[CRIU] [PATCH 1/1] zdtm: skip 'tcp-close' tests if restore is skipped

[Pavel Emelyanov]

On 06/30/2017 11:00 PM, Pavel Begunkov wrote:
> On Thu, Jun 29, 2017 at 9:59 AM, Andrey Vagin <avagin at virtuozzo.com> wrote:
>> On Wed, Jun 28, 2017 at 01:52:07PM +0300, Pavel Emelyanov wrote:
>>> On 06/28/2017 03:41 AM, silence wrote:
>>>> From: Pavel Begunkov <asml.silence at gmail.com>
>>>>
>>>> 'tcp-close' is a option that replaces connected tcp-sockets with
>>>> 'closed' ones during restoring, therefore it have no sense without
>>>> restore stage. Thus let's skip related tests if the 'norst' flag was
>>>> specified.
>>>
>>> Andrey, I leave it up to you, but methinks, that we'd better tune up the
>>> zdtm lib to let test know whether it was C/R-ed or not, so that after the
>>> wakeup it can perform different checks on its state.
>>
>> I'm agree with you. Unfortunately I didn't notice your comment and
>> committed this patch.
>>
>> Pavel (the Second one;), could you fix the zdtm library accoding
>> with Pavel's comment.
> 
> Ok. Any suggestions how this 'argument passing' should be implemented?
> Should it be cli parameters, environment vars, or maybe some already
> existing in CRIU mechanism?

Currently there's no such mechanisms. CLI/env won't work, as you have to
notify the running process about whether or not to perform post-checks.

I'd suggest doing this via kill with some different signal (not TERM) or
adding an "info" to the TERM signal with the sigqueue() glibc call.

-- Pavel