usb with fast interrupts

[M. Warner Losh]

In message: <Pine.NEB.3.96L.1041112220719.26498C-100000 at fledge.watson.org>
            Robert Watson <rwatson at freebsd.org> writes:
: Sounds good to me.  Scott's comments on possibly wanting a non-task worker
: may well be valid, though.  Or you can do what acpi and other components
: do: use the task queue mechanism but provide their own kernel thread.

I'll do that if performance changes in any significant way.  Given
that we're taking GIANT out unconditionally anyway in the taskqueue,
I'm not sure that timing really matters...  It will matter more once
we push Giant further down into usb.

Warner