fs/btrfs/backref.c

   1 /*
   2  * Copyright (C) 2011 STRATO.  All rights reserved.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU General Public
   6  * License v2 as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public
  14  * License along with this program; if not, write to the
  15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  16  * Boston, MA 021110-1307, USA.
  17  */
  18
  19 #include <linux/vmalloc.h>
  20 #include "ctree.h"
  21 #include "disk-io.h"
  22 #include "backref.h"
  23 #include "ulist.h"
  24 #include "transaction.h"
  25 #include "delayed-ref.h"
  26 #include "locking.h"
  27
  28 struct extent_inode_elem {
  29         u64 inum;
  30         u64 offset;
  31         struct extent_inode_elem *next;
  32 };
  33
  34 static int check_extent_in_eb(struct btrfs_key *key, struct extent_buffer *eb,
  35                                 struct btrfs_file_extent_item *fi,
  36                                 u64 extent_item_pos,
  37                                 struct extent_inode_elem **eie)
  38 {
  39         u64 data_offset;
  40         u64 data_len;
  41         struct extent_inode_elem *e;
  42
  43         data_offset = btrfs_file_extent_offset(eb, fi);
  44         data_len = btrfs_file_extent_num_bytes(eb, fi);
  45
  46         if (extent_item_pos < data_offset ||
  47             extent_item_pos >= data_offset + data_len)
  48                 return 1;
  49
  50         e = kmalloc(sizeof(*e), GFP_NOFS);
  51         if (!e)
  52                 return -ENOMEM;
  53
  54         e->next = *eie;
  55         e->inum = key->objectid;
  56         e->offset = key->offset + (extent_item_pos - data_offset);
  57         *eie = e;
  58
  59         return 0;
  60 }
  61
  62 static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
  63                                 u64 extent_item_pos,
  64                                 struct extent_inode_elem **eie)
  65 {
  66         u64 disk_byte;
  67         struct btrfs_key key;
  68         struct btrfs_file_extent_item *fi;
  69         int slot;
  70         int nritems;
  71         int extent_type;
  72         int ret;
  73
  74         /*
  75          * from the shared data ref, we only have the leaf but we need
  76          * the key. thus, we must look into all items and see that we
  77          * find one (some) with a reference to our extent item.
  78          */
  79         nritems = btrfs_header_nritems(eb);
  80         for (slot = 0; slot < nritems; ++slot) {
  81                 btrfs_item_key_to_cpu(eb, &key, slot);
  82                 if (key.type != BTRFS_EXTENT_DATA_KEY)
  83                         continue;
  84                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
  85                 extent_type = btrfs_file_extent_type(eb, fi);
  86                 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
  87                         continue;
  88                 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
  89                 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
  90                 if (disk_byte != wanted_disk_byte)
  91                         continue;
  92
  93                 ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
  94                 if (ret < 0)
  95                         return ret;
  96         }
  97
  98         return 0;
  99 }
 100
 101 /*
 102  * this structure records all encountered refs on the way up to the root
 103  */
 104 struct __prelim_ref {
 105         struct list_head list;
 106         u64 root_id;
 107         struct btrfs_key key_for_search;
 108         int level;
 109         int count;
 110         struct extent_inode_elem *inode_list;
 111         u64 parent;
 112         u64 wanted_disk_byte;
 113 };
 114
 115 /*
 116  * the rules for all callers of this function are:
 117  * - obtaining the parent is the goal
 118  * - if you add a key, you must know that it is a correct key
 119  * - if you cannot add the parent or a correct key, then we will look into the
 120  *   block later to set a correct key
 121  *
 122  * delayed refs
 123  * ============
 124  *        backref type | shared | indirect | shared | indirect
 125  * information         |   tree |     tree |   data |     data
 126  * --------------------+--------+----------+--------+----------
 127  *      parent logical |    y   |     -    |    -   |     -
 128  *      key to resolve |    -   |     y    |    y   |     y
 129  *  tree block logical |    -   |     -    |    -   |     -
 130  *  root for resolving |    y   |     y    |    y   |     y
 131  *
 132  * - column 1:       we've the parent -> done
 133  * - column 2, 3, 4: we use the key to find the parent
 134  *
 135  * on disk refs (inline or keyed)
 136  * ==============================
 137  *        backref type | shared | indirect | shared | indirect
 138  * information         |   tree |     tree |   data |     data
 139  * --------------------+--------+----------+--------+----------
 140  *      parent logical |    y   |     -    |    y   |     -
 141  *      key to resolve |    -   |     -    |    -   |     y
 142  *  tree block logical |    y   |     y    |    y   |     y
 143  *  root for resolving |    -   |     y    |    y   |     y
 144  *
 145  * - column 1, 3: we've the parent -> done
 146  * - column 2:    we take the first key from the block to find the parent
 147  *                (see __add_missing_keys)
 148  * - column 4:    we use the key to find the parent
 149  *
 150  * additional information that's available but not required to find the parent
 151  * block might help in merging entries to gain some speed.
 152  */
 153
 154 static int __add_prelim_ref(struct list_head *head, u64 root_id,
 155                             struct btrfs_key *key, int level,
 156                             u64 parent, u64 wanted_disk_byte, int count)
 157 {
 158         struct __prelim_ref *ref;
 159
 160         /* in case we're adding delayed refs, we're holding the refs spinlock */
 161         ref = kmalloc(sizeof(*ref), GFP_ATOMIC);
 162         if (!ref)
 163                 return -ENOMEM;
 164
 165         ref->root_id = root_id;
 166         if (key)
 167                 ref->key_for_search = *key;
 168         else
 169                 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
 170
 171         ref->inode_list = NULL;
 172         ref->level = level;
 173         ref->count = count;
 174         ref->parent = parent;
 175         ref->wanted_disk_byte = wanted_disk_byte;
 176         list_add_tail(&ref->list, head);
 177
 178         return 0;
 179 }
 180
 181 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 182                                 struct ulist *parents, int level,
 183                                 struct btrfs_key *key_for_search, u64 time_seq,
 184                                 u64 wanted_disk_byte,
 185                                 const u64 *extent_item_pos)
 186 {
 187         int ret = 0;
 188         int slot;
 189         struct extent_buffer *eb;
 190         struct btrfs_key key;
 191         struct btrfs_file_extent_item *fi;
 192         struct extent_inode_elem *eie = NULL;
 193         u64 disk_byte;
 194
 195         if (level != 0) {
 196                 eb = path->nodes[level];
 197                 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
 198                 if (ret < 0)
 199                         return ret;
 200                 return 0;
 201         }
 202
 203         /*
 204          * We normally enter this function with the path already pointing to
 205          * the first item to check. But sometimes, we may enter it with
 206          * slot==nritems. In that case, go to the next leaf before we continue.
