Messages in this thread | | | Subject | Re: Deadlock scenario in regulator core | From | Peter Zijlstra <> | Date | Fri, 25 Mar 2011 11:59:16 +0100 |
| |
On Wed, 2011-03-23 at 10:42 +0000, Mark Brown wrote: > > That's helpful, though it doesn't really say anything about how one > picks subclass?
With care ;-)
Its most useful for recursive use of the same lock class, typically we only have 2 locks held and we can use the SINGLE_DEPTH_NESTING thing. Some few sites use subclasses for larger things, some are ok, some are dubious (ext4 comes to mind).
And like said in the other email, be careful with lockdep annotations, you can actually annotate real deadlocks away.
| |