projects/linux·blob master · refs

1406 lines · 40.2 KiB

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 "messages.h"
10#include "ctree.h"
11#include "delayed-ref.h"
12#include "extent-tree.h"
13#include "transaction.h"
14#include "qgroup.h"
15#include "space-info.h"
16#include "tree-mod-log.h"
17#include "fs.h"
18
19struct kmem_cache *btrfs_delayed_ref_head_cachep;
20struct kmem_cache *btrfs_delayed_ref_node_cachep;
21struct kmem_cache *btrfs_delayed_extent_op_cachep;
22/*
23 * delayed back reference update tracking. For subvolume trees
24 * we queue up extent allocations and backref maintenance for
25 * delayed processing. This avoids deep call chains where we
26 * add extents in the middle of btrfs_search_slot, and it allows
27 * us to buffer up frequently modified backrefs in an rb tree instead
28 * of hammering updates on the extent allocation tree.
29 */
30
31bool btrfs_check_space_for_delayed_refs(struct btrfs_fs_info *fs_info)
32{
33 struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
34 struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
35 bool ret = false;
36 u64 reserved;
37
38 spin_lock(&global_rsv->lock);
39 reserved = global_rsv->reserved;
40 spin_unlock(&global_rsv->lock);
41
42 /*
43 * Since the global reserve is just kind of magic we don't really want
44 * to rely on it to save our bacon, so if our size is more than the
45 * delayed_refs_rsv and the global rsv then it's time to think about
46 * bailing.
47 */
48 spin_lock(&delayed_refs_rsv->lock);
49 reserved += delayed_refs_rsv->reserved;
50 if (delayed_refs_rsv->size >= reserved)
51 ret = true;
52 spin_unlock(&delayed_refs_rsv->lock);
53 return ret;
54}
55
56/*
57 * Release a ref head's reservation.
58 *
59 * @fs_info: the filesystem
60 * @nr_refs: number of delayed refs to drop
61 * @nr_csums: number of csum items to drop
62 *
63 * Drops the delayed ref head's count from the delayed refs rsv and free any
64 * excess reservation we had.
65 */
66void btrfs_delayed_refs_rsv_release(struct btrfs_fs_info *fs_info, int nr_refs, int nr_csums)
67{
68 struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
69 u64 num_bytes;
70 u64 released;
71
72 num_bytes = btrfs_calc_delayed_ref_bytes(fs_info, nr_refs);
73 num_bytes += btrfs_calc_delayed_ref_csum_bytes(fs_info, nr_csums);
74
75 released = btrfs_block_rsv_release(fs_info, block_rsv, num_bytes, NULL);
76 if (released)
77 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
78 0, released, 0);
79}
80
81/*
82 * Adjust the size of the delayed refs rsv.
83 *
84 * This is to be called anytime we may have adjusted trans->delayed_ref_updates
85 * or trans->delayed_ref_csum_deletions, it'll calculate the additional size and
86 * add it to the delayed_refs_rsv.
87 */
88void btrfs_update_delayed_refs_rsv(struct btrfs_trans_handle *trans)
89{
90 struct btrfs_fs_info *fs_info = trans->fs_info;
91 struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
92 struct btrfs_block_rsv *local_rsv = &trans->delayed_rsv;
93 u64 num_bytes;
94 u64 reserved_bytes;
95
96 if (btrfs_is_testing(fs_info))
97 return;
98
99 num_bytes = btrfs_calc_delayed_ref_bytes(fs_info, trans->delayed_ref_updates);
100 num_bytes += btrfs_calc_delayed_ref_csum_bytes(fs_info,
101 trans->delayed_ref_csum_deletions);
102
103 if (num_bytes == 0)
104 return;
105
106 /*
107 * Try to take num_bytes from the transaction's local delayed reserve.
108 * If not possible, try to take as much as it's available. If the local
109 * reserve doesn't have enough reserved space, the delayed refs reserve
110 * will be refilled next time btrfs_delayed_refs_rsv_refill() is called
111 * by someone or if a transaction commit is triggered before that, the
112 * global block reserve will be used. We want to minimize using the
113 * global block reserve for cases we can account for in advance, to
114 * avoid exhausting it and reach -ENOSPC during a transaction commit.
115 */
116 spin_lock(&local_rsv->lock);
117 reserved_bytes = min(num_bytes, local_rsv->reserved);
118 local_rsv->reserved -= reserved_bytes;
119 local_rsv->full = (local_rsv->reserved >= local_rsv->size);
120 spin_unlock(&local_rsv->lock);
121
122 spin_lock(&delayed_rsv->lock);
123 delayed_rsv->size += num_bytes;
124 delayed_rsv->reserved += reserved_bytes;
125 delayed_rsv->full = (delayed_rsv->reserved >= delayed_rsv->size);
126 spin_unlock(&delayed_rsv->lock);
127 trans->delayed_ref_updates = 0;
128 trans->delayed_ref_csum_deletions = 0;
129}
130
131/*
132 * Adjust the size of the delayed refs block reserve for 1 block group item
133 * insertion, used after allocating a block group.
134 */
135void btrfs_inc_delayed_refs_rsv_bg_inserts(struct btrfs_fs_info *fs_info)
136{
137 struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
138
139 spin_lock(&delayed_rsv->lock);
140 /*
141 * Inserting a block group item does not require changing the free space
142 * tree, only the extent tree or the block group tree, so this is all we
143 * need.
144 */
145 delayed_rsv->size += btrfs_calc_insert_metadata_size(fs_info, 1);
146 delayed_rsv->full = false;
147 spin_unlock(&delayed_rsv->lock);
148}
149
150/*
151 * Adjust the size of the delayed refs block reserve to release space for 1
152 * block group item insertion.
153 */
154void btrfs_dec_delayed_refs_rsv_bg_inserts(struct btrfs_fs_info *fs_info)
155{
156 struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
157 const u64 num_bytes = btrfs_calc_insert_metadata_size(fs_info, 1);
158 u64 released;
159
160 released = btrfs_block_rsv_release(fs_info, delayed_rsv, num_bytes, NULL);
161 if (released > 0)
162 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
163 0, released, 0);
164}
165
166/*
167 * Adjust the size of the delayed refs block reserve for 1 block group item
168 * update.
169 */
170void btrfs_inc_delayed_refs_rsv_bg_updates(struct btrfs_fs_info *fs_info)
171{
172 struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
173
174 spin_lock(&delayed_rsv->lock);
175 /*
176 * Updating a block group item does not result in new nodes/leaves and
177 * does not require changing the free space tree, only the extent tree
178 * or the block group tree, so this is all we need.
