 1872 struct btrfs_delayed_ref_root *delayed_refs; 1873 int ret; 1874 1875 delayed_refs = &trans->transaction->delayed_refs; 1876 1877 ret = run_and_cleanup_extent_op(trans, head); 1878 if (ret < 0) { 1879 unselect_delayed_ref_head(delayed_refs, head); 1880 btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret); 1881 return ret; 1882 } else if (ret) { 1883 return ret; 1884 } 1885 1886 /* 1887 * Need to drop our head ref lock and re-acquire the delayed ref lock 1888 * and then re-check to make sure nobody got added. 1889 */ 1890 spin_unlock(&head->lock); 1891 spin_lock(&delayed_refs->lock); 1892 spin_lock(&head->lock); 1893 if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) { 1894 spin_unlock(&head->lock); 1895 spin_unlock(&delayed_refs->lock); 1896 return 1; 1897 } 1898 btrfs_delete_ref_head(delayed_refs, head); 1899 spin_unlock(&head->lock); 1900 spin_unlock(&delayed_refs->lock); 1901 1902 if (head->must_insert_reserved) { 1903 btrfs_pin_extent(trans, head->bytenr, head->num_bytes, 1); 1904 if (head->is_data) { 1905 struct btrfs_root *csum_root; 1906 1907 csum_root = btrfs_csum_root(fs_info, head->bytenr); 1908 ret = btrfs_del_csums(trans, csum_root, head->bytenr, 1909 head->num_bytes); 1910 } 1911 } 1912 1913 btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head); 1914 1915 trace_run_delayed_ref_head(fs_info, head, 0); 1916 btrfs_delayed_ref_unlock(head); 1917 btrfs_put_delayed_ref_head(head); 1918 return ret; 1919 } 1920 1921 static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head( 1922 struct btrfs_trans_handle *trans) 1923 { 1924 struct btrfs_delayed_ref_root *delayed_refs = 1925 &trans->transaction->delayed_refs; 1926 struct btrfs_delayed_ref_head *head = NULL; 1927 int ret; 1928 1929 spin_lock(&delayed_refs->lock); 1930 head = btrfs_select_ref_head(delayed_refs); 1931 if (!head) { 1932 spin_unlock(&delayed_refs->lock); 1933 return head; 1934 } 1935 1936 /* 1937 * Grab the lock that says we are going to process all the refs for 1938 * this head 1939 */ 1940 ret = btrfs_delayed_ref_lock(delayed_refs, head); 1941 spin_unlock(&delayed_refs->lock); 1942 1943 /* 1944 * We may have dropped the spin lock to get the head mutex lock, and 1945 * that might have given someone else time to free the head. If that's 1946 * true, it has been removed from our list and we can move on. 1947 */ 1948 if (ret == -EAGAIN) 1949 head = ERR_PTR(-EAGAIN); 1950 1951 return head; 1952 } 1953 1954 static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans, 1955 struct btrfs_delayed_ref_head *locked_ref) 1956 { 1957 struct btrfs_fs_info *fs_info = trans->fs_info; 1958 struct btrfs_delayed_ref_root *delayed_refs; 1959 struct btrfs_delayed_extent_op *extent_op; 1960 struct btrfs_delayed_ref_node *ref; 1961 bool must_insert_reserved; 1962 int ret; 1963 1964 delayed_refs = &trans->transaction->delayed_refs; 1965 1966 lockdep_assert_held(&locked_ref->mutex); 1967 lockdep_assert_held(&locked_ref->lock); 1968 1969 while ((ref = select_delayed_ref(locked_ref))) { 1970 if (ref->seq && 1971 btrfs_check_delayed_seq(fs_info, ref->seq)) { 1972 spin_unlock(&locked_ref->lock); 1973 unselect_delayed_ref_head(delayed_refs, locked_ref); 1974 return -EAGAIN; 1975 } 1976 1977 rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree); 1978 RB_CLEAR_NODE(&ref->ref_node); 1979 if (!list_empty(&ref->add_list)) 1980 list_del(&ref->add_list); 1981 /* 1982 * When we play the delayed ref, also correct the ref_mod on 1983 * head 1984 */ 1985 switch (ref->action) { 1986 case BTRFS_ADD_DELAYED_REF: 1987 case BTRFS_ADD_DELAYED_EXTENT: 1988 locked_ref->ref_mod -= ref->ref_mod; 1989 break; 1990 case BTRFS_DROP_DELAYED_REF: 1991 locked_ref->ref_mod += ref->ref_mod; 1992 break; 1993 default: 1994 WARN_ON(1); 1995 } 1996 atomic_dec(&delayed_refs->num_entries); 1997 1998 /* 1999 * Record the must_insert_reserved flag before we drop the 2000 * spin lock. 2001 */ 2002 must_insert_reserved = locked_ref->must_insert_reserved; 2003 locked_ref->must_insert_reserved = false; 2004 2005 extent_op = locked_ref->extent_op; 2006 locked_ref->extent_op = NULL; 2007 spin_unlock(&locked_ref->lock); 2008 2009 ret = run_one_delayed_ref(trans, ref, extent_op, 2010 must_insert_reserved); 2011 2012 btrfs_free_delayed_extent_op(extent_op); 2013 if (ret) { 2014 unselect_delayed_ref_head(delayed_refs, locked_ref); 2015 btrfs_put_delayed_ref(ref); 2016 return ret; 2017 } 2018 2019 btrfs_put_delayed_ref(ref); 2020 cond_resched(); 2021 2022 spin_lock(&locked_ref->lock); 2023 btrfs_merge_delayed_refs(fs_info, delayed_refs, locked_ref); 2024 } 2025 2026 return 0; 2027 } 2028 2029 /* 2030 * Returns 0 on success or if called with an already aborted transaction. 2031 * Returns -ENOMEM or -EIO on failure and will abort the transaction. 2032 */ 2033 static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans, 2034 unsigned long nr) 2035 { 2036 struct btrfs_fs_info *fs_info = trans->fs_info; 2037 struct btrfs_delayed_ref_root *delayed_refs; 2038 struct btrfs_delayed_ref_head *locked_ref = NULL; 2039 int ret; 2040 unsigned long count = 0; 2041 2042 delayed_refs = &trans->transaction->delayed_refs; 2043 do { 2044 if (!locked_ref) { 2045 locked_ref = btrfs_obtain_ref_head(trans); 2046 if (IS_ERR_OR_NULL(locked_ref)) { 2047 if (PTR_ERR(locked_ref) == -EAGAIN) { 2048 continue; 2049 } else { 2050 break; 2051 } 2052 } 2053 count++; 2054 } 2055 /* 2056 * We need to try and merge add/drops of the same ref since we 2057 * can run into issues with relocate dropping the implicit ref 2058 * and then it being added back again before the drop can 2059 * finish. If we merged anything we need to re-loop so we can 2060 * get a good ref. 2061 * Or we can get node references of the same type that weren't 2062 * merged when created due to bumps in the tree mod seq, and 2063 * we need to merge them to prevent adding an inline extent 2064 * backref before dropping it (triggering a BUG_ON at 2065 * insert_inline_extent_backref()). 2066 */ 2067 spin_lock(&locked_ref->lock); 2068 btrfs_merge_delayed_refs(fs_info, delayed_refs, locked_ref); 2069 2070 ret = btrfs_run_delayed_refs_for_head(trans, locked_ref); 2071 if (ret < 0 && ret != -EAGAIN) { 2072 /* 2073 * Error, btrfs_run_delayed_refs_for_head already 2074 * unlocked everything so just bail out 2075 */ 2076 return ret; 2077 } else if (!ret) { 2078 /* 2079 * Success, perform the usual cleanup of a processed 2080 * head 2081 */ 2082 ret = cleanup_ref_head(trans, locked_ref); 2083 if (ret > 0 ) { 2084 /* We dropped our lock, we need to loop. */ 2085 ret = 0; 2086 continue; 2087 } else if (ret) { 2088 return ret; 2089 } 2090 } 2091 2092 /* 2093 * Either success case or btrfs_run_delayed_refs_for_head 2094 * returned -EAGAIN, meaning we need to select another head 2095 */ 2096 2097 locked_ref = NULL; 2098 cond_resched(); 2099 } while ((nr != -1 && count < nr) || locked_ref); 2100 2101 return 0; 2102 } 2103 2104 #ifdef SCRAMBLE_DELAYED_REFS 2105 /* 2106 * Normally delayed refs get processed in ascending bytenr order. This 2107 * correlates in most cases to the order added. To expose dependencies on this 2108 * order, we start to process the tree in the middle instead of the beginning 2109 */ 2110 static u64 find_middle(struct rb_root *root) 2111 { 2112 struct rb_node *n = root->rb_node; 2113 struct btrfs_delayed_ref_node *entry; 2114 int alt = 1; 2115 u64 middle; 2116 u64 first = 0, last = 0; 2117 2118 n = rb_first(root); 2119 if (n) { 2120 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node); 2121 first = entry->bytenr; 2122 } 2123 n = rb_last(root); 2124 if (n) { 2125 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node); 2126 last = entry->bytenr; 2127 } 2128 n = root->rb_node; 2129 2130 while (n) { 2131 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node); 2132 WARN_ON(!entry->in_tree); 2133 2134 middle = entry->bytenr; 2135 2136 if (alt) 2137 n = n->rb_left; 2138 else 2139 n = n->rb_right; 2140 2141 alt = 1 - alt; 2142 } 2143 return middle; 2144 } 2145 #endif 2146 2147 /* 2148 * this starts processing the delayed reference count updates and 2149 * extent insertions we have queued up so far. count can be 2150 * 0, which means to process everything in the tree at the start 2151 * of the run (but not newly added entries), or it can be some target 2152 * number you'd like to process. 2153 * 2154 * Returns 0 on success or if called with an aborted transaction 2155 * Returns <0 on error and aborts the transaction 2156 */ 2157 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans, 2158 unsigned long count) 2159 { 2160 struct btrfs_fs_info *fs_info = trans->fs_info; 2161 struct rb_node *node; 2162 struct btrfs_delayed_ref_root *delayed_refs; 2163 struct btrfs_delayed_ref_head *head; 2164 int ret; 2165 int run_all = count == (unsigned long)-1; 2166 2167 /* We'll clean this up in btrfs_cleanup_transaction */ 2168 if (TRANS_ABORTED(trans)) 2169 return 0; 2170 2171 if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags)) 2172 return 0; 2173 2174 delayed_refs = &trans->transaction->delayed_refs; 2175 if (count == 0) 2176 count = delayed_refs->num_heads_ready; 2177 2178 again: 2179 #ifdef SCRAMBLE_DELAYED_REFS 2180 delayed_refs->run_delayed_start = find_middle(&delayed_refs->root); 2181 #endif 2182 ret = __btrfs_run_delayed_refs(trans, count); 2183 if (ret < 0) { 2184 btrfs_abort_transaction(trans, ret); 2185 return ret; 2186 } 2187 2188 if (run_all) { 2189 btrfs_create_pending_block_groups(trans); 2190 2191 spin_lock(&delayed_refs->lock); 2192 node = rb_first_cached(&delayed_refs->href_root); 2193 if (!node) { 2194 spin_unlock(&delayed_refs->lock); 2195 goto out; 2196 } 2197 head = rb_entry(node, struct btrfs_delayed_ref_head, 2198 href_node); 2199 refcount_inc(&head->refs); 2200 spin_unlock(&delayed_refs->lock); 2201 2202 /* Mutex was contended, block until it's released and retry. */ 2203 mutex_lock(&head->mutex); 2204 mutex_unlock(&head->mutex); 2205 2206 btrfs_put_delayed_ref_head(head); 2207 cond_resched(); 2208 goto again; 2209 } 2210 out: 2211 return 0; 2212 } 2213 2214 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans, 2215 struct extent_buffer *eb, u64 flags) 2216 { 2217 struct btrfs_delayed_extent_op *extent_op; 2218 int level = btrfs_header_level(eb); 2219 int ret; 2220 2221 extent_op = btrfs_alloc_delayed_extent_op(); 2222 if (!extent_op) 2223 return -ENOMEM; 2224 2225 extent_op->flags_to_set = flags; 2226 extent_op->update_flags = true; 2227 extent_op->update_key = false; 2228 extent_op->level = level; 2229 2230 ret = btrfs_add_delayed_extent_op(trans, eb->start, eb->len, extent_op); 2231 if (ret) 2232 btrfs_free_delayed_extent_op(extent_op); 2233 return ret; 2234 } 2235 2236 static noinline int check_delayed_ref(struct btrfs_root *root, 2237 struct btrfs_path *path, 2238 u64 objectid, u64 offset, u64 bytenr) 2239 { 2240 struct btrfs_delayed_ref_head *head; 2241 struct btrfs_delayed_ref_node *ref; 2242 struct btrfs_delayed_data_ref *data_ref; 2243 struct btrfs_delayed_ref_root *delayed_refs; 2244 struct btrfs_transaction *cur_trans; 2245 struct rb_node *node; 2246 int ret = 0; 2247 2248 spin_lock(&root->fs_info->trans_lock); 2249 cur_trans = root->fs_info->running_transaction; 2250 if (cur_trans) 2251 refcount_inc(&cur_trans->use_count); 2252 spin_unlock(&root->fs_info->trans_lock); 2253 if (!cur_trans) 2254 return 0; 2255 2256 delayed_refs = &cur_trans->delayed_refs; 2257 spin_lock(&delayed_refs->lock); 2258 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr); 2259 if (!head) { 2260 spin_unlock(&delayed_refs->lock); 2261 btrfs_put_transaction(cur_trans); 2262 return 0; 2263 } 2264 2265 if (!mutex_trylock(&head->mutex)) { 2266 if (path->nowait) { 2267 spin_unlock(&delayed_refs->lock); 2268 btrfs_put_transaction(cur_trans); 2269 return -EAGAIN; 2270 } 2271 2272 refcount_inc(&head->refs); 2273 spin_unlock(&delayed_refs->lock); 2274 2275 btrfs_release_path(path); 2276 2277 /* 2278 * Mutex was contended, block until it's released and let 2279 * caller try again 2280 */ 2281 mutex_lock(&head->mutex); 2282 mutex_unlock(&head->mutex); 2283 btrfs_put_delayed_ref_head(head); 2284 btrfs_put_transaction(cur_trans); 2285 return -EAGAIN; 2286 } 2287 spin_unlock(&delayed_refs->lock); 2288 2289 spin_lock(&head->lock); 2290 /* 2291 * XXX: We should replace this with a proper search function in the 2292 * future. 2293 */ 2294 for (node = rb_first_cached(&head->ref_tree); node; 2295 node = rb_next(node)) { 2296 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node); 2297 /* If it's a shared ref we know a cross reference exists */ 2298 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) { 2299 ret = 1; 2300 break; 2301 } 2302 2303 data_ref = btrfs_delayed_node_to_data_ref(ref); 2304 2305 /* 2306 * If our ref doesn't match the one we're currently looking at 2307 * then we have a cross reference. 2308 */ 2309 if (data_ref->root != root->root_key.objectid || 2310 data_ref->objectid != objectid || 2311 data_ref->offset != offset) { 2312 ret = 1; 2313 break; 2314 } 2315 } 2316 spin_unlock(&head->lock); 2317 mutex_unlock(&head->mutex); 2318 btrfs_put_transaction(cur_trans); 2319 return ret; 2320 } 2321 2322 static noinline int check_committed_ref(struct btrfs_root *root, 2323 struct btrfs_path *path, 2324 u64 objectid, u64 offset, u64 bytenr, 2325 bool strict) 2326 { 2327 struct btrfs_fs_info *fs_info = root->fs_info; 2328 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, bytenr); 2329 struct extent_buffer *leaf; 2330 struct btrfs_extent_data_ref *ref; 2331 struct btrfs_extent_inline_ref *iref; 2332 struct btrfs_extent_item *ei; 2333 struct btrfs_key key; 2334 u32 item_size; 2335 int type; 2336 int ret; 2337 2338 key.objectid = bytenr; 2339 key.offset = (u64)-1; 2340 key.type = BTRFS_EXTENT_ITEM_KEY; 2341 2342 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0); 2343 if (ret < 0) 2344 goto out; 2345 BUG_ON(ret == 0); /* Corruption */ 2346 2347 ret = -ENOENT; 2348 if (path->slots[0] == 0) 2349 goto out; 2350 2351 path->slots[0]--; 2352 leaf = path->nodes[0]; 2353 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]); 2354 2355 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY) 2356 goto out; 2357 2358 ret = 1; 2359 item_size = btrfs_item_size(leaf, path->slots[0]); 2360 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item); 2361 2362 /* If extent item has more than 1 inline ref then it's shared */ 2363 if (item_size != sizeof(*ei) + 2364 btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY)) 2365 goto out; 2366 2367 /* 2368 * If extent created before last snapshot => it's shared unless the 2369 * snapshot has been deleted. Use the heuristic if strict is false. 