Forcing DEBUG_UART_{PHYS, VIRT} changes when switching between v7 platforms?

[Florian Fainelli]

2014-03-20 0:07 GMT-07:00 Arnd Bergmann <arnd at arndb.de>:
> On Wednesday 19 March 2014 16:38:45 Florian Fainelli wrote:
>> Hi,
>>
>> When switching between different Multi v7 platform which all select
>> DEBUG_UART_8250, whichever platform managed to set
>> DEBUG_UART_{PHYS,VIRT} first will end-up forcing its value to the
>> others.
>>
>> What I am seeing at the moment is for instance enabling BCM_MOBILE
>> will set DEBUG_UART_PHYS to 0x3e000000, disabling BCM_MOBILE and now
>> enabling MVEBU won't force a new DEBUG_UART_PHYS address to be
>> re-computed.
>>
>> Should we add some sort of specific annotation to the Kconfig symbol
>> to force a recalculation of the UART PHYS address?
>
> Russell has plans to change the Konfig logic in the future, removing the
> platform specific options. With the current code, you could probably
> do something along the lines of
>
> diff --git a/arch/arm/Kconfig.debug b/arch/arm/Kconfig.debug
> index f72147f..aa20929 100644
> --- a/arch/arm/Kconfig.debug
> +++ b/arch/arm/Kconfig.debug
> @@ -1013,7 +1013,7 @@ config DEBUG_UART_8250
>                 ARCH_LPC32XX || ARCH_MV78XX0 || ARCH_ORION5X || ARCH_RPC
>
>  config DEBUG_UART_PHYS
> -       hex "Physical base address of debug UART"
> +       hex
>         default 0x01c20000 if DEBUG_DAVINCI_DMx_UART0
>         default 0x01c28000 if DEBUG_SUNXI_UART0
>         default 0x01c28400 if DEBUG_SUNXI_UART1
> @@ -1074,7 +1074,7 @@ config DEBUG_UART_PHYS
>                 DEBUG_UART_8250 || DEBUG_UART_PL01X
>
>  config DEBUG_UART_VIRT
> -       hex "Virtual base address of debug UART"
> +       hex
>         default 0xe0010fe0 if ARCH_RPC
>         default 0xf0000be0 if ARCH_EBSA110
>         default 0xf0009000 if DEBUG_CNS3XXX
>
> with some extra logic to still allow you to enter the values
> manually if not set by default, but that probably wouldn't work
> any more when the rework is done.

Your change works great, thanks! Should we make that the current
behavior until Russell revamps the Kconfig logic?
-- 
Florian