[PATCH v2 net-next 0/2] Add software rx timestamp for TCP.

[PATCH v2 net-next 0/2] Add software rx timestamp for TCP.

[Mike Maloney]

From: Mike Maloney <maloney@google.com>

Add software rx timestamps for TCP, and a test to ensure consistency of
behavior between IP, UDP, and TCP implementation.

Changes since v1:
  -Initialize tss->ts[1] to 0 if caller requested any timestamps.
  -Fix test case to validate that tss->ts[1] is zero.
  -Fix tests to actually use a raw socket.
  -Fix --tcp flag to work on the test.

Mike Maloney (2):
  tcp: Extend SOF_TIMESTAMPING_RX_SOFTWARE to TCP recvmsg
  selftests/net: Add a test to validate behavior of rx timestamps

 include/net/tcp.h                                  |   9 +-
 net/ipv4/tcp.c                                     |  65 ++++
 net/ipv4/tcp_input.c                               |  35 +-
 net/ipv4/tcp_ipv4.c                                |   2 +
 net/ipv6/tcp_ipv6.c                                |   2 +
 .../selftests/networking/timestamping/.gitignore   |   1 +
 .../selftests/networking/timestamping/Makefile     |   4 +-
 .../networking/timestamping/rxtimestamp.c          | 389 +++++++++++++++++++++
 8 files changed, 501 insertions(+), 6 deletions(-)
 create mode 100644 tools/testing/selftests/networking/timestamping/rxtimestamp.c

-- 
2.14.1.480.gb18f417b89-goog

[PATCH v2 net-next 1/2] tcp: Extend SOF_TIMESTAMPING_RX_SOFTWARE to TCP recvmsg

[Mike Maloney]

From: Mike Maloney <maloney@google.com>

When SOF_TIMESTAMPING_RX_SOFTWARE is enabled for tcp sockets, return the
timestamp corresponding to the highest sequence number data returned.

Previously the skb->tstamp is overwritten when a TCP packet is placed
in the out of order queue.  While the packet is in the ooo queue, save the
timestamp in the TCB_SKB_CB.  This space is shared with the gso_*
options which are only used on the tx path, and a previously unused 4
byte hole.

When skbs are coalesced either in the sk_receive_queue or the
out_of_order_queue always choose the timestamp of the appended skb to
maintain the invariant of returning the timestamp of the last byte in
the recvmsg buffer.

Signed-off-by: Mike Maloney <maloney@google.com>
---
 include/net/tcp.h    |  9 +++++++-
 net/ipv4/tcp.c       | 65 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_input.c | 35 ++++++++++++++++++++++++----
 net/ipv4/tcp_ipv4.c  |  2 ++
 net/ipv6/tcp_ipv6.c  |  2 ++
 5 files changed, 108 insertions(+), 5 deletions(-)

diff --git a/include/net/tcp.h b/include/net/tcp.h
index afdab3781425..f26d20e9760d 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -774,6 +774,12 @@ struct tcp_skb_cb {
 			u16	tcp_gso_segs;
 			u16	tcp_gso_size;
 		};
+
+		/* Used to stash the receive timestamp while this skb is in the
+		 * out of order queue, as skb->tstamp is overwritten by the
+		 * rbnode.
+		 */
+		ktime_t		swtstamp;
 	};
 	__u8		tcp_flags;	/* TCP header flags. (tcp[13])	*/
 
@@ -790,7 +796,8 @@ struct tcp_skb_cb {
 	__u8		ip_dsfield;	/* IPv4 tos or IPv6 dsfield	*/
 	__u8		txstamp_ack:1,	/* Record TX timestamp for ack? */
 			eor:1,		/* Is skb MSG_EOR marked? */
-			unused:6;
+			has_rxtstamp:1,	/* SKB has a RX timestamp	*/
+			unused:5;
 	__u32		ack_seq;	/* Sequence number ACK'd	*/
 	union {
 		struct {
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d25e3bcca66b..0cce4472b4a1 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -269,6 +269,7 @@
 #include <linux/err.h>
 #include <linux/time.h>
 #include <linux/slab.h>
+#include <linux/errqueue.h>
 
