From: Kent Overstreet Date: Thu, 11 Jan 2024 04:47:04 +0000 (-0500) Subject: bcachefs: Reduce would_deadlock restarts X-Git-Url: http://git.maquefel.me/?a=commitdiff_plain;h=57f2d2097603fea102330e8cfe6be4a8db24809e;p=linux.git bcachefs: Reduce would_deadlock restarts We don't have to take locks in any particular ordering - we'll make forward progress just fine - but if we try to stick to an ordering, it can help to avoid excessive would_deadlock transaction restarts. This tweaks the reflink path to take extents btree locks in the right order. Signed-off-by: Kent Overstreet --- diff --git a/fs/bcachefs/reflink.c b/fs/bcachefs/reflink.c index 6070109174210..98255aa64e226 100644 --- a/fs/bcachefs/reflink.c +++ b/fs/bcachefs/reflink.c @@ -486,6 +486,13 @@ s64 bch2_remap_range(struct bch_fs *c, bch2_btree_iter_set_snapshot(&dst_iter, dst_snapshot); + if (dst_inum.inum < src_inum.inum) { + /* Avoid some lock cycle transaction restarts */ + ret = bch2_btree_iter_traverse(&dst_iter); + if (ret) + continue; + } + dst_done = dst_iter.pos.offset - dst_start.offset; src_want = POS(src_start.inode, src_start.offset + dst_done); bch2_btree_iter_set_pos(&src_iter, src_want);