[CRIU] [PATCH] arch, x86: int80 -- Clobber all rX registers

[Pavel Tikhomirov]

On 1/15/19 5:49 PM, Pavel Tikhomirov wrote:
> On 1/15/19 5:08 PM, Cyrill Gorcunov wrote:
>> On Tue, Jan 15, 2019 at 02:06:30PM +0000, Dmitry Safonov wrote:
>>>
>>> Some more details would be appreciated.
>>> What kind of kernels is it about?
>>> Is there a redhat patch that changes this ABI?
>>> Worth a reference?
>>>
>>> It's not that I'm against - probably performance penalty on
>>> mainstream from pushing a couple of registers more on the stack
>>> apriori syscalls is small(?), but the commit message could
>>> be improved.
>>
>> Pasha has to apply it, so it is vz kernel I think. Pasha?
>>
> 
> I saw these problem running criu-dev zdtm on VZ7 kernel (it was year ago 
> so it's hard to find exact version), it showed up as:
> 
> ========================== Run zdtm/static/env00 in h 
> ==========================
> Start test
> ./env00 --pidfile=env00.pid --outfile=env00.out --envname=ENV_00_TEST
> Run criu dump
> =[log]=> dump/zdtm/static/env00/36/1/dump.log
> ------------------------ grep Error ------------------------
> (00.006926) Obtaining personality ...
> (00.006953) Sent msg to daemon 64 0 0
> (00.006966) Wait for ack 64 on daemon socket
> pie: 36: __fetched msg: 64 0 0
> (00.007040) Error (criu/parasite-syscall.c:92): si_code=4 si_pid=36 
> si_status=11
> (00.007052) Error (criu/parasite-syscall.c:100): 36 was stopped by 11 
> unexpectedly
> ------------------------ ERROR OVER ------------------------
> ################### Test zdtm/static/env00 FAIL at CRIU dump 
> ###################
> 
> And these patch applied on top of criu-dev unblocked the test.
> 
> Will try to run without these patch to check if it is still relevant.

On latest VZ7 kernel without the patch tests passes with no error: 
https://ci.openvz.org/job/CRIU/job/CRIU-virtuozzo/job/criu-dev/4631/consoleFull

-- 
Best regards, Tikhomirov Pavel
Software Developer, Virtuozzo.