lkml.org 
[lkml]   [2008]   [Jun]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[RFC PATCH 1/5] memcg VM overcommit documentation
Date
Documentation of the VM overcommit memory controller included in the generic
memory controller documentation: basic description and usage.

Signed-off-by: Andrea Righi <righi.andrea@gmail.com>
---
Documentation/controllers/memory.txt | 29 +++++++++++++++++++++++++++++
1 files changed, 29 insertions(+), 0 deletions(-)

diff --git a/Documentation/controllers/memory.txt b/Documentation/controllers/memory.txt
index 866b9cd..e984bfb 100644
--- a/Documentation/controllers/memory.txt
+++ b/Documentation/controllers/memory.txt
@@ -12,6 +12,7 @@ c. Provides *zero overhead* for non memory controller users
d. Provides a double LRU: global memory pressure causes reclaim from the
global LRU; a cgroup on hitting a limit, reclaims from the per
cgroup LRU
+e. Provide distinct cgroup VM overcommit accounting and handling

NOTE: Swap Cache (unmapped) is not accounted now.

@@ -142,6 +143,31 @@ The reclaim algorithm has not been modified for cgroups, except that
pages that are selected for reclaiming come from the per cgroup LRU
list.

+2.5 VM overcommit accounting and handling
+
+The concept of committed VM is replicated within each cgroup as well as global
+committed memory concept. Each cgroup can set its own overcommit policy using
+the files:
+
+memory.overcommit_memory
+memory.overcommit_ratio
+
+These settings override the system sysctl settings (`vm.overcommit_memory` and
+`vm.overcommit_ratio`) and they apply locally to the cgroup they refer.
+
+Global sysctl settings are initially used by the root level cgroups. Child
+cgroups initially inherit the parent's settings. Each cgroup can change its own
+overcommit parameters at any time simply modifying the files
+`memory.overcommit_memory` and/or `memory.overcommit_ratio`.
+
+Statistics about the current committed space and limit are reported in
+`memory.overcommit_as` for each cgroup.
+
+Per-cgroup overcommit limit depends of the local cgroup overcommit settings and
+memory limit (RSS + cache) imposed by the memory controller.
+
+See "Documentation/vm/overcommit-accounting" for additional details.
+
2. Locking

The memory controller uses the following hierarchy
@@ -230,6 +256,9 @@ carried forward. The pages allocated from the original cgroup still
remain charged to it, the charge is dropped when the page is freed or
reclaimed.

+The amount of the task's committed VM, instead, is uncharged from the old
+cgroup and accounted to the newer.
+
4.3 Removing a cgroup

A cgroup can be removed by rmdir, but as discussed in sections 4.1 and 4.2, a
--
1.5.4.3


\
 
 \ /
  Last update: 2008-06-10 01:45    [W:0.193 / U:0.528 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site