amdgpu: suspend-to-idle on a 4350G desktop

amdgpu: suspend-to-idle on a 4350G desktop

[Rafael Ávila de Espíndola]

Hi,

I have a desktop that doubles as a NAS, and would like to suspend it
when not in use. It works, but when it comes back it is using about 7W
more than before going to sleep.

It looks like a6ed2035878e5ad2e43ed175d8812ac9399d6c40 is the
culprit. The system doesn't even support S3:

% cat /sys/power/mem_sleep
[s2idle]

And looks like the vendor decided that they should not define
ACPI_FADT_LOW_POWER_S0. The MB is a Gigabyte B550I AORUS PRO AX with
bios version F16e.

Which systems had crashes when ACPI_FADT_LOW_POWER_S0 was not defined?
Would it be possible to limit the check only to those systems or at
least only to systems that have S3?

I can build the amdgpu module without that check and see if the problem
goes away if that would help.

Thanks,
Rafael

Re: amdgpu: suspend-to-idle on a 4350G desktop

[Limonciello, Mario]

On 1/21/2023 10:37, Rafael Ávila de Espíndola wrote:
> Hi,
> 
> I have a desktop that doubles as a NAS, and would like to suspend it
> when not in use. It works, but when it comes back it is using about 7W
> more than before going to sleep.
> 
> It looks like a6ed2035878e5ad2e43ed175d8812ac9399d6c40 is the
> culprit. The system doesn't even support S3:
> 
> % cat /sys/power/mem_sleep
> [s2idle]
> 
> And looks like the vendor decided that they should not define
> ACPI_FADT_LOW_POWER_S0. The MB is a Gigabyte B550I AORUS PRO AX with
> bios version F16e.
> 
> Which systems had crashes when ACPI_FADT_LOW_POWER_S0 was not defined?
> Would it be possible to limit the check only to those systems or at
> least only to systems that have S3?
> 
> I can build the amdgpu module without that check and see if the problem
> goes away if that would help.
> 
> Thanks,
> Rafael

Rafael,

Can you please open a bug report here
https://gitlab.freedesktop.org/drm/amd/-/issues

And attach the log from:
https://gitlab.freedesktop.org/drm/amd/-/blob/master/scripts/amd_s2idle.py

Thanks,

Re: amdgpu: suspend-to-idle on a 4350G desktop

[Rafael Ávila de Espíndola]

Will do.

Thanks,
Rafael

"Limonciello, Mario" <Mario.Limonciello@amd.com> writes:

> [AMD Official Use Only - General]
>
> Rafael,
>
> Can you please open a bug report here
> https://gitlab.freedesktop.org/drm/amd/-/issues
>
> And attach the log from:
> https://gitlab.freedesktop.org/drm/amd/-/blob/master/scripts/amd_s2idle.py
>
> If it fails prerequisites please also include a full dmesg including a suspend cycle.
>
> Thanks,
> ________________________________
> From: Rafael Ávila de Espíndola <rafael@espindo.la>
> Sent: Saturday, January 21, 2023 10:37:23 AM
> To: Limonciello, Mario <Mario.Limonciello@amd.com>
> Cc: linux-kernel@vger.kernel.org <linux-kernel@vger.kernel.org>
> Subject: amdgpu: suspend-to-idle on a 4350G desktop
>
> Hi,
>
> I have a desktop that doubles as a NAS, and would like to suspend it
> when not in use. It works, but when it comes back it is using about 7W
> more than before going to sleep.
>
> It looks like a6ed2035878e5ad2e43ed175d8812ac9399d6c40 is the
> culprit. The system doesn't even support S3:
>
> % cat /sys/power/mem_sleep
> [s2idle]
>
> And looks like the vendor decided that they should not define
> ACPI_FADT_LOW_POWER_S0. The MB is a Gigabyte B550I AORUS PRO AX with
> bios version F16e.
>
> Which systems had crashes when ACPI_FADT_LOW_POWER_S0 was not defined?
> Would it be possible to limit the check only to those systems or at
> least only to systems that have S3?
>
> I can build the amdgpu module without that check and see if the problem
> goes away if that would help.
>
> Thanks,
> Rafael

Re: amdgpu: suspend-to-idle on a 4350G desktop

[Rafael Ávila de Espíndola]

"Limonciello, Mario" <Mario.Limonciello@amd.com> writes:

> [AMD Official Use Only - General]
>
> Rafael,
>
> Can you please open a bug report here
> https://gitlab.freedesktop.org/drm/amd/-/issues

I reported the bug:
https://gitlab.freedesktop.org/drm/amd/-/issues/2364

Cheers,
Rafael