Hi Johannes, Michal, Christoph, Recently I posted my thoughts on how we can handle kmem caches of dead memcgs: https://lkml.org/lkml/2014/4/20/38 The only feedback I got then was from Johannes who voted for migrating slabs of such caches to the parent memcg's cache (so called reparenting), so in this RFC I'd like to propose a draft of possible implementation of slab reparenting. I'd appreciate if you could look through it and post if it's worth developing in this direction or not. The implementation of reparenting is given in patch 3, which is the most important part of this set. Patch 1 just makes slub keep full slabs on list, and patch 2 a bit extends percpu-refcnt interface. NOTE the implementation is given only for slub, though it should be easy to implement the same hack for slab. Thanks, Vladimir Davydov (3): slub: keep full slabs on list for per memcg caches percpu-refcount: allow to get dead reference slub: reparent memcg caches' slabs on memcg offline include/linux/memcontrol.h | 4 +- include/linux/percpu-refcount.h | 11 +- include/linux/slab.h | 7 +- mm/memcontrol.c | 54 ++++--- mm/slab.h | 7 +- mm/slub.c | 299 ++++++++++++++++++++++++++++++++++----- 6 files changed, 318 insertions(+), 64 deletions(-) -- 1.7.10.4 -- 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>