fs/btrfs/delayed-inode.c

   1 /*
   2  * Copyright (C) 2011 Fujitsu.  All rights reserved.
   3  * Written by Miao Xie <miaox@cn.fujitsu.com>
   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
  25 #define BTRFS_DELAYED_WRITEBACK         400
  26 #define BTRFS_DELAYED_BACKGROUND        100
  27
  28 static struct kmem_cache *delayed_node_cache;
  29
  30 int __init btrfs_delayed_inode_init(void)
  31 {
  32         delayed_node_cache = kmem_cache_create("delayed_node",
  33                                         sizeof(struct btrfs_delayed_node),
  34                                         0,
  35                                         SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
  36                                         NULL);
  37         if (!delayed_node_cache)
  38                 return -ENOMEM;
  39         return 0;
  40 }
  41
  42 void btrfs_delayed_inode_exit(void)
  43 {
  44         if (delayed_node_cache)
  45                 kmem_cache_destroy(delayed_node_cache);
  46 }
  47
  48 static inline void btrfs_init_delayed_node(
  49                                 struct btrfs_delayed_node *delayed_node,
  50                                 struct btrfs_root *root, u64 inode_id)
  51 {
  52         delayed_node->root = root;
  53         delayed_node->inode_id = inode_id;
  54         atomic_set(&delayed_node->refs, 0);
  55         delayed_node->count = 0;
  56         delayed_node->in_list = 0;
  57         delayed_node->inode_dirty = 0;
  58         delayed_node->ins_root = RB_ROOT;
  59         delayed_node->del_root = RB_ROOT;
  60         mutex_init(&delayed_node->mutex);
  61         delayed_node->index_cnt = 0;
  62         INIT_LIST_HEAD(&delayed_node->n_list);
  63         INIT_LIST_HEAD(&delayed_node->p_list);
  64         delayed_node->bytes_reserved = 0;
  65 }
  66
  67 static inline int btrfs_is_continuous_delayed_item(
  68                                         struct btrfs_delayed_item *item1,
  69                                         struct btrfs_delayed_item *item2)
  70 {
  71         if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
  72             item1->key.objectid == item2->key.objectid &&
  73             item1->key.type == item2->key.type &&
  74             item1->key.offset + 1 == item2->key.offset)
  75                 return 1;
  76         return 0;
  77 }
  78
  79 static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
  80                                                         struct btrfs_root *root)
  81 {
  82         return root->fs_info->delayed_root;
  83 }
  84
  85 static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
  86                                                         struct inode *inode)
  87 {
  88         struct btrfs_delayed_node *node;
  89         struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
  90         struct btrfs_root *root = btrfs_inode->root;
  91         int ret;
  92
  93 again:
  94         node = ACCESS_ONCE(btrfs_inode->delayed_node);
  95         if (node) {
  96                 atomic_inc(&node->refs);        /* can be accessed */
  97                 return node;
  98         }
  99
 100         spin_lock(&root->inode_lock);
 101         node = radix_tree_lookup(&root->delayed_nodes_tree, inode->i_ino);
 102         if (node) {
 103                 if (btrfs_inode->delayed_node) {
 104                         spin_unlock(&root->inode_lock);
 105                         goto again;
 106                 }
 107                 btrfs_inode->delayed_node = node;
 108                 atomic_inc(&node->refs);        /* can be accessed */
 109                 atomic_inc(&node->refs);        /* cached in the inode */
 110                 spin_unlock(&root->inode_lock);
 111                 return node;
 112         }
 113         spin_unlock(&root->inode_lock);
 114
 115         node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
 116         if (!node)
 117                 return ERR_PTR(-ENOMEM);
 118         btrfs_init_delayed_node(node, root, inode->i_ino);
 119
 120         atomic_inc(&node->refs);        /* cached in the btrfs inode */
 121         atomic_inc(&node->refs);        /* can be accessed */
 122
 123         ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
 124         if (ret) {
 125                 kmem_cache_free(delayed_node_cache, node);
 126                 return ERR_PTR(ret);
 127         }
 128
 129         spin_lock(&root->inode_lock);
 130         ret = radix_tree_insert(&root->delayed_nodes_tree, inode->i_ino, node);
 131         if (ret == -EEXIST) {
 132                 kmem_cache_free(delayed_node_cache, node);
 133                 spin_unlock(&root->inode_lock);
 134                 radix_tree_preload_end();
 135                 goto again;
 136         }
 137         btrfs_inode->delayed_node = node;
 138         spin_unlock(&root->inode_lock);
 139         radix_tree_preload_end();
 140
 141         return node;
 142 }
 143
 144 /*
 145  * Call it when holding delayed_node->mutex
 146  *
 147  * If mod = 1, add this node into the prepared list.
 148  */
 149 static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
 150                                      struct btrfs_delayed_node *node,
 151                                      int mod)
 152 {
 153         spin_lock(&root->lock);
 154         if (node->in_list) {
 155                 if (!list_empty(&node->p_list))
 156                         list_move_tail(&node->p_list, &root->prepare_list);
 157                 else if (mod)
 158                         list_add_tail(&node->p_list, &root->prepare_list);
 159         } else {
 160                 list_add_tail(&node->n_list, &root->node_list);
 161                 list_add_tail(&node->p_list, &root->prepare_list);
 162                 atomic_inc(&node->refs);        /* inserted into list */
 163                 root->nodes++;
 164                 node->in_list = 1;
 165         }
 166         spin_unlock(&root->lock);
 167 }
 168
 169 /* Call it when holding delayed_node->mutex */
 170 static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
 171                                        struct btrfs_delayed_node *node)
 172 {
 173         spin_lock(&root->lock);
 174         if (node->in_list) {
 175                 root->nodes--;
 176                 atomic_dec(&node->refs);        /* not in the list */
 177                 list_del_init(&node->n_list);
 178                 if (!list_empty(&node->p_list))
 179                         list_del_init(&node->p_list);
 180                 node->in_list = 0;
 181         }
 182         spin_unlock(&root->lock);
 183 }
 184
 185 struct btrfs_delayed_node *btrfs_first_delayed_node(
 186                         struct btrfs_delayed_root *delayed_root)
 187 {
 188         struct list_head *p;
 189         struct btrfs_delayed_node *node = NULL;
 190
 191         spin_lock(&delayed_root->lock);
 192         if (list_empty(&delayed_root->node_list))
 193                 goto out;
 194
 195         p = delayed_root->node_list.next;
 196         node = list_entry(p, struct btrfs_delayed_node, n_list);
 197         atomic_inc(&node->refs);
 198 out:
 199         spin_unlock(&delayed_root->lock);
 200
 201         return node;
 202 }
 203
 204 struct btrfs_delayed_node *btrfs_next_delayed_node(
 205                                                 struct btrfs_delayed_node *node)
 206 {
 207         struct btrfs_delayed_root *delayed_root;
 208         struct list_head *p;
 209         struct btrfs_delayed_node *next = NULL;
 210
 211         delayed_root = node->root->fs_info->delayed_root;
 212         spin_lock(&delayed_root->lock);
 213         if (!node->in_list) {   /* not in the list */
 214                 if (list_empty(&delayed_root->node_list))
 215                         goto out;
 216                 p = delayed_root->node_list.next;
 217         } else if (list_is_last(&node->n_list, &delayed_root->node_list))
 218                 goto out;
 219         else
 220                 p = node->n_list.next;
 221
