From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:57543) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1RUP67-00024Q-9D for qemu-devel@nongnu.org; Sat, 26 Nov 2011 15:43:44 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1RUP66-0005Jh-Bu for qemu-devel@nongnu.org; Sat, 26 Nov 2011 15:43:43 -0500 Received: from mx1.redhat.com ([209.132.183.28]:40408) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1RUP66-0005Im-4q for qemu-devel@nongnu.org; Sat, 26 Nov 2011 15:43:42 -0500 Date: Sat, 26 Nov 2011 22:45:17 +0200 From: "Michael S. Tsirkin" Message-ID: <20111126204517.GA4388@redhat.com> References: <1322137732-30840-1-git-send-email-pbonzini@redhat.com> <20111124164203.GE26770@redhat.com> <4ECE7593.5090902@redhat.com> <20111124175436.GA29199@redhat.com> <4ECF508E.5050709@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4ECF508E.5050709@redhat.com> Subject: Re: [Qemu-devel] [PATCH 1.0] virtio: add and use virtio_set_features List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini Cc: qemu-devel@nongnu.org, agraf@suse.de On Fri, Nov 25, 2011 at 09:23:42AM +0100, Paolo Bonzini wrote: > On 11/24/2011 06:55 PM, Michael S. Tsirkin wrote: > >> Could some backend make it a hard failure? > > > >I don't see how, there's no way to report a failure from > >an io port write. > > You can exit(1), or fall back to a restricted set of features like > we do for BAD_FEATURE. BAD_FEATURE is a special case of features > that is not exposed by the host, but requested by the guest. I donn't think we want to exit on BAD_FEATURE... > >> If I understand > >> correctly, this would have prevented the BAD_FEATURE bug too. > > > >Which bug? > > VIRTIO_F_BAD_FEATURE(30) > This feature should never be negotiated by the guest; doing so is an > indication that the guest is faulty. An experimental virtio PCI > driver contained in Linux version 2.6.25 had this problem, and this > feature bit can be used to detect it. Ah, I understand, you mean helping debugging. OK, but if we add exit() in the future we still do not need to return a status from this function. > > what would have prevented it. > > exit(1) on unsupported features. > > Paolo This is somewhat problematic, for example this does not flush out the cache and so might cause data corruption. Yet I think it would be nice to have a 'safe_exit' functionality to invoke on a guest bug. Or maybe it's better to stop the VM, this way the monitor is available and debugging is easier. -- MST