fs/btrfs/extent-tree.c

   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 "misc.h"
  20 #include "tree-log.h"
  21 #include "disk-io.h"
  22 #include "print-tree.h"
  23 #include "volumes.h"
  24 #include "raid56.h"
  25 #include "locking.h"
  26 #include "free-space-cache.h"
  27 #include "free-space-tree.h"
  28 #include "sysfs.h"
  29 #include "qgroup.h"
  30 #include "ref-verify.h"
  31 #include "space-info.h"
  32 #include "block-rsv.h"
  33 #include "delalloc-space.h"
  34 #include "block-group.h"
  35 #include "discard.h"
  36 #include "rcu-string.h"
  37
  38 #undef SCRAMBLE_DELAYED_REFS
  39
  40
  41 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
  42                                struct btrfs_delayed_ref_node *node, u64 parent,
  43                                u64 root_objectid, u64 owner_objectid,
  44                                u64 owner_offset, int refs_to_drop,
  45                                struct btrfs_delayed_extent_op *extra_op);
  46 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
  47                                     struct extent_buffer *leaf,
  48                                     struct btrfs_extent_item *ei);
  49 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
  50                                       u64 parent, u64 root_objectid,
  51                                       u64 flags, u64 owner, u64 offset,
  52                                       struct btrfs_key *ins, int ref_mod);
  53 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
  54                                      struct btrfs_delayed_ref_node *node,
  55                                      struct btrfs_delayed_extent_op *extent_op);
  56 static int find_next_key(struct btrfs_path *path, int level,
  57                          struct btrfs_key *key);
  58
  59 static int block_group_bits(struct btrfs_block_group *cache, u64 bits)
  60 {
  61         return (cache->flags & bits) == bits;
  62 }
  63
  64 int btrfs_add_excluded_extent(struct btrfs_fs_info *fs_info,
  65                               u64 start, u64 num_bytes)
  66 {
  67         u64 end = start + num_bytes - 1;
  68         set_extent_bits(&fs_info->excluded_extents, start, end,
  69                         EXTENT_UPTODATE);
  70         return 0;
  71 }
  72
  73 void btrfs_free_excluded_extents(struct btrfs_block_group *cache)
  74 {
  75         struct btrfs_fs_info *fs_info = cache->fs_info;
  76         u64 start, end;
  77
  78         start = cache->start;
  79         end = start + cache->length - 1;
  80
  81         clear_extent_bits(&fs_info->excluded_extents, start, end,
  82                           EXTENT_UPTODATE);
  83 }
  84
  85 static u64 generic_ref_to_space_flags(struct btrfs_ref *ref)
  86 {
  87         if (ref->type == BTRFS_REF_METADATA) {
  88                 if (ref->tree_ref.root == BTRFS_CHUNK_TREE_OBJECTID)
  89                         return BTRFS_BLOCK_GROUP_SYSTEM;
  90                 else
  91                         return BTRFS_BLOCK_GROUP_METADATA;
  92         }
  93         return BTRFS_BLOCK_GROUP_DATA;
  94 }
  95
  96 static void add_pinned_bytes(struct btrfs_fs_info *fs_info,
  97                              struct btrfs_ref *ref)
  98 {
  99         struct btrfs_space_info *space_info;
 100         u64 flags = generic_ref_to_space_flags(ref);
 101
 102         space_info = btrfs_find_space_info(fs_info, flags);
 103         ASSERT(space_info);
 104         percpu_counter_add_batch(&space_info->total_bytes_pinned, ref->len,
 105                     BTRFS_TOTAL_BYTES_PINNED_BATCH);
 106 }
 107
 108 static void sub_pinned_bytes(struct btrfs_fs_info *fs_info,
 109                              struct btrfs_ref *ref)
 110 {
 111         struct btrfs_space_info *space_info;
 112         u64 flags = generic_ref_to_space_flags(ref);
 113
 114         space_info = btrfs_find_space_info(fs_info, flags);
 115         ASSERT(space_info);
 116         percpu_counter_add_batch(&space_info->total_bytes_pinned, -ref->len,
 117                     BTRFS_TOTAL_BYTES_PINNED_BATCH);
 118 }
 119
 120 /* simple helper to search for an existing data extent at a given offset */
 121 int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
 122 {
 123         int ret;
 124         struct btrfs_key key;
 125         struct btrfs_path *path;
 126
 127         path = btrfs_alloc_path();
 128         if (!path)
 129                 return -ENOMEM;
 130
 131         key.objectid = start;
 132         key.offset = len;
 133         key.type = BTRFS_EXTENT_ITEM_KEY;
 134         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
 135         btrfs_free_path(path);
 136         return ret;
 137 }
 138
 139 /*
 140  * helper function to lookup reference count and flags of a tree block.
 141  *
 142  * the head node for delayed ref is used to store the sum of all the
 143  * reference count modifications queued up in the rbtree. the head
 144  * node may also store the extent flags to set. This way you can check
 145  * to see what the reference count and extent flags would be if all of
 146  * the delayed refs are not processed.
 147  */
 148 int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 149                              struct btrfs_fs_info *fs_info, u64 bytenr,
 150                              u64 offset, int metadata, u64 *refs, u64 *flags)
 151 {
 152         struct btrfs_delayed_ref_head *head;
 153         struct btrfs_delayed_ref_root *delayed_refs;
 154         struct btrfs_path *path;
 155         struct btrfs_extent_item *ei;
 156         struct extent_buffer *leaf;
 157         struct btrfs_key key;
 158         u32 item_size;
 159         u64 num_refs;
 160         u64 extent_flags;
 161         int ret;
 162
 163         /*
 164          * If we don't have skinny metadata, don't bother doing anything
 165          * different
 166          */
 167         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
 168                 offset = fs_info->nodesize;
 169                 metadata = 0;
 170         }
 171
 172         path = btrfs_alloc_path();
 173         if (!path)
 174                 return -ENOMEM;
 175
 176         if (!trans) {
 177                 path->skip_locking = 1;
 178                 path->search_commit_root = 1;
 179         }
 180
 181 search_again:
 182         key.objectid = bytenr;
 183         key.offset = offset;
 184         if (metadata)
 185                 key.type = BTRFS_METADATA_ITEM_KEY;
 186         else
 187                 key.type = BTRFS_EXTENT_ITEM_KEY;
 188
 189         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
 190         if (ret < 0)
 191                 goto out_free;
 192
 193         if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
 194                 if (path->slots[0]) {
 195                         path->slots[0]--;
 196                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 197                                               path->slots[0]);
 198                         if (key.objectid == bytenr &&
 199                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 200                             key.offset == fs_info->nodesize)
 201                                 ret = 0;
 202                 }
 203         }
 204
 205         if (ret == 0) {
 206                 leaf = path->nodes[0];
 207                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 208                 if (item_size >= sizeof(*ei)) {
 209                         ei = btrfs_item_ptr(leaf, path->slots[0],
 210                                             struct btrfs_extent_item);
 211                         num_refs = btrfs_extent_refs(leaf, ei);
 212                         extent_flags = btrfs_extent_flags(leaf, ei);
 213                 } else {
 214                         ret = -EINVAL;
 215                         btrfs_print_v0_err(fs_info);
 216                         if (trans)
 217                                 btrfs_abort_transaction(trans, ret);
 218                         else
 219                                 btrfs_handle_fs_error(fs_info, ret, NULL);
 220
 221                         goto out_free;
 222                 }
 223
 224                 BUG_ON(num_refs == 0);
 225         } else {
 226                 num_refs = 0;
 227                 extent_flags = 0;
 228                 ret = 0;
 229         }
 230
 231         if (!trans)
 232                 goto out;
 233
 234         delayed_refs = &trans->transaction->delayed_refs;
 235         spin_lock(&delayed_refs->lock);
 236         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
 237         if (head) {
 238                 if (!mutex_trylock(&head->mutex)) {
 239                         refcount_inc(&head->refs);
 240                         spin_unlock(&delayed_refs->lock);
 241
 242                         btrfs_release_path(path);
 243
 244                         /*
 245                          * Mutex was contended, block until it's released and try
 246                          * again
 247                          */
 248                         mutex_lock(&head->mutex);
 249                         mutex_unlock(&head->mutex);
 250                         btrfs_put_delayed_ref_head(head);
 251                         goto search_again;
 252                 }
 253                 spin_lock(&head->lock);
 254                 if (head->extent_op && head->extent_op->update_flags)
 255                         extent_flags |= head->extent_op->flags_to_set;
 256                 else
 257                         BUG_ON(num_refs == 0);
 258
 259                 num_refs += head->ref_mod;
 260                 spin_unlock(&head->lock);
 261                 mutex_unlock(&head->mutex);
 262         }
 263         spin_unlock(&delayed_refs->lock);
 264 out:
 265         WARN_ON(num_refs == 0);
 266         if (refs)
 267                 *refs = num_refs;
 268         if (flags)
 269                 *flags = extent_flags;
 270 out_free:
 271         btrfs_free_path(path);
 272         return ret;
 273 }
 274
 275 /*
 276  * Back reference rules.  Back refs have three main goals:
 277  *
 278  * 1) differentiate between all holders of references to an extent so that
 279  *    when a reference is dropped we can make sure it was a valid reference
 280  *    before freeing the extent.
 281  *
 282  * 2) Provide enough information to quickly find the holders of an extent
 283  *    if we notice a given block is corrupted or bad.
 284  *
 285  * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 286  *    maintenance.  This is actually the same as #2, but with a slightly
 287  *    different use case.
 288  *
 289  * There are two kinds of back refs. The implicit back refs is optimized
 290  * for pointers in non-shared tree blocks. For a given pointer in a block,
 291  * back refs of this kind provide information about the block's owner tree
 292  * and the pointer's key. These information allow us to find the block by
 293  * b-tree searching. The full back refs is for pointers in tree blocks not
 294  * referenced by their owner trees. The location of tree block is recorded
 295  * in the back refs. Actually the full back refs is generic, and can be
 296  * used in all cases the implicit back refs is used. The major shortcoming
 297  * of the full back refs is its overhead. Every time a tree block gets
 298  * COWed, we have to update back refs entry for all pointers in it.
 299  *
 300  * For a newly allocated tree block, we use implicit back refs for
 301  * pointers in it. This means most tree related operations only involve
 302  * implicit back refs. For a tree block created in old transaction, the
 303  * only way to drop a reference to it is COW it. So we can detect the
 304  * event that tree block loses its owner tree's reference and do the
 305  * back refs conversion.
 306  *
 307  * When a tree block is COWed through a tree, there are four cases:
 308  *
 309  * The reference count of the block is one and the tree is the block's
 310  * owner tree. Nothing to do in this case.
 311  *
 312  * The reference count of the block is one and the tree is not the
 313  * block's owner tree. In this case, full back refs is used for pointers
 314  * in the block. Remove these full back refs, add implicit back refs for
 315  * every pointers in the new block.
 316  *
 317  * The reference count of the block is greater than one and the tree is
 318  * the block's owner tree. In this case, implicit back refs is used for
 319  * pointers in the block. Add full back refs for every pointers in the
 320  * block, increase lower level extents' reference counts. The original
 321  * implicit back refs are entailed to the new block.
 322  *
 323  * The reference count of the block is greater than one and the tree is
 324  * not the block's owner tree. Add implicit back refs for every pointer in
 325  * the new block, increase lower level extents' reference count.
 326  *
 327  * Back Reference Key composing:
 328  *
 329  * The key objectid corresponds to the first byte in the extent,
 330  * The key type is used to differentiate between types of back refs.
 331  * There are different meanings of the key offset for different types
 332  * of back refs.
 333  *
 334  * File extents can be referenced by:
 335  *
 336  * - multiple snapshots, subvolumes, or different generations in one subvol
 337  * - different files inside a single subvolume
 338  * - different offsets inside a file (bookend extents in file.c)
 339  *
 340  * The extent ref structure for the implicit back refs has fields for:
 341  *
 342  * - Objectid of the subvolume root
 343  * - objectid of the file holding the reference
 344  * - original offset in the file
 345  * - how many bookend extents
 346  *
 347  * The key offset for the implicit back refs is hash of the first
 348  * three fields.
 349  *
 350  * The extent ref structure for the full back refs has field for:
 351  *
 352  * - number of pointers in the tree leaf
 353  *
 354  * The key offset for the implicit back refs is the first byte of
 355  * the tree leaf
 356  *
 357  * When a file extent is allocated, The implicit back refs is used.
 358  * the fields are filled in:
 359  *
 360  *     (root_key.objectid, inode objectid, offset in file, 1)
 361  *
 362  * When a file extent is removed file truncation, we find the
 363  * corresponding implicit back refs and check the following fields:
 364  *
 365  *     (btrfs_header_owner(leaf), inode objectid, offset in file)
 366  *
 367  * Btree extents can be referenced by:
 368  *
 369  * - Different subvolumes
 370  *
 371  * Both the implicit back refs and the full back refs for tree blocks
 372  * only consist of key. The key offset for the implicit back refs is
 373  * objectid of block's owner tree. The key offset for the full back refs
 374  * is the first byte of parent block.
 375  *
 376  * When implicit back refs is used, information about the lowest key and
 377  * level of the tree block are required. These information are stored in
 378  * tree block info structure.
 379  */
 380
 381 /*
 382  * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
 383  * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
 384  * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
 385  */
 386 int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
 387                                      struct btrfs_extent_inline_ref *iref,
 388                                      enum btrfs_inline_ref_type is_data)
 389 {
 390         int type = btrfs_extent_inline_ref_type(eb, iref);
 391         u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
 392
 393         if (type == BTRFS_TREE_BLOCK_REF_KEY ||
 394             type == BTRFS_SHARED_BLOCK_REF_KEY ||
 395             type == BTRFS_SHARED_DATA_REF_KEY ||
 396             type == BTRFS_EXTENT_DATA_REF_KEY) {
 397                 if (is_data == BTRFS_REF_TYPE_BLOCK) {
 398                         if (type == BTRFS_TREE_BLOCK_REF_KEY)
 399                                 return type;
 400                         if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
 401                                 ASSERT(eb->fs_info);
 402                                 /*
 403                                  * Every shared one has parent tree block,
 404                                  * which must be aligned to sector size.
 405                                  */
 406                                 if (offset &&
 407                                     IS_ALIGNED(offset, eb->fs_info->sectorsize))
 408                                         return type;
 409                         }
 410                 } else if (is_data == BTRFS_REF_TYPE_DATA) {
 411                         if (type == BTRFS_EXTENT_DATA_REF_KEY)
 412                                 return type;
 413                         if (type == BTRFS_SHARED_DATA_REF_KEY) {
 414                                 ASSERT(eb->fs_info);
 415                                 /*
 416                                  * Every shared one has parent tree block,
 417                                  * which must be aligned to sector size.
 418                                  */
 419                                 if (offset &&
 420                                     IS_ALIGNED(offset, eb->fs_info->sectorsize))
 421                                         return type;
 422                         }
 423                 } else {
 424                         ASSERT(is_data == BTRFS_REF_TYPE_ANY);
 425                         return type;
 426                 }
 427         }
 428
 429         btrfs_print_leaf((struct extent_buffer *)eb);
 430         btrfs_err(eb->fs_info,
 431                   "eb %llu iref 0x%lx invalid extent inline ref type %d",
 432                   eb->start, (unsigned long)iref, type);
 433         WARN_ON(1);
 434
 435         return BTRFS_REF_TYPE_INVALID;
 436 }
 437
 438 u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
 439 {
 440         u32 high_crc = ~(u32)0;
 441         u32 low_crc = ~(u32)0;
 442         __le64 lenum;
 443
 444         lenum = cpu_to_le64(root_objectid);
 445         high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
 446         lenum = cpu_to_le64(owner);
 447         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 448         lenum = cpu_to_le64(offset);
 449         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 450
 451         return ((u64)high_crc << 31) ^ (u64)low_crc;
 452 }
 453
 454 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
 455                                      struct btrfs_extent_data_ref *ref)
 456 {
 457         return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
 458                                     btrfs_extent_data_ref_objectid(leaf, ref),
 459                                     btrfs_extent_data_ref_offset(leaf, ref));
 460 }
 461
 462 static int match_extent_data_ref(struct extent_buffer *leaf,
 463                                  struct btrfs_extent_data_ref *ref,
 464                                  u64 root_objectid, u64 owner, u64 offset)
 465 {
 466         if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
 467             btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
 468             btrfs_extent_data_ref_offset(leaf, ref) != offset)
 469                 return 0;
 470         return 1;
 471 }
 472
 473 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
 474                                            struct btrfs_path *path,
 475                                            u64 bytenr, u64 parent,
 476                                            u64 root_objectid,
 477                                            u64 owner, u64 offset)
 478 {
 479         struct btrfs_root *root = trans->fs_info->extent_root;
 480         struct btrfs_key key;
 481         struct btrfs_extent_data_ref *ref;
 482         struct extent_buffer *leaf;
 483         u32 nritems;
 484         int ret;
 485         int recow;
 486         int err = -ENOENT;
 487
 488         key.objectid = bytenr;
 489         if (parent) {
 490                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 491                 key.offset = parent;
 492         } else {
 493                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 494                 key.offset = hash_extent_data_ref(root_objectid,
 495                                                   owner, offset);
 496         }
 497 again:
 498         recow = 0;
 499         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 500         if (ret < 0) {
 501                 err = ret;
 502                 goto fail;
 503         }
 504
 505         if (parent) {
 506                 if (!ret)
 507                         return 0;
 508                 goto fail;
 509         }
 510
 511         leaf = path->nodes[0];
 512         nritems = btrfs_header_nritems(leaf);
 513         while (1) {
 514                 if (path->slots[0] >= nritems) {
 515                         ret = btrfs_next_leaf(root, path);
 516                         if (ret < 0)
 517                                 err = ret;
 518                         if (ret)
 519                                 goto fail;
 520
 521                         leaf = path->nodes[0];
 522                         nritems = btrfs_header_nritems(leaf);
 523                         recow = 1;
 524                 }
 525
 526                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 527                 if (key.objectid != bytenr ||
 528                     key.type != BTRFS_EXTENT_DATA_REF_KEY)
 529                         goto fail;
 530
 531                 ref = btrfs_item_ptr(leaf, path->slots[0],
 532                                      struct btrfs_extent_data_ref);
 533
 534                 if (match_extent_data_ref(leaf, ref, root_objectid,
 535                                           owner, offset)) {
 536                         if (recow) {
 537                                 btrfs_release_path(path);
 538                                 goto again;
 539                         }
 540                         err = 0;
 541                         break;
 542                 }
 543                 path->slots[0]++;
 544         }
 545 fail:
 546         return err;
 547 }
 548
 549 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 550                                            struct btrfs_path *path,
 551                                            u64 bytenr, u64 parent,
 552                                            u64 root_objectid, u64 owner,
 553                                            u64 offset, int refs_to_add)
 554 {
 555         struct btrfs_root *root = trans->fs_info->extent_root;
 556         struct btrfs_key key;
 557         struct extent_buffer *leaf;
 558         u32 size;
 559         u32 num_refs;
 560         int ret;
 561
 562         key.objectid = bytenr;
 563         if (parent) {
 564                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 565                 key.offset = parent;
 566                 size = sizeof(struct btrfs_shared_data_ref);
 567         } else {
 568                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 569                 key.offset = hash_extent_data_ref(root_objectid,
 570                                                   owner, offset);
 571                 size = sizeof(struct btrfs_extent_data_ref);
 572         }
 573
 574         ret = btrfs_insert_empty_item(trans, root, path, &key, size);
 575         if (ret && ret != -EEXIST)
 576                 goto fail;
 577
 578         leaf = path->nodes[0];
 579         if (parent) {
 580                 struct btrfs_shared_data_ref *ref;
 581                 ref = btrfs_item_ptr(leaf, path->slots[0],
 582                                      struct btrfs_shared_data_ref);
 583                 if (ret == 0) {
 584                         btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
 585                 } else {
 586                         num_refs = btrfs_shared_data_ref_count(leaf, ref);
 587                         num_refs += refs_to_add;
 588                         btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
 589                 }
 590         } else {
 591                 struct btrfs_extent_data_ref *ref;
 592                 while (ret == -EEXIST) {
 593                         ref = btrfs_item_ptr(leaf, path->slots[0],
 594                                              struct btrfs_extent_data_ref);
 595                         if (match_extent_data_ref(leaf, ref, root_objectid,
 596                                                   owner, offset))
 597                                 break;
 598                         btrfs_release_path(path);
 599                         key.offset++;
 600                         ret = btrfs_insert_empty_item(trans, root, path, &key,
 601                                                       size);
 602                         if (ret && ret != -EEXIST)
 603                                 goto fail;
 604
 605                         leaf = path->nodes[0];
 606                 }
 607                 ref = btrfs_item_ptr(leaf, path->slots[0],
 608                                      struct btrfs_extent_data_ref);
 609                 if (ret == 0) {
 610                         btrfs_set_extent_data_ref_root(leaf, ref,
 611                                                        root_objectid);
 612                         btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
 613                         btrfs_set_extent_data_ref_offset(leaf, ref, offset);
 614                         btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
 615                 } else {
 616                         num_refs = btrfs_extent_data_ref_count(leaf, ref);
 617                         num_refs += refs_to_add;
 618                         btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
 619                 }
 620         }
 621         btrfs_mark_buffer_dirty(leaf);
 622         ret = 0;
 623 fail:
 624         btrfs_release_path(path);
 625         return ret;
 626 }
 627
 628 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 629                                            struct btrfs_path *path,
 630                                            int refs_to_drop, int *last_ref)
 631 {
 632         struct btrfs_key key;
 633         struct btrfs_extent_data_ref *ref1 = NULL;
 634         struct btrfs_shared_data_ref *ref2 = NULL;
 635         struct extent_buffer *leaf;
 636         u32 num_refs = 0;
 637         int ret = 0;
 638
 639         leaf = path->nodes[0];
 640         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 641
 642         if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 643                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 644                                       struct btrfs_extent_data_ref);
 645                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 646         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 647                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 648                                       struct btrfs_shared_data_ref);
 649                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 650         } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
 651                 btrfs_print_v0_err(trans->fs_info);
 652                 btrfs_abort_transaction(trans, -EINVAL);
 653                 return -EINVAL;
 654         } else {
 655                 BUG();
 656         }
 657
 658         BUG_ON(num_refs < refs_to_drop);
 659         num_refs -= refs_to_drop;
 660
 661         if (num_refs == 0) {
 662                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
 663                 *last_ref = 1;
 664         } else {
 665                 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
 666                         btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
 667                 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
 668                         btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
 669                 btrfs_mark_buffer_dirty(leaf);
 670         }
 671         return ret;
 672 }
 673
 674 static noinline u32 extent_data_ref_count(struct btrfs_path *path,
 675                                           struct btrfs_extent_inline_ref *iref)
 676 {
 677         struct btrfs_key key;
 678         struct extent_buffer *leaf;
 679         struct btrfs_extent_data_ref *ref1;
 680         struct btrfs_shared_data_ref *ref2;
 681         u32 num_refs = 0;
 682         int type;
 683
 684         leaf = path->nodes[0];
 685         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 686
 687         BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
 688         if (iref) {
 689                 /*
 690                  * If type is invalid, we should have bailed out earlier than
 691                  * this call.
 692                  */
 693                 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
 694                 ASSERT(type != BTRFS_REF_TYPE_INVALID);
 695                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 696                         ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
 697                         num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 698                 } else {
 699                         ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
 700                         num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 701                 }
 702         } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 703                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 704                                       struct btrfs_extent_data_ref);
 705                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 706         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 707                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 708                                       struct btrfs_shared_data_ref);
 709                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 710         } else {
 711                 WARN_ON(1);
 712         }
 713         return num_refs;
 714 }
 715
 716 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
 717                                           struct btrfs_path *path,
 718                                           u64 bytenr, u64 parent,
 719                                           u64 root_objectid)
 720 {
 721         struct btrfs_root *root = trans->fs_info->extent_root;
 722         struct btrfs_key key;
 723         int ret;
 724
 725         key.objectid = bytenr;
 726         if (parent) {
 727                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 728                 key.offset = parent;
 729         } else {
 730                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 731                 key.offset = root_objectid;
 732         }
 733
 734         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 735         if (ret > 0)
 736                 ret = -ENOENT;
 737         return ret;
 738 }
 739
 740 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
 741                                           struct btrfs_path *path,
 742                                           u64 bytenr, u64 parent,
 743                                           u64 root_objectid)
 744 {
 745         struct btrfs_key key;
 746         int ret;
 747
 748         key.objectid = bytenr;
 749         if (parent) {
 750                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 751                 key.offset = parent;
 752         } else {
 753                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 754                 key.offset = root_objectid;
 755         }
 756
 757         ret = btrfs_insert_empty_item(trans, trans->fs_info->extent_root,
 758                                       path, &key, 0);
 759         btrfs_release_path(path);
 760         return ret;
 761 }
 762
 763 static inline int extent_ref_type(u64 parent, u64 owner)
 764 {
 765         int type;
 766         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
 767                 if (parent > 0)
 768                         type = BTRFS_SHARED_BLOCK_REF_KEY;
 769                 else
 770                         type = BTRFS_TREE_BLOCK_REF_KEY;
 771         } else {
 772                 if (parent > 0)
 773                         type = BTRFS_SHARED_DATA_REF_KEY;
 774                 else
 775                         type = BTRFS_EXTENT_DATA_REF_KEY;
 776         }
 777         return type;
 778 }
 779
 780 static int find_next_key(struct btrfs_path *path, int level,
 781                          struct btrfs_key *key)
 782
 783 {
 784         for (; level < BTRFS_MAX_LEVEL; level++) {
 785                 if (!path->nodes[level])
 786                         break;
 787                 if (path->slots[level] + 1 >=
 788                     btrfs_header_nritems(path->nodes[level]))
 789                         continue;
 790                 if (level == 0)
 791                         btrfs_item_key_to_cpu(path->nodes[level], key,
 792                                               path->slots[level] + 1);
 793                 else
 794                         btrfs_node_key_to_cpu(path->nodes[level], key,
 795                                               path->slots[level] + 1);
 796                 return 0;
 797         }
 798         return 1;
 799 }
 800
 801 /*
 802  * look for inline back ref. if back ref is found, *ref_ret is set
 803  * to the address of inline back ref, and 0 is returned.
