Merge branch 'selftests-net-packetdrill-netns-and-two-imports'

Willem de Bruijn says:

====================
selftests/net: packetdrill: netns and two imports

From: Willem de Bruijn <willemb@google.com>

1/3: run in nets, as discussed, and add missing CONFIGs
2/3: import tcp/zerocopy
3/3: import tcp/slow_start
====================

Link: https://patch.msgid.link/20240912005317.1253001-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

25 files changed, 1108 insertions, 2 deletions

diff --git a/tools/testing/selftests/net/packetdrill/Makefile b/tools/testing/selftests/net/packetdrill/Makefile
index 870f7258dc8d..31cfb666ba8b 100644
--- a/tools/testing/selftests/net/packetdrill/Makefile
+++ b/tools/testing/selftests/net/packetdrill/Makefile
@@ -2,6 +2,7 @@
 
 TEST_INCLUDES := ksft_runner.sh \
 		 defaults.sh \
+		 set_sysctls.py \
 		 ../../kselftest/ktap_helpers.sh
 
 TEST_PROGS := $(wildcard *.pkt)
diff --git a/tools/testing/selftests/net/packetdrill/config b/tools/testing/selftests/net/packetdrill/config
index 0d402830f18d..0237ed98f3c0 100644
--- a/tools/testing/selftests/net/packetdrill/config
+++ b/tools/testing/selftests/net/packetdrill/config
@@ -1,5 +1,11 @@
 CONFIG_IPV6=y
+CONFIG_HZ_1000=y
+CONFIG_HZ=1000
+CONFIG_NET_NS=y
 CONFIG_NET_SCH_FIFO=y
+CONFIG_NET_SCH_FQ=y
 CONFIG_PROC_SYSCTL=y
+CONFIG_SYN_COOKIES=y
+CONFIG_TCP_CONG_CUBIC=y
 CONFIG_TCP_MD5SIG=y
 CONFIG_TUN=y
diff --git a/tools/testing/selftests/net/packetdrill/ksft_runner.sh b/tools/testing/selftests/net/packetdrill/ksft_runner.sh
index 2f62caccbbbc..7478c0c0c9aa 100755
--- a/tools/testing/selftests/net/packetdrill/ksft_runner.sh
+++ b/tools/testing/selftests/net/packetdrill/ksft_runner.sh
@@ -33,9 +33,9 @@ fi
 ktap_print_header
 ktap_set_plan 2
 
-packetdrill ${ipv4_args[@]} $(basename $script) > /dev/null \
+unshare -n packetdrill ${ipv4_args[@]} $(basename $script) > /dev/null \
 	&& ktap_test_pass "ipv4" || ktap_test_fail "ipv4"
-packetdrill ${ipv6_args[@]} $(basename $script) > /dev/null \
+unshare -n packetdrill ${ipv6_args[@]} $(basename $script) > /dev/null \
 	&& ktap_test_pass "ipv6" || ktap_test_fail "ipv6"
 
