Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2009 Oracle. All rights reserved.
4 */
5
6#include <linux/sched.h>
7#include <linux/pagemap.h>
8#include <linux/writeback.h>
9#include <linux/blkdev.h>
10#include <linux/rbtree.h>
11#include <linux/slab.h>
12#include <linux/error-injection.h>
13#include "ctree.h"
14#include "disk-io.h"
15#include "transaction.h"
16#include "volumes.h"
17#include "locking.h"
18#include "btrfs_inode.h"
19#include "async-thread.h"
20#include "free-space-cache.h"
21#include "qgroup.h"
22#include "print-tree.h"
23#include "delalloc-space.h"
24#include "block-group.h"
25#include "backref.h"
26#include "misc.h"
27#include "subpage.h"
28#include "zoned.h"
29#include "inode-item.h"
30#include "space-info.h"
31#include "fs.h"
32#include "accessors.h"
33#include "extent-tree.h"
34#include "root-tree.h"
35#include "file-item.h"
36#include "relocation.h"
37#include "super.h"
38#include "tree-checker.h"
39
40/*
41 * Relocation overview
42 *
43 * [What does relocation do]
44 *
45 * The objective of relocation is to relocate all extents of the target block
46 * group to other block groups.
47 * This is utilized by resize (shrink only), profile converting, compacting
48 * space, or balance routine to spread chunks over devices.
49 *
50 * Before | After
51 * ------------------------------------------------------------------
52 * BG A: 10 data extents | BG A: deleted
53 * BG B: 2 data extents | BG B: 10 data extents (2 old + 8 relocated)
54 * BG C: 1 extents | BG C: 3 data extents (1 old + 2 relocated)
55 *
56 * [How does relocation work]
57 *
58 * 1. Mark the target block group read-only
59 * New extents won't be allocated from the target block group.
60 *
61 * 2.1 Record each extent in the target block group
62 * To build a proper map of extents to be relocated.
63 *
64 * 2.2 Build data reloc tree and reloc trees
65 * Data reloc tree will contain an inode, recording all newly relocated
66 * data extents.
67 * There will be only one data reloc tree for one data block group.
68 *
69 * Reloc tree will be a special snapshot of its source tree, containing
70 * relocated tree blocks.
71 * Each tree referring to a tree block in target block group will get its
72 * reloc tree built.
73 *
74 * 2.3 Swap source tree with its corresponding reloc tree
75 * Each involved tree only refers to new extents after swap.
76 *
77 * 3. Cleanup reloc trees and data reloc tree.
78 * As old extents in the target block group are still referenced by reloc
79 * trees, we need to clean them up before really freeing the target block
80 * group.
81 *
82 * The main complexity is in steps 2.2 and 2.3.
83 *
84 * The entry point of relocation is relocate_block_group() function.
85 */
86
87#define RELOCATION_RESERVED_NODES 256
88/*
89 * map address of tree root to tree
90 */
91struct mapping_node {
92 struct {
93 struct rb_node rb_node;
94 u64 bytenr;
95 }; /* Use rb_simle_node for search/insert */
96 void *data;
97};
98
99struct mapping_tree {
100 struct rb_root rb_root;
101 spinlock_t lock;
102};
103
104/*
105 * present a tree block to process
106 */
107struct tree_block {
108 struct {
109 struct rb_node rb_node;
110 u64 bytenr;
111 }; /* Use rb_simple_node for search/insert */
112 u64 owner;
113 struct btrfs_key key;
114 unsigned int level:8;
115 unsigned int key_ready:1;
116};
117
118#define MAX_EXTENTS 128
119
120struct file_extent_cluster {
121 u64 start;
122 u64 end;
123 u64 boundary[MAX_EXTENTS];
124 unsigned int nr;
125};
126
127struct reloc_control {
128 /* block group to relocate */
129 struct btrfs_block_group *block_group;
130 /* extent tree */
131 struct btrfs_root *extent_root;
132 /* inode for moving data */
133 struct inode *data_inode;
134
135 struct btrfs_block_rsv *block_rsv;
136
137 struct btrfs_backref_cache backref_cache;
138
139 struct file_extent_cluster cluster;
140 /* tree blocks have been processed */
141 struct extent_io_tree processed_blocks;
142 /* map start of tree root to corresponding reloc tree */
143 struct mapping_tree reloc_root_tree;
144 /* list of reloc trees */
145 struct list_head reloc_roots;
146 /* list of subvolume trees that get relocated */
147 struct list_head dirty_subvol_roots;
148 /* size of metadata reservation for merging reloc trees */
149 u64 merging_rsv_size;
150 /* size of relocated tree nodes */
151 u64 nodes_relocated;
152 /* reserved size for block group relocation*/
153 u64 reserved_bytes;
154
155 u64 search_start;
156 u64 extents_found;
157
158 unsigned int stage:8;
159 unsigned int create_reloc_tree:1;
160 unsigned int merge_reloc_tree:1;
161 unsigned int found_file_extent:1;
162};
163
164/* stages of data relocation */
165#define MOVE_DATA_EXTENTS 0
166#define UPDATE_DATA_PTRS 1
167
168static void mark_block_processed(struct reloc_control *rc,
169 struct btrfs_backref_node *node)
170{
171 u32 blocksize;
172
173 if (node->level == 0 ||
174 in_range(node->bytenr, rc->block_group->start,
175 rc->block_group->length)) {
176 blocksize = rc->extent_root->fs_info->nodesize;
177 set_extent_bits(&rc->processed_blocks, node->bytenr,
178 node->bytenr + blocksize - 1, EXTENT_DIRTY);
179 }
180 node->processed = 1;
181}
182
183
184static void mapping_tree_init(struct mapping_tree *tree)
185{
186 tree->rb_root = RB_ROOT;
187 spin_lock_init(&tree->lock);
188}
189
190/*
191 * walk up backref nodes until reach node presents tree root
192 */
193static struct btrfs_backref_node *walk_up_backref(
194 struct btrfs_backref_node *node,
195 struct btrfs_backref_edge *edges[], int *index)
196{
197 struct btrfs_backref_edge *edge;
198 int idx = *index;
199
200 while (!list_empty(&node->upper)) {
201 edge = list_entry(node->upper.next,
202 struct btrfs_backref_edge, list[LOWER]);
203 edges[idx++] = edge;
204 node = edge->node[UPPER];
205 }
206 BUG_ON(node->detached);
207 *index = idx;
208 return node;
209}
210
211/*
212 * walk down backref nodes to find start of next reference path
213 */
214static struct btrfs_backref_node *walk_down_backref(
215 struct btrfs_backref_edge *edges[], int *index)
216{
217 struct btrfs_backref_edge *edge;
218 struct btrfs_backref_node *lower;
219 int idx = *index;
220
221 while (idx > 0) {
222 edge = edges[idx - 1];
223 lower = edge->node[LOWER];
224 if (list_is_last(&edge->list[LOWER], &lower->upper)) {
225 idx--;
226 continue;
227 }
228 edge = list_entry(edge->list[LOWER].next,
229 struct btrfs_backref_edge, list[LOWER]);
230 edges[idx - 1] = edge;
231 *index = idx;
232 return edge->node[UPPER];
233 }
234 *index = 0;
235 return NULL;
236}
237
238static void update_backref_node(struct btrfs_backref_cache *cache,
239 struct btrfs_backref_node *node, u64 bytenr)
240{
241 struct rb_node *rb_node;
242 rb_erase(&node->rb_node, &cache->rb_root);
243 node->bytenr = bytenr;
244 rb_node = rb_simple_insert(&cache->rb_root, node->bytenr, &node->rb_node);
245 if (rb_node)
246 btrfs_backref_panic(cache->fs_info, bytenr, -EEXIST);
247}
248
249/*
250 * update backref cache after a transaction commit
251 */
252static int update_backref_cache(struct btrfs_trans_handle *trans,
253 struct btrfs_backref_cache *cache)
254{
255 struct btrfs_backref_node *node;
256 int level = 0;
257
258 if (cache->last_trans == 0) {
259 cache->last_trans = trans->transid;
260 return 0;
261 }
262
263 if (cache->last_trans == trans->transid)
264 return 0;
265
266 /*
267 * detached nodes are used to avoid unnecessary backref
268 * lookup. transaction commit changes the extent tree.
269 * so the detached nodes are no longer useful.
270 */
271 while (!list_empty(&cache->detached)) {
272 node = list_entry(cache->detached.next,
273 struct btrfs_backref_node, list);
274 btrfs_backref_cleanup_node(cache, node);
275 }
276
277 while (!list_empty(&cache->changed)) {
278 node = list_entry(cache->changed.next,
279 struct btrfs_backref_node, list);
280 list_del_init(&node->list);
281 BUG_ON(node->pending);
282 update_backref_node(cache, node, node->new_bytenr);
283 }
284
285 /*
286 * some nodes can be left in the pending list if there were
287 * errors during processing the pending nodes.
288 */
289 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
290 list_for_each_entry(node, &cache->pending[level], list) {
291 BUG_ON(!node->pending);
292 if (node->bytenr == node->new_bytenr)
293 continue;
294 update_backref_node(cache, node, node->new_bytenr);
295 }
296 }
297
298 cache->last_trans = 0;
299 return 1;
300}
301
302static bool reloc_root_is_dead(struct btrfs_root *root)
303{
304 /*
305 * Pair with set_bit/clear_bit in clean_dirty_subvols and
306 * btrfs_update_reloc_root. We need to see the updated bit before
307 * trying to access reloc_root
308 */
309 smp_rmb();
310 if (test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state))
311 return true;
312 return false;
313}
314
315/*
316 * Check if this subvolume tree has valid reloc tree.
317 *
318 * Reloc tree after swap is considered dead, thus not considered as valid.
319 * This is enough for most callers, as they don't distinguish dead reloc root
320 * from no reloc root. But btrfs_should_ignore_reloc_root() below is a
321 * special case.
322 */
323static bool have_reloc_root(struct btrfs_root *root)
324{
325 if (reloc_root_is_dead(root))
326 return false;
327 if (!root->reloc_root)
328 return false;
329 return true;
330}
331
332int btrfs_should_ignore_reloc_root(struct btrfs_root *root)
333{
334 struct btrfs_root *reloc_root;
335
336 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
337 return 0;
338
339 /* This root has been merged with its reloc tree, we can ignore it */
340 if (reloc_root_is_dead(root))
341 return 1;
342
343 reloc_root = root->reloc_root;
344 if (!reloc_root)
345 return 0;
346
347 if (btrfs_header_generation(reloc_root->commit_root) ==
348 root->fs_info->running_transaction->transid)
349 return 0;
350 /*
351 * if there is reloc tree and it was created in previous
352 * transaction backref lookup can find the reloc tree,
353 * so backref node for the fs tree root is useless for
354 * relocation.
355 */
356 return 1;
357}
358
359/*
360 * find reloc tree by address of tree root
361 */
362struct btrfs_root *find_reloc_root(struct btrfs_fs_info *fs_info, u64 bytenr)
363{
364 struct reloc_control *rc = fs_info->reloc_ctl;
365 struct rb_node *rb_node;
366 struct mapping_node *node;
367 struct btrfs_root *root = NULL;
368
369 ASSERT(rc);
370 spin_lock(&rc->reloc_root_tree.lock);
371 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root, bytenr);
372 if (rb_node) {
373 node = rb_entry(rb_node, struct mapping_node, rb_node);
374 root = node->data;
375 }
376 spin_unlock(&rc->reloc_root_tree.lock);
377 return btrfs_grab_root(root);
378}
379
380/*
381 * For useless nodes, do two major clean ups:
382 *
383 * - Cleanup the children edges and nodes
384 * If child node is also orphan (no parent) during cleanup, then the child
385 * node will also be cleaned up.
386 *
387 * - Freeing up leaves (level 0), keeps nodes detached
388 * For nodes, the node is still cached as "detached"
389 *
390 * Return false if @node is not in the @useless_nodes list.
391 * Return true if @node is in the @useless_nodes list.
392 */
393static bool handle_useless_nodes(struct reloc_control *rc,
394 struct btrfs_backref_node *node)
395{
396 struct btrfs_backref_cache *cache = &rc->backref_cache;
397 struct list_head *useless_node = &cache->useless_node;
398 bool ret = false;
399
400 while (!list_empty(useless_node)) {
401 struct btrfs_backref_node *cur;
402
403 cur = list_first_entry(useless_node, struct btrfs_backref_node,
404 list);
405 list_del_init(&cur->list);
406
407 /* Only tree root nodes can be added to @useless_nodes */
408 ASSERT(list_empty(&cur->upper));
409
410 if (cur == node)
411 ret = true;
412
413 /* The node is the lowest node */
414 if (cur->lowest) {
415 list_del_init(&cur->lower);
416 cur->lowest = 0;
417 }
418
419 /* Cleanup the lower edges */
420 while (!list_empty(&cur->lower)) {
421 struct btrfs_backref_edge *edge;
422 struct btrfs_backref_node *lower;
423
424 edge = list_entry(cur->lower.next,
425 struct btrfs_backref_edge, list[UPPER]);
426 list_del(&edge->list[UPPER]);
427 list_del(&edge->list[LOWER]);
428 lower = edge->node[LOWER];
429 btrfs_backref_free_edge(cache, edge);
430
431 /* Child node is also orphan, queue for cleanup */
432 if (list_empty(&lower->upper))
433 list_add(&lower->list, useless_node);
434 }
435 /* Mark this block processed for relocation */
436 mark_block_processed(rc, cur);
437
438 /*
439 * Backref nodes for tree leaves are deleted from the cache.
440 * Backref nodes for upper level tree blocks are left in the
441 * cache to avoid unnecessary backref lookup.
442 */
443 if (cur->level > 0) {
444 list_add(&cur->list, &cache->detached);
445 cur->detached = 1;
446 } else {
447 rb_erase(&cur->rb_node, &cache->rb_root);
448 btrfs_backref_free_node(cache, cur);
449 }
450 }
451 return ret;
452}
453
454/*
455 * Build backref tree for a given tree block. Root of the backref tree
456 * corresponds the tree block, leaves of the backref tree correspond roots of
457 * b-trees that reference the tree block.
458 *
459 * The basic idea of this function is check backrefs of a given block to find
460 * upper level blocks that reference the block, and then check backrefs of
461 * these upper level blocks recursively. The recursion stops when tree root is
462 * reached or backrefs for the block is cached.
463 *
464 * NOTE: if we find that backrefs for a block are cached, we know backrefs for
465 * all upper level blocks that directly/indirectly reference the block are also
466 * cached.
467 */
468static noinline_for_stack struct btrfs_backref_node *build_backref_tree(
469 struct reloc_control *rc, struct btrfs_key *node_key,
470 int level, u64 bytenr)
471{
472 struct btrfs_backref_iter *iter;
473 struct btrfs_backref_cache *cache = &rc->backref_cache;
474 /* For searching parent of TREE_BLOCK_REF */
475 struct btrfs_path *path;
476 struct btrfs_backref_node *cur;
477 struct btrfs_backref_node *node = NULL;
478 struct btrfs_backref_edge *edge;
479 int ret;
480 int err = 0;
481
482 iter = btrfs_backref_iter_alloc(rc->extent_root->fs_info);
483 if (!iter)
484 return ERR_PTR(-ENOMEM);
485 path = btrfs_alloc_path();
486 if (!path) {
487 err = -ENOMEM;
488 goto out;
489 }
490
491 node = btrfs_backref_alloc_node(cache, bytenr, level);
492 if (!node) {
493 err = -ENOMEM;
494 goto out;
495 }
496
497 node->lowest = 1;
498 cur = node;
499
500 /* Breadth-first search to build backref cache */
501 do {
502 ret = btrfs_backref_add_tree_node(cache, path, iter, node_key,
503 cur);
504 if (ret < 0) {
505 err = ret;
506 goto out;
507 }
508 edge = list_first_entry_or_null(&cache->pending_edge,
509 struct btrfs_backref_edge, list[UPPER]);
510 /*
511 * The pending list isn't empty, take the first block to
512 * process
513 */
514 if (edge) {
515 list_del_init(&edge->list[UPPER]);
516 cur = edge->node[UPPER];
517 }
518 } while (edge);
519
520 /* Finish the upper linkage of newly added edges/nodes */
521 ret = btrfs_backref_finish_upper_links(cache, node);
522 if (ret < 0) {
523 err = ret;
524 goto out;
525 }
526
527 if (handle_useless_nodes(rc, node))
528 node = NULL;
529out:
530 btrfs_backref_iter_free(iter);
531 btrfs_free_path(path);
532 if (err) {
533 btrfs_backref_error_cleanup(cache, node);
534 return ERR_PTR(err);
535 }
536 ASSERT(!node || !node->detached);
537 ASSERT(list_empty(&cache->useless_node) &&
538 list_empty(&cache->pending_edge));
539 return node;
540}
541
542/*
543 * helper to add backref node for the newly created snapshot.
544 * the backref node is created by cloning backref node that
545 * corresponds to root of source tree
546 */
547static int clone_backref_node(struct btrfs_trans_handle *trans,
548 struct reloc_control *rc,
549 struct btrfs_root *src,
550 struct btrfs_root *dest)
551{
552 struct btrfs_root *reloc_root = src->reloc_root;
553 struct btrfs_backref_cache *cache = &rc->backref_cache;
554 struct btrfs_backref_node *node = NULL;
555 struct btrfs_backref_node *new_node;
556 struct btrfs_backref_edge *edge;
557 struct btrfs_backref_edge *new_edge;
558 struct rb_node *rb_node;
559
560 if (cache->last_trans > 0)
561 update_backref_cache(trans, cache);
562
563 rb_node = rb_simple_search(&cache->rb_root, src->commit_root->start);
564 if (rb_node) {
565 node = rb_entry(rb_node, struct btrfs_backref_node, rb_node);
566 if (node->detached)
567 node = NULL;
568 else
569 BUG_ON(node->new_bytenr != reloc_root->node->start);
570 }
571
572 if (!node) {
573 rb_node = rb_simple_search(&cache->rb_root,
574 reloc_root->commit_root->start);
575 if (rb_node) {
576 node = rb_entry(rb_node, struct btrfs_backref_node,
577 rb_node);
578 BUG_ON(node->detached);
579 }
580 }
581
582 if (!node)
583 return 0;
584
585 new_node = btrfs_backref_alloc_node(cache, dest->node->start,
586 node->level);
587 if (!new_node)
588 return -ENOMEM;
589
590 new_node->lowest = node->lowest;
591 new_node->checked = 1;
592 new_node->root = btrfs_grab_root(dest);
593 ASSERT(new_node->root);
594
595 if (!node->lowest) {
596 list_for_each_entry(edge, &node->lower, list[UPPER]) {
597 new_edge = btrfs_backref_alloc_edge(cache);
598 if (!new_edge)
599 goto fail;
600
601 btrfs_backref_link_edge(new_edge, edge->node[LOWER],
602 new_node, LINK_UPPER);
603 }
604 } else {
605 list_add_tail(&new_node->lower, &cache->leaves);
606 }
607
608 rb_node = rb_simple_insert(&cache->rb_root, new_node->bytenr,
609 &new_node->rb_node);
610 if (rb_node)
611 btrfs_backref_panic(trans->fs_info, new_node->bytenr, -EEXIST);
612
613 if (!new_node->lowest) {
614 list_for_each_entry(new_edge, &new_node->lower, list[UPPER]) {
615 list_add_tail(&new_edge->list[LOWER],
616 &new_edge->node[LOWER]->upper);
617 }
618 }
619 return 0;
620fail:
621 while (!list_empty(&new_node->lower)) {
622 new_edge = list_entry(new_node->lower.next,
623 struct btrfs_backref_edge, list[UPPER]);
624 list_del(&new_edge->list[UPPER]);
625 btrfs_backref_free_edge(cache, new_edge);
626 }
627 btrfs_backref_free_node(cache, new_node);
628 return -ENOMEM;
629}
630
631/*
632 * helper to add 'address of tree root -> reloc tree' mapping
633 */
634static int __must_check __add_reloc_root(struct btrfs_root *root)
635{
636 struct btrfs_fs_info *fs_info = root->fs_info;
637 struct rb_node *rb_node;
638 struct mapping_node *node;
639 struct reloc_control *rc = fs_info->reloc_ctl;
640
641 node = kmalloc(sizeof(*node), GFP_NOFS);
642 if (!node)
643 return -ENOMEM;
644
645 node->bytenr = root->commit_root->start;
646 node->data = root;
647
648 spin_lock(&rc->reloc_root_tree.lock);
649 rb_node = rb_simple_insert(&rc->reloc_root_tree.rb_root,
650 node->bytenr, &node->rb_node);
651 spin_unlock(&rc->reloc_root_tree.lock);
652 if (rb_node) {
653 btrfs_err(fs_info,
654 "Duplicate root found for start=%llu while inserting into relocation tree",
655 node->bytenr);
656 return -EEXIST;
657 }
658
659 list_add_tail(&root->root_list, &rc->reloc_roots);
660 return 0;
661}
662
663/*
664 * helper to delete the 'address of tree root -> reloc tree'
665 * mapping
666 */
667static void __del_reloc_root(struct btrfs_root *root)
668{
669 struct btrfs_fs_info *fs_info = root->fs_info;
670 struct rb_node *rb_node;
671 struct mapping_node *node = NULL;
672 struct reloc_control *rc = fs_info->reloc_ctl;
673 bool put_ref = false;
674
675 if (rc && root->node) {
676 spin_lock(&rc->reloc_root_tree.lock);
677 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root,
678 root->commit_root->start);
679 if (rb_node) {
680 node = rb_entry(rb_node, struct mapping_node, rb_node);
681 rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
682 RB_CLEAR_NODE(&node->rb_node);
683 }
684 spin_unlock(&rc->reloc_root_tree.lock);
685 ASSERT(!node || (struct btrfs_root *)node->data == root);
686 }
687
688 /*
689 * We only put the reloc root here if it's on the list. There's a lot
690 * of places where the pattern is to splice the rc->reloc_roots, process
691 * the reloc roots, and then add the reloc root back onto
692 * rc->reloc_roots. If we call __del_reloc_root while it's off of the
693 * list we don't want the reference being dropped, because the guy
694 * messing with the list is in charge of the reference.
695 */
696 spin_lock(&fs_info->trans_lock);
697 if (!list_empty(&root->root_list)) {
698 put_ref = true;
699 list_del_init(&root->root_list);
700 }
701 spin_unlock(&fs_info->trans_lock);
702 if (put_ref)
703 btrfs_put_root(root);
704 kfree(node);
705}
706
707/*
708 * helper to update the 'address of tree root -> reloc tree'
709 * mapping
710 */
711static int __update_reloc_root(struct btrfs_root *root)
712{
713 struct btrfs_fs_info *fs_info = root->fs_info;
714 struct rb_node *rb_node;
715 struct mapping_node *node = NULL;
716 struct reloc_control *rc = fs_info->reloc_ctl;
717
718 spin_lock(&rc->reloc_root_tree.lock);
719 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root,
720 root->commit_root->start);
721 if (rb_node) {
722 node = rb_entry(rb_node, struct mapping_node, rb_node);
723 rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
724 }
725 spin_unlock(&rc->reloc_root_tree.lock);
726
727 if (!node)
728 return 0;
729 BUG_ON((struct btrfs_root *)node->data != root);
730
731 spin_lock(&rc->reloc_root_tree.lock);
732 node->bytenr = root->node->start;
733 rb_node = rb_simple_insert(&rc->reloc_root_tree.rb_root,
734 node->bytenr, &node->rb_node);
735 spin_unlock(&rc->reloc_root_tree.lock);
736 if (rb_node)
737 btrfs_backref_panic(fs_info, node->bytenr, -EEXIST);
738 return 0;
739}
740
741static struct btrfs_root *create_reloc_root(struct btrfs_trans_handle *trans,
742 struct btrfs_root *root, u64 objectid)
743{
744 struct btrfs_fs_info *fs_info = root->fs_info;
745 struct btrfs_root *reloc_root;
746 struct extent_buffer *eb;
747 struct btrfs_root_item *root_item;
748 struct btrfs_key root_key;
749 int ret = 0;
750 bool must_abort = false;
751
752 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
753 if (!root_item)
754 return ERR_PTR(-ENOMEM);
755
756 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
757 root_key.type = BTRFS_ROOT_ITEM_KEY;
758 root_key.offset = objectid;
759
760 if (root->root_key.objectid == objectid) {
761 u64 commit_root_gen;
762
763 /* called by btrfs_init_reloc_root */
764 ret = btrfs_copy_root(trans, root, root->commit_root, &eb,
765 BTRFS_TREE_RELOC_OBJECTID);
766 if (ret)
767 goto fail;
768
769 /*
770 * Set the last_snapshot field to the generation of the commit
771 * root - like this ctree.c:btrfs_block_can_be_shared() behaves
772 * correctly (returns true) when the relocation root is created
773 * either inside the critical section of a transaction commit
774 * (through transaction.c:qgroup_account_snapshot()) and when
775 * it's created before the transaction commit is started.
776 */
777 commit_root_gen = btrfs_header_generation(root->commit_root);
778 btrfs_set_root_last_snapshot(&root->root_item, commit_root_gen);
779 } else {
780 /*
781 * called by btrfs_reloc_post_snapshot_hook.
782 * the source tree is a reloc tree, all tree blocks
783 * modified after it was created have RELOC flag
784 * set in their headers. so it's OK to not update
785 * the 'last_snapshot'.
786 */
787 ret = btrfs_copy_root(trans, root, root->node, &eb,
788 BTRFS_TREE_RELOC_OBJECTID);
789 if (ret)
790 goto fail;
791 }
792
793 /*
794 * We have changed references at this point, we must abort the
795 * transaction if anything fails.
796 */
797 must_abort = true;
798
799 memcpy(root_item, &root->root_item, sizeof(*root_item));
800 btrfs_set_root_bytenr(root_item, eb->start);
801 btrfs_set_root_level(root_item, btrfs_header_level(eb));
802 btrfs_set_root_generation(root_item, trans->transid);
803
804 if (root->root_key.objectid == objectid) {
805 btrfs_set_root_refs(root_item, 0);
806 memset(&root_item->drop_progress, 0,
807 sizeof(struct btrfs_disk_key));
808 btrfs_set_root_drop_level(root_item, 0);
809 }
810
811 btrfs_tree_unlock(eb);
812 free_extent_buffer(eb);
813
814 ret = btrfs_insert_root(trans, fs_info->tree_root,
815 &root_key, root_item);
816 if (ret)
817 goto fail;
818
819 kfree(root_item);
820
821 reloc_root = btrfs_read_tree_root(fs_info->tree_root, &root_key);
822 if (IS_ERR(reloc_root)) {
823 ret = PTR_ERR(reloc_root);
824 goto abort;
825 }
826 set_bit(BTRFS_ROOT_SHAREABLE, &reloc_root->state);
827 reloc_root->last_trans = trans->transid;
828 return reloc_root;
829fail:
830 kfree(root_item);
831abort:
832 if (must_abort)
833 btrfs_abort_transaction(trans, ret);
834 return ERR_PTR(ret);
835}
836
837/*
838 * create reloc tree for a given fs tree. reloc tree is just a
839 * snapshot of the fs tree with special root objectid.
840 *
841 * The reloc_root comes out of here with two references, one for
842 * root->reloc_root, and another for being on the rc->reloc_roots list.
843 */
844int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
845 struct btrfs_root *root)
846{
847 struct btrfs_fs_info *fs_info = root->fs_info;
848 struct btrfs_root *reloc_root;
849 struct reloc_control *rc = fs_info->reloc_ctl;
850 struct btrfs_block_rsv *rsv;
851 int clear_rsv = 0;
852 int ret;
853
854 if (!rc)
855 return 0;
856
857 /*
858 * The subvolume has reloc tree but the swap is finished, no need to
859 * create/update the dead reloc tree
860 */
861 if (reloc_root_is_dead(root))
862 return 0;
863
864 /*
865 * This is subtle but important. We do not do
866 * record_root_in_transaction for reloc roots, instead we record their
867 * corresponding fs root, and then here we update the last trans for the
868 * reloc root. This means that we have to do this for the entire life
869 * of the reloc root, regardless of which stage of the relocation we are
870 * in.
871 */
872 if (root->reloc_root) {
873 reloc_root = root->reloc_root;
874 reloc_root->last_trans = trans->transid;
875 return 0;
876 }
877
878 /*
879 * We are merging reloc roots, we do not need new reloc trees. Also
880 * reloc trees never need their own reloc tree.
881 */
882 if (!rc->create_reloc_tree ||
883 root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
884 return 0;
885
886 if (!trans->reloc_reserved) {
887 rsv = trans->block_rsv;
888 trans->block_rsv = rc->block_rsv;
889 clear_rsv = 1;
890 }
891 reloc_root = create_reloc_root(trans, root, root->root_key.objectid);
892 if (clear_rsv)
893 trans->block_rsv = rsv;
894 if (IS_ERR(reloc_root))
895 return PTR_ERR(reloc_root);
896
897 ret = __add_reloc_root(reloc_root);
898 ASSERT(ret != -EEXIST);
899 if (ret) {
900 /* Pairs with create_reloc_root */
901 btrfs_put_root(reloc_root);
902 return ret;
903 }
904 root->reloc_root = btrfs_grab_root(reloc_root);
905 return 0;
906}
907
908/*
909 * update root item of reloc tree
910 */
911int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
912 struct btrfs_root *root)
913{
914 struct btrfs_fs_info *fs_info = root->fs_info;
915 struct btrfs_root *reloc_root;
916 struct btrfs_root_item *root_item;
917 int ret;
918
919 if (!have_reloc_root(root))
920 return 0;
921
922 reloc_root = root->reloc_root;
923 root_item = &reloc_root->root_item;
924
925 /*
926 * We are probably ok here, but __del_reloc_root() will drop its ref of
927 * the root. We have the ref for root->reloc_root, but just in case
928 * hold it while we update the reloc root.
929 */
930 btrfs_grab_root(reloc_root);
931
932 /* root->reloc_root will stay until current relocation finished */
933 if (fs_info->reloc_ctl->merge_reloc_tree &&
934 btrfs_root_refs(root_item) == 0) {
935 set_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
936 /*
937 * Mark the tree as dead before we change reloc_root so
938 * have_reloc_root will not touch it from now on.
939 */
940 smp_wmb();
941 __del_reloc_root(reloc_root);
942 }
943
944 if (reloc_root->commit_root != reloc_root->node) {
945 __update_reloc_root(reloc_root);
946 btrfs_set_root_node(root_item, reloc_root->node);
947 free_extent_buffer(reloc_root->commit_root);
948 reloc_root->commit_root = btrfs_root_node(reloc_root);
949 }
950
951 ret = btrfs_update_root(trans, fs_info->tree_root,
952 &reloc_root->root_key, root_item);
953 btrfs_put_root(reloc_root);
954 return ret;
955}
956
957/*
958 * helper to find first cached inode with inode number >= objectid
959 * in a subvolume
960 */
961static struct inode *find_next_inode(struct btrfs_root *root, u64 objectid)
962{
963 struct rb_node *node;
964 struct rb_node *prev;
965 struct btrfs_inode *entry;
966 struct inode *inode;
967
968 spin_lock(&root->inode_lock);
969again:
970 node = root->inode_tree.rb_node;
971 prev = NULL;
972 while (node) {
973 prev = node;
974 entry = rb_entry(node, struct btrfs_inode, rb_node);
975
976 if (objectid < btrfs_ino(entry))
977 node = node->rb_left;
978 else if (objectid > btrfs_ino(entry))
979 node = node->rb_right;
980 else
981 break;
982 }
983 if (!node) {
984 while (prev) {
985 entry = rb_entry(prev, struct btrfs_inode, rb_node);
986 if (objectid <= btrfs_ino(entry)) {
987 node = prev;
988 break;
989 }
990 prev = rb_next(prev);
991 }
992 }
993 while (node) {
994 entry = rb_entry(node, struct btrfs_inode, rb_node);
995 inode = igrab(&entry->vfs_inode);
996 if (inode) {
997 spin_unlock(&root->inode_lock);
998 return inode;
999 }
1000
1001 objectid = btrfs_ino(entry) + 1;
1002 if (cond_resched_lock(&root->inode_lock))
1003 goto again;
1004
1005 node = rb_next(node);
1006 }
1007 spin_unlock(&root->inode_lock);
1008 return NULL;
1009}
1010
1011/*
1012 * get new location of data
1013 */
1014static int get_new_location(struct inode *reloc_inode, u64 *new_bytenr,
1015 u64 bytenr, u64 num_bytes)
1016{
1017 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
1018 struct btrfs_path *path;
1019 struct btrfs_file_extent_item *fi;
1020 struct extent_buffer *leaf;
1021 int ret;
1022
1023 path = btrfs_alloc_path();
1024 if (!path)
1025 return -ENOMEM;
1026
1027 bytenr -= BTRFS_I(reloc_inode)->index_cnt;
1028 ret = btrfs_lookup_file_extent(NULL, root, path,
1029 btrfs_ino(BTRFS_I(reloc_inode)), bytenr, 0);
1030 if (ret < 0)
1031 goto out;
1032 if (ret > 0) {
1033 ret = -ENOENT;
1034 goto out;
1035 }
1036
1037 leaf = path->nodes[0];
1038 fi = btrfs_item_ptr(leaf, path->slots[0],
1039 struct btrfs_file_extent_item);
1040
1041 BUG_ON(btrfs_file_extent_offset(leaf, fi) ||
1042 btrfs_file_extent_compression(leaf, fi) ||
1043 btrfs_file_extent_encryption(leaf, fi) ||
1044 btrfs_file_extent_other_encoding(leaf, fi));
1045
1046 if (num_bytes != btrfs_file_extent_disk_num_bytes(leaf, fi)) {
1047 ret = -EINVAL;
1048 goto out;
1049 }
1050
1051 *new_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1052 ret = 0;
1053out:
1054 btrfs_free_path(path);
1055 return ret;
1056}
1057
1058/*
1059 * update file extent items in the tree leaf to point to
1060 * the new locations.
1061 */
1062static noinline_for_stack
1063int replace_file_extents(struct btrfs_trans_handle *trans,
1064 struct reloc_control *rc,
1065 struct btrfs_root *root,
1066 struct extent_buffer *leaf)
1067{
1068 struct btrfs_fs_info *fs_info = root->fs_info;
1069 struct btrfs_key key;
1070 struct btrfs_file_extent_item *fi;
1071 struct inode *inode = NULL;
1072 u64 parent;
1073 u64 bytenr;
1074 u64 new_bytenr = 0;
1075 u64 num_bytes;
1076 u64 end;
1077 u32 nritems;
1078 u32 i;
1079 int ret = 0;
1080 int first = 1;
1081 int dirty = 0;
1082
1083 if (rc->stage != UPDATE_DATA_PTRS)
1084 return 0;
1085
1086 /* reloc trees always use full backref */
1087 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
1088 parent = leaf->start;
1089 else
1090 parent = 0;
1091
1092 nritems = btrfs_header_nritems(leaf);
1093 for (i = 0; i < nritems; i++) {
1094 struct btrfs_ref ref = { 0 };
1095
1096 cond_resched();
1097 btrfs_item_key_to_cpu(leaf, &key, i);
1098 if (key.type != BTRFS_EXTENT_DATA_KEY)
1099 continue;
1100 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
1101 if (btrfs_file_extent_type(leaf, fi) ==
1102 BTRFS_FILE_EXTENT_INLINE)
1103 continue;
1104 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1105 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
1106 if (bytenr == 0)
1107 continue;
1108 if (!in_range(bytenr, rc->block_group->start,
1109 rc->block_group->length))
1110 continue;
1111
1112 /*
1113 * if we are modifying block in fs tree, wait for read_folio
1114 * to complete and drop the extent cache
1115 */
1116 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
1117 if (first) {
1118 inode = find_next_inode(root, key.objectid);
1119 first = 0;
1120 } else if (inode && btrfs_ino(BTRFS_I(inode)) < key.objectid) {
1121 btrfs_add_delayed_iput(BTRFS_I(inode));
1122 inode = find_next_inode(root, key.objectid);
1123 }
1124 if (inode && btrfs_ino(BTRFS_I(inode)) == key.objectid) {
1125 struct extent_state *cached_state = NULL;
1126
1127 end = key.offset +
1128 btrfs_file_extent_num_bytes(leaf, fi);
1129 WARN_ON(!IS_ALIGNED(key.offset,
1130 fs_info->sectorsize));
1131 WARN_ON(!IS_ALIGNED(end, fs_info->sectorsize));
1132 end--;
1133 ret = try_lock_extent(&BTRFS_I(inode)->io_tree,
1134 key.offset, end,
1135 &cached_state);
1136 if (!ret)
1137 continue;
1138
1139 btrfs_drop_extent_map_range(BTRFS_I(inode),
1140 key.offset, end, true);
1141 unlock_extent(&BTRFS_I(inode)->io_tree,
1142 key.offset, end, &cached_state);
1143 }
1144 }
1145
1146 ret = get_new_location(rc->data_inode, &new_bytenr,
1147 bytenr, num_bytes);
1148 if (ret) {
1149 /*
1150 * Don't have to abort since we've not changed anything
1151 * in the file extent yet.
1152 */
1153 break;
1154 }
1155
1156 btrfs_set_file_extent_disk_bytenr(leaf, fi, new_bytenr);
1157 dirty = 1;
1158
1159 key.offset -= btrfs_file_extent_offset(leaf, fi);
1160 btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
1161 num_bytes, parent);
1162 btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
1163 key.objectid, key.offset,
1164 root->root_key.objectid, false);
1165 ret = btrfs_inc_extent_ref(trans, &ref);
1166 if (ret) {
1167 btrfs_abort_transaction(trans, ret);
1168 break;
1169 }
1170
1171 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, bytenr,
1172 num_bytes, parent);
1173 btrfs_init_data_ref(&ref, btrfs_header_owner(leaf),
1174 key.objectid, key.offset,
1175 root->root_key.objectid, false);
1176 ret = btrfs_free_extent(trans, &ref);
1177 if (ret) {
1178 btrfs_abort_transaction(trans, ret);
1179 break;
1180 }
1181 }
1182 if (dirty)
1183 btrfs_mark_buffer_dirty(leaf);
1184 if (inode)
1185 btrfs_add_delayed_iput(BTRFS_I(inode));
1186 return ret;
1187}
1188
1189static noinline_for_stack
1190int memcmp_node_keys(struct extent_buffer *eb, int slot,
1191 struct btrfs_path *path, int level)
1192{
1193 struct btrfs_disk_key key1;
1194 struct btrfs_disk_key key2;
1195 btrfs_node_key(eb, &key1, slot);
1196 btrfs_node_key(path->nodes[level], &key2, path->slots[level]);
1197 return memcmp(&key1, &key2, sizeof(key1));
1198}
1199
1200/*
1201 * try to replace tree blocks in fs tree with the new blocks
1202 * in reloc tree. tree blocks haven't been modified since the
1203 * reloc tree was create can be replaced.
