fs/btrfs/extent-tree.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/sched.h>
   7 #include <linux/sched/signal.h>
   8 #include <linux/pagemap.h>
   9 #include <linux/writeback.h>
  10 #include <linux/blkdev.h>
  11 #include <linux/sort.h>
  12 #include <linux/rcupdate.h>
  13 #include <linux/kthread.h>
  14 #include <linux/slab.h>
  15 #include <linux/ratelimit.h>
  16 #include <linux/percpu_counter.h>
  17 #include <linux/lockdep.h>
  18 #include <linux/crc32c.h>
  19 #include "misc.h"
  20 #include "tree-log.h"
  21 #include "disk-io.h"
  22 #include "print-tree.h"
  23 #include "volumes.h"
  24 #include "raid56.h"
  25 #include "locking.h"
  26 #include "free-space-cache.h"
  27 #include "free-space-tree.h"
  28 #include "sysfs.h"
  29 #include "qgroup.h"
  30 #include "ref-verify.h"
  31 #include "space-info.h"
  32 #include "block-rsv.h"
  33 #include "delalloc-space.h"
  34 #include "block-group.h"
  35 #include "discard.h"
  36 #include "rcu-string.h"
  37
  38 #undef SCRAMBLE_DELAYED_REFS
  39
  40
  41 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
  42                                struct btrfs_delayed_ref_node *node, u64 parent,
  43                                u64 root_objectid, u64 owner_objectid,
  44                                u64 owner_offset, int refs_to_drop,
  45                                struct btrfs_delayed_extent_op *extra_op);
  46 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
  47                                     struct extent_buffer *leaf,
  48                                     struct btrfs_extent_item *ei);
  49 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
  50                                       u64 parent, u64 root_objectid,
  51                                       u64 flags, u64 owner, u64 offset,
  52                                       struct btrfs_key *ins, int ref_mod);
  53 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
  54                                      struct btrfs_delayed_ref_node *node,
  55                                      struct btrfs_delayed_extent_op *extent_op);
  56 static int find_next_key(struct btrfs_path *path, int level,
  57                          struct btrfs_key *key);
  58
  59 static int block_group_bits(struct btrfs_block_group *cache, u64 bits)
  60 {
  61         return (cache->flags & bits) == bits;
  62 }
  63
  64 int btrfs_add_excluded_extent(struct btrfs_fs_info *fs_info,
  65                               u64 start, u64 num_bytes)
  66 {
  67         u64 end = start + num_bytes - 1;
  68         set_extent_bits(&fs_info->excluded_extents, start, end,
  69                         EXTENT_UPTODATE);
  70         return 0;
  71 }
  72
  73 void btrfs_free_excluded_extents(struct btrfs_block_group *cache)
  74 {
  75         struct btrfs_fs_info *fs_info = cache->fs_info;
  76         u64 start, end;
  77
  78         start = cache->start;
  79         end = start + cache->length - 1;
  80
  81         clear_extent_bits(&fs_info->excluded_extents, start, end,
  82                           EXTENT_UPTODATE);
  83 }
  84
  85 /* simple helper to search for an existing data extent at a given offset */
  86 int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
  87 {
  88         int ret;
  89         struct btrfs_key key;
  90         struct btrfs_path *path;
  91
  92         path = btrfs_alloc_path();
  93         if (!path)
  94                 return -ENOMEM;
  95
  96         key.objectid = start;
  97         key.offset = len;
  98         key.type = BTRFS_EXTENT_ITEM_KEY;
  99         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
 100         btrfs_free_path(path);
 101         return ret;
 102 }
 103
 104 /*
 105  * helper function to lookup reference count and flags of a tree block.
 106  *
 107  * the head node for delayed ref is used to store the sum of all the
 108  * reference count modifications queued up in the rbtree. the head
 109  * node may also store the extent flags to set. This way you can check
 110  * to see what the reference count and extent flags would be if all of
 111  * the delayed refs are not processed.
 112  */
 113 int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 114                              struct btrfs_fs_info *fs_info, u64 bytenr,
 115                              u64 offset, int metadata, u64 *refs, u64 *flags)
 116 {
 117         struct btrfs_delayed_ref_head *head;
 118         struct btrfs_delayed_ref_root *delayed_refs;
 119         struct btrfs_path *path;
 120         struct btrfs_extent_item *ei;
 121         struct extent_buffer *leaf;
 122         struct btrfs_key key;
 123         u32 item_size;
 124         u64 num_refs;
 125         u64 extent_flags;
 126         int ret;
 127
 128         /*
 129          * If we don't have skinny metadata, don't bother doing anything
 130          * different
 131          */
 132         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
 133                 offset = fs_info->nodesize;
 134                 metadata = 0;
 135         }
 136
 137         path = btrfs_alloc_path();
 138         if (!path)
 139                 return -ENOMEM;
 140
 141         if (!trans) {
 142                 path->skip_locking = 1;
 143                 path->search_commit_root = 1;
 144         }
 145
 146 search_again:
 147         key.objectid = bytenr;
 148         key.offset = offset;
 149         if (metadata)
 150                 key.type = BTRFS_METADATA_ITEM_KEY;
 151         else
 152                 key.type = BTRFS_EXTENT_ITEM_KEY;
 153
 154         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
 155         if (ret < 0)
 156                 goto out_free;
 157
 158         if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
 159                 if (path->slots[0]) {
 160                         path->slots[0]--;
 161                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 162                                               path->slots[0]);
 163                         if (key.objectid == bytenr &&
 164                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 165                             key.offset == fs_info->nodesize)
 166                                 ret = 0;
 167                 }
 168         }
 169
 170         if (ret == 0) {
 171                 leaf = path->nodes[0];
 172                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 173                 if (item_size >= sizeof(*ei)) {
 174                         ei = btrfs_item_ptr(leaf, path->slots[0],
 175                                             struct btrfs_extent_item);
 176                         num_refs = btrfs_extent_refs(leaf, ei);
 177                         extent_flags = btrfs_extent_flags(leaf, ei);
 178                 } else {
 179                         ret = -EINVAL;
 180                         btrfs_print_v0_err(fs_info);
 181                         if (trans)
 182                                 btrfs_abort_transaction(trans, ret);
 183                         else
 184                                 btrfs_handle_fs_error(fs_info, ret, NULL);
 185
 186                         goto out_free;
 187                 }
 188
 189                 BUG_ON(num_refs == 0);
 190         } else {
 191                 num_refs = 0;
 192                 extent_flags = 0;
 193                 ret = 0;
 194         }
 195
 196         if (!trans)
 197                 goto out;
 198
 199         delayed_refs = &trans->transaction->delayed_refs;
 200         spin_lock(&delayed_refs->lock);
 201         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
 202         if (head) {
 203                 if (!mutex_trylock(&head->mutex)) {
 204                         refcount_inc(&head->refs);
 205                         spin_unlock(&delayed_refs->lock);
 206
 207                         btrfs_release_path(path);
 208
 209                         /*
 210                          * Mutex was contended, block until it's released and try
 211                          * again
 212                          */
 213                         mutex_lock(&head->mutex);
 214                         mutex_unlock(&head->mutex);
 215                         btrfs_put_delayed_ref_head(head);
 216                         goto search_again;
 217                 }
 218                 spin_lock(&head->lock);
 219                 if (head->extent_op && head->extent_op->update_flags)
 220                         extent_flags |= head->extent_op->flags_to_set;
 221                 else
 222                         BUG_ON(num_refs == 0);
 223
 224                 num_refs += head->ref_mod;
 225                 spin_unlock(&head->lock);
 226                 mutex_unlock(&head->mutex);
 227         }
 228         spin_unlock(&delayed_refs->lock);
 229 out:
 230         WARN_ON(num_refs == 0);
 231         if (refs)
 232                 *refs = num_refs;
 233         if (flags)
 234                 *flags = extent_flags;
 235 out_free:
 236         btrfs_free_path(path);
 237         return ret;
 238 }
 239
 240 /*
 241  * Back reference rules.  Back refs have three main goals:
 242  *
 243  * 1) differentiate between all holders of references to an extent so that
 244  *    when a reference is dropped we can make sure it was a valid reference
 245  *    before freeing the extent.
 246  *
 247  * 2) Provide enough information to quickly find the holders of an extent
 248  *    if we notice a given block is corrupted or bad.
 249  *
 250  * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 251  *    maintenance.  This is actually the same as #2, but with a slightly
 252  *    different use case.
 253  *
 254  * There are two kinds of back refs. The implicit back refs is optimized
 255  * for pointers in non-shared tree blocks. For a given pointer in a block,
 256  * back refs of this kind provide information about the block's owner tree
 257  * and the pointer's key. These information allow us to find the block by
 258  * b-tree searching. The full back refs is for pointers in tree blocks not
 259  * referenced by their owner trees. The location of tree block is recorded
 260  * in the back refs. Actually the full back refs is generic, and can be
 261  * used in all cases the implicit back refs is used. The major shortcoming
 262  * of the full back refs is its overhead. Every time a tree block gets
 263  * COWed, we have to update back refs entry for all pointers in it.
 264  *
 265  * For a newly allocated tree block, we use implicit back refs for
 266  * pointers in it. This means most tree related operations only involve
 267  * implicit back refs. For a tree block created in old transaction, the
 268  * only way to drop a reference to it is COW it. So we can detect the
 269  * event that tree block loses its owner tree's reference and do the
 270  * back refs conversion.
 271  *
 272  * When a tree block is COWed through a tree, there are four cases:
 273  *
 274  * The reference count of the block is one and the tree is the block's
 275  * owner tree. Nothing to do in this case.
 276  *
 277  * The reference count of the block is one and the tree is not the
 278  * block's owner tree. In this case, full back refs is used for pointers
 279  * in the block. Remove these full back refs, add implicit back refs for
 280  * every pointers in the new block.
 281  *
 282  * The reference count of the block is greater than one and the tree is
 283  * the block's owner tree. In this case, implicit back refs is used for
 284  * pointers in the block. Add full back refs for every pointers in the
 285  * block, increase lower level extents' reference counts. The original
 286  * implicit back refs are entailed to the new block.
 287  *
 288  * The reference count of the block is greater than one and the tree is
 289  * not the block's owner tree. Add implicit back refs for every pointer in
 290  * the new block, increase lower level extents' reference count.
 291  *
 292  * Back Reference Key composing:
 293  *
 294  * The key objectid corresponds to the first byte in the extent,
 295  * The key type is used to differentiate between types of back refs.
 296  * There are different meanings of the key offset for different types
 297  * of back refs.
 298  *
 299  * File extents can be referenced by:
 300  *
 301  * - multiple snapshots, subvolumes, or different generations in one subvol
 302  * - different files inside a single subvolume
 303  * - different offsets inside a file (bookend extents in file.c)
 304  *
 305  * The extent ref structure for the implicit back refs has fields for:
 306  *
 307  * - Objectid of the subvolume root
 308  * - objectid of the file holding the reference
 309  * - original offset in the file
 310  * - how many bookend extents
 311  *
 312  * The key offset for the implicit back refs is hash of the first
 313  * three fields.
 314  *
 315  * The extent ref structure for the full back refs has field for:
 316  *
 317  * - number of pointers in the tree leaf
 318  *
 319  * The key offset for the implicit back refs is the first byte of
 320  * the tree leaf
 321  *
 322  * When a file extent is allocated, The implicit back refs is used.
 323  * the fields are filled in:
 324  *
 325  *     (root_key.objectid, inode objectid, offset in file, 1)
 326  *
 327  * When a file extent is removed file truncation, we find the
 328  * corresponding implicit back refs and check the following fields:
 329  *
 330  *     (btrfs_header_owner(leaf), inode objectid, offset in file)
 331  *
 332  * Btree extents can be referenced by:
 333  *
 334  * - Different subvolumes
 335  *
 336  * Both the implicit back refs and the full back refs for tree blocks
 337  * only consist of key. The key offset for the implicit back refs is
 338  * objectid of block's owner tree. The key offset for the full back refs
 339  * is the first byte of parent block.
 340  *
 341  * When implicit back refs is used, information about the lowest key and
 342  * level of the tree block are required. These information are stored in
 343  * tree block info structure.
 344  */
 345
 346 /*
 347  * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
 348  * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
 349  * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
 350  */
 351 int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
 352                                      struct btrfs_extent_inline_ref *iref,
 353                                      enum btrfs_inline_ref_type is_data)
 354 {
 355         int type = btrfs_extent_inline_ref_type(eb, iref);
 356         u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
 357
 358         if (type == BTRFS_TREE_BLOCK_REF_KEY ||
 359             type == BTRFS_SHARED_BLOCK_REF_KEY ||
 360             type == BTRFS_SHARED_DATA_REF_KEY ||
 361             type == BTRFS_EXTENT_DATA_REF_KEY) {
 362                 if (is_data == BTRFS_REF_TYPE_BLOCK) {
 363                         if (type == BTRFS_TREE_BLOCK_REF_KEY)
 364                                 return type;
 365                         if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
 366                                 ASSERT(eb->fs_info);
 367                                 /*
 368                                  * Every shared one has parent tree block,
 369                                  * which must be aligned to sector size.
 370                                  */
 371                                 if (offset &&
 372                                     IS_ALIGNED(offset, eb->fs_info->sectorsize))
 373                                         return type;
 374                         }
 375                 } else if (is_data == BTRFS_REF_TYPE_DATA) {
 376                         if (type == BTRFS_EXTENT_DATA_REF_KEY)
 377                                 return type;
 378                         if (type == BTRFS_SHARED_DATA_REF_KEY) {
 379                                 ASSERT(eb->fs_info);
 380                                 /*
 381                                  * Every shared one has parent tree block,
 382                                  * which must be aligned to sector size.
 383                                  */
 384                                 if (offset &&
 385                                     IS_ALIGNED(offset, eb->fs_info->sectorsize))
 386                                         return type;
 387                         }
 388                 } else {
 389                         ASSERT(is_data == BTRFS_REF_TYPE_ANY);
 390                         return type;
 391                 }
 392         }
 393
 394         btrfs_print_leaf((struct extent_buffer *)eb);
 395         btrfs_err(eb->fs_info,
 396                   "eb %llu iref 0x%lx invalid extent inline ref type %d",
 397                   eb->start, (unsigned long)iref, type);
 398         WARN_ON(1);
 399
 400         return BTRFS_REF_TYPE_INVALID;
 401 }
 402
 403 u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
 404 {
 405         u32 high_crc = ~(u32)0;
 406         u32 low_crc = ~(u32)0;
 407         __le64 lenum;
 408
 409         lenum = cpu_to_le64(root_objectid);
 410         high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
 411         lenum = cpu_to_le64(owner);
 412         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 413         lenum = cpu_to_le64(offset);
 414         low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
 415
 416         return ((u64)high_crc << 31) ^ (u64)low_crc;
 417 }
 418
 419 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
 420                                      struct btrfs_extent_data_ref *ref)
 421 {
 422         return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
 423                                     btrfs_extent_data_ref_objectid(leaf, ref),
 424                                     btrfs_extent_data_ref_offset(leaf, ref));
 425 }
 426
 427 static int match_extent_data_ref(struct extent_buffer *leaf,
 428                                  struct btrfs_extent_data_ref *ref,
 429                                  u64 root_objectid, u64 owner, u64 offset)
 430 {
 431         if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
 432             btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
 433             btrfs_extent_data_ref_offset(leaf, ref) != offset)
 434                 return 0;
 435         return 1;
 436 }
 437
 438 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
 439                                            struct btrfs_path *path,
 440                                            u64 bytenr, u64 parent,
 441                                            u64 root_objectid,
 442                                            u64 owner, u64 offset)
 443 {
 444         struct btrfs_root *root = trans->fs_info->extent_root;
 445         struct btrfs_key key;
 446         struct btrfs_extent_data_ref *ref;
 447         struct extent_buffer *leaf;
 448         u32 nritems;
 449         int ret;
 450         int recow;
 451         int err = -ENOENT;
 452
 453         key.objectid = bytenr;
 454         if (parent) {
 455                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 456                 key.offset = parent;
 457         } else {
 458                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 459                 key.offset = hash_extent_data_ref(root_objectid,
 460                                                   owner, offset);
 461         }
 462 again:
 463         recow = 0;
 464         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 465         if (ret < 0) {
 466                 err = ret;
 467                 goto fail;
 468         }
 469
 470         if (parent) {
 471                 if (!ret)
 472                         return 0;
 473                 goto fail;
 474         }
 475
 476         leaf = path->nodes[0];
 477         nritems = btrfs_header_nritems(leaf);
 478         while (1) {
 479                 if (path->slots[0] >= nritems) {
 480                         ret = btrfs_next_leaf(root, path);
 481                         if (ret < 0)
 482                                 err = ret;
 483                         if (ret)
 484                                 goto fail;
 485
 486                         leaf = path->nodes[0];
 487                         nritems = btrfs_header_nritems(leaf);
 488                         recow = 1;
 489                 }
 490
 491                 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 492                 if (key.objectid != bytenr ||
 493                     key.type != BTRFS_EXTENT_DATA_REF_KEY)
 494                         goto fail;
 495
 496                 ref = btrfs_item_ptr(leaf, path->slots[0],
 497                                      struct btrfs_extent_data_ref);
 498
 499                 if (match_extent_data_ref(leaf, ref, root_objectid,
 500                                           owner, offset)) {
 501                         if (recow) {
 502                                 btrfs_release_path(path);
 503                                 goto again;
 504                         }
 505                         err = 0;
 506                         break;
 507                 }
 508                 path->slots[0]++;
 509         }
 510 fail:
 511         return err;
 512 }
 513
 514 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
 515                                            struct btrfs_path *path,
 516                                            u64 bytenr, u64 parent,
 517                                            u64 root_objectid, u64 owner,
 518                                            u64 offset, int refs_to_add)
 519 {
 520         struct btrfs_root *root = trans->fs_info->extent_root;
 521         struct btrfs_key key;
 522         struct extent_buffer *leaf;
 523         u32 size;
 524         u32 num_refs;
 525         int ret;
 526
 527         key.objectid = bytenr;
 528         if (parent) {
 529                 key.type = BTRFS_SHARED_DATA_REF_KEY;
 530                 key.offset = parent;
 531                 size = sizeof(struct btrfs_shared_data_ref);
 532         } else {
 533                 key.type = BTRFS_EXTENT_DATA_REF_KEY;
 534                 key.offset = hash_extent_data_ref(root_objectid,
 535                                                   owner, offset);
 536                 size = sizeof(struct btrfs_extent_data_ref);
 537         }
 538
 539         ret = btrfs_insert_empty_item(trans, root, path, &key, size);
 540         if (ret && ret != -EEXIST)
 541                 goto fail;
 542
 543         leaf = path->nodes[0];
 544         if (parent) {
 545                 struct btrfs_shared_data_ref *ref;
 546                 ref = btrfs_item_ptr(leaf, path->slots[0],
 547                                      struct btrfs_shared_data_ref);
 548                 if (ret == 0) {
 549                         btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
 550                 } else {
 551                         num_refs = btrfs_shared_data_ref_count(leaf, ref);
 552                         num_refs += refs_to_add;
 553                         btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
 554                 }
 555         } else {
 556                 struct btrfs_extent_data_ref *ref;
 557                 while (ret == -EEXIST) {
 558                         ref = btrfs_item_ptr(leaf, path->slots[0],
 559                                              struct btrfs_extent_data_ref);
 560                         if (match_extent_data_ref(leaf, ref, root_objectid,
 561                                                   owner, offset))
 562                                 break;
 563                         btrfs_release_path(path);
 564                         key.offset++;
 565                         ret = btrfs_insert_empty_item(trans, root, path, &key,
 566                                                       size);
 567                         if (ret && ret != -EEXIST)
 568                                 goto fail;
 569
 570                         leaf = path->nodes[0];
 571                 }
 572                 ref = btrfs_item_ptr(leaf, path->slots[0],
 573                                      struct btrfs_extent_data_ref);
 574                 if (ret == 0) {
 575                         btrfs_set_extent_data_ref_root(leaf, ref,
 576                                                        root_objectid);
 577                         btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
 578                         btrfs_set_extent_data_ref_offset(leaf, ref, offset);
 579                         btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
 580                 } else {
 581                         num_refs = btrfs_extent_data_ref_count(leaf, ref);
 582                         num_refs += refs_to_add;
 583                         btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
 584                 }
 585         }
 586         btrfs_mark_buffer_dirty(leaf);
 587         ret = 0;
 588 fail:
 589         btrfs_release_path(path);
 590         return ret;
 591 }
 592
 593 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
 594                                            struct btrfs_path *path,
 595                                            int refs_to_drop, int *last_ref)
 596 {
 597         struct btrfs_key key;
 598         struct btrfs_extent_data_ref *ref1 = NULL;
 599         struct btrfs_shared_data_ref *ref2 = NULL;
 600         struct extent_buffer *leaf;
 601         u32 num_refs = 0;
 602         int ret = 0;
 603
 604         leaf = path->nodes[0];
 605         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 606
 607         if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 608                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 609                                       struct btrfs_extent_data_ref);
 610                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 611         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 612                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 613                                       struct btrfs_shared_data_ref);
 614                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 615         } else if (unlikely(key.type == BTRFS_EXTENT_REF_V0_KEY)) {
 616                 btrfs_print_v0_err(trans->fs_info);
 617                 btrfs_abort_transaction(trans, -EINVAL);
 618                 return -EINVAL;
 619         } else {
 620                 BUG();
 621         }
 622
 623         BUG_ON(num_refs < refs_to_drop);
 624         num_refs -= refs_to_drop;
 625
 626         if (num_refs == 0) {
 627                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
 628                 *last_ref = 1;
 629         } else {
 630                 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
 631                         btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
 632                 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
 633                         btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
 634                 btrfs_mark_buffer_dirty(leaf);
 635         }
 636         return ret;
 637 }
 638
 639 static noinline u32 extent_data_ref_count(struct btrfs_path *path,
 640                                           struct btrfs_extent_inline_ref *iref)
 641 {
 642         struct btrfs_key key;
 643         struct extent_buffer *leaf;
 644         struct btrfs_extent_data_ref *ref1;
 645         struct btrfs_shared_data_ref *ref2;
 646         u32 num_refs = 0;
 647         int type;
 648
 649         leaf = path->nodes[0];
 650         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
 651
 652         BUG_ON(key.type == BTRFS_EXTENT_REF_V0_KEY);
 653         if (iref) {
 654                 /*
 655                  * If type is invalid, we should have bailed out earlier than
 656                  * this call.
 657                  */
 658                 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
 659                 ASSERT(type != BTRFS_REF_TYPE_INVALID);
 660                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 661                         ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
 662                         num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 663                 } else {
 664                         ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
 665                         num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 666                 }
 667         } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
 668                 ref1 = btrfs_item_ptr(leaf, path->slots[0],
 669                                       struct btrfs_extent_data_ref);
 670                 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
 671         } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
 672                 ref2 = btrfs_item_ptr(leaf, path->slots[0],
 673                                       struct btrfs_shared_data_ref);
 674                 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
 675         } else {
 676                 WARN_ON(1);
 677         }
 678         return num_refs;
 679 }
 680
 681 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
 682                                           struct btrfs_path *path,
 683                                           u64 bytenr, u64 parent,
 684                                           u64 root_objectid)
 685 {
 686         struct btrfs_root *root = trans->fs_info->extent_root;
 687         struct btrfs_key key;
 688         int ret;
 689
 690         key.objectid = bytenr;
 691         if (parent) {
 692                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 693                 key.offset = parent;
 694         } else {
 695                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 696                 key.offset = root_objectid;
 697         }
 698
 699         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
 700         if (ret > 0)
 701                 ret = -ENOENT;
 702         return ret;
 703 }
 704
 705 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
 706                                           struct btrfs_path *path,
 707                                           u64 bytenr, u64 parent,
 708                                           u64 root_objectid)
 709 {
 710         struct btrfs_key key;
 711         int ret;
 712
 713         key.objectid = bytenr;
 714         if (parent) {
 715                 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
 716                 key.offset = parent;
 717         } else {
 718                 key.type = BTRFS_TREE_BLOCK_REF_KEY;
 719                 key.offset = root_objectid;
 720         }
 721
 722         ret = btrfs_insert_empty_item(trans, trans->fs_info->extent_root,
 723                                       path, &key, 0);
 724         btrfs_release_path(path);
 725         return ret;
 726 }
 727
 728 static inline int extent_ref_type(u64 parent, u64 owner)
 729 {
 730         int type;
 731         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
 732                 if (parent > 0)
 733                         type = BTRFS_SHARED_BLOCK_REF_KEY;
 734                 else
 735                         type = BTRFS_TREE_BLOCK_REF_KEY;
 736         } else {
 737                 if (parent > 0)
 738                         type = BTRFS_SHARED_DATA_REF_KEY;
 739                 else
 740                         type = BTRFS_EXTENT_DATA_REF_KEY;
 741         }
 742         return type;
 743 }
 744
 745 static int find_next_key(struct btrfs_path *path, int level,
 746                          struct btrfs_key *key)
 747
 748 {
 749         for (; level < BTRFS_MAX_LEVEL; level++) {
 750                 if (!path->nodes[level])
 751                         break;
 752                 if (path->slots[level] + 1 >=
 753                     btrfs_header_nritems(path->nodes[level]))
 754                         continue;
 755                 if (level == 0)
 756                         btrfs_item_key_to_cpu(path->nodes[level], key,
 757                                               path->slots[level] + 1);
 758                 else
 759                         btrfs_node_key_to_cpu(path->nodes[level], key,
 760                                               path->slots[level] + 1);
 761                 return 0;
 762         }
 763         return 1;
 764 }
 765
 766 /*
 767  * look for inline back ref. if back ref is found, *ref_ret is set
 768  * to the address of inline back ref, and 0 is returned.
 769  *
 770  * if back ref isn't found, *ref_ret is set to the address where it
 771  * should be inserted, and -ENOENT is returned.
 772  *
 773  * if insert is true and there are too many inline back refs, the path
 774  * points to the extent item, and -EAGAIN is returned.
 775  *
 776  * NOTE: inline back refs are ordered in the same way that back ref
 777  *       items in the tree are ordered.
 778  */
 779 static noinline_for_stack
 780 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 781                                  struct btrfs_path *path,
 782                                  struct btrfs_extent_inline_ref **ref_ret,
 783                                  u64 bytenr, u64 num_bytes,
 784                                  u64 parent, u64 root_objectid,
 785                                  u64 owner, u64 offset, int insert)
 786 {
 787         struct btrfs_fs_info *fs_info = trans->fs_info;
 788         struct btrfs_root *root = fs_info->extent_root;
 789         struct btrfs_key key;
 790         struct extent_buffer *leaf;
 791         struct btrfs_extent_item *ei;
 792         struct btrfs_extent_inline_ref *iref;
 793         u64 flags;
 794         u64 item_size;
 795         unsigned long ptr;
 796         unsigned long end;
 797         int extra_size;
 798         int type;
 799         int want;
 800         int ret;
 801         int err = 0;
 802         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
 803         int needed;
 804
 805         key.objectid = bytenr;
 806         key.type = BTRFS_EXTENT_ITEM_KEY;
 807         key.offset = num_bytes;
 808
 809         want = extent_ref_type(parent, owner);
 810         if (insert) {
 811                 extra_size = btrfs_extent_inline_ref_size(want);
 812                 path->keep_locks = 1;
 813         } else
 814                 extra_size = -1;
 815
 816         /*
 817          * Owner is our level, so we can just add one to get the level for the
 818          * block we are interested in.
 819          */
 820         if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
 821                 key.type = BTRFS_METADATA_ITEM_KEY;
 822                 key.offset = owner;
 823         }
 824
 825 again:
 826         ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
 827         if (ret < 0) {
 828                 err = ret;
 829                 goto out;
 830         }
 831
 832         /*
 833          * We may be a newly converted file system which still has the old fat
 834          * extent entries for metadata, so try and see if we have one of those.
 835          */
 836         if (ret > 0 && skinny_metadata) {
 837                 skinny_metadata = false;
 838                 if (path->slots[0]) {
 839                         path->slots[0]--;
 840                         btrfs_item_key_to_cpu(path->nodes[0], &key,
 841                                               path->slots[0]);
 842                         if (key.objectid == bytenr &&
 843                             key.type == BTRFS_EXTENT_ITEM_KEY &&
 844                             key.offset == num_bytes)
 845                                 ret = 0;
 846                 }
 847                 if (ret) {
 848                         key.objectid = bytenr;
 849                         key.type = BTRFS_EXTENT_ITEM_KEY;
 850                         key.offset = num_bytes;
 851                         btrfs_release_path(path);
 852                         goto again;
 853                 }
 854         }
 855
 856         if (ret && !insert) {
 857                 err = -ENOENT;
 858                 goto out;
 859         } else if (WARN_ON(ret)) {
 860                 btrfs_print_leaf(path->nodes[0]);
 861                 btrfs_err(fs_info,
 862 "extent item not found for insert, bytenr %llu num_bytes %llu parent %llu root_objectid %llu owner %llu offset %llu",
 863                           bytenr, num_bytes, parent, root_objectid, owner,
 864                           offset);
 865                 err = -EIO;
 866                 goto out;
 867         }
 868
 869         leaf = path->nodes[0];
 870         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
 871         if (unlikely(item_size < sizeof(*ei))) {
 872                 err = -EINVAL;
 873                 btrfs_print_v0_err(fs_info);
 874                 btrfs_abort_transaction(trans, err);
 875                 goto out;
 876         }
 877
 878         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 879         flags = btrfs_extent_flags(leaf, ei);
 880
 881         ptr = (unsigned long)(ei + 1);
 882         end = (unsigned long)ei + item_size;
 883
 884         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
 885                 ptr += sizeof(struct btrfs_tree_block_info);
 886                 BUG_ON(ptr > end);
 887         }
 888
 889         if (owner >= BTRFS_FIRST_FREE_OBJECTID)
 890                 needed = BTRFS_REF_TYPE_DATA;
 891         else
 892                 needed = BTRFS_REF_TYPE_BLOCK;
 893
 894         err = -ENOENT;
 895         while (1) {
 896                 if (ptr >= end) {
 897                         WARN_ON(ptr > end);
 898                         break;
 899                 }
 900                 iref = (struct btrfs_extent_inline_ref *)ptr;
 901                 type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
 902                 if (type == BTRFS_REF_TYPE_INVALID) {
 903                         err = -EUCLEAN;
 904                         goto out;
 905                 }
 906
 907                 if (want < type)
 908                         break;
 909                 if (want > type) {
 910                         ptr += btrfs_extent_inline_ref_size(type);
 911                         continue;
 912                 }
 913
 914                 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
 915                         struct btrfs_extent_data_ref *dref;
 916                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
 917                         if (match_extent_data_ref(leaf, dref, root_objectid,
 918                                                   owner, offset)) {
 919                                 err = 0;
 920                                 break;
 921                         }
 922                         if (hash_extent_data_ref_item(leaf, dref) <
 923                             hash_extent_data_ref(root_objectid, owner, offset))
 924                                 break;
 925                 } else {
 926                         u64 ref_offset;
 927                         ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
 928                         if (parent > 0) {
 929                                 if (parent == ref_offset) {
 930                                         err = 0;
 931                                         break;
 932                                 }
 933                                 if (ref_offset < parent)
 934                                         break;
 935                         } else {
 936                                 if (root_objectid == ref_offset) {
 937                                         err = 0;
 938                                         break;
 939                                 }
 940                                 if (ref_offset < root_objectid)
 941                                         break;
 942                         }
 943                 }
 944                 ptr += btrfs_extent_inline_ref_size(type);
 945         }
 946         if (err == -ENOENT && insert) {
 947                 if (item_size + extra_size >=
 948                     BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
 949                         err = -EAGAIN;
 950                         goto out;
 951                 }
 952                 /*
 953                  * To add new inline back ref, we have to make sure
 954                  * there is no corresponding back ref item.
 955                  * For simplicity, we just do not add new inline back
 956                  * ref if there is any kind of item for this block
 957                  */
 958                 if (find_next_key(path, 0, &key) == 0 &&
 959                     key.objectid == bytenr &&
 960                     key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
 961                         err = -EAGAIN;
 962                         goto out;
 963                 }
 964         }
 965         *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
 966 out:
 967         if (insert) {
 968                 path->keep_locks = 0;
 969                 btrfs_unlock_up_safe(path, 1);
 970         }
 971         return err;
 972 }
 973
 974 /*
 975  * helper to add new inline back ref
 976  */
 977 static noinline_for_stack
 978 void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
 979                                  struct btrfs_path *path,
 980                                  struct btrfs_extent_inline_ref *iref,
 981                                  u64 parent, u64 root_objectid,
 982                                  u64 owner, u64 offset, int refs_to_add,
 983                                  struct btrfs_delayed_extent_op *extent_op)
 984 {
 985         struct extent_buffer *leaf;
 986         struct btrfs_extent_item *ei;
 987         unsigned long ptr;
 988         unsigned long end;
 989         unsigned long item_offset;
 990         u64 refs;
 991         int size;
 992         int type;
 993
 994         leaf = path->nodes[0];
 995         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
 996         item_offset = (unsigned long)iref - (unsigned long)ei;
 997
 998         type = extent_ref_type(parent, owner);
 999         size = btrfs_extent_inline_ref_size(type);
1000
1001         btrfs_extend_item(path, size);
1002
1003         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1004         refs = btrfs_extent_refs(leaf, ei);
1005         refs += refs_to_add;
1006         btrfs_set_extent_refs(leaf, ei, refs);
1007         if (extent_op)
1008                 __run_delayed_extent_op(extent_op, leaf, ei);
1009
1010         ptr = (unsigned long)ei + item_offset;
1011         end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1012         if (ptr < end - size)
1013                 memmove_extent_buffer(leaf, ptr + size, ptr,
1014                                       end - size - ptr);
1015
1016         iref = (struct btrfs_extent_inline_ref *)ptr;
1017         btrfs_set_extent_inline_ref_type(leaf, iref, type);
1018         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1019                 struct btrfs_extent_data_ref *dref;
1020                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1021                 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1022                 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1023                 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1024                 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1025         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1026                 struct btrfs_shared_data_ref *sref;
1027                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1028                 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1029                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1030         } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1031                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1032         } else {
1033                 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1034         }
1035         btrfs_mark_buffer_dirty(leaf);
1036 }
1037
1038 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1039                                  struct btrfs_path *path,
1040                                  struct btrfs_extent_inline_ref **ref_ret,
1041                                  u64 bytenr, u64 num_bytes, u64 parent,
1042                                  u64 root_objectid, u64 owner, u64 offset)
1043 {
1044         int ret;
1045
1046         ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
1047                                            num_bytes, parent, root_objectid,
1048                                            owner, offset, 0);
1049         if (ret != -ENOENT)
1050                 return ret;
1051
1052         btrfs_release_path(path);
1053         *ref_ret = NULL;
1054
1055         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1056                 ret = lookup_tree_block_ref(trans, path, bytenr, parent,
1057                                             root_objectid);
1058         } else {
1059                 ret = lookup_extent_data_ref(trans, path, bytenr, parent,
1060                                              root_objectid, owner, offset);
1061         }
1062         return ret;
1063 }
1064
1065 /*
1066  * helper to update/remove inline back ref
1067  */
1068 static noinline_for_stack
1069 void update_inline_extent_backref(struct btrfs_path *path,
1070                                   struct btrfs_extent_inline_ref *iref,
1071                                   int refs_to_mod,
1072                                   struct btrfs_delayed_extent_op *extent_op,
1073                                   int *last_ref)
1074 {
1075         struct extent_buffer *leaf = path->nodes[0];
1076         struct btrfs_extent_item *ei;
1077         struct btrfs_extent_data_ref *dref = NULL;
1078         struct btrfs_shared_data_ref *sref = NULL;
1079         unsigned long ptr;
1080         unsigned long end;
1081         u32 item_size;
1082         int size;
1083         int type;
1084         u64 refs;
1085
1086         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1087         refs = btrfs_extent_refs(leaf, ei);
1088         WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1089         refs += refs_to_mod;
1090         btrfs_set_extent_refs(leaf, ei, refs);
1091         if (extent_op)
1092                 __run_delayed_extent_op(extent_op, leaf, ei);
1093
1094         /*
1095          * If type is invalid, we should have bailed out after
1096          * lookup_inline_extent_backref().
1097          */
1098         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
1099         ASSERT(type != BTRFS_REF_TYPE_INVALID);
1100
1101         if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1102                 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1103                 refs = btrfs_extent_data_ref_count(leaf, dref);
1104         } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1105                 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1106                 refs = btrfs_shared_data_ref_count(leaf, sref);
1107         } else {
1108                 refs = 1;
1109                 BUG_ON(refs_to_mod != -1);
1110         }
1111
1112         BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1113         refs += refs_to_mod;
1114
1115         if (refs > 0) {
1116                 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1117                         btrfs_set_extent_data_ref_count(leaf, dref, refs);
1118                 else
1119                         btrfs_set_shared_data_ref_count(leaf, sref, refs);
1120         } else {
1121                 *last_ref = 1;
1122                 size =  btrfs_extent_inline_ref_size(type);
1123                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1124                 ptr = (unsigned long)iref;
1125                 end = (unsigned long)ei + item_size;
1126                 if (ptr + size < end)
1127                         memmove_extent_buffer(leaf, ptr, ptr + size,
1128                                               end - ptr - size);
1129                 item_size -= size;
1130                 btrfs_truncate_item(path, item_size, 1);
1131         }
1132         btrfs_mark_buffer_dirty(leaf);
1133 }
1134
1135 static noinline_for_stack
1136 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1137                                  struct btrfs_path *path,
1138                                  u64 bytenr, u64 num_bytes, u64 parent,
1139                                  u64 root_objectid, u64 owner,
1140                                  u64 offset, int refs_to_add,
1141                                  struct btrfs_delayed_extent_op *extent_op)
1142 {
1143         struct btrfs_extent_inline_ref *iref;
1144         int ret;
1145
1146         ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
1147                                            num_bytes, parent, root_objectid,
1148                                            owner, offset, 1);
1149         if (ret == 0) {
1150                 /*
1151                  * We're adding refs to a tree block we already own, this
1152                  * should not happen at all.
1153                  */
1154                 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1155                         btrfs_crit(trans->fs_info,
1156 "adding refs to an existing tree ref, bytenr %llu num_bytes %llu root_objectid %llu",
1157                                    bytenr, num_bytes, root_objectid);
1158                         if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) {
1159                                 WARN_ON(1);
1160                                 btrfs_crit(trans->fs_info,
1161                         "path->slots[0]=%d path->nodes[0]:", path->slots[0]);
1162                                 btrfs_print_leaf(path->nodes[0]);
1163                         }
1164                         return -EUCLEAN;
1165                 }
1166                 update_inline_extent_backref(path, iref, refs_to_add,
1167                                              extent_op, NULL);
1168         } else if (ret == -ENOENT) {
1169                 setup_inline_extent_backref(trans->fs_info, path, iref, parent,
1170                                             root_objectid, owner, offset,
1171                                             refs_to_add, extent_op);
1172                 ret = 0;
1173         }
1174         return ret;
1175 }
1176
1177 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1178                                  struct btrfs_path *path,
1179                                  struct btrfs_extent_inline_ref *iref,
1180                                  int refs_to_drop, int is_data, int *last_ref)
1181 {
1182         int ret = 0;
1183
1184         BUG_ON(!is_data && refs_to_drop != 1);
1185         if (iref) {
1186                 update_inline_extent_backref(path, iref, -refs_to_drop, NULL,
1187                                              last_ref);
1188         } else if (is_data) {
1189                 ret = remove_extent_data_ref(trans, path, refs_to_drop,
1190                                              last_ref);
1191         } else {
1192                 *last_ref = 1;
1193                 ret = btrfs_del_item(trans, trans->fs_info->extent_root, path);
1194         }
1195         return ret;
1196 }
1197
1198 static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
1199                                u64 *discarded_bytes)
1200 {
1201         int j, ret = 0;
1202         u64 bytes_left, end;
1203         u64 aligned_start = ALIGN(start, 1 << 9);
1204
1205         /* Adjust the range to be aligned to 512B sectors if necessary. */
1206         if (start != aligned_start) {
1207                 len -= aligned_start - start;
1208                 len = round_down(len, 1 << 9);
1209                 start = aligned_start;
1210         }
1211
1212         *discarded_bytes = 0;
1213
1214         if (!len)
1215                 return 0;
1216
1217         end = start + len;
1218         bytes_left = len;
1219
1220         /* Skip any superblocks on this device. */
1221         for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
1222                 u64 sb_start = btrfs_sb_offset(j);
1223                 u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
1224                 u64 size = sb_start - start;
1225
1226                 if (!in_range(sb_start, start, bytes_left) &&
1227                     !in_range(sb_end, start, bytes_left) &&
1228                     !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
1229                         continue;
1230
1231                 /*
1232                  * Superblock spans beginning of range.  Adjust start and
1233                  * try again.
1234                  */
1235                 if (sb_start <= start) {
1236                         start += sb_end - start;
1237                         if (start > end) {
1238                                 bytes_left = 0;
1239                                 break;
1240                         }
1241                         bytes_left = end - start;
1242                         continue;
1243                 }
1244
1245                 if (size) {
1246                         ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
1247                                                    GFP_NOFS, 0);
1248                         if (!ret)
1249                                 *discarded_bytes += size;
1250                         else if (ret != -EOPNOTSUPP)
1251                                 return ret;
1252                 }
1253
1254                 start = sb_end;
1255                 if (start > end) {
1256                         bytes_left = 0;
1257                         break;
1258                 }
1259                 bytes_left = end - start;
1260         }
1261
1262         if (bytes_left) {
1263                 ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
1264                                            GFP_NOFS, 0);
1265                 if (!ret)
1266                         *discarded_bytes += bytes_left;
1267         }
1268         return ret;
1269 }
1270
1271 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
1272                          u64 num_bytes, u64 *actual_bytes)
1273 {
1274         int ret = 0;
1275         u64 discarded_bytes = 0;
1276         u64 end = bytenr + num_bytes;
1277         u64 cur = bytenr;
1278         struct btrfs_bio *bbio = NULL;
1279
1280
1281         /*
1282          * Avoid races with device replace and make sure our bbio has devices
1283          * associated to its stripes that don't go away while we are discarding.
1284          */
1285         btrfs_bio_counter_inc_blocked(fs_info);
1286         while (cur < end) {
1287                 struct btrfs_bio_stripe *stripe;
1288                 int i;
1289
1290                 num_bytes = end - cur;
1291                 /* Tell the block device(s) that the sectors can be discarded */
1292                 ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, cur,
1293                                       &num_bytes, &bbio, 0);
1294                 /*
1295                  * Error can be -ENOMEM, -ENOENT (no such chunk mapping) or
1296                  * -EOPNOTSUPP. For any such error, @num_bytes is not updated,
1297                  * thus we can't continue anyway.
1298                  */
1299                 if (ret < 0)
1300                         goto out;
1301
1302                 stripe = bbio->stripes;
1303                 for (i = 0; i < bbio->num_stripes; i++, stripe++) {
1304                         u64 bytes;
1305                         struct request_queue *req_q;
1306                         struct btrfs_device *device = stripe->dev;
1307
1308                         if (!device->bdev) {
1309                                 ASSERT(btrfs_test_opt(fs_info, DEGRADED));
1310                                 continue;
1311                         }
1312                         req_q = bdev_get_queue(device->bdev);
1313                         if (!blk_queue_discard(req_q))
1314                                 continue;
1315
1316                         if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state))
1317                                 continue;
1318
1319                         ret = btrfs_issue_discard(device->bdev,
1320                                                   stripe->physical,
1321                                                   stripe->length,
1322                                                   &bytes);
1323                         if (!ret) {
1324                                 discarded_bytes += bytes;
1325                         } else if (ret != -EOPNOTSUPP) {
1326                                 /*
1327                                  * Logic errors or -ENOMEM, or -EIO, but
1328                                  * unlikely to happen.
1329                                  *
1330                                  * And since there are two loops, explicitly
1331                                  * go to out to avoid confusion.
1332                                  */
1333                                 btrfs_put_bbio(bbio);
1334                                 goto out;
1335                         }
1336
1337                         /*
1338                          * Just in case we get back EOPNOTSUPP for some reason,
1339                          * just ignore the return value so we don't screw up
1340                          * people calling discard_extent.
1341                          */
1342                         ret = 0;
1343                 }
1344                 btrfs_put_bbio(bbio);
1345                 cur += num_bytes;
1346         }
1347 out:
1348         btrfs_bio_counter_dec(fs_info);
1349
1350         if (actual_bytes)
1351                 *actual_bytes = discarded_bytes;
1352
1353
1354         if (ret == -EOPNOTSUPP)
1355                 ret = 0;
1356         return ret;
1357 }
1358
1359 /* Can return -ENOMEM */
1360 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1361                          struct btrfs_ref *generic_ref)
1362 {
1363         struct btrfs_fs_info *fs_info = trans->fs_info;
1364         int ret;
1365
1366         ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
1367                generic_ref->action);
1368         BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
1369                generic_ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID);
1370
1371         if (generic_ref->type == BTRFS_REF_METADATA)
1372                 ret = btrfs_add_delayed_tree_ref(trans, generic_ref, NULL);
1373         else
1374                 ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0);
1375
1376         btrfs_ref_tree_mod(fs_info, generic_ref);
1377
1378         return ret;
1379 }
1380
1381 /*
1382  * __btrfs_inc_extent_ref - insert backreference for a given extent
1383  *
1384  * The counterpart is in __btrfs_free_extent(), with examples and more details
1385  * how it works.
1386  *
1387  * @trans:          Handle of transaction
1388  *
1389  * @node:           The delayed ref node used to get the bytenr/length for
1390  *                  extent whose references are incremented.
1391  *
1392  * @parent:         If this is a shared extent (BTRFS_SHARED_DATA_REF_KEY/
1393  *                  BTRFS_SHARED_BLOCK_REF_KEY) then it holds the logical
1394  *                  bytenr of the parent block. Since new extents are always
1395  *                  created with indirect references, this will only be the case
1396  *                  when relocating a shared extent. In that case, root_objectid
1397  *                  will be BTRFS_TREE_RELOC_OBJECTID. Otheriwse, parent must
1398  *                  be 0
1399  *
1400  * @root_objectid:  The id of the root where this modification has originated,
1401  *                  this can be either one of the well-known metadata trees or
1402  *                  the subvolume id which references this extent.
1403  *
1404  * @owner:          For data extents it is the inode number of the owning file.
1405  *                  For metadata extents this parameter holds the level in the
1406  *                  tree of the extent.
1407  *
1408  * @offset:         For metadata extents the offset is ignored and is currently
1409  *                  always passed as 0. For data extents it is the fileoffset
1410  *                  this extent belongs to.
1411  *
1412  * @refs_to_add     Number of references to add
1413  *
1414  * @extent_op       Pointer to a structure, holding information necessary when
1415  *                  updating a tree block's flags
1416  *
1417  */
1418 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1419                                   struct btrfs_delayed_ref_node *node,
1420                                   u64 parent, u64 root_objectid,
1421                                   u64 owner, u64 offset, int refs_to_add,
1422                                   struct btrfs_delayed_extent_op *extent_op)
1423 {
1424         struct btrfs_path *path;
1425         struct extent_buffer *leaf;
1426         struct btrfs_extent_item *item;
1427         struct btrfs_key key;
1428         u64 bytenr = node->bytenr;
1429         u64 num_bytes = node->num_bytes;
1430         u64 refs;
1431         int ret;
1432
1433         path = btrfs_alloc_path();
1434         if (!path)
1435                 return -ENOMEM;
1436
1437         path->leave_spinning = 1;
1438         /* this will setup the path even if it fails to insert the back ref */
1439         ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
1440                                            parent, root_objectid, owner,
1441                                            offset, refs_to_add, extent_op);
1442         if ((ret < 0 && ret != -EAGAIN) || !ret)
1443                 goto out;
1444
1445         /*
1446          * Ok we had -EAGAIN which means we didn't have space to insert and
1447          * inline extent ref, so just update the reference count and add a
1448          * normal backref.
