From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Michael S. Tsirkin" Subject: qemu-kvm: plan for MSI-X Date: Tue, 30 Jun 2009 13:53:41 +0300 Message-ID: <20090630105341.GC30065@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii To: Anthony Liguori , kvm@vger.kernel.org, avi@redhat.com, dlaor@redhat.com, mtosatti@redhat.com Return-path: Received: from mx2.redhat.com ([66.187.237.31]:39740 "EHLO mx2.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753165AbZF3KyL (ORCPT ); Tue, 30 Jun 2009 06:54:11 -0400 Content-Disposition: inline Sender: kvm-owner@vger.kernel.org List-ID: So, now MSI-X is in upstream qemu. The issue here is that irqchip in kernel is still unsupported. So here's the plan for qemu-kvm: 1. on next merge from qemu.git to qemu.kvm, virtio net will be broken 2. a quick work-around is to replace this line in hw/apic.c: msix_supported = 1; with if (!kvm_enabled() || !qemu_kvm_irqchip_in_kernel()) msix_supported = 1; 3. once that is merged, I'll add a patch for kernel msi-x that is qemu-kvm specific. 4. Overall it is unfortunate that we still need to add qemu-kvm specific hacks in new code. Is someone working on upstreaming irqchip btw? Comments? Avi? -- MST