1204 *
1205 * if a block was replaced, level of the block + 1 is returned.
1206 * if no block got replaced, 0 is returned. if there are other
1207 * errors, a negative error number is returned.
1208 */
1209static noinline_for_stack
1210int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
1211 struct btrfs_root *dest, struct btrfs_root *src,
1212 struct btrfs_path *path, struct btrfs_key *next_key,
1213 int lowest_level, int max_level)
1214{
1215 struct btrfs_fs_info *fs_info = dest->fs_info;
1216 struct extent_buffer *eb;
1217 struct extent_buffer *parent;
1218 struct btrfs_ref ref = { 0 };
1219 struct btrfs_key key;
1220 u64 old_bytenr;
1221 u64 new_bytenr;
1222 u64 old_ptr_gen;
1223 u64 new_ptr_gen;
1224 u64 last_snapshot;
1225 u32 blocksize;
1226 int cow = 0;
1227 int level;
1228 int ret;
1229 int slot;
1230
1231 ASSERT(src->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID);
1232 ASSERT(dest->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
1233
1234 last_snapshot = btrfs_root_last_snapshot(&src->root_item);
1235again:
1236 slot = path->slots[lowest_level];
1237 btrfs_node_key_to_cpu(path->nodes[lowest_level], &key, slot);
1238
1239 eb = btrfs_lock_root_node(dest);
1240 level = btrfs_header_level(eb);
1241
1242 if (level < lowest_level) {
1243 btrfs_tree_unlock(eb);
1244 free_extent_buffer(eb);
1245 return 0;
1246 }
1247
1248 if (cow) {
1249 ret = btrfs_cow_block(trans, dest, eb, NULL, 0, &eb,
1250 BTRFS_NESTING_COW);
1251 if (ret) {
1252 btrfs_tree_unlock(eb);
1253 free_extent_buffer(eb);
1254 return ret;
1255 }
1256 }
1257
1258 if (next_key) {
1259 next_key->objectid = (u64)-1;
1260 next_key->type = (u8)-1;
1261 next_key->offset = (u64)-1;
1262 }
1263
1264 parent = eb;
1265 while (1) {
1266 level = btrfs_header_level(parent);
1267 ASSERT(level >= lowest_level);
1268
1269 ret = btrfs_bin_search(parent, &key, &slot);
1270 if (ret < 0)
1271 break;
1272 if (ret && slot > 0)
1273 slot--;
1274
1275 if (next_key && slot + 1 < btrfs_header_nritems(parent))
1276 btrfs_node_key_to_cpu(parent, next_key, slot + 1);
1277
1278 old_bytenr = btrfs_node_blockptr(parent, slot);
1279 blocksize = fs_info->nodesize;
1280 old_ptr_gen = btrfs_node_ptr_generation(parent, slot);
1281
1282 if (level <= max_level) {
1283 eb = path->nodes[level];
1284 new_bytenr = btrfs_node_blockptr(eb,
1285 path->slots[level]);
1286 new_ptr_gen = btrfs_node_ptr_generation(eb,
1287 path->slots[level]);
1288 } else {
1289 new_bytenr = 0;
1290 new_ptr_gen = 0;
1291 }
1292
1293 if (WARN_ON(new_bytenr > 0 && new_bytenr == old_bytenr)) {
1294 ret = level;
1295 break;
1296 }
1297
1298 if (new_bytenr == 0 || old_ptr_gen > last_snapshot ||
1299 memcmp_node_keys(parent, slot, path, level)) {
1300 if (level <= lowest_level) {
1301 ret = 0;
1302 break;
1303 }
1304
1305 eb = btrfs_read_node_slot(parent, slot);
1306 if (IS_ERR(eb)) {
1307 ret = PTR_ERR(eb);
1308 break;
1309 }
1310 btrfs_tree_lock(eb);
1311 if (cow) {
1312 ret = btrfs_cow_block(trans, dest, eb, parent,
1313 slot, &eb,
1314 BTRFS_NESTING_COW);
1315 if (ret) {
1316 btrfs_tree_unlock(eb);
1317 free_extent_buffer(eb);
1318 break;
1319 }
1320 }
1321
1322 btrfs_tree_unlock(parent);
1323 free_extent_buffer(parent);
1324
1325 parent = eb;
1326 continue;
1327 }
1328
1329 if (!cow) {
1330 btrfs_tree_unlock(parent);
1331 free_extent_buffer(parent);
1332 cow = 1;
1333 goto again;
1334 }
1335
1336 btrfs_node_key_to_cpu(path->nodes[level], &key,
1337 path->slots[level]);
1338 btrfs_release_path(path);
1339
1340 path->lowest_level = level;
1341 set_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &src->state);
1342 ret = btrfs_search_slot(trans, src, &key, path, 0, 1);
1343 clear_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &src->state);
1344 path->lowest_level = 0;
1345 if (ret) {
1346 if (ret > 0)
1347 ret = -ENOENT;
1348 break;
1349 }
1350
1351 /*
1352 * Info qgroup to trace both subtrees.
1353 *
1354 * We must trace both trees.
1355 * 1) Tree reloc subtree
1356 * If not traced, we will leak data numbers
1357 * 2) Fs subtree
1358 * If not traced, we will double count old data
1359 *
1360 * We don't scan the subtree right now, but only record
1361 * the swapped tree blocks.
1362 * The real subtree rescan is delayed until we have new
1363 * CoW on the subtree root node before transaction commit.
1364 */
1365 ret = btrfs_qgroup_add_swapped_blocks(trans, dest,
1366 rc->block_group, parent, slot,
1367 path->nodes[level], path->slots[level],
1368 last_snapshot);
1369 if (ret < 0)
1370 break;
1371 /*
1372 * swap blocks in fs tree and reloc tree.
1373 */
1374 btrfs_set_node_blockptr(parent, slot, new_bytenr);
1375 btrfs_set_node_ptr_generation(parent, slot, new_ptr_gen);
1376 btrfs_mark_buffer_dirty(parent);
1377
1378 btrfs_set_node_blockptr(path->nodes[level],
1379 path->slots[level], old_bytenr);
1380 btrfs_set_node_ptr_generation(path->nodes[level],
1381 path->slots[level], old_ptr_gen);
1382 btrfs_mark_buffer_dirty(path->nodes[level]);
1383
1384 btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, old_bytenr,
1385 blocksize, path->nodes[level]->start);
1386 btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
1387 0, true);
1388 ret = btrfs_inc_extent_ref(trans, &ref);
1389 if (ret) {
1390 btrfs_abort_transaction(trans, ret);
1391 break;
1392 }
1393 btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF, new_bytenr,
1394 blocksize, 0);
1395 btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid, 0,
1396 true);
1397 ret = btrfs_inc_extent_ref(trans, &ref);
1398 if (ret) {
1399 btrfs_abort_transaction(trans, ret);
1400 break;
1401 }
1402
1403 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, new_bytenr,
1404 blocksize, path->nodes[level]->start);
1405 btrfs_init_tree_ref(&ref, level - 1, src->root_key.objectid,
1406 0, true);
1407 ret = btrfs_free_extent(trans, &ref);
1408 if (ret) {
1409 btrfs_abort_transaction(trans, ret);
1410 break;
1411 }
1412
1413 btrfs_init_generic_ref(&ref, BTRFS_DROP_DELAYED_REF, old_bytenr,
1414 blocksize, 0);
1415 btrfs_init_tree_ref(&ref, level - 1, dest->root_key.objectid,
1416 0, true);
1417 ret = btrfs_free_extent(trans, &ref);
1418 if (ret) {
1419 btrfs_abort_transaction(trans, ret);
1420 break;
1421 }
1422
1423 btrfs_unlock_up_safe(path, 0);
1424
1425 ret = level;
1426 break;
1427 }
1428 btrfs_tree_unlock(parent);
1429 free_extent_buffer(parent);
1430 return ret;
1431}
1432
1433/*
1434 * helper to find next relocated block in reloc tree
1435 */
1436static noinline_for_stack
1437int walk_up_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
1438 int *level)
1439{
1440 struct extent_buffer *eb;
1441 int i;
1442 u64 last_snapshot;
1443 u32 nritems;
1444
1445 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1446
1447 for (i = 0; i < *level; i++) {
1448 free_extent_buffer(path->nodes[i]);
1449 path->nodes[i] = NULL;
1450 }
1451
1452 for (i = *level; i < BTRFS_MAX_LEVEL && path->nodes[i]; i++) {
1453 eb = path->nodes[i];
1454 nritems = btrfs_header_nritems(eb);
1455 while (path->slots[i] + 1 < nritems) {
1456 path->slots[i]++;
1457 if (btrfs_node_ptr_generation(eb, path->slots[i]) <=
1458 last_snapshot)
1459 continue;
1460
1461 *level = i;
1462 return 0;
1463 }
1464 free_extent_buffer(path->nodes[i]);
1465 path->nodes[i] = NULL;
1466 }
1467 return 1;
1468}
1469
1470/*
1471 * walk down reloc tree to find relocated block of lowest level
1472 */
1473static noinline_for_stack
1474int walk_down_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
1475 int *level)
1476{
1477 struct extent_buffer *eb = NULL;
1478 int i;
1479 u64 ptr_gen = 0;
1480 u64 last_snapshot;
1481 u32 nritems;
1482
1483 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1484
1485 for (i = *level; i > 0; i--) {
1486 eb = path->nodes[i];
1487 nritems = btrfs_header_nritems(eb);
1488 while (path->slots[i] < nritems) {
1489 ptr_gen = btrfs_node_ptr_generation(eb, path->slots[i]);
1490 if (ptr_gen > last_snapshot)
1491 break;
1492 path->slots[i]++;
1493 }
1494 if (path->slots[i] >= nritems) {
1495 if (i == *level)
1496 break;
1497 *level = i + 1;
1498 return 0;
1499 }
1500 if (i == 1) {
1501 *level = i;
1502 return 0;
1503 }
1504
1505 eb = btrfs_read_node_slot(eb, path->slots[i]);
1506 if (IS_ERR(eb))
1507 return PTR_ERR(eb);
1508 BUG_ON(btrfs_header_level(eb) != i - 1);
1509 path->nodes[i - 1] = eb;
1510 path->slots[i - 1] = 0;
1511 }
1512 return 1;
1513}
1514
1515/*
1516 * invalidate extent cache for file extents whose key in range of
1517 * [min_key, max_key)
1518 */
1519static int invalidate_extent_cache(struct btrfs_root *root,
1520 struct btrfs_key *min_key,
1521 struct btrfs_key *max_key)
1522{
1523 struct btrfs_fs_info *fs_info = root->fs_info;
1524 struct inode *inode = NULL;
1525 u64 objectid;
1526 u64 start, end;
1527 u64 ino;
1528
1529 objectid = min_key->objectid;
1530 while (1) {
1531 struct extent_state *cached_state = NULL;
1532
1533 cond_resched();
1534 iput(inode);
1535
1536 if (objectid > max_key->objectid)
1537 break;
1538
1539 inode = find_next_inode(root, objectid);
1540 if (!inode)
1541 break;
1542 ino = btrfs_ino(BTRFS_I(inode));
1543
1544 if (ino > max_key->objectid) {
1545 iput(inode);
1546 break;
1547 }
1548
1549 objectid = ino + 1;
1550 if (!S_ISREG(inode->i_mode))
1551 continue;
1552
1553 if (unlikely(min_key->objectid == ino)) {
1554 if (min_key->type > BTRFS_EXTENT_DATA_KEY)
1555 continue;
1556 if (min_key->type < BTRFS_EXTENT_DATA_KEY)
1557 start = 0;
1558 else {
1559 start = min_key->offset;
1560 WARN_ON(!IS_ALIGNED(start, fs_info->sectorsize));
1561 }
1562 } else {
1563 start = 0;
1564 }
1565
1566 if (unlikely(max_key->objectid == ino)) {
1567 if (max_key->type < BTRFS_EXTENT_DATA_KEY)
1568 continue;
1569 if (max_key->type > BTRFS_EXTENT_DATA_KEY) {
1570 end = (u64)-1;
1571 } else {
1572 if (max_key->offset == 0)
1573 continue;
1574 end = max_key->offset;
1575 WARN_ON(!IS_ALIGNED(end, fs_info->sectorsize));
1576 end--;
1577 }
1578 } else {
1579 end = (u64)-1;
1580 }
1581
1582 /* the lock_extent waits for read_folio to complete */
1583 lock_extent(&BTRFS_I(inode)->io_tree, start, end, &cached_state);
1584 btrfs_drop_extent_map_range(BTRFS_I(inode), start, end, true);
1585 unlock_extent(&BTRFS_I(inode)->io_tree, start, end, &cached_state);
1586 }
1587 return 0;
1588}
1589
1590static int find_next_key(struct btrfs_path *path, int level,
1591 struct btrfs_key *key)
1592
1593{
1594 while (level < BTRFS_MAX_LEVEL) {
1595 if (!path->nodes[level])
1596 break;
1597 if (path->slots[level] + 1 <
1598 btrfs_header_nritems(path->nodes[level])) {
1599 btrfs_node_key_to_cpu(path->nodes[level], key,
1600 path->slots[level] + 1);
1601 return 0;
1602 }
1603 level++;
1604 }
1605 return 1;
1606}
1607
1608/*
1609 * Insert current subvolume into reloc_control::dirty_subvol_roots
1610 */
1611static int insert_dirty_subvol(struct btrfs_trans_handle *trans,
1612 struct reloc_control *rc,
1613 struct btrfs_root *root)
1614{
1615 struct btrfs_root *reloc_root = root->reloc_root;
1616 struct btrfs_root_item *reloc_root_item;
1617 int ret;
1618
1619 /* @root must be a subvolume tree root with a valid reloc tree */
1620 ASSERT(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
1621 ASSERT(reloc_root);
1622
1623 reloc_root_item = &reloc_root->root_item;
1624 memset(&reloc_root_item->drop_progress, 0,
1625 sizeof(reloc_root_item->drop_progress));
1626 btrfs_set_root_drop_level(reloc_root_item, 0);
1627 btrfs_set_root_refs(reloc_root_item, 0);
1628 ret = btrfs_update_reloc_root(trans, root);
1629 if (ret)
1630 return ret;
1631
1632 if (list_empty(&root->reloc_dirty_list)) {
1633 btrfs_grab_root(root);
1634 list_add_tail(&root->reloc_dirty_list, &rc->dirty_subvol_roots);
1635 }
1636
1637 return 0;
1638}
1639
1640static int clean_dirty_subvols(struct reloc_control *rc)
1641{
1642 struct btrfs_root *root;
1643 struct btrfs_root *next;
1644 int ret = 0;
1645 int ret2;
1646
1647 list_for_each_entry_safe(root, next, &rc->dirty_subvol_roots,
1648 reloc_dirty_list) {
1649 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
1650 /* Merged subvolume, cleanup its reloc root */
1651 struct btrfs_root *reloc_root = root->reloc_root;
1652
1653 list_del_init(&root->reloc_dirty_list);
1654 root->reloc_root = NULL;
1655 /*
1656 * Need barrier to ensure clear_bit() only happens after
1657 * root->reloc_root = NULL. Pairs with have_reloc_root.
1658 */
1659 smp_wmb();
1660 clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
1661 if (reloc_root) {
1662 /*
1663 * btrfs_drop_snapshot drops our ref we hold for
1664 * ->reloc_root. If it fails however we must
1665 * drop the ref ourselves.
1666 */
1667 ret2 = btrfs_drop_snapshot(reloc_root, 0, 1);
1668 if (ret2 < 0) {
1669 btrfs_put_root(reloc_root);
1670 if (!ret)
1671 ret = ret2;
1672 }
1673 }
1674 btrfs_put_root(root);
1675 } else {
1676 /* Orphan reloc tree, just clean it up */
1677 ret2 = btrfs_drop_snapshot(root, 0, 1);
1678 if (ret2 < 0) {
1679 btrfs_put_root(root);
1680 if (!ret)
1681 ret = ret2;
1682 }
1683 }
1684 }
1685 return ret;
1686}
1687
1688/*
1689 * merge the relocated tree blocks in reloc tree with corresponding
1690 * fs tree.
1691 */
1692static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
1693 struct btrfs_root *root)
1694{
1695 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
1696 struct btrfs_key key;
1697 struct btrfs_key next_key;
1698 struct btrfs_trans_handle *trans = NULL;
1699 struct btrfs_root *reloc_root;
1700 struct btrfs_root_item *root_item;
1701 struct btrfs_path *path;
1702 struct extent_buffer *leaf;
1703 int reserve_level;
1704 int level;
1705 int max_level;
1706 int replaced = 0;
1707 int ret = 0;
1708 u32 min_reserved;
1709
1710 path = btrfs_alloc_path();
1711 if (!path)
1712 return -ENOMEM;
1713 path->reada = READA_FORWARD;
1714
1715 reloc_root = root->reloc_root;
1716 root_item = &reloc_root->root_item;
1717
1718 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1719 level = btrfs_root_level(root_item);
1720 atomic_inc(&reloc_root->node->refs);
1721 path->nodes[level] = reloc_root->node;
1722 path->slots[level] = 0;
1723 } else {
1724 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1725
1726 level = btrfs_root_drop_level(root_item);
1727 BUG_ON(level == 0);
1728 path->lowest_level = level;
1729 ret = btrfs_search_slot(NULL, reloc_root, &key, path, 0, 0);
1730 path->lowest_level = 0;
1731 if (ret < 0) {
1732 btrfs_free_path(path);
1733 return ret;
1734 }
1735
1736 btrfs_node_key_to_cpu(path->nodes[level], &next_key,
1737 path->slots[level]);
1738 WARN_ON(memcmp(&key, &next_key, sizeof(key)));
1739
1740 btrfs_unlock_up_safe(path, 0);
1741 }
1742
1743 /*
1744 * In merge_reloc_root(), we modify the upper level pointer to swap the
1745 * tree blocks between reloc tree and subvolume tree. Thus for tree
1746 * block COW, we COW at most from level 1 to root level for each tree.
1747 *
1748 * Thus the needed metadata size is at most root_level * nodesize,
1749 * and * 2 since we have two trees to COW.
1750 */
1751 reserve_level = max_t(int, 1, btrfs_root_level(root_item));
1752 min_reserved = fs_info->nodesize * reserve_level * 2;
1753 memset(&next_key, 0, sizeof(next_key));
1754
1755 while (1) {
1756 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv,
1757 min_reserved,
1758 BTRFS_RESERVE_FLUSH_LIMIT);
1759 if (ret)
1760 goto out;
1761 trans = btrfs_start_transaction(root, 0);
1762 if (IS_ERR(trans)) {
1763 ret = PTR_ERR(trans);
1764 trans = NULL;
1765 goto out;
1766 }
1767
1768 /*
1769 * At this point we no longer have a reloc_control, so we can't
1770 * depend on btrfs_init_reloc_root to update our last_trans.
1771 *
1772 * But that's ok, we started the trans handle on our
1773 * corresponding fs_root, which means it's been added to the
1774 * dirty list. At commit time we'll still call
1775 * btrfs_update_reloc_root() and update our root item
1776 * appropriately.
1777 */
1778 reloc_root->last_trans = trans->transid;
1779 trans->block_rsv = rc->block_rsv;
1780
1781 replaced = 0;
1782 max_level = level;
1783
1784 ret = walk_down_reloc_tree(reloc_root, path, &level);
1785 if (ret < 0)
1786 goto out;
1787 if (ret > 0)
1788 break;
1789
1790 if (!find_next_key(path, level, &key) &&
1791 btrfs_comp_cpu_keys(&next_key, &key) >= 0) {
1792 ret = 0;
1793 } else {
1794 ret = replace_path(trans, rc, root, reloc_root, path,
1795 &next_key, level, max_level);
1796 }
1797 if (ret < 0)
1798 goto out;
1799 if (ret > 0) {
1800 level = ret;
1801 btrfs_node_key_to_cpu(path->nodes[level], &key,
1802 path->slots[level]);
1803 replaced = 1;
1804 }
1805
1806 ret = walk_up_reloc_tree(reloc_root, path, &level);
1807 if (ret > 0)
1808 break;
1809
1810 BUG_ON(level == 0);
1811 /*
1812 * save the merging progress in the drop_progress.
1813 * this is OK since root refs == 1 in this case.
1814 */
1815 btrfs_node_key(path->nodes[level], &root_item->drop_progress,
1816 path->slots[level]);
1817 btrfs_set_root_drop_level(root_item, level);
1818
1819 btrfs_end_transaction_throttle(trans);
1820 trans = NULL;
1821
1822 btrfs_btree_balance_dirty(fs_info);
1823
1824 if (replaced && rc->stage == UPDATE_DATA_PTRS)
1825 invalidate_extent_cache(root, &key, &next_key);
1826 }
1827
1828 /*
1829 * handle the case only one block in the fs tree need to be
1830 * relocated and the block is tree root.
1831 */
1832 leaf = btrfs_lock_root_node(root);
1833 ret = btrfs_cow_block(trans, root, leaf, NULL, 0, &leaf,
1834 BTRFS_NESTING_COW);
1835 btrfs_tree_unlock(leaf);
1836 free_extent_buffer(leaf);
1837out:
1838 btrfs_free_path(path);
1839
1840 if (ret == 0) {
1841 ret = insert_dirty_subvol(trans, rc, root);
1842 if (ret)
1843 btrfs_abort_transaction(trans, ret);
1844 }
1845
1846 if (trans)
1847 btrfs_end_transaction_throttle(trans);
1848
1849 btrfs_btree_balance_dirty(fs_info);
1850
1851 if (replaced && rc->stage == UPDATE_DATA_PTRS)
1852 invalidate_extent_cache(root, &key, &next_key);
1853
1854 return ret;
1855}
1856
1857static noinline_for_stack
1858int prepare_to_merge(struct reloc_control *rc, int err)
1859{
1860 struct btrfs_root *root = rc->extent_root;
1861 struct btrfs_fs_info *fs_info = root->fs_info;
1862 struct btrfs_root *reloc_root;
1863 struct btrfs_trans_handle *trans;
1864 LIST_HEAD(reloc_roots);
1865 u64 num_bytes = 0;
1866 int ret;
1867
1868 mutex_lock(&fs_info->reloc_mutex);
1869 rc->merging_rsv_size += fs_info->nodesize * (BTRFS_MAX_LEVEL - 1) * 2;
1870 rc->merging_rsv_size += rc->nodes_relocated * 2;
1871 mutex_unlock(&fs_info->reloc_mutex);
1872
1873again:
1874 if (!err) {
1875 num_bytes = rc->merging_rsv_size;
1876 ret = btrfs_block_rsv_add(fs_info, rc->block_rsv, num_bytes,
1877 BTRFS_RESERVE_FLUSH_ALL);
1878 if (ret)
1879 err = ret;
1880 }
1881
1882 trans = btrfs_join_transaction(rc->extent_root);
1883 if (IS_ERR(trans)) {
1884 if (!err)
1885 btrfs_block_rsv_release(fs_info, rc->block_rsv,
1886 num_bytes, NULL);
1887 return PTR_ERR(trans);
1888 }
1889
1890 if (!err) {
1891 if (num_bytes != rc->merging_rsv_size) {
1892 btrfs_end_transaction(trans);
1893 btrfs_block_rsv_release(fs_info, rc->block_rsv,
1894 num_bytes, NULL);
1895 goto again;
1896 }
1897 }
1898
1899 rc->merge_reloc_tree = 1;
1900
1901 while (!list_empty(&rc->reloc_roots)) {
1902 reloc_root = list_entry(rc->reloc_roots.next,
1903 struct btrfs_root, root_list);
1904 list_del_init(&reloc_root->root_list);
1905
1906 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
1907 false);
1908 if (IS_ERR(root)) {
1909 /*
1910 * Even if we have an error we need this reloc root
1911 * back on our list so we can clean up properly.
1912 */
1913 list_add(&reloc_root->root_list, &reloc_roots);
1914 btrfs_abort_transaction(trans, (int)PTR_ERR(root));
1915 if (!err)
1916 err = PTR_ERR(root);
1917 break;
1918 }
1919 ASSERT(root->reloc_root == reloc_root);
1920
1921 /*
1922 * set reference count to 1, so btrfs_recover_relocation
1923 * knows it should resumes merging
1924 */
1925 if (!err)
1926 btrfs_set_root_refs(&reloc_root->root_item, 1);
1927 ret = btrfs_update_reloc_root(trans, root);
1928
1929 /*
1930 * Even if we have an error we need this reloc root back on our
1931 * list so we can clean up properly.
1932 */
1933 list_add(&reloc_root->root_list, &reloc_roots);
1934 btrfs_put_root(root);
1935
1936 if (ret) {
1937 btrfs_abort_transaction(trans, ret);
1938 if (!err)
1939 err = ret;
1940 break;
1941 }
1942 }
1943
1944 list_splice(&reloc_roots, &rc->reloc_roots);
1945
1946 if (!err)
1947 err = btrfs_commit_transaction(trans);
1948 else
1949 btrfs_end_transaction(trans);
1950 return err;
1951}
1952
1953static noinline_for_stack
1954void free_reloc_roots(struct list_head *list)
1955{
1956 struct btrfs_root *reloc_root, *tmp;
1957
1958 list_for_each_entry_safe(reloc_root, tmp, list, root_list)
1959 __del_reloc_root(reloc_root);
1960}
1961
1962static noinline_for_stack
1963void merge_reloc_roots(struct reloc_control *rc)
1964{
1965 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
1966 struct btrfs_root *root;
1967 struct btrfs_root *reloc_root;
1968 LIST_HEAD(reloc_roots);
1969 int found = 0;
1970 int ret = 0;
1971again:
1972 root = rc->extent_root;
1973
1974 /*
1975 * this serializes us with btrfs_record_root_in_transaction,
1976 * we have to make sure nobody is in the middle of
1977 * adding their roots to the list while we are
1978 * doing this splice
1979 */
1980 mutex_lock(&fs_info->reloc_mutex);
1981 list_splice_init(&rc->reloc_roots, &reloc_roots);
1982 mutex_unlock(&fs_info->reloc_mutex);
1983
1984 while (!list_empty(&reloc_roots)) {
1985 found = 1;
1986 reloc_root = list_entry(reloc_roots.next,
1987 struct btrfs_root, root_list);
1988
1989 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
1990 false);
1991 if (btrfs_root_refs(&reloc_root->root_item) > 0) {
1992 if (IS_ERR(root)) {
1993 /*
1994 * For recovery we read the fs roots on mount,
1995 * and if we didn't find the root then we marked
1996 * the reloc root as a garbage root. For normal
1997 * relocation obviously the root should exist in
1998 * memory. However there's no reason we can't
1999 * handle the error properly here just in case.
2000 */
2001 ASSERT(0);
2002 ret = PTR_ERR(root);
2003 goto out;
2004 }
2005 if (root->reloc_root != reloc_root) {
2006 /*
2007 * This is actually impossible without something
2008 * going really wrong (like weird race condition
2009 * or cosmic rays).
2010 */
2011 ASSERT(0);
2012 ret = -EINVAL;
2013 goto out;
2014 }
2015 ret = merge_reloc_root(rc, root);
2016 btrfs_put_root(root);
2017 if (ret) {
2018 if (list_empty(&reloc_root->root_list))
2019 list_add_tail(&reloc_root->root_list,
2020 &reloc_roots);
2021 goto out;
2022 }
2023 } else {
2024 if (!IS_ERR(root)) {
2025 if (root->reloc_root == reloc_root) {
2026 root->reloc_root = NULL;
2027 btrfs_put_root(reloc_root);
2028 }
2029 clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE,
2030 &root->state);
2031 btrfs_put_root(root);
2032 }
2033
2034 list_del_init(&reloc_root->root_list);
2035 /* Don't forget to queue this reloc root for cleanup */
2036 list_add_tail(&reloc_root->reloc_dirty_list,
2037 &rc->dirty_subvol_roots);
2038 }
2039 }
2040
2041 if (found) {
2042 found = 0;
2043 goto again;
2044 }
2045out:
2046 if (ret) {
2047 btrfs_handle_fs_error(fs_info, ret, NULL);
2048 free_reloc_roots(&reloc_roots);
2049
2050 /* new reloc root may be added */
2051 mutex_lock(&fs_info->reloc_mutex);
2052 list_splice_init(&rc->reloc_roots, &reloc_roots);
2053 mutex_unlock(&fs_info->reloc_mutex);
2054 free_reloc_roots(&reloc_roots);
2055 }
2056
2057 /*
2058 * We used to have
2059 *
2060 * BUG_ON(!RB_EMPTY_ROOT(&rc->reloc_root_tree.rb_root));
2061 *
2062 * here, but it's wrong. If we fail to start the transaction in
2063 * prepare_to_merge() we will have only 0 ref reloc roots, none of which
2064 * have actually been removed from the reloc_root_tree rb tree. This is
2065 * fine because we're bailing here, and we hold a reference on the root
2066 * for the list that holds it, so these roots will be cleaned up when we
2067 * do the reloc_dirty_list afterwards. Meanwhile the root->reloc_root
2068 * will be cleaned up on unmount.
2069 *
2070 * The remaining nodes will be cleaned up by free_reloc_control.
2071 */
2072}
2073
2074static void free_block_list(struct rb_root *blocks)
2075{
2076 struct tree_block *block;
2077 struct rb_node *rb_node;
2078 while ((rb_node = rb_first(blocks))) {
2079 block = rb_entry(rb_node, struct tree_block, rb_node);
2080 rb_erase(rb_node, blocks);
2081 kfree(block);
2082 }
2083}
2084
2085static int record_reloc_root_in_trans(struct btrfs_trans_handle *trans,
2086 struct btrfs_root *reloc_root)
2087{
2088 struct btrfs_fs_info *fs_info = reloc_root->fs_info;
2089 struct btrfs_root *root;
2090 int ret;
2091
2092 if (reloc_root->last_trans == trans->transid)
2093 return 0;
2094
2095 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset, false);
2096
2097 /*
2098 * This should succeed, since we can't have a reloc root without having
2099 * already looked up the actual root and created the reloc root for this
2100 * root.
2101 *
2102 * However if there's some sort of corruption where we have a ref to a
2103 * reloc root without a corresponding root this could return ENOENT.
2104 */
2105 if (IS_ERR(root)) {
2106 ASSERT(0);
2107 return PTR_ERR(root);
2108 }
2109 if (root->reloc_root != reloc_root) {
2110 ASSERT(0);
2111 btrfs_err(fs_info,
2112 "root %llu has two reloc roots associated with it",
2113 reloc_root->root_key.offset);
2114 btrfs_put_root(root);
2115 return -EUCLEAN;
2116 }
2117 ret = btrfs_record_root_in_trans(trans, root);
2118 btrfs_put_root(root);
2119
2120 return ret;
2121}
2122
2123static noinline_for_stack
2124struct btrfs_root *select_reloc_root(struct btrfs_trans_handle *trans,
2125 struct reloc_control *rc,
2126 struct btrfs_backref_node *node,
2127 struct btrfs_backref_edge *edges[])
2128{
2129 struct btrfs_backref_node *next;
2130 struct btrfs_root *root;
2131 int index = 0;
2132 int ret;
2133
2134 next = node;
2135 while (1) {
2136 cond_resched();
2137 next = walk_up_backref(next, edges, &index);
2138 root = next->root;
2139
2140 /*
2141 * If there is no root, then our references for this block are
2142 * incomplete, as we should be able to walk all the way up to a
2143 * block that is owned by a root.
2144 *
2145 * This path is only for SHAREABLE roots, so if we come upon a
2146 * non-SHAREABLE root then we have backrefs that resolve
2147 * improperly.
2148 *
2149 * Both of these cases indicate file system corruption, or a bug
2150 * in the backref walking code.
2151 */
2152 if (!root) {
2153 ASSERT(0);
2154 btrfs_err(trans->fs_info,
2155 "bytenr %llu doesn't have a backref path ending in a root",
2156 node->bytenr);
2157 return ERR_PTR(-EUCLEAN);
2158 }
2159 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) {
2160 ASSERT(0);
2161 btrfs_err(trans->fs_info,
2162 "bytenr %llu has multiple refs with one ending in a non-shareable root",
2163 node->bytenr);
2164 return ERR_PTR(-EUCLEAN);
2165 }
2166
2167 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) {
2168 ret = record_reloc_root_in_trans(trans, root);
2169 if (ret)
2170 return ERR_PTR(ret);
2171 break;
2172 }
2173
2174 ret = btrfs_record_root_in_trans(trans, root);
2175 if (ret)
2176 return ERR_PTR(ret);
2177 root = root->reloc_root;
2178
2179 /*
2180 * We could have raced with another thread which failed, so
2181 * root->reloc_root may not be set, return ENOENT in this case.
2182 */
2183 if (!root)
2184 return ERR_PTR(-ENOENT);
2185
2186 if (next->new_bytenr != root->node->start) {
2187 /*
2188 * We just created the reloc root, so we shouldn't have
2189 * ->new_bytenr set and this shouldn't be in the changed
2190 * list. If it is then we have multiple roots pointing
2191 * at the same bytenr which indicates corruption, or
2192 * we've made a mistake in the backref walking code.
2193 */
2194 ASSERT(next->new_bytenr == 0);
2195 ASSERT(list_empty(&next->list));
2196 if (next->new_bytenr || !list_empty(&next->list)) {
2197 btrfs_err(trans->fs_info,
2198 "bytenr %llu possibly has multiple roots pointing at the same bytenr %llu",
2199 node->bytenr, next->bytenr);
2200 return ERR_PTR(-EUCLEAN);
2201 }
2202
2203 next->new_bytenr = root->node->start;
2204 btrfs_put_root(next->root);
2205 next->root = btrfs_grab_root(root);
2206 ASSERT(next->root);
2207 list_add_tail(&next->list,
2208 &rc->backref_cache.changed);
2209 mark_block_processed(rc, next);
2210 break;
2211 }
2212
2213 WARN_ON(1);
2214 root = NULL;
2215 next = walk_down_backref(edges, &index);
2216 if (!next || next->level <= node->level)
2217 break;
2218 }
2219 if (!root) {
2220 /*
2221 * This can happen if there's fs corruption or if there's a bug
2222 * in the backref lookup code.
2223 */
2224 ASSERT(0);
2225 return ERR_PTR(-ENOENT);
2226 }
2227
2228 next = node;
2229 /* setup backref node path for btrfs_reloc_cow_block */
2230 while (1) {
2231 rc->backref_cache.path[next->level] = next;
2232 if (--index < 0)
2233 break;
2234 next = edges[index]->node[UPPER];
2235 }
2236 return root;
2237}
2238
2239/*
2240 * Select a tree root for relocation.
2241 *
2242 * Return NULL if the block is not shareable. We should use do_relocation() in
2243 * this case.
2244 *
2245 * Return a tree root pointer if the block is shareable.
2246 * Return -ENOENT if the block is root of reloc tree.
2247 */
2248static noinline_for_stack
2249struct btrfs_root *select_one_root(struct btrfs_backref_node *node)
2250{
2251 struct btrfs_backref_node *next;
2252 struct btrfs_root *root;
2253 struct btrfs_root *fs_root = NULL;
2254 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2255 int index = 0;
2256
2257 next = node;
2258 while (1) {
2259 cond_resched();
2260 next = walk_up_backref(next, edges, &index);
2261 root = next->root;
2262
2263 /*
2264 * This can occur if we have incomplete extent refs leading all
2265 * the way up a particular path, in this case return -EUCLEAN.
2266 */
2267 if (!root)
2268 return ERR_PTR(-EUCLEAN);
2269
2270 /* No other choice for non-shareable tree */
2271 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
2272 return root;
2273
2274 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID)
2275 fs_root = root;
2276
2277 if (next != node)
2278 return NULL;
2279
2280 next = walk_down_backref(edges, &index);
2281 if (!next || next->level <= node->level)
2282 break;
2283 }
2284
2285 if (!fs_root)
2286 return ERR_PTR(-ENOENT);
2287 return fs_root;
2288}
2289
2290static noinline_for_stack
2291u64 calcu_metadata_size(struct reloc_control *rc,
2292 struct btrfs_backref_node *node, int reserve)
2293{
2294 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
2295 struct btrfs_backref_node *next = node;
2296 struct btrfs_backref_edge *edge;
2297 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2298 u64 num_bytes = 0;
2299 int index = 0;
2300
2301 BUG_ON(reserve && node->processed);
2302
2303 while (next) {
2304 cond_resched();
2305 while (1) {
2306 if (next->processed && (reserve || next != node))
2307 break;
2308
2309 num_bytes += fs_info->nodesize;
2310
2311 if (list_empty(&next->upper))
2312 break;
2313
2314 edge = list_entry(next->upper.next,
2315 struct btrfs_backref_edge, list[LOWER]);
2316 edges[index++] = edge;
2317 next = edge->node[UPPER];
2318 }
2319 next = walk_down_backref(edges, &index);
2320 }
2321 return num_bytes;
2322}
2323
2324static int reserve_metadata_space(struct btrfs_trans_handle *trans,
2325 struct reloc_control *rc,
2326 struct btrfs_backref_node *node)
2327{
2328 struct btrfs_root *root = rc->extent_root;
2329 struct btrfs_fs_info *fs_info = root->fs_info;
2330 u64 num_bytes;
2331 int ret;
2332 u64 tmp;
2333
2334 num_bytes = calcu_metadata_size(rc, node, 1) * 2;
2335
2336 trans->block_rsv = rc->block_rsv;
2337 rc->reserved_bytes += num_bytes;
2338
2339 /*
2340 * We are under a transaction here so we can only do limited flushing.
2341 * If we get an enospc just kick back -EAGAIN so we know to drop the
2342 * transaction and try to refill when we can flush all the things.
2343 */
2344 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv, num_bytes,
2345 BTRFS_RESERVE_FLUSH_LIMIT);
2346 if (ret) {
2347 tmp = fs_info->nodesize * RELOCATION_RESERVED_NODES;
2348 while (tmp <= rc->reserved_bytes)
2349 tmp <<= 1;
2350 /*
2351 * only one thread can access block_rsv at this point,
2352 * so we don't need hold lock to protect block_rsv.
2353 * we expand more reservation size here to allow enough
2354 * space for relocation and we will return earlier in
2355 * enospc case.
