On 08/24/2012 08:12 PM, Christoph Lameter wrote:
> What is done there can be done in __kmem_cache_shutdown.
>
> This affects RCU handling somewhat. On rcu free all slab allocators
> do not refer to other management structures than the kmem_cache structure.
> Therefore these other structures can be freed before the rcu deferred
> free to the page allocator occurs.
>
> Reviewed-by: Joonsoo Kim <js1304@xxxxxxxxx>
> Signed-off-by: Christoph Lameter <cl@xxxxxxxxx>
Here is the code for that in slab_common.c:
if (!__kmem_cache_shutdown(s)) {
if (s->flags & SLAB_DESTROY_BY_RCU)
rcu_barrier();
__kmem_cache_destroy(s);
} ...
All that code that used to belong in __kmem_cache_destroy(), will not be
executed in kmem_cache_shutdown() without an rcu_barrier.
You need at least Paul's ack here to guarantee it is safe, but I believe
it is not. Take a look for instance at 7ed9f7e5db5, which describes a
subtle bug arising from such a situation.
--
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>