misc xfstests fixes

misc xfstests fixes

[Christoph Hellwig]

Hi all,

this fixes a few failures I'm seeing on unusual configurations not
expected by the tests.

[PATCH 1/3] generic/484: force I/O to the data device for XFS

[Christoph Hellwig]

Otherwise the error injection to the data device might not work as
expected.  For example in some zoned setups I see the failures in
a slightly different spot than expected without this.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 tests/generic/484 | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tests/generic/484 b/tests/generic/484
index ec50735a5b58..0455efcb6000 100755
--- a/tests/generic/484
+++ b/tests/generic/484
@@ -42,6 +42,9 @@ _scratch_mkfs > $seqres.full 2>&1
 _dmerror_init
 _dmerror_mount
 
+# ensure we are on the data device, as dm error inject the error there
+[ "$FSTYP" == "xfs" ] && _xfs_force_bdev data $SCRATCH_MNT
+
 # create file
 testfile=$SCRATCH_MNT/syncfs-reports-errors
 touch $testfile
-- 
2.47.3

[PATCH 2/3] xfs/049: create the nested XFS file systems on the loop device

[Christoph Hellwig]

Without this I see failures on 4k sector size RT devices, as mkfs.xfs
can't pick up the logical block size on files.  Note that the test
already does this for the nested ext2 image as well.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 tests/xfs/049 | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/tests/xfs/049 b/tests/xfs/049
index 07feb58c9ad6..a3f478fa9351 100755
--- a/tests/xfs/049
+++ b/tests/xfs/049
@@ -58,7 +58,9 @@ mount -t ext2 $SCRATCH_DEV $SCRATCH_MNT >> $seqres.full 2>&1 \
     || _fail "!!! failed to mount"
 
 _log "Create xfs fs in file on scratch"
-${MKFS_XFS_PROG} -f -dfile,name=$SCRATCH_MNT/test.xfs,size=40m \
+truncate -s 40m $SCRATCH_MNT/test.xfs
+loop_dev1=$(_create_loop_device $SCRATCH_MNT/test.xfs)
+${MKFS_XFS_PROG} -f $loop_dev1 \
     >> $seqres.full 2>&1 \
     || _fail "!!! failed to mkfs xfs"
 
@@ -67,7 +69,6 @@ mkdir $SCRATCH_MNT/test $SCRATCH_MNT/test2 >> $seqres.full 2>&1 \
     || _fail "!!! failed to make mount points"
 
 _log "Mount xfs via loop"
-loop_dev1=$(_create_loop_device $SCRATCH_MNT/test.xfs)
 _mount $loop_dev1 $SCRATCH_MNT/test >> $seqres.full 2>&1 \
     || _fail "!!! failed to loop mount xfs"
 
-- 
2.47.3

[PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Christoph Hellwig]

I get an "inode btree counters not supported without finobt support"
with some zoned setups with the latests xfsprogs.  Just _notrun the
test if we can't create the original file system feature combination
that we're trying to upgrade from.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 tests/xfs/158 | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/tests/xfs/158 b/tests/xfs/158
index 89bf8c851659..02ab39ffda0b 100755
--- a/tests/xfs/158
+++ b/tests/xfs/158
@@ -22,12 +22,14 @@ _scratch_mkfs -m crc=1,inobtcount=1,finobt=0 &> $seqres.full && \
 	echo "Should not be able to format with inobtcount but not finobt."
 
