Skip to content

memcg: make memcg's life cycle the same as cgroup

Li Zefan cgroups linux-mm lkml 4 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v4 Latest9 patches 2 repliesIn review
  2. v30 of 12 patches 1 replySuperseded
  3. v212 patches 17 repliesSuperseded
Earlier versions · 1
VersionPostedPatchesRepliesStatus
v1 RFC7 patches40 repliesSuperseded

Changes in v4 from the cover letter

- rebased against mmotm 2013-06-06-16-19
- changed wmb() to smp_wmb() and moved it to memcg_kmem_mark_dead() and added more comment.

Patches in v4 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/9Revert "memcg: avoid dangling reference count in creation failure."mm/memcontrol.c+0 -20 replies
2/9memcg, kmem: fix reference count handling on the error pathmm/memcontrol.c+0 -80 replies
3/9memcg: use css_get() in sock_update_memcg()mm/memcontrol.c+4 -40 replies
4/9memcg: don't use mem_cgroup_get() when creating a kmemcg cachemm/memcontrol.c+5 -50 replies
5/9memcg: use css_get/put when charging/uncharging kmemmm/memcontrol.c+45 -251 reply
6/9memcg: use css_get/put for swap memcgmm/memcontrol.c+16 -100 replies
7/9memcg: don't need to get a reference to the parentmm/memcontrol.c+3 -160 replies
8/9memcg: kill memcg refcntmm/memcontrol.c+1 -170 replies
9/9memcg: don't need to free memcg via RCU or workqueuemm/memcontrol.c+5 -460 replies

Reviews trailers per patch

All 9 patches have a Reviewed-by or Acked-by.

PatchKAMEZAWA HiroyukiMichal HockoTejun Heo
1/9A-b
2/9A-b
3/9A-b A-b
4/9A-b A-b
5/9A-b A-b R-b
6/9A-b A-b
7/9A-b A-b
8/9A-b A-b
9/9A-b A-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes. Scroll the table sideways for all 3 reviewers.

Read the v4 thread

Cover letter [PATCH v4 0/9] memcg: make memcg's life cycle the same as cgroup

Hi Andrew,

All the patches in this patchset has been acked by Michal and Kamezawa-san, and it's ready to be merged into -mm.

I have another pending patchset that kills css_id, which depends on this one.

Changes since v3:
- rebased against mmotm 2013-06-06-16-19
- changed wmb() to smp_wmb() and moved it to memcg_kmem_mark_dead() and added more comment.

Changes since v2:

- rebased against 3.10-rc1
- collected some acks
- the two memcg bug fixes has been merged into mainline
- the cgroup core patch has been merged into mainline

Changes since v1:

- wrote better changelog and added acked-by and reviewed-by tags
- revised some comments as suggested by Michal
- added a wmb() in kmem_cgroup_css_offline(), pointed out by Michal
- fixed a bug which causes a css_put() never be called

Now memcg has its own refcnt, so when a cgroup is destroyed, the memcg can still be alive. This patchset converts memcg to always use css_get/put, so memcg will have the same life cycle as its corresponding cgroup.

The historical reason that memcg didn't use css_get in some cases, is that cgroup couldn't be removed if there're still css refs. The situation has changed so that rmdir a cgroup will succeed regardless css refs, but won't be freed until css refs goes down to 0.

Since the introduction of kmemcg, the memcg refcnt handling grows even more complicated. This patchset greately simplifies memcg's life cycle management.

Also, after those changes, we can convert memcg to use cgroup->id, and then we can kill css_id.

Li Zefan (7):
memcg: use css_get() in sock_update_memcg()
memcg: don't use mem_cgroup_get() when creating a kmemcg cache
memcg: use css_get/put when charging/uncharging kmem
memcg: use css_get/put for swap memcg
memcg: don't need to get a reference to the parent
memcg: kill memcg refcnt
memcg: don't need to free memcg via RCU or workqueue

Michal Hocko (2):
Revert "memcg: avoid dangling reference count in creation failure." memcg, kmem: fix reference count handling on the error path

 mm/memcontrol.c | 208 +++++++++++++++++++++-----------------------------------
 1 file changed, 77 insertions(+), 131 deletions(-)
Signature · 7 lines
1.8.0.2

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>