fs/btrfs/ctree.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2007,2008 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/sched.h>
   7 #include <linux/slab.h>
   8 #include <linux/rbtree.h>
   9 #include <linux/mm.h>
  10 #include "ctree.h"
  11 #include "disk-io.h"
  12 #include "transaction.h"
  13 #include "print-tree.h"
  14 #include "locking.h"
  15 #include "volumes.h"
  16 #include "qgroup.h"
  17
  18 static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
  19                       *root, struct btrfs_path *path, int level);
  20 static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root,
  21                       const struct btrfs_key *ins_key, struct btrfs_path *path,
  22                       int data_size, int extend);
  23 static int push_node_left(struct btrfs_trans_handle *trans,
  24                           struct extent_buffer *dst,
  25                           struct extent_buffer *src, int empty);
  26 static int balance_node_right(struct btrfs_trans_handle *trans,
  27                               struct extent_buffer *dst_buf,
  28                               struct extent_buffer *src_buf);
  29 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
  30                     int level, int slot);
  31
  32 static const struct btrfs_csums {
  33         u16             size;
  34         const char      name[10];
  35         const char      driver[12];
  36 } btrfs_csums[] = {
  37         [BTRFS_CSUM_TYPE_CRC32] = { .size = 4, .name = "crc32c" },
  38         [BTRFS_CSUM_TYPE_XXHASH] = { .size = 8, .name = "xxhash64" },
  39         [BTRFS_CSUM_TYPE_SHA256] = { .size = 32, .name = "sha256" },
  40         [BTRFS_CSUM_TYPE_BLAKE2] = { .size = 32, .name = "blake2b",
  41                                      .driver = "blake2b-256" },
  42 };
  43
  44 int btrfs_super_csum_size(const struct btrfs_super_block *s)
  45 {
  46         u16 t = btrfs_super_csum_type(s);
  47         /*
  48          * csum type is validated at mount time
  49          */
  50         return btrfs_csums[t].size;
  51 }
  52
  53 const char *btrfs_super_csum_name(u16 csum_type)
  54 {
  55         /* csum type is validated at mount time */
  56         return btrfs_csums[csum_type].name;
  57 }
  58
  59 /*
  60  * Return driver name if defined, otherwise the name that's also a valid driver
  61  * name
  62  */
  63 const char *btrfs_super_csum_driver(u16 csum_type)
  64 {
  65         /* csum type is validated at mount time */
  66         return btrfs_csums[csum_type].driver[0] ?
  67                 btrfs_csums[csum_type].driver :
  68                 btrfs_csums[csum_type].name;
  69 }
  70
  71 size_t __const btrfs_get_num_csums(void)
  72 {
  73         return ARRAY_SIZE(btrfs_csums);
  74 }
  75
  76 struct btrfs_path *btrfs_alloc_path(void)
  77 {
  78         return kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS);
  79 }
  80
  81 /* this also releases the path */
  82 void btrfs_free_path(struct btrfs_path *p)
  83 {
  84         if (!p)
  85                 return;
  86         btrfs_release_path(p);
  87         kmem_cache_free(btrfs_path_cachep, p);
  88 }
  89
  90 /*
  91  * path release drops references on the extent buffers in the path
  92  * and it drops any locks held by this path
  93  *
  94  * It is safe to call this on paths that no locks or extent buffers held.
  95  */
  96 noinline void btrfs_release_path(struct btrfs_path *p)
  97 {
  98         int i;
  99
 100         for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
 101                 p->slots[i] = 0;
 102                 if (!p->nodes[i])
 103                         continue;
 104                 if (p->locks[i]) {
 105                         btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]);
 106                         p->locks[i] = 0;
 107                 }
 108                 free_extent_buffer(p->nodes[i]);
 109                 p->nodes[i] = NULL;
 110         }
 111 }
 112
 113 /*
 114  * safely gets a reference on the root node of a tree.  A lock
 115  * is not taken, so a concurrent writer may put a different node
 116  * at the root of the tree.  See btrfs_lock_root_node for the
 117  * looping required.
 118  *
 119  * The extent buffer returned by this has a reference taken, so
 120  * it won't disappear.  It may stop being the root of the tree
 121  * at any time because there are no locks held.
 122  */
 123 struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
 124 {
 125         struct extent_buffer *eb;
 126
 127         while (1) {
 128                 rcu_read_lock();
 129                 eb = rcu_dereference(root->node);
 130
 131                 /*
 132                  * RCU really hurts here, we could free up the root node because
 133                  * it was COWed but we may not get the new root node yet so do
 134                  * the inc_not_zero dance and if it doesn't work then
 135                  * synchronize_rcu and try again.
 136                  */
 137                 if (atomic_inc_not_zero(&eb->refs)) {
 138                         rcu_read_unlock();
 139                         break;
 140                 }
 141                 rcu_read_unlock();
 142                 synchronize_rcu();
 143         }
 144         return eb;
 145 }
 146
 147 /* cowonly root (everything not a reference counted cow subvolume), just get
 148  * put onto a simple dirty list.  transaction.c walks this to make sure they
 149  * get properly updated on disk.
 150  */
 151 static void add_root_to_dirty_list(struct btrfs_root *root)
 152 {
 153         struct btrfs_fs_info *fs_info = root->fs_info;
 154
 155         if (test_bit(BTRFS_ROOT_DIRTY, &root->state) ||
 156             !test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state))
 157                 return;
 158
 159         spin_lock(&fs_info->trans_lock);
 160         if (!test_and_set_bit(BTRFS_ROOT_DIRTY, &root->state)) {
 161                 /* Want the extent tree to be the last on the list */
 162                 if (root->root_key.objectid == BTRFS_EXTENT_TREE_OBJECTID)
 163                         list_move_tail(&root->dirty_list,
 164                                        &fs_info->dirty_cowonly_roots);
 165                 else
 166                         list_move(&root->dirty_list,
 167                                   &fs_info->dirty_cowonly_roots);
 168         }
 169         spin_unlock(&fs_info->trans_lock);
 170 }
 171
 172 /*
 173  * used by snapshot creation to make a copy of a root for a tree with
 174  * a given objectid.  The buffer with the new root node is returned in
 175  * cow_ret, and this func returns zero on success or a negative error code.
 176  */
 177 int btrfs_copy_root(struct btrfs_trans_handle *trans,
 178                       struct btrfs_root *root,
 179                       struct extent_buffer *buf,
 180                       struct extent_buffer **cow_ret, u64 new_root_objectid)
 181 {
 182         struct btrfs_fs_info *fs_info = root->fs_info;
 183         struct extent_buffer *cow;
 184         int ret = 0;
 185         int level;
 186         struct btrfs_disk_key disk_key;
 187
 188         WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 189                 trans->transid != fs_info->running_transaction->transid);
 190         WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 191                 trans->transid != root->last_trans);
 192
 193         level = btrfs_header_level(buf);
 194         if (level == 0)
 195                 btrfs_item_key(buf, &disk_key, 0);
 196         else
 197                 btrfs_node_key(buf, &disk_key, 0);
 198
 199         cow = btrfs_alloc_tree_block(trans, root, 0, new_root_objectid,
 200                         &disk_key, level, buf->start, 0);
 201         if (IS_ERR(cow))
 202                 return PTR_ERR(cow);
 203
 204         copy_extent_buffer_full(cow, buf);
 205         btrfs_set_header_bytenr(cow, cow->start);
 206         btrfs_set_header_generation(cow, trans->transid);
 207         btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
 208         btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
 209                                      BTRFS_HEADER_FLAG_RELOC);
 210         if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
 211                 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
 212         else
 213                 btrfs_set_header_owner(cow, new_root_objectid);
 214
 215         write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);
 216
 217         WARN_ON(btrfs_header_generation(buf) > trans->transid);
 218         if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
 219                 ret = btrfs_inc_ref(trans, root, cow, 1);
 220         else
 221                 ret = btrfs_inc_ref(trans, root, cow, 0);
 222
 223         if (ret)
 224                 return ret;
 225
 226         btrfs_mark_buffer_dirty(cow);
 227         *cow_ret = cow;
 228         return 0;
 229 }
 230
 231 enum mod_log_op {
 232         MOD_LOG_KEY_REPLACE,
 233         MOD_LOG_KEY_ADD,
 234         MOD_LOG_KEY_REMOVE,
 235         MOD_LOG_KEY_REMOVE_WHILE_FREEING,
 236         MOD_LOG_KEY_REMOVE_WHILE_MOVING,
 237         MOD_LOG_MOVE_KEYS,
 238         MOD_LOG_ROOT_REPLACE,
 239 };
 240
 241 struct tree_mod_root {
 242         u64 logical;
 243         u8 level;
 244 };
 245
 246 struct tree_mod_elem {
 247         struct rb_node node;
 248         u64 logical;
 249         u64 seq;
 250         enum mod_log_op op;
 251
 252         /* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */
 253         int slot;
 254
 255         /* this is used for MOD_LOG_KEY* and MOD_LOG_ROOT_REPLACE */
 256         u64 generation;
 257
 258         /* those are used for op == MOD_LOG_KEY_{REPLACE,REMOVE} */
 259         struct btrfs_disk_key key;
 260         u64 blockptr;
 261
 262         /* this is used for op == MOD_LOG_MOVE_KEYS */
 263         struct {
 264                 int dst_slot;
 265                 int nr_items;
 266         } move;
 267
 268         /* this is used for op == MOD_LOG_ROOT_REPLACE */
 269         struct tree_mod_root old_root;
 270 };
 271
 272 /*
 273  * Pull a new tree mod seq number for our operation.
 274  */
 275 static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info)
 276 {
 277         return atomic64_inc_return(&fs_info->tree_mod_seq);
 278 }
 279
 280 /*
 281  * This adds a new blocker to the tree mod log's blocker list if the @elem
 282  * passed does not already have a sequence number set. So when a caller expects
 283  * to record tree modifications, it should ensure to set elem->seq to zero
 284  * before calling btrfs_get_tree_mod_seq.
 285  * Returns a fresh, unused tree log modification sequence number, even if no new
 286  * blocker was added.
 287  */
 288 u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
 289                            struct seq_list *elem)
 290 {
 291         write_lock(&fs_info->tree_mod_log_lock);
 292         if (!elem->seq) {
 293                 elem->seq = btrfs_inc_tree_mod_seq(fs_info);
 294                 list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
 295         }
 296         write_unlock(&fs_info->tree_mod_log_lock);
 297
 298         return elem->seq;
 299 }
 300
 301 void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
 302                             struct seq_list *elem)
 303 {
 304         struct rb_root *tm_root;
 305         struct rb_node *node;
 306         struct rb_node *next;
 307         struct tree_mod_elem *tm;
 308         u64 min_seq = (u64)-1;
 309         u64 seq_putting = elem->seq;
 310
 311         if (!seq_putting)
 312                 return;
 313
 314         write_lock(&fs_info->tree_mod_log_lock);
 315         list_del(&elem->list);
 316         elem->seq = 0;
 317
 318         if (!list_empty(&fs_info->tree_mod_seq_list)) {
 319                 struct seq_list *first;
 320
 321                 first = list_first_entry(&fs_info->tree_mod_seq_list,
 322                                          struct seq_list, list);
 323                 if (seq_putting > first->seq) {
 324                         /*
 325                          * Blocker with lower sequence number exists, we
 326                          * cannot remove anything from the log.
 327                          */
 328                         write_unlock(&fs_info->tree_mod_log_lock);
 329                         return;
 330                 }
 331                 min_seq = first->seq;
 332         }
 333
 334         /*
 335          * anything that's lower than the lowest existing (read: blocked)
 336          * sequence number can be removed from the tree.
 337          */
 338         tm_root = &fs_info->tree_mod_log;
 339         for (node = rb_first(tm_root); node; node = next) {
 340                 next = rb_next(node);
 341                 tm = rb_entry(node, struct tree_mod_elem, node);
 342                 if (tm->seq >= min_seq)
 343                         continue;
 344                 rb_erase(node, tm_root);
 345                 kfree(tm);
 346         }
 347         write_unlock(&fs_info->tree_mod_log_lock);
 348 }
 349
 350 /*
 351  * key order of the log:
 352  *       node/leaf start address -> sequence
 353  *
 354  * The 'start address' is the logical address of the *new* root node
 355  * for root replace operations, or the logical address of the affected
 356  * block for all other operations.
 357  */
 358 static noinline int
 359 __tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm)
 360 {
 361         struct rb_root *tm_root;
 362         struct rb_node **new;
 363         struct rb_node *parent = NULL;
 364         struct tree_mod_elem *cur;
 365
 366         lockdep_assert_held_write(&fs_info->tree_mod_log_lock);
 367
 368         tm->seq = btrfs_inc_tree_mod_seq(fs_info);
 369
 370         tm_root = &fs_info->tree_mod_log;
 371         new = &tm_root->rb_node;
 372         while (*new) {
 373                 cur = rb_entry(*new, struct tree_mod_elem, node);
 374                 parent = *new;
 375                 if (cur->logical < tm->logical)
 376                         new = &((*new)->rb_left);
 377                 else if (cur->logical > tm->logical)
 378                         new = &((*new)->rb_right);
 379                 else if (cur->seq < tm->seq)
 380                         new = &((*new)->rb_left);
 381                 else if (cur->seq > tm->seq)
 382                         new = &((*new)->rb_right);
 383                 else
 384                         return -EEXIST;
 385         }
 386
 387         rb_link_node(&tm->node, parent, new);
 388         rb_insert_color(&tm->node, tm_root);
 389         return 0;
 390 }
 391
 392 /*
 393  * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it
 394  * returns zero with the tree_mod_log_lock acquired. The caller must hold
 395  * this until all tree mod log insertions are recorded in the rb tree and then
 396  * write unlock fs_info::tree_mod_log_lock.
 397  */
 398 static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info,
 399                                     struct extent_buffer *eb) {
 400         smp_mb();
 401         if (list_empty(&(fs_info)->tree_mod_seq_list))
 402                 return 1;
 403         if (eb && btrfs_header_level(eb) == 0)
 404                 return 1;
 405
 406         write_lock(&fs_info->tree_mod_log_lock);
 407         if (list_empty(&(fs_info)->tree_mod_seq_list)) {
 408                 write_unlock(&fs_info->tree_mod_log_lock);
 409                 return 1;
 410         }
 411
 412         return 0;
 413 }
 414
 415 /* Similar to tree_mod_dont_log, but doesn't acquire any locks. */
 416 static inline int tree_mod_need_log(const struct btrfs_fs_info *fs_info,
 417                                     struct extent_buffer *eb)
 418 {
 419         smp_mb();
 420         if (list_empty(&(fs_info)->tree_mod_seq_list))
 421                 return 0;
 422         if (eb && btrfs_header_level(eb) == 0)
 423                 return 0;
 424
 425         return 1;
 426 }
 427
 428 static struct tree_mod_elem *
 429 alloc_tree_mod_elem(struct extent_buffer *eb, int slot,
 430                     enum mod_log_op op, gfp_t flags)
 431 {
 432         struct tree_mod_elem *tm;
 433
 434         tm = kzalloc(sizeof(*tm), flags);
 435         if (!tm)
 436                 return NULL;
 437
 438         tm->logical = eb->start;
 439         if (op != MOD_LOG_KEY_ADD) {
 440                 btrfs_node_key(eb, &tm->key, slot);
 441                 tm->blockptr = btrfs_node_blockptr(eb, slot);
 442         }
 443         tm->op = op;
 444         tm->slot = slot;
 445         tm->generation = btrfs_node_ptr_generation(eb, slot);
 446         RB_CLEAR_NODE(&tm->node);
 447
 448         return tm;
 449 }
 450
 451 static noinline int tree_mod_log_insert_key(struct extent_buffer *eb, int slot,
 452                 enum mod_log_op op, gfp_t flags)
 453 {
 454         struct tree_mod_elem *tm;
 455         int ret;
 456
 457         if (!tree_mod_need_log(eb->fs_info, eb))
 458                 return 0;
 459
 460         tm = alloc_tree_mod_elem(eb, slot, op, flags);
 461         if (!tm)
 462                 return -ENOMEM;
 463
 464         if (tree_mod_dont_log(eb->fs_info, eb)) {
 465                 kfree(tm);
 466                 return 0;
 467         }
 468
 469         ret = __tree_mod_log_insert(eb->fs_info, tm);
 470         write_unlock(&eb->fs_info->tree_mod_log_lock);
 471         if (ret)
 472                 kfree(tm);
 473
 474         return ret;
 475 }
 476
 477 static noinline int tree_mod_log_insert_move(struct extent_buffer *eb,
 478                 int dst_slot, int src_slot, int nr_items)
 479 {
 480         struct tree_mod_elem *tm = NULL;
 481         struct tree_mod_elem **tm_list = NULL;
 482         int ret = 0;
 483         int i;
 484         int locked = 0;
 485
 486         if (!tree_mod_need_log(eb->fs_info, eb))
 487                 return 0;
 488
 489         tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS);
 490         if (!tm_list)
 491                 return -ENOMEM;
 492
 493         tm = kzalloc(sizeof(*tm), GFP_NOFS);
 494         if (!tm) {
 495                 ret = -ENOMEM;
 496                 goto free_tms;
 497         }
 498
 499         tm->logical = eb->start;
 500         tm->slot = src_slot;
 501         tm->move.dst_slot = dst_slot;
 502         tm->move.nr_items = nr_items;
 503         tm->op = MOD_LOG_MOVE_KEYS;
 504
 505         for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
 506                 tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot,
 507                     MOD_LOG_KEY_REMOVE_WHILE_MOVING, GFP_NOFS);
 508                 if (!tm_list[i]) {
 509                         ret = -ENOMEM;
 510                         goto free_tms;
 511                 }
 512         }
 513
 514         if (tree_mod_dont_log(eb->fs_info, eb))
 515                 goto free_tms;
 516         locked = 1;
 517
 518         /*
 519          * When we override something during the move, we log these removals.
 520          * This can only happen when we move towards the beginning of the
 521          * buffer, i.e. dst_slot < src_slot.
 522          */
 523         for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
 524                 ret = __tree_mod_log_insert(eb->fs_info, tm_list[i]);
 525                 if (ret)
 526                         goto free_tms;
 527         }
 528
 529         ret = __tree_mod_log_insert(eb->fs_info, tm);
 530         if (ret)
 531                 goto free_tms;
 532         write_unlock(&eb->fs_info->tree_mod_log_lock);
 533         kfree(tm_list);
 534
 535         return 0;
 536 free_tms:
 537         for (i = 0; i < nr_items; i++) {
 538                 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
 539                         rb_erase(&tm_list[i]->node, &eb->fs_info->tree_mod_log);
 540                 kfree(tm_list[i]);
 541         }
 542         if (locked)
 543                 write_unlock(&eb->fs_info->tree_mod_log_lock);
 544         kfree(tm_list);
 545         kfree(tm);
 546
 547         return ret;
 548 }
 549
 550 static inline int
 551 __tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
 552                        struct tree_mod_elem **tm_list,
 553                        int nritems)
 554 {
 555         int i, j;
 556         int ret;
 557
 558         for (i = nritems - 1; i >= 0; i--) {
 559                 ret = __tree_mod_log_insert(fs_info, tm_list[i]);
 560                 if (ret) {
 561                         for (j = nritems - 1; j > i; j--)
 562                                 rb_erase(&tm_list[j]->node,
 563                                          &fs_info->tree_mod_log);
 564                         return ret;
 565                 }
 566         }
 567
 568         return 0;
 569 }
 570
 571 static noinline int tree_mod_log_insert_root(struct extent_buffer *old_root,
 572                          struct extent_buffer *new_root, int log_removal)
 573 {
 574         struct btrfs_fs_info *fs_info = old_root->fs_info;
 575         struct tree_mod_elem *tm = NULL;
 576         struct tree_mod_elem **tm_list = NULL;
 577         int nritems = 0;
 578         int ret = 0;
 579         int i;
 580
 581         if (!tree_mod_need_log(fs_info, NULL))
 582                 return 0;
 583
 584         if (log_removal && btrfs_header_level(old_root) > 0) {
 585                 nritems = btrfs_header_nritems(old_root);
 586                 tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *),
 587                                   GFP_NOFS);
 588                 if (!tm_list) {
 589                         ret = -ENOMEM;
 590                         goto free_tms;
 591                 }
 592                 for (i = 0; i < nritems; i++) {
 593                         tm_list[i] = alloc_tree_mod_elem(old_root, i,
 594                             MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS);
 595                         if (!tm_list[i]) {
 596                                 ret = -ENOMEM;
 597                                 goto free_tms;
 598                         }
 599                 }
 600         }
 601
 602         tm = kzalloc(sizeof(*tm), GFP_NOFS);
 603         if (!tm) {
 604                 ret = -ENOMEM;
 605                 goto free_tms;
 606         }
 607
 608         tm->logical = new_root->start;
 609         tm->old_root.logical = old_root->start;
 610         tm->old_root.level = btrfs_header_level(old_root);
 611         tm->generation = btrfs_header_generation(old_root);
 612         tm->op = MOD_LOG_ROOT_REPLACE;
 613
 614         if (tree_mod_dont_log(fs_info, NULL))
 615                 goto free_tms;
 616
 617         if (tm_list)
 618                 ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems);
 619         if (!ret)
 620                 ret = __tree_mod_log_insert(fs_info, tm);
 621
 622         write_unlock(&fs_info->tree_mod_log_lock);
 623         if (ret)
 624                 goto free_tms;
 625         kfree(tm_list);
 626
 627         return ret;
 628
 629 free_tms:
 630         if (tm_list) {
 631                 for (i = 0; i < nritems; i++)
 632                         kfree(tm_list[i]);
 633                 kfree(tm_list);
 634         }
 635         kfree(tm);
 636
 637         return ret;
 638 }
 639
 640 static struct tree_mod_elem *
 641 __tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
 642                       int smallest)
 643 {
 644         struct rb_root *tm_root;
 645         struct rb_node *node;
 646         struct tree_mod_elem *cur = NULL;
 647         struct tree_mod_elem *found = NULL;
 648
 649         read_lock(&fs_info->tree_mod_log_lock);
 650         tm_root = &fs_info->tree_mod_log;
 651         node = tm_root->rb_node;
 652         while (node) {
 653                 cur = rb_entry(node, struct tree_mod_elem, node);
 654                 if (cur->logical < start) {
 655                         node = node->rb_left;
 656                 } else if (cur->logical > start) {
 657                         node = node->rb_right;
 658                 } else if (cur->seq < min_seq) {
 659                         node = node->rb_left;
 660                 } else if (!smallest) {
 661                         /* we want the node with the highest seq */
 662                         if (found)
 663                                 BUG_ON(found->seq > cur->seq);
 664                         found = cur;
 665                         node = node->rb_left;
 666                 } else if (cur->seq > min_seq) {
 667                         /* we want the node with the smallest seq */
 668                         if (found)
 669                                 BUG_ON(found->seq < cur->seq);
 670                         found = cur;
 671                         node = node->rb_right;
 672                 } else {
 673                         found = cur;
 674                         break;
 675                 }
 676         }
 677         read_unlock(&fs_info->tree_mod_log_lock);
 678
 679         return found;
 680 }
 681
 682 /*
 683  * this returns the element from the log with the smallest time sequence
 684  * value that's in the log (the oldest log item). any element with a time
 685  * sequence lower than min_seq will be ignored.
 686  */
 687 static struct tree_mod_elem *
 688 tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info, u64 start,
 689                            u64 min_seq)
 690 {
 691         return __tree_mod_log_search(fs_info, start, min_seq, 1);
 692 }
 693
 694 /*
 695  * this returns the element from the log with the largest time sequence
 696  * value that's in the log (the most recent log item). any element with
 697  * a time sequence lower than min_seq will be ignored.