 # Make sure we can't upgrade a filesystem to inobtcount without finobt.
-_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
+try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \
+	_notrun "invalid feature combination" >> $seqres.full
 _scratch_xfs_admin -O inobtcount=1 2>> $seqres.full
 _check_scratch_xfs_features INOBTCNT
 
 # Format V5 filesystem without inode btree counter support and populate it.
-_scratch_mkfs -m crc=1,inobtcount=0 >> $seqres.full
+_scratch_mkfs -m crc=1,inobtcount=0 || \
+	_notrun "invalid feature combination" >> $seqres.full
 _scratch_mount
 
 mkdir $SCRATCH_MNT/stress
-- 
2.47.3

Re: [PATCH 1/3] generic/484: force I/O to the data device for XFS

[Darrick J. Wong]

On Fri, Nov 21, 2025 at 08:10:05AM +0100, Christoph Hellwig wrote:
> Otherwise the error injection to the data device might not work as
> expected.  For example in some zoned setups I see the failures in
> a slightly different spot than expected without this.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Wow, how did I miss this all these years?

Ohhh, because this is an internal-rtdev failure isn't it?

How about removing the "unset SCRATCH_RTDEV" a few lines up too?

With that,
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D

> ---
>  tests/generic/484 | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/tests/generic/484 b/tests/generic/484
> index ec50735a5b58..0455efcb6000 100755
> --- a/tests/generic/484
> +++ b/tests/generic/484
> @@ -42,6 +42,9 @@ _scratch_mkfs > $seqres.full 2>&1
>  _dmerror_init
>  _dmerror_mount
>  
> +# ensure we are on the data device, as dm error inject the error there
> +[ "$FSTYP" == "xfs" ] && _xfs_force_bdev data $SCRATCH_MNT
> +
>  # create file
>  testfile=$SCRATCH_MNT/syncfs-reports-errors
>  touch $testfile
> -- 
> 2.47.3
> 
> 

Re: [PATCH 2/3] xfs/049: create the nested XFS file systems on the loop device

[Darrick J. Wong]

On Fri, Nov 21, 2025 at 08:10:06AM +0100, Christoph Hellwig wrote:
> Without this I see failures on 4k sector size RT devices, as mkfs.xfs
> can't pick up the logical block size on files.  Note that the test
> already does this for the nested ext2 image as well.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

LGTM,
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D

> ---
>  tests/xfs/049 | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/049 b/tests/xfs/049
> index 07feb58c9ad6..a3f478fa9351 100755
> --- a/tests/xfs/049
> +++ b/tests/xfs/049
> @@ -58,7 +58,9 @@ mount -t ext2 $SCRATCH_DEV $SCRATCH_MNT >> $seqres.full 2>&1 \
>      || _fail "!!! failed to mount"
>  
>  _log "Create xfs fs in file on scratch"
> -${MKFS_XFS_PROG} -f -dfile,name=$SCRATCH_MNT/test.xfs,size=40m \
> +truncate -s 40m $SCRATCH_MNT/test.xfs
> +loop_dev1=$(_create_loop_device $SCRATCH_MNT/test.xfs)
> +${MKFS_XFS_PROG} -f $loop_dev1 \
>      >> $seqres.full 2>&1 \
>      || _fail "!!! failed to mkfs xfs"
>  
> @@ -67,7 +69,6 @@ mkdir $SCRATCH_MNT/test $SCRATCH_MNT/test2 >> $seqres.full 2>&1 \
>      || _fail "!!! failed to make mount points"
>  
>  _log "Mount xfs via loop"
> -loop_dev1=$(_create_loop_device $SCRATCH_MNT/test.xfs)
>  _mount $loop_dev1 $SCRATCH_MNT/test >> $seqres.full 2>&1 \
>      || _fail "!!! failed to loop mount xfs"
>  
> -- 
> 2.47.3
> 
> 

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Darrick J. Wong]

On Fri, Nov 21, 2025 at 08:10:07AM +0100, Christoph Hellwig wrote:
> I get an "inode btree counters not supported without finobt support"
> with some zoned setups with the latests xfsprogs.  Just _notrun the
> test if we can't create the original file system feature combination
> that we're trying to upgrade from.