 222         next = list_entry(p, struct btrfs_delayed_node, n_list);
 223         atomic_inc(&next->refs);
 224 out:
 225         spin_unlock(&delayed_root->lock);
 226
 227         return next;
 228 }
 229
 230 static void __btrfs_release_delayed_node(
 231                                 struct btrfs_delayed_node *delayed_node,
 232                                 int mod)
 233 {
 234         struct btrfs_delayed_root *delayed_root;
 235
 236         if (!delayed_node)
 237                 return;
 238
 239         delayed_root = delayed_node->root->fs_info->delayed_root;
 240
 241         mutex_lock(&delayed_node->mutex);
 242         if (delayed_node->count)
 243                 btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
 244         else
 245                 btrfs_dequeue_delayed_node(delayed_root, delayed_node);
 246         mutex_unlock(&delayed_node->mutex);
 247
 248         if (atomic_dec_and_test(&delayed_node->refs)) {
 249                 struct btrfs_root *root = delayed_node->root;
 250                 spin_lock(&root->inode_lock);
 251                 if (atomic_read(&delayed_node->refs) == 0) {
 252                         radix_tree_delete(&root->delayed_nodes_tree,
 253                                           delayed_node->inode_id);
 254                         kmem_cache_free(delayed_node_cache, delayed_node);
 255                 }
 256                 spin_unlock(&root->inode_lock);
 257         }
 258 }
 259
 260 static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
 261 {
 262         __btrfs_release_delayed_node(node, 0);
 263 }
 264
 265 struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
 266                                         struct btrfs_delayed_root *delayed_root)
 267 {
 268         struct list_head *p;
 269         struct btrfs_delayed_node *node = NULL;
 270
 271         spin_lock(&delayed_root->lock);
 272         if (list_empty(&delayed_root->prepare_list))
 273                 goto out;
 274
 275         p = delayed_root->prepare_list.next;
 276         list_del_init(p);
 277         node = list_entry(p, struct btrfs_delayed_node, p_list);
 278         atomic_inc(&node->refs);
 279 out:
 280         spin_unlock(&delayed_root->lock);
 281
 282         return node;
 283 }
 284
 285 static inline void btrfs_release_prepared_delayed_node(
 286                                         struct btrfs_delayed_node *node)
 287 {
 288         __btrfs_release_delayed_node(node, 1);
 289 }
 290
 291 struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
 292 {
 293         struct btrfs_delayed_item *item;
 294         item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
 295         if (item) {
 296                 item->data_len = data_len;
 297                 item->ins_or_del = 0;
 298                 item->bytes_reserved = 0;
 299                 item->block_rsv = NULL;
 300                 item->delayed_node = NULL;
 301                 atomic_set(&item->refs, 1);
 302         }
 303         return item;
 304 }
 305
 306 /*
 307  * __btrfs_lookup_delayed_item - look up the delayed item by key
 308  * @delayed_node: pointer to the delayed node
 309  * @key:          the key to look up
 310  * @prev:         used to store the prev item if the right item isn't found
 311  * @next:         used to store the next item if the right item isn't found
 312  *
 313  * Note: if we don't find the right item, we will return the prev item and
 314  * the next item.
 315  */
 316 static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
 317                                 struct rb_root *root,
 318                                 struct btrfs_key *key,
 319                                 struct btrfs_delayed_item **prev,
 320                                 struct btrfs_delayed_item **next)
 321 {
 322         struct rb_node *node, *prev_node = NULL;
 323         struct btrfs_delayed_item *delayed_item = NULL;
 324         int ret = 0;
 325
 326         node = root->rb_node;
 327
 328         while (node) {
 329                 delayed_item = rb_entry(node, struct btrfs_delayed_item,
 330                                         rb_node);
 331                 prev_node = node;
 332                 ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
 333                 if (ret < 0)
 334                         node = node->rb_right;
 335                 else if (ret > 0)
 336                         node = node->rb_left;
 337                 else
 338                         return delayed_item;
 339         }
 340
 341         if (prev) {
 342                 if (!prev_node)
 343                         *prev = NULL;
 344                 else if (ret < 0)
 345                         *prev = delayed_item;
 346                 else if ((node = rb_prev(prev_node)) != NULL) {
 347                         *prev = rb_entry(node, struct btrfs_delayed_item,
 348                                          rb_node);
 349                 } else
 350                         *prev = NULL;
 351         }
 352
 353         if (next) {
 354                 if (!prev_node)
 355                         *next = NULL;
 356                 else if (ret > 0)
 357                         *next = delayed_item;
 358                 else if ((node = rb_next(prev_node)) != NULL) {
 359                         *next = rb_entry(node, struct btrfs_delayed_item,
 360                                          rb_node);
 361                 } else
 362                         *next = NULL;
 363         }
 364         return NULL;
 365 }
 366
 367 struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
 368                                         struct btrfs_delayed_node *delayed_node,
 369                                         struct btrfs_key *key)
 370 {
 371         struct btrfs_delayed_item *item;
 372
 373         item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
 374                                            NULL, NULL);
 375         return item;
 376 }
 377
 378 struct btrfs_delayed_item *__btrfs_lookup_delayed_deletion_item(
 379                                         struct btrfs_delayed_node *delayed_node,
 380                                         struct btrfs_key *key)
 381 {
 382         struct btrfs_delayed_item *item;
 383
 384         item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
 385                                            NULL, NULL);
 386         return item;
 387 }
 388
 389 struct btrfs_delayed_item *__btrfs_search_delayed_insertion_item(
 390                                         struct btrfs_delayed_node *delayed_node,
 391                                         struct btrfs_key *key)
 392 {
 393         struct btrfs_delayed_item *item, *next;
 394
 395         item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
 396                                            NULL, &next);
 397         if (!item)
 398                 item = next;
 399
 400         return item;
 401 }
 402
 403 struct btrfs_delayed_item *__btrfs_search_delayed_deletion_item(
 404                                         struct btrfs_delayed_node *delayed_node,
 405                                         struct btrfs_key *key)
 406 {
 407         struct btrfs_delayed_item *item, *next;
 408
 409         item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
 410                                            NULL, &next);
 411         if (!item)
 412                 item = next;
 413
 414         return item;
 415 }
 416
 417 static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
 418                                     struct btrfs_delayed_item *ins,
 419                                     int action)
 420 {
 421         struct rb_node **p, *node;
 422         struct rb_node *parent_node = NULL;
 423         struct rb_root *root;
 424         struct btrfs_delayed_item *item;
 425         int cmp;
 426
 427         if (action == BTRFS_DELAYED_INSERTION_ITEM)
 428                 root = &delayed_node->ins_root;
 429         else if (action == BTRFS_DELAYED_DELETION_ITEM)
 430                 root = &delayed_node->del_root;
 431         else
