[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [Xen-users] [REQUEST] Request for Xen Users to Attempt Jean David Techer's Xen 4.2-unstable VGA Passthrough Documentation



On 03/29/2012 04:29 PM, Casey DeLorme wrote:
David,

XenServer VGA Passthrough requires a paid/licensed copy, which costs $2500, a bit out of my price range for experimentation.  Important to note that the feature is not a part of the 30-day trial license.

However, Citrix recently visited my college and I was able to preview hardware access on a laptop one of the employees had, where they swapped between Ubuntu and Windows with a hotkey, and various hardware components including onboard GPU and the WebCam were accessible.

In testing XenServer, I can say that if I had a business, that's the product I would use.  In the past month having tried Xen and ESXi, I was astonished with the ease of use for XenServer.

As for Catalyst, version 12.2 (the latest currently) worked for me.

Important to note that until I followed Andrews advice to omit the Catalyst Control Center, the installation resulted in a BSOD.
I saw similar issue whiling playing with XenClient. After discussing with AMD GPU driver team, the conclusion was that the installer has a bug. But I have not received any further update from them. Also manual driver installation (after many tries) did fix problem for me.

The solution, select "Custom" installation and uncheck the CCC.  After the installation your first reboot should run some follow-up updates via cmd, you need to reboot a second time for fully functional drivers.

Also, I had underscan on my monitor so I went out on a limb and re-ran the setup for Catalyst, and was able to get CCC installed with a second run through, which allowed me to fix my underscan issue.

My conclusion is that the CCC requires some driver functionality that isn't available until after you install the drivers, this could be on all systems or it might be related to how HVM's handle the PCI devices, that much I can't say.



Teo,

I could be spouting nonsense, and if so I'm sure Wei can correct me, but I am pretty sure AMD engineers have been contributing to Xen for a while, and some patches have already been applied.  Obviously it isn't flawless, I myself haven't gotten video at boot time, only at the login screen.

This is because VBIOS patch wasn't applied. But as I said before, my VBIOS wasn't universal enough to put it as a production patch. So I am hesitant to put it out.
Mine works on 4.1.2, but it is possible that 4.1.0 had less of these "patches" hence Sebastien's post.

Also, I apologize as I did not properly word my opinion before.  VGA Passthrough is new "for consumer components".  In 2010 the number of desktop (not server) boards boasting VT-d functionality could probably be counted on one hand.  To my understanding that means the technology is at most 3 years old, still a baby in my opinion.

I didn't mean that the technology hadn't been implemented into various Hypervisors, just that it is clearly not a perfected feature.  If you consider 3 years of consumer availability, dates become important when researching.  Sebastien's post was May 2011, just shy of one year ago, and Thomas's was in 2010.

There are newer patches still for ATI in Xen 4.2, which I intend to test over the next week.  I have NOT gotten ATI to work at boot time, video starts at the login screen.

I agree with Wei that drivers can contribute to BSOD's and errors, but when an install doesn't fail but the hardware reacts the same as before, I would like to assume the driver is irrelative.

Have you looked at XenClient project? It has a customized ATI component which allows you to switch between VMs flawlessly. I think it is the most mature Xen solution for GPU passthru in client area.

~Casey

On Thu, Mar 29, 2012 at 4:11 PM, Wei Huang <wei.huang2@xxxxxxx> wrote:
On 03/29/2012 01:35 PM, Teo En Ming (Zhang Enming) wrote:
Dear Casey DeLorme,

This guy, Sebastien Gauthier, also has the same problems as us. He was using Xen 4.1.0 and an ATI Radeon 4550. He applied Mr. Wei Huang's patch. After installing the latest ATI/AMD Catalyst drivers, he got a BSOD with Xen VGA Passthrough to Windows 7.

Please read Sebastien Gauthier's case here:

http://readlist.com/lists/lists.xensource.com/xen-users/11/59090.html

Hence Sebastien Gauthier reported Xen VGA Passthrough with PARTIAL SUCCESS, like us.

Dear Tobias Geiger,

You were saying that with ATI VGA cards, you do not need to apply any Xen VGA Passthrough patches. But this guy Sebastien Gauthier applied Mr. Wei Huang's patches to Xen 4.1.0 and got a BSOD after installing ATI Catalyst drivers. Sebastien Gauthier did not get 100% success with Xen VGA Passthrough to Windows 7 using an ATI VGA card.

