fs/btrfs/extent-tree.c

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
block_group_bits
btrfs_add_excluded_extent
btrfs_free_excluded_extents
generic_ref_to_space_flags
add_pinned_bytes
sub_pinned_bytes
btrfs_lookup_data_extent
btrfs_lookup_extent_info
btrfs_get_extent_inline_ref_type
hash_extent_data_ref
hash_extent_data_ref_item
match_extent_data_ref
lookup_extent_data_ref
insert_extent_data_ref
remove_extent_data_ref
extent_data_ref_count
lookup_tree_block_ref
insert_tree_block_ref
extent_ref_type
find_next_key
lookup_inline_extent_backref
setup_inline_extent_backref
lookup_extent_backref
update_inline_extent_backref
insert_inline_extent_backref
insert_extent_backref
remove_extent_backref
btrfs_issue_discard
btrfs_discard_extent
btrfs_inc_extent_ref
__btrfs_inc_extent_ref
run_delayed_data_ref
__run_delayed_extent_op
run_delayed_extent_op
run_delayed_tree_ref
run_one_delayed_ref
select_delayed_ref
unselect_delayed_ref_head
cleanup_extent_op
run_and_cleanup_extent_op
btrfs_cleanup_ref_head_accounting
cleanup_ref_head
btrfs_obtain_ref_head
btrfs_run_delayed_refs_for_head
__btrfs_run_delayed_refs
find_middle
heads_to_leaves
btrfs_csum_bytes_to_leaves
btrfs_run_delayed_refs
btrfs_set_disk_extent_flags
check_delayed_ref
check_committed_ref
btrfs_cross_ref_exist
__btrfs_mod_ref
btrfs_inc_ref
btrfs_dec_ref
btrfs_extent_readonly
get_alloc_profile_by_root
first_logical_byte
pin_down_extent
btrfs_pin_extent
btrfs_pin_extent_for_log_replay
__exclude_logged_extent
btrfs_exclude_logged_extents
btrfs_inc_block_group_reservations
btrfs_prepare_extent_commit
fetch_cluster_info
unpin_extent_range
btrfs_finish_extent_commit
__btrfs_free_extent
check_ref_cleanup
btrfs_free_tree_block
btrfs_free_extent
btrfs_lock_block_group
btrfs_grab_block_group
btrfs_lock_cluster
btrfs_release_block_group
find_free_extent_clustered
find_free_extent_unclustered
find_free_extent_update_loop
find_free_extent
btrfs_reserve_extent
__btrfs_free_reserved_extent
btrfs_free_reserved_extent
btrfs_free_and_pin_reserved_extent
alloc_reserved_file_extent
alloc_reserved_tree_block
btrfs_alloc_reserved_file_extent
btrfs_alloc_logged_file_extent
btrfs_init_new_buffer
btrfs_alloc_tree_block
reada_walk_down
walk_down_proc
check_ref_exists
do_walk_down
walk_up_proc
walk_down_tree
walk_up_tree
btrfs_drop_snapshot
btrfs_drop_subtree
btrfs_account_ro_block_groups_free_space
btrfs_error_unpin_extent_range
btrfs_trim_free_extents
btrfs_trim_fs
btrfs_end_write_no_snapshotting
btrfs_start_write_no_snapshotting
btrfs_wait_for_snapshot_creation
   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 
  36 #undef SCRAMBLE_DELAYED_REFS
  37 
  38 
  39 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
  40                                struct btrfs_delayed_ref_node *node, u64 parent,
  41                                u64 root_objectid, u64 owner_objectid,
  42                                u64 owner_offset, int refs_to_drop,
  43                                struct btrfs_delayed_extent_op *extra_op);
  44 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
  45                                     struct extent_buffer *leaf,
  46                                     struct btrfs_extent_item *ei);
  47 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
  48                                       u64 parent, u64 root_objectid,
  49                                       u64 flags, u64 owner, u64 offset,
  50                                       struct btrfs_key *ins, int ref_mod);
  51 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
  52                                      struct btrfs_delayed_ref_node *node,
  53                                      struct btrfs_delayed_extent_op *extent_op);
  54 static int find_next_key(struct btrfs_path *path, int level,
  55                          struct btrfs_key *key);
  56 
  57 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
  58 {
  59         return (cache->flags & bits) == bits;
  60 }
  61 
  62 int btrfs_add_excluded_extent(struct btrfs_fs_info *fs_info,
  63                               u64 start, u64 num_bytes)
  64 {
  65         u64 end = start + num_bytes - 1;
  66         set_extent_bits(&fs_info->freed_extents[0],
  67                         start, end, EXTENT_UPTODATE);
  68         set_extent_bits(&fs_info->freed_extents[1],
  69                         start, end, EXTENT_UPTODATE);
  70         return 0;
  71 }
  72 
  73 void btrfs_free_excluded_extents(struct btrfs_block_group_cache *cache)
  74 {
  75         struct btrfs_fs_info *fs_info = cache->fs_info;
  76         u64 start, end;
  77 
  78         start = cache->key.objectid;
  79         end = start + cache->key.offset - 1;
  80 
  81         clear_extent_bits(&fs_info->freed_extents[0],
  82                           start, end, EXTENT_UPTODATE);
  83         clear_extent_bits(&fs_info->freed_extents[1],
  84                           start, end, EXTENT_UPTODATE);
  85 }
  86 
  87 static u64 generic_ref_to_space_flags(struct btrfs_ref *ref)
  88 {
  89         if (ref->type == BTRFS_REF_METADATA) {
  90                 if (ref->tree_ref.root == BTRFS_CHUNK_TREE_OBJECTID)
  91                         return BTRFS_BLOCK_GROUP_SYSTEM;
  92                 else
  93                         return BTRFS_BLOCK_GROUP_METADATA;
  94         }
  95         return BTRFS_BLOCK_GROUP_DATA;
  96 }
  97 
  98 static void add_pinned_bytes(struct btrfs_fs_info *fs_info,
  99                              struct btrfs_ref *ref)
 100 {
 101         struct btrfs_space_info *space_info;
 102         u64 flags = generic_ref_to_space_flags(ref);
 103 
 104         space_info = btrfs_find_space_info(fs_info, flags);
 105         ASSERT(space_info);
 106         percpu_counter_add_batch(&space_info->total_bytes_pinned, ref->len,
 107                     BTRFS_TOTAL_BYTES_PINNED_BATCH);
 108 }
 109 
 110 static void sub_pinned_bytes(struct btrfs_fs_info *fs_info,
 111                              struct btrfs_ref *ref)
 112 {
 113         struct btrfs_space_info *space_info;
 114         u64 flags = generic_ref_to_space_flags(ref);
 115 
 116         space_info = btrfs_find_space_info(fs_info, flags);
 117         ASSERT(space_info);
 118         percpu_counter_add_batch(&space_info->total_bytes_pinned, -ref->len,
 119                     BTRFS_TOTAL_BYTES_PINNED_BATCH);
 120 }
 121 
 122 /* simple helper to search for an existing data extent at a given offset */
 123 int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
 124 {
 125         int ret;
 126         struct btrfs_key key;
 127         struct btrfs_path *path;
 128 
 129         path = btrfs_alloc_path();
 130         if (!path)
 131                 return -ENOMEM;
 132 
 133         key.objectid = start;
 134         key.offset = len;
 135         key.type = BTRFS_EXTENT_ITEM_KEY;
 136         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
 137         btrfs_free_path(path);
 138         return ret;
 139 }
 140 
 141 /*
 142  * helper function to lookup reference count and flags of a tree block.
 143  *
 144  * the head node for delayed ref is used to store the sum of all the
 145  * reference count modifications queued up in the rbtree. the head
 146  * node may also store the extent flags to set. This way you can check
 147  * to see what the reference count and extent flags would be if all of
 148  * the delayed refs are not processed.
 149  */
 150 int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 151                              struct btrfs_fs_info *fs_info, u64 bytenr,
 152                              u64 offset, int metadata, u64 *refs, u64 *flags)
 153 {
 154         struct btrfs_delayed_ref_head *head;
 155         struct btrfs_delayed_ref_root *delayed_refs;
 156         struct btrfs_path *path;
 157         struct btrfs_extent_item *ei;
 158         struct extent_buffer *leaf;
 159         struct btrfs_key key;
 160         u32 item_size;
 161         u64 num_refs;
 162         u64 extent_flags;
 163         int ret;
 164 
 165         /*
 166          * If we don't have skinny metadata, don't bother doing anything
 167          * different
 168          */
 169         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
 170                 offset = fs_info->nodesize;
 171                 metadata = 0;
 172         }
 173 
 174         path = btrfs_alloc_path();
 175         if (!path)
 176                 return -ENOMEM;
 177 
 178         if (!trans) {
 179                 path->skip_locking = 1;
 180                 path->search_commit_root = 1;
 181         }
 182 
 183 search_again:
 184         key.objectid = bytenr;
 185         key.offset = offset;
 186         if (metadata)
 187                 key.type = BTRFS_METADATA_ITEM_KEY;
 188         else
 189                 key.type = BTRFS_EXTENT_ITEM_KEY;
 190 
 191         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
 192         if (ret < 0)
 193                 goto out_free;
 194 
 195         if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
 196                 if (path->slots[0]) {
 197                         path->slots[0]--;
 198                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 199                                               path->slots[0]);
 200                         if (key.objectid == bytenr &&
 201                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 202                             key.offset == fs_info->nodesize)
 203                                 ret = 0;
 204                 }
 205         }
 206 
 207         if (ret == 0) {
 208                 leaf = path->nodes[0];
 209                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 210                 if (item_size >= sizeof(*ei)) {
 211                         ei = btrfs_item_ptr(leaf, path->slots[0],
 212                                             struct btrfs_extent_item);
 213                         num_refs = btrfs_extent_refs(leaf, ei);
 214                         extent_flags = btrfs_extent_flags(leaf, ei);
 215                 } else {
 216                         ret = -EINVAL;
 217                         btrfs_print_v0_err(fs_info);
 218                         if (trans)
 219                                 btrfs_abort_transaction(trans, ret);
 220                         else
 221                                 btrfs_handle_fs_error(fs_info, ret, NULL);
 222 
 223                         goto out_free;
 224                 }
 225 
 226                 BUG_ON(num_refs == 0);
 227         } else {
 228                 num_refs = 0;
 229                 extent_flags = 0;
 230                 ret = 0;
 231         }
 232 
 233         if (!trans)
 234                 goto out;
 235 
 236         delayed_refs = &trans->transaction->delayed_refs;
 237         spin_lock(&delayed_refs->lock);
 238         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
 239         if (head) {
 240                 if (!mutex_trylock(&head->mutex)) {
 241                         refcount_inc(&head->refs);
 242                         spin_unlock(&delayed_refs->lock);
 243 
 244                         btrfs_release_path(path);
 245 
 246                         /*
 247                          * Mutex was contended, block until it's released and try
 248                          * again
 249                          */
 250                         mutex_lock(&head->mutex);
 251                         mutex_unlock(&head->mutex);
 252                         btrfs_put_delayed_ref_head(head);
 253                         goto search_again;
 254                 }
 255                 spin_lock(&head->lock);
 256                 if (head->extent_op && head->extent_op->update_flags)
 257                         extent_flags |= head->extent_op->flags_to_set;
 258                 else
 259                         BUG_ON(num_refs == 0);
 260 
 261                 num_refs += head->ref_mod;
 262                 spin_unlock(&head->lock);
 263                 mutex_unlock(&head->mutex);
 264         }
 265         spin_unlock(&delayed_refs->lock);
 266 out:
 267         WARN_ON(num_refs == 0);
 268         if (refs)
 269                 *refs = num_refs;
 270         if (flags)
 271                 *flags = extent_flags;
 272 out_free:
 273         btrfs_free_path(path);
 274         return ret;
 275 }
 276 
 277 /*
 278  * Back reference rules.  Back refs have three main goals:
 279  *
 280  * 1) differentiate between all holders of references to an extent so that
 281  *    when a reference is dropped we can make sure it was a valid reference
 282  *    before freeing the extent.
 283  *
 284  * 2) Provide enough information to quickly find the holders of an extent
 285  *    if we notice a given block is corrupted or bad.
 286  *
 287  * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 288  *    maintenance.  This is actually the same as #2, but with a slightly
 289  *    different use case.
 290  *
 291  * There are two kinds of back refs. The implicit back refs is optimized
 292  * for pointers in non-shared tree blocks. For a given pointer in a block,
 293  * back refs of this kind provide information about the block's owner tree
 294  * and the pointer's key. These information allow us to find the block by
 295  * b-tree searching. The full back refs is for pointers in tree blocks not
 296  * referenced by their owner trees. The location of tree block is recorded
 297  * in the back refs. Actually the full back refs is generic, and can be
 298  * used in all cases the implicit back refs is used. The major shortcoming
 299  * of the full back refs is its overhead. Every time a tree block gets
 300  * COWed, we have to update back refs entry for all pointers in it.
 301  *
 302  * For a newly allocated tree block, we use implicit back refs for
 303  * pointers in it. This means most tree related operations only involve
 304  * implicit back refs. For a tree block created in old transaction, the
 305  * only way to drop a reference to it is COW it. So we can detect the
 306  * event that tree block loses its owner tree's reference and do the
 307  * back refs conversion.
 308  *
 309  * When a tree block is COWed through a tree, there are four cases:
 310  *
 311  * The reference count of the block is one and the tree is the block's
 312  * owner tree. Nothing to do in this case.
 313  *
 314  * The reference count of the block is one and the tree is not the
 315  * block's owner tree. In this case, full back refs is used for pointers
 316  * in the block. Remove these full back refs, add implicit back refs for
 317  * every pointers in the new block.
 318  *
 319  * The reference count of the block is greater than one and the tree is
 320  * the block's owner tree. In this case, implicit back refs is used for
 321  * pointers in the block. Add full back refs for every pointers in the
 322  * block, increase lower level extents' reference counts. The original
 323  * implicit back refs are entailed to the new block.
 324  *
 325  * The reference count of the block is greater than one and the tree is
 326  * not the block's owner tree. Add implicit back refs for every pointer in
 327  * the new block, increase lower level extents' reference count.
 328  *
 329  * Back Reference Key composing:
 330  *
 331  * The key objectid corresponds to the first byte in the extent,
 332  * The key type is used to differentiate between types of back refs.
 333  * There are different meanings of the key offset for different types
 334  * of back refs.
 335  *
 336  * File extents can be referenced by:
 337  *
 338  * - multiple snapshots, subvolumes, or different generations in one subvol
 339  * - different files inside a single subvolume
 340  * - different offsets inside a file (bookend extents in file.c)
 341  *
 342  * The extent ref structure for the implicit back refs has fields for:
 343  *
 344  * - Objectid of the subvolume root
 345  * - objectid of the file holding the reference
 346  * - original offset in the file
 347  * - how many bookend extents
 348  *
 349  * The key offset for the implicit back refs is hash of the first
 350  * three fields.
 351  *
 352  * The extent ref structure for the full back refs has field for:
 353  *
 354  * - number of pointers in the tree leaf
 355  *
 356  * The key offset for the implicit back refs is the first byte of
 357  * the tree leaf
 358  *
 359  * When a file extent is allocated, The implicit back refs is used.
 360  * the fields are filled in:
 361  *
 362  *     (root_key.objectid, inode objectid, offset in file, 1)
 363  *
 364  * When a file extent is removed file truncation, we find the
 365  * corresponding implicit back refs and check the following fields:
 366  *
 367  *     (btrfs_header_owner(leaf), inode objectid, offset in file)
 368  *
 369  * Btree extents can be referenced by:
 370  *
 371  * - Different subvolumes
 372  *
 373  * Both the implicit back refs and the full back refs for tree blocks
 374  * only consist of key. The key offset for the implicit back refs is
 375  * objectid of block's owner tree. The key offset for the full back refs
 376  * is the first byte of parent block.
 377  *
 378  * When implicit back refs is used, information about the lowest key and
 379  * level of the tree block are required. These information are stored in
 380  * tree block info structure.
 381  */
 382 
 383 /*
 384  * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
 385  * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
 386  * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
 387  */
 388 int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
 389                                      struct btrfs_extent_inline_ref *iref,
 390                                      enum btrfs_inline_ref_type is_data)
 391 {
 392         int type = btrfs_extent_inline_ref_type(eb, iref);
 393         u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
 394 
 395         if (type == BTRFS_TREE_BLOCK_REF_KEY ||
 396             type == BTRFS_SHARED_BLOCK_REF_KEY ||
 397             type == BTRFS_SHARED_DATA_REF_KEY ||
 398             type == BTRFS_EXTENT_DATA_REF_KEY) {
 399                 if (is_data == BTRFS_REF_TYPE_BLOCK) {
 400                         if (type == BTRFS_TREE_BLOCK_REF_KEY)
 401                                 return type;
 402                         if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
 403                                 ASSERT(eb->fs_info);
 404                                 /*
 405                                  * Every shared one has parent tree
 406                                  * block, which must be aligned to
 407                                  * nodesize.
 408                                  */
 409                                 if (offset &&
 410                                     IS_ALIGNED(offset, eb->fs_info->nodesize))
 411                                         return type;
 412                         }
 413                 } else if (is_data == BTRFS_REF_TYPE_DATA) {
 414                         if (type == BTRFS_EXTENT_DATA_REF_KEY)
 415                                 return type;
 416                         if (type == BTRFS_SHARED_DATA_REF_KEY) {
 417                                 ASSERT(eb->fs_info);
 418                                 /*
 419                                  * Every shared one has parent tree
 420                                  * block, which must be aligned to
 421                                  * nodesize.
 422                                  */
 423                                 if (offset &&
 424                                     IS_ALIGNED(offset, eb->fs_info->nodesize))
 425                                         return type;
 426                         }
 427                 } else {
 428                         ASSERT(is_data == BTRFS_REF_TYPE_ANY);
 429                         return type;
 430                 }
 431         }
 432 
 433         btrfs_print_leaf((struct extent_buffer *)eb);
 434         btrfs_err(eb->fs_info, "eb %llu invalid extent inline ref type %d",
 435                   eb->start, type);
 436         WARN_ON(1);
 437 
 438         return BTRFS_REF_TYPE_INVALID;
 439 }
 440 
 441 u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
 442 {
 443         u32 high_crc = ~(u32)0;
 444         u32 low_crc = ~(u32)0;
 445         __le64 lenum;
 446 
 447         lenum = cpu_to_le64(root_objectid);
 448         high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
 449         lenum = cpu_to_le64(owner);
 450         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 451         lenum = cpu_to_le64(offset);
 452         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 453 
 454         return ((u64)high_crc << 31) ^ (u64)low_crc;
 455 }
 456 
 457 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
 458                                      struct btrfs_extent_data_ref *ref)
 459 {
 460         return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
 461                                     btrfs_extent_data_ref_objectid(leaf, ref),
 462                                     btrfs_extent_data_ref_offset(leaf, ref));
 463 }
 464 
 465 static int match_extent_data_ref(struct extent_buffer *leaf,
 466                                  struct btrfs_extent_data_ref *ref,
 467                                  u64 root_objectid, u64 owner, u64 offset)
 468 {
 469         if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
 470             btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
 471             btrfs_extent_data_ref_offset(leaf, ref) != offset)
 472                 return 0;
 473         return 1;
 474 }
 475 
 476 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
 477                                            struct btrfs_path *path,
 478                                            u64 bytenr, u64 parent,
 479                                            u64 root_objectid,
 480                                            u64 owner, u64 offset)
 481 {
 482         struct btrfs_root *root = trans->fs_info->extent_root;
 483         struct btrfs_key key;
 484         struct btrfs_extent_data_ref *ref;
 485         struct extent_buffer *leaf;
 486         u32 nritems;
 487         int ret;
 488         int recow;
 489         int err = -ENOENT;
 490 
 491         key.objectid = bytenr;
 492         if (parent) {
 493                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 494                 key.offset = parent;
 495         } else {
 496                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 497                 key.offset = hash_extent_data_ref(root_objectid,
 498                                                   owner, offset);
 499         }
 500 again:
 501         recow = 0;
 502         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 503         if (ret < 0) {
 504                 err = ret;
 505                 goto fail;
 506         }
 507 
 508         if (parent) {
 509                 if (!ret)
 510                         return 0;
 511                 goto fail;
 512         }
 513 
 514         leaf = path->nodes[0];
 515         nritems = btrfs_header_nritems(leaf);
 516         while (1) {
 517                 if (path->slots[0] >= nritems) {
 518                         ret = btrfs_next_leaf(root, path);
 519                         if (ret < 0)
 520                                 err = ret;
 521                         if (ret)
 522                                 goto fail;
 523 
 524                         leaf = path->nodes[0];
 525                         nritems = btrfs_header_nritems(leaf);
 526                         recow = 1;
 527                 }
 528 
 529                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 530                 if (key.objectid != bytenr ||
 531                     key.type != BTRFS_EXTENT_DATA_REF_KEY)
 532                         goto fail;
 533 
 534                 ref = btrfs_item_ptr(leaf, path->slots[0],
 535                                      struct btrfs_extent_data_ref);
 536 
 537                 if (match_extent_data_ref(leaf, ref, root_objectid,
 538                                           owner, offset)) {
 539                         if (recow) {
 540                                 btrfs_release_path(path);
 541                                 goto again;
 542                         }
 543                         err = 0;
 544                         break;
 545                 }
 546                 path->slots[0]++;
 547         }
 548 fail:
 549         return err;
 550 }
 551 
 552 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 553                                            struct btrfs_path *path,
 554                                            u64 bytenr, u64 parent,
 555                                            u64 root_objectid, u64 owner,
 556                                            u64 offset, int refs_to_add)
 557 {
 558         struct btrfs_root *root = trans->fs_info->extent_root;
 559         struct btrfs_key key;
 560         struct extent_buffer *leaf;
 561         u32 size;
 562         u32 num_refs;
 563         int ret;
 564 
 565         key.objectid = bytenr;
 566         if (parent) {
 567                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 568                 key.offset = parent;
 569                 size = sizeof(struct btrfs_shared_data_ref);
 570         } else {
 571                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 572                 key.offset = hash_extent_data_ref(root_objectid,
 573                                                   owner, offset);
 574                 size = sizeof(struct btrfs_extent_data_ref);
 575         }
 576 
 577         ret = btrfs_insert_empty_item(trans, root, path, &key, size);
 578         if (ret && ret != -EEXIST)
 579                 goto fail;
 580 
 581         leaf = path->nodes[0];
 582         if (parent) {
 583                 struct btrfs_shared_data_ref *ref;
 584                 ref = btrfs_item_ptr(leaf, path->slots[0],
 585                                      struct btrfs_shared_data_ref);
 586                 if (ret == 0) {
 587                         btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
 588                 } else {
 589                         num_refs = btrfs_shared_data_ref_count(leaf, ref);
 590                         num_refs += refs_to_add;
 591                         btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
 592                 }
 593         } else {
 594                 struct btrfs_extent_data_ref *ref;
 595                 while (ret == -EEXIST) {
 596                         ref = btrfs_item_ptr(leaf, path->slots[0],
 597                                              struct btrfs_extent_data_ref);
 598                         if (match_extent_data_ref(leaf, ref, root_objectid,
 599                                                   owner, offset))
 600                                 break;
 601                         btrfs_release_path(path);
 602                         key.offset++;
 603                         ret = btrfs_insert_empty_item(trans, root, path, &key,
 604                                                       size);
 605                         if (ret && ret != -EEXIST)
 606                                 goto fail;
 607 
 608                         leaf = path->nodes[0];
 609                 }
 610                 ref = btrfs_item_ptr(leaf, path->slots[0],
 611                                      struct btrfs_extent_data_ref);
 612                 if (ret == 0) {
 613                         btrfs_set_extent_data_ref_root(leaf, ref,
 614                                                        root_objectid);
 615                         btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
 616                         btrfs_set_extent_data_ref_offset(leaf, ref, offset);
 617                         btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
 618                 } else {
 619                         num_refs = btrfs_extent_data_ref_count(leaf, ref);
 620                         num_refs += refs_to_add;
 621                         btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
 622                 }
 623         }
 624         btrfs_mark_buffer_dirty(leaf);
 625         ret = 0;
 626 fail:
 627         btrfs_release_path(path);
 628         return ret;
 629 }
 630 
 631 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 632                                            struct btrfs_path *path,
 633                                            int refs_to_drop, int *last_ref)
 634 {
 635         struct btrfs_key key;
 636         struct btrfs_extent_data_ref *ref1 = NULL;
 637         struct btrfs_shared_data_ref *ref2 = NULL;
 638         struct extent_buffer *leaf;
 639         u32 num_refs = 0;
 640         int ret = 0;
 641 
 642         leaf = path->nodes[0];
 643         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 644 
 645         if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 646                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 647                                       struct btrfs_extent_data_ref);
 648                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 649         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 650                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 651                                       struct btrfs_shared_data_ref);
 652                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 653         } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
 654                 btrfs_print_v0_err(trans->fs_info);
 655                 btrfs_abort_transaction(trans, -EINVAL);
 656                 return -EINVAL;
 657         } else {
 658                 BUG();
 659         }
 660 
 661         BUG_ON(num_refs < refs_to_drop);
 662         num_refs -= refs_to_drop;
 663 
 664         if (num_refs == 0) {
 665                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
 666                 *last_ref = 1;
 667         } else {
 668                 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
 669                         btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
 670                 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
 671                         btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
 672                 btrfs_mark_buffer_dirty(leaf);
 673         }
 674         return ret;
 675 }
 676 
 677 static noinline u32 extent_data_ref_count(struct btrfs_path *path,
 678                                           struct btrfs_extent_inline_ref *iref)
 679 {
 680         struct btrfs_key key;
 681         struct extent_buffer *leaf;
 682         struct btrfs_extent_data_ref *ref1;
 683         struct btrfs_shared_data_ref *ref2;
 684         u32 num_refs = 0;
 685         int type;
 686 
 687         leaf = path->nodes[0];
 688         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 689 
 690         BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
 691         if (iref) {
 692                 /*
 693                  * If type is invalid, we should have bailed out earlier than
 694                  * this call.
