[CRIU] Re: [PATCH 2/6] parasite: Dump blocked thread signals mask

[Pavel Emelyanov]

On 11/01/2012 03:04 PM, Cyrill Gorcunov wrote:
> On Thu, Nov 01, 2012 at 02:59:10PM +0400, Pavel Emelyanov wrote:
>> On 10/31/2012 02:44 PM, Cyrill Gorcunov wrote:
>>> ---
>>>  include/parasite.h |    1 +
>>>  parasite.c         |    8 ++++++--
>>>  2 files changed, 7 insertions(+), 2 deletions(-)
>>>
>>
>>> @@ -354,11 +354,15 @@ static int dump_thread_info(struct parasite_dump_thread_info *args)
>>>  
>>>  	ret = sys_prctl(PR_GET_TID_ADDRESS, (unsigned long) &args->tid_addr, 0, 0, 0);
>>>  	if (ret)
>>> -		return ret;
>>> +		goto err;
>>>  
>>>  	args->tid = sys_gettid();
>>>  
>>> -	return 0;
>>> +	ksigfillset(&args->blocked);
>>> +	ret = sys_sigprocmask(SIG_SETMASK, NULL, &args->blocked, sizeof(k_rtsigset_t));
>>> +
>>> +err:
>>> +	return ret;
>>>  }
>>>  
>>>  static int drain_fds(struct parasite_drain_fd *args)
>>
>> No, this is wrong. Threads' blocked signals should be handled the same way as
>> tasks' are.
> 
> Not sure if I follow you here. At moment task's signals are obtained from
> parasite code, where they are saved once the task get infected with our
> blob. Now when we run parasite code inside thread space we fetch blocked
> signals, so what i'm missing?
> .
> 

Threads should also block their signals once seized and infected.