Re: [PATCH v16 6/6] soc: amd: Add support for AMD Pensando SoC Controller

[Brad Larson <blarson@amd.com>]

Hi Arnd,

Thanks for the reply and guidance!

On Fri, Sep 22, 2023, at 06:24:00 -0400, Arnd Bergmann wrote:
> On Wed, Sep 13, 2023, at 17:49, Brad Larson wrote:
>> The Pensando SoC controller is a SPI connected companion device
>> that is present in all Pensando SoC board designs.  The essential
>> board management registers are accessed on chip select 0 with
>> board mgmt IO support accessed using additional chip selects.
>>
>> Signed-off-by: Brad Larson <blarson@amd.com>
>> ---
>>
>> v15 changes:
>> - Drop custom ioctl and use existing miscdevice interface.
>> - Delete unused definitions in amd-pensando-ctrl.h 
>> - Makefile change to compile for ARCH_PENSANDO
>
> Hi Brad,
> 
> I'm sorry I've been out of the loop for so long, and I hope
> we can find a way to manage your SoC support soon. My impression
> is that the normal support patches (1, 3, 4, and 5) are largely
> uncontroversial, while the SoC controller support seems like
> we are still not converging onto something that is ready to
> merge, so I would suggest you split the two parts and send
> the basic support for inclusion in linux-6.7 while we continue
> to discuss the soc controller driver.

I've sent a patchset with only patches 1, 3, 4, and 5.  Correctness
is the priority for upstream inclusion.

> Please remove any references to the soc controller from the
> dts files and send that first series to:soc@kernel.org
> cc:linux-arm-kernel (and the other interested parties) so
> I can pick those up.

Yes, I've removed the node for which the compatible driver is the problem.

> 
> Regarding the soc controller driver, let me try to give
> you my impression of where we are:
> 
> - you have gone through 16 revisions already, which is way
>   too much for a public review, we should have been able
>   to find a solution earlier than that, and this is partly
>   our fault on the reviewer side, and I'm sorry about that.
> 
> - Andy's latest comments and a lot of the earlier reviews
>   were focused on implementation details. While those comments
>   are helpful suggestions for improving the code, they miss
>   the larger point about the system design that I'm worried
>   about and probably don't help you actually get it merged.
> 
> - The main problem I still see is that this driver completely
>   bypasses our normal kernel abstractions and instead creates
>   a low-level passthrough interface for handling kernel
>   functionality in userspace. This creates a liability both
>   for the user ABI and the kernel implementation and prevents
>   any 
> 
> - There is a chance that your design is in fact the
>   best way to handle this particular hardware, but it is
>   your job to write a convincing explanation of why this
>   platform is different from all the others in the patch
>   description. Your current one-paragraph text does not
>   explain this at all.
>  
> I would suggest you prioritize getting the other patches
> included for the moment, but we can keep discussion the
> API design for this driver either in this thread or on the
> #armlinux IRC channel (irc.libera.chat) in parallel if you
> like. In order to help you here, I would need either
> the documentation of the SPI software interface, or the
> source code for the userspace tool.
> 
>      Arnd 

I'll redirect and reframe what the SoC driver is doing to
#armlinux IRC to find an appropriate solution.

Regards,
Brad