Re: Reboot crash at msm_atomic_commit_tail

On Tue, Jan 19, 2021 at 6:36 PM Fabio Estevam festevam@gmail.com wrote:

Hi Rob,

On Tue, Jan 19, 2021 at 1:40 PM Rob Clark robdclark@gmail.com wrote:

...

...

I suppose we should do the drm_atomic_helper_shutdown() conditionally?

This suggestion works, thanks. I will submit a patch shortly.

I think the cleanest way to do this is 2 patches: - add checks for DRIVER_MODESET to these helpers (there's a function to check these flags) - filter out DRIVER_MODESET flag in drm_device.driver_features (this is used to substract features at load time so that the drm_device struct can stay const)

There's probably other drivers that'd need the same treatment (but they don't use these helpers yet). -Daniel