I was talking together with Emanuele on this topic and we are confused on how to proceed.
From the various comments and tests in this thread, to me the actual
code change is correct, the dp83867 does not support EEE and we have to explicitly disable it in the dp83867 driver.
As of now we do not have a clear shared understanding on what is going on in the stmmac driver. And the commit message is not correct on this regard.
This patch is already merged [1] in netdev tree, should we send a series reverting this commit and another commit with just the same change and a different commit message?
No, if its merged, its merged. What should come out of this is a learning, please be precise with commit messages, more detail rather than less, and provide as much supporting evidence as you can for any claims you make.
In parallel, unrelated to the dp83867 topic, Emanuele is trying to help figuring out why the actual behavior of the stmmac changed after Russell refactoring. And it's clear that this change in behavior is not expected.
Great.
Andrew