2356 */
2357 rc->block_rsv->size = tmp + fs_info->nodesize *
2358 RELOCATION_RESERVED_NODES;
2359 return -EAGAIN;
2360 }
2361
2362 return 0;
2363}
2364
2365/*
2366 * relocate a block tree, and then update pointers in upper level
2367 * blocks that reference the block to point to the new location.
2368 *
2369 * if called by link_to_upper, the block has already been relocated.
2370 * in that case this function just updates pointers.
2371 */
2372static int do_relocation(struct btrfs_trans_handle *trans,
2373 struct reloc_control *rc,
2374 struct btrfs_backref_node *node,
2375 struct btrfs_key *key,
2376 struct btrfs_path *path, int lowest)
2377{
2378 struct btrfs_backref_node *upper;
2379 struct btrfs_backref_edge *edge;
2380 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2381 struct btrfs_root *root;
2382 struct extent_buffer *eb;
2383 u32 blocksize;
2384 u64 bytenr;
2385 int slot;
2386 int ret = 0;
2387
2388 /*
2389 * If we are lowest then this is the first time we're processing this
2390 * block, and thus shouldn't have an eb associated with it yet.
2391 */
2392 ASSERT(!lowest || !node->eb);
2393
2394 path->lowest_level = node->level + 1;
2395 rc->backref_cache.path[node->level] = node;
2396 list_for_each_entry(edge, &node->upper, list[LOWER]) {
2397 struct btrfs_ref ref = { 0 };
2398
2399 cond_resched();
2400
2401 upper = edge->node[UPPER];
2402 root = select_reloc_root(trans, rc, upper, edges);
2403 if (IS_ERR(root)) {
2404 ret = PTR_ERR(root);
2405 goto next;
2406 }
2407
2408 if (upper->eb && !upper->locked) {
2409 if (!lowest) {
2410 ret = btrfs_bin_search(upper->eb, key, &slot);
2411 if (ret < 0)
2412 goto next;
2413 BUG_ON(ret);
2414 bytenr = btrfs_node_blockptr(upper->eb, slot);
2415 if (node->eb->start == bytenr)
2416 goto next;
2417 }
2418 btrfs_backref_drop_node_buffer(upper);
2419 }
2420
2421 if (!upper->eb) {
2422 ret = btrfs_search_slot(trans, root, key, path, 0, 1);
2423 if (ret) {
2424 if (ret > 0)
2425 ret = -ENOENT;
2426
2427 btrfs_release_path(path);
2428 break;
2429 }
2430
2431 if (!upper->eb) {
2432 upper->eb = path->nodes[upper->level];
2433 path->nodes[upper->level] = NULL;
2434 } else {
2435 BUG_ON(upper->eb != path->nodes[upper->level]);
2436 }
2437
2438 upper->locked = 1;
2439 path->locks[upper->level] = 0;
2440
2441 slot = path->slots[upper->level];
2442 btrfs_release_path(path);
2443 } else {
2444 ret = btrfs_bin_search(upper->eb, key, &slot);
2445 if (ret < 0)
2446 goto next;
2447 BUG_ON(ret);
2448 }
2449
2450 bytenr = btrfs_node_blockptr(upper->eb, slot);
2451 if (lowest) {
2452 if (bytenr != node->bytenr) {
2453 btrfs_err(root->fs_info,
2454 "lowest leaf/node mismatch: bytenr %llu node->bytenr %llu slot %d upper %llu",
2455 bytenr, node->bytenr, slot,
2456 upper->eb->start);
2457 ret = -EIO;
2458 goto next;
2459 }
2460 } else {
2461 if (node->eb->start == bytenr)
2462 goto next;
2463 }
2464
2465 blocksize = root->fs_info->nodesize;
2466 eb = btrfs_read_node_slot(upper->eb, slot);
2467 if (IS_ERR(eb)) {
2468 ret = PTR_ERR(eb);
2469 goto next;
2470 }
2471 btrfs_tree_lock(eb);
2472
2473 if (!node->eb) {
2474 ret = btrfs_cow_block(trans, root, eb, upper->eb,
2475 slot, &eb, BTRFS_NESTING_COW);
2476 btrfs_tree_unlock(eb);
2477 free_extent_buffer(eb);
2478 if (ret < 0)
2479 goto next;
2480 /*
2481 * We've just COWed this block, it should have updated
2482 * the correct backref node entry.
2483 */
2484 ASSERT(node->eb == eb);
2485 } else {
2486 btrfs_set_node_blockptr(upper->eb, slot,
2487 node->eb->start);
2488 btrfs_set_node_ptr_generation(upper->eb, slot,
2489 trans->transid);
2490 btrfs_mark_buffer_dirty(upper->eb);
2491
2492 btrfs_init_generic_ref(&ref, BTRFS_ADD_DELAYED_REF,
2493 node->eb->start, blocksize,
2494 upper->eb->start);
2495 btrfs_init_tree_ref(&ref, node->level,
2496 btrfs_header_owner(upper->eb),
2497 root->root_key.objectid, false);
2498 ret = btrfs_inc_extent_ref(trans, &ref);
2499 if (!ret)
2500 ret = btrfs_drop_subtree(trans, root, eb,
2501 upper->eb);
2502 if (ret)
2503 btrfs_abort_transaction(trans, ret);
2504 }
2505next:
2506 if (!upper->pending)
2507 btrfs_backref_drop_node_buffer(upper);
2508 else
2509 btrfs_backref_unlock_node_buffer(upper);
2510 if (ret)
2511 break;
2512 }
2513
2514 if (!ret && node->pending) {
2515 btrfs_backref_drop_node_buffer(node);
2516 list_move_tail(&node->list, &rc->backref_cache.changed);
2517 node->pending = 0;
2518 }
2519
2520 path->lowest_level = 0;
2521
2522 /*
2523 * We should have allocated all of our space in the block rsv and thus
2524 * shouldn't ENOSPC.
2525 */
2526 ASSERT(ret != -ENOSPC);
2527 return ret;
2528}
2529
2530static int link_to_upper(struct btrfs_trans_handle *trans,
2531 struct reloc_control *rc,
2532 struct btrfs_backref_node *node,
2533 struct btrfs_path *path)
2534{
2535 struct btrfs_key key;
2536
2537 btrfs_node_key_to_cpu(node->eb, &key, 0);
2538 return do_relocation(trans, rc, node, &key, path, 0);
2539}
2540
2541static int finish_pending_nodes(struct btrfs_trans_handle *trans,
2542 struct reloc_control *rc,
2543 struct btrfs_path *path, int err)
2544{
2545 LIST_HEAD(list);
2546 struct btrfs_backref_cache *cache = &rc->backref_cache;
2547 struct btrfs_backref_node *node;
2548 int level;
2549 int ret;
2550
2551 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2552 while (!list_empty(&cache->pending[level])) {
2553 node = list_entry(cache->pending[level].next,
2554 struct btrfs_backref_node, list);
2555 list_move_tail(&node->list, &list);
2556 BUG_ON(!node->pending);
2557
2558 if (!err) {
2559 ret = link_to_upper(trans, rc, node, path);
2560 if (ret < 0)
2561 err = ret;
2562 }
2563 }
2564 list_splice_init(&list, &cache->pending[level]);
2565 }
2566 return err;
2567}
2568
2569/*
2570 * mark a block and all blocks directly/indirectly reference the block
2571 * as processed.
2572 */
2573static void update_processed_blocks(struct reloc_control *rc,
2574 struct btrfs_backref_node *node)
2575{
2576 struct btrfs_backref_node *next = node;
2577 struct btrfs_backref_edge *edge;
2578 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2579 int index = 0;
2580
2581 while (next) {
2582 cond_resched();
2583 while (1) {
2584 if (next->processed)
2585 break;
2586
2587 mark_block_processed(rc, next);
2588
2589 if (list_empty(&next->upper))
2590 break;
2591
2592 edge = list_entry(next->upper.next,
2593 struct btrfs_backref_edge, list[LOWER]);
2594 edges[index++] = edge;
2595 next = edge->node[UPPER];
2596 }
2597 next = walk_down_backref(edges, &index);
2598 }
2599}
2600
2601static int tree_block_processed(u64 bytenr, struct reloc_control *rc)
2602{
2603 u32 blocksize = rc->extent_root->fs_info->nodesize;
2604
2605 if (test_range_bit(&rc->processed_blocks, bytenr,
2606 bytenr + blocksize - 1, EXTENT_DIRTY, 1, NULL))
2607 return 1;
2608 return 0;
2609}
2610
2611static int get_tree_block_key(struct btrfs_fs_info *fs_info,
2612 struct tree_block *block)
2613{
2614 struct btrfs_tree_parent_check check = {
2615 .level = block->level,
2616 .owner_root = block->owner,
2617 .transid = block->key.offset
2618 };
2619 struct extent_buffer *eb;
2620
2621 eb = read_tree_block(fs_info, block->bytenr, &check);
2622 if (IS_ERR(eb))
2623 return PTR_ERR(eb);
2624 if (!extent_buffer_uptodate(eb)) {
2625 free_extent_buffer(eb);
2626 return -EIO;
2627 }
2628 if (block->level == 0)
2629 btrfs_item_key_to_cpu(eb, &block->key, 0);
2630 else
2631 btrfs_node_key_to_cpu(eb, &block->key, 0);
2632 free_extent_buffer(eb);
2633 block->key_ready = 1;
2634 return 0;
2635}
2636
2637/*
2638 * helper function to relocate a tree block
2639 */
2640static int relocate_tree_block(struct btrfs_trans_handle *trans,
2641 struct reloc_control *rc,
2642 struct btrfs_backref_node *node,
2643 struct btrfs_key *key,
2644 struct btrfs_path *path)
2645{
2646 struct btrfs_root *root;
2647 int ret = 0;
2648
2649 if (!node)
2650 return 0;
2651
2652 /*
2653 * If we fail here we want to drop our backref_node because we are going
2654 * to start over and regenerate the tree for it.
2655 */
2656 ret = reserve_metadata_space(trans, rc, node);
2657 if (ret)
2658 goto out;
2659
2660 BUG_ON(node->processed);
2661 root = select_one_root(node);
2662 if (IS_ERR(root)) {
2663 ret = PTR_ERR(root);
2664
2665 /* See explanation in select_one_root for the -EUCLEAN case. */
2666 ASSERT(ret == -ENOENT);
2667 if (ret == -ENOENT) {
2668 ret = 0;
2669 update_processed_blocks(rc, node);
2670 }
2671 goto out;
2672 }
2673
2674 if (root) {
2675 if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) {
2676 /*
2677 * This block was the root block of a root, and this is
2678 * the first time we're processing the block and thus it
2679 * should not have had the ->new_bytenr modified and
2680 * should have not been included on the changed list.
2681 *
2682 * However in the case of corruption we could have
2683 * multiple refs pointing to the same block improperly,
2684 * and thus we would trip over these checks. ASSERT()
2685 * for the developer case, because it could indicate a
2686 * bug in the backref code, however error out for a
2687 * normal user in the case of corruption.
2688 */
2689 ASSERT(node->new_bytenr == 0);
2690 ASSERT(list_empty(&node->list));
2691 if (node->new_bytenr || !list_empty(&node->list)) {
2692 btrfs_err(root->fs_info,
2693 "bytenr %llu has improper references to it",
2694 node->bytenr);
2695 ret = -EUCLEAN;
2696 goto out;
2697 }
2698 ret = btrfs_record_root_in_trans(trans, root);
2699 if (ret)
2700 goto out;
2701 /*
2702 * Another thread could have failed, need to check if we
2703 * have reloc_root actually set.
2704 */
2705 if (!root->reloc_root) {
2706 ret = -ENOENT;
2707 goto out;
2708 }
2709 root = root->reloc_root;
2710 node->new_bytenr = root->node->start;
2711 btrfs_put_root(node->root);
2712 node->root = btrfs_grab_root(root);
2713 ASSERT(node->root);
2714 list_add_tail(&node->list, &rc->backref_cache.changed);
2715 } else {
2716 path->lowest_level = node->level;
2717 if (root == root->fs_info->chunk_root)
2718 btrfs_reserve_chunk_metadata(trans, false);
2719 ret = btrfs_search_slot(trans, root, key, path, 0, 1);
2720 btrfs_release_path(path);
2721 if (root == root->fs_info->chunk_root)
2722 btrfs_trans_release_chunk_metadata(trans);
2723 if (ret > 0)
2724 ret = 0;
2725 }
2726 if (!ret)
2727 update_processed_blocks(rc, node);
2728 } else {
2729 ret = do_relocation(trans, rc, node, key, path, 1);
2730 }
2731out:
2732 if (ret || node->level == 0 || node->cowonly)
2733 btrfs_backref_cleanup_node(&rc->backref_cache, node);
2734 return ret;
2735}
2736
2737/*
2738 * relocate a list of blocks
2739 */
2740static noinline_for_stack
2741int relocate_tree_blocks(struct btrfs_trans_handle *trans,
2742 struct reloc_control *rc, struct rb_root *blocks)
2743{
2744 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
2745 struct btrfs_backref_node *node;
2746 struct btrfs_path *path;
2747 struct tree_block *block;
2748 struct tree_block *next;
2749 int ret;
2750 int err = 0;
2751
2752 path = btrfs_alloc_path();
2753 if (!path) {
2754 err = -ENOMEM;
2755 goto out_free_blocks;
2756 }
2757
2758 /* Kick in readahead for tree blocks with missing keys */
2759 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2760 if (!block->key_ready)
2761 btrfs_readahead_tree_block(fs_info, block->bytenr,
2762 block->owner, 0,
2763 block->level);
2764 }
2765
2766 /* Get first keys */
2767 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2768 if (!block->key_ready) {
2769 err = get_tree_block_key(fs_info, block);
2770 if (err)
2771 goto out_free_path;
2772 }
2773 }
2774
2775 /* Do tree relocation */
2776 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2777 node = build_backref_tree(rc, &block->key,
2778 block->level, block->bytenr);
2779 if (IS_ERR(node)) {
2780 err = PTR_ERR(node);
2781 goto out;
2782 }
2783
2784 ret = relocate_tree_block(trans, rc, node, &block->key,
2785 path);
2786 if (ret < 0) {
2787 err = ret;
2788 break;
2789 }
2790 }
2791out:
2792 err = finish_pending_nodes(trans, rc, path, err);
2793
2794out_free_path:
2795 btrfs_free_path(path);
2796out_free_blocks:
2797 free_block_list(blocks);
2798 return err;
2799}
2800
2801static noinline_for_stack int prealloc_file_extent_cluster(
2802 struct btrfs_inode *inode,
2803 struct file_extent_cluster *cluster)
2804{
2805 u64 alloc_hint = 0;
2806 u64 start;
2807 u64 end;
2808 u64 offset = inode->index_cnt;
2809 u64 num_bytes;
2810 int nr;
2811 int ret = 0;
2812 u64 i_size = i_size_read(&inode->vfs_inode);
2813 u64 prealloc_start = cluster->start - offset;
2814 u64 prealloc_end = cluster->end - offset;
2815 u64 cur_offset = prealloc_start;
2816
2817 /*
2818 * For subpage case, previous i_size may not be aligned to PAGE_SIZE.
2819 * This means the range [i_size, PAGE_END + 1) is filled with zeros by
2820 * btrfs_do_readpage() call of previously relocated file cluster.
2821 *
2822 * If the current cluster starts in the above range, btrfs_do_readpage()
2823 * will skip the read, and relocate_one_page() will later writeback
2824 * the padding zeros as new data, causing data corruption.
2825 *
2826 * Here we have to manually invalidate the range (i_size, PAGE_END + 1).
2827 */
2828 if (!IS_ALIGNED(i_size, PAGE_SIZE)) {
2829 struct address_space *mapping = inode->vfs_inode.i_mapping;
2830 struct btrfs_fs_info *fs_info = inode->root->fs_info;
2831 const u32 sectorsize = fs_info->sectorsize;
2832 struct page *page;
2833
2834 ASSERT(sectorsize < PAGE_SIZE);
2835 ASSERT(IS_ALIGNED(i_size, sectorsize));
2836
2837 /*
2838 * Subpage can't handle page with DIRTY but without UPTODATE
2839 * bit as it can lead to the following deadlock:
2840 *
2841 * btrfs_read_folio()
2842 * | Page already *locked*
2843 * |- btrfs_lock_and_flush_ordered_range()
2844 * |- btrfs_start_ordered_extent()
2845 * |- extent_write_cache_pages()
2846 * |- lock_page()
2847 * We try to lock the page we already hold.
2848 *
2849 * Here we just writeback the whole data reloc inode, so that
2850 * we will be ensured to have no dirty range in the page, and
2851 * are safe to clear the uptodate bits.
2852 *
2853 * This shouldn't cause too much overhead, as we need to write
2854 * the data back anyway.
2855 */
2856 ret = filemap_write_and_wait(mapping);
2857 if (ret < 0)
2858 return ret;
2859
2860 clear_extent_bits(&inode->io_tree, i_size,
2861 round_up(i_size, PAGE_SIZE) - 1,
2862 EXTENT_UPTODATE);
2863 page = find_lock_page(mapping, i_size >> PAGE_SHIFT);
2864 /*
2865 * If page is freed we don't need to do anything then, as we
2866 * will re-read the whole page anyway.
2867 */
2868 if (page) {
2869 btrfs_subpage_clear_uptodate(fs_info, page, i_size,
2870 round_up(i_size, PAGE_SIZE) - i_size);
2871 unlock_page(page);
2872 put_page(page);
2873 }
2874 }
2875
2876 BUG_ON(cluster->start != cluster->boundary[0]);
2877 ret = btrfs_alloc_data_chunk_ondemand(inode,
2878 prealloc_end + 1 - prealloc_start);
2879 if (ret)
2880 return ret;
2881
2882 btrfs_inode_lock(inode, 0);
2883 for (nr = 0; nr < cluster->nr; nr++) {
2884 struct extent_state *cached_state = NULL;
2885
2886 start = cluster->boundary[nr] - offset;
2887 if (nr + 1 < cluster->nr)
2888 end = cluster->boundary[nr + 1] - 1 - offset;
2889 else
2890 end = cluster->end - offset;
2891
2892 lock_extent(&inode->io_tree, start, end, &cached_state);
2893 num_bytes = end + 1 - start;
2894 ret = btrfs_prealloc_file_range(&inode->vfs_inode, 0, start,
2895 num_bytes, num_bytes,
2896 end + 1, &alloc_hint);
2897 cur_offset = end + 1;
2898 unlock_extent(&inode->io_tree, start, end, &cached_state);
2899 if (ret)
2900 break;
2901 }
2902 btrfs_inode_unlock(inode, 0);
2903
2904 if (cur_offset < prealloc_end)
2905 btrfs_free_reserved_data_space_noquota(inode->root->fs_info,
2906 prealloc_end + 1 - cur_offset);
2907 return ret;
2908}
2909
2910static noinline_for_stack int setup_relocation_extent_mapping(struct inode *inode,
2911 u64 start, u64 end, u64 block_start)
2912{
2913 struct extent_map *em;
2914 struct extent_state *cached_state = NULL;
2915 int ret = 0;
2916
2917 em = alloc_extent_map();
2918 if (!em)
2919 return -ENOMEM;
2920
2921 em->start = start;
2922 em->len = end + 1 - start;
2923 em->block_len = em->len;
2924 em->block_start = block_start;
2925 set_bit(EXTENT_FLAG_PINNED, &em->flags);
2926
2927 lock_extent(&BTRFS_I(inode)->io_tree, start, end, &cached_state);
2928 ret = btrfs_replace_extent_map_range(BTRFS_I(inode), em, false);
2929 unlock_extent(&BTRFS_I(inode)->io_tree, start, end, &cached_state);
2930 free_extent_map(em);
2931
2932 return ret;
2933}
2934
2935/*
2936 * Allow error injection to test balance/relocation cancellation
2937 */
2938noinline int btrfs_should_cancel_balance(struct btrfs_fs_info *fs_info)
2939{
2940 return atomic_read(&fs_info->balance_cancel_req) ||
2941 atomic_read(&fs_info->reloc_cancel_req) ||
2942 fatal_signal_pending(current);
2943}
2944ALLOW_ERROR_INJECTION(btrfs_should_cancel_balance, TRUE);
2945
2946static u64 get_cluster_boundary_end(struct file_extent_cluster *cluster,
2947 int cluster_nr)
2948{
2949 /* Last extent, use cluster end directly */
2950 if (cluster_nr >= cluster->nr - 1)
2951 return cluster->end;
2952
2953 /* Use next boundary start*/
2954 return cluster->boundary[cluster_nr + 1] - 1;
2955}
2956
2957static int relocate_one_page(struct inode *inode, struct file_ra_state *ra,
2958 struct file_extent_cluster *cluster,
2959 int *cluster_nr, unsigned long page_index)
2960{
2961 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2962 u64 offset = BTRFS_I(inode)->index_cnt;
2963 const unsigned long last_index = (cluster->end - offset) >> PAGE_SHIFT;
2964 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
2965 struct page *page;
2966 u64 page_start;
2967 u64 page_end;
2968 u64 cur;
2969 int ret;
2970
2971 ASSERT(page_index <= last_index);
2972 page = find_lock_page(inode->i_mapping, page_index);
2973 if (!page) {
2974 page_cache_sync_readahead(inode->i_mapping, ra, NULL,
2975 page_index, last_index + 1 - page_index);
2976 page = find_or_create_page(inode->i_mapping, page_index, mask);
2977 if (!page)
2978 return -ENOMEM;
2979 }
2980 ret = set_page_extent_mapped(page);
2981 if (ret < 0)
2982 goto release_page;
2983
2984 if (PageReadahead(page))
2985 page_cache_async_readahead(inode->i_mapping, ra, NULL,
2986 page_folio(page), page_index,
2987 last_index + 1 - page_index);
2988
2989 if (!PageUptodate(page)) {
2990 btrfs_read_folio(NULL, page_folio(page));
2991 lock_page(page);
2992 if (!PageUptodate(page)) {
2993 ret = -EIO;
2994 goto release_page;
2995 }
2996 }
2997
2998 page_start = page_offset(page);
2999 page_end = page_start + PAGE_SIZE - 1;
3000
3001 /*
3002 * Start from the cluster, as for subpage case, the cluster can start
3003 * inside the page.
3004 */
3005 cur = max(page_start, cluster->boundary[*cluster_nr] - offset);
3006 while (cur <= page_end) {
3007 struct extent_state *cached_state = NULL;
3008 u64 extent_start = cluster->boundary[*cluster_nr] - offset;
3009 u64 extent_end = get_cluster_boundary_end(cluster,
3010 *cluster_nr) - offset;
3011 u64 clamped_start = max(page_start, extent_start);
3012 u64 clamped_end = min(page_end, extent_end);
3013 u32 clamped_len = clamped_end + 1 - clamped_start;
3014
3015 /* Reserve metadata for this range */
3016 ret = btrfs_delalloc_reserve_metadata(BTRFS_I(inode),
3017 clamped_len, clamped_len,
3018 false);
3019 if (ret)
3020 goto release_page;
3021
3022 /* Mark the range delalloc and dirty for later writeback */
3023 lock_extent(&BTRFS_I(inode)->io_tree, clamped_start, clamped_end,
3024 &cached_state);
3025 ret = btrfs_set_extent_delalloc(BTRFS_I(inode), clamped_start,
3026 clamped_end, 0, &cached_state);
3027 if (ret) {
3028 clear_extent_bit(&BTRFS_I(inode)->io_tree,
3029 clamped_start, clamped_end,
3030 EXTENT_LOCKED | EXTENT_BOUNDARY,
3031 &cached_state);
3032 btrfs_delalloc_release_metadata(BTRFS_I(inode),
3033 clamped_len, true);
3034 btrfs_delalloc_release_extents(BTRFS_I(inode),
3035 clamped_len);
3036 goto release_page;
3037 }
3038 btrfs_page_set_dirty(fs_info, page, clamped_start, clamped_len);
3039
3040 /*
3041 * Set the boundary if it's inside the page.
3042 * Data relocation requires the destination extents to have the
3043 * same size as the source.
3044 * EXTENT_BOUNDARY bit prevents current extent from being merged
3045 * with previous extent.
3046 */
3047 if (in_range(cluster->boundary[*cluster_nr] - offset,
3048 page_start, PAGE_SIZE)) {
3049 u64 boundary_start = cluster->boundary[*cluster_nr] -
3050 offset;
3051 u64 boundary_end = boundary_start +
3052 fs_info->sectorsize - 1;
3053
3054 set_extent_bits(&BTRFS_I(inode)->io_tree,
3055 boundary_start, boundary_end,
3056 EXTENT_BOUNDARY);
3057 }
3058 unlock_extent(&BTRFS_I(inode)->io_tree, clamped_start, clamped_end,
3059 &cached_state);
3060 btrfs_delalloc_release_extents(BTRFS_I(inode), clamped_len);
3061 cur += clamped_len;
3062
3063 /* Crossed extent end, go to next extent */
3064 if (cur >= extent_end) {
3065 (*cluster_nr)++;
3066 /* Just finished the last extent of the cluster, exit. */
3067 if (*cluster_nr >= cluster->nr)
3068 break;
3069 }
3070 }
3071 unlock_page(page);
3072 put_page(page);
3073
3074 balance_dirty_pages_ratelimited(inode->i_mapping);
3075 btrfs_throttle(fs_info);
3076 if (btrfs_should_cancel_balance(fs_info))
3077 ret = -ECANCELED;
3078 return ret;
3079
3080release_page:
3081 unlock_page(page);
3082 put_page(page);
3083 return ret;
3084}
3085
3086static int relocate_file_extent_cluster(struct inode *inode,
3087 struct file_extent_cluster *cluster)
3088{
3089 u64 offset = BTRFS_I(inode)->index_cnt;
3090 unsigned long index;
3091 unsigned long last_index;
3092 struct file_ra_state *ra;
3093 int cluster_nr = 0;
3094 int ret = 0;
3095
3096 if (!cluster->nr)
3097 return 0;
3098
3099 ra = kzalloc(sizeof(*ra), GFP_NOFS);
3100 if (!ra)
3101 return -ENOMEM;
3102
3103 ret = prealloc_file_extent_cluster(BTRFS_I(inode), cluster);
3104 if (ret)
3105 goto out;
3106
3107 file_ra_state_init(ra, inode->i_mapping);
3108
3109 ret = setup_relocation_extent_mapping(inode, cluster->start - offset,
3110 cluster->end - offset, cluster->start);
3111 if (ret)
3112 goto out;
3113
3114 last_index = (cluster->end - offset) >> PAGE_SHIFT;
3115 for (index = (cluster->start - offset) >> PAGE_SHIFT;
3116 index <= last_index && !ret; index++)
3117 ret = relocate_one_page(inode, ra, cluster, &cluster_nr, index);
3118 if (ret == 0)
3119 WARN_ON(cluster_nr != cluster->nr);
3120out:
3121 kfree(ra);
3122 return ret;
3123}
3124
3125static noinline_for_stack
3126int relocate_data_extent(struct inode *inode, struct btrfs_key *extent_key,
3127 struct file_extent_cluster *cluster)
3128{
3129 int ret;
3130
3131 if (cluster->nr > 0 && extent_key->objectid != cluster->end + 1) {
3132 ret = relocate_file_extent_cluster(inode, cluster);
3133 if (ret)
3134 return ret;
3135 cluster->nr = 0;
3136 }
3137
3138 if (!cluster->nr)
3139 cluster->start = extent_key->objectid;
3140 else
3141 BUG_ON(cluster->nr >= MAX_EXTENTS);
3142 cluster->end = extent_key->objectid + extent_key->offset - 1;
3143 cluster->boundary[cluster->nr] = extent_key->objectid;
3144 cluster->nr++;
3145
3146 if (cluster->nr >= MAX_EXTENTS) {
3147 ret = relocate_file_extent_cluster(inode, cluster);
3148 if (ret)
3149 return ret;
3150 cluster->nr = 0;
3151 }
3152 return 0;
3153}
3154
3155/*
3156 * helper to add a tree block to the list.
3157 * the major work is getting the generation and level of the block
3158 */
3159static int add_tree_block(struct reloc_control *rc,
3160 struct btrfs_key *extent_key,
3161 struct btrfs_path *path,
3162 struct rb_root *blocks)
3163{
3164 struct extent_buffer *eb;
3165 struct btrfs_extent_item *ei;
3166 struct btrfs_tree_block_info *bi;
3167 struct tree_block *block;
3168 struct rb_node *rb_node;
3169 u32 item_size;
3170 int level = -1;
3171 u64 generation;
3172 u64 owner = 0;
3173
3174 eb = path->nodes[0];
3175 item_size = btrfs_item_size(eb, path->slots[0]);
3176
3177 if (extent_key->type == BTRFS_METADATA_ITEM_KEY ||
3178 item_size >= sizeof(*ei) + sizeof(*bi)) {
3179 unsigned long ptr = 0, end;
3180
3181 ei = btrfs_item_ptr(eb, path->slots[0],
3182 struct btrfs_extent_item);
3183 end = (unsigned long)ei + item_size;
3184 if (extent_key->type == BTRFS_EXTENT_ITEM_KEY) {
3185 bi = (struct btrfs_tree_block_info *)(ei + 1);
3186 level = btrfs_tree_block_level(eb, bi);
3187 ptr = (unsigned long)(bi + 1);
3188 } else {
3189 level = (int)extent_key->offset;
3190 ptr = (unsigned long)(ei + 1);
3191 }
3192 generation = btrfs_extent_generation(eb, ei);
3193
3194 /*
3195 * We're reading random blocks without knowing their owner ahead
3196 * of time. This is ok most of the time, as all reloc roots and
3197 * fs roots have the same lock type. However normal trees do
3198 * not, and the only way to know ahead of time is to read the
3199 * inline ref offset. We know it's an fs root if
3200 *
3201 * 1. There's more than one ref.
3202 * 2. There's a SHARED_DATA_REF_KEY set.
3203 * 3. FULL_BACKREF is set on the flags.
3204 *
3205 * Otherwise it's safe to assume that the ref offset == the
3206 * owner of this block, so we can use that when calling
3207 * read_tree_block.
3208 */
3209 if (btrfs_extent_refs(eb, ei) == 1 &&
3210 !(btrfs_extent_flags(eb, ei) &
3211 BTRFS_BLOCK_FLAG_FULL_BACKREF) &&
3212 ptr < end) {
3213 struct btrfs_extent_inline_ref *iref;
3214 int type;
3215
3216 iref = (struct btrfs_extent_inline_ref *)ptr;
3217 type = btrfs_get_extent_inline_ref_type(eb, iref,
3218 BTRFS_REF_TYPE_BLOCK);
3219 if (type == BTRFS_REF_TYPE_INVALID)
3220 return -EINVAL;
3221 if (type == BTRFS_TREE_BLOCK_REF_KEY)
3222 owner = btrfs_extent_inline_ref_offset(eb, iref);
3223 }
3224 } else if (unlikely(item_size == sizeof(struct btrfs_extent_item_v0))) {
3225 btrfs_print_v0_err(eb->fs_info);
3226 btrfs_handle_fs_error(eb->fs_info, -EINVAL, NULL);
3227 return -EINVAL;
3228 } else {
3229 BUG();
3230 }
3231
3232 btrfs_release_path(path);
3233
3234 BUG_ON(level == -1);
3235
3236 block = kmalloc(sizeof(*block), GFP_NOFS);
3237 if (!block)
3238 return -ENOMEM;
3239
3240 block->bytenr = extent_key->objectid;
3241 block->key.objectid = rc->extent_root->fs_info->nodesize;
3242 block->key.offset = generation;
3243 block->level = level;
3244 block->key_ready = 0;
3245 block->owner = owner;
3246
3247 rb_node = rb_simple_insert(blocks, block->bytenr, &block->rb_node);
3248 if (rb_node)
3249 btrfs_backref_panic(rc->extent_root->fs_info, block->bytenr,
3250 -EEXIST);
3251
3252 return 0;
3253}
3254
3255/*
3256 * helper to add tree blocks for backref of type BTRFS_SHARED_DATA_REF_KEY
3257 */
3258static int __add_tree_block(struct reloc_control *rc,
3259 u64 bytenr, u32 blocksize,
3260 struct rb_root *blocks)
3261{
3262 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3263 struct btrfs_path *path;
3264 struct btrfs_key key;
3265 int ret;
3266 bool skinny = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
3267
3268 if (tree_block_processed(bytenr, rc))
3269 return 0;
3270
3271 if (rb_simple_search(blocks, bytenr))
3272 return 0;
3273
3274 path = btrfs_alloc_path();
3275 if (!path)
3276 return -ENOMEM;
3277again:
3278 key.objectid = bytenr;
3279 if (skinny) {
3280 key.type = BTRFS_METADATA_ITEM_KEY;
3281 key.offset = (u64)-1;
3282 } else {
3283 key.type = BTRFS_EXTENT_ITEM_KEY;
3284 key.offset = blocksize;
3285 }
3286
3287 path->search_commit_root = 1;
3288 path->skip_locking = 1;
3289 ret = btrfs_search_slot(NULL, rc->extent_root, &key, path, 0, 0);
3290 if (ret < 0)
3291 goto out;
3292
3293 if (ret > 0 && skinny) {
3294 if (path->slots[0]) {
3295 path->slots[0]--;
3296 btrfs_item_key_to_cpu(path->nodes[0], &key,
3297 path->slots[0]);
3298 if (key.objectid == bytenr &&
3299 (key.type == BTRFS_METADATA_ITEM_KEY ||
3300 (key.type == BTRFS_EXTENT_ITEM_KEY &&
3301 key.offset == blocksize)))
3302 ret = 0;
3303 }
3304
3305 if (ret) {
3306 skinny = false;
3307 btrfs_release_path(path);
3308 goto again;
3309 }
3310 }
3311 if (ret) {
3312 ASSERT(ret == 1);
3313 btrfs_print_leaf(path->nodes[0]);
3314 btrfs_err(fs_info,
3315 "tree block extent item (%llu) is not found in extent tree",
3316 bytenr);
3317 WARN_ON(1);
3318 ret = -EINVAL;
3319 goto out;
3320 }
3321
3322 ret = add_tree_block(rc, &key, path, blocks);
3323out:
3324 btrfs_free_path(path);
3325 return ret;
3326}
3327
3328static int delete_block_group_cache(struct btrfs_fs_info *fs_info,
3329 struct btrfs_block_group *block_group,
3330 struct inode *inode,
3331 u64 ino)
3332{
3333 struct btrfs_root *root = fs_info->tree_root;
3334 struct btrfs_trans_handle *trans;
3335 int ret = 0;
3336
3337 if (inode)
3338 goto truncate;
3339
3340 inode = btrfs_iget(fs_info->sb, ino, root);
3341 if (IS_ERR(inode))
3342 return -ENOENT;
3343
3344truncate:
3345 ret = btrfs_check_trunc_cache_free_space(fs_info,
3346 &fs_info->global_block_rsv);
3347 if (ret)
3348 goto out;
3349
3350 trans = btrfs_join_transaction(root);
3351 if (IS_ERR(trans)) {
3352 ret = PTR_ERR(trans);
3353 goto out;
3354 }
3355
3356 ret = btrfs_truncate_free_space_cache(trans, block_group, inode);
3357
3358 btrfs_end_transaction(trans);
3359 btrfs_btree_balance_dirty(fs_info);
3360out:
3361 iput(inode);
3362 return ret;
3363}
3364
3365/*
3366 * Locate the free space cache EXTENT_DATA in root tree leaf and delete the
3367 * cache inode, to avoid free space cache data extent blocking data relocation.
