bcachefs: Copygc allocations shouldn't be nowait
authorKent Overstreet <kent.overstreet@gmail.com>
Tue, 15 Mar 2022 20:40:55 +0000 (16:40 -0400)
committerKent Overstreet <kent.overstreet@linux.dev>
Sun, 22 Oct 2023 21:09:29 +0000 (17:09 -0400)
We don't actually want copygc allocations to be nowait - an allocation
for copygc might fail and then later succeed due to a bucket needing to
wait on journal commit, or to be discarded.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
fs/bcachefs/move.c

index 3a5c81f3697bad7ce922905be1908cec6a055263..a219c10a71357eb41b07c9dc697e58aa48299e22 100644 (file)
@@ -352,7 +352,6 @@ int bch2_migrate_write_init(struct bch_fs *c, struct migrate_write *m,
 
        if (m->data_opts.btree_insert_flags & BTREE_INSERT_USE_RESERVE) {
                m->op.alloc_reserve = RESERVE_movinggc;
-               m->op.flags |= BCH_WRITE_ALLOC_NOWAIT;
        } else {
                /* XXX: this should probably be passed in */
                m->op.flags |= BCH_WRITE_ONLY_SPECIFIED_DEVS;