1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) 2007 Oracle. All rights reserved.
4 */
5
6 #include <linux/sched.h>
7 #include <linux/sched/signal.h>
8 #include <linux/pagemap.h>
9 #include <linux/writeback.h>
10 #include <linux/blkdev.h>
11 #include <linux/sort.h>
12 #include <linux/rcupdate.h>
13 #include <linux/kthread.h>
14 #include <linux/slab.h>
15 #include <linux/ratelimit.h>
16 #include <linux/percpu_counter.h>
17 #include <linux/lockdep.h>
18 #include <linux/crc32c.h>
19 #include "ctree.h"
20 #include "extent-tree.h"
21 #include "transaction.h"
22 #include "disk-io.h"
23 #include "print-tree.h"
24 #include "volumes.h"
25 #include "raid56.h"
26 #include "locking.h"
27 #include "free-space-cache.h"
28 #include "free-space-tree.h"
29 #include "qgroup.h"
30 #include "ref-verify.h"
31 #include "space-info.h"
32 #include "block-rsv.h"
33 #include "discard.h"
34 #include "zoned.h"
35 #include "dev-replace.h"
36 #include "fs.h"
37 #include "accessors.h"
38 #include "root-tree.h"
39 #include "file-item.h"
40 #include "orphan.h"
41 #include "tree-checker.h"
42 #include "raid-stripe-tree.h"
43
44 #undef SCRAMBLE_DELAYED_REFS
45
46
47 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
48 struct btrfs_delayed_ref_head *href,
49 struct btrfs_delayed_ref_node *node,
50 struct btrfs_delayed_extent_op *extra_op);
51 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
52 struct extent_buffer *leaf,
53 struct btrfs_extent_item *ei);
54 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
55 u64 parent, u64 root_objectid,
56 u64 flags, u64 owner, u64 offset,
57 struct btrfs_key *ins, int ref_mod, u64 oref_root);
58 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
59 struct btrfs_delayed_ref_node *node,
60 struct btrfs_delayed_extent_op *extent_op);
61 static int find_next_key(struct btrfs_path *path, int level,
62 struct btrfs_key *key);
63
64 static int block_group_bits(struct btrfs_block_group *cache, u64 bits)
65 {
66 return (cache->flags & bits) == bits;
67 }
68
69 /* simple helper to search for an existing data extent at a given offset */
70 int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
71 {
72 struct btrfs_root *root = btrfs_extent_root(fs_info, start);
73 int ret;
74 struct btrfs_key key;
75 struct btrfs_path *path;
76
77 path = btrfs_alloc_path();
78 if (!path)
79 return -ENOMEM;
80
81 key.objectid = start;
82 key.offset = len;
83 key.type = BTRFS_EXTENT_ITEM_KEY;
84 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
85 btrfs_free_path(path);
86 return ret;
87 }
88
89 /*
90 * helper function to lookup reference count and flags of a tree block.
91 *
92 * the head node for delayed ref is used to store the sum of all the
93 * reference count modifications queued up in the rbtree. the head
94 * node may also store the extent flags to set. This way you can check
95 * to see what the reference count and extent flags would be if all of
96 * the delayed refs are not processed.
97 */
98 int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
99 struct btrfs_fs_info *fs_info, u64 bytenr,
100 u64 offset, int metadata, u64 *refs, u64 *flags,
101 u64 *owning_root)
102 {
103 struct btrfs_root *extent_root;
104 struct btrfs_delayed_ref_head *head;
105 struct btrfs_delayed_ref_root *delayed_refs;
106 struct btrfs_path *path;
107 struct btrfs_key key;
108 u64 num_refs;
109 u64 extent_flags;
110 u64 owner = 0;
111 int ret;
112
113 /*
114 * If we don't have skinny metadata, don't bother doing anything
115 * different
116 */
117 if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
118 offset = fs_info->nodesize;
119 metadata = 0;
120 }
121
122 path = btrfs_alloc_path();
123 if (!path)
124 return -ENOMEM;
125
126 search_again:
127 key.objectid = bytenr;
128 key.offset = offset;
129 if (metadata)
130 key.type = BTRFS_METADATA_ITEM_KEY;
131 else
132 key.type = BTRFS_EXTENT_ITEM_KEY;
133
134 extent_root = btrfs_extent_root(fs_info, bytenr);
135 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
136 if (ret < 0)
137 goto out_free;
138
139 if (ret > 0 && key.type == BTRFS_METADATA_ITEM_KEY) {
140 if (path->slots[0]) {
141 path->slots[0]--;
142 btrfs_item_key_to_cpu(path->nodes[0], &key,
143 path->slots[0]);
144 if (key.objectid == bytenr &&
145 key.type == BTRFS_EXTENT_ITEM_KEY &&
146 key.offset == fs_info->nodesize)
147 ret = 0;
148 }
149 }
150
151 if (ret == 0) {
152 struct extent_buffer *leaf = path->nodes[0];
153 struct btrfs_extent_item *ei;
154 const u32 item_size = btrfs_item_size(leaf, path->slots[0]);
155
156 if (unlikely(item_size < sizeof(*ei))) {
157 ret = -EUCLEAN;
158 btrfs_err(fs_info,
159 "unexpected extent item size, has %u expect >= %zu",
160 item_size, sizeof(*ei));
161 btrfs_abort_transaction(trans, ret);
162 goto out_free;
163 }
164
165 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
166 num_refs = btrfs_extent_refs(leaf, ei);
167 if (unlikely(num_refs == 0)) {
168 ret = -EUCLEAN;
169 btrfs_err(fs_info,
170 "unexpected zero reference count for extent item (%llu %u %llu)",
171 key.objectid, key.type, key.offset);
172 btrfs_abort_transaction(trans, ret);
173 goto out_free;
174 }
175 extent_flags = btrfs_extent_flags(leaf, ei);
176 owner = btrfs_get_extent_owner_root(fs_info, leaf, path->slots[0]);
177 } else {
178 num_refs = 0;
179 extent_flags = 0;
180 ret = 0;
181 }
182
183 delayed_refs = &trans->transaction->delayed_refs;
184 spin_lock(&delayed_refs->lock);
185 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
186 if (head) {
187 if (!mutex_trylock(&head->mutex)) {
188 refcount_inc(&head->refs);
189 spin_unlock(&delayed_refs->lock);
190
191 btrfs_release_path(path);
192
193 /*
194 * Mutex was contended, block until it's released and try
195 * again
196 */
197 mutex_lock(&head->mutex);
198 mutex_unlock(&head->mutex);
199 btrfs_put_delayed_ref_head(head);
200 goto search_again;
201 }
202 spin_lock(&head->lock);
203 if (head->extent_op && head->extent_op->update_flags)
204 extent_flags |= head->extent_op->flags_to_set;
205
206 num_refs += head->ref_mod;
207 spin_unlock(&head->lock);
208 mutex_unlock(&head->mutex);
209 }
210 spin_unlock(&delayed_refs->lock);
211
212 WARN_ON(num_refs == 0);
213 if (refs)
214 *refs = num_refs;
215 if (flags)
216 *flags = extent_flags;
217 if (owning_root)
218 *owning_root = owner;
219 out_free:
220 btrfs_free_path(path);
221 return ret;
222 }
223
224 /*
225 * Back reference rules. Back refs have three main goals:
226 *
227 * 1) differentiate between all holders of references to an extent so that
228 * when a reference is dropped we can make sure it was a valid reference
229 * before freeing the extent.
230 *
231 * 2) Provide enough information to quickly find the holders of an extent
232 * if we notice a given block is corrupted or bad.
233 *
234 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
235 * maintenance. This is actually the same as #2, but with a slightly
236 * different use case.
237 *
238 * There are two kinds of back refs. The implicit back refs is optimized
239 * for pointers in non-shared tree blocks. For a given pointer in a block,
240 * back refs of this kind provide information about the block's owner tree
241 * and the pointer's key. These information allow us to find the block by
242 * b-tree searching. The full back refs is for pointers in tree blocks not
243 * referenced by their owner trees. The location of tree block is recorded
244 * in the back refs. Actually the full back refs is generic, and can be
245 * used in all cases the implicit back refs is used. The major shortcoming
246 * of the full back refs is its overhead. Every time a tree block gets
247 * COWed, we have to update back refs entry for all pointers in it.
248 *
249 * For a newly allocated tree block, we use implicit back refs for
250 * pointers in it. This means most tree related operations only involve
251 * implicit back refs. For a tree block created in old transaction, the
252 * only way to drop a reference to it is COW it. So we can detect the
253 * event that tree block loses its owner tree's reference and do the
254 * back refs conversion.
255 *
256 * When a tree block is COWed through a tree, there are four cases:
257 *
258 * The reference count of the block is one and the tree is the block's
259 * owner tree. Nothing to do in this case.
260 *
261 * The reference count of the block is one and the tree is not the
262 * block's owner tree. In this case, full back refs is used for pointers
263 * in the block. Remove these full back refs, add implicit back refs for
264 * every pointers in the new block.
265 *
266 * The reference count of the block is greater than one and the tree is
267 * the block's owner tree. In this case, implicit back refs is used for
268 * pointers in the block. Add full back refs for every pointers in the
269 * block, increase lower level extents' reference counts. The original
270 * implicit back refs are entailed to the new block.
271 *
272 * The reference count of the block is greater than one and the tree is
273 * not the block's owner tree. Add implicit back refs for every pointer in
274 * the new block, increase lower level extents' reference count.
275 *
276 * Back Reference Key composing:
277 *
278 * The key objectid corresponds to the first byte in the extent,
279 * The key type is used to differentiate between types of back refs.
280 * There are different meanings of the key offset for different types
281 * of back refs.
282 *
283 * File extents can be referenced by:
284 *
285 * - multiple snapshots, subvolumes, or different generations in one subvol
286 * - different files inside a single subvolume
287 * - different offsets inside a file (bookend extents in file.c)
288 *
289 * The extent ref structure for the implicit back refs has fields for:
290 *
291 * - Objectid of the subvolume root
292 * - objectid of the file holding the reference
293 * - original offset in the file
294 * - how many bookend extents
295 *
296 * The key offset for the implicit back refs is hash of the first
297 * three fields.
298 *
299 * The extent ref structure for the full back refs has field for:
300 *
301 * - number of pointers in the tree leaf
302 *
303 * The key offset for the implicit back refs is the first byte of
304 * the tree leaf
305 *
306 * When a file extent is allocated, The implicit back refs is used.
307 * the fields are filled in:
308 *
309 * (root_key.objectid, inode objectid, offset in file, 1)
310 *
311 * When a file extent is removed file truncation, we find the
312 * corresponding implicit back refs and check the following fields:
313 *
314 * (btrfs_header_owner(leaf), inode objectid, offset in file)
315 *
316 * Btree extents can be referenced by:
317 *
318 * - Different subvolumes
319 *
320 * Both the implicit back refs and the full back refs for tree blocks
321 * only consist of key. The key offset for the implicit back refs is
322 * objectid of block's owner tree. The key offset for the full back refs
323 * is the first byte of parent block.
324 *
325 * When implicit back refs is used, information about the lowest key and
326 * level of the tree block are required. These information are stored in
327 * tree block info structure.
328 */
329
330 /*
331 * is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
332 * is_data == BTRFS_REF_TYPE_DATA, data type is requiried,
333 * is_data == BTRFS_REF_TYPE_ANY, either type is OK.
334 */
335 int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
336 struct btrfs_extent_inline_ref *iref,
337 enum btrfs_inline_ref_type is_data)
338 {
339 struct btrfs_fs_info *fs_info = eb->fs_info;
340 int type = btrfs_extent_inline_ref_type(eb, iref);
341 u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
342
343 if (type == BTRFS_EXTENT_OWNER_REF_KEY) {
344 ASSERT(btrfs_fs_incompat(fs_info, SIMPLE_QUOTA));
345 return type;
346 }
347
348 if (type == BTRFS_TREE_BLOCK_REF_KEY ||
349 type == BTRFS_SHARED_BLOCK_REF_KEY ||
350 type == BTRFS_SHARED_DATA_REF_KEY ||
351 type == BTRFS_EXTENT_DATA_REF_KEY) {
352 if (is_data == BTRFS_REF_TYPE_BLOCK) {
353 if (type == BTRFS_TREE_BLOCK_REF_KEY)
354 return type;
355 if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
356 ASSERT(fs_info);
357 /*
358 * Every shared one has parent tree block,
359 * which must be aligned to sector size.
360 */
361 if (offset && IS_ALIGNED(offset, fs_info->sectorsize))
362 return type;
363 }
364 } else if (is_data == BTRFS_REF_TYPE_DATA) {
365 if (type == BTRFS_EXTENT_DATA_REF_KEY)
366 return type;
367 if (type == BTRFS_SHARED_DATA_REF_KEY) {
368 ASSERT(fs_info);
369 /*
370 * Every shared one has parent tree block,
371 * which must be aligned to sector size.
372 */
373 if (offset &&
374 IS_ALIGNED(offset, fs_info->sectorsize))
375 return type;
376 }
377 } else {
378 ASSERT(is_data == BTRFS_REF_TYPE_ANY);
379 return type;
380 }
381 }
382
383 WARN_ON(1);
384 btrfs_print_leaf(eb);
385 btrfs_err(fs_info,
386 "eb %llu iref 0x%lx invalid extent inline ref type %d",
387 eb->start, (unsigned long)iref, type);
388
389 return BTRFS_REF_TYPE_INVALID;
390 }
391
392 u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
393 {
394 u32 high_crc = ~(u32)0;
395 u32 low_crc = ~(u32)0;
396 __le64 lenum;
397
398 lenum = cpu_to_le64(root_objectid);
399 high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
400 lenum = cpu_to_le64(owner);
401 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
402 lenum = cpu_to_le64(offset);
403 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
404
405 return ((u64)high_crc << 31) ^ (u64)low_crc;
406 }
407
408 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
409 struct btrfs_extent_data_ref *ref)
410 {
411 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
412 btrfs_extent_data_ref_objectid(leaf, ref),
413 btrfs_extent_data_ref_offset(leaf, ref));
414 }
415
416 static int match_extent_data_ref(struct extent_buffer *leaf,
417 struct btrfs_extent_data_ref *ref,
418 u64 root_objectid, u64 owner, u64 offset)
419 {
420 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
421 btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
422 btrfs_extent_data_ref_offset(leaf, ref) != offset)
423 return 0;
424 return 1;
425 }
426
427 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
428 struct btrfs_path *path,
429 u64 bytenr, u64 parent,
430 u64 root_objectid,
431 u64 owner, u64 offset)
432 {
433 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr);
434 struct btrfs_key key;
435 struct btrfs_extent_data_ref *ref;
436 struct extent_buffer *leaf;
437 u32 nritems;
438 int recow;
439 int ret;
440
441 key.objectid = bytenr;
442 if (parent) {
443 key.type = BTRFS_SHARED_DATA_REF_KEY;
444 key.offset = parent;
445 } else {
446 key.type = BTRFS_EXTENT_DATA_REF_KEY;
447 key.offset = hash_extent_data_ref(root_objectid,
448 owner, offset);
449 }
450 again:
451 recow = 0;
452 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
453 if (ret < 0)
454 return ret;
455
456 if (parent) {
457 if (ret)
458 return -ENOENT;
459 return 0;
460 }
461
462 ret = -ENOENT;
463 leaf = path->nodes[0];
464 nritems = btrfs_header_nritems(leaf);
465 while (1) {
466 if (path->slots[0] >= nritems) {
467 ret = btrfs_next_leaf(root, path);
468 if (ret) {
469 if (ret > 0)
470 return -ENOENT;
471 return ret;
472 }
473
474 leaf = path->nodes[0];
475 nritems = btrfs_header_nritems(leaf);
476 recow = 1;
477 }
478
479 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
480 if (key.objectid != bytenr ||
481 key.type != BTRFS_EXTENT_DATA_REF_KEY)
482 goto fail;
483
484 ref = btrfs_item_ptr(leaf, path->slots[0],
485 struct btrfs_extent_data_ref);
486
487 if (match_extent_data_ref(leaf, ref, root_objectid,
488 owner, offset)) {
489 if (recow) {
490 btrfs_release_path(path);
491 goto again;
492 }
493 ret = 0;
494 break;
495 }
496 path->slots[0]++;
497 }
498 fail:
499 return ret;
500 }
501
502 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
503 struct btrfs_path *path,
504 struct btrfs_delayed_ref_node *node,
505 u64 bytenr)
506 {
507 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr);
508 struct btrfs_key key;
509 struct extent_buffer *leaf;
510 u64 owner = btrfs_delayed_ref_owner(node);
511 u64 offset = btrfs_delayed_ref_offset(node);
512 u32 size;
513 u32 num_refs;
514 int ret;
515
516 key.objectid = bytenr;
517 if (node->parent) {
518 key.type = BTRFS_SHARED_DATA_REF_KEY;
519 key.offset = node->parent;
520 size = sizeof(struct btrfs_shared_data_ref);
521 } else {
522 key.type = BTRFS_EXTENT_DATA_REF_KEY;
523 key.offset = hash_extent_data_ref(node->ref_root, owner, offset);
524 size = sizeof(struct btrfs_extent_data_ref);
525 }
526
527 ret = btrfs_insert_empty_item(trans, root, path, &key, size);
528 if (ret && ret != -EEXIST)
529 goto fail;
530
531 leaf = path->nodes[0];
532 if (node->parent) {
533 struct btrfs_shared_data_ref *ref;
534 ref = btrfs_item_ptr(leaf, path->slots[0],
535 struct btrfs_shared_data_ref);
536 if (ret == 0) {
537 btrfs_set_shared_data_ref_count(leaf, ref, node->ref_mod);
538 } else {
539 num_refs = btrfs_shared_data_ref_count(leaf, ref);
540 num_refs += node->ref_mod;
541 btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
542 }
543 } else {
544 struct btrfs_extent_data_ref *ref;
545 while (ret == -EEXIST) {
546 ref = btrfs_item_ptr(leaf, path->slots[0],
547 struct btrfs_extent_data_ref);
548 if (match_extent_data_ref(leaf, ref, node->ref_root,
549 owner, offset))
550 break;
551 btrfs_release_path(path);
552 key.offset++;
553 ret = btrfs_insert_empty_item(trans, root, path, &key,
554 size);
555 if (ret && ret != -EEXIST)
556 goto fail;
557
558 leaf = path->nodes[0];
559 }
560 ref = btrfs_item_ptr(leaf, path->slots[0],
561 struct btrfs_extent_data_ref);
562 if (ret == 0) {
563 btrfs_set_extent_data_ref_root(leaf, ref, node->ref_root);
564 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
565 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
566 btrfs_set_extent_data_ref_count(leaf, ref, node->ref_mod);
567 } else {
568 num_refs = btrfs_extent_data_ref_count(leaf, ref);
569 num_refs += node->ref_mod;
570 btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
571 }
572 }
573 btrfs_mark_buffer_dirty(trans, leaf);
574 ret = 0;
575 fail:
576 btrfs_release_path(path);
577 return ret;
578 }
579
580 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
581 struct btrfs_root *root,
582 struct btrfs_path *path,
583 int refs_to_drop)
584 {
585 struct btrfs_key key;
586 struct btrfs_extent_data_ref *ref1 = NULL;
587 struct btrfs_shared_data_ref *ref2 = NULL;
588 struct extent_buffer *leaf;
589 u32 num_refs = 0;
590 int ret = 0;
591
592 leaf = path->nodes[0];
593 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
594
595 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
596 ref1 = btrfs_item_ptr(leaf, path->slots[0],
597 struct btrfs_extent_data_ref);
598 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
599 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
600 ref2 = btrfs_item_ptr(leaf, path->slots[0],
601 struct btrfs_shared_data_ref);
602 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
603 } else {
604 btrfs_err(trans->fs_info,
605 "unrecognized backref key (%llu %u %llu)",
606 key.objectid, key.type, key.offset);
607 btrfs_abort_transaction(trans, -EUCLEAN);
608 return -EUCLEAN;
609 }
610
611 BUG_ON(num_refs < refs_to_drop);
612 num_refs -= refs_to_drop;
613
614 if (num_refs == 0) {
615 ret = btrfs_del_item(trans, root, path);
616 } else {
617 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
618 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
619 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
620 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
621 btrfs_mark_buffer_dirty(trans, leaf);
622 }
623 return ret;
624 }
625
626 static noinline u32 extent_data_ref_count(struct btrfs_path *path,
627 struct btrfs_extent_inline_ref *iref)
628 {
629 struct btrfs_key key;
630 struct extent_buffer *leaf;
631 struct btrfs_extent_data_ref *ref1;
632 struct btrfs_shared_data_ref *ref2;
633 u32 num_refs = 0;
634 int type;
635
636 leaf = path->nodes[0];
637 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
638
639 if (iref) {
640 /*
641 * If type is invalid, we should have bailed out earlier than
642 * this call.
643 */
644 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
645 ASSERT(type != BTRFS_REF_TYPE_INVALID);
646 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
647 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
648 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
649 } else {
650 ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
651 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
652 }
653 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
654 ref1 = btrfs_item_ptr(leaf, path->slots[0],
655 struct btrfs_extent_data_ref);
656 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
657 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
658 ref2 = btrfs_item_ptr(leaf, path->slots[0],
659 struct btrfs_shared_data_ref);
660 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
661 } else {
662 WARN_ON(1);
663 }
664 return num_refs;
665 }
666
667 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
668 struct btrfs_path *path,
669 u64 bytenr, u64 parent,
670 u64 root_objectid)
671 {
672 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr);
673 struct btrfs_key key;
674 int ret;
675
676 key.objectid = bytenr;
677 if (parent) {
678 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
679 key.offset = parent;
680 } else {
681 key.type = BTRFS_TREE_BLOCK_REF_KEY;
682 key.offset = root_objectid;
683 }
684
685 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
686 if (ret > 0)
687 ret = -ENOENT;
688 return ret;
689 }
690
691 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
692 struct btrfs_path *path,
693 struct btrfs_delayed_ref_node *node,
694 u64 bytenr)
695 {
696 struct btrfs_root *root = btrfs_extent_root(trans->fs_info, bytenr);
697 struct btrfs_key key;
698 int ret;
699
700 key.objectid = bytenr;
701 if (node->parent) {
702 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
703 key.offset = node->parent;
704 } else {
705 key.type = BTRFS_TREE_BLOCK_REF_KEY;
706 key.offset = node->ref_root;
707 }
708
709 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
710 btrfs_release_path(path);
711 return ret;
712 }
713
714 static inline int extent_ref_type(u64 parent, u64 owner)
715 {
716 int type;
717 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
718 if (parent > 0)
719 type = BTRFS_SHARED_BLOCK_REF_KEY;
720 else
721 type = BTRFS_TREE_BLOCK_REF_KEY;
722 } else {
723 if (parent > 0)
724 type = BTRFS_SHARED_DATA_REF_KEY;
725 else
726 type = BTRFS_EXTENT_DATA_REF_KEY;
727 }
728 return type;
729 }
730
731 static int find_next_key(struct btrfs_path *path, int level,
732 struct btrfs_key *key)
733
734 {
735 for (; level < BTRFS_MAX_LEVEL; level++) {
736 if (!path->nodes[level])
737 break;
738 if (path->slots[level] + 1 >=
739 btrfs_header_nritems(path->nodes[level]))
740 continue;
741 if (level == 0)
742 btrfs_item_key_to_cpu(path->nodes[level], key,
743 path->slots[level] + 1);
744 else
745 btrfs_node_key_to_cpu(path->nodes[level], key,
746 path->slots[level] + 1);
747 return 0;
748 }
749 return 1;
750 }
751
752 /*
753 * look for inline back ref. if back ref is found, *ref_ret is set
754 * to the address of inline back ref, and 0 is returned.
755 *
756 * if back ref isn't found, *ref_ret is set to the address where it
757 * should be inserted, and -ENOENT is returned.
758 *
759 * if insert is true and there are too many inline back refs, the path
760 * points to the extent item, and -EAGAIN is returned.
761 *
762 * NOTE: inline back refs are ordered in the same way that back ref
763 * items in the tree are ordered.
764 */
765 static noinline_for_stack
766 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
767 struct btrfs_path *path,
768 struct btrfs_extent_inline_ref **ref_ret,
769 u64 bytenr, u64 num_bytes,
770 u64 parent, u64 root_objectid,
771 u64 owner, u64 offset, int insert)
772 {
773 struct btrfs_fs_info *fs_info = trans->fs_info;
774 struct btrfs_root *root = btrfs_extent_root(fs_info, bytenr);
775 struct btrfs_key key;
776 struct extent_buffer *leaf;
777 struct btrfs_extent_item *ei;
778 struct btrfs_extent_inline_ref *iref;
779 u64 flags;
780 u64 item_size;
781 unsigned long ptr;
782 unsigned long end;
783 int extra_size;
784 int type;
785 int want;
786 int ret;
787 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
788 int needed;
789
790 key.objectid = bytenr;
791 key.type = BTRFS_EXTENT_ITEM_KEY;
792 key.offset = num_bytes;
793
794 want = extent_ref_type(parent, owner);
795 if (insert) {
796 extra_size = btrfs_extent_inline_ref_size(want);
797 path->search_for_extension = 1;
798 path->keep_locks = 1;
799 } else
800 extra_size = -1;
801
802 /*
803 * Owner is our level, so we can just add one to get the level for the
804 * block we are interested in.
805 */
806 if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
807 key.type = BTRFS_METADATA_ITEM_KEY;
808 key.offset = owner;
809 }
810
811 again:
812 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
813 if (ret < 0)
814 goto out;
815
816 /*
817 * We may be a newly converted file system which still has the old fat
818 * extent entries for metadata, so try and see if we have one of those.
819 */
820 if (ret > 0 && skinny_metadata) {
821 skinny_metadata = false;
822 if (path->slots[0]) {
823 path->slots[0]--;
824 btrfs_item_key_to_cpu(path->nodes[0], &key,
825 path->slots[0]);
826 if (key.objectid == bytenr &&
827 key.type == BTRFS_EXTENT_ITEM_KEY &&
828 key.offset == num_bytes)
829 ret = 0;
830 }
831 if (ret) {
832 key.objectid = bytenr;
833 key.type = BTRFS_EXTENT_ITEM_KEY;
834 key.offset = num_bytes;
835 btrfs_release_path(path);
836 goto again;
837 }
838 }
839
840 if (ret && !insert) {
841 ret = -ENOENT;
842 goto out;
843 } else if (WARN_ON(ret)) {
844 btrfs_print_leaf(path->nodes[0]);
845 btrfs_err(fs_info,
846 "extent item not found for insert, bytenr %llu num_bytes %llu parent %llu root_objectid %llu owner %llu offset %llu",
847 bytenr, num_bytes, parent, root_objectid, owner,
848 offset);
849 ret = -EUCLEAN;
850 goto out;
851 }
852
853 leaf = path->nodes[0];
854 item_size = btrfs_item_size(leaf, path->slots[0]);
855 if (unlikely(item_size < sizeof(*ei))) {
856 ret = -EUCLEAN;
857 btrfs_err(fs_info,
858 "unexpected extent item size, has %llu expect >= %zu",
859 item_size, sizeof(*ei));
860 btrfs_abort_transaction(trans, ret);
861 goto out;
862 }
863
864 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
865 flags = btrfs_extent_flags(leaf, ei);
866
867 ptr = (unsigned long)(ei + 1);
868 end = (unsigned long)ei + item_size;
869
870 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
871 ptr += sizeof(struct btrfs_tree_block_info);
872 BUG_ON(ptr > end);
873 }
874
875 if (owner >= BTRFS_FIRST_FREE_OBJECTID)
876 needed = BTRFS_REF_TYPE_DATA;
877 else
878 needed = BTRFS_REF_TYPE_BLOCK;
879
880 ret = -ENOENT;
881 while (ptr < end) {
882 iref = (struct btrfs_extent_inline_ref *)ptr;
883 type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
884 if (type == BTRFS_EXTENT_OWNER_REF_KEY) {
885 ASSERT(btrfs_fs_incompat(fs_info, SIMPLE_QUOTA));
886 ptr += btrfs_extent_inline_ref_size(type);
887 continue;
888 }
889 if (type == BTRFS_REF_TYPE_INVALID) {
890 ret = -EUCLEAN;
891 goto out;
892 }
893
894 if (want < type)
895 break;
896 if (want > type) {
897 ptr += btrfs_extent_inline_ref_size(type);
898 continue;
899 }
900
901 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
902 struct btrfs_extent_data_ref *dref;
903 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
904 if (match_extent_data_ref(leaf, dref, root_objectid,
905 owner, offset)) {
906 ret = 0;
907 break;
908 }
909 if (hash_extent_data_ref_item(leaf, dref) <
910 hash_extent_data_ref(root_objectid, owner, offset))
911 break;
912 } else {
913 u64 ref_offset;
914 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
915 if (parent > 0) {
916 if (parent == ref_offset) {
917 ret = 0;
918 break;
919 }
920 if (ref_offset < parent)
921 break;
922 } else {
923 if (root_objectid == ref_offset) {
924 ret = 0;
925 break;
926 }
927 if (ref_offset < root_objectid)
928 break;
929 }
930 }
931 ptr += btrfs_extent_inline_ref_size(type);
932 }
933
934 if (unlikely(ptr > end)) {
935 ret = -EUCLEAN;
936 btrfs_print_leaf(path->nodes[0]);
937 btrfs_crit(fs_info,
938 "overrun extent record at slot %d while looking for inline extent for root %llu owner %llu offset %llu parent %llu",
939 path->slots[0], root_objectid, owner, offset, parent);
940 goto out;
941 }
942
943 if (ret == -ENOENT && insert) {
944 if (item_size + extra_size >=
945 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
946 ret = -EAGAIN;
947 goto out;
948 }
949 /*
950 * To add new inline back ref, we have to make sure
951 * there is no corresponding back ref item.
