net: sched: Merge Qdisc::bstats and Qdisc::cpu_bstats data types

The only factor differentiating per-CPU bstats data type (struct
gnet_stats_basic_cpu) from the packed non-per-CPU one (struct
gnet_stats_basic_packed) was a u64_stats sync point inside the former.
The two data types are now equivalent: earlier commits added a u64_stats
sync point to the latter.

Combine both data types into "struct gnet_stats_basic_sync". This
eliminates redundancy and simplifies the bstats read/write APIs.

Use u64_stats_t for bstats "packets" and "bytes" data types. On 64-bit
architectures, u64_stats sync points do not use sequence counter
protection.

Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

Author: a-darwish

drivers/net/ethernet/netronome/nfp/abm/qdisc.c

@@ -458,7 +458,7 @@ nfp_abm_qdisc_graft(struct nfp_abm_link *alink, u32 handle, u32 child_handle,
 static void
 nfp_abm_stats_calculate(struct nfp_alink_stats *new,
 			struct nfp_alink_stats *old,
-			struct gnet_stats_basic_packed *bstats,
+			struct gnet_stats_basic_sync *bstats,
 			struct gnet_stats_queue *qstats)
 {
 	_bstats_update(bstats, new->tx_bytes - old->tx_bytes,

include/net/act_api.h

@@ -30,13 +30,13 @@ struct tc_action {
 	atomic_t			tcfa_bindcnt;
 	int				tcfa_action;
 	struct tcf_t			tcfa_tm;
-	struct gnet_stats_basic_packed	tcfa_bstats;
-	struct gnet_stats_basic_packed	tcfa_bstats_hw;
+	struct gnet_stats_basic_sync	tcfa_bstats;
+	struct gnet_stats_basic_sync	tcfa_bstats_hw;
 	struct gnet_stats_queue		tcfa_qstats;
 	struct net_rate_estimator __rcu *tcfa_rate_est;
 	spinlock_t			tcfa_lock;
-	struct gnet_stats_basic_cpu __percpu *cpu_bstats;
-	struct gnet_stats_basic_cpu __percpu *cpu_bstats_hw;
+	struct gnet_stats_basic_sync __percpu *cpu_bstats;
+	struct gnet_stats_basic_sync __percpu *cpu_bstats_hw;
 	struct gnet_stats_queue __percpu *cpu_qstats;
 	struct tc_cookie	__rcu *act_cookie;
 	struct tcf_chain	__rcu *goto_chain;
@@ -206,7 +206,7 @@ static inline void tcf_action_update_bstats(struct tc_action *a,
 					    struct sk_buff *skb)
 {
 	if (likely(a->cpu_bstats)) {
-		bstats_cpu_update(this_cpu_ptr(a->cpu_bstats), skb);
+		bstats_update(this_cpu_ptr(a->cpu_bstats), skb);
 		return;
 	}
 	spin_lock(&a->tcfa_lock);

include/net/gen_stats.h

@@ -7,15 +7,17 @@
 #include <linux/rtnetlink.h>
 #include <linux/pkt_sched.h>
 
-/* Note: this used to be in include/uapi/linux/gen_stats.h */
-struct gnet_stats_basic_packed {
-	__u64	bytes;
-	__u64	packets;
-	struct u64_stats_sync syncp;
-};
-
-struct gnet_stats_basic_cpu {
-	struct gnet_stats_basic_packed bstats;
+/* Throughput stats.
+ * Must be initialized beforehand with gnet_stats_basic_sync_init().
+ *
+ * If no reads can ever occur parallel to writes (e.g. stack-allocated
+ * bstats), then the internal stat values can be written to and read
+ * from directly. Otherwise, use _bstats_set/update() for writes and
+ * gnet_stats_add_basic() for reads.
+ */
+struct gnet_stats_basic_sync {
+	u64_stats_t bytes;
+	u64_stats_t packets;
 	struct u64_stats_sync syncp;
 } __aligned(2 * sizeof(u64));
 
@@ -35,7 +37,7 @@ struct gnet_dump {
 	struct tc_stats   tc_stats;
 };
 