 804  *
 805  * if back ref isn't found, *ref_ret is set to the address where it
 806  * should be inserted, and -ENOENT is returned.
 807  *
 808  * if insert is true and there are too many inline back refs, the path
 809  * points to the extent item, and -EAGAIN is returned.
 810  *
 811  * NOTE: inline back refs are ordered in the same way that back ref
 812  *       items in the tree are ordered.
 813  */
 814 static noinline_for_stack
 815 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 816                                  struct btrfs_path *path,
 817                                  struct btrfs_extent_inline_ref **ref_ret,
 818                                  u64 bytenr, u64 num_bytes,
 819                                  u64 parent, u64 root_objectid,
 820                                  u64 owner, u64 offset, int insert)
 821 {
 822         struct btrfs_fs_info *fs_info = trans->fs_info;
 823         struct btrfs_root *root = fs_info->extent_root;
 824         struct btrfs_key key;
 825         struct extent_buffer *leaf;
 826         struct btrfs_extent_item *ei;
 827         struct btrfs_extent_inline_ref *iref;
 828         u64 flags;
 829         u64 item_size;
 830         unsigned long ptr;
 831         unsigned long end;
 832         int extra_size;
 833         int type;
 834         int want;
 835         int ret;
 836         int err = 0;
 837         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
 838         int needed;
 839
 840         key.objectid = bytenr;
 841         key.type = BTRFS_EXTENT_ITEM_KEY;
 842         key.offset = num_bytes;
 843
 844         want = extent_ref_type(parent, owner);
 845         if (insert) {
 846                 extra_size = btrfs_extent_inline_ref_size(want);
 847                 path->keep_locks = 1;
 848         } else
 849                 extra_size = -1;
 850
 851         /*
 852          * Owner is our level, so we can just add one to get the level for the
 853          * block we are interested in.
 854          */
 855         if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
 856                 key.type = BTRFS_METADATA_ITEM_KEY;
 857                 key.offset = owner;
 858         }
 859
 860 again:
 861         ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
 862         if (ret < 0) {
 863                 err = ret;
 864                 goto out;
 865         }
 866
 867         /*
 868          * We may be a newly converted file system which still has the old fat
 869          * extent entries for metadata, so try and see if we have one of those.
 870          */
 871         if (ret > 0 && skinny_metadata) {
 872                 skinny_metadata = false;
 873                 if (path->slots[0]) {
 874                         path->slots[0]--;
 875                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 876                                               path->slots[0]);
 877                         if (key.objectid == bytenr &&
 878                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 879                             key.offset == num_bytes)
 880                                 ret = 0;
 881                 }
 882                 if (ret) {
 883                         key.objectid = bytenr;
 884                         key.type = BTRFS_EXTENT_ITEM_KEY;
 885                         key.offset = num_bytes;
 886                         btrfs_release_path(path);
 887                         goto again;
 888                 }
 889         }
 890
 891         if (ret && !insert) {
 892                 err = -ENOENT;
 893                 goto out;
 894         } else if (WARN_ON(ret)) {
 895                 err = -EIO;
 896                 goto out;
 897         }
 898
 899         leaf = path->nodes[0];
 900         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 901         if (unlikely(item_size < sizeof(*ei))) {
 902                 err = -EINVAL;
 903                 btrfs_print_v0_err(fs_info);
 904                 btrfs_abort_transaction(trans, err);
 905                 goto out;
 906         }
 907
 908         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 909         flags = btrfs_extent_flags(leaf, ei);
 910
 911         ptr = (unsigned long)(ei + 1);
 912         end = (unsigned long)ei + item_size;
 913
 914         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
 915                 ptr += sizeof(struct btrfs_tree_block_info);
 916                 BUG_ON(ptr > end);
 917         }
 918
 919         if (owner >= BTRFS_FIRST_FREE_OBJECTID)
 920                 needed = BTRFS_REF_TYPE_DATA;
 921         else
 922                 needed = BTRFS_REF_TYPE_BLOCK;
 923
 924         err = -ENOENT;
 925         while (1) {
 926                 if (ptr >= end) {
 927                         WARN_ON(ptr > end);
 928                         break;
 929                 }
 930                 iref = (struct btrfs_extent_inline_ref *)ptr;
 931                 type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
 932                 if (type == BTRFS_REF_TYPE_INVALID) {
 933                         err = -EUCLEAN;
 934                         goto out;
 935                 }
 936
 937                 if (want < type)
 938                         break;
 939                 if (want > type) {
 940                         ptr += btrfs_extent_inline_ref_size(type);
 941                         continue;
 942                 }
 943
 944                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 945                         struct btrfs_extent_data_ref *dref;
 946                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
 947                         if (match_extent_data_ref(leaf, dref, root_objectid,
 948                                                   owner, offset)) {
 949                                 err = 0;
 950                                 break;
 951                         }
 952                         if (hash_extent_data_ref_item(leaf, dref) <
 953                             hash_extent_data_ref(root_objectid, owner, offset))
 954                                 break;
 955                 } else {
 956                         u64 ref_offset;
 957                         ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
 958                         if (parent > 0) {
 959                                 if (parent == ref_offset) {
 960                                         err = 0;
 961                                         break;
 962                                 }
 963                                 if (ref_offset < parent)
 964                                         break;
 965                         } else {
 966                                 if (root_objectid == ref_offset) {
 967                                         err = 0;
 968                                         break;
 969                                 }
 970                                 if (ref_offset < root_objectid)
 971                                         break;
 972                         }
 973                 }
 974                 ptr += btrfs_extent_inline_ref_size(type);
 975         }
 976         if (err == -ENOENT && insert) {
 977                 if (item_size + extra_size >=
 978                     BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
 979                         err = -EAGAIN;
 980                         goto out;
 981                 }
 982                 /*
 983                  * To add new inline back ref, we have to make sure
 984                  * there is no corresponding back ref item.
 985                  * For simplicity, we just do not add new inline back
 986                  * ref if there is any kind of item for this block
 987                  */
 988                 if (find_next_key(path, 0, &key) == 0 &&
 989                     key.objectid == bytenr &&
 990                     key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
 991                         err = -EAGAIN;
 992                         goto out;
 993                 }
 994         }
 995         *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
 996 out:
 997         if (insert) {
 998                 path->keep_locks = 0;
 999                 btrfs_unlock_up_safe(path, 1);
1000         }
1001         return err;
1002 }
1003
1004 /*
1005  * helper to add new inline back ref
1006  */
1007 static noinline_for_stack
1008 void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
1009                                  struct btrfs_path *path,
1010                                  struct btrfs_extent_inline_ref *iref,
1011                                  u64 parent, u64 root_objectid,
1012                                  u64 owner, u64 offset, int refs_to_add,
1013                                  struct btrfs_delayed_extent_op *extent_op)
1014 {
1015         struct extent_buffer *leaf;
1016         struct btrfs_extent_item *ei;
1017         unsigned long ptr;
1018         unsigned long end;
1019         unsigned long item_offset;
1020         u64 refs;
1021         int size;
1022         int type;
1023
1024         leaf = path->nodes[0];
1025         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1026         item_offset = (unsigned long)iref - (unsigned long)ei;
1027
1028         type = extent_ref_type(parent, owner);
1029         size = btrfs_extent_inline_ref_size(type);
1030
1031         btrfs_extend_item(path, size);
1032
1033         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1034         refs = btrfs_extent_refs(leaf, ei);
1035         refs += refs_to_add;
1036         btrfs_set_extent_refs(leaf, ei, refs);
1037         if (extent_op)
1038                 __run_delayed_extent_op(extent_op, leaf, ei);
1039
1040         ptr = (unsigned long)ei + item_offset;
1041         end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1042         if (ptr < end - size)
1043                 memmove_extent_buffer(leaf, ptr + size, ptr,
1044                                       end - size - ptr);
1045
1046         iref = (struct btrfs_extent_inline_ref *)ptr;
1047         btrfs_set_extent_inline_ref_type(leaf, iref, type);
1048         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1049                 struct btrfs_extent_data_ref *dref;
1050                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1051                 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1052                 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1053                 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1054                 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1055         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1056                 struct btrfs_shared_data_ref *sref;
1057                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1058                 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1059                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1060         } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1061                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1062         } else {
1063                 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1064         }
1065         btrfs_mark_buffer_dirty(leaf);
1066 }
1067
1068 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1069                                  struct btrfs_path *path,
1070                                  struct btrfs_extent_inline_ref **ref_ret,
1071                                  u64 bytenr, u64 num_bytes, u64 parent,
1072                                  u64 root_objectid, u64 owner, u64 offset)
1073 {
1074         int ret;
1075
1076         ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
1077                                            num_bytes, parent, root_objectid,
1078                                            owner, offset, 0);
1079         if (ret != -ENOENT)
1080                 return ret;
1081
1082         btrfs_release_path(path);
1083         *ref_ret = NULL;
1084
1085         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1086                 ret = lookup_tree_block_ref(trans, path, bytenr, parent,
1087                                             root_objectid);
1088         } else {
1089                 ret = lookup_extent_data_ref(trans, path, bytenr, parent,
1090                                              root_objectid, owner, offset);
1091         }
1092         return ret;
1093 }
1094
1095 /*
1096  * helper to update/remove inline back ref
1097  */
1098 static noinline_for_stack
1099 void update_inline_extent_backref(struct btrfs_path *path,
1100                                   struct btrfs_extent_inline_ref *iref,
1101                                   int refs_to_mod,
1102                                   struct btrfs_delayed_extent_op *extent_op,
1103                                   int *last_ref)
1104 {
1105         struct extent_buffer *leaf = path->nodes[0];
1106         struct btrfs_extent_item *ei;
1107         struct btrfs_extent_data_ref *dref = NULL;
1108         struct btrfs_shared_data_ref *sref = NULL;
1109         unsigned long ptr;
1110         unsigned long end;
1111         u32 item_size;
1112         int size;
1113         int type;
1114         u64 refs;
1115
1116         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1117         refs = btrfs_extent_refs(leaf, ei);
1118         WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1119         refs += refs_to_mod;
1120         btrfs_set_extent_refs(leaf, ei, refs);
1121         if (extent_op)
1122                 __run_delayed_extent_op(extent_op, leaf, ei);
1123
1124         /*
1125          * If type is invalid, we should have bailed out after
1126          * lookup_inline_extent_backref().
1127          */
1128         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
1129         ASSERT(type != BTRFS_REF_TYPE_INVALID);
1130
1131         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1132                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1133                 refs = btrfs_extent_data_ref_count(leaf, dref);
1134         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1135                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1136                 refs = btrfs_shared_data_ref_count(leaf, sref);
1137         } else {
1138                 refs = 1;
1139                 BUG_ON(refs_to_mod != -1);
1140         }
1141
1142         BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1143         refs += refs_to_mod;
1144
1145         if (refs > 0) {
1146                 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1147                         btrfs_set_extent_data_ref_count(leaf, dref, refs);
1148                 else
1149                         btrfs_set_shared_data_ref_count(leaf, sref, refs);
1150         } else {
1151                 *last_ref = 1;
1152                 size =  btrfs_extent_inline_ref_size(type);
1153                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1154                 ptr = (unsigned long)iref;
1155                 end = (unsigned long)ei + item_size;
1156                 if (ptr + size < end)
1157                         memmove_extent_buffer(leaf, ptr, ptr + size,
1158                                               end - ptr - size);
1159                 item_size -= size;
1160                 btrfs_truncate_item(path, item_size, 1);
1161         }
1162         btrfs_mark_buffer_dirty(leaf);
1163 }
1164
1165 static noinline_for_stack
1166 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1167                                  struct btrfs_path *path,
1168                                  u64 bytenr, u64 num_bytes, u64 parent,
1169                                  u64 root_objectid, u64 owner,
1170                                  u64 offset, int refs_to_add,
1171                                  struct btrfs_delayed_extent_op *extent_op)
1172 {
1173         struct btrfs_extent_inline_ref *iref;
1174         int ret;
1175
1176         ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
1177                                            num_bytes, parent, root_objectid,
1178                                            owner, offset, 1);
1179         if (ret == 0) {
1180                 /*
1181                  * We're adding refs to a tree block we already own, this
1182                  * should not happen at all.
1183                  */
1184                 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1185                         btrfs_crit(trans->fs_info,
1186 "adding refs to an existing tree ref, bytenr %llu num_bytes %llu root_objectid %llu",
1187                                    bytenr, num_bytes, root_objectid);
1188                         if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) {
1189                                 WARN_ON(1);
1190                                 btrfs_crit(trans->fs_info,
1191                         "path->slots[0]=%d path->nodes[0]:", path->slots[0]);
1192                                 btrfs_print_leaf(path->nodes[0]);
1193                         }
1194                         return -EUCLEAN;
1195                 }
1196                 update_inline_extent_backref(path, iref, refs_to_add,
1197                                              extent_op, NULL);
1198         } else if (ret == -ENOENT) {
1199                 setup_inline_extent_backref(trans->fs_info, path, iref, parent,
1200                                             root_objectid, owner, offset,
1201                                             refs_to_add, extent_op);
1202                 ret = 0;
1203         }
1204         return ret;
1205 }
1206
1207 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1208                                  struct btrfs_path *path,
1209                                  struct btrfs_extent_inline_ref *iref,
1210                                  int refs_to_drop, int is_data, int *last_ref)
1211 {
1212         int ret = 0;
1213
1214         BUG_ON(!is_data && refs_to_drop != 1);
1215         if (iref) {
1216                 update_inline_extent_backref(path, iref, -refs_to_drop, NULL,
1217                                              last_ref);
1218         } else if (is_data) {
1219                 ret = remove_extent_data_ref(trans, path, refs_to_drop,
1220                                              last_ref);
1221         } else {
1222                 *last_ref = 1;
1223                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
1224         }
1225         return ret;
1226 }
1227
1228 static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
1229                                u64 *discarded_bytes)
1230 {
1231         int j, ret = 0;
1232         u64 bytes_left, end;
1233         u64 aligned_start = ALIGN(start, 1 << 9);
1234
1235         if (WARN_ON(start != aligned_start)) {
1236                 len -= aligned_start - start;
1237                 len = round_down(len, 1 << 9);
1238                 start = aligned_start;
1239         }
1240
1241         *discarded_bytes = 0;
1242
1243         if (!len)
1244                 return 0;
1245
1246         end = start + len;
1247         bytes_left = len;
1248
1249         /* Skip any superblocks on this device. */
1250         for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
1251                 u64 sb_start = btrfs_sb_offset(j);
1252                 u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
1253                 u64 size = sb_start - start;
1254
1255                 if (!in_range(sb_start, start, bytes_left) &&
1256                     !in_range(sb_end, start, bytes_left) &&
1257                     !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
1258                         continue;
1259
1260                 /*
1261                  * Superblock spans beginning of range.  Adjust start and
1262                  * try again.
1263                  */
1264                 if (sb_start <= start) {
1265                         start += sb_end - start;
1266                         if (start > end) {
1267                                 bytes_left = 0;
1268                                 break;
1269                         }
1270                         bytes_left = end - start;
1271                         continue;
1272                 }
1273
1274                 if (size) {
1275                         ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
1276                                                    GFP_NOFS, 0);
1277                         if (!ret)
1278                                 *discarded_bytes += size;
1279                         else if (ret != -EOPNOTSUPP)
1280                                 return ret;
1281                 }
1282
1283                 start = sb_end;
1284                 if (start > end) {
1285                         bytes_left = 0;
1286                         break;
1287                 }
1288                 bytes_left = end - start;
1289         }
1290
1291         if (bytes_left) {
1292                 ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
1293                                            GFP_NOFS, 0);
1294                 if (!ret)
1295                         *discarded_bytes += bytes_left;
1296         }
1297         return ret;
1298 }
1299
1300 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
1301                          u64 num_bytes, u64 *actual_bytes)
1302 {
1303         int ret = 0;
1304         u64 discarded_bytes = 0;
1305         u64 end = bytenr + num_bytes;
1306         u64 cur = bytenr;
1307         struct btrfs_bio *bbio = NULL;
1308
1309
1310         /*
1311          * Avoid races with device replace and make sure our bbio has devices
1312          * associated to its stripes that don't go away while we are discarding.
1313          */
1314         btrfs_bio_counter_inc_blocked(fs_info);
1315         while (cur < end) {
1316                 struct btrfs_bio_stripe *stripe;
1317                 int i;
1318
1319                 num_bytes = end - cur;
1320                 /* Tell the block device(s) that the sectors can be discarded */
1321                 ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, cur,
1322                                       &num_bytes, &bbio, 0);
1323                 /*
1324                  * Error can be -ENOMEM, -ENOENT (no such chunk mapping) or
1325                  * -EOPNOTSUPP. For any such error, @num_bytes is not updated,
1326                  * thus we can't continue anyway.
1327                  */
1328                 if (ret < 0)
1329                         goto out;
1330
1331                 stripe = bbio->stripes;
1332                 for (i = 0; i < bbio->num_stripes; i++, stripe++) {
1333                         u64 bytes;
1334                         struct request_queue *req_q;
1335
1336                         if (!stripe->dev->bdev) {
1337                                 ASSERT(btrfs_test_opt(fs_info, DEGRADED));
1338                                 continue;
1339                         }
1340                         req_q = bdev_get_queue(stripe->dev->bdev);
1341                         if (!blk_queue_discard(req_q))
1342                                 continue;
1343
1344                         ret = btrfs_issue_discard(stripe->dev->bdev,
1345                                                   stripe->physical,
1346                                                   stripe->length,
1347                                                   &bytes);
1348                         if (!ret) {
1349                                 discarded_bytes += bytes;
1350                         } else if (ret != -EOPNOTSUPP) {
1351                                 /*
1352                                  * Logic errors or -ENOMEM, or -EIO, but
1353                                  * unlikely to happen.
1354                                  *
1355                                  * And since there are two loops, explicitly
1356                                  * go to out to avoid confusion.
1357                                  */
1358                                 btrfs_put_bbio(bbio);
1359                                 goto out;
1360                         }
1361
1362                         /*
1363                          * Just in case we get back EOPNOTSUPP for some reason,
1364                          * just ignore the return value so we don't screw up
1365                          * people calling discard_extent.
1366                          */
1367                         ret = 0;
1368                 }
1369                 btrfs_put_bbio(bbio);
1370                 cur += num_bytes;
1371         }
1372 out:
1373         btrfs_bio_counter_dec(fs_info);
1374
1375         if (actual_bytes)
1376                 *actual_bytes = discarded_bytes;
1377
1378
1379         if (ret == -EOPNOTSUPP)
1380                 ret = 0;
1381         return ret;
1382 }
1383
1384 /* Can return -ENOMEM */
1385 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1386                          struct btrfs_ref *generic_ref)
1387 {
1388         struct btrfs_fs_info *fs_info = trans->fs_info;
1389         int old_ref_mod, new_ref_mod;
1390         int ret;
1391
1392         ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
1393                generic_ref->action);
1394         BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
1395                generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
1396
1397         if (generic_ref->type == BTRFS_REF_METADATA)
1398                 ret = btrfs_add_delayed_tree_ref(trans, generic_ref,
1399                                 NULL, &old_ref_mod, &new_ref_mod);
1400         else
1401                 ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0,
1402                                                  &old_ref_mod, &new_ref_mod);
1403
1404         btrfs_ref_tree_mod(fs_info, generic_ref);
1405
1406         if (ret == 0 && old_ref_mod < 0 && new_ref_mod >= 0)
1407                 sub_pinned_bytes(fs_info, generic_ref);
1408
1409         return ret;
1410 }
1411
1412 /*
1413  * __btrfs_inc_extent_ref - insert backreference for a given extent
1414  *
1415  * The counterpart is in __btrfs_free_extent(), with examples and more details
1416  * how it works.
1417  *
1418  * @trans:          Handle of transaction
1419  *
1420  * @node:           The delayed ref node used to get the bytenr/length for
1421  *                  extent whose references are incremented.
1422  *
1423  * @parent:         If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/
1424  *                  BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical
1425  *                  bytenr of the parent block. Since new extents are always
1426  *                  created with indirect references, this will only be the case
1427  *                  when relocating a shared extent. In that case, root_objectid
1428  *                  will be BTRFS_TREE_RELOC_OBJECTID. Otheriwse, parent must
1429  *                  be 0
1430  *
1431  * @root_objectid:  The id of the root where this modification has originated,
1432  *                  this can be either one of the well-known metadata trees or
1433  *                  the subvolume id which references this extent.
1434  *
1435  * @owner:          For data extents it is the inode number of the owning file.
1436  *                  For metadata extents this parameter holds the level in the
1437  *                  tree of the extent.
1438  *
1439  * @offset:         For metadata extents the offset is ignored and is currently
1440  *                  always passed as 0. For data extents it is the fileoffset
1441  *                  this extent belongs to.
1442  *
1443  * @refs_to_add     Number of references to add
1444  *
1445  * @extent_op       Pointer to a structure, holding information necessary when
1446  *                  updating a tree block's flags
1447  *
1448  */
1449 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1450                                   struct btrfs_delayed_ref_node *node,
1451                                   u64 parent, u64 root_objectid,
1452                                   u64 owner, u64 offset, int refs_to_add,
1453                                   struct btrfs_delayed_extent_op *extent_op)
1454 {
1455         struct btrfs_path *path;
1456         struct extent_buffer *leaf;
1457         struct btrfs_extent_item *item;
1458         struct btrfs_key key;
1459         u64 bytenr = node->bytenr;
1460         u64 num_bytes = node->num_bytes;
1461         u64 refs;
1462         int ret;
1463
1464         path = btrfs_alloc_path();
1465         if (!path)
1466                 return -ENOMEM;
1467
1468         path->leave_spinning = 1;
1469         /* this will setup the path even if it fails to insert the back ref */
1470         ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
1471                                            parent, root_objectid, owner,
1472                                            offset, refs_to_add, extent_op);
1473         if ((ret < 0 && ret != -EAGAIN) || !ret)
1474                 goto out;
1475
1476         /*
1477          * Ok we had -EAGAIN which means we didn't have space to insert and
1478          * inline extent ref, so just update the reference count and add a
1479          * normal backref.