952 * For simplicity, we just do not add new inline back
953 * ref if there is any kind of item for this block
954 */
955 if (find_next_key(path, 0, &key) == 0 &&
956 key.objectid == bytenr &&
957 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
958 ret = -EAGAIN;
959 goto out;
960 }
961 }
962 *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
963 out:
964 if (insert) {
965 path->keep_locks = 0;
966 path->search_for_extension = 0;
967 btrfs_unlock_up_safe(path, 1);
968 }
969 return ret;
970 }
971
972 /*
973 * helper to add new inline back ref
974 */
975 static noinline_for_stack
976 void setup_inline_extent_backref(struct btrfs_trans_handle *trans,
977 struct btrfs_path *path,
978 struct btrfs_extent_inline_ref *iref,
979 u64 parent, u64 root_objectid,
980 u64 owner, u64 offset, int refs_to_add,
981 struct btrfs_delayed_extent_op *extent_op)
982 {
983 struct extent_buffer *leaf;
984 struct btrfs_extent_item *ei;
985 unsigned long ptr;
986 unsigned long end;
987 unsigned long item_offset;
988 u64 refs;
989 int size;
990 int type;
991
992 leaf = path->nodes[0];
993 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
994 item_offset = (unsigned long)iref - (unsigned long)ei;
995
996 type = extent_ref_type(parent, owner);
997 size = btrfs_extent_inline_ref_size(type);
998
999 btrfs_extend_item(trans, path, size);
1000
1001 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1002 refs = btrfs_extent_refs(leaf, ei);
1003 refs += refs_to_add;
1004 btrfs_set_extent_refs(leaf, ei, refs);
1005 if (extent_op)
1006 __run_delayed_extent_op(extent_op, leaf, ei);
1007
1008 ptr = (unsigned long)ei + item_offset;
1009 end = (unsigned long)ei + btrfs_item_size(leaf, path->slots[0]);
1010 if (ptr < end - size)
1011 memmove_extent_buffer(leaf, ptr + size, ptr,
1012 end - size - ptr);
1013
1014 iref = (struct btrfs_extent_inline_ref *)ptr;
1015 btrfs_set_extent_inline_ref_type(leaf, iref, type);
1016 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1017 struct btrfs_extent_data_ref *dref;
1018 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1019 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1020 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1021 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1022 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1023 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1024 struct btrfs_shared_data_ref *sref;
1025 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1026 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1027 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1028 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1029 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1030 } else {
1031 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1032 }
1033 btrfs_mark_buffer_dirty(trans, leaf);
1034 }
1035
1036 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1037 struct btrfs_path *path,
1038 struct btrfs_extent_inline_ref **ref_ret,
1039 u64 bytenr, u64 num_bytes, u64 parent,
1040 u64 root_objectid, u64 owner, u64 offset)
1041 {
1042 int ret;
1043
1044 ret = lookup_inline_extent_backref(trans, path, ref_ret, bytenr,
1045 num_bytes, parent, root_objectid,
1046 owner, offset, 0);
1047 if (ret != -ENOENT)
1048 return ret;
1049
1050 btrfs_release_path(path);
1051 *ref_ret = NULL;
1052
1053 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1054 ret = lookup_tree_block_ref(trans, path, bytenr, parent,
1055 root_objectid);
1056 } else {
1057 ret = lookup_extent_data_ref(trans, path, bytenr, parent,
1058 root_objectid, owner, offset);
1059 }
1060 return ret;
1061 }
1062
1063 /*
1064 * helper to update/remove inline back ref
1065 */
1066 static noinline_for_stack int update_inline_extent_backref(
1067 struct btrfs_trans_handle *trans,
1068 struct btrfs_path *path,
1069 struct btrfs_extent_inline_ref *iref,
1070 int refs_to_mod,
1071 struct btrfs_delayed_extent_op *extent_op)
1072 {
1073 struct extent_buffer *leaf = path->nodes[0];
1074 struct btrfs_fs_info *fs_info = leaf->fs_info;
1075 struct btrfs_extent_item *ei;
1076 struct btrfs_extent_data_ref *dref = NULL;
1077 struct btrfs_shared_data_ref *sref = NULL;
1078 unsigned long ptr;
1079 unsigned long end;
1080 u32 item_size;
1081 int size;
1082 int type;
1083 u64 refs;
1084
1085 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1086 refs = btrfs_extent_refs(leaf, ei);
1087 if (unlikely(refs_to_mod < 0 && refs + refs_to_mod <= 0)) {
1088 struct btrfs_key key;
1089 u32 extent_size;
1090
1091 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1092 if (key.type == BTRFS_METADATA_ITEM_KEY)
1093 extent_size = fs_info->nodesize;
1094 else
1095 extent_size = key.offset;
1096 btrfs_print_leaf(leaf);
1097 btrfs_err(fs_info,
1098 "invalid refs_to_mod for extent %llu num_bytes %u, has %d expect >= -%llu",
1099 key.objectid, extent_size, refs_to_mod, refs);
1100 return -EUCLEAN;
1101 }
1102 refs += refs_to_mod;
1103 btrfs_set_extent_refs(leaf, ei, refs);
1104 if (extent_op)
1105 __run_delayed_extent_op(extent_op, leaf, ei);
1106
1107 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
1108 /*
1109 * Function btrfs_get_extent_inline_ref_type() has already printed
1110 * error messages.
1111 */
1112 if (unlikely(type == BTRFS_REF_TYPE_INVALID))
1113 return -EUCLEAN;
1114
1115 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1116 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1117 refs = btrfs_extent_data_ref_count(leaf, dref);
1118 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1119 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1120 refs = btrfs_shared_data_ref_count(leaf, sref);
1121 } else {
1122 refs = 1;
1123 /*
1124 * For tree blocks we can only drop one ref for it, and tree
1125 * blocks should not have refs > 1.
1126 *
1127 * Furthermore if we're inserting a new inline backref, we
1128 * won't reach this path either. That would be
1129 * setup_inline_extent_backref().
1130 */
1131 if (unlikely(refs_to_mod != -1)) {
1132 struct btrfs_key key;
1133
1134 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1135
1136 btrfs_print_leaf(leaf);
1137 btrfs_err(fs_info,
1138 "invalid refs_to_mod for tree block %llu, has %d expect -1",
1139 key.objectid, refs_to_mod);
1140 return -EUCLEAN;
1141 }
1142 }
1143
1144 if (unlikely(refs_to_mod < 0 && refs < -refs_to_mod)) {
1145 struct btrfs_key key;
1146 u32 extent_size;
1147
1148 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1149 if (key.type == BTRFS_METADATA_ITEM_KEY)
1150 extent_size = fs_info->nodesize;
1151 else
1152 extent_size = key.offset;
1153 btrfs_print_leaf(leaf);
1154 btrfs_err(fs_info,
1155 "invalid refs_to_mod for backref entry, iref %lu extent %llu num_bytes %u, has %d expect >= -%llu",
1156 (unsigned long)iref, key.objectid, extent_size,
1157 refs_to_mod, refs);
1158 return -EUCLEAN;
1159 }
1160 refs += refs_to_mod;
1161
1162 if (refs > 0) {
1163 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1164 btrfs_set_extent_data_ref_count(leaf, dref, refs);
1165 else
1166 btrfs_set_shared_data_ref_count(leaf, sref, refs);
1167 } else {
1168 size = btrfs_extent_inline_ref_size(type);
1169 item_size = btrfs_item_size(leaf, path->slots[0]);
1170 ptr = (unsigned long)iref;
1171 end = (unsigned long)ei + item_size;
1172 if (ptr + size < end)
1173 memmove_extent_buffer(leaf, ptr, ptr + size,
1174 end - ptr - size);
1175 item_size -= size;
1176 btrfs_truncate_item(trans, path, item_size, 1);
1177 }
1178 btrfs_mark_buffer_dirty(trans, leaf);
1179 return 0;
1180 }
1181
1182 static noinline_for_stack
1183 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1184 struct btrfs_path *path,
1185 u64 bytenr, u64 num_bytes, u64 parent,
1186 u64 root_objectid, u64 owner,
1187 u64 offset, int refs_to_add,
1188 struct btrfs_delayed_extent_op *extent_op)
1189 {
1190 struct btrfs_extent_inline_ref *iref;
1191 int ret;
1192
1193 ret = lookup_inline_extent_backref(trans, path, &iref, bytenr,
1194 num_bytes, parent, root_objectid,
1195 owner, offset, 1);
1196 if (ret == 0) {
1197 /*
1198 * We're adding refs to a tree block we already own, this
1199 * should not happen at all.
1200 */
1201 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1202 btrfs_print_leaf(path->nodes[0]);
1203 btrfs_crit(trans->fs_info,
1204 "adding refs to an existing tree ref, bytenr %llu num_bytes %llu root_objectid %llu slot %u",
1205 bytenr, num_bytes, root_objectid, path->slots[0]);
1206 return -EUCLEAN;
1207 }
1208 ret = update_inline_extent_backref(trans, path, iref,
1209 refs_to_add, extent_op);
1210 } else if (ret == -ENOENT) {
1211 setup_inline_extent_backref(trans, path, iref, parent,
1212 root_objectid, owner, offset,
1213 refs_to_add, extent_op);
1214 ret = 0;
1215 }
1216 return ret;
1217 }
1218
1219 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1220 struct btrfs_root *root,
1221 struct btrfs_path *path,
1222 struct btrfs_extent_inline_ref *iref,
1223 int refs_to_drop, int is_data)
1224 {
1225 int ret = 0;
1226
1227 BUG_ON(!is_data && refs_to_drop != 1);
1228 if (iref)
1229 ret = update_inline_extent_backref(trans, path, iref,
1230 -refs_to_drop, NULL);
1231 else if (is_data)
1232 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1233 else
1234 ret = btrfs_del_item(trans, root, path);
1235 return ret;
1236 }
1237
1238 static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
1239 u64 *discarded_bytes)
1240 {
1241 int j, ret = 0;
1242 u64 bytes_left, end;
1243 u64 aligned_start = ALIGN(start, 1 << SECTOR_SHIFT);
1244
1245 /* Adjust the range to be aligned to 512B sectors if necessary. */
1246 if (start != aligned_start) {
1247 len -= aligned_start - start;
1248 len = round_down(len, 1 << SECTOR_SHIFT);
1249 start = aligned_start;
1250 }
1251
1252 *discarded_bytes = 0;
1253
1254 if (!len)
1255 return 0;
1256
1257 end = start + len;
1258 bytes_left = len;
1259
1260 /* Skip any superblocks on this device. */
1261 for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
1262 u64 sb_start = btrfs_sb_offset(j);
1263 u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
1264 u64 size = sb_start - start;
1265
1266 if (!in_range(sb_start, start, bytes_left) &&
1267 !in_range(sb_end, start, bytes_left) &&
1268 !in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
1269 continue;
1270
1271 /*
1272 * Superblock spans beginning of range. Adjust start and
1273 * try again.
1274 */
1275 if (sb_start <= start) {
1276 start += sb_end - start;
1277 if (start > end) {
1278 bytes_left = 0;
1279 break;
1280 }
1281 bytes_left = end - start;
1282 continue;
1283 }
1284
1285 if (size) {
1286 ret = blkdev_issue_discard(bdev, start >> SECTOR_SHIFT,
1287 size >> SECTOR_SHIFT,
1288 GFP_NOFS);
1289 if (!ret)
1290 *discarded_bytes += size;
1291 else if (ret != -EOPNOTSUPP)
1292 return ret;
1293 }
1294
1295 start = sb_end;
1296 if (start > end) {
1297 bytes_left = 0;
1298 break;
1299 }
1300 bytes_left = end - start;
1301 }
1302
1303 if (bytes_left) {
1304 ret = blkdev_issue_discard(bdev, start >> SECTOR_SHIFT,
1305 bytes_left >> SECTOR_SHIFT,
1306 GFP_NOFS);
1307 if (!ret)
1308 *discarded_bytes += bytes_left;
1309 }
1310 return ret;
1311 }
1312
1313 static int do_discard_extent(struct btrfs_discard_stripe *stripe, u64 *bytes)
1314 {
1315 struct btrfs_device *dev = stripe->dev;
1316 struct btrfs_fs_info *fs_info = dev->fs_info;
1317 struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
1318 u64 phys = stripe->physical;
1319 u64 len = stripe->length;
1320 u64 discarded = 0;
1321 int ret = 0;
1322
1323 /* Zone reset on a zoned filesystem */
1324 if (btrfs_can_zone_reset(dev, phys, len)) {
1325 u64 src_disc;
1326
1327 ret = btrfs_reset_device_zone(dev, phys, len, &discarded);
1328 if (ret)
1329 goto out;
1330
1331 if (!btrfs_dev_replace_is_ongoing(dev_replace) ||
1332 dev != dev_replace->srcdev)
1333 goto out;
1334
1335 src_disc = discarded;
1336
1337 /* Send to replace target as well */
1338 ret = btrfs_reset_device_zone(dev_replace->tgtdev, phys, len,
1339 &discarded);
1340 discarded += src_disc;
1341 } else if (bdev_max_discard_sectors(stripe->dev->bdev)) {
1342 ret = btrfs_issue_discard(dev->bdev, phys, len, &discarded);
1343 } else {
1344 ret = 0;
1345 *bytes = 0;
1346 }
1347
1348 out:
1349 *bytes = discarded;
1350 return ret;
1351 }
1352
1353 int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
1354 u64 num_bytes, u64 *actual_bytes)
1355 {
1356 int ret = 0;
1357 u64 discarded_bytes = 0;
1358 u64 end = bytenr + num_bytes;
1359 u64 cur = bytenr;
1360
1361 /*
1362 * Avoid races with device replace and make sure the devices in the
1363 * stripes don't go away while we are discarding.
1364 */
1365 btrfs_bio_counter_inc_blocked(fs_info);
1366 while (cur < end) {
1367 struct btrfs_discard_stripe *stripes;
1368 unsigned int num_stripes;
1369 int i;
1370
1371 num_bytes = end - cur;
1372 stripes = btrfs_map_discard(fs_info, cur, &num_bytes, &num_stripes);
1373 if (IS_ERR(stripes)) {
1374 ret = PTR_ERR(stripes);
1375 if (ret == -EOPNOTSUPP)
1376 ret = 0;
1377 break;
1378 }
1379
1380 for (i = 0; i < num_stripes; i++) {
1381 struct btrfs_discard_stripe *stripe = stripes + i;
1382 u64 bytes;
1383
1384 if (!stripe->dev->bdev) {
1385 ASSERT(btrfs_test_opt(fs_info, DEGRADED));
1386 continue;
1387 }
1388
1389 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE,
1390 &stripe->dev->dev_state))
1391 continue;
1392
1393 ret = do_discard_extent(stripe, &bytes);
1394 if (ret) {
1395 /*
1396 * Keep going if discard is not supported by the
1397 * device.
1398 */
1399 if (ret != -EOPNOTSUPP)
1400 break;
1401 ret = 0;
1402 } else {
1403 discarded_bytes += bytes;
1404 }
1405 }
1406 kfree(stripes);
1407 if (ret)
1408 break;
1409 cur += num_bytes;
1410 }
1411 btrfs_bio_counter_dec(fs_info);
1412 if (actual_bytes)
1413 *actual_bytes = discarded_bytes;
1414 return ret;
1415 }
1416
1417 /* Can return -ENOMEM */
1418 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1419 struct btrfs_ref *generic_ref)
1420 {
1421 struct btrfs_fs_info *fs_info = trans->fs_info;
1422 int ret;
1423
1424 ASSERT(generic_ref->type != BTRFS_REF_NOT_SET &&
1425 generic_ref->action);
1426 BUG_ON(generic_ref->type == BTRFS_REF_METADATA &&
1427 generic_ref->ref_root == BTRFS_TREE_LOG_OBJECTID);
1428
1429 if (generic_ref->type == BTRFS_REF_METADATA)
1430 ret = btrfs_add_delayed_tree_ref(trans, generic_ref, NULL);
1431 else
1432 ret = btrfs_add_delayed_data_ref(trans, generic_ref, 0);
1433
1434 btrfs_ref_tree_mod(fs_info, generic_ref);
1435
1436 return ret;
1437 }
1438
1439 /*
1440 * Insert backreference for a given extent.
1441 *
1442 * The counterpart is in __btrfs_free_extent(), with examples and more details
1443 * how it works.
1444 *
1445 * @trans: Handle of transaction
1446 *
1447 * @node: The delayed ref node used to get the bytenr/length for
1448 * extent whose references are incremented.
1449 *
1450 * @extent_op Pointer to a structure, holding information necessary when
1451 * updating a tree block's flags
1452 *
1453 */
1454 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1455 struct btrfs_delayed_ref_node *node,
1456 struct btrfs_delayed_extent_op *extent_op)
1457 {
1458 struct btrfs_path *path;
1459 struct extent_buffer *leaf;
1460 struct btrfs_extent_item *item;
1461 struct btrfs_key key;
1462 u64 bytenr = node->bytenr;
1463 u64 num_bytes = node->num_bytes;
1464 u64 owner = btrfs_delayed_ref_owner(node);
1465 u64 offset = btrfs_delayed_ref_offset(node);
1466 u64 refs;
1467 int refs_to_add = node->ref_mod;
1468 int ret;
1469
1470 path = btrfs_alloc_path();
1471 if (!path)
1472 return -ENOMEM;
1473
1474 /* this will setup the path even if it fails to insert the back ref */
1475 ret = insert_inline_extent_backref(trans, path, bytenr, num_bytes,
1476 node->parent, node->ref_root, owner,
1477 offset, refs_to_add, extent_op);
1478 if ((ret < 0 && ret != -EAGAIN) || !ret)
1479 goto out;
1480
1481 /*
1482 * Ok we had -EAGAIN which means we didn't have space to insert and
1483 * inline extent ref, so just update the reference count and add a
1484 * normal backref.
1485 */
1486 leaf = path->nodes[0];
1487 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1488 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1489 refs = btrfs_extent_refs(leaf, item);
1490 btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1491 if (extent_op)
1492 __run_delayed_extent_op(extent_op, leaf, item);
1493
1494 btrfs_mark_buffer_dirty(trans, leaf);
1495 btrfs_release_path(path);
1496
1497 /* now insert the actual backref */
1498 if (owner < BTRFS_FIRST_FREE_OBJECTID)
1499 ret = insert_tree_block_ref(trans, path, node, bytenr);
1500 else
1501 ret = insert_extent_data_ref(trans, path, node, bytenr);
1502
1503 if (ret)
1504 btrfs_abort_transaction(trans, ret);
1505 out:
1506 btrfs_free_path(path);
1507 return ret;
1508 }
1509
1510 static void free_head_ref_squota_rsv(struct btrfs_fs_info *fs_info,
1511 struct btrfs_delayed_ref_head *href)
1512 {
1513 u64 root = href->owning_root;
1514
1515 /*
1516 * Don't check must_insert_reserved, as this is called from contexts
1517 * where it has already been unset.
1518 */
1519 if (btrfs_qgroup_mode(fs_info) != BTRFS_QGROUP_MODE_SIMPLE ||
1520 !href->is_data || !is_fstree(root))
1521 return;
1522
1523 btrfs_qgroup_free_refroot(fs_info, root, href->reserved_bytes,
1524 BTRFS_QGROUP_RSV_DATA);
1525 }
1526
1527 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1528 struct btrfs_delayed_ref_head *href,
1529 struct btrfs_delayed_ref_node *node,
1530 struct btrfs_delayed_extent_op *extent_op,
1531 bool insert_reserved)
1532 {
1533 int ret = 0;
1534 u64 parent = 0;
1535 u64 flags = 0;
1536
1537 trace_run_delayed_data_ref(trans->fs_info, node);
1538
1539 if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1540 parent = node->parent;
1541
1542 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1543 struct btrfs_key key;
1544 struct btrfs_squota_delta delta = {
1545 .root = href->owning_root,
1546 .num_bytes = node->num_bytes,
1547 .is_data = true,
1548 .is_inc = true,
1549 .generation = trans->transid,
1550 };
1551 u64 owner = btrfs_delayed_ref_owner(node);
1552 u64 offset = btrfs_delayed_ref_offset(node);
1553
1554 if (extent_op)
1555 flags |= extent_op->flags_to_set;
1556
1557 key.objectid = node->bytenr;
1558 key.type = BTRFS_EXTENT_ITEM_KEY;
1559 key.offset = node->num_bytes;
1560
1561 ret = alloc_reserved_file_extent(trans, parent, node->ref_root,
1562 flags, owner, offset, &key,
1563 node->ref_mod,
1564 href->owning_root);
1565 free_head_ref_squota_rsv(trans->fs_info, href);
1566 if (!ret)
1567 ret = btrfs_record_squota_delta(trans->fs_info, &delta);
1568 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1569 ret = __btrfs_inc_extent_ref(trans, node, extent_op);
1570 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1571 ret = __btrfs_free_extent(trans, href, node, extent_op);
1572 } else {
1573 BUG();
1574 }
1575 return ret;
1576 }
1577
1578 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1579 struct extent_buffer *leaf,
1580 struct btrfs_extent_item *ei)
1581 {
1582 u64 flags = btrfs_extent_flags(leaf, ei);
1583 if (extent_op->update_flags) {
1584 flags |= extent_op->flags_to_set;
1585 btrfs_set_extent_flags(leaf, ei, flags);
1586 }
1587
1588 if (extent_op->update_key) {
1589 struct btrfs_tree_block_info *bi;
1590 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1591 bi = (struct btrfs_tree_block_info *)(ei + 1);
1592 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1593 }
1594 }
1595
1596 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1597 struct btrfs_delayed_ref_head *head,
1598 struct btrfs_delayed_extent_op *extent_op)
1599 {
1600 struct btrfs_fs_info *fs_info = trans->fs_info;
1601 struct btrfs_root *root;
1602 struct btrfs_key key;
1603 struct btrfs_path *path;
1604 struct btrfs_extent_item *ei;
1605 struct extent_buffer *leaf;
1606 u32 item_size;
1607 int ret;
1608 int metadata = 1;
1609
1610 if (TRANS_ABORTED(trans))
1611 return 0;
1612
1613 if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA))
1614 metadata = 0;
1615
1616 path = btrfs_alloc_path();
1617 if (!path)
1618 return -ENOMEM;
1619
1620 key.objectid = head->bytenr;
1621
1622 if (metadata) {
1623 key.type = BTRFS_METADATA_ITEM_KEY;
1624 key.offset = head->level;
1625 } else {
1626 key.type = BTRFS_EXTENT_ITEM_KEY;
1627 key.offset = head->num_bytes;
1628 }
1629
1630 root = btrfs_extent_root(fs_info, key.objectid);
1631 again:
1632 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
1633 if (ret < 0) {
1634 goto out;
1635 } else if (ret > 0) {
1636 if (metadata) {
1637 if (path->slots[0] > 0) {
1638 path->slots[0]--;
1639 btrfs_item_key_to_cpu(path->nodes[0], &key,
1640 path->slots[0]);
1641 if (key.objectid == head->bytenr &&
1642 key.type == BTRFS_EXTENT_ITEM_KEY &&
1643 key.offset == head->num_bytes)
1644 ret = 0;
1645 }
1646 if (ret > 0) {
1647 btrfs_release_path(path);
1648 metadata = 0;
1649
1650 key.objectid = head->bytenr;
1651 key.offset = head->num_bytes;
1652 key.type = BTRFS_EXTENT_ITEM_KEY;
1653 goto again;
1654 }
1655 } else {
1656 ret = -EUCLEAN;
1657 btrfs_err(fs_info,
1658 "missing extent item for extent %llu num_bytes %llu level %d",
1659 head->bytenr, head->num_bytes, head->level);
1660 goto out;
1661 }
1662 }
1663
1664 leaf = path->nodes[0];
1665 item_size = btrfs_item_size(leaf, path->slots[0]);
1666
1667 if (unlikely(item_size < sizeof(*ei))) {
1668 ret = -EUCLEAN;
1669 btrfs_err(fs_info,
1670 "unexpected extent item size, has %u expect >= %zu",
1671 item_size, sizeof(*ei));
1672 btrfs_abort_transaction(trans, ret);
1673 goto out;
1674 }
1675
1676 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1677 __run_delayed_extent_op(extent_op, leaf, ei);
1678
1679 btrfs_mark_buffer_dirty(trans, leaf);
1680 out:
1681 btrfs_free_path(path);
1682 return ret;
1683 }
1684
1685 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1686 struct btrfs_delayed_ref_head *href,
1687 struct btrfs_delayed_ref_node *node,
1688 struct btrfs_delayed_extent_op *extent_op,
1689 bool insert_reserved)
1690 {
1691 int ret = 0;
1692 struct btrfs_fs_info *fs_info = trans->fs_info;
1693 u64 parent = 0;
1694 u64 ref_root = 0;
1695
1696 trace_run_delayed_tree_ref(trans->fs_info, node);
1697
1698 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1699 parent = node->parent;
1700 ref_root = node->ref_root;
1701
1702 if (unlikely(node->ref_mod != 1)) {
1703 btrfs_err(trans->fs_info,
1704 "btree block %llu has %d references rather than 1: action %d ref_root %llu parent %llu",
1705 node->bytenr, node->ref_mod, node->action, ref_root,
1706 parent);
1707 return -EUCLEAN;
1708 }
1709 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1710 struct btrfs_squota_delta delta = {
1711 .root = href->owning_root,
1712 .num_bytes = fs_info->nodesize,
1713 .is_data = false,
1714 .is_inc = true,
1715 .generation = trans->transid,
1716 };
1717
1718 ret = alloc_reserved_tree_block(trans, node, extent_op);
1719 if (!ret)
1720 btrfs_record_squota_delta(fs_info, &delta);
1721 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1722 ret = __btrfs_inc_extent_ref(trans, node, extent_op);
1723 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1724 ret = __btrfs_free_extent(trans, href, node, extent_op);
1725 } else {
1726 BUG();
1727 }
1728 return ret;
1729 }
1730
1731 /* helper function to actually process a single delayed ref entry */
1732 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1733 struct btrfs_delayed_ref_head *href,
1734 struct btrfs_delayed_ref_node *node,
1735 struct btrfs_delayed_extent_op *extent_op,
1736 bool insert_reserved)
1737 {
1738 int ret = 0;
1739
1740 if (TRANS_ABORTED(trans)) {
1741 if (insert_reserved) {
1742 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1743 free_head_ref_squota_rsv(trans->fs_info, href);
1744 }
1745 return 0;
1746 }
1747
1748 if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1749 node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1750 ret = run_delayed_tree_ref(trans, href, node, extent_op,
1751 insert_reserved);
1752 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1753 node->type == BTRFS_SHARED_DATA_REF_KEY)
1754 ret = run_delayed_data_ref(trans, href, node, extent_op,
1755 insert_reserved);
1756 else if (node->type == BTRFS_EXTENT_OWNER_REF_KEY)
1757 ret = 0;
1758 else
1759 BUG();
1760 if (ret && insert_reserved)
1761 btrfs_pin_extent(trans, node->bytenr, node->num_bytes, 1);
1762 if (ret < 0)
1763 btrfs_err(trans->fs_info,
1764 "failed to run delayed ref for logical %llu num_bytes %llu type %u action %u ref_mod %d: %d",
1765 node->bytenr, node->num_bytes, node->type,
1766 node->action, node->ref_mod, ret);
1767 return ret;
1768 }
1769
1770 static inline struct btrfs_delayed_ref_node *
1771 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1772 {
1773 struct btrfs_delayed_ref_node *ref;
1774
1775 if (RB_EMPTY_ROOT(&head->ref_tree.rb_root))
1776 return NULL;
1777
1778 /*
1779 * Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
1780 * This is to prevent a ref count from going down to zero, which deletes
1781 * the extent item from the extent tree, when there still are references
1782 * to add, which would fail because they would not find the extent item.
