schedule_work and speedtch.c

[Duncan Sands]

> Hmm, I didn't realize keventd runs everything synchronously...

Yes, it's a nasty flaw.

> > I guess the speedtouch (or the usbatm core) could
> > get its own workqueue.
> 
> Agreed.  I think doing it in the usbatm core is better: it can be useful
> to other drivers too, and to the core itself for doing usb_clear_halt()s
> (e.g. like in the patch I've sent).

In that case, maybe we could use the workqueue for heavy_init too.

Ciao,

D.