2370 */ 2371 if (!strict && 2372 (btrfs_extent_generation(leaf, ei) <= 2373 btrfs_root_last_snapshot(&root->root_item))) 2374 goto out; 2375 2376 iref = (struct btrfs_extent_inline_ref *)(ei + 1); 2377 2378 /* If this extent has SHARED_DATA_REF then it's shared */ 2379 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA); 2380 if (type != BTRFS_EXTENT_DATA_REF_KEY) 2381 goto out; 2382 2383 ref = (struct btrfs_extent_data_ref *)(&iref->offset); 2384 if (btrfs_extent_refs(leaf, ei) != 2385 btrfs_extent_data_ref_count(leaf, ref) || 2386 btrfs_extent_data_ref_root(leaf, ref) != 2387 root->root_key.objectid || 2388 btrfs_extent_data_ref_objectid(leaf, ref) != objectid || 2389 btrfs_extent_data_ref_offset(leaf, ref) != offset) 2390 goto out; 2391 2392 ret = 0; 2393 out: 2394 return ret; 2395 } 2396 2397 int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset, 2398 u64 bytenr, bool strict, struct btrfs_path *path) 2399 { 2400 int ret; 2401 2402 do { 2403 ret = check_committed_ref(root, path, objectid, 2404 offset, bytenr, strict); 2405 if (ret && ret != -ENOENT) 2406 goto out; 2407 2408 ret = check_delayed_ref(root, path, objectid, offset, bytenr); 2409 } while (ret == -EAGAIN && !path->nowait); 2410 2411 out: 2412 btrfs_release_path(path); 2413 if (btrfs_is_data_reloc_root(root)) 2414 WARN_ON(ret > 0); 2415 return ret; 2416 } 2417 2418 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans, 2419 struct btrfs_root *root, 2420 struct extent_buffer *buf, 2421 int full_backref, int inc) 2422 { 2423 struct btrfs_fs_info *fs_info = root->fs_info; 2424 u64 bytenr; 2425 u64 num_bytes; 2426 u64 parent; 2427 u64 ref_root; 2428 u32 nritems; 2429 struct btrfs_key key; 2430 struct btrfs_file_extent_item *fi; 2431 struct btrfs_ref generic_ref = { 0 }; 2432 bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC); 2433 int i; 2434 int action; 2435 int level; 2436 int ret = 0; 2437 2438 if (btrfs_is_testing(fs_info)) 2439 return 0; 2440 2441 ref_root = btrfs_header_owner(buf); 2442 nritems = btrfs_header_nritems(buf); 2443 level = btrfs_header_level(buf); 2444 2445 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && level == 0) 2446 return 0; 2447 2448 if (full_backref) 2449 parent = buf->start; 2450 else 2451 parent = 0; 2452 if (inc) 2453 action = BTRFS_ADD_DELAYED_REF; 2454 else 2455 action = BTRFS_DROP_DELAYED_REF; 2456 2457 for (i = 0; i < nritems; i++) { 2458 if (level == 0) { 2459 btrfs_item_key_to_cpu(buf, &key, i); 2460 if (key.type != BTRFS_EXTENT_DATA_KEY) 2461 continue; 2462 fi = btrfs_item_ptr(buf, i, 2463 struct btrfs_file_extent_item); 2464 if (btrfs_file_extent_type(buf, fi) == 2465 BTRFS_FILE_EXTENT_INLINE) 2466 continue; 2467 bytenr = btrfs_file_extent_disk_bytenr(buf, fi); 2468 if (bytenr == 0) 2469 continue; 2470 2471 num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi); 2472 key.offset -= btrfs_file_extent_offset(buf, fi); 2473 btrfs_init_generic_ref(&generic_ref, action, bytenr, 2474 num_bytes, parent); 2475 btrfs_init_data_ref(&generic_ref, ref_root, key.objectid, 2476 key.offset, root->root_key.objectid, 2477 for_reloc); 2478 if (inc) 2479 ret = btrfs_inc_extent_ref(trans, &generic_ref); 2480 else 2481 ret = btrfs_free_extent(trans, &generic_ref); 2482 if (ret) 2483 goto fail; 2484 } else { 2485 bytenr = btrfs_node_blockptr(buf, i); 2486 num_bytes = fs_info->nodesize; 2487 btrfs_init_generic_ref(&generic_ref, action, bytenr, 2488 num_bytes, parent); 2489 btrfs_init_tree_ref(&generic_ref, level - 1, ref_root, 2490 root->root_key.objectid, for_reloc); 2491 if (inc) 2492 ret = btrfs_inc_extent_ref(trans, &generic_ref); 2493 else 2494 ret = btrfs_free_extent(trans, &generic_ref); 2495 if (ret) 2496 goto fail; 2497 } 2498 } 2499 return 0; 2500 fail: 2501 return ret; 2502 } 2503 2504 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, 2505 struct extent_buffer *buf, int full_backref) 2506 { 2507 return __btrfs_mod_ref(trans, root, buf, full_backref, 1); 2508 } 2509 2510 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root, 2511 struct extent_buffer *buf, int full_backref) 2512 { 2513 return __btrfs_mod_ref(trans, root, buf, full_backref, 0); 2514 } 2515 2516 static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data) 2517 { 2518 struct btrfs_fs_info *fs_info = root->fs_info; 2519 u64 flags; 2520 u64 ret; 2521 2522 if (data) 2523 flags = BTRFS_BLOCK_GROUP_DATA; 2524 else if (root == fs_info->chunk_root) 2525 flags = BTRFS_BLOCK_GROUP_SYSTEM; 2526 else 2527 flags = BTRFS_BLOCK_GROUP_METADATA; 2528 2529 ret = btrfs_get_alloc_profile(fs_info, flags); 2530 return ret; 2531 } 2532 2533 static u64 first_logical_byte(struct btrfs_fs_info *fs_info) 2534 { 2535 struct rb_node *leftmost; 2536 u64 bytenr = 0; 2537 2538 read_lock(&fs_info->block_group_cache_lock); 2539 /* Get the block group with the lowest logical start address. */ 2540 leftmost = rb_first_cached(&fs_info->block_group_cache_tree); 2541 if (leftmost) { 2542 struct btrfs_block_group *bg; 2543 2544 bg = rb_entry(leftmost, struct btrfs_block_group, cache_node); 2545 bytenr = bg->start; 2546 } 2547 read_unlock(&fs_info->block_group_cache_lock); 2548 2549 return bytenr; 2550 } 2551 2552 static int pin_down_extent(struct btrfs_trans_handle *trans, 2553 struct btrfs_block_group *cache, 2554 u64 bytenr, u64 num_bytes, int reserved) 2555 { 2556 struct btrfs_fs_info *fs_info = cache->fs_info; 2557 2558 spin_lock(&cache->space_info->lock); 2559 spin_lock(&cache->lock); 2560 cache->pinned += num_bytes; 2561 btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info, 2562 num_bytes); 2563 if (reserved) { 2564 cache->reserved -= num_bytes; 2565 cache->space_info->bytes_reserved -= num_bytes; 2566 } 2567 spin_unlock(&cache->lock); 2568 spin_unlock(&cache->space_info->lock); 2569 2570 set_extent_bit(&trans->transaction->pinned_extents, bytenr, 2571 bytenr + num_bytes - 1, EXTENT_DIRTY, NULL); 2572 return 0; 2573 } 2574 2575 int btrfs_pin_extent(struct btrfs_trans_handle *trans, 2576 u64 bytenr, u64 num_bytes, int reserved) 2577 { 2578 struct btrfs_block_group *cache; 2579 2580 cache = btrfs_lookup_block_group(trans->fs_info, bytenr); 2581 BUG_ON(!cache); /* Logic error */ 2582 2583 pin_down_extent(trans, cache, bytenr, num_bytes, reserved); 2584 2585 btrfs_put_block_group(cache); 2586 return 0; 2587 } 2588 2589 /* 2590 * this function must be called within transaction 2591 */ 2592 int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans, 2593 u64 bytenr, u64 num_bytes) 2594 { 2595 struct btrfs_block_group *cache; 2596 int ret; 2597 2598 cache = btrfs_lookup_block_group(trans->fs_info, bytenr); 2599 if (!cache) 2600 return -EINVAL; 2601 2602 /* 2603 * Fully cache the free space first so that our pin removes the free space 2604 * from the cache. 2605 */ 2606 ret = btrfs_cache_block_group(cache, true); 2607 if (ret) 2608 goto out; 2609 2610 pin_down_extent(trans, cache, bytenr, num_bytes, 0); 2611 2612 /* remove us from the free space cache (if we're there at all) */ 2613 ret = btrfs_remove_free_space(cache, bytenr, num_bytes); 2614 out: 2615 btrfs_put_block_group(cache); 2616 return ret; 2617 } 2618 2619 static int __exclude_logged_extent(struct btrfs_fs_info *fs_info, 2620 u64 start, u64 num_bytes) 2621 { 2622 int ret; 2623 struct btrfs_block_group *block_group; 2624 2625 block_group = btrfs_lookup_block_group(fs_info, start); 2626 if (!block_group) 2627 return -EINVAL; 2628 2629 ret = btrfs_cache_block_group(block_group, true); 2630 if (ret) 2631 goto out; 2632 2633 ret = btrfs_remove_free_space(block_group, start, num_bytes); 2634 out: 2635 btrfs_put_block_group(block_group); 2636 return ret; 2637 } 2638 2639 int btrfs_exclude_logged_extents(struct extent_buffer *eb) 2640 { 2641 struct btrfs_fs_info *fs_info = eb->fs_info; 2642 struct btrfs_file_extent_item *item; 2643 struct btrfs_key key; 2644 int found_type; 2645 int i; 2646 int ret = 0; 2647 2648 if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) 2649 return 0; 2650 2651 for (i = 0; i < btrfs_header_nritems(eb); i++) { 2652 btrfs_item_key_to_cpu(eb, &key, i); 2653 if (key.type != BTRFS_EXTENT_DATA_KEY) 2654 continue; 2655 item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item); 2656 found_type = btrfs_file_extent_type(eb, item); 2657 if (found_type == BTRFS_FILE_EXTENT_INLINE) 2658 continue; 2659 if (btrfs_file_extent_disk_bytenr(eb, item) == 0) 2660 continue; 2661 key.objectid = btrfs_file_extent_disk_bytenr(eb, item); 2662 key.offset = btrfs_file_extent_disk_num_bytes(eb, item); 2663 ret = __exclude_logged_extent(fs_info, key.objectid, key.offset); 2664 if (ret) 2665 break; 2666 } 2667 2668 return ret; 2669 } 2670 2671 static void 2672 btrfs_inc_block_group_reservations(struct btrfs_block_group *bg) 2673 { 2674 atomic_inc(&bg->reservations); 2675 } 2676 2677 /* 2678 * Returns the free cluster for the given space info and sets empty_cluster to 2679 * what it should be based on the mount options. 2680 */ 2681 static struct btrfs_free_cluster * 2682 fetch_cluster_info(struct btrfs_fs_info *fs_info, 2683 struct btrfs_space_info *space_info, u64 *empty_cluster) 2684 { 2685 struct btrfs_free_cluster *ret = NULL; 2686 2687 *empty_cluster = 0; 2688 if (btrfs_mixed_space_info(space_info)) 2689 return ret; 2690 2691 if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) { 2692 ret = &fs_info->meta_alloc_cluster; 2693 if (btrfs_test_opt(fs_info, SSD)) 2694 *empty_cluster = SZ_2M; 2695 else 2696 *empty_cluster = SZ_64K; 2697 } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) && 2698 btrfs_test_opt(fs_info, SSD_SPREAD)) { 2699 *empty_cluster = SZ_2M; 2700 ret = &fs_info->data_alloc_cluster; 2701 } 2702 2703 return ret; 2704 } 2705 2706 static int unpin_extent_range(struct btrfs_fs_info *fs_info, 2707 u64 start, u64 end, 2708 const bool return_free_space) 2709 { 2710 struct btrfs_block_group *cache = NULL; 2711 struct btrfs_space_info *space_info; 2712 struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv; 2713 struct btrfs_free_cluster *cluster = NULL; 2714 u64 len; 2715 u64 total_unpinned = 0; 2716 u64 empty_cluster = 0; 2717 bool readonly; 2718 2719 while (start <= end) { 2720 readonly = false; 2721 if (!cache || 2722 start >= cache->start + cache->length) { 2723 if (cache) 2724 btrfs_put_block_group(cache); 2725 total_unpinned = 0; 2726 cache = btrfs_lookup_block_group(fs_info, start); 2727 BUG_ON(!cache); /* Logic error */ 2728 2729 cluster = fetch_cluster_info(fs_info, 2730 cache->space_info, 2731 &empty_cluster); 2732 empty_cluster <<= 1; 2733 } 2734 2735 len = cache->start + cache->length - start; 2736 len = min(len, end + 1 - start); 2737 2738 if (return_free_space) 2739 btrfs_add_free_space(cache, start, len); 2740 2741 start += len; 2742 total_unpinned += len; 2743 space_info = cache->space_info; 2744 2745 /* 2746 * If this space cluster has been marked as fragmented and we've 2747 * unpinned enough in this block group to potentially allow a 2748 * cluster to be created inside of it go ahead and clear the 2749 * fragmented check. 2750 */ 2751 if (cluster && cluster->fragmented && 2752 total_unpinned > empty_cluster) { 2753 spin_lock(&cluster->lock); 2754 cluster->fragmented = 0; 2755 spin_unlock(&cluster->lock); 2756 } 2757 2758 spin_lock(&space_info->lock); 2759 spin_lock(&cache->lock); 2760 cache->pinned -= len; 2761 btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len); 2762 space_info->max_extent_size = 0; 2763 if (cache->ro) { 2764 space_info->bytes_readonly += len; 2765 readonly = true; 2766 } else if (btrfs_is_zoned(fs_info)) { 2767 /* Need reset before reusing in a zoned block group */ 2768 btrfs_space_info_update_bytes_zone_unusable(fs_info, space_info, 2769 len); 2770 readonly = true; 2771 } 2772 spin_unlock(&cache->lock); 2773 if (!readonly && return_free_space && 2774 global_rsv->space_info == space_info) { 2775 spin_lock(&global_rsv->lock); 2776 if (!global_rsv->full) { 2777 u64 to_add = min(len, global_rsv->size - 2778 global_rsv->reserved); 2779 2780 global_rsv->reserved += to_add; 2781 btrfs_space_info_update_bytes_may_use(fs_info, 2782 space_info, to_add); 2783 if (global_rsv->reserved >= global_rsv->size) 2784 global_rsv->full = 1; 2785 len -= to_add; 2786 } 2787 spin_unlock(&global_rsv->lock); 2788 } 2789 /* Add to any tickets we may have */ 2790 if (!readonly && return_free_space && len) 2791 btrfs_try_granting_tickets(fs_info, space_info); 2792 spin_unlock(&space_info->lock); 2793 } 2794 2795 if (cache) 2796 btrfs_put_block_group(cache); 2797 return 0; 2798 } 2799 2800 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans) 2801 { 2802 struct btrfs_fs_info *fs_info = trans->fs_info; 2803 struct btrfs_block_group *block_group, *tmp; 2804 struct list_head *deleted_bgs; 2805 struct extent_io_tree *unpin; 2806 u64 start; 2807 u64 end; 2808 int ret; 2809 2810 unpin = &trans->transaction->pinned_extents; 2811 2812 while (!TRANS_ABORTED(trans)) { 2813 struct extent_state *cached_state = NULL; 2814 2815 mutex_lock(&fs_info->unused_bg_unpin_mutex); 2816 if (!find_first_extent_bit(unpin, 0, &start, &end, 2817 EXTENT_DIRTY, &cached_state)) { 2818 mutex_unlock(&fs_info->unused_bg_unpin_mutex); 2819 break; 2820 } 2821 2822 if (btrfs_test_opt(fs_info, DISCARD_SYNC)) 2823 ret = btrfs_discard_extent(fs_info, start, 2824 end + 1 - start, NULL); 2825 2826 clear_extent_dirty(unpin, start, end, &cached_state); 2827 unpin_extent_range(fs_info, start, end, true); 2828 mutex_unlock(&fs_info->unused_bg_unpin_mutex); 2829 free_extent_state(cached_state); 2830 cond_resched(); 2831 } 2832 2833 if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) { 2834 btrfs_discard_calc_delay(&fs_info->discard_ctl); 2835 btrfs_discard_schedule_work(&fs_info->discard_ctl, true); 2836 } 2837 2838 /* 2839 * Transaction is finished. We don't need the lock anymore. We 2840 * do need to clean up the block groups in case of a transaction 2841 * abort. 2842 */ 2843 deleted_bgs = &trans->transaction->deleted_bgs; 2844 list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) { 2845 u64 trimmed = 0; 2846 2847 ret = -EROFS; 2848 if (!TRANS_ABORTED(trans)) 2849 ret = btrfs_discard_extent(fs_info, 2850 block_group->start, 2851 block_group->length, 2852 &trimmed); 2853 2854 list_del_init(&block_group->bg_list); 2855 btrfs_unfreeze_block_group(block_group); 2856 btrfs_put_block_group(block_group); 2857 2858 if (ret) { 2859 const char *errstr = btrfs_decode_error(ret); 2860 btrfs_warn(fs_info, 2861 "discard failed while removing blockgroup: errno=%d %s", 2862 ret, errstr); 2863 } 2864 } 2865 2866 return 0; 2867 } 2868 2869 static int do_free_extent_accounting(struct btrfs_trans_handle *trans, 2870 u64 bytenr, u64 num_bytes, bool is_data) 2871 { 2872 int ret; 2873 2874 if (is_data) { 2875 struct btrfs_root *csum_root; 2876 2877 csum_root = btrfs_csum_root(trans->fs_info, bytenr); 2878 ret = btrfs_del_csums(trans, csum_root, bytenr, num_bytes); 2879 if (ret) { 2880 btrfs_abort_transaction(trans, ret); 2881 return ret; 2882 } 2883 } 2884 2885 ret = add_to_free_space_tree(trans, bytenr, num_bytes); 2886 if (ret) { 2887 btrfs_abort_transaction(trans, ret); 2888 return ret; 2889 } 2890 2891 ret = btrfs_update_block_group(trans, bytenr, num_bytes, false); 2892 if (ret) 2893 btrfs_abort_transaction(trans, ret); 2894 2895 return ret; 2896 } 2897 2898 #define abort_and_dump(trans, path, fmt, args...) \ 2899 ({ \ 2900 btrfs_abort_transaction(trans, -EUCLEAN); \ 2901 btrfs_print_leaf(path->nodes[0]); \ 2902 btrfs_crit(trans->fs_info, fmt, ##args); \ 2903 }) 2904 2905 /* 2906 * Drop one or more refs of @node. 2907 * 2908 * 1. Locate the extent refs. 2909 * It's either inline in EXTENT/METADATA_ITEM or in keyed SHARED_* item. 2910 * Locate it, then reduce the refs number or remove the ref line completely. 2911 * 2912 * 2. Update the refs count in EXTENT/METADATA_ITEM 2913 * 2914 * Inline backref case: 2915 * 2916 * in extent tree we have: 2917 * 2918 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82 2919 * refs 2 gen 6 flags DATA 2920 * extent data backref root FS_TREE objectid 258 offset 0 count 1 2921 * extent data backref root FS_TREE objectid 257 offset 0 count 1 2922 * 2923 * This function gets called with: 2924 * 2925 * node->bytenr = 13631488 2926 * node->num_bytes = 1048576 2927 * root_objectid = FS_TREE 2928 * owner_objectid = 257 2929 * owner_offset = 0 2930 * refs_to_drop = 1 2931 * 2932 * Then we should get some like: 2933 * 2934 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82 2935 * refs 1 gen 6 flags DATA 2936 * extent data backref root FS_TREE objectid 258 offset 0 count 1 2937 * 2938 * Keyed backref case: 2939 * 2940 * in extent tree we have: 2941 * 2942 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24 2943 * refs 754 gen 6 flags DATA 2944 * [...] 2945 * item 2 key (13631488 EXTENT_DATA_REF <HASH>) itemoff 3915 itemsize 28 2946 * extent data backref root FS_TREE objectid 866 offset 0 count 1 2947 * 2948 * This function get called with: 2949 * 2950 * node->bytenr = 13631488 2951 * node->num_bytes = 1048576 2952 * root_objectid = FS_TREE 2953 * owner_objectid = 866 2954 * owner_offset = 0 2955 * refs_to_drop = 1 2956 * 2957 * Then we should get some like: 2958 * 2959 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24 2960 * refs 753 gen 6 flags DATA 2961 * 2962 * And that (13631488 EXTENT_DATA_REF <HASH>) gets removed. 