From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cooper Subject: Re: [PATCH v3] AMD/intremap: Prevent use of per-device vector maps until irq logic is fixed Date: Mon, 17 Jun 2013 11:01:14 +0100 Message-ID: <51BEDE6A.1080103@citrix.com> References: <51B5E1CD02000078000DCA59@nat28.tlf.novell.com> <51BBBF9D.9060404@amd.com> <51BECF1F.7050104@eu.citrix.com> <51BEEC3402000078000DEBCF@nat28.tlf.novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <51BEEC3402000078000DEBCF@nat28.tlf.novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Jan Beulich Cc: "Keir (Xen.org)" , George Dunlap , Jacob Shin , "xen-devel@lists.xen.org" , Suravee Suthikulanit , Sherry Hurwitz List-Id: xen-devel@lists.xenproject.org On 17/06/13 10:00, Jan Beulich wrote: >>>> On 17.06.13 at 10:55, George Dunlap wrote: >> On 15/06/13 02:13, Suravee Suthikulanit wrote: >>> On 6/10/2013 7:25 AM, Jan Beulich wrote: >>>>>>> On 04.06.13 at 18:38, Andrew Cooper >>>>>>> wrote: >>>>> XSA-36 changed the default vector map mode from global to >>>>> per-device. This is >>>>> because a global vector map does not prevent one PCI device from >>>>> impersonating >>>>> another and launching a DoS on the system. >>>>> >>>>> However, the per-device vector map logic is broken for devices with >>>>> multiple >>>>> MSI-X vectors, which can either result in a failed ASSERT() or >>>>> misprogramming >>>>> of a guests interrupt remapping tables. The core problem is not >>>>> trivial to >>>>> fix. >>>>> >>>>> In an effort to get AMD systems back to a non-regressed state, >>>>> introduce a >>>>> new >>>>> type of vector map called per-device-global. This uses per-device >>>>> vector maps >>>>> in the IOMMU, but uses a single used_vector map for the core IRQ logic. >>>>> >>>>> This patch is intended to be removed as soon as the per-device logic >>>>> is fixed >>>>> correctly. >>>>> >>>>> Signed-off-by: Andrew Cooper >>>> Suravee, Jacob, >>>> >>>> no opinion on this at all? I've been talked into considering this >>>> acceptable >>> Sorry for late reply, and for having missed this conversation previously. >>> >>> If we have to go with this solution temporary until we have the >>> permanent fix. >>> I think that is okay with me. Although, would you mind pointing out >>> the affect >>> of having "per-device" vs. "global" irq vector map? I am not quite >>> familiar >>> with the differences. >>> >>>> (with a small coding style fixup, and with the question on >>>> the usefulness of the final warning message - imo redundant with the >>>> immediately preceding message that is being left untouched) >>> I also think the messages are quite confusing. Actually, now that we >>> can have >>> irq vector map and intremap map with different mode, we should be more >>> explicit >>> in the message. >>> >>> Also, the message "Not overriding irq_vector_map setting" is confusing >>> to me. >>> >>> Would you mind considering the attached patch? Here is the sample output >>> >>> (XEN) AMD-Vi: IOMMU 0 Enabled. >>> (XEN) AMD-Vi BUG: per-device vector map logic is broken. Using >>> per-device-global maps instead until a fix is found >> At the very least it can't say BUG -- that needs to be reserved for >> things that actually cause the host to crash (a la BUG_ON()). > That was worded this way in Andrew's original version of the patch > too, and I had also already noted that this wording is too strong. > > Jan > I am not overly attached to the current wording, so feel free to tweak if you wish. As for my opinions of the revised patch: The explicit print of the IOMMU mode is nice, although those in the know could already work it out given the reference or lackthereof to XSA-36. The explicit print of the vector map is wrong, and is liable to be disappearing in 4.4 anyway. On the face of it, the revised patch seems much more like a general cleanup of the printing, rather than the temporary bugfix it is indented to be. I dont have a problem with the cleanup persay, but it should be part of a separate patch. ~Andrew