xref: /linux/fs/bcachefs/btree_update.c (revision ab475966455ce285c2c9978a3e3bfe97d75ff8d4)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include "bcachefs.h"
4 #include "btree_update.h"
5 #include "btree_iter.h"
6 #include "btree_journal_iter.h"
7 #include "btree_locking.h"
8 #include "buckets.h"
9 #include "debug.h"
10 #include "errcode.h"
11 #include "error.h"
12 #include "extents.h"
13 #include "keylist.h"
14 #include "snapshot.h"
15 #include "trace.h"
16 
17 static inline int btree_insert_entry_cmp(const struct btree_insert_entry *l,
18 					 const struct btree_insert_entry *r)
19 {
20 	return   cmp_int(l->btree_id,	r->btree_id) ?:
21 		 cmp_int(l->cached,	r->cached) ?:
22 		 -cmp_int(l->level,	r->level) ?:
23 		 bpos_cmp(l->k->k.p,	r->k->k.p);
24 }
25 
26 static int __must_check
27 bch2_trans_update_by_path(struct btree_trans *, struct btree_path *,
28 			  struct bkey_i *, enum btree_update_flags,
29 			  unsigned long ip);
30 
31 static noinline int extent_front_merge(struct btree_trans *trans,
32 				       struct btree_iter *iter,
33 				       struct bkey_s_c k,
34 				       struct bkey_i **insert,
35 				       enum btree_update_flags flags)
36 {
37 	struct bch_fs *c = trans->c;
38 	struct bkey_i *update;
39 	int ret;
40 
41 	update = bch2_bkey_make_mut_noupdate(trans, k);
42 	ret = PTR_ERR_OR_ZERO(update);
43 	if (ret)
44 		return ret;
45 
46 	if (!bch2_bkey_merge(c, bkey_i_to_s(update), bkey_i_to_s_c(*insert)))
47 		return 0;
48 
49 	ret =   bch2_key_has_snapshot_overwrites(trans, iter->btree_id, k.k->p) ?:
50 		bch2_key_has_snapshot_overwrites(trans, iter->btree_id, (*insert)->k.p);
51 	if (ret < 0)
52 		return ret;
53 	if (ret)
54 		return 0;
55 
56 	ret = bch2_btree_delete_at(trans, iter, flags);
57 	if (ret)
58 		return ret;
59 
60 	*insert = update;
61 	return 0;
62 }
63 
64 static noinline int extent_back_merge(struct btree_trans *trans,
65 				      struct btree_iter *iter,
66 				      struct bkey_i *insert,
67 				      struct bkey_s_c k)
68 {
69 	struct bch_fs *c = trans->c;
70 	int ret;
71 
72 	ret =   bch2_key_has_snapshot_overwrites(trans, iter->btree_id, insert->k.p) ?:
73 		bch2_key_has_snapshot_overwrites(trans, iter->btree_id, k.k->p);
74 	if (ret < 0)
75 		return ret;
76 	if (ret)
77 		return 0;
78 
79 	bch2_bkey_merge(c, bkey_i_to_s(insert), k);
80 	return 0;
81 }
82 
83 /*
84  * When deleting, check if we need to emit a whiteout (because we're overwriting
85  * something in an ancestor snapshot)
86  */
87 static int need_whiteout_for_snapshot(struct btree_trans *trans,
88 				      enum btree_id btree_id, struct bpos pos)
89 {
90 	struct btree_iter iter;
91 	struct bkey_s_c k;
92 	u32 snapshot = pos.snapshot;
93 	int ret;
94 
95 	if (!bch2_snapshot_parent(trans->c, pos.snapshot))
96 		return 0;
97 
98 	pos.snapshot++;
99 
100 	for_each_btree_key_norestart(trans, iter, btree_id, pos,
101 			   BTREE_ITER_ALL_SNAPSHOTS|
102 			   BTREE_ITER_NOPRESERVE, k, ret) {
103 		if (!bkey_eq(k.k->p, pos))
104 			break;
105 
106 		if (bch2_snapshot_is_ancestor(trans->c, snapshot,
107 					      k.k->p.snapshot)) {
108 			ret = !bkey_whiteout(k.k);
109 			break;
110 		}
111 	}
112 	bch2_trans_iter_exit(trans, &iter);
113 
114 	return ret;
115 }
116 
117 int __bch2_insert_snapshot_whiteouts(struct btree_trans *trans,
118 				   enum btree_id id,
119 				   struct bpos old_pos,
120 				   struct bpos new_pos)
121 {
122 	struct bch_fs *c = trans->c;
123 	struct btree_iter old_iter, new_iter = { NULL };
124 	struct bkey_s_c old_k, new_k;
125 	snapshot_id_list s;
126 	struct bkey_i *update;
127 	int ret = 0;
128 
129 	if (!bch2_snapshot_has_children(c, old_pos.snapshot))
