Re: [PATCH] tg: add cpu's wait_count of a task group

Hello tejun:
On Sat, Jan 16, 2021 at 4:20 AM Tejun Heo [off-list ref] wrote:

Hello,

On Fri, Jan 15, 2021 at 10:30:05PM +0800, wu860403-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org wrote:

quoted

-     seq_printf(sf, "throttled_time %llu\n", cfs_b->throttled_time);

...

quoted

+     seq_printf(sf, "nr_periods %d\n"
+                "nr_throttled %d\n"
+                "throttled_usec %llu\n",
+                cfs_b->nr_periods, cfs_b->nr_throttled,
+                throttled_usec);

This is interface breaking change. I don't think we can do this at this
point.

Thanks for your reply, agree with it.

quoted

@@ -8255,6 +8265,19 @@ static int cpu_extra_stat_show(struct seq_file *sf,
                         "throttled_usec %llu\n",
                         cfs_b->nr_periods, cfs_b->nr_throttled,
                         throttled_usec);
+             if (schedstat_enabled() && tg != &root_task_group) {
+                     u64 ws = 0;
+                     u64 wc = 0;
+                     int i;
+
+                     for_each_possible_cpu(i) {
+                             ws += schedstat_val(tg->se[i]->statistics.wait_sum);
+                             wc += schedstat_val(tg->se[i]->statistics.wait_count);
+                     }
+
+                     seq_printf(sf, "wait_sum %llu\n"
+                             "wait_count %llu\n", ws, wc);
+             }

What does sum/count tell?

It can tell the task group average latency of every context switch
wait_sum is equivalent to sched_info->run_delay (the second parameter
of /proc/$pid/schedstat)
wait_count is equivalent to sched_info->pcount(the third parameter of
/proc/$pid/schedstat)

Thanks

Liming