net: sched: Protect Qdisc::bstats with u64_stats

The not-per-CPU variant of qdisc tc (traffic control) statistics,
Qdisc::gnet_stats_basic_packed bstats, is protected with Qdisc::running
sequence counter.

This sequence counter is used for reliably protecting bstats reads from
parallel writes. Meanwhile, the seqcount's write section covers a much
wider area than bstats update: qdisc_run_begin() => qdisc_run_end().

That read/write section asymmetry can lead to needless retries of the
read section. To prepare for removing the Qdisc::running sequence
counter altogether, introduce a u64_stats sync point inside bstats
instead.

Modify _bstats_update() to start/end the bstats u64_stats write
section.

For bisectability, and finer commits granularity, the bstats read
section is still protected with a Qdisc::running read/retry loop and
qdisc_run_begin/end() still starts/ends that seqcount write section.
Once all call sites are modified to use _bstats_update(), the
Qdisc::running seqcount will be removed and bstats read/retry loop will
be modified to utilize the internal u64_stats sync point.

Note, using u64_stats implies no sequence counter protection for 64-bit
architectures. This can lead to the statistics "packets" vs. "bytes"
values getting out of sync on rare occasions. The individual values will
still be valid.

[bigeasy: Minor commit message edits, init all gnet_stats_basic_packed.]

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>

12 files changed, 21 insertions, 7 deletions

diff --git a/net/sched/act_api.c b/net/sched/act_api.c
index 7dd3a2dc5fa4..0302dad42df1 100644
--- a/net/sched/act_api.c
+++ b/net/sched/act_api.c
@@ -490,6 +490,8 @@ int tcf_idr_create(struct tc_action_net *tn, u32 index, struct nlattr *est,
 		if (!p->cpu_qstats)
 			goto err3;
 	}
+	gnet_stats_basic_packed_init(&p->tcfa_bstats);
+	gnet_stats_basic_packed_init(&p->tcfa_bstats_hw);
 	spin_lock_init(&p->tcfa_lock);
 	p->tcfa_index = index;
 	p->tcfa_tm.install = jiffies;
diff --git a/net/sched/sch_atm.c b/net/sched/sch_atm.c
index 7d8518176b45..c8e1771383f9 100644
--- a/net/sched/sch_atm.c
+++ b/net/sched/sch_atm.c
@@ -548,6 +548,7 @@ static int atm_tc_init(struct Qdisc *sch, struct nlattr *opt,
 	pr_debug("atm_tc_init(sch %p,[qdisc %p],opt %p)\n", sch, p, opt);
 	INIT_LIST_HEAD(&p->flows);
 	INIT_LIST_HEAD(&p->link.list);
+	gnet_stats_basic_packed_init(&p->link.bstats);
 	list_add(&p->link.list, &p->flows);
 	p->link.q = qdisc_create_dflt(sch->dev_queue,
 				      &pfifo_qdisc_ops, sch->handle, extack);
diff --git a/net/sched/sch_cbq.c b/net/sched/sch_cbq.c
index e0da15530f0e..d01f6ec315f8 100644
--- a/net/sched/sch_cbq.c
+++ b/net/sched/sch_cbq.c
@@ -1611,6 +1611,7 @@ cbq_change_class(struct Qdisc *sch, u32 classid, u32 parentid, struct nlattr **t
 	if (cl == NULL)
 		goto failure;
 
+	gnet_stats_basic_packed_init(&cl->bstats);
 	err = tcf_block_get(&cl->block, &cl->filter_list, sch, extack);
 	if (err) {
 		kfree(cl);
diff --git a/net/sched/sch_drr.c b/net/sched/sch_drr.c
index 642cd179b7a7..319906e19a6b 100644
--- a/net/sched/sch_drr.c
+++ b/net/sched/sch_drr.c
@@ -106,6 +106,7 @@ static int drr_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
 	if (cl == NULL)
 		return -ENOBUFS;
 
+	gnet_stats_basic_packed_init(&cl->bstats);
 	cl->common.classid = classid;
 	cl->quantum	   = quantum;
 	cl->qdisc	   = qdisc_create_dflt(sch->dev_queue,
diff --git a/net/sched/sch_ets.c b/net/sched/sch_ets.c
index ed86b7021f6d..83693107371f 100644
--- a/net/sched/sch_ets.c
+++ b/net/sched/sch_ets.c
@@ -689,7 +689,7 @@ static int ets_qdisc_change(struct Qdisc *sch, struct nlattr *opt,
 		q->classes[i].qdisc = NULL;
 		q->classes[i].quantum = 0;
 		q->classes[i].deficit = 0;
-		memset(&q->classes[i].bstats, 0, sizeof(q->classes[i].bstats));
+		gnet_stats_basic_packed_init(&q->classes[i].bstats);
 		memset(&q->classes[i].qstats, 0, sizeof(q->classes[i].qstats));
 	}
 	return 0;
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index 8c64a552a64f..ef27ff3ddee4 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -892,6 +892,7 @@ struct Qdisc *qdisc_alloc(struct netdev_queue *dev_queue,
 	__skb_queue_head_init(&sch->gso_skb);
 	__skb_queue_head_init(&sch->skb_bad_txq);
 	qdisc_skb_head_init(&sch->q);
+	gnet_stats_basic_packed_init(&sch->bstats);
 	spin_lock_init(&sch->q.lock);
 
 	if (ops->static_flags & TCQ_F_CPUSTATS) {
diff --git a/net/sched/sch_gred.c b/net/sched/sch_gred.c
index 621dc6afde8f..2ddcbb2efdbb 100644
--- a/net/sched/sch_gred.c
+++ b/net/sched/sch_gred.c
@@ -364,9 +364,11 @@ static int gred_offload_dump_stats(struct Qdisc *sch)
 	hw_stats->handle = sch->handle;
 	hw_stats->parent = sch->parent;
 
