Re: [PATCH] sched: cpuacct: Use bigger percpu counter batch values for stats counters

[Balbir Singh <balbir@linux.vnet.ibm.com>]

On Tuesday 26 January 2010 04:44 AM, Andrew Morton wrote:
> On Mon, 18 Jan 2010 15:41:42 +1100
> Anton Blanchard <anton@samba.org> wrote:
> 
>> When CONFIG_VIRT_CPU_ACCOUNTING and CONFIG_CGROUP_CPUACCT are enabled we can
>> call cpuacct_update_stats with values much larger than percpu_counter_batch.
>> This means the call to percpu_counter_add will always add to the global count
>> which is protected by a spinlock and we end up with a global spinlock in
>> the scheduler.
> 
> When one looks at the end result:
> 
> : static void cpuacct_update_stats(struct task_struct *tsk,
> : 		enum cpuacct_stat_index idx, cputime_t val)
> : {
> : 	struct cpuacct *ca;
> : 	int batch;
> : 
> : 	if (unlikely(!cpuacct_subsys.active))
> : 		return;
> : 
> : 	rcu_read_lock();
> : 	ca = task_ca(tsk);
> : 
> : 	batch = min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX);
> : 	do {
> : 		__percpu_counter_add(&ca->cpustat[idx], val, batch);
> : 		ca = ca->parent;
> : 	} while (ca);
> : 	rcu_read_unlock();
> : }
> 
> the code (which used to be quite obvious) becomes pretty unobvious.  In
> fact it looks quite wrong.
> 
> Shouldn't there be a comment there explaining wtf is going on?

Andrew,

I guess a lot of the changelog and comments are in the email history,
but your point on the comment is valid. Why does it look quite wrong to you?

cputime_one_jiffy tells us how many cputime_t's we've gotten in one
jiffy. If virtual accounting is enabled, this number is quite large, and
1 if virtual accounting is not enabled. Overall the value is set to 32
for non-virtual accounting enabled systems. On systems that support
virtual accounting, the value is set to 32*cputime_per_jifffy, so the
per cpu counter syncs up roughly once in 32 jiffies assuming
cpuacct_update_stats is called once per jiffy for non-virtual machines.

If the above comment, pleases you I'll polish it and send it across.
Anton, could you please confirm what I've said above is indeed correct.



-- 
Three Cheers,
Balbir Singh