 ktap_finished
diff --git a/tools/testing/selftests/net/packetdrill/set_sysctls.py b/tools/testing/selftests/net/packetdrill/set_sysctls.py
new file mode 100755
index 000000000000..5ddf456ae973
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/set_sysctls.py
@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: GPL-2.0
+
+"""Sets sysctl values and writes a file that restores them.
+
+The arguments are of the form "<proc-file>=<val>" separated by spaces.
+The program first reads the current value of the proc-file and creates
+a shell script named "/tmp/sysctl_restore_${PACKETDRILL_PID}.sh" which
+restores the values when executed. It then sets the new values.
+
+PACKETDRILL_PID is set by packetdrill to the pid of itself, so a .pkt
+file could restore sysctls by running `/tmp/sysctl_restore_${PPID}.sh`
+at the end.
+"""
+
+import os
+import subprocess
+import sys
+
+filename = '/tmp/sysctl_restore_%s.sh' % os.environ['PACKETDRILL_PID']
+
+# Open file for restoring sysctl values
+restore_file = open(filename, 'w')
+print('#!/bin/bash', file=restore_file)
+
+for a in sys.argv[1:]:
+  sysctl = a.split('=')
+  # sysctl[0] contains the proc-file name, sysctl[1] the new value
+
+  # read current value and add restore command to file
+  cur_val = subprocess.check_output(['cat', sysctl[0]], universal_newlines=True)
+  print('echo "%s" > %s' % (cur_val.strip(), sysctl[0]), file=restore_file)
+
+  # set new value
+  cmd = 'echo "%s" > %s' % (sysctl[1], sysctl[0])
+  os.system(cmd)
+
+os.system('chmod u+x %s' % filename)
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-1pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-1pkt.pkt
new file mode 100644
index 000000000000..795c476d222d
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-1pkt.pkt
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when not application-limited, so that
+// the cwnd continues to grow.
+// In this variant, the receiver ACKs every packet.
+
+// Set up config. To keep things simple, disable the
+// mechanism that defers sending in order to send bigger TSO packets.
+`./defaults.sh
+sysctl -q net.ipv4.tcp_tso_win_divisor=100`
+
+    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>
+  +.1 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 30000) = 30000
+   +0 > P. 1:10001(10000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
++.105 < . 1:1(0) ack 1001 win 257
+   +0 > P. 10001:12001(2000) ack 1
+
+   +0 < . 1:1(0) ack 2001 win 257
+   +0 > P. 12001:14001(2000) ack 1
+
++.005 < . 1:1(0) ack 3001 win 257
+   +0 > P. 14001:16001(2000) ack 1
+
+   +0 < . 1:1(0) ack 4001 win 257
+   +0 > P. 16001:18001(2000) ack 1
+
++.005 < . 1:1(0) ack 5001 win 257
+   +0 > P. 18001:20001(2000) ack 1
+
+   +0 < . 1:1(0) ack 6001 win 257
+   +0 > P. 20001:22001(2000) ack 1
+
++.005 < . 1:1(0) ack 7001 win 257
+   +0 > P. 22001:24001(2000) ack 1
+
+   +0 < . 1:1(0) ack 8001 win 257
+   +0 > P. 24001:26001(2000) ack 1
+
++.005 < . 1:1(0) ack 9001 win 257
+   +0 > P. 26001:28001(2000) ack 1
+
+   +0 < . 1:1(0) ack 10001 win 257
+   +0 > P. 28001:30001(2000) ack 1
+
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-5pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-5pkt.pkt
new file mode 100644
index 000000000000..9212ae1fd0f2
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-5pkt.pkt
@@ -0,0 +1,33 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when an outstanding flight of packets is
+// less than the current cwnd, and not big enough to bump up cwnd.
+//
+// In this variant, the receiver ACKs every other packet,
+// approximating standard delayed ACKs.
+
+// Set up config.
+`./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,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+   +0 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+
+// Only send 5 packets.
+   +0 write(4, ..., 5000) = 5000
+   +0 > P. 1:5001(5000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 2001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 10, 'cwnd=%d' % tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 4001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 10, 'cwnd=%d' % tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 5001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 10, 'cwnd=%d' % tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-6pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-6pkt.pkt
new file mode 100644
index 000000000000..416c901ddf51
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt-send-6pkt.pkt
@@ -0,0 +1,34 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when an outstanding flight of packets is
+// less than the current cwnd, but still big enough that in slow
+// start we want to increase our cwnd a little.
+//
+// In this variant, the receiver ACKs every other packet,
+// approximating standard delayed ACKs.
+
+// Set up config.
+`./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,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+   +0 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+
+// Only send 6 packets.
+   +0 write(4, ..., 6000) = 6000
+   +0 > P. 1:6001(6000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 2001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 12, 'cwnd=%d' % tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 4001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 12, 'cwnd=%d' % tcpi_snd_cwnd }%
+
+   +0 < . 1:1(0) ack 6001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 12, 'cwnd=%d' % tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt.pkt
new file mode 100644
index 000000000000..a894b7d4559c
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-2pkt.pkt
@@ -0,0 +1,42 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when not application-limited, so that