-void gnet_stats_basic_packed_init(struct gnet_stats_basic_packed *b);
+void gnet_stats_basic_sync_init(struct gnet_stats_basic_sync *b);
 int gnet_stats_start_copy(struct sk_buff *skb, int type, spinlock_t *lock,
 			  struct gnet_dump *d, int padattr);
 
@@ -46,16 +48,16 @@ int gnet_stats_start_copy_compat(struct sk_buff *skb, int type,
 
 int gnet_stats_copy_basic(const seqcount_t *running,
 			  struct gnet_dump *d,
-			  struct gnet_stats_basic_cpu __percpu *cpu,
-			  struct gnet_stats_basic_packed *b);
+			  struct gnet_stats_basic_sync __percpu *cpu,
+			  struct gnet_stats_basic_sync *b);
 void gnet_stats_add_basic(const seqcount_t *running,
-			  struct gnet_stats_basic_packed *bstats,
-			  struct gnet_stats_basic_cpu __percpu *cpu,
-			  struct gnet_stats_basic_packed *b);
+			  struct gnet_stats_basic_sync *bstats,
+			  struct gnet_stats_basic_sync __percpu *cpu,
+			  struct gnet_stats_basic_sync *b);
 int gnet_stats_copy_basic_hw(const seqcount_t *running,
 			     struct gnet_dump *d,
-			     struct gnet_stats_basic_cpu __percpu *cpu,
-			     struct gnet_stats_basic_packed *b);
+			     struct gnet_stats_basic_sync __percpu *cpu,
+			     struct gnet_stats_basic_sync *b);
 int gnet_stats_copy_rate_est(struct gnet_dump *d,
 			     struct net_rate_estimator __rcu **ptr);
 int gnet_stats_copy_queue(struct gnet_dump *d,
@@ -68,14 +70,14 @@ int gnet_stats_copy_app(struct gnet_dump *d, void *st, int len);
 
 int gnet_stats_finish_copy(struct gnet_dump *d);
 
-int gen_new_estimator(struct gnet_stats_basic_packed *bstats,
-		      struct gnet_stats_basic_cpu __percpu *cpu_bstats,
+int gen_new_estimator(struct gnet_stats_basic_sync *bstats,
+		      struct gnet_stats_basic_sync __percpu *cpu_bstats,
 		      struct net_rate_estimator __rcu **rate_est,
 		      spinlock_t *lock,
 		      seqcount_t *running, struct nlattr *opt);
 void gen_kill_estimator(struct net_rate_estimator __rcu **ptr);
-int gen_replace_estimator(struct gnet_stats_basic_packed *bstats,
-			  struct gnet_stats_basic_cpu __percpu *cpu_bstats,
+int gen_replace_estimator(struct gnet_stats_basic_sync *bstats,
+			  struct gnet_stats_basic_sync __percpu *cpu_bstats,
 			  struct net_rate_estimator __rcu **ptr,
 			  spinlock_t *lock,
 			  seqcount_t *running, struct nlattr *opt);

include/net/netfilter/xt_rateest.h

@@ -6,7 +6,7 @@
 
 struct xt_rateest {
 	/* keep lock and bstats on same cache line to speedup xt_rateest_tg() */
-	struct gnet_stats_basic_packed	bstats;
+	struct gnet_stats_basic_sync	bstats;
 	spinlock_t			lock;
 
 

include/net/pkt_cls.h

@@ -765,7 +765,7 @@ struct tc_cookie {
 };
 
 struct tc_qopt_offload_stats {
-	struct gnet_stats_basic_packed *bstats;
+	struct gnet_stats_basic_sync *bstats;
 	struct gnet_stats_queue *qstats;
 };
 