1480          */
1481         leaf = path->nodes[0];
1482         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1483         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1484         refs = btrfs_extent_refs(leaf, item);
1485         btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1486         if (extent_op)
1487                 __run_delayed_extent_op(extent_op, leaf, item);
1488
1489         btrfs_mark_buffer_dirty(leaf);
1490         btrfs_release_path(path);
1491
1492         path->leave_spinning = 1;
1493         /* now insert the actual backref */
1494         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1495                 BUG_ON(refs_to_add != 1);
1496                 ret = insert_tree_block_ref(trans, path, bytenr, parent,
1497                                             root_objectid);
1498         } else {
1499                 ret = insert_extent_data_ref(trans, path, bytenr, parent,
1500                                              root_objectid, owner, offset,
1501                                              refs_to_add);
1502         }
1503         if (ret)
1504                 btrfs_abort_transaction(trans, ret);
1505 out:
1506         btrfs_free_path(path);
1507         return ret;
1508 }
1509
1510 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1511                                 struct btrfs_delayed_ref_node *node,
1512                                 struct btrfs_delayed_extent_op *extent_op,
1513                                 int insert_reserved)
1514 {
1515         int ret = 0;
1516         struct btrfs_delayed_data_ref *ref;
1517         struct btrfs_key ins;
1518         u64 parent = 0;
1519         u64 ref_root = 0;
1520         u64 flags = 0;
1521
1522         ins.objectid = node->bytenr;
1523         ins.offset = node->num_bytes;
1524         ins.type = BTRFS_EXTENT_ITEM_KEY;
1525
1526         ref = btrfs_delayed_node_to_data_ref(node);
1527         trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action);
1528
1529         if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1530                 parent = ref->parent;
1531         ref_root = ref->root;
1532
1533         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1534                 if (extent_op)
1535                         flags |= extent_op->flags_to_set;
1536                 ret = alloc_reserved_file_extent(trans, parent, ref_root,
1537                                                  flags, ref->objectid,
1538                                                  ref->offset, &ins,
1539                                                  node->ref_mod);
1540         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1541                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1542                                              ref->objectid, ref->offset,
1543                                              node->ref_mod, extent_op);
1544         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1545                 ret = __btrfs_free_extent(trans, node, parent,
1546                                           ref_root, ref->objectid,
1547                                           ref->offset, node->ref_mod,
1548                                           extent_op);
1549         } else {
1550                 BUG();
1551         }
1552         return ret;
1553 }
1554
1555 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1556                                     struct extent_buffer *leaf,
1557                                     struct btrfs_extent_item *ei)
1558 {
1559         u64 flags = btrfs_extent_flags(leaf, ei);
1560         if (extent_op->update_flags) {
1561                 flags |= extent_op->flags_to_set;
1562                 btrfs_set_extent_flags(leaf, ei, flags);
1563         }
1564
1565         if (extent_op->update_key) {
1566                 struct btrfs_tree_block_info *bi;
1567                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1568                 bi = (struct btrfs_tree_block_info *)(ei + 1);
1569                 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1570         }
1571 }
1572
1573 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1574                                  struct btrfs_delayed_ref_head *head,
1575                                  struct btrfs_delayed_extent_op *extent_op)
1576 {
1577         struct btrfs_fs_info *fs_info = trans->fs_info;
1578         struct btrfs_key key;
1579         struct btrfs_path *path;
1580         struct btrfs_extent_item *ei;
1581         struct extent_buffer *leaf;
1582         u32 item_size;
1583         int ret;
1584         int err = 0;
1585         int metadata = !extent_op->is_data;
1586
1587         if (TRANS_ABORTED(trans))
1588                 return 0;
1589
1590         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1591                 metadata = 0;
1592
1593         path = btrfs_alloc_path();
1594         if (!path)
1595                 return -ENOMEM;
1596
1597         key.objectid = head->bytenr;
1598
1599         if (metadata) {
1600                 key.type = BTRFS_METADATA_ITEM_KEY;
1601                 key.offset = extent_op->level;
1602         } else {
1603                 key.type = BTRFS_EXTENT_ITEM_KEY;
1604                 key.offset = head->num_bytes;
1605         }
1606
1607 again:
1608         path->leave_spinning = 1;
1609         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
1610         if (ret < 0) {
1611                 err = ret;
1612                 goto out;
1613         }
1614         if (ret > 0) {
1615                 if (metadata) {
1616                         if (path->slots[0] > 0) {
1617                                 path->slots[0]--;
1618                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
1619                                                       path->slots[0]);
1620                                 if (key.objectid == head->bytenr &&
1621                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
1622                                     key.offset == head->num_bytes)
1623                                         ret = 0;
1624                         }
1625                         if (ret > 0) {
1626                                 btrfs_release_path(path);
1627                                 metadata = 0;
1628
1629                                 key.objectid = head->bytenr;
1630                                 key.offset = head->num_bytes;
1631                                 key.type = BTRFS_EXTENT_ITEM_KEY;
1632                                 goto again;
1633                         }
1634                 } else {
1635                         err = -EIO;
1636                         goto out;
1637                 }
1638         }
1639
1640         leaf = path->nodes[0];
1641         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1642
1643         if (unlikely(item_size < sizeof(*ei))) {
1644                 err = -EINVAL;
1645                 btrfs_print_v0_err(fs_info);
1646                 btrfs_abort_transaction(trans, err);
1647                 goto out;
1648         }
1649
1650         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1651         __run_delayed_extent_op(extent_op, leaf, ei);
1652
1653         btrfs_mark_buffer_dirty(leaf);
1654 out:
1655         btrfs_free_path(path);
1656         return err;
1657 }
1658
1659 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1660                                 struct btrfs_delayed_ref_node *node,
1661                                 struct btrfs_delayed_extent_op *extent_op,
1662                                 int insert_reserved)
1663 {
1664         int ret = 0;
1665         struct btrfs_delayed_tree_ref *ref;
1666         u64 parent = 0;
1667         u64 ref_root = 0;
1668
1669         ref = btrfs_delayed_node_to_tree_ref(node);
1670         trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action);
1671
1672         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1673                 parent = ref->parent;
1674         ref_root = ref->root;
1675
1676         if (node->ref_mod != 1) {
1677                 btrfs_err(trans->fs_info,
1678         "btree block(%llu) has %d references rather than 1: action %d ref_root %llu parent %llu",
1679                           node->bytenr, node->ref_mod, node->action, ref_root,
1680                           parent);
1681                 return -EIO;
1682         }
1683         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1684                 BUG_ON(!extent_op || !extent_op->update_flags);
1685                 ret = alloc_reserved_tree_block(trans, node, extent_op);
1686         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1687                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1688                                              ref->level, 0, 1, extent_op);
1689         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1690                 ret = __btrfs_free_extent(trans, node, parent, ref_root,
1691                                           ref->level, 0, 1, extent_op);
1692         } else {
1693                 BUG();
1694         }
1695         return ret;
1696 }
1697
1698 /* helper function to actually process a single delayed ref entry */
1699 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1700                                struct btrfs_delayed_ref_node *node,
1701                                struct btrfs_delayed_extent_op *extent_op,
1702                                int insert_reserved)
1703 {
1704         int ret = 0;
1705
1706         if (TRANS_ABORTED(trans)) {
1707                 if (insert_reserved)
1708                         btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1709                 return 0;
1710         }
1711
1712         if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1713             node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1714                 ret = run_delayed_tree_ref(trans, node, extent_op,
1715                                            insert_reserved);
1716         else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1717                  node->type == BTRFS_SHARED_DATA_REF_KEY)
1718                 ret = run_delayed_data_ref(trans, node, extent_op,
1719                                            insert_reserved);
1720         else
1721                 BUG();
1722         if (ret && insert_reserved)
1723                 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1724         return ret;
1725 }
1726
1727 static inline struct btrfs_delayed_ref_node *
1728 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1729 {
1730         struct btrfs_delayed_ref_node *ref;
1731
1732         if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
1733                 return NULL;
1734
1735         /*
1736          * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
1737          * This is to prevent a ref count from going down to zero, which deletes
1738          * the extent item from the extent tree, when there still are references
1739          * to add, which would fail because they would not find the extent item.
1740          */
1741         if (!list_empty(&head->ref_add_list))
1742                 return list_first_entry(&head->ref_add_list,
1743                                 struct btrfs_delayed_ref_node, add_list);
1744
1745         ref = rb_entry(rb_first_cached(&head->ref_tree),
1746                        struct btrfs_delayed_ref_node, ref_node);
1747         ASSERT(list_empty(&ref->add_list));
1748         return ref;
1749 }
1750
1751 static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
1752                                       struct btrfs_delayed_ref_head *head)
1753 {
1754         spin_lock(&delayed_refs->lock);
1755         head->processing = 0;
1756         delayed_refs->num_heads_ready++;
1757         spin_unlock(&delayed_refs->lock);
1758         btrfs_delayed_ref_unlock(head);
1759 }
1760
1761 static struct btrfs_delayed_extent_op *cleanup_extent_op(
1762                                 struct btrfs_delayed_ref_head *head)
1763 {
1764         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
1765
1766         if (!extent_op)
1767                 return NULL;
1768
1769         if (head->must_insert_reserved) {
1770                 head->extent_op = NULL;
1771                 btrfs_free_delayed_extent_op(extent_op);
1772                 return NULL;
1773         }
1774         return extent_op;
1775 }
1776
1777 static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
1778                                      struct btrfs_delayed_ref_head *head)
1779 {
1780         struct btrfs_delayed_extent_op *extent_op;
1781         int ret;
1782
1783         extent_op = cleanup_extent_op(head);
1784         if (!extent_op)
1785                 return 0;
1786         head->extent_op = NULL;
1787         spin_unlock(&head->lock);
1788         ret = run_delayed_extent_op(trans, head, extent_op);
1789         btrfs_free_delayed_extent_op(extent_op);
1790         return ret ? ret : 1;
1791 }
1792
1793 void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
1794                                   struct btrfs_delayed_ref_root *delayed_refs,
1795                                   struct btrfs_delayed_ref_head *head)
1796 {
1797         int nr_items = 1;       /* Dropping this ref head update. */
1798
1799         if (head->total_ref_mod < 0) {
1800                 struct btrfs_space_info *space_info;
1801                 u64 flags;
1802
1803                 if (head->is_data)
1804                         flags = BTRFS_BLOCK_GROUP_DATA;
1805                 else if (head->is_system)
1806                         flags = BTRFS_BLOCK_GROUP_SYSTEM;
1807                 else
1808                         flags = BTRFS_BLOCK_GROUP_METADATA;
1809                 space_info = btrfs_find_space_info(fs_info, flags);
1810                 ASSERT(space_info);
1811                 percpu_counter_add_batch(&space_info->total_bytes_pinned,
1812                                    -head->num_bytes,
1813                                    BTRFS_TOTAL_BYTES_PINNED_BATCH);
1814
1815                 /*
1816                  * We had csum deletions accounted for in our delayed refs rsv,
1817                  * we need to drop the csum leaves for this update from our
1818                  * delayed_refs_rsv.
1819                  */
1820                 if (head->is_data) {
1821                         spin_lock(&delayed_refs->lock);
1822                         delayed_refs->pending_csums -= head->num_bytes;
1823                         spin_unlock(&delayed_refs->lock);
1824                         nr_items += btrfs_csum_bytes_to_leaves(fs_info,
1825                                 head->num_bytes);
1826                 }
1827         }
1828
1829         btrfs_delayed_refs_rsv_release(fs_info, nr_items);
1830 }
1831
1832 static int cleanup_ref_head(struct btrfs_trans_handle *trans,
1833                             struct btrfs_delayed_ref_head *head)
1834 {
1835
1836         struct btrfs_fs_info *fs_info = trans->fs_info;
1837         struct btrfs_delayed_ref_root *delayed_refs;
1838         int ret;
1839
1840         delayed_refs = &trans->transaction->delayed_refs;
1841
1842         ret = run_and_cleanup_extent_op(trans, head);
1843         if (ret < 0) {
1844                 unselect_delayed_ref_head(delayed_refs, head);
1845                 btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
1846                 return ret;
1847         } else if (ret) {
1848                 return ret;
1849         }
1850
1851         /*
1852          * Need to drop our head ref lock and re-acquire the delayed ref lock
1853          * and then re-check to make sure nobody got added.
1854          */
1855         spin_unlock(&head->lock);
1856         spin_lock(&delayed_refs->lock);
1857         spin_lock(&head->lock);
1858         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
1859                 spin_unlock(&head->lock);
1860                 spin_unlock(&delayed_refs->lock);
1861                 return 1;
1862         }
1863         btrfs_delete_ref_head(delayed_refs, head);
1864         spin_unlock(&head->lock);
1865         spin_unlock(&delayed_refs->lock);
1866
1867         if (head->must_insert_reserved) {
1868                 btrfs_pin_extent(trans, head->bytenr, head->num_bytes, 1);
1869                 if (head->is_data) {
1870                         ret = btrfs_del_csums(trans, fs_info->csum_root,
1871                                               head->bytenr, head->num_bytes);
1872                 }
1873         }
1874
1875         btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
1876
1877         trace_run_delayed_ref_head(fs_info, head, 0);
1878         btrfs_delayed_ref_unlock(head);
1879         btrfs_put_delayed_ref_head(head);
1880         return 0;
1881 }
1882
1883 static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
1884                                         struct btrfs_trans_handle *trans)
1885 {
1886         struct btrfs_delayed_ref_root *delayed_refs =
1887                 &trans->transaction->delayed_refs;
1888         struct btrfs_delayed_ref_head *head = NULL;
1889         int ret;
1890
1891         spin_lock(&delayed_refs->lock);
1892         head = btrfs_select_ref_head(delayed_refs);
1893         if (!head) {
1894                 spin_unlock(&delayed_refs->lock);
1895                 return head;
1896         }
1897
1898         /*
1899          * Grab the lock that says we are going to process all the refs for
1900          * this head
1901          */
1902         ret = btrfs_delayed_ref_lock(delayed_refs, head);
1903         spin_unlock(&delayed_refs->lock);
1904
1905         /*
1906          * We may have dropped the spin lock to get the head mutex lock, and
1907          * that might have given someone else time to free the head.  If that's
1908          * true, it has been removed from our list and we can move on.
1909          */
1910         if (ret == -EAGAIN)
1911                 head = ERR_PTR(-EAGAIN);
1912
1913         return head;
1914 }
1915
1916 static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
1917                                     struct btrfs_delayed_ref_head *locked_ref,
1918                                     unsigned long *run_refs)
1919 {
1920         struct btrfs_fs_info *fs_info = trans->fs_info;
1921         struct btrfs_delayed_ref_root *delayed_refs;
1922         struct btrfs_delayed_extent_op *extent_op;
1923         struct btrfs_delayed_ref_node *ref;
1924         int must_insert_reserved = 0;
1925         int ret;
1926
1927         delayed_refs = &trans->transaction->delayed_refs;
1928
1929         lockdep_assert_held(&locked_ref->mutex);
1930         lockdep_assert_held(&locked_ref->lock);
1931
1932         while ((ref = select_delayed_ref(locked_ref))) {
1933                 if (ref->seq &&
1934                     btrfs_check_delayed_seq(fs_info, ref->seq)) {
1935                         spin_unlock(&locked_ref->lock);
1936                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1937                         return -EAGAIN;
1938                 }
1939
1940                 (*run_refs)++;
1941                 ref->in_tree = 0;
1942                 rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
1943                 RB_CLEAR_NODE(&ref->ref_node);
1944                 if (!list_empty(&ref->add_list))
1945                         list_del(&ref->add_list);
1946                 /*
1947                  * When we play the delayed ref, also correct the ref_mod on
1948                  * head
1949                  */
1950                 switch (ref->action) {
1951                 case BTRFS_ADD_DELAYED_REF:
1952                 case BTRFS_ADD_DELAYED_EXTENT:
1953                         locked_ref->ref_mod -= ref->ref_mod;
1954                         break;
1955                 case BTRFS_DROP_DELAYED_REF:
1956                         locked_ref->ref_mod += ref->ref_mod;
1957                         break;
1958                 default:
1959                         WARN_ON(1);
1960                 }
1961                 atomic_dec(&delayed_refs->num_entries);
1962
1963                 /*
1964                  * Record the must_insert_reserved flag before we drop the
1965                  * spin lock.
1966                  */
1967                 must_insert_reserved = locked_ref->must_insert_reserved;
1968                 locked_ref->must_insert_reserved = 0;
1969
1970                 extent_op = locked_ref->extent_op;
1971                 locked_ref->extent_op = NULL;
1972                 spin_unlock(&locked_ref->lock);
1973
1974                 ret = run_one_delayed_ref(trans, ref, extent_op,
1975                                           must_insert_reserved);
1976
1977                 btrfs_free_delayed_extent_op(extent_op);
1978                 if (ret) {
1979                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1980                         btrfs_put_delayed_ref(ref);
1981                         btrfs_debug(fs_info, "run_one_delayed_ref returned %d",
1982                                     ret);
1983                         return ret;
1984                 }
1985
1986                 btrfs_put_delayed_ref(ref);
1987                 cond_resched();
1988
1989                 spin_lock(&locked_ref->lock);
1990                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
1991         }
1992
1993         return 0;
1994 }
1995
1996 /*
1997  * Returns 0 on success or if called with an already aborted transaction.
1998  * Returns -ENOMEM or -EIO on failure and will abort the transaction.
1999  */
2000 static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2001                                              unsigned long nr)
2002 {
2003         struct btrfs_fs_info *fs_info = trans->fs_info;
2004         struct btrfs_delayed_ref_root *delayed_refs;
2005         struct btrfs_delayed_ref_head *locked_ref = NULL;
2006         ktime_t start = ktime_get();
2007         int ret;
2008         unsigned long count = 0;
2009         unsigned long actual_count = 0;
2010
2011         delayed_refs = &trans->transaction->delayed_refs;
2012         do {
2013                 if (!locked_ref) {
2014                         locked_ref = btrfs_obtain_ref_head(trans);
2015                         if (IS_ERR_OR_NULL(locked_ref)) {
2016                                 if (PTR_ERR(locked_ref) == -EAGAIN) {
2017                                         continue;
2018                                 } else {
2019                                         break;
2020                                 }
2021                         }
2022                         count++;
2023                 }
2024                 /*
2025                  * We need to try and merge add/drops of the same ref since we
2026                  * can run into issues with relocate dropping the implicit ref
2027                  * and then it being added back again before the drop can
2028                  * finish.  If we merged anything we need to re-loop so we can
2029                  * get a good ref.
2030                  * Or we can get node references of the same type that weren't
2031                  * merged when created due to bumps in the tree mod seq, and
2032                  * we need to merge them to prevent adding an inline extent
2033                  * backref before dropping it (triggering a BUG_ON at
2034                  * insert_inline_extent_backref()).
2035                  */
2036                 spin_lock(&locked_ref->lock);
2037                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
2038
2039                 ret = btrfs_run_delayed_refs_for_head(trans, locked_ref,
2040                                                       &actual_count);
2041                 if (ret < 0 && ret != -EAGAIN) {
2042                         /*
2043                          * Error, btrfs_run_delayed_refs_for_head already
2044                          * unlocked everything so just bail out
2045                          */
2046                         return ret;
2047                 } else if (!ret) {
2048                         /*
2049                          * Success, perform the usual cleanup of a processed
2050                          * head
2051                          */
2052                         ret = cleanup_ref_head(trans, locked_ref);
2053                         if (ret > 0 ) {
2054                                 /* We dropped our lock, we need to loop. */
2055                                 ret = 0;
2056                                 continue;
2057                         } else if (ret) {
2058                                 return ret;
2059                         }
2060                 }
2061
2062                 /*
2063                  * Either success case or btrfs_run_delayed_refs_for_head
2064                  * returned -EAGAIN, meaning we need to select another head
2065                  */
2066
2067                 locked_ref = NULL;
2068                 cond_resched();
2069         } while ((nr != -1 && count < nr) || locked_ref);
2070
2071         /*
2072          * We don't want to include ref heads since we can have empty ref heads
2073          * and those will drastically skew our runtime down since we just do
2074          * accounting, no actual extent tree updates.
2075          */
2076         if (actual_count > 0) {
2077                 u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
2078                 u64 avg;
2079
2080                 /*
2081                  * We weigh the current average higher than our current runtime
2082                  * to avoid large swings in the average.
2083                  */
2084                 spin_lock(&delayed_refs->lock);
2085                 avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
2086                 fs_info->avg_delayed_ref_runtime = avg >> 2;    /* div by 4 */
2087                 spin_unlock(&delayed_refs->lock);
2088         }
2089         return 0;
2090 }
2091
2092 #ifdef SCRAMBLE_DELAYED_REFS
2093 /*
2094  * Normally delayed refs get processed in ascending bytenr order. This
2095  * correlates in most cases to the order added. To expose dependencies on this
2096  * order, we start to process the tree in the middle instead of the beginning
2097  */
2098 static u64 find_middle(struct rb_root *root)
2099 {
2100         struct rb_node *n = root->rb_node;
2101         struct btrfs_delayed_ref_node *entry;
2102         int alt = 1;
2103         u64 middle;
2104         u64 first = 0, last = 0;
2105
2106         n = rb_first(root);
2107         if (n) {
2108                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2109                 first = entry->bytenr;
2110         }
2111         n = rb_last(root);
2112         if (n) {
2113                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2114                 last = entry->bytenr;
2115         }
2116         n = root->rb_node;
2117
2118         while (n) {
2119                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2120                 WARN_ON(!entry->in_tree);
2121
2122                 middle = entry->bytenr;
2123
2124                 if (alt)
2125                         n = n->rb_left;
2126                 else
2127                         n = n->rb_right;
2128
2129                 alt = 1 - alt;
2130         }
2131         return middle;
2132 }
2133 #endif
2134
2135 /*
2136  * Takes the number of bytes to be csumm'ed and figures out how many leaves it
2137  * would require to store the csums for that many bytes.
2138  */
2139 u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
2140 {
2141         u64 csum_size;
2142         u64 num_csums_per_leaf;
2143         u64 num_csums;
2144
2145         csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
2146         num_csums_per_leaf = div64_u64(csum_size,
2147                         (u64)btrfs_super_csum_size(fs_info->super_copy));
2148         num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
2149         num_csums += num_csums_per_leaf - 1;
2150         num_csums = div64_u64(num_csums, num_csums_per_leaf);
2151         return num_csums;
2152 }
2153
2154 /*
2155  * this starts processing the delayed reference count updates and
2156  * extent insertions we have queued up so far.  count can be
2157  * 0, which means to process everything in the tree at the start
2158  * of the run (but not newly added entries), or it can be some target
2159  * number you'd like to process.
2160  *
2161  * Returns 0 on success or if called with an aborted transaction
2162  * Returns <0 on error and aborts the transaction
2163  */
2164 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2165                            unsigned long count)
2166 {
2167         struct btrfs_fs_info *fs_info = trans->fs_info;
2168         struct rb_node *node;
2169         struct btrfs_delayed_ref_root *delayed_refs;
2170         struct btrfs_delayed_ref_head *head;
2171         int ret;
2172         int run_all = count == (unsigned long)-1;
2173
2174         /* We'll clean this up in btrfs_cleanup_transaction */
2175         if (TRANS_ABORTED(trans))
2176                 return 0;
2177
2178         if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
2179                 return 0;
2180
2181         delayed_refs = &trans->transaction->delayed_refs;
2182         if (count == 0)
2183                 count = atomic_read(&delayed_refs->num_entries) * 2;
2184
2185 again:
2186 #ifdef SCRAMBLE_DELAYED_REFS
2187         delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
2188 #endif
2189         ret = __btrfs_run_delayed_refs(trans, count);
2190         if (ret < 0) {
2191                 btrfs_abort_transaction(trans, ret);
2192                 return ret;
2193         }
2194
2195         if (run_all) {
2196                 btrfs_create_pending_block_groups(trans);
2197
2198                 spin_lock(&delayed_refs->lock);
2199                 node = rb_first_cached(&delayed_refs->href_root);
2200                 if (!node) {
2201                         spin_unlock(&delayed_refs->lock);
2202                         goto out;
2203                 }
2204                 head = rb_entry(node, struct btrfs_delayed_ref_head,
2205                                 href_node);
2206                 refcount_inc(&head->refs);
2207                 spin_unlock(&delayed_refs->lock);
2208
2209                 /* Mutex was contended, block until it's released and retry. */
2210                 mutex_lock(&head->mutex);
2211                 mutex_unlock(&head->mutex);
2212
2213                 btrfs_put_delayed_ref_head(head);
2214                 cond_resched();
2215                 goto again;
2216         }
2217 out:
2218         return 0;
2219 }
2220
2221 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2222                                 struct extent_buffer *eb, u64 flags,
2223                                 int level, int is_data)
2224 {
2225         struct btrfs_delayed_extent_op *extent_op;
2226         int ret;
2227
2228         extent_op = btrfs_alloc_delayed_extent_op();
2229         if (!extent_op)
2230                 return -ENOMEM;
2231
2232         extent_op->flags_to_set = flags;
2233         extent_op->update_flags = true;
2234         extent_op->update_key = false;
2235         extent_op->is_data = is_data ? true : false;
2236         extent_op->level = level;
2237
2238         ret = btrfs_add_delayed_extent_op(trans, eb->start, eb->len, extent_op);
2239         if (ret)
2240                 btrfs_free_delayed_extent_op(extent_op);
2241         return ret;
2242 }
2243
2244 static noinline int check_delayed_ref(struct btrfs_root *root,
2245                                       struct btrfs_path *path,
2246                                       u64 objectid, u64 offset, u64 bytenr)
2247 {
2248         struct btrfs_delayed_ref_head *head;
2249         struct btrfs_delayed_ref_node *ref;
2250         struct btrfs_delayed_data_ref *data_ref;
2251         struct btrfs_delayed_ref_root *delayed_refs;
2252         struct btrfs_transaction *cur_trans;
2253         struct rb_node *node;
2254         int ret = 0;
2255
2256         spin_lock(&root->fs_info->trans_lock);
2257         cur_trans = root->fs_info->running_transaction;
2258         if (cur_trans)
2259                 refcount_inc(&cur_trans->use_count);
2260         spin_unlock(&root->fs_info->trans_lock);
2261         if (!cur_trans)
2262                 return 0;
2263
2264         delayed_refs = &cur_trans->delayed_refs;
2265         spin_lock(&delayed_refs->lock);
2266         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
2267         if (!head) {
2268                 spin_unlock(&delayed_refs->lock);
2269                 btrfs_put_transaction(cur_trans);
2270                 return 0;
2271         }
2272
2273         if (!mutex_trylock(&head->mutex)) {
2274                 refcount_inc(&head->refs);
2275                 spin_unlock(&delayed_refs->lock);
2276
2277                 btrfs_release_path(path);
2278
2279                 /*
2280                  * Mutex was contended, block until it's released and let
2281                  * caller try again
2282                  */
2283                 mutex_lock(&head->mutex);
2284                 mutex_unlock(&head->mutex);
2285                 btrfs_put_delayed_ref_head(head);
2286                 btrfs_put_transaction(cur_trans);
2287                 return -EAGAIN;
2288         }
2289         spin_unlock(&delayed_refs->lock);
2290
2291         spin_lock(&head->lock);
2292         /*
2293          * XXX: We should replace this with a proper search function in the
2294          * future.
2295          */
2296         for (node = rb_first_cached(&head->ref_tree); node;
2297              node = rb_next(node)) {
2298                 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
2299                 /* If it's a shared ref we know a cross reference exists */
2300                 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
2301                         ret = 1;
2302                         break;
2303                 }
2304
2305                 data_ref = btrfs_delayed_node_to_data_ref(ref);
2306
2307                 /*
2308                  * If our ref doesn't match the one we're currently looking at
2309                  * then we have a cross reference.
