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
  15 struct kmem_cache *btrfs_delayed_ref_head_cachep;
  16 struct kmem_cache *btrfs_delayed_tree_ref_cachep;
  17 struct kmem_cache *btrfs_delayed_data_ref_cachep;
  18 struct kmem_cache *btrfs_delayed_extent_op_cachep;
  19 /*
  20  * delayed back reference update tracking.  For subvolume trees
  21  * we queue up extent allocations and backref maintenance for
  22  * delayed processing.   This avoids deep call chains where we
  23  * add extents in the middle of btrfs_search_slot, and it allows
  24  * us to buffer up frequently modified backrefs in an rb tree instead
  25  * of hammering updates on the extent allocation tree.
  26  */
  27
  28 bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info)
  29 {
  30         struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
  31         struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
  32         bool ret = false;
  33         u64 reserved;
  34
  35         spin_lock(&global_rsv->lock);
  36         reserved = global_rsv->reserved;
  37         spin_unlock(&global_rsv->lock);
  38
  39         /*
  40          * Since the global reserve is just kind of magic we don't really want
  41          * to rely on it to save our bacon, so if our size is more than the
  42          * delayed_refs_rsv and the global rsv then it's time to think about
  43          * bailing.
  44          */
  45         spin_lock(&delayed_refs_rsv->lock);
  46         reserved += delayed_refs_rsv->reserved;
  47         if (delayed_refs_rsv->size >= reserved)
  48                 ret = true;
  49         spin_unlock(&delayed_refs_rsv->lock);
  50         return ret;
  51 }
  52
  53 int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans)
  54 {
  55         u64 num_entries =
  56                 atomic_read(&trans->transaction->delayed_refs.num_entries);
  57         u64 avg_runtime;
  58         u64 val;
  59
  60         smp_mb();
  61         avg_runtime = trans->fs_info->avg_delayed_ref_runtime;
  62         val = num_entries * avg_runtime;
  63         if (val >= NSEC_PER_SEC)
  64                 return 1;
  65         if (val >= NSEC_PER_SEC / 2)
  66                 return 2;
  67
  68         return btrfs_check_space_for_delayed_refs(trans->fs_info);
  69 }
  70
  71 /**
  72  * Release a ref head's reservation
  73  *
  74  * @fs_info:  the filesystem
  75  * @nr:       number of items to drop
  76  *
  77  * This drops the delayed ref head's count from the delayed refs rsv and frees
  78  * any excess reservation we had.
  79  */
  80 void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr)
  81 {
  82         struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
  83         u64 num_bytes = btrfs_calc_insert_metadata_size(fs_info, nr);
  84         u64 released = 0;
  85
  86         released = btrfs_block_rsv_release(fs_info, block_rsv, num_bytes, NULL);
  87         if (released)
  88                 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
  89                                               0, released, 0);
  90 }
  91
  92 /*
  93  * btrfs_update_delayed_refs_rsv - adjust the size of the delayed refs rsv
  94  * @trans - the trans that may have generated delayed refs
  95  *
  96  * This is to be called anytime we may have adjusted trans->delayed_ref_updates,
  97  * it'll calculate the additional size and add it to the delayed_refs_rsv.
  98  */
  99 void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans)
 100 {
 101         struct btrfs_fs_info *fs_info = trans->fs_info;
 102         struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
 103         u64 num_bytes;
 104
 105         if (!trans->delayed_ref_updates)
 106                 return;
 107
 108         num_bytes = btrfs_calc_insert_metadata_size(fs_info,
 109                                                     trans->delayed_ref_updates);
 110         spin_lock(&delayed_rsv->lock);
 111         delayed_rsv->size += num_bytes;
 112         delayed_rsv->full = 0;
 113         spin_unlock(&delayed_rsv->lock);
 114         trans->delayed_ref_updates = 0;
 115 }
 116
 117 /**
 118  * Transfer bytes to our delayed refs rsv
 119  *
 120  * @fs_info:   the filesystem
 121  * @src:       source block rsv to transfer from
 122  * @num_bytes: number of bytes to transfer
 123  *
 124  * This transfers up to the num_bytes amount from the src rsv to the
 125  * delayed_refs_rsv.  Any extra bytes are returned to the space info.
 126  */
 127 void btrfs_migrate_to_delayed_refs_rsv(struct btrfs_fs_info *fs_info,
 128                                        struct btrfs_block_rsv *src,
 129                                        u64 num_bytes)
 130 {
 131         struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
 132         u64 to_free = 0;
 133
 134         spin_lock(&src->lock);
 135         src->reserved -= num_bytes;
 136         src->size -= num_bytes;
 137         spin_unlock(&src->lock);
 138
 139         spin_lock(&delayed_refs_rsv->lock);
 140         if (delayed_refs_rsv->size > delayed_refs_rsv->reserved) {
 141                 u64 delta = delayed_refs_rsv->size -
 142                         delayed_refs_rsv->reserved;
 143                 if (num_bytes > delta) {
 144                         to_free = num_bytes - delta;
 145                         num_bytes = delta;
 146                 }
 147         } else {
 148                 to_free = num_bytes;
 149                 num_bytes = 0;
 150         }
 151
 152         if (num_bytes)
 153                 delayed_refs_rsv->reserved += num_bytes;
 154         if (delayed_refs_rsv->reserved >= delayed_refs_rsv->size)
 155                 delayed_refs_rsv->full = 1;
 156         spin_unlock(&delayed_refs_rsv->lock);
 157
 158         if (num_bytes)
 159                 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
 160                                               0, num_bytes, 1);
 161         if (to_free)
 162                 btrfs_space_info_free_bytes_may_use(fs_info,
 163                                 delayed_refs_rsv->space_info, to_free);
 164 }
 165
 166 /**
 167  * Refill based on our delayed refs usage
 168  *
 169  * @fs_info: the filesystem
 170  * @flush:   control how we can flush for this reservation.