 #include <net/icmp.h>
 #include <net/inet_common.h>
@@ -1695,6 +1696,61 @@ int tcp_peek_len(struct socket *sock)
 }
 EXPORT_SYMBOL(tcp_peek_len);
 
+static void tcp_update_recv_tstamps(struct sk_buff *skb,
+				    struct scm_timestamping *tss)
+{
+	if (skb->tstamp)
+		tss->ts[0] = ktime_to_timespec(skb->tstamp);
+	else
+		tss->ts[0] = (struct timespec) {0};
+
+	if (skb_hwtstamps(skb)->hwtstamp)
+		tss->ts[2] = ktime_to_timespec(skb_hwtstamps(skb)->hwtstamp);
+	else
+		tss->ts[2] = (struct timespec) {0};
+}
+
+/* Similar to __sock_recv_timestamp, but does not require an skb */
+void tcp_recv_timestamp(struct msghdr *msg, const struct sock *sk,
+			struct scm_timestamping *tss)
+{
+	struct timeval tv;
+	bool has_timestamping = false;
+
+	if (tss->ts[0].tv_sec || tss->ts[0].tv_nsec) {
+		if (sock_flag(sk, SOCK_RCVTSTAMP)) {
+			if (sock_flag(sk, SOCK_RCVTSTAMPNS)) {
+				put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
+					 sizeof(tss->ts[0]), &tss->ts[0]);
+			} else {
+				tv.tv_sec = tss->ts[0].tv_sec;
+				tv.tv_usec = tss->ts[0].tv_nsec / 1000;
+
+				put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
+					 sizeof(tv), &tv);
+			}
+		}
+
+		if (sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE)
+			has_timestamping = true;
+		else
+			tss->ts[0] = (struct timespec) {0};
+	}
+
+	if (tss->ts[2].tv_sec || tss->ts[2].tv_nsec) {
+		if (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)
+			has_timestamping = true;
+		else
+			tss->ts[2] = (struct timespec) {0};
+	}
+
+	if (has_timestamping) {
+		tss->ts[1] = (struct timespec) {0};
+		put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING,
+			 sizeof(*tss), tss);
+	}
+}
+
 /*
  *	This routine copies from a sock struct into the user buffer.
  *
@@ -1716,6 +1772,8 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 	long timeo;
 	struct sk_buff *skb, *last;
 	u32 urg_hole = 0;
+	struct scm_timestamping tss;
+	bool has_tss = false;
 
 	if (unlikely(flags & MSG_ERRQUEUE))
 		return inet_recv_error(sk, msg, len, addr_len);
@@ -1911,6 +1969,10 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 		if (used + offset < skb->len)
 			continue;
 
+		if (TCP_SKB_CB(skb)->has_rxtstamp) {
+			tcp_update_recv_tstamps(skb, &tss);
+			has_tss = true;
+		}
 		if (TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN)
 			goto found_fin_ok;
 		if (!(flags & MSG_PEEK))
@@ -1929,6 +1991,9 @@ int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
 	 * on connected socket. I was just happy when found this 8) --ANK
 	 */
 
+	if (has_tss)
+		tcp_recv_timestamp(msg, sk, &tss);
+
 	/* Clean up data we have read: This will do ACK frames. */
 	tcp_cleanup_rbuf(sk, copied);
 
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index ddc854728a60..66abcbf6f381 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -4246,9 +4246,15 @@ static void tcp_sack_remove(struct tcp_sock *tp)
 	tp->rx_opt.num_sacks = num_sacks;
 }
 
