Hi,
On 10.11.2023 05:35, Jeffrey Hugo wrote:
> On 11/7/2023 5:35 AM, Jacek Lawrynowicz wrote:
>> From: Stanislaw Gruszka <stanislaw.gruszka@xxxxxxxxxxxxxxx>
>>
>> Change to use work for timeout detection. Needed for thread_irq
>> conversion.
>>
>> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@xxxxxxxxxxxxxxx>
>
> Missing SOB.
Yeah, I've missed it. I will add SOB to the whole patchset in v2.
>> void ivpu_pm_cancel_recovery(struct ivpu_device *vdev)
>> {
>> + drm_WARN_ON(&vdev->drm, delayed_work_pending(&vdev->pm->job_timeout_work));
>
> This seems odd. Looks like this function is only called from the dev_fini. It's a non-fatal error to tear down the device (hotplug? fatal error?) if a job is pending?
This check is used to detect driver bug where recovery is started just after ivpu_pm_cancel_recovery() was called (job_timeout_work schedules recovery_work).
Jobs should all be done/cancelled at this stage. It is more an assert()-like thing then an error that should be reported to the user.
Regards,
Jacek
