xref: /openbmc/linux/fs/btrfs/tree-mod-log.c (revision 2fa5ebe3)
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 
10 struct tree_mod_root {
11 	u64 logical;
12 	u8 level;
13 };
14 
15 struct 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  */
47 static inline 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  */
60 u64 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 
74 void 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  */
133 static 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  */
173 static inline bool tree_mod_dont_log(struct btrfs_fs_info *fs_info,
174 				    struct extent_buffer *eb)
175 {
176 	if (!test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags))
177 		return true;
178 	if (eb && btrfs_header_level(eb) == 0)
179 		return true;
180 
181 	write_lock(&fs_info->tree_mod_log_lock);
182 	if (list_empty(&(fs_info)->tree_mod_seq_list)) {
183 		write_unlock(&fs_info->tree_mod_log_lock);
184 		return true;
185 	}
186 
187 	return false;
188 }
189 
190 /* Similar to tree_mod_dont_log, but doesn't acquire any locks. */
191 static inline bool tree_mod_need_log(const struct btrfs_fs_info *fs_info,
192 				    struct extent_buffer *eb)
193 {
194 	if (!test_bit(BTRFS_FS_TREE_MOD_LOG_USERS, &fs_info->flags))
195 		return false;
196 	if (eb && btrfs_header_level(eb) == 0)
197 		return false;
198 
199 	return true;
200 }
201 
202 static struct tree_mod_elem *alloc_tree_mod_elem(struct extent_buffer *eb,
203 						 int slot,
204 						 enum btrfs_mod_log_op op)
205 {
206 	struct tree_mod_elem *tm;
207 
208 	tm = kzalloc(sizeof(*tm), GFP_NOFS);
209 	if (!tm)
210 		return NULL;
211 
212 	tm->logical = eb->start;
213 	if (op != BTRFS_MOD_LOG_KEY_ADD) {
214 		btrfs_node_key(eb, &tm->key, slot);
215 		tm->blockptr = btrfs_node_blockptr(eb, slot);
216 	}
217 	tm->op = op;
218 	tm->slot = slot;
219 	tm->generation = btrfs_node_ptr_generation(eb, slot);
220 	RB_CLEAR_NODE(&tm->node);
221 
222 	return tm;
223 }
224 
225 int btrfs_tree_mod_log_insert_key(struct extent_buffer *eb, int slot,
226 				  enum btrfs_mod_log_op op)
227 {
228 	struct tree_mod_elem *tm;
229 	int ret;
230 
231 	if (!tree_mod_need_log(eb->fs_info, eb))
232 		return 0;
233 
234 	tm = alloc_tree_mod_elem(eb, slot, op);
235 	if (!tm)
236 		return -ENOMEM;
237 
238 	if (tree_mod_dont_log(eb->fs_info, eb)) {
239 		kfree(tm);
240 		return 0;
241 	}
242 
243 	ret = tree_mod_log_insert(eb->fs_info, tm);
244 	write_unlock(&eb->fs_info->tree_mod_log_lock);
245 	if (ret)
246 		kfree(tm);
247 
248 	return ret;
249 }
250 
251 int btrfs_tree_mod_log_insert_move(struct extent_buffer *eb,
252 				   int dst_slot, int src_slot,
253 				   int nr_items)
254 {
255 	struct tree_mod_elem *tm = NULL;
256 	struct tree_mod_elem **tm_list = NULL;
257 	int ret = 0;
258 	int i;
259 	bool locked = false;
260 
261 	if (!tree_mod_need_log(eb->fs_info, eb))
262 		return 0;
263 
264 	tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS);
265 	if (!tm_list)
266 		return -ENOMEM;
267 
268 	tm = kzalloc(sizeof(*tm), GFP_NOFS);
269 	if (!tm) {
270 		ret = -ENOMEM;
271 		goto free_tms;
272 	}
273 
274 	tm->logical = eb->start;
275 	tm->slot = src_slot;
276 	tm->move.dst_slot = dst_slot;
277 	tm->move.nr_items = nr_items;
278 	tm->op = BTRFS_MOD_LOG_MOVE_KEYS;
279 
280 	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
281 		tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot,
282 				BTRFS_MOD_LOG_KEY_REMOVE_WHILE_MOVING);
283 		if (!tm_list[i]) {
284 			ret = -ENOMEM;
285 			goto free_tms;
286 		}
287 	}
288 
289 	if (tree_mod_dont_log(eb->fs_info, eb))
290 		goto free_tms;
291 	locked = true;
292 
293 	/*
294 	 * When we override something during the move, we log these removals.
