1// SPDX-License-Identifier: GPL-2.0
2
3#include "messages.h"
4#include "tree-mod-log.h"
5#include "disk-io.h"
6#include "fs.h"
7#include "accessors.h"
8#include "tree-checker.h"
9
10struct tree_mod_root {
11	u64 logical;
12	u8 level;
13};
14
15struct tree_mod_elem {
16	struct rb_node node;
17	u64 logical;
18	u64 seq;
19	enum btrfs_mod_log_op op;
20
21	/*
22	 * This is used for BTRFS_MOD_LOG_KEY_* and BTRFS_MOD_LOG_MOVE_KEYS
23	 * operations.
24	 */
25	int slot;
26
27	/* This is used for BTRFS_MOD_LOG_KEY* and BTRFS_MOD_LOG_ROOT_REPLACE. */
28	u64 generation;
29
30	/* Those are used for op == BTRFS_MOD_LOG_KEY_{REPLACE,REMOVE}. */
31	struct btrfs_disk_key key;
32	u64 blockptr;
33
34	/* This is used for op == BTRFS_MOD_LOG_MOVE_KEYS. */
35	struct {
36		int dst_slot;
37		int nr_items;
38	} move;
39
40	/* This is used for op == BTRFS_MOD_LOG_ROOT_REPLACE. */
41	struct tree_mod_root old_root;
42};
43
44/*
45 * Pull a new tree mod seq number for our operation.
46 */
47static u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info)
48{
49	return atomic64_inc_return(&fs_info->tree_mod_seq);
50}
51
52/*
53 * This adds a new blocker to the tree mod log's blocker list if the @elem
54 * passed does not already have a sequence number set. So when a caller expects
55 * to record tree modifications, it should ensure to set elem->seq to zero
56 * before calling btrfs_get_tree_mod_seq.
57 * Returns a fresh, unused tree log modification sequence number, even if no new
58 * blocker was added.
59 */
60u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
61			   struct btrfs_seq_list *elem)
62{
63	write_lock(&fs_info->tree_mod_log_lock);
64	if (!elem->seq) {
65		elem->seq = btrfs_inc_tree_mod_seq(fs_info);
66		list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
67		set_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags);
68	}
69	write_unlock(&fs_info->tree_mod_log_lock);
70
71	return elem->seq;
72}
73
74void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
75			    struct btrfs_seq_list *elem)
76{
77	struct rb_root *tm_root;
78	struct rb_node *node;
79	struct rb_node *next;
80	struct tree_mod_elem *tm;
81	u64 min_seq = BTRFS_SEQ_LAST;
82	u64 seq_putting = elem->seq;
83
84	if (!seq_putting)
85		return;
86
87	write_lock(&fs_info->tree_mod_log_lock);
88	list_del(&elem->list);
89	elem->seq = 0;
90
91	if (list_empty(&fs_info->tree_mod_seq_list)) {
92		clear_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags);
93	} else {
94		struct btrfs_seq_list *first;
95
96		first = list_first_entry(&fs_info->tree_mod_seq_list,
97					 struct btrfs_seq_list, list);
98		if (seq_putting > first->seq) {
99			/*
100			 * Blocker with lower sequence number exists, we cannot
101			 * remove anything from the log.
102			 */
103			write_unlock(&fs_info->tree_mod_log_lock);
104			return;
105		}
106		min_seq = first->seq;
107	}
108
109	/*
110	 * Anything that's lower than the lowest existing (read: blocked)
111	 * sequence number can be removed from the tree.
112	 */
113	tm_root = &fs_info->tree_mod_log;
114	for (node = rb_first(tm_root); node; node = next) {
115		next = rb_next(node);
116		tm = rb_entry(node, struct tree_mod_elem, node);
117		if (tm->seq >= min_seq)
118			continue;
119		rb_erase(node, tm_root);
120		kfree(tm);
121	}
122	write_unlock(&fs_info->tree_mod_log_lock);
123}
124
125/*
126 * Key order of the log:
127 *       node/leaf start address -> sequence
128 *
129 * The 'start address' is the logical address of the *new* root node for root
130 * replace operations, or the logical address of the affected block for all
131 * other operations.
