[PATCH 0/2] xfs: add regression test for small zone capacity

[PATCH 0/2] xfs: add regression test for small zone capacity

[cem]

From: Carlos Maiolino <cem@kernel.org>

This adds a regression test for the smaller sequential zone capacities
problem.

For that, we need to be able to create a zoned loop device with a custom
capacity, so the first patch extents _create_zloop to accept a custom
zone capacity.

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>

Carlos Maiolino (2):
  common/zoned: enable passing a custom capacity
  xfs: Add test for mkfs with smaller zone capacity

 common/zoned      |  8 ++++++--
 tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
 tests/xfs/333.out |  2 ++
 3 files changed, 45 insertions(+), 2 deletions(-)
 create mode 100755 tests/xfs/333
 create mode 100644 tests/xfs/333.out

-- 
2.51.1

[PATCH 1/2] common/zoned: enable passing a custom capacity

[cem]

From: Carlos Maiolino <cem@kernel.org>

Extend _create_zloop() to accept a custom zone capacity.

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
---
 common/zoned | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/common/zoned b/common/zoned
index 88b81de5db4d..51c011b247d2 100644
--- a/common/zoned
+++ b/common/zoned
@@ -61,7 +61,7 @@ _find_next_zloop()
 }
 
 # Create a zloop device
-# usage: _create_zloop <base_dir> <zone_size> <nr_conv_zones>
+# usage: _create_zloop <base_dir> <zone_size> <nr_conv_zones> <zone_capacity>
 _create_zloop()
 {
     local id="$(_find_next_zloop)"
@@ -80,9 +80,13 @@ _create_zloop()
         local conv_zones=",conv_zones=$3"
     fi
 
+    if [ -n "$4" ]; then
+	local zone_capacity=",zone_capacity_mb=$4"
+    fi
+
     mkdir -p "$zloop_base/$id"
 
-    local zloop_args="add id=$id,base_dir=$zloop_base$zone_size$conv_zones"
+    local zloop_args="add id=$id,base_dir=$zloop_base$zone_size$conv_zones$zone_capacity"
 
     echo "$zloop_args" > /dev/zloop-control || \
         _fail "cannot create zloop device"
-- 
2.51.1

[PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[cem]

From: Carlos Maiolino <cem@kernel.org>

Add a regression test for initializing zoned block devices with
sequential zones with a capacity smaller than the conventional
zones capacity.

Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
---
 tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
 tests/xfs/333.out |  2 ++
 2 files changed, 39 insertions(+)
 create mode 100755 tests/xfs/333
 create mode 100644 tests/xfs/333.out

diff --git a/tests/xfs/333 b/tests/xfs/333
new file mode 100755
index 000000000000..f045b13c73ee
--- /dev/null
+++ b/tests/xfs/333
@@ -0,0 +1,37 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2025 Red Hat, Inc.  All Rights Reserved.
+#
+# FS QA Test 333
+#
+# Test that mkfs can properly initialize zoned devices
+# with a sequential zone capacity smaller than the conventional zone.
+#
+. ./common/preamble
+. ./common/zoned
+
+_begin_fstest auto zone mkfs quick
+_cleanup()
+{
+	_destroy_zloop $zloop
+}
+
+_require_scratch
+_require_zloop
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount >> $seqres.full
+
+zloopdir="$SCRATCH_MNT/zloop"
+zone_size=64
+conv_zones=2
+zone_capacity=63
+
+zloop=$(_create_zloop $zloopdir $zone_size $conv_zones $zone_capacity)
+
+_try_mkfs_dev $zloop >> $seqres.full 2>&1 || \
+	_fail "Cannot mkfs zoned filesystem"
+
+echo Silence is golden
+# success, all done
+_exit 0
diff --git a/tests/xfs/333.out b/tests/xfs/333.out
new file mode 100644
index 000000000000..60a158987a22
--- /dev/null
+++ b/tests/xfs/333.out
@@ -0,0 +1,2 @@
+QA output created by 333
+Silence is golden
-- 
2.51.1

Re: [PATCH 1/2] common/zoned: enable passing a custom capacity

[Johannes Thumshirn]

Looks good,

Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>

Re: [PATCH 1/2] common/zoned: enable passing a custom capacity

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Christoph Hellwig]

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Johannes Thumshirn]

On 11/20/25 5:09 PM, cem@kernel.org wrote:
> From: Carlos Maiolino <cem@kernel.org>
>
> Add a regression test for initializing zoned block devices with
> sequential zones with a capacity smaller than the conventional
> zones capacity.


