[PATCH] shared/011: run on all file system that support cgroup writeback

[PATCH] shared/011: run on all file system that support cgroup writeback

[Christoph Hellwig]

Run the cgroup writeback test on xfs, for which I've just posted
a patch to support cgroup writeback as well as ext2 and f2fs, which
have supported cgroup writeback for a while now.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 tests/shared/011 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/shared/011 b/tests/shared/011
index a0ac375d..96ce9d1c 100755
--- a/tests/shared/011
+++ b/tests/shared/011
@@ -39,7 +39,7 @@ rm -f $seqres.full
 # real QA test starts here
 
 # Modify as appropriate.
-_supported_fs ext4 btrfs
+_supported_fs ext2 ext4 f2fs btrfs xfs
 _supported_os Linux
 _require_scratch
 _require_cgroup2 io
-- 
2.20.1

Re: [PATCH] shared/011: run on all file system that support cgroup writeback

[Theodore Ts'o]

On Mon, Jun 24, 2019 at 03:44:07PM +0200, Christoph Hellwig wrote:
> Run the cgroup writeback test on xfs, for which I've just posted
> a patch to support cgroup writeback as well as ext2 and f2fs, which
> have supported cgroup writeback for a while now.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  tests/shared/011 | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tests/shared/011 b/tests/shared/011
> index a0ac375d..96ce9d1c 100755
> --- a/tests/shared/011
> +++ b/tests/shared/011
> @@ -39,7 +39,7 @@ rm -f $seqres.full
>  # real QA test starts here
>  
>  # Modify as appropriate.
> -_supported_fs ext4 btrfs
> +_supported_fs ext2 ext4 f2fs btrfs xfs

Per my comments in another e-mail thread, given how many of the
primary file systems support cgroup-aware writeback, maybe we should
just remove the _supported_fs line and move this test to generic?

Whether we like it or not, there are more and more userspace tools
which assume that cgroup-aware writeback is a thing.

Alternatively, maybe we should have some standardized way so the
kernel can signal whether or not a particular mounted file system
supports cgroup-aware writeback?

						- Ted

Re: [PATCH] shared/011: run on all file system that support cgroup writeback

[Darrick J. Wong]

On Mon, Jun 24, 2019 at 11:08:39AM -0400, Theodore Ts'o wrote:
> On Mon, Jun 24, 2019 at 03:44:07PM +0200, Christoph Hellwig wrote:
> > Run the cgroup writeback test on xfs, for which I've just posted
> > a patch to support cgroup writeback as well as ext2 and f2fs, which
> > have supported cgroup writeback for a while now.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  tests/shared/011 | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/tests/shared/011 b/tests/shared/011
> > index a0ac375d..96ce9d1c 100755
> > --- a/tests/shared/011
> > +++ b/tests/shared/011
> > @@ -39,7 +39,7 @@ rm -f $seqres.full
> >  # real QA test starts here
> >  
> >  # Modify as appropriate.
> > -_supported_fs ext4 btrfs
> > +_supported_fs ext2 ext4 f2fs btrfs xfs
> 
> Per my comments in another e-mail thread, given how many of the
> primary file systems support cgroup-aware writeback, maybe we should
> just remove the _supported_fs line and move this test to generic?
> 
> Whether we like it or not, there are more and more userspace tools
> which assume that cgroup-aware writeback is a thing.
> 
> Alternatively, maybe we should have some standardized way so the
> kernel can signal whether or not a particular mounted file system
> supports cgroup-aware writeback?

I prefer this second option because I'd rather the test suite do the
work to figure out if cgroup aware writeback is enabled and therefore
worth testing rather than making everyone's QA department to add another
conditional known-failure entry for when they want to run new fstests on
some old LTS/distro kernel.

--D

> 						- Ted

Re: [PATCH] shared/011: run on all file system that support cgroup writeback

[Eryu Guan]

On Mon, Jun 24, 2019 at 08:19:10AM -0700, Darrick J. Wong wrote:
> On Mon, Jun 24, 2019 at 11:08:39AM -0400, Theodore Ts'o wrote:
> > On Mon, Jun 24, 2019 at 03:44:07PM +0200, Christoph Hellwig wrote:
> > > Run the cgroup writeback test on xfs, for which I've just posted
> > > a patch to support cgroup writeback as well as ext2 and f2fs, which
> > > have supported cgroup writeback for a while now.
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > ---
> > >  tests/shared/011 | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/tests/shared/011 b/tests/shared/011
> > > index a0ac375d..96ce9d1c 100755
> > > --- a/tests/shared/011
> > > +++ b/tests/shared/011
> > > @@ -39,7 +39,7 @@ rm -f $seqres.full
> > >  # real QA test starts here
> > >  
> > >  # Modify as appropriate.
> > > -_supported_fs ext4 btrfs
> > > +_supported_fs ext2 ext4 f2fs btrfs xfs
> > 
> > Per my comments in another e-mail thread, given how many of the
> > primary file systems support cgroup-aware writeback, maybe we should
> > just remove the _supported_fs line and move this test to generic?
> > 
> > Whether we like it or not, there are more and more userspace tools
> > which assume that cgroup-aware writeback is a thing.
> > 
> > Alternatively, maybe we should have some standardized way so the
> > kernel can signal whether or not a particular mounted file system
> > supports cgroup-aware writeback?
> 
> I prefer this second option because I'd rather the test suite do the
> work to figure out if cgroup aware writeback is enabled and therefore
> worth testing rather than making everyone's QA department to add another
> conditional known-failure entry for when they want to run new fstests on
> some old LTS/distro kernel.

Agreed, a standard way to query cgroup-aware writeback support status
would be the best.

Thanks,
Eryu

> 
> --D
> 
> > 						- Ted

Re: [PATCH] shared/011: run on all file system that support cgroup writeback

[Christoph Hellwig]

On Mon, Jun 24, 2019 at 11:08:39AM -0400, Theodore Ts'o wrote:
> Per my comments in another e-mail thread, given how many of the
> primary file systems support cgroup-aware writeback, maybe we should
> just remove the _supported_fs line and move this test to generic?
> 
> Whether we like it or not, there are more and more userspace tools
> which assume that cgroup-aware writeback is a thing.
> 
> Alternatively, maybe we should have some standardized way so the
> kernel can signal whether or not a particular mounted file system
> supports cgroup-aware writeback?

I'm fine with moving the patch to generic as said.  I'm fine saying
we assume cgroup wb support for block device based file systems,
xfs was the only major one missing, and the support is pretty trivial
as well.