3368 */
3369static int delete_v1_space_cache(struct extent_buffer *leaf,
3370 struct btrfs_block_group *block_group,
3371 u64 data_bytenr)
3372{
3373 u64 space_cache_ino;
3374 struct btrfs_file_extent_item *ei;
3375 struct btrfs_key key;
3376 bool found = false;
3377 int i;
3378 int ret;
3379
3380 if (btrfs_header_owner(leaf) != BTRFS_ROOT_TREE_OBJECTID)
3381 return 0;
3382
3383 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
3384 u8 type;
3385
3386 btrfs_item_key_to_cpu(leaf, &key, i);
3387 if (key.type != BTRFS_EXTENT_DATA_KEY)
3388 continue;
3389 ei = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
3390 type = btrfs_file_extent_type(leaf, ei);
3391
3392 if ((type == BTRFS_FILE_EXTENT_REG ||
3393 type == BTRFS_FILE_EXTENT_PREALLOC) &&
3394 btrfs_file_extent_disk_bytenr(leaf, ei) == data_bytenr) {
3395 found = true;
3396 space_cache_ino = key.objectid;
3397 break;
3398 }
3399 }
3400 if (!found)
3401 return -ENOENT;
3402 ret = delete_block_group_cache(leaf->fs_info, block_group, NULL,
3403 space_cache_ino);
3404 return ret;
3405}
3406
3407/*
3408 * helper to find all tree blocks that reference a given data extent
3409 */
3410static noinline_for_stack
3411int add_data_references(struct reloc_control *rc,
3412 struct btrfs_key *extent_key,
3413 struct btrfs_path *path,
3414 struct rb_root *blocks)
3415{
3416 struct btrfs_backref_walk_ctx ctx = { 0 };
3417 struct ulist_iterator leaf_uiter;
3418 struct ulist_node *ref_node = NULL;
3419 const u32 blocksize = rc->extent_root->fs_info->nodesize;
3420 int ret = 0;
3421
3422 btrfs_release_path(path);
3423
3424 ctx.bytenr = extent_key->objectid;
3425 ctx.ignore_extent_item_pos = true;
3426 ctx.fs_info = rc->extent_root->fs_info;
3427
3428 ret = btrfs_find_all_leafs(&ctx);
3429 if (ret < 0)
3430 return ret;
3431
3432 ULIST_ITER_INIT(&leaf_uiter);
3433 while ((ref_node = ulist_next(ctx.refs, &leaf_uiter))) {
3434 struct btrfs_tree_parent_check check = { 0 };
3435 struct extent_buffer *eb;
3436
3437 eb = read_tree_block(ctx.fs_info, ref_node->val, &check);
3438 if (IS_ERR(eb)) {
3439 ret = PTR_ERR(eb);
3440 break;
3441 }
3442 ret = delete_v1_space_cache(eb, rc->block_group,
3443 extent_key->objectid);
3444 free_extent_buffer(eb);
3445 if (ret < 0)
3446 break;
3447 ret = __add_tree_block(rc, ref_node->val, blocksize, blocks);
3448 if (ret < 0)
3449 break;
3450 }
3451 if (ret < 0)
3452 free_block_list(blocks);
3453 ulist_free(ctx.refs);
3454 return ret;
3455}
3456
3457/*
3458 * helper to find next unprocessed extent
3459 */
3460static noinline_for_stack
3461int find_next_extent(struct reloc_control *rc, struct btrfs_path *path,
3462 struct btrfs_key *extent_key)
3463{
3464 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3465 struct btrfs_key key;
3466 struct extent_buffer *leaf;
3467 u64 start, end, last;
3468 int ret;
3469
3470 last = rc->block_group->start + rc->block_group->length;
3471 while (1) {
3472 cond_resched();
3473 if (rc->search_start >= last) {
3474 ret = 1;
3475 break;
3476 }
3477
3478 key.objectid = rc->search_start;
3479 key.type = BTRFS_EXTENT_ITEM_KEY;
3480 key.offset = 0;
3481
3482 path->search_commit_root = 1;
3483 path->skip_locking = 1;
3484 ret = btrfs_search_slot(NULL, rc->extent_root, &key, path,
3485 0, 0);
3486 if (ret < 0)
3487 break;
3488next:
3489 leaf = path->nodes[0];
3490 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
3491 ret = btrfs_next_leaf(rc->extent_root, path);
3492 if (ret != 0)
3493 break;
3494 leaf = path->nodes[0];
3495 }
3496
3497 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3498 if (key.objectid >= last) {
3499 ret = 1;
3500 break;
3501 }
3502
3503 if (key.type != BTRFS_EXTENT_ITEM_KEY &&
3504 key.type != BTRFS_METADATA_ITEM_KEY) {
3505 path->slots[0]++;
3506 goto next;
3507 }
3508
3509 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3510 key.objectid + key.offset <= rc->search_start) {
3511 path->slots[0]++;
3512 goto next;
3513 }
3514
3515 if (key.type == BTRFS_METADATA_ITEM_KEY &&
3516 key.objectid + fs_info->nodesize <=
3517 rc->search_start) {
3518 path->slots[0]++;
3519 goto next;
3520 }
3521
3522 ret = find_first_extent_bit(&rc->processed_blocks,
3523 key.objectid, &start, &end,
3524 EXTENT_DIRTY, NULL);
3525
3526 if (ret == 0 && start <= key.objectid) {
3527 btrfs_release_path(path);
3528 rc->search_start = end + 1;
3529 } else {
3530 if (key.type == BTRFS_EXTENT_ITEM_KEY)
3531 rc->search_start = key.objectid + key.offset;
3532 else
3533 rc->search_start = key.objectid +
3534 fs_info->nodesize;
3535 memcpy(extent_key, &key, sizeof(key));
3536 return 0;
3537 }
3538 }
3539 btrfs_release_path(path);
3540 return ret;
3541}
3542
3543static void set_reloc_control(struct reloc_control *rc)
3544{
3545 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3546
3547 mutex_lock(&fs_info->reloc_mutex);
3548 fs_info->reloc_ctl = rc;
3549 mutex_unlock(&fs_info->reloc_mutex);
3550}
3551
3552static void unset_reloc_control(struct reloc_control *rc)
3553{
3554 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3555
3556 mutex_lock(&fs_info->reloc_mutex);
3557 fs_info->reloc_ctl = NULL;
3558 mutex_unlock(&fs_info->reloc_mutex);
3559}
3560
3561static noinline_for_stack
3562int prepare_to_relocate(struct reloc_control *rc)
3563{
3564 struct btrfs_trans_handle *trans;
3565 int ret;
3566
3567 rc->block_rsv = btrfs_alloc_block_rsv(rc->extent_root->fs_info,
3568 BTRFS_BLOCK_RSV_TEMP);
3569 if (!rc->block_rsv)
3570 return -ENOMEM;
3571
3572 memset(&rc->cluster, 0, sizeof(rc->cluster));
3573 rc->search_start = rc->block_group->start;
3574 rc->extents_found = 0;
3575 rc->nodes_relocated = 0;
3576 rc->merging_rsv_size = 0;
3577 rc->reserved_bytes = 0;
3578 rc->block_rsv->size = rc->extent_root->fs_info->nodesize *
3579 RELOCATION_RESERVED_NODES;
3580 ret = btrfs_block_rsv_refill(rc->extent_root->fs_info,
3581 rc->block_rsv, rc->block_rsv->size,
3582 BTRFS_RESERVE_FLUSH_ALL);
3583 if (ret)
3584 return ret;
3585
3586 rc->create_reloc_tree = 1;
3587 set_reloc_control(rc);
3588
3589 trans = btrfs_join_transaction(rc->extent_root);
3590 if (IS_ERR(trans)) {
3591 unset_reloc_control(rc);
3592 /*
3593 * extent tree is not a ref_cow tree and has no reloc_root to
3594 * cleanup. And callers are responsible to free the above
3595 * block rsv.
3596 */
3597 return PTR_ERR(trans);
3598 }
3599
3600 ret = btrfs_commit_transaction(trans);
3601 if (ret)
3602 unset_reloc_control(rc);
3603
3604 return ret;
3605}
3606
3607static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
3608{
3609 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3610 struct rb_root blocks = RB_ROOT;
3611 struct btrfs_key key;
3612 struct btrfs_trans_handle *trans = NULL;
3613 struct btrfs_path *path;
3614 struct btrfs_extent_item *ei;
3615 u64 flags;
3616 int ret;
3617 int err = 0;
3618 int progress = 0;
3619
3620 path = btrfs_alloc_path();
3621 if (!path)
3622 return -ENOMEM;
3623 path->reada = READA_FORWARD;
3624
3625 ret = prepare_to_relocate(rc);
3626 if (ret) {
3627 err = ret;
3628 goto out_free;
3629 }
3630
3631 while (1) {
3632 rc->reserved_bytes = 0;
3633 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv,
3634 rc->block_rsv->size,
3635 BTRFS_RESERVE_FLUSH_ALL);
3636 if (ret) {
3637 err = ret;
3638 break;
3639 }
3640 progress++;
3641 trans = btrfs_start_transaction(rc->extent_root, 0);
3642 if (IS_ERR(trans)) {
3643 err = PTR_ERR(trans);
3644 trans = NULL;
3645 break;
3646 }
3647restart:
3648 if (update_backref_cache(trans, &rc->backref_cache)) {
3649 btrfs_end_transaction(trans);
3650 trans = NULL;
3651 continue;
3652 }
3653
3654 ret = find_next_extent(rc, path, &key);
3655 if (ret < 0)
3656 err = ret;
3657 if (ret != 0)
3658 break;
3659
3660 rc->extents_found++;
3661
3662 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3663 struct btrfs_extent_item);
3664 flags = btrfs_extent_flags(path->nodes[0], ei);
3665
3666 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
3667 ret = add_tree_block(rc, &key, path, &blocks);
3668 } else if (rc->stage == UPDATE_DATA_PTRS &&
3669 (flags & BTRFS_EXTENT_FLAG_DATA)) {
3670 ret = add_data_references(rc, &key, path, &blocks);
3671 } else {
3672 btrfs_release_path(path);
3673 ret = 0;
3674 }
3675 if (ret < 0) {
3676 err = ret;
3677 break;
3678 }
3679
3680 if (!RB_EMPTY_ROOT(&blocks)) {
3681 ret = relocate_tree_blocks(trans, rc, &blocks);
3682 if (ret < 0) {
3683 if (ret != -EAGAIN) {
3684 err = ret;
3685 break;
3686 }
3687 rc->extents_found--;
3688 rc->search_start = key.objectid;
3689 }
3690 }
3691
3692 btrfs_end_transaction_throttle(trans);
3693 btrfs_btree_balance_dirty(fs_info);
3694 trans = NULL;
3695
3696 if (rc->stage == MOVE_DATA_EXTENTS &&
3697 (flags & BTRFS_EXTENT_FLAG_DATA)) {
3698 rc->found_file_extent = 1;
3699 ret = relocate_data_extent(rc->data_inode,
3700 &key, &rc->cluster);
3701 if (ret < 0) {
3702 err = ret;
3703 break;
3704 }
3705 }
3706 if (btrfs_should_cancel_balance(fs_info)) {
3707 err = -ECANCELED;
3708 break;
3709 }
3710 }
3711 if (trans && progress && err == -ENOSPC) {
3712 ret = btrfs_force_chunk_alloc(trans, rc->block_group->flags);
3713 if (ret == 1) {
3714 err = 0;
3715 progress = 0;
3716 goto restart;
3717 }
3718 }
3719
3720 btrfs_release_path(path);
3721 clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY);
3722
3723 if (trans) {
3724 btrfs_end_transaction_throttle(trans);
3725 btrfs_btree_balance_dirty(fs_info);
3726 }
3727
3728 if (!err) {
3729 ret = relocate_file_extent_cluster(rc->data_inode,
3730 &rc->cluster);
3731 if (ret < 0)
3732 err = ret;
3733 }
3734
3735 rc->create_reloc_tree = 0;
3736 set_reloc_control(rc);
3737
3738 btrfs_backref_release_cache(&rc->backref_cache);
3739 btrfs_block_rsv_release(fs_info, rc->block_rsv, (u64)-1, NULL);
3740
3741 /*
3742 * Even in the case when the relocation is cancelled, we should all go
3743 * through prepare_to_merge() and merge_reloc_roots().
3744 *
3745 * For error (including cancelled balance), prepare_to_merge() will
3746 * mark all reloc trees orphan, then queue them for cleanup in
3747 * merge_reloc_roots()
3748 */
3749 err = prepare_to_merge(rc, err);
3750
3751 merge_reloc_roots(rc);
3752
3753 rc->merge_reloc_tree = 0;
3754 unset_reloc_control(rc);
3755 btrfs_block_rsv_release(fs_info, rc->block_rsv, (u64)-1, NULL);
3756
3757 /* get rid of pinned extents */
3758 trans = btrfs_join_transaction(rc->extent_root);
3759 if (IS_ERR(trans)) {
3760 err = PTR_ERR(trans);
3761 goto out_free;
3762 }
3763 ret = btrfs_commit_transaction(trans);
3764 if (ret && !err)
3765 err = ret;
3766out_free:
3767 ret = clean_dirty_subvols(rc);
3768 if (ret < 0 && !err)
3769 err = ret;
3770 btrfs_free_block_rsv(fs_info, rc->block_rsv);
3771 btrfs_free_path(path);
3772 return err;
3773}
3774
3775static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
3776 struct btrfs_root *root, u64 objectid)
3777{
3778 struct btrfs_path *path;
3779 struct btrfs_inode_item *item;
3780 struct extent_buffer *leaf;
3781 int ret;
3782
3783 path = btrfs_alloc_path();
3784 if (!path)
3785 return -ENOMEM;
3786
3787 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
3788 if (ret)
3789 goto out;
3790
3791 leaf = path->nodes[0];
3792 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
3793 memzero_extent_buffer(leaf, (unsigned long)item, sizeof(*item));
3794 btrfs_set_inode_generation(leaf, item, 1);
3795 btrfs_set_inode_size(leaf, item, 0);
3796 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
3797 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS |
3798 BTRFS_INODE_PREALLOC);
3799 btrfs_mark_buffer_dirty(leaf);
3800out:
3801 btrfs_free_path(path);
3802 return ret;
3803}
3804
3805static void delete_orphan_inode(struct btrfs_trans_handle *trans,
3806 struct btrfs_root *root, u64 objectid)
3807{
3808 struct btrfs_path *path;
3809 struct btrfs_key key;
3810 int ret = 0;
3811
3812 path = btrfs_alloc_path();
3813 if (!path) {
3814 ret = -ENOMEM;
3815 goto out;
3816 }
3817
3818 key.objectid = objectid;
3819 key.type = BTRFS_INODE_ITEM_KEY;
3820 key.offset = 0;
3821 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3822 if (ret) {
3823 if (ret > 0)
3824 ret = -ENOENT;
3825 goto out;
3826 }
3827 ret = btrfs_del_item(trans, root, path);
3828out:
3829 if (ret)
3830 btrfs_abort_transaction(trans, ret);
3831 btrfs_free_path(path);
3832}
3833
3834/*
3835 * helper to create inode for data relocation.
3836 * the inode is in data relocation tree and its link count is 0
3837 */
3838static noinline_for_stack
3839struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
3840 struct btrfs_block_group *group)
3841{
3842 struct inode *inode = NULL;
3843 struct btrfs_trans_handle *trans;
3844 struct btrfs_root *root;
3845 u64 objectid;
3846 int err = 0;
3847
3848 root = btrfs_grab_root(fs_info->data_reloc_root);
3849 trans = btrfs_start_transaction(root, 6);
3850 if (IS_ERR(trans)) {
3851 btrfs_put_root(root);
3852 return ERR_CAST(trans);
3853 }
3854
3855 err = btrfs_get_free_objectid(root, &objectid);
3856 if (err)
3857 goto out;
3858
3859 err = __insert_orphan_inode(trans, root, objectid);
3860 if (err)
3861 goto out;
3862
3863 inode = btrfs_iget(fs_info->sb, objectid, root);
3864 if (IS_ERR(inode)) {
3865 delete_orphan_inode(trans, root, objectid);
3866 err = PTR_ERR(inode);
3867 inode = NULL;
3868 goto out;
3869 }
3870 BTRFS_I(inode)->index_cnt = group->start;
3871
3872 err = btrfs_orphan_add(trans, BTRFS_I(inode));
3873out:
3874 btrfs_put_root(root);
3875 btrfs_end_transaction(trans);
3876 btrfs_btree_balance_dirty(fs_info);
3877 if (err) {
3878 iput(inode);
3879 inode = ERR_PTR(err);
3880 }
3881 return inode;
3882}
3883
3884/*
3885 * Mark start of chunk relocation that is cancellable. Check if the cancellation
3886 * has been requested meanwhile and don't start in that case.
3887 *
3888 * Return:
3889 * 0 success
3890 * -EINPROGRESS operation is already in progress, that's probably a bug
3891 * -ECANCELED cancellation request was set before the operation started
3892 */
3893static int reloc_chunk_start(struct btrfs_fs_info *fs_info)
3894{
3895 if (test_and_set_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) {
3896 /* This should not happen */
3897 btrfs_err(fs_info, "reloc already running, cannot start");
3898 return -EINPROGRESS;
3899 }
3900
3901 if (atomic_read(&fs_info->reloc_cancel_req) > 0) {
3902 btrfs_info(fs_info, "chunk relocation canceled on start");
3903 /*
3904 * On cancel, clear all requests but let the caller mark
3905 * the end after cleanup operations.
3906 */
3907 atomic_set(&fs_info->reloc_cancel_req, 0);
3908 return -ECANCELED;
3909 }
3910 return 0;
3911}
3912
3913/*
3914 * Mark end of chunk relocation that is cancellable and wake any waiters.
3915 */
3916static void reloc_chunk_end(struct btrfs_fs_info *fs_info)
3917{
3918 /* Requested after start, clear bit first so any waiters can continue */
3919 if (atomic_read(&fs_info->reloc_cancel_req) > 0)
3920 btrfs_info(fs_info, "chunk relocation canceled during operation");
3921 clear_and_wake_up_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags);
3922 atomic_set(&fs_info->reloc_cancel_req, 0);
3923}
3924
3925static struct reloc_control *alloc_reloc_control(struct btrfs_fs_info *fs_info)
3926{
3927 struct reloc_control *rc;
3928
3929 rc = kzalloc(sizeof(*rc), GFP_NOFS);
3930 if (!rc)
3931 return NULL;
3932
3933 INIT_LIST_HEAD(&rc->reloc_roots);
3934 INIT_LIST_HEAD(&rc->dirty_subvol_roots);
3935 btrfs_backref_init_cache(fs_info, &rc->backref_cache, 1);
3936 mapping_tree_init(&rc->reloc_root_tree);
3937 extent_io_tree_init(fs_info, &rc->processed_blocks, IO_TREE_RELOC_BLOCKS);
3938 return rc;
3939}
3940
3941static void free_reloc_control(struct reloc_control *rc)
3942{
3943 struct mapping_node *node, *tmp;
3944
3945 free_reloc_roots(&rc->reloc_roots);
3946 rbtree_postorder_for_each_entry_safe(node, tmp,
3947 &rc->reloc_root_tree.rb_root, rb_node)
3948 kfree(node);
3949
3950 kfree(rc);
3951}
3952
3953/*
3954 * Print the block group being relocated
3955 */
3956static void describe_relocation(struct btrfs_fs_info *fs_info,
3957 struct btrfs_block_group *block_group)
3958{
3959 char buf[128] = {'\0'};
3960
3961 btrfs_describe_block_groups(block_group->flags, buf, sizeof(buf));
3962
3963 btrfs_info(fs_info,
3964 "relocating block group %llu flags %s",
3965 block_group->start, buf);
3966}
3967
3968static const char *stage_to_string(int stage)
3969{
3970 if (stage == MOVE_DATA_EXTENTS)
3971 return "move data extents";
3972 if (stage == UPDATE_DATA_PTRS)
3973 return "update data pointers";
3974 return "unknown";
3975}
3976
3977/*
3978 * function to relocate all extents in a block group.
3979 */
3980int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
3981{
3982 struct btrfs_block_group *bg;
3983 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, group_start);
3984 struct reloc_control *rc;
3985 struct inode *inode;
3986 struct btrfs_path *path;
3987 int ret;
3988 int rw = 0;
3989 int err = 0;
3990
3991 /*
3992 * This only gets set if we had a half-deleted snapshot on mount. We
3993 * cannot allow relocation to start while we're still trying to clean up
3994 * these pending deletions.
3995 */
3996 ret = wait_on_bit(&fs_info->flags, BTRFS_FS_UNFINISHED_DROPS, TASK_INTERRUPTIBLE);
3997 if (ret)
3998 return ret;
3999
4000 /* We may have been woken up by close_ctree, so bail if we're closing. */
4001 if (btrfs_fs_closing(fs_info))
4002 return -EINTR;
4003
4004 bg = btrfs_lookup_block_group(fs_info, group_start);
4005 if (!bg)
4006 return -ENOENT;
4007
4008 /*
4009 * Relocation of a data block group creates ordered extents. Without
4010 * sb_start_write(), we can freeze the filesystem while unfinished
4011 * ordered extents are left. Such ordered extents can cause a deadlock
4012 * e.g. when syncfs() is waiting for their completion but they can't
4013 * finish because they block when joining a transaction, due to the
4014 * fact that the freeze locks are being held in write mode.
4015 */
4016 if (bg->flags & BTRFS_BLOCK_GROUP_DATA)
4017 ASSERT(sb_write_started(fs_info->sb));
4018
4019 if (btrfs_pinned_by_swapfile(fs_info, bg)) {
4020 btrfs_put_block_group(bg);
4021 return -ETXTBSY;
4022 }
4023
4024 rc = alloc_reloc_control(fs_info);
4025 if (!rc) {
4026 btrfs_put_block_group(bg);
4027 return -ENOMEM;
4028 }
4029
4030 ret = reloc_chunk_start(fs_info);
4031 if (ret < 0) {
4032 err = ret;
4033 goto out_put_bg;
4034 }
4035
4036 rc->extent_root = extent_root;
4037 rc->block_group = bg;
4038
4039 ret = btrfs_inc_block_group_ro(rc->block_group, true);
4040 if (ret) {
4041 err = ret;
4042 goto out;
4043 }
4044 rw = 1;
4045
4046 path = btrfs_alloc_path();
4047 if (!path) {
4048 err = -ENOMEM;
4049 goto out;
4050 }
4051
4052 inode = lookup_free_space_inode(rc->block_group, path);
4053 btrfs_free_path(path);
4054
4055 if (!IS_ERR(inode))
4056 ret = delete_block_group_cache(fs_info, rc->block_group, inode, 0);
4057 else
4058 ret = PTR_ERR(inode);
4059
4060 if (ret && ret != -ENOENT) {
4061 err = ret;
4062 goto out;
4063 }
4064
4065 rc->data_inode = create_reloc_inode(fs_info, rc->block_group);
4066 if (IS_ERR(rc->data_inode)) {
4067 err = PTR_ERR(rc->data_inode);
4068 rc->data_inode = NULL;
4069 goto out;
4070 }
4071
4072 describe_relocation(fs_info, rc->block_group);
4073
4074 btrfs_wait_block_group_reservations(rc->block_group);
4075 btrfs_wait_nocow_writers(rc->block_group);
4076 btrfs_wait_ordered_roots(fs_info, U64_MAX,
4077 rc->block_group->start,
4078 rc->block_group->length);
4079
4080 ret = btrfs_zone_finish(rc->block_group);
4081 WARN_ON(ret && ret != -EAGAIN);
4082
4083 while (1) {
4084 int finishes_stage;
4085
4086 mutex_lock(&fs_info->cleaner_mutex);
4087 ret = relocate_block_group(rc);
4088 mutex_unlock(&fs_info->cleaner_mutex);
4089 if (ret < 0)
4090 err = ret;
4091
4092 finishes_stage = rc->stage;
4093 /*
4094 * We may have gotten ENOSPC after we already dirtied some
4095 * extents. If writeout happens while we're relocating a
4096 * different block group we could end up hitting the
4097 * BUG_ON(rc->stage == UPDATE_DATA_PTRS) in
4098 * btrfs_reloc_cow_block. Make sure we write everything out
4099 * properly so we don't trip over this problem, and then break
4100 * out of the loop if we hit an error.
4101 */
4102 if (rc->stage == MOVE_DATA_EXTENTS && rc->found_file_extent) {
4103 ret = btrfs_wait_ordered_range(rc->data_inode, 0,
4104 (u64)-1);
4105 if (ret)
4106 err = ret;
4107 invalidate_mapping_pages(rc->data_inode->i_mapping,
4108 0, -1);
4109 rc->stage = UPDATE_DATA_PTRS;
4110 }
4111
4112 if (err < 0)
4113 goto out;
4114
4115 if (rc->extents_found == 0)
4116 break;
4117
4118 btrfs_info(fs_info, "found %llu extents, stage: %s",
4119 rc->extents_found, stage_to_string(finishes_stage));
4120 }
4121
4122 WARN_ON(rc->block_group->pinned > 0);
4123 WARN_ON(rc->block_group->reserved > 0);
4124 WARN_ON(rc->block_group->used > 0);
4125out:
4126 if (err && rw)
4127 btrfs_dec_block_group_ro(rc->block_group);
4128 iput(rc->data_inode);
4129out_put_bg:
4130 btrfs_put_block_group(bg);
4131 reloc_chunk_end(fs_info);
4132 free_reloc_control(rc);
4133 return err;
4134}
4135
4136static noinline_for_stack int mark_garbage_root(struct btrfs_root *root)
4137{
4138 struct btrfs_fs_info *fs_info = root->fs_info;
4139 struct btrfs_trans_handle *trans;
4140 int ret, err;
4141
4142 trans = btrfs_start_transaction(fs_info->tree_root, 0);
4143 if (IS_ERR(trans))
4144 return PTR_ERR(trans);
4145
4146 memset(&root->root_item.drop_progress, 0,
4147 sizeof(root->root_item.drop_progress));
4148 btrfs_set_root_drop_level(&root->root_item, 0);
4149 btrfs_set_root_refs(&root->root_item, 0);
4150 ret = btrfs_update_root(trans, fs_info->tree_root,
4151 &root->root_key, &root->root_item);
4152
4153 err = btrfs_end_transaction(trans);
4154 if (err)
4155 return err;
4156 return ret;
4157}
4158
4159/*
4160 * recover relocation interrupted by system crash.
4161 *
4162 * this function resumes merging reloc trees with corresponding fs trees.
4163 * this is important for keeping the sharing of tree blocks
4164 */
4165int btrfs_recover_relocation(struct btrfs_fs_info *fs_info)
4166{
4167 LIST_HEAD(reloc_roots);
4168 struct btrfs_key key;
4169 struct btrfs_root *fs_root;
4170 struct btrfs_root *reloc_root;
4171 struct btrfs_path *path;
4172 struct extent_buffer *leaf;
4173 struct reloc_control *rc = NULL;
4174 struct btrfs_trans_handle *trans;
4175 int ret;
4176 int err = 0;
4177
4178 path = btrfs_alloc_path();
4179 if (!path)
4180 return -ENOMEM;
4181 path->reada = READA_BACK;
4182
4183 key.objectid = BTRFS_TREE_RELOC_OBJECTID;
4184 key.type = BTRFS_ROOT_ITEM_KEY;
4185 key.offset = (u64)-1;
4186
4187 while (1) {
4188 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key,
4189 path, 0, 0);
4190 if (ret < 0) {
4191 err = ret;
4192 goto out;
4193 }
4194 if (ret > 0) {
4195 if (path->slots[0] == 0)
4196 break;
4197 path->slots[0]--;
4198 }
4199 leaf = path->nodes[0];
4200 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4201 btrfs_release_path(path);
4202
4203 if (key.objectid != BTRFS_TREE_RELOC_OBJECTID ||
4204 key.type != BTRFS_ROOT_ITEM_KEY)
4205 break;
4206
4207 reloc_root = btrfs_read_tree_root(fs_info->tree_root, &key);
4208 if (IS_ERR(reloc_root)) {
4209 err = PTR_ERR(reloc_root);
4210 goto out;
4211 }
4212
4213 set_bit(BTRFS_ROOT_SHAREABLE, &reloc_root->state);
4214 list_add(&reloc_root->root_list, &reloc_roots);
4215
4216 if (btrfs_root_refs(&reloc_root->root_item) > 0) {
4217 fs_root = btrfs_get_fs_root(fs_info,
4218 reloc_root->root_key.offset, false);
4219 if (IS_ERR(fs_root)) {
4220 ret = PTR_ERR(fs_root);
4221 if (ret != -ENOENT) {
4222 err = ret;
4223 goto out;
4224 }
4225 ret = mark_garbage_root(reloc_root);
4226 if (ret < 0) {
4227 err = ret;
4228 goto out;
4229 }
4230 } else {
4231 btrfs_put_root(fs_root);
4232 }
4233 }
4234
4235 if (key.offset == 0)
4236 break;
4237
4238 key.offset--;
4239 }
4240 btrfs_release_path(path);
4241
4242 if (list_empty(&reloc_roots))
4243 goto out;
4244
4245 rc = alloc_reloc_control(fs_info);
4246 if (!rc) {
4247 err = -ENOMEM;
4248 goto out;
4249 }
4250
4251 ret = reloc_chunk_start(fs_info);
4252 if (ret < 0) {
4253 err = ret;
4254 goto out_end;
4255 }
4256
4257 rc->extent_root = btrfs_extent_root(fs_info, 0);
4258
4259 set_reloc_control(rc);
4260
4261 trans = btrfs_join_transaction(rc->extent_root);
4262 if (IS_ERR(trans)) {
4263 err = PTR_ERR(trans);
4264 goto out_unset;
4265 }
4266
4267 rc->merge_reloc_tree = 1;
4268
4269 while (!list_empty(&reloc_roots)) {
4270 reloc_root = list_entry(reloc_roots.next,
4271 struct btrfs_root, root_list);
4272 list_del(&reloc_root->root_list);
4273
4274 if (btrfs_root_refs(&reloc_root->root_item) == 0) {
4275 list_add_tail(&reloc_root->root_list,
4276 &rc->reloc_roots);
4277 continue;
4278 }
4279
4280 fs_root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
4281 false);
4282 if (IS_ERR(fs_root)) {
4283 err = PTR_ERR(fs_root);
4284 list_add_tail(&reloc_root->root_list, &reloc_roots);
4285 btrfs_end_transaction(trans);
4286 goto out_unset;
4287 }
4288
4289 err = __add_reloc_root(reloc_root);
4290 ASSERT(err != -EEXIST);
4291 if (err) {
4292 list_add_tail(&reloc_root->root_list, &reloc_roots);
4293 btrfs_put_root(fs_root);
4294 btrfs_end_transaction(trans);
4295 goto out_unset;
4296 }
4297 fs_root->reloc_root = btrfs_grab_root(reloc_root);
4298 btrfs_put_root(fs_root);
4299 }
4300
4301 err = btrfs_commit_transaction(trans);
4302 if (err)
4303 goto out_unset;
4304
4305 merge_reloc_roots(rc);
4306
4307 unset_reloc_control(rc);
4308
4309 trans = btrfs_join_transaction(rc->extent_root);
4310 if (IS_ERR(trans)) {
4311 err = PTR_ERR(trans);
4312 goto out_clean;
4313 }
4314 err = btrfs_commit_transaction(trans);
4315out_clean:
4316 ret = clean_dirty_subvols(rc);
4317 if (ret < 0 && !err)
4318 err = ret;
4319out_unset:
4320 unset_reloc_control(rc);
4321out_end:
4322 reloc_chunk_end(fs_info);
4323 free_reloc_control(rc);
4324out:
4325 free_reloc_roots(&reloc_roots);
4326
4327 btrfs_free_path(path);
4328
4329 if (err == 0) {
4330 /* cleanup orphan inode in data relocation tree */
4331 fs_root = btrfs_grab_root(fs_info->data_reloc_root);
4332 ASSERT(fs_root);
4333 err = btrfs_orphan_cleanup(fs_root);
4334 btrfs_put_root(fs_root);
4335 }
4336 return err;
4337}
4338
4339/*
4340 * helper to add ordered checksum for data relocation.
4341 *
4342 * cloning checksum properly handles the nodatasum extents.
4343 * it also saves CPU time to re-calculate the checksum.
4344 */
4345int btrfs_reloc_clone_csums(struct btrfs_inode *inode, u64 file_pos, u64 len)
4346{
4347 struct btrfs_fs_info *fs_info = inode->root->fs_info;
4348 struct btrfs_root *csum_root;
4349 struct btrfs_ordered_sum *sums;
4350 struct btrfs_ordered_extent *ordered;
4351 int ret;
4352 u64 disk_bytenr;
4353 u64 new_bytenr;
4354 LIST_HEAD(list);
4355
4356 ordered = btrfs_lookup_ordered_extent(inode, file_pos);
4357 BUG_ON(ordered->file_offset != file_pos || ordered->num_bytes != len);
4358
4359 disk_bytenr = file_pos + inode->index_cnt;
4360 csum_root = btrfs_csum_root(fs_info, disk_bytenr);
4361 ret = btrfs_lookup_csums_list(csum_root, disk_bytenr,
4362 disk_bytenr + len - 1, &list, 0, false);
4363 if (ret)
4364 goto out;
4365
4366 while (!list_empty(&list)) {
4367 sums = list_entry(list.next, struct btrfs_ordered_sum, list);
4368 list_del_init(&sums->list);
4369
4370 /*
4371 * We need to offset the new_bytenr based on where the csum is.
4372 * We need to do this because we will read in entire prealloc
4373 * extents but we may have written to say the middle of the
4374 * prealloc extent, so we need to make sure the csum goes with
4375 * the right disk offset.
4376 *
4377 * We can do this because the data reloc inode refers strictly
4378 * to the on disk bytes, so we don't have to worry about
4379 * disk_len vs real len like with real inodes since it's all
4380 * disk length.
4381 */
4382 new_bytenr = ordered->disk_bytenr + sums->bytenr - disk_bytenr;
4383 sums->bytenr = new_bytenr;
4384
4385 btrfs_add_ordered_sum(ordered, sums);
4386 }
4387out:
4388 btrfs_put_ordered_extent(ordered);
4389 return ret;
4390}
4391
4392int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
4393 struct btrfs_root *root, struct extent_buffer *buf,
4394 struct extent_buffer *cow)
4395{
4396 struct btrfs_fs_info *fs_info = root->fs_info;
4397 struct reloc_control *rc;
4398 struct btrfs_backref_node *node;
4399 int first_cow = 0;
4400 int level;
4401 int ret = 0;
4402
4403 rc = fs_info->reloc_ctl;
4404 if (!rc)
4405 return 0;
4406
4407 BUG_ON(rc->stage == UPDATE_DATA_PTRS && btrfs_is_data_reloc_root(root));
4408
4409 level = btrfs_header_level(buf);
4410 if (btrfs_header_generation(buf) <=
4411 btrfs_root_last_snapshot(&root->root_item))
4412 first_cow = 1;
4413
4414 if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID &&
4415 rc->create_reloc_tree) {
4416 WARN_ON(!first_cow && level == 0);
4417
4418 node = rc->backref_cache.path[level];
4419 BUG_ON(node->bytenr != buf->start &&
4420 node->new_bytenr != buf->start);
4421
4422 btrfs_backref_drop_node_buffer(node);
4423 atomic_inc(&cow->refs);
4424 node->eb = cow;
4425 node->new_bytenr = cow->start;
4426
4427 if (!node->pending) {
4428 list_move_tail(&node->list,
4429 &rc->backref_cache.pending[level]);
4430 node->pending = 1;
4431 }
4432
4433 if (first_cow)
4434 mark_block_processed(rc, node);
4435
4436 if (first_cow && level > 0)
4437 rc->nodes_relocated += buf->len;
4438 }
4439
4440 if (level == 0 && first_cow && rc->stage == UPDATE_DATA_PTRS)
4441 ret = replace_file_extents(trans, rc, root, cow);
4442 return ret;
4443}
4444
4445/*
4446 * called before creating snapshot. it calculates metadata reservation
4447 * required for relocating tree blocks in the snapshot
4448 */
4449void btrfs_reloc_pre_snapshot(struct btrfs_pending_snapshot *pending,
4450 u64 *bytes_to_reserve)
4451{
4452 struct btrfs_root *root = pending->root;
4453 struct reloc_control *rc = root->fs_info->reloc_ctl;
4454
4455 if (!rc || !have_reloc_root(root))
4456 return;
4457
4458 if (!rc->merge_reloc_tree)
4459 return;
4460
4461 root = root->reloc_root;
4462 BUG_ON(btrfs_root_refs(&root->root_item) == 0);
4463 /*
4464 * relocation is in the stage of merging trees. the space
4465 * used by merging a reloc tree is twice the size of
4466 * relocated tree nodes in the worst case. half for cowing
4467 * the reloc tree, half for cowing the fs tree. the space
4468 * used by cowing the reloc tree will be freed after the
4469 * tree is dropped. if we create snapshot, cowing the fs
4470 * tree may use more space than it frees. so we need
4471 * reserve extra space.
4472 */
4473 *bytes_to_reserve += rc->nodes_relocated;
4474}
4475
4476/*
4477 * called after snapshot is created. migrate block reservation
4478 * and create reloc root for the newly created snapshot
4479 *
4480 * This is similar to btrfs_init_reloc_root(), we come out of here with two
4481 * references held on the reloc_root, one for root->reloc_root and one for
4482 * rc->reloc_roots.
4483 */
4484int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
4485 struct btrfs_pending_snapshot *pending)
4486{
4487 struct btrfs_root *root = pending->root;
4488 struct btrfs_root *reloc_root;
4489 struct btrfs_root *new_root;
4490 struct reloc_control *rc = root->fs_info->reloc_ctl;
4491 int ret;
4492
4493 if (!rc || !have_reloc_root(root))
4494 return 0;
4495
4496 rc = root->fs_info->reloc_ctl;
4497 rc->merging_rsv_size += rc->nodes_relocated;
4498
4499 if (rc->merge_reloc_tree) {
4500 ret = btrfs_block_rsv_migrate(&pending->block_rsv,
4501 rc->block_rsv,
4502 rc->nodes_relocated, true);
4503 if (ret)
4504 return ret;
4505 }
4506
4507 new_root = pending->snap;
4508 reloc_root = create_reloc_root(trans, root->reloc_root,
4509 new_root->root_key.objectid);
4510 if (IS_ERR(reloc_root))
4511 return PTR_ERR(reloc_root);
4512
4513 ret = __add_reloc_root(reloc_root);
4514 ASSERT(ret != -EEXIST);
4515 if (ret) {
4516 /* Pairs with create_reloc_root */
4517 btrfs_put_root(reloc_root);
4518 return ret;
4519 }
4520 new_root->reloc_root = btrfs_grab_root(reloc_root);
4521
4522 if (rc->create_reloc_tree)
4523 ret = clone_backref_node(trans, rc, root, reloc_root);
4524 return ret;
4525}
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (C) 2009 Oracle. All rights reserved.
4 */
5
6#include <linux/sched.h>
7#include <linux/pagemap.h>
8#include <linux/writeback.h>
9#include <linux/blkdev.h>
10#include <linux/rbtree.h>
11#include <linux/slab.h>
12#include <linux/error-injection.h>
13#include "ctree.h"
14#include "disk-io.h"
15#include "transaction.h"
16#include "volumes.h"
17#include "locking.h"
18#include "btrfs_inode.h"
19#include "async-thread.h"
20#include "free-space-cache.h"
21#include "qgroup.h"
22#include "print-tree.h"
23#include "delalloc-space.h"
24#include "block-group.h"
25#include "backref.h"
26#include "misc.h"
27#include "subpage.h"
28#include "zoned.h"
29#include "inode-item.h"
30#include "space-info.h"
31#include "fs.h"
32#include "accessors.h"
33#include "extent-tree.h"
34#include "root-tree.h"
35#include "file-item.h"
36#include "relocation.h"
37#include "super.h"
38#include "tree-checker.h"
39#include "raid-stripe-tree.h"
40
41/*
42 * Relocation overview
43 *
44 * [What does relocation do]
45 *
46 * The objective of relocation is to relocate all extents of the target block
47 * group to other block groups.
48 * This is utilized by resize (shrink only), profile converting, compacting
49 * space, or balance routine to spread chunks over devices.
50 *
51 * Before | After
52 * ------------------------------------------------------------------
53 * BG A: 10 data extents | BG A: deleted
54 * BG B: 2 data extents | BG B: 10 data extents (2 old + 8 relocated)
55 * BG C: 1 extents | BG C: 3 data extents (1 old + 2 relocated)
56 *
57 * [How does relocation work]
58 *
59 * 1. Mark the target block group read-only
60 * New extents won't be allocated from the target block group.
61 *
62 * 2.1 Record each extent in the target block group
63 * To build a proper map of extents to be relocated.
64 *
65 * 2.2 Build data reloc tree and reloc trees
66 * Data reloc tree will contain an inode, recording all newly relocated
67 * data extents.
68 * There will be only one data reloc tree for one data block group.
69 *
70 * Reloc tree will be a special snapshot of its source tree, containing
71 * relocated tree blocks.
72 * Each tree referring to a tree block in target block group will get its
73 * reloc tree built.
74 *
75 * 2.3 Swap source tree with its corresponding reloc tree
76 * Each involved tree only refers to new extents after swap.
77 *
78 * 3. Cleanup reloc trees and data reloc tree.
79 * As old extents in the target block group are still referenced by reloc
80 * trees, we need to clean them up before really freeing the target block
81 * group.
82 *
83 * The main complexity is in steps 2.2 and 2.3.
84 *
85 * The entry point of relocation is relocate_block_group() function.