2310                  */
2311                 if (data_ref->root != root->root_key.objectid ||
2312                     data_ref->objectid != objectid ||
2313                     data_ref->offset != offset) {
2314                         ret = 1;
2315                         break;
2316                 }
2317         }
2318         spin_unlock(&head->lock);
2319         mutex_unlock(&head->mutex);
2320         btrfs_put_transaction(cur_trans);
2321         return ret;
2322 }
2323
2324 static noinline int check_committed_ref(struct btrfs_root *root,
2325                                         struct btrfs_path *path,
2326                                         u64 objectid, u64 offset, u64 bytenr,
2327                                         bool strict)
2328 {
2329         struct btrfs_fs_info *fs_info = root->fs_info;
2330         struct btrfs_root *extent_root = fs_info->extent_root;
2331         struct extent_buffer *leaf;
2332         struct btrfs_extent_data_ref *ref;
2333         struct btrfs_extent_inline_ref *iref;
2334         struct btrfs_extent_item *ei;
2335         struct btrfs_key key;
2336         u32 item_size;
2337         int type;
2338         int ret;
2339
2340         key.objectid = bytenr;
2341         key.offset = (u64)-1;
2342         key.type = BTRFS_EXTENT_ITEM_KEY;
2343
2344         ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2345         if (ret < 0)
2346                 goto out;
2347         BUG_ON(ret == 0); /* Corruption */
2348
2349         ret = -ENOENT;
2350         if (path->slots[0] == 0)
2351                 goto out;
2352
2353         path->slots[0]--;
2354         leaf = path->nodes[0];
2355         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2356
2357         if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2358                 goto out;
2359
2360         ret = 1;
2361         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2362         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2363
2364         /* If extent item has more than 1 inline ref then it's shared */
2365         if (item_size != sizeof(*ei) +
2366             btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2367                 goto out;
2368
2369         /*
2370          * If extent created before last snapshot => it's shared unless the
2371          * snapshot has been deleted. Use the heuristic if strict is false.
2372          */
2373         if (!strict &&
2374             (btrfs_extent_generation(leaf, ei) <=
2375              btrfs_root_last_snapshot(&root->root_item)))
2376                 goto out;
2377
2378         iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2379
2380         /* If this extent has SHARED_DATA_REF then it's shared */
2381         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
2382         if (type != BTRFS_EXTENT_DATA_REF_KEY)
2383                 goto out;
2384
2385         ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2386         if (btrfs_extent_refs(leaf, ei) !=
2387             btrfs_extent_data_ref_count(leaf, ref) ||
2388             btrfs_extent_data_ref_root(leaf, ref) !=
2389             root->root_key.objectid ||
2390             btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2391             btrfs_extent_data_ref_offset(leaf, ref) != offset)
2392                 goto out;
2393
2394         ret = 0;
2395 out:
2396         return ret;
2397 }
2398
2399 int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
2400                           u64 bytenr, bool strict)
2401 {
2402         struct btrfs_path *path;
2403         int ret;
2404
2405         path = btrfs_alloc_path();
2406         if (!path)
2407                 return -ENOMEM;
2408
2409         do {
2410                 ret = check_committed_ref(root, path, objectid,
2411                                           offset, bytenr, strict);
2412                 if (ret && ret != -ENOENT)
2413                         goto out;
2414
2415                 ret = check_delayed_ref(root, path, objectid, offset, bytenr);
2416         } while (ret == -EAGAIN);
2417
2418 out:
2419         btrfs_free_path(path);
2420         if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
2421                 WARN_ON(ret > 0);
2422         return ret;
2423 }
2424
2425 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2426                            struct btrfs_root *root,
2427                            struct extent_buffer *buf,
2428                            int full_backref, int inc)
2429 {
2430         struct btrfs_fs_info *fs_info = root->fs_info;
2431         u64 bytenr;
2432         u64 num_bytes;
2433         u64 parent;
2434         u64 ref_root;
2435         u32 nritems;
2436         struct btrfs_key key;
2437         struct btrfs_file_extent_item *fi;
2438         struct btrfs_ref generic_ref = { 0 };
2439         bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
2440         int i;
2441         int action;
2442         int level;
2443         int ret = 0;
2444
2445         if (btrfs_is_testing(fs_info))
2446                 return 0;
2447
2448         ref_root = btrfs_header_owner(buf);
2449         nritems = btrfs_header_nritems(buf);
2450         level = btrfs_header_level(buf);
2451
2452         if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && level == 0)
2453                 return 0;
2454
2455         if (full_backref)
2456                 parent = buf->start;
2457         else
2458                 parent = 0;
2459         if (inc)
2460                 action = BTRFS_ADD_DELAYED_REF;
2461         else
2462                 action = BTRFS_DROP_DELAYED_REF;
2463
2464         for (i = 0; i < nritems; i++) {
2465                 if (level == 0) {
2466                         btrfs_item_key_to_cpu(buf, &key, i);
2467                         if (key.type != BTRFS_EXTENT_DATA_KEY)
2468                                 continue;
2469                         fi = btrfs_item_ptr(buf, i,
2470                                             struct btrfs_file_extent_item);
2471                         if (btrfs_file_extent_type(buf, fi) ==
2472                             BTRFS_FILE_EXTENT_INLINE)
2473                                 continue;
2474                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2475                         if (bytenr == 0)
2476                                 continue;
2477
2478                         num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2479                         key.offset -= btrfs_file_extent_offset(buf, fi);
2480                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2481                                                num_bytes, parent);
2482                         generic_ref.real_root = root->root_key.objectid;
2483                         btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
2484                                             key.offset);
2485                         generic_ref.skip_qgroup = for_reloc;
2486                         if (inc)
2487                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2488                         else
2489                                 ret = btrfs_free_extent(trans, &generic_ref);
2490                         if (ret)
2491                                 goto fail;
2492                 } else {
2493                         bytenr = btrfs_node_blockptr(buf, i);
2494                         num_bytes = fs_info->nodesize;
2495                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2496                                                num_bytes, parent);
2497                         generic_ref.real_root = root->root_key.objectid;
2498                         btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
2499                         generic_ref.skip_qgroup = for_reloc;
2500                         if (inc)
2501                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2502                         else
2503                                 ret = btrfs_free_extent(trans, &generic_ref);
2504                         if (ret)
2505                                 goto fail;
2506                 }
2507         }
2508         return 0;
2509 fail:
2510         return ret;
2511 }
2512
2513 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2514                   struct extent_buffer *buf, int full_backref)
2515 {
2516         return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2517 }
2518
2519 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2520                   struct extent_buffer *buf, int full_backref)
2521 {
2522         return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2523 }
2524
2525 int btrfs_extent_readonly(struct btrfs_fs_info *fs_info, u64 bytenr)
2526 {
2527         struct btrfs_block_group *block_group;
2528         int readonly = 0;
2529
2530         block_group = btrfs_lookup_block_group(fs_info, bytenr);
2531         if (!block_group || block_group->ro)
2532                 readonly = 1;
2533         if (block_group)
2534                 btrfs_put_block_group(block_group);
2535         return readonly;
2536 }
2537
2538 static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
2539 {
2540         struct btrfs_fs_info *fs_info = root->fs_info;
2541         u64 flags;
2542         u64 ret;
2543
2544         if (data)
2545                 flags = BTRFS_BLOCK_GROUP_DATA;
2546         else if (root == fs_info->chunk_root)
2547                 flags = BTRFS_BLOCK_GROUP_SYSTEM;
2548         else
2549                 flags = BTRFS_BLOCK_GROUP_METADATA;
2550
2551         ret = btrfs_get_alloc_profile(fs_info, flags);
2552         return ret;
2553 }
2554
2555 static u64 first_logical_byte(struct btrfs_fs_info *fs_info, u64 search_start)
2556 {
2557         struct btrfs_block_group *cache;
2558         u64 bytenr;
2559
2560         spin_lock(&fs_info->block_group_cache_lock);
2561         bytenr = fs_info->first_logical_byte;
2562         spin_unlock(&fs_info->block_group_cache_lock);
2563
2564         if (bytenr < (u64)-1)
2565                 return bytenr;
2566
2567         cache = btrfs_lookup_first_block_group(fs_info, search_start);
2568         if (!cache)
2569                 return 0;
2570
2571         bytenr = cache->start;
2572         btrfs_put_block_group(cache);
2573
2574         return bytenr;
2575 }
2576
2577 static int pin_down_extent(struct btrfs_trans_handle *trans,
2578                            struct btrfs_block_group *cache,
2579                            u64 bytenr, u64 num_bytes, int reserved)
2580 {
2581         struct btrfs_fs_info *fs_info = cache->fs_info;
2582
2583         spin_lock(&cache->space_info->lock);
2584         spin_lock(&cache->lock);
2585         cache->pinned += num_bytes;
2586         btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
2587                                              num_bytes);
2588         if (reserved) {
2589                 cache->reserved -= num_bytes;
2590                 cache->space_info->bytes_reserved -= num_bytes;
2591         }
2592         spin_unlock(&cache->lock);
2593         spin_unlock(&cache->space_info->lock);
2594
2595         percpu_counter_add_batch(&cache->space_info->total_bytes_pinned,
2596                     num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH);
2597         set_extent_dirty(&trans->transaction->pinned_extents, bytenr,
2598                          bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
2599         return 0;
2600 }
2601
2602 int btrfs_pin_extent(struct btrfs_trans_handle *trans,
2603                      u64 bytenr, u64 num_bytes, int reserved)
2604 {
2605         struct btrfs_block_group *cache;
2606
2607         cache = btrfs_lookup_block_group(trans->fs_info, bytenr);
2608         BUG_ON(!cache); /* Logic error */
2609
2610         pin_down_extent(trans, cache, bytenr, num_bytes, reserved);
2611
2612         btrfs_put_block_group(cache);
2613         return 0;
2614 }
2615
2616 /*
2617  * this function must be called within transaction
2618  */
2619 int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans,
2620                                     u64 bytenr, u64 num_bytes)
2621 {
2622         struct btrfs_block_group *cache;
2623         int ret;
2624
2625         btrfs_add_excluded_extent(trans->fs_info, bytenr, num_bytes);
2626
2627         cache = btrfs_lookup_block_group(trans->fs_info, bytenr);
2628         if (!cache)
2629                 return -EINVAL;
2630
2631         /*
2632          * pull in the free space cache (if any) so that our pin
2633          * removes the free space from the cache.  We have load_only set
2634          * to one because the slow code to read in the free extents does check
2635          * the pinned extents.
2636          */
2637         btrfs_cache_block_group(cache, 1);
2638
2639         pin_down_extent(trans, cache, bytenr, num_bytes, 0);
2640
2641         /* remove us from the free space cache (if we're there at all) */
2642         ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
2643         btrfs_put_block_group(cache);
2644         return ret;
2645 }
2646
2647 static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
2648                                    u64 start, u64 num_bytes)
2649 {
2650         int ret;
2651         struct btrfs_block_group *block_group;
2652         struct btrfs_caching_control *caching_ctl;
2653
2654         block_group = btrfs_lookup_block_group(fs_info, start);
2655         if (!block_group)
2656                 return -EINVAL;
2657
2658         btrfs_cache_block_group(block_group, 0);
2659         caching_ctl = btrfs_get_caching_control(block_group);
2660
2661         if (!caching_ctl) {
2662                 /* Logic error */
2663                 BUG_ON(!btrfs_block_group_done(block_group));
2664                 ret = btrfs_remove_free_space(block_group, start, num_bytes);
2665         } else {
2666                 mutex_lock(&caching_ctl->mutex);
2667
2668                 if (start >= caching_ctl->progress) {
2669                         ret = btrfs_add_excluded_extent(fs_info, start,
2670                                                         num_bytes);
2671                 } else if (start + num_bytes <= caching_ctl->progress) {
2672                         ret = btrfs_remove_free_space(block_group,
2673                                                       start, num_bytes);
2674                 } else {
2675                         num_bytes = caching_ctl->progress - start;
2676                         ret = btrfs_remove_free_space(block_group,
2677                                                       start, num_bytes);
2678                         if (ret)
2679                                 goto out_lock;
2680
2681                         num_bytes = (start + num_bytes) -
2682                                 caching_ctl->progress;
2683                         start = caching_ctl->progress;
2684                         ret = btrfs_add_excluded_extent(fs_info, start,
2685                                                         num_bytes);
2686                 }
2687 out_lock:
2688                 mutex_unlock(&caching_ctl->mutex);
2689                 btrfs_put_caching_control(caching_ctl);
2690         }
2691         btrfs_put_block_group(block_group);
2692         return ret;
2693 }
2694
2695 int btrfs_exclude_logged_extents(struct extent_buffer *eb)
2696 {
2697         struct btrfs_fs_info *fs_info = eb->fs_info;
2698         struct btrfs_file_extent_item *item;
2699         struct btrfs_key key;
2700         int found_type;
2701         int i;
2702         int ret = 0;
2703
2704         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
2705                 return 0;
2706
2707         for (i = 0; i < btrfs_header_nritems(eb); i++) {
2708                 btrfs_item_key_to_cpu(eb, &key, i);
2709                 if (key.type != BTRFS_EXTENT_DATA_KEY)
2710                         continue;
2711                 item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
2712                 found_type = btrfs_file_extent_type(eb, item);
2713                 if (found_type == BTRFS_FILE_EXTENT_INLINE)
2714                         continue;
2715                 if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
2716                         continue;
2717                 key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
2718                 key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
2719                 ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
2720                 if (ret)
2721                         break;
2722         }
2723
2724         return ret;
2725 }
2726
2727 static void
2728 btrfs_inc_block_group_reservations(struct btrfs_block_group *bg)
2729 {
2730         atomic_inc(&bg->reservations);
2731 }
2732
2733 void btrfs_prepare_extent_commit(struct btrfs_fs_info *fs_info)
2734 {
2735         struct btrfs_caching_control *next;
2736         struct btrfs_caching_control *caching_ctl;
2737         struct btrfs_block_group *cache;
2738
2739         down_write(&fs_info->commit_root_sem);
2740
2741         list_for_each_entry_safe(caching_ctl, next,
2742                                  &fs_info->caching_block_groups, list) {
2743                 cache = caching_ctl->block_group;
2744                 if (btrfs_block_group_done(cache)) {
2745                         cache->last_byte_to_unpin = (u64)-1;
2746                         list_del_init(&caching_ctl->list);
2747                         btrfs_put_caching_control(caching_ctl);
2748                 } else {
2749                         cache->last_byte_to_unpin = caching_ctl->progress;
2750                 }
2751         }
2752
2753         up_write(&fs_info->commit_root_sem);
2754
2755         btrfs_update_global_block_rsv(fs_info);
2756 }
2757
2758 /*
2759  * Returns the free cluster for the given space info and sets empty_cluster to
2760  * what it should be based on the mount options.
2761  */
2762 static struct btrfs_free_cluster *
2763 fetch_cluster_info(struct btrfs_fs_info *fs_info,
2764                    struct btrfs_space_info *space_info, u64 *empty_cluster)
2765 {
2766         struct btrfs_free_cluster *ret = NULL;
2767
2768         *empty_cluster = 0;
2769         if (btrfs_mixed_space_info(space_info))
2770                 return ret;
2771
2772         if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
2773                 ret = &fs_info->meta_alloc_cluster;
2774                 if (btrfs_test_opt(fs_info, SSD))
2775                         *empty_cluster = SZ_2M;
2776                 else
2777                         *empty_cluster = SZ_64K;
2778         } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
2779                    btrfs_test_opt(fs_info, SSD_SPREAD)) {
2780                 *empty_cluster = SZ_2M;
2781                 ret = &fs_info->data_alloc_cluster;
2782         }
2783
2784         return ret;
2785 }
2786
2787 static int unpin_extent_range(struct btrfs_fs_info *fs_info,
2788                               u64 start, u64 end,
2789                               const bool return_free_space)
2790 {
2791         struct btrfs_block_group *cache = NULL;
2792         struct btrfs_space_info *space_info;
2793         struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
2794         struct btrfs_free_cluster *cluster = NULL;
2795         u64 len;
2796         u64 total_unpinned = 0;
2797         u64 empty_cluster = 0;
2798         bool readonly;
2799
2800         while (start <= end) {
2801                 readonly = false;
2802                 if (!cache ||
2803                     start >= cache->start + cache->length) {
2804                         if (cache)
2805                                 btrfs_put_block_group(cache);
2806                         total_unpinned = 0;
2807                         cache = btrfs_lookup_block_group(fs_info, start);
2808                         BUG_ON(!cache); /* Logic error */
2809
2810                         cluster = fetch_cluster_info(fs_info,
2811                                                      cache->space_info,
2812                                                      &empty_cluster);
2813                         empty_cluster <<= 1;
2814                 }
2815
2816                 len = cache->start + cache->length - start;
2817                 len = min(len, end + 1 - start);
2818
2819                 if (start < cache->last_byte_to_unpin) {
2820                         len = min(len, cache->last_byte_to_unpin - start);
2821                         if (return_free_space)
2822                                 btrfs_add_free_space(cache, start, len);
2823                 }
2824
2825                 start += len;
2826                 total_unpinned += len;
2827                 space_info = cache->space_info;
2828
2829                 /*
2830                  * If this space cluster has been marked as fragmented and we've
2831                  * unpinned enough in this block group to potentially allow a
2832                  * cluster to be created inside of it go ahead and clear the
2833                  * fragmented check.
2834                  */
2835                 if (cluster && cluster->fragmented &&
2836                     total_unpinned > empty_cluster) {
2837                         spin_lock(&cluster->lock);
2838                         cluster->fragmented = 0;
2839                         spin_unlock(&cluster->lock);
2840                 }
2841
2842                 spin_lock(&space_info->lock);
2843                 spin_lock(&cache->lock);
2844                 cache->pinned -= len;
2845                 btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
2846                 space_info->max_extent_size = 0;
2847                 percpu_counter_add_batch(&space_info->total_bytes_pinned,
2848                             -len, BTRFS_TOTAL_BYTES_PINNED_BATCH);
2849                 if (cache->ro) {
2850                         space_info->bytes_readonly += len;
2851                         readonly = true;
2852                 }
2853                 spin_unlock(&cache->lock);
2854                 if (!readonly && return_free_space &&
2855                     global_rsv->space_info == space_info) {
2856                         u64 to_add = len;
2857
2858                         spin_lock(&global_rsv->lock);
2859                         if (!global_rsv->full) {
2860                                 to_add = min(len, global_rsv->size -
2861                                              global_rsv->reserved);
2862                                 global_rsv->reserved += to_add;
2863                                 btrfs_space_info_update_bytes_may_use(fs_info,
2864                                                 space_info, to_add);
2865                                 if (global_rsv->reserved >= global_rsv->size)
2866                                         global_rsv->full = 1;
2867                                 len -= to_add;
2868                         }
2869                         spin_unlock(&global_rsv->lock);
2870                 }
2871                 /* Add to any tickets we may have */
2872                 if (!readonly && return_free_space && len)
2873                         btrfs_try_granting_tickets(fs_info, space_info);
2874                 spin_unlock(&space_info->lock);
2875         }
2876
2877         if (cache)
2878                 btrfs_put_block_group(cache);
2879         return 0;
2880 }
2881
2882 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
2883 {
2884         struct btrfs_fs_info *fs_info = trans->fs_info;
2885         struct btrfs_block_group *block_group, *tmp;
2886         struct list_head *deleted_bgs;
2887         struct extent_io_tree *unpin;
2888         u64 start;
2889         u64 end;
2890         int ret;
2891
2892         unpin = &trans->transaction->pinned_extents;
2893
2894         while (!TRANS_ABORTED(trans)) {
2895                 struct extent_state *cached_state = NULL;
2896
2897                 mutex_lock(&fs_info->unused_bg_unpin_mutex);
2898                 ret = find_first_extent_bit(unpin, 0, &start, &end,
2899                                             EXTENT_DIRTY, &cached_state);
2900                 if (ret) {
2901                         mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2902                         break;
2903                 }
2904                 if (test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags))
2905                         clear_extent_bits(&fs_info->excluded_extents, start,
2906                                           end, EXTENT_UPTODATE);
2907
2908                 if (btrfs_test_opt(fs_info, DISCARD_SYNC))
2909                         ret = btrfs_discard_extent(fs_info, start,
2910                                                    end + 1 - start, NULL);
2911
2912                 clear_extent_dirty(unpin, start, end, &cached_state);
2913                 unpin_extent_range(fs_info, start, end, true);
2914                 mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2915                 free_extent_state(cached_state);
2916                 cond_resched();
2917         }
2918
2919         if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
2920                 btrfs_discard_calc_delay(&fs_info->discard_ctl);
2921                 btrfs_discard_schedule_work(&fs_info->discard_ctl, true);
2922         }
2923
2924         /*
2925          * Transaction is finished.  We don't need the lock anymore.  We
2926          * do need to clean up the block groups in case of a transaction
2927          * abort.
2928          */
2929         deleted_bgs = &trans->transaction->deleted_bgs;
2930         list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
2931                 u64 trimmed = 0;
2932
2933                 ret = -EROFS;
2934                 if (!TRANS_ABORTED(trans))
2935                         ret = btrfs_discard_extent(fs_info,
2936                                                    block_group->start,
2937                                                    block_group->length,
2938                                                    &trimmed);
2939
2940                 list_del_init(&block_group->bg_list);
2941                 btrfs_unfreeze_block_group(block_group);
2942                 btrfs_put_block_group(block_group);
2943
2944                 if (ret) {
2945                         const char *errstr = btrfs_decode_error(ret);
2946                         btrfs_warn(fs_info,
2947                            "discard failed while removing blockgroup: errno=%d %s",
2948                                    ret, errstr);
2949                 }
2950         }
2951
2952         return 0;
2953 }
2954
2955 /*
2956  * Drop one or more refs of @node.
2957  *
2958  * 1. Locate the extent refs.
2959  *    It's either inline in EXTENT/METADATA_ITEM or in keyed SHARED_* item.
2960  *    Locate it, then reduce the refs number or remove the ref line completely.
2961  *
2962  * 2. Update the refs count in EXTENT/METADATA_ITEM
2963  *
2964  * Inline backref case:
2965  *
2966  * in extent tree we have:
2967  *
2968  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
2969  *              refs 2 gen 6 flags DATA
2970  *              extent data backref root FS_TREE objectid 258 offset 0 count 1
2971  *              extent data backref root FS_TREE objectid 257 offset 0 count 1
2972  *
2973  * This function gets called with:
2974  *
2975  *    node->bytenr = 13631488
2976  *    node->num_bytes = 1048576
2977  *    root_objectid = FS_TREE
2978  *    owner_objectid = 257
2979  *    owner_offset = 0
2980  *    refs_to_drop = 1
2981  *
2982  * Then we should get some like:
2983  *
2984  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
2985  *              refs 1 gen 6 flags DATA
2986  *              extent data backref root FS_TREE objectid 258 offset 0 count 1
2987  *
2988  * Keyed backref case:
2989  *
2990  * in extent tree we have:
2991  *
2992  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
2993  *              refs 754 gen 6 flags DATA
2994  *      [...]
2995  *      item 2 key (13631488 EXTENT_DATA_REF <HASH>) itemoff 3915 itemsize 28
2996  *              extent data backref root FS_TREE objectid 866 offset 0 count 1
2997  *
2998  * This function get called with:
2999  *
3000  *    node->bytenr = 13631488
3001  *    node->num_bytes = 1048576
3002  *    root_objectid = FS_TREE
3003  *    owner_objectid = 866
3004  *    owner_offset = 0
3005  *    refs_to_drop = 1
3006  *
3007  * Then we should get some like:
3008  *
3009  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
3010  *              refs 753 gen 6 flags DATA
3011  *
3012  * And that (13631488 EXTENT_DATA_REF <HASH>) gets removed.
3013  */
3014 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3015                                struct btrfs_delayed_ref_node *node, u64 parent,
3016                                u64 root_objectid, u64 owner_objectid,
3017                                u64 owner_offset, int refs_to_drop,
3018                                struct btrfs_delayed_extent_op *extent_op)
3019 {
3020         struct btrfs_fs_info *info = trans->fs_info;
3021         struct btrfs_key key;
3022         struct btrfs_path *path;
3023         struct btrfs_root *extent_root = info->extent_root;
3024         struct extent_buffer *leaf;
3025         struct btrfs_extent_item *ei;
3026         struct btrfs_extent_inline_ref *iref;
3027         int ret;
3028         int is_data;
3029         int extent_slot = 0;
3030         int found_extent = 0;
3031         int num_to_del = 1;
3032         u32 item_size;
3033         u64 refs;
3034         u64 bytenr = node->bytenr;
3035         u64 num_bytes = node->num_bytes;
3036         int last_ref = 0;
3037         bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
3038
3039         path = btrfs_alloc_path();
3040         if (!path)
3041                 return -ENOMEM;
3042
3043         path->leave_spinning = 1;
3044
3045         is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3046
3047         if (!is_data && refs_to_drop != 1) {
3048                 btrfs_crit(info,
3049 "invalid refs_to_drop, dropping more than 1 refs for tree block %llu refs_to_drop %u",
3050                            node->bytenr, refs_to_drop);
3051                 ret = -EINVAL;
3052                 btrfs_abort_transaction(trans, ret);
3053                 goto out;
3054         }
3055
3056         if (is_data)
3057                 skinny_metadata = false;
3058
3059         ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
3060                                     parent, root_objectid, owner_objectid,
3061                                     owner_offset);
3062         if (ret == 0) {
3063                 /*
3064                  * Either the inline backref or the SHARED_DATA_REF/
3065                  * SHARED_BLOCK_REF is found
3066                  *
3067                  * Here is a quick path to locate EXTENT/METADATA_ITEM.
