From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752216AbcGYOJR (ORCPT ); Mon, 25 Jul 2016 10:09:17 -0400 Received: from smtp.citrix.com ([66.165.176.89]:23624 "EHLO SMTP.CITRIX.COM" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751292AbcGYOJP (ORCPT ); Mon, 25 Jul 2016 10:09:15 -0400 X-IronPort-AV: E=Sophos;i="5.28,419,1464652800"; d="scan'208";a="368207712" Subject: Re: [Xen-devel] [PATCH linux v2 0/9] xen: pvhvm: support bootup on secondary vCPUs To: Julien Grall , Vitaly Kuznetsov References: <1467302203-22399-1-git-send-email-vkuznets@redhat.com> <57960843.9070602@citrix.com> <87bn1lx2fd.fsf@vitty.brq.redhat.com> <70e013df-550b-3071-71fe-1a618e0a27d7@arm.com> CC: Juergen Gross , Stefano Stabellini , Andrew Cooper , , , Ingo Molnar , David Vrabel , Jan Beulich , "H. Peter Anvin" , , Boris Ostrovsky , Thomas Gleixner From: David Vrabel Message-ID: <57961D84.1090503@citrix.com> Date: Mon, 25 Jul 2016 15:09:08 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Icedove/38.5.0 MIME-Version: 1.0 In-Reply-To: <70e013df-550b-3071-71fe-1a618e0a27d7@arm.com> Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit X-DLP: MIA1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 25/07/16 15:01, Julien Grall wrote: > Hello, > > On 25/07/16 14:39, Vitaly Kuznetsov wrote: >> Julien Grall writes: >> >>> Hi David, >>> >>> On 25/07/16 13:38, David Vrabel wrote: >>>> On 30/06/16 16:56, Vitaly Kuznetsov wrote: >>>>> It may happen that Xen's and Linux's ideas of vCPU id diverge. In >>>>> particular, when we crash on a secondary vCPU we may want to do kdump >>>>> and unlike plain kexec where we do migrate_to_reboot_cpu() we try >>>>> booting >>>>> on the vCPU which crashed. This doesn't work very well for PVHVM >>>>> guests as >>>>> we have a number of hypercalls where we pass vCPU id as a >>>>> parameter. These >>>>> hypercalls either fail or do something unexpected. To solve the >>>>> issue we >>>>> need to have a mapping between Linux's and Xen's vCPU ids. >>>>> >>>>> This series solves the issue for x86 PVHVM guests. PV guests don't >>>>> (and >>>>> probably won't) support kdump so I always assume Xen's vCPU id == >>>>> Linux's >>>>> vCPU id. ARM guests will probably need to get proper mapping once >>>>> we start >>>>> supporting kexec/kdump there. >>>> >>>> Applied to for-linus-4.8, thanks. >>> >>> It would have been nice to send a ping before applying. This patch >>> series is containing Xen ARM code which has not been acked by Stefano, >>> nor had feedback from ARM side. >>> >>> For instance given that all the hypercalls are representing a "vcpu >>> id" using "uint32_t" it is a bit weird to use "int" to define >>> xen_vcpu_id (see patch #3). >> >> CPU id is usually 'int' in linux and now we pass it to all >> hypercalls as it is. > > Well, we need to differentiate between the internal representation of > the CPU which is based on the boot order and the logical CPU ID. For > instance on ARM, the logical CPU ID may not be contiguous nor 0 for the > first CPU. > > From my understanding, the macros in patch #3 will be used at the last > minute when prepare the hypercall data. IHMO this is very similar to a > logical ID and defined as uint32_t by the hypercall ABI. > > Although, I agree that currently we use the internal CPU id on ARM which > is very unfortunate because this value is based on the order of the > nodes in the device tree. > > One way to abolish it on ARM would be to use the MPIDR (or at least a > part) for the VCPU ID. > >> It is a bit more convenient in the mapping I >> introduce as we can set it to a negative value to indicate there is no >> mapping available. I can definitely change that and use something like >> U32_MAX-1 to instead but I'm not sure it is worth it... > > I looked at the definition of cpu_acpi_id on x86 which return > x86_cpu_to_acpiid that has been defined to an uint32_t. > > So you are assuming that it will never be possible to have an ID > > 0x80000000. > > Also, this may not be true on ARM depending how we define the VCPU > mapping. We could decide to use the MPIDR which is in this case may be > considered as "negative". If you want to change how you number vCPUs on ARM in the future, you can (if necessary) change the type of this per-cpu variable as well. David