Hi Carlos,

Two quick questions:

1) Is there a specific reason this is a xfs only test? I think checking 
this on btrfs and f2fs would make sense as well, like with generic/781.

2)  I would also mount the FS and perform some IO on it.



> Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> ---
>   tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
>   tests/xfs/333.out |  2 ++
>   2 files changed, 39 insertions(+)
>   create mode 100755 tests/xfs/333
>   create mode 100644 tests/xfs/333.out
>
> diff --git a/tests/xfs/333 b/tests/xfs/333
> new file mode 100755
> index 000000000000..f045b13c73ee
> --- /dev/null
> +++ b/tests/xfs/333
> @@ -0,0 +1,37 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2025 Red Hat, Inc.  All Rights Reserved.
> +#
> +# FS QA Test 333
> +#
> +# Test that mkfs can properly initialize zoned devices
> +# with a sequential zone capacity smaller than the conventional zone.
> +#
> +. ./common/preamble
> +. ./common/zoned
> +
> +_begin_fstest auto zone mkfs quick
> +_cleanup()
> +{
> +	_destroy_zloop $zloop
> +}
> +
> +_require_scratch
> +_require_zloop
> +
> +_scratch_mkfs > /dev/null 2>&1
> +_scratch_mount >> $seqres.full
> +
> +zloopdir="$SCRATCH_MNT/zloop"
> +zone_size=64
> +conv_zones=2
> +zone_capacity=63
> +
> +zloop=$(_create_zloop $zloopdir $zone_size $conv_zones $zone_capacity)
> +
> +_try_mkfs_dev $zloop >> $seqres.full 2>&1 || \
> +	_fail "Cannot mkfs zoned filesystem"
> +
> +echo Silence is golden
> +# success, all done
> +_exit 0
> diff --git a/tests/xfs/333.out b/tests/xfs/333.out
> new file mode 100644
> index 000000000000..60a158987a22
> --- /dev/null
> +++ b/tests/xfs/333.out
> @@ -0,0 +1,2 @@
> +QA output created by 333
> +Silence is golden

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[hch]

On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
> On 11/20/25 5:09 PM, cem@kernel.org wrote:
> > From: Carlos Maiolino <cem@kernel.org>
> >
> > Add a regression test for initializing zoned block devices with
> > sequential zones with a capacity smaller than the conventional
> > zones capacity.
> 
> 
> Hi Carlos,
> 
> Two quick questions:
> 
> 1) Is there a specific reason this is a xfs only test? I think checking 
> this on btrfs and f2fs would make sense as well, like with generic/781.

Didn't f2fs drop zone_capacity < zone_size support because they only
care about their android out of tree use case?

But otherwise I'd agree.

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Johannes Thumshirn]

On 11/21/25 7:57 AM, hch wrote:
> On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
>> On 11/20/25 5:09 PM, cem@kernel.org wrote:
>>> From: Carlos Maiolino <cem@kernel.org>
>>>
>>> Add a regression test for initializing zoned block devices with
>>> sequential zones with a capacity smaller than the conventional
>>> zones capacity.
>>
>> Hi Carlos,
>>
>> Two quick questions:
>>
>> 1) Is there a specific reason this is a xfs only test? I think checking
>> this on btrfs and f2fs would make sense as well, like with generic/781.
> Didn't f2fs drop zone_capacity < zone_size support because they only
> care about their android out of tree use case?
>
> But otherwise I'd agree.
>
That one I actually don't know

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Hans Holmberg]

On 21/11/2025 08:00, Johannes Thumshirn wrote:
> On 11/21/25 7:57 AM, hch wrote:
>> On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
>>> On 11/20/25 5:09 PM, cem@kernel.org wrote:
>>>> From: Carlos Maiolino <cem@kernel.org>
>>>>
>>>> Add a regression test for initializing zoned block devices with
>>>> sequential zones with a capacity smaller than the conventional
>>>> zones capacity.
>>>
>>> Hi Carlos,
>>>
>>> Two quick questions:
>>>
>>> 1) Is there a specific reason this is a xfs only test? I think checking
>>> this on btrfs and f2fs would make sense as well, like with generic/781.
>> Didn't f2fs drop zone_capacity < zone_size support because they only
>> care about their android out of tree use case?
>>
>> But otherwise I'd agree.
>>
> That one I actually don't know
> 
> 
+ Bart

