> Maybe a better solution now would be to restore phy_resume()'s lock- > taking behaviour, and provide a lockless __phy_resume() which can be > used internally within phylib. This means drivers using phy_resume() > would see no change. Maybe something like (untested):
Hi Russell
I was thinking the same, and have a pretty much identical untested
patch. This gets things 'fixed' and we can then later come back and
look at the overall architecture.
Andrew
