selftests/net: packetdrill: import tcp/fast_recovery, tcp/nagle, tcp/timestamping

Use the standard import and testing method, as described in the
import of tcp/ecn , tcp/close , tcp/sack , tcp/tcp_info.

Signed-off-by: Willem de Bruijn <[e-mail]>
Signed-off-by: Soham Chakradeo <[e-mail]>
Link: https://patch.msgid.link/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>

tools/testing/selftests/net/packetdrill/tcp_fast_recovery_prr-ss-10pkt-lost-1.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test PRR-slowstart implementation.
+// In this variant we test a simple case where in-flight == ssthresh
+// all the way through recovery, so during fast recovery we send one segment
+// for each segment SACKed/ACKed.
+
+// Set up config.
+`./defaults.sh`
+
+// Establish a connection.
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+  +.1 < S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+// RTT 100ms
+  +.1 < . 1:1(0) ack 1 win 320
+   +0 accept(3, ..., ...) = 4
+
+// Send 10 data segments.
+   +0 write(4, ..., 10000) = 10000
+   +0 > P. 1:10001(10000) ack 1
+
+// Lost packet 1:1001.
+ +.11 < . 1:1(0) ack 1 win 320 <sack 1001:2001,nop,nop>
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:3001,nop,nop>
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:4001,nop,nop>
+// Enter fast recovery.
+   +0 > . 1:1001(1000) ack 1
+ +.01 %{
+assert tcpi_ca_state == TCP_CA_Recovery, tcpi_ca_state
+assert tcpi_snd_cwnd == 7, tcpi_snd_cwnd
+assert tcpi_snd_ssthresh == 7, tcpi_snd_ssthresh
+}%
+
+// Write some more, which we will send 1 MSS at a time,
+// as in-flight segments are SACKed or ACKed.
+ +.01 write(4, ..., 7000) = 7000
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:5001,nop,nop>
+   +0 > . 10001:11001(1000) ack 1
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:6001,nop,nop>
+   +0 > . 11001:12001(1000) ack 1
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:7001,nop,nop>
+   +0 > . 12001:13001(1000) ack 1
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:8001,nop,nop>
+   +0 > . 13001:14001(1000) ack 1
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:9001,nop,nop>
+   +0 > . 14001:15001(1000) ack 1
+
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:10001,nop,nop>
+   +0 > . 15001:16001(1000) ack 1
+
+ +.02 < . 1:1(0) ack 10001 win 320
+   +0 > P. 16001:17001(1000) ack 1
+// Leave fast recovery.
+ +.01 %{
+assert tcpi_ca_state == TCP_CA_Open, tcpi_ca_state
+assert tcpi_snd_cwnd == 7, tcpi_snd_cwnd
+assert tcpi_snd_ssthresh == 7, tcpi_snd_ssthresh
+}%
+
+ +.03 < . 1:1(0) ack 12001 win 320
+ +.02 < . 1:1(0) ack 14001 win 320
+ +.02 < . 1:1(0) ack 16001 win 320
+ +.02 < . 1:1(0) ack 17001 win 320

