Re: [PATCH 0/1] futex: check PF_KTHREAD rather than !p->mm to filter out kthreads

From: Peter Zijlstra <peterz@infradead.org>
To: Oleg Nesterov <oleg@redhat.com>
Cc: Darren Hart <darren@dvhart.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Jerome Marchand <jmarchan@redhat.com>,
	Larry Woodman <lwoodman@redhat.com>,
	Mateusz Guzik <mguzik@redhat.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 0/1] futex: check PF_KTHREAD rather than !p->mm to filter out kthreads
Date: Mon, 2 Feb 2015 16:13:38 +0100	[thread overview]
Message-ID: <20150202151338.GE24151@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <20150202151159.GE26304@twins.programming.kicks-ass.net>

(private noaw)

On Mon, Feb 02, 2015 at 04:11:59PM +0100, Peter Zijlstra wrote:
> On Mon, Feb 02, 2015 at 03:05:15PM +0100, Oleg Nesterov wrote:
> 
> > First of all, why exactly do we need this mm/PF_KTHREAD check added by
> > f0d71b3dcb8332f7971 ? Of course, it is simply wrong to declare a random
> > kernel thread to be the owner as the changelog says. But why kthread is
> > worse than a random user-space task, say, /sbin/init?
> 
> As the changelog says, we _should_ equally disallow other userspace
> tasks that do not share the futex value with us, its just that at the
> time we could not come up with a sensible (and cheap) way of testing for
> this.
> 
> > IIUC, the fact that we can abuse ->pi_state_list is not that bad, no matter
> > if this (k)thread will exit or not. AFAICS, the only problem is that we can
> > boost the prio of this thread. Or I missed another problem?
> 
> No that's it.

Prio leaks allow (local) DoS attacks. It allows an unpriv user to gain
FIFO and burn silly amounts of cycles.

We should really plug that hole entirely.