 432                 BUG();
 433         p = &root->rb_node;
 434         node = &ins->rb_node;
 435
 436         while (*p) {
 437                 parent_node = *p;
 438                 item = rb_entry(parent_node, struct btrfs_delayed_item,
 439                                  rb_node);
 440
 441                 cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
 442                 if (cmp < 0)
 443                         p = &(*p)->rb_right;
 444                 else if (cmp > 0)
 445                         p = &(*p)->rb_left;
 446                 else
 447                         return -EEXIST;
 448         }
 449
 450         rb_link_node(node, parent_node, p);
 451         rb_insert_color(node, root);
 452         ins->delayed_node = delayed_node;
 453         ins->ins_or_del = action;
 454
 455         if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
 456             action == BTRFS_DELAYED_INSERTION_ITEM &&
 457             ins->key.offset >= delayed_node->index_cnt)
 458                         delayed_node->index_cnt = ins->key.offset + 1;
 459
 460         delayed_node->count++;
 461         atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
 462         return 0;
 463 }
 464
 465 static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
 466                                               struct btrfs_delayed_item *item)
 467 {
 468         return __btrfs_add_delayed_item(node, item,
 469                                         BTRFS_DELAYED_INSERTION_ITEM);
 470 }
 471
 472 static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
 473                                              struct btrfs_delayed_item *item)
 474 {
 475         return __btrfs_add_delayed_item(node, item,
 476                                         BTRFS_DELAYED_DELETION_ITEM);
 477 }
 478
 479 static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
 480 {
 481         struct rb_root *root;
 482         struct btrfs_delayed_root *delayed_root;
 483
 484         delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
 485
 486         BUG_ON(!delayed_root);
 487         BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
 488                delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
 489
 490         if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
 491                 root = &delayed_item->delayed_node->ins_root;
 492         else
 493                 root = &delayed_item->delayed_node->del_root;
 494
 495         rb_erase(&delayed_item->rb_node, root);
 496         delayed_item->delayed_node->count--;
 497         atomic_dec(&delayed_root->items);
 498         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND &&
 499             waitqueue_active(&delayed_root->wait))
 500                 wake_up(&delayed_root->wait);
 501 }
 502
 503 static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
 504 {
 505         if (item) {
 506                 __btrfs_remove_delayed_item(item);
 507                 if (atomic_dec_and_test(&item->refs))
 508                         kfree(item);
 509         }
 510 }
 511
 512 struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
 513                                         struct btrfs_delayed_node *delayed_node)
 514 {
 515         struct rb_node *p;
 516         struct btrfs_delayed_item *item = NULL;
 517
 518         p = rb_first(&delayed_node->ins_root);
 519         if (p)
 520                 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
 521
 522         return item;
 523 }
 524
 525 struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
 526                                         struct btrfs_delayed_node *delayed_node)
 527 {
 528         struct rb_node *p;
 529         struct btrfs_delayed_item *item = NULL;
 530
 531         p = rb_first(&delayed_node->del_root);
 532         if (p)
 533                 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
 534
 535         return item;
 536 }
 537
 538 struct btrfs_delayed_item *__btrfs_next_delayed_item(
 539                                                 struct btrfs_delayed_item *item)
 540 {
 541         struct rb_node *p;
 542         struct btrfs_delayed_item *next = NULL;
 543
 544         p = rb_next(&item->rb_node);
 545         if (p)
 546                 next = rb_entry(p, struct btrfs_delayed_item, rb_node);
 547
 548         return next;
 549 }
 550
 551 static inline struct btrfs_delayed_node *btrfs_get_delayed_node(
 552                                                         struct inode *inode)
 553 {
 554         struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
 555         struct btrfs_delayed_node *delayed_node;
 556
 557         delayed_node = btrfs_inode->delayed_node;
 558         if (delayed_node)
 559                 atomic_inc(&delayed_node->refs);
 560
 561         return delayed_node;
 562 }
 563
 564 static inline struct btrfs_root *btrfs_get_fs_root(struct btrfs_root *root,
 565                                                    u64 root_id)
 566 {
 567         struct btrfs_key root_key;
 568
 569         if (root->objectid == root_id)
 570                 return root;
 571
 572         root_key.objectid = root_id;
 573         root_key.type = BTRFS_ROOT_ITEM_KEY;
 574         root_key.offset = (u64)-1;
 575         return btrfs_read_fs_root_no_name(root->fs_info, &root_key);
 576 }
 577
 578 static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
 579                                                struct btrfs_root *root,
 580                                                struct btrfs_delayed_item *item)
 581 {
 582         struct btrfs_block_rsv *src_rsv;
 583         struct btrfs_block_rsv *dst_rsv;
 584         u64 num_bytes;
 585         int ret;
 586
 587         if (!trans->bytes_reserved)
 588                 return 0;
 589
 590         src_rsv = trans->block_rsv;
 591         dst_rsv = &root->fs_info->global_block_rsv;
 592
 593         num_bytes = btrfs_calc_trans_metadata_size(root, 1);
 594         ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
 595         if (!ret) {
 596                 item->bytes_reserved = num_bytes;
 597                 item->block_rsv = dst_rsv;
 598         }
 599
 600         return ret;
 601 }
 602
 603 static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
 604                                                 struct btrfs_delayed_item *item)
 605 {
 606         if (!item->bytes_reserved)
 607                 return;
 608
 609         btrfs_block_rsv_release(root, item->block_rsv,
 610                                 item->bytes_reserved);
 611 }
 612
 613 static int btrfs_delayed_inode_reserve_metadata(
 614                                         struct btrfs_trans_handle *trans,
 615                                         struct btrfs_root *root,
 616                                         struct btrfs_delayed_node *node)
 617 {
 618         struct btrfs_block_rsv *src_rsv;
 619         struct btrfs_block_rsv *dst_rsv;
 620         u64 num_bytes;
 621         int ret;
 622
 623         if (!trans->bytes_reserved)
 624                 return 0;
 625
 626         src_rsv = trans->block_rsv;
 627         dst_rsv = &root->fs_info->global_block_rsv;
 628
 629         num_bytes = btrfs_calc_trans_metadata_size(root, 1);
 630         ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
 631         if (!ret)
 632                 node->bytes_reserved = num_bytes;
 633
 634         return ret;
 635 }
 636
 637 static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
 638                                                 struct btrfs_delayed_node *node)
 639 {
 640         struct btrfs_block_rsv *rsv;
 641
 642         if (!node->bytes_reserved)
 643                 return;
 644
 645         rsv = &root->fs_info->global_block_rsv;
 646         btrfs_block_rsv_release(root, rsv,
 647                                 node->bytes_reserved);
 648         node->bytes_reserved = 0;
 649 }
 650
 651 /*
 652  * This helper will insert some continuous items into the same leaf according
 653  * to the free space of the leaf.