130 		return 0;
131 
132 	darray_init(&s);
133 
134 	bch2_trans_iter_init(trans, &old_iter, id, old_pos,
135 			     BTREE_ITER_NOT_EXTENTS|
136 			     BTREE_ITER_ALL_SNAPSHOTS);
137 	while ((old_k = bch2_btree_iter_prev(&old_iter)).k &&
138 	       !(ret = bkey_err(old_k)) &&
139 	       bkey_eq(old_pos, old_k.k->p)) {
140 		struct bpos whiteout_pos =
141 			SPOS(new_pos.inode, new_pos.offset, old_k.k->p.snapshot);;
142 
143 		if (!bch2_snapshot_is_ancestor(c, old_k.k->p.snapshot, old_pos.snapshot) ||
144 		    snapshot_list_has_ancestor(c, &s, old_k.k->p.snapshot))
145 			continue;
146 
147 		new_k = bch2_bkey_get_iter(trans, &new_iter, id, whiteout_pos,
148 					   BTREE_ITER_NOT_EXTENTS|
149 					   BTREE_ITER_INTENT);
150 		ret = bkey_err(new_k);
151 		if (ret)
152 			break;
153 
154 		if (new_k.k->type == KEY_TYPE_deleted) {
155 			update = bch2_trans_kmalloc(trans, sizeof(struct bkey_i));
156 			ret = PTR_ERR_OR_ZERO(update);
157 			if (ret)
158 				break;
159 
160 			bkey_init(&update->k);
161 			update->k.p		= whiteout_pos;
162 			update->k.type		= KEY_TYPE_whiteout;
163 
164 			ret = bch2_trans_update(trans, &new_iter, update,
165 						BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
166 		}
167 		bch2_trans_iter_exit(trans, &new_iter);
168 
169 		ret = snapshot_list_add(c, &s, old_k.k->p.snapshot);
170 		if (ret)
171 			break;
172 	}
173 	bch2_trans_iter_exit(trans, &new_iter);
174 	bch2_trans_iter_exit(trans, &old_iter);
175 	darray_exit(&s);
176 
177 	return ret;
178 }
179 
180 int bch2_trans_update_extent_overwrite(struct btree_trans *trans,
181 				       struct btree_iter *iter,
182 				       enum btree_update_flags flags,
183 				       struct bkey_s_c old,
184 				       struct bkey_s_c new)
185 {
186 	enum btree_id btree_id = iter->btree_id;
187 	struct bkey_i *update;
188 	struct bpos new_start = bkey_start_pos(new.k);
189 	bool front_split = bkey_lt(bkey_start_pos(old.k), new_start);
190 	bool back_split  = bkey_gt(old.k->p, new.k->p);
191 	int ret = 0, compressed_sectors;
192 
193 	/*
194 	 * If we're going to be splitting a compressed extent, note it
195 	 * so that __bch2_trans_commit() can increase our disk
196 	 * reservation:
197 	 */
198 	if (((front_split && back_split) ||
199 	     ((front_split || back_split) && old.k->p.snapshot != new.k->p.snapshot)) &&
200 	    (compressed_sectors = bch2_bkey_sectors_compressed(old)))
201 		trans->extra_journal_res += compressed_sectors;
202 
203 	if (front_split) {
204 		update = bch2_bkey_make_mut_noupdate(trans, old);
205 		if ((ret = PTR_ERR_OR_ZERO(update)))
206 			return ret;
207 
208 		bch2_cut_back(new_start, update);
209 
210 		ret =   bch2_insert_snapshot_whiteouts(trans, btree_id,
211 					old.k->p, update->k.p) ?:
212 			bch2_btree_insert_nonextent(trans, btree_id, update,
213 					BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
214 		if (ret)
215 			return ret;
216 	}
217 
218 	/* If we're overwriting in a different snapshot - middle split: */
219 	if (old.k->p.snapshot != new.k->p.snapshot &&
220 	    (front_split || back_split)) {
221 		update = bch2_bkey_make_mut_noupdate(trans, old);
222 		if ((ret = PTR_ERR_OR_ZERO(update)))
223 			return ret;
224 
225 		bch2_cut_front(new_start, update);
226 		bch2_cut_back(new.k->p, update);
227 
228 		ret =   bch2_insert_snapshot_whiteouts(trans, btree_id,
229 					old.k->p, update->k.p) ?:
230 			bch2_btree_insert_nonextent(trans, btree_id, update,
231 					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
232 		if (ret)
233 			return ret;
234 	}
235 
236 	if (bkey_le(old.k->p, new.k->p)) {
237 		update = bch2_trans_kmalloc(trans, sizeof(*update));