 698  */
 699 static struct tree_mod_elem *
 700 tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq)
 701 {
 702         return __tree_mod_log_search(fs_info, start, min_seq, 0);
 703 }
 704
 705 static noinline int tree_mod_log_eb_copy(struct extent_buffer *dst,
 706                      struct extent_buffer *src, unsigned long dst_offset,
 707                      unsigned long src_offset, int nr_items)
 708 {
 709         struct btrfs_fs_info *fs_info = dst->fs_info;
 710         int ret = 0;
 711         struct tree_mod_elem **tm_list = NULL;
 712         struct tree_mod_elem **tm_list_add, **tm_list_rem;
 713         int i;
 714         int locked = 0;
 715
 716         if (!tree_mod_need_log(fs_info, NULL))
 717                 return 0;
 718
 719         if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
 720                 return 0;
 721
 722         tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *),
 723                           GFP_NOFS);
 724         if (!tm_list)
 725                 return -ENOMEM;
 726
 727         tm_list_add = tm_list;
 728         tm_list_rem = tm_list + nr_items;
 729         for (i = 0; i < nr_items; i++) {
 730                 tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset,
 731                     MOD_LOG_KEY_REMOVE, GFP_NOFS);
 732                 if (!tm_list_rem[i]) {
 733                         ret = -ENOMEM;
 734                         goto free_tms;
 735                 }
 736
 737                 tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset,
 738                     MOD_LOG_KEY_ADD, GFP_NOFS);
 739                 if (!tm_list_add[i]) {
 740                         ret = -ENOMEM;
 741                         goto free_tms;
 742                 }
 743         }
 744
 745         if (tree_mod_dont_log(fs_info, NULL))
 746                 goto free_tms;
 747         locked = 1;
 748
 749         for (i = 0; i < nr_items; i++) {
 750                 ret = __tree_mod_log_insert(fs_info, tm_list_rem[i]);
 751                 if (ret)
 752                         goto free_tms;
 753                 ret = __tree_mod_log_insert(fs_info, tm_list_add[i]);
 754                 if (ret)
 755                         goto free_tms;
 756         }
 757
 758         write_unlock(&fs_info->tree_mod_log_lock);
 759         kfree(tm_list);
 760
 761         return 0;
 762
 763 free_tms:
 764         for (i = 0; i < nr_items * 2; i++) {
 765                 if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
 766                         rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
 767                 kfree(tm_list[i]);
 768         }
 769         if (locked)
 770                 write_unlock(&fs_info->tree_mod_log_lock);
 771         kfree(tm_list);
 772
 773         return ret;
 774 }
 775
 776 static noinline int tree_mod_log_free_eb(struct extent_buffer *eb)
 777 {
 778         struct tree_mod_elem **tm_list = NULL;
 779         int nritems = 0;
 780         int i;
 781         int ret = 0;
 782
 783         if (btrfs_header_level(eb) == 0)
 784                 return 0;
 785
 786         if (!tree_mod_need_log(eb->fs_info, NULL))
 787                 return 0;
 788
 789         nritems = btrfs_header_nritems(eb);
 790         tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS);
 791         if (!tm_list)
 792                 return -ENOMEM;
 793
 794         for (i = 0; i < nritems; i++) {
 795                 tm_list[i] = alloc_tree_mod_elem(eb, i,
 796                     MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS);
 797                 if (!tm_list[i]) {
 798                         ret = -ENOMEM;
 799                         goto free_tms;
 800                 }
 801         }
 802
 803         if (tree_mod_dont_log(eb->fs_info, eb))
 804                 goto free_tms;
 805
 806         ret = __tree_mod_log_free_eb(eb->fs_info, tm_list, nritems);
 807         write_unlock(&eb->fs_info->tree_mod_log_lock);
 808         if (ret)
 809                 goto free_tms;
 810         kfree(tm_list);
 811
 812         return 0;
 813
 814 free_tms:
 815         for (i = 0; i < nritems; i++)
 816                 kfree(tm_list[i]);
 817         kfree(tm_list);
 818
 819         return ret;
 820 }
 821
 822 /*
 823  * check if the tree block can be shared by multiple trees
 824  */
 825 int btrfs_block_can_be_shared(struct btrfs_root *root,
 826                               struct extent_buffer *buf)
 827 {
 828         /*
 829          * Tree blocks not in reference counted trees and tree roots
 830          * are never shared. If a block was allocated after the last
 831          * snapshot and the block was not allocated by tree relocation,
 832          * we know the block is not shared.
 833          */
 834         if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
 835             buf != root->node && buf != root->commit_root &&
 836             (btrfs_header_generation(buf) <=
 837              btrfs_root_last_snapshot(&root->root_item) ||
 838              btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
 839                 return 1;
 840
 841         return 0;
 842 }
 843
 844 static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
 845                                        struct btrfs_root *root,
 846                                        struct extent_buffer *buf,
 847                                        struct extent_buffer *cow,
 848                                        int *last_ref)
 849 {
 850         struct btrfs_fs_info *fs_info = root->fs_info;
 851         u64 refs;
 852         u64 owner;
 853         u64 flags;
 854         u64 new_flags = 0;
 855         int ret;
 856
 857         /*
 858          * Backrefs update rules:
 859          *
 860          * Always use full backrefs for extent pointers in tree block
 861          * allocated by tree relocation.
 862          *
 863          * If a shared tree block is no longer referenced by its owner
 864          * tree (btrfs_header_owner(buf) == root->root_key.objectid),
 865          * use full backrefs for extent pointers in tree block.
 866          *
 867          * If a tree block is been relocating
 868          * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID),
 869          * use full backrefs for extent pointers in tree block.
 870          * The reason for this is some operations (such as drop tree)
 871          * are only allowed for blocks use full backrefs.
 872          */
 873
 874         if (btrfs_block_can_be_shared(root, buf)) {
 875                 ret = btrfs_lookup_extent_info(trans, fs_info, buf->start,
 876                                                btrfs_header_level(buf), 1,
 877                                                &refs, &flags);
 878                 if (ret)
 879                         return ret;
 880                 if (refs == 0) {
 881                         ret = -EROFS;
 882                         btrfs_handle_fs_error(fs_info, ret, NULL);
 883                         return ret;
 884                 }
 885         } else {
 886                 refs = 1;
 887                 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
 888                     btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
 889                         flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
 890                 else
 891                         flags = 0;
 892         }
 893
 894         owner = btrfs_header_owner(buf);
 895         BUG_ON(owner == BTRFS_TREE_RELOC_OBJECTID &&
 896                !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
 897
 898         if (refs > 1) {
 899                 if ((owner == root->root_key.objectid ||
 900                      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) &&
 901                     !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
 902                         ret = btrfs_inc_ref(trans, root, buf, 1);
 903                         if (ret)
 904                                 return ret;
 905
 906                         if (root->root_key.objectid ==
 907                             BTRFS_TREE_RELOC_OBJECTID) {
 908                                 ret = btrfs_dec_ref(trans, root, buf, 0);
 909                                 if (ret)
 910                                         return ret;
 911                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 912                                 if (ret)
 913                                         return ret;
 914                         }
 915                         new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
 916                 } else {
 917
 918                         if (root->root_key.objectid ==
 919                             BTRFS_TREE_RELOC_OBJECTID)
 920                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 921                         else
 922                                 ret = btrfs_inc_ref(trans, root, cow, 0);
 923                         if (ret)
 924                                 return ret;
 925                 }
 926                 if (new_flags != 0) {
 927                         int level = btrfs_header_level(buf);
 928
 929                         ret = btrfs_set_disk_extent_flags(trans, buf,
 930                                                           new_flags, level, 0);
 931                         if (ret)
 932                                 return ret;
 933                 }
 934         } else {
 935                 if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
 936                         if (root->root_key.objectid ==
 937                             BTRFS_TREE_RELOC_OBJECTID)
 938                                 ret = btrfs_inc_ref(trans, root, cow, 1);
 939                         else
 940                                 ret = btrfs_inc_ref(trans, root, cow, 0);
 941                         if (ret)
 942                                 return ret;
 943                         ret = btrfs_dec_ref(trans, root, buf, 1);
 944                         if (ret)
 945                                 return ret;
 946                 }
 947                 btrfs_clean_tree_block(buf);
 948                 *last_ref = 1;
 949         }
 950         return 0;
 951 }
 952
 953 static struct extent_buffer *alloc_tree_block_no_bg_flush(
 954                                           struct btrfs_trans_handle *trans,
 955                                           struct btrfs_root *root,
 956                                           u64 parent_start,
 957                                           const struct btrfs_disk_key *disk_key,
 958                                           int level,
 959                                           u64 hint,
 960                                           u64 empty_size)
 961 {
 962         struct btrfs_fs_info *fs_info = root->fs_info;
 963         struct extent_buffer *ret;
 964
 965         /*
 966          * If we are COWing a node/leaf from the extent, chunk, device or free
 967          * space trees, make sure that we do not finish block group creation of
 968          * pending block groups. We do this to avoid a deadlock.
 969          * COWing can result in allocation of a new chunk, and flushing pending
 970          * block groups (btrfs_create_pending_block_groups()) can be triggered
 971          * when finishing allocation of a new chunk. Creation of a pending block
 972          * group modifies the extent, chunk, device and free space trees,
 973          * therefore we could deadlock with ourselves since we are holding a
 974          * lock on an extent buffer that btrfs_create_pending_block_groups() may
 975          * try to COW later.
 976          * For similar reasons, we also need to delay flushing pending block
 977          * groups when splitting a leaf or node, from one of those trees, since
 978          * we are holding a write lock on it and its parent or when inserting a
 979          * new root node for one of those trees.
 980          */
 981         if (root == fs_info->extent_root ||
 982             root == fs_info->chunk_root ||
 983             root == fs_info->dev_root ||
 984             root == fs_info->free_space_root)
 985                 trans->can_flush_pending_bgs = false;
 986
 987         ret = btrfs_alloc_tree_block(trans, root, parent_start,
 988                                      root->root_key.objectid, disk_key, level,
 989                                      hint, empty_size);
 990         trans->can_flush_pending_bgs = true;
 991
 992         return ret;
 993 }
 994
 995 /*
 996  * does the dirty work in cow of a single block.  The parent block (if
 997  * supplied) is updated to point to the new cow copy.  The new buffer is marked
 998  * dirty and returned locked.  If you modify the block it needs to be marked
 999  * dirty again.
1000  *
1001  * search_start -- an allocation hint for the new block
1002  *
1003  * empty_size -- a hint that you plan on doing more cow.  This is the size in
1004  * bytes the allocator should try to find free next to the block it returns.
1005  * This is just a hint and may be ignored by the allocator.
1006  */
1007 static noinline int __btrfs_cow_block(struct btrfs_trans_handle *trans,
1008                              struct btrfs_root *root,
1009                              struct extent_buffer *buf,
1010                              struct extent_buffer *parent, int parent_slot,
1011                              struct extent_buffer **cow_ret,
1012                              u64 search_start, u64 empty_size)
1013 {
1014         struct btrfs_fs_info *fs_info = root->fs_info;
1015         struct btrfs_disk_key disk_key;
1016         struct extent_buffer *cow;
1017         int level, ret;
1018         int last_ref = 0;
1019         int unlock_orig = 0;
1020         u64 parent_start = 0;
1021
1022         if (*cow_ret == buf)
1023                 unlock_orig = 1;
1024
1025         btrfs_assert_tree_locked(buf);
1026
1027         WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
1028                 trans->transid != fs_info->running_transaction->transid);
1029         WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
1030                 trans->transid != root->last_trans);
1031
1032         level = btrfs_header_level(buf);
1033
1034         if (level == 0)
1035                 btrfs_item_key(buf, &disk_key, 0);
1036         else
1037                 btrfs_node_key(buf, &disk_key, 0);
1038
1039         if ((root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && parent)
1040                 parent_start = parent->start;
1041
1042         cow = alloc_tree_block_no_bg_flush(trans, root, parent_start, &disk_key,
1043                                            level, search_start, empty_size);
1044         if (IS_ERR(cow))
1045                 return PTR_ERR(cow);
1046
1047         /* cow is set to blocking by btrfs_init_new_buffer */
1048
1049         copy_extent_buffer_full(cow, buf);
1050         btrfs_set_header_bytenr(cow, cow->start);
1051         btrfs_set_header_generation(cow, trans->transid);
1052         btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
1053         btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
1054                                      BTRFS_HEADER_FLAG_RELOC);
1055         if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
1056                 btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
1057         else
1058                 btrfs_set_header_owner(cow, root->root_key.objectid);
1059
1060         write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);
1061
1062         ret = update_ref_for_cow(trans, root, buf, cow, &last_ref);
1063         if (ret) {
1064                 btrfs_abort_transaction(trans, ret);
1065                 return ret;
1066         }
1067
1068         if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) {
1069                 ret = btrfs_reloc_cow_block(trans, root, buf, cow);
1070                 if (ret) {
1071                         btrfs_abort_transaction(trans, ret);
1072                         return ret;
1073                 }
1074         }
1075
1076         if (buf == root->node) {
1077                 WARN_ON(parent && parent != buf);
1078                 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
1079                     btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
1080                         parent_start = buf->start;
1081
1082                 atomic_inc(&cow->refs);
1083                 ret = tree_mod_log_insert_root(root->node, cow, 1);
1084                 BUG_ON(ret < 0);
1085                 rcu_assign_pointer(root->node, cow);
1086
1087                 btrfs_free_tree_block(trans, root, buf, parent_start,
1088                                       last_ref);
1089                 free_extent_buffer(buf);
1090                 add_root_to_dirty_list(root);
1091         } else {
1092                 WARN_ON(trans->transid != btrfs_header_generation(parent));
1093                 tree_mod_log_insert_key(parent, parent_slot,
1094                                         MOD_LOG_KEY_REPLACE, GFP_NOFS);
1095                 btrfs_set_node_blockptr(parent, parent_slot,
1096                                         cow->start);
1097                 btrfs_set_node_ptr_generation(parent, parent_slot,
1098                                               trans->transid);
1099                 btrfs_mark_buffer_dirty(parent);
1100                 if (last_ref) {
1101                         ret = tree_mod_log_free_eb(buf);
1102                         if (ret) {
1103                                 btrfs_abort_transaction(trans, ret);
1104                                 return ret;
1105                         }
1106                 }
1107                 btrfs_free_tree_block(trans, root, buf, parent_start,
1108                                       last_ref);
1109         }
1110         if (unlock_orig)
1111                 btrfs_tree_unlock(buf);
1112         free_extent_buffer_stale(buf);
1113         btrfs_mark_buffer_dirty(cow);
1114         *cow_ret = cow;
1115         return 0;
1116 }
1117
1118 /*
1119  * returns the logical address of the oldest predecessor of the given root.
1120  * entries older than time_seq are ignored.
1121  */
1122 static struct tree_mod_elem *__tree_mod_log_oldest_root(
1123                 struct extent_buffer *eb_root, u64 time_seq)
1124 {
1125         struct tree_mod_elem *tm;
1126         struct tree_mod_elem *found = NULL;
1127         u64 root_logical = eb_root->start;
1128         int looped = 0;
1129
1130         if (!time_seq)
1131                 return NULL;
1132
1133         /*
1134          * the very last operation that's logged for a root is the
1135          * replacement operation (if it is replaced at all). this has
1136          * the logical address of the *new* root, making it the very
1137          * first operation that's logged for this root.
1138          */
1139         while (1) {
1140                 tm = tree_mod_log_search_oldest(eb_root->fs_info, root_logical,
1141                                                 time_seq);
1142                 if (!looped && !tm)
1143                         return NULL;
1144                 /*
1145                  * if there are no tree operation for the oldest root, we simply
1146                  * return it. this should only happen if that (old) root is at
1147                  * level 0.
1148                  */
1149                 if (!tm)
1150                         break;
1151
1152                 /*
1153                  * if there's an operation that's not a root replacement, we
1154                  * found the oldest version of our root. normally, we'll find a
1155                  * MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here.
1156                  */
1157                 if (tm->op != MOD_LOG_ROOT_REPLACE)
1158                         break;
1159
1160                 found = tm;
1161                 root_logical = tm->old_root.logical;
1162                 looped = 1;
1163         }
1164
1165         /* if there's no old root to return, return what we found instead */
1166         if (!found)
1167                 found = tm;
1168
1169         return found;
1170 }
1171
1172 /*
1173  * tm is a pointer to the first operation to rewind within eb. then, all
1174  * previous operations will be rewound (until we reach something older than
1175  * time_seq).
1176  */
1177 static void
1178 __tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct extent_buffer *eb,
1179                       u64 time_seq, struct tree_mod_elem *first_tm)
1180 {
1181         u32 n;
1182         struct rb_node *next;
1183         struct tree_mod_elem *tm = first_tm;
1184         unsigned long o_dst;
1185         unsigned long o_src;
1186         unsigned long p_size = sizeof(struct btrfs_key_ptr);
1187
1188         n = btrfs_header_nritems(eb);
1189         read_lock(&fs_info->tree_mod_log_lock);
1190         while (tm && tm->seq >= time_seq) {
1191                 /*
1192                  * all the operations are recorded with the operator used for
1193                  * the modification. as we're going backwards, we do the
1194                  * opposite of each operation here.
1195                  */
1196                 switch (tm->op) {
1197                 case MOD_LOG_KEY_REMOVE_WHILE_FREEING:
1198                         BUG_ON(tm->slot < n);
1199                         /* Fallthrough */
1200                 case MOD_LOG_KEY_REMOVE_WHILE_MOVING:
1201                 case MOD_LOG_KEY_REMOVE:
1202                         btrfs_set_node_key(eb, &tm->key, tm->slot);
1203                         btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
1204                         btrfs_set_node_ptr_generation(eb, tm->slot,
1205                                                       tm->generation);
1206                         n++;
1207                         break;
1208                 case MOD_LOG_KEY_REPLACE:
1209                         BUG_ON(tm->slot >= n);
1210                         btrfs_set_node_key(eb, &tm->key, tm->slot);
1211                         btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
1212                         btrfs_set_node_ptr_generation(eb, tm->slot,
1213                                                       tm->generation);
1214                         break;
1215                 case MOD_LOG_KEY_ADD:
1216                         /* if a move operation is needed it's in the log */
1217                         n--;
1218                         break;
1219                 case MOD_LOG_MOVE_KEYS:
1220                         o_dst = btrfs_node_key_ptr_offset(tm->slot);
1221                         o_src = btrfs_node_key_ptr_offset(tm->move.dst_slot);
1222                         memmove_extent_buffer(eb, o_dst, o_src,
1223                                               tm->move.nr_items * p_size);
1224                         break;
1225                 case MOD_LOG_ROOT_REPLACE:
1226                         /*
1227                          * this operation is special. for roots, this must be
1228                          * handled explicitly before rewinding.
1229                          * for non-roots, this operation may exist if the node
1230                          * was a root: root A -> child B; then A gets empty and
1231                          * B is promoted to the new root. in the mod log, we'll
1232                          * have a root-replace operation for B, a tree block
1233                          * that is no root. we simply ignore that operation.
1234                          */
1235                         break;
1236                 }
1237                 next = rb_next(&tm->node);
1238                 if (!next)
1239                         break;
1240                 tm = rb_entry(next, struct tree_mod_elem, node);
1241                 if (tm->logical != first_tm->logical)
1242                         break;
1243         }
1244         read_unlock(&fs_info->tree_mod_log_lock);
1245         btrfs_set_header_nritems(eb, n);
1246 }
1247
1248 /*
1249  * Called with eb read locked. If the buffer cannot be rewound, the same buffer
1250  * is returned. If rewind operations happen, a fresh buffer is returned. The
1251  * returned buffer is always read-locked. If the returned buffer is not the
1252  * input buffer, the lock on the input buffer is released and the input buffer
1253  * is freed (its refcount is decremented).
1254  */
1255 static struct extent_buffer *
1256 tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path,
1257                     struct extent_buffer *eb, u64 time_seq)
1258 {
1259         struct extent_buffer *eb_rewin;
1260         struct tree_mod_elem *tm;
1261
1262         if (!time_seq)
1263                 return eb;
1264
1265         if (btrfs_header_level(eb) == 0)
1266                 return eb;
1267
1268         tm = tree_mod_log_search(fs_info, eb->start, time_seq);
1269         if (!tm)
1270                 return eb;
1271
1272         btrfs_set_path_blocking(path);
1273         btrfs_set_lock_blocking_read(eb);
1274
1275         if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
1276                 BUG_ON(tm->slot != 0);
1277                 eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start);
1278                 if (!eb_rewin) {
1279                         btrfs_tree_read_unlock_blocking(eb);
1280                         free_extent_buffer(eb);
1281                         return NULL;
1282                 }
1283                 btrfs_set_header_bytenr(eb_rewin, eb->start);
1284                 btrfs_set_header_backref_rev(eb_rewin,
1285                                              btrfs_header_backref_rev(eb));
1286                 btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb));
1287                 btrfs_set_header_level(eb_rewin, btrfs_header_level(eb));
1288         } else {
1289                 eb_rewin = btrfs_clone_extent_buffer(eb);
1290                 if (!eb_rewin) {
1291                         btrfs_tree_read_unlock_blocking(eb);
1292                         free_extent_buffer(eb);
1293                         return NULL;
1294                 }
1295         }
1296
1297         btrfs_tree_read_unlock_blocking(eb);
1298         free_extent_buffer(eb);
1299
1300         btrfs_tree_read_lock(eb_rewin);
1301         __tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm);
1302         WARN_ON(btrfs_header_nritems(eb_rewin) >
1303                 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1304
1305         return eb_rewin;
1306 }
1307
1308 /*
1309  * get_old_root() rewinds the state of @root's root node to the given @time_seq
1310  * value. If there are no changes, the current root->root_node is returned. If
1311  * anything changed in between, there's a fresh buffer allocated on which the
1312  * rewind operations are done. In any case, the returned buffer is read locked.
1313  * Returns NULL on error (with no locks held).
1314  */
1315 static inline struct extent_buffer *
1316 get_old_root(struct btrfs_root *root, u64 time_seq)
1317 {
1318         struct btrfs_fs_info *fs_info = root->fs_info;
1319         struct tree_mod_elem *tm;
1320         struct extent_buffer *eb = NULL;
1321         struct extent_buffer *eb_root;
1322         u64 eb_root_owner = 0;
1323         struct extent_buffer *old;
1324         struct tree_mod_root *old_root = NULL;
1325         u64 old_generation = 0;
1326         u64 logical;
1327         int level;
1328
1329         eb_root = btrfs_read_lock_root_node(root);
1330         tm = __tree_mod_log_oldest_root(eb_root, time_seq);
1331         if (!tm)
1332                 return eb_root;
1333
1334         if (tm->op == MOD_LOG_ROOT_REPLACE) {
1335                 old_root = &tm->old_root;
1336                 old_generation = tm->generation;
1337                 logical = old_root->logical;
1338                 level = old_root->level;
1339         } else {
1340                 logical = eb_root->start;
1341                 level = btrfs_header_level(eb_root);
1342         }
1343
1344         tm = tree_mod_log_search(fs_info, logical, time_seq);
1345         if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
1346                 btrfs_tree_read_unlock(eb_root);
1347                 free_extent_buffer(eb_root);
1348                 old = read_tree_block(fs_info, logical, 0, level, NULL);
1349                 if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) {
1350                         if (!IS_ERR(old))
1351                                 free_extent_buffer(old);
1352                         btrfs_warn(fs_info,
1353                                    "failed to read tree block %llu from get_old_root",
1354                                    logical);
1355                 } else {
1356                         eb = btrfs_clone_extent_buffer(old);
1357                         free_extent_buffer(old);
1358                 }
1359         } else if (old_root) {
1360                 eb_root_owner = btrfs_header_owner(eb_root);
1361                 btrfs_tree_read_unlock(eb_root);
1362                 free_extent_buffer(eb_root);
1363                 eb = alloc_dummy_extent_buffer(fs_info, logical);
1364         } else {
1365                 btrfs_set_lock_blocking_read(eb_root);
1366                 eb = btrfs_clone_extent_buffer(eb_root);
1367                 btrfs_tree_read_unlock_blocking(eb_root);
1368                 free_extent_buffer(eb_root);
1369         }
1370
1371         if (!eb)
1372                 return NULL;
1373         btrfs_tree_read_lock(eb);
1374         if (old_root) {
1375                 btrfs_set_header_bytenr(eb, eb->start);
1376                 btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV);
1377                 btrfs_set_header_owner(eb, eb_root_owner);
1378                 btrfs_set_header_level(eb, old_root->level);
1379                 btrfs_set_header_generation(eb, old_generation);
1380         }
1381         if (tm)
1382                 __tree_mod_log_rewind(fs_info, eb, time_seq, tm);
1383         else
1384                 WARN_ON(btrfs_header_level(eb) != 0);
1385         WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1386
1387         return eb;
1388 }
1389
1390 int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq)
1391 {
1392         struct tree_mod_elem *tm;
1393         int level;
1394         struct extent_buffer *eb_root = btrfs_root_node(root);
1395
1396         tm = __tree_mod_log_oldest_root(eb_root, time_seq);
1397         if (tm && tm->op == MOD_LOG_ROOT_REPLACE) {
1398                 level = tm->old_root.level;
1399         } else {
1400                 level = btrfs_header_level(eb_root);
1401         }
1402         free_extent_buffer(eb_root);
1403
1404         return level;
1405 }
1406
1407 static inline int should_cow_block(struct btrfs_trans_handle *trans,
1408                                    struct btrfs_root *root,
1409                                    struct extent_buffer *buf)
1410 {
1411         if (btrfs_is_testing(root->fs_info))
1412                 return 0;
1413
1414         /* Ensure we can see the FORCE_COW bit */
1415         smp_mb__before_atomic();
1416
1417         /*
1418          * We do not need to cow a block if
1419          * 1) this block is not created or changed in this transaction;
1420          * 2) this block does not belong to TREE_RELOC tree;
1421          * 3) the root is not forced COW.
