Re: [PATCH v2 1/4] firmware: arm_scmi: fix deps arm-smccc-discovery deps in Kconfig

[Cristian Marussi]

Hi Etienne,

On Fri, May 21, 2021 at 03:40:51PM +0200, Etienne Carriere wrote:
> ARM_SCMI_PROTOCOL depends one either MAILBOX or HAVE_ARM_SMCCC_DISCOVERY,
> not MAILBOX only. This change fixes this in Kconfig file and driver
> private header file.
> 
> Signed-off-by: Etienne Carriere <etienne.carriere@xxxxxxxxxx>
> ---
> No change since v1
> ---
>  drivers/firmware/Kconfig           | 2 +-
>  drivers/firmware/arm_scmi/common.h | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/firmware/Kconfig b/drivers/firmware/Kconfig
> index db0ea2d2d75a..a9c613c32282 100644
> --- a/drivers/firmware/Kconfig
> +++ b/drivers/firmware/Kconfig
> @@ -9,7 +9,7 @@ menu "Firmware Drivers"
>  config ARM_SCMI_PROTOCOL
>  	tristate "ARM System Control and Management Interface (SCMI) Message Protocol"
>  	depends on ARM || ARM64 || COMPILE_TEST
> -	depends on MAILBOX
> +	depends on MAILBOX || HAVE_ARM_SMCCC_DISCOVERY
>  	help
>  	  ARM System Control and Management Interface (SCMI) protocol is a
>  	  set of operating system-independent software interfaces that are
> diff --git a/drivers/firmware/arm_scmi/common.h b/drivers/firmware/arm_scmi/common.h
> index 228bf4a71d23..8685619d38f9 100644
> --- a/drivers/firmware/arm_scmi/common.h
> +++ b/drivers/firmware/arm_scmi/common.h
> @@ -331,7 +331,7 @@ struct scmi_desc {
>  };
>  
>  extern const struct scmi_desc scmi_mailbox_desc;
> -#ifdef CONFIG_HAVE_ARM_SMCCC
> +#ifdef CONFIG_HAVE_ARM_SMCCC_DISCOVERY
>  extern const struct scmi_desc scmi_smc_desc;
>  #endif
>  

This seems reasonable to me as of now, but I have to say in the original 
virtio-scmi series from Peter this issue is tackled a bit differently:

(https://lore.kernel.org/linux-arm-kernel/20210511002040.802226-1-peter.hilber@xxxxxxxxxxxxxxx/)

adding something like ARM_SCMI_HAVE_SHMEM and ARM_SCMI_HAVE_MSG and then
make mailbox and smc transport select HAVE_SHMEM while virtio selects
HAVE_MSG and keeping ARM_SCMI_PROTOCOL depends on either of those.

I found such approach a bit better in term of abstraction indeed, but
given that such virtio-scmi series I'm reworking is still probably a bit
further down the line in terms of posting, I think that your solution is
good enough for now if Sudeep agrees with it, and we can switch to the
above mentioned HAVE_* abstractions once virtio-scmi is pushed forward.

Reviewed-by: Cristian Marussi <cristian.marussi@xxxxxxx>

Thanks,
Cristian

> -- 
> 2.17.1
>