Pileus Git - ~andy/linux/blob - 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 offset = 0;
  40         struct extent_inode_elem *e;
  41
  42         if (!btrfs_file_extent_compression(eb, fi) &&
  43             !btrfs_file_extent_encryption(eb, fi) &&
  44             !btrfs_file_extent_other_encoding(eb, fi)) {
  45                 u64 data_offset;
  46                 u64 data_len;
  47
  48                 data_offset = btrfs_file_extent_offset(eb, fi);
  49                 data_len = btrfs_file_extent_num_bytes(eb, fi);
  50
  51                 if (extent_item_pos < data_offset ||
  52                     extent_item_pos >= data_offset + data_len)
  53                         return 1;
  54                 offset = extent_item_pos - data_offset;
  55         }
  56
  57         e = kmalloc(sizeof(*e), GFP_NOFS);
  58         if (!e)
  59                 return -ENOMEM;
  60
  61         e->next = *eie;
  62         e->inum = key->objectid;
  63         e->offset = key->offset + offset;
  64         *eie = e;
  65
  66         return 0;
  67 }
  68
  69 static int find_extent_in_eb(struct extent_buffer *eb, u64 wanted_disk_byte,
  70                                 u64 extent_item_pos,
  71                                 struct extent_inode_elem **eie)
  72 {
  73         u64 disk_byte;
  74         struct btrfs_key key;
  75         struct btrfs_file_extent_item *fi;
  76         int slot;
  77         int nritems;
  78         int extent_type;
  79         int ret;
  80
  81         /*
  82          * from the shared data ref, we only have the leaf but we need
  83          * the key. thus, we must look into all items and see that we
  84          * find one (some) with a reference to our extent item.
  85          */
  86         nritems = btrfs_header_nritems(eb);
  87         for (slot = 0; slot < nritems; ++slot) {
  88                 btrfs_item_key_to_cpu(eb, &key, slot);
  89                 if (key.type != BTRFS_EXTENT_DATA_KEY)
  90                         continue;
  91                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
  92                 extent_type = btrfs_file_extent_type(eb, fi);
  93                 if (extent_type == BTRFS_FILE_EXTENT_INLINE)
  94                         continue;
  95                 /* don't skip BTRFS_FILE_EXTENT_PREALLOC, we can handle that */
  96                 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
  97                 if (disk_byte != wanted_disk_byte)
  98                         continue;
  99
 100                 ret = check_extent_in_eb(&key, eb, fi, extent_item_pos, eie);
 101                 if (ret < 0)
 102                         return ret;
 103         }
 104
 105         return 0;
 106 }
 107
 108 /*
 109  * this structure records all encountered refs on the way up to the root
 110  */
 111 struct __prelim_ref {
 112         struct list_head list;
 113         u64 root_id;
 114         struct btrfs_key key_for_search;
 115         int level;
 116         int count;
 117         struct extent_inode_elem *inode_list;
 118         u64 parent;
 119         u64 wanted_disk_byte;
 120 };
 121
 122 static struct kmem_cache *btrfs_prelim_ref_cache;
 123
 124 int __init btrfs_prelim_ref_init(void)
 125 {
 126         btrfs_prelim_ref_cache = kmem_cache_create("btrfs_prelim_ref",
 127                                         sizeof(struct __prelim_ref),
 128                                         0,
 129                                         SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
 130                                         NULL);
 131         if (!btrfs_prelim_ref_cache)
 132                 return -ENOMEM;
 133         return 0;
 134 }
 135
 136 void btrfs_prelim_ref_exit(void)
 137 {
 138         if (btrfs_prelim_ref_cache)
 139                 kmem_cache_destroy(btrfs_prelim_ref_cache);
 140 }
 141
 142 /*
 143  * the rules for all callers of this function are:
 144  * - obtaining the parent is the goal
 145  * - if you add a key, you must know that it is a correct key
 146  * - if you cannot add the parent or a correct key, then we will look into the
 147  *   block later to set a correct key
 148  *
 149  * delayed refs
 150  * ============
 151  *        backref type | shared | indirect | shared | indirect
 152  * information         |   tree |     tree |   data |     data
 153  * --------------------+--------+----------+--------+----------
 154  *      parent logical |    y   |     -    |    -   |     -
 155  *      key to resolve |    -   |     y    |    y   |     y
 156  *  tree block logical |    -   |     -    |    -   |     -
 157  *  root for resolving |    y   |     y    |    y   |     y
 158  *
 159  * - column 1:       we've the parent -> done
 160  * - column 2, 3, 4: we use the key to find the parent
 161  *
 162  * on disk refs (inline or keyed)
 163  * ==============================
 164  *        backref type | shared | indirect | shared | indirect
 165  * information         |   tree |     tree |   data |     data
 166  * --------------------+--------+----------+--------+----------
 167  *      parent logical |    y   |     -    |    y   |     -
 168  *      key to resolve |    -   |     -    |    -   |     y
 169  *  tree block logical |    y   |     y    |    y   |     y
 170  *  root for resolving |    -   |     y    |    y   |     y
 171  *
 172  * - column 1, 3: we've the parent -> done
 173  * - column 2:    we take the first key from the block to find the parent
 174  *                (see __add_missing_keys)
 175  * - column 4:    we use the key to find the parent
 176  *
 177  * additional information that's available but not required to find the parent
 178  * block might help in merging entries to gain some speed.
 179  */
 180
 181 static int __add_prelim_ref(struct list_head *head, u64 root_id,
 182                             struct btrfs_key *key, int level,
 183                             u64 parent, u64 wanted_disk_byte, int count,
 184                             gfp_t gfp_mask)
 185 {
 186         struct __prelim_ref *ref;
 187
 188         if (root_id == BTRFS_DATA_RELOC_TREE_OBJECTID)
 189                 return 0;
 190
 191         ref = kmem_cache_alloc(btrfs_prelim_ref_cache, gfp_mask);
 192         if (!ref)
 193                 return -ENOMEM;
 194
 195         ref->root_id = root_id;
 196         if (key)
 197                 ref->key_for_search = *key;
 198         else
 199                 memset(&ref->key_for_search, 0, sizeof(ref->key_for_search));
 200
 201         ref->inode_list = NULL;
 202         ref->level = level;
 203         ref->count = count;
 204         ref->parent = parent;
 205         ref->wanted_disk_byte = wanted_disk_byte;
 206         list_add_tail(&ref->list, head);
 207
 208         return 0;
 209 }
 210
 211 static int add_all_parents(struct btrfs_root *root, struct btrfs_path *path,
 212                            struct ulist *parents, struct __prelim_ref *ref,
 213                            int level, u64 time_seq, const u64 *extent_item_pos)
 214 {
 215         int ret = 0;
 216         int slot;
 217         struct extent_buffer *eb;
 218         struct btrfs_key key;
 219         struct btrfs_key *key_for_search = &ref->key_for_search;
 220         struct btrfs_file_extent_item *fi;
 221         struct extent_inode_elem *eie = NULL, *old = NULL;
 222         u64 disk_byte;
 223         u64 wanted_disk_byte = ref->wanted_disk_byte;
 224         u64 count = 0;
 225
 226         if (level != 0) {
 227                 eb = path->nodes[level];
 228                 ret = ulist_add(parents, eb->start, 0, GFP_NOFS);
 229                 if (ret < 0)
 230                         return ret;
 231                 return 0;
 232         }
 233
 234         /*
 235          * We normally enter this function with the path already pointing to
 236          * the first item to check. But sometimes, we may enter it with
 237          * slot==nritems. In that case, go to the next leaf before we continue.