238 		if ((ret = PTR_ERR_OR_ZERO(update)))
239 			return ret;
240 
241 		bkey_init(&update->k);
242 		update->k.p = old.k->p;
243 		update->k.p.snapshot = new.k->p.snapshot;
244 
245 		if (new.k->p.snapshot != old.k->p.snapshot) {
246 			update->k.type = KEY_TYPE_whiteout;
247 		} else if (btree_type_has_snapshots(btree_id)) {
248 			ret = need_whiteout_for_snapshot(trans, btree_id, update->k.p);
249 			if (ret < 0)
250 				return ret;
251 			if (ret)
252 				update->k.type = KEY_TYPE_whiteout;
253 		}
254 
255 		ret = bch2_btree_insert_nonextent(trans, btree_id, update,
256 					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|flags);
257 		if (ret)
258 			return ret;
259 	}
260 
261 	if (back_split) {
262 		update = bch2_bkey_make_mut_noupdate(trans, old);
263 		if ((ret = PTR_ERR_OR_ZERO(update)))
264 			return ret;
265 
266 		bch2_cut_front(new.k->p, update);
267 
268 		ret = bch2_trans_update_by_path(trans, iter->path, update,
269 					  BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE|
270 					  flags, _RET_IP_);
271 		if (ret)
272 			return ret;
273 	}
274 
275 	return 0;
276 }
277 
278 static int bch2_trans_update_extent(struct btree_trans *trans,
279 				    struct btree_iter *orig_iter,
280 				    struct bkey_i *insert,
281 				    enum btree_update_flags flags)
282 {
283 	struct btree_iter iter;
284 	struct bkey_s_c k;
285 	enum btree_id btree_id = orig_iter->btree_id;
286 	int ret = 0;
287 
288 	bch2_trans_iter_init(trans, &iter, btree_id, bkey_start_pos(&insert->k),
289 			     BTREE_ITER_INTENT|
290 			     BTREE_ITER_WITH_UPDATES|
291 			     BTREE_ITER_NOT_EXTENTS);
292 	k = bch2_btree_iter_peek_upto(&iter, POS(insert->k.p.inode, U64_MAX));
293 	if ((ret = bkey_err(k)))
294 		goto err;
295 	if (!k.k)
296 		goto out;
297 
298 	if (bkey_eq(k.k->p, bkey_start_pos(&insert->k))) {
299 		if (bch2_bkey_maybe_mergable(k.k, &insert->k)) {
300 			ret = extent_front_merge(trans, &iter, k, &insert, flags);
301 			if (ret)
302 				goto err;
303 		}
304 
305 		goto next;
306 	}
307 
308 	while (bkey_gt(insert->k.p, bkey_start_pos(k.k))) {
309 		bool done = bkey_lt(insert->k.p, k.k->p);
310 
311 		ret = bch2_trans_update_extent_overwrite(trans, &iter, flags, k, bkey_i_to_s_c(insert));
312 		if (ret)
313 			goto err;
314 
315 		if (done)
316 			goto out;
317 next:
318 		bch2_btree_iter_advance(&iter);
319 		k = bch2_btree_iter_peek_upto(&iter, POS(insert->k.p.inode, U64_MAX));
320 		if ((ret = bkey_err(k)))
321 			goto err;
322 		if (!k.k)
323 			goto out;
324 	}
325 
326 	if (bch2_bkey_maybe_mergable(&insert->k, k.k)) {
327 		ret = extent_back_merge(trans, &iter, insert, k);
328 		if (ret)
329 			goto err;
330 	}
331 out:
332 	if (!bkey_deleted(&insert->k))
333 		ret = bch2_btree_insert_nonextent(trans, btree_id, insert, flags);
334 err:
335 	bch2_trans_iter_exit(trans, &iter);
336 
337 	return ret;
338 }
339 
340 static noinline int flush_new_cached_update(struct btree_trans *trans,
341 					    struct btree_path *path,
342 					    struct btree_insert_entry *i,
343 					    enum btree_update_flags flags,
344 					    unsigned long ip)
345 {
346 	struct btree_path *btree_path;
347 	struct bkey k;
348 	int ret;
349 
350 	btree_path = bch2_path_get(trans, path->btree_id, path->pos, 1, 0,
351 				   BTREE_ITER_INTENT, _THIS_IP_);
352 	ret = bch2_btree_path_traverse(trans, btree_path, 0);
353 	if (ret)
354 		goto out;
355 
356 	/*
357 	 * The old key in the insert entry might actually refer to an existing
358 	 * key in the btree that has been deleted from cache and not yet
359 	 * flushed. Check for this and skip the flush so we don't run triggers
360 	 * against a stale key.
