NACK: [PATCH 0/5][SRU][Lunar] System hangs at S3 for amdgpu Navi 31/Navi33

[Roxana Nicolescu]

On 07/07/2023 09:54, You-Sheng Yang wrote:
> BugLink: https://bugs.launchpad.net/bugs/2024427
>
> [Impact]
>
> Under stress testing it was reported that the amdgpu Navi31/Navi33 platforms
> will sometimes fail to wake from S3.
>
> [Fix]
>
> kernel patches:
> ac2f5739fdca drm/amdgpu/mes11: enable reg active poll
> a2fe4534bb38 drm/amd/amdgpu: update mes11 api def
> da9a8dc33da2 drm/amdgpu: reserve the old gc_11_0_*_mes.bin
> 616843d5a11b drm/amd/amdgpu: introduce gc_*_mes_2.bin v2
> 09bf14907d86 drm/amdgpu: declare firmware for new MES 11.0.4
>
> firmware patches:
> * Navi31: ffe1a41e2ddb amdgpu: update GC 11.0.0 firmware for amd.5.5 release
> * Navi33: a5d7b4df1a76 amdgpu: update GC 11.0.2 firmware for amd.5.5 release
>
> Three prerequisite commits for backporting to v6.2 (linux/lunar):
> * cc42e76e7de5 drm/amd: Load MES microcode during early_init
> * 2210af50ae7f drm/amd: Add a new helper for loading/validating microcode
> * 11e0b0067ec0 drm/amd: Use `amdgpu_ucode_*` helpers for MES
>
> [Test Case]
>
> $ checkbox-cli run com.canonical.certification::stress-suspend-30-cycles-with-reboots-automated
>
> [Where problems could occur]
>
> Little we know about the firmware fixes. However, while with these commits have
> been pulled via stable kernel fixes, the driver begins to request new firmware
> blobs of a different filename.
>
> [Other Info]
>
> The kernel driver commits are in v6.4-rc1, backported to v6.3.4, v6.1.31, and
> partially (missing da9a8dc33da2, 616843d5a11b) v6.2.16. Only linux/lunar has to
> be fixed.
>
> For the firmware parts, they have been included in linux-firmware/mantic,
> leaving linux-firmware/lunar and linux-firmware/jammy to be fixed.
>
> Jack Xiao (1):
>    drm/amd/amdgpu: introduce gc_*_mes_2.bin v2
>
> Li Ma (1):
>    drm/amdgpu: reserve the old gc_11_0_*_mes.bin
>
> Mario Limonciello (3):
>    drm/amd: Load MES microcode during early_init
>    drm/amd: Add a new helper for loading/validating microcode
>    drm/amd: Use `amdgpu_ucode_*` helpers for MES
>
>   drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c   |  77 ++++++++++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h   |   1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c |  36 ++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h |   3 +
>   drivers/gpu/drm/amd/amdgpu/mes_v10_1.c    | 107 ++++------------------
>   drivers/gpu/drm/amd/amdgpu/mes_v11_0.c    | 103 +++++----------------
>   6 files changed, 158 insertions(+), 169 deletions(-)
>
These were already applied with upstream stable patches.

Roxana