132 */
133static noinline int tree_mod_log_insert(struct btrfs_fs_info *fs_info,
134					struct tree_mod_elem *tm)
135{
136	struct rb_root *tm_root;
137	struct rb_node **new;
138	struct rb_node *parent = NULL;
139	struct tree_mod_elem *cur;
140
141	lockdep_assert_held_write(&fs_info->tree_mod_log_lock);
142
143	tm->seq = btrfs_inc_tree_mod_seq(fs_info);
144
145	tm_root = &fs_info->tree_mod_log;
146	new = &tm_root->rb_node;
147	while (*new) {
148		cur = rb_entry(*new, struct tree_mod_elem, node);
149		parent = *new;
150		if (cur->logical < tm->logical)
151			new = &((*new)->rb_left);
152		else if (cur->logical > tm->logical)
153			new = &((*new)->rb_right);
154		else if (cur->seq < tm->seq)
155			new = &((*new)->rb_left);
156		else if (cur->seq > tm->seq)
157			new = &((*new)->rb_right);
158		else
159			return -EEXIST;
160	}
161
162	rb_link_node(&tm->node, parent, new);
163	rb_insert_color(&tm->node, tm_root);
164	return 0;
165}
166
167/*
168 * Determines if logging can be omitted. Returns true if it can. Otherwise, it
169 * returns false with the tree_mod_log_lock acquired. The caller must hold
170 * this until all tree mod log insertions are recorded in the rb tree and then
171 * write unlock fs_info::tree_mod_log_lock.
172 */
173static bool tree_mod_dont_log(struct btrfs_fs_info *fs_info, struct extent_buffer *eb)
174{
175	if (!test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags))
176		return true;
177	if (eb && btrfs_header_level(eb) == 0)
178		return true;
179
180	write_lock(&fs_info->tree_mod_log_lock);
181	if (list_empty(&(fs_info)->tree_mod_seq_list)) {
182		write_unlock(&fs_info->tree_mod_log_lock);
183		return true;
184	}
185
186	return false;
187}
188
189/* Similar to tree_mod_dont_log, but doesn't acquire any locks. */
190static bool tree_mod_need_log(const struct btrfs_fs_info *fs_info,
191				    struct extent_buffer *eb)
192{
193	if (!test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags))
194		return false;
195	if (eb && btrfs_header_level(eb) == 0)
196		return false;
197
198	return true;
199}
200
201static struct tree_mod_elem *alloc_tree_mod_elem(struct extent_buffer *eb,
202						 int slot,
203						 enum btrfs_mod_log_op op)
204{
205	struct tree_mod_elem *tm;
206
207	tm = kzalloc(sizeof(*tm), GFP_NOFS);
208	if (!tm)
209		return NULL;
210
211	tm->logical = eb->start;
212	if (op != BTRFS_MOD_LOG_KEY_ADD) {
213		btrfs_node_key(eb, &tm->key, slot);
214		tm->blockptr = btrfs_node_blockptr(eb, slot);
215	}
216	tm->op = op;
217	tm->slot = slot;
218	tm->generation = btrfs_node_ptr_generation(eb, slot);
219	RB_CLEAR_NODE(&tm->node);
220
221	return tm;
222}
223
224int btrfs_tree_mod_log_insert_key(struct extent_buffer *eb, int slot,
225				  enum btrfs_mod_log_op op)
226{
227	struct tree_mod_elem *tm;
228	int ret = 0;
229
230	if (!tree_mod_need_log(eb->fs_info, eb))
231		return 0;
232
233	tm = alloc_tree_mod_elem(eb, slot, op);
234	if (!tm)
235		ret = -ENOMEM;
236
237	if (tree_mod_dont_log(eb->fs_info, eb)) {
238		kfree(tm);
239		/*
240		 * Don't error if we failed to allocate memory because we don't
241		 * need to log.
242		 */
243		return 0;
244	} else if (ret != 0) {
245		/*
246		 * We previously failed to allocate memory and we need to log,
247		 * so we have to fail.
248		 */
249		goto out_unlock;
250	}
251
252	ret = tree_mod_log_insert(eb->fs_info, tm);
253out_unlock:
254	write_unlock(&eb->fs_info->tree_mod_log_lock);
255	if (ret)
256		kfree(tm);
257
258	return ret;
259}
260
261static struct tree_mod_elem *tree_mod_log_alloc_move(struct extent_buffer *eb,
262						     int dst_slot, int src_slot,
263						     int nr_items)
264{
265	struct tree_mod_elem *tm;
266
267	tm = kzalloc(sizeof(*tm), GFP_NOFS);
268	if (!tm)
269		return ERR_PTR(-ENOMEM);
270
271	tm->logical = eb->start;
272	tm->slot = src_slot;
273	tm->move.dst_slot = dst_slot;
274	tm->move.nr_items = nr_items;
275	tm->op = BTRFS_MOD_LOG_MOVE_KEYS;
276	RB_CLEAR_NODE(&tm->node);
277
278	return tm;
279}
280
281int btrfs_tree_mod_log_insert_move(struct extent_buffer *eb,
282				   int dst_slot, int src_slot,
283				   int nr_items)
284{
285	struct tree_mod_elem *tm = NULL;
286	struct tree_mod_elem **tm_list = NULL;
287	int ret = 0;
288	int i;
289	bool locked = false;
290
291	if (!tree_mod_need_log(eb->fs_info, eb))
292		return 0;
293
294	tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS);
295	if (!tm_list) {
296		ret = -ENOMEM;
297		goto lock;
298	}
299
300	tm = tree_mod_log_alloc_move(eb, dst_slot, src_slot, nr_items);
301	if (IS_ERR(tm)) {
302		ret = PTR_ERR(tm);
303		tm = NULL;
304		goto lock;
305	}
306
307	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
308		tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot,
309				BTRFS_MOD_LOG_KEY_REMOVE_WHILE_MOVING);
310		if (!tm_list[i]) {
311			ret = -ENOMEM;
312			goto lock;
313		}
314	}
315
316lock:
317	if (tree_mod_dont_log(eb->fs_info, eb)) {
318		/*
319		 * Don't error if we failed to allocate memory because we don't
320		 * need to log.
