Re: [net-next.git 1/4 (v5)] phy: add the EEE support and the way to access to the MMD registers.

[Giuseppe CAVALLARO <peppe.cavallaro@st.com>]

On 6/15/2012 6:37 PM, Ben Hutchings wrote:
> On Fri, 2012-06-15 at 08:06 +0200, Giuseppe CAVALLARO wrote:
>> Hello Ben
>>
>> On 6/14/2012 1:28 AM, Ben Hutchings wrote:
> [...]
>>> But you also use this condition to decide whether to enable TX LPI, so
>>> it's important that it does match the specification (§78.3) for whether
>>> EEE is supported - but it doesn't.  You need to work out what mode was
>>> autonegotiated, then check that the relevant bit is set in both our EEE
>>> advertising (*not* supported) and the LP EEE advertising masks.
>>
>> I've some doubts and, before resending the patch, I kindly ask you some
>> further details just on this point.
>>
>> In the code, I check if the EEE is supported and on GMII, MII and RGMII
>> and duplex mode; in case of success the Ethernet driver can enable the
>> TX LPI.
>> Indeed, I am only using the 3.20 and 7.61 registers w/o looking at the
>> 7.60. So this should be fixed, shouldn't it?
>> Am I missing anything else?
>> What do you mean when say that it doesn't match the specification
>> (§78.3)? I'm pointing to the '78.3 Capabilities Negotiation' chapter of
>> the IEEE802-3az, is it ok?
> 
> Yes that's what I mean.  As I read it, you need to check which link mode
> was autonegotiated, then the corresponding bit in 7.60 and 7.61.  If
> they're both set then EEE is supported on the current link.  (But, let
> me repeat, I have not done any work on implementing EEE, so it's
> entirely possible that I have misunderstood some things.)

Ben, you are right, the code needs this kind of check.
For example, my phy device only supports the 100BASE-TX and, with the
current implementation, the phy_init_eee could enable the EEE on 10/full
link mode and it is not good.

I'll send the new patch asap.

Thanks
Peppe

> 
> Ben.
>