Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Thu, 24 Oct 2019 12:31:34 +0200
From:   Steffen Klassert <steffen.klassert@secunet.com>
To:     Tom Rix <trix@redhat.com>
CC:     <herbert@gondor.apana.org.au>, <davem@davemloft.net>,
        <netdev@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        Joerg Vehlow <lkml@jv-coder.de>
Subject: Re: [PATCH v2 1/1] xfrm : lock input tasklet skb queue
Message-ID: <20191024103134.GD13225@gauss3.secunet.de>
References: <CACVy4SUkfn4642Vne=c1yuWhne=2cutPZQ5XeXz_QBz1g67CrA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
In-Reply-To: <CACVy4SUkfn4642Vne=c1yuWhne=2cutPZQ5XeXz_QBz1g67CrA@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Oct 22, 2019 at 05:22:04PM -0700, Tom Rix wrote:
> On PREEMPT_RT_FULL while running netperf, a corruption
> of the skb queue causes an oops.
> 
> This appears to be caused by a race condition here
>         __skb_queue_tail(&trans->queue, skb);
>         tasklet_schedule(&trans->tasklet);
> Where the queue is changed before the tasklet is locked by
> tasklet_schedule.
> 
> The fix is to use the skb queue lock.
> 
> This is the original work of Joerg Vehlow <joerg.vehlow@aox-tech.de>
> https://lkml.org/lkml/2019/9/9/111
>   xfrm_input: Protect queue with lock
> 
>   During the skb_queue_splice_init the tasklet could have been preempted
>   and __skb_queue_tail called, which led to an inconsistent queue.
> 
> ifdefs for CONFIG_PREEMPT_RT_FULL added to reduce runtime effects
> on the normal kernel.

Has Herbert commented on your initial patch, please
fix PREEMPT_RT_FULL instead. There are certainly many
more codepaths that take such assumptions. You can not
fix this by distributing a spin_lock_irqsave here
and there.