 207          */
 208         if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
 209                 ret = btrfs_next_old_leaf(root, path, time_seq);
 210
 211         while (!ret) {
 212                 eb = path->nodes[0];
 213                 slot = path->slots[0];
 214
 215                 btrfs_item_key_to_cpu(eb, &key, slot);
 216
 217                 if (key.objectid != key_for_search->objectid ||
 218                     key.type != BTRFS_EXTENT_DATA_KEY)
 219                         break;
 220
 221                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
 222                 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
 223
 224                 if (disk_byte == wanted_disk_byte) {
 225                         eie = NULL;
 226                         if (extent_item_pos) {
 227                                 ret = check_extent_in_eb(&key, eb, fi,
 228                                                 *extent_item_pos,
 229                                                 &eie);
 230                                 if (ret < 0)
 231                                         break;
 232                         }
 233                         if (!ret) {
 234                                 ret = ulist_add(parents, eb->start,
 235                                                 (uintptr_t)eie, GFP_NOFS);
 236                                 if (ret < 0)
 237                                         break;
 238                                 if (!extent_item_pos) {
 239                                         ret = btrfs_next_old_leaf(root, path,
 240                                                         time_seq);
 241                                         continue;
 242                                 }
 243                         }
 244                 }
 245                 ret = btrfs_next_old_item(root, path, time_seq);
 246         }
 247
 248         if (ret > 0)
 249                 ret = 0;
 250         return ret;
 251 }
 252
 253 /*
 254  * resolve an indirect backref in the form (root_id, key, level)
 255  * to a logical address
 256  */
 257 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 258                                         int search_commit_root,
 259                                         u64 time_seq,
 260                                         struct __prelim_ref *ref,
 261                                         struct ulist *parents,
 262                                         const u64 *extent_item_pos)
 263 {
 264         struct btrfs_path *path;
 265         struct btrfs_root *root;
 266         struct btrfs_key root_key;
 267         struct extent_buffer *eb;
 268         int ret = 0;
 269         int root_level;
 270         int level = ref->level;
 271
 272         path = btrfs_alloc_path();
 273         if (!path)
 274                 return -ENOMEM;
 275         path->search_commit_root = !!search_commit_root;
 276
 277         root_key.objectid = ref->root_id;
 278         root_key.type = BTRFS_ROOT_ITEM_KEY;
 279         root_key.offset = (u64)-1;
 280         root = btrfs_read_fs_root_no_name(fs_info, &root_key);
 281         if (IS_ERR(root)) {
 282                 ret = PTR_ERR(root);
 283                 goto out;
 284         }
 285
 286         root_level = btrfs_old_root_level(root, time_seq);
 287
 288         if (root_level + 1 == level)
 289                 goto out;
 290
 291         path->lowest_level = level;
 292         ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
 293         pr_debug("search slot in root %llu (level %d, ref count %d) returned "
 294                  "%d for key (%llu %u %llu)\n",
 295                  (unsigned long long)ref->root_id, level, ref->count, ret,
 296                  (unsigned long long)ref->key_for_search.objectid,
 297                  ref->key_for_search.type,
 298                  (unsigned long long)ref->key_for_search.offset);
 299         if (ret < 0)
 300                 goto out;
 301
 302         eb = path->nodes[level];
 303         while (!eb) {
 304                 if (!level) {
 305                         WARN_ON(1);
 306                         ret = 1;
 307                         goto out;
 308                 }
 309                 level--;
 310                 eb = path->nodes[level];
 311         }
 312
 313         ret = add_all_parents(root, path, parents, level, &ref->key_for_search,
 314                                 time_seq, ref->wanted_disk_byte,
 315                                 extent_item_pos);
 316 out:
 317         btrfs_free_path(path);
 318         return ret;
 319 }
 320
 321 /*
 322  * resolve all indirect backrefs from the list
 323  */
 324 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
 325                                    int search_commit_root, u64 time_seq,
 326                                    struct list_head *head,
 327                                    const u64 *extent_item_pos)
 328 {
 329         int err;
 330         int ret = 0;
 331         struct __prelim_ref *ref;
 332         struct __prelim_ref *ref_safe;
 333         struct __prelim_ref *new_ref;
 334         struct ulist *parents;
 335         struct ulist_node *node;
 336         struct ulist_iterator uiter;
 337
 338         parents = ulist_alloc(GFP_NOFS);
 339         if (!parents)
 340                 return -ENOMEM;
 341
 342         /*
 343          * _safe allows us to insert directly after the current item without
 344          * iterating over the newly inserted items.
 345          * we're also allowed to re-assign ref during iteration.
 346          */
 347         list_for_each_entry_safe(ref, ref_safe, head, list) {
 348                 if (ref->parent)        /* already direct */
 349                         continue;
 350                 if (ref->count == 0)
 351                         continue;
 352                 err = __resolve_indirect_ref(fs_info, search_commit_root,
 353                                              time_seq, ref, parents,
 354                                              extent_item_pos);
 355                 if (err)
 356                         continue;
 357
 358                 /* we put the first parent into the ref at hand */
 359                 ULIST_ITER_INIT(&uiter);
 360                 node = ulist_next(parents, &uiter);
 361                 ref->parent = node ? node->val : 0;
 362                 ref->inode_list = node ?
 363                         (struct extent_inode_elem *)(uintptr_t)node->aux : 0;
 364
 365                 /* additional parents require new refs being added here */
 366                 while ((node = ulist_next(parents, &uiter))) {
 367                         new_ref = kmalloc(sizeof(*new_ref), GFP_NOFS);
 368                         if (!new_ref) {
 369                                 ret = -ENOMEM;
 370                                 break;
 371                         }
 372                         memcpy(new_ref, ref, sizeof(*ref));
 373                         new_ref->parent = node->val;
 374                         new_ref->inode_list = (struct extent_inode_elem *)
 375                                                         (uintptr_t)node->aux;
 376                         list_add(&new_ref->list, &ref->list);
 377                 }
 378                 ulist_reinit(parents);
 379         }
 380
 381         ulist_free(parents);
 382         return ret;
 383 }
 384
 385 static inline int ref_for_same_block(struct __prelim_ref *ref1,
 386                                      struct __prelim_ref *ref2)
 387 {
 388         if (ref1->level != ref2->level)
 389                 return 0;
 390         if (ref1->root_id != ref2->root_id)
 391                 return 0;
 392         if (ref1->key_for_search.type != ref2->key_for_search.type)
 393                 return 0;
 394         if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
 395                 return 0;
 396         if (ref1->key_for_search.offset != ref2->key_for_search.offset)
 397                 return 0;
 398         if (ref1->parent != ref2->parent)
 399                 return 0;
 400
 401         return 1;
 402 }
 403
 404 /*
 405  * read tree blocks and add keys where required.
 406  */
 407 static int __add_missing_keys(struct btrfs_fs_info *fs_info,
 408                               struct list_head *head)
 409 {
 410         struct list_head *pos;
 411         struct extent_buffer *eb;
 412
 413         list_for_each(pos, head) {
 414                 struct __prelim_ref *ref;
 415                 ref = list_entry(pos, struct __prelim_ref, list);
 416
 417                 if (ref->parent)
 418                         continue;
 419                 if (ref->key_for_search.type)
 420                         continue;
 421                 BUG_ON(!ref->wanted_disk_byte);
 422                 eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
 423                                      fs_info->tree_root->leafsize, 0);
 424                 BUG_ON(!eb);
 425                 btrfs_tree_read_lock(eb);
 426                 if (btrfs_header_level(eb) == 0)
 427                         btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
 428                 else
 429                         btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
 430                 btrfs_tree_read_unlock(eb);
 431                 free_extent_buffer(eb);
 432         }
 433         return 0;
 434 }
 435
 436 /*
 437  * merge two lists of backrefs and adjust counts accordingly
 438  *
 439  * mode = 1: merge identical keys, if key is set
 440  *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
 441  *           additionally, we could even add a key range for the blocks we
 442  *           looked into to merge even more (-> replace unresolved refs by those
 443  *           having a parent).