1422          *
1423          * What is forced COW:
1424          *    when we create snapshot during committing the transaction,
1425          *    after we've finished copying src root, we must COW the shared
1426          *    block to ensure the metadata consistency.
1427          */
1428         if (btrfs_header_generation(buf) == trans->transid &&
1429             !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) &&
1430             !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
1431               btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) &&
1432             !test_bit(BTRFS_ROOT_FORCE_COW, &root->state))
1433                 return 0;
1434         return 1;
1435 }
1436
1437 /*
1438  * cows a single block, see __btrfs_cow_block for the real work.
1439  * This version of it has extra checks so that a block isn't COWed more than
1440  * once per transaction, as long as it hasn't been written yet
1441  */
1442 noinline int btrfs_cow_block(struct btrfs_trans_handle *trans,
1443                     struct btrfs_root *root, struct extent_buffer *buf,
1444                     struct extent_buffer *parent, int parent_slot,
1445                     struct extent_buffer **cow_ret)
1446 {
1447         struct btrfs_fs_info *fs_info = root->fs_info;
1448         u64 search_start;
1449         int ret;
1450
1451         if (test_bit(BTRFS_ROOT_DELETING, &root->state))
1452                 btrfs_err(fs_info,
1453                         "COW'ing blocks on a fs root that's being dropped");
1454
1455         if (trans->transaction != fs_info->running_transaction)
1456                 WARN(1, KERN_CRIT "trans %llu running %llu\n",
1457                        trans->transid,
1458                        fs_info->running_transaction->transid);
1459
1460         if (trans->transid != fs_info->generation)
1461                 WARN(1, KERN_CRIT "trans %llu running %llu\n",
1462                        trans->transid, fs_info->generation);
1463
1464         if (!should_cow_block(trans, root, buf)) {
1465                 trans->dirty = true;
1466                 *cow_ret = buf;
1467                 return 0;
1468         }
1469
1470         search_start = buf->start & ~((u64)SZ_1G - 1);
1471
1472         if (parent)
1473                 btrfs_set_lock_blocking_write(parent);
1474         btrfs_set_lock_blocking_write(buf);
1475
1476         /*
1477          * Before CoWing this block for later modification, check if it's
1478          * the subtree root and do the delayed subtree trace if needed.
1479          *
1480          * Also We don't care about the error, as it's handled internally.
1481          */
1482         btrfs_qgroup_trace_subtree_after_cow(trans, root, buf);
1483         ret = __btrfs_cow_block(trans, root, buf, parent,
1484                                  parent_slot, cow_ret, search_start, 0);
1485
1486         trace_btrfs_cow_block(root, buf, *cow_ret);
1487
1488         return ret;
1489 }
1490
1491 /*
1492  * helper function for defrag to decide if two blocks pointed to by a
1493  * node are actually close by
1494  */
1495 static int close_blocks(u64 blocknr, u64 other, u32 blocksize)
1496 {
1497         if (blocknr < other && other - (blocknr + blocksize) < 32768)
1498                 return 1;
1499         if (blocknr > other && blocknr - (other + blocksize) < 32768)
1500                 return 1;
1501         return 0;
1502 }
1503
1504 /*
1505  * compare two keys in a memcmp fashion
1506  */
1507 static int comp_keys(const struct btrfs_disk_key *disk,
1508                      const struct btrfs_key *k2)
1509 {
1510         struct btrfs_key k1;
1511
1512         btrfs_disk_key_to_cpu(&k1, disk);
1513
1514         return btrfs_comp_cpu_keys(&k1, k2);
1515 }
1516
1517 /*
1518  * same as comp_keys only with two btrfs_key's
1519  */
1520 int __pure btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2)
1521 {
1522         if (k1->objectid > k2->objectid)
1523                 return 1;
1524         if (k1->objectid < k2->objectid)
1525                 return -1;
1526         if (k1->type > k2->type)
1527                 return 1;
1528         if (k1->type < k2->type)
1529                 return -1;
1530         if (k1->offset > k2->offset)
1531                 return 1;
1532         if (k1->offset < k2->offset)
1533                 return -1;
1534         return 0;
1535 }
1536
1537 /*
1538  * this is used by the defrag code to go through all the
1539  * leaves pointed to by a node and reallocate them so that
1540  * disk order is close to key order
1541  */
1542 int btrfs_realloc_node(struct btrfs_trans_handle *trans,
1543                        struct btrfs_root *root, struct extent_buffer *parent,
1544                        int start_slot, u64 *last_ret,
1545                        struct btrfs_key *progress)
1546 {
1547         struct btrfs_fs_info *fs_info = root->fs_info;
1548         struct extent_buffer *cur;
1549         u64 blocknr;
1550         u64 gen;
1551         u64 search_start = *last_ret;
1552         u64 last_block = 0;
1553         u64 other;
1554         u32 parent_nritems;
1555         int end_slot;
1556         int i;
1557         int err = 0;
1558         int parent_level;
1559         int uptodate;
1560         u32 blocksize;
1561         int progress_passed = 0;
1562         struct btrfs_disk_key disk_key;
1563
1564         parent_level = btrfs_header_level(parent);
1565
1566         WARN_ON(trans->transaction != fs_info->running_transaction);
1567         WARN_ON(trans->transid != fs_info->generation);
1568
1569         parent_nritems = btrfs_header_nritems(parent);
1570         blocksize = fs_info->nodesize;
1571         end_slot = parent_nritems - 1;
1572
1573         if (parent_nritems <= 1)
1574                 return 0;
1575
1576         btrfs_set_lock_blocking_write(parent);
1577
1578         for (i = start_slot; i <= end_slot; i++) {
1579                 struct btrfs_key first_key;
1580                 int close = 1;
1581
1582                 btrfs_node_key(parent, &disk_key, i);
1583                 if (!progress_passed && comp_keys(&disk_key, progress) < 0)
1584                         continue;
1585
1586                 progress_passed = 1;
1587                 blocknr = btrfs_node_blockptr(parent, i);
1588                 gen = btrfs_node_ptr_generation(parent, i);
1589                 btrfs_node_key_to_cpu(parent, &first_key, i);
1590                 if (last_block == 0)
1591                         last_block = blocknr;
1592
1593                 if (i > 0) {
1594                         other = btrfs_node_blockptr(parent, i - 1);
1595                         close = close_blocks(blocknr, other, blocksize);
1596                 }
1597                 if (!close && i < end_slot) {
1598                         other = btrfs_node_blockptr(parent, i + 1);
1599                         close = close_blocks(blocknr, other, blocksize);
1600                 }
1601                 if (close) {
1602                         last_block = blocknr;
1603                         continue;
1604                 }
1605
1606                 cur = find_extent_buffer(fs_info, blocknr);
1607                 if (cur)
1608                         uptodate = btrfs_buffer_uptodate(cur, gen, 0);
1609                 else
1610                         uptodate = 0;
1611                 if (!cur || !uptodate) {
1612                         if (!cur) {
1613                                 cur = read_tree_block(fs_info, blocknr, gen,
1614                                                       parent_level - 1,
1615                                                       &first_key);
1616                                 if (IS_ERR(cur)) {
1617                                         return PTR_ERR(cur);
1618                                 } else if (!extent_buffer_uptodate(cur)) {
1619                                         free_extent_buffer(cur);
1620                                         return -EIO;
1621                                 }
1622                         } else if (!uptodate) {
1623                                 err = btrfs_read_buffer(cur, gen,
1624                                                 parent_level - 1,&first_key);
1625                                 if (err) {
1626                                         free_extent_buffer(cur);
1627                                         return err;
1628                                 }
1629                         }
1630                 }
1631                 if (search_start == 0)
1632                         search_start = last_block;
1633
1634                 btrfs_tree_lock(cur);
1635                 btrfs_set_lock_blocking_write(cur);
1636                 err = __btrfs_cow_block(trans, root, cur, parent, i,
1637                                         &cur, search_start,
1638                                         min(16 * blocksize,
1639                                             (end_slot - i) * blocksize));
1640                 if (err) {
1641                         btrfs_tree_unlock(cur);
1642                         free_extent_buffer(cur);
1643                         break;
1644                 }
1645                 search_start = cur->start;
1646                 last_block = cur->start;
1647                 *last_ret = search_start;
1648                 btrfs_tree_unlock(cur);
1649                 free_extent_buffer(cur);
1650         }
1651         return err;
1652 }
1653
1654 /*
1655  * search for key in the extent_buffer.  The items start at offset p,
1656  * and they are item_size apart.  There are 'max' items in p.
1657  *
1658  * the slot in the array is returned via slot, and it points to
1659  * the place where you would insert key if it is not found in
1660  * the array.
1661  *
1662  * slot may point to max if the key is bigger than all of the keys
1663  */
1664 static noinline int generic_bin_search(struct extent_buffer *eb,
1665                                        unsigned long p, int item_size,
1666                                        const struct btrfs_key *key,
1667                                        int max, int *slot)
1668 {
1669         int low = 0;
1670         int high = max;
1671         int mid;
1672         int ret;
1673         struct btrfs_disk_key *tmp = NULL;
1674         struct btrfs_disk_key unaligned;
1675         unsigned long offset;
1676         char *kaddr = NULL;
1677         unsigned long map_start = 0;
1678         unsigned long map_len = 0;
1679         int err;
1680
1681         if (low > high) {
1682                 btrfs_err(eb->fs_info,
1683                  "%s: low (%d) > high (%d) eb %llu owner %llu level %d",
1684                           __func__, low, high, eb->start,
1685                           btrfs_header_owner(eb), btrfs_header_level(eb));
1686                 return -EINVAL;
1687         }
1688
1689         while (low < high) {
1690                 mid = (low + high) / 2;
1691                 offset = p + mid * item_size;
1692
1693                 if (!kaddr || offset < map_start ||
1694                     (offset + sizeof(struct btrfs_disk_key)) >
1695                     map_start + map_len) {
1696
1697                         err = map_private_extent_buffer(eb, offset,
1698                                                 sizeof(struct btrfs_disk_key),
1699                                                 &kaddr, &map_start, &map_len);
1700
1701                         if (!err) {
1702                                 tmp = (struct btrfs_disk_key *)(kaddr + offset -
1703                                                         map_start);
1704                         } else if (err == 1) {
1705                                 read_extent_buffer(eb, &unaligned,
1706                                                    offset, sizeof(unaligned));
1707                                 tmp = &unaligned;
1708                         } else {
1709                                 return err;
1710                         }
1711
1712                 } else {
1713                         tmp = (struct btrfs_disk_key *)(kaddr + offset -
1714                                                         map_start);
1715                 }
1716                 ret = comp_keys(tmp, key);
1717
1718                 if (ret < 0)
1719                         low = mid + 1;
1720                 else if (ret > 0)
1721                         high = mid;
1722                 else {
1723                         *slot = mid;
1724                         return 0;
1725                 }
1726         }
1727         *slot = low;
1728         return 1;
1729 }
1730
1731 /*
1732  * simple bin_search frontend that does the right thing for
1733  * leaves vs nodes
1734  */
1735 int btrfs_bin_search(struct extent_buffer *eb, const struct btrfs_key *key,
1736                      int level, int *slot)
1737 {
1738         if (level == 0)
1739                 return generic_bin_search(eb,
1740                                           offsetof(struct btrfs_leaf, items),
1741                                           sizeof(struct btrfs_item),
1742                                           key, btrfs_header_nritems(eb),
1743                                           slot);
1744         else
1745                 return generic_bin_search(eb,
1746                                           offsetof(struct btrfs_node, ptrs),
1747                                           sizeof(struct btrfs_key_ptr),
1748                                           key, btrfs_header_nritems(eb),
1749                                           slot);
1750 }
1751
1752 static void root_add_used(struct btrfs_root *root, u32 size)
1753 {
1754         spin_lock(&root->accounting_lock);
1755         btrfs_set_root_used(&root->root_item,
1756                             btrfs_root_used(&root->root_item) + size);
1757         spin_unlock(&root->accounting_lock);
1758 }
1759
1760 static void root_sub_used(struct btrfs_root *root, u32 size)
1761 {
1762         spin_lock(&root->accounting_lock);
1763         btrfs_set_root_used(&root->root_item,
1764                             btrfs_root_used(&root->root_item) - size);
1765         spin_unlock(&root->accounting_lock);
1766 }
1767
1768 /* given a node and slot number, this reads the blocks it points to.  The
1769  * extent buffer is returned with a reference taken (but unlocked).
1770  */
1771 struct extent_buffer *btrfs_read_node_slot(struct extent_buffer *parent,
1772                                            int slot)
1773 {
1774         int level = btrfs_header_level(parent);
1775         struct extent_buffer *eb;
1776         struct btrfs_key first_key;
1777
1778         if (slot < 0 || slot >= btrfs_header_nritems(parent))
1779                 return ERR_PTR(-ENOENT);
1780
1781         BUG_ON(level == 0);
1782
1783         btrfs_node_key_to_cpu(parent, &first_key, slot);
1784         eb = read_tree_block(parent->fs_info, btrfs_node_blockptr(parent, slot),
1785                              btrfs_node_ptr_generation(parent, slot),
1786                              level - 1, &first_key);
1787         if (!IS_ERR(eb) && !extent_buffer_uptodate(eb)) {
1788                 free_extent_buffer(eb);
1789                 eb = ERR_PTR(-EIO);
1790         }
1791
1792         return eb;
1793 }
1794
1795 /*
1796  * node level balancing, used to make sure nodes are in proper order for
1797  * item deletion.  We balance from the top down, so we have to make sure
1798  * that a deletion won't leave an node completely empty later on.
1799  */
1800 static noinline int balance_level(struct btrfs_trans_handle *trans,
1801                          struct btrfs_root *root,
1802                          struct btrfs_path *path, int level)
1803 {
1804         struct btrfs_fs_info *fs_info = root->fs_info;
1805         struct extent_buffer *right = NULL;
1806         struct extent_buffer *mid;
1807         struct extent_buffer *left = NULL;
1808         struct extent_buffer *parent = NULL;
1809         int ret = 0;
1810         int wret;
1811         int pslot;
1812         int orig_slot = path->slots[level];
1813         u64 orig_ptr;
1814
1815         ASSERT(level > 0);
1816
1817         mid = path->nodes[level];
1818
1819         WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK &&
1820                 path->locks[level] != BTRFS_WRITE_LOCK_BLOCKING);
1821         WARN_ON(btrfs_header_generation(mid) != trans->transid);
1822
1823         orig_ptr = btrfs_node_blockptr(mid, orig_slot);
1824
1825         if (level < BTRFS_MAX_LEVEL - 1) {
1826                 parent = path->nodes[level + 1];
1827                 pslot = path->slots[level + 1];
1828         }
1829
1830         /*
1831          * deal with the case where there is only one pointer in the root
1832          * by promoting the node below to a root
1833          */
1834         if (!parent) {
1835                 struct extent_buffer *child;
1836
1837                 if (btrfs_header_nritems(mid) != 1)
1838                         return 0;
1839
1840                 /* promote the child to a root */
1841                 child = btrfs_read_node_slot(mid, 0);
1842                 if (IS_ERR(child)) {
1843                         ret = PTR_ERR(child);
1844                         btrfs_handle_fs_error(fs_info, ret, NULL);
1845                         goto enospc;
1846                 }
1847
1848                 btrfs_tree_lock(child);
1849                 btrfs_set_lock_blocking_write(child);
1850                 ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
1851                 if (ret) {
1852                         btrfs_tree_unlock(child);
1853                         free_extent_buffer(child);
1854                         goto enospc;
1855                 }
1856
1857                 ret = tree_mod_log_insert_root(root->node, child, 1);
1858                 BUG_ON(ret < 0);
1859                 rcu_assign_pointer(root->node, child);
1860
1861                 add_root_to_dirty_list(root);
1862                 btrfs_tree_unlock(child);
1863
1864                 path->locks[level] = 0;
1865                 path->nodes[level] = NULL;
1866                 btrfs_clean_tree_block(mid);
1867                 btrfs_tree_unlock(mid);
1868                 /* once for the path */
1869                 free_extent_buffer(mid);
1870
1871                 root_sub_used(root, mid->len);
1872                 btrfs_free_tree_block(trans, root, mid, 0, 1);
1873                 /* once for the root ptr */
1874                 free_extent_buffer_stale(mid);
1875                 return 0;
1876         }
1877         if (btrfs_header_nritems(mid) >
1878             BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 4)
1879                 return 0;
1880
1881         left = btrfs_read_node_slot(parent, pslot - 1);
1882         if (IS_ERR(left))
1883                 left = NULL;
1884
1885         if (left) {
1886                 btrfs_tree_lock(left);
1887                 btrfs_set_lock_blocking_write(left);
1888                 wret = btrfs_cow_block(trans, root, left,
1889                                        parent, pslot - 1, &left);
1890                 if (wret) {
1891                         ret = wret;
1892                         goto enospc;
1893                 }
1894         }
1895
1896         right = btrfs_read_node_slot(parent, pslot + 1);
1897         if (IS_ERR(right))
1898                 right = NULL;
1899
1900         if (right) {
1901                 btrfs_tree_lock(right);
1902                 btrfs_set_lock_blocking_write(right);
1903                 wret = btrfs_cow_block(trans, root, right,
1904                                        parent, pslot + 1, &right);
1905                 if (wret) {
1906                         ret = wret;
1907                         goto enospc;
1908                 }
1909         }
1910
1911         /* first, try to make some room in the middle buffer */
1912         if (left) {
1913                 orig_slot += btrfs_header_nritems(left);
1914                 wret = push_node_left(trans, left, mid, 1);
1915                 if (wret < 0)
1916                         ret = wret;
1917         }
1918
1919         /*
1920          * then try to empty the right most buffer into the middle
1921          */
1922         if (right) {
1923                 wret = push_node_left(trans, mid, right, 1);
1924                 if (wret < 0 && wret != -ENOSPC)
1925                         ret = wret;
1926                 if (btrfs_header_nritems(right) == 0) {
1927                         btrfs_clean_tree_block(right);
1928                         btrfs_tree_unlock(right);
1929                         del_ptr(root, path, level + 1, pslot + 1);
1930                         root_sub_used(root, right->len);
1931                         btrfs_free_tree_block(trans, root, right, 0, 1);
1932                         free_extent_buffer_stale(right);
1933                         right = NULL;
1934                 } else {
1935                         struct btrfs_disk_key right_key;
1936                         btrfs_node_key(right, &right_key, 0);
1937                         ret = tree_mod_log_insert_key(parent, pslot + 1,
1938                                         MOD_LOG_KEY_REPLACE, GFP_NOFS);
1939                         BUG_ON(ret < 0);
1940                         btrfs_set_node_key(parent, &right_key, pslot + 1);
1941                         btrfs_mark_buffer_dirty(parent);
1942                 }
1943         }
1944         if (btrfs_header_nritems(mid) == 1) {
1945                 /*
1946                  * we're not allowed to leave a node with one item in the
1947                  * tree during a delete.  A deletion from lower in the tree
1948                  * could try to delete the only pointer in this node.
1949                  * So, pull some keys from the left.
1950                  * There has to be a left pointer at this point because
1951                  * otherwise we would have pulled some pointers from the
1952                  * right
1953                  */
1954                 if (!left) {
1955                         ret = -EROFS;
1956                         btrfs_handle_fs_error(fs_info, ret, NULL);
1957                         goto enospc;
1958                 }
1959                 wret = balance_node_right(trans, mid, left);
1960                 if (wret < 0) {
1961                         ret = wret;
1962                         goto enospc;
1963                 }
1964                 if (wret == 1) {
1965                         wret = push_node_left(trans, left, mid, 1);
1966                         if (wret < 0)
1967                                 ret = wret;
1968                 }
1969                 BUG_ON(wret == 1);
1970         }
1971         if (btrfs_header_nritems(mid) == 0) {
1972                 btrfs_clean_tree_block(mid);
1973                 btrfs_tree_unlock(mid);
1974                 del_ptr(root, path, level + 1, pslot);
1975                 root_sub_used(root, mid->len);
1976                 btrfs_free_tree_block(trans, root, mid, 0, 1);
1977                 free_extent_buffer_stale(mid);
1978                 mid = NULL;
1979         } else {
1980                 /* update the parent key to reflect our changes */
1981                 struct btrfs_disk_key mid_key;
1982                 btrfs_node_key(mid, &mid_key, 0);
1983                 ret = tree_mod_log_insert_key(parent, pslot,
1984                                 MOD_LOG_KEY_REPLACE, GFP_NOFS);
1985                 BUG_ON(ret < 0);
1986                 btrfs_set_node_key(parent, &mid_key, pslot);
1987                 btrfs_mark_buffer_dirty(parent);
1988         }
1989
1990         /* update the path */
1991         if (left) {
1992                 if (btrfs_header_nritems(left) > orig_slot) {
1993                         atomic_inc(&left->refs);
1994                         /* left was locked after cow */
1995                         path->nodes[level] = left;
1996                         path->slots[level + 1] -= 1;
1997                         path->slots[level] = orig_slot;
1998                         if (mid) {
1999                                 btrfs_tree_unlock(mid);
2000                                 free_extent_buffer(mid);
2001                         }
2002                 } else {
2003                         orig_slot -= btrfs_header_nritems(left);
2004                         path->slots[level] = orig_slot;
2005                 }
2006         }
2007         /* double check we haven't messed things up */
2008         if (orig_ptr !=
2009             btrfs_node_blockptr(path->nodes[level], path->slots[level]))
2010                 BUG();
2011 enospc:
2012         if (right) {
2013                 btrfs_tree_unlock(right);
2014                 free_extent_buffer(right);
2015         }
2016         if (left) {
2017                 if (path->nodes[level] != left)
2018                         btrfs_tree_unlock(left);
2019                 free_extent_buffer(left);
2020         }
2021         return ret;
2022 }
2023
2024 /* Node balancing for insertion.  Here we only split or push nodes around
2025  * when they are completely full.  This is also done top down, so we
2026  * have to be pessimistic.
