[PATCH] block/028: check if T10 verification fails

[PATCH] block/028: check if T10 verification fails

[Ming Lei]

When T10 verification fails, the error code of BLK_STS_PROTECTION
may not be propagated to user space, see mpage_end_io().

So seems the only reliable way for detecting the failure is to
check dmesg.

Cc: Martin K . Petersen <martin.petersen@oracle.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 tests/block/028 | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/tests/block/028 b/tests/block/028
index 7bee691c0515..a0ab9ff208db 100755
--- a/tests/block/028
+++ b/tests/block/028
@@ -38,6 +38,14 @@ test() {
 			test_pi "$dix" "$dif"
 			echo "Test(dix:$dix dif:$dif) complete"
 		done
+		if dmesg | grep -q "guard tag error at sector"; then
+			echo "Fail"
+			break
+		fi
+		if dmesg | grep -q "ref tag error at location"; then
+			echo "Fail"
+			break
+		fi
 	done
 
 	rm -f "$FULL"
-- 
2.9.5

Re: [PATCH] block/028: check if T10 verification fails

[Christoph Hellwig]

On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
> When T10 verification fails, the error code of BLK_STS_PROTECTION
> may not be propagated to user space, see mpage_end_io().

Well, we should probably fix that instead of papering over it..

Re: [PATCH] block/028: check if T10 verification fails

[Chaitanya Kulkarni]

Until we fix this, I'm okay to have this change in the blktest, if this 
is breaking any exiting behavior.

On 4/26/19 7:57 AM, Christoph Hellwig wrote:
> On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
>> When T10 verification fails, the error code of BLK_STS_PROTECTION
>> may not be propagated to user space, see mpage_end_io().
> 
> Well, we should probably fix that instead of papering over it..
> 

Re: [PATCH] block/028: check if T10 verification fails

[Ming Lei]

On Fri, Apr 26, 2019 at 07:57:11AM -0700, Christoph Hellwig wrote:
> On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
> > When T10 verification fails, the error code of BLK_STS_PROTECTION
> > may not be propagated to user space, see mpage_end_io().
> 
> Well, we should probably fix that instead of papering over it..

That might be a big change, and not sure when they can be fixed
completely.

This patch at least can expose the issue to user first.

Anyway, CC filesystem guys.

Thanks, 
Ming

Re: [PATCH] block/028: check if T10 verification fails

[Ming Lei]

On Fri, Apr 26, 2019 at 06:24:08PM +0000, Chaitanya Kulkarni wrote:
> Until we fix this, I'm okay to have this change in the blktest, if this 
> is breaking any exiting behavior.

This is a patch for enhancing/fixing block/028 test case, so that we
can't hide the kernel issue. Now when the issue is triggered, block/028
still shows 'passed'.

It is nothing do with fix in kernel side.

Even though it is fixed upstream, there may be other old kernel releases
which needs to expose the issue.

Even it is fixed completely, sometimes regression may come up again.


Thanks,
Ming

Re: [PATCH] block/028: check if T10 verification fails

[Darrick J. Wong]

On Sat, Apr 27, 2019 at 06:16:32AM +0800, Ming Lei wrote:
> On Fri, Apr 26, 2019 at 07:57:11AM -0700, Christoph Hellwig wrote:
> > On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
> > > When T10 verification fails, the error code of BLK_STS_PROTECTION
> > > may not be propagated to user space, see mpage_end_io().
> > 
> > Well, we should probably fix that instead of papering over it..
> 
> That might be a big change, and not sure when they can be fixed
> completely.
> 
> This patch at least can expose the issue to user first.

I don't see a patch anywhere ... ?

--D

> Anyway, CC filesystem guys.
> 
> Thanks, 
> Ming

Re: [PATCH] block/028: check if T10 verification fails

[Ming Lei]

On Fri, Apr 26, 2019 at 06:24:49PM -0700, Darrick J. Wong wrote:
> On Sat, Apr 27, 2019 at 06:16:32AM +0800, Ming Lei wrote:
> > On Fri, Apr 26, 2019 at 07:57:11AM -0700, Christoph Hellwig wrote:
> > > On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
> > > > When T10 verification fails, the error code of BLK_STS_PROTECTION
> > > > may not be propagated to user space, see mpage_end_io().
> > > 
> > > Well, we should probably fix that instead of papering over it..
> > 
> > That might be a big change, and not sure when they can be fixed
> > completely.
> > 
> > This patch at least can expose the issue to user first.
> 
> I don't see a patch anywhere ... ?

It is one blktest patch, please see the following link:

https://lore.kernel.org/linux-block/20190427012449.GC178347@magnolia/T/#t

Thanks,
Ming

Re: [PATCH] block/028: check if T10 verification fails

[Omar Sandoval]

On Fri, Apr 26, 2019 at 10:39:37AM +0800, Ming Lei wrote:
> When T10 verification fails, the error code of BLK_STS_PROTECTION
> may not be propagated to user space, see mpage_end_io().
> 
> So seems the only reliable way for detecting the failure is to
> check dmesg.
> 
> Cc: Martin K . Petersen <martin.petersen@oracle.com>
> Signed-off-by: Ming Lei <ming.lei@redhat.com>
> ---
>  tests/block/028 | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/tests/block/028 b/tests/block/028
> index 7bee691c0515..a0ab9ff208db 100755
> --- a/tests/block/028
> +++ b/tests/block/028
> @@ -38,6 +38,14 @@ test() {
>  			test_pi "$dix" "$dif"
>  			echo "Test(dix:$dix dif:$dif) complete"
>  		done
> +		if dmesg | grep -q "guard tag error at sector"; then
> +			echo "Fail"
> +			break
> +		fi
> +		if dmesg | grep -q "ref tag error at location"; then
> +			echo "Fail"
> +			break
> +		fi

This will also catch messages that didn't occur during the test run. Not
a huge deal, but I reworked it to use DMESG_FILTER instead. Thanks!