On Sun, May 26, 2013 at 10:20:21PM +0900, Simon Horman wrote: > On Wed, May 22, 2013 at 10:57:36AM +0200, Thomas Renninger wrote: > > cgroup memory subsystem pre-allocates quite some memory at boot up. > > Example dmesg: > > allocated 469762048 bytes of page_cgroup > > please try 'cgroup_disable=memory' option if you don't want memory cgroups > > > > Due to maxcpu=1 it is by far not that much than in a productive kernel. > > It still seem to be around 4M statically as no Numa overhead exists with > > only one active CPU. It is still worth to disable cgroup memory > > pre-allocating in crash kernel environment. > > > > Signed-off-by: Thomas Renninger <trenn at suse.de> > > Signed-off-by: Thomas Renninger <Thomas Renninger" trenn at suse.de> > > I'm not really sure this kind of logic should be hard-coded into > kexec-tools. --append should allow callers of kexec to add this > option if it is appropriate. Agreed. cgroup_disable=memory is better not hard coded here. In F19 default config we have cgroup_disable=memory specified which is passed to kexec-tools using --append option. Thanks Vivek