1783 */
1784 if (!list_empty(&head->ref_add_list))
1785 return list_first_entry(&head->ref_add_list,
1786 struct btrfs_delayed_ref_node, add_list);
1787
1788 ref = rb_entry(rb_first_cached(&head->ref_tree),
1789 struct btrfs_delayed_ref_node, ref_node);
1790 ASSERT(list_empty(&ref->add_list));
1791 return ref;
1792 }
1793
1794 static void unselect_delayed_ref_head(struct btrfs_delayed_ref_root *delayed_refs,
1795 struct btrfs_delayed_ref_head *head)
1796 {
1797 spin_lock(&delayed_refs->lock);
1798 head->processing = false;
1799 delayed_refs->num_heads_ready++;
1800 spin_unlock(&delayed_refs->lock);
1801 btrfs_delayed_ref_unlock(head);
1802 }
1803
1804 static struct btrfs_delayed_extent_op *cleanup_extent_op(
1805 struct btrfs_delayed_ref_head *head)
1806 {
1807 struct btrfs_delayed_extent_op *extent_op = head->extent_op;
1808
1809 if (!extent_op)
1810 return NULL;
1811
1812 if (head->must_insert_reserved) {
1813 head->extent_op = NULL;
1814 btrfs_free_delayed_extent_op(extent_op);
1815 return NULL;
1816 }
1817 return extent_op;
1818 }
1819
1820 static int run_and_cleanup_extent_op(struct btrfs_trans_handle *trans,
1821 struct btrfs_delayed_ref_head *head)
1822 {
1823 struct btrfs_delayed_extent_op *extent_op;
1824 int ret;
1825
1826 extent_op = cleanup_extent_op(head);
1827 if (!extent_op)
1828 return 0;
1829 head->extent_op = NULL;
1830 spin_unlock(&head->lock);
1831 ret = run_delayed_extent_op(trans, head, extent_op);
1832 btrfs_free_delayed_extent_op(extent_op);
1833 return ret ? ret : 1;
1834 }
1835
1836 u64 btrfs_cleanup_ref_head_accounting(struct btrfs_fs_info *fs_info,
1837 struct btrfs_delayed_ref_root *delayed_refs,
1838 struct btrfs_delayed_ref_head *head)
1839 {
1840 u64 ret = 0;
1841
1842 /*
1843 * We had csum deletions accounted for in our delayed refs rsv, we need
1844 * to drop the csum leaves for this update from our delayed_refs_rsv.
1845 */
1846 if (head->total_ref_mod < 0 && head->is_data) {
1847 int nr_csums;
1848
1849 spin_lock(&delayed_refs->lock);
1850 delayed_refs->pending_csums -= head->num_bytes;
1851 spin_unlock(&delayed_refs->lock);
1852 nr_csums = btrfs_csum_bytes_to_leaves(fs_info, head->num_bytes);
1853
1854 btrfs_delayed_refs_rsv_release(fs_info, 0, nr_csums);
1855
1856 ret = btrfs_calc_delayed_ref_csum_bytes(fs_info, nr_csums);
1857 }
1858 /* must_insert_reserved can be set only if we didn't run the head ref. */
1859 if (head->must_insert_reserved)
1860 free_head_ref_squota_rsv(fs_info, head);
1861
1862 return ret;
1863 }
1864
1865 static int cleanup_ref_head(struct btrfs_trans_handle *trans,
1866 struct btrfs_delayed_ref_head *head,
1867 u64 *bytes_released)
1868 {
1869
1870 struct btrfs_fs_info *fs_info = trans->fs_info;
1871 struct btrfs_delayed_ref_root *delayed_refs;
1872 int ret;
1873
1874 delayed_refs = &trans->transaction->delayed_refs;
1875
1876 ret = run_and_cleanup_extent_op(trans, head);
1877 if (ret < 0) {
1878 unselect_delayed_ref_head(delayed_refs, head);
1879 btrfs_debug(fs_info, "run_delayed_extent_op returned %d", ret);
1880 return ret;
1881 } else if (ret) {
1882 return ret;
1883 }
1884
1885 /*
1886 * Need to drop our head ref lock and re-acquire the delayed ref lock
1887 * and then re-check to make sure nobody got added.
1888 */
1889 spin_unlock(&head->lock);
1890 spin_lock(&delayed_refs->lock);
1891 spin_lock(&head->lock);
1892 if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root) || head->extent_op) {
1893 spin_unlock(&head->lock);
1894 spin_unlock(&delayed_refs->lock);
1895 return 1;
1896 }
1897 btrfs_delete_ref_head(delayed_refs, head);
1898 spin_unlock(&head->lock);
1899 spin_unlock(&delayed_refs->lock);
1900
1901 if (head->must_insert_reserved) {
1902 btrfs_pin_extent(trans, head->bytenr, head->num_bytes, 1);
1903 if (head->is_data) {
1904 struct btrfs_root *csum_root;
1905
1906 csum_root = btrfs_csum_root(fs_info, head->bytenr);
1907 ret = btrfs_del_csums(trans, csum_root, head->bytenr,
1908 head->num_bytes);
1909 }
1910 }
1911
1912 *bytes_released += btrfs_cleanup_ref_head_accounting(fs_info, delayed_refs, head);
1913
1914 trace_run_delayed_ref_head(fs_info, head, 0);
1915 btrfs_delayed_ref_unlock(head);
1916 btrfs_put_delayed_ref_head(head);
1917 return ret;
1918 }
1919
1920 static struct btrfs_delayed_ref_head *btrfs_obtain_ref_head(
1921 struct btrfs_trans_handle *trans)
1922 {
1923 struct btrfs_delayed_ref_root *delayed_refs =
1924 &trans->transaction->delayed_refs;
1925 struct btrfs_delayed_ref_head *head = NULL;
1926 int ret;
1927
1928 spin_lock(&delayed_refs->lock);
1929 head = btrfs_select_ref_head(delayed_refs);
1930 if (!head) {
1931 spin_unlock(&delayed_refs->lock);
1932 return head;
1933 }
1934
1935 /*
1936 * Grab the lock that says we are going to process all the refs for
1937 * this head
1938 */
1939 ret = btrfs_delayed_ref_lock(delayed_refs, head);
1940 spin_unlock(&delayed_refs->lock);
1941
1942 /*
1943 * We may have dropped the spin lock to get the head mutex lock, and
1944 * that might have given someone else time to free the head. If that's
1945 * true, it has been removed from our list and we can move on.
1946 */
1947 if (ret == -EAGAIN)
1948 head = ERR_PTR(-EAGAIN);
1949
1950 return head;
1951 }
1952
1953 static int btrfs_run_delayed_refs_for_head(struct btrfs_trans_handle *trans,
1954 struct btrfs_delayed_ref_head *locked_ref,
1955 u64 *bytes_released)
1956 {
1957 struct btrfs_fs_info *fs_info = trans->fs_info;
1958 struct btrfs_delayed_ref_root *delayed_refs;
1959 struct btrfs_delayed_extent_op *extent_op;
1960 struct btrfs_delayed_ref_node *ref;
1961 bool must_insert_reserved;
1962 int ret;
1963
1964 delayed_refs = &trans->transaction->delayed_refs;
1965
1966 lockdep_assert_held(&locked_ref->mutex);
1967 lockdep_assert_held(&locked_ref->lock);
1968
1969 while ((ref = select_delayed_ref(locked_ref))) {
1970 if (ref->seq &&
1971 btrfs_check_delayed_seq(fs_info, ref->seq)) {
1972 spin_unlock(&locked_ref->lock);
1973 unselect_delayed_ref_head(delayed_refs, locked_ref);
1974 return -EAGAIN;
1975 }
1976
1977 rb_erase_cached(&ref->ref_node, &locked_ref->ref_tree);
1978 RB_CLEAR_NODE(&ref->ref_node);
1979 if (!list_empty(&ref->add_list))
1980 list_del(&ref->add_list);
1981 /*
1982 * When we play the delayed ref, also correct the ref_mod on
1983 * head
1984 */
1985 switch (ref->action) {
1986 case BTRFS_ADD_DELAYED_REF:
1987 case BTRFS_ADD_DELAYED_EXTENT:
1988 locked_ref->ref_mod -= ref->ref_mod;
1989 break;
1990 case BTRFS_DROP_DELAYED_REF:
1991 locked_ref->ref_mod += ref->ref_mod;
1992 break;
1993 default:
1994 WARN_ON(1);
1995 }
1996 atomic_dec(&delayed_refs->num_entries);
1997
1998 /*
1999 * Record the must_insert_reserved flag before we drop the
2000 * spin lock.
2001 */
2002 must_insert_reserved = locked_ref->must_insert_reserved;
2003 /*
2004 * Unsetting this on the head ref relinquishes ownership of
2005 * the rsv_bytes, so it is critical that every possible code
2006 * path from here forward frees all reserves including qgroup
2007 * reserve.
2008 */
2009 locked_ref->must_insert_reserved = false;
2010
2011 extent_op = locked_ref->extent_op;
2012 locked_ref->extent_op = NULL;
2013 spin_unlock(&locked_ref->lock);
2014
2015 ret = run_one_delayed_ref(trans, locked_ref, ref, extent_op,
2016 must_insert_reserved);
2017 btrfs_delayed_refs_rsv_release(fs_info, 1, 0);
2018 *bytes_released += btrfs_calc_delayed_ref_bytes(fs_info, 1);
2019
2020 btrfs_free_delayed_extent_op(extent_op);
2021 if (ret) {
2022 unselect_delayed_ref_head(delayed_refs, locked_ref);
2023 btrfs_put_delayed_ref(ref);
2024 return ret;
2025 }
2026
2027 btrfs_put_delayed_ref(ref);
2028 cond_resched();
2029
2030 spin_lock(&locked_ref->lock);
2031 btrfs_merge_delayed_refs(fs_info, delayed_refs, locked_ref);
2032 }
2033
2034 return 0;
2035 }
2036
2037 /*
2038 * Returns 0 on success or if called with an already aborted transaction.
2039 * Returns -ENOMEM or -EIO on failure and will abort the transaction.
2040 */
2041 static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2042 u64 min_bytes)
2043 {
2044 struct btrfs_fs_info *fs_info = trans->fs_info;
2045 struct btrfs_delayed_ref_root *delayed_refs;
2046 struct btrfs_delayed_ref_head *locked_ref = NULL;
2047 int ret;
2048 unsigned long count = 0;
2049 unsigned long max_count = 0;
2050 u64 bytes_processed = 0;
2051
2052 delayed_refs = &trans->transaction->delayed_refs;
2053 if (min_bytes == 0) {
2054 max_count = delayed_refs->num_heads_ready;
2055 min_bytes = U64_MAX;
2056 }
2057
2058 do {
2059 if (!locked_ref) {
2060 locked_ref = btrfs_obtain_ref_head(trans);
2061 if (IS_ERR_OR_NULL(locked_ref)) {
2062 if (PTR_ERR(locked_ref) == -EAGAIN) {
2063 continue;
2064 } else {
2065 break;
2066 }
2067 }
2068 count++;
2069 }
2070 /*
2071 * We need to try and merge add/drops of the same ref since we
2072 * can run into issues with relocate dropping the implicit ref
2073 * and then it being added back again before the drop can
2074 * finish. If we merged anything we need to re-loop so we can
2075 * get a good ref.
2076 * Or we can get node references of the same type that weren't
2077 * merged when created due to bumps in the tree mod seq, and
2078 * we need to merge them to prevent adding an inline extent
2079 * backref before dropping it (triggering a BUG_ON at
2080 * insert_inline_extent_backref()).
2081 */
2082 spin_lock(&locked_ref->lock);
2083 btrfs_merge_delayed_refs(fs_info, delayed_refs, locked_ref);
2084
2085 ret = btrfs_run_delayed_refs_for_head(trans, locked_ref, &bytes_processed);
2086 if (ret < 0 && ret != -EAGAIN) {
2087 /*
2088 * Error, btrfs_run_delayed_refs_for_head already
2089 * unlocked everything so just bail out
2090 */
2091 return ret;
2092 } else if (!ret) {
2093 /*
2094 * Success, perform the usual cleanup of a processed
2095 * head
2096 */
2097 ret = cleanup_ref_head(trans, locked_ref, &bytes_processed);
2098 if (ret > 0 ) {
2099 /* We dropped our lock, we need to loop. */
2100 ret = 0;
2101 continue;
2102 } else if (ret) {
2103 return ret;
2104 }
2105 }
2106
2107 /*
2108 * Either success case or btrfs_run_delayed_refs_for_head
2109 * returned -EAGAIN, meaning we need to select another head
2110 */
2111
2112 locked_ref = NULL;
2113 cond_resched();
2114 } while ((min_bytes != U64_MAX && bytes_processed < min_bytes) ||
2115 (max_count > 0 && count < max_count) ||
2116 locked_ref);
2117
2118 return 0;
2119 }
2120
2121 #ifdef SCRAMBLE_DELAYED_REFS
2122 /*
2123 * Normally delayed refs get processed in ascending bytenr order. This
2124 * correlates in most cases to the order added. To expose dependencies on this
2125 * order, we start to process the tree in the middle instead of the beginning
2126 */
2127 static u64 find_middle(struct rb_root *root)
2128 {
2129 struct rb_node *n = root->rb_node;
2130 struct btrfs_delayed_ref_node *entry;
2131 int alt = 1;
2132 u64 middle;
2133 u64 first = 0, last = 0;
2134
2135 n = rb_first(root);
2136 if (n) {
2137 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2138 first = entry->bytenr;
2139 }
2140 n = rb_last(root);
2141 if (n) {
2142 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2143 last = entry->bytenr;
2144 }
2145 n = root->rb_node;
2146
2147 while (n) {
2148 entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
2149 WARN_ON(!entry->in_tree);
2150
2151 middle = entry->bytenr;
2152
2153 if (alt)
2154 n = n->rb_left;
2155 else
2156 n = n->rb_right;
2157
2158 alt = 1 - alt;
2159 }
2160 return middle;
2161 }
2162 #endif
2163
2164 /*
2165 * Start processing the delayed reference count updates and extent insertions
2166 * we have queued up so far.
2167 *
2168 * @trans: Transaction handle.
2169 * @min_bytes: How many bytes of delayed references to process. After this
2170 * many bytes we stop processing delayed references if there are
2171 * any more. If 0 it means to run all existing delayed references,
2172 * but not new ones added after running all existing ones.
2173 * Use (u64)-1 (U64_MAX) to run all existing delayed references
2174 * plus any new ones that are added.
2175 *
2176 * Returns 0 on success or if called with an aborted transaction
2177 * Returns <0 on error and aborts the transaction
2178 */
2179 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans, u64 min_bytes)
2180 {
2181 struct btrfs_fs_info *fs_info = trans->fs_info;
2182 struct btrfs_delayed_ref_root *delayed_refs;
2183 int ret;
2184
2185 /* We'll clean this up in btrfs_cleanup_transaction */
2186 if (TRANS_ABORTED(trans))
2187 return 0;
2188
2189 if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
2190 return 0;
2191
2192 delayed_refs = &trans->transaction->delayed_refs;
2193 again:
2194 #ifdef SCRAMBLE_DELAYED_REFS
2195 delayed_refs->run_delayed_start = find_middle(&delayed_refs->root);
2196 #endif
2197 ret = __btrfs_run_delayed_refs(trans, min_bytes);
2198 if (ret < 0) {
2199 btrfs_abort_transaction(trans, ret);
2200 return ret;
2201 }
2202
2203 if (min_bytes == U64_MAX) {
2204 btrfs_create_pending_block_groups(trans);
2205
2206 spin_lock(&delayed_refs->lock);
2207 if (RB_EMPTY_ROOT(&delayed_refs->href_root.rb_root)) {
2208 spin_unlock(&delayed_refs->lock);
2209 return 0;
2210 }
2211 spin_unlock(&delayed_refs->lock);
2212
2213 cond_resched();
2214 goto again;
2215 }
2216
2217 return 0;
2218 }
2219
2220 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2221 struct extent_buffer *eb, u64 flags)
2222 {
2223 struct btrfs_delayed_extent_op *extent_op;
2224 int ret;
2225
2226 extent_op = btrfs_alloc_delayed_extent_op();
2227 if (!extent_op)
2228 return -ENOMEM;
2229
2230 extent_op->flags_to_set = flags;
2231 extent_op->update_flags = true;
2232 extent_op->update_key = false;
2233
2234 ret = btrfs_add_delayed_extent_op(trans, eb->start, eb->len,
2235 btrfs_header_level(eb), extent_op);
2236 if (ret)
2237 btrfs_free_delayed_extent_op(extent_op);
2238 return ret;
2239 }
2240
2241 static noinline int check_delayed_ref(struct btrfs_root *root,
2242 struct btrfs_path *path,
2243 u64 objectid, u64 offset, u64 bytenr)
2244 {
2245 struct btrfs_delayed_ref_head *head;
2246 struct btrfs_delayed_ref_node *ref;
2247 struct btrfs_delayed_ref_root *delayed_refs;
2248 struct btrfs_transaction *cur_trans;
2249 struct rb_node *node;
2250 int ret = 0;
2251
2252 spin_lock(&root->fs_info->trans_lock);
2253 cur_trans = root->fs_info->running_transaction;
2254 if (cur_trans)
2255 refcount_inc(&cur_trans->use_count);
2256 spin_unlock(&root->fs_info->trans_lock);
2257 if (!cur_trans)
2258 return 0;
2259
2260 delayed_refs = &cur_trans->delayed_refs;
2261 spin_lock(&delayed_refs->lock);
2262 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
2263 if (!head) {
2264 spin_unlock(&delayed_refs->lock);
2265 btrfs_put_transaction(cur_trans);
2266 return 0;
2267 }
2268
2269 if (!mutex_trylock(&head->mutex)) {
2270 if (path->nowait) {
2271 spin_unlock(&delayed_refs->lock);
2272 btrfs_put_transaction(cur_trans);
2273 return -EAGAIN;
2274 }
2275
2276 refcount_inc(&head->refs);
2277 spin_unlock(&delayed_refs->lock);
2278
2279 btrfs_release_path(path);
2280
2281 /*
2282 * Mutex was contended, block until it's released and let
2283 * caller try again
2284 */
2285 mutex_lock(&head->mutex);
2286 mutex_unlock(&head->mutex);
2287 btrfs_put_delayed_ref_head(head);
2288 btrfs_put_transaction(cur_trans);
2289 return -EAGAIN;
2290 }
2291 spin_unlock(&delayed_refs->lock);
2292
2293 spin_lock(&head->lock);
2294 /*
2295 * XXX: We should replace this with a proper search function in the
2296 * future.
2297 */
2298 for (node = rb_first_cached(&head->ref_tree); node;
2299 node = rb_next(node)) {
2300 u64 ref_owner;
2301 u64 ref_offset;
2302
2303 ref = rb_entry(node, struct btrfs_delayed_ref_node, ref_node);
2304 /* If it's a shared ref we know a cross reference exists */
2305 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY) {
2306 ret = 1;
2307 break;
2308 }
2309
2310 ref_owner = btrfs_delayed_ref_owner(ref);
2311 ref_offset = btrfs_delayed_ref_offset(ref);
2312
2313 /*
2314 * If our ref doesn't match the one we're currently looking at
2315 * then we have a cross reference.
2316 */
2317 if (ref->ref_root != btrfs_root_id(root) ||
2318 ref_owner != objectid || ref_offset != offset) {
2319 ret = 1;
2320 break;
2321 }
2322 }
2323 spin_unlock(&head->lock);
2324 mutex_unlock(&head->mutex);
2325 btrfs_put_transaction(cur_trans);
2326 return ret;
2327 }
2328
2329 static noinline int check_committed_ref(struct btrfs_root *root,
2330 struct btrfs_path *path,
2331 u64 objectid, u64 offset, u64 bytenr,
2332 bool strict)
2333 {
2334 struct btrfs_fs_info *fs_info = root->fs_info;
2335 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, bytenr);
2336 struct extent_buffer *leaf;
2337 struct btrfs_extent_data_ref *ref;
2338 struct btrfs_extent_inline_ref *iref;
2339 struct btrfs_extent_item *ei;
2340 struct btrfs_key key;
2341 u32 item_size;
2342 u32 expected_size;
2343 int type;
2344 int ret;
2345
2346 key.objectid = bytenr;
2347 key.offset = (u64)-1;
2348 key.type = BTRFS_EXTENT_ITEM_KEY;
2349
2350 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2351 if (ret < 0)
2352 goto out;
2353 if (ret == 0) {
2354 /*
2355 * Key with offset -1 found, there would have to exist an extent
2356 * item with such offset, but this is out of the valid range.
2357 */
2358 ret = -EUCLEAN;
2359 goto out;
2360 }
2361
2362 ret = -ENOENT;
2363 if (path->slots[0] == 0)
2364 goto out;
2365
2366 path->slots[0]--;
2367 leaf = path->nodes[0];
2368 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2369
2370 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2371 goto out;
2372
2373 ret = 1;
2374 item_size = btrfs_item_size(leaf, path->slots[0]);
2375 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2376 expected_size = sizeof(*ei) + btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY);
2377
2378 /* No inline refs; we need to bail before checking for owner ref. */
2379 if (item_size == sizeof(*ei))
2380 goto out;
2381
2382 /* Check for an owner ref; skip over it to the real inline refs. */
2383 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2384 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
2385 if (btrfs_fs_incompat(fs_info, SIMPLE_QUOTA) && type == BTRFS_EXTENT_OWNER_REF_KEY) {
2386 expected_size += btrfs_extent_inline_ref_size(BTRFS_EXTENT_OWNER_REF_KEY);
2387 iref = (struct btrfs_extent_inline_ref *)(iref + 1);
2388 }
2389
2390 /* If extent item has more than 1 inline ref then it's shared */
2391 if (item_size != expected_size)
2392 goto out;
2393
2394 /*
2395 * If extent created before last snapshot => it's shared unless the
2396 * snapshot has been deleted. Use the heuristic if strict is false.
2397 */
2398 if (!strict &&
2399 (btrfs_extent_generation(leaf, ei) <=
2400 btrfs_root_last_snapshot(&root->root_item)))
2401 goto out;
2402
2403 /* If this extent has SHARED_DATA_REF then it's shared */
2404 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
2405 if (type != BTRFS_EXTENT_DATA_REF_KEY)
2406 goto out;
2407
2408 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2409 if (btrfs_extent_refs(leaf, ei) !=
2410 btrfs_extent_data_ref_count(leaf, ref) ||
2411 btrfs_extent_data_ref_root(leaf, ref) != btrfs_root_id(root) ||
2412 btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2413 btrfs_extent_data_ref_offset(leaf, ref) != offset)
2414 goto out;
2415
2416 ret = 0;
2417 out:
2418 return ret;
2419 }
2420
2421 int btrfs_cross_ref_exist(struct btrfs_root *root, u64 objectid, u64 offset,
2422 u64 bytenr, bool strict, struct btrfs_path *path)
2423 {
2424 int ret;
2425
2426 do {
2427 ret = check_committed_ref(root, path, objectid,
2428 offset, bytenr, strict);
2429 if (ret && ret != -ENOENT)
2430 goto out;
2431
2432 ret = check_delayed_ref(root, path, objectid, offset, bytenr);
2433 } while (ret == -EAGAIN);
2434
2435 out:
2436 btrfs_release_path(path);
2437 if (btrfs_is_data_reloc_root(root))
2438 WARN_ON(ret > 0);
2439 return ret;
2440 }
2441
2442 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2443 struct btrfs_root *root,
2444 struct extent_buffer *buf,
2445 int full_backref, int inc)
2446 {
2447 struct btrfs_fs_info *fs_info = root->fs_info;
2448 u64 parent;
2449 u64 ref_root;
2450 u32 nritems;
2451 struct btrfs_key key;
2452 struct btrfs_file_extent_item *fi;
2453 bool for_reloc = btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC);
2454 int i;
2455 int action;
2456 int level;
2457 int ret = 0;
2458
2459 if (btrfs_is_testing(fs_info))
2460 return 0;
2461
2462 ref_root = btrfs_header_owner(buf);
2463 nritems = btrfs_header_nritems(buf);
2464 level = btrfs_header_level(buf);
2465
2466 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state) && level == 0)
2467 return 0;
2468
2469 if (full_backref)
2470 parent = buf->start;
2471 else
2472 parent = 0;
2473 if (inc)
2474 action = BTRFS_ADD_DELAYED_REF;
2475 else
2476 action = BTRFS_DROP_DELAYED_REF;
2477
2478 for (i = 0; i < nritems; i++) {
2479 struct btrfs_ref ref = {
2480 .action = action,
2481 .parent = parent,
2482 .ref_root = ref_root,
2483 };
2484
2485 if (level == 0) {
2486 btrfs_item_key_to_cpu(buf, &key, i);
2487 if (key.type != BTRFS_EXTENT_DATA_KEY)
2488 continue;
2489 fi = btrfs_item_ptr(buf, i,
2490 struct btrfs_file_extent_item);
2491 if (btrfs_file_extent_type(buf, fi) ==
2492 BTRFS_FILE_EXTENT_INLINE)
2493 continue;
2494 ref.bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2495 if (ref.bytenr == 0)
2496 continue;
2497
2498 ref.num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2499 ref.owning_root = ref_root;
2500
2501 key.offset -= btrfs_file_extent_offset(buf, fi);
2502 btrfs_init_data_ref(&ref, key.objectid, key.offset,
2503 btrfs_root_id(root), for_reloc);
2504 if (inc)
2505 ret = btrfs_inc_extent_ref(trans, &ref);
2506 else
2507 ret = btrfs_free_extent(trans, &ref);
2508 if (ret)
2509 goto fail;
2510 } else {
2511 /* We don't know the owning_root, leave as 0. */
2512 ref.bytenr = btrfs_node_blockptr(buf, i);
2513 ref.num_bytes = fs_info->nodesize;
2514
2515 btrfs_init_tree_ref(&ref, level - 1,
2516 btrfs_root_id(root), for_reloc);
2517 if (inc)
2518 ret = btrfs_inc_extent_ref(trans, &ref);
2519 else
2520 ret = btrfs_free_extent(trans, &ref);
2521 if (ret)
2522 goto fail;
2523 }
2524 }
2525 return 0;
2526 fail:
2527 return ret;
2528 }
2529
2530 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2531 struct extent_buffer *buf, int full_backref)
2532 {
2533 return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2534 }
2535
2536 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2537 struct extent_buffer *buf, int full_backref)
2538 {
2539 return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2540 }
2541
2542 static u64 get_alloc_profile_by_root(struct btrfs_root *root, int data)
2543 {
2544 struct btrfs_fs_info *fs_info = root->fs_info;
2545 u64 flags;
2546 u64 ret;
2547
2548 if (data)
2549 flags = BTRFS_BLOCK_GROUP_DATA;
2550 else if (root == fs_info->chunk_root)
2551 flags = BTRFS_BLOCK_GROUP_SYSTEM;
2552 else
2553 flags = BTRFS_BLOCK_GROUP_METADATA;
2554
2555 ret = btrfs_get_alloc_profile(fs_info, flags);
2556 return ret;
2557 }
2558
2559 static u64 first_logical_byte(struct btrfs_fs_info *fs_info)
2560 {
2561 struct rb_node *leftmost;
2562 u64 bytenr = 0;
2563
2564 read_lock(&fs_info->block_group_cache_lock);
2565 /* Get the block group with the lowest logical start address. */
2566 leftmost = rb_first_cached(&fs_info->block_group_cache_tree);
2567 if (leftmost) {
2568 struct btrfs_block_group *bg;
2569
2570 bg = rb_entry(leftmost, struct btrfs_block_group, cache_node);
2571 bytenr = bg->start;
2572 }
2573 read_unlock(&fs_info->block_group_cache_lock);
2574
2575 return bytenr;
2576 }
2577
2578 static int pin_down_extent(struct btrfs_trans_handle *trans,
2579 struct btrfs_block_group *cache,
2580 u64 bytenr, u64 num_bytes, int reserved)
2581 {
2582 struct btrfs_fs_info *fs_info = cache->fs_info;
2583
2584 spin_lock(&cache->space_info->lock);
2585 spin_lock(&cache->lock);
2586 cache->pinned += num_bytes;
2587 btrfs_space_info_update_bytes_pinned(fs_info, cache->space_info,
2588 num_bytes);
2589 if (reserved) {
2590 cache->reserved -= num_bytes;
2591 cache->space_info->bytes_reserved -= num_bytes;
2592 }
2593 spin_unlock(&cache->lock);
2594 spin_unlock(&cache->space_info->lock);
2595
2596 set_extent_bit(&trans->transaction->pinned_extents, bytenr,
2597 bytenr + num_bytes - 1, EXTENT_DIRTY, NULL);
2598 return 0;
2599 }
2600
2601 int btrfs_pin_extent(struct btrfs_trans_handle *trans,
2602 u64 bytenr, u64 num_bytes, int reserved)
2603 {
2604 struct btrfs_block_group *cache;
2605
2606 cache = btrfs_lookup_block_group(trans->fs_info, bytenr);
2607 BUG_ON(!cache); /* Logic error */
2608
2609 pin_down_extent(trans, cache, bytenr, num_bytes, reserved);
2610
2611 btrfs_put_block_group(cache);
2612 return 0;
2613 }
2614
2615 int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans,
2616 const struct extent_buffer *eb)
2617 {
2618 struct btrfs_block_group *cache;
2619 int ret;
2620
2621 cache = btrfs_lookup_block_group(trans->fs_info, eb->start);
2622 if (!cache)
2623 return -EINVAL;
2624
2625 /*
2626 * Fully cache the free space first so that our pin removes the free space
2627 * from the cache.
2628 */
2629 ret = btrfs_cache_block_group(cache, true);
2630 if (ret)
2631 goto out;
2632
2633 pin_down_extent(trans, cache, eb->start, eb->len, 0);
2634
2635 /* remove us from the free space cache (if we're there at all) */
2636 ret = btrfs_remove_free_space(cache, eb->start, eb->len);
2637 out:
2638 btrfs_put_block_group(cache);
2639 return ret;
2640 }
2641
2642 static int __exclude_logged_extent(struct btrfs_fs_info *fs_info,
2643 u64 start, u64 num_bytes)
2644 {
2645 int ret;
2646 struct btrfs_block_group *block_group;
2647
2648 block_group = btrfs_lookup_block_group(fs_info, start);
2649 if (!block_group)
2650 return -EINVAL;
2651
2652 ret = btrfs_cache_block_group(block_group, true);
2653 if (ret)
2654 goto out;
2655
2656 ret = btrfs_remove_free_space(block_group, start, num_bytes);
2657 out:
2658 btrfs_put_block_group(block_group);
2659 return ret;
2660 }
2661
2662 int btrfs_exclude_logged_extents(struct extent_buffer *eb)
2663 {
2664 struct btrfs_fs_info *fs_info = eb->fs_info;
2665 struct btrfs_file_extent_item *item;
2666 struct btrfs_key key;
2667 int found_type;
2668 int i;
2669 int ret = 0;
2670
2671 if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS))
2672 return 0;
2673
2674 for (i = 0; i < btrfs_header_nritems(eb); i++) {
2675 btrfs_item_key_to_cpu(eb, &key, i);
2676 if (key.type != BTRFS_EXTENT_DATA_KEY)
2677 continue;
2678 item = btrfs_item_ptr(eb, i, struct btrfs_file_extent_item);
2679 found_type = btrfs_file_extent_type(eb, item);
2680 if (found_type == BTRFS_FILE_EXTENT_INLINE)
2681 continue;
2682 if (btrfs_file_extent_disk_bytenr(eb, item) == 0)
2683 continue;
2684 key.objectid = btrfs_file_extent_disk_bytenr(eb, item);
2685 key.offset = btrfs_file_extent_disk_num_bytes(eb, item);
2686 ret = __exclude_logged_extent(fs_info, key.objectid, key.offset);
2687 if (ret)
2688 break;
2689 }
2690
2691 return ret;
2692 }
2693
2694 static void
2695 btrfs_inc_block_group_reservations(struct btrfs_block_group *bg)
2696 {
2697 atomic_inc(&bg->reservations);
2698 }
2699
2700 /*
2701 * Returns the free cluster for the given space info and sets empty_cluster to
2702 * what it should be based on the mount options.
2703 */
2704 static struct btrfs_free_cluster *
2705 fetch_cluster_info(struct btrfs_fs_info *fs_info,
2706 struct btrfs_space_info *space_info, u64 *empty_cluster)
2707 {
2708 struct btrfs_free_cluster *ret = NULL;
2709
2710 *empty_cluster = 0;
2711 if (btrfs_mixed_space_info(space_info))
2712 return ret;
2713
2714 if (space_info->flags & BTRFS_BLOCK_GROUP_METADATA) {
2715 ret = &fs_info->meta_alloc_cluster;
2716 if (btrfs_test_opt(fs_info, SSD))
2717 *empty_cluster = SZ_2M;
2718 else
2719 *empty_cluster = SZ_64K;
2720 } else if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
2721 btrfs_test_opt(fs_info, SSD_SPREAD)) {
2722 *empty_cluster = SZ_2M;
2723 ret = &fs_info->data_alloc_cluster;
2724 }
2725
2726 return ret;
2727 }
2728
2729 static int unpin_extent_range(struct btrfs_fs_info *fs_info,
2730 u64 start, u64 end,
2731 const bool return_free_space)
2732 {
2733 struct btrfs_block_group *cache = NULL;
2734 struct btrfs_space_info *space_info;
2735 struct btrfs_block_rsv *global_rsv = &fs_info->global_block_rsv;
2736 struct btrfs_free_cluster *cluster = NULL;
2737 u64 len;
2738 u64 total_unpinned = 0;
2739 u64 empty_cluster = 0;
2740 bool readonly;
2741 int ret = 0;
2742
2743 while (start <= end) {
2744 readonly = false;
2745 if (!cache ||
2746 start >= cache->start + cache->length) {
2747 if (cache)
2748 btrfs_put_block_group(cache);
2749 total_unpinned = 0;
2750 cache = btrfs_lookup_block_group(fs_info, start);
2751 if (cache == NULL) {
2752 /* Logic error, something removed the block group. */
2753 ret = -EUCLEAN;
2754 goto out;
2755 }
2756
2757 cluster = fetch_cluster_info(fs_info,
2758 cache->space_info,
2759 &empty_cluster);
2760 empty_cluster <<= 1;
2761 }
2762
2763 len = cache->start + cache->length - start;
2764 len = min(len, end + 1 - start);
2765
2766 if (return_free_space)
2767 btrfs_add_free_space(cache, start, len);
2768
2769 start += len;
2770 total_unpinned += len;
2771 space_info = cache->space_info;
2772
2773 /*
2774 * If this space cluster has been marked as fragmented and we've
2775 * unpinned enough in this block group to potentially allow a
2776 * cluster to be created inside of it go ahead and clear the
2777 * fragmented check.
2778 */
2779 if (cluster && cluster->fragmented &&
2780 total_unpinned > empty_cluster) {
2781 spin_lock(&cluster->lock);
2782 cluster->fragmented = 0;
2783 spin_unlock(&cluster->lock);
2784 }
2785
2786 spin_lock(&space_info->lock);
2787 spin_lock(&cache->lock);
2788 cache->pinned -= len;
2789 btrfs_space_info_update_bytes_pinned(fs_info, space_info, -len);
2790 space_info->max_extent_size = 0;
2791 if (cache->ro) {
2792 space_info->bytes_readonly += len;
2793 readonly = true;
2794 } else if (btrfs_is_zoned(fs_info)) {
2795 /* Need reset before reusing in a zoned block group */
2796 btrfs_space_info_update_bytes_zone_unusable(fs_info, space_info,
2797 len);
2798 readonly = true;
2799 }
2800 spin_unlock(&cache->lock);
2801 if (!readonly && return_free_space &&
2802 global_rsv->space_info == space_info) {
2803 spin_lock(&global_rsv->lock);
2804 if (!global_rsv->full) {
2805 u64 to_add = min(len, global_rsv->size -
2806 global_rsv->reserved);
2807
2808 global_rsv->reserved += to_add;
2809 btrfs_space_info_update_bytes_may_use(fs_info,
2810 space_info, to_add);
2811 if (global_rsv->reserved >= global_rsv->size)
2812 global_rsv->full = 1;
2813 len -= to_add;
2814 }
2815 spin_unlock(&global_rsv->lock);
2816 }
2817 /* Add to any tickets we may have */
2818 if (!readonly && return_free_space && len)
2819 btrfs_try_granting_tickets(fs_info, space_info);
2820 spin_unlock(&space_info->lock);
2821 }
2822
2823 if (cache)
2824 btrfs_put_block_group(cache);
2825 out:
2826 return ret;
2827 }
2828
2829 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
2830 {
2831 struct btrfs_fs_info *fs_info = trans->fs_info;
2832 struct btrfs_block_group *block_group, *tmp;
2833 struct list_head *deleted_bgs;
2834 struct extent_io_tree *unpin;
2835 u64 start;
2836 u64 end;
2837 int ret;
2838
2839 unpin = &trans->transaction->pinned_extents;
2840
2841 while (!TRANS_ABORTED(trans)) {
2842 struct extent_state *cached_state = NULL;
2843
2844 mutex_lock(&fs_info->unused_bg_unpin_mutex);
2845 if (!find_first_extent_bit(unpin, 0, &start, &end,
2846 EXTENT_DIRTY, &cached_state)) {
2847 mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2848 break;
2849 }
2850
2851 if (btrfs_test_opt(fs_info, DISCARD_SYNC))
2852 ret = btrfs_discard_extent(fs_info, start,
2853 end + 1 - start, NULL);
2854
2855 clear_extent_dirty(unpin, start, end, &cached_state);
2856 ret = unpin_extent_range(fs_info, start, end, true);
2857 BUG_ON(ret);
2858 mutex_unlock(&fs_info->unused_bg_unpin_mutex);
2859 free_extent_state(cached_state);
2860 cond_resched();
2861 }
2862
2863 if (btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
2864 btrfs_discard_calc_delay(&fs_info->discard_ctl);
2865 btrfs_discard_schedule_work(&fs_info->discard_ctl, true);
2866 }
2867
2868 /*
2869 * Transaction is finished. We don't need the lock anymore. We
2870 * do need to clean up the block groups in case of a transaction
2871 * abort.
2872 */
2873 deleted_bgs = &trans->transaction->deleted_bgs;
2874 list_for_each_entry_safe(block_group, tmp, deleted_bgs, bg_list) {
2875 u64 trimmed = 0;
2876
2877 ret = -EROFS;
2878 if (!TRANS_ABORTED(trans))
2879 ret = btrfs_discard_extent(fs_info,
2880 block_group->start,
2881 block_group->length,
2882 &trimmed);
2883
2884 list_del_init(&block_group->bg_list);
2885 btrfs_unfreeze_block_group(block_group);
2886 btrfs_put_block_group(block_group);
2887
2888 if (ret) {
2889 const char *errstr = btrfs_decode_error(ret);
2890 btrfs_warn(fs_info,
2891 "discard failed while removing blockgroup: errno=%d %s",
2892 ret, errstr);
2893 }
2894 }
2895
2896 return 0;
2897 }
2898
2899 /*
2900 * Parse an extent item's inline extents looking for a simple quotas owner ref.
2901 *
2902 * @fs_info: the btrfs_fs_info for this mount
2903 * @leaf: a leaf in the extent tree containing the extent item
2904 * @slot: the slot in the leaf where the extent item is found
2905 *
2906 * Returns the objectid of the root that originally allocated the extent item
2907 * if the inline owner ref is expected and present, otherwise 0.
2908 *
2909 * If an extent item has an owner ref item, it will be the first inline ref
2910 * item. Therefore the logic is to check whether there are any inline ref
2911 * items, then check the type of the first one.
2912 */
2913 u64 btrfs_get_extent_owner_root(struct btrfs_fs_info *fs_info,
2914 struct extent_buffer *leaf, int slot)
2915 {
2916 struct btrfs_extent_item *ei;
2917 struct btrfs_extent_inline_ref *iref;
2918 struct btrfs_extent_owner_ref *oref;
2919 unsigned long ptr;
2920 unsigned long end;
2921 int type;
2922
2923 if (!btrfs_fs_incompat(fs_info, SIMPLE_QUOTA))
2924 return 0;
2925
2926 ei = btrfs_item_ptr(leaf, slot, struct btrfs_extent_item);
2927 ptr = (unsigned long)(ei + 1);
2928 end = (unsigned long)ei + btrfs_item_size(leaf, slot);
2929
2930 /* No inline ref items of any kind, can't check type. */
2931 if (ptr == end)
2932 return 0;
2933
2934 iref = (struct btrfs_extent_inline_ref *)ptr;
2935 type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
2936
2937 /* We found an owner ref, get the root out of it. */
2938 if (type == BTRFS_EXTENT_OWNER_REF_KEY) {
2939 oref = (struct btrfs_extent_owner_ref *)(&iref->offset);
2940 return btrfs_extent_owner_ref_root_id(leaf, oref);
2941 }
2942
2943 /* We have inline refs, but not an owner ref. */
2944 return 0;
2945 }
2946
2947 static int do_free_extent_accounting(struct btrfs_trans_handle *trans,
2948 u64 bytenr, struct btrfs_squota_delta *delta)
2949 {
2950 int ret;
2951 u64 num_bytes = delta->num_bytes;
2952
2953 if (delta->is_data) {
2954 struct btrfs_root *csum_root;
2955
2956 csum_root = btrfs_csum_root(trans->fs_info, bytenr);
2957 ret = btrfs_del_csums(trans, csum_root, bytenr, num_bytes);
2958 if (ret) {
2959 btrfs_abort_transaction(trans, ret);
2960 return ret;
2961 }
2962
2963 ret = btrfs_delete_raid_extent(trans, bytenr, num_bytes);
2964 if (ret) {
2965 btrfs_abort_transaction(trans, ret);
2966 return ret;
2967 }
2968 }
2969
2970 ret = btrfs_record_squota_delta(trans->fs_info, delta);
2971 if (ret) {
2972 btrfs_abort_transaction(trans, ret);
2973 return ret;
2974 }
2975
2976 ret = add_to_free_space_tree(trans, bytenr, num_bytes);
2977 if (ret) {
2978 btrfs_abort_transaction(trans, ret);
2979 return ret;
2980 }
2981
2982 ret = btrfs_update_block_group(trans, bytenr, num_bytes, false);
2983 if (ret)
2984 btrfs_abort_transaction(trans, ret);
2985
2986 return ret;
2987 }
2988
2989 #define abort_and_dump(trans, path, fmt, args...) \
2990 ({ \
2991 btrfs_abort_transaction(trans, -EUCLEAN); \
2992 btrfs_print_leaf(path->nodes[0]); \
2993 btrfs_crit(trans->fs_info, fmt, ##args); \
2994 })
2995
2996 /*
2997 * Drop one or more refs of @node.
2998 *
2999 * 1. Locate the extent refs.
3000 * It's either inline in EXTENT/METADATA_ITEM or in keyed SHARED_* item.
3001 * Locate it, then reduce the refs number or remove the ref line completely.
3002 *
3003 * 2. Update the refs count in EXTENT/METADATA_ITEM
3004 *
3005 * Inline backref case:
3006 *
3007 * in extent tree we have:
3008 *
3009 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
3010 * refs 2 gen 6 flags DATA
3011 * extent data backref root FS_TREE objectid 258 offset 0 count 1
3012 * extent data backref root FS_TREE objectid 257 offset 0 count 1
3013 *
3014 * This function gets called with:
3015 *
3016 * node->bytenr = 13631488
3017 * node->num_bytes = 1048576
3018 * root_objectid = FS_TREE
3019 * owner_objectid = 257
3020 * owner_offset = 0
3021 * refs_to_drop = 1
3022 *
3023 * Then we should get some like:
3024 *
3025 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 16201 itemsize 82
3026 * refs 1 gen 6 flags DATA
3027 * extent data backref root FS_TREE objectid 258 offset 0 count 1
3028 *
3029 * Keyed backref case:
3030 *
3031 * in extent tree we have:
3032 *
3033 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
3034 * refs 754 gen 6 flags DATA
3035 * [...]
3036 * item 2 key (13631488 EXTENT_DATA_REF <HASH>) itemoff 3915 itemsize 28
3037 * extent data backref root FS_TREE objectid 866 offset 0 count 1
3038 *
3039 * This function get called with:
3040 *
3041 * node->bytenr = 13631488
3042 * node->num_bytes = 1048576
3043 * root_objectid = FS_TREE
3044 * owner_objectid = 866
3045 * owner_offset = 0
3046 * refs_to_drop = 1
3047 *
3048 * Then we should get some like:
3049 *
3050 * item 0 key (13631488 EXTENT_ITEM 1048576) itemoff 3971 itemsize 24
3051 * refs 753 gen 6 flags DATA
3052 *
3053 * And that (13631488 EXTENT_DATA_REF <HASH>) gets removed.
3054 */
3055 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3056 struct btrfs_delayed_ref_head *href,
3057 struct btrfs_delayed_ref_node *node,
3058 struct btrfs_delayed_extent_op *extent_op)
3059 {
3060 struct btrfs_fs_info *info = trans->fs_info;
3061 struct btrfs_key key;
3062 struct btrfs_path *path;
3063 struct btrfs_root *extent_root;
3064 struct extent_buffer *leaf;
3065 struct btrfs_extent_item *ei;
3066 struct btrfs_extent_inline_ref *iref;
3067 int ret;
3068 int is_data;
3069 int extent_slot = 0;
3070 int found_extent = 0;
3071 int num_to_del = 1;
3072 int refs_to_drop = node->ref_mod;
3073 u32 item_size;
3074 u64 refs;
3075 u64 bytenr = node->bytenr;
3076 u64 num_bytes = node->num_bytes;
3077 u64 owner_objectid = btrfs_delayed_ref_owner(node);
3078 u64 owner_offset = btrfs_delayed_ref_offset(node);
3079 bool skinny_metadata = btrfs_fs_incompat(info, SKINNY_METADATA);
3080 u64 delayed_ref_root = href->owning_root;
3081
3082 extent_root = btrfs_extent_root(info, bytenr);
3083 ASSERT(extent_root);
3084
3085 path = btrfs_alloc_path();
3086 if (!path)
3087 return -ENOMEM;
3088
3089 is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3090
3091 if (!is_data && refs_to_drop != 1) {
3092 btrfs_crit(info,
3093 "invalid refs_to_drop, dropping more than 1 refs for tree block %llu refs_to_drop %u",
3094 node->bytenr, refs_to_drop);
3095 ret = -EINVAL;
3096 btrfs_abort_transaction(trans, ret);
3097 goto out;
3098 }
3099
3100 if (is_data)
3101 skinny_metadata = false;
3102
3103 ret = lookup_extent_backref(trans, path, &iref, bytenr, num_bytes,
3104 node->parent, node->ref_root, owner_objectid,
3105 owner_offset);
3106 if (ret == 0) {
3107 /*
3108 * Either the inline backref or the SHARED_DATA_REF/
3109 * SHARED_BLOCK_REF is found
3110 *
3111 * Here is a quick path to locate EXTENT/METADATA_ITEM.
3112 * It's possible the EXTENT/METADATA_ITEM is near current slot.
3113 */
3114 extent_slot = path->slots[0];
3115 while (extent_slot >= 0) {
3116 btrfs_item_key_to_cpu(path->nodes[0], &key,
3117 extent_slot);
3118 if (key.objectid != bytenr)
3119 break;
3120 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3121 key.offset == num_bytes) {
3122 found_extent = 1;
3123 break;
3124 }
3125 if (key.type == BTRFS_METADATA_ITEM_KEY &&
3126 key.offset == owner_objectid) {
3127 found_extent = 1;
3128 break;
3129 }
3130
3131 /* Quick path didn't find the EXTEMT/METADATA_ITEM */
3132 if (path->slots[0] - extent_slot > 5)
3133 break;
3134 extent_slot--;
3135 }
3136
3137 if (!found_extent) {
3138 if (iref) {
3139 abort_and_dump(trans, path,
3140 "invalid iref slot %u, no EXTENT/METADATA_ITEM found but has inline extent ref",
3141 path->slots[0]);
3142 ret = -EUCLEAN;
3143 goto out;
3144 }
3145 /* Must be SHARED_* item, remove the backref first */
3146 ret = remove_extent_backref(trans, extent_root, path,
3147 NULL, refs_to_drop, is_data);
3148 if (ret) {
3149 btrfs_abort_transaction(trans, ret);
3150 goto out;
3151 }
3152 btrfs_release_path(path);
3153
3154 /* Slow path to locate EXTENT/METADATA_ITEM */
3155 key.objectid = bytenr;
3156 key.type = BTRFS_EXTENT_ITEM_KEY;
3157 key.offset = num_bytes;
3158
3159 if (!is_data && skinny_metadata) {
3160 key.type = BTRFS_METADATA_ITEM_KEY;
3161 key.offset = owner_objectid;
3162 }
3163
3164 ret = btrfs_search_slot(trans, extent_root,
3165 &key, path, -1, 1);
3166 if (ret > 0 && skinny_metadata && path->slots[0]) {
3167 /*
3168 * Couldn't find our skinny metadata item,
3169 * see if we have ye olde extent item.
3170 */
3171 path->slots[0]--;
3172 btrfs_item_key_to_cpu(path->nodes[0], &key,
3173 path->slots[0]);
3174 if (key.objectid == bytenr &&
3175 key.type == BTRFS_EXTENT_ITEM_KEY &&
3176 key.offset == num_bytes)
3177 ret = 0;
3178 }
3179
3180 if (ret > 0 && skinny_metadata) {
3181 skinny_metadata = false;
3182 key.objectid = bytenr;
3183 key.type = BTRFS_EXTENT_ITEM_KEY;
3184 key.offset = num_bytes;
3185 btrfs_release_path(path);
3186 ret = btrfs_search_slot(trans, extent_root,
3187 &key, path, -1, 1);
3188 }
3189
3190 if (ret) {
3191 if (ret > 0)
3192 btrfs_print_leaf(path->nodes[0]);
3193 btrfs_err(info,
3194 "umm, got %d back from search, was looking for %llu, slot %d",
3195 ret, bytenr, path->slots[0]);
3196 }
3197 if (ret < 0) {
3198 btrfs_abort_transaction(trans, ret);
3199 goto out;
3200 }
3201 extent_slot = path->slots[0];
3202 }
3203 } else if (WARN_ON(ret == -ENOENT)) {
3204 abort_and_dump(trans, path,
3205 "unable to find ref byte nr %llu parent %llu root %llu owner %llu offset %llu slot %d",
3206 bytenr, node->parent, node->ref_root, owner_objectid,
3207 owner_offset, path->slots[0]);
3208 goto out;
3209 } else {
3210 btrfs_abort_transaction(trans, ret);
3211 goto out;
3212 }
3213
3214 leaf = path->nodes[0];
3215 item_size = btrfs_item_size(leaf, extent_slot);
3216 if (unlikely(item_size < sizeof(*ei))) {
3217 ret = -EUCLEAN;
3218 btrfs_err(trans->fs_info,
3219 "unexpected extent item size, has %u expect >= %zu",
3220 item_size, sizeof(*ei));
3221 btrfs_abort_transaction(trans, ret);
3222 goto out;
3223 }
3224 ei = btrfs_item_ptr(leaf, extent_slot,
3225 struct btrfs_extent_item);
3226 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID &&
3227 key.type == BTRFS_EXTENT_ITEM_KEY) {
3228 struct btrfs_tree_block_info *bi;
3229
3230 if (item_size < sizeof(*ei) + sizeof(*bi)) {
3231 abort_and_dump(trans, path,
3232 "invalid extent item size for key (%llu, %u, %llu) slot %u owner %llu, has %u expect >= %zu",
3233 key.objectid, key.type, key.offset,
3234 path->slots[0], owner_objectid, item_size,
3235 sizeof(*ei) + sizeof(*bi));
3236 ret = -EUCLEAN;
3237 goto out;
3238 }
3239 bi = (struct btrfs_tree_block_info *)(ei + 1);
3240 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3241 }
3242
3243 refs = btrfs_extent_refs(leaf, ei);
3244 if (refs < refs_to_drop) {
3245 abort_and_dump(trans, path,
3246 "trying to drop %d refs but we only have %llu for bytenr %llu slot %u",
3247 refs_to_drop, refs, bytenr, path->slots[0]);
3248 ret = -EUCLEAN;
3249 goto out;
3250 }
3251 refs -= refs_to_drop;
3252
3253 if (refs > 0) {
3254 if (extent_op)
3255 __run_delayed_extent_op(extent_op, leaf, ei);
3256 /*
3257 * In the case of inline back ref, reference count will
3258 * be updated by remove_extent_backref
3259 */
3260 if (iref) {
3261 if (!found_extent) {
3262 abort_and_dump(trans, path,
3263 "invalid iref, got inlined extent ref but no EXTENT/METADATA_ITEM found, slot %u",
3264 path->slots[0]);
3265 ret = -EUCLEAN;
3266 goto out;
3267 }
3268 } else {
3269 btrfs_set_extent_refs(leaf, ei, refs);
3270 btrfs_mark_buffer_dirty(trans, leaf);
3271 }
3272 if (found_extent) {
3273 ret = remove_extent_backref(trans, extent_root, path,
3274 iref, refs_to_drop, is_data);
3275 if (ret) {
3276 btrfs_abort_transaction(trans, ret);
3277 goto out;
3278 }
3279 }
3280 } else {
3281 struct btrfs_squota_delta delta = {
3282 .root = delayed_ref_root,
3283 .num_bytes = num_bytes,
3284 .is_data = is_data,
3285 .is_inc = false,
3286 .generation = btrfs_extent_generation(leaf, ei),
3287 };
3288
3289 /* In this branch refs == 1 */
3290 if (found_extent) {
3291 if (is_data && refs_to_drop !=
3292 extent_data_ref_count(path, iref)) {
3293 abort_and_dump(trans, path,
3294 "invalid refs_to_drop, current refs %u refs_to_drop %u slot %u",
3295 extent_data_ref_count(path, iref),
3296 refs_to_drop, path->slots[0]);
3297 ret = -EUCLEAN;
3298 goto out;
3299 }
3300 if (iref) {
3301 if (path->slots[0] != extent_slot) {
3302 abort_and_dump(trans, path,
3303 "invalid iref, extent item key (%llu %u %llu) slot %u doesn't have wanted iref",
3304 key.objectid, key.type,
3305 key.offset, path->slots[0]);
3306 ret = -EUCLEAN;
3307 goto out;
3308 }
3309 } else {
3310 /*
3311 * No inline ref, we must be at SHARED_* item,
3312 * And it's single ref, it must be:
3313 * | extent_slot ||extent_slot + 1|
3314 * [ EXTENT/METADATA_ITEM ][ SHARED_* ITEM ]
3315 */
3316 if (path->slots[0] != extent_slot + 1) {
3317 abort_and_dump(trans, path,
3318 "invalid SHARED_* item slot %u, previous item is not EXTENT/METADATA_ITEM",
3319 path->slots[0]);
3320 ret = -EUCLEAN;
3321 goto out;
3322 }
3323 path->slots[0] = extent_slot;
3324 num_to_del = 2;
3325 }
3326 }
3327 /*
3328 * We can't infer the data owner from the delayed ref, so we need
3329 * to try to get it from the owning ref item.
3330 *
3331 * If it is not present, then that extent was not written under
3332 * simple quotas mode, so we don't need to account for its deletion.
3333 */
3334 if (is_data)
3335 delta.root = btrfs_get_extent_owner_root(trans->fs_info,
3336 leaf, extent_slot);
3337
3338 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3339 num_to_del);
3340 if (ret) {
3341 btrfs_abort_transaction(trans, ret);
3342 goto out;
3343 }
3344 btrfs_release_path(path);
3345
3346 ret = do_free_extent_accounting(trans, bytenr, &delta);
3347 }
3348 btrfs_release_path(path);
3349
3350 out:
3351 btrfs_free_path(path);
3352 return ret;
3353 }
3354
3355 /*
3356 * when we free an block, it is possible (and likely) that we free the last
3357 * delayed ref for that extent as well. This searches the delayed ref tree for
3358 * a given extent, and if there are no other delayed refs to be processed, it
3359 * removes it from the tree.
3360 */
3361 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3362 u64 bytenr)
3363 {
3364 struct btrfs_delayed_ref_head *head;
3365 struct btrfs_delayed_ref_root *delayed_refs;
3366 int ret = 0;
3367
3368 delayed_refs = &trans->transaction->delayed_refs;
3369 spin_lock(&delayed_refs->lock);
3370 head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
3371 if (!head)
3372 goto out_delayed_unlock;
3373
3374 spin_lock(&head->lock);
3375 if (!RB_EMPTY_ROOT(&head->ref_tree.rb_root))
3376 goto out;
3377
3378 if (cleanup_extent_op(head) != NULL)
3379 goto out;
3380
3381 /*
3382 * waiting for the lock here would deadlock. If someone else has it
3383 * locked they are already in the process of dropping it anyway
3384 */
3385 if (!mutex_trylock(&head->mutex))
3386 goto out;
3387
3388 btrfs_delete_ref_head(delayed_refs, head);
3389 head->processing = false;
3390
3391 spin_unlock(&head->lock);
3392 spin_unlock(&delayed_refs->lock);
3393
3394 BUG_ON(head->extent_op);
3395 if (head->must_insert_reserved)
3396 ret = 1;
3397
3398 btrfs_cleanup_ref_head_accounting(trans->fs_info, delayed_refs, head);
3399 mutex_unlock(&head->mutex);
3400 btrfs_put_delayed_ref_head(head);
3401 return ret;
3402 out:
3403 spin_unlock(&head->lock);
3404
3405 out_delayed_unlock:
3406 spin_unlock(&delayed_refs->lock);
3407 return 0;
3408 }
3409
3410 int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
3411 u64 root_id,
3412 struct extent_buffer *buf,
3413 u64 parent, int last_ref)
3414 {
3415 struct btrfs_fs_info *fs_info = trans->fs_info;
3416 struct btrfs_block_group *bg;
3417 int ret;
3418
3419 if (root_id != BTRFS_TREE_LOG_OBJECTID) {
3420 struct btrfs_ref generic_ref = {
3421 .action = BTRFS_DROP_DELAYED_REF,
3422 .bytenr = buf->start,
3423 .num_bytes = buf->len,
3424 .parent = parent,
3425 .owning_root = btrfs_header_owner(buf),
3426 .ref_root = root_id,
3427 };
3428
3429 /*
3430 * Assert that the extent buffer is not cleared due to
3431 * EXTENT_BUFFER_ZONED_ZEROOUT. Please refer
3432 * btrfs_clear_buffer_dirty() and btree_csum_one_bio() for
3433 * detail.
