Re: [PATCH 2/2] ptrace: Return the signal to continue with from ptrace_stop

["Eric W. Biederman" <ebiederm@xxxxxxxxxxxx>]

Oleg Nesterov <oleg@xxxxxxxxxx> writes:

> Not sure I understand this patch, I can't apply it. I guess I need to
> clone your tree first, will do later.
>
> Just one question right now,
>
> On 03/15, Eric W. Biederman wrote:
>>
>> +static int ptrace_stop(int exit_code, int why, int clear_code,
>>  			unsigned long message, kernel_siginfo_t *info)
>>  	__releases(&current->sighand->siglock)
>>  	__acquires(&current->sighand->siglock)
>>  {
>>  	bool gstop_done = false;
>> +	bool read_code = true;
>>  
>>  	if (arch_ptrace_stop_needed()) {
>>  		/*
>> @@ -2305,8 +2307,9 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>>  
>>  		/* tasklist protects us from ptrace_freeze_traced() */
>>  		__set_current_state(TASK_RUNNING);
>> +		read_code = false;
>>  		if (clear_code)
>> -			current->exit_code = 0;
>> +			exit_code = 0;
>>  		read_unlock(&tasklist_lock);
>>  	}
>>  
>> @@ -2316,8 +2319,10 @@ static void ptrace_stop(int exit_code, int why, int clear_code,
>>  	 * any signal-sending on another CPU that wants to examine it.
>>  	 */
>>  	spin_lock_irq(&current->sighand->siglock);
>> +	if (read_code) exit_code = current->exit_code;
>
> style ;)
>
>>  	current->last_siginfo = NULL;
>>  	current->ptrace_message = 0;
>> +	current->exit_code = 0;
>
> OK, I like it... but can't we remove the ugly "int clear_code" arg?

The flag clear_code controls what happens if a ptrace_stop does not
stop.  In particular clear_code means do not continue with
a signal if we can not stop.

For do_jobctl_trap calling ptrace_stop it clearly does not matter.

For ptrace_signal it would be a change in behavior, that would
cause the signal not to be delivered.

For ptrace_do_notify we don't have a signal to deliver unless userspace
gives us one so clear_code makes sense at that point.

Which is a long way of saying that no we can not remove clear_stop
because the behavior it is used to implement makes sense and userspace
can reasonably depend upon it.

Eric