fs/btrfs/delayed-ref.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  * Copyright (C) 2009 Oracle.  All rights reserved.
   4  */
   5
   6 #include <linux/sched.h>
   7 #include <linux/slab.h>
   8 #include <linux/sort.h>
   9 #include "ctree.h"
  10 #include "delayed-ref.h"
  11 #include "transaction.h"
  12 #include "qgroup.h"
  13 #include "space-info.h"
  14 #include "tree-mod-log.h"
  15
  16 struct kmem_cache *btrfs_delayed_ref_head_cachep;
  17 struct kmem_cache *btrfs_delayed_tree_ref_cachep;
  18 struct kmem_cache *btrfs_delayed_data_ref_cachep;
  19 struct kmem_cache *btrfs_delayed_extent_op_cachep;
  20 /*
  21  * delayed back reference update tracking.  For subvolume trees
  22  * we queue up extent allocations and backref maintenance for
  23  * delayed processing.   This avoids deep call chains where we
  24  * add extents in the middle of btrfs_search_slot, and it allows
  25  * us to buffer up frequently modified backrefs in an rb tree instead
  26  * of hammering updates on the extent allocation tree.
  27  */
  28
  29 bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info)
  30 {
  31         struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
  32         struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
  33         bool ret = false;
  34         u64 reserved;
  35
  36         spin_lock(&global_rsv->lock);
  37         reserved = global_rsv->reserved;
  38         spin_unlock(&global_rsv->lock);
  39
  40         /*
  41          * Since the global reserve is just kind of magic we don't really want
  42          * to rely on it to save our bacon, so if our size is more than the
  43          * delayed_refs_rsv and the global rsv then it's time to think about
  44          * bailing.
  45          */
  46         spin_lock(&delayed_refs_rsv->lock);
  47         reserved += delayed_refs_rsv->reserved;
  48         if (delayed_refs_rsv->size >= reserved)
  49                 ret = true;
  50         spin_unlock(&delayed_refs_rsv->lock);
  51         return ret;
  52 }
  53
  54 int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans)
  55 {
  56         u64 num_entries =
  57                 atomic_read(&trans->transaction->delayed_refs.num_entries);
  58         u64 avg_runtime;
  59         u64 val;
  60
  61         smp_mb();
  62         avg_runtime = trans->fs_info->avg_delayed_ref_runtime;
  63         val = num_entries * avg_runtime;
  64         if (val >= NSEC_PER_SEC)
  65                 return 1;
  66         if (val >= NSEC_PER_SEC / 2)
  67                 return 2;
  68
  69         return btrfs_check_space_for_delayed_refs(trans->fs_info);
  70 }
  71
  72 /**
  73  * Release a ref head's reservation
  74  *
  75  * @fs_info:  the filesystem
  76  * @nr:       number of items to drop
  77  *
  78  * This drops the delayed ref head's count from the delayed refs rsv and frees
  79  * any excess reservation we had.
  80  */
  81 void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr)
  82 {
  83         struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
  84         u64 num_bytes = btrfs_calc_insert_metadata_size(fs_info, nr);
  85         u64 released = 0;
  86
  87         /*
  88          * We have to check the mount option here because we could be enabling
  89          * the free space tree for the first time and don't have the compat_ro
  90          * option set yet.
  91          *
  92          * We need extra reservations if we have the free space tree because
  93          * we'll have to modify that tree as well.
  94          */
  95         if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
  96                 num_bytes *= 2;
  97
  98         released = btrfs_block_rsv_release(fs_info, block_rsv, num_bytes, NULL);
  99         if (released)
 100                 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
 101                                               0, released, 0);
 102 }
 103
 104 /*
 105  * btrfs_update_delayed_refs_rsv - adjust the size of the delayed refs rsv
 106  * @trans - the trans that may have generated delayed refs
 107  *
 108  * This is to be called anytime we may have adjusted trans->delayed_ref_updates,
 109  * it'll calculate the additional size and add it to the delayed_refs_rsv.
 110  */
 111 void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans)
 112 {
 113         struct btrfs_fs_info *fs_info = trans->fs_info;
 114         struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
 115         u64 num_bytes;
 116
 117         if (!trans->delayed_ref_updates)
 118                 return;
 119
 120         num_bytes = btrfs_calc_insert_metadata_size(fs_info,
 121                                                     trans->delayed_ref_updates);
 122         /*
 123          * We have to check the mount option here because we could be enabling
 124          * the free space tree for the first time and don't have the compat_ro
 125          * option set yet.
 126          *
 127          * We need extra reservations if we have the free space tree because
 128          * we'll have to modify that tree as well.
 129          */
 130         if (btrfs_test_opt(fs_info, FREE_SPACE_TREE))
 131                 num_bytes *= 2;
 132
 133         spin_lock(&delayed_rsv->lock);
 134         delayed_rsv->size += num_bytes;
 135         delayed_rsv->full = 0;
 136         spin_unlock(&delayed_rsv->lock);
 137         trans->delayed_ref_updates = 0;
 138 }
 139
 140 /**
 141  * Transfer bytes to our delayed refs rsv
 142  *
 143  * @fs_info:   the filesystem
 144  * @src:       source block rsv to transfer from
 145  * @num_bytes: number of bytes to transfer
 146  *
 147  * This transfers up to the num_bytes amount from the src rsv to the
 148  * delayed_refs_rsv.  Any extra bytes are returned to the space info.
 149  */
 150 void btrfs_migrate_to_delayed_refs_rsv(struct btrfs_fs_info *fs_info,
 151                                        struct btrfs_block_rsv *src,
 152                                        u64 num_bytes)
 153 {
 154         struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
 155         u64 to_free = 0;
 156
 157         spin_lock(&src->lock);
 158         src->reserved -= num_bytes;
 159         src->size -= num_bytes;
 160         spin_unlock(&src->lock);
 161
 162         spin_lock(&delayed_refs_rsv->lock);
 163         if (delayed_refs_rsv->size > delayed_refs_rsv->reserved) {
 164                 u64 delta = delayed_refs_rsv->size -
 165                         delayed_refs_rsv->reserved;
 166                 if (num_bytes > delta) {
 167                         to_free = num_bytes - delta;
 168                         num_bytes = delta;
 169                 }
 170         } else {
 171                 to_free = num_bytes;
 172                 num_bytes = 0;
 173         }
 174
 175         if (num_bytes)
 176                 delayed_refs_rsv->reserved += num_bytes;
 177         if (delayed_refs_rsv->reserved >= delayed_refs_rsv->size)
 178                 delayed_refs_rsv->full = 1;
 179         spin_unlock(&delayed_refs_rsv->lock);
 180
 181         if (num_bytes)
 182                 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
 183                                               0, num_bytes, 1);
 184         if (to_free)
 185                 btrfs_space_info_free_bytes_may_use(fs_info,
 186                                 delayed_refs_rsv->space_info, to_free);
 187 }
 188
 189 /**
 190  * Refill based on our delayed refs usage
 191  *
 192  * @fs_info: the filesystem
 193  * @flush:   control how we can flush for this reservation.
 194  *
 195  * This will refill the delayed block_rsv up to 1 items size worth of space and
 196  * will return -ENOSPC if we can't make the reservation.
 197  */
 198 int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
 199                                   enum btrfs_reserve_flush_enum flush)
 200 {
 201         struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
 202         u64 limit = btrfs_calc_insert_metadata_size(fs_info, 1);
 203         u64 num_bytes = 0;
 204         int ret = -ENOSPC;
 205
 206         spin_lock(&block_rsv->lock);
 207         if (block_rsv->reserved < block_rsv->size) {
 208                 num_bytes = block_rsv->size - block_rsv->reserved;
 209                 num_bytes = min(num_bytes, limit);
 210         }
 211         spin_unlock(&block_rsv->lock);
 212
 213         if (!num_bytes)
 214                 return 0;
 215
 216         ret = btrfs_reserve_metadata_bytes(fs_info, block_rsv, num_bytes, flush);
 217         if (ret)
 218                 return ret;
 219         btrfs_block_rsv_add_bytes(block_rsv, num_bytes, 0);
 220         trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
 221                                       0, num_bytes, 1);
 222         return 0;
 223 }
 224
 225 /*
 226  * compare two delayed tree backrefs with same bytenr and type
 227  */
 228 static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref1,
 229                           struct btrfs_delayed_tree_ref *ref2)
 230 {
 231         if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
 232                 if (ref1->root < ref2->root)
 233                         return -1;
 234                 if (ref1->root > ref2->root)
 235                         return 1;
 236         } else {
 237                 if (ref1->parent < ref2->parent)
 238                         return -1;
 239                 if (ref1->parent > ref2->parent)
 240                         return 1;
 241         }
 242         return 0;
 243 }
 244
 245 /*
 246  * compare two delayed data backrefs with same bytenr and type
 247  */
 248 static int comp_data_refs(struct btrfs_delayed_data_ref *ref1,
 249                           struct btrfs_delayed_data_ref *ref2)
 250 {
 251         if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) {
 252                 if (ref1->root < ref2->root)
 253                         return -1;
 254                 if (ref1->root > ref2->root)
 255                         return 1;
 256                 if (ref1->objectid < ref2->objectid)
 257                         return -1;
 258                 if (ref1->objectid > ref2->objectid)
 259                         return 1;
 260                 if (ref1->offset < ref2->offset)
 261                         return -1;
 262                 if (ref1->offset > ref2->offset)
 263                         return 1;
 264         } else {
 265                 if (ref1->parent < ref2->parent)
 266                         return -1;
 267                 if (ref1->parent > ref2->parent)
 268                         return 1;
 269         }
 270         return 0;
 271 }
 272
 273 static int comp_refs(struct btrfs_delayed_ref_node *ref1,
 274                      struct btrfs_delayed_ref_node *ref2,
 275                      bool check_seq)
 276 {
 277         int ret = 0;
 278
 279         if (ref1->type < ref2->type)
 280                 return -1;
 281         if (ref1->type > ref2->type)
 282                 return 1;
 283         if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
 284             ref1->type == BTRFS_SHARED_BLOCK_REF_KEY)
 285                 ret = comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref1),
 286                                      btrfs_delayed_node_to_tree_ref(ref2));
 287         else
 288                 ret = comp_data_refs(btrfs_delayed_node_to_data_ref(ref1),
 289                                      btrfs_delayed_node_to_data_ref(ref2));
 290         if (ret)
 291                 return ret;
 292         if (check_seq) {
 293                 if (ref1->seq < ref2->seq)
 294                         return -1;
 295                 if (ref1->seq > ref2->seq)
 296                         return 1;
 297         }
 298         return 0;
 299 }
 300
 301 /* insert a new ref to head ref rbtree */
 302 static struct btrfs_delayed_ref_head *htree_insert(struct rb_root_cached *root,
 303                                                    struct rb_node *node)
 304 {
 305         struct rb_node **p = &root->rb_root.rb_node;
 306         struct rb_node *parent_node = NULL;
 307         struct btrfs_delayed_ref_head *entry;
 308         struct btrfs_delayed_ref_head *ins;
 309         u64 bytenr;
 310         bool leftmost = true;
 311
 312         ins = rb_entry(node, struct btrfs_delayed_ref_head, href_node);
 313         bytenr = ins->bytenr;
 314         while (*p) {
 315                 parent_node = *p;
 316                 entry = rb_entry(parent_node, struct btrfs_delayed_ref_head,
