diff options
author | Darrick J. Wong <djwong@kernel.org> | 2022-12-26 21:11:17 +0300 |
---|---|---|
committer | Darrick J. Wong <djwong@kernel.org> | 2022-12-26 21:11:17 +0300 |
commit | 26870c3f5b15187268bf183055c7b9f29fe66079 (patch) | |
tree | 04540c702a6bbcdb7554df763225ae9c9a2f3d67 /fs/xfs | |
parent | 1b929c02afd37871d5afb9d498426f83432e71c2 (diff) | |
download | linux-26870c3f5b15187268bf183055c7b9f29fe66079.tar.xz |
xfs: don't assert if cmap covers imap after cycling lock
In xfs_reflink_fill_cow_hole, there's a debugging assertion that trips
if (after cycling the ILOCK to get a transaction) the requeried cow
mapping overlaps the start of the area being written. IOWs, it trips if
the hole in the cow fork that it's supposed to fill has been filled.
This is trivially possible since we cycled ILOCK_EXCL. If we trip the
assertion, then we know that cmap is a delalloc extent because @found is
false. Fortunately, the bmapi_write call below will convert the
delalloc extent to a real unwritten cow fork extent, so all we need to
do here is remove the assertion.
It turns out that generic/095 trips this pretty regularly with alwayscow
mode enabled.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Diffstat (limited to 'fs/xfs')
-rw-r--r-- | fs/xfs/xfs_reflink.c | 2 |
1 files changed, 0 insertions, 2 deletions
diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c index fe46bce8cae6..5535778a98f9 100644 --- a/fs/xfs/xfs_reflink.c +++ b/fs/xfs/xfs_reflink.c @@ -416,8 +416,6 @@ xfs_reflink_fill_cow_hole( goto convert; } - ASSERT(cmap->br_startoff > imap->br_startoff); - /* Allocate the entire reservation as unwritten blocks. */ nimaps = 1; error = xfs_bmapi_write(tp, ip, imap->br_startoff, imap->br_blockcount, |