 695                  */
 696                 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
 697                 ASSERT(type != BTRFS_REF_TYPE_INVALID);
 698                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 699                         ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
 700                         num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 701                 } else {
 702                         ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
 703                         num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 704                 }
 705         } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 706                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 707                                       struct btrfs_extent_data_ref);
 708                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 709         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 710                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 711                                       struct btrfs_shared_data_ref);
 712                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 713         } else {
 714                 WARN_ON(1);
 715         }
 716         return num_refs;
 717 }
 718 
 719 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
 720                                           struct btrfs_path *path,
 721                                           u64 bytenr, u64 parent,
 722                                           u64 root_objectid)
 723 {
 724         struct btrfs_root *root = trans->fs_info->extent_root;
 725         struct btrfs_key key;
 726         int ret;
 727 
 728         key.objectid = bytenr;
 729         if (parent) {
 730                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 731                 key.offset = parent;
 732         } else {
 733                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 734                 key.offset = root_objectid;
 735         }
 736 
 737         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 738         if (ret > 0)
 739                 ret = -ENOENT;
 740         return ret;
 741 }
 742 
 743 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
 744                                           struct btrfs_path *path,
 745                                           u64 bytenr, u64 parent,
 746                                           u64 root_objectid)
 747 {
 748         struct btrfs_key key;
 749         int ret;
 750 
 751         key.objectid = bytenr;
 752         if (parent) {
 753                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 754                 key.offset = parent;
 755         } else {
 756                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 757                 key.offset = root_objectid;
 758         }
 759 
 760         ret = btrfs_insert_empty_item(trans, trans->fs_info->extent_root,
 761                                       path, &key, 0);
 762         btrfs_release_path(path);
 763         return ret;
 764 }
 765 
 766 static inline int extent_ref_type(u64 parent, u64 owner)
 767 {
 768         int type;
 769         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
 770                 if (parent > 0)
 771                         type = BTRFS_SHARED_BLOCK_REF_KEY;
 772                 else
 773                         type = BTRFS_TREE_BLOCK_REF_KEY;
 774         } else {
 775                 if (parent > 0)
 776                         type = BTRFS_SHARED_DATA_REF_KEY;
 777                 else
 778                         type = BTRFS_EXTENT_DATA_REF_KEY;
 779         }
 780         return type;
 781 }
 782 
 783 static int find_next_key(struct btrfs_path *path, int level,
 784                          struct btrfs_key *key)
 785 
 786 {
 787         for (; level < BTRFS_MAX_LEVEL; level++) {
 788                 if (!path->nodes[level])
 789                         break;
 790                 if (path->slots[level] + 1 >=
 791                     btrfs_header_nritems(path->nodes[level]))
 792                         continue;
 793                 if (level == 0)
 794                         btrfs_item_key_to_cpu(path->nodes[level], key,
 795                                               path->slots[level] + 1);
 796                 else
 797                         btrfs_node_key_to_cpu(path->nodes[level], key,
 798                                               path->slots[level] + 1);
 799                 return 0;
 800         }
 801         return 1;
 802 }
 803 
 804 /*
 805  * look for inline back ref. if back ref is found, *ref_ret is set
 806  * to the address of inline back ref, and 0 is returned.
 807  *
 808  * if back ref isn't found, *ref_ret is set to the address where it
 809  * should be inserted, and -ENOENT is returned.
 810  *
 811  * if insert is true and there are too many inline back refs, the path
 812  * points to the extent item, and -EAGAIN is returned.
 813  *
 814  * NOTE: inline back refs are ordered in the same way that back ref
 815  *       items in the tree are ordered.
 816  */
 817 static noinline_for_stack
 818 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 819                                  struct btrfs_path *path,
 820                                  struct btrfs_extent_inline_ref **ref_ret,
 821                                  u64 bytenr, u64 num_bytes,
 822                                  u64 parent, u64 root_objectid,
 823                                  u64 owner, u64 offset, int insert)
 824 {
 825         struct btrfs_fs_info *fs_info = trans->fs_info;
 826         struct btrfs_root *root = fs_info->extent_root;
 827         struct btrfs_key key;
 828         struct extent_buffer *leaf;
 829         struct btrfs_extent_item *ei;
 830         struct btrfs_extent_inline_ref *iref;
 831         u64 flags;
 832         u64 item_size;
 833         unsigned long ptr;
 834         unsigned long end;
 835         int extra_size;
 836         int type;
 837         int want;
 838         int ret;
 839         int err = 0;
 840         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
 841         int needed;
 842 
 843         key.objectid = bytenr;
 844         key.type = BTRFS_EXTENT_ITEM_KEY;
 845         key.offset = num_bytes;
 846 
 847         want = extent_ref_type(parent, owner);
 848         if (insert) {
 849                 extra_size = btrfs_extent_inline_ref_size(want);
 850                 path->keep_locks = 1;
 851         } else
 852                 extra_size = -1;
 853 
 854         /*
 855          * Owner is our level, so we can just add one to get the level for the
 856          * block we are interested in.
 857          */
 858         if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
 859                 key.type = BTRFS_METADATA_ITEM_KEY;
 860                 key.offset = owner;
 861         }
 862 
 863 again:
 864         ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
 865         if (ret < 0) {
 866                 err = ret;
 867                 goto out;
 868         }
 869 
 870         /*
 871          * We may be a newly converted file system which still has the old fat
 872          * extent entries for metadata, so try and see if we have one of those.
 873          */
 874         if (ret > 0 && skinny_metadata) {
 875                 skinny_metadata = false;
 876                 if (path->slots[0]) {
 877                         path->slots[0]--;
 878                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 879                                               path->slots[0]);
 880                         if (key.objectid == bytenr &&
 881                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 882                             key.offset == num_bytes)
 883                                 ret = 0;
 884                 }
 885                 if (ret) {
 886                         key.objectid = bytenr;
 887                         key.type = BTRFS_EXTENT_ITEM_KEY;
 888                         key.offset = num_bytes;
 889                         btrfs_release_path(path);
 890                         goto again;
 891                 }
 892         }
 893 
 894         if (ret && !insert) {
 895                 err = -ENOENT;
 896                 goto out;
 897         } else if (WARN_ON(ret)) {
 898                 err = -EIO;
 899                 goto out;
 900         }
 901 
 902         leaf = path->nodes[0];
 903         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 904         if (unlikely(item_size < sizeof(*ei))) {
 905                 err = -EINVAL;
 906                 btrfs_print_v0_err(fs_info);
 907                 btrfs_abort_transaction(trans, err);
 908                 goto out;
 909         }
 910 
 911         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 912         flags = btrfs_extent_flags(leaf, ei);
 913 
 914         ptr = (unsigned long)(ei + 1);
 915         end = (unsigned long)ei + item_size;
 916 
 917         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
 918                 ptr += sizeof(struct btrfs_tree_block_info);
 919                 BUG_ON(ptr > end);
 920         }
 921 
 922         if (owner >= BTRFS_FIRST_FREE_OBJECTID)
 923                 needed = BTRFS_REF_TYPE_DATA;
 924         else
 925                 needed = BTRFS_REF_TYPE_BLOCK;
 926 
 927         err = -ENOENT;
 928         while (1) {
 929                 if (ptr >= end) {
 930                         WARN_ON(ptr > end);
 931                         break;
 932                 }
 933                 iref = (struct btrfs_extent_inline_ref *)ptr;
 934                 type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
 935                 if (type == BTRFS_REF_TYPE_INVALID) {
 936                         err = -EUCLEAN;
 937                         goto out;
 938                 }
 939 
 940                 if (want < type)
 941                         break;
 942                 if (want > type) {
 943                         ptr += btrfs_extent_inline_ref_size(type);
 944                         continue;
 945                 }
 946 
 947                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 948                         struct btrfs_extent_data_ref *dref;
 949                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
 950                         if (match_extent_data_ref(leaf, dref, root_objectid,
 951                                                   owner, offset)) {
 952                                 err = 0;
 953                                 break;
 954                         }
 955                         if (hash_extent_data_ref_item(leaf, dref) <
 956                             hash_extent_data_ref(root_objectid, owner, offset))
 957                                 break;
 958                 } else {
 959                         u64 ref_offset;
 960                         ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
 961                         if (parent > 0) {
 962                                 if (parent == ref_offset) {
 963                                         err = 0;
 964                                         break;
 965                                 }
 966                                 if (ref_offset < parent)
 967                                         break;
 968                         } else {
 969                                 if (root_objectid == ref_offset) {
 970                                         err = 0;
 971                                         break;
 972                                 }
 973                                 if (ref_offset < root_objectid)
 974                                         break;
 975                         }
 976                 }
 977                 ptr += btrfs_extent_inline_ref_size(type);
 978         }
 979         if (err == -ENOENT && insert) {
 980                 if (item_size + extra_size >=
 981                     BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
 982                         err = -EAGAIN;
 983                         goto out;
 984                 }
 985                 /*
 986                  * To add new inline back ref, we have to make sure
 987                  * there is no corresponding back ref item.
 988                  * For simplicity, we just do not add new inline back
 989                  * ref if there is any kind of item for this block
 990                  */
 991                 if (find_next_key(path, 0, &key) == 0 &&
 992                     key.objectid == bytenr &&
 993                     key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
 994                         err = -EAGAIN;
 995                         goto out;
 996                 }
 997         }
 998         *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
 999 out:
1000         if (insert) {
1001                 path->keep_locks = 0;
1002                 btrfs_unlock_up_safe(path, 1);
1003         }
1004         return err;
1005 }
1006 
1007 /*
1008  * helper to add new inline back ref
1009  */
1010 static noinline_for_stack
1011 void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
1012                                  struct btrfs_path *path,
1013                                  struct btrfs_extent_inline_ref *iref,
1014                                  u64 parent, u64 root_objectid,
1015                                  u64 owner, u64 offset, int refs_to_add,
1016                                  struct btrfs_delayed_extent_op *extent_op)
1017 {
1018         struct extent_buffer *leaf;
1019         struct btrfs_extent_item *ei;
1020         unsigned long ptr;
1021         unsigned long end;
1022         unsigned long item_offset;
1023         u64 refs;
1024         int size;
1025         int type;
1026 
1027         leaf = path->nodes[0];
1028         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1029         item_offset = (unsigned long)iref - (unsigned long)ei;
1030 
1031         type = extent_ref_type(parent, owner);
1032         size = btrfs_extent_inline_ref_size(type);
1033 
1034         btrfs_extend_item(path, size);
1035 
1036         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1037         refs = btrfs_extent_refs(leaf, ei);
1038         refs += refs_to_add;
1039         btrfs_set_extent_refs(leaf, ei, refs);
1040         if (extent_op)
1041                 __run_delayed_extent_op(extent_op, leaf, ei);
1042 
1043         ptr = (unsigned long)ei + item_offset;
1044         end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1045         if (ptr < end - size)
1046                 memmove_extent_buffer(leaf, ptr + size, ptr,
1047                                       end - size - ptr);
1048 
1049         iref = (struct btrfs_extent_inline_ref *)ptr;
1050         btrfs_set_extent_inline_ref_type(leaf, iref, type);
1051         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1052                 struct btrfs_extent_data_ref *dref;
1053                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1054                 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1055                 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1056                 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1057                 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1058         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1059                 struct btrfs_shared_data_ref *sref;
1060                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1061                 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1062                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1063         } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1064                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1065         } else {
1066                 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1067         }
1068         btrfs_mark_buffer_dirty(leaf);
1069 }
1070 
1071 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1072                                  struct btrfs_path *path,
1073                                  struct btrfs_extent_inline_ref **ref_ret,
1074                                  u64 bytenr, u64 num_bytes, u64 parent,
1075                                  u64 root_objectid, u64 owner, u64 offset)
1076 {
1077         int ret;
1078 
1079         ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
1080                                            num_bytes, parent, root_objectid,
1081                                            owner, offset, 0);
1082         if (ret != -ENOENT)
1083                 return ret;
1084 
1085         btrfs_release_path(path);
1086         *ref_ret = NULL;
1087 
1088         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1089                 ret = lookup_tree_block_ref(trans, path, bytenr, parent,
1090                                             root_objectid);
1091         } else {
1092                 ret = lookup_extent_data_ref(trans, path, bytenr, parent,
1093                                              root_objectid, owner, offset);
1094         }
1095         return ret;
1096 }
1097 
1098 /*
1099  * helper to update/remove inline back ref
1100  */
1101 static noinline_for_stack
1102 void update_inline_extent_backref(struct btrfs_path *path,
1103                                   struct btrfs_extent_inline_ref *iref,
1104                                   int refs_to_mod,
1105                                   struct btrfs_delayed_extent_op *extent_op,
1106                                   int *last_ref)
1107 {
1108         struct extent_buffer *leaf = path->nodes[0];
1109         struct btrfs_extent_item *ei;
1110         struct btrfs_extent_data_ref *dref = NULL;
1111         struct btrfs_shared_data_ref *sref = NULL;
1112         unsigned long ptr;
1113         unsigned long end;
1114         u32 item_size;
1115         int size;
1116         int type;
1117         u64 refs;
1118 
1119         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1120         refs = btrfs_extent_refs(leaf, ei);
1121         WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1122         refs += refs_to_mod;
1123         btrfs_set_extent_refs(leaf, ei, refs);
1124         if (extent_op)
1125                 __run_delayed_extent_op(extent_op, leaf, ei);
1126 
1127         /*
1128          * If type is invalid, we should have bailed out after
1129          * lookup_inline_extent_backref().
1130          */
1131         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
1132         ASSERT(type != BTRFS_REF_TYPE_INVALID);
1133 
1134         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1135                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1136                 refs = btrfs_extent_data_ref_count(leaf, dref);
1137         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1138                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1139                 refs = btrfs_shared_data_ref_count(leaf, sref);
1140         } else {
1141                 refs = 1;
1142                 BUG_ON(refs_to_mod != -1);
1143         }
1144 
1145         BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1146         refs += refs_to_mod;
1147 
1148         if (refs > 0) {
1149                 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1150                         btrfs_set_extent_data_ref_count(leaf, dref, refs);
1151                 else
1152                         btrfs_set_shared_data_ref_count(leaf, sref, refs);
1153         } else {
1154                 *last_ref = 1;
1155                 size =  btrfs_extent_inline_ref_size(type);
1156                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1157                 ptr = (unsigned long)iref;
1158                 end = (unsigned long)ei + item_size;
1159                 if (ptr + size < end)
1160                         memmove_extent_buffer(leaf, ptr, ptr + size,
1161                                               end - ptr - size);
1162                 item_size -= size;
1163                 btrfs_truncate_item(path, item_size, 1);
1164         }
1165         btrfs_mark_buffer_dirty(leaf);
1166 }
1167 
1168 static noinline_for_stack
1169 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1170                                  struct btrfs_path *path,
1171                                  u64 bytenr, u64 num_bytes, u64 parent,
1172                                  u64 root_objectid, u64 owner,
1173                                  u64 offset, int refs_to_add,
1174                                  struct btrfs_delayed_extent_op *extent_op)
1175 {
1176         struct btrfs_extent_inline_ref *iref;
1177         int ret;
1178 
1179         ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
1180                                            num_bytes, parent, root_objectid,
1181                                            owner, offset, 1);
1182         if (ret == 0) {
1183                 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1184                 update_inline_extent_backref(path, iref, refs_to_add,
1185                                              extent_op, NULL);
1186         } else if (ret == -ENOENT) {
1187                 setup_inline_extent_backref(trans->fs_info, path, iref, parent,
1188                                             root_objectid, owner, offset,
1189                                             refs_to_add, extent_op);
1190                 ret = 0;
1191         }
1192         return ret;
1193 }
1194 
1195 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1196                                  struct btrfs_path *path,
1197                                  u64 bytenr, u64 parent, u64 root_objectid,
1198                                  u64 owner, u64 offset, int refs_to_add)
1199 {
1200         int ret;
1201         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1202                 BUG_ON(refs_to_add != 1);
1203                 ret = insert_tree_block_ref(trans, path, bytenr, parent,
1204                                             root_objectid);
1205         } else {
1206                 ret = insert_extent_data_ref(trans, path, bytenr, parent,
1207                                              root_objectid, owner, offset,
1208                                              refs_to_add);
1209         }
1210         return ret;
1211 }
1212 
1213 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1214                                  struct btrfs_path *path,
1215                                  struct btrfs_extent_inline_ref *iref,
1216                                  int refs_to_drop, int is_data, int *last_ref)
1217 {
1218         int ret = 0;
1219 
1220         BUG_ON(!is_data && refs_to_drop != 1);
1221         if (iref) {
1222                 update_inline_extent_backref(path, iref, -refs_to_drop, NULL,
1223                                              last_ref);
1224         } else if (is_data) {
1225                 ret = remove_extent_data_ref(trans, path, refs_to_drop,
1226                                              last_ref);
1227         } else {
1228                 *last_ref = 1;
1229                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
1230         }
1231         return ret;
1232 }
1233 
1234 static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
1235                                u64 *discarded_bytes)
1236 {
1237         int j, ret = 0;
1238         u64 bytes_left, end;
1239         u64 aligned_start = ALIGN(start, 1 << 9);
1240 
1241         if (WARN_ON(start != aligned_start)) {
1242                 len -= aligned_start - start;
1243                 len = round_down(len, 1 << 9);
1244                 start = aligned_start;
1245         }
1246 
1247         *discarded_bytes = 0;
1248 
1249         if (!len)
1250                 return 0;
1251 
1252         end = start + len;
1253         bytes_left = len;
1254 
1255         /* Skip any superblocks on this device. */
1256         for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
1257                 u64 sb_start = btrfs_sb_offset(j);
1258                 u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
1259                 u64 size = sb_start - start;
1260 
1261                 if (!in_range(sb_start, start, bytes_left) &&
1262                     !in_range(sb_end, start, bytes_left) &&
1263                     !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
1264                         continue;
1265 
1266                 /*
1267                  * Superblock spans beginning of range.  Adjust start and
1268                  * try again.
1269                  */
1270                 if (sb_start <= start) {
1271                         start += sb_end - start;
1272                         if (start > end) {
1273                                 bytes_left = 0;
1274                                 break;
1275                         }
1276                         bytes_left = end - start;
1277                         continue;
1278                 }
1279 
1280                 if (size) {
1281                         ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
1282                                                    GFP_NOFS, 0);
1283                         if (!ret)
1284                                 *discarded_bytes += size;
1285                         else if (ret != -EOPNOTSUPP)
1286                                 return ret;
1287                 }
1288 
1289                 start = sb_end;
1290                 if (start > end) {
1291                         bytes_left = 0;
1292                         break;
1293                 }
1294                 bytes_left = end - start;
1295         }
1296 
1297         if (bytes_left) {
1298                 ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
1299                                            GFP_NOFS, 0);
1300                 if (!ret)
1301                         *discarded_bytes += bytes_left;
1302         }
1303         return ret;
1304 }
1305 
1306 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
1307                          u64 num_bytes, u64 *actual_bytes)
1308 {
1309         int ret;
1310         u64 discarded_bytes = 0;
1311         struct btrfs_bio *bbio = NULL;
1312 
1313 
1314         /*
1315          * Avoid races with device replace and make sure our bbio has devices
1316          * associated to its stripes that don't go away while we are discarding.
1317          */
1318         btrfs_bio_counter_inc_blocked(fs_info);
1319         /* Tell the block device(s) that the sectors can be discarded */
1320         ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, bytenr, &num_bytes,
1321                               &bbio, 0);
1322         /* Error condition is -ENOMEM */
1323         if (!ret) {
1324                 struct btrfs_bio_stripe *stripe = bbio->stripes;
1325                 int i;
1326 
1327 
1328                 for (i = 0; i < bbio->num_stripes; i++, stripe++) {
1329                         u64 bytes;
1330                         struct request_queue *req_q;
1331 
1332                         if (!stripe->dev->bdev) {
1333                                 ASSERT(btrfs_test_opt(fs_info, DEGRADED));
1334                                 continue;
1335                         }
1336                         req_q = bdev_get_queue(stripe->dev->bdev);
1337                         if (!blk_queue_discard(req_q))
1338                                 continue;
1339 
1340                         ret = btrfs_issue_discard(stripe->dev->bdev,
1341                                                   stripe->physical,
1342                                                   stripe->length,
1343                                                   &bytes);
1344                         if (!ret)
1345                                 discarded_bytes += bytes;
1346                         else if (ret != -EOPNOTSUPP)
1347                                 break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */
1348 
1349                         /*
1350                          * Just in case we get back EOPNOTSUPP for some reason,
1351                          * just ignore the return value so we don't screw up
1352                          * people calling discard_extent.
1353                          */
1354                         ret = 0;
1355                 }
1356                 btrfs_put_bbio(bbio);
1357         }
1358         btrfs_bio_counter_dec(fs_info);
1359 
1360         if (actual_bytes)
1361                 *actual_bytes = discarded_bytes;
1362 
1363 
1364         if (ret == -EOPNOTSUPP)
1365                 ret = 0;
1366         return ret;
1367 }
1368 
1369 /* Can return -ENOMEM */
1370 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1371                          struct btrfs_ref *generic_ref)
1372 {
1373         struct btrfs_fs_info *fs_info = trans->fs_info;
1374         int old_ref_mod, new_ref_mod;
1375         int ret;
1376 
1377         ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
1378                generic_ref->action);
1379         BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
1380                generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
1381 
1382         if (generic_ref->type == BTRFS_REF_METADATA)
1383                 ret = btrfs_add_delayed_tree_ref(trans, generic_ref,
1384                                 NULL, &old_ref_mod, &new_ref_mod);
1385         else
1386                 ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0,
1387                                                  &old_ref_mod, &new_ref_mod);
1388 
1389         btrfs_ref_tree_mod(fs_info, generic_ref);
1390 
1391         if (ret == 0 && old_ref_mod < 0 && new_ref_mod >= 0)
1392                 sub_pinned_bytes(fs_info, generic_ref);
1393 
1394         return ret;
1395 }
1396 
1397 /*
1398  * __btrfs_inc_extent_ref - insert backreference for a given extent
1399  *
1400  * @trans:          Handle of transaction
1401  *
1402  * @node:           The delayed ref node used to get the bytenr/length for
1403  *                  extent whose references are incremented.
1404  *
1405  * @parent:         If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/
1406  *                  BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical
1407  *                  bytenr of the parent block. Since new extents are always
1408  *                  created with indirect references, this will only be the case
1409  *                  when relocating a shared extent. In that case, root_objectid
1410  *                  will be BTRFS_TREE_RELOC_OBJECTID. Otheriwse, parent must
1411  *                  be 0
1412  *
1413  * @root_objectid:  The id of the root where this modification has originated,
1414  *                  this can be either one of the well-known metadata trees or
1415  *                  the subvolume id which references this extent.
1416  *
1417  * @owner:          For data extents it is the inode number of the owning file.
1418  *                  For metadata extents this parameter holds the level in the
1419  *                  tree of the extent.
1420  *
1421  * @offset:         For metadata extents the offset is ignored and is currently
1422  *                  always passed as 0. For data extents it is the fileoffset
1423  *                  this extent belongs to.
1424  *
1425  * @refs_to_add     Number of references to add
1426  *
1427  * @extent_op       Pointer to a structure, holding information necessary when
1428  *                  updating a tree block's flags
1429  *
1430  */
1431 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1432                                   struct btrfs_delayed_ref_node *node,
1433                                   u64 parent, u64 root_objectid,
1434                                   u64 owner, u64 offset, int refs_to_add,
1435                                   struct btrfs_delayed_extent_op *extent_op)
1436 {
1437         struct btrfs_path *path;
1438         struct extent_buffer *leaf;
1439         struct btrfs_extent_item *item;
1440         struct btrfs_key key;
1441         u64 bytenr = node->bytenr;
1442         u64 num_bytes = node->num_bytes;
1443         u64 refs;
1444         int ret;
1445 
1446         path = btrfs_alloc_path();
1447         if (!path)
1448                 return -ENOMEM;
1449 
1450         path->reada = READA_FORWARD;
1451         path->leave_spinning = 1;
1452         /* this will setup the path even if it fails to insert the back ref */
1453         ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
1454                                            parent, root_objectid, owner,
1455                                            offset, refs_to_add, extent_op);
1456         if ((ret < 0 && ret != -EAGAIN) || !ret)
1457                 goto out;
1458 
1459         /*
1460          * Ok we had -EAGAIN which means we didn't have space to insert and
1461          * inline extent ref, so just update the reference count and add a
1462          * normal backref.
