net-timestamp: ACK timestamp for bytestreams

Add SOF_TIMESTAMPING_TX_ACK, a request for a tstamp when the last byte
in the send() call is acknowledged. It implements the feature for TCP.

The timestamp is generated when the TCP socket cumulative ACK is moved
beyond the tracked seqno for the first time. The feature ignores SACK
and FACK, because those acknowledge the specific byte, but not
necessarily the entire contents of the buffer up to that byte.

Signed-off-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/skbuff.h

@@ -250,9 +250,14 @@ enum {
 
 	/* generate software time stamp when entering packet scheduling */
 	SKBTX_SCHED_TSTAMP = 1 << 6,
+
+	/* generate software timestamp on peer data acknowledgment */
+	SKBTX_ACK_TSTAMP = 1 << 7,
 };
 
-#define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP | SKBTX_SCHED_TSTAMP)
+#define SKBTX_ANY_SW_TSTAMP	(SKBTX_SW_TSTAMP    | \
+				 SKBTX_SCHED_TSTAMP | \
+				 SKBTX_ACK_TSTAMP)
 #define SKBTX_ANY_TSTAMP	(SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
 
 /*

include/uapi/linux/errqueue.h

@@ -40,6 +40,7 @@ struct scm_timestamping {
 enum {
 	SCM_TSTAMP_SND,		/* driver passed skb to NIC, or HW */
 	SCM_TSTAMP_SCHED,	/* data entered the packet scheduler */
+	SCM_TSTAMP_ACK,		/* data acknowledged by peer */
 };
 
 #endif /* _UAPI_LINUX_ERRQUEUE_H */

include/uapi/linux/net_tstamp.h

@@ -22,8 +22,9 @@ enum {
 	SOF_TIMESTAMPING_RAW_HARDWARE = (1<<6),
 	SOF_TIMESTAMPING_OPT_ID = (1<<7),
 	SOF_TIMESTAMPING_TX_SCHED = (1<<8),
+	SOF_TIMESTAMPING_TX_ACK = (1<<9),
 
-	SOF_TIMESTAMPING_LAST = SOF_TIMESTAMPING_TX_SCHED,
+	SOF_TIMESTAMPING_LAST = SOF_TIMESTAMPING_TX_ACK,
 	SOF_TIMESTAMPING_MASK = (SOF_TIMESTAMPING_LAST - 1) |
 				 SOF_TIMESTAMPING_LAST
 };

net/ipv4/tcp_input.c

@@ -74,6 +74,7 @@
 #include <linux/ipsec.h>
 #include <asm/unaligned.h>
 #include <net/netdma.h>
+#include <linux/errqueue.h>
 
 int sysctl_tcp_timestamps __read_mostly = 1;
 int sysctl_tcp_window_scaling __read_mostly = 1;
@@ -3106,6 +3107,11 @@ static int tcp_clean_rtx_queue(struct sock *sk, int prior_fackets,
 			tp->retrans_stamp = 0;
 		}
 
+		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_ACK_TSTAMP) &&
+		    between(skb_shinfo(skb)->tskey, prior_snd_una,
+			    tp->snd_una + 1))
+			__skb_tstamp_tx(skb, NULL, sk, SCM_TSTAMP_ACK);
+
 		if (!fully_acked)
 			break;
 

net/socket.c

@@ -619,6 +619,8 @@ void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 		*tx_flags |= SKBTX_SW_TSTAMP;
 	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
 		*tx_flags |= SKBTX_SCHED_TSTAMP;
+	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
+		*tx_flags |= SKBTX_ACK_TSTAMP;
 
 	if (sock_flag(sk, SOCK_WIFI_STATUS))
 		*tx_flags |= SKBTX_WIFI_STATUS;