 238          */
 239         if (path->slots[0] >= btrfs_header_nritems(path->nodes[0]))
 240                 ret = btrfs_next_old_leaf(root, path, time_seq);
 241
 242         while (!ret && count < ref->count) {
 243                 eb = path->nodes[0];
 244                 slot = path->slots[0];
 245
 246                 btrfs_item_key_to_cpu(eb, &key, slot);
 247
 248                 if (key.objectid != key_for_search->objectid ||
 249                     key.type != BTRFS_EXTENT_DATA_KEY)
 250                         break;
 251
 252                 fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
 253                 disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
 254
 255                 if (disk_byte == wanted_disk_byte) {
 256                         eie = NULL;
 257                         old = NULL;
 258                         count++;
 259                         if (extent_item_pos) {
 260                                 ret = check_extent_in_eb(&key, eb, fi,
 261                                                 *extent_item_pos,
 262                                                 &eie);
 263                                 if (ret < 0)
 264                                         break;
 265                         }
 266                         if (ret > 0)
 267                                 goto next;
 268                         ret = ulist_add_merge(parents, eb->start,
 269                                               (uintptr_t)eie,
 270                                               (u64 *)&old, GFP_NOFS);
 271                         if (ret < 0)
 272                                 break;
 273                         if (!ret && extent_item_pos) {
 274                                 while (old->next)
 275                                         old = old->next;
 276                                 old->next = eie;
 277                         }
 278                 }
 279 next:
 280                 ret = btrfs_next_old_item(root, path, time_seq);
 281         }
 282
 283         if (ret > 0)
 284                 ret = 0;
 285         return ret;
 286 }
 287
 288 /*
 289  * resolve an indirect backref in the form (root_id, key, level)
 290  * to a logical address
 291  */
 292 static int __resolve_indirect_ref(struct btrfs_fs_info *fs_info,
 293                                   struct btrfs_path *path, u64 time_seq,
 294                                   struct __prelim_ref *ref,
 295                                   struct ulist *parents,
 296                                   const u64 *extent_item_pos)
 297 {
 298         struct btrfs_root *root;
 299         struct btrfs_key root_key;
 300         struct extent_buffer *eb;
 301         int ret = 0;
 302         int root_level;
 303         int level = ref->level;
 304         int index;
 305
 306         root_key.objectid = ref->root_id;
 307         root_key.type = BTRFS_ROOT_ITEM_KEY;
 308         root_key.offset = (u64)-1;
 309
 310         index = srcu_read_lock(&fs_info->subvol_srcu);
 311
 312         root = btrfs_read_fs_root_no_name(fs_info, &root_key);
 313         if (IS_ERR(root)) {
 314                 srcu_read_unlock(&fs_info->subvol_srcu, index);
 315                 ret = PTR_ERR(root);
 316                 goto out;
 317         }
 318
 319         root_level = btrfs_old_root_level(root, time_seq);
 320
 321         if (root_level + 1 == level) {
 322                 srcu_read_unlock(&fs_info->subvol_srcu, index);
 323                 goto out;
 324         }
 325
 326         path->lowest_level = level;
 327         ret = btrfs_search_old_slot(root, &ref->key_for_search, path, time_seq);
 328
 329         /* root node has been locked, we can release @subvol_srcu safely here */
 330         srcu_read_unlock(&fs_info->subvol_srcu, index);
 331
 332         pr_debug("search slot in root %llu (level %d, ref count %d) returned "
 333                  "%d for key (%llu %u %llu)\n",
 334                  ref->root_id, level, ref->count, ret,
 335                  ref->key_for_search.objectid, ref->key_for_search.type,
 336                  ref->key_for_search.offset);
 337         if (ret < 0)
 338                 goto out;
 339
 340         eb = path->nodes[level];
 341         while (!eb) {
 342                 if (WARN_ON(!level)) {
 343                         ret = 1;
 344                         goto out;
 345                 }
 346                 level--;
 347                 eb = path->nodes[level];
 348         }
 349
 350         ret = add_all_parents(root, path, parents, ref, level, time_seq,
 351                               extent_item_pos);
 352 out:
 353         path->lowest_level = 0;
 354         btrfs_release_path(path);
 355         return ret;
 356 }
 357
 358 /*
 359  * resolve all indirect backrefs from the list
 360  */
 361 static int __resolve_indirect_refs(struct btrfs_fs_info *fs_info,
 362                                    struct btrfs_path *path, u64 time_seq,
 363                                    struct list_head *head,
 364                                    const u64 *extent_item_pos)
 365 {
 366         int err;
 367         int ret = 0;
 368         struct __prelim_ref *ref;
 369         struct __prelim_ref *ref_safe;
 370         struct __prelim_ref *new_ref;
 371         struct ulist *parents;
 372         struct ulist_node *node;
 373         struct ulist_iterator uiter;
 374
 375         parents = ulist_alloc(GFP_NOFS);
 376         if (!parents)
 377                 return -ENOMEM;
 378
 379         /*
 380          * _safe allows us to insert directly after the current item without
 381          * iterating over the newly inserted items.
 382          * we're also allowed to re-assign ref during iteration.
 383          */
 384         list_for_each_entry_safe(ref, ref_safe, head, list) {
 385                 if (ref->parent)        /* already direct */
 386                         continue;
 387                 if (ref->count == 0)
 388                         continue;
 389                 err = __resolve_indirect_ref(fs_info, path, time_seq, ref,
 390                                              parents, extent_item_pos);
 391                 if (err == -ENOMEM)
 392                         goto out;
 393                 if (err)
 394                         continue;
 395
 396                 /* we put the first parent into the ref at hand */
 397                 ULIST_ITER_INIT(&uiter);
 398                 node = ulist_next(parents, &uiter);
 399                 ref->parent = node ? node->val : 0;
 400                 ref->inode_list = node ?
 401                         (struct extent_inode_elem *)(uintptr_t)node->aux : NULL;
 402
 403                 /* additional parents require new refs being added here */
 404                 while ((node = ulist_next(parents, &uiter))) {
 405                         new_ref = kmem_cache_alloc(btrfs_prelim_ref_cache,
 406                                                    GFP_NOFS);
 407                         if (!new_ref) {
 408                                 ret = -ENOMEM;
 409                                 goto out;
 410                         }
 411                         memcpy(new_ref, ref, sizeof(*ref));
 412                         new_ref->parent = node->val;
 413                         new_ref->inode_list = (struct extent_inode_elem *)
 414                                                         (uintptr_t)node->aux;
 415                         list_add(&new_ref->list, &ref->list);
 416                 }
 417                 ulist_reinit(parents);
 418         }
 419 out:
 420         ulist_free(parents);
 421         return ret;
 422 }
 423
 424 static inline int ref_for_same_block(struct __prelim_ref *ref1,
 425                                      struct __prelim_ref *ref2)
 426 {
 427         if (ref1->level != ref2->level)
 428                 return 0;
 429         if (ref1->root_id != ref2->root_id)
 430                 return 0;
 431         if (ref1->key_for_search.type != ref2->key_for_search.type)
 432                 return 0;
 433         if (ref1->key_for_search.objectid != ref2->key_for_search.objectid)
 434                 return 0;
 435         if (ref1->key_for_search.offset != ref2->key_for_search.offset)
 436                 return 0;
 437         if (ref1->parent != ref2->parent)
 438                 return 0;
 439
 440         return 1;
 441 }
 442
 443 /*
 444  * read tree blocks and add keys where required.
 445  */
 446 static int __add_missing_keys(struct btrfs_fs_info *fs_info,
 447                               struct list_head *head)
 448 {
 449         struct list_head *pos;
 450         struct extent_buffer *eb;
 451
 452         list_for_each(pos, head) {
 453                 struct __prelim_ref *ref;
 454                 ref = list_entry(pos, struct __prelim_ref, list);
 455
 456                 if (ref->parent)
 457                         continue;
 458                 if (ref->key_for_search.type)
 459                         continue;
 460                 BUG_ON(!ref->wanted_disk_byte);
 461                 eb = read_tree_block(fs_info->tree_root, ref->wanted_disk_byte,
 462                                      fs_info->tree_root->leafsize, 0);
 463                 if (!eb || !extent_buffer_uptodate(eb)) {
 464                         free_extent_buffer(eb);
 465                         return -EIO;
 466                 }
 467                 btrfs_tree_read_lock(eb);
 468                 if (btrfs_header_level(eb) == 0)
 469                         btrfs_item_key_to_cpu(eb, &ref->key_for_search, 0);
 470                 else
 471                         btrfs_node_key_to_cpu(eb, &ref->key_for_search, 0);
 472                 btrfs_tree_read_unlock(eb);
 473                 free_extent_buffer(eb);
 474         }
 475         return 0;
 476 }
 477
 478 /*
 479  * merge two lists of backrefs and adjust counts accordingly
 480  *
 481  * mode = 1: merge identical keys, if key is set
 482  *    FIXME: if we add more keys in __add_prelim_ref, we can merge more here.
 483  *           additionally, we could even add a key range for the blocks we
 484  *           looked into to merge even more (-> replace unresolved refs by those
 485  *           having a parent).
