From: Greg Edwards Subject: Re: [PATCH] iommu/vt-d: fix race between free_irte() and get_irte() Date: Wed, 23 Jul 2014 08:49:17 -0600 Message-ID: <20140723144917.GA26986@psuche.datadirectnet.com> References: <20140722142719.GA28143@psuche.datadirectnet.com> <20140723144024.GA14017@8bytes.org> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <20140723144024.GA14017-zLv9SwRftAIdnm+yROfE0A@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Errors-To: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org To: Joerg Roedel Cc: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org, David Woodhouse , linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: iommu@lists.linux-foundation.org On Wed, Jul 23, 2014 at 04:40:24PM +0200, Joerg Roedel wrote: > On Tue, Jul 22, 2014 at 08:27:19AM -0600, Greg Edwards wrote: >> get_irte() can race with free_irte() and dereference a NULL iommu >> pointer. > > Have you seen any real occurance of this race? Get_irte is called in the > set_affinity path, how can that race with the irq being freed? Yes, that's how we hit it. A process was setting the CPU affinity while QEMU was releasing the IRQ. We have a CI stress test that turned this up. Greg