On 04/04/2013 07:40 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange(a)redhat.com>
Currently when getting an instance of virCgroupPtr we will
create the path in all cgroup controllers. Only at the virt
driver layer are we attempting to filter controllers. This
is bad because the mere act of creating the dirs in the
controllers can have a functional impact on the kernel,
particularly for performance.
Update the virCgroupForDriver() method to accept a bitmask
of controllers to use. Only create dirs in the controllers
that are requested. When creating cgroups for domains,
respect the active controller list from the parent cgroup
Signed-off-by: Daniel P. Berrange <berrange(a)redhat.com>
---
+++ b/src/lxc/lxc_cgroup.c
@@ -530,7 +530,7 @@ virCgroupPtr virLXCCgroupCreate(virDomainDefPtr def)
int ret = -1;
int rc;
- rc = virCgroupForDriver("lxc", &driver, 1, 0);
+ rc = virCgroupForDriver("lxc", &driver, 1, 0, -1);
Do we want -1 to always mean all possible groups, or should we define a
symbolic constant which contains the mask of the maximum groups and
leave remaining bits 0? Then again, this is never passed over the wire,
so it is not an RPC compatibility concern, so I guess it doesn't matter.
+++ b/src/qemu/qemu_conf.c
@@ -134,14 +134,7 @@ virQEMUDriverConfigPtr virQEMUDriverConfigNew(bool privileged)
}
cfg->dynamicOwnership = privileged;
- cfg->cgroupControllers =
- (1 << VIR_CGROUP_CONTROLLER_CPU) |
- (1 << VIR_CGROUP_CONTROLLER_DEVICES) |
- (1 << VIR_CGROUP_CONTROLLER_MEMORY) |
- (1 << VIR_CGROUP_CONTROLLER_BLKIO) |
- (1 << VIR_CGROUP_CONTROLLER_CPUSET) |
- (1 << VIR_CGROUP_CONTROLLER_CPUACCT);
-
+ cfg->cgroupControllers = -1; /* -1 == auto-detect */
Oh, -1 isn't _all_ controllers, but auto-detect. Now it makes more sense.
ACK.
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org