net_sched: gen_estimator: complete rewrite of rate estimators

1) Old code was hard to maintain, due to complex lock chains.
   (We probably will be able to remove some kfree_rcu() in callers)

2) Using a single timer to update all estimators does not scale.

3) Code was buggy on 32bit kernel (WRITE_ONCE() on 64bit quantity
   is not supposed to work well)

In this rewrite :

- I removed the RB tree that had to be scanned in
  gen_estimator_active(). qdisc dumps should be much faster.

- Each estimator has its own timer.

- Estimations are maintained in net_rate_estimator structure,
  instead of dirtying the qdisc. Minor, but part of the simplification.

- Reading the estimator uses RCU and a seqcount to provide proper
  support for 32bit kernels.

- We reduce memory need when estimators are not used, since
  we store a pointer, instead of the bytes/packets counters.

- xt_rateest_mt() no longer has to grab a spinlock.
  (In the future, xt_rateest_tg() could be switched to per cpu counters)

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 15 insertions, 17 deletions

diff --git a/net/netfilter/xt_RATEEST.c b/net/netfilter/xt_RATEEST.c
index dbd6c4a12b97..91a373a3f534 100644
--- a/net/netfilter/xt_RATEEST.c
+++ b/net/netfilter/xt_RATEEST.c
@@ -63,7 +63,7 @@ void xt_rateest_put(struct xt_rateest *est)
 	mutex_lock(&xt_rateest_mutex);
 	if (--est->refcnt == 0) {
 		hlist_del(&est->list);
-		gen_kill_estimator(&est->bstats, &est->rstats);
+		gen_kill_estimator(&est->rate_est);
 		/*
 		 * gen_estimator est_timer() might access est->lock or bstats,
 		 * wait a RCU grace period before freeing 'est'
@@ -132,7 +132,7 @@ static int xt_rateest_tg_checkentry(const struct xt_tgchk_param *par)
 	cfg.est.interval	= info->interval;
 	cfg.est.ewma_log	= info->ewma_log;
 
-	ret = gen_new_estimator(&est->bstats, NULL, &est->rstats,
+	ret = gen_new_estimator(&est->bstats, NULL, &est->rate_est,
 				&est->lock, NULL, &cfg.opt);
 	if (ret < 0)
 		goto err2;
diff --git a/net/netfilter/xt_rateest.c b/net/netfilter/xt_rateest.c
index 7720b036d76a..1db02f6fca54 100644
--- a/net/netfilter/xt_rateest.c
+++ b/net/netfilter/xt_rateest.c
@@ -18,35 +18,33 @@ static bool
 xt_rateest_mt(const struct sk_buff *skb, struct xt_action_param *par)
 {
 	const struct xt_rateest_match_info *info = par->matchinfo;
-	struct gnet_stats_rate_est64 *r;
+	struct gnet_stats_rate_est64 sample = {0};
 	u_int32_t bps1, bps2, pps1, pps2;
 	bool ret = true;
 
-	spin_lock_bh(&info->est1->lock);
-	r = &info->est1->rstats;
+	gen_estimator_read(&info->est1->rate_est, &sample);
+
 	if (info->flags & XT_RATEEST_MATCH_DELTA) {
-		bps1 = info->bps1 >= r->bps ? info->bps1 - r->bps : 0;
-		pps1 = info->pps1 >= r->pps ? info->pps1 - r->pps : 0;
+		bps1 = info->bps1 >= sample.bps ? info->bps1 - sample.bps : 0;
+		pps1 = info->pps1 >= sample.pps ? info->pps1 - sample.pps : 0;
 	} else {
-		bps1 = r->bps;
-		pps1 = r->pps;
+		bps1 = sample.bps;
+		pps1 = sample.pps;
 	}
-	spin_unlock_bh(&info->est1->lock);
 
 	if (info->flags & XT_RATEEST_MATCH_ABS) {
 		bps2 = info->bps2;
 		pps2 = info->pps2;
 	} else {
-		spin_lock_bh(&info->est2->lock);
-		r = &info->est2->rstats;
+		gen_estimator_read(&info->est2->rate_est, &sample);
+
 		if (info->flags & XT_RATEEST_MATCH_DELTA) {
-			bps2 = info->bps2 >= r->bps ? info->bps2 - r->bps : 0;
-			pps2 = info->pps2 >= r->pps ? info->pps2 - r->pps : 0;
+			bps2 = info->bps2 >= sample.bps ? info->bps2 - sample.bps : 0;
+			pps2 = info->pps2 >= sample.pps ? info->pps2 - sample.pps : 0;
 		} else {
-			bps2 = r->bps;
-			pps2 = r->pps;
+			bps2 = sample.bps;
+			pps2 = sample.pps;
 		}
-		spin_unlock_bh(&info->est2->lock);
 	}
 
 	switch (info->mode) {