Jammy: NVIDIA Turing architecture firmware for Nouveau missing (TU117)

[Lassi Väätämöinen]

Black screen when logging out in KDE Neon.
Original bug report for KDE here:

https://bugs.kde.org/show_bug.cgi?id=473144

-------

Key points in the last comments:

I wonder could it be related to this linux-firmware issue with NVIDIA Turing arch:

https://gitlab.freedesktop.org/drm/nouveau/-/issues/207

https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/commit/?id=2e92a49f90f73c8edc44b25c6e669d5e70893c90

Current linux-firmware package in Jammy is:
20220329.git681281e4-0ubuntu3.17

So that would be slightly older than the one where the NVIDIA fix was introduced (3rd April)

---------

https://gitlab.freedesktop.org/drm/nouveau/-/issues/207
>Versions with this issue can boot with nomodeset kernel parameter.

I added 'nomodeset' to the "linux" -line in Grub for kernel 6.2 and now the logout works. So this looks like it very likely is the firmware issue with Ubuntu Jammy.

---------

So, this linux-firmware change would be needed:

https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmware.git/commit/?id=2e92a49f90f73c8edc44b25c6e669d5e70893c90

[Lassi Väätämöinen (lvaatamoinen)]

Syslog for the issue with 6.2 kernel. And testing 5.19 kernel as workaround.

[Juerg Haefliger (juergh)]

Have you verified, that the updated firmware from 2e92a49f90f73c8edc44b25c6e669d5e70893c90 actually fixes your problem (without 'nomodeset')?

[Juerg Haefliger (juergh)]

I don't think this upstream commit helps. You have NVIDIA TU117 (167000a1) and that commits updates tu102 and tu116 to support new HW.

[Lassi Väätämöinen (lvaatamoinen)]

>I don't think this upstream commit helps. You have NVIDIA TU117 (167000a1) and that commits updates tu102 and tu116

Unfortunately you are right. I just tested the FW from the commit and I would indeed need tu117 firmware. So this needs upstream solution.

[Lassi Väätämöinen (lvaatamoinen)]

Juerg, which information do you need? The bug is marked 'incomplete', but there is no information request pending.

The upstream commit 2e92a49f90f73c8edc44b25c6e669d5e70893c90 does not fix the issue, so new firmware seems to be needed.

[Timo Aaltonen (tjaalton)]

there are no updated firmware for tu117, not upstream or anywhere

[Juerg Haefliger (juergh)]

Clarification:

Some of the TU117 files are symlinks to TU116 so commit 2e92a49f90f73c8edc44b25c6e669d5e70893c90 does affect TU117:

$ grep tu117 WHENCE
Link: nvidia/tu117/acr/bl.bin -> ../../tu116/acr/bl.bin
Link: nvidia/tu117/acr/ucode_ahesasc.bin -> ../../tu116/acr/ucode_ahesasc.bin
Link: nvidia/tu117/acr/ucode_asb.bin -> ../../tu116/acr/ucode_asb.bin
Link: nvidia/tu117/acr/ucode_unload.bin -> ../../tu116/acr/ucode_unload.bin
Link: nvidia/tu117/acr/unload_bl.bin -> ../../tu116/acr/unload_bl.bin
Link: nvidia/tu117/gr/fecs_bl.bin -> ../../tu116/gr/fecs_bl.bin
File: nvidia/tu117/gr/fecs_data.bin
File: nvidia/tu117/gr/fecs_inst.bin
File: nvidia/tu117/gr/fecs_sig.bin
Link: nvidia/tu117/gr/gpccs_bl.bin -> ../../tu116/gr/gpccs_bl.bin
File: nvidia/tu117/gr/gpccs_data.bin
File: nvidia/tu117/gr/gpccs_inst.bin
File: nvidia/tu117/gr/gpccs_sig.bin
File: nvidia/tu117/gr/sw_bundle_init.bin
File: nvidia/tu117/gr/sw_ctx.bin
File: nvidia/tu117/gr/sw_method_init.bin
File: nvidia/tu117/gr/sw_nonctx.bin
File: nvidia/tu117/gr/sw_veid_bundle_init.bin
Link: nvidia/tu117/nvdec/scrubber.bin -> ../../tu116/nvdec/scrubber.bin
Link: nvidia/tu117/sec2/desc.bin -> ../../tu116/sec2/desc.bin
Link: nvidia/tu117/sec2/image.bin -> ../../tu116/sec2/image.bin
Link: nvidia/tu117/sec2/sig.bin -> ../../tu116/sec2/sig.bin