321		 */
322		ret = 0;
323		goto free_tms;
324	}
325	locked = true;
326
327	/*
328	 * We previously failed to allocate memory and we need to log, so we
329	 * have to fail.
330	 */
331	if (ret != 0)
332		goto free_tms;
333
334	/*
335	 * When we override something during the move, we log these removals.
336	 * This can only happen when we move towards the beginning of the
337	 * buffer, i.e. dst_slot < src_slot.
338	 */
339	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
340		ret = tree_mod_log_insert(eb->fs_info, tm_list[i]);
341		if (ret)
342			goto free_tms;
343	}
344
345	ret = tree_mod_log_insert(eb->fs_info, tm);
346	if (ret)
347		goto free_tms;
348	write_unlock(&eb->fs_info->tree_mod_log_lock);
349	kfree(tm_list);
350
351	return 0;
352
353free_tms:
354	if (tm_list) {
355		for (i = 0; i < nr_items; i++) {
356			if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
357				rb_erase(&tm_list[i]->node, &eb->fs_info->tree_mod_log);
358			kfree(tm_list[i]);
359		}
360	}
361	if (locked)
362		write_unlock(&eb->fs_info->tree_mod_log_lock);
363	kfree(tm_list);
364	kfree(tm);
365
366	return ret;
367}
368
369static int tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
370				struct tree_mod_elem **tm_list,
371				int nritems)
372{
373	int i, j;
374	int ret;
375
376	for (i = nritems - 1; i >= 0; i--) {
377		ret = tree_mod_log_insert(fs_info, tm_list[i]);
378		if (ret) {
379			for (j = nritems - 1; j > i; j--)
380				rb_erase(&tm_list[j]->node,
381					 &fs_info->tree_mod_log);
382			return ret;
383		}
384	}
385
386	return 0;
387}
388
389int btrfs_tree_mod_log_insert_root(struct extent_buffer *old_root,
390				   struct extent_buffer *new_root,
391				   bool log_removal)
392{
393	struct btrfs_fs_info *fs_info = old_root->fs_info;
394	struct tree_mod_elem *tm = NULL;
395	struct tree_mod_elem **tm_list = NULL;
396	int nritems = 0;
397	int ret = 0;
398	int i;
399
400	if (!tree_mod_need_log(fs_info, NULL))
401		return 0;
402
403	if (log_removal && btrfs_header_level(old_root) > 0) {
404		nritems = btrfs_header_nritems(old_root);
405		tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *),
406				  GFP_NOFS);
407		if (!tm_list) {
408			ret = -ENOMEM;
409			goto lock;
410		}
411		for (i = 0; i < nritems; i++) {
412			tm_list[i] = alloc_tree_mod_elem(old_root, i,
413			    BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING);
414			if (!tm_list[i]) {
415				ret = -ENOMEM;
416				goto lock;
417			}
418		}
419	}
420
421	tm = kzalloc(sizeof(*tm), GFP_NOFS);
422	if (!tm) {
423		ret = -ENOMEM;
424		goto lock;
425	}
426
427	tm->logical = new_root->start;
428	tm->old_root.logical = old_root->start;
429	tm->old_root.level = btrfs_header_level(old_root);
430	tm->generation = btrfs_header_generation(old_root);
431	tm->op = BTRFS_MOD_LOG_ROOT_REPLACE;
432
433lock:
434	if (tree_mod_dont_log(fs_info, NULL)) {
435		/*
436		 * Don't error if we failed to allocate memory because we don't
437		 * need to log.
438		 */
439		ret = 0;
440		goto free_tms;
441	} else if (ret != 0) {
442		/*
443		 * We previously failed to allocate memory and we need to log,
444		 * so we have to fail.