Hi Teo,

The VBIOS patch I sent out did not work for all ATI cards. The patch itself assumed certain behavior of GPU VBIOS. But this doesn't apply to every GPU generation. From this perspective, my patch isn't universal. Also there are many factors, some of which are not in control by us (like graphics driver), can contribute to BSOD you mentioned. I am not in a position to debug it for everyone's card (as I don't have all cards).

Thanks,
-Wei

-- 
Yours sincerely,

Mr. Teo En Ming (Zhang Enming)
Singapore




On 29/03/2012 23:50, Teo En Ming (Zhang Enming) wrote:
On 29/03/2012 12:29, Casey DeLorme wrote:
My mistake for not hitting "Reply to All", sorry.

No worries Casey.


It might be of some value to mention that my tests were with Windows 7, I have no interest in using XP anymore.  Also, I did all of my testing through remote VNC, not once did I actually get video, even 2D, working. 

I bought 2 copies of Windows XP Home Edition in the past because it is cheap, at S$145. I did not buy Windows 7 because it is expensive. I got video working all right, but only 2D. I cannot get 3D graphics working.

However, my errors are exactly as you described:

1.  Windows recognizes the model (GTX 460), not "unknown PCI device".
2.  Device code 43.
3.  No resources assigned to the device.

We have the same set of errors.


Might be worth mentioning, I was able to run the latest nVidia driver installation without errors, but after rebooting there was no change, same error code, still no video.

Same situation here with the latest NVIDIA drivers. But I only have 2D video working. I can't get 3D video to work.

 To me this would be evidence that driver version isn't a problem, but then again I didn't have anything actually working.

I agree that driver version isn't a problem. Something is wrong somewhere.



I am an IT student in college, so my experience is limited to mostly programming with very little knowledge of the inner-workings of hardware.  So, forgive me for being unable to help with regards to memory addresses on the cards.

I am hoping that Intel engineers and Xen developers would be able to help.


I've been using *nix for about 4 years, and Windows since I my age was a single digit.  I have had experience setting up server features such as web, database, and application servers but I am still green when it comes to kernel or hardware configuration.

I started learning Linux/UNIX since the year 2005, that is about 7 years ago. I started using Windows when it was version 3.1. I love compiling the latest Linux kernel and assembling my own computer hardware.


My Xen adventure began 35 days ago, and it is no exaggeration to say that in those 35 days I have learned more about linux than I had in the past two years.  I like challenges as much as the next IT person, but I ran out of time and ideas for debugging my nVidia problems.  The nVidia stretch of my adventure lasted for 24 days through 54 fresh linux installations accompanied by over 200~ pages worth of documented failures and not a single pixel sighted.

Why not make your documentation into PDF? It is a very popular document format.


The ATI card took me a day, less than 12 hours of relatively easy debugging by comparison to the aforementioned testing.  I do fully understanding financial constraints, but it is a working solution and worth mentioning.  I do not know what the prices are like in Singapore, but in the states I was able to buy an ATI Radeon HD 6870 for $160.  For me it came down to weighing my objectives against my curiosity.

That ATI Radeon HD 6870 would cost me $279, I think.


If you do continue your nVidia endeavors I wish you success, but as in my former email VGA passthrough is a new frontier, not even Guru's like David may not be able to help beyond their own hardware experiences.

I don't think VGA passthrough is a new frontier. Oracle VirtualBox and Linux KVM supports VGA passthrough as well. Xen ATI VGA Passthrough works out of the box, as Tobias Geiger suggested, but NVIDIA requires patches.



Documentation is one problem I agree with you on 100%.  I came into this knowing relatively little about Linux & hardware, and nothing about Xen, and most every guide I found assumed I had been in a deep relationship with Linux for many years and had a basic understanding of Xen commands.

I started learning Xen since the year 2007, which is about 5 years ago.


Like you, I intend to use those documented failures as well as my recent success to create a comprehensive guide with photographs, screen captures, and perhaps even videos going from "assembled computer" to "Complete Xen Dom0 /w HVM & VGA Passthrough".  Provided the wiki will allow me to upload the screenshots, I'll be certain to post it there.

I will be looking forward to your documentation and videos. Right now Xen wiki allows uploading image files and PDF files. Why not create a PDF document and share it with all of us? It is known as portable document format and is very popular, but it appears that xen mailing lists don't like PDF format.

