projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: use a slab for ordered extents allocation
[linux-block.git] / fs / btrfs / delayed-inode.c
CommitLineData
16cdcec7
MX		 1/*
2 * Copyright (C) 2011 Fujitsu. All rights reserved.
3 * Written by Miao Xie <miaox@cn.fujitsu.com>
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
18 */
19
20#include <linux/slab.h>
21#include "delayed-inode.h"
22#include "disk-io.h"
23#include "transaction.h"
24
25#define BTRFS_DELAYED_WRITEBACK 400
26#define BTRFS_DELAYED_BACKGROUND 100
27
28static struct kmem_cache *delayed_node_cache;
29
30int __init btrfs_delayed_inode_init(void)
31{
32 delayed_node_cache = kmem_cache_create("delayed_node",
33 sizeof(struct btrfs_delayed_node),
34 0,
35 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
36 NULL);
37 if (!delayed_node_cache)
38 return -ENOMEM;
39 return 0;
40}
41
42void btrfs_delayed_inode_exit(void)
43{
44 if (delayed_node_cache)
45 kmem_cache_destroy(delayed_node_cache);
46}
47
48static inline void btrfs_init_delayed_node(
49 struct btrfs_delayed_node *delayed_node,
50 struct btrfs_root *root, u64 inode_id)
51{
52 delayed_node->root = root;
53 delayed_node->inode_id = inode_id;
54 atomic_set(&delayed_node->refs, 0);
55 delayed_node->count = 0;
56 delayed_node->in_list = 0;
57 delayed_node->inode_dirty = 0;
58 delayed_node->ins_root = RB_ROOT;
59 delayed_node->del_root = RB_ROOT;
60 mutex_init(&delayed_node->mutex);
61 delayed_node->index_cnt = 0;
62 INIT_LIST_HEAD(&delayed_node->n_list);
63 INIT_LIST_HEAD(&delayed_node->p_list);
64 delayed_node->bytes_reserved = 0;
293f7e07 65 memset(&delayed_node->inode_item, 0, sizeof(delayed_node->inode_item));
16cdcec7
MX		 66}
67
68static inline int btrfs_is_continuous_delayed_item(
69 struct btrfs_delayed_item *item1,
70 struct btrfs_delayed_item *item2)
71{
72 if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
73 item1->key.objectid == item2->key.objectid &&
74 item1->key.type == item2->key.type &&
75 item1->key.offset + 1 == item2->key.offset)
76 return 1;
77 return 0;
78}
79
80static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
81 struct btrfs_root *root)
82{
83 return root->fs_info->delayed_root;
84}
85
2f7e33d4 86static struct btrfs_delayed_node *btrfs_get_delayed_node(struct inode *inode)
16cdcec7 87{
16cdcec7
MX		 88 struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
89 struct btrfs_root *root = btrfs_inode->root;
0d0ca30f 90 u64 ino = btrfs_ino(inode);
2f7e33d4 91 struct btrfs_delayed_node *node;
16cdcec7 92
16cdcec7
MX		 93 node = ACCESS_ONCE(btrfs_inode->delayed_node);
94 if (node) {
2f7e33d4 95 atomic_inc(&node->refs);
16cdcec7
MX		 96 return node;
97 }
98
99 spin_lock(&root->inode_lock);
0d0ca30f 100 node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
16cdcec7
MX		 101 if (node) {
102 if (btrfs_inode->delayed_node) {
2f7e33d4
MX		 103 atomic_inc(&node->refs); /* can be accessed */
104 BUG_ON(btrfs_inode->delayed_node != node);
16cdcec7 105 spin_unlock(&root->inode_lock);
2f7e33d4 106 return node;
16cdcec7
MX		 107 }
108 btrfs_inode->delayed_node = node;
109 atomic_inc(&node->refs); /* can be accessed */
110 atomic_inc(&node->refs); /* cached in the inode */
111 spin_unlock(&root->inode_lock);
112 return node;
113 }
114 spin_unlock(&root->inode_lock);
115
2f7e33d4
MX		 116 return NULL;
117}
118
79787eaa 119/* Will return either the node or PTR_ERR(-ENOMEM) */
2f7e33d4
MX		 120static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
121 struct inode *inode)
122{
123 struct btrfs_delayed_node *node;
124 struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
125 struct btrfs_root *root = btrfs_inode->root;
126 u64 ino = btrfs_ino(inode);
127 int ret;
128
129again:
130 node = btrfs_get_delayed_node(inode);
131 if (node)
132 return node;
133
16cdcec7
MX		 134 node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
135 if (!node)
136 return ERR_PTR(-ENOMEM);
0d0ca30f 137 btrfs_init_delayed_node(node, root, ino);
16cdcec7
MX		 138
139 atomic_inc(&node->refs); /* cached in the btrfs inode */
140 atomic_inc(&node->refs); /* can be accessed */
141
142 ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
143 if (ret) {
144 kmem_cache_free(delayed_node_cache, node);
145 return ERR_PTR(ret);
146 }
147
148 spin_lock(&root->inode_lock);
0d0ca30f 149 ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
16cdcec7
MX		 150 if (ret == -EEXIST) {
151 kmem_cache_free(delayed_node_cache, node);
152 spin_unlock(&root->inode_lock);
153 radix_tree_preload_end();
154 goto again;
155 }
156 btrfs_inode->delayed_node = node;
157 spin_unlock(&root->inode_lock);
158 radix_tree_preload_end();
159
160 return node;
161}
162
163/*
164 * Call it when holding delayed_node->mutex
165 *
166 * If mod = 1, add this node into the prepared list.
167 */
168static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
169 struct btrfs_delayed_node *node,
170 int mod)
171{
172 spin_lock(&root->lock);
173 if (node->in_list) {
174 if (!list_empty(&node->p_list))
175 list_move_tail(&node->p_list, &root->prepare_list);
176 else if (mod)
177 list_add_tail(&node->p_list, &root->prepare_list);
178 } else {
179 list_add_tail(&node->n_list, &root->node_list);
180 list_add_tail(&node->p_list, &root->prepare_list);
181 atomic_inc(&node->refs); /* inserted into list */
182 root->nodes++;
183 node->in_list = 1;
184 }
185 spin_unlock(&root->lock);
186}
187
188/* Call it when holding delayed_node->mutex */
189static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
190 struct btrfs_delayed_node *node)
191{
192 spin_lock(&root->lock);
193 if (node->in_list) {
194 root->nodes--;
195 atomic_dec(&node->refs); /* not in the list */
196 list_del_init(&node->n_list);
197 if (!list_empty(&node->p_list))
198 list_del_init(&node->p_list);
199 node->in_list = 0;
200 }
201 spin_unlock(&root->lock);
202}
203
204struct btrfs_delayed_node *btrfs_first_delayed_node(
205 struct btrfs_delayed_root *delayed_root)
206{
207 struct list_head *p;
208 struct btrfs_delayed_node *node = NULL;
209
210 spin_lock(&delayed_root->lock);
211 if (list_empty(&delayed_root->node_list))
212 goto out;
213
214 p = delayed_root->node_list.next;
215 node = list_entry(p, struct btrfs_delayed_node, n_list);
216 atomic_inc(&node->refs);
217out:
218 spin_unlock(&delayed_root->lock);
219
220 return node;
221}
222
223struct btrfs_delayed_node *btrfs_next_delayed_node(
224 struct btrfs_delayed_node *node)
225{
226 struct btrfs_delayed_root *delayed_root;
227 struct list_head *p;
228 struct btrfs_delayed_node *next = NULL;
229
230 delayed_root = node->root->fs_info->delayed_root;
231 spin_lock(&delayed_root->lock);
232 if (!node->in_list) { /* not in the list */
233 if (list_empty(&delayed_root->node_list))
234 goto out;
235 p = delayed_root->node_list.next;
236 } else if (list_is_last(&node->n_list, &delayed_root->node_list))
237 goto out;
238 else
239 p = node->n_list.next;
240
241 next = list_entry(p, struct btrfs_delayed_node, n_list);
242 atomic_inc(&next->refs);
243out:
244 spin_unlock(&delayed_root->lock);
245
246 return next;
247}
248
249static void __btrfs_release_delayed_node(
250 struct btrfs_delayed_node *delayed_node,
251 int mod)
252{
253 struct btrfs_delayed_root *delayed_root;
254
255 if (!delayed_node)
256 return;
257
258 delayed_root = delayed_node->root->fs_info->delayed_root;
259
260 mutex_lock(&delayed_node->mutex);
261 if (delayed_node->count)
262 btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
263 else
264 btrfs_dequeue_delayed_node(delayed_root, delayed_node);
265 mutex_unlock(&delayed_node->mutex);
266
267 if (atomic_dec_and_test(&delayed_node->refs)) {
268 struct btrfs_root *root = delayed_node->root;
269 spin_lock(&root->inode_lock);
270 if (atomic_read(&delayed_node->refs) == 0) {
271 radix_tree_delete(&root->delayed_nodes_tree,
272 delayed_node->inode_id);
273 kmem_cache_free(delayed_node_cache, delayed_node);
274 }
275 spin_unlock(&root->inode_lock);
276 }
277}
278
279static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
280{
281 __btrfs_release_delayed_node(node, 0);
282}
283
284struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
285 struct btrfs_delayed_root *delayed_root)
286{
287 struct list_head *p;
288 struct btrfs_delayed_node *node = NULL;
289
290 spin_lock(&delayed_root->lock);
291 if (list_empty(&delayed_root->prepare_list))
292 goto out;
293
294 p = delayed_root->prepare_list.next;
295 list_del_init(p);
296 node = list_entry(p, struct btrfs_delayed_node, p_list);
297 atomic_inc(&node->refs);
298out:
299 spin_unlock(&delayed_root->lock);
300
301 return node;
302}
303
304static inline void btrfs_release_prepared_delayed_node(
305 struct btrfs_delayed_node *node)
306{
307 __btrfs_release_delayed_node(node, 1);
308}
309
310struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
311{
312 struct btrfs_delayed_item *item;
313 item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
314 if (item) {
315 item->data_len = data_len;
316 item->ins_or_del = 0;
317 item->bytes_reserved = 0;
16cdcec7
MX		 318 item->delayed_node = NULL;
319 atomic_set(&item->refs, 1);
320 }
321 return item;
322}
323
324/*
325 * __btrfs_lookup_delayed_item - look up the delayed item by key
326 * @delayed_node: pointer to the delayed node
327 * @key: the key to look up
328 * @prev: used to store the prev item if the right item isn't found
329 * @next: used to store the next item if the right item isn't found
330 *
331 * Note: if we don't find the right item, we will return the prev item and
332 * the next item.
