Re: [PATCH v3 06/15] drm/tidss: Remove extra pm_runtime_mark_last_busy

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] This problem wasn't introduced by this patch, but the WARN_ON(r < 0) following pm_runtime_put_autosuspend() incorrectly treats valid Runtime PM API returns as driver bugs. According to the Runtime PM subsystem API contracts, pm_runtime_put_autosuspend() can legitimately return negative error codes during normal operation. For example, if another thread concurrently increments the usage counter between the usage drop and the idle check, the internal PM checks will return -EAGAIN. If the driver has active children (like the newly added OLDI auxiliary devices) and ignore_children is not set, it can return -EBUSY. Furthermore, if CONFIG_PM is disabled, it unconditionally returns -ENOSYS. Asserting WARN_ON(r < 0) on these valid state transitions leads to spurious warnings in the kernel log.
--

commit d589ba6abd338a264b1bc27a4829172f13fddbbc
Author: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>

drm/tidss: Remove extra pm_runtime_mark_last_busy

This commit removes the explicit call to pm_runtime_mark_last_busy() in
tidss_runtime_put(), as pm_runtime_put_autosuspend() already handles this
internally.

> diff --git a/drivers/gpu/drm/tidss/tidss_drv.c b/drivers/gpu/drm/tidss/tidss_drv.c
> index 1c8cc18bc53c3..5cb3e746aeb37 100644
> --- a/drivers/gpu/drm/tidss/tidss_drv.c
> +++ b/drivers/gpu/drm/tidss/tidss_drv.c
> @@ -42,8 +42,6 @@ void tidss_runtime_put(struct tidss_device *tidss)
>  {
>  	int r;
>  
> -	pm_runtime_mark_last_busy(tidss->dev);
> -
>  	r = pm_runtime_put_autosuspend(tidss->dev);
>  	WARN_ON(r < 0);

[Severity: High]
This isn't a bug introduced by this patch, but does this WARN_ON() incorrectly
treat valid Runtime PM API returns as driver bugs?

According to the Runtime PM subsystem API contracts, 
pm_runtime_put_autosuspend() can legitimately return negative error codes during
normal operation. For example, if another thread concurrently increments the
usage counter between the usage drop and the idle check, the internal PM checks
will return -EAGAIN. 

If the driver has active children and ignore_children is not set, it can return
-EBUSY. Furthermore, if CONFIG_PM is disabled, it unconditionally returns
-ENOSYS.

Asserting WARN_ON(r < 0) on these valid state transitions could lead to spurious
warnings in the kernel log.

>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260529-beagley-ai-display-v3-0-7fefdc5d1adf@ideasonboard.com?part=6