295 	 * This can only happen when we move towards the beginning of the
296 	 * buffer, i.e. dst_slot < src_slot.
297 	 */
298 	for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
299 		ret = tree_mod_log_insert(eb->fs_info, tm_list[i]);
300 		if (ret)
301 			goto free_tms;
302 	}
303 
304 	ret = tree_mod_log_insert(eb->fs_info, tm);
305 	if (ret)
306 		goto free_tms;
307 	write_unlock(&eb->fs_info->tree_mod_log_lock);
308 	kfree(tm_list);
309 
310 	return 0;
311 
312 free_tms:
313 	for (i = 0; i < nr_items; i++) {
314 		if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
315 			rb_erase(&tm_list[i]->node, &eb->fs_info->tree_mod_log);
316 		kfree(tm_list[i]);
317 	}
318 	if (locked)
319 		write_unlock(&eb->fs_info->tree_mod_log_lock);
320 	kfree(tm_list);
321 	kfree(tm);
322 
323 	return ret;
324 }
325 
326 static inline int tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
327 				       struct tree_mod_elem **tm_list,
328 				       int nritems)
329 {
330 	int i, j;
331 	int ret;
332 
333 	for (i = nritems - 1; i >= 0; i--) {
334 		ret = tree_mod_log_insert(fs_info, tm_list[i]);
335 		if (ret) {
336 			for (j = nritems - 1; j > i; j--)
337 				rb_erase(&tm_list[j]->node,
338 					 &fs_info->tree_mod_log);
339 			return ret;
340 		}
341 	}
342 
343 	return 0;
344 }
345 
346 int btrfs_tree_mod_log_insert_root(struct extent_buffer *old_root,
347 				   struct extent_buffer *new_root,
348 				   bool log_removal)
349 {
350 	struct btrfs_fs_info *fs_info = old_root->fs_info;
351 	struct tree_mod_elem *tm = NULL;
352 	struct tree_mod_elem **tm_list = NULL;
353 	int nritems = 0;
354 	int ret = 0;
355 	int i;
356 
357 	if (!tree_mod_need_log(fs_info, NULL))
358 		return 0;
359 
360 	if (log_removal && btrfs_header_level(old_root) > 0) {
361 		nritems = btrfs_header_nritems(old_root);
362 		tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *),
363 				  GFP_NOFS);
364 		if (!tm_list) {
365 			ret = -ENOMEM;
366 			goto free_tms;
367 		}
368 		for (i = 0; i < nritems; i++) {
369 			tm_list[i] = alloc_tree_mod_elem(old_root, i,
370 			    BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING);
371 			if (!tm_list[i]) {
372 				ret = -ENOMEM;
373 				goto free_tms;
374 			}
375 		}
376 	}
377 
378 	tm = kzalloc(sizeof(*tm), GFP_NOFS);
379 	if (!tm) {
380 		ret = -ENOMEM;
381 		goto free_tms;
382 	}
383 
384 	tm->logical = new_root->start;
385 	tm->old_root.logical = old_root->start;
386 	tm->old_root.level = btrfs_header_level(old_root);
387 	tm->generation = btrfs_header_generation(old_root);
388 	tm->op = BTRFS_MOD_LOG_ROOT_REPLACE;
389 
390 	if (tree_mod_dont_log(fs_info, NULL))
391 		goto free_tms;
392 
393 	if (tm_list)
394 		ret = tree_mod_log_free_eb(fs_info, tm_list, nritems);
395 	if (!ret)
396 		ret = tree_mod_log_insert(fs_info, tm);
397 
398 	write_unlock(&fs_info->tree_mod_log_lock);
399 	if (ret)
400 		goto free_tms;
401 	kfree(tm_list);
402 
403 	return ret;
404 
405 free_tms:
406 	if (tm_list) {
407 		for (i = 0; i < nritems; i++)
408 			kfree(tm_list[i]);
409 		kfree(tm_list);
410 	}
411 	kfree(tm);
412 
413 	return ret;
414 }
415 
416 static struct tree_mod_elem *__tree_mod_log_search(struct btrfs_fs_info *fs_info,