2027  */
2028 static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
2029                                           struct btrfs_root *root,
2030                                           struct btrfs_path *path, int level)
2031 {
2032         struct btrfs_fs_info *fs_info = root->fs_info;
2033         struct extent_buffer *right = NULL;
2034         struct extent_buffer *mid;
2035         struct extent_buffer *left = NULL;
2036         struct extent_buffer *parent = NULL;
2037         int ret = 0;
2038         int wret;
2039         int pslot;
2040         int orig_slot = path->slots[level];
2041
2042         if (level == 0)
2043                 return 1;
2044
2045         mid = path->nodes[level];
2046         WARN_ON(btrfs_header_generation(mid) != trans->transid);
2047
2048         if (level < BTRFS_MAX_LEVEL - 1) {
2049                 parent = path->nodes[level + 1];
2050                 pslot = path->slots[level + 1];
2051         }
2052
2053         if (!parent)
2054                 return 1;
2055
2056         left = btrfs_read_node_slot(parent, pslot - 1);
2057         if (IS_ERR(left))
2058                 left = NULL;
2059
2060         /* first, try to make some room in the middle buffer */
2061         if (left) {
2062                 u32 left_nr;
2063
2064                 btrfs_tree_lock(left);
2065                 btrfs_set_lock_blocking_write(left);
2066
2067                 left_nr = btrfs_header_nritems(left);
2068                 if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
2069                         wret = 1;
2070                 } else {
2071                         ret = btrfs_cow_block(trans, root, left, parent,
2072                                               pslot - 1, &left);
2073                         if (ret)
2074                                 wret = 1;
2075                         else {
2076                                 wret = push_node_left(trans, left, mid, 0);
2077                         }
2078                 }
2079                 if (wret < 0)
2080                         ret = wret;
2081                 if (wret == 0) {
2082                         struct btrfs_disk_key disk_key;
2083                         orig_slot += left_nr;
2084                         btrfs_node_key(mid, &disk_key, 0);
2085                         ret = tree_mod_log_insert_key(parent, pslot,
2086                                         MOD_LOG_KEY_REPLACE, GFP_NOFS);
2087                         BUG_ON(ret < 0);
2088                         btrfs_set_node_key(parent, &disk_key, pslot);
2089                         btrfs_mark_buffer_dirty(parent);
2090                         if (btrfs_header_nritems(left) > orig_slot) {
2091                                 path->nodes[level] = left;
2092                                 path->slots[level + 1] -= 1;
2093                                 path->slots[level] = orig_slot;
2094                                 btrfs_tree_unlock(mid);
2095                                 free_extent_buffer(mid);
2096                         } else {
2097                                 orig_slot -=
2098                                         btrfs_header_nritems(left);
2099                                 path->slots[level] = orig_slot;
2100                                 btrfs_tree_unlock(left);
2101                                 free_extent_buffer(left);
2102                         }
2103                         return 0;
2104                 }
2105                 btrfs_tree_unlock(left);
2106                 free_extent_buffer(left);
2107         }
2108         right = btrfs_read_node_slot(parent, pslot + 1);
2109         if (IS_ERR(right))
2110                 right = NULL;
2111
2112         /*
2113          * then try to empty the right most buffer into the middle
2114          */
2115         if (right) {
2116                 u32 right_nr;
2117
2118                 btrfs_tree_lock(right);
2119                 btrfs_set_lock_blocking_write(right);
2120
2121                 right_nr = btrfs_header_nritems(right);
2122                 if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
2123                         wret = 1;
2124                 } else {
2125                         ret = btrfs_cow_block(trans, root, right,
2126                                               parent, pslot + 1,
2127                                               &right);
2128                         if (ret)
2129                                 wret = 1;
2130                         else {
2131                                 wret = balance_node_right(trans, right, mid);
2132                         }
2133                 }
2134                 if (wret < 0)
2135                         ret = wret;
2136                 if (wret == 0) {
2137                         struct btrfs_disk_key disk_key;
2138
2139                         btrfs_node_key(right, &disk_key, 0);
2140                         ret = tree_mod_log_insert_key(parent, pslot + 1,
2141                                         MOD_LOG_KEY_REPLACE, GFP_NOFS);
2142                         BUG_ON(ret < 0);
2143                         btrfs_set_node_key(parent, &disk_key, pslot + 1);
2144                         btrfs_mark_buffer_dirty(parent);
2145
2146                         if (btrfs_header_nritems(mid) <= orig_slot) {
2147                                 path->nodes[level] = right;
2148                                 path->slots[level + 1] += 1;
2149                                 path->slots[level] = orig_slot -
2150                                         btrfs_header_nritems(mid);
2151                                 btrfs_tree_unlock(mid);
2152                                 free_extent_buffer(mid);
2153                         } else {
2154                                 btrfs_tree_unlock(right);
2155                                 free_extent_buffer(right);
2156                         }
2157                         return 0;
2158                 }
2159                 btrfs_tree_unlock(right);
2160                 free_extent_buffer(right);
2161         }
2162         return 1;
2163 }
2164
2165 /*
2166  * readahead one full node of leaves, finding things that are close
2167  * to the block in 'slot', and triggering ra on them.
2168  */
2169 static void reada_for_search(struct btrfs_fs_info *fs_info,
2170                              struct btrfs_path *path,
2171                              int level, int slot, u64 objectid)
2172 {
2173         struct extent_buffer *node;
2174         struct btrfs_disk_key disk_key;
2175         u32 nritems;
2176         u64 search;
2177         u64 target;
2178         u64 nread = 0;
2179         struct extent_buffer *eb;
2180         u32 nr;
2181         u32 blocksize;
2182         u32 nscan = 0;
2183
2184         if (level != 1)
2185                 return;
2186
2187         if (!path->nodes[level])
2188                 return;
2189
2190         node = path->nodes[level];
2191
2192         search = btrfs_node_blockptr(node, slot);
2193         blocksize = fs_info->nodesize;
2194         eb = find_extent_buffer(fs_info, search);
2195         if (eb) {
2196                 free_extent_buffer(eb);
2197                 return;
2198         }
2199
2200         target = search;
2201
2202         nritems = btrfs_header_nritems(node);
2203         nr = slot;
2204
2205         while (1) {
2206                 if (path->reada == READA_BACK) {
2207                         if (nr == 0)
2208                                 break;
2209                         nr--;
2210                 } else if (path->reada == READA_FORWARD) {
2211                         nr++;
2212                         if (nr >= nritems)
2213                                 break;
2214                 }
2215                 if (path->reada == READA_BACK && objectid) {
2216                         btrfs_node_key(node, &disk_key, nr);
2217                         if (btrfs_disk_key_objectid(&disk_key) != objectid)
2218                                 break;
2219                 }
2220                 search = btrfs_node_blockptr(node, nr);
2221                 if ((search <= target && target - search <= 65536) ||
2222                     (search > target && search - target <= 65536)) {
2223                         readahead_tree_block(fs_info, search);
2224                         nread += blocksize;
2225                 }
2226                 nscan++;
2227                 if ((nread > 65536 || nscan > 32))
2228                         break;
2229         }
2230 }
2231
2232 static noinline void reada_for_balance(struct btrfs_fs_info *fs_info,
2233                                        struct btrfs_path *path, int level)
2234 {
2235         int slot;
2236         int nritems;
2237         struct extent_buffer *parent;
2238         struct extent_buffer *eb;
2239         u64 gen;
2240         u64 block1 = 0;
2241         u64 block2 = 0;
2242
2243         parent = path->nodes[level + 1];
2244         if (!parent)
2245                 return;
2246
2247         nritems = btrfs_header_nritems(parent);
2248         slot = path->slots[level + 1];
2249
2250         if (slot > 0) {
2251                 block1 = btrfs_node_blockptr(parent, slot - 1);
2252                 gen = btrfs_node_ptr_generation(parent, slot - 1);
2253                 eb = find_extent_buffer(fs_info, block1);
2254                 /*
2255                  * if we get -eagain from btrfs_buffer_uptodate, we
2256                  * don't want to return eagain here.  That will loop
2257                  * forever
2258                  */
2259                 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
2260                         block1 = 0;
2261                 free_extent_buffer(eb);
2262         }
2263         if (slot + 1 < nritems) {
2264                 block2 = btrfs_node_blockptr(parent, slot + 1);
2265                 gen = btrfs_node_ptr_generation(parent, slot + 1);
2266                 eb = find_extent_buffer(fs_info, block2);
2267                 if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
2268                         block2 = 0;
2269                 free_extent_buffer(eb);
2270         }
2271
2272         if (block1)
2273                 readahead_tree_block(fs_info, block1);
2274         if (block2)
2275                 readahead_tree_block(fs_info, block2);
2276 }
2277
2278
2279 /*
2280  * when we walk down the tree, it is usually safe to unlock the higher layers
2281  * in the tree.  The exceptions are when our path goes through slot 0, because
2282  * operations on the tree might require changing key pointers higher up in the
2283  * tree.
2284  *
2285  * callers might also have set path->keep_locks, which tells this code to keep
2286  * the lock if the path points to the last slot in the block.  This is part of
2287  * walking through the tree, and selecting the next slot in the higher block.
2288  *
2289  * lowest_unlock sets the lowest level in the tree we're allowed to unlock.  so
2290  * if lowest_unlock is 1, level 0 won't be unlocked
2291  */
2292 static noinline void unlock_up(struct btrfs_path *path, int level,
2293                                int lowest_unlock, int min_write_lock_level,
2294                                int *write_lock_level)
2295 {
2296         int i;
2297         int skip_level = level;
2298         int no_skips = 0;
2299         struct extent_buffer *t;
2300
2301         for (i = level; i < BTRFS_MAX_LEVEL; i++) {
2302                 if (!path->nodes[i])
2303                         break;
2304                 if (!path->locks[i])
2305                         break;
2306                 if (!no_skips && path->slots[i] == 0) {
2307                         skip_level = i + 1;
2308                         continue;
2309                 }
2310                 if (!no_skips && path->keep_locks) {
2311                         u32 nritems;
2312                         t = path->nodes[i];
2313                         nritems = btrfs_header_nritems(t);
2314                         if (nritems < 1 || path->slots[i] >= nritems - 1) {
2315                                 skip_level = i + 1;
2316                                 continue;
2317                         }
2318                 }
2319                 if (skip_level < i && i >= lowest_unlock)
2320                         no_skips = 1;
2321
2322                 t = path->nodes[i];
2323                 if (i >= lowest_unlock && i > skip_level) {
2324                         btrfs_tree_unlock_rw(t, path->locks[i]);
2325                         path->locks[i] = 0;
2326                         if (write_lock_level &&
2327                             i > min_write_lock_level &&
2328                             i <= *write_lock_level) {
2329                                 *write_lock_level = i - 1;
2330                         }
2331                 }
2332         }
2333 }
2334
2335 /*
2336  * helper function for btrfs_search_slot.  The goal is to find a block
2337  * in cache without setting the path to blocking.  If we find the block
2338  * we return zero and the path is unchanged.
2339  *
2340  * If we can't find the block, we set the path blocking and do some
2341  * reada.  -EAGAIN is returned and the search must be repeated.
2342  */
2343 static int
2344 read_block_for_search(struct btrfs_root *root, struct btrfs_path *p,
2345                       struct extent_buffer **eb_ret, int level, int slot,
2346                       const struct btrfs_key *key)
2347 {
2348         struct btrfs_fs_info *fs_info = root->fs_info;
2349         u64 blocknr;
2350         u64 gen;
2351         struct extent_buffer *b = *eb_ret;
2352         struct extent_buffer *tmp;
2353         struct btrfs_key first_key;
2354         int ret;
2355         int parent_level;
2356
2357         blocknr = btrfs_node_blockptr(b, slot);
2358         gen = btrfs_node_ptr_generation(b, slot);
2359         parent_level = btrfs_header_level(b);
2360         btrfs_node_key_to_cpu(b, &first_key, slot);
2361
2362         tmp = find_extent_buffer(fs_info, blocknr);
2363         if (tmp) {
2364                 /* first we do an atomic uptodate check */
2365                 if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
2366                         /*
2367                          * Do extra check for first_key, eb can be stale due to
2368                          * being cached, read from scrub, or have multiple
2369                          * parents (shared tree blocks).
2370                          */
2371                         if (btrfs_verify_level_key(tmp,
2372                                         parent_level - 1, &first_key, gen)) {
2373                                 free_extent_buffer(tmp);
2374                                 return -EUCLEAN;
2375                         }
2376                         *eb_ret = tmp;
2377                         return 0;
2378                 }
2379
2380                 /* the pages were up to date, but we failed
2381                  * the generation number check.  Do a full
2382                  * read for the generation number that is correct.
2383                  * We must do this without dropping locks so
2384                  * we can trust our generation number
2385                  */
2386                 btrfs_set_path_blocking(p);
2387
2388                 /* now we're allowed to do a blocking uptodate check */
2389                 ret = btrfs_read_buffer(tmp, gen, parent_level - 1, &first_key);
2390                 if (!ret) {
2391                         *eb_ret = tmp;
2392                         return 0;
2393                 }
2394                 free_extent_buffer(tmp);
2395                 btrfs_release_path(p);
2396                 return -EIO;
2397         }
2398
2399         /*
2400          * reduce lock contention at high levels
2401          * of the btree by dropping locks before
2402          * we read.  Don't release the lock on the current
2403          * level because we need to walk this node to figure
2404          * out which blocks to read.
2405          */
2406         btrfs_unlock_up_safe(p, level + 1);
2407         btrfs_set_path_blocking(p);
2408
2409         if (p->reada != READA_NONE)
2410                 reada_for_search(fs_info, p, level, slot, key->objectid);
2411
2412         ret = -EAGAIN;
2413         tmp = read_tree_block(fs_info, blocknr, gen, parent_level - 1,
2414                               &first_key);
2415         if (!IS_ERR(tmp)) {
2416                 /*
2417                  * If the read above didn't mark this buffer up to date,
2418                  * it will never end up being up to date.  Set ret to EIO now
2419                  * and give up so that our caller doesn't loop forever
2420                  * on our EAGAINs.
2421                  */
2422                 if (!extent_buffer_uptodate(tmp))
2423                         ret = -EIO;
2424                 free_extent_buffer(tmp);
2425         } else {
2426                 ret = PTR_ERR(tmp);
2427         }
2428
2429         btrfs_release_path(p);
2430         return ret;
2431 }
2432
2433 /*
2434  * helper function for btrfs_search_slot.  This does all of the checks
2435  * for node-level blocks and does any balancing required based on
2436  * the ins_len.
2437  *
2438  * If no extra work was required, zero is returned.  If we had to
2439  * drop the path, -EAGAIN is returned and btrfs_search_slot must
2440  * start over
2441  */
2442 static int
2443 setup_nodes_for_search(struct btrfs_trans_handle *trans,
2444                        struct btrfs_root *root, struct btrfs_path *p,
2445                        struct extent_buffer *b, int level, int ins_len,
2446                        int *write_lock_level)
2447 {
2448         struct btrfs_fs_info *fs_info = root->fs_info;
2449         int ret;
2450
2451         if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
2452             BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) {
2453                 int sret;
2454
2455                 if (*write_lock_level < level + 1) {
2456                         *write_lock_level = level + 1;
2457                         btrfs_release_path(p);
2458                         goto again;
2459                 }
2460
2461                 btrfs_set_path_blocking(p);
2462                 reada_for_balance(fs_info, p, level);
2463                 sret = split_node(trans, root, p, level);
2464
2465                 BUG_ON(sret > 0);
2466                 if (sret) {
2467                         ret = sret;
2468                         goto done;
2469                 }
2470                 b = p->nodes[level];
2471         } else if (ins_len < 0 && btrfs_header_nritems(b) <
2472                    BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 2) {
2473                 int sret;
2474
2475                 if (*write_lock_level < level + 1) {
2476                         *write_lock_level = level + 1;
2477                         btrfs_release_path(p);
2478                         goto again;
2479                 }
2480
2481                 btrfs_set_path_blocking(p);
2482                 reada_for_balance(fs_info, p, level);
2483                 sret = balance_level(trans, root, p, level);
2484
2485                 if (sret) {
2486                         ret = sret;
2487                         goto done;
2488                 }
2489                 b = p->nodes[level];
2490                 if (!b) {
2491                         btrfs_release_path(p);
2492                         goto again;
2493                 }
2494                 BUG_ON(btrfs_header_nritems(b) == 1);
2495         }
2496         return 0;
2497
2498 again:
2499         ret = -EAGAIN;
2500 done:
2501         return ret;
2502 }
2503
2504 static int key_search(struct extent_buffer *b, const struct btrfs_key *key,
2505                       int level, int *prev_cmp, int *slot)
2506 {
2507         if (*prev_cmp != 0) {
2508                 *prev_cmp = btrfs_bin_search(b, key, level, slot);
2509                 return *prev_cmp;
2510         }
2511
2512         *slot = 0;
2513
2514         return 0;
2515 }
2516
2517 int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
2518                 u64 iobjectid, u64 ioff, u8 key_type,
2519                 struct btrfs_key *found_key)
2520 {
2521         int ret;
2522         struct btrfs_key key;
2523         struct extent_buffer *eb;
2524
2525         ASSERT(path);
2526         ASSERT(found_key);
2527
2528         key.type = key_type;
2529         key.objectid = iobjectid;
2530         key.offset = ioff;
2531
2532         ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
2533         if (ret < 0)
2534                 return ret;
2535
2536         eb = path->nodes[0];
2537         if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
2538                 ret = btrfs_next_leaf(fs_root, path);
2539                 if (ret)
2540                         return ret;
2541                 eb = path->nodes[0];
2542         }
2543
2544         btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
2545         if (found_key->type != key.type ||
2546                         found_key->objectid != key.objectid)
2547                 return 1;
2548
2549         return 0;
2550 }
2551
2552 static struct extent_buffer *btrfs_search_slot_get_root(struct btrfs_root *root,
2553                                                         struct btrfs_path *p,
2554                                                         int write_lock_level)
2555 {
2556         struct btrfs_fs_info *fs_info = root->fs_info;
2557         struct extent_buffer *b;
2558         int root_lock;
2559         int level = 0;
2560
2561         /* We try very hard to do read locks on the root */
2562         root_lock = BTRFS_READ_LOCK;
2563
2564         if (p->search_commit_root) {
2565                 /*
2566                  * The commit roots are read only so we always do read locks,
2567                  * and we always must hold the commit_root_sem when doing
2568                  * searches on them, the only exception is send where we don't
2569                  * want to block transaction commits for a long time, so
2570                  * we need to clone the commit root in order to avoid races
2571                  * with transaction commits that create a snapshot of one of
2572                  * the roots used by a send operation.
2573                  */
2574                 if (p->need_commit_sem) {
2575                         down_read(&fs_info->commit_root_sem);
2576                         b = btrfs_clone_extent_buffer(root->commit_root);
2577                         up_read(&fs_info->commit_root_sem);
2578                         if (!b)
2579                                 return ERR_PTR(-ENOMEM);
2580
2581                 } else {
2582                         b = root->commit_root;
2583                         atomic_inc(&b->refs);
2584                 }
2585                 level = btrfs_header_level(b);
2586                 /*
2587                  * Ensure that all callers have set skip_locking when
2588                  * p->search_commit_root = 1.
2589                  */
2590                 ASSERT(p->skip_locking == 1);
2591
2592                 goto out;
2593         }
2594
2595         if (p->skip_locking) {
2596                 b = btrfs_root_node(root);
2597                 level = btrfs_header_level(b);
2598                 goto out;
2599         }
2600
2601         /*
2602          * If the level is set to maximum, we can skip trying to get the read
2603          * lock.
2604          */
2605         if (write_lock_level < BTRFS_MAX_LEVEL) {
2606                 /*
2607                  * We don't know the level of the root node until we actually
2608                  * have it read locked
2609                  */
2610                 b = btrfs_read_lock_root_node(root);
2611                 level = btrfs_header_level(b);
2612                 if (level > write_lock_level)
2613                         goto out;
2614
2615                 /* Whoops, must trade for write lock */
2616                 btrfs_tree_read_unlock(b);
2617                 free_extent_buffer(b);
2618         }
2619
2620         b = btrfs_lock_root_node(root);
2621         root_lock = BTRFS_WRITE_LOCK;
2622
2623         /* The level might have changed, check again */
2624         level = btrfs_header_level(b);
2625
2626 out:
2627         p->nodes[level] = b;
2628         if (!p->skip_locking)
2629                 p->locks[level] = root_lock;
2630         /*
2631          * Callers are responsible for dropping b's references.
2632          */
2633         return b;
2634 }
2635
2636
2637 /*
2638  * btrfs_search_slot - look for a key in a tree and perform necessary
2639  * modifications to preserve tree invariants.
2640  *
2641  * @trans:      Handle of transaction, used when modifying the tree
2642  * @p:          Holds all btree nodes along the search path
2643  * @root:       The root node of the tree
2644  * @key:        The key we are looking for
2645  * @ins_len:    Indicates purpose of search, for inserts it is 1, for
2646  *              deletions it's -1. 0 for plain searches
2647  * @cow:        boolean should CoW operations be performed. Must always be 1
2648  *              when modifying the tree.
2649  *
2650  * If @ins_len > 0, nodes and leaves will be split as we walk down the tree.
2651  * If @ins_len < 0, nodes will be merged as we walk down the tree (if possible)
2652  *
2653  * If @key is found, 0 is returned and you can find the item in the leaf level
2654  * of the path (level 0)
2655  *
2656  * If @key isn't found, 1 is returned and the leaf level of the path (level 0)
2657  * points to the slot where it should be inserted
2658  *
2659  * If an error is encountered while searching the tree a negative error number
2660  * is returned
2661  */
2662 int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2663                       const struct btrfs_key *key, struct btrfs_path *p,
2664                       int ins_len, int cow)
2665 {
2666         struct extent_buffer *b;
2667         int slot;
2668         int ret;
2669         int err;
2670         int level;
2671         int lowest_unlock = 1;
2672         /* everything at write_lock_level or lower must be write locked */
2673         int write_lock_level = 0;
2674         u8 lowest_level = 0;
2675         int min_write_lock_level;
2676         int prev_cmp;
2677
2678         lowest_level = p->lowest_level;
2679         WARN_ON(lowest_level && ins_len > 0);
2680         WARN_ON(p->nodes[0] != NULL);
2681         BUG_ON(!cow && ins_len);
2682
2683         if (ins_len < 0) {
2684                 lowest_unlock = 2;
2685
2686                 /* when we are removing items, we might have to go up to level
2687                  * two as we update tree pointers  Make sure we keep write
2688                  * for those levels as well
2689                  */
2690                 write_lock_level = 2;
2691         } else if (ins_len > 0) {
2692                 /*
2693                  * for inserting items, make sure we have a write lock on
2694                  * level 1 so we can update keys
2695                  */
2696                 write_lock_level = 1;
2697         }
2698
2699         if (!cow)
2700                 write_lock_level = -1;
2701
2702         if (cow && (p->keep_locks || p->lowest_level))
2703                 write_lock_level = BTRFS_MAX_LEVEL;
2704
2705         min_write_lock_level = write_lock_level;
2706
2707 again:
2708         prev_cmp = -1;
2709         b = btrfs_search_slot_get_root(root, p, write_lock_level);
2710         if (IS_ERR(b)) {
2711                 ret = PTR_ERR(b);
2712                 goto done;
2713         }
2714
2715         while (b) {
2716                 int dec = 0;
2717
2718                 level = btrfs_header_level(b);
2719
2720                 if (cow) {
2721                         bool last_level = (level == (BTRFS_MAX_LEVEL - 1));
2722
2723                         /*
2724                          * if we don't really need to cow this block
2725                          * then we don't want to set the path blocking,
2726                          * so we test it here
2727                          */
2728                         if (!should_cow_block(trans, root, b)) {
2729                                 trans->dirty = true;
2730                                 goto cow_done;
2731                         }
2732
2733                         /*
2734                          * must have write locks on this node and the
2735                          * parent
2736                          */
2737                         if (level > write_lock_level ||
2738                             (level + 1 > write_lock_level &&
2739                             level + 1 < BTRFS_MAX_LEVEL &&
2740                             p->nodes[level + 1])) {
2741                                 write_lock_level = level + 1;
2742                                 btrfs_release_path(p);
2743                                 goto again;
2744                         }
2745
2746                         btrfs_set_path_blocking(p);
2747                         if (last_level)
2748                                 err = btrfs_cow_block(trans, root, b, NULL, 0,
2749                                                       &b);
2750                         else
2751                                 err = btrfs_cow_block(trans, root, b,
2752                                                       p->nodes[level + 1],
2753                                                       p->slots[level + 1], &b);
2754                         if (err) {
2755                                 ret = err;
2756                                 goto done;
2757                         }
2758                 }
2759 cow_done:
2760                 p->nodes[level] = b;
2761                 /*
2762                  * Leave path with blocking locks to avoid massive
2763                  * lock context switch, this is made on purpose.
2764                  */
2765
2766                 /*
2767                  * we have a lock on b and as long as we aren't changing
2768                  * the tree, there is no way to for the items in b to change.
2769                  * It is safe to drop the lock on our parent before we
2770                  * go through the expensive btree search on b.
2771                  *
2772                  * If we're inserting or deleting (ins_len != 0), then we might
2773                  * be changing slot zero, which may require changing the parent.
2774                  * So, we can't drop the lock until after we know which slot
2775                  * we're operating on.