tools/testing/selftests/net/packetdrill/tcp_fast_recovery_prr-ss-30pkt-lost-1_4-11_16.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test PRR-slowstart implementation. The sender sends 20 packets. Packet
+// 1 to 4, and 11 to 16 are dropped.
+`./defaults.sh`
+
+// Establish a connection.
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+
+ +.01 < . 1:1(0) ack 1 win 320
+   +0 accept(3, ..., ...) = 4
+
+// Write 20 data segments.
+   +0 write(4, ..., 20000) = 20000
+   +0 > P. 1:10001(10000) ack 1
+
+// Receive first DUPACK, entering PRR part
+ +.01 < . 1:1(0) ack 1 win 320 <sack 4001:5001,nop,nop>
+   +0 > . 10001:11001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:6001,nop,nop>
+   +0 > . 11001:12001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:7001,nop,nop>
+   +0 > . 1:1001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:8001,nop,nop>
+   +0 > . 1001:2001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:9001,nop,nop>
+   +0 > . 2001:3001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:10001,nop,nop>
+   +0 > . 3001:4001(1000) ack 1
+// Enter PRR CRB
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:11001,nop,nop>
+   +0 > . 12001:13001(1000) ack 1
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:12001,nop,nop>
+   +0 > . 13001:14001(1000) ack 1
+// Enter PRR slow start
+ +.01 < . 1:1(0) ack 1001 win 320 <sack 4001:12001,nop,nop>
+   +0 > P. 14001:16001(2000) ack 1
++.002 < . 1:1(0) ack 1001 win 320 <sack 2001:12001,nop,nop>
+   +0 > . 1001:2001(1000) ack 1
+   +0 > . 16001:17001(1000) ack 1
+// inflight reaches ssthresh, goes into packet conservation mode
++.002 < . 1:1(0) ack 1001 win 320 <sack 2001:13001,nop,nop>
+   +0 > . 17001:18001(1000) ack 1
++.002 < . 1:1(0) ack 1001 win 320 <sack 2001:14001,nop,nop>
+   +0 > . 18001:19001(1000) ack 1

tools/testing/selftests/net/packetdrill/tcp_fast_recovery_prr-ss-30pkt-lost1_4.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test PRR-slowstart implementation. The sender sends 20 packets. Packet
+// 1 to 4 are lost. The sender writes another 10 packets.
+`./defaults.sh`
+
+// Establish a connection.
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+
+ +.01 < . 1:1(0) ack 1 win 320
+   +0 accept(3, ..., ...) = 4
+
+// Send 20 data segments.
+   +0 write(4, ..., 10000) = 10000
+   +0 > P. 1:10001(10000) ack 1
+
+// Lost packet 1,2,3,4
+ +.01 < . 1:1(0) ack 1 win 320 <sack 4001:5001,nop,nop>
++.002 < . 1:1(0) ack 1 win 320 <sack 4001:6001,nop,nop>
+   +0 < . 1:1(0) ack 1 win 320 <sack 4001:7001,nop,nop>
+   +0 > . 1:1001(1000) ack 1
+   +0 < . 1:1(0) ack 1 win 320 <sack 4001:8001,nop,nop>
+   +0 > . 1001:2001(1000) ack 1
+   +0 < . 1:1(0) ack 1 win 320 <sack 4001:9001,nop,nop>
+   +0 > . 2001:3001(1000) ack 1
+   +0 < . 1:1(0) ack 1 win 320 <sack 4001:10001,nop,nop>
+   +0 > . 3001:4001(1000) ack 1
+
+// Receiver ACKs all data.
+ +.01 < . 1:1(0) ack 1001 win 320 <sack 4001:10001,nop,nop>
+   +0 < . 1:1(0) ack 2001 win 320 <sack 4001:10001,nop,nop>
+   +0 < . 1:1(0) ack 3001 win 320 <sack 4001:10001,nop,nop>
+   +0 < . 1:1(0) ack 10001 win 320
+
+// Writes another 10 packets, which the ssthresh*mss amount
+// should be sent right away
+ +.01 write(4, ..., 10000) = 10000
+   +0 > . 10001:17001(7000) ack 1