361 	 */
362 	bch2_btree_path_peek_slot_exact(btree_path, &k);
363 	if (!bkey_deleted(&k))
364 		goto out;
365 
366 	i->key_cache_already_flushed = true;
367 	i->flags |= BTREE_TRIGGER_NORUN;
368 
369 	btree_path_set_should_be_locked(btree_path);
370 	ret = bch2_trans_update_by_path(trans, btree_path, i->k, flags, ip);
371 out:
372 	bch2_path_put(trans, btree_path, true);
373 	return ret;
374 }
375 
376 static int __must_check
377 bch2_trans_update_by_path(struct btree_trans *trans, struct btree_path *path,
378 			  struct bkey_i *k, enum btree_update_flags flags,
379 			  unsigned long ip)
380 {
381 	struct bch_fs *c = trans->c;
382 	struct btree_insert_entry *i, n;
383 	u64 seq = 0;
384 	int cmp;
385 
386 	EBUG_ON(!path->should_be_locked);
387 	EBUG_ON(trans->nr_updates >= BTREE_ITER_MAX);
388 	EBUG_ON(!bpos_eq(k->k.p, path->pos));
389 
390 	/*
391 	 * The transaction journal res hasn't been allocated at this point.
392 	 * That occurs at commit time. Reuse the seq field to pass in the seq
393 	 * of a prejournaled key.
394 	 */
395 	if (flags & BTREE_UPDATE_PREJOURNAL)
396 		seq = trans->journal_res.seq;
397 
398 	n = (struct btree_insert_entry) {
399 		.flags		= flags,
400 		.bkey_type	= __btree_node_type(path->level, path->btree_id),
401 		.btree_id	= path->btree_id,
402 		.level		= path->level,
403 		.cached		= path->cached,
404 		.path		= path,
405 		.k		= k,
406 		.seq		= seq,
407 		.ip_allocated	= ip,
408 	};
409 
410 #ifdef CONFIG_BCACHEFS_DEBUG
411 	trans_for_each_update(trans, i)
412 		BUG_ON(i != trans->updates &&
413 		       btree_insert_entry_cmp(i - 1, i) >= 0);
414 #endif
415 
416 	/*
417 	 * Pending updates are kept sorted: first, find position of new update,
418 	 * then delete/trim any updates the new update overwrites:
419 	 */
420 	trans_for_each_update(trans, i) {
421 		cmp = btree_insert_entry_cmp(&n, i);
422 		if (cmp <= 0)
423 			break;
424 	}
425 
426 	if (!cmp && i < trans->updates + trans->nr_updates) {
427 		EBUG_ON(i->insert_trigger_run || i->overwrite_trigger_run);
428 
429 		bch2_path_put(trans, i->path, true);
430 		i->flags	= n.flags;
431 		i->cached	= n.cached;
432 		i->k		= n.k;
433 		i->path		= n.path;
434 		i->seq		= n.seq;
435 		i->ip_allocated	= n.ip_allocated;
436 	} else {
437 		array_insert_item(trans->updates, trans->nr_updates,
438 				  i - trans->updates, n);
439 
440 		i->old_v = bch2_btree_path_peek_slot_exact(path, &i->old_k).v;
441 		i->old_btree_u64s = !bkey_deleted(&i->old_k) ? i->old_k.u64s : 0;
442 
443 		if (unlikely(trans->journal_replay_not_finished)) {
444 			struct bkey_i *j_k =
445 				bch2_journal_keys_peek_slot(c, n.btree_id, n.level, k->k.p);
446 
447 			if (j_k) {
448 				i->old_k = j_k->k;
449 				i->old_v = &j_k->v;
450 			}
451 		}
452 	}
453 
454 	__btree_path_get(i->path, true);
455 
456 	/*
457 	 * If a key is present in the key cache, it must also exist in the