417 						   u64 start, u64 min_seq,
418 						   bool smallest)
419 {
420 	struct rb_root *tm_root;
421 	struct rb_node *node;
422 	struct tree_mod_elem *cur = NULL;
423 	struct tree_mod_elem *found = NULL;
424 
425 	read_lock(&fs_info->tree_mod_log_lock);
426 	tm_root = &fs_info->tree_mod_log;
427 	node = tm_root->rb_node;
428 	while (node) {
429 		cur = rb_entry(node, struct tree_mod_elem, node);
430 		if (cur->logical < start) {
431 			node = node->rb_left;
432 		} else if (cur->logical > start) {
433 			node = node->rb_right;
434 		} else if (cur->seq < min_seq) {
435 			node = node->rb_left;
436 		} else if (!smallest) {
437 			/* We want the node with the highest seq */
438 			if (found)
439 				BUG_ON(found->seq > cur->seq);
440 			found = cur;
441 			node = node->rb_left;
442 		} else if (cur->seq > min_seq) {
443 			/* We want the node with the smallest seq */
444 			if (found)
445 				BUG_ON(found->seq < cur->seq);
446 			found = cur;
447 			node = node->rb_right;
448 		} else {
449 			found = cur;
450 			break;
451 		}
452 	}
453 	read_unlock(&fs_info->tree_mod_log_lock);
454 
455 	return found;
456 }
457 
458 /*
459  * This returns the element from the log with the smallest time sequence
460  * value that's in the log (the oldest log item). Any element with a time
461  * sequence lower than min_seq will be ignored.
462  */
463 static struct tree_mod_elem *tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info,
464 							u64 start, u64 min_seq)
465 {
466 	return __tree_mod_log_search(fs_info, start, min_seq, true);
467 }
468 
469 /*
470  * This returns the element from the log with the largest time sequence
471  * value that's in the log (the most recent log item). Any element with
472  * a time sequence lower than min_seq will be ignored.
473  */
474 static struct tree_mod_elem *tree_mod_log_search(struct btrfs_fs_info *fs_info,
475 						 u64 start, u64 min_seq)
476 {
477 	return __tree_mod_log_search(fs_info, start, min_seq, false);
478 }
479 
480 int btrfs_tree_mod_log_eb_copy(struct extent_buffer *dst,
481 			       struct extent_buffer *src,
482 			       unsigned long dst_offset,
483 			       unsigned long src_offset,
484 			       int nr_items)
485 {
486 	struct btrfs_fs_info *fs_info = dst->fs_info;
487 	int ret = 0;
488 	struct tree_mod_elem **tm_list = NULL;
489 	struct tree_mod_elem **tm_list_add, **tm_list_rem;
490 	int i;
491 	bool locked = false;
492 
493 	if (!tree_mod_need_log(fs_info, NULL))
494 		return 0;
495 
496 	if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
497 		return 0;
498 
499 	tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *),
500 			  GFP_NOFS);
501 	if (!tm_list)
502 		return -ENOMEM;
503 
504 	tm_list_add = tm_list;
505 	tm_list_rem = tm_list + nr_items;
506 	for (i = 0; i < nr_items; i++) {
507 		tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset,
508 						     BTRFS_MOD_LOG_KEY_REMOVE);
509 		if (!tm_list_rem[i]) {
510 			ret = -ENOMEM;
511 			goto free_tms;
512 		}
513 
514 		tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset,
515 						     BTRFS_MOD_LOG_KEY_ADD);
