No drivers detected for TU104GL [T4G] card

[Joao Andre Simioni]

ubuntu-drivers is not returning any driver for g5g instances in AWS, which use Tesla T4G cards:

# lspci -tv
-[0000:00]-+-00.0 Amazon.com, Inc. Device 0200
           +-01.0 Amazon.com, Inc. Device 8250
           +-04.0 Amazon.com, Inc. NVMe EBS Controller
           +-05.0 Amazon.com, Inc. Elastic Network Adapter (ENA)
           \-1f.0 NVIDIA Corporation TU104GL [T4G]

# ubuntu-drivers --gpgpu list
This is gpgpu mode

# uname -a
Linux ip-172-31-16-212 6.8.0-1008-aws #8-Ubuntu SMP Sat Apr 20 02:43:14 UTC 2024 aarch64 aarch64 aarch64
GNU/Linux

# lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 24.04 LTS
Release: 24.04
Codename: noble

Processor:
Handle 0x0004, DMI type 4, 42 bytes
Processor Information
        Socket Designation: CPU00
        Type: Central Processor
        Family: ARMv8
        Manufacturer: AWS
        ID: C1 D0 3F 41 00 00 00 00
        Signature: Implementor 0x41, Variant 0x3, Architecture 15, Part 0xd0c, Revision 1
        Version: AWS Graviton2

I tried Noble and Jammy and both have the same behavior.

Manually installing the drivers works as expected (below is the output in Jammy, as Noble is hitting https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-550-server/+bug/2062380 and freezes on module loading)

$ sudo nvidia-smi
Tue Jun 4 14:18:15 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.161.08 Driver Version: 535.161.08 CUDA Version: 12.2 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 NVIDIA T4G Off | 00000000:00:1F.0 Off | 0 |
| N/A 36C P0 27W / 70W | 2MiB / 15360MiB | 4% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+

+---------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=======================================================================================|
| No running processes found |
+---------------------------------------------------------------------------------------+

[Joao Andre Simioni (jasimioni)]

sos report of the jammy host