amd-sfh: 10s resume delay from s2idle due to sensor status timeout

amd-sfh: 10s resume delay from s2idle due to sensor status timeout

[Regölö Maass]

When closing the lid and reopening it again the screen is frozen for around 10s seconds before any keyboard input is recognized.
That results in every sensor being in a broken state i.e. "monitor-sensor" shows the accelerometer orientation as "bottom-up" and tilt as "undefined" and the ALS shows very low values.
The broken states of the sensors are: status 0x5 for sid 0x0 (accelerometer), status 0x4 for sid 0xf (operating-mode) and status 0x4 for sid 0x16 (ALS).

As on my current system the operating-mode is not recognized in userspace (No SW_TABLET_MODE Event) I can not be sure about it's state.
This resume bug happens consistently on every try based on my experience.

Here's the filtered dmes output:
[So, 16. Aug 2026, 13:30:49] PM: suspend entry (s2idle)
[So, 16. Aug 2026, 13:30:53] amdgpu 0000:03:00.0: dpm has been disabled
[So, 16. Aug 2026, 13:30:53] amdgpu 0000:03:00.0: SMU is resumed successfully!
[So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 0x0 (accelerometer) status 0x5
[So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 0xf (operating-mode) status 0x4
[So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 0x16 (ALS) status 0x4
[So, 16. Aug 2026, 13:31:03] PM: suspend exit

As the dmesg output shows the resume operations for the sensor are delayed every time by exactly 10 seconds.

I managed to solve the issue about the frozen screen by disabling the amd_sfh driver via "modprobe -r amd_sfh", after which the resume operation was nearly instant and no freezing occurred.

I updated the BIOS version to the latest available version: TN3604YA.306.

Importantly I'm using the "mem_sleep_default=s2idle" flag as a grub argument.

I also opened a bug-report in bugzilla as a reference:
https://bugzilla.kernel.org/show_bug.cgi?id=221895

Kernel infos: Linux xxx-xxx 7.1.8-arch1-3 #1 SMP PREEMPT_DYNAMIC Tue, 11 Aug 2026 09:16:08 +0000 x86_64 GNU/Linux
HW: Asus Vivobook S 16 Flip (TN3604YA), AMD Ryzen 5 7530U

Happy to provide more info if needed or test patches.

Re: amd-sfh: 10s resume delay from s2idle due to sensor status timeout

[Regölö Maass]

Update: As I've mentioned before the operating-mode was also affected. 
But the operating-mode (SW_TABLET_MODE) is now working correctly in 
userspace after testing 7.3-rc-1.

But the resume-delay still is present, including the operating-mode 
which comes unavailable after resume.

I'm willing to test new patches if they come available.


On 8/19/26 13:37, Regölö Maass wrote:
> When closing the lid and reopening it again the screen is frozen for 
> around 10s seconds before any keyboard input is recognized.
> That results in every sensor being in a broken state i.e. 
> "monitor-sensor" shows the accelerometer orientation as "bottom-up" 
> and tilt as "undefined" and the ALS shows very low values.
> The broken states of the sensors are: status 0x5 for sid 0x0 
> (accelerometer), status 0x4 for sid 0xf (operating-mode) and status 
> 0x4 for sid 0x16 (ALS).
>
> As on my current system the operating-mode is not recognized in 
> userspace (No SW_TABLET_MODE Event) I can not be sure about it's state.
> This resume bug happens consistently on every try based on my experience.
>
> Here's the filtered dmes output:
> [So, 16. Aug 2026, 13:30:49] PM: suspend entry (s2idle)
> [So, 16. Aug 2026, 13:30:53] amdgpu 0000:03:00.0: dpm has been disabled
> [So, 16. Aug 2026, 13:30:53] amdgpu 0000:03:00.0: SMU is resumed 
> successfully!
> [So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 0x0 
> (accelerometer) status 0x5
> [So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 0xf 
> (operating-mode) status 0x4
> [So, 16. Aug 2026, 13:31:03] pcie_mp2_amd 0000:03:00.7: resume sid 
> 0x16 (ALS) status 0x4
> [So, 16. Aug 2026, 13:31:03] PM: suspend exit
>
> As the dmesg output shows the resume operations for the sensor are 
> delayed every time by exactly 10 seconds.
>
> I managed to solve the issue about the frozen screen by disabling the 
> amd_sfh driver via "modprobe -r amd_sfh", after which the resume 
> operation was nearly instant and no freezing occurred.
>
> I updated the BIOS version to the latest available version: TN3604YA.306.
>
> Importantly I'm using the "mem_sleep_default=s2idle" flag as a grub 
> argument.
>
> I also opened a bug-report in bugzilla as a reference:
> https://bugzilla.kernel.org/show_bug.cgi?id=221895
>
> Kernel infos: Linux xxx-xxx 7.1.8-arch1-3 #1 SMP PREEMPT_DYNAMIC Tue, 
> 11 Aug 2026 09:16:08 +0000 x86_64 GNU/Linux
> HW: Asus Vivobook S 16 Flip (TN3604YA), AMD Ryzen 5 7530U
>
> Happy to provide more info if needed or test patches.
>