458 	 * btree - this is necessary for cache coherency. When iterating over
459 	 * a btree that's cached in the key cache, the btree iter code checks
460 	 * the key cache - but the key has to exist in the btree for that to
461 	 * work:
462 	 */
463 	if (path->cached && bkey_deleted(&i->old_k))
464 		return flush_new_cached_update(trans, path, i, flags, ip);
465 
466 	return 0;
467 }
468 
469 static noinline int bch2_trans_update_get_key_cache(struct btree_trans *trans,
470 						    struct btree_iter *iter,
471 						    struct btree_path *path)
472 {
473 	if (!iter->key_cache_path ||
474 	    !iter->key_cache_path->should_be_locked ||
475 	    !bpos_eq(iter->key_cache_path->pos, iter->pos)) {
476 		struct bkey_cached *ck;
477 		int ret;
478 
479 		if (!iter->key_cache_path)
480 			iter->key_cache_path =
481 				bch2_path_get(trans, path->btree_id, path->pos, 1, 0,
482 					      BTREE_ITER_INTENT|
483 					      BTREE_ITER_CACHED, _THIS_IP_);
484 
485 		iter->key_cache_path =
486 			bch2_btree_path_set_pos(trans, iter->key_cache_path, path->pos,
487 						iter->flags & BTREE_ITER_INTENT,
488 						_THIS_IP_);
489 
490 		ret = bch2_btree_path_traverse(trans, iter->key_cache_path,
491 					       BTREE_ITER_CACHED);
492 		if (unlikely(ret))
493 			return ret;
494 
495 		ck = (void *) iter->key_cache_path->l[0].b;
496 
497 		if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
498 			trace_and_count(trans->c, trans_restart_key_cache_raced, trans, _RET_IP_);
499 			return btree_trans_restart(trans, BCH_ERR_transaction_restart_key_cache_raced);
500 		}
501 
502 		btree_path_set_should_be_locked(iter->key_cache_path);
503 	}
504 
505 	return 0;
506 }
507 
508 int __must_check bch2_trans_update(struct btree_trans *trans, struct btree_iter *iter,
509 				   struct bkey_i *k, enum btree_update_flags flags)
510 {
511 	struct btree_path *path = iter->update_path ?: iter->path;
512 	int ret;
513 
514 	if (iter->flags & BTREE_ITER_IS_EXTENTS)
515 		return bch2_trans_update_extent(trans, iter, k, flags);
516 
517 	if (bkey_deleted(&k->k) &&
518 	    !(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
519 	    (iter->flags & BTREE_ITER_FILTER_SNAPSHOTS)) {
520 		ret = need_whiteout_for_snapshot(trans, iter->btree_id, k->k.p);
521 		if (unlikely(ret < 0))
522 			return ret;
523 
524 		if (ret)
525 			k->k.type = KEY_TYPE_whiteout;
526 	}
527 
528 	/*
529 	 * Ensure that updates to cached btrees go to the key cache:
530 	 */
531 	if (!(flags & BTREE_UPDATE_KEY_CACHE_RECLAIM) &&
532 	    !path->cached &&
533 	    !path->level &&
534 	    btree_id_cached(trans->c, path->btree_id)) {
535 		ret = bch2_trans_update_get_key_cache(trans, iter, path);
536 		if (ret)
537 			return ret;
538 
539 		path = iter->key_cache_path;
540 	}
541 
542 	return bch2_trans_update_by_path(trans, path, k, flags, _RET_IP_);
543 }
544 
545 /*
546  * Add a transaction update for a key that has already been journaled.