86 */
87
88#define RELOCATION_RESERVED_NODES 256
89/*
90 * map address of tree root to tree
91 */
92struct mapping_node {
93 struct {
94 struct rb_node rb_node;
95 u64 bytenr;
96 }; /* Use rb_simle_node for search/insert */
97 void *data;
98};
99
100struct mapping_tree {
101 struct rb_root rb_root;
102 spinlock_t lock;
103};
104
105/*
106 * present a tree block to process
107 */
108struct tree_block {
109 struct {
110 struct rb_node rb_node;
111 u64 bytenr;
112 }; /* Use rb_simple_node for search/insert */
113 u64 owner;
114 struct btrfs_key key;
115 u8 level;
116 bool key_ready;
117};
118
119#define MAX_EXTENTS 128
120
121struct file_extent_cluster {
122 u64 start;
123 u64 end;
124 u64 boundary[MAX_EXTENTS];
125 unsigned int nr;
126 u64 owning_root;
127};
128
129/* Stages of data relocation. */
130enum reloc_stage {
131 MOVE_DATA_EXTENTS,
132 UPDATE_DATA_PTRS
133};
134
135struct reloc_control {
136 /* block group to relocate */
137 struct btrfs_block_group *block_group;
138 /* extent tree */
139 struct btrfs_root *extent_root;
140 /* inode for moving data */
141 struct inode *data_inode;
142
143 struct btrfs_block_rsv *block_rsv;
144
145 struct btrfs_backref_cache backref_cache;
146
147 struct file_extent_cluster cluster;
148 /* tree blocks have been processed */
149 struct extent_io_tree processed_blocks;
150 /* map start of tree root to corresponding reloc tree */
151 struct mapping_tree reloc_root_tree;
152 /* list of reloc trees */
153 struct list_head reloc_roots;
154 /* list of subvolume trees that get relocated */
155 struct list_head dirty_subvol_roots;
156 /* size of metadata reservation for merging reloc trees */
157 u64 merging_rsv_size;
158 /* size of relocated tree nodes */
159 u64 nodes_relocated;
160 /* reserved size for block group relocation*/
161 u64 reserved_bytes;
162
163 u64 search_start;
164 u64 extents_found;
165
166 enum reloc_stage stage;
167 bool create_reloc_tree;
168 bool merge_reloc_tree;
169 bool found_file_extent;
170};
171
172static void mark_block_processed(struct reloc_control *rc,
173 struct btrfs_backref_node *node)
174{
175 u32 blocksize;
176
177 if (node->level == 0 ||
178 in_range(node->bytenr, rc->block_group->start,
179 rc->block_group->length)) {
180 blocksize = rc->extent_root->fs_info->nodesize;
181 set_extent_bit(&rc->processed_blocks, node->bytenr,
182 node->bytenr + blocksize - 1, EXTENT_DIRTY, NULL);
183 }
184 node->processed = 1;
185}
186
187/*
188 * walk up backref nodes until reach node presents tree root
189 */
190static struct btrfs_backref_node *walk_up_backref(
191 struct btrfs_backref_node *node,
192 struct btrfs_backref_edge *edges[], int *index)
193{
194 struct btrfs_backref_edge *edge;
195 int idx = *index;
196
197 while (!list_empty(&node->upper)) {
198 edge = list_entry(node->upper.next,
199 struct btrfs_backref_edge, list[LOWER]);
200 edges[idx++] = edge;
201 node = edge->node[UPPER];
202 }
203 BUG_ON(node->detached);
204 *index = idx;
205 return node;
206}
207
208/*
209 * walk down backref nodes to find start of next reference path
210 */
211static struct btrfs_backref_node *walk_down_backref(
212 struct btrfs_backref_edge *edges[], int *index)
213{
214 struct btrfs_backref_edge *edge;
215 struct btrfs_backref_node *lower;
216 int idx = *index;
217
218 while (idx > 0) {
219 edge = edges[idx - 1];
220 lower = edge->node[LOWER];
221 if (list_is_last(&edge->list[LOWER], &lower->upper)) {
222 idx--;
223 continue;
224 }
225 edge = list_entry(edge->list[LOWER].next,
226 struct btrfs_backref_edge, list[LOWER]);
227 edges[idx - 1] = edge;
228 *index = idx;
229 return edge->node[UPPER];
230 }
231 *index = 0;
232 return NULL;
233}
234
235static bool reloc_root_is_dead(const struct btrfs_root *root)
236{
237 /*
238 * Pair with set_bit/clear_bit in clean_dirty_subvols and
239 * btrfs_update_reloc_root. We need to see the updated bit before
240 * trying to access reloc_root
241 */
242 smp_rmb();
243 if (test_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state))
244 return true;
245 return false;
246}
247
248/*
249 * Check if this subvolume tree has valid reloc tree.
250 *
251 * Reloc tree after swap is considered dead, thus not considered as valid.
252 * This is enough for most callers, as they don't distinguish dead reloc root
253 * from no reloc root. But btrfs_should_ignore_reloc_root() below is a
254 * special case.
255 */
256static bool have_reloc_root(const struct btrfs_root *root)
257{
258 if (reloc_root_is_dead(root))
259 return false;
260 if (!root->reloc_root)
261 return false;
262 return true;
263}
264
265bool btrfs_should_ignore_reloc_root(const struct btrfs_root *root)
266{
267 struct btrfs_root *reloc_root;
268
269 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
270 return false;
271
272 /* This root has been merged with its reloc tree, we can ignore it */
273 if (reloc_root_is_dead(root))
274 return true;
275
276 reloc_root = root->reloc_root;
277 if (!reloc_root)
278 return false;
279
280 if (btrfs_header_generation(reloc_root->commit_root) ==
281 root->fs_info->running_transaction->transid)
282 return false;
283 /*
284 * If there is reloc tree and it was created in previous transaction
285 * backref lookup can find the reloc tree, so backref node for the fs
286 * tree root is useless for relocation.
287 */
288 return true;
289}
290
291/*
292 * find reloc tree by address of tree root
293 */
294struct btrfs_root *find_reloc_root(struct btrfs_fs_info *fs_info, u64 bytenr)
295{
296 struct reloc_control *rc = fs_info->reloc_ctl;
297 struct rb_node *rb_node;
298 struct mapping_node *node;
299 struct btrfs_root *root = NULL;
300
301 ASSERT(rc);
302 spin_lock(&rc->reloc_root_tree.lock);
303 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root, bytenr);
304 if (rb_node) {
305 node = rb_entry(rb_node, struct mapping_node, rb_node);
306 root = node->data;
307 }
308 spin_unlock(&rc->reloc_root_tree.lock);
309 return btrfs_grab_root(root);
310}
311
312/*
313 * For useless nodes, do two major clean ups:
314 *
315 * - Cleanup the children edges and nodes
316 * If child node is also orphan (no parent) during cleanup, then the child
317 * node will also be cleaned up.
318 *
319 * - Freeing up leaves (level 0), keeps nodes detached
320 * For nodes, the node is still cached as "detached"
321 *
322 * Return false if @node is not in the @useless_nodes list.
323 * Return true if @node is in the @useless_nodes list.
324 */
325static bool handle_useless_nodes(struct reloc_control *rc,
326 struct btrfs_backref_node *node)
327{
328 struct btrfs_backref_cache *cache = &rc->backref_cache;
329 struct list_head *useless_node = &cache->useless_node;
330 bool ret = false;
331
332 while (!list_empty(useless_node)) {
333 struct btrfs_backref_node *cur;
334
335 cur = list_first_entry(useless_node, struct btrfs_backref_node,
336 list);
337 list_del_init(&cur->list);
338
339 /* Only tree root nodes can be added to @useless_nodes */
340 ASSERT(list_empty(&cur->upper));
341
342 if (cur == node)
343 ret = true;
344
345 /* The node is the lowest node */
346 if (cur->lowest) {
347 list_del_init(&cur->lower);
348 cur->lowest = 0;
349 }
350
351 /* Cleanup the lower edges */
352 while (!list_empty(&cur->lower)) {
353 struct btrfs_backref_edge *edge;
354 struct btrfs_backref_node *lower;
355
356 edge = list_entry(cur->lower.next,
357 struct btrfs_backref_edge, list[UPPER]);
358 list_del(&edge->list[UPPER]);
359 list_del(&edge->list[LOWER]);
360 lower = edge->node[LOWER];
361 btrfs_backref_free_edge(cache, edge);
362
363 /* Child node is also orphan, queue for cleanup */
364 if (list_empty(&lower->upper))
365 list_add(&lower->list, useless_node);
366 }
367 /* Mark this block processed for relocation */
368 mark_block_processed(rc, cur);
369
370 /*
371 * Backref nodes for tree leaves are deleted from the cache.
372 * Backref nodes for upper level tree blocks are left in the
373 * cache to avoid unnecessary backref lookup.
374 */
375 if (cur->level > 0) {
376 list_add(&cur->list, &cache->detached);
377 cur->detached = 1;
378 } else {
379 rb_erase(&cur->rb_node, &cache->rb_root);
380 btrfs_backref_free_node(cache, cur);
381 }
382 }
383 return ret;
384}
385
386/*
387 * Build backref tree for a given tree block. Root of the backref tree
388 * corresponds the tree block, leaves of the backref tree correspond roots of
389 * b-trees that reference the tree block.
390 *
391 * The basic idea of this function is check backrefs of a given block to find
392 * upper level blocks that reference the block, and then check backrefs of
393 * these upper level blocks recursively. The recursion stops when tree root is
394 * reached or backrefs for the block is cached.
395 *
396 * NOTE: if we find that backrefs for a block are cached, we know backrefs for
397 * all upper level blocks that directly/indirectly reference the block are also
398 * cached.
399 */
400static noinline_for_stack struct btrfs_backref_node *build_backref_tree(
401 struct btrfs_trans_handle *trans,
402 struct reloc_control *rc, struct btrfs_key *node_key,
403 int level, u64 bytenr)
404{
405 struct btrfs_backref_iter *iter;
406 struct btrfs_backref_cache *cache = &rc->backref_cache;
407 /* For searching parent of TREE_BLOCK_REF */
408 struct btrfs_path *path;
409 struct btrfs_backref_node *cur;
410 struct btrfs_backref_node *node = NULL;
411 struct btrfs_backref_edge *edge;
412 int ret;
413
414 iter = btrfs_backref_iter_alloc(rc->extent_root->fs_info);
415 if (!iter)
416 return ERR_PTR(-ENOMEM);
417 path = btrfs_alloc_path();
418 if (!path) {
419 ret = -ENOMEM;
420 goto out;
421 }
422
423 node = btrfs_backref_alloc_node(cache, bytenr, level);
424 if (!node) {
425 ret = -ENOMEM;
426 goto out;
427 }
428
429 node->lowest = 1;
430 cur = node;
431
432 /* Breadth-first search to build backref cache */
433 do {
434 ret = btrfs_backref_add_tree_node(trans, cache, path, iter,
435 node_key, cur);
436 if (ret < 0)
437 goto out;
438
439 edge = list_first_entry_or_null(&cache->pending_edge,
440 struct btrfs_backref_edge, list[UPPER]);
441 /*
442 * The pending list isn't empty, take the first block to
443 * process
444 */
445 if (edge) {
446 list_del_init(&edge->list[UPPER]);
447 cur = edge->node[UPPER];
448 }
449 } while (edge);
450
451 /* Finish the upper linkage of newly added edges/nodes */
452 ret = btrfs_backref_finish_upper_links(cache, node);
453 if (ret < 0)
454 goto out;
455
456 if (handle_useless_nodes(rc, node))
457 node = NULL;
458out:
459 btrfs_free_path(iter->path);
460 kfree(iter);
461 btrfs_free_path(path);
462 if (ret) {
463 btrfs_backref_error_cleanup(cache, node);
464 return ERR_PTR(ret);
465 }
466 ASSERT(!node || !node->detached);
467 ASSERT(list_empty(&cache->useless_node) &&
468 list_empty(&cache->pending_edge));
469 return node;
470}
471
472/*
473 * helper to add backref node for the newly created snapshot.
474 * the backref node is created by cloning backref node that
475 * corresponds to root of source tree
476 */
477static int clone_backref_node(struct btrfs_trans_handle *trans,
478 struct reloc_control *rc,
479 const struct btrfs_root *src,
480 struct btrfs_root *dest)
481{
482 struct btrfs_root *reloc_root = src->reloc_root;
483 struct btrfs_backref_cache *cache = &rc->backref_cache;
484 struct btrfs_backref_node *node = NULL;
485 struct btrfs_backref_node *new_node;
486 struct btrfs_backref_edge *edge;
487 struct btrfs_backref_edge *new_edge;
488 struct rb_node *rb_node;
489
490 rb_node = rb_simple_search(&cache->rb_root, src->commit_root->start);
491 if (rb_node) {
492 node = rb_entry(rb_node, struct btrfs_backref_node, rb_node);
493 if (node->detached)
494 node = NULL;
495 else
496 BUG_ON(node->new_bytenr != reloc_root->node->start);
497 }
498
499 if (!node) {
500 rb_node = rb_simple_search(&cache->rb_root,
501 reloc_root->commit_root->start);
502 if (rb_node) {
503 node = rb_entry(rb_node, struct btrfs_backref_node,
504 rb_node);
505 BUG_ON(node->detached);
506 }
507 }
508
509 if (!node)
510 return 0;
511
512 new_node = btrfs_backref_alloc_node(cache, dest->node->start,
513 node->level);
514 if (!new_node)
515 return -ENOMEM;
516
517 new_node->lowest = node->lowest;
518 new_node->checked = 1;
519 new_node->root = btrfs_grab_root(dest);
520 ASSERT(new_node->root);
521
522 if (!node->lowest) {
523 list_for_each_entry(edge, &node->lower, list[UPPER]) {
524 new_edge = btrfs_backref_alloc_edge(cache);
525 if (!new_edge)
526 goto fail;
527
528 btrfs_backref_link_edge(new_edge, edge->node[LOWER],
529 new_node, LINK_UPPER);
530 }
531 } else {
532 list_add_tail(&new_node->lower, &cache->leaves);
533 }
534
535 rb_node = rb_simple_insert(&cache->rb_root, new_node->bytenr,
536 &new_node->rb_node);
537 if (rb_node)
538 btrfs_backref_panic(trans->fs_info, new_node->bytenr, -EEXIST);
539
540 if (!new_node->lowest) {
541 list_for_each_entry(new_edge, &new_node->lower, list[UPPER]) {
542 list_add_tail(&new_edge->list[LOWER],
543 &new_edge->node[LOWER]->upper);
544 }
545 }
546 return 0;
547fail:
548 while (!list_empty(&new_node->lower)) {
549 new_edge = list_entry(new_node->lower.next,
550 struct btrfs_backref_edge, list[UPPER]);
551 list_del(&new_edge->list[UPPER]);
552 btrfs_backref_free_edge(cache, new_edge);
553 }
554 btrfs_backref_free_node(cache, new_node);
555 return -ENOMEM;
556}
557
558/*
559 * helper to add 'address of tree root -> reloc tree' mapping
560 */
561static int __add_reloc_root(struct btrfs_root *root)
562{
563 struct btrfs_fs_info *fs_info = root->fs_info;
564 struct rb_node *rb_node;
565 struct mapping_node *node;
566 struct reloc_control *rc = fs_info->reloc_ctl;
567
568 node = kmalloc(sizeof(*node), GFP_NOFS);
569 if (!node)
570 return -ENOMEM;
571
572 node->bytenr = root->commit_root->start;
573 node->data = root;
574
575 spin_lock(&rc->reloc_root_tree.lock);
576 rb_node = rb_simple_insert(&rc->reloc_root_tree.rb_root,
577 node->bytenr, &node->rb_node);
578 spin_unlock(&rc->reloc_root_tree.lock);
579 if (rb_node) {
580 btrfs_err(fs_info,
581 "Duplicate root found for start=%llu while inserting into relocation tree",
582 node->bytenr);
583 return -EEXIST;
584 }
585
586 list_add_tail(&root->root_list, &rc->reloc_roots);
587 return 0;
588}
589
590/*
591 * helper to delete the 'address of tree root -> reloc tree'
592 * mapping
593 */
594static void __del_reloc_root(struct btrfs_root *root)
595{
596 struct btrfs_fs_info *fs_info = root->fs_info;
597 struct rb_node *rb_node;
598 struct mapping_node *node = NULL;
599 struct reloc_control *rc = fs_info->reloc_ctl;
600 bool put_ref = false;
601
602 if (rc && root->node) {
603 spin_lock(&rc->reloc_root_tree.lock);
604 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root,
605 root->commit_root->start);
606 if (rb_node) {
607 node = rb_entry(rb_node, struct mapping_node, rb_node);
608 rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
609 RB_CLEAR_NODE(&node->rb_node);
610 }
611 spin_unlock(&rc->reloc_root_tree.lock);
612 ASSERT(!node || (struct btrfs_root *)node->data == root);
613 }
614
615 /*
616 * We only put the reloc root here if it's on the list. There's a lot
617 * of places where the pattern is to splice the rc->reloc_roots, process
618 * the reloc roots, and then add the reloc root back onto
619 * rc->reloc_roots. If we call __del_reloc_root while it's off of the
620 * list we don't want the reference being dropped, because the guy
621 * messing with the list is in charge of the reference.
622 */
623 spin_lock(&fs_info->trans_lock);
624 if (!list_empty(&root->root_list)) {
625 put_ref = true;
626 list_del_init(&root->root_list);
627 }
628 spin_unlock(&fs_info->trans_lock);
629 if (put_ref)
630 btrfs_put_root(root);
631 kfree(node);
632}
633
634/*
635 * helper to update the 'address of tree root -> reloc tree'
636 * mapping
637 */
638static int __update_reloc_root(struct btrfs_root *root)
639{
640 struct btrfs_fs_info *fs_info = root->fs_info;
641 struct rb_node *rb_node;
642 struct mapping_node *node = NULL;
643 struct reloc_control *rc = fs_info->reloc_ctl;
644
645 spin_lock(&rc->reloc_root_tree.lock);
646 rb_node = rb_simple_search(&rc->reloc_root_tree.rb_root,
647 root->commit_root->start);
648 if (rb_node) {
649 node = rb_entry(rb_node, struct mapping_node, rb_node);
650 rb_erase(&node->rb_node, &rc->reloc_root_tree.rb_root);
651 }
652 spin_unlock(&rc->reloc_root_tree.lock);
653
654 if (!node)
655 return 0;
656 BUG_ON((struct btrfs_root *)node->data != root);
657
658 spin_lock(&rc->reloc_root_tree.lock);
659 node->bytenr = root->node->start;
660 rb_node = rb_simple_insert(&rc->reloc_root_tree.rb_root,
661 node->bytenr, &node->rb_node);
662 spin_unlock(&rc->reloc_root_tree.lock);
663 if (rb_node)
664 btrfs_backref_panic(fs_info, node->bytenr, -EEXIST);
665 return 0;
666}
667
668static struct btrfs_root *create_reloc_root(struct btrfs_trans_handle *trans,
669 struct btrfs_root *root, u64 objectid)
670{
671 struct btrfs_fs_info *fs_info = root->fs_info;
672 struct btrfs_root *reloc_root;
673 struct extent_buffer *eb;
674 struct btrfs_root_item *root_item;
675 struct btrfs_key root_key;
676 int ret = 0;
677 bool must_abort = false;
678
679 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
680 if (!root_item)
681 return ERR_PTR(-ENOMEM);
682
683 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
684 root_key.type = BTRFS_ROOT_ITEM_KEY;
685 root_key.offset = objectid;
686
687 if (btrfs_root_id(root) == objectid) {
688 u64 commit_root_gen;
689
690 /* called by btrfs_init_reloc_root */
691 ret = btrfs_copy_root(trans, root, root->commit_root, &eb,
692 BTRFS_TREE_RELOC_OBJECTID);
693 if (ret)
694 goto fail;
695
696 /*
697 * Set the last_snapshot field to the generation of the commit
698 * root - like this ctree.c:btrfs_block_can_be_shared() behaves
699 * correctly (returns true) when the relocation root is created
700 * either inside the critical section of a transaction commit
701 * (through transaction.c:qgroup_account_snapshot()) and when
702 * it's created before the transaction commit is started.
703 */
704 commit_root_gen = btrfs_header_generation(root->commit_root);
705 btrfs_set_root_last_snapshot(&root->root_item, commit_root_gen);
706 } else {
707 /*
708 * called by btrfs_reloc_post_snapshot_hook.
709 * the source tree is a reloc tree, all tree blocks
710 * modified after it was created have RELOC flag
711 * set in their headers. so it's OK to not update
712 * the 'last_snapshot'.
713 */
714 ret = btrfs_copy_root(trans, root, root->node, &eb,
715 BTRFS_TREE_RELOC_OBJECTID);
716 if (ret)
717 goto fail;
718 }
719
720 /*
721 * We have changed references at this point, we must abort the
722 * transaction if anything fails.
723 */
724 must_abort = true;
725
726 memcpy(root_item, &root->root_item, sizeof(*root_item));
727 btrfs_set_root_bytenr(root_item, eb->start);
728 btrfs_set_root_level(root_item, btrfs_header_level(eb));
729 btrfs_set_root_generation(root_item, trans->transid);
730
731 if (btrfs_root_id(root) == objectid) {
732 btrfs_set_root_refs(root_item, 0);
733 memset(&root_item->drop_progress, 0,
734 sizeof(struct btrfs_disk_key));
735 btrfs_set_root_drop_level(root_item, 0);
736 }
737
738 btrfs_tree_unlock(eb);
739 free_extent_buffer(eb);
740
741 ret = btrfs_insert_root(trans, fs_info->tree_root,
742 &root_key, root_item);
743 if (ret)
744 goto fail;
745
746 kfree(root_item);
747
748 reloc_root = btrfs_read_tree_root(fs_info->tree_root, &root_key);
749 if (IS_ERR(reloc_root)) {
750 ret = PTR_ERR(reloc_root);
751 goto abort;
752 }
753 set_bit(BTRFS_ROOT_SHAREABLE, &reloc_root->state);
754 btrfs_set_root_last_trans(reloc_root, trans->transid);
755 return reloc_root;
756fail:
757 kfree(root_item);
758abort:
759 if (must_abort)
760 btrfs_abort_transaction(trans, ret);
761 return ERR_PTR(ret);
762}
763
764/*
765 * create reloc tree for a given fs tree. reloc tree is just a
766 * snapshot of the fs tree with special root objectid.
767 *
768 * The reloc_root comes out of here with two references, one for
769 * root->reloc_root, and another for being on the rc->reloc_roots list.
770 */
771int btrfs_init_reloc_root(struct btrfs_trans_handle *trans,
772 struct btrfs_root *root)
773{
774 struct btrfs_fs_info *fs_info = root->fs_info;
775 struct btrfs_root *reloc_root;
776 struct reloc_control *rc = fs_info->reloc_ctl;
777 struct btrfs_block_rsv *rsv;
778 int clear_rsv = 0;
779 int ret;
780
781 if (!rc)
782 return 0;
783
784 /*
785 * The subvolume has reloc tree but the swap is finished, no need to
786 * create/update the dead reloc tree
787 */
788 if (reloc_root_is_dead(root))
789 return 0;
790
791 /*
792 * This is subtle but important. We do not do
793 * record_root_in_transaction for reloc roots, instead we record their
794 * corresponding fs root, and then here we update the last trans for the
795 * reloc root. This means that we have to do this for the entire life
796 * of the reloc root, regardless of which stage of the relocation we are
797 * in.
798 */
799 if (root->reloc_root) {
800 reloc_root = root->reloc_root;
801 btrfs_set_root_last_trans(reloc_root, trans->transid);
802 return 0;
803 }
804
805 /*
806 * We are merging reloc roots, we do not need new reloc trees. Also
807 * reloc trees never need their own reloc tree.
808 */
809 if (!rc->create_reloc_tree || btrfs_root_id(root) == BTRFS_TREE_RELOC_OBJECTID)
810 return 0;
811
812 if (!trans->reloc_reserved) {
813 rsv = trans->block_rsv;
814 trans->block_rsv = rc->block_rsv;
815 clear_rsv = 1;
816 }
817 reloc_root = create_reloc_root(trans, root, btrfs_root_id(root));
818 if (clear_rsv)
819 trans->block_rsv = rsv;
820 if (IS_ERR(reloc_root))
821 return PTR_ERR(reloc_root);
822
823 ret = __add_reloc_root(reloc_root);
824 ASSERT(ret != -EEXIST);
825 if (ret) {
826 /* Pairs with create_reloc_root */
827 btrfs_put_root(reloc_root);
828 return ret;
829 }
830 root->reloc_root = btrfs_grab_root(reloc_root);
831 return 0;
832}
833
834/*
835 * update root item of reloc tree
836 */
837int btrfs_update_reloc_root(struct btrfs_trans_handle *trans,
838 struct btrfs_root *root)
839{
840 struct btrfs_fs_info *fs_info = root->fs_info;
841 struct btrfs_root *reloc_root;
842 struct btrfs_root_item *root_item;
843 int ret;
844
845 if (!have_reloc_root(root))
846 return 0;
847
848 reloc_root = root->reloc_root;
849 root_item = &reloc_root->root_item;
850
851 /*
852 * We are probably ok here, but __del_reloc_root() will drop its ref of
853 * the root. We have the ref for root->reloc_root, but just in case
854 * hold it while we update the reloc root.
855 */
856 btrfs_grab_root(reloc_root);
857
858 /* root->reloc_root will stay until current relocation finished */
859 if (fs_info->reloc_ctl && fs_info->reloc_ctl->merge_reloc_tree &&
860 btrfs_root_refs(root_item) == 0) {
861 set_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
862 /*
863 * Mark the tree as dead before we change reloc_root so
864 * have_reloc_root will not touch it from now on.
865 */
866 smp_wmb();
867 __del_reloc_root(reloc_root);
868 }
869
870 if (reloc_root->commit_root != reloc_root->node) {
871 __update_reloc_root(reloc_root);
872 btrfs_set_root_node(root_item, reloc_root->node);
873 free_extent_buffer(reloc_root->commit_root);
874 reloc_root->commit_root = btrfs_root_node(reloc_root);
875 }
876
877 ret = btrfs_update_root(trans, fs_info->tree_root,
878 &reloc_root->root_key, root_item);
879 btrfs_put_root(reloc_root);
880 return ret;
881}
882
883/*
884 * get new location of data
885 */
886static int get_new_location(struct inode *reloc_inode, u64 *new_bytenr,
887 u64 bytenr, u64 num_bytes)
888{
889 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
890 struct btrfs_path *path;
891 struct btrfs_file_extent_item *fi;
892 struct extent_buffer *leaf;
893 int ret;
894
895 path = btrfs_alloc_path();
896 if (!path)
897 return -ENOMEM;
898
899 bytenr -= BTRFS_I(reloc_inode)->reloc_block_group_start;
900 ret = btrfs_lookup_file_extent(NULL, root, path,
901 btrfs_ino(BTRFS_I(reloc_inode)), bytenr, 0);
902 if (ret < 0)
903 goto out;
904 if (ret > 0) {
905 ret = -ENOENT;
906 goto out;
907 }
908
909 leaf = path->nodes[0];
910 fi = btrfs_item_ptr(leaf, path->slots[0],
911 struct btrfs_file_extent_item);
912
913 BUG_ON(btrfs_file_extent_offset(leaf, fi) ||
914 btrfs_file_extent_compression(leaf, fi) ||
915 btrfs_file_extent_encryption(leaf, fi) ||
916 btrfs_file_extent_other_encoding(leaf, fi));
917
918 if (num_bytes != btrfs_file_extent_disk_num_bytes(leaf, fi)) {
919 ret = -EINVAL;
920 goto out;
921 }
922
923 *new_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
924 ret = 0;
925out:
926 btrfs_free_path(path);
927 return ret;
928}
929
930/*
931 * update file extent items in the tree leaf to point to
932 * the new locations.
933 */
934static noinline_for_stack
935int replace_file_extents(struct btrfs_trans_handle *trans,
936 struct reloc_control *rc,
937 struct btrfs_root *root,
938 struct extent_buffer *leaf)
939{
940 struct btrfs_fs_info *fs_info = root->fs_info;
941 struct btrfs_key key;
942 struct btrfs_file_extent_item *fi;
943 struct btrfs_inode *inode = NULL;
944 u64 parent;
945 u64 bytenr;
946 u64 new_bytenr = 0;
947 u64 num_bytes;
948 u64 end;
949 u32 nritems;
950 u32 i;
951 int ret = 0;
952 int first = 1;
953 int dirty = 0;
954
955 if (rc->stage != UPDATE_DATA_PTRS)
956 return 0;
957
958 /* reloc trees always use full backref */
959 if (btrfs_root_id(root) == BTRFS_TREE_RELOC_OBJECTID)
960 parent = leaf->start;
961 else
962 parent = 0;
963
964 nritems = btrfs_header_nritems(leaf);
965 for (i = 0; i < nritems; i++) {
966 struct btrfs_ref ref = { 0 };
967
968 cond_resched();
969 btrfs_item_key_to_cpu(leaf, &key, i);
970 if (key.type != BTRFS_EXTENT_DATA_KEY)
971 continue;
972 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
973 if (btrfs_file_extent_type(leaf, fi) ==
974 BTRFS_FILE_EXTENT_INLINE)
975 continue;
976 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
977 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
978 if (bytenr == 0)
979 continue;
980 if (!in_range(bytenr, rc->block_group->start,
981 rc->block_group->length))
982 continue;
983
984 /*
985 * if we are modifying block in fs tree, wait for read_folio
986 * to complete and drop the extent cache
987 */
988 if (btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID) {
989 if (first) {
990 inode = btrfs_find_first_inode(root, key.objectid);
991 first = 0;
992 } else if (inode && btrfs_ino(inode) < key.objectid) {
993 btrfs_add_delayed_iput(inode);
994 inode = btrfs_find_first_inode(root, key.objectid);
995 }
996 if (inode && btrfs_ino(inode) == key.objectid) {
997 struct extent_state *cached_state = NULL;
998
999 end = key.offset +
1000 btrfs_file_extent_num_bytes(leaf, fi);
1001 WARN_ON(!IS_ALIGNED(key.offset,
1002 fs_info->sectorsize));
1003 WARN_ON(!IS_ALIGNED(end, fs_info->sectorsize));
1004 end--;
1005 /* Take mmap lock to serialize with reflinks. */
1006 if (!down_read_trylock(&inode->i_mmap_lock))
1007 continue;
1008 ret = try_lock_extent(&inode->io_tree, key.offset,
1009 end, &cached_state);
1010 if (!ret) {
1011 up_read(&inode->i_mmap_lock);
1012 continue;
1013 }
1014
1015 btrfs_drop_extent_map_range(inode, key.offset, end, true);
1016 unlock_extent(&inode->io_tree, key.offset, end,
1017 &cached_state);
1018 up_read(&inode->i_mmap_lock);
1019 }
1020 }
1021
1022 ret = get_new_location(rc->data_inode, &new_bytenr,
1023 bytenr, num_bytes);
1024 if (ret) {
1025 /*
1026 * Don't have to abort since we've not changed anything
1027 * in the file extent yet.
1028 */
1029 break;
1030 }
1031
1032 btrfs_set_file_extent_disk_bytenr(leaf, fi, new_bytenr);
1033 dirty = 1;
1034
1035 key.offset -= btrfs_file_extent_offset(leaf, fi);
1036 ref.action = BTRFS_ADD_DELAYED_REF;
1037 ref.bytenr = new_bytenr;
1038 ref.num_bytes = num_bytes;
1039 ref.parent = parent;
1040 ref.owning_root = btrfs_root_id(root);
1041 ref.ref_root = btrfs_header_owner(leaf);
1042 btrfs_init_data_ref(&ref, key.objectid, key.offset,
1043 btrfs_root_id(root), false);
1044 ret = btrfs_inc_extent_ref(trans, &ref);
1045 if (ret) {
1046 btrfs_abort_transaction(trans, ret);
1047 break;
1048 }
1049
1050 ref.action = BTRFS_DROP_DELAYED_REF;
1051 ref.bytenr = bytenr;
1052 ref.num_bytes = num_bytes;
1053 ref.parent = parent;
1054 ref.owning_root = btrfs_root_id(root);
1055 ref.ref_root = btrfs_header_owner(leaf);
1056 btrfs_init_data_ref(&ref, key.objectid, key.offset,
1057 btrfs_root_id(root), false);
1058 ret = btrfs_free_extent(trans, &ref);
1059 if (ret) {
1060 btrfs_abort_transaction(trans, ret);
1061 break;
1062 }
1063 }
1064 if (dirty)
1065 btrfs_mark_buffer_dirty(trans, leaf);
1066 if (inode)
1067 btrfs_add_delayed_iput(inode);
1068 return ret;
1069}
1070
1071static noinline_for_stack int memcmp_node_keys(const struct extent_buffer *eb,
1072 int slot, const struct btrfs_path *path,
1073 int level)
1074{
1075 struct btrfs_disk_key key1;
1076 struct btrfs_disk_key key2;
1077 btrfs_node_key(eb, &key1, slot);
1078 btrfs_node_key(path->nodes[level], &key2, path->slots[level]);
1079 return memcmp(&key1, &key2, sizeof(key1));
1080}
1081
1082/*
1083 * try to replace tree blocks in fs tree with the new blocks
1084 * in reloc tree. tree blocks haven't been modified since the
1085 * reloc tree was create can be replaced.
1086 *
1087 * if a block was replaced, level of the block + 1 is returned.
1088 * if no block got replaced, 0 is returned. if there are other
1089 * errors, a negative error number is returned.
1090 */
1091static noinline_for_stack
1092int replace_path(struct btrfs_trans_handle *trans, struct reloc_control *rc,
1093 struct btrfs_root *dest, struct btrfs_root *src,
1094 struct btrfs_path *path, struct btrfs_key *next_key,
1095 int lowest_level, int max_level)
1096{
1097 struct btrfs_fs_info *fs_info = dest->fs_info;
1098 struct extent_buffer *eb;
1099 struct extent_buffer *parent;
1100 struct btrfs_ref ref = { 0 };
1101 struct btrfs_key key;
1102 u64 old_bytenr;
1103 u64 new_bytenr;
1104 u64 old_ptr_gen;
1105 u64 new_ptr_gen;
1106 u64 last_snapshot;
1107 u32 blocksize;
1108 int cow = 0;
1109 int level;
1110 int ret;
1111 int slot;
1112
1113 ASSERT(btrfs_root_id(src) == BTRFS_TREE_RELOC_OBJECTID);
1114 ASSERT(btrfs_root_id(dest) != BTRFS_TREE_RELOC_OBJECTID);
1115
1116 last_snapshot = btrfs_root_last_snapshot(&src->root_item);
1117again:
1118 slot = path->slots[lowest_level];
1119 btrfs_node_key_to_cpu(path->nodes[lowest_level], &key, slot);
1120
1121 eb = btrfs_lock_root_node(dest);
1122 level = btrfs_header_level(eb);
1123
1124 if (level < lowest_level) {
1125 btrfs_tree_unlock(eb);
1126 free_extent_buffer(eb);
1127 return 0;
1128 }
1129
1130 if (cow) {
1131 ret = btrfs_cow_block(trans, dest, eb, NULL, 0, &eb,
1132 BTRFS_NESTING_COW);
1133 if (ret) {
1134 btrfs_tree_unlock(eb);
1135 free_extent_buffer(eb);
1136 return ret;
1137 }
1138 }
1139
1140 if (next_key) {
1141 next_key->objectid = (u64)-1;
1142 next_key->type = (u8)-1;
1143 next_key->offset = (u64)-1;
1144 }
1145
1146 parent = eb;
1147 while (1) {
1148 level = btrfs_header_level(parent);
1149 ASSERT(level >= lowest_level);
1150
1151 ret = btrfs_bin_search(parent, 0, &key, &slot);
1152 if (ret < 0)
1153 break;
1154 if (ret && slot > 0)
1155 slot--;
1156
1157 if (next_key && slot + 1 < btrfs_header_nritems(parent))
1158 btrfs_node_key_to_cpu(parent, next_key, slot + 1);
1159
1160 old_bytenr = btrfs_node_blockptr(parent, slot);
1161 blocksize = fs_info->nodesize;
1162 old_ptr_gen = btrfs_node_ptr_generation(parent, slot);
1163
1164 if (level <= max_level) {
1165 eb = path->nodes[level];
1166 new_bytenr = btrfs_node_blockptr(eb,
1167 path->slots[level]);
1168 new_ptr_gen = btrfs_node_ptr_generation(eb,
1169 path->slots[level]);
1170 } else {
1171 new_bytenr = 0;
1172 new_ptr_gen = 0;
1173 }
1174
1175 if (WARN_ON(new_bytenr > 0 && new_bytenr == old_bytenr)) {
1176 ret = level;
1177 break;
1178 }
1179
1180 if (new_bytenr == 0 || old_ptr_gen > last_snapshot ||
1181 memcmp_node_keys(parent, slot, path, level)) {
1182 if (level <= lowest_level) {
1183 ret = 0;
1184 break;
1185 }
1186
1187 eb = btrfs_read_node_slot(parent, slot);
1188 if (IS_ERR(eb)) {
1189 ret = PTR_ERR(eb);
1190 break;
1191 }
1192 btrfs_tree_lock(eb);
1193 if (cow) {
1194 ret = btrfs_cow_block(trans, dest, eb, parent,
1195 slot, &eb,
1196 BTRFS_NESTING_COW);
1197 if (ret) {
1198 btrfs_tree_unlock(eb);
1199 free_extent_buffer(eb);
1200 break;
1201 }
1202 }
1203
1204 btrfs_tree_unlock(parent);
1205 free_extent_buffer(parent);
1206
1207 parent = eb;
1208 continue;
1209 }
1210
1211 if (!cow) {
1212 btrfs_tree_unlock(parent);
1213 free_extent_buffer(parent);
1214 cow = 1;
1215 goto again;
1216 }
1217
1218 btrfs_node_key_to_cpu(path->nodes[level], &key,
1219 path->slots[level]);
1220 btrfs_release_path(path);
1221
1222 path->lowest_level = level;
1223 set_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &src->state);
1224 ret = btrfs_search_slot(trans, src, &key, path, 0, 1);
1225 clear_bit(BTRFS_ROOT_RESET_LOCKDEP_CLASS, &src->state);
1226 path->lowest_level = 0;
1227 if (ret) {
1228 if (ret > 0)
1229 ret = -ENOENT;
1230 break;
1231 }
1232
1233 /*
1234 * Info qgroup to trace both subtrees.
1235 *
1236 * We must trace both trees.