3068                  * It's possible the EXTENT/METADATA_ITEM is near current slot.
3069                  */
3070                 extent_slot = path->slots[0];
3071                 while (extent_slot >= 0) {
3072                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3073                                               extent_slot);
3074                         if (key.objectid != bytenr)
3075                                 break;
3076                         if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3077                             key.offset == num_bytes) {
3078                                 found_extent = 1;
3079                                 break;
3080                         }
3081                         if (key.type == BTRFS_METADATA_ITEM_KEY &&
3082                             key.offset == owner_objectid) {
3083                                 found_extent = 1;
3084                                 break;
3085                         }
3086
3087                         /* Quick path didn't find the EXTEMT/METADATA_ITEM */
3088                         if (path->slots[0] - extent_slot > 5)
3089                                 break;
3090                         extent_slot--;
3091                 }
3092
3093                 if (!found_extent) {
3094                         if (iref) {
3095                                 btrfs_crit(info,
3096 "invalid iref, no EXTENT/METADATA_ITEM found but has inline extent ref");
3097                                 btrfs_abort_transaction(trans, -EUCLEAN);
3098                                 goto err_dump;
3099                         }
3100                         /* Must be SHARED_* item, remove the backref first */
3101                         ret = remove_extent_backref(trans, path, NULL,
3102                                                     refs_to_drop,
3103                                                     is_data, &last_ref);
3104                         if (ret) {
3105                                 btrfs_abort_transaction(trans, ret);
3106                                 goto out;
3107                         }
3108                         btrfs_release_path(path);
3109                         path->leave_spinning = 1;
3110
3111                         /* Slow path to locate EXTENT/METADATA_ITEM */
3112                         key.objectid = bytenr;
3113                         key.type = BTRFS_EXTENT_ITEM_KEY;
3114                         key.offset = num_bytes;
3115
3116                         if (!is_data && skinny_metadata) {
3117                                 key.type = BTRFS_METADATA_ITEM_KEY;
3118                                 key.offset = owner_objectid;
3119                         }
3120
3121                         ret = btrfs_search_slot(trans, extent_root,
3122                                                 &key, path, -1, 1);
3123                         if (ret > 0 && skinny_metadata && path->slots[0]) {
3124                                 /*
3125                                  * Couldn't find our skinny metadata item,
3126                                  * see if we have ye olde extent item.
3127                                  */
3128                                 path->slots[0]--;
3129                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
3130                                                       path->slots[0]);
3131                                 if (key.objectid == bytenr &&
3132                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
3133                                     key.offset == num_bytes)
3134                                         ret = 0;
3135                         }
3136
3137                         if (ret > 0 && skinny_metadata) {
3138                                 skinny_metadata = false;
3139                                 key.objectid = bytenr;
3140                                 key.type = BTRFS_EXTENT_ITEM_KEY;
3141                                 key.offset = num_bytes;
3142                                 btrfs_release_path(path);
3143                                 ret = btrfs_search_slot(trans, extent_root,
3144                                                         &key, path, -1, 1);
3145                         }
3146
3147                         if (ret) {
3148                                 btrfs_err(info,
3149                                           "umm, got %d back from search, was looking for %llu",
3150                                           ret, bytenr);
3151                                 if (ret > 0)
3152                                         btrfs_print_leaf(path->nodes[0]);
3153                         }
3154                         if (ret < 0) {
3155                                 btrfs_abort_transaction(trans, ret);
3156                                 goto out;
3157                         }
3158                         extent_slot = path->slots[0];
3159                 }
3160         } else if (WARN_ON(ret == -ENOENT)) {
3161                 btrfs_print_leaf(path->nodes[0]);
3162                 btrfs_err(info,
3163                         "unable to find ref byte nr %llu parent %llu root %llu  owner %llu offset %llu",
3164                         bytenr, parent, root_objectid, owner_objectid,
3165                         owner_offset);
3166                 btrfs_abort_transaction(trans, ret);
3167                 goto out;
3168         } else {
3169                 btrfs_abort_transaction(trans, ret);
3170                 goto out;
3171         }
3172
3173         leaf = path->nodes[0];
3174         item_size = btrfs_item_size_nr(leaf, extent_slot);
3175         if (unlikely(item_size < sizeof(*ei))) {
3176                 ret = -EINVAL;
3177                 btrfs_print_v0_err(info);
3178                 btrfs_abort_transaction(trans, ret);
3179                 goto out;
3180         }
3181         ei = btrfs_item_ptr(leaf, extent_slot,
3182                             struct btrfs_extent_item);
3183         if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
3184             key.type == BTRFS_EXTENT_ITEM_KEY) {
3185                 struct btrfs_tree_block_info *bi;
3186                 if (item_size < sizeof(*ei) + sizeof(*bi)) {
3187                         btrfs_crit(info,
3188 "invalid extent item size for key (%llu, %u, %llu) owner %llu, has %u expect >= %lu",
3189                                    key.objectid, key.type, key.offset,
3190                                    owner_objectid, item_size,
3191                                    sizeof(*ei) + sizeof(*bi));
3192                         btrfs_abort_transaction(trans, -EUCLEAN);
3193                         goto err_dump;
3194                 }
3195                 bi = (struct btrfs_tree_block_info *)(ei + 1);
3196                 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3197         }
3198
3199         refs = btrfs_extent_refs(leaf, ei);
3200         if (refs < refs_to_drop) {
3201                 btrfs_crit(info,
3202                 "trying to drop %d refs but we only have %llu for bytenr %llu",
3203                           refs_to_drop, refs, bytenr);
3204                 btrfs_abort_transaction(trans, -EUCLEAN);
3205                 goto err_dump;
3206         }
3207         refs -= refs_to_drop;
3208
3209         if (refs > 0) {
3210                 if (extent_op)
3211                         __run_delayed_extent_op(extent_op, leaf, ei);
3212                 /*
3213                  * In the case of inline back ref, reference count will
3214                  * be updated by remove_extent_backref
3215                  */
3216                 if (iref) {
3217                         if (!found_extent) {
3218                                 btrfs_crit(info,
3219 "invalid iref, got inlined extent ref but no EXTENT/METADATA_ITEM found");
3220                                 btrfs_abort_transaction(trans, -EUCLEAN);
3221                                 goto err_dump;
3222                         }
3223                 } else {
3224                         btrfs_set_extent_refs(leaf, ei, refs);
3225                         btrfs_mark_buffer_dirty(leaf);
3226                 }
3227                 if (found_extent) {
3228                         ret = remove_extent_backref(trans, path, iref,
3229                                                     refs_to_drop, is_data,
3230                                                     &last_ref);
3231                         if (ret) {
3232                                 btrfs_abort_transaction(trans, ret);
3233                                 goto out;
3234                         }
3235                 }
3236         } else {
3237                 /* In this branch refs == 1 */
3238                 if (found_extent) {
3239                         if (is_data && refs_to_drop !=
3240                             extent_data_ref_count(path, iref)) {
3241                                 btrfs_crit(info,
3242                 "invalid refs_to_drop, current refs %u refs_to_drop %u",
3243                                            extent_data_ref_count(path, iref),
3244                                            refs_to_drop);
3245                                 btrfs_abort_transaction(trans, -EUCLEAN);
3246                                 goto err_dump;
3247                         }
3248                         if (iref) {
3249                                 if (path->slots[0] != extent_slot) {
3250                                         btrfs_crit(info,
3251 "invalid iref, extent item key (%llu %u %llu) doesn't have wanted iref",
3252                                                    key.objectid, key.type,
3253                                                    key.offset);
3254                                         btrfs_abort_transaction(trans, -EUCLEAN);
3255                                         goto err_dump;
3256                                 }
3257                         } else {
3258                                 /*
3259                                  * No inline ref, we must be at SHARED_* item,
3260                                  * And it's single ref, it must be:
3261                                  * |    extent_slot       ||extent_slot + 1|
3262                                  * [ EXTENT/METADATA_ITEM ][ SHARED_* ITEM ]
3263                                  */
3264                                 if (path->slots[0] != extent_slot + 1) {
3265                                         btrfs_crit(info,
3266         "invalid SHARED_* item, previous item is not EXTENT/METADATA_ITEM");
3267                                         btrfs_abort_transaction(trans, -EUCLEAN);
3268                                         goto err_dump;
3269                                 }
3270                                 path->slots[0] = extent_slot;
3271                                 num_to_del = 2;
3272                         }
3273                 }
3274
3275                 last_ref = 1;
3276                 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3277                                       num_to_del);
3278                 if (ret) {
3279                         btrfs_abort_transaction(trans, ret);
3280                         goto out;
3281                 }
3282                 btrfs_release_path(path);
3283
3284                 if (is_data) {
3285                         ret = btrfs_del_csums(trans, info->csum_root, bytenr,
3286                                               num_bytes);
3287                         if (ret) {
3288                                 btrfs_abort_transaction(trans, ret);
3289                                 goto out;
3290                         }
3291                 }
3292
3293                 ret = add_to_free_space_tree(trans, bytenr, num_bytes);
3294                 if (ret) {
3295                         btrfs_abort_transaction(trans, ret);
3296                         goto out;
3297                 }
3298
3299                 ret = btrfs_update_block_group(trans, bytenr, num_bytes, 0);
3300                 if (ret) {
3301                         btrfs_abort_transaction(trans, ret);
3302                         goto out;
3303                 }
3304         }
3305         btrfs_release_path(path);
3306
3307 out:
3308         btrfs_free_path(path);
3309         return ret;
3310 err_dump:
3311         /*
3312          * Leaf dump can take up a lot of log buffer, so we only do full leaf
3313          * dump for debug build.
3314          */
3315         if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) {
3316                 btrfs_crit(info, "path->slots[0]=%d extent_slot=%d",
3317                            path->slots[0], extent_slot);
3318                 btrfs_print_leaf(path->nodes[0]);
3319         }
3320
3321         btrfs_free_path(path);
3322         return -EUCLEAN;
3323 }
3324
3325 /*
3326  * when we free an block, it is possible (and likely) that we free the last
3327  * delayed ref for that extent as well.  This searches the delayed ref tree for
3328  * a given extent, and if there are no other delayed refs to be processed, it
3329  * removes it from the tree.
3330  */
3331 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3332                                       u64 bytenr)
3333 {
3334         struct btrfs_delayed_ref_head *head;
3335         struct btrfs_delayed_ref_root *delayed_refs;
3336         int ret = 0;
3337
3338         delayed_refs = &trans->transaction->delayed_refs;
3339         spin_lock(&delayed_refs->lock);
3340         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
3341         if (!head)
3342                 goto out_delayed_unlock;
3343
3344         spin_lock(&head->lock);
3345         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
3346                 goto out;
3347
3348         if (cleanup_extent_op(head) != NULL)
3349                 goto out;
3350
3351         /*
3352          * waiting for the lock here would deadlock.  If someone else has it
3353          * locked they are already in the process of dropping it anyway
3354          */
3355         if (!mutex_trylock(&head->mutex))
3356                 goto out;
3357
3358         btrfs_delete_ref_head(delayed_refs, head);
3359         head->processing = 0;
3360
3361         spin_unlock(&head->lock);
3362         spin_unlock(&delayed_refs->lock);
3363
3364         BUG_ON(head->extent_op);
3365         if (head->must_insert_reserved)
3366                 ret = 1;
3367
3368         btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
3369         mutex_unlock(&head->mutex);
3370         btrfs_put_delayed_ref_head(head);
3371         return ret;
3372 out:
3373         spin_unlock(&head->lock);
3374
3375 out_delayed_unlock:
3376         spin_unlock(&delayed_refs->lock);
3377         return 0;
3378 }
3379
3380 void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
3381                            struct btrfs_root *root,
3382                            struct extent_buffer *buf,
3383                            u64 parent, int last_ref)
3384 {
3385         struct btrfs_fs_info *fs_info = root->fs_info;
3386         struct btrfs_ref generic_ref = { 0 };
3387         int pin = 1;
3388         int ret;
3389
3390         btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
3391                                buf->start, buf->len, parent);
3392         btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
3393                             root->root_key.objectid);
3394
3395         if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3396                 int old_ref_mod, new_ref_mod;
3397
3398                 btrfs_ref_tree_mod(fs_info, &generic_ref);
3399                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL,
3400                                                  &old_ref_mod, &new_ref_mod);
3401                 BUG_ON(ret); /* -ENOMEM */
3402                 pin = old_ref_mod >= 0 && new_ref_mod < 0;
3403         }
3404
3405         if (last_ref && btrfs_header_generation(buf) == trans->transid) {
3406                 struct btrfs_block_group *cache;
3407
3408                 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3409                         ret = check_ref_cleanup(trans, buf->start);
3410                         if (!ret)
3411                                 goto out;
3412                 }
3413
3414                 pin = 0;
3415                 cache = btrfs_lookup_block_group(fs_info, buf->start);
3416
3417                 if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3418                         pin_down_extent(trans, cache, buf->start, buf->len, 1);
3419                         btrfs_put_block_group(cache);
3420                         goto out;
3421                 }
3422
3423                 WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
3424
3425                 btrfs_add_free_space(cache, buf->start, buf->len);
3426                 btrfs_free_reserved_bytes(cache, buf->len, 0);
3427                 btrfs_put_block_group(cache);
3428                 trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
3429         }
3430 out:
3431         if (pin)
3432                 add_pinned_bytes(fs_info, &generic_ref);
3433
3434         if (last_ref) {
3435                 /*
3436                  * Deleting the buffer, clear the corrupt flag since it doesn't
3437                  * matter anymore.
3438                  */
3439                 clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
3440         }
3441 }
3442
3443 /* Can return -ENOMEM */
3444 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
3445 {
3446         struct btrfs_fs_info *fs_info = trans->fs_info;
3447         int old_ref_mod, new_ref_mod;
3448         int ret;
3449
3450         if (btrfs_is_testing(fs_info))
3451                 return 0;
3452
3453         /*
3454          * tree log blocks never actually go into the extent allocation
3455          * tree, just update pinning info and exit early.
3456          */
3457         if ((ref->type == BTRFS_REF_METADATA &&
3458              ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3459             (ref->type == BTRFS_REF_DATA &&
3460              ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
3461                 /* unlocks the pinned mutex */
3462                 btrfs_pin_extent(trans, ref->bytenr, ref->len, 1);
3463                 old_ref_mod = new_ref_mod = 0;
3464                 ret = 0;
3465         } else if (ref->type == BTRFS_REF_METADATA) {
3466                 ret = btrfs_add_delayed_tree_ref(trans, ref, NULL,
3467                                                  &old_ref_mod, &new_ref_mod);
3468         } else {
3469                 ret = btrfs_add_delayed_data_ref(trans, ref, 0,
3470                                                  &old_ref_mod, &new_ref_mod);
3471         }
3472
3473         if (!((ref->type == BTRFS_REF_METADATA &&
3474                ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3475               (ref->type == BTRFS_REF_DATA &&
3476                ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
3477                 btrfs_ref_tree_mod(fs_info, ref);
3478
3479         if (ret == 0 && old_ref_mod >= 0 && new_ref_mod < 0)
3480                 add_pinned_bytes(fs_info, ref);
3481
3482         return ret;
3483 }
3484
3485 enum btrfs_loop_type {
3486         LOOP_CACHING_NOWAIT,
3487         LOOP_CACHING_WAIT,
3488         LOOP_ALLOC_CHUNK,
3489         LOOP_NO_EMPTY_SIZE,
3490 };
3491
3492 static inline void
3493 btrfs_lock_block_group(struct btrfs_block_group *cache,
3494                        int delalloc)
3495 {
3496         if (delalloc)
3497                 down_read(&cache->data_rwsem);
3498 }
3499
3500 static inline void btrfs_grab_block_group(struct btrfs_block_group *cache,
3501                        int delalloc)
3502 {
3503         btrfs_get_block_group(cache);
3504         if (delalloc)
3505                 down_read(&cache->data_rwsem);
3506 }
3507
3508 static struct btrfs_block_group *btrfs_lock_cluster(
3509                    struct btrfs_block_group *block_group,
3510                    struct btrfs_free_cluster *cluster,
3511                    int delalloc)
3512         __acquires(&cluster->refill_lock)
3513 {
3514         struct btrfs_block_group *used_bg = NULL;
3515
3516         spin_lock(&cluster->refill_lock);
3517         while (1) {
3518                 used_bg = cluster->block_group;
3519                 if (!used_bg)
3520                         return NULL;
3521
3522                 if (used_bg == block_group)
3523                         return used_bg;
3524
3525                 btrfs_get_block_group(used_bg);
3526
3527                 if (!delalloc)
3528                         return used_bg;
3529
3530                 if (down_read_trylock(&used_bg->data_rwsem))
3531                         return used_bg;
3532
3533                 spin_unlock(&cluster->refill_lock);
3534
3535                 /* We should only have one-level nested. */
3536                 down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
3537
3538                 spin_lock(&cluster->refill_lock);
3539                 if (used_bg == cluster->block_group)
3540                         return used_bg;
3541
3542                 up_read(&used_bg->data_rwsem);
3543                 btrfs_put_block_group(used_bg);
3544         }
3545 }
3546
3547 static inline void
3548 btrfs_release_block_group(struct btrfs_block_group *cache,
3549                          int delalloc)
3550 {
3551         if (delalloc)
3552                 up_read(&cache->data_rwsem);
3553         btrfs_put_block_group(cache);
3554 }
3555
3556 enum btrfs_extent_allocation_policy {
3557         BTRFS_EXTENT_ALLOC_CLUSTERED,
3558 };
3559
3560 /*
3561  * Structure used internally for find_free_extent() function.  Wraps needed
3562  * parameters.
3563  */
3564 struct find_free_extent_ctl {
3565         /* Basic allocation info */
3566         u64 num_bytes;
3567         u64 empty_size;
3568         u64 flags;
3569         int delalloc;
3570
3571         /* Where to start the search inside the bg */
3572         u64 search_start;
3573
3574         /* For clustered allocation */
3575         u64 empty_cluster;
3576         struct btrfs_free_cluster *last_ptr;
3577         bool use_cluster;
3578
3579         bool have_caching_bg;
3580         bool orig_have_caching_bg;
3581
3582         /* RAID index, converted from flags */
3583         int index;
3584
3585         /*
3586          * Current loop number, check find_free_extent_update_loop() for details
3587          */
3588         int loop;
3589
3590         /*
3591          * Whether we're refilling a cluster, if true we need to re-search
3592          * current block group but don't try to refill the cluster again.
3593          */
3594         bool retry_clustered;
3595
3596         /*
3597          * Whether we're updating free space cache, if true we need to re-search
3598          * current block group but don't try updating free space cache again.
3599          */
3600         bool retry_unclustered;
3601
3602         /* If current block group is cached */
3603         int cached;
3604
3605         /* Max contiguous hole found */
3606         u64 max_extent_size;
3607
3608         /* Total free space from free space cache, not always contiguous */
3609         u64 total_free_space;
3610
3611         /* Found result */
3612         u64 found_offset;
3613
3614         /* Hint where to start looking for an empty space */
3615         u64 hint_byte;
3616
3617         /* Allocation policy */
3618         enum btrfs_extent_allocation_policy policy;
3619 };
3620
3621
3622 /*
3623  * Helper function for find_free_extent().
3624  *
3625  * Return -ENOENT to inform caller that we need fallback to unclustered mode.
3626  * Return -EAGAIN to inform caller that we need to re-search this block group
3627  * Return >0 to inform caller that we find nothing
3628  * Return 0 means we have found a location and set ffe_ctl->found_offset.
3629  */
3630 static int find_free_extent_clustered(struct btrfs_block_group *bg,
3631                                       struct find_free_extent_ctl *ffe_ctl,
3632                                       struct btrfs_block_group **cluster_bg_ret)
3633 {
3634         struct btrfs_block_group *cluster_bg;
3635         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3636         u64 aligned_cluster;
3637         u64 offset;
3638         int ret;
3639
3640         cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
3641         if (!cluster_bg)
3642                 goto refill_cluster;
3643         if (cluster_bg != bg && (cluster_bg->ro ||
3644             !block_group_bits(cluster_bg, ffe_ctl->flags)))
3645                 goto release_cluster;
3646
3647         offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
3648                         ffe_ctl->num_bytes, cluster_bg->start,
3649                         &ffe_ctl->max_extent_size);
3650         if (offset) {
3651                 /* We have a block, we're done */
3652                 spin_unlock(&last_ptr->refill_lock);
3653                 trace_btrfs_reserve_extent_cluster(cluster_bg,
3654                                 ffe_ctl->search_start, ffe_ctl->num_bytes);
3655                 *cluster_bg_ret = cluster_bg;
3656                 ffe_ctl->found_offset = offset;
3657                 return 0;
3658         }
3659         WARN_ON(last_ptr->block_group != cluster_bg);
3660
3661 release_cluster:
3662         /*
3663          * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
3664          * lets just skip it and let the allocator find whatever block it can
3665          * find. If we reach this point, we will have tried the cluster
3666          * allocator plenty of times and not have found anything, so we are
3667          * likely way too fragmented for the clustering stuff to find anything.
3668          *
3669          * However, if the cluster is taken from the current block group,
3670          * release the cluster first, so that we stand a better chance of
3671          * succeeding in the unclustered allocation.
3672          */
3673         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
3674                 spin_unlock(&last_ptr->refill_lock);
3675                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3676                 return -ENOENT;
3677         }
3678
3679         /* This cluster didn't work out, free it and start over */
3680         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3681
3682         if (cluster_bg != bg)
3683                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3684
3685 refill_cluster:
3686         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
3687                 spin_unlock(&last_ptr->refill_lock);
3688                 return -ENOENT;
3689         }
3690
3691         aligned_cluster = max_t(u64,
3692                         ffe_ctl->empty_cluster + ffe_ctl->empty_size,
3693                         bg->full_stripe_len);
3694         ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
3695                         ffe_ctl->num_bytes, aligned_cluster);
3696         if (ret == 0) {
3697                 /* Now pull our allocation out of this cluster */
3698                 offset = btrfs_alloc_from_cluster(bg, last_ptr,
3699                                 ffe_ctl->num_bytes, ffe_ctl->search_start,
3700                                 &ffe_ctl->max_extent_size);
3701                 if (offset) {
3702                         /* We found one, proceed */
3703                         spin_unlock(&last_ptr->refill_lock);
3704                         trace_btrfs_reserve_extent_cluster(bg,
3705                                         ffe_ctl->search_start,
3706                                         ffe_ctl->num_bytes);
3707                         ffe_ctl->found_offset = offset;
3708                         return 0;
3709                 }
3710         } else if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
3711                    !ffe_ctl->retry_clustered) {
3712                 spin_unlock(&last_ptr->refill_lock);
3713
3714                 ffe_ctl->retry_clustered = true;
3715                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3716                                 ffe_ctl->empty_cluster + ffe_ctl->empty_size);
3717                 return -EAGAIN;
3718         }
3719         /*
3720          * At this point we either didn't find a cluster or we weren't able to
3721          * allocate a block from our cluster.  Free the cluster we've been
3722          * trying to use, and go to the next block group.
3723          */
3724         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3725         spin_unlock(&last_ptr->refill_lock);
3726         return 1;
3727 }
3728
3729 /*
3730  * Return >0 to inform caller that we find nothing
3731  * Return 0 when we found an free extent and set ffe_ctrl->found_offset
3732  * Return -EAGAIN to inform caller that we need to re-search this block group
3733  */
3734 static int find_free_extent_unclustered(struct btrfs_block_group *bg,
3735                                         struct find_free_extent_ctl *ffe_ctl)
3736 {
3737         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3738         u64 offset;
3739
3740         /*
3741          * We are doing an unclustered allocation, set the fragmented flag so
3742          * we don't bother trying to setup a cluster again until we get more
3743          * space.
3744          */
3745         if (unlikely(last_ptr)) {
3746                 spin_lock(&last_ptr->lock);
3747                 last_ptr->fragmented = 1;
3748                 spin_unlock(&last_ptr->lock);
3749         }
3750         if (ffe_ctl->cached) {
3751                 struct btrfs_free_space_ctl *free_space_ctl;
3752
3753                 free_space_ctl = bg->free_space_ctl;
3754                 spin_lock(&free_space_ctl->tree_lock);
3755                 if (free_space_ctl->free_space <
3756                     ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
3757                     ffe_ctl->empty_size) {
3758                         ffe_ctl->total_free_space = max_t(u64,
3759                                         ffe_ctl->total_free_space,
3760                                         free_space_ctl->free_space);
3761                         spin_unlock(&free_space_ctl->tree_lock);
3762                         return 1;
3763                 }
3764                 spin_unlock(&free_space_ctl->tree_lock);
3765         }
3766
3767         offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
3768                         ffe_ctl->num_bytes, ffe_ctl->empty_size,
3769                         &ffe_ctl->max_extent_size);
3770
3771         /*
3772          * If we didn't find a chunk, and we haven't failed on this block group
3773          * before, and this block group is in the middle of caching and we are
3774          * ok with waiting, then go ahead and wait for progress to be made, and
3775          * set @retry_unclustered to true.
