Messages in this thread | | | Date | Fri, 1 Aug 2008 14:10:43 -0700 | From | "Paul E. McKenney" <> | Subject | Re: Kernel WARNING: at net/core/dev.c:1330 __netif_schedule+0x2c/0x98() |
| |
On Thu, Jul 24, 2008 at 09:06:51PM +1000, Nick Piggin wrote: > On Thursday 24 July 2008 20:55, Miklos Szeredi wrote: > > On Thu, 24 Jul 2008, Nick Piggin wrote: > > > Hey, something kind of cool (and OT) I've just thought of that we can > > > do with ticket locks is to take tickets for 2 (or 64K) nested locks, > > > and then wait for them both (all), so the cost is N*lock + longest spin, > > > rather than N*lock + N*avg spin. > > > > Isn't this deadlocky? > > > > E.g. one task takes ticket x=1, then other task comes in and takes x=2 > > and y=1, then first task takes y=2. Then neither can actually > > complete both locks. > > Oh duh of course you still need mutual exclusion from the first lock > to order the subsequent :P > > So yeah it only works for N > 2 locks, and you have to spin_lock the > first one... so unsuitable for scheduler.
Or sort the locks by address or some such.
Thanx, Paul
| |