516 		if (!tm_list_add[i]) {
517 			ret = -ENOMEM;
518 			goto free_tms;
519 		}
520 	}
521 
522 	if (tree_mod_dont_log(fs_info, NULL))
523 		goto free_tms;
524 	locked = true;
525 
526 	for (i = 0; i < nr_items; i++) {
527 		ret = tree_mod_log_insert(fs_info, tm_list_rem[i]);
528 		if (ret)
529 			goto free_tms;
530 		ret = tree_mod_log_insert(fs_info, tm_list_add[i]);
531 		if (ret)
532 			goto free_tms;
533 	}
534 
535 	write_unlock(&fs_info->tree_mod_log_lock);
536 	kfree(tm_list);
537 
538 	return 0;
539 
540 free_tms:
541 	for (i = 0; i < nr_items * 2; i++) {
542 		if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
543 			rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
544 		kfree(tm_list[i]);
545 	}
546 	if (locked)
547 		write_unlock(&fs_info->tree_mod_log_lock);
548 	kfree(tm_list);
549 
550 	return ret;
551 }
552 
553 int btrfs_tree_mod_log_free_eb(struct extent_buffer *eb)
554 {
555 	struct tree_mod_elem **tm_list = NULL;
556 	int nritems = 0;
557 	int i;
558 	int ret = 0;
559 
560 	if (!tree_mod_need_log(eb->fs_info, eb))
561 		return 0;
562 
563 	nritems = btrfs_header_nritems(eb);
564 	tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS);
565 	if (!tm_list)
566 		return -ENOMEM;
567 
568 	for (i = 0; i < nritems; i++) {
569 		tm_list[i] = alloc_tree_mod_elem(eb, i,
570 				    BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING);
571 		if (!tm_list[i]) {
572 			ret = -ENOMEM;
573 			goto free_tms;
574 		}
575 	}
576 
577 	if (tree_mod_dont_log(eb->fs_info, eb))
578 		goto free_tms;
579 
580 	ret = tree_mod_log_free_eb(eb->fs_info, tm_list, nritems);
581 	write_unlock(&eb->fs_info->tree_mod_log_lock);
582 	if (ret)
583 		goto free_tms;
584 	kfree(tm_list);
585 
586 	return 0;
587 
588 free_tms:
589 	for (i = 0; i < nritems; i++)
590 		kfree(tm_list[i]);
591 	kfree(tm_list);
592 
593 	return ret;
594 }
595 
596 /*
597  * Returns the logical address of the oldest predecessor of the given root.
598  * Entries older than time_seq are ignored.
599  */
600 static struct tree_mod_elem *tree_mod_log_oldest_root(struct extent_buffer *eb_root,
601 						      u64 time_seq)
602 {
603 	struct tree_mod_elem *tm;
604 	struct tree_mod_elem *found = NULL;
605 	u64 root_logical = eb_root->start;
606 	bool looped = false;
607 
608 	if (!time_seq)
609 		return NULL;
610 
611 	/*
612 	 * The very last operation that's logged for a root is the replacement
613 	 * operation (if it is replaced at all). This has the logical address
614 	 * of the *new* root, making it the very first operation that's logged
615 	 * for this root.
616 	 */
617 	while (1) {
618 		tm = tree_mod_log_search_oldest(eb_root->fs_info, root_logical,
619 						time_seq);
620 		if (!looped && !tm)
621 			return NULL;
622 		/*
623 		 * If there are no tree operation for the oldest root, we simply
624 		 * return it. This should only happen if that (old) root is at
625 		 * level 0.
626 		 */
627 		if (!tm)
628 			break;
629 
630 		/*
631 		 * If there's an operation that's not a root replacement, we
632 		 * found the oldest version of our root. Normally, we'll find a
633 		 * BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here.
