Re: [RFC][PATCH] sched: take rt_rq->rt_runtime_lock around rt_rq->rt_runtime modification

[Peter Zijlstra <peterz@infradead.org>]

On Thu, 2011-07-28 at 22:13 +0200, Jesper Juhl wrote:
> Everywhere (that I could find) where we modify rt_rq->rt_runtime we hold 
> rt_rq->rt_runtime_lock except in alloc_rt_sched_group(). Shouldn't we do 
> so - as per this patch. ???
> 
> Signed-off-by: Jesper Juhl <jj@chaosbits.net>
> ---
>  kernel/sched.c |    2 ++
>  1 files changed, 2 insertions(+), 0 deletions(-)
> 
> diff --git a/kernel/sched.c b/kernel/sched.c
> index ccacdbd..d5a3737 100644
> --- a/kernel/sched.c
> +++ b/kernel/sched.c
> @@ -8488,7 +8488,9 @@ int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
>  			goto err_free_rq;
>  
>  		init_rt_rq(rt_rq, cpu_rq(i));
> +		raw_spin_lock(&rt_rq->rt_runtime_lock);
>  		rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
> +		raw_spin_unlock(&rt_rq->rt_runtime_lock);
>  		init_tg_rt_entry(tg, rt_rq, rt_se, i, parent->rt_se[i]);

This is init code, the rt_rq is fresh and isn't exposed yet. There isn't
any concurrency.