fs/btrfs/delayed-inode.c

   1 /*
   2  * Copyright (C) 2011 Fujitsu.  All rights reserved.
   3  * Written by Miao Xie <[email protected]>
   4  *
   5  * This program is free software; you can redistribute it and/or
   6  * modify it under the terms of the GNU General Public
   7  * License v2 as published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12  * General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public
  15  * License along with this program; if not, write to the
  16  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  17  * Boston, MA 021110-1307, USA.
  18  */
  19
  20 #include <linux/slab.h>
  21 #include "delayed-inode.h"
  22 #include "disk-io.h"
  23 #include "transaction.h"
  24 #include "ctree.h"
  25
  26 #define BTRFS_DELAYED_WRITEBACK         512
  27 #define BTRFS_DELAYED_BACKGROUND        128
  28 #define BTRFS_DELAYED_BATCH             16
  29
  30 static struct kmem_cache *delayed_node_cache;
  31
  32 int __init btrfs_delayed_inode_init(void)
  33 {
  34         delayed_node_cache = kmem_cache_create("btrfs_delayed_node",
  35                                         sizeof(struct btrfs_delayed_node),
  36                                         0,
  37                                         SLAB_MEM_SPREAD,
  38                                         NULL);
  39         if (!delayed_node_cache)
  40                 return -ENOMEM;
  41         return 0;
  42 }
  43
  44 void btrfs_delayed_inode_exit(void)
  45 {
  46         kmem_cache_destroy(delayed_node_cache);
  47 }
  48
  49 static inline void btrfs_init_delayed_node(
  50                                 struct btrfs_delayed_node *delayed_node,
  51                                 struct btrfs_root *root, u64 inode_id)
  52 {
  53         delayed_node->root = root;
  54         delayed_node->inode_id = inode_id;
  55         atomic_set(&delayed_node->refs, 0);
  56         delayed_node->ins_root = RB_ROOT;
  57         delayed_node->del_root = RB_ROOT;
  58         mutex_init(&delayed_node->mutex);
  59         INIT_LIST_HEAD(&delayed_node->n_list);
  60         INIT_LIST_HEAD(&delayed_node->p_list);
  61 }
  62
  63 static inline int btrfs_is_continuous_delayed_item(
  64                                         struct btrfs_delayed_item *item1,
  65                                         struct btrfs_delayed_item *item2)
  66 {
  67         if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
  68             item1->key.objectid == item2->key.objectid &&
  69             item1->key.type == item2->key.type &&
  70             item1->key.offset + 1 == item2->key.offset)
  71                 return 1;
  72         return 0;
  73 }
  74
  75 static struct btrfs_delayed_node *btrfs_get_delayed_node(
  76                 struct btrfs_inode *btrfs_inode)
  77 {
  78         struct btrfs_root *root = btrfs_inode->root;
  79         u64 ino = btrfs_ino(btrfs_inode);
  80         struct btrfs_delayed_node *node;
  81
  82         node = READ_ONCE(btrfs_inode->delayed_node);
  83         if (node) {
  84                 atomic_inc(&node->refs);
  85                 return node;
  86         }
  87
  88         spin_lock(&root->inode_lock);
  89         node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
  90         if (node) {
  91                 if (btrfs_inode->delayed_node) {
  92                         atomic_inc(&node->refs);        /* can be accessed */
  93                         BUG_ON(btrfs_inode->delayed_node != node);
  94                         spin_unlock(&root->inode_lock);
  95                         return node;
  96                 }
  97                 btrfs_inode->delayed_node = node;
  98                 /* can be accessed and cached in the inode */
  99                 atomic_add(2, &node->refs);
 100                 spin_unlock(&root->inode_lock);
 101                 return node;
 102         }
 103         spin_unlock(&root->inode_lock);
 104
 105         return NULL;
 106 }
 107
 108 /* Will return either the node or PTR_ERR(-ENOMEM) */
 109 static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
 110                 struct btrfs_inode *btrfs_inode)
 111 {
 112         struct btrfs_delayed_node *node;
 113         struct btrfs_root *root = btrfs_inode->root;
 114         u64 ino = btrfs_ino(btrfs_inode);
 115         int ret;
 116
 117 again:
 118         node = btrfs_get_delayed_node(btrfs_inode);
 119         if (node)
 120                 return node;
 121
 122         node = kmem_cache_zalloc(delayed_node_cache, GFP_NOFS);
 123         if (!node)
 124                 return ERR_PTR(-ENOMEM);
 125         btrfs_init_delayed_node(node, root, ino);
 126
 127         /* cached in the btrfs inode and can be accessed */
 128         atomic_add(2, &node->refs);
 129
 130         ret = radix_tree_preload(GFP_NOFS);
 131         if (ret) {
 132                 kmem_cache_free(delayed_node_cache, node);
 133                 return ERR_PTR(ret);
 134         }
 135
 136         spin_lock(&root->inode_lock);
 137         ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
 138         if (ret == -EEXIST) {
 139                 spin_unlock(&root->inode_lock);
 140                 kmem_cache_free(delayed_node_cache, node);
 141                 radix_tree_preload_end();
 142                 goto again;
 143         }
 144         btrfs_inode->delayed_node = node;
 145         spin_unlock(&root->inode_lock);
 146         radix_tree_preload_end();
 147
 148         return node;
 149 }
 150
 151 /*
 152  * Call it when holding delayed_node->mutex
 153  *
 154  * If mod = 1, add this node into the prepared list.
 155  */
 156 static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
 157                                      struct btrfs_delayed_node *node,
 158                                      int mod)
 159 {
 160         spin_lock(&root->lock);
 161         if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
 162                 if (!list_empty(&node->p_list))
 163                         list_move_tail(&node->p_list, &root->prepare_list);
 164                 else if (mod)
 165                         list_add_tail(&node->p_list, &root->prepare_list);
 166         } else {
 167                 list_add_tail(&node->n_list, &root->node_list);
 168                 list_add_tail(&node->p_list, &root->prepare_list);
 169                 atomic_inc(&node->refs);        /* inserted into list */
 170                 root->nodes++;
 171                 set_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
 172         }
 173         spin_unlock(&root->lock);
 174 }
 175
 176 /* Call it when holding delayed_node->mutex */
 177 static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
 178                                        struct btrfs_delayed_node *node)
 179 {
 180         spin_lock(&root->lock);
 181         if (test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
 182                 root->nodes--;
 183                 atomic_dec(&node->refs);        /* not in the list */
 184                 list_del_init(&node->n_list);
 185                 if (!list_empty(&node->p_list))
 186                         list_del_init(&node->p_list);
 187                 clear_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags);
 188         }
 189         spin_unlock(&root->lock);
 190 }
 191
 192 static struct btrfs_delayed_node *btrfs_first_delayed_node(
 193                         struct btrfs_delayed_root *delayed_root)
 194 {
 195         struct list_head *p;
 196         struct btrfs_delayed_node *node = NULL;
 197
 198         spin_lock(&delayed_root->lock);
 199         if (list_empty(&delayed_root->node_list))
 200                 goto out;
 201
 202         p = delayed_root->node_list.next;
 203         node = list_entry(p, struct btrfs_delayed_node, n_list);
 204         atomic_inc(&node->refs);
 205 out:
 206         spin_unlock(&delayed_root->lock);
 207
 208         return node;
 209 }
 210
 211 static struct btrfs_delayed_node *btrfs_next_delayed_node(
 212                                                 struct btrfs_delayed_node *node)
 213 {
 214         struct btrfs_delayed_root *delayed_root;
 215         struct list_head *p;
 216         struct btrfs_delayed_node *next = NULL;
 217
 218         delayed_root = node->root->fs_info->delayed_root;
 219         spin_lock(&delayed_root->lock);
 220         if (!test_bit(BTRFS_DELAYED_NODE_IN_LIST, &node->flags)) {
 221                 /* not in the list */
 222                 if (list_empty(&delayed_root->node_list))
 223                         goto out;
 224                 p = delayed_root->node_list.next;
 225         } else if (list_is_last(&node->n_list, &delayed_root->node_list))
 226                 goto out;
 227         else
 228                 p = node->n_list.next;
 229
 230         next = list_entry(p, struct btrfs_delayed_node, n_list);
 231         atomic_inc(&next->refs);
 232 out:
 233         spin_unlock(&delayed_root->lock);
 234
 235         return next;
 236 }
 237
 238 static void __btrfs_release_delayed_node(
 239                                 struct btrfs_delayed_node *delayed_node,
 240                                 int mod)
 241 {
 242         struct btrfs_delayed_root *delayed_root;
 243
 244         if (!delayed_node)
 245                 return;
 246
 247         delayed_root = delayed_node->root->fs_info->delayed_root;
 248
 249         mutex_lock(&delayed_node->mutex);
 250         if (delayed_node->count)
 251                 btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
 252         else
 253                 btrfs_dequeue_delayed_node(delayed_root, delayed_node);
 254         mutex_unlock(&delayed_node->mutex);
 255
 256         if (atomic_dec_and_test(&delayed_node->refs)) {
 257                 bool free = false;
 258                 struct btrfs_root *root = delayed_node->root;
 259                 spin_lock(&root->inode_lock);
 260                 if (atomic_read(&delayed_node->refs) == 0) {
 261                         radix_tree_delete(&root->delayed_nodes_tree,
 262                                           delayed_node->inode_id);
 263                         free = true;
 264                 }
 265                 spin_unlock(&root->inode_lock);
 266                 if (free)
 267                         kmem_cache_free(delayed_node_cache, delayed_node);
 268         }
 269 }
 270
 271 static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
 272 {
 273         __btrfs_release_delayed_node(node, 0);
 274 }
 275
 276 static struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
 277                                         struct btrfs_delayed_root *delayed_root)
 278 {
 279         struct list_head *p;
 280         struct btrfs_delayed_node *node = NULL;
 281
 282         spin_lock(&delayed_root->lock);
 283         if (list_empty(&delayed_root->prepare_list))
 284                 goto out;
 285
 286         p = delayed_root->prepare_list.next;
 287         list_del_init(p);
 288         node = list_entry(p, struct btrfs_delayed_node, p_list);
 289         atomic_inc(&node->refs);
 290 out:
 291         spin_unlock(&delayed_root->lock);
 292
 293         return node;
 294 }
 295
 296 static inline void btrfs_release_prepared_delayed_node(
 297                                         struct btrfs_delayed_node *node)
 298 {
 299         __btrfs_release_delayed_node(node, 1);
 300 }
 301
 302 static struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
 303 {
 304         struct btrfs_delayed_item *item;
 305         item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
 306         if (item) {
 307                 item->data_len = data_len;
 308                 item->ins_or_del = 0;
 309                 item->bytes_reserved = 0;
 310                 item->delayed_node = NULL;
 311                 atomic_set(&item->refs, 1);
 312         }
 313         return item;
 314 }
 315
 316 /*
 317  * __btrfs_lookup_delayed_item - look up the delayed item by key
 318  * @delayed_node: pointer to the delayed node
 319  * @key:          the key to look up
 320  * @prev:         used to store the prev item if the right item isn't found
 321  * @next:         used to store the next item if the right item isn't found
 322  *
 323  * Note: if we don't find the right item, we will return the prev item and
 324  * the next item.
 325  */
 326 static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
 327                                 struct rb_root *root,
 328                                 struct btrfs_key *key,
 329                                 struct btrfs_delayed_item **prev,
 330                                 struct btrfs_delayed_item **next)
 331 {
 332         struct rb_node *node, *prev_node = NULL;
 333         struct btrfs_delayed_item *delayed_item = NULL;
 334         int ret = 0;
 335
 336         node = root->rb_node;
 337
 338         while (node) {
 339                 delayed_item = rb_entry(node, struct btrfs_delayed_item,
 340                                         rb_node);
 341                 prev_node = node;
 342                 ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
 343                 if (ret < 0)
 344                         node = node->rb_right;
 345                 else if (ret > 0)
 346                         node = node->rb_left;
 347                 else
 348                         return delayed_item;
 349         }
 350
 351         if (prev) {
 352                 if (!prev_node)
 353                         *prev = NULL;
 354                 else if (ret < 0)
 355                         *prev = delayed_item;
 356                 else if ((node = rb_prev(prev_node)) != NULL) {
 357                         *prev = rb_entry(node, struct btrfs_delayed_item,
 358                                          rb_node);
 359                 } else
 360                         *prev = NULL;
 361         }
 362
 363         if (next) {
 364                 if (!prev_node)
 365                         *next = NULL;
 366                 else if (ret > 0)
 367                         *next = delayed_item;
 368                 else if ((node = rb_next(prev_node)) != NULL) {
 369                         *next = rb_entry(node, struct btrfs_delayed_item,
 370                                          rb_node);
 371                 } else
 372                         *next = NULL;
 373         }
 374         return NULL;
 375 }
 376
 377 static struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
 378                                         struct btrfs_delayed_node *delayed_node,
 379                                         struct btrfs_key *key)
 380 {
 381         return __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
 382                                            NULL, NULL);
 383 }
 384
 385 static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
 386                                     struct btrfs_delayed_item *ins,
 387                                     int action)
 388 {
 389         struct rb_node **p, *node;
 390         struct rb_node *parent_node = NULL;
 391         struct rb_root *root;
 392         struct btrfs_delayed_item *item;
 393         int cmp;
 394
 395         if (action == BTRFS_DELAYED_INSERTION_ITEM)
 396                 root = &delayed_node->ins_root;
 397         else if (action == BTRFS_DELAYED_DELETION_ITEM)
 398                 root = &delayed_node->del_root;
 399         else
 400                 BUG();
 401         p = &root->rb_node;
 402         node = &ins->rb_node;
 403
 404         while (*p) {
 405                 parent_node = *p;
 406                 item = rb_entry(parent_node, struct btrfs_delayed_item,
 407                                  rb_node);
 408
 409                 cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
 410                 if (cmp < 0)
 411                         p = &(*p)->rb_right;
 412                 else if (cmp > 0)
 413                         p = &(*p)->rb_left;
 414                 else
 415                         return -EEXIST;
 416         }
 417
 418         rb_link_node(node, parent_node, p);
 419         rb_insert_color(node, root);
 420         ins->delayed_node = delayed_node;
 421         ins->ins_or_del = action;
 422
 423         if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
 424             action == BTRFS_DELAYED_INSERTION_ITEM &&
 425             ins->key.offset >= delayed_node->index_cnt)
 426                         delayed_node->index_cnt = ins->key.offset + 1;
 427
 428         delayed_node->count++;
 429         atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
 430         return 0;
 431 }
 432
 433 static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
 434                                               struct btrfs_delayed_item *item)
 435 {
 436         return __btrfs_add_delayed_item(node, item,
 437                                         BTRFS_DELAYED_INSERTION_ITEM);
 438 }
 439
 440 static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
 441                                              struct btrfs_delayed_item *item)
 442 {
 443         return __btrfs_add_delayed_item(node, item,
 444                                         BTRFS_DELAYED_DELETION_ITEM);
 445 }
 446
 447 static void finish_one_item(struct btrfs_delayed_root *delayed_root)
 448 {
 449         int seq = atomic_inc_return(&delayed_root->items_seq);
 450
 451         /*
 452          * atomic_dec_return implies a barrier for waitqueue_active
 453          */
 454         if ((atomic_dec_return(&delayed_root->items) <
 455             BTRFS_DELAYED_BACKGROUND || seq % BTRFS_DELAYED_BATCH == 0) &&
 456             waitqueue_active(&delayed_root->wait))
 457                 wake_up(&delayed_root->wait);
 458 }
 459
 460 static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
 461 {
 462         struct rb_root *root;
 463         struct btrfs_delayed_root *delayed_root;
 464
 465         delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
 466
 467         BUG_ON(!delayed_root);
 468         BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
 469                delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
 470
 471         if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
 472                 root = &delayed_item->delayed_node->ins_root;
 473         else
 474                 root = &delayed_item->delayed_node->del_root;
 475
 476         rb_erase(&delayed_item->rb_node, root);
 477         delayed_item->delayed_node->count--;
 478
 479         finish_one_item(delayed_root);
 480 }
 481
 482 static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
 483 {
 484         if (item) {
 485                 __btrfs_remove_delayed_item(item);
 486                 if (atomic_dec_and_test(&item->refs))
 487                         kfree(item);
 488         }
 489 }
 490
 491 static struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
 492                                         struct btrfs_delayed_node *delayed_node)
 493 {
 494         struct rb_node *p;
 495         struct btrfs_delayed_item *item = NULL;
 496
 497         p = rb_first(&delayed_node->ins_root);
 498         if (p)
 499                 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
 500
 501         return item;
 502 }
 503
 504 static struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
 505                                         struct btrfs_delayed_node *delayed_node)
 506 {
 507         struct rb_node *p;
 508         struct btrfs_delayed_item *item = NULL;
 509
 510         p = rb_first(&delayed_node->del_root);
 511         if (p)
 512                 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
 513
 514         return item;
 515 }
 516
 517 static struct btrfs_delayed_item *__btrfs_next_delayed_item(
 518                                                 struct btrfs_delayed_item *item)
 519 {
 520         struct rb_node *p;
 521         struct btrfs_delayed_item *next = NULL;
 522
 523         p = rb_next(&item->rb_node);
 524         if (p)
 525                 next = rb_entry(p, struct btrfs_delayed_item, rb_node);
 526
 527         return next;
 528 }
 529
 530 static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
 531                                                struct btrfs_fs_info *fs_info,
 532                                                struct btrfs_delayed_item *item)
 533 {
 534         struct btrfs_block_rsv *src_rsv;
 535         struct btrfs_block_rsv *dst_rsv;
 536         u64 num_bytes;
 537         int ret;
 538
 539         if (!trans->bytes_reserved)
 540                 return 0;
 541
 542         src_rsv = trans->block_rsv;
 543         dst_rsv = &fs_info->delayed_block_rsv;
 544
 545         num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
 546         ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, 1);
 547         if (!ret) {
 548                 trace_btrfs_space_reservation(fs_info, "delayed_item",
 549                                               item->key.objectid,
 550                                               num_bytes, 1);
 551                 item->bytes_reserved = num_bytes;
 552         }
 553
 554         return ret;
 555 }
 556
 557 static void btrfs_delayed_item_release_metadata(struct btrfs_fs_info *fs_info,
 558                                                 struct btrfs_delayed_item *item)
 559 {
 560         struct btrfs_block_rsv *rsv;
 561
 562         if (!item->bytes_reserved)
 563                 return;
 564
 565         rsv = &fs_info->delayed_block_rsv;
 566         trace_btrfs_space_reservation(fs_info, "delayed_item",
 567                                       item->key.objectid, item->bytes_reserved,
 568                                       0);
 569         btrfs_block_rsv_release(fs_info, rsv,
 570                                 item->bytes_reserved);
 571 }
 572
 573 static int btrfs_delayed_inode_reserve_metadata(
 574                                         struct btrfs_trans_handle *trans,
 575                                         struct btrfs_root *root,
 576                                         struct btrfs_inode *inode,
 577                                         struct btrfs_delayed_node *node)
 578 {
 579         struct btrfs_fs_info *fs_info = root->fs_info;
 580         struct btrfs_block_rsv *src_rsv;
 581         struct btrfs_block_rsv *dst_rsv;
 582         u64 num_bytes;
 583         int ret;
 584         bool release = false;
 585
 586         src_rsv = trans->block_rsv;
 587         dst_rsv = &fs_info->delayed_block_rsv;
 588
 589         num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
 590
 591         /*
 592          * If our block_rsv is the delalloc block reserve then check and see if
 593          * we have our extra reservation for updating the inode.  If not fall
 594          * through and try to reserve space quickly.
 595          *
 596          * We used to try and steal from the delalloc block rsv or the global
 597          * reserve, but we'd steal a full reservation, which isn't kind.  We are
 598          * here through delalloc which means we've likely just cowed down close
 599          * to the leaf that contains the inode, so we would steal less just
 600          * doing the fallback inode update, so if we do end up having to steal
 601          * from the global block rsv we hopefully only steal one or two blocks
 602          * worth which is less likely to hurt us.
 603          */
 604         if (src_rsv && src_rsv->type == BTRFS_BLOCK_RSV_DELALLOC) {
 605                 spin_lock(&inode->lock);
 606                 if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
 607                                        &inode->runtime_flags))
 608                         release = true;
 609                 else
 610                         src_rsv = NULL;
 611                 spin_unlock(&inode->lock);
 612         }
 613
 614         /*
 615          * btrfs_dirty_inode will update the inode under btrfs_join_transaction
 616          * which doesn't reserve space for speed.  This is a problem since we
 617          * still need to reserve space for this update, so try to reserve the
 618          * space.
 619          *
 620          * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
 621          * we're accounted for.
 622          */
 623         if (!src_rsv || (!trans->bytes_reserved &&
 624                          src_rsv->type != BTRFS_BLOCK_RSV_DELALLOC)) {
 625                 ret = btrfs_block_rsv_add(root, dst_rsv, num_bytes,
 626                                           BTRFS_RESERVE_NO_FLUSH);
 627                 /*
 628                  * Since we're under a transaction reserve_metadata_bytes could
 629                  * try to commit the transaction which will make it return
 630                  * EAGAIN to make us stop the transaction we have, so return
 631                  * ENOSPC instead so that btrfs_dirty_inode knows what to do.
 632                  */
 633                 if (ret == -EAGAIN)
 634                         ret = -ENOSPC;
 635                 if (!ret) {
 636                         node->bytes_reserved = num_bytes;
 637                         trace_btrfs_space_reservation(fs_info,
 638                                                       "delayed_inode",
 639                                                       btrfs_ino(inode),
 640                                                       num_bytes, 1);
 641                 }
 642                 return ret;
 643         }
 644
 645         ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes, 1);
 646
 647         /*
 648          * Migrate only takes a reservation, it doesn't touch the size of the
 649          * block_rsv.  This is to simplify people who don't normally have things
 650          * migrated from their block rsv.  If they go to release their
 651          * reservation, that will decrease the size as well, so if migrate
 652          * reduced size we'd end up with a negative size.  But for the
 653          * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
 654          * but we could in fact do this reserve/migrate dance several times
 655          * between the time we did the original reservation and we'd clean it
 656          * up.  So to take care of this, release the space for the meta
 657          * reservation here.  I think it may be time for a documentation page on
 658          * how block rsvs. work.
 659          */
 660         if (!ret) {
 661                 trace_btrfs_space_reservation(fs_info, "delayed_inode",
 662                                               btrfs_ino(inode), num_bytes, 1);
 663                 node->bytes_reserved = num_bytes;
 664         }
 665
 666         if (release) {
 667                 trace_btrfs_space_reservation(fs_info, "delalloc",
 668                                               btrfs_ino(inode), num_bytes, 0);
 669                 btrfs_block_rsv_release(fs_info, src_rsv, num_bytes);
 670         }
 671
 672         return ret;
 673 }
 674
 675 static void btrfs_delayed_inode_release_metadata(struct btrfs_fs_info *fs_info,
 676                                                 struct btrfs_delayed_node *node)
 677 {
 678         struct btrfs_block_rsv *rsv;
 679
 680         if (!node->bytes_reserved)
 681                 return;
 682
 683         rsv = &fs_info->delayed_block_rsv;
 684         trace_btrfs_space_reservation(fs_info, "delayed_inode",
 685                                       node->inode_id, node->bytes_reserved, 0);
 686         btrfs_block_rsv_release(fs_info, rsv,
 687                                 node->bytes_reserved);
 688         node->bytes_reserved = 0;
 689 }
 690
 691 /*
 692  * This helper will insert some continuous items into the same leaf according
 693  * to the free space of the leaf.
 694  */
 695 static int btrfs_batch_insert_items(struct btrfs_root *root,
 696                                     struct btrfs_path *path,
 697                                     struct btrfs_delayed_item *item)
 698 {
 699         struct btrfs_fs_info *fs_info = root->fs_info;
 700         struct btrfs_delayed_item *curr, *next;
 701         int free_space;
 702         int total_data_size = 0, total_size = 0;
 703         struct extent_buffer *leaf;
 704         char *data_ptr;
 705         struct btrfs_key *keys;
 706         u32 *data_size;
 707         struct list_head head;
 708         int slot;
 709         int nitems;
 710         int i;
 711         int ret = 0;
 712
 713         BUG_ON(!path->nodes[0]);
 714
 715         leaf = path->nodes[0];
 716         free_space = btrfs_leaf_free_space(fs_info, leaf);
 717         INIT_LIST_HEAD(&head);
 718
 719         next = item;
 720         nitems = 0;
 721
 722         /*
 723          * count the number of the continuous items that we can insert in batch
 724          */
 725         while (total_size + next->data_len + sizeof(struct btrfs_item) <=
 726                free_space) {
 727                 total_data_size += next->data_len;
 728                 total_size += next->data_len + sizeof(struct btrfs_item);
 729                 list_add_tail(&next->tree_list, &head);
 730                 nitems++;
 731
 732                 curr = next;
 733                 next = __btrfs_next_delayed_item(curr);
 734                 if (!next)
 735                         break;
 736
 737                 if (!btrfs_is_continuous_delayed_item(curr, next))
 738                         break;
 739         }
 740
 741         if (!nitems) {
 742                 ret = 0;
 743                 goto out;
 744         }
 745
 746         /*
 747          * we need allocate some memory space, but it might cause the task
 748          * to sleep, so we set all locked nodes in the path to blocking locks
 749          * first.