333 */
334static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
335 struct rb_root *root,
336 struct btrfs_key *key,
337 struct btrfs_delayed_item **prev,
338 struct btrfs_delayed_item **next)
339{
340 struct rb_node *node, *prev_node = NULL;
341 struct btrfs_delayed_item *delayed_item = NULL;
342 int ret = 0;
343
344 node = root->rb_node;
345
346 while (node) {
347 delayed_item = rb_entry(node, struct btrfs_delayed_item,
348 rb_node);
349 prev_node = node;
350 ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
351 if (ret < 0)
352 node = node->rb_right;
353 else if (ret > 0)
354 node = node->rb_left;
355 else
356 return delayed_item;
357 }
358
359 if (prev) {
360 if (!prev_node)
361 *prev = NULL;
362 else if (ret < 0)
363 *prev = delayed_item;
364 else if ((node = rb_prev(prev_node)) != NULL) {
365 *prev = rb_entry(node, struct btrfs_delayed_item,
366 rb_node);
367 } else
368 *prev = NULL;
369 }
370
371 if (next) {
372 if (!prev_node)
373 *next = NULL;
374 else if (ret > 0)
375 *next = delayed_item;
376 else if ((node = rb_next(prev_node)) != NULL) {
377 *next = rb_entry(node, struct btrfs_delayed_item,
378 rb_node);
379 } else
380 *next = NULL;
381 }
382 return NULL;
383}
384
385struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
386 struct btrfs_delayed_node *delayed_node,
387 struct btrfs_key *key)
388{
389 struct btrfs_delayed_item *item;
390
391 item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
392 NULL, NULL);
393 return item;
394}
395
396struct btrfs_delayed_item *__btrfs_lookup_delayed_deletion_item(
397 struct btrfs_delayed_node *delayed_node,
398 struct btrfs_key *key)
399{
400 struct btrfs_delayed_item *item;
401
402 item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
403 NULL, NULL);
404 return item;
405}
406
407struct btrfs_delayed_item *__btrfs_search_delayed_insertion_item(
408 struct btrfs_delayed_node *delayed_node,
409 struct btrfs_key *key)
410{
411 struct btrfs_delayed_item *item, *next;
412
413 item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
414 NULL, &next);
415 if (!item)
416 item = next;
417
418 return item;
419}
420
421struct btrfs_delayed_item *__btrfs_search_delayed_deletion_item(
422 struct btrfs_delayed_node *delayed_node,
423 struct btrfs_key *key)
424{
425 struct btrfs_delayed_item *item, *next;
426
427 item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
428 NULL, &next);
429 if (!item)
430 item = next;
431
432 return item;
433}
434
435static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
436 struct btrfs_delayed_item *ins,
437 int action)
438{
439 struct rb_node **p, *node;
440 struct rb_node *parent_node = NULL;
441 struct rb_root *root;
442 struct btrfs_delayed_item *item;
443 int cmp;
444
445 if (action == BTRFS_DELAYED_INSERTION_ITEM)
446 root = &delayed_node->ins_root;
447 else if (action == BTRFS_DELAYED_DELETION_ITEM)
448 root = &delayed_node->del_root;
449 else
450 BUG();
451 p = &root->rb_node;
452 node = &ins->rb_node;
453
454 while (*p) {
455 parent_node = *p;
456 item = rb_entry(parent_node, struct btrfs_delayed_item,
457 rb_node);
458
459 cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
460 if (cmp < 0)
461 p = &(*p)->rb_right;
462 else if (cmp > 0)
463 p = &(*p)->rb_left;
464 else
465 return -EEXIST;
466 }
467
468 rb_link_node(node, parent_node, p);
469 rb_insert_color(node, root);
470 ins->delayed_node = delayed_node;
471 ins->ins_or_del = action;
472
473 if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
474 action == BTRFS_DELAYED_INSERTION_ITEM &&
475 ins->key.offset >= delayed_node->index_cnt)
476 delayed_node->index_cnt = ins->key.offset + 1;
477
478 delayed_node->count++;
479 atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
480 return 0;
481}
482
483static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
484 struct btrfs_delayed_item *item)
485{
486 return __btrfs_add_delayed_item(node, item,
487 BTRFS_DELAYED_INSERTION_ITEM);
488}
489
490static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
491 struct btrfs_delayed_item *item)
492{
493 return __btrfs_add_delayed_item(node, item,
494 BTRFS_DELAYED_DELETION_ITEM);
495}
496
497static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
498{
499 struct rb_root *root;
500 struct btrfs_delayed_root *delayed_root;
501
502 delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;
503
504 BUG_ON(!delayed_root);
505 BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
506 delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);
507
508 if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
509 root = &delayed_item->delayed_node->ins_root;
510 else
511 root = &delayed_item->delayed_node->del_root;
512
513 rb_erase(&delayed_item->rb_node, root);
514 delayed_item->delayed_node->count--;
66657b31
JB		 515 if (atomic_dec_return(&delayed_root->items) <
516 BTRFS_DELAYED_BACKGROUND &&
16cdcec7
MX		 517 waitqueue_active(&delayed_root->wait))
518 wake_up(&delayed_root->wait);
519}
520
521static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
522{
523 if (item) {
524 __btrfs_remove_delayed_item(item);
525 if (atomic_dec_and_test(&item->refs))
526 kfree(item);
527 }
528}
529
530struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
531 struct btrfs_delayed_node *delayed_node)
532{
533 struct rb_node *p;
534 struct btrfs_delayed_item *item = NULL;
535
536 p = rb_first(&delayed_node->ins_root);
537 if (p)
538 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
539
540 return item;
541}
542
543struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
544 struct btrfs_delayed_node *delayed_node)
545{
546 struct rb_node *p;
547 struct btrfs_delayed_item *item = NULL;
548
549 p = rb_first(&delayed_node->del_root);
550 if (p)
551 item = rb_entry(p, struct btrfs_delayed_item, rb_node);
552
553 return item;
554}
555
556struct btrfs_delayed_item *__btrfs_next_delayed_item(
557 struct btrfs_delayed_item *item)
558{
559 struct rb_node *p;
560 struct btrfs_delayed_item *next = NULL;
561
562 p = rb_next(&item->rb_node);
563 if (p)
564 next = rb_entry(p, struct btrfs_delayed_item, rb_node);
565
566 return next;
567}
568
16cdcec7
MX		 569static inline struct btrfs_root *btrfs_get_fs_root(struct btrfs_root *root,
570 u64 root_id)
571{
572 struct btrfs_key root_key;
573
574 if (root->objectid == root_id)
575 return root;
576
577 root_key.objectid = root_id;
578 root_key.type = BTRFS_ROOT_ITEM_KEY;
579 root_key.offset = (u64)-1;
580 return btrfs_read_fs_root_no_name(root->fs_info, &root_key);
581}
582
583static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
584 struct btrfs_root *root,
585 struct btrfs_delayed_item *item)
586{
587 struct btrfs_block_rsv *src_rsv;
588 struct btrfs_block_rsv *dst_rsv;
589 u64 num_bytes;
590 int ret;
591
592 if (!trans->bytes_reserved)
593 return 0;
594
595 src_rsv = trans->block_rsv;
6d668dda 596 dst_rsv = &root->fs_info->delayed_block_rsv;
16cdcec7
MX		 597
598 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
599 ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