2776                  */
2777                 if (!ins_len && !p->keep_locks) {
2778                         int u = level + 1;
2779
2780                         if (u < BTRFS_MAX_LEVEL && p->locks[u]) {
2781                                 btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]);
2782                                 p->locks[u] = 0;
2783                         }
2784                 }
2785
2786                 ret = key_search(b, key, level, &prev_cmp, &slot);
2787                 if (ret < 0)
2788                         goto done;
2789
2790                 if (level == 0) {
2791                         p->slots[level] = slot;
2792                         if (ins_len > 0 &&
2793                             btrfs_leaf_free_space(b) < ins_len) {
2794                                 if (write_lock_level < 1) {
2795                                         write_lock_level = 1;
2796                                         btrfs_release_path(p);
2797                                         goto again;
2798                                 }
2799
2800                                 btrfs_set_path_blocking(p);
2801                                 err = split_leaf(trans, root, key,
2802                                                  p, ins_len, ret == 0);
2803
2804                                 BUG_ON(err > 0);
2805                                 if (err) {
2806                                         ret = err;
2807                                         goto done;
2808                                 }
2809                         }
2810                         if (!p->search_for_split)
2811                                 unlock_up(p, level, lowest_unlock,
2812                                           min_write_lock_level, NULL);
2813                         goto done;
2814                 }
2815                 if (ret && slot > 0) {
2816                         dec = 1;
2817                         slot--;
2818                 }
2819                 p->slots[level] = slot;
2820                 err = setup_nodes_for_search(trans, root, p, b, level, ins_len,
2821                                              &write_lock_level);
2822                 if (err == -EAGAIN)
2823                         goto again;
2824                 if (err) {
2825                         ret = err;
2826                         goto done;
2827                 }
2828                 b = p->nodes[level];
2829                 slot = p->slots[level];
2830
2831                 /*
2832                  * Slot 0 is special, if we change the key we have to update
2833                  * the parent pointer which means we must have a write lock on
2834                  * the parent
2835                  */
2836                 if (slot == 0 && ins_len && write_lock_level < level + 1) {
2837                         write_lock_level = level + 1;
2838                         btrfs_release_path(p);
2839                         goto again;
2840                 }
2841
2842                 unlock_up(p, level, lowest_unlock, min_write_lock_level,
2843                           &write_lock_level);
2844
2845                 if (level == lowest_level) {
2846                         if (dec)
2847                                 p->slots[level]++;
2848                         goto done;
2849                 }
2850
2851                 err = read_block_for_search(root, p, &b, level, slot, key);
2852                 if (err == -EAGAIN)
2853                         goto again;
2854                 if (err) {
2855                         ret = err;
2856                         goto done;
2857                 }
2858
2859                 if (!p->skip_locking) {
2860                         level = btrfs_header_level(b);
2861                         if (level <= write_lock_level) {
2862                                 if (!btrfs_try_tree_write_lock(b)) {
2863                                         btrfs_set_path_blocking(p);
2864                                         btrfs_tree_lock(b);
2865                                 }
2866                                 p->locks[level] = BTRFS_WRITE_LOCK;
2867                         } else {
2868                                 if (!btrfs_tree_read_lock_atomic(b)) {
2869                                         btrfs_set_path_blocking(p);
2870                                         btrfs_tree_read_lock(b);
2871                                 }
2872                                 p->locks[level] = BTRFS_READ_LOCK;
2873                         }
2874                         p->nodes[level] = b;
2875                 }
2876         }
2877         ret = 1;
2878 done:
2879         /*
2880          * we don't really know what they plan on doing with the path
2881          * from here on, so for now just mark it as blocking
2882          */
2883         if (!p->leave_spinning)
2884                 btrfs_set_path_blocking(p);
2885         if (ret < 0 && !p->skip_release_on_error)
2886                 btrfs_release_path(p);
2887         return ret;
2888 }
2889
2890 /*
2891  * Like btrfs_search_slot, this looks for a key in the given tree. It uses the
2892  * current state of the tree together with the operations recorded in the tree
2893  * modification log to search for the key in a previous version of this tree, as
2894  * denoted by the time_seq parameter.
2895  *
2896  * Naturally, there is no support for insert, delete or cow operations.
2897  *
2898  * The resulting path and return value will be set up as if we called
2899  * btrfs_search_slot at that point in time with ins_len and cow both set to 0.
2900  */
2901 int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key,
2902                           struct btrfs_path *p, u64 time_seq)
2903 {
2904         struct btrfs_fs_info *fs_info = root->fs_info;
2905         struct extent_buffer *b;
2906         int slot;
2907         int ret;
2908         int err;
2909         int level;
2910         int lowest_unlock = 1;
2911         u8 lowest_level = 0;
2912         int prev_cmp = -1;
2913
2914         lowest_level = p->lowest_level;
2915         WARN_ON(p->nodes[0] != NULL);
2916
2917         if (p->search_commit_root) {
2918                 BUG_ON(time_seq);
2919                 return btrfs_search_slot(NULL, root, key, p, 0, 0);
2920         }
2921
2922 again:
2923         b = get_old_root(root, time_seq);
2924         if (!b) {
2925                 ret = -EIO;
2926                 goto done;
2927         }
2928         level = btrfs_header_level(b);
2929         p->locks[level] = BTRFS_READ_LOCK;
2930
2931         while (b) {
2932                 int dec = 0;
2933
2934                 level = btrfs_header_level(b);
2935                 p->nodes[level] = b;
2936
2937                 /*
2938                  * we have a lock on b and as long as we aren't changing
2939                  * the tree, there is no way to for the items in b to change.
2940                  * It is safe to drop the lock on our parent before we
2941                  * go through the expensive btree search on b.
2942                  */
2943                 btrfs_unlock_up_safe(p, level + 1);
2944
2945                 /*
2946                  * Since we can unwind ebs we want to do a real search every
2947                  * time.
2948                  */
2949                 prev_cmp = -1;
2950                 ret = key_search(b, key, level, &prev_cmp, &slot);
2951                 if (ret < 0)
2952                         goto done;
2953
2954                 if (level == 0) {
2955                         p->slots[level] = slot;
2956                         unlock_up(p, level, lowest_unlock, 0, NULL);
2957                         goto done;
2958                 }
2959
2960                 if (ret && slot > 0) {
2961                         dec = 1;
2962                         slot--;
2963                 }
2964                 p->slots[level] = slot;
2965                 unlock_up(p, level, lowest_unlock, 0, NULL);
2966
2967                 if (level == lowest_level) {
2968                         if (dec)
2969                                 p->slots[level]++;
2970                         goto done;
2971                 }
2972
2973                 err = read_block_for_search(root, p, &b, level, slot, key);
2974                 if (err == -EAGAIN)
2975                         goto again;
2976                 if (err) {
2977                         ret = err;
2978                         goto done;
2979                 }
2980
2981                 level = btrfs_header_level(b);
2982                 if (!btrfs_tree_read_lock_atomic(b)) {
2983                         btrfs_set_path_blocking(p);
2984                         btrfs_tree_read_lock(b);
2985                 }
2986                 b = tree_mod_log_rewind(fs_info, p, b, time_seq);
2987                 if (!b) {
2988                         ret = -ENOMEM;
2989                         goto done;
2990                 }
2991                 p->locks[level] = BTRFS_READ_LOCK;
2992                 p->nodes[level] = b;
2993         }
2994         ret = 1;
2995 done:
2996         if (!p->leave_spinning)
2997                 btrfs_set_path_blocking(p);
2998         if (ret < 0)
2999                 btrfs_release_path(p);
3000
3001         return ret;
3002 }
3003
3004 /*
3005  * helper to use instead of search slot if no exact match is needed but
3006  * instead the next or previous item should be returned.
3007  * When find_higher is true, the next higher item is returned, the next lower
3008  * otherwise.
3009  * When return_any and find_higher are both true, and no higher item is found,
3010  * return the next lower instead.
3011  * When return_any is true and find_higher is false, and no lower item is found,
3012  * return the next higher instead.
3013  * It returns 0 if any item is found, 1 if none is found (tree empty), and
3014  * < 0 on error
3015  */
3016 int btrfs_search_slot_for_read(struct btrfs_root *root,
3017                                const struct btrfs_key *key,
3018                                struct btrfs_path *p, int find_higher,
3019                                int return_any)
3020 {
3021         int ret;
3022         struct extent_buffer *leaf;
3023
3024 again:
3025         ret = btrfs_search_slot(NULL, root, key, p, 0, 0);
3026         if (ret <= 0)
3027                 return ret;
3028         /*
3029          * a return value of 1 means the path is at the position where the
3030          * item should be inserted. Normally this is the next bigger item,
3031          * but in case the previous item is the last in a leaf, path points
3032          * to the first free slot in the previous leaf, i.e. at an invalid
3033          * item.
3034          */
3035         leaf = p->nodes[0];
3036
3037         if (find_higher) {
3038                 if (p->slots[0] >= btrfs_header_nritems(leaf)) {
3039                         ret = btrfs_next_leaf(root, p);
3040                         if (ret <= 0)
3041                                 return ret;
3042                         if (!return_any)
3043                                 return 1;
3044                         /*
3045                          * no higher item found, return the next
3046                          * lower instead
3047                          */
3048                         return_any = 0;
3049                         find_higher = 0;
3050                         btrfs_release_path(p);
3051                         goto again;
3052                 }
3053         } else {
3054                 if (p->slots[0] == 0) {
3055                         ret = btrfs_prev_leaf(root, p);
3056                         if (ret < 0)
3057                                 return ret;
3058                         if (!ret) {
3059                                 leaf = p->nodes[0];
3060                                 if (p->slots[0] == btrfs_header_nritems(leaf))
3061                                         p->slots[0]--;
3062                                 return 0;
3063                         }
3064                         if (!return_any)
3065                                 return 1;
3066                         /*
3067                          * no lower item found, return the next
3068                          * higher instead
3069                          */
3070                         return_any = 0;
3071                         find_higher = 1;
3072                         btrfs_release_path(p);
3073                         goto again;
3074                 } else {
3075                         --p->slots[0];
3076                 }
3077         }
3078         return 0;
3079 }
3080
3081 /*
3082  * adjust the pointers going up the tree, starting at level
3083  * making sure the right key of each node is points to 'key'.
3084  * This is used after shifting pointers to the left, so it stops
3085  * fixing up pointers when a given leaf/node is not in slot 0 of the
3086  * higher levels
3087  *
3088  */
3089 static void fixup_low_keys(struct btrfs_path *path,
3090                            struct btrfs_disk_key *key, int level)
3091 {
3092         int i;
3093         struct extent_buffer *t;
3094         int ret;
3095
3096         for (i = level; i < BTRFS_MAX_LEVEL; i++) {
3097                 int tslot = path->slots[i];
3098
3099                 if (!path->nodes[i])
3100                         break;
3101                 t = path->nodes[i];
3102                 ret = tree_mod_log_insert_key(t, tslot, MOD_LOG_KEY_REPLACE,
3103                                 GFP_ATOMIC);
3104                 BUG_ON(ret < 0);
3105                 btrfs_set_node_key(t, key, tslot);
3106                 btrfs_mark_buffer_dirty(path->nodes[i]);
3107                 if (tslot != 0)
3108                         break;
3109         }
3110 }
3111
3112 /*
3113  * update item key.
3114  *
3115  * This function isn't completely safe. It's the caller's responsibility
3116  * that the new key won't break the order
3117  */
3118 void btrfs_set_item_key_safe(struct btrfs_fs_info *fs_info,
3119                              struct btrfs_path *path,
3120                              const struct btrfs_key *new_key)
3121 {
3122         struct btrfs_disk_key disk_key;
3123         struct extent_buffer *eb;
3124         int slot;
3125
3126         eb = path->nodes[0];
3127         slot = path->slots[0];
3128         if (slot > 0) {
3129                 btrfs_item_key(eb, &disk_key, slot - 1);
3130                 if (unlikely(comp_keys(&disk_key, new_key) >= 0)) {
3131                         btrfs_crit(fs_info,
3132                 "slot %u key (%llu %u %llu) new key (%llu %u %llu)",
3133                                    slot, btrfs_disk_key_objectid(&disk_key),
3134                                    btrfs_disk_key_type(&disk_key),
3135                                    btrfs_disk_key_offset(&disk_key),
3136                                    new_key->objectid, new_key->type,
3137                                    new_key->offset);
3138                         btrfs_print_leaf(eb);
3139                         BUG();
3140                 }
3141         }
3142         if (slot < btrfs_header_nritems(eb) - 1) {
3143                 btrfs_item_key(eb, &disk_key, slot + 1);
3144                 if (unlikely(comp_keys(&disk_key, new_key) <= 0)) {
3145                         btrfs_crit(fs_info,
3146                 "slot %u key (%llu %u %llu) new key (%llu %u %llu)",
3147                                    slot, btrfs_disk_key_objectid(&disk_key),
3148                                    btrfs_disk_key_type(&disk_key),
3149                                    btrfs_disk_key_offset(&disk_key),
3150                                    new_key->objectid, new_key->type,
3151                                    new_key->offset);
3152                         btrfs_print_leaf(eb);
3153                         BUG();
3154                 }
3155         }
3156
3157         btrfs_cpu_key_to_disk(&disk_key, new_key);
3158         btrfs_set_item_key(eb, &disk_key, slot);
3159         btrfs_mark_buffer_dirty(eb);
3160         if (slot == 0)
3161                 fixup_low_keys(path, &disk_key, 1);
3162 }
3163
3164 /*
3165  * try to push data from one node into the next node left in the
3166  * tree.
3167  *
3168  * returns 0 if some ptrs were pushed left, < 0 if there was some horrible
3169  * error, and > 0 if there was no room in the left hand block.
3170  */
3171 static int push_node_left(struct btrfs_trans_handle *trans,
3172                           struct extent_buffer *dst,
3173                           struct extent_buffer *src, int empty)
3174 {
3175         struct btrfs_fs_info *fs_info = trans->fs_info;
3176         int push_items = 0;
3177         int src_nritems;
3178         int dst_nritems;
3179         int ret = 0;
3180
3181         src_nritems = btrfs_header_nritems(src);
3182         dst_nritems = btrfs_header_nritems(dst);
3183         push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems;
3184         WARN_ON(btrfs_header_generation(src) != trans->transid);
3185         WARN_ON(btrfs_header_generation(dst) != trans->transid);
3186
3187         if (!empty && src_nritems <= 8)
3188                 return 1;
3189
3190         if (push_items <= 0)
3191                 return 1;
3192
3193         if (empty) {
3194                 push_items = min(src_nritems, push_items);
3195                 if (push_items < src_nritems) {
3196                         /* leave at least 8 pointers in the node if
3197                          * we aren't going to empty it
3198                          */
3199                         if (src_nritems - push_items < 8) {
3200                                 if (push_items <= 8)
3201                                         return 1;
3202                                 push_items -= 8;
3203                         }
3204                 }
3205         } else
3206                 push_items = min(src_nritems - 8, push_items);
3207
3208         ret = tree_mod_log_eb_copy(dst, src, dst_nritems, 0, push_items);
3209         if (ret) {
3210                 btrfs_abort_transaction(trans, ret);
3211                 return ret;
3212         }
3213         copy_extent_buffer(dst, src,
3214                            btrfs_node_key_ptr_offset(dst_nritems),
3215                            btrfs_node_key_ptr_offset(0),
3216                            push_items * sizeof(struct btrfs_key_ptr));
3217
3218         if (push_items < src_nritems) {
3219                 /*
3220                  * Don't call tree_mod_log_insert_move here, key removal was
3221                  * already fully logged by tree_mod_log_eb_copy above.
3222                  */
3223                 memmove_extent_buffer(src, btrfs_node_key_ptr_offset(0),
3224                                       btrfs_node_key_ptr_offset(push_items),
3225                                       (src_nritems - push_items) *
3226                                       sizeof(struct btrfs_key_ptr));
3227         }
3228         btrfs_set_header_nritems(src, src_nritems - push_items);
3229         btrfs_set_header_nritems(dst, dst_nritems + push_items);
3230         btrfs_mark_buffer_dirty(src);
3231         btrfs_mark_buffer_dirty(dst);
3232
3233         return ret;
3234 }
3235
3236 /*
3237  * try to push data from one node into the next node right in the
3238  * tree.
3239  *
3240  * returns 0 if some ptrs were pushed, < 0 if there was some horrible
3241  * error, and > 0 if there was no room in the right hand block.
3242  *
3243  * this will  only push up to 1/2 the contents of the left node over
3244  */
3245 static int balance_node_right(struct btrfs_trans_handle *trans,
3246                               struct extent_buffer *dst,
3247                               struct extent_buffer *src)
3248 {
3249         struct btrfs_fs_info *fs_info = trans->fs_info;
3250         int push_items = 0;
3251         int max_push;
3252         int src_nritems;
3253         int dst_nritems;
3254         int ret = 0;
3255
3256         WARN_ON(btrfs_header_generation(src) != trans->transid);
3257         WARN_ON(btrfs_header_generation(dst) != trans->transid);
3258
3259         src_nritems = btrfs_header_nritems(src);
3260         dst_nritems = btrfs_header_nritems(dst);
3261         push_items = BTRFS_NODEPTRS_PER_BLOCK(fs_info) - dst_nritems;
3262         if (push_items <= 0)
3263                 return 1;
3264
3265         if (src_nritems < 4)
3266                 return 1;
3267
3268         max_push = src_nritems / 2 + 1;
3269         /* don't try to empty the node */
3270         if (max_push >= src_nritems)
3271                 return 1;
3272
3273         if (max_push < push_items)
3274                 push_items = max_push;
3275
3276         ret = tree_mod_log_insert_move(dst, push_items, 0, dst_nritems);
3277         BUG_ON(ret < 0);
3278         memmove_extent_buffer(dst, btrfs_node_key_ptr_offset(push_items),
3279                                       btrfs_node_key_ptr_offset(0),
3280                                       (dst_nritems) *
3281                                       sizeof(struct btrfs_key_ptr));
3282
3283         ret = tree_mod_log_eb_copy(dst, src, 0, src_nritems - push_items,
3284                                    push_items);
3285         if (ret) {
3286                 btrfs_abort_transaction(trans, ret);
3287                 return ret;
3288         }
3289         copy_extent_buffer(dst, src,
3290                            btrfs_node_key_ptr_offset(0),
3291                            btrfs_node_key_ptr_offset(src_nritems - push_items),
3292                            push_items * sizeof(struct btrfs_key_ptr));
3293
3294         btrfs_set_header_nritems(src, src_nritems - push_items);
3295         btrfs_set_header_nritems(dst, dst_nritems + push_items);
3296
3297         btrfs_mark_buffer_dirty(src);
3298         btrfs_mark_buffer_dirty(dst);
3299
3300         return ret;
3301 }
3302
3303 /*
3304  * helper function to insert a new root level in the tree.
3305  * A new node is allocated, and a single item is inserted to
3306  * point to the existing root
3307  *
3308  * returns zero on success or < 0 on failure.
3309  */
3310 static noinline int insert_new_root(struct btrfs_trans_handle *trans,
3311                            struct btrfs_root *root,
3312                            struct btrfs_path *path, int level)
3313 {
3314         struct btrfs_fs_info *fs_info = root->fs_info;
3315         u64 lower_gen;
3316         struct extent_buffer *lower;
3317         struct extent_buffer *c;
3318         struct extent_buffer *old;
3319         struct btrfs_disk_key lower_key;
3320         int ret;
3321
3322         BUG_ON(path->nodes[level]);
3323         BUG_ON(path->nodes[level-1] != root->node);
3324
3325         lower = path->nodes[level-1];
3326         if (level == 1)
3327                 btrfs_item_key(lower, &lower_key, 0);
3328         else
3329                 btrfs_node_key(lower, &lower_key, 0);
3330
3331         c = alloc_tree_block_no_bg_flush(trans, root, 0, &lower_key, level,
3332                                          root->node->start, 0);
3333         if (IS_ERR(c))
3334                 return PTR_ERR(c);
3335
3336         root_add_used(root, fs_info->nodesize);
3337
3338         btrfs_set_header_nritems(c, 1);
3339         btrfs_set_node_key(c, &lower_key, 0);
3340         btrfs_set_node_blockptr(c, 0, lower->start);
3341         lower_gen = btrfs_header_generation(lower);
3342         WARN_ON(lower_gen != trans->transid);
3343
3344         btrfs_set_node_ptr_generation(c, 0, lower_gen);
3345
3346         btrfs_mark_buffer_dirty(c);
3347
3348         old = root->node;
3349         ret = tree_mod_log_insert_root(root->node, c, 0);
3350         BUG_ON(ret < 0);
3351         rcu_assign_pointer(root->node, c);
3352
3353         /* the super has an extra ref to root->node */
3354         free_extent_buffer(old);
3355
3356         add_root_to_dirty_list(root);
3357         atomic_inc(&c->refs);
3358         path->nodes[level] = c;
3359         path->locks[level] = BTRFS_WRITE_LOCK_BLOCKING;
3360         path->slots[level] = 0;
3361         return 0;
3362 }
3363
3364 /*
3365  * worker function to insert a single pointer in a node.
3366  * the node should have enough room for the pointer already
3367  *
3368  * slot and level indicate where you want the key to go, and
3369  * blocknr is the block the key points to.
3370  */
3371 static void insert_ptr(struct btrfs_trans_handle *trans,
3372                        struct btrfs_path *path,
3373                        struct btrfs_disk_key *key, u64 bytenr,
3374                        int slot, int level)
3375 {
3376         struct extent_buffer *lower;
3377         int nritems;
3378         int ret;
3379
3380         BUG_ON(!path->nodes[level]);
3381         btrfs_assert_tree_locked(path->nodes[level]);
3382         lower = path->nodes[level];
3383         nritems = btrfs_header_nritems(lower);
3384         BUG_ON(slot > nritems);
3385         BUG_ON(nritems == BTRFS_NODEPTRS_PER_BLOCK(trans->fs_info));
3386         if (slot != nritems) {
3387                 if (level) {
3388                         ret = tree_mod_log_insert_move(lower, slot + 1, slot,
3389                                         nritems - slot);
3390                         BUG_ON(ret < 0);
3391                 }
3392                 memmove_extent_buffer(lower,
3393                               btrfs_node_key_ptr_offset(slot + 1),
3394                               btrfs_node_key_ptr_offset(slot),
3395                               (nritems - slot) * sizeof(struct btrfs_key_ptr));
3396         }
3397         if (level) {
3398                 ret = tree_mod_log_insert_key(lower, slot, MOD_LOG_KEY_ADD,
3399                                 GFP_NOFS);
3400                 BUG_ON(ret < 0);
3401         }
3402         btrfs_set_node_key(lower, key, slot);
3403         btrfs_set_node_blockptr(lower, slot, bytenr);
3404         WARN_ON(trans->transid == 0);
3405         btrfs_set_node_ptr_generation(lower, slot, trans->transid);
3406         btrfs_set_header_nritems(lower, nritems + 1);
3407         btrfs_mark_buffer_dirty(lower);
3408 }
3409
3410 /*
3411  * split the node at the specified level in path in two.
3412  * The path is corrected to point to the appropriate node after the split
3413  *
3414  * Before splitting this tries to make some room in the node by pushing
3415  * left and right, if either one works, it returns right away.
3416  *
3417  * returns 0 on success and < 0 on failure
3418  */
3419 static noinline int split_node(struct btrfs_trans_handle *trans,
3420                                struct btrfs_root *root,
3421                                struct btrfs_path *path, int level)
3422 {
3423         struct btrfs_fs_info *fs_info = root->fs_info;
3424         struct extent_buffer *c;
3425         struct extent_buffer *split;
3426         struct btrfs_disk_key disk_key;
3427         int mid;
3428         int ret;
3429         u32 c_nritems;
3430
3431         c = path->nodes[level];
3432         WARN_ON(btrfs_header_generation(c) != trans->transid);
3433         if (c == root->node) {
3434                 /*
3435                  * trying to split the root, lets make a new one
3436                  *
3437                  * tree mod log: We don't log_removal old root in
3438                  * insert_new_root, because that root buffer will be kept as a
3439                  * normal node. We are going to log removal of half of the
3440                  * elements below with tree_mod_log_eb_copy. We're holding a
3441                  * tree lock on the buffer, which is why we cannot race with
3442                  * other tree_mod_log users.
