| From: |
| KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> |
| To: |
| LKML <linux-kernel@vger.kernel.org> |
| Subject: |
| [PATCH -mm][preview] memcg: a patch series for next [0/9] |
| Date: |
| Tue, 19 Aug 2008 17:30:14 +0900 |
| Message-ID: |
| <20080819173014.17358c17.kamezawa.hiroyu@jp.fujitsu.com> |
| Cc: |
| "balbir@linux.vnet.ibm.com" <balbir@linux.vnet.ibm.com>,
"yamamoto@valinux.co.jp" <yamamoto@valinux.co.jp>,
"nishimura@mxp.nes.nec.co.jp" <nishimura@mxp.nes.nec.co.jp>,
ryov@valinux.co.jp |
| Archive-link: |
| Article,
Thread
|
Hi,
This post is for showing what I'm trying now.
This patch set is for memory resource controller.
4 purposes here.
- improve performance of memcg.
- remove lock_page_cgroup()
- making page_cgroup->flags to be atomic_ops.
- support mem+swap controller.
But this is still under test and the series is not well organised.
and base tree is old. (2.6.27-rc1-mm1) I'll rebase this set to newer mmtom tree.
Maybe this set have some troubles/objections but I think the direction is not bad.
Patch description. (patch ordering is bad. I'll fix in the next post.)
[1/9] ... private_counter ...replace res_counter with my own counter.
This is for supporting mem+swap controller.
(And I think memcg has a bit different characteristics from other
users of res_counter....)
[2/9] ... change-order-uncharge ...
This patch is for making it easy to handle swap-cache.
[3/9] ... atomic_flags
This patch changes operations for page_cgroup->flags to be atomic_ops.
[4/9] ... delayed freeing.
delaying to free page_cgroup at uncharge.
[5/9] ... RCU freeing of page_cgroup
free page_cgroup by RCU.
[6/9] ... lockress page cgroup.
remove lock_page_cgroup() and use RCU semantics.
[7/9] ... add preftech
add prefetch() macro
[8/9] ... mem+swap controller base.
introduce mem+swap controller. A bit big patch....but have tons of TODO.
and have troubles. (it seems it's difficult to cause OOM killer.)
[9/9] ... mem+swap controller control files.
add mem+swap controller's control files.
I'd like to push patch [2,3,4,5,6,7] first.
Thanks,
-Kame