3434 */
3435 ASSERT(btrfs_header_bytenr(buf) != 0);
3436
3437 btrfs_init_tree_ref(&generic_ref, btrfs_header_level(buf), 0, false);
3438 btrfs_ref_tree_mod(fs_info, &generic_ref);
3439 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, NULL);
3440 if (ret < 0)
3441 return ret;
3442 }
3443
3444 if (!last_ref)
3445 return 0;
3446
3447 if (btrfs_header_generation(buf) != trans->transid)
3448 goto out;
3449
3450 if (root_id != BTRFS_TREE_LOG_OBJECTID) {
3451 ret = check_ref_cleanup(trans, buf->start);
3452 if (!ret)
3453 goto out;
3454 }
3455
3456 bg = btrfs_lookup_block_group(fs_info, buf->start);
3457
3458 if (btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3459 pin_down_extent(trans, bg, buf->start, buf->len, 1);
3460 btrfs_put_block_group(bg);
3461 goto out;
3462 }
3463
3464 /*
3465 * If there are tree mod log users we may have recorded mod log
3466 * operations for this node. If we re-allocate this node we
3467 * could replay operations on this node that happened when it
3468 * existed in a completely different root. For example if it
3469 * was part of root A, then was reallocated to root B, and we
3470 * are doing a btrfs_old_search_slot(root b), we could replay
3471 * operations that happened when the block was part of root A,
3472 * giving us an inconsistent view of the btree.
3473 *
3474 * We are safe from races here because at this point no other
3475 * node or root points to this extent buffer, so if after this
3476 * check a new tree mod log user joins we will not have an
3477 * existing log of operations on this node that we have to
3478 * contend with.
3479 */
3480
3481 if (test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags)
3482 || btrfs_is_zoned(fs_info)) {
3483 pin_down_extent(trans, bg, buf->start, buf->len, 1);
3484 btrfs_put_block_group(bg);
3485 goto out;
3486 }
3487
3488 WARN_ON(test_bit(EXTENT_BUFFER_DIRTY, &buf->bflags));
3489
3490 btrfs_add_free_space(bg, buf->start, buf->len);
3491 btrfs_free_reserved_bytes(bg, buf->len, 0);
3492 btrfs_put_block_group(bg);
3493 trace_btrfs_reserved_extent_free(fs_info, buf->start, buf->len);
3494
3495 out:
3496
3497 /*
3498 * Deleting the buffer, clear the corrupt flag since it doesn't
3499 * matter anymore.
3500 */
3501 clear_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags);
3502 return 0;
3503 }
3504
3505 /* Can return -ENOMEM */
3506 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_ref *ref)
3507 {
3508 struct btrfs_fs_info *fs_info = trans->fs_info;
3509 int ret;
3510
3511 if (btrfs_is_testing(fs_info))
3512 return 0;
3513
3514 /*
3515 * tree log blocks never actually go into the extent allocation
3516 * tree, just update pinning info and exit early.
3517 */
3518 if (ref->ref_root == BTRFS_TREE_LOG_OBJECTID) {
3519 btrfs_pin_extent(trans, ref->bytenr, ref->num_bytes, 1);
3520 ret = 0;
3521 } else if (ref->type == BTRFS_REF_METADATA) {
3522 ret = btrfs_add_delayed_tree_ref(trans, ref, NULL);
3523 } else {
3524 ret = btrfs_add_delayed_data_ref(trans, ref, 0);
3525 }
3526
3527 if (ref->ref_root != BTRFS_TREE_LOG_OBJECTID)
3528 btrfs_ref_tree_mod(fs_info, ref);
3529
3530 return ret;
3531 }
3532
3533 enum btrfs_loop_type {
3534 /*
3535 * Start caching block groups but do not wait for progress or for them
3536 * to be done.
3537 */
3538 LOOP_CACHING_NOWAIT,
3539
3540 /*
3541 * Wait for the block group free_space >= the space we're waiting for if
3542 * the block group isn't cached.
3543 */
3544 LOOP_CACHING_WAIT,
3545
3546 /*
3547 * Allow allocations to happen from block groups that do not yet have a
3548 * size classification.
3549 */
3550 LOOP_UNSET_SIZE_CLASS,
3551
3552 /*
3553 * Allocate a chunk and then retry the allocation.
3554 */
3555 LOOP_ALLOC_CHUNK,
3556
3557 /*
3558 * Ignore the size class restrictions for this allocation.
3559 */
3560 LOOP_WRONG_SIZE_CLASS,
3561
3562 /*
3563 * Ignore the empty size, only try to allocate the number of bytes
3564 * needed for this allocation.
3565 */
3566 LOOP_NO_EMPTY_SIZE,
3567 };
3568
3569 static inline void
3570 btrfs_lock_block_group(struct btrfs_block_group *cache,
3571 int delalloc)
3572 {
3573 if (delalloc)
3574 down_read(&cache->data_rwsem);
3575 }
3576
3577 static inline void btrfs_grab_block_group(struct btrfs_block_group *cache,
3578 int delalloc)
3579 {
3580 btrfs_get_block_group(cache);
3581 if (delalloc)
3582 down_read(&cache->data_rwsem);
3583 }
3584
3585 static struct btrfs_block_group *btrfs_lock_cluster(
3586 struct btrfs_block_group *block_group,
3587 struct btrfs_free_cluster *cluster,
3588 int delalloc)
3589 __acquires(&cluster->refill_lock)
3590 {
3591 struct btrfs_block_group *used_bg = NULL;
3592
3593 spin_lock(&cluster->refill_lock);
3594 while (1) {
3595 used_bg = cluster->block_group;
3596 if (!used_bg)
3597 return NULL;
3598
3599 if (used_bg == block_group)
3600 return used_bg;
3601
3602 btrfs_get_block_group(used_bg);
3603
3604 if (!delalloc)
3605 return used_bg;
3606
3607 if (down_read_trylock(&used_bg->data_rwsem))
3608 return used_bg;
3609
3610 spin_unlock(&cluster->refill_lock);
3611
3612 /* We should only have one-level nested. */
3613 down_read_nested(&used_bg->data_rwsem, SINGLE_DEPTH_NESTING);
3614
3615 spin_lock(&cluster->refill_lock);
3616 if (used_bg == cluster->block_group)
3617 return used_bg;
3618
3619 up_read(&used_bg->data_rwsem);
3620 btrfs_put_block_group(used_bg);
3621 }
3622 }
3623
3624 static inline void
3625 btrfs_release_block_group(struct btrfs_block_group *cache,
3626 int delalloc)
3627 {
3628 if (delalloc)
3629 up_read(&cache->data_rwsem);
3630 btrfs_put_block_group(cache);
3631 }
3632
3633 /*
3634 * Helper function for find_free_extent().
3635 *
3636 * Return -ENOENT to inform caller that we need fallback to unclustered mode.
3637 * Return >0 to inform caller that we find nothing
3638 * Return 0 means we have found a location and set ffe_ctl->found_offset.
3639 */
3640 static int find_free_extent_clustered(struct btrfs_block_group *bg,
3641 struct find_free_extent_ctl *ffe_ctl,
3642 struct btrfs_block_group **cluster_bg_ret)
3643 {
3644 struct btrfs_block_group *cluster_bg;
3645 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3646 u64 aligned_cluster;
3647 u64 offset;
3648 int ret;
3649
3650 cluster_bg = btrfs_lock_cluster(bg, last_ptr, ffe_ctl->delalloc);
3651 if (!cluster_bg)
3652 goto refill_cluster;
3653 if (cluster_bg != bg && (cluster_bg->ro ||
3654 !block_group_bits(cluster_bg, ffe_ctl->flags)))
3655 goto release_cluster;
3656
3657 offset = btrfs_alloc_from_cluster(cluster_bg, last_ptr,
3658 ffe_ctl->num_bytes, cluster_bg->start,
3659 &ffe_ctl->max_extent_size);
3660 if (offset) {
3661 /* We have a block, we're done */
3662 spin_unlock(&last_ptr->refill_lock);
3663 trace_btrfs_reserve_extent_cluster(cluster_bg, ffe_ctl);
3664 *cluster_bg_ret = cluster_bg;
3665 ffe_ctl->found_offset = offset;
3666 return 0;
3667 }
3668 WARN_ON(last_ptr->block_group != cluster_bg);
3669
3670 release_cluster:
3671 /*
3672 * If we are on LOOP_NO_EMPTY_SIZE, we can't set up a new clusters, so
3673 * lets just skip it and let the allocator find whatever block it can
3674 * find. If we reach this point, we will have tried the cluster
3675 * allocator plenty of times and not have found anything, so we are
3676 * likely way too fragmented for the clustering stuff to find anything.
3677 *
3678 * However, if the cluster is taken from the current block group,
3679 * release the cluster first, so that we stand a better chance of
3680 * succeeding in the unclustered allocation.
3681 */
3682 if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE && cluster_bg != bg) {
3683 spin_unlock(&last_ptr->refill_lock);
3684 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3685 return -ENOENT;
3686 }
3687
3688 /* This cluster didn't work out, free it and start over */
3689 btrfs_return_cluster_to_free_space(NULL, last_ptr);
3690
3691 if (cluster_bg != bg)
3692 btrfs_release_block_group(cluster_bg, ffe_ctl->delalloc);
3693
3694 refill_cluster:
3695 if (ffe_ctl->loop >= LOOP_NO_EMPTY_SIZE) {
3696 spin_unlock(&last_ptr->refill_lock);
3697 return -ENOENT;
3698 }
3699
3700 aligned_cluster = max_t(u64,
3701 ffe_ctl->empty_cluster + ffe_ctl->empty_size,
3702 bg->full_stripe_len);
3703 ret = btrfs_find_space_cluster(bg, last_ptr, ffe_ctl->search_start,
3704 ffe_ctl->num_bytes, aligned_cluster);
3705 if (ret == 0) {
3706 /* Now pull our allocation out of this cluster */
3707 offset = btrfs_alloc_from_cluster(bg, last_ptr,
3708 ffe_ctl->num_bytes, ffe_ctl->search_start,
3709 &ffe_ctl->max_extent_size);
3710 if (offset) {
3711 /* We found one, proceed */
3712 spin_unlock(&last_ptr->refill_lock);
3713 ffe_ctl->found_offset = offset;
3714 trace_btrfs_reserve_extent_cluster(bg, ffe_ctl);
3715 return 0;
3716 }
3717 }
3718 /*
3719 * At this point we either didn't find a cluster or we weren't able to
3720 * allocate a block from our cluster. Free the cluster we've been
3721 * trying to use, and go to the next block group.
3722 */
3723 btrfs_return_cluster_to_free_space(NULL, last_ptr);
3724 spin_unlock(&last_ptr->refill_lock);
3725 return 1;
3726 }
3727
3728 /*
3729 * Return >0 to inform caller that we find nothing
3730 * Return 0 when we found an free extent and set ffe_ctrl->found_offset
3731 */
3732 static int find_free_extent_unclustered(struct btrfs_block_group *bg,
3733 struct find_free_extent_ctl *ffe_ctl)
3734 {
3735 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
3736 u64 offset;
3737
3738 /*
3739 * We are doing an unclustered allocation, set the fragmented flag so
3740 * we don't bother trying to setup a cluster again until we get more
3741 * space.
3742 */
3743 if (unlikely(last_ptr)) {
3744 spin_lock(&last_ptr->lock);
3745 last_ptr->fragmented = 1;
3746 spin_unlock(&last_ptr->lock);
3747 }
3748 if (ffe_ctl->cached) {
3749 struct btrfs_free_space_ctl *free_space_ctl;
3750
3751 free_space_ctl = bg->free_space_ctl;
3752 spin_lock(&free_space_ctl->tree_lock);
3753 if (free_space_ctl->free_space <
3754 ffe_ctl->num_bytes + ffe_ctl->empty_cluster +
3755 ffe_ctl->empty_size) {
3756 ffe_ctl->total_free_space = max_t(u64,
3757 ffe_ctl->total_free_space,
3758 free_space_ctl->free_space);
3759 spin_unlock(&free_space_ctl->tree_lock);
3760 return 1;
3761 }
3762 spin_unlock(&free_space_ctl->tree_lock);
3763 }
3764
3765 offset = btrfs_find_space_for_alloc(bg, ffe_ctl->search_start,
3766 ffe_ctl->num_bytes, ffe_ctl->empty_size,
3767 &ffe_ctl->max_extent_size);
3768 if (!offset)
3769 return 1;
3770 ffe_ctl->found_offset = offset;
3771 return 0;
3772 }
3773
3774 static int do_allocation_clustered(struct btrfs_block_group *block_group,
3775 struct find_free_extent_ctl *ffe_ctl,
3776 struct btrfs_block_group **bg_ret)
3777 {
3778 int ret;
3779
3780 /* We want to try and use the cluster allocator, so lets look there */
3781 if (ffe_ctl->last_ptr && ffe_ctl->use_cluster) {
3782 ret = find_free_extent_clustered(block_group, ffe_ctl, bg_ret);
3783 if (ret >= 0)
3784 return ret;
3785 /* ret == -ENOENT case falls through */
3786 }
3787
3788 return find_free_extent_unclustered(block_group, ffe_ctl);
3789 }
3790
3791 /*
3792 * Tree-log block group locking
3793 * ============================
3794 *
3795 * fs_info::treelog_bg_lock protects the fs_info::treelog_bg which
3796 * indicates the starting address of a block group, which is reserved only
3797 * for tree-log metadata.
3798 *
3799 * Lock nesting
3800 * ============
3801 *
3802 * space_info::lock
3803 * block_group::lock
3804 * fs_info::treelog_bg_lock
3805 */
3806
3807 /*
3808 * Simple allocator for sequential-only block group. It only allows sequential
3809 * allocation. No need to play with trees. This function also reserves the
3810 * bytes as in btrfs_add_reserved_bytes.
3811 */
3812 static int do_allocation_zoned(struct btrfs_block_group *block_group,
3813 struct find_free_extent_ctl *ffe_ctl,
3814 struct btrfs_block_group **bg_ret)
3815 {
3816 struct btrfs_fs_info *fs_info = block_group->fs_info;
3817 struct btrfs_space_info *space_info = block_group->space_info;
3818 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
3819 u64 start = block_group->start;
3820 u64 num_bytes = ffe_ctl->num_bytes;
3821 u64 avail;
3822 u64 bytenr = block_group->start;
3823 u64 log_bytenr;
3824 u64 data_reloc_bytenr;
3825 int ret = 0;
3826 bool skip = false;
3827
3828 ASSERT(btrfs_is_zoned(block_group->fs_info));
3829
3830 /*
3831 * Do not allow non-tree-log blocks in the dedicated tree-log block
3832 * group, and vice versa.
3833 */
3834 spin_lock(&fs_info->treelog_bg_lock);
3835 log_bytenr = fs_info->treelog_bg;
3836 if (log_bytenr && ((ffe_ctl->for_treelog && bytenr != log_bytenr) ||
3837 (!ffe_ctl->for_treelog && bytenr == log_bytenr)))
3838 skip = true;
3839 spin_unlock(&fs_info->treelog_bg_lock);
3840 if (skip)
3841 return 1;
3842
3843 /*
3844 * Do not allow non-relocation blocks in the dedicated relocation block
3845 * group, and vice versa.
3846 */
3847 spin_lock(&fs_info->relocation_bg_lock);
3848 data_reloc_bytenr = fs_info->data_reloc_bg;
3849 if (data_reloc_bytenr &&
3850 ((ffe_ctl->for_data_reloc && bytenr != data_reloc_bytenr) ||
3851 (!ffe_ctl->for_data_reloc && bytenr == data_reloc_bytenr)))
3852 skip = true;
3853 spin_unlock(&fs_info->relocation_bg_lock);
3854 if (skip)
3855 return 1;
3856
3857 /* Check RO and no space case before trying to activate it */
3858 spin_lock(&block_group->lock);
3859 if (block_group->ro || btrfs_zoned_bg_is_full(block_group)) {
3860 ret = 1;
3861 /*
3862 * May need to clear fs_info->{treelog,data_reloc}_bg.
3863 * Return the error after taking the locks.
3864 */
3865 }
3866 spin_unlock(&block_group->lock);
3867
3868 /* Metadata block group is activated at write time. */
3869 if (!ret && (block_group->flags & BTRFS_BLOCK_GROUP_DATA) &&
3870 !btrfs_zone_activate(block_group)) {
3871 ret = 1;
3872 /*
3873 * May need to clear fs_info->{treelog,data_reloc}_bg.
3874 * Return the error after taking the locks.
3875 */
3876 }
3877
3878 spin_lock(&space_info->lock);
3879 spin_lock(&block_group->lock);
3880 spin_lock(&fs_info->treelog_bg_lock);
3881 spin_lock(&fs_info->relocation_bg_lock);
3882
3883 if (ret)
3884 goto out;
3885
3886 ASSERT(!ffe_ctl->for_treelog ||
3887 block_group->start == fs_info->treelog_bg ||
3888 fs_info->treelog_bg == 0);
3889 ASSERT(!ffe_ctl->for_data_reloc ||
3890 block_group->start == fs_info->data_reloc_bg ||
3891 fs_info->data_reloc_bg == 0);
3892
3893 if (block_group->ro ||
3894 (!ffe_ctl->for_data_reloc &&
3895 test_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC, &block_group->runtime_flags))) {
3896 ret = 1;
3897 goto out;
3898 }
3899
3900 /*
3901 * Do not allow currently using block group to be tree-log dedicated
3902 * block group.
3903 */
3904 if (ffe_ctl->for_treelog && !fs_info->treelog_bg &&
3905 (block_group->used || block_group->reserved)) {
3906 ret = 1;
3907 goto out;
3908 }
3909
3910 /*
3911 * Do not allow currently used block group to be the data relocation
3912 * dedicated block group.
3913 */
3914 if (ffe_ctl->for_data_reloc && !fs_info->data_reloc_bg &&
3915 (block_group->used || block_group->reserved)) {
3916 ret = 1;
3917 goto out;
3918 }
3919
3920 WARN_ON_ONCE(block_group->alloc_offset > block_group->zone_capacity);
3921 avail = block_group->zone_capacity - block_group->alloc_offset;
3922 if (avail < num_bytes) {
3923 if (ffe_ctl->max_extent_size < avail) {
3924 /*
3925 * With sequential allocator, free space is always
3926 * contiguous
3927 */
3928 ffe_ctl->max_extent_size = avail;
3929 ffe_ctl->total_free_space = avail;
3930 }
3931 ret = 1;
3932 goto out;
3933 }
3934
3935 if (ffe_ctl->for_treelog && !fs_info->treelog_bg)
3936 fs_info->treelog_bg = block_group->start;
3937
3938 if (ffe_ctl->for_data_reloc) {
3939 if (!fs_info->data_reloc_bg)
3940 fs_info->data_reloc_bg = block_group->start;
3941 /*
3942 * Do not allow allocations from this block group, unless it is
3943 * for data relocation. Compared to increasing the ->ro, setting
3944 * the ->zoned_data_reloc_ongoing flag still allows nocow
3945 * writers to come in. See btrfs_inc_nocow_writers().
3946 *
3947 * We need to disable an allocation to avoid an allocation of
3948 * regular (non-relocation data) extent. With mix of relocation
3949 * extents and regular extents, we can dispatch WRITE commands
3950 * (for relocation extents) and ZONE APPEND commands (for
3951 * regular extents) at the same time to the same zone, which
3952 * easily break the write pointer.
3953 *
3954 * Also, this flag avoids this block group to be zone finished.
3955 */
3956 set_bit(BLOCK_GROUP_FLAG_ZONED_DATA_RELOC, &block_group->runtime_flags);
3957 }
3958
3959 ffe_ctl->found_offset = start + block_group->alloc_offset;
3960 block_group->alloc_offset += num_bytes;
3961 spin_lock(&ctl->tree_lock);
3962 ctl->free_space -= num_bytes;
3963 spin_unlock(&ctl->tree_lock);
3964
3965 /*
3966 * We do not check if found_offset is aligned to stripesize. The
3967 * address is anyway rewritten when using zone append writing.
3968 */
3969
3970 ffe_ctl->search_start = ffe_ctl->found_offset;
3971
3972 out:
3973 if (ret && ffe_ctl->for_treelog)
3974 fs_info->treelog_bg = 0;
3975 if (ret && ffe_ctl->for_data_reloc)
3976 fs_info->data_reloc_bg = 0;
3977 spin_unlock(&fs_info->relocation_bg_lock);
3978 spin_unlock(&fs_info->treelog_bg_lock);
3979 spin_unlock(&block_group->lock);
3980 spin_unlock(&space_info->lock);
3981 return ret;
3982 }
3983
3984 static int do_allocation(struct btrfs_block_group *block_group,
3985 struct find_free_extent_ctl *ffe_ctl,
3986 struct btrfs_block_group **bg_ret)
3987 {
3988 switch (ffe_ctl->policy) {
3989 case BTRFS_EXTENT_ALLOC_CLUSTERED:
3990 return do_allocation_clustered(block_group, ffe_ctl, bg_ret);
3991 case BTRFS_EXTENT_ALLOC_ZONED:
3992 return do_allocation_zoned(block_group, ffe_ctl, bg_ret);
3993 default:
3994 BUG();
3995 }
3996 }
3997
3998 static void release_block_group(struct btrfs_block_group *block_group,
3999 struct find_free_extent_ctl *ffe_ctl,
4000 int delalloc)
4001 {
4002 switch (ffe_ctl->policy) {
4003 case BTRFS_EXTENT_ALLOC_CLUSTERED:
4004 ffe_ctl->retry_uncached = false;
4005 break;
4006 case BTRFS_EXTENT_ALLOC_ZONED:
4007 /* Nothing to do */
4008 break;
4009 default:
4010 BUG();
4011 }
4012
4013 BUG_ON(btrfs_bg_flags_to_raid_index(block_group->flags) !=
4014 ffe_ctl->index);
4015 btrfs_release_block_group(block_group, delalloc);
4016 }
4017
4018 static void found_extent_clustered(struct find_free_extent_ctl *ffe_ctl,
4019 struct btrfs_key *ins)
4020 {
4021 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
4022
4023 if (!ffe_ctl->use_cluster && last_ptr) {
4024 spin_lock(&last_ptr->lock);
4025 last_ptr->window_start = ins->objectid;
4026 spin_unlock(&last_ptr->lock);
4027 }
4028 }
4029
4030 static void found_extent(struct find_free_extent_ctl *ffe_ctl,
4031 struct btrfs_key *ins)
4032 {
4033 switch (ffe_ctl->policy) {
4034 case BTRFS_EXTENT_ALLOC_CLUSTERED:
4035 found_extent_clustered(ffe_ctl, ins);
4036 break;
4037 case BTRFS_EXTENT_ALLOC_ZONED:
4038 /* Nothing to do */
4039 break;
4040 default:
4041 BUG();
4042 }
4043 }
4044
4045 static int can_allocate_chunk_zoned(struct btrfs_fs_info *fs_info,
4046 struct find_free_extent_ctl *ffe_ctl)
4047 {
4048 /* Block group's activeness is not a requirement for METADATA block groups. */
4049 if (!(ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA))
4050 return 0;
4051
4052 /* If we can activate new zone, just allocate a chunk and use it */
4053 if (btrfs_can_activate_zone(fs_info->fs_devices, ffe_ctl->flags))
4054 return 0;
4055
4056 /*
4057 * We already reached the max active zones. Try to finish one block
4058 * group to make a room for a new block group. This is only possible
4059 * for a data block group because btrfs_zone_finish() may need to wait
4060 * for a running transaction which can cause a deadlock for metadata
4061 * allocation.
4062 */
4063 if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA) {
4064 int ret = btrfs_zone_finish_one_bg(fs_info);
4065
4066 if (ret == 1)
4067 return 0;
4068 else if (ret < 0)
4069 return ret;
4070 }
4071
4072 /*
4073 * If we have enough free space left in an already active block group
4074 * and we can't activate any other zone now, do not allow allocating a
4075 * new chunk and let find_free_extent() retry with a smaller size.
4076 */
4077 if (ffe_ctl->max_extent_size >= ffe_ctl->min_alloc_size)
4078 return -ENOSPC;
4079
4080 /*
4081 * Even min_alloc_size is not left in any block groups. Since we cannot
4082 * activate a new block group, allocating it may not help. Let's tell a
4083 * caller to try again and hope it progress something by writing some
4084 * parts of the region. That is only possible for data block groups,
4085 * where a part of the region can be written.
4086 */
4087 if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA)
4088 return -EAGAIN;
4089
4090 /*
4091 * We cannot activate a new block group and no enough space left in any
4092 * block groups. So, allocating a new block group may not help. But,
4093 * there is nothing to do anyway, so let's go with it.
4094 */
4095 return 0;
4096 }
4097
4098 static int can_allocate_chunk(struct btrfs_fs_info *fs_info,
4099 struct find_free_extent_ctl *ffe_ctl)
4100 {
4101 switch (ffe_ctl->policy) {
4102 case BTRFS_EXTENT_ALLOC_CLUSTERED:
4103 return 0;
4104 case BTRFS_EXTENT_ALLOC_ZONED:
4105 return can_allocate_chunk_zoned(fs_info, ffe_ctl);
4106 default:
4107 BUG();
4108 }
4109 }
4110
4111 /*
4112 * Return >0 means caller needs to re-search for free extent
4113 * Return 0 means we have the needed free extent.
4114 * Return <0 means we failed to locate any free extent.
4115 */
4116 static int find_free_extent_update_loop(struct btrfs_fs_info *fs_info,
4117 struct btrfs_key *ins,
4118 struct find_free_extent_ctl *ffe_ctl,
4119 bool full_search)
4120 {
4121 struct btrfs_root *root = fs_info->chunk_root;
4122 int ret;
4123
4124 if ((ffe_ctl->loop == LOOP_CACHING_NOWAIT) &&
4125 ffe_ctl->have_caching_bg && !ffe_ctl->orig_have_caching_bg)
4126 ffe_ctl->orig_have_caching_bg = true;
4127
4128 if (ins->objectid) {
4129 found_extent(ffe_ctl, ins);
4130 return 0;
4131 }
4132
4133 if (ffe_ctl->loop >= LOOP_CACHING_WAIT && ffe_ctl->have_caching_bg)
4134 return 1;
4135
4136 ffe_ctl->index++;
4137 if (ffe_ctl->index < BTRFS_NR_RAID_TYPES)
4138 return 1;
4139
4140 /* See the comments for btrfs_loop_type for an explanation of the phases. */
4141 if (ffe_ctl->loop < LOOP_NO_EMPTY_SIZE) {
4142 ffe_ctl->index = 0;
4143 /*
4144 * We want to skip the LOOP_CACHING_WAIT step if we don't have
4145 * any uncached bgs and we've already done a full search
4146 * through.
4147 */
4148 if (ffe_ctl->loop == LOOP_CACHING_NOWAIT &&
4149 (!ffe_ctl->orig_have_caching_bg && full_search))
4150 ffe_ctl->loop++;
4151 ffe_ctl->loop++;
4152
4153 if (ffe_ctl->loop == LOOP_ALLOC_CHUNK) {
4154 struct btrfs_trans_handle *trans;
4155 int exist = 0;
4156
4157 /* Check if allocation policy allows to create a new chunk */
4158 ret = can_allocate_chunk(fs_info, ffe_ctl);
4159 if (ret)
4160 return ret;
4161
4162 trans = current->journal_info;
4163 if (trans)
4164 exist = 1;
4165 else
4166 trans = btrfs_join_transaction(root);
4167
4168 if (IS_ERR(trans)) {
4169 ret = PTR_ERR(trans);
4170 return ret;
4171 }
4172
4173 ret = btrfs_chunk_alloc(trans, ffe_ctl->flags,
4174 CHUNK_ALLOC_FORCE_FOR_EXTENT);
4175
4176 /* Do not bail out on ENOSPC since we can do more. */
4177 if (ret == -ENOSPC) {
4178 ret = 0;
4179 ffe_ctl->loop++;
4180 }
4181 else if (ret < 0)
4182 btrfs_abort_transaction(trans, ret);
4183 else
4184 ret = 0;
4185 if (!exist)
4186 btrfs_end_transaction(trans);
4187 if (ret)
4188 return ret;
4189 }
4190
4191 if (ffe_ctl->loop == LOOP_NO_EMPTY_SIZE) {
4192 if (ffe_ctl->policy != BTRFS_EXTENT_ALLOC_CLUSTERED)
4193 return -ENOSPC;
4194
4195 /*
4196 * Don't loop again if we already have no empty_size and
4197 * no empty_cluster.
