Re: [RFC PATCH] powerpc: udbg: export udbg_putc

[Randy Dunlap]

Hi Pali,

On 5/13/23 02:13, Pali Rohár wrote:
> On Friday 12 May 2023 22:58:04 Randy Dunlap wrote:
>> In a randconfig with CONFIG_SERIAL_CPM=m and
>> CONFIG_PPC_EARLY_DEBUG_CPM=y, there is a build error:
>> ERROR: modpost: "udbg_putc" [drivers/tty/serial/cpm_uart/cpm_uart.ko] undefined!
>>
>> The build can be fixed by exporting "udbg_putc" in udbg.c.
>>
>> OTOH, maybe something like this is more appropriate in
>> arch/powerpc/Kconfig.debug?
>>
>> config PPC_EARLY_DEBUG_CPM
>> 	bool "Early serial debugging for Freescale CPM-based serial ports"
>> -	depends on SERIAL_CPM
>> +	depends on SERIAL_CPM=y
>>
>> Signed-off-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
>> Cc: Michael Ellerman <mpe@xxxxxxxxxxxxxx>
>> Cc: Nicholas Piggin <npiggin@xxxxxxxxx>
>> Cc: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
>> Cc: linuxppc-dev@xxxxxxxxxxxxxxxx
>> Cc: Arnd Bergmann <arnd@xxxxxxxx>
>> Cc: "Pali Rohár" <pali@xxxxxxxxxx>
>> ---
>>  arch/powerpc/kernel/udbg.c |    1 +
>>  1 file changed, 1 insertion(+)
>>
>> diff -- a/arch/powerpc/kernel/udbg.c b/arch/powerpc/kernel/udbg.c
>> --- a/arch/powerpc/kernel/udbg.c
>> +++ b/arch/powerpc/kernel/udbg.c
>> @@ -14,6 +14,7 @@
>>  #include <asm/udbg.h>
>>  
>>  void (*udbg_putc)(char c);
>> +EXPORT_SYMBOL(udbg_putc);
>>  void (*udbg_flush)(void);
>>  int (*udbg_getc)(void);
>>  int (*udbg_getc_poll)(void);
> 
> Hello! I do not think that it is a good idea to export udbg functions
> for kernel modules. I have quickly looked at the cpm_uart driver file

Sure, that was just a conversation starter.

> drivers/tty/serial/cpm_uart/cpm_uart_core.c and it looks like that once
> udbg from it is registered then there is no code for unregistering it.
> So I have feeling that compiling cpm_uart driver as module should not be
> allowed when CONFIG_SERIAL_CPM_CONSOLE is enabled, and early debug
> should depend on CONFIG_CONSOLE_POLL || CONFIG_SERIAL_CPM_CONSOLE.

The other change above also fixes the build error: (Option 2)

>> config PPC_EARLY_DEBUG_CPM
>> 	bool "Early serial debugging for Freescale CPM-based serial ports"
>> -	depends on SERIAL_CPM
>> +	depends on SERIAL_CPM=y


Also, making SERIAL_CPM (cpm_uart driver) be restricted to not allow it
as a module when SERIAL_CPM_CONSOLE is enabled [how does one express that
in Kconfig language?] will cause a circular dependency since SERIAL_CPM_CONSOLE
depends on SERIAL_CPM=y. It looks like you are suggesting reversing the
dependencies.

> Any other opinion?

Yes, please. Otherwise I prefer my option 2.

-- 
~Randy