From mboxrd@z Thu Jan 1 00:00:00 1970 From: Yang Zhang Subject: Re: X58 Virtualization w/ Linux Date: Mon, 20 Jun 2016 10:07:48 +0800 Message-ID: References: <03f27bbf-f8ad-b377-c194-adaefe808077@stevenovakov.com> <2884c8b8-9ad3-cba7-d9ba-b94946775087@gmail.com> <0cc56e39-829c-144d-5e30-7d5505fe896a@stevenovakov.com> <3b5e8af3-c0da-8fab-8f40-05fa27ef2866@stevenovakov.com> <6c76b555-f9af-d651-d5e2-490cf899ee93@stevenovakov.com> <9a909eb1-23ef-f29f-a332-359fa842afd4@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit To: Paolo Bonzini , Steve Novakov , kvm@vger.kernel.org Return-path: Received: from mail-ob0-f182.google.com ([209.85.214.182]:33872 "EHLO mail-ob0-f182.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751450AbcFTDzX (ORCPT ); Sun, 19 Jun 2016 23:55:23 -0400 Received: by mail-ob0-f182.google.com with SMTP id y7so8078297obt.1 for ; Sun, 19 Jun 2016 20:54:49 -0700 (PDT) In-Reply-To: Sender: kvm-owner@vger.kernel.org List-ID: On 2016/6/16 15:17, Paolo Bonzini wrote: > > > On 16/06/2016 03:19, Yang Zhang wrote: >>> intel_iommu=on >>> intel_iommu=on vfio_iommu_type1.allow_unsafe_interrupts=1 >>> >>> all sorts of combinations. I've verified that all of the relevant >>> modules load (vfio* kvm kvm_intel pci_stub) through dmesg. That that >>> "DMAR-IR" message is suspicious though, no? That there is some setting >>> in the kernel that disables remapping by default for my BIOS. I've seen >>> posts with working vt-d even with an i7-950. I wonder if they have the >>> C3 stepping of the northbridge. Does that sound like it might be an >>> issue? >>> >>> I can't believe that there is not some way to make this work, as it >>> works flawlessly with ESXi (which sits on top of a linux kernel, though >>> it's vmware's proprietary fork I think). If my system boots in "Plug and >>> play" mode, the BIOS configuration for non critical devices shouldn't >>> matter, right? >> >> It seems there are something wrong with vfio. Have you tried with legacy >> device pass through? It should be disabled on new distribution(like >> ubuntu 16.04) but RHEL6 may still support it. > > There shouldn't be any difference. He gets a freeze with > intel_iommu=on, even before going to device assignment. Yes, if system freeze on boot, then it seems something wrong in linux kernel. Steve, can you post the dmesg log and check whether there are some additional devices on your board? I remember some legacy device drivers are not friendly to IOMMU. -- best regards yang