sfb_dump_stats() only runs with RTNL held,
reading fields that can be changed in qdisc fast path.

Add READ_ONCE()/WRITE_ONCE() annotations.

Alternative would be to acquire the qdisc spinlock, but our long-term
goal is to make qdisc dump operations lockless as much as we can.

tc_sfb_xstats fields don't need to be latched atomically,
otherwise this bug would have been caught earlier.

Fixes: edb09eb17ed8 ("net: sched: do not acquire qdisc spinlock in qdisc/class stats dump")
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/sched/sch_sfb.c | 46 +++++++++++++++++++++++++++------------------
 1 file changed, 28 insertions(+), 18 deletions(-)

diff --git a/net/sched/sch_sfb.c b/net/sched/sch_sfb.c
index 2258567cbcaf70863eace85d347efda882a00145..315edd7f87fcf1600d69a3a92733ddb9fee55e99 100644
--- a/net/sched/sch_sfb.c
+++ b/net/sched/sch_sfb.c
@@ -202,11 +202,14 @@ static u32 sfb_compute_qlen(u32 *prob_r, u32 *avgpm_r, const struct sfb_sched_da
 	const struct sfb_bucket *b = &q->bins[q->slot].bins[0][0];
 
 	for (i = 0; i < SFB_LEVELS * SFB_NUMBUCKETS; i++) {
-		if (qlen < b->qlen)
-			qlen = b->qlen;
-		totalpm += b->p_mark;
-		if (prob < b->p_mark)
-			prob = b->p_mark;
+		u32 b_qlen = READ_ONCE(b->qlen);
+		u32 b_mark = READ_ONCE(b->p_mark);
+
+		if (qlen < b_qlen)
+			qlen = b_qlen;
+		totalpm += b_mark;
+		if (prob < b_mark)
+			prob = b_mark;
 		b++;
 	}
 	*prob_r = prob;
@@ -295,7 +298,8 @@ static int sfb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 
 	if (unlikely(sch->q.qlen >= q->limit)) {
 		qdisc_qstats_overlimit(sch);
-		q->stats.queuedrop++;
+		WRITE_ONCE(q->stats.queuedrop,
+			   q->stats.queuedrop + 1);
 		goto drop;
 	}
 
@@ -348,7 +352,8 @@ static int sfb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 
 	if (unlikely(minqlen >= q->max)) {
 		qdisc_qstats_overlimit(sch);
-		q->stats.bucketdrop++;
+		WRITE_ONCE(q->stats.bucketdrop,
+			   q->stats.bucketdrop + 1);
 		goto drop;
 	}
 
@@ -374,7 +379,8 @@ static int sfb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		}
 		if (sfb_rate_limit(skb, q)) {
 			qdisc_qstats_overlimit(sch);
-			q->stats.penaltydrop++;
+			WRITE_ONCE(q->stats.penaltydrop,
+				   q->stats.penaltydrop + 1);
 			goto drop;
 		}
 		goto enqueue;
@@ -390,14 +396,17 @@ static int sfb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 			 * In either case, we want to start dropping packets.
 			 */
 			if (r < (p_min - SFB_MAX_PROB / 2) * 2) {
-				q->stats.earlydrop++;
+				WRITE_ONCE(q->stats.earlydrop,
+					   q->stats.earlydrop + 1);
 				goto drop;
 			}
 		}
 		if (INET_ECN_set_ce(skb)) {
-			q->stats.marked++;
+			WRITE_ONCE(q->stats.marked,
+				   q->stats.marked + 1);
 		} else {
-			q->stats.earlydrop++;
+			WRITE_ONCE(q->stats.earlydrop,
+				   q->stats.earlydrop + 1);
 			goto drop;
 		}
 	}
@@ -410,7 +419,8 @@ static int sfb_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		qdisc_qlen_inc(sch);
 		increment_qlen(&cb, q);
 	} else if (net_xmit_drop_count(ret)) {
-		q->stats.childdrop++;
+		WRITE_ONCE(q->stats.childdrop,
+			   q->stats.childdrop + 1);
 		qdisc_qstats_drop(sch);
 	}
 	return ret;
@@ -599,12 +609,12 @@ static int sfb_dump_stats(struct Qdisc *sch, struct gnet_dump *d)
 {
 	struct sfb_sched_data *q = qdisc_priv(sch);
 	struct tc_sfb_xstats st = {
-		.earlydrop = q->stats.earlydrop,
-		.penaltydrop = q->stats.penaltydrop,
-		.bucketdrop = q->stats.bucketdrop,
-		.queuedrop = q->stats.queuedrop,
-		.childdrop = q->stats.childdrop,
-		.marked = q->stats.marked,
+		.earlydrop = READ_ONCE(q->stats.earlydrop),
+		.penaltydrop = READ_ONCE(q->stats.penaltydrop),
+		.bucketdrop = READ_ONCE(q->stats.bucketdrop),
+		.queuedrop = READ_ONCE(q->stats.queuedrop),
+		.childdrop = READ_ONCE(q->stats.childdrop),
+		.marked = READ_ONCE(q->stats.marked),
 	};
 
 	st.maxqlen = sfb_compute_qlen(&st.maxprob, &st.avgprob, q);
-- 
2.53.0.1213.gd9a14994de-goog