tools/testing/selftests/net/packetdrill/tcp_fast_recovery_prr-ss-ack-below-snd_una-cubic.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test PRR-slowstart implementation.
+// In this variant we verify that the sender uses SACK info on an ACK
+// below snd_una.
+
+// Set up config.
+`./defaults.sh`
+
+// Establish a connection.
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 8>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+// RTT 10ms
+ +.01 < . 1:1(0) ack 1 win 320
+   +0 accept(3, ..., ...) = 4
+
+// Send 10 data segments.
+   +0 write(4, ..., 10000) = 10000
+   +0 > P. 1:10001(10000) ack 1
+
+// Lost packet 1:1001,4001:5001,7001:8001.
+ +.01 < . 1:1(0) ack 1 win 320 <sack 1001:2001,nop,nop>
+   +0 < . 1:1(0) ack 1 win 320 <sack 1001:3001,nop,nop>
+   +0 < . 1:1(0) ack 1 win 320 <sack 1001:3001 8001:9001,nop,nop>
+   +0 > . 1:1001(1000) ack 1
+
++.012 < . 1:1(0) ack 4001 win 320 <sack 8001:9001,nop,nop>
+   +0 > . 4001:7001(3000) ack 1
+
+   +0 write(4, ..., 10000) = 10000
+
+// The following ACK was reordered - delayed so that it arrives with
+// an ACK field below snd_una. Here we check that the newly-SACKed
+// 2MSS at 5001:7001 cause us to send out 2 more MSS.
++.002 < . 1:1(0) ack 3001 win 320 <sack 5001:7001,nop,nop>
+   +0 > . 7001:8001(1000) ack 1
+   +0 > . 10001:11001(1000) ack 1

tools/testing/selftests/net/packetdrill/tcp_nagle_https_client.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// This is a test inspired by an Android client app using SSL. This
+// test verifies using TCP_NODELAY would save application latency
+// (Perhaps even better with TCP_NAGLE).
+//
+`./defaults.sh
+ethtool -K tun0 tso off gso off
+./set_sysctls.py /proc/sys/net/ipv4/tcp_timestamps=0`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 4
+   +0 fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0
+
+   +0 connect(4, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+   +0 > S 0:0(0) <mss 1460,nop,nop,sackOK,nop,wscale 8>
+  +.1 < S. 0:0(0) ack 1 win 5792 <mss 974,nop,nop,sackOK,nop,wscale 7>
+   +0 > . 1:1(0) ack 1
+
+// SSL handshake (resumed session)
+   +0 write(4, ..., 517) = 517
+   +0 > P. 1:518(517) ack 1
+  +.1 < . 1:1(0) ack 518 win 229
+
+   +0 < P. 1:144(143) ack 1 win 229
+   +0 > . 518:518(0) ack 144
+   +0 read(4, ..., 1000) = 143
+
+// Application POST header (51B) and body (2002B)
+   +0 write(4, ..., 51) = 51
+   +0 > P. 518:569(51) ack 144
+ +.03 write(4, ..., 2002) = 2002
+   +0 > . 569:1543(974) ack 144
+   +0 > P. 1543:2517(974) ack 144
+// Without disabling Nagle, this packet will not happen until the remote ACK.
+   +0 > P. 2517:2571(54) ack 144
+
+  +.1 < . 1:1(0) ack 2571 win 229
+
+// Reset sysctls
+`/tmp/sysctl_restore_${PPID}.sh`