 750          */
 751         btrfs_set_path_blocking(path);
 752
 753         keys = kmalloc_array(nitems, sizeof(struct btrfs_key), GFP_NOFS);
 754         if (!keys) {
 755                 ret = -ENOMEM;
 756                 goto out;
 757         }
 758
 759         data_size = kmalloc_array(nitems, sizeof(u32), GFP_NOFS);
 760         if (!data_size) {
 761                 ret = -ENOMEM;
 762                 goto error;
 763         }
 764
 765         /* get keys of all the delayed items */
 766         i = 0;
 767         list_for_each_entry(next, &head, tree_list) {
 768                 keys[i] = next->key;
 769                 data_size[i] = next->data_len;
 770                 i++;
 771         }
 772
 773         /* reset all the locked nodes in the patch to spinning locks. */
 774         btrfs_clear_path_blocking(path, NULL, 0);
 775
 776         /* insert the keys of the items */
 777         setup_items_for_insert(root, path, keys, data_size,
 778                                total_data_size, total_size, nitems);
 779
 780         /* insert the dir index items */
 781         slot = path->slots[0];
 782         list_for_each_entry_safe(curr, next, &head, tree_list) {
 783                 data_ptr = btrfs_item_ptr(leaf, slot, char);
 784                 write_extent_buffer(leaf, &curr->data,
 785                                     (unsigned long)data_ptr,
 786                                     curr->data_len);
 787                 slot++;
 788
 789                 btrfs_delayed_item_release_metadata(fs_info, curr);
 790
 791                 list_del(&curr->tree_list);
 792                 btrfs_release_delayed_item(curr);
 793         }
 794
 795 error:
 796         kfree(data_size);
 797         kfree(keys);
 798 out:
 799         return ret;
 800 }
 801
 802 /*
 803  * This helper can just do simple insertion that needn't extend item for new
 804  * data, such as directory name index insertion, inode insertion.
 805  */
 806 static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
 807                                      struct btrfs_root *root,
 808                                      struct btrfs_path *path,
 809                                      struct btrfs_delayed_item *delayed_item)
 810 {
 811         struct btrfs_fs_info *fs_info = root->fs_info;
 812         struct extent_buffer *leaf;
 813         char *ptr;
 814         int ret;
 815
 816         ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
 817                                       delayed_item->data_len);
 818         if (ret < 0 && ret != -EEXIST)
 819                 return ret;
 820
 821         leaf = path->nodes[0];
 822
 823         ptr = btrfs_item_ptr(leaf, path->slots[0], char);
 824
 825         write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
 826                             delayed_item->data_len);
 827         btrfs_mark_buffer_dirty(leaf);
 828
 829         btrfs_delayed_item_release_metadata(fs_info, delayed_item);
 830         return 0;
 831 }
 832
 833 /*
 834  * we insert an item first, then if there are some continuous items, we try
 835  * to insert those items into the same leaf.
 836  */
 837 static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
 838                                       struct btrfs_path *path,
 839                                       struct btrfs_root *root,
 840                                       struct btrfs_delayed_node *node)
 841 {
 842         struct btrfs_delayed_item *curr, *prev;
 843         int ret = 0;
 844
 845 do_again:
 846         mutex_lock(&node->mutex);
 847         curr = __btrfs_first_delayed_insertion_item(node);
 848         if (!curr)
 849                 goto insert_end;
 850
 851         ret = btrfs_insert_delayed_item(trans, root, path, curr);
 852         if (ret < 0) {
 853                 btrfs_release_path(path);
 854                 goto insert_end;
 855         }
 856
 857         prev = curr;
 858         curr = __btrfs_next_delayed_item(prev);
 859         if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
 860                 /* insert the continuous items into the same leaf */
 861                 path->slots[0]++;
 862                 btrfs_batch_insert_items(root, path, curr);
 863         }
 864         btrfs_release_delayed_item(prev);
 865         btrfs_mark_buffer_dirty(path->nodes[0]);
 866
 867         btrfs_release_path(path);
 868         mutex_unlock(&node->mutex);
 869         goto do_again;
 870
 871 insert_end:
 872         mutex_unlock(&node->mutex);
 873         return ret;
 874 }
 875
 876 static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
 877                                     struct btrfs_root *root,
 878                                     struct btrfs_path *path,
 879                                     struct btrfs_delayed_item *item)
 880 {
 881         struct btrfs_fs_info *fs_info = root->fs_info;
 882         struct btrfs_delayed_item *curr, *next;
 883         struct extent_buffer *leaf;
 884         struct btrfs_key key;
 885         struct list_head head;
 886         int nitems, i, last_item;
 887         int ret = 0;
 888
 889         BUG_ON(!path->nodes[0]);
 890
 891         leaf = path->nodes[0];
 892
 893         i = path->slots[0];
 894         last_item = btrfs_header_nritems(leaf) - 1;
 895         if (i > last_item)
 896                 return -ENOENT; /* FIXME: Is errno suitable? */
 897
 898         next = item;
 899         INIT_LIST_HEAD(&head);
 900         btrfs_item_key_to_cpu(leaf, &key, i);
 901         nitems = 0;
 902         /*
 903          * count the number of the dir index items that we can delete in batch
 904          */
 905         while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
 906                 list_add_tail(&next->tree_list, &head);
 907                 nitems++;
 908
 909                 curr = next;
 910                 next = __btrfs_next_delayed_item(curr);
 911                 if (!next)
 912                         break;
 913
 914                 if (!btrfs_is_continuous_delayed_item(curr, next))
 915                         break;
 916
 917                 i++;
 918                 if (i > last_item)
 919                         break;
 920                 btrfs_item_key_to_cpu(leaf, &key, i);
 921         }
 922
 923         if (!nitems)
 924                 return 0;
 925
 926         ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
 927         if (ret)
 928                 goto out;
 929
 930         list_for_each_entry_safe(curr, next, &head, tree_list) {
 931                 btrfs_delayed_item_release_metadata(fs_info, curr);
 932                 list_del(&curr->tree_list);
 933                 btrfs_release_delayed_item(curr);
 934         }
 935
 936 out:
 937         return ret;
 938 }
 939
 940 static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
 941                                       struct btrfs_path *path,
 942                                       struct btrfs_root *root,
 943                                       struct btrfs_delayed_node *node)
 944 {
 945         struct btrfs_delayed_item *curr, *prev;
 946         int ret = 0;
 947
 948 do_again:
 949         mutex_lock(&node->mutex);
 950         curr = __btrfs_first_delayed_deletion_item(node);
 951         if (!curr)
 952                 goto delete_fail;
 953
 954         ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
 955         if (ret < 0)
 956                 goto delete_fail;
 957         else if (ret > 0) {
 958                 /*
 959                  * can't find the item which the node points to, so this node
 960                  * is invalid, just drop it.
 961                  */
 962                 prev = curr;
 963                 curr = __btrfs_next_delayed_item(prev);
 964                 btrfs_release_delayed_item(prev);
 965                 ret = 0;
 966                 btrfs_release_path(path);
 967                 if (curr) {
 968                         mutex_unlock(&node->mutex);
 969                         goto do_again;
 970                 } else
 971                         goto delete_fail;
 972         }
 973
 974         btrfs_batch_delete_items(trans, root, path, curr);
 975         btrfs_release_path(path);
 976         mutex_unlock(&node->mutex);
 977         goto do_again;
 978
 979 delete_fail:
 980         btrfs_release_path(path);
 981         mutex_unlock(&node->mutex);
 982         return ret;
 983 }
 984
 985 static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
 986 {
 987         struct btrfs_delayed_root *delayed_root;
 988
 989         if (delayed_node &&
 990             test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
 991                 BUG_ON(!delayed_node->root);
 992                 clear_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
 993                 delayed_node->count--;
 994
 995                 delayed_root = delayed_node->root->fs_info->delayed_root;
 996                 finish_one_item(delayed_root);
 997         }
 998 }
 999
1000 static void btrfs_release_delayed_iref(struct btrfs_delayed_node *delayed_node)
1001 {
1002         struct btrfs_delayed_root *delayed_root;
1003
1004         ASSERT(delayed_node->root);
1005         clear_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
1006         delayed_node->count--;
1007
1008         delayed_root = delayed_node->root->fs_info->delayed_root;
1009         finish_one_item(delayed_root);
1010 }
1011
1012 static int __btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1013                                         struct btrfs_root *root,
1014                                         struct btrfs_path *path,
1015                                         struct btrfs_delayed_node *node)
1016 {
1017         struct btrfs_fs_info *fs_info = root->fs_info;
1018         struct btrfs_key key;
1019         struct btrfs_inode_item *inode_item;
1020         struct extent_buffer *leaf;
1021         int mod;
1022         int ret;
1023
1024         key.objectid = node->inode_id;
1025         key.type = BTRFS_INODE_ITEM_KEY;
1026         key.offset = 0;
1027
1028         if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1029                 mod = -1;
1030         else
1031                 mod = 1;
1032
1033         ret = btrfs_lookup_inode(trans, root, path, &key, mod);
1034         if (ret > 0) {
1035                 btrfs_release_path(path);
1036                 return -ENOENT;
1037         } else if (ret < 0) {
1038                 return ret;
1039         }
1040
1041         leaf = path->nodes[0];
1042         inode_item = btrfs_item_ptr(leaf, path->slots[0],
1043                                     struct btrfs_inode_item);
1044         write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
1045                             sizeof(struct btrfs_inode_item));
1046         btrfs_mark_buffer_dirty(leaf);
1047
1048         if (!test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &node->flags))
1049                 goto no_iref;
1050
1051         path->slots[0]++;
1052         if (path->slots[0] >= btrfs_header_nritems(leaf))
1053                 goto search;
1054 again:
1055         btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1056         if (key.objectid != node->inode_id)
1057                 goto out;
1058
1059         if (key.type != BTRFS_INODE_REF_KEY &&
1060             key.type != BTRFS_INODE_EXTREF_KEY)
1061                 goto out;
1062
1063         /*
1064          * Delayed iref deletion is for the inode who has only one link,
1065          * so there is only one iref. The case that several irefs are
1066          * in the same item doesn't exist.
1067          */
1068         btrfs_del_item(trans, root, path);
1069 out:
1070         btrfs_release_delayed_iref(node);
1071 no_iref:
1072         btrfs_release_path(path);
1073 err_out:
1074         btrfs_delayed_inode_release_metadata(fs_info, node);
1075         btrfs_release_delayed_inode(node);
1076
1077         return ret;
1078
1079 search:
1080         btrfs_release_path(path);
1081
1082         key.type = BTRFS_INODE_EXTREF_KEY;
1083         key.offset = -1;
1084         ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1085         if (ret < 0)
1086                 goto err_out;
1087         ASSERT(ret);
1088
1089         ret = 0;
1090         leaf = path->nodes[0];
1091         path->slots[0]--;
1092         goto again;
1093 }
1094
1095 static inline int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1096                                              struct btrfs_root *root,
1097                                              struct btrfs_path *path,
1098                                              struct btrfs_delayed_node *node)
1099 {
1100         int ret;
1101
1102         mutex_lock(&node->mutex);
1103         if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &node->flags)) {
1104                 mutex_unlock(&node->mutex);
1105                 return 0;
1106         }
1107
1108         ret = __btrfs_update_delayed_inode(trans, root, path, node);
1109         mutex_unlock(&node->mutex);
1110         return ret;
1111 }
1112
1113 static inline int
1114 __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1115                                    struct btrfs_path *path,
1116                                    struct btrfs_delayed_node *node)
1117 {
1118         int ret;
1119
1120         ret = btrfs_insert_delayed_items(trans, path, node->root, node);
1121         if (ret)
1122                 return ret;
1123
1124         ret = btrfs_delete_delayed_items(trans, path, node->root, node);
1125         if (ret)
1126                 return ret;
1127
1128         ret = btrfs_update_delayed_inode(trans, node->root, path, node);
1129         return ret;
1130 }
1131
1132 /*
1133  * Called when committing the transaction.
1134  * Returns 0 on success.
1135  * Returns < 0 on error and returns with an aborted transaction with any
1136  * outstanding delayed items cleaned up.
1137  */
1138 static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1139                                      struct btrfs_fs_info *fs_info, int nr)
1140 {
1141         struct btrfs_delayed_root *delayed_root;
1142         struct btrfs_delayed_node *curr_node, *prev_node;
1143         struct btrfs_path *path;
1144         struct btrfs_block_rsv *block_rsv;
1145         int ret = 0;
1146         bool count = (nr > 0);
1147
1148         if (trans->aborted)
1149                 return -EIO;
1150
1151         path = btrfs_alloc_path();
1152         if (!path)
1153                 return -ENOMEM;
1154         path->leave_spinning = 1;
1155
1156         block_rsv = trans->block_rsv;
1157         trans->block_rsv = &fs_info->delayed_block_rsv;
1158
1159         delayed_root = fs_info->delayed_root;
1160
1161         curr_node = btrfs_first_delayed_node(delayed_root);
1162         while (curr_node && (!count || (count && nr--))) {
1163                 ret = __btrfs_commit_inode_delayed_items(trans, path,
1164                                                          curr_node);
1165                 if (ret) {
1166                         btrfs_release_delayed_node(curr_node);
1167                         curr_node = NULL;
1168                         btrfs_abort_transaction(trans, ret);
1169                         break;
1170                 }
1171
1172                 prev_node = curr_node;
1173                 curr_node = btrfs_next_delayed_node(curr_node);
1174                 btrfs_release_delayed_node(prev_node);
1175         }
1176
1177         if (curr_node)
1178                 btrfs_release_delayed_node(curr_node);
1179         btrfs_free_path(path);
1180         trans->block_rsv = block_rsv;
1181
1182         return ret;
1183 }
1184
1185 int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1186                             struct btrfs_fs_info *fs_info)
1187 {
1188         return __btrfs_run_delayed_items(trans, fs_info, -1);
1189 }
1190
1191 int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans,
1192                                struct btrfs_fs_info *fs_info, int nr)
1193 {
1194         return __btrfs_run_delayed_items(trans, fs_info, nr);
1195 }
1196
1197 int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1198                                      struct btrfs_inode *inode)
1199 {
1200         struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1201         struct btrfs_path *path;
1202         struct btrfs_block_rsv *block_rsv;
1203         int ret;
1204
1205         if (!delayed_node)
1206                 return 0;
1207
1208         mutex_lock(&delayed_node->mutex);
1209         if (!delayed_node->count) {
1210                 mutex_unlock(&delayed_node->mutex);
1211                 btrfs_release_delayed_node(delayed_node);
1212                 return 0;
1213         }
1214         mutex_unlock(&delayed_node->mutex);
1215
1216         path = btrfs_alloc_path();
1217         if (!path) {
1218                 btrfs_release_delayed_node(delayed_node);
1219                 return -ENOMEM;
1220         }
1221         path->leave_spinning = 1;
1222
1223         block_rsv = trans->block_rsv;
1224         trans->block_rsv = &delayed_node->root->fs_info->delayed_block_rsv;
1225
1226         ret = __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1227
1228         btrfs_release_delayed_node(delayed_node);
1229         btrfs_free_path(path);
1230         trans->block_rsv = block_rsv;
1231
1232         return ret;
1233 }
1234
1235 int btrfs_commit_inode_delayed_inode(struct btrfs_inode *inode)
1236 {
1237         struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
1238         struct btrfs_trans_handle *trans;
1239         struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1240         struct btrfs_path *path;
1241         struct btrfs_block_rsv *block_rsv;
1242         int ret;
1243
1244         if (!delayed_node)
1245                 return 0;
1246
1247         mutex_lock(&delayed_node->mutex);
1248         if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1249                 mutex_unlock(&delayed_node->mutex);
1250                 btrfs_release_delayed_node(delayed_node);
1251                 return 0;
1252         }
1253         mutex_unlock(&delayed_node->mutex);
1254
1255         trans = btrfs_join_transaction(delayed_node->root);
1256         if (IS_ERR(trans)) {
1257                 ret = PTR_ERR(trans);
1258                 goto out;
1259         }
1260
1261         path = btrfs_alloc_path();
1262         if (!path) {
1263                 ret = -ENOMEM;
1264                 goto trans_out;
1265         }
1266         path->leave_spinning = 1;
1267
1268         block_rsv = trans->block_rsv;
1269         trans->block_rsv = &fs_info->delayed_block_rsv;
1270
1271         mutex_lock(&delayed_node->mutex);
1272         if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags))
1273                 ret = __btrfs_update_delayed_inode(trans, delayed_node->root,
1274                                                    path, delayed_node);
1275         else
1276                 ret = 0;
1277         mutex_unlock(&delayed_node->mutex);
1278
1279         btrfs_free_path(path);
1280         trans->block_rsv = block_rsv;
1281 trans_out:
1282         btrfs_end_transaction(trans);
1283         btrfs_btree_balance_dirty(fs_info);
1284 out:
1285         btrfs_release_delayed_node(delayed_node);
1286
1287         return ret;
1288 }
1289
1290 void btrfs_remove_delayed_node(struct btrfs_inode *inode)
1291 {
1292         struct btrfs_delayed_node *delayed_node;
1293
1294         delayed_node = READ_ONCE(inode->delayed_node);
1295         if (!delayed_node)
1296                 return;
1297
1298         inode->delayed_node = NULL;
1299         btrfs_release_delayed_node(delayed_node);
1300 }
1301
1302 struct btrfs_async_delayed_work {
1303         struct btrfs_delayed_root *delayed_root;
1304         int nr;
1305         struct btrfs_work work;
1306 };
1307
1308 static void btrfs_async_run_delayed_root(struct btrfs_work *work)
1309 {
1310         struct btrfs_async_delayed_work *async_work;
1311         struct btrfs_delayed_root *delayed_root;
1312         struct btrfs_trans_handle *trans;
1313         struct btrfs_path *path;
