[vfio-users] IGD primary pci passthrough troubles - can anyone help?

Paul Handy paul.d.handy at gmail.com
Fri Nov 11 16:52:49 UTC 2016


Taking cues from
https://vfio.blogspot.com/2016/07/intel-graphics-assignment.html, and big
thanks to Alex for getting me this far, I had put it upon myself to try
passthrough of my HD graphics on my intel i7-6700k processor.
Unfortunately, I'm not getting it quite right, and I'm afraid that I've
missed something still, despite much searching.

*lspci -vvv:*
http://pastebin.com/2WF3pVk5

*uname -a:*
http://pastebin.com/5Te7Kbcc

*relevant kernel parameters I'm using:*
intel_iommu=on video=efifb:off,vesafb:off,simplefb:off

*modprobe.d:*
blacklist i915
options vfio-pci ids=8086:1912,8086:a170

*modules I've added to the initramfs:*
vfio
vfio_iommu_type1
vfio_pci
vfio_virqfd

*dmesg | grep -e DMAR -e IOMMU*:
http://pastebin.com/fpm0rywx

*dmesg | grep 0000:00:02.0*:
http://pastebin.com/VCV2sw9i

*cat /proc/iomem | grep -A 2 02.0:*
http://pastebin.com/Tg9AWmPe

*cat /proc/ioports*:
http://pastebin.com/eMNccFCc


*qemu-system-x86_64 --version:*QEMU emulator version 2.7.0(Debian
1:2.7+dfsg-1), Copyright (c) 2003-2016 Fabrice Bellard and the QEMU Project
developers


*my xml file:*http://pastebin.com/usY87nmN


As others have reported, I've gotten this message despite best efforts:
Failed to mmap 0000:00:02.0 BAR 2. Performance may be slow

This indicated to me that the at least part of graphics device (or its
memory) is still being grabbed by *something*, even though I have gone to
the extra effort of making sure that I turned off efifb,vesafb, and
simplefb. I should probably note that in my boot process, it gets to a
certain point, when the last change to the screen is adding the line:
device changed decodes:
PCI:0000:00:02.0,olddecodes=io+mem,decodes=io+mem:owns=io+mem

Along with that, my logs show a lot of these types of messages:
2016-11-10T21:03:29.361455Z qemu-system-x86_64:
vfio_region_write(0000:00:02.0:region2+0x4cbd4, 0x0,4) failed: Device or
resource busy

Surprisingly, SeaBIOS still shows on the display, with the text:
SeaBIOS (version 1.9.3-20161025_171382-gandalf)

Followed by a garbled mess:
https://imgur.com/rc9lK8x

And eventually, if the screen doesn't just go blank, I get this:
https://imgur.com/3gAQPDS

which is quite heartening, considering that means my VM is actually
somewhat able to detect and write to the integrated graphics, despite it
being the only graphics device available, and the mmap problem. Any help is
much appreciated!

Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20161111/60fca0a9/attachment.htm>


More information about the vfio-users mailing list