Received-SPF: pass (google.com: domain of linux-kernel+bounces-73347-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Tue, 20 Feb 2024 17:22:02 +0100
From: Oleg Nesterov <oleg@redhat.com>
To: Christian Brauner <brauner@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>,
	"Eric W. Biederman" <ebiederm@xmission.com>,
	Tycho Andersen <tycho@tycho.pizza>, linux-api@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 2/2] pidfd: change pidfd_send_signal() to respect
 PIDFD_THREAD
Message-ID: <20240220162201.GD7783@redhat.com>
References: <20240214123655.GB16265@redhat.com>
 <20240216-albern-aufwiegen-1de327c7dafd@brauner>
 <20240216130625.GA8723@redhat.com>
 <20240216-ohnedies-improvisieren-58edcc102b6a@brauner>
 <20240216181214.GA10393@redhat.com>
 <20240220-einwurf-depesche-d8682be0370c@brauner>
 <20240220090255.GA7783@redhat.com>
 <20240220-pragmatisch-parzelle-8a1d10a94fae@brauner>
 <20240220110012.GB7783@redhat.com>
 <20240220-anlegen-feinmechaniker-3c2cfcc3ec01@brauner>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240220-anlegen-feinmechaniker-3c2cfcc3ec01@brauner>
User-Agent: Mutt/1.5.24 (2015-08-30)

On 02/20, Christian Brauner wrote:
>
> On Tue, Feb 20, 2024 at 12:00:12PM +0100, Oleg Nesterov wrote:
> >
> > Perhaps we can kill the "task_pid(current) != pid" check and just return
> > EPERM if "kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL", I don't think
> > anobody needs pidfd_send_send_signal() to signal yourself. See below.
>
> Yeah.

You have my ack in advance

> > > +       /* Currently unused. */
> > > +       if (info)
> > > +               return -EINVAL;
> >
> > Well, to me this looks like the unnecessary restriction... And why?
>
> Because right now we aren't sure that it's used

Yes, but...

> and we aren't sure what use-cases are there.

the same use-cases as for rt_sigqueueinfo() ?

Christian, I won't really argue but I still disagree.

Let me first repeat once again, I do not know what people do with pidfd
and pidfd_send_signal() in particular, so I won't be surprised if this
change won't cause any regression report.

But at the same time, I can easily imagine the following scenario: a
userspace programmer tries to use pidfd_send_signal(info != NULL), gets
-EINVAL, decides it can't/shouldn't work, and switches to sigqueueinfo()
without any report to lkml.

> Yes, absolutely. That was always the plan. See appended patch I put on top.
> I put you as author since you did spot this. Let me know if you don't
> want that.

Ah. Thanks Christian. I am fine either way, whatever is more convenient
for you.

But just in case, I won't mind at all if you simply fold this minor fix
into your PIDFD_SEND_PROCESS_GROUP patch, I certainly don't care about
the "From" tag ;)

A really, really minor/cosmetic nit below, feel free to ignore:

> -		if ((task_pid(current) != pid) &&
> +		if (((task_pid(current) != pid) || type > PIDTYPE_TGID) &&

we can remove the unnecessary parens around "task_pid(current) != pid"
or add the extra parens aroung "type > PIDTYPE_TGID".

I mean, the 1st operand of "&&" is

	(task_pid(current) != pid) || type > PIDTYPE_TGID

and this looks a bit inconsistent to me.

Oleg.