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>

include/net/netfilter/xt_rateest.h

@@ -1,19 +1,23 @@
 #ifndef _XT_RATEEST_H
 #define _XT_RATEEST_H
 
+#include <net/gen_stats.h>
+
 struct xt_rateest {
 	/* keep lock and bstats on same cache line to speedup xt_rateest_tg() */
 	struct gnet_stats_basic_packed	bstats;
 	spinlock_t			lock;
-	/* keep rstats and lock on same cache line to speedup xt_rateest_mt() */
-	struct gnet_stats_rate_est64	rstats;
+
 
 	/* following fields not accessed in hot path */
+	unsigned int			refcnt;
 	struct hlist_node		list;
 	char				name[IFNAMSIZ];
-	unsigned int			refcnt;
 	struct gnet_estimator		params;
 	struct rcu_head			rcu;
+
+	/* keep this field far away to speedup xt_rateest_mt() */
+	struct net_rate_estimator __rcu *rate_est;
 };
 
 struct xt_rateest *xt_rateest_lookup(const char *name);