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