 486  * mode = 2: merge identical parents
 487  */
 488 static void __merge_refs(struct list_head *head, int mode)
 489 {
 490         struct list_head *pos1;
 491
 492         list_for_each(pos1, head) {
 493                 struct list_head *n2;
 494                 struct list_head *pos2;
 495                 struct __prelim_ref *ref1;
 496
 497                 ref1 = list_entry(pos1, struct __prelim_ref, list);
 498
 499                 for (pos2 = pos1->next, n2 = pos2->next; pos2 != head;
 500                      pos2 = n2, n2 = pos2->next) {
 501                         struct __prelim_ref *ref2;
 502                         struct __prelim_ref *xchg;
 503                         struct extent_inode_elem *eie;
 504
 505                         ref2 = list_entry(pos2, struct __prelim_ref, list);
 506
 507                         if (mode == 1) {
 508                                 if (!ref_for_same_block(ref1, ref2))
 509                                         continue;
 510                                 if (!ref1->parent && ref2->parent) {
 511                                         xchg = ref1;
 512                                         ref1 = ref2;
 513                                         ref2 = xchg;
 514                                 }
 515                         } else {
 516                                 if (ref1->parent != ref2->parent)
 517                                         continue;
 518                         }
 519
 520                         eie = ref1->inode_list;
 521                         while (eie && eie->next)
 522                                 eie = eie->next;
 523                         if (eie)
 524                                 eie->next = ref2->inode_list;
 525                         else
 526                                 ref1->inode_list = ref2->inode_list;
 527                         ref1->count += ref2->count;
 528
 529                         list_del(&ref2->list);
 530                         kmem_cache_free(btrfs_prelim_ref_cache, ref2);
 531                 }
 532
 533         }
 534 }
 535
 536 /*
 537  * add all currently queued delayed refs from this head whose seq nr is
 538  * smaller or equal that seq to the list
 539  */
 540 static int __add_delayed_refs(struct btrfs_delayed_ref_head *head, u64 seq,
 541                               struct list_head *prefs)
 542 {
 543         struct btrfs_delayed_extent_op *extent_op = head->extent_op;
 544         struct rb_node *n = &head->node.rb_node;
 545         struct btrfs_key key;
 546         struct btrfs_key op_key = {0};
 547         int sgn;
 548         int ret = 0;
 549
 550         if (extent_op && extent_op->update_key)
 551                 btrfs_disk_key_to_cpu(&op_key, &extent_op->key);
 552
 553         spin_lock(&head->lock);
 554         n = rb_first(&head->ref_root);
 555         while (n) {
 556                 struct btrfs_delayed_ref_node *node;
 557                 node = rb_entry(n, struct btrfs_delayed_ref_node,
 558                                 rb_node);
 559                 n = rb_next(n);
 560                 if (node->seq > seq)
 561                         continue;
 562
 563                 switch (node->action) {
 564                 case BTRFS_ADD_DELAYED_EXTENT:
 565                 case BTRFS_UPDATE_DELAYED_HEAD:
 566                         WARN_ON(1);
 567                         continue;
 568                 case BTRFS_ADD_DELAYED_REF:
 569                         sgn = 1;
 570                         break;
 571                 case BTRFS_DROP_DELAYED_REF:
 572                         sgn = -1;
 573                         break;
 574                 default:
 575                         BUG_ON(1);
 576                 }
 577                 switch (node->type) {
 578                 case BTRFS_TREE_BLOCK_REF_KEY: {
 579                         struct btrfs_delayed_tree_ref *ref;
 580
 581                         ref = btrfs_delayed_node_to_tree_ref(node);
 582                         ret = __add_prelim_ref(prefs, ref->root, &op_key,
 583                                                ref->level + 1, 0, node->bytenr,
 584                                                node->ref_mod * sgn, GFP_ATOMIC);
 585                         break;
 586                 }
 587                 case BTRFS_SHARED_BLOCK_REF_KEY: {
 588                         struct btrfs_delayed_tree_ref *ref;
 589
 590                         ref = btrfs_delayed_node_to_tree_ref(node);
 591                         ret = __add_prelim_ref(prefs, ref->root, NULL,
 592                                                ref->level + 1, ref->parent,
 593                                                node->bytenr,
 594                                                node->ref_mod * sgn, GFP_ATOMIC);
 595                         break;
 596                 }
 597                 case BTRFS_EXTENT_DATA_REF_KEY: {
 598                         struct btrfs_delayed_data_ref *ref;
 599                         ref = btrfs_delayed_node_to_data_ref(node);
 600
 601                         key.objectid = ref->objectid;
 602                         key.type = BTRFS_EXTENT_DATA_KEY;
 603                         key.offset = ref->offset;
 604                         ret = __add_prelim_ref(prefs, ref->root, &key, 0, 0,
 605                                                node->bytenr,
 606                                                node->ref_mod * sgn, GFP_ATOMIC);
 607                         break;
 608                 }
 609                 case BTRFS_SHARED_DATA_REF_KEY: {
 610                         struct btrfs_delayed_data_ref *ref;
 611
 612                         ref = btrfs_delayed_node_to_data_ref(node);
 613
 614                         key.objectid = ref->objectid;
 615                         key.type = BTRFS_EXTENT_DATA_KEY;
 616                         key.offset = ref->offset;
 617                         ret = __add_prelim_ref(prefs, ref->root, &key, 0,
 618                                                ref->parent, node->bytenr,
 619                                                node->ref_mod * sgn, GFP_ATOMIC);
 620                         break;
 621                 }
 622                 default:
 623                         WARN_ON(1);
 624                 }
 625                 if (ret)
 626                         break;
 627         }
 628         spin_unlock(&head->lock);
 629         return ret;
 630 }
 631
 632 /*
 633  * add all inline backrefs for bytenr to the list
 634  */
 635 static int __add_inline_refs(struct btrfs_fs_info *fs_info,
 636                              struct btrfs_path *path, u64 bytenr,
 637                              int *info_level, struct list_head *prefs)
 638 {
 639         int ret = 0;
 640         int slot;
 641         struct extent_buffer *leaf;
 642         struct btrfs_key key;
 643         struct btrfs_key found_key;
 644         unsigned long ptr;
 645         unsigned long end;
 646         struct btrfs_extent_item *ei;
 647         u64 flags;
 648         u64 item_size;
 649
 650         /*
 651          * enumerate all inline refs
 652          */
 653         leaf = path->nodes[0];
 654         slot = path->slots[0];
 655
 656         item_size = btrfs_item_size_nr(leaf, slot);
 657         BUG_ON(item_size < sizeof(*ei));
 658
 659         ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
 660         flags = btrfs_extent_flags(leaf, ei);
 661         btrfs_item_key_to_cpu(leaf, &found_key, slot);
 662
 663         ptr = (unsigned long)(ei + 1);
 664         end = (unsigned long)ei + item_size;
 665
 666         if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
 667             flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
 668                 struct btrfs_tree_block_info *info;
 669
 670                 info = (struct btrfs_tree_block_info *)ptr;
 671                 *info_level = btrfs_tree_block_level(leaf, info);
 672                 ptr += sizeof(struct btrfs_tree_block_info);
 673                 BUG_ON(ptr > end);
 674         } else if (found_key.type == BTRFS_METADATA_ITEM_KEY) {
 675                 *info_level = found_key.offset;
 676         } else {
 677                 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
