Re: [PATCH] pidfd: prevent creation of pidfds for kthreads

[Christian Brauner <brauner@kernel.org>]

On Thu, Aug 01, 2024 at 10:01:20AM GMT, Oleg Nesterov wrote:
> OK, I won't argue, but ....
> 
> On 08/01, Christian Brauner wrote:
> >
> > On Wed, Jul 31, 2024 at 04:51:33PM GMT, Oleg Nesterov wrote:
> > > On 07/31, Christian Brauner wrote:
> > > >
> > > > It's currently possible to create pidfds for kthreads but it is unclear
> > > > what that is supposed to mean. Until we have use-cases for it and we
> > > > figured out what behavior we want block the creation of pidfds for
> > > > kthreads.
> > >
> > > Hmm... could you explain your concerns? Why do you think we should disallow
> > > pidfd_open(pid-of-kthread) ?
> >
> > It basically just works now and it's not intentional - at least not on
> > my part. You can't send signals to them,
> 
> Yes, you can't send signals to kthread. So what?
> 
> You can't send signals to the normal processes if check_kill_permission()
> fails. And even if you are root, you can't send an unhandled signal via
> pidfd = pidfd_open(1).
> 
> > you may or may not get notified
> > via poll when a kthread exits.
> 
> Why? the exiting kthread should not differ in this respect?

Why do you want to allow it? I see zero reason to get a reference to a
kthread if there's no use-case for it. kthreads are mostly a kernel
thing so why give userspace handles to it. And as I said before, there's
userspace out there that's already confused why they can get references
to them in the first place.

> 
> > (So imho this causes more confusion then it is actually helpful. If we
> > add supports for kthreads I'd also like pidfs to gain a way to identify
> > them via statx() or fdinfo.)
> 
> /proc/$pid/status has a "Kthread" field...

Going forward, I don't want to force people to parse basic stuff out of
procfs. Ideally, they'll be able to mostly rely on pidfd operations
only.