From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:47227) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VpxSi-0005P2-W1 for qemu-devel@nongnu.org; Mon, 09 Dec 2013 04:49:18 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1VpxSc-0000qV-0X for qemu-devel@nongnu.org; Mon, 09 Dec 2013 04:49:12 -0500 Received: from mx1.redhat.com ([209.132.183.28]:33843) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1VpxSb-0000qJ-OU for qemu-devel@nongnu.org; Mon, 09 Dec 2013 04:49:05 -0500 Message-ID: <52A5920B.2040601@redhat.com> Date: Mon, 09 Dec 2013 10:48:59 +0100 From: Paolo Bonzini MIME-Version: 1.0 References: <20131206204715.16731.12627.stgit@bling.home> <20131206204842.16731.29998.stgit@bling.home> <52A24A6A.6000809@redhat.com> <1386369360.25738.523.camel@ul30vt.home> In-Reply-To: <1386369360.25738.523.camel@ul30vt.home> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PULL 4/5] vfio-pci: Add debug config options to disable MSI/X KVM support List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Alex Williamson Cc: qemu-devel@nongnu.org, aliguori@amazon.com, kvm@vger.kernel.org Il 06/12/2013 23:36, Alex Williamson ha scritto: >>> > > #define VFIO_ALLOW_MMAP 1 >>> > > #define VFIO_ALLOW_KVM_INTX 1 >>> > > +#define VFIO_ALLOW_KVM_MSI 1 >>> > > +#define VFIO_ALLOW_KVM_MSIX 1 >> > >> > Why not make these device properties instead? > Honestly, I don't think they're that useful to the average or even to > the advanced user. Each of these disables an acceleration/bypass path > and gives more logging output when debug is enabled. Otherwise we don't > get debugging to see when mmap'd BARs are accessed and we can't tell > when interrupts happen relative to other accesses. For instance it was > pretty obvious to find the Nvidia MSI ACK when seeing a config space > write immediately following an MSI injection. Without debugging > enabled, these just slow things down and I don't know of any cases where > a device behaves better with these features disabled. Yeah, behaving better is unlikely. I was thinking of them as a debugging features. As to debug messages, have you considered replacing them with tracepoints? Paolo