Re: [PATCH v2 32/37] drm/omap: sdi: Switch to atomic bridge callbacks

[sashiko-bot@kernel.org]

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

Pre-existing issues:
- [High] `sdi_bridge_enable` and `sdi_bridge_disable` call sleeping functions in atomic context.
--

--- Patch [32]: [PATCH v2 32/37] drm/omap: sdi: Switch to atomic bridge callbacks ---
commit fe567fd05157eb0a22dd0fa9bede0ea86a947f4e
Author: Maxime Ripard <mripard@kernel.org>

drm/omap: sdi: Switch to atomic bridge callbacks

This commit updates the omap sdi bridge to use atomic bridge callbacks instead
of legacy non-atomic ones. It modifies the enable and disable bridge callbacks
to their atomic counterparts and adds atomic state handlers.

> diff --git a/drivers/gpu/drm/omapdrm/dss/sdi.c b/drivers/gpu/drm/omapdrm/dss/sdi.c
> --- a/drivers/gpu/drm/omapdrm/dss/sdi.c
> +++ b/drivers/gpu/drm/omapdrm/dss/sdi.c
[ ... ]
> -static void sdi_bridge_enable(struct drm_bridge *bridge)
> +static void sdi_bridge_enable(struct drm_bridge *bridge,
> +			      struct drm_atomic_commit *commit)
>  {
>  	struct sdi_device *sdi = drm_bridge_to_sdi(bridge);
>  	struct dispc_clock_info dispc_cinfo;

[Severity: High]
This isn't a bug introduced by this patch, but could this lead to system
deadlocks by calling sleeping functions from an atomic context?

By wiring this to the atomic_enable callback, it is now executed in the
atomic commit tail path where sleeping is forbidden. Looking at the contents
of sdi_bridge_enable(), it calls:

	r = regulator_enable(sdi->vdds_sdi_reg);
	if (r)
		return;

	r = dispc_runtime_get(sdi->dss->dispc);

Both regulator_enable() (which acquires a mutex) and dispc_runtime_get()
(which calls pm_runtime_get_sync()) can sleep.

[ ... ]
> -static void sdi_bridge_disable(struct drm_bridge *bridge)
> +static void sdi_bridge_disable(struct drm_bridge *bridge,
> +			       struct drm_atomic_commit *commit)
>  {
>  	struct sdi_device *sdi = drm_bridge_to_sdi(bridge);

[Severity: High]
This is a pre-existing issue, but does switching this function to be an
atomic callback introduce sleeping in atomic context warnings here as well?

Similar to the enable path above, sdi_bridge_disable() calls:

	dispc_runtime_put(sdi->dss->dispc);

	regulator_disable(sdi->vdds_sdi_reg);

Could these sleeping functions trigger kernel warnings when called during
display atomic modeset operations?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260707-drm-all-atomic-bridges-v2-0-21d03cbca446@kernel.org?part=32