On Wed, May 04, 2005 at 11:32:10PM +0400, Roman Kagan wrote: > On Wed, May 04, 2005 at 11:19:15AM -0400, Alan Stern wrote: > > Here's an idea I had recently to improve the klist library; tell me what > > you think. As a space optimization, instead of storing a struct > > completion in every klist_node, just put a wait_queue_header in struct > > klist. Under normal usage it's very rare to remove more than one > > klist_node from a klist at any time, so false wakeups won't be a problem. > > I personally wouldn't care much about optimizations, especially minor, > before correctness is well proven. And as a matter of fact I'm yet to > be convinced that klists, with their rather heavy and complex structure > and non-intuitive semantics, provide any usability improvement over > traditional lists. Maybe I'm just not seeing the big picture...
Removal of the use of the big, huge, nasty subsystem-wide rw-semaphore is the goal here. klists solve that, and now, we can start doing the other things that people have been wanting for a long time (binding/unbinding from userspace, multi-threaded probing, creating new devices from within a probe() call, etc.) thanks, greg k-h ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. Get your fingers limbered up and give it your best shot. 4 great events, 4 opportunities to win big! Highest score wins.NEC IT Guy Games. Play to win an NEC 61 plasma display. Visit http://www.necitguy.com/?r=20 _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel