Re: [PATCH 04/10] Display /proc/stat information per cgroup

[Glauber Costa <glommer@parallels.com>]

On 10/05/2011 04:38 PM, Peter Zijlstra wrote:
> On Wed, 2011-10-05 at 16:10 +0400, Glauber Costa wrote:
>> On 10/05/2011 12:56 PM, Peter Zijlstra wrote:
>>> On Sun, 2011-10-02 at 23:21 +0400, Glauber Costa wrote:
>>>> +struct kernel_stat *task_group_kstat(struct task_struct *p)
>>>> +{
>>>> +       struct task_group *tg;
>>>> +       struct kernel_stat *kstat;
>>>> +
>>>> +       rcu_read_lock();
>>>> +       tg = task_group(p);
>>>> +       kstat = tg->cpustat;
>>>> +       rcu_read_unlock();
>>>> +       return kstat;
>>>> +}
>>>
>>> Who keeps tg alive and kicking while you poke at its (cpustat) member?
>>
>> * All calls to this function currently pass current as a parameter
>> (Okay, maybe it is too generic and we should pass nothing at all, and
>> grab current within it)
>> * rcu_read_lock() guarantees that current will exist during this call,
>> and task_group won't change. (right?)
>
> The thing I worry about is:
>
> A (pid n)				B
>
> kstat = task_group_kstat()
> 					echo n>  /cgroup/something-else/pid
> 					rmdir /cgroup/group-that-had-A
> <timer interrupt>
>    RCU complete
>    <softirq>
>      kfree(tg) etc..
>
> kstat->foo++;<-- *BOOM*
>
>
> The only way to avoid someone moving you around is by holding some
> cgroup lock, task->alloc_lock, task->pi_lock or the rq->lock where task
> runs. Alternatively keep rcu_read_lock() around the entire kstat usage.
>
>
Well, if I understand it correctly, we'd have to hold the lock around 
the entire kstat usge as well, right?

Otherwise it can just explode. So rcu does seem the way to go. I do, 
however, see the problem you are describing.

Maybe we can remove the rcu_read_lock() call and replace with a rcu 
validation. Then patch all callers. (Including of course the current 
users of kstat_cpu() all over). What do you think?