I don't like wiki pages because anybody can edit and fundamentally mess up the wiki pages, even providing bogus and erroneous information. That's why I don't like creating wiki pages. Anybody can edit and mess up the information you have painstakingly created on the wiki pages. So please take note.

-- 
Yours sincerely,

Mr. Teo En Ming (Zhang Enming)
Singapore



~Casey

On Wed, Mar 28, 2012 at 10:21 PM, Teo En Ming (Zhang Enming) <singapore.mr.teo.en.ming@xxxxxxxxx> wrote:
Dearest Casey DeLorme,

Thank you very very much for your kind feedback and input. I would also like to thank Mr. Tobias Geiger, again, for providing his suggestion on exposing the fourth memory region in tools/firmware/hvmloader/acpi/dsdt.asl. In any case, either exposing the first 3 memory regions only or exposing all the 4 memory regions does not work. Sadly, Tobias Geiger is unable to help me further.

I have asked Jean David Techer, what about the 4th PCI memory region? Why only expose the first 3 PCI memory regions? I don't understand, of course. Jean David Techer did not reply to my question.

I have decided to post your prompt reply to the xen-users and xen-devel mailing lists, in case people think that I am finding fault with Jean David Techer, or trying to irritate him, or trying to make him angry, or trying to aggravate him. Jean David Techer replied me with an email saying that I spent too much time and too bent on solving the yellow exclamation mark glitch for my NVIDIA Geforce 8400GS in Device Manager in Windows 8 Consumer Preview and Windows XP Home Edition, and that I sent stupid requests. Stupid requests? Did he read my emails carefully, word by word?

Casey DeLorme, please, can I confirm with you again that you are getting the following errors after applying Jean David Techer's Xen 4.2-unstable VGA Passthrough patches:

(1) Yellow exclamation mark besides your NVIDIA GTX 460 in Device Manager
(2) Windows has stopped this device because it has reported problems. (Code 43)
(3) This device isn't using any resources because it has a problem.


Jean David Techer insists that our technical issues are due to a NVIDIA driver problem. He insists that you have to install NVIDIA driver versions 275.33 WHQL and 275.50 BETA. Any other NVIDIA driver versions (above 280.XX) will not work, according to Jean David Techer. However, I have tried installing NVIDIA driver versions 275.33 and 275.50 from www.softpedia.com, as he suggested, but it caused my Windows XP Home Edition HVM virtual machine to be destroyed/terminated/crash after a few minutes and my dom0 to crash as well. NVIDIA driver versions 275.33 and 275.50 for Windows XP 32-bit is not available from the official NVIDIA website.

So it is definitely not a NVIDIA driver problem. I suspect that the technical issue has to do with MMIO BARs pBAR:vBAR 1:1 matching. I don't think there is any problem with vgabios-pt.bin extracted out from our NVIDIA VGA cards, because I have performed a "hexdump -C" on my extracted VGA BIOS EEPROM, or Electrically Erasable Programmable Read Only Memory.

Secondly, it does seem strange that Jean David Techer was able to attain 100%, ie. perfect success with Xen 4.2-unstable VGA Passthrough to his Windows XP 32-bit and 64-bit HVM domU. Have you watched his Youtube video? It is only 4 minutes. Please do watch Jean David Techer's Youtube video at the following URL:

Jean David Techer's Xen 4.2-unstable VGA Passthrough to Windows XP x64 HVM domU Youtube video link: http://www.youtube.com/watch?v=3SaYO0ERW44

I am appalled and baffled that he has attained 100% success while both of us have only attained partial success (i.e. less than 100%) on Xen 4.2-unstable VGA Passthrough to Windows 8 Consumer Preview and Windows XP.

Solving the yellow exclamation mark issue is important because we would not be able to run 3D graphics benchmarks and play 3D games without solving it. I am not sending silly emails about some yellow marks, as Jean David Techer suggested. I can't even run Unigine Heaven DX11, and 3dmark11 3D display benchmarks, because of the yellow exclamation mark for NVIDIA Geforce 8400 GS in Device Manager.

Casey DeLorme, with your report on relatively easy success with ATI VGA cards, I think I would go the ATI way, but I would have to spend a few hundred dollars compared to my cheap SGD$44 NVIDIA Geforce 8400 GS card. And while deciding to go the ATI way, I would also like to continue troubleshooting with the NVIDIA problem, because I consider it to be a technical challenge.