 444  * mode = 2: merge identical parents
 445  */
 446 static int __merge_refs(struct list_head *head, int mode)
 447 {
 448         struct list_head *pos1;
 449
 450         list_for_each(pos1, head) {
 451                 struct list_head *n2;
 452                 struct list_head *pos2;
 453                 struct __prelim_ref *ref1;
 454
 455                 ref1 = list_entry(pos1, struct __prelim_ref, list);
 456
 457                 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
 458                      pos2 = n2, n2 = pos2->next) {
 459                         struct __prelim_ref *ref2;
 460                         struct __prelim_ref *xchg;
 461                         struct extent_inode_elem *eie;
 462
 463                         ref2 = list_entry(pos2, struct __prelim_ref, list);
 464
 465                         if (mode == 1) {
 466                                 if (!ref_for_same_block(ref1, ref2))
 467                                         continue;
 468                                 if (!ref1->parent && ref2->parent) {
 469                                         xchg = ref1;
 470                                         ref1 = ref2;
 471                                         ref2 = xchg;
 472                                 }
 473                         } else {
 474                                 if (ref1->parent != ref2->parent)
 475                                         continue;
 476                         }
 477
 478                         eie = ref1->inode_list;
 479                         while (eie && eie->next)
 480                                 eie = eie->next;
 481                         if (eie)
 482                                 eie->next = ref2->inode_list;
 483                         else
 484                                 ref1->inode_list = ref2->inode_list;
 485                         ref1->count += ref2->count;
 486
 487                         list_del(&ref2->list);
 488                         kfree(ref2);
 489                 }
 490
 491         }
 492         return 0;
 493 }
 494
 495 /*
 496  * add all currently queued delayed refs from this head whose seq nr is
 497  * smaller or equal that seq to the list
 498  */
 499 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
 500                               struct list_head *prefs)
 501 {
 502         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
 503         struct rb_node *n = &head->node.rb_node;
 504         struct btrfs_key key;
 505         struct btrfs_key op_key = {0};
 506         int sgn;
 507         int ret = 0;
 508
 509         if (extent_op && extent_op->update_key)
 510                 btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
 511
 512         while ((n = rb_prev(n))) {
 513                 struct btrfs_delayed_ref_node *node;
 514                 node = rb_entry(n, struct btrfs_delayed_ref_node,
 515                                 rb_node);
 516                 if (node->bytenr != head->node.bytenr)
 517                         break;
 518                 WARN_ON(node->is_head);
 519
 520                 if (node->seq > seq)
 521                         continue;
 522
 523                 switch (node->action) {
 524                 case BTRFS_ADD_DELAYED_EXTENT:
 525                 case BTRFS_UPDATE_DELAYED_HEAD:
 526                         WARN_ON(1);
 527                         continue;
 528                 case BTRFS_ADD_DELAYED_REF:
 529                         sgn = 1;
 530                         break;
 531                 case BTRFS_DROP_DELAYED_REF:
 532                         sgn = -1;
 533                         break;
 534                 default:
 535                         BUG_ON(1);
 536                 }
 537                 switch (node->type) {
 538                 case BTRFS_TREE_BLOCK_REF_KEY: {
 539                         struct btrfs_delayed_tree_ref *ref;
 540
 541                         ref = btrfs_delayed_node_to_tree_ref(node);
 542                         ret = __add_prelim_ref(prefs, ref->root, &op_key,
 543                                                ref->level + 1, 0, node->bytenr,
 544                                                node->ref_mod * sgn);
 545                         break;
 546                 }
 547                 case BTRFS_SHARED_BLOCK_REF_KEY: {
 548                         struct btrfs_delayed_tree_ref *ref;
 549
 550                         ref = btrfs_delayed_node_to_tree_ref(node);
 551                         ret = __add_prelim_ref(prefs, ref->root, NULL,
 552                                                ref->level + 1, ref->parent,
 553                                                node->bytenr,
 554                                                node->ref_mod * sgn);
 555                         break;
 556                 }
 557                 case BTRFS_EXTENT_DATA_REF_KEY: {
 558                         struct btrfs_delayed_data_ref *ref;
 559                         ref = btrfs_delayed_node_to_data_ref(node);
 560
 561                         key.objectid = ref->objectid;
 562                         key.type = BTRFS_EXTENT_DATA_KEY;
 563                         key.offset = ref->offset;
 564                         ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
 565                                                node->bytenr,
 566                                                node->ref_mod * sgn);
 567                         break;
 568                 }
 569                 case BTRFS_SHARED_DATA_REF_KEY: {
 570                         struct btrfs_delayed_data_ref *ref;
 571
 572                         ref = btrfs_delayed_node_to_data_ref(node);
 573
 574                         key.objectid = ref->objectid;
 575                         key.type = BTRFS_EXTENT_DATA_KEY;
 576                         key.offset = ref->offset;
 577                         ret = __add_prelim_ref(prefs, ref->root, &key, 0,
 578                                                ref->parent, node->bytenr,
 579                                                node->ref_mod * sgn);
 580                         break;
 581                 }
 582                 default:
 583                         WARN_ON(1);
 584                 }
 585                 BUG_ON(ret);
 586         }
 587
 588         return 0;
 589 }
 590
 591 /*
 592  * add all inline backrefs for bytenr to the list
 593  */
 594 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
 595                              struct btrfs_path *path, u64 bytenr,
 596                              int *info_level, struct list_head *prefs)
 597 {
 598         int ret = 0;
 599         int slot;
 600         struct extent_buffer *leaf;
 601         struct btrfs_key key;
 602         unsigned long ptr;
 603         unsigned long end;
 604         struct btrfs_extent_item *ei;
 605         u64 flags;
 606         u64 item_size;
 607
 608         /*
 609          * enumerate all inline refs
 610          */
 611         leaf = path->nodes[0];
 612         slot = path->slots[0];
 613
 614         item_size = btrfs_item_size_nr(leaf, slot);
 615         BUG_ON(item_size < sizeof(*ei));
 616
 617         ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
 618         flags = btrfs_extent_flags(leaf, ei);
 619
 620         ptr = (unsigned long)(ei + 1);
 621         end = (unsigned long)ei + item_size;
 622
 623         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
 624                 struct btrfs_tree_block_info *info;
 625
 626                 info = (struct btrfs_tree_block_info *)ptr;
 627                 *info_level = btrfs_tree_block_level(leaf, info);
 628                 ptr += sizeof(struct btrfs_tree_block_info);
 629                 BUG_ON(ptr > end);
 630         } else {
 631                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
 632         }
 633
 634         while (ptr < end) {
 635                 struct btrfs_extent_inline_ref *iref;
 636                 u64 offset;
 637                 int type;
 638
 639                 iref = (struct btrfs_extent_inline_ref *)ptr;
 640                 type = btrfs_extent_inline_ref_type(leaf, iref);
 641                 offset = btrfs_extent_inline_ref_offset(leaf, iref);
 642
 643                 switch (type) {
 644                 case BTRFS_SHARED_BLOCK_REF_KEY:
 645                         ret = __add_prelim_ref(prefs, 0, NULL,
 646                                                 *info_level + 1, offset,
 647                                                 bytenr, 1);
 648                         break;
 649                 case BTRFS_SHARED_DATA_REF_KEY: {
 650                         struct btrfs_shared_data_ref *sdref;
 651                         int count;
 652
 653                         sdref = (struct btrfs_shared_data_ref *)(iref + 1);
 654                         count = btrfs_shared_data_ref_count(leaf, sdref);
 655                         ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
 656                                                bytenr, count);
 657                         break;
 658                 }
 659                 case BTRFS_TREE_BLOCK_REF_KEY:
 660                         ret = __add_prelim_ref(prefs, offset, NULL,
 661                                                *info_level + 1, 0,
 662                                                bytenr, 1);
 663                         break;
 664                 case BTRFS_EXTENT_DATA_REF_KEY: {
 665                         struct btrfs_extent_data_ref *dref;
 666                         int count;
 667                         u64 root;
 668
 669                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
 670                         count = btrfs_extent_data_ref_count(leaf, dref);
 671                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
 672                                                                       dref);
