[PATCH] generic: require fibmap for testcases using filefrag

[PATCH] generic: require fibmap for testcases using filefrag

[Murphy Zhou]

To mute false alarm on NFS, which is reporting FIBMAP unsupported.

Signed-off-by: Murphy Zhou <jencce.kernel@gmail.com>
---
 tests/generic/516 | 1 +
 tests/generic/540 | 1 +
 tests/generic/541 | 1 +
 tests/generic/542 | 1 +
 tests/generic/543 | 1 +
 tests/generic/578 | 1 +
 6 files changed, 6 insertions(+)

diff --git a/tests/generic/516 b/tests/generic/516
index ceaa4151..906e38ec 100755
--- a/tests/generic/516
+++ b/tests/generic/516
@@ -31,6 +31,7 @@ _cleanup()
 # real QA test starts here
 _supported_os Linux
 _require_test_dedupe
+_require_fibmap
 
 rm -f $seqres.full
 
diff --git a/tests/generic/540 b/tests/generic/540
index dc4d9485..242ab805 100755
--- a/tests/generic/540
+++ b/tests/generic/540
@@ -40,6 +40,7 @@ _cleanup()
 _supported_os Linux
 _supported_fs generic
 _require_scratch_reflink
+_require_fibmap
 _require_xfs_io_command "falloc"
 
 rm -f $seqres.full
diff --git a/tests/generic/541 b/tests/generic/541
index fcb5d567..ec7ed52a 100755
--- a/tests/generic/541
+++ b/tests/generic/541
@@ -40,6 +40,7 @@ _cleanup()
 _supported_os Linux
 _supported_fs generic
 _require_scratch_reflink
+_require_fibmap
 _require_xfs_io_command "falloc"
 
 rm -f $seqres.full
diff --git a/tests/generic/542 b/tests/generic/542
index 62b32cb5..62fc55ea 100755
--- a/tests/generic/542
+++ b/tests/generic/542
@@ -40,6 +40,7 @@ _cleanup()
 _supported_os Linux
 _supported_fs generic
 _require_scratch_reflink
+_require_fibmap
 _require_xfs_io_command "falloc"
 
 rm -f $seqres.full
diff --git a/tests/generic/543 b/tests/generic/543
index cc7ee532..75c2f7ef 100755
--- a/tests/generic/543
+++ b/tests/generic/543
@@ -40,6 +40,7 @@ _cleanup()
 _supported_os Linux
 _supported_fs generic
 _require_scratch_reflink
+_require_fibmap
 _require_xfs_io_command "falloc"
 
 rm -f $seqres.full
diff --git a/tests/generic/578 b/tests/generic/578
index 798cccf0..6e160142 100755
--- a/tests/generic/578
+++ b/tests/generic/578
@@ -29,6 +29,7 @@ _cleanup()
 # real QA test starts here
 _supported_os Linux
 _supported_fs generic
+_require_fibmap
 _require_test_program "mmap-write-concurrent"
 _require_command "$FILEFRAG_PROG" filefrag
 _require_test_reflink
-- 
2.20.1

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Christoph Hellwig]

On Fri, Jan 17, 2020 at 04:38:24PM +0800, Murphy Zhou wrote:
> To mute false alarm on NFS, which is reporting FIBMAP unsupported.

How do these tests end up using FIBMAP?  We really should not use FIBMAP
in any tests except those exercising FIBMAP itself.

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Murphy Zhou]

On Fri, Jan 17, 2020 at 12:45:17AM -0800, Christoph Hellwig wrote:
> On Fri, Jan 17, 2020 at 04:38:24PM +0800, Murphy Zhou wrote:
> > To mute false alarm on NFS, which is reporting FIBMAP unsupported.
> 
> How do these tests end up using FIBMAP?  We really should not use FIBMAP
> in any tests except those exercising FIBMAP itself.
> 
Through filefrag(8)?

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Darrick J. Wong]

On Fri, Jan 17, 2020 at 06:02:30PM +0800, Murphy Zhou wrote:
> On Fri, Jan 17, 2020 at 12:45:17AM -0800, Christoph Hellwig wrote:
> > On Fri, Jan 17, 2020 at 04:38:24PM +0800, Murphy Zhou wrote:
> > > To mute false alarm on NFS, which is reporting FIBMAP unsupported.
> > 
> > How do these tests end up using FIBMAP?  We really should not use FIBMAP
> > in any tests except those exercising FIBMAP itself.
> > 
> Through filefrag(8)?

But do those tests require filefrag -B??

--D

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Murphy Zhou]

On Fri, Jan 17, 2020 at 04:47:00PM -0800, Darrick J. Wong wrote:
> On Fri, Jan 17, 2020 at 06:02:30PM +0800, Murphy Zhou wrote:
> > On Fri, Jan 17, 2020 at 12:45:17AM -0800, Christoph Hellwig wrote:
> > > On Fri, Jan 17, 2020 at 04:38:24PM +0800, Murphy Zhou wrote:
> > > > To mute false alarm on NFS, which is reporting FIBMAP unsupported.
> > > 
> > > How do these tests end up using FIBMAP?  We really should not use FIBMAP
> > > in any tests except those exercising FIBMAP itself.
> > > 
> > Through filefrag(8)?
> 
> But do those tests require filefrag -B??

Not really. But filefrag will fall back to use fibmap.

> 
> --D

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Christoph Hellwig]

On Sat, Jan 18, 2020 at 05:38:51PM +0800, Murphy Zhou wrote:
> > But do those tests require filefrag -B??
> 
> Not really. But filefrag will fall back to use fibmap.

But that means it requires either FIEMAP or FIBMAP, and your patch
will cause regressions for file systems only implementing FIEMAP.
So the fix isn't to ask for FIBMAP, but to implement FIEMAP in
any file system missing that support.

Re: [PATCH] generic: require fibmap for testcases using filefrag

[Murphy Zhou]

On Sun, Jan 19, 2020 at 02:02:52PM -0800, Christoph Hellwig wrote:
> On Sat, Jan 18, 2020 at 05:38:51PM +0800, Murphy Zhou wrote:
> > > But do those tests require filefrag -B??
> > 
> > Not really. But filefrag will fall back to use fibmap.
> 
> But that means it requires either FIEMAP or FIBMAP, and your patch
> will cause regressions for file systems only implementing FIEMAP.
> So the fix isn't to ask for FIBMAP, but to implement FIEMAP in
> any file system missing that support.

Hmm, you are right :-)