Re: [PATCH] ceph: check negative offsets on ceph_llseek()

[Ilya Dryomov]

On Thu, Aug 17, 2017 at 8:40 PM, Luis Henriques <lhenriques@xxxxxxxx> wrote:
> Ilya Dryomov <idryomov@xxxxxxxxx> writes:
>
>> On Thu, Aug 17, 2017 at 3:45 PM, Luis Henriques <lhenriques@xxxxxxxx> wrote:
>>> Luis Henriques <lhenriques@xxxxxxxx> writes:
>>>
>>>> When a user requests SEEK_HOLE or SEEK_DATA with a negative offset
>>>> ceph_llseek should return -ENXIO.  Currently -EINVAL is being returned for
>>>> SEEK_DATA and 0 for SEEK_HOLE.
>>>
>>> Ping
>>>
>>> This patch should make xfstest generic/448 happy.
>>
>> It should or it does? ;)
>>
>
> Heh, it *does* make generic/448 happy.
>
>>
>> I saw generic/448 failures on ext4 with Darrick's recent test
>>change,
>> haven't tried ceph yet.
>
> The generic/448 test currently fails with the error codes described in
> the commit log.  I didn't found it useful to include in the commit
> log, but here's the full test output:
>
> File system supports the default behavior.
> File system does not support unwritten extents.
> File system magic#: 0xc36400
> Allocation size: 4194304
> 18. Test file with negative SEEK_{HOLE,DATA} offsets
> 18.01 SEEK_HOLE expected -1 or -1, got 0.                         FAIL
> 18.02 SEEK_DATA expected -1 with errno -6, got -22.               FAIL
>
> seek sanity check failed!

Applied.

Thanks,

                Ilya