4198 */
4199 if (ffe_ctl->empty_size == 0 &&
4200 ffe_ctl->empty_cluster == 0)
4201 return -ENOSPC;
4202 ffe_ctl->empty_size = 0;
4203 ffe_ctl->empty_cluster = 0;
4204 }
4205 return 1;
4206 }
4207 return -ENOSPC;
4208 }
4209
4210 static bool find_free_extent_check_size_class(struct find_free_extent_ctl *ffe_ctl,
4211 struct btrfs_block_group *bg)
4212 {
4213 if (ffe_ctl->policy == BTRFS_EXTENT_ALLOC_ZONED)
4214 return true;
4215 if (!btrfs_block_group_should_use_size_class(bg))
4216 return true;
4217 if (ffe_ctl->loop >= LOOP_WRONG_SIZE_CLASS)
4218 return true;
4219 if (ffe_ctl->loop >= LOOP_UNSET_SIZE_CLASS &&
4220 bg->size_class == BTRFS_BG_SZ_NONE)
4221 return true;
4222 return ffe_ctl->size_class == bg->size_class;
4223 }
4224
4225 static int prepare_allocation_clustered(struct btrfs_fs_info *fs_info,
4226 struct find_free_extent_ctl *ffe_ctl,
4227 struct btrfs_space_info *space_info,
4228 struct btrfs_key *ins)
4229 {
4230 /*
4231 * If our free space is heavily fragmented we may not be able to make
4232 * big contiguous allocations, so instead of doing the expensive search
4233 * for free space, simply return ENOSPC with our max_extent_size so we
4234 * can go ahead and search for a more manageable chunk.
4235 *
4236 * If our max_extent_size is large enough for our allocation simply
4237 * disable clustering since we will likely not be able to find enough
4238 * space to create a cluster and induce latency trying.
4239 */
4240 if (space_info->max_extent_size) {
4241 spin_lock(&space_info->lock);
4242 if (space_info->max_extent_size &&
4243 ffe_ctl->num_bytes > space_info->max_extent_size) {
4244 ins->offset = space_info->max_extent_size;
4245 spin_unlock(&space_info->lock);
4246 return -ENOSPC;
4247 } else if (space_info->max_extent_size) {
4248 ffe_ctl->use_cluster = false;
4249 }
4250 spin_unlock(&space_info->lock);
4251 }
4252
4253 ffe_ctl->last_ptr = fetch_cluster_info(fs_info, space_info,
4254 &ffe_ctl->empty_cluster);
4255 if (ffe_ctl->last_ptr) {
4256 struct btrfs_free_cluster *last_ptr = ffe_ctl->last_ptr;
4257
4258 spin_lock(&last_ptr->lock);
4259 if (last_ptr->block_group)
4260 ffe_ctl->hint_byte = last_ptr->window_start;
4261 if (last_ptr->fragmented) {
4262 /*
4263 * We still set window_start so we can keep track of the
4264 * last place we found an allocation to try and save
4265 * some time.
4266 */
4267 ffe_ctl->hint_byte = last_ptr->window_start;
4268 ffe_ctl->use_cluster = false;
4269 }
4270 spin_unlock(&last_ptr->lock);
4271 }
4272
4273 return 0;
4274 }
4275
4276 static int prepare_allocation_zoned(struct btrfs_fs_info *fs_info,
4277 struct find_free_extent_ctl *ffe_ctl)
4278 {
4279 if (ffe_ctl->for_treelog) {
4280 spin_lock(&fs_info->treelog_bg_lock);
4281 if (fs_info->treelog_bg)
4282 ffe_ctl->hint_byte = fs_info->treelog_bg;
4283 spin_unlock(&fs_info->treelog_bg_lock);
4284 } else if (ffe_ctl->for_data_reloc) {
4285 spin_lock(&fs_info->relocation_bg_lock);
4286 if (fs_info->data_reloc_bg)
4287 ffe_ctl->hint_byte = fs_info->data_reloc_bg;
4288 spin_unlock(&fs_info->relocation_bg_lock);
4289 } else if (ffe_ctl->flags & BTRFS_BLOCK_GROUP_DATA) {
4290 struct btrfs_block_group *block_group;
4291
4292 spin_lock(&fs_info->zone_active_bgs_lock);
4293 list_for_each_entry(block_group, &fs_info->zone_active_bgs, active_bg_list) {
4294 /*
4295 * No lock is OK here because avail is monotinically
4296 * decreasing, and this is just a hint.
4297 */
4298 u64 avail = block_group->zone_capacity - block_group->alloc_offset;
4299
4300 if (block_group_bits(block_group, ffe_ctl->flags) &&
4301 avail >= ffe_ctl->num_bytes) {
4302 ffe_ctl->hint_byte = block_group->start;
4303 break;
4304 }
4305 }
4306 spin_unlock(&fs_info->zone_active_bgs_lock);
4307 }
4308
4309 return 0;
4310 }
4311
4312 static int prepare_allocation(struct btrfs_fs_info *fs_info,
4313 struct find_free_extent_ctl *ffe_ctl,
4314 struct btrfs_space_info *space_info,
4315 struct btrfs_key *ins)
4316 {
4317 switch (ffe_ctl->policy) {
4318 case BTRFS_EXTENT_ALLOC_CLUSTERED:
4319 return prepare_allocation_clustered(fs_info, ffe_ctl,
4320 space_info, ins);
4321 case BTRFS_EXTENT_ALLOC_ZONED:
4322 return prepare_allocation_zoned(fs_info, ffe_ctl);
4323 default:
4324 BUG();
4325 }
4326 }
4327
4328 /*
4329 * walks the btree of allocated extents and find a hole of a given size.
4330 * The key ins is changed to record the hole:
4331 * ins->objectid == start position
4332 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4333 * ins->offset == the size of the hole.
4334 * Any available blocks before search_start are skipped.
4335 *
4336 * If there is no suitable free space, we will record the max size of
4337 * the free space extent currently.
4338 *
4339 * The overall logic and call chain:
4340 *
4341 * find_free_extent()
4342 * |- Iterate through all block groups
4343 * | |- Get a valid block group
4344 * | |- Try to do clustered allocation in that block group
4345 * | |- Try to do unclustered allocation in that block group
4346 * | |- Check if the result is valid
4347 * | | |- If valid, then exit
4348 * | |- Jump to next block group
4349 * |
4350 * |- Push harder to find free extents
4351 * |- If not found, re-iterate all block groups
4352 */
4353 static noinline int find_free_extent(struct btrfs_root *root,
4354 struct btrfs_key *ins,
4355 struct find_free_extent_ctl *ffe_ctl)
4356 {
4357 struct btrfs_fs_info *fs_info = root->fs_info;
4358 int ret = 0;
4359 int cache_block_group_error = 0;
4360 struct btrfs_block_group *block_group = NULL;
4361 struct btrfs_space_info *space_info;
4362 bool full_search = false;
4363
4364 WARN_ON(ffe_ctl->num_bytes < fs_info->sectorsize);
4365
4366 ffe_ctl->search_start = 0;
4367 /* For clustered allocation */
4368 ffe_ctl->empty_cluster = 0;
4369 ffe_ctl->last_ptr = NULL;
4370 ffe_ctl->use_cluster = true;
4371 ffe_ctl->have_caching_bg = false;
4372 ffe_ctl->orig_have_caching_bg = false;
4373 ffe_ctl->index = btrfs_bg_flags_to_raid_index(ffe_ctl->flags);
4374 ffe_ctl->loop = 0;
4375 ffe_ctl->retry_uncached = false;
4376 ffe_ctl->cached = 0;
4377 ffe_ctl->max_extent_size = 0;
4378 ffe_ctl->total_free_space = 0;
4379 ffe_ctl->found_offset = 0;
4380 ffe_ctl->policy = BTRFS_EXTENT_ALLOC_CLUSTERED;
4381 ffe_ctl->size_class = btrfs_calc_block_group_size_class(ffe_ctl->num_bytes);
4382
4383 if (btrfs_is_zoned(fs_info))
4384 ffe_ctl->policy = BTRFS_EXTENT_ALLOC_ZONED;
4385
4386 ins->type = BTRFS_EXTENT_ITEM_KEY;
4387 ins->objectid = 0;
4388 ins->offset = 0;
4389
4390 trace_find_free_extent(root, ffe_ctl);
4391
4392 space_info = btrfs_find_space_info(fs_info, ffe_ctl->flags);
4393 if (!space_info) {
4394 btrfs_err(fs_info, "No space info for %llu", ffe_ctl->flags);
4395 return -ENOSPC;
4396 }
4397
4398 ret = prepare_allocation(fs_info, ffe_ctl, space_info, ins);
4399 if (ret < 0)
4400 return ret;
4401
4402 ffe_ctl->search_start = max(ffe_ctl->search_start,
4403 first_logical_byte(fs_info));
4404 ffe_ctl->search_start = max(ffe_ctl->search_start, ffe_ctl->hint_byte);
4405 if (ffe_ctl->search_start == ffe_ctl->hint_byte) {
4406 block_group = btrfs_lookup_block_group(fs_info,
4407 ffe_ctl->search_start);
4408 /*
4409 * we don't want to use the block group if it doesn't match our
4410 * allocation bits, or if its not cached.
4411 *
4412 * However if we are re-searching with an ideal block group
4413 * picked out then we don't care that the block group is cached.
4414 */
4415 if (block_group && block_group_bits(block_group, ffe_ctl->flags) &&
4416 block_group->cached != BTRFS_CACHE_NO) {
4417 down_read(&space_info->groups_sem);
4418 if (list_empty(&block_group->list) ||
4419 block_group->ro) {
4420 /*
4421 * someone is removing this block group,
4422 * we can't jump into the have_block_group
4423 * target because our list pointers are not
4424 * valid
4425 */
4426 btrfs_put_block_group(block_group);
4427 up_read(&space_info->groups_sem);
4428 } else {
4429 ffe_ctl->index = btrfs_bg_flags_to_raid_index(
4430 block_group->flags);
4431 btrfs_lock_block_group(block_group,
4432 ffe_ctl->delalloc);
4433 ffe_ctl->hinted = true;
4434 goto have_block_group;
4435 }
4436 } else if (block_group) {
4437 btrfs_put_block_group(block_group);
4438 }
4439 }
4440 search:
4441 trace_find_free_extent_search_loop(root, ffe_ctl);
4442 ffe_ctl->have_caching_bg = false;
4443 if (ffe_ctl->index == btrfs_bg_flags_to_raid_index(ffe_ctl->flags) ||
4444 ffe_ctl->index == 0)
4445 full_search = true;
4446 down_read(&space_info->groups_sem);
4447 list_for_each_entry(block_group,
4448 &space_info->block_groups[ffe_ctl->index], list) {
4449 struct btrfs_block_group *bg_ret;
4450
4451 ffe_ctl->hinted = false;
4452 /* If the block group is read-only, we can skip it entirely. */
4453 if (unlikely(block_group->ro)) {
4454 if (ffe_ctl->for_treelog)
4455 btrfs_clear_treelog_bg(block_group);
4456 if (ffe_ctl->for_data_reloc)
4457 btrfs_clear_data_reloc_bg(block_group);
4458 continue;
4459 }
4460
4461 btrfs_grab_block_group(block_group, ffe_ctl->delalloc);
4462 ffe_ctl->search_start = block_group->start;
4463
4464 /*
4465 * this can happen if we end up cycling through all the
4466 * raid types, but we want to make sure we only allocate
4467 * for the proper type.
4468 */
4469 if (!block_group_bits(block_group, ffe_ctl->flags)) {
4470 u64 extra = BTRFS_BLOCK_GROUP_DUP |
4471 BTRFS_BLOCK_GROUP_RAID1_MASK |
4472 BTRFS_BLOCK_GROUP_RAID56_MASK |
4473 BTRFS_BLOCK_GROUP_RAID10;
4474
4475 /*
4476 * if they asked for extra copies and this block group
4477 * doesn't provide them, bail. This does allow us to
4478 * fill raid0 from raid1.
4479 */
4480 if ((ffe_ctl->flags & extra) && !(block_group->flags & extra))
4481 goto loop;
4482
4483 /*
4484 * This block group has different flags than we want.
4485 * It's possible that we have MIXED_GROUP flag but no
4486 * block group is mixed. Just skip such block group.
4487 */
4488 btrfs_release_block_group(block_group, ffe_ctl->delalloc);
4489 continue;
4490 }
4491
4492 have_block_group:
4493 trace_find_free_extent_have_block_group(root, ffe_ctl, block_group);
4494 ffe_ctl->cached = btrfs_block_group_done(block_group);
4495 if (unlikely(!ffe_ctl->cached)) {
4496 ffe_ctl->have_caching_bg = true;
4497 ret = btrfs_cache_block_group(block_group, false);
4498
4499 /*
4500 * If we get ENOMEM here or something else we want to
4501 * try other block groups, because it may not be fatal.
4502 * However if we can't find anything else we need to
4503 * save our return here so that we return the actual
4504 * error that caused problems, not ENOSPC.
4505 */
4506 if (ret < 0) {
4507 if (!cache_block_group_error)
4508 cache_block_group_error = ret;
4509 ret = 0;
4510 goto loop;
4511 }
4512 ret = 0;
4513 }
4514
4515 if (unlikely(block_group->cached == BTRFS_CACHE_ERROR)) {
4516 if (!cache_block_group_error)
4517 cache_block_group_error = -EIO;
4518 goto loop;
4519 }
4520
4521 if (!find_free_extent_check_size_class(ffe_ctl, block_group))
4522 goto loop;
4523
4524 bg_ret = NULL;
4525 ret = do_allocation(block_group, ffe_ctl, &bg_ret);
4526 if (ret > 0)
4527 goto loop;
4528
4529 if (bg_ret && bg_ret != block_group) {
4530 btrfs_release_block_group(block_group, ffe_ctl->delalloc);
4531 block_group = bg_ret;
4532 }
4533
4534 /* Checks */
4535 ffe_ctl->search_start = round_up(ffe_ctl->found_offset,
4536 fs_info->stripesize);
4537
4538 /* move on to the next group */
4539 if (ffe_ctl->search_start + ffe_ctl->num_bytes >
4540 block_group->start + block_group->length) {
4541 btrfs_add_free_space_unused(block_group,
4542 ffe_ctl->found_offset,
4543 ffe_ctl->num_bytes);
4544 goto loop;
4545 }
4546
4547 if (ffe_ctl->found_offset < ffe_ctl->search_start)
4548 btrfs_add_free_space_unused(block_group,
4549 ffe_ctl->found_offset,
4550 ffe_ctl->search_start - ffe_ctl->found_offset);
4551
4552 ret = btrfs_add_reserved_bytes(block_group, ffe_ctl->ram_bytes,
4553 ffe_ctl->num_bytes,
4554 ffe_ctl->delalloc,
4555 ffe_ctl->loop >= LOOP_WRONG_SIZE_CLASS);
4556 if (ret == -EAGAIN) {
4557 btrfs_add_free_space_unused(block_group,
4558 ffe_ctl->found_offset,
4559 ffe_ctl->num_bytes);
4560 goto loop;
4561 }
4562 btrfs_inc_block_group_reservations(block_group);
4563
4564 /* we are all good, lets return */
4565 ins->objectid = ffe_ctl->search_start;
4566 ins->offset = ffe_ctl->num_bytes;
4567
4568 trace_btrfs_reserve_extent(block_group, ffe_ctl);
4569 btrfs_release_block_group(block_group, ffe_ctl->delalloc);
4570 break;
4571 loop:
4572 if (!ffe_ctl->cached && ffe_ctl->loop > LOOP_CACHING_NOWAIT &&
4573 !ffe_ctl->retry_uncached) {
4574 ffe_ctl->retry_uncached = true;
4575 btrfs_wait_block_group_cache_progress(block_group,
4576 ffe_ctl->num_bytes +
4577 ffe_ctl->empty_cluster +
4578 ffe_ctl->empty_size);
4579 goto have_block_group;
4580 }
4581 release_block_group(block_group, ffe_ctl, ffe_ctl->delalloc);
4582 cond_resched();
4583 }
4584 up_read(&space_info->groups_sem);
4585
4586 ret = find_free_extent_update_loop(fs_info, ins, ffe_ctl, full_search);
4587 if (ret > 0)
4588 goto search;
4589
4590 if (ret == -ENOSPC && !cache_block_group_error) {
4591 /*
4592 * Use ffe_ctl->total_free_space as fallback if we can't find
4593 * any contiguous hole.
4594 */
4595 if (!ffe_ctl->max_extent_size)
4596 ffe_ctl->max_extent_size = ffe_ctl->total_free_space;
4597 spin_lock(&space_info->lock);
4598 space_info->max_extent_size = ffe_ctl->max_extent_size;
4599 spin_unlock(&space_info->lock);
4600 ins->offset = ffe_ctl->max_extent_size;
4601 } else if (ret == -ENOSPC) {
4602 ret = cache_block_group_error;
4603 }
4604 return ret;
4605 }
4606
4607 /*
4608 * Entry point to the extent allocator. Tries to find a hole that is at least
4609 * as big as @num_bytes.
4610 *
4611 * @root - The root that will contain this extent
4612 *
4613 * @ram_bytes - The amount of space in ram that @num_bytes take. This
4614 * is used for accounting purposes. This value differs
4615 * from @num_bytes only in the case of compressed extents.
4616 *
4617 * @num_bytes - Number of bytes to allocate on-disk.
4618 *
4619 * @min_alloc_size - Indicates the minimum amount of space that the
4620 * allocator should try to satisfy. In some cases
4621 * @num_bytes may be larger than what is required and if
4622 * the filesystem is fragmented then allocation fails.
4623 * However, the presence of @min_alloc_size gives a
4624 * chance to try and satisfy the smaller allocation.
4625 *
4626 * @empty_size - A hint that you plan on doing more COW. This is the
4627 * size in bytes the allocator should try to find free
4628 * next to the block it returns. This is just a hint and
4629 * may be ignored by the allocator.
4630 *
4631 * @hint_byte - Hint to the allocator to start searching above the byte
4632 * address passed. It might be ignored.
4633 *
4634 * @ins - This key is modified to record the found hole. It will
4635 * have the following values:
4636 * ins->objectid == start position
4637 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4638 * ins->offset == the size of the hole.
4639 *
4640 * @is_data - Boolean flag indicating whether an extent is
4641 * allocated for data (true) or metadata (false)
4642 *
4643 * @delalloc - Boolean flag indicating whether this allocation is for
4644 * delalloc or not. If 'true' data_rwsem of block groups
4645 * is going to be acquired.
4646 *
4647 *
4648 * Returns 0 when an allocation succeeded or < 0 when an error occurred. In
4649 * case -ENOSPC is returned then @ins->offset will contain the size of the
4650 * largest available hole the allocator managed to find.
4651 */
4652 int btrfs_reserve_extent(struct btrfs_root *root, u64 ram_bytes,
4653 u64 num_bytes, u64 min_alloc_size,
4654 u64 empty_size, u64 hint_byte,
4655 struct btrfs_key *ins, int is_data, int delalloc)
4656 {
4657 struct btrfs_fs_info *fs_info = root->fs_info;
4658 struct find_free_extent_ctl ffe_ctl = {};
4659 bool final_tried = num_bytes == min_alloc_size;
4660 u64 flags;
4661 int ret;
4662 bool for_treelog = (btrfs_root_id(root) == BTRFS_TREE_LOG_OBJECTID);
4663 bool for_data_reloc = (btrfs_is_data_reloc_root(root) && is_data);
4664
4665 flags = get_alloc_profile_by_root(root, is_data);
4666 again:
4667 WARN_ON(num_bytes < fs_info->sectorsize);
4668
4669 ffe_ctl.ram_bytes = ram_bytes;
4670 ffe_ctl.num_bytes = num_bytes;
4671 ffe_ctl.min_alloc_size = min_alloc_size;
4672 ffe_ctl.empty_size = empty_size;
4673 ffe_ctl.flags = flags;
4674 ffe_ctl.delalloc = delalloc;
4675 ffe_ctl.hint_byte = hint_byte;
4676 ffe_ctl.for_treelog = for_treelog;
4677 ffe_ctl.for_data_reloc = for_data_reloc;
4678
4679 ret = find_free_extent(root, ins, &ffe_ctl);
4680 if (!ret && !is_data) {
4681 btrfs_dec_block_group_reservations(fs_info, ins->objectid);
4682 } else if (ret == -ENOSPC) {
4683 if (!final_tried && ins->offset) {
4684 num_bytes = min(num_bytes >> 1, ins->offset);
4685 num_bytes = round_down(num_bytes,
4686 fs_info->sectorsize);
4687 num_bytes = max(num_bytes, min_alloc_size);
4688 ram_bytes = num_bytes;
4689 if (num_bytes == min_alloc_size)
4690 final_tried = true;
4691 goto again;
4692 } else if (btrfs_test_opt(fs_info, ENOSPC_DEBUG)) {
4693 struct btrfs_space_info *sinfo;
4694
4695 sinfo = btrfs_find_space_info(fs_info, flags);
4696 btrfs_err(fs_info,
4697 "allocation failed flags %llu, wanted %llu tree-log %d, relocation: %d",
4698 flags, num_bytes, for_treelog, for_data_reloc);
4699 if (sinfo)
4700 btrfs_dump_space_info(fs_info, sinfo,
4701 num_bytes, 1);
4702 }
4703 }
4704
4705 return ret;
4706 }
4707
4708 int btrfs_free_reserved_extent(struct btrfs_fs_info *fs_info,
4709 u64 start, u64 len, int delalloc)
4710 {
4711 struct btrfs_block_group *cache;
4712
4713 cache = btrfs_lookup_block_group(fs_info, start);
4714 if (!cache) {
4715 btrfs_err(fs_info, "Unable to find block group for %llu",
4716 start);
4717 return -ENOSPC;
4718 }
4719
4720 btrfs_add_free_space(cache, start, len);
4721 btrfs_free_reserved_bytes(cache, len, delalloc);
4722 trace_btrfs_reserved_extent_free(fs_info, start, len);
4723
4724 btrfs_put_block_group(cache);
4725 return 0;
4726 }
4727
4728 int btrfs_pin_reserved_extent(struct btrfs_trans_handle *trans,
4729 const struct extent_buffer *eb)
4730 {
4731 struct btrfs_block_group *cache;
4732 int ret = 0;
4733
4734 cache = btrfs_lookup_block_group(trans->fs_info, eb->start);
4735 if (!cache) {
4736 btrfs_err(trans->fs_info, "unable to find block group for %llu",
4737 eb->start);
4738 return -ENOSPC;
4739 }
4740
4741 ret = pin_down_extent(trans, cache, eb->start, eb->len, 1);
4742 btrfs_put_block_group(cache);
4743 return ret;
4744 }
4745
4746 static int alloc_reserved_extent(struct btrfs_trans_handle *trans, u64 bytenr,
4747 u64 num_bytes)
4748 {
4749 struct btrfs_fs_info *fs_info = trans->fs_info;
4750 int ret;
4751
4752 ret = remove_from_free_space_tree(trans, bytenr, num_bytes);
4753 if (ret)
4754 return ret;
4755
4756 ret = btrfs_update_block_group(trans, bytenr, num_bytes, true);
4757 if (ret) {
4758 ASSERT(!ret);
4759 btrfs_err(fs_info, "update block group failed for %llu %llu",
4760 bytenr, num_bytes);
4761 return ret;
4762 }
4763
4764 trace_btrfs_reserved_extent_alloc(fs_info, bytenr, num_bytes);
4765 return 0;
4766 }
4767
4768 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4769 u64 parent, u64 root_objectid,
4770 u64 flags, u64 owner, u64 offset,
4771 struct btrfs_key *ins, int ref_mod, u64 oref_root)
4772 {
4773 struct btrfs_fs_info *fs_info = trans->fs_info;
4774 struct btrfs_root *extent_root;
4775 int ret;
4776 struct btrfs_extent_item *extent_item;
4777 struct btrfs_extent_owner_ref *oref;
4778 struct btrfs_extent_inline_ref *iref;
4779 struct btrfs_path *path;
4780 struct extent_buffer *leaf;
4781 int type;
4782 u32 size;
4783 const bool simple_quota = (btrfs_qgroup_mode(fs_info) == BTRFS_QGROUP_MODE_SIMPLE);
4784
4785 if (parent > 0)
4786 type = BTRFS_SHARED_DATA_REF_KEY;
4787 else
4788 type = BTRFS_EXTENT_DATA_REF_KEY;
4789
4790 size = sizeof(*extent_item);
4791 if (simple_quota)
4792 size += btrfs_extent_inline_ref_size(BTRFS_EXTENT_OWNER_REF_KEY);
4793 size += btrfs_extent_inline_ref_size(type);
4794
4795 path = btrfs_alloc_path();
4796 if (!path)
4797 return -ENOMEM;
4798
4799 extent_root = btrfs_extent_root(fs_info, ins->objectid);
4800 ret = btrfs_insert_empty_item(trans, extent_root, path, ins, size);
4801 if (ret) {
4802 btrfs_free_path(path);
4803 return ret;
4804 }
4805
4806 leaf = path->nodes[0];
4807 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4808 struct btrfs_extent_item);
4809 btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4810 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4811 btrfs_set_extent_flags(leaf, extent_item,
4812 flags | BTRFS_EXTENT_FLAG_DATA);
4813
4814 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4815 if (simple_quota) {
4816 btrfs_set_extent_inline_ref_type(leaf, iref, BTRFS_EXTENT_OWNER_REF_KEY);
4817 oref = (struct btrfs_extent_owner_ref *)(&iref->offset);
4818 btrfs_set_extent_owner_ref_root_id(leaf, oref, oref_root);
4819 iref = (struct btrfs_extent_inline_ref *)(oref + 1);
4820 }
4821 btrfs_set_extent_inline_ref_type(leaf, iref, type);
4822
4823 if (parent > 0) {
4824 struct btrfs_shared_data_ref *ref;
4825 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4826 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4827 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4828 } else {
4829 struct btrfs_extent_data_ref *ref;
4830 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4831 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4832 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4833 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4834 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4835 }
4836
4837 btrfs_mark_buffer_dirty(trans, path->nodes[0]);
4838 btrfs_free_path(path);
4839
4840 return alloc_reserved_extent(trans, ins->objectid, ins->offset);
4841 }
4842
4843 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4844 struct btrfs_delayed_ref_node *node,
4845 struct btrfs_delayed_extent_op *extent_op)
4846 {
4847 struct btrfs_fs_info *fs_info = trans->fs_info;
4848 struct btrfs_root *extent_root;
4849 int ret;
4850 struct btrfs_extent_item *extent_item;
4851 struct btrfs_key extent_key;
4852 struct btrfs_tree_block_info *block_info;
4853 struct btrfs_extent_inline_ref *iref;
4854 struct btrfs_path *path;
4855 struct extent_buffer *leaf;
4856 u32 size = sizeof(*extent_item) + sizeof(*iref);
4857 const u64 flags = (extent_op ? extent_op->flags_to_set : 0);
4858 /* The owner of a tree block is the level. */
4859 int level = btrfs_delayed_ref_owner(node);
4860 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
4861
4862 extent_key.objectid = node->bytenr;
4863 if (skinny_metadata) {
4864 /* The owner of a tree block is the level. */
4865 extent_key.offset = level;
4866 extent_key.type = BTRFS_METADATA_ITEM_KEY;
4867 } else {
4868 extent_key.offset = node->num_bytes;
4869 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4870 size += sizeof(*block_info);
4871 }
4872
4873 path = btrfs_alloc_path();
4874 if (!path)
4875 return -ENOMEM;
4876
4877 extent_root = btrfs_extent_root(fs_info, extent_key.objectid);
4878 ret = btrfs_insert_empty_item(trans, extent_root, path, &extent_key,
4879 size);
4880 if (ret) {
4881 btrfs_free_path(path);
4882 return ret;
4883 }
4884
4885 leaf = path->nodes[0];
4886 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4887 struct btrfs_extent_item);
4888 btrfs_set_extent_refs(leaf, extent_item, 1);
4889 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4890 btrfs_set_extent_flags(leaf, extent_item,
4891 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4892
4893 if (skinny_metadata) {
4894 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4895 } else {
4896 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4897 btrfs_set_tree_block_key(leaf, block_info, &extent_op->key);
4898 btrfs_set_tree_block_level(leaf, block_info, level);
4899 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4900 }
4901
4902 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) {
4903 btrfs_set_extent_inline_ref_type(leaf, iref,
4904 BTRFS_SHARED_BLOCK_REF_KEY);
4905 btrfs_set_extent_inline_ref_offset(leaf, iref, node->parent);
4906 } else {
4907 btrfs_set_extent_inline_ref_type(leaf, iref,
4908 BTRFS_TREE_BLOCK_REF_KEY);
4909 btrfs_set_extent_inline_ref_offset(leaf, iref, node->ref_root);
4910 }
4911
4912 btrfs_mark_buffer_dirty(trans, leaf);
4913 btrfs_free_path(path);
4914
4915 return alloc_reserved_extent(trans, node->bytenr, fs_info->nodesize);
4916 }
4917
4918 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4919 struct btrfs_root *root, u64 owner,
4920 u64 offset, u64 ram_bytes,
4921 struct btrfs_key *ins)
4922 {
4923 struct btrfs_ref generic_ref = {
4924 .action = BTRFS_ADD_DELAYED_EXTENT,
4925 .bytenr = ins->objectid,
4926 .num_bytes = ins->offset,
4927 .owning_root = btrfs_root_id(root),
4928 .ref_root = btrfs_root_id(root),
4929 };
4930
4931 ASSERT(generic_ref.ref_root != BTRFS_TREE_LOG_OBJECTID);
4932
4933 if (btrfs_is_data_reloc_root(root) && is_fstree(root->relocation_src_root))
4934 generic_ref.owning_root = root->relocation_src_root;
4935
4936 btrfs_init_data_ref(&generic_ref, owner, offset, 0, false);
4937 btrfs_ref_tree_mod(root->fs_info, &generic_ref);
4938
4939 return btrfs_add_delayed_data_ref(trans, &generic_ref, ram_bytes);
4940 }
4941
4942 /*
4943 * this is used by the tree logging recovery code. It records that
4944 * an extent has been allocated and makes sure to clear the free
4945 * space cache bits as well
4946 */
4947 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4948 u64 root_objectid, u64 owner, u64 offset,
4949 struct btrfs_key *ins)
4950 {
4951 struct btrfs_fs_info *fs_info = trans->fs_info;
4952 int ret;
4953 struct btrfs_block_group *block_group;
4954 struct btrfs_space_info *space_info;
4955 struct btrfs_squota_delta delta = {
4956 .root = root_objectid,
4957 .num_bytes = ins->offset,
4958 .generation = trans->transid,
4959 .is_data = true,
4960 .is_inc = true,
4961 };
4962
4963 /*
4964 * Mixed block groups will exclude before processing the log so we only
4965 * need to do the exclude dance if this fs isn't mixed.
