[PATCH] generic/805: skip this test if we can't mount the filesystem

[PATCH] generic/805: skip this test if we can't mount the filesystem

[Darrick J. Wong]

From: Darrick J. Wong <djwong@kernel.org>

This test doesn't work on filesystems with external devices (e.g. xfs
realtime) because the SCRATCH_DEV gets formatted with regular scratch fs
options but the filesystem is mounted with a bare _mount call, not
_scratch_mount.  Since SCRATCH_DEV is not the dm-linear device created
earlier in the test, we can't really use that either.

In any case, if we can't mount the filesystem immediately after
formatting it, then we're not satisfying a test precondition.  Change
the _fail to _notrun.

While we're at it, fix the cleanup function so that unmount failures
don't pollute the output.

Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
 tests/generic/805 |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tests/generic/805 b/tests/generic/805
index ae498f5c669808..403d36d383a973 100755
--- a/tests/generic/805
+++ b/tests/generic/805
@@ -25,7 +25,7 @@ _cleanup()
 {
 	xfs_freeze -u $SCRATCH_MNT 2>/dev/null
 	$DMSETUP_PROG resume $lvdev >> $seqres.full 2>&1
-	_unmount -q $SCRATCH_MNT
+	_unmount -q $SCRATCH_MNT &>> $seqres.full
 	_dmsetup_remove $node
 	cd /
 	rm -f $tmp.*
@@ -47,7 +47,7 @@ table="0 $size_in_sector linear $SCRATCH_DEV 0"
 _dmsetup_create $node --table "$table" || \
 	_fail "failed to create dm device"
 
-_mount $lvdev $SCRATCH_MNT || _fail "failed to mount dm device"
+_mount $lvdev $SCRATCH_MNT || _notrun "failed to mount dm device"
 
 # Freeze the filesystem from userspace first, then nest a block device
 # initiated freeze/thaw cycle inside it via dm suspend/resume.

Re: [PATCH] generic/805: skip this test if we can't mount the filesystem

[Andrey Albershteyn]

On 2026-09-23 23:02:45, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
> 
> This test doesn't work on filesystems with external devices (e.g. xfs
> realtime) because the SCRATCH_DEV gets formatted with regular scratch fs
> options but the filesystem is mounted with a bare _mount call, not
> _scratch_mount.  Since SCRATCH_DEV is not the dm-linear device created
> earlier in the test, we can't really use that either.
> 
> In any case, if we can't mount the filesystem immediately after
> formatting it, then we're not satisfying a test precondition.  Change
> the _fail to _notrun.
> 
> While we're at it, fix the cleanup function so that unmount failures
> don't pollute the output.
> 
> Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>

Looks good to me
Reviewed-by: Andrey Albershteyn <aalbersh@kernel.org>

-- 
- Andrey

Re: [PATCH] generic/805: skip this test if we can't mount the filesystem

[Christoph Hellwig]

I think my version of this is better - we really should never
use the scratch helpers unless using a real scratch fs.

Re: [PATCH] generic/805: skip this test if we can't mount the filesystem

[Darrick J. Wong]

On Thu, Sep 24, 2026 at 09:58:09PM -0700, Christoph Hellwig wrote:
> I think my version of this is better - we really should never
> use the scratch helpers unless using a real scratch fs.

<nod> I think your patch works too, though if we can't _mkfs the
synthesized device then there's no point in continuing the test.

--D