 654  */
 655 static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans,
 656                                 struct btrfs_root *root,
 657                                 struct btrfs_path *path,
 658                                 struct btrfs_delayed_item *item)
 659 {
 660         struct btrfs_delayed_item *curr, *next;
 661         int free_space;
 662         int total_data_size = 0, total_size = 0;
 663         struct extent_buffer *leaf;
 664         char *data_ptr;
 665         struct btrfs_key *keys;
 666         u32 *data_size;
 667         struct list_head head;
 668         int slot;
 669         int nitems;
 670         int i;
 671         int ret = 0;
 672
 673         BUG_ON(!path->nodes[0]);
 674
 675         leaf = path->nodes[0];
 676         free_space = btrfs_leaf_free_space(root, leaf);
 677         INIT_LIST_HEAD(&head);
 678
 679         next = item;
 680
 681         /*
 682          * count the number of the continuous items that we can insert in batch
 683          */
 684         while (total_size + next->data_len + sizeof(struct btrfs_item) <=
 685                free_space) {
 686                 total_data_size += next->data_len;
 687                 total_size += next->data_len + sizeof(struct btrfs_item);
 688                 list_add_tail(&next->tree_list, &head);
 689                 nitems++;
 690
 691                 curr = next;
 692                 next = __btrfs_next_delayed_item(curr);
 693                 if (!next)
 694                         break;
 695
 696                 if (!btrfs_is_continuous_delayed_item(curr, next))
 697                         break;
 698         }
 699
 700         if (!nitems) {
 701                 ret = 0;
 702                 goto out;
 703         }
 704
 705         /*
 706          * we need allocate some memory space, but it might cause the task
 707          * to sleep, so we set all locked nodes in the path to blocking locks
 708          * first.
 709          */
 710         btrfs_set_path_blocking(path);
 711
 712         keys = kmalloc(sizeof(struct btrfs_key) * nitems, GFP_NOFS);
 713         if (!keys) {
 714                 ret = -ENOMEM;
 715                 goto out;
 716         }
 717
 718         data_size = kmalloc(sizeof(u32) * nitems, GFP_NOFS);
 719         if (!data_size) {
 720                 ret = -ENOMEM;
 721                 goto error;
 722         }
 723
 724         /* get keys of all the delayed items */
 725         i = 0;
 726         list_for_each_entry(next, &head, tree_list) {
 727                 keys[i] = next->key;
 728                 data_size[i] = next->data_len;
 729                 i++;
 730         }
 731
 732         /* reset all the locked nodes in the patch to spinning locks. */
 733         btrfs_clear_path_blocking(path, NULL);
 734
 735         /* insert the keys of the items */
 736         ret = setup_items_for_insert(trans, root, path, keys, data_size,
 737                                      total_data_size, total_size, nitems);
 738         if (ret)
 739                 goto error;
 740
 741         /* insert the dir index items */
 742         slot = path->slots[0];
 743         list_for_each_entry_safe(curr, next, &head, tree_list) {
 744                 data_ptr = btrfs_item_ptr(leaf, slot, char);
 745                 write_extent_buffer(leaf, &curr->data,
 746                                     (unsigned long)data_ptr,
 747                                     curr->data_len);
 748                 slot++;
 749
 750                 btrfs_delayed_item_release_metadata(root, curr);
 751
 752                 list_del(&curr->tree_list);
 753                 btrfs_release_delayed_item(curr);
 754         }
 755
 756 error:
 757         kfree(data_size);
 758         kfree(keys);
 759 out:
 760         return ret;
 761 }
 762
 763 /*
 764  * This helper can just do simple insertion that needn't extend item for new
 765  * data, such as directory name index insertion, inode insertion.
 766  */
 767 static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
 768                                      struct btrfs_root *root,
 769                                      struct btrfs_path *path,
 770                                      struct btrfs_delayed_item *delayed_item)
 771 {
 772         struct extent_buffer *leaf;
 773         struct btrfs_item *item;
 774         char *ptr;
 775         int ret;
 776
 777         ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
 778                                       delayed_item->data_len);
 779         if (ret < 0 && ret != -EEXIST)
 780                 return ret;
 781
 782         leaf = path->nodes[0];
 783
 784         item = btrfs_item_nr(leaf, path->slots[0]);
 785         ptr = btrfs_item_ptr(leaf, path->slots[0], char);
 786
 787         write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
 788                             delayed_item->data_len);
 789         btrfs_mark_buffer_dirty(leaf);
 790
 791         btrfs_delayed_item_release_metadata(root, delayed_item);
 792         return 0;
 793 }
 794
 795 /*
 796  * we insert an item first, then if there are some continuous items, we try
 797  * to insert those items into the same leaf.
 798  */
 799 static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
 800                                       struct btrfs_path *path,
 801                                       struct btrfs_root *root,
 802                                       struct btrfs_delayed_node *node)
 803 {
 804         struct btrfs_delayed_item *curr, *prev;
 805         int ret = 0;
 806
 807 do_again:
 808         mutex_lock(&node->mutex);
 809         curr = __btrfs_first_delayed_insertion_item(node);
 810         if (!curr)
 811                 goto insert_end;
 812
 813         ret = btrfs_insert_delayed_item(trans, root, path, curr);
 814         if (ret < 0) {
 815                 btrfs_release_path(root, path);
 816                 goto insert_end;
 817         }
 818
 819         prev = curr;
 820         curr = __btrfs_next_delayed_item(prev);
 821         if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
 822                 /* insert the continuous items into the same leaf */
 823                 path->slots[0]++;
 824                 btrfs_batch_insert_items(trans, root, path, curr);
 825         }
 826         btrfs_release_delayed_item(prev);
 827         btrfs_mark_buffer_dirty(path->nodes[0]);
 828
 829         btrfs_release_path(root, path);
 830         mutex_unlock(&node->mutex);
 831         goto do_again;
 832
 833 insert_end:
 834         mutex_unlock(&node->mutex);
 835         return ret;
 836 }
 837
 838 static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
 839                                     struct btrfs_root *root,
 840                                     struct btrfs_path *path,
 841                                     struct btrfs_delayed_item *item)
 842 {
 843         struct btrfs_delayed_item *curr, *next;
 844         struct extent_buffer *leaf;
 845         struct btrfs_key key;
 846         struct list_head head;
 847         int nitems, i, last_item;
 848         int ret = 0;
 849
 850         BUG_ON(!path->nodes[0]);
 851
 852         leaf = path->nodes[0];
 853
 854         i = path->slots[0];
 855         last_item = btrfs_header_nritems(leaf) - 1;
 856         if (i > last_item)
 857                 return -ENOENT; /* FIXME: Is errno suitable? */
 858
 859         next = item;
 860         INIT_LIST_HEAD(&head);
 861         btrfs_item_key_to_cpu(leaf, &key, i);
 862         nitems = 0;
 863         /*
 864          * count the number of the dir index items that we can delete in batch
 865          */
 866         while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
 867                 list_add_tail(&next->tree_list, &head);
 868                 nitems++;
 869
 870                 curr = next;
 871                 next = __btrfs_next_delayed_item(curr);
 872                 if (!next)
 873                         break;
 874
 875                 if (!btrfs_is_continuous_delayed_item(curr, next))
 876                         break;
 877
 878                 i++;
 879                 if (i > last_item)
 880                         break;
 881                 btrfs_item_key_to_cpu(leaf, &key, i);
 882         }
 883
 884         if (!nitems)
 885                 return 0;
 886
 887         ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
 888         if (ret)
 889                 goto out;
 890
 891         list_for_each_entry_safe(curr, next, &head, tree_list) {
 892                 btrfs_delayed_item_release_metadata(root, curr);
 893                 list_del(&curr->tree_list);
 894                 btrfs_release_delayed_item(curr);
 895         }
 896
 897 out:
 898         return ret;
 899 }
 900
 901 static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
 902                                       struct btrfs_path *path,
 903                                       struct btrfs_root *root,
 904                                       struct btrfs_delayed_node *node)
 905 {
 906         struct btrfs_delayed_item *curr, *prev;
 907         int ret = 0;
 908
 909 do_again:
 910         mutex_lock(&node->mutex);
 911         curr = __btrfs_first_delayed_deletion_item(node);
 912         if (!curr)
 913                 goto delete_fail;
 914
 915         ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
 916         if (ret < 0)
 917                 goto delete_fail;
 918         else if (ret > 0) {
 919                 /*
 920                  * can't find the item which the node points to, so this node
 921                  * is invalid, just drop it.
