From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:38725) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1e2G0X-0006ek-71 for qemu-devel@nongnu.org; Wed, 11 Oct 2017 08:21:06 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1e2G0U-0001c4-2v for qemu-devel@nongnu.org; Wed, 11 Oct 2017 08:21:05 -0400 Received: from mx1.redhat.com ([209.132.183.28]:47958) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1e2G0T-0001bW-So for qemu-devel@nongnu.org; Wed, 11 Oct 2017 08:21:02 -0400 Date: Wed, 11 Oct 2017 14:20:53 +0200 From: Cornelia Huck Message-ID: <20171011142053.1f1e2723.cohuck@redhat.com> In-Reply-To: <1952d9c1-6ef8-0f1d-d89a-08e1236ad275@linux.vnet.ibm.com> References: <1507124979-8880-1-git-send-email-pmorel@linux.vnet.ibm.com> <1507124979-8880-5-git-send-email-pmorel@linux.vnet.ibm.com> <20171009084506.77da9d5c@t450s.home> <20171010113520.6c3d03e3.cohuck@redhat.com> <1952d9c1-6ef8-0f1d-d89a-08e1236ad275@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH v1 4/5] s390x/pci: Refuse to realize VFIO-PCI if AIS needed but supported List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Pierre Morel Cc: Alex Williamson , qemu-devel@nongnu.org, borntraeger@de.ibm.com, pasic@linux.vnet.ibm.com, zyimin@linux.vnet.ibm.com, agraf@suse.de On Tue, 10 Oct 2017 18:01:53 +0200 Pierre Morel wrote: > Since in the case we have no AIS in KVM only PCI is impacted, do you > think we can fence IRQFD for MSI with > > kvm_msi_via_irqfd_allowed = false; > > in kvm_arch_init_irq_routing() ? Hm, it seems we never set that for s390x... does that imply that we never support irqfd for pci anyway? > > In fact, if we also verify the AIS availability in the same routine we > can avoid the ugly include of vfio/pci.h If we can get the serialization right (and setting that value to false does indeed have the desired effect), that's probably the way to go.