Re: [EXTERNAL] Re: [PATCH v10 1/1] misc: mrvl-cn10k-dpi: add Octeon CN10K DPI administrative driver

[Arnd Bergmann]

On Wed, Jul 10, 2024, at 15:27, Greg KH wrote:
> On Wed, Jul 10, 2024 at 01:19:56PM +0000, Vamsi Krishna Attunuru wrote:
>> >
>> >Why does ARM64 matter here?  I don't see any dependency required of it.
>> >
>> Thanks, Greg, for your time. This DPI device is an on-chip PCIe device and only present on
>> Marvell's CN10K platforms(which are 64-bit ARM SoC processors), so added those dependency.
>
> Then perhaps keep the ARM64 and add a COMPILE_TEST option as well so
> that we can build this as part of normal testing?
>
> So that would be:
> 	depends on PCI && (ARM64 || COMPILE_TEST)
> right?
>

That looks correct to me, though for specific ARM64 SoC we tend
to also have a more specific option, which would make this
something like

       depends on PCI
       depends on ARCH_THUNDER|| COMPILE_TEST

This is what we use e.g. for the PMU driver. I had at some
point submitted a patch to rename ARCH_THUNDER to ARCH_OCTEON,
since we seem to have some drivers that get this wrong, e.g.

config CRYPTO_DEV_OCTEONTX2_CPT
        tristate "Marvell OcteonTX2 CPT driver"
        depends on ARCH_THUNDER2 || COMPILE_TEST

which should also be ARCH_THUNDER instead of the discontinued
ARCH_THUNDER2.

For reference, we have five lineages of arm64 SoCs from Marvell:

- ARCH_THUNDER: Cavium ThunderX CN88xx, Marvell OcteonTX
     CN80xx/CN81xx/CN82xx/CN83xx, Marvell OcteonTX2
     CN92xx/CN93xx/CN96xx, Marvell Octeon Fusion, Marvell
     Octeon CN10xxx

- ARCH_THUNDER2: Broadcom Vulcan, Marvell ThunderX2 CN99xx

- ARCH_MVEBU: Marvell ThunderX2 CN91xx, Marvell Armada 7K,
     Marvell Armada 8K, Marvell Armada 37xx, Marvell
     Prestera 98DX25xx

- ARCH_MMP: PXA19x8

- ARCH_BERLIN: Armada 1500 Ultra, Synaptics BG4CT/BG5CT,
     Synaptics VisionSmart VS6xx, Synaptics Astra
     SL16xx, SenaryTech SN36xx.


      Arnd