Re: [PATCH 1/2] pid: add pidfd_open()

[Oleg Nesterov <oleg@xxxxxxxxxx>]

On 05/15, Christian Brauner wrote:
>
> On Wed, May 15, 2019 at 04:38:58PM +0200, Oleg Nesterov wrote:
> >
> > it seems that you can do a single check
> >
> > 	tsk = pid_task(p, PIDTYPE_TGID);
> > 	if (!tsk)
> > 		ret = -ESRCH;
> >
> > this even looks more correct if we race with exec changing the leader.
>
> The logic here being that you can only reach the thread_group leader
> from struct pid if PIDTYPE_PID == PIDTYPE_TGID for this struct pid?

Not exactly... it is not that PIDTYPE_PID == PIDTYPE_TGID for this pid,
struct pid has no "type" or something like this.

The logic is that pid->tasks[PIDTYPE_XXX] is the list of task which use
this pid as "XXX" type.

For example, clone(CLONE_THREAD) creates a pid which has a single non-
empty list, pid->tasks[PIDTYPE_PID]. This pid can't be used as TGID or
SID.

So if pid_task(PIDTYPE_TGID) returns non-NULL we know that this pid was
used for a group-leader, see copy_process() which does

	if (thread_group_leader(p))
		attach_pid(p, PIDTYPE_TGID);


If we race with exec which changes the leader pid_task(TGID) can return
the old leader. We do not care, but this means that we should not check
thread_group_leader().

Oleg.