1463          */
1464         leaf = path->nodes[0];
1465         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1466         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1467         refs = btrfs_extent_refs(leaf, item);
1468         btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1469         if (extent_op)
1470                 __run_delayed_extent_op(extent_op, leaf, item);
1471 
1472         btrfs_mark_buffer_dirty(leaf);
1473         btrfs_release_path(path);
1474 
1475         path->reada = READA_FORWARD;
1476         path->leave_spinning = 1;
1477         /* now insert the actual backref */
1478         ret = insert_extent_backref(trans, path, bytenr, parent, root_objectid,
1479                                     owner, offset, refs_to_add);
1480         if (ret)
1481                 btrfs_abort_transaction(trans, ret);
1482 out:
1483         btrfs_free_path(path);
1484         return ret;
1485 }
1486 
1487 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1488                                 struct btrfs_delayed_ref_node *node,
1489                                 struct btrfs_delayed_extent_op *extent_op,
1490                                 int insert_reserved)
1491 {
1492         int ret = 0;
1493         struct btrfs_delayed_data_ref *ref;
1494         struct btrfs_key ins;
1495         u64 parent = 0;
1496         u64 ref_root = 0;
1497         u64 flags = 0;
1498 
1499         ins.objectid = node->bytenr;
1500         ins.offset = node->num_bytes;
1501         ins.type = BTRFS_EXTENT_ITEM_KEY;
1502 
1503         ref = btrfs_delayed_node_to_data_ref(node);
1504         trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action);
1505 
1506         if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1507                 parent = ref->parent;
1508         ref_root = ref->root;
1509 
1510         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1511                 if (extent_op)
1512                         flags |= extent_op->flags_to_set;
1513                 ret = alloc_reserved_file_extent(trans, parent, ref_root,
1514                                                  flags, ref->objectid,
1515                                                  ref->offset, &ins,
1516                                                  node->ref_mod);
1517         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1518                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1519                                              ref->objectid, ref->offset,
1520                                              node->ref_mod, extent_op);
1521         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1522                 ret = __btrfs_free_extent(trans, node, parent,
1523                                           ref_root, ref->objectid,
1524                                           ref->offset, node->ref_mod,
1525                                           extent_op);
1526         } else {
1527                 BUG();
1528         }
1529         return ret;
1530 }
1531 
1532 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1533                                     struct extent_buffer *leaf,
1534                                     struct btrfs_extent_item *ei)
1535 {
1536         u64 flags = btrfs_extent_flags(leaf, ei);
1537         if (extent_op->update_flags) {
1538                 flags |= extent_op->flags_to_set;
1539                 btrfs_set_extent_flags(leaf, ei, flags);
1540         }
1541 
1542         if (extent_op->update_key) {
1543                 struct btrfs_tree_block_info *bi;
1544                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1545                 bi = (struct btrfs_tree_block_info *)(ei + 1);
1546                 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1547         }
1548 }
1549 
1550 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1551                                  struct btrfs_delayed_ref_head *head,
1552                                  struct btrfs_delayed_extent_op *extent_op)
1553 {
1554         struct btrfs_fs_info *fs_info = trans->fs_info;
1555         struct btrfs_key key;
1556         struct btrfs_path *path;
1557         struct btrfs_extent_item *ei;
1558         struct extent_buffer *leaf;
1559         u32 item_size;
1560         int ret;
1561         int err = 0;
1562         int metadata = !extent_op->is_data;
1563 
1564         if (trans->aborted)
1565                 return 0;
1566 
1567         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1568                 metadata = 0;
1569 
1570         path = btrfs_alloc_path();
1571         if (!path)
1572                 return -ENOMEM;
1573 
1574         key.objectid = head->bytenr;
1575 
1576         if (metadata) {
1577                 key.type = BTRFS_METADATA_ITEM_KEY;
1578                 key.offset = extent_op->level;
1579         } else {
1580                 key.type = BTRFS_EXTENT_ITEM_KEY;
1581                 key.offset = head->num_bytes;
1582         }
1583 
1584 again:
1585         path->reada = READA_FORWARD;
1586         path->leave_spinning = 1;
1587         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
1588         if (ret < 0) {
1589                 err = ret;
1590                 goto out;
1591         }
1592         if (ret > 0) {
1593                 if (metadata) {
1594                         if (path->slots[0] > 0) {
1595                                 path->slots[0]--;
1596                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
1597                                                       path->slots[0]);
1598                                 if (key.objectid == head->bytenr &&
1599                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
1600                                     key.offset == head->num_bytes)
1601                                         ret = 0;
1602                         }
1603                         if (ret > 0) {
1604                                 btrfs_release_path(path);
1605                                 metadata = 0;
1606 
1607                                 key.objectid = head->bytenr;
1608                                 key.offset = head->num_bytes;
1609                                 key.type = BTRFS_EXTENT_ITEM_KEY;
1610                                 goto again;
1611                         }
1612                 } else {
1613                         err = -EIO;
1614                         goto out;
1615                 }
1616         }
1617 
1618         leaf = path->nodes[0];
1619         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1620 
1621         if (unlikely(item_size < sizeof(*ei))) {
1622                 err = -EINVAL;
1623                 btrfs_print_v0_err(fs_info);
1624                 btrfs_abort_transaction(trans, err);
1625                 goto out;
1626         }
1627 
1628         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1629         __run_delayed_extent_op(extent_op, leaf, ei);
1630 
1631         btrfs_mark_buffer_dirty(leaf);
1632 out:
1633         btrfs_free_path(path);
1634         return err;
1635 }
1636 
1637 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1638                                 struct btrfs_delayed_ref_node *node,
1639                                 struct btrfs_delayed_extent_op *extent_op,
1640                                 int insert_reserved)
1641 {
1642         int ret = 0;
1643         struct btrfs_delayed_tree_ref *ref;
1644         u64 parent = 0;
1645         u64 ref_root = 0;
1646 
1647         ref = btrfs_delayed_node_to_tree_ref(node);
1648         trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action);
1649 
1650         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1651                 parent = ref->parent;
1652         ref_root = ref->root;
1653 
1654         if (node->ref_mod != 1) {
1655                 btrfs_err(trans->fs_info,
1656         "btree block(%llu) has %d references rather than 1: action %d ref_root %llu parent %llu",
1657                           node->bytenr, node->ref_mod, node->action, ref_root,
1658                           parent);
1659                 return -EIO;
1660         }
1661         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1662                 BUG_ON(!extent_op || !extent_op->update_flags);
1663                 ret = alloc_reserved_tree_block(trans, node, extent_op);
1664         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1665                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1666                                              ref->level, 0, 1, extent_op);
1667         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1668                 ret = __btrfs_free_extent(trans, node, parent, ref_root,
1669                                           ref->level, 0, 1, extent_op);
1670         } else {
1671                 BUG();
1672         }
1673         return ret;
1674 }
1675 
1676 /* helper function to actually process a single delayed ref entry */
1677 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1678                                struct btrfs_delayed_ref_node *node,
1679                                struct btrfs_delayed_extent_op *extent_op,
1680                                int insert_reserved)
1681 {
1682         int ret = 0;
1683 
1684         if (trans->aborted) {
1685                 if (insert_reserved)
1686                         btrfs_pin_extent(trans->fs_info, node->bytenr,
1687                                          node->num_bytes, 1);
1688                 return 0;
1689         }
1690 
1691         if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1692             node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1693                 ret = run_delayed_tree_ref(trans, node, extent_op,
1694                                            insert_reserved);
1695         else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1696                  node->type == BTRFS_SHARED_DATA_REF_KEY)
1697                 ret = run_delayed_data_ref(trans, node, extent_op,
1698                                            insert_reserved);
1699         else
1700                 BUG();
1701         if (ret && insert_reserved)
1702                 btrfs_pin_extent(trans->fs_info, node->bytenr,
1703                                  node->num_bytes, 1);
1704         return ret;
1705 }
1706 
1707 static inline struct btrfs_delayed_ref_node *
1708 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1709 {
1710         struct btrfs_delayed_ref_node *ref;
1711 
1712         if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
1713                 return NULL;
1714 
1715         /*
1716          * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
1717          * This is to prevent a ref count from going down to zero, which deletes
1718          * the extent item from the extent tree, when there still are references
1719          * to add, which would fail because they would not find the extent item.
1720          */
1721         if (!list_empty(&head->ref_add_list))
1722                 return list_first_entry(&head->ref_add_list,
1723                                 struct btrfs_delayed_ref_node, add_list);
1724 
1725         ref = rb_entry(rb_first_cached(&head->ref_tree),
1726                        struct btrfs_delayed_ref_node, ref_node);
1727         ASSERT(list_empty(&ref->add_list));
1728         return ref;
1729 }
1730 
1731 static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
1732                                       struct btrfs_delayed_ref_head *head)
1733 {
1734         spin_lock(&delayed_refs->lock);
1735         head->processing = 0;
1736         delayed_refs->num_heads_ready++;
1737         spin_unlock(&delayed_refs->lock);
1738         btrfs_delayed_ref_unlock(head);
1739 }
1740 
1741 static struct btrfs_delayed_extent_op *cleanup_extent_op(
1742                                 struct btrfs_delayed_ref_head *head)
1743 {
1744         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
1745 
1746         if (!extent_op)
1747                 return NULL;
1748 
1749         if (head->must_insert_reserved) {
1750                 head->extent_op = NULL;
1751                 btrfs_free_delayed_extent_op(extent_op);
1752                 return NULL;
1753         }
1754         return extent_op;
1755 }
1756 
1757 static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
1758                                      struct btrfs_delayed_ref_head *head)
1759 {
1760         struct btrfs_delayed_extent_op *extent_op;
1761         int ret;
1762 
1763         extent_op = cleanup_extent_op(head);
1764         if (!extent_op)
1765                 return 0;
1766         head->extent_op = NULL;
1767         spin_unlock(&head->lock);
1768         ret = run_delayed_extent_op(trans, head, extent_op);
1769         btrfs_free_delayed_extent_op(extent_op);
1770         return ret ? ret : 1;
1771 }
1772 
1773 void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
1774                                   struct btrfs_delayed_ref_root *delayed_refs,
1775                                   struct btrfs_delayed_ref_head *head)
1776 {
1777         int nr_items = 1;       /* Dropping this ref head update. */
1778 
1779         if (head->total_ref_mod < 0) {
1780                 struct btrfs_space_info *space_info;
1781                 u64 flags;
1782 
1783                 if (head->is_data)
1784                         flags = BTRFS_BLOCK_GROUP_DATA;
1785                 else if (head->is_system)
1786                         flags = BTRFS_BLOCK_GROUP_SYSTEM;
1787                 else
1788                         flags = BTRFS_BLOCK_GROUP_METADATA;
1789                 space_info = btrfs_find_space_info(fs_info, flags);
1790                 ASSERT(space_info);
1791                 percpu_counter_add_batch(&space_info->total_bytes_pinned,
1792                                    -head->num_bytes,
1793                                    BTRFS_TOTAL_BYTES_PINNED_BATCH);
1794 
1795                 /*
1796                  * We had csum deletions accounted for in our delayed refs rsv,
1797                  * we need to drop the csum leaves for this update from our
1798                  * delayed_refs_rsv.
1799                  */
1800                 if (head->is_data) {
1801                         spin_lock(&delayed_refs->lock);
1802                         delayed_refs->pending_csums -= head->num_bytes;
1803                         spin_unlock(&delayed_refs->lock);
1804                         nr_items += btrfs_csum_bytes_to_leaves(fs_info,
1805                                 head->num_bytes);
1806                 }
1807         }
1808 
1809         btrfs_delayed_refs_rsv_release(fs_info, nr_items);
1810 }
1811 
1812 static int cleanup_ref_head(struct btrfs_trans_handle *trans,
1813                             struct btrfs_delayed_ref_head *head)
1814 {
1815 
1816         struct btrfs_fs_info *fs_info = trans->fs_info;
1817         struct btrfs_delayed_ref_root *delayed_refs;
1818         int ret;
1819 
1820         delayed_refs = &trans->transaction->delayed_refs;
1821 
1822         ret = run_and_cleanup_extent_op(trans, head);
1823         if (ret < 0) {
1824                 unselect_delayed_ref_head(delayed_refs, head);
1825                 btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
1826                 return ret;
1827         } else if (ret) {
1828                 return ret;
1829         }
1830 
1831         /*
1832          * Need to drop our head ref lock and re-acquire the delayed ref lock
1833          * and then re-check to make sure nobody got added.
1834          */
1835         spin_unlock(&head->lock);
1836         spin_lock(&delayed_refs->lock);
1837         spin_lock(&head->lock);
1838         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
1839                 spin_unlock(&head->lock);
1840                 spin_unlock(&delayed_refs->lock);
1841                 return 1;
1842         }
1843         btrfs_delete_ref_head(delayed_refs, head);
1844         spin_unlock(&head->lock);
1845         spin_unlock(&delayed_refs->lock);
1846 
1847         if (head->must_insert_reserved) {
1848                 btrfs_pin_extent(fs_info, head->bytenr,
1849                                  head->num_bytes, 1);
1850                 if (head->is_data) {
1851                         ret = btrfs_del_csums(trans, fs_info->csum_root,
1852                                               head->bytenr, head->num_bytes);
1853                 }
1854         }
1855 
1856         btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
1857 
1858         trace_run_delayed_ref_head(fs_info, head, 0);
1859         btrfs_delayed_ref_unlock(head);
1860         btrfs_put_delayed_ref_head(head);
1861         return 0;
1862 }
1863 
1864 static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
1865                                         struct btrfs_trans_handle *trans)
1866 {
1867         struct btrfs_delayed_ref_root *delayed_refs =
1868                 &trans->transaction->delayed_refs;
1869         struct btrfs_delayed_ref_head *head = NULL;
1870         int ret;
1871 
1872         spin_lock(&delayed_refs->lock);
1873         head = btrfs_select_ref_head(delayed_refs);
1874         if (!head) {
1875                 spin_unlock(&delayed_refs->lock);
1876                 return head;
1877         }
1878 
1879         /*
1880          * Grab the lock that says we are going to process all the refs for
1881          * this head
1882          */
1883         ret = btrfs_delayed_ref_lock(delayed_refs, head);
1884         spin_unlock(&delayed_refs->lock);
1885 
1886         /*
1887          * We may have dropped the spin lock to get the head mutex lock, and
1888          * that might have given someone else time to free the head.  If that's
1889          * true, it has been removed from our list and we can move on.
1890          */
1891         if (ret == -EAGAIN)
1892                 head = ERR_PTR(-EAGAIN);
1893 
1894         return head;
1895 }
1896 
1897 static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
1898                                     struct btrfs_delayed_ref_head *locked_ref,
1899                                     unsigned long *run_refs)
1900 {
1901         struct btrfs_fs_info *fs_info = trans->fs_info;
1902         struct btrfs_delayed_ref_root *delayed_refs;
1903         struct btrfs_delayed_extent_op *extent_op;
1904         struct btrfs_delayed_ref_node *ref;
1905         int must_insert_reserved = 0;
1906         int ret;
1907 
1908         delayed_refs = &trans->transaction->delayed_refs;
1909 
1910         lockdep_assert_held(&locked_ref->mutex);
1911         lockdep_assert_held(&locked_ref->lock);
1912 
1913         while ((ref = select_delayed_ref(locked_ref))) {
1914                 if (ref->seq &&
1915                     btrfs_check_delayed_seq(fs_info, ref->seq)) {
1916                         spin_unlock(&locked_ref->lock);
1917                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1918                         return -EAGAIN;
1919                 }
1920 
1921                 (*run_refs)++;
1922                 ref->in_tree = 0;
1923                 rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
1924                 RB_CLEAR_NODE(&ref->ref_node);
1925                 if (!list_empty(&ref->add_list))
1926                         list_del(&ref->add_list);
1927                 /*
1928                  * When we play the delayed ref, also correct the ref_mod on
1929                  * head
1930                  */
1931                 switch (ref->action) {
1932                 case BTRFS_ADD_DELAYED_REF:
1933                 case BTRFS_ADD_DELAYED_EXTENT:
1934                         locked_ref->ref_mod -= ref->ref_mod;
1935                         break;
1936                 case BTRFS_DROP_DELAYED_REF:
1937                         locked_ref->ref_mod += ref->ref_mod;
1938                         break;
1939                 default:
1940                         WARN_ON(1);
1941                 }
1942                 atomic_dec(&delayed_refs->num_entries);
1943 
1944                 /*
1945                  * Record the must_insert_reserved flag before we drop the
1946                  * spin lock.
1947                  */
1948                 must_insert_reserved = locked_ref->must_insert_reserved;
1949                 locked_ref->must_insert_reserved = 0;
1950 
1951                 extent_op = locked_ref->extent_op;
1952                 locked_ref->extent_op = NULL;
1953                 spin_unlock(&locked_ref->lock);
1954 
1955                 ret = run_one_delayed_ref(trans, ref, extent_op,
1956                                           must_insert_reserved);
1957 
1958                 btrfs_free_delayed_extent_op(extent_op);
1959                 if (ret) {
1960                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1961                         btrfs_put_delayed_ref(ref);
1962                         btrfs_debug(fs_info, "run_one_delayed_ref returned %d",
1963                                     ret);
1964                         return ret;
1965                 }
1966 
1967                 btrfs_put_delayed_ref(ref);
1968                 cond_resched();
1969 
1970                 spin_lock(&locked_ref->lock);
1971                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
1972         }
1973 
1974         return 0;
1975 }
1976 
1977 /*
1978  * Returns 0 on success or if called with an already aborted transaction.
1979  * Returns -ENOMEM or -EIO on failure and will abort the transaction.
1980  */
1981 static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
1982                                              unsigned long nr)
1983 {
1984         struct btrfs_fs_info *fs_info = trans->fs_info;
1985         struct btrfs_delayed_ref_root *delayed_refs;
1986         struct btrfs_delayed_ref_head *locked_ref = NULL;
1987         ktime_t start = ktime_get();
1988         int ret;
1989         unsigned long count = 0;
1990         unsigned long actual_count = 0;
1991 
1992         delayed_refs = &trans->transaction->delayed_refs;
1993         do {
1994                 if (!locked_ref) {
1995                         locked_ref = btrfs_obtain_ref_head(trans);
1996                         if (IS_ERR_OR_NULL(locked_ref)) {
1997                                 if (PTR_ERR(locked_ref) == -EAGAIN) {
1998                                         continue;
1999                                 } else {
2000                                         break;
2001                                 }
2002                         }
2003                         count++;
2004                 }
2005                 /*
2006                  * We need to try and merge add/drops of the same ref since we
2007                  * can run into issues with relocate dropping the implicit ref
2008                  * and then it being added back again before the drop can
2009                  * finish.  If we merged anything we need to re-loop so we can
2010                  * get a good ref.
2011                  * Or we can get node references of the same type that weren't
2012                  * merged when created due to bumps in the tree mod seq, and
2013                  * we need to merge them to prevent adding an inline extent
2014                  * backref before dropping it (triggering a BUG_ON at
2015                  * insert_inline_extent_backref()).
2016                  */
2017                 spin_lock(&locked_ref->lock);
2018                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
2019 
2020                 ret = btrfs_run_delayed_refs_for_head(trans, locked_ref,
2021                                                       &actual_count);
2022                 if (ret < 0 && ret != -EAGAIN) {
2023                         /*
2024                          * Error, btrfs_run_delayed_refs_for_head already
2025                          * unlocked everything so just bail out
2026                          */
2027                         return ret;
2028                 } else if (!ret) {
2029                         /*
2030                          * Success, perform the usual cleanup of a processed
2031                          * head
2032                          */
2033                         ret = cleanup_ref_head(trans, locked_ref);
2034                         if (ret > 0 ) {
2035                                 /* We dropped our lock, we need to loop. */
2036                                 ret = 0;
2037                                 continue;
2038                         } else if (ret) {
2039                                 return ret;
2040                         }
2041                 }
2042 
2043                 /*
2044                  * Either success case or btrfs_run_delayed_refs_for_head
2045                  * returned -EAGAIN, meaning we need to select another head
2046                  */
2047 
2048                 locked_ref = NULL;
2049                 cond_resched();
2050         } while ((nr != -1 && count < nr) || locked_ref);
2051 
2052         /*
2053          * We don't want to include ref heads since we can have empty ref heads
2054          * and those will drastically skew our runtime down since we just do
2055          * accounting, no actual extent tree updates.
2056          */
2057         if (actual_count > 0) {
2058                 u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
2059                 u64 avg;
2060 
2061                 /*
2062                  * We weigh the current average higher than our current runtime
2063                  * to avoid large swings in the average.
2064                  */
2065                 spin_lock(&delayed_refs->lock);
2066                 avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
2067                 fs_info->avg_delayed_ref_runtime = avg >> 2;    /* div by 4 */
2068                 spin_unlock(&delayed_refs->lock);
2069         }
2070         return 0;
2071 }
2072 
2073 #ifdef SCRAMBLE_DELAYED_REFS
2074 /*
2075  * Normally delayed refs get processed in ascending bytenr order. This
2076  * correlates in most cases to the order added. To expose dependencies on this
2077  * order, we start to process the tree in the middle instead of the beginning
2078  */
2079 static u64 find_middle(struct rb_root *root)
2080 {
2081         struct rb_node *n = root->rb_node;
2082         struct btrfs_delayed_ref_node *entry;
2083         int alt = 1;
2084         u64 middle;
2085         u64 first = 0, last = 0;
2086 
2087         n = rb_first(root);
2088         if (n) {
2089                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2090                 first = entry->bytenr;
2091         }
2092         n = rb_last(root);
2093         if (n) {
2094                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2095                 last = entry->bytenr;
2096         }
2097         n = root->rb_node;
2098 
2099         while (n) {
2100                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2101                 WARN_ON(!entry->in_tree);
2102 
2103                 middle = entry->bytenr;
2104 
2105                 if (alt)
2106                         n = n->rb_left;
2107                 else
2108                         n = n->rb_right;
2109 
2110                 alt = 1 - alt;
2111         }
2112         return middle;
2113 }
2114 #endif
2115 
2116 static inline u64 heads_to_leaves(struct btrfs_fs_info *fs_info, u64 heads)
2117 {
2118         u64 num_bytes;
2119 
2120         num_bytes = heads * (sizeof(struct btrfs_extent_item) +
2121                              sizeof(struct btrfs_extent_inline_ref));
2122         if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA))
2123                 num_bytes += heads * sizeof(struct btrfs_tree_block_info);
2124 
2125         /*
2126          * We don't ever fill up leaves all the way so multiply by 2 just to be
2127          * closer to what we're really going to want to use.
2128          */
2129         return div_u64(num_bytes, BTRFS_LEAF_DATA_SIZE(fs_info));
2130 }
2131 
2132 /*
2133  * Takes the number of bytes to be csumm'ed and figures out how many leaves it
2134  * would require to store the csums for that many bytes.
2135  */
2136 u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
2137 {
2138         u64 csum_size;
2139         u64 num_csums_per_leaf;
2140         u64 num_csums;
2141 
2142         csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
2143         num_csums_per_leaf = div64_u64(csum_size,
2144                         (u64)btrfs_super_csum_size(fs_info->super_copy));
2145         num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
2146         num_csums += num_csums_per_leaf - 1;
2147         num_csums = div64_u64(num_csums, num_csums_per_leaf);
2148         return num_csums;
2149 }
2150 
2151 /*
2152  * this starts processing the delayed reference count updates and
2153  * extent insertions we have queued up so far.  count can be
2154  * 0, which means to process everything in the tree at the start
2155  * of the run (but not newly added entries), or it can be some target
2156  * number you'd like to process.
2157  *
2158  * Returns 0 on success or if called with an aborted transaction
2159  * Returns <0 on error and aborts the transaction
2160  */
2161 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2162                            unsigned long count)
2163 {
2164         struct btrfs_fs_info *fs_info = trans->fs_info;
2165         struct rb_node *node;
2166         struct btrfs_delayed_ref_root *delayed_refs;
2167         struct btrfs_delayed_ref_head *head;
2168         int ret;
2169         int run_all = count == (unsigned long)-1;
2170 
2171         /* We'll clean this up in btrfs_cleanup_transaction */
2172         if (trans->aborted)
2173                 return 0;
2174 
2175         if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
2176                 return 0;
2177 
2178         delayed_refs = &trans->transaction->delayed_refs;
2179         if (count == 0)
2180                 count = atomic_read(&delayed_refs->num_entries) * 2;
2181 
2182 again:
2183 #ifdef SCRAMBLE_DELAYED_REFS
2184         delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
2185 #endif
2186         ret = __btrfs_run_delayed_refs(trans, count);
2187         if (ret < 0) {
2188                 btrfs_abort_transaction(trans, ret);
2189                 return ret;
2190         }
2191 
2192         if (run_all) {
2193                 btrfs_create_pending_block_groups(trans);
2194 
2195                 spin_lock(&delayed_refs->lock);
2196                 node = rb_first_cached(&delayed_refs->href_root);
2197                 if (!node) {
2198                         spin_unlock(&delayed_refs->lock);
2199                         goto out;
2200                 }
2201                 head = rb_entry(node, struct btrfs_delayed_ref_head,
2202                                 href_node);
2203                 refcount_inc(&head->refs);
2204                 spin_unlock(&delayed_refs->lock);
2205 
2206                 /* Mutex was contended, block until it's released and retry. */
2207                 mutex_lock(&head->mutex);
2208                 mutex_unlock(&head->mutex);
2209 
2210                 btrfs_put_delayed_ref_head(head);
2211                 cond_resched();
2212                 goto again;
2213         }
2214 out:
2215         return 0;
2216 }
2217 
2218 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2219                                 u64 bytenr, u64 num_bytes, u64 flags,
2220                                 int level, int is_data)
2221 {
2222         struct btrfs_delayed_extent_op *extent_op;
2223         int ret;
2224 
2225         extent_op = btrfs_alloc_delayed_extent_op();
2226         if (!extent_op)
2227                 return -ENOMEM;
2228 
2229         extent_op->flags_to_set = flags;
2230         extent_op->update_flags = true;
2231         extent_op->update_key = false;
2232         extent_op->is_data = is_data ? true : false;
2233         extent_op->level = level;
2234 
2235         ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
2236         if (ret)
2237                 btrfs_free_delayed_extent_op(extent_op);
2238         return ret;
2239 }
2240 
2241 static noinline int check_delayed_ref(struct btrfs_root *root,
2242                                       struct btrfs_path *path,
2243                                       u64 objectid, u64 offset, u64 bytenr)
2244 {
2245         struct btrfs_delayed_ref_head *head;
2246         struct btrfs_delayed_ref_node *ref;
2247         struct btrfs_delayed_data_ref *data_ref;
2248         struct btrfs_delayed_ref_root *delayed_refs;
2249         struct btrfs_transaction *cur_trans;
2250         struct rb_node *node;
2251         int ret = 0;
2252 
2253         spin_lock(&root->fs_info->trans_lock);
2254         cur_trans = root->fs_info->running_transaction;
2255         if (cur_trans)
2256                 refcount_inc(&cur_trans->use_count);
2257         spin_unlock(&root->fs_info->trans_lock);
2258         if (!cur_trans)
2259                 return 0;
2260 
2261         delayed_refs = &cur_trans->delayed_refs;
2262         spin_lock(&delayed_refs->lock);
2263         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
2264         if (!head) {
2265                 spin_unlock(&delayed_refs->lock);
2266                 btrfs_put_transaction(cur_trans);
2267                 return 0;
2268         }
2269 
2270         if (!mutex_trylock(&head->mutex)) {
2271                 refcount_inc(&head->refs);
2272                 spin_unlock(&delayed_refs->lock);
2273 
2274                 btrfs_release_path(path);
2275 
2276                 /*
2277                  * Mutex was contended, block until it's released and let
2278                  * caller try again
2279                  */
2280                 mutex_lock(&head->mutex);
2281                 mutex_unlock(&head->mutex);
2282                 btrfs_put_delayed_ref_head(head);
2283                 btrfs_put_transaction(cur_trans);
2284                 return -EAGAIN;
2285         }
2286         spin_unlock(&delayed_refs->lock);
2287 
2288         spin_lock(&head->lock);
2289         /*
2290          * XXX: We should replace this with a proper search function in the
2291          * future.
2292          */
2293         for (node = rb_first_cached(&head->ref_tree); node;
2294              node = rb_next(node)) {
2295                 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
2296                 /* If it's a shared ref we know a cross reference exists */
2297                 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
2298                         ret = 1;
2299                         break;
2300                 }
2301 
2302                 data_ref = btrfs_delayed_node_to_data_ref(ref);
2303 
2304                 /*
2305                  * If our ref doesn't match the one we're currently looking at
2306                  * then we have a cross reference.
2307                  */
2308                 if (data_ref->root != root->root_key.objectid ||
2309                     data_ref->objectid != objectid ||
2310                     data_ref->offset != offset) {
2311                         ret = 1;
2312                         break;
2313                 }
2314         }
2315         spin_unlock(&head->lock);
2316         mutex_unlock(&head->mutex);
2317         btrfs_put_transaction(cur_trans);
2318         return ret;
2319 }
2320 
2321 static noinline int check_committed_ref(struct btrfs_root *root,
2322                                         struct btrfs_path *path,
2323                                         u64 objectid, u64 offset, u64 bytenr)
2324 {
2325         struct btrfs_fs_info *fs_info = root->fs_info;
2326         struct btrfs_root *extent_root = fs_info->extent_root;
2327         struct extent_buffer *leaf;
2328         struct btrfs_extent_data_ref *ref;
2329         struct btrfs_extent_inline_ref *iref;
2330         struct btrfs_extent_item *ei;
2331         struct btrfs_key key;
2332         u32 item_size;
2333         int type;
2334         int ret;
2335 
2336         key.objectid = bytenr;
2337         key.offset = (u64)-1;
2338         key.type = BTRFS_EXTENT_ITEM_KEY;
2339 
2340         ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2341         if (ret < 0)
2342                 goto out;
2343         BUG_ON(ret == 0); /* Corruption */
2344 
2345         ret = -ENOENT;
2346         if (path->slots[0] == 0)
2347                 goto out;
2348 
2349         path->slots[0]--;
2350         leaf = path->nodes[0];
2351         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2352 
2353         if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2354                 goto out;
2355 
2356         ret = 1;
2357         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2358         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2359 
2360         /* If extent item has more than 1 inline ref then it's shared */
2361         if (item_size != sizeof(*ei) +
2362             btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2363                 goto out;
2364 
2365         /* If extent created before last snapshot => it's definitely shared */
2366         if (btrfs_extent_generation(leaf, ei) <=
2367             btrfs_root_last_snapshot(&root->root_item))
2368                 goto out;
2369 
2370         iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2371 
2372         /* If this extent has SHARED_DATA_REF then it's shared */
2373         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
2374         if (type != BTRFS_EXTENT_DATA_REF_KEY)
2375                 goto out;
2376 
2377         ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2378         if (btrfs_extent_refs(leaf, ei) !=
2379             btrfs_extent_data_ref_count(leaf, ref) ||
2380             btrfs_extent_data_ref_root(leaf, ref) !=
2381             root->root_key.objectid ||
2382             btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2383             btrfs_extent_data_ref_offset(leaf, ref) != offset)
2384                 goto out;
2385 
2386         ret = 0;
2387 out:
2388         return ret;
2389 }
2390 
2391 int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
2392                           u64 bytenr)
2393 {
2394         struct btrfs_path *path;
2395         int ret;
2396 
2397         path = btrfs_alloc_path();
2398         if (!path)
2399                 return -ENOMEM;
2400 
2401         do {
2402                 ret = check_committed_ref(root, path, objectid,
2403                                           offset, bytenr);
2404                 if (ret && ret != -ENOENT)
2405                         goto out;
2406 
2407                 ret = check_delayed_ref(root, path, objectid, offset, bytenr);
2408         } while (ret == -EAGAIN);
2409 
2410 out:
2411         btrfs_free_path(path);
2412         if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
2413                 WARN_ON(ret > 0);
2414         return ret;
2415 }
2416 
2417 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2418                            struct btrfs_root *root,
2419                            struct extent_buffer *buf,
2420                            int full_backref, int inc)
2421 {
2422         struct btrfs_fs_info *fs_info = root->fs_info;
2423         u64 bytenr;
2424         u64 num_bytes;
2425         u64 parent;
2426         u64 ref_root;
2427         u32 nritems;
2428         struct btrfs_key key;
2429         struct btrfs_file_extent_item *fi;
2430         struct btrfs_ref generic_ref = { 0 };
2431         bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
2432         int i;
2433         int action;
2434         int level;
2435         int ret = 0;
2436 
2437         if (btrfs_is_testing(fs_info))
2438                 return 0;
2439 
2440         ref_root = btrfs_header_owner(buf);
2441         nritems = btrfs_header_nritems(buf);
2442         level = btrfs_header_level(buf);
2443 
2444         if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state) && level == 0)
2445                 return 0;
2446 
2447         if (full_backref)
2448                 parent = buf->start;
2449         else
2450                 parent = 0;
2451         if (inc)
2452                 action = BTRFS_ADD_DELAYED_REF;
2453         else
2454                 action = BTRFS_DROP_DELAYED_REF;
2455 
2456         for (i = 0; i < nritems; i++) {
2457                 if (level == 0) {
2458                         btrfs_item_key_to_cpu(buf, &key, i);
2459                         if (key.type != BTRFS_EXTENT_DATA_KEY)
2460                                 continue;
2461                         fi = btrfs_item_ptr(buf, i,
2462                                             struct btrfs_file_extent_item);
2463                         if (btrfs_file_extent_type(buf, fi) ==
2464                             BTRFS_FILE_EXTENT_INLINE)
2465                                 continue;
2466                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2467                         if (bytenr == 0)
2468                                 continue;
2469 
2470                         num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2471                         key.offset -= btrfs_file_extent_offset(buf, fi);
2472                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2473                                                num_bytes, parent);
2474                         generic_ref.real_root = root->root_key.objectid;
2475                         btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
2476                                             key.offset);
2477                         generic_ref.skip_qgroup = for_reloc;
2478                         if (inc)
2479                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2480                         else
2481                                 ret = btrfs_free_extent(trans, &generic_ref);
2482                         if (ret)
2483                                 goto fail;
2484                 } else {
2485                         bytenr = btrfs_node_blockptr(buf, i);
2486                         num_bytes = fs_info->nodesize;
2487                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2488                                                num_bytes, parent);
2489                         generic_ref.real_root = root->root_key.objectid;
2490                         btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
2491                         generic_ref.skip_qgroup = for_reloc;
2492                         if (inc)
2493                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2494                         else
2495                                 ret = btrfs_free_extent(trans, &generic_ref);
2496                         if (ret)
2497                                 goto fail;
2498                 }
2499         }
2500         return 0;
2501 fail:
2502         return ret;
2503 }
2504 
2505 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2506                   struct extent_buffer *buf, int full_backref)
2507 {
2508         return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2509 }
2510 
2511 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2512                   struct extent_buffer *buf, int full_backref)
2513 {
2514         return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2515 }
2516 
2517 int btrfs_extent_readonly(struct btrfs_fs_info *fs_info, u64 bytenr)
2518 {
2519         struct btrfs_block_group_cache *block_group;
2520         int readonly = 0;
2521 
2522         block_group = btrfs_lookup_block_group(fs_info, bytenr);
2523         if (!block_group || block_group->ro)
2524                 readonly = 1;
2525         if (block_group)
2526                 btrfs_put_block_group(block_group);
2527         return readonly;
2528 }
2529 
2530 static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
2531 {
2532         struct btrfs_fs_info *fs_info = root->fs_info;
2533         u64 flags;
2534         u64 ret;
2535 
2536         if (data)
2537                 flags = BTRFS_BLOCK_GROUP_DATA;
2538         else if (root == fs_info->chunk_root)
2539                 flags = BTRFS_BLOCK_GROUP_SYSTEM;
2540         else
2541                 flags = BTRFS_BLOCK_GROUP_METADATA;
2542 
2543         ret = btrfs_get_alloc_profile(fs_info, flags);
2544         return ret;
2545 }
2546 
2547 static u64 first_logical_byte(struct btrfs_fs_info *fs_info, u64 search_start)
2548 {
2549         struct btrfs_block_group_cache *cache;
2550         u64 bytenr;
2551 
2552         spin_lock(&fs_info->block_group_cache_lock);
2553         bytenr = fs_info->first_logical_byte;
2554         spin_unlock(&fs_info->block_group_cache_lock);
2555 
2556         if (bytenr < (u64)-1)
2557                 return bytenr;
2558 
2559         cache = btrfs_lookup_first_block_group(fs_info, search_start);
2560         if (!cache)
2561                 return 0;
2562 
2563         bytenr = cache->key.objectid;
2564         btrfs_put_block_group(cache);
2565 
2566         return bytenr;
2567 }
2568 
2569 static int pin_down_extent(struct btrfs_block_group_cache *cache,
2570                            u64 bytenr, u64 num_bytes, int reserved)
2571 {
2572         struct btrfs_fs_info *fs_info = cache->fs_info;
2573 
2574         spin_lock(&cache->space_info->lock);
2575         spin_lock(&cache->lock);
2576         cache->pinned += num_bytes;
2577         btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
2578                                              num_bytes);
2579         if (reserved) {
2580                 cache->reserved -= num_bytes;
2581                 cache->space_info->bytes_reserved -= num_bytes;
2582         }
2583         spin_unlock(&cache->lock);
2584         spin_unlock(&cache->space_info->lock);
2585 
2586         percpu_counter_add_batch(&cache->space_info->total_bytes_pinned,
2587                     num_bytes, BTRFS_TOTAL_BYTES_PINNED_BATCH);
2588         set_extent_dirty(fs_info->pinned_extents, bytenr,
2589                          bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
2590         return 0;
2591 }
2592 
2593 /*
2594  * this function must be called within transaction
2595  */
2596 int btrfs_pin_extent(struct btrfs_fs_info *fs_info,
2597                      u64 bytenr, u64 num_bytes, int reserved)
2598 {
2599         struct btrfs_block_group_cache *cache;
2600 
2601         cache = btrfs_lookup_block_group(fs_info, bytenr);
2602         BUG_ON(!cache); /* Logic error */
2603 
2604         pin_down_extent(cache, bytenr, num_bytes, reserved);
2605 
2606         btrfs_put_block_group(cache);
2607         return 0;
2608 }
2609 
2610 /*
2611  * this function must be called within transaction
2612  */
2613 int btrfs_pin_extent_for_log_replay(struct btrfs_fs_info *fs_info,
2614                                     u64 bytenr, u64 num_bytes)
2615 {
2616         struct btrfs_block_group_cache *cache;
2617         int ret;
2618 
2619         cache = btrfs_lookup_block_group(fs_info, bytenr);
2620         if (!cache)
2621                 return -EINVAL;
2622 
2623         /*
2624          * pull in the free space cache (if any) so that our pin
2625          * removes the free space from the cache.  We have load_only set
2626          * to one because the slow code to read in the free extents does check
2627          * the pinned extents.
2628          */
2629         btrfs_cache_block_group(cache, 1);
2630 
2631         pin_down_extent(cache, bytenr, num_bytes, 0);
2632 
2633         /* remove us from the free space cache (if we're there at all) */
2634         ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
2635         btrfs_put_block_group(cache);
2636         return ret;
2637 }
2638 
2639 static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
2640                                    u64 start, u64 num_bytes)
2641 {
2642         int ret;
2643         struct btrfs_block_group_cache *block_group;
2644         struct btrfs_caching_control *caching_ctl;
2645 
2646         block_group = btrfs_lookup_block_group(fs_info, start);
2647         if (!block_group)
2648                 return -EINVAL;
2649 
2650         btrfs_cache_block_group(block_group, 0);
2651         caching_ctl = btrfs_get_caching_control(block_group);
2652 
2653         if (!caching_ctl) {
2654                 /* Logic error */
2655                 BUG_ON(!btrfs_block_group_cache_done(block_group));
2656                 ret = btrfs_remove_free_space(block_group, start, num_bytes);
2657         } else {
2658                 mutex_lock(&caching_ctl->mutex);
2659 
2660                 if (start >= caching_ctl->progress) {
2661                         ret = btrfs_add_excluded_extent(fs_info, start,
2662                                                         num_bytes);
2663                 } else if (start + num_bytes <= caching_ctl->progress) {
2664                         ret = btrfs_remove_free_space(block_group,
2665                                                       start, num_bytes);
2666                 } else {
2667                         num_bytes = caching_ctl->progress - start;
2668                         ret = btrfs_remove_free_space(block_group,
2669                                                       start, num_bytes);
2670                         if (ret)
2671                                 goto out_lock;
2672 
2673                         num_bytes = (start + num_bytes) -
2674                                 caching_ctl->progress;
2675                         start = caching_ctl->progress;
2676                         ret = btrfs_add_excluded_extent(fs_info, start,
2677                                                         num_bytes);
2678                 }
2679 out_lock:
2680                 mutex_unlock(&caching_ctl->mutex);
2681                 btrfs_put_caching_control(caching_ctl);
2682         }
2683         btrfs_put_block_group(block_group);
2684         return ret;
2685 }
2686 
2687 int btrfs_exclude_logged_extents(struct extent_buffer *eb)
2688 {
2689         struct btrfs_fs_info *fs_info = eb->fs_info;
2690         struct btrfs_file_extent_item *item;
2691         struct btrfs_key key;
2692         int found_type;
2693         int i;
2694         int ret = 0;
2695 
2696         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
2697                 return 0;
2698 
2699         for (i = 0; i < btrfs_header_nritems(eb); i++) {
2700                 btrfs_item_key_to_cpu(eb, &key, i);
2701                 if (key.type != BTRFS_EXTENT_DATA_KEY)
2702                         continue;
2703                 item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
2704                 found_type = btrfs_file_extent_type(eb, item);
2705                 if (found_type == BTRFS_FILE_EXTENT_INLINE)
2706                         continue;
2707                 if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
2708                         continue;
2709                 key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
2710                 key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
2711                 ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
2712                 if (ret)
2713                         break;
2714         }
2715 
2716         return ret;
2717 }
2718 
2719 static void
2720 btrfs_inc_block_group_reservations(struct btrfs_block_group_cache *bg)
2721 {
2722         atomic_inc(&bg->reservations);
2723 }
2724 
2725 void btrfs_prepare_extent_commit(struct btrfs_fs_info *fs_info)
2726 {
2727         struct btrfs_caching_control *next;
2728         struct btrfs_caching_control *caching_ctl;
2729         struct btrfs_block_group_cache *cache;
2730 
2731         down_write(&fs_info->commit_root_sem);
2732 
2733         list_for_each_entry_safe(caching_ctl, next,
2734                                  &fs_info->caching_block_groups, list) {
2735                 cache = caching_ctl->block_group;
2736                 if (btrfs_block_group_cache_done(cache)) {
2737                         cache->last_byte_to_unpin = (u64)-1;
2738                         list_del_init(&caching_ctl->list);
2739                         btrfs_put_caching_control(caching_ctl);
2740                 } else {
2741                         cache->last_byte_to_unpin = caching_ctl->progress;
2742                 }
2743         }
2744 
2745         if (fs_info->pinned_extents == &fs_info->freed_extents[0])
2746                 fs_info->pinned_extents = &fs_info->freed_extents[1];
2747         else
2748                 fs_info->pinned_extents = &fs_info->freed_extents[0];
2749 
2750         up_write(&fs_info->commit_root_sem);
2751 
2752         btrfs_update_global_block_rsv(fs_info);
2753 }
2754 
2755 /*
2756  * Returns the free cluster for the given space info and sets empty_cluster to
2757  * what it should be based on the mount options.
2758  */
2759 static struct btrfs_free_cluster *
2760 fetch_cluster_info(struct btrfs_fs_info *fs_info,
2761                    struct btrfs_space_info *space_info, u64 *empty_cluster)
2762 {
2763         struct btrfs_free_cluster *ret = NULL;
2764 
2765         *empty_cluster = 0;
2766         if (btrfs_mixed_space_info(space_info))
2767                 return ret;
2768 
2769         if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
2770                 ret = &fs_info->meta_alloc_cluster;
2771                 if (btrfs_test_opt(fs_info, SSD))
2772                         *empty_cluster = SZ_2M;
2773                 else
2774                         *empty_cluster = SZ_64K;
2775         } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
2776                    btrfs_test_opt(fs_info, SSD_SPREAD)) {
2777                 *empty_cluster = SZ_2M;
2778                 ret = &fs_info->data_alloc_cluster;
2779         }
2780 
2781         return ret;
2782 }
2783 
2784 static int unpin_extent_range(struct btrfs_fs_info *fs_info,
2785                               u64 start, u64 end,
2786                               const bool return_free_space)
2787 {
2788         struct btrfs_block_group_cache *cache = NULL;
2789         struct btrfs_space_info *space_info;
2790         struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
2791         struct btrfs_free_cluster *cluster = NULL;
2792         u64 len;
2793         u64 total_unpinned = 0;
2794         u64 empty_cluster = 0;
2795         bool readonly;
2796 
2797         while (start <= end) {
2798                 readonly = false;
2799                 if (!cache ||
2800                     start >= cache->key.objectid + cache->key.offset) {
2801                         if (cache)
2802                                 btrfs_put_block_group(cache);
2803                         total_unpinned = 0;
2804                         cache = btrfs_lookup_block_group(fs_info, start);
2805                         BUG_ON(!cache); /* Logic error */
2806 
2807                         cluster = fetch_cluster_info(fs_info,
2808                                                      cache->space_info,
2809                                                      &empty_cluster);
2810                         empty_cluster <<= 1;
2811                 }
2812 
2813                 len = cache->key.objectid + cache->key.offset - start;
2814                 len = min(len, end + 1 - start);
2815 
2816                 if (start < cache->last_byte_to_unpin) {
2817                         len = min(len, cache->last_byte_to_unpin - start);
2818                         if (return_free_space)
2819                                 btrfs_add_free_space(cache, start, len);
2820                 }
2821 
2822                 start += len;
2823                 total_unpinned += len;
2824                 space_info = cache->space_info;
2825 
2826                 /*
2827                  * If this space cluster has been marked as fragmented and we've
2828                  * unpinned enough in this block group to potentially allow a
2829                  * cluster to be created inside of it go ahead and clear the
2830                  * fragmented check.
2831                  */
2832                 if (cluster && cluster->fragmented &&
2833                     total_unpinned > empty_cluster) {
2834                         spin_lock(&cluster->lock);
2835                         cluster->fragmented = 0;
2836                         spin_unlock(&cluster->lock);
2837                 }
2838 
2839                 spin_lock(&space_info->lock);
2840                 spin_lock(&cache->lock);
2841                 cache->pinned -= len;
2842                 btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
2843                 space_info->max_extent_size = 0;
2844                 percpu_counter_add_batch(&space_info->total_bytes_pinned,
2845                             -len, BTRFS_TOTAL_BYTES_PINNED_BATCH);
2846                 if (cache->ro) {
2847                         space_info->bytes_readonly += len;
2848                         readonly = true;
2849                 }
2850                 spin_unlock(&cache->lock);
2851                 if (!readonly && return_free_space &&
2852                     global_rsv->space_info == space_info) {
2853                         u64 to_add = len;
2854 
2855                         spin_lock(&global_rsv->lock);
2856                         if (!global_rsv->full) {
2857                                 to_add = min(len, global_rsv->size -
2858                                              global_rsv->reserved);
2859                                 global_rsv->reserved += to_add;
2860                                 btrfs_space_info_update_bytes_may_use(fs_info,
2861                                                 space_info, to_add);
2862                                 if (global_rsv->reserved >= global_rsv->size)
2863                                         global_rsv->full = 1;
2864                                 len -= to_add;
2865                         }
2866                         spin_unlock(&global_rsv->lock);
2867                         /* Add to any tickets we may have */
2868                         if (len)
2869                                 btrfs_try_granting_tickets(fs_info,
2870                                                            space_info);
2871                 }
2872                 spin_unlock(&space_info->lock);
2873         }
2874 
2875         if (cache)
2876                 btrfs_put_block_group(cache);
2877         return 0;
2878 }
2879 
2880 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
2881 {
2882         struct btrfs_fs_info *fs_info = trans->fs_info;
2883         struct btrfs_block_group_cache *block_group, *tmp;
2884         struct list_head *deleted_bgs;
2885         struct extent_io_tree *unpin;
2886         u64 start;
2887         u64 end;
2888         int ret;
2889 
2890         if (fs_info->pinned_extents == &fs_info->freed_extents[0])
2891                 unpin = &fs_info->freed_extents[1];
2892         else
2893                 unpin = &fs_info->freed_extents[0];
2894 
2895         while (!trans->aborted) {
2896                 struct extent_state *cached_state = NULL;
2897 
2898                 mutex_lock(&fs_info->unused_bg_unpin_mutex);
2899                 ret = find_first_extent_bit(unpin, 0, &start, &end,
2900                                             EXTENT_DIRTY, &cached_state);
2901                 if (ret) {
2902                         mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2903                         break;
2904                 }
2905 
2906                 if (btrfs_test_opt(fs_info, DISCARD))
2907                         ret = btrfs_discard_extent(fs_info, start,
2908                                                    end + 1 - start, NULL);
2909 
2910                 clear_extent_dirty(unpin, start, end, &cached_state);
2911                 unpin_extent_range(fs_info, start, end, true);
2912                 mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2913                 free_extent_state(cached_state);
2914                 cond_resched();
2915         }
2916 
2917         /*
2918          * Transaction is finished.  We don't need the lock anymore.  We
2919          * do need to clean up the block groups in case of a transaction
2920          * abort.
