TLB Invalidation time out on i915 SR-IOV passthrough

[MARDI Youness]

Hello,

Could you help us on this issue: https://github.com/intel/linux-intel-lts/issues/54

Host environment
Operating system: Gentoo Base System release 2.14
OS/kernel version: https://github.com/intel/linux-intel-lts/tree/lts-v6.6.34-linux-240626T131354Z
Architecture: x86_64
QEMU flavor: qemu-system-x86_64
QEMU version: latest qemu (master branch)
CPU: 12th Gen Intel(R) Core(TM) i7-1270P
igpu: Alder Lake-P
firmware: https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/snapshot/linux-firmware-20241110.tar.gz

Emulated/Virtualized environment
Operating system: Windows 10 21H1


Description of problem
After setting up SR-IOV (kernel compilation, kernel cmdline, vfio-pci driver attribution to the new pci..)
I've got my two new pci.


00:02.0 VGA compatible controller: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c)
DeviceName: Onboard IGD

Subsystem: Hewlett-Packard Company Alder Lake-P Integrated Graphics Controller
Kernel driver in use: i915

00:02.1 VGA compatible controller: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c)
Subsystem: Hewlett-Packard Company Alder Lake-P Integrated Graphics Controller
Kernel driver in use: vfio-pci

00:02.2 VGA compatible controller: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c)
Subsystem: Hewlett-Packard Company Alder Lake-P Integrated Graphics Controller
Kernel driver in use: vfio-pci

I gave one of those pci to my VM with this qemu cmdline:

-cpu host,migratable=on,hv-time,hv-relaxed,hv-vapic,hv-spinlocks=0x1fff,hv-passthrough,hv-vendor-id=IrisXE
...
-device vfio-pci-nohotplug,host=0000:00:02.1,id=hostdev0,bus=pci.4,addr=0x0

Sometimes it working properly when I start the qemu cmdline but most of the time I've got those kernel errors and a GPU hang:

    kernel [ 2252.208134] i915 0000:00:02.0: [drm] ERROR GT0: GUC: TLB invalidation response timed out for seqno 9679
    kernel [ 2252.208134] i915 0000:00:02.0: [drm] ERROR GT0: GUC: TLB invalidation response timed out for seqno 9679
    kernel i915 0000:00:02.0: [drm] ERROR GT0: GUC: TLB invalidation response timed out for seqno 9679
    kernel i915 0000:00:02.0: [drm] ERROR GT0: GUC: TLB invalidation response timed out for seqno 9679
    ....
    kernel Fence expiration time out i915-0000:00:02.0:renderThread22381:6e0!
    kernel i915 0000:00:02.0: [drm] GT0: GuC firmware i915/adlp_guc_70.bin version 70.13.1
    kernel i915 0000:00:02.0: [drm] GT0: HuC firmware i915/tgl_huc.bin version 7.9.3
    kernel i915 0000:00:02.0: [drm] GT0: HuC: authenticated for all workloads
    kernel i915 0000:00:02.0: [drm] GT0: GUC: submission enabled
    kernel i915 0000:00:02.0: [drm] GT0: GUC: SLPC enabled
    kernel [ 2730.991019] i915 0000:00:02.0: [drm] GPU HANG: ecode 12:1:85dfbfff, in renderThread [22381]
    kernel [ 2730.991084] i915 0000:00:02.0: [drm] renderThread22381 context reset due to GPU hang


It mostly appears when Qemu is starting..
Any help would be appreciated, thanks a lot

Best Regards,

Youness MARDI



C2 - Usage restreint
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx/attachments/20250128/a95e8750/attachment-0001.htm>