Darren Reed wrote:
> Looking at this webrev vs the first, you've brought the
> PHY code out into specific files - "natsemi", "intel", etc.
>
> Can you explain what the architecture is that you're
> pursuing here?
>
> Do you see an end goal being an API that supports
> PHYs to be delivered as modules, seperate to NICs?
Possibly. At the moment I don't think we need to do so, but I've tried
to design a reasonably narrow interface for it to be possible -- but not
as narrow (for example) as the BSD interface. I want to make better
code reuse possible than BSD does.
At the moment, this is more like the AC'97 code -- by putting vendor
specific pieces together in one file, it makes it possible to quickly
determine how a PHY deviates from the standard without complicating the
rest of the MII core.
So put another way, this division (at present) is mostly about sound
software engineering, and not about (at this time) a pluggable interface.
If PSARC says that a pluggable interface is desired, I can create one
for sure, by creating a new "mii" module class. I'd hoped to avoid that
though -- I don't think the problem warrants the extra complication.
- Garrett
