sendfile: Use asynchronized pru_send when ever possible

On Phenom 9550 (4 core, 2.2GHz):
8 parallel netperf -H 127.0.0.1 -t TCP_SENDFILE -P0 (4 runs, unit: Mbps)

old  10509.48  12364.60  11930.55  11104.94
new  21031.34  20165.39  19888.42  19896.47

This give 70% ~ 90% performance improvement

diff --git a/sys/kern/uipc_syscalls.c b/sys/kern/uipc_syscalls.c
index 561233b..9b87fe2 100644 (file)
--- a/sys/kern/uipc_syscalls.c
+++ b/sys/kern/uipc_syscalls.c
@@ -1768,7 +1768,7 @@ retry_space:
                        }
                        goto retry_space;
                }
-               error = so_pru_send(so, 0, m, NULL, NULL, td);
+               error = so_pru_senda(so, 0, m, NULL, NULL, td);
                crit_exit();
                if (error) {
                        ssb_unlock(&so->so_snd);
@@ -1777,7 +1777,7 @@ retry_space:
        }
        if (mheader != NULL) {
                *sbytes += mheader->m_pkthdr.len;
-               error = so_pru_send(so, 0, mheader, NULL, NULL, td);
+               error = so_pru_senda(so, 0, mheader, NULL, NULL, td);
                mheader = NULL;
        }
        ssb_unlock(&so->so_snd);

diff --git a/sys/sys/socketops.h b/sys/sys/socketops.h
index 6159c8d..ea43075 100644 (file)
--- a/sys/sys/socketops.h
+++ b/sys/sys/socketops.h
@@ -106,5 +106,17 @@ int so_pr_ctloutput(struct socket *so, struct sockopt *sopt);
 void so_pru_ctlinput(struct protosw *pr, int cmd,
                struct sockaddr *arg, void *extra);
 
+static __inline int
+so_pru_senda(struct socket *so, int flags, struct mbuf *m,
+            struct sockaddr *addr, struct mbuf *control, struct thread *td)
+{
+       if (so->so_proto->pr_flags & PR_ASYNC_SEND) {
+               so_pru_send_async(so, flags, m, addr, control, td);
+               return 0;
+       } else {
+               return so_pru_send(so, flags, m, addr, control, td);
+       }
+}
+
 #endif /* _KERNEL */
 #endif /* _SYS_SOCKETOPS_H_ */