@@ -885,7 +885,7 @@ struct tc_gred_qopt_offload_params {
 };
 
 struct tc_gred_qopt_offload_stats {
-	struct gnet_stats_basic_packed bstats[MAX_DPs];
+	struct gnet_stats_basic_sync bstats[MAX_DPs];
 	struct gnet_stats_queue qstats[MAX_DPs];
 	struct red_stats *xstats[MAX_DPs];
 };

include/net/sch_generic.h

@@ -97,7 +97,7 @@ struct Qdisc {
 	struct netdev_queue	*dev_queue;
 
 	struct net_rate_estimator __rcu *rate_est;
-	struct gnet_stats_basic_cpu __percpu *cpu_bstats;
+	struct gnet_stats_basic_sync __percpu *cpu_bstats;
 	struct gnet_stats_queue	__percpu *cpu_qstats;
 	int			pad;
 	refcount_t		refcnt;
@@ -107,7 +107,7 @@ struct Qdisc {
 	 */
 	struct sk_buff_head	gso_skb ____cacheline_aligned_in_smp;
 	struct qdisc_skb_head	q;
-	struct gnet_stats_basic_packed bstats;
+	struct gnet_stats_basic_sync bstats;
 	seqcount_t		running;
 	struct gnet_stats_queue	qstats;
 	unsigned long		state;
@@ -849,43 +849,27 @@ static inline int qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	return sch->enqueue(skb, sch, to_free);
 }
 
-static inline void _bstats_update(struct gnet_stats_basic_packed *bstats,
+static inline void _bstats_update(struct gnet_stats_basic_sync *bstats,
 				  __u64 bytes, __u32 packets)
 {
 	u64_stats_update_begin(&bstats->syncp);
-	bstats->bytes += bytes;
-	bstats->packets += packets;
+	u64_stats_add(&bstats->bytes, bytes);
+	u64_stats_add(&bstats->packets, packets);
 	u64_stats_update_end(&bstats->syncp);
 }
 
-static inline void bstats_update(struct gnet_stats_basic_packed *bstats,
+static inline void bstats_update(struct gnet_stats_basic_sync *bstats,
 				 const struct sk_buff *skb)
 {
 	_bstats_update(bstats,
 		       qdisc_pkt_len(skb),
 		       skb_is_gso(skb) ? skb_shinfo(skb)->gso_segs : 1);
 }
 
-static inline void _bstats_cpu_update(struct gnet_stats_basic_cpu *bstats,
-				      __u64 bytes, __u32 packets)
-{
-	u64_stats_update_begin(&bstats->syncp);
-	_bstats_update(&bstats->bstats, bytes, packets);
-	u64_stats_update_end(&bstats->syncp);
-}
-
-static inline void bstats_cpu_update(struct gnet_stats_basic_cpu *bstats,
-				     const struct sk_buff *skb)
-{
-	u64_stats_update_begin(&bstats->syncp);
-	bstats_update(&bstats->bstats, skb);
-	u64_stats_update_end(&bstats->syncp);
-}
-
 static inline void qdisc_bstats_cpu_update(struct Qdisc *sch,
 					   const struct sk_buff *skb)
 {
-	bstats_cpu_update(this_cpu_ptr(sch->cpu_bstats), skb);
+	bstats_update(this_cpu_ptr(sch->cpu_bstats), skb);
 }
 
 static inline void qdisc_bstats_update(struct Qdisc *sch,
@@ -1317,15 +1301,15 @@ void psched_ppscfg_precompute(struct psched_pktrate *r, u64 pktrate64);
 struct mini_Qdisc {
 	struct tcf_proto *filter_list;
 	struct tcf_block *block;
-	struct gnet_stats_basic_cpu __percpu *cpu_bstats;
+	struct gnet_stats_basic_sync __percpu *cpu_bstats;
 	struct gnet_stats_queue	__percpu *cpu_qstats;
 	struct rcu_head rcu;
 };
 
 static inline void mini_qdisc_bstats_cpu_update(struct mini_Qdisc *miniq,
 						const struct sk_buff *skb)
 {
-	bstats_cpu_update(this_cpu_ptr(miniq->cpu_bstats), skb);
+	bstats_update(this_cpu_ptr(miniq->cpu_bstats), skb);
 }
 
 static inline void mini_qdisc_qstats_cpu_drop(struct mini_Qdisc *miniq)

net/core/gen_estimator.c

@@ -40,10 +40,10 @@
  */
 
 struct net_rate_estimator {
-	struct gnet_stats_basic_packed	*bstats;
+	struct gnet_stats_basic_sync	*bstats;
 	spinlock_t		*stats_lock;
 	seqcount_t		*running;
-	struct gnet_stats_basic_cpu __percpu *cpu_bstats;
+	struct gnet_stats_basic_sync __percpu *cpu_bstats;
 	u8			ewma_log;
 	u8			intvl_log; /* period : (250ms << intvl_log) */
 
