[Pkg-xen-devel] Bug#988333: Bug#988333: Bug#988333: libxenmisc4.16: libxl fails to grant necessary I/O memory access for gfx_passthru of Intel IGD

[Chuck Zmudzinski]

On 3/8/2022 6:00 AM, Hans van Kranenburg wrote:
> On 3/7/22 18:30, Chuck Zmudzinski wrote:
>> [...]
>
> Thanks for adding all the info and researching this, Chuck!
>
> Hans

Thanks for the encouragement. Now I am working on two things:

1. I want to understand the problem better so I can Improve the patch. 
The patch I proposed for this bug currently adds two pages to the 
permitted I/O addresses for the domain starting at 0xcc490, but I don't 
know the best way to determine that memory address. Probably a 
hard-coded constant address is not good enough to handle all possible 
hardware affected by this bug. It might be different for different Intel 
revisions of the Intel IGD, and it probably can be read from somewhere, 
but where? Also, what is that memory for? I would like to understand 
that before finalizing what the patch should be. I am researching these 
things, and then I will have a better version of the patch to propose 
for Xen upstream. I have found places where Qemu/vfio deals with what 
they call Intel IGD quirks, and that information can shed light on how 
to solve this bug in Xen. Hopefully upstream will accept a patch when we 
have a patch to propose to them. I am fairly sure there really is a bug 
that should be patched upstream to improve support for the Intel IGD 
passthrough feature on Xen.

2. Now that I have IGD passthrough of the Intel IGD to Linux working on 
my box with the traditional Qemu device model and Debian's Xen 4.16 for 
unstable, I am working on getting passthrough to Linux working with the 
upstream Qemu 6.2 device model that comes with Debian, so this bug can 
more easily be tested and verified without needing to build the 
traditional Qemu device model for Debian.

Chuck