3776          *
3777          * If @retry_unclustered is true then we've already waited on this
3778          * block group once and should move on to the next block group.
3779          */
3780         if (!offset && !ffe_ctl->retry_unclustered && !ffe_ctl->cached &&
3781             ffe_ctl->loop > LOOP_CACHING_NOWAIT) {
3782                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3783                                                       ffe_ctl->empty_size);
3784                 ffe_ctl->retry_unclustered = true;
3785                 return -EAGAIN;
3786         } else if (!offset) {
3787                 return 1;
3788         }
3789         ffe_ctl->found_offset = offset;
3790         return 0;
3791 }
3792
3793 static int do_allocation_clustered(struct btrfs_block_group *block_group,
3794                                    struct find_free_extent_ctl *ffe_ctl,
3795                                    struct btrfs_block_group **bg_ret)
3796 {
3797         int ret;
3798
3799         /* We want to try and use the cluster allocator, so lets look there */
3800         if (ffe_ctl->last_ptr && ffe_ctl->use_cluster) {
3801                 ret = find_free_extent_clustered(block_group, ffe_ctl, bg_ret);
3802                 if (ret >= 0 || ret == -EAGAIN)
3803                         return ret;
3804                 /* ret == -ENOENT case falls through */
3805         }
3806
3807         return find_free_extent_unclustered(block_group, ffe_ctl);
3808 }
3809
3810 static int do_allocation(struct btrfs_block_group *block_group,
3811                          struct find_free_extent_ctl *ffe_ctl,
3812                          struct btrfs_block_group **bg_ret)
3813 {
3814         switch (ffe_ctl->policy) {
3815         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3816                 return do_allocation_clustered(block_group, ffe_ctl, bg_ret);
3817         default:
3818                 BUG();
3819         }
3820 }
3821
3822 static void release_block_group(struct btrfs_block_group *block_group,
3823                                 struct find_free_extent_ctl *ffe_ctl,
3824                                 int delalloc)
3825 {
3826         switch (ffe_ctl->policy) {
3827         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3828                 ffe_ctl->retry_clustered = false;
3829                 ffe_ctl->retry_unclustered = false;
3830                 break;
3831         default:
3832                 BUG();
3833         }
3834
3835         BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
3836                ffe_ctl->index);
3837         btrfs_release_block_group(block_group, delalloc);
3838 }
3839
3840 static void found_extent_clustered(struct find_free_extent_ctl *ffe_ctl,
3841                                    struct btrfs_key *ins)
3842 {
3843         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3844
3845         if (!ffe_ctl->use_cluster && last_ptr) {
3846                 spin_lock(&last_ptr->lock);
3847                 last_ptr->window_start = ins->objectid;
3848                 spin_unlock(&last_ptr->lock);
3849         }
3850 }
3851
3852 static void found_extent(struct find_free_extent_ctl *ffe_ctl,
3853                          struct btrfs_key *ins)
3854 {
3855         switch (ffe_ctl->policy) {
3856         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3857                 found_extent_clustered(ffe_ctl, ins);
3858                 break;
3859         default:
3860                 BUG();
3861         }
3862 }
3863
3864 static int chunk_allocation_failed(struct find_free_extent_ctl *ffe_ctl)
3865 {
3866         switch (ffe_ctl->policy) {
3867         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3868                 /*
3869                  * If we can't allocate a new chunk we've already looped through
3870                  * at least once, move on to the NO_EMPTY_SIZE case.
3871                  */
3872                 ffe_ctl->loop = LOOP_NO_EMPTY_SIZE;
3873                 return 0;
3874         default:
3875                 BUG();
3876         }
3877 }
3878
3879 /*
3880  * Return >0 means caller needs to re-search for free extent
3881  * Return 0 means we have the needed free extent.
3882  * Return <0 means we failed to locate any free extent.
3883  */
3884 static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
3885                                         struct btrfs_key *ins,
3886                                         struct find_free_extent_ctl *ffe_ctl,
3887                                         bool full_search)
3888 {
3889         struct btrfs_root *root = fs_info->extent_root;
3890         int ret;
3891
3892         if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
3893             ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
3894                 ffe_ctl->orig_have_caching_bg = true;
3895
3896         if (!ins->objectid && ffe_ctl->loop >= LOOP_CACHING_WAIT &&
3897             ffe_ctl->have_caching_bg)
3898                 return 1;
3899
3900         if (!ins->objectid && ++(ffe_ctl->index) < BTRFS_NR_RAID_TYPES)
3901                 return 1;
3902
3903         if (ins->objectid) {
3904                 found_extent(ffe_ctl, ins);
3905                 return 0;
3906         }
3907
3908         /*
3909          * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
3910          *                      caching kthreads as we move along
3911          * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
3912          * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
3913          * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
3914          *                     again
3915          */
3916         if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
3917                 ffe_ctl->index = 0;
3918                 if (ffe_ctl->loop == LOOP_CACHING_NOWAIT) {
3919                         /*
3920                          * We want to skip the LOOP_CACHING_WAIT step if we
3921                          * don't have any uncached bgs and we've already done a
3922                          * full search through.
3923                          */
3924                         if (ffe_ctl->orig_have_caching_bg || !full_search)
3925                                 ffe_ctl->loop = LOOP_CACHING_WAIT;
3926                         else
3927                                 ffe_ctl->loop = LOOP_ALLOC_CHUNK;
3928                 } else {
3929                         ffe_ctl->loop++;
3930                 }
3931
3932                 if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
3933                         struct btrfs_trans_handle *trans;
3934                         int exist = 0;
3935
3936                         trans = current->journal_info;
3937                         if (trans)
3938                                 exist = 1;
3939                         else
3940                                 trans = btrfs_join_transaction(root);
3941
3942                         if (IS_ERR(trans)) {
3943                                 ret = PTR_ERR(trans);
3944                                 return ret;
3945                         }
3946
3947                         ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
3948                                                 CHUNK_ALLOC_FORCE);
3949
3950                         /* Do not bail out on ENOSPC since we can do more. */
3951                         if (ret == -ENOSPC)
3952                                 ret = chunk_allocation_failed(ffe_ctl);
3953                         else if (ret < 0)
3954                                 btrfs_abort_transaction(trans, ret);
3955                         else
3956                                 ret = 0;
3957                         if (!exist)
3958                                 btrfs_end_transaction(trans);
3959                         if (ret)
3960                                 return ret;
3961                 }
3962
3963                 if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
3964                         if (ffe_ctl->policy != BTRFS_EXTENT_ALLOC_CLUSTERED)
3965                                 return -ENOSPC;
3966
3967                         /*
3968                          * Don't loop again if we already have no empty_size and
3969                          * no empty_cluster.
3970                          */
3971                         if (ffe_ctl->empty_size == 0 &&
3972                             ffe_ctl->empty_cluster == 0)
3973                                 return -ENOSPC;
3974                         ffe_ctl->empty_size = 0;
3975                         ffe_ctl->empty_cluster = 0;
3976                 }
3977                 return 1;
3978         }
3979         return -ENOSPC;
3980 }
3981
3982 static int prepare_allocation_clustered(struct btrfs_fs_info *fs_info,
3983                                         struct find_free_extent_ctl *ffe_ctl,
3984                                         struct btrfs_space_info *space_info,
3985                                         struct btrfs_key *ins)
3986 {
3987         /*
3988          * If our free space is heavily fragmented we may not be able to make
3989          * big contiguous allocations, so instead of doing the expensive search
3990          * for free space, simply return ENOSPC with our max_extent_size so we
3991          * can go ahead and search for a more manageable chunk.
3992          *
3993          * If our max_extent_size is large enough for our allocation simply
3994          * disable clustering since we will likely not be able to find enough
3995          * space to create a cluster and induce latency trying.
3996          */
3997         if (space_info->max_extent_size) {
3998                 spin_lock(&space_info->lock);
3999                 if (space_info->max_extent_size &&
4000                     ffe_ctl->num_bytes > space_info->max_extent_size) {
4001                         ins->offset = space_info->max_extent_size;
4002                         spin_unlock(&space_info->lock);
4003                         return -ENOSPC;
4004                 } else if (space_info->max_extent_size) {
4005                         ffe_ctl->use_cluster = false;
4006                 }
4007                 spin_unlock(&space_info->lock);
4008         }
4009
4010         ffe_ctl->last_ptr = fetch_cluster_info(fs_info, space_info,
4011                                                &ffe_ctl->empty_cluster);
4012         if (ffe_ctl->last_ptr) {
4013                 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
4014
4015                 spin_lock(&last_ptr->lock);
4016                 if (last_ptr->block_group)
4017                         ffe_ctl->hint_byte = last_ptr->window_start;
4018                 if (last_ptr->fragmented) {
4019                         /*
4020                          * We still set window_start so we can keep track of the
4021                          * last place we found an allocation to try and save
4022                          * some time.
4023                          */
4024                         ffe_ctl->hint_byte = last_ptr->window_start;
4025                         ffe_ctl->use_cluster = false;
4026                 }
4027                 spin_unlock(&last_ptr->lock);
4028         }
4029
4030         return 0;
4031 }
4032
4033 static int prepare_allocation(struct btrfs_fs_info *fs_info,
4034                               struct find_free_extent_ctl *ffe_ctl,
4035                               struct btrfs_space_info *space_info,
4036                               struct btrfs_key *ins)
4037 {
4038         switch (ffe_ctl->policy) {
4039         case BTRFS_EXTENT_ALLOC_CLUSTERED:
4040                 return prepare_allocation_clustered(fs_info, ffe_ctl,
4041                                                     space_info, ins);
4042         default:
4043                 BUG();
4044         }
4045 }
4046
4047 /*
4048  * walks the btree of allocated extents and find a hole of a given size.
4049  * The key ins is changed to record the hole:
4050  * ins->objectid == start position
4051  * ins->flags = BTRFS_EXTENT_ITEM_KEY
4052  * ins->offset == the size of the hole.
4053  * Any available blocks before search_start are skipped.
4054  *
4055  * If there is no suitable free space, we will record the max size of
4056  * the free space extent currently.
4057  *
4058  * The overall logic and call chain:
4059  *
4060  * find_free_extent()
4061  * |- Iterate through all block groups
4062  * |  |- Get a valid block group
4063  * |  |- Try to do clustered allocation in that block group
4064  * |  |- Try to do unclustered allocation in that block group
4065  * |  |- Check if the result is valid
4066  * |  |  |- If valid, then exit
4067  * |  |- Jump to next block group
4068  * |
4069  * |- Push harder to find free extents
4070  *    |- If not found, re-iterate all block groups
4071  */
4072 static noinline int find_free_extent(struct btrfs_root *root,
4073                                 u64 ram_bytes, u64 num_bytes, u64 empty_size,
4074                                 u64 hint_byte_orig, struct btrfs_key *ins,
4075                                 u64 flags, int delalloc)
4076 {
4077         struct btrfs_fs_info *fs_info = root->fs_info;
4078         int ret = 0;
4079         int cache_block_group_error = 0;
4080         struct btrfs_block_group *block_group = NULL;
4081         struct find_free_extent_ctl ffe_ctl = {0};
4082         struct btrfs_space_info *space_info;
4083         bool full_search = false;
4084
4085         WARN_ON(num_bytes < fs_info->sectorsize);
4086
4087         ffe_ctl.num_bytes = num_bytes;
4088         ffe_ctl.empty_size = empty_size;
4089         ffe_ctl.flags = flags;
4090         ffe_ctl.search_start = 0;
4091         ffe_ctl.delalloc = delalloc;
4092         ffe_ctl.index = btrfs_bg_flags_to_raid_index(flags);
4093         ffe_ctl.have_caching_bg = false;
4094         ffe_ctl.orig_have_caching_bg = false;
4095         ffe_ctl.found_offset = 0;
4096         ffe_ctl.hint_byte = hint_byte_orig;
4097         ffe_ctl.policy = BTRFS_EXTENT_ALLOC_CLUSTERED;
4098
4099         /* For clustered allocation */
4100         ffe_ctl.retry_clustered = false;
4101         ffe_ctl.retry_unclustered = false;
4102         ffe_ctl.last_ptr = NULL;
4103         ffe_ctl.use_cluster = true;
4104
4105         ins->type = BTRFS_EXTENT_ITEM_KEY;
4106         ins->objectid = 0;
4107         ins->offset = 0;
4108
4109         trace_find_free_extent(root, num_bytes, empty_size, flags);
4110
4111         space_info = btrfs_find_space_info(fs_info, flags);
4112         if (!space_info) {
4113                 btrfs_err(fs_info, "No space info for %llu", flags);
4114                 return -ENOSPC;
4115         }
4116
4117         ret = prepare_allocation(fs_info, &ffe_ctl, space_info, ins);
4118         if (ret < 0)
4119                 return ret;
4120
4121         ffe_ctl.search_start = max(ffe_ctl.search_start,
4122                                    first_logical_byte(fs_info, 0));
4123         ffe_ctl.search_start = max(ffe_ctl.search_start, ffe_ctl.hint_byte);
4124         if (ffe_ctl.search_start == ffe_ctl.hint_byte) {
4125                 block_group = btrfs_lookup_block_group(fs_info,
4126                                                        ffe_ctl.search_start);
4127                 /*
4128                  * we don't want to use the block group if it doesn't match our
4129                  * allocation bits, or if its not cached.
4130                  *
4131                  * However if we are re-searching with an ideal block group
4132                  * picked out then we don't care that the block group is cached.
4133                  */
4134                 if (block_group && block_group_bits(block_group, flags) &&
4135                     block_group->cached != BTRFS_CACHE_NO) {
4136                         down_read(&space_info->groups_sem);
4137                         if (list_empty(&block_group->list) ||
4138                             block_group->ro) {
4139                                 /*
4140                                  * someone is removing this block group,
4141                                  * we can't jump into the have_block_group
4142                                  * target because our list pointers are not
4143                                  * valid
4144                                  */
4145                                 btrfs_put_block_group(block_group);
4146                                 up_read(&space_info->groups_sem);
4147                         } else {
4148                                 ffe_ctl.index = btrfs_bg_flags_to_raid_index(
4149                                                 block_group->flags);
4150                                 btrfs_lock_block_group(block_group, delalloc);
4151                                 goto have_block_group;
4152                         }
4153                 } else if (block_group) {
4154                         btrfs_put_block_group(block_group);
4155                 }
4156         }
4157 search:
4158         ffe_ctl.have_caching_bg = false;
4159         if (ffe_ctl.index == btrfs_bg_flags_to_raid_index(flags) ||
4160             ffe_ctl.index == 0)
4161                 full_search = true;
4162         down_read(&space_info->groups_sem);
4163         list_for_each_entry(block_group,
4164                             &space_info->block_groups[ffe_ctl.index], list) {
4165                 struct btrfs_block_group *bg_ret;
4166
4167                 /* If the block group is read-only, we can skip it entirely. */
4168                 if (unlikely(block_group->ro))
4169                         continue;
4170
4171                 btrfs_grab_block_group(block_group, delalloc);
4172                 ffe_ctl.search_start = block_group->start;
4173
4174                 /*
4175                  * this can happen if we end up cycling through all the
4176                  * raid types, but we want to make sure we only allocate
4177                  * for the proper type.
4178                  */
4179                 if (!block_group_bits(block_group, flags)) {
4180                         u64 extra = BTRFS_BLOCK_GROUP_DUP |
4181                                 BTRFS_BLOCK_GROUP_RAID1_MASK |
4182                                 BTRFS_BLOCK_GROUP_RAID56_MASK |
4183                                 BTRFS_BLOCK_GROUP_RAID10;
4184
4185                         /*
4186                          * if they asked for extra copies and this block group
4187                          * doesn't provide them, bail.  This does allow us to
4188                          * fill raid0 from raid1.
4189                          */
4190                         if ((flags & extra) && !(block_group->flags & extra))
4191                                 goto loop;
4192
4193                         /*
4194                          * This block group has different flags than we want.
4195                          * It's possible that we have MIXED_GROUP flag but no
4196                          * block group is mixed.  Just skip such block group.
4197                          */
4198                         btrfs_release_block_group(block_group, delalloc);
4199                         continue;
4200                 }
4201
4202 have_block_group:
4203                 ffe_ctl.cached = btrfs_block_group_done(block_group);
4204                 if (unlikely(!ffe_ctl.cached)) {
4205                         ffe_ctl.have_caching_bg = true;
4206                         ret = btrfs_cache_block_group(block_group, 0);
4207
4208                         /*
4209                          * If we get ENOMEM here or something else we want to
4210                          * try other block groups, because it may not be fatal.
4211                          * However if we can't find anything else we need to
4212                          * save our return here so that we return the actual
4213                          * error that caused problems, not ENOSPC.
4214                          */
4215                         if (ret < 0) {
4216                                 if (!cache_block_group_error)
4217                                         cache_block_group_error = ret;
4218                                 ret = 0;
4219                                 goto loop;
4220                         }
4221                         ret = 0;
4222                 }
4223
4224                 if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
4225                         goto loop;
4226
4227                 bg_ret = NULL;
4228                 ret = do_allocation(block_group, &ffe_ctl, &bg_ret);
4229                 if (ret == 0) {
4230                         if (bg_ret && bg_ret != block_group) {
4231                                 btrfs_release_block_group(block_group, delalloc);
4232                                 block_group = bg_ret;
4233                         }
4234                 } else if (ret == -EAGAIN) {
4235                         goto have_block_group;
4236                 } else if (ret > 0) {
4237                         goto loop;
4238                 }
4239
4240                 /* Checks */
4241                 ffe_ctl.search_start = round_up(ffe_ctl.found_offset,
4242                                              fs_info->stripesize);
4243
4244                 /* move on to the next group */
4245                 if (ffe_ctl.search_start + num_bytes >
4246                     block_group->start + block_group->length) {
4247                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4248                                              num_bytes);
4249                         goto loop;
4250                 }
4251
4252                 if (ffe_ctl.found_offset < ffe_ctl.search_start)
4253                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4254                                 ffe_ctl.search_start - ffe_ctl.found_offset);
4255
4256                 ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
4257                                 num_bytes, delalloc);
4258                 if (ret == -EAGAIN) {
4259                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4260                                              num_bytes);
4261                         goto loop;
4262                 }
4263                 btrfs_inc_block_group_reservations(block_group);
4264
4265                 /* we are all good, lets return */
4266                 ins->objectid = ffe_ctl.search_start;
4267                 ins->offset = num_bytes;
4268
4269                 trace_btrfs_reserve_extent(block_group, ffe_ctl.search_start,
4270                                            num_bytes);
4271                 btrfs_release_block_group(block_group, delalloc);
4272                 break;
4273 loop:
4274                 release_block_group(block_group, &ffe_ctl, delalloc);
4275                 cond_resched();
4276         }
4277         up_read(&space_info->groups_sem);
4278
4279         ret = find_free_extent_update_loop(fs_info, ins, &ffe_ctl, full_search);
4280         if (ret > 0)
4281                 goto search;
4282
4283         if (ret == -ENOSPC && !cache_block_group_error) {
4284                 /*
4285                  * Use ffe_ctl->total_free_space as fallback if we can't find
4286                  * any contiguous hole.
4287                  */
4288                 if (!ffe_ctl.max_extent_size)
4289                         ffe_ctl.max_extent_size = ffe_ctl.total_free_space;
4290                 spin_lock(&space_info->lock);
4291                 space_info->max_extent_size = ffe_ctl.max_extent_size;
4292                 spin_unlock(&space_info->lock);
4293                 ins->offset = ffe_ctl.max_extent_size;
4294         } else if (ret == -ENOSPC) {
4295                 ret = cache_block_group_error;
4296         }
4297         return ret;
4298 }
4299
4300 /*
4301  * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a
4302  *                        hole that is at least as big as @num_bytes.
4303  *
4304  * @root           -    The root that will contain this extent
4305  *
4306  * @ram_bytes      -    The amount of space in ram that @num_bytes take. This
4307  *                      is used for accounting purposes. This value differs
4308  *                      from @num_bytes only in the case of compressed extents.
4309  *
4310  * @num_bytes      -    Number of bytes to allocate on-disk.
4311  *
4312  * @min_alloc_size -    Indicates the minimum amount of space that the
4313  *                      allocator should try to satisfy. In some cases
4314  *                      @num_bytes may be larger than what is required and if
4315  *                      the filesystem is fragmented then allocation fails.
4316  *                      However, the presence of @min_alloc_size gives a
4317  *                      chance to try and satisfy the smaller allocation.
4318  *
4319  * @empty_size     -    A hint that you plan on doing more COW. This is the
4320  *                      size in bytes the allocator should try to find free
4321  *                      next to the block it returns.  This is just a hint and
4322  *                      may be ignored by the allocator.
4323  *
4324  * @hint_byte      -    Hint to the allocator to start searching above the byte
4325  *                      address passed. It might be ignored.
4326  *
4327  * @ins            -    This key is modified to record the found hole. It will
4328  *                      have the following values:
4329  *                      ins->objectid == start position
4330  *                      ins->flags = BTRFS_EXTENT_ITEM_KEY
4331  *                      ins->offset == the size of the hole.
4332  *
4333  * @is_data        -    Boolean flag indicating whether an extent is
4334  *                      allocated for data (true) or metadata (false)
4335  *
4336  * @delalloc       -    Boolean flag indicating whether this allocation is for
4337  *                      delalloc or not. If 'true' data_rwsem of block groups
4338  *                      is going to be acquired.
4339  *
4340  *
4341  * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
4342  * case -ENOSPC is returned then @ins->offset will contain the size of the
4343  * largest available hole the allocator managed to find.
