Re: [PATCH] get rid of read/write lock in gen_estimator

[Eric Dumazet <eric.dumazet@gmail.com>]

Stephen Hemminger a écrit :
> Don't need a read/write lock here sinc there already is a spin lock that
> is being acquired.
> 
> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
> 
> --- a/net/core/gen_estimator.c	2009-09-24 16:27:01.239755070 -0700
> +++ b/net/core/gen_estimator.c	2009-09-24 16:41:09.290751273 -0700
> @@ -101,9 +101,6 @@ struct gen_estimator_head
>  
>  static struct gen_estimator_head elist[EST_MAX_INTERVAL+1];
>  
> -/* Protects against NULL dereference */
> -static DEFINE_RWLOCK(est_lock);
> -
>  /* Protects against soft lockup during large deletion */
>  static struct rb_root est_root = RB_ROOT;
>  
> @@ -118,9 +115,8 @@ static void est_timer(unsigned long arg)
>  		u64 brate;
>  		u32 npackets;
>  		u32 rate;
> -
> +
>  		spin_lock(e->stats_lock);
> -		read_lock(&est_lock);
>  		if (e->bstats == NULL)
>  			goto skip;
>  
> @@ -136,7 +132,6 @@ static void est_timer(unsigned long arg)
>  		e->avpps += (rate >> e->ewma_log) - (e->avpps >> e->ewma_log);
>  		e->rate_est->pps = (e->avpps+0x1FF)>>10;
>  skip:
> -		read_unlock(&est_lock);
>  		spin_unlock(e->stats_lock);
>  	}
>  
> @@ -270,9 +265,9 @@ void gen_kill_estimator(struct gnet_stat
>  	while ((e = gen_find_node(bstats, rate_est))) {
>  		rb_erase(&e->node, &est_root);
>  
> -		write_lock_bh(&est_lock);
> +		spin_lock(e->stats_lock);

Are you sure _bh() variant is not needed here ?

>  		e->bstats = NULL;
> -		write_unlock_bh(&est_lock);
> +		spin_unlock(e->stats_lock);
>  
>  		list_del_rcu(&e->list);
>  		call_rcu(&e->e_rcu, __gen_kill_estimator);