(ANother) stall fixed, please update to HEAD

[Adrian Chadd]

Cool, thanks.

I think this particular bug (kern/170513) is different to what you've
been seeing.

Those descriptors don't look like they actually _have_ the aggregate
bit set. It's almost like what I'm queueing to the hardware is broken.

Please update to -HEAD and retest. If it's still broken (read: my
recent fixes to the restart DMA code didn't fix it) then I'll have to
start digging into things to see how it is there's only one buffer in
the aggregate list but bf_state.bfs_nframes > 1.

Thanks! Let's try to nail this one down quickly so I can shift onto
trying to design a working power save management framework and
filtered frames support so we can finally advertise 11n support by
default.



Adrian