 922                  */
 923                 prev = curr;
 924                 curr = __btrfs_next_delayed_item(prev);
 925                 btrfs_release_delayed_item(prev);
 926                 ret = 0;
 927                 btrfs_release_path(root, path);
 928                 if (curr)
 929                         goto do_again;
 930                 else
 931                         goto delete_fail;
 932         }
 933
 934         btrfs_batch_delete_items(trans, root, path, curr);
 935         btrfs_release_path(root, path);
 936         mutex_unlock(&node->mutex);
 937         goto do_again;
 938
 939 delete_fail:
 940         btrfs_release_path(root, path);
 941         mutex_unlock(&node->mutex);
 942         return ret;
 943 }
 944
 945 static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
 946 {
 947         struct btrfs_delayed_root *delayed_root;
 948
 949         if (delayed_node && delayed_node->inode_dirty) {
 950                 BUG_ON(!delayed_node->root);
 951                 delayed_node->inode_dirty = 0;
 952                 delayed_node->count--;
 953
 954                 delayed_root = delayed_node->root->fs_info->delayed_root;
 955                 atomic_dec(&delayed_root->items);
 956                 if (atomic_read(&delayed_root->items) <
 957                     BTRFS_DELAYED_BACKGROUND &&
 958                     waitqueue_active(&delayed_root->wait))
 959                         wake_up(&delayed_root->wait);
 960         }
 961 }
 962
 963 static int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
 964                                       struct btrfs_root *root,
 965                                       struct btrfs_path *path,
 966                                       struct btrfs_delayed_node *node)
 967 {
 968         struct btrfs_key key;
 969         struct btrfs_inode_item *inode_item;
 970         struct extent_buffer *leaf;
 971         int ret;
 972
 973         mutex_lock(&node->mutex);
 974         if (!node->inode_dirty) {
 975                 mutex_unlock(&node->mutex);
 976                 return 0;
 977         }
 978
 979         key.objectid = node->inode_id;
 980         btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
 981         key.offset = 0;
 982         ret = btrfs_lookup_inode(trans, root, path, &key, 1);
 983         if (ret > 0) {
 984                 btrfs_release_path(root, path);
 985                 mutex_unlock(&node->mutex);
 986                 return -ENOENT;
 987         } else if (ret < 0) {
 988                 mutex_unlock(&node->mutex);
 989                 return ret;
 990         }
 991
 992         btrfs_unlock_up_safe(path, 1);
 993         leaf = path->nodes[0];
 994         inode_item = btrfs_item_ptr(leaf, path->slots[0],
 995                                     struct btrfs_inode_item);
 996         write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
 997                             sizeof(struct btrfs_inode_item));
 998         btrfs_mark_buffer_dirty(leaf);
 999         btrfs_release_path(root, path);
1000
1001         btrfs_delayed_inode_release_metadata(root, node);
1002         btrfs_release_delayed_inode(node);
1003         mutex_unlock(&node->mutex);
1004
1005         return 0;
1006 }
1007
1008 /* Called when committing the transaction. */
1009 int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1010                             struct btrfs_root *root)
1011 {
1012         struct btrfs_delayed_root *delayed_root;
1013         struct btrfs_delayed_node *curr_node, *prev_node;
1014         struct btrfs_path *path;
1015         int ret = 0;
1016
1017         path = btrfs_alloc_path();
1018         if (!path)
1019                 return -ENOMEM;
1020         path->leave_spinning = 1;
1021
1022         delayed_root = btrfs_get_delayed_root(root);
1023
1024         curr_node = btrfs_first_delayed_node(delayed_root);
1025         while (curr_node) {
1026                 root = curr_node->root;
1027                 ret = btrfs_insert_delayed_items(trans, path, root,
1028                                                  curr_node);
1029                 if (!ret)
1030                         ret = btrfs_delete_delayed_items(trans, path, root,
1031                                                          curr_node);
1032                 if (!ret)
1033                         ret = btrfs_update_delayed_inode(trans, root, path,
1034                                                          curr_node);
1035                 if (ret) {
1036                         btrfs_release_delayed_node(curr_node);
1037                         break;
1038                 }
1039
1040                 prev_node = curr_node;
1041                 curr_node = btrfs_next_delayed_node(curr_node);
1042                 btrfs_release_delayed_node(prev_node);
1043         }
1044
1045         btrfs_free_path(path);
1046         return ret;
1047 }
1048
1049 static int __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1050                                               struct btrfs_delayed_node *node)
1051 {
1052         struct btrfs_path *path;
1053         int ret;
1054
1055         path = btrfs_alloc_path();
1056         if (!path)
1057                 return -ENOMEM;
1058         path->leave_spinning = 1;
1059
1060         ret = btrfs_insert_delayed_items(trans, path, node->root, node);
1061         if (!ret)
1062                 ret = btrfs_delete_delayed_items(trans, path, node->root, node);
1063         if (!ret)
1064                 ret = btrfs_update_delayed_inode(trans, node->root, path, node);
1065         btrfs_free_path(path);
1066
1067         return ret;
1068 }
1069
1070 int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1071                                      struct inode *inode)
1072 {
1073         struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1074         int ret;
1075
1076         if (!delayed_node)
1077                 return 0;
1078
1079         mutex_lock(&delayed_node->mutex);
1080         if (!delayed_node->count) {
1081                 mutex_unlock(&delayed_node->mutex);
1082                 btrfs_release_delayed_node(delayed_node);
1083                 return 0;
1084         }
1085         mutex_unlock(&delayed_node->mutex);
1086
1087         ret = __btrfs_commit_inode_delayed_items(trans, delayed_node);
1088         btrfs_release_delayed_node(delayed_node);
1089         return ret;
1090 }
1091
1092 void btrfs_remove_delayed_node(struct inode *inode)
1093 {
1094         struct btrfs_delayed_node *delayed_node;
1095
1096         delayed_node = ACCESS_ONCE(BTRFS_I(inode)->delayed_node);
1097         if (!delayed_node)
1098                 return;
1099
1100         BTRFS_I(inode)->delayed_node = NULL;
1101         btrfs_release_delayed_node(delayed_node);
1102 }
1103
1104 struct btrfs_async_delayed_node {
1105         struct btrfs_root *root;
1106         struct btrfs_delayed_node *delayed_node;
1107         struct btrfs_work work;
1108 };
1109
1110 static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
1111 {
1112         struct btrfs_async_delayed_node *async_node;
1113         struct btrfs_trans_handle *trans;
1114         struct btrfs_path *path;
1115         struct btrfs_delayed_node *delayed_node = NULL;
1116         struct btrfs_root *root;
1117         unsigned long nr = 0;
1118         int need_requeue = 0;
1119         int ret;
1120
1121         async_node = container_of(work, struct btrfs_async_delayed_node, work);
1122
1123         path = btrfs_alloc_path();
1124         if (!path)
1125                 goto out;
1126         path->leave_spinning = 1;
1127
1128         delayed_node = async_node->delayed_node;
1129         root = delayed_node->root;
1130
1131         trans = btrfs_join_transaction(root, 0);
1132         if (IS_ERR(trans))
1133                 goto free_path;
1134
1135         ret = btrfs_insert_delayed_items(trans, path, root, delayed_node);
1136         if (!ret)
1137                 ret = btrfs_delete_delayed_items(trans, path, root,
1138                                                  delayed_node);
1139
1140         if (!ret)
1141                 btrfs_update_delayed_inode(trans, root, path, delayed_node);
1142
1143         /*
1144          * Maybe new delayed items have been inserted, so we need requeue
1145          * the work. Besides that, we must dequeue the empty delayed nodes
1146          * to avoid the race between delayed items balance and the worker.