 673                         key.type = BTRFS_EXTENT_DATA_KEY;
 674                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
 675                         root = btrfs_extent_data_ref_root(leaf, dref);
 676                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
 677                                                bytenr, count);
 678                         break;
 679                 }
 680                 default:
 681                         WARN_ON(1);
 682                 }
 683                 BUG_ON(ret);
 684                 ptr += btrfs_extent_inline_ref_size(type);
 685         }
 686
 687         return 0;
 688 }
 689
 690 /*
 691  * add all non-inline backrefs for bytenr to the list
 692  */
 693 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
 694                             struct btrfs_path *path, u64 bytenr,
 695                             int info_level, struct list_head *prefs)
 696 {
 697         struct btrfs_root *extent_root = fs_info->extent_root;
 698         int ret;
 699         int slot;
 700         struct extent_buffer *leaf;
 701         struct btrfs_key key;
 702
 703         while (1) {
 704                 ret = btrfs_next_item(extent_root, path);
 705                 if (ret < 0)
 706                         break;
 707                 if (ret) {
 708                         ret = 0;
 709                         break;
 710                 }
 711
 712                 slot = path->slots[0];
 713                 leaf = path->nodes[0];
 714                 btrfs_item_key_to_cpu(leaf, &key, slot);
 715
 716                 if (key.objectid != bytenr)
 717                         break;
 718                 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
 719                         continue;
 720                 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
 721                         break;
 722
 723                 switch (key.type) {
 724                 case BTRFS_SHARED_BLOCK_REF_KEY:
 725                         ret = __add_prelim_ref(prefs, 0, NULL,
 726                                                 info_level + 1, key.offset,
 727                                                 bytenr, 1);
 728                         break;
 729                 case BTRFS_SHARED_DATA_REF_KEY: {
 730                         struct btrfs_shared_data_ref *sdref;
 731                         int count;
 732
 733                         sdref = btrfs_item_ptr(leaf, slot,
 734                                               struct btrfs_shared_data_ref);
 735                         count = btrfs_shared_data_ref_count(leaf, sdref);
 736                         ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
 737                                                 bytenr, count);
 738                         break;
 739                 }
 740                 case BTRFS_TREE_BLOCK_REF_KEY:
 741                         ret = __add_prelim_ref(prefs, key.offset, NULL,
 742                                                info_level + 1, 0,
 743                                                bytenr, 1);
 744                         break;
 745                 case BTRFS_EXTENT_DATA_REF_KEY: {
 746                         struct btrfs_extent_data_ref *dref;
 747                         int count;
 748                         u64 root;
 749
 750                         dref = btrfs_item_ptr(leaf, slot,
 751                                               struct btrfs_extent_data_ref);
 752                         count = btrfs_extent_data_ref_count(leaf, dref);
 753                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
 754                                                                       dref);
 755                         key.type = BTRFS_EXTENT_DATA_KEY;
 756                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
 757                         root = btrfs_extent_data_ref_root(leaf, dref);
 758                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
 759                                                bytenr, count);
 760                         break;
 761                 }
 762                 default:
 763                         WARN_ON(1);
 764                 }
 765                 BUG_ON(ret);
 766         }
 767
 768         return ret;
 769 }
 770
 771 /*
 772  * this adds all existing backrefs (inline backrefs, backrefs and delayed
 773  * refs) for the given bytenr to the refs list, merges duplicates and resolves
 774  * indirect refs to their parent bytenr.
 775  * When roots are found, they're added to the roots list
 776  *
 777  * FIXME some caching might speed things up
 778  */
 779 static int find_parent_nodes(struct btrfs_trans_handle *trans,
 780                              struct btrfs_fs_info *fs_info, u64 bytenr,
 781                              u64 time_seq, struct ulist *refs,
 782                              struct ulist *roots, const u64 *extent_item_pos)
 783 {
 784         struct btrfs_key key;
 785         struct btrfs_path *path;
 786         struct btrfs_delayed_ref_root *delayed_refs = NULL;
 787         struct btrfs_delayed_ref_head *head;
 788         int info_level = 0;
 789         int ret;
 790         int search_commit_root = (trans == BTRFS_BACKREF_SEARCH_COMMIT_ROOT);
 791         struct list_head prefs_delayed;
 792         struct list_head prefs;
 793         struct __prelim_ref *ref;
 794
 795         INIT_LIST_HEAD(&prefs);
 796         INIT_LIST_HEAD(&prefs_delayed);
 797
 798         key.objectid = bytenr;
 799         key.type = BTRFS_EXTENT_ITEM_KEY;
 800         key.offset = (u64)-1;
 801
 802         path = btrfs_alloc_path();
 803         if (!path)
 804                 return -ENOMEM;
 805         path->search_commit_root = !!search_commit_root;
 806
 807         /*
 808          * grab both a lock on the path and a lock on the delayed ref head.
 809          * We need both to get a consistent picture of how the refs look
 810          * at a specified point in time
 811          */
 812 again:
 813         head = NULL;
 814
 815         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
 816         if (ret < 0)
 817                 goto out;
 818         BUG_ON(ret == 0);
 819
 820         if (trans != BTRFS_BACKREF_SEARCH_COMMIT_ROOT) {
 821                 /*
 822                  * look if there are updates for this ref queued and lock the
 823                  * head
 824                  */
 825                 delayed_refs = &trans->transaction->delayed_refs;
 826                 spin_lock(&delayed_refs->lock);
 827                 head = btrfs_find_delayed_ref_head(trans, bytenr);
 828                 if (head) {
 829                         if (!mutex_trylock(&head->mutex)) {
 830                                 atomic_inc(&head->node.refs);
 831                                 spin_unlock(&delayed_refs->lock);
 832
 833                                 btrfs_release_path(path);
 834
 835                                 /*
 836                                  * Mutex was contended, block until it's
 837                                  * released and try again
 838                                  */
 839                                 mutex_lock(&head->mutex);
 840                                 mutex_unlock(&head->mutex);
 841                                 btrfs_put_delayed_ref(&head->node);
 842                                 goto again;
 843                         }
 844                         ret = __add_delayed_refs(head, time_seq,
 845                                                  &prefs_delayed);
 846                         mutex_unlock(&head->mutex);
 847                         if (ret) {
 848                                 spin_unlock(&delayed_refs->lock);
 849                                 goto out;
 850                         }
 851                 }
 852                 spin_unlock(&delayed_refs->lock);
 853         }
 854
 855         if (path->slots[0]) {
 856                 struct extent_buffer *leaf;
 857                 int slot;
 858
 859                 path->slots[0]--;
 860                 leaf = path->nodes[0];
 861                 slot = path->slots[0];
 862                 btrfs_item_key_to_cpu(leaf, &key, slot);
 863                 if (key.objectid == bytenr &&
 864                     key.type == BTRFS_EXTENT_ITEM_KEY) {
 865                         ret = __add_inline_refs(fs_info, path, bytenr,
 866                                                 &info_level, &prefs);
 867                         if (ret)
 868                                 goto out;
 869                         ret = __add_keyed_refs(fs_info, path, bytenr,
 870                                                info_level, &prefs);
 871                         if (ret)
 872                                 goto out;
 873                 }
 874         }
 875         btrfs_release_path(path);
 876
 877         list_splice_init(&prefs_delayed, &prefs);
 878
 879         ret = __add_missing_keys(fs_info, &prefs);
 880         if (ret)
 881                 goto out;
 882
 883         ret = __merge_refs(&prefs, 1);
 884         if (ret)
 885                 goto out;
 886
 887         ret = __resolve_indirect_refs(fs_info, search_commit_root, time_seq,
 888                                       &prefs, extent_item_pos);
 889         if (ret)
 890                 goto out;
 891
 892         ret = __merge_refs(&prefs, 2);
 893         if (ret)
 894                 goto out;
 895
 896         while (!list_empty(&prefs)) {
 897                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
 898                 list_del(&ref->list);
 899                 WARN_ON(ref->count < 0);
 900                 if (ref->count && ref->root_id && ref->parent == 0) {
 901                         /* no parent == root of tree */
 902                         ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