 171  *
 172  * This will refill the delayed block_rsv up to 1 items size worth of space and
 173  * will return -ENOSPC if we can't make the reservation.
 174  */
 175 int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
 176                                   enum btrfs_reserve_flush_enum flush)
 177 {
 178         struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
 179         u64 limit = btrfs_calc_insert_metadata_size(fs_info, 1);
 180         u64 num_bytes = 0;
 181         int ret = -ENOSPC;
 182
 183         spin_lock(&block_rsv->lock);
 184         if (block_rsv->reserved < block_rsv->size) {
 185                 num_bytes = block_rsv->size - block_rsv->reserved;
 186                 num_bytes = min(num_bytes, limit);
 187         }
 188         spin_unlock(&block_rsv->lock);
 189
 190         if (!num_bytes)
 191                 return 0;
 192
 193         ret = btrfs_reserve_metadata_bytes(fs_info->extent_root, block_rsv,
 194                                            num_bytes, flush);
 195         if (ret)
 196                 return ret;
 197         btrfs_block_rsv_add_bytes(block_rsv, num_bytes, 0);
 198         trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
 199                                       0, num_bytes, 1);
 200         return 0;
 201 }
 202
 203 /*
 204  * compare two delayed tree backrefs with same bytenr and type
 205  */
 206 static int comp_tree_refs(struct btrfs_delayed_tree_ref *ref1,
 207                           struct btrfs_delayed_tree_ref *ref2)
 208 {
 209         if (ref1->node.type == BTRFS_TREE_BLOCK_REF_KEY) {
 210                 if (ref1->root < ref2->root)
 211                         return -1;
 212                 if (ref1->root > ref2->root)
 213                         return 1;
 214         } else {
 215                 if (ref1->parent < ref2->parent)
 216                         return -1;
 217                 if (ref1->parent > ref2->parent)
 218                         return 1;
 219         }
 220         return 0;
 221 }
 222
 223 /*
 224  * compare two delayed data backrefs with same bytenr and type
 225  */
 226 static int comp_data_refs(struct btrfs_delayed_data_ref *ref1,
 227                           struct btrfs_delayed_data_ref *ref2)
 228 {
 229         if (ref1->node.type == BTRFS_EXTENT_DATA_REF_KEY) {
 230                 if (ref1->root < ref2->root)
 231                         return -1;
 232                 if (ref1->root > ref2->root)
 233                         return 1;
 234                 if (ref1->objectid < ref2->objectid)
 235                         return -1;
 236                 if (ref1->objectid > ref2->objectid)
 237                         return 1;
 238                 if (ref1->offset < ref2->offset)
 239                         return -1;
 240                 if (ref1->offset > ref2->offset)
 241                         return 1;
 242         } else {
 243                 if (ref1->parent < ref2->parent)
 244                         return -1;
 245                 if (ref1->parent > ref2->parent)
 246                         return 1;
 247         }
 248         return 0;
 249 }
 250
 251 static int comp_refs(struct btrfs_delayed_ref_node *ref1,
 252                      struct btrfs_delayed_ref_node *ref2,
 253                      bool check_seq)
 254 {
 255         int ret = 0;
 256
 257         if (ref1->type < ref2->type)
 258                 return -1;
 259         if (ref1->type > ref2->type)
 260                 return 1;
 261         if (ref1->type == BTRFS_TREE_BLOCK_REF_KEY ||
 262             ref1->type == BTRFS_SHARED_BLOCK_REF_KEY)
 263                 ret = comp_tree_refs(btrfs_delayed_node_to_tree_ref(ref1),
 264                                      btrfs_delayed_node_to_tree_ref(ref2));
 265         else
 266                 ret = comp_data_refs(btrfs_delayed_node_to_data_ref(ref1),
 267                                      btrfs_delayed_node_to_data_ref(ref2));
 268         if (ret)
 269                 return ret;
 270         if (check_seq) {
 271                 if (ref1->seq < ref2->seq)
 272                         return -1;
 273                 if (ref1->seq > ref2->seq)
 274                         return 1;
 275         }
 276         return 0;
 277 }
 278
 279 /* insert a new ref to head ref rbtree */
 280 static struct btrfs_delayed_ref_head *htree_insert(struct rb_root_cached *root,
 281                                                    struct rb_node *node)
 282 {
 283         struct rb_node **p = &root->rb_root.rb_node;
 284         struct rb_node *parent_node = NULL;
 285         struct btrfs_delayed_ref_head *entry;
 286         struct btrfs_delayed_ref_head *ins;
 287         u64 bytenr;
 288         bool leftmost = true;
 289
 290         ins = rb_entry(node, struct btrfs_delayed_ref_head, href_node);
 291         bytenr = ins->bytenr;
 292         while (*p) {
 293                 parent_node = *p;
 294                 entry = rb_entry(parent_node, struct btrfs_delayed_ref_head,
 295                                  href_node);
 296
 297                 if (bytenr < entry->bytenr) {
 298                         p = &(*p)->rb_left;
 299                 } else if (bytenr > entry->bytenr) {
 300                         p = &(*p)->rb_right;
 301                         leftmost = false;
 302                 } else {
 303                         return entry;
 304                 }
 305         }
 306
 307         rb_link_node(node, parent_node, p);
 308         rb_insert_color_cached(node, root, leftmost);
 309         return NULL;
 310 }
 311
 312 static struct btrfs_delayed_ref_node* tree_insert(struct rb_root_cached *root,
 313                 struct btrfs_delayed_ref_node *ins)
 314 {
 315         struct rb_node **p = &root->rb_root.rb_node;
 316         struct rb_node *node = &ins->ref_node;
 317         struct rb_node *parent_node = NULL;
 318         struct btrfs_delayed_ref_node *entry;
 319         bool leftmost = true;
 320
 321         while (*p) {
 322                 int comp;
 323
 324                 parent_node = *p;
 325                 entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,
 326                                  ref_node);
 327                 comp = comp_refs(ins, entry, true);
 328                 if (comp < 0) {
 329                         p = &(*p)->rb_left;
 330                 } else if (comp > 0) {
 331                         p = &(*p)->rb_right;
 332                         leftmost = false;
 333                 } else {
 334                         return entry;
 335                 }
 336         }
 337
 338         rb_link_node(node, parent_node, p);
 339         rb_insert_color_cached(node, root, leftmost);
 340         return NULL;
 341 }
 342
 343 static struct btrfs_delayed_ref_head *find_first_ref_head(
 344                 struct btrfs_delayed_ref_root *dr)
 345 {
 346         struct rb_node *n;
 347         struct btrfs_delayed_ref_head *entry;
 348
 349         n = rb_first_cached(&dr->href_root);
 350         if (!n)
 351                 return NULL;
 352
 353         entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
 354
 355         return entry;
 356 }
 357
 358 /*
 359  * Find a head entry based on bytenr. This returns the delayed ref head if it
 360  * was able to find one, or NULL if nothing was in that spot.  If return_bigger
 361  * is given, the next bigger entry is returned if no exact match is found.