1449          */
1450         leaf = path->nodes[0];
1451         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1452         item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1453         refs = btrfs_extent_refs(leaf, item);
1454         btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1455         if (extent_op)
1456                 __run_delayed_extent_op(extent_op, leaf, item);
1457
1458         btrfs_mark_buffer_dirty(leaf);
1459         btrfs_release_path(path);
1460
1461         path->leave_spinning = 1;
1462         /* now insert the actual backref */
1463         if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1464                 BUG_ON(refs_to_add != 1);
1465                 ret = insert_tree_block_ref(trans, path, bytenr, parent,
1466                                             root_objectid);
1467         } else {
1468                 ret = insert_extent_data_ref(trans, path, bytenr, parent,
1469                                              root_objectid, owner, offset,
1470                                              refs_to_add);
1471         }
1472         if (ret)
1473                 btrfs_abort_transaction(trans, ret);
1474 out:
1475         btrfs_free_path(path);
1476         return ret;
1477 }
1478
1479 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1480                                 struct btrfs_delayed_ref_node *node,
1481                                 struct btrfs_delayed_extent_op *extent_op,
1482                                 int insert_reserved)
1483 {
1484         int ret = 0;
1485         struct btrfs_delayed_data_ref *ref;
1486         struct btrfs_key ins;
1487         u64 parent = 0;
1488         u64 ref_root = 0;
1489         u64 flags = 0;
1490
1491         ins.objectid = node->bytenr;
1492         ins.offset = node->num_bytes;
1493         ins.type = BTRFS_EXTENT_ITEM_KEY;
1494
1495         ref = btrfs_delayed_node_to_data_ref(node);
1496         trace_run_delayed_data_ref(trans->fs_info, node, ref, node->action);
1497
1498         if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1499                 parent = ref->parent;
1500         ref_root = ref->root;
1501
1502         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1503                 if (extent_op)
1504                         flags |= extent_op->flags_to_set;
1505                 ret = alloc_reserved_file_extent(trans, parent, ref_root,
1506                                                  flags, ref->objectid,
1507                                                  ref->offset, &ins,
1508                                                  node->ref_mod);
1509         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1510                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1511                                              ref->objectid, ref->offset,
1512                                              node->ref_mod, extent_op);
1513         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1514                 ret = __btrfs_free_extent(trans, node, parent,
1515                                           ref_root, ref->objectid,
1516                                           ref->offset, node->ref_mod,
1517                                           extent_op);
1518         } else {
1519                 BUG();
1520         }
1521         return ret;
1522 }
1523
1524 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1525                                     struct extent_buffer *leaf,
1526                                     struct btrfs_extent_item *ei)
1527 {
1528         u64 flags = btrfs_extent_flags(leaf, ei);
1529         if (extent_op->update_flags) {
1530                 flags |= extent_op->flags_to_set;
1531                 btrfs_set_extent_flags(leaf, ei, flags);
1532         }
1533
1534         if (extent_op->update_key) {
1535                 struct btrfs_tree_block_info *bi;
1536                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1537                 bi = (struct btrfs_tree_block_info *)(ei + 1);
1538                 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1539         }
1540 }
1541
1542 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1543                                  struct btrfs_delayed_ref_head *head,
1544                                  struct btrfs_delayed_extent_op *extent_op)
1545 {
1546         struct btrfs_fs_info *fs_info = trans->fs_info;
1547         struct btrfs_key key;
1548         struct btrfs_path *path;
1549         struct btrfs_extent_item *ei;
1550         struct extent_buffer *leaf;
1551         u32 item_size;
1552         int ret;
1553         int err = 0;
1554         int metadata = !extent_op->is_data;
1555
1556         if (TRANS_ABORTED(trans))
1557                 return 0;
1558
1559         if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1560                 metadata = 0;
1561
1562         path = btrfs_alloc_path();
1563         if (!path)
1564                 return -ENOMEM;
1565
1566         key.objectid = head->bytenr;
1567
1568         if (metadata) {
1569                 key.type = BTRFS_METADATA_ITEM_KEY;
1570                 key.offset = extent_op->level;
1571         } else {
1572                 key.type = BTRFS_EXTENT_ITEM_KEY;
1573                 key.offset = head->num_bytes;
1574         }
1575
1576 again:
1577         path->leave_spinning = 1;
1578         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
1579         if (ret < 0) {
1580                 err = ret;
1581                 goto out;
1582         }
1583         if (ret > 0) {
1584                 if (metadata) {
1585                         if (path->slots[0] > 0) {
1586                                 path->slots[0]--;
1587                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
1588                                                       path->slots[0]);
1589                                 if (key.objectid == head->bytenr &&
1590                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
1591                                     key.offset == head->num_bytes)
1592                                         ret = 0;
1593                         }
1594                         if (ret > 0) {
1595                                 btrfs_release_path(path);
1596                                 metadata = 0;
1597
1598                                 key.objectid = head->bytenr;
1599                                 key.offset = head->num_bytes;
1600                                 key.type = BTRFS_EXTENT_ITEM_KEY;
1601                                 goto again;
1602                         }
1603                 } else {
1604                         err = -EIO;
1605                         goto out;
1606                 }
1607         }
1608
1609         leaf = path->nodes[0];
1610         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1611
1612         if (unlikely(item_size < sizeof(*ei))) {
1613                 err = -EINVAL;
1614                 btrfs_print_v0_err(fs_info);
1615                 btrfs_abort_transaction(trans, err);
1616                 goto out;
1617         }
1618
1619         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1620         __run_delayed_extent_op(extent_op, leaf, ei);
1621
1622         btrfs_mark_buffer_dirty(leaf);
1623 out:
1624         btrfs_free_path(path);
1625         return err;
1626 }
1627
1628 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1629                                 struct btrfs_delayed_ref_node *node,
1630                                 struct btrfs_delayed_extent_op *extent_op,
1631                                 int insert_reserved)
1632 {
1633         int ret = 0;
1634         struct btrfs_delayed_tree_ref *ref;
1635         u64 parent = 0;
1636         u64 ref_root = 0;
1637
1638         ref = btrfs_delayed_node_to_tree_ref(node);
1639         trace_run_delayed_tree_ref(trans->fs_info, node, ref, node->action);
1640
1641         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1642                 parent = ref->parent;
1643         ref_root = ref->root;
1644
1645         if (unlikely(node->ref_mod != 1)) {
1646                 btrfs_err(trans->fs_info,
1647         "btree block %llu has %d references rather than 1: action %d ref_root %llu parent %llu",
1648                           node->bytenr, node->ref_mod, node->action, ref_root,
1649                           parent);
1650                 return -EUCLEAN;
1651         }
1652         if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1653                 BUG_ON(!extent_op || !extent_op->update_flags);
1654                 ret = alloc_reserved_tree_block(trans, node, extent_op);
1655         } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1656                 ret = __btrfs_inc_extent_ref(trans, node, parent, ref_root,
1657                                              ref->level, 0, 1, extent_op);
1658         } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1659                 ret = __btrfs_free_extent(trans, node, parent, ref_root,
1660                                           ref->level, 0, 1, extent_op);
1661         } else {
1662                 BUG();
1663         }
1664         return ret;
1665 }
1666
1667 /* helper function to actually process a single delayed ref entry */
1668 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1669                                struct btrfs_delayed_ref_node *node,
1670                                struct btrfs_delayed_extent_op *extent_op,
1671                                int insert_reserved)
1672 {
1673         int ret = 0;
1674
1675         if (TRANS_ABORTED(trans)) {
1676                 if (insert_reserved)
1677                         btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1678                 return 0;
1679         }
1680
1681         if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1682             node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1683                 ret = run_delayed_tree_ref(trans, node, extent_op,
1684                                            insert_reserved);
1685         else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1686                  node->type == BTRFS_SHARED_DATA_REF_KEY)
1687                 ret = run_delayed_data_ref(trans, node, extent_op,
1688                                            insert_reserved);
1689         else
1690                 BUG();
1691         if (ret && insert_reserved)
1692                 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1693         if (ret < 0)
1694                 btrfs_err(trans->fs_info,
1695 "failed to run delayed ref for logical %llu num_bytes %llu type %u action %u ref_mod %d: %d",
1696                           node->bytenr, node->num_bytes, node->type,
1697                           node->action, node->ref_mod, ret);
1698         return ret;
1699 }
1700
1701 static inline struct btrfs_delayed_ref_node *
1702 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1703 {
1704         struct btrfs_delayed_ref_node *ref;
1705
1706         if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
1707                 return NULL;
1708
1709         /*
1710          * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
1711          * This is to prevent a ref count from going down to zero, which deletes
1712          * the extent item from the extent tree, when there still are references
1713          * to add, which would fail because they would not find the extent item.
1714          */
1715         if (!list_empty(&head->ref_add_list))
1716                 return list_first_entry(&head->ref_add_list,
1717                                 struct btrfs_delayed_ref_node, add_list);
1718
1719         ref = rb_entry(rb_first_cached(&head->ref_tree),
1720                        struct btrfs_delayed_ref_node, ref_node);
1721         ASSERT(list_empty(&ref->add_list));
1722         return ref;
1723 }
1724
1725 static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
1726                                       struct btrfs_delayed_ref_head *head)
1727 {
1728         spin_lock(&delayed_refs->lock);
1729         head->processing = 0;
1730         delayed_refs->num_heads_ready++;
1731         spin_unlock(&delayed_refs->lock);
1732         btrfs_delayed_ref_unlock(head);
1733 }
1734
1735 static struct btrfs_delayed_extent_op *cleanup_extent_op(
1736                                 struct btrfs_delayed_ref_head *head)
1737 {
1738         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
1739
1740         if (!extent_op)
1741                 return NULL;
1742
1743         if (head->must_insert_reserved) {
1744                 head->extent_op = NULL;
1745                 btrfs_free_delayed_extent_op(extent_op);
1746                 return NULL;
1747         }
1748         return extent_op;
1749 }
1750
1751 static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
1752                                      struct btrfs_delayed_ref_head *head)
1753 {
1754         struct btrfs_delayed_extent_op *extent_op;
1755         int ret;
1756
1757         extent_op = cleanup_extent_op(head);
1758         if (!extent_op)
1759                 return 0;
1760         head->extent_op = NULL;
1761         spin_unlock(&head->lock);
1762         ret = run_delayed_extent_op(trans, head, extent_op);
1763         btrfs_free_delayed_extent_op(extent_op);
1764         return ret ? ret : 1;
1765 }
1766
1767 void btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
1768                                   struct btrfs_delayed_ref_root *delayed_refs,
1769                                   struct btrfs_delayed_ref_head *head)
1770 {
1771         int nr_items = 1;       /* Dropping this ref head update. */
1772
1773         /*
1774          * We had csum deletions accounted for in our delayed refs rsv, we need
1775          * to drop the csum leaves for this update from our delayed_refs_rsv.
1776          */
1777         if (head->total_ref_mod < 0 && head->is_data) {
1778                 spin_lock(&delayed_refs->lock);
1779                 delayed_refs->pending_csums -= head->num_bytes;
1780                 spin_unlock(&delayed_refs->lock);
1781                 nr_items += btrfs_csum_bytes_to_leaves(fs_info, head->num_bytes);
1782         }
1783
1784         /*
1785          * We were dropping refs, or had a new ref and dropped it, and thus must
1786          * adjust down our total_bytes_pinned, the space may or may not have
1787          * been pinned and so is accounted for properly in the pinned space by
1788          * now.
1789          */
1790         if (head->total_ref_mod < 0 ||
1791             (head->total_ref_mod == 0 && head->must_insert_reserved)) {
1792                 u64 flags = btrfs_ref_head_to_space_flags(head);
1793
1794                 btrfs_mod_total_bytes_pinned(fs_info, flags, -head->num_bytes);
1795         }
1796
1797         btrfs_delayed_refs_rsv_release(fs_info, nr_items);
1798 }
1799
1800 static int cleanup_ref_head(struct btrfs_trans_handle *trans,
1801                             struct btrfs_delayed_ref_head *head)
1802 {
1803
1804         struct btrfs_fs_info *fs_info = trans->fs_info;
1805         struct btrfs_delayed_ref_root *delayed_refs;
1806         int ret;
1807
1808         delayed_refs = &trans->transaction->delayed_refs;
1809
1810         ret = run_and_cleanup_extent_op(trans, head);
1811         if (ret < 0) {
1812                 unselect_delayed_ref_head(delayed_refs, head);
1813                 btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
1814                 return ret;
1815         } else if (ret) {
1816                 return ret;
1817         }
1818
1819         /*
1820          * Need to drop our head ref lock and re-acquire the delayed ref lock
1821          * and then re-check to make sure nobody got added.
1822          */
1823         spin_unlock(&head->lock);
1824         spin_lock(&delayed_refs->lock);
1825         spin_lock(&head->lock);
1826         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
1827                 spin_unlock(&head->lock);
1828                 spin_unlock(&delayed_refs->lock);
1829                 return 1;
1830         }
1831         btrfs_delete_ref_head(delayed_refs, head);
1832         spin_unlock(&head->lock);
1833         spin_unlock(&delayed_refs->lock);
1834
1835         if (head->must_insert_reserved) {
1836                 btrfs_pin_extent(trans, head->bytenr, head->num_bytes, 1);
1837                 if (head->is_data) {
1838                         ret = btrfs_del_csums(trans, fs_info->csum_root,
1839                                               head->bytenr, head->num_bytes);
1840                 }
1841         }
1842
1843         btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
1844
1845         trace_run_delayed_ref_head(fs_info, head, 0);
1846         btrfs_delayed_ref_unlock(head);
1847         btrfs_put_delayed_ref_head(head);
1848         return ret;
1849 }
1850
1851 static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
1852                                         struct btrfs_trans_handle *trans)
1853 {
1854         struct btrfs_delayed_ref_root *delayed_refs =
1855                 &trans->transaction->delayed_refs;
1856         struct btrfs_delayed_ref_head *head = NULL;
1857         int ret;
1858
1859         spin_lock(&delayed_refs->lock);
1860         head = btrfs_select_ref_head(delayed_refs);
1861         if (!head) {
1862                 spin_unlock(&delayed_refs->lock);
1863                 return head;
1864         }
1865
1866         /*
1867          * Grab the lock that says we are going to process all the refs for
1868          * this head
1869          */
1870         ret = btrfs_delayed_ref_lock(delayed_refs, head);
1871         spin_unlock(&delayed_refs->lock);
1872
1873         /*
1874          * We may have dropped the spin lock to get the head mutex lock, and
1875          * that might have given someone else time to free the head.  If that's
1876          * true, it has been removed from our list and we can move on.
1877          */
1878         if (ret == -EAGAIN)
1879                 head = ERR_PTR(-EAGAIN);
1880
1881         return head;
1882 }
1883
1884 static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
1885                                     struct btrfs_delayed_ref_head *locked_ref,
1886                                     unsigned long *run_refs)
1887 {
1888         struct btrfs_fs_info *fs_info = trans->fs_info;
1889         struct btrfs_delayed_ref_root *delayed_refs;
1890         struct btrfs_delayed_extent_op *extent_op;
1891         struct btrfs_delayed_ref_node *ref;
1892         int must_insert_reserved = 0;
1893         int ret;
1894
1895         delayed_refs = &trans->transaction->delayed_refs;
1896
1897         lockdep_assert_held(&locked_ref->mutex);
1898         lockdep_assert_held(&locked_ref->lock);
1899
1900         while ((ref = select_delayed_ref(locked_ref))) {
1901                 if (ref->seq &&
1902                     btrfs_check_delayed_seq(fs_info, ref->seq)) {
1903                         spin_unlock(&locked_ref->lock);
1904                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1905                         return -EAGAIN;
1906                 }
1907
1908                 (*run_refs)++;
1909                 ref->in_tree = 0;
1910                 rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
1911                 RB_CLEAR_NODE(&ref->ref_node);
1912                 if (!list_empty(&ref->add_list))
1913                         list_del(&ref->add_list);
1914                 /*
1915                  * When we play the delayed ref, also correct the ref_mod on
1916                  * head
1917                  */
1918                 switch (ref->action) {
1919                 case BTRFS_ADD_DELAYED_REF:
1920                 case BTRFS_ADD_DELAYED_EXTENT:
1921                         locked_ref->ref_mod -= ref->ref_mod;
1922                         break;
1923                 case BTRFS_DROP_DELAYED_REF:
1924                         locked_ref->ref_mod += ref->ref_mod;
1925                         break;
1926                 default:
1927                         WARN_ON(1);
1928                 }
1929                 atomic_dec(&delayed_refs->num_entries);
1930
1931                 /*
1932                  * Record the must_insert_reserved flag before we drop the
1933                  * spin lock.
1934                  */
1935                 must_insert_reserved = locked_ref->must_insert_reserved;
1936                 locked_ref->must_insert_reserved = 0;
1937
1938                 extent_op = locked_ref->extent_op;
1939                 locked_ref->extent_op = NULL;
1940                 spin_unlock(&locked_ref->lock);
1941
1942                 ret = run_one_delayed_ref(trans, ref, extent_op,
1943                                           must_insert_reserved);
1944
1945                 btrfs_free_delayed_extent_op(extent_op);
1946                 if (ret) {
1947                         unselect_delayed_ref_head(delayed_refs, locked_ref);
1948                         btrfs_put_delayed_ref(ref);
1949                         return ret;
1950                 }
1951
1952                 btrfs_put_delayed_ref(ref);
1953                 cond_resched();
1954
1955                 spin_lock(&locked_ref->lock);
1956                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
1957         }
1958
1959         return 0;
1960 }
1961
1962 /*
1963  * Returns 0 on success or if called with an already aborted transaction.
1964  * Returns -ENOMEM or -EIO on failure and will abort the transaction.
1965  */
1966 static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
1967                                              unsigned long nr)
1968 {
1969         struct btrfs_fs_info *fs_info = trans->fs_info;
1970         struct btrfs_delayed_ref_root *delayed_refs;
1971         struct btrfs_delayed_ref_head *locked_ref = NULL;
1972         ktime_t start = ktime_get();
1973         int ret;
1974         unsigned long count = 0;
1975         unsigned long actual_count = 0;
1976
1977         delayed_refs = &trans->transaction->delayed_refs;
1978         do {
1979                 if (!locked_ref) {
1980                         locked_ref = btrfs_obtain_ref_head(trans);
1981                         if (IS_ERR_OR_NULL(locked_ref)) {
1982                                 if (PTR_ERR(locked_ref) == -EAGAIN) {
1983                                         continue;
1984                                 } else {
1985                                         break;
1986                                 }
1987                         }
1988                         count++;
1989                 }
1990                 /*
1991                  * We need to try and merge add/drops of the same ref since we
1992                  * can run into issues with relocate dropping the implicit ref
1993                  * and then it being added back again before the drop can
1994                  * finish.  If we merged anything we need to re-loop so we can
1995                  * get a good ref.
1996                  * Or we can get node references of the same type that weren't
1997                  * merged when created due to bumps in the tree mod seq, and
1998                  * we need to merge them to prevent adding an inline extent
1999                  * backref before dropping it (triggering a BUG_ON at
2000                  * insert_inline_extent_backref()).
2001                  */
2002                 spin_lock(&locked_ref->lock);
2003                 btrfs_merge_delayed_refs(trans, delayed_refs, locked_ref);
2004
2005                 ret = btrfs_run_delayed_refs_for_head(trans, locked_ref,
2006                                                       &actual_count);
2007                 if (ret < 0 && ret != -EAGAIN) {
2008                         /*
2009                          * Error, btrfs_run_delayed_refs_for_head already
2010                          * unlocked everything so just bail out
2011                          */
2012                         return ret;
2013                 } else if (!ret) {
2014                         /*
2015                          * Success, perform the usual cleanup of a processed
2016                          * head
2017                          */
2018                         ret = cleanup_ref_head(trans, locked_ref);
2019                         if (ret > 0 ) {
2020                                 /* We dropped our lock, we need to loop. */
2021                                 ret = 0;
2022                                 continue;
2023                         } else if (ret) {
2024                                 return ret;
2025                         }
2026                 }
2027
2028                 /*
2029                  * Either success case or btrfs_run_delayed_refs_for_head
2030                  * returned -EAGAIN, meaning we need to select another head
2031                  */
2032
2033                 locked_ref = NULL;
2034                 cond_resched();
2035         } while ((nr != -1 && count < nr) || locked_ref);
2036
2037         /*
2038          * We don't want to include ref heads since we can have empty ref heads
2039          * and those will drastically skew our runtime down since we just do
2040          * accounting, no actual extent tree updates.
2041          */
2042         if (actual_count > 0) {
2043                 u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
2044                 u64 avg;
2045
2046                 /*
2047                  * We weigh the current average higher than our current runtime
2048                  * to avoid large swings in the average.
2049                  */
2050                 spin_lock(&delayed_refs->lock);
2051                 avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
2052                 fs_info->avg_delayed_ref_runtime = avg >> 2;    /* div by 4 */
2053                 spin_unlock(&delayed_refs->lock);
2054         }
2055         return 0;
2056 }
2057
2058 #ifdef SCRAMBLE_DELAYED_REFS
2059 /*
2060  * Normally delayed refs get processed in ascending bytenr order. This
2061  * correlates in most cases to the order added. To expose dependencies on this
2062  * order, we start to process the tree in the middle instead of the beginning
2063  */
2064 static u64 find_middle(struct rb_root *root)
2065 {
2066         struct rb_node *n = root->rb_node;
2067         struct btrfs_delayed_ref_node *entry;
2068         int alt = 1;
2069         u64 middle;
2070         u64 first = 0, last = 0;
2071
2072         n = rb_first(root);
2073         if (n) {
2074                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2075                 first = entry->bytenr;
2076         }
2077         n = rb_last(root);
2078         if (n) {
2079                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2080                 last = entry->bytenr;
2081         }
2082         n = root->rb_node;
2083
2084         while (n) {
2085                 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2086                 WARN_ON(!entry->in_tree);
2087
2088                 middle = entry->bytenr;
2089
2090                 if (alt)
2091                         n = n->rb_left;
2092                 else
2093                         n = n->rb_right;
2094
2095                 alt = 1 - alt;
2096         }
2097         return middle;
2098 }
2099 #endif
2100
2101 /*
2102  * Takes the number of bytes to be csumm'ed and figures out how many leaves it
2103  * would require to store the csums for that many bytes.
2104  */
2105 u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
2106 {
2107         u64 csum_size;
2108         u64 num_csums_per_leaf;
2109         u64 num_csums;
2110
2111         csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
2112         num_csums_per_leaf = div64_u64(csum_size,
2113                         (u64)btrfs_super_csum_size(fs_info->super_copy));
2114         num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
2115         num_csums += num_csums_per_leaf - 1;
2116         num_csums = div64_u64(num_csums, num_csums_per_leaf);
2117         return num_csums;
2118 }
2119
2120 /*
2121  * this starts processing the delayed reference count updates and
2122  * extent insertions we have queued up so far.  count can be
2123  * 0, which means to process everything in the tree at the start
2124  * of the run (but not newly added entries), or it can be some target
2125  * number you'd like to process.
2126  *
2127  * Returns 0 on success or if called with an aborted transaction
2128  * Returns <0 on error and aborts the transaction
2129  */
2130 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2131                            unsigned long count)
2132 {
2133         struct btrfs_fs_info *fs_info = trans->fs_info;
2134         struct rb_node *node;
2135         struct btrfs_delayed_ref_root *delayed_refs;
2136         struct btrfs_delayed_ref_head *head;
2137         int ret;
2138         int run_all = count == (unsigned long)-1;
2139
2140         /* We'll clean this up in btrfs_cleanup_transaction */
2141         if (TRANS_ABORTED(trans))
2142                 return 0;
2143
2144         if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
2145                 return 0;
2146
2147         delayed_refs = &trans->transaction->delayed_refs;
2148         if (count == 0)
2149                 count = atomic_read(&delayed_refs->num_entries) * 2;
2150
2151 again:
2152 #ifdef SCRAMBLE_DELAYED_REFS
2153         delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
2154 #endif
2155         ret = __btrfs_run_delayed_refs(trans, count);
2156         if (ret < 0) {
2157                 btrfs_abort_transaction(trans, ret);
2158                 return ret;
2159         }
2160
2161         if (run_all) {
2162                 btrfs_create_pending_block_groups(trans);
2163
2164                 spin_lock(&delayed_refs->lock);
2165                 node = rb_first_cached(&delayed_refs->href_root);
2166                 if (!node) {
2167                         spin_unlock(&delayed_refs->lock);
2168                         goto out;
2169                 }
2170                 head = rb_entry(node, struct btrfs_delayed_ref_head,
2171                                 href_node);
2172                 refcount_inc(&head->refs);
2173                 spin_unlock(&delayed_refs->lock);
2174
2175                 /* Mutex was contended, block until it's released and retry. */
2176                 mutex_lock(&head->mutex);
2177                 mutex_unlock(&head->mutex);
2178
2179                 btrfs_put_delayed_ref_head(head);
2180                 cond_resched();
2181                 goto again;
2182         }
2183 out:
2184         return 0;
2185 }
2186
2187 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2188                                 struct extent_buffer *eb, u64 flags,
2189                                 int level, int is_data)
2190 {
2191         struct btrfs_delayed_extent_op *extent_op;
2192         int ret;
2193
2194         extent_op = btrfs_alloc_delayed_extent_op();
2195         if (!extent_op)
2196                 return -ENOMEM;
2197
2198         extent_op->flags_to_set = flags;
2199         extent_op->update_flags = true;
2200         extent_op->update_key = false;
2201         extent_op->is_data = is_data ? true : false;
2202         extent_op->level = level;
2203
2204         ret = btrfs_add_delayed_extent_op(trans, eb->start, eb->len, extent_op);
2205         if (ret)
2206                 btrfs_free_delayed_extent_op(extent_op);
2207         return ret;
2208 }
2209
2210 static noinline int check_delayed_ref(struct btrfs_root *root,
2211                                       struct btrfs_path *path,
2212                                       u64 objectid, u64 offset, u64 bytenr)
2213 {
2214         struct btrfs_delayed_ref_head *head;
2215         struct btrfs_delayed_ref_node *ref;
2216         struct btrfs_delayed_data_ref *data_ref;
2217         struct btrfs_delayed_ref_root *delayed_refs;
2218         struct btrfs_transaction *cur_trans;
2219         struct rb_node *node;
2220         int ret = 0;
2221
2222         spin_lock(&root->fs_info->trans_lock);
2223         cur_trans = root->fs_info->running_transaction;
2224         if (cur_trans)
2225                 refcount_inc(&cur_trans->use_count);
2226         spin_unlock(&root->fs_info->trans_lock);
2227         if (!cur_trans)
2228                 return 0;
2229
2230         delayed_refs = &cur_trans->delayed_refs;
2231         spin_lock(&delayed_refs->lock);
2232         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
2233         if (!head) {
2234                 spin_unlock(&delayed_refs->lock);
2235                 btrfs_put_transaction(cur_trans);
2236                 return 0;
2237         }
2238
2239         if (!mutex_trylock(&head->mutex)) {
2240                 refcount_inc(&head->refs);
2241                 spin_unlock(&delayed_refs->lock);
2242
2243                 btrfs_release_path(path);
2244
2245                 /*
2246                  * Mutex was contended, block until it's released and let
2247                  * caller try again
2248                  */
2249                 mutex_lock(&head->mutex);
2250                 mutex_unlock(&head->mutex);
2251                 btrfs_put_delayed_ref_head(head);
2252                 btrfs_put_transaction(cur_trans);
2253                 return -EAGAIN;
2254         }
2255         spin_unlock(&delayed_refs->lock);
2256
2257         spin_lock(&head->lock);
2258         /*
2259          * XXX: We should replace this with a proper search function in the
2260          * future.
2261          */
2262         for (node = rb_first_cached(&head->ref_tree); node;
2263              node = rb_next(node)) {
2264                 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
2265                 /* If it's a shared ref we know a cross reference exists */
2266                 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
2267                         ret = 1;
2268                         break;
2269                 }
2270
2271                 data_ref = btrfs_delayed_node_to_data_ref(ref);
2272
2273                 /*
2274                  * If our ref doesn't match the one we're currently looking at
2275                  * then we have a cross reference.
2276                  */
2277                 if (data_ref->root != root->root_key.objectid ||
2278                     data_ref->objectid != objectid ||
2279                     data_ref->offset != offset) {
2280                         ret = 1;
2281                         break;
2282                 }
2283         }
2284         spin_unlock(&head->lock);
2285         mutex_unlock(&head->mutex);
2286         btrfs_put_transaction(cur_trans);
2287         return ret;
2288 }
2289
2290 static noinline int check_committed_ref(struct btrfs_root *root,
2291                                         struct btrfs_path *path,
2292                                         u64 objectid, u64 offset, u64 bytenr,
2293                                         bool strict)
2294 {
2295         struct btrfs_fs_info *fs_info = root->fs_info;
2296         struct btrfs_root *extent_root = fs_info->extent_root;
2297         struct extent_buffer *leaf;
2298         struct btrfs_extent_data_ref *ref;
2299         struct btrfs_extent_inline_ref *iref;
2300         struct btrfs_extent_item *ei;
2301         struct btrfs_key key;
2302         u32 item_size;
2303         int type;
2304         int ret;
2305
2306         key.objectid = bytenr;
2307         key.offset = (u64)-1;
2308         key.type = BTRFS_EXTENT_ITEM_KEY;
2309
2310         ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2311         if (ret < 0)
2312                 goto out;
2313         BUG_ON(ret == 0); /* Corruption */
2314
2315         ret = -ENOENT;
2316         if (path->slots[0] == 0)
2317                 goto out;
2318
2319         path->slots[0]--;
2320         leaf = path->nodes[0];
2321         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2322
2323         if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2324                 goto out;
2325
2326         ret = 1;
2327         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2328         ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2329
2330         /* If extent item has more than 1 inline ref then it's shared */
2331         if (item_size != sizeof(*ei) +
2332             btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2333                 goto out;
2334
2335         /*
2336          * If extent created before last snapshot => it's shared unless the
2337          * snapshot has been deleted. Use the heuristic if strict is false.
2338          */
2339         if (!strict &&
2340             (btrfs_extent_generation(leaf, ei) <=
2341              btrfs_root_last_snapshot(&root->root_item)))
2342                 goto out;
2343
2344         iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2345
2346         /* If this extent has SHARED_DATA_REF then it's shared */
2347         type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
2348         if (type != BTRFS_EXTENT_DATA_REF_KEY)
2349                 goto out;
2350
2351         ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2352         if (btrfs_extent_refs(leaf, ei) !=
2353             btrfs_extent_data_ref_count(leaf, ref) ||
2354             btrfs_extent_data_ref_root(leaf, ref) !=
2355             root->root_key.objectid ||
2356             btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2357             btrfs_extent_data_ref_offset(leaf, ref) != offset)
2358                 goto out;
2359
2360         ret = 0;
2361 out:
2362         return ret;
2363 }
2364
2365 int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
2366                           u64 bytenr, bool strict)
2367 {
2368         struct btrfs_path *path;
2369         int ret;
2370
2371         path = btrfs_alloc_path();
2372         if (!path)
2373                 return -ENOMEM;
2374
2375         do {
2376                 ret = check_committed_ref(root, path, objectid,
2377                                           offset, bytenr, strict);
2378                 if (ret && ret != -ENOENT)
2379                         goto out;
2380
2381                 ret = check_delayed_ref(root, path, objectid, offset, bytenr);
2382         } while (ret == -EAGAIN);
2383
2384 out:
2385         btrfs_free_path(path);
2386         if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
2387                 WARN_ON(ret > 0);
2388         return ret;
2389 }
2390
2391 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2392                            struct btrfs_root *root,
2393                            struct extent_buffer *buf,
2394                            int full_backref, int inc)
2395 {
2396         struct btrfs_fs_info *fs_info = root->fs_info;
2397         u64 bytenr;
2398         u64 num_bytes;
2399         u64 parent;
2400         u64 ref_root;
2401         u32 nritems;
2402         struct btrfs_key key;
2403         struct btrfs_file_extent_item *fi;
2404         struct btrfs_ref generic_ref = { 0 };
2405         bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
2406         int i;
2407         int action;
2408         int level;
2409         int ret = 0;
2410
2411         if (btrfs_is_testing(fs_info))
2412                 return 0;
2413
2414         ref_root = btrfs_header_owner(buf);
2415         nritems = btrfs_header_nritems(buf);
2416         level = btrfs_header_level(buf);
2417
2418         if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && level == 0)
2419                 return 0;
2420
2421         if (full_backref)
2422                 parent = buf->start;
2423         else
2424                 parent = 0;
2425         if (inc)
2426                 action = BTRFS_ADD_DELAYED_REF;
2427         else
2428                 action = BTRFS_DROP_DELAYED_REF;
2429
2430         for (i = 0; i < nritems; i++) {
2431                 if (level == 0) {
2432                         btrfs_item_key_to_cpu(buf, &key, i);
2433                         if (key.type != BTRFS_EXTENT_DATA_KEY)
2434                                 continue;
2435                         fi = btrfs_item_ptr(buf, i,
2436                                             struct btrfs_file_extent_item);
2437                         if (btrfs_file_extent_type(buf, fi) ==
2438                             BTRFS_FILE_EXTENT_INLINE)
2439                                 continue;
2440                         bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2441                         if (bytenr == 0)
2442                                 continue;
2443
2444                         num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2445                         key.offset -= btrfs_file_extent_offset(buf, fi);
2446                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2447                                                num_bytes, parent);
2448                         generic_ref.real_root = root->root_key.objectid;
2449                         btrfs_init_data_ref(&generic_ref, ref_root, key.objectid,
2450                                             key.offset);
2451                         generic_ref.skip_qgroup = for_reloc;
2452                         if (inc)
2453                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2454                         else
2455                                 ret = btrfs_free_extent(trans, &generic_ref);
2456                         if (ret)
2457                                 goto fail;
2458                 } else {
2459                         bytenr = btrfs_node_blockptr(buf, i);
2460                         num_bytes = fs_info->nodesize;
2461                         btrfs_init_generic_ref(&generic_ref, action, bytenr,
2462                                                num_bytes, parent);
2463                         generic_ref.real_root = root->root_key.objectid;
2464                         btrfs_init_tree_ref(&generic_ref, level - 1, ref_root);
2465                         generic_ref.skip_qgroup = for_reloc;
2466                         if (inc)
2467                                 ret = btrfs_inc_extent_ref(trans, &generic_ref);
2468                         else
2469                                 ret = btrfs_free_extent(trans, &generic_ref);
2470                         if (ret)
2471                                 goto fail;
2472                 }
2473         }
2474         return 0;
2475 fail:
2476         return ret;
2477 }
2478
2479 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2480                   struct extent_buffer *buf, int full_backref)
2481 {
2482         return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2483 }
2484
2485 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2486                   struct extent_buffer *buf, int full_backref)
2487 {
2488         return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2489 }
2490
2491 int btrfs_extent_readonly(struct btrfs_fs_info *fs_info, u64 bytenr)
2492 {
2493         struct btrfs_block_group *block_group;
2494         int readonly = 0;
2495
2496         block_group = btrfs_lookup_block_group(fs_info, bytenr);
2497         if (!block_group || block_group->ro)
2498                 readonly = 1;
2499         if (block_group)
2500                 btrfs_put_block_group(block_group);
2501         return readonly;
2502 }
2503
2504 static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
2505 {
2506         struct btrfs_fs_info *fs_info = root->fs_info;
2507         u64 flags;
2508         u64 ret;
2509
2510         if (data)
2511                 flags = BTRFS_BLOCK_GROUP_DATA;
2512         else if (root == fs_info->chunk_root)
2513                 flags = BTRFS_BLOCK_GROUP_SYSTEM;
2514         else
2515                 flags = BTRFS_BLOCK_GROUP_METADATA;
2516
2517         ret = btrfs_get_alloc_profile(fs_info, flags);
2518         return ret;
2519 }
2520
2521 static u64 first_logical_byte(struct btrfs_fs_info *fs_info, u64 search_start)
2522 {
2523         struct btrfs_block_group *cache;
2524         u64 bytenr;
2525
2526         spin_lock(&fs_info->block_group_cache_lock);
2527         bytenr = fs_info->first_logical_byte;
2528         spin_unlock(&fs_info->block_group_cache_lock);
2529
2530         if (bytenr < (u64)-1)
2531                 return bytenr;
2532
2533         cache = btrfs_lookup_first_block_group(fs_info, search_start);
2534         if (!cache)
2535                 return 0;
2536
2537         bytenr = cache->start;
2538         btrfs_put_block_group(cache);
2539
2540         return bytenr;
2541 }
2542
2543 static int pin_down_extent(struct btrfs_trans_handle *trans,
2544                            struct btrfs_block_group *cache,
2545                            u64 bytenr, u64 num_bytes, int reserved)
2546 {
2547         struct btrfs_fs_info *fs_info = cache->fs_info;
2548
2549         spin_lock(&cache->space_info->lock);
2550         spin_lock(&cache->lock);
2551         cache->pinned += num_bytes;
2552         btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
2553                                              num_bytes);
2554         if (reserved) {
2555                 cache->reserved -= num_bytes;
2556                 cache->space_info->bytes_reserved -= num_bytes;
2557         }
2558         spin_unlock(&cache->lock);
2559         spin_unlock(&cache->space_info->lock);
2560
2561         __btrfs_mod_total_bytes_pinned(cache->space_info, num_bytes);
2562         set_extent_dirty(&trans->transaction->pinned_extents, bytenr,
2563                          bytenr + num_bytes - 1, GFP_NOFS | __GFP_NOFAIL);
2564         return 0;
2565 }
2566
2567 int btrfs_pin_extent(struct btrfs_trans_handle *trans,
2568                      u64 bytenr, u64 num_bytes, int reserved)
2569 {
2570         struct btrfs_block_group *cache;
2571
2572         cache = btrfs_lookup_block_group(trans->fs_info, bytenr);
2573         BUG_ON(!cache); /* Logic error */
2574
2575         pin_down_extent(trans, cache, bytenr, num_bytes, reserved);
2576
2577         btrfs_put_block_group(cache);
2578         return 0;
2579 }
2580
2581 /*
2582  * this function must be called within transaction
2583  */
2584 int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans,
2585                                     u64 bytenr, u64 num_bytes)
2586 {
2587         struct btrfs_block_group *cache;
2588         int ret;
2589
2590         btrfs_add_excluded_extent(trans->fs_info, bytenr, num_bytes);
2591
2592         cache = btrfs_lookup_block_group(trans->fs_info, bytenr);
2593         if (!cache)
2594                 return -EINVAL;
2595
2596         /*
2597          * pull in the free space cache (if any) so that our pin
2598          * removes the free space from the cache.  We have load_only set
2599          * to one because the slow code to read in the free extents does check
2600          * the pinned extents.
2601          */
2602         btrfs_cache_block_group(cache, 1);
2603
2604         pin_down_extent(trans, cache, bytenr, num_bytes, 0);
2605
2606         /* remove us from the free space cache (if we're there at all) */
2607         ret = btrfs_remove_free_space(cache, bytenr, num_bytes);
2608         btrfs_put_block_group(cache);
2609         return ret;
2610 }
2611
2612 static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
2613                                    u64 start, u64 num_bytes)
2614 {
2615         int ret;
2616         struct btrfs_block_group *block_group;
2617         struct btrfs_caching_control *caching_ctl;
2618
2619         block_group = btrfs_lookup_block_group(fs_info, start);
2620         if (!block_group)
2621                 return -EINVAL;
2622
2623         btrfs_cache_block_group(block_group, 0);
2624         caching_ctl = btrfs_get_caching_control(block_group);
2625
2626         if (!caching_ctl) {
2627                 /* Logic error */
2628                 BUG_ON(!btrfs_block_group_done(block_group));
2629                 ret = btrfs_remove_free_space(block_group, start, num_bytes);
2630         } else {
2631                 mutex_lock(&caching_ctl->mutex);
2632
2633                 if (start >= caching_ctl->progress) {
2634                         ret = btrfs_add_excluded_extent(fs_info, start,
2635                                                         num_bytes);
2636                 } else if (start + num_bytes <= caching_ctl->progress) {
2637                         ret = btrfs_remove_free_space(block_group,
2638                                                       start, num_bytes);
2639                 } else {
2640                         num_bytes = caching_ctl->progress - start;
2641                         ret = btrfs_remove_free_space(block_group,
2642                                                       start, num_bytes);
2643                         if (ret)
2644                                 goto out_lock;
2645
2646                         num_bytes = (start + num_bytes) -
2647                                 caching_ctl->progress;
2648                         start = caching_ctl->progress;
2649                         ret = btrfs_add_excluded_extent(fs_info, start,
2650                                                         num_bytes);
2651                 }
2652 out_lock:
2653                 mutex_unlock(&caching_ctl->mutex);
2654                 btrfs_put_caching_control(caching_ctl);
2655         }
2656         btrfs_put_block_group(block_group);
2657         return ret;
2658 }
2659
2660 int btrfs_exclude_logged_extents(struct extent_buffer *eb)
2661 {
2662         struct btrfs_fs_info *fs_info = eb->fs_info;
2663         struct btrfs_file_extent_item *item;
2664         struct btrfs_key key;
2665         int found_type;
2666         int i;
2667         int ret = 0;
2668
2669         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
2670                 return 0;
2671
2672         for (i = 0; i < btrfs_header_nritems(eb); i++) {
2673                 btrfs_item_key_to_cpu(eb, &key, i);
2674                 if (key.type != BTRFS_EXTENT_DATA_KEY)
2675                         continue;
2676                 item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
2677                 found_type = btrfs_file_extent_type(eb, item);
2678                 if (found_type == BTRFS_FILE_EXTENT_INLINE)
2679                         continue;
2680                 if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
2681                         continue;
2682                 key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
2683                 key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
2684                 ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
2685                 if (ret)
2686                         break;
2687         }
2688
2689         return ret;
2690 }
2691
2692 static void
2693 btrfs_inc_block_group_reservations(struct btrfs_block_group *bg)
2694 {
2695         atomic_inc(&bg->reservations);
2696 }
2697
2698 /*
2699  * Returns the free cluster for the given space info and sets empty_cluster to
2700  * what it should be based on the mount options.
2701  */
2702 static struct btrfs_free_cluster *
2703 fetch_cluster_info(struct btrfs_fs_info *fs_info,
2704                    struct btrfs_space_info *space_info, u64 *empty_cluster)
2705 {
2706         struct btrfs_free_cluster *ret = NULL;
2707
2708         *empty_cluster = 0;
2709         if (btrfs_mixed_space_info(space_info))
2710                 return ret;
2711
2712         if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
2713                 ret = &fs_info->meta_alloc_cluster;
2714                 if (btrfs_test_opt(fs_info, SSD))
2715                         *empty_cluster = SZ_2M;
2716                 else
2717                         *empty_cluster = SZ_64K;
2718         } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
2719                    btrfs_test_opt(fs_info, SSD_SPREAD)) {
2720                 *empty_cluster = SZ_2M;
2721                 ret = &fs_info->data_alloc_cluster;
2722         }
2723
2724         return ret;
2725 }
2726
2727 static int unpin_extent_range(struct btrfs_fs_info *fs_info,
2728                               u64 start, u64 end,
2729                               const bool return_free_space)
2730 {
2731         struct btrfs_block_group *cache = NULL;
2732         struct btrfs_space_info *space_info;
2733         struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
2734         struct btrfs_free_cluster *cluster = NULL;
2735         u64 len;
2736         u64 total_unpinned = 0;
2737         u64 empty_cluster = 0;
2738         bool readonly;
2739
2740         while (start <= end) {
2741                 readonly = false;
2742                 if (!cache ||
2743                     start >= cache->start + cache->length) {
2744                         if (cache)
2745                                 btrfs_put_block_group(cache);
2746                         total_unpinned = 0;
2747                         cache = btrfs_lookup_block_group(fs_info, start);
2748                         BUG_ON(!cache); /* Logic error */
2749
2750                         cluster = fetch_cluster_info(fs_info,
2751                                                      cache->space_info,
2752                                                      &empty_cluster);
2753                         empty_cluster <<= 1;
2754                 }
2755
2756                 len = cache->start + cache->length - start;
2757                 len = min(len, end + 1 - start);
2758
2759                 if (start < cache->last_byte_to_unpin && return_free_space) {
2760                         u64 add_len = min(len, cache->last_byte_to_unpin - start);
2761
2762                         btrfs_add_free_space(cache, start, add_len);
2763                 }
2764
2765                 start += len;
2766                 total_unpinned += len;
2767                 space_info = cache->space_info;
2768
2769                 /*
2770                  * If this space cluster has been marked as fragmented and we've
2771                  * unpinned enough in this block group to potentially allow a
2772                  * cluster to be created inside of it go ahead and clear the
2773                  * fragmented check.
2774                  */
2775                 if (cluster && cluster->fragmented &&
2776                     total_unpinned > empty_cluster) {
2777                         spin_lock(&cluster->lock);
2778                         cluster->fragmented = 0;
2779                         spin_unlock(&cluster->lock);
2780                 }
2781
2782                 spin_lock(&space_info->lock);
2783                 spin_lock(&cache->lock);
2784                 cache->pinned -= len;
2785                 btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
2786                 space_info->max_extent_size = 0;
2787                 __btrfs_mod_total_bytes_pinned(space_info, -len);
2788                 if (cache->ro) {
2789                         space_info->bytes_readonly += len;
2790                         readonly = true;
2791                 }
2792                 spin_unlock(&cache->lock);
2793                 if (!readonly && return_free_space &&
2794                     global_rsv->space_info == space_info) {
2795                         u64 to_add = len;
2796
2797                         spin_lock(&global_rsv->lock);
2798                         if (!global_rsv->full) {
2799                                 to_add = min(len, global_rsv->size -
2800                                              global_rsv->reserved);
2801                                 global_rsv->reserved += to_add;
2802                                 btrfs_space_info_update_bytes_may_use(fs_info,
2803                                                 space_info, to_add);
2804                                 if (global_rsv->reserved >= global_rsv->size)
2805                                         global_rsv->full = 1;
2806                                 len -= to_add;
2807                         }
2808                         spin_unlock(&global_rsv->lock);
2809                 }
2810                 /* Add to any tickets we may have */
2811                 if (!readonly && return_free_space && len)
2812                         btrfs_try_granting_tickets(fs_info, space_info);
2813                 spin_unlock(&space_info->lock);
2814         }
2815
2816         if (cache)
2817                 btrfs_put_block_group(cache);
2818         return 0;
2819 }
2820
2821 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
2822 {
2823         struct btrfs_fs_info *fs_info = trans->fs_info;
2824         struct btrfs_block_group *block_group, *tmp;
2825         struct list_head *deleted_bgs;
2826         struct extent_io_tree *unpin;
2827         u64 start;
2828         u64 end;
2829         int ret;
2830
2831         unpin = &trans->transaction->pinned_extents;
2832
2833         while (!TRANS_ABORTED(trans)) {
2834                 struct extent_state *cached_state = NULL;
2835
2836                 mutex_lock(&fs_info->unused_bg_unpin_mutex);
2837                 ret = find_first_extent_bit(unpin, 0, &start, &end,
2838                                             EXTENT_DIRTY, &cached_state);
2839                 if (ret) {
2840                         mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2841                         break;
2842                 }
2843                 if (test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags))
2844                         clear_extent_bits(&fs_info->excluded_extents, start,
2845                                           end, EXTENT_UPTODATE);
2846
2847                 if (btrfs_test_opt(fs_info, DISCARD_SYNC))
2848                         ret = btrfs_discard_extent(fs_info, start,
2849                                                    end + 1 - start, NULL);
2850
2851                 clear_extent_dirty(unpin, start, end, &cached_state);
2852                 unpin_extent_range(fs_info, start, end, true);
2853                 mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2854                 free_extent_state(cached_state);
2855                 cond_resched();
2856         }
2857
2858         if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
2859                 btrfs_discard_calc_delay(&fs_info->discard_ctl);
2860                 btrfs_discard_schedule_work(&fs_info->discard_ctl, true);
2861         }
2862
2863         /*
2864          * Transaction is finished.  We don't need the lock anymore.  We
2865          * do need to clean up the block groups in case of a transaction
2866          * abort.
2867          */
2868         deleted_bgs = &trans->transaction->deleted_bgs;
2869         list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
2870                 u64 trimmed = 0;
2871
2872                 ret = -EROFS;
2873                 if (!TRANS_ABORTED(trans))
2874                         ret = btrfs_discard_extent(fs_info,
2875                                                    block_group->start,
2876                                                    block_group->length,
2877                                                    &trimmed);
2878
2879                 list_del_init(&block_group->bg_list);
2880                 btrfs_unfreeze_block_group(block_group);
2881                 btrfs_put_block_group(block_group);
2882
2883                 if (ret) {
2884                         const char *errstr = btrfs_decode_error(ret);
2885                         btrfs_warn(fs_info,
2886                            "discard failed while removing blockgroup: errno=%d %s",
2887                                    ret, errstr);
2888                 }
2889         }
2890
2891         return 0;
2892 }
2893
2894 /*
2895  * Drop one or more refs of @node.
2896  *
2897  * 1. Locate the extent refs.
2898  *    It's either inline in EXTENT/METADATA_ITEM or in keyed SHARED_* item.
2899  *    Locate it, then reduce the refs number or remove the ref line completely.
2900  *
2901  * 2. Update the refs count in EXTENT/METADATA_ITEM
2902  *
2903  * Inline backref case:
2904  *
2905  * in extent tree we have:
2906  *
2907  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
2908  *              refs 2 gen 6 flags DATA
2909  *              extent data backref root FS_TREE objectid 258 offset 0 count 1
2910  *              extent data backref root FS_TREE objectid 257 offset 0 count 1
2911  *
2912  * This function gets called with:
2913  *
2914  *    node->bytenr = 13631488
2915  *    node->num_bytes = 1048576
2916  *    root_objectid = FS_TREE
2917  *    owner_objectid = 257
2918  *    owner_offset = 0
2919  *    refs_to_drop = 1
2920  *
2921  * Then we should get some like:
2922  *
2923  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
2924  *              refs 1 gen 6 flags DATA
2925  *              extent data backref root FS_TREE objectid 258 offset 0 count 1
2926  *
2927  * Keyed backref case:
2928  *
2929  * in extent tree we have:
2930  *
2931  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
2932  *              refs 754 gen 6 flags DATA
2933  *      [...]
2934  *      item 2 key (13631488 EXTENT_DATA_REF <HASH>) itemoff 3915 itemsize 28
2935  *              extent data backref root FS_TREE objectid 866 offset 0 count 1
2936  *
2937  * This function get called with:
2938  *
2939  *    node->bytenr = 13631488
2940  *    node->num_bytes = 1048576
2941  *    root_objectid = FS_TREE
2942  *    owner_objectid = 866
2943  *    owner_offset = 0
2944  *    refs_to_drop = 1
2945  *
2946  * Then we should get some like:
2947  *
2948  *      item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
2949  *              refs 753 gen 6 flags DATA
2950  *
2951  * And that (13631488 EXTENT_DATA_REF <HASH>) gets removed.
2952  */
2953 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
2954                                struct btrfs_delayed_ref_node *node, u64 parent,
2955                                u64 root_objectid, u64 owner_objectid,
2956                                u64 owner_offset, int refs_to_drop,
2957                                struct btrfs_delayed_extent_op *extent_op)
2958 {
2959         struct btrfs_fs_info *info = trans->fs_info;
2960         struct btrfs_key key;
2961         struct btrfs_path *path;
2962         struct btrfs_root *extent_root = info->extent_root;
2963         struct extent_buffer *leaf;
2964         struct btrfs_extent_item *ei;
2965         struct btrfs_extent_inline_ref *iref;
2966         int ret;
2967         int is_data;
2968         int extent_slot = 0;
2969         int found_extent = 0;
2970         int num_to_del = 1;
2971         u32 item_size;
2972         u64 refs;
2973         u64 bytenr = node->bytenr;
2974         u64 num_bytes = node->num_bytes;
2975         int last_ref = 0;
2976         bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
2977
2978         path = btrfs_alloc_path();
2979         if (!path)
2980                 return -ENOMEM;
2981
2982         path->leave_spinning = 1;
2983
2984         is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
2985
2986         if (!is_data && refs_to_drop != 1) {
2987                 btrfs_crit(info,
2988 "invalid refs_to_drop, dropping more than 1 refs for tree block %llu refs_to_drop %u",
2989                            node->bytenr, refs_to_drop);
2990                 ret = -EINVAL;
2991                 btrfs_abort_transaction(trans, ret);
2992                 goto out;
2993         }
2994
2995         if (is_data)
2996                 skinny_metadata = false;
2997
2998         ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
2999                                     parent, root_objectid, owner_objectid,
3000                                     owner_offset);
3001         if (ret == 0) {
3002                 /*
3003                  * Either the inline backref or the SHARED_DATA_REF/
3004                  * SHARED_BLOCK_REF is found
3005                  *
3006                  * Here is a quick path to locate EXTENT/METADATA_ITEM.
3007                  * It's possible the EXTENT/METADATA_ITEM is near current slot.