8c2a3ca2
JB		 600 if (!ret) {
601 trace_btrfs_space_reservation(root->fs_info, "delayed_item",
602 item->key.objectid,
603 num_bytes, 1);
16cdcec7 604 item->bytes_reserved = num_bytes;
8c2a3ca2 605 }
16cdcec7
MX		 606
607 return ret;
608}
609
610static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
611 struct btrfs_delayed_item *item)
612{
19fd2949
MX		 613 struct btrfs_block_rsv *rsv;
614
16cdcec7
MX		 615 if (!item->bytes_reserved)
616 return;
617
6d668dda 618 rsv = &root->fs_info->delayed_block_rsv;
8c2a3ca2
JB		 619 trace_btrfs_space_reservation(root->fs_info, "delayed_item",
620 item->key.objectid, item->bytes_reserved,
621 0);
19fd2949 622 btrfs_block_rsv_release(root, rsv,
16cdcec7
MX		 623 item->bytes_reserved);
624}
625
626static int btrfs_delayed_inode_reserve_metadata(
627 struct btrfs_trans_handle *trans,
628 struct btrfs_root *root,
7fd2ae21 629 struct inode *inode,
16cdcec7
MX		 630 struct btrfs_delayed_node *node)
631{
632 struct btrfs_block_rsv *src_rsv;
633 struct btrfs_block_rsv *dst_rsv;
634 u64 num_bytes;
635 int ret;
8c2a3ca2 636 bool release = false;
16cdcec7 637
16cdcec7 638 src_rsv = trans->block_rsv;
6d668dda 639 dst_rsv = &root->fs_info->delayed_block_rsv;
16cdcec7
MX		 640
641 num_bytes = btrfs_calc_trans_metadata_size(root, 1);
c06a0e12
JB		 642
643 /*
644 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
645 * which doesn't reserve space for speed. This is a problem since we
646 * still need to reserve space for this update, so try to reserve the
647 * space.
648 *
649 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
650 * we're accounted for.
651 */
e755d9ab
CM		 652 if (!src_rsv || (!trans->bytes_reserved &&
653 src_rsv != &root->fs_info->delalloc_block_rsv)) {
c06a0e12
JB		 654 ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes);
655 /*
656 * Since we're under a transaction reserve_metadata_bytes could
657 * try to commit the transaction which will make it return
658 * EAGAIN to make us stop the transaction we have, so return
659 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
660 */
661 if (ret == -EAGAIN)
662 ret = -ENOSPC;
8c2a3ca2 663 if (!ret) {
c06a0e12 664 node->bytes_reserved = num_bytes;
8c2a3ca2
JB		 665 trace_btrfs_space_reservation(root->fs_info,
666 "delayed_inode",
667 btrfs_ino(inode),
668 num_bytes, 1);
669 }
c06a0e12 670 return ret;
7fd2ae21
JB		 671 } else if (src_rsv == &root->fs_info->delalloc_block_rsv) {
672 spin_lock(&BTRFS_I(inode)->lock);
72ac3c0d
JB		 673 if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
674 &BTRFS_I(inode)->runtime_flags)) {
7fd2ae21
JB		 675 spin_unlock(&BTRFS_I(inode)->lock);
676 release = true;
677 goto migrate;
678 }
679 spin_unlock(&BTRFS_I(inode)->lock);
680
681 /* Ok we didn't have space pre-reserved. This shouldn't happen
682 * too often but it can happen if we do delalloc to an existing
683 * inode which gets dirtied because of the time update, and then
684 * isn't touched again until after the transaction commits and
685 * then we try to write out the data. First try to be nice and
686 * reserve something strictly for us. If not be a pain and try
687 * to steal from the delalloc block rsv.
688 */
689 ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes);
690 if (!ret)
691 goto out;
692
693 ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
694 if (!ret)
695 goto out;
696
697 /*
698 * Ok this is a problem, let's just steal from the global rsv
699 * since this really shouldn't happen that often.
700 */
701 WARN_ON(1);
702 ret = btrfs_block_rsv_migrate(&root->fs_info->global_block_rsv,
703 dst_rsv, num_bytes);
704 goto out;
c06a0e12
JB		 705 }
706
7fd2ae21 707migrate:
16cdcec7 708 ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
7fd2ae21
JB		 709
710out:
711 /*
712 * Migrate only takes a reservation, it doesn't touch the size of the
713 * block_rsv. This is to simplify people who don't normally have things
714 * migrated from their block rsv. If they go to release their
715 * reservation, that will decrease the size as well, so if migrate
716 * reduced size we'd end up with a negative size. But for the
717 * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
718 * but we could in fact do this reserve/migrate dance several times
719 * between the time we did the original reservation and we'd clean it
720 * up. So to take care of this, release the space for the meta
721 * reservation here. I think it may be time for a documentation page on
722 * how block rsvs. work.
723 */
8c2a3ca2
JB		 724 if (!ret) {
725 trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
726 btrfs_ino(inode), num_bytes, 1);
16cdcec7 727 node->bytes_reserved = num_bytes;
8c2a3ca2 728 }
16cdcec7 729
8c2a3ca2
JB		 730 if (release) {
731 trace_btrfs_space_reservation(root->fs_info, "delalloc",
732 btrfs_ino(inode), num_bytes, 0);
7fd2ae21 733 btrfs_block_rsv_release(root, src_rsv, num_bytes);
8c2a3ca2 734 }
16cdcec7
MX		 735
736 return ret;
737}
738
739static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
740 struct btrfs_delayed_node *node)
741{
742 struct btrfs_block_rsv *rsv;
743
744 if (!node->bytes_reserved)
745 return;
746
6d668dda 747 rsv = &root->fs_info->delayed_block_rsv;
8c2a3ca2
JB		 748 trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
749 node->inode_id, node->bytes_reserved, 0);
16cdcec7
MX		 750 btrfs_block_rsv_release(root, rsv,
751 node->bytes_reserved);
752 node->bytes_reserved = 0;
753}
754
755/*
756 * This helper will insert some continuous items into the same leaf according
757 * to the free space of the leaf.
758 */
759static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans,
760 struct btrfs_root *root,
761 struct btrfs_path *path,
762 struct btrfs_delayed_item *item)
763{
764 struct btrfs_delayed_item *curr, *next;
765 int free_space;
766 int total_data_size = 0, total_size = 0;
767 struct extent_buffer *leaf;
768 char *data_ptr;
769 struct btrfs_key *keys;
770 u32 *data_size;
771 struct list_head head;
772 int slot;
773 int nitems;
774 int i;
775 int ret = 0;
776
777 BUG_ON(!path->nodes[0]);
778
779 leaf = path->nodes[0];
780 free_space = btrfs_leaf_free_space(root, leaf);
781 INIT_LIST_HEAD(&head);
782
783 next = item;
17aca1c9 784 nitems = 0;
16cdcec7
MX		 785
786 /*
787 * count the number of the continuous items that we can insert in batch
788 */
789 while (total_size + next->data_len + sizeof(struct btrfs_item) <=
790 free_space) {
791 total_data_size += next->data_len;
792 total_size += next->data_len + sizeof(struct btrfs_item);
793 list_add_tail(&next->tree_list, &head);
794 nitems++;
795
796 curr = next;
797 next = __btrfs_next_delayed_item(curr);
798 if (!next)
799 break;
800
801 if (!btrfs_is_continuous_delayed_item(curr, next))
802 break;
803 }
804
805 if (!nitems) {
806 ret = 0;
807 goto out;
808 }
809
810 /*
811 * we need allocate some memory space, but it might cause the task
812 * to sleep, so we set all locked nodes in the path to blocking locks
813 * first.