 317                                  href_node);
 318
 319                 if (bytenr < entry->bytenr) {
 320                         p = &(*p)->rb_left;
 321                 } else if (bytenr > entry->bytenr) {
 322                         p = &(*p)->rb_right;
 323                         leftmost = false;
 324                 } else {
 325                         return entry;
 326                 }
 327         }
 328
 329         rb_link_node(node, parent_node, p);
 330         rb_insert_color_cached(node, root, leftmost);
 331         return NULL;
 332 }
 333
 334 static struct btrfs_delayed_ref_node* tree_insert(struct rb_root_cached *root,
 335                 struct btrfs_delayed_ref_node *ins)
 336 {
 337         struct rb_node **p = &root->rb_root.rb_node;
 338         struct rb_node *node = &ins->ref_node;
 339         struct rb_node *parent_node = NULL;
 340         struct btrfs_delayed_ref_node *entry;
 341         bool leftmost = true;
 342
 343         while (*p) {
 344                 int comp;
 345
 346                 parent_node = *p;
 347                 entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
 348                                  ref_node);
 349                 comp = comp_refs(ins, entry, true);
 350                 if (comp < 0) {
 351                         p = &(*p)->rb_left;
 352                 } else if (comp > 0) {
 353                         p = &(*p)->rb_right;
 354                         leftmost = false;
 355                 } else {
 356                         return entry;
 357                 }
 358         }
 359
 360         rb_link_node(node, parent_node, p);
 361         rb_insert_color_cached(node, root, leftmost);
 362         return NULL;
 363 }
 364
 365 static struct btrfs_delayed_ref_head *find_first_ref_head(
 366                 struct btrfs_delayed_ref_root *dr)
 367 {
 368         struct rb_node *n;
 369         struct btrfs_delayed_ref_head *entry;
 370
 371         n = rb_first_cached(&dr->href_root);
 372         if (!n)
 373                 return NULL;
 374
 375         entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
 376
 377         return entry;
 378 }
 379
 380 /*
 381  * Find a head entry based on bytenr. This returns the delayed ref head if it
 382  * was able to find one, or NULL if nothing was in that spot.  If return_bigger
 383  * is given, the next bigger entry is returned if no exact match is found.
 384  */
 385 static struct btrfs_delayed_ref_head *find_ref_head(
 386                 struct btrfs_delayed_ref_root *dr, u64 bytenr,
 387                 bool return_bigger)
 388 {
 389         struct rb_root *root = &dr->href_root.rb_root;
 390         struct rb_node *n;
 391         struct btrfs_delayed_ref_head *entry;
 392
 393         n = root->rb_node;
 394         entry = NULL;
 395         while (n) {
 396                 entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
 397
 398                 if (bytenr < entry->bytenr)
 399                         n = n->rb_left;
 400                 else if (bytenr > entry->bytenr)
 401                         n = n->rb_right;
 402                 else
 403                         return entry;
 404         }
 405         if (entry && return_bigger) {
 406                 if (bytenr > entry->bytenr) {
 407                         n = rb_next(&entry->href_node);
 408                         if (!n)
 409                                 return NULL;
 410                         entry = rb_entry(n, struct btrfs_delayed_ref_head,
 411                                          href_node);
 412                 }
 413                 return entry;
 414         }
 415         return NULL;
 416 }
 417
 418 int btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
 419                            struct btrfs_delayed_ref_head *head)
 420 {
 421         lockdep_assert_held(&delayed_refs->lock);
 422         if (mutex_trylock(&head->mutex))
 423                 return 0;
 424
 425         refcount_inc(&head->refs);
 426         spin_unlock(&delayed_refs->lock);
 427
 428         mutex_lock(&head->mutex);
 429         spin_lock(&delayed_refs->lock);
 430         if (RB_EMPTY_NODE(&head->href_node)) {
 431                 mutex_unlock(&head->mutex);
 432                 btrfs_put_delayed_ref_head(head);
 433                 return -EAGAIN;
 434         }
 435         btrfs_put_delayed_ref_head(head);
 436         return 0;
 437 }
 438
 439 static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,
 440                                     struct btrfs_delayed_ref_root *delayed_refs,
 441                                     struct btrfs_delayed_ref_head *head,
 442                                     struct btrfs_delayed_ref_node *ref)
 443 {
 444         lockdep_assert_held(&head->lock);
 445         rb_erase_cached(&ref->ref_node, &head->ref_tree);
 446         RB_CLEAR_NODE(&ref->ref_node);
 447         if (!list_empty(&ref->add_list))
 448                 list_del(&ref->add_list);
 449         ref->in_tree = 0;
 450         btrfs_put_delayed_ref(ref);
 451         atomic_dec(&delayed_refs->num_entries);
 452 }
 453
 454 static bool merge_ref(struct btrfs_trans_handle *trans,
 455                       struct btrfs_delayed_ref_root *delayed_refs,
 456                       struct btrfs_delayed_ref_head *head,
 457                       struct btrfs_delayed_ref_node *ref,
 458                       u64 seq)
 459 {
 460         struct btrfs_delayed_ref_node *next;
 461         struct rb_node *node = rb_next(&ref->ref_node);
 462         bool done = false;
 463
 464         while (!done && node) {
 465                 int mod;
 466
 467                 next = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
 468                 node = rb_next(node);
 469                 if (seq && next->seq >= seq)
 470                         break;
 471                 if (comp_refs(ref, next, false))
 472                         break;
 473
 474                 if (ref->action == next->action) {
 475                         mod = next->ref_mod;
 476                 } else {
 477                         if (ref->ref_mod < next->ref_mod) {
 478                                 swap(ref, next);
 479                                 done = true;
 480                         }
 481                         mod = -next->ref_mod;
 482                 }
 483
 484                 drop_delayed_ref(trans, delayed_refs, head, next);
 485                 ref->ref_mod += mod;
 486                 if (ref->ref_mod == 0) {
 487                         drop_delayed_ref(trans, delayed_refs, head, ref);
 488                         done = true;
 489                 } else {
 490                         /*
 491                          * Can't have multiples of the same ref on a tree block.
