summaryrefslogtreecommitdiff
path: root/net/core/sock.c
diff options
context:
space:
mode:
authorEric Dumazet <edumazet@google.com>2016-05-02 10:56:27 -0700
committerDavid S. Miller <davem@davemloft.net>2016-05-03 16:02:36 -0400
commit1d2077ac0165c0d173a2255e37cf4dc5033d92c7 (patch)
treefc18bebb7e5ffe2109fd61c56230c5f47d5f313b /net/core/sock.c
parente34b1638d02bef8c3278af30ee73077c5babc082 (diff)
net: add __sock_wfree() helper
Hosts sending lot of ACK packets exhibit high sock_wfree() cost because of cache line miss to test SOCK_USE_WRITE_QUEUE We could move this flag close to sk_wmem_alloc but it is better to perform the atomic_sub_and_test() on a clean cache line, as it avoid one extra bus transaction. skb_orphan_partial() can also have a fast track for packets that either are TCP acks, or already went through another skb_orphan_partial() Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/core/sock.c')
-rw-r--r--net/core/sock.c24
1 files changed, 24 insertions, 0 deletions
diff --git a/net/core/sock.c b/net/core/sock.c
index f615e9391170..08bf97eceeb3 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -1655,6 +1655,17 @@ void sock_wfree(struct sk_buff *skb)
}
EXPORT_SYMBOL(sock_wfree);
+/* This variant of sock_wfree() is used by TCP,
+ * since it sets SOCK_USE_WRITE_QUEUE.
+ */
+void __sock_wfree(struct sk_buff *skb)
+{
+ struct sock *sk = skb->sk;
+
+ if (atomic_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
+ __sk_free(sk);
+}
+
void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
{
skb_orphan(skb);
@@ -1677,8 +1688,21 @@ void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
}
EXPORT_SYMBOL(skb_set_owner_w);
+/* This helper is used by netem, as it can hold packets in its
+ * delay queue. We want to allow the owner socket to send more
+ * packets, as if they were already TX completed by a typical driver.
+ * But we also want to keep skb->sk set because some packet schedulers
+ * rely on it (sch_fq for example). So we set skb->truesize to a small
+ * amount (1) and decrease sk_wmem_alloc accordingly.
+ */
void skb_orphan_partial(struct sk_buff *skb)
{
+ /* If this skb is a TCP pure ACK or already went here,
+ * we have nothing to do. 2 is already a very small truesize.
+ */
+ if (skb->truesize <= 2)
+ return;
+
/* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
* so we do not completely orphan skb, but transfert all
* accounted bytes but one, to avoid unexpected reorders.