+enum tcp_queue {
+	OOO_QUEUE,
+	RCV_QUEUE,
+};
+
 /**
  * tcp_try_coalesce - try to merge skb to prior one
  * @sk: socket
+ * @dest: destination queue
  * @to: prior buffer
  * @from: buffer to add in queue
  * @fragstolen: pointer to boolean
@@ -4260,6 +4266,7 @@ static void tcp_sack_remove(struct tcp_sock *tp)
  * Returns true if caller should free @from instead of queueing it
  */
 static bool tcp_try_coalesce(struct sock *sk,
+			     enum tcp_queue dest,
 			     struct sk_buff *to,
 			     struct sk_buff *from,
 			     bool *fragstolen)
@@ -4281,6 +4288,15 @@ static bool tcp_try_coalesce(struct sock *sk,
 	TCP_SKB_CB(to)->end_seq = TCP_SKB_CB(from)->end_seq;
 	TCP_SKB_CB(to)->ack_seq = TCP_SKB_CB(from)->ack_seq;
 	TCP_SKB_CB(to)->tcp_flags |= TCP_SKB_CB(from)->tcp_flags;
+
+	if (TCP_SKB_CB(from)->has_rxtstamp) {
+		TCP_SKB_CB(to)->has_rxtstamp = true;
+		if (dest == OOO_QUEUE)
+			TCP_SKB_CB(to)->swtstamp = TCP_SKB_CB(from)->swtstamp;
+		else
+			to->tstamp = from->tstamp;
+	}
+
 	return true;
 }
 
@@ -4315,6 +4331,9 @@ static void tcp_ofo_queue(struct sock *sk)
 		}
 		p = rb_next(p);
 		rb_erase(&skb->rbnode, &tp->out_of_order_queue);
+		/* Replace tstamp which was stomped by rbnode */
+		if (TCP_SKB_CB(skb)->has_rxtstamp)
+			skb->tstamp = TCP_SKB_CB(skb)->swtstamp;
 
 		if (unlikely(!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt))) {
 			SOCK_DEBUG(sk, "ofo packet was already received\n");
@@ -4326,7 +4345,8 @@ static void tcp_ofo_queue(struct sock *sk)
 			   TCP_SKB_CB(skb)->end_seq);
 
 		tail = skb_peek_tail(&sk->sk_receive_queue);
-		eaten = tail && tcp_try_coalesce(sk, tail, skb, &fragstolen);
+		eaten = tail && tcp_try_coalesce(sk, RCV_QUEUE,
+						 tail, skb, &fragstolen);
 		tcp_rcv_nxt_update(tp, TCP_SKB_CB(skb)->end_seq);
 		fin = TCP_SKB_CB(skb)->tcp_flags & TCPHDR_FIN;
 		if (!eaten)
@@ -4380,6 +4400,10 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 		return;
 	}
 
+	/* Stash tstamp to avoid being stomped on by rbnode */
+	if (TCP_SKB_CB(skb)->has_rxtstamp)
+		TCP_SKB_CB(skb)->swtstamp = skb->tstamp;
+
 	inet_csk_schedule_ack(sk);
 
 	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPOFOQUEUE);
@@ -4405,7 +4429,8 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 	/* In the typical case, we are adding an skb to the end of the list.
 	 * Use of ooo_last_skb avoids the O(Log(N)) rbtree lookup.
 	 */
-	if (tcp_try_coalesce(sk, tp->ooo_last_skb, skb, &fragstolen)) {
+	if (tcp_try_coalesce(sk, OOO_QUEUE, tp->ooo_last_skb,
+			     skb, &fragstolen)) {
 coalesce_done:
 		tcp_grow_window(sk, skb);
 		kfree_skb_partial(skb, fragstolen);
@@ -4455,7 +4480,8 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 				__kfree_skb(skb1);
 				goto merge_right;
 			}
-		} else if (tcp_try_coalesce(sk, skb1, skb, &fragstolen)) {
+		} else if (tcp_try_coalesce(sk, OOO_QUEUE, skb1,
+					    skb, &fragstolen)) {
 			goto coalesce_done;
 		}
 		p = &parent->rb_right;
@@ -4506,7 +4532,8 @@ static int __must_check tcp_queue_rcv(struct sock *sk, struct sk_buff *skb, int
 