 678         }
 679
 680         while (ptr < end) {
 681                 struct btrfs_extent_inline_ref *iref;
 682                 u64 offset;
 683                 int type;
 684
 685                 iref = (struct btrfs_extent_inline_ref *)ptr;
 686                 type = btrfs_extent_inline_ref_type(leaf, iref);
 687                 offset = btrfs_extent_inline_ref_offset(leaf, iref);
 688
 689                 switch (type) {
 690                 case BTRFS_SHARED_BLOCK_REF_KEY:
 691                         ret = __add_prelim_ref(prefs, 0, NULL,
 692                                                 *info_level + 1, offset,
 693                                                 bytenr, 1, GFP_NOFS);
 694                         break;
 695                 case BTRFS_SHARED_DATA_REF_KEY: {
 696                         struct btrfs_shared_data_ref *sdref;
 697                         int count;
 698
 699                         sdref = (struct btrfs_shared_data_ref *)(iref + 1);
 700                         count = btrfs_shared_data_ref_count(leaf, sdref);
 701                         ret = __add_prelim_ref(prefs, 0, NULL, 0, offset,
 702                                                bytenr, count, GFP_NOFS);
 703                         break;
 704                 }
 705                 case BTRFS_TREE_BLOCK_REF_KEY:
 706                         ret = __add_prelim_ref(prefs, offset, NULL,
 707                                                *info_level + 1, 0,
 708                                                bytenr, 1, GFP_NOFS);
 709                         break;
 710                 case BTRFS_EXTENT_DATA_REF_KEY: {
 711                         struct btrfs_extent_data_ref *dref;
 712                         int count;
 713                         u64 root;
 714
 715                         dref = (struct btrfs_extent_data_ref *)(&iref->offset);
 716                         count = btrfs_extent_data_ref_count(leaf, dref);
 717                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
 718                                                                       dref);
 719                         key.type = BTRFS_EXTENT_DATA_KEY;
 720                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
 721                         root = btrfs_extent_data_ref_root(leaf, dref);
 722                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
 723                                                bytenr, count, GFP_NOFS);
 724                         break;
 725                 }
 726                 default:
 727                         WARN_ON(1);
 728                 }
 729                 if (ret)
 730                         return ret;
 731                 ptr += btrfs_extent_inline_ref_size(type);
 732         }
 733
 734         return 0;
 735 }
 736
 737 /*
 738  * add all non-inline backrefs for bytenr to the list
 739  */
 740 static int __add_keyed_refs(struct btrfs_fs_info *fs_info,
 741                             struct btrfs_path *path, u64 bytenr,
 742                             int info_level, struct list_head *prefs)
 743 {
 744         struct btrfs_root *extent_root = fs_info->extent_root;
 745         int ret;
 746         int slot;
 747         struct extent_buffer *leaf;
 748         struct btrfs_key key;
 749
 750         while (1) {
 751                 ret = btrfs_next_item(extent_root, path);
 752                 if (ret < 0)
 753                         break;
 754                 if (ret) {
 755                         ret = 0;
 756                         break;
 757                 }
 758
 759                 slot = path->slots[0];
 760                 leaf = path->nodes[0];
 761                 btrfs_item_key_to_cpu(leaf, &key, slot);
 762
 763                 if (key.objectid != bytenr)
 764                         break;
 765                 if (key.type < BTRFS_TREE_BLOCK_REF_KEY)
 766                         continue;
 767                 if (key.type > BTRFS_SHARED_DATA_REF_KEY)
 768                         break;
 769
 770                 switch (key.type) {
 771                 case BTRFS_SHARED_BLOCK_REF_KEY:
 772                         ret = __add_prelim_ref(prefs, 0, NULL,
 773                                                 info_level + 1, key.offset,
 774                                                 bytenr, 1, GFP_NOFS);
 775                         break;
 776                 case BTRFS_SHARED_DATA_REF_KEY: {
 777                         struct btrfs_shared_data_ref *sdref;
 778                         int count;
 779
 780                         sdref = btrfs_item_ptr(leaf, slot,
 781                                               struct btrfs_shared_data_ref);
 782                         count = btrfs_shared_data_ref_count(leaf, sdref);
 783                         ret = __add_prelim_ref(prefs, 0, NULL, 0, key.offset,
 784                                                 bytenr, count, GFP_NOFS);
 785                         break;
 786                 }
 787                 case BTRFS_TREE_BLOCK_REF_KEY:
 788                         ret = __add_prelim_ref(prefs, key.offset, NULL,
 789                                                info_level + 1, 0,
 790                                                bytenr, 1, GFP_NOFS);
 791                         break;
 792                 case BTRFS_EXTENT_DATA_REF_KEY: {
 793                         struct btrfs_extent_data_ref *dref;
 794                         int count;
 795                         u64 root;
 796
 797                         dref = btrfs_item_ptr(leaf, slot,
 798                                               struct btrfs_extent_data_ref);
 799                         count = btrfs_extent_data_ref_count(leaf, dref);
 800                         key.objectid = btrfs_extent_data_ref_objectid(leaf,
 801                                                                       dref);
 802                         key.type = BTRFS_EXTENT_DATA_KEY;
 803                         key.offset = btrfs_extent_data_ref_offset(leaf, dref);
 804                         root = btrfs_extent_data_ref_root(leaf, dref);
 805                         ret = __add_prelim_ref(prefs, root, &key, 0, 0,
 806                                                bytenr, count, GFP_NOFS);
 807                         break;
 808                 }
 809                 default:
 810                         WARN_ON(1);
 811                 }
 812                 if (ret)
 813                         return ret;
 814
 815         }
 816
 817         return ret;
 818 }
 819
 820 /*
 821  * this adds all existing backrefs (inline backrefs, backrefs and delayed
 822  * refs) for the given bytenr to the refs list, merges duplicates and resolves
 823  * indirect refs to their parent bytenr.
 824  * When roots are found, they're added to the roots list
 825  *
 826  * FIXME some caching might speed things up
 827  */
 828 static int find_parent_nodes(struct btrfs_trans_handle *trans,
 829                              struct btrfs_fs_info *fs_info, u64 bytenr,
 830                              u64 time_seq, struct ulist *refs,
 831                              struct ulist *roots, const u64 *extent_item_pos)
 832 {
 833         struct btrfs_key key;
 834         struct btrfs_path *path;
 835         struct btrfs_delayed_ref_root *delayed_refs = NULL;
 836         struct btrfs_delayed_ref_head *head;
 837         int info_level = 0;
 838         int ret;
 839         struct list_head prefs_delayed;
 840         struct list_head prefs;
 841         struct __prelim_ref *ref;
 842
 843         INIT_LIST_HEAD(&prefs);
 844         INIT_LIST_HEAD(&prefs_delayed);
 845
 846         key.objectid = bytenr;
 847         key.offset = (u64)-1;
 848         if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
 849                 key.type = BTRFS_METADATA_ITEM_KEY;
 850         else
 851                 key.type = BTRFS_EXTENT_ITEM_KEY;
 852
 853         path = btrfs_alloc_path();
 854         if (!path)
 855                 return -ENOMEM;
 856         if (!trans)
 857                 path->search_commit_root = 1;
 858
 859         /*
 860          * grab both a lock on the path and a lock on the delayed ref head.