-	for (i = 0; i < MAX_DPs; i++)
+	for (i = 0; i < MAX_DPs; i++) {
+		gnet_stats_basic_packed_init(&hw_stats->stats.bstats[i]);
 		if (table->tab[i])
 			hw_stats->stats.xstats[i] = &table->tab[i]->stats;
+	}
 
 	ret = qdisc_offload_dump_helper(sch, TC_SETUP_QDISC_GRED, hw_stats);
 	/* Even if driver returns failure adjust the stats - in case offload
diff --git a/net/sched/sch_hfsc.c b/net/sched/sch_hfsc.c
index b7ac30cca035..ff6ff54806fc 100644
--- a/net/sched/sch_hfsc.c
+++ b/net/sched/sch_hfsc.c
@@ -1406,6 +1406,7 @@ hfsc_init_qdisc(struct Qdisc *sch, struct nlattr *opt,
 	if (err)
 		return err;
 
+	gnet_stats_basic_packed_init(&q->root.bstats);
 	q->root.cl_common.classid = sch->handle;
 	q->root.sched   = q;
 	q->root.qdisc = qdisc_create_dflt(sch->dev_queue, &pfifo_qdisc_ops,
diff --git a/net/sched/sch_htb.c b/net/sched/sch_htb.c
index 5067a6e5d4fd..2e805b17efcf 100644
--- a/net/sched/sch_htb.c
+++ b/net/sched/sch_htb.c
@@ -1311,7 +1311,7 @@ static void htb_offload_aggregate_stats(struct htb_sched *q,
 	struct htb_class *c;
 	unsigned int i;
 
-	memset(&cl->bstats, 0, sizeof(cl->bstats));
+	gnet_stats_basic_packed_init(&cl->bstats);
 
 	for (i = 0; i < q->clhash.hashsize; i++) {
 		hlist_for_each_entry(c, &q->clhash.hash[i], common.hnode) {
@@ -1357,7 +1357,7 @@ htb_dump_class_stats(struct Qdisc *sch, unsigned long arg, struct gnet_dump *d)
 			if (cl->leaf.q)
 				cl->bstats = cl->leaf.q->bstats;
 			else
-				memset(&cl->bstats, 0, sizeof(cl->bstats));
+				gnet_stats_basic_packed_init(&cl->bstats);
 			cl->bstats.bytes += cl->bstats_bias.bytes;
 			cl->bstats.packets += cl->bstats_bias.packets;
 		} else {
@@ -1849,6 +1849,9 @@ static int htb_change_class(struct Qdisc *sch, u32 classid,
 		if (!cl)
 			goto failure;
 
+		gnet_stats_basic_packed_init(&cl->bstats);
+		gnet_stats_basic_packed_init(&cl->bstats_bias);
+
 		err = tcf_block_get(&cl->block, &cl->filter_list, sch, extack);
 		if (err) {
 			kfree(cl);
diff --git a/net/sched/sch_mq.c b/net/sched/sch_mq.c
index 9d58ecb4e80c..704e14a58f09 100644
--- a/net/sched/sch_mq.c
+++ b/net/sched/sch_mq.c
@@ -132,7 +132,7 @@ static int mq_dump(struct Qdisc *sch, struct sk_buff *skb)
 	unsigned int ntx;
 
 	sch->q.qlen = 0;
-	memset(&sch->bstats, 0, sizeof(sch->bstats));
+	gnet_stats_basic_packed_init(&sch->bstats);
 	memset(&sch->qstats, 0, sizeof(sch->qstats));
 
 	/* MQ supports lockless qdiscs. However, statistics accounting needs
diff --git a/net/sched/sch_mqprio.c b/net/sched/sch_mqprio.c
index 57427b40f0d2..fe6b4a178fc9 100644
--- a/net/sched/sch_mqprio.c
+++ b/net/sched/sch_mqprio.c
@@ -390,7 +390,7 @@ static int mqprio_dump(struct Qdisc *sch, struct sk_buff *skb)
 	unsigned int ntx, tc;
 
 	sch->q.qlen = 0;
-	memset(&sch->bstats, 0, sizeof(sch->bstats));
+	gnet_stats_basic_packed_init(&sch->bstats);
 	memset(&sch->qstats, 0, sizeof(sch->qstats));
 
 	/* MQ supports lockless qdiscs. However, statistics accounting needs
@@ -500,10 +500,11 @@ static int mqprio_dump_class_stats(struct Qdisc *sch, unsigned long cl,
 		int i;
 		__u32 qlen;
 		struct gnet_stats_queue qstats = {0};
-		struct gnet_stats_basic_packed bstats = {0};
+		struct gnet_stats_basic_packed bstats;
 		struct net_device *dev = qdisc_dev(sch);
 		struct netdev_tc_txq tc = dev->tc_to_txq[cl & TC_BITMASK];
 
+		gnet_stats_basic_packed_init(&bstats);
 		/* Drop lock here it will be reclaimed before touching
 		 * statistics this is required because the d->lock we
 		 * hold here is the look on dev_queue->qdisc_sleeping
diff --git a/net/sched/sch_qfq.c b/net/sched/sch_qfq.c
index 58a9d42b52b8..b6d989b69324 100644
--- a/net/sched/sch_qfq.c
+++ b/net/sched/sch_qfq.c
@@ -465,6 +465,7 @@ static int qfq_change_class(struct Qdisc *sch, u32 classid, u32 parentid,
 	if (cl == NULL)
 		return -ENOBUFS;
 
+	gnet_stats_basic_packed_init(&cl->bstats);
 	cl->common.classid = classid;
 	cl->deficit = lmax;