1147          * The race like this:
1148          *      Task1                           Worker thread
1149          *                                      count == 0, needn't requeue
1150          *                                        also needn't insert the
1151          *                                        delayed node into prepare
1152          *                                        list again.
1153          *      add lots of delayed items
1154          *      queue the delayed node
1155          *        already in the list,
1156          *        and not in the prepare
1157          *        list, it means the delayed
1158          *        node is being dealt with
1159          *        by the worker.
1160          *      do delayed items balance
1161          *        the delayed node is being
1162          *        dealt with by the worker
1163          *        now, just wait.
1164          *                                      the worker goto idle.
1165          * Task1 will sleep until the transaction is commited.
1166          */
1167         mutex_lock(&delayed_node->mutex);
1168         if (delayed_node->count)
1169                 need_requeue = 1;
1170         else
1171                 btrfs_dequeue_delayed_node(root->fs_info->delayed_root,
1172                                            delayed_node);
1173         mutex_unlock(&delayed_node->mutex);
1174
1175         nr = trans->blocks_used;
1176
1177         btrfs_end_transaction_dmeta(trans, root);
1178         __btrfs_btree_balance_dirty(root, nr);
1179 free_path:
1180         btrfs_free_path(path);
1181 out:
1182         if (need_requeue)
1183                 btrfs_requeue_work(&async_node->work);
1184         else {
1185                 btrfs_release_prepared_delayed_node(delayed_node);
1186                 kfree(async_node);
1187         }
1188 }
1189
1190 static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
1191                                      struct btrfs_root *root, int all)
1192 {
1193         struct btrfs_async_delayed_node *async_node;
1194         struct btrfs_delayed_node *curr;
1195         int count = 0;
1196
1197 again:
1198         curr = btrfs_first_prepared_delayed_node(delayed_root);
1199         if (!curr)
1200                 return 0;
1201
1202         async_node = kmalloc(sizeof(*async_node), GFP_NOFS);
1203         if (!async_node) {
1204                 btrfs_release_prepared_delayed_node(curr);
1205                 return -ENOMEM;
1206         }
1207
1208         async_node->root = root;
1209         async_node->delayed_node = curr;
1210
1211         async_node->work.func = btrfs_async_run_delayed_node_done;
1212         async_node->work.flags = 0;
1213
1214         btrfs_queue_worker(&root->fs_info->delayed_workers, &async_node->work);
1215         count++;
1216
1217         if (all || count < 4)
1218                 goto again;
1219
1220         return 0;
1221 }
1222
1223 void btrfs_balance_delayed_items(struct btrfs_root *root)
1224 {
1225         struct btrfs_delayed_root *delayed_root;
1226
1227         delayed_root = btrfs_get_delayed_root(root);
1228
1229         if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1230                 return;
1231
1232         if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
1233                 int ret;
1234                 ret = btrfs_wq_run_delayed_node(delayed_root, root, 1);
1235                 if (ret)
1236                         return;
1237
1238                 wait_event_interruptible_timeout(
1239                                 delayed_root->wait,
1240                                 (atomic_read(&delayed_root->items) <
1241                                  BTRFS_DELAYED_BACKGROUND),
1242                                 HZ);
1243                 return;
1244         }
1245
1246         btrfs_wq_run_delayed_node(delayed_root, root, 0);
1247 }
1248
1249 int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
1250                                    struct btrfs_root *root, const char *name,
1251                                    int name_len, struct inode *dir,
1252                                    struct btrfs_disk_key *disk_key, u8 type,
1253                                    u64 index)
1254 {
1255         struct btrfs_delayed_node *delayed_node;
1256         struct btrfs_delayed_item *delayed_item;
1257         struct btrfs_dir_item *dir_item;
1258         int ret;
1259
1260         delayed_node = btrfs_get_or_create_delayed_node(dir);
1261         if (IS_ERR(delayed_node))
1262                 return PTR_ERR(delayed_node);
1263
1264         delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
1265         if (!delayed_item) {
1266                 ret = -ENOMEM;
1267                 goto release_node;
1268         }
1269
1270         ret = btrfs_delayed_item_reserve_metadata(trans, root, delayed_item);
1271         /*
1272          * we have reserved enough space when we start a new transaction,
1273          * so reserving metadata failure is impossible
1274          */
1275         BUG_ON(ret);
1276
1277         delayed_item->key.objectid = dir->i_ino;
1278         btrfs_set_key_type(&delayed_item->key, BTRFS_DIR_INDEX_KEY);
1279         delayed_item->key.offset = index;
1280
1281         dir_item = (struct btrfs_dir_item *)delayed_item->data;
1282         dir_item->location = *disk_key;
1283         dir_item->transid = cpu_to_le64(trans->transid);
1284         dir_item->data_len = 0;
1285         dir_item->name_len = cpu_to_le16(name_len);
1286         dir_item->type = type;
1287         memcpy((char *)(dir_item + 1), name, name_len);
1288
1289         mutex_lock(&delayed_node->mutex);
1290         ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
1291         if (unlikely(ret)) {
1292                 printk(KERN_ERR "err add delayed dir index item(name: %s) into "
1293                                 "the insertion tree of the delayed node"
1294                                 "(root id: %llu, inode id: %llu, errno: %d)\n",
1295                                 name,
1296                                 (unsigned long long)delayed_node->root->objectid,
1297                                 (unsigned long long)delayed_node->inode_id,
1298                                 ret);
1299                 BUG();
1300         }
1301         mutex_unlock(&delayed_node->mutex);
1302
1303 release_node:
1304         btrfs_release_delayed_node(delayed_node);
1305         return ret;
1306 }
1307
1308 static int btrfs_delete_delayed_insertion_item(struct btrfs_root *root,
1309                                                struct btrfs_delayed_node *node,
1310                                                struct btrfs_key *key)
1311 {
1312         struct btrfs_delayed_item *item;
1313
1314         mutex_lock(&node->mutex);
1315         item = __btrfs_lookup_delayed_insertion_item(node, key);
1316         if (!item) {
1317                 mutex_unlock(&node->mutex);
1318                 return 1;
1319         }
1320
1321         btrfs_delayed_item_release_metadata(root, item);
1322         btrfs_release_delayed_item(item);
1323         mutex_unlock(&node->mutex);
1324         return 0;
1325 }
1326
1327 int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
1328                                    struct btrfs_root *root, struct inode *dir,
1329                                    u64 index)
1330 {
1331         struct btrfs_delayed_node *node;
1332         struct btrfs_delayed_item *item;
1333         struct btrfs_key item_key;
1334         int ret;
1335
1336         node = btrfs_get_or_create_delayed_node(dir);
1337         if (IS_ERR(node))
1338                 return PTR_ERR(node);
1339
1340         item_key.objectid = dir->i_ino;
1341         btrfs_set_key_type(&item_key, BTRFS_DIR_INDEX_KEY);
1342         item_key.offset = index;
1343
1344         ret = btrfs_delete_delayed_insertion_item(root, node, &item_key);
1345         if (!ret)
1346                 goto end;
1347
1348         item = btrfs_alloc_delayed_item(0);
1349         if (!item) {
1350                 ret = -ENOMEM;
1351                 goto end;
1352         }
1353
1354         item->key = item_key;
1355
1356         ret = btrfs_delayed_item_reserve_metadata(trans, root, item);
1357         /*
1358          * we have reserved enough space when we start a new transaction,
1359          * so reserving metadata failure is impossible.
