This patch adds support to usbnet for devices based on the ASIX AX88172
USB 2.0 Ethernet chipset.

The patch frees the bulk write urb's transfer buffer in the write
bulk callback (the buffer is allocated dynamically when the urb is
submitted).  The patch also tidies up the write bulk callback
function a little bit - removing some unnecessary paranoid checks
and scheduling a soft interrupt regardless of the port's open count
(for consistency with other usb serial drivers).

One tester reported that disconnecting in the middle of a scatterlist
operation would give a lot of confusing messages.  This should improve
that situation a bunch:  most of the messages were pointless.  It also
resolves some related FIXMEs for both queue cleanup modes (cancelation,
and after transport fault).

  - now only urbs in flight have non-null urb->dev, so we can avoid
    trying to unlink unsubmitted or (common) already-completed urbs.

  - it's ok to get -EBUSY unlink status, it just means the urb's
    already being given back (though since it's in flight, it's
    not an error to try unlinking).

Also, the relevant diagnostics now use dev_*() driver model calls,
making the kernel strings smaller and the output more useful.

fix drivers/usb/misc/usbtest.c:438: warning: passing arg 1 of
`__swab32s' from incompatible pointer type by using le16_to_cpus instead
of le32_to_cpus .

It's not needed.

of requiring users to select SCSI in order to see it.

into fys.uio.no:/home/linux/bitkeeper/nfsclient-2.5

From Christoph Hellwig

From David Mosberger:

> We definitely are seeing some strange crashes in get_ksymbol_core().
> I think the reason is the lack of iterator initialization.

Yes, looks good: reset the iterator at allocation.

Remove the PCMCIA release timer that got missed in earlier patch.  This
is required for successful compilation of the driver.

into fys.uio.no:/home/linux/bitkeeper/nfsclient-2.5

nfsroot_mount().