 903                         BUG_ON(ret < 0);
 904                 }
 905                 if (ref->count && ref->parent) {
 906                         struct extent_inode_elem *eie = NULL;
 907                         if (extent_item_pos && !ref->inode_list) {
 908                                 u32 bsz;
 909                                 struct extent_buffer *eb;
 910                                 bsz = btrfs_level_size(fs_info->extent_root,
 911                                                         info_level);
 912                                 eb = read_tree_block(fs_info->extent_root,
 913                                                            ref->parent, bsz, 0);
 914                                 BUG_ON(!eb);
 915                                 ret = find_extent_in_eb(eb, bytenr,
 916                                                         *extent_item_pos, &eie);
 917                                 ref->inode_list = eie;
 918                                 free_extent_buffer(eb);
 919                         }
 920                         ret = ulist_add_merge(refs, ref->parent,
 921                                               (uintptr_t)ref->inode_list,
 922                                               (u64 *)&eie, GFP_NOFS);
 923                         if (!ret && extent_item_pos) {
 924                                 /*
 925                                  * we've recorded that parent, so we must extend
 926                                  * its inode list here
 927                                  */
 928                                 BUG_ON(!eie);
 929                                 while (eie->next)
 930                                         eie = eie->next;
 931                                 eie->next = ref->inode_list;
 932                         }
 933                         BUG_ON(ret < 0);
 934                 }
 935                 kfree(ref);
 936         }
 937
 938 out:
 939         btrfs_free_path(path);
 940         while (!list_empty(&prefs)) {
 941                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
 942                 list_del(&ref->list);
 943                 kfree(ref);
 944         }
 945         while (!list_empty(&prefs_delayed)) {
 946                 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
 947                                        list);
 948                 list_del(&ref->list);
 949                 kfree(ref);
 950         }
 951
 952         return ret;
 953 }
 954
 955 static void free_leaf_list(struct ulist *blocks)
 956 {
 957         struct ulist_node *node = NULL;
 958         struct extent_inode_elem *eie;
 959         struct extent_inode_elem *eie_next;
 960         struct ulist_iterator uiter;
 961
 962         ULIST_ITER_INIT(&uiter);
 963         while ((node = ulist_next(blocks, &uiter))) {
 964                 if (!node->aux)
 965                         continue;
 966                 eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
 967                 for (; eie; eie = eie_next) {
 968                         eie_next = eie->next;
 969                         kfree(eie);
 970                 }
 971                 node->aux = 0;
 972         }
 973
 974         ulist_free(blocks);
 975 }
 976
 977 /*
 978  * Finds all leafs with a reference to the specified combination of bytenr and
 979  * offset. key_list_head will point to a list of corresponding keys (caller must
 980  * free each list element). The leafs will be stored in the leafs ulist, which
 981  * must be freed with ulist_free.
 982  *
 983  * returns 0 on success, <0 on error
 984  */
 985 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
 986                                 struct btrfs_fs_info *fs_info, u64 bytenr,
 987                                 u64 time_seq, struct ulist **leafs,
 988                                 const u64 *extent_item_pos)
 989 {
 990         struct ulist *tmp;
 991         int ret;
 992
 993         tmp = ulist_alloc(GFP_NOFS);
 994         if (!tmp)
 995                 return -ENOMEM;
 996         *leafs = ulist_alloc(GFP_NOFS);
 997         if (!*leafs) {
 998                 ulist_free(tmp);
 999                 return -ENOMEM;
1000         }
1001
1002         ret = find_parent_nodes(trans, fs_info, bytenr,
1003                                 time_seq, *leafs, tmp, extent_item_pos);
1004         ulist_free(tmp);
1005
1006         if (ret < 0 && ret != -ENOENT) {
1007                 free_leaf_list(*leafs);
1008                 return ret;
1009         }
1010
1011         return 0;
1012 }
1013
1014 /*
1015  * walk all backrefs for a given extent to find all roots that reference this
1016  * extent. Walking a backref means finding all extents that reference this
1017  * extent and in turn walk the backrefs of those, too. Naturally this is a
1018  * recursive process, but here it is implemented in an iterative fashion: We
1019  * find all referencing extents for the extent in question and put them on a
1020  * list. In turn, we find all referencing extents for those, further appending
1021  * to the list. The way we iterate the list allows adding more elements after
1022  * the current while iterating. The process stops when we reach the end of the
1023  * list. Found roots are added to the roots list.
1024  *
1025  * returns 0 on success, < 0 on error.
1026  */
1027 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1028                                 struct btrfs_fs_info *fs_info, u64 bytenr,
1029                                 u64 time_seq, struct ulist **roots)
1030 {
1031         struct ulist *tmp;
1032         struct ulist_node *node = NULL;
1033         struct ulist_iterator uiter;
1034         int ret;
1035
1036         tmp = ulist_alloc(GFP_NOFS);
1037         if (!tmp)
1038                 return -ENOMEM;
1039         *roots = ulist_alloc(GFP_NOFS);
1040         if (!*roots) {
1041                 ulist_free(tmp);
1042                 return -ENOMEM;
1043         }
1044
1045         ULIST_ITER_INIT(&uiter);
1046         while (1) {
1047                 ret = find_parent_nodes(trans, fs_info, bytenr,
1048                                         time_seq, tmp, *roots, NULL);
1049                 if (ret < 0 && ret != -ENOENT) {
1050                         ulist_free(tmp);
1051                         ulist_free(*roots);
1052                         return ret;
1053                 }
1054                 node = ulist_next(tmp, &uiter);
1055                 if (!node)
1056                         break;
1057                 bytenr = node->val;
1058         }
1059
1060         ulist_free(tmp);
1061         return 0;
1062 }
1063
1064
1065 static int __inode_info(u64 inum, u64 ioff, u8 key_type,
1066                         struct btrfs_root *fs_root, struct btrfs_path *path,
1067                         struct btrfs_key *found_key)
1068 {
1069         int ret;
1070         struct btrfs_key key;
1071         struct extent_buffer *eb;
1072
1073         key.type = key_type;
1074         key.objectid = inum;
1075         key.offset = ioff;
1076
1077         ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
1078         if (ret < 0)
1079                 return ret;
1080
1081         eb = path->nodes[0];
1082         if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
1083                 ret = btrfs_next_leaf(fs_root, path);
1084                 if (ret)
1085                         return ret;
1086                 eb = path->nodes[0];
1087         }
1088
1089         btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
1090         if (found_key->type != key.type || found_key->objectid != key.objectid)
1091                 return 1;
1092
1093         return 0;
1094 }
1095
1096 /*
1097  * this makes the path point to (inum INODE_ITEM ioff)
1098  */
1099 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1100                         struct btrfs_path *path)
1101 {
1102         struct btrfs_key key;
1103         return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
1104                                 &key);
1105 }
1106
1107 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1108                                 struct btrfs_path *path,
1109                                 struct btrfs_key *found_key)
1110 {
1111         return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
1112                                 found_key);
1113 }
1114
1115 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1116                           u64 start_off, struct btrfs_path *path,
1117                           struct btrfs_inode_extref **ret_extref,
1118                           u64 *found_off)
1119 {
1120         int ret, slot;
1121         struct btrfs_key key;
1122         struct btrfs_key found_key;
1123         struct btrfs_inode_extref *extref;
1124         struct extent_buffer *leaf;
1125         unsigned long ptr;
1126
1127         key.objectid = inode_objectid;
1128         btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
1129         key.offset = start_off;
1130
1131         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1132         if (ret < 0)
1133                 return ret;
1134
1135         while (1) {
1136                 leaf = path->nodes[0];
1137                 slot = path->slots[0];
1138                 if (slot >= btrfs_header_nritems(leaf)) {
1139                         /*
1140                          * If the item at offset is not found,
1141                          * btrfs_search_slot will point us to the slot
1142                          * where it should be inserted. In our case
1143                          * that will be the slot directly before the
1144                          * next INODE_REF_KEY_V2 item. In the case
1145                          * that we're pointing to the last slot in a
1146                          * leaf, we must move one leaf over.
