linux-kernel - Re: [PATCH v2 3/3] tty: Use RCU read lock to iterate tasks and threads in __do

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20180117173955.GB7964@redhat.com>
Date:   Wed, 17 Jan 2018 18:39:55 +0100
From:   Oleg Nesterov <oleg@...hat.com>
To:     "Eric W. Biederman" <ebiederm@...ssion.com>
Cc:     Kirill Tkhai <ktkhai@...tuozzo.com>, gregkh@...uxfoundation.org,
        jslaby@...e.com, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 3/3] tty: Use RCU read lock to iterate tasks and
 threads in __do_SAK()

On 01/17, Eric W. Biederman wrote:
>
> __do_SAK() needs to be 100% accurate.

But it can't. A process/thread can open tty right after the check.

> I do not see the rcu_read_lock
> guaranteeing that new processes created while the process list is being
> iterated that happen to have a reference to the tty will be seen.

We can't miss the new child if its parent has this tty opened at fork() time,
__do_SAK() sends SIGKILL and ->siglock serializes __do_SAK() with copy_process()
which checks signal_pending() under the same ->siglock. So either fork() should
fail or for_each_process() should see the new child.

Right?

Otherwise we do not care. The child can open tty later but this doesn't differ
from the "race" above.

Oleg.