814 */
815 btrfs_set_path_blocking(path);
816
817 keys = kmalloc(sizeof(struct btrfs_key) * nitems, GFP_NOFS);
818 if (!keys) {
819 ret = -ENOMEM;
820 goto out;
821 }
822
823 data_size = kmalloc(sizeof(u32) * nitems, GFP_NOFS);
824 if (!data_size) {
825 ret = -ENOMEM;
826 goto error;
827 }
828
829 /* get keys of all the delayed items */
830 i = 0;
831 list_for_each_entry(next, &head, tree_list) {
832 keys[i] = next->key;
833 data_size[i] = next->data_len;
834 i++;
835 }
836
837 /* reset all the locked nodes in the patch to spinning locks. */
bd681513 838 btrfs_clear_path_blocking(path, NULL, 0);
16cdcec7
MX		 839
840 /* insert the keys of the items */
143bede5
JM		 841 setup_items_for_insert(trans, root, path, keys, data_size,
842 total_data_size, total_size, nitems);
16cdcec7
MX		 843
844 /* insert the dir index items */
845 slot = path->slots[0];
846 list_for_each_entry_safe(curr, next, &head, tree_list) {
847 data_ptr = btrfs_item_ptr(leaf, slot, char);
848 write_extent_buffer(leaf, &curr->data,
849 (unsigned long)data_ptr,
850 curr->data_len);
851 slot++;
852
853 btrfs_delayed_item_release_metadata(root, curr);
854
855 list_del(&curr->tree_list);
856 btrfs_release_delayed_item(curr);
857 }
858
859error:
860 kfree(data_size);
861 kfree(keys);
862out:
863 return ret;
864}
865
866/*
867 * This helper can just do simple insertion that needn't extend item for new
868 * data, such as directory name index insertion, inode insertion.
869 */
870static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
871 struct btrfs_root *root,
872 struct btrfs_path *path,
873 struct btrfs_delayed_item *delayed_item)
874{
875 struct extent_buffer *leaf;
876 struct btrfs_item *item;
877 char *ptr;
878 int ret;
879
880 ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
881 delayed_item->data_len);
882 if (ret < 0 && ret != -EEXIST)
883 return ret;
884
885 leaf = path->nodes[0];
886
887 item = btrfs_item_nr(leaf, path->slots[0]);
888 ptr = btrfs_item_ptr(leaf, path->slots[0], char);
889
890 write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
891 delayed_item->data_len);
892 btrfs_mark_buffer_dirty(leaf);
893
894 btrfs_delayed_item_release_metadata(root, delayed_item);
895 return 0;
896}
897
898/*
899 * we insert an item first, then if there are some continuous items, we try
900 * to insert those items into the same leaf.
901 */
902static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
903 struct btrfs_path *path,
904 struct btrfs_root *root,
905 struct btrfs_delayed_node *node)
906{
907 struct btrfs_delayed_item *curr, *prev;
908 int ret = 0;
909
910do_again:
911 mutex_lock(&node->mutex);
912 curr = __btrfs_first_delayed_insertion_item(node);
913 if (!curr)
914 goto insert_end;
915
916 ret = btrfs_insert_delayed_item(trans, root, path, curr);
917 if (ret < 0) {
945d8962 918 btrfs_release_path(path);
16cdcec7
MX		 919 goto insert_end;
920 }
921
922 prev = curr;
923 curr = __btrfs_next_delayed_item(prev);
924 if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
925 /* insert the continuous items into the same leaf */
926 path->slots[0]++;
927 btrfs_batch_insert_items(trans, root, path, curr);
928 }
929 btrfs_release_delayed_item(prev);
930 btrfs_mark_buffer_dirty(path->nodes[0]);
931
945d8962 932 btrfs_release_path(path);
16cdcec7
MX		 933 mutex_unlock(&node->mutex);
934 goto do_again;
935
936insert_end:
937 mutex_unlock(&node->mutex);
938 return ret;
939}
940
941static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
942 struct btrfs_root *root,
943 struct btrfs_path *path,
944 struct btrfs_delayed_item *item)
945{
946 struct btrfs_delayed_item *curr, *next;
947 struct extent_buffer *leaf;
948 struct btrfs_key key;
949 struct list_head head;
950 int nitems, i, last_item;
951 int ret = 0;
952
953 BUG_ON(!path->nodes[0]);
954
955 leaf = path->nodes[0];
956
957 i = path->slots[0];
958 last_item = btrfs_header_nritems(leaf) - 1;
959 if (i > last_item)
960 return -ENOENT; /* FIXME: Is errno suitable? */
961
962 next = item;
963 INIT_LIST_HEAD(&head);
964 btrfs_item_key_to_cpu(leaf, &key, i);
965 nitems = 0;
966 /*
967 * count the number of the dir index items that we can delete in batch
968 */
969 while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
970 list_add_tail(&next->tree_list, &head);
971 nitems++;
972
973 curr = next;
974 next = __btrfs_next_delayed_item(curr);
975 if (!next)
976 break;
977
978 if (!btrfs_is_continuous_delayed_item(curr, next))
979 break;
980
981 i++;
982 if (i > last_item)
983 break;
984 btrfs_item_key_to_cpu(leaf, &key, i);
985 }
986
987 if (!nitems)
988 return 0;
989
990 ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
991 if (ret)
992 goto out;
993
994 list_for_each_entry_safe(curr, next, &head, tree_list) {
995 btrfs_delayed_item_release_metadata(root, curr);
996 list_del(&curr->tree_list);
997 btrfs_release_delayed_item(curr);
998 }
999
1000out:
1001 return ret;
1002}
1003
1004static int btrfs_delete_delayed_items(struct btrfs_trans_handle *trans,
1005 struct btrfs_path *path,
1006 struct btrfs_root *root,
1007 struct btrfs_delayed_node *node)
1008{
1009 struct btrfs_delayed_item *curr, *prev;
1010 int ret = 0;
1011
1012do_again:
1013 mutex_lock(&node->mutex);
1014 curr = __btrfs_first_delayed_deletion_item(node);
1015 if (!curr)
1016 goto delete_fail;
1017
1018 ret = btrfs_search_slot(trans, root, &curr->key, path, -1, 1);
1019 if (ret < 0)
1020 goto delete_fail;
1021 else if (ret > 0) {
1022 /*
1023 * can't find the item which the node points to, so this node
1024 * is invalid, just drop it.
1025 */
1026 prev = curr;
1027 curr = __btrfs_next_delayed_item(prev);
1028 btrfs_release_delayed_item(prev);
1029 ret = 0;
945d8962 1030 btrfs_release_path(path);
62095265
FW		 1031 if (curr) {
1032 mutex_unlock(&node->mutex);
16cdcec7 1033 goto do_again;
62095265 1034 } else
16cdcec7
MX		 1035 goto delete_fail;
1036 }
1037
1038 btrfs_batch_delete_items(trans, root, path, curr);
945d8962 1039 btrfs_release_path(path);
16cdcec7
MX		 1040 mutex_unlock(&node->mutex);
1041 goto do_again;
1042
1043delete_fail:
945d8962 1044 btrfs_release_path(path);
16cdcec7
MX		 1045 mutex_unlock(&node->mutex);
1046 return ret;
1047}
1048
1049static void btrfs_release_delayed_inode(struct btrfs_delayed_node *delayed_node)
1050{
1051 struct btrfs_delayed_root *delayed_root;
1052
1053 if (delayed_node && delayed_node->inode_dirty) {
1054 BUG_ON(!delayed_node->root);
1055 delayed_node->inode_dirty = 0;
1056 delayed_node->count--;
1057
1058 delayed_root = delayed_node->root->fs_info->delayed_root;
66657b31 1059 if (atomic_dec_return(&delayed_root->items) <
16cdcec7
MX		 1060 BTRFS_DELAYED_BACKGROUND &&
1061 waitqueue_active(&delayed_root->wait))
1062 wake_up(&delayed_root->wait);
1063 }
1064}
1065
1066static int btrfs_update_delayed_inode(struct btrfs_trans_handle *trans,
1067 struct btrfs_root *root,
1068 struct btrfs_path *path,
1069 struct btrfs_delayed_node *node)
1070{
1071 struct btrfs_key key;
1072 struct btrfs_inode_item *inode_item;
1073 struct extent_buffer *leaf;
1074 int ret;
1075
1076 mutex_lock(&node->mutex);
1077 if (!node->inode_dirty) {
1078 mutex_unlock(&node->mutex);
1079 return 0;
1080 }
1081
1082 key.objectid = node->inode_id;
1083 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1084 key.offset = 0;
1085 ret = btrfs_lookup_inode(trans, root, path, &key, 1);
1086 if (ret > 0) {
945d8962 1087 btrfs_release_path(path);
16cdcec7
MX		 1088 mutex_unlock(&node->mutex);
1089 return -ENOENT;
1090 } else if (ret < 0) {
1091 mutex_unlock(&node->mutex);
1092 return ret;
1093 }
1094
1095 btrfs_unlock_up_safe(path, 1);
1096 leaf = path->nodes[0];
1097 inode_item = btrfs_item_ptr(leaf, path->slots[0],
1098 struct btrfs_inode_item);
1099 write_extent_buffer(leaf, &node->inode_item, (unsigned long)inode_item,
1100 sizeof(struct btrfs_inode_item));
1101 btrfs_mark_buffer_dirty(leaf);
945d8962 1102 btrfs_release_path(path);
16cdcec7
MX		 1103
1104 btrfs_delayed_inode_release_metadata(root, node);
1105 btrfs_release_delayed_inode(node);
1106 mutex_unlock(&node->mutex);
1107
1108 return 0;
1109}
1110
79787eaa
JM		 1111/*
1112 * Called when committing the transaction.