 	__skb_pull(skb, hdrlen);
 	eaten = (tail &&
-		 tcp_try_coalesce(sk, tail, skb, fragstolen)) ? 1 : 0;
+		 tcp_try_coalesce(sk, RCV_QUEUE, tail,
+				  skb, fragstolen)) ? 1 : 0;
 	tcp_rcv_nxt_update(tcp_sk(sk), TCP_SKB_CB(skb)->end_seq);
 	if (!eaten) {
 		__skb_queue_tail(&sk->sk_receive_queue, skb);
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 5af8b809dfbc..a63486afa7a7 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1637,6 +1637,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
 	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
 	TCP_SKB_CB(skb)->ip_dsfield = ipv4_get_dsfield(iph);
 	TCP_SKB_CB(skb)->sacked	 = 0;
+	TCP_SKB_CB(skb)->has_rxtstamp =
+			skb->tstamp || skb_hwtstamps(skb)->hwtstamp;
 
 lookup:
 	sk = __inet_lookup_skb(&tcp_hashinfo, skb, __tcp_hdrlen(th), th->source,
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index d79a1af3252e..abba3bc2a3d9 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -1394,6 +1394,8 @@ static void tcp_v6_fill_cb(struct sk_buff *skb, const struct ipv6hdr *hdr,
 	TCP_SKB_CB(skb)->tcp_tw_isn = 0;
 	TCP_SKB_CB(skb)->ip_dsfield = ipv6_get_dsfield(hdr);
 	TCP_SKB_CB(skb)->sacked = 0;
+	TCP_SKB_CB(skb)->has_rxtstamp =
+			skb->tstamp || skb_hwtstamps(skb)->hwtstamp;
 }
 
 static int tcp_v6_rcv(struct sk_buff *skb)
-- 
2.14.1.480.gb18f417b89-goog

[PATCH v2 net-next 2/2] selftests/net: Add a test to validate behavior of rx timestamps

[Mike Maloney]

From: Mike Maloney <maloney@google.com>

Validate the behavior of the combination of various timestamp socket
options, and ensure consistency across ip, udp, and tcp.

Signed-off-by: Mike Maloney <maloney@google.com>
---
 .../selftests/networking/timestamping/.gitignore   |   1 +
 .../selftests/networking/timestamping/Makefile     |   4 +-
 .../networking/timestamping/rxtimestamp.c          | 389 +++++++++++++++++++++
 3 files changed, 393 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/networking/timestamping/rxtimestamp.c

diff --git a/tools/testing/selftests/networking/timestamping/.gitignore b/tools/testing/selftests/networking/timestamping/.gitignore
index 9e69e982fb38..d9355035e746 100644
--- a/tools/testing/selftests/networking/timestamping/.gitignore
+++ b/tools/testing/selftests/networking/timestamping/.gitignore
@@ -1,3 +1,4 @@
 timestamping
+rxtimestamp
 txtimestamp
 hwtstamp_config
diff --git a/tools/testing/selftests/networking/timestamping/Makefile b/tools/testing/selftests/networking/timestamping/Makefile
index ccbb9edbbbb9..92fb8ee917c5 100644
--- a/tools/testing/selftests/networking/timestamping/Makefile
+++ b/tools/testing/selftests/networking/timestamping/Makefile
@@ -1,4 +1,6 @@
-TEST_PROGS := hwtstamp_config timestamping txtimestamp
+CFLAGS += -I../../../../../usr/include
+
+TEST_PROGS := hwtstamp_config rxtimestamp timestamping txtimestamp
 
 all: $(TEST_PROGS)
 