2963 */ 2964 static int __btrfs_free_extent(struct btrfs_trans_handle *trans, 2965 struct btrfs_delayed_ref_node *node, u64 parent, 2966 u64 root_objectid, u64 owner_objectid, 2967 u64 owner_offset, int refs_to_drop, 2968 struct btrfs_delayed_extent_op *extent_op) 2969 { 2970 struct btrfs_fs_info *info = trans->fs_info; 2971 struct btrfs_key key; 2972 struct btrfs_path *path; 2973 struct btrfs_root *extent_root; 2974 struct extent_buffer *leaf; 2975 struct btrfs_extent_item *ei; 2976 struct btrfs_extent_inline_ref *iref; 2977 int ret; 2978 int is_data; 2979 int extent_slot = 0; 2980 int found_extent = 0; 2981 int num_to_del = 1; 2982 u32 item_size; 2983 u64 refs; 2984 u64 bytenr = node->bytenr; 2985 u64 num_bytes = node->num_bytes; 2986 bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA); 2987 2988 extent_root = btrfs_extent_root(info, bytenr); 2989 ASSERT(extent_root); 2990 2991 path = btrfs_alloc_path(); 2992 if (!path) 2993 return -ENOMEM; 2994 2995 is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID; 2996 2997 if (!is_data && refs_to_drop != 1) { 2998 btrfs_crit(info, 2999 "invalid refs_to_drop, dropping more than 1 refs for tree block %llu refs_to_drop %u", 3000 node->bytenr, refs_to_drop); 3001 ret = -EINVAL; 3002 btrfs_abort_transaction(trans, ret); 3003 goto out; 3004 } 3005 3006 if (is_data) 3007 skinny_metadata = false; 3008 3009 ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes, 3010 parent, root_objectid, owner_objectid, 3011 owner_offset); 3012 if (ret == 0) { 3013 /* 3014 * Either the inline backref or the SHARED_DATA_REF/ 3015 * SHARED_BLOCK_REF is found 3016 * 3017 * Here is a quick path to locate EXTENT/METADATA_ITEM. 3018 * It's possible the EXTENT/METADATA_ITEM is near current slot. 3019 */ 3020 extent_slot = path->slots[0]; 3021 while (extent_slot >= 0) { 3022 btrfs_item_key_to_cpu(path->nodes[0], &key, 3023 extent_slot); 3024 if (key.objectid != bytenr) 3025 break; 3026 if (key.type == BTRFS_EXTENT_ITEM_KEY && 3027 key.offset == num_bytes) { 3028 found_extent = 1; 3029 break; 3030 } 3031 if (key.type == BTRFS_METADATA_ITEM_KEY && 3032 key.offset == owner_objectid) { 3033 found_extent = 1; 3034 break; 3035 } 3036 3037 /* Quick path didn't find the EXTEMT/METADATA_ITEM */ 3038 if (path->slots[0] - extent_slot > 5) 3039 break; 3040 extent_slot--; 3041 } 3042 3043 if (!found_extent) { 3044 if (iref) { 3045 abort_and_dump(trans, path, 3046 "invalid iref slot %u, no EXTENT/METADATA_ITEM found but has inline extent ref", 3047 path->slots[0]); 3048 ret = -EUCLEAN; 3049 goto out; 3050 } 3051 /* Must be SHARED_* item, remove the backref first */ 3052 ret = remove_extent_backref(trans, extent_root, path, 3053 NULL, refs_to_drop, is_data); 3054 if (ret) { 3055 btrfs_abort_transaction(trans, ret); 3056 goto out; 3057 } 3058 btrfs_release_path(path); 3059 3060 /* Slow path to locate EXTENT/METADATA_ITEM */ 3061 key.objectid = bytenr; 3062 key.type = BTRFS_EXTENT_ITEM_KEY; 3063 key.offset = num_bytes; 3064 3065 if (!is_data && skinny_metadata) { 3066 key.type = BTRFS_METADATA_ITEM_KEY; 3067 key.offset = owner_objectid; 3068 } 3069 3070 ret = btrfs_search_slot(trans, extent_root, 3071 &key, path, -1, 1); 3072 if (ret > 0 && skinny_metadata && path->slots[0]) { 3073 /* 3074 * Couldn't find our skinny metadata item, 3075 * see if we have ye olde extent item. 3076 */ 3077 path->slots[0]--; 3078 btrfs_item_key_to_cpu(path->nodes[0], &key, 3079 path->slots[0]); 3080 if (key.objectid == bytenr && 3081 key.type == BTRFS_EXTENT_ITEM_KEY && 3082 key.offset == num_bytes) 3083 ret = 0; 3084 } 3085 3086 if (ret > 0 && skinny_metadata) { 3087 skinny_metadata = false; 3088 key.objectid = bytenr; 3089 key.type = BTRFS_EXTENT_ITEM_KEY; 3090 key.offset = num_bytes; 3091 btrfs_release_path(path); 3092 ret = btrfs_search_slot(trans, extent_root, 3093 &key, path, -1, 1); 3094 } 3095 3096 if (ret) { 3097 if (ret > 0) 3098 btrfs_print_leaf(path->nodes[0]); 3099 btrfs_err(info, 3100 "umm, got %d back from search, was looking for %llu, slot %d", 3101 ret, bytenr, path->slots[0]); 3102 } 3103 if (ret < 0) { 3104 btrfs_abort_transaction(trans, ret); 3105 goto out; 3106 } 3107 extent_slot = path->slots[0]; 3108 } 3109 } else if (WARN_ON(ret == -ENOENT)) { 3110 abort_and_dump(trans, path, 3111 "unable to find ref byte nr %llu parent %llu root %llu owner %llu offset %llu slot %d", 3112 bytenr, parent, root_objectid, owner_objectid, 3113 owner_offset, path->slots[0]); 3114 goto out; 3115 } else { 3116 btrfs_abort_transaction(trans, ret); 3117 goto out; 3118 } 3119 3120 leaf = path->nodes[0]; 3121 item_size = btrfs_item_size(leaf, extent_slot); 3122 if (unlikely(item_size < sizeof(*ei))) { 3123 ret = -EUCLEAN; 3124 btrfs_err(trans->fs_info, 3125 "unexpected extent item size, has %u expect >= %zu", 3126 item_size, sizeof(*ei)); 3127 btrfs_abort_transaction(trans, ret); 3128 goto out; 3129 } 3130 ei = btrfs_item_ptr(leaf, extent_slot, 3131 struct btrfs_extent_item); 3132 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID && 3133 key.type == BTRFS_EXTENT_ITEM_KEY) { 3134 struct btrfs_tree_block_info *bi; 3135 3136 if (item_size < sizeof(*ei) + sizeof(*bi)) { 3137 abort_and_dump(trans, path, 3138 "invalid extent item size for key (%llu, %u, %llu) slot %u owner %llu, has %u expect >= %zu", 3139 key.objectid, key.type, key.offset, 3140 path->slots[0], owner_objectid, item_size, 3141 sizeof(*ei) + sizeof(*bi)); 3142 ret = -EUCLEAN; 3143 goto out; 3144 } 3145 bi = (struct btrfs_tree_block_info *)(ei + 1); 3146 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi)); 3147 } 3148 3149 refs = btrfs_extent_refs(leaf, ei); 3150 if (refs < refs_to_drop) { 3151 abort_and_dump(trans, path, 3152 "trying to drop %d refs but we only have %llu for bytenr %llu slot %u", 3153 refs_to_drop, refs, bytenr, path->slots[0]); 3154 ret = -EUCLEAN; 3155 goto out; 3156 } 3157 refs -= refs_to_drop; 3158 3159 if (refs > 0) { 3160 if (extent_op) 3161 __run_delayed_extent_op(extent_op, leaf, ei); 3162 /* 3163 * In the case of inline back ref, reference count will 3164 * be updated by remove_extent_backref 3165 */ 3166 if (iref) { 3167 if (!found_extent) { 3168 abort_and_dump(trans, path, 3169 "invalid iref, got inlined extent ref but no EXTENT/METADATA_ITEM found, slot %u", 3170 path->slots[0]); 3171 ret = -EUCLEAN; 3172 goto out; 3173 } 3174 } else { 3175 btrfs_set_extent_refs(leaf, ei, refs); 3176 btrfs_mark_buffer_dirty(trans, leaf); 3177 } 3178 if (found_extent) { 3179 ret = remove_extent_backref(trans, extent_root, path, 3180 iref, refs_to_drop, is_data); 3181 if (ret) { 3182 btrfs_abort_transaction(trans, ret); 3183 goto out; 3184 } 3185 } 3186 } else { 3187 /* In this branch refs == 1 */ 3188 if (found_extent) { 3189 if (is_data && refs_to_drop != 3190 extent_data_ref_count(path, iref)) { 3191 abort_and_dump(trans, path, 3192 "invalid refs_to_drop, current refs %u refs_to_drop %u slot %u", 3193 extent_data_ref_count(path, iref), 3194 refs_to_drop, path->slots[0]); 3195 ret = -EUCLEAN; 3196 goto out; 3197 } 3198 if (iref) { 3199 if (path->slots[0] != extent_slot) { 3200 abort_and_dump(trans, path, 3201 "invalid iref, extent item key (%llu %u %llu) slot %u doesn't have wanted iref", 3202 key.objectid, key.type, 3203 key.offset, path->slots[0]); 3204 ret = -EUCLEAN; 3205 goto out; 3206 } 3207 } else { 3208 /* 3209 * No inline ref, we must be at SHARED_* item, 3210 * And it's single ref, it must be: 3211 * | extent_slot ||extent_slot + 1| 3212 * [ EXTENT/METADATA_ITEM ][ SHARED_* ITEM ] 3213 */ 3214 if (path->slots[0] != extent_slot + 1) { 3215 abort_and_dump(trans, path, 3216 "invalid SHARED_* item slot %u, previous item is not EXTENT/METADATA_ITEM", 3217 path->slots[0]); 3218 ret = -EUCLEAN; 3219 goto out; 3220 } 3221 path->slots[0] = extent_slot; 3222 num_to_del = 2; 3223 } 3224 } 3225 3226 ret = btrfs_del_items(trans, extent_root, path, path->slots[0], 3227 num_to_del); 3228 if (ret) { 3229 btrfs_abort_transaction(trans, ret); 3230 goto out; 3231 } 3232 btrfs_release_path(path); 3233 3234 ret = do_free_extent_accounting(trans, bytenr, num_bytes, is_data); 3235 } 3236 btrfs_release_path(path); 3237 3238 out: 3239 btrfs_free_path(path); 3240 return ret; 3241 } 3242 3243 /* 3244 * when we free an block, it is possible (and likely) that we free the last 3245 * delayed ref for that extent as well. This searches the delayed ref tree for 3246 * a given extent, and if there are no other delayed refs to be processed, it 3247 * removes it from the tree. 3248 */ 3249 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans, 3250 u64 bytenr) 3251 { 3252 struct btrfs_delayed_ref_head *head; 3253 struct btrfs_delayed_ref_root *delayed_refs; 3254 int ret = 0; 3255 3256 delayed_refs = &trans->transaction->delayed_refs; 3257 spin_lock(&delayed_refs->lock); 3258 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr); 3259 if (!head) 3260 goto out_delayed_unlock; 3261 3262 spin_lock(&head->lock); 3263 if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root)) 3264 goto out; 3265 3266 if (cleanup_extent_op(head) != NULL) 3267 goto out; 3268 3269 /* 3270 * waiting for the lock here would deadlock. If someone else has it 3271 * locked they are already in the process of dropping it anyway 3272 */ 3273 if (!mutex_trylock(&head->mutex)) 3274 goto out; 3275 3276 btrfs_delete_ref_head(delayed_refs, head); 3277 head->processing = false; 3278 3279 spin_unlock(&head->lock); 3280 spin_unlock(&delayed_refs->lock); 3281 3282 BUG_ON(head->extent_op); 3283 if (head->must_insert_reserved) 3284 ret = 1; 3285 3286 btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head); 3287 mutex_unlock(&head->mutex); 3288 btrfs_put_delayed_ref_head(head); 3289 return ret; 3290 out: 3291 spin_unlock(&head->lock); 3292 3293 out_delayed_unlock: 3294 spin_unlock(&delayed_refs->lock); 3295 return 0; 3296 } 3297 3298 int btrfs_free_tree_block(struct btrfs_trans_handle *trans, 3299 u64 root_id, 3300 struct extent_buffer *buf, 3301 u64 parent, int last_ref) 3302 { 3303 struct btrfs_fs_info *fs_info = trans->fs_info; 3304 struct btrfs_ref generic_ref = { 0 }; 3305 int ret; 3306 3307 btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF, 3308 buf->start, buf->len, parent); 3309 btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf), 3310 root_id, 0, false); 3311 3312 if (root_id != BTRFS_TREE_LOG_OBJECTID) { 3313 btrfs_ref_tree_mod(fs_info, &generic_ref); 3314 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL); 3315 if (ret < 0) 3316 return ret; 3317 } 3318 3319 if (last_ref && btrfs_header_generation(buf) == trans->transid) { 3320 struct btrfs_block_group *cache; 3321 bool must_pin = false; 3322 3323 if (root_id != BTRFS_TREE_LOG_OBJECTID) { 3324 ret = check_ref_cleanup(trans, buf->start); 3325 if (!ret) { 3326 btrfs_redirty_list_add(trans->transaction, buf); 3327 goto out; 3328 } 3329 } 3330 3331 cache = btrfs_lookup_block_group(fs_info, buf->start); 3332 3333 if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) { 3334 pin_down_extent(trans, cache, buf->start, buf->len, 1); 3335 btrfs_put_block_group(cache); 3336 goto out; 3337 } 3338 3339 /* 3340 * If there are tree mod log users we may have recorded mod log 3341 * operations for this node. If we re-allocate this node we 3342 * could replay operations on this node that happened when it 3343 * existed in a completely different root. For example if it 3344 * was part of root A, then was reallocated to root B, and we 3345 * are doing a btrfs_old_search_slot(root b), we could replay 3346 * operations that happened when the block was part of root A, 3347 * giving us an inconsistent view of the btree. 3348 * 3349 * We are safe from races here because at this point no other 3350 * node or root points to this extent buffer, so if after this 3351 * check a new tree mod log user joins we will not have an 3352 * existing log of operations on this node that we have to 3353 * contend with. 3354 */ 3355 if (test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags)) 3356 must_pin = true; 3357 3358 if (must_pin || btrfs_is_zoned(fs_info)) { 3359 btrfs_redirty_list_add(trans->transaction, buf); 3360 pin_down_extent(trans, cache, buf->start, buf->len, 1); 3361 btrfs_put_block_group(cache); 3362 goto out; 3363 } 3364 3365 WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags)); 3366 3367 btrfs_add_free_space(cache, buf->start, buf->len); 3368 btrfs_free_reserved_bytes(cache, buf->len, 0); 3369 btrfs_put_block_group(cache); 3370 trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len); 3371 } 3372 out: 3373 if (last_ref) { 3374 /* 3375 * Deleting the buffer, clear the corrupt flag since it doesn't 3376 * matter anymore. 3377 */ 3378 clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags); 3379 } 3380 return 0; 3381 } 3382 3383 /* Can return -ENOMEM */ 3384 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref) 3385 { 3386 struct btrfs_fs_info *fs_info = trans->fs_info; 3387 int ret; 3388 3389 if (btrfs_is_testing(fs_info)) 3390 return 0; 3391 3392 /* 3393 * tree log blocks never actually go into the extent allocation 3394 * tree, just update pinning info and exit early. 3395 */ 3396 if ((ref->type == BTRFS_REF_METADATA && 3397 ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID) || 3398 (ref->type == BTRFS_REF_DATA && 3399 ref->data_ref.owning_root == BTRFS_TREE_LOG_OBJECTID)) { 3400 /* unlocks the pinned mutex */ 3401 btrfs_pin_extent(trans, ref->bytenr, ref->len, 1); 3402 ret = 0; 3403 } else if (ref->type == BTRFS_REF_METADATA) { 3404 ret = btrfs_add_delayed_tree_ref(trans, ref, NULL); 3405 } else { 3406 ret = btrfs_add_delayed_data_ref(trans, ref, 0); 3407 } 3408 3409 if (!((ref->type == BTRFS_REF_METADATA && 3410 ref->tree_ref.owning_root == BTRFS_TREE_LOG_OBJECTID) || 3411 (ref->type == BTRFS_REF_DATA && 3412 ref->data_ref.owning_root == BTRFS_TREE_LOG_OBJECTID))) 3413 btrfs_ref_tree_mod(fs_info, ref); 3414 3415 return ret; 3416 } 3417 3418 enum btrfs_loop_type { 3419 /* 3420 * Start caching block groups but do not wait for progress or for them 3421 * to be done. 3422 */ 3423 LOOP_CACHING_NOWAIT, 3424 3425 /* 3426 * Wait for the block group free_space >= the space we're waiting for if 3427 * the block group isn't cached. 3428 */ 3429 LOOP_CACHING_WAIT, 3430 3431 /* 3432 * Allow allocations to happen from block groups that do not yet have a 3433 * size classification. 3434 */ 3435 LOOP_UNSET_SIZE_CLASS, 3436 3437 /* 3438 * Allocate a chunk and then retry the allocation. 3439 */ 3440 LOOP_ALLOC_CHUNK, 3441 3442 /* 3443 * Ignore the size class restrictions for this allocation. 3444 */ 3445 LOOP_WRONG_SIZE_CLASS, 3446 3447 /* 3448 * Ignore the empty size, only try to allocate the number of bytes 3449 * needed for this allocation. 3450 */ 3451 LOOP_NO_EMPTY_SIZE, 3452 }; 3453 3454 static inline void 3455 btrfs_lock_block_group(struct btrfs_block_group *cache, 3456 int delalloc) 3457 { 3458 if (delalloc) 3459 down_read(&cache->data_rwsem); 3460 } 3461 3462 static inline void btrfs_grab_block_group(struct btrfs_block_group *cache, 3463 int delalloc) 3464 { 3465 btrfs_get_block_group(cache); 3466 if (delalloc) 3467 down_read(&cache->data_rwsem); 3468 } 3469 3470 static struct btrfs_block_group *btrfs_lock_cluster( 3471 struct btrfs_block_group *block_group, 3472 struct btrfs_free_cluster *cluster, 3473 int delalloc) 3474 __acquires(&cluster->refill_lock) 3475 { 3476 struct btrfs_block_group *used_bg = NULL; 3477 3478 spin_lock(&cluster->refill_lock); 3479 while (1) { 3480 used_bg = cluster->block_group; 3481 if (!used_bg) 3482 return NULL; 3483 3484 if (used_bg == block_group) 3485 return used_bg; 3486 3487 btrfs_get_block_group(used_bg); 3488 3489 if (!delalloc) 3490 return used_bg; 3491 3492 if (down_read_trylock(&used_bg->data_rwsem)) 3493 return used_bg; 3494 3495 spin_unlock(&cluster->refill_lock); 3496 3497 /* We should only have one-level nested. */ 3498 down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING); 3499 3500 spin_lock(&cluster->refill_lock); 3501 if (used_bg == cluster->block_group) 3502 return used_bg; 3503 3504 up_read(&used_bg->data_rwsem); 3505 btrfs_put_block_group(used_bg); 3506 } 3507 } 3508 3509 static inline void 3510 btrfs_release_block_group(struct btrfs_block_group *cache, 3511 int delalloc) 3512 { 3513 if (delalloc) 3514 up_read(&cache->data_rwsem); 3515 btrfs_put_block_group(cache); 3516 } 3517 3518 /* 3519 * Helper function for find_free_extent(). 3520 * 3521 * Return -ENOENT to inform caller that we need fallback to unclustered mode. 3522 * Return >0 to inform caller that we find nothing 3523 * Return 0 means we have found a location and set ffe_ctl->found_offset. 3524 */ 3525 static int find_free_extent_clustered(struct btrfs_block_group *bg, 3526 struct find_free_extent_ctl *ffe_ctl, 3527 struct btrfs_block_group **cluster_bg_ret) 3528 { 3529 struct btrfs_block_group *cluster_bg; 3530 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr; 3531 u64 aligned_cluster; 3532 u64 offset; 3533 int ret; 3534 3535 cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc); 3536 if (!cluster_bg) 3537 goto refill_cluster; 3538 if (cluster_bg != bg && (cluster_bg->ro || 3539 !block_group_bits(cluster_bg, ffe_ctl->flags))) 3540 goto release_cluster; 3541 3542 offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr, 3543 ffe_ctl->num_bytes, cluster_bg->start, 3544 &ffe_ctl->max_extent_size); 3545 if (offset) { 3546 /* We have a block, we're done */ 3547 spin_unlock(&last_ptr->refill_lock); 3548 trace_btrfs_reserve_extent_cluster(cluster_bg, ffe_ctl); 3549 *cluster_bg_ret = cluster_bg; 3550 ffe_ctl->found_offset = offset; 3551 return 0; 3552 } 3553 WARN_ON(last_ptr->block_group != cluster_bg); 3554 3555 release_cluster: 3556 /* 3557 * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so 3558 * lets just skip it and let the allocator find whatever block it can 3559 * find. If we reach this point, we will have tried the cluster 3560 * allocator plenty of times and not have found anything, so we are 3561 * likely way too fragmented for the clustering stuff to find anything. 