Hi,

On Sun, Jan 30, 2022 at 03:57:07PM +0100, Michael Stapelberg wrote:

Hello again,

I am facing the exact same symptoms as last September: https://lists.freedesktop.org/archives/dri-devel/2021-September/324008.html

…meaning my (headless) Raspberry Pi 3 B and Raspberry Pi 3 B+ no longer boot with Linux 5.16.3, while the (headless) Raspberry Pi 4 still boots.

I'm not entirely sure what you expect by insisting on things getting broken *again*, but we got it, thanks.

This issue is not fixed with the latest Linux 5.16.4 either.

If needed, I can git-bisect it, but perhaps you have a hunch already?

With 5.16.2 (working), I see the following serial console output: [ 2.493456] fb0: switching to vc4 from simple [ 2.500306] Console: switching to colour dummy device 80x25 [ 2.507204] [drm] Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0 [ 2.517617] vc4-drm soc:gpu: [drm] Cannot find any crtc or sizes [ 12.517962] vc4-drm soc:gpu: [drm] Cannot find any crtc or sizes

With 5.16.3 (broken), I see: [ 2.485311] fb0: switching to vc4 from simple [ 2.492167] Console: switching to colour dummy device 80x25 [ 2.499148] [drm] Initialized vc4 0.0.0 20140616 for soc:gpu on minor 0

<boot hangs>

I bisected it, and it looks like it's because of a bad conflict resolution in stable.

The original commit is here: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

And the stable commit is here: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=...

For some reason, in the stable commit, pm_runtime_put ended up in vc4_hdmi_cec_init while it should have been at the end of vc4_hdmi_cec_adap_enable.

I'll send a patch to fix it.

Maxime