net: dynamic ingress_queue allocation

ingress being not used very much, and net_device->ingress_queue being
quite a big object (128 or 256 bytes), use a dynamic allocation if
needed (tc qdisc add dev eth0 ingress ...)

dev_ingress_queue(dev) helper should be used only with RTNL taken.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 26 insertions, 8 deletions

diff --git a/net/core/dev.c b/net/core/dev.c
index a313bab1b75..ce6ad88c980 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2702,11 +2702,10 @@ EXPORT_SYMBOL_GPL(br_fdb_test_addr_hook);
  * the ingress scheduler, you just cant add policies on ingress.
  *
  */
-static int ing_filter(struct sk_buff *skb)
+static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
 {
         struct net_device *dev = skb->dev;
         u32 ttl = G_TC_RTTL(skb->tc_verd);
-        struct netdev_queue *rxq;
         int result = TC_ACT_OK;
         struct Qdisc *q;
 
@@ -2720,8 +2719,6 @@ static int ing_filter(struct sk_buff *skb)
         skb->tc_verd = SET_TC_RTTL(skb->tc_verd, ttl);
         skb->tc_verd = SET_TC_AT(skb->tc_verd, AT_INGRESS);
 
-        rxq = &dev->ingress_queue;
-
         q = rxq->qdisc;
         if (q != &noop_qdisc) {
                 spin_lock(qdisc_lock(q));
@@ -2737,7 +2734,9 @@ static inline struct sk_buff *handle_ing(struct sk_buff *skb,
                                          struct packet_type **pt_prev,
                                          int *ret, struct net_device *orig_dev)
 {
-        if (skb->dev->ingress_queue.qdisc == &noop_qdisc)
+        struct netdev_queue *rxq = rcu_dereference(skb->dev->ingress_queue);
+
+        if (!rxq || rxq->qdisc == &noop_qdisc)
                 goto out;
 
         if (*pt_prev) {
@@ -2745,7 +2744,7 @@ static inline struct sk_buff *handle_ing(struct sk_buff *skb,
                 *pt_prev = NULL;
         }
 
-        switch (ing_filter(skb)) {
+        switch (ing_filter(skb, rxq)) {
         case TC_ACT_SHOT:
         case TC_ACT_STOLEN:
                 kfree_skb(skb);
@@ -4940,7 +4939,6 @@ static void __netdev_init_queue_locks_one(struct net_device *dev,
 static void netdev_init_queue_locks(struct net_device *dev)
 {
         netdev_for_each_tx_queue(dev, __netdev_init_queue_locks_one, NULL);
-        __netdev_init_queue_locks_one(dev, &dev->ingress_queue, NULL);
 }
 
 unsigned long netdev_fix_features(unsigned long features, const char *name)
@@ -5452,11 +5450,29 @@ static void netdev_init_one_queue(struct net_device *dev,
 
 static void netdev_init_queues(struct net_device *dev)
 {
-        netdev_init_one_queue(dev, &dev->ingress_queue, NULL);
         netdev_for_each_tx_queue(dev, netdev_init_one_queue, NULL);
         spin_lock_init(&dev->tx_global_lock);
 }
 
+struct netdev_queue *dev_ingress_queue_create(struct net_device *dev)
+{
+        struct netdev_queue *queue = dev_ingress_queue(dev);
+
+#ifdef CONFIG_NET_CLS_ACT
+        if (queue)
+                return queue;
+        queue = kzalloc(sizeof(*queue), GFP_KERNEL);
+        if (!queue)
+                return NULL;
+        netdev_init_one_queue(dev, queue, NULL);
+        __netdev_init_queue_locks_one(dev, queue, NULL);
+        queue->qdisc = &noop_qdisc;
+        queue->qdisc_sleeping = &noop_qdisc;
+        rcu_assign_pointer(dev->ingress_queue, queue);
+#endif
+        return queue;
+}
+
 /**
  *      alloc_netdev_mq - allocate network device
  *      @sizeof_priv:   size of private data to allocate space for
@@ -5559,6 +5575,8 @@ void free_netdev(struct net_device *dev)
 
         kfree(dev->_tx);
 
+        kfree(rcu_dereference_raw(dev->ingress_queue));
+
         /* Flush device addresses */
         dev_addr_flush(dev);