 492                          */
 493                         WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
 494                                 ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
 495                 }
 496         }
 497
 498         return done;
 499 }
 500
 501 void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
 502                               struct btrfs_delayed_ref_root *delayed_refs,
 503                               struct btrfs_delayed_ref_head *head)
 504 {
 505         struct btrfs_fs_info *fs_info = trans->fs_info;
 506         struct btrfs_delayed_ref_node *ref;
 507         struct rb_node *node;
 508         u64 seq = 0;
 509
 510         lockdep_assert_held(&head->lock);
 511
 512         if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
 513                 return;
 514
 515         /* We don't have too many refs to merge for data. */
 516         if (head->is_data)
 517                 return;
 518
 519         seq = btrfs_tree_mod_log_lowest_seq(fs_info);
 520 again:
 521         for (node = rb_first_cached(&head->ref_tree); node;
 522              node = rb_next(node)) {
 523                 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
 524                 if (seq && ref->seq >= seq)
 525                         continue;
 526                 if (merge_ref(trans, delayed_refs, head, ref, seq))
 527                         goto again;
 528         }
 529 }
 530
 531 int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq)
 532 {
 533         int ret = 0;
 534         u64 min_seq = btrfs_tree_mod_log_lowest_seq(fs_info);
 535
 536         if (min_seq != 0 && seq >= min_seq) {
 537                 btrfs_debug(fs_info,
 538                             "holding back delayed_ref %llu, lowest is %llu",
 539                             seq, min_seq);
 540                 ret = 1;
 541         }
 542
 543         return ret;
 544 }
 545
 546 struct btrfs_delayed_ref_head *btrfs_select_ref_head(
 547                 struct btrfs_delayed_ref_root *delayed_refs)
 548 {
 549         struct btrfs_delayed_ref_head *head;
 550
 551 again:
 552         head = find_ref_head(delayed_refs, delayed_refs->run_delayed_start,
 553                              true);
 554         if (!head && delayed_refs->run_delayed_start != 0) {
 555                 delayed_refs->run_delayed_start = 0;
 556                 head = find_first_ref_head(delayed_refs);
 557         }
 558         if (!head)
 559                 return NULL;
 560
 561         while (head->processing) {
 562                 struct rb_node *node;
 563
 564                 node = rb_next(&head->href_node);
 565                 if (!node) {
 566                         if (delayed_refs->run_delayed_start == 0)
 567                                 return NULL;
 568                         delayed_refs->run_delayed_start = 0;
 569                         goto again;
 570                 }
 571                 head = rb_entry(node, struct btrfs_delayed_ref_head,
 572                                 href_node);
 573         }
 574
 575         head->processing = 1;
 576         WARN_ON(delayed_refs->num_heads_ready == 0);
 577         delayed_refs->num_heads_ready--;
 578         delayed_refs->run_delayed_start = head->bytenr +
 579                 head->num_bytes;
 580         return head;
 581 }
 582
 583 void btrfs_delete_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
 584                            struct btrfs_delayed_ref_head *head)
 585 {
 586         lockdep_assert_held(&delayed_refs->lock);
 587         lockdep_assert_held(&head->lock);
 588
 589         rb_erase_cached(&head->href_node, &delayed_refs->href_root);
 590         RB_CLEAR_NODE(&head->href_node);
 591         atomic_dec(&delayed_refs->num_entries);
 592         delayed_refs->num_heads--;
 593         if (head->processing == 0)
 594                 delayed_refs->num_heads_ready--;
 595 }
 596
 597 /*
 598  * Helper to insert the ref_node to the tail or merge with tail.
 599  *
 600  * Return 0 for insert.
 601  * Return >0 for merge.
