[PATCH] Revert "usb: kbd: destroy device after console is stopped"

[Marek Vasut]

On 4/22/21 1:30 PM, Andy Shevchenko wrote:
> On Thu, Apr 22, 2021 at 12:11:44PM +0100, Peter Robinson wrote:
>> On Thu, Apr 22, 2021 at 11:52 AM Andy Shevchenko
>> <andriy.shevchenko at linux.intel.com> wrote:
>>>
>>> On Thu, Apr 22, 2021 at 10:19:10AM +0100, Peter Robinson wrote:
>>>
>>> Thanks for the report.
>>>
>>>> Reverts commit eb5fd9e46c1, it causes ARMv7 devices to stop booting
>>>> Linux when a USB keyboard is attached. The kernels starts but there's
>>>> no output. Reverting it makes things work again.
>>>
>>> It's not good. When we revert, we introduce another (I consider more serious)
>>> bug.
>>
>> Sure, looking at the fixes tags I figured as much, but it also causes
>> regressions itself. I figured sending a revert would at least start
>> the discussion off.
>>
>>>> Signed-off-by: Peter Robinson <pbrobinson at gmail.com>
>>>> Cc: Nicolas Saenz Julienne <nsaenzjulienne at suse.de>
>>>> Cc: Andy Shevchenko <andriy.shevchenko at linux.intel.com>
>>>> ---
>>>>
>>>> This has caused us issues on a number of ARMv7 deviices. I'm not sure
>>>> why specifically ARMv7 because it happens on a RPi3 running Fedora
>>>> armhfp, but doesn't happen on aarch64. Issue seen on RPis, Cubietruck,
>>>> and others.
>>>
>>> Can we conduct a bit of investigation here?
>>> Is it using ATAGs or command line? How Linux kernel gets its parameters?
>>
>> I believe command line, I've been testing booting using UEFI. TBH I'm
>> unsure why I see the problem on ARMv7 and not using aarchc64 on the
>> same device.
> 
> And what is the U-Boot environment in both cases? (Can you share it somewhere?)

Could it be related to CONFIG_SYS_DEVICE_NULLDEV / CONFIG_NULLDEV_SERIAL 
not being set in one of the cases ?