 861          * We need both to get a consistent picture of how the refs look
 862          * at a specified point in time
 863          */
 864 again:
 865         head = NULL;
 866
 867         ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
 868         if (ret < 0)
 869                 goto out;
 870         BUG_ON(ret == 0);
 871
 872         if (trans) {
 873                 /*
 874                  * look if there are updates for this ref queued and lock the
 875                  * head
 876                  */
 877                 delayed_refs = &trans->transaction->delayed_refs;
 878                 spin_lock(&delayed_refs->lock);
 879                 head = btrfs_find_delayed_ref_head(trans, bytenr);
 880                 if (head) {
 881                         if (!mutex_trylock(&head->mutex)) {
 882                                 atomic_inc(&head->node.refs);
 883                                 spin_unlock(&delayed_refs->lock);
 884
 885                                 btrfs_release_path(path);
 886
 887                                 /*
 888                                  * Mutex was contended, block until it's
 889                                  * released and try again
 890                                  */
 891                                 mutex_lock(&head->mutex);
 892                                 mutex_unlock(&head->mutex);
 893                                 btrfs_put_delayed_ref(&head->node);
 894                                 goto again;
 895                         }
 896                         spin_unlock(&delayed_refs->lock);
 897                         ret = __add_delayed_refs(head, time_seq,
 898                                                  &prefs_delayed);
 899                         mutex_unlock(&head->mutex);
 900                         if (ret)
 901                                 goto out;
 902                 } else {
 903                         spin_unlock(&delayed_refs->lock);
 904                 }
 905         }
 906
 907         if (path->slots[0]) {
 908                 struct extent_buffer *leaf;
 909                 int slot;
 910
 911                 path->slots[0]--;
 912                 leaf = path->nodes[0];
 913                 slot = path->slots[0];
 914                 btrfs_item_key_to_cpu(leaf, &key, slot);
 915                 if (key.objectid == bytenr &&
 916                     (key.type == BTRFS_EXTENT_ITEM_KEY ||
 917                      key.type == BTRFS_METADATA_ITEM_KEY)) {
 918                         ret = __add_inline_refs(fs_info, path, bytenr,
 919                                                 &info_level, &prefs);
 920                         if (ret)
 921                                 goto out;
 922                         ret = __add_keyed_refs(fs_info, path, bytenr,
 923                                                info_level, &prefs);
 924                         if (ret)
 925                                 goto out;
 926                 }
 927         }
 928         btrfs_release_path(path);
 929
 930         list_splice_init(&prefs_delayed, &prefs);
 931
 932         ret = __add_missing_keys(fs_info, &prefs);
 933         if (ret)
 934                 goto out;
 935
 936         __merge_refs(&prefs, 1);
 937
 938         ret = __resolve_indirect_refs(fs_info, path, time_seq, &prefs,
 939                                       extent_item_pos);
 940         if (ret)
 941                 goto out;
 942
 943         __merge_refs(&prefs, 2);
 944
 945         while (!list_empty(&prefs)) {
 946                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
 947                 WARN_ON(ref->count < 0);
 948                 if (ref->count && ref->root_id && ref->parent == 0) {
 949                         /* no parent == root of tree */
 950                         ret = ulist_add(roots, ref->root_id, 0, GFP_NOFS);
 951                         if (ret < 0)
 952                                 goto out;
 953                 }
 954                 if (ref->count && ref->parent) {
 955                         struct extent_inode_elem *eie = NULL;
 956                         if (extent_item_pos && !ref->inode_list) {
 957                                 u32 bsz;
 958                                 struct extent_buffer *eb;
 959                                 bsz = btrfs_level_size(fs_info->extent_root,
 960                                                         info_level);
 961                                 eb = read_tree_block(fs_info->extent_root,
 962                                                            ref->parent, bsz, 0);
 963                                 if (!eb || !extent_buffer_uptodate(eb)) {
 964                                         free_extent_buffer(eb);
 965                                         ret = -EIO;
 966                                         goto out;
 967                                 }
 968                                 ret = find_extent_in_eb(eb, bytenr,
 969                                                         *extent_item_pos, &eie);
 970                                 free_extent_buffer(eb);
 971                                 if (ret < 0)
 972                                         goto out;
 973                                 ref->inode_list = eie;
 974                         }
 975                         ret = ulist_add_merge(refs, ref->parent,
 976                                               (uintptr_t)ref->inode_list,
 977                                               (u64 *)&eie, GFP_NOFS);
 978                         if (ret < 0)
 979                                 goto out;
 980                         if (!ret && extent_item_pos) {
 981                                 /*
 982                                  * we've recorded that parent, so we must extend
 983                                  * its inode list here
 984                                  */
 985                                 BUG_ON(!eie);
 986                                 while (eie->next)
 987                                         eie = eie->next;
 988                                 eie->next = ref->inode_list;
 989                         }
 990                 }
 991                 list_del(&ref->list);
 992                 kmem_cache_free(btrfs_prelim_ref_cache, ref);
 993         }
 994
 995 out:
 996         btrfs_free_path(path);
 997         while (!list_empty(&prefs)) {
 998                 ref = list_first_entry(&prefs, struct __prelim_ref, list);
 999                 list_del(&ref->list);
1000                 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1001         }
1002         while (!list_empty(&prefs_delayed)) {
1003                 ref = list_first_entry(&prefs_delayed, struct __prelim_ref,
1004                                        list);
1005                 list_del(&ref->list);
1006                 kmem_cache_free(btrfs_prelim_ref_cache, ref);
1007         }
1008
1009         return ret;
1010 }
1011
1012 static void free_leaf_list(struct ulist *blocks)
1013 {
1014         struct ulist_node *node = NULL;
1015         struct extent_inode_elem *eie;
1016         struct extent_inode_elem *eie_next;
1017         struct ulist_iterator uiter;
1018
1019         ULIST_ITER_INIT(&uiter);
1020         while ((node = ulist_next(blocks, &uiter))) {
1021                 if (!node->aux)
1022                         continue;
1023                 eie = (struct extent_inode_elem *)(uintptr_t)node->aux;
1024                 for (; eie; eie = eie_next) {
1025                         eie_next = eie->next;
1026                         kfree(eie);
1027                 }
1028                 node->aux = 0;
1029         }
1030
1031         ulist_free(blocks);
1032 }
1033
1034 /*
1035  * Finds all leafs with a reference to the specified combination of bytenr and
1036  * offset. key_list_head will point to a list of corresponding keys (caller must
1037  * free each list element). The leafs will be stored in the leafs ulist, which
1038  * must be freed with ulist_free.
1039  *
1040  * returns 0 on success, <0 on error
1041  */
1042 static int btrfs_find_all_leafs(struct btrfs_trans_handle *trans,
1043                                 struct btrfs_fs_info *fs_info, u64 bytenr,
1044                                 u64 time_seq, struct ulist **leafs,
1045                                 const u64 *extent_item_pos)
1046 {
1047         struct ulist *tmp;
1048         int ret;
1049
1050         tmp = ulist_alloc(GFP_NOFS);
1051         if (!tmp)
1052                 return -ENOMEM;
1053         *leafs = ulist_alloc(GFP_NOFS);
1054         if (!*leafs) {
1055                 ulist_free(tmp);
1056                 return -ENOMEM;
1057         }
1058
1059         ret = find_parent_nodes(trans, fs_info, bytenr,
1060                                 time_seq, *leafs, tmp, extent_item_pos);
1061         ulist_free(tmp);
1062
1063         if (ret < 0 && ret != -ENOENT) {
1064                 free_leaf_list(*leafs);
1065                 return ret;
1066         }
1067
1068         return 0;
1069 }
1070
1071 /*
1072  * walk all backrefs for a given extent to find all roots that reference this
1073  * extent. Walking a backref means finding all extents that reference this
1074  * extent and in turn walk the backrefs of those, too. Naturally this is a
1075  * recursive process, but here it is implemented in an iterative fashion: We
1076  * find all referencing extents for the extent in question and put them on a
1077  * list. In turn, we find all referencing extents for those, further appending
1078  * to the list. The way we iterate the list allows adding more elements after
1079  * the current while iterating. The process stops when we reach the end of the
1080  * list. Found roots are added to the roots list.
1081  *
1082  * returns 0 on success, < 0 on error.
1083  */
1084 int btrfs_find_all_roots(struct btrfs_trans_handle *trans,
1085                                 struct btrfs_fs_info *fs_info, u64 bytenr,
1086                                 u64 time_seq, struct ulist **roots)
1087 {
1088         struct ulist *tmp;
1089         struct ulist_node *node = NULL;
1090         struct ulist_iterator uiter;
1091         int ret;
1092
1093         tmp = ulist_alloc(GFP_NOFS);
1094         if (!tmp)
1095                 return -ENOMEM;
1096         *roots = ulist_alloc(GFP_NOFS);
1097         if (!*roots) {
1098                 ulist_free(tmp);
1099                 return -ENOMEM;
1100         }
1101
1102         ULIST_ITER_INIT(&uiter);
1103         while (1) {
1104                 ret = find_parent_nodes(trans, fs_info, bytenr,
1105                                         time_seq, tmp, *roots, NULL);
1106                 if (ret < 0 && ret != -ENOENT) {
1107                         ulist_free(tmp);
1108                         ulist_free(*roots);
1109                         return ret;
1110                 }
1111                 node = ulist_next(tmp, &uiter);
1112                 if (!node)
1113                         break;
1114                 bytenr = node->val;
1115         }
1116
1117         ulist_free(tmp);
1118         return 0;
1119 }
1120
1121 /*
1122  * this makes the path point to (inum INODE_ITEM ioff)
1123  */
1124 int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1125                         struct btrfs_path *path)
1126 {
1127         struct btrfs_key key;
1128         return btrfs_find_item(fs_root, path, inum, ioff,
1129                         BTRFS_INODE_ITEM_KEY, &key);
1130 }
1131
1132 static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
1133                                 struct btrfs_path *path,
1134                                 struct btrfs_key *found_key)
1135 {
1136         return btrfs_find_item(fs_root, path, inum, ioff,
1137                         BTRFS_INODE_REF_KEY, found_key);
1138 }
1139
1140 int btrfs_find_one_extref(struct btrfs_root *root, u64 inode_objectid,
1141                           u64 start_off, struct btrfs_path *path,
1142                           struct btrfs_inode_extref **ret_extref,
1143                           u64 *found_off)
1144 {
1145         int ret, slot;
1146         struct btrfs_key key;
1147         struct btrfs_key found_key;
1148         struct btrfs_inode_extref *extref;
1149         struct extent_buffer *leaf;
1150         unsigned long ptr;
1151
1152         key.objectid = inode_objectid;
1153         btrfs_set_key_type(&key, BTRFS_INODE_EXTREF_KEY);
1154         key.offset = start_off;
1155
1156         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1157         if (ret < 0)
1158                 return ret;
1159
1160         while (1) {
1161                 leaf = path->nodes[0];
1162                 slot = path->slots[0];
1163                 if (slot >= btrfs_header_nritems(leaf)) {
1164                         /*
1165                          * If the item at offset is not found,
1166                          * btrfs_search_slot will point us to the slot
1167                          * where it should be inserted. In our case
1168                          * that will be the slot directly before the
1169                          * next INODE_REF_KEY_V2 item. In the case
1170                          * that we're pointing to the last slot in a
1171                          * leaf, we must move one leaf over.