3443                  */
3444                 ret = insert_new_root(trans, root, path, level + 1);
3445                 if (ret)
3446                         return ret;
3447         } else {
3448                 ret = push_nodes_for_insert(trans, root, path, level);
3449                 c = path->nodes[level];
3450                 if (!ret && btrfs_header_nritems(c) <
3451                     BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3)
3452                         return 0;
3453                 if (ret < 0)
3454                         return ret;
3455         }
3456
3457         c_nritems = btrfs_header_nritems(c);
3458         mid = (c_nritems + 1) / 2;
3459         btrfs_node_key(c, &disk_key, mid);
3460
3461         split = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, level,
3462                                              c->start, 0);
3463         if (IS_ERR(split))
3464                 return PTR_ERR(split);
3465
3466         root_add_used(root, fs_info->nodesize);
3467         ASSERT(btrfs_header_level(c) == level);
3468
3469         ret = tree_mod_log_eb_copy(split, c, 0, mid, c_nritems - mid);
3470         if (ret) {
3471                 btrfs_abort_transaction(trans, ret);
3472                 return ret;
3473         }
3474         copy_extent_buffer(split, c,
3475                            btrfs_node_key_ptr_offset(0),
3476                            btrfs_node_key_ptr_offset(mid),
3477                            (c_nritems - mid) * sizeof(struct btrfs_key_ptr));
3478         btrfs_set_header_nritems(split, c_nritems - mid);
3479         btrfs_set_header_nritems(c, mid);
3480         ret = 0;
3481
3482         btrfs_mark_buffer_dirty(c);
3483         btrfs_mark_buffer_dirty(split);
3484
3485         insert_ptr(trans, path, &disk_key, split->start,
3486                    path->slots[level + 1] + 1, level + 1);
3487
3488         if (path->slots[level] >= mid) {
3489                 path->slots[level] -= mid;
3490                 btrfs_tree_unlock(c);
3491                 free_extent_buffer(c);
3492                 path->nodes[level] = split;
3493                 path->slots[level + 1] += 1;
3494         } else {
3495                 btrfs_tree_unlock(split);
3496                 free_extent_buffer(split);
3497         }
3498         return ret;
3499 }
3500
3501 /*
3502  * how many bytes are required to store the items in a leaf.  start
3503  * and nr indicate which items in the leaf to check.  This totals up the
3504  * space used both by the item structs and the item data
3505  */
3506 static int leaf_space_used(struct extent_buffer *l, int start, int nr)
3507 {
3508         struct btrfs_item *start_item;
3509         struct btrfs_item *end_item;
3510         struct btrfs_map_token token;
3511         int data_len;
3512         int nritems = btrfs_header_nritems(l);
3513         int end = min(nritems, start + nr) - 1;
3514
3515         if (!nr)
3516                 return 0;
3517         btrfs_init_map_token(&token, l);
3518         start_item = btrfs_item_nr(start);
3519         end_item = btrfs_item_nr(end);
3520         data_len = btrfs_token_item_offset(l, start_item, &token) +
3521                 btrfs_token_item_size(l, start_item, &token);
3522         data_len = data_len - btrfs_token_item_offset(l, end_item, &token);
3523         data_len += sizeof(struct btrfs_item) * nr;
3524         WARN_ON(data_len < 0);
3525         return data_len;
3526 }
3527
3528 /*
3529  * The space between the end of the leaf items and
3530  * the start of the leaf data.  IOW, how much room
3531  * the leaf has left for both items and data
3532  */
3533 noinline int btrfs_leaf_free_space(struct extent_buffer *leaf)
3534 {
3535         struct btrfs_fs_info *fs_info = leaf->fs_info;
3536         int nritems = btrfs_header_nritems(leaf);
3537         int ret;
3538
3539         ret = BTRFS_LEAF_DATA_SIZE(fs_info) - leaf_space_used(leaf, 0, nritems);
3540         if (ret < 0) {
3541                 btrfs_crit(fs_info,
3542                            "leaf free space ret %d, leaf data size %lu, used %d nritems %d",
3543                            ret,
3544                            (unsigned long) BTRFS_LEAF_DATA_SIZE(fs_info),
3545                            leaf_space_used(leaf, 0, nritems), nritems);
3546         }
3547         return ret;
3548 }
3549
3550 /*
3551  * min slot controls the lowest index we're willing to push to the
3552  * right.  We'll push up to and including min_slot, but no lower
3553  */
3554 static noinline int __push_leaf_right(struct btrfs_path *path,
3555                                       int data_size, int empty,
3556                                       struct extent_buffer *right,
3557                                       int free_space, u32 left_nritems,
3558                                       u32 min_slot)
3559 {
3560         struct btrfs_fs_info *fs_info = right->fs_info;
3561         struct extent_buffer *left = path->nodes[0];
3562         struct extent_buffer *upper = path->nodes[1];
3563         struct btrfs_map_token token;
3564         struct btrfs_disk_key disk_key;
3565         int slot;
3566         u32 i;
3567         int push_space = 0;
3568         int push_items = 0;
3569         struct btrfs_item *item;
3570         u32 nr;
3571         u32 right_nritems;
3572         u32 data_end;
3573         u32 this_item_size;
3574
3575         if (empty)
3576                 nr = 0;
3577         else
3578                 nr = max_t(u32, 1, min_slot);
3579
3580         if (path->slots[0] >= left_nritems)
3581                 push_space += data_size;
3582
3583         slot = path->slots[1];
3584         i = left_nritems - 1;
3585         while (i >= nr) {
3586                 item = btrfs_item_nr(i);
3587
3588                 if (!empty && push_items > 0) {
3589                         if (path->slots[0] > i)
3590                                 break;
3591                         if (path->slots[0] == i) {
3592                                 int space = btrfs_leaf_free_space(left);
3593
3594                                 if (space + push_space * 2 > free_space)
3595                                         break;
3596                         }
3597                 }
3598
3599                 if (path->slots[0] == i)
3600                         push_space += data_size;
3601
3602                 this_item_size = btrfs_item_size(left, item);
3603                 if (this_item_size + sizeof(*item) + push_space > free_space)
3604                         break;
3605
3606                 push_items++;
3607                 push_space += this_item_size + sizeof(*item);
3608                 if (i == 0)
3609                         break;
3610                 i--;
3611         }
3612
3613         if (push_items == 0)
3614                 goto out_unlock;
3615
3616         WARN_ON(!empty && push_items == left_nritems);
3617
3618         /* push left to right */
3619         right_nritems = btrfs_header_nritems(right);
3620
3621         push_space = btrfs_item_end_nr(left, left_nritems - push_items);
3622         push_space -= leaf_data_end(left);
3623
3624         /* make room in the right data area */
3625         data_end = leaf_data_end(right);
3626         memmove_extent_buffer(right,
3627                               BTRFS_LEAF_DATA_OFFSET + data_end - push_space,
3628                               BTRFS_LEAF_DATA_OFFSET + data_end,
3629                               BTRFS_LEAF_DATA_SIZE(fs_info) - data_end);
3630
3631         /* copy from the left data area */
3632         copy_extent_buffer(right, left, BTRFS_LEAF_DATA_OFFSET +
3633                      BTRFS_LEAF_DATA_SIZE(fs_info) - push_space,
3634                      BTRFS_LEAF_DATA_OFFSET + leaf_data_end(left),
3635                      push_space);
3636
3637         memmove_extent_buffer(right, btrfs_item_nr_offset(push_items),
3638                               btrfs_item_nr_offset(0),
3639                               right_nritems * sizeof(struct btrfs_item));
3640
3641         /* copy the items from left to right */
3642         copy_extent_buffer(right, left, btrfs_item_nr_offset(0),
3643                    btrfs_item_nr_offset(left_nritems - push_items),
3644                    push_items * sizeof(struct btrfs_item));
3645
3646         /* update the item pointers */
3647         btrfs_init_map_token(&token, right);
3648         right_nritems += push_items;
3649         btrfs_set_header_nritems(right, right_nritems);
3650         push_space = BTRFS_LEAF_DATA_SIZE(fs_info);
3651         for (i = 0; i < right_nritems; i++) {
3652                 item = btrfs_item_nr(i);
3653                 push_space -= btrfs_token_item_size(right, item, &token);
3654                 btrfs_set_token_item_offset(right, item, push_space, &token);
3655         }
3656
3657         left_nritems -= push_items;
3658         btrfs_set_header_nritems(left, left_nritems);
3659
3660         if (left_nritems)
3661                 btrfs_mark_buffer_dirty(left);
3662         else
3663                 btrfs_clean_tree_block(left);
3664
3665         btrfs_mark_buffer_dirty(right);
3666
3667         btrfs_item_key(right, &disk_key, 0);
3668         btrfs_set_node_key(upper, &disk_key, slot + 1);
3669         btrfs_mark_buffer_dirty(upper);
3670
3671         /* then fixup the leaf pointer in the path */
3672         if (path->slots[0] >= left_nritems) {
3673                 path->slots[0] -= left_nritems;
3674                 if (btrfs_header_nritems(path->nodes[0]) == 0)
3675                         btrfs_clean_tree_block(path->nodes[0]);
3676                 btrfs_tree_unlock(path->nodes[0]);
3677                 free_extent_buffer(path->nodes[0]);
3678                 path->nodes[0] = right;
3679                 path->slots[1] += 1;
3680         } else {
3681                 btrfs_tree_unlock(right);
3682                 free_extent_buffer(right);
3683         }
3684         return 0;
3685
3686 out_unlock:
3687         btrfs_tree_unlock(right);
3688         free_extent_buffer(right);
3689         return 1;
3690 }
3691
3692 /*
3693  * push some data in the path leaf to the right, trying to free up at
3694  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
3695  *
3696  * returns 1 if the push failed because the other node didn't have enough
3697  * room, 0 if everything worked out and < 0 if there were major errors.
3698  *
3699  * this will push starting from min_slot to the end of the leaf.  It won't
3700  * push any slot lower than min_slot
3701  */
3702 static int push_leaf_right(struct btrfs_trans_handle *trans, struct btrfs_root
3703                            *root, struct btrfs_path *path,
3704                            int min_data_size, int data_size,
3705                            int empty, u32 min_slot)
3706 {
3707         struct extent_buffer *left = path->nodes[0];
3708         struct extent_buffer *right;
3709         struct extent_buffer *upper;
3710         int slot;
3711         int free_space;
3712         u32 left_nritems;
3713         int ret;
3714
3715         if (!path->nodes[1])
3716                 return 1;
3717
3718         slot = path->slots[1];
3719         upper = path->nodes[1];
3720         if (slot >= btrfs_header_nritems(upper) - 1)
3721                 return 1;
3722
3723         btrfs_assert_tree_locked(path->nodes[1]);
3724
3725         right = btrfs_read_node_slot(upper, slot + 1);
3726         /*
3727          * slot + 1 is not valid or we fail to read the right node,
3728          * no big deal, just return.
3729          */
3730         if (IS_ERR(right))
3731                 return 1;
3732
3733         btrfs_tree_lock(right);
3734         btrfs_set_lock_blocking_write(right);
3735
3736         free_space = btrfs_leaf_free_space(right);
3737         if (free_space < data_size)
3738                 goto out_unlock;
3739
3740         /* cow and double check */
3741         ret = btrfs_cow_block(trans, root, right, upper,
3742                               slot + 1, &right);
3743         if (ret)
3744                 goto out_unlock;
3745
3746         free_space = btrfs_leaf_free_space(right);
3747         if (free_space < data_size)
3748                 goto out_unlock;
3749
3750         left_nritems = btrfs_header_nritems(left);
3751         if (left_nritems == 0)
3752                 goto out_unlock;
3753
3754         if (path->slots[0] == left_nritems && !empty) {
3755                 /* Key greater than all keys in the leaf, right neighbor has
3756                  * enough room for it and we're not emptying our leaf to delete
3757                  * it, therefore use right neighbor to insert the new item and
3758                  * no need to touch/dirty our left leaf. */
3759                 btrfs_tree_unlock(left);
3760                 free_extent_buffer(left);
3761                 path->nodes[0] = right;
3762                 path->slots[0] = 0;
3763                 path->slots[1]++;
3764                 return 0;
3765         }
3766
3767         return __push_leaf_right(path, min_data_size, empty,
3768                                 right, free_space, left_nritems, min_slot);
3769 out_unlock:
3770         btrfs_tree_unlock(right);
3771         free_extent_buffer(right);
3772         return 1;
3773 }
3774
3775 /*
3776  * push some data in the path leaf to the left, trying to free up at
3777  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
3778  *
3779  * max_slot can put a limit on how far into the leaf we'll push items.  The
3780  * item at 'max_slot' won't be touched.  Use (u32)-1 to make us do all the
3781  * items
3782  */
3783 static noinline int __push_leaf_left(struct btrfs_path *path, int data_size,
3784                                      int empty, struct extent_buffer *left,
3785                                      int free_space, u32 right_nritems,
3786                                      u32 max_slot)
3787 {
3788         struct btrfs_fs_info *fs_info = left->fs_info;
3789         struct btrfs_disk_key disk_key;
3790         struct extent_buffer *right = path->nodes[0];
3791         int i;
3792         int push_space = 0;
3793         int push_items = 0;
3794         struct btrfs_item *item;
3795         u32 old_left_nritems;
3796         u32 nr;
3797         int ret = 0;
3798         u32 this_item_size;
3799         u32 old_left_item_size;
3800         struct btrfs_map_token token;
3801
3802         if (empty)
3803                 nr = min(right_nritems, max_slot);
3804         else
3805                 nr = min(right_nritems - 1, max_slot);
3806
3807         for (i = 0; i < nr; i++) {
3808                 item = btrfs_item_nr(i);
3809
3810                 if (!empty && push_items > 0) {
3811                         if (path->slots[0] < i)
3812                                 break;
3813                         if (path->slots[0] == i) {
3814                                 int space = btrfs_leaf_free_space(right);
3815
3816                                 if (space + push_space * 2 > free_space)
3817                                         break;
3818                         }
3819                 }
3820
3821                 if (path->slots[0] == i)
3822                         push_space += data_size;
3823
3824                 this_item_size = btrfs_item_size(right, item);
3825                 if (this_item_size + sizeof(*item) + push_space > free_space)
3826                         break;
3827
3828                 push_items++;
3829                 push_space += this_item_size + sizeof(*item);
3830         }
3831
3832         if (push_items == 0) {
3833                 ret = 1;
3834                 goto out;
3835         }
3836         WARN_ON(!empty && push_items == btrfs_header_nritems(right));
3837
3838         /* push data from right to left */
3839         copy_extent_buffer(left, right,
3840                            btrfs_item_nr_offset(btrfs_header_nritems(left)),
3841                            btrfs_item_nr_offset(0),
3842                            push_items * sizeof(struct btrfs_item));
3843
3844         push_space = BTRFS_LEAF_DATA_SIZE(fs_info) -
3845                      btrfs_item_offset_nr(right, push_items - 1);
3846
3847         copy_extent_buffer(left, right, BTRFS_LEAF_DATA_OFFSET +
3848                      leaf_data_end(left) - push_space,
3849                      BTRFS_LEAF_DATA_OFFSET +
3850                      btrfs_item_offset_nr(right, push_items - 1),
3851                      push_space);
3852         old_left_nritems = btrfs_header_nritems(left);
3853         BUG_ON(old_left_nritems <= 0);
3854
3855         btrfs_init_map_token(&token, left);
3856         old_left_item_size = btrfs_item_offset_nr(left, old_left_nritems - 1);
3857         for (i = old_left_nritems; i < old_left_nritems + push_items; i++) {
3858                 u32 ioff;
3859
3860                 item = btrfs_item_nr(i);
3861
3862                 ioff = btrfs_token_item_offset(left, item, &token);
3863                 btrfs_set_token_item_offset(left, item,
3864                       ioff - (BTRFS_LEAF_DATA_SIZE(fs_info) - old_left_item_size),
3865                       &token);
3866         }
3867         btrfs_set_header_nritems(left, old_left_nritems + push_items);
3868
3869         /* fixup right node */
3870         if (push_items > right_nritems)
3871                 WARN(1, KERN_CRIT "push items %d nr %u\n", push_items,
3872                        right_nritems);
3873
3874         if (push_items < right_nritems) {
3875                 push_space = btrfs_item_offset_nr(right, push_items - 1) -
3876                                                   leaf_data_end(right);
3877                 memmove_extent_buffer(right, BTRFS_LEAF_DATA_OFFSET +
3878                                       BTRFS_LEAF_DATA_SIZE(fs_info) - push_space,
3879                                       BTRFS_LEAF_DATA_OFFSET +
3880                                       leaf_data_end(right), push_space);
3881
3882                 memmove_extent_buffer(right, btrfs_item_nr_offset(0),
3883                               btrfs_item_nr_offset(push_items),
3884                              (btrfs_header_nritems(right) - push_items) *
3885                              sizeof(struct btrfs_item));
3886         }
3887
3888         btrfs_init_map_token(&token, right);
3889         right_nritems -= push_items;
3890         btrfs_set_header_nritems(right, right_nritems);
3891         push_space = BTRFS_LEAF_DATA_SIZE(fs_info);
3892         for (i = 0; i < right_nritems; i++) {
3893                 item = btrfs_item_nr(i);
3894
3895                 push_space = push_space - btrfs_token_item_size(right,
3896                                                                 item, &token);
3897                 btrfs_set_token_item_offset(right, item, push_space, &token);
3898         }
3899
3900         btrfs_mark_buffer_dirty(left);
3901         if (right_nritems)
3902                 btrfs_mark_buffer_dirty(right);
3903         else
3904                 btrfs_clean_tree_block(right);
3905
3906         btrfs_item_key(right, &disk_key, 0);
3907         fixup_low_keys(path, &disk_key, 1);
3908
3909         /* then fixup the leaf pointer in the path */
3910         if (path->slots[0] < push_items) {
3911                 path->slots[0] += old_left_nritems;
3912                 btrfs_tree_unlock(path->nodes[0]);
3913                 free_extent_buffer(path->nodes[0]);
3914                 path->nodes[0] = left;
3915                 path->slots[1] -= 1;
3916         } else {
3917                 btrfs_tree_unlock(left);
3918                 free_extent_buffer(left);
3919                 path->slots[0] -= push_items;
3920         }
3921         BUG_ON(path->slots[0] < 0);
3922         return ret;
3923 out:
3924         btrfs_tree_unlock(left);
3925         free_extent_buffer(left);
3926         return ret;
3927 }
3928
3929 /*
3930  * push some data in the path leaf to the left, trying to free up at
3931  * least data_size bytes.  returns zero if the push worked, nonzero otherwise
3932  *
3933  * max_slot can put a limit on how far into the leaf we'll push items.  The
3934  * item at 'max_slot' won't be touched.  Use (u32)-1 to make us push all the
3935  * items
3936  */
3937 static int push_leaf_left(struct btrfs_trans_handle *trans, struct btrfs_root
3938                           *root, struct btrfs_path *path, int min_data_size,
3939                           int data_size, int empty, u32 max_slot)
3940 {
3941         struct extent_buffer *right = path->nodes[0];
3942         struct extent_buffer *left;
3943         int slot;
3944         int free_space;
3945         u32 right_nritems;
3946         int ret = 0;
3947
3948         slot = path->slots[1];
3949         if (slot == 0)
3950                 return 1;
3951         if (!path->nodes[1])
3952                 return 1;
3953
3954         right_nritems = btrfs_header_nritems(right);
3955         if (right_nritems == 0)
3956                 return 1;
3957
3958         btrfs_assert_tree_locked(path->nodes[1]);
3959
3960         left = btrfs_read_node_slot(path->nodes[1], slot - 1);
3961         /*
3962          * slot - 1 is not valid or we fail to read the left node,
3963          * no big deal, just return.
3964          */
3965         if (IS_ERR(left))
3966                 return 1;
3967
3968         btrfs_tree_lock(left);
3969         btrfs_set_lock_blocking_write(left);
3970
3971         free_space = btrfs_leaf_free_space(left);
3972         if (free_space < data_size) {
3973                 ret = 1;
3974                 goto out;
3975         }
3976
3977         /* cow and double check */
3978         ret = btrfs_cow_block(trans, root, left,
3979                               path->nodes[1], slot - 1, &left);
3980         if (ret) {
3981                 /* we hit -ENOSPC, but it isn't fatal here */
3982                 if (ret == -ENOSPC)
3983                         ret = 1;
3984                 goto out;
3985         }
3986
3987         free_space = btrfs_leaf_free_space(left);
3988         if (free_space < data_size) {
3989                 ret = 1;
3990                 goto out;
3991         }
3992
3993         return __push_leaf_left(path, min_data_size,
3994                                empty, left, free_space, right_nritems,
3995                                max_slot);
3996 out:
3997         btrfs_tree_unlock(left);
3998         free_extent_buffer(left);
3999         return ret;
4000 }
4001
4002 /*
4003  * split the path's leaf in two, making sure there is at least data_size
4004  * available for the resulting leaf level of the path.
4005  */
4006 static noinline void copy_for_split(struct btrfs_trans_handle *trans,
4007                                     struct btrfs_path *path,
4008                                     struct extent_buffer *l,
4009                                     struct extent_buffer *right,
4010                                     int slot, int mid, int nritems)
4011 {
4012         struct btrfs_fs_info *fs_info = trans->fs_info;
4013         int data_copy_size;
4014         int rt_data_off;
4015         int i;
4016         struct btrfs_disk_key disk_key;
4017         struct btrfs_map_token token;
4018
4019         nritems = nritems - mid;
4020         btrfs_set_header_nritems(right, nritems);
4021         data_copy_size = btrfs_item_end_nr(l, mid) - leaf_data_end(l);
4022
4023         copy_extent_buffer(right, l, btrfs_item_nr_offset(0),
4024                            btrfs_item_nr_offset(mid),
4025                            nritems * sizeof(struct btrfs_item));
4026
4027         copy_extent_buffer(right, l,
4028                      BTRFS_LEAF_DATA_OFFSET + BTRFS_LEAF_DATA_SIZE(fs_info) -
4029                      data_copy_size, BTRFS_LEAF_DATA_OFFSET +
4030                      leaf_data_end(l), data_copy_size);
4031
4032         rt_data_off = BTRFS_LEAF_DATA_SIZE(fs_info) - btrfs_item_end_nr(l, mid);
4033
4034         btrfs_init_map_token(&token, right);
4035         for (i = 0; i < nritems; i++) {
4036                 struct btrfs_item *item = btrfs_item_nr(i);
4037                 u32 ioff;
4038
4039                 ioff = btrfs_token_item_offset(right, item, &token);
4040                 btrfs_set_token_item_offset(right, item,
4041                                             ioff + rt_data_off, &token);
4042         }
4043
4044         btrfs_set_header_nritems(l, mid);
4045         btrfs_item_key(right, &disk_key, 0);
4046         insert_ptr(trans, path, &disk_key, right->start, path->slots[1] + 1, 1);
4047
4048         btrfs_mark_buffer_dirty(right);
4049         btrfs_mark_buffer_dirty(l);
4050         BUG_ON(path->slots[0] != slot);
4051
4052         if (mid <= slot) {
4053                 btrfs_tree_unlock(path->nodes[0]);
4054                 free_extent_buffer(path->nodes[0]);
4055                 path->nodes[0] = right;
4056                 path->slots[0] -= mid;
4057                 path->slots[1] += 1;
4058         } else {
4059                 btrfs_tree_unlock(right);
4060                 free_extent_buffer(right);
4061         }
4062
4063         BUG_ON(path->slots[0] < 0);
4064 }
4065
4066 /*
4067  * double splits happen when we need to insert a big item in the middle
4068  * of a leaf.  A double split can leave us with 3 mostly empty leaves:
4069  * leaf: [ slots 0 - N] [ our target ] [ N + 1 - total in leaf ]
4070  *          A                 B                 C
4071  *
4072  * We avoid this by trying to push the items on either side of our target
4073  * into the adjacent leaves.  If all goes well we can avoid the double split
4074  * completely.
4075  */
4076 static noinline int push_for_double_split(struct btrfs_trans_handle *trans,
4077                                           struct btrfs_root *root,
4078                                           struct btrfs_path *path,
4079                                           int data_size)
4080 {
4081         int ret;
4082         int progress = 0;
4083         int slot;
4084         u32 nritems;
4085         int space_needed = data_size;
4086
4087         slot = path->slots[0];
4088         if (slot < btrfs_header_nritems(path->nodes[0]))
4089                 space_needed -= btrfs_leaf_free_space(path->nodes[0]);
4090
4091         /*
4092          * try to push all the items after our slot into the
4093          * right leaf
4094          */
4095         ret = push_leaf_right(trans, root, path, 1, space_needed, 0, slot);
4096         if (ret < 0)
4097                 return ret;
4098
4099         if (ret == 0)
4100                 progress++;
4101
4102         nritems = btrfs_header_nritems(path->nodes[0]);
4103         /*
4104          * our goal is to get our slot at the start or end of a leaf.  If
4105          * we've done so we're done
4106          */
4107         if (path->slots[0] == 0 || path->slots[0] == nritems)
4108                 return 0;
4109
4110         if (btrfs_leaf_free_space(path->nodes[0]) >= data_size)
4111                 return 0;
4112
4113         /* try to push all the items before our slot into the next leaf */
4114         slot = path->slots[0];
4115         space_needed = data_size;
4116         if (slot > 0)
4117                 space_needed -= btrfs_leaf_free_space(path->nodes[0]);
4118         ret = push_leaf_left(trans, root, path, 1, space_needed, 0, slot);
4119         if (ret < 0)
4120                 return ret;
4121
4122         if (ret == 0)
4123                 progress++;
4124
4125         if (progress)
4126                 return 0;
4127         return 1;
4128 }
4129
4130 /*
4131  * split the path's leaf in two, making sure there is at least data_size
4132  * available for the resulting leaf level of the path.
4133  *
4134  * returns 0 if all went well and < 0 on failure.
