[PATCH 1/2] Fix the imx8mm gpu hang due to adb400 reset wrongly
Jacky Bai
ping.bai at nxp.com
Mon Jan 19 02:04:35 PST 2026
> Subject: Re: [PATCH 1/2] Fix the imx8mm gpu hang due to adb400 reset
> wrongly
>
> On 19/01/2026 09:53, Jacky Bai wrote:
> > Due to the HW limitation on i.MX8MM, the gpumix, gpu2d and gpu3d share
> > the same reset domain, that means when gpu2d/3d go through the power
> > off/on cycle, the gpu2d/3d reset will also reset the gpumix domain,
> > The GPUMIX ADB400 port also be reset. But the ADB400 must be put into
> > power down before reset it.
> >
> > currently, gpumix, gpu2d/3d power domain use the pm runtime_pm to
> > handle these power domain dependency, but in some corner case, the
> > gpumix power off will be skipped, then the ADB400 port will be in
> > active while gpu2d/3d doing the reset. The GPUMIX the ADB400 port will
> > be reset wrongly, so lead to unpredictable bus behavior.
> >
> > To simplify these domain on/off order & dependency, refine the code to
> > directly handle GPUMIX domain on/off along with the gpu2d/3d power
> > on/off flow.
> >
> > Signed-off-by: Jacky Bai <ping.bai at nxp.com>
> > ---
>
> Please use subject prefixes matching the subsystem. You can get them for
> example with `git log --oneline -- DIRECTORY_OR_FILE` on the directory your
> patch is touching. For bindings, the preferred subjects are explained here:
Yes, the prefix should be added. Thank you to point out this problem. Missed when cleanup the patch, will fix it in V2.
BR
Jacky Bai
>
> Best regards,
> Krzysztof
More information about the linux-arm-kernel
mailing list