634 		 */
635 		if (tm->op != BTRFS_MOD_LOG_ROOT_REPLACE)
636 			break;
637 
638 		found = tm;
639 		root_logical = tm->old_root.logical;
640 		looped = true;
641 	}
642 
643 	/* If there's no old root to return, return what we found instead */
644 	if (!found)
645 		found = tm;
646 
647 	return found;
648 }
649 
650 
651 /*
652  * tm is a pointer to the first operation to rewind within eb. Then, all
653  * previous operations will be rewound (until we reach something older than
654  * time_seq).
655  */
656 static void tree_mod_log_rewind(struct btrfs_fs_info *fs_info,
657 				struct extent_buffer *eb,
658 				u64 time_seq,
659 				struct tree_mod_elem *first_tm)
660 {
661 	u32 n;
662 	struct rb_node *next;
663 	struct tree_mod_elem *tm = first_tm;
664 	unsigned long o_dst;
665 	unsigned long o_src;
666 	unsigned long p_size = sizeof(struct btrfs_key_ptr);
667 
668 	n = btrfs_header_nritems(eb);
669 	read_lock(&fs_info->tree_mod_log_lock);
670 	while (tm && tm->seq >= time_seq) {
671 		/*
672 		 * All the operations are recorded with the operator used for
673 		 * the modification. As we're going backwards, we do the
674 		 * opposite of each operation here.
675 		 */
676 		switch (tm->op) {
677 		case BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING:
678 			BUG_ON(tm->slot < n);
679 			fallthrough;
680 		case BTRFS_MOD_LOG_KEY_REMOVE_WHILE_MOVING:
681 		case BTRFS_MOD_LOG_KEY_REMOVE:
682 			btrfs_set_node_key(eb, &tm->key, tm->slot);
683 			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
684 			btrfs_set_node_ptr_generation(eb, tm->slot,
685 						      tm->generation);
686 			n++;
687 			break;
688 		case BTRFS_MOD_LOG_KEY_REPLACE:
689 			BUG_ON(tm->slot >= n);
690 			btrfs_set_node_key(eb, &tm->key, tm->slot);
691 			btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
692 			btrfs_set_node_ptr_generation(eb, tm->slot,
693 						      tm->generation);
694 			break;
695 		case BTRFS_MOD_LOG_KEY_ADD:
696 			/* if a move operation is needed it's in the log */
697 			n--;
698 			break;
699 		case BTRFS_MOD_LOG_MOVE_KEYS:
700 			o_dst = btrfs_node_key_ptr_offset(eb, tm->slot);
701 			o_src = btrfs_node_key_ptr_offset(eb, tm->move.dst_slot);
702 			memmove_extent_buffer(eb, o_dst, o_src,
703 					      tm->move.nr_items * p_size);
704 			break;
705 		case BTRFS_MOD_LOG_ROOT_REPLACE:
706 			/*
707 			 * This operation is special. For roots, this must be
708 			 * handled explicitly before rewinding.
709 			 * For non-roots, this operation may exist if the node
710 			 * was a root: root A -> child B; then A gets empty and
711 			 * B is promoted to the new root. In the mod log, we'll
712 			 * have a root-replace operation for B, a tree block
713 			 * that is no root. We simply ignore that operation.
714 			 */
715 			break;
716 		}
717 		next = rb_next(&tm->node);
718 		if (!next)
719 			break;
720 		tm = rb_entry(next, struct tree_mod_elem, node);
721 		if (tm->logical != first_tm->logical)
722 			break;
723 	}
724 	read_unlock(&fs_info->tree_mod_log_lock);
725 	btrfs_set_header_nritems(eb, n);
726 }
727 
728 /*
729  * Called with eb read locked. If the buffer cannot be rewound, the same buffer
730  * is returned. If rewind operations happen, a fresh buffer is returned. The
731  * returned buffer is always read-locked. If the returned buffer is not the
732  * input buffer, the lock on the input buffer is released and the input buffer
733  * is freed (its refcount is decremented).