diff --git a/tools/testing/selftests/networking/timestamping/rxtimestamp.c b/tools/testing/selftests/networking/timestamping/rxtimestamp.c
new file mode 100644
index 000000000000..00f286661dcd
--- /dev/null
+++ b/tools/testing/selftests/networking/timestamping/rxtimestamp.c
@@ -0,0 +1,389 @@
+#include <errno.h>
+#include <error.h>
+#include <getopt.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include <sys/time.h>
+#include <sys/socket.h>
+#include <sys/select.h>
+#include <sys/ioctl.h>
+#include <arpa/inet.h>
+#include <net/if.h>
+
+#include <asm/types.h>
+#include <linux/net_tstamp.h>
+#include <linux/errqueue.h>
+
+#define ARRAY_SIZE(arr) (sizeof(arr) / sizeof((arr)[0]))
+
+struct options {
+	int so_timestamp;
+	int so_timestampns;
+	int so_timestamping;
+};
+
+struct tstamps {
+	bool tstamp;
+	bool tstampns;
+	bool swtstamp;
+	bool hwtstamp;
+};
+
+struct socket_type {
+	char *friendly_name;
+	int type;
+	int protocol;
+	bool enabled;
+};
+
+struct test_case {
+	struct options sockopt;
+	struct tstamps expected;
+	bool enabled;
+};
+
+struct sof_flag {
+	int mask;
+	char *name;
+};
+
+static struct sof_flag sof_flags[] = {
+#define SOF_FLAG(f) { f, #f }
+	SOF_FLAG(SOF_TIMESTAMPING_SOFTWARE),
+	SOF_FLAG(SOF_TIMESTAMPING_RX_SOFTWARE),
+	SOF_FLAG(SOF_TIMESTAMPING_RX_HARDWARE),
+};
+
+static struct socket_type socket_types[] = {
+	{ "ip",		SOCK_RAW,	IPPROTO_EGP },
+	{ "udp",	SOCK_DGRAM,	IPPROTO_UDP },
+	{ "tcp",	SOCK_STREAM,	IPPROTO_TCP },
+};
+
+static struct test_case test_cases[] = {
+	{ {}, {} },
+	{
+		{ so_timestamp: 1 },
+		{ tstamp: true }
+	},
+	{
+		{ so_timestampns: 1 },
+		{ tstampns: true }
+	},
+	{
+		{ so_timestamp: 1, so_timestampns: 1 },
+		{ tstampns: true }
+	},
+	{
+		{ so_timestamping: SOF_TIMESTAMPING_RX_SOFTWARE },
+		{}
+	},
+	{
+		/* Loopback device does not support hw timestamps. */
+		{ so_timestamping: SOF_TIMESTAMPING_RX_HARDWARE },
+		{}
+	},
+	{
+		{ so_timestamping: SOF_TIMESTAMPING_SOFTWARE },
+		{}
+	},
+	{
+		{ so_timestamping: SOF_TIMESTAMPING_RX_SOFTWARE
+			| SOF_TIMESTAMPING_RX_HARDWARE },
+		{}
+	},
+	{
+		{ so_timestamping: SOF_TIMESTAMPING_SOFTWARE
+			| SOF_TIMESTAMPING_RX_SOFTWARE },
+		{ swtstamp: true }
+	},
+	{
+		{ so_timestamp: 1, so_timestamping: SOF_TIMESTAMPING_SOFTWARE
+			| SOF_TIMESTAMPING_RX_SOFTWARE },
+		{ tstamp: true, swtstamp: true }
+	},
+};
+
+static struct option long_options[] = {
+	{ "list_tests", no_argument, 0, 'l' },
+	{ "test_num", required_argument, 0, 'n' },
+	{ "op_size", required_argument, 0, 's' },
+	{ "tcp", no_argument, 0, 't' },
+	{ "udp", no_argument, 0, 'u' },
+	{ "ip", no_argument, 0, 'i' },
+};
+
+static int next_port = 19999;
+static int op_size = 10 * 1024;
+
+void print_test_case(struct test_case *t)
+{
+	int f = 0;
+
+	printf("sockopts {");
+	if (t->sockopt.so_timestamp)
+		printf(" SO_TIMESTAMP ");
+	if (t->sockopt.so_timestampns)
+		printf(" SO_TIMESTAMPNS ");
+	if (t->sockopt.so_timestamping) {
+		printf(" SO_TIMESTAMPING: {");
+		for (f = 0; f < ARRAY_SIZE(sof_flags); f++)
+			if (t->sockopt.so_timestamping & sof_flags[f].mask)
+				printf(" %s |", sof_flags[f].name);
+		printf("}");
+	}
