Skip to contentSkip to conversation

linux-stable 6.1.53 kernel crash on COLOR_ID_MULTI handling change

DISCUSSION Da Xue 3 participants 3 messages Sep 13, 2023 – Sep 18, 2023

Subsystems (2)
lore
Share & tools
lore

Da Xue

author
#0 lore raw

Hi Pavel,

We have running systems that use COLOR_ID_MULTI. The GPIO toggles between two colors and we have used the identifier. RGB is not a good fit since it is not a RGB LED. Please provide guidance.

This patch causes the system to not start: f741121a2251 leds: Fix BUG_ON check for LED_COLOR_ID_MULTI that is always false

It was also backported to stable causing previously booting systems to no longer boot.

Best,
Da Xue

#1 lore raw

Hi!

Quoted 9 lines

We have running systems that use COLOR_ID_MULTI. The GPIO toggles between two colors and we have used the identifier. RGB is not a good fit since it is not a RGB LED. Please provide guidance.

This patch causes the system to not start: f741121a2251 leds: Fix BUG_ON check for LED_COLOR_ID_MULTI that is always false

It was also backported to stable causing previously booting systems to no longer boot.

Lets cc patch authors.

And please drop this from stable, it does not belong there.

Best regards,
Pavel

Signature · 1 line
People of Russia, stop Putin before his war on Ukraine escalates.
#2 lore raw
Quoted 12 lines · On Thu, 14 Sep 2023 12:21:09 +0200 Pavel Machek <[email protected]> wrote:

Hi!

Quoted 9 lines

We have running systems that use COLOR_ID_MULTI. The GPIO toggles between two colors and we have used the identifier. RGB is not a good fit since it is not a RGB LED. Please provide guidance.

This patch causes the system to not start: f741121a2251 leds: Fix BUG_ON check for LED_COLOR_ID_MULTI that is always false

It was also backported to stable causing previously booting systems to no longer boot.

Quoted 6 lines

Lets cc patch authors.

And please drop this from stable, it does not belong there.

Best regards,
Pavel

The BUG_ON was a no-op since it was introduced. It's purpose was to prevent people from using LED_COLOR_ID_MULTI since it was thought that non-RGB LEDs are not yet completely agreed on, or something.

But since the BUG_ON was a no-op, someone started using LED_COLOR_ID_MULTI without noticing that they should not. There are now even some in-tree device-tree files using LED_COLOR_ID_MULTI.

My patch that fixes the BUG_ON uncovered this and caused a regression for some people.

I think we have to drop the BUG_ON altoghether now.

I've sent a patch to linux-leds doing just that.

Sorry for the incovenience.

Marek