linux-kernel - Re: [PATCH] kill_something_info: don't take tasklist

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: <m1d4ng74gm.fsf@frodo.ebiederm.org>
Date:	Tue, 20 May 2008 19:53:29 -0700
From:	ebiederm@...ssion.com (Eric W. Biederman)
To:	Atsushi Tsuji <a-tsuji@...jp.nec.com>
Cc:	Oleg Nesterov <oleg@...sign.ru>, linux-kernel@...r.kernel.org,
	Roland McGrath <roland@...hat.com>,
	Andrew Morton <akpm@...ux-foundation.org>
Subject: Re: [PATCH] kill_something_info: don't take tasklist_lock for pid==-1 case

Atsushi Tsuji <a-tsuji@...jp.nec.com> writes:

> Sorry for late reply and thank you for your comment. I understood the
> mechanism that kill(-1, SIGKILL) can miss the tasks forked by init
> (and the thread group of the current process, because we don't also
> send the signal to them). If kill(-1, SIGKILL) finish before the
> forking init process does list_add_tail_rcu(p->tasks) in
> copy_process(), the process forked by init appears on the ->tasks list
> after that.  Is that right?

No because of this from fork.c:copy_process()
       /*
         * Process group and session signals need to be delivered to just the
         * parent before the fork or both the parent and the child after the
         * fork. Restart if a signal comes in before we add the new process to
         * it's process group.
         * A fatal signal pending means that current will exit, so the new
         * thread can't slip out of an OOM kill (or normal SIGKILL).
         */
        recalc_sigpending();
        if (signal_pending(current)) {
                spin_unlock(&current->sighand->siglock);
                write_unlock_irq(&tasklist_lock);
                retval = -ERESTARTNOINTR;
                goto bad_fork_free_pid;
        }

We closed that whole a while ago, and in doing so reviewed the semantics
and verify that the behavior is required.
>
> Now, I noticed the important problem. I found the tasklist lock in
> kill_something_info() can cause stall when some processes execute
> kill(-1,SIGCONT) concurrently. It can happen even if a system has
> only 4 CPUs (and even if a user is not privileged (not root)).  This is
> because the writer cannot take the tasklist lock when a lot of readers
> exist and keep holding it.
>
> This allows a local DoS. So we have to avoid that stall. The
> conversion from the tasklist lock to rcu_read_lock() can solve this
> problem. I think my patch doesn't make the new problem because the
> problem that kill can miss the tasks have originally occurred without
> my one. If there is no problem, could you ack it?

There are problems.  It would be nice to avoid the local DOS.  How is
a good question, given the atomic definition of signal delivery.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/