1314         struct btrfs_delayed_node *delayed_node = NULL;
1315         struct btrfs_root *root;
1316         struct btrfs_block_rsv *block_rsv;
1317         int total_done = 0;
1318
1319         async_work = container_of(work, struct btrfs_async_delayed_work, work);
1320         delayed_root = async_work->delayed_root;
1321
1322         path = btrfs_alloc_path();
1323         if (!path)
1324                 goto out;
1325
1326 again:
1327         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND / 2)
1328                 goto free_path;
1329
1330         delayed_node = btrfs_first_prepared_delayed_node(delayed_root);
1331         if (!delayed_node)
1332                 goto free_path;
1333
1334         path->leave_spinning = 1;
1335         root = delayed_node->root;
1336
1337         trans = btrfs_join_transaction(root);
1338         if (IS_ERR(trans))
1339                 goto release_path;
1340
1341         block_rsv = trans->block_rsv;
1342         trans->block_rsv = &root->fs_info->delayed_block_rsv;
1343
1344         __btrfs_commit_inode_delayed_items(trans, path, delayed_node);
1345
1346         trans->block_rsv = block_rsv;
1347         btrfs_end_transaction(trans);
1348         btrfs_btree_balance_dirty_nodelay(root->fs_info);
1349
1350 release_path:
1351         btrfs_release_path(path);
1352         total_done++;
1353
1354         btrfs_release_prepared_delayed_node(delayed_node);
1355         if ((async_work->nr == 0 && total_done < BTRFS_DELAYED_WRITEBACK) ||
1356             total_done < async_work->nr)
1357                 goto again;
1358
1359 free_path:
1360         btrfs_free_path(path);
1361 out:
1362         wake_up(&delayed_root->wait);
1363         kfree(async_work);
1364 }
1365
1366
1367 static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
1368                                      struct btrfs_fs_info *fs_info, int nr)
1369 {
1370         struct btrfs_async_delayed_work *async_work;
1371
1372         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND ||
1373             btrfs_workqueue_normal_congested(fs_info->delayed_workers))
1374                 return 0;
1375
1376         async_work = kmalloc(sizeof(*async_work), GFP_NOFS);
1377         if (!async_work)
1378                 return -ENOMEM;
1379
1380         async_work->delayed_root = delayed_root;
1381         btrfs_init_work(&async_work->work, btrfs_delayed_meta_helper,
1382                         btrfs_async_run_delayed_root, NULL, NULL);
1383         async_work->nr = nr;
1384
1385         btrfs_queue_work(fs_info->delayed_workers, &async_work->work);
1386         return 0;
1387 }
1388
1389 void btrfs_assert_delayed_root_empty(struct btrfs_fs_info *fs_info)
1390 {
1391         WARN_ON(btrfs_first_delayed_node(fs_info->delayed_root));
1392 }
1393
1394 static int could_end_wait(struct btrfs_delayed_root *delayed_root, int seq)
1395 {
1396         int val = atomic_read(&delayed_root->items_seq);
1397
1398         if (val < seq || val >= seq + BTRFS_DELAYED_BATCH)
1399                 return 1;
1400
1401         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1402                 return 1;
1403
1404         return 0;
1405 }
1406
1407 void btrfs_balance_delayed_items(struct btrfs_fs_info *fs_info)
1408 {
1409         struct btrfs_delayed_root *delayed_root = fs_info->delayed_root;
1410
1411         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1412                 return;
1413
1414         if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
1415                 int seq;
1416                 int ret;
1417
1418                 seq = atomic_read(&delayed_root->items_seq);
1419
1420                 ret = btrfs_wq_run_delayed_node(delayed_root, fs_info, 0);
1421                 if (ret)
1422                         return;
1423
1424                 wait_event_interruptible(delayed_root->wait,
1425                                          could_end_wait(delayed_root, seq));
1426                 return;
1427         }
1428
1429         btrfs_wq_run_delayed_node(delayed_root, fs_info, BTRFS_DELAYED_BATCH);
1430 }
1431
1432 /* Will return 0 or -ENOMEM */
1433 int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
1434                                    struct btrfs_fs_info *fs_info,
1435                                    const char *name, int name_len,
1436                                    struct btrfs_inode *dir,
1437                                    struct btrfs_disk_key *disk_key, u8 type,
1438                                    u64 index)
1439 {
1440         struct btrfs_delayed_node *delayed_node;
1441         struct btrfs_delayed_item *delayed_item;
1442         struct btrfs_dir_item *dir_item;
1443         int ret;
1444
1445         delayed_node = btrfs_get_or_create_delayed_node(dir);
1446         if (IS_ERR(delayed_node))
1447                 return PTR_ERR(delayed_node);
1448
1449         delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
1450         if (!delayed_item) {
1451                 ret = -ENOMEM;
1452                 goto release_node;
1453         }
1454
1455         delayed_item->key.objectid = btrfs_ino(dir);
1456         delayed_item->key.type = BTRFS_DIR_INDEX_KEY;
1457         delayed_item->key.offset = index;
1458
1459         dir_item = (struct btrfs_dir_item *)delayed_item->data;
1460         dir_item->location = *disk_key;
1461         btrfs_set_stack_dir_transid(dir_item, trans->transid);
1462         btrfs_set_stack_dir_data_len(dir_item, 0);
1463         btrfs_set_stack_dir_name_len(dir_item, name_len);
1464         btrfs_set_stack_dir_type(dir_item, type);
1465         memcpy((char *)(dir_item + 1), name, name_len);
1466
1467         ret = btrfs_delayed_item_reserve_metadata(trans, fs_info, delayed_item);
1468         /*
1469          * we have reserved enough space when we start a new transaction,
1470          * so reserving metadata failure is impossible
1471          */
1472         BUG_ON(ret);
1473
1474
1475         mutex_lock(&delayed_node->mutex);
1476         ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
1477         if (unlikely(ret)) {
1478                 btrfs_err(fs_info,
1479                           "err add delayed dir index item(name: %.*s) into the insertion tree of the delayed node(root id: %llu, inode id: %llu, errno: %d)",
1480                           name_len, name, delayed_node->root->objectid,
1481                           delayed_node->inode_id, ret);
1482                 BUG();
1483         }
1484         mutex_unlock(&delayed_node->mutex);
1485
1486 release_node:
1487         btrfs_release_delayed_node(delayed_node);
1488         return ret;
1489 }
1490
1491 static int btrfs_delete_delayed_insertion_item(struct btrfs_fs_info *fs_info,
1492                                                struct btrfs_delayed_node *node,
1493                                                struct btrfs_key *key)
1494 {
1495         struct btrfs_delayed_item *item;
1496
1497         mutex_lock(&node->mutex);
1498         item = __btrfs_lookup_delayed_insertion_item(node, key);
1499         if (!item) {
1500                 mutex_unlock(&node->mutex);
1501                 return 1;
1502         }
1503
1504         btrfs_delayed_item_release_metadata(fs_info, item);
1505         btrfs_release_delayed_item(item);
1506         mutex_unlock(&node->mutex);
1507         return 0;
1508 }
1509
1510 int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
1511                                    struct btrfs_fs_info *fs_info,
1512                                    struct btrfs_inode *dir, u64 index)
1513 {
1514         struct btrfs_delayed_node *node;
1515         struct btrfs_delayed_item *item;
1516         struct btrfs_key item_key;
1517         int ret;
1518
1519         node = btrfs_get_or_create_delayed_node(dir);
1520         if (IS_ERR(node))
1521                 return PTR_ERR(node);
1522
1523         item_key.objectid = btrfs_ino(dir);
1524         item_key.type = BTRFS_DIR_INDEX_KEY;
1525         item_key.offset = index;
1526
1527         ret = btrfs_delete_delayed_insertion_item(fs_info, node, &item_key);
1528         if (!ret)
1529                 goto end;
1530
1531         item = btrfs_alloc_delayed_item(0);
1532         if (!item) {
1533                 ret = -ENOMEM;
1534                 goto end;
1535         }
1536
1537         item->key = item_key;
1538
1539         ret = btrfs_delayed_item_reserve_metadata(trans, fs_info, item);
1540         /*
1541          * we have reserved enough space when we start a new transaction,
1542          * so reserving metadata failure is impossible.
