From mboxrd@z Thu Jan 1 00:00:00 1970 From: Keir Fraser Subject: Re: [PATCH][SVM] HVM fix for SWINT event injection Date: Tue, 31 Jul 2007 19:02:53 +0100 Message-ID: References: <683860AD674C7348A0BF0DE3918482F605801BE9@SAUSEXMB1.amd.com> Mime-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <683860AD674C7348A0BF0DE3918482F605801BE9@SAUSEXMB1.amd.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: "Woller, Thomas" , xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org On 31/7/07 16:30, "Woller, Thomas" wrote: >> What I'll do is reorder SVM's intr.c core code to be the same >> as for VMX (it's currently bogusly different) but with the >> added constraint that we do not propagate exitintinfo.swint >> into eventinj. I'll add a comment to explain why. > Ok. Shoot me the patch first if you have time, I'll test with all the > HVM guests I have on hand. http://xenbits.xensource.com/staging/xen-unstable.hg, changeset 15662. It's rather large, and only run-tested on VT. :-) I'll try and do some SVM testing at home, but it'd be great if you can give it a spin. My only question is whether vmcb->exitintinfo is written on every vmexit, even those where no event was being delivered. I assumed so, and hence do not bother to zap the field before vmentry, but that'll need adding back in if I'm wrong. > Re: 15652 > 15658 looks good initially, appreciate the quick fixes. Yes, that was an embarrassing bug and not in Eric Liu's original patch! -- Keir