4966 */
4967 if (!btrfs_fs_incompat(fs_info, MIXED_GROUPS)) {
4968 ret = __exclude_logged_extent(fs_info, ins->objectid,
4969 ins->offset);
4970 if (ret)
4971 return ret;
4972 }
4973
4974 block_group = btrfs_lookup_block_group(fs_info, ins->objectid);
4975 if (!block_group)
4976 return -EINVAL;
4977
4978 space_info = block_group->space_info;
4979 spin_lock(&space_info->lock);
4980 spin_lock(&block_group->lock);
4981 space_info->bytes_reserved += ins->offset;
4982 block_group->reserved += ins->offset;
4983 spin_unlock(&block_group->lock);
4984 spin_unlock(&space_info->lock);
4985
4986 ret = alloc_reserved_file_extent(trans, 0, root_objectid, 0, owner,
4987 offset, ins, 1, root_objectid);
4988 if (ret)
4989 btrfs_pin_extent(trans, ins->objectid, ins->offset, 1);
4990 ret = btrfs_record_squota_delta(fs_info, &delta);
4991 btrfs_put_block_group(block_group);
4992 return ret;
4993 }
4994
4995 #ifdef CONFIG_BTRFS_DEBUG
4996 /*
4997 * Extra safety check in case the extent tree is corrupted and extent allocator
4998 * chooses to use a tree block which is already used and locked.
4999 */
5000 static bool check_eb_lock_owner(const struct extent_buffer *eb)
5001 {
5002 if (eb->lock_owner == current->pid) {
5003 btrfs_err_rl(eb->fs_info,
5004 "tree block %llu owner %llu already locked by pid=%d, extent tree corruption detected",
5005 eb->start, btrfs_header_owner(eb), current->pid);
5006 return true;
5007 }
5008 return false;
5009 }
5010 #else
5011 static bool check_eb_lock_owner(struct extent_buffer *eb)
5012 {
5013 return false;
5014 }
5015 #endif
5016
5017 static struct extent_buffer *
5018 btrfs_init_new_buffer(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5019 u64 bytenr, int level, u64 owner,
5020 enum btrfs_lock_nesting nest)
5021 {
5022 struct btrfs_fs_info *fs_info = root->fs_info;
5023 struct extent_buffer *buf;
5024 u64 lockdep_owner = owner;
5025
5026 buf = btrfs_find_create_tree_block(fs_info, bytenr, owner, level);
5027 if (IS_ERR(buf))
5028 return buf;
5029
5030 if (check_eb_lock_owner(buf)) {
5031 free_extent_buffer(buf);
5032 return ERR_PTR(-EUCLEAN);
5033 }
5034
5035 /*
5036 * The reloc trees are just snapshots, so we need them to appear to be
5037 * just like any other fs tree WRT lockdep.
5038 *
5039 * The exception however is in replace_path() in relocation, where we
5040 * hold the lock on the original fs root and then search for the reloc
5041 * root. At that point we need to make sure any reloc root buffers are
5042 * set to the BTRFS_TREE_RELOC_OBJECTID lockdep class in order to make
5043 * lockdep happy.
5044 */
5045 if (lockdep_owner == BTRFS_TREE_RELOC_OBJECTID &&
5046 !test_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &root->state))
5047 lockdep_owner = BTRFS_FS_TREE_OBJECTID;
5048
5049 /* btrfs_clear_buffer_dirty() accesses generation field. */
5050 btrfs_set_header_generation(buf, trans->transid);
5051
5052 /*
5053 * This needs to stay, because we could allocate a freed block from an
5054 * old tree into a new tree, so we need to make sure this new block is
5055 * set to the appropriate level and owner.
5056 */
5057 btrfs_set_buffer_lockdep_class(lockdep_owner, buf, level);
5058
5059 btrfs_tree_lock_nested(buf, nest);
5060 btrfs_clear_buffer_dirty(trans, buf);
5061 clear_bit(EXTENT_BUFFER_STALE, &buf->bflags);
5062 clear_bit(EXTENT_BUFFER_ZONED_ZEROOUT, &buf->bflags);
5063
5064 set_extent_buffer_uptodate(buf);
5065
5066 memzero_extent_buffer(buf, 0, sizeof(struct btrfs_header));
5067 btrfs_set_header_level(buf, level);
5068 btrfs_set_header_bytenr(buf, buf->start);
5069 btrfs_set_header_generation(buf, trans->transid);
5070 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
5071 btrfs_set_header_owner(buf, owner);
5072 write_extent_buffer_fsid(buf, fs_info->fs_devices->metadata_uuid);
5073 write_extent_buffer_chunk_tree_uuid(buf, fs_info->chunk_tree_uuid);
5074 if (btrfs_root_id(root) == BTRFS_TREE_LOG_OBJECTID) {
5075 buf->log_index = root->log_transid % 2;
5076 /*
5077 * we allow two log transactions at a time, use different
5078 * EXTENT bit to differentiate dirty pages.
5079 */
5080 if (buf->log_index == 0)
5081 set_extent_bit(&root->dirty_log_pages, buf->start,
5082 buf->start + buf->len - 1,
5083 EXTENT_DIRTY, NULL);
5084 else
5085 set_extent_bit(&root->dirty_log_pages, buf->start,
5086 buf->start + buf->len - 1,
5087 EXTENT_NEW, NULL);
5088 } else {
5089 buf->log_index = -1;
5090 set_extent_bit(&trans->transaction->dirty_pages, buf->start,
5091 buf->start + buf->len - 1, EXTENT_DIRTY, NULL);
5092 }
5093 /* this returns a buffer locked for blocking */
5094 return buf;
5095 }
5096
5097 /*
5098 * finds a free extent and does all the dirty work required for allocation
5099 * returns the tree buffer or an ERR_PTR on error.
5100 */
5101 struct extent_buffer *btrfs_alloc_tree_block(struct btrfs_trans_handle *trans,
5102 struct btrfs_root *root,
5103 u64 parent, u64 root_objectid,
5104 const struct btrfs_disk_key *key,
5105 int level, u64 hint,
5106 u64 empty_size,
5107 u64 reloc_src_root,
5108 enum btrfs_lock_nesting nest)
5109 {
5110 struct btrfs_fs_info *fs_info = root->fs_info;
5111 struct btrfs_key ins;
5112 struct btrfs_block_rsv *block_rsv;
5113 struct extent_buffer *buf;
5114 u64 flags = 0;
5115 int ret;
5116 u32 blocksize = fs_info->nodesize;
5117 bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
5118 u64 owning_root;
5119
5120 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
5121 if (btrfs_is_testing(fs_info)) {
5122 buf = btrfs_init_new_buffer(trans, root, root->alloc_bytenr,
5123 level, root_objectid, nest);
5124 if (!IS_ERR(buf))
5125 root->alloc_bytenr += blocksize;
5126 return buf;
5127 }
5128 #endif
5129
5130 block_rsv = btrfs_use_block_rsv(trans, root, blocksize);
5131 if (IS_ERR(block_rsv))
5132 return ERR_CAST(block_rsv);
5133
5134 ret = btrfs_reserve_extent(root, blocksize, blocksize, blocksize,
5135 empty_size, hint, &ins, 0, 0);
5136 if (ret)
5137 goto out_unuse;
5138
5139 buf = btrfs_init_new_buffer(trans, root, ins.objectid, level,
5140 root_objectid, nest);
5141 if (IS_ERR(buf)) {
5142 ret = PTR_ERR(buf);
5143 goto out_free_reserved;
5144 }
5145 owning_root = btrfs_header_owner(buf);
5146
5147 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
5148 if (parent == 0)
5149 parent = ins.objectid;
5150 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
5151 owning_root = reloc_src_root;
5152 } else
5153 BUG_ON(parent > 0);
5154
5155 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
5156 struct btrfs_delayed_extent_op *extent_op;
5157 struct btrfs_ref generic_ref = {
5158 .action = BTRFS_ADD_DELAYED_EXTENT,
5159 .bytenr = ins.objectid,
5160 .num_bytes = ins.offset,
5161 .parent = parent,
5162 .owning_root = owning_root,
5163 .ref_root = root_objectid,
5164 };
5165
5166 if (!skinny_metadata || flags != 0) {
5167 extent_op = btrfs_alloc_delayed_extent_op();
5168 if (!extent_op) {
5169 ret = -ENOMEM;
5170 goto out_free_buf;
5171 }
5172 if (key)
5173 memcpy(&extent_op->key, key, sizeof(extent_op->key));
5174 else
5175 memset(&extent_op->key, 0, sizeof(extent_op->key));
5176 extent_op->flags_to_set = flags;
5177 extent_op->update_key = (skinny_metadata ? false : true);
5178 extent_op->update_flags = (flags != 0);
5179 } else {
5180 extent_op = NULL;
5181 }
5182
5183 btrfs_init_tree_ref(&generic_ref, level, btrfs_root_id(root), false);
5184 btrfs_ref_tree_mod(fs_info, &generic_ref);
5185 ret = btrfs_add_delayed_tree_ref(trans, &generic_ref, extent_op);
5186 if (ret) {
5187 btrfs_free_delayed_extent_op(extent_op);
5188 goto out_free_buf;
5189 }
5190 }
5191 return buf;
5192
5193 out_free_buf:
5194 btrfs_tree_unlock(buf);
5195 free_extent_buffer(buf);
5196 out_free_reserved:
5197 btrfs_free_reserved_extent(fs_info, ins.objectid, ins.offset, 0);
5198 out_unuse:
5199 btrfs_unuse_block_rsv(fs_info, block_rsv, blocksize);
5200 return ERR_PTR(ret);
5201 }
5202
5203 struct walk_control {
5204 u64 refs[BTRFS_MAX_LEVEL];
5205 u64 flags[BTRFS_MAX_LEVEL];
5206 struct btrfs_key update_progress;
5207 struct btrfs_key drop_progress;
5208 int drop_level;
5209 int stage;
5210 int level;
5211 int shared_level;
5212 int update_ref;
5213 int keep_locks;
5214 int reada_slot;
5215 int reada_count;
5216 int restarted;
5217 /* Indicate that extent info needs to be looked up when walking the tree. */
5218 int lookup_info;
5219 };
5220
5221 /*
5222 * This is our normal stage. We are traversing blocks the current snapshot owns
5223 * and we are dropping any of our references to any children we are able to, and
5224 * then freeing the block once we've processed all of the children.
5225 */
5226 #define DROP_REFERENCE 1
5227
5228 /*
5229 * We enter this stage when we have to walk into a child block (meaning we can't
5230 * simply drop our reference to it from our current parent node) and there are
5231 * more than one reference on it. If we are the owner of any of the children
5232 * blocks from the current parent node then we have to do the FULL_BACKREF dance
5233 * on them in order to drop our normal ref and add the shared ref.
5234 */
5235 #define UPDATE_BACKREF 2
5236
5237 /*
5238 * Decide if we need to walk down into this node to adjust the references.
5239 *
5240 * @root: the root we are currently deleting
5241 * @wc: the walk control for this deletion
5242 * @eb: the parent eb that we're currently visiting
5243 * @refs: the number of refs for wc->level - 1
5244 * @flags: the flags for wc->level - 1
5245 * @slot: the slot in the eb that we're currently checking
5246 *
5247 * This is meant to be called when we're evaluating if a node we point to at
5248 * wc->level should be read and walked into, or if we can simply delete our
5249 * reference to it. We return true if we should walk into the node, false if we
5250 * can skip it.
5251 *
5252 * We have assertions in here to make sure this is called correctly. We assume
5253 * that sanity checking on the blocks read to this point has been done, so any
5254 * corrupted file systems must have been caught before calling this function.
5255 */
5256 static bool visit_node_for_delete(struct btrfs_root *root, struct walk_control *wc,
5257 struct extent_buffer *eb, u64 refs, u64 flags, int slot)
5258 {
5259 struct btrfs_key key;
5260 u64 generation;
5261 int level = wc->level;
5262
5263 ASSERT(level > 0);
5264 ASSERT(wc->refs[level - 1] > 0);
5265
5266 /*
5267 * The update backref stage we only want to skip if we already have
5268 * FULL_BACKREF set, otherwise we need to read.
5269 */
5270 if (wc->stage == UPDATE_BACKREF) {
5271 if (level == 1 && flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5272 return false;
5273 return true;
5274 }
5275
5276 /*
5277 * We're the last ref on this block, we must walk into it and process
5278 * any refs it's pointing at.
5279 */
5280 if (wc->refs[level - 1] == 1)
5281 return true;
5282
5283 /*
5284 * If we're already FULL_BACKREF then we know we can just drop our
5285 * current reference.
5286 */
5287 if (level == 1 && flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5288 return false;
5289
5290 /*
5291 * This block is older than our creation generation, we can drop our
5292 * reference to it.
5293 */
5294 generation = btrfs_node_ptr_generation(eb, slot);
5295 if (!wc->update_ref || generation <= root->root_key.offset)
5296 return false;
5297
5298 /*
5299 * This block was processed from a previous snapshot deletion run, we
5300 * can skip it.
5301 */
5302 btrfs_node_key_to_cpu(eb, &key, slot);
5303 if (btrfs_comp_cpu_keys(&key, &wc->update_progress) < 0)
5304 return false;
5305
5306 /* All other cases we need to wander into the node. */
5307 return true;
5308 }
5309
5310 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
5311 struct btrfs_root *root,
5312 struct walk_control *wc,
5313 struct btrfs_path *path)
5314 {
5315 struct btrfs_fs_info *fs_info = root->fs_info;
5316 u64 bytenr;
5317 u64 generation;
5318 u64 refs;
5319 u64 flags;
5320 u32 nritems;
5321 struct extent_buffer *eb;
5322 int ret;
5323 int slot;
5324 int nread = 0;
5325
5326 if (path->slots[wc->level] < wc->reada_slot) {
5327 wc->reada_count = wc->reada_count * 2 / 3;
5328 wc->reada_count = max(wc->reada_count, 2);
5329 } else {
5330 wc->reada_count = wc->reada_count * 3 / 2;
5331 wc->reada_count = min_t(int, wc->reada_count,
5332 BTRFS_NODEPTRS_PER_BLOCK(fs_info));
5333 }
5334
5335 eb = path->nodes[wc->level];
5336 nritems = btrfs_header_nritems(eb);
5337
5338 for (slot = path->slots[wc->level]; slot < nritems; slot++) {
5339 if (nread >= wc->reada_count)
5340 break;
5341
5342 cond_resched();
5343 bytenr = btrfs_node_blockptr(eb, slot);
5344 generation = btrfs_node_ptr_generation(eb, slot);
5345
5346 if (slot == path->slots[wc->level])
5347 goto reada;
5348
5349 if (wc->stage == UPDATE_BACKREF &&
5350 generation <= root->root_key.offset)
5351 continue;
5352
5353 /* We don't lock the tree block, it's OK to be racy here */
5354 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr,
5355 wc->level - 1, 1, &refs,
5356 &flags, NULL);
5357 /* We don't care about errors in readahead. */
5358 if (ret < 0)
5359 continue;
5360
5361 /*
5362 * This could be racey, it's conceivable that we raced and end
5363 * up with a bogus refs count, if that's the case just skip, if
5364 * we are actually corrupt we will notice when we look up
5365 * everything again with our locks.
5366 */
5367 if (refs == 0)
5368 continue;
5369
5370 /* If we don't need to visit this node don't reada. */
5371 if (!visit_node_for_delete(root, wc, eb, refs, flags, slot))
5372 continue;
5373 reada:
5374 btrfs_readahead_node_child(eb, slot);
5375 nread++;
5376 }
5377 wc->reada_slot = slot;
5378 }
5379
5380 /*
5381 * helper to process tree block while walking down the tree.
5382 *
5383 * when wc->stage == UPDATE_BACKREF, this function updates
5384 * back refs for pointers in the block.
5385 *
5386 * NOTE: return value 1 means we should stop walking down.
5387 */
5388 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
5389 struct btrfs_root *root,
5390 struct btrfs_path *path,
5391 struct walk_control *wc)
5392 {
5393 struct btrfs_fs_info *fs_info = root->fs_info;
5394 int level = wc->level;
5395 struct extent_buffer *eb = path->nodes[level];
5396 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5397 int ret;
5398
5399 if (wc->stage == UPDATE_BACKREF && btrfs_header_owner(eb) != btrfs_root_id(root))
5400 return 1;
5401
5402 /*
5403 * when reference count of tree block is 1, it won't increase
5404 * again. once full backref flag is set, we never clear it.
5405 */
5406 if (wc->lookup_info &&
5407 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
5408 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
5409 ASSERT(path->locks[level]);
5410 ret = btrfs_lookup_extent_info(trans, fs_info,
5411 eb->start, level, 1,
5412 &wc->refs[level],
5413 &wc->flags[level],
5414 NULL);
5415 if (ret)
5416 return ret;
5417 if (unlikely(wc->refs[level] == 0)) {
5418 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0",
5419 eb->start);
5420 return -EUCLEAN;
5421 }
5422 }
5423
5424 if (wc->stage == DROP_REFERENCE) {
5425 if (wc->refs[level] > 1)
5426 return 1;
5427
5428 if (path->locks[level] && !wc->keep_locks) {
5429 btrfs_tree_unlock_rw(eb, path->locks[level]);
5430 path->locks[level] = 0;
5431 }
5432 return 0;
5433 }
5434
5435 /* wc->stage == UPDATE_BACKREF */
5436 if (!(wc->flags[level] & flag)) {
5437 ASSERT(path->locks[level]);
5438 ret = btrfs_inc_ref(trans, root, eb, 1);
5439 if (ret) {
5440 btrfs_abort_transaction(trans, ret);
5441 return ret;
5442 }
5443 ret = btrfs_dec_ref(trans, root, eb, 0);
5444 if (ret) {
5445 btrfs_abort_transaction(trans, ret);
5446 return ret;
5447 }
5448 ret = btrfs_set_disk_extent_flags(trans, eb, flag);
5449 if (ret) {
5450 btrfs_abort_transaction(trans, ret);
5451 return ret;
5452 }
5453 wc->flags[level] |= flag;
5454 }
5455
5456 /*
5457 * the block is shared by multiple trees, so it's not good to
5458 * keep the tree lock
5459 */
5460 if (path->locks[level] && level > 0) {
5461 btrfs_tree_unlock_rw(eb, path->locks[level]);
5462 path->locks[level] = 0;
5463 }
5464 return 0;
5465 }
5466
5467 /*
5468 * This is used to verify a ref exists for this root to deal with a bug where we
5469 * would have a drop_progress key that hadn't been updated properly.
5470 */
5471 static int check_ref_exists(struct btrfs_trans_handle *trans,
5472 struct btrfs_root *root, u64 bytenr, u64 parent,
5473 int level)
5474 {
5475 struct btrfs_path *path;
5476 struct btrfs_extent_inline_ref *iref;
5477 int ret;
5478
5479 path = btrfs_alloc_path();
5480 if (!path)
5481 return -ENOMEM;
5482
5483 ret = lookup_extent_backref(trans, path, &iref, bytenr,
5484 root->fs_info->nodesize, parent,
5485 btrfs_root_id(root), level, 0);
5486 btrfs_free_path(path);
5487 if (ret == -ENOENT)
5488 return 0;
5489 if (ret < 0)
5490 return ret;
5491 return 1;
5492 }
5493
5494 /*
5495 * We may not have an uptodate block, so if we are going to walk down into this
5496 * block we need to drop the lock, read it off of the disk, re-lock it and
5497 * return to continue dropping the snapshot.
5498 */
5499 static int check_next_block_uptodate(struct btrfs_trans_handle *trans,
5500 struct btrfs_root *root,
5501 struct btrfs_path *path,
5502 struct walk_control *wc,
5503 struct extent_buffer *next)
5504 {
5505 struct btrfs_tree_parent_check check = { 0 };
5506 u64 generation;
5507 int level = wc->level;
5508 int ret;
5509
5510 btrfs_assert_tree_write_locked(next);
5511
5512 generation = btrfs_node_ptr_generation(path->nodes[level], path->slots[level]);
5513
5514 if (btrfs_buffer_uptodate(next, generation, 0))
5515 return 0;
5516
5517 check.level = level - 1;
5518 check.transid = generation;
5519 check.owner_root = btrfs_root_id(root);
5520 check.has_first_key = true;
5521 btrfs_node_key_to_cpu(path->nodes[level], &check.first_key, path->slots[level]);
5522
5523 btrfs_tree_unlock(next);
5524 if (level == 1)
5525 reada_walk_down(trans, root, wc, path);
5526 ret = btrfs_read_extent_buffer(next, &check);
5527 if (ret) {
5528 free_extent_buffer(next);
5529 return ret;
5530 }
5531 btrfs_tree_lock(next);
5532 wc->lookup_info = 1;
5533 return 0;
5534 }
5535
5536 /*
5537 * If we determine that we don't have to visit wc->level - 1 then we need to
5538 * determine if we can drop our reference.
5539 *
5540 * If we are UPDATE_BACKREF then we will not, we need to update our backrefs.
5541 *
5542 * If we are DROP_REFERENCE this will figure out if we need to drop our current
5543 * reference, skipping it if we dropped it from a previous incompleted drop, or
5544 * dropping it if we still have a reference to it.
5545 */
5546 static int maybe_drop_reference(struct btrfs_trans_handle *trans, struct btrfs_root *root,
5547 struct btrfs_path *path, struct walk_control *wc,
5548 struct extent_buffer *next, u64 owner_root)
5549 {
5550 struct btrfs_ref ref = {
5551 .action = BTRFS_DROP_DELAYED_REF,
5552 .bytenr = next->start,
5553 .num_bytes = root->fs_info->nodesize,
5554 .owning_root = owner_root,
5555 .ref_root = btrfs_root_id(root),
5556 };
5557 int level = wc->level;
5558 int ret;
5559
5560 /* We are UPDATE_BACKREF, we're not dropping anything. */
5561 if (wc->stage == UPDATE_BACKREF)
5562 return 0;
5563
5564 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5565 ref.parent = path->nodes[level]->start;
5566 } else {
5567 ASSERT(btrfs_root_id(root) == btrfs_header_owner(path->nodes[level]));
5568 if (btrfs_root_id(root) != btrfs_header_owner(path->nodes[level])) {
5569 btrfs_err(root->fs_info, "mismatched block owner");
5570 return -EIO;
5571 }
5572 }
5573
5574 /*
5575 * If we had a drop_progress we need to verify the refs are set as
5576 * expected. If we find our ref then we know that from here on out
5577 * everything should be correct, and we can clear the
5578 * ->restarted flag.
5579 */
5580 if (wc->restarted) {
5581 ret = check_ref_exists(trans, root, next->start, ref.parent,
5582 level - 1);
5583 if (ret <= 0)
5584 return ret;
5585 ret = 0;
5586 wc->restarted = 0;
5587 }
5588
5589 /*
5590 * Reloc tree doesn't contribute to qgroup numbers, and we have already
5591 * accounted them at merge time (replace_path), thus we could skip
5592 * expensive subtree trace here.
5593 */
5594 if (btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID &&
5595 wc->refs[level - 1] > 1) {
5596 u64 generation = btrfs_node_ptr_generation(path->nodes[level],
5597 path->slots[level]);
5598
5599 ret = btrfs_qgroup_trace_subtree(trans, next, generation, level - 1);
5600 if (ret) {
5601 btrfs_err_rl(root->fs_info,
5602 "error %d accounting shared subtree, quota is out of sync, rescan required",
5603 ret);
5604 }
5605 }
5606
5607 /*
5608 * We need to update the next key in our walk control so we can update
5609 * the drop_progress key accordingly. We don't care if find_next_key
5610 * doesn't find a key because that means we're at the end and are going
5611 * to clean up now.
5612 */
5613 wc->drop_level = level;
5614 find_next_key(path, level, &wc->drop_progress);
5615
5616 btrfs_init_tree_ref(&ref, level - 1, 0, false);
5617 return btrfs_free_extent(trans, &ref);
5618 }
5619
5620 /*
5621 * helper to process tree block pointer.
5622 *
5623 * when wc->stage == DROP_REFERENCE, this function checks
5624 * reference count of the block pointed to. if the block
5625 * is shared and we need update back refs for the subtree
5626 * rooted at the block, this function changes wc->stage to
5627 * UPDATE_BACKREF. if the block is shared and there is no
5628 * need to update back, this function drops the reference
5629 * to the block.
5630 *
5631 * NOTE: return value 1 means we should stop walking down.
5632 */
5633 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5634 struct btrfs_root *root,
5635 struct btrfs_path *path,
5636 struct walk_control *wc)
5637 {
5638 struct btrfs_fs_info *fs_info = root->fs_info;
5639 u64 bytenr;
5640 u64 generation;
5641 u64 owner_root = 0;
5642 struct extent_buffer *next;
5643 int level = wc->level;
5644 int ret = 0;
5645
5646 generation = btrfs_node_ptr_generation(path->nodes[level],
5647 path->slots[level]);
5648 /*
5649 * if the lower level block was created before the snapshot
5650 * was created, we know there is no need to update back refs
5651 * for the subtree
5652 */
5653 if (wc->stage == UPDATE_BACKREF &&
5654 generation <= root->root_key.offset) {
5655 wc->lookup_info = 1;
5656 return 1;
5657 }
5658
5659 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
5660
5661 next = btrfs_find_create_tree_block(fs_info, bytenr, btrfs_root_id(root),
5662 level - 1);
5663 if (IS_ERR(next))
5664 return PTR_ERR(next);
5665
5666 btrfs_tree_lock(next);
5667
5668 ret = btrfs_lookup_extent_info(trans, fs_info, bytenr, level - 1, 1,
5669 &wc->refs[level - 1],
5670 &wc->flags[level - 1],
5671 &owner_root);
5672 if (ret < 0)
5673 goto out_unlock;
5674
5675 if (unlikely(wc->refs[level - 1] == 0)) {
5676 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0",
5677 bytenr);
5678 ret = -EUCLEAN;
5679 goto out_unlock;
5680 }
5681 wc->lookup_info = 0;
5682
5683 /* If we don't have to walk into this node skip it. */
5684 if (!visit_node_for_delete(root, wc, path->nodes[level],
5685 wc->refs[level - 1], wc->flags[level - 1],
5686 path->slots[level]))
5687 goto skip;
5688
5689 /*
5690 * We have to walk down into this node, and if we're currently at the
5691 * DROP_REFERNCE stage and this block is shared then we need to switch
5692 * to the UPDATE_BACKREF stage in order to convert to FULL_BACKREF.