What configuration, specifically?  I haven't noticed any problems with
xfs/158 and:

MKFS_OPTIONS  -- -f -m metadir=1 -r zoned=1 /dev/sdf
MKFS_OPTIONS  -- -f -rrtdev=/dev/sdd -m metadir=1 -d rtinherit=1 -r zoned=1 /dev/sdf

--D

> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  tests/xfs/158 | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/158 b/tests/xfs/158
> index 89bf8c851659..02ab39ffda0b 100755
> --- a/tests/xfs/158
> +++ b/tests/xfs/158
> @@ -22,12 +22,14 @@ _scratch_mkfs -m crc=1,inobtcount=1,finobt=0 &> $seqres.full && \
>  	echo "Should not be able to format with inobtcount but not finobt."
>  
>  # Make sure we can't upgrade a filesystem to inobtcount without finobt.
> -_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
> +try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \
> +	_notrun "invalid feature combination" >> $seqres.full
>  _scratch_xfs_admin -O inobtcount=1 2>> $seqres.full
>  _check_scratch_xfs_features INOBTCNT
>  
>  # Format V5 filesystem without inode btree counter support and populate it.
> -_scratch_mkfs -m crc=1,inobtcount=0 >> $seqres.full
> +_scratch_mkfs -m crc=1,inobtcount=0 || \
> +	_notrun "invalid feature combination" >> $seqres.full
>  _scratch_mount
>  
>  mkdir $SCRATCH_MNT/stress
> -- 
> 2.47.3
> 
> 

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Zorro Lang]

On Fri, Nov 21, 2025 at 08:10:07AM +0100, Christoph Hellwig wrote:
> I get an "inode btree counters not supported without finobt support"
> with some zoned setups with the latests xfsprogs.  Just _notrun the
> test if we can't create the original file system feature combination
> that we're trying to upgrade from.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  tests/xfs/158 | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/158 b/tests/xfs/158
> index 89bf8c851659..02ab39ffda0b 100755
> --- a/tests/xfs/158
> +++ b/tests/xfs/158
> @@ -22,12 +22,14 @@ _scratch_mkfs -m crc=1,inobtcount=1,finobt=0 &> $seqres.full && \
>  	echo "Should not be able to format with inobtcount but not finobt."
>  
>  # Make sure we can't upgrade a filesystem to inobtcount without finobt.
> -_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
> +try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \

Hmm... is "try_scratch_mkfs" a function in your personal repo ?

> +	_notrun "invalid feature combination" >> $seqres.full
>  _scratch_xfs_admin -O inobtcount=1 2>> $seqres.full
>  _check_scratch_xfs_features INOBTCNT
>  
>  # Format V5 filesystem without inode btree counter support and populate it.
> -_scratch_mkfs -m crc=1,inobtcount=0 >> $seqres.full
> +_scratch_mkfs -m crc=1,inobtcount=0 || \
> +	_notrun "invalid feature combination" >> $seqres.full
>  _scratch_mount
>  
>  mkdir $SCRATCH_MNT/stress
> -- 
> 2.47.3
> 

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Christoph Hellwig]

On Tue, Dec 02, 2025 at 12:49:00PM +0800, Zorro Lang wrote:
> >  # Make sure we can't upgrade a filesystem to inobtcount without finobt.
> > -_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
> > +try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \
> 
> Hmm... is "try_scratch_mkfs" a function in your personal repo ?

This was supposed to be _try_scratch_mkfs.  But I guess that typo
is a good way to make the test never fail :)

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Zorro Lang]

On Tue, Dec 02, 2025 at 08:36:09AM +0100, Christoph Hellwig wrote:
> On Tue, Dec 02, 2025 at 12:49:00PM +0800, Zorro Lang wrote:
> > >  # Make sure we can't upgrade a filesystem to inobtcount without finobt.
> > > -_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
> > > +try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \
> > 
> > Hmm... is "try_scratch_mkfs" a function in your personal repo ?
> 
> This was supposed to be _try_scratch_mkfs.  But I guess that typo
> is a good way to make the test never fail :)

Hah, but there's not _try_scratch_mkfs in mainline fstests currently, there's
_try_scratch_mkfs_xfs .

