[LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Petr Vorel]

tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
trivial support for direct IO"). There is no reason for tmpfs to reject
direct IO of any size therefore follow the approach already used for
Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
and 14 (read, write with non-aligned buffer).

Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Petr Vorel <pvorel@suse.cz>
---
 testcases/kernel/io/direct_io/diotest4.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/testcases/kernel/io/direct_io/diotest4.c b/testcases/kernel/io/direct_io/diotest4.c
index 45c677b5b..ad00fa3e0 100644
--- a/testcases/kernel/io/direct_io/diotest4.c
+++ b/testcases/kernel/io/direct_io/diotest4.c
@@ -270,6 +270,7 @@ int main(int argc, char *argv[])
 	case TST_NFS_MAGIC:
 	case TST_BTRFS_MAGIC:
 	case TST_FUSE_MAGIC:
+	case TST_TMPFS_MAGIC:
 		tst_resm(TCONF, "%s supports odd count IO",
 			 tst_fs_type_name(fs_type));
 	break;
@@ -443,6 +444,7 @@ int main(int argc, char *argv[])
 	case TST_NFS_MAGIC:
 	case TST_BTRFS_MAGIC:
 	case TST_FUSE_MAGIC:
+	case TST_TMPFS_MAGIC:
 		tst_resm(TCONF, "%s supports non-aligned buffer",
 			 tst_fs_type_name(fs_type));
 	break;
-- 
2.42.0


-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Hugh Dickins via ltp]

On Wed, 22 Nov 2023, Petr Vorel wrote:

> tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
> trivial support for direct IO"). There is no reason for tmpfs to reject
> direct IO of any size therefore follow the approach already used for
> Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
> and 14 (read, write with non-aligned buffer).
> 
> Suggested-by: Jan Kara <jack@suse.cz>
> Signed-off-by: Petr Vorel <pvorel@suse.cz>

Acked-by: Hugh Dickins <hughd@google.com>

Great, thanks a lot. I've never minded tmpfs failing those LTP tests,
but had never looked at them closely enough to notice that there is
this established way to avoid the failures...

... but hold on: I can't look at the LTP source at the moment,
but records tell me that the two tests failing were dio04 and dio10,
whereas you say test 3 and test 14 - please clarify or correct, thanks.

> ---
>  testcases/kernel/io/direct_io/diotest4.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/testcases/kernel/io/direct_io/diotest4.c b/testcases/kernel/io/direct_io/diotest4.c
> index 45c677b5b..ad00fa3e0 100644
> --- a/testcases/kernel/io/direct_io/diotest4.c
> +++ b/testcases/kernel/io/direct_io/diotest4.c
> @@ -270,6 +270,7 @@ int main(int argc, char *argv[])
>  	case TST_NFS_MAGIC:
>  	case TST_BTRFS_MAGIC:
>  	case TST_FUSE_MAGIC:
> +	case TST_TMPFS_MAGIC:
>  		tst_resm(TCONF, "%s supports odd count IO",
>  			 tst_fs_type_name(fs_type));
>  	break;
> @@ -443,6 +444,7 @@ int main(int argc, char *argv[])
>  	case TST_NFS_MAGIC:
>  	case TST_BTRFS_MAGIC:
>  	case TST_FUSE_MAGIC:
> +	case TST_TMPFS_MAGIC:
>  		tst_resm(TCONF, "%s supports non-aligned buffer",
>  			 tst_fs_type_name(fs_type));
>  	break;
> -- 
> 2.42.0

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Petr Vorel]

> On Wed, 22 Nov 2023, Petr Vorel wrote:

> > tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
> > trivial support for direct IO"). There is no reason for tmpfs to reject
> > direct IO of any size therefore follow the approach already used for
> > Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
> > and 14 (read, write with non-aligned buffer).

> > Suggested-by: Jan Kara <jack@suse.cz>
> > Signed-off-by: Petr Vorel <pvorel@suse.cz>

> Acked-by: Hugh Dickins <hughd@google.com>

> Great, thanks a lot. I've never minded tmpfs failing those LTP tests,
> but had never looked at them closely enough to notice that there is
> this established way to avoid the failures...

I was surprised as well, Jan Kara noticed.

> ... but hold on: I can't look at the LTP source at the moment,
> but records tell me that the two tests failing were dio04 and dio10,
> whereas you say test 3 and test 14 - please clarify or correct, thanks.

These 2 tests are both running diotest4.c, just with different block size:

$ git grep -e ^dio04 -e ^dio10 runtest/
runtest/dio:dio04 diotest4
runtest/dio:dio10 diotest4 -b 65536

NOTE: dio tests are currently being rewritten, but diotest4.c have not been done
yet. Thus the code is using legacy LTP API and needs cleanup.

