[Devel] Re: [PATCH v3 6/6] expose per-taskgroup schedstats in cgroup

[Peter Zijlstra]

On Wed, 2012-05-30 at 13:48 +0400, Glauber Costa wrote:
> +static u64 tg_idle(struct task_group *tg, int cpu)
> +{
> +       u64 val;
> +
> +       if (tg != &root_task_group) {
> +               val = cfs_read_sleep(tg->se[cpu]);
> +               /* If we have rt tasks running, we're not really idle */
> +               val -= rt_rq(exec_clock, tg, cpu);
> +       } else
> +               /*
> +                * There are many errors here that we are accumulating.
> +                * However, we only provide this in the interest of having
> +                * a consistent interface for all cgroups. Everybody
> +                * probing the root cgroup should be getting its figures
> +                * from system-wide files as /proc/stat. That would be faster
> +                * to begin with...
> +                *
> +                * Ditto for steal.
> +                */
> +               val = kcpustat_cpu(cpu).cpustat[CPUTIME_IDLE] * TICK_NSEC;

You just violated 2 coding style rules in one go :-)

If one side of the if-else has braces, the other side should have too.
If a block is multi-line (regardless of multi-stmt) it should have
braces.

/me hands you a bucket full of {}.

> +       return val;
> +}