Merge tag 'bcachefs-2025-05-15' of git://evilpiepirate.org/bcachefsHEADmaster

Pull bcachefs fixes from Kent Overstreet:
 "The main user reported ones are:

   - Fix a btree iterator locking inconsistency that's been causing us
     to go emergency read-only in evacuate: "Fix broken btree_path lock
     invariants in next_node()"

   - Minor btree node cache reclaim tweak that should help with OOMs:
     don't set btree nodes as accessed on fill

   - Fix a bch2_bkey_clear_rebalance() issue that was causing rebalance
     to do needless work"

* tag 'bcachefs-2025-05-15' of git://evilpiepirate.org/bcachefs:
  bcachefs: fix wrong arg to fsck_err()
  bcachefs: Fix missing commit in backpointer to missing target
  bcachefs: Fix accidental O(n^2) in fiemap
  bcachefs: Fix set_should_be_locked() call in peek_slot()
  bcachefs: Fix self deadlock
  bcachefs: Don't set btree nodes as accessed on fill
  bcachefs: Fix livelock in journal_entry_open()
  bcachefs: Fix broken btree_path lock invariants in next_node()
  bcachefs: Don't strip rebalance_opts from indirect extents

9 files changed, 140 insertions, 66 deletions

diff --git a/fs/bcachefs/backpointers.c b/fs/bcachefs/backpointers.c
index ff26bb515150..5f195d2280a4 100644
--- a/fs/bcachefs/backpointers.c
+++ b/fs/bcachefs/backpointers.c
@@ -192,7 +192,8 @@ static inline int bch2_backpointers_maybe_flush(struct btree_trans *trans,
 static int backpointer_target_not_found(struct btree_trans *trans,
 				  struct bkey_s_c_backpointer bp,
 				  struct bkey_s_c target_k,
-				  struct bkey_buf *last_flushed)
+				  struct bkey_buf *last_flushed,
+				  bool commit)
 {
 	struct bch_fs *c = trans->c;
 	struct printbuf buf = PRINTBUF;
@@ -228,18 +229,77 @@ static int backpointer_target_not_found(struct btree_trans *trans,
 		}
 
 	if (fsck_err(trans, backpointer_to_missing_ptr,
-		     "%s", buf.buf))
+		     "%s", buf.buf)) {
 		ret = bch2_backpointer_del(trans, bp.k->p);
+		if (ret || !commit)
+			goto out;
+
+		/*
+		 * Normally, on transaction commit from inside a transaction,
+		 * we'll return -BCH_ERR_transaction_restart_nested, since a
+		 * transaction commit invalidates pointers given out by peek().
+		 *
+		 * However, since we're updating a write buffer btree, if we
+		 * return a transaction restart and loop we won't see that the
+		 * backpointer has been deleted without an additional write
+		 * buffer flush - and those are expensive.
+		 *
+		 * So we're relying on the caller immediately advancing to the
+		 * next backpointer and starting a new transaction immediately
+		 * after backpointer_get_key() returns NULL:
+		 */
+		ret = bch2_trans_commit(trans, NULL, NULL, BCH_TRANS_COMMIT_no_enospc);
+	}
+out:
 fsck_err:
 	printbuf_exit(&buf);
 	return ret;
 }
 
-struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
-					 struct bkey_s_c_backpointer bp,
-					 struct btree_iter *iter,
-					 unsigned iter_flags,
-					 struct bkey_buf *last_flushed)
+static struct btree *__bch2_backpointer_get_node(struct btree_trans *trans,
+						 struct bkey_s_c_backpointer bp,
+						 struct btree_iter *iter,
+						 struct bkey_buf *last_flushed,
+						 bool commit)
+{
+	struct bch_fs *c = trans->c;
+
+	BUG_ON(!bp.v->level);
+
+	bch2_trans_node_iter_init(trans, iter,
+				  bp.v->btree_id,
+				  bp.v->pos,
+				  0,
+				  bp.v->level - 1,
+				  0);
+	struct btree *b = bch2_btree_iter_peek_node(trans, iter);
+	if (IS_ERR_OR_NULL(b))
+		goto err;
+
+	BUG_ON(b->c.level != bp.v->level - 1);
+
+	if (extent_matches_bp(c, bp.v->btree_id, bp.v->level,
+			      bkey_i_to_s_c(&b->key), bp))
+		return b;
+
+	if (btree_node_will_make_reachable(b)) {
+		b = ERR_PTR(-BCH_ERR_backpointer_to_overwritten_btree_node);
+	} else {
+		int ret = backpointer_target_not_found(trans, bp, bkey_i_to_s_c(&b->key),
+						       last_flushed, commit);
+		b = ret ? ERR_PTR(ret) : NULL;
+	}
+err:
+	bch2_trans_iter_exit(trans, iter);
+	return b;
+}
+
+static struct bkey_s_c __bch2_backpointer_get_key(struct btree_trans *trans,
+						  struct bkey_s_c_backpointer bp,
+						  struct btree_iter *iter,
+						  unsigned iter_flags,
+						  struct bkey_buf *last_flushed,
+						  bool commit)
 {
 	struct bch_fs *c = trans->c;
 
@@ -277,10 +337,10 @@ struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
 	bch2_trans_iter_exit(trans, iter);
 
 	if (!bp.v->level) {
-		int ret = backpointer_target_not_found(trans, bp, k, last_flushed);
+		int ret = backpointer_target_not_found(trans, bp, k, last_flushed, commit);
 		return ret ? bkey_s_c_err(ret) : bkey_s_c_null;
 	} else {
-		struct btree *b = bch2_backpointer_get_node(trans, bp, iter, last_flushed);
+		struct btree *b = __bch2_backpointer_get_node(trans, bp, iter, last_flushed, commit);
 		if (b == ERR_PTR(-BCH_ERR_backpointer_to_overwritten_btree_node))
 			return bkey_s_c_null;
 		if (IS_ERR_OR_NULL(b))
@@ -295,35 +355,16 @@ struct btree *bch2_backpointer_get_node(struct btree_trans *trans,
 					struct btree_iter *iter,
 					struct bkey_buf *last_flushed)
 {
-	struct bch_fs *c = trans->c;
-
-	BUG_ON(!bp.v->level);
-
-	bch2_trans_node_iter_init(trans, iter,
-				  bp.v->btree_id,
-				  bp.v->pos,
-				  0,
-				  bp.v->level - 1,
-				  0);
-	struct btree *b = bch2_btree_iter_peek_node(trans, iter);
-	if (IS_ERR_OR_NULL(b))
-		goto err;
-
-	BUG_ON(b->c.level != bp.v->level - 1);
-
-	if (extent_matches_bp(c, bp.v->btree_id, bp.v->level,
-			      bkey_i_to_s_c(&b->key), bp))
-		return b;
+	return __bch2_backpointer_get_node(trans, bp, iter, last_flushed, true);
+}
 
-	if (btree_node_will_make_reachable(b)) {
-		b = ERR_PTR(-BCH_ERR_backpointer_to_overwritten_btree_node);
-	} else {
-		int ret = backpointer_target_not_found(trans, bp, bkey_i_to_s_c(&b->key), last_flushed);
-		b = ret ? ERR_PTR(ret) : NULL;
-	}
-err:
-	bch2_trans_iter_exit(trans, iter);
-	return b;
+struct bkey_s_c bch2_backpointer_get_key(struct btree_trans *trans,
+					 struct bkey_s_c_backpointer bp,
+					 struct btree_iter *iter,
+					 unsigned iter_flags,
+					 struct bkey_buf *last_flushed)
+{
+	return __bch2_backpointer_get_key(trans, bp, iter, iter_flags, last_flushed, true);
 }
 