445		 */
446		goto out_unlock;
447	}
448
449	if (tm_list)
450		ret = tree_mod_log_free_eb(fs_info, tm_list, nritems);
451	if (!ret)
452		ret = tree_mod_log_insert(fs_info, tm);
453
454out_unlock:
455	write_unlock(&fs_info->tree_mod_log_lock);
456	if (ret)
457		goto free_tms;
458	kfree(tm_list);
459
460	return ret;
461
462free_tms:
463	if (tm_list) {
464		for (i = 0; i < nritems; i++)
465			kfree(tm_list[i]);
466		kfree(tm_list);
467	}
468	kfree(tm);
469
470	return ret;
471}
472
473static struct tree_mod_elem *__tree_mod_log_search(struct btrfs_fs_info *fs_info,
474						   u64 start, u64 min_seq,
475						   bool smallest)
476{
477	struct rb_root *tm_root;
478	struct rb_node *node;
479	struct tree_mod_elem *cur = NULL;
480	struct tree_mod_elem *found = NULL;
481
482	read_lock(&fs_info->tree_mod_log_lock);
483	tm_root = &fs_info->tree_mod_log;
484	node = tm_root->rb_node;
485	while (node) {
486		cur = rb_entry(node, struct tree_mod_elem, node);
487		if (cur->logical < start) {
488			node = node->rb_left;
489		} else if (cur->logical > start) {
490			node = node->rb_right;
491		} else if (cur->seq < min_seq) {
492			node = node->rb_left;
493		} else if (!smallest) {
494			/* We want the node with the highest seq */
495			if (found)
496				BUG_ON(found->seq > cur->seq);
497			found = cur;
498			node = node->rb_left;
499		} else if (cur->seq > min_seq) {
500			/* We want the node with the smallest seq */
501			if (found)
502				BUG_ON(found->seq < cur->seq);
503			found = cur;
504			node = node->rb_right;
505		} else {
506			found = cur;
507			break;
508		}
509	}
510	read_unlock(&fs_info->tree_mod_log_lock);
511
512	return found;
513}
514
515/*
516 * This returns the element from the log with the smallest time sequence
517 * value that's in the log (the oldest log item). Any element with a time
518 * sequence lower than min_seq will be ignored.
519 */
520static struct tree_mod_elem *tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info,
521							u64 start, u64 min_seq)
522{
523	return __tree_mod_log_search(fs_info, start, min_seq, true);
524}
525
526/*
527 * This returns the element from the log with the largest time sequence
528 * value that's in the log (the most recent log item). Any element with
529 * a time sequence lower than min_seq will be ignored.
530 */
531static struct tree_mod_elem *tree_mod_log_search(struct btrfs_fs_info *fs_info,
532						 u64 start, u64 min_seq)
533{
534	return __tree_mod_log_search(fs_info, start, min_seq, false);
535}
536
537int btrfs_tree_mod_log_eb_copy(struct extent_buffer *dst,
538			       struct extent_buffer *src,
539			       unsigned long dst_offset,
540			       unsigned long src_offset,
541			       int nr_items)
542{
543	struct btrfs_fs_info *fs_info = dst->fs_info;
544	int ret = 0;
545	struct tree_mod_elem **tm_list = NULL;
546	struct tree_mod_elem **tm_list_add = NULL;
547	struct tree_mod_elem **tm_list_rem = NULL;
548	int i;
549	bool locked = false;
550	struct tree_mod_elem *dst_move_tm = NULL;
551	struct tree_mod_elem *src_move_tm = NULL;
552	u32 dst_move_nr_items = btrfs_header_nritems(dst) - dst_offset;
553	u32 src_move_nr_items = btrfs_header_nritems(src) - (src_offset + nr_items);
554
555	if (!tree_mod_need_log(fs_info, NULL))
556		return 0;
557
558	if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
559		return 0;
560
561	tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *),
562			  GFP_NOFS);
563	if (!tm_list) {
564		ret = -ENOMEM;
565		goto lock;
566	}
567
568	if (dst_move_nr_items) {
569		dst_move_tm = tree_mod_log_alloc_move(dst, dst_offset + nr_items,
570						      dst_offset, dst_move_nr_items);
571		if (IS_ERR(dst_move_tm)) {
572			ret = PTR_ERR(dst_move_tm);
573			dst_move_tm = NULL;
574			goto lock;
575		}
576	}
577	if (src_move_nr_items) {
578		src_move_tm = tree_mod_log_alloc_move(src, src_offset,
579						      src_offset + nr_items,
580						      src_move_nr_items);
581		if (IS_ERR(src_move_tm)) {
582			ret = PTR_ERR(src_move_tm);
583			src_move_tm = NULL;
584			goto lock;
585		}
586	}
587
588	tm_list_add = tm_list;
589	tm_list_rem = tm_list + nr_items;
590	for (i = 0; i < nr_items; i++) {
591		tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset,
592						     BTRFS_MOD_LOG_KEY_REMOVE);
593		if (!tm_list_rem[i]) {
594			ret = -ENOMEM;
595			goto lock;
596		}
597
598		tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset,
599						     BTRFS_MOD_LOG_KEY_ADD);
600		if (!tm_list_add[i]) {
601			ret = -ENOMEM;
602			goto lock;
603		}
604	}
605
606lock:
607	if (tree_mod_dont_log(fs_info, NULL)) {
608		/*
609		 * Don't error if we failed to allocate memory because we don't
610		 * need to log.
