[U-Boot] [PATCH] mii: miiphy register address width change

[Andy Fleming]

On Thu, Oct 20, 2011 at 5:42 PM, Mike Frysinger <vapier at gentoo.org> wrote:
> On Thursday 20 October 2011 17:45:33 Andy Fleming wrote:
>> On Thu, Oct 20, 2011 at 8:58 AM, Mike Frysinger <vapier at gentoo.org> wrote:
>> > On Thursday 20 October 2011 09:55:34 Kumar Gala wrote:
>> >> On Oct 20, 2011, at 8:26 AM, Mike Frysinger wrote:
>> >> > On Thursday 20 October 2011 06:34:33 Kumar Nath, Chandan wrote:
>> >> >> This patch was acked on September 21, but in latest code base I could
>> >> >> not find this patch. Is there anything left which I need to take care
>> >> >> in my patch. If so, please let me know so that this can be picked up.
>> >> >
>> >> > i don't merge net patches.  wolfgang does.  i don't think there's
>> >> > anything left for you to do.
>> >>
>> >> I think this breaks 10g support.  I'm pretty sure Andy made the data
>> >> type a short for a reason.
>> >
>> > the data type in mainline is 8bits (char).  Chandan is fixing it to be
>> > 16bits (short).  if 10g breaks with a short, that sounds like a bug in
>> > the 10g code we should figure out + fix.  Linux is using a short just
>> > fine afaict.
>>
>> Actually, there's some confusion, here. The function being updated by
>> this patch isn't part of phylib. It's part of the legacy miiphy code.
>> I don't think there's any reason to update it. If you are writing a
>> new driver with 10G support (which would, admittedly, require a 16-bit
>> register argument), then don't use miiphy_register. Call
>> mdio_register, and register proper phylib support. The mii command
>> isn't capable of dealing with Clause-45 MDIO transactions, anyway (no
>> devad)
>
> so the patch is correct, but we don't want to merge it because we don't want
> to encourage the old miiphy interface ?  and leaving it broken forces people
> to migrate to the new phy layer ?

I don't think the patch is correct or incorrect. It attempts to update
the miiphy_* API to support larger register offsets. I'm suggesting
that we shouldn't change the legacy API to support new features.
There's no way to specify via miiphy_write that you want to write to a
register with a 16-bit register offset, because there's no way to
specify which device to write.

So I'm not saying we shouldn't apply this patch. I'm questioning why
it was written in a hope that we can discourage new use of the old
API. I'm also not convinced the patch helps anything.

Andy