 362  */
 363 static struct btrfs_delayed_ref_head *find_ref_head(
 364                 struct btrfs_delayed_ref_root *dr, u64 bytenr,
 365                 bool return_bigger)
 366 {
 367         struct rb_root *root = &dr->href_root.rb_root;
 368         struct rb_node *n;
 369         struct btrfs_delayed_ref_head *entry;
 370
 371         n = root->rb_node;
 372         entry = NULL;
 373         while (n) {
 374                 entry = rb_entry(n, struct btrfs_delayed_ref_head, href_node);
 375
 376                 if (bytenr < entry->bytenr)
 377                         n = n->rb_left;
 378                 else if (bytenr > entry->bytenr)
 379                         n = n->rb_right;
 380                 else
 381                         return entry;
 382         }
 383         if (entry && return_bigger) {
 384                 if (bytenr > entry->bytenr) {
 385                         n = rb_next(&entry->href_node);
 386                         if (!n)
 387                                 return NULL;
 388                         entry = rb_entry(n, struct btrfs_delayed_ref_head,
 389                                          href_node);
 390                 }
 391                 return entry;
 392         }
 393         return NULL;
 394 }
 395
 396 int btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
 397                            struct btrfs_delayed_ref_head *head)
 398 {
 399         lockdep_assert_held(&delayed_refs->lock);
 400         if (mutex_trylock(&head->mutex))
 401                 return 0;
 402
 403         refcount_inc(&head->refs);
 404         spin_unlock(&delayed_refs->lock);
 405
 406         mutex_lock(&head->mutex);
 407         spin_lock(&delayed_refs->lock);
 408         if (RB_EMPTY_NODE(&head->href_node)) {
 409                 mutex_unlock(&head->mutex);
 410                 btrfs_put_delayed_ref_head(head);
 411                 return -EAGAIN;
 412         }
 413         btrfs_put_delayed_ref_head(head);
 414         return 0;
 415 }
 416
 417 static inline void drop_delayed_ref(struct btrfs_trans_handle *trans,
 418                                     struct btrfs_delayed_ref_root *delayed_refs,
 419                                     struct btrfs_delayed_ref_head *head,
 420                                     struct btrfs_delayed_ref_node *ref)
 421 {
 422         lockdep_assert_held(&head->lock);
 423         rb_erase_cached(&ref->ref_node, &head->ref_tree);
 424         RB_CLEAR_NODE(&ref->ref_node);
 425         if (!list_empty(&ref->add_list))
 426                 list_del(&ref->add_list);
 427         ref->in_tree = 0;
 428         btrfs_put_delayed_ref(ref);
 429         atomic_dec(&delayed_refs->num_entries);
 430 }
 431
 432 static bool merge_ref(struct btrfs_trans_handle *trans,
 433                       struct btrfs_delayed_ref_root *delayed_refs,
 434                       struct btrfs_delayed_ref_head *head,
 435                       struct btrfs_delayed_ref_node *ref,
 436                       u64 seq)
 437 {
 438         struct btrfs_delayed_ref_node *next;
 439         struct rb_node *node = rb_next(&ref->ref_node);
 440         bool done = false;
 441
 442         while (!done && node) {
 443                 int mod;
 444
 445                 next = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
 446                 node = rb_next(node);
 447                 if (seq && next->seq >= seq)
 448                         break;
 449                 if (comp_refs(ref, next, false))
 450                         break;
 451
 452                 if (ref->action == next->action) {
 453                         mod = next->ref_mod;
 454                 } else {
 455                         if (ref->ref_mod < next->ref_mod) {
 456                                 swap(ref, next);
 457                                 done = true;
 458                         }
 459                         mod = -next->ref_mod;
 460                 }
 461
 462                 drop_delayed_ref(trans, delayed_refs, head, next);
 463                 ref->ref_mod += mod;
 464                 if (ref->ref_mod == 0) {
 465                         drop_delayed_ref(trans, delayed_refs, head, ref);
 466                         done = true;
 467                 } else {
 468                         /*
 469                          * Can't have multiples of the same ref on a tree block.