1172                          */
1173                         ret = btrfs_next_leaf(root, path);
1174                         if (ret) {
1175                                 if (ret >= 1)
1176                                         ret = -ENOENT;
1177                                 break;
1178                         }
1179                         continue;
1180                 }
1181
1182                 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1183
1184                 /*
1185                  * Check that we're still looking at an extended ref key for
1186                  * this particular objectid. If we have different
1187                  * objectid or type then there are no more to be found
1188                  * in the tree and we can exit.
1189                  */
1190                 ret = -ENOENT;
1191                 if (found_key.objectid != inode_objectid)
1192                         break;
1193                 if (btrfs_key_type(&found_key) != BTRFS_INODE_EXTREF_KEY)
1194                         break;
1195
1196                 ret = 0;
1197                 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
1198                 extref = (struct btrfs_inode_extref *)ptr;
1199                 *ret_extref = extref;
1200                 if (found_off)
1201                         *found_off = found_key.offset;
1202                 break;
1203         }
1204
1205         return ret;
1206 }
1207
1208 /*
1209  * this iterates to turn a name (from iref/extref) into a full filesystem path.
1210  * Elements of the path are separated by '/' and the path is guaranteed to be
1211  * 0-terminated. the path is only given within the current file system.
1212  * Therefore, it never starts with a '/'. the caller is responsible to provide
1213  * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
1214  * the start point of the resulting string is returned. this pointer is within
1215  * dest, normally.
1216  * in case the path buffer would overflow, the pointer is decremented further
1217  * as if output was written to the buffer, though no more output is actually
1218  * generated. that way, the caller can determine how much space would be
1219  * required for the path to fit into the buffer. in that case, the returned
1220  * value will be smaller than dest. callers must check this!
1221  */
1222 char *btrfs_ref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
1223                         u32 name_len, unsigned long name_off,
1224                         struct extent_buffer *eb_in, u64 parent,
1225                         char *dest, u32 size)
1226 {
1227         int slot;
1228         u64 next_inum;
1229         int ret;
1230         s64 bytes_left = ((s64)size) - 1;
1231         struct extent_buffer *eb = eb_in;
1232         struct btrfs_key found_key;
1233         int leave_spinning = path->leave_spinning;
1234         struct btrfs_inode_ref *iref;
1235
1236         if (bytes_left >= 0)
1237                 dest[bytes_left] = '\0';
1238
1239         path->leave_spinning = 1;
1240         while (1) {
1241                 bytes_left -= name_len;
1242                 if (bytes_left >= 0)
1243                         read_extent_buffer(eb, dest + bytes_left,
1244                                            name_off, name_len);
1245                 if (eb != eb_in) {
1246                         btrfs_tree_read_unlock_blocking(eb);
1247                         free_extent_buffer(eb);
1248                 }
1249                 ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
1250                 if (ret > 0)
1251                         ret = -ENOENT;
1252                 if (ret)
1253                         break;
1254
1255                 next_inum = found_key.offset;
1256
1257                 /* regular exit ahead */
1258                 if (parent == next_inum)
1259                         break;
1260
1261                 slot = path->slots[0];
1262                 eb = path->nodes[0];
1263                 /* make sure we can use eb after releasing the path */
1264                 if (eb != eb_in) {
1265                         atomic_inc(&eb->refs);
1266                         btrfs_tree_read_lock(eb);
1267                         btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1268                 }
1269                 btrfs_release_path(path);
1270                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1271
1272                 name_len = btrfs_inode_ref_name_len(eb, iref);
1273                 name_off = (unsigned long)(iref + 1);
1274
1275                 parent = next_inum;
1276                 --bytes_left;
1277                 if (bytes_left >= 0)
1278                         dest[bytes_left] = '/';
1279         }
1280
1281         btrfs_release_path(path);
1282         path->leave_spinning = leave_spinning;
1283
1284         if (ret)
1285                 return ERR_PTR(ret);
1286
1287         return dest + bytes_left;
1288 }
1289
1290 /*
1291  * this makes the path point to (logical EXTENT_ITEM *)
1292  * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
1293  * tree blocks and <0 on error.
1294  */
1295 int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
1296                         struct btrfs_path *path, struct btrfs_key *found_key,
1297                         u64 *flags_ret)
1298 {
1299         int ret;
1300         u64 flags;
1301         u64 size = 0;
1302         u32 item_size;
1303         struct extent_buffer *eb;
1304         struct btrfs_extent_item *ei;
1305         struct btrfs_key key;
1306
1307         if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1308                 key.type = BTRFS_METADATA_ITEM_KEY;
1309         else
1310                 key.type = BTRFS_EXTENT_ITEM_KEY;
1311         key.objectid = logical;
1312         key.offset = (u64)-1;
1313
1314         ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
1315         if (ret < 0)
1316                 return ret;
1317
1318         while (1) {
1319                 u32 nritems;
1320                 if (path->slots[0] == 0) {
1321                         btrfs_set_path_blocking(path);
1322                         ret = btrfs_prev_leaf(fs_info->extent_root, path);
1323                         if (ret != 0) {
1324                                 if (ret > 0) {
1325                                         pr_debug("logical %llu is not within "
1326                                                  "any extent\n", logical);
1327                                         ret = -ENOENT;
1328                                 }
1329                                 return ret;
1330                         }
1331                 } else {
1332                         path->slots[0]--;
1333                 }
1334                 nritems = btrfs_header_nritems(path->nodes[0]);
1335                 if (nritems == 0) {
1336                         pr_debug("logical %llu is not within any extent\n",
1337                                  logical);
1338                         return -ENOENT;
1339                 }
1340                 if (path->slots[0] == nritems)
1341                         path->slots[0]--;
1342
1343                 btrfs_item_key_to_cpu(path->nodes[0], found_key,
1344                                       path->slots[0]);
1345                 if (found_key->type == BTRFS_EXTENT_ITEM_KEY ||
1346                     found_key->type == BTRFS_METADATA_ITEM_KEY)
1347                         break;
1348         }
1349
1350         if (found_key->type == BTRFS_METADATA_ITEM_KEY)
1351                 size = fs_info->extent_root->leafsize;
1352         else if (found_key->type == BTRFS_EXTENT_ITEM_KEY)
1353                 size = found_key->offset;
1354
1355         if (found_key->objectid > logical ||
1356             found_key->objectid + size <= logical) {
1357                 pr_debug("logical %llu is not within any extent\n", logical);
1358                 return -ENOENT;
1359         }
1360
1361         eb = path->nodes[0];
1362         item_size = btrfs_item_size_nr(eb, path->slots[0]);
1363         BUG_ON(item_size < sizeof(*ei));
1364
1365         ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
1366         flags = btrfs_extent_flags(eb, ei);
1367
1368         pr_debug("logical %llu is at position %llu within the extent (%llu "
1369                  "EXTENT_ITEM %llu) flags %#llx size %u\n",
1370                  logical, logical - found_key->objectid, found_key->objectid,
1371                  found_key->offset, flags, item_size);
1372
1373         WARN_ON(!flags_ret);
1374         if (flags_ret) {
1375                 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1376                         *flags_ret = BTRFS_EXTENT_FLAG_TREE_BLOCK;
1377                 else if (flags & BTRFS_EXTENT_FLAG_DATA)
1378                         *flags_ret = BTRFS_EXTENT_FLAG_DATA;
1379                 else
1380                         BUG_ON(1);
1381                 return 0;
1382         }
1383
1384         return -EIO;
1385 }
1386
1387 /*
1388  * helper function to iterate extent inline refs. ptr must point to a 0 value
1389  * for the first call and may be modified. it is used to track state.
