During the 4.14.214 timeframe, we submitted the following cherry-picks/backports:
mm: memcontrol: eliminate raw access to stat and event counters mm: memcontrol: implement lruvec stat functions on top of each other mm: memcontrol: fix excessive complexity in memory.stat reporting
They were accepted and part of linux-4.14.y since 4.14.214.
These were valid fixes, but needed a few follow-ups, or regressions could occur. The regressions are described in the follow-up commits.
This 4.14 series adds these follow-ups, fixing these regressions.
Apologies for missing these commits. Two of them even had the proper Fixes tags. The other two didn't, but still mentioned the original commit in the description, so we should have caught them.
They are all clean cherry-picks, with the exception of "mm: fix oom_kill event handling", which needed minor contextual massaging.
----
Aaron Lu (1): mem_cgroup: make sure moving_account, move_lock_task and stat_cpu in the same cacheline
Greg Thelen (1): mm: writeback: use exact memcg dirty counts
Johannes Weiner (2): mm: memcontrol: fix NR_WRITEBACK leak in memcg and system stats mm: memcg: make sure memory.events is uptodate when waking pollers
Roman Gushchin (1): mm: fix oom_kill event handling
include/linux/memcontrol.h | 111 ++++++++++++++++++++++++++----------- mm/memcontrol.c | 54 ++++++++++++------ mm/oom_kill.c | 2 +- mm/vmscan.c | 2 +- 4 files changed, 118 insertions(+), 51 deletions(-)