1237 * 1) Tree reloc subtree
1238 * If not traced, we will leak data numbers
1239 * 2) Fs subtree
1240 * If not traced, we will double count old data
1241 *
1242 * We don't scan the subtree right now, but only record
1243 * the swapped tree blocks.
1244 * The real subtree rescan is delayed until we have new
1245 * CoW on the subtree root node before transaction commit.
1246 */
1247 ret = btrfs_qgroup_add_swapped_blocks(dest,
1248 rc->block_group, parent, slot,
1249 path->nodes[level], path->slots[level],
1250 last_snapshot);
1251 if (ret < 0)
1252 break;
1253 /*
1254 * swap blocks in fs tree and reloc tree.
1255 */
1256 btrfs_set_node_blockptr(parent, slot, new_bytenr);
1257 btrfs_set_node_ptr_generation(parent, slot, new_ptr_gen);
1258 btrfs_mark_buffer_dirty(trans, parent);
1259
1260 btrfs_set_node_blockptr(path->nodes[level],
1261 path->slots[level], old_bytenr);
1262 btrfs_set_node_ptr_generation(path->nodes[level],
1263 path->slots[level], old_ptr_gen);
1264 btrfs_mark_buffer_dirty(trans, path->nodes[level]);
1265
1266 ref.action = BTRFS_ADD_DELAYED_REF;
1267 ref.bytenr = old_bytenr;
1268 ref.num_bytes = blocksize;
1269 ref.parent = path->nodes[level]->start;
1270 ref.owning_root = btrfs_root_id(src);
1271 ref.ref_root = btrfs_root_id(src);
1272 btrfs_init_tree_ref(&ref, level - 1, 0, true);
1273 ret = btrfs_inc_extent_ref(trans, &ref);
1274 if (ret) {
1275 btrfs_abort_transaction(trans, ret);
1276 break;
1277 }
1278
1279 ref.action = BTRFS_ADD_DELAYED_REF;
1280 ref.bytenr = new_bytenr;
1281 ref.num_bytes = blocksize;
1282 ref.parent = 0;
1283 ref.owning_root = btrfs_root_id(dest);
1284 ref.ref_root = btrfs_root_id(dest);
1285 btrfs_init_tree_ref(&ref, level - 1, 0, true);
1286 ret = btrfs_inc_extent_ref(trans, &ref);
1287 if (ret) {
1288 btrfs_abort_transaction(trans, ret);
1289 break;
1290 }
1291
1292 /* We don't know the real owning_root, use 0. */
1293 ref.action = BTRFS_DROP_DELAYED_REF;
1294 ref.bytenr = new_bytenr;
1295 ref.num_bytes = blocksize;
1296 ref.parent = path->nodes[level]->start;
1297 ref.owning_root = 0;
1298 ref.ref_root = btrfs_root_id(src);
1299 btrfs_init_tree_ref(&ref, level - 1, 0, true);
1300 ret = btrfs_free_extent(trans, &ref);
1301 if (ret) {
1302 btrfs_abort_transaction(trans, ret);
1303 break;
1304 }
1305
1306 /* We don't know the real owning_root, use 0. */
1307 ref.action = BTRFS_DROP_DELAYED_REF;
1308 ref.bytenr = old_bytenr;
1309 ref.num_bytes = blocksize;
1310 ref.parent = 0;
1311 ref.owning_root = 0;
1312 ref.ref_root = btrfs_root_id(dest);
1313 btrfs_init_tree_ref(&ref, level - 1, 0, true);
1314 ret = btrfs_free_extent(trans, &ref);
1315 if (ret) {
1316 btrfs_abort_transaction(trans, ret);
1317 break;
1318 }
1319
1320 btrfs_unlock_up_safe(path, 0);
1321
1322 ret = level;
1323 break;
1324 }
1325 btrfs_tree_unlock(parent);
1326 free_extent_buffer(parent);
1327 return ret;
1328}
1329
1330/*
1331 * helper to find next relocated block in reloc tree
1332 */
1333static noinline_for_stack
1334int walk_up_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
1335 int *level)
1336{
1337 struct extent_buffer *eb;
1338 int i;
1339 u64 last_snapshot;
1340 u32 nritems;
1341
1342 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1343
1344 for (i = 0; i < *level; i++) {
1345 free_extent_buffer(path->nodes[i]);
1346 path->nodes[i] = NULL;
1347 }
1348
1349 for (i = *level; i < BTRFS_MAX_LEVEL && path->nodes[i]; i++) {
1350 eb = path->nodes[i];
1351 nritems = btrfs_header_nritems(eb);
1352 while (path->slots[i] + 1 < nritems) {
1353 path->slots[i]++;
1354 if (btrfs_node_ptr_generation(eb, path->slots[i]) <=
1355 last_snapshot)
1356 continue;
1357
1358 *level = i;
1359 return 0;
1360 }
1361 free_extent_buffer(path->nodes[i]);
1362 path->nodes[i] = NULL;
1363 }
1364 return 1;
1365}
1366
1367/*
1368 * walk down reloc tree to find relocated block of lowest level
1369 */
1370static noinline_for_stack
1371int walk_down_reloc_tree(struct btrfs_root *root, struct btrfs_path *path,
1372 int *level)
1373{
1374 struct extent_buffer *eb = NULL;
1375 int i;
1376 u64 ptr_gen = 0;
1377 u64 last_snapshot;
1378 u32 nritems;
1379
1380 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1381
1382 for (i = *level; i > 0; i--) {
1383 eb = path->nodes[i];
1384 nritems = btrfs_header_nritems(eb);
1385 while (path->slots[i] < nritems) {
1386 ptr_gen = btrfs_node_ptr_generation(eb, path->slots[i]);
1387 if (ptr_gen > last_snapshot)
1388 break;
1389 path->slots[i]++;
1390 }
1391 if (path->slots[i] >= nritems) {
1392 if (i == *level)
1393 break;
1394 *level = i + 1;
1395 return 0;
1396 }
1397 if (i == 1) {
1398 *level = i;
1399 return 0;
1400 }
1401
1402 eb = btrfs_read_node_slot(eb, path->slots[i]);
1403 if (IS_ERR(eb))
1404 return PTR_ERR(eb);
1405 BUG_ON(btrfs_header_level(eb) != i - 1);
1406 path->nodes[i - 1] = eb;
1407 path->slots[i - 1] = 0;
1408 }
1409 return 1;
1410}
1411
1412/*
1413 * invalidate extent cache for file extents whose key in range of
1414 * [min_key, max_key)
1415 */
1416static int invalidate_extent_cache(struct btrfs_root *root,
1417 const struct btrfs_key *min_key,
1418 const struct btrfs_key *max_key)
1419{
1420 struct btrfs_fs_info *fs_info = root->fs_info;
1421 struct btrfs_inode *inode = NULL;
1422 u64 objectid;
1423 u64 start, end;
1424 u64 ino;
1425
1426 objectid = min_key->objectid;
1427 while (1) {
1428 struct extent_state *cached_state = NULL;
1429
1430 cond_resched();
1431 if (inode)
1432 iput(&inode->vfs_inode);
1433
1434 if (objectid > max_key->objectid)
1435 break;
1436
1437 inode = btrfs_find_first_inode(root, objectid);
1438 if (!inode)
1439 break;
1440 ino = btrfs_ino(inode);
1441
1442 if (ino > max_key->objectid) {
1443 iput(&inode->vfs_inode);
1444 break;
1445 }
1446
1447 objectid = ino + 1;
1448 if (!S_ISREG(inode->vfs_inode.i_mode))
1449 continue;
1450
1451 if (unlikely(min_key->objectid == ino)) {
1452 if (min_key->type > BTRFS_EXTENT_DATA_KEY)
1453 continue;
1454 if (min_key->type < BTRFS_EXTENT_DATA_KEY)
1455 start = 0;
1456 else {
1457 start = min_key->offset;
1458 WARN_ON(!IS_ALIGNED(start, fs_info->sectorsize));
1459 }
1460 } else {
1461 start = 0;
1462 }
1463
1464 if (unlikely(max_key->objectid == ino)) {
1465 if (max_key->type < BTRFS_EXTENT_DATA_KEY)
1466 continue;
1467 if (max_key->type > BTRFS_EXTENT_DATA_KEY) {
1468 end = (u64)-1;
1469 } else {
1470 if (max_key->offset == 0)
1471 continue;
1472 end = max_key->offset;
1473 WARN_ON(!IS_ALIGNED(end, fs_info->sectorsize));
1474 end--;
1475 }
1476 } else {
1477 end = (u64)-1;
1478 }
1479
1480 /* the lock_extent waits for read_folio to complete */
1481 lock_extent(&inode->io_tree, start, end, &cached_state);
1482 btrfs_drop_extent_map_range(inode, start, end, true);
1483 unlock_extent(&inode->io_tree, start, end, &cached_state);
1484 }
1485 return 0;
1486}
1487
1488static int find_next_key(struct btrfs_path *path, int level,
1489 struct btrfs_key *key)
1490
1491{
1492 while (level < BTRFS_MAX_LEVEL) {
1493 if (!path->nodes[level])
1494 break;
1495 if (path->slots[level] + 1 <
1496 btrfs_header_nritems(path->nodes[level])) {
1497 btrfs_node_key_to_cpu(path->nodes[level], key,
1498 path->slots[level] + 1);
1499 return 0;
1500 }
1501 level++;
1502 }
1503 return 1;
1504}
1505
1506/*
1507 * Insert current subvolume into reloc_control::dirty_subvol_roots
1508 */
1509static int insert_dirty_subvol(struct btrfs_trans_handle *trans,
1510 struct reloc_control *rc,
1511 struct btrfs_root *root)
1512{
1513 struct btrfs_root *reloc_root = root->reloc_root;
1514 struct btrfs_root_item *reloc_root_item;
1515 int ret;
1516
1517 /* @root must be a subvolume tree root with a valid reloc tree */
1518 ASSERT(btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID);
1519 ASSERT(reloc_root);
1520
1521 reloc_root_item = &reloc_root->root_item;
1522 memset(&reloc_root_item->drop_progress, 0,
1523 sizeof(reloc_root_item->drop_progress));
1524 btrfs_set_root_drop_level(reloc_root_item, 0);
1525 btrfs_set_root_refs(reloc_root_item, 0);
1526 ret = btrfs_update_reloc_root(trans, root);
1527 if (ret)
1528 return ret;
1529
1530 if (list_empty(&root->reloc_dirty_list)) {
1531 btrfs_grab_root(root);
1532 list_add_tail(&root->reloc_dirty_list, &rc->dirty_subvol_roots);
1533 }
1534
1535 return 0;
1536}
1537
1538static int clean_dirty_subvols(struct reloc_control *rc)
1539{
1540 struct btrfs_root *root;
1541 struct btrfs_root *next;
1542 int ret = 0;
1543 int ret2;
1544
1545 list_for_each_entry_safe(root, next, &rc->dirty_subvol_roots,
1546 reloc_dirty_list) {
1547 if (btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID) {
1548 /* Merged subvolume, cleanup its reloc root */
1549 struct btrfs_root *reloc_root = root->reloc_root;
1550
1551 list_del_init(&root->reloc_dirty_list);
1552 root->reloc_root = NULL;
1553 /*
1554 * Need barrier to ensure clear_bit() only happens after
1555 * root->reloc_root = NULL. Pairs with have_reloc_root.
1556 */
1557 smp_wmb();
1558 clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE, &root->state);
1559 if (reloc_root) {
1560 /*
1561 * btrfs_drop_snapshot drops our ref we hold for
1562 * ->reloc_root. If it fails however we must
1563 * drop the ref ourselves.
1564 */
1565 ret2 = btrfs_drop_snapshot(reloc_root, 0, 1);
1566 if (ret2 < 0) {
1567 btrfs_put_root(reloc_root);
1568 if (!ret)
1569 ret = ret2;
1570 }
1571 }
1572 btrfs_put_root(root);
1573 } else {
1574 /* Orphan reloc tree, just clean it up */
1575 ret2 = btrfs_drop_snapshot(root, 0, 1);
1576 if (ret2 < 0) {
1577 btrfs_put_root(root);
1578 if (!ret)
1579 ret = ret2;
1580 }
1581 }
1582 }
1583 return ret;
1584}
1585
1586/*
1587 * merge the relocated tree blocks in reloc tree with corresponding
1588 * fs tree.
1589 */
1590static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
1591 struct btrfs_root *root)
1592{
1593 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
1594 struct btrfs_key key;
1595 struct btrfs_key next_key;
1596 struct btrfs_trans_handle *trans = NULL;
1597 struct btrfs_root *reloc_root;
1598 struct btrfs_root_item *root_item;
1599 struct btrfs_path *path;
1600 struct extent_buffer *leaf;
1601 int reserve_level;
1602 int level;
1603 int max_level;
1604 int replaced = 0;
1605 int ret = 0;
1606 u32 min_reserved;
1607
1608 path = btrfs_alloc_path();
1609 if (!path)
1610 return -ENOMEM;
1611 path->reada = READA_FORWARD;
1612
1613 reloc_root = root->reloc_root;
1614 root_item = &reloc_root->root_item;
1615
1616 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1617 level = btrfs_root_level(root_item);
1618 atomic_inc(&reloc_root->node->refs);
1619 path->nodes[level] = reloc_root->node;
1620 path->slots[level] = 0;
1621 } else {
1622 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1623
1624 level = btrfs_root_drop_level(root_item);
1625 BUG_ON(level == 0);
1626 path->lowest_level = level;
1627 ret = btrfs_search_slot(NULL, reloc_root, &key, path, 0, 0);
1628 path->lowest_level = 0;
1629 if (ret < 0) {
1630 btrfs_free_path(path);
1631 return ret;
1632 }
1633
1634 btrfs_node_key_to_cpu(path->nodes[level], &next_key,
1635 path->slots[level]);
1636 WARN_ON(memcmp(&key, &next_key, sizeof(key)));
1637
1638 btrfs_unlock_up_safe(path, 0);
1639 }
1640
1641 /*
1642 * In merge_reloc_root(), we modify the upper level pointer to swap the
1643 * tree blocks between reloc tree and subvolume tree. Thus for tree
1644 * block COW, we COW at most from level 1 to root level for each tree.
1645 *
1646 * Thus the needed metadata size is at most root_level * nodesize,
1647 * and * 2 since we have two trees to COW.
1648 */
1649 reserve_level = max_t(int, 1, btrfs_root_level(root_item));
1650 min_reserved = fs_info->nodesize * reserve_level * 2;
1651 memset(&next_key, 0, sizeof(next_key));
1652
1653 while (1) {
1654 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv,
1655 min_reserved,
1656 BTRFS_RESERVE_FLUSH_LIMIT);
1657 if (ret)
1658 goto out;
1659 trans = btrfs_start_transaction(root, 0);
1660 if (IS_ERR(trans)) {
1661 ret = PTR_ERR(trans);
1662 trans = NULL;
1663 goto out;
1664 }
1665
1666 /*
1667 * At this point we no longer have a reloc_control, so we can't
1668 * depend on btrfs_init_reloc_root to update our last_trans.
1669 *
1670 * But that's ok, we started the trans handle on our
1671 * corresponding fs_root, which means it's been added to the
1672 * dirty list. At commit time we'll still call
1673 * btrfs_update_reloc_root() and update our root item
1674 * appropriately.
1675 */
1676 btrfs_set_root_last_trans(reloc_root, trans->transid);
1677 trans->block_rsv = rc->block_rsv;
1678
1679 replaced = 0;
1680 max_level = level;
1681
1682 ret = walk_down_reloc_tree(reloc_root, path, &level);
1683 if (ret < 0)
1684 goto out;
1685 if (ret > 0)
1686 break;
1687
1688 if (!find_next_key(path, level, &key) &&
1689 btrfs_comp_cpu_keys(&next_key, &key) >= 0) {
1690 ret = 0;
1691 } else {
1692 ret = replace_path(trans, rc, root, reloc_root, path,
1693 &next_key, level, max_level);
1694 }
1695 if (ret < 0)
1696 goto out;
1697 if (ret > 0) {
1698 level = ret;
1699 btrfs_node_key_to_cpu(path->nodes[level], &key,
1700 path->slots[level]);
1701 replaced = 1;
1702 }
1703
1704 ret = walk_up_reloc_tree(reloc_root, path, &level);
1705 if (ret > 0)
1706 break;
1707
1708 BUG_ON(level == 0);
1709 /*
1710 * save the merging progress in the drop_progress.
1711 * this is OK since root refs == 1 in this case.
1712 */
1713 btrfs_node_key(path->nodes[level], &root_item->drop_progress,
1714 path->slots[level]);
1715 btrfs_set_root_drop_level(root_item, level);
1716
1717 btrfs_end_transaction_throttle(trans);
1718 trans = NULL;
1719
1720 btrfs_btree_balance_dirty(fs_info);
1721
1722 if (replaced && rc->stage == UPDATE_DATA_PTRS)
1723 invalidate_extent_cache(root, &key, &next_key);
1724 }
1725
1726 /*
1727 * handle the case only one block in the fs tree need to be
1728 * relocated and the block is tree root.
1729 */
1730 leaf = btrfs_lock_root_node(root);
1731 ret = btrfs_cow_block(trans, root, leaf, NULL, 0, &leaf,
1732 BTRFS_NESTING_COW);
1733 btrfs_tree_unlock(leaf);
1734 free_extent_buffer(leaf);
1735out:
1736 btrfs_free_path(path);
1737
1738 if (ret == 0) {
1739 ret = insert_dirty_subvol(trans, rc, root);
1740 if (ret)
1741 btrfs_abort_transaction(trans, ret);
1742 }
1743
1744 if (trans)
1745 btrfs_end_transaction_throttle(trans);
1746
1747 btrfs_btree_balance_dirty(fs_info);
1748
1749 if (replaced && rc->stage == UPDATE_DATA_PTRS)
1750 invalidate_extent_cache(root, &key, &next_key);
1751
1752 return ret;
1753}
1754
1755static noinline_for_stack
1756int prepare_to_merge(struct reloc_control *rc, int err)
1757{
1758 struct btrfs_root *root = rc->extent_root;
1759 struct btrfs_fs_info *fs_info = root->fs_info;
1760 struct btrfs_root *reloc_root;
1761 struct btrfs_trans_handle *trans;
1762 LIST_HEAD(reloc_roots);
1763 u64 num_bytes = 0;
1764 int ret;
1765
1766 mutex_lock(&fs_info->reloc_mutex);
1767 rc->merging_rsv_size += fs_info->nodesize * (BTRFS_MAX_LEVEL - 1) * 2;
1768 rc->merging_rsv_size += rc->nodes_relocated * 2;
1769 mutex_unlock(&fs_info->reloc_mutex);
1770
1771again:
1772 if (!err) {
1773 num_bytes = rc->merging_rsv_size;
1774 ret = btrfs_block_rsv_add(fs_info, rc->block_rsv, num_bytes,
1775 BTRFS_RESERVE_FLUSH_ALL);
1776 if (ret)
1777 err = ret;
1778 }
1779
1780 trans = btrfs_join_transaction(rc->extent_root);
1781 if (IS_ERR(trans)) {
1782 if (!err)
1783 btrfs_block_rsv_release(fs_info, rc->block_rsv,
1784 num_bytes, NULL);
1785 return PTR_ERR(trans);
1786 }
1787
1788 if (!err) {
1789 if (num_bytes != rc->merging_rsv_size) {
1790 btrfs_end_transaction(trans);
1791 btrfs_block_rsv_release(fs_info, rc->block_rsv,
1792 num_bytes, NULL);
1793 goto again;
1794 }
1795 }
1796
1797 rc->merge_reloc_tree = true;
1798
1799 while (!list_empty(&rc->reloc_roots)) {
1800 reloc_root = list_entry(rc->reloc_roots.next,
1801 struct btrfs_root, root_list);
1802 list_del_init(&reloc_root->root_list);
1803
1804 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
1805 false);
1806 if (IS_ERR(root)) {
1807 /*
1808 * Even if we have an error we need this reloc root
1809 * back on our list so we can clean up properly.
1810 */
1811 list_add(&reloc_root->root_list, &reloc_roots);
1812 btrfs_abort_transaction(trans, (int)PTR_ERR(root));
1813 if (!err)
1814 err = PTR_ERR(root);
1815 break;
1816 }
1817
1818 if (unlikely(root->reloc_root != reloc_root)) {
1819 if (root->reloc_root) {
1820 btrfs_err(fs_info,
1821"reloc tree mismatch, root %lld has reloc root key (%lld %u %llu) gen %llu, expect reloc root key (%lld %u %llu) gen %llu",
1822 btrfs_root_id(root),
1823 btrfs_root_id(root->reloc_root),
1824 root->reloc_root->root_key.type,
1825 root->reloc_root->root_key.offset,
1826 btrfs_root_generation(
1827 &root->reloc_root->root_item),
1828 btrfs_root_id(reloc_root),
1829 reloc_root->root_key.type,
1830 reloc_root->root_key.offset,
1831 btrfs_root_generation(
1832 &reloc_root->root_item));
1833 } else {
1834 btrfs_err(fs_info,
1835"reloc tree mismatch, root %lld has no reloc root, expect reloc root key (%lld %u %llu) gen %llu",
1836 btrfs_root_id(root),
1837 btrfs_root_id(reloc_root),
1838 reloc_root->root_key.type,
1839 reloc_root->root_key.offset,
1840 btrfs_root_generation(
1841 &reloc_root->root_item));
1842 }
1843 list_add(&reloc_root->root_list, &reloc_roots);
1844 btrfs_put_root(root);
1845 btrfs_abort_transaction(trans, -EUCLEAN);
1846 if (!err)
1847 err = -EUCLEAN;
1848 break;
1849 }
1850
1851 /*
1852 * set reference count to 1, so btrfs_recover_relocation
1853 * knows it should resumes merging
1854 */
1855 if (!err)
1856 btrfs_set_root_refs(&reloc_root->root_item, 1);
1857 ret = btrfs_update_reloc_root(trans, root);
1858
1859 /*
1860 * Even if we have an error we need this reloc root back on our
1861 * list so we can clean up properly.
1862 */
1863 list_add(&reloc_root->root_list, &reloc_roots);
1864 btrfs_put_root(root);
1865
1866 if (ret) {
1867 btrfs_abort_transaction(trans, ret);
1868 if (!err)
1869 err = ret;
1870 break;
1871 }
1872 }
1873
1874 list_splice(&reloc_roots, &rc->reloc_roots);
1875
1876 if (!err)
1877 err = btrfs_commit_transaction(trans);
1878 else
1879 btrfs_end_transaction(trans);
1880 return err;
1881}
1882
1883static noinline_for_stack
1884void free_reloc_roots(struct list_head *list)
1885{
1886 struct btrfs_root *reloc_root, *tmp;
1887
1888 list_for_each_entry_safe(reloc_root, tmp, list, root_list)
1889 __del_reloc_root(reloc_root);
1890}
1891
1892static noinline_for_stack
1893void merge_reloc_roots(struct reloc_control *rc)
1894{
1895 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
1896 struct btrfs_root *root;
1897 struct btrfs_root *reloc_root;
1898 LIST_HEAD(reloc_roots);
1899 int found = 0;
1900 int ret = 0;
1901again:
1902 root = rc->extent_root;
1903
1904 /*
1905 * this serializes us with btrfs_record_root_in_transaction,
1906 * we have to make sure nobody is in the middle of
1907 * adding their roots to the list while we are
1908 * doing this splice
1909 */
1910 mutex_lock(&fs_info->reloc_mutex);
1911 list_splice_init(&rc->reloc_roots, &reloc_roots);
1912 mutex_unlock(&fs_info->reloc_mutex);
1913
1914 while (!list_empty(&reloc_roots)) {
1915 found = 1;
1916 reloc_root = list_entry(reloc_roots.next,
1917 struct btrfs_root, root_list);
1918
1919 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
1920 false);
1921 if (btrfs_root_refs(&reloc_root->root_item) > 0) {
1922 if (WARN_ON(IS_ERR(root))) {
1923 /*
1924 * For recovery we read the fs roots on mount,
1925 * and if we didn't find the root then we marked
1926 * the reloc root as a garbage root. For normal
1927 * relocation obviously the root should exist in
1928 * memory. However there's no reason we can't
1929 * handle the error properly here just in case.
1930 */
1931 ret = PTR_ERR(root);
1932 goto out;
1933 }
1934 if (WARN_ON(root->reloc_root != reloc_root)) {
1935 /*
1936 * This can happen if on-disk metadata has some
1937 * corruption, e.g. bad reloc tree key offset.
1938 */
1939 ret = -EINVAL;
1940 goto out;
1941 }
1942 ret = merge_reloc_root(rc, root);
1943 btrfs_put_root(root);
1944 if (ret) {
1945 if (list_empty(&reloc_root->root_list))
1946 list_add_tail(&reloc_root->root_list,
1947 &reloc_roots);
1948 goto out;
1949 }
1950 } else {
1951 if (!IS_ERR(root)) {
1952 if (root->reloc_root == reloc_root) {
1953 root->reloc_root = NULL;
1954 btrfs_put_root(reloc_root);
1955 }
1956 clear_bit(BTRFS_ROOT_DEAD_RELOC_TREE,
1957 &root->state);
1958 btrfs_put_root(root);
1959 }
1960
1961 list_del_init(&reloc_root->root_list);
1962 /* Don't forget to queue this reloc root for cleanup */
1963 list_add_tail(&reloc_root->reloc_dirty_list,
1964 &rc->dirty_subvol_roots);
1965 }
1966 }
1967
1968 if (found) {
1969 found = 0;
1970 goto again;
1971 }
1972out:
1973 if (ret) {
1974 btrfs_handle_fs_error(fs_info, ret, NULL);
1975 free_reloc_roots(&reloc_roots);
1976
1977 /* new reloc root may be added */
1978 mutex_lock(&fs_info->reloc_mutex);
1979 list_splice_init(&rc->reloc_roots, &reloc_roots);
1980 mutex_unlock(&fs_info->reloc_mutex);
1981 free_reloc_roots(&reloc_roots);
1982 }
1983
1984 /*
1985 * We used to have
1986 *
1987 * BUG_ON(!RB_EMPTY_ROOT(&rc->reloc_root_tree.rb_root));
1988 *
1989 * here, but it's wrong. If we fail to start the transaction in
1990 * prepare_to_merge() we will have only 0 ref reloc roots, none of which
1991 * have actually been removed from the reloc_root_tree rb tree. This is
1992 * fine because we're bailing here, and we hold a reference on the root
1993 * for the list that holds it, so these roots will be cleaned up when we
1994 * do the reloc_dirty_list afterwards. Meanwhile the root->reloc_root
1995 * will be cleaned up on unmount.
1996 *
1997 * The remaining nodes will be cleaned up by free_reloc_control.
1998 */
1999}
2000
2001static void free_block_list(struct rb_root *blocks)
2002{
2003 struct tree_block *block;
2004 struct rb_node *rb_node;
2005 while ((rb_node = rb_first(blocks))) {
2006 block = rb_entry(rb_node, struct tree_block, rb_node);
2007 rb_erase(rb_node, blocks);
2008 kfree(block);
2009 }
2010}
2011
2012static int record_reloc_root_in_trans(struct btrfs_trans_handle *trans,
2013 struct btrfs_root *reloc_root)
2014{
2015 struct btrfs_fs_info *fs_info = reloc_root->fs_info;
2016 struct btrfs_root *root;
2017 int ret;
2018
2019 if (btrfs_get_root_last_trans(reloc_root) == trans->transid)
2020 return 0;
2021
2022 root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset, false);
2023
2024 /*
2025 * This should succeed, since we can't have a reloc root without having
2026 * already looked up the actual root and created the reloc root for this
2027 * root.
2028 *
2029 * However if there's some sort of corruption where we have a ref to a
2030 * reloc root without a corresponding root this could return ENOENT.
2031 */
2032 if (IS_ERR(root)) {
2033 ASSERT(0);
2034 return PTR_ERR(root);
2035 }
2036 if (root->reloc_root != reloc_root) {
2037 ASSERT(0);
2038 btrfs_err(fs_info,
2039 "root %llu has two reloc roots associated with it",
2040 reloc_root->root_key.offset);
2041 btrfs_put_root(root);
2042 return -EUCLEAN;
2043 }
2044 ret = btrfs_record_root_in_trans(trans, root);
2045 btrfs_put_root(root);
2046
2047 return ret;
2048}
2049
2050static noinline_for_stack
2051struct btrfs_root *select_reloc_root(struct btrfs_trans_handle *trans,
2052 struct reloc_control *rc,
2053 struct btrfs_backref_node *node,
2054 struct btrfs_backref_edge *edges[])
2055{
2056 struct btrfs_backref_node *next;
2057 struct btrfs_root *root;
2058 int index = 0;
2059 int ret;
2060
2061 next = node;
2062 while (1) {
2063 cond_resched();
2064 next = walk_up_backref(next, edges, &index);
2065 root = next->root;
2066
2067 /*
2068 * If there is no root, then our references for this block are
2069 * incomplete, as we should be able to walk all the way up to a
2070 * block that is owned by a root.
2071 *
2072 * This path is only for SHAREABLE roots, so if we come upon a
2073 * non-SHAREABLE root then we have backrefs that resolve
2074 * improperly.
2075 *
2076 * Both of these cases indicate file system corruption, or a bug
2077 * in the backref walking code.
2078 */
2079 if (!root) {
2080 ASSERT(0);
2081 btrfs_err(trans->fs_info,
2082 "bytenr %llu doesn't have a backref path ending in a root",
2083 node->bytenr);
2084 return ERR_PTR(-EUCLEAN);
2085 }
2086 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) {
2087 ASSERT(0);
2088 btrfs_err(trans->fs_info,
2089 "bytenr %llu has multiple refs with one ending in a non-shareable root",
2090 node->bytenr);
2091 return ERR_PTR(-EUCLEAN);
2092 }
2093
2094 if (btrfs_root_id(root) == BTRFS_TREE_RELOC_OBJECTID) {
2095 ret = record_reloc_root_in_trans(trans, root);
2096 if (ret)
2097 return ERR_PTR(ret);
2098 break;
2099 }
2100
2101 ret = btrfs_record_root_in_trans(trans, root);
2102 if (ret)
2103 return ERR_PTR(ret);
2104 root = root->reloc_root;
2105
2106 /*
2107 * We could have raced with another thread which failed, so
2108 * root->reloc_root may not be set, return ENOENT in this case.
2109 */
2110 if (!root)
2111 return ERR_PTR(-ENOENT);
2112
2113 if (next->new_bytenr != root->node->start) {
2114 /*
2115 * We just created the reloc root, so we shouldn't have
2116 * ->new_bytenr set and this shouldn't be in the changed
2117 * list. If it is then we have multiple roots pointing
2118 * at the same bytenr which indicates corruption, or
2119 * we've made a mistake in the backref walking code.
2120 */
2121 ASSERT(next->new_bytenr == 0);
2122 ASSERT(list_empty(&next->list));
2123 if (next->new_bytenr || !list_empty(&next->list)) {
2124 btrfs_err(trans->fs_info,
2125 "bytenr %llu possibly has multiple roots pointing at the same bytenr %llu",
2126 node->bytenr, next->bytenr);
2127 return ERR_PTR(-EUCLEAN);
2128 }
2129
2130 next->new_bytenr = root->node->start;
2131 btrfs_put_root(next->root);
2132 next->root = btrfs_grab_root(root);
2133 ASSERT(next->root);
2134 list_add_tail(&next->list,
2135 &rc->backref_cache.changed);
2136 mark_block_processed(rc, next);
2137 break;
2138 }
2139
2140 WARN_ON(1);
2141 root = NULL;
2142 next = walk_down_backref(edges, &index);
2143 if (!next || next->level <= node->level)
2144 break;
2145 }
2146 if (!root) {
2147 /*
2148 * This can happen if there's fs corruption or if there's a bug
2149 * in the backref lookup code.
2150 */
2151 ASSERT(0);
2152 return ERR_PTR(-ENOENT);
2153 }
2154
2155 next = node;
2156 /* setup backref node path for btrfs_reloc_cow_block */
2157 while (1) {
2158 rc->backref_cache.path[next->level] = next;
2159 if (--index < 0)
2160 break;
2161 next = edges[index]->node[UPPER];
2162 }
2163 return root;
2164}
2165
2166/*
2167 * Select a tree root for relocation.
2168 *
2169 * Return NULL if the block is not shareable. We should use do_relocation() in
2170 * this case.
2171 *
2172 * Return a tree root pointer if the block is shareable.
2173 * Return -ENOENT if the block is root of reloc tree.
2174 */
2175static noinline_for_stack
2176struct btrfs_root *select_one_root(struct btrfs_backref_node *node)
2177{
2178 struct btrfs_backref_node *next;
2179 struct btrfs_root *root;
2180 struct btrfs_root *fs_root = NULL;
2181 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2182 int index = 0;
2183
2184 next = node;
2185 while (1) {
2186 cond_resched();
2187 next = walk_up_backref(next, edges, &index);
2188 root = next->root;
2189
2190 /*
2191 * This can occur if we have incomplete extent refs leading all
2192 * the way up a particular path, in this case return -EUCLEAN.
2193 */
2194 if (!root)
2195 return ERR_PTR(-EUCLEAN);
2196
2197 /* No other choice for non-shareable tree */
2198 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
2199 return root;
2200
2201 if (btrfs_root_id(root) != BTRFS_TREE_RELOC_OBJECTID)
2202 fs_root = root;
2203
2204 if (next != node)
2205 return NULL;
2206
2207 next = walk_down_backref(edges, &index);
2208 if (!next || next->level <= node->level)
2209 break;
2210 }
2211
2212 if (!fs_root)
2213 return ERR_PTR(-ENOENT);
2214 return fs_root;
2215}
2216
2217static noinline_for_stack u64 calcu_metadata_size(struct reloc_control *rc,
2218 struct btrfs_backref_node *node)
2219{
2220 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
2221 struct btrfs_backref_node *next = node;
2222 struct btrfs_backref_edge *edge;
2223 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2224 u64 num_bytes = 0;
2225 int index = 0;
2226
2227 BUG_ON(node->processed);
2228
2229 while (next) {
2230 cond_resched();
2231 while (1) {
2232 if (next->processed)
2233 break;
2234
2235 num_bytes += fs_info->nodesize;
2236
2237 if (list_empty(&next->upper))
2238 break;
2239
2240 edge = list_entry(next->upper.next,
2241 struct btrfs_backref_edge, list[LOWER]);
2242 edges[index++] = edge;
2243 next = edge->node[UPPER];
2244 }
2245 next = walk_down_backref(edges, &index);
2246 }
2247 return num_bytes;
2248}
2249
2250static int reserve_metadata_space(struct btrfs_trans_handle *trans,
2251 struct reloc_control *rc,
2252 struct btrfs_backref_node *node)
2253{
2254 struct btrfs_root *root = rc->extent_root;
2255 struct btrfs_fs_info *fs_info = root->fs_info;
2256 u64 num_bytes;
2257 int ret;
2258 u64 tmp;
2259
2260 num_bytes = calcu_metadata_size(rc, node) * 2;
2261
2262 trans->block_rsv = rc->block_rsv;
2263 rc->reserved_bytes += num_bytes;
2264
2265 /*
2266 * We are under a transaction here so we can only do limited flushing.
2267 * If we get an enospc just kick back -EAGAIN so we know to drop the
2268 * transaction and try to refill when we can flush all the things.
2269 */
2270 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv, num_bytes,
2271 BTRFS_RESERVE_FLUSH_LIMIT);
2272 if (ret) {
2273 tmp = fs_info->nodesize * RELOCATION_RESERVED_NODES;
2274 while (tmp <= rc->reserved_bytes)
2275 tmp <<= 1;
2276 /*
2277 * only one thread can access block_rsv at this point,
2278 * so we don't need hold lock to protect block_rsv.
2279 * we expand more reservation size here to allow enough
2280 * space for relocation and we will return earlier in
2281 * enospc case.
2282 */
2283 rc->block_rsv->size = tmp + fs_info->nodesize *
2284 RELOCATION_RESERVED_NODES;
2285 return -EAGAIN;
2286 }
2287
2288 return 0;
2289}
2290
2291/*
2292 * relocate a block tree, and then update pointers in upper level
2293 * blocks that reference the block to point to the new location.
2294 *
2295 * if called by link_to_upper, the block has already been relocated.
2296 * in that case this function just updates pointers.
2297 */
2298static int do_relocation(struct btrfs_trans_handle *trans,
2299 struct reloc_control *rc,
2300 struct btrfs_backref_node *node,
2301 struct btrfs_key *key,
2302 struct btrfs_path *path, int lowest)
2303{
2304 struct btrfs_backref_node *upper;
2305 struct btrfs_backref_edge *edge;
2306 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2307 struct btrfs_root *root;
2308 struct extent_buffer *eb;
2309 u32 blocksize;
2310 u64 bytenr;
2311 int slot;
2312 int ret = 0;
2313
2314 /*
2315 * If we are lowest then this is the first time we're processing this
2316 * block, and thus shouldn't have an eb associated with it yet.
2317 */
2318 ASSERT(!lowest || !node->eb);
2319
2320 path->lowest_level = node->level + 1;
2321 rc->backref_cache.path[node->level] = node;
2322 list_for_each_entry(edge, &node->upper, list[LOWER]) {
2323 cond_resched();
2324
2325 upper = edge->node[UPPER];
2326 root = select_reloc_root(trans, rc, upper, edges);
2327 if (IS_ERR(root)) {
2328 ret = PTR_ERR(root);
2329 goto next;
2330 }
2331
2332 if (upper->eb && !upper->locked) {
2333 if (!lowest) {
2334 ret = btrfs_bin_search(upper->eb, 0, key, &slot);
2335 if (ret < 0)
2336 goto next;
2337 BUG_ON(ret);
2338 bytenr = btrfs_node_blockptr(upper->eb, slot);
2339 if (node->eb->start == bytenr)
2340 goto next;
2341 }
2342 btrfs_backref_drop_node_buffer(upper);
2343 }
2344
2345 if (!upper->eb) {
2346 ret = btrfs_search_slot(trans, root, key, path, 0, 1);
2347 if (ret) {
2348 if (ret > 0)
2349 ret = -ENOENT;
2350
2351 btrfs_release_path(path);
2352 break;
2353 }
2354
2355 if (!upper->eb) {
2356 upper->eb = path->nodes[upper->level];
2357 path->nodes[upper->level] = NULL;
2358 } else {
2359 BUG_ON(upper->eb != path->nodes[upper->level]);
2360 }
2361
2362 upper->locked = 1;
2363 path->locks[upper->level] = 0;
2364
2365 slot = path->slots[upper->level];
2366 btrfs_release_path(path);
2367 } else {
2368 ret = btrfs_bin_search(upper->eb, 0, key, &slot);
2369 if (ret < 0)
2370 goto next;
2371 BUG_ON(ret);
2372 }
2373
2374 bytenr = btrfs_node_blockptr(upper->eb, slot);
2375 if (lowest) {
2376 if (bytenr != node->bytenr) {
2377 btrfs_err(root->fs_info,
2378 "lowest leaf/node mismatch: bytenr %llu node->bytenr %llu slot %d upper %llu",
2379 bytenr, node->bytenr, slot,
2380 upper->eb->start);
2381 ret = -EIO;
2382 goto next;
2383 }
2384 } else {
2385 if (node->eb->start == bytenr)
2386 goto next;
2387 }
2388
2389 blocksize = root->fs_info->nodesize;
2390 eb = btrfs_read_node_slot(upper->eb, slot);
2391 if (IS_ERR(eb)) {
2392 ret = PTR_ERR(eb);
2393 goto next;
2394 }
2395 btrfs_tree_lock(eb);
2396
2397 if (!node->eb) {
2398 ret = btrfs_cow_block(trans, root, eb, upper->eb,
2399 slot, &eb, BTRFS_NESTING_COW);
2400 btrfs_tree_unlock(eb);
2401 free_extent_buffer(eb);
2402 if (ret < 0)
2403 goto next;
2404 /*
2405 * We've just COWed this block, it should have updated
2406 * the correct backref node entry.