 470                          */
 471                         WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
 472                                 ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
 473                 }
 474         }
 475
 476         return done;
 477 }
 478
 479 void btrfs_merge_delayed_refs(struct btrfs_trans_handle *trans,
 480                               struct btrfs_delayed_ref_root *delayed_refs,
 481                               struct btrfs_delayed_ref_head *head)
 482 {
 483         struct btrfs_fs_info *fs_info = trans->fs_info;
 484         struct btrfs_delayed_ref_node *ref;
 485         struct rb_node *node;
 486         u64 seq = 0;
 487
 488         lockdep_assert_held(&head->lock);
 489
 490         if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
 491                 return;
 492
 493         /* We don't have too many refs to merge for data. */
 494         if (head->is_data)
 495                 return;
 496
 497         read_lock(&fs_info->tree_mod_log_lock);
 498         if (!list_empty(&fs_info->tree_mod_seq_list)) {
 499                 struct seq_list *elem;
 500
 501                 elem = list_first_entry(&fs_info->tree_mod_seq_list,
 502                                         struct seq_list, list);
 503                 seq = elem->seq;
 504         }
 505         read_unlock(&fs_info->tree_mod_log_lock);
 506
 507 again:
 508         for (node = rb_first_cached(&head->ref_tree); node;
 509              node = rb_next(node)) {
 510                 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
 511                 if (seq && ref->seq >= seq)
 512                         continue;
 513                 if (merge_ref(trans, delayed_refs, head, ref, seq))
 514                         goto again;
 515         }
 516 }
 517
 518 int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq)
 519 {
 520         struct seq_list *elem;
 521         int ret = 0;
 522
 523         read_lock(&fs_info->tree_mod_log_lock);
 524         if (!list_empty(&fs_info->tree_mod_seq_list)) {
 525                 elem = list_first_entry(&fs_info->tree_mod_seq_list,
 526                                         struct seq_list, list);
 527                 if (seq >= elem->seq) {
 528                         btrfs_debug(fs_info,
 529                                 "holding back delayed_ref %#x.%x, lowest is %#x.%x",
 530                                 (u32)(seq >> 32), (u32)seq,
 531                                 (u32)(elem->seq >> 32), (u32)elem->seq);
 532                         ret = 1;
 533                 }
 534         }
 535
 536         read_unlock(&fs_info->tree_mod_log_lock);
 537         return ret;
 538 }
 539
 540 struct btrfs_delayed_ref_head *btrfs_select_ref_head(
 541                 struct btrfs_delayed_ref_root *delayed_refs)
 542 {
 543         struct btrfs_delayed_ref_head *head;
 544
 545 again:
 546         head = find_ref_head(delayed_refs, delayed_refs->run_delayed_start,
 547                              true);
 548         if (!head && delayed_refs->run_delayed_start != 0) {
 549                 delayed_refs->run_delayed_start = 0;
 550                 head = find_first_ref_head(delayed_refs);
 551         }
 552         if (!head)
 553                 return NULL;
 554
 555         while (head->processing) {
 556                 struct rb_node *node;
 557
 558                 node = rb_next(&head->href_node);
 559                 if (!node) {
 560                         if (delayed_refs->run_delayed_start == 0)
 561                                 return NULL;
 562                         delayed_refs->run_delayed_start = 0;
 563                         goto again;
 564                 }
 565                 head = rb_entry(node, struct btrfs_delayed_ref_head,
 566                                 href_node);
 567         }
 568
 569         head->processing = 1;
 570         WARN_ON(delayed_refs->num_heads_ready == 0);
 571         delayed_refs->num_heads_ready--;
 572         delayed_refs->run_delayed_start = head->bytenr +
 573                 head->num_bytes;
 574         return head;
 575 }
 576
 577 void btrfs_delete_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
 578                            struct btrfs_delayed_ref_head *head)
 579 {
 580         lockdep_assert_held(&delayed_refs->lock);
 581         lockdep_assert_held(&head->lock);
 582
 583         rb_erase_cached(&head->href_node, &delayed_refs->href_root);
 584         RB_CLEAR_NODE(&head->href_node);
 585         atomic_dec(&delayed_refs->num_entries);
 586         delayed_refs->num_heads--;
 587         if (head->processing == 0)
 588                 delayed_refs->num_heads_ready--;
 589 }
 590
 591 /*
 592  * Helper to insert the ref_node to the tail or merge with tail.
 593  *
 594  * Return 0 for insert.
 595  * Return >0 for merge.