734  */
735 struct extent_buffer *btrfs_tree_mod_log_rewind(struct btrfs_fs_info *fs_info,
736 						struct btrfs_path *path,
737 						struct extent_buffer *eb,
738 						u64 time_seq)
739 {
740 	struct extent_buffer *eb_rewin;
741 	struct tree_mod_elem *tm;
742 
743 	if (!time_seq)
744 		return eb;
745 
746 	if (btrfs_header_level(eb) == 0)
747 		return eb;
748 
749 	tm = tree_mod_log_search(fs_info, eb->start, time_seq);
750 	if (!tm)
751 		return eb;
752 
753 	if (tm->op == BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
754 		BUG_ON(tm->slot != 0);
755 		eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start);
756 		if (!eb_rewin) {
757 			btrfs_tree_read_unlock(eb);
758 			free_extent_buffer(eb);
759 			return NULL;
760 		}
761 		btrfs_set_header_bytenr(eb_rewin, eb->start);
762 		btrfs_set_header_backref_rev(eb_rewin,
763 					     btrfs_header_backref_rev(eb));
764 		btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb));
765 		btrfs_set_header_level(eb_rewin, btrfs_header_level(eb));
766 	} else {
767 		eb_rewin = btrfs_clone_extent_buffer(eb);
768 		if (!eb_rewin) {
769 			btrfs_tree_read_unlock(eb);
770 			free_extent_buffer(eb);
771 			return NULL;
772 		}
773 	}
774 
775 	btrfs_tree_read_unlock(eb);
776 	free_extent_buffer(eb);
777 
778 	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb_rewin),
779 				       eb_rewin, btrfs_header_level(eb_rewin));
780 	btrfs_tree_read_lock(eb_rewin);
781 	tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm);
782 	WARN_ON(btrfs_header_nritems(eb_rewin) >
783 		BTRFS_NODEPTRS_PER_BLOCK(fs_info));
784 
785 	return eb_rewin;
786 }
787 
788 /*
789  * Rewind the state of @root's root node to the given @time_seq value.
790  * If there are no changes, the current root->root_node is returned. If anything
791  * changed in between, there's a fresh buffer allocated on which the rewind
792  * operations are done. In any case, the returned buffer is read locked.
793  * Returns NULL on error (with no locks held).
794  */
795 struct extent_buffer *btrfs_get_old_root(struct btrfs_root *root, u64 time_seq)
796 {
797 	struct btrfs_fs_info *fs_info = root->fs_info;
798 	struct tree_mod_elem *tm;
799 	struct extent_buffer *eb = NULL;
800 	struct extent_buffer *eb_root;
801 	u64 eb_root_owner = 0;
802 	struct extent_buffer *old;
803 	struct tree_mod_root *old_root = NULL;
804 	u64 old_generation = 0;
805 	u64 logical;
806 	int level;
807 
808 	eb_root = btrfs_read_lock_root_node(root);
809 	tm = tree_mod_log_oldest_root(eb_root, time_seq);
810 	if (!tm)
811 		return eb_root;
812 
813 	if (tm->op == BTRFS_MOD_LOG_ROOT_REPLACE) {
814 		old_root = &tm->old_root;
815 		old_generation = tm->generation;
816 		logical = old_root->logical;
817 		level = old_root->level;
818 	} else {
819 		logical = eb_root->start;
820 		level = btrfs_header_level(eb_root);
821 	}
822 
823 	tm = tree_mod_log_search(fs_info, logical, time_seq);
824 	if (old_root && tm && tm->op != BTRFS_MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
825 		struct btrfs_tree_parent_check check = { 0 };
826 
827 		btrfs_tree_read_unlock(eb_root);
828 		free_extent_buffer(eb_root);
829 
830 		check.level = level;
831 		check.owner_root = root->root_key.objectid;
832 
833 		old = read_tree_block(fs_info, logical, &check);
834 		if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) {
835 			if (!IS_ERR(old))
836 				free_extent_buffer(old);
837 			btrfs_warn(fs_info,
838 				   "failed to read tree block %llu from get_old_root",
839 				   logical);
840 		} else {
841 			struct tree_mod_elem *tm2;
842 
843 			btrfs_tree_read_lock(old);
844 			eb = btrfs_clone_extent_buffer(old);
845 			/*
846 			 * After the lookup for the most recent tree mod operation
847 			 * above and before we locked and cloned the extent buffer
848 			 * 'old', a new tree mod log operation may have been added.
