libvirt users - Aug 2013 - How to deal with LXC cgroup access control with apparmor ?

I am playing with libvirt 1.1.1 (lxc)
when I was starting a LXC container,  the process location of cgroup is pretty ,
just the root directory
from the process. But I could tune the cgroup in a container as an user that
logged, This is not accepted...

I wonder how to restrict it with apparmor ,so one can not modify files in  the
cgroup fs, e.g  the cpus or mem,
if i restrict it with "deny /sys/fs/cgroup/** wrklx,"  in apparmor
,the container woulld not start up .
"Permission denied", because that a process would mount the cgroup, it
seems done by libvirt_lxc,
Any way to restrict the cgroup in the container or just not mount cgroup in the
container ??

Any help would be appreciated, thanks .

On 08/26/2013 03:42 PM, 止语 wrote:
> I am playing with libvirt 1.1.1 (lxc)
> when I was starting a LXC container,  the process location of cgroup is
pretty ,  just the root directory
> from the process. But I could tune the cgroup in a container as an user
that logged, This is not accepted...
> 
> I wonder how to restrict it with apparmor ,so one can not modify files in 
the cgroup fs, e.g  the cpus or mem,
> if i restrict it with "deny /sys/fs/cgroup/** wrklx,"  in
apparmor ,the container woulld not start up .
> "Permission denied", because that a process would mount the
cgroup, it seems done by libvirt_lxc,
> Any way to restrict the cgroup in the container or just not mount cgroup in
the container ??
> 
> Any help would be appreciated, thanks .
> 
The simplest way is to enable user namespace for libvirt.


the below is the configuration you should do to enable user namespace

[quote]

If you want to enable user namespace,set the idmap element. the uid and gid
elements have three attributes:

start
First user id in container.
target
The first user id in container will be mapped to this target user id in host.
count
How many users in container being allowed to map to host's user.
  <idmap>
    <uid start='0' target='1000' count='10'/>
    <gid start='0' target='1000' count='10'/>
  </idmap>


[/quote]

thx, Gao feng,
If I do not want to disable the cgroup in container , is there any config file ?
or  do i have to do something to the libvirt source code
to skip it ?


 




------------------ 原始邮件 ------------------
发件人: "Gao feng"<gaofeng@cn.fujitsu.com>;
发送时间: 2013年8月26日(星期一) 下午4:06
收件人: "止语"<zhongjj@foxmail.com>; 
抄送: "libvirt-users"<libvirt-users@redhat.com>; 
主题: Re: [libvirt-users] How to deal with LXC cgroup access control withapparmor
?



On 08/26/2013 03:42 PM, 止语 wrote:
> I am playing with libvirt 1.1.1 (lxc)
> when I was starting a LXC container,  the process location of cgroup is
pretty ,  just the root directory
> from the process. But I could tune the cgroup in a container as an user
that logged, This is not accepted...
> 
> I wonder how to restrict it with apparmor ,so one can not modify files in 
the cgroup fs, e.g  the cpus or mem,
> if i restrict it with "deny /sys/fs/cgroup/** wrklx,"  in
apparmor ,the container woulld not start up .
> "Permission denied", because that a process would mount the
cgroup, it seems done by libvirt_lxc,
> Any way to restrict the cgroup in the container or just not mount cgroup in
the container ??
> 
> Any help would be appreciated, thanks .
> 
The simplest way is to enable user namespace for libvirt.


the below is the configuration you should do to enable user namespace

[quote]

If you want to enable user namespace,set the idmap element. the uid and gid
elements have three attributes:

start
First user id in container.
target
The first user id in container will be mapped to this target user id in host.
count
How many users in container being allowed to map to host's user.
  <idmap>
    <uid start='0' target='1000' count='10'/>
    <gid start='0' target='1000' count='10'/>
  </idmap>


[/quote]