working on EHCI bug in 4.x

[Julian Elischer]

general report to the list..
I'm working on a bug that is hitting my employer in FBSD4.10
and may (not sure yet) also exist in 5.x/6. (Need to get a 5.x
test machine with EHCI).

The symptom is that an IO timeout removes the 'xfer' from the interface
but fails to handshake with the hardware correctly, resulting in
the hardware DMA-ing to the buffer which has been added back to the
free list. The code in question is the same in 4/5/6 so it may
be that it is everywhere.. I now undersand the EHCI interface
enough to say it is not an OBVIOUS bug as teh code SEEMS to be doing
the handshake correctly. *sigh*

anyhow.. back to your regularly scheduled hacking..

Anyone with ideas should feel free to speak up :-)