Re: [LTP] [PATCH] ltp cgroup: Create a hierarchy with no bound subsystems

[Garrett C. <yan...@gm...>]

On Wed, Apr 6, 2011 at 11:22 PM, Garrett Cooper <yan...@gm...> wrote:
> On Wed, Apr 6, 2011 at 10:24 PM, Gui Jianfeng
> <gui...@cn...> wrote:
>> Hi Cyril
>>
>> Any objection of this fix, if no, would you push it?
>>
>> Thanks
>> Gui
>>
>> Gui Jianfeng wrote:
>>> Some testcases are not subsystem specific, So there's no need
>>> to mount all subsystems at "cgroup/". Furthre more, if some
>>> subsystem is already mounted, it would fail if we're trying
>>> to mount all subsystems at a mountpoint.
>>> This patch creates a hierarchy with no bound subsystems.
>>>
>>> Signed-off-by: Gui Jianfeng <gui...@cn...>
>>> ---
>>> diff --git a/testcases/kernel/controllers/cgroup/cgroup_regression_test.sh b/testcases/kernel/controllers/cgroup/cgroup_regression_test.sh
>>> index 5527afc..878f91a 100755
>>> --- a/testcases/kernel/controllers/cgroup/cgroup_regression_test.sh
>>> +++ b/testcases/kernel/controllers/cgroup/cgroup_regression_test.sh
>>> @@ -97,7 +97,7 @@ test_1()
>>>       ./fork_processes &
>>>       sleep 1
>>>
>>> -     mount -t cgroup xxx cgroup/
>>> +     mount -t cgroup -o none,name=foo cgroup cgroup/
>>>       if [ $? -ne 0 ]; then
>>>               tst_resm TFAIL "failed to mount cgroup filesystem"
>>>               failed=1
>>> @@ -124,7 +124,7 @@ test_1()
>>>  #---------------------------------------------------------------------------
>>>  test_2()
>>>  {
>>> -     mount -t cgroup xxx cgroup/
>>> +     mount -t cgroup -o none,name=foo cgroup cgroup/
>>>       if [ $? -ne 0 ]; then
>>>               tst_resm TFAIL "Failed to mount cgroup filesystem"
>>>               failed=1
>>> @@ -220,7 +220,7 @@ test_4()
>>>               return
>>>       fi
>>>
>>> -     mount -t cgroup xxx cgroup/
>>> +     mount -t cgroup -o none,name=foo cgroup cgroup/
>>>       mkdir cgroup/0
>>>       rmdir cgroup/0
>>>       umount cgroup/
>>> @@ -360,7 +360,7 @@ test_7_1()
>>>
>>>  test_7_2()
>>>  {
>>> -     mount -t cgroup xxx cgroup/
>>> +     mount -t cgroup -o none,name=foo cgroup cgroup/
>>>       if [ $? -ne 0 ]; then
>>>               tst_resm TFAIL "failed to mount $subsys"
>>>               failed=1
>>> @@ -429,7 +429,7 @@ test_7()
>>>  #---------------------------------------------------------------------------
>>>  test_8()
>>>  {
>>> -     mount -t cgroup xxx cgroup/
>>> +     mount -t cgroup -o none,name=foo cgroup cgroup/
>>>       if [ $? -ne 0 ]; then
>>>               tst_resm TFAIL "failed to mount cgroup filesystem"
>>>               failed=1
>
> Tried this out in my Fedora 13 VM and it looked ok. Would be nice if
> these options were properly documented in mount(8).

    One other thing I failed to mention: it would be good to provide a
uniquifier (the script pid?) to avoid race conditions with multiple
instances of the script as name=foo is not unique.
Thanks,
-Garrett