> Hmm, would it be appropriate for the ep_match to verify a PHY exists
> before allowing ep_attach?

No.  The driver is sure that that it's got the right device for the
driver, so the match should return success.  If some other driver
knows more, it can return a higher match value.

"weird" diagnostic output from the known-correct driver is much, much
better than the device not being attached and the configuration code
simply reporting that no driver was "configured" for the device.
Ideally, the "weird" diagnostic output could be made better, but it
seems it was enough to let _somebody_ tell you what was going on.  If
all you had was a "not configured" message, and you _knew_ you had the
'right' driver in the kernel, all you'd get was a useless message and
a lot of frustration.



cgd