2921          */
2922         deleted_bgs = &trans->transaction->deleted_bgs;
2923         list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
2924                 u64 trimmed = 0;
2925 
2926                 ret = -EROFS;
2927                 if (!trans->aborted)
2928                         ret = btrfs_discard_extent(fs_info,
2929                                                    block_group->key.objectid,
2930                                                    block_group->key.offset,
2931                                                    &trimmed);
2932 
2933                 list_del_init(&block_group->bg_list);
2934                 btrfs_put_block_group_trimming(block_group);
2935                 btrfs_put_block_group(block_group);
2936 
2937                 if (ret) {
2938                         const char *errstr = btrfs_decode_error(ret);
2939                         btrfs_warn(fs_info,
2940                            "discard failed while removing blockgroup: errno=%d %s",
2941                                    ret, errstr);
2942                 }
2943         }
2944 
2945         return 0;
2946 }
2947 
2948 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
2949                                struct btrfs_delayed_ref_node *node, u64 parent,
2950                                u64 root_objectid, u64 owner_objectid,
2951                                u64 owner_offset, int refs_to_drop,
2952                                struct btrfs_delayed_extent_op *extent_op)
2953 {
2954         struct btrfs_fs_info *info = trans->fs_info;
2955         struct btrfs_key key;
2956         struct btrfs_path *path;
2957         struct btrfs_root *extent_root = info->extent_root;
2958         struct extent_buffer *leaf;
2959         struct btrfs_extent_item *ei;
2960         struct btrfs_extent_inline_ref *iref;
2961         int ret;
2962         int is_data;
2963         int extent_slot = 0;
2964         int found_extent = 0;
2965         int num_to_del = 1;
2966         u32 item_size;
2967         u64 refs;
2968         u64 bytenr = node->bytenr;
2969         u64 num_bytes = node->num_bytes;
2970         int last_ref = 0;
2971         bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
2972 
2973         path = btrfs_alloc_path();
2974         if (!path)
2975                 return -ENOMEM;
2976 
2977         path->reada = READA_FORWARD;
2978         path->leave_spinning = 1;
2979 
2980         is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
2981         BUG_ON(!is_data && refs_to_drop != 1);
2982 
2983         if (is_data)
2984                 skinny_metadata = false;
2985 
2986         ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
2987                                     parent, root_objectid, owner_objectid,
2988                                     owner_offset);
2989         if (ret == 0) {
2990                 extent_slot = path->slots[0];
2991                 while (extent_slot >= 0) {
2992                         btrfs_item_key_to_cpu(path->nodes[0], &key,
2993                                               extent_slot);
2994                         if (key.objectid != bytenr)
2995                                 break;
2996                         if (key.type == BTRFS_EXTENT_ITEM_KEY &&
2997                             key.offset == num_bytes) {
2998                                 found_extent = 1;
2999                                 break;
3000                         }
3001                         if (key.type == BTRFS_METADATA_ITEM_KEY &&
3002                             key.offset == owner_objectid) {
3003                                 found_extent = 1;
3004                                 break;
3005                         }
3006                         if (path->slots[0] - extent_slot > 5)
3007                                 break;
3008                         extent_slot--;
3009                 }
3010 
3011                 if (!found_extent) {
3012                         BUG_ON(iref);
3013                         ret = remove_extent_backref(trans, path, NULL,
3014                                                     refs_to_drop,
3015                                                     is_data, &last_ref);
3016                         if (ret) {
3017                                 btrfs_abort_transaction(trans, ret);
3018                                 goto out;
3019                         }
3020                         btrfs_release_path(path);
3021                         path->leave_spinning = 1;
3022 
3023                         key.objectid = bytenr;
3024                         key.type = BTRFS_EXTENT_ITEM_KEY;
3025                         key.offset = num_bytes;
3026 
3027                         if (!is_data && skinny_metadata) {
3028                                 key.type = BTRFS_METADATA_ITEM_KEY;
3029                                 key.offset = owner_objectid;
3030                         }
3031 
3032                         ret = btrfs_search_slot(trans, extent_root,
3033                                                 &key, path, -1, 1);
3034                         if (ret > 0 && skinny_metadata && path->slots[0]) {
3035                                 /*
3036                                  * Couldn't find our skinny metadata item,
3037                                  * see if we have ye olde extent item.
3038                                  */
3039                                 path->slots[0]--;
3040                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
3041                                                       path->slots[0]);
3042                                 if (key.objectid == bytenr &&
3043                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
3044                                     key.offset == num_bytes)
3045                                         ret = 0;
3046                         }
3047 
3048                         if (ret > 0 && skinny_metadata) {
3049                                 skinny_metadata = false;
3050                                 key.objectid = bytenr;
3051                                 key.type = BTRFS_EXTENT_ITEM_KEY;
3052                                 key.offset = num_bytes;
3053                                 btrfs_release_path(path);
3054                                 ret = btrfs_search_slot(trans, extent_root,
3055                                                         &key, path, -1, 1);
3056                         }
3057 
3058                         if (ret) {
3059                                 btrfs_err(info,
3060                                           "umm, got %d back from search, was looking for %llu",
3061                                           ret, bytenr);
3062                                 if (ret > 0)
3063                                         btrfs_print_leaf(path->nodes[0]);
3064                         }
3065                         if (ret < 0) {
3066                                 btrfs_abort_transaction(trans, ret);
3067                                 goto out;
3068                         }
3069                         extent_slot = path->slots[0];
3070                 }
3071         } else if (WARN_ON(ret == -ENOENT)) {
3072                 btrfs_print_leaf(path->nodes[0]);
3073                 btrfs_err(info,
3074                         "unable to find ref byte nr %llu parent %llu root %llu  owner %llu offset %llu",
3075                         bytenr, parent, root_objectid, owner_objectid,
3076                         owner_offset);
3077                 btrfs_abort_transaction(trans, ret);
3078                 goto out;
3079         } else {
3080                 btrfs_abort_transaction(trans, ret);
3081                 goto out;
3082         }
3083 
3084         leaf = path->nodes[0];
3085         item_size = btrfs_item_size_nr(leaf, extent_slot);
3086         if (unlikely(item_size < sizeof(*ei))) {
3087                 ret = -EINVAL;
3088                 btrfs_print_v0_err(info);
3089                 btrfs_abort_transaction(trans, ret);
3090                 goto out;
3091         }
3092         ei = btrfs_item_ptr(leaf, extent_slot,
3093                             struct btrfs_extent_item);
3094         if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
3095             key.type == BTRFS_EXTENT_ITEM_KEY) {
3096                 struct btrfs_tree_block_info *bi;
3097                 BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
3098                 bi = (struct btrfs_tree_block_info *)(ei + 1);
3099                 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3100         }
3101 
3102         refs = btrfs_extent_refs(leaf, ei);
3103         if (refs < refs_to_drop) {
3104                 btrfs_err(info,
3105                           "trying to drop %d refs but we only have %Lu for bytenr %Lu",
3106                           refs_to_drop, refs, bytenr);
3107                 ret = -EINVAL;
3108                 btrfs_abort_transaction(trans, ret);
3109                 goto out;
3110         }
3111         refs -= refs_to_drop;
3112 
3113         if (refs > 0) {
3114                 if (extent_op)
3115                         __run_delayed_extent_op(extent_op, leaf, ei);
3116                 /*
3117                  * In the case of inline back ref, reference count will
3118                  * be updated by remove_extent_backref
3119                  */
3120                 if (iref) {
3121                         BUG_ON(!found_extent);
3122                 } else {
3123                         btrfs_set_extent_refs(leaf, ei, refs);
3124                         btrfs_mark_buffer_dirty(leaf);
3125                 }
3126                 if (found_extent) {
3127                         ret = remove_extent_backref(trans, path, iref,
3128                                                     refs_to_drop, is_data,
3129                                                     &last_ref);
3130                         if (ret) {
3131                                 btrfs_abort_transaction(trans, ret);
3132                                 goto out;
3133                         }
3134                 }
3135         } else {
3136                 if (found_extent) {
3137                         BUG_ON(is_data && refs_to_drop !=
3138                                extent_data_ref_count(path, iref));
3139                         if (iref) {
3140                                 BUG_ON(path->slots[0] != extent_slot);
3141                         } else {
3142                                 BUG_ON(path->slots[0] != extent_slot + 1);
3143                                 path->slots[0] = extent_slot;
3144                                 num_to_del = 2;
3145                         }
3146                 }
3147 
3148                 last_ref = 1;
3149                 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3150                                       num_to_del);
3151                 if (ret) {
3152                         btrfs_abort_transaction(trans, ret);
3153                         goto out;
3154                 }
3155                 btrfs_release_path(path);
3156 
3157                 if (is_data) {
3158                         ret = btrfs_del_csums(trans, info->csum_root, bytenr,
3159                                               num_bytes);
3160                         if (ret) {
3161                                 btrfs_abort_transaction(trans, ret);
3162                                 goto out;
3163                         }
3164                 }
3165 
3166                 ret = add_to_free_space_tree(trans, bytenr, num_bytes);
3167                 if (ret) {
3168                         btrfs_abort_transaction(trans, ret);
3169                         goto out;
3170                 }
3171 
3172                 ret = btrfs_update_block_group(trans, bytenr, num_bytes, 0);
3173                 if (ret) {
3174                         btrfs_abort_transaction(trans, ret);
3175                         goto out;
3176                 }
3177         }
3178         btrfs_release_path(path);
3179 
3180 out:
3181         btrfs_free_path(path);
3182         return ret;
3183 }
3184 
3185 /*
3186  * when we free an block, it is possible (and likely) that we free the last
3187  * delayed ref for that extent as well.  This searches the delayed ref tree for
3188  * a given extent, and if there are no other delayed refs to be processed, it
3189  * removes it from the tree.
3190  */
3191 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3192                                       u64 bytenr)
3193 {
3194         struct btrfs_delayed_ref_head *head;
3195         struct btrfs_delayed_ref_root *delayed_refs;
3196         int ret = 0;
3197 
3198         delayed_refs = &trans->transaction->delayed_refs;
3199         spin_lock(&delayed_refs->lock);
3200         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
3201         if (!head)
3202                 goto out_delayed_unlock;
3203 
3204         spin_lock(&head->lock);
3205         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
3206                 goto out;
3207 
3208         if (cleanup_extent_op(head) != NULL)
3209                 goto out;
3210 
3211         /*
3212          * waiting for the lock here would deadlock.  If someone else has it
3213          * locked they are already in the process of dropping it anyway
3214          */
3215         if (!mutex_trylock(&head->mutex))
3216                 goto out;
3217 
3218         btrfs_delete_ref_head(delayed_refs, head);
3219         head->processing = 0;
3220 
3221         spin_unlock(&head->lock);
3222         spin_unlock(&delayed_refs->lock);
3223 
3224         BUG_ON(head->extent_op);
3225         if (head->must_insert_reserved)
3226                 ret = 1;
3227 
3228         btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
3229         mutex_unlock(&head->mutex);
3230         btrfs_put_delayed_ref_head(head);
3231         return ret;
3232 out:
3233         spin_unlock(&head->lock);
3234 
3235 out_delayed_unlock:
3236         spin_unlock(&delayed_refs->lock);
3237         return 0;
3238 }
3239 
3240 void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
3241                            struct btrfs_root *root,
3242                            struct extent_buffer *buf,
3243                            u64 parent, int last_ref)
3244 {
3245         struct btrfs_fs_info *fs_info = root->fs_info;
3246         struct btrfs_ref generic_ref = { 0 };
3247         int pin = 1;
3248         int ret;
3249 
3250         btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
3251                                buf->start, buf->len, parent);
3252         btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
3253                             root->root_key.objectid);
3254 
3255         if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3256                 int old_ref_mod, new_ref_mod;
3257 
3258                 btrfs_ref_tree_mod(fs_info, &generic_ref);
3259                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL,
3260                                                  &old_ref_mod, &new_ref_mod);
3261                 BUG_ON(ret); /* -ENOMEM */
3262                 pin = old_ref_mod >= 0 && new_ref_mod < 0;
3263         }
3264 
3265         if (last_ref && btrfs_header_generation(buf) == trans->transid) {
3266                 struct btrfs_block_group_cache *cache;
3267 
3268                 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3269                         ret = check_ref_cleanup(trans, buf->start);
3270                         if (!ret)
3271                                 goto out;
3272                 }
3273 
3274                 pin = 0;
3275                 cache = btrfs_lookup_block_group(fs_info, buf->start);
3276 
3277                 if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3278                         pin_down_extent(cache, buf->start, buf->len, 1);
3279                         btrfs_put_block_group(cache);
3280                         goto out;
3281                 }
3282 
3283                 WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
3284 
3285                 btrfs_add_free_space(cache, buf->start, buf->len);
3286                 btrfs_free_reserved_bytes(cache, buf->len, 0);
3287                 btrfs_put_block_group(cache);
3288                 trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
3289         }
3290 out:
3291         if (pin)
3292                 add_pinned_bytes(fs_info, &generic_ref);
3293 
3294         if (last_ref) {
3295                 /*
3296                  * Deleting the buffer, clear the corrupt flag since it doesn't
3297                  * matter anymore.
3298                  */
3299                 clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
3300         }
3301 }
3302 
3303 /* Can return -ENOMEM */
3304 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
3305 {
3306         struct btrfs_fs_info *fs_info = trans->fs_info;
3307         int old_ref_mod, new_ref_mod;
3308         int ret;
3309 
3310         if (btrfs_is_testing(fs_info))
3311                 return 0;
3312 
3313         /*
3314          * tree log blocks never actually go into the extent allocation
3315          * tree, just update pinning info and exit early.
3316          */
3317         if ((ref->type == BTRFS_REF_METADATA &&
3318              ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3319             (ref->type == BTRFS_REF_DATA &&
3320              ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
3321                 /* unlocks the pinned mutex */
3322                 btrfs_pin_extent(fs_info, ref->bytenr, ref->len, 1);
3323                 old_ref_mod = new_ref_mod = 0;
3324                 ret = 0;
3325         } else if (ref->type == BTRFS_REF_METADATA) {
3326                 ret = btrfs_add_delayed_tree_ref(trans, ref, NULL,
3327                                                  &old_ref_mod, &new_ref_mod);
3328         } else {
3329                 ret = btrfs_add_delayed_data_ref(trans, ref, 0,
3330                                                  &old_ref_mod, &new_ref_mod);
3331         }
3332 
3333         if (!((ref->type == BTRFS_REF_METADATA &&
3334                ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3335               (ref->type == BTRFS_REF_DATA &&
3336                ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
3337                 btrfs_ref_tree_mod(fs_info, ref);
3338 
3339         if (ret == 0 && old_ref_mod >= 0 && new_ref_mod < 0)
3340                 add_pinned_bytes(fs_info, ref);
3341 
3342         return ret;
3343 }
3344 
3345 enum btrfs_loop_type {
3346         LOOP_CACHING_NOWAIT,
3347         LOOP_CACHING_WAIT,
3348         LOOP_ALLOC_CHUNK,
3349         LOOP_NO_EMPTY_SIZE,
3350 };
3351 
3352 static inline void
3353 btrfs_lock_block_group(struct btrfs_block_group_cache *cache,
3354                        int delalloc)
3355 {
3356         if (delalloc)
3357                 down_read(&cache->data_rwsem);
3358 }
3359 
3360 static inline void
3361 btrfs_grab_block_group(struct btrfs_block_group_cache *cache,
3362                        int delalloc)
3363 {
3364         btrfs_get_block_group(cache);
3365         if (delalloc)
3366                 down_read(&cache->data_rwsem);
3367 }
3368 
3369 static struct btrfs_block_group_cache *
3370 btrfs_lock_cluster(struct btrfs_block_group_cache *block_group,
3371                    struct btrfs_free_cluster *cluster,
3372                    int delalloc)
3373 {
3374         struct btrfs_block_group_cache *used_bg = NULL;
3375 
3376         spin_lock(&cluster->refill_lock);
3377         while (1) {
3378                 used_bg = cluster->block_group;
3379                 if (!used_bg)
3380                         return NULL;
3381 
3382                 if (used_bg == block_group)
3383                         return used_bg;
3384 
3385                 btrfs_get_block_group(used_bg);
3386 
3387                 if (!delalloc)
3388                         return used_bg;
3389 
3390                 if (down_read_trylock(&used_bg->data_rwsem))
3391                         return used_bg;
3392 
3393                 spin_unlock(&cluster->refill_lock);
3394 
3395                 /* We should only have one-level nested. */
3396                 down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
3397 
3398                 spin_lock(&cluster->refill_lock);
3399                 if (used_bg == cluster->block_group)
3400                         return used_bg;
3401 
3402                 up_read(&used_bg->data_rwsem);
3403                 btrfs_put_block_group(used_bg);
3404         }
3405 }
3406 
3407 static inline void
3408 btrfs_release_block_group(struct btrfs_block_group_cache *cache,
3409                          int delalloc)
3410 {
3411         if (delalloc)
3412                 up_read(&cache->data_rwsem);
3413         btrfs_put_block_group(cache);
3414 }
3415 
3416 /*
3417  * Structure used internally for find_free_extent() function.  Wraps needed
3418  * parameters.
3419  */
3420 struct find_free_extent_ctl {
3421         /* Basic allocation info */
3422         u64 ram_bytes;
3423         u64 num_bytes;
3424         u64 empty_size;
3425         u64 flags;
3426         int delalloc;
3427 
3428         /* Where to start the search inside the bg */
3429         u64 search_start;
3430 
3431         /* For clustered allocation */
3432         u64 empty_cluster;
3433 
3434         bool have_caching_bg;
3435         bool orig_have_caching_bg;
3436 
3437         /* RAID index, converted from flags */
3438         int index;
3439 
3440         /*
3441          * Current loop number, check find_free_extent_update_loop() for details
3442          */
3443         int loop;
3444 
3445         /*
3446          * Whether we're refilling a cluster, if true we need to re-search
3447          * current block group but don't try to refill the cluster again.
3448          */
3449         bool retry_clustered;
3450 
3451         /*
3452          * Whether we're updating free space cache, if true we need to re-search
3453          * current block group but don't try updating free space cache again.
3454          */
3455         bool retry_unclustered;
3456 
3457         /* If current block group is cached */
3458         int cached;
3459 
3460         /* Max contiguous hole found */
3461         u64 max_extent_size;
3462 
3463         /* Total free space from free space cache, not always contiguous */
3464         u64 total_free_space;
3465 
3466         /* Found result */
3467         u64 found_offset;
3468 };
3469 
3470 
3471 /*
3472  * Helper function for find_free_extent().
3473  *
3474  * Return -ENOENT to inform caller that we need fallback to unclustered mode.
3475  * Return -EAGAIN to inform caller that we need to re-search this block group
3476  * Return >0 to inform caller that we find nothing
3477  * Return 0 means we have found a location and set ffe_ctl->found_offset.
3478  */
3479 static int find_free_extent_clustered(struct btrfs_block_group_cache *bg,
3480                 struct btrfs_free_cluster *last_ptr,
3481                 struct find_free_extent_ctl *ffe_ctl,
3482                 struct btrfs_block_group_cache **cluster_bg_ret)
3483 {
3484         struct btrfs_block_group_cache *cluster_bg;
3485         u64 aligned_cluster;
3486         u64 offset;
3487         int ret;
3488 
3489         cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
3490         if (!cluster_bg)
3491                 goto refill_cluster;
3492         if (cluster_bg != bg && (cluster_bg->ro ||
3493             !block_group_bits(cluster_bg, ffe_ctl->flags)))
3494                 goto release_cluster;
3495 
3496         offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
3497                         ffe_ctl->num_bytes, cluster_bg->key.objectid,
3498                         &ffe_ctl->max_extent_size);
3499         if (offset) {
3500                 /* We have a block, we're done */
3501                 spin_unlock(&last_ptr->refill_lock);
3502                 trace_btrfs_reserve_extent_cluster(cluster_bg,
3503                                 ffe_ctl->search_start, ffe_ctl->num_bytes);
3504                 *cluster_bg_ret = cluster_bg;
3505                 ffe_ctl->found_offset = offset;
3506                 return 0;
3507         }
3508         WARN_ON(last_ptr->block_group != cluster_bg);
3509 
3510 release_cluster:
3511         /*
3512          * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
3513          * lets just skip it and let the allocator find whatever block it can
3514          * find. If we reach this point, we will have tried the cluster
3515          * allocator plenty of times and not have found anything, so we are
3516          * likely way too fragmented for the clustering stuff to find anything.
3517          *
3518          * However, if the cluster is taken from the current block group,
3519          * release the cluster first, so that we stand a better chance of
3520          * succeeding in the unclustered allocation.
3521          */
3522         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
3523                 spin_unlock(&last_ptr->refill_lock);
3524                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3525                 return -ENOENT;
3526         }
3527 
3528         /* This cluster didn't work out, free it and start over */
3529         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3530 
3531         if (cluster_bg != bg)
3532                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3533 
3534 refill_cluster:
3535         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
3536                 spin_unlock(&last_ptr->refill_lock);
3537                 return -ENOENT;
3538         }
3539 
3540         aligned_cluster = max_t(u64,
3541                         ffe_ctl->empty_cluster + ffe_ctl->empty_size,
3542                         bg->full_stripe_len);
3543         ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
3544                         ffe_ctl->num_bytes, aligned_cluster);
3545         if (ret == 0) {
3546                 /* Now pull our allocation out of this cluster */
3547                 offset = btrfs_alloc_from_cluster(bg, last_ptr,
3548                                 ffe_ctl->num_bytes, ffe_ctl->search_start,
3549                                 &ffe_ctl->max_extent_size);
3550                 if (offset) {
3551                         /* We found one, proceed */
3552                         spin_unlock(&last_ptr->refill_lock);
3553                         trace_btrfs_reserve_extent_cluster(bg,
3554                                         ffe_ctl->search_start,
3555                                         ffe_ctl->num_bytes);
3556                         ffe_ctl->found_offset = offset;
3557                         return 0;
3558                 }
3559         } else if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
3560                    !ffe_ctl->retry_clustered) {
3561                 spin_unlock(&last_ptr->refill_lock);
3562 
3563                 ffe_ctl->retry_clustered = true;
3564                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3565                                 ffe_ctl->empty_cluster + ffe_ctl->empty_size);
3566                 return -EAGAIN;
3567         }
3568         /*
3569          * At this point we either didn't find a cluster or we weren't able to
3570          * allocate a block from our cluster.  Free the cluster we've been
3571          * trying to use, and go to the next block group.
3572          */
3573         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3574         spin_unlock(&last_ptr->refill_lock);
3575         return 1;
3576 }
3577 
3578 /*
3579  * Return >0 to inform caller that we find nothing
3580  * Return 0 when we found an free extent and set ffe_ctrl->found_offset
3581  * Return -EAGAIN to inform caller that we need to re-search this block group
3582  */
3583 static int find_free_extent_unclustered(struct btrfs_block_group_cache *bg,
3584                 struct btrfs_free_cluster *last_ptr,
3585                 struct find_free_extent_ctl *ffe_ctl)
3586 {
3587         u64 offset;
3588 
3589         /*
3590          * We are doing an unclustered allocation, set the fragmented flag so
3591          * we don't bother trying to setup a cluster again until we get more
3592          * space.
3593          */
3594         if (unlikely(last_ptr)) {
3595                 spin_lock(&last_ptr->lock);
3596                 last_ptr->fragmented = 1;
3597                 spin_unlock(&last_ptr->lock);
3598         }
3599         if (ffe_ctl->cached) {
3600                 struct btrfs_free_space_ctl *free_space_ctl;
3601 
3602                 free_space_ctl = bg->free_space_ctl;
3603                 spin_lock(&free_space_ctl->tree_lock);
3604                 if (free_space_ctl->free_space <
3605                     ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
3606                     ffe_ctl->empty_size) {
3607                         ffe_ctl->total_free_space = max_t(u64,
3608                                         ffe_ctl->total_free_space,
3609                                         free_space_ctl->free_space);
3610                         spin_unlock(&free_space_ctl->tree_lock);
3611                         return 1;
3612                 }
3613                 spin_unlock(&free_space_ctl->tree_lock);
3614         }
3615 
3616         offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
3617                         ffe_ctl->num_bytes, ffe_ctl->empty_size,
3618                         &ffe_ctl->max_extent_size);
3619 
3620         /*
3621          * If we didn't find a chunk, and we haven't failed on this block group
3622          * before, and this block group is in the middle of caching and we are
3623          * ok with waiting, then go ahead and wait for progress to be made, and
3624          * set @retry_unclustered to true.
