GSoC2020: RTEMS OpenFirmWare FDT Implementation (outline)

[Niteesh G. S.]

Ping.

Sorry for pinging so quickly but please understand that this is hampering
my progress.

Thanks,
Niteesh.

On Sun, Aug 9, 2020 at 12:52 AM Niteesh G. S. <niteesh.gs at gmail.com> wrote:

> Hello,
>
> I have created a basic outline of how the RTEMS openfirm API would
> look like. The implementation is inspired by the openfirm.h implementation
> in FreeBSD.
>
> Please a have look at the last 2 commits
>
> https://github.com/gs-niteesh/rtems/commits/ofw-rtems6-fdt-implementation-v1
>
> Since we are concerned only about the FDT implementation (also libBSD)
> I have removed all functions that are not supported by the FDT
> implementation
> for eg: OF_write, OF_call etc.
>
> I have a few questions regarding the API.
>
> 1) Should we 100% conform with the OFW API? As mentioned above the
> current idea will not implement functions like OF_call, OF_write etc.
>
> 2) For libBSD drivers to work we need to support the following functions
>
> device_t OF_device_from_xref(phandle_t xref);
> phandle_t OF_xref_from_device(device_t dev);
> int OF_device_register_xref(phandle_t xref, device_t dev);
>
> And implementing these functons in RTEMS is not possible.
>
> One way to handle this, mentioned in the past is to let the openfirm.c
> in libBSD handle the implementations and openfirm.h in RTEMS provide
> the interface. This caused us to forward declare the device_t structure
> in openfirm.h in RTEMS.
> Is this approach OK? Is there any other better approach?
>
> 3) There are a few functions unimplemented in libBSD should we implement
> them in the RTEMS API?
>
> Thanks,
> Niteesh.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20200811/b3593cd2/attachment.html>