1113 * Returns 0 on success.
1114 * Returns < 0 on error and returns with an aborted transaction with any
1115 * outstanding delayed items cleaned up.
1116 */
96c3f433
JB		 1117static int __btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1118 struct btrfs_root *root, int nr)
16cdcec7 1119{
79787eaa 1120 struct btrfs_root *curr_root = root;
16cdcec7
MX		 1121 struct btrfs_delayed_root *delayed_root;
1122 struct btrfs_delayed_node *curr_node, *prev_node;
1123 struct btrfs_path *path;
19fd2949 1124 struct btrfs_block_rsv *block_rsv;
16cdcec7 1125 int ret = 0;
96c3f433 1126 bool count = (nr > 0);
16cdcec7 1127
79787eaa
JM		 1128 if (trans->aborted)
1129 return -EIO;
1130
16cdcec7
MX		 1131 path = btrfs_alloc_path();
1132 if (!path)
1133 return -ENOMEM;
1134 path->leave_spinning = 1;
1135
19fd2949 1136 block_rsv = trans->block_rsv;
6d668dda 1137 trans->block_rsv = &root->fs_info->delayed_block_rsv;
19fd2949 1138
16cdcec7
MX		 1139 delayed_root = btrfs_get_delayed_root(root);
1140
1141 curr_node = btrfs_first_delayed_node(delayed_root);
96c3f433 1142 while (curr_node && (!count || (count && nr--))) {
79787eaa
JM		 1143 curr_root = curr_node->root;
1144 ret = btrfs_insert_delayed_items(trans, path, curr_root,
16cdcec7
MX		 1145 curr_node);
1146 if (!ret)
79787eaa
JM		 1147 ret = btrfs_delete_delayed_items(trans, path,
1148 curr_root, curr_node);
16cdcec7 1149 if (!ret)
79787eaa
JM		 1150 ret = btrfs_update_delayed_inode(trans, curr_root,
1151 path, curr_node);
16cdcec7
MX		 1152 if (ret) {
1153 btrfs_release_delayed_node(curr_node);
96c3f433 1154 curr_node = NULL;
79787eaa 1155 btrfs_abort_transaction(trans, root, ret);
16cdcec7
MX		 1156 break;
1157 }
1158
1159 prev_node = curr_node;
1160 curr_node = btrfs_next_delayed_node(curr_node);
1161 btrfs_release_delayed_node(prev_node);
1162 }
1163
96c3f433
JB		 1164 if (curr_node)
1165 btrfs_release_delayed_node(curr_node);
16cdcec7 1166 btrfs_free_path(path);
19fd2949 1167 trans->block_rsv = block_rsv;
79787eaa 1168
16cdcec7
MX		 1169 return ret;
1170}
1171
96c3f433
JB		 1172int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
1173 struct btrfs_root *root)
1174{
1175 return __btrfs_run_delayed_items(trans, root, -1);
1176}
1177
1178int btrfs_run_delayed_items_nr(struct btrfs_trans_handle *trans,
1179 struct btrfs_root *root, int nr)
1180{
1181 return __btrfs_run_delayed_items(trans, root, nr);
1182}
1183
16cdcec7
MX		 1184static int __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1185 struct btrfs_delayed_node *node)
1186{
1187 struct btrfs_path *path;
19fd2949 1188 struct btrfs_block_rsv *block_rsv;
16cdcec7
MX		 1189 int ret;
1190
1191 path = btrfs_alloc_path();
1192 if (!path)
1193 return -ENOMEM;
1194 path->leave_spinning = 1;
1195
19fd2949 1196 block_rsv = trans->block_rsv;
6d668dda 1197 trans->block_rsv = &node->root->fs_info->delayed_block_rsv;
19fd2949 1198
16cdcec7
MX		 1199 ret = btrfs_insert_delayed_items(trans, path, node->root, node);
1200 if (!ret)
1201 ret = btrfs_delete_delayed_items(trans, path, node->root, node);
1202 if (!ret)
1203 ret = btrfs_update_delayed_inode(trans, node->root, path, node);
1204 btrfs_free_path(path);
1205
19fd2949 1206 trans->block_rsv = block_rsv;
16cdcec7
MX		 1207 return ret;
1208}
1209
1210int btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
1211 struct inode *inode)
1212{
1213 struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
1214 int ret;
1215
1216 if (!delayed_node)
1217 return 0;
1218
1219 mutex_lock(&delayed_node->mutex);
1220 if (!delayed_node->count) {
1221 mutex_unlock(&delayed_node->mutex);
1222 btrfs_release_delayed_node(delayed_node);
1223 return 0;
1224 }
1225 mutex_unlock(&delayed_node->mutex);
1226
1227 ret = __btrfs_commit_inode_delayed_items(trans, delayed_node);
1228 btrfs_release_delayed_node(delayed_node);
1229 return ret;
1230}
1231
1232void btrfs_remove_delayed_node(struct inode *inode)
1233{
1234 struct btrfs_delayed_node *delayed_node;
1235
1236 delayed_node = ACCESS_ONCE(BTRFS_I(inode)->delayed_node);
1237 if (!delayed_node)
1238 return;
1239
1240 BTRFS_I(inode)->delayed_node = NULL;
1241 btrfs_release_delayed_node(delayed_node);
1242}
1243
1244struct btrfs_async_delayed_node {
1245 struct btrfs_root *root;
1246 struct btrfs_delayed_node *delayed_node;
1247 struct btrfs_work work;
1248};
1249
1250static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
1251{
1252 struct btrfs_async_delayed_node *async_node;
1253 struct btrfs_trans_handle *trans;
1254 struct btrfs_path *path;
1255 struct btrfs_delayed_node *delayed_node = NULL;
1256 struct btrfs_root *root;
19fd2949 1257 struct btrfs_block_rsv *block_rsv;
16cdcec7
MX		 1258 unsigned long nr = 0;
1259 int need_requeue = 0;
1260 int ret;
1261
1262 async_node = container_of(work, struct btrfs_async_delayed_node, work);
1263
1264 path = btrfs_alloc_path();
1265 if (!path)
1266 goto out;
1267 path->leave_spinning = 1;
1268
1269 delayed_node = async_node->delayed_node;
1270 root = delayed_node->root;
1271
ff5714cc 1272 trans = btrfs_join_transaction(root);
16cdcec7
MX		 1273 if (IS_ERR(trans))
1274 goto free_path;
1275
19fd2949 1276 block_rsv = trans->block_rsv;
6d668dda 1277 trans->block_rsv = &root->fs_info->delayed_block_rsv;
19fd2949 1278
16cdcec7
MX		 1279 ret = btrfs_insert_delayed_items(trans, path, root, delayed_node);
1280 if (!ret)
1281 ret = btrfs_delete_delayed_items(trans, path, root,
1282 delayed_node);
1283
1284 if (!ret)
1285 btrfs_update_delayed_inode(trans, root, path, delayed_node);
1286
1287 /*
1288 * Maybe new delayed items have been inserted, so we need requeue
1289 * the work. Besides that, we must dequeue the empty delayed nodes
1290 * to avoid the race between delayed items balance and the worker.
1291 * The race like this:
1292 * Task1 Worker thread
1293 * count == 0, needn't requeue
1294 * also needn't insert the
1295 * delayed node into prepare
1296 * list again.
1297 * add lots of delayed items
1298 * queue the delayed node
1299 * already in the list,
1300 * and not in the prepare
1301 * list, it means the delayed
1302 * node is being dealt with
1303 * by the worker.