Kind regards,
Petr

> > ---
> >  testcases/kernel/io/direct_io/diotest4.c | 2 ++
> >  1 file changed, 2 insertions(+)

> > diff --git a/testcases/kernel/io/direct_io/diotest4.c b/testcases/kernel/io/direct_io/diotest4.c
> > index 45c677b5b..ad00fa3e0 100644
> > --- a/testcases/kernel/io/direct_io/diotest4.c
> > +++ b/testcases/kernel/io/direct_io/diotest4.c
> > @@ -270,6 +270,7 @@ int main(int argc, char *argv[])
> >  	case TST_NFS_MAGIC:
> >  	case TST_BTRFS_MAGIC:
> >  	case TST_FUSE_MAGIC:
> > +	case TST_TMPFS_MAGIC:
> >  		tst_resm(TCONF, "%s supports odd count IO",
> >  			 tst_fs_type_name(fs_type));
> >  	break;
> > @@ -443,6 +444,7 @@ int main(int argc, char *argv[])
> >  	case TST_NFS_MAGIC:
> >  	case TST_BTRFS_MAGIC:
> >  	case TST_FUSE_MAGIC:
> > +	case TST_TMPFS_MAGIC:
> >  		tst_resm(TCONF, "%s supports non-aligned buffer",
> >  			 tst_fs_type_name(fs_type));
> >  	break;
> > -- 
> > 2.42.0

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Jan Kara]

On Wed 22-11-23 22:13:18, Petr Vorel wrote:
> tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
> trivial support for direct IO"). There is no reason for tmpfs to reject
> direct IO of any size therefore follow the approach already used for
> Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
> and 14 (read, write with non-aligned buffer).
> 
> Suggested-by: Jan Kara <jack@suse.cz>
> Signed-off-by: Petr Vorel <pvorel@suse.cz>

Looks good, thanks for writing this! Feel free to add:

Reviewed-by: Jan Kara <jack@suse.cz>

								Honza

> ---
>  testcases/kernel/io/direct_io/diotest4.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/testcases/kernel/io/direct_io/diotest4.c b/testcases/kernel/io/direct_io/diotest4.c
> index 45c677b5b..ad00fa3e0 100644
> --- a/testcases/kernel/io/direct_io/diotest4.c
> +++ b/testcases/kernel/io/direct_io/diotest4.c
> @@ -270,6 +270,7 @@ int main(int argc, char *argv[])
>  	case TST_NFS_MAGIC:
>  	case TST_BTRFS_MAGIC:
>  	case TST_FUSE_MAGIC:
> +	case TST_TMPFS_MAGIC:
>  		tst_resm(TCONF, "%s supports odd count IO",
>  			 tst_fs_type_name(fs_type));
>  	break;
> @@ -443,6 +444,7 @@ int main(int argc, char *argv[])
>  	case TST_NFS_MAGIC:
>  	case TST_BTRFS_MAGIC:
>  	case TST_FUSE_MAGIC:
> +	case TST_TMPFS_MAGIC:
>  		tst_resm(TCONF, "%s supports non-aligned buffer",
>  			 tst_fs_type_name(fs_type));
>  	break;
> -- 
> 2.42.0
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Christian Brauner]

On Wed, Nov 22, 2023 at 10:13:18PM +0100, Petr Vorel wrote:
> tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
> trivial support for direct IO"). There is no reason for tmpfs to reject
> direct IO of any size therefore follow the approach already used for
> Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
> and 14 (read, write with non-aligned buffer).
> 
> Suggested-by: Jan Kara <jack@suse.cz>
> Signed-off-by: Petr Vorel <pvorel@suse.cz>
> ---

Looks good to me,
Acked-by: Christian Brauner <brauner@kernel.org>

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp

Re: [LTP] [PATCH 1/1] diotest4: Skip test 3 and 14 for tmpfs

[Petr Vorel]

Hi Christian,

> On Wed, Nov 22, 2023 at 10:13:18PM +0100, Petr Vorel wrote:
> > tmpfs in kernel v6.6 got support for O_DIRECT in e88e0d366f9c ("tmpfs:
> > trivial support for direct IO"). There is no reason for tmpfs to reject
> > direct IO of any size therefore follow the approach already used for
> > Btrfs, NFS and FUSE: skipping test 3 (odd count of read and write)
> > and 14 (read, write with non-aligned buffer).

> > Suggested-by: Jan Kara <jack@suse.cz>
> > Signed-off-by: Petr Vorel <pvorel@suse.cz>
> > ---

> Looks good to me,
> Acked-by: Christian Brauner <brauner@kernel.org>

Thanks a lot for your ack. I'm sorry, already merged, but I was not able to
reply due planned mailserver outage.

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp