ipv6: Don't depend on per socket memory for neighbour discovery messages

Allocating skbs when sending out neighbour discovery messages
currently uses sock_alloc_send_skb() based on a per net namespace
socket and thus share a socket wmem buffer space.

If a netdevice is temporarily unable to transmit due to carrier
loss or for other reasons, the queued up ndisc messages will cosnume
all of the wmem space and will thus prevent from any more skbs to
be allocated even for netdevices that are able to transmit packets.

The number of neighbour discovery messages sent is very limited,
simply use alloc_skb() and don't depend on any socket wmem space any
longer.

This patch has orginally been posted by Eric Dumazet in a modified
form.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/ipv6/ndisc.c

@@ -370,16 +370,12 @@ static struct sk_buff *ndisc_alloc_skb(struct net_device *dev,
 {
 	int hlen = LL_RESERVED_SPACE(dev);
 	int tlen = dev->needed_tailroom;
-	struct sock *sk = dev_net(dev)->ipv6.ndisc_sk;
 	struct sk_buff *skb;
-	int err;
 
-	skb = sock_alloc_send_skb(sk,
-				  hlen + sizeof(struct ipv6hdr) + len + tlen,
-				  1, &err);
+	skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC);
 	if (!skb) {
-		ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb, err=%d\n",
-			  __func__, err);
+		ND_PRINTK(0, err, "ndisc: %s failed to allocate an skb\n",
+			  __func__);
 		return NULL;
 	}