[PATCH 2/2] drm: Schedule the HPD poll work on the system unbound workqueue

[Gabriele Monaco]

On Fri, 2023-09-01 at 17:04 +0300, Imre Deak wrote:
> On some i915 platforms at least the HPD poll work involves I2C
> bit-banging using udelay()s to probe for monitor EDIDs. This in turn
> may trigger the
> 
>  workqueue: output_poll_execute [drm_kms_helper] hogged CPU for
> >10000us 4 times, consider switching to WQ_UNBOUND
> 
> warning. Fix this by scheduling drm_mode_config::output_poll_work on
> a
> WQ_UNBOUND workqueue.
> 
> Cc: Tejun Heo <tj at kernel.org>
> Cc: Heiner Kallweit <hkallweit1 at gmail.com>
> CC: stable at vger.kernel.org # 6.5
> Cc: dri-devel at lists.freedesktop.org
> Suggested-by: Tejun Heo <tj at kernel.org>
> Suggested-by: Heiner Kallweit <hkallweit1 at gmail.com>
> Reported-by: Heiner Kallweit <hkallweit1 at gmail.com>
> Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/9245
> Link:
> https://lore.kernel.org/all/f7e21caa-e98d-e5b5-932a-fe12d27fde9b@gmail.com
> Signed-off-by: Imre Deak <imre.deak at intel.com>
> ---
>  drivers/gpu/drm/drm_probe_helper.c | 8 +++++---
>  1 file changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_probe_helper.c
> b/drivers/gpu/drm/drm_probe_helper.c
> index 3f479483d7d80..72eac0cd25e74 100644
> --- a/drivers/gpu/drm/drm_probe_helper.c
> +++ b/drivers/gpu/drm/drm_probe_helper.c
> @@ -279,7 +279,8 @@ static void reschedule_output_poll_work(struct
> drm_device *dev)
>  		 */
>  		delay = HZ;
>  
> -	schedule_delayed_work(&dev->mode_config.output_poll_work,
> delay);
> +	queue_delayed_work(system_unbound_wq,
> +			   &dev->mode_config.output_poll_work,
> delay);
>  }
>  
>  /**
> @@ -614,7 +615,7 @@ int
> drm_helper_probe_single_connector_modes(struct drm_connector
> *connector,
>  		 */
>  		dev->mode_config.delayed_event = true;
>  		if (dev->mode_config.poll_enabled)
> -			mod_delayed_work(system_wq,
> +			mod_delayed_work(system_unbound_wq,
>  					 &dev-
> >mode_config.output_poll_work,
>  					 0);
>  	}
> @@ -838,7 +839,8 @@ static void output_poll_execute(struct
> work_struct *work)
>  		drm_kms_helper_hotplug_event(dev);
>  
>  	if (repoll)
> -		schedule_delayed_work(delayed_work,
> DRM_OUTPUT_POLL_PERIOD);
> +		queue_delayed_work(system_unbound_wq,
> +				   delayed_work,
> DRM_OUTPUT_POLL_PERIOD);
>  }
>  
>  /**

Hello all,

why was this patch never included?
Especially this 2/2 seems to solve a latency issue we are observing:
the work can last milliseconds and run on isolated cores, affecting
latency requirements in some real time scenarios (e.g. oslat).

Thanks,
Gabriele