1304 * do delayed items balance
1305 * the delayed node is being
1306 * dealt with by the worker
1307 * now, just wait.
1308 * the worker goto idle.
1309 * Task1 will sleep until the transaction is commited.
1310 */
1311 mutex_lock(&delayed_node->mutex);
1312 if (delayed_node->count)
1313 need_requeue = 1;
1314 else
1315 btrfs_dequeue_delayed_node(root->fs_info->delayed_root,
1316 delayed_node);
1317 mutex_unlock(&delayed_node->mutex);
1318
1319 nr = trans->blocks_used;
1320
19fd2949 1321 trans->block_rsv = block_rsv;
16cdcec7
MX		 1322 btrfs_end_transaction_dmeta(trans, root);
1323 __btrfs_btree_balance_dirty(root, nr);
1324free_path:
1325 btrfs_free_path(path);
1326out:
1327 if (need_requeue)
1328 btrfs_requeue_work(&async_node->work);
1329 else {
1330 btrfs_release_prepared_delayed_node(delayed_node);
1331 kfree(async_node);
1332 }
1333}
1334
1335static int btrfs_wq_run_delayed_node(struct btrfs_delayed_root *delayed_root,
1336 struct btrfs_root *root, int all)
1337{
1338 struct btrfs_async_delayed_node *async_node;
1339 struct btrfs_delayed_node *curr;
1340 int count = 0;
1341
1342again:
1343 curr = btrfs_first_prepared_delayed_node(delayed_root);
1344 if (!curr)
1345 return 0;
1346
1347 async_node = kmalloc(sizeof(*async_node), GFP_NOFS);
1348 if (!async_node) {
1349 btrfs_release_prepared_delayed_node(curr);
1350 return -ENOMEM;
1351 }
1352
1353 async_node->root = root;
1354 async_node->delayed_node = curr;
1355
1356 async_node->work.func = btrfs_async_run_delayed_node_done;
1357 async_node->work.flags = 0;
1358
1359 btrfs_queue_worker(&root->fs_info->delayed_workers, &async_node->work);
1360 count++;
1361
1362 if (all || count < 4)
1363 goto again;
1364
1365 return 0;
1366}
1367
e999376f
CM		 1368void btrfs_assert_delayed_root_empty(struct btrfs_root *root)
1369{
1370 struct btrfs_delayed_root *delayed_root;
1371 delayed_root = btrfs_get_delayed_root(root);
1372 WARN_ON(btrfs_first_delayed_node(delayed_root));
1373}
1374
16cdcec7
MX		 1375void btrfs_balance_delayed_items(struct btrfs_root *root)
1376{
1377 struct btrfs_delayed_root *delayed_root;
1378
1379 delayed_root = btrfs_get_delayed_root(root);
1380
1381 if (atomic_read(&delayed_root->items) < BTRFS_DELAYED_BACKGROUND)
1382 return;
1383
1384 if (atomic_read(&delayed_root->items) >= BTRFS_DELAYED_WRITEBACK) {
1385 int ret;
1386 ret = btrfs_wq_run_delayed_node(delayed_root, root, 1);
1387 if (ret)
1388 return;
1389
1390 wait_event_interruptible_timeout(
1391 delayed_root->wait,
1392 (atomic_read(&delayed_root->items) <
1393 BTRFS_DELAYED_BACKGROUND),
1394 HZ);
1395 return;
1396 }
1397
1398 btrfs_wq_run_delayed_node(delayed_root, root, 0);
1399}
1400
79787eaa 1401/* Will return 0 or -ENOMEM */
16cdcec7
MX		 1402int btrfs_insert_delayed_dir_index(struct btrfs_trans_handle *trans,
1403 struct btrfs_root *root, const char *name,
1404 int name_len, struct inode *dir,
1405 struct btrfs_disk_key *disk_key, u8 type,
1406 u64 index)
1407{
1408 struct btrfs_delayed_node *delayed_node;
1409 struct btrfs_delayed_item *delayed_item;
1410 struct btrfs_dir_item *dir_item;
1411 int ret;
1412
1413 delayed_node = btrfs_get_or_create_delayed_node(dir);
1414 if (IS_ERR(delayed_node))
1415 return PTR_ERR(delayed_node);
1416
1417 delayed_item = btrfs_alloc_delayed_item(sizeof(*dir_item) + name_len);
1418 if (!delayed_item) {
1419 ret = -ENOMEM;
1420 goto release_node;
1421 }
1422
0d0ca30f 1423 delayed_item->key.objectid = btrfs_ino(dir);
16cdcec7
MX		 1424 btrfs_set_key_type(&delayed_item->key, BTRFS_DIR_INDEX_KEY);
1425 delayed_item->key.offset = index;
1426
1427 dir_item = (struct btrfs_dir_item *)delayed_item->data;
1428 dir_item->location = *disk_key;
1429 dir_item->transid = cpu_to_le64(trans->transid);
1430 dir_item->data_len = 0;
1431 dir_item->name_len = cpu_to_le16(name_len);
1432 dir_item->type = type;
1433 memcpy((char *)(dir_item + 1), name, name_len);
1434
8c2a3ca2
JB		 1435 ret = btrfs_delayed_item_reserve_metadata(trans, root, delayed_item);
1436 /*
1437 * we have reserved enough space when we start a new transaction,
1438 * so reserving metadata failure is impossible
1439 */
1440 BUG_ON(ret);
1441
1442
16cdcec7
MX		 1443 mutex_lock(&delayed_node->mutex);
1444 ret = __btrfs_add_delayed_insertion_item(delayed_node, delayed_item);
1445 if (unlikely(ret)) {
1446 printk(KERN_ERR "err add delayed dir index item(name: %s) into "
1447 "the insertion tree of the delayed node"
1448 "(root id: %llu, inode id: %llu, errno: %d)\n",
1449 name,
1450 (unsigned long long)delayed_node->root->objectid,
1451 (unsigned long long)delayed_node->inode_id,
1452 ret);
1453 BUG();
1454 }
1455 mutex_unlock(&delayed_node->mutex);
1456
1457release_node:
1458 btrfs_release_delayed_node(delayed_node);
1459 return ret;
1460}
1461
1462static int btrfs_delete_delayed_insertion_item(struct btrfs_root *root,
1463 struct btrfs_delayed_node *node,
1464 struct btrfs_key *key)
1465{
1466 struct btrfs_delayed_item *item;
1467
1468 mutex_lock(&node->mutex);
1469 item = __btrfs_lookup_delayed_insertion_item(node, key);
1470 if (!item) {
1471 mutex_unlock(&node->mutex);
1472 return 1;
1473 }
1474
1475 btrfs_delayed_item_release_metadata(root, item);
1476 btrfs_release_delayed_item(item);
1477 mutex_unlock(&node->mutex);
1478 return 0;
1479}
1480
1481int btrfs_delete_delayed_dir_index(struct btrfs_trans_handle *trans,
1482 struct btrfs_root *root, struct inode *dir,
1483 u64 index)
1484{
1485 struct btrfs_delayed_node *node;
1486 struct btrfs_delayed_item *item;
1487 struct btrfs_key item_key;
1488 int ret;
1489
1490 node = btrfs_get_or_create_delayed_node(dir);
1491 if (IS_ERR(node))
1492 return PTR_ERR(node);
1493
0d0ca30f 1494 item_key.objectid = btrfs_ino(dir);
16cdcec7
MX		 1495 btrfs_set_key_type(&item_key, BTRFS_DIR_INDEX_KEY);
1496 item_key.offset = index;
1497
1498 ret = btrfs_delete_delayed_insertion_item(root, node, &item_key);
1499 if (!ret)
1500 goto end;
1501
1502 item = btrfs_alloc_delayed_item(0);
1503 if (!item) {
1504 ret = -ENOMEM;
1505 goto end;
1506 }
1507
1508 item->key = item_key;
1509
1510 ret = btrfs_delayed_item_reserve_metadata(trans, root, item);
1511 /*
1512 * we have reserved enough space when we start a new transaction,
1513 * so reserving metadata failure is impossible.