 596  */
 597 static int insert_delayed_ref(struct btrfs_trans_handle *trans,
 598                               struct btrfs_delayed_ref_root *root,
 599                               struct btrfs_delayed_ref_head *href,
 600                               struct btrfs_delayed_ref_node *ref)
 601 {
 602         struct btrfs_delayed_ref_node *exist;
 603         int mod;
 604         int ret = 0;
 605
 606         spin_lock(&href->lock);
 607         exist = tree_insert(&href->ref_tree, ref);
 608         if (!exist)
 609                 goto inserted;
 610
 611         /* Now we are sure we can merge */
 612         ret = 1;
 613         if (exist->action == ref->action) {
 614                 mod = ref->ref_mod;
 615         } else {
 616                 /* Need to change action */
 617                 if (exist->ref_mod < ref->ref_mod) {
 618                         exist->action = ref->action;
 619                         mod = -exist->ref_mod;
 620                         exist->ref_mod = ref->ref_mod;
 621                         if (ref->action == BTRFS_ADD_DELAYED_REF)
 622                                 list_add_tail(&exist->add_list,
 623                                               &href->ref_add_list);
 624                         else if (ref->action == BTRFS_DROP_DELAYED_REF) {
 625                                 ASSERT(!list_empty(&exist->add_list));
 626                                 list_del(&exist->add_list);
 627                         } else {
 628                                 ASSERT(0);
 629                         }
 630                 } else
 631                         mod = -ref->ref_mod;
 632         }
 633         exist->ref_mod += mod;
 634
 635         /* remove existing tail if its ref_mod is zero */
 636         if (exist->ref_mod == 0)
 637                 drop_delayed_ref(trans, root, href, exist);
 638         spin_unlock(&href->lock);
 639         return ret;
 640 inserted:
 641         if (ref->action == BTRFS_ADD_DELAYED_REF)
 642                 list_add_tail(&ref->add_list, &href->ref_add_list);
 643         atomic_inc(&root->num_entries);
 644         spin_unlock(&href->lock);
 645         return ret;
 646 }
 647
 648 /*
 649  * helper function to update the accounting in the head ref
 650  * existing and update must have the same bytenr
 651  */
 652 static noinline void update_existing_head_ref(struct btrfs_trans_handle *trans,
 653                          struct btrfs_delayed_ref_head *existing,
 654                          struct btrfs_delayed_ref_head *update,
 655                          int *old_ref_mod_ret)
 656 {
 657         struct btrfs_delayed_ref_root *delayed_refs =
 658                 &trans->transaction->delayed_refs;
 659         struct btrfs_fs_info *fs_info = trans->fs_info;
 660         int old_ref_mod;
 661
 662         BUG_ON(existing->is_data != update->is_data);
 663
 664         spin_lock(&existing->lock);
 665         if (update->must_insert_reserved) {
 666                 /* if the extent was freed and then
 667                  * reallocated before the delayed ref
 668                  * entries were processed, we can end up
 669                  * with an existing head ref without
 670                  * the must_insert_reserved flag set.
 671                  * Set it again here
 672                  */
 673                 existing->must_insert_reserved = update->must_insert_reserved;
 674
 675                 /*
 676                  * update the num_bytes so we make sure the accounting
 677                  * is done correctly
 678                  */
 679                 existing->num_bytes = update->num_bytes;
 680
 681         }
 682
 683         if (update->extent_op) {
 684                 if (!existing->extent_op) {
 685                         existing->extent_op = update->extent_op;
 686                 } else {
 687                         if (update->extent_op->update_key) {
 688                                 memcpy(&existing->extent_op->key,
 689                                        &update->extent_op->key,
 690                                        sizeof(update->extent_op->key));
 691                                 existing->extent_op->update_key = true;
 692                         }
 693                         if (update->extent_op->update_flags) {
 694                                 existing->extent_op->flags_to_set |=
 695                                         update->extent_op->flags_to_set;
 696                                 existing->extent_op->update_flags = true;
 697                         }
 698                         btrfs_free_delayed_extent_op(update->extent_op);
 699                 }
 700         }
 701         /*
 702          * update the reference mod on the head to reflect this new operation,
 703          * only need the lock for this case cause we could be processing it
 704          * currently, for refs we just added we know we're a-ok.
 705          */
 706         old_ref_mod = existing->total_ref_mod;
 707         if (old_ref_mod_ret)
 708                 *old_ref_mod_ret = old_ref_mod;
 709         existing->ref_mod += update->ref_mod;
 710         existing->total_ref_mod += update->ref_mod;
 711
 712         /*
 713          * If we are going to from a positive ref mod to a negative or vice
 714          * versa we need to make sure to adjust pending_csums accordingly.
 715          */
 716         if (existing->is_data) {
 717                 u64 csum_leaves =
 718                         btrfs_csum_bytes_to_leaves(fs_info,
 719                                                    existing->num_bytes);
 720
 721                 if (existing->total_ref_mod >= 0 && old_ref_mod < 0) {
 722                         delayed_refs->pending_csums -= existing->num_bytes;
 723                         btrfs_delayed_refs_rsv_release(fs_info, csum_leaves);
 724                 }
 725                 if (existing->total_ref_mod < 0 && old_ref_mod >= 0) {
 726                         delayed_refs->pending_csums += existing->num_bytes;
 727                         trans->delayed_ref_updates += csum_leaves;
 728                 }
 729         }
 730         spin_unlock(&existing->lock);
 731 }
 732
 733 static void init_delayed_ref_head(struct btrfs_delayed_ref_head *head_ref,
 734                                   struct btrfs_qgroup_extent_record *qrecord,
 735                                   u64 bytenr, u64 num_bytes, u64 ref_root,
 736                                   u64 reserved, int action, bool is_data,
 737                                   bool is_system)
 738 {
 739         int count_mod = 1;
 740         int must_insert_reserved = 0;
 741
 742         /* If reserved is provided, it must be a data extent. */
 743         BUG_ON(!is_data && reserved);
 744
 745         /*
 746          * The head node stores the sum of all the mods, so dropping a ref
 747          * should drop the sum in the head node by one.
 748          */
 749         if (action == BTRFS_UPDATE_DELAYED_HEAD)
 750                 count_mod = 0;
 751         else if (action == BTRFS_DROP_DELAYED_REF)
 752                 count_mod = -1;
 753
 754         /*
 755          * BTRFS_ADD_DELAYED_EXTENT means that we need to update the reserved
 756          * accounting when the extent is finally added, or if a later
 757          * modification deletes the delayed ref without ever inserting the
 758          * extent into the extent allocation tree.  ref->must_insert_reserved
 759          * is the flag used to record that accounting mods are required.
 760          *
 761          * Once we record must_insert_reserved, switch the action to
 762          * BTRFS_ADD_DELAYED_REF because other special casing is not required.
