try_fiemap() uses FIEMAP_FLAG_SYNC which has a significant performance
impact.
Prefer seek_hole() over fiemap() to avoid this impact where possible.
seek_hole is more widely used and, arguably, has potential to be
optimised in the kernel.
Reported-By: Michael Steffens <michael_steff...@posteo.de>
Signed-off-by: Tony Breeds <t...@bakeyournoodle.com>
Cc: Kevin Wolf <kw...@redhat.com>
Cc: Markus Armbruster <arm...@redhat.com>
Cc: Stefan Hajnoczi <stefa...@redhat.com>
Cc: Max Reitz <mre...@redhat.com>
Cc: Pádraig Brady <pbr...@redhat.com>
Cc: Eric Blake <ebl...@redhat.com>
---
Changes since v1:
- split in to 2 patches
- tried to make the commit messages better
block/raw-posix.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/block/raw-posix.c b/block/raw-posix.c
index b8203e9..b438c54 100644
--- a/block/raw-posix.c
+++ b/block/raw-posix.c
@@ -1561,9 +1561,9 @@ static int64_t coroutine_fn
raw_co_get_block_status(BlockDriverState *bs,
start = sector_num * BDRV_SECTOR_SIZE;
- ret = try_fiemap(bs, start, &data, &hole, nb_sectors, pnum);
+ ret = try_seek_hole(bs, start, &data, &hole, pnum);
if (ret < 0) {
- ret = try_seek_hole(bs, start, &data, &hole, pnum);
+ ret = try_fiemap(bs, start, &data, &hole, nb_sectors, pnum);
if (ret < 0) {
/* Assume everything is allocated. */
data = 0;
--
1.9.3