5693 */
5694 if (wc->stage == DROP_REFERENCE && wc->refs[level - 1] > 1) {
5695 wc->stage = UPDATE_BACKREF;
5696 wc->shared_level = level - 1;
5697 }
5698
5699 ret = check_next_block_uptodate(trans, root, path, wc, next);
5700 if (ret)
5701 return ret;
5702
5703 level--;
5704 ASSERT(level == btrfs_header_level(next));
5705 if (level != btrfs_header_level(next)) {
5706 btrfs_err(root->fs_info, "mismatched level");
5707 ret = -EIO;
5708 goto out_unlock;
5709 }
5710 path->nodes[level] = next;
5711 path->slots[level] = 0;
5712 path->locks[level] = BTRFS_WRITE_LOCK;
5713 wc->level = level;
5714 if (wc->level == 1)
5715 wc->reada_slot = 0;
5716 return 0;
5717 skip:
5718 ret = maybe_drop_reference(trans, root, path, wc, next, owner_root);
5719 if (ret)
5720 goto out_unlock;
5721 wc->refs[level - 1] = 0;
5722 wc->flags[level - 1] = 0;
5723 wc->lookup_info = 1;
5724 ret = 1;
5725
5726 out_unlock:
5727 btrfs_tree_unlock(next);
5728 free_extent_buffer(next);
5729
5730 return ret;
5731 }
5732
5733 /*
5734 * helper to process tree block while walking up the tree.
5735 *
5736 * when wc->stage == DROP_REFERENCE, this function drops
5737 * reference count on the block.
5738 *
5739 * when wc->stage == UPDATE_BACKREF, this function changes
5740 * wc->stage back to DROP_REFERENCE if we changed wc->stage
5741 * to UPDATE_BACKREF previously while processing the block.
5742 *
5743 * NOTE: return value 1 means we should stop walking up.
5744 */
5745 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5746 struct btrfs_root *root,
5747 struct btrfs_path *path,
5748 struct walk_control *wc)
5749 {
5750 struct btrfs_fs_info *fs_info = root->fs_info;
5751 int ret = 0;
5752 int level = wc->level;
5753 struct extent_buffer *eb = path->nodes[level];
5754 u64 parent = 0;
5755
5756 if (wc->stage == UPDATE_BACKREF) {
5757 ASSERT(wc->shared_level >= level);
5758 if (level < wc->shared_level)
5759 goto out;
5760
5761 ret = find_next_key(path, level + 1, &wc->update_progress);
5762 if (ret > 0)
5763 wc->update_ref = 0;
5764
5765 wc->stage = DROP_REFERENCE;
5766 wc->shared_level = -1;
5767 path->slots[level] = 0;
5768
5769 /*
5770 * check reference count again if the block isn't locked.
5771 * we should start walking down the tree again if reference
5772 * count is one.
5773 */
5774 if (!path->locks[level]) {
5775 ASSERT(level > 0);
5776 btrfs_tree_lock(eb);
5777 path->locks[level] = BTRFS_WRITE_LOCK;
5778
5779 ret = btrfs_lookup_extent_info(trans, fs_info,
5780 eb->start, level, 1,
5781 &wc->refs[level],
5782 &wc->flags[level],
5783 NULL);
5784 if (ret < 0) {
5785 btrfs_tree_unlock_rw(eb, path->locks[level]);
5786 path->locks[level] = 0;
5787 return ret;
5788 }
5789 if (unlikely(wc->refs[level] == 0)) {
5790 btrfs_tree_unlock_rw(eb, path->locks[level]);
5791 btrfs_err(fs_info, "bytenr %llu has 0 references, expect > 0",
5792 eb->start);
5793 return -EUCLEAN;
5794 }
5795 if (wc->refs[level] == 1) {
5796 btrfs_tree_unlock_rw(eb, path->locks[level]);
5797 path->locks[level] = 0;
5798 return 1;
5799 }
5800 }
5801 }
5802
5803 /* wc->stage == DROP_REFERENCE */
5804 ASSERT(path->locks[level] || wc->refs[level] == 1);
5805
5806 if (wc->refs[level] == 1) {
5807 if (level == 0) {
5808 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5809 ret = btrfs_dec_ref(trans, root, eb, 1);
5810 else
5811 ret = btrfs_dec_ref(trans, root, eb, 0);
5812 if (ret) {
5813 btrfs_abort_transaction(trans, ret);
5814 return ret;
5815 }
5816 if (is_fstree(btrfs_root_id(root))) {
5817 ret = btrfs_qgroup_trace_leaf_items(trans, eb);
5818 if (ret) {
5819 btrfs_err_rl(fs_info,
5820 "error %d accounting leaf items, quota is out of sync, rescan required",
5821 ret);
5822 }
5823 }
5824 }
5825 /* Make block locked assertion in btrfs_clear_buffer_dirty happy. */
5826 if (!path->locks[level]) {
5827 btrfs_tree_lock(eb);
5828 path->locks[level] = BTRFS_WRITE_LOCK;
5829 }
5830 btrfs_clear_buffer_dirty(trans, eb);
5831 }
5832
5833 if (eb == root->node) {
5834 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5835 parent = eb->start;
5836 else if (btrfs_root_id(root) != btrfs_header_owner(eb))
5837 goto owner_mismatch;
5838 } else {
5839 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5840 parent = path->nodes[level + 1]->start;
5841 else if (btrfs_root_id(root) !=
5842 btrfs_header_owner(path->nodes[level + 1]))
5843 goto owner_mismatch;
5844 }
5845
5846 ret = btrfs_free_tree_block(trans, btrfs_root_id(root), eb, parent,
5847 wc->refs[level] == 1);
5848 if (ret < 0)
5849 btrfs_abort_transaction(trans, ret);
5850 out:
5851 wc->refs[level] = 0;
5852 wc->flags[level] = 0;
5853 return ret;
5854
5855 owner_mismatch:
5856 btrfs_err_rl(fs_info, "unexpected tree owner, have %llu expect %llu",
5857 btrfs_header_owner(eb), btrfs_root_id(root));
5858 return -EUCLEAN;
5859 }
5860
5861 /*
5862 * walk_down_tree consists of two steps.
5863 *
5864 * walk_down_proc(). Look up the reference count and reference of our current
5865 * wc->level. At this point path->nodes[wc->level] should be populated and
5866 * uptodate, and in most cases should already be locked. If we are in
5867 * DROP_REFERENCE and our refcount is > 1 then we've entered a shared node and
5868 * we can walk back up the tree. If we are UPDATE_BACKREF we have to set
5869 * FULL_BACKREF on this node if it's not already set, and then do the
5870 * FULL_BACKREF conversion dance, which is to drop the root reference and add
5871 * the shared reference to all of this nodes children.
5872 *
5873 * do_walk_down(). This is where we actually start iterating on the children of
5874 * our current path->nodes[wc->level]. For DROP_REFERENCE that means dropping
5875 * our reference to the children that return false from visit_node_for_delete(),
5876 * which has various conditions where we know we can just drop our reference
5877 * without visiting the node. For UPDATE_BACKREF we will skip any children that
5878 * visit_node_for_delete() returns false for, only walking down when necessary.
5879 * The bulk of the work for UPDATE_BACKREF occurs in the walk_up_tree() part of
5880 * snapshot deletion.
5881 */
5882 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5883 struct btrfs_root *root,
5884 struct btrfs_path *path,
5885 struct walk_control *wc)
5886 {
5887 int level = wc->level;
5888 int ret = 0;
5889
5890 wc->lookup_info = 1;
5891 while (level >= 0) {
5892 ret = walk_down_proc(trans, root, path, wc);
5893 if (ret)
5894 break;
5895
5896 if (level == 0)
5897 break;
5898
5899 if (path->slots[level] >=
5900 btrfs_header_nritems(path->nodes[level]))
5901 break;
5902
5903 ret = do_walk_down(trans, root, path, wc);
5904 if (ret > 0) {
5905 path->slots[level]++;
5906 continue;
5907 } else if (ret < 0)
5908 break;
5909 level = wc->level;
5910 }
5911 return (ret == 1) ? 0 : ret;
5912 }
5913
5914 /*
5915 * walk_up_tree() is responsible for making sure we visit every slot on our
5916 * current node, and if we're at the end of that node then we call
5917 * walk_up_proc() on our current node which will do one of a few things based on
5918 * our stage.
5919 *
5920 * UPDATE_BACKREF. If we wc->level is currently less than our wc->shared_level
5921 * then we need to walk back up the tree, and then going back down into the
5922 * other slots via walk_down_tree to update any other children from our original
5923 * wc->shared_level. Once we're at or above our wc->shared_level we can switch
5924 * back to DROP_REFERENCE, lookup the current nodes refs and flags, and carry on.
5925 *
5926 * DROP_REFERENCE. If our refs == 1 then we're going to free this tree block.
5927 * If we're level 0 then we need to btrfs_dec_ref() on all of the data extents
5928 * in our current leaf. After that we call btrfs_free_tree_block() on the
5929 * current node and walk up to the next node to walk down the next slot.
5930 */
5931 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5932 struct btrfs_root *root,
5933 struct btrfs_path *path,
5934 struct walk_control *wc, int max_level)
5935 {
5936 int level = wc->level;
5937 int ret;
5938
5939 path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5940 while (level < max_level && path->nodes[level]) {
5941 wc->level = level;
5942 if (path->slots[level] + 1 <
5943 btrfs_header_nritems(path->nodes[level])) {
5944 path->slots[level]++;
5945 return 0;
5946 } else {
5947 ret = walk_up_proc(trans, root, path, wc);
5948 if (ret > 0)
5949 return 0;
5950 if (ret < 0)
5951 return ret;
5952
5953 if (path->locks[level]) {
5954 btrfs_tree_unlock_rw(path->nodes[level],
5955 path->locks[level]);
5956 path->locks[level] = 0;
5957 }
5958 free_extent_buffer(path->nodes[level]);
5959 path->nodes[level] = NULL;
5960 level++;
5961 }
5962 }
5963 return 1;
5964 }
5965
5966 /*
5967 * drop a subvolume tree.
5968 *
5969 * this function traverses the tree freeing any blocks that only
5970 * referenced by the tree.
5971 *
5972 * when a shared tree block is found. this function decreases its
5973 * reference count by one. if update_ref is true, this function
5974 * also make sure backrefs for the shared block and all lower level
5975 * blocks are properly updated.
5976 *
5977 * If called with for_reloc == 0, may exit early with -EAGAIN
5978 */
5979 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref, int for_reloc)
5980 {
5981 const bool is_reloc_root = (btrfs_root_id(root) == BTRFS_TREE_RELOC_OBJECTID);
5982 struct btrfs_fs_info *fs_info = root->fs_info;
5983 struct btrfs_path *path;
5984 struct btrfs_trans_handle *trans;
5985 struct btrfs_root *tree_root = fs_info->tree_root;
5986 struct btrfs_root_item *root_item = &root->root_item;
5987 struct walk_control *wc;
5988 struct btrfs_key key;
5989 const u64 rootid = btrfs_root_id(root);
5990 int ret = 0;
5991 int level;
5992 bool root_dropped = false;
5993 bool unfinished_drop = false;
5994
5995 btrfs_debug(fs_info, "Drop subvolume %llu", btrfs_root_id(root));
5996
5997 path = btrfs_alloc_path();
5998 if (!path) {
5999 ret = -ENOMEM;
6000 goto out;
6001 }
6002
6003 wc = kzalloc(sizeof(*wc), GFP_NOFS);
6004 if (!wc) {
6005 btrfs_free_path(path);
6006 ret = -ENOMEM;
6007 goto out;
6008 }
6009
6010 /*
6011 * Use join to avoid potential EINTR from transaction start. See
6012 * wait_reserve_ticket and the whole reservation callchain.
6013 */
6014 if (for_reloc)
6015 trans = btrfs_join_transaction(tree_root);
6016 else
6017 trans = btrfs_start_transaction(tree_root, 0);
6018 if (IS_ERR(trans)) {
6019 ret = PTR_ERR(trans);
6020 goto out_free;
6021 }
6022
6023 ret = btrfs_run_delayed_items(trans);
6024 if (ret)
6025 goto out_end_trans;
6026
6027 /*
6028 * This will help us catch people modifying the fs tree while we're
6029 * dropping it. It is unsafe to mess with the fs tree while it's being
6030 * dropped as we unlock the root node and parent nodes as we walk down
6031 * the tree, assuming nothing will change. If something does change
6032 * then we'll have stale information and drop references to blocks we've
6033 * already dropped.
6034 */
6035 set_bit(BTRFS_ROOT_DELETING, &root->state);
6036 unfinished_drop = test_bit(BTRFS_ROOT_UNFINISHED_DROP, &root->state);
6037
6038 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
6039 level = btrfs_header_level(root->node);
6040 path->nodes[level] = btrfs_lock_root_node(root);
6041 path->slots[level] = 0;
6042 path->locks[level] = BTRFS_WRITE_LOCK;
6043 memset(&wc->update_progress, 0,
6044 sizeof(wc->update_progress));
6045 } else {
6046 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
6047 memcpy(&wc->update_progress, &key,
6048 sizeof(wc->update_progress));
6049
6050 level = btrfs_root_drop_level(root_item);
6051 BUG_ON(level == 0);
6052 path->lowest_level = level;
6053 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
6054 path->lowest_level = 0;
6055 if (ret < 0)
6056 goto out_end_trans;
6057
6058 WARN_ON(ret > 0);
6059 ret = 0;
6060
6061 /*
6062 * unlock our path, this is safe because only this
6063 * function is allowed to delete this snapshot
6064 */
6065 btrfs_unlock_up_safe(path, 0);
6066
6067 level = btrfs_header_level(root->node);
6068 while (1) {
6069 btrfs_tree_lock(path->nodes[level]);
6070 path->locks[level] = BTRFS_WRITE_LOCK;
6071
6072 /*
6073 * btrfs_lookup_extent_info() returns 0 for success,
6074 * or < 0 for error.
6075 */
6076 ret = btrfs_lookup_extent_info(trans, fs_info,
6077 path->nodes[level]->start,
6078 level, 1, &wc->refs[level],
6079 &wc->flags[level], NULL);
6080 if (ret < 0)
6081 goto out_end_trans;
6082
6083 BUG_ON(wc->refs[level] == 0);
6084
6085 if (level == btrfs_root_drop_level(root_item))
6086 break;
6087
6088 btrfs_tree_unlock(path->nodes[level]);
6089 path->locks[level] = 0;
6090 WARN_ON(wc->refs[level] != 1);
6091 level--;
6092 }
6093 }
6094
6095 wc->restarted = test_bit(BTRFS_ROOT_DEAD_TREE, &root->state);
6096 wc->level = level;
6097 wc->shared_level = -1;
6098 wc->stage = DROP_REFERENCE;
6099 wc->update_ref = update_ref;
6100 wc->keep_locks = 0;
6101 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
6102
6103 while (1) {
6104
6105 ret = walk_down_tree(trans, root, path, wc);
6106 if (ret < 0) {
6107 btrfs_abort_transaction(trans, ret);
6108 break;
6109 }
6110
6111 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
6112 if (ret < 0) {
6113 btrfs_abort_transaction(trans, ret);
6114 break;
6115 }
6116
6117 if (ret > 0) {
6118 BUG_ON(wc->stage != DROP_REFERENCE);
6119 ret = 0;
6120 break;
6121 }
6122
6123 if (wc->stage == DROP_REFERENCE) {
6124 wc->drop_level = wc->level;
6125 btrfs_node_key_to_cpu(path->nodes[wc->drop_level],
6126 &wc->drop_progress,
6127 path->slots[wc->drop_level]);
6128 }
6129 btrfs_cpu_key_to_disk(&root_item->drop_progress,
6130 &wc->drop_progress);
6131 btrfs_set_root_drop_level(root_item, wc->drop_level);
6132
6133 BUG_ON(wc->level == 0);
6134 if (btrfs_should_end_transaction(trans) ||
6135 (!for_reloc && btrfs_need_cleaner_sleep(fs_info))) {
6136 ret = btrfs_update_root(trans, tree_root,
6137 &root->root_key,
6138 root_item);
6139 if (ret) {
6140 btrfs_abort_transaction(trans, ret);
6141 goto out_end_trans;
6142 }
6143
6144 if (!is_reloc_root)
6145 btrfs_set_last_root_drop_gen(fs_info, trans->transid);
6146
6147 btrfs_end_transaction_throttle(trans);
6148 if (!for_reloc && btrfs_need_cleaner_sleep(fs_info)) {
6149 btrfs_debug(fs_info,
6150 "drop snapshot early exit");
6151 ret = -EAGAIN;
6152 goto out_free;
6153 }
6154
6155 /*
6156 * Use join to avoid potential EINTR from transaction
6157 * start. See wait_reserve_ticket and the whole
6158 * reservation callchain.
6159 */
6160 if (for_reloc)
6161 trans = btrfs_join_transaction(tree_root);
6162 else
6163 trans = btrfs_start_transaction(tree_root, 0);
6164 if (IS_ERR(trans)) {
6165 ret = PTR_ERR(trans);
6166 goto out_free;
6167 }
6168 }
6169 }
6170 btrfs_release_path(path);
6171 if (ret)
6172 goto out_end_trans;
6173
6174 ret = btrfs_del_root(trans, &root->root_key);
6175 if (ret) {
6176 btrfs_abort_transaction(trans, ret);
6177 goto out_end_trans;
6178 }
6179
6180 if (!is_reloc_root) {
6181 ret = btrfs_find_root(tree_root, &root->root_key, path,
6182 NULL, NULL);
6183 if (ret < 0) {
6184 btrfs_abort_transaction(trans, ret);
6185 goto out_end_trans;
6186 } else if (ret > 0) {
6187 ret = 0;
6188 /*
6189 * If we fail to delete the orphan item this time
6190 * around, it'll get picked up the next time.
6191 *
6192 * The most common failure here is just -ENOENT.
6193 */
6194 btrfs_del_orphan_item(trans, tree_root, btrfs_root_id(root));
6195 }
6196 }
6197
6198 /*
6199 * This subvolume is going to be completely dropped, and won't be
6200 * recorded as dirty roots, thus pertrans meta rsv will not be freed at
6201 * commit transaction time. So free it here manually.
6202 */
6203 btrfs_qgroup_convert_reserved_meta(root, INT_MAX);
6204 btrfs_qgroup_free_meta_all_pertrans(root);
6205
6206 if (test_bit(BTRFS_ROOT_IN_RADIX, &root->state))
6207 btrfs_add_dropped_root(trans, root);
6208 else
6209 btrfs_put_root(root);
6210 root_dropped = true;
6211 out_end_trans:
6212 if (!is_reloc_root)
6213 btrfs_set_last_root_drop_gen(fs_info, trans->transid);
6214
6215 btrfs_end_transaction_throttle(trans);
6216 out_free:
6217 kfree(wc);
6218 btrfs_free_path(path);
6219 out:
6220 if (!ret && root_dropped) {
6221 ret = btrfs_qgroup_cleanup_dropped_subvolume(fs_info, rootid);
6222 if (ret < 0)
6223 btrfs_warn_rl(fs_info,
6224 "failed to cleanup qgroup 0/%llu: %d",
6225 rootid, ret);
6226 ret = 0;
6227 }
6228 /*
6229 * We were an unfinished drop root, check to see if there are any
6230 * pending, and if not clear and wake up any waiters.
6231 */
6232 if (!ret && unfinished_drop)
6233 btrfs_maybe_wake_unfinished_drop(fs_info);
6234
6235 /*
6236 * So if we need to stop dropping the snapshot for whatever reason we
6237 * need to make sure to add it back to the dead root list so that we
6238 * keep trying to do the work later. This also cleans up roots if we
6239 * don't have it in the radix (like when we recover after a power fail
6240 * or unmount) so we don't leak memory.
6241 */
6242 if (!for_reloc && !root_dropped)
6243 btrfs_add_dead_root(root);
6244 return ret;
6245 }
6246
6247 /*
6248 * drop subtree rooted at tree block 'node'.
6249 *
6250 * NOTE: this function will unlock and release tree block 'node'
6251 * only used by relocation code
6252 */
6253 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
6254 struct btrfs_root *root,
6255 struct extent_buffer *node,
6256 struct extent_buffer *parent)
6257 {
6258 struct btrfs_fs_info *fs_info = root->fs_info;
6259 struct btrfs_path *path;
6260 struct walk_control *wc;
6261 int level;
6262 int parent_level;
6263 int ret = 0;
6264
6265 BUG_ON(btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID);
6266
6267 path = btrfs_alloc_path();
6268 if (!path)
6269 return -ENOMEM;
6270
6271 wc = kzalloc(sizeof(*wc), GFP_NOFS);
6272 if (!wc) {
6273 btrfs_free_path(path);
6274 return -ENOMEM;
6275 }
6276
6277 btrfs_assert_tree_write_locked(parent);
6278 parent_level = btrfs_header_level(parent);
6279 atomic_inc(&parent->refs);
6280 path->nodes[parent_level] = parent;
6281 path->slots[parent_level] = btrfs_header_nritems(parent);
6282
6283 btrfs_assert_tree_write_locked(node);
6284 level = btrfs_header_level(node);
6285 path->nodes[level] = node;
6286 path->slots[level] = 0;
6287 path->locks[level] = BTRFS_WRITE_LOCK;
6288
6289 wc->refs[parent_level] = 1;
6290 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
6291 wc->level = level;
6292 wc->shared_level = -1;
6293 wc->stage = DROP_REFERENCE;
6294 wc->update_ref = 0;
6295 wc->keep_locks = 1;
6296 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(fs_info);
6297
6298 while (1) {
6299 ret = walk_down_tree(trans, root, path, wc);
6300 if (ret < 0)
6301 break;
6302
6303 ret = walk_up_tree(trans, root, path, wc, parent_level);
6304 if (ret) {
6305 if (ret > 0)
6306 ret = 0;
6307 break;
6308 }
6309 }
6310
6311 kfree(wc);
6312 btrfs_free_path(path);
6313 return ret;
6314 }
6315
6316 /*
6317 * Unpin the extent range in an error context and don't add the space back.
6318 * Errors are not propagated further.
6319 */
6320 void btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info, u64 start, u64 end)
6321 {
6322 unpin_extent_range(fs_info, start, end, false);
6323 }
6324
6325 /*
6326 * It used to be that old block groups would be left around forever.
6327 * Iterating over them would be enough to trim unused space. Since we
6328 * now automatically remove them, we also need to iterate over unallocated
6329 * space.
6330 *
6331 * We don't want a transaction for this since the discard may take a
6332 * substantial amount of time. We don't require that a transaction be
6333 * running, but we do need to take a running transaction into account
6334 * to ensure that we're not discarding chunks that were released or
6335 * allocated in the current transaction.
6336 *
6337 * Holding the chunks lock will prevent other threads from allocating
6338 * or releasing chunks, but it won't prevent a running transaction
6339 * from committing and releasing the memory that the pending chunks
6340 * list head uses. For that, we need to take a reference to the
6341 * transaction and hold the commit root sem. We only need to hold
6342 * it while performing the free space search since we have already
6343 * held back allocations.
6344 */
6345 static int btrfs_trim_free_extents(struct btrfs_device *device, u64 *trimmed)
6346 {
6347 u64 start = BTRFS_DEVICE_RANGE_RESERVED, len = 0, end = 0;
6348 int ret;
6349
6350 *trimmed = 0;
6351
6352 /* Discard not supported = nothing to do. */
6353 if (!bdev_max_discard_sectors(device->bdev))
6354 return 0;
6355
6356 /* Not writable = nothing to do. */
6357 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state))
6358 return 0;
6359
6360 /* No free space = nothing to do. */
6361 if (device->total_bytes <= device->bytes_used)
6362 return 0;
6363
6364 ret = 0;
6365
6366 while (1) {
6367 struct btrfs_fs_info *fs_info = device->fs_info;
6368 u64 bytes;
6369
6370 ret = mutex_lock_interruptible(&fs_info->chunk_mutex);
6371 if (ret)
6372 break;
6373
6374 find_first_clear_extent_bit(&device->alloc_state, start,
6375 &start, &end,
6376 CHUNK_TRIMMED | CHUNK_ALLOCATED);
6377
6378 /* Check if there are any CHUNK_* bits left */
6379 if (start > device->total_bytes) {
6380 WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG));
6381 btrfs_warn_in_rcu(fs_info,
6382 "ignoring attempt to trim beyond device size: offset %llu length %llu device %s device size %llu",
6383 start, end - start + 1,
6384 btrfs_dev_name(device),
6385 device->total_bytes);
6386 mutex_unlock(&fs_info->chunk_mutex);
6387 ret = 0;
6388 break;
6389 }
6390
6391 /* Ensure we skip the reserved space on each device. */
6392 start = max_t(u64, start, BTRFS_DEVICE_RANGE_RESERVED);
6393
6394 /*
6395 * If find_first_clear_extent_bit find a range that spans the
6396 * end of the device it will set end to -1, in this case it's up
6397 * to the caller to trim the value to the size of the device.
6398 */
6399 end = min(end, device->total_bytes - 1);
6400
6401 len = end - start + 1;
6402
6403 /* We didn't find any extents */
6404 if (!len) {
6405 mutex_unlock(&fs_info->chunk_mutex);
6406 ret = 0;
6407 break;
6408 }
6409
6410 ret = btrfs_issue_discard(device->bdev, start, len,
6411 &bytes);
6412 if (!ret)
6413 set_extent_bit(&device->alloc_state, start,
6414 start + bytes - 1, CHUNK_TRIMMED, NULL);
6415 mutex_unlock(&fs_info->chunk_mutex);
6416
6417 if (ret)
6418 break;
6419
6420 start += len;
6421 *trimmed += bytes;
6422
6423 if (fatal_signal_pending(current)) {
6424 ret = -ERESTARTSYS;
6425 break;
6426 }
6427
6428 cond_resched();
6429 }
6430
6431 return ret;
6432 }
6433
6434 /*
6435 * Trim the whole filesystem by:
6436 * 1) trimming the free space in each block group
6437 * 2) trimming the unallocated space on each device
6438 *
6439 * This will also continue trimming even if a block group or device encounters
6440 * an error. The return value will be the last error, or 0 if nothing bad
6441 * happens.
6442 */
6443 int btrfs_trim_fs(struct btrfs_fs_info *fs_info, struct fstrim_range *range)
6444 {
6445 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
6446 struct btrfs_block_group *cache = NULL;
6447 struct btrfs_device *device;
6448 u64 group_trimmed;
6449 u64 range_end = U64_MAX;
6450 u64 start;
6451 u64 end;
6452 u64 trimmed = 0;
6453 u64 bg_failed = 0;
6454 u64 dev_failed = 0;
6455 int bg_ret = 0;
6456 int dev_ret = 0;
6457 int ret = 0;
6458
6459 if (range->start == U64_MAX)
6460 return -EINVAL;
6461
6462 /*
6463 * Check range overflow if range->len is set.
6464 * The default range->len is U64_MAX.
6465 */
6466 if (range->len != U64_MAX &&
6467 check_add_overflow(range->start, range->len, &range_end))
6468 return -EINVAL;
6469
6470 cache = btrfs_lookup_first_block_group(fs_info, range->start);
6471 for (; cache; cache = btrfs_next_block_group(cache)) {
6472 if (cache->start >= range_end) {
6473 btrfs_put_block_group(cache);
6474 break;
6475 }
6476
6477 start = max(range->start, cache->start);
6478 end = min(range_end, cache->start + cache->length);
6479
6480 if (end - start >= range->minlen) {
6481 if (!btrfs_block_group_done(cache)) {
6482 ret = btrfs_cache_block_group(cache, true);
6483 if (ret) {
6484 bg_failed++;
6485 bg_ret = ret;
6486 continue;
6487 }
6488 }
6489 ret = btrfs_trim_block_group(cache,
6490 &group_trimmed,
6491 start,
6492 end,
6493 range->minlen);
6494
6495 trimmed += group_trimmed;
6496 if (ret) {
6497 bg_failed++;
6498 bg_ret = ret;
6499 continue;
6500 }
6501 }
6502 }
6503
6504 if (bg_failed)
6505 btrfs_warn(fs_info,
6506 "failed to trim %llu block group(s), last error %d",
6507 bg_failed, bg_ret);
6508
6509 mutex_lock(&fs_devices->device_list_mutex);
6510 list_for_each_entry(device, &fs_devices->devices, dev_list) {
6511 if (test_bit(BTRFS_DEV_STATE_MISSING, &device->dev_state))
6512 continue;
6513
6514 ret = btrfs_trim_free_extents(device, &group_trimmed);
6515 if (ret) {
6516 dev_failed++;
6517 dev_ret = ret;
6518 break;
6519 }
6520
6521 trimmed += group_trimmed;
6522 }
6523 mutex_unlock(&fs_devices->device_list_mutex);
6524
6525 if (dev_failed)
6526 btrfs_warn(fs_info,
6527 "failed to trim %llu device(s), last error %d",
6528 dev_failed, dev_ret);
6529 range->len = trimmed;
6530 if (bg_ret)
6531 return bg_ret;
6532 return dev_ret;
6533 }
6534
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.