1360          */
1361         BUG_ON(ret);
1362
1363         mutex_lock(&node->mutex);
1364         ret = __btrfs_add_delayed_deletion_item(node, item);
1365         if (unlikely(ret)) {
1366                 printk(KERN_ERR "err add delayed dir index item(index: %llu) "
1367                                 "into the deletion tree of the delayed node"
1368                                 "(root id: %llu, inode id: %llu, errno: %d)\n",
1369                                 (unsigned long long)index,
1370                                 (unsigned long long)node->root->objectid,
1371                                 (unsigned long long)node->inode_id,
1372                                 ret);
1373                 BUG();
1374         }
1375         mutex_unlock(&node->mutex);
1376 end:
1377         btrfs_release_delayed_node(node);
1378         return ret;
1379 }
1380
1381 int btrfs_inode_delayed_dir_index_count(struct inode *inode)
1382 {
1383         struct btrfs_delayed_node *delayed_node = BTRFS_I(inode)->delayed_node;
1384         int ret = 0;
1385
1386         if (!delayed_node)
1387                 return -ENOENT;
1388
1389         /*
1390          * Since we have held i_mutex of this directory, it is impossible that
1391          * a new directory index is added into the delayed node and index_cnt
1392          * is updated now. So we needn't lock the delayed node.
1393          */
1394         if (!delayed_node->index_cnt)
1395                 return -EINVAL;
1396
1397         BTRFS_I(inode)->index_cnt = delayed_node->index_cnt;
1398         return ret;
1399 }
1400
1401 void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
1402                              struct list_head *del_list)
1403 {
1404         struct btrfs_delayed_node *delayed_node;
1405         struct btrfs_delayed_item *item;
1406
1407         delayed_node = btrfs_get_delayed_node(inode);
1408         if (!delayed_node)
1409                 return;
1410
1411         mutex_lock(&delayed_node->mutex);
1412         item = __btrfs_first_delayed_insertion_item(delayed_node);
1413         while (item) {
1414                 atomic_inc(&item->refs);
1415                 list_add_tail(&item->readdir_list, ins_list);
1416                 item = __btrfs_next_delayed_item(item);
1417         }
1418
1419         item = __btrfs_first_delayed_deletion_item(delayed_node);
1420         while (item) {
1421                 atomic_inc(&item->refs);
1422                 list_add_tail(&item->readdir_list, del_list);
1423                 item = __btrfs_next_delayed_item(item);
1424         }
1425         mutex_unlock(&delayed_node->mutex);
1426         /*
1427          * This delayed node is still cached in the btrfs inode, so refs
1428          * must be > 1 now, and we needn't check it is going to be freed
1429          * or not.
1430          *
1431          * Besides that, this function is used to read dir, we do not
1432          * insert/delete delayed items in this period. So we also needn't
1433          * requeue or dequeue this delayed node.
1434          */
1435         atomic_dec(&delayed_node->refs);
1436 }
1437
1438 void btrfs_put_delayed_items(struct list_head *ins_list,
1439                              struct list_head *del_list)
1440 {
1441         struct btrfs_delayed_item *curr, *next;
1442
1443         list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1444                 list_del(&curr->readdir_list);
1445                 if (atomic_dec_and_test(&curr->refs))
1446                         kfree(curr);
1447         }
1448
1449         list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1450                 list_del(&curr->readdir_list);
1451                 if (atomic_dec_and_test(&curr->refs))
1452                         kfree(curr);
1453         }
1454 }
1455
1456 int btrfs_should_delete_dir_index(struct list_head *del_list,
1457                                   u64 index)
1458 {
1459         struct btrfs_delayed_item *curr, *next;
1460         int ret;
1461
1462         if (list_empty(del_list))
1463                 return 0;
1464
1465         list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1466                 if (curr->key.offset > index)
1467                         break;
1468
1469                 list_del(&curr->readdir_list);
1470                 ret = (curr->key.offset == index);
1471
1472                 if (atomic_dec_and_test(&curr->refs))
1473                         kfree(curr);
1474
1475                 if (ret)
1476                         return 1;
1477                 else
1478                         continue;
1479         }
1480         return 0;
1481 }
1482
1483 /*
1484  * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
1485  *
1486  */
1487 int btrfs_readdir_delayed_dir_index(struct file *filp, void *dirent,
1488                                     filldir_t filldir,
1489                                     struct list_head *ins_list)
1490 {
1491         struct btrfs_dir_item *di;
1492         struct btrfs_delayed_item *curr, *next;
1493         struct btrfs_key location;
1494         char *name;
1495         int name_len;
1496         int over = 0;
1497         unsigned char d_type;
1498
1499         if (list_empty(ins_list))
1500                 return 0;
1501
1502         /*
1503          * Changing the data of the delayed item is impossible. So
1504          * we needn't lock them. And we have held i_mutex of the
1505          * directory, nobody can delete any directory indexes now.
