net/sched: mq: no longer acquire qdisc spinlocks in dump operations

Prepare mq_dump_common() for RTNL avoidance.

Use RCU instead of RTNL, and no longer acquire each children spinlock.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Toke Høiland-Jørgensen <toke@toke.dk>
Link: https://patch.msgid.link/20260510091455.4039245-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 2ce84b2c9423ee3b945c0a537620b2568279e1ca..9e6ed92729d282642e17a72cc578f25e1d22e4d9 100644 (file)
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -952,6 +952,15 @@ static inline void _bstats_update(struct gnet_stats_basic_sync *bstats,
        u64_stats_update_end(&bstats->syncp);
 }
 
+static inline void _bstats_set(struct gnet_stats_basic_sync *bstats,
+                              u64 bytes, u64 packets)
+{
+       u64_stats_update_begin(&bstats->syncp);
+       u64_stats_set(&bstats->bytes, bytes);
+       u64_stats_set(&bstats->packets, packets);
+       u64_stats_update_end(&bstats->syncp);
+}
+
 static inline void bstats_update(struct gnet_stats_basic_sync *bstats,
                                 const struct sk_buff *skb)
 {

diff --git a/net/sched/sch_mq.c b/net/sched/sch_mq.c
index 4172ec24a43d1c2fe56789986a46da93eb522721..5a6b14557d85fdc6c2462ee9e926155438f597b8 100644 (file)
--- a/net/sched/sch_mq.c
+++ b/net/sched/sch_mq.c
@@ -143,30 +143,42 @@ EXPORT_SYMBOL_NS_GPL(mq_attach, "NET_SCHED_INTERNAL");
 void mq_dump_common(struct Qdisc *sch, struct sk_buff *skb)
 {
        struct net_device *dev = qdisc_dev(sch);
+       struct gnet_stats_queue qstats = { 0 };
+       struct gnet_stats_basic_sync bstats;
+       const struct Qdisc *qdisc;
        unsigned int qlen = 0;
-       struct Qdisc *qdisc;
        unsigned int ntx;
 
-       gnet_stats_basic_sync_init(&sch->bstats);
-       memset(&sch->qstats, 0, sizeof(sch->qstats));
+       gnet_stats_basic_sync_init(&bstats);
 
        /* MQ supports lockless qdiscs. However, statistics accounting needs
         * to account for all, none, or a mix of locked and unlocked child
         * qdiscs. Percpu stats are added to counters in-band and locking
         * qdisc totals are added at end.
         */
+       rcu_read_lock();
        for (ntx = 0; ntx < dev->num_tx_queues; ntx++) {
-               qdisc = rtnl_dereference(netdev_get_tx_queue(dev, ntx)->qdisc_sleeping);
-               spin_lock_bh(qdisc_lock(qdisc));
+               qdisc = rcu_dereference(netdev_get_tx_queue(dev, ntx)->qdisc_sleeping);
 
-               gnet_stats_add_basic(&sch->bstats, qdisc->cpu_bstats,
-                                    &qdisc->bstats, false);
-               gnet_stats_add_queue(&sch->qstats, qdisc->cpu_qstats,
+               gnet_stats_add_basic(&bstats, qdisc->cpu_bstats,
+                                    &qdisc->bstats, true);
+               gnet_stats_add_queue(&qstats, qdisc->cpu_qstats,
                                     &qdisc->qstats);
                qlen += qdisc_qlen_lockless(qdisc);
-
-               spin_unlock_bh(qdisc_lock(qdisc));
        }
+       rcu_read_unlock();
+
+       spin_lock_bh(qdisc_lock(sch));
+       _bstats_set(&sch->bstats, u64_stats_read(&bstats.bytes),
+                   u64_stats_read(&bstats.packets));
+       spin_unlock_bh(qdisc_lock(sch));
+
+       WRITE_ONCE(sch->qstats.qlen, qstats.qlen);
+       WRITE_ONCE(sch->qstats.backlog, qstats.backlog);
+       WRITE_ONCE(sch->qstats.drops, qstats.drops);
+       WRITE_ONCE(sch->qstats.requeues, qstats.requeues);
+       WRITE_ONCE(sch->qstats.overlimits, qstats.overlimits);
+
        WRITE_ONCE(sch->q.qlen, qlen);
 }
 EXPORT_SYMBOL_NS_GPL(mq_dump_common, "NET_SCHED_INTERNAL");