On Thu, 2015-11-26 at 18:03 +0100, Hannes Frederic Sowa wrote: > Also isn't the reason why slub exists so it can track memory regions > per-cpu.
call_rcu() and kfree_rcu() will add a grace period (multiple ms) where the cpu will likely evict from its caches the data contained in the 'about to be freed' objects, defeating the SLUB/SLAB ability to quickly reuse a freed and hot object (LIFO) This is one of the major RCU drawback : Force a FIFO behavior in object reuse while LIFO one is much better for data locality, especially with per-cpu lists. Another problem is a slightly bigger working set size, which can hurt some workloads that used to exactly fit cpu caches. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
