Re: [PATCH] spi: cadence-xspi: Revert COMPILE_TEST support

[Rosen Penev]

On Sun Jun 14, 2026 at 9:30 PM PDT, Jisheng Zhang wrote:
> On Mon, Jun 08, 2026 at 11:09:27AM -0700, Nathan Chancellor wrote:
>> On Mon, Jun 08, 2026 at 11:31:33AM +0100, Mark Brown wrote:
>> > On Sat, Jun 06, 2026 at 03:26:04PM -0700, Nathan Chancellor wrote:
>> > > Commit 0c5b5c40dc31 ("spi: cadence-xspi: Add COMPILE_TEST support")
>> > > allows this driver to be built for 32-bit platforms, which causes a
>> > > semantic conflict with commit 4954d4eca469 ("spi: cadence-xspi: Support
>> > > 32bit and 64bit slave dma interface"), as readsq() and writesq() are
>> > > only available when targeting 64-bit platforms:
>> >
>> > >  config SPI_CADENCE_XSPI
>> > >  	tristate "Cadence XSPI controller"
>> > > -	depends on HAS_IOMEM || COMPILE_TEST
>> > > -	depends on OF
>> > > +	depends on OF && HAS_IOMEM && 64BIT
>> > >  	depends on SPI_MEM
>> >
>> > Why combine all these, they're not obviously related and HAS_IOMEM can
>> > be turned off.
>>
>> This is just a straight revert of 0c5b5c40dc31, so it is just restoring
>> how it was before that change.
>
> For the long run, it seems we need to add guard for the marvell support
> code, then I will solve the readsq/writesq on 64BIT. But it's too late
> to add new patches now, let me do this in next development window.
This sounds like a great solution.
>
> BTW: d58ecc54bb09 ("spi: cadence: Add 64BIT Kconfig dependency")
> introduced the 64BIT dependency.