axe0: read PHY failed

[Luke]

> Comparing to the NetBSD driver, I added the USBD_FORCE_SHORT_XFER flag and now
> it is working perfectly. Can someone please review the patch below and commit
> if appropriate.

Does this patch limit the speed?

I fought with this same problem for a couple of months and discovered that 
limiting the speed would stop the panics, but it made the device so 
slow as to be useless for my purposes.

For example, if I attached the device to an ohci controller, I never had 
any problems.
If I attched it to an ehci controller, but set the debugging sysctl 
variables high enough to force the traffic to slow down, I never had any 
problems (except for filling up my system message log with garbage).

If your patch can stop the panics without limiting the speed of the device 
to something like 2Mbs then great!!  :)