2407 */
2408 ASSERT(node->eb == eb);
2409 } else {
2410 struct btrfs_ref ref = {
2411 .action = BTRFS_ADD_DELAYED_REF,
2412 .bytenr = node->eb->start,
2413 .num_bytes = blocksize,
2414 .parent = upper->eb->start,
2415 .owning_root = btrfs_header_owner(upper->eb),
2416 .ref_root = btrfs_header_owner(upper->eb),
2417 };
2418
2419 btrfs_set_node_blockptr(upper->eb, slot,
2420 node->eb->start);
2421 btrfs_set_node_ptr_generation(upper->eb, slot,
2422 trans->transid);
2423 btrfs_mark_buffer_dirty(trans, upper->eb);
2424
2425 btrfs_init_tree_ref(&ref, node->level,
2426 btrfs_root_id(root), false);
2427 ret = btrfs_inc_extent_ref(trans, &ref);
2428 if (!ret)
2429 ret = btrfs_drop_subtree(trans, root, eb,
2430 upper->eb);
2431 if (ret)
2432 btrfs_abort_transaction(trans, ret);
2433 }
2434next:
2435 if (!upper->pending)
2436 btrfs_backref_drop_node_buffer(upper);
2437 else
2438 btrfs_backref_unlock_node_buffer(upper);
2439 if (ret)
2440 break;
2441 }
2442
2443 if (!ret && node->pending) {
2444 btrfs_backref_drop_node_buffer(node);
2445 list_move_tail(&node->list, &rc->backref_cache.changed);
2446 node->pending = 0;
2447 }
2448
2449 path->lowest_level = 0;
2450
2451 /*
2452 * We should have allocated all of our space in the block rsv and thus
2453 * shouldn't ENOSPC.
2454 */
2455 ASSERT(ret != -ENOSPC);
2456 return ret;
2457}
2458
2459static int link_to_upper(struct btrfs_trans_handle *trans,
2460 struct reloc_control *rc,
2461 struct btrfs_backref_node *node,
2462 struct btrfs_path *path)
2463{
2464 struct btrfs_key key;
2465
2466 btrfs_node_key_to_cpu(node->eb, &key, 0);
2467 return do_relocation(trans, rc, node, &key, path, 0);
2468}
2469
2470static int finish_pending_nodes(struct btrfs_trans_handle *trans,
2471 struct reloc_control *rc,
2472 struct btrfs_path *path, int err)
2473{
2474 LIST_HEAD(list);
2475 struct btrfs_backref_cache *cache = &rc->backref_cache;
2476 struct btrfs_backref_node *node;
2477 int level;
2478 int ret;
2479
2480 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2481 while (!list_empty(&cache->pending[level])) {
2482 node = list_entry(cache->pending[level].next,
2483 struct btrfs_backref_node, list);
2484 list_move_tail(&node->list, &list);
2485 BUG_ON(!node->pending);
2486
2487 if (!err) {
2488 ret = link_to_upper(trans, rc, node, path);
2489 if (ret < 0)
2490 err = ret;
2491 }
2492 }
2493 list_splice_init(&list, &cache->pending[level]);
2494 }
2495 return err;
2496}
2497
2498/*
2499 * mark a block and all blocks directly/indirectly reference the block
2500 * as processed.
2501 */
2502static void update_processed_blocks(struct reloc_control *rc,
2503 struct btrfs_backref_node *node)
2504{
2505 struct btrfs_backref_node *next = node;
2506 struct btrfs_backref_edge *edge;
2507 struct btrfs_backref_edge *edges[BTRFS_MAX_LEVEL - 1];
2508 int index = 0;
2509
2510 while (next) {
2511 cond_resched();
2512 while (1) {
2513 if (next->processed)
2514 break;
2515
2516 mark_block_processed(rc, next);
2517
2518 if (list_empty(&next->upper))
2519 break;
2520
2521 edge = list_entry(next->upper.next,
2522 struct btrfs_backref_edge, list[LOWER]);
2523 edges[index++] = edge;
2524 next = edge->node[UPPER];
2525 }
2526 next = walk_down_backref(edges, &index);
2527 }
2528}
2529
2530static int tree_block_processed(u64 bytenr, struct reloc_control *rc)
2531{
2532 u32 blocksize = rc->extent_root->fs_info->nodesize;
2533
2534 if (test_range_bit(&rc->processed_blocks, bytenr,
2535 bytenr + blocksize - 1, EXTENT_DIRTY, NULL))
2536 return 1;
2537 return 0;
2538}
2539
2540static int get_tree_block_key(struct btrfs_fs_info *fs_info,
2541 struct tree_block *block)
2542{
2543 struct btrfs_tree_parent_check check = {
2544 .level = block->level,
2545 .owner_root = block->owner,
2546 .transid = block->key.offset
2547 };
2548 struct extent_buffer *eb;
2549
2550 eb = read_tree_block(fs_info, block->bytenr, &check);
2551 if (IS_ERR(eb))
2552 return PTR_ERR(eb);
2553 if (!extent_buffer_uptodate(eb)) {
2554 free_extent_buffer(eb);
2555 return -EIO;
2556 }
2557 if (block->level == 0)
2558 btrfs_item_key_to_cpu(eb, &block->key, 0);
2559 else
2560 btrfs_node_key_to_cpu(eb, &block->key, 0);
2561 free_extent_buffer(eb);
2562 block->key_ready = true;
2563 return 0;
2564}
2565
2566/*
2567 * helper function to relocate a tree block
2568 */
2569static int relocate_tree_block(struct btrfs_trans_handle *trans,
2570 struct reloc_control *rc,
2571 struct btrfs_backref_node *node,
2572 struct btrfs_key *key,
2573 struct btrfs_path *path)
2574{
2575 struct btrfs_root *root;
2576 int ret = 0;
2577
2578 if (!node)
2579 return 0;
2580
2581 /*
2582 * If we fail here we want to drop our backref_node because we are going
2583 * to start over and regenerate the tree for it.
2584 */
2585 ret = reserve_metadata_space(trans, rc, node);
2586 if (ret)
2587 goto out;
2588
2589 BUG_ON(node->processed);
2590 root = select_one_root(node);
2591 if (IS_ERR(root)) {
2592 ret = PTR_ERR(root);
2593
2594 /* See explanation in select_one_root for the -EUCLEAN case. */
2595 ASSERT(ret == -ENOENT);
2596 if (ret == -ENOENT) {
2597 ret = 0;
2598 update_processed_blocks(rc, node);
2599 }
2600 goto out;
2601 }
2602
2603 if (root) {
2604 if (test_bit(BTRFS_ROOT_SHAREABLE, &root->state)) {
2605 /*
2606 * This block was the root block of a root, and this is
2607 * the first time we're processing the block and thus it
2608 * should not have had the ->new_bytenr modified and
2609 * should have not been included on the changed list.
2610 *
2611 * However in the case of corruption we could have
2612 * multiple refs pointing to the same block improperly,
2613 * and thus we would trip over these checks. ASSERT()
2614 * for the developer case, because it could indicate a
2615 * bug in the backref code, however error out for a
2616 * normal user in the case of corruption.
2617 */
2618 ASSERT(node->new_bytenr == 0);
2619 ASSERT(list_empty(&node->list));
2620 if (node->new_bytenr || !list_empty(&node->list)) {
2621 btrfs_err(root->fs_info,
2622 "bytenr %llu has improper references to it",
2623 node->bytenr);
2624 ret = -EUCLEAN;
2625 goto out;
2626 }
2627 ret = btrfs_record_root_in_trans(trans, root);
2628 if (ret)
2629 goto out;
2630 /*
2631 * Another thread could have failed, need to check if we
2632 * have reloc_root actually set.
2633 */
2634 if (!root->reloc_root) {
2635 ret = -ENOENT;
2636 goto out;
2637 }
2638 root = root->reloc_root;
2639 node->new_bytenr = root->node->start;
2640 btrfs_put_root(node->root);
2641 node->root = btrfs_grab_root(root);
2642 ASSERT(node->root);
2643 list_add_tail(&node->list, &rc->backref_cache.changed);
2644 } else {
2645 path->lowest_level = node->level;
2646 if (root == root->fs_info->chunk_root)
2647 btrfs_reserve_chunk_metadata(trans, false);
2648 ret = btrfs_search_slot(trans, root, key, path, 0, 1);
2649 btrfs_release_path(path);
2650 if (root == root->fs_info->chunk_root)
2651 btrfs_trans_release_chunk_metadata(trans);
2652 if (ret > 0)
2653 ret = 0;
2654 }
2655 if (!ret)
2656 update_processed_blocks(rc, node);
2657 } else {
2658 ret = do_relocation(trans, rc, node, key, path, 1);
2659 }
2660out:
2661 if (ret || node->level == 0 || node->cowonly)
2662 btrfs_backref_cleanup_node(&rc->backref_cache, node);
2663 return ret;
2664}
2665
2666/*
2667 * relocate a list of blocks
2668 */
2669static noinline_for_stack
2670int relocate_tree_blocks(struct btrfs_trans_handle *trans,
2671 struct reloc_control *rc, struct rb_root *blocks)
2672{
2673 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
2674 struct btrfs_backref_node *node;
2675 struct btrfs_path *path;
2676 struct tree_block *block;
2677 struct tree_block *next;
2678 int ret = 0;
2679
2680 path = btrfs_alloc_path();
2681 if (!path) {
2682 ret = -ENOMEM;
2683 goto out_free_blocks;
2684 }
2685
2686 /* Kick in readahead for tree blocks with missing keys */
2687 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2688 if (!block->key_ready)
2689 btrfs_readahead_tree_block(fs_info, block->bytenr,
2690 block->owner, 0,
2691 block->level);
2692 }
2693
2694 /* Get first keys */
2695 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2696 if (!block->key_ready) {
2697 ret = get_tree_block_key(fs_info, block);
2698 if (ret)
2699 goto out_free_path;
2700 }
2701 }
2702
2703 /* Do tree relocation */
2704 rbtree_postorder_for_each_entry_safe(block, next, blocks, rb_node) {
2705 node = build_backref_tree(trans, rc, &block->key,
2706 block->level, block->bytenr);
2707 if (IS_ERR(node)) {
2708 ret = PTR_ERR(node);
2709 goto out;
2710 }
2711
2712 ret = relocate_tree_block(trans, rc, node, &block->key,
2713 path);
2714 if (ret < 0)
2715 break;
2716 }
2717out:
2718 ret = finish_pending_nodes(trans, rc, path, ret);
2719
2720out_free_path:
2721 btrfs_free_path(path);
2722out_free_blocks:
2723 free_block_list(blocks);
2724 return ret;
2725}
2726
2727static noinline_for_stack int prealloc_file_extent_cluster(struct reloc_control *rc)
2728{
2729 const struct file_extent_cluster *cluster = &rc->cluster;
2730 struct btrfs_inode *inode = BTRFS_I(rc->data_inode);
2731 u64 alloc_hint = 0;
2732 u64 start;
2733 u64 end;
2734 u64 offset = inode->reloc_block_group_start;
2735 u64 num_bytes;
2736 int nr;
2737 int ret = 0;
2738 u64 i_size = i_size_read(&inode->vfs_inode);
2739 u64 prealloc_start = cluster->start - offset;
2740 u64 prealloc_end = cluster->end - offset;
2741 u64 cur_offset = prealloc_start;
2742
2743 /*
2744 * For subpage case, previous i_size may not be aligned to PAGE_SIZE.
2745 * This means the range [i_size, PAGE_END + 1) is filled with zeros by
2746 * btrfs_do_readpage() call of previously relocated file cluster.
2747 *
2748 * If the current cluster starts in the above range, btrfs_do_readpage()
2749 * will skip the read, and relocate_one_folio() will later writeback
2750 * the padding zeros as new data, causing data corruption.
2751 *
2752 * Here we have to manually invalidate the range (i_size, PAGE_END + 1).
2753 */
2754 if (!PAGE_ALIGNED(i_size)) {
2755 struct address_space *mapping = inode->vfs_inode.i_mapping;
2756 struct btrfs_fs_info *fs_info = inode->root->fs_info;
2757 const u32 sectorsize = fs_info->sectorsize;
2758 struct folio *folio;
2759
2760 ASSERT(sectorsize < PAGE_SIZE);
2761 ASSERT(IS_ALIGNED(i_size, sectorsize));
2762
2763 /*
2764 * Subpage can't handle page with DIRTY but without UPTODATE
2765 * bit as it can lead to the following deadlock:
2766 *
2767 * btrfs_read_folio()
2768 * | Page already *locked*
2769 * |- btrfs_lock_and_flush_ordered_range()
2770 * |- btrfs_start_ordered_extent()
2771 * |- extent_write_cache_pages()
2772 * |- lock_page()
2773 * We try to lock the page we already hold.
2774 *
2775 * Here we just writeback the whole data reloc inode, so that
2776 * we will be ensured to have no dirty range in the page, and
2777 * are safe to clear the uptodate bits.
2778 *
2779 * This shouldn't cause too much overhead, as we need to write
2780 * the data back anyway.
2781 */
2782 ret = filemap_write_and_wait(mapping);
2783 if (ret < 0)
2784 return ret;
2785
2786 clear_extent_bits(&inode->io_tree, i_size,
2787 round_up(i_size, PAGE_SIZE) - 1,
2788 EXTENT_UPTODATE);
2789 folio = filemap_lock_folio(mapping, i_size >> PAGE_SHIFT);
2790 /*
2791 * If page is freed we don't need to do anything then, as we
2792 * will re-read the whole page anyway.
2793 */
2794 if (!IS_ERR(folio)) {
2795 btrfs_subpage_clear_uptodate(fs_info, folio, i_size,
2796 round_up(i_size, PAGE_SIZE) - i_size);
2797 folio_unlock(folio);
2798 folio_put(folio);
2799 }
2800 }
2801
2802 BUG_ON(cluster->start != cluster->boundary[0]);
2803 ret = btrfs_alloc_data_chunk_ondemand(inode,
2804 prealloc_end + 1 - prealloc_start);
2805 if (ret)
2806 return ret;
2807
2808 btrfs_inode_lock(inode, 0);
2809 for (nr = 0; nr < cluster->nr; nr++) {
2810 struct extent_state *cached_state = NULL;
2811
2812 start = cluster->boundary[nr] - offset;
2813 if (nr + 1 < cluster->nr)
2814 end = cluster->boundary[nr + 1] - 1 - offset;
2815 else
2816 end = cluster->end - offset;
2817
2818 lock_extent(&inode->io_tree, start, end, &cached_state);
2819 num_bytes = end + 1 - start;
2820 ret = btrfs_prealloc_file_range(&inode->vfs_inode, 0, start,
2821 num_bytes, num_bytes,
2822 end + 1, &alloc_hint);
2823 cur_offset = end + 1;
2824 unlock_extent(&inode->io_tree, start, end, &cached_state);
2825 if (ret)
2826 break;
2827 }
2828 btrfs_inode_unlock(inode, 0);
2829
2830 if (cur_offset < prealloc_end)
2831 btrfs_free_reserved_data_space_noquota(inode->root->fs_info,
2832 prealloc_end + 1 - cur_offset);
2833 return ret;
2834}
2835
2836static noinline_for_stack int setup_relocation_extent_mapping(struct reloc_control *rc)
2837{
2838 struct btrfs_inode *inode = BTRFS_I(rc->data_inode);
2839 struct extent_map *em;
2840 struct extent_state *cached_state = NULL;
2841 u64 offset = inode->reloc_block_group_start;
2842 u64 start = rc->cluster.start - offset;
2843 u64 end = rc->cluster.end - offset;
2844 int ret = 0;
2845
2846 em = alloc_extent_map();
2847 if (!em)
2848 return -ENOMEM;
2849
2850 em->start = start;
2851 em->len = end + 1 - start;
2852 em->disk_bytenr = rc->cluster.start;
2853 em->disk_num_bytes = em->len;
2854 em->ram_bytes = em->len;
2855 em->flags |= EXTENT_FLAG_PINNED;
2856
2857 lock_extent(&inode->io_tree, start, end, &cached_state);
2858 ret = btrfs_replace_extent_map_range(inode, em, false);
2859 unlock_extent(&inode->io_tree, start, end, &cached_state);
2860 free_extent_map(em);
2861
2862 return ret;
2863}
2864
2865/*
2866 * Allow error injection to test balance/relocation cancellation
2867 */
2868noinline int btrfs_should_cancel_balance(const struct btrfs_fs_info *fs_info)
2869{
2870 return atomic_read(&fs_info->balance_cancel_req) ||
2871 atomic_read(&fs_info->reloc_cancel_req) ||
2872 fatal_signal_pending(current);
2873}
2874ALLOW_ERROR_INJECTION(btrfs_should_cancel_balance, TRUE);
2875
2876static u64 get_cluster_boundary_end(const struct file_extent_cluster *cluster,
2877 int cluster_nr)
2878{
2879 /* Last extent, use cluster end directly */
2880 if (cluster_nr >= cluster->nr - 1)
2881 return cluster->end;
2882
2883 /* Use next boundary start*/
2884 return cluster->boundary[cluster_nr + 1] - 1;
2885}
2886
2887static int relocate_one_folio(struct reloc_control *rc,
2888 struct file_ra_state *ra,
2889 int *cluster_nr, unsigned long index)
2890{
2891 const struct file_extent_cluster *cluster = &rc->cluster;
2892 struct inode *inode = rc->data_inode;
2893 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2894 u64 offset = BTRFS_I(inode)->reloc_block_group_start;
2895 const unsigned long last_index = (cluster->end - offset) >> PAGE_SHIFT;
2896 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
2897 struct folio *folio;
2898 u64 folio_start;
2899 u64 folio_end;
2900 u64 cur;
2901 int ret;
2902 const bool use_rst = btrfs_need_stripe_tree_update(fs_info, rc->block_group->flags);
2903
2904 ASSERT(index <= last_index);
2905again:
2906 folio = filemap_lock_folio(inode->i_mapping, index);
2907 if (IS_ERR(folio)) {
2908
2909 /*
2910 * On relocation we're doing readahead on the relocation inode,
2911 * but if the filesystem is backed by a RAID stripe tree we can
2912 * get ENOENT (e.g. due to preallocated extents not being
2913 * mapped in the RST) from the lookup.
2914 *
2915 * But readahead doesn't handle the error and submits invalid
2916 * reads to the device, causing a assertion failures.
2917 */
2918 if (!use_rst)
2919 page_cache_sync_readahead(inode->i_mapping, ra, NULL,
2920 index, last_index + 1 - index);
2921 folio = __filemap_get_folio(inode->i_mapping, index,
2922 FGP_LOCK | FGP_ACCESSED | FGP_CREAT,
2923 mask);
2924 if (IS_ERR(folio))
2925 return PTR_ERR(folio);
2926 }
2927
2928 WARN_ON(folio_order(folio));
2929
2930 if (folio_test_readahead(folio) && !use_rst)
2931 page_cache_async_readahead(inode->i_mapping, ra, NULL,
2932 folio, last_index + 1 - index);
2933
2934 if (!folio_test_uptodate(folio)) {
2935 btrfs_read_folio(NULL, folio);
2936 folio_lock(folio);
2937 if (!folio_test_uptodate(folio)) {
2938 ret = -EIO;
2939 goto release_folio;
2940 }
2941 if (folio->mapping != inode->i_mapping) {
2942 folio_unlock(folio);
2943 folio_put(folio);
2944 goto again;
2945 }
2946 }
2947
2948 /*
2949 * We could have lost folio private when we dropped the lock to read the
2950 * folio above, make sure we set_page_extent_mapped here so we have any
2951 * of the subpage blocksize stuff we need in place.
2952 */
2953 ret = set_folio_extent_mapped(folio);
2954 if (ret < 0)
2955 goto release_folio;
2956
2957 folio_start = folio_pos(folio);
2958 folio_end = folio_start + PAGE_SIZE - 1;
2959
2960 /*
2961 * Start from the cluster, as for subpage case, the cluster can start
2962 * inside the folio.
2963 */
2964 cur = max(folio_start, cluster->boundary[*cluster_nr] - offset);
2965 while (cur <= folio_end) {
2966 struct extent_state *cached_state = NULL;
2967 u64 extent_start = cluster->boundary[*cluster_nr] - offset;
2968 u64 extent_end = get_cluster_boundary_end(cluster,
2969 *cluster_nr) - offset;
2970 u64 clamped_start = max(folio_start, extent_start);
2971 u64 clamped_end = min(folio_end, extent_end);
2972 u32 clamped_len = clamped_end + 1 - clamped_start;
2973
2974 /* Reserve metadata for this range */
2975 ret = btrfs_delalloc_reserve_metadata(BTRFS_I(inode),
2976 clamped_len, clamped_len,
2977 false);
2978 if (ret)
2979 goto release_folio;
2980
2981 /* Mark the range delalloc and dirty for later writeback */
2982 lock_extent(&BTRFS_I(inode)->io_tree, clamped_start, clamped_end,
2983 &cached_state);
2984 ret = btrfs_set_extent_delalloc(BTRFS_I(inode), clamped_start,
2985 clamped_end, 0, &cached_state);
2986 if (ret) {
2987 clear_extent_bit(&BTRFS_I(inode)->io_tree,
2988 clamped_start, clamped_end,
2989 EXTENT_LOCKED | EXTENT_BOUNDARY,
2990 &cached_state);
2991 btrfs_delalloc_release_metadata(BTRFS_I(inode),
2992 clamped_len, true);
2993 btrfs_delalloc_release_extents(BTRFS_I(inode),
2994 clamped_len);
2995 goto release_folio;
2996 }
2997 btrfs_folio_set_dirty(fs_info, folio, clamped_start, clamped_len);
2998
2999 /*
3000 * Set the boundary if it's inside the folio.
3001 * Data relocation requires the destination extents to have the
3002 * same size as the source.
3003 * EXTENT_BOUNDARY bit prevents current extent from being merged
3004 * with previous extent.
3005 */
3006 if (in_range(cluster->boundary[*cluster_nr] - offset, folio_start, PAGE_SIZE)) {
3007 u64 boundary_start = cluster->boundary[*cluster_nr] -
3008 offset;
3009 u64 boundary_end = boundary_start +
3010 fs_info->sectorsize - 1;
3011
3012 set_extent_bit(&BTRFS_I(inode)->io_tree,
3013 boundary_start, boundary_end,
3014 EXTENT_BOUNDARY, NULL);
3015 }
3016 unlock_extent(&BTRFS_I(inode)->io_tree, clamped_start, clamped_end,
3017 &cached_state);
3018 btrfs_delalloc_release_extents(BTRFS_I(inode), clamped_len);
3019 cur += clamped_len;
3020
3021 /* Crossed extent end, go to next extent */
3022 if (cur >= extent_end) {
3023 (*cluster_nr)++;
3024 /* Just finished the last extent of the cluster, exit. */
3025 if (*cluster_nr >= cluster->nr)
3026 break;
3027 }
3028 }
3029 folio_unlock(folio);
3030 folio_put(folio);
3031
3032 balance_dirty_pages_ratelimited(inode->i_mapping);
3033 btrfs_throttle(fs_info);
3034 if (btrfs_should_cancel_balance(fs_info))
3035 ret = -ECANCELED;
3036 return ret;
3037
3038release_folio:
3039 folio_unlock(folio);
3040 folio_put(folio);
3041 return ret;
3042}
3043
3044static int relocate_file_extent_cluster(struct reloc_control *rc)
3045{
3046 struct inode *inode = rc->data_inode;
3047 const struct file_extent_cluster *cluster = &rc->cluster;
3048 u64 offset = BTRFS_I(inode)->reloc_block_group_start;
3049 unsigned long index;
3050 unsigned long last_index;
3051 struct file_ra_state *ra;
3052 int cluster_nr = 0;
3053 int ret = 0;
3054
3055 if (!cluster->nr)
3056 return 0;
3057
3058 ra = kzalloc(sizeof(*ra), GFP_NOFS);
3059 if (!ra)
3060 return -ENOMEM;
3061
3062 ret = prealloc_file_extent_cluster(rc);
3063 if (ret)
3064 goto out;
3065
3066 file_ra_state_init(ra, inode->i_mapping);
3067
3068 ret = setup_relocation_extent_mapping(rc);
3069 if (ret)
3070 goto out;
3071
3072 last_index = (cluster->end - offset) >> PAGE_SHIFT;
3073 for (index = (cluster->start - offset) >> PAGE_SHIFT;
3074 index <= last_index && !ret; index++)
3075 ret = relocate_one_folio(rc, ra, &cluster_nr, index);
3076 if (ret == 0)
3077 WARN_ON(cluster_nr != cluster->nr);
3078out:
3079 kfree(ra);
3080 return ret;
3081}
3082
3083static noinline_for_stack int relocate_data_extent(struct reloc_control *rc,
3084 const struct btrfs_key *extent_key)
3085{
3086 struct inode *inode = rc->data_inode;
3087 struct file_extent_cluster *cluster = &rc->cluster;
3088 int ret;
3089 struct btrfs_root *root = BTRFS_I(inode)->root;
3090
3091 if (cluster->nr > 0 && extent_key->objectid != cluster->end + 1) {
3092 ret = relocate_file_extent_cluster(rc);
3093 if (ret)
3094 return ret;
3095 cluster->nr = 0;
3096 }
3097
3098 /*
3099 * Under simple quotas, we set root->relocation_src_root when we find
3100 * the extent. If adjacent extents have different owners, we can't merge
3101 * them while relocating. Handle this by storing the owning root that
3102 * started a cluster and if we see an extent from a different root break
3103 * cluster formation (just like the above case of non-adjacent extents).
3104 *
3105 * Without simple quotas, relocation_src_root is always 0, so we should
3106 * never see a mismatch, and it should have no effect on relocation
3107 * clusters.
3108 */
3109 if (cluster->nr > 0 && cluster->owning_root != root->relocation_src_root) {
3110 u64 tmp = root->relocation_src_root;
3111
3112 /*
3113 * root->relocation_src_root is the state that actually affects
3114 * the preallocation we do here, so set it to the root owning
3115 * the cluster we need to relocate.
3116 */
3117 root->relocation_src_root = cluster->owning_root;
3118 ret = relocate_file_extent_cluster(rc);
3119 if (ret)
3120 return ret;
3121 cluster->nr = 0;
3122 /* And reset it back for the current extent's owning root. */
3123 root->relocation_src_root = tmp;
3124 }
3125
3126 if (!cluster->nr) {
3127 cluster->start = extent_key->objectid;
3128 cluster->owning_root = root->relocation_src_root;
3129 }
3130 else
3131 BUG_ON(cluster->nr >= MAX_EXTENTS);
3132 cluster->end = extent_key->objectid + extent_key->offset - 1;
3133 cluster->boundary[cluster->nr] = extent_key->objectid;
3134 cluster->nr++;
3135
3136 if (cluster->nr >= MAX_EXTENTS) {
3137 ret = relocate_file_extent_cluster(rc);
3138 if (ret)
3139 return ret;
3140 cluster->nr = 0;
3141 }
3142 return 0;
3143}
3144
3145/*
3146 * helper to add a tree block to the list.
3147 * the major work is getting the generation and level of the block
3148 */
3149static int add_tree_block(struct reloc_control *rc,
3150 const struct btrfs_key *extent_key,
3151 struct btrfs_path *path,
3152 struct rb_root *blocks)
3153{
3154 struct extent_buffer *eb;
3155 struct btrfs_extent_item *ei;
3156 struct btrfs_tree_block_info *bi;
3157 struct tree_block *block;
3158 struct rb_node *rb_node;
3159 u32 item_size;
3160 int level = -1;
3161 u64 generation;
3162 u64 owner = 0;
3163
3164 eb = path->nodes[0];
3165 item_size = btrfs_item_size(eb, path->slots[0]);
3166
3167 if (extent_key->type == BTRFS_METADATA_ITEM_KEY ||
3168 item_size >= sizeof(*ei) + sizeof(*bi)) {
3169 unsigned long ptr = 0, end;
3170
3171 ei = btrfs_item_ptr(eb, path->slots[0],
3172 struct btrfs_extent_item);
3173 end = (unsigned long)ei + item_size;
3174 if (extent_key->type == BTRFS_EXTENT_ITEM_KEY) {
3175 bi = (struct btrfs_tree_block_info *)(ei + 1);
3176 level = btrfs_tree_block_level(eb, bi);
3177 ptr = (unsigned long)(bi + 1);
3178 } else {
3179 level = (int)extent_key->offset;
3180 ptr = (unsigned long)(ei + 1);
3181 }
3182 generation = btrfs_extent_generation(eb, ei);
3183
3184 /*
3185 * We're reading random blocks without knowing their owner ahead
3186 * of time. This is ok most of the time, as all reloc roots and
3187 * fs roots have the same lock type. However normal trees do
3188 * not, and the only way to know ahead of time is to read the
3189 * inline ref offset. We know it's an fs root if
3190 *
3191 * 1. There's more than one ref.
3192 * 2. There's a SHARED_DATA_REF_KEY set.
3193 * 3. FULL_BACKREF is set on the flags.
3194 *
3195 * Otherwise it's safe to assume that the ref offset == the
3196 * owner of this block, so we can use that when calling
3197 * read_tree_block.
3198 */
3199 if (btrfs_extent_refs(eb, ei) == 1 &&
3200 !(btrfs_extent_flags(eb, ei) &
3201 BTRFS_BLOCK_FLAG_FULL_BACKREF) &&
3202 ptr < end) {
3203 struct btrfs_extent_inline_ref *iref;
3204 int type;
3205
3206 iref = (struct btrfs_extent_inline_ref *)ptr;
3207 type = btrfs_get_extent_inline_ref_type(eb, iref,
3208 BTRFS_REF_TYPE_BLOCK);
3209 if (type == BTRFS_REF_TYPE_INVALID)
3210 return -EINVAL;
3211 if (type == BTRFS_TREE_BLOCK_REF_KEY)
3212 owner = btrfs_extent_inline_ref_offset(eb, iref);
3213 }
3214 } else {
3215 btrfs_print_leaf(eb);
3216 btrfs_err(rc->block_group->fs_info,
3217 "unrecognized tree backref at tree block %llu slot %u",
3218 eb->start, path->slots[0]);
3219 btrfs_release_path(path);
3220 return -EUCLEAN;
3221 }
3222
3223 btrfs_release_path(path);
3224
3225 BUG_ON(level == -1);
3226
3227 block = kmalloc(sizeof(*block), GFP_NOFS);
3228 if (!block)
3229 return -ENOMEM;
3230
3231 block->bytenr = extent_key->objectid;
3232 block->key.objectid = rc->extent_root->fs_info->nodesize;
3233 block->key.offset = generation;
3234 block->level = level;
3235 block->key_ready = false;
3236 block->owner = owner;
3237
3238 rb_node = rb_simple_insert(blocks, block->bytenr, &block->rb_node);
3239 if (rb_node)
3240 btrfs_backref_panic(rc->extent_root->fs_info, block->bytenr,
3241 -EEXIST);
3242
3243 return 0;
3244}
3245
3246/*
3247 * helper to add tree blocks for backref of type BTRFS_SHARED_DATA_REF_KEY
3248 */
3249static int __add_tree_block(struct reloc_control *rc,
3250 u64 bytenr, u32 blocksize,
3251 struct rb_root *blocks)
3252{
3253 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3254 struct btrfs_path *path;
3255 struct btrfs_key key;
3256 int ret;
3257 bool skinny = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
3258
3259 if (tree_block_processed(bytenr, rc))
3260 return 0;
3261
3262 if (rb_simple_search(blocks, bytenr))
3263 return 0;
3264
3265 path = btrfs_alloc_path();
3266 if (!path)
3267 return -ENOMEM;
3268again:
3269 key.objectid = bytenr;
3270 if (skinny) {
3271 key.type = BTRFS_METADATA_ITEM_KEY;
3272 key.offset = (u64)-1;
3273 } else {
3274 key.type = BTRFS_EXTENT_ITEM_KEY;
3275 key.offset = blocksize;
3276 }
3277
3278 path->search_commit_root = 1;
3279 path->skip_locking = 1;
3280 ret = btrfs_search_slot(NULL, rc->extent_root, &key, path, 0, 0);
3281 if (ret < 0)
3282 goto out;
3283
3284 if (ret > 0 && skinny) {
3285 if (path->slots[0]) {
3286 path->slots[0]--;
3287 btrfs_item_key_to_cpu(path->nodes[0], &key,
3288 path->slots[0]);
3289 if (key.objectid == bytenr &&
3290 (key.type == BTRFS_METADATA_ITEM_KEY ||
3291 (key.type == BTRFS_EXTENT_ITEM_KEY &&
3292 key.offset == blocksize)))
3293 ret = 0;
3294 }
3295
3296 if (ret) {
3297 skinny = false;
3298 btrfs_release_path(path);
3299 goto again;
3300 }
3301 }
3302 if (ret) {
3303 ASSERT(ret == 1);
3304 btrfs_print_leaf(path->nodes[0]);
3305 btrfs_err(fs_info,
3306 "tree block extent item (%llu) is not found in extent tree",
3307 bytenr);
3308 WARN_ON(1);
3309 ret = -EINVAL;
3310 goto out;
3311 }
3312
3313 ret = add_tree_block(rc, &key, path, blocks);
3314out:
3315 btrfs_free_path(path);
3316 return ret;
3317}
3318
3319static int delete_block_group_cache(struct btrfs_fs_info *fs_info,
3320 struct btrfs_block_group *block_group,
3321 struct inode *inode,
3322 u64 ino)
3323{
3324 struct btrfs_root *root = fs_info->tree_root;
3325 struct btrfs_trans_handle *trans;
3326 int ret = 0;
3327
3328 if (inode)
3329 goto truncate;
3330
3331 inode = btrfs_iget(ino, root);
3332 if (IS_ERR(inode))
3333 return -ENOENT;
3334
3335truncate:
3336 ret = btrfs_check_trunc_cache_free_space(fs_info,
3337 &fs_info->global_block_rsv);
3338 if (ret)
3339 goto out;
3340
3341 trans = btrfs_join_transaction(root);
3342 if (IS_ERR(trans)) {
3343 ret = PTR_ERR(trans);
3344 goto out;
3345 }
3346
3347 ret = btrfs_truncate_free_space_cache(trans, block_group, inode);
3348
3349 btrfs_end_transaction(trans);
3350 btrfs_btree_balance_dirty(fs_info);
3351out:
3352 iput(inode);
3353 return ret;
3354}
3355
3356/*
3357 * Locate the free space cache EXTENT_DATA in root tree leaf and delete the
3358 * cache inode, to avoid free space cache data extent blocking data relocation.