611		 */
612		ret = 0;
613		goto free_tms;
614	}
615	locked = true;
616
617	/*
618	 * We previously failed to allocate memory and we need to log, so we
619	 * have to fail.
620	 */
621	if (ret != 0)
622		goto free_tms;
623
624	if (dst_move_tm) {
625		ret = tree_mod_log_insert(fs_info, dst_move_tm);
626		if (ret)
627			goto free_tms;
628	}
629	for (i = 0; i < nr_items; i++) {
630		ret = tree_mod_log_insert(fs_info, tm_list_rem[i]);
631		if (ret)
632			goto free_tms;
633		ret = tree_mod_log_insert(fs_info, tm_list_add[i]);
634		if (ret)
635			goto free_tms;
636	}
637	if (src_move_tm) {
638		ret = tree_mod_log_insert(fs_info, src_move_tm);
639		if (ret)
640			goto free_tms;
641	}
642
643	write_unlock(&fs_info->tree_mod_log_lock);
644	kfree(tm_list);
645
646	return 0;
647
648free_tms:
649	if (dst_move_tm && !RB_EMPTY_NODE(&dst_move_tm->node))
650		rb_erase(&dst_move_tm->node, &fs_info->tree_mod_log);
651	kfree(dst_move_tm);
652	if (src_move_tm && !RB_EMPTY_NODE(&src_move_tm->node))
653		rb_erase(&src_move_tm->node, &fs_info->tree_mod_log);
654	kfree(src_move_tm);
655	if (tm_list) {
656		for (i = 0; i < nr_items * 2; i++) {
657			if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
658				rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
659			kfree(tm_list[i]);
660		}
661	}
662	if (locked)
663		write_unlock(&fs_info->tree_mod_log_lock);
664	kfree(tm_list);
665
666	return ret;
667}
668
669int btrfs_tree_mod_log_free_eb(struct extent_buffer *eb)
670{
671	struct tree_mod_elem **tm_list = NULL;
672	int nritems = 0;
673	int i;
674	int ret = 0;
675
676	if (!tree_mod_need_log(eb->fs_info, eb))
677		return 0;
678
679	nritems = btrfs_header_nritems(eb);
680	tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS);
681	if (!tm_list) {
682		ret = -ENOMEM;
683		goto lock;
684	}
685
686	for (i = 0; i < nritems; i++) {
687		tm_list[i] = alloc_tree_mod_elem(eb, i,
688				    BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING);
689		if (!tm_list[i]) {
690			ret = -ENOMEM;
691			goto lock;
692		}
693	}
694
695lock:
696	if (tree_mod_dont_log(eb->fs_info, eb)) {
697		/*
698		 * Don't error if we failed to allocate memory because we don't
699		 * need to log.
700		 */
701		ret = 0;
702		goto free_tms;
703	} else if (ret != 0) {
704		/*
705		 * We previously failed to allocate memory and we need to log,
706		 * so we have to fail.
707		 */
708		goto out_unlock;
709	}
710
711	ret = tree_mod_log_free_eb(eb->fs_info, tm_list, nritems);
712out_unlock:
713	write_unlock(&eb->fs_info->tree_mod_log_lock);
714	if (ret)
715		goto free_tms;
716	kfree(tm_list);
717
718	return 0;
719
720free_tms:
721	if (tm_list) {
722		for (i = 0; i < nritems; i++)
723			kfree(tm_list[i]);
724		kfree(tm_list);
725	}
726
727	return ret;
728}
729
730/*
731 * Returns the logical address of the oldest predecessor of the given root.
732 * Entries older than time_seq are ignored.
733 */
734static struct tree_mod_elem *tree_mod_log_oldest_root(struct extent_buffer *eb_root,
735						      u64 time_seq)
736{
737	struct tree_mod_elem *tm;
738	struct tree_mod_elem *found = NULL;
739	u64 root_logical = eb_root->start;
740	bool looped = false;
741
742	if (!time_seq)
743		return NULL;
744
745	/*
746	 * The very last operation that's logged for a root is the replacement
747	 * operation (if it is replaced at all). This has the logical address
748	 * of the *new* root, making it the very first operation that's logged
749	 * for this root.