In essence, Jean David Techner is considered to be a "boss", or business owner, or proprietor, or technopreneur, or entrepreneur, or technical support officer, or customer support officer, or IT helpdesk engineer, providing services like his forward-ported Xen 4.2-unstable VGA Passthrough patches and the documentation on his blog. I repost Jean David Techer's official website here:

Jean David Techer's Xen 4.2-unstable VGA Passthrough blog: http://www.davidgis.fr/blog/index.php?2011/12/07/860-xen-42unstable-patches-for-vga-pass-through

Jean David Techer's official website is his business venture.

Basically, I am Jean David Techer's "customer", trying to obtain technical support from him. Of course, he is not obliged to provide technical support to me since he is providing free services. It is, after all, an open source software project. Nobody is obliged to provide anybody with technical support. To do Jean David Techer justice, he replied most of my questions while avoiding some of my questions.

Finally, I have also failed to obtain technical support from Xen developers like Ian Campbell from Citrix Corporation and Konrad Wilk from Oracle Corporation. I have always provided all the steps which I have taken, the configuration files and necessary documentation, and kernel messages and error logs to xen-users and xen-devel mailing lists, but they keep insisting I did not provide the information they required. I wondered why. I think they did not read my emails carefully. They told me they would not reply to me any more if I do not provide the information they requested. But the problem is that I have always provided information they requested! I think they missed some of my emails, or did not read my emails carefully enough. I am an ardent supporter and SERIOUS software tester for open source Xen virtualization/hypervisor but they treated me lightly. I always read my emails WORD BY WORD. I have even went to the point of making a video on the BUG and uploading my video to Youtube. The video is only THREE minutes.

As everybody says, a picture is worth a thousand words. A video is worth a BILLION words!

I have also failed to obtain technical support from Xen developers regarding Xen 4.2-unstable VGA Passthrough.

I am hoping Xen 4.2 would have official support for Xen VGA Passthrough for both NVIDIA and ATI cards.

Casey DeLorme, thank you very much once again. I will be making changes to my Xen, Linux Kernel and Xen VGA Passthrough Documentation and will be releasing Version 1.7 shortly. Jean David Techer's documentation assumes some level of advanced Linux technical knowledge, so I am writing documentation on my own so that everybody, not just advanced Linux and Xen users, can follow. I have made references to Jean David Techer's documentation in my own documentation.

I would be very happy if people would use my documentation. Of course, it satisfies my ego and my vanity. Haha.

I have been un-employed for nearly three years now, and I would hesitate to spend a few hundred dollars on an ATI VGA card. I quit my job as an IT engineer 3 years ago because my father suffered from lacunar infarct, or more commonly known as stroke. My NVIDIA Geforce 8400 GS costs only S$44. Please understand why I hesitate to buy an ATI VGA card. The cheapest one costs SGD$279.

I have a diploma in Mechanical+Electronics engineering from Singapore Polytechnic and a Bachelor's degree in Mechanical Engineering from the National University of Singapore. But I do not have qualifications in Computer Science or Information Technology. I have worked as an Information Technology engineer in Defense Science and Technology Agency, Ministry of Defense, Singapore, National Computer Systems Pte Ltd, Asiasoft Online Pte Ltd, and Ishinemax Singapore Pte Ltd.

Google search terms: Frenchman Jean David Techer, Singaporean Teo En Ming's Xen, Linux Kernel and Xen VGA Passthrough Documentation, Xen 4.2-unstable VGA Passthrough to Windows 8 Consumer Preview and Windows XP HVM Virtual Machines

Thank you very much for reading my lengthy email. I am always courteous, saying "Please help me. Please. Please. Please." and "Thank you very much for your kind assistance" in my emails.

Thank you very much.

--
Yours sincerely,
Mr. Teo En Ming (Zhang Enming)
Singapore Citizen

cc: His Excellency The Prime Minister Mr. Lee Hsien Loong, Prime Minister's Office, Republic of Singapore

On 29/03/2012 03:53, Casey DeLorme wrote:
Hi Teo,

I tried David's patch files a while ago without success.  I had Xen compiled with the patch files and my GTX 460 VGA BIOS rom, but I got the same as you, either a BSOD or Code 43 in Device Manager.

You sound plenty competent, but it's important to remember that you are pioneering a technology that for consumers is still in its infancy.  Very few people are testing this with consumer equipment, so finding results seems to be a rarity.





_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users








_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.