On Wed, 9 Sep 2015, Paul E. McKenney wrote: > Either way, Dmitry's tool got a hit on real code using the slab > allocators. If that hit is a false positive, then clearly Dmitry > needs to fix his tool, however, I am not (yet) convinced that it is a > false positive. If it is not a false positive, we might well need to > articulate the rules for use of the slab allocators. Could I get a clear definiton as to what exactly is positive? Was this using SLAB, SLUB or SLOB? > > This would all use per cpu data. As soon as a handoff is required within > > the allocators locks are being used. So I would say no. > > As in "no, it is not necessary for the caller of kfree() to invoke barrier() > in this example", right? Actually SLUB contains a barrier already in kfree(). Has to be there because of the way the per cpu pointer is being handled. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>