 static int bch2_check_backpointer_has_valid_bucket(struct btree_trans *trans, struct bkey_s_c k,
@@ -521,7 +562,7 @@ check_existing_bp:
 	struct bkey_s_c_backpointer other_bp = bkey_s_c_to_backpointer(bp_k);
 
 	struct bkey_s_c other_extent =
-		bch2_backpointer_get_key(trans, other_bp, &other_extent_iter, 0, NULL);
+		__bch2_backpointer_get_key(trans, other_bp, &other_extent_iter, 0, NULL, false);
 	ret = bkey_err(other_extent);
 	if (ret == -BCH_ERR_backpointer_to_overwritten_btree_node)
 		ret = 0;
diff --git a/fs/bcachefs/btree_cache.c b/fs/bcachefs/btree_cache.c
index 9b80201c7982..899891295797 100644
--- a/fs/bcachefs/btree_cache.c
+++ b/fs/bcachefs/btree_cache.c
@@ -852,7 +852,6 @@ out:
 	b->sib_u64s[1]		= 0;
 	b->whiteout_u64s	= 0;
 	bch2_btree_keys_init(b);
-	set_btree_node_accessed(b);
 
 	bch2_time_stats_update(&c->times[BCH_TIME_btree_node_mem_alloc],
 			       start_time);
@@ -1286,6 +1285,10 @@ lock_node:
 			six_unlock_read(&b->c.lock);
 			goto retry;
 		}
+
+		/* avoid atomic set bit if it's not needed: */
+		if (!btree_node_accessed(b))
+			set_btree_node_accessed(b);
 	}
 
 	/* XXX: waiting on IO with btree locks held: */
@@ -1301,10 +1304,6 @@ lock_node:
 		prefetch(p + L1_CACHE_BYTES * 2);
 	}
 
-	/* avoid atomic set bit if it's not needed: */
-	if (!btree_node_accessed(b))
-		set_btree_node_accessed(b);
-
 	if (unlikely(btree_node_read_error(b))) {
 		six_unlock_read(&b->c.lock);
 		b = ERR_PTR(-BCH_ERR_btree_node_read_err_cached);
diff --git a/fs/bcachefs/btree_iter.c b/fs/bcachefs/btree_iter.c
index 59fa527ac685..a873ec1baf58 100644
--- a/fs/bcachefs/btree_iter.c
+++ b/fs/bcachefs/btree_iter.c
@@ -1971,6 +1971,12 @@ struct btree *bch2_btree_iter_next_node(struct btree_trans *trans, struct btree_
 		return NULL;
 	}
 
+	/*
+	 * We don't correctly handle nodes with extra intent locks here:
+	 * downgrade so we don't violate locking invariants
+	 */
+	bch2_btree_path_downgrade(trans, path);
+
 	if (!bch2_btree_node_relock(trans, path, path->level + 1)) {
 		__bch2_btree_path_unlock(trans, path);
 		path->l[path->level].b		= ERR_PTR(-BCH_ERR_no_btree_node_relock);
@@ -2743,7 +2749,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_trans *trans, struct btre
 	ret = trans_maybe_inject_restart(trans, _RET_IP_);
 	if (unlikely(ret)) {
 		k = bkey_s_c_err(ret);
-		goto out_no_locked;
+		goto out;
 	}
 
 	/* extents can't span inode numbers: */
@@ -2763,13 +2769,15 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_trans *trans, struct btre
 	ret = bch2_btree_path_traverse(trans, iter->path, iter->flags);
 	if (unlikely(ret)) {
 		k = bkey_s_c_err(ret);
-		goto out_no_locked;
+		goto out;
 	}
 
 	struct btree_path *path = btree_iter_path(trans, iter);
 	if (unlikely(!btree_path_node(path, path->level)))
 		return bkey_s_c_null;
 