+// the cwnd continues to grow.
+// In this variant, the receiver ACKs every other packet,
+// approximating standard delayed ACKs.
+
+// Set up config. To keep things simple, disable the
+// mechanism that defers sending in order to send bigger TSO packets.
+`./defaults.sh
+sysctl -q net.ipv4.tcp_tso_win_divisor=100`
+
+    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>
+  +.1 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 30000) = 30000
+   +0 > P. 1:10001(10000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
++.105 < . 1:1(0) ack 2001 win 257
+   +0 > P. 10001:14001(4000) ack 1
+
++.005 < . 1:1(0) ack 4001 win 257
+   +0 > P. 14001:18001(4000) ack 1
+
++.005 < . 1:1(0) ack 6001 win 257
+   +0 > P. 18001:22001(4000) ack 1
+
++.005 < . 1:1(0) ack 8001 win 257
+   +0 > P. 22001:26001(4000) ack 1
+
++.005 < . 1:1(0) ack 10001 win 257
+   +0 > P. 26001:30001(4000) ack 1
+
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-4pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-4pkt.pkt
new file mode 100644
index 000000000000..065fae9e9abd
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-ack-per-4pkt.pkt
@@ -0,0 +1,35 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when not application-limited, so that
+// the cwnd continues to grow.
+// In this variant, the receiver sends one ACK per 4 packets.
+
+// Set up config. To keep things simple, disable the
+// mechanism that defers sending in order to send bigger TSO packets.
+`./defaults.sh
+sysctl -q net.ipv4.tcp_tso_win_divisor=100`
+
+    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>
+  +.1 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 30000) = 30000
+   +0 > P. 1:10001(10000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+ +.11 < . 1:1(0) ack 4001 win 257
+   +0 > P. 10001:18001(8000) ack 1
+
+ +.01 < . 1:1(0) ack 8001 win 257
+   +0 > P. 18001:26001(8000) ack 1
+
++.005 < . 1:1(0) ack 10001 win 257
+   +0 > P. 26001:30001(4000) ack 1
+
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-idle.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-idle.pkt
new file mode 100644
index 000000000000..11b213be1138
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-idle.pkt
@@ -0,0 +1,39 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start after idle
+// This test expects tso size to be at least initial cwnd * mss
+
+`./defaults.sh
+./set_sysctls.py /proc/sys/net/ipv4/tcp_slow_start_after_idle=1 \
+		 /proc/sys/net/ipv4/tcp_min_tso_segs=10`
+
+    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 65535 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+  +.1 < . 1:1(0) ack 1 win 511
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 26000) = 26000
+   +0 > P. 1:5001(5000) ack 1
+   +0 > P. 5001:10001(5000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+  +.1 < . 1:1(0) ack 10001 win 511
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
+   +0 > P. 10001:20001(10000) ack 1
+   +0 > P. 20001:26001(6000) ack 1
+
+  +.1 < . 1:1(0) ack 26001 win 511
+   +0 %{ assert tcpi_snd_cwnd == 36, tcpi_snd_cwnd }%
+
+   +2 write(4, ..., 20000) = 20000
+// If slow start after idle works properly, we should send 5 MSS here (cwnd/2)
+   +0 > P. 26001:31001(5000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+// Reset sysctls
+`/tmp/sysctl_restore_${PPID}.sh`
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-win-update.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-win-update.pkt
new file mode 100644
index 000000000000..577ed8c8852c
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-after-win-update.pkt
@@ -0,0 +1,50 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start after window update
+// This test expects tso size to be at least initial cwnd * mss
+
+`./defaults.sh
+./set_sysctls.py /proc/sys/net/ipv4/tcp_slow_start_after_idle=1 \
+		 /proc/sys/net/ipv4/tcp_min_tso_segs=10`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 65535 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+  +.1 < . 1:1(0) ack 1 win 511
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 26000) = 26000
+   +0 > P. 1:5001(5000) ack 1
+   +0 > P. 5001:10001(5000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
+  +.1 < . 1:1(0) ack 10001 win 511
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
+   +0 > P. 10001:20001(10000) ack 1
+   +0 > P. 20001:26001(6000) ack 1
+
+  +.1 < . 1:1(0) ack 26001 win 0
+   +0 %{ assert tcpi_snd_cwnd == 36, tcpi_snd_cwnd }%
+
+   +0 write(4, ..., 20000) = 20000
+// 1st win0 probe
++.3~+.310 > . 26000:26000(0) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 36, tcpi_snd_cwnd }%
+
+// 2nd win0 probe
++.6~+.620 > . 26000:26000(0) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 36, tcpi_snd_cwnd }%
+
+// 3rd win0 probe
++1.2~+1.240 > . 26000:26000(0) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 36, tcpi_snd_cwnd }%
+
+  +.9 < . 1:1(0) ack 26001 win 511
+   +0 > P. 26001:31001(5000) ack 1
+
+// Reset sysctls
+`/tmp/sysctl_restore_${PPID}.sh`
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited-9-packets-out.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited-9-packets-out.pkt
new file mode 100644
index 000000000000..869f32c35a2a
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited-9-packets-out.pkt
@@ -0,0 +1,38 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when application-limited: in this case,
+// with IW10, if we don't fully use our cwnd but instead
+// send just 9 packets, then cwnd should grow to twice that
+// value, or 18 packets.
+
+// Set up config.
+`./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
+
+  +.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>
+  +.1 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 9000) = 9000