3008                  */
3009                 extent_slot = path->slots[0];
3010                 while (extent_slot >= 0) {
3011                         btrfs_item_key_to_cpu(path->nodes[0], &key,
3012                                               extent_slot);
3013                         if (key.objectid != bytenr)
3014                                 break;
3015                         if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3016                             key.offset == num_bytes) {
3017                                 found_extent = 1;
3018                                 break;
3019                         }
3020                         if (key.type == BTRFS_METADATA_ITEM_KEY &&
3021                             key.offset == owner_objectid) {
3022                                 found_extent = 1;
3023                                 break;
3024                         }
3025
3026                         /* Quick path didn't find the EXTEMT/METADATA_ITEM */
3027                         if (path->slots[0] - extent_slot > 5)
3028                                 break;
3029                         extent_slot--;
3030                 }
3031
3032                 if (!found_extent) {
3033                         if (iref) {
3034                                 btrfs_crit(info,
3035 "invalid iref, no EXTENT/METADATA_ITEM found but has inline extent ref");
3036                                 btrfs_abort_transaction(trans, -EUCLEAN);
3037                                 goto err_dump;
3038                         }
3039                         /* Must be SHARED_* item, remove the backref first */
3040                         ret = remove_extent_backref(trans, path, NULL,
3041                                                     refs_to_drop,
3042                                                     is_data, &last_ref);
3043                         if (ret) {
3044                                 btrfs_abort_transaction(trans, ret);
3045                                 goto out;
3046                         }
3047                         btrfs_release_path(path);
3048                         path->leave_spinning = 1;
3049
3050                         /* Slow path to locate EXTENT/METADATA_ITEM */
3051                         key.objectid = bytenr;
3052                         key.type = BTRFS_EXTENT_ITEM_KEY;
3053                         key.offset = num_bytes;
3054
3055                         if (!is_data && skinny_metadata) {
3056                                 key.type = BTRFS_METADATA_ITEM_KEY;
3057                                 key.offset = owner_objectid;
3058                         }
3059
3060                         ret = btrfs_search_slot(trans, extent_root,
3061                                                 &key, path, -1, 1);
3062                         if (ret > 0 && skinny_metadata && path->slots[0]) {
3063                                 /*
3064                                  * Couldn't find our skinny metadata item,
3065                                  * see if we have ye olde extent item.
3066                                  */
3067                                 path->slots[0]--;
3068                                 btrfs_item_key_to_cpu(path->nodes[0], &key,
3069                                                       path->slots[0]);
3070                                 if (key.objectid == bytenr &&
3071                                     key.type == BTRFS_EXTENT_ITEM_KEY &&
3072                                     key.offset == num_bytes)
3073                                         ret = 0;
3074                         }
3075
3076                         if (ret > 0 && skinny_metadata) {
3077                                 skinny_metadata = false;
3078                                 key.objectid = bytenr;
3079                                 key.type = BTRFS_EXTENT_ITEM_KEY;
3080                                 key.offset = num_bytes;
3081                                 btrfs_release_path(path);
3082                                 ret = btrfs_search_slot(trans, extent_root,
3083                                                         &key, path, -1, 1);
3084                         }
3085
3086                         if (ret) {
3087                                 btrfs_err(info,
3088                                           "umm, got %d back from search, was looking for %llu",
3089                                           ret, bytenr);
3090                                 if (ret > 0)
3091                                         btrfs_print_leaf(path->nodes[0]);
3092                         }
3093                         if (ret < 0) {
3094                                 btrfs_abort_transaction(trans, ret);
3095                                 goto out;
3096                         }
3097                         extent_slot = path->slots[0];
3098                 }
3099         } else if (WARN_ON(ret == -ENOENT)) {
3100                 btrfs_print_leaf(path->nodes[0]);
3101                 btrfs_err(info,
3102                         "unable to find ref byte nr %llu parent %llu root %llu  owner %llu offset %llu",
3103                         bytenr, parent, root_objectid, owner_objectid,
3104                         owner_offset);
3105                 btrfs_abort_transaction(trans, ret);
3106                 goto out;
3107         } else {
3108                 btrfs_abort_transaction(trans, ret);
3109                 goto out;
3110         }
3111
3112         leaf = path->nodes[0];
3113         item_size = btrfs_item_size_nr(leaf, extent_slot);
3114         if (unlikely(item_size < sizeof(*ei))) {
3115                 ret = -EINVAL;
3116                 btrfs_print_v0_err(info);
3117                 btrfs_abort_transaction(trans, ret);
3118                 goto out;
3119         }
3120         ei = btrfs_item_ptr(leaf, extent_slot,
3121                             struct btrfs_extent_item);
3122         if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
3123             key.type == BTRFS_EXTENT_ITEM_KEY) {
3124                 struct btrfs_tree_block_info *bi;
3125                 if (item_size < sizeof(*ei) + sizeof(*bi)) {
3126                         btrfs_crit(info,
3127 "invalid extent item size for key (%llu, %u, %llu) owner %llu, has %u expect >= %zu",
3128                                    key.objectid, key.type, key.offset,
3129                                    owner_objectid, item_size,
3130                                    sizeof(*ei) + sizeof(*bi));
3131                         btrfs_abort_transaction(trans, -EUCLEAN);
3132                         goto err_dump;
3133                 }
3134                 bi = (struct btrfs_tree_block_info *)(ei + 1);
3135                 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3136         }
3137
3138         refs = btrfs_extent_refs(leaf, ei);
3139         if (refs < refs_to_drop) {
3140                 btrfs_crit(info,
3141                 "trying to drop %d refs but we only have %llu for bytenr %llu",
3142                           refs_to_drop, refs, bytenr);
3143                 btrfs_abort_transaction(trans, -EUCLEAN);
3144                 goto err_dump;
3145         }
3146         refs -= refs_to_drop;
3147
3148         if (refs > 0) {
3149                 if (extent_op)
3150                         __run_delayed_extent_op(extent_op, leaf, ei);
3151                 /*
3152                  * In the case of inline back ref, reference count will
3153                  * be updated by remove_extent_backref
3154                  */
3155                 if (iref) {
3156                         if (!found_extent) {
3157                                 btrfs_crit(info,
3158 "invalid iref, got inlined extent ref but no EXTENT/METADATA_ITEM found");
3159                                 btrfs_abort_transaction(trans, -EUCLEAN);
3160                                 goto err_dump;
3161                         }
3162                 } else {
3163                         btrfs_set_extent_refs(leaf, ei, refs);
3164                         btrfs_mark_buffer_dirty(leaf);
3165                 }
3166                 if (found_extent) {
3167                         ret = remove_extent_backref(trans, path, iref,
3168                                                     refs_to_drop, is_data,
3169                                                     &last_ref);
3170                         if (ret) {
3171                                 btrfs_abort_transaction(trans, ret);
3172                                 goto out;
3173                         }
3174                 }
3175         } else {
3176                 /* In this branch refs == 1 */
3177                 if (found_extent) {
3178                         if (is_data && refs_to_drop !=
3179                             extent_data_ref_count(path, iref)) {
3180                                 btrfs_crit(info,
3181                 "invalid refs_to_drop, current refs %u refs_to_drop %u",
3182                                            extent_data_ref_count(path, iref),
3183                                            refs_to_drop);
3184                                 btrfs_abort_transaction(trans, -EUCLEAN);
3185                                 goto err_dump;
3186                         }
3187                         if (iref) {
3188                                 if (path->slots[0] != extent_slot) {
3189                                         btrfs_crit(info,
3190 "invalid iref, extent item key (%llu %u %llu) doesn't have wanted iref",
3191                                                    key.objectid, key.type,
3192                                                    key.offset);
3193                                         btrfs_abort_transaction(trans, -EUCLEAN);
3194                                         goto err_dump;
3195                                 }
3196                         } else {
3197                                 /*
3198                                  * No inline ref, we must be at SHARED_* item,
3199                                  * And it's single ref, it must be:
3200                                  * |    extent_slot       ||extent_slot + 1|
3201                                  * [ EXTENT/METADATA_ITEM ][ SHARED_* ITEM ]
3202                                  */
3203                                 if (path->slots[0] != extent_slot + 1) {
3204                                         btrfs_crit(info,
3205         "invalid SHARED_* item, previous item is not EXTENT/METADATA_ITEM");
3206                                         btrfs_abort_transaction(trans, -EUCLEAN);
3207                                         goto err_dump;
3208                                 }
3209                                 path->slots[0] = extent_slot;
3210                                 num_to_del = 2;
3211                         }
3212                 }
3213
3214                 last_ref = 1;
3215                 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3216                                       num_to_del);
3217                 if (ret) {
3218                         btrfs_abort_transaction(trans, ret);
3219                         goto out;
3220                 }
3221                 btrfs_release_path(path);
3222
3223                 if (is_data) {
3224                         ret = btrfs_del_csums(trans, info->csum_root, bytenr,
3225                                               num_bytes);
3226                         if (ret) {
3227                                 btrfs_abort_transaction(trans, ret);
3228                                 goto out;
3229                         }
3230                 }
3231
3232                 ret = add_to_free_space_tree(trans, bytenr, num_bytes);
3233                 if (ret) {
3234                         btrfs_abort_transaction(trans, ret);
3235                         goto out;
3236                 }
3237
3238                 ret = btrfs_update_block_group(trans, bytenr, num_bytes, 0);
3239                 if (ret) {
3240                         btrfs_abort_transaction(trans, ret);
3241                         goto out;
3242                 }
3243         }
3244         btrfs_release_path(path);
3245
3246 out:
3247         btrfs_free_path(path);
3248         return ret;
3249 err_dump:
3250         /*
3251          * Leaf dump can take up a lot of log buffer, so we only do full leaf
3252          * dump for debug build.
3253          */
3254         if (IS_ENABLED(CONFIG_BTRFS_DEBUG)) {
3255                 btrfs_crit(info, "path->slots[0]=%d extent_slot=%d",
3256                            path->slots[0], extent_slot);
3257                 btrfs_print_leaf(path->nodes[0]);
3258         }
3259
3260         btrfs_free_path(path);
3261         return -EUCLEAN;
3262 }
3263
3264 /*
3265  * when we free an block, it is possible (and likely) that we free the last
3266  * delayed ref for that extent as well.  This searches the delayed ref tree for
3267  * a given extent, and if there are no other delayed refs to be processed, it
3268  * removes it from the tree.
3269  */
3270 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3271                                       u64 bytenr)
3272 {
3273         struct btrfs_delayed_ref_head *head;
3274         struct btrfs_delayed_ref_root *delayed_refs;
3275         int ret = 0;
3276
3277         delayed_refs = &trans->transaction->delayed_refs;
3278         spin_lock(&delayed_refs->lock);
3279         head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
3280         if (!head)
3281                 goto out_delayed_unlock;
3282
3283         spin_lock(&head->lock);
3284         if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
3285                 goto out;
3286
3287         if (cleanup_extent_op(head) != NULL)
3288                 goto out;
3289
3290         /*
3291          * waiting for the lock here would deadlock.  If someone else has it
3292          * locked they are already in the process of dropping it anyway
3293          */
3294         if (!mutex_trylock(&head->mutex))
3295                 goto out;
3296
3297         btrfs_delete_ref_head(delayed_refs, head);
3298         head->processing = 0;
3299
3300         spin_unlock(&head->lock);
3301         spin_unlock(&delayed_refs->lock);
3302
3303         BUG_ON(head->extent_op);
3304         if (head->must_insert_reserved)
3305                 ret = 1;
3306
3307         btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
3308         mutex_unlock(&head->mutex);
3309         btrfs_put_delayed_ref_head(head);
3310         return ret;
3311 out:
3312         spin_unlock(&head->lock);
3313
3314 out_delayed_unlock:
3315         spin_unlock(&delayed_refs->lock);
3316         return 0;
3317 }
3318
3319 void btrfs_free_tree_block(struct btrfs_trans_handle *trans,
3320                            struct btrfs_root *root,
3321                            struct extent_buffer *buf,
3322                            u64 parent, int last_ref)
3323 {
3324         struct btrfs_fs_info *fs_info = root->fs_info;
3325         struct btrfs_ref generic_ref = { 0 };
3326         int ret;
3327
3328         btrfs_init_generic_ref(&generic_ref, BTRFS_DROP_DELAYED_REF,
3329                                buf->start, buf->len, parent);
3330         btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf),
3331                             root->root_key.objectid);
3332
3333         if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3334                 btrfs_ref_tree_mod(fs_info, &generic_ref);
3335                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL);
3336                 BUG_ON(ret); /* -ENOMEM */
3337         }
3338
3339         if (last_ref && btrfs_header_generation(buf) == trans->transid) {
3340                 struct btrfs_block_group *cache;
3341
3342                 if (root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID) {
3343                         ret = check_ref_cleanup(trans, buf->start);
3344                         if (!ret)
3345                                 goto out;
3346                 }
3347
3348                 cache = btrfs_lookup_block_group(fs_info, buf->start);
3349
3350                 if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3351                         pin_down_extent(trans, cache, buf->start, buf->len, 1);
3352                         btrfs_put_block_group(cache);
3353                         goto out;
3354                 }
3355
3356                 WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
3357
3358                 btrfs_add_free_space(cache, buf->start, buf->len);
3359                 btrfs_free_reserved_bytes(cache, buf->len, 0);
3360                 btrfs_put_block_group(cache);
3361                 trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
3362         }
3363 out:
3364         if (last_ref) {
3365                 /*
3366                  * Deleting the buffer, clear the corrupt flag since it doesn't
3367                  * matter anymore.
3368                  */
3369                 clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
3370         }
3371 }
3372
3373 /* Can return -ENOMEM */
3374 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
3375 {
3376         struct btrfs_fs_info *fs_info = trans->fs_info;
3377         int ret;
3378
3379         if (btrfs_is_testing(fs_info))
3380                 return 0;
3381
3382         /*
3383          * tree log blocks never actually go into the extent allocation
3384          * tree, just update pinning info and exit early.
3385          */
3386         if ((ref->type == BTRFS_REF_METADATA &&
3387              ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3388             (ref->type == BTRFS_REF_DATA &&
3389              ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)) {
3390                 /* unlocks the pinned mutex */
3391                 btrfs_pin_extent(trans, ref->bytenr, ref->len, 1);
3392                 ret = 0;
3393         } else if (ref->type == BTRFS_REF_METADATA) {
3394                 ret = btrfs_add_delayed_tree_ref(trans, ref, NULL);
3395         } else {
3396                 ret = btrfs_add_delayed_data_ref(trans, ref, 0);
3397         }
3398
3399         if (!((ref->type == BTRFS_REF_METADATA &&
3400                ref->tree_ref.root == BTRFS_TREE_LOG_OBJECTID) ||
3401               (ref->type == BTRFS_REF_DATA &&
3402                ref->data_ref.ref_root == BTRFS_TREE_LOG_OBJECTID)))
3403                 btrfs_ref_tree_mod(fs_info, ref);
3404
3405         return ret;
3406 }
3407
3408 enum btrfs_loop_type {
3409         LOOP_CACHING_NOWAIT,
3410         LOOP_CACHING_WAIT,
3411         LOOP_ALLOC_CHUNK,
3412         LOOP_NO_EMPTY_SIZE,
3413 };
3414
3415 static inline void
3416 btrfs_lock_block_group(struct btrfs_block_group *cache,
3417                        int delalloc)
3418 {
3419         if (delalloc)
3420                 down_read(&cache->data_rwsem);
3421 }
3422
3423 static inline void btrfs_grab_block_group(struct btrfs_block_group *cache,
3424                        int delalloc)
3425 {
3426         btrfs_get_block_group(cache);
3427         if (delalloc)
3428                 down_read(&cache->data_rwsem);
3429 }
3430
3431 static struct btrfs_block_group *btrfs_lock_cluster(
3432                    struct btrfs_block_group *block_group,
3433                    struct btrfs_free_cluster *cluster,
3434                    int delalloc)
3435         __acquires(&cluster->refill_lock)
3436 {
3437         struct btrfs_block_group *used_bg = NULL;
3438
3439         spin_lock(&cluster->refill_lock);
3440         while (1) {
3441                 used_bg = cluster->block_group;
3442                 if (!used_bg)
3443                         return NULL;
3444
3445                 if (used_bg == block_group)
3446                         return used_bg;
3447
3448                 btrfs_get_block_group(used_bg);
3449
3450                 if (!delalloc)
3451                         return used_bg;
3452
3453                 if (down_read_trylock(&used_bg->data_rwsem))
3454                         return used_bg;
3455
3456                 spin_unlock(&cluster->refill_lock);
3457
3458                 /* We should only have one-level nested. */
3459                 down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
3460
3461                 spin_lock(&cluster->refill_lock);
3462                 if (used_bg == cluster->block_group)
3463                         return used_bg;
3464
3465                 up_read(&used_bg->data_rwsem);
3466                 btrfs_put_block_group(used_bg);
3467         }
3468 }
3469
3470 static inline void
3471 btrfs_release_block_group(struct btrfs_block_group *cache,
3472                          int delalloc)
3473 {
3474         if (delalloc)
3475                 up_read(&cache->data_rwsem);
3476         btrfs_put_block_group(cache);
3477 }
3478
3479 enum btrfs_extent_allocation_policy {
3480         BTRFS_EXTENT_ALLOC_CLUSTERED,
3481 };
3482
3483 /*
3484  * Structure used internally for find_free_extent() function.  Wraps needed
3485  * parameters.
3486  */
3487 struct find_free_extent_ctl {
3488         /* Basic allocation info */
3489         u64 num_bytes;
3490         u64 empty_size;
3491         u64 flags;
3492         int delalloc;
3493
3494         /* Where to start the search inside the bg */
3495         u64 search_start;
3496
3497         /* For clustered allocation */
3498         u64 empty_cluster;
3499         struct btrfs_free_cluster *last_ptr;
3500         bool use_cluster;
3501
3502         bool have_caching_bg;
3503         bool orig_have_caching_bg;
3504
3505         /* RAID index, converted from flags */
3506         int index;
3507
3508         /*
3509          * Current loop number, check find_free_extent_update_loop() for details
3510          */
3511         int loop;
3512
3513         /*
3514          * Whether we're refilling a cluster, if true we need to re-search
3515          * current block group but don't try to refill the cluster again.
3516          */
3517         bool retry_clustered;
3518
3519         /*
3520          * Whether we're updating free space cache, if true we need to re-search
3521          * current block group but don't try updating free space cache again.
3522          */
3523         bool retry_unclustered;
3524
3525         /* If current block group is cached */
3526         int cached;
3527
3528         /* Max contiguous hole found */
3529         u64 max_extent_size;
3530
3531         /* Total free space from free space cache, not always contiguous */
3532         u64 total_free_space;
3533
3534         /* Found result */
3535         u64 found_offset;
3536
3537         /* Hint where to start looking for an empty space */
3538         u64 hint_byte;
3539
3540         /* Allocation policy */
3541         enum btrfs_extent_allocation_policy policy;
3542 };
3543
3544
3545 /*
3546  * Helper function for find_free_extent().
3547  *
3548  * Return -ENOENT to inform caller that we need fallback to unclustered mode.
3549  * Return -EAGAIN to inform caller that we need to re-search this block group
3550  * Return >0 to inform caller that we find nothing
3551  * Return 0 means we have found a location and set ffe_ctl->found_offset.
3552  */
3553 static int find_free_extent_clustered(struct btrfs_block_group *bg,
3554                                       struct find_free_extent_ctl *ffe_ctl,
3555                                       struct btrfs_block_group **cluster_bg_ret)
3556 {
3557         struct btrfs_block_group *cluster_bg;
3558         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3559         u64 aligned_cluster;
3560         u64 offset;
3561         int ret;
3562
3563         cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
3564         if (!cluster_bg)
3565                 goto refill_cluster;
3566         if (cluster_bg != bg && (cluster_bg->ro ||
3567             !block_group_bits(cluster_bg, ffe_ctl->flags)))
3568                 goto release_cluster;
3569
3570         offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
3571                         ffe_ctl->num_bytes, cluster_bg->start,
3572                         &ffe_ctl->max_extent_size);
3573         if (offset) {
3574                 /* We have a block, we're done */
3575                 spin_unlock(&last_ptr->refill_lock);
3576                 trace_btrfs_reserve_extent_cluster(cluster_bg,
3577                                 ffe_ctl->search_start, ffe_ctl->num_bytes);
3578                 *cluster_bg_ret = cluster_bg;
3579                 ffe_ctl->found_offset = offset;
3580                 return 0;
3581         }
3582         WARN_ON(last_ptr->block_group != cluster_bg);
3583
3584 release_cluster:
3585         /*
3586          * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
3587          * lets just skip it and let the allocator find whatever block it can
3588          * find. If we reach this point, we will have tried the cluster
3589          * allocator plenty of times and not have found anything, so we are
3590          * likely way too fragmented for the clustering stuff to find anything.
3591          *
3592          * However, if the cluster is taken from the current block group,
3593          * release the cluster first, so that we stand a better chance of
3594          * succeeding in the unclustered allocation.
3595          */
3596         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
3597                 spin_unlock(&last_ptr->refill_lock);
3598                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3599                 return -ENOENT;
3600         }
3601
3602         /* This cluster didn't work out, free it and start over */
3603         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3604
3605         if (cluster_bg != bg)
3606                 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3607
3608 refill_cluster:
3609         if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
3610                 spin_unlock(&last_ptr->refill_lock);
3611                 return -ENOENT;
3612         }
3613
3614         aligned_cluster = max_t(u64,
3615                         ffe_ctl->empty_cluster + ffe_ctl->empty_size,
3616                         bg->full_stripe_len);
3617         ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
3618                         ffe_ctl->num_bytes, aligned_cluster);
3619         if (ret == 0) {
3620                 /* Now pull our allocation out of this cluster */
3621                 offset = btrfs_alloc_from_cluster(bg, last_ptr,
3622                                 ffe_ctl->num_bytes, ffe_ctl->search_start,
3623                                 &ffe_ctl->max_extent_size);
3624                 if (offset) {
3625                         /* We found one, proceed */
3626                         spin_unlock(&last_ptr->refill_lock);
3627                         trace_btrfs_reserve_extent_cluster(bg,
3628                                         ffe_ctl->search_start,
3629                                         ffe_ctl->num_bytes);
3630                         ffe_ctl->found_offset = offset;
3631                         return 0;
3632                 }
3633         } else if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
3634                    !ffe_ctl->retry_clustered) {
3635                 spin_unlock(&last_ptr->refill_lock);
3636
3637                 ffe_ctl->retry_clustered = true;
3638                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3639                                 ffe_ctl->empty_cluster + ffe_ctl->empty_size);
3640                 return -EAGAIN;
3641         }
3642         /*
3643          * At this point we either didn't find a cluster or we weren't able to
3644          * allocate a block from our cluster.  Free the cluster we've been
3645          * trying to use, and go to the next block group.
3646          */
3647         btrfs_return_cluster_to_free_space(NULL, last_ptr);
3648         spin_unlock(&last_ptr->refill_lock);
3649         return 1;
3650 }
3651
3652 /*
3653  * Return >0 to inform caller that we find nothing
3654  * Return 0 when we found an free extent and set ffe_ctrl->found_offset
3655  * Return -EAGAIN to inform caller that we need to re-search this block group
3656  */
3657 static int find_free_extent_unclustered(struct btrfs_block_group *bg,
3658                                         struct find_free_extent_ctl *ffe_ctl)
3659 {
3660         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3661         u64 offset;
3662
3663         /*
3664          * We are doing an unclustered allocation, set the fragmented flag so
3665          * we don't bother trying to setup a cluster again until we get more
3666          * space.
3667          */
3668         if (unlikely(last_ptr)) {
3669                 spin_lock(&last_ptr->lock);
3670                 last_ptr->fragmented = 1;
3671                 spin_unlock(&last_ptr->lock);
3672         }
3673         if (ffe_ctl->cached) {
3674                 struct btrfs_free_space_ctl *free_space_ctl;
3675
3676                 free_space_ctl = bg->free_space_ctl;
3677                 spin_lock(&free_space_ctl->tree_lock);
3678                 if (free_space_ctl->free_space <
3679                     ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
3680                     ffe_ctl->empty_size) {
3681                         ffe_ctl->total_free_space = max_t(u64,
3682                                         ffe_ctl->total_free_space,
3683                                         free_space_ctl->free_space);
3684                         spin_unlock(&free_space_ctl->tree_lock);
3685                         return 1;
3686                 }
3687                 spin_unlock(&free_space_ctl->tree_lock);
3688         }
3689
3690         offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
3691                         ffe_ctl->num_bytes, ffe_ctl->empty_size,
3692                         &ffe_ctl->max_extent_size);
3693
3694         /*
3695          * If we didn't find a chunk, and we haven't failed on this block group
3696          * before, and this block group is in the middle of caching and we are
3697          * ok with waiting, then go ahead and wait for progress to be made, and
3698          * set @retry_unclustered to true.
3699          *
3700          * If @retry_unclustered is true then we've already waited on this
3701          * block group once and should move on to the next block group.
3702          */
3703         if (!offset && !ffe_ctl->retry_unclustered && !ffe_ctl->cached &&
3704             ffe_ctl->loop > LOOP_CACHING_NOWAIT) {
3705                 btrfs_wait_block_group_cache_progress(bg, ffe_ctl->num_bytes +
3706                                                       ffe_ctl->empty_size);
3707                 ffe_ctl->retry_unclustered = true;
3708                 return -EAGAIN;
3709         } else if (!offset) {
3710                 return 1;
3711         }
3712         ffe_ctl->found_offset = offset;
3713         return 0;
3714 }
3715
3716 static int do_allocation_clustered(struct btrfs_block_group *block_group,
3717                                    struct find_free_extent_ctl *ffe_ctl,
3718                                    struct btrfs_block_group **bg_ret)
3719 {
3720         int ret;
3721
3722         /* We want to try and use the cluster allocator, so lets look there */
3723         if (ffe_ctl->last_ptr && ffe_ctl->use_cluster) {
3724                 ret = find_free_extent_clustered(block_group, ffe_ctl, bg_ret);
3725                 if (ret >= 0 || ret == -EAGAIN)
3726                         return ret;
3727                 /* ret == -ENOENT case falls through */
3728         }
3729
3730         return find_free_extent_unclustered(block_group, ffe_ctl);
3731 }
3732
3733 static int do_allocation(struct btrfs_block_group *block_group,
3734                          struct find_free_extent_ctl *ffe_ctl,
3735                          struct btrfs_block_group **bg_ret)
3736 {
3737         switch (ffe_ctl->policy) {
3738         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3739                 return do_allocation_clustered(block_group, ffe_ctl, bg_ret);
3740         default:
3741                 BUG();
3742         }
3743 }
3744
3745 static void release_block_group(struct btrfs_block_group *block_group,
3746                                 struct find_free_extent_ctl *ffe_ctl,
3747                                 int delalloc)
3748 {
3749         switch (ffe_ctl->policy) {
3750         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3751                 ffe_ctl->retry_clustered = false;
3752                 ffe_ctl->retry_unclustered = false;
3753                 break;
3754         default:
3755                 BUG();
3756         }
3757
3758         BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
3759                ffe_ctl->index);
3760         btrfs_release_block_group(block_group, delalloc);
3761 }
3762
3763 static void found_extent_clustered(struct find_free_extent_ctl *ffe_ctl,
3764                                    struct btrfs_key *ins)
3765 {
3766         struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3767
3768         if (!ffe_ctl->use_cluster && last_ptr) {
3769                 spin_lock(&last_ptr->lock);
3770                 last_ptr->window_start = ins->objectid;
3771                 spin_unlock(&last_ptr->lock);
3772         }
3773 }
3774
3775 static void found_extent(struct find_free_extent_ctl *ffe_ctl,
3776                          struct btrfs_key *ins)
3777 {
3778         switch (ffe_ctl->policy) {
3779         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3780                 found_extent_clustered(ffe_ctl, ins);
3781                 break;
3782         default:
3783                 BUG();
3784         }
3785 }
3786
3787 static int chunk_allocation_failed(struct find_free_extent_ctl *ffe_ctl)
3788 {
3789         switch (ffe_ctl->policy) {
3790         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3791                 /*
3792                  * If we can't allocate a new chunk we've already looped through
3793                  * at least once, move on to the NO_EMPTY_SIZE case.
3794                  */
3795                 ffe_ctl->loop = LOOP_NO_EMPTY_SIZE;
3796                 return 0;
3797         default:
3798                 BUG();
3799         }
3800 }
3801
3802 /*
3803  * Return >0 means caller needs to re-search for free extent
3804  * Return 0 means we have the needed free extent.
3805  * Return <0 means we failed to locate any free extent.
3806  */
3807 static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
3808                                         struct btrfs_key *ins,
3809                                         struct find_free_extent_ctl *ffe_ctl,
3810                                         bool full_search)
3811 {
3812         struct btrfs_root *root = fs_info->extent_root;
3813         int ret;
3814
3815         if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
3816             ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
3817                 ffe_ctl->orig_have_caching_bg = true;
3818
3819         if (!ins->objectid && ffe_ctl->loop >= LOOP_CACHING_WAIT &&
3820             ffe_ctl->have_caching_bg)
3821                 return 1;
3822
3823         if (!ins->objectid && ++(ffe_ctl->index) < BTRFS_NR_RAID_TYPES)
3824                 return 1;
3825
3826         if (ins->objectid) {
3827                 found_extent(ffe_ctl, ins);
3828                 return 0;
3829         }
3830
3831         /*
3832          * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
3833          *                      caching kthreads as we move along
3834          * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
3835          * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
3836          * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
3837          *                     again
3838          */
3839         if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
3840                 ffe_ctl->index = 0;
3841                 if (ffe_ctl->loop == LOOP_CACHING_NOWAIT) {
3842                         /*
3843                          * We want to skip the LOOP_CACHING_WAIT step if we
3844                          * don't have any uncached bgs and we've already done a
3845                          * full search through.
3846                          */
3847                         if (ffe_ctl->orig_have_caching_bg || !full_search)
3848                                 ffe_ctl->loop = LOOP_CACHING_WAIT;
3849                         else
3850                                 ffe_ctl->loop = LOOP_ALLOC_CHUNK;
3851                 } else {
3852                         ffe_ctl->loop++;
3853                 }
3854
3855                 if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
3856                         struct btrfs_trans_handle *trans;
3857                         int exist = 0;
3858
3859                         trans = current->journal_info;
3860                         if (trans)
3861                                 exist = 1;
3862                         else
3863                                 trans = btrfs_join_transaction(root);
3864
3865                         if (IS_ERR(trans)) {
3866                                 ret = PTR_ERR(trans);
3867                                 return ret;
3868                         }
3869
3870                         ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
3871                                                 CHUNK_ALLOC_FORCE);
3872
3873                         /* Do not bail out on ENOSPC since we can do more. */
3874                         if (ret == -ENOSPC)
3875                                 ret = chunk_allocation_failed(ffe_ctl);
3876                         else if (ret < 0)
3877                                 btrfs_abort_transaction(trans, ret);
3878                         else
3879                                 ret = 0;
3880                         if (!exist)
3881                                 btrfs_end_transaction(trans);
3882                         if (ret)
3883                                 return ret;
3884                 }
3885
3886                 if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
3887                         if (ffe_ctl->policy != BTRFS_EXTENT_ALLOC_CLUSTERED)
3888                                 return -ENOSPC;
3889
3890                         /*
3891                          * Don't loop again if we already have no empty_size and
3892                          * no empty_cluster.
3893                          */
3894                         if (ffe_ctl->empty_size == 0 &&
3895                             ffe_ctl->empty_cluster == 0)
3896                                 return -ENOSPC;
3897                         ffe_ctl->empty_size = 0;
3898                         ffe_ctl->empty_cluster = 0;
3899                 }
3900                 return 1;
3901         }
3902         return -ENOSPC;
3903 }
3904
3905 static int prepare_allocation_clustered(struct btrfs_fs_info *fs_info,
3906                                         struct find_free_extent_ctl *ffe_ctl,
3907                                         struct btrfs_space_info *space_info,
3908                                         struct btrfs_key *ins)
3909 {
3910         /*
3911          * If our free space is heavily fragmented we may not be able to make
3912          * big contiguous allocations, so instead of doing the expensive search
3913          * for free space, simply return ENOSPC with our max_extent_size so we
3914          * can go ahead and search for a more manageable chunk.
3915          *
3916          * If our max_extent_size is large enough for our allocation simply
3917          * disable clustering since we will likely not be able to find enough
3918          * space to create a cluster and induce latency trying.
3919          */
3920         if (space_info->max_extent_size) {
3921                 spin_lock(&space_info->lock);
3922                 if (space_info->max_extent_size &&
3923                     ffe_ctl->num_bytes > space_info->max_extent_size) {
3924                         ins->offset = space_info->max_extent_size;
3925                         spin_unlock(&space_info->lock);
3926                         return -ENOSPC;
3927                 } else if (space_info->max_extent_size) {
3928                         ffe_ctl->use_cluster = false;
3929                 }
3930                 spin_unlock(&space_info->lock);
3931         }
3932
3933         ffe_ctl->last_ptr = fetch_cluster_info(fs_info, space_info,
3934                                                &ffe_ctl->empty_cluster);
3935         if (ffe_ctl->last_ptr) {
3936                 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3937
3938                 spin_lock(&last_ptr->lock);
3939                 if (last_ptr->block_group)
3940                         ffe_ctl->hint_byte = last_ptr->window_start;
3941                 if (last_ptr->fragmented) {
3942                         /*
3943                          * We still set window_start so we can keep track of the
3944                          * last place we found an allocation to try and save
3945                          * some time.
3946                          */
3947                         ffe_ctl->hint_byte = last_ptr->window_start;
3948                         ffe_ctl->use_cluster = false;
3949                 }
3950                 spin_unlock(&last_ptr->lock);
3951         }
3952
3953         return 0;
3954 }
3955
3956 static int prepare_allocation(struct btrfs_fs_info *fs_info,
3957                               struct find_free_extent_ctl *ffe_ctl,
3958                               struct btrfs_space_info *space_info,
3959                               struct btrfs_key *ins)
3960 {
3961         switch (ffe_ctl->policy) {
3962         case BTRFS_EXTENT_ALLOC_CLUSTERED:
3963                 return prepare_allocation_clustered(fs_info, ffe_ctl,
3964                                                     space_info, ins);
3965         default:
3966                 BUG();
3967         }
3968 }
3969
3970 /*
3971  * walks the btree of allocated extents and find a hole of a given size.
3972  * The key ins is changed to record the hole:
3973  * ins->objectid == start position
3974  * ins->flags = BTRFS_EXTENT_ITEM_KEY
3975  * ins->offset == the size of the hole.
3976  * Any available blocks before search_start are skipped.
3977  *
3978  * If there is no suitable free space, we will record the max size of
3979  * the free space extent currently.
3980  *
3981  * The overall logic and call chain:
3982  *
3983  * find_free_extent()
3984  * |- Iterate through all block groups
3985  * |  |- Get a valid block group
3986  * |  |- Try to do clustered allocation in that block group
3987  * |  |- Try to do unclustered allocation in that block group
3988  * |  |- Check if the result is valid
3989  * |  |  |- If valid, then exit
3990  * |  |- Jump to next block group
3991  * |
3992  * |- Push harder to find free extents
3993  *    |- If not found, re-iterate all block groups
3994  */
3995 static noinline int find_free_extent(struct btrfs_root *root,
3996                                 u64 ram_bytes, u64 num_bytes, u64 empty_size,
3997                                 u64 hint_byte_orig, struct btrfs_key *ins,
3998                                 u64 flags, int delalloc)
3999 {
4000         struct btrfs_fs_info *fs_info = root->fs_info;
4001         int ret = 0;
4002         int cache_block_group_error = 0;
4003         struct btrfs_block_group *block_group = NULL;
4004         struct find_free_extent_ctl ffe_ctl = {0};
4005         struct btrfs_space_info *space_info;
4006         bool full_search = false;
4007
4008         WARN_ON(num_bytes < fs_info->sectorsize);
4009
4010         ffe_ctl.num_bytes = num_bytes;
4011         ffe_ctl.empty_size = empty_size;
4012         ffe_ctl.flags = flags;
4013         ffe_ctl.search_start = 0;
4014         ffe_ctl.delalloc = delalloc;
4015         ffe_ctl.index = btrfs_bg_flags_to_raid_index(flags);
4016         ffe_ctl.have_caching_bg = false;
4017         ffe_ctl.orig_have_caching_bg = false;
4018         ffe_ctl.found_offset = 0;
4019         ffe_ctl.hint_byte = hint_byte_orig;
4020         ffe_ctl.policy = BTRFS_EXTENT_ALLOC_CLUSTERED;
4021
4022         /* For clustered allocation */
4023         ffe_ctl.retry_clustered = false;
4024         ffe_ctl.retry_unclustered = false;
4025         ffe_ctl.last_ptr = NULL;
4026         ffe_ctl.use_cluster = true;
4027
4028         ins->type = BTRFS_EXTENT_ITEM_KEY;
4029         ins->objectid = 0;
4030         ins->offset = 0;
4031
4032         trace_find_free_extent(root, num_bytes, empty_size, flags);
4033
4034         space_info = btrfs_find_space_info(fs_info, flags);
4035         if (!space_info) {
4036                 btrfs_err(fs_info, "No space info for %llu", flags);
4037                 return -ENOSPC;
4038         }
4039
4040         ret = prepare_allocation(fs_info, &ffe_ctl, space_info, ins);
4041         if (ret < 0)
4042                 return ret;
4043
4044         ffe_ctl.search_start = max(ffe_ctl.search_start,
4045                                    first_logical_byte(fs_info, 0));
4046         ffe_ctl.search_start = max(ffe_ctl.search_start, ffe_ctl.hint_byte);
4047         if (ffe_ctl.search_start == ffe_ctl.hint_byte) {
4048                 block_group = btrfs_lookup_block_group(fs_info,
4049                                                        ffe_ctl.search_start);
4050                 /*
4051                  * we don't want to use the block group if it doesn't match our
4052                  * allocation bits, or if its not cached.
4053                  *
4054                  * However if we are re-searching with an ideal block group
4055                  * picked out then we don't care that the block group is cached.
4056                  */
4057                 if (block_group && block_group_bits(block_group, flags) &&
4058                     block_group->cached != BTRFS_CACHE_NO) {
4059                         down_read(&space_info->groups_sem);
4060                         if (list_empty(&block_group->list) ||
4061                             block_group->ro) {
4062                                 /*
4063                                  * someone is removing this block group,
4064                                  * we can't jump into the have_block_group
4065                                  * target because our list pointers are not
4066                                  * valid
4067                                  */
4068                                 btrfs_put_block_group(block_group);
4069                                 up_read(&space_info->groups_sem);
4070                         } else {
4071                                 ffe_ctl.index = btrfs_bg_flags_to_raid_index(
4072                                                 block_group->flags);
4073                                 btrfs_lock_block_group(block_group, delalloc);
4074                                 goto have_block_group;
4075                         }
4076                 } else if (block_group) {
4077                         btrfs_put_block_group(block_group);
4078                 }
4079         }
4080 search:
4081         ffe_ctl.have_caching_bg = false;
4082         if (ffe_ctl.index == btrfs_bg_flags_to_raid_index(flags) ||
4083             ffe_ctl.index == 0)
4084                 full_search = true;
4085         down_read(&space_info->groups_sem);
4086         list_for_each_entry(block_group,
4087                             &space_info->block_groups[ffe_ctl.index], list) {
4088                 struct btrfs_block_group *bg_ret;
4089
4090                 /* If the block group is read-only, we can skip it entirely. */
4091                 if (unlikely(block_group->ro))
4092                         continue;
4093
4094                 btrfs_grab_block_group(block_group, delalloc);
4095                 ffe_ctl.search_start = block_group->start;
4096
4097                 /*
4098                  * this can happen if we end up cycling through all the
4099                  * raid types, but we want to make sure we only allocate
4100                  * for the proper type.
4101                  */
4102                 if (!block_group_bits(block_group, flags)) {
4103                         u64 extra = BTRFS_BLOCK_GROUP_DUP |
4104                                 BTRFS_BLOCK_GROUP_RAID1_MASK |
4105                                 BTRFS_BLOCK_GROUP_RAID56_MASK |
4106                                 BTRFS_BLOCK_GROUP_RAID10;
4107
4108                         /*
4109                          * if they asked for extra copies and this block group
4110                          * doesn't provide them, bail.  This does allow us to
4111                          * fill raid0 from raid1.
4112                          */
4113                         if ((flags & extra) && !(block_group->flags & extra))
4114                                 goto loop;
4115
4116                         /*
4117                          * This block group has different flags than we want.
4118                          * It's possible that we have MIXED_GROUP flag but no
4119                          * block group is mixed.  Just skip such block group.
4120                          */
4121                         btrfs_release_block_group(block_group, delalloc);
4122                         continue;
4123                 }
4124
4125 have_block_group:
4126                 ffe_ctl.cached = btrfs_block_group_done(block_group);
4127                 if (unlikely(!ffe_ctl.cached)) {
4128                         ffe_ctl.have_caching_bg = true;
4129                         ret = btrfs_cache_block_group(block_group, 0);
4130
4131                         /*
4132                          * If we get ENOMEM here or something else we want to
4133                          * try other block groups, because it may not be fatal.
4134                          * However if we can't find anything else we need to
4135                          * save our return here so that we return the actual
4136                          * error that caused problems, not ENOSPC.
4137                          */
4138                         if (ret < 0) {
4139                                 if (!cache_block_group_error)
4140                                         cache_block_group_error = ret;
4141                                 ret = 0;
4142                                 goto loop;
4143                         }
4144                         ret = 0;
4145                 }
4146
4147                 if (unlikely(block_group->cached == BTRFS_CACHE_ERROR)) {
4148                         if (!cache_block_group_error)
4149                                 cache_block_group_error = -EIO;
4150                         goto loop;
4151                 }
4152
4153                 bg_ret = NULL;
4154                 ret = do_allocation(block_group, &ffe_ctl, &bg_ret);
4155                 if (ret == 0) {
4156                         if (bg_ret && bg_ret != block_group) {
4157                                 btrfs_release_block_group(block_group, delalloc);
4158                                 block_group = bg_ret;
4159                         }
4160                 } else if (ret == -EAGAIN) {
4161                         goto have_block_group;
4162                 } else if (ret > 0) {
4163                         goto loop;
4164                 }
4165
4166                 /* Checks */
4167                 ffe_ctl.search_start = round_up(ffe_ctl.found_offset,
4168                                              fs_info->stripesize);
4169
4170                 /* move on to the next group */
4171                 if (ffe_ctl.search_start + num_bytes >
4172                     block_group->start + block_group->length) {
4173                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4174                                              num_bytes);
4175                         goto loop;
4176                 }
4177
4178                 if (ffe_ctl.found_offset < ffe_ctl.search_start)
4179                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4180                                 ffe_ctl.search_start - ffe_ctl.found_offset);
4181
4182                 ret = btrfs_add_reserved_bytes(block_group, ram_bytes,
4183                                 num_bytes, delalloc);
4184                 if (ret == -EAGAIN) {
4185                         btrfs_add_free_space(block_group, ffe_ctl.found_offset,
4186                                              num_bytes);
4187                         goto loop;
4188                 }
4189                 btrfs_inc_block_group_reservations(block_group);
4190
4191                 /* we are all good, lets return */
4192                 ins->objectid = ffe_ctl.search_start;
4193                 ins->offset = num_bytes;
4194
4195                 trace_btrfs_reserve_extent(block_group, ffe_ctl.search_start,
4196                                            num_bytes);
4197                 btrfs_release_block_group(block_group, delalloc);
4198                 break;
4199 loop:
4200                 release_block_group(block_group, &ffe_ctl, delalloc);
4201                 cond_resched();
4202         }
4203         up_read(&space_info->groups_sem);
4204
4205         ret = find_free_extent_update_loop(fs_info, ins, &ffe_ctl, full_search);
4206         if (ret > 0)
4207                 goto search;
4208
4209         if (ret == -ENOSPC && !cache_block_group_error) {
4210                 /*
4211                  * Use ffe_ctl->total_free_space as fallback if we can't find
4212                  * any contiguous hole.
4213                  */
4214                 if (!ffe_ctl.max_extent_size)
4215                         ffe_ctl.max_extent_size = ffe_ctl.total_free_space;
4216                 spin_lock(&space_info->lock);
4217                 space_info->max_extent_size = ffe_ctl.max_extent_size;
4218                 spin_unlock(&space_info->lock);
4219                 ins->offset = ffe_ctl.max_extent_size;
4220         } else if (ret == -ENOSPC) {
4221                 ret = cache_block_group_error;
4222         }
4223         return ret;
4224 }
4225
4226 /*
4227  * btrfs_reserve_extent - entry point to the extent allocator. Tries to find a
4228  *                        hole that is at least as big as @num_bytes.
4229  *
4230  * @root           -    The root that will contain this extent
4231  *
4232  * @ram_bytes      -    The amount of space in ram that @num_bytes take. This
4233  *                      is used for accounting purposes. This value differs
4234  *                      from @num_bytes only in the case of compressed extents.
4235  *
4236  * @num_bytes      -    Number of bytes to allocate on-disk.
4237  *
4238  * @min_alloc_size -    Indicates the minimum amount of space that the
4239  *                      allocator should try to satisfy. In some cases
4240  *                      @num_bytes may be larger than what is required and if
4241  *                      the filesystem is fragmented then allocation fails.
4242  *                      However, the presence of @min_alloc_size gives a
4243  *                      chance to try and satisfy the smaller allocation.
4244  *
4245  * @empty_size     -    A hint that you plan on doing more COW. This is the
4246  *                      size in bytes the allocator should try to find free
4247  *                      next to the block it returns.  This is just a hint and
4248  *                      may be ignored by the allocator.
4249  *
4250  * @hint_byte      -    Hint to the allocator to start searching above the byte
4251  *                      address passed. It might be ignored.
4252  *
4253  * @ins            -    This key is modified to record the found hole. It will
4254  *                      have the following values:
4255  *                      ins->objectid == start position
4256  *                      ins->flags = BTRFS_EXTENT_ITEM_KEY
4257  *                      ins->offset == the size of the hole.
4258  *
4259  * @is_data        -    Boolean flag indicating whether an extent is
4260  *                      allocated for data (true) or metadata (false)
4261  *
4262  * @delalloc       -    Boolean flag indicating whether this allocation is for
4263  *                      delalloc or not. If 'true' data_rwsem of block groups
4264  *                      is going to be acquired.
4265  *
4266  *
4267  * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
4268  * case -ENOSPC is returned then @ins->offset will contain the size of the
4269  * largest available hole the allocator managed to find.