179 */
180 delayed_rsv->size += btrfs_calc_metadata_size(fs_info, 1);
181 delayed_rsv->full = false;
182 spin_unlock(&delayed_rsv->lock);
183}
184
185/*
186 * Adjust the size of the delayed refs block reserve to release space for 1
187 * block group item update.
188 */
189void btrfs_dec_delayed_refs_rsv_bg_updates(struct btrfs_fs_info *fs_info)
190{
191 struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_refs_rsv;
192 const u64 num_bytes = btrfs_calc_metadata_size(fs_info, 1);
193 u64 released;
194
195 released = btrfs_block_rsv_release(fs_info, delayed_rsv, num_bytes, NULL);
196 if (released > 0)
197 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv",
198 0, released, 0);
199}
200
201/*
202 * Refill based on our delayed refs usage.
203 *
204 * @fs_info: the filesystem
205 * @flush: control how we can flush for this reservation.
206 *
207 * This will refill the delayed block_rsv up to 1 items size worth of space and
208 * will return -ENOSPC if we can't make the reservation.
209 */
210static int btrfs_zoned_cap_metadata_reservation(struct btrfs_space_info *space_info)
211{
212 struct btrfs_fs_info *fs_info = space_info->fs_info;
213 struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
214 u64 usable;
215 u64 cap;
216 int ret = 0;
217
218 if (!btrfs_is_zoned(fs_info))
219 return 0;
220
221 spin_lock(&space_info->lock);
222 usable = space_info->total_bytes - space_info->bytes_zone_unusable;
223 spin_unlock(&space_info->lock);
224 cap = usable >> 1;
225
226 spin_lock(&block_rsv->lock);
227 if (block_rsv->size > cap)
228 ret = -EAGAIN;
229 spin_unlock(&block_rsv->lock);
230
231 return ret;
232}
233
234int btrfs_delayed_refs_rsv_refill(struct btrfs_fs_info *fs_info,
235 enum btrfs_reserve_flush_enum flush)
236{
237 struct btrfs_block_rsv *block_rsv = &fs_info->delayed_refs_rsv;
238 struct btrfs_space_info *space_info = block_rsv->space_info;
239 u64 limit = btrfs_calc_delayed_ref_bytes(fs_info, 1);
240 u64 num_bytes = 0;
241 u64 refilled_bytes;
242 u64 to_free;
243 int ret = -ENOSPC;
244
245 spin_lock(&block_rsv->lock);
246 if (block_rsv->reserved < block_rsv->size) {
247 num_bytes = block_rsv->size - block_rsv->reserved;
248 num_bytes = min(num_bytes, limit);
249 }
250 spin_unlock(&block_rsv->lock);
251
252 if (!num_bytes)
253 return 0;
254
255 ret = btrfs_zoned_cap_metadata_reservation(space_info);
256 if (ret)
257 return ret;
258
259 ret = btrfs_reserve_metadata_bytes(space_info, num_bytes, flush);
260 if (ret)
261 return ret;
262
263 /*
264 * We may have raced with someone else, so check again if we the block
265 * reserve is still not full and release any excess space.
266 */
267 spin_lock(&block_rsv->lock);
268 if (block_rsv->reserved < block_rsv->size) {
269 u64 needed = block_rsv->size - block_rsv->reserved;
270
271 if (num_bytes >= needed) {
272 block_rsv->reserved += needed;
273 block_rsv->full = true;
274 to_free = num_bytes - needed;
275 refilled_bytes = needed;
276 } else {
277 block_rsv->reserved += num_bytes;
278 to_free = 0;
279 refilled_bytes = num_bytes;
280 }
281 } else {
282 to_free = num_bytes;
283 refilled_bytes = 0;
284 }
285 spin_unlock(&block_rsv->lock);
286
287 if (to_free > 0)
288 btrfs_space_info_free_bytes_may_use(space_info, to_free);
289
290 if (refilled_bytes > 0)
291 trace_btrfs_space_reservation(fs_info, "delayed_refs_rsv", 0,
292 refilled_bytes, 1);
293 return 0;
294}
295
296/*
297 * compare two delayed data backrefs with same bytenr and type
298 */
299static int comp_data_refs(const struct btrfs_delayed_ref_node *ref1,
300 const struct btrfs_delayed_ref_node *ref2)
301{
302 if (ref1->data_ref.objectid < ref2->data_ref.objectid)
303 return -1;
304 if (ref1->data_ref.objectid > ref2->data_ref.objectid)
305 return 1;
306 if (ref1->data_ref.offset < ref2->data_ref.offset)
307 return -1;
308 if (ref1->data_ref.offset > ref2->data_ref.offset)
309 return 1;
310 return 0;
311}
312
313static int comp_refs(const struct btrfs_delayed_ref_node *ref1,
314 const struct btrfs_delayed_ref_node *ref2,
315 bool check_seq)
316{
317 int ret = 0;
318
319 if (ref1->type < ref2->type)
320 return -1;
321 if (ref1->type > ref2->type)
322 return 1;
323 if (ref1->type == BTRFS_SHARED_BLOCK_REF_KEY ||
324 ref1->type == BTRFS_SHARED_DATA_REF_KEY) {
325 if (ref1->parent < ref2->parent)
326 return -1;
327 if (ref1->parent > ref2->parent)
328 return 1;
329 } else {
330 if (ref1->ref_root < ref2->ref_root)
331 return -1;
332 if (ref1->ref_root > ref2->ref_root)
333 return 1;
334 if (ref1->type == BTRFS_EXTENT_DATA_REF_KEY)
335 ret = comp_data_refs(ref1, ref2);
336 }
337 if (ret)
338 return ret;
339 if (check_seq) {
340 if (ref1->seq < ref2->seq)
341 return -1;
342 if (ref1->seq > ref2->seq)
343 return 1;
344 }
345 return 0;
346}
347
348static int cmp_refs_node(const struct rb_node *new, const struct rb_node *exist)
349{
350 const struct btrfs_delayed_ref_node *new_node =
351 rb_entry(new, struct btrfs_delayed_ref_node, ref_node);
352 const struct btrfs_delayed_ref_node *exist_node =
353 rb_entry(exist, struct btrfs_delayed_ref_node, ref_node);
354
355 return comp_refs(new_node, exist_node, true);
356}
357
358static struct btrfs_delayed_ref_node* tree_insert(struct rb_root_cached *root,
359 struct btrfs_delayed_ref_node *ins)
360{
361 struct rb_node *node = &ins->ref_node;
362 struct rb_node *exist = rb_find_add_cached(node, root, cmp_refs_node);
363
364 return rb_entry_safe(exist, struct btrfs_delayed_ref_node, ref_node);
365}
366
367static struct btrfs_delayed_ref_head *find_first_ref_head(
368 struct btrfs_delayed_ref_root *dr)
369{
370 unsigned long from = 0;
371
372 lockdep_assert_held(&dr->lock);
373
374 return xa_find(&dr->head_refs, &from, ULONG_MAX, XA_PRESENT);
375}
376
377static bool btrfs_delayed_ref_lock(struct btrfs_delayed_ref_root *delayed_refs,
378 struct btrfs_delayed_ref_head *head)
379{
380 lockdep_assert_held(&delayed_refs->lock);
381 if (mutex_trylock(&head->mutex))
382 return true;
383
384 refcount_inc(&head->refs);
385 spin_unlock(&delayed_refs->lock);
386
387 mutex_lock(&head->mutex);
388 spin_lock(&delayed_refs->lock);
389 if (!head->tracked) {
390 mutex_unlock(&head->mutex);
391 btrfs_put_delayed_ref_head(head);
392 return false;
393 }
394 btrfs_put_delayed_ref_head(head);
395 return true;
396}
397
398static inline void drop_delayed_ref(struct btrfs_fs_info *fs_info,
399 struct btrfs_delayed_ref_root *delayed_refs,
400 struct btrfs_delayed_ref_head *head,
401 struct btrfs_delayed_ref_node *ref)
402{
403 lockdep_assert_held(&head->lock);
404 rb_erase_cached(&ref->ref_node, &head->ref_tree);
405 RB_CLEAR_NODE(&ref->ref_node);
406 if (!list_empty(&ref->add_list))
407 list_del(&ref->add_list);
408 btrfs_put_delayed_ref(ref);
409 btrfs_delayed_refs_rsv_release(fs_info, 1, 0);
410}
411
412static bool merge_ref(struct btrfs_fs_info *fs_info,
413 struct btrfs_delayed_ref_root *delayed_refs,
414 struct btrfs_delayed_ref_head *head,
415 struct btrfs_delayed_ref_node *ref,
416 u64 seq)
417{
418 struct btrfs_delayed_ref_node *next;
419 struct rb_node *node = rb_next(&ref->ref_node);
420 bool done = false;
421
422 while (!done && node) {
423 int mod;
424
425 next = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
426 node = rb_next(node);
427 if (seq && next->seq >= seq)
428 break;
429 if (comp_refs(ref, next, false))
430 break;
431
432 if (ref->action == next->action) {
433 mod = next->ref_mod;
434 } else {
435 if (ref->ref_mod < next->ref_mod) {
436 swap(ref, next);
437 done = true;
438 }
439 mod = -next->ref_mod;
440 }
441
442 drop_delayed_ref(fs_info, delayed_refs, head, next);
443 ref->ref_mod += mod;
444 if (ref->ref_mod == 0) {
445 drop_delayed_ref(fs_info, delayed_refs, head, ref);
446 done = true;
447 } else {
448 /*
449 * Can't have multiples of the same ref on a tree block.
450 */
451 WARN_ON(ref->type == BTRFS_TREE_BLOCK_REF_KEY ||
452 ref->type == BTRFS_SHARED_BLOCK_REF_KEY);
453 }
454 }
455
456 return done;
457}
458
459void btrfs_merge_delayed_refs(struct btrfs_fs_info *fs_info,
460 struct btrfs_delayed_ref_root *delayed_refs,
461 struct btrfs_delayed_ref_head *head)
462{
463 struct btrfs_delayed_ref_node *ref;
464 struct rb_node *node;
465 u64 seq = 0;
466
467 lockdep_assert_held(&head->lock);
468
469 if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
470 return;
471
472 /* We don't have too many refs to merge for data. */
473 if (head->is_data)
474 return;
475
476 seq = btrfs_tree_mod_log_lowest_seq(fs_info);
477again:
478 for (node = rb_first_cached(&head->ref_tree); node;
479 node = rb_next(node)) {
480 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
481 if (seq && ref->seq >= seq)
482 continue;
483 if (merge_ref(fs_info, delayed_refs, head, ref, seq))
484 goto again;
485 }
486}
487
488int btrfs_check_delayed_seq(struct btrfs_fs_info *fs_info, u64 seq)
489{
490 int ret = 0;
491 u64 min_seq = btrfs_tree_mod_log_lowest_seq(fs_info);
492
493 if (min_seq != 0 && seq >= min_seq) {
494 btrfs_debug(fs_info,
495 "holding back delayed_ref %llu, lowest is %llu",
496 seq, min_seq);
497 ret = 1;
498 }
499
500 return ret;
501}
502
503struct btrfs_delayed_ref_head *btrfs_select_ref_head(
504 const struct btrfs_fs_info *fs_info,
505 struct btrfs_delayed_ref_root *delayed_refs)
506{
507 struct btrfs_delayed_ref_head *head;
508 unsigned long start_index;
509 unsigned long found_index;
510 bool found_head = false;
511 bool locked;
512
513 spin_lock(&delayed_refs->lock);
514again:
515 start_index = (delayed_refs->run_delayed_start >> fs_info->sectorsize_bits);
516 xa_for_each_start(&delayed_refs->head_refs, found_index, head, start_index) {
517 if (!head->processing) {
518 found_head = true;
519 break;
520 }
521 }
522 if (!found_head) {
523 if (delayed_refs->run_delayed_start == 0) {
524 spin_unlock(&delayed_refs->lock);
525 return NULL;
526 }
527 delayed_refs->run_delayed_start = 0;
528 goto again;
529 }
530
531 head->processing = true;
532 WARN_ON(delayed_refs->num_heads_ready == 0);
533 delayed_refs->num_heads_ready--;
534 delayed_refs->run_delayed_start = head->bytenr +
535 head->num_bytes;
536
537 locked = btrfs_delayed_ref_lock(delayed_refs, head);
538 spin_unlock(&delayed_refs->lock);
539
540 /*
541 * We may have dropped the spin lock to get the head mutex lock, and
542 * that might have given someone else time to free the head. If that's
543 * true, it has been removed from our list and we can move on.
