Re: [PATCH] media: amd: isp4: add system suspend/resume support

[Kinn Julião]

will download the kernel again tonight. I've downloaded the rc2 and it's not booting anymore. will nomodeset later and check the logs.

That said, I just realized the driver in 6.19.6 is kinda flaky too (Kamoso screenshot attached).

On Thu, Mar 5, 2026 at 5:22 PM Nirujogi, Pratap <pnirujog@xxxxxxx> wrote:

Hi Kinn,

Thanks for reporting this issue. 

I tried reproducing it on kernel v7.0‑rc1 but ran into other problems, where the amdisp device does not probe during boot.

Have you modified isp_genpd_add_device() [1] to avoid the kernel panic and boot the system successfully? And then explicitly loaded amd_isp4_capture device using modprobe? If so, this may bring up the camera, but suspend-resume may still not work, as the amdisp device registration with genpd may not be completing correctly.

Can you please share the exact steps you are using to reproduce suspend-resume issue on v7.0‑rc1.

[1] https://github.com/torvalds/linux/blob/master/drivers/gpu/drm/amd/amdgpu/isp_v4_1_1.c#L132

Thanks,

Pratap

On 3/3/2026 7:28 PM, Kinn Coelho Juliao wrote:

[You don't often get email from kinncj@xxxxxxxxx. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.


Thanks Pratap, Mario.

I reviewed commit 0288a345f19b ("drm/amdgpu/isp: fix SMU warning
during ISP suspend-resume"). I understand it marks ISP MFD children as
syscore devices and manages their suspend/resume from the amdgpu parent
via pm_runtime_force_suspend/resume.

I'm curious about one thing though; on my current system (7.0-rc1),
s2idle resume still hangs with the in-tree V4L2 module loaded. My
out-of-tree module, which adds dev_pm_ops that call
isp4sd_pwroff_and_deinit() before suspend, resolves the hang.

pm_runtime_force_suspend() manages the power domain but doesn't tear
down firmware ring buffers or V4L2 streaming state. If the camera was
active at any point during the session when suspend occurs, the driver
would have stale state on resume. That seems to be what's causing the
hang on my hardware (HP ZBook Ultra G1a, Strix Halo).

Is the firmware/streaming state teardown handled elsewhere that I'm
missing, or does the V4L2 driver still need its own suspend path?

Thanks,
Kinn

--

--

Kinn Coelho Juliao

Ottawa- ON/Canada

Attachment: Screenshot_20260305_211812.png
Description: PNG image