+	btree_path_set_should_be_locked(trans, path);
+
 	if ((iter->flags & BTREE_ITER_cached) ||
 	    !(iter->flags & (BTREE_ITER_is_extents|BTREE_ITER_filter_snapshots))) {
 		k = bkey_s_c_null;
@@ -2790,12 +2798,12 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_trans *trans, struct btre
 			if (!bkey_err(k))
 				iter->k = *k.k;
 			/* We're not returning a key from iter->path: */
-			goto out_no_locked;
+			goto out;
 		}
 
-		k = bch2_btree_path_peek_slot(trans->paths + iter->path, &iter->k);
+		k = bch2_btree_path_peek_slot(btree_iter_path(trans, iter), &iter->k);
 		if (unlikely(!k.k))
-			goto out_no_locked;
+			goto out;
 
 		if (unlikely(k.k->type == KEY_TYPE_whiteout &&
 			     (iter->flags & BTREE_ITER_filter_snapshots) &&
@@ -2833,7 +2841,7 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_trans *trans, struct btre
 		}
 
 		if (unlikely(bkey_err(k)))
-			goto out_no_locked;
+			goto out;
 
 		next = k.k ? bkey_start_pos(k.k) : POS_MAX;
 
@@ -2855,8 +2863,6 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_trans *trans, struct btre
 		}
 	}
 out:
-	btree_path_set_should_be_locked(trans, btree_iter_path(trans, iter));
-out_no_locked:
 	bch2_btree_iter_verify_entry_exit(iter);
 	bch2_btree_iter_verify(trans, iter);
 	ret = bch2_btree_iter_verify_ret(trans, iter, k);
diff --git a/fs/bcachefs/disk_accounting.c b/fs/bcachefs/disk_accounting.c
index b007319b72e9..1f0422bfae35 100644
--- a/fs/bcachefs/disk_accounting.c
+++ b/fs/bcachefs/disk_accounting.c
@@ -376,6 +376,19 @@ int bch2_accounting_mem_insert(struct bch_fs *c, struct bkey_s_c_accounting a,
 	return ret;
 }
 
+int bch2_accounting_mem_insert_locked(struct bch_fs *c, struct bkey_s_c_accounting a,
+			       enum bch_accounting_mode mode)
+{
+	struct bch_replicas_padded r;
+
+	if (mode != BCH_ACCOUNTING_read &&
+	    accounting_to_replicas(&r.e, a.k->p) &&
+	    !bch2_replicas_marked_locked(c, &r.e))
+		return -BCH_ERR_btree_insert_need_mark_replicas;
+
+	return __bch2_accounting_mem_insert(c, a);
+}
+
 static bool accounting_mem_entry_is_zero(struct accounting_mem_entry *e)
 {
 	for (unsigned i = 0; i < e->nr_counters; i++)
@@ -583,7 +596,7 @@ int bch2_gc_accounting_done(struct bch_fs *c)
 					accounting_key_init(&k_i.k, &acc_k, src_v, nr);
 					bch2_accounting_mem_mod_locked(trans,
 								bkey_i_to_s_c_accounting(&k_i.k),
-								BCH_ACCOUNTING_normal);
+								BCH_ACCOUNTING_normal, true);
 
 					preempt_disable();
 					struct bch_fs_usage_base *dst = this_cpu_ptr(c->usage);
@@ -612,7 +625,7 @@ static int accounting_read_key(struct btree_trans *trans, struct bkey_s_c k)
 
 	percpu_down_read(&c->mark_lock);
 	int ret = bch2_accounting_mem_mod_locked(trans, bkey_s_c_to_accounting(k),
-						 BCH_ACCOUNTING_read);
+						 BCH_ACCOUNTING_read, false);
 	percpu_up_read(&c->mark_lock);
 	return ret;
 }
diff --git a/fs/bcachefs/disk_accounting.h b/fs/bcachefs/disk_accounting.h
index abb1f6206fe9..d557b99b3c0a 100644
--- a/fs/bcachefs/disk_accounting.h
+++ b/fs/bcachefs/disk_accounting.h
@@ -136,6 +136,7 @@ enum bch_accounting_mode {
 };
 
