This is a follow-up to `[PATCH] memcg: fix slab accounting in
refill_obj_stock() trylock path`. The way the slab stat cache and the
objcg charge cache interact appears a bit too fragile. This series
factors those paths apart as much as practical.

Based on today's mm-new.

 mm/memcontrol.c | 146 +++++++++++++++++++++++++++++++++++-------------------
 1 file changed, 96 insertions(+), 50 deletions(-)