Re: Q: check_unsafe_exec() races (Was: [PATCH 2/4] fix setuid sometimes doesn't)

[Oleg Nesterov <oleg@redhat.com>]

On 03/29, Al Viro wrote:
>
> On Sun, Mar 29, 2009 at 01:53:43AM +0100, Oleg Nesterov wrote:
>
> > Let's suppose we have two threads T1 and T2. T1 exits, and calls
> > exit_fs().
> >
> > 	exit_fs:
> >
> > 		tsk->fs = NULL;
> > 		// WINDOW
> > 		put_fs_struct(fs);
> >
> > Now, if T2 does exec() and check_unsafe_exec() happens in the WINDOW
> > above, we set LSM_UNSAFE_SHARE.
> >
> > Or we can race with sub-thread doing clone(CLONE_FS|CLONE_THREAD),
> > the new thread is not visible in ->thread_group, buy copy_fs()
> > can already increment fs->count.
>
> Frankly, I don't think we really care.  Note that having several sub-threads
> and doing execve() in one of them will kill the rest, so you really want
> to do some kind of synchronization to get something similar to reasonable
> behaviour anyway.

OK.

Let's suppose that check_unsafe_exec() does not set LSM_UNSAFE_SHARE and
drops ->siglock. After that, another sub-thread does clone(CLONE_FS) without
CLONE_THREAD.

Unless we killed other threads, I can't see how we can check ->fs is not
shared with another process, we can fool ->bprm_set_creds() anyway.

Confused.

Oleg.