 763          */
 764         if (action == BTRFS_ADD_DELAYED_EXTENT)
 765                 must_insert_reserved = 1;
 766         else
 767                 must_insert_reserved = 0;
 768
 769         refcount_set(&head_ref->refs, 1);
 770         head_ref->bytenr = bytenr;
 771         head_ref->num_bytes = num_bytes;
 772         head_ref->ref_mod = count_mod;
 773         head_ref->must_insert_reserved = must_insert_reserved;
 774         head_ref->is_data = is_data;
 775         head_ref->is_system = is_system;
 776         head_ref->ref_tree = RB_ROOT_CACHED;
 777         INIT_LIST_HEAD(&head_ref->ref_add_list);
 778         RB_CLEAR_NODE(&head_ref->href_node);
 779         head_ref->processing = 0;
 780         head_ref->total_ref_mod = count_mod;
 781         spin_lock_init(&head_ref->lock);
 782         mutex_init(&head_ref->mutex);
 783
 784         if (qrecord) {
 785                 if (ref_root && reserved) {
 786                         qrecord->data_rsv = reserved;
 787                         qrecord->data_rsv_refroot = ref_root;
 788                 }
 789                 qrecord->bytenr = bytenr;
 790                 qrecord->num_bytes = num_bytes;
 791                 qrecord->old_roots = NULL;
 792         }
 793 }
 794
 795 /*
 796  * helper function to actually insert a head node into the rbtree.
 797  * this does all the dirty work in terms of maintaining the correct
 798  * overall modification count.
 799  */
 800 static noinline struct btrfs_delayed_ref_head *
 801 add_delayed_ref_head(struct btrfs_trans_handle *trans,
 802                      struct btrfs_delayed_ref_head *head_ref,
 803                      struct btrfs_qgroup_extent_record *qrecord,
 804                      int action, int *qrecord_inserted_ret,
 805                      int *old_ref_mod, int *new_ref_mod)
 806 {
 807         struct btrfs_delayed_ref_head *existing;
 808         struct btrfs_delayed_ref_root *delayed_refs;
 809         int qrecord_inserted = 0;
 810
 811         delayed_refs = &trans->transaction->delayed_refs;
 812
 813         /* Record qgroup extent info if provided */
 814         if (qrecord) {
 815                 if (btrfs_qgroup_trace_extent_nolock(trans->fs_info,
 816                                         delayed_refs, qrecord))
 817                         kfree(qrecord);
 818                 else
 819                         qrecord_inserted = 1;
 820         }
 821
 822         trace_add_delayed_ref_head(trans->fs_info, head_ref, action);
 823
 824         existing = htree_insert(&delayed_refs->href_root,
 825                                 &head_ref->href_node);
 826         if (existing) {
 827                 update_existing_head_ref(trans, existing, head_ref,
 828                                          old_ref_mod);
 829                 /*
 830                  * we've updated the existing ref, free the newly
 831                  * allocated ref
 832                  */
 833                 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
 834                 head_ref = existing;
 835         } else {
 836                 if (old_ref_mod)
 837                         *old_ref_mod = 0;
 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         if (new_ref_mod)
 852                 *new_ref_mod = head_ref->total_ref_mod;
 853
 854         return head_ref;
 855 }
 856
 857 /*
 858  * init_delayed_ref_common - Initialize the structure which represents a
 859  *                           modification to a an extent.
 860  *
 861  * @fs_info:    Internal to the mounted filesystem mount structure.
 862  *
 863  * @ref:        The structure which is going to be initialized.
 864  *
 865  * @bytenr:     The logical address of the extent for which a modification is
 866  *              going to be recorded.
 867  *
 868  * @num_bytes:  Size of the extent whose modification is being recorded.
 869  *
 870  * @ref_root:   The id of the root where this modification has originated, this
 871  *              can be either one of the well-known metadata trees or the
 872  *              subvolume id which references this extent.
 873  *
 874  * @action:     Can be one of BTRFS_ADD_DELAYED_REF/BTRFS_DROP_DELAYED_REF or
 875  *              BTRFS_ADD_DELAYED_EXTENT
 876  *
 877  * @ref_type:   Holds the type of the extent which is being recorded, can be
 878  *              one of BTRFS_SHARED_BLOCK_REF_KEY/BTRFS_TREE_BLOCK_REF_KEY
 879  *              when recording a metadata extent or BTRFS_SHARED_DATA_REF_KEY/
 880  *              BTRFS_EXTENT_DATA_REF_KEY when recording data extent
 881  */
 882 static void init_delayed_ref_common(struct btrfs_fs_info *fs_info,
 883                                     struct btrfs_delayed_ref_node *ref,
 884                                     u64 bytenr, u64 num_bytes, u64 ref_root,
 885                                     int action, u8 ref_type)
 886 {
 887         u64 seq = 0;
 888
 889         if (action == BTRFS_ADD_DELAYED_EXTENT)
 890                 action = BTRFS_ADD_DELAYED_REF;
 891
 892         if (is_fstree(ref_root))
 893                 seq = atomic64_read(&fs_info->tree_mod_seq);
 894
 895         refcount_set(&ref->refs, 1);
 896         ref->bytenr = bytenr;
 897         ref->num_bytes = num_bytes;
 898         ref->ref_mod = 1;
 899         ref->action = action;
 900         ref->is_head = 0;
 901         ref->in_tree = 1;
 902         ref->seq = seq;
 903         ref->type = ref_type;
 904         RB_CLEAR_NODE(&ref->ref_node);
 905         INIT_LIST_HEAD(&ref->add_list);
 906 }
 907
 908 /*
 909  * add a delayed tree ref.  This does all of the accounting required
 910  * to make sure the delayed ref is eventually processed before this
 911  * transaction commits.