547  */
548 int __must_check bch2_trans_update_seq(struct btree_trans *trans, u64 seq,
549 				       struct btree_iter *iter, struct bkey_i *k,
550 				       enum btree_update_flags flags)
551 {
552 	trans->journal_res.seq = seq;
553 	return bch2_trans_update(trans, iter, k, flags|BTREE_UPDATE_NOJOURNAL|
554 						 BTREE_UPDATE_PREJOURNAL);
555 }
556 
557 int __must_check bch2_trans_update_buffered(struct btree_trans *trans,
558 					    enum btree_id btree,
559 					    struct bkey_i *k)
560 {
561 	struct btree_write_buffered_key *i;
562 	int ret;
563 
564 	EBUG_ON(trans->nr_wb_updates > trans->wb_updates_size);
565 	EBUG_ON(k->k.u64s > BTREE_WRITE_BUFERED_U64s_MAX);
566 
567 	trans_for_each_wb_update(trans, i) {
568 		if (i->btree == btree && bpos_eq(i->k.k.p, k->k.p)) {
569 			bkey_copy(&i->k, k);
570 			return 0;
571 		}
572 	}
573 
574 	if (!trans->wb_updates ||
575 	    trans->nr_wb_updates == trans->wb_updates_size) {
576 		struct btree_write_buffered_key *u;
577 
578 		if (trans->nr_wb_updates == trans->wb_updates_size) {
579 			struct btree_transaction_stats *s = btree_trans_stats(trans);
580 
581 			BUG_ON(trans->wb_updates_size > U8_MAX / 2);
582 			trans->wb_updates_size = max(1, trans->wb_updates_size * 2);
583 			if (s)
584 				s->wb_updates_size = trans->wb_updates_size;
585 		}
586 
587 		u = bch2_trans_kmalloc_nomemzero(trans,
588 					trans->wb_updates_size *
589 					sizeof(struct btree_write_buffered_key));
590 		ret = PTR_ERR_OR_ZERO(u);
591 		if (ret)
592 			return ret;
593 
594 		if (trans->nr_wb_updates)
595 			memcpy(u, trans->wb_updates, trans->nr_wb_updates *
596 			       sizeof(struct btree_write_buffered_key));
597 		trans->wb_updates = u;
598 	}
599 
600 	trans->wb_updates[trans->nr_wb_updates] = (struct btree_write_buffered_key) {
601 		.btree	= btree,
602 	};
603 
604 	bkey_copy(&trans->wb_updates[trans->nr_wb_updates].k, k);
605 	trans->nr_wb_updates++;
606 
607 	return 0;
608 }
609 
610 int bch2_bkey_get_empty_slot(struct btree_trans *trans, struct btree_iter *iter,
611 			     enum btree_id btree, struct bpos end)
612 {
613 	struct bkey_s_c k;
614 	int ret = 0;
615 
616 	bch2_trans_iter_init(trans, iter, btree, POS_MAX, BTREE_ITER_INTENT);
617 	k = bch2_btree_iter_prev(iter);
618 	ret = bkey_err(k);
619 	if (ret)
620 		goto err;
621 
622 	bch2_btree_iter_advance(iter);
623 	k = bch2_btree_iter_peek_slot(iter);
624 	ret = bkey_err(k);
625 	if (ret)
626 		goto err;
627 
628 	BUG_ON(k.k->type != KEY_TYPE_deleted);
629 
630 	if (bkey_gt(k.k->p, end)) {
631 		ret = -BCH_ERR_ENOSPC_btree_slot;
632 		goto err;
633 	}
634 
635 	return 0;
636 err:
637 	bch2_trans_iter_exit(trans, iter);
638 	return ret;
639 }
640 
641 void bch2_trans_commit_hook(struct btree_trans *trans,
642 			    struct btree_trans_commit_hook *h)
643 {
644 	h->next = trans->hooks;
645 	trans->hooks = h;
646 }
647 
648 int bch2_btree_insert_nonextent(struct btree_trans *trans,
649 				enum btree_id btree, struct bkey_i *k,
650 				enum btree_update_flags flags)
651 {
652 	struct btree_iter iter;
653 	int ret;
654 
655 	bch2_trans_iter_init(trans, &iter, btree, k->k.p,
656 			     BTREE_ITER_CACHED|
657 			     BTREE_ITER_NOT_EXTENTS|
658 			     BTREE_ITER_INTENT);
659 	ret   = bch2_btree_iter_traverse(&iter) ?:
660 		bch2_trans_update(trans, &iter, k, flags);
661 	bch2_trans_iter_exit(trans, &iter);
662 	return ret;
663 }
664 
665 int bch2_btree_insert_trans(struct btree_trans *trans, enum btree_id id,
666 			    struct bkey_i *k, enum btree_update_flags flags)
667 {
668 	struct btree_iter iter;
669 	int ret;
670 
671 	bch2_trans_iter_init(trans, &iter, id, bkey_start_pos(&k->k),
672 			     BTREE_ITER_CACHED|
673 			     BTREE_ITER_INTENT);
674 	ret   = bch2_btree_iter_traverse(&iter) ?:
675 		bch2_trans_update(trans, &iter, k, flags);
676 	bch2_trans_iter_exit(trans, &iter);
677 	return ret;
678 }
679 
680 /**
681  * bch2_btree_insert - insert keys into the extent btree
682  * @c:			pointer to struct bch_fs
683  * @id:			btree to insert into
684  * @k:			key to insert
685  * @disk_res:		must be non-NULL whenever inserting or potentially
686  *			splitting data extents
687  * @flags:		transaction commit flags
688  *
689  * Returns:		0 on success, error code on failure
690  */
691 int bch2_btree_insert(struct bch_fs *c, enum btree_id id, struct bkey_i *k,
692 		      struct disk_reservation *disk_res, int flags)
693 {
694 	return bch2_trans_do(c, disk_res, NULL, flags,
695 			     bch2_btree_insert_trans(trans, id, k, 0));
696 }
697 
698 int bch2_btree_delete_extent_at(struct btree_trans *trans, struct btree_iter *iter,
699 				unsigned len, unsigned update_flags)
700 {
701 	struct bkey_i *k;
702 
703 	k = bch2_trans_kmalloc(trans, sizeof(*k));
704 	if (IS_ERR(k))
705 		return PTR_ERR(k);
706 
707 	bkey_init(&k->k);
708 	k->k.p = iter->pos;
709 	bch2_key_resize(&k->k, len);
710 	return bch2_trans_update(trans, iter, k, update_flags);
711 }
712 
713 int bch2_btree_delete_at(struct btree_trans *trans,
714 			 struct btree_iter *iter, unsigned update_flags)
715 {
716 	return bch2_btree_delete_extent_at(trans, iter, 0, update_flags);
717 }
718 
719 int bch2_btree_delete_at_buffered(struct btree_trans *trans,
720 				  enum btree_id btree, struct bpos pos)
721 {
722 	struct bkey_i *k;
723 
724 	k = bch2_trans_kmalloc(trans, sizeof(*k));
725 	if (IS_ERR(k))
726 		return PTR_ERR(k);
727 
728 	bkey_init(&k->k);
729 	k->k.p = pos;
730 	return bch2_trans_update_buffered(trans, btree, k);
731 }
732 
733 int bch2_btree_delete(struct btree_trans *trans,
734 		      enum btree_id btree, struct bpos pos,
735 		      unsigned update_flags)
736 {
737 	struct btree_iter iter;
738 	int ret;
739 
740 	bch2_trans_iter_init(trans, &iter, btree, pos,
741 			     BTREE_ITER_CACHED|
742 			     BTREE_ITER_INTENT);
743 	ret   = bch2_btree_iter_traverse(&iter) ?:
744 		bch2_btree_delete_at(trans, &iter, update_flags);
745 	bch2_trans_iter_exit(trans, &iter);
746 
747 	return ret;
748 }
749 
750 int bch2_btree_delete_range_trans(struct btree_trans *trans, enum btree_id id,
751 				  struct bpos start, struct bpos end,
752 				  unsigned update_flags,
753 				  u64 *journal_seq)
754 {
755 	u32 restart_count = trans->restart_count;
756 	struct btree_iter iter;
757 	struct bkey_s_c k;
758 	int ret = 0;
759 
760 	bch2_trans_iter_init(trans, &iter, id, start, BTREE_ITER_INTENT);
761 	while ((k = bch2_btree_iter_peek_upto(&iter, end)).k) {
762 		struct disk_reservation disk_res =
763 			bch2_disk_reservation_init(trans->c, 0);
764 		struct bkey_i delete;
765 
766 		ret = bkey_err(k);
767 		if (ret)
768 			goto err;
769 
770 		bkey_init(&delete.k);
771 
772 		/*
773 		 * This could probably be more efficient for extents:
774 		 */
775 
776 		/*
777 		 * For extents, iter.pos won't necessarily be the same as
778 		 * bkey_start_pos(k.k) (for non extents they always will be the
779 		 * same). It's important that we delete starting from iter.pos
780 		 * because the range we want to delete could start in the middle
781 		 * of k.
782 		 *
783 		 * (bch2_btree_iter_peek() does guarantee that iter.pos >=
784 		 * bkey_start_pos(k.k)).