 int bch2_accounting_mem_insert(struct bch_fs *, struct bkey_s_c_accounting, enum bch_accounting_mode);
+int bch2_accounting_mem_insert_locked(struct bch_fs *, struct bkey_s_c_accounting, enum bch_accounting_mode);
 void bch2_accounting_mem_gc(struct bch_fs *);
 
 static inline bool bch2_accounting_is_mem(struct disk_accounting_pos acc)
@@ -150,7 +151,8 @@ static inline bool bch2_accounting_is_mem(struct disk_accounting_pos acc)
  */
 static inline int bch2_accounting_mem_mod_locked(struct btree_trans *trans,
 						 struct bkey_s_c_accounting a,
-						 enum bch_accounting_mode mode)
+						 enum bch_accounting_mode mode,
+						 bool write_locked)
 {
 	struct bch_fs *c = trans->c;
 	struct bch_accounting_mem *acc = &c->accounting;
@@ -189,7 +191,11 @@ static inline int bch2_accounting_mem_mod_locked(struct btree_trans *trans,
 
 	while ((idx = eytzinger0_find(acc->k.data, acc->k.nr, sizeof(acc->k.data[0]),
 				      accounting_pos_cmp, &a.k->p)) >= acc->k.nr) {
-		int ret = bch2_accounting_mem_insert(c, a, mode);
+		int ret = 0;
+		if (unlikely(write_locked))
+			ret = bch2_accounting_mem_insert_locked(c, a, mode);
+		else
+			ret = bch2_accounting_mem_insert(c, a, mode);
 		if (ret)
 			return ret;
 	}
@@ -206,7 +212,7 @@ static inline int bch2_accounting_mem_mod_locked(struct btree_trans *trans,
 static inline int bch2_accounting_mem_add(struct btree_trans *trans, struct bkey_s_c_accounting a, bool gc)
 {
 	percpu_down_read(&trans->c->mark_lock);
-	int ret = bch2_accounting_mem_mod_locked(trans, a, gc ? BCH_ACCOUNTING_gc : BCH_ACCOUNTING_normal);
+	int ret = bch2_accounting_mem_mod_locked(trans, a, gc ? BCH_ACCOUNTING_gc : BCH_ACCOUNTING_normal, false);
 	percpu_up_read(&trans->c->mark_lock);
 	return ret;
 }
@@ -259,7 +265,7 @@ static inline int bch2_accounting_trans_commit_hook(struct btree_trans *trans,
 	EBUG_ON(bversion_zero(a->k.bversion));
 
 	return likely(!(commit_flags & BCH_TRANS_COMMIT_skip_accounting_apply))
-		? bch2_accounting_mem_mod_locked(trans, accounting_i_to_s_c(a), BCH_ACCOUNTING_normal)
+		? bch2_accounting_mem_mod_locked(trans, accounting_i_to_s_c(a), BCH_ACCOUNTING_normal, false)
 		: 0;
 }
 
@@ -271,7 +277,7 @@ static inline void bch2_accounting_trans_commit_revert(struct btree_trans *trans
 		struct bkey_s_accounting a = accounting_i_to_s(a_i);
 
 		bch2_accounting_neg(a);
-		bch2_accounting_mem_mod_locked(trans, a.c, BCH_ACCOUNTING_normal);
+		bch2_accounting_mem_mod_locked(trans, a.c, BCH_ACCOUNTING_normal, false);
 		bch2_accounting_neg(a);
 	}
 }
diff --git a/fs/bcachefs/fs.c b/fs/bcachefs/fs.c
index b6801861c66f..4b742e62255b 100644
--- a/fs/bcachefs/fs.c
+++ b/fs/bcachefs/fs.c
@@ -1429,7 +1429,9 @@ static int bch2_next_fiemap_extent(struct btree_trans *trans,
 	if (ret)
 		goto err;
 