I've been benchmarking f2fs with zone capacity < zone size lately on SSD
devices, so i'd be surprised if that has stopped working (but that requires
a separate blockdev for metadata (e.g. a separate name space)

Single-zoned-device f2fs file systems also works:

mkfs.f2fs -m /dev/<host managed smr disk> 
(but that has zone_size==cone_capacity)

..so i think this should be applicable for f2fs as well

Re: [PATCH 1/2] common/zoned: enable passing a custom capacity

[Zorro Lang]

On Thu, Nov 20, 2025 at 05:08:29PM +0100, cem@kernel.org wrote:
> From: Carlos Maiolino <cem@kernel.org>
> 
> Extend _create_zloop() to accept a custom zone capacity.
> 
> Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> ---
>  common/zoned | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/common/zoned b/common/zoned
> index 88b81de5db4d..51c011b247d2 100644
> --- a/common/zoned
> +++ b/common/zoned
> @@ -61,7 +61,7 @@ _find_next_zloop()
>  }
>  
>  # Create a zloop device
> -# usage: _create_zloop <base_dir> <zone_size> <nr_conv_zones>
> +# usage: _create_zloop <base_dir> <zone_size> <nr_conv_zones> <zone_capacity>

Actually from the logic of _create_zloop, these 4 arguments are optional, not
necessary, so the "<...>" should be "[...]"

>  _create_zloop()
>  {
>      local id="$(_find_next_zloop)"
> @@ -80,9 +80,13 @@ _create_zloop()
>          local conv_zones=",conv_zones=$3"
>      fi
>  
> +    if [ -n "$4" ]; then
> +	local zone_capacity=",zone_capacity_mb=$4"
> +    fi
> +
>      mkdir -p "$zloop_base/$id"
>  
> -    local zloop_args="add id=$id,base_dir=$zloop_base$zone_size$conv_zones"
> +    local zloop_args="add id=$id,base_dir=$zloop_base$zone_size$conv_zones$zone_capacity"

If the "zone_capacity" isn't a new feature which need a _require_ helper to check
it's supported, then this patch looks good to me.

Reviewed-by: Zorro Lang <zlang@redhat.com>

>  
>      echo "$zloop_args" > /dev/zloop-control || \
>          _fail "cannot create zloop device"
> -- 
> 2.51.1
> 

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Zorro Lang]

On Thu, Nov 20, 2025 at 05:08:30PM +0100, cem@kernel.org wrote:
> From: Carlos Maiolino <cem@kernel.org>
> 
> Add a regression test for initializing zoned block devices with
> sequential zones with a capacity smaller than the conventional
> zones capacity.
> 
> Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> ---
>  tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
>  tests/xfs/333.out |  2 ++
>  2 files changed, 39 insertions(+)
>  create mode 100755 tests/xfs/333
>  create mode 100644 tests/xfs/333.out
> 
> diff --git a/tests/xfs/333 b/tests/xfs/333
> new file mode 100755
> index 000000000000..f045b13c73ee
> --- /dev/null
> +++ b/tests/xfs/333
> @@ -0,0 +1,37 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2025 Red Hat, Inc.  All Rights Reserved.
> +#
> +# FS QA Test 333
> +#
> +# Test that mkfs can properly initialize zoned devices
> +# with a sequential zone capacity smaller than the conventional zone.
> +#
> +. ./common/preamble
> +. ./common/zoned
> +
> +_begin_fstest auto zone mkfs quick
> +_cleanup()
> +{
> +	_destroy_zloop $zloop

        cd /
        rm -r -f $tmp.*

> +}
> +
> +_require_scratch

_require_block_device $SCRATCH_DEV

> +_require_zloop

g/781 checks if current $FSTYP supports zoned filesystem, and _notrun if it's
not supported:

        _try_mkfs_dev $zloop >> $seqres.full 2>&1 ||\
                _notrun "cannot mkfs zoned filesystem"

I'm wondering if we could have a common helper for that, then this case can be
a generic test case too.