1390  * if more refs exist, 0 is returned and the next call to
1391  * __get_extent_inline_ref must pass the modified ptr parameter to get the
1392  * next ref. after the last ref was processed, 1 is returned.
1393  * returns <0 on error
1394  */
1395 static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
1396                                 struct btrfs_extent_item *ei, u32 item_size,
1397                                 struct btrfs_extent_inline_ref **out_eiref,
1398                                 int *out_type)
1399 {
1400         unsigned long end;
1401         u64 flags;
1402         struct btrfs_tree_block_info *info;
1403
1404         if (!*ptr) {
1405                 /* first call */
1406                 flags = btrfs_extent_flags(eb, ei);
1407                 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1408                         info = (struct btrfs_tree_block_info *)(ei + 1);
1409                         *out_eiref =
1410                                 (struct btrfs_extent_inline_ref *)(info + 1);
1411                 } else {
1412                         *out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
1413                 }
1414                 *ptr = (unsigned long)*out_eiref;
1415                 if ((void *)*ptr >= (void *)ei + item_size)
1416                         return -ENOENT;
1417         }
1418
1419         end = (unsigned long)ei + item_size;
1420         *out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
1421         *out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
1422
1423         *ptr += btrfs_extent_inline_ref_size(*out_type);
1424         WARN_ON(*ptr > end);
1425         if (*ptr == end)
1426                 return 1; /* last */
1427
1428         return 0;
1429 }
1430
1431 /*
1432  * reads the tree block backref for an extent. tree level and root are returned
1433  * through out_level and out_root. ptr must point to a 0 value for the first
1434  * call and may be modified (see __get_extent_inline_ref comment).
1435  * returns 0 if data was provided, 1 if there was no more data to provide or
1436  * <0 on error.
1437  */
1438 int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
1439                                 struct btrfs_extent_item *ei, u32 item_size,
1440                                 u64 *out_root, u8 *out_level)
1441 {
1442         int ret;
1443         int type;
1444         struct btrfs_tree_block_info *info;
1445         struct btrfs_extent_inline_ref *eiref;
1446
1447         if (*ptr == (unsigned long)-1)
1448                 return 1;
1449
1450         while (1) {
1451                 ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
1452                                                 &eiref, &type);
1453                 if (ret < 0)
1454                         return ret;
1455
1456                 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
1457                     type == BTRFS_SHARED_BLOCK_REF_KEY)
1458                         break;
1459
1460                 if (ret == 1)
1461                         return 1;
1462         }
1463
1464         /* we can treat both ref types equally here */
1465         info = (struct btrfs_tree_block_info *)(ei + 1);
1466         *out_root = btrfs_extent_inline_ref_offset(eb, eiref);
1467         *out_level = btrfs_tree_block_level(eb, info);
1468
1469         if (ret == 1)
1470                 *ptr = (unsigned long)-1;
1471
1472         return 0;
1473 }
1474
1475 static int iterate_leaf_refs(struct extent_inode_elem *inode_list,
1476                                 u64 root, u64 extent_item_objectid,
1477                                 iterate_extent_inodes_t *iterate, void *ctx)
1478 {
1479         struct extent_inode_elem *eie;
1480         int ret = 0;
1481
1482         for (eie = inode_list; eie; eie = eie->next) {
1483                 pr_debug("ref for %llu resolved, key (%llu EXTEND_DATA %llu), "
1484                          "root %llu\n", extent_item_objectid,
1485                          eie->inum, eie->offset, root);
1486                 ret = iterate(eie->inum, eie->offset, root, ctx);
1487                 if (ret) {
1488                         pr_debug("stopping iteration for %llu due to ret=%d\n",
1489                                  extent_item_objectid, ret);
1490                         break;
1491                 }
1492         }
1493
1494         return ret;
1495 }
1496
1497 /*
1498  * calls iterate() for every inode that references the extent identified by
1499  * the given parameters.
1500  * when the iterator function returns a non-zero value, iteration stops.
1501  */
1502 int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
1503                                 u64 extent_item_objectid, u64 extent_item_pos,
1504                                 int search_commit_root,
1505                                 iterate_extent_inodes_t *iterate, void *ctx)
1506 {
1507         int ret;
1508         struct btrfs_trans_handle *trans = NULL;
1509         struct ulist *refs = NULL;
1510         struct ulist *roots = NULL;
1511         struct ulist_node *ref_node = NULL;
1512         struct ulist_node *root_node = NULL;
1513         struct seq_list tree_mod_seq_elem = {};
1514         struct ulist_iterator ref_uiter;
1515         struct ulist_iterator root_uiter;
1516
1517         pr_debug("resolving all inodes for extent %llu\n",
1518                         extent_item_objectid);
1519
1520         if (!search_commit_root) {
1521                 trans = btrfs_join_transaction(fs_info->extent_root);
1522                 if (IS_ERR(trans))
1523                         return PTR_ERR(trans);
1524                 btrfs_get_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1525         }
1526
1527         ret = btrfs_find_all_leafs(trans, fs_info, extent_item_objectid,
1528                                    tree_mod_seq_elem.seq, &refs,
1529                                    &extent_item_pos);
1530         if (ret)
1531                 goto out;
1532
1533         ULIST_ITER_INIT(&ref_uiter);
1534         while (!ret && (ref_node = ulist_next(refs, &ref_uiter))) {
1535                 ret = btrfs_find_all_roots(trans, fs_info, ref_node->val,
1536                                            tree_mod_seq_elem.seq, &roots);
1537                 if (ret)
1538                         break;
1539                 ULIST_ITER_INIT(&root_uiter);
1540                 while (!ret && (root_node = ulist_next(roots, &root_uiter))) {
1541                         pr_debug("root %llu references leaf %llu, data list "
1542                                  "%#llx\n", root_node->val, ref_node->val,
1543                                  ref_node->aux);
1544                         ret = iterate_leaf_refs((struct extent_inode_elem *)
1545                                                 (uintptr_t)ref_node->aux,
1546                                                 root_node->val,
1547                                                 extent_item_objectid,
1548                                                 iterate, ctx);
1549                 }
1550                 ulist_free(roots);
1551         }
1552
1553         free_leaf_list(refs);
1554 out:
1555         if (!search_commit_root) {
1556                 btrfs_put_tree_mod_seq(fs_info, &tree_mod_seq_elem);
1557                 btrfs_end_transaction(trans, fs_info->extent_root);
1558         }
1559
1560         return ret;
1561 }
1562
1563 int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
1564                                 struct btrfs_path *path,
1565                                 iterate_extent_inodes_t *iterate, void *ctx)
1566 {
1567         int ret;
1568         u64 extent_item_pos;
1569         u64 flags = 0;
1570         struct btrfs_key found_key;
1571         int search_commit_root = path->search_commit_root;
1572
1573         ret = extent_from_logical(fs_info, logical, path, &found_key, &flags);
1574         btrfs_release_path(path);
1575         if (ret < 0)
1576                 return ret;
1577         if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
1578                 return -EINVAL;
1579
1580         extent_item_pos = logical - found_key.objectid;
1581         ret = iterate_extent_inodes(fs_info, found_key.objectid,
1582                                         extent_item_pos, search_commit_root,
1583                                         iterate, ctx);
1584
1585         return ret;
1586 }
1587
1588 typedef int (iterate_irefs_t)(u64 parent, u32 name_len, unsigned long name_off,
1589                               struct extent_buffer *eb, void *ctx);
1590
1591 static int iterate_inode_refs(u64 inum, struct btrfs_root *fs_root,
1592                               struct btrfs_path *path,
1593                               iterate_irefs_t *iterate, void *ctx)
1594 {
1595         int ret = 0;
1596         int slot;
1597         u32 cur;
1598         u32 len;
1599         u32 name_len;
1600         u64 parent = 0;
1601         int found = 0;
1602         struct extent_buffer *eb;
1603         struct btrfs_item *item;
1604         struct btrfs_inode_ref *iref;
1605         struct btrfs_key found_key;
1606
1607         while (!ret) {
1608                 ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
1609                                      &found_key);
1610                 if (ret < 0)
1611                         break;
1612                 if (ret) {
1613                         ret = found ? 0 : -ENOENT;
1614                         break;
1615                 }
1616                 ++found;
1617
1618                 parent = found_key.offset;
1619                 slot = path->slots[0];
1620                 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1621                 if (!eb) {
1622                         ret = -ENOMEM;
1623                         break;
1624                 }
1625                 extent_buffer_get(eb);
1626                 btrfs_tree_read_lock(eb);
1627                 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1628                 btrfs_release_path(path);
1629
1630                 item = btrfs_item_nr(slot);
1631                 iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
1632
1633                 for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
1634                         name_len = btrfs_inode_ref_name_len(eb, iref);
1635                         /* path must be released before calling iterate()! */
1636                         pr_debug("following ref at offset %u for inode %llu in "
1637                                  "tree %llu\n", cur, found_key.objectid,
1638                                  fs_root->objectid);
1639                         ret = iterate(parent, name_len,
1640                                       (unsigned long)(iref + 1), eb, ctx);
1641                         if (ret)
1642                                 break;
1643                         len = sizeof(*iref) + name_len;
1644                         iref = (struct btrfs_inode_ref *)((char *)iref + len);
1645                 }
1646                 btrfs_tree_read_unlock_blocking(eb);
1647                 free_extent_buffer(eb);
1648         }
1649
1650         btrfs_release_path(path);
1651
1652         return ret;
1653 }
1654
1655 static int iterate_inode_extrefs(u64 inum, struct btrfs_root *fs_root,
1656                                  struct btrfs_path *path,
1657                                  iterate_irefs_t *iterate, void *ctx)
1658 {
1659         int ret;
1660         int slot;
1661         u64 offset = 0;
1662         u64 parent;
1663         int found = 0;
1664         struct extent_buffer *eb;
1665         struct btrfs_inode_extref *extref;
1666         struct extent_buffer *leaf;
1667         u32 item_size;
1668         u32 cur_offset;
1669         unsigned long ptr;
1670
1671         while (1) {
1672                 ret = btrfs_find_one_extref(fs_root, inum, offset, path, &extref,
1673                                             &offset);
1674                 if (ret < 0)
1675                         break;
1676                 if (ret) {
1677                         ret = found ? 0 : -ENOENT;
1678                         break;
1679                 }
1680                 ++found;
1681
1682                 slot = path->slots[0];
1683                 eb = btrfs_clone_extent_buffer(path->nodes[0]);
1684                 if (!eb) {
1685                         ret = -ENOMEM;
1686                         break;
1687                 }
1688                 extent_buffer_get(eb);
1689
1690                 btrfs_tree_read_lock(eb);
1691                 btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);
1692                 btrfs_release_path(path);
1693
1694                 leaf = path->nodes[0];
1695                 item_size = btrfs_item_size_nr(leaf, slot);
1696                 ptr = btrfs_item_ptr_offset(leaf, slot);
1697                 cur_offset = 0;
1698
1699                 while (cur_offset < item_size) {
1700                         u32 name_len;
1701
1702                         extref = (struct btrfs_inode_extref *)(ptr + cur_offset);
1703                         parent = btrfs_inode_extref_parent(eb, extref);
1704                         name_len = btrfs_inode_extref_name_len(eb, extref);
1705                         ret = iterate(parent, name_len,
1706                                       (unsigned long)&extref->name, eb, ctx);
1707                         if (ret)
1708                                 break;
1709
1710                         cur_offset += btrfs_inode_extref_name_len(leaf, extref);
1711                         cur_offset += sizeof(*extref);
1712                 }
1713                 btrfs_tree_read_unlock_blocking(eb);
1714                 free_extent_buffer(eb);
1715
1716                 offset++;
1717         }
1718
1719         btrfs_release_path(path);
1720
1721         return ret;
1722 }
1723
1724 static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
1725                          struct btrfs_path *path, iterate_irefs_t *iterate,
1726                          void *ctx)
1727 {
1728         int ret;
1729         int found_refs = 0;
1730
1731         ret = iterate_inode_refs(inum, fs_root, path, iterate, ctx);
1732         if (!ret)
1733                 ++found_refs;
1734         else if (ret != -ENOENT)
1735                 return ret;
1736
1737         ret = iterate_inode_extrefs(inum, fs_root, path, iterate, ctx);
1738         if (ret == -ENOENT && found_refs)
1739                 return 0;
1740
1741         return ret;
1742 }
1743
1744 /*
1745  * returns 0 if the path could be dumped (probably truncated)
1746  * returns <0 in case of an error
1747  */
1748 static int inode_to_path(u64 inum, u32 name_len, unsigned long name_off,
1749                          struct extent_buffer *eb, void *ctx)
1750 {
1751         struct inode_fs_paths *ipath = ctx;
1752         char *fspath;
1753         char *fspath_min;
1754         int i = ipath->fspath->elem_cnt;
1755         const int s_ptr = sizeof(char *);
1756         u32 bytes_left;
1757
1758         bytes_left = ipath->fspath->bytes_left > s_ptr ?
1759                                         ipath->fspath->bytes_left - s_ptr : 0;
1760
1761         fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
1762         fspath = btrfs_ref_to_path(ipath->fs_root, ipath->btrfs_path, name_len,
1763                                    name_off, eb, inum, fspath_min, bytes_left);
1764         if (IS_ERR(fspath))
1765                 return PTR_ERR(fspath);
1766
1767         if (fspath > fspath_min) {
1768                 ipath->fspath->val[i] = (u64)(unsigned long)fspath;
1769                 ++ipath->fspath->elem_cnt;
1770                 ipath->fspath->bytes_left = fspath - fspath_min;
1771         } else {
1772                 ++ipath->fspath->elem_missed;
1773                 ipath->fspath->bytes_missing += fspath_min - fspath;
1774                 ipath->fspath->bytes_left = 0;
1775         }
1776
1777         return 0;
1778 }
1779
1780 /*
1781  * this dumps all file system paths to the inode into the ipath struct, provided
1782  * is has been created large enough. each path is zero-terminated and accessed
1783  * from ipath->fspath->val[i].
1784  * when it returns, there are ipath->fspath->elem_cnt number of paths available
1785  * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
1786  * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
1787  * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
1788  * have been needed to return all paths.
1789  */
1790 int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
1791 {
1792         return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
1793                              inode_to_path, ipath);
1794 }
1795
1796 struct btrfs_data_container *init_data_container(u32 total_bytes)
1797 {
1798         struct btrfs_data_container *data;
1799         size_t alloc_bytes;
1800
1801         alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
1802         data = vmalloc(alloc_bytes);
1803         if (!data)
1804                 return ERR_PTR(-ENOMEM);
1805
1806         if (total_bytes >= sizeof(*data)) {
1807                 data->bytes_left = total_bytes - sizeof(*data);
1808                 data->bytes_missing = 0;
1809         } else {
1810                 data->bytes_missing = sizeof(*data) - total_bytes;
1811                 data->bytes_left = 0;
1812         }
1813
1814         data->elem_cnt = 0;
1815         data->elem_missed = 0;
1816
1817         return data;
1818 }
1819
1820 /*
1821  * allocates space to return multiple file system paths for an inode.
1822  * total_bytes to allocate are passed, note that space usable for actual path
1823  * information will be total_bytes - sizeof(struct inode_fs_paths).
1824  * the returned pointer must be freed with free_ipath() in the end.
1825  */
1826 struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
1827                                         struct btrfs_path *path)
1828 {
1829         struct inode_fs_paths *ifp;
1830         struct btrfs_data_container *fspath;
1831
1832         fspath = init_data_container(total_bytes);
1833         if (IS_ERR(fspath))
1834                 return (void *)fspath;
1835
1836         ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
1837         if (!ifp) {
1838                 kfree(fspath);
1839                 return ERR_PTR(-ENOMEM);
1840         }
1841
1842         ifp->btrfs_path = path;
1843         ifp->fspath = fspath;
1844         ifp->fs_root = fs_root;
1845
1846         return ifp;
1847 }
1848
1849 void free_ipath(struct inode_fs_paths *ipath)
1850 {
1851         if (!ipath)
1852                 return;
1853         vfree(ipath->fspath);
1854         kfree(ipath);
1855 }