Re: [PATCH v10 12/69] drm/bridge: Fix unlocked list_del in drm_bridge_add()

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

Pre-existing issues:
- [High] Legacy zero-initialized bridges not allocated via devm_drm_bridge_alloc() will cause a kernel panic due to an uninitialized list_head and trigger refcount warnings.
--

--- Patch [12]: [PATCH v10 12/69] drm/bridge: Fix unlocked list_del in drm_bridge_add() ---
commit e0f10da4c05d8a75c537ce94edb147bda1eb2e78
Author: Cristian Ciocaltea <[email protected]>

drm/bridge: Fix unlocked list_del in drm_bridge_add()

This commit addresses a list corruption issue in drm_bridge_add() by wrapping 
the bridge_lingering_list removal sequence with the bridge_lock mutex. It 
prevents concurrent modifications from drm_bridge_remove() and debugfs walks.

Closes: https://lore.kernel.org/all/[email protected]/

> diff --git a/drivers/gpu/drm/drm_bridge.c b/drivers/gpu/drm/drm_bridge.c
> index afaae272347c0..8b8f71c3ccff9 100644
> --- a/drivers/gpu/drm/drm_bridge.c
> +++ b/drivers/gpu/drm/drm_bridge.c

[ ... ]

> @@ -454,8 +455,10 @@ void drm_bridge_add(struct drm_bridge *bridge)
>  	 * in bridge_lingering_list. Remove it or bridge_lingering_list will be
>  	 * corrupted when adding this bridge to bridge_list below.
>  	 */
> -	if (!list_empty(&bridge->list))
> -		list_del_init(&bridge->list);
> +	scoped_guard(mutex, &bridge_lock) {
> +		if (!list_empty(&bridge->list))
> +			list_del_init(&bridge->list);
> +	}

[Severity: High]
This is a pre-existing issue, but will this cause a kernel panic for legacy 
bridge drivers that allocate struct drm_bridge using kzalloc() instead of 
devm_drm_bridge_alloc()?

For a zero-initialized bridge struct, the list_empty() check evaluates to 
false because the uninitialized list_head pointers are NULL, and NULL is not 
equal to &bridge->list. 

This means list_del_init() will execute and unconditionally dereference the 
NULL prev and next pointers, crashing the kernel during device probe.

Is it possible to check if the list_head is actually initialized before 
testing if it is empty, or handle legacy zero-initialized bridges safely 
to avoid breaking the probe path on affected platforms?

>  
>  	mutex_init(&bridge->hpd_state_mutex);
>  	mutex_init(&bridge->hpd_mutex);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=12