@@ -60,9 +60,9 @@ struct net_rate_estimator {
 };
 
 static void est_fetch_counters(struct net_rate_estimator *e,
-			       struct gnet_stats_basic_packed *b)
+			       struct gnet_stats_basic_sync *b)
 {
-	gnet_stats_basic_packed_init(b);
+	gnet_stats_basic_sync_init(b);
 	if (e->stats_lock)
 		spin_lock(e->stats_lock);
 
@@ -76,23 +76,27 @@ static void est_fetch_counters(struct net_rate_estimator *e,
 static void est_timer(struct timer_list *t)
 {
 	struct net_rate_estimator *est = from_timer(est, t, timer);
-	struct gnet_stats_basic_packed b;
+	struct gnet_stats_basic_sync b;
+	u64 b_bytes, b_packets;
 	u64 rate, brate;
 
 	est_fetch_counters(est, &b);
-	brate = (b.bytes - est->last_bytes) << (10 - est->intvl_log);
+	b_bytes = u64_stats_read(&b.bytes);
+	b_packets = u64_stats_read(&b.packets);
+
+	brate = (b_bytes - est->last_bytes) << (10 - est->intvl_log);
 	brate = (brate >> est->ewma_log) - (est->avbps >> est->ewma_log);
 
-	rate = (b.packets - est->last_packets) << (10 - est->intvl_log);
+	rate = (b_packets - est->last_packets) << (10 - est->intvl_log);
 	rate = (rate >> est->ewma_log) - (est->avpps >> est->ewma_log);
 
 	write_seqcount_begin(&est->seq);
 	est->avbps += brate;
 	est->avpps += rate;
 	write_seqcount_end(&est->seq);
 
-	est->last_bytes = b.bytes;
-	est->last_packets = b.packets;
+	est->last_bytes = b_bytes;
+	est->last_packets = b_packets;
 
 	est->next_jiffies += ((HZ/4) << est->intvl_log);
 
@@ -121,16 +125,16 @@ static void est_timer(struct timer_list *t)
  * Returns 0 on success or a negative error code.
  *
  */
-int gen_new_estimator(struct gnet_stats_basic_packed *bstats,
-		      struct gnet_stats_basic_cpu __percpu *cpu_bstats,
+int gen_new_estimator(struct gnet_stats_basic_sync *bstats,
+		      struct gnet_stats_basic_sync __percpu *cpu_bstats,
 		      struct net_rate_estimator __rcu **rate_est,
 		      spinlock_t *lock,
 		      seqcount_t *running,
 		      struct nlattr *opt)
 {
 	struct gnet_estimator *parm = nla_data(opt);
 	struct net_rate_estimator *old, *est;
-	struct gnet_stats_basic_packed b;
+	struct gnet_stats_basic_sync b;
 	int intvl_log;
 
 	if (nla_len(opt) < sizeof(*parm))
@@ -164,8 +168,8 @@ int gen_new_estimator(struct gnet_stats_basic_packed *bstats,
 	est_fetch_counters(est, &b);
 	if (lock)
 		local_bh_enable();
-	est->last_bytes = b.bytes;
-	est->last_packets = b.packets;
+	est->last_bytes = u64_stats_read(&b.bytes);
+	est->last_packets = u64_stats_read(&b.packets);
 
 	if (lock)
 		spin_lock_bh(lock);
@@ -222,8 +226,8 @@ EXPORT_SYMBOL(gen_kill_estimator);
  *
  * Returns 0 on success or a negative error code.
  */
-int gen_replace_estimator(struct gnet_stats_basic_packed *bstats,
-			  struct gnet_stats_basic_cpu __percpu *cpu_bstats,
+int gen_replace_estimator(struct gnet_stats_basic_sync *bstats,
+			  struct gnet_stats_basic_sync __percpu *cpu_bstats,
 			  struct net_rate_estimator __rcu **rate_est,
 			  spinlock_t *lock,
 			  seqcount_t *running, struct nlattr *opt)