[Devel] Re: [PATCH 3/5] Use task_pid() to find leader's pid

[Oleg Nesterov]

Sukadev Bhattiprolu wrote:
>
> Use task_pid() to get leader's pid since find_pid() cannot be used
> after detach_pid(). See comments in the code below for more details.
>
> ...
>
> +		 * Note: With multiple pid namespaces, active pid namespace of
> +		 * 	 a process is stored in its struct pid. The detach_pid
> +		 * 	 below frees the struct pid, so we will have no notion
> +		 * 	 of an active pid namespace until we complete the
> +		 * 	 subsequent attach_pid(). Which means - calls like
> +		 * 	 find_pid()/pid_to_nr() return NULL and cannot be used
> +		 * 	 between the detach_pid() and attach_pid() calls.

I think both the changelog and the comment are confusing,

>  		detach_pid(tsk, PIDTYPE_PID);
>  		tsk->pid = leader->pid;
> -		attach_pid(tsk, PIDTYPE_PID,  find_pid(tsk->pid));
> +		attach_pid(tsk, PIDTYPE_PID,  task_pid(leader));

because the change itself looks like an obvious performance fix, even
we don't use multiple pid namespaces. I don't think it is good idea to
add a fat comment which doesn't match the current reality, and find_pid()
should be avoided anyway.

Stupid question: why do we need to put the pid namespace into the struct
pid? Isn't it better if the user of the struct pid should know its ns?
For example, if /proc does put_pid(), that pid should be from the active
namespace.

Sukadev, could you cc me if you do that kind of changes?

Oleg.

_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers