<div dir="ltr">Please see inline<br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Feb 19, 2015 at 4:43 PM, Steve Gordon <span dir="ltr"><<a href="mailto:sgordon@redhat.com" target="_blank">sgordon@redhat.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">----- Original Message -----<br>
> From: "Irena Berezovsky" <<a href="mailto:irenab.dev@gmail.com">irenab.dev@gmail.com</a>><br>
> To: "OpenStack Development Mailing List (not for usage questions)" <<a href="mailto:openstack-dev@lists.openstack.org">openstack-dev@lists.openstack.org</a>>,<br>
><br>
> On Thu, Feb 5, 2015 at 9:01 PM, Steve Gordon <<a href="mailto:sgordon@redhat.com">sgordon@redhat.com</a>> wrote:<br>
><br>
> > ----- Original Message -----<br>
> > > From: "Przemyslaw Czesnowicz" <<a href="mailto:przemyslaw.czesnowicz@intel.com">przemyslaw.czesnowicz@intel.com</a>><br>
> > > To: "OpenStack Development Mailing List (not for usage questions)" <<br>
> > <a href="mailto:openstack-dev@lists.openstack.org">openstack-dev@lists.openstack.org</a>><br>
> > ><br>
> > > Hi<br>
> > ><br>
> > > > 1) If the device is a "normal" PCI device, but is a network card, am I<br>
> > > > still able to<br>
> > > > take advantage of the advanced syntax added circa Juno to define the<br>
> > > > relationship between that card and a given physical network so that the<br>
> > > > scheduler can place accordingly (and does this still use the ML2 mech<br>
> > > > drvier for<br>
> > > > SR-IOV even though it's a "normal" device.<br>
> > ><br>
> > > Actually libvirt won't allow using "normal" PCI devices for network<br>
> > > interfaces into VM.<br>
> > > Following error is thrown by libvirt <a href="http://1.2.9.1" target="_blank">1.2.9.1</a>:<br>
> > > libvirtError: unsupported configuration: Interface type hostdev is<br>
> > currently<br>
> > > supported on SR-IOV Virtual Functions only<br>
> > ><br>
> > > I don't know why libvirt prohibits that. But we should prohibit that on<br>
> > > Openstack side as well.<br>
> ><br>
> > This is true for hostdev"> style configuration, "normal" PCI devices are<br>
> > still valid in Libvirt for passthrough using <hostdev> though. The former<br>
> > having been specifically created for handling passthrough of VFs, the<br>
> > latter being the more generic passthrough functionality and what was used<br>
> > with the original PCI passthrough functionality introduced circa Havana.<br>
> ><br>
> > I guess what I'm really asking in this particular question is what is the<br>
> > intersection of these two implementations - if any, as on face value it<br>
> > seems that to passthrough a physical PCI device I must use the older syntax<br>
> > and thus can't have the scheduler be aware of its external network<br>
> > connectivity.<br>
> ><br>
> Support for "normal" PCI device passthrough for networking in SR-IOV like<br>
> way will require new VIF Driver support for hostdev style device guest XML<br>
> being created and some call invocation to set MAC address and VLAN tag.<br>
><br>
> ><br>
> > > > 2) There is no functional reason from a Libvirt/Qemu perspective that I<br>
> > > > couldn't<br>
> > > > pass through a PF to a guest, and some users have expressed surprise<br>
> > to me<br>
> > > > when they have run into this check in the Nova driver. I assume in the<br>
> > > > initial<br>
> > > > implementation this was prevented to avoid a whole heap of fun<br>
> > additional<br>
> > > > logic<br>
> > > > that is required if this is allowed (e.g. check that no VFs from the PF<br>
> > > > being<br>
> > > > requested are already in use, remove all the associated VFs from the<br>
> > pool<br>
> > > > when<br>
> > > > assigning the PF, who gets allowed to use PFs versus VFs etc.). Am I<br>
> > > > correct here<br>
> > > > or is there another reason that this would be undesirable to allow in<br>
> > > > future -<br>
> > > > assuming such checks can also be designed - that I am missing?<br>
> > > ><br>
> > > I think that is correct. But even if the additional logic was<br>
> > implemented it<br>
> > > wouldn't work because of how libvirt behaves currently.<br>
> ><br>
> > Again though, in the code we have a distinction between a physical device<br>
> > (as I was asking about in Q1) and a physical function (as I am asking about<br>
> > in Q2) and similarly whether libvirt allows or not depends on how you<br>
> > configure in the guest XML. Though I wouldn't be surprised on the PF case<br>
> > if it is in fact not allowed in Libvirt (even with <hostdev>) it is again<br>
> > important to consider this distinctly separate from passing through the<br>
> > physical device case which we DO allow currently in the code I'm asking<br>
> > about.<br>
> ><br>
> I think what you suggest is not difficult to support, but current (since<br>
> Juno) PCI device passthrough for networking is all about SR-IOV PCI device<br>
> passthrough. As I mentioned, to support "normal" PCI device will require<br>
> libvirt VIF Driver adjustment. I think its possible to make this work with<br>
> existing neutron ML2 SRIOV Mechanism Driver.<br>
<br>
</div></div>Understood, was just trying to understand if there was an explicit reason *not* to do this. How should we track this, keep adding to <a href="https://etherpad.openstack.org/p/kilo_sriov_pci_passthrough" target="_blank">https://etherpad.openstack.org/p/kilo_sriov_pci_passthrough</a> ?<br></blockquote><div><br></div><div>I think that probably new etherpad for Liberty should be created in order to track SR-IOV and PCI features. Most of the features proposed for Kilo were rejected due to the nova and neutron priorities focus on other areas. All listed and rejected features and new features priorities should be evaluated and probably picked by people willing to drive it. For Kilo we started this work during the pci_passthrough weekly meetings and finalized at the summit. I think it worked well. I would suggest to do the same for Liberty. </div><div><br></div><div>BR,</div><div>Irena</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Thanks,<br>
<br>
Steve<br>
</blockquote></div><br></div></div>