3625          *
3626          * If @retry_unclustered is true then we've already waited on this
3627          * block group once and should move on to the next block group.
3628          */
3629         if (!offset && !ffe_ctl->retry_unclustered && !ffe_ctl->cached &&
3630             ffe_ctl->loop > LOOP_CACHING_NOWAIT) {
3631                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3632                                                       ffe_ctl->empty_size);
3633                 ffe_ctl->retry_unclustered = true;
3634                 return -EAGAIN;
3635         } else if (!offset) {
3636                 return 1;
3637         }
3638         ffe_ctl->found_offset = offset;
3639         return 0;
3640 }
3641 
3642 /*
3643  * Return >0 means caller needs to re-search for free extent
3644  * Return 0 means we have the needed free extent.
3645  * Return <0 means we failed to locate any free extent.
3646  */
3647 static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
3648                                         struct btrfs_free_cluster *last_ptr,
3649                                         struct btrfs_key *ins,
3650                                         struct find_free_extent_ctl *ffe_ctl,
3651                                         int full_search, bool use_cluster)
3652 {
3653         struct btrfs_root *root = fs_info->extent_root;
3654         int ret;
3655 
3656         if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
3657             ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
3658                 ffe_ctl->orig_have_caching_bg = true;
3659 
3660         if (!ins->objectid && ffe_ctl->loop >= LOOP_CACHING_WAIT &&
3661             ffe_ctl->have_caching_bg)
3662                 return 1;
3663 
3664         if (!ins->objectid && ++(ffe_ctl->index) < BTRFS_NR_RAID_TYPES)
3665                 return 1;
3666 
3667         if (ins->objectid) {
3668                 if (!use_cluster && last_ptr) {
3669                         spin_lock(&last_ptr->lock);
3670                         last_ptr->window_start = ins->objectid;
3671                         spin_unlock(&last_ptr->lock);
3672                 }
3673                 return 0;
3674         }
3675 
3676         /*
3677          * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
3678          *                      caching kthreads as we move along
3679          * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
3680          * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
3681          * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
3682          *                     again
3683          */
3684         if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
3685                 ffe_ctl->index = 0;
3686                 if (ffe_ctl->loop == LOOP_CACHING_NOWAIT) {
3687                         /*
3688                          * We want to skip the LOOP_CACHING_WAIT step if we
3689                          * don't have any uncached bgs and we've already done a
3690                          * full search through.
3691                          */
3692                         if (ffe_ctl->orig_have_caching_bg || !full_search)
3693                                 ffe_ctl->loop = LOOP_CACHING_WAIT;
3694                         else
3695                                 ffe_ctl->loop = LOOP_ALLOC_CHUNK;
3696                 } else {
3697                         ffe_ctl->loop++;
3698                 }
3699 
3700                 if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
3701                         struct btrfs_trans_handle *trans;
3702                         int exist = 0;
3703 
3704                         trans = current->journal_info;
3705                         if (trans)
3706                                 exist = 1;
3707                         else
3708                                 trans = btrfs_join_transaction(root);
3709 
3710                         if (IS_ERR(trans)) {
3711                                 ret = PTR_ERR(trans);
3712                                 return ret;
3713                         }
3714 
3715                         ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
3716                                                 CHUNK_ALLOC_FORCE);
3717 
3718                         /*
3719                          * If we can't allocate a new chunk we've already looped
3720                          * through at least once, move on to the NO_EMPTY_SIZE
3721                          * case.
3722                          */
3723                         if (ret == -ENOSPC)
3724                                 ffe_ctl->loop = LOOP_NO_EMPTY_SIZE;
3725 
3726                         /* Do not bail out on ENOSPC since we can do more. */
3727                         if (ret < 0 && ret != -ENOSPC)
3728                                 btrfs_abort_transaction(trans, ret);
3729                         else
3730                                 ret = 0;
3731                         if (!exist)
3732                                 btrfs_end_transaction(trans);
3733                         if (ret)
3734                                 return ret;
3735                 }
3736 
3737                 if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
3738                         /*
3739                          * Don't loop again if we already have no empty_size and
3740                          * no empty_cluster.
3741                          */
3742                         if (ffe_ctl->empty_size == 0 &&
3743                             ffe_ctl->empty_cluster == 0)
3744                                 return -ENOSPC;
3745                         ffe_ctl->empty_size = 0;
3746                         ffe_ctl->empty_cluster = 0;
3747                 }
3748                 return 1;
3749         }
3750         return -ENOSPC;
3751 }
3752 
3753 /*
3754  * walks the btree of allocated extents and find a hole of a given size.
3755  * The key ins is changed to record the hole:
3756  * ins->objectid == start position
3757  * ins->flags = BTRFS_EXTENT_ITEM_KEY
3758  * ins->offset == the size of the hole.
3759  * Any available blocks before search_start are skipped.
3760  *
3761  * If there is no suitable free space, we will record the max size of
3762  * the free space extent currently.
3763  *
3764  * The overall logic and call chain:
3765  *
3766  * find_free_extent()
3767  * |- Iterate through all block groups
3768  * |  |- Get a valid block group
3769  * |  |- Try to do clustered allocation in that block group
3770  * |  |- Try to do unclustered allocation in that block group
3771  * |  |- Check if the result is valid
3772  * |  |  |- If valid, then exit
3773  * |  |- Jump to next block group
3774  * |
3775  * |- Push harder to find free extents
3776  *    |- If not found, re-iterate all block groups
3777  */
3778 static noinline int find_free_extent(struct btrfs_fs_info *fs_info,
3779                                 u64 ram_bytes, u64 num_bytes, u64 empty_size,
3780                                 u64 hint_byte, struct btrfs_key *ins,
3781                                 u64 flags, int delalloc)
3782 {
3783         int ret = 0;
3784         int cache_block_group_error = 0;
3785         struct btrfs_free_cluster *last_ptr = NULL;
3786         struct btrfs_block_group_cache *block_group = NULL;
3787         struct find_free_extent_ctl ffe_ctl = {0};
3788         struct btrfs_space_info *space_info;
3789         bool use_cluster = true;
3790         bool full_search = false;
3791 
3792         WARN_ON(num_bytes < fs_info->sectorsize);
3793 
3794         ffe_ctl.ram_bytes = ram_bytes;
3795         ffe_ctl.num_bytes = num_bytes;
3796         ffe_ctl.empty_size = empty_size;
3797         ffe_ctl.flags = flags;
3798         ffe_ctl.search_start = 0;
3799         ffe_ctl.retry_clustered = false;
3800         ffe_ctl.retry_unclustered = false;
3801         ffe_ctl.delalloc = delalloc;
3802         ffe_ctl.index = btrfs_bg_flags_to_raid_index(flags);
3803         ffe_ctl.have_caching_bg = false;
3804         ffe_ctl.orig_have_caching_bg = false;
3805         ffe_ctl.found_offset = 0;
3806 
3807         ins->type = BTRFS_EXTENT_ITEM_KEY;
3808         ins->objectid = 0;
3809         ins->offset = 0;
3810 
3811         trace_find_free_extent(fs_info, num_bytes, empty_size, flags);
3812 
3813         space_info = btrfs_find_space_info(fs_info, flags);
3814         if (!space_info) {
3815                 btrfs_err(fs_info, "No space info for %llu", flags);
3816                 return -ENOSPC;
3817         }
3818 
3819         /*
3820          * If our free space is heavily fragmented we may not be able to make
3821          * big contiguous allocations, so instead of doing the expensive search
3822          * for free space, simply return ENOSPC with our max_extent_size so we
3823          * can go ahead and search for a more manageable chunk.
3824          *
3825          * If our max_extent_size is large enough for our allocation simply
3826          * disable clustering since we will likely not be able to find enough
3827          * space to create a cluster and induce latency trying.
3828          */
3829         if (unlikely(space_info->max_extent_size)) {
3830                 spin_lock(&space_info->lock);
3831                 if (space_info->max_extent_size &&
3832                     num_bytes > space_info->max_extent_size) {
3833                         ins->offset = space_info->max_extent_size;
3834                         spin_unlock(&space_info->lock);
3835                         return -ENOSPC;
3836                 } else if (space_info->max_extent_size) {
3837                         use_cluster = false;
3838                 }
3839                 spin_unlock(&space_info->lock);
3840         }
3841 
3842         last_ptr = fetch_cluster_info(fs_info, space_info,
3843                                       &ffe_ctl.empty_cluster);
3844         if (last_ptr) {
3845                 spin_lock(&last_ptr->lock);
3846                 if (last_ptr->block_group)
3847                         hint_byte = last_ptr->window_start;
3848                 if (last_ptr->fragmented) {
3849                         /*
3850                          * We still set window_start so we can keep track of the
3851                          * last place we found an allocation to try and save
3852                          * some time.
3853                          */
3854                         hint_byte = last_ptr->window_start;
3855                         use_cluster = false;
3856                 }
3857                 spin_unlock(&last_ptr->lock);
3858         }
3859 
3860         ffe_ctl.search_start = max(ffe_ctl.search_start,
3861                                    first_logical_byte(fs_info, 0));
3862         ffe_ctl.search_start = max(ffe_ctl.search_start, hint_byte);
3863         if (ffe_ctl.search_start == hint_byte) {
3864                 block_group = btrfs_lookup_block_group(fs_info,
3865                                                        ffe_ctl.search_start);
3866                 /*
3867                  * we don't want to use the block group if it doesn't match our
3868                  * allocation bits, or if its not cached.
3869                  *
3870                  * However if we are re-searching with an ideal block group
3871                  * picked out then we don't care that the block group is cached.
3872                  */
3873                 if (block_group && block_group_bits(block_group, flags) &&
3874                     block_group->cached != BTRFS_CACHE_NO) {
3875                         down_read(&space_info->groups_sem);
3876                         if (list_empty(&block_group->list) ||
3877                             block_group->ro) {
3878                                 /*
3879                                  * someone is removing this block group,
3880                                  * we can't jump into the have_block_group
3881                                  * target because our list pointers are not
3882                                  * valid
3883                                  */
3884                                 btrfs_put_block_group(block_group);
3885                                 up_read(&space_info->groups_sem);
3886                         } else {
3887                                 ffe_ctl.index = btrfs_bg_flags_to_raid_index(
3888                                                 block_group->flags);
3889                                 btrfs_lock_block_group(block_group, delalloc);
3890                                 goto have_block_group;
3891                         }
3892                 } else if (block_group) {
3893                         btrfs_put_block_group(block_group);
3894                 }
3895         }
3896 search:
3897         ffe_ctl.have_caching_bg = false;
3898         if (ffe_ctl.index == btrfs_bg_flags_to_raid_index(flags) ||
3899             ffe_ctl.index == 0)
3900                 full_search = true;
3901         down_read(&space_info->groups_sem);
3902         list_for_each_entry(block_group,
3903                             &space_info->block_groups[ffe_ctl.index], list) {
3904                 /* If the block group is read-only, we can skip it entirely. */
3905                 if (unlikely(block_group->ro))
3906                         continue;
3907 
3908                 btrfs_grab_block_group(block_group, delalloc);
3909                 ffe_ctl.search_start = block_group->key.objectid;
3910 
3911                 /*
3912                  * this can happen if we end up cycling through all the
3913                  * raid types, but we want to make sure we only allocate
3914                  * for the proper type.
3915                  */
3916                 if (!block_group_bits(block_group, flags)) {
3917                         u64 extra = BTRFS_BLOCK_GROUP_DUP |
3918                                 BTRFS_BLOCK_GROUP_RAID1_MASK |
3919                                 BTRFS_BLOCK_GROUP_RAID56_MASK |
3920                                 BTRFS_BLOCK_GROUP_RAID10;
3921 
3922                         /*
3923                          * if they asked for extra copies and this block group
3924                          * doesn't provide them, bail.  This does allow us to
3925                          * fill raid0 from raid1.
3926                          */
3927                         if ((flags & extra) && !(block_group->flags & extra))
3928                                 goto loop;
3929 
3930                         /*
3931                          * This block group has different flags than we want.
3932                          * It's possible that we have MIXED_GROUP flag but no
3933                          * block group is mixed.  Just skip such block group.
3934                          */
3935                         btrfs_release_block_group(block_group, delalloc);
3936                         continue;
3937                 }
3938 
3939 have_block_group:
3940                 ffe_ctl.cached = btrfs_block_group_cache_done(block_group);
3941                 if (unlikely(!ffe_ctl.cached)) {
3942                         ffe_ctl.have_caching_bg = true;
3943                         ret = btrfs_cache_block_group(block_group, 0);
3944 
3945                         /*
3946                          * If we get ENOMEM here or something else we want to
3947                          * try other block groups, because it may not be fatal.
3948                          * However if we can't find anything else we need to
3949                          * save our return here so that we return the actual
3950                          * error that caused problems, not ENOSPC.
3951                          */
3952                         if (ret < 0) {
3953                                 if (!cache_block_group_error)
3954                                         cache_block_group_error = ret;
3955                                 ret = 0;
3956                                 goto loop;
3957                         }
3958                         ret = 0;
3959                 }
3960 
3961                 if (unlikely(block_group->cached == BTRFS_CACHE_ERROR))
3962                         goto loop;
3963 
3964                 /*
3965                  * Ok we want to try and use the cluster allocator, so
3966                  * lets look there
3967                  */
3968                 if (last_ptr && use_cluster) {
3969                         struct btrfs_block_group_cache *cluster_bg = NULL;
3970 
3971                         ret = find_free_extent_clustered(block_group, last_ptr,
3972                                                          &ffe_ctl, &cluster_bg);
3973 
3974                         if (ret == 0) {
3975                                 if (cluster_bg && cluster_bg != block_group) {
3976                                         btrfs_release_block_group(block_group,
3977                                                                   delalloc);
3978                                         block_group = cluster_bg;
3979                                 }
3980                                 goto checks;
3981                         } else if (ret == -EAGAIN) {
3982                                 goto have_block_group;
3983                         } else if (ret > 0) {
3984                                 goto loop;
3985                         }
3986                         /* ret == -ENOENT case falls through */
3987                 }
3988 
3989                 ret = find_free_extent_unclustered(block_group, last_ptr,
3990                                                    &ffe_ctl);
3991                 if (ret == -EAGAIN)
3992                         goto have_block_group;
3993                 else if (ret > 0)
3994                         goto loop;
3995                 /* ret == 0 case falls through */
3996 checks:
3997                 ffe_ctl.search_start = round_up(ffe_ctl.found_offset,
3998                                              fs_info->stripesize);
3999 
4000                 /* move on to the next group */
4001                 if (ffe_ctl.search_start + num_bytes >
4002                     block_group->key.objectid + block_group->key.offset) {
4003                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4004                                              num_bytes);
4005                         goto loop;
4006                 }
4007 
4008                 if (ffe_ctl.found_offset < ffe_ctl.search_start)
4009                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4010                                 ffe_ctl.search_start - ffe_ctl.found_offset);
4011 
4012                 ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
4013                                 num_bytes, delalloc);
4014                 if (ret == -EAGAIN) {
4015                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4016                                              num_bytes);
4017                         goto loop;
4018                 }
4019                 btrfs_inc_block_group_reservations(block_group);
4020 
4021                 /* we are all good, lets return */
4022                 ins->objectid = ffe_ctl.search_start;
4023                 ins->offset = num_bytes;
4024 
4025                 trace_btrfs_reserve_extent(block_group, ffe_ctl.search_start,
4026                                            num_bytes);
4027                 btrfs_release_block_group(block_group, delalloc);
4028                 break;
4029 loop:
4030                 ffe_ctl.retry_clustered = false;
4031                 ffe_ctl.retry_unclustered = false;
4032                 BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
4033                        ffe_ctl.index);
4034                 btrfs_release_block_group(block_group, delalloc);
4035                 cond_resched();
4036         }
4037         up_read(&space_info->groups_sem);
4038 
4039         ret = find_free_extent_update_loop(fs_info, last_ptr, ins, &ffe_ctl,
4040                                            full_search, use_cluster);
4041         if (ret > 0)
4042                 goto search;
4043 
4044         if (ret == -ENOSPC && !cache_block_group_error) {
4045                 /*
4046                  * Use ffe_ctl->total_free_space as fallback if we can't find
4047                  * any contiguous hole.
4048                  */
4049                 if (!ffe_ctl.max_extent_size)
4050                         ffe_ctl.max_extent_size = ffe_ctl.total_free_space;
4051                 spin_lock(&space_info->lock);
4052                 space_info->max_extent_size = ffe_ctl.max_extent_size;
4053                 spin_unlock(&space_info->lock);
4054                 ins->offset = ffe_ctl.max_extent_size;
4055         } else if (ret == -ENOSPC) {
4056                 ret = cache_block_group_error;
4057         }
4058         return ret;
4059 }
4060 
4061 /*
4062  * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a
4063  *                        hole that is at least as big as @num_bytes.
4064  *
4065  * @root           -    The root that will contain this extent
4066  *
4067  * @ram_bytes      -    The amount of space in ram that @num_bytes take. This
4068  *                      is used for accounting purposes. This value differs
4069  *                      from @num_bytes only in the case of compressed extents.
4070  *
4071  * @num_bytes      -    Number of bytes to allocate on-disk.
4072  *
4073  * @min_alloc_size -    Indicates the minimum amount of space that the
4074  *                      allocator should try to satisfy. In some cases
4075  *                      @num_bytes may be larger than what is required and if
4076  *                      the filesystem is fragmented then allocation fails.
4077  *                      However, the presence of @min_alloc_size gives a
4078  *                      chance to try and satisfy the smaller allocation.
4079  *
4080  * @empty_size     -    A hint that you plan on doing more COW. This is the
4081  *                      size in bytes the allocator should try to find free
4082  *                      next to the block it returns.  This is just a hint and
4083  *                      may be ignored by the allocator.
4084  *
4085  * @hint_byte      -    Hint to the allocator to start searching above the byte
4086  *                      address passed. It might be ignored.
4087  *
4088  * @ins            -    This key is modified to record the found hole. It will
4089  *                      have the following values:
4090  *                      ins->objectid == start position
4091  *                      ins->flags = BTRFS_EXTENT_ITEM_KEY
4092  *                      ins->offset == the size of the hole.
4093  *
4094  * @is_data        -    Boolean flag indicating whether an extent is
4095  *                      allocated for data (true) or metadata (false)
4096  *
4097  * @delalloc       -    Boolean flag indicating whether this allocation is for
4098  *                      delalloc or not. If 'true' data_rwsem of block groups
4099  *                      is going to be acquired.
4100  *
4101  *
4102  * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
4103  * case -ENOSPC is returned then @ins->offset will contain the size of the
4104  * largest available hole the allocator managed to find.
4105  */
4106 int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
4107                          u64 num_bytes, u64 min_alloc_size,
4108                          u64 empty_size, u64 hint_byte,
4109                          struct btrfs_key *ins, int is_data, int delalloc)
4110 {
4111         struct btrfs_fs_info *fs_info = root->fs_info;
4112         bool final_tried = num_bytes == min_alloc_size;
4113         u64 flags;
4114         int ret;
4115 
4116         flags = get_alloc_profile_by_root(root, is_data);
4117 again:
4118         WARN_ON(num_bytes < fs_info->sectorsize);
4119         ret = find_free_extent(fs_info, ram_bytes, num_bytes, empty_size,
4120                                hint_byte, ins, flags, delalloc);
4121         if (!ret && !is_data) {
4122                 btrfs_dec_block_group_reservations(fs_info, ins->objectid);
4123         } else if (ret == -ENOSPC) {
4124                 if (!final_tried && ins->offset) {
4125                         num_bytes = min(num_bytes >> 1, ins->offset);
4126                         num_bytes = round_down(num_bytes,
4127                                                fs_info->sectorsize);
4128                         num_bytes = max(num_bytes, min_alloc_size);
4129                         ram_bytes = num_bytes;
4130                         if (num_bytes == min_alloc_size)
4131                                 final_tried = true;
4132                         goto again;
4133                 } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
4134                         struct btrfs_space_info *sinfo;
4135 
4136                         sinfo = btrfs_find_space_info(fs_info, flags);
4137                         btrfs_err(fs_info,
4138                                   "allocation failed flags %llu, wanted %llu",
4139                                   flags, num_bytes);
4140                         if (sinfo)
4141                                 btrfs_dump_space_info(fs_info, sinfo,
4142                                                       num_bytes, 1);
4143                 }
4144         }
4145 
4146         return ret;
4147 }
4148 
4149 static int __btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
4150                                         u64 start, u64 len,
4151                                         int pin, int delalloc)
4152 {
4153         struct btrfs_block_group_cache *cache;
4154         int ret = 0;
4155 
4156         cache = btrfs_lookup_block_group(fs_info, start);
4157         if (!cache) {
4158                 btrfs_err(fs_info, "Unable to find block group for %llu",
4159                           start);
4160                 return -ENOSPC;
4161         }
4162 
4163         if (pin)
4164                 pin_down_extent(cache, start, len, 1);
4165         else {
4166                 if (btrfs_test_opt(fs_info, DISCARD))
4167                         ret = btrfs_discard_extent(fs_info, start, len, NULL);
4168                 btrfs_add_free_space(cache, start, len);
4169                 btrfs_free_reserved_bytes(cache, len, delalloc);
4170                 trace_btrfs_reserved_extent_free(fs_info, start, len);
4171         }
4172 
4173         btrfs_put_block_group(cache);
4174         return ret;
4175 }
4176 
4177 int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
4178                                u64 start, u64 len, int delalloc)
4179 {
4180         return __btrfs_free_reserved_extent(fs_info, start, len, 0, delalloc);
4181 }
4182 
4183 int btrfs_free_and_pin_reserved_extent(struct btrfs_fs_info *fs_info,
4184                                        u64 start, u64 len)
4185 {
4186         return __btrfs_free_reserved_extent(fs_info, start, len, 1, 0);
4187 }
4188 
4189 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4190                                       u64 parent, u64 root_objectid,
4191                                       u64 flags, u64 owner, u64 offset,
4192                                       struct btrfs_key *ins, int ref_mod)
4193 {
4194         struct btrfs_fs_info *fs_info = trans->fs_info;
4195         int ret;
4196         struct btrfs_extent_item *extent_item;
4197         struct btrfs_extent_inline_ref *iref;
4198         struct btrfs_path *path;
4199         struct extent_buffer *leaf;
4200         int type;
4201         u32 size;
4202 
4203         if (parent > 0)
4204                 type = BTRFS_SHARED_DATA_REF_KEY;
4205         else
4206                 type = BTRFS_EXTENT_DATA_REF_KEY;
4207 
4208         size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4209 
4210         path = btrfs_alloc_path();
4211         if (!path)
4212                 return -ENOMEM;
4213 
4214         path->leave_spinning = 1;
4215         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4216                                       ins, size);
4217         if (ret) {
4218                 btrfs_free_path(path);
4219                 return ret;
4220         }
4221 
4222         leaf = path->nodes[0];
4223         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4224                                      struct btrfs_extent_item);
4225         btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4226         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4227         btrfs_set_extent_flags(leaf, extent_item,
4228                                flags | BTRFS_EXTENT_FLAG_DATA);
4229 
4230         iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4231         btrfs_set_extent_inline_ref_type(leaf, iref, type);
4232         if (parent > 0) {
4233                 struct btrfs_shared_data_ref *ref;
4234                 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4235                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4236                 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4237         } else {
4238                 struct btrfs_extent_data_ref *ref;
4239                 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4240                 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4241                 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4242                 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4243                 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4244         }
4245 
4246         btrfs_mark_buffer_dirty(path->nodes[0]);
4247         btrfs_free_path(path);
4248 
4249         ret = remove_from_free_space_tree(trans, ins->objectid, ins->offset);
4250         if (ret)
4251                 return ret;
4252 
4253         ret = btrfs_update_block_group(trans, ins->objectid, ins->offset, 1);
4254         if (ret) { /* -ENOENT, logic error */
4255                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4256                         ins->objectid, ins->offset);
4257                 BUG();
4258         }
4259         trace_btrfs_reserved_extent_alloc(fs_info, ins->objectid, ins->offset);
4260         return ret;
4261 }
4262 
4263 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4264                                      struct btrfs_delayed_ref_node *node,
4265                                      struct btrfs_delayed_extent_op *extent_op)
4266 {
4267         struct btrfs_fs_info *fs_info = trans->fs_info;
4268         int ret;
4269         struct btrfs_extent_item *extent_item;
4270         struct btrfs_key extent_key;
4271         struct btrfs_tree_block_info *block_info;
4272         struct btrfs_extent_inline_ref *iref;
4273         struct btrfs_path *path;
4274         struct extent_buffer *leaf;
4275         struct btrfs_delayed_tree_ref *ref;
4276         u32 size = sizeof(*extent_item) + sizeof(*iref);
4277         u64 num_bytes;
4278         u64 flags = extent_op->flags_to_set;
4279         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4280 
4281         ref = btrfs_delayed_node_to_tree_ref(node);
4282 
4283         extent_key.objectid = node->bytenr;
4284         if (skinny_metadata) {
4285                 extent_key.offset = ref->level;
4286                 extent_key.type = BTRFS_METADATA_ITEM_KEY;
4287                 num_bytes = fs_info->nodesize;
4288         } else {
4289                 extent_key.offset = node->num_bytes;
4290                 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4291                 size += sizeof(*block_info);
4292                 num_bytes = node->num_bytes;
4293         }
4294 
4295         path = btrfs_alloc_path();
4296         if (!path)
4297                 return -ENOMEM;
4298 
4299         path->leave_spinning = 1;
4300         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4301                                       &extent_key, size);
4302         if (ret) {
4303                 btrfs_free_path(path);
4304                 return ret;
4305         }
4306 
4307         leaf = path->nodes[0];
4308         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4309                                      struct btrfs_extent_item);
4310         btrfs_set_extent_refs(leaf, extent_item, 1);
4311         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4312         btrfs_set_extent_flags(leaf, extent_item,
4313                                flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4314 
4315         if (skinny_metadata) {
4316                 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4317         } else {
4318                 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4319                 btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
4320                 btrfs_set_tree_block_level(leaf, block_info, ref->level);
4321                 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4322         }
4323 
4324         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
4325                 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4326                 btrfs_set_extent_inline_ref_type(leaf, iref,
4327                                                  BTRFS_SHARED_BLOCK_REF_KEY);
4328                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent);
4329         } else {
4330                 btrfs_set_extent_inline_ref_type(leaf, iref,
4331                                                  BTRFS_TREE_BLOCK_REF_KEY);
4332                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root);
4333         }
4334 
4335         btrfs_mark_buffer_dirty(leaf);
4336         btrfs_free_path(path);
4337 
4338         ret = remove_from_free_space_tree(trans, extent_key.objectid,
4339                                           num_bytes);
4340         if (ret)
4341                 return ret;
4342 
4343         ret = btrfs_update_block_group(trans, extent_key.objectid,
4344                                        fs_info->nodesize, 1);
4345         if (ret) { /* -ENOENT, logic error */
4346                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4347                         extent_key.objectid, extent_key.offset);
4348                 BUG();
4349         }
4350 
4351         trace_btrfs_reserved_extent_alloc(fs_info, extent_key.objectid,
4352                                           fs_info->nodesize);
4353         return ret;
4354 }
4355 
4356 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4357                                      struct btrfs_root *root, u64 owner,
4358                                      u64 offset, u64 ram_bytes,
4359                                      struct btrfs_key *ins)
4360 {
4361         struct btrfs_ref generic_ref = { 0 };
4362         int ret;
4363 
4364         BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
4365 
4366         btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4367                                ins->objectid, ins->offset, 0);
4368         btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
4369         btrfs_ref_tree_mod(root->fs_info, &generic_ref);
4370         ret = btrfs_add_delayed_data_ref(trans, &generic_ref,
4371                                          ram_bytes, NULL, NULL);
4372         return ret;
4373 }
4374 
4375 /*
4376  * this is used by the tree logging recovery code.  It records that
4377  * an extent has been allocated and makes sure to clear the free
4378  * space cache bits as well
4379  */
4380 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4381                                    u64 root_objectid, u64 owner, u64 offset,
4382                                    struct btrfs_key *ins)
4383 {
4384         struct btrfs_fs_info *fs_info = trans->fs_info;
4385         int ret;
4386         struct btrfs_block_group_cache *block_group;
4387         struct btrfs_space_info *space_info;
4388 
4389         /*
4390          * Mixed block groups will exclude before processing the log so we only
4391          * need to do the exclude dance if this fs isn't mixed.