544 */
545 if (!locked)
546 return ERR_PTR(-EAGAIN);
547
548 return head;
549}
550
551void btrfs_unselect_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
552 struct btrfs_delayed_ref_head *head)
553{
554 spin_lock(&delayed_refs->lock);
555 head->processing = false;
556 delayed_refs->num_heads_ready++;
557 spin_unlock(&delayed_refs->lock);
558 btrfs_delayed_ref_unlock(head);
559}
560
561void btrfs_delete_ref_head(const struct btrfs_fs_info *fs_info,
562 struct btrfs_delayed_ref_root *delayed_refs,
563 struct btrfs_delayed_ref_head *head)
564{
565 const unsigned long index = (head->bytenr >> fs_info->sectorsize_bits);
566
567 lockdep_assert_held(&delayed_refs->lock);
568 lockdep_assert_held(&head->lock);
569
570 xa_erase(&delayed_refs->head_refs, index);
571 head->tracked = false;
572 delayed_refs->num_heads--;
573 if (!head->processing)
574 delayed_refs->num_heads_ready--;
575}
576
577struct btrfs_delayed_ref_node *btrfs_select_delayed_ref(struct btrfs_delayed_ref_head *head)
578{
579 struct btrfs_delayed_ref_node *ref;
580
581 lockdep_assert_held(&head->mutex);
582 lockdep_assert_held(&head->lock);
583
584 if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
585 return NULL;
586
587 /*
588 * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
589 * This is to prevent a ref count from going down to zero, which deletes
590 * the extent item from the extent tree, when there still are references
591 * to add, which would fail because they would not find the extent item.
592 */
593 if (!list_empty(&head->ref_add_list))
594 return list_first_entry(&head->ref_add_list,
595 struct btrfs_delayed_ref_node, add_list);
596
597 ref = rb_entry(rb_first_cached(&head->ref_tree),
598 struct btrfs_delayed_ref_node, ref_node);
599 ASSERT(list_empty(&ref->add_list));
600 return ref;
601}
602
603/*
604 * Helper to insert the ref_node to the tail or merge with tail.
605 *
606 * Return false if the ref was inserted.
607 * Return true if the ref was merged into an existing one (and therefore can be
608 * freed by the caller).
609 */
610static bool insert_delayed_ref(struct btrfs_trans_handle *trans,
611 struct btrfs_delayed_ref_head *href,
612 struct btrfs_delayed_ref_node *ref)
613{
614 struct btrfs_delayed_ref_root *root = &trans->transaction->delayed_refs;
615 struct btrfs_delayed_ref_node *exist;
616 int mod;
617
618 ASSERT(ref->action == BTRFS_ADD_DELAYED_REF ||
619 ref->action == BTRFS_DROP_DELAYED_REF);
620
621 spin_lock(&href->lock);
622 exist = tree_insert(&href->ref_tree, ref);
623 if (!exist) {
624 if (ref->action == BTRFS_ADD_DELAYED_REF)
625 list_add_tail(&ref->add_list, &href->ref_add_list);
626 spin_unlock(&href->lock);
627 trans->delayed_ref_updates++;
628 return false;
629 }
630
631 /* Now we are sure we can merge */
632 if (exist->action == ref->action) {
633 mod = ref->ref_mod;
634 } else {
635 /* Need to change action */
636 if (exist->ref_mod < ref->ref_mod) {
637 exist->action = ref->action;
638 mod = -exist->ref_mod;
639 exist->ref_mod = ref->ref_mod;
640 if (ref->action == BTRFS_ADD_DELAYED_REF)
641 list_add_tail(&exist->add_list,
642 &href->ref_add_list);
643 else if (ref->action == BTRFS_DROP_DELAYED_REF) {
644 ASSERT(!list_empty(&exist->add_list));
645 list_del_init(&exist->add_list);
646 } else {
647 DEBUG_WARN();
648 }
649 } else
650 mod = -ref->ref_mod;
651 }
652 exist->ref_mod += mod;
653
654 /* remove existing tail if its ref_mod is zero */
655 if (exist->ref_mod == 0)
656 drop_delayed_ref(trans->fs_info, root, href, exist);
657 spin_unlock(&href->lock);
658 return true;
659}
660
661/*
662 * helper function to update the accounting in the head ref
663 * existing and update must have the same bytenr
664 */
665static noinline void update_existing_head_ref(struct btrfs_trans_handle *trans,
666 struct btrfs_delayed_ref_head *existing,
667 struct btrfs_delayed_ref_head *update)
668{
669 struct btrfs_delayed_ref_root *delayed_refs =
670 &trans->transaction->delayed_refs;
671 struct btrfs_fs_info *fs_info = trans->fs_info;
672 int old_ref_mod;
673
674 BUG_ON(existing->is_data != update->is_data);
675
676 spin_lock(&existing->lock);
677
678 /*
679 * When freeing an extent, we may not know the owning root when we
680 * first create the head_ref. However, some deref before the last deref
681 * will know it, so we just need to update the head_ref accordingly.
682 */
683 if (!existing->owning_root)
684 existing->owning_root = update->owning_root;
685
686 if (update->must_insert_reserved) {
687 /* if the extent was freed and then
688 * reallocated before the delayed ref
689 * entries were processed, we can end up
690 * with an existing head ref without
691 * the must_insert_reserved flag set.
692 * Set it again here
693 */
694 existing->must_insert_reserved = update->must_insert_reserved;
695 existing->owning_root = update->owning_root;
696
697 /*
698 * update the num_bytes so we make sure the accounting
699 * is done correctly
700 */
701 existing->num_bytes = update->num_bytes;
702
703 }
704
705 if (update->extent_op) {
706 if (!existing->extent_op) {
707 existing->extent_op = update->extent_op;
708 } else {
709 if (update->extent_op->update_key) {
710 memcpy(&existing->extent_op->key,
711 &update->extent_op->key,
712 sizeof(update->extent_op->key));
713 existing->extent_op->update_key = true;
714 }
715 if (update->extent_op->update_flags) {
716 existing->extent_op->flags_to_set |=
717 update->extent_op->flags_to_set;
718 existing->extent_op->update_flags = true;
719 }
720 btrfs_free_delayed_extent_op(update->extent_op);
721 }
722 }
723 /*
724 * update the reference mod on the head to reflect this new operation,
725 * only need the lock for this case cause we could be processing it
726 * currently, for refs we just added we know we're a-ok.
727 */
728 old_ref_mod = existing->total_ref_mod;
729 existing->ref_mod += update->ref_mod;
730 existing->total_ref_mod += update->ref_mod;
731
732 /*
733 * If we are going to from a positive ref mod to a negative or vice
734 * versa we need to make sure to adjust pending_csums accordingly.
735 * We reserve bytes for csum deletion when adding or updating a ref head
736 * see add_delayed_ref_head() for more details.