3562 * 3563 * However, if the cluster is taken from the current block group, 3564 * release the cluster first, so that we stand a better chance of 3565 * succeeding in the unclustered allocation. 3566 */ 3567 if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) { 3568 spin_unlock(&last_ptr->refill_lock); 3569 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc); 3570 return -ENOENT; 3571 } 3572 3573 /* This cluster didn't work out, free it and start over */ 3574 btrfs_return_cluster_to_free_space(NULL, last_ptr); 3575 3576 if (cluster_bg != bg) 3577 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc); 3578 3579 refill_cluster: 3580 if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) { 3581 spin_unlock(&last_ptr->refill_lock); 3582 return -ENOENT; 3583 } 3584 3585 aligned_cluster = max_t(u64, 3586 ffe_ctl->empty_cluster + ffe_ctl->empty_size, 3587 bg->full_stripe_len); 3588 ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start, 3589 ffe_ctl->num_bytes, aligned_cluster); 3590 if (ret == 0) { 3591 /* Now pull our allocation out of this cluster */ 3592 offset = btrfs_alloc_from_cluster(bg, last_ptr, 3593 ffe_ctl->num_bytes, ffe_ctl->search_start, 3594 &ffe_ctl->max_extent_size); 3595 if (offset) { 3596 /* We found one, proceed */ 3597 spin_unlock(&last_ptr->refill_lock); 3598 ffe_ctl->found_offset = offset; 3599 trace_btrfs_reserve_extent_cluster(bg, ffe_ctl); 3600 return 0; 3601 } 3602 } 3603 /* 3604 * At this point we either didn't find a cluster or we weren't able to 3605 * allocate a block from our cluster. Free the cluster we've been 3606 * trying to use, and go to the next block group. 3607 */ 3608 btrfs_return_cluster_to_free_space(NULL, last_ptr); 3609 spin_unlock(&last_ptr->refill_lock); 3610 return 1; 3611 } 3612 3613 /* 3614 * Return >0 to inform caller that we find nothing 3615 * Return 0 when we found an free extent and set ffe_ctrl->found_offset 3616 */ 3617 static int find_free_extent_unclustered(struct btrfs_block_group *bg, 3618 struct find_free_extent_ctl *ffe_ctl) 3619 { 3620 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr; 3621 u64 offset; 3622 3623 /* 3624 * We are doing an unclustered allocation, set the fragmented flag so 3625 * we don't bother trying to setup a cluster again until we get more 3626 * space. 3627 */ 3628 if (unlikely(last_ptr)) { 3629 spin_lock(&last_ptr->lock); 3630 last_ptr->fragmented = 1; 3631 spin_unlock(&last_ptr->lock); 3632 } 3633 if (ffe_ctl->cached) { 3634 struct btrfs_free_space_ctl *free_space_ctl; 3635 3636 free_space_ctl = bg->free_space_ctl; 3637 spin_lock(&free_space_ctl->tree_lock); 3638 if (free_space_ctl->free_space < 3639 ffe_ctl->num_bytes + ffe_ctl->empty_cluster + 3640 ffe_ctl->empty_size) { 3641 ffe_ctl->total_free_space = max_t(u64, 3642 ffe_ctl->total_free_space, 3643 free_space_ctl->free_space); 3644 spin_unlock(&free_space_ctl->tree_lock); 3645 return 1; 3646 } 3647 spin_unlock(&free_space_ctl->tree_lock); 3648 } 3649 3650 offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start, 3651 ffe_ctl->num_bytes, ffe_ctl->empty_size, 3652 &ffe_ctl->max_extent_size); 3653 if (!offset) 3654 return 1; 3655 ffe_ctl->found_offset = offset; 3656 return 0; 3657 } 3658 3659 static int do_allocation_clustered(struct btrfs_block_group *block_group, 3660 struct find_free_extent_ctl *ffe_ctl, 3661 struct btrfs_block_group **bg_ret) 3662 { 3663 int ret; 3664 3665 /* We want to try and use the cluster allocator, so lets look there */ 3666 if (ffe_ctl->last_ptr && ffe_ctl->use_cluster) { 3667 ret = find_free_extent_clustered(block_group, ffe_ctl, bg_ret); 3668 if (ret >= 0) 3669 return ret; 3670 /* ret == -ENOENT case falls through */ 3671 } 3672 3673 return find_free_extent_unclustered(block_group, ffe_ctl); 3674 } 3675 3676 /* 3677 * Tree-log block group locking 3678 * ============================ 3679 * 3680 * fs_info::treelog_bg_lock protects the fs_info::treelog_bg which 3681 * indicates the starting address of a block group, which is reserved only 3682 * for tree-log metadata. 3683 * 3684 * Lock nesting 3685 * ============ 3686 * 3687 * space_info::lock 3688 * block_group::lock 3689 * fs_info::treelog_bg_lock 3690 */ 3691 3692 /* 3693 * Simple allocator for sequential-only block group. It only allows sequential 3694 * allocation. No need to play with trees. This function also reserves the 3695 * bytes as in btrfs_add_reserved_bytes. 3696 */ 3697 static int do_allocation_zoned(struct btrfs_block_group *block_group, 3698 struct find_free_extent_ctl *ffe_ctl, 3699 struct btrfs_block_group **bg_ret) 3700 { 3701 struct btrfs_fs_info *fs_info = block_group->fs_info; 3702 struct btrfs_space_info *space_info = block_group->space_info; 3703 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl; 3704 u64 start = block_group->start; 3705 u64 num_bytes = ffe_ctl->num_bytes; 3706 u64 avail; 3707 u64 bytenr = block_group->start; 3708 u64 log_bytenr; 3709 u64 data_reloc_bytenr; 3710 int ret = 0; 3711 bool skip = false; 3712 3713 ASSERT(btrfs_is_zoned(block_group->fs_info)); 3714 3715 /* 3716 * Do not allow non-tree-log blocks in the dedicated tree-log block 3717 * group, and vice versa. 3718 */ 3719 spin_lock(&fs_info->treelog_bg_lock); 3720 log_bytenr = fs_info->treelog_bg; 3721 if (log_bytenr && ((ffe_ctl->for_treelog && bytenr != log_bytenr) || 3722 (!ffe_ctl->for_treelog && bytenr == log_bytenr))) 3723 skip = true; 3724 spin_unlock(&fs_info->treelog_bg_lock); 3725 if (skip) 3726 return 1; 3727 3728 /* 3729 * Do not allow non-relocation blocks in the dedicated relocation block 3730 * group, and vice versa. 3731 */ 3732 spin_lock(&fs_info->relocation_bg_lock); 3733 data_reloc_bytenr = fs_info->data_reloc_bg; 3734 if (data_reloc_bytenr && 3735 ((ffe_ctl->for_data_reloc && bytenr != data_reloc_bytenr) || 3736 (!ffe_ctl->for_data_reloc && bytenr == data_reloc_bytenr))) 3737 skip = true; 3738 spin_unlock(&fs_info->relocation_bg_lock); 3739 if (skip) 3740 return 1; 3741 3742 /* Check RO and no space case before trying to activate it */ 3743 spin_lock(&block_group->lock); 3744 if (block_group->ro || btrfs_zoned_bg_is_full(block_group)) { 3745 ret = 1; 3746 /* 3747 * May need to clear fs_info->{treelog,data_reloc}_bg. 3748 * Return the error after taking the locks. 3749 */ 3750 } 3751 spin_unlock(&block_group->lock); 3752 3753 /* Metadata block group is activated at write time. */ 3754 if (!ret && (block_group->flags & BTRFS_BLOCK_GROUP_DATA) && 3755 !btrfs_zone_activate(block_group)) { 3756 ret = 1; 3757 /* 3758 * May need to clear fs_info->{treelog,data_reloc}_bg. 3759 * Return the error after taking the locks. 3760 */ 3761 } 3762 3763 spin_lock(&space_info->lock); 3764 spin_lock(&block_group->lock); 3765 spin_lock(&fs_info->treelog_bg_lock); 3766 spin_lock(&fs_info->relocation_bg_lock); 3767 3768 if (ret) 3769 goto out; 3770 3771 ASSERT(!ffe_ctl->for_treelog || 3772 block_group->start == fs_info->treelog_bg || 3773 fs_info->treelog_bg == 0); 3774 ASSERT(!ffe_ctl->for_data_reloc || 3775 block_group->start == fs_info->data_reloc_bg || 3776 fs_info->data_reloc_bg == 0); 3777 3778 if (block_group->ro || 3779 (!ffe_ctl->for_data_reloc && 3780 test_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC, &block_group->runtime_flags))) { 3781 ret = 1; 3782 goto out; 3783 } 3784 3785 /* 3786 * Do not allow currently using block group to be tree-log dedicated 3787 * block group. 3788 */ 3789 if (ffe_ctl->for_treelog && !fs_info->treelog_bg && 3790 (block_group->used || block_group->reserved)) { 3791 ret = 1; 3792 goto out; 3793 } 3794 3795 /* 3796 * Do not allow currently used block group to be the data relocation 3797 * dedicated block group. 3798 */ 3799 if (ffe_ctl->for_data_reloc && !fs_info->data_reloc_bg && 3800 (block_group->used || block_group->reserved)) { 3801 ret = 1; 3802 goto out; 3803 } 3804 3805 WARN_ON_ONCE(block_group->alloc_offset > block_group->zone_capacity); 3806 avail = block_group->zone_capacity - block_group->alloc_offset; 3807 if (avail < num_bytes) { 3808 if (ffe_ctl->max_extent_size < avail) { 3809 /* 3810 * With sequential allocator, free space is always 3811 * contiguous 3812 */ 3813 ffe_ctl->max_extent_size = avail; 3814 ffe_ctl->total_free_space = avail; 3815 } 3816 ret = 1; 3817 goto out; 3818 } 3819 3820 if (ffe_ctl->for_treelog && !fs_info->treelog_bg) 3821 fs_info->treelog_bg = block_group->start; 3822 3823 if (ffe_ctl->for_data_reloc) { 3824 if (!fs_info->data_reloc_bg) 3825 fs_info->data_reloc_bg = block_group->start; 3826 /* 3827 * Do not allow allocations from this block group, unless it is 3828 * for data relocation. Compared to increasing the ->ro, setting 3829 * the ->zoned_data_reloc_ongoing flag still allows nocow 3830 * writers to come in. See btrfs_inc_nocow_writers(). 3831 * 3832 * We need to disable an allocation to avoid an allocation of 3833 * regular (non-relocation data) extent. With mix of relocation 3834 * extents and regular extents, we can dispatch WRITE commands 3835 * (for relocation extents) and ZONE APPEND commands (for 3836 * regular extents) at the same time to the same zone, which 3837 * easily break the write pointer. 3838 * 3839 * Also, this flag avoids this block group to be zone finished. 3840 */ 3841 set_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC, &block_group->runtime_flags); 3842 } 3843 3844 ffe_ctl->found_offset = start + block_group->alloc_offset; 3845 block_group->alloc_offset += num_bytes; 3846 spin_lock(&ctl->tree_lock); 3847 ctl->free_space -= num_bytes; 3848 spin_unlock(&ctl->tree_lock); 3849 3850 /* 3851 * We do not check if found_offset is aligned to stripesize. The 3852 * address is anyway rewritten when using zone append writing. 3853 */ 3854 3855 ffe_ctl->search_start = ffe_ctl->found_offset; 3856 3857 out: 3858 if (ret && ffe_ctl->for_treelog) 3859 fs_info->treelog_bg = 0; 3860 if (ret && ffe_ctl->for_data_reloc) 3861 fs_info->data_reloc_bg = 0; 3862 spin_unlock(&fs_info->relocation_bg_lock); 3863 spin_unlock(&fs_info->treelog_bg_lock); 3864 spin_unlock(&block_group->lock); 3865 spin_unlock(&space_info->lock); 3866 return ret; 3867 } 3868 3869 static int do_allocation(struct btrfs_block_group *block_group, 3870 struct find_free_extent_ctl *ffe_ctl, 3871 struct btrfs_block_group **bg_ret) 3872 { 3873 switch (ffe_ctl->policy) { 3874 case BTRFS_EXTENT_ALLOC_CLUSTERED: 3875 return do_allocation_clustered(block_group, ffe_ctl, bg_ret); 3876 case BTRFS_EXTENT_ALLOC_ZONED: 3877 return do_allocation_zoned(block_group, ffe_ctl, bg_ret); 3878 default: 3879 BUG(); 3880 } 3881 } 3882 3883 static void release_block_group(struct btrfs_block_group *block_group, 3884 struct find_free_extent_ctl *ffe_ctl, 3885 int delalloc) 3886 { 3887 switch (ffe_ctl->policy) { 3888 case BTRFS_EXTENT_ALLOC_CLUSTERED: 3889 ffe_ctl->retry_uncached = false; 3890 break; 3891 case BTRFS_EXTENT_ALLOC_ZONED: 3892 /* Nothing to do */ 3893 break; 3894 default: 3895 BUG(); 3896 } 3897 3898 BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) != 3899 ffe_ctl->index); 3900 btrfs_release_block_group(block_group, delalloc); 3901 } 3902 3903 static void found_extent_clustered(struct find_free_extent_ctl *ffe_ctl, 3904 struct btrfs_key *ins) 3905 { 3906 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr; 3907 3908 if (!ffe_ctl->use_cluster && last_ptr) { 3909 spin_lock(&last_ptr->lock); 3910 last_ptr->window_start = ins->objectid; 3911 spin_unlock(&last_ptr->lock); 3912 } 3913 } 3914 3915 static void found_extent(struct find_free_extent_ctl *ffe_ctl, 3916 struct btrfs_key *ins) 3917 { 3918 switch (ffe_ctl->policy) { 3919 case BTRFS_EXTENT_ALLOC_CLUSTERED: 3920 found_extent_clustered(ffe_ctl, ins); 3921 break; 3922 case BTRFS_EXTENT_ALLOC_ZONED: 3923 /* Nothing to do */ 3924 break; 3925 default: 3926 BUG(); 3927 } 3928 } 3929 3930 static int can_allocate_chunk_zoned(struct btrfs_fs_info *fs_info, 3931 struct find_free_extent_ctl *ffe_ctl) 3932 { 3933 /* Block group's activeness is not a requirement for METADATA block groups. */ 3934 if (!(ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA)) 3935 return 0; 3936 3937 /* If we can activate new zone, just allocate a chunk and use it */ 3938 if (btrfs_can_activate_zone(fs_info->fs_devices, ffe_ctl->flags)) 3939 return 0; 3940 3941 /* 3942 * We already reached the max active zones. Try to finish one block 3943 * group to make a room for a new block group. This is only possible 3944 * for a data block group because btrfs_zone_finish() may need to wait 3945 * for a running transaction which can cause a deadlock for metadata 3946 * allocation. 3947 */ 3948 if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA) { 3949 int ret = btrfs_zone_finish_one_bg(fs_info); 3950 3951 if (ret == 1) 3952 return 0; 3953 else if (ret < 0) 3954 return ret; 3955 } 3956 3957 /* 3958 * If we have enough free space left in an already active block group 3959 * and we can't activate any other zone now, do not allow allocating a 3960 * new chunk and let find_free_extent() retry with a smaller size. 3961 */ 3962 if (ffe_ctl->max_extent_size >= ffe_ctl->min_alloc_size) 3963 return -ENOSPC; 3964 3965 /* 3966 * Even min_alloc_size is not left in any block groups. Since we cannot 3967 * activate a new block group, allocating it may not help. Let's tell a 3968 * caller to try again and hope it progress something by writing some 3969 * parts of the region. That is only possible for data block groups, 3970 * where a part of the region can be written. 3971 */ 3972 if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA) 3973 return -EAGAIN; 3974 3975 /* 3976 * We cannot activate a new block group and no enough space left in any 3977 * block groups. So, allocating a new block group may not help. But, 3978 * there is nothing to do anyway, so let's go with it. 3979 */ 3980 return 0; 3981 } 3982 3983 static int can_allocate_chunk(struct btrfs_fs_info *fs_info, 3984 struct find_free_extent_ctl *ffe_ctl) 3985 { 3986 switch (ffe_ctl->policy) { 3987 case BTRFS_EXTENT_ALLOC_CLUSTERED: 3988 return 0; 3989 case BTRFS_EXTENT_ALLOC_ZONED: 3990 return can_allocate_chunk_zoned(fs_info, ffe_ctl); 3991 default: 3992 BUG(); 3993 } 3994 } 3995 3996 /* 3997 * Return >0 means caller needs to re-search for free extent 3998 * Return 0 means we have the needed free extent. 3999 * Return <0 means we failed to locate any free extent. 4000 */ 4001 static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info, 4002 struct btrfs_key *ins, 4003 struct find_free_extent_ctl *ffe_ctl, 4004 bool full_search) 4005 { 4006 struct btrfs_root *root = fs_info->chunk_root; 4007 int ret; 4008 4009 if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) && 4010 ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg) 4011 ffe_ctl->orig_have_caching_bg = true; 4012 4013 if (ins->objectid) { 4014 found_extent(ffe_ctl, ins); 4015 return 0; 4016 } 4017 4018 if (ffe_ctl->loop >= LOOP_CACHING_WAIT && ffe_ctl->have_caching_bg) 4019 return 1; 4020 4021 ffe_ctl->index++; 4022 if (ffe_ctl->index < BTRFS_NR_RAID_TYPES) 4023 return 1; 4024 4025 /* See the comments for btrfs_loop_type for an explanation of the phases. */ 4026 if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) { 4027 ffe_ctl->index = 0; 4028 /* 4029 * We want to skip the LOOP_CACHING_WAIT step if we don't have 4030 * any uncached bgs and we've already done a full search 4031 * through. 4032 */ 4033 if (ffe_ctl->loop == LOOP_CACHING_NOWAIT && 4034 (!ffe_ctl->orig_have_caching_bg && full_search)) 4035 ffe_ctl->loop++; 4036 ffe_ctl->loop++; 4037 4038 if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) { 4039 struct btrfs_trans_handle *trans; 4040 int exist = 0; 4041 4042 /* Check if allocation policy allows to create a new chunk */ 4043 ret = can_allocate_chunk(fs_info, ffe_ctl); 4044 if (ret) 4045 return ret; 4046 4047 trans = current->journal_info; 4048 if (trans) 4049 exist = 1; 4050 else 4051 trans = btrfs_join_transaction(root); 4052 4053 if (IS_ERR(trans)) { 4054 ret = PTR_ERR(trans); 4055 return ret; 4056 } 4057 4058 ret = btrfs_chunk_alloc(trans, ffe_ctl->flags, 4059 CHUNK_ALLOC_FORCE_FOR_EXTENT); 4060 4061 /* Do not bail out on ENOSPC since we can do more. */ 4062 if (ret == -ENOSPC) { 4063 ret = 0; 4064 ffe_ctl->loop++; 4065 } 4066 else if (ret < 0) 4067 btrfs_abort_transaction(trans, ret); 4068 else 4069 ret = 0; 4070 if (!exist) 4071 btrfs_end_transaction(trans); 4072 if (ret) 4073 return ret; 4074 } 4075 4076 if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) { 4077 if (ffe_ctl->policy != BTRFS_EXTENT_ALLOC_CLUSTERED) 4078 return -ENOSPC; 4079 4080 /* 4081 * Don't loop again if we already have no empty_size and 4082 * no empty_cluster. 4083 */ 4084 if (ffe_ctl->empty_size == 0 && 4085 ffe_ctl->empty_cluster == 0) 4086 return -ENOSPC; 4087 ffe_ctl->empty_size = 0; 4088 ffe_ctl->empty_cluster = 0; 4089 } 4090 return 1; 4091 } 4092 return -ENOSPC; 4093 } 4094 4095 static bool find_free_extent_check_size_class(struct find_free_extent_ctl *ffe_ctl, 4096 struct btrfs_block_group *bg) 4097 { 4098 if (ffe_ctl->policy == BTRFS_EXTENT_ALLOC_ZONED) 4099 return true; 4100 if (!btrfs_block_group_should_use_size_class(bg)) 4101 return true; 4102 if (ffe_ctl->loop >= LOOP_WRONG_SIZE_CLASS) 4103 return true; 4104 if (ffe_ctl->loop >= LOOP_UNSET_SIZE_CLASS && 4105 bg->size_class == BTRFS_BG_SZ_NONE) 4106 return true; 4107 return ffe_ctl->size_class == bg->size_class; 4108 } 4109 4110 static int prepare_allocation_clustered(struct btrfs_fs_info *fs_info, 4111 struct find_free_extent_ctl *ffe_ctl, 4112 struct btrfs_space_info *space_info, 4113 struct btrfs_key *ins) 4114 { 4115 /* 4116 * If our free space is heavily fragmented we may not be able to make 4117 * big contiguous allocations, so instead of doing the expensive search 4118 * for free space, simply return ENOSPC with our max_extent_size so we 4119 * can go ahead and search for a more manageable chunk. 4120 * 4121 * If our max_extent_size is large enough for our allocation simply 4122 * disable clustering since we will likely not be able to find enough 4123 * space to create a cluster and induce latency trying. 4124 */ 4125 if (space_info->max_extent_size) { 4126 spin_lock(&space_info->lock); 4127 if (space_info->max_extent_size && 4128 ffe_ctl->num_bytes > space_info->max_extent_size) { 4129 ins->offset = space_info->max_extent_size; 4130 spin_unlock(&space_info->lock); 4131 return -ENOSPC; 4132 } else if (space_info->max_extent_size) { 4133 ffe_ctl->use_cluster = false; 4134 } 4135 spin_unlock(&space_info->lock); 4136 } 4137 4138 ffe_ctl->last_ptr = fetch_cluster_info(fs_info, space_info, 4139 &ffe_ctl->empty_cluster); 4140 if (ffe_ctl->last_ptr) { 4141 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr; 4142 4143 spin_lock(&last_ptr->lock); 4144 if (last_ptr->block_group) 4145 ffe_ctl->hint_byte = last_ptr->window_start; 4146 if (last_ptr->fragmented) { 4147 /* 4148 * We still set window_start so we can keep track of the 4149 * last place we found an allocation to try and save 4150 * some time. 4151 */ 4152 ffe_ctl->hint_byte = last_ptr->window_start; 4153 ffe_ctl->use_cluster = false; 4154 } 4155 spin_unlock(&last_ptr->lock); 4156 } 4157 4158 return 0; 4159 } 4160 4161 static int prepare_allocation_zoned(struct btrfs_fs_info *fs_info, 4162 struct find_free_extent_ctl *ffe_ctl) 4163 { 4164 if (ffe_ctl->for_treelog) { 4165 spin_lock(&fs_info->treelog_bg_lock); 4166 if (fs_info->treelog_bg) 4167 ffe_ctl->hint_byte = fs_info->treelog_bg; 4168 spin_unlock(&fs_info->treelog_bg_lock); 4169 } else if (ffe_ctl->for_data_reloc) { 4170 spin_lock(&fs_info->relocation_bg_lock); 4171 if (fs_info->data_reloc_bg) 4172 ffe_ctl->hint_byte = fs_info->data_reloc_bg; 4173 spin_unlock(&fs_info->relocation_bg_lock); 4174 } else if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA) { 4175 struct btrfs_block_group *block_group; 4176 4177 spin_lock(&fs_info->zone_active_bgs_lock); 4178 list_for_each_entry(block_group, &fs_info->zone_active_bgs, active_bg_list) { 4179 /* 4180 * No lock is OK here because avail is monotinically 4181 * decreasing, and this is just a hint. 4182 */ 4183 u64 avail = block_group->zone_capacity - block_group->alloc_offset; 4184 4185 if (block_group_bits(block_group, ffe_ctl->flags) && 4186 avail >= ffe_ctl->num_bytes) { 4187 ffe_ctl->hint_byte = block_group->start; 4188 break; 4189 } 4190 } 4191 spin_unlock(&fs_info->zone_active_bgs_lock); 4192 } 4193 4194 return 0; 4195 } 4196 4197 static int prepare_allocation(struct btrfs_fs_info *fs_info, 4198 struct find_free_extent_ctl *ffe_ctl, 4199 struct btrfs_space_info *space_info, 4200 struct btrfs_key *ins) 4201 { 4202 switch (ffe_ctl->policy) { 4203 case BTRFS_EXTENT_ALLOC_CLUSTERED: 4204 return prepare_allocation_clustered(fs_info, ffe_ctl, 4205 space_info, ins); 4206 case BTRFS_EXTENT_ALLOC_ZONED: 4207 return prepare_allocation_zoned(fs_info, ffe_ctl); 4208 default: 4209 BUG(); 4210 } 4211 } 4212 4213 /* 4214 * walks the btree of allocated extents and find a hole of a given size. 4215 * The key ins is changed to record the hole: 4216 * ins->objectid == start position 4217 * ins->flags = BTRFS_EXTENT_ITEM_KEY 4218 * ins->offset == the size of the hole. 4219 * Any available blocks before search_start are skipped. 4220 * 4221 * If there is no suitable free space, we will record the max size of 4222 * the free space extent currently. 4223 * 4224 * The overall logic and call chain: 4225 * 4226 * find_free_extent() 4227 * |- Iterate through all block groups 4228 * | |- Get a valid block group 4229 * | |- Try to do clustered allocation in that block group 4230 * | |- Try to do unclustered allocation in that block group 4231 * | |- Check if the result is valid 4232 * | | |- If valid, then exit 4233 * | |- Jump to next block group 4234 * | 4235 * |- Push harder to find free extents 4236 * |- If not found, re-iterate all block groups 4237 */ 4238 static noinline int find_free_extent(struct btrfs_root *root, 4239 struct btrfs_key *ins, 4240 struct find_free_extent_ctl *ffe_ctl) 4241 { 4242 struct btrfs_fs_info *fs_info = root->fs_info; 4243 int ret = 0; 4244 int cache_block_group_error = 0; 4245 struct btrfs_block_group *block_group = NULL; 4246 struct btrfs_space_info *space_info; 4247 bool full_search = false; 4248 4249 WARN_ON(ffe_ctl->num_bytes < fs_info->sectorsize); 4250 4251 ffe_ctl->search_start = 0; 4252 /* For clustered allocation */ 4253 ffe_ctl->empty_cluster = 0; 4254 ffe_ctl->last_ptr = NULL; 4255 ffe_ctl->use_cluster = true; 4256 ffe_ctl->have_caching_bg = false; 4257 ffe_ctl->orig_have_caching_bg = false; 4258 ffe_ctl->index = btrfs_bg_flags_to_raid_index(ffe_ctl->flags); 4259 ffe_ctl->loop = 0; 4260 ffe_ctl->retry_uncached = false; 4261 ffe_ctl->cached = 0; 4262 ffe_ctl->max_extent_size = 0; 4263 ffe_ctl->total_free_space = 0; 4264 ffe_ctl->found_offset = 0; 4265 ffe_ctl->policy = BTRFS_EXTENT_ALLOC_CLUSTERED; 4266 ffe_ctl->size_class = btrfs_calc_block_group_size_class(ffe_ctl->num_bytes); 4267 4268 if (btrfs_is_zoned(fs_info)) 4269 ffe_ctl->policy = BTRFS_EXTENT_ALLOC_ZONED; 4270 4271 ins->type = BTRFS_EXTENT_ITEM_KEY; 4272 ins->objectid = 0; 4273 ins->offset = 0; 4274 4275 trace_find_free_extent(root, ffe_ctl); 4276 4277 space_info = btrfs_find_space_info(fs_info, ffe_ctl->flags); 4278 if (!space_info) { 4279 btrfs_err(fs_info, "No space info for %llu", ffe_ctl->flags); 4280 return -ENOSPC; 4281 } 4282 4283 ret = prepare_allocation(fs_info, ffe_ctl, space_info, ins); 4284 if (ret < 0) 4285 return ret; 4286 4287 ffe_ctl->search_start = max(ffe_ctl->search_start, 4288 first_logical_byte(fs_info)); 4289 ffe_ctl->search_start = max(ffe_ctl->search_start, ffe_ctl->hint_byte); 4290 if (ffe_ctl->search_start == ffe_ctl->hint_byte) { 4291 block_group = btrfs_lookup_block_group(fs_info, 4292 ffe_ctl->search_start); 4293 /* 4294 * we don't want to use the block group if it doesn't match our 4295 * allocation bits, or if its not cached. 4296 * 4297 * However if we are re-searching with an ideal block group 4298 * picked out then we don't care that the block group is cached. 4299 */ 4300 if (block_group && block_group_bits(block_group, ffe_ctl->flags) && 4301 block_group->cached != BTRFS_CACHE_NO) { 4302 down_read(&space_info->groups_sem); 4303 if (list_empty(&block_group->list) || 4304 block_group->ro) { 4305 /* 4306 * someone is removing this block group, 4307 * we can't jump into the have_block_group 4308 * target because our list pointers are not 4309 * valid 4310 */ 4311 btrfs_put_block_group(block_group); 4312 up_read(&space_info->groups_sem); 4313 } else { 4314 ffe_ctl->index = btrfs_bg_flags_to_raid_index( 4315 block_group->flags); 4316 btrfs_lock_block_group(block_group, 4317 ffe_ctl->delalloc); 4318 ffe_ctl->hinted = true; 4319 goto have_block_group; 4320 } 4321 } else if (block_group) { 4322 btrfs_put_block_group(block_group); 4323 } 4324 } 4325 search: 4326 trace_find_free_extent_search_loop(root, ffe_ctl); 4327 ffe_ctl->have_caching_bg = false; 4328 if (ffe_ctl->index == btrfs_bg_flags_to_raid_index(ffe_ctl->flags) || 4329 ffe_ctl->index == 0) 4330 full_search = true; 4331 down_read(&space_info->groups_sem); 4332 list_for_each_entry(block_group, 4333 &space_info->block_groups[ffe_ctl->index], list) { 4334 struct btrfs_block_group *bg_ret; 4335 4336 ffe_ctl->hinted = false; 4337 /* If the block group is read-only, we can skip it entirely. */ 4338 if (unlikely(block_group->ro)) { 4339 if (ffe_ctl->for_treelog) 4340 btrfs_clear_treelog_bg(block_group); 4341 if (ffe_ctl->for_data_reloc) 4342 btrfs_clear_data_reloc_bg(block_group); 4343 continue; 4344 } 4345 4346 btrfs_grab_block_group(block_group, ffe_ctl->delalloc); 4347 ffe_ctl->search_start = block_group->start; 4348 4349 /* 4350 * this can happen if we end up cycling through all the 4351 * raid types, but we want to make sure we only allocate 4352 * for the proper type. 4353 */ 4354 if (!block_group_bits(block_group, ffe_ctl->flags)) { 4355 u64 extra = BTRFS_BLOCK_GROUP_DUP | 4356 BTRFS_BLOCK_GROUP_RAID1_MASK | 4357 BTRFS_BLOCK_GROUP_RAID56_MASK | 4358 BTRFS_BLOCK_GROUP_RAID10; 4359 4360 /* 4361 * if they asked for extra copies and this block group 4362 * doesn't provide them, bail. This does allow us to 4363 * fill raid0 from raid1. 4364 */ 4365 if ((ffe_ctl->flags & extra) && !(block_group->flags & extra)) 4366 goto loop; 4367 4368 /* 4369 * This block group has different flags than we want. 4370 * It's possible that we have MIXED_GROUP flag but no 4371 * block group is mixed. Just skip such block group. 4372 */ 4373 btrfs_release_block_group(block_group, ffe_ctl->delalloc); 4374 continue; 4375 } 4376 4377 have_block_group: 4378 trace_find_free_extent_have_block_group(root, ffe_ctl, block_group); 4379 ffe_ctl->cached = btrfs_block_group_done(block_group); 4380 if (unlikely(!ffe_ctl->cached)) { 4381 ffe_ctl->have_caching_bg = true; 4382 ret = btrfs_cache_block_group(block_group, false); 4383 4384 /* 4385 * If we get ENOMEM here or something else we want to 4386 * try other block groups, because it may not be fatal. 4387 * However if we can't find anything else we need to 4388 * save our return here so that we return the actual 4389 * error that caused problems, not ENOSPC. 4390 */ 4391 if (ret < 0) { 4392 if (!cache_block_group_error) 4393 cache_block_group_error = ret; 4394 ret = 0; 4395 goto loop; 4396 } 4397 ret = 0; 4398 } 4399 4400 if (unlikely(block_group->cached == BTRFS_CACHE_ERROR)) { 4401 if (!cache_block_group_error) 4402 cache_block_group_error = -EIO; 4403 goto loop; 4404 } 4405 4406 if (!find_free_extent_check_size_class(ffe_ctl, block_group)) 4407 goto loop; 4408 4409 bg_ret = NULL; 4410 ret = do_allocation(block_group, ffe_ctl, &bg_ret); 4411 if (ret > 0) 4412 goto loop; 4413 4414 if (bg_ret && bg_ret != block_group) { 4415 btrfs_release_block_group(block_group, ffe_ctl->delalloc); 4416 block_group = bg_ret; 4417 } 4418 4419 /* Checks */ 4420 ffe_ctl->search_start = round_up(ffe_ctl->found_offset, 4421 fs_info->stripesize); 4422 4423 /* move on to the next group */ 4424 if (ffe_ctl->search_start + ffe_ctl->num_bytes > 4425 block_group->start + block_group->length) { 4426 btrfs_add_free_space_unused(block_group, 4427 ffe_ctl->found_offset, 4428 ffe_ctl->num_bytes); 4429 goto loop; 4430 } 4431 4432 if (ffe_ctl->found_offset < ffe_ctl->search_start) 4433 btrfs_add_free_space_unused(block_group, 4434 ffe_ctl->found_offset, 4435 ffe_ctl->search_start - ffe_ctl->found_offset); 4436 4437 ret = btrfs_add_reserved_bytes(block_group, ffe_ctl->ram_bytes, 4438 ffe_ctl->num_bytes, 4439 ffe_ctl->delalloc, 4440 ffe_ctl->loop >= LOOP_WRONG_SIZE_CLASS); 4441 if (ret == -EAGAIN) { 4442 btrfs_add_free_space_unused(block_group, 4443 ffe_ctl->found_offset, 4444 ffe_ctl->num_bytes); 4445 goto loop; 4446 } 4447 btrfs_inc_block_group_reservations(block_group); 4448 4449 /* we are all good, lets return */ 4450 ins->objectid = ffe_ctl->search_start; 4451 ins->offset = ffe_ctl->num_bytes; 4452 4453 trace_btrfs_reserve_extent(block_group, ffe_ctl); 4454 btrfs_release_block_group(block_group, ffe_ctl->delalloc); 4455 break; 4456 loop: 4457 if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT && 4458 !ffe_ctl->retry_uncached) { 4459 ffe_ctl->retry_uncached = true; 4460 btrfs_wait_block_group_cache_progress(block_group, 4461 ffe_ctl->num_bytes + 4462 ffe_ctl->empty_cluster + 4463 ffe_ctl->empty_size); 4464 goto have_block_group; 4465 } 4466 release_block_group(block_group, ffe_ctl, ffe_ctl->delalloc); 4467 cond_resched(); 4468 } 4469 up_read(&space_info->groups_sem); 4470 4471 ret = find_free_extent_update_loop(fs_info, ins, ffe_ctl, full_search); 4472 if (ret > 0) 4473 goto search; 4474 4475 if (ret == -ENOSPC && !cache_block_group_error) { 4476 /* 4477 * Use ffe_ctl->total_free_space as fallback if we can't find 4478 * any contiguous hole. 4479 */ 4480 if (!ffe_ctl->max_extent_size) 4481 ffe_ctl->max_extent_size = ffe_ctl->total_free_space; 4482 spin_lock(&space_info->lock); 4483 space_info->max_extent_size = ffe_ctl->max_extent_size; 4484 spin_unlock(&space_info->lock); 4485 ins->offset = ffe_ctl->max_extent_size; 4486 } else if (ret == -ENOSPC) { 4487 ret = cache_block_group_error; 4488 } 4489 return ret; 4490 } 4491 4492 /* 4493 * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a 4494 * hole that is at least as big as @num_bytes. 4495 * 4496 * @root - The root that will contain this extent 4497 * 4498 * @ram_bytes - The amount of space in ram that @num_bytes take. This 4499 * is used for accounting purposes. This value differs 4500 * from @num_bytes only in the case of compressed extents. 4501 * 4502 * @num_bytes - Number of bytes to allocate on-disk. 4503 * 4504 * @min_alloc_size - Indicates the minimum amount of space that the 4505 * allocator should try to satisfy. In some cases 4506 * @num_bytes may be larger than what is required and if 4507 * the filesystem is fragmented then allocation fails. 4508 * However, the presence of @min_alloc_size gives a 4509 * chance to try and satisfy the smaller allocation. 4510 * 4511 * @empty_size - A hint that you plan on doing more COW. This is the 4512 * size in bytes the allocator should try to find free 4513 * next to the block it returns. This is just a hint and 4514 * may be ignored by the allocator. 4515 * 4516 * @hint_byte - Hint to the allocator to start searching above the byte 4517 * address passed. It might be ignored. 4518 * 4519 * @ins - This key is modified to record the found hole. It will 4520 * have the following values: 4521 * ins->objectid == start position 4522 * ins->flags = BTRFS_EXTENT_ITEM_KEY 4523 * ins->offset == the size of the hole. 4524 * 4525 * @is_data - Boolean flag indicating whether an extent is 4526 * allocated for data (true) or metadata (false) 4527 * 4528 * @delalloc - Boolean flag indicating whether this allocation is for 4529 * delalloc or not. If 'true' data_rwsem of block groups 4530 * is going to be acquired. 4531 * 4532 * 4533 * Returns 0 when an allocation succeeded or < 0 when an error occurred. In 4534 * case -ENOSPC is returned then @ins->offset will contain the size of the 4535 * largest available hole the allocator managed to find. 4536 */ 4537 int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes, 4538 u64 num_bytes, u64 min_alloc_size, 4539 u64 empty_size, u64 hint_byte, 4540 struct btrfs_key *ins, int is_data, int delalloc) 4541 { 4542 struct btrfs_fs_info *fs_info = root->fs_info; 4543 struct find_free_extent_ctl ffe_ctl = {}; 4544 bool final_tried = num_bytes == min_alloc_size; 4545 u64 flags; 4546 int ret; 4547 bool for_treelog = (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID); 4548 bool for_data_reloc = (btrfs_is_data_reloc_root(root) && is_data); 4549 4550 flags = get_alloc_profile_by_root(root, is_data); 4551 again: 4552 WARN_ON(num_bytes < fs_info->sectorsize); 4553 4554 ffe_ctl.ram_bytes = ram_bytes; 4555 ffe_ctl.num_bytes = num_bytes; 4556 ffe_ctl.min_alloc_size = min_alloc_size; 4557 ffe_ctl.empty_size = empty_size; 4558 ffe_ctl.flags = flags; 4559 ffe_ctl.delalloc = delalloc; 4560 ffe_ctl.hint_byte = hint_byte; 4561 ffe_ctl.for_treelog = for_treelog; 4562 ffe_ctl.for_data_reloc = for_data_reloc; 4563 4564 ret = find_free_extent(root, ins, &ffe_ctl); 4565 if (!ret && !is_data) { 4566 btrfs_dec_block_group_reservations(fs_info, ins->objectid); 4567 } else if (ret == -ENOSPC) { 4568 if (!final_tried && ins->offset) { 4569 num_bytes = min(num_bytes >> 1, ins->offset); 4570 num_bytes = round_down(num_bytes, 4571 fs_info->sectorsize); 4572 num_bytes = max(num_bytes, min_alloc_size); 4573 ram_bytes = num_bytes; 4574 if (num_bytes == min_alloc_size) 4575 final_tried = true; 4576 goto again; 4577 } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) { 4578 struct btrfs_space_info *sinfo; 4579 4580 sinfo = btrfs_find_space_info(fs_info, flags); 4581 btrfs_err(fs_info, 4582 "allocation failed flags %llu, wanted %llu tree-log %d, relocation: %d", 4583 flags, num_bytes, for_treelog, for_data_reloc); 4584 if (sinfo) 4585 btrfs_dump_space_info(fs_info, sinfo, 4586 num_bytes, 1); 4587 } 4588 } 4589 4590 return ret; 4591 } 4592 4593 int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info, 4594 u64 start, u64 len, int delalloc) 4595 { 4596 struct btrfs_block_group *cache; 4597 4598 cache = btrfs_lookup_block_group(fs_info, start); 4599 if (!cache) { 4600 btrfs_err(fs_info, "Unable to find block group for %llu", 4601 start); 4602 return -ENOSPC; 4603 } 4604 4605 btrfs_add_free_space(cache, start, len); 4606 btrfs_free_reserved_bytes(cache, len, delalloc); 4607 trace_btrfs_reserved_extent_free(fs_info, start, len); 4608 4609 btrfs_put_block_group(cache); 4610 return 0; 4611 } 4612 4613 int btrfs_pin_reserved_extent(struct btrfs_trans_handle *trans, u64 start, 4614 u64 len) 4615 { 4616 struct btrfs_block_group *cache; 4617 int ret = 0; 4618 4619 cache = btrfs_lookup_block_group(trans->fs_info, start); 4620 if (!cache) { 4621 btrfs_err(trans->fs_info, "unable to find block group for %llu", 4622 start); 4623 return -ENOSPC; 4624 } 4625 4626 ret = pin_down_extent(trans, cache, start, len, 1); 4627 btrfs_put_block_group(cache); 4628 return ret; 4629 } 4630 4631 static int alloc_reserved_extent(struct btrfs_trans_handle *trans, u64 bytenr, 4632 u64 num_bytes) 4633 { 4634 struct btrfs_fs_info *fs_info = trans->fs_info; 4635 int ret; 4636 4637 ret = remove_from_free_space_tree(trans, bytenr, num_bytes); 4638 if (ret) 4639 return ret; 4640 4641 ret = btrfs_update_block_group(trans, bytenr, num_bytes, true); 4642 if (ret) { 4643 ASSERT(!ret); 4644 btrfs_err(fs_info, "update block group failed for %llu %llu", 4645 bytenr, num_bytes); 4646 return ret; 4647 } 4648 4649 trace_btrfs_reserved_extent_alloc(fs_info, bytenr, num_bytes); 4650 return 0; 4651 } 4652 4653 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans, 4654 u64 parent, u64 root_objectid, 4655 u64 flags, u64 owner, u64 offset, 4656 struct btrfs_key *ins, int ref_mod) 4657 { 4658 struct btrfs_fs_info *fs_info = trans->fs_info; 4659 struct btrfs_root *extent_root; 4660 int ret; 4661 struct btrfs_extent_item *extent_item; 4662 struct btrfs_extent_inline_ref *iref; 4663 struct btrfs_path *path; 4664 struct extent_buffer *leaf; 4665 int type; 4666 u32 size; 4667 4668 if (parent > 0) 4669 type = BTRFS_SHARED_DATA_REF_KEY; 4670 else 4671 type = BTRFS_EXTENT_DATA_REF_KEY; 4672 4673 size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type); 4674 4675 path = btrfs_alloc_path(); 4676 if (!path) 4677 return -ENOMEM; 4678 4679 extent_root = btrfs_extent_root(fs_info, ins->objectid); 4680 ret = btrfs_insert_empty_item(trans, extent_root, path, ins, size); 4681 if (ret) { 4682 btrfs_free_path(path); 4683 return ret; 4684 } 4685 4686 leaf = path->nodes[0]; 4687 extent_item = btrfs_item_ptr(leaf, path->slots[0], 4688 struct btrfs_extent_item); 4689 btrfs_set_extent_refs(leaf, extent_item, ref_mod); 4690 btrfs_set_extent_generation(leaf, extent_item, trans->transid); 4691 btrfs_set_extent_flags(leaf, extent_item, 4692 flags | BTRFS_EXTENT_FLAG_DATA); 4693 4694 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1); 4695 btrfs_set_extent_inline_ref_type(leaf, iref, type); 4696 if (parent > 0) { 4697 struct btrfs_shared_data_ref *ref; 4698 ref = (struct btrfs_shared_data_ref *)(iref + 1); 4699 btrfs_set_extent_inline_ref_offset(leaf, iref, parent); 4700 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod); 4701 } else { 4702 struct btrfs_extent_data_ref *ref; 4703 ref = (struct btrfs_extent_data_ref *)(&iref->offset); 4704 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid); 4705 btrfs_set_extent_data_ref_objectid(leaf, ref, owner); 4706 btrfs_set_extent_data_ref_offset(leaf, ref, offset); 4707 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod); 4708 } 4709 4710 btrfs_mark_buffer_dirty(trans, path->nodes[0]); 4711 btrfs_free_path(path); 4712 4713 return alloc_reserved_extent(trans, ins->objectid, ins->offset); 4714 } 4715 4716 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans, 4717 struct btrfs_delayed_ref_node *node, 4718 struct btrfs_delayed_extent_op *extent_op) 4719 { 4720 struct btrfs_fs_info *fs_info = trans->fs_info; 4721 struct btrfs_root *extent_root; 4722 int ret; 4723 struct btrfs_extent_item *extent_item; 4724 struct btrfs_key extent_key; 4725 struct btrfs_tree_block_info *block_info; 4726 struct btrfs_extent_inline_ref *iref; 4727 struct btrfs_path *path; 4728 struct extent_buffer *leaf; 4729 struct btrfs_delayed_tree_ref *ref; 4730 u32 size = sizeof(*extent_item) + sizeof(*iref); 4731 u64 flags = extent_op->flags_to_set; 4732 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA); 4733 4734 ref = btrfs_delayed_node_to_tree_ref(node); 4735 4736 extent_key.objectid = node->bytenr; 4737 if (skinny_metadata) { 4738 extent_key.offset = ref->level; 4739 extent_key.type = BTRFS_METADATA_ITEM_KEY; 4740 } else { 4741 extent_key.offset = node->num_bytes; 4742 extent_key.type = BTRFS_EXTENT_ITEM_KEY; 4743 size += sizeof(*block_info); 4744 } 4745 4746 path = btrfs_alloc_path(); 4747 if (!path) 4748 return -ENOMEM; 4749 4750 extent_root = btrfs_extent_root(fs_info, extent_key.objectid); 4751 ret = btrfs_insert_empty_item(trans, extent_root, path, &extent_key, 4752 size); 4753 if (ret) { 4754 btrfs_free_path(path); 4755 return ret; 4756 } 4757 4758 leaf = path->nodes[0]; 4759 extent_item = btrfs_item_ptr(leaf, path->slots[0], 4760 struct btrfs_extent_item); 4761 btrfs_set_extent_refs(leaf, extent_item, 1); 4762 btrfs_set_extent_generation(leaf, extent_item, trans->transid); 4763 btrfs_set_extent_flags(leaf, extent_item, 4764 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK); 4765 4766 if (skinny_metadata) { 4767 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1); 4768 } else { 4769 block_info = (struct btrfs_tree_block_info *)(extent_item + 1); 4770 btrfs_set_tree_block_key(leaf, block_info, &extent_op->key); 4771 btrfs_set_tree_block_level(leaf, block_info, ref->level); 4772 iref = (struct btrfs_extent_inline_ref *)(block_info + 1); 4773 } 4774 4775 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) { 4776 btrfs_set_extent_inline_ref_type(leaf, iref, 4777 BTRFS_SHARED_BLOCK_REF_KEY); 4778 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent); 4779 } else { 4780 btrfs_set_extent_inline_ref_type(leaf, iref, 4781 BTRFS_TREE_BLOCK_REF_KEY); 4782 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root); 4783 } 4784 4785 btrfs_mark_buffer_dirty(trans, leaf); 4786 btrfs_free_path(path); 4787 4788 return alloc_reserved_extent(trans, node->bytenr, fs_info->nodesize); 4789 } 4790 4791 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans, 4792 struct btrfs_root *root, u64 owner, 4793 u64 offset, u64 ram_bytes, 4794 struct btrfs_key *ins) 4795 { 4796 struct btrfs_ref generic_ref = { 0 }; 4797 4798 BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID); 4799 4800 btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT, 4801 ins->objectid, ins->offset, 0); 4802 btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, 4803 offset, 0, false); 4804 btrfs_ref_tree_mod(root->fs_info, &generic_ref); 4805 4806 return btrfs_add_delayed_data_ref(trans, &generic_ref, ram_bytes); 4807 } 4808 4809 /* 4810 * this is used by the tree logging recovery code. It records that 4811 * an extent has been allocated and makes sure to clear the free 4812 * space cache bits as well 4813 */ 4814 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans, 4815 u64 root_objectid, u64 owner, u64 offset, 4816 struct btrfs_key *ins) 4817 { 4818 struct btrfs_fs_info *fs_info = trans->fs_info; 4819 int ret; 4820 struct btrfs_block_group *block_group; 4821 struct btrfs_space_info *space_info; 4822 4823 /* 4824 * Mixed block groups will exclude before processing the log so we only 4825 * need to do the exclude dance if this fs isn't mixed. 4826 */ 4827 if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) { 4828 ret = __exclude_logged_extent(fs_info, ins->objectid, 4829 ins->offset); 4830 if (ret) 4831 return ret; 4832 } 4833 4834 block_group = btrfs_lookup_block_group(fs_info, ins->objectid); 4835 if (!block_group) 4836 return -EINVAL; 4837 4838 space_info = block_group->space_info; 4839 spin_lock(&space_info->lock); 4840 spin_lock(&block_group->lock); 4841 space_info->bytes_reserved += ins->offset; 4842 block_group->reserved += ins->offset; 4843 spin_unlock(&block_group->lock); 4844 spin_unlock(&space_info->lock); 4845 4846 ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner, 4847 offset, ins, 1); 4848 if (ret) 4849 btrfs_pin_extent(trans, ins->objectid, ins->offset, 1); 4850 btrfs_put_block_group(block_group); 4851 return ret; 4852 } 4853 4854 static struct extent_buffer * 4855 btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root, 4856 u64 bytenr, int level, u64 owner, 4857 enum btrfs_lock_nesting nest) 4858 { 4859 struct btrfs_fs_info *fs_info = root->fs_info; 4860 struct extent_buffer *buf; 4861 u64 lockdep_owner = owner; 4862 4863 buf = btrfs_find_create_tree_block(fs_info, bytenr, owner, level); 4864 if (IS_ERR(buf)) 4865 return buf; 4866 4867 /* 4868 * Extra safety check in case the extent tree is corrupted and extent 4869 * allocator chooses to use a tree block which is already used and 4870 * locked. 4871 */ 4872 if (buf->lock_owner == current->pid) { 4873 btrfs_err_rl(fs_info, 4874 "tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected", 4875 buf->start, btrfs_header_owner(buf), current->pid); 4876 free_extent_buffer(buf); 4877 return ERR_PTR(-EUCLEAN); 4878 } 4879 4880 /* 4881 * The reloc trees are just snapshots, so we need them to appear to be 4882 * just like any other fs tree WRT lockdep. 4883 * 4884 * The exception however is in replace_path() in relocation, where we 4885 * hold the lock on the original fs root and then search for the reloc 4886 * root. At that point we need to make sure any reloc root buffers are 4887 * set to the BTRFS_TREE_RELOC_OBJECTID lockdep class in order to make 4888 * lockdep happy. 4889 */ 4890 if (lockdep_owner == BTRFS_TREE_RELOC_OBJECTID && 4891 !test_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &root->state)) 4892 lockdep_owner = BTRFS_FS_TREE_OBJECTID; 4893 4894 /* btrfs_clear_buffer_dirty() accesses generation field. */ 4895 btrfs_set_header_generation(buf, trans->transid); 4896 4897 /* 4898 * This needs to stay, because we could allocate a freed block from an 4899 * old tree into a new tree, so we need to make sure this new block is 4900 * set to the appropriate level and owner. 4901 */ 4902 btrfs_set_buffer_lockdep_class(lockdep_owner, buf, level); 4903 4904 __btrfs_tree_lock(buf, nest); 4905 btrfs_clear_buffer_dirty(trans, buf); 4906 clear_bit(EXTENT_BUFFER_STALE, &buf->bflags); 4907 clear_bit(EXTENT_BUFFER_NO_CHECK, &buf->bflags); 4908 4909 set_extent_buffer_uptodate(buf); 4910 4911 memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header)); 4912 btrfs_set_header_level(buf, level); 4913 btrfs_set_header_bytenr(buf, buf->start); 4914 btrfs_set_header_generation(buf, trans->transid); 4915 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV); 4916 btrfs_set_header_owner(buf, owner); 4917 write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid); 4918 write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid); 4919 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) { 4920 buf->log_index = root->log_transid % 2; 4921 /* 4922 * we allow two log transactions at a time, use different 4923 * EXTENT bit to differentiate dirty pages. 4924 */ 4925 if (buf->log_index == 0) 4926 set_extent_bit(&root->dirty_log_pages, buf->start, 4927 buf->start + buf->len - 1, 4928 EXTENT_DIRTY, NULL); 4929 else 4930 set_extent_bit(&root->dirty_log_pages, buf->start, 4931 buf->start + buf->len - 1, 4932 EXTENT_NEW, NULL); 4933 } else { 4934 buf->log_index = -1; 4935 set_extent_bit(&trans->transaction->dirty_pages, buf->start, 4936 buf->start + buf->len - 1, EXTENT_DIRTY, NULL); 4937 } 4938 /* this returns a buffer locked for blocking */ 4939 return buf; 4940 } 4941 4942 /* 4943 * finds a free extent and does all the dirty work required for allocation 4944 * returns the tree buffer or an ERR_PTR on error. 4945 */ 4946 struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans, 4947 struct btrfs_root *root, 4948 u64 parent, u64 root_objectid, 4949 const struct btrfs_disk_key *key, 4950 int level, u64 hint, 4951 u64 empty_size, 4952 enum btrfs_lock_nesting nest) 4953 { 4954 struct btrfs_fs_info *fs_info = root->fs_info; 4955 struct btrfs_key ins; 4956 struct btrfs_block_rsv *block_rsv; 4957 struct extent_buffer *buf; 4958 struct btrfs_delayed_extent_op *extent_op; 4959 struct btrfs_ref generic_ref = { 0 }; 4960 u64 flags = 0; 4961 int ret; 4962 u32 blocksize = fs_info->nodesize; 4963 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA); 4964 4965 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS 4966 if (btrfs_is_testing(fs_info)) { 4967 buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr, 4968 level, root_objectid, nest); 4969 if (!IS_ERR(buf)) 4970 root->alloc_bytenr += blocksize; 4971 return buf; 4972 } 4973 #endif 4974 4975 block_rsv = btrfs_use_block_rsv(trans, root, blocksize); 4976 if (IS_ERR(block_rsv)) 4977 return ERR_CAST(block_rsv); 4978 4979 ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize, 4980 empty_size, hint, &ins, 0, 0); 4981 if (ret) 4982 goto out_unuse; 4983 4984 buf = btrfs_init_new_buffer(trans, root, ins.objectid, level, 4985 root_objectid, nest); 4986 if (IS_ERR(buf)) { 4987 ret = PTR_ERR(buf); 4988 goto out_free_reserved; 4989 } 4990 4991 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) { 4992 if (parent == 0) 4993 parent = ins.objectid; 4994 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF; 4995 } else 4996 BUG_ON(parent > 0); 4997 4998 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) { 4999 extent_op = btrfs_alloc_delayed_extent_op(); 5000 if (!extent_op) { 5001 ret = -ENOMEM; 5002 goto out_free_buf; 5003 } 5004 if (key) 5005 memcpy(&extent_op->key, key, sizeof(extent_op->key)); 5006 else 5007 memset(&extent_op->key, 0, sizeof(extent_op->key)); 5008 extent_op->flags_to_set = flags; 5009 extent_op->update_key = skinny_metadata ? false : true; 5010 extent_op->update_flags = true; 5011 extent_op->level = level; 5012 5013 btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT, 5014 ins.objectid, ins.offset, parent); 5015 btrfs_init_tree_ref(&generic_ref, level, root_objectid, 5016 root->root_key.objectid, false); 5017 btrfs_ref_tree_mod(fs_info, &generic_ref); 5018 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, extent_op); 5019 if (ret) 5020 goto out_free_delayed; 5021 } 5022 return buf; 5023 5024 out_free_delayed: 5025 btrfs_free_delayed_extent_op(extent_op); 5026 out_free_buf: 5027 btrfs_tree_unlock(buf); 5028 free_extent_buffer(buf); 5029 out_free_reserved: 5030 btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0); 5031 out_unuse: 5032 btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize); 5033 return ERR_PTR(ret); 5034 } 5035 5036 struct walk_control { 5037 u64 refs[BTRFS_MAX_LEVEL]; 5038 u64 flags[BTRFS_MAX_LEVEL]; 5039 struct btrfs_key update_progress; 5040 struct btrfs_key drop_progress; 5041 int drop_level; 5042 int stage; 5043 int level; 5044 int shared_level; 5045 int update_ref; 5046 int keep_locks; 5047 int reada_slot; 5048 int reada_count; 5049 int restarted; 5050 }; 5051 5052 #define DROP_REFERENCE 1 5053 #define UPDATE_BACKREF 2 5054 5055 static noinline void reada_walk_down(struct btrfs_trans_handle *trans, 5056 struct btrfs_root *root, 5057 struct walk_control *wc, 5058 struct btrfs_path *path) 5059 { 5060 struct btrfs_fs_info *fs_info = root->fs_info; 5061 u64 bytenr; 5062 u64 generation; 5063 u64 refs; 5064 u64 flags; 5065 u32 nritems; 5066 struct btrfs_key key; 5067 struct extent_buffer *eb; 5068 int ret; 5069 int slot; 5070 int nread = 0; 5071 5072 if (path->slots[wc->level] < wc->reada_slot) { 5073 wc->reada_count = wc->reada_count * 2 / 3; 5074 wc->reada_count = max(wc->reada_count, 2); 5075 } else { 5076 wc->reada_count = wc->reada_count * 3 / 2; 5077 wc->reada_count = min_t(int, wc->reada_count, 5078 BTRFS_NODEPTRS_PER_BLOCK(fs_info)); 5079 } 5080 5081 eb = path->nodes[wc->level]; 5082 nritems = btrfs_header_nritems(eb); 5083 5084 for (slot = path->slots[wc->level]; slot < nritems; slot++) { 5085 if (nread >= wc->reada_count) 5086 break; 5087 5088 cond_resched(); 5089 bytenr = btrfs_node_blockptr(eb, slot); 5090 generation = btrfs_node_ptr_generation(eb, slot); 5091 5092 if (slot == path->slots[wc->level]) 5093 goto reada; 5094 5095 if (wc->stage == UPDATE_BACKREF && 5096 generation <= root->root_key.offset) 5097 continue; 5098 5099 /* We don't lock the tree block, it's OK to be racy here */ 5100 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, 5101 wc->level - 1, 1, &refs, 5102 &flags); 5103 /* We don't care about errors in readahead. */ 5104 if (ret < 0) 5105 continue; 5106 5107 /* 5108 * This could be racey, it's conceivable that we raced and end 5109 * up with a bogus refs count, if that's the case just skip, if 5110 * we are actually corrupt we will notice when we look up 5111 * everything again with our locks. 