From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joby Poriyath Subject: Re: (XEN) [2013-08-23 12:23:26] Assertion 'entry->pirq == pirq' failed at vmsi.c Date: Fri, 23 Aug 2013 21:58:33 +0100 Message-ID: <20130823205833.GA16674@citrix.com> References: <1269238736.20130823143420@eikelenboom.it> <5217780502000078000EE00B@nat28.tlf.novell.com> <18710329292.20130823152657@eikelenboom.it> <521785F402000078000EE061@nat28.tlf.novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <521785F402000078000EE061@nat28.tlf.novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Jan Beulich Cc: Sander Eikelenboom , malcolm.crossley@citrix.com, xen-devel List-Id: xen-devel@lists.xenproject.org On Fri, Aug 23, 2013 at 02:55:32PM +0100, Jan Beulich wrote: > >>> On 23.08.13 at 15:26, Sander Eikelenboom wrote: > > I have adjusted the debug patch, only letting it reuse the entry if ( pdev > > == entry->pdev && entry->pirq == pirq){ > > so the assertion should be always true now, would this perhaps be the proper > > fix or should it also check if it is also masked ? > > Yes, that might be a possible solution, albeit I don't like it - it points > out that the original patch from Joby is - beyond the bug here - at > least inefficient (in that we'd then allocate an entry per MSI-X table > entry, i.e. per IRQ, instead of per device). I therefore think we > rather ought to revert that change, and have him redo it properly. Ok, we ought to revisit the original patch. Let us revert the code for now. Thanks Sander for spotting this. It's a certainly an oversight from my side. A PCIe function can have multiple MSIx entries. I'm on holidays. I'll be back on 29th Aug. I shall pick this up, then. Thanks, Joby