+	printf("} expected cmsgs: {");
+	if (t->expected.tstamp)
+		printf(" SCM_TIMESTAMP ");
+	if (t->expected.tstampns)
+		printf(" SCM_TIMESTAMPNS ");
+	if (t->expected.swtstamp || t->expected.hwtstamp) {
+		printf(" SCM_TIMESTAMPING {");
+		if (t->expected.swtstamp)
+			printf("0");
+		if (t->expected.swtstamp && t->expected.hwtstamp)
+			printf(",");
+		if (t->expected.hwtstamp)
+			printf("2");
+		printf("}");
+	}
+	printf("}\n");
+}
+
+void do_send(int src)
+{
+	int r;
+	char *buf = malloc(op_size);
+
+	memset(buf, 'z', op_size);
+	r = write(src, buf, op_size);
+	if (r < 0)
+		error(1, errno, "Failed to sendmsg");
+
+	free(buf);
+}
+
+bool do_recv(int rcv, int read_size, struct tstamps expected)
+{
+	const int CMSG_SIZE = 1024;
+
+	struct scm_timestamping *ts;
+	struct tstamps actual = {};
+	char cmsg_buf[CMSG_SIZE];
+	struct iovec recv_iov;
+	struct cmsghdr *cmsg;
+	bool failed = false;
+	struct msghdr hdr;
+	int flags = 0;
+	int r;
+
+	memset(&hdr, 0, sizeof(hdr));
+	hdr.msg_iov = &recv_iov;
+	hdr.msg_iovlen = 1;
+	recv_iov.iov_base = malloc(read_size);
+	recv_iov.iov_len = read_size;
+
+	hdr.msg_control = cmsg_buf;
+	hdr.msg_controllen = sizeof(cmsg_buf);
+
+	r = recvmsg(rcv, &hdr, flags);
+	if (r < 0)
+		error(1, errno, "Failed to recvmsg");
+	if (r != read_size)
+		error(1, 0, "Only received %d bytes of payload.", r);
+
+	if (hdr.msg_flags & (MSG_TRUNC | MSG_CTRUNC))
+		error(1, 0, "Message was truncated.");
+
+	for (cmsg = CMSG_FIRSTHDR(&hdr); cmsg != NULL;
+	     cmsg = CMSG_NXTHDR(&hdr, cmsg)) {
+		if (cmsg->cmsg_level != SOL_SOCKET)
+			error(1, 0, "Unexpected cmsg_level %d",
+			      cmsg->cmsg_level);
+		switch (cmsg->cmsg_type) {
+		case SCM_TIMESTAMP:
+			actual.tstamp = true;
+			break;
+		case SCM_TIMESTAMPNS:
+			actual.tstampns = true;
+			break;
+		case SCM_TIMESTAMPING:
+			ts = (struct scm_timestamping *)CMSG_DATA(cmsg);
+			actual.swtstamp = !!ts->ts[0].tv_sec;
+			if (ts->ts[1].tv_sec != 0)
+				error(0, 0, "ts[1] should not be set.");
+			actual.hwtstamp = !!ts->ts[2].tv_sec;
+			break;
+		default:
+			error(1, 0, "Unexpected cmsg_type %d", cmsg->cmsg_type);
+		}
+	}
+
+#define VALIDATE(field) \
+	do { \
+		if (expected.field != actual.field) { \
+			if (expected.field) \
+				error(0, 0, "Expected " #field " to be set."); \
+			else \
+				error(0, 0, \
+				      "Expected " #field " to not be set."); \
+			failed = true; \
+		} \
+	} while (0)
+
+	VALIDATE(tstamp);
+	VALIDATE(tstampns);
+	VALIDATE(swtstamp);
+	VALIDATE(hwtstamp);
+#undef VALIDATE
+
+	free(recv_iov.iov_base);
+
+	return failed;
+}
+
+void config_so_flags(int rcv, struct options o)
+{
+	int on = 1;
+
+	if (setsockopt(rcv, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on)) < 0)
+		error(1, errno, "Failed to enable SO_REUSEADDR");
+
+	if (o.so_timestamp &&
+	    setsockopt(rcv, SOL_SOCKET, SO_TIMESTAMP,
+		       &o.so_timestamp, sizeof(o.so_timestamp)) < 0)
+		error(1, errno, "Failed to enable SO_TIMESTAMP");
+
+	if (o.so_timestampns &&
+	    setsockopt(rcv, SOL_SOCKET, SO_TIMESTAMPNS,
+		       &o.so_timestampns, sizeof(o.so_timestampns)) < 0)
+		error(1, errno, "Failed to enable SO_TIMESTAMPNS");
+
+	if (o.so_timestamping &&
+	    setsockopt(rcv, SOL_SOCKET, SO_TIMESTAMPING,
+		       &o.so_timestamping, sizeof(o.so_timestamping)) < 0)
+		error(1, errno, "Failed to set SO_TIMESTAMPING");
+}
+
+bool run_test_case(struct socket_type s, struct test_case t)
+{
+	int port = (s.type == SOCK_RAW) ? 0 : next_port++;
+	int read_size = op_size;
+	struct sockaddr_in addr;
+	bool failed = false;
+	int src, dst, rcv;
+
+	src = socket(AF_INET, s.type, s.protocol);
+	if (src < 0)
+		error(1, errno, "Failed to open src socket");
+
+	dst = socket(AF_INET, s.type, s.protocol);
+	if (dst < 0)
+		error(1, errno, "Failed to open dst socket");
+
+	memset(&addr, 0, sizeof(addr));
+	addr.sin_family = AF_INET;
+	addr.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
+	addr.sin_port = htons(port);
+
+	if (bind(dst, (struct sockaddr *)&addr, sizeof(addr)) < 0)
+		error(1, errno, "Failed to bind to port %d", port);
+
+	if (s.type == SOCK_STREAM && (listen(dst, 1) < 0))
+		error(1, errno, "Failed to listen");
+
+	if (connect(src, (struct sockaddr *)&addr, sizeof(addr)) < 0)
+		error(1, errno, "Failed to connect");
+
+	if (s.type == SOCK_STREAM) {
+		rcv = accept(dst, NULL, NULL);
+		if (rcv < 0)
+			error(1, errno, "Failed to accept");
+		close(dst);
+	} else {
+		rcv = dst;
+	}
+
+	config_so_flags(rcv, t.sockopt);
+	usleep(20000); /* setsockopt for SO_TIMESTAMPING is asynchronous */
+	do_send(src);
+
+	if (s.type == SOCK_RAW)
+		read_size += 20;  /* for IP header */
+	failed = do_recv(rcv, read_size, t.expected);
+
+	close(rcv);
+	close(src);
+
+	return failed;
+}
+
+int main(int argc, char **argv)
+{
+	bool all_protocols = true;
+	bool all_tests = true;
+	int arg_index = 0;
+	int failures = 0;
+	int s, t;
+	char opt;
+
+	while ((opt = getopt_long(argc, argv, "", long_options,
+				  &arg_index)) != -1) {
+		switch (opt) {
+		case 'l':
+			for (t = 0; t < ARRAY_SIZE(test_cases); t++) {
+				printf("%d\t", t);
+				print_test_case(&test_cases[t]);
+			}
+			return 0;
+		case 'n':
+			t = atoi(optarg);
+			if (t > ARRAY_SIZE(test_cases))
+				error(1, 0, "Invalid test case: %d", t);
+			all_tests = false;
+			test_cases[t].enabled = true;
+			break;
+		case 's':
+			op_size = atoi(optarg);
+			break;
+		case 't':
+			all_protocols = false;
+			socket_types[2].enabled = true;
+			break;
+		case 'u':
+			all_protocols = false;
+			socket_types[1].enabled = true;
+			break;
+		case 'i':
+			all_protocols = false;
+			socket_types[0].enabled = true;
+			break;
+		default:
+			error(1, 0, "Failed to parse parameters.");
+		}
+	}
+
+	for (s = 0; s < ARRAY_SIZE(socket_types); s++) {
+		if (!all_protocols && !socket_types[s].enabled)
+			continue;
+
+		printf("Testing %s...\n", socket_types[s].friendly_name);
+		for (t = 0; t < ARRAY_SIZE(test_cases); t++) {
+			if (!all_tests && !test_cases[t].enabled)
+				continue;
+
+			printf("Starting testcase %d...\n", t);
+			if (run_test_case(socket_types[s], test_cases[t])) {
+				failures++;
+				printf("FAILURE in test case ");
+				print_test_case(&test_cases[t]);
+			}
+		}
+	}
+	if (!failures)
+		printf("PASSED.\n");
+	return failures;
+}
-- 
2.14.1.480.gb18f417b89-goog

Re: [PATCH v2 net-next 1/2] tcp: Extend SOF_TIMESTAMPING_RX_SOFTWARE to TCP recvmsg

[Willem de Bruijn]

On Tue, Aug 22, 2017 at 5:08 PM, Mike Maloney <maloneykernel@gmail.com> wrote:
> From: Mike Maloney <maloney@google.com>
>
> When SOF_TIMESTAMPING_RX_SOFTWARE is enabled for tcp sockets, return the
> timestamp corresponding to the highest sequence number data returned.
>
> Previously the skb->tstamp is overwritten when a TCP packet is placed
> in the out of order queue.  While the packet is in the ooo queue, save the
> timestamp in the TCB_SKB_CB.  This space is shared with the gso_*
> options which are only used on the tx path, and a previously unused 4
> byte hole.
>
> When skbs are coalesced either in the sk_receive_queue or the
> out_of_order_queue always choose the timestamp of the appended skb to
> maintain the invariant of returning the timestamp of the last byte in
> the recvmsg buffer.
>
> Signed-off-by: Mike Maloney <maloney@google.com>

Acked-by: Willem de Bruijn <willemb@google.com>

Re: [PATCH v2 net-next 2/2] selftests/net: Add a test to validate behavior of rx timestamps

[Willem de Bruijn]

On Tue, Aug 22, 2017 at 5:08 PM, Mike Maloney <maloneykernel@gmail.com> wrote:
> From: Mike Maloney <maloney@google.com>
>
> Validate the behavior of the combination of various timestamp socket
> options, and ensure consistency across ip, udp, and tcp.
>
> Signed-off-by: Mike Maloney <maloney@google.com>

Acked-by: Willem de Bruijn <willemb@google.com>

Thanks for adding these tests.

For tcp, it may be useful to eventually also add packetdrill tests
to validate the more subtle semantics, such as that the timestamp is
returned only when reading the last byte of the skb.

> +       do_send(src);
> +
> +       if (s.type == SOCK_RAW)
> +               read_size += 20;  /* for IP header */

No need to revise just for this, but if submitting a v3:

please use sizeof(struct iphdr). And maybe test PF_INET6, too.

Re: [PATCH v2 net-next 0/2] Add software rx timestamp for TCP.

[David Miller]

From: Mike Maloney <maloneykernel@gmail.com>
Date: Tue, 22 Aug 2017 17:08:47 -0400

> From: Mike Maloney <maloney@google.com>
> 
> Add software rx timestamps for TCP, and a test to ensure consistency of
> behavior between IP, UDP, and TCP implementation.
> 
> Changes since v1:
>   -Initialize tss->ts[1] to 0 if caller requested any timestamps.
>   -Fix test case to validate that tss->ts[1] is zero.
>   -Fix tests to actually use a raw socket.
>   -Fix --tcp flag to work on the test.

Series applied, thanks.