785 		 */
786 		delete.k.p = iter.pos;
787 
788 		if (iter.flags & BTREE_ITER_IS_EXTENTS)
789 			bch2_key_resize(&delete.k,
790 					bpos_min(end, k.k->p).offset -
791 					iter.pos.offset);
792 
793 		ret   = bch2_trans_update(trans, &iter, &delete, update_flags) ?:
794 			bch2_trans_commit(trans, &disk_res, journal_seq,
795 					  BTREE_INSERT_NOFAIL);
796 		bch2_disk_reservation_put(trans->c, &disk_res);
797 err:
798 		/*
799 		 * the bch2_trans_begin() call is in a weird place because we
800 		 * need to call it after every transaction commit, to avoid path
801 		 * overflow, but don't want to call it if the delete operation
802 		 * is a no-op and we have no work to do:
803 		 */
804 		bch2_trans_begin(trans);
805 
806 		if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
807 			ret = 0;
808 		if (ret)
809 			break;
810 	}
811 	bch2_trans_iter_exit(trans, &iter);
812 
813 	return ret ?: trans_was_restarted(trans, restart_count);
814 }
815 
816 /*
817  * bch_btree_delete_range - delete everything within a given range
818  *
819  * Range is a half open interval - [start, end)
820  */
821 int bch2_btree_delete_range(struct bch_fs *c, enum btree_id id,
822 			    struct bpos start, struct bpos end,
823 			    unsigned update_flags,
824 			    u64 *journal_seq)
825 {
826 	int ret = bch2_trans_run(c,
827 			bch2_btree_delete_range_trans(trans, id, start, end,
828 						      update_flags, journal_seq));
829 	if (ret == -BCH_ERR_transaction_restart_nested)
830 		ret = 0;
831 	return ret;
832 }
833 
834 int bch2_btree_bit_mod(struct btree_trans *trans, enum btree_id btree,
835 		       struct bpos pos, bool set)
836 {
837 	struct bkey_i *k;
838 	int ret = 0;
839 
840 	k = bch2_trans_kmalloc_nomemzero(trans, sizeof(*k));
841 	ret = PTR_ERR_OR_ZERO(k);
842 	if (unlikely(ret))
843 		return ret;
844 
845 	bkey_init(&k->k);
846 	k->k.type = set ? KEY_TYPE_set : KEY_TYPE_deleted;
847 	k->k.p = pos;
848 
849 	return bch2_trans_update_buffered(trans, btree, k);
850 }
851 
852 __printf(2, 0)
853 static int __bch2_trans_log_msg(darray_u64 *entries, const char *fmt, va_list args)
854 {
855 	struct printbuf buf = PRINTBUF;
856 	struct jset_entry_log *l;
857 	unsigned u64s;
858 	int ret;
859 
860 	prt_vprintf(&buf, fmt, args);
861 	ret = buf.allocation_failure ? -BCH_ERR_ENOMEM_trans_log_msg : 0;
862 	if (ret)
863 		goto err;
864 
865 	u64s = DIV_ROUND_UP(buf.pos, sizeof(u64));
866 
867 	ret = darray_make_room(entries, jset_u64s(u64s));
868 	if (ret)
869 		goto err;
870 
871 	l = (void *) &darray_top(*entries);
872 	l->entry.u64s		= cpu_to_le16(u64s);
873 	l->entry.btree_id	= 0;
874 	l->entry.level		= 1;
875 	l->entry.type		= BCH_JSET_ENTRY_log;
876 	l->entry.pad[0]		= 0;
877 	l->entry.pad[1]		= 0;
878 	l->entry.pad[2]		= 0;
879 	memcpy(l->d, buf.buf, buf.pos);
880 	while (buf.pos & 7)
881 		l->d[buf.pos++] = '\0';
882 
883 	entries->nr += jset_u64s(u64s);
884 err:
885 	printbuf_exit(&buf);
886 	return ret;
887 }
888 
889 __printf(3, 0)
890 static int
891 __bch2_fs_log_msg(struct bch_fs *c, unsigned commit_flags, const char *fmt,
892 		  va_list args)
893 {
894 	int ret;
895 
896 	if (!test_bit(JOURNAL_STARTED, &c->journal.flags)) {
897 		ret = __bch2_trans_log_msg(&c->journal.early_journal_entries, fmt, args);
898 	} else {
899 		ret = bch2_trans_do(c, NULL, NULL,
900 			BTREE_INSERT_LAZY_RW|commit_flags,
901 			__bch2_trans_log_msg(&trans->extra_journal_entries, fmt, args));
902 	}
903 
904 	return ret;
905 }
906 
907 __printf(2, 3)
908 int bch2_fs_log_msg(struct bch_fs *c, const char *fmt, ...)
909 {
910 	va_list args;
911 	int ret;
912 
913 	va_start(args, fmt);
914 	ret = __bch2_fs_log_msg(c, 0, fmt, args);
915 	va_end(args);
916 	return ret;
917 }
918 
919 /*
920  * Use for logging messages during recovery to enable reserved space and avoid
921  * blocking.
922  */
923 __printf(2, 3)
924 int bch2_journal_log_msg(struct bch_fs *c, const char *fmt, ...)
925 {
926 	va_list args;
927 	int ret;
928 
929 	va_start(args, fmt);
930 	ret = __bch2_fs_log_msg(c, BCH_WATERMARK_reclaim, fmt, args);
931 	va_end(args);
932 	return ret;
933 }
934