1543          */
1544         BUG_ON(ret);
1545
1546         mutex_lock(&node->mutex);
1547         ret = __btrfs_add_delayed_deletion_item(node, item);
1548         if (unlikely(ret)) {
1549                 btrfs_err(fs_info,
1550                           "err add delayed dir index item(index: %llu) into the deletion tree of the delayed node(root id: %llu, inode id: %llu, errno: %d)",
1551                           index, node->root->objectid, node->inode_id, ret);
1552                 BUG();
1553         }
1554         mutex_unlock(&node->mutex);
1555 end:
1556         btrfs_release_delayed_node(node);
1557         return ret;
1558 }
1559
1560 int btrfs_inode_delayed_dir_index_count(struct btrfs_inode *inode)
1561 {
1562         struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1563
1564         if (!delayed_node)
1565                 return -ENOENT;
1566
1567         /*
1568          * Since we have held i_mutex of this directory, it is impossible that
1569          * a new directory index is added into the delayed node and index_cnt
1570          * is updated now. So we needn't lock the delayed node.
1571          */
1572         if (!delayed_node->index_cnt) {
1573                 btrfs_release_delayed_node(delayed_node);
1574                 return -EINVAL;
1575         }
1576
1577         inode->index_cnt = delayed_node->index_cnt;
1578         btrfs_release_delayed_node(delayed_node);
1579         return 0;
1580 }
1581
1582 bool btrfs_readdir_get_delayed_items(struct inode *inode,
1583                                      struct list_head *ins_list,
1584                                      struct list_head *del_list)
1585 {
1586         struct btrfs_delayed_node *delayed_node;
1587         struct btrfs_delayed_item *item;
1588
1589         delayed_node = btrfs_get_delayed_node(BTRFS_I(inode));
1590         if (!delayed_node)
1591                 return false;
1592
1593         /*
1594          * We can only do one readdir with delayed items at a time because of
1595          * item->readdir_list.
1596          */
1597         inode_unlock_shared(inode);
1598         inode_lock(inode);
1599
1600         mutex_lock(&delayed_node->mutex);
1601         item = __btrfs_first_delayed_insertion_item(delayed_node);
1602         while (item) {
1603                 atomic_inc(&item->refs);
1604                 list_add_tail(&item->readdir_list, ins_list);
1605                 item = __btrfs_next_delayed_item(item);
1606         }
1607
1608         item = __btrfs_first_delayed_deletion_item(delayed_node);
1609         while (item) {
1610                 atomic_inc(&item->refs);
1611                 list_add_tail(&item->readdir_list, del_list);
1612                 item = __btrfs_next_delayed_item(item);
1613         }
1614         mutex_unlock(&delayed_node->mutex);
1615         /*
1616          * This delayed node is still cached in the btrfs inode, so refs
1617          * must be > 1 now, and we needn't check it is going to be freed
1618          * or not.
1619          *
1620          * Besides that, this function is used to read dir, we do not
1621          * insert/delete delayed items in this period. So we also needn't
1622          * requeue or dequeue this delayed node.
1623          */
1624         atomic_dec(&delayed_node->refs);
1625
1626         return true;
1627 }
1628
1629 void btrfs_readdir_put_delayed_items(struct inode *inode,
1630                                      struct list_head *ins_list,
1631                                      struct list_head *del_list)
1632 {
1633         struct btrfs_delayed_item *curr, *next;
1634
1635         list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1636                 list_del(&curr->readdir_list);
1637                 if (atomic_dec_and_test(&curr->refs))
1638                         kfree(curr);
1639         }
1640
1641         list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1642                 list_del(&curr->readdir_list);
1643                 if (atomic_dec_and_test(&curr->refs))
1644                         kfree(curr);
1645         }
1646
1647         /*
1648          * The VFS is going to do up_read(), so we need to downgrade back to a
1649          * read lock.
1650          */
1651         downgrade_write(&inode->i_rwsem);
1652 }
1653
1654 int btrfs_should_delete_dir_index(struct list_head *del_list,
1655                                   u64 index)
1656 {
1657         struct btrfs_delayed_item *curr, *next;
1658         int ret;
1659
1660         if (list_empty(del_list))
1661                 return 0;
1662
1663         list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1664                 if (curr->key.offset > index)
1665                         break;
1666
1667                 list_del(&curr->readdir_list);
1668                 ret = (curr->key.offset == index);
1669
1670                 if (atomic_dec_and_test(&curr->refs))
1671                         kfree(curr);
1672
1673                 if (ret)
1674                         return 1;
1675                 else
1676                         continue;
1677         }
1678         return 0;
1679 }
1680
1681 /*
1682  * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
1683  *
1684  */
1685 int btrfs_readdir_delayed_dir_index(struct dir_context *ctx,
1686                                     struct list_head *ins_list)
1687 {
1688         struct btrfs_dir_item *di;
1689         struct btrfs_delayed_item *curr, *next;
1690         struct btrfs_key location;
1691         char *name;
1692         int name_len;
1693         int over = 0;
1694         unsigned char d_type;
1695
1696         if (list_empty(ins_list))
1697                 return 0;
1698
1699         /*
1700          * Changing the data of the delayed item is impossible. So
1701          * we needn't lock them. And we have held i_mutex of the
1702          * directory, nobody can delete any directory indexes now.
1703          */
1704         list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1705                 list_del(&curr->readdir_list);
1706
1707                 if (curr->key.offset < ctx->pos) {
1708                         if (atomic_dec_and_test(&curr->refs))
1709                                 kfree(curr);
1710                         continue;
1711                 }
1712
1713                 ctx->pos = curr->key.offset;
1714
1715                 di = (struct btrfs_dir_item *)curr->data;
1716                 name = (char *)(di + 1);
1717                 name_len = btrfs_stack_dir_name_len(di);
1718
1719                 d_type = btrfs_filetype_table[di->type];
1720                 btrfs_disk_key_to_cpu(&location, &di->location);
1721
1722                 over = !dir_emit(ctx, name, name_len,
1723                                location.objectid, d_type);
1724
1725                 if (atomic_dec_and_test(&curr->refs))
1726                         kfree(curr);
1727
1728                 if (over)
1729                         return 1;
1730         }
1731         return 0;
1732 }
1733
1734 static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
1735                                   struct btrfs_inode_item *inode_item,
1736                                   struct inode *inode)
1737 {
1738         btrfs_set_stack_inode_uid(inode_item, i_uid_read(inode));
1739         btrfs_set_stack_inode_gid(inode_item, i_gid_read(inode));
1740         btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
1741         btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
1742         btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
1743         btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
1744         btrfs_set_stack_inode_generation(inode_item,
1745                                          BTRFS_I(inode)->generation);
1746         btrfs_set_stack_inode_sequence(inode_item, inode->i_version);
1747         btrfs_set_stack_inode_transid(inode_item, trans->transid);
1748         btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
1749         btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
1750         btrfs_set_stack_inode_block_group(inode_item, 0);
1751
1752         btrfs_set_stack_timespec_sec(&inode_item->atime,
1753                                      inode->i_atime.tv_sec);
1754         btrfs_set_stack_timespec_nsec(&inode_item->atime,
1755                                       inode->i_atime.tv_nsec);
1756
1757         btrfs_set_stack_timespec_sec(&inode_item->mtime,
1758                                      inode->i_mtime.tv_sec);
1759         btrfs_set_stack_timespec_nsec(&inode_item->mtime,
1760                                       inode->i_mtime.tv_nsec);
1761
1762         btrfs_set_stack_timespec_sec(&inode_item->ctime,
1763                                      inode->i_ctime.tv_sec);
1764         btrfs_set_stack_timespec_nsec(&inode_item->ctime,
1765                                       inode->i_ctime.tv_nsec);
1766
1767         btrfs_set_stack_timespec_sec(&inode_item->otime,
1768                                      BTRFS_I(inode)->i_otime.tv_sec);
1769         btrfs_set_stack_timespec_nsec(&inode_item->otime,
1770                                      BTRFS_I(inode)->i_otime.tv_nsec);
1771 }
1772
1773 int btrfs_fill_inode(struct inode *inode, u32 *rdev)
1774 {
1775         struct btrfs_delayed_node *delayed_node;