1514 */
1515 BUG_ON(ret);
1516
1517 mutex_lock(&node->mutex);
1518 ret = __btrfs_add_delayed_deletion_item(node, item);
1519 if (unlikely(ret)) {
1520 printk(KERN_ERR "err add delayed dir index item(index: %llu) "
1521 "into the deletion tree of the delayed node"
1522 "(root id: %llu, inode id: %llu, errno: %d)\n",
1523 (unsigned long long)index,
1524 (unsigned long long)node->root->objectid,
1525 (unsigned long long)node->inode_id,
1526 ret);
1527 BUG();
1528 }
1529 mutex_unlock(&node->mutex);
1530end:
1531 btrfs_release_delayed_node(node);
1532 return ret;
1533}
1534
1535int btrfs_inode_delayed_dir_index_count(struct inode *inode)
1536{
2f7e33d4 1537 struct btrfs_delayed_node *delayed_node = btrfs_get_delayed_node(inode);
16cdcec7
MX		 1538
1539 if (!delayed_node)
1540 return -ENOENT;
1541
1542 /*
1543 * Since we have held i_mutex of this directory, it is impossible that
1544 * a new directory index is added into the delayed node and index_cnt
1545 * is updated now. So we needn't lock the delayed node.
1546 */
2f7e33d4
MX		 1547 if (!delayed_node->index_cnt) {
1548 btrfs_release_delayed_node(delayed_node);
16cdcec7 1549 return -EINVAL;
2f7e33d4 1550 }
16cdcec7
MX		 1551
1552 BTRFS_I(inode)->index_cnt = delayed_node->index_cnt;
2f7e33d4
MX		 1553 btrfs_release_delayed_node(delayed_node);
1554 return 0;
16cdcec7
MX		 1555}
1556
1557void btrfs_get_delayed_items(struct inode *inode, struct list_head *ins_list,
1558 struct list_head *del_list)
1559{
1560 struct btrfs_delayed_node *delayed_node;
1561 struct btrfs_delayed_item *item;
1562
1563 delayed_node = btrfs_get_delayed_node(inode);
1564 if (!delayed_node)
1565 return;
1566
1567 mutex_lock(&delayed_node->mutex);
1568 item = __btrfs_first_delayed_insertion_item(delayed_node);
1569 while (item) {
1570 atomic_inc(&item->refs);
1571 list_add_tail(&item->readdir_list, ins_list);
1572 item = __btrfs_next_delayed_item(item);
1573 }
1574
1575 item = __btrfs_first_delayed_deletion_item(delayed_node);
1576 while (item) {
1577 atomic_inc(&item->refs);
1578 list_add_tail(&item->readdir_list, del_list);
1579 item = __btrfs_next_delayed_item(item);
1580 }
1581 mutex_unlock(&delayed_node->mutex);
1582 /*
1583 * This delayed node is still cached in the btrfs inode, so refs
1584 * must be > 1 now, and we needn't check it is going to be freed
1585 * or not.
1586 *
1587 * Besides that, this function is used to read dir, we do not
1588 * insert/delete delayed items in this period. So we also needn't
1589 * requeue or dequeue this delayed node.
1590 */
1591 atomic_dec(&delayed_node->refs);
1592}
1593
1594void btrfs_put_delayed_items(struct list_head *ins_list,
1595 struct list_head *del_list)
1596{
1597 struct btrfs_delayed_item *curr, *next;
1598
1599 list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1600 list_del(&curr->readdir_list);
1601 if (atomic_dec_and_test(&curr->refs))
1602 kfree(curr);
1603 }
1604
1605 list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1606 list_del(&curr->readdir_list);
1607 if (atomic_dec_and_test(&curr->refs))
1608 kfree(curr);
1609 }
1610}
1611
1612int btrfs_should_delete_dir_index(struct list_head *del_list,
1613 u64 index)
1614{
1615 struct btrfs_delayed_item *curr, *next;
1616 int ret;
1617
1618 if (list_empty(del_list))
1619 return 0;
1620
1621 list_for_each_entry_safe(curr, next, del_list, readdir_list) {
1622 if (curr->key.offset > index)
1623 break;
1624
1625 list_del(&curr->readdir_list);
1626 ret = (curr->key.offset == index);
1627
1628 if (atomic_dec_and_test(&curr->refs))
1629 kfree(curr);
1630
1631 if (ret)
1632 return 1;
1633 else
1634 continue;
1635 }
1636 return 0;
1637}
1638
1639/*
1640 * btrfs_readdir_delayed_dir_index - read dir info stored in the delayed tree
1641 *
1642 */
1643int btrfs_readdir_delayed_dir_index(struct file *filp, void *dirent,
1644 filldir_t filldir,
1645 struct list_head *ins_list)
1646{
1647 struct btrfs_dir_item *di;
1648 struct btrfs_delayed_item *curr, *next;
1649 struct btrfs_key location;
1650 char *name;
1651 int name_len;
1652 int over = 0;
1653 unsigned char d_type;
1654
1655 if (list_empty(ins_list))
1656 return 0;
1657
1658 /*
1659 * Changing the data of the delayed item is impossible. So
1660 * we needn't lock them. And we have held i_mutex of the
1661 * directory, nobody can delete any directory indexes now.
1662 */
1663 list_for_each_entry_safe(curr, next, ins_list, readdir_list) {
1664 list_del(&curr->readdir_list);
1665
1666 if (curr->key.offset < filp->f_pos) {
1667 if (atomic_dec_and_test(&curr->refs))
1668 kfree(curr);
1669 continue;
1670 }
1671
1672 filp->f_pos = curr->key.offset;
1673
1674 di = (struct btrfs_dir_item *)curr->data;
1675 name = (char *)(di + 1);
1676 name_len = le16_to_cpu(di->name_len);
1677
1678 d_type = btrfs_filetype_table[di->type];
1679 btrfs_disk_key_to_cpu(&location, &di->location);
1680
1681 over = filldir(dirent, name, name_len, curr->key.offset,
1682 location.objectid, d_type);
1683
1684 if (atomic_dec_and_test(&curr->refs))
1685 kfree(curr);
1686
1687 if (over)
1688 return 1;
1689 }
1690 return 0;
1691}
1692
1693BTRFS_SETGET_STACK_FUNCS(stack_inode_generation, struct btrfs_inode_item,
1694 generation, 64);
1695BTRFS_SETGET_STACK_FUNCS(stack_inode_sequence, struct btrfs_inode_item,
1696 sequence, 64);
1697BTRFS_SETGET_STACK_FUNCS(stack_inode_transid, struct btrfs_inode_item,
1698 transid, 64);
1699BTRFS_SETGET_STACK_FUNCS(stack_inode_size, struct btrfs_inode_item, size, 64);
1700BTRFS_SETGET_STACK_FUNCS(stack_inode_nbytes, struct btrfs_inode_item,
1701 nbytes, 64);
1702BTRFS_SETGET_STACK_FUNCS(stack_inode_block_group, struct btrfs_inode_item,
1703 block_group, 64);
1704BTRFS_SETGET_STACK_FUNCS(stack_inode_nlink, struct btrfs_inode_item, nlink, 32);
1705BTRFS_SETGET_STACK_FUNCS(stack_inode_uid, struct btrfs_inode_item, uid, 32);
1706BTRFS_SETGET_STACK_FUNCS(stack_inode_gid, struct btrfs_inode_item, gid, 32);
1707BTRFS_SETGET_STACK_FUNCS(stack_inode_mode, struct btrfs_inode_item, mode, 32);
1708BTRFS_SETGET_STACK_FUNCS(stack_inode_rdev, struct btrfs_inode_item, rdev, 64);
1709BTRFS_SETGET_STACK_FUNCS(stack_inode_flags, struct btrfs_inode_item, flags, 64);
1710
1711BTRFS_SETGET_STACK_FUNCS(stack_timespec_sec, struct btrfs_timespec, sec, 64);
1712BTRFS_SETGET_STACK_FUNCS(stack_timespec_nsec, struct btrfs_timespec, nsec, 32);
1713
1714static void fill_stack_inode_item(struct btrfs_trans_handle *trans,
1715 struct btrfs_inode_item *inode_item,
1716 struct inode *inode)
1717{
1718 btrfs_set_stack_inode_uid(inode_item, inode->i_uid);
1719 btrfs_set_stack_inode_gid(inode_item, inode->i_gid);
1720 btrfs_set_stack_inode_size(inode_item, BTRFS_I(inode)->disk_i_size);
1721 btrfs_set_stack_inode_mode(inode_item, inode->i_mode);
1722 btrfs_set_stack_inode_nlink(inode_item, inode->i_nlink);
