Re: [PATCH v3 1/1] pidfd: implement PIDFD_THREAD flag for pidfd_open()

[Oleg Nesterov]

On 01/31, Christian Brauner wrote:
>
> On Wed, Jan 31, 2024 at 02:26:02PM +0100, Oleg Nesterov wrote:
> > With this flag:
> >
> > 	- pidfd_open() doesn't require that the target task must be
> > 	  a thread-group leader
> >
> > 	- pidfd_poll() succeeds when the task exits and becomes a
> > 	  zombie (iow, passes exit_notify()), even if it is a leader
> > 	  and thread-group is not empty.
> >
> > 	  This means that the behaviour of pidfd_poll(PIDFD_THREAD,
> > 	  pid-of-group-leader) is not well defined if it races with
> > 	  exec() from its sub-thread; pidfd_poll() can succeed or not
> > 	  depending on whether pidfd_task_exited() is called before
> > 	  or after exchange_tids().
>
> I think that we can live with that for now. If it doesn't hinder Tycho's
> use-case it's fine.

OK, good.

> (I often wish that we could report custom data from poll similar to
> kqueue on bsd then we could attach file specific information to the
> notification. That would enable userspace to handle this the way they
> want.)

Or at least perhaps we can change do_notify_pidfd() to use the keyed
wakeups... I'll try to take a look later.

> > thread_group_exited() is no longer used, perhaps it can die.
>
> Feel free to add a patch for that on top of it.

Yes, will do.

Oleg.