5112 */ 5113 if (refs == 0) 5114 continue; 5115 5116 if (wc->stage == DROP_REFERENCE) { 5117 if (refs == 1) 5118 goto reada; 5119 5120 if (wc->level == 1 && 5121 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) 5122 continue; 5123 if (!wc->update_ref || 5124 generation <= root->root_key.offset) 5125 continue; 5126 btrfs_node_key_to_cpu(eb, &key, slot); 5127 ret = btrfs_comp_cpu_keys(&key, 5128 &wc->update_progress); 5129 if (ret < 0) 5130 continue; 5131 } else { 5132 if (wc->level == 1 && 5133 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) 5134 continue; 5135 } 5136 reada: 5137 btrfs_readahead_node_child(eb, slot); 5138 nread++; 5139 } 5140 wc->reada_slot = slot; 5141 } 5142 5143 /* 5144 * helper to process tree block while walking down the tree. 5145 * 5146 * when wc->stage == UPDATE_BACKREF, this function updates 5147 * back refs for pointers in the block. 5148 * 5149 * NOTE: return value 1 means we should stop walking down. 5150 */ 5151 static noinline int walk_down_proc(struct btrfs_trans_handle *trans, 5152 struct btrfs_root *root, 5153 struct btrfs_path *path, 5154 struct walk_control *wc, int lookup_info) 5155 { 5156 struct btrfs_fs_info *fs_info = root->fs_info; 5157 int level = wc->level; 5158 struct extent_buffer *eb = path->nodes[level]; 5159 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF; 5160 int ret; 5161 5162 if (wc->stage == UPDATE_BACKREF && 5163 btrfs_header_owner(eb) != root->root_key.objectid) 5164 return 1; 5165 5166 /* 5167 * when reference count of tree block is 1, it won't increase 5168 * again. once full backref flag is set, we never clear it. 5169 */ 5170 if (lookup_info && 5171 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) || 5172 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) { 5173 ASSERT(path->locks[level]); 5174 ret = btrfs_lookup_extent_info(trans, fs_info, 5175 eb->start, level, 1, 5176 &wc->refs[level], 5177 &wc->flags[level]); 5178 if (ret) 5179 return ret; 5180 if (unlikely(wc->refs[level] == 0)) { 5181 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0", 5182 eb->start); 5183 return -EUCLEAN; 5184 } 5185 } 5186 5187 if (wc->stage == DROP_REFERENCE) { 5188 if (wc->refs[level] > 1) 5189 return 1; 5190 5191 if (path->locks[level] && !wc->keep_locks) { 5192 btrfs_tree_unlock_rw(eb, path->locks[level]); 5193 path->locks[level] = 0; 5194 } 5195 return 0; 5196 } 5197 5198 /* wc->stage == UPDATE_BACKREF */ 5199 if (!(wc->flags[level] & flag)) { 5200 ASSERT(path->locks[level]); 5201 ret = btrfs_inc_ref(trans, root, eb, 1); 5202 BUG_ON(ret); /* -ENOMEM */ 5203 ret = btrfs_dec_ref(trans, root, eb, 0); 5204 BUG_ON(ret); /* -ENOMEM */ 5205 ret = btrfs_set_disk_extent_flags(trans, eb, flag); 5206 BUG_ON(ret); /* -ENOMEM */ 5207 wc->flags[level] |= flag; 5208 } 5209 5210 /* 5211 * the block is shared by multiple trees, so it's not good to 5212 * keep the tree lock 5213 */ 5214 if (path->locks[level] && level > 0) { 5215 btrfs_tree_unlock_rw(eb, path->locks[level]); 5216 path->locks[level] = 0; 5217 } 5218 return 0; 5219 } 5220 5221 /* 5222 * This is used to verify a ref exists for this root to deal with a bug where we 5223 * would have a drop_progress key that hadn't been updated properly. 5224 */ 5225 static int check_ref_exists(struct btrfs_trans_handle *trans, 5226 struct btrfs_root *root, u64 bytenr, u64 parent, 5227 int level) 5228 { 5229 struct btrfs_path *path; 5230 struct btrfs_extent_inline_ref *iref; 5231 int ret; 5232 5233 path = btrfs_alloc_path(); 5234 if (!path) 5235 return -ENOMEM; 5236 5237 ret = lookup_extent_backref(trans, path, &iref, bytenr, 5238 root->fs_info->nodesize, parent, 5239 root->root_key.objectid, level, 0); 5240 btrfs_free_path(path); 5241 if (ret == -ENOENT) 5242 return 0; 5243 if (ret < 0) 5244 return ret; 5245 return 1; 5246 } 5247 5248 /* 5249 * helper to process tree block pointer. 5250 * 5251 * when wc->stage == DROP_REFERENCE, this function checks 5252 * reference count of the block pointed to. if the block 5253 * is shared and we need update back refs for the subtree 5254 * rooted at the block, this function changes wc->stage to 5255 * UPDATE_BACKREF. if the block is shared and there is no 5256 * need to update back, this function drops the reference 5257 * to the block. 5258 * 5259 * NOTE: return value 1 means we should stop walking down. 5260 */ 5261 static noinline int do_walk_down(struct btrfs_trans_handle *trans, 5262 struct btrfs_root *root, 5263 struct btrfs_path *path, 5264 struct walk_control *wc, int *lookup_info) 5265 { 5266 struct btrfs_fs_info *fs_info = root->fs_info; 5267 u64 bytenr; 5268 u64 generation; 5269 u64 parent; 5270 struct btrfs_tree_parent_check check = { 0 }; 5271 struct btrfs_key key; 5272 struct btrfs_ref ref = { 0 }; 5273 struct extent_buffer *next; 5274 int level = wc->level; 5275 int reada = 0; 5276 int ret = 0; 5277 bool need_account = false; 5278 5279 generation = btrfs_node_ptr_generation(path->nodes[level], 5280 path->slots[level]); 5281 /* 5282 * if the lower level block was created before the snapshot 5283 * was created, we know there is no need to update back refs 5284 * for the subtree 5285 */ 5286 if (wc->stage == UPDATE_BACKREF && 5287 generation <= root->root_key.offset) { 5288 *lookup_info = 1; 5289 return 1; 5290 } 5291 5292 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]); 5293 5294 check.level = level - 1; 5295 check.transid = generation; 5296 check.owner_root = root->root_key.objectid; 5297 check.has_first_key = true; 5298 btrfs_node_key_to_cpu(path->nodes[level], &check.first_key, 5299 path->slots[level]); 5300 5301 next = find_extent_buffer(fs_info, bytenr); 5302 if (!next) { 5303 next = btrfs_find_create_tree_block(fs_info, bytenr, 5304 root->root_key.objectid, level - 1); 5305 if (IS_ERR(next)) 5306 return PTR_ERR(next); 5307 reada = 1; 5308 } 5309 btrfs_tree_lock(next); 5310 5311 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1, 5312 &wc->refs[level - 1], 5313 &wc->flags[level - 1]); 5314 if (ret < 0) 5315 goto out_unlock; 5316 5317 if (unlikely(wc->refs[level - 1] == 0)) { 5318 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0", 5319 bytenr); 5320 ret = -EUCLEAN; 5321 goto out_unlock; 5322 } 5323 *lookup_info = 0; 5324 5325 if (wc->stage == DROP_REFERENCE) { 5326 if (wc->refs[level - 1] > 1) { 5327 need_account = true; 5328 if (level == 1 && 5329 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF)) 5330 goto skip; 5331 5332 if (!wc->update_ref || 5333 generation <= root->root_key.offset) 5334 goto skip; 5335 5336 btrfs_node_key_to_cpu(path->nodes[level], &key, 5337 path->slots[level]); 5338 ret = btrfs_comp_cpu_keys(&key, &wc->update_progress); 5339 if (ret < 0) 5340 goto skip; 5341 5342 wc->stage = UPDATE_BACKREF; 5343 wc->shared_level = level - 1; 5344 } 5345 } else { 5346 if (level == 1 && 5347 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF)) 5348 goto skip; 5349 } 5350 5351 if (!btrfs_buffer_uptodate(next, generation, 0)) { 5352 btrfs_tree_unlock(next); 5353 free_extent_buffer(next); 5354 next = NULL; 5355 *lookup_info = 1; 5356 } 5357 5358 if (!next) { 5359 if (reada && level == 1) 5360 reada_walk_down(trans, root, wc, path); 5361 next = read_tree_block(fs_info, bytenr, &check); 5362 if (IS_ERR(next)) { 5363 return PTR_ERR(next); 5364 } else if (!extent_buffer_uptodate(next)) { 5365 free_extent_buffer(next); 5366 return -EIO; 5367 } 5368 btrfs_tree_lock(next); 5369 } 5370 5371 level--; 5372 ASSERT(level == btrfs_header_level(next)); 5373 if (level != btrfs_header_level(next)) { 5374 btrfs_err(root->fs_info, "mismatched level"); 5375 ret = -EIO; 5376 goto out_unlock; 5377 } 5378 path->nodes[level] = next; 5379 path->slots[level] = 0; 5380 path->locks[level] = BTRFS_WRITE_LOCK; 5381 wc->level = level; 5382 if (wc->level == 1) 5383 wc->reada_slot = 0; 5384 return 0; 5385 skip: 5386 wc->refs[level - 1] = 0; 5387 wc->flags[level - 1] = 0; 5388 if (wc->stage == DROP_REFERENCE) { 5389 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) { 5390 parent = path->nodes[level]->start; 5391 } else { 5392 ASSERT(root->root_key.objectid == 5393 btrfs_header_owner(path->nodes[level])); 5394 if (root->root_key.objectid != 5395 btrfs_header_owner(path->nodes[level])) { 5396 btrfs_err(root->fs_info, 5397 "mismatched block owner"); 5398 ret = -EIO; 5399 goto out_unlock; 5400 } 5401 parent = 0; 5402 } 5403 5404 /* 5405 * If we had a drop_progress we need to verify the refs are set 5406 * as expected. If we find our ref then we know that from here 5407 * on out everything should be correct, and we can clear the 5408 * ->restarted flag. 5409 */ 5410 if (wc->restarted) { 5411 ret = check_ref_exists(trans, root, bytenr, parent, 5412 level - 1); 5413 if (ret < 0) 5414 goto out_unlock; 5415 if (ret == 0) 5416 goto no_delete; 5417 ret = 0; 5418 wc->restarted = 0; 5419 } 5420 5421 /* 5422 * Reloc tree doesn't contribute to qgroup numbers, and we have 5423 * already accounted them at merge time (replace_path), 5424 * thus we could skip expensive subtree trace here. 5425 */ 5426 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID && 5427 need_account) { 5428 ret = btrfs_qgroup_trace_subtree(trans, next, 5429 generation, level - 1); 5430 if (ret) { 5431 btrfs_err_rl(fs_info, 5432 "Error %d accounting shared subtree. Quota is out of sync, rescan required.", 5433 ret); 5434 } 5435 } 5436 5437 /* 5438 * We need to update the next key in our walk control so we can 5439 * update the drop_progress key accordingly. We don't care if 5440 * find_next_key doesn't find a key because that means we're at 5441 * the end and are going to clean up now. 5442 */ 5443 wc->drop_level = level; 5444 find_next_key(path, level, &wc->drop_progress); 5445 5446 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr, 5447 fs_info->nodesize, parent); 5448 btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid, 5449 0, false); 5450 ret = btrfs_free_extent(trans, &ref); 5451 if (ret) 5452 goto out_unlock; 5453 } 5454 no_delete: 5455 *lookup_info = 1; 5456 ret = 1; 5457 5458 out_unlock: 5459 btrfs_tree_unlock(next); 5460 free_extent_buffer(next); 5461 5462 return ret; 5463 } 5464 5465 /* 5466 * helper to process tree block while walking up the tree. 5467 * 5468 * when wc->stage == DROP_REFERENCE, this function drops 5469 * reference count on the block. 5470 * 5471 * when wc->stage == UPDATE_BACKREF, this function changes 5472 * wc->stage back to DROP_REFERENCE if we changed wc->stage 5473 * to UPDATE_BACKREF previously while processing the block. 5474 * 5475 * NOTE: return value 1 means we should stop walking up. 5476 */ 5477 static noinline int walk_up_proc(struct btrfs_trans_handle *trans, 5478 struct btrfs_root *root, 5479 struct btrfs_path *path, 5480 struct walk_control *wc) 5481 { 5482 struct btrfs_fs_info *fs_info = root->fs_info; 5483 int ret = 0; 5484 int level = wc->level; 5485 struct extent_buffer *eb = path->nodes[level]; 5486 u64 parent = 0; 5487 5488 if (wc->stage == UPDATE_BACKREF) { 5489 BUG_ON(wc->shared_level < level); 5490 if (level < wc->shared_level) 5491 goto out; 5492 5493 ret = find_next_key(path, level + 1, &wc->update_progress); 5494 if (ret > 0) 5495 wc->update_ref = 0; 5496 5497 wc->stage = DROP_REFERENCE; 5498 wc->shared_level = -1; 5499 path->slots[level] = 0; 5500 5501 /* 5502 * check reference count again if the block isn't locked. 5503 * we should start walking down the tree again if reference 5504 * count is one. 5505 */ 5506 if (!path->locks[level]) { 5507 BUG_ON(level == 0); 5508 btrfs_tree_lock(eb); 5509 path->locks[level] = BTRFS_WRITE_LOCK; 5510 5511 ret = btrfs_lookup_extent_info(trans, fs_info, 5512 eb->start, level, 1, 5513 &wc->refs[level], 5514 &wc->flags[level]); 5515 if (ret < 0) { 5516 btrfs_tree_unlock_rw(eb, path->locks[level]); 5517 path->locks[level] = 0; 5518 return ret; 5519 } 5520 if (unlikely(wc->refs[level] == 0)) { 5521 btrfs_tree_unlock_rw(eb, path->locks[level]); 5522 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0", 5523 eb->start); 5524 return -EUCLEAN; 5525 } 5526 if (wc->refs[level] == 1) { 5527 btrfs_tree_unlock_rw(eb, path->locks[level]); 5528 path->locks[level] = 0; 5529 return 1; 5530 } 5531 } 5532 } 5533 5534 /* wc->stage == DROP_REFERENCE */ 5535 BUG_ON(wc->refs[level] > 1 && !path->locks[level]); 5536 5537 if (wc->refs[level] == 1) { 5538 if (level == 0) { 5539 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) 5540 ret = btrfs_dec_ref(trans, root, eb, 1); 5541 else 5542 ret = btrfs_dec_ref(trans, root, eb, 0); 5543 BUG_ON(ret); /* -ENOMEM */ 5544 if (is_fstree(root->root_key.objectid)) { 5545 ret = btrfs_qgroup_trace_leaf_items(trans, eb); 5546 if (ret) { 5547 btrfs_err_rl(fs_info, 5548 "error %d accounting leaf items, quota is out of sync, rescan required", 5549 ret); 5550 } 5551 } 5552 } 5553 /* Make block locked assertion in btrfs_clear_buffer_dirty happy. */ 5554 if (!path->locks[level]) { 5555 btrfs_tree_lock(eb); 5556 path->locks[level] = BTRFS_WRITE_LOCK; 5557 } 5558 btrfs_clear_buffer_dirty(trans, eb); 5559 } 5560 5561 if (eb == root->node) { 5562 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) 5563 parent = eb->start; 5564 else if (root->root_key.objectid != btrfs_header_owner(eb)) 5565 goto owner_mismatch; 5566 } else { 5567 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF) 5568 parent = path->nodes[level + 1]->start; 5569 else if (root->root_key.objectid != 5570 btrfs_header_owner(path->nodes[level + 1])) 5571 goto owner_mismatch; 5572 } 5573 5574 ret = btrfs_free_tree_block(trans, btrfs_root_id(root), eb, parent, 5575 wc->refs[level] == 1); 5576 if (ret < 0) 5577 btrfs_abort_transaction(trans, ret); 5578 out: 5579 wc->refs[level] = 0; 5580 wc->flags[level] = 0; 5581 return ret; 5582 5583 owner_mismatch: 5584 btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu", 5585 btrfs_header_owner(eb), root->root_key.objectid); 5586 return -EUCLEAN; 5587 } 5588 5589 static noinline int walk_down_tree(struct btrfs_trans_handle *trans, 5590 struct btrfs_root *root, 5591 struct btrfs_path *path, 5592 struct walk_control *wc) 5593 { 5594 int level = wc->level; 5595 int lookup_info = 1; 5596 int ret = 0; 5597 5598 while (level >= 0) { 5599 ret = walk_down_proc(trans, root, path, wc, lookup_info); 5600 if (ret) 5601 break; 5602 5603 if (level == 0) 5604 break; 5605 5606 if (path->slots[level] >= 5607 btrfs_header_nritems(path->nodes[level])) 5608 break; 5609 5610 ret = do_walk_down(trans, root, path, wc, &lookup_info); 5611 if (ret > 0) { 5612 path->slots[level]++; 5613 continue; 5614 } else if (ret < 0) 5615 break; 5616 level = wc->level; 5617 } 5618 return (ret == 1) ? 