 912  */
 913 int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
 914                                struct btrfs_ref *generic_ref,
 915                                struct btrfs_delayed_extent_op *extent_op,
 916                                int *old_ref_mod, int *new_ref_mod)
 917 {
 918         struct btrfs_fs_info *fs_info = trans->fs_info;
 919         struct btrfs_delayed_tree_ref *ref;
 920         struct btrfs_delayed_ref_head *head_ref;
 921         struct btrfs_delayed_ref_root *delayed_refs;
 922         struct btrfs_qgroup_extent_record *record = NULL;
 923         int qrecord_inserted;
 924         bool is_system;
 925         int action = generic_ref->action;
 926         int level = generic_ref->tree_ref.level;
 927         int ret;
 928         u64 bytenr = generic_ref->bytenr;
 929         u64 num_bytes = generic_ref->len;
 930         u64 parent = generic_ref->parent;
 931         u8 ref_type;
 932
 933         is_system = (generic_ref->real_root == BTRFS_CHUNK_TREE_OBJECTID);
 934
 935         ASSERT(generic_ref->type == BTRFS_REF_METADATA && generic_ref->action);
 936         BUG_ON(extent_op && extent_op->is_data);
 937         ref = kmem_cache_alloc(btrfs_delayed_tree_ref_cachep, GFP_NOFS);
 938         if (!ref)
 939                 return -ENOMEM;
 940
 941         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
 942         if (!head_ref) {
 943                 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 944                 return -ENOMEM;
 945         }
 946
 947         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
 948             is_fstree(generic_ref->real_root) &&
 949             is_fstree(generic_ref->tree_ref.root) &&
 950             !generic_ref->skip_qgroup) {
 951                 record = kzalloc(sizeof(*record), GFP_NOFS);
 952                 if (!record) {
 953                         kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
 954                         kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
 955                         return -ENOMEM;
 956                 }
 957         }
 958
 959         if (parent)
 960                 ref_type = BTRFS_SHARED_BLOCK_REF_KEY;
 961         else
 962                 ref_type = BTRFS_TREE_BLOCK_REF_KEY;
 963
 964         init_delayed_ref_common(fs_info, &ref->node, bytenr, num_bytes,
 965                                 generic_ref->tree_ref.root, action, ref_type);
 966         ref->root = generic_ref->tree_ref.root;
 967         ref->parent = parent;
 968         ref->level = level;
 969
 970         init_delayed_ref_head(head_ref, record, bytenr, num_bytes,
 971                               generic_ref->tree_ref.root, 0, action, false,
 972                               is_system);
 973         head_ref->extent_op = extent_op;
 974
 975         delayed_refs = &trans->transaction->delayed_refs;
 976         spin_lock(&delayed_refs->lock);
 977
 978         /*
 979          * insert both the head node and the new ref without dropping
 980          * the spin lock
 981          */
 982         head_ref = add_delayed_ref_head(trans, head_ref, record,
 983                                         action, &qrecord_inserted,
 984                                         old_ref_mod, new_ref_mod);
 985
 986         ret = insert_delayed_ref(trans, delayed_refs, head_ref, &ref->node);
 987         spin_unlock(&delayed_refs->lock);
 988
 989         /*
 990          * Need to update the delayed_refs_rsv with any changes we may have
 991          * made.
 992          */
 993         btrfs_update_delayed_refs_rsv(trans);
 994
 995         trace_add_delayed_tree_ref(fs_info, &ref->node, ref,
 996                                    action == BTRFS_ADD_DELAYED_EXTENT ?
 997                                    BTRFS_ADD_DELAYED_REF : action);
 998         if (ret > 0)
 999                 kmem_cache_free(btrfs_delayed_tree_ref_cachep, ref);
1000
1001         if (qrecord_inserted)
1002                 btrfs_qgroup_trace_extent_post(fs_info, record);
1003
1004         return 0;
1005 }
1006
1007 /*
1008  * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
1009  */
1010 int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
1011                                struct btrfs_ref *generic_ref,
1012                                u64 reserved, int *old_ref_mod,
1013                                int *new_ref_mod)
1014 {
1015         struct btrfs_fs_info *fs_info = trans->fs_info;
1016         struct btrfs_delayed_data_ref *ref;
1017         struct btrfs_delayed_ref_head *head_ref;
1018         struct btrfs_delayed_ref_root *delayed_refs;
1019         struct btrfs_qgroup_extent_record *record = NULL;
1020         int qrecord_inserted;
1021         int action = generic_ref->action;
1022         int ret;
1023         u64 bytenr = generic_ref->bytenr;
1024         u64 num_bytes = generic_ref->len;
1025         u64 parent = generic_ref->parent;
1026         u64 ref_root = generic_ref->data_ref.ref_root;
1027         u64 owner = generic_ref->data_ref.ino;
1028         u64 offset = generic_ref->data_ref.offset;
1029         u8 ref_type;
1030
1031         ASSERT(generic_ref->type == BTRFS_REF_DATA && action);
1032         ref = kmem_cache_alloc(btrfs_delayed_data_ref_cachep, GFP_NOFS);
1033         if (!ref)
1034                 return -ENOMEM;
1035
1036         if (parent)
1037                 ref_type = BTRFS_SHARED_DATA_REF_KEY;
1038         else
1039                 ref_type = BTRFS_EXTENT_DATA_REF_KEY;
1040         init_delayed_ref_common(fs_info, &ref->node, bytenr, num_bytes,
1041                                 ref_root, action, ref_type);
1042         ref->root = ref_root;
1043         ref->parent = parent;
1044         ref->objectid = owner;
1045         ref->offset = offset;
1046
1047
1048         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1049         if (!head_ref) {
1050                 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1051                 return -ENOMEM;
1052         }
1053
1054         if (test_bit(BTRFS_FS_QUOTA_ENABLED, &fs_info->flags) &&
1055             is_fstree(ref_root) &&
1056             is_fstree(generic_ref->real_root) &&
1057             !generic_ref->skip_qgroup) {
1058                 record = kzalloc(sizeof(*record), GFP_NOFS);
1059                 if (!record) {
1060                         kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1061                         kmem_cache_free(btrfs_delayed_ref_head_cachep,
1062                                         head_ref);
1063                         return -ENOMEM;
1064                 }
1065         }
1066
1067         init_delayed_ref_head(head_ref, record, bytenr, num_bytes, ref_root,
1068                               reserved, action, true, false);
1069         head_ref->extent_op = NULL;
1070
1071         delayed_refs = &trans->transaction->delayed_refs;
1072         spin_lock(&delayed_refs->lock);
1073
1074         /*
1075          * insert both the head node and the new ref without dropping
1076          * the spin lock
1077          */
1078         head_ref = add_delayed_ref_head(trans, head_ref, record,
1079                                         action, &qrecord_inserted,
1080                                         old_ref_mod, new_ref_mod);
1081
1082         ret = insert_delayed_ref(trans, delayed_refs, head_ref, &ref->node);
1083         spin_unlock(&delayed_refs->lock);
1084
1085         /*
1086          * Need to update the delayed_refs_rsv with any changes we may have
1087          * made.