4344  */
4345 int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
4346                          u64 num_bytes, u64 min_alloc_size,
4347                          u64 empty_size, u64 hint_byte,
4348                          struct btrfs_key *ins, int is_data, int delalloc)
4349 {
4350         struct btrfs_fs_info *fs_info = root->fs_info;
4351         bool final_tried = num_bytes == min_alloc_size;
4352         u64 flags;
4353         int ret;
4354
4355         flags = get_alloc_profile_by_root(root, is_data);
4356 again:
4357         WARN_ON(num_bytes < fs_info->sectorsize);
4358         ret = find_free_extent(root, ram_bytes, num_bytes, empty_size,
4359                                hint_byte, ins, flags, delalloc);
4360         if (!ret && !is_data) {
4361                 btrfs_dec_block_group_reservations(fs_info, ins->objectid);
4362         } else if (ret == -ENOSPC) {
4363                 if (!final_tried && ins->offset) {
4364                         num_bytes = min(num_bytes >> 1, ins->offset);
4365                         num_bytes = round_down(num_bytes,
4366                                                fs_info->sectorsize);
4367                         num_bytes = max(num_bytes, min_alloc_size);
4368                         ram_bytes = num_bytes;
4369                         if (num_bytes == min_alloc_size)
4370                                 final_tried = true;
4371                         goto again;
4372                 } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
4373                         struct btrfs_space_info *sinfo;
4374
4375                         sinfo = btrfs_find_space_info(fs_info, flags);
4376                         btrfs_err(fs_info,
4377                                   "allocation failed flags %llu, wanted %llu",
4378                                   flags, num_bytes);
4379                         if (sinfo)
4380                                 btrfs_dump_space_info(fs_info, sinfo,
4381                                                       num_bytes, 1);
4382                 }
4383         }
4384
4385         return ret;
4386 }
4387
4388 int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
4389                                u64 start, u64 len, int delalloc)
4390 {
4391         struct btrfs_block_group *cache;
4392
4393         cache = btrfs_lookup_block_group(fs_info, start);
4394         if (!cache) {
4395                 btrfs_err(fs_info, "Unable to find block group for %llu",
4396                           start);
4397                 return -ENOSPC;
4398         }
4399
4400         btrfs_add_free_space(cache, start, len);
4401         btrfs_free_reserved_bytes(cache, len, delalloc);
4402         trace_btrfs_reserved_extent_free(fs_info, start, len);
4403
4404         btrfs_put_block_group(cache);
4405         return 0;
4406 }
4407
4408 int btrfs_pin_reserved_extent(struct btrfs_trans_handle *trans, u64 start,
4409                               u64 len)
4410 {
4411         struct btrfs_block_group *cache;
4412         int ret = 0;
4413
4414         cache = btrfs_lookup_block_group(trans->fs_info, start);
4415         if (!cache) {
4416                 btrfs_err(trans->fs_info, "unable to find block group for %llu",
4417                           start);
4418                 return -ENOSPC;
4419         }
4420
4421         ret = pin_down_extent(trans, cache, start, len, 1);
4422         btrfs_put_block_group(cache);
4423         return ret;
4424 }
4425
4426 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4427                                       u64 parent, u64 root_objectid,
4428                                       u64 flags, u64 owner, u64 offset,
4429                                       struct btrfs_key *ins, int ref_mod)
4430 {
4431         struct btrfs_fs_info *fs_info = trans->fs_info;
4432         int ret;
4433         struct btrfs_extent_item *extent_item;
4434         struct btrfs_extent_inline_ref *iref;
4435         struct btrfs_path *path;
4436         struct extent_buffer *leaf;
4437         int type;
4438         u32 size;
4439
4440         if (parent > 0)
4441                 type = BTRFS_SHARED_DATA_REF_KEY;
4442         else
4443                 type = BTRFS_EXTENT_DATA_REF_KEY;
4444
4445         size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4446
4447         path = btrfs_alloc_path();
4448         if (!path)
4449                 return -ENOMEM;
4450
4451         path->leave_spinning = 1;
4452         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4453                                       ins, size);
4454         if (ret) {
4455                 btrfs_free_path(path);
4456                 return ret;
4457         }
4458
4459         leaf = path->nodes[0];
4460         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4461                                      struct btrfs_extent_item);
4462         btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4463         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4464         btrfs_set_extent_flags(leaf, extent_item,
4465                                flags | BTRFS_EXTENT_FLAG_DATA);
4466
4467         iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4468         btrfs_set_extent_inline_ref_type(leaf, iref, type);
4469         if (parent > 0) {
4470                 struct btrfs_shared_data_ref *ref;
4471                 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4472                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4473                 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4474         } else {
4475                 struct btrfs_extent_data_ref *ref;
4476                 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4477                 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4478                 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4479                 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4480                 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4481         }
4482
4483         btrfs_mark_buffer_dirty(path->nodes[0]);
4484         btrfs_free_path(path);
4485
4486         ret = remove_from_free_space_tree(trans, ins->objectid, ins->offset);
4487         if (ret)
4488                 return ret;
4489
4490         ret = btrfs_update_block_group(trans, ins->objectid, ins->offset, 1);
4491         if (ret) { /* -ENOENT, logic error */
4492                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4493                         ins->objectid, ins->offset);
4494                 BUG();
4495         }
4496         trace_btrfs_reserved_extent_alloc(fs_info, ins->objectid, ins->offset);
4497         return ret;
4498 }
4499
4500 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4501                                      struct btrfs_delayed_ref_node *node,
4502                                      struct btrfs_delayed_extent_op *extent_op)
4503 {
4504         struct btrfs_fs_info *fs_info = trans->fs_info;
4505         int ret;
4506         struct btrfs_extent_item *extent_item;
4507         struct btrfs_key extent_key;
4508         struct btrfs_tree_block_info *block_info;
4509         struct btrfs_extent_inline_ref *iref;
4510         struct btrfs_path *path;
4511         struct extent_buffer *leaf;
4512         struct btrfs_delayed_tree_ref *ref;
4513         u32 size = sizeof(*extent_item) + sizeof(*iref);
4514         u64 num_bytes;
4515         u64 flags = extent_op->flags_to_set;
4516         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4517
4518         ref = btrfs_delayed_node_to_tree_ref(node);
4519
4520         extent_key.objectid = node->bytenr;
4521         if (skinny_metadata) {
4522                 extent_key.offset = ref->level;
4523                 extent_key.type = BTRFS_METADATA_ITEM_KEY;
4524                 num_bytes = fs_info->nodesize;
4525         } else {
4526                 extent_key.offset = node->num_bytes;
4527                 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4528                 size += sizeof(*block_info);
4529                 num_bytes = node->num_bytes;
4530         }
4531
4532         path = btrfs_alloc_path();
4533         if (!path)
4534                 return -ENOMEM;
4535
4536         path->leave_spinning = 1;
4537         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4538                                       &extent_key, size);
4539         if (ret) {
4540                 btrfs_free_path(path);
4541                 return ret;
4542         }
4543
4544         leaf = path->nodes[0];
4545         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4546                                      struct btrfs_extent_item);
4547         btrfs_set_extent_refs(leaf, extent_item, 1);
4548         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4549         btrfs_set_extent_flags(leaf, extent_item,
4550                                flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4551
4552         if (skinny_metadata) {
4553                 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4554         } else {
4555                 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4556                 btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
4557                 btrfs_set_tree_block_level(leaf, block_info, ref->level);
4558                 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4559         }
4560
4561         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
4562                 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4563                 btrfs_set_extent_inline_ref_type(leaf, iref,
4564                                                  BTRFS_SHARED_BLOCK_REF_KEY);
4565                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent);
4566         } else {
4567                 btrfs_set_extent_inline_ref_type(leaf, iref,
4568                                                  BTRFS_TREE_BLOCK_REF_KEY);
4569                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root);
4570         }
4571
4572         btrfs_mark_buffer_dirty(leaf);
4573         btrfs_free_path(path);
4574
4575         ret = remove_from_free_space_tree(trans, extent_key.objectid,
4576                                           num_bytes);
4577         if (ret)
4578                 return ret;
4579
4580         ret = btrfs_update_block_group(trans, extent_key.objectid,
4581                                        fs_info->nodesize, 1);
4582         if (ret) { /* -ENOENT, logic error */
4583                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4584                         extent_key.objectid, extent_key.offset);
4585                 BUG();
4586         }
4587
4588         trace_btrfs_reserved_extent_alloc(fs_info, extent_key.objectid,
4589                                           fs_info->nodesize);
4590         return ret;
4591 }
4592
4593 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4594                                      struct btrfs_root *root, u64 owner,
4595                                      u64 offset, u64 ram_bytes,
4596                                      struct btrfs_key *ins)
4597 {
4598         struct btrfs_ref generic_ref = { 0 };
4599         int ret;
4600
4601         BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
4602
4603         btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4604                                ins->objectid, ins->offset, 0);
4605         btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
4606         btrfs_ref_tree_mod(root->fs_info, &generic_ref);
4607         ret = btrfs_add_delayed_data_ref(trans, &generic_ref,
4608                                          ram_bytes, NULL, NULL);
4609         return ret;
4610 }
4611
4612 /*
4613  * this is used by the tree logging recovery code.  It records that
4614  * an extent has been allocated and makes sure to clear the free
4615  * space cache bits as well
4616  */
4617 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4618                                    u64 root_objectid, u64 owner, u64 offset,
4619                                    struct btrfs_key *ins)
4620 {
4621         struct btrfs_fs_info *fs_info = trans->fs_info;
4622         int ret;
4623         struct btrfs_block_group *block_group;
4624         struct btrfs_space_info *space_info;
4625
4626         /*
4627          * Mixed block groups will exclude before processing the log so we only
4628          * need to do the exclude dance if this fs isn't mixed.
4629          */
4630         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
4631                 ret = __exclude_logged_extent(fs_info, ins->objectid,
4632                                               ins->offset);
4633                 if (ret)
4634                         return ret;
4635         }
4636
4637         block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
4638         if (!block_group)
4639                 return -EINVAL;
4640
4641         space_info = block_group->space_info;
4642         spin_lock(&space_info->lock);
4643         spin_lock(&block_group->lock);
4644         space_info->bytes_reserved += ins->offset;
4645         block_group->reserved += ins->offset;
4646         spin_unlock(&block_group->lock);
4647         spin_unlock(&space_info->lock);
4648
4649         ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
4650                                          offset, ins, 1);
4651         if (ret)
4652                 btrfs_pin_extent(trans, ins->objectid, ins->offset, 1);
4653         btrfs_put_block_group(block_group);
4654         return ret;
4655 }
4656
4657 static struct extent_buffer *
4658 btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4659                       u64 bytenr, int level, u64 owner,
4660                       enum btrfs_lock_nesting nest)
4661 {
4662         struct btrfs_fs_info *fs_info = root->fs_info;
4663         struct extent_buffer *buf;
4664
4665         buf = btrfs_find_create_tree_block(fs_info, bytenr);
4666         if (IS_ERR(buf))
4667                 return buf;
4668
4669         /*
4670          * Extra safety check in case the extent tree is corrupted and extent
4671          * allocator chooses to use a tree block which is already used and
4672          * locked.
4673          */
4674         if (buf->lock_owner == current->pid) {
4675                 btrfs_err_rl(fs_info,
4676 "tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
4677                         buf->start, btrfs_header_owner(buf), current->pid);
4678                 free_extent_buffer(buf);
4679                 return ERR_PTR(-EUCLEAN);
4680         }
4681
4682         btrfs_set_buffer_lockdep_class(owner, buf, level);
4683         __btrfs_tree_lock(buf, nest);
4684         btrfs_clean_tree_block(buf);
4685         clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
4686
4687         btrfs_set_lock_blocking_write(buf);
4688         set_extent_buffer_uptodate(buf);
4689
4690         memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
4691         btrfs_set_header_level(buf, level);
4692         btrfs_set_header_bytenr(buf, buf->start);
4693         btrfs_set_header_generation(buf, trans->transid);
4694         btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
4695         btrfs_set_header_owner(buf, owner);
4696         write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
4697         write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
4698         if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4699                 buf->log_index = root->log_transid % 2;
4700                 /*
4701                  * we allow two log transactions at a time, use different
4702                  * EXTENT bit to differentiate dirty pages.
4703                  */
4704                 if (buf->log_index == 0)
4705                         set_extent_dirty(&root->dirty_log_pages, buf->start,
4706                                         buf->start + buf->len - 1, GFP_NOFS);
4707                 else
4708                         set_extent_new(&root->dirty_log_pages, buf->start,
4709                                         buf->start + buf->len - 1);
4710         } else {
4711                 buf->log_index = -1;
4712                 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4713                          buf->start + buf->len - 1, GFP_NOFS);
4714         }
4715         trans->dirty = true;
4716         /* this returns a buffer locked for blocking */
4717         return buf;
4718 }
4719
4720 /*
4721  * finds a free extent and does all the dirty work required for allocation
4722  * returns the tree buffer or an ERR_PTR on error.
4723  */
4724 struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
4725                                              struct btrfs_root *root,
4726                                              u64 parent, u64 root_objectid,
4727                                              const struct btrfs_disk_key *key,
4728                                              int level, u64 hint,
4729                                              u64 empty_size,
4730                                              enum btrfs_lock_nesting nest)
4731 {
4732         struct btrfs_fs_info *fs_info = root->fs_info;
4733         struct btrfs_key ins;
4734         struct btrfs_block_rsv *block_rsv;
4735         struct extent_buffer *buf;
4736         struct btrfs_delayed_extent_op *extent_op;
4737         struct btrfs_ref generic_ref = { 0 };
4738         u64 flags = 0;
4739         int ret;
4740         u32 blocksize = fs_info->nodesize;
4741         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4742
4743 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4744         if (btrfs_is_testing(fs_info)) {
4745                 buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
4746                                             level, root_objectid, nest);
4747                 if (!IS_ERR(buf))
4748                         root->alloc_bytenr += blocksize;
4749                 return buf;
4750         }
4751 #endif
4752
4753         block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
4754         if (IS_ERR(block_rsv))
4755                 return ERR_CAST(block_rsv);
4756
4757         ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
4758                                    empty_size, hint, &ins, 0, 0);
4759         if (ret)
4760                 goto out_unuse;
4761
4762         buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
4763                                     root_objectid, nest);
4764         if (IS_ERR(buf)) {
4765                 ret = PTR_ERR(buf);
4766                 goto out_free_reserved;
4767         }
4768
4769         if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4770                 if (parent == 0)
4771                         parent = ins.objectid;
4772                 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4773         } else
4774                 BUG_ON(parent > 0);
4775
4776         if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4777                 extent_op = btrfs_alloc_delayed_extent_op();
4778                 if (!extent_op) {
4779                         ret = -ENOMEM;
4780                         goto out_free_buf;
4781                 }
4782                 if (key)
4783                         memcpy(&extent_op->key, key, sizeof(extent_op->key));
4784                 else
4785                         memset(&extent_op->key, 0, sizeof(extent_op->key));
4786                 extent_op->flags_to_set = flags;
4787                 extent_op->update_key = skinny_metadata ? false : true;
4788                 extent_op->update_flags = true;
4789                 extent_op->is_data = false;
4790                 extent_op->level = level;
4791
4792                 btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4793                                        ins.objectid, ins.offset, parent);
4794                 generic_ref.real_root = root->root_key.objectid;
4795                 btrfs_init_tree_ref(&generic_ref, level, root_objectid);
4796                 btrfs_ref_tree_mod(fs_info, &generic_ref);
4797                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref,
4798                                                  extent_op, NULL, NULL);
4799                 if (ret)
4800                         goto out_free_delayed;
4801         }
4802         return buf;
4803
4804 out_free_delayed:
4805         btrfs_free_delayed_extent_op(extent_op);
4806 out_free_buf:
4807         free_extent_buffer(buf);
4808 out_free_reserved:
4809         btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
4810 out_unuse:
4811         btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
4812         return ERR_PTR(ret);
4813 }
4814
4815 struct walk_control {
4816         u64 refs[BTRFS_MAX_LEVEL];
4817         u64 flags[BTRFS_MAX_LEVEL];
4818         struct btrfs_key update_progress;
4819         struct btrfs_key drop_progress;
4820         int drop_level;
4821         int stage;
4822         int level;
4823         int shared_level;
4824         int update_ref;
4825         int keep_locks;
4826         int reada_slot;
4827         int reada_count;
4828         int restarted;
4829 };
4830
4831 #define DROP_REFERENCE  1
4832 #define UPDATE_BACKREF  2
4833
4834 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4835                                      struct btrfs_root *root,
4836                                      struct walk_control *wc,
4837                                      struct btrfs_path *path)
4838 {
4839         struct btrfs_fs_info *fs_info = root->fs_info;
4840         u64 bytenr;
4841         u64 generation;
4842         u64 refs;
4843         u64 flags;
4844         u32 nritems;
4845         struct btrfs_key key;
4846         struct extent_buffer *eb;
4847         int ret;
4848         int slot;
4849         int nread = 0;
4850
4851         if (path->slots[wc->level] < wc->reada_slot) {
4852                 wc->reada_count = wc->reada_count * 2 / 3;
4853                 wc->reada_count = max(wc->reada_count, 2);
4854         } else {
4855                 wc->reada_count = wc->reada_count * 3 / 2;
4856                 wc->reada_count = min_t(int, wc->reada_count,
4857                                         BTRFS_NODEPTRS_PER_BLOCK(fs_info));
4858         }
4859
4860         eb = path->nodes[wc->level];
4861         nritems = btrfs_header_nritems(eb);
4862
4863         for (slot = path->slots[wc->level]; slot < nritems; slot++) {
4864                 if (nread >= wc->reada_count)
4865                         break;
4866
4867                 cond_resched();
4868                 bytenr = btrfs_node_blockptr(eb, slot);
4869                 generation = btrfs_node_ptr_generation(eb, slot);
4870
4871                 if (slot == path->slots[wc->level])
4872                         goto reada;
4873
4874                 if (wc->stage == UPDATE_BACKREF &&
4875                     generation <= root->root_key.offset)
4876                         continue;
4877
4878                 /* We don't lock the tree block, it's OK to be racy here */
4879                 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
4880                                                wc->level - 1, 1, &refs,
4881                                                &flags);
4882                 /* We don't care about errors in readahead. */
4883                 if (ret < 0)
4884                         continue;
4885                 BUG_ON(refs == 0);
4886
4887                 if (wc->stage == DROP_REFERENCE) {
4888                         if (refs == 1)
4889                                 goto reada;
4890
4891                         if (wc->level == 1 &&
4892                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4893                                 continue;
4894                         if (!wc->update_ref ||
4895                             generation <= root->root_key.offset)
4896                                 continue;
4897                         btrfs_node_key_to_cpu(eb, &key, slot);
4898                         ret = btrfs_comp_cpu_keys(&key,
4899                                                   &wc->update_progress);
4900                         if (ret < 0)
4901                                 continue;
4902                 } else {
4903                         if (wc->level == 1 &&
4904                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4905                                 continue;
4906                 }
4907 reada:
4908                 readahead_tree_block(fs_info, bytenr);
4909                 nread++;
4910         }
4911         wc->reada_slot = slot;
4912 }
4913
4914 /*
4915  * helper to process tree block while walking down the tree.
4916  *
4917  * when wc->stage == UPDATE_BACKREF, this function updates
4918  * back refs for pointers in the block.
4919  *
4920  * NOTE: return value 1 means we should stop walking down.
4921  */
4922 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4923                                    struct btrfs_root *root,
4924                                    struct btrfs_path *path,
4925                                    struct walk_control *wc, int lookup_info)
4926 {
4927         struct btrfs_fs_info *fs_info = root->fs_info;
4928         int level = wc->level;
4929         struct extent_buffer *eb = path->nodes[level];
4930         u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
4931         int ret;
4932
4933         if (wc->stage == UPDATE_BACKREF &&
4934             btrfs_header_owner(eb) != root->root_key.objectid)
4935                 return 1;
4936
4937         /*
4938          * when reference count of tree block is 1, it won't increase
4939          * again. once full backref flag is set, we never clear it.
4940          */
4941         if (lookup_info &&
4942             ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
4943              (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
4944                 BUG_ON(!path->locks[level]);
4945                 ret = btrfs_lookup_extent_info(trans, fs_info,
4946                                                eb->start, level, 1,
4947                                                &wc->refs[level],
4948                                                &wc->flags[level]);
4949                 BUG_ON(ret == -ENOMEM);
4950                 if (ret)
4951                         return ret;
4952                 BUG_ON(wc->refs[level] == 0);
4953         }
4954
4955         if (wc->stage == DROP_REFERENCE) {
4956                 if (wc->refs[level] > 1)
4957                         return 1;
4958
4959                 if (path->locks[level] && !wc->keep_locks) {
4960                         btrfs_tree_unlock_rw(eb, path->locks[level]);
4961                         path->locks[level] = 0;
4962                 }
4963                 return 0;
4964         }
4965
4966         /* wc->stage == UPDATE_BACKREF */
4967         if (!(wc->flags[level] & flag)) {
4968                 BUG_ON(!path->locks[level]);
4969                 ret = btrfs_inc_ref(trans, root, eb, 1);
4970                 BUG_ON(ret); /* -ENOMEM */
4971                 ret = btrfs_dec_ref(trans, root, eb, 0);
4972                 BUG_ON(ret); /* -ENOMEM */
4973                 ret = btrfs_set_disk_extent_flags(trans, eb, flag,
4974                                                   btrfs_header_level(eb), 0);
4975                 BUG_ON(ret); /* -ENOMEM */
4976                 wc->flags[level] |= flag;
4977         }
4978
4979         /*
4980          * the block is shared by multiple trees, so it's not good to
4981          * keep the tree lock
4982          */
4983         if (path->locks[level] && level > 0) {
4984                 btrfs_tree_unlock_rw(eb, path->locks[level]);
4985                 path->locks[level] = 0;
4986         }
4987         return 0;
4988 }
4989
4990 /*
4991  * This is used to verify a ref exists for this root to deal with a bug where we
4992  * would have a drop_progress key that hadn't been updated properly.
4993  */
4994 static int check_ref_exists(struct btrfs_trans_handle *trans,
4995                             struct btrfs_root *root, u64 bytenr, u64 parent,
4996                             int level)
4997 {
4998         struct btrfs_path *path;
4999         struct btrfs_extent_inline_ref *iref;
5000         int ret;
5001
5002         path = btrfs_alloc_path();
5003         if (!path)
5004                 return -ENOMEM;
5005
5006         ret = lookup_extent_backref(trans, path, &iref, bytenr,
5007                                     root->fs_info->nodesize, parent,
5008                                     root->root_key.objectid, level, 0);
5009         btrfs_free_path(path);
5010         if (ret == -ENOENT)
5011                 return 0;
5012         if (ret < 0)
5013                 return ret;
5014         return 1;
5015 }
5016
5017 /*
5018  * helper to process tree block pointer.
5019  *
5020  * when wc->stage == DROP_REFERENCE, this function checks
5021  * reference count of the block pointed to. if the block
5022  * is shared and we need update back refs for the subtree
5023  * rooted at the block, this function changes wc->stage to
5024  * UPDATE_BACKREF. if the block is shared and there is no
5025  * need to update back, this function drops the reference
5026  * to the block.
5027  *
5028  * NOTE: return value 1 means we should stop walking down.
5029  */
5030 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5031                                  struct btrfs_root *root,
5032                                  struct btrfs_path *path,
5033                                  struct walk_control *wc, int *lookup_info)
5034 {
5035         struct btrfs_fs_info *fs_info = root->fs_info;
5036         u64 bytenr;
5037         u64 generation;
5038         u64 parent;
5039         struct btrfs_key key;
5040         struct btrfs_key first_key;
5041         struct btrfs_ref ref = { 0 };
5042         struct extent_buffer *next;
5043         int level = wc->level;
5044         int reada = 0;
5045         int ret = 0;
5046         bool need_account = false;
5047
5048         generation = btrfs_node_ptr_generation(path->nodes[level],
5049                                                path->slots[level]);
5050         /*
5051          * if the lower level block was created before the snapshot
5052          * was created, we know there is no need to update back refs
5053          * for the subtree
5054          */
5055         if (wc->stage == UPDATE_BACKREF &&
5056             generation <= root->root_key.offset) {
5057                 *lookup_info = 1;
5058                 return 1;
5059         }
5060
5061         bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
5062         btrfs_node_key_to_cpu(path->nodes[level], &first_key,
5063                               path->slots[level]);
5064
5065         next = find_extent_buffer(fs_info, bytenr);
5066         if (!next) {
5067                 next = btrfs_find_create_tree_block(fs_info, bytenr);
5068                 if (IS_ERR(next))
5069                         return PTR_ERR(next);
5070
5071                 btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
5072                                                level - 1);
5073                 reada = 1;
5074         }
5075         btrfs_tree_lock(next);
5076         btrfs_set_lock_blocking_write(next);
5077
5078         ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
5079                                        &wc->refs[level - 1],
5080                                        &wc->flags[level - 1]);
5081         if (ret < 0)
5082                 goto out_unlock;
5083
5084         if (unlikely(wc->refs[level - 1] == 0)) {
5085                 btrfs_err(fs_info, "Missing references.");
5086                 ret = -EIO;
5087                 goto out_unlock;
5088         }
5089         *lookup_info = 0;
5090
5091         if (wc->stage == DROP_REFERENCE) {
5092                 if (wc->refs[level - 1] > 1) {
5093                         need_account = true;
5094                         if (level == 1 &&
5095                             (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5096                                 goto skip;
5097
5098                         if (!wc->update_ref ||
5099                             generation <= root->root_key.offset)
5100                                 goto skip;
5101
5102                         btrfs_node_key_to_cpu(path->nodes[level], &key,
5103                                               path->slots[level]);
5104                         ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
5105                         if (ret < 0)
5106                                 goto skip;
5107
5108                         wc->stage = UPDATE_BACKREF;
5109                         wc->shared_level = level - 1;
5110                 }
5111         } else {
5112                 if (level == 1 &&
5113                     (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5114                         goto skip;
5115         }
5116
5117         if (!btrfs_buffer_uptodate(next, generation, 0)) {
5118                 btrfs_tree_unlock(next);
5119                 free_extent_buffer(next);
5120                 next = NULL;
5121                 *lookup_info = 1;
5122         }
5123
5124         if (!next) {
5125                 if (reada && level == 1)
5126                         reada_walk_down(trans, root, wc, path);
5127                 next = read_tree_block(fs_info, bytenr, generation, level - 1,
5128                                        &first_key);
5129                 if (IS_ERR(next)) {
5130                         return PTR_ERR(next);
5131                 } else if (!extent_buffer_uptodate(next)) {
5132                         free_extent_buffer(next);
5133                         return -EIO;
5134                 }
5135                 btrfs_tree_lock(next);
5136                 btrfs_set_lock_blocking_write(next);
5137         }
5138
5139         level--;
5140         ASSERT(level == btrfs_header_level(next));
5141         if (level != btrfs_header_level(next)) {
5142                 btrfs_err(root->fs_info, "mismatched level");
5143                 ret = -EIO;
5144                 goto out_unlock;
5145         }
5146         path->nodes[level] = next;
5147         path->slots[level] = 0;
5148         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5149         wc->level = level;
5150         if (wc->level == 1)
5151                 wc->reada_slot = 0;
5152         return 0;
5153 skip:
5154         wc->refs[level - 1] = 0;
5155         wc->flags[level - 1] = 0;
5156         if (wc->stage == DROP_REFERENCE) {
5157                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5158                         parent = path->nodes[level]->start;
5159                 } else {
5160                         ASSERT(root->root_key.objectid ==
5161                                btrfs_header_owner(path->nodes[level]));
5162                         if (root->root_key.objectid !=
5163                             btrfs_header_owner(path->nodes[level])) {
5164                                 btrfs_err(root->fs_info,
5165                                                 "mismatched block owner");
5166                                 ret = -EIO;
5167                                 goto out_unlock;
5168                         }
5169                         parent = 0;
5170                 }
5171
5172                 /*
5173                  * If we had a drop_progress we need to verify the refs are set
5174                  * as expected.  If we find our ref then we know that from here
5175                  * on out everything should be correct, and we can clear the
5176                  * ->restarted flag.
5177                  */
5178                 if (wc->restarted) {
5179                         ret = check_ref_exists(trans, root, bytenr, parent,
5180                                                level - 1);
5181                         if (ret < 0)
5182                                 goto out_unlock;
5183                         if (ret == 0)
5184                                 goto no_delete;
5185                         ret = 0;
5186                         wc->restarted = 0;
5187                 }
5188
5189                 /*
5190                  * Reloc tree doesn't contribute to qgroup numbers, and we have
5191                  * already accounted them at merge time (replace_path),
5192                  * thus we could skip expensive subtree trace here.
