Re: [PATCH] proc: make proc_fd_permission() thread-friendly

[Oleg Nesterov <oleg@redhat.com>]

On 08/26, Eric W. Biederman wrote:
>
> Oleg Nesterov <oleg@redhat.com> writes:
> >
> > And I also assume that you agree with this change ;)
>
> I don't disagree.  Comparing tgid to pids is goofy and my brain is
> elsewhere so I have no thought through the implications.
>
> Actually thinking I think the check should really be.  In which case we
> are comparing what we really care about.
>
> int proc_fd_permission(struct inode *inode, int mask)
> {
> 	int rv = generic_permission(inode, mask);
> 	if (rv == 0)
> 		return 0;
>
>         rcu_read_lock();
>         struct task *task = pid_task(proc_pid(inode));
>         if (task && (current->files == task->files))

But for what?

To me, this looks like the unnecessary semantic complication. It
looks as if we actually need to restrict the access to /proc/self/fd
or /proc/<tgid>/fd or /proc/<subthread-tid>/fd.

I do not think there is any security reason to deny this. They share
->mm, a sub-thread can do "everything" with its leader or vice versa.

same_thread_group() looks more simple and natural to me. And note
that __ptrace_may_access() was recently changed (in -mm) to use
same_thread_group() instead of "task == current".

So personally I'd prefer to not change this patch and I think it
makes sense even with "make /proc/self point to thread" I sent.

But. please tell me if you really dislike it. You are maintainer,
I won't argue.

Oleg.