 602  */
 603 static int insert_delayed_ref(struct btrfs_trans_handle *trans,
 604                               struct btrfs_delayed_ref_root *root,
 605                               struct btrfs_delayed_ref_head *href,
 606                               struct btrfs_delayed_ref_node *ref)
 607 {
 608         struct btrfs_delayed_ref_node *exist;
 609         int mod;
 610         int ret = 0;
 611
 612         spin_lock(&href->lock);
 613         exist = tree_insert(&href->ref_tree, ref);
 614         if (!exist)
 615                 goto inserted;
 616
 617         /* Now we are sure we can merge */
 618         ret = 1;
 619         if (exist->action == ref->action) {
 620                 mod = ref->ref_mod;
 621         } else {
 622                 /* Need to change action */
 623                 if (exist->ref_mod < ref->ref_mod) {
 624                         exist->action = ref->action;
 625                         mod = -exist->ref_mod;
 626                         exist->ref_mod = ref->ref_mod;
 627                         if (ref->action == BTRFS_ADD_DELAYED_REF)
 628                                 list_add_tail(&exist->add_list,
 629                                               &href->ref_add_list);
 630                         else if (ref->action == BTRFS_DROP_DELAYED_REF) {
 631                                 ASSERT(!list_empty(&exist->add_list));
 632                                 list_del(&exist->add_list);
 633                         } else {
 634                                 ASSERT(0);
 635                         }
 636                 } else
 637                         mod = -ref->ref_mod;
 638         }
 639         exist->ref_mod += mod;
 640
 641         /* remove existing tail if its ref_mod is zero */
 642         if (exist->ref_mod == 0)
 643                 drop_delayed_ref(trans, root, href, exist);
 644         spin_unlock(&href->lock);
 645         return ret;
 646 inserted:
 647         if (ref->action == BTRFS_ADD_DELAYED_REF)
 648                 list_add_tail(&ref->add_list, &href->ref_add_list);
 649         atomic_inc(&root->num_entries);
 650         spin_unlock(&href->lock);
 651         return ret;
 652 }
 653
 654 /*
 655  * helper function to update the accounting in the head ref
 656  * existing and update must have the same bytenr
 657  */
 658 static noinline void update_existing_head_ref(struct btrfs_trans_handle *trans,
 659                          struct btrfs_delayed_ref_head *existing,
 660                          struct btrfs_delayed_ref_head *update)
 661 {
 662         struct btrfs_delayed_ref_root *delayed_refs =
 663                 &trans->transaction->delayed_refs;
 664         struct btrfs_fs_info *fs_info = trans->fs_info;
 665         int old_ref_mod;
 666
 667         BUG_ON(existing->is_data != update->is_data);
 668
 669         spin_lock(&existing->lock);
 670         if (update->must_insert_reserved) {
 671                 /* if the extent was freed and then
 672                  * reallocated before the delayed ref
 673                  * entries were processed, we can end up
 674                  * with an existing head ref without
 675                  * the must_insert_reserved flag set.
 676                  * Set it again here
 677                  */
 678                 existing->must_insert_reserved = update->must_insert_reserved;
 679
 680                 /*
 681                  * update the num_bytes so we make sure the accounting
 682                  * is done correctly
 683                  */
 684                 existing->num_bytes = update->num_bytes;
 685
 686         }
 687
 688         if (update->extent_op) {
 689                 if (!existing->extent_op) {
 690                         existing->extent_op = update->extent_op;
 691                 } else {
 692                         if (update->extent_op->update_key) {
 693                                 memcpy(&existing->extent_op->key,
 694                                        &update->extent_op->key,
 695                                        sizeof(update->extent_op->key));
 696                                 existing->extent_op->update_key = true;
 697                         }
 698                         if (update->extent_op->update_flags) {
 699                                 existing->extent_op->flags_to_set |=
 700                                         update->extent_op->flags_to_set;
 701                                 existing->extent_op->update_flags = true;
 702                         }
 703                         btrfs_free_delayed_extent_op(update->extent_op);
 704                 }
 705         }
 706         /*
 707          * update the reference mod on the head to reflect this new operation,
 708          * only need the lock for this case cause we could be processing it
 709          * currently, for refs we just added we know we're a-ok.
 710          */
 711         old_ref_mod = existing->total_ref_mod;
 712         existing->ref_mod += update->ref_mod;
 713         existing->total_ref_mod += update->ref_mod;
 714
 715         /*
 716          * If we are going to from a positive ref mod to a negative or vice
 717          * versa we need to make sure to adjust pending_csums accordingly.
 718          */
 719         if (existing->is_data) {
 720                 u64 csum_leaves =
 721                         btrfs_csum_bytes_to_leaves(fs_info,
 722                                                    existing->num_bytes);
 723
 724                 if (existing->total_ref_mod >= 0 && old_ref_mod < 0) {
 725                         delayed_refs->pending_csums -= existing->num_bytes;
 726                         btrfs_delayed_refs_rsv_release(fs_info, csum_leaves);
 727                 }
 728                 if (existing->total_ref_mod < 0 && old_ref_mod >= 0) {
 729                         delayed_refs->pending_csums += existing->num_bytes;
 730                         trans->delayed_ref_updates += csum_leaves;
 731                 }
 732         }
 733
 734         spin_unlock(&existing->lock);
 735 }
 736
 737 static void init_delayed_ref_head(struct btrfs_delayed_ref_head *head_ref,
 738                                   struct btrfs_qgroup_extent_record *qrecord,
 739                                   u64 bytenr, u64 num_bytes, u64 ref_root,
 740                                   u64 reserved, int action, bool is_data,
 741                                   bool is_system)
 742 {
 743         int count_mod = 1;
 744         int must_insert_reserved = 0;
 745
 746         /* If reserved is provided, it must be a data extent. */
 747         BUG_ON(!is_data && reserved);
 748
 749         /*
 750          * The head node stores the sum of all the mods, so dropping a ref
 751          * should drop the sum in the head node by one.
 752          */
 753         if (action == BTRFS_UPDATE_DELAYED_HEAD)
 754                 count_mod = 0;
 755         else if (action == BTRFS_DROP_DELAYED_REF)
 756                 count_mod = -1;
 757
 758         /*
 759          * BTRFS_ADD_DELAYED_EXTENT means that we need to update the reserved
 760          * accounting when the extent is finally added, or if a later
 761          * modification deletes the delayed ref without ever inserting the
 762          * extent into the extent allocation tree.  ref->must_insert_reserved
 763          * is the flag used to record that accounting mods are required.
 764          *
 765          * Once we record must_insert_reserved, switch the action to
 766          * BTRFS_ADD_DELAYED_REF because other special casing is not required.
 767          */
 768         if (action == BTRFS_ADD_DELAYED_EXTENT)
 769                 must_insert_reserved = 1;
 770         else
 771                 must_insert_reserved = 0;
 772
 773         refcount_set(&head_ref->refs, 1);
 774         head_ref->bytenr = bytenr;
 775         head_ref->num_bytes = num_bytes;
 776         head_ref->ref_mod = count_mod;
 777         head_ref->must_insert_reserved = must_insert_reserved;
 778         head_ref->is_data = is_data;
 779         head_ref->is_system = is_system;
 780         head_ref->ref_tree = RB_ROOT_CACHED;
 781         INIT_LIST_HEAD(&head_ref->ref_add_list);
 782         RB_CLEAR_NODE(&head_ref->href_node);
 783         head_ref->processing = 0;
 784         head_ref->total_ref_mod = count_mod;
 785         spin_lock_init(&head_ref->lock);
 786         mutex_init(&head_ref->mutex);
 787
 788         if (qrecord) {
 789                 if (ref_root && reserved) {
 790                         qrecord->data_rsv = reserved;
 791                         qrecord->data_rsv_refroot = ref_root;
 792                 }
 793                 qrecord->bytenr = bytenr;
 794                 qrecord->num_bytes = num_bytes;
 795                 qrecord->old_roots = NULL;
 796         }
 797 }
 798
 799 /*
 800  * helper function to actually insert a head node into the rbtree.
 801  * this does all the dirty work in terms of maintaining the correct
 802  * overall modification count.
 803  */
 804 static noinline struct btrfs_delayed_ref_head *
 805 add_delayed_ref_head(struct btrfs_trans_handle *trans,
 806                      struct btrfs_delayed_ref_head *head_ref,
 807                      struct btrfs_qgroup_extent_record *qrecord,
 808                      int action, int *qrecord_inserted_ret)
 809 {
 810         struct btrfs_delayed_ref_head *existing;
 811         struct btrfs_delayed_ref_root *delayed_refs;
 812         int qrecord_inserted = 0;
 813
 814         delayed_refs = &trans->transaction->delayed_refs;
 815
 816         /* Record qgroup extent info if provided */
 817         if (qrecord) {
 818                 if (btrfs_qgroup_trace_extent_nolock(trans->fs_info,
 819                                         delayed_refs, qrecord))
 820                         kfree(qrecord);
 821                 else
 822                         qrecord_inserted = 1;
 823         }
 824
 825         trace_add_delayed_ref_head(trans->fs_info, head_ref, action);
 826
 827         existing = htree_insert(&delayed_refs->href_root,
 828                                 &head_ref->href_node);
 829         if (existing) {
 830                 update_existing_head_ref(trans, existing, head_ref);
 831                 /*
 832                  * we've updated the existing ref, free the newly
 833                  * allocated ref
 834                  */
 835                 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
 836                 head_ref = existing;
 837         } else {
 838                 if (head_ref->is_data && head_ref->ref_mod < 0) {
 839                         delayed_refs->pending_csums += head_ref->num_bytes;
 840                         trans->delayed_ref_updates +=
 841                                 btrfs_csum_bytes_to_leaves(trans->fs_info,
 842                                                            head_ref->num_bytes);
 843                 }
 844                 delayed_refs->num_heads++;
 845                 delayed_refs->num_heads_ready++;
 846                 atomic_inc(&delayed_refs->num_entries);
 847                 trans->delayed_ref_updates++;
 848         }
 849         if (qrecord_inserted_ret)
 850                 *qrecord_inserted_ret = qrecord_inserted;
 851
 852         return head_ref;
 853 }
 854
 855 /*
 856  * init_delayed_ref_common - Initialize the structure which represents a
 857  *                           modification to a an extent.
 858  *
 859  * @fs_info:    Internal to the mounted filesystem mount structure.
 860  *
 861  * @ref:        The structure which is going to be initialized.
 862  *
 863  * @bytenr:     The logical address of the extent for which a modification is
 864  *              going to be recorded.
 865  *
 866  * @num_bytes:  Size of the extent whose modification is being recorded.
 867  *
 868  * @ref_root:   The id of the root where this modification has originated, this
 869  *              can be either one of the well-known metadata trees or the
 870  *              subvolume id which references this extent.
 871  *
 872  * @action:     Can be one of BTRFS_ADD_DELAYED_REF/BTRFS_DROP_DELAYED_REF or
 873  *              BTRFS_ADD_DELAYED_EXTENT
 874  *
 875  * @ref_type:   Holds the type of the extent which is being recorded, can be
 876  *              one of BTRFS_SHARED_BLOCK_REF_KEY/BTRFS_TREE_BLOCK_REF_KEY
 877  *              when recording a metadata extent or BTRFS_SHARED_DATA_REF_KEY/
 878  *              BTRFS_EXTENT_DATA_REF_KEY when recording data extent
 879  */
 880 static void init_delayed_ref_common(struct btrfs_fs_info *fs_info,
 881                                     struct btrfs_delayed_ref_node *ref,
 882                                     u64 bytenr, u64 num_bytes, u64 ref_root,
 883                                     int action, u8 ref_type)
 884 {
 885         u64 seq = 0;
 886
 887         if (action == BTRFS_ADD_DELAYED_EXTENT)
 888                 action = BTRFS_ADD_DELAYED_REF;
 889
 890         if (is_fstree(ref_root))
 891                 seq = atomic64_read(&fs_info->tree_mod_seq);
 892
 893         refcount_set(&ref->refs, 1);
 894         ref->bytenr = bytenr;
 895         ref->num_bytes = num_bytes;
 896         ref->ref_mod = 1;
 897         ref->action = action;
 898         ref->is_head = 0;
 899         ref->in_tree = 1;
 900         ref->seq = seq;
 901         ref->type = ref_type;
 902         RB_CLEAR_NODE(&ref->ref_node);
 903         INIT_LIST_HEAD(&ref->add_list);
 904 }
 905
 906 /*
 907  * add a delayed tree ref.  This does all of the accounting required
 908  * to make sure the delayed ref is eventually processed before this
 909  * transaction commits.
 910  */
 911 int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 912                                struct btrfs_ref *generic_ref,
 913                                struct btrfs_delayed_extent_op *extent_op)
 914 {
 915         struct btrfs_fs_info *fs_info = trans->fs_info;
 916         struct btrfs_delayed_tree_ref *ref;
 917         struct btrfs_delayed_ref_head *head_ref;
 918         struct btrfs_delayed_ref_root *delayed_refs;
 919         struct btrfs_qgroup_extent_record *record = NULL;
 920         int qrecord_inserted;
 921         bool is_system;
 922         int action = generic_ref->action;
 923         int level = generic_ref->tree_ref.level;
 924         int ret;
 925         u64 bytenr = generic_ref->bytenr;
 926         u64 num_bytes = generic_ref->len;
 927         u64 parent = generic_ref->parent;
 928         u8 ref_type;
 929
 930         is_system = (generic_ref->tree_ref.owning_root == BTRFS_CHUNK_TREE_OBJECTID);
 931
 932         ASSERT(generic_ref->type == BTRFS_REF_METADATA && generic_ref->action);
 933         ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS);
 934         if (!ref)
 935                 return -ENOMEM;
 936
 937         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
 938         if (!head_ref) {
 939                 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 940                 return -ENOMEM;
 941         }
 942
 943         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
 944             !generic_ref->skip_qgroup) {
 945                 record = kzalloc(sizeof(*record), GFP_NOFS);
 946                 if (!record) {
 947                         kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 948                         kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
 949                         return -ENOMEM;
 950                 }
 951         }
 952
 953         if (parent)
 954                 ref_type = BTRFS_SHARED_BLOCK_REF_KEY;
 955         else
 956                 ref_type = BTRFS_TREE_BLOCK_REF_KEY;
 957
 958         init_delayed_ref_common(fs_info, &ref->node, bytenr, num_bytes,
 959                                 generic_ref->tree_ref.owning_root, action,
 960                                 ref_type);
 961         ref->root = generic_ref->tree_ref.owning_root;
 962         ref->parent = parent;
 963         ref->level = level;
 964
 965         init_delayed_ref_head(head_ref, record, bytenr, num_bytes,
 966                               generic_ref->tree_ref.owning_root, 0, action,
 967                               false, is_system);
 968         head_ref->extent_op = extent_op;
 969
 970         delayed_refs = &trans->transaction->delayed_refs;
 971         spin_lock(&delayed_refs->lock);
 972
 973         /*
 974          * insert both the head node and the new ref without dropping
 975          * the spin lock
 976          */
 977         head_ref = add_delayed_ref_head(trans, head_ref, record,
 978                                         action, &qrecord_inserted);
 979
 980         ret = insert_delayed_ref(trans, delayed_refs, head_ref, &ref->node);
 981         spin_unlock(&delayed_refs->lock);
 982
 983         /*
 984          * Need to update the delayed_refs_rsv with any changes we may have
 985          * made.
 986          */
 987         btrfs_update_delayed_refs_rsv(trans);
 988
 989         trace_add_delayed_tree_ref(fs_info, &ref->node, ref,
 990                                    action == BTRFS_ADD_DELAYED_EXTENT ?
 991                                    BTRFS_ADD_DELAYED_REF : action);
 992         if (ret > 0)
 993                 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 994
 995         if (qrecord_inserted)
 996                 btrfs_qgroup_trace_extent_post(trans, record);
 997
 998         return 0;
 999 }
1000
1001 /*
1002  * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
1003  */
1004 int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
1005                                struct btrfs_ref *generic_ref,
1006                                u64 reserved)
1007 {
1008         struct btrfs_fs_info *fs_info = trans->fs_info;
1009         struct btrfs_delayed_data_ref *ref;
1010         struct btrfs_delayed_ref_head *head_ref;
1011         struct btrfs_delayed_ref_root *delayed_refs;
1012         struct btrfs_qgroup_extent_record *record = NULL;
1013         int qrecord_inserted;
1014         int action = generic_ref->action;
1015         int ret;
1016         u64 bytenr = generic_ref->bytenr;
1017         u64 num_bytes = generic_ref->len;
1018         u64 parent = generic_ref->parent;
1019         u64 ref_root = generic_ref->data_ref.owning_root;
1020         u64 owner = generic_ref->data_ref.ino;
1021         u64 offset = generic_ref->data_ref.offset;
1022         u8 ref_type;
1023
1024         ASSERT(generic_ref->type == BTRFS_REF_DATA && action);
1025         ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS);
1026         if (!ref)
1027                 return -ENOMEM;
1028
1029         if (parent)
1030                 ref_type = BTRFS_SHARED_DATA_REF_KEY;
1031         else
1032                 ref_type = BTRFS_EXTENT_DATA_REF_KEY;
1033         init_delayed_ref_common(fs_info, &ref->node, bytenr, num_bytes,
1034                                 ref_root, action, ref_type);
1035         ref->root = ref_root;
1036         ref->parent = parent;
1037         ref->objectid = owner;
1038         ref->offset = offset;
1039
1040
1041         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1042         if (!head_ref) {
1043                 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1044                 return -ENOMEM;
1045         }
1046
1047         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
1048             !generic_ref->skip_qgroup) {
1049                 record = kzalloc(sizeof(*record), GFP_NOFS);
1050                 if (!record) {
1051                         kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1052                         kmem_cache_free(btrfs_delayed_ref_head_cachep,
1053                                         head_ref);
1054                         return -ENOMEM;
1055                 }
1056         }
1057
1058         init_delayed_ref_head(head_ref, record, bytenr, num_bytes, ref_root,
1059                               reserved, action, true, false);
1060         head_ref->extent_op = NULL;
1061
1062         delayed_refs = &trans->transaction->delayed_refs;
1063         spin_lock(&delayed_refs->lock);
1064
1065         /*
1066          * insert both the head node and the new ref without dropping
1067          * the spin lock
1068          */
1069         head_ref = add_delayed_ref_head(trans, head_ref, record,
1070                                         action, &qrecord_inserted);
1071
1072         ret = insert_delayed_ref(trans, delayed_refs, head_ref, &ref->node);
1073         spin_unlock(&delayed_refs->lock);
1074
1075         /*
1076          * Need to update the delayed_refs_rsv with any changes we may have
1077          * made.
1078          */
1079         btrfs_update_delayed_refs_rsv(trans);
1080
1081         trace_add_delayed_data_ref(trans->fs_info, &ref->node, ref,
1082                                    action == BTRFS_ADD_DELAYED_EXTENT ?
1083                                    BTRFS_ADD_DELAYED_REF : action);
1084         if (ret > 0)
1085                 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1086
1087
1088         if (qrecord_inserted)
1089                 return btrfs_qgroup_trace_extent_post(trans, record);
1090         return 0;
1091 }
1092
1093 int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
1094                                 u64 bytenr, u64 num_bytes,
1095                                 struct btrfs_delayed_extent_op *extent_op)
1096 {
1097         struct btrfs_delayed_ref_head *head_ref;
1098         struct btrfs_delayed_ref_root *delayed_refs;
1099
1100         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1101         if (!head_ref)
1102                 return -ENOMEM;
1103
1104         init_delayed_ref_head(head_ref, NULL, bytenr, num_bytes, 0, 0,
1105                               BTRFS_UPDATE_DELAYED_HEAD, false, false);
1106         head_ref->extent_op = extent_op;
1107
1108         delayed_refs = &trans->transaction->delayed_refs;
1109         spin_lock(&delayed_refs->lock);
1110
1111         add_delayed_ref_head(trans, head_ref, NULL, BTRFS_UPDATE_DELAYED_HEAD,
1112                              NULL);
1113
1114         spin_unlock(&delayed_refs->lock);
1115
1116         /*
1117          * Need to update the delayed_refs_rsv with any changes we may have
1118          * made.
1119          */
1120         btrfs_update_delayed_refs_rsv(trans);
1121         return 0;
1122 }
1123
1124 /*
1125  * This does a simple search for the head node for a given extent.  Returns the
1126  * head node if found, or NULL if not.
1127  */
1128 struct btrfs_delayed_ref_head *
1129 btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs, u64 bytenr)
1130 {
1131         lockdep_assert_held(&delayed_refs->lock);
1132
1133         return find_ref_head(delayed_refs, bytenr, false);
1134 }
1135
1136 void __cold btrfs_delayed_ref_exit(void)
1137 {
1138         kmem_cache_destroy(btrfs_delayed_ref_head_cachep);
1139         kmem_cache_destroy(btrfs_delayed_tree_ref_cachep);
1140         kmem_cache_destroy(btrfs_delayed_data_ref_cachep);
1141         kmem_cache_destroy(btrfs_delayed_extent_op_cachep);
1142 }
1143
1144 int __init btrfs_delayed_ref_init(void)
1145 {
1146         btrfs_delayed_ref_head_cachep = kmem_cache_create(
1147                                 "btrfs_delayed_ref_head",
1148                                 sizeof(struct btrfs_delayed_ref_head), 0,
1149                                 SLAB_MEM_SPREAD, NULL);
1150         if (!btrfs_delayed_ref_head_cachep)
1151                 goto fail;
1152
1153         btrfs_delayed_tree_ref_cachep = kmem_cache_create(
1154                                 "btrfs_delayed_tree_ref",
1155                                 sizeof(struct btrfs_delayed_tree_ref), 0,
1156                                 SLAB_MEM_SPREAD, NULL);
1157         if (!btrfs_delayed_tree_ref_cachep)
1158                 goto fail;
1159
1160         btrfs_delayed_data_ref_cachep = kmem_cache_create(
1161                                 "btrfs_delayed_data_ref",
1162                                 sizeof(struct btrfs_delayed_data_ref), 0,
1163                                 SLAB_MEM_SPREAD, NULL);
1164         if (!btrfs_delayed_data_ref_cachep)
1165                 goto fail;
1166
1167         btrfs_delayed_extent_op_cachep = kmem_cache_create(
1168                                 "btrfs_delayed_extent_op",
1169                                 sizeof(struct btrfs_delayed_extent_op), 0,
1170                                 SLAB_MEM_SPREAD, NULL);
1171         if (!btrfs_delayed_extent_op_cachep)
1172                 goto fail;
1173
1174         return 0;
1175 fail:
1176         btrfs_delayed_ref_exit();
1177         return -ENOMEM;
1178 }