tools/testing/selftests/net/packetdrill/tcp_nagle_sendmsg_msg_more.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test the MSG_MORE flag will correctly corks the tiny writes
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 32792 <mss 1000,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,wscale 8>
+ +.01 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+// Disable Nagle by default on this socket.
+   +0 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0
+
+// Test the basic case: MSG_MORE overwrites TCP_NODELAY and enables Nagle.
+   +0 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 40}], msg_flags=0}, MSG_MORE) = 40
+ +.21~+.215 > P. 1:41(40) ack 1
+ +.01 < . 1:1(0) ack 41 win 257
+
+// Test unsetting MSG_MORE releases the packet
+   +0 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 100}], msg_flags=0}, MSG_MORE) = 100
++.005 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 160}], msg_flags=0}, MSG_MORE) = 160
+ +.01 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(3)=[{..., 100}, {..., 200}, {..., 195}],
+		  msg_flags=0}, MSG_MORE) = 495
++.008 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 5}], msg_flags=0}, 0) = 5
+   +0 > P. 41:801(760) ack 1
+ +.02 < . 1:1(0) ack 801 win 257
+
+
+// Test >MSS write will unleash MSS packets but hold on the remaining data.
+  +.1 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 3100}], msg_flags=0}, MSG_MORE) = 3100
+   +0 > . 801:3801(3000) ack 1
++.003 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 50}], msg_flags=0}, MSG_MORE) = 50
+
+ +.01 < . 1:1(0) ack 2801 win 257
+// Err... we relase the remaining right after the ACK? note that PUSH is reset
+   +0 > . 3801:3951(150) ack 1
+
+// Test we'll hold on the subsequent writes when inflight (3801:3951) > 0
++.001 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 1}], msg_flags=0}, MSG_MORE) = 1
++.002 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 2}], msg_flags=0}, MSG_MORE) = 2
++.003 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 3}], msg_flags=0}, MSG_MORE) = 3
++.004 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 4}], msg_flags=0}, MSG_MORE) = 4
+ +.02 < . 1:1(0) ack 3951 win 257
+   +0 > . 3951:3961(10) ack 1
+ +.02 < . 1:1(0) ack 3961 win 257
+
+
+// Test the case a MSG_MORE send followed by a write flushes the data
+   +0 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{..., 20}], msg_flags=0}, MSG_MORE) = 20
+ +.05 write(4, ..., 20) = 20
+   +0 > P. 3961:4001(40) ack 1

tools/testing/selftests/net/packetdrill/tcp_nagle_sockopt_cork_nodelay.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test TCP_CORK and TCP_NODELAY sockopt behavior
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 32792 <mss 1000,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,wscale 8>
+ +.01 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+// Set TCP_CORK sockopt to hold small packets
+   +0 setsockopt(4, SOL_TCP, TCP_CORK, [1], 4) = 0
+
+   +0 write(4, ..., 40) = 40
+ +.05 write(4, ..., 40) = 40
+
+// Unset TCP_CORK should push pending bytes out
+ +.01 setsockopt(4, SOL_TCP, TCP_CORK, [0], 4) = 0
+   +0 > P. 1:81(80) ack 1
+ +.01 < . 1:1(0) ack 81 win 257
+
+// Set TCP_CORK sockopt to hold small packets
+   +0 setsockopt(4, SOL_TCP, TCP_CORK, [1], 4) = 0
+
+   +0 write(4, ..., 40) = 40
+ +.05 write(4, ..., 40) = 40
+
+// Set TCP_NODELAY sockopt should push pending bytes out
+   +0 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0
+   +0 > P. 81:161(80) ack 1
+ +.01 < . 1:1(0) ack 161 win 257
+
+// Set MSG_MORE to hold small packets
+   +0 send(4, ..., 40, MSG_MORE) = 40
+ +.05 send(4, ..., 40, MSG_MORE) = 40
+
+// Set TCP_NODELAY sockopt should push pending bytes out
+ +.01 setsockopt(4, SOL_TCP, TCP_NODELAY, [1], 4) = 0
+   +0 > . 161:241(80) ack 1
+ +.01 < . 1:1(0) ack 241 win 257

