From mboxrd@z Thu Jan 1 00:00:00 1970 From: Avi Kivity Subject: Re: [PATCH 2/2][RFC] KVM: Emulate MSI-X table and PBA in kernel Date: Thu, 30 Dec 2010 13:27:15 +0200 Message-ID: <4D1C6C93.8090703@redhat.com> References: <1293007495-32325-1-git-send-email-sheng@linux.intel.com> <201012291655.19535.sheng@linux.intel.com> <20101229092824.GA2876@redhat.com> <201012301532.42341.sheng@linux.intel.com> <20101230074748.GB7889@redhat.com> <4D1C5124.2090409@redhat.com> <20101230103256.GB6441@redhat.com> <4D1C60DE.4020800@redhat.com> <20101230110728.GC6441@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Sheng Yang , Marcelo Tosatti , kvm@vger.kernel.org, Alex Williamson To: "Michael S. Tsirkin" Return-path: Received: from mx1.redhat.com ([209.132.183.28]:29431 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753420Ab0L3L1U (ORCPT ); Thu, 30 Dec 2010 06:27:20 -0500 In-Reply-To: <20101230110728.GC6441@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: On 12/30/2010 01:07 PM, Michael S. Tsirkin wrote: > On Thu, Dec 30, 2010 at 12:37:18PM +0200, Avi Kivity wrote: > > On 12/30/2010 12:32 PM, Michael S. Tsirkin wrote: > > >On Thu, Dec 30, 2010 at 11:30:12AM +0200, Avi Kivity wrote: > > >> On 12/30/2010 09:47 AM, Michael S. Tsirkin wrote: > > >> >I am not really suggesting this. What I say is PBA is unimpl= emented > > >> >let us not commit to an interface yet. > > >> > > >> What happens to a guest that tries to use PBA? > > >> It's a mandatory part of MSI-X, no? > > > > > >Yes. Unfortunately the pending bit is in fact a communication cha= nnel > > >used for function specific purposes when mask bit is set, > > >and 0 when unset. The spec even seems to *require* this use: > > > > > >I refer to this: > > > > > > For MSI and MSI-X, while a vector is masked, the function is pro= hibited > > > from sending the associated message, and the function must set t= he > > > associated Pending bit whenever the function would otherwise sen= d the > > > message. When software unmasks a vector whose associated Pending= bit is > > > set, the function must schedule sending the associated message, = and > > > clear the Pending bit as soon as the message has been sent. Note= that > > > clearing the MSI-X Function Mask bit may result in many messages= needing > > > to be sent. > > > > > > > > > If a masked vector has its Pending bit set, and the associated > > > underlying interrupt events are somehow satisfied (usually by so= ftware > > > though the exact manner is function-specific), the function must= clear > > > the Pending bit, to avoid sending a spurious interrupt message l= ater > > > when software unmasks the vector. However, if a subsequent inter= rupt > > > event occurs while the vector is still masked, the function must= again > > > set the Pending bit. > > > > > > > > > Software is permitted to mask one or more vectors indefinitely, = and > > > service their associated interrupt events strictly based on poll= ing > > > their Pending bits. A function must set and clear its Pending bi= ts as > > > necessary to support this =E2=80=9Cpure polling=E2=80=9D mode of= operation. > > > > > >For assigned devices, supporting this would require > > >that the mask bits on the device are set if the mask bit in > > >guest is set (otherwise pending bits are disabled). > > > > Can't this be done by setting the real mask bit when the guest rea= ds > > the virtual pending bit, then reading the real pending bit? > > Function specific is function-specific, but most likely not, > by that time the pending bit in the device might be clear: > 'clear the Pending bit as soon as the message has been sent' But when we set the mask bit, it must change the pending bit back to th= e=20 function-specific condition? > > >Existing code does not support PBA in assigned devices, so at lea= st it's > > >not a regression there, and the virtio spec says nothing about th= is so > > >we should be fine. > > > > Why isn't it subject to the pci spec? > > > > If an interrupt condition exits, the bit should be set. > > I wish. But this is not what the spec says above. It says if vector i= s > unmasked, bit must be cleared. If interrupt condition exists, and the vector is masked, the pending bi= t=20 is set. Otherwise the pending bit is clear. Better? --=20 error compiling committee.c: too many arguments to function