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

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

On Monday 18 January 2010 10:11 AM, Anton Blanchard wrote:
> 
> Hi,
> 
> Another try at this percpu_counter batch issue with CONFIG_VIRT_CPU_ACCOUNTING
> and CONFIG_CGROUP_CPUACCT enabled. Thoughts?
> 
> --
> 
> 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.
> 
> Based on an idea by KOSAKI Motohiro, this patch scales the batch value by
> cputime_one_jiffy such that we have the same batch limit as we would
> if CONFIG_VIRT_CPU_ACCOUNTING was disabled. His patch did this once at boot
> but that initialisation happened too early on PowerPC (before time_init)
> and it was never updated at runtime as a result of a hotplug cpu add/remove.
> 
> This patch instead scales percpu_counter_batch by cputime_one_jiffy at
> runtime, which keeps the batch correct even after cpu hotplug operations.
> We cap it at INT_MAX in case of overflow.
> 
> For architectures that do not support CONFIG_VIRT_CPU_ACCOUNTING,
> cputime_one_jiffy is the constant 1 and gcc is smart enough to
> optimise min(s32 percpu_counter_batch, INT_MAX) to just percpu_counter_batch
> at least on x86 and PowerPC. So there is no need to add an #ifdef.
> 
> On a 64 thread PowerPC box with CONFIG_VIRT_CPU_ACCOUNTING and 
> CONFIG_CGROUP_CPUACCT enabled, a context switch microbenchmark is 234x faster
> and almost matches a CONFIG_CGROUP_CPUACCT disabled kernel:
> 
> CONFIG_CGROUP_CPUACCT disabled:		16906698 ctx switches/sec
> CONFIG_CGROUP_CPUACCT enabled:		   61720 ctx switches/sec
> CONFIG_CGROUP_CPUACCT + patch:		16663217 ctx switches/sec
> 
> Tested with:
> 
> wget http://ozlabs.org/~anton/junkcode/context_switch.c
> make context_switch
> for i in `seq 0 63`; do taskset -c $i ./context_switch & done
> vmstat 1
> 
> Signed-off-by: Anton Blanchard <anton@samba.org>
> ---
> 
> Note: ccing ia64 and s390 who have not yet added code to statically
> initialise cputime_one_jiffy at boot. 
> See a42548a18866e87092db93b771e6c5b060d78401 (cputime: Optimize
> jiffies_to_cputime(1) for details). Adding this would help optimise not only
> this patch but many other areas of the scheduler when
> CONFIG_VIRT_CPU_ACCOUNTING is enabled.
> 
> Index: linux.trees.git/kernel/sched.c
> ===================================================================
> --- linux.trees.git.orig/kernel/sched.c	2010-01-18 14:27:12.000000000 +1100
> +++ linux.trees.git/kernel/sched.c	2010-01-18 15:21:59.000000000 +1100
> @@ -10894,6 +10894,7 @@ static void cpuacct_update_stats(struct 
>  		enum cpuacct_stat_index idx, cputime_t val)
>  {
>  	struct cpuacct *ca;
> +	int batch;
> 
>  	if (unlikely(!cpuacct_subsys.active))
>  		return;
> @@ -10901,8 +10902,9 @@ static void cpuacct_update_stats(struct 
>  	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);
> +		__percpu_counter_add(&ca->cpustat[idx], val, batch);
>  		ca = ca->parent;
>  	} while (ca);
>  	rcu_read_unlock();

Looks good to me, but I'll test it as well and revert back. I think we
might need to look at the call side where we do the percpu_counter_read().

Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>

Balbir