[CRIU] [PATCH] compel: new catch and "release at" helpers added

[Stanislav Kinsburskiy]

17.03.2017 16:03, Pavel Emelyanov пишет:
> On 03/17/2017 05:59 PM, Stanislav Kinsburskiy wrote:
>>
>> 17.03.2017 15:43, Pavel Emelyanov пишет:
>>> On 03/17/2017 04:54 PM, Stanislav Kinsburskiy wrote:
>>>> 17.03.2017 14:41, Pavel Emelyanov пишет:
>>>>> On 03/17/2017 04:31 PM, Stanislav Kinsburskiy wrote:
>>>>>> 17.03.2017 14:28, Pavel Emelyanov пишет:
>>>>>>> On 03/16/2017 02:43 PM, Stanislav Kinsburskiy wrote:
>>>>>>>> These helpers are compel_run_at split result.
>>>>>>> BTW, the compel_run_at itself seems to be unused.
>>>>>> I use it.
>>>>> Ah :) OK, can we then give this trio some better names?
>>>>> Because run_at, release_at and catch do not give a valid
>>>>> impression on what they really do.
>>>> Well, naming is not one of my good skills. But, frankly, in this case I
>>>> like the names. There are pretty clear (and meaningful) to me.
>>>> What might be other options&?
>>> release_at -> step_start
>>> catch -> step_wait
>>> run_at -> step = step_start + step_wait
>>>
>>> ?
>> Emmm... What does "step" here mean?
>> It remains me GDB "step" and this is very obfuscating.
> OK, makes sense. What's your proposal? "Keeping the name as is" is not nice,
> try to come up with smth else.

What about "compel_jump", "compel_wait" and "compel_exec" respectively?

> -- Pavel
>
>>>>> -- Pavel
>>>>>
>>>>>>>> IOW, it's an asynchronious way
>>>>>>>> of calling some binary code in remote process and cathing it after it's done.
>>>>>>>>
>>>>>>>> Signed-off-by: Stanislav Kinsburskiy <skinsbursky at virtuozzo.com>
>>>>>>>> ---
>>>>>>>>      compel/include/uapi/infect.h |    2 ++
>>>>>>>>      compel/src/lib/infect.c      |   14 ++++++++++++++
>>>>>>>>      2 files changed, 16 insertions(+)
>>>>>>>>
>>>>>>>> diff --git a/compel/include/uapi/infect.h b/compel/include/uapi/infect.h
>>>>>>>> index da0ca3a..aebb77c 100644
>>>>>>>> --- a/compel/include/uapi/infect.h
>>>>>>>> +++ b/compel/include/uapi/infect.h
>>>>>>>> @@ -66,6 +66,8 @@ extern int compel_syscall(struct parasite_ctl *ctl, int nr, long *ret,
>>>>>>>>      		unsigned long arg6);
>>>>>>>>      extern int compel_run_in_thread(struct parasite_thread_ctl *tctl, unsigned int cmd);
>>>>>>>>      extern int compel_run_at(struct parasite_ctl *ctl, unsigned long ip, user_regs_struct_t *ret_regs);
>>>>>>>> +extern int compel_release_at(struct parasite_ctl *ctl, unsigned long ip);
>>>>>>>> +extern int compel_catch(struct parasite_ctl *ctl, user_regs_struct_t *ret_regs);
>>>>>>>>      
>>>>>>>>      /*
>>>>>>>>       * The PTRACE_SYSCALL will trap task twice -- on
>>>>>>>> diff --git a/compel/src/lib/infect.c b/compel/src/lib/infect.c
>>>>>>>> index c024250..a2cb5d6 100644
>>>>>>>> --- a/compel/src/lib/infect.c
>>>>>>>> +++ b/compel/src/lib/infect.c
>>>>>>>> @@ -581,6 +581,20 @@ int compel_run_at(struct parasite_ctl *ctl, unsigned long ip, user_regs_struct_t
>>>>>>>>      	return ret;
>>>>>>>>      }
>>>>>>>>      
>>>>>>>> +int compel_release_at(struct parasite_ctl *ctl, unsigned long ip)
>>>>>>>> +{
>>>>>>>> +	user_regs_struct_t regs = ctl->orig.regs;
>>>>>>>> +
>>>>>>>> +	return parasite_run(ctl->rpid, PTRACE_CONT, ip, 0, &regs, &ctl->orig);
>>>>>>>> +}
>>>>>>>> +
>>>>>>>> +int compel_catch(struct parasite_ctl *ctl, user_regs_struct_t *ret_regs)
>>>>>>>> +{
>>>>>>>> +	user_regs_struct_t regs;
>>>>>>>> +
>>>>>>>> +	return parasite_trap(ctl, ctl->rpid, ret_regs ? ret_regs : &regs, &ctl->orig);
>>>>>>>> +}
>>>>>>>> +
>>>>>>>>      static int accept_tsock(struct parasite_ctl *ctl)
>>>>>>>>      {
>>>>>>>>      	int sock;
>>>>>>>>
>>>>>>>> .
>>>>>>>>
>>>>>> .
>>>>>>
>>>> .
>>>>
>> .
>>