4270  */
4271 int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
4272                          u64 num_bytes, u64 min_alloc_size,
4273                          u64 empty_size, u64 hint_byte,
4274                          struct btrfs_key *ins, int is_data, int delalloc)
4275 {
4276         struct btrfs_fs_info *fs_info = root->fs_info;
4277         bool final_tried = num_bytes == min_alloc_size;
4278         u64 flags;
4279         int ret;
4280
4281         flags = get_alloc_profile_by_root(root, is_data);
4282 again:
4283         WARN_ON(num_bytes < fs_info->sectorsize);
4284         ret = find_free_extent(root, ram_bytes, num_bytes, empty_size,
4285                                hint_byte, ins, flags, delalloc);
4286         if (!ret && !is_data) {
4287                 btrfs_dec_block_group_reservations(fs_info, ins->objectid);
4288         } else if (ret == -ENOSPC) {
4289                 if (!final_tried && ins->offset) {
4290                         num_bytes = min(num_bytes >> 1, ins->offset);
4291                         num_bytes = round_down(num_bytes,
4292                                                fs_info->sectorsize);
4293                         num_bytes = max(num_bytes, min_alloc_size);
4294                         ram_bytes = num_bytes;
4295                         if (num_bytes == min_alloc_size)
4296                                 final_tried = true;
4297                         goto again;
4298                 } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
4299                         struct btrfs_space_info *sinfo;
4300
4301                         sinfo = btrfs_find_space_info(fs_info, flags);
4302                         btrfs_err(fs_info,
4303                                   "allocation failed flags %llu, wanted %llu",
4304                                   flags, num_bytes);
4305                         if (sinfo)
4306                                 btrfs_dump_space_info(fs_info, sinfo,
4307                                                       num_bytes, 1);
4308                 }
4309         }
4310
4311         return ret;
4312 }
4313
4314 int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
4315                                u64 start, u64 len, int delalloc)
4316 {
4317         struct btrfs_block_group *cache;
4318
4319         cache = btrfs_lookup_block_group(fs_info, start);
4320         if (!cache) {
4321                 btrfs_err(fs_info, "Unable to find block group for %llu",
4322                           start);
4323                 return -ENOSPC;
4324         }
4325
4326         btrfs_add_free_space(cache, start, len);
4327         btrfs_free_reserved_bytes(cache, len, delalloc);
4328         trace_btrfs_reserved_extent_free(fs_info, start, len);
4329
4330         btrfs_put_block_group(cache);
4331         return 0;
4332 }
4333
4334 int btrfs_pin_reserved_extent(struct btrfs_trans_handle *trans, u64 start,
4335                               u64 len)
4336 {
4337         struct btrfs_block_group *cache;
4338         int ret = 0;
4339
4340         cache = btrfs_lookup_block_group(trans->fs_info, start);
4341         if (!cache) {
4342                 btrfs_err(trans->fs_info, "unable to find block group for %llu",
4343                           start);
4344                 return -ENOSPC;
4345         }
4346
4347         ret = pin_down_extent(trans, cache, start, len, 1);
4348         btrfs_put_block_group(cache);
4349         return ret;
4350 }
4351
4352 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4353                                       u64 parent, u64 root_objectid,
4354                                       u64 flags, u64 owner, u64 offset,
4355                                       struct btrfs_key *ins, int ref_mod)
4356 {
4357         struct btrfs_fs_info *fs_info = trans->fs_info;
4358         int ret;
4359         struct btrfs_extent_item *extent_item;
4360         struct btrfs_extent_inline_ref *iref;
4361         struct btrfs_path *path;
4362         struct extent_buffer *leaf;
4363         int type;
4364         u32 size;
4365
4366         if (parent > 0)
4367                 type = BTRFS_SHARED_DATA_REF_KEY;
4368         else
4369                 type = BTRFS_EXTENT_DATA_REF_KEY;
4370
4371         size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4372
4373         path = btrfs_alloc_path();
4374         if (!path)
4375                 return -ENOMEM;
4376
4377         path->leave_spinning = 1;
4378         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4379                                       ins, size);
4380         if (ret) {
4381                 btrfs_free_path(path);
4382                 return ret;
4383         }
4384
4385         leaf = path->nodes[0];
4386         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4387                                      struct btrfs_extent_item);
4388         btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4389         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4390         btrfs_set_extent_flags(leaf, extent_item,
4391                                flags | BTRFS_EXTENT_FLAG_DATA);
4392
4393         iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4394         btrfs_set_extent_inline_ref_type(leaf, iref, type);
4395         if (parent > 0) {
4396                 struct btrfs_shared_data_ref *ref;
4397                 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4398                 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4399                 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4400         } else {
4401                 struct btrfs_extent_data_ref *ref;
4402                 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4403                 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4404                 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4405                 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4406                 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4407         }
4408
4409         btrfs_mark_buffer_dirty(path->nodes[0]);
4410         btrfs_free_path(path);
4411
4412         ret = remove_from_free_space_tree(trans, ins->objectid, ins->offset);
4413         if (ret)
4414                 return ret;
4415
4416         ret = btrfs_update_block_group(trans, ins->objectid, ins->offset, 1);
4417         if (ret) { /* -ENOENT, logic error */
4418                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4419                         ins->objectid, ins->offset);
4420                 BUG();
4421         }
4422         trace_btrfs_reserved_extent_alloc(fs_info, ins->objectid, ins->offset);
4423         return ret;
4424 }
4425
4426 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4427                                      struct btrfs_delayed_ref_node *node,
4428                                      struct btrfs_delayed_extent_op *extent_op)
4429 {
4430         struct btrfs_fs_info *fs_info = trans->fs_info;
4431         int ret;
4432         struct btrfs_extent_item *extent_item;
4433         struct btrfs_key extent_key;
4434         struct btrfs_tree_block_info *block_info;
4435         struct btrfs_extent_inline_ref *iref;
4436         struct btrfs_path *path;
4437         struct extent_buffer *leaf;
4438         struct btrfs_delayed_tree_ref *ref;
4439         u32 size = sizeof(*extent_item) + sizeof(*iref);
4440         u64 num_bytes;
4441         u64 flags = extent_op->flags_to_set;
4442         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4443
4444         ref = btrfs_delayed_node_to_tree_ref(node);
4445
4446         extent_key.objectid = node->bytenr;
4447         if (skinny_metadata) {
4448                 extent_key.offset = ref->level;
4449                 extent_key.type = BTRFS_METADATA_ITEM_KEY;
4450                 num_bytes = fs_info->nodesize;
4451         } else {
4452                 extent_key.offset = node->num_bytes;
4453                 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4454                 size += sizeof(*block_info);
4455                 num_bytes = node->num_bytes;
4456         }
4457
4458         path = btrfs_alloc_path();
4459         if (!path)
4460                 return -ENOMEM;
4461
4462         path->leave_spinning = 1;
4463         ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4464                                       &extent_key, size);
4465         if (ret) {
4466                 btrfs_free_path(path);
4467                 return ret;
4468         }
4469
4470         leaf = path->nodes[0];
4471         extent_item = btrfs_item_ptr(leaf, path->slots[0],
4472                                      struct btrfs_extent_item);
4473         btrfs_set_extent_refs(leaf, extent_item, 1);
4474         btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4475         btrfs_set_extent_flags(leaf, extent_item,
4476                                flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4477
4478         if (skinny_metadata) {
4479                 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4480         } else {
4481                 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4482                 btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
4483                 btrfs_set_tree_block_level(leaf, block_info, ref->level);
4484                 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4485         }
4486
4487         if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
4488                 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4489                 btrfs_set_extent_inline_ref_type(leaf, iref,
4490                                                  BTRFS_SHARED_BLOCK_REF_KEY);
4491                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->parent);
4492         } else {
4493                 btrfs_set_extent_inline_ref_type(leaf, iref,
4494                                                  BTRFS_TREE_BLOCK_REF_KEY);
4495                 btrfs_set_extent_inline_ref_offset(leaf, iref, ref->root);
4496         }
4497
4498         btrfs_mark_buffer_dirty(leaf);
4499         btrfs_free_path(path);
4500
4501         ret = remove_from_free_space_tree(trans, extent_key.objectid,
4502                                           num_bytes);
4503         if (ret)
4504                 return ret;
4505
4506         ret = btrfs_update_block_group(trans, extent_key.objectid,
4507                                        fs_info->nodesize, 1);
4508         if (ret) { /* -ENOENT, logic error */
4509                 btrfs_err(fs_info, "update block group failed for %llu %llu",
4510                         extent_key.objectid, extent_key.offset);
4511                 BUG();
4512         }
4513
4514         trace_btrfs_reserved_extent_alloc(fs_info, extent_key.objectid,
4515                                           fs_info->nodesize);
4516         return ret;
4517 }
4518
4519 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4520                                      struct btrfs_root *root, u64 owner,
4521                                      u64 offset, u64 ram_bytes,
4522                                      struct btrfs_key *ins)
4523 {
4524         struct btrfs_ref generic_ref = { 0 };
4525
4526         BUG_ON(root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID);
4527
4528         btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4529                                ins->objectid, ins->offset, 0);
4530         btrfs_init_data_ref(&generic_ref, root->root_key.objectid, owner, offset);
4531         btrfs_ref_tree_mod(root->fs_info, &generic_ref);
4532
4533         return btrfs_add_delayed_data_ref(trans, &generic_ref, ram_bytes);
4534 }
4535
4536 /*
4537  * this is used by the tree logging recovery code.  It records that
4538  * an extent has been allocated and makes sure to clear the free
4539  * space cache bits as well
4540  */
4541 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4542                                    u64 root_objectid, u64 owner, u64 offset,
4543                                    struct btrfs_key *ins)
4544 {
4545         struct btrfs_fs_info *fs_info = trans->fs_info;
4546         int ret;
4547         struct btrfs_block_group *block_group;
4548         struct btrfs_space_info *space_info;
4549
4550         /*
4551          * Mixed block groups will exclude before processing the log so we only
4552          * need to do the exclude dance if this fs isn't mixed.
4553          */
4554         if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
4555                 ret = __exclude_logged_extent(fs_info, ins->objectid,
4556                                               ins->offset);
4557                 if (ret)
4558                         return ret;
4559         }
4560
4561         block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
4562         if (!block_group)
4563                 return -EINVAL;
4564
4565         space_info = block_group->space_info;
4566         spin_lock(&space_info->lock);
4567         spin_lock(&block_group->lock);
4568         space_info->bytes_reserved += ins->offset;
4569         block_group->reserved += ins->offset;
4570         spin_unlock(&block_group->lock);
4571         spin_unlock(&space_info->lock);
4572
4573         ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
4574                                          offset, ins, 1);
4575         if (ret)
4576                 btrfs_pin_extent(trans, ins->objectid, ins->offset, 1);
4577         btrfs_put_block_group(block_group);
4578         return ret;
4579 }
4580
4581 static struct extent_buffer *
4582 btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4583                       u64 bytenr, int level, u64 owner,
4584                       enum btrfs_lock_nesting nest)
4585 {
4586         struct btrfs_fs_info *fs_info = root->fs_info;
4587         struct extent_buffer *buf;
4588
4589         buf = btrfs_find_create_tree_block(fs_info, bytenr);
4590         if (IS_ERR(buf))
4591                 return buf;
4592
4593         /*
4594          * Extra safety check in case the extent tree is corrupted and extent
4595          * allocator chooses to use a tree block which is already used and
4596          * locked.
4597          */
4598         if (buf->lock_owner == current->pid) {
4599                 btrfs_err_rl(fs_info,
4600 "tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
4601                         buf->start, btrfs_header_owner(buf), current->pid);
4602                 free_extent_buffer(buf);
4603                 return ERR_PTR(-EUCLEAN);
4604         }
4605
4606         btrfs_set_buffer_lockdep_class(owner, buf, level);
4607         __btrfs_tree_lock(buf, nest);
4608         btrfs_clean_tree_block(buf);
4609         clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
4610
4611         btrfs_set_lock_blocking_write(buf);
4612         set_extent_buffer_uptodate(buf);
4613
4614         memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
4615         btrfs_set_header_level(buf, level);
4616         btrfs_set_header_bytenr(buf, buf->start);
4617         btrfs_set_header_generation(buf, trans->transid);
4618         btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
4619         btrfs_set_header_owner(buf, owner);
4620         write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
4621         write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
4622         if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4623                 buf->log_index = root->log_transid % 2;
4624                 /*
4625                  * we allow two log transactions at a time, use different
4626                  * EXTENT bit to differentiate dirty pages.
4627                  */
4628                 if (buf->log_index == 0)
4629                         set_extent_dirty(&root->dirty_log_pages, buf->start,
4630                                         buf->start + buf->len - 1, GFP_NOFS);
4631                 else
4632                         set_extent_new(&root->dirty_log_pages, buf->start,
4633                                         buf->start + buf->len - 1);
4634         } else {
4635                 buf->log_index = -1;
4636                 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4637                          buf->start + buf->len - 1, GFP_NOFS);
4638         }
4639         trans->dirty = true;
4640         /* this returns a buffer locked for blocking */
4641         return buf;
4642 }
4643
4644 /*
4645  * finds a free extent and does all the dirty work required for allocation
4646  * returns the tree buffer or an ERR_PTR on error.
4647  */
4648 struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
4649                                              struct btrfs_root *root,
4650                                              u64 parent, u64 root_objectid,
4651                                              const struct btrfs_disk_key *key,
4652                                              int level, u64 hint,
4653                                              u64 empty_size,
4654                                              enum btrfs_lock_nesting nest)
4655 {
4656         struct btrfs_fs_info *fs_info = root->fs_info;
4657         struct btrfs_key ins;
4658         struct btrfs_block_rsv *block_rsv;
4659         struct extent_buffer *buf;
4660         struct btrfs_delayed_extent_op *extent_op;
4661         struct btrfs_ref generic_ref = { 0 };
4662         u64 flags = 0;
4663         int ret;
4664         u32 blocksize = fs_info->nodesize;
4665         bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4666
4667 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4668         if (btrfs_is_testing(fs_info)) {
4669                 buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
4670                                             level, root_objectid, nest);
4671                 if (!IS_ERR(buf))
4672                         root->alloc_bytenr += blocksize;
4673                 return buf;
4674         }
4675 #endif
4676
4677         block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
4678         if (IS_ERR(block_rsv))
4679                 return ERR_CAST(block_rsv);
4680
4681         ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
4682                                    empty_size, hint, &ins, 0, 0);
4683         if (ret)
4684                 goto out_unuse;
4685
4686         buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
4687                                     root_objectid, nest);
4688         if (IS_ERR(buf)) {
4689                 ret = PTR_ERR(buf);
4690                 goto out_free_reserved;
4691         }
4692
4693         if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4694                 if (parent == 0)
4695                         parent = ins.objectid;
4696                 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4697         } else
4698                 BUG_ON(parent > 0);
4699
4700         if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4701                 extent_op = btrfs_alloc_delayed_extent_op();
4702                 if (!extent_op) {
4703                         ret = -ENOMEM;
4704                         goto out_free_buf;
4705                 }
4706                 if (key)
4707                         memcpy(&extent_op->key, key, sizeof(extent_op->key));
4708                 else
4709                         memset(&extent_op->key, 0, sizeof(extent_op->key));
4710                 extent_op->flags_to_set = flags;
4711                 extent_op->update_key = skinny_metadata ? false : true;
4712                 extent_op->update_flags = true;
4713                 extent_op->is_data = false;
4714                 extent_op->level = level;
4715
4716                 btrfs_init_generic_ref(&generic_ref, BTRFS_ADD_DELAYED_EXTENT,
4717                                        ins.objectid, ins.offset, parent);
4718                 generic_ref.real_root = root->root_key.objectid;
4719                 btrfs_init_tree_ref(&generic_ref, level, root_objectid);
4720                 btrfs_ref_tree_mod(fs_info, &generic_ref);
4721                 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, extent_op);
4722                 if (ret)
4723                         goto out_free_delayed;
4724         }
4725         return buf;
4726
4727 out_free_delayed:
4728         btrfs_free_delayed_extent_op(extent_op);
4729 out_free_buf:
4730         btrfs_tree_unlock(buf);
4731         free_extent_buffer(buf);
4732 out_free_reserved:
4733         btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
4734 out_unuse:
4735         btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
4736         return ERR_PTR(ret);
4737 }
4738
4739 struct walk_control {
4740         u64 refs[BTRFS_MAX_LEVEL];
4741         u64 flags[BTRFS_MAX_LEVEL];
4742         struct btrfs_key update_progress;
4743         struct btrfs_key drop_progress;
4744         int drop_level;
4745         int stage;
4746         int level;
4747         int shared_level;
4748         int update_ref;
4749         int keep_locks;
4750         int reada_slot;
4751         int reada_count;
4752         int restarted;
4753 };
4754
4755 #define DROP_REFERENCE  1
4756 #define UPDATE_BACKREF  2
4757
4758 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4759                                      struct btrfs_root *root,
4760                                      struct walk_control *wc,
4761                                      struct btrfs_path *path)
4762 {
4763         struct btrfs_fs_info *fs_info = root->fs_info;
4764         u64 bytenr;
4765         u64 generation;
4766         u64 refs;
4767         u64 flags;
4768         u32 nritems;
4769         struct btrfs_key key;
4770         struct extent_buffer *eb;
4771         int ret;
4772         int slot;
4773         int nread = 0;
4774
4775         if (path->slots[wc->level] < wc->reada_slot) {
4776                 wc->reada_count = wc->reada_count * 2 / 3;
4777                 wc->reada_count = max(wc->reada_count, 2);
4778         } else {
4779                 wc->reada_count = wc->reada_count * 3 / 2;
4780                 wc->reada_count = min_t(int, wc->reada_count,
4781                                         BTRFS_NODEPTRS_PER_BLOCK(fs_info));
4782         }
4783
4784         eb = path->nodes[wc->level];
4785         nritems = btrfs_header_nritems(eb);
4786
4787         for (slot = path->slots[wc->level]; slot < nritems; slot++) {
4788                 if (nread >= wc->reada_count)
4789                         break;
4790
4791                 cond_resched();
4792                 bytenr = btrfs_node_blockptr(eb, slot);
4793                 generation = btrfs_node_ptr_generation(eb, slot);
4794
4795                 if (slot == path->slots[wc->level])
4796                         goto reada;
4797
4798                 if (wc->stage == UPDATE_BACKREF &&
4799                     generation <= root->root_key.offset)
4800                         continue;
4801
4802                 /* We don't lock the tree block, it's OK to be racy here */
4803                 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
4804                                                wc->level - 1, 1, &refs,
4805                                                &flags);
4806                 /* We don't care about errors in readahead. */
4807                 if (ret < 0)
4808                         continue;
4809                 BUG_ON(refs == 0);
4810
4811                 if (wc->stage == DROP_REFERENCE) {
4812                         if (refs == 1)
4813                                 goto reada;
4814
4815                         if (wc->level == 1 &&
4816                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4817                                 continue;
4818                         if (!wc->update_ref ||
4819                             generation <= root->root_key.offset)
4820                                 continue;
4821                         btrfs_node_key_to_cpu(eb, &key, slot);
4822                         ret = btrfs_comp_cpu_keys(&key,
4823                                                   &wc->update_progress);
4824                         if (ret < 0)
4825                                 continue;
4826                 } else {
4827                         if (wc->level == 1 &&
4828                             (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
4829                                 continue;
4830                 }
4831 reada:
4832                 readahead_tree_block(fs_info, bytenr);
4833                 nread++;
4834         }
4835         wc->reada_slot = slot;
4836 }
4837
4838 /*
4839  * helper to process tree block while walking down the tree.
4840  *
4841  * when wc->stage == UPDATE_BACKREF, this function updates
4842  * back refs for pointers in the block.
4843  *
4844  * NOTE: return value 1 means we should stop walking down.
4845  */
4846 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4847                                    struct btrfs_root *root,
4848                                    struct btrfs_path *path,
4849                                    struct walk_control *wc, int lookup_info)
4850 {
4851         struct btrfs_fs_info *fs_info = root->fs_info;
4852         int level = wc->level;
4853         struct extent_buffer *eb = path->nodes[level];
4854         u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
4855         int ret;
4856
4857         if (wc->stage == UPDATE_BACKREF &&
4858             btrfs_header_owner(eb) != root->root_key.objectid)
4859                 return 1;
4860
4861         /*
4862          * when reference count of tree block is 1, it won't increase
4863          * again. once full backref flag is set, we never clear it.
4864          */
4865         if (lookup_info &&
4866             ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
4867              (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
4868                 BUG_ON(!path->locks[level]);
4869                 ret = btrfs_lookup_extent_info(trans, fs_info,
4870                                                eb->start, level, 1,
4871                                                &wc->refs[level],
4872                                                &wc->flags[level]);
4873                 BUG_ON(ret == -ENOMEM);
4874                 if (ret)
4875                         return ret;
4876                 BUG_ON(wc->refs[level] == 0);
4877         }
4878
4879         if (wc->stage == DROP_REFERENCE) {
4880                 if (wc->refs[level] > 1)
4881                         return 1;
4882
4883                 if (path->locks[level] && !wc->keep_locks) {
4884                         btrfs_tree_unlock_rw(eb, path->locks[level]);
4885                         path->locks[level] = 0;
4886                 }
4887                 return 0;
4888         }
4889
4890         /* wc->stage == UPDATE_BACKREF */
4891         if (!(wc->flags[level] & flag)) {
4892                 BUG_ON(!path->locks[level]);
4893                 ret = btrfs_inc_ref(trans, root, eb, 1);
4894                 BUG_ON(ret); /* -ENOMEM */
4895                 ret = btrfs_dec_ref(trans, root, eb, 0);
4896                 BUG_ON(ret); /* -ENOMEM */
4897                 ret = btrfs_set_disk_extent_flags(trans, eb, flag,
4898                                                   btrfs_header_level(eb), 0);
4899                 BUG_ON(ret); /* -ENOMEM */
4900                 wc->flags[level] |= flag;
4901         }
4902
4903         /*
4904          * the block is shared by multiple trees, so it's not good to
4905          * keep the tree lock
4906          */
4907         if (path->locks[level] && level > 0) {
4908                 btrfs_tree_unlock_rw(eb, path->locks[level]);
4909                 path->locks[level] = 0;
4910         }
4911         return 0;
4912 }
4913
4914 /*
4915  * This is used to verify a ref exists for this root to deal with a bug where we
4916  * would have a drop_progress key that hadn't been updated properly.
4917  */
4918 static int check_ref_exists(struct btrfs_trans_handle *trans,
4919                             struct btrfs_root *root, u64 bytenr, u64 parent,
4920                             int level)
4921 {
4922         struct btrfs_path *path;
4923         struct btrfs_extent_inline_ref *iref;
4924         int ret;
4925
4926         path = btrfs_alloc_path();
4927         if (!path)
4928                 return -ENOMEM;
4929
4930         ret = lookup_extent_backref(trans, path, &iref, bytenr,
4931                                     root->fs_info->nodesize, parent,
4932                                     root->root_key.objectid, level, 0);
4933         btrfs_free_path(path);
4934         if (ret == -ENOENT)
4935                 return 0;
4936         if (ret < 0)
4937                 return ret;
4938         return 1;
4939 }
4940
4941 /*
4942  * helper to process tree block pointer.
4943  *
4944  * when wc->stage == DROP_REFERENCE, this function checks
4945  * reference count of the block pointed to. if the block
4946  * is shared and we need update back refs for the subtree
4947  * rooted at the block, this function changes wc->stage to
4948  * UPDATE_BACKREF. if the block is shared and there is no
4949  * need to update back, this function drops the reference
4950  * to the block.
4951  *
4952  * NOTE: return value 1 means we should stop walking down.
4953  */
4954 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
4955                                  struct btrfs_root *root,
4956                                  struct btrfs_path *path,
4957                                  struct walk_control *wc, int *lookup_info)
4958 {
4959         struct btrfs_fs_info *fs_info = root->fs_info;
4960         u64 bytenr;
4961         u64 generation;
4962         u64 parent;
4963         struct btrfs_key key;
4964         struct btrfs_key first_key;
4965         struct btrfs_ref ref = { 0 };
4966         struct extent_buffer *next;
4967         int level = wc->level;
4968         int reada = 0;
4969         int ret = 0;
4970         bool need_account = false;
4971
4972         generation = btrfs_node_ptr_generation(path->nodes[level],
4973                                                path->slots[level]);
4974         /*
4975          * if the lower level block was created before the snapshot
4976          * was created, we know there is no need to update back refs
4977          * for the subtree
4978          */
4979         if (wc->stage == UPDATE_BACKREF &&
4980             generation <= root->root_key.offset) {
4981                 *lookup_info = 1;
4982                 return 1;
4983         }
4984
4985         bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
4986         btrfs_node_key_to_cpu(path->nodes[level], &first_key,
4987                               path->slots[level]);
4988
4989         next = find_extent_buffer(fs_info, bytenr);
4990         if (!next) {
4991                 next = btrfs_find_create_tree_block(fs_info, bytenr);
4992                 if (IS_ERR(next))
4993                         return PTR_ERR(next);
4994
4995                 btrfs_set_buffer_lockdep_class(root->root_key.objectid, next,
4996                                                level - 1);
4997                 reada = 1;
4998         }
4999         btrfs_tree_lock(next);
5000         btrfs_set_lock_blocking_write(next);
5001
5002         ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
5003                                        &wc->refs[level - 1],
5004                                        &wc->flags[level - 1]);
5005         if (ret < 0)
5006                 goto out_unlock;
5007
5008         if (unlikely(wc->refs[level - 1] == 0)) {
5009                 btrfs_err(fs_info, "Missing references.");
5010                 ret = -EIO;
5011                 goto out_unlock;
5012         }
5013         *lookup_info = 0;
5014
5015         if (wc->stage == DROP_REFERENCE) {
5016                 if (wc->refs[level - 1] > 1) {
5017                         need_account = true;
5018                         if (level == 1 &&
5019                             (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5020                                 goto skip;
5021
5022                         if (!wc->update_ref ||
5023                             generation <= root->root_key.offset)
5024                                 goto skip;
5025
5026                         btrfs_node_key_to_cpu(path->nodes[level], &key,
5027                                               path->slots[level]);
5028                         ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
5029                         if (ret < 0)
5030                                 goto skip;
5031
5032                         wc->stage = UPDATE_BACKREF;
5033                         wc->shared_level = level - 1;
5034                 }
5035         } else {
5036                 if (level == 1 &&
5037                     (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5038                         goto skip;
5039         }
5040
5041         if (!btrfs_buffer_uptodate(next, generation, 0)) {
5042                 btrfs_tree_unlock(next);
5043                 free_extent_buffer(next);
5044                 next = NULL;
5045                 *lookup_info = 1;
5046         }
5047
5048         if (!next) {
5049                 if (reada && level == 1)
5050                         reada_walk_down(trans, root, wc, path);
5051                 next = read_tree_block(fs_info, bytenr, generation, level - 1,
5052                                        &first_key);
5053                 if (IS_ERR(next)) {
5054                         return PTR_ERR(next);
5055                 } else if (!extent_buffer_uptodate(next)) {
5056                         free_extent_buffer(next);
5057                         return -EIO;
5058                 }
5059                 btrfs_tree_lock(next);
5060                 btrfs_set_lock_blocking_write(next);
5061         }
5062
5063         level--;
5064         ASSERT(level == btrfs_header_level(next));
5065         if (level != btrfs_header_level(next)) {
5066                 btrfs_err(root->fs_info, "mismatched level");
5067                 ret = -EIO;
5068                 goto out_unlock;
5069         }
5070         path->nodes[level] = next;
5071         path->slots[level] = 0;
5072         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5073         wc->level = level;
5074         if (wc->level == 1)
5075                 wc->reada_slot = 0;
5076         return 0;
5077 skip:
5078         wc->refs[level - 1] = 0;
5079         wc->flags[level - 1] = 0;
5080         if (wc->stage == DROP_REFERENCE) {
5081                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5082                         parent = path->nodes[level]->start;
5083                 } else {
5084                         ASSERT(root->root_key.objectid ==
5085                                btrfs_header_owner(path->nodes[level]));
5086                         if (root->root_key.objectid !=
5087                             btrfs_header_owner(path->nodes[level])) {
5088                                 btrfs_err(root->fs_info,
5089                                                 "mismatched block owner");
5090                                 ret = -EIO;
5091                                 goto out_unlock;
5092                         }
5093                         parent = 0;
5094                 }
5095
5096                 /*
5097                  * If we had a drop_progress we need to verify the refs are set
5098                  * as expected.  If we find our ref then we know that from here
5099                  * on out everything should be correct, and we can clear the
5100                  * ->restarted flag.
5101                  */
5102                 if (wc->restarted) {
5103                         ret = check_ref_exists(trans, root, bytenr, parent,
5104                                                level - 1);
5105                         if (ret < 0)
5106                                 goto out_unlock;
5107                         if (ret == 0)
5108                                 goto no_delete;
5109                         ret = 0;
5110                         wc->restarted = 0;
5111                 }
5112
5113                 /*
5114                  * Reloc tree doesn't contribute to qgroup numbers, and we have
5115                  * already accounted them at merge time (replace_path),
5116                  * thus we could skip expensive subtree trace here.
5117                  */
5118                 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
5119                     need_account) {
5120                         ret = btrfs_qgroup_trace_subtree(trans, next,
5121                                                          generation, level - 1);
5122                         if (ret) {
5123                                 btrfs_err_rl(fs_info,
5124                                              "Error %d accounting shared subtree. Quota is out of sync, rescan required.",
5125                                              ret);
5126                         }
5127                 }
5128
5129                 /*
5130                  * We need to update the next key in our walk control so we can
5131                  * update the drop_progress key accordingly.  We don't care if
5132                  * find_next_key doesn't find a key because that means we're at
5133                  * the end and are going to clean up now.
