Re: [PATCH 4/7] drm/bridge: lock the encoder chain in scoped for_each loops

[Luca Ceresoli]

Hello,

On Tue, 30 Sep 2025 14:16:23 +0800
kernel test robot <oliver.sang@xxxxxxxxx> wrote:

> [  674.477504][    C1] watchdog: BUG: soft lockup - CPU#1 stuck for 626s! [swapper/0:1]
> [  674.477539][    C1] CPU#1 Utilization every 96s during lockup:
> [  674.477543][    C1] 	#1: 100% system,	  1% softirq,	  1% hardirq,	  0% idle
> [  674.477549][    C1] 	#2: 100% system,	  1% softirq,	  1% hardirq,	  0% idle
> [  674.477553][    C1] 	#3: 100% system,	  0% softirq,	  1% hardirq,	  0% idle
> [  674.477557][    C1] 	#4: 100% system,	  1% softirq,	  1% hardirq,	  0% idle
> [  674.477561][    C1] 	#5: 100% system,	  1% softirq,	  1% hardirq,	  0% idle
...
> [ 674.477713][ C1] drm_atomic_add_encoder_bridges (drivers/gpu/drm/drm_atomic.c:1319 (discriminator 1025)) 

The one reported is an actual bug, causing an infinite loop when the
encoder bridge chain is empty. Took a while to reproduce and debug, but
the fix is very simple:

 #define drm_for_each_bridge_in_chain_scoped(encoder, bridge)                           \
        for (struct drm_bridge *bridge __free(drm_bridge_encoder_chain_unlock) =        \
-                    list_first_entry(&drm_encoder_chain_lock(encoder)->bridge_chain,   \
+                    list_first_entry_or_null(&drm_encoder_chain_lock(encoder)->bridge_chain,   \

Fix queued for v2.

Luca

-- 
Luca Ceresoli, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com