For example:

_require_zloop_filesystem()
{
        _require_zloop

        local zloopdir="$TEST_DIR/zloop_test"
        local zloop=$(_create_zloop $zloopdir 64 2)

	_try_mkfs_dev $zloop >/dev/null 2>&1 || \
		_notrun "cannot make $FSTYP on zloop"
	_destroy_zloop $zloop
}

But this method takes too much time to run, does anyone have a better idea to help it
to be finished in several seconds?


> +
> +_scratch_mkfs > /dev/null 2>&1
> +_scratch_mount >> $seqres.full
> +
> +zloopdir="$SCRATCH_MNT/zloop"
> +zone_size=64
> +conv_zones=2
> +zone_capacity=63

Better to add a comment to explain what are these numbers for.

> +
> +zloop=$(_create_zloop $zloopdir $zone_size $conv_zones $zone_capacity)
> +
> +_try_mkfs_dev $zloop >> $seqres.full 2>&1 || \
> +	_fail "Cannot mkfs zoned filesystem"
> +
> +echo Silence is golden

Is this done? If such zloop device can be created, should we expect it works as usual?

Thanks,
Zorro

> +# success, all done
> +_exit 0
> diff --git a/tests/xfs/333.out b/tests/xfs/333.out
> new file mode 100644
> index 000000000000..60a158987a22
> --- /dev/null
> +++ b/tests/xfs/333.out
> @@ -0,0 +1,2 @@
> +QA output created by 333
> +Silence is golden
> -- 
> 2.51.1
> 

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Carlos Maiolino]

On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
> On 11/20/25 5:09 PM, cem@kernel.org wrote:
> > From: Carlos Maiolino <cem@kernel.org>
> >
> > Add a regression test for initializing zoned block devices with
> > sequential zones with a capacity smaller than the conventional
> > zones capacity.
> 
> 
> Hi Carlos,
> 
> Two quick questions:
> 
> 1) Is there a specific reason this is a xfs only test? I think checking
> this on btrfs and f2fs would make sense as well, like with generic/781.

I wrote this mostly as a regression test for xfs's mkfs, but yeah, I don't
think there is any reason for this to be xfs-specific.

> 
> 2)  I would also mount the FS and perform some IO on it.

I'm not sure about this. Do you have any purpose in mind? This is
specifically to test mkfs is able to properly format the filesystem, not
to try the kernel module per-se.
One could argue that something 'could' go wrong in the mkfs that might
be found out only via some IO, but that would require much more than
just 'some IO'.

I do think a mount/unmount might add some value to the test, but I fail
to see why issuing a random amount of I/O would prove the correctness of
mkfs properly dealing with small capacities.

Cheers.
Carlos

> 
> 
> 
> > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > ---
> >   tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
> >   tests/xfs/333.out |  2 ++
> >   2 files changed, 39 insertions(+)
> >   create mode 100755 tests/xfs/333
> >   create mode 100644 tests/xfs/333.out
> >
> > diff --git a/tests/xfs/333 b/tests/xfs/333
> > new file mode 100755
> > index 000000000000..f045b13c73ee
> > --- /dev/null
> > +++ b/tests/xfs/333
> > @@ -0,0 +1,37 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2025 Red Hat, Inc.  All Rights Reserved.
> > +#
> > +# FS QA Test 333
> > +#
> > +# Test that mkfs can properly initialize zoned devices
> > +# with a sequential zone capacity smaller than the conventional zone.
> > +#
> > +. ./common/preamble
> > +. ./common/zoned
> > +
> > +_begin_fstest auto zone mkfs quick
> > +_cleanup()
> > +{
> > +	_destroy_zloop $zloop
> > +}
> > +
> > +_require_scratch
> > +_require_zloop
> > +
> > +_scratch_mkfs > /dev/null 2>&1
> > +_scratch_mount >> $seqres.full
> > +
> > +zloopdir="$SCRATCH_MNT/zloop"
> > +zone_size=64
> > +conv_zones=2
> > +zone_capacity=63
> > +
> > +zloop=$(_create_zloop $zloopdir $zone_size $conv_zones $zone_capacity)
> > +
> > +_try_mkfs_dev $zloop >> $seqres.full 2>&1 || \
> > +	_fail "Cannot mkfs zoned filesystem"
> > +
> > +echo Silence is golden
> > +# success, all done
> > +_exit 0
> > diff --git a/tests/xfs/333.out b/tests/xfs/333.out
> > new file mode 100644
> > index 000000000000..60a158987a22
> > --- /dev/null
> > +++ b/tests/xfs/333.out
> > @@ -0,0 +1,2 @@
> > +QA output created by 333
> > +Silence is golden
> 
> 

