inet: Decrease overhead of on-stack inet_cork.

When we fast path datagram sends to avoid locking by putting
the inet_cork on the stack we use up lots of space that isn't
necessary.

This is because inet_cork contains a "struct flowi" which isn't
used in these code paths.

Split inet_cork to two parts, "inet_cork" and "inet_cork_full".
Only the latter of which has the "struct flowi" and is what is
stored in inet_sock.

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

1 files changed, 12 insertions, 10 deletions

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index db38c1822de8..eb0647a2f073 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -1096,14 +1096,14 @@ int ip_append_data(struct sock *sk,
                 return 0;
 
         if (skb_queue_empty(&sk->sk_write_queue)) {
-                err = ip_setup_cork(sk, &inet->cork, ipc, rtp);
+                err = ip_setup_cork(sk, &inet->cork.base, ipc, rtp);
                 if (err)
                         return err;
         } else {
                 transhdrlen = 0;
         }
 
-        return __ip_append_data(sk, &sk->sk_write_queue, &inet->cork, getfrag,
+        return __ip_append_data(sk, &sk->sk_write_queue, &inet->cork.base, getfrag,
                                 from, length, transhdrlen, flags);
 }
 
@@ -1114,6 +1114,7 @@ ssize_t	ip_append_page(struct sock *sk, struct page *page,
         struct sk_buff *skb;
         struct rtable *rt;
         struct ip_options *opt = NULL;
+        struct inet_cork *cork;
         int hh_len;
         int mtu;
         int len;
@@ -1129,20 +1130,21 @@ ssize_t	ip_append_page(struct sock *sk, struct page *page,
         if (skb_queue_empty(&sk->sk_write_queue))
                 return -EINVAL;
 
-        rt = (struct rtable *)inet->cork.dst;
-        if (inet->cork.flags & IPCORK_OPT)
-                opt = inet->cork.opt;
+        cork = &inet->cork.base;
+        rt = (struct rtable *)cork->dst;
+        if (cork->flags & IPCORK_OPT)
+                opt = cork->opt;
 
         if (!(rt->dst.dev->features&NETIF_F_SG))
                 return -EOPNOTSUPP;
 
         hh_len = LL_RESERVED_SPACE(rt->dst.dev);
-        mtu = inet->cork.fragsize;
+        mtu = cork->fragsize;
 
         fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
         maxfraglen = ((mtu - fragheaderlen) & ~7) + fragheaderlen;
 
-        if (inet->cork.length + size > 0xFFFF - fragheaderlen) {
+        if (cork->length + size > 0xFFFF - fragheaderlen) {
                 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->inet_dport, mtu);
                 return -EMSGSIZE;
         }
@@ -1150,7 +1152,7 @@ ssize_t	ip_append_page(struct sock *sk, struct page *page,
         if ((skb = skb_peek_tail(&sk->sk_write_queue)) == NULL)
                 return -EINVAL;
 
-        inet->cork.length += size;
+        cork->length += size;
         if ((size + skb->len > mtu) &&
             (sk->sk_protocol == IPPROTO_UDP) &&
             (rt->dst.dev->features & NETIF_F_UFO)) {
@@ -1245,7 +1247,7 @@ ssize_t	ip_append_page(struct sock *sk, struct page *page,
         return 0;
 
 error:
-        inet->cork.length -= size;
+        cork->length -= size;
         IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
         return err;
 }
@@ -1396,7 +1398,7 @@ static void __ip_flush_pending_frames(struct sock *sk,
 
 void ip_flush_pending_frames(struct sock *sk)
 {
-        __ip_flush_pending_frames(sk, &sk->sk_write_queue, &inet_sk(sk)->cork);
+        __ip_flush_pending_frames(sk, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
 }
 
 struct sk_buff *ip_make_skb(struct sock *sk,