kernel panic in xhci on boot

[Brett Lymn <blymn%internode.on.net@localhost>]

On my desktop machine I am pretty much guaranteed a kernel panic when I
reboot from windows 8 into NetBSD, the traceback is:

softint_schedule()
usb_schedsoftintr()
xhci_intr1()

The assert in softint_schedule is firing because offset is 0, this is
passed in by usb_schedsoftintr, it should be the soft intr cookie.  It
looks to me like there is a race there that a softintr can try to be
queued before the softintr queue has been set up.  Usually, if I reboot
again I don't see the assertion and things work fine from that point on.
This is on an amd64 system with sources from a few days ago.

-- 
Brett Lymn