[U-Boot] [PATCH 12/14] fdt: eth_fixup: Add hook for board to override MAC

[Olliver Schinagl]

Hey Simon,

On 01-12-16 03:20, Simon Glass wrote:
> Hi,
>
> On 25 November 2016 at 08:30, Olliver Schinagl <oliver at schinagl.nl> wrote:
>> This patch adds a method for the board to set the MAC address if the
>> environment is not yet set. The environment based MAC addresses are not
>> touched, but if the fdt has an alias set, it is parsed and put into the
>> environment.
>>
>> E.g. The environment contains ethaddr and eth1addr, and the fdt contains
>> an ethernet1 nothing happens. If the fdt contains ethernet2 however, it
>> is parsed and inserted into the environment as eth2addr.
>>
>> Signed-off-by: Olliver Schinagl <oliver at schinagl.nl>
>> ---
>>  common/fdt_support.c | 8 +++++++-
>>  1 file changed, 7 insertions(+), 1 deletion(-)
>>
>> diff --git a/common/fdt_support.c b/common/fdt_support.c
>> index c34a13c..f127392 100644
>> --- a/common/fdt_support.c
>> +++ b/common/fdt_support.c
>> @@ -465,6 +465,11 @@ int fdt_fixup_memory(void *blob, u64 start, u64 size)
>>         return fdt_fixup_memory_banks(blob, &start, &size, 1);
>>  }
>>
>> +__weak int board_get_enetaddr(const int i, unsigned char *mac_addr)
>> +{
>> +       return -ENOSYS;
>> +}
>> +
>>  void fdt_fixup_ethernet(void *fdt)
>>  {
>>         int i, prop;
>> @@ -507,7 +512,8 @@ void fdt_fixup_ethernet(void *fdt)
>>                         if (fdt_eth_addr)
>>                                 eth_parse_enetaddr(fdt_eth_addr, mac_addr);
>>                         else
>> -                               continue;
>> +                               if (board_get_enetaddr(i, mac_addr) < 0)
>> +                                       continue;
>>
>>                         do_fixup_by_path(fdt, path, "mac-address",
>>                                          &mac_addr, 6, 0);
>> --
>> 2.10.2
>>
>
> Much as I don't want to pin this on any one patch, but I feel that our
> DT fixup stuff is a bit out of control. We have so many functions that
> do this, and they are called from various places. At some point we
> need to think about the infrastructure.
>
> IMO we should move to a linker list approach a bit like SPL did
> recently (SPL_LOAD_IMAGE_METHOD). Then boards and subsystems can
> register (at build-time) a fixup routine and they all get called one
> after the other.
>
> We could also (later) add run-time support for registering fixups,
> that drivers could use.

We kinda left this out in the cold last time (or I did?) and I just sent 
my v5 of this patch-series (without this one).

The gist from what I remember here was, to not introduce a weak function 
here, and have boards go back-forth to other subsystems. I belive one 
suggestion was to use a more specific name, lets say sunxi_get_hwaddr() 
for example and use that. Quite obviously this being the 
common/fdt_support, it would have to be generic.

So I'm still at loss as to how to handle this here.

Some background here, what the idea was I belive, that certain network 
adapters rely on the fdt to supply the MAC address to the driver, and so 
u-boot injects the mac address into the fdt.

We currently hack this together I belive, by generating environment 
variables for non-existant devices (if the fdt has ethernet aliases) and 
when doing this bit of code, inject the generated environment variables 
into the fdt.

I can't rewrite/fixup the fdt stuff, but am also not sure what an 
acceptable solution for this would be. Granted we really are hacking 
this into u-boot but I suppose we cannot escape this for legacy reasons. 
Devices expect the same generated mac address as before.

Again, bar rewriting this stuff, what can we do as the current code in 
board/sunxi/board.c is just as ugly a wart.

Olliver

>
> Regards,
> Simon
>