737 */
738 if (existing->is_data) {
739 u64 csum_leaves =
740 btrfs_csum_bytes_to_leaves(fs_info,
741 existing->num_bytes);
742
743 if (existing->total_ref_mod >= 0 && old_ref_mod < 0) {
744 delayed_refs->pending_csums -= existing->num_bytes;
745 btrfs_delayed_refs_rsv_release(fs_info, 0, csum_leaves);
746 }
747 if (existing->total_ref_mod < 0 && old_ref_mod >= 0) {
748 delayed_refs->pending_csums += existing->num_bytes;
749 trans->delayed_ref_csum_deletions += csum_leaves;
750 }
751 }
752
753 spin_unlock(&existing->lock);
754}
755
756static void init_delayed_ref_head(struct btrfs_delayed_ref_head *head_ref,
757 struct btrfs_ref *generic_ref,
758 struct btrfs_qgroup_extent_record *qrecord,
759 u64 reserved)
760{
761 int count_mod = 1;
762 bool must_insert_reserved = false;
763
764 /* If reserved is provided, it must be a data extent. */
765 BUG_ON(generic_ref->type != BTRFS_REF_DATA && reserved);
766
767 switch (generic_ref->action) {
768 case BTRFS_ADD_DELAYED_REF:
769 /* count_mod is already set to 1. */
770 break;
771 case BTRFS_UPDATE_DELAYED_HEAD:
772 count_mod = 0;
773 break;
774 case BTRFS_DROP_DELAYED_REF:
775 /*
776 * The head node stores the sum of all the mods, so dropping a ref
777 * should drop the sum in the head node by one.
778 */
779 count_mod = -1;
780 break;
781 case BTRFS_ADD_DELAYED_EXTENT:
782 /*
783 * BTRFS_ADD_DELAYED_EXTENT means that we need to update the
784 * reserved accounting when the extent is finally added, or if a
785 * later modification deletes the delayed ref without ever
786 * inserting the extent into the extent allocation tree.
787 * ref->must_insert_reserved is the flag used to record that
788 * accounting mods are required.
789 *
790 * Once we record must_insert_reserved, switch the action to
791 * BTRFS_ADD_DELAYED_REF because other special casing is not
792 * required.
793 */
794 must_insert_reserved = true;
795 break;
796 }
797
798 refcount_set(&head_ref->refs, 1);
799 head_ref->bytenr = generic_ref->bytenr;
800 head_ref->num_bytes = generic_ref->num_bytes;
801 head_ref->ref_mod = count_mod;
802 head_ref->reserved_bytes = reserved;
803 head_ref->must_insert_reserved = must_insert_reserved;
804 head_ref->owning_root = generic_ref->owning_root;
805 head_ref->is_data = (generic_ref->type == BTRFS_REF_DATA);
806 head_ref->is_system = (generic_ref->ref_root == BTRFS_CHUNK_TREE_OBJECTID);
807 head_ref->ref_tree = RB_ROOT_CACHED;
808 INIT_LIST_HEAD(&head_ref->ref_add_list);
809 head_ref->tracked = false;
810 head_ref->processing = false;
811 head_ref->total_ref_mod = count_mod;
812 spin_lock_init(&head_ref->lock);
813 mutex_init(&head_ref->mutex);
814
815 /* If not metadata set an impossible level to help debugging. */
816 if (generic_ref->type == BTRFS_REF_METADATA)
817 head_ref->level = generic_ref->tree_ref.level;
818 else
819 head_ref->level = U8_MAX;
820
821 if (qrecord) {
822 if (generic_ref->ref_root && reserved) {
823 qrecord->data_rsv = reserved;
824 qrecord->data_rsv_refroot = generic_ref->ref_root;
825 }
826 qrecord->num_bytes = generic_ref->num_bytes;
827 qrecord->old_roots = NULL;
828 }
829}
830
831/*
832 * Helper function to actually insert a head node into the xarray. This does all
833 * the dirty work in terms of maintaining the correct overall modification
834 * count.
835 *
836 * The caller is responsible for calling kfree() on @qrecord. More specifically,
837 * if this function reports that it did not insert it as noted in
838 * @qrecord_inserted_ret, then it's safe to call kfree() on it.
839 *
840 * Returns an error pointer in case of an error.
841 */
842static noinline struct btrfs_delayed_ref_head *
843add_delayed_ref_head(struct btrfs_trans_handle *trans,
844 struct btrfs_delayed_ref_head *head_ref,
845 struct btrfs_qgroup_extent_record *qrecord,
846 int action, bool *qrecord_inserted_ret)
847{
848 struct btrfs_fs_info *fs_info = trans->fs_info;
849 struct btrfs_delayed_ref_head *existing;
850 struct btrfs_delayed_ref_root *delayed_refs;
851 const unsigned long index = (head_ref->bytenr >> fs_info->sectorsize_bits);
852
853 /*
854 * If 'qrecord_inserted_ret' is provided, then the first thing we need
855 * to do is to initialize it to false just in case we have an exit
856 * before trying to insert the record.
857 */
858 if (qrecord_inserted_ret)
859 *qrecord_inserted_ret = false;
860
861 delayed_refs = &trans->transaction->delayed_refs;
862 lockdep_assert_held(&delayed_refs->lock);
863
864#if BITS_PER_LONG == 32
865 if (head_ref->bytenr >= MAX_LFS_FILESIZE) {
866 if (qrecord)
867 xa_release(&delayed_refs->dirty_extents, index);
868 btrfs_err_rl(fs_info,
869"delayed ref head %llu is beyond 32bit page cache and xarray index limit",
870 head_ref->bytenr);
871 btrfs_err_32bit_limit(fs_info);
872 return ERR_PTR(-EOVERFLOW);
873 }
874#endif
875
876 /* Record qgroup extent info if provided */
877 if (qrecord) {
878 /*
879 * Setting 'qrecord' but not 'qrecord_inserted_ret' will likely
880 * result in a memory leakage.
881 */
882 ASSERT(qrecord_inserted_ret != NULL);
883
884 int ret;
885
886 ret = btrfs_qgroup_trace_extent_nolock(fs_info, delayed_refs, qrecord,
887 head_ref->bytenr);
888 if (ret) {
889 /* Clean up if insertion fails or item exists. */
890 xa_release(&delayed_refs->dirty_extents, index);
891 if (ret < 0)
892 return ERR_PTR(ret);
893 } else if (qrecord_inserted_ret) {
894 *qrecord_inserted_ret = true;
895 }
896 }
897
898 trace_add_delayed_ref_head(fs_info, head_ref, action);
899
900 existing = xa_load(&delayed_refs->head_refs, index);
901 if (existing) {
902 update_existing_head_ref(trans, existing, head_ref);
903 /*
904 * we've updated the existing ref, free the newly
905 * allocated ref
906 */
907 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
908 head_ref = existing;
909 } else {
910 existing = xa_store(&delayed_refs->head_refs, index, head_ref, GFP_ATOMIC);
911 if (xa_is_err(existing)) {
912 /* Memory was preallocated by the caller. */
913 ASSERT(xa_err(existing) != -ENOMEM);
914 return ERR_PTR(xa_err(existing));
915 } else if (WARN_ON(existing)) {
916 /*
917 * Shouldn't happen we just did a lookup before under
918 * delayed_refs->lock.