4392          */
4393         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
4394                 ret = __exclude_logged_extent(fs_info, ins->objectid,
4395                                               ins->offset);
4396                 if (ret)
4397                         return ret;
4398         }
4399 
4400         block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
4401         if (!block_group)
4402                 return -EINVAL;
4403 
4404         space_info = block_group->space_info;
4405         spin_lock(&space_info->lock);
4406         spin_lock(&block_group->lock);
4407         space_info->bytes_reserved += ins->offset;
4408         block_group->reserved += ins->offset;
4409         spin_unlock(&block_group->lock);
4410         spin_unlock(&space_info->lock);
4411 
4412         ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
4413                                          offset, ins, 1);
4414         if (ret)
4415                 btrfs_pin_extent(fs_info, ins->objectid, ins->offset, 1);
4416         btrfs_put_block_group(block_group);
4417         return ret;
4418 }
4419 
4420 static struct extent_buffer *
4421 btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4422                       u64 bytenr, int level, u64 owner)
4423 {
4424         struct btrfs_fs_info *fs_info = root->fs_info;
4425         struct extent_buffer *buf;
4426 
4427         buf = btrfs_find_create_tree_block(fs_info, bytenr);
4428         if (IS_ERR(buf))
4429                 return buf;
4430 
4431         /*
4432          * Extra safety check in case the extent tree is corrupted and extent
4433          * allocator chooses to use a tree block which is already used and
4434          * locked.
4435          */
4436         if (buf->lock_owner == current->pid) {
4437                 btrfs_err_rl(fs_info,
4438 "tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
4439                         buf->start, btrfs_header_owner(buf), current->pid);
4440                 free_extent_buffer(buf);
4441                 return ERR_PTR(-EUCLEAN);
4442         }
4443 
4444         btrfs_set_buffer_lockdep_class(root->root_key.objectid, buf, level);
4445         btrfs_tree_lock(buf);
4446         btrfs_clean_tree_block(buf);
4447         clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
4448 
4449         btrfs_set_lock_blocking_write(buf);
4450         set_extent_buffer_uptodate(buf);
4451 
4452         memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
4453         btrfs_set_header_level(buf, level);
4454         btrfs_set_header_bytenr(buf, buf->start);
4455         btrfs_set_header_generation(buf, trans->transid);
4456         btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
4457         btrfs_set_header_owner(buf, owner);
4458         write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
4459         write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
4460         if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4461                 buf->log_index = root->log_transid % 2;
4462                 /*
4463                  * we allow two log transactions at a time, use different
4464                  * EXTENT bit to differentiate dirty pages.
4465                  */
4466                 if (buf->log_index == 0)
4467                         set_extent_dirty(&root->dirty_log_pages, buf->start,
4468                                         buf->start + buf->len - 1, GFP_NOFS);
4469                 else
4470                         set_extent_new(&root->dirty_log_pages, buf->start,
4471                                         buf->start + buf->len - 1);
4472         } else {
4473                 buf->log_index = -1;
4474                 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4475                          buf->start + buf->len - 1, GFP_NOFS);
4476         }
4477         trans->dirty = true;
4478         /* this returns a buffer locked for blocking */
4479         return buf;
4480 }
4481 
4482 /*
4483  * finds a free extent and does all the dirty work required for allocation
4484  * returns the tree buffer or an ERR_PTR on error.
4485  */
4486 struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
4487                                              struct btrfs_root *root,
4488                                              u64 parent, u64 root_objectid,
4489                                              const struct btrfs_disk_key *key,
4490                                              int level, u64 hint,
4491                                              u64 empty_size)
4492 {
4493         struct btrfs_fs_info *fs_info = root->fs_info;
4494         struct btrfs_key ins;
4495         struct btrfs_block_rsv *block_rsv;
4496         struct extent_buffer *buf;
4497         struct btrfs_delayed_extent_op *extent_op;
4498         struct btrfs_ref generic_ref = { 0 };
4499         u64 flags = 0;
4500         int ret;
4501         u32 blocksize = fs_info->nodesize;
4502         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4503 
4504 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4505         if (btrfs_is_testing(fs_info)) {
4506                 buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
4507                                             level, root_objectid);
4508                 if (!IS_ERR(buf))
4509                         root->alloc_bytenr += blocksize;
4510                 return buf;
4511         }
4512 #endif
4513 
4514         block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
4515         if (IS_ERR(block_rsv))
4516                 return ERR_CAST(block_rsv);
4517 
4518         ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
4519                                    empty_size, hint, &ins, 0, 0);
4520         if (ret)
4521                 goto out_unuse;
4522 
4523         buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
4524                                     root_objectid);
4525         if (IS_ERR(buf)) {
4526                 ret = PTR_ERR(buf);
4527                 goto out_free_reserved;
4528         }
4529 
4530         if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4531                 if (parent == 0)
4532                         parent = ins.objectid;
4533                 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4534         } else
4535                 BUG_ON(parent > 0);
4536 
4537         if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4538                 extent_op = btrfs_alloc_delayed_extent_op();
4539                 if (!extent_op) {
4540                         ret = -ENOMEM;
4541                         goto out_free_buf;
4542                 }
4543                 if (key)
4544                         memcpy(&extent_op->key, key, sizeof(extent_op->key));
4545                 else
4546                         memset(&extent_op->key, 0, sizeof(extent_op->key));
4547                 extent_op->flags_to_set = flags;
4548                 extent_op->update_key = skinny_metadata ? false : true;
4549                 extent_op->update_flags = true;
4550                 extent_op->is_data = false;
4551                 extent_op->level = level;
4552 
4553                 btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4554                                        ins.objectid, ins.offset, parent);
4555                 generic_ref.real_root = root->root_key.objectid;
4556                 btrfs_init_tree_ref(&generic_ref, level, root_objectid);
4557                 btrfs_ref_tree_mod(fs_info, &generic_ref);
4558                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref,
4559                                                  extent_op, NULL, NULL);
4560                 if (ret)
4561                         goto out_free_delayed;
4562         }
4563         return buf;
4564 
4565 out_free_delayed:
4566         btrfs_free_delayed_extent_op(extent_op);
4567 out_free_buf:
4568         free_extent_buffer(buf);
4569 out_free_reserved:
4570         btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
4571 out_unuse:
4572         btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
4573         return ERR_PTR(ret);
4574 }
4575 
4576 struct walk_control {
4577         u64 refs[BTRFS_MAX_LEVEL];
4578         u64 flags[BTRFS_MAX_LEVEL];
4579         struct btrfs_key update_progress;
4580         struct btrfs_key drop_progress;
4581         int drop_level;
4582         int stage;
4583         int level;
4584         int shared_level;
4585         int update_ref;
4586         int keep_locks;
4587         int reada_slot;
4588         int reada_count;
4589         int restarted;
4590 };
4591 
4592 #define DROP_REFERENCE  1
4593 #define UPDATE_BACKREF  2
4594 
4595 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4596                                      struct btrfs_root *root,
4597                                      struct walk_control *wc,
4598                                      struct btrfs_path *path)
4599 {
4600         struct btrfs_fs_info *fs_info = root->fs_info;
4601         u64 bytenr;
4602         u64 generation;
4603         u64 refs;
4604         u64 flags;
4605         u32 nritems;
4606         struct btrfs_key key;
4607         struct extent_buffer *eb;
4608         int ret;
4609         int slot;
4610         int nread = 0;
4611 
4612         if (path->slots[wc->level] < wc->reada_slot) {
4613                 wc->reada_count = wc->reada_count * 2 / 3;
4614                 wc->reada_count = max(wc->reada_count, 2);
4615         } else {
4616                 wc->reada_count = wc->reada_count * 3 / 2;
4617                 wc->reada_count = min_t(int, wc->reada_count,
4618                                         BTRFS_NODEPTRS_PER_BLOCK(fs_info));
4619         }
4620 
4621         eb = path->nodes[wc->level];
4622         nritems = btrfs_header_nritems(eb);
4623 
4624         for (slot = path->slots[wc->level]; slot < nritems; slot++) {
4625                 if (nread >= wc->reada_count)
4626                         break;
4627 
4628                 cond_resched();
4629                 bytenr = btrfs_node_blockptr(eb, slot);
4630                 generation = btrfs_node_ptr_generation(eb, slot);
4631 
4632                 if (slot == path->slots[wc->level])
4633                         goto reada;
4634 
4635                 if (wc->stage == UPDATE_BACKREF &&
4636                     generation <= root->root_key.offset)
4637                         continue;
4638 
4639                 /* We don't lock the tree block, it's OK to be racy here */
4640                 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
4641                                                wc->level - 1, 1, &refs,
4642                                                &flags);
4643                 /* We don't care about errors in readahead. */
4644                 if (ret < 0)
4645                         continue;
4646                 BUG_ON(refs == 0);
4647 
4648                 if (wc->stage == DROP_REFERENCE) {
4649                         if (refs == 1)
4650                                 goto reada;
4651 
4652                         if (wc->level == 1 &&
4653                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4654                                 continue;
4655                         if (!wc->update_ref ||
4656                             generation <= root->root_key.offset)
4657                                 continue;
4658                         btrfs_node_key_to_cpu(eb, &key, slot);
4659                         ret = btrfs_comp_cpu_keys(&key,
4660                                                   &wc->update_progress);
4661                         if (ret < 0)
4662                                 continue;
4663                 } else {
4664                         if (wc->level == 1 &&
4665                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4666                                 continue;
4667                 }
4668 reada:
4669                 readahead_tree_block(fs_info, bytenr);
4670                 nread++;
4671         }
4672         wc->reada_slot = slot;
4673 }
4674 
4675 /*
4676  * helper to process tree block while walking down the tree.
4677  *
4678  * when wc->stage == UPDATE_BACKREF, this function updates
4679  * back refs for pointers in the block.
4680  *
4681  * NOTE: return value 1 means we should stop walking down.
4682  */
4683 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4684                                    struct btrfs_root *root,
4685                                    struct btrfs_path *path,
4686                                    struct walk_control *wc, int lookup_info)
4687 {
4688         struct btrfs_fs_info *fs_info = root->fs_info;
4689         int level = wc->level;
4690         struct extent_buffer *eb = path->nodes[level];
4691         u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
4692         int ret;
4693 
4694         if (wc->stage == UPDATE_BACKREF &&
4695             btrfs_header_owner(eb) != root->root_key.objectid)
4696                 return 1;
4697 
4698         /*
4699          * when reference count of tree block is 1, it won't increase
4700          * again. once full backref flag is set, we never clear it.
4701          */
4702         if (lookup_info &&
4703             ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
4704              (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
4705                 BUG_ON(!path->locks[level]);
4706                 ret = btrfs_lookup_extent_info(trans, fs_info,
4707                                                eb->start, level, 1,
4708                                                &wc->refs[level],
4709                                                &wc->flags[level]);
4710                 BUG_ON(ret == -ENOMEM);
4711                 if (ret)
4712                         return ret;
4713                 BUG_ON(wc->refs[level] == 0);
4714         }
4715 
4716         if (wc->stage == DROP_REFERENCE) {
4717                 if (wc->refs[level] > 1)
4718                         return 1;
4719 
4720                 if (path->locks[level] && !wc->keep_locks) {
4721                         btrfs_tree_unlock_rw(eb, path->locks[level]);
4722                         path->locks[level] = 0;
4723                 }
4724                 return 0;
4725         }
4726 
4727         /* wc->stage == UPDATE_BACKREF */
4728         if (!(wc->flags[level] & flag)) {
4729                 BUG_ON(!path->locks[level]);
4730                 ret = btrfs_inc_ref(trans, root, eb, 1);
4731                 BUG_ON(ret); /* -ENOMEM */
4732                 ret = btrfs_dec_ref(trans, root, eb, 0);
4733                 BUG_ON(ret); /* -ENOMEM */
4734                 ret = btrfs_set_disk_extent_flags(trans, eb->start,
4735                                                   eb->len, flag,
4736                                                   btrfs_header_level(eb), 0);
4737                 BUG_ON(ret); /* -ENOMEM */
4738                 wc->flags[level] |= flag;
4739         }
4740 
4741         /*
4742          * the block is shared by multiple trees, so it's not good to
4743          * keep the tree lock
4744          */
4745         if (path->locks[level] && level > 0) {
4746                 btrfs_tree_unlock_rw(eb, path->locks[level]);
4747                 path->locks[level] = 0;
4748         }
4749         return 0;
4750 }
4751 
4752 /*
4753  * This is used to verify a ref exists for this root to deal with a bug where we
4754  * would have a drop_progress key that hadn't been updated properly.
4755  */
4756 static int check_ref_exists(struct btrfs_trans_handle *trans,
4757                             struct btrfs_root *root, u64 bytenr, u64 parent,
4758                             int level)
4759 {
4760         struct btrfs_path *path;
4761         struct btrfs_extent_inline_ref *iref;
4762         int ret;
4763 
4764         path = btrfs_alloc_path();
4765         if (!path)
4766                 return -ENOMEM;
4767 
4768         ret = lookup_extent_backref(trans, path, &iref, bytenr,
4769                                     root->fs_info->nodesize, parent,
4770                                     root->root_key.objectid, level, 0);
4771         btrfs_free_path(path);
4772         if (ret == -ENOENT)
4773                 return 0;
4774         if (ret < 0)
4775                 return ret;
4776         return 1;
4777 }
4778 
4779 /*
4780  * helper to process tree block pointer.
4781  *
4782  * when wc->stage == DROP_REFERENCE, this function checks
4783  * reference count of the block pointed to. if the block
4784  * is shared and we need update back refs for the subtree
4785  * rooted at the block, this function changes wc->stage to
4786  * UPDATE_BACKREF. if the block is shared and there is no
4787  * need to update back, this function drops the reference
4788  * to the block.
4789  *
4790  * NOTE: return value 1 means we should stop walking down.
4791  */
4792 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
4793                                  struct btrfs_root *root,
4794                                  struct btrfs_path *path,
4795                                  struct walk_control *wc, int *lookup_info)
4796 {
4797         struct btrfs_fs_info *fs_info = root->fs_info;
4798         u64 bytenr;
4799         u64 generation;
4800         u64 parent;
4801         struct btrfs_key key;
4802         struct btrfs_key first_key;
4803         struct btrfs_ref ref = { 0 };
4804         struct extent_buffer *next;
4805         int level = wc->level;
4806         int reada = 0;
4807         int ret = 0;
4808         bool need_account = false;
4809 
4810         generation = btrfs_node_ptr_generation(path->nodes[level],
4811                                                path->slots[level]);
4812         /*
4813          * if the lower level block was created before the snapshot
4814          * was created, we know there is no need to update back refs
4815          * for the subtree
4816          */
4817         if (wc->stage == UPDATE_BACKREF &&
4818             generation <= root->root_key.offset) {
4819                 *lookup_info = 1;
4820                 return 1;
4821         }
4822 
4823         bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
4824         btrfs_node_key_to_cpu(path->nodes[level], &first_key,
4825                               path->slots[level]);
4826 
4827         next = find_extent_buffer(fs_info, bytenr);
4828         if (!next) {
4829                 next = btrfs_find_create_tree_block(fs_info, bytenr);
4830                 if (IS_ERR(next))
4831                         return PTR_ERR(next);
4832 
4833                 btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
4834                                                level - 1);
4835                 reada = 1;
4836         }
4837         btrfs_tree_lock(next);
4838         btrfs_set_lock_blocking_write(next);
4839 
4840         ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
4841                                        &wc->refs[level - 1],
4842                                        &wc->flags[level - 1]);
4843         if (ret < 0)
4844                 goto out_unlock;
4845 
4846         if (unlikely(wc->refs[level - 1] == 0)) {
4847                 btrfs_err(fs_info, "Missing references.");
4848                 ret = -EIO;
4849                 goto out_unlock;
4850         }
4851         *lookup_info = 0;
4852 
4853         if (wc->stage == DROP_REFERENCE) {
4854                 if (wc->refs[level - 1] > 1) {
4855                         need_account = true;
4856                         if (level == 1 &&
4857                             (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4858                                 goto skip;
4859 
4860                         if (!wc->update_ref ||
4861                             generation <= root->root_key.offset)
4862                                 goto skip;
4863 
4864                         btrfs_node_key_to_cpu(path->nodes[level], &key,
4865                                               path->slots[level]);
4866                         ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
4867                         if (ret < 0)
4868                                 goto skip;
4869 
4870                         wc->stage = UPDATE_BACKREF;
4871                         wc->shared_level = level - 1;
4872                 }
4873         } else {
4874                 if (level == 1 &&
4875                     (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4876                         goto skip;
4877         }
4878 
4879         if (!btrfs_buffer_uptodate(next, generation, 0)) {
4880                 btrfs_tree_unlock(next);
4881                 free_extent_buffer(next);
4882                 next = NULL;
4883                 *lookup_info = 1;
4884         }
4885 
4886         if (!next) {
4887                 if (reada && level == 1)
4888                         reada_walk_down(trans, root, wc, path);
4889                 next = read_tree_block(fs_info, bytenr, generation, level - 1,
4890                                        &first_key);
4891                 if (IS_ERR(next)) {
4892                         return PTR_ERR(next);
4893                 } else if (!extent_buffer_uptodate(next)) {
4894                         free_extent_buffer(next);
4895                         return -EIO;
4896                 }
4897                 btrfs_tree_lock(next);
4898                 btrfs_set_lock_blocking_write(next);
4899         }
4900 
4901         level--;
4902         ASSERT(level == btrfs_header_level(next));
4903         if (level != btrfs_header_level(next)) {
4904                 btrfs_err(root->fs_info, "mismatched level");
4905                 ret = -EIO;
4906                 goto out_unlock;
4907         }
4908         path->nodes[level] = next;
4909         path->slots[level] = 0;
4910         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
4911         wc->level = level;
4912         if (wc->level == 1)
4913                 wc->reada_slot = 0;
4914         return 0;
4915 skip:
4916         wc->refs[level - 1] = 0;
4917         wc->flags[level - 1] = 0;
4918         if (wc->stage == DROP_REFERENCE) {
4919                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
4920                         parent = path->nodes[level]->start;
4921                 } else {
4922                         ASSERT(root->root_key.objectid ==
4923                                btrfs_header_owner(path->nodes[level]));
4924                         if (root->root_key.objectid !=
4925                             btrfs_header_owner(path->nodes[level])) {
4926                                 btrfs_err(root->fs_info,
4927                                                 "mismatched block owner");
4928                                 ret = -EIO;
4929                                 goto out_unlock;
4930                         }
4931                         parent = 0;
4932                 }
4933 
4934                 /*
4935                  * If we had a drop_progress we need to verify the refs are set
4936                  * as expected.  If we find our ref then we know that from here
4937                  * on out everything should be correct, and we can clear the
4938                  * ->restarted flag.
4939                  */
4940                 if (wc->restarted) {
4941                         ret = check_ref_exists(trans, root, bytenr, parent,
4942                                                level - 1);
4943                         if (ret < 0)
4944                                 goto out_unlock;
4945                         if (ret == 0)
4946                                 goto no_delete;
4947                         ret = 0;
4948                         wc->restarted = 0;
4949                 }
4950 
4951                 /*
4952                  * Reloc tree doesn't contribute to qgroup numbers, and we have
4953                  * already accounted them at merge time (replace_path),
4954                  * thus we could skip expensive subtree trace here.
4955                  */
4956                 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
4957                     need_account) {
4958                         ret = btrfs_qgroup_trace_subtree(trans, next,
4959                                                          generation, level - 1);
4960                         if (ret) {
4961                                 btrfs_err_rl(fs_info,
4962                                              "Error %d accounting shared subtree. Quota is out of sync, rescan required.",
4963                                              ret);
4964                         }
4965                 }
4966 
4967                 /*
4968                  * We need to update the next key in our walk control so we can
4969                  * update the drop_progress key accordingly.  We don't care if
4970                  * find_next_key doesn't find a key because that means we're at
4971                  * the end and are going to clean up now.
4972                  */
4973                 wc->drop_level = level;
4974                 find_next_key(path, level, &wc->drop_progress);
4975 
4976                 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
4977                                        fs_info->nodesize, parent);
4978                 btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
4979                 ret = btrfs_free_extent(trans, &ref);
4980                 if (ret)
4981                         goto out_unlock;
4982         }
4983 no_delete:
4984         *lookup_info = 1;
4985         ret = 1;
4986 
4987 out_unlock:
4988         btrfs_tree_unlock(next);
4989         free_extent_buffer(next);
4990 
4991         return ret;
4992 }
4993 
4994 /*
4995  * helper to process tree block while walking up the tree.
4996  *
4997  * when wc->stage == DROP_REFERENCE, this function drops
4998  * reference count on the block.
4999  *
5000  * when wc->stage == UPDATE_BACKREF, this function changes
5001  * wc->stage back to DROP_REFERENCE if we changed wc->stage
5002  * to UPDATE_BACKREF previously while processing the block.
5003  *
5004  * NOTE: return value 1 means we should stop walking up.