tools/testing/selftests/net/packetdrill/tcp_timestamping_client-only-last-byte.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test that tx timestamping sends timestamps only for
+// the last byte of each sendmsg.
+`./defaults.sh
+`
+
+// Create a socket and set it to non-blocking.
+    0	socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0	fcntl(3, F_GETFL) = 0x2 (flags O_RDWR)
+   +0	fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+
+// Establish connection and verify that there was no error.
+   +0	connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+   +0 > S 0:0(0) <mss 1460,sackOK,TS val 100 ecr 0,nop,wscale 8>
+ +.01 < S. 0:0(0) ack 1 win 20000 <mss 1000,nop,nop,sackOK>
+   +0 > . 1:1(0) ack 1
+   +0	getsockopt(3, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
+   +0	fcntl(3, F_SETFL, O_RDWR) = 0   // set back to blocking
+
+   +0	setsockopt(3, SOL_SOCKET, SO_TIMESTAMPING,
+		   [SOF_TIMESTAMPING_TX_SCHED | SOF_TIMESTAMPING_TX_SOFTWARE |
+		    SOF_TIMESTAMPING_TX_ACK | SOF_TIMESTAMPING_SOFTWARE |
+		    SOF_TIMESTAMPING_OPT_ID], 4) = 0
+
+   +0	write(3, ..., 11000) = 11000
+   +0	> P. 1:10001(10000) ack 1
+ +.01	< . 1:1(0) ack 10001 win 4000
+   +0	> P. 10001:11001(1000) ack 1
+ +.01	< . 1:1(0) ack 11001 win 4000
+
+// Make sure that internal TCP timestamps are not overwritten and we have sane
+// RTT measurement.
+   +0	%{
+assert 5000 <= tcpi_rtt <= 20000, 'srtt=%d us' % tcpi_rtt
+}%
+
+// SCM_TSTAMP_SCHED for the last byte should be received almost immediately
+// once 10001 is acked at t=20ms.
+// setsockopt(..., [SOF_TIMESTAMPING_SOFTWARE | SOF_TIMESTAMPING_OPT_ID], ...)
+// is called after when SYN is acked. So, we expect the last byte of the first
+// chunk to have a timestamp key of 10999 (i.e., 11000 - 1).
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=20000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SCHED,
+				    ee_data=10999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_SND for the last byte should be received almost immediately
+// once 10001 is acked at t=20ms.
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=20000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SND,
+				    ee_data=10999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_ACK for the last byte should be received at t=30ms.
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=30000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_ACK,
+				    ee_data=10999}}
+		    ]}, MSG_ERRQUEUE) = 0

tools/testing/selftests/net/packetdrill/tcp_timestamping_partial.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test tx timestamping for partial writes (IPv4).
+`./defaults.sh
+`
+
+// Create a socket and set it to non-blocking.
+    0	socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0	fcntl(3, F_GETFL) = 0x2 (flags O_RDWR)
+   +0	fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+
+// Establish connection and verify that there was no error.
+   +0	connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+   +0	> S 0:0(0) <mss 1460,sackOK,TS val 100 ecr 0,nop,wscale 8>
+ +.01	< S. 0:0(0) ack 1 win 2000 <mss 1000,sackOK,TS val 700 ecr 100,nop,wscale 7>
+   +0	> . 1:1(0) ack 1 <nop,nop,TS val 200 ecr 700>
+   +0	getsockopt(3, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
+
+   +0	setsockopt(3, SOL_SOCKET, SO_SNDBUF, [1000], 4) = 0
+   +0	setsockopt(3, SOL_SOCKET, SO_TIMESTAMPING,
+		   [SOF_TIMESTAMPING_TX_SCHED | SOF_TIMESTAMPING_TX_SOFTWARE |
+		    SOF_TIMESTAMPING_TX_ACK | SOF_TIMESTAMPING_SOFTWARE |
+		    SOF_TIMESTAMPING_OPT_ID], 4) = 0
+
+// We have a partial write.
+   +0	write(3, ..., 10000) = 2964
+   +0	> . 1:989(988) ack 1 <nop,nop,TS val 110 ecr 700>
+   +0	> P. 989:1977(988) ack 1 <nop,nop,TS val 110 ecr 700>
+ +.01	< . 1:1(0) ack 1977 win 92 <nop,nop,TS val 800 ecr 200>
+   +0	> P. 1977:2965(988) ack 1 <nop,nop,TS val 114 ecr 800>
+ +.01	< . 1:1(0) ack 2965 win 92 <nop,nop,TS val 800 ecr 200>
+
+// Make sure that internal TCP timestamps are not overwritten and we have sane
+// RTT measurement.
+   +0	%{
+assert 5000 <= tcpi_rtt <= 20000, 'srtt=%d us' % tcpi_rtt
+}%
+
+// SCM_TSTAMP_SCHED for the first chunk should be received almost immediately
+// after the first ack at t=20ms.
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=20000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SCHED,
+				    ee_data=2963}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_SND for the first chunk should be received almost immediately
+// after the first ack at t=20ms.
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=20000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SND,
+				    ee_data=2963}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_ACK for the first chunk should be received after the last ack at
+// t=30ms.
+   +0	recvmsg(3, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=30000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_ACK,
+				    ee_data=2963}}
+		    ]}, MSG_ERRQUEUE) = 0

