From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Beng Heng, Ng" Subject: Re: Experimental results for VGA passthrough Date: Mon, 08 Jun 2009 16:59:45 -0400 Message-ID: <4A2D7BC1.2090200@eecs.umich.edu> References: <4A27E395.1000705@eecs.umich.edu> <0511662A905F904FB91411FF69F1052B250B07@mnetexch1.adamapps.host> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <0511662A905F904FB91411FF69F1052B250B07@mnetexch1.adamapps.host> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: djmagee@mageenet.net Cc: xen-devel List-Id: xen-devel@lists.xenproject.org I noticed that Xen does not boot if only the Intel IGD is present and selected, i.e. only one graphics device. Also, I'm not able to get high resolution if I passthrough Intel IGD and install the Intel Graphics Media Accelerator driver. Ironically, I can get higher resolution and color if I remove the driver. Has anyone encountered similar observations before? - Beng Heng djmagee@mageenet.net wrote: > The DQ45CB does not have enough PCI Express lanes to drive the IGD and a > x16 graphics card at the same time. > > -----Original Message----- > From: xen-devel-bounces@lists.xensource.com > [mailto:xen-devel-bounces@lists.xensource.com] On Behalf Of Beng Heng, > Ng > Sent: Thursday, June 04, 2009 11:09 AM > To: xen-devel > Subject: [Xen-devel] Experimental results for VGA passthrough > > Hi, > > I tried patching Xen 3.4 Testing c/s 19628 with vga passthrough code > (less HID passthrough), and did some tests. I made some modifications in > pci_setup to skip loading the emulated VGA bios and pci_load_option_roms > to load the passthrough VGA bios. My setup uses Intel DQ45CB and ATI > HD2600XT. > > The BDFs for the relevant devices are: > Intel IGD 00:02.0 > ATI gfx 01:00.0, 01:00.1 > There's 2 BDFs for ATI because one of them is for the Audio. Both have > to be passed through simultaneously. > > In the first scenario, Intel IGD is set up as the primary for Dom0. The > ATI gfx is passed through to DomU. Interestingly, it is the screen that > is connected to Intel's output that displays DomU (albeit with few > colors and low resolution). However, in the Device Manager, ATI is > reported. I believe this is because Intel's address space is being > mapped, instead of ATI's. Shutting down DomU does not return the display > to Dom0. The display remains at the last screen showing DomU shutting > down. Except for the shutting down phase, I've recorded the observations > in the first attachment. > > In the second scenario, Intel IGD is passed through to DomU. The ATI gfx > is thus absent. However, when DomU is booted, there is an IOMMU page > fault. This is shown in the second attachment starting from Line 659. > The system subsequently hangs. Does anyone have an explanation to what's > going on in this case? > > So, VGA passthrough works, but seems extremely flaky. Any > comments/suggestions to correct the problems observed will be greatly > appreciated. > > Thanks. > > - Beng Heng