5005  */
5006 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5007                                  struct btrfs_root *root,
5008                                  struct btrfs_path *path,
5009                                  struct walk_control *wc)
5010 {
5011         struct btrfs_fs_info *fs_info = root->fs_info;
5012         int ret;
5013         int level = wc->level;
5014         struct extent_buffer *eb = path->nodes[level];
5015         u64 parent = 0;
5016 
5017         if (wc->stage == UPDATE_BACKREF) {
5018                 BUG_ON(wc->shared_level < level);
5019                 if (level < wc->shared_level)
5020                         goto out;
5021 
5022                 ret = find_next_key(path, level + 1, &wc->update_progress);
5023                 if (ret > 0)
5024                         wc->update_ref = 0;
5025 
5026                 wc->stage = DROP_REFERENCE;
5027                 wc->shared_level = -1;
5028                 path->slots[level] = 0;
5029 
5030                 /*
5031                  * check reference count again if the block isn't locked.
5032                  * we should start walking down the tree again if reference
5033                  * count is one.
5034                  */
5035                 if (!path->locks[level]) {
5036                         BUG_ON(level == 0);
5037                         btrfs_tree_lock(eb);
5038                         btrfs_set_lock_blocking_write(eb);
5039                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5040 
5041                         ret = btrfs_lookup_extent_info(trans, fs_info,
5042                                                        eb->start, level, 1,
5043                                                        &wc->refs[level],
5044                                                        &wc->flags[level]);
5045                         if (ret < 0) {
5046                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5047                                 path->locks[level] = 0;
5048                                 return ret;
5049                         }
5050                         BUG_ON(wc->refs[level] == 0);
5051                         if (wc->refs[level] == 1) {
5052                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5053                                 path->locks[level] = 0;
5054                                 return 1;
5055                         }
5056                 }
5057         }
5058 
5059         /* wc->stage == DROP_REFERENCE */
5060         BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5061 
5062         if (wc->refs[level] == 1) {
5063                 if (level == 0) {
5064                         if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5065                                 ret = btrfs_dec_ref(trans, root, eb, 1);
5066                         else
5067                                 ret = btrfs_dec_ref(trans, root, eb, 0);
5068                         BUG_ON(ret); /* -ENOMEM */
5069                         if (is_fstree(root->root_key.objectid)) {
5070                                 ret = btrfs_qgroup_trace_leaf_items(trans, eb);
5071                                 if (ret) {
5072                                         btrfs_err_rl(fs_info,
5073         "error %d accounting leaf items, quota is out of sync, rescan required",
5074                                              ret);
5075                                 }
5076                         }
5077                 }
5078                 /* make block locked assertion in btrfs_clean_tree_block happy */
5079                 if (!path->locks[level] &&
5080                     btrfs_header_generation(eb) == trans->transid) {
5081                         btrfs_tree_lock(eb);
5082                         btrfs_set_lock_blocking_write(eb);
5083                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5084                 }
5085                 btrfs_clean_tree_block(eb);
5086         }
5087 
5088         if (eb == root->node) {
5089                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5090                         parent = eb->start;
5091                 else if (root->root_key.objectid != btrfs_header_owner(eb))
5092                         goto owner_mismatch;
5093         } else {
5094                 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5095                         parent = path->nodes[level + 1]->start;
5096                 else if (root->root_key.objectid !=
5097                          btrfs_header_owner(path->nodes[level + 1]))
5098                         goto owner_mismatch;
5099         }
5100 
5101         btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
5102 out:
5103         wc->refs[level] = 0;
5104         wc->flags[level] = 0;
5105         return 0;
5106 
5107 owner_mismatch:
5108         btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
5109                      btrfs_header_owner(eb), root->root_key.objectid);
5110         return -EUCLEAN;
5111 }
5112 
5113 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5114                                    struct btrfs_root *root,
5115                                    struct btrfs_path *path,
5116                                    struct walk_control *wc)
5117 {
5118         int level = wc->level;
5119         int lookup_info = 1;
5120         int ret;
5121 
5122         while (level >= 0) {
5123                 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5124                 if (ret > 0)
5125                         break;
5126 
5127                 if (level == 0)
5128                         break;
5129 
5130                 if (path->slots[level] >=
5131                     btrfs_header_nritems(path->nodes[level]))
5132                         break;
5133 
5134                 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5135                 if (ret > 0) {
5136                         path->slots[level]++;
5137                         continue;
5138                 } else if (ret < 0)
5139                         return ret;
5140                 level = wc->level;
5141         }
5142         return 0;
5143 }
5144 
5145 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5146                                  struct btrfs_root *root,
5147                                  struct btrfs_path *path,
5148                                  struct walk_control *wc, int max_level)
5149 {
5150         int level = wc->level;
5151         int ret;
5152 
5153         path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5154         while (level < max_level && path->nodes[level]) {
5155                 wc->level = level;
5156                 if (path->slots[level] + 1 <
5157                     btrfs_header_nritems(path->nodes[level])) {
5158                         path->slots[level]++;
5159                         return 0;
5160                 } else {
5161                         ret = walk_up_proc(trans, root, path, wc);
5162                         if (ret > 0)
5163                                 return 0;
5164                         if (ret < 0)
5165                                 return ret;
5166 
5167                         if (path->locks[level]) {
5168                                 btrfs_tree_unlock_rw(path->nodes[level],
5169                                                      path->locks[level]);
5170                                 path->locks[level] = 0;
5171                         }
5172                         free_extent_buffer(path->nodes[level]);
5173                         path->nodes[level] = NULL;
5174                         level++;
5175                 }
5176         }
5177         return 1;
5178 }
5179 
5180 /*
5181  * drop a subvolume tree.
5182  *
5183  * this function traverses the tree freeing any blocks that only
5184  * referenced by the tree.
5185  *
5186  * when a shared tree block is found. this function decreases its
5187  * reference count by one. if update_ref is true, this function
5188  * also make sure backrefs for the shared block and all lower level
5189  * blocks are properly updated.
5190  *
5191  * If called with for_reloc == 0, may exit early with -EAGAIN
5192  */
5193 int btrfs_drop_snapshot(struct btrfs_root *root,
5194                          struct btrfs_block_rsv *block_rsv, int update_ref,
5195                          int for_reloc)
5196 {
5197         struct btrfs_fs_info *fs_info = root->fs_info;
5198         struct btrfs_path *path;
5199         struct btrfs_trans_handle *trans;
5200         struct btrfs_root *tree_root = fs_info->tree_root;
5201         struct btrfs_root_item *root_item = &root->root_item;
5202         struct walk_control *wc;
5203         struct btrfs_key key;
5204         int err = 0;
5205         int ret;
5206         int level;
5207         bool root_dropped = false;
5208 
5209         btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
5210 
5211         path = btrfs_alloc_path();
5212         if (!path) {
5213                 err = -ENOMEM;
5214                 goto out;
5215         }
5216 
5217         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5218         if (!wc) {
5219                 btrfs_free_path(path);
5220                 err = -ENOMEM;
5221                 goto out;
5222         }
5223 
5224         trans = btrfs_start_transaction(tree_root, 0);
5225         if (IS_ERR(trans)) {
5226                 err = PTR_ERR(trans);
5227                 goto out_free;
5228         }
5229 
5230         err = btrfs_run_delayed_items(trans);
5231         if (err)
5232                 goto out_end_trans;
5233 
5234         if (block_rsv)
5235                 trans->block_rsv = block_rsv;
5236 
5237         /*
5238          * This will help us catch people modifying the fs tree while we're
5239          * dropping it.  It is unsafe to mess with the fs tree while it's being
5240          * dropped as we unlock the root node and parent nodes as we walk down
5241          * the tree, assuming nothing will change.  If something does change
5242          * then we'll have stale information and drop references to blocks we've
5243          * already dropped.
5244          */
5245         set_bit(BTRFS_ROOT_DELETING, &root->state);
5246         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5247                 level = btrfs_header_level(root->node);
5248                 path->nodes[level] = btrfs_lock_root_node(root);
5249                 btrfs_set_lock_blocking_write(path->nodes[level]);
5250                 path->slots[level] = 0;
5251                 path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5252                 memset(&wc->update_progress, 0,
5253                        sizeof(wc->update_progress));
5254         } else {
5255                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5256                 memcpy(&wc->update_progress, &key,
5257                        sizeof(wc->update_progress));
5258 
5259                 level = root_item->drop_level;
5260                 BUG_ON(level == 0);
5261                 path->lowest_level = level;
5262                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5263                 path->lowest_level = 0;
5264                 if (ret < 0) {
5265                         err = ret;
5266                         goto out_end_trans;
5267                 }
5268                 WARN_ON(ret > 0);
5269 
5270                 /*
5271                  * unlock our path, this is safe because only this
5272                  * function is allowed to delete this snapshot
5273                  */
5274                 btrfs_unlock_up_safe(path, 0);
5275 
5276                 level = btrfs_header_level(root->node);
5277                 while (1) {
5278                         btrfs_tree_lock(path->nodes[level]);
5279                         btrfs_set_lock_blocking_write(path->nodes[level]);
5280                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5281 
5282                         ret = btrfs_lookup_extent_info(trans, fs_info,
5283                                                 path->nodes[level]->start,
5284                                                 level, 1, &wc->refs[level],
5285                                                 &wc->flags[level]);
5286                         if (ret < 0) {
5287                                 err = ret;
5288                                 goto out_end_trans;
5289                         }
5290                         BUG_ON(wc->refs[level] == 0);
5291 
5292                         if (level == root_item->drop_level)
5293                                 break;
5294 
5295                         btrfs_tree_unlock(path->nodes[level]);
5296                         path->locks[level] = 0;
5297                         WARN_ON(wc->refs[level] != 1);
5298                         level--;
5299                 }
5300         }
5301 
5302         wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
5303         wc->level = level;
5304         wc->shared_level = -1;
5305         wc->stage = DROP_REFERENCE;
5306         wc->update_ref = update_ref;
5307         wc->keep_locks = 0;
5308         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5309 
5310         while (1) {
5311 
5312                 ret = walk_down_tree(trans, root, path, wc);
5313                 if (ret < 0) {
5314                         err = ret;
5315                         break;
5316                 }
5317 
5318                 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5319                 if (ret < 0) {
5320                         err = ret;
5321                         break;
5322                 }
5323 
5324                 if (ret > 0) {
5325                         BUG_ON(wc->stage != DROP_REFERENCE);
5326                         break;
5327                 }
5328 
5329                 if (wc->stage == DROP_REFERENCE) {
5330                         wc->drop_level = wc->level;
5331                         btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
5332                                               &wc->drop_progress,
5333                                               path->slots[wc->drop_level]);
5334                 }
5335                 btrfs_cpu_key_to_disk(&root_item->drop_progress,
5336                                       &wc->drop_progress);
5337                 root_item->drop_level = wc->drop_level;
5338 
5339                 BUG_ON(wc->level == 0);
5340                 if (btrfs_should_end_transaction(trans) ||
5341                     (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
5342                         ret = btrfs_update_root(trans, tree_root,
5343                                                 &root->root_key,
5344                                                 root_item);
5345                         if (ret) {
5346                                 btrfs_abort_transaction(trans, ret);
5347                                 err = ret;
5348                                 goto out_end_trans;
5349                         }
5350 
5351                         btrfs_end_transaction_throttle(trans);
5352                         if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
5353                                 btrfs_debug(fs_info,
5354                                             "drop snapshot early exit");
5355                                 err = -EAGAIN;
5356                                 goto out_free;
5357                         }
5358 
5359                         trans = btrfs_start_transaction(tree_root, 0);
5360                         if (IS_ERR(trans)) {
5361                                 err = PTR_ERR(trans);
5362                                 goto out_free;
5363                         }
5364                         if (block_rsv)
5365                                 trans->block_rsv = block_rsv;
5366                 }
5367         }
5368         btrfs_release_path(path);
5369         if (err)
5370                 goto out_end_trans;
5371 
5372         ret = btrfs_del_root(trans, &root->root_key);
5373         if (ret) {
5374                 btrfs_abort_transaction(trans, ret);
5375                 err = ret;
5376                 goto out_end_trans;
5377         }
5378 
5379         if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5380                 ret = btrfs_find_root(tree_root, &root->root_key, path,
5381                                       NULL, NULL);
5382                 if (ret < 0) {
5383                         btrfs_abort_transaction(trans, ret);
5384                         err = ret;
5385                         goto out_end_trans;
5386                 } else if (ret > 0) {
5387                         /* if we fail to delete the orphan item this time
5388                          * around, it'll get picked up the next time.
5389                          *
5390                          * The most common failure here is just -ENOENT.
5391                          */
5392                         btrfs_del_orphan_item(trans, tree_root,
5393                                               root->root_key.objectid);
5394                 }
5395         }
5396 
5397         if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state)) {
5398                 btrfs_add_dropped_root(trans, root);
5399         } else {
5400                 free_extent_buffer(root->node);
5401                 free_extent_buffer(root->commit_root);
5402                 btrfs_put_fs_root(root);
5403         }
5404         root_dropped = true;
5405 out_end_trans:
5406         btrfs_end_transaction_throttle(trans);
5407 out_free:
5408         kfree(wc);
5409         btrfs_free_path(path);
5410 out:
5411         /*
5412          * So if we need to stop dropping the snapshot for whatever reason we
5413          * need to make sure to add it back to the dead root list so that we
5414          * keep trying to do the work later.  This also cleans up roots if we
5415          * don't have it in the radix (like when we recover after a power fail
5416          * or unmount) so we don't leak memory.
5417          */
5418         if (!for_reloc && !root_dropped)
5419                 btrfs_add_dead_root(root);
5420         if (err && err != -EAGAIN)
5421                 btrfs_handle_fs_error(fs_info, err, NULL);
5422         return err;
5423 }
5424 
5425 /*
5426  * drop subtree rooted at tree block 'node'.
5427  *
5428  * NOTE: this function will unlock and release tree block 'node'
5429  * only used by relocation code
5430  */
5431 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5432                         struct btrfs_root *root,
5433                         struct extent_buffer *node,
5434                         struct extent_buffer *parent)
5435 {
5436         struct btrfs_fs_info *fs_info = root->fs_info;
5437         struct btrfs_path *path;
5438         struct walk_control *wc;
5439         int level;
5440         int parent_level;
5441         int ret = 0;
5442         int wret;
5443 
5444         BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5445 
5446         path = btrfs_alloc_path();
5447         if (!path)
5448                 return -ENOMEM;
5449 
5450         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5451         if (!wc) {
5452                 btrfs_free_path(path);
5453                 return -ENOMEM;
5454         }
5455 
5456         btrfs_assert_tree_locked(parent);
5457         parent_level = btrfs_header_level(parent);
5458         extent_buffer_get(parent);
5459         path->nodes[parent_level] = parent;
5460         path->slots[parent_level] = btrfs_header_nritems(parent);
5461 
5462         btrfs_assert_tree_locked(node);
5463         level = btrfs_header_level(node);
5464         path->nodes[level] = node;
5465         path->slots[level] = 0;
5466         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5467 
5468         wc->refs[parent_level] = 1;
5469         wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5470         wc->level = level;
5471         wc->shared_level = -1;
5472         wc->stage = DROP_REFERENCE;
5473         wc->update_ref = 0;
5474         wc->keep_locks = 1;
5475         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5476 
5477         while (1) {
5478                 wret = walk_down_tree(trans, root, path, wc);
5479                 if (wret < 0) {
5480                         ret = wret;
5481                         break;
5482                 }
5483 
5484                 wret = walk_up_tree(trans, root, path, wc, parent_level);
5485                 if (wret < 0)
5486                         ret = wret;
5487                 if (wret != 0)
5488                         break;
5489         }
5490 
5491         kfree(wc);
5492         btrfs_free_path(path);
5493         return ret;
5494 }
5495 
5496 /*
5497  * helper to account the unused space of all the readonly block group in the
5498  * space_info. takes mirrors into account.
5499  */
5500 u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
5501 {
5502         struct btrfs_block_group_cache *block_group;
5503         u64 free_bytes = 0;
5504         int factor;
5505 
5506         /* It's df, we don't care if it's racy */
5507         if (list_empty(&sinfo->ro_bgs))
5508                 return 0;
5509 
5510         spin_lock(&sinfo->lock);
5511         list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
5512                 spin_lock(&block_group->lock);
5513 
5514                 if (!block_group->ro) {
5515                         spin_unlock(&block_group->lock);
5516                         continue;
5517                 }
5518 
5519                 factor = btrfs_bg_type_to_factor(block_group->flags);
5520                 free_bytes += (block_group->key.offset -
5521                                btrfs_block_group_used(&block_group->item)) *
5522                                factor;
5523 
5524                 spin_unlock(&block_group->lock);
5525         }
5526         spin_unlock(&sinfo->lock);
5527 
5528         return free_bytes;
5529 }
5530 
5531 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
5532                                    u64 start, u64 end)
5533 {
5534         return unpin_extent_range(fs_info, start, end, false);
5535 }
5536 
5537 /*
5538  * It used to be that old block groups would be left around forever.
5539  * Iterating over them would be enough to trim unused space.  Since we
5540  * now automatically remove them, we also need to iterate over unallocated
5541  * space.
5542  *
5543  * We don't want a transaction for this since the discard may take a
5544  * substantial amount of time.  We don't require that a transaction be
5545  * running, but we do need to take a running transaction into account
5546  * to ensure that we're not discarding chunks that were released or
5547  * allocated in the current transaction.
5548  *
5549  * Holding the chunks lock will prevent other threads from allocating
5550  * or releasing chunks, but it won't prevent a running transaction
5551  * from committing and releasing the memory that the pending chunks
5552  * list head uses.  For that, we need to take a reference to the
5553  * transaction and hold the commit root sem.  We only need to hold
5554  * it while performing the free space search since we have already
5555  * held back allocations.
5556  */
5557 static int btrfs_trim_free_extents(struct btrfs_device *device, u64 *trimmed)
5558 {
5559         u64 start = SZ_1M, len = 0, end = 0;
5560         int ret;
5561 
5562         *trimmed = 0;
5563 
5564         /* Discard not supported = nothing to do. */
5565         if (!blk_queue_discard(bdev_get_queue(device->bdev)))
5566                 return 0;
5567 
5568         /* Not writable = nothing to do. */
5569         if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state))
5570                 return 0;
5571 
5572         /* No free space = nothing to do. */
5573         if (device->total_bytes <= device->bytes_used)
5574                 return 0;
5575 
5576         ret = 0;
5577 
5578         while (1) {
5579                 struct btrfs_fs_info *fs_info = device->fs_info;
5580                 u64 bytes;
5581 
5582                 ret = mutex_lock_interruptible(&fs_info->chunk_mutex);
5583                 if (ret)
5584                         break;
5585 
5586                 find_first_clear_extent_bit(&device->alloc_state, start,
5587                                             &start, &end,
5588                                             CHUNK_TRIMMED | CHUNK_ALLOCATED);
5589 
5590                 /* Ensure we skip the reserved area in the first 1M */
5591                 start = max_t(u64, start, SZ_1M);
5592 
5593                 /*
5594                  * If find_first_clear_extent_bit find a range that spans the
5595                  * end of the device it will set end to -1, in this case it's up
5596                  * to the caller to trim the value to the size of the device.
5597                  */
5598                 end = min(end, device->total_bytes - 1);
5599 
5600                 len = end - start + 1;
5601 
5602                 /* We didn't find any extents */
5603                 if (!len) {
5604                         mutex_unlock(&fs_info->chunk_mutex);
5605                         ret = 0;
5606                         break;
5607                 }
5608 
5609                 ret = btrfs_issue_discard(device->bdev, start, len,
5610                                           &bytes);
5611                 if (!ret)
5612                         set_extent_bits(&device->alloc_state, start,
5613                                         start + bytes - 1,
5614                                         CHUNK_TRIMMED);
5615                 mutex_unlock(&fs_info->chunk_mutex);
5616 
5617                 if (ret)
5618                         break;
5619 
5620                 start += len;
5621                 *trimmed += bytes;
5622 
5623                 if (fatal_signal_pending(current)) {
5624                         ret = -ERESTARTSYS;
5625                         break;
5626                 }
5627 
5628                 cond_resched();
5629         }
5630 
5631         return ret;
5632 }
5633 
5634 /*
5635  * Trim the whole filesystem by:
5636  * 1) trimming the free space in each block group
5637  * 2) trimming the unallocated space on each device
5638  *
5639  * This will also continue trimming even if a block group or device encounters
5640  * an error.  The return value will be the last error, or 0 if nothing bad
5641  * happens.
5642  */
5643 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range)
5644 {
5645         struct btrfs_block_group_cache *cache = NULL;
5646         struct btrfs_device *device;
5647         struct list_head *devices;
5648         u64 group_trimmed;
5649         u64 range_end = U64_MAX;
5650         u64 start;
5651         u64 end;
5652         u64 trimmed = 0;
5653         u64 bg_failed = 0;
5654         u64 dev_failed = 0;
5655         int bg_ret = 0;
5656         int dev_ret = 0;
5657         int ret = 0;
5658 
5659         /*
5660          * Check range overflow if range->len is set.
5661          * The default range->len is U64_MAX.
5662          */
5663         if (range->len != U64_MAX &&
5664             check_add_overflow(range->start, range->len, &range_end))
5665                 return -EINVAL;
5666 
5667         cache = btrfs_lookup_first_block_group(fs_info, range->start);
5668         for (; cache; cache = btrfs_next_block_group(cache)) {
5669                 if (cache->key.objectid >= range_end) {
5670                         btrfs_put_block_group(cache);
5671                         break;
5672                 }
5673 
5674                 start = max(range->start, cache->key.objectid);
5675                 end = min(range_end, cache->key.objectid + cache->key.offset);
5676 
5677                 if (end - start >= range->minlen) {
5678                         if (!btrfs_block_group_cache_done(cache)) {
5679                                 ret = btrfs_cache_block_group(cache, 0);
5680                                 if (ret) {
5681                                         bg_failed++;
5682                                         bg_ret = ret;
5683                                         continue;
5684                                 }
5685                                 ret = btrfs_wait_block_group_cache_done(cache);
5686                                 if (ret) {
5687                                         bg_failed++;
5688                                         bg_ret = ret;
5689                                         continue;
5690                                 }
5691                         }
5692                         ret = btrfs_trim_block_group(cache,
5693                                                      &group_trimmed,
5694                                                      start,
5695                                                      end,
5696                                                      range->minlen);
5697 
5698                         trimmed += group_trimmed;
5699                         if (ret) {
5700                                 bg_failed++;
5701                                 bg_ret = ret;
5702                                 continue;
5703                         }
5704                 }
5705         }
5706 
5707         if (bg_failed)
5708                 btrfs_warn(fs_info,
5709                         "failed to trim %llu block group(s), last error %d",
5710                         bg_failed, bg_ret);
5711         mutex_lock(&fs_info->fs_devices->device_list_mutex);
5712         devices = &fs_info->fs_devices->devices;
5713         list_for_each_entry(device, devices, dev_list) {
5714                 ret = btrfs_trim_free_extents(device, &group_trimmed);
5715                 if (ret) {
5716                         dev_failed++;
5717                         dev_ret = ret;
5718                         break;
5719                 }
5720 
5721                 trimmed += group_trimmed;
5722         }
5723         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
5724 
5725         if (dev_failed)
5726                 btrfs_warn(fs_info,
5727                         "failed to trim %llu device(s), last error %d",
5728                         dev_failed, dev_ret);
5729         range->len = trimmed;
5730         if (bg_ret)
5731                 return bg_ret;
5732         return dev_ret;
5733 }
5734 
5735 /*
5736  * btrfs_{start,end}_write_no_snapshotting() are similar to
5737  * mnt_{want,drop}_write(), they are used to prevent some tasks from writing
5738  * data into the page cache through nocow before the subvolume is snapshoted,
5739  * but flush the data into disk after the snapshot creation, or to prevent
5740  * operations while snapshotting is ongoing and that cause the snapshot to be
5741  * inconsistent (writes followed by expanding truncates for example).
5742  */
5743 void btrfs_end_write_no_snapshotting(struct btrfs_root *root)
5744 {
5745         percpu_counter_dec(&root->subv_writers->counter);
5746         cond_wake_up(&root->subv_writers->wait);
5747 }
5748 
5749 int btrfs_start_write_no_snapshotting(struct btrfs_root *root)
5750 {
5751         if (atomic_read(&root->will_be_snapshotted))
5752                 return 0;
5753 
5754         percpu_counter_inc(&root->subv_writers->counter);
5755         /*
5756          * Make sure counter is updated before we check for snapshot creation.
5757          */
5758         smp_mb();
5759         if (atomic_read(&root->will_be_snapshotted)) {
5760                 btrfs_end_write_no_snapshotting(root);
5761                 return 0;
5762         }
5763         return 1;
5764 }
5765 
5766 void btrfs_wait_for_snapshot_creation(struct btrfs_root *root)
5767 {
5768         while (true) {
5769                 int ret;
5770 
5771                 ret = btrfs_start_write_no_snapshotting(root);
5772                 if (ret)
5773                         break;
5774                 wait_var_event(&root->will_be_snapshotted,
5775                                !atomic_read(&root->will_be_snapshotted));
5776         }
5777 }
/* [<][>][^][v][top][bottom][index][help] */
root/fs/btrfs/extent-tree.c

DEFINITIONS