1723 btrfs_set_stack_inode_nbytes(inode_item, inode_get_bytes(inode));
1724 btrfs_set_stack_inode_generation(inode_item,
1725 BTRFS_I(inode)->generation);
0c4d2d95 1726 btrfs_set_stack_inode_sequence(inode_item, inode->i_version);
16cdcec7
MX		 1727 btrfs_set_stack_inode_transid(inode_item, trans->transid);
1728 btrfs_set_stack_inode_rdev(inode_item, inode->i_rdev);
1729 btrfs_set_stack_inode_flags(inode_item, BTRFS_I(inode)->flags);
ff5714cc 1730 btrfs_set_stack_inode_block_group(inode_item, 0);
16cdcec7
MX		 1731
1732 btrfs_set_stack_timespec_sec(btrfs_inode_atime(inode_item),
1733 inode->i_atime.tv_sec);
1734 btrfs_set_stack_timespec_nsec(btrfs_inode_atime(inode_item),
1735 inode->i_atime.tv_nsec);
1736
1737 btrfs_set_stack_timespec_sec(btrfs_inode_mtime(inode_item),
1738 inode->i_mtime.tv_sec);
1739 btrfs_set_stack_timespec_nsec(btrfs_inode_mtime(inode_item),
1740 inode->i_mtime.tv_nsec);
1741
1742 btrfs_set_stack_timespec_sec(btrfs_inode_ctime(inode_item),
1743 inode->i_ctime.tv_sec);
1744 btrfs_set_stack_timespec_nsec(btrfs_inode_ctime(inode_item),
1745 inode->i_ctime.tv_nsec);
1746}
1747
2f7e33d4
MX		 1748int btrfs_fill_inode(struct inode *inode, u32 *rdev)
1749{
1750 struct btrfs_delayed_node *delayed_node;
1751 struct btrfs_inode_item *inode_item;
1752 struct btrfs_timespec *tspec;
1753
1754 delayed_node = btrfs_get_delayed_node(inode);
1755 if (!delayed_node)
1756 return -ENOENT;
1757
1758 mutex_lock(&delayed_node->mutex);
1759 if (!delayed_node->inode_dirty) {
1760 mutex_unlock(&delayed_node->mutex);
1761 btrfs_release_delayed_node(delayed_node);
1762 return -ENOENT;
1763 }
1764
1765 inode_item = &delayed_node->inode_item;
1766
1767 inode->i_uid = btrfs_stack_inode_uid(inode_item);
1768 inode->i_gid = btrfs_stack_inode_gid(inode_item);
1769 btrfs_i_size_write(inode, btrfs_stack_inode_size(inode_item));
1770 inode->i_mode = btrfs_stack_inode_mode(inode_item);
bfe86848 1771 set_nlink(inode, btrfs_stack_inode_nlink(inode_item));
2f7e33d4
MX		 1772 inode_set_bytes(inode, btrfs_stack_inode_nbytes(inode_item));
1773 BTRFS_I(inode)->generation = btrfs_stack_inode_generation(inode_item);
0c4d2d95 1774 inode->i_version = btrfs_stack_inode_sequence(inode_item);
2f7e33d4
MX		 1775 inode->i_rdev = 0;
1776 *rdev = btrfs_stack_inode_rdev(inode_item);
1777 BTRFS_I(inode)->flags = btrfs_stack_inode_flags(inode_item);
1778
1779 tspec = btrfs_inode_atime(inode_item);
1780 inode->i_atime.tv_sec = btrfs_stack_timespec_sec(tspec);
1781 inode->i_atime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
1782
1783 tspec = btrfs_inode_mtime(inode_item);
1784 inode->i_mtime.tv_sec = btrfs_stack_timespec_sec(tspec);
1785 inode->i_mtime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
1786
1787 tspec = btrfs_inode_ctime(inode_item);
1788 inode->i_ctime.tv_sec = btrfs_stack_timespec_sec(tspec);
1789 inode->i_ctime.tv_nsec = btrfs_stack_timespec_nsec(tspec);
1790
1791 inode->i_generation = BTRFS_I(inode)->generation;
1792 BTRFS_I(inode)->index_cnt = (u64)-1;
1793
1794 mutex_unlock(&delayed_node->mutex);
1795 btrfs_release_delayed_node(delayed_node);
1796 return 0;
1797}
1798
16cdcec7
MX		 1799int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
1800 struct btrfs_root *root, struct inode *inode)
1801{
1802 struct btrfs_delayed_node *delayed_node;
aa0467d8 1803 int ret = 0;
16cdcec7
MX		 1804
1805 delayed_node = btrfs_get_or_create_delayed_node(inode);
1806 if (IS_ERR(delayed_node))
1807 return PTR_ERR(delayed_node);
1808
1809 mutex_lock(&delayed_node->mutex);
1810 if (delayed_node->inode_dirty) {
1811 fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1812 goto release_node;
1813 }
1814
7fd2ae21
JB		 1815 ret = btrfs_delayed_inode_reserve_metadata(trans, root, inode,
1816 delayed_node);
c06a0e12
JB		 1817 if (ret)
1818 goto release_node;
16cdcec7
MX		 1819
1820 fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
1821 delayed_node->inode_dirty = 1;
1822 delayed_node->count++;
1823 atomic_inc(&root->fs_info->delayed_root->items);
1824release_node:
1825 mutex_unlock(&delayed_node->mutex);
1826 btrfs_release_delayed_node(delayed_node);
1827 return ret;
1828}
1829
1830static void __btrfs_kill_delayed_node(struct btrfs_delayed_node *delayed_node)
1831{
1832 struct btrfs_root *root = delayed_node->root;
1833 struct btrfs_delayed_item *curr_item, *prev_item;
1834
1835 mutex_lock(&delayed_node->mutex);
1836 curr_item = __btrfs_first_delayed_insertion_item(delayed_node);
1837 while (curr_item) {
1838 btrfs_delayed_item_release_metadata(root, curr_item);
1839 prev_item = curr_item;
1840 curr_item = __btrfs_next_delayed_item(prev_item);
1841 btrfs_release_delayed_item(prev_item);
1842 }
1843
1844 curr_item = __btrfs_first_delayed_deletion_item(delayed_node);
1845 while (curr_item) {
1846 btrfs_delayed_item_release_metadata(root, curr_item);
1847 prev_item = curr_item;
1848 curr_item = __btrfs_next_delayed_item(prev_item);
1849 btrfs_release_delayed_item(prev_item);
1850 }
1851
1852 if (delayed_node->inode_dirty) {
1853 btrfs_delayed_inode_release_metadata(root, delayed_node);
1854 btrfs_release_delayed_inode(delayed_node);
1855 }
1856 mutex_unlock(&delayed_node->mutex);
1857}
1858
1859void btrfs_kill_delayed_inode_items(struct inode *inode)
1860{
1861 struct btrfs_delayed_node *delayed_node;
1862
1863 delayed_node = btrfs_get_delayed_node(inode);
1864 if (!delayed_node)
1865 return;
1866
1867 __btrfs_kill_delayed_node(delayed_node);
1868 btrfs_release_delayed_node(delayed_node);
1869}
1870
1871void btrfs_kill_all_delayed_nodes(struct btrfs_root *root)
1872{
1873 u64 inode_id = 0;
1874 struct btrfs_delayed_node *delayed_nodes[8];
1875 int i, n;
1876
1877 while (1) {
1878 spin_lock(&root->inode_lock);
1879 n = radix_tree_gang_lookup(&root->delayed_nodes_tree,
1880 (void **)delayed_nodes, inode_id,
1881 ARRAY_SIZE(delayed_nodes));
1882 if (!n) {
1883 spin_unlock(&root->inode_lock);
1884 break;
1885 }
1886
1887 inode_id = delayed_nodes[n - 1]->inode_id + 1;
1888
1889 for (i = 0; i < n; i++)
1890 atomic_inc(&delayed_nodes[i]->refs);
1891 spin_unlock(&root->inode_lock);
1892
1893 for (i = 0; i < n; i++) {
1894 __btrfs_kill_delayed_node(delayed_nodes[i]);
1895 btrfs_release_delayed_node(delayed_nodes[i]);
1896 }
1897 }
1898}
67cde344
MX		 1899
1900void btrfs_destroy_delayed_inodes(struct btrfs_root *root)
1901{
1902 struct btrfs_delayed_root *delayed_root;
1903 struct btrfs_delayed_node *curr_node, *prev_node;
1904
1905 delayed_root = btrfs_get_delayed_root(root);
1906
1907 curr_node = btrfs_first_delayed_node(delayed_root);
1908 while (curr_node) {
1909 __btrfs_kill_delayed_node(curr_node);
1910
1911 prev_node = curr_node;
1912 curr_node = btrfs_next_delayed_node(curr_node);
1913 btrfs_release_delayed_node(prev_node);
1914 }
1915}
1916
Linux 6.x block layer and io_uring tree(s)