Ethernet PHY drivers

[John Teller]

I've been working lately with the MII stuff from NetBSD - basically it's 
what FreeBSD uses.  It's not so complex that it couldn't be a reasonable 
port to RTEMS.  There's some auto-detection stuff that may not be 
desirable and the hooks into the O/S are different.  The nice thing 
about that design is that it can be left out entirely for smaller 
deployments.

For testing I can set up at least two Blackfin environments, one of 
which is the BF527 EZ-Kit, the other is a BF537 Stamp.

---JST



On 12/06/2011 06:27 PM, Chris Johns wrote:
> On 27/11/11 3:41 AM, John Teller wrote:
>> Hi,
>>
>> I'm mucking about with Blackfin networking stuff and was able to get the
>> Ethernet driver running on a BF527. At this point it only talks to the
>> PHY using the basic IEEE register set, which is OK for getting the board
>> up and running. Unfortunately, all the good stuff like interrupt
>> generation on link changes etc. are in the extended vendor-specific 
>> area.
>>
>
> Nice.
>
>> After looking around the 4.10 codebase, I see that every architecture
>> has its own PHY drivers, which are pretty much tied into the
>> architecture's built-in PHY, or a small subset of available parts.
>>
>> I've also worked with uClinux on the Blackfin, where the PHY drivers are
>> separate from the MAC. It's a big and complicated system, very much
>> overkill for what I believe RTEMS is aimed at, namely a small embedded
>> system with a predefined hardware configuration. Politics and licensing
>> aside, there is still plenty in there to be learned from.
>>
>> Are there any projects afoot to add a subset of this kind of
>> functionality? If not, I'm willing to give it a go.
>>
>
> No there is no specific MII project happening.
>
> I suppose the only comment I can make is the TCP/IP stack needs to be 
> updated to the latest version and for this we want a much closer match 
> to the FreeBSD code base, including drivers. I would expect we bring 
> across the MII layer with a new stack.
>
> Chris
>