919 */
920 return ERR_PTR(-EEXIST);
921 }
922 head_ref->tracked = true;
923 /*
924 * We reserve the amount of bytes needed to delete csums when
925 * adding the ref head and not when adding individual drop refs
926 * since the csum items are deleted only after running the last
927 * delayed drop ref (the data extent's ref count drops to 0).
928 */
929 if (head_ref->is_data && head_ref->ref_mod < 0) {
930 delayed_refs->pending_csums += head_ref->num_bytes;
931 trans->delayed_ref_csum_deletions +=
932 btrfs_csum_bytes_to_leaves(fs_info, head_ref->num_bytes);
933 }
934 delayed_refs->num_heads++;
935 delayed_refs->num_heads_ready++;
936 }
937
938 return head_ref;
939}
940
941/*
942 * Initialize the structure which represents a modification to an extent.
943 *
944 * @fs_info: Internal to the mounted filesystem mount structure.
945 *
946 * @ref: The structure which is going to be initialized.
947 *
948 * @bytenr: The logical address of the extent for which a modification is
949 * going to be recorded.
950 *
951 * @num_bytes: Size of the extent whose modification is being recorded.
952 *
953 * @ref_root: The id of the root where this modification has originated, this
954 * can be either one of the well-known metadata trees or the
955 * subvolume id which references this extent.
956 *
957 * @action: Can be one of BTRFS_ADD_DELAYED_REF/BTRFS_DROP_DELAYED_REF or
958 * BTRFS_ADD_DELAYED_EXTENT
959 *
960 * @ref_type: Holds the type of the extent which is being recorded, can be
961 * one of BTRFS_SHARED_BLOCK_REF_KEY/BTRFS_TREE_BLOCK_REF_KEY
962 * when recording a metadata extent or BTRFS_SHARED_DATA_REF_KEY/
963 * BTRFS_EXTENT_DATA_REF_KEY when recording data extent
964 */
965static void init_delayed_ref_common(struct btrfs_fs_info *fs_info,
966 struct btrfs_delayed_ref_node *ref,
967 struct btrfs_ref *generic_ref)
968{
969 int action = generic_ref->action;
970 u64 seq = 0;
971
972 if (action == BTRFS_ADD_DELAYED_EXTENT)
973 action = BTRFS_ADD_DELAYED_REF;
974
975 if (btrfs_is_fstree(generic_ref->ref_root))
976 seq = atomic64_read(&fs_info->tree_mod_seq);
977
978 refcount_set(&ref->refs, 1);
979 ref->bytenr = generic_ref->bytenr;
980 ref->num_bytes = generic_ref->num_bytes;
981 ref->ref_mod = 1;
982 ref->action = action;
983 ref->seq = seq;
984 ref->type = btrfs_ref_type(generic_ref);
985 ref->ref_root = generic_ref->ref_root;
986 ref->parent = generic_ref->parent;
987 RB_CLEAR_NODE(&ref->ref_node);
988 INIT_LIST_HEAD(&ref->add_list);
989
990 if (generic_ref->type == BTRFS_REF_DATA)
991 ref->data_ref = generic_ref->data_ref;
992 else
993 ref->tree_ref = generic_ref->tree_ref;
994}
995
996void btrfs_init_tree_ref(struct btrfs_ref *generic_ref, int level, u64 mod_root,
997 bool skip_qgroup)
998{
999#ifdef CONFIG_BTRFS_DEBUG
1000 /* If @real_root not set, use @root as fallback */
1001 generic_ref->real_root = mod_root ?: generic_ref->ref_root;
1002#endif
1003 generic_ref->tree_ref.level = level;
1004 generic_ref->type = BTRFS_REF_METADATA;
1005 if (skip_qgroup || !(btrfs_is_fstree(generic_ref->ref_root) &&
1006 (!mod_root || btrfs_is_fstree(mod_root))))
1007 generic_ref->skip_qgroup = true;
1008 else
1009 generic_ref->skip_qgroup = false;
1010
1011}
1012
1013void btrfs_init_data_ref(struct btrfs_ref *generic_ref, u64 ino, u64 offset,
1014 u64 mod_root, bool skip_qgroup)
1015{
1016#ifdef CONFIG_BTRFS_DEBUG
1017 /* If @real_root not set, use @root as fallback */
1018 generic_ref->real_root = mod_root ?: generic_ref->ref_root;
1019#endif
1020 generic_ref->data_ref.objectid = ino;
1021 generic_ref->data_ref.offset = offset;
1022 generic_ref->type = BTRFS_REF_DATA;
1023 if (skip_qgroup || !(btrfs_is_fstree(generic_ref->ref_root) &&
1024 (!mod_root || btrfs_is_fstree(mod_root))))
1025 generic_ref->skip_qgroup = true;
1026 else
1027 generic_ref->skip_qgroup = false;
1028}
1029
1030static int add_delayed_ref(struct btrfs_trans_handle *trans,
1031 struct btrfs_ref *generic_ref,
1032 struct btrfs_delayed_extent_op *extent_op,
1033 u64 reserved)
1034{
1035 struct btrfs_fs_info *fs_info = trans->fs_info;
1036 struct btrfs_delayed_ref_node *node;
1037 struct btrfs_delayed_ref_head *head_ref;
1038 struct btrfs_delayed_ref_head *new_head_ref;
1039 struct btrfs_delayed_ref_root *delayed_refs;
1040 struct btrfs_qgroup_extent_record *record = NULL;
1041 const unsigned long index = (generic_ref->bytenr >> fs_info->sectorsize_bits);
1042 bool qrecord_reserved = false;
1043 bool qrecord_inserted;
1044 int action = generic_ref->action;
1045 bool merged;
1046 int ret;
1047
1048 node = kmem_cache_alloc(btrfs_delayed_ref_node_cachep, GFP_NOFS);
1049 if (!node)
1050 return -ENOMEM;
1051
1052 head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1053 if (!head_ref) {
1054 ret = -ENOMEM;
1055 goto free_node;
1056 }
1057
1058 delayed_refs = &trans->transaction->delayed_refs;
1059
1060 if (btrfs_qgroup_full_accounting(fs_info) && !generic_ref->skip_qgroup) {
1061 record = kzalloc_obj(*record, GFP_NOFS);
1062 if (!record) {
1063 ret = -ENOMEM;
1064 goto free_head_ref;
1065 }
1066 if (xa_reserve(&delayed_refs->dirty_extents, index, GFP_NOFS)) {
1067 ret = -ENOMEM;
1068 goto free_record;
1069 }
1070 qrecord_reserved = true;
1071 }
1072
1073 ret = xa_reserve(&delayed_refs->head_refs, index, GFP_NOFS);
1074 if (ret) {
1075 if (qrecord_reserved)
1076 xa_release(&delayed_refs->dirty_extents, index);
1077 goto free_record;
1078 }
1079
1080 init_delayed_ref_common(fs_info, node, generic_ref);
1081 init_delayed_ref_head(head_ref, generic_ref, record, reserved);
1082 head_ref->extent_op = extent_op;
1083
1084 spin_lock(&delayed_refs->lock);
1085
1086 /*
1087 * insert both the head node and the new ref without dropping
1088 * the spin lock
1089 */
1090 new_head_ref = add_delayed_ref_head(trans, head_ref, record,
1091 action, &qrecord_inserted);
1092 if (IS_ERR(new_head_ref)) {
1093 xa_release(&delayed_refs->head_refs, index);
1094 spin_unlock(&delayed_refs->lock);
1095 ret = PTR_ERR(new_head_ref);
1096
1097 /*
1098 * It's only safe to call kfree() on 'qrecord' if
1099 * add_delayed_ref_head() has _not_ inserted it for
1100 * tracing. Otherwise we need to handle this here.