750	 */
751	while (1) {
752		tm = tree_mod_log_search_oldest(eb_root->fs_info, root_logical,
753						time_seq);
754		if (!looped && !tm)
755			return NULL;
756		/*
757		 * If there are no tree operation for the oldest root, we simply
758		 * return it. This should only happen if that (old) root is at
759		 * level 0.
760		 */
761		if (!tm)
762			break;
763
764		/*
765		 * If there's an operation that's not a root replacement, we
766		 * found the oldest version of our root. Normally, we'll find a
767		 * BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here.
768		 */
769		if (tm->op != BTRFS_MOD_LOG_ROOT_REPLACE)
770			break;
771
772		found = tm;
773		root_logical = tm->old_root.logical;
774		looped = true;
775	}
776
777	/* If there's no old root to return, return what we found instead */
778	if (!found)
779		found = tm;
780
781	return found;
782}
783
784
785/*
786 * tm is a pointer to the first operation to rewind within eb. Then, all
787 * previous operations will be rewound (until we reach something older than
788 * time_seq).
789 */
790static void tree_mod_log_rewind(struct btrfs_fs_info *fs_info,
791				struct extent_buffer *eb,
792				u64 time_seq,
793				struct tree_mod_elem *first_tm)
794{
795	u32 n;
796	struct rb_node *next;
797	struct tree_mod_elem *tm = first_tm;
798	unsigned long o_dst;
799	unsigned long o_src;
800	unsigned long p_size = sizeof(struct btrfs_key_ptr);
801	/*
802	 * max_slot tracks the maximum valid slot of the rewind eb at every
803	 * step of the rewind. This is in contrast with 'n' which eventually
804	 * matches the number of items, but can be wrong during moves or if
805	 * removes overlap on already valid slots (which is probably separately
806	 * a bug). We do this to validate the offsets of memmoves for rewinding
807	 * moves and detect invalid memmoves.
808	 *
809	 * Since a rewind eb can start empty, max_slot is a signed integer with
810	 * a special meaning for -1, which is that no slot is valid to move out
811	 * of. Any other negative value is invalid.
812	 */
813	int max_slot;
814	int move_src_end_slot;
815	int move_dst_end_slot;
816
817	n = btrfs_header_nritems(eb);
818	max_slot = n - 1;
819	read_lock(&fs_info->tree_mod_log_lock);
820	while (tm && tm->seq >= time_seq) {
821		ASSERT(max_slot >= -1);
822		/*
823		 * All the operations are recorded with the operator used for
824		 * the modification. As we're going backwards, we do the
825		 * opposite of each operation here.
826		 */
827		switch (tm->op) {
828		case BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING:
829			BUG_ON(tm->slot < n);
830			fallthrough;
831		case BTRFS_MOD_LOG_KEY_REMOVE_WHILE_MOVING:
832		case BTRFS_MOD_LOG_KEY_REMOVE:
833			btrfs_set_node_key(eb, &tm->key, tm->slot);
834			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
835			btrfs_set_node_ptr_generation(eb, tm->slot,
836						      tm->generation);
837			n++;
838			if (tm->slot > max_slot)
839				max_slot = tm->slot;
840			break;
841		case BTRFS_MOD_LOG_KEY_REPLACE:
842			BUG_ON(tm->slot >= n);
843			btrfs_set_node_key(eb, &tm->key, tm->slot);
844			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
845			btrfs_set_node_ptr_generation(eb, tm->slot,
846						      tm->generation);
847			break;
848		case BTRFS_MOD_LOG_KEY_ADD:
849			/*
850			 * It is possible we could have already removed keys
851			 * behind the known max slot, so this will be an
852			 * overestimate. In practice, the copy operation
853			 * inserts them in increasing order, and overestimating
854			 * just means we miss some warnings, so it's OK. It
855			 * isn't worth carefully tracking the full array of
856			 * valid slots to check against when moving.
