Re: [PATCH v2 06/15] pidfs: allow to retrieve exit information

[Christian Brauner <brauner@kernel.org>]

On Tue, Mar 04, 2025 at 06:34:56PM +0100, Oleg Nesterov wrote:
> On 03/04, Christian Brauner wrote:
> >
> > +	task = get_pid_task(pid, PIDTYPE_PID);
> > +	if (!task) {
> > +		if (!(mask & PIDFD_INFO_EXIT))
> > +			return -ESRCH;
> > +
> > +		if (!current_in_pidns(pid))
> > +			return -ESRCH;
> 
> Damn ;) could you explain the current_in_pidns() check to me ?
> I am puzzled...

So we currently restrict interactions with pidfd by pid namespace
hierarchy. Meaning that we ensure that the pidfd is part of the caller's
pid namespace hierarchy. So this check is similar to:

pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
{
        struct upid *upid;
        pid_t nr = 0;

        if (pid && ns->level <= pid->level) {
                upid = &pid->numbers[ns->level];
                if (upid->ns == ns)
                        nr = upid->nr;
        }
        return nr;
}

IOW, we verify that the caller's pid namespace can be found within the
pid namespace hierarchy that the struct pid had a pid numbers allocated
in.

Only that by the time we perform this check the pid numbers have already
been freed so we can't use that function directly. But the pid namespace
hierarchy is still alive as that won't be released until the pidfd has
put the reference on struct @pid. So we can use current_in_pidns().

Is that understandable?