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

[Christian Brauner <brauner@kernel.org>]

On Mon, Feb 05, 2024 at 04:22:41PM +0100, Oleg Nesterov wrote:
> On 02/02, Christian Brauner wrote:
> >
> > > @@ -2050,6 +2051,8 @@ static void pidfd_show_fdinfo(struct seq_file *m, struct file *f)
> > >
> > >  	seq_put_decimal_ll(m, "Pid:\t", nr);
> > >
> > > +	/* TODO: report PIDFD_THREAD */
> >
> > So I think we don't need to do anything here. Since PIDFD_THREAD sets
> > O_EXCL in file->f_flags and in contrast to do_dentry_open() it isn't
> > dropped. So userspace can already detect PIDFD_NONBLOCK as O_NONBLOCK
> > and PIDFD_THREAD as O_EXCL.
> 
> Ah, indeed, I didn't know that fs/proc/fd.c:seq_show() reports ->f_flags.
> Thanks.
> 
> 
> 
> OK, what about another TODO in sys_pidfd_send_signal() ?
> 
> I mean, should I send a simple patch which changes pidfd_send_signal()
> to use do_send_specific() if PIDFD_THREAD ? Or do you think this should

Yes, by default the type of pidfd should determine the scope of the
signal as discussed elsewhere.

(And later we can add flags that allow the caller to change the scope of
the signal. So that Andy's use-cases are covered as well.)