5134                  */
5135                 wc->drop_level = level;
5136                 find_next_key(path, level, &wc->drop_progress);
5137
5138                 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
5139                                        fs_info->nodesize, parent);
5140                 btrfs_init_tree_ref(&ref, level - 1, root->root_key.objectid);
5141                 ret = btrfs_free_extent(trans, &ref);
5142                 if (ret)
5143                         goto out_unlock;
5144         }
5145 no_delete:
5146         *lookup_info = 1;
5147         ret = 1;
5148
5149 out_unlock:
5150         btrfs_tree_unlock(next);
5151         free_extent_buffer(next);
5152
5153         return ret;
5154 }
5155
5156 /*
5157  * helper to process tree block while walking up the tree.
5158  *
5159  * when wc->stage == DROP_REFERENCE, this function drops
5160  * reference count on the block.
5161  *
5162  * when wc->stage == UPDATE_BACKREF, this function changes
5163  * wc->stage back to DROP_REFERENCE if we changed wc->stage
5164  * to UPDATE_BACKREF previously while processing the block.
5165  *
5166  * NOTE: return value 1 means we should stop walking up.
5167  */
5168 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5169                                  struct btrfs_root *root,
5170                                  struct btrfs_path *path,
5171                                  struct walk_control *wc)
5172 {
5173         struct btrfs_fs_info *fs_info = root->fs_info;
5174         int ret;
5175         int level = wc->level;
5176         struct extent_buffer *eb = path->nodes[level];
5177         u64 parent = 0;
5178
5179         if (wc->stage == UPDATE_BACKREF) {
5180                 BUG_ON(wc->shared_level < level);
5181                 if (level < wc->shared_level)
5182                         goto out;
5183
5184                 ret = find_next_key(path, level + 1, &wc->update_progress);
5185                 if (ret > 0)
5186                         wc->update_ref = 0;
5187
5188                 wc->stage = DROP_REFERENCE;
5189                 wc->shared_level = -1;
5190                 path->slots[level] = 0;
5191
5192                 /*
5193                  * check reference count again if the block isn't locked.
5194                  * we should start walking down the tree again if reference
5195                  * count is one.
5196                  */
5197                 if (!path->locks[level]) {
5198                         BUG_ON(level == 0);
5199                         btrfs_tree_lock(eb);
5200                         btrfs_set_lock_blocking_write(eb);
5201                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5202
5203                         ret = btrfs_lookup_extent_info(trans, fs_info,
5204                                                        eb->start, level, 1,
5205                                                        &wc->refs[level],
5206                                                        &wc->flags[level]);
5207                         if (ret < 0) {
5208                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5209                                 path->locks[level] = 0;
5210                                 return ret;
5211                         }
5212                         BUG_ON(wc->refs[level] == 0);
5213                         if (wc->refs[level] == 1) {
5214                                 btrfs_tree_unlock_rw(eb, path->locks[level]);
5215                                 path->locks[level] = 0;
5216                                 return 1;
5217                         }
5218                 }
5219         }
5220
5221         /* wc->stage == DROP_REFERENCE */
5222         BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5223
5224         if (wc->refs[level] == 1) {
5225                 if (level == 0) {
5226                         if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5227                                 ret = btrfs_dec_ref(trans, root, eb, 1);
5228                         else
5229                                 ret = btrfs_dec_ref(trans, root, eb, 0);
5230                         BUG_ON(ret); /* -ENOMEM */
5231                         if (is_fstree(root->root_key.objectid)) {
5232                                 ret = btrfs_qgroup_trace_leaf_items(trans, eb);
5233                                 if (ret) {
5234                                         btrfs_err_rl(fs_info,
5235         "error %d accounting leaf items, quota is out of sync, rescan required",
5236                                              ret);
5237                                 }
5238                         }
5239                 }
5240                 /* make block locked assertion in btrfs_clean_tree_block happy */
5241                 if (!path->locks[level] &&
5242                     btrfs_header_generation(eb) == trans->transid) {
5243                         btrfs_tree_lock(eb);
5244                         btrfs_set_lock_blocking_write(eb);
5245                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5246                 }
5247                 btrfs_clean_tree_block(eb);
5248         }
5249
5250         if (eb == root->node) {
5251                 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5252                         parent = eb->start;
5253                 else if (root->root_key.objectid != btrfs_header_owner(eb))
5254                         goto owner_mismatch;
5255         } else {
5256                 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5257                         parent = path->nodes[level + 1]->start;
5258                 else if (root->root_key.objectid !=
5259                          btrfs_header_owner(path->nodes[level + 1]))
5260                         goto owner_mismatch;
5261         }
5262
5263         btrfs_free_tree_block(trans, root, eb, parent, wc->refs[level] == 1);
5264 out:
5265         wc->refs[level] = 0;
5266         wc->flags[level] = 0;
5267         return 0;
5268
5269 owner_mismatch:
5270         btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
5271                      btrfs_header_owner(eb), root->root_key.objectid);
5272         return -EUCLEAN;
5273 }
5274
5275 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5276                                    struct btrfs_root *root,
5277                                    struct btrfs_path *path,
5278                                    struct walk_control *wc)
5279 {
5280         int level = wc->level;
5281         int lookup_info = 1;
5282         int ret;
5283
5284         while (level >= 0) {
5285                 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5286                 if (ret > 0)
5287                         break;
5288
5289                 if (level == 0)
5290                         break;
5291
5292                 if (path->slots[level] >=
5293                     btrfs_header_nritems(path->nodes[level]))
5294                         break;
5295
5296                 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5297                 if (ret > 0) {
5298                         path->slots[level]++;
5299                         continue;
5300                 } else if (ret < 0)
5301                         return ret;
5302                 level = wc->level;
5303         }
5304         return 0;
5305 }
5306
5307 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5308                                  struct btrfs_root *root,
5309                                  struct btrfs_path *path,
5310                                  struct walk_control *wc, int max_level)
5311 {
5312         int level = wc->level;
5313         int ret;
5314
5315         path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5316         while (level < max_level && path->nodes[level]) {
5317                 wc->level = level;
5318                 if (path->slots[level] + 1 <
5319                     btrfs_header_nritems(path->nodes[level])) {
5320                         path->slots[level]++;
5321                         return 0;
5322                 } else {
5323                         ret = walk_up_proc(trans, root, path, wc);
5324                         if (ret > 0)
5325                                 return 0;
5326                         if (ret < 0)
5327                                 return ret;
5328
5329                         if (path->locks[level]) {
5330                                 btrfs_tree_unlock_rw(path->nodes[level],
5331                                                      path->locks[level]);
5332                                 path->locks[level] = 0;
5333                         }
5334                         free_extent_buffer(path->nodes[level]);
5335                         path->nodes[level] = NULL;
5336                         level++;
5337                 }
5338         }
5339         return 1;
5340 }
5341
5342 /*
5343  * drop a subvolume tree.
5344  *
5345  * this function traverses the tree freeing any blocks that only
5346  * referenced by the tree.
5347  *
5348  * when a shared tree block is found. this function decreases its
5349  * reference count by one. if update_ref is true, this function
5350  * also make sure backrefs for the shared block and all lower level
5351  * blocks are properly updated.
5352  *
5353  * If called with for_reloc == 0, may exit early with -EAGAIN
5354  */
5355 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc)
5356 {
5357         struct btrfs_fs_info *fs_info = root->fs_info;
5358         struct btrfs_path *path;
5359         struct btrfs_trans_handle *trans;
5360         struct btrfs_root *tree_root = fs_info->tree_root;
5361         struct btrfs_root_item *root_item = &root->root_item;
5362         struct walk_control *wc;
5363         struct btrfs_key key;
5364         int err = 0;
5365         int ret;
5366         int level;
5367         bool root_dropped = false;
5368
5369         btrfs_debug(fs_info, "Drop subvolume %llu", root->root_key.objectid);
5370
5371         path = btrfs_alloc_path();
5372         if (!path) {
5373                 err = -ENOMEM;
5374                 goto out;
5375         }
5376
5377         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5378         if (!wc) {
5379                 btrfs_free_path(path);
5380                 err = -ENOMEM;
5381                 goto out;
5382         }
5383
5384         /*
5385          * Use join to avoid potential EINTR from transaction start. See
5386          * wait_reserve_ticket and the whole reservation callchain.
5387          */
5388         if (for_reloc)
5389                 trans = btrfs_join_transaction(tree_root);
5390         else
5391                 trans = btrfs_start_transaction(tree_root, 0);
5392         if (IS_ERR(trans)) {
5393                 err = PTR_ERR(trans);
5394                 goto out_free;
5395         }
5396
5397         err = btrfs_run_delayed_items(trans);
5398         if (err)
5399                 goto out_end_trans;
5400
5401         /*
5402          * This will help us catch people modifying the fs tree while we're
5403          * dropping it.  It is unsafe to mess with the fs tree while it's being
5404          * dropped as we unlock the root node and parent nodes as we walk down
5405          * the tree, assuming nothing will change.  If something does change
5406          * then we'll have stale information and drop references to blocks we've
5407          * already dropped.
5408          */
5409         set_bit(BTRFS_ROOT_DELETING, &root->state);
5410         if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5411                 level = btrfs_header_level(root->node);
5412                 path->nodes[level] = btrfs_lock_root_node(root);
5413                 btrfs_set_lock_blocking_write(path->nodes[level]);
5414                 path->slots[level] = 0;
5415                 path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5416                 memset(&wc->update_progress, 0,
5417                        sizeof(wc->update_progress));
5418         } else {
5419                 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5420                 memcpy(&wc->update_progress, &key,
5421                        sizeof(wc->update_progress));
5422
5423                 level = root_item->drop_level;
5424                 BUG_ON(level == 0);
5425                 path->lowest_level = level;
5426                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5427                 path->lowest_level = 0;
5428                 if (ret < 0) {
5429                         err = ret;
5430                         goto out_end_trans;
5431                 }
5432                 WARN_ON(ret > 0);
5433
5434                 /*
5435                  * unlock our path, this is safe because only this
5436                  * function is allowed to delete this snapshot
5437                  */
5438                 btrfs_unlock_up_safe(path, 0);
5439
5440                 level = btrfs_header_level(root->node);
5441                 while (1) {
5442                         btrfs_tree_lock(path->nodes[level]);
5443                         btrfs_set_lock_blocking_write(path->nodes[level]);
5444                         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5445
5446                         ret = btrfs_lookup_extent_info(trans, fs_info,
5447                                                 path->nodes[level]->start,
5448                                                 level, 1, &wc->refs[level],
5449                                                 &wc->flags[level]);
5450                         if (ret < 0) {
5451                                 err = ret;
5452                                 goto out_end_trans;
5453                         }
5454                         BUG_ON(wc->refs[level] == 0);
5455
5456                         if (level == root_item->drop_level)
5457                                 break;
5458
5459                         btrfs_tree_unlock(path->nodes[level]);
5460                         path->locks[level] = 0;
5461                         WARN_ON(wc->refs[level] != 1);
5462                         level--;
5463                 }
5464         }
5465
5466         wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
5467         wc->level = level;
5468         wc->shared_level = -1;
5469         wc->stage = DROP_REFERENCE;
5470         wc->update_ref = update_ref;
5471         wc->keep_locks = 0;
5472         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5473
5474         while (1) {
5475
5476                 ret = walk_down_tree(trans, root, path, wc);
5477                 if (ret < 0) {
5478                         err = ret;
5479                         break;
5480                 }
5481
5482                 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5483                 if (ret < 0) {
5484                         err = ret;
5485                         break;
5486                 }
5487
5488                 if (ret > 0) {
5489                         BUG_ON(wc->stage != DROP_REFERENCE);
5490                         break;
5491                 }
5492
5493                 if (wc->stage == DROP_REFERENCE) {
5494                         wc->drop_level = wc->level;
5495                         btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
5496                                               &wc->drop_progress,
5497                                               path->slots[wc->drop_level]);
5498                 }
5499                 btrfs_cpu_key_to_disk(&root_item->drop_progress,
5500                                       &wc->drop_progress);
5501                 root_item->drop_level = wc->drop_level;
5502
5503                 BUG_ON(wc->level == 0);
5504                 if (btrfs_should_end_transaction(trans) ||
5505                     (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
5506                         ret = btrfs_update_root(trans, tree_root,
5507                                                 &root->root_key,
5508                                                 root_item);
5509                         if (ret) {
5510                                 btrfs_abort_transaction(trans, ret);
5511                                 err = ret;
5512                                 goto out_end_trans;
5513                         }
5514
5515                         btrfs_end_transaction_throttle(trans);
5516                         if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
5517                                 btrfs_debug(fs_info,
5518                                             "drop snapshot early exit");
5519                                 err = -EAGAIN;
5520                                 goto out_free;
5521                         }
5522
5523                        /*
5524                         * Use join to avoid potential EINTR from transaction
5525                         * start. See wait_reserve_ticket and the whole
5526                         * reservation callchain.
5527                         */
5528                         if (for_reloc)
5529                                 trans = btrfs_join_transaction(tree_root);
5530                         else
5531                                 trans = btrfs_start_transaction(tree_root, 0);
5532                         if (IS_ERR(trans)) {
5533                                 err = PTR_ERR(trans);
5534                                 goto out_free;
5535                         }
5536                 }
5537         }
5538         btrfs_release_path(path);
5539         if (err)
5540                 goto out_end_trans;
5541
5542         ret = btrfs_del_root(trans, &root->root_key);
5543         if (ret) {
5544                 btrfs_abort_transaction(trans, ret);
5545                 err = ret;
5546                 goto out_end_trans;
5547         }
5548
5549         if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5550                 ret = btrfs_find_root(tree_root, &root->root_key, path,
5551                                       NULL, NULL);
5552                 if (ret < 0) {
5553                         btrfs_abort_transaction(trans, ret);
5554                         err = ret;
5555                         goto out_end_trans;
5556                 } else if (ret > 0) {
5557                         /* if we fail to delete the orphan item this time
5558                          * around, it'll get picked up the next time.
5559                          *
5560                          * The most common failure here is just -ENOENT.
5561                          */
5562                         btrfs_del_orphan_item(trans, tree_root,
5563                                               root->root_key.objectid);
5564                 }
5565         }
5566
5567         /*
5568          * This subvolume is going to be completely dropped, and won't be
5569          * recorded as dirty roots, thus pertrans meta rsv will not be freed at
5570          * commit transaction time.  So free it here manually.
5571          */
5572         btrfs_qgroup_convert_reserved_meta(root, INT_MAX);
5573         btrfs_qgroup_free_meta_all_pertrans(root);
5574
5575         if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state))
5576                 btrfs_add_dropped_root(trans, root);
5577         else
5578                 btrfs_put_root(root);
5579         root_dropped = true;
5580 out_end_trans:
5581         btrfs_end_transaction_throttle(trans);
5582 out_free:
5583         kfree(wc);
5584         btrfs_free_path(path);
5585 out:
5586         /*
5587          * So if we need to stop dropping the snapshot for whatever reason we
5588          * need to make sure to add it back to the dead root list so that we
5589          * keep trying to do the work later.  This also cleans up roots if we
5590          * don't have it in the radix (like when we recover after a power fail
5591          * or unmount) so we don't leak memory.
5592          */
5593         if (!for_reloc && !root_dropped)
5594                 btrfs_add_dead_root(root);
5595         return err;
5596 }
5597
5598 /*
5599  * drop subtree rooted at tree block 'node'.
5600  *
5601  * NOTE: this function will unlock and release tree block 'node'
5602  * only used by relocation code
5603  */
5604 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5605                         struct btrfs_root *root,
5606                         struct extent_buffer *node,
5607                         struct extent_buffer *parent)
5608 {
5609         struct btrfs_fs_info *fs_info = root->fs_info;
5610         struct btrfs_path *path;
5611         struct walk_control *wc;
5612         int level;
5613         int parent_level;
5614         int ret = 0;
5615         int wret;
5616
5617         BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5618
5619         path = btrfs_alloc_path();
5620         if (!path)
5621                 return -ENOMEM;
5622
5623         wc = kzalloc(sizeof(*wc), GFP_NOFS);
5624         if (!wc) {
5625                 btrfs_free_path(path);
5626                 return -ENOMEM;
5627         }
5628
5629         btrfs_assert_tree_locked(parent);
5630         parent_level = btrfs_header_level(parent);
5631         atomic_inc(&parent->refs);
5632         path->nodes[parent_level] = parent;
5633         path->slots[parent_level] = btrfs_header_nritems(parent);
5634
5635         btrfs_assert_tree_locked(node);
5636         level = btrfs_header_level(node);
5637         path->nodes[level] = node;
5638         path->slots[level] = 0;
5639         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
5640
5641         wc->refs[parent_level] = 1;
5642         wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5643         wc->level = level;
5644         wc->shared_level = -1;
5645         wc->stage = DROP_REFERENCE;
5646         wc->update_ref = 0;
5647         wc->keep_locks = 1;
5648         wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
5649
5650         while (1) {
5651                 wret = walk_down_tree(trans, root, path, wc);
5652                 if (wret < 0) {
5653                         ret = wret;
5654                         break;
5655                 }
5656
5657                 wret = walk_up_tree(trans, root, path, wc, parent_level);
5658                 if (wret < 0)
5659                         ret = wret;
5660                 if (wret != 0)
5661                         break;
5662         }
5663
5664         kfree(wc);
5665         btrfs_free_path(path);
5666         return ret;
5667 }
5668
5669 /*
5670  * helper to account the unused space of all the readonly block group in the
5671  * space_info. takes mirrors into account.
5672  */
5673 u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
5674 {
5675         struct btrfs_block_group *block_group;
5676         u64 free_bytes = 0;
5677         int factor;
5678
5679         /* It's df, we don't care if it's racy */
5680         if (list_empty(&sinfo->ro_bgs))
5681                 return 0;
5682
5683         spin_lock(&sinfo->lock);
5684         list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
5685                 spin_lock(&block_group->lock);
5686
5687                 if (!block_group->ro) {
5688                         spin_unlock(&block_group->lock);
5689                         continue;
5690                 }
5691
5692                 factor = btrfs_bg_type_to_factor(block_group->flags);
5693                 free_bytes += (block_group->length -
5694                                block_group->used) * factor;
5695
5696                 spin_unlock(&block_group->lock);
5697         }
5698         spin_unlock(&sinfo->lock);
5699
5700         return free_bytes;
5701 }
5702
5703 int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
5704                                    u64 start, u64 end)
5705 {
5706         return unpin_extent_range(fs_info, start, end, false);
5707 }
5708
5709 /*
5710  * It used to be that old block groups would be left around forever.
5711  * Iterating over them would be enough to trim unused space.  Since we
5712  * now automatically remove them, we also need to iterate over unallocated
5713  * space.
5714  *
5715  * We don't want a transaction for this since the discard may take a
5716  * substantial amount of time.  We don't require that a transaction be
5717  * running, but we do need to take a running transaction into account
5718  * to ensure that we're not discarding chunks that were released or
5719  * allocated in the current transaction.
5720  *
5721  * Holding the chunks lock will prevent other threads from allocating
5722  * or releasing chunks, but it won't prevent a running transaction
5723  * from committing and releasing the memory that the pending chunks
5724  * list head uses.  For that, we need to take a reference to the
5725  * transaction and hold the commit root sem.  We only need to hold
5726  * it while performing the free space search since we have already
5727  * held back allocations.
5728  */
5729 static int btrfs_trim_free_extents(struct btrfs_device *device, u64 *trimmed)
5730 {
5731         u64 start = SZ_1M, len = 0, end = 0;
5732         int ret;
5733
5734         *trimmed = 0;
5735
5736         /* Discard not supported = nothing to do. */
5737         if (!blk_queue_discard(bdev_get_queue(device->bdev)))
5738                 return 0;
5739
5740         /* Not writable = nothing to do. */
5741         if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state))
5742                 return 0;
5743
5744         /* No free space = nothing to do. */
5745         if (device->total_bytes <= device->bytes_used)
5746                 return 0;
5747
5748         ret = 0;
5749
5750         while (1) {
5751                 struct btrfs_fs_info *fs_info = device->fs_info;
5752                 u64 bytes;
5753
5754                 ret = mutex_lock_interruptible(&fs_info->chunk_mutex);
5755                 if (ret)
5756                         break;
5757
5758                 find_first_clear_extent_bit(&device->alloc_state, start,
5759                                             &start, &end,
5760                                             CHUNK_TRIMMED | CHUNK_ALLOCATED);
5761
5762                 /* Check if there are any CHUNK_* bits left */
5763                 if (start > device->total_bytes) {
5764                         WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
5765                         btrfs_warn_in_rcu(fs_info,
5766 "ignoring attempt to trim beyond device size: offset %llu length %llu device %s device size %llu",
5767                                           start, end - start + 1,
5768                                           rcu_str_deref(device->name),
5769                                           device->total_bytes);
5770                         mutex_unlock(&fs_info->chunk_mutex);
5771                         ret = 0;
5772                         break;
5773                 }
5774
5775                 /* Ensure we skip the reserved area in the first 1M */
5776                 start = max_t(u64, start, SZ_1M);
5777
5778                 /*
5779                  * If find_first_clear_extent_bit find a range that spans the
5780                  * end of the device it will set end to -1, in this case it's up
5781                  * to the caller to trim the value to the size of the device.
5782                  */
5783                 end = min(end, device->total_bytes - 1);
5784
5785                 len = end - start + 1;
5786
5787                 /* We didn't find any extents */
5788                 if (!len) {
5789                         mutex_unlock(&fs_info->chunk_mutex);
5790                         ret = 0;
5791                         break;
5792                 }
5793
5794                 ret = btrfs_issue_discard(device->bdev, start, len,
5795                                           &bytes);
5796                 if (!ret)
5797                         set_extent_bits(&device->alloc_state, start,
5798                                         start + bytes - 1,
5799                                         CHUNK_TRIMMED);
5800                 mutex_unlock(&fs_info->chunk_mutex);
5801
5802                 if (ret)
5803                         break;
5804
5805                 start += len;
5806                 *trimmed += bytes;
5807
5808                 if (fatal_signal_pending(current)) {
5809                         ret = -ERESTARTSYS;
5810                         break;
5811                 }
5812
5813                 cond_resched();
5814         }
5815
5816         return ret;
5817 }
5818
5819 /*
5820  * Trim the whole filesystem by:
5821  * 1) trimming the free space in each block group
5822  * 2) trimming the unallocated space on each device
5823  *
5824  * This will also continue trimming even if a block group or device encounters
5825  * an error.  The return value will be the last error, or 0 if nothing bad
5826  * happens.
5827  */
5828 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range)
5829 {
5830         struct btrfs_block_group *cache = NULL;
5831         struct btrfs_device *device;
5832         struct list_head *devices;
5833         u64 group_trimmed;
5834         u64 range_end = U64_MAX;
5835         u64 start;
5836         u64 end;
5837         u64 trimmed = 0;
5838         u64 bg_failed = 0;
5839         u64 dev_failed = 0;
5840         int bg_ret = 0;
5841         int dev_ret = 0;
5842         int ret = 0;
5843
5844         /*
5845          * Check range overflow if range->len is set.
5846          * The default range->len is U64_MAX.
5847          */
5848         if (range->len != U64_MAX &&
5849             check_add_overflow(range->start, range->len, &range_end))
5850                 return -EINVAL;
5851
5852         cache = btrfs_lookup_first_block_group(fs_info, range->start);
5853         for (; cache; cache = btrfs_next_block_group(cache)) {
5854                 if (cache->start >= range_end) {
5855                         btrfs_put_block_group(cache);
5856                         break;
5857                 }
5858
5859                 start = max(range->start, cache->start);
5860                 end = min(range_end, cache->start + cache->length);
5861
5862                 if (end - start >= range->minlen) {
5863                         if (!btrfs_block_group_done(cache)) {
5864                                 ret = btrfs_cache_block_group(cache, 0);
5865                                 if (ret) {
5866                                         bg_failed++;
5867                                         bg_ret = ret;
5868                                         continue;
5869                                 }
5870                                 ret = btrfs_wait_block_group_cache_done(cache);
5871                                 if (ret) {
5872                                         bg_failed++;
5873                                         bg_ret = ret;
5874                                         continue;
5875                                 }
5876                         }
5877                         ret = btrfs_trim_block_group(cache,
5878                                                      &group_trimmed,
5879                                                      start,
5880                                                      end,
5881                                                      range->minlen);
5882
5883                         trimmed += group_trimmed;
5884                         if (ret) {
5885                                 bg_failed++;
5886                                 bg_ret = ret;
5887                                 continue;
5888                         }
5889                 }
5890         }
5891
5892         if (bg_failed)
5893                 btrfs_warn(fs_info,
5894                         "failed to trim %llu block group(s), last error %d",
5895                         bg_failed, bg_ret);
5896         mutex_lock(&fs_info->fs_devices->device_list_mutex);
5897         devices = &fs_info->fs_devices->devices;
5898         list_for_each_entry(device, devices, dev_list) {
5899                 if (test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
5900                         continue;
5901
5902                 ret = btrfs_trim_free_extents(device, &group_trimmed);
5903                 if (ret) {
5904                         dev_failed++;
5905                         dev_ret = ret;
5906                         break;
5907                 }
5908
5909                 trimmed += group_trimmed;
5910         }
5911         mutex_unlock(&fs_info->fs_devices->device_list_mutex);
5912
5913         if (dev_failed)
5914                 btrfs_warn(fs_info,
5915                         "failed to trim %llu device(s), last error %d",
5916                         dev_failed, dev_ret);
5917         range->len = trimmed;
5918         if (bg_ret)
5919                 return bg_ret;
5920         return dev_ret;
5921 }