tools/testing/selftests/net/packetdrill/tcp_timestamping_server.pkt

···
+// SPDX-License-Identifier: GPL-2.0
+// Test tx timestamping for server-side (IPv4).
+`./defaults.sh
+`
+
+// Initialize connection
+    0	socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0	setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0	bind(3, ..., ...) = 0
+   +0	listen(3, 1) = 0
+
+   +0	< S 0:0(0) win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 10>
+   +0	> S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+ +.01	< . 1:1(0) ack 1 win 514
+
+   +0	accept(3, ..., ...) = 4
+   +0	setsockopt(4, SOL_SOCKET, SO_TIMESTAMPING,
+		   [SOF_TIMESTAMPING_TX_SCHED | SOF_TIMESTAMPING_TX_SOFTWARE |
+		    SOF_TIMESTAMPING_TX_ACK | SOF_TIMESTAMPING_SOFTWARE |
+		    SOF_TIMESTAMPING_OPT_ID], 4) = 0
+
+// Write two 2KB chunks.
+// setsockopt(..., [SOF_TIMESTAMPING_SOFTWARE | SOF_TIMESTAMPING_OPT_ID], ...)
+// is called after when SYN is acked. So, we expect the last byte of the first
+// and the second chunks to have timestamp keys of 1999 (i.e., 2000 - 1) and
+// 3999 (i.e., 4000 - 1) respectively.
+   +0	write(4, ..., 2000) = 2000
+   +0	write(4, ..., 2000) = 2000
+   +0	> P. 1:2001(2000) ack 1
+   +0	> P. 2001:4001(2000) ack 1
+ +.01	< .  1:1(0) ack 2001 win 514
+ +.01	< .  1:1(0) ack 4001 win 514
+
+// Make sure that internal TCP timestamps are not overwritten and we have sane
+// RTT measurement.
+   +0	%{
+assert 5000 <= tcpi_rtt <= 20000, 'srtt=%d us' % tcpi_rtt
+}%
+
+// SCM_TSTAMP_SCHED for the first chunk should be received almost immediately
+// after write at t=10ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=10000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SCHED,
+				    ee_data=1999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_SND for the first chunk should be received almost immediately
+// after write at t=10ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=10000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SND,
+				    ee_data=1999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_SCHED for the second chunk should be received almost immediately
+// after that at t=10ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=10000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SCHED,
+				    ee_data=3999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_SND for the second chunk should be received almost immediately
+// after that at t=10ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=10000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_SND,
+				    ee_data=3999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_ACK for the first chunk should be received at t=20ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=20000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_ACK,
+				    ee_data=1999}}
+		    ]}, MSG_ERRQUEUE) = 0
+// SCM_TSTAMP_ACK for the second chunk should be received at t=30ms.
+   +0	recvmsg(4, {msg_name(...)=...,
+		    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE|MSG_TRUNC,
+                    msg_control=[
+			{cmsg_level=SOL_SOCKET,
+			 cmsg_type=SCM_TIMESTAMPING,
+			 cmsg_data={scm_sec=0,scm_nsec=30000000}},
+			{cmsg_level=CMSG_LEVEL_IP,
+			 cmsg_type=CMSG_TYPE_RECVERR,
+			 cmsg_data={ee_errno=ENOMSG,
+				    ee_origin=SO_EE_ORIGIN_TIMESTAMPING,
+				    ee_type=0,
+				    ee_code=0,
+				    ee_info=SCM_TSTAMP_ACK,
+				    ee_data=3999}}
+		    ]}, MSG_ERRQUEUE) = 0