1506          */
1507         list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1508                 list_del(&curr->readdir_list);
1509
1510                 if (curr->key.offset < filp->f_pos) {
1511                         if (atomic_dec_and_test(&curr->refs))
1512                                 kfree(curr);
1513                         continue;
1514                 }
1515
1516                 filp->f_pos = curr->key.offset;
1517
1518                 di = (struct btrfs_dir_item *)curr->data;
1519                 name = (char *)(di + 1);
1520                 name_len = le16_to_cpu(di->name_len);
1521
1522                 d_type = btrfs_filetype_table[di->type];
1523                 btrfs_disk_key_to_cpu(&location, &di->location);
1524
1525                 over = filldir(dirent, name, name_len, curr->key.offset,
1526                                location.objectid, d_type);
1527
1528                 if (atomic_dec_and_test(&curr->refs))
1529                         kfree(curr);
1530
1531                 if (over)
1532                         return 1;
1533         }
1534         return 0;
1535 }
1536
1537 BTRFS_SETGET_STACK_FUNCS(stack_inode_generation, struct btrfs_inode_item,
1538                          generation, 64);
1539 BTRFS_SETGET_STACK_FUNCS(stack_inode_sequence, struct btrfs_inode_item,
1540                          sequence, 64);
1541 BTRFS_SETGET_STACK_FUNCS(stack_inode_transid, struct btrfs_inode_item,
1542                          transid, 64);
1543 BTRFS_SETGET_STACK_FUNCS(stack_inode_size, struct btrfs_inode_item, size, 64);
1544 BTRFS_SETGET_STACK_FUNCS(stack_inode_nbytes, struct btrfs_inode_item,
1545                          nbytes, 64);
1546 BTRFS_SETGET_STACK_FUNCS(stack_inode_block_group, struct btrfs_inode_item,
1547                          block_group, 64);
1548 BTRFS_SETGET_STACK_FUNCS(stack_inode_nlink, struct btrfs_inode_item, nlink, 32);
1549 BTRFS_SETGET_STACK_FUNCS(stack_inode_uid, struct btrfs_inode_item, uid, 32);
1550 BTRFS_SETGET_STACK_FUNCS(stack_inode_gid, struct btrfs_inode_item, gid, 32);
1551 BTRFS_SETGET_STACK_FUNCS(stack_inode_mode, struct btrfs_inode_item, mode, 32);
1552 BTRFS_SETGET_STACK_FUNCS(stack_inode_rdev, struct btrfs_inode_item, rdev, 64);
1553 BTRFS_SETGET_STACK_FUNCS(stack_inode_flags, struct btrfs_inode_item, flags, 64);
1554
1555 BTRFS_SETGET_STACK_FUNCS(stack_timespec_sec, struct btrfs_timespec, sec, 64);
1556 BTRFS_SETGET_STACK_FUNCS(stack_timespec_nsec, struct btrfs_timespec, nsec, 32);
1557
1558 static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
1559                                   struct btrfs_inode_item *inode_item,
1560                                   struct inode *inode)
1561 {
1562         btrfs_set_stack_inode_uid(inode_item, inode->i_uid);
1563         btrfs_set_stack_inode_gid(inode_item, inode->i_gid);
1564         btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
1565         btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
1566         btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
1567         btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
1568         btrfs_set_stack_inode_generation(inode_item,
1569                                          BTRFS_I(inode)->generation);
1570         btrfs_set_stack_inode_sequence(inode_item, BTRFS_I(inode)->sequence);
1571         btrfs_set_stack_inode_transid(inode_item, trans->transid);
1572         btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
1573         btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
1574         btrfs_set_stack_inode_block_group(inode_item,
1575                                           BTRFS_I(inode)->block_group);
1576
1577         btrfs_set_stack_timespec_sec(btrfs_inode_atime(inode_item),
1578                                      inode->i_atime.tv_sec);
1579         btrfs_set_stack_timespec_nsec(btrfs_inode_atime(inode_item),
1580                                       inode->i_atime.tv_nsec);
1581
1582         btrfs_set_stack_timespec_sec(btrfs_inode_mtime(inode_item),
1583                                      inode->i_mtime.tv_sec);
1584         btrfs_set_stack_timespec_nsec(btrfs_inode_mtime(inode_item),
1585                                       inode->i_mtime.tv_nsec);
1586
1587         btrfs_set_stack_timespec_sec(btrfs_inode_ctime(inode_item),
1588                                      inode->i_ctime.tv_sec);
1589         btrfs_set_stack_timespec_nsec(btrfs_inode_ctime(inode_item),
1590                                       inode->i_ctime.tv_nsec);
1591 }
1592
1593 int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
1594                                struct btrfs_root *root, struct inode *inode)
1595 {
1596         struct btrfs_delayed_node *delayed_node;
1597         int ret;
1598
1599         delayed_node = btrfs_get_or_create_delayed_node(inode);
1600         if (IS_ERR(delayed_node))
1601                 return PTR_ERR(delayed_node);
1602
1603         mutex_lock(&delayed_node->mutex);
1604         if (delayed_node->inode_dirty) {
1605                 fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1606                 goto release_node;
1607         }
1608
1609         ret = btrfs_delayed_inode_reserve_metadata(trans, root, delayed_node);
1610         /*
1611          * we must reserve enough space when we start a new transaction,
1612          * so reserving metadata failure is impossible
1613          */
1614         BUG_ON(ret);
1615
1616         fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1617         delayed_node->inode_dirty = 1;
1618         delayed_node->count++;
1619         atomic_inc(&root->fs_info->delayed_root->items);
1620 release_node:
1621         mutex_unlock(&delayed_node->mutex);
1622         btrfs_release_delayed_node(delayed_node);
1623         return ret;
1624 }
1625
1626 static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
1627 {
1628         struct btrfs_root *root = delayed_node->root;
1629         struct btrfs_delayed_item *curr_item, *prev_item;
1630
1631         mutex_lock(&delayed_node->mutex);
1632         curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
1633         while (curr_item) {
1634                 btrfs_delayed_item_release_metadata(root, curr_item);
1635                 prev_item = curr_item;
1636                 curr_item = __btrfs_next_delayed_item(prev_item);
1637                 btrfs_release_delayed_item(prev_item);
1638         }
1639
1640         curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
1641         while (curr_item) {
1642                 btrfs_delayed_item_release_metadata(root, curr_item);
1643                 prev_item = curr_item;
1644                 curr_item = __btrfs_next_delayed_item(prev_item);
1645                 btrfs_release_delayed_item(prev_item);
1646         }
1647
1648         if (delayed_node->inode_dirty) {
1649                 btrfs_delayed_inode_release_metadata(root, delayed_node);
1650                 btrfs_release_delayed_inode(delayed_node);
1651         }
1652         mutex_unlock(&delayed_node->mutex);
1653 }
1654
1655 void btrfs_kill_delayed_inode_items(struct inode *inode)
1656 {
1657         struct btrfs_delayed_node *delayed_node;
1658
1659         delayed_node = btrfs_get_delayed_node(inode);
1660         if (!delayed_node)
1661                 return;
1662
1663         __btrfs_kill_delayed_node(delayed_node);
1664         btrfs_release_delayed_node(delayed_node);
1665 }
1666
1667 void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
1668 {
1669         u64 inode_id = 0;
1670         struct btrfs_delayed_node *delayed_nodes[8];
1671         int i, n;
1672
1673         while (1) {
1674                 spin_lock(&root->inode_lock);
1675                 n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
1676                                            (void **)delayed_nodes, inode_id,
1677                                            ARRAY_SIZE(delayed_nodes));
1678                 if (!n) {
1679                         spin_unlock(&root->inode_lock);
1680                         break;
1681                 }
1682
1683                 inode_id = delayed_nodes[n - 1]->inode_id + 1;
1684
1685                 for (i = 0; i < n; i++)
1686                         atomic_inc(&delayed_nodes[i]->refs);
1687                 spin_unlock(&root->inode_lock);
1688
1689                 for (i = 0; i < n; i++) {
1690                         __btrfs_kill_delayed_node(delayed_nodes[i]);
1691                         btrfs_release_delayed_node(delayed_nodes[i]);
1692                 }
1693         }
1694 }