1776         struct btrfs_inode_item *inode_item;
1777
1778         delayed_node = btrfs_get_delayed_node(BTRFS_I(inode));
1779         if (!delayed_node)
1780                 return -ENOENT;
1781
1782         mutex_lock(&delayed_node->mutex);
1783         if (!test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1784                 mutex_unlock(&delayed_node->mutex);
1785                 btrfs_release_delayed_node(delayed_node);
1786                 return -ENOENT;
1787         }
1788
1789         inode_item = &delayed_node->inode_item;
1790
1791         i_uid_write(inode, btrfs_stack_inode_uid(inode_item));
1792         i_gid_write(inode, btrfs_stack_inode_gid(inode_item));
1793         btrfs_i_size_write(BTRFS_I(inode), btrfs_stack_inode_size(inode_item));
1794         inode->i_mode = btrfs_stack_inode_mode(inode_item);
1795         set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
1796         inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
1797         BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
1798         BTRFS_I(inode)->last_trans = btrfs_stack_inode_transid(inode_item);
1799
1800         inode->i_version = btrfs_stack_inode_sequence(inode_item);
1801         inode->i_rdev = 0;
1802         *rdev = btrfs_stack_inode_rdev(inode_item);
1803         BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
1804
1805         inode->i_atime.tv_sec = btrfs_stack_timespec_sec(&inode_item->atime);
1806         inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->atime);
1807
1808         inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(&inode_item->mtime);
1809         inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->mtime);
1810
1811         inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(&inode_item->ctime);
1812         inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(&inode_item->ctime);
1813
1814         BTRFS_I(inode)->i_otime.tv_sec =
1815                 btrfs_stack_timespec_sec(&inode_item->otime);
1816         BTRFS_I(inode)->i_otime.tv_nsec =
1817                 btrfs_stack_timespec_nsec(&inode_item->otime);
1818
1819         inode->i_generation = BTRFS_I(inode)->generation;
1820         BTRFS_I(inode)->index_cnt = (u64)-1;
1821
1822         mutex_unlock(&delayed_node->mutex);
1823         btrfs_release_delayed_node(delayed_node);
1824         return 0;
1825 }
1826
1827 int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
1828                                struct btrfs_root *root, struct inode *inode)
1829 {
1830         struct btrfs_delayed_node *delayed_node;
1831         int ret = 0;
1832
1833         delayed_node = btrfs_get_or_create_delayed_node(BTRFS_I(inode));
1834         if (IS_ERR(delayed_node))
1835                 return PTR_ERR(delayed_node);
1836
1837         mutex_lock(&delayed_node->mutex);
1838         if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1839                 fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1840                 goto release_node;
1841         }
1842
1843         ret = btrfs_delayed_inode_reserve_metadata(trans, root, BTRFS_I(inode),
1844                                                    delayed_node);
1845         if (ret)
1846                 goto release_node;
1847
1848         fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1849         set_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags);
1850         delayed_node->count++;
1851         atomic_inc(&root->fs_info->delayed_root->items);
1852 release_node:
1853         mutex_unlock(&delayed_node->mutex);
1854         btrfs_release_delayed_node(delayed_node);
1855         return ret;
1856 }
1857
1858 int btrfs_delayed_delete_inode_ref(struct btrfs_inode *inode)
1859 {
1860         struct btrfs_fs_info *fs_info = btrfs_sb(inode->vfs_inode.i_sb);
1861         struct btrfs_delayed_node *delayed_node;
1862
1863         /*
1864          * we don't do delayed inode updates during log recovery because it
1865          * leads to enospc problems.  This means we also can't do
1866          * delayed inode refs
1867          */
1868         if (test_bit(BTRFS_FS_LOG_RECOVERING, &fs_info->flags))
1869                 return -EAGAIN;
1870
1871         delayed_node = btrfs_get_or_create_delayed_node(inode);
1872         if (IS_ERR(delayed_node))
1873                 return PTR_ERR(delayed_node);
1874
1875         /*
1876          * We don't reserve space for inode ref deletion is because:
1877          * - We ONLY do async inode ref deletion for the inode who has only
1878          *   one link(i_nlink == 1), it means there is only one inode ref.
1879          *   And in most case, the inode ref and the inode item are in the
1880          *   same leaf, and we will deal with them at the same time.
1881          *   Since we are sure we will reserve the space for the inode item,
1882          *   it is unnecessary to reserve space for inode ref deletion.
1883          * - If the inode ref and the inode item are not in the same leaf,
1884          *   We also needn't worry about enospc problem, because we reserve
1885          *   much more space for the inode update than it needs.
1886          * - At the worst, we can steal some space from the global reservation.
1887          *   It is very rare.
1888          */
1889         mutex_lock(&delayed_node->mutex);
1890         if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1891                 goto release_node;
1892
1893         set_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags);
1894         delayed_node->count++;
1895         atomic_inc(&fs_info->delayed_root->items);
1896 release_node:
1897         mutex_unlock(&delayed_node->mutex);
1898         btrfs_release_delayed_node(delayed_node);
1899         return 0;
1900 }
1901
1902 static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
1903 {
1904         struct btrfs_root *root = delayed_node->root;
1905         struct btrfs_fs_info *fs_info = root->fs_info;
1906         struct btrfs_delayed_item *curr_item, *prev_item;
1907
1908         mutex_lock(&delayed_node->mutex);
1909         curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
1910         while (curr_item) {
1911                 btrfs_delayed_item_release_metadata(fs_info, curr_item);
1912                 prev_item = curr_item;
1913                 curr_item = __btrfs_next_delayed_item(prev_item);
1914                 btrfs_release_delayed_item(prev_item);
1915         }
1916
1917         curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
1918         while (curr_item) {
1919                 btrfs_delayed_item_release_metadata(fs_info, curr_item);
1920                 prev_item = curr_item;
1921                 curr_item = __btrfs_next_delayed_item(prev_item);
1922                 btrfs_release_delayed_item(prev_item);
1923         }
1924
1925         if (test_bit(BTRFS_DELAYED_NODE_DEL_IREF, &delayed_node->flags))
1926                 btrfs_release_delayed_iref(delayed_node);
1927
1928         if (test_bit(BTRFS_DELAYED_NODE_INODE_DIRTY, &delayed_node->flags)) {
1929                 btrfs_delayed_inode_release_metadata(fs_info, delayed_node);
1930                 btrfs_release_delayed_inode(delayed_node);
1931         }
1932         mutex_unlock(&delayed_node->mutex);
1933 }
1934
1935 void btrfs_kill_delayed_inode_items(struct btrfs_inode *inode)
1936 {
1937         struct btrfs_delayed_node *delayed_node;
1938
1939         delayed_node = btrfs_get_delayed_node(inode);
1940         if (!delayed_node)
1941                 return;
1942
1943         __btrfs_kill_delayed_node(delayed_node);
1944         btrfs_release_delayed_node(delayed_node);
1945 }
1946
1947 void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
1948 {
1949         u64 inode_id = 0;
1950         struct btrfs_delayed_node *delayed_nodes[8];
1951         int i, n;
1952
1953         while (1) {
1954                 spin_lock(&root->inode_lock);
1955                 n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
1956                                            (void **)delayed_nodes, inode_id,
1957                                            ARRAY_SIZE(delayed_nodes));
1958                 if (!n) {
1959                         spin_unlock(&root->inode_lock);
1960                         break;
1961                 }
1962
1963                 inode_id = delayed_nodes[n - 1]->inode_id + 1;
1964
1965                 for (i = 0; i < n; i++)
1966                         atomic_inc(&delayed_nodes[i]->refs);
1967                 spin_unlock(&root->inode_lock);
1968
1969                 for (i = 0; i < n; i++) {
1970                         __btrfs_kill_delayed_node(delayed_nodes[i]);
1971                         btrfs_release_delayed_node(delayed_nodes[i]);
1972                 }
1973         }
1974 }
1975
1976 void btrfs_destroy_delayed_inodes(struct btrfs_fs_info *fs_info)
1977 {
1978         struct btrfs_delayed_node *curr_node, *prev_node;
1979
1980         curr_node = btrfs_first_delayed_node(fs_info->delayed_root);
1981         while (curr_node) {
1982                 __btrfs_kill_delayed_node(curr_node);
1983
1984                 prev_node = curr_node;
1985                 curr_node = btrfs_next_delayed_node(curr_node);
1986                 btrfs_release_delayed_node(prev_node);
1987         }
1988 }
1989