lttng-instrumentation/lttng-instrumentation-socket

LTTng instrumentation - socket

Instrument socket creation and activity (msg send, receive). Socket "call"
instrumentation should be completed.

Those tracepoints are used by LTTng.

About the performance impact of tracepoints (which is comparable to markers),
even without immediate values optimizations, tests done by Hideo Aoki on ia64
show no regression. His test case was using hackbench on a kernel where
scheduler instrumentation (about 5 events in code scheduler code) was added.
See the "Tracepoints" patch header for performance result detail.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
CC: netdev@vger.kernel.org
CC: David S. Miller <davem@davemloft.net>
CC: Masami Hiramatsu <mhiramat@redhat.com>
CC: 'Peter Zijlstra' <peterz@infradead.org>
CC: "Frank Ch. Eigler" <fche@redhat.com>
CC: 'Ingo Molnar' <mingo@elte.hu>
CC: 'Hideo AOKI' <haoki@redhat.com>
CC: Takashi Nishiie <t-nishiie@np.css.fujitsu.com>
CC: 'Steven Rostedt' <rostedt@goodmis.org>
CC: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>

2 files changed, 37 insertions, 0 deletions

diff --git a/include/trace/socket.h b/include/trace/socket.h
new file mode 100644
index 00000000000..228b5a7c944
--- /dev/null
+++ b/include/trace/socket.h
@@ -0,0 +1,26 @@
+#ifndef _TRACE_SOCKET_H
+#define _TRACE_SOCKET_H
+
+#include <net/sock.h>
+#include <linux/tracepoint.h>
+
+DECLARE_TRACE(socket_sendmsg,
+	TP_PROTO(struct socket *sock, struct msghdr *msg, size_t size, int ret),
+	TP_ARGS(sock, msg, size, ret));
+DECLARE_TRACE(socket_recvmsg,
+	TP_PROTO(struct socket *sock, struct msghdr *msg, size_t size, int flags,
+		int ret),
+	TP_ARGS(sock, msg, size, flags, ret));
+DECLARE_TRACE(socket_create,
+	TP_PROTO(struct socket *sock, int fd),
+	TP_ARGS(sock, fd));
+/*
+ * socket_call
+ *
+ * TODO : This tracepoint should be expanded to cover each element of the
+ * switch in sys_socketcall().
+ */
+DECLARE_TRACE(socket_call,
+	TP_PROTO(int call, unsigned long a0),
+	TP_ARGS(call, a0));
+#endif
diff --git a/net/socket.c b/net/socket.c
index ac2219f90d5..319a3b8f241 100644
--- a/net/socket.c
+++ b/net/socket.c
@@ -98,6 +98,7 @@
 
 #include <net/sock.h>
 #include <linux/netfilter.h>
+#include <trace/socket.h>
 
 #include <linux/if_tun.h>
 #include <linux/ipv6_route.h>
@@ -164,6 +165,11 @@ static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 
 static DEFINE_PER_CPU(int, sockets_in_use);
 
+DEFINE_TRACE(socket_sendmsg);
+DEFINE_TRACE(socket_recvmsg);
+DEFINE_TRACE(socket_create);
+DEFINE_TRACE(socket_call);
+
 /*
  * Support routines.
  * Move socket addresses back and forth across the kernel/user
@@ -578,6 +584,7 @@ int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 	ret = __sock_sendmsg(&iocb, sock, msg, size);
 	if (-EIOCBQUEUED == ret)
 		ret = wait_on_sync_kiocb(&iocb);
+	trace_socket_sendmsg(sock, msg, size, ret);
 	return ret;
 }
 EXPORT_SYMBOL(sock_sendmsg);
@@ -714,6 +721,7 @@ int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 	if (-EIOCBQUEUED == ret)
 		ret = wait_on_sync_kiocb(&iocb);
+	trace_socket_recvmsg(sock, msg, size, flags, ret);
 	return ret;
 }
 EXPORT_SYMBOL(sock_recvmsg);
@@ -1314,6 +1322,7 @@ SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
 	if (retval < 0)
 		goto out_release;
 
+	trace_socket_create(sock, retval);
 out:
 	/* It may be already another descriptor 8) Not kernel problem. */
 	return retval;
@@ -2249,6 +2258,8 @@ SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
 	a0 = a[0];
 	a1 = a[1];
 
+	trace_socket_call(call, a0);
+
 	switch (call) {
 	case SYS_SOCKET:
 		err = sys_socket(a0, a1, a[2]);