[PATCH v2] Revert v6.2-rc1 and later "ARM: dts: bcm2835-rpi: Use firmware clocks for display"

Florian Fainelli florian.fainelli at broadcom.com
Mon Jan 20 08:34:14 PST 2025



On 1/19/2025 7:04 AM, 'H. Nikolaus Schaller' via 
BCM-KERNEL-FEEDBACK-LIST,PDL wrote:
> Hi Stefan,
> 
>> Am 19.01.2025 um 01:36 schrieb Stefan Wahren <wahrenst at gmx.net>:
>>
>> Hi,
>>
>> Am 18.01.25 um 17:27 schrieb H. Nikolaus Schaller:
>>> This reverts commit 27ab05e1b7e5c5ec9b4f658e1b2464c0908298a6.
>>>
>>> I tried to upgrade a RasPi 3B+ with Waveshare 7inch HDMI LCD
>>> from 6.1.y to 6.6.y but found that the display is broken with
>>> this log message:
>>>
>>> [   17.776315] vc4-drm soc:gpu: bound 3f400000.hvs (ops vc4_drm_unregister [vc4])
>>> [   17.784034] platform 3f806000.vec: deferred probe pending
>>>
>>> Some tests revealed that while 6.1.y works, 6.2-rc1 is already broken but all
>>> newer kernels as well. And a bisect did lead me to this patch.
>> I successfully tested every Kernel release until Linux 6.13-rc with the
>> Raspberry Pi 3B+, so i prefer to step back and analyze this issue further.
> 
> Yes, I would be happy with any solution.
> 
>> What kernel config do you use ?
> 
> a private one which enables application specific drivers.
> 
>> What is the value of CONFIG_CLK_RASPBERRYPI ?
> 
> CONFIG_CLK_RASPBERRYPI is not set
> 
> I checked where this is defined and it is in bcm2835_defconfig and
> multi_v7_defconfig by
> 
> 4c6f5d4038af2c ("ARM: defconfig: enable cpufreq driver for RPi")
> 
> which hides this requirement quite well and got therefore unnoticed...
> 
> Setting CONFIG_CLK_RASPBERRYPI=y makes HDMI work without my proposed revert.
> Tested with v6.2.16, v6.6.72, v6.12.10 and v6.13-rc7.

I have been burned before by something similar and came up with this 
patch series that I should resubmit after addressing Conor's comment:

https://lore.kernel.org/all/20240513235234.1474619-1-florian.fainelli@broadcom.com/

Essentially, it removes the guess work, all you have to do is enable 
CONFIG_ARCH_BCM2835 and it just works, which is how it should be IMHO.
-- 
Florian




More information about the linux-arm-kernel mailing list