5193                  */
5194                 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
5195                     need_account) {
5196                         ret = btrfs_qgroup_trace_subtree(trans, next,
5197                                                          generation, level - 1);
5198                         if (ret) {
5199                                 btrfs_err_rl(fs_info,
5200                                              "Error %d accounting shared subtree. Quota is out of sync, rescan required.",
5201                                              ret);
5202                         }
5203                 }
5204
5205                 /*
5206                  * We need to update the next key in our walk control so we can
5207                  * update the drop_progress key accordingly.  We don't care if
5208                  * find_next_key doesn't find a key because that means we're at
5209                  * the end and are going to clean up now.
5210                  */
5211                 wc->drop_level = level;
5212                 find_next_key(path, level, &wc->drop_progress);
5213
5214                 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
5215                                        fs_info->nodesize, parent);
5216                 btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
5217                 ret = btrfs_free_extent(trans, &ref);
5218                 if (ret)
5219                         goto out_unlock;
5220         }
5221 no_delete:
5222         *lookup_info = 1;
5223         ret = 1;
5224
5225 out_unlock:
5226         btrfs_tree_unlock(next);
5227         free_extent_buffer(next);
5228
5229         return ret;
5230 }
5231
5232 /*
5233  * helper to process tree block while walking up the tree.
5234  *
5235  * when wc->stage == DROP_REFERENCE, this function drops
5236  * reference count on the block.
5237  *
5238  * when wc->stage == UPDATE_BACKREF, this function changes
5239  * wc->stage back to DROP_REFERENCE if we changed wc->stage
5240  * to UPDATE_BACKREF previously while processing the block.
5241  *
5242  * NOTE: return value 1 means we should stop walking up.
5243  */
5244 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5245                                  struct btrfs_root *root,
5246                                  struct btrfs_path *path,
5247                                  struct walk_control *wc)
5248 {
5249         struct btrfs_fs_info *fs_info = root->fs_info;
5250         int ret;
5251         int level = wc->level;
5252         struct extent_buffer *eb = path->nodes[level];
5253         u64 parent = 0;
5254
5255         if (wc->stage == UPDATE_BACKREF) {
5256                 BUG_ON(wc->shared_level < level);
5257                 if (level < wc->shared_level)
5258                         goto out;
5259
5260                 ret = find_next_key(path, level + 1, &wc->update_progress);
5261                 if (ret > 0)
5262                         wc->update_ref = 0;
5263
5264                 wc->stage = DROP_REFERENCE;
5265                 wc->shared_level = -1;
5266                 path->slots[level] = 0;
5267
5268                 /*
5269                  * check reference count again if the block isn't locked.
5270                  * we should start walking down the tree again if reference
5271                  * count is one.
5272                  */
5273                 if (!path->locks[level]) {
5274                         BUG_ON(level == 0);
5275                         btrfs_tree_lock(eb);
5276                         btrfs_set_lock_blocking_write(eb);
5277                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5278
5279                         ret = btrfs_lookup_extent_info(trans, fs_info,
5280                                                        eb->start, level, 1,
5281                                                        &wc->refs[level],
5282                                                        &wc->flags[level]);
5283                         if (ret < 0) {
5284                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5285                                 path->locks[level] = 0;
5286                                 return ret;
5287                         }
5288                         BUG_ON(wc->refs[level] == 0);
5289                         if (wc->refs[level] == 1) {
5290                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5291                                 path->locks[level] = 0;
5292                                 return 1;
5293                         }
5294                 }
5295         }
5296
5297         /* wc->stage == DROP_REFERENCE */
5298         BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5299
5300         if (wc->refs[level] == 1) {
5301                 if (level == 0) {
5302                         if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5303                                 ret = btrfs_dec_ref(trans, root, eb, 1);
5304                         else
5305                                 ret = btrfs_dec_ref(trans, root, eb, 0);
5306                         BUG_ON(ret); /* -ENOMEM */
5307                         if (is_fstree(root->root_key.objectid)) {
5308                                 ret = btrfs_qgroup_trace_leaf_items(trans, eb);
5309                                 if (ret) {
5310                                         btrfs_err_rl(fs_info,
5311         "error %d accounting leaf items, quota is out of sync, rescan required",
5312                                              ret);
5313                                 }
5314                         }
5315                 }
5316                 /* make block locked assertion in btrfs_clean_tree_block happy */
5317                 if (!path->locks[level] &&
5318                     btrfs_header_generation(eb) == trans->transid) {
5319                         btrfs_tree_lock(eb);
5320                         btrfs_set_lock_blocking_write(eb);
5321                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5322                 }
5323                 btrfs_clean_tree_block(eb);
5324         }
5325
5326         if (eb == root->node) {
5327                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5328                         parent = eb->start;
5329                 else if (root->root_key.objectid != btrfs_header_owner(eb))
5330                         goto owner_mismatch;
5331         } else {
5332                 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5333                         parent = path->nodes[level + 1]->start;
5334                 else if (root->root_key.objectid !=
5335                          btrfs_header_owner(path->nodes[level + 1]))
5336                         goto owner_mismatch;
5337         }
5338
5339         btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
5340 out:
5341         wc->refs[level] = 0;
5342         wc->flags[level] = 0;
5343         return 0;
5344
5345 owner_mismatch:
5346         btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
5347                      btrfs_header_owner(eb), root->root_key.objectid);
5348         return -EUCLEAN;
5349 }
5350
5351 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5352                                    struct btrfs_root *root,
5353                                    struct btrfs_path *path,
5354                                    struct walk_control *wc)
5355 {
5356         int level = wc->level;
5357         int lookup_info = 1;
5358         int ret;
5359
5360         while (level >= 0) {
5361                 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5362                 if (ret > 0)
5363                         break;
5364
5365                 if (level == 0)
5366                         break;
5367
5368                 if (path->slots[level] >=
5369                     btrfs_header_nritems(path->nodes[level]))
5370                         break;
5371
5372                 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5373                 if (ret > 0) {
5374                         path->slots[level]++;
5375                         continue;
5376                 } else if (ret < 0)
5377                         return ret;
5378                 level = wc->level;
5379         }
5380         return 0;
5381 }
5382
5383 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5384                                  struct btrfs_root *root,
5385                                  struct btrfs_path *path,
5386                                  struct walk_control *wc, int max_level)
5387 {
5388         int level = wc->level;
5389         int ret;
5390
5391         path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5392         while (level < max_level && path->nodes[level]) {
5393                 wc->level = level;
5394                 if (path->slots[level] + 1 <
5395                     btrfs_header_nritems(path->nodes[level])) {
5396                         path->slots[level]++;
5397                         return 0;
5398                 } else {
5399                         ret = walk_up_proc(trans, root, path, wc);
5400                         if (ret > 0)
5401                                 return 0;
5402                         if (ret < 0)
5403                                 return ret;
5404
5405                         if (path->locks[level]) {
5406                                 btrfs_tree_unlock_rw(path->nodes[level],
5407                                                      path->locks[level]);
5408                                 path->locks[level] = 0;
5409                         }
5410                         free_extent_buffer(path->nodes[level]);
5411                         path->nodes[level] = NULL;
5412                         level++;
5413                 }
5414         }
5415         return 1;
5416 }
5417
5418 /*
5419  * drop a subvolume tree.
5420  *
5421  * this function traverses the tree freeing any blocks that only
5422  * referenced by the tree.
5423  *
5424  * when a shared tree block is found. this function decreases its
5425  * reference count by one. if update_ref is true, this function
5426  * also make sure backrefs for the shared block and all lower level
5427  * blocks are properly updated.
5428  *
5429  * If called with for_reloc == 0, may exit early with -EAGAIN
5430  */
5431 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc)
5432 {
5433         struct btrfs_fs_info *fs_info = root->fs_info;
5434         struct btrfs_path *path;
5435         struct btrfs_trans_handle *trans;
5436         struct btrfs_root *tree_root = fs_info->tree_root;
5437         struct btrfs_root_item *root_item = &root->root_item;
5438         struct walk_control *wc;
5439         struct btrfs_key key;
5440         int err = 0;
5441         int ret;
5442         int level;
5443         bool root_dropped = false;
5444
5445         btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
5446
5447         path = btrfs_alloc_path();
5448         if (!path) {
5449                 err = -ENOMEM;
5450                 goto out;
5451         }
5452
5453         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5454         if (!wc) {
5455                 btrfs_free_path(path);
5456                 err = -ENOMEM;
5457                 goto out;
5458         }
5459
5460         /*
5461          * Use join to avoid potential EINTR from transaction start. See
5462          * wait_reserve_ticket and the whole reservation callchain.
5463          */
5464         if (for_reloc)
5465                 trans = btrfs_join_transaction(tree_root);
5466         else
5467                 trans = btrfs_start_transaction(tree_root, 0);
5468         if (IS_ERR(trans)) {
5469                 err = PTR_ERR(trans);
5470                 goto out_free;
5471         }
5472
5473         err = btrfs_run_delayed_items(trans);
5474         if (err)
5475                 goto out_end_trans;
5476
5477         /*
5478          * This will help us catch people modifying the fs tree while we're
5479          * dropping it.  It is unsafe to mess with the fs tree while it's being
5480          * dropped as we unlock the root node and parent nodes as we walk down
5481          * the tree, assuming nothing will change.  If something does change
5482          * then we'll have stale information and drop references to blocks we've
5483          * already dropped.
5484          */
5485         set_bit(BTRFS_ROOT_DELETING, &root->state);
5486         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5487                 level = btrfs_header_level(root->node);
5488                 path->nodes[level] = btrfs_lock_root_node(root);
5489                 btrfs_set_lock_blocking_write(path->nodes[level]);
5490                 path->slots[level] = 0;
5491                 path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5492                 memset(&wc->update_progress, 0,
5493                        sizeof(wc->update_progress));
5494         } else {
5495                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5496                 memcpy(&wc->update_progress, &key,
5497                        sizeof(wc->update_progress));
5498
5499                 level = root_item->drop_level;
5500                 BUG_ON(level == 0);
5501                 path->lowest_level = level;
5502                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5503                 path->lowest_level = 0;
5504                 if (ret < 0) {
5505                         err = ret;
5506                         goto out_end_trans;
5507                 }
5508                 WARN_ON(ret > 0);
5509
5510                 /*
5511                  * unlock our path, this is safe because only this
5512                  * function is allowed to delete this snapshot
5513                  */
5514                 btrfs_unlock_up_safe(path, 0);
5515
5516                 level = btrfs_header_level(root->node);
5517                 while (1) {
5518                         btrfs_tree_lock(path->nodes[level]);
5519                         btrfs_set_lock_blocking_write(path->nodes[level]);
5520                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5521
5522                         ret = btrfs_lookup_extent_info(trans, fs_info,
5523                                                 path->nodes[level]->start,
5524                                                 level, 1, &wc->refs[level],
5525                                                 &wc->flags[level]);
5526                         if (ret < 0) {
5527                                 err = ret;
5528                                 goto out_end_trans;
5529                         }
5530                         BUG_ON(wc->refs[level] == 0);
5531
5532                         if (level == root_item->drop_level)
5533                                 break;
5534
5535                         btrfs_tree_unlock(path->nodes[level]);
5536                         path->locks[level] = 0;
5537                         WARN_ON(wc->refs[level] != 1);
5538                         level--;
5539                 }
5540         }
5541
5542         wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
5543         wc->level = level;
5544         wc->shared_level = -1;
5545         wc->stage = DROP_REFERENCE;
5546         wc->update_ref = update_ref;
5547         wc->keep_locks = 0;
5548         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5549
5550         while (1) {
5551
5552                 ret = walk_down_tree(trans, root, path, wc);
5553                 if (ret < 0) {
5554                         err = ret;
5555                         break;
5556                 }
5557
5558                 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5559                 if (ret < 0) {
5560                         err = ret;
5561                         break;
5562                 }
5563
5564                 if (ret > 0) {
5565                         BUG_ON(wc->stage != DROP_REFERENCE);
5566                         break;
5567                 }
5568
5569                 if (wc->stage == DROP_REFERENCE) {
5570                         wc->drop_level = wc->level;
5571                         btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
5572                                               &wc->drop_progress,
5573                                               path->slots[wc->drop_level]);
5574                 }
5575                 btrfs_cpu_key_to_disk(&root_item->drop_progress,
5576                                       &wc->drop_progress);
5577                 root_item->drop_level = wc->drop_level;
5578
5579                 BUG_ON(wc->level == 0);
5580                 if (btrfs_should_end_transaction(trans) ||
5581                     (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
5582                         ret = btrfs_update_root(trans, tree_root,
5583                                                 &root->root_key,
5584                                                 root_item);
5585                         if (ret) {
5586                                 btrfs_abort_transaction(trans, ret);
5587                                 err = ret;
5588                                 goto out_end_trans;
5589                         }
5590
5591                         btrfs_end_transaction_throttle(trans);
5592                         if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
5593                                 btrfs_debug(fs_info,
5594                                             "drop snapshot early exit");
5595                                 err = -EAGAIN;
5596                                 goto out_free;
5597                         }
5598
5599                         trans = btrfs_start_transaction(tree_root, 0);
5600                         if (IS_ERR(trans)) {
5601                                 err = PTR_ERR(trans);
5602                                 goto out_free;
5603                         }
5604                 }
5605         }
5606         btrfs_release_path(path);
5607         if (err)
5608                 goto out_end_trans;
5609
5610         ret = btrfs_del_root(trans, &root->root_key);
5611         if (ret) {
5612                 btrfs_abort_transaction(trans, ret);
5613                 err = ret;
5614                 goto out_end_trans;
5615         }
5616
5617         if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5618                 ret = btrfs_find_root(tree_root, &root->root_key, path,
5619                                       NULL, NULL);
5620                 if (ret < 0) {
5621                         btrfs_abort_transaction(trans, ret);
5622                         err = ret;
5623                         goto out_end_trans;
5624                 } else if (ret > 0) {
5625                         /* if we fail to delete the orphan item this time
5626                          * around, it'll get picked up the next time.
5627                          *
5628                          * The most common failure here is just -ENOENT.
5629                          */
5630                         btrfs_del_orphan_item(trans, tree_root,
5631                                               root->root_key.objectid);
5632                 }
5633         }
5634
5635         /*
5636          * This subvolume is going to be completely dropped, and won't be
5637          * recorded as dirty roots, thus pertrans meta rsv will not be freed at
5638          * commit transaction time.  So free it here manually.
5639          */
5640         btrfs_qgroup_convert_reserved_meta(root, INT_MAX);
5641         btrfs_qgroup_free_meta_all_pertrans(root);
5642
5643         if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state))
5644                 btrfs_add_dropped_root(trans, root);
5645         else
5646                 btrfs_put_root(root);
5647         root_dropped = true;
5648 out_end_trans:
5649         btrfs_end_transaction_throttle(trans);
5650 out_free:
5651         kfree(wc);
5652         btrfs_free_path(path);
5653 out:
5654         /*
5655          * So if we need to stop dropping the snapshot for whatever reason we
5656          * need to make sure to add it back to the dead root list so that we
5657          * keep trying to do the work later.  This also cleans up roots if we
5658          * don't have it in the radix (like when we recover after a power fail
5659          * or unmount) so we don't leak memory.
5660          */
5661         if (!for_reloc && !root_dropped)
5662                 btrfs_add_dead_root(root);
5663         return err;
5664 }
5665
5666 /*
5667  * drop subtree rooted at tree block 'node'.
5668  *
5669  * NOTE: this function will unlock and release tree block 'node'
5670  * only used by relocation code
5671  */
5672 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5673                         struct btrfs_root *root,
5674                         struct extent_buffer *node,
5675                         struct extent_buffer *parent)
5676 {
5677         struct btrfs_fs_info *fs_info = root->fs_info;
5678         struct btrfs_path *path;
5679         struct walk_control *wc;
5680         int level;
5681         int parent_level;
5682         int ret = 0;
5683         int wret;
5684
5685         BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5686
5687         path = btrfs_alloc_path();
5688         if (!path)
5689                 return -ENOMEM;
5690
5691         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5692         if (!wc) {
5693                 btrfs_free_path(path);
5694                 return -ENOMEM;
5695         }
5696
5697         btrfs_assert_tree_locked(parent);
5698         parent_level = btrfs_header_level(parent);
5699         atomic_inc(&parent->refs);
5700         path->nodes[parent_level] = parent;
5701         path->slots[parent_level] = btrfs_header_nritems(parent);
5702
5703         btrfs_assert_tree_locked(node);
5704         level = btrfs_header_level(node);
5705         path->nodes[level] = node;
5706         path->slots[level] = 0;
5707         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5708
5709         wc->refs[parent_level] = 1;
5710         wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5711         wc->level = level;
5712         wc->shared_level = -1;
5713         wc->stage = DROP_REFERENCE;
5714         wc->update_ref = 0;
5715         wc->keep_locks = 1;
5716         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5717
5718         while (1) {
5719                 wret = walk_down_tree(trans, root, path, wc);
5720                 if (wret < 0) {
5721                         ret = wret;
5722                         break;
5723                 }
5724
5725                 wret = walk_up_tree(trans, root, path, wc, parent_level);
5726                 if (wret < 0)
5727                         ret = wret;
5728                 if (wret != 0)
5729                         break;
5730         }
5731
5732         kfree(wc);
5733         btrfs_free_path(path);
5734         return ret;
5735 }
5736
5737 /*
5738  * helper to account the unused space of all the readonly block group in the
5739  * space_info. takes mirrors into account.
5740  */
5741 u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
5742 {
5743         struct btrfs_block_group *block_group;
5744         u64 free_bytes = 0;
5745         int factor;
5746
5747         /* It's df, we don't care if it's racy */
5748         if (list_empty(&sinfo->ro_bgs))
5749                 return 0;
5750
5751         spin_lock(&sinfo->lock);
5752         list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
5753                 spin_lock(&block_group->lock);
5754
5755                 if (!block_group->ro) {
5756                         spin_unlock(&block_group->lock);
5757                         continue;
5758                 }
5759
5760                 factor = btrfs_bg_type_to_factor(block_group->flags);
5761                 free_bytes += (block_group->length -
5762                                block_group->used) * factor;
5763
5764                 spin_unlock(&block_group->lock);
5765         }
5766         spin_unlock(&sinfo->lock);
5767
5768         return free_bytes;
5769 }
5770
5771 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
5772                                    u64 start, u64 end)
5773 {
5774         return unpin_extent_range(fs_info, start, end, false);
5775 }
5776
5777 /*
5778  * It used to be that old block groups would be left around forever.
5779  * Iterating over them would be enough to trim unused space.  Since we
5780  * now automatically remove them, we also need to iterate over unallocated
5781  * space.
5782  *
5783  * We don't want a transaction for this since the discard may take a
5784  * substantial amount of time.  We don't require that a transaction be
5785  * running, but we do need to take a running transaction into account
5786  * to ensure that we're not discarding chunks that were released or
5787  * allocated in the current transaction.
5788  *
5789  * Holding the chunks lock will prevent other threads from allocating
5790  * or releasing chunks, but it won't prevent a running transaction
5791  * from committing and releasing the memory that the pending chunks
5792  * list head uses.  For that, we need to take a reference to the
5793  * transaction and hold the commit root sem.  We only need to hold
5794  * it while performing the free space search since we have already
5795  * held back allocations.
5796  */
5797 static int btrfs_trim_free_extents(struct btrfs_device *device, u64 *trimmed)
5798 {
5799         u64 start = SZ_1M, len = 0, end = 0;
5800         int ret;
5801
5802         *trimmed = 0;
5803
5804         /* Discard not supported = nothing to do. */
5805         if (!blk_queue_discard(bdev_get_queue(device->bdev)))
5806                 return 0;
5807
5808         /* Not writable = nothing to do. */
5809         if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state))
5810                 return 0;
5811
5812         /* No free space = nothing to do. */
5813         if (device->total_bytes <= device->bytes_used)
5814                 return 0;
5815
5816         ret = 0;
5817
5818         while (1) {
5819                 struct btrfs_fs_info *fs_info = device->fs_info;
5820                 u64 bytes;
5821
5822                 ret = mutex_lock_interruptible(&fs_info->chunk_mutex);
5823                 if (ret)
5824                         break;
5825
5826                 find_first_clear_extent_bit(&device->alloc_state, start,
5827                                             &start, &end,
5828                                             CHUNK_TRIMMED | CHUNK_ALLOCATED);
5829
5830                 /* Check if there are any CHUNK_* bits left */
5831                 if (start > device->total_bytes) {
5832                         WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
5833                         btrfs_warn_in_rcu(fs_info,
5834 "ignoring attempt to trim beyond device size: offset %llu length %llu device %s device size %llu",
5835                                           start, end - start + 1,
5836                                           rcu_str_deref(device->name),
5837                                           device->total_bytes);
5838                         mutex_unlock(&fs_info->chunk_mutex);
5839                         ret = 0;
5840                         break;
5841                 }
5842
5843                 /* Ensure we skip the reserved area in the first 1M */
5844                 start = max_t(u64, start, SZ_1M);
5845
5846                 /*
5847                  * If find_first_clear_extent_bit find a range that spans the
5848                  * end of the device it will set end to -1, in this case it's up
5849                  * to the caller to trim the value to the size of the device.
5850                  */
5851                 end = min(end, device->total_bytes - 1);
5852
5853                 len = end - start + 1;
5854
5855                 /* We didn't find any extents */
5856                 if (!len) {
5857                         mutex_unlock(&fs_info->chunk_mutex);
5858                         ret = 0;
5859                         break;
5860                 }
5861
5862                 ret = btrfs_issue_discard(device->bdev, start, len,
5863                                           &bytes);
5864                 if (!ret)
5865                         set_extent_bits(&device->alloc_state, start,
5866                                         start + bytes - 1,
5867                                         CHUNK_TRIMMED);
5868                 mutex_unlock(&fs_info->chunk_mutex);
5869
5870                 if (ret)
5871                         break;
5872
5873                 start += len;
5874                 *trimmed += bytes;
5875
5876                 if (fatal_signal_pending(current)) {
5877                         ret = -ERESTARTSYS;
5878                         break;
5879                 }
5880
5881                 cond_resched();
5882         }
5883
5884         return ret;
5885 }
5886
5887 /*
5888  * Trim the whole filesystem by:
5889  * 1) trimming the free space in each block group
5890  * 2) trimming the unallocated space on each device
5891  *
5892  * This will also continue trimming even if a block group or device encounters
5893  * an error.  The return value will be the last error, or 0 if nothing bad
5894  * happens.
5895  */
5896 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range)
5897 {
5898         struct btrfs_block_group *cache = NULL;
5899         struct btrfs_device *device;
5900         struct list_head *devices;
5901         u64 group_trimmed;
5902         u64 range_end = U64_MAX;
5903         u64 start;
5904         u64 end;
5905         u64 trimmed = 0;
5906         u64 bg_failed = 0;
5907         u64 dev_failed = 0;
5908         int bg_ret = 0;
5909         int dev_ret = 0;
5910         int ret = 0;
5911
5912         /*
5913          * Check range overflow if range->len is set.
5914          * The default range->len is U64_MAX.
5915          */
5916         if (range->len != U64_MAX &&
5917             check_add_overflow(range->start, range->len, &range_end))
5918                 return -EINVAL;
5919
5920         cache = btrfs_lookup_first_block_group(fs_info, range->start);
5921         for (; cache; cache = btrfs_next_block_group(cache)) {
5922                 if (cache->start >= range_end) {
5923                         btrfs_put_block_group(cache);
5924                         break;
5925                 }
5926
5927                 start = max(range->start, cache->start);
5928                 end = min(range_end, cache->start + cache->length);
5929
5930                 if (end - start >= range->minlen) {
5931                         if (!btrfs_block_group_done(cache)) {
5932                                 ret = btrfs_cache_block_group(cache, 0);
5933                                 if (ret) {
5934                                         bg_failed++;
5935                                         bg_ret = ret;
5936                                         continue;
5937                                 }
5938                                 ret = btrfs_wait_block_group_cache_done(cache);
5939                                 if (ret) {
5940                                         bg_failed++;
5941                                         bg_ret = ret;
5942                                         continue;
5943                                 }
5944                         }
5945                         ret = btrfs_trim_block_group(cache,
5946                                                      &group_trimmed,
5947                                                      start,
5948                                                      end,
5949                                                      range->minlen);
5950
5951                         trimmed += group_trimmed;
5952                         if (ret) {
5953                                 bg_failed++;
5954                                 bg_ret = ret;
5955                                 continue;
5956                         }
5957                 }
5958         }
5959
5960         if (bg_failed)
5961                 btrfs_warn(fs_info,
5962                         "failed to trim %llu block group(s), last error %d",
5963                         bg_failed, bg_ret);
5964         mutex_lock(&fs_info->fs_devices->device_list_mutex);
5965         devices = &fs_info->fs_devices->devices;
5966         list_for_each_entry(device, devices, dev_list) {
5967                 ret = btrfs_trim_free_extents(device, &group_trimmed);
5968                 if (ret) {
5969                         dev_failed++;
5970                         dev_ret = ret;
5971                         break;
5972                 }
5973
5974                 trimmed += group_trimmed;
5975         }
5976         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
5977
5978         if (dev_failed)
5979                 btrfs_warn(fs_info,
5980                         "failed to trim %llu device(s), last error %d",
5981                         dev_failed, dev_ret);
5982         range->len = trimmed;
5983         if (bg_ret)
5984                 return bg_ret;
5985         return dev_ret;
5986 }