1147                          */
1148                         ret = btrfs_next_leaf(root, path);
1149                         if (ret) {
1150                                 if (ret >= 1)
1151                                         ret = -ENOENT;
1152                                 break;
1153                         }
1154                         continue;
1155                 }
1156
1157                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1158
1159                 /*
1160                  * Check that we're still looking at an extended ref key for
1161                  * this particular objectid. If we have different
1162                  * objectid or type then there are no more to be found
1163                  * in the tree and we can exit.
1164                  */
1165                 ret = -ENOENT;
1166                 if (found_key.objectid != inode_objectid)
1167                         break;
1168                 if (btrfs_key_type(&found_key) != BTRFS_INODE_EXTREF_KEY)
1169                         break;
1170
1171                 ret = 0;
1172                 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1173                 extref = (struct btrfs_inode_extref *)ptr;
1174                 *ret_extref = extref;
1175                 if (found_off)
1176                         *found_off = found_key.offset;
1177                 break;
1178         }
1179
1180         return ret;
1181 }
1182
1183 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1184                         u32 name_len, unsigned long name_off,
1185                         struct extent_buffer *eb_in, u64 parent,
1186                         char *dest, u32 size)
1187 {
1188         int slot;
1189         u64 next_inum;
1190         int ret;
1191         s64 bytes_left = ((s64)size) - 1;
1192         struct extent_buffer *eb = eb_in;
1193         struct btrfs_key found_key;
1194         int leave_spinning = path->leave_spinning;
1195         struct btrfs_inode_ref *iref;
1196
1197         if (bytes_left >= 0)
1198                 dest[bytes_left] = '\0';
1199
1200         path->leave_spinning = 1;
1201         while (1) {
1202                 bytes_left -= name_len;
1203                 if (bytes_left >= 0)
1204                         read_extent_buffer(eb, dest + bytes_left,
1205                                            name_off, name_len);
1206                 if (eb != eb_in) {
1207                         btrfs_tree_read_unlock_blocking(eb);
1208                         free_extent_buffer(eb);
1209                 }
1210                 ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
1211                 if (ret > 0)
1212                         ret = -ENOENT;
1213                 if (ret)
1214                         break;
1215
1216                 next_inum = found_key.offset;
1217
1218                 /* regular exit ahead */
1219                 if (parent == next_inum)
1220                         break;
1221
1222                 slot = path->slots[0];
1223                 eb = path->nodes[0];
1224                 /* make sure we can use eb after releasing the path */
1225                 if (eb != eb_in) {
1226                         atomic_inc(&eb->refs);
1227                         btrfs_tree_read_lock(eb);
1228                         btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1229                 }
1230                 btrfs_release_path(path);
1231                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1232
1233                 name_len = btrfs_inode_ref_name_len(eb, iref);
1234                 name_off = (unsigned long)(iref + 1);
1235
1236                 parent = next_inum;
1237                 --bytes_left;
1238                 if (bytes_left >= 0)
1239                         dest[bytes_left] = '/';
1240         }
1241
1242         btrfs_release_path(path);
1243         path->leave_spinning = leave_spinning;
1244
1245         if (ret)
1246                 return ERR_PTR(ret);
1247
1248         return dest + bytes_left;
1249 }
1250
1251 /*
1252  * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
1253  * of the path are separated by '/' and the path is guaranteed to be
1254  * 0-terminated. the path is only given within the current file system.
1255  * Therefore, it never starts with a '/'. the caller is responsible to provide
1256  * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1257  * the start point of the resulting string is returned. this pointer is within
1258  * dest, normally.
1259  * in case the path buffer would overflow, the pointer is decremented further
1260  * as if output was written to the buffer, though no more output is actually
1261  * generated. that way, the caller can determine how much space would be
1262  * required for the path to fit into the buffer. in that case, the returned
1263  * value will be smaller than dest. callers must check this!
1264  */
1265 char *btrfs_iref_to_path(struct btrfs_root *fs_root,
1266                          struct btrfs_path *path,
1267                          struct btrfs_inode_ref *iref,
1268                          struct extent_buffer *eb_in, u64 parent,
1269                          char *dest, u32 size)
1270 {
1271         return btrfs_ref_to_path(fs_root, path,
1272                                  btrfs_inode_ref_name_len(eb_in, iref),
1273                                  (unsigned long)(iref + 1),
1274                                  eb_in, parent, dest, size);
1275 }
1276
1277 /*
1278  * this makes the path point to (logical EXTENT_ITEM *)
1279  * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1280  * tree blocks and <0 on error.
1281  */
1282 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1283                         struct btrfs_path *path, struct btrfs_key *found_key,
1284                         u64 *flags_ret)
1285 {
1286         int ret;
1287         u64 flags;
1288         u32 item_size;
1289         struct extent_buffer *eb;
1290         struct btrfs_extent_item *ei;
1291         struct btrfs_key key;
1292
1293         key.type = BTRFS_EXTENT_ITEM_KEY;
1294         key.objectid = logical;
1295         key.offset = (u64)-1;
1296
1297         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1298         if (ret < 0)
1299                 return ret;
1300         ret = btrfs_previous_item(fs_info->extent_root, path,
1301                                         0, BTRFS_EXTENT_ITEM_KEY);
1302         if (ret < 0)
1303                 return ret;
1304
1305         btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
1306         if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
1307             found_key->objectid > logical ||
1308             found_key->objectid + found_key->offset <= logical) {
1309                 pr_debug("logical %llu is not within any extent\n",
1310                          (unsigned long long)logical);
1311                 return -ENOENT;
1312         }
1313
1314         eb = path->nodes[0];
1315         item_size = btrfs_item_size_nr(eb, path->slots[0]);
1316         BUG_ON(item_size < sizeof(*ei));
1317
1318         ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1319         flags = btrfs_extent_flags(eb, ei);
1320
1321         pr_debug("logical %llu is at position %llu within the extent (%llu "
1322                  "EXTENT_ITEM %llu) flags %#llx size %u\n",
1323                  (unsigned long long)logical,
1324                  (unsigned long long)(logical - found_key->objectid),
1325                  (unsigned long long)found_key->objectid,
1326                  (unsigned long long)found_key->offset,
1327                  (unsigned long long)flags, item_size);
1328
1329         WARN_ON(!flags_ret);
1330         if (flags_ret) {
1331                 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1332                         *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1333                 else if (flags & BTRFS_EXTENT_FLAG_DATA)
1334                         *flags_ret = BTRFS_EXTENT_FLAG_DATA;
1335                 else
1336                         BUG_ON(1);
1337                 return 0;
1338         }
1339
1340         return -EIO;
1341 }
1342
1343 /*
1344  * helper function to iterate extent inline refs. ptr must point to a 0 value
1345  * for the first call and may be modified. it is used to track state.
