diff options
| author | Kent Overstreet <kent.overstreet@linux.dev> | 2025-06-25 12:45:11 -0400 |
|---|---|---|
| committer | Kent Overstreet <kent.overstreet@linux.dev> | 2025-06-26 00:01:16 -0400 |
| commit | 3e72acb78b73ccffeaf929c039dc5a0a7a147535 (patch) | |
| tree | 5a32bd0ae84c907d929172f5f883f4c71a5e3025 /fs/bcachefs/btree_node_scan.c | |
| parent | bcachefs: btree_root_unreadable_and_scan_found_nothing should not be autofix (diff) | |
| download | linux-3e72acb78b73ccffeaf929c039dc5a0a7a147535.tar.gz linux-3e72acb78b73ccffeaf929c039dc5a0a7a147535.zip | |
bcachefs: Ensure btree node scan runs before checking for scanned nodes
Previously, calling bch2_btree_has_scanned_nodes() when btree node
scan hadn't actually run would erroniously return false - causing us to
think a btree was entirely gone.
This fixes a 6.16 regression from moving the scheduling of btree node
scan out of bch2_btree_lost_data() (fixing the bug where we'd schedule
it persistently in the superblock) and only scheduling it when
check_toploogy() is asking for scanned btree nodes.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
Diffstat (limited to 'fs/bcachefs/btree_node_scan.c')
| -rw-r--r-- | fs/bcachefs/btree_node_scan.c | 6 |
1 files changed, 5 insertions, 1 deletions
diff --git a/fs/bcachefs/btree_node_scan.c b/fs/bcachefs/btree_node_scan.c index a35847734a60..23d8c62ea4b6 100644 --- a/fs/bcachefs/btree_node_scan.c +++ b/fs/bcachefs/btree_node_scan.c @@ -521,8 +521,12 @@ bool bch2_btree_node_is_stale(struct bch_fs *c, struct btree *b) return false; } -bool bch2_btree_has_scanned_nodes(struct bch_fs *c, enum btree_id btree) +int bch2_btree_has_scanned_nodes(struct bch_fs *c, enum btree_id btree) { + int ret = bch2_run_print_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes); + if (ret) + return ret; + struct found_btree_node search = { .btree_id = btree, .level = 0, |