[u-boot][PATCH 1/3] scripts: Makefile.spl: Enable memory drivers to be built for SPL

[Simon Glass]

Hi Roger,

On Sat, 1 Oct 2022 at 02:37, Roger Quadros <rogerq at kernel.org> wrote:
>
> Hi,
>
> On 01/10/2022 02:48, Simon Glass wrote:
> > Hi,
> >
> > On Fri, 30 Sept 2022 at 14:18, Roger Quadros <rogerq at kernel.org> wrote:
> >>
> >> On 30/09/2022 17:00, Tom Rini wrote:
> >>> On Fri, Sep 30, 2022 at 07:28:51AM -0600, Simon Glass wrote:
> >>>> Hi Roger,
> >>>>
> >>>> On Fri, 30 Sept 2022 at 06:47, Roger Quadros <rogerq at kernel.org> wrote:
> >>>>>
> >>>>> Simon,
> >>>>>
> >>>>> On 29/09/2022 21:06, Simon Glass wrote:
> >>>>>> Hi Roger,
> >>>>>>
> >>>>>> On Thu, 29 Sept 2022 at 01:03, Roger Quadros <rogerq at kernel.org> wrote:
> >>>>>>>
> >>>>>>> Hi Simon,
> >>>>>>>
> >>>>>>> On 28/09/2022 19:27, Simon Glass wrote:
> >>>>>>>> Hi Roger,
> >>>>>>>>
> >>>>>>>> On Wed, 28 Sept 2022 at 06:12, Roger Quadros <rogerq at kernel.org> wrote:
> >>>>>>>>>
> >>>>>>>>> We will need ti-gpmc driver for SPL. Allow memory drivers
> >>>>>>>>> do be built for SPL.
> >>>>>>>>>
> >>>>>>>>> Signed-off-by: Roger Quadros <rogerq at kernel.org>
> >>>>>>>>> ---
> >>>>>>>>>  scripts/Makefile.spl | 1 +
> >>>>>>>>>  1 file changed, 1 insertion(+)
> >>>>>>>>
> >>>>>>>> Please can you use the existing drivers/ram directory?
> >>>>>>>
> >>>>>>> The ti-gpmc driver is not actually a RAM only controller. Although it can support SRAM.
> >>>>>>> It is a more general purpose controller that can support different peripherals.
> >>>>>>> It is similar to the drivers already existing in the divers/memory directory.
> >>>>>>>
> >>>>>>> I was just trying to keep the file layout similar to that in the Linux kernel.
> >>>>>>>
> >>>>>>> Do you still see a problem with it?
> >>>>>>
> >>>>>> Well in this case perhaps the RAM device would be a child of this one?
> >>>>>
> >>>>> That's right.
> >>>>>>
> >>>>>> But there is no uclass for your new device. One of the drivers in that
> >>>>>> dir uses UCLASS_NOP and your one seems to use UCLASS_SIMPLE_BUS
> >>>>>>
> >>>>>> So let's add a uclass for it and describe exactly what it is for.
> >>>>>
> >>>>> Why isn't UCLASS_SIMPLE_BUS sufficient?
> >>>>> By itself, the GPMC driver doesn't offer any usable functionality.
> >>>>> It just configures the bus interface and then populates the children.
> >>>>
> >>>> That's OK, but in that case it should go in drivers/bus and perhaps
> >>>> drivers/memory should go away?
> >>>
> >>> No, drivers/memory/stm32-fmc2-ebi.c is there and is the equivalent (more
> >>> or less) of drivers/memory/stm32-fmc2-ebi.c in the linux kernel. So
> >>> maybe a question is, are we talking about the equivalent of
> >>> drivers/memory/omap-gpmc.c here? Or something else?
> >>>
> >>
> >> Yes, I just picked up that file and adapted it for u-boot.
> >> As GPMC is not only for OMAP devices anymore I renamed it to ti-gpmc.c.
> >> If it helps we can retain the kernel naming.
> >
> > In the fullness of time all driver/xxx directories should have a
> > uclass associated with them. So in this case, do we want to use
> > UCLASS_BUS and put it in drivers/bus or add a UCLASS_MEMORY and put it
> > in drivers/memory?
>
> I'd opt for UCLASS_MEMORY.

OK, here is an example patch for adding a new uclass:

https://patchwork.ozlabs.org/project/uboot/patch/20220930120430.42307-2-post@lespocky.de/

It is the bare minimum as it has not methods, but it seems like yours
won't either.

Regards,
Simon