3359 */
3360static int delete_v1_space_cache(struct extent_buffer *leaf,
3361 struct btrfs_block_group *block_group,
3362 u64 data_bytenr)
3363{
3364 u64 space_cache_ino;
3365 struct btrfs_file_extent_item *ei;
3366 struct btrfs_key key;
3367 bool found = false;
3368 int i;
3369 int ret;
3370
3371 if (btrfs_header_owner(leaf) != BTRFS_ROOT_TREE_OBJECTID)
3372 return 0;
3373
3374 for (i = 0; i < btrfs_header_nritems(leaf); i++) {
3375 u8 type;
3376
3377 btrfs_item_key_to_cpu(leaf, &key, i);
3378 if (key.type != BTRFS_EXTENT_DATA_KEY)
3379 continue;
3380 ei = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
3381 type = btrfs_file_extent_type(leaf, ei);
3382
3383 if ((type == BTRFS_FILE_EXTENT_REG ||
3384 type == BTRFS_FILE_EXTENT_PREALLOC) &&
3385 btrfs_file_extent_disk_bytenr(leaf, ei) == data_bytenr) {
3386 found = true;
3387 space_cache_ino = key.objectid;
3388 break;
3389 }
3390 }
3391 if (!found)
3392 return -ENOENT;
3393 ret = delete_block_group_cache(leaf->fs_info, block_group, NULL,
3394 space_cache_ino);
3395 return ret;
3396}
3397
3398/*
3399 * helper to find all tree blocks that reference a given data extent
3400 */
3401static noinline_for_stack int add_data_references(struct reloc_control *rc,
3402 const struct btrfs_key *extent_key,
3403 struct btrfs_path *path,
3404 struct rb_root *blocks)
3405{
3406 struct btrfs_backref_walk_ctx ctx = { 0 };
3407 struct ulist_iterator leaf_uiter;
3408 struct ulist_node *ref_node = NULL;
3409 const u32 blocksize = rc->extent_root->fs_info->nodesize;
3410 int ret = 0;
3411
3412 btrfs_release_path(path);
3413
3414 ctx.bytenr = extent_key->objectid;
3415 ctx.skip_inode_ref_list = true;
3416 ctx.fs_info = rc->extent_root->fs_info;
3417
3418 ret = btrfs_find_all_leafs(&ctx);
3419 if (ret < 0)
3420 return ret;
3421
3422 ULIST_ITER_INIT(&leaf_uiter);
3423 while ((ref_node = ulist_next(ctx.refs, &leaf_uiter))) {
3424 struct btrfs_tree_parent_check check = { 0 };
3425 struct extent_buffer *eb;
3426
3427 eb = read_tree_block(ctx.fs_info, ref_node->val, &check);
3428 if (IS_ERR(eb)) {
3429 ret = PTR_ERR(eb);
3430 break;
3431 }
3432 ret = delete_v1_space_cache(eb, rc->block_group,
3433 extent_key->objectid);
3434 free_extent_buffer(eb);
3435 if (ret < 0)
3436 break;
3437 ret = __add_tree_block(rc, ref_node->val, blocksize, blocks);
3438 if (ret < 0)
3439 break;
3440 }
3441 if (ret < 0)
3442 free_block_list(blocks);
3443 ulist_free(ctx.refs);
3444 return ret;
3445}
3446
3447/*
3448 * helper to find next unprocessed extent
3449 */
3450static noinline_for_stack
3451int find_next_extent(struct reloc_control *rc, struct btrfs_path *path,
3452 struct btrfs_key *extent_key)
3453{
3454 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3455 struct btrfs_key key;
3456 struct extent_buffer *leaf;
3457 u64 start, end, last;
3458 int ret;
3459
3460 last = rc->block_group->start + rc->block_group->length;
3461 while (1) {
3462 bool block_found;
3463
3464 cond_resched();
3465 if (rc->search_start >= last) {
3466 ret = 1;
3467 break;
3468 }
3469
3470 key.objectid = rc->search_start;
3471 key.type = BTRFS_EXTENT_ITEM_KEY;
3472 key.offset = 0;
3473
3474 path->search_commit_root = 1;
3475 path->skip_locking = 1;
3476 ret = btrfs_search_slot(NULL, rc->extent_root, &key, path,
3477 0, 0);
3478 if (ret < 0)
3479 break;
3480next:
3481 leaf = path->nodes[0];
3482 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
3483 ret = btrfs_next_leaf(rc->extent_root, path);
3484 if (ret != 0)
3485 break;
3486 leaf = path->nodes[0];
3487 }
3488
3489 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3490 if (key.objectid >= last) {
3491 ret = 1;
3492 break;
3493 }
3494
3495 if (key.type != BTRFS_EXTENT_ITEM_KEY &&
3496 key.type != BTRFS_METADATA_ITEM_KEY) {
3497 path->slots[0]++;
3498 goto next;
3499 }
3500
3501 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3502 key.objectid + key.offset <= rc->search_start) {
3503 path->slots[0]++;
3504 goto next;
3505 }
3506
3507 if (key.type == BTRFS_METADATA_ITEM_KEY &&
3508 key.objectid + fs_info->nodesize <=
3509 rc->search_start) {
3510 path->slots[0]++;
3511 goto next;
3512 }
3513
3514 block_found = find_first_extent_bit(&rc->processed_blocks,
3515 key.objectid, &start, &end,
3516 EXTENT_DIRTY, NULL);
3517
3518 if (block_found && start <= key.objectid) {
3519 btrfs_release_path(path);
3520 rc->search_start = end + 1;
3521 } else {
3522 if (key.type == BTRFS_EXTENT_ITEM_KEY)
3523 rc->search_start = key.objectid + key.offset;
3524 else
3525 rc->search_start = key.objectid +
3526 fs_info->nodesize;
3527 memcpy(extent_key, &key, sizeof(key));
3528 return 0;
3529 }
3530 }
3531 btrfs_release_path(path);
3532 return ret;
3533}
3534
3535static void set_reloc_control(struct reloc_control *rc)
3536{
3537 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3538
3539 mutex_lock(&fs_info->reloc_mutex);
3540 fs_info->reloc_ctl = rc;
3541 mutex_unlock(&fs_info->reloc_mutex);
3542}
3543
3544static void unset_reloc_control(struct reloc_control *rc)
3545{
3546 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3547
3548 mutex_lock(&fs_info->reloc_mutex);
3549 fs_info->reloc_ctl = NULL;
3550 mutex_unlock(&fs_info->reloc_mutex);
3551}
3552
3553static noinline_for_stack
3554int prepare_to_relocate(struct reloc_control *rc)
3555{
3556 struct btrfs_trans_handle *trans;
3557 int ret;
3558
3559 rc->block_rsv = btrfs_alloc_block_rsv(rc->extent_root->fs_info,
3560 BTRFS_BLOCK_RSV_TEMP);
3561 if (!rc->block_rsv)
3562 return -ENOMEM;
3563
3564 memset(&rc->cluster, 0, sizeof(rc->cluster));
3565 rc->search_start = rc->block_group->start;
3566 rc->extents_found = 0;
3567 rc->nodes_relocated = 0;
3568 rc->merging_rsv_size = 0;
3569 rc->reserved_bytes = 0;
3570 rc->block_rsv->size = rc->extent_root->fs_info->nodesize *
3571 RELOCATION_RESERVED_NODES;
3572 ret = btrfs_block_rsv_refill(rc->extent_root->fs_info,
3573 rc->block_rsv, rc->block_rsv->size,
3574 BTRFS_RESERVE_FLUSH_ALL);
3575 if (ret)
3576 return ret;
3577
3578 rc->create_reloc_tree = true;
3579 set_reloc_control(rc);
3580
3581 trans = btrfs_join_transaction(rc->extent_root);
3582 if (IS_ERR(trans)) {
3583 unset_reloc_control(rc);
3584 /*
3585 * extent tree is not a ref_cow tree and has no reloc_root to
3586 * cleanup. And callers are responsible to free the above
3587 * block rsv.
3588 */
3589 return PTR_ERR(trans);
3590 }
3591
3592 ret = btrfs_commit_transaction(trans);
3593 if (ret)
3594 unset_reloc_control(rc);
3595
3596 return ret;
3597}
3598
3599static noinline_for_stack int relocate_block_group(struct reloc_control *rc)
3600{
3601 struct btrfs_fs_info *fs_info = rc->extent_root->fs_info;
3602 struct rb_root blocks = RB_ROOT;
3603 struct btrfs_key key;
3604 struct btrfs_trans_handle *trans = NULL;
3605 struct btrfs_path *path;
3606 struct btrfs_extent_item *ei;
3607 u64 flags;
3608 int ret;
3609 int err = 0;
3610 int progress = 0;
3611
3612 path = btrfs_alloc_path();
3613 if (!path)
3614 return -ENOMEM;
3615 path->reada = READA_FORWARD;
3616
3617 ret = prepare_to_relocate(rc);
3618 if (ret) {
3619 err = ret;
3620 goto out_free;
3621 }
3622
3623 while (1) {
3624 rc->reserved_bytes = 0;
3625 ret = btrfs_block_rsv_refill(fs_info, rc->block_rsv,
3626 rc->block_rsv->size,
3627 BTRFS_RESERVE_FLUSH_ALL);
3628 if (ret) {
3629 err = ret;
3630 break;
3631 }
3632 progress++;
3633 trans = btrfs_start_transaction(rc->extent_root, 0);
3634 if (IS_ERR(trans)) {
3635 err = PTR_ERR(trans);
3636 trans = NULL;
3637 break;
3638 }
3639restart:
3640 if (rc->backref_cache.last_trans != trans->transid)
3641 btrfs_backref_release_cache(&rc->backref_cache);
3642 rc->backref_cache.last_trans = trans->transid;
3643
3644 ret = find_next_extent(rc, path, &key);
3645 if (ret < 0)
3646 err = ret;
3647 if (ret != 0)
3648 break;
3649
3650 rc->extents_found++;
3651
3652 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3653 struct btrfs_extent_item);
3654 flags = btrfs_extent_flags(path->nodes[0], ei);
3655
3656 /*
3657 * If we are relocating a simple quota owned extent item, we
3658 * need to note the owner on the reloc data root so that when
3659 * we allocate the replacement item, we can attribute it to the
3660 * correct eventual owner (rather than the reloc data root).
3661 */
3662 if (btrfs_qgroup_mode(fs_info) == BTRFS_QGROUP_MODE_SIMPLE) {
3663 struct btrfs_root *root = BTRFS_I(rc->data_inode)->root;
3664 u64 owning_root_id = btrfs_get_extent_owner_root(fs_info,
3665 path->nodes[0],
3666 path->slots[0]);
3667
3668 root->relocation_src_root = owning_root_id;
3669 }
3670
3671 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
3672 ret = add_tree_block(rc, &key, path, &blocks);
3673 } else if (rc->stage == UPDATE_DATA_PTRS &&
3674 (flags & BTRFS_EXTENT_FLAG_DATA)) {
3675 ret = add_data_references(rc, &key, path, &blocks);
3676 } else {
3677 btrfs_release_path(path);
3678 ret = 0;
3679 }
3680 if (ret < 0) {
3681 err = ret;
3682 break;
3683 }
3684
3685 if (!RB_EMPTY_ROOT(&blocks)) {
3686 ret = relocate_tree_blocks(trans, rc, &blocks);
3687 if (ret < 0) {
3688 if (ret != -EAGAIN) {
3689 err = ret;
3690 break;
3691 }
3692 rc->extents_found--;
3693 rc->search_start = key.objectid;
3694 }
3695 }
3696
3697 btrfs_end_transaction_throttle(trans);
3698 btrfs_btree_balance_dirty(fs_info);
3699 trans = NULL;
3700
3701 if (rc->stage == MOVE_DATA_EXTENTS &&
3702 (flags & BTRFS_EXTENT_FLAG_DATA)) {
3703 rc->found_file_extent = true;
3704 ret = relocate_data_extent(rc, &key);
3705 if (ret < 0) {
3706 err = ret;
3707 break;
3708 }
3709 }
3710 if (btrfs_should_cancel_balance(fs_info)) {
3711 err = -ECANCELED;
3712 break;
3713 }
3714 }
3715 if (trans && progress && err == -ENOSPC) {
3716 ret = btrfs_force_chunk_alloc(trans, rc->block_group->flags);
3717 if (ret == 1) {
3718 err = 0;
3719 progress = 0;
3720 goto restart;
3721 }
3722 }
3723
3724 btrfs_release_path(path);
3725 clear_extent_bits(&rc->processed_blocks, 0, (u64)-1, EXTENT_DIRTY);
3726
3727 if (trans) {
3728 btrfs_end_transaction_throttle(trans);
3729 btrfs_btree_balance_dirty(fs_info);
3730 }
3731
3732 if (!err) {
3733 ret = relocate_file_extent_cluster(rc);
3734 if (ret < 0)
3735 err = ret;
3736 }
3737
3738 rc->create_reloc_tree = false;
3739 set_reloc_control(rc);
3740
3741 btrfs_backref_release_cache(&rc->backref_cache);
3742 btrfs_block_rsv_release(fs_info, rc->block_rsv, (u64)-1, NULL);
3743
3744 /*
3745 * Even in the case when the relocation is cancelled, we should all go
3746 * through prepare_to_merge() and merge_reloc_roots().
3747 *
3748 * For error (including cancelled balance), prepare_to_merge() will
3749 * mark all reloc trees orphan, then queue them for cleanup in
3750 * merge_reloc_roots()
3751 */
3752 err = prepare_to_merge(rc, err);
3753
3754 merge_reloc_roots(rc);
3755
3756 rc->merge_reloc_tree = false;
3757 unset_reloc_control(rc);
3758 btrfs_block_rsv_release(fs_info, rc->block_rsv, (u64)-1, NULL);
3759
3760 /* get rid of pinned extents */
3761 trans = btrfs_join_transaction(rc->extent_root);
3762 if (IS_ERR(trans)) {
3763 err = PTR_ERR(trans);
3764 goto out_free;
3765 }
3766 ret = btrfs_commit_transaction(trans);
3767 if (ret && !err)
3768 err = ret;
3769out_free:
3770 ret = clean_dirty_subvols(rc);
3771 if (ret < 0 && !err)
3772 err = ret;
3773 btrfs_free_block_rsv(fs_info, rc->block_rsv);
3774 btrfs_free_path(path);
3775 return err;
3776}
3777
3778static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
3779 struct btrfs_root *root, u64 objectid)
3780{
3781 struct btrfs_path *path;
3782 struct btrfs_inode_item *item;
3783 struct extent_buffer *leaf;
3784 int ret;
3785
3786 path = btrfs_alloc_path();
3787 if (!path)
3788 return -ENOMEM;
3789
3790 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
3791 if (ret)
3792 goto out;
3793
3794 leaf = path->nodes[0];
3795 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
3796 memzero_extent_buffer(leaf, (unsigned long)item, sizeof(*item));
3797 btrfs_set_inode_generation(leaf, item, 1);
3798 btrfs_set_inode_size(leaf, item, 0);
3799 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
3800 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS |
3801 BTRFS_INODE_PREALLOC);
3802 btrfs_mark_buffer_dirty(trans, leaf);
3803out:
3804 btrfs_free_path(path);
3805 return ret;
3806}
3807
3808static void delete_orphan_inode(struct btrfs_trans_handle *trans,
3809 struct btrfs_root *root, u64 objectid)
3810{
3811 struct btrfs_path *path;
3812 struct btrfs_key key;
3813 int ret = 0;
3814
3815 path = btrfs_alloc_path();
3816 if (!path) {
3817 ret = -ENOMEM;
3818 goto out;
3819 }
3820
3821 key.objectid = objectid;
3822 key.type = BTRFS_INODE_ITEM_KEY;
3823 key.offset = 0;
3824 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3825 if (ret) {
3826 if (ret > 0)
3827 ret = -ENOENT;
3828 goto out;
3829 }
3830 ret = btrfs_del_item(trans, root, path);
3831out:
3832 if (ret)
3833 btrfs_abort_transaction(trans, ret);
3834 btrfs_free_path(path);
3835}
3836
3837/*
3838 * helper to create inode for data relocation.
3839 * the inode is in data relocation tree and its link count is 0
3840 */
3841static noinline_for_stack struct inode *create_reloc_inode(
3842 struct btrfs_fs_info *fs_info,
3843 const struct btrfs_block_group *group)
3844{
3845 struct inode *inode = NULL;
3846 struct btrfs_trans_handle *trans;
3847 struct btrfs_root *root;
3848 u64 objectid;
3849 int ret = 0;
3850
3851 root = btrfs_grab_root(fs_info->data_reloc_root);
3852 trans = btrfs_start_transaction(root, 6);
3853 if (IS_ERR(trans)) {
3854 btrfs_put_root(root);
3855 return ERR_CAST(trans);
3856 }
3857
3858 ret = btrfs_get_free_objectid(root, &objectid);
3859 if (ret)
3860 goto out;
3861
3862 ret = __insert_orphan_inode(trans, root, objectid);
3863 if (ret)
3864 goto out;
3865
3866 inode = btrfs_iget(objectid, root);
3867 if (IS_ERR(inode)) {
3868 delete_orphan_inode(trans, root, objectid);
3869 ret = PTR_ERR(inode);
3870 inode = NULL;
3871 goto out;
3872 }
3873 BTRFS_I(inode)->reloc_block_group_start = group->start;
3874
3875 ret = btrfs_orphan_add(trans, BTRFS_I(inode));
3876out:
3877 btrfs_put_root(root);
3878 btrfs_end_transaction(trans);
3879 btrfs_btree_balance_dirty(fs_info);
3880 if (ret) {
3881 iput(inode);
3882 inode = ERR_PTR(ret);
3883 }
3884 return inode;
3885}
3886
3887/*
3888 * Mark start of chunk relocation that is cancellable. Check if the cancellation
3889 * has been requested meanwhile and don't start in that case.
3890 *
3891 * Return:
3892 * 0 success
3893 * -EINPROGRESS operation is already in progress, that's probably a bug
3894 * -ECANCELED cancellation request was set before the operation started
3895 */
3896static int reloc_chunk_start(struct btrfs_fs_info *fs_info)
3897{
3898 if (test_and_set_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags)) {
3899 /* This should not happen */
3900 btrfs_err(fs_info, "reloc already running, cannot start");
3901 return -EINPROGRESS;
3902 }
3903
3904 if (atomic_read(&fs_info->reloc_cancel_req) > 0) {
3905 btrfs_info(fs_info, "chunk relocation canceled on start");
3906 /*
3907 * On cancel, clear all requests but let the caller mark
3908 * the end after cleanup operations.
3909 */
3910 atomic_set(&fs_info->reloc_cancel_req, 0);
3911 return -ECANCELED;
3912 }
3913 return 0;
3914}
3915
3916/*
3917 * Mark end of chunk relocation that is cancellable and wake any waiters.
3918 */
3919static void reloc_chunk_end(struct btrfs_fs_info *fs_info)
3920{
3921 /* Requested after start, clear bit first so any waiters can continue */
3922 if (atomic_read(&fs_info->reloc_cancel_req) > 0)
3923 btrfs_info(fs_info, "chunk relocation canceled during operation");
3924 clear_and_wake_up_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags);
3925 atomic_set(&fs_info->reloc_cancel_req, 0);
3926}
3927
3928static struct reloc_control *alloc_reloc_control(struct btrfs_fs_info *fs_info)
3929{
3930 struct reloc_control *rc;
3931
3932 rc = kzalloc(sizeof(*rc), GFP_NOFS);
3933 if (!rc)
3934 return NULL;
3935
3936 INIT_LIST_HEAD(&rc->reloc_roots);
3937 INIT_LIST_HEAD(&rc->dirty_subvol_roots);
3938 btrfs_backref_init_cache(fs_info, &rc->backref_cache, true);
3939 rc->reloc_root_tree.rb_root = RB_ROOT;
3940 spin_lock_init(&rc->reloc_root_tree.lock);
3941 extent_io_tree_init(fs_info, &rc->processed_blocks, IO_TREE_RELOC_BLOCKS);
3942 return rc;
3943}
3944
3945static void free_reloc_control(struct reloc_control *rc)
3946{
3947 struct mapping_node *node, *tmp;
3948
3949 free_reloc_roots(&rc->reloc_roots);
3950 rbtree_postorder_for_each_entry_safe(node, tmp,
3951 &rc->reloc_root_tree.rb_root, rb_node)
3952 kfree(node);
3953
3954 kfree(rc);
3955}
3956
3957/*
3958 * Print the block group being relocated
3959 */
3960static void describe_relocation(struct btrfs_block_group *block_group)
3961{
3962 char buf[128] = {'\0'};
3963
3964 btrfs_describe_block_groups(block_group->flags, buf, sizeof(buf));
3965
3966 btrfs_info(block_group->fs_info, "relocating block group %llu flags %s",
3967 block_group->start, buf);
3968}
3969
3970static const char *stage_to_string(enum reloc_stage stage)
3971{
3972 if (stage == MOVE_DATA_EXTENTS)
3973 return "move data extents";
3974 if (stage == UPDATE_DATA_PTRS)
3975 return "update data pointers";
3976 return "unknown";
3977}
3978
3979/*
3980 * function to relocate all extents in a block group.
3981 */
3982int btrfs_relocate_block_group(struct btrfs_fs_info *fs_info, u64 group_start)
3983{
3984 struct btrfs_block_group *bg;
3985 struct btrfs_root *extent_root = btrfs_extent_root(fs_info, group_start);
3986 struct reloc_control *rc;
3987 struct inode *inode;
3988 struct btrfs_path *path;
3989 int ret;
3990 int rw = 0;
3991 int err = 0;
3992
3993 /*
3994 * This only gets set if we had a half-deleted snapshot on mount. We
3995 * cannot allow relocation to start while we're still trying to clean up
3996 * these pending deletions.
3997 */
3998 ret = wait_on_bit(&fs_info->flags, BTRFS_FS_UNFINISHED_DROPS, TASK_INTERRUPTIBLE);
3999 if (ret)
4000 return ret;
4001
4002 /* We may have been woken up by close_ctree, so bail if we're closing. */
4003 if (btrfs_fs_closing(fs_info))
4004 return -EINTR;
4005
4006 bg = btrfs_lookup_block_group(fs_info, group_start);
4007 if (!bg)
4008 return -ENOENT;
4009
4010 /*
4011 * Relocation of a data block group creates ordered extents. Without
4012 * sb_start_write(), we can freeze the filesystem while unfinished
4013 * ordered extents are left. Such ordered extents can cause a deadlock
4014 * e.g. when syncfs() is waiting for their completion but they can't
4015 * finish because they block when joining a transaction, due to the
4016 * fact that the freeze locks are being held in write mode.
4017 */
4018 if (bg->flags & BTRFS_BLOCK_GROUP_DATA)
4019 ASSERT(sb_write_started(fs_info->sb));
4020
4021 if (btrfs_pinned_by_swapfile(fs_info, bg)) {
4022 btrfs_put_block_group(bg);
4023 return -ETXTBSY;
4024 }
4025
4026 rc = alloc_reloc_control(fs_info);
4027 if (!rc) {
4028 btrfs_put_block_group(bg);
4029 return -ENOMEM;
4030 }
4031
4032 ret = reloc_chunk_start(fs_info);
4033 if (ret < 0) {
4034 err = ret;
4035 goto out_put_bg;
4036 }
4037
4038 rc->extent_root = extent_root;
4039 rc->block_group = bg;
4040
4041 ret = btrfs_inc_block_group_ro(rc->block_group, true);
4042 if (ret) {
4043 err = ret;
4044 goto out;
4045 }
4046 rw = 1;
4047
4048 path = btrfs_alloc_path();
4049 if (!path) {
4050 err = -ENOMEM;
4051 goto out;
4052 }
4053
4054 inode = lookup_free_space_inode(rc->block_group, path);
4055 btrfs_free_path(path);
4056
4057 if (!IS_ERR(inode))
4058 ret = delete_block_group_cache(fs_info, rc->block_group, inode, 0);
4059 else
4060 ret = PTR_ERR(inode);
4061
4062 if (ret && ret != -ENOENT) {
4063 err = ret;
4064 goto out;
4065 }
4066
4067 rc->data_inode = create_reloc_inode(fs_info, rc->block_group);
4068 if (IS_ERR(rc->data_inode)) {
4069 err = PTR_ERR(rc->data_inode);
4070 rc->data_inode = NULL;
4071 goto out;
4072 }
4073
4074 describe_relocation(rc->block_group);
4075
4076 btrfs_wait_block_group_reservations(rc->block_group);
4077 btrfs_wait_nocow_writers(rc->block_group);
4078 btrfs_wait_ordered_roots(fs_info, U64_MAX, rc->block_group);
4079
4080 ret = btrfs_zone_finish(rc->block_group);
4081 WARN_ON(ret && ret != -EAGAIN);
4082
4083 while (1) {
4084 enum reloc_stage finishes_stage;
4085
4086 mutex_lock(&fs_info->cleaner_mutex);
4087 ret = relocate_block_group(rc);
4088 mutex_unlock(&fs_info->cleaner_mutex);
4089 if (ret < 0)
4090 err = ret;
4091
4092 finishes_stage = rc->stage;
4093 /*
4094 * We may have gotten ENOSPC after we already dirtied some
4095 * extents. If writeout happens while we're relocating a
4096 * different block group we could end up hitting the
4097 * BUG_ON(rc->stage == UPDATE_DATA_PTRS) in
4098 * btrfs_reloc_cow_block. Make sure we write everything out
4099 * properly so we don't trip over this problem, and then break
4100 * out of the loop if we hit an error.
4101 */
4102 if (rc->stage == MOVE_DATA_EXTENTS && rc->found_file_extent) {
4103 ret = btrfs_wait_ordered_range(BTRFS_I(rc->data_inode), 0,
4104 (u64)-1);
4105 if (ret)
4106 err = ret;
4107 invalidate_mapping_pages(rc->data_inode->i_mapping,
4108 0, -1);
4109 rc->stage = UPDATE_DATA_PTRS;
4110 }
4111
4112 if (err < 0)
4113 goto out;
4114
4115 if (rc->extents_found == 0)
4116 break;
4117
4118 btrfs_info(fs_info, "found %llu extents, stage: %s",
4119 rc->extents_found, stage_to_string(finishes_stage));
4120 }
4121
4122 WARN_ON(rc->block_group->pinned > 0);
4123 WARN_ON(rc->block_group->reserved > 0);
4124 WARN_ON(rc->block_group->used > 0);
4125out:
4126 if (err && rw)
4127 btrfs_dec_block_group_ro(rc->block_group);
4128 iput(rc->data_inode);
4129out_put_bg:
4130 btrfs_put_block_group(bg);
4131 reloc_chunk_end(fs_info);
4132 free_reloc_control(rc);
4133 return err;
4134}
4135
4136static noinline_for_stack int mark_garbage_root(struct btrfs_root *root)
4137{
4138 struct btrfs_fs_info *fs_info = root->fs_info;
4139 struct btrfs_trans_handle *trans;
4140 int ret, err;
4141
4142 trans = btrfs_start_transaction(fs_info->tree_root, 0);
4143 if (IS_ERR(trans))
4144 return PTR_ERR(trans);
4145
4146 memset(&root->root_item.drop_progress, 0,
4147 sizeof(root->root_item.drop_progress));
4148 btrfs_set_root_drop_level(&root->root_item, 0);
4149 btrfs_set_root_refs(&root->root_item, 0);
4150 ret = btrfs_update_root(trans, fs_info->tree_root,
4151 &root->root_key, &root->root_item);
4152
4153 err = btrfs_end_transaction(trans);
4154 if (err)
4155 return err;
4156 return ret;
4157}
4158
4159/*
4160 * recover relocation interrupted by system crash.
4161 *
4162 * this function resumes merging reloc trees with corresponding fs trees.
4163 * this is important for keeping the sharing of tree blocks
4164 */
4165int btrfs_recover_relocation(struct btrfs_fs_info *fs_info)
4166{
4167 LIST_HEAD(reloc_roots);
4168 struct btrfs_key key;
4169 struct btrfs_root *fs_root;
4170 struct btrfs_root *reloc_root;
4171 struct btrfs_path *path;
4172 struct extent_buffer *leaf;
4173 struct reloc_control *rc = NULL;
4174 struct btrfs_trans_handle *trans;
4175 int ret2;
4176 int ret = 0;
4177
4178 path = btrfs_alloc_path();
4179 if (!path)
4180 return -ENOMEM;
4181 path->reada = READA_BACK;
4182
4183 key.objectid = BTRFS_TREE_RELOC_OBJECTID;
4184 key.type = BTRFS_ROOT_ITEM_KEY;
4185 key.offset = (u64)-1;
4186
4187 while (1) {
4188 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key,
4189 path, 0, 0);
4190 if (ret < 0)
4191 goto out;
4192 if (ret > 0) {
4193 if (path->slots[0] == 0)
4194 break;
4195 path->slots[0]--;
4196 }
4197 ret = 0;
4198 leaf = path->nodes[0];
4199 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4200 btrfs_release_path(path);
4201
4202 if (key.objectid != BTRFS_TREE_RELOC_OBJECTID ||
4203 key.type != BTRFS_ROOT_ITEM_KEY)
4204 break;
4205
4206 reloc_root = btrfs_read_tree_root(fs_info->tree_root, &key);
4207 if (IS_ERR(reloc_root)) {
4208 ret = PTR_ERR(reloc_root);
4209 goto out;
4210 }
4211
4212 set_bit(BTRFS_ROOT_SHAREABLE, &reloc_root->state);
4213 list_add(&reloc_root->root_list, &reloc_roots);
4214
4215 if (btrfs_root_refs(&reloc_root->root_item) > 0) {
4216 fs_root = btrfs_get_fs_root(fs_info,
4217 reloc_root->root_key.offset, false);
4218 if (IS_ERR(fs_root)) {
4219 ret = PTR_ERR(fs_root);
4220 if (ret != -ENOENT)
4221 goto out;
4222 ret = mark_garbage_root(reloc_root);
4223 if (ret < 0)
4224 goto out;
4225 ret = 0;
4226 } else {
4227 btrfs_put_root(fs_root);
4228 }
4229 }
4230
4231 if (key.offset == 0)
4232 break;
4233
4234 key.offset--;
4235 }
4236 btrfs_release_path(path);
4237
4238 if (list_empty(&reloc_roots))
4239 goto out;
4240
4241 rc = alloc_reloc_control(fs_info);
4242 if (!rc) {
4243 ret = -ENOMEM;
4244 goto out;
4245 }
4246
4247 ret = reloc_chunk_start(fs_info);
4248 if (ret < 0)
4249 goto out_end;
4250
4251 rc->extent_root = btrfs_extent_root(fs_info, 0);
4252
4253 set_reloc_control(rc);
4254
4255 trans = btrfs_join_transaction(rc->extent_root);
4256 if (IS_ERR(trans)) {
4257 ret = PTR_ERR(trans);
4258 goto out_unset;
4259 }
4260
4261 rc->merge_reloc_tree = true;
4262
4263 while (!list_empty(&reloc_roots)) {
4264 reloc_root = list_entry(reloc_roots.next,
4265 struct btrfs_root, root_list);
4266 list_del(&reloc_root->root_list);
4267
4268 if (btrfs_root_refs(&reloc_root->root_item) == 0) {
4269 list_add_tail(&reloc_root->root_list,
4270 &rc->reloc_roots);
4271 continue;
4272 }
4273
4274 fs_root = btrfs_get_fs_root(fs_info, reloc_root->root_key.offset,
4275 false);
4276 if (IS_ERR(fs_root)) {
4277 ret = PTR_ERR(fs_root);
4278 list_add_tail(&reloc_root->root_list, &reloc_roots);
4279 btrfs_end_transaction(trans);
4280 goto out_unset;
4281 }
4282
4283 ret = __add_reloc_root(reloc_root);
4284 ASSERT(ret != -EEXIST);
4285 if (ret) {
4286 list_add_tail(&reloc_root->root_list, &reloc_roots);
4287 btrfs_put_root(fs_root);
4288 btrfs_end_transaction(trans);
4289 goto out_unset;
4290 }
4291 fs_root->reloc_root = btrfs_grab_root(reloc_root);
4292 btrfs_put_root(fs_root);
4293 }
4294
4295 ret = btrfs_commit_transaction(trans);
4296 if (ret)
4297 goto out_unset;
4298
4299 merge_reloc_roots(rc);
4300
4301 unset_reloc_control(rc);
4302
4303 trans = btrfs_join_transaction(rc->extent_root);
4304 if (IS_ERR(trans)) {
4305 ret = PTR_ERR(trans);
4306 goto out_clean;
4307 }
4308 ret = btrfs_commit_transaction(trans);
4309out_clean:
4310 ret2 = clean_dirty_subvols(rc);
4311 if (ret2 < 0 && !ret)
4312 ret = ret2;
4313out_unset:
4314 unset_reloc_control(rc);
4315out_end:
4316 reloc_chunk_end(fs_info);
4317 free_reloc_control(rc);
4318out:
4319 free_reloc_roots(&reloc_roots);
4320
4321 btrfs_free_path(path);
4322
4323 if (ret == 0) {
4324 /* cleanup orphan inode in data relocation tree */
4325 fs_root = btrfs_grab_root(fs_info->data_reloc_root);
4326 ASSERT(fs_root);
4327 ret = btrfs_orphan_cleanup(fs_root);
4328 btrfs_put_root(fs_root);
4329 }
4330 return ret;
4331}
4332
4333/*
4334 * helper to add ordered checksum for data relocation.
4335 *
4336 * cloning checksum properly handles the nodatasum extents.
4337 * it also saves CPU time to re-calculate the checksum.
4338 */
4339int btrfs_reloc_clone_csums(struct btrfs_ordered_extent *ordered)
4340{
4341 struct btrfs_inode *inode = ordered->inode;
4342 struct btrfs_fs_info *fs_info = inode->root->fs_info;
4343 u64 disk_bytenr = ordered->file_offset + inode->reloc_block_group_start;
4344 struct btrfs_root *csum_root = btrfs_csum_root(fs_info, disk_bytenr);
4345 LIST_HEAD(list);
4346 int ret;
4347
4348 ret = btrfs_lookup_csums_list(csum_root, disk_bytenr,
4349 disk_bytenr + ordered->num_bytes - 1,
4350 &list, false);
4351 if (ret < 0) {
4352 btrfs_mark_ordered_extent_error(ordered);
4353 return ret;
4354 }
4355
4356 while (!list_empty(&list)) {
4357 struct btrfs_ordered_sum *sums =
4358 list_entry(list.next, struct btrfs_ordered_sum, list);
4359
4360 list_del_init(&sums->list);
4361
4362 /*
4363 * We need to offset the new_bytenr based on where the csum is.
4364 * We need to do this because we will read in entire prealloc
4365 * extents but we may have written to say the middle of the
4366 * prealloc extent, so we need to make sure the csum goes with
4367 * the right disk offset.
4368 *
4369 * We can do this because the data reloc inode refers strictly
4370 * to the on disk bytes, so we don't have to worry about
4371 * disk_len vs real len like with real inodes since it's all
4372 * disk length.
4373 */
4374 sums->logical = ordered->disk_bytenr + sums->logical - disk_bytenr;
4375 btrfs_add_ordered_sum(ordered, sums);
4376 }
4377
4378 return 0;
4379}
4380
4381int btrfs_reloc_cow_block(struct btrfs_trans_handle *trans,
4382 struct btrfs_root *root,
4383 const struct extent_buffer *buf,
4384 struct extent_buffer *cow)
4385{
4386 struct btrfs_fs_info *fs_info = root->fs_info;
4387 struct reloc_control *rc;
4388 struct btrfs_backref_node *node;
4389 int first_cow = 0;
4390 int level;
4391 int ret = 0;
4392
4393 rc = fs_info->reloc_ctl;
4394 if (!rc)
4395 return 0;
4396
4397 BUG_ON(rc->stage == UPDATE_DATA_PTRS && btrfs_is_data_reloc_root(root));
4398
4399 level = btrfs_header_level(buf);
4400 if (btrfs_header_generation(buf) <=
4401 btrfs_root_last_snapshot(&root->root_item))
4402 first_cow = 1;
4403
4404 if (btrfs_root_id(root) == BTRFS_TREE_RELOC_OBJECTID && rc->create_reloc_tree) {
4405 WARN_ON(!first_cow && level == 0);
4406
4407 node = rc->backref_cache.path[level];
4408
4409 /*
4410 * If node->bytenr != buf->start and node->new_bytenr !=
4411 * buf->start then we've got the wrong backref node for what we
4412 * expected to see here and the cache is incorrect.
4413 */
4414 if (unlikely(node->bytenr != buf->start && node->new_bytenr != buf->start)) {
4415 btrfs_err(fs_info,
4416"bytenr %llu was found but our backref cache was expecting %llu or %llu",
4417 buf->start, node->bytenr, node->new_bytenr);
4418 return -EUCLEAN;
4419 }
4420
4421 btrfs_backref_drop_node_buffer(node);
4422 atomic_inc(&cow->refs);
4423 node->eb = cow;
4424 node->new_bytenr = cow->start;
4425
4426 if (!node->pending) {
4427 list_move_tail(&node->list,
4428 &rc->backref_cache.pending[level]);
4429 node->pending = 1;
4430 }
4431
4432 if (first_cow)
4433 mark_block_processed(rc, node);
4434
4435 if (first_cow && level > 0)
4436 rc->nodes_relocated += buf->len;
4437 }
4438
4439 if (level == 0 && first_cow && rc->stage == UPDATE_DATA_PTRS)
4440 ret = replace_file_extents(trans, rc, root, cow);
4441 return ret;
4442}
4443
4444/*
4445 * called before creating snapshot. it calculates metadata reservation
4446 * required for relocating tree blocks in the snapshot
4447 */
4448void btrfs_reloc_pre_snapshot(struct btrfs_pending_snapshot *pending,
4449 u64 *bytes_to_reserve)
4450{
4451 struct btrfs_root *root = pending->root;
4452 struct reloc_control *rc = root->fs_info->reloc_ctl;
4453
4454 if (!rc || !have_reloc_root(root))
4455 return;
4456
4457 if (!rc->merge_reloc_tree)
4458 return;
4459
4460 root = root->reloc_root;
4461 BUG_ON(btrfs_root_refs(&root->root_item) == 0);
4462 /*
4463 * relocation is in the stage of merging trees. the space
4464 * used by merging a reloc tree is twice the size of
4465 * relocated tree nodes in the worst case. half for cowing
4466 * the reloc tree, half for cowing the fs tree. the space
4467 * used by cowing the reloc tree will be freed after the
4468 * tree is dropped. if we create snapshot, cowing the fs
4469 * tree may use more space than it frees. so we need
4470 * reserve extra space.
4471 */
4472 *bytes_to_reserve += rc->nodes_relocated;
4473}
4474
4475/*
4476 * called after snapshot is created. migrate block reservation
4477 * and create reloc root for the newly created snapshot
4478 *
4479 * This is similar to btrfs_init_reloc_root(), we come out of here with two
4480 * references held on the reloc_root, one for root->reloc_root and one for
4481 * rc->reloc_roots.
4482 */
4483int btrfs_reloc_post_snapshot(struct btrfs_trans_handle *trans,
4484 struct btrfs_pending_snapshot *pending)
4485{
4486 struct btrfs_root *root = pending->root;
4487 struct btrfs_root *reloc_root;
4488 struct btrfs_root *new_root;
4489 struct reloc_control *rc = root->fs_info->reloc_ctl;
4490 int ret;
4491
4492 if (!rc || !have_reloc_root(root))
4493 return 0;
4494
4495 rc = root->fs_info->reloc_ctl;
4496 rc->merging_rsv_size += rc->nodes_relocated;
4497
4498 if (rc->merge_reloc_tree) {
4499 ret = btrfs_block_rsv_migrate(&pending->block_rsv,
4500 rc->block_rsv,
4501 rc->nodes_relocated, true);
4502 if (ret)
4503 return ret;
4504 }
4505
4506 new_root = pending->snap;
4507 reloc_root = create_reloc_root(trans, root->reloc_root, btrfs_root_id(new_root));
4508 if (IS_ERR(reloc_root))
4509 return PTR_ERR(reloc_root);
4510
4511 ret = __add_reloc_root(reloc_root);
4512 ASSERT(ret != -EEXIST);
4513 if (ret) {
4514 /* Pairs with create_reloc_root */
4515 btrfs_put_root(reloc_root);
4516 return ret;
4517 }
4518 new_root->reloc_root = btrfs_grab_root(reloc_root);
4519
4520 if (rc->create_reloc_tree)
4521 ret = clone_backref_node(trans, rc, root, reloc_root);
4522 return ret;
4523}
4524
4525/*
4526 * Get the current bytenr for the block group which is being relocated.
4527 *
4528 * Return U64_MAX if no running relocation.
4529 */
4530u64 btrfs_get_reloc_bg_bytenr(const struct btrfs_fs_info *fs_info)
4531{
4532 u64 logical = U64_MAX;
4533
4534 lockdep_assert_held(&fs_info->reloc_mutex);
4535
4536 if (fs_info->reloc_ctl && fs_info->reloc_ctl->block_group)
4537 logical = fs_info->reloc_ctl->block_group->start;
4538 return logical;
4539}