1101 */
1102 if (!qrecord_reserved || qrecord_inserted)
1103 goto free_head_ref;
1104 goto free_record;
1105 }
1106 head_ref = new_head_ref;
1107
1108 merged = insert_delayed_ref(trans, head_ref, node);
1109 spin_unlock(&delayed_refs->lock);
1110
1111 /*
1112 * Need to update the delayed_refs_rsv with any changes we may have
1113 * made.
1114 */
1115 btrfs_update_delayed_refs_rsv(trans);
1116
1117 if (generic_ref->type == BTRFS_REF_DATA)
1118 trace_add_delayed_data_ref(trans->fs_info, node);
1119 else
1120 trace_add_delayed_tree_ref(trans->fs_info, node);
1121 if (merged)
1122 kmem_cache_free(btrfs_delayed_ref_node_cachep, node);
1123
1124 if (qrecord_inserted)
1125 return btrfs_qgroup_trace_extent_post(trans, record, generic_ref->bytenr);
1126
1127 kfree(record);
1128 return 0;
1129
1130free_record:
1131 kfree(record);
1132free_head_ref:
1133 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
1134free_node:
1135 kmem_cache_free(btrfs_delayed_ref_node_cachep, node);
1136 return ret;
1137}
1138
1139/*
1140 * Add a delayed tree ref. This does all of the accounting required to make sure
1141 * the delayed ref is eventually processed before this transaction commits.
1142 */
1143int btrfs_add_delayed_tree_ref(struct btrfs_trans_handle *trans,
1144 struct btrfs_ref *generic_ref,
1145 struct btrfs_delayed_extent_op *extent_op)
1146{
1147 ASSERT(generic_ref->type == BTRFS_REF_METADATA && generic_ref->action);
1148 return add_delayed_ref(trans, generic_ref, extent_op, 0);
1149}
1150
1151/*
1152 * add a delayed data ref. it's similar to btrfs_add_delayed_tree_ref.
1153 */
1154int btrfs_add_delayed_data_ref(struct btrfs_trans_handle *trans,
1155 struct btrfs_ref *generic_ref,
1156 u64 reserved)
1157{
1158 ASSERT(generic_ref->type == BTRFS_REF_DATA && generic_ref->action);
1159 return add_delayed_ref(trans, generic_ref, NULL, reserved);
1160}
1161
1162int btrfs_add_delayed_extent_op(struct btrfs_trans_handle *trans,
1163 u64 bytenr, u64 num_bytes, u8 level,
1164 struct btrfs_delayed_extent_op *extent_op)
1165{
1166 const unsigned long index = (bytenr >> trans->fs_info->sectorsize_bits);
1167 struct btrfs_delayed_ref_head *head_ref;
1168 struct btrfs_delayed_ref_head *head_ref_ret;
1169 struct btrfs_delayed_ref_root *delayed_refs;
1170 struct btrfs_ref generic_ref = {
1171 .type = BTRFS_REF_METADATA,
1172 .action = BTRFS_UPDATE_DELAYED_HEAD,
1173 .bytenr = bytenr,
1174 .num_bytes = num_bytes,
1175 .tree_ref.level = level,
1176 };
1177 int ret;
1178
1179 head_ref = kmem_cache_alloc(btrfs_delayed_ref_head_cachep, GFP_NOFS);
1180 if (!head_ref)
1181 return -ENOMEM;
1182
1183 init_delayed_ref_head(head_ref, &generic_ref, NULL, 0);
1184 head_ref->extent_op = extent_op;
1185
1186 delayed_refs = &trans->transaction->delayed_refs;
1187
1188 ret = xa_reserve(&delayed_refs->head_refs, index, GFP_NOFS);
1189 if (ret) {
1190 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
1191 return ret;
1192 }
1193
1194 spin_lock(&delayed_refs->lock);
1195 head_ref_ret = add_delayed_ref_head(trans, head_ref, NULL,
1196 BTRFS_UPDATE_DELAYED_HEAD, NULL);
1197 if (IS_ERR(head_ref_ret)) {
1198 xa_release(&delayed_refs->head_refs, index);
1199 spin_unlock(&delayed_refs->lock);
1200 kmem_cache_free(btrfs_delayed_ref_head_cachep, head_ref);
1201 return PTR_ERR(head_ref_ret);
1202 }
1203 spin_unlock(&delayed_refs->lock);
1204
1205 /*
1206 * Need to update the delayed_refs_rsv with any changes we may have
1207 * made.
1208 */
1209 btrfs_update_delayed_refs_rsv(trans);
1210 return 0;
1211}
1212
1213void btrfs_put_delayed_ref(struct btrfs_delayed_ref_node *ref)
1214{
1215 if (refcount_dec_and_test(&ref->refs)) {
1216 WARN_ON(!RB_EMPTY_NODE(&ref->ref_node));
1217 kmem_cache_free(btrfs_delayed_ref_node_cachep, ref);
1218 }
1219}
1220
1221/*
1222 * This does a simple search for the head node for a given extent. Returns the
1223 * head node if found, or NULL if not.