-	ret = bch2_next_fiemap_pagecache_extent(trans, inode, start, end, cur);
+	u64 pagecache_end = k.k ? max(start, bkey_start_offset(k.k)) : end;
+
+	ret = bch2_next_fiemap_pagecache_extent(trans, inode, start, pagecache_end, cur);
 	if (ret)
 		goto err;
 
diff --git a/fs/bcachefs/fsck.c b/fs/bcachefs/fsck.c
index 7b25cedd3e40..71d428f376a5 100644
--- a/fs/bcachefs/fsck.c
+++ b/fs/bcachefs/fsck.c
@@ -2446,7 +2446,7 @@ static int check_subvol_path(struct btree_trans *trans, struct btree_iter *iter,
 		u32 parent = le32_to_cpu(s.v->fs_path_parent);
 
 		if (darray_u32_has(&subvol_path, parent)) {
-			if (fsck_err(c, subvol_loop, "subvolume loop"))
+			if (fsck_err(trans, subvol_loop, "subvolume loop"))
 				ret = reattach_subvol(trans, s);
 			break;
 		}
diff --git a/fs/bcachefs/journal_reclaim.c b/fs/bcachefs/journal_reclaim.c
index 976464d8a695..cc00b0fc40d8 100644
--- a/fs/bcachefs/journal_reclaim.c
+++ b/fs/bcachefs/journal_reclaim.c
@@ -17,6 +17,8 @@
 #include <linux/kthread.h>
 #include <linux/sched/mm.h>
 
+static bool __should_discard_bucket(struct journal *, struct journal_device *);
+
 /* Free space calculations: */
 
 static unsigned journal_space_from(struct journal_device *ja,
@@ -203,8 +205,7 @@ void bch2_journal_space_available(struct journal *j)
 		       ja->bucket_seq[ja->dirty_idx_ondisk] < j->last_seq_ondisk)
 			ja->dirty_idx_ondisk = (ja->dirty_idx_ondisk + 1) % ja->nr;
 
-		if (ja->discard_idx != ja->dirty_idx_ondisk)
-			can_discard = true;
+		can_discard |= __should_discard_bucket(j, ja);
 
 		max_entry_size = min_t(unsigned, max_entry_size, ca->mi.bucket_size);
 		nr_online++;
@@ -264,13 +265,19 @@ out:
 
 /* Discards - last part of journal reclaim: */
 
-static bool should_discard_bucket(struct journal *j, struct journal_device *ja)
+static bool __should_discard_bucket(struct journal *j, struct journal_device *ja)
 {
-	spin_lock(&j->lock);
 	unsigned min_free = max(4, ja->nr / 8);
 
-	bool ret = bch2_journal_dev_buckets_available(j, ja, journal_space_discarded) < min_free &&
+	return bch2_journal_dev_buckets_available(j, ja, journal_space_discarded) <
+		min_free &&
 		ja->discard_idx != ja->dirty_idx_ondisk;
+}
+
+static bool should_discard_bucket(struct journal *j, struct journal_device *ja)
+{
+	spin_lock(&j->lock);
+	bool ret = __should_discard_bucket(j, ja);
 	spin_unlock(&j->lock);
 
 	return ret;
diff --git a/fs/bcachefs/rebalance.c b/fs/bcachefs/rebalance.c
index 4ccdfc1f34aa..623273556aa9 100644
--- a/fs/bcachefs/rebalance.c
+++ b/fs/bcachefs/rebalance.c
@@ -309,7 +309,7 @@ static int bch2_bkey_clear_needs_rebalance(struct btree_trans *trans,
 					   struct btree_iter *iter,
 					   struct bkey_s_c k)
 {
-	if (!bch2_bkey_rebalance_opts(k))
+	if (k.k->type == KEY_TYPE_reflink_v || !bch2_bkey_rebalance_opts(k))
 		return 0;
 
 	struct bkey_i *n = bch2_bkey_make_mut(trans, iter, &k, 0);