0 : ret; 5619 } 5620 5621 static noinline int walk_up_tree(struct btrfs_trans_handle *trans, 5622 struct btrfs_root *root, 5623 struct btrfs_path *path, 5624 struct walk_control *wc, int max_level) 5625 { 5626 int level = wc->level; 5627 int ret; 5628 5629 path->slots[level] = btrfs_header_nritems(path->nodes[level]); 5630 while (level < max_level && path->nodes[level]) { 5631 wc->level = level; 5632 if (path->slots[level] + 1 < 5633 btrfs_header_nritems(path->nodes[level])) { 5634 path->slots[level]++; 5635 return 0; 5636 } else { 5637 ret = walk_up_proc(trans, root, path, wc); 5638 if (ret > 0) 5639 return 0; 5640 if (ret < 0) 5641 return ret; 5642 5643 if (path->locks[level]) { 5644 btrfs_tree_unlock_rw(path->nodes[level], 5645 path->locks[level]); 5646 path->locks[level] = 0; 5647 } 5648 free_extent_buffer(path->nodes[level]); 5649 path->nodes[level] = NULL; 5650 level++; 5651 } 5652 } 5653 return 1; 5654 } 5655 5656 /* 5657 * drop a subvolume tree. 5658 * 5659 * this function traverses the tree freeing any blocks that only 5660 * referenced by the tree. 5661 * 5662 * when a shared tree block is found. this function decreases its 5663 * reference count by one. if update_ref is true, this function 5664 * also make sure backrefs for the shared block and all lower level 5665 * blocks are properly updated. 5666 * 5667 * If called with for_reloc == 0, may exit early with -EAGAIN 5668 */ 5669 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc) 5670 { 5671 const bool is_reloc_root = (root->root_key.objectid == 5672 BTRFS_TREE_RELOC_OBJECTID); 5673 struct btrfs_fs_info *fs_info = root->fs_info; 5674 struct btrfs_path *path; 5675 struct btrfs_trans_handle *trans; 5676 struct btrfs_root *tree_root = fs_info->tree_root; 5677 struct btrfs_root_item *root_item = &root->root_item; 5678 struct walk_control *wc; 5679 struct btrfs_key key; 5680 int err = 0; 5681 int ret; 5682 int level; 5683 bool root_dropped = false; 5684 bool unfinished_drop = false; 5685 5686 btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid); 5687 5688 path = btrfs_alloc_path(); 5689 if (!path) { 5690 err = -ENOMEM; 5691 goto out; 5692 } 5693 5694 wc = kzalloc(sizeof(*wc), GFP_NOFS); 5695 if (!wc) { 5696 btrfs_free_path(path); 5697 err = -ENOMEM; 5698 goto out; 5699 } 5700 5701 /* 5702 * Use join to avoid potential EINTR from transaction start. See 5703 * wait_reserve_ticket and the whole reservation callchain. 5704 */ 5705 if (for_reloc) 5706 trans = btrfs_join_transaction(tree_root); 5707 else 5708 trans = btrfs_start_transaction(tree_root, 0); 5709 if (IS_ERR(trans)) { 5710 err = PTR_ERR(trans); 5711 goto out_free; 5712 } 5713 5714 err = btrfs_run_delayed_items(trans); 5715 if (err) 5716 goto out_end_trans; 5717 5718 /* 5719 * This will help us catch people modifying the fs tree while we're 5720 * dropping it. It is unsafe to mess with the fs tree while it's being 5721 * dropped as we unlock the root node and parent nodes as we walk down 5722 * the tree, assuming nothing will change. If something does change 5723 * then we'll have stale information and drop references to blocks we've 5724 * already dropped. 5725 */ 5726 set_bit(BTRFS_ROOT_DELETING, &root->state); 5727 unfinished_drop = test_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state); 5728 5729 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) { 5730 level = btrfs_header_level(root->node); 5731 path->nodes[level] = btrfs_lock_root_node(root); 5732 path->slots[level] = 0; 5733 path->locks[level] = BTRFS_WRITE_LOCK; 5734 memset(&wc->update_progress, 0, 5735 sizeof(wc->update_progress)); 5736 } else { 5737 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress); 5738 memcpy(&wc->update_progress, &key, 5739 sizeof(wc->update_progress)); 5740 5741 level = btrfs_root_drop_level(root_item); 5742 BUG_ON(level == 0); 5743 path->lowest_level = level; 5744 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); 5745 path->lowest_level = 0; 5746 if (ret < 0) { 5747 err = ret; 5748 goto out_end_trans; 5749 } 5750 WARN_ON(ret > 0); 5751 5752 /* 5753 * unlock our path, this is safe because only this 5754 * function is allowed to delete this snapshot 5755 */ 5756 btrfs_unlock_up_safe(path, 0); 5757 5758 level = btrfs_header_level(root->node); 5759 while (1) { 5760 btrfs_tree_lock(path->nodes[level]); 5761 path->locks[level] = BTRFS_WRITE_LOCK; 5762 5763 ret = btrfs_lookup_extent_info(trans, fs_info, 5764 path->nodes[level]->start, 5765 level, 1, &wc->refs[level], 5766 &wc->flags[level]); 5767 if (ret < 0) { 5768 err = ret; 5769 goto out_end_trans; 5770 } 5771 BUG_ON(wc->refs[level] == 0); 5772 5773 if (level == btrfs_root_drop_level(root_item)) 5774 break; 5775 5776 btrfs_tree_unlock(path->nodes[level]); 5777 path->locks[level] = 0; 5778 WARN_ON(wc->refs[level] != 1); 5779 level--; 5780 } 5781 } 5782 5783 wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state); 5784 wc->level = level; 5785 wc->shared_level = -1; 5786 wc->stage = DROP_REFERENCE; 5787 wc->update_ref = update_ref; 5788 wc->keep_locks = 0; 5789 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info); 5790 5791 while (1) { 5792 5793 ret = walk_down_tree(trans, root, path, wc); 5794 if (ret < 0) { 5795 btrfs_abort_transaction(trans, ret); 5796 err = ret; 5797 break; 5798 } 5799 5800 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL); 5801 if (ret < 0) { 5802 btrfs_abort_transaction(trans, ret); 5803 err = ret; 5804 break; 5805 } 5806 5807 if (ret > 0) { 5808 BUG_ON(wc->stage != DROP_REFERENCE); 5809 break; 5810 } 5811 5812 if (wc->stage == DROP_REFERENCE) { 5813 wc->drop_level = wc->level; 5814 btrfs_node_key_to_cpu(path->nodes[wc->drop_level], 5815 &wc->drop_progress, 5816 path->slots[wc->drop_level]); 5817 } 5818 btrfs_cpu_key_to_disk(&root_item->drop_progress, 5819 &wc->drop_progress); 5820 btrfs_set_root_drop_level(root_item, wc->drop_level); 5821 5822 BUG_ON(wc->level == 0); 5823 if (btrfs_should_end_transaction(trans) || 5824 (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) { 5825 ret = btrfs_update_root(trans, tree_root, 5826 &root->root_key, 5827 root_item); 5828 if (ret) { 5829 btrfs_abort_transaction(trans, ret); 5830 err = ret; 5831 goto out_end_trans; 5832 } 5833 5834 if (!is_reloc_root) 5835 btrfs_set_last_root_drop_gen(fs_info, trans->transid); 5836 5837 btrfs_end_transaction_throttle(trans); 5838 if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) { 5839 btrfs_debug(fs_info, 5840 "drop snapshot early exit"); 5841 err = -EAGAIN; 5842 goto out_free; 5843 } 5844 5845 /* 5846 * Use join to avoid potential EINTR from transaction 5847 * start. See wait_reserve_ticket and the whole 5848 * reservation callchain. 5849 */ 5850 if (for_reloc) 5851 trans = btrfs_join_transaction(tree_root); 5852 else 5853 trans = btrfs_start_transaction(tree_root, 0); 5854 if (IS_ERR(trans)) { 5855 err = PTR_ERR(trans); 5856 goto out_free; 5857 } 5858 } 5859 } 5860 btrfs_release_path(path); 5861 if (err) 5862 goto out_end_trans; 5863 5864 ret = btrfs_del_root(trans, &root->root_key); 5865 if (ret) { 5866 btrfs_abort_transaction(trans, ret); 5867 err = ret; 5868 goto out_end_trans; 5869 } 5870 5871 if (!is_reloc_root) { 5872 ret = btrfs_find_root(tree_root, &root->root_key, path, 5873 NULL, NULL); 5874 if (ret < 0) { 5875 btrfs_abort_transaction(trans, ret); 5876 err = ret; 5877 goto out_end_trans; 5878 } else if (ret > 0) { 5879 /* if we fail to delete the orphan item this time 5880 * around, it'll get picked up the next time. 5881 * 5882 * The most common failure here is just -ENOENT. 5883 */ 5884 btrfs_del_orphan_item(trans, tree_root, 5885 root->root_key.objectid); 5886 } 5887 } 5888 5889 /* 5890 * This subvolume is going to be completely dropped, and won't be 5891 * recorded as dirty roots, thus pertrans meta rsv will not be freed at 5892 * commit transaction time. So free it here manually. 5893 */ 5894 btrfs_qgroup_convert_reserved_meta(root, INT_MAX); 5895 btrfs_qgroup_free_meta_all_pertrans(root); 5896 5897 if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state)) 5898 btrfs_add_dropped_root(trans, root); 5899 else 5900 btrfs_put_root(root); 5901 root_dropped = true; 5902 out_end_trans: 5903 if (!is_reloc_root) 5904 btrfs_set_last_root_drop_gen(fs_info, trans->transid); 5905 5906 btrfs_end_transaction_throttle(trans); 5907 out_free: 5908 kfree(wc); 5909 btrfs_free_path(path); 5910 out: 5911 /* 5912 * We were an unfinished drop root, check to see if there are any 5913 * pending, and if not clear and wake up any waiters. 5914 */ 5915 if (!err && unfinished_drop) 5916 btrfs_maybe_wake_unfinished_drop(fs_info); 5917 5918 /* 5919 * So if we need to stop dropping the snapshot for whatever reason we 5920 * need to make sure to add it back to the dead root list so that we 5921 * keep trying to do the work later. This also cleans up roots if we 5922 * don't have it in the radix (like when we recover after a power fail 5923 * or unmount) so we don't leak memory. 5924 */ 5925 if (!for_reloc && !root_dropped) 5926 btrfs_add_dead_root(root); 5927 return err; 5928 } 5929 5930 /* 5931 * drop subtree rooted at tree block 'node'. 5932 * 5933 * NOTE: this function will unlock and release tree block 'node' 5934 * only used by relocation code 5935 */ 5936 int btrfs_drop_subtree(struct btrfs_trans_handle *trans, 5937 struct btrfs_root *root, 5938 struct extent_buffer *node, 5939 struct extent_buffer *parent) 5940 { 5941 struct btrfs_fs_info *fs_info = root->fs_info; 5942 struct btrfs_path *path; 5943 struct walk_control *wc; 5944 int level; 5945 int parent_level; 5946 int ret = 0; 5947 int wret; 5948 5949 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID); 5950 5951 path = btrfs_alloc_path(); 5952 if (!path) 5953 return -ENOMEM; 5954 5955 wc = kzalloc(sizeof(*wc), GFP_NOFS); 5956 if (!wc) { 5957 btrfs_free_path(path); 5958 return -ENOMEM; 5959 } 5960 5961 btrfs_assert_tree_write_locked(parent); 5962 parent_level = btrfs_header_level(parent); 5963 atomic_inc(&parent->refs); 5964 path->nodes[parent_level] = parent; 5965 path->slots[parent_level] = btrfs_header_nritems(parent); 5966 5967 btrfs_assert_tree_write_locked(node); 5968 level = btrfs_header_level(node); 5969 path->nodes[level] = node; 5970 path->slots[level] = 0; 5971 path->locks[level] = BTRFS_WRITE_LOCK; 5972 5973 wc->refs[parent_level] = 1; 5974 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF; 5975 wc->level = level; 5976 wc->shared_level = -1; 5977 wc->stage = DROP_REFERENCE; 5978 wc->update_ref = 0; 5979 wc->keep_locks = 1; 5980 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info); 5981 5982 while (1) { 5983 wret = walk_down_tree(trans, root, path, wc); 5984 if (wret < 0) { 5985 ret = wret; 5986 break; 5987 } 5988 5989 wret = walk_up_tree(trans, root, path, wc, parent_level); 5990 if (wret < 0) 5991 ret = wret; 5992 if (wret != 0) 5993 break; 5994 } 5995 5996 kfree(wc); 5997 btrfs_free_path(path); 5998 return ret; 5999 } 6000 6001 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info, 6002 u64 start, u64 end) 6003 { 6004 return unpin_extent_range(fs_info, start, end, false); 6005 } 6006 6007 /* 6008 * It used to be that old block groups would be left around forever. 6009 * Iterating over them would be enough to trim unused space. Since we 6010 * now automatically remove them, we also need to iterate over unallocated 6011 * space. 6012 * 6013 * We don't want a transaction for this since the discard may take a 6014 * substantial amount of time. We don't require that a transaction be 6015 * running, but we do need to take a running transaction into account 6016 * to ensure that we're not discarding chunks that were released or 6017 * allocated in the current transaction. 6018 * 6019 * Holding the chunks lock will prevent other threads from allocating 6020 * or releasing chunks, but it won't prevent a running transaction 6021 * from committing and releasing the memory that the pending chunks 6022 * list head uses. For that, we need to take a reference to the 6023 * transaction and hold the commit root sem. We only need to hold 6024 * it while performing the free space search since we have already 6025 * held back allocations. 6026 */ 6027 static int btrfs_trim_free_extents(struct btrfs_device *device, u64 *trimmed) 6028 { 6029 u64 start = BTRFS_DEVICE_RANGE_RESERVED, len = 0, end = 0; 6030 int ret; 6031 6032 *trimmed = 0; 6033 6034 /* Discard not supported = nothing to do. */ 6035 if (!bdev_max_discard_sectors(device->bdev)) 6036 return 0; 6037 6038 /* Not writable = nothing to do. */ 6039 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) 6040 return 0; 6041 6042 /* No free space = nothing to do. */ 6043 if (device->total_bytes <= device->bytes_used) 6044 return 0; 6045 6046 ret = 0; 6047 6048 while (1) { 6049 struct btrfs_fs_info *fs_info = device->fs_info; 6050 u64 bytes; 6051 6052 ret = mutex_lock_interruptible(&fs_info->chunk_mutex); 6053 if (ret) 6054 break; 6055 6056 find_first_clear_extent_bit(&device->alloc_state, start, 6057 &start, &end, 6058 CHUNK_TRIMMED | CHUNK_ALLOCATED); 6059 6060 /* Check if there are any CHUNK_* bits left */ 6061 if (start > device->total_bytes) { 6062 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)); 6063 btrfs_warn_in_rcu(fs_info, 6064 "ignoring attempt to trim beyond device size: offset %llu length %llu device %s device size %llu", 6065 start, end - start + 1, 6066 btrfs_dev_name(device), 6067 device->total_bytes); 6068 mutex_unlock(&fs_info->chunk_mutex); 6069 ret = 0; 6070 break; 6071 } 6072 6073 /* Ensure we skip the reserved space on each device. */ 6074 start = max_t(u64, start, BTRFS_DEVICE_RANGE_RESERVED); 6075 6076 /* 6077 * If find_first_clear_extent_bit find a range that spans the 6078 * end of the device it will set end to -1, in this case it's up 6079 * to the caller to trim the value to the size of the device. 6080 */ 6081 end = min(end, device->total_bytes - 1); 6082 6083 len = end - start + 1; 6084 6085 /* We didn't find any extents */ 6086 if (!len) { 6087 mutex_unlock(&fs_info->chunk_mutex); 6088 ret = 0; 6089 break; 6090 } 6091 6092 ret = btrfs_issue_discard(device->bdev, start, len, 6093 &bytes); 6094 if (!ret) 6095 set_extent_bit(&device->alloc_state, start, 6096 start + bytes - 1, CHUNK_TRIMMED, NULL); 6097 mutex_unlock(&fs_info->chunk_mutex); 6098 6099 if (ret) 6100 break; 6101 6102 start += len; 6103 *trimmed += bytes; 6104 6105 if (btrfs_trim_interrupted()) { 6106 ret = -ERESTARTSYS; 6107 break; 6108 } 6109 6110 cond_resched(); 6111 } 6112 6113 return ret; 6114 } 6115 6116 /* 6117 * Trim the whole filesystem by: 6118 * 1) trimming the free space in each block group 6119 * 2) trimming the unallocated space on each device 6120 * 6121 * This will also continue trimming even if a block group or device encounters 6122 * an error. The return value will be the last error, or 0 if nothing bad 6123 * happens. 6124 */ 6125 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range) 6126 { 6127 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices; 6128 struct btrfs_block_group *cache = NULL; 6129 struct btrfs_device *device; 6130 u64 group_trimmed; 6131 u64 range_end = U64_MAX; 6132 u64 start; 6133 u64 end; 6134 u64 trimmed = 0; 6135 u64 bg_failed = 0; 6136 u64 dev_failed = 0; 6137 int bg_ret = 0; 6138 int dev_ret = 0; 6139 int ret = 0; 6140 6141 if (range->start == U64_MAX) 6142 return -EINVAL; 6143 6144 /* 6145 * Check range overflow if range->len is set. 6146 * The default range->len is U64_MAX. 6147 */ 6148 if (range->len != U64_MAX && 6149 check_add_overflow(range->start, range->len, &range_end)) 6150 return -EINVAL; 6151 6152 cache = btrfs_lookup_first_block_group(fs_info, range->start); 6153 for (; cache; cache = btrfs_next_block_group(cache)) { 6154 if (cache->start >= range_end) { 6155 btrfs_put_block_group(cache); 6156 break; 6157 } 6158 6159 start = max(range->start, cache->start); 6160 end = min(range_end, cache->start + cache->length); 6161 6162 if (end - start >= range->minlen) { 6163 if (!btrfs_block_group_done(cache)) { 6164 ret = btrfs_cache_block_group(cache, true); 6165 if (ret) { 6166 bg_failed++; 6167 bg_ret = ret; 6168 continue; 6169 } 6170 } 6171 ret = btrfs_trim_block_group(cache, 6172 &group_trimmed, 6173 start, 6174 end, 6175 range->minlen); 6176 6177 trimmed += group_trimmed; 6178 if (ret) { 6179 bg_failed++; 6180 bg_ret = ret; 6181 continue; 6182 } 6183 } 6184 } 6185 6186 if (bg_failed) 6187 btrfs_warn(fs_info, 6188 "failed to trim %llu block group(s), last error %d", 6189 bg_failed, bg_ret); 6190 6191 mutex_lock(&fs_devices->device_list_mutex); 6192 list_for_each_entry(device, &fs_devices->devices, dev_list) { 6193 if (test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state)) 6194 continue; 6195 6196 ret = btrfs_trim_free_extents(device, &group_trimmed); 6197 6198 trimmed += group_trimmed; 6199 if (ret) { 6200 dev_failed++; 6201 dev_ret = ret; 6202 break; 6203 } 6204 } 6205 mutex_unlock(&fs_devices->device_list_mutex); 6206 6207 if (dev_failed) 6208 btrfs_warn(fs_info, 6209 "failed to trim %llu device(s), last error %d", 6210 dev_failed, dev_ret); 6211 range->len = trimmed; 6212 if (bg_ret) 6213 return bg_ret; 6214 return dev_ret; 6215 } 6216