On Thu, Oct 7, 2010 at 3:45 PM, stephane eranian <eranian@googlemail.com> wrote:
quoted text
> On Thu, Oct 7, 2010 at 3:20 AM, Li Zefan <lizf@cn.fujitsu.com> wrote:
>> Stephane Eranian wrote:
>>> This kernel patch adds the ability to filter monitoring based on
>>> container groups (cgroups). This is for use in per-cpu mode only.
>>>
>>> The cgroup to monitor is passed as a file descriptor in the pid
>>> argument to the syscall. The file descriptor must be opened to
>>> the cgroup name in the cgroup filesystem. For instance, if the
>>> cgroup name is foo and cgroupfs is mounted in /cgroup, then the
>>> file descriptor is opened to /cgroup/foo. Cgroup mode is
>>> activated by passing PERF_FLAG_PID_CGROUP into the flags argument
>>> to the syscall.
>>>
>>> Signed-off-by: Stephane Eranian <eranian@google.com>
>>>
>>> ---
>>>
>>> diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
>>> index 709dfb9..67cf276 100644
>>> --- a/include/linux/cgroup.h
>>> +++ b/include/linux/cgroup.h
>>> @@ -623,6 +623,8 @@ bool css_is_ancestor(struct cgroup_subsys_state *cg,
>>>  unsigned short css_id(struct cgroup_subsys_state *css);
>>>  unsigned short css_depth(struct cgroup_subsys_state *css);
>>>
>>> +struct cgroup_subsys_state *cgroup_css_from_dir(struct file *f, int id);
>>> +
>>>  #else /* !CONFIG_CGROUPS */
>>>
>>>  static inline int cgroup_init_early(void) { return 0; }
>>> diff --git a/include/linux/cgroup_subsys.h b/include/linux/cgroup_subsys.h
>>> index ccefff0..93f86b7 100644
>>> --- a/include/linux/cgroup_subsys.h
>>> +++ b/include/linux/cgroup_subsys.h
>>> @@ -65,4 +65,8 @@ SUBSYS(net_cls)
>>>  SUBSYS(blkio)
>>>  #endif
>>>
>>> +#ifdef CONFIG_PERF_EVENTS
>>> +SUBSYS(perf)
>>> +#endif
>>> +
>>>  /* */
>>> diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
>>> index 61b1e2d..ad79f0a 100644
>>> --- a/include/linux/perf_event.h
>>> +++ b/include/linux/perf_event.h
>>> @@ -454,6 +454,7 @@ enum perf_callchain_context {
>>>
>>>  #define PERF_FLAG_FD_NO_GROUP        (1U << 0)
>>>  #define PERF_FLAG_FD_OUTPUT  (1U << 1)
>>> +#define PERF_FLAG_PID_CGROUP (1U << 2) /* pid=cgroup id, per-cpu mode */
>>>
>>>  #ifdef __KERNEL__
>>>  /*
>>> @@ -461,6 +462,7 @@ enum perf_callchain_context {
>>>   */
>>>
>>>  #ifdef CONFIG_PERF_EVENTS
>>> +# include <linux/cgroup.h>
>>>  # include <asm/perf_event.h>
>>>  # include <asm/local64.h>
>>>  #endif
>>> @@ -698,6 +700,18 @@ struct swevent_hlist {
>>>  #define PERF_ATTACH_CONTEXT  0x01
>>>  #define PERF_ATTACH_GROUP    0x02
>>>
>>> +#ifdef CONFIG_CGROUPS
>>> +struct perf_cgroup_time {
>>> +     u64 time;
>>> +     u64 timestamp;
>>> +};
>>> +
>>> +struct perf_cgroup {
>>> +     struct cgroup_subsys_state css;
>>> +     struct perf_cgroup_time *time;
>>> +};
>>
>> Can we avoid adding this perf cgroup subsystem? It has 2 disavantages:
>>
> Well, I need to maintain some timing information for each cgroup. This has
> to be stored somewhere.
>
>> - If one mounted cgroup fs without perf cgroup subsys, he can't monitor it.
>
> That's unfortunately true ;-)
>
>> - If there are several different cgroup mount points, only one can be
>>  monitored.
>>
>> To choose which cgroup hierarchy to monitor, hierarchy id can be passed
>> from userspace, which is the 2nd column below:
>>
> Ok, I will investigate this. As long as the hierarchy id is unique AND it can be
> searched, then we can use it. Using /proc is fine with me.
>
>> $ cat /proc/cgroups
>> #subsys_name    hierarchy       num_cgroups     enabled
>> debug   0       1       1
>> net_cls 0       1       1
>>

If I mount all subsystems:
mount -t cgroup none /dev/cgroup
Then, I get:
#subsys_name	hierarchy	num_cgroups	enabled
cpuset	1	1	1
cpu	        1	1	1
perf_event	1	1	1

In other words, the hierarchy id is not unique.
If the perf_event is not mounted, then hierarchy id = 0.

When I compare with my approach, if perf_event is
not mounted, then the file descriptor won't lead to the
css, and therefore you will fail and that is fine because
it means the perf_event subsystem is not instantiated
therefore it cannot be used.

In my patch, there was a missing check for a NULL
css. I fixed that now, and it works fine.

As for multiple mount points, it seems like the first
mount determines the restrictions for all mounts.
In other words, if you mount only cpuset, then no
other mount can provide more than cpuset, and vice-versa.

I have tried mounting cgroupfs in multiple places at the same
time. Whatever directory I used, I got to the right css.

Am I missing your point here?
--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/