> 

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Zorro Lang]

On Fri, Nov 21, 2025 at 08:10:07AM +0100, Christoph Hellwig wrote:
> I get an "inode btree counters not supported without finobt support"
> with some zoned setups with the latests xfsprogs.  Just _notrun the
> test if we can't create the original file system feature combination
> that we're trying to upgrade from.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  tests/xfs/158 | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/tests/xfs/158 b/tests/xfs/158
> index 89bf8c851659..02ab39ffda0b 100755
> --- a/tests/xfs/158
> +++ b/tests/xfs/158
> @@ -22,12 +22,14 @@ _scratch_mkfs -m crc=1,inobtcount=1,finobt=0 &> $seqres.full && \
>  	echo "Should not be able to format with inobtcount but not finobt."
>  
>  # Make sure we can't upgrade a filesystem to inobtcount without finobt.
> -_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 >> $seqres.full
> +try_scratch_mkfs -m crc=1,inobtcount=0,finobt=0 || \
> +	_notrun "invalid feature combination" >> $seqres.full
>  _scratch_xfs_admin -O inobtcount=1 2>> $seqres.full
>  _check_scratch_xfs_features INOBTCNT
>  
>  # Format V5 filesystem without inode btree counter support and populate it.
> -_scratch_mkfs -m crc=1,inobtcount=0 >> $seqres.full
> +_scratch_mkfs -m crc=1,inobtcount=0 || \
> +	_notrun "invalid feature combination" >> $seqres.full

Hi Christoph,

BTW, shouldn't we do ">> $seqres.full" behind the _scratch_mkfs, why does that
for _notrun?

Two patches of this patchset have been acked. As this patchset is a random fix,
so I'd like to merge those 2 "acked" patches at first, then you can re-send
the 3rd patch with other patches later. Is that good to you?

Thanks,
Zorro

>  _scratch_mount
>  
>  mkdir $SCRATCH_MNT/stress
> -- 
> 2.47.3
> 

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Christoph Hellwig]

On Fri, Dec 05, 2025 at 03:31:25PM +0800, Zorro Lang wrote:
> > +_scratch_mkfs -m crc=1,inobtcount=0 || \
> > +	_notrun "invalid feature combination" >> $seqres.full
> 
> Hi Christoph,
> 
> BTW, shouldn't we do ">> $seqres.full" behind the _scratch_mkfs, why does that
> for _notrun?
> 
> Two patches of this patchset have been acked. As this patchset is a random fix,
> so I'd like to merge those 2 "acked" patches at first, then you can re-send
> the 3rd patch with other patches later. Is that good to you?

Please drop it.  I have some ideas to attack the whole set of issues with
combining flags in a more systematic way, but I'll need some time to try
and test them.

Re: [PATCH 3/3] xfs/158: _notrun when the file system can't be created

[Zorro Lang]

On Fri, Dec 05, 2025 at 09:08:11AM +0100, Christoph Hellwig wrote:
> On Fri, Dec 05, 2025 at 03:31:25PM +0800, Zorro Lang wrote:
> > > +_scratch_mkfs -m crc=1,inobtcount=0 || \
> > > +	_notrun "invalid feature combination" >> $seqres.full
> > 
> > Hi Christoph,
> > 
> > BTW, shouldn't we do ">> $seqres.full" behind the _scratch_mkfs, why does that
> > for _notrun?
> > 
> > Two patches of this patchset have been acked. As this patchset is a random fix,
> > so I'd like to merge those 2 "acked" patches at first, then you can re-send
> > the 3rd patch with other patches later. Is that good to you?
> 
> Please drop it.  I have some ideas to attack the whole set of issues with
> combining flags in a more systematic way, but I'll need some time to try
> and test them.

OK, thanks Christoph, I'll merge other 2 patches in next fstests release.

>