David Miller wrote:

I think he's saying that he doesn't think your code is yet a
reasonable way to solve the problem, and therefore doesn't belong
upstream.

That is why it has not yet been submitted upstream.  Respectfully, I
do not think that jgarzik has yet put in the work to know if this anti
deadlock technique is reasonable or not, and he was only commenting
on some superficial blemish.  I still don't get his point, if there
was one.  He seems to be arguing in favor of a jump-off-the-cliff
approach to driver conversion.  If he wants to do the work and take
the blame when some driver inevitably breaks because of being edited
in a hurry then he is welcome to submit the necessary additional
patches.  Until then, there are about 3 nics that actually matter to
network storage at the moment, all of them GigE.

The layer 2 blemishes can be fixed easily, including avoiding the
atomic op stall and the ->dev volatility .  Thankyou for pointing
those out.

Regards,

Daniel