intel a750 on linux, module loads but no file in /dev

[ct_clmsn]

I've recently installed ubuntu and wanted to get an A750 gpu working with the environment.

Great news! The ubuntu installer detected the accelerator and automatically added the `xe` driver and the associated intel audio driver.

Ran into a blind spot! I was expecting to see a `/dev/dri` file and that file doesn't exist. Was wondering if anyone had advice or guidance on how to debug and resolve this issue?

Additionally, should I bail on the `xe` kernel driver and fall back to the `i915` driver with dkms support? Specifically the driver at this link -> https://github.com/intel-gpu/intel-gpu-i915-backports

If there's a "better path" to follow I'm open to hear about it!

I've included output from uname, lsb_release, lspci, and a modified version of the lsmod output.

uname

6.8.0-55-generic

lsb_release

Distributor ID: Ubuntu
Description: Ubuntu 24.04.1 LTS
Release: 24.04
Codename: noble

lspci

09:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A750] (rev 08)

lsmod

Module Size Used by
xe 3649536 0
drm_gpuvm 49152 1 xe
drm_exec 16384 2 drm_gpuvm,xe
gpu_sched 73728 1 xe
snd_hda_intel 73728 0
drm_buddy 36864 1 xe
snd_intel_dspcfg 32768 1 snd_hda_intel
video 57344 1 xe
drm_suballoc_helper 16384 1 xe
drm_ttm_helper 12288 1 xe
snd_hda_codec 253952 1 snd_hda_intel
ttm 118784 2 drm_ttm_helper,xe
drm_display_helper 266240 1 xe
snd_hda_core 167936 2 snd_hda_intel,snd_hda_codec
snd_hwdep 20480 1 snd_hda_codec
snd_pcm 225280 3 snd_hda_intel,snd_hda_codec,snd_hda_core
snd_timer 53248 1 snd_pcm
snd 159744 5 snd_hwdep,snd_hda_intel,snd_hda_codec,snd_timer,snd_pcm
i2c_algo_bit 16384 1 xe
backlight 32768 2 video,xe
soundcore 20480 1 snd​

[VonM_Intel]

Hi, ct_clmsn.

It sounds like you're running into an issue where the "/dev/dri" directory is not showing up, even though the system has detected the Intel Arc A750 and loaded the xe driver. Could you confirm whether the "/dev/dri" directory exists at all, or if it’s just missing the specific files (e.g., "/dev/dri.card0")? Are there any error messages related to the GPU or display drivers when you boot up the system?

By getting answers to these questions, I will have a better idea of where the problem lies, whether it's a driver issue, kernel compatibility, or something else.

Additionally, I'd like to delve deeper and verify specifics like the precise driver versions, and any system errors. Utilizing the Intel SSU tool will enable us to collect detailed information about your system setup, which could aid in identifying the underlying cause of this issue. To ensure your privacy and security, I recommend not sharing the Intel SSU logs on public forums, as they contain sensitive information about your system. Instead, please send the SSU logs to me via email. Once I receive the logs, I'll be able to further investigate the issue and assist you more effectively. After that, we can continue our communication in the community forum if needed.

Thank you. Have a nice day!

Best regards,
Von M.
Intel Customer Support Technician

[ct_clmsn]

Von,

Thank you for the quick response! I'm not able to see a `/dev/dri` or a `/dev/drm` directory.

Just ran `journalctl` and saw this in the log:

kernel: xe 0000:09:00.0: Your graphics device 56a1 is not officially supported
by xe driver in this kernel version. To force Xe probe,
use xe.force_probe='56a1' and i915.force_probe='!56a1'
module parameters or CONFIG_DRM_XE_FORCE_PROBE='56a1' and
CONFIG_DRM_I915_FORCE_PROBE='!56a1' configuration options.

I take it updating the kernel module's parameter is as simple as the following?

modprobe xe xe.force_probe='56a1'

I tried creating a file `/etc/modprobe.d/xe.conf` with the following contents:

options xe force_probe='56a1'

Just rebooted the system and managed to get the same error message about the device (56a1) not being supported by the ubuntu version's kernel xe module. As you'd expect `/dev/dri/` and/or `/dev/drm` are still not present.

On the upside, I did see this in the `journalctl` log

kernel: pci 0000:09:00.0: [8086:56a1] type 00 class 0x030000 PCIe Endpoint

which confirms the device information the kernel is communicating.

Would you advise installing the i915 dkms kernel module/driver? i915 isn't installed on my system currently. Or would a kernel recompile with an updated version of the xe driver be a better path forward?

v/r,

ct

[ct_clmsn · Accepted solution]

Quick update.

I was able to get a directory `/dev/dri` by updating `/etc/default/grub` with the following modification:

GRUB_CMDLINE_LINUX_DEFAULT="xe.force_probe='56a1'"

Then ran `update-grub` and rebooted the machine.

After the reboot, the machine is now reporting the following:

$ ls /dev/dri/
by-path  card0  renderD128

Think this means the machine is now "in business"!

v/r,

ct

[VonM_Intel]

Hello, ct_clmsn.

Thank you for the detailed update. It's great to hear that you're making progress. It looks like updating the kernel parameter via /etc/default/grub was a good approach, and now that you've got /dev/dri/ showing up, it seems like the system is starting to properly detect the graphics device. Thank you for taking the time to share this fix, and hopefully, it may help other community members experiencing similar behavior. Since the thread is now solved, I will proceed to close it. Please keep in mind that this thread will no longer be monitored by Intel.

Thank you. Have a nice day!

Best regards,

Von M.

Intel Customer Support Technician