[PATCH v2 1/8] dm: fpga: Introduce new uclass

[Simon Glass]

Hi Michal,

On Thu, 22 Sept 2022 at 05:45, Michal Simek <michal.simek at amd.com> wrote:
>
>
>
> On 9/22/22 13:35, Simon Glass wrote:
> > Hi,
> >
> > On Thu, 22 Sept 2022 at 12:27, Michal Simek <michal.simek at amd.com> wrote:
> >>
> >>
> >>
> >> On 9/21/22 15:22, Alexander Dahl wrote:
> >>> For future DM based FPGA drivers and for now to have a meaningful
> >>> logging class for old FPGA drivers.
> >>>
> >>> Suggested-by: Michal Simek <michal.simek at amd.com>
> >>> Signed-off-by: Alexander Dahl <ada at thorsis.com>
> >>> ---
> >>>    include/dm/uclass-id.h | 1 +
> >>>    1 file changed, 1 insertion(+)
> >>>
> >>> diff --git a/include/dm/uclass-id.h b/include/dm/uclass-id.h
> >>> index a432e43871..c2b15881ba 100644
> >>> --- a/include/dm/uclass-id.h
> >>> +++ b/include/dm/uclass-id.h
> >>> @@ -56,6 +56,7 @@ enum uclass_id {
> >>>        UCLASS_ETH,             /* Ethernet device */
> >>>        UCLASS_ETH_PHY,         /* Ethernet PHY device */
> >>>        UCLASS_FIRMWARE,        /* Firmware */
> >>> +     UCLASS_FPGA,            /* FPGA device */
> >>>        UCLASS_FUZZING_ENGINE,  /* Fuzzing engine */
> >>>        UCLASS_FS_FIRMWARE_LOADER,              /* Generic loader */
> >>>        UCLASS_GPIO,            /* Bank of general-purpose I/O pins */
> >>
> >> Simon: the whole series look good to me. I am happy to take it via my tree when
> >> you ACK it. Also no problem if you want to take it via your tree.
> >> Please let me know which way you want to go.
> >
> > This is a good step forward but needs a lot more work.
> >
> > Please add a uclass file for the FPGA - i.e.
> > drivers/fpga/fpga-uclass.c - see other such files for examples.
> >
> > The FPGA uclass should have methods that match the non-DM interface.
> > You will likely need a DM_FPGA config to allow enabling the uclass.
> >
> > Also this needs a simple sandbox driver/emulator pair, so that it can
> > be tested, with tests in test/dm/fpga.c that use the driver.
> >
> > Admittedly this should have been done ages ago. I vaguely remember
> > mentioning it at the time, but perhaps I missed it. In any case, all
> > uclasses must have an API, implementation and tests that run in CI
> > with sandbox. Testing is a vital part of U-Boot and lack of testing is
> > the main reason why we went back to the 3-month release cycle.
>
> It can be done in steps for sure. Issues which Alex is addressing are there for
> quite some time and I think we shouldn't gate them by adding requirement to
> create the whole fpga uclass. It can be done on the top of this series.
> We know that it has to happen but I wouldn't push Alex to do it as condition for
> applying this series.
>  From my perspective if he has time to do, let's start with it. If not it can be
> done later.

Well if this is a start, then let's make it a real start. At minimum:

- add a uclass file with the uclass driver
- we can skip having any methods for now
- add a sandbox driver which does nothing
- add a test which probes the sandbox device

That is about 50 lines of code and people can then add to it over time.

Without that, I'd rather not have the UCLASS_FPGA.

Regards,
Simon