Re: [PATCH 5/5] ARM: make ARCH_MULTIPLATFORM user-visible

[Arnd Bergmann]

On Wed, Sep 28, 2022, at 8:48 AM, Geert Uytterhoeven wrote:
> On Tue, Sep 27, 2022 at 10:19 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
>> On Tue, Sep 27, 2022, at 3:31 PM, Geert Uytterhoeven wrote:
>> > On Thu, Aug 18, 2022 at 4:56 PM Arnd Bergmann <arnd@xxxxxxxxxx> wrote:
>> >> --- a/arch/arm/Kconfig.debug
>> >> +++ b/arch/arm/Kconfig.debug
>> >> @@ -1904,6 +1904,7 @@ config DEBUG_UART_8250_PALMCHIP
>> >>
>> >>  config DEBUG_UNCOMPRESS
>> >>         bool "Enable decompressor debugging via DEBUG_LL output"
>> >> +       depends on !ARCH_MULTIPLATFORM
>> >
>> > Shouldn't DEBUG_LL itself depend on !ARCH_MULTIPLATFORM instead?
>>
>> That would also be possible, but I prefer to keep limiting only
>> the DEBUG_UNCOMPRESS. The idea of DEBUG_LL is that while it's
>> hardwired to a particular hardware address, it does not actually
>> access this address unless you specify the 'earlyprintk'
>> argument on the command line.
>
> ... or unless something goes really wrong, and the kernel tries to
> inform the user using early_print()?

I don't think this matters either: without DEBUG_LL, you get a non-booting
kernel and no diagnostics, while with DEBUG_LL, you might get some
diagnostic if you have configured the right debug address, and otherwise
the user gets the same as before: a crash without any output ;-)

> Note that the I/O region for the debug serial port is mapped regardless.
> Any chance this can cause conflicts?

Not sure. The early debug mapping should only be used in for
the earlyprintk output, but if a platform has a conflicting mapping
at the same address, it just never gets used before it gets replaced.

     Arnd