4135  */
4136 static noinline int split_leaf(struct btrfs_trans_handle *trans,
4137                                struct btrfs_root *root,
4138                                const struct btrfs_key *ins_key,
4139                                struct btrfs_path *path, int data_size,
4140                                int extend)
4141 {
4142         struct btrfs_disk_key disk_key;
4143         struct extent_buffer *l;
4144         u32 nritems;
4145         int mid;
4146         int slot;
4147         struct extent_buffer *right;
4148         struct btrfs_fs_info *fs_info = root->fs_info;
4149         int ret = 0;
4150         int wret;
4151         int split;
4152         int num_doubles = 0;
4153         int tried_avoid_double = 0;
4154
4155         l = path->nodes[0];
4156         slot = path->slots[0];
4157         if (extend && data_size + btrfs_item_size_nr(l, slot) +
4158             sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(fs_info))
4159                 return -EOVERFLOW;
4160
4161         /* first try to make some room by pushing left and right */
4162         if (data_size && path->nodes[1]) {
4163                 int space_needed = data_size;
4164
4165                 if (slot < btrfs_header_nritems(l))
4166                         space_needed -= btrfs_leaf_free_space(l);
4167
4168                 wret = push_leaf_right(trans, root, path, space_needed,
4169                                        space_needed, 0, 0);
4170                 if (wret < 0)
4171                         return wret;
4172                 if (wret) {
4173                         space_needed = data_size;
4174                         if (slot > 0)
4175                                 space_needed -= btrfs_leaf_free_space(l);
4176                         wret = push_leaf_left(trans, root, path, space_needed,
4177                                               space_needed, 0, (u32)-1);
4178                         if (wret < 0)
4179                                 return wret;
4180                 }
4181                 l = path->nodes[0];
4182
4183                 /* did the pushes work? */
4184                 if (btrfs_leaf_free_space(l) >= data_size)
4185                         return 0;
4186         }
4187
4188         if (!path->nodes[1]) {
4189                 ret = insert_new_root(trans, root, path, 1);
4190                 if (ret)
4191                         return ret;
4192         }
4193 again:
4194         split = 1;
4195         l = path->nodes[0];
4196         slot = path->slots[0];
4197         nritems = btrfs_header_nritems(l);
4198         mid = (nritems + 1) / 2;
4199
4200         if (mid <= slot) {
4201                 if (nritems == 1 ||
4202                     leaf_space_used(l, mid, nritems - mid) + data_size >
4203                         BTRFS_LEAF_DATA_SIZE(fs_info)) {
4204                         if (slot >= nritems) {
4205                                 split = 0;
4206                         } else {
4207                                 mid = slot;
4208                                 if (mid != nritems &&
4209                                     leaf_space_used(l, mid, nritems - mid) +
4210                                     data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) {
4211                                         if (data_size && !tried_avoid_double)
4212                                                 goto push_for_double;
4213                                         split = 2;
4214                                 }
4215                         }
4216                 }
4217         } else {
4218                 if (leaf_space_used(l, 0, mid) + data_size >
4219                         BTRFS_LEAF_DATA_SIZE(fs_info)) {
4220                         if (!extend && data_size && slot == 0) {
4221                                 split = 0;
4222                         } else if ((extend || !data_size) && slot == 0) {
4223                                 mid = 1;
4224                         } else {
4225                                 mid = slot;
4226                                 if (mid != nritems &&
4227                                     leaf_space_used(l, mid, nritems - mid) +
4228                                     data_size > BTRFS_LEAF_DATA_SIZE(fs_info)) {
4229                                         if (data_size && !tried_avoid_double)
4230                                                 goto push_for_double;
4231                                         split = 2;
4232                                 }
4233                         }
4234                 }
4235         }
4236
4237         if (split == 0)
4238                 btrfs_cpu_key_to_disk(&disk_key, ins_key);
4239         else
4240                 btrfs_item_key(l, &disk_key, mid);
4241
4242         right = alloc_tree_block_no_bg_flush(trans, root, 0, &disk_key, 0,
4243                                              l->start, 0);
4244         if (IS_ERR(right))
4245                 return PTR_ERR(right);
4246
4247         root_add_used(root, fs_info->nodesize);
4248
4249         if (split == 0) {
4250                 if (mid <= slot) {
4251                         btrfs_set_header_nritems(right, 0);
4252                         insert_ptr(trans, path, &disk_key,
4253                                    right->start, path->slots[1] + 1, 1);
4254                         btrfs_tree_unlock(path->nodes[0]);
4255                         free_extent_buffer(path->nodes[0]);
4256                         path->nodes[0] = right;
4257                         path->slots[0] = 0;
4258                         path->slots[1] += 1;
4259                 } else {
4260                         btrfs_set_header_nritems(right, 0);
4261                         insert_ptr(trans, path, &disk_key,
4262                                    right->start, path->slots[1], 1);
4263                         btrfs_tree_unlock(path->nodes[0]);
4264                         free_extent_buffer(path->nodes[0]);
4265                         path->nodes[0] = right;
4266                         path->slots[0] = 0;
4267                         if (path->slots[1] == 0)
4268                                 fixup_low_keys(path, &disk_key, 1);
4269                 }
4270                 /*
4271                  * We create a new leaf 'right' for the required ins_len and
4272                  * we'll do btrfs_mark_buffer_dirty() on this leaf after copying
4273                  * the content of ins_len to 'right'.
4274                  */
4275                 return ret;
4276         }
4277
4278         copy_for_split(trans, path, l, right, slot, mid, nritems);
4279
4280         if (split == 2) {
4281                 BUG_ON(num_doubles != 0);
4282                 num_doubles++;
4283                 goto again;
4284         }
4285
4286         return 0;
4287
4288 push_for_double:
4289         push_for_double_split(trans, root, path, data_size);
4290         tried_avoid_double = 1;
4291         if (btrfs_leaf_free_space(path->nodes[0]) >= data_size)
4292                 return 0;
4293         goto again;
4294 }
4295
4296 static noinline int setup_leaf_for_split(struct btrfs_trans_handle *trans,
4297                                          struct btrfs_root *root,
4298                                          struct btrfs_path *path, int ins_len)
4299 {
4300         struct btrfs_key key;
4301         struct extent_buffer *leaf;
4302         struct btrfs_file_extent_item *fi;
4303         u64 extent_len = 0;
4304         u32 item_size;
4305         int ret;
4306
4307         leaf = path->nodes[0];
4308         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4309
4310         BUG_ON(key.type != BTRFS_EXTENT_DATA_KEY &&
4311                key.type != BTRFS_EXTENT_CSUM_KEY);
4312
4313         if (btrfs_leaf_free_space(leaf) >= ins_len)
4314                 return 0;
4315
4316         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
4317         if (key.type == BTRFS_EXTENT_DATA_KEY) {
4318                 fi = btrfs_item_ptr(leaf, path->slots[0],
4319                                     struct btrfs_file_extent_item);
4320                 extent_len = btrfs_file_extent_num_bytes(leaf, fi);
4321         }
4322         btrfs_release_path(path);
4323
4324         path->keep_locks = 1;
4325         path->search_for_split = 1;
4326         ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
4327         path->search_for_split = 0;
4328         if (ret > 0)
4329                 ret = -EAGAIN;
4330         if (ret < 0)
4331                 goto err;
4332
4333         ret = -EAGAIN;
4334         leaf = path->nodes[0];
4335         /* if our item isn't there, return now */
4336         if (item_size != btrfs_item_size_nr(leaf, path->slots[0]))
4337                 goto err;
4338
4339         /* the leaf has  changed, it now has room.  return now */
4340         if (btrfs_leaf_free_space(path->nodes[0]) >= ins_len)
4341                 goto err;
4342
4343         if (key.type == BTRFS_EXTENT_DATA_KEY) {
4344                 fi = btrfs_item_ptr(leaf, path->slots[0],
4345                                     struct btrfs_file_extent_item);
4346                 if (extent_len != btrfs_file_extent_num_bytes(leaf, fi))
4347                         goto err;
4348         }
4349
4350         btrfs_set_path_blocking(path);
4351         ret = split_leaf(trans, root, &key, path, ins_len, 1);
4352         if (ret)
4353                 goto err;
4354
4355         path->keep_locks = 0;
4356         btrfs_unlock_up_safe(path, 1);
4357         return 0;
4358 err:
4359         path->keep_locks = 0;
4360         return ret;
4361 }
4362
4363 static noinline int split_item(struct btrfs_path *path,
4364                                const struct btrfs_key *new_key,
4365                                unsigned long split_offset)
4366 {
4367         struct extent_buffer *leaf;
4368         struct btrfs_item *item;
4369         struct btrfs_item *new_item;
4370         int slot;
4371         char *buf;
4372         u32 nritems;
4373         u32 item_size;
4374         u32 orig_offset;
4375         struct btrfs_disk_key disk_key;
4376
4377         leaf = path->nodes[0];
4378         BUG_ON(btrfs_leaf_free_space(leaf) < sizeof(struct btrfs_item));
4379
4380         btrfs_set_path_blocking(path);
4381
4382         item = btrfs_item_nr(path->slots[0]);
4383         orig_offset = btrfs_item_offset(leaf, item);
4384         item_size = btrfs_item_size(leaf, item);
4385
4386         buf = kmalloc(item_size, GFP_NOFS);
4387         if (!buf)
4388                 return -ENOMEM;
4389
4390         read_extent_buffer(leaf, buf, btrfs_item_ptr_offset(leaf,
4391                             path->slots[0]), item_size);
4392
4393         slot = path->slots[0] + 1;
4394         nritems = btrfs_header_nritems(leaf);
4395         if (slot != nritems) {
4396                 /* shift the items */
4397                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + 1),
4398                                 btrfs_item_nr_offset(slot),
4399                                 (nritems - slot) * sizeof(struct btrfs_item));
4400         }
4401
4402         btrfs_cpu_key_to_disk(&disk_key, new_key);
4403         btrfs_set_item_key(leaf, &disk_key, slot);
4404
4405         new_item = btrfs_item_nr(slot);
4406
4407         btrfs_set_item_offset(leaf, new_item, orig_offset);
4408         btrfs_set_item_size(leaf, new_item, item_size - split_offset);
4409
4410         btrfs_set_item_offset(leaf, item,
4411                               orig_offset + item_size - split_offset);
4412         btrfs_set_item_size(leaf, item, split_offset);
4413
4414         btrfs_set_header_nritems(leaf, nritems + 1);
4415
4416         /* write the data for the start of the original item */
4417         write_extent_buffer(leaf, buf,
4418                             btrfs_item_ptr_offset(leaf, path->slots[0]),
4419                             split_offset);
4420
4421         /* write the data for the new item */
4422         write_extent_buffer(leaf, buf + split_offset,
4423                             btrfs_item_ptr_offset(leaf, slot),
4424                             item_size - split_offset);
4425         btrfs_mark_buffer_dirty(leaf);
4426
4427         BUG_ON(btrfs_leaf_free_space(leaf) < 0);
4428         kfree(buf);
4429         return 0;
4430 }
4431
4432 /*
4433  * This function splits a single item into two items,
4434  * giving 'new_key' to the new item and splitting the
4435  * old one at split_offset (from the start of the item).
4436  *
4437  * The path may be released by this operation.  After
4438  * the split, the path is pointing to the old item.  The
4439  * new item is going to be in the same node as the old one.
4440  *
4441  * Note, the item being split must be smaller enough to live alone on
4442  * a tree block with room for one extra struct btrfs_item
4443  *
4444  * This allows us to split the item in place, keeping a lock on the
4445  * leaf the entire time.
4446  */
4447 int btrfs_split_item(struct btrfs_trans_handle *trans,
4448                      struct btrfs_root *root,
4449                      struct btrfs_path *path,
4450                      const struct btrfs_key *new_key,
4451                      unsigned long split_offset)
4452 {
4453         int ret;
4454         ret = setup_leaf_for_split(trans, root, path,
4455                                    sizeof(struct btrfs_item));
4456         if (ret)
4457                 return ret;
4458
4459         ret = split_item(path, new_key, split_offset);
4460         return ret;
4461 }
4462
4463 /*
4464  * This function duplicate a item, giving 'new_key' to the new item.
4465  * It guarantees both items live in the same tree leaf and the new item
4466  * is contiguous with the original item.
4467  *
4468  * This allows us to split file extent in place, keeping a lock on the
4469  * leaf the entire time.
4470  */
4471 int btrfs_duplicate_item(struct btrfs_trans_handle *trans,
4472                          struct btrfs_root *root,
4473                          struct btrfs_path *path,
4474                          const struct btrfs_key *new_key)
4475 {
4476         struct extent_buffer *leaf;
4477         int ret;
4478         u32 item_size;
4479
4480         leaf = path->nodes[0];
4481         item_size = btrfs_item_size_nr(leaf, path->slots[0]);
4482         ret = setup_leaf_for_split(trans, root, path,
4483                                    item_size + sizeof(struct btrfs_item));
4484         if (ret)
4485                 return ret;
4486
4487         path->slots[0]++;
4488         setup_items_for_insert(root, path, new_key, &item_size,
4489                                item_size, item_size +
4490                                sizeof(struct btrfs_item), 1);
4491         leaf = path->nodes[0];
4492         memcpy_extent_buffer(leaf,
4493                              btrfs_item_ptr_offset(leaf, path->slots[0]),
4494                              btrfs_item_ptr_offset(leaf, path->slots[0] - 1),
4495                              item_size);
4496         return 0;
4497 }
4498
4499 /*
4500  * make the item pointed to by the path smaller.  new_size indicates
4501  * how small to make it, and from_end tells us if we just chop bytes
4502  * off the end of the item or if we shift the item to chop bytes off
4503  * the front.
4504  */
4505 void btrfs_truncate_item(struct btrfs_path *path, u32 new_size, int from_end)
4506 {
4507         int slot;
4508         struct extent_buffer *leaf;
4509         struct btrfs_item *item;
4510         u32 nritems;
4511         unsigned int data_end;
4512         unsigned int old_data_start;
4513         unsigned int old_size;
4514         unsigned int size_diff;
4515         int i;
4516         struct btrfs_map_token token;
4517
4518         leaf = path->nodes[0];
4519         slot = path->slots[0];
4520
4521         old_size = btrfs_item_size_nr(leaf, slot);
4522         if (old_size == new_size)
4523                 return;
4524
4525         nritems = btrfs_header_nritems(leaf);
4526         data_end = leaf_data_end(leaf);
4527
4528         old_data_start = btrfs_item_offset_nr(leaf, slot);
4529
4530         size_diff = old_size - new_size;
4531
4532         BUG_ON(slot < 0);
4533         BUG_ON(slot >= nritems);
4534
4535         /*
4536          * item0..itemN ... dataN.offset..dataN.size .. data0.size
4537          */
4538         /* first correct the data pointers */
4539         btrfs_init_map_token(&token, leaf);
4540         for (i = slot; i < nritems; i++) {
4541                 u32 ioff;
4542                 item = btrfs_item_nr(i);
4543
4544                 ioff = btrfs_token_item_offset(leaf, item, &token);
4545                 btrfs_set_token_item_offset(leaf, item,
4546                                             ioff + size_diff, &token);
4547         }
4548
4549         /* shift the data */
4550         if (from_end) {
4551                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4552                               data_end + size_diff, BTRFS_LEAF_DATA_OFFSET +
4553                               data_end, old_data_start + new_size - data_end);
4554         } else {
4555                 struct btrfs_disk_key disk_key;
4556                 u64 offset;
4557
4558                 btrfs_item_key(leaf, &disk_key, slot);
4559
4560                 if (btrfs_disk_key_type(&disk_key) == BTRFS_EXTENT_DATA_KEY) {
4561                         unsigned long ptr;
4562                         struct btrfs_file_extent_item *fi;
4563
4564                         fi = btrfs_item_ptr(leaf, slot,
4565                                             struct btrfs_file_extent_item);
4566                         fi = (struct btrfs_file_extent_item *)(
4567                              (unsigned long)fi - size_diff);
4568
4569                         if (btrfs_file_extent_type(leaf, fi) ==
4570                             BTRFS_FILE_EXTENT_INLINE) {
4571                                 ptr = btrfs_item_ptr_offset(leaf, slot);
4572                                 memmove_extent_buffer(leaf, ptr,
4573                                       (unsigned long)fi,
4574                                       BTRFS_FILE_EXTENT_INLINE_DATA_START);
4575                         }
4576                 }
4577
4578                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4579                               data_end + size_diff, BTRFS_LEAF_DATA_OFFSET +
4580                               data_end, old_data_start - data_end);
4581
4582                 offset = btrfs_disk_key_offset(&disk_key);
4583                 btrfs_set_disk_key_offset(&disk_key, offset + size_diff);
4584                 btrfs_set_item_key(leaf, &disk_key, slot);
4585                 if (slot == 0)
4586                         fixup_low_keys(path, &disk_key, 1);
4587         }
4588
4589         item = btrfs_item_nr(slot);
4590         btrfs_set_item_size(leaf, item, new_size);
4591         btrfs_mark_buffer_dirty(leaf);
4592
4593         if (btrfs_leaf_free_space(leaf) < 0) {
4594                 btrfs_print_leaf(leaf);
4595                 BUG();
4596         }
4597 }
4598
4599 /*
4600  * make the item pointed to by the path bigger, data_size is the added size.
4601  */
4602 void btrfs_extend_item(struct btrfs_path *path, u32 data_size)
4603 {
4604         int slot;
4605         struct extent_buffer *leaf;
4606         struct btrfs_item *item;
4607         u32 nritems;
4608         unsigned int data_end;
4609         unsigned int old_data;
4610         unsigned int old_size;
4611         int i;
4612         struct btrfs_map_token token;
4613
4614         leaf = path->nodes[0];
4615
4616         nritems = btrfs_header_nritems(leaf);
4617         data_end = leaf_data_end(leaf);
4618
4619         if (btrfs_leaf_free_space(leaf) < data_size) {
4620                 btrfs_print_leaf(leaf);
4621                 BUG();
4622         }
4623         slot = path->slots[0];
4624         old_data = btrfs_item_end_nr(leaf, slot);
4625
4626         BUG_ON(slot < 0);
4627         if (slot >= nritems) {
4628                 btrfs_print_leaf(leaf);
4629                 btrfs_crit(leaf->fs_info, "slot %d too large, nritems %d",
4630                            slot, nritems);
4631                 BUG();
4632         }
4633
4634         /*
4635          * item0..itemN ... dataN.offset..dataN.size .. data0.size
4636          */
4637         /* first correct the data pointers */
4638         btrfs_init_map_token(&token, leaf);
4639         for (i = slot; i < nritems; i++) {
4640                 u32 ioff;
4641                 item = btrfs_item_nr(i);
4642
4643                 ioff = btrfs_token_item_offset(leaf, item, &token);
4644                 btrfs_set_token_item_offset(leaf, item,
4645                                             ioff - data_size, &token);
4646         }
4647
4648         /* shift the data */
4649         memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4650                       data_end - data_size, BTRFS_LEAF_DATA_OFFSET +
4651                       data_end, old_data - data_end);
4652
4653         data_end = old_data;
4654         old_size = btrfs_item_size_nr(leaf, slot);
4655         item = btrfs_item_nr(slot);
4656         btrfs_set_item_size(leaf, item, old_size + data_size);
4657         btrfs_mark_buffer_dirty(leaf);
4658
4659         if (btrfs_leaf_free_space(leaf) < 0) {
4660                 btrfs_print_leaf(leaf);
4661                 BUG();
4662         }
4663 }
4664
4665 /*
4666  * this is a helper for btrfs_insert_empty_items, the main goal here is
4667  * to save stack depth by doing the bulk of the work in a function
4668  * that doesn't call btrfs_search_slot
4669  */
4670 void setup_items_for_insert(struct btrfs_root *root, struct btrfs_path *path,
4671                             const struct btrfs_key *cpu_key, u32 *data_size,
4672                             u32 total_data, u32 total_size, int nr)
4673 {
4674         struct btrfs_fs_info *fs_info = root->fs_info;
4675         struct btrfs_item *item;
4676         int i;
4677         u32 nritems;
4678         unsigned int data_end;
4679         struct btrfs_disk_key disk_key;
4680         struct extent_buffer *leaf;
4681         int slot;
4682         struct btrfs_map_token token;
4683
4684         if (path->slots[0] == 0) {
4685                 btrfs_cpu_key_to_disk(&disk_key, cpu_key);
4686                 fixup_low_keys(path, &disk_key, 1);
4687         }
4688         btrfs_unlock_up_safe(path, 1);
4689
4690         leaf = path->nodes[0];
4691         slot = path->slots[0];
4692
4693         nritems = btrfs_header_nritems(leaf);
4694         data_end = leaf_data_end(leaf);
4695
4696         if (btrfs_leaf_free_space(leaf) < total_size) {
4697                 btrfs_print_leaf(leaf);
4698                 btrfs_crit(fs_info, "not enough freespace need %u have %d",
4699                            total_size, btrfs_leaf_free_space(leaf));
4700                 BUG();
4701         }
4702
4703         btrfs_init_map_token(&token, leaf);
4704         if (slot != nritems) {
4705                 unsigned int old_data = btrfs_item_end_nr(leaf, slot);
4706
4707                 if (old_data < data_end) {
4708                         btrfs_print_leaf(leaf);
4709                         btrfs_crit(fs_info, "slot %d old_data %d data_end %d",
4710                                    slot, old_data, data_end);
4711                         BUG();
4712                 }
4713                 /*
4714                  * item0..itemN ... dataN.offset..dataN.size .. data0.size
4715                  */
4716                 /* first correct the data pointers */
4717                 for (i = slot; i < nritems; i++) {
4718                         u32 ioff;
4719
4720                         item = btrfs_item_nr(i);
4721                         ioff = btrfs_token_item_offset(leaf, item, &token);
4722                         btrfs_set_token_item_offset(leaf, item,
4723                                                     ioff - total_data, &token);
4724                 }
4725                 /* shift the items */
4726                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot + nr),
4727                               btrfs_item_nr_offset(slot),
4728                               (nritems - slot) * sizeof(struct btrfs_item));
4729
4730                 /* shift the data */
4731                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4732                               data_end - total_data, BTRFS_LEAF_DATA_OFFSET +
4733                               data_end, old_data - data_end);
4734                 data_end = old_data;
4735         }
4736
4737         /* setup the item for the new data */
4738         for (i = 0; i < nr; i++) {
4739                 btrfs_cpu_key_to_disk(&disk_key, cpu_key + i);
4740                 btrfs_set_item_key(leaf, &disk_key, slot + i);
4741                 item = btrfs_item_nr(slot + i);
4742                 btrfs_set_token_item_offset(leaf, item,
4743                                             data_end - data_size[i], &token);
4744                 data_end -= data_size[i];
4745                 btrfs_set_token_item_size(leaf, item, data_size[i], &token);
4746         }
4747
4748         btrfs_set_header_nritems(leaf, nritems + nr);
4749         btrfs_mark_buffer_dirty(leaf);
4750
4751         if (btrfs_leaf_free_space(leaf) < 0) {
4752                 btrfs_print_leaf(leaf);
4753                 BUG();
4754         }
4755 }
4756
4757 /*
4758  * Given a key and some data, insert items into the tree.
4759  * This does all the path init required, making room in the tree if needed.
4760  */
4761 int btrfs_insert_empty_items(struct btrfs_trans_handle *trans,
4762                             struct btrfs_root *root,
4763                             struct btrfs_path *path,
4764                             const struct btrfs_key *cpu_key, u32 *data_size,
4765                             int nr)
4766 {
4767         int ret = 0;
4768         int slot;
4769         int i;
4770         u32 total_size = 0;
4771         u32 total_data = 0;
4772
4773         for (i = 0; i < nr; i++)
4774                 total_data += data_size[i];
4775
4776         total_size = total_data + (nr * sizeof(struct btrfs_item));
4777         ret = btrfs_search_slot(trans, root, cpu_key, path, total_size, 1);
4778         if (ret == 0)
4779                 return -EEXIST;
4780         if (ret < 0)
4781                 return ret;
4782
4783         slot = path->slots[0];
4784         BUG_ON(slot < 0);
4785
4786         setup_items_for_insert(root, path, cpu_key, data_size,
4787                                total_data, total_size, nr);
4788         return 0;
4789 }
4790
4791 /*
4792  * Given a key and some data, insert an item into the tree.
4793  * This does all the path init required, making room in the tree if needed.
4794  */
4795 int btrfs_insert_item(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4796                       const struct btrfs_key *cpu_key, void *data,
4797                       u32 data_size)
4798 {
4799         int ret = 0;
4800         struct btrfs_path *path;
4801         struct extent_buffer *leaf;
4802         unsigned long ptr;
4803
4804         path = btrfs_alloc_path();
4805         if (!path)
4806                 return -ENOMEM;
4807         ret = btrfs_insert_empty_item(trans, root, path, cpu_key, data_size);
4808         if (!ret) {
4809                 leaf = path->nodes[0];
4810                 ptr = btrfs_item_ptr_offset(leaf, path->slots[0]);
4811                 write_extent_buffer(leaf, data, ptr, data_size);
4812                 btrfs_mark_buffer_dirty(leaf);
4813         }
4814         btrfs_free_path(path);
4815         return ret;
4816 }
4817
4818 /*
4819  * delete the pointer from a given node.
4820  *
4821  * the tree should have been previously balanced so the deletion does not
4822  * empty a node.