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Carlos Maiolino]

On Sat, Nov 22, 2025 at 04:36:46PM +0800, Zorro Lang wrote:
> On Thu, Nov 20, 2025 at 05:08:30PM +0100, cem@kernel.org wrote:
> > From: Carlos Maiolino <cem@kernel.org>
> >
> > Add a regression test for initializing zoned block devices with
> > sequential zones with a capacity smaller than the conventional
> > zones capacity.
> >
> > Signed-off-by: Carlos Maiolino <cmaiolino@redhat.com>
> > ---
> >  tests/xfs/333     | 37 +++++++++++++++++++++++++++++++++++++
> >  tests/xfs/333.out |  2 ++
> >  2 files changed, 39 insertions(+)
> >  create mode 100755 tests/xfs/333
> >  create mode 100644 tests/xfs/333.out
> >
> > diff --git a/tests/xfs/333 b/tests/xfs/333
> > new file mode 100755
> > index 000000000000..f045b13c73ee
> > --- /dev/null
> > +++ b/tests/xfs/333
> > @@ -0,0 +1,37 @@
> > +#! /bin/bash
> > +# SPDX-License-Identifier: GPL-2.0
> > +# Copyright (c) 2025 Red Hat, Inc.  All Rights Reserved.
> > +#
> > +# FS QA Test 333
> > +#
> > +# Test that mkfs can properly initialize zoned devices
> > +# with a sequential zone capacity smaller than the conventional zone.
> > +#
> > +. ./common/preamble
> > +. ./common/zoned
> > +
> > +_begin_fstest auto zone mkfs quick
> > +_cleanup()
> > +{
> > +	_destroy_zloop $zloop
> 
>         cd /
>         rm -r -f $tmp.*
> 
> > +}
> > +
> > +_require_scratch
> 
> _require_block_device $SCRATCH_DEV

I'll add to the V2.

> 
> > +_require_zloop
> 
> g/781 checks if current $FSTYP supports zoned filesystem, and _notrun if it's
> not supported:
> 
>         _try_mkfs_dev $zloop >> $seqres.full 2>&1 ||\
>                 _notrun "cannot mkfs zoned filesystem"
> 
> I'm wondering if we could have a common helper for that, then this case can be
> a generic test case too.
> 
> For example:
> 
> _require_zloop_filesystem()
> {
>         _require_zloop
> 
>         local zloopdir="$TEST_DIR/zloop_test"
>         local zloop=$(_create_zloop $zloopdir 64 2)
> 
> 	_try_mkfs_dev $zloop >/dev/null 2>&1 || \
> 		_notrun "cannot make $FSTYP on zloop"
> 	_destroy_zloop $zloop
> }
> 
> But this method takes too much time to run, does anyone have a better idea to help it
> to be finished in several seconds?
> 
> 
> > +
> > +_scratch_mkfs > /dev/null 2>&1
> > +_scratch_mount >> $seqres.full
> > +
> > +zloopdir="$SCRATCH_MNT/zloop"
> > +zone_size=64
> > +conv_zones=2
> > +zone_capacity=63
> 
> Better to add a comment to explain what are these numbers for.

Hmm, why? The variable names are very self-descriptive. I could add a
comment, but it's just redundant IMHO, ex:

zone_size=64		# Set zloop zone size to 64MiB

I added the variables instead of passing the numbers directly to
_create_zloop to avoid needing to go back and forth to _create_zloop
definition and make the meaning clear.
I'm fine adding the comments, I just don't see the point.


> 
> > +
> > +zloop=$(_create_zloop $zloopdir $zone_size $conv_zones $zone_capacity)
> > +
> > +_try_mkfs_dev $zloop >> $seqres.full 2>&1 || \
> > +	_fail "Cannot mkfs zoned filesystem"
> > +
> > +echo Silence is golden
> 
> Is this done? If such zloop device can be created, should we expect it works as usual?

Yes, the whole point is to ensure mkfs can initialize the filesystem on
top of the zoned block device.

Cheers,
Carlos

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Johannes Thumshirn]

On 11/24/25 4:27 PM, Carlos Maiolino wrote:
> On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
>> On 11/20/25 5:09 PM, cem@kernel.org wrote:
>>> From: Carlos Maiolino <cem@kernel.org>
>>>
>>> Add a regression test for initializing zoned block devices with
>>> sequential zones with a capacity smaller than the conventional
>>> zones capacity.
>>
>> Hi Carlos,
>>
>> Two quick questions:
>>
>> 1) Is there a specific reason this is a xfs only test? I think checking
>> this on btrfs and f2fs would make sense as well, like with generic/781.
> I wrote this mostly as a regression test for xfs's mkfs, but yeah, I don't
> think there is any reason for this to be xfs-specific.
>
>> 2)  I would also mount the FS and perform some IO on it.
> I'm not sure about this. Do you have any purpose in mind? This is
> specifically to test mkfs is able to properly format the filesystem, not
> to try the kernel module per-se.
> One could argue that something 'could' go wrong in the mkfs that might
> be found out only via some IO, but that would require much more than
> just 'some IO'.


Yep that's what I had in mind.


> I do think a mount/unmount might add some value to the test, but I fail
> to see why issuing a random amount of I/O would prove the correctness of
> mkfs properly dealing with small capacities.


fstests does a fsck after each test, doesn't it? So that should be 
sufficient as well.

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[hch]

On Mon, Nov 24, 2025 at 05:31:03PM +0000, Johannes Thumshirn wrote:
> > I do think a mount/unmount might add some value to the test, but I fail
> > to see why issuing a random amount of I/O would prove the correctness of
> > mkfs properly dealing with small capacities.
> 
> 
> fstests does a fsck after each test, doesn't it? So that should be 
> sufficient as well.

But only for well defined devices, not those made up from thin
air.  So we'd need to do that manually here.

Re: [PATCH 2/2] xfs: Add test for mkfs with smaller zone capacity

[Carlos Maiolino]

On Mon, Nov 24, 2025 at 05:31:03PM +0000, Johannes Thumshirn wrote:
> On 11/24/25 4:27 PM, Carlos Maiolino wrote:
> > On Fri, Nov 21, 2025 at 06:51:31AM +0000, Johannes Thumshirn wrote:
> >> On 11/20/25 5:09 PM, cem@kernel.org wrote:
> >>> From: Carlos Maiolino <cem@kernel.org>
> >>>
> >>> Add a regression test for initializing zoned block devices with
> >>> sequential zones with a capacity smaller than the conventional
> >>> zones capacity.
> >>
> >> Hi Carlos,
> >>
> >> Two quick questions:
> >>
> >> 1) Is there a specific reason this is a xfs only test? I think checking
> >> this on btrfs and f2fs would make sense as well, like with generic/781.
> > I wrote this mostly as a regression test for xfs's mkfs, but yeah, I don't
> > think there is any reason for this to be xfs-specific.
> >
> >> 2)  I would also mount the FS and perform some IO on it.
> > I'm not sure about this. Do you have any purpose in mind? This is
> > specifically to test mkfs is able to properly format the filesystem, not
> > to try the kernel module per-se.
> > One could argue that something 'could' go wrong in the mkfs that might
> > be found out only via some IO, but that would require much more than
> > just 'some IO'.
> 
> 
> Yep that's what I had in mind.
> 
> 
> > I do think a mount/unmount might add some value to the test, but I fail
> > to see why issuing a random amount of I/O would prove the correctness of
> > mkfs properly dealing with small capacities.
> 
> 
> fstests does a fsck after each test, doesn't it? So that should be
> sufficient as well.
> 

I particularly don't agree with mounting/unmounting the FS here, that
extends the test to more than it is supposed to test, i.e. ensuring mkfs
can initialize the filesystem.
Although I agree running fsck on it to certify nothing obviously wrong
happened with mkfs.

I think I've got another bug though which I started to investigate a few
minutes ago. Reason why I think we should split mkfs and mount/unmount
testing into two different tests.

Anyway, I'll see what's going on an update this test appropriately for a
new version.

Thanks for the reviews.