[CRIU] [PATCH 00/19] Parasite daemon mode, v1

Pavel Emelyanov xemul at parallels.com
Wed Feb 27 03:19:32 EST 2013


On 02/27/2013 11:55 AM, Cyrill Gorcunov wrote:
> On Wed, Feb 27, 2013 at 11:28:54AM +0400, Andrew Vagin wrote:
>> On Tue, Feb 26, 2013 at 05:08:42PM +0400, Cyrill Gorcunov wrote:
>>>
>>> Hi guys, I'm pleased to announce that named "parasite daemon mode"
>>> where we switch parasite code to a daemon which waits for commands
>>> from main crtools module.
>>
>> I think a dumped application should be resumed, if crtools dump crashed.
>>
>> I applied the next patch:
>>
>> diff --git a/cr-dump.c b/cr-dump.c
>> index 3e8ef05..4edaf41 100644
>> --- a/cr-dump.c
>> +++ b/cr-dump.c
>> @@ -418,6 +418,7 @@ static int dump_task_mm(pid_t pid, const struct
>> proc_pid_stat *stat,
>>         MmEntry mme = MM_ENTRY__INIT;
>>         int ret = -1;
>>  
>> +       BUG();
>>         mme.mm_start_code = stat->start_code;
>>         mme.mm_end_code = stat->end_code;
>>         mme.mm_start_data = stat->start_data;
>>
>> bash -x test/zdtm.sh static/busyloop00
> 
> OK, so as being discussed face to face, seems switching
> to unix sockets should be more flexible than using futexes
> here.

Why did you made futexes here at all? We've discussed it several times
and decided, that master thread hangs on socket, sub-threads are controlled
by futexes. WTF?

> I'll prepare a new patch for that (this changes
> last patch only so rest of series remains valid).
> .
> 




More information about the CRIU mailing list