4823  */
4824 static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
4825                     int level, int slot)
4826 {
4827         struct extent_buffer *parent = path->nodes[level];
4828         u32 nritems;
4829         int ret;
4830
4831         nritems = btrfs_header_nritems(parent);
4832         if (slot != nritems - 1) {
4833                 if (level) {
4834                         ret = tree_mod_log_insert_move(parent, slot, slot + 1,
4835                                         nritems - slot - 1);
4836                         BUG_ON(ret < 0);
4837                 }
4838                 memmove_extent_buffer(parent,
4839                               btrfs_node_key_ptr_offset(slot),
4840                               btrfs_node_key_ptr_offset(slot + 1),
4841                               sizeof(struct btrfs_key_ptr) *
4842                               (nritems - slot - 1));
4843         } else if (level) {
4844                 ret = tree_mod_log_insert_key(parent, slot, MOD_LOG_KEY_REMOVE,
4845                                 GFP_NOFS);
4846                 BUG_ON(ret < 0);
4847         }
4848
4849         nritems--;
4850         btrfs_set_header_nritems(parent, nritems);
4851         if (nritems == 0 && parent == root->node) {
4852                 BUG_ON(btrfs_header_level(root->node) != 1);
4853                 /* just turn the root into a leaf and break */
4854                 btrfs_set_header_level(root->node, 0);
4855         } else if (slot == 0) {
4856                 struct btrfs_disk_key disk_key;
4857
4858                 btrfs_node_key(parent, &disk_key, 0);
4859                 fixup_low_keys(path, &disk_key, level + 1);
4860         }
4861         btrfs_mark_buffer_dirty(parent);
4862 }
4863
4864 /*
4865  * a helper function to delete the leaf pointed to by path->slots[1] and
4866  * path->nodes[1].
4867  *
4868  * This deletes the pointer in path->nodes[1] and frees the leaf
4869  * block extent.  zero is returned if it all worked out, < 0 otherwise.
4870  *
4871  * The path must have already been setup for deleting the leaf, including
4872  * all the proper balancing.  path->nodes[1] must be locked.
4873  */
4874 static noinline void btrfs_del_leaf(struct btrfs_trans_handle *trans,
4875                                     struct btrfs_root *root,
4876                                     struct btrfs_path *path,
4877                                     struct extent_buffer *leaf)
4878 {
4879         WARN_ON(btrfs_header_generation(leaf) != trans->transid);
4880         del_ptr(root, path, 1, path->slots[1]);
4881
4882         /*
4883          * btrfs_free_extent is expensive, we want to make sure we
4884          * aren't holding any locks when we call it
4885          */
4886         btrfs_unlock_up_safe(path, 0);
4887
4888         root_sub_used(root, leaf->len);
4889
4890         atomic_inc(&leaf->refs);
4891         btrfs_free_tree_block(trans, root, leaf, 0, 1);
4892         free_extent_buffer_stale(leaf);
4893 }
4894 /*
4895  * delete the item at the leaf level in path.  If that empties
4896  * the leaf, remove it from the tree
4897  */
4898 int btrfs_del_items(struct btrfs_trans_handle *trans, struct btrfs_root *root,
4899                     struct btrfs_path *path, int slot, int nr)
4900 {
4901         struct btrfs_fs_info *fs_info = root->fs_info;
4902         struct extent_buffer *leaf;
4903         struct btrfs_item *item;
4904         u32 last_off;
4905         u32 dsize = 0;
4906         int ret = 0;
4907         int wret;
4908         int i;
4909         u32 nritems;
4910
4911         leaf = path->nodes[0];
4912         last_off = btrfs_item_offset_nr(leaf, slot + nr - 1);
4913
4914         for (i = 0; i < nr; i++)
4915                 dsize += btrfs_item_size_nr(leaf, slot + i);
4916
4917         nritems = btrfs_header_nritems(leaf);
4918
4919         if (slot + nr != nritems) {
4920                 int data_end = leaf_data_end(leaf);
4921                 struct btrfs_map_token token;
4922
4923                 memmove_extent_buffer(leaf, BTRFS_LEAF_DATA_OFFSET +
4924                               data_end + dsize,
4925                               BTRFS_LEAF_DATA_OFFSET + data_end,
4926                               last_off - data_end);
4927
4928                 btrfs_init_map_token(&token, leaf);
4929                 for (i = slot + nr; i < nritems; i++) {
4930                         u32 ioff;
4931
4932                         item = btrfs_item_nr(i);
4933                         ioff = btrfs_token_item_offset(leaf, item, &token);
4934                         btrfs_set_token_item_offset(leaf, item,
4935                                                     ioff + dsize, &token);
4936                 }
4937
4938                 memmove_extent_buffer(leaf, btrfs_item_nr_offset(slot),
4939                               btrfs_item_nr_offset(slot + nr),
4940                               sizeof(struct btrfs_item) *
4941                               (nritems - slot - nr));
4942         }
4943         btrfs_set_header_nritems(leaf, nritems - nr);
4944         nritems -= nr;
4945
4946         /* delete the leaf if we've emptied it */
4947         if (nritems == 0) {
4948                 if (leaf == root->node) {
4949                         btrfs_set_header_level(leaf, 0);
4950                 } else {
4951                         btrfs_set_path_blocking(path);
4952                         btrfs_clean_tree_block(leaf);
4953                         btrfs_del_leaf(trans, root, path, leaf);
4954                 }
4955         } else {
4956                 int used = leaf_space_used(leaf, 0, nritems);
4957                 if (slot == 0) {
4958                         struct btrfs_disk_key disk_key;
4959
4960                         btrfs_item_key(leaf, &disk_key, 0);
4961                         fixup_low_keys(path, &disk_key, 1);
4962                 }
4963
4964                 /* delete the leaf if it is mostly empty */
4965                 if (used < BTRFS_LEAF_DATA_SIZE(fs_info) / 3) {
4966                         /* push_leaf_left fixes the path.
4967                          * make sure the path still points to our leaf
4968                          * for possible call to del_ptr below
4969                          */
4970                         slot = path->slots[1];
4971                         atomic_inc(&leaf->refs);
4972
4973                         btrfs_set_path_blocking(path);
4974                         wret = push_leaf_left(trans, root, path, 1, 1,
4975                                               1, (u32)-1);
4976                         if (wret < 0 && wret != -ENOSPC)
4977                                 ret = wret;
4978
4979                         if (path->nodes[0] == leaf &&
4980                             btrfs_header_nritems(leaf)) {
4981                                 wret = push_leaf_right(trans, root, path, 1,
4982                                                        1, 1, 0);
4983                                 if (wret < 0 && wret != -ENOSPC)
4984                                         ret = wret;
4985                         }
4986
4987                         if (btrfs_header_nritems(leaf) == 0) {
4988                                 path->slots[1] = slot;
4989                                 btrfs_del_leaf(trans, root, path, leaf);
4990                                 free_extent_buffer(leaf);
4991                                 ret = 0;
4992                         } else {
4993                                 /* if we're still in the path, make sure
4994                                  * we're dirty.  Otherwise, one of the
4995                                  * push_leaf functions must have already
4996                                  * dirtied this buffer
4997                                  */
4998                                 if (path->nodes[0] == leaf)
4999                                         btrfs_mark_buffer_dirty(leaf);
5000                                 free_extent_buffer(leaf);
5001                         }
5002                 } else {
5003                         btrfs_mark_buffer_dirty(leaf);
5004                 }
5005         }
5006         return ret;
5007 }
5008
5009 /*
5010  * search the tree again to find a leaf with lesser keys
5011  * returns 0 if it found something or 1 if there are no lesser leaves.
5012  * returns < 0 on io errors.
5013  *
5014  * This may release the path, and so you may lose any locks held at the
5015  * time you call it.
5016  */
5017 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path)
5018 {
5019         struct btrfs_key key;
5020         struct btrfs_disk_key found_key;
5021         int ret;
5022
5023         btrfs_item_key_to_cpu(path->nodes[0], &key, 0);
5024
5025         if (key.offset > 0) {
5026                 key.offset--;
5027         } else if (key.type > 0) {
5028                 key.type--;
5029                 key.offset = (u64)-1;
5030         } else if (key.objectid > 0) {
5031                 key.objectid--;
5032                 key.type = (u8)-1;
5033                 key.offset = (u64)-1;
5034         } else {
5035                 return 1;
5036         }
5037
5038         btrfs_release_path(path);
5039         ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5040         if (ret < 0)
5041                 return ret;
5042         btrfs_item_key(path->nodes[0], &found_key, 0);
5043         ret = comp_keys(&found_key, &key);
5044         /*
5045          * We might have had an item with the previous key in the tree right
5046          * before we released our path. And after we released our path, that
5047          * item might have been pushed to the first slot (0) of the leaf we
5048          * were holding due to a tree balance. Alternatively, an item with the
5049          * previous key can exist as the only element of a leaf (big fat item).
5050          * Therefore account for these 2 cases, so that our callers (like
5051          * btrfs_previous_item) don't miss an existing item with a key matching
5052          * the previous key we computed above.
5053          */
5054         if (ret <= 0)
5055                 return 0;
5056         return 1;
5057 }
5058
5059 /*
5060  * A helper function to walk down the tree starting at min_key, and looking
5061  * for nodes or leaves that are have a minimum transaction id.
5062  * This is used by the btree defrag code, and tree logging
5063  *
5064  * This does not cow, but it does stuff the starting key it finds back
5065  * into min_key, so you can call btrfs_search_slot with cow=1 on the
5066  * key and get a writable path.
5067  *
5068  * This honors path->lowest_level to prevent descent past a given level
5069  * of the tree.
5070  *
5071  * min_trans indicates the oldest transaction that you are interested
5072  * in walking through.  Any nodes or leaves older than min_trans are
5073  * skipped over (without reading them).
5074  *
5075  * returns zero if something useful was found, < 0 on error and 1 if there
5076  * was nothing in the tree that matched the search criteria.
5077  */
5078 int btrfs_search_forward(struct btrfs_root *root, struct btrfs_key *min_key,
5079                          struct btrfs_path *path,
5080                          u64 min_trans)
5081 {
5082         struct extent_buffer *cur;
5083         struct btrfs_key found_key;
5084         int slot;
5085         int sret;
5086         u32 nritems;
5087         int level;
5088         int ret = 1;
5089         int keep_locks = path->keep_locks;
5090
5091         path->keep_locks = 1;
5092 again:
5093         cur = btrfs_read_lock_root_node(root);
5094         level = btrfs_header_level(cur);
5095         WARN_ON(path->nodes[level]);
5096         path->nodes[level] = cur;
5097         path->locks[level] = BTRFS_READ_LOCK;
5098
5099         if (btrfs_header_generation(cur) < min_trans) {
5100                 ret = 1;
5101                 goto out;
5102         }
5103         while (1) {
5104                 nritems = btrfs_header_nritems(cur);
5105                 level = btrfs_header_level(cur);
5106                 sret = btrfs_bin_search(cur, min_key, level, &slot);
5107                 if (sret < 0) {
5108                         ret = sret;
5109                         goto out;
5110                 }
5111
5112                 /* at the lowest level, we're done, setup the path and exit */
5113                 if (level == path->lowest_level) {
5114                         if (slot >= nritems)
5115                                 goto find_next_key;
5116                         ret = 0;
5117                         path->slots[level] = slot;
5118                         btrfs_item_key_to_cpu(cur, &found_key, slot);
5119                         goto out;
5120                 }
5121                 if (sret && slot > 0)
5122                         slot--;
5123                 /*
5124                  * check this node pointer against the min_trans parameters.
5125                  * If it is too old, old, skip to the next one.
5126                  */
5127                 while (slot < nritems) {
5128                         u64 gen;
5129
5130                         gen = btrfs_node_ptr_generation(cur, slot);
5131                         if (gen < min_trans) {
5132                                 slot++;
5133                                 continue;
5134                         }
5135                         break;
5136                 }
5137 find_next_key:
5138                 /*
5139                  * we didn't find a candidate key in this node, walk forward
5140                  * and find another one
5141                  */
5142                 if (slot >= nritems) {
5143                         path->slots[level] = slot;
5144                         btrfs_set_path_blocking(path);
5145                         sret = btrfs_find_next_key(root, path, min_key, level,
5146                                                   min_trans);
5147                         if (sret == 0) {
5148                                 btrfs_release_path(path);
5149                                 goto again;
5150                         } else {
5151                                 goto out;
5152                         }
5153                 }
5154                 /* save our key for returning back */
5155                 btrfs_node_key_to_cpu(cur, &found_key, slot);
5156                 path->slots[level] = slot;
5157                 if (level == path->lowest_level) {
5158                         ret = 0;
5159                         goto out;
5160                 }
5161                 btrfs_set_path_blocking(path);
5162                 cur = btrfs_read_node_slot(cur, slot);
5163                 if (IS_ERR(cur)) {
5164                         ret = PTR_ERR(cur);
5165                         goto out;
5166                 }
5167
5168                 btrfs_tree_read_lock(cur);
5169
5170                 path->locks[level - 1] = BTRFS_READ_LOCK;
5171                 path->nodes[level - 1] = cur;
5172                 unlock_up(path, level, 1, 0, NULL);
5173         }
5174 out:
5175         path->keep_locks = keep_locks;
5176         if (ret == 0) {
5177                 btrfs_unlock_up_safe(path, path->lowest_level + 1);
5178                 btrfs_set_path_blocking(path);
5179                 memcpy(min_key, &found_key, sizeof(found_key));
5180         }
5181         return ret;
5182 }
5183
5184 /*
5185  * this is similar to btrfs_next_leaf, but does not try to preserve
5186  * and fixup the path.  It looks for and returns the next key in the
5187  * tree based on the current path and the min_trans parameters.
5188  *
5189  * 0 is returned if another key is found, < 0 if there are any errors
5190  * and 1 is returned if there are no higher keys in the tree
5191  *
5192  * path->keep_locks should be set to 1 on the search made before
5193  * calling this function.
5194  */
5195 int btrfs_find_next_key(struct btrfs_root *root, struct btrfs_path *path,
5196                         struct btrfs_key *key, int level, u64 min_trans)
5197 {
5198         int slot;
5199         struct extent_buffer *c;
5200
5201         WARN_ON(!path->keep_locks && !path->skip_locking);
5202         while (level < BTRFS_MAX_LEVEL) {
5203                 if (!path->nodes[level])
5204                         return 1;
5205
5206                 slot = path->slots[level] + 1;
5207                 c = path->nodes[level];
5208 next:
5209                 if (slot >= btrfs_header_nritems(c)) {
5210                         int ret;
5211                         int orig_lowest;
5212                         struct btrfs_key cur_key;
5213                         if (level + 1 >= BTRFS_MAX_LEVEL ||
5214                             !path->nodes[level + 1])
5215                                 return 1;
5216
5217                         if (path->locks[level + 1] || path->skip_locking) {
5218                                 level++;
5219                                 continue;
5220                         }
5221
5222                         slot = btrfs_header_nritems(c) - 1;
5223                         if (level == 0)
5224                                 btrfs_item_key_to_cpu(c, &cur_key, slot);
5225                         else
5226                                 btrfs_node_key_to_cpu(c, &cur_key, slot);
5227
5228                         orig_lowest = path->lowest_level;
5229                         btrfs_release_path(path);
5230                         path->lowest_level = level;
5231                         ret = btrfs_search_slot(NULL, root, &cur_key, path,
5232                                                 0, 0);
5233                         path->lowest_level = orig_lowest;
5234                         if (ret < 0)
5235                                 return ret;
5236
5237                         c = path->nodes[level];
5238                         slot = path->slots[level];
5239                         if (ret == 0)
5240                                 slot++;
5241                         goto next;
5242                 }
5243
5244                 if (level == 0)
5245                         btrfs_item_key_to_cpu(c, key, slot);
5246                 else {
5247                         u64 gen = btrfs_node_ptr_generation(c, slot);
5248
5249                         if (gen < min_trans) {
5250                                 slot++;
5251                                 goto next;
5252                         }
5253                         btrfs_node_key_to_cpu(c, key, slot);
5254                 }
5255                 return 0;
5256         }
5257         return 1;
5258 }
5259
5260 /*
5261  * search the tree again to find a leaf with greater keys
5262  * returns 0 if it found something or 1 if there are no greater leaves.
5263  * returns < 0 on io errors.
5264  */
5265 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path)
5266 {
5267         return btrfs_next_old_leaf(root, path, 0);
5268 }
5269
5270 int btrfs_next_old_leaf(struct btrfs_root *root, struct btrfs_path *path,
5271                         u64 time_seq)
5272 {
5273         int slot;
5274         int level;
5275         struct extent_buffer *c;
5276         struct extent_buffer *next;
5277         struct btrfs_key key;
5278         u32 nritems;
5279         int ret;
5280         int old_spinning = path->leave_spinning;
5281         int next_rw_lock = 0;
5282
5283         nritems = btrfs_header_nritems(path->nodes[0]);
5284         if (nritems == 0)
5285                 return 1;
5286
5287         btrfs_item_key_to_cpu(path->nodes[0], &key, nritems - 1);
5288 again:
5289         level = 1;
5290         next = NULL;
5291         next_rw_lock = 0;
5292         btrfs_release_path(path);
5293
5294         path->keep_locks = 1;
5295         path->leave_spinning = 1;
5296
5297         if (time_seq)
5298                 ret = btrfs_search_old_slot(root, &key, path, time_seq);
5299         else
5300                 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5301         path->keep_locks = 0;
5302
5303         if (ret < 0)
5304                 return ret;
5305
5306         nritems = btrfs_header_nritems(path->nodes[0]);
5307         /*
5308          * by releasing the path above we dropped all our locks.  A balance
5309          * could have added more items next to the key that used to be
5310          * at the very end of the block.  So, check again here and
5311          * advance the path if there are now more items available.
5312          */
5313         if (nritems > 0 && path->slots[0] < nritems - 1) {
5314                 if (ret == 0)
5315                         path->slots[0]++;
5316                 ret = 0;
5317                 goto done;
5318         }
5319         /*
5320          * So the above check misses one case:
5321          * - after releasing the path above, someone has removed the item that
5322          *   used to be at the very end of the block, and balance between leafs
5323          *   gets another one with bigger key.offset to replace it.
5324          *
5325          * This one should be returned as well, or we can get leaf corruption
5326          * later(esp. in __btrfs_drop_extents()).
5327          *
5328          * And a bit more explanation about this check,
5329          * with ret > 0, the key isn't found, the path points to the slot
5330          * where it should be inserted, so the path->slots[0] item must be the
5331          * bigger one.
5332          */
5333         if (nritems > 0 && ret > 0 && path->slots[0] == nritems - 1) {
5334                 ret = 0;
5335                 goto done;
5336         }
5337
5338         while (level < BTRFS_MAX_LEVEL) {
5339                 if (!path->nodes[level]) {
5340                         ret = 1;
5341                         goto done;
5342                 }
5343
5344                 slot = path->slots[level] + 1;
5345                 c = path->nodes[level];
5346                 if (slot >= btrfs_header_nritems(c)) {
5347                         level++;
5348                         if (level == BTRFS_MAX_LEVEL) {
5349                                 ret = 1;
5350                                 goto done;
5351                         }
5352                         continue;
5353                 }
5354
5355                 if (next) {
5356                         btrfs_tree_unlock_rw(next, next_rw_lock);
5357                         free_extent_buffer(next);
5358                 }
5359
5360                 next = c;
5361                 next_rw_lock = path->locks[level];
5362                 ret = read_block_for_search(root, path, &next, level,
5363                                             slot, &key);
5364                 if (ret == -EAGAIN)
5365                         goto again;
5366
5367                 if (ret < 0) {
5368                         btrfs_release_path(path);
5369                         goto done;
5370                 }
5371
5372                 if (!path->skip_locking) {
5373                         ret = btrfs_try_tree_read_lock(next);
5374                         if (!ret && time_seq) {
5375                                 /*
5376                                  * If we don't get the lock, we may be racing
5377                                  * with push_leaf_left, holding that lock while
5378                                  * itself waiting for the leaf we've currently
5379                                  * locked. To solve this situation, we give up
5380                                  * on our lock and cycle.
5381                                  */
5382                                 free_extent_buffer(next);
5383                                 btrfs_release_path(path);
5384                                 cond_resched();
5385                                 goto again;
5386                         }
5387                         if (!ret) {
5388                                 btrfs_set_path_blocking(path);
5389                                 btrfs_tree_read_lock(next);
5390                         }
5391                         next_rw_lock = BTRFS_READ_LOCK;
5392                 }
5393                 break;
5394         }
5395         path->slots[level] = slot;
5396         while (1) {
5397                 level--;
5398                 c = path->nodes[level];
5399                 if (path->locks[level])
5400                         btrfs_tree_unlock_rw(c, path->locks[level]);
5401
5402                 free_extent_buffer(c);
5403                 path->nodes[level] = next;
5404                 path->slots[level] = 0;
5405                 if (!path->skip_locking)
5406                         path->locks[level] = next_rw_lock;
5407                 if (!level)
5408                         break;
5409
5410                 ret = read_block_for_search(root, path, &next, level,
5411                                             0, &key);
5412                 if (ret == -EAGAIN)
5413                         goto again;
5414
5415                 if (ret < 0) {
5416                         btrfs_release_path(path);
5417                         goto done;
5418                 }
5419
5420                 if (!path->skip_locking) {
5421                         ret = btrfs_try_tree_read_lock(next);
5422                         if (!ret) {
5423                                 btrfs_set_path_blocking(path);
5424                                 btrfs_tree_read_lock(next);
5425                         }
5426                         next_rw_lock = BTRFS_READ_LOCK;
5427                 }
5428         }
5429         ret = 0;
5430 done:
5431         unlock_up(path, 0, 1, 0, NULL);
5432         path->leave_spinning = old_spinning;
5433         if (!old_spinning)
5434                 btrfs_set_path_blocking(path);
5435
5436         return ret;
5437 }
5438
5439 /*
5440  * this uses btrfs_prev_leaf to walk backwards in the tree, and keeps
5441  * searching until it gets past min_objectid or finds an item of 'type'
5442  *
5443  * returns 0 if something is found, 1 if nothing was found and < 0 on error
5444  */
5445 int btrfs_previous_item(struct btrfs_root *root,
5446                         struct btrfs_path *path, u64 min_objectid,
5447                         int type)
5448 {
5449         struct btrfs_key found_key;
5450         struct extent_buffer *leaf;
5451         u32 nritems;
5452         int ret;
5453
5454         while (1) {
5455                 if (path->slots[0] == 0) {
5456                         btrfs_set_path_blocking(path);
5457                         ret = btrfs_prev_leaf(root, path);
5458                         if (ret != 0)
5459                                 return ret;
5460                 } else {
5461                         path->slots[0]--;
5462                 }
5463                 leaf = path->nodes[0];
5464                 nritems = btrfs_header_nritems(leaf);
5465                 if (nritems == 0)
5466                         return 1;
5467                 if (path->slots[0] == nritems)
5468                         path->slots[0]--;
5469
5470                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5471                 if (found_key.objectid < min_objectid)
5472                         break;
5473                 if (found_key.type == type)
5474                         return 0;
5475                 if (found_key.objectid == min_objectid &&
5476                     found_key.type < type)
5477                         break;
5478         }
5479         return 1;
5480 }
5481
5482 /*
5483  * search in extent tree to find a previous Metadata/Data extent item with
5484  * min objecitd.
5485  *
5486  * returns 0 if something is found, 1 if nothing was found and < 0 on error
5487  */
5488 int btrfs_previous_extent_item(struct btrfs_root *root,
5489                         struct btrfs_path *path, u64 min_objectid)
5490 {
5491         struct btrfs_key found_key;
5492         struct extent_buffer *leaf;
5493         u32 nritems;
5494         int ret;
5495
5496         while (1) {
5497                 if (path->slots[0] == 0) {
5498                         btrfs_set_path_blocking(path);
5499                         ret = btrfs_prev_leaf(root, path);
5500                         if (ret != 0)
5501                                 return ret;
5502                 } else {
5503                         path->slots[0]--;
5504                 }
5505                 leaf = path->nodes[0];
5506                 nritems = btrfs_header_nritems(leaf);
5507                 if (nritems == 0)
5508                         return 1;
5509                 if (path->slots[0] == nritems)
5510                         path->slots[0]--;
5511
5512                 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5513                 if (found_key.objectid < min_objectid)
5514                         break;
5515                 if (found_key.type == BTRFS_EXTENT_ITEM_KEY ||
5516                     found_key.type == BTRFS_METADATA_ITEM_KEY)
5517                         return 0;
5518                 if (found_key.objectid == min_objectid &&
5519                     found_key.type < BTRFS_EXTENT_ITEM_KEY)
5520                         break;
5521         }
5522         return 1;
5523 }