[PATCH] Fix bug #13121 - Non-smbd processes using kernel oplocks can hang smbd

[Gary Lockyer]

Have been able to get this to fail by running
make TESTS="kernel-oplock" test
in a loop on my dev box.

Linux 4.10.0-40-generic #44~16.04.1-Ubuntu SMP Thu Nov 9 15:37:44 UTC
2017 x86_64 x86_64 x86_64 GNU/Linux



On 24/11/17 07:12, Gary Lockyer via samba-technical wrote:
> Ok this appears to be flapping, Ran 13 builds on master overnight 5
> failures 8 successes.
> 
> Ubuntu 14.04 Kernel 4.4.0-98-generic
> 
> And failing with.
> 
> 
> Bad child exit code 10
> UNEXPECTED(error): samba3.smb2.kernel-oplocks.kernel_oplocks8(nt4_dc)
> REASON: Exception: Exception: Unknown error/failure. Missing
> torture_fail() or torture_assert_*() call?
> 
> Will kick off some local builds over the weekend on Ubuntu 16.04 and a
> 4.10.0 kernel.
> 
> Gary
> 
> On 17/11/17 06:06, Jeremy Allison via samba-technical wrote:
>> On Thu, Nov 16, 2017 at 04:16:11PM +1300, Gary Lockyer wrote:
>>>
>>>
>>> On 16/11/17 10:42, Jeremy Allison via samba-technical wrote:
>>>> On Thu, Nov 16, 2017 at 10:38:26AM +1300, Andrew Bartlett wrote:
>>>>> On Wed, 2017-11-15 at 13:29 -0800, Jeremy Allison via samba-technical
>>>>> wrote:
>>>>>> On Thu, Nov 16, 2017 at 10:18:26AM +1300, Gary Lockyer wrote:
>>>>>>
>>>>>>> Ok just failed logs attached.
>>>>>>> [558(3552)/2212 at 1h9m32s] samba3.smb2.kernel-oplocks(nt4_dc)
>>>>>>> Bad child exit code 10
>>>>>>
>>>>>> Right - the key is the above. It comes from here in my new
>>>>>> patch:
>>>>>>
>>>>>> +       if (got_alarm == 1) {
>>>>>> +               return 10;
>>>>>> +       }
>>>>>>
>>>>>> which means that the alarm(5) is firing in the client.
>>>>>>
>>>>>> This means your kernel is not sending a RT_SIGNAL_LEASE
>>>>>> signal correctly to the client process holding the lease.
>>>>>>
>>>>>> AKA - you're running with a broken kernel.
>>>>>
>>>>> I've just worked with Gary to use a more recent 14.04 image for the
>>>>> cloud builds, that should clear this up.
>>>>
>>>> Great, thanks ! If one of your Team could also +1 the
>>>> additional patch I'll push it to master so at least
>>>> we won't hang anymore if running on a broken kernel
>>>> and give some feedback as to what is wrong.
>>>>
>>>> Cheers,
>>>>
>>>> 	Jeremy.
>>>>
>>> Passes our auto builds on a Ubuntu 14.04 with a more current kernel
>>
>> Thanks for confirming ! Now at least we can know when a specific Linux
>> kernel version is broken :-).
>>
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20171124/90d739e4/signature.sig>