1088          */
1089         btrfs_update_delayed_refs_rsv(trans);
1090
1091         trace_add_delayed_data_ref(trans->fs_info, &ref->node, ref,
1092                                    action == BTRFS_ADD_DELAYED_EXTENT ?
1093                                    BTRFS_ADD_DELAYED_REF : action);
1094         if (ret > 0)
1095                 kmem_cache_free(btrfs_delayed_data_ref_cachep, ref);
1096
1097
1098         if (qrecord_inserted)
1099                 return btrfs_qgroup_trace_extent_post(fs_info, record);
1100         return 0;
1101 }
1102
1103 int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
1104                                 u64 bytenr, u64 num_bytes,
1105                                 struct btrfs_delayed_extent_op *extent_op)
1106 {
1107         struct btrfs_delayed_ref_head *head_ref;
1108         struct btrfs_delayed_ref_root *delayed_refs;
1109
1110         head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1111         if (!head_ref)
1112                 return -ENOMEM;
1113
1114         init_delayed_ref_head(head_ref, NULL, bytenr, num_bytes, 0, 0,
1115                               BTRFS_UPDATE_DELAYED_HEAD, extent_op->is_data,
1116                               false);
1117         head_ref->extent_op = extent_op;
1118
1119         delayed_refs = &trans->transaction->delayed_refs;
1120         spin_lock(&delayed_refs->lock);
1121
1122         add_delayed_ref_head(trans, head_ref, NULL, BTRFS_UPDATE_DELAYED_HEAD,
1123                              NULL, NULL, NULL);
1124
1125         spin_unlock(&delayed_refs->lock);
1126
1127         /*
1128          * Need to update the delayed_refs_rsv with any changes we may have
1129          * made.
1130          */
1131         btrfs_update_delayed_refs_rsv(trans);
1132         return 0;
1133 }
1134
1135 /*
1136  * This does a simple search for the head node for a given extent.  Returns the
1137  * head node if found, or NULL if not.
1138  */
1139 struct btrfs_delayed_ref_head *
1140 btrfs_find_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs, u64 bytenr)
1141 {
1142         lockdep_assert_held(&delayed_refs->lock);
1143
1144         return find_ref_head(delayed_refs, bytenr, false);
1145 }
1146
1147 void __cold btrfs_delayed_ref_exit(void)
1148 {
1149         kmem_cache_destroy(btrfs_delayed_ref_head_cachep);
1150         kmem_cache_destroy(btrfs_delayed_tree_ref_cachep);
1151         kmem_cache_destroy(btrfs_delayed_data_ref_cachep);
1152         kmem_cache_destroy(btrfs_delayed_extent_op_cachep);
1153 }
1154
1155 int __init btrfs_delayed_ref_init(void)
1156 {
1157         btrfs_delayed_ref_head_cachep = kmem_cache_create(
1158                                 "btrfs_delayed_ref_head",
1159                                 sizeof(struct btrfs_delayed_ref_head), 0,
1160                                 SLAB_MEM_SPREAD, NULL);
1161         if (!btrfs_delayed_ref_head_cachep)
1162                 goto fail;
1163
1164         btrfs_delayed_tree_ref_cachep = kmem_cache_create(
1165                                 "btrfs_delayed_tree_ref",
1166                                 sizeof(struct btrfs_delayed_tree_ref), 0,
1167                                 SLAB_MEM_SPREAD, NULL);
1168         if (!btrfs_delayed_tree_ref_cachep)
1169                 goto fail;
1170
1171         btrfs_delayed_data_ref_cachep = kmem_cache_create(
1172                                 "btrfs_delayed_data_ref",
1173                                 sizeof(struct btrfs_delayed_data_ref), 0,
1174                                 SLAB_MEM_SPREAD, NULL);
1175         if (!btrfs_delayed_data_ref_cachep)
1176                 goto fail;
1177
1178         btrfs_delayed_extent_op_cachep = kmem_cache_create(
1179                                 "btrfs_delayed_extent_op",
1180                                 sizeof(struct btrfs_delayed_extent_op), 0,
1181                                 SLAB_MEM_SPREAD, NULL);
1182         if (!btrfs_delayed_extent_op_cachep)
1183                 goto fail;
1184
1185         return 0;
1186 fail:
1187         btrfs_delayed_ref_exit();
1188         return -ENOMEM;
1189 }