1224 */
1225struct btrfs_delayed_ref_head *
1226btrfs_find_delayed_ref_head(const struct btrfs_fs_info *fs_info,
1227 struct btrfs_delayed_ref_root *delayed_refs,
1228 u64 bytenr)
1229{
1230 const unsigned long index = (bytenr >> fs_info->sectorsize_bits);
1231
1232 lockdep_assert_held(&delayed_refs->lock);
1233
1234 return xa_load(&delayed_refs->head_refs, index);
1235}
1236
1237static int find_comp(struct btrfs_delayed_ref_node *entry, u64 root, u64 parent)
1238{
1239 int type = parent ? BTRFS_SHARED_BLOCK_REF_KEY : BTRFS_TREE_BLOCK_REF_KEY;
1240
1241 if (type < entry->type)
1242 return -1;
1243 if (type > entry->type)
1244 return 1;
1245
1246 if (type == BTRFS_TREE_BLOCK_REF_KEY) {
1247 if (root < entry->ref_root)
1248 return -1;
1249 if (root > entry->ref_root)
1250 return 1;
1251 } else {
1252 if (parent < entry->parent)
1253 return -1;
1254 if (parent > entry->parent)
1255 return 1;
1256 }
1257 return 0;
1258}
1259
1260/*
1261 * Check to see if a given root/parent reference is attached to the head. This
1262 * only checks for BTRFS_ADD_DELAYED_REF references that match, as that
1263 * indicates the reference exists for the given root or parent. This is for
1264 * tree blocks only.
1265 *
1266 * @head: the head of the bytenr we're searching.
1267 * @root: the root objectid of the reference if it is a normal reference.
1268 * @parent: the parent if this is a shared backref.
1269 */
1270bool btrfs_find_delayed_tree_ref(struct btrfs_delayed_ref_head *head,
1271 u64 root, u64 parent)
1272{
1273 struct rb_node *node;
1274 bool found = false;
1275
1276 lockdep_assert_held(&head->mutex);
1277
1278 spin_lock(&head->lock);
1279 node = head->ref_tree.rb_root.rb_node;
1280 while (node) {
1281 struct btrfs_delayed_ref_node *entry;
1282 int ret;
1283
1284 entry = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
1285 ret = find_comp(entry, root, parent);
1286 if (ret < 0) {
1287 node = node->rb_left;
1288 } else if (ret > 0) {
1289 node = node->rb_right;
1290 } else {
1291 /*
1292 * We only want to count ADD actions, as drops mean the
1293 * ref doesn't exist.
1294 */
1295 if (entry->action == BTRFS_ADD_DELAYED_REF)
1296 found = true;
1297 break;
1298 }
1299 }
1300 spin_unlock(&head->lock);
1301 return found;
1302}
1303
1304void btrfs_destroy_delayed_refs(struct btrfs_transaction *trans)
1305{
1306 struct btrfs_delayed_ref_root *delayed_refs = &trans->delayed_refs;
1307 struct btrfs_fs_info *fs_info = trans->fs_info;
1308
1309 spin_lock(&delayed_refs->lock);
1310 while (true) {
1311 struct btrfs_delayed_ref_head *head;
1312 struct rb_node *n;
1313 bool pin_bytes = false;
1314
1315 head = find_first_ref_head(delayed_refs);
1316 if (!head)
1317 break;
1318
1319 if (!btrfs_delayed_ref_lock(delayed_refs, head))
1320 continue;
1321
1322 spin_lock(&head->lock);
1323 while ((n = rb_first_cached(&head->ref_tree)) != NULL) {
1324 struct btrfs_delayed_ref_node *ref;
1325
1326 ref = rb_entry(n, struct btrfs_delayed_ref_node, ref_node);
1327 drop_delayed_ref(fs_info, delayed_refs, head, ref);
1328 }
1329 if (head->must_insert_reserved)
1330 pin_bytes = true;
1331 btrfs_free_delayed_extent_op(head->extent_op);
1332 btrfs_delete_ref_head(fs_info, delayed_refs, head);
1333 spin_unlock(&head->lock);
1334 spin_unlock(&delayed_refs->lock);
1335 mutex_unlock(&head->mutex);
1336
1337 if (!btrfs_is_testing(fs_info) && pin_bytes) {
1338 struct btrfs_block_group *bg;
1339
1340 bg = btrfs_lookup_block_group(fs_info, head->bytenr);
1341 if (WARN_ON_ONCE(bg == NULL)) {
1342 /*
1343 * Unexpected and there's nothing we can do here
1344 * because we are in a transaction abort path,
1345 * so any errors can only be ignored or reported
1346 * while attempting to cleanup all resources.
1347 */
1348 btrfs_err(fs_info,
1349"block group for delayed ref at %llu was not found while destroying ref head",
1350 head->bytenr);
1351 } else {
1352 spin_lock(&bg->space_info->lock);
1353 spin_lock(&bg->lock);
1354 bg->pinned += head->num_bytes;
1355 btrfs_space_info_update_bytes_pinned(bg->space_info,
1356 head->num_bytes);
1357 bg->reserved -= head->num_bytes;
1358 bg->space_info->bytes_reserved -= head->num_bytes;
1359 spin_unlock(&bg->lock);
1360 spin_unlock(&bg->space_info->lock);
1361
1362 btrfs_put_block_group(bg);
1363 }
1364
1365 btrfs_error_unpin_extent_range(fs_info, head->bytenr,
1366 head->bytenr + head->num_bytes - 1);
1367 }
1368 if (!btrfs_is_testing(fs_info))
1369 btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
1370 btrfs_put_delayed_ref_head(head);
1371 cond_resched();
1372 spin_lock(&delayed_refs->lock);
1373 }
1374
1375 if (!btrfs_is_testing(fs_info))
1376 btrfs_qgroup_destroy_extent_records(trans);
1377
1378 spin_unlock(&delayed_refs->lock);
1379}
1380
1381void __cold btrfs_delayed_ref_exit(void)
1382{
1383 kmem_cache_destroy(btrfs_delayed_ref_head_cachep);
1384 kmem_cache_destroy(btrfs_delayed_ref_node_cachep);
1385 kmem_cache_destroy(btrfs_delayed_extent_op_cachep);
1386}
1387
1388int __init btrfs_delayed_ref_init(void)
1389{
1390 btrfs_delayed_ref_head_cachep = KMEM_CACHE(btrfs_delayed_ref_head, 0);
1391 if (!btrfs_delayed_ref_head_cachep)
1392 return -ENOMEM;
1393
1394 btrfs_delayed_ref_node_cachep = KMEM_CACHE(btrfs_delayed_ref_node, 0);
1395 if (!btrfs_delayed_ref_node_cachep)
1396 goto fail;
1397
1398 btrfs_delayed_extent_op_cachep = KMEM_CACHE(btrfs_delayed_extent_op, 0);
1399 if (!btrfs_delayed_extent_op_cachep)
1400 goto fail;
1401
1402 return 0;
1403fail:
1404 btrfs_delayed_ref_exit();
1405 return -ENOMEM;
1406}