Changelog v1
1. Split cgroup_disable into cgroup_disable and cgroup_enable
2. Remove cgroup_toggleDue to the overhead of the memory controller. The
memory controller is now disabled by default. This patch adds cgroup_enable.If everyone agrees on this approach and likes it, should we push this
into 2.6.25?Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
---Documentation/kernel-parameters.txt | 3 +++
kernel/cgroup.c | 28 ++++++++++++++++++++++++++--
mm/memcontrol.c | 1 +
3 files changed, 30 insertions(+), 2 deletions(-)diff -puN kernel/cgroup.c~memory-controller-default-option-off kernel/cgroup.c
--- linux-2.6.25-rc8/kernel/cgroup.c~memory-controller-default-option-off 2008-04-07 16:24:28.000000000 +0530
+++ linux-2.6.25-rc8-balbir/kernel/cgroup.c 2008-04-07 18:30:31.000000000 +0530
@@ -3077,8 +3077,8 @@ static int __init cgroup_disable(char *sif (!strcmp(token, ss->name)) {
ss->disabled = 1;
- printk(KERN_INFO "Disabling %s control group"
- " subsystem\n", ss->name);
+ printk(KERN_INFO "%s control group "
+ "is disabled\n", ss->name);
break;
}
}
@@ -3086,3 +3086,27 @@ static int __init cgroup_disable(char *s
return 1;
}
__setup("cgroup_disable=", cgroup_disable);
+
+static int __init cgroup_enable(char *str)
+{
+ int i;
+ char *token;
+
+ while ((token = strsep(&str, ",")) != NULL) {
+ if (!*token)
+ continue;
+
+ for (i = 0; i < CGROUP_SUBSYS_COUNT; i++) {
+ struct cgroup_subsys *ss = subsys[i];
+
+ if (!strcmp(token, ss->name)) {
+ ss->disabled = 0;
+ printk(KERN_INFO "%s control group "
+ "is enabled\n", ss->name);
+ break;
+ }
+ }
+ }
+ return 1;
+}
+__setup("cgroup_enable=", cgroup_enable);
diff -puN mm/memcontrol.c~memory-controller-default-option-off mm/memcontrol.c
--- linux-2.6.25-rc8/mm/memcontrol.c~memory-controller-default-option-off 2008-04-07 16:24:28.000000000 +0530...
On Mon, 07 Apr 2008 18:32:15 +0530
BTW, how the user can know which controllers are on/off at default ?
All controllers are off ?Thanks,
-Kame--
/proc/cgroups has an enabled field (fourth one). That should show what is
enabled/disabled. I've also documented it in
Documentation/kernel-parameters.txt. I intend to enable the memory controller
again, once we bring down the overhead.--
Warm Regards,
Balbir Singh
Linux Technology Center
IBM, ISTL
--
Good idea - but you could just use the same handler function for both
of these (with a one-line wrapper for each to pass disabled=1 or
disabled=0)Paul
--
Acked-by: KOSAKI Motohiro<kosaki.motohiro@jp.fujitsu.com>
--
| Tarkan Erimer | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Andi Kleen | [PATCH x86] [0/16] Various i386/x86-64 changes |
| Vladislav Bolkhovitin | Re: Integration of SCST in the mainstream Linux kernel |
| Pavel Roskin | ndiswrapper and GPL-only symbols redux |
git: | |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| Natalie Protasevich | [BUG] New Kernel Bugs |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Arjan van de Ven | Re: [GIT]: Networking |
