[soc:soc/drivers 20/24] ERROR: modpost: "__ffsdi2" [drivers/platform/cznic/turris-omnia-mcu.ko] undefined!

[Arnd Bergmann]

On Fri, Mar 21, 2025, at 10:36, Geert Uytterhoeven wrote:
>> >
>> > ERROR: modpost: "__delay" [drivers/net/mdio/mdio-cavium.ko] undefined!
>
> Bug in driver (must not use __delay() directly).
>
>> > ERROR: modpost: "devm_of_clk_add_hw_provider"
>> > [drivers/media/i2c/tc358746.ko] undefined!
>> > ERROR: modpost: "devm_clk_hw_register" [drivers/media/i2c/tc358746.ko]
>> > undefined!
>> > ERROR: modpost: "of_clk_hw_simple_get" [drivers/media/i2c/tc358746.ko]
>> > undefined!
>
> Missing dependencies for driver.

Agreed, but both of these only break the build on sh because
the architecture code is different from the others. It's obviously
best to fix these in the drivers, I just wouldn't expect the
driver maintainers to care unless someone sends them a patch/

>> >>> ERROR: modpost: "__ffsdi2" [drivers/platform/cznic/turris-omnia-mcu.ko] undefined!
>>
>> This comes from __bf_shf(spec) in omnia_mcu_request_irq().
>>
>> As far as I can tell, this is not your problem but in the
>> SH architecture missing one of the libgcc functions that are
>> called by gcc.
>>
>> Since this is only for compile-testing sh allmodconfig, and
>> that is already broken because of other bugs, I'm not going
>> to care either.
>>
>> Adding the linux-sh list to Cc, it would be nice to get a clean
>> build again. All four problems should be trivial to address,
>> and some of these have been broken for many years.
>
> So one SH-specific issue to fix...

There is also the missing clone3 syscall that has been
causing a warning for many years.

      Arnd