857			 */
858			if (tm->slot == max_slot)
859				max_slot--;
860			/* if a move operation is needed it's in the log */
861			n--;
862			break;
863		case BTRFS_MOD_LOG_MOVE_KEYS:
864			ASSERT(tm->move.nr_items > 0);
865			move_src_end_slot = tm->move.dst_slot + tm->move.nr_items - 1;
866			move_dst_end_slot = tm->slot + tm->move.nr_items - 1;
867			o_dst = btrfs_node_key_ptr_offset(eb, tm->slot);
868			o_src = btrfs_node_key_ptr_offset(eb, tm->move.dst_slot);
869			if (WARN_ON(move_src_end_slot > max_slot ||
870				    tm->move.nr_items <= 0)) {
871				btrfs_warn(fs_info,
872"move from invalid tree mod log slot eb %llu slot %d dst_slot %d nr_items %d seq %llu n %u max_slot %d",
873					   eb->start, tm->slot,
874					   tm->move.dst_slot, tm->move.nr_items,
875					   tm->seq, n, max_slot);
876			}
877			memmove_extent_buffer(eb, o_dst, o_src,
878					      tm->move.nr_items * p_size);
879			max_slot = move_dst_end_slot;
880			break;
881		case BTRFS_MOD_LOG_ROOT_REPLACE:
882			/*
883			 * This operation is special. For roots, this must be
884			 * handled explicitly before rewinding.
885			 * For non-roots, this operation may exist if the node
886			 * was a root: root A -> child B; then A gets empty and
887			 * B is promoted to the new root. In the mod log, we'll
888			 * have a root-replace operation for B, a tree block
889			 * that is no root. We simply ignore that operation.
890			 */
891			break;
892		}
893		next = rb_next(&tm->node);
894		if (!next)
895			break;
896		tm = rb_entry(next, struct tree_mod_elem, node);
897		if (tm->logical != first_tm->logical)
898			break;
899	}
900	read_unlock(&fs_info->tree_mod_log_lock);
901	btrfs_set_header_nritems(eb, n);
902}
903
904/*
905 * Called with eb read locked. If the buffer cannot be rewound, the same buffer
906 * is returned. If rewind operations happen, a fresh buffer is returned. The
907 * returned buffer is always read-locked. If the returned buffer is not the
908 * input buffer, the lock on the input buffer is released and the input buffer
909 * is freed (its refcount is decremented).
910 */
911struct extent_buffer *btrfs_tree_mod_log_rewind(struct btrfs_fs_info *fs_info,
912						struct btrfs_path *path,
913						struct extent_buffer *eb,
914						u64 time_seq)
915{
916	struct extent_buffer *eb_rewin;
917	struct tree_mod_elem *tm;
918
919	if (!time_seq)
920		return eb;
921
922	if (btrfs_header_level(eb) == 0)
923		return eb;
924
925	tm = tree_mod_log_search(fs_info, eb->start, time_seq);
926	if (!tm)
927		return eb;
928
929	if (tm->op == BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
930		BUG_ON(tm->slot != 0);
931		eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start);
932		if (!eb_rewin) {
933			btrfs_tree_read_unlock(eb);
934			free_extent_buffer(eb);
935			return NULL;
936		}
937		btrfs_set_header_bytenr(eb_rewin, eb->start);
938		btrfs_set_header_backref_rev(eb_rewin,
939					     btrfs_header_backref_rev(eb));
940		btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb));
941		btrfs_set_header_level(eb_rewin, btrfs_header_level(eb));
942	} else {
943		eb_rewin = btrfs_clone_extent_buffer(eb);
944		if (!eb_rewin) {
945			btrfs_tree_read_unlock(eb);
946			free_extent_buffer(eb);
947			return NULL;
948		}
949	}
950
951	btrfs_tree_read_unlock(eb);
952	free_extent_buffer(eb);
953
954	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb_rewin),
955				       eb_rewin, btrfs_header_level(eb_rewin));
956	btrfs_tree_read_lock(eb_rewin);
957	tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm);
958	WARN_ON(btrfs_header_nritems(eb_rewin) >
959		BTRFS_NODEPTRS_PER_BLOCK(fs_info));
960
961	return eb_rewin;
962}
963
964/*
965 * Rewind the state of @root's root node to the given @time_seq value.
966 * If there are no changes, the current root->root_node is returned. If anything
967 * changed in between, there's a fresh buffer allocated on which the rewind
968 * operations are done. In any case, the returned buffer is read locked.
969 * Returns NULL on error (with no locks held).