1346  * if more refs exist, 0 is returned and the next call to
1347  * __get_extent_inline_ref must pass the modified ptr parameter to get the
1348  * next ref. after the last ref was processed, 1 is returned.
1349  * returns <0 on error
1350  */
1351 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1352                                 struct btrfs_extent_item *ei, u32 item_size,
1353                                 struct btrfs_extent_inline_ref **out_eiref,
1354                                 int *out_type)
1355 {
1356         unsigned long end;
1357         u64 flags;
1358         struct btrfs_tree_block_info *info;
1359
1360         if (!*ptr) {
1361                 /* first call */
1362                 flags = btrfs_extent_flags(eb, ei);
1363                 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1364                         info = (struct btrfs_tree_block_info *)(ei + 1);
1365                         *out_eiref =
1366                                 (struct btrfs_extent_inline_ref *)(info + 1);
1367                 } else {
1368                         *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1369                 }
1370                 *ptr = (unsigned long)*out_eiref;
1371                 if ((void *)*ptr >= (void *)ei + item_size)
1372                         return -ENOENT;
1373         }
1374
1375         end = (unsigned long)ei + item_size;
1376         *out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1377         *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1378
1379         *ptr += btrfs_extent_inline_ref_size(*out_type);
1380         WARN_ON(*ptr > end);
1381         if (*ptr == end)
1382                 return 1; /* last */
1383
1384         return 0;
1385 }
1386
1387 /*
1388  * reads the tree block backref for an extent. tree level and root are returned
1389  * through out_level and out_root. ptr must point to a 0 value for the first
1390  * call and may be modified (see __get_extent_inline_ref comment).
1391  * returns 0 if data was provided, 1 if there was no more data to provide or
1392  * <0 on error.
1393  */
1394 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1395                                 struct btrfs_extent_item *ei, u32 item_size,
1396                                 u64 *out_root, u8 *out_level)
1397 {
1398         int ret;
1399         int type;
1400         struct btrfs_tree_block_info *info;
1401         struct btrfs_extent_inline_ref *eiref;
1402
1403         if (*ptr == (unsigned long)-1)
1404                 return 1;
1405
1406         while (1) {
1407                 ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1408                                                 &eiref, &type);
1409                 if (ret < 0)
1410                         return ret;
1411
1412                 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1413                     type == BTRFS_SHARED_BLOCK_REF_KEY)
1414                         break;
1415
1416                 if (ret == 1)
1417                         return 1;
1418         }
1419
1420         /* we can treat both ref types equally here */
1421         info = (struct btrfs_tree_block_info *)(ei + 1);
1422         *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1423         *out_level = btrfs_tree_block_level(eb, info);
1424
1425         if (ret == 1)
1426                 *ptr = (unsigned long)-1;
1427
1428         return 0;
1429 }
1430
1431 static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1432                                 u64 root, u64 extent_item_objectid,
1433                                 iterate_extent_inodes_t *iterate, void *ctx)
1434 {
1435         struct extent_inode_elem *eie;
1436         int ret = 0;
1437
1438         for (eie = inode_list; eie; eie = eie->next) {
1439                 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1440                          "root %llu\n", extent_item_objectid,
1441                          eie->inum, eie->offset, root);
1442                 ret = iterate(eie->inum, eie->offset, root, ctx);
1443                 if (ret) {
1444                         pr_debug("stopping iteration for %llu due to ret=%d\n",
1445                                  extent_item_objectid, ret);
1446                         break;
1447                 }
1448         }
1449
1450         return ret;
1451 }
1452
1453 /*
1454  * calls iterate() for every inode that references the extent identified by
1455  * the given parameters.
1456  * when the iterator function returns a non-zero value, iteration stops.
1457  */
1458 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1459                                 u64 extent_item_objectid, u64 extent_item_pos,
1460                                 int search_commit_root,
1461                                 iterate_extent_inodes_t *iterate, void *ctx)
1462 {
1463         int ret;
1464         struct list_head data_refs = LIST_HEAD_INIT(data_refs);
1465         struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
1466         struct btrfs_trans_handle *trans;
1467         struct ulist *refs = NULL;
1468         struct ulist *roots = NULL;
1469         struct ulist_node *ref_node = NULL;
1470         struct ulist_node *root_node = NULL;
1471         struct seq_list tree_mod_seq_elem = {};
1472         struct ulist_iterator ref_uiter;
1473         struct ulist_iterator root_uiter;
1474
1475         pr_debug("resolving all inodes for extent %llu\n",
1476                         extent_item_objectid);
1477
1478         if (search_commit_root) {
1479                 trans = BTRFS_BACKREF_SEARCH_COMMIT_ROOT;
1480         } else {
1481                 trans = btrfs_join_transaction(fs_info->extent_root);
1482                 if (IS_ERR(trans))
1483                         return PTR_ERR(trans);
1484                 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1485         }
1486
1487         ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1488                                    tree_mod_seq_elem.seq, &refs,
1489                                    &extent_item_pos);
1490         if (ret)
1491                 goto out;
1492
1493         ULIST_ITER_INIT(&ref_uiter);
1494         while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1495                 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
1496                                            tree_mod_seq_elem.seq, &roots);
1497                 if (ret)
1498                         break;
1499                 ULIST_ITER_INIT(&root_uiter);
1500                 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1501                         pr_debug("root %llu references leaf %llu, data list "
1502                                  "%#llx\n", root_node->val, ref_node->val,
1503                                  (long long)ref_node->aux);
1504                         ret = iterate_leaf_refs((struct extent_inode_elem *)
1505                                                 (uintptr_t)ref_node->aux,
1506                                                 root_node->val,
1507                                                 extent_item_objectid,
1508                                                 iterate, ctx);
1509                 }
1510                 ulist_free(roots);
1511                 roots = NULL;
1512         }
1513
1514         free_leaf_list(refs);
1515         ulist_free(roots);
1516 out:
1517         if (!search_commit_root) {
1518                 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1519                 btrfs_end_transaction(trans, fs_info->extent_root);
1520         }
1521
1522         return ret;
1523 }
1524
1525 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1526                                 struct btrfs_path *path,
1527                                 iterate_extent_inodes_t *iterate, void *ctx)
1528 {
1529         int ret;
1530         u64 extent_item_pos;
1531         u64 flags = 0;
1532         struct btrfs_key found_key;
1533         int search_commit_root = path->search_commit_root;
1534
1535         ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
1536         btrfs_release_path(path);
1537         if (ret < 0)
1538                 return ret;
1539         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1540                 return -EINVAL;
1541
1542         extent_item_pos = logical - found_key.objectid;
1543         ret = iterate_extent_inodes(fs_info, found_key.objectid,
1544                                         extent_item_pos, search_commit_root,
1545                                         iterate, ctx);
1546
1547         return ret;
1548 }
1549
1550 typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
1551                               struct extent_buffer *eb, void *ctx);
1552
1553 static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
1554                               struct btrfs_path *path,
1555                               iterate_irefs_t *iterate, void *ctx)
1556 {
1557         int ret = 0;
1558         int slot;
1559         u32 cur;
1560         u32 len;
1561         u32 name_len;
1562         u64 parent = 0;
1563         int found = 0;
1564         struct extent_buffer *eb;
1565         struct btrfs_item *item;
1566         struct btrfs_inode_ref *iref;
1567         struct btrfs_key found_key;
1568
1569         while (!ret) {
1570                 path->leave_spinning = 1;