849 			 * So lookup for a more recent one to make sure the number
850 			 * of mod log operations we replay is consistent with the
851 			 * number of items we have in the cloned extent buffer,
852 			 * otherwise we can hit a BUG_ON when rewinding the extent
853 			 * buffer.
854 			 */
855 			tm2 = tree_mod_log_search(fs_info, logical, time_seq);
856 			btrfs_tree_read_unlock(old);
857 			free_extent_buffer(old);
858 			ASSERT(tm2);
859 			ASSERT(tm2 == tm || tm2->seq > tm->seq);
860 			if (!tm2 || tm2->seq < tm->seq) {
861 				free_extent_buffer(eb);
862 				return NULL;
863 			}
864 			tm = tm2;
865 		}
866 	} else if (old_root) {
867 		eb_root_owner = btrfs_header_owner(eb_root);
868 		btrfs_tree_read_unlock(eb_root);
869 		free_extent_buffer(eb_root);
870 		eb = alloc_dummy_extent_buffer(fs_info, logical);
871 	} else {
872 		eb = btrfs_clone_extent_buffer(eb_root);
873 		btrfs_tree_read_unlock(eb_root);
874 		free_extent_buffer(eb_root);
875 	}
876 
877 	if (!eb)
878 		return NULL;
879 	if (old_root) {
880 		btrfs_set_header_bytenr(eb, eb->start);
881 		btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV);
882 		btrfs_set_header_owner(eb, eb_root_owner);
883 		btrfs_set_header_level(eb, old_root->level);
884 		btrfs_set_header_generation(eb, old_generation);
885 	}
886 	btrfs_set_buffer_lockdep_class(btrfs_header_owner(eb), eb,
887 				       btrfs_header_level(eb));
888 	btrfs_tree_read_lock(eb);
889 	if (tm)
890 		tree_mod_log_rewind(fs_info, eb, time_seq, tm);
891 	else
892 		WARN_ON(btrfs_header_level(eb) != 0);
893 	WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info));
894 
895 	return eb;
896 }
897 
898 int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq)
899 {
900 	struct tree_mod_elem *tm;
901 	int level;
902 	struct extent_buffer *eb_root = btrfs_root_node(root);
903 
904 	tm = tree_mod_log_oldest_root(eb_root, time_seq);
905 	if (tm && tm->op == BTRFS_MOD_LOG_ROOT_REPLACE)
906 		level = tm->old_root.level;
907 	else
908 		level = btrfs_header_level(eb_root);
909 
910 	free_extent_buffer(eb_root);
911 
912 	return level;
913 }
914 
915 /*
916  * Return the lowest sequence number in the tree modification log.
917  *
918  * Return the sequence number of the oldest tree modification log user, which
919  * corresponds to the lowest sequence number of all existing users. If there are
920  * no users it returns 0.
921  */
922 u64 btrfs_tree_mod_log_lowest_seq(struct btrfs_fs_info *fs_info)
923 {
924 	u64 ret = 0;
925 
926 	read_lock(&fs_info->tree_mod_log_lock);
927 	if (!list_empty(&fs_info->tree_mod_seq_list)) {
928 		struct btrfs_seq_list *elem;
929 
930 		elem = list_first_entry(&fs_info->tree_mod_seq_list,
931 					struct btrfs_seq_list, list);
932 		ret = elem->seq;
933 	}
934 	read_unlock(&fs_info->tree_mod_log_lock);
935 
936 	return ret;
937 }
938