970 */
971struct extent_buffer *btrfs_get_old_root(struct btrfs_root *root, u64 time_seq)
972{
973	struct btrfs_fs_info *fs_info = root->fs_info;
974	struct tree_mod_elem *tm;
975	struct extent_buffer *eb = NULL;
976	struct extent_buffer *eb_root;
977	u64 eb_root_owner = 0;
978	struct extent_buffer *old;
979	struct tree_mod_root *old_root = NULL;
980	u64 old_generation = 0;
981	u64 logical;
982	int level;
983
984	eb_root = btrfs_read_lock_root_node(root);
985	tm = tree_mod_log_oldest_root(eb_root, time_seq);
986	if (!tm)
987		return eb_root;
988
989	if (tm->op == BTRFS_MOD_LOG_ROOT_REPLACE) {
990		old_root = &tm->old_root;
991		old_generation = tm->generation;
992		logical = old_root->logical;
993		level = old_root->level;
994	} else {
995		logical = eb_root->start;
996		level = btrfs_header_level(eb_root);
997	}
998
999	tm = tree_mod_log_search(fs_info, logical, time_seq);
1000	if (old_root && tm && tm->op != BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
1001		struct btrfs_tree_parent_check check = { 0 };
1002
1003		btrfs_tree_read_unlock(eb_root);
1004		free_extent_buffer(eb_root);
1005
1006		check.level = level;
1007		check.owner_root = root->root_key.objectid;
1008
1009		old = read_tree_block(fs_info, logical, &check);
1010		if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) {
1011			if (!IS_ERR(old))
1012				free_extent_buffer(old);
1013			btrfs_warn(fs_info,
1014				   "failed to read tree block %llu from get_old_root",
1015				   logical);
1016		} else {
1017			struct tree_mod_elem *tm2;
1018
1019			btrfs_tree_read_lock(old);
1020			eb = btrfs_clone_extent_buffer(old);
1021			/*
1022			 * After the lookup for the most recent tree mod operation
1023			 * above and before we locked and cloned the extent buffer
1024			 * 'old', a new tree mod log operation may have been added.
1025			 * So lookup for a more recent one to make sure the number
1026			 * of mod log operations we replay is consistent with the
1027			 * number of items we have in the cloned extent buffer,
1028			 * otherwise we can hit a BUG_ON when rewinding the extent
1029			 * buffer.
1030			 */
1031			tm2 = tree_mod_log_search(fs_info, logical, time_seq);
1032			btrfs_tree_read_unlock(old);
1033			free_extent_buffer(old);
1034			ASSERT(tm2);
1035			ASSERT(tm2 == tm || tm2->seq > tm->seq);
1036			if (!tm2 || tm2->seq < tm->seq) {
1037				free_extent_buffer(eb);
1038				return NULL;
1039			}
1040			tm = tm2;
1041		}
1042	} else if (old_root) {
1043		eb_root_owner = btrfs_header_owner(eb_root);
1044		btrfs_tree_read_unlock(eb_root);
1045		free_extent_buffer(eb_root);
1046		eb = alloc_dummy_extent_buffer(fs_info, logical);
1047	} else {
1048		eb = btrfs_clone_extent_buffer(eb_root);
1049		btrfs_tree_read_unlock(eb_root);
1050		free_extent_buffer(eb_root);
1051	}
1052
1053	if (!eb)
1054		return NULL;
1055	if (old_root) {
1056		btrfs_set_header_bytenr(eb, eb->start);
1057		btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV);
1058		btrfs_set_header_owner(eb, eb_root_owner);
1059		btrfs_set_header_level(eb, old_root->level);
1060		btrfs_set_header_generation(eb, old_generation);
1061	}
1062	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), eb,
1063				       btrfs_header_level(eb));
1064	btrfs_tree_read_lock(eb);
1065	if (tm)
1066		tree_mod_log_rewind(fs_info, eb, time_seq, tm);
1067	else
1068		WARN_ON(btrfs_header_level(eb) != 0);
1069	WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info));
1070
1071	return eb;
1072}
1073
1074int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq)
1075{
1076	struct tree_mod_elem *tm;
1077	int level;
1078	struct extent_buffer *eb_root = btrfs_root_node(root);
1079
1080	tm = tree_mod_log_oldest_root(eb_root, time_seq);
1081	if (tm && tm->op == BTRFS_MOD_LOG_ROOT_REPLACE)
1082		level = tm->old_root.level;
1083	else
1084		level = btrfs_header_level(eb_root);
1085
1086	free_extent_buffer(eb_root);
1087
1088	return level;
1089}
1090
1091/*
1092 * Return the lowest sequence number in the tree modification log.
1093 *
1094 * Return the sequence number of the oldest tree modification log user, which
1095 * corresponds to the lowest sequence number of all existing users. If there are
1096 * no users it returns 0.
1097 */
1098u64 btrfs_tree_mod_log_lowest_seq(struct btrfs_fs_info *fs_info)
1099{
1100	u64 ret = 0;
1101
1102	read_lock(&fs_info->tree_mod_log_lock);
1103	if (!list_empty(&fs_info->tree_mod_seq_list)) {
1104		struct btrfs_seq_list *elem;
1105
1106		elem = list_first_entry(&fs_info->tree_mod_seq_list,
1107					struct btrfs_seq_list, list);
1108		ret = elem->seq;
1109	}
1110	read_unlock(&fs_info->tree_mod_log_lock);
1111
1112	return ret;
1113}
1114