1571                 ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1572                                      &found_key);
1573                 if (ret < 0)
1574                         break;
1575                 if (ret) {
1576                         ret = found ? 0 : -ENOENT;
1577                         break;
1578                 }
1579                 ++found;
1580
1581                 parent = found_key.offset;
1582                 slot = path->slots[0];
1583                 eb = path->nodes[0];
1584                 /* make sure we can use eb after releasing the path */
1585                 atomic_inc(&eb->refs);
1586                 btrfs_tree_read_lock(eb);
1587                 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1588                 btrfs_release_path(path);
1589
1590                 item = btrfs_item_nr(eb, slot);
1591                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1592
1593                 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1594                         name_len = btrfs_inode_ref_name_len(eb, iref);
1595                         /* path must be released before calling iterate()! */
1596                         pr_debug("following ref at offset %u for inode %llu in "
1597                                  "tree %llu\n", cur,
1598                                  (unsigned long long)found_key.objectid,
1599                                  (unsigned long long)fs_root->objectid);
1600                         ret = iterate(parent, name_len,
1601                                       (unsigned long)(iref + 1), eb, ctx);
1602                         if (ret)
1603                                 break;
1604                         len = sizeof(*iref) + name_len;
1605                         iref = (struct btrfs_inode_ref *)((char *)iref + len);
1606                 }
1607                 btrfs_tree_read_unlock_blocking(eb);
1608                 free_extent_buffer(eb);
1609         }
1610
1611         btrfs_release_path(path);
1612
1613         return ret;
1614 }
1615
1616 static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
1617                                  struct btrfs_path *path,
1618                                  iterate_irefs_t *iterate, void *ctx)
1619 {
1620         int ret;
1621         int slot;
1622         u64 offset = 0;
1623         u64 parent;
1624         int found = 0;
1625         struct extent_buffer *eb;
1626         struct btrfs_inode_extref *extref;
1627         struct extent_buffer *leaf;
1628         u32 item_size;
1629         u32 cur_offset;
1630         unsigned long ptr;
1631
1632         while (1) {
1633                 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
1634                                             &offset);
1635                 if (ret < 0)
1636                         break;
1637                 if (ret) {
1638                         ret = found ? 0 : -ENOENT;
1639                         break;
1640                 }
1641                 ++found;
1642
1643                 slot = path->slots[0];
1644                 eb = path->nodes[0];
1645                 /* make sure we can use eb after releasing the path */
1646                 atomic_inc(&eb->refs);
1647
1648                 btrfs_tree_read_lock(eb);
1649                 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1650                 btrfs_release_path(path);
1651
1652                 leaf = path->nodes[0];
1653                 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1654                 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1655                 cur_offset = 0;
1656
1657                 while (cur_offset < item_size) {
1658                         u32 name_len;
1659
1660                         extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
1661                         parent = btrfs_inode_extref_parent(eb, extref);
1662                         name_len = btrfs_inode_extref_name_len(eb, extref);
1663                         ret = iterate(parent, name_len,
1664                                       (unsigned long)&extref->name, eb, ctx);
1665                         if (ret)
1666                                 break;
1667
1668                         cur_offset += btrfs_inode_extref_name_len(leaf, extref);
1669                         cur_offset += sizeof(*extref);
1670                 }
1671                 btrfs_tree_read_unlock_blocking(eb);
1672                 free_extent_buffer(eb);
1673
1674                 offset++;
1675         }
1676
1677         btrfs_release_path(path);
1678
1679         return ret;
1680 }
1681
1682 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1683                          struct btrfs_path *path, iterate_irefs_t *iterate,
1684                          void *ctx)
1685 {
1686         int ret;
1687         int found_refs = 0;
1688
1689         ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
1690         if (!ret)
1691                 ++found_refs;
1692         else if (ret != -ENOENT)
1693                 return ret;
1694
1695         ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
1696         if (ret == -ENOENT && found_refs)
1697                 return 0;
1698
1699         return ret;
1700 }
1701
1702 /*
1703  * returns 0 if the path could be dumped (probably truncated)
1704  * returns <0 in case of an error
1705  */
1706 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
1707                          struct extent_buffer *eb, void *ctx)
1708 {
1709         struct inode_fs_paths *ipath = ctx;
1710         char *fspath;
1711         char *fspath_min;
1712         int i = ipath->fspath->elem_cnt;
1713         const int s_ptr = sizeof(char *);
1714         u32 bytes_left;
1715
1716         bytes_left = ipath->fspath->bytes_left > s_ptr ?
1717                                         ipath->fspath->bytes_left - s_ptr : 0;
1718
1719         fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1720         fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
1721                                    name_off, eb, inum, fspath_min, bytes_left);
1722         if (IS_ERR(fspath))
1723                 return PTR_ERR(fspath);
1724
1725         if (fspath > fspath_min) {
1726                 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1727                 ++ipath->fspath->elem_cnt;
1728                 ipath->fspath->bytes_left = fspath - fspath_min;
1729         } else {
1730                 ++ipath->fspath->elem_missed;
1731                 ipath->fspath->bytes_missing += fspath_min - fspath;
1732                 ipath->fspath->bytes_left = 0;
1733         }
1734
1735         return 0;
1736 }
1737
1738 /*
1739  * this dumps all file system paths to the inode into the ipath struct, provided
1740  * is has been created large enough. each path is zero-terminated and accessed
1741  * from ipath->fspath->val[i].
1742  * when it returns, there are ipath->fspath->elem_cnt number of paths available
1743  * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1744  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1745  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1746  * have been needed to return all paths.
1747  */
1748 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1749 {
1750         return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1751                              inode_to_path, ipath);
1752 }
1753
1754 struct btrfs_data_container *init_data_container(u32 total_bytes)
1755 {
1756         struct btrfs_data_container *data;
1757         size_t alloc_bytes;
1758
1759         alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1760         data = vmalloc(alloc_bytes);
1761         if (!data)
1762                 return ERR_PTR(-ENOMEM);
1763
1764         if (total_bytes >= sizeof(*data)) {
1765                 data->bytes_left = total_bytes - sizeof(*data);
1766                 data->bytes_missing = 0;
1767         } else {
1768                 data->bytes_missing = sizeof(*data) - total_bytes;
1769                 data->bytes_left = 0;
1770         }
1771
1772         data->elem_cnt = 0;
1773         data->elem_missed = 0;
1774
1775         return data;
1776 }
1777
1778 /*
1779  * allocates space to return multiple file system paths for an inode.
1780  * total_bytes to allocate are passed, note that space usable for actual path
1781  * information will be total_bytes - sizeof(struct inode_fs_paths).
1782  * the returned pointer must be freed with free_ipath() in the end.
1783  */
1784 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1785                                         struct btrfs_path *path)
1786 {
1787         struct inode_fs_paths *ifp;
1788         struct btrfs_data_container *fspath;
1789
1790         fspath = init_data_container(total_bytes);
1791         if (IS_ERR(fspath))
1792                 return (void *)fspath;
1793
1794         ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1795         if (!ifp) {
1796                 kfree(fspath);
1797                 return ERR_PTR(-ENOMEM);
1798         }
1799
1800         ifp->btrfs_path = path;
1801         ifp->fspath = fspath;
1802         ifp->fs_root = fs_root;
1803
1804         return ifp;
1805 }
1806
1807 void free_ipath(struct inode_fs_paths *ipath)
1808 {
1809         if (!ipath)
1810                 return;
1811         vfree(ipath->fspath);
1812         kfree(ipath);
1813 }