[RFC 0/1] ext4/054: Should we remove auto and quick group?

[RFC 0/1] ext4/054: Should we remove auto and quick group?

[Ritesh Harjani]

Hello Zhang/Ted,

Looks like the issue fixed by patches at [1], were observed with fault injection
testing and with errors=continue mount option. But were not cc'd to stable.

Do you think those should be cc'd to stable tree?

Meanwhile, I was thinking we should anyway remove auto and quick group from this
test as it could trigger a bug on in older kernel targets. Thoughts?


[1]: https://lore.kernel.org/all/20210908120850.4012324-1-yi.zhang@huawei.com/

Ritesh Harjani (1):
  ext4/054: Remove auto and quick group

 tests/ext4/054 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--
2.31.1

[RFC 1/1] ext4/054: Remove auto and quick group

[Ritesh Harjani]

It seems this test creates a crafted corrupted image by modifying ext4
extent block structure of an inode to test some ext4 extent consistency
fixes done at [1].
This IMO, should not be in auto and quick group, since it could cause BUG_ON()
and happens only with some crafted corrupted image (or with fault injection
testing with errors=continue mount option).

[1]: https://lore.kernel.org/all/20210908120850.4012324-1-yi.zhang@huawei.com/
Signed-off-by: Ritesh Harjani <riteshh@linux.ibm.com>
---
 tests/ext4/054 | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/ext4/054 b/tests/ext4/054
index 9a11719f..21fa4e0a 100755
--- a/tests/ext4/054
+++ b/tests/ext4/054
@@ -12,7 +12,7 @@
 #    ext4_valid_extent_entries())

 . ./common/preamble
-_begin_fstest auto quick dangerous_fuzzers
+_begin_fstest dangerous_fuzzers

 # Import common functions
 . ./common/filter
--
2.31.1

Re: [RFC 0/1] ext4/054: Should we remove auto and quick group?

[Eryu Guan]

On Tue, Jan 25, 2022 at 11:32:01AM +0530, Ritesh Harjani wrote:
> Hello Zhang/Ted,
> 
> Looks like the issue fixed by patches at [1], were observed with fault injection
> testing and with errors=continue mount option. But were not cc'd to stable.
> 
> Do you think those should be cc'd to stable tree?
> 
> Meanwhile, I was thinking we should anyway remove auto and quick group from this
> test as it could trigger a bug on in older kernel targets. Thoughts?

IMO, ext4/054 is a targeted regression test and should be in auto group,
which ensures the bug doesn't get re-introduced in future.

I think you could just skip this test to fit your kernel version, e.g.

echo ext4/054 > ext4.exclude
./check -X ext4.exclude

Thanks,
Eryu

> 
> 
> [1]: https://lore.kernel.org/all/20210908120850.4012324-1-yi.zhang@huawei.com/
> 
> Ritesh Harjani (1):
>   ext4/054: Remove auto and quick group
> 
>  tests/ext4/054 | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> --
> 2.31.1

Re: [RFC 0/1] ext4/054: Should we remove auto and quick group?

[Theodore Y. Ts'o]

On Tue, Jan 25, 2022 at 11:32:01AM +0530, Ritesh Harjani wrote:
> Hello Zhang/Ted,
> 
> Looks like the issue fixed by patches at [1], were observed with fault injection
> testing and with errors=continue mount option. But were not cc'd to stable.
> 
> Do you think those should be cc'd to stable tree?

I already requested that they be backported, and they are in 5.10.89+
and 5.15.12+.  Unfortunately the patches don't backport cleanly into
5.4, and while I did the manual backport for 5.10, I haven't gotten
around to backporting them into 5.4 or older kernels.

       	  	      	   	       - Ted

Re: [RFC 0/1] ext4/054: Should we remove auto and quick group?

[Ritesh Harjani]

On 22/01/25 03:08PM, Theodore Y. Ts'o wrote:
> On Tue, Jan 25, 2022 at 11:32:01AM +0530, Ritesh Harjani wrote:
> > Hello Zhang/Ted,
> >
> > Looks like the issue fixed by patches at [1], were observed with fault injection
> > testing and with errors=continue mount option. But were not cc'd to stable.
> >
> > Do you think those should be cc'd to stable tree?
>
> I already requested that they be backported, and they are in 5.10.89+
> and 5.15.12+.  Unfortunately the patches don't backport cleanly into
> 5.4, and while I did the manual backport for 5.10, I haven't gotten
> around to backporting them into 5.4 or older kernels.
>

Sure Ted, thanks a lot for the backport and for providing above information.

-ritesh

Re: [RFC 0/1] ext4/054: Should we remove auto and quick group?

[Ritesh Harjani]

On 22/01/25 03:43PM, Eryu Guan wrote:
> On Tue, Jan 25, 2022 at 11:32:01AM +0530, Ritesh Harjani wrote:
> > Hello Zhang/Ted,
> >
> > Looks like the issue fixed by patches at [1], were observed with fault injection
> > testing and with errors=continue mount option. But were not cc'd to stable.
> >
> > Do you think those should be cc'd to stable tree?
> >
> > Meanwhile, I was thinking we should anyway remove auto and quick group from this
> > test as it could trigger a bug on in older kernel targets. Thoughts?
>
> IMO, ext4/054 is a targeted regression test and should be in auto group,
> which ensures the bug doesn't get re-introduced in future.

Yes, I agree with it.

>
> I think you could just skip this test to fit your kernel version, e.g.
>
> echo ext4/054 > ext4.exclude
> ./check -X ext4.exclude

Sure, thanks Eryu.

-ritesh