+   +0 > P. 1:9001(9000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
++.105 < . 1:1(0) ack 2001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 12, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 4001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 14, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 6001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 16, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 8001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 18, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 9001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 18, tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited.pkt
new file mode 100644
index 000000000000..0f77b7955db6
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-app-limited.pkt
@@ -0,0 +1,36 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when application-limited: in this case,
+// with IW10, if we send exactly 10 packets then cwnd should grow to 20.
+
+// Set up config.
+`./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
+
+  +.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>
+  +.1 < . 1:1(0) ack 1 win 257
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 10000) = 10000
+   +0 > P. 1:10001(10000) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
++.105 < . 1:1(0) ack 2001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 12, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 4001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 14, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 6001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 16, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 8001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 18, tcpi_snd_cwnd }%
+
++.005 < . 1:1(0) ack 10001 win 257
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-fq-ack-per-2pkt.pkt b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-fq-ack-per-2pkt.pkt
new file mode 100644
index 000000000000..7e9c83d617c2
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_slow_start_slow-start-fq-ack-per-2pkt.pkt
@@ -0,0 +1,63 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test of slow start when not application-limited, so that
+// the cwnd continues to grow, even if TSQ triggers.
+// In this variant, the receiver ACKs every other packet,
+// approximating standard delayed ACKs.
+
+// Note we use FQ/pacing to check if TCP Small Queues is not hurting
+
+`./defaults.sh
+tc qdisc replace dev tun0 root fq
+sysctl -q net/ipv4/tcp_pacing_ss_ratio=200
+sysctl -e -q net.ipv4.tcp_min_tso_segs=2`
+
+    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 1460,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+  +.1 < . 1:1(0) ack 1 win 500
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_SNDBUF, [200000], 4) = 0
+
+   +0 write(4, ..., 40000) = 40000
+// This might change if we cook the initial packet with 10 MSS.
+   +0 > P. 1:2921(2920) ack 1
+   +0 > P. 2921:5841(2920) ack 1
+   +0 > P. 5841:8761(2920) ack 1
+   +0 > P. 8761:11681(2920) ack 1
+   +0 > P. 11681:14601(2920) ack 1
+   +0 %{ assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd }%
+
++.105 < . 1:1(0) ack 2921 win 500
+   +0 %{ assert tcpi_snd_cwnd == 12, tcpi_snd_cwnd }%
+
+// Note: after this commit : "net_sched: sch_fq: account for schedule/timers drifts"
+// FQ notices that this packet missed the 'time to send next packet' computed
+// when prior packet (11681:14601(2920)) was sent.
+// So FQ will allow following packet to be sent a bit earlier (quantum/2)
+// (FQ commit allows an application/cwnd limited flow to get at most quantum/2 extra credit)
+   +0 > P. 14601:17521(2920) ack 1
+
++.003 < . 1:1(0) ack 5841 win 500
+   +0 %{ assert tcpi_snd_cwnd == 14, tcpi_snd_cwnd }%
+
++.001 > P. 17521:20441(2920) ack 1
+
++.001 < . 1:1(0) ack 8761 win 500
+   +0 %{ assert tcpi_snd_cwnd == 16, tcpi_snd_cwnd }%
+
+// remaining packets are delivered at a constant rate.
++.007 > P. 20441:23361(2920) ack 1
+
++.002 < . 1:1(0) ack 11681 win 500
+   +0 %{ assert tcpi_snd_cwnd == 18, tcpi_snd_cwnd }%
++.001 < . 1:1(0) ack 14601 win 500
+
++.004 > P. 23361:26281(2920) ack 1
+
++.007 > P. 26281:29201(2920) ack 1
+
+   +0 %{ assert tcpi_snd_cwnd == 20, 'cwnd=%d' % tcpi_snd_cwnd }%
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_basic.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_basic.pkt
new file mode 100644
index 000000000000..a82c8899d36b
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_basic.pkt
@@ -0,0 +1,55 @@
+// SPDX-License-Identifier: GPL-2.0
+// basic zerocopy test:
+//
+// send a packet with MSG_ZEROCOPY and receive the notification ID
+// repeat and verify IDs are consecutive
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 4001:8001(4000) ack 1
+   +0 < . 1:1(0) ack 8001 win 257
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=1,
+                                    ee_data=1}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_batch.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_batch.pkt
new file mode 100644
index 000000000000..c01915e7f4a1
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_batch.pkt
@@ -0,0 +1,41 @@
+// SPDX-License-Identifier: GPL-2.0
+// batch zerocopy test:
+//
+// send multiple packets, then read one range of all notifications.
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_MARK, [666], 4) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 4001:8001(4000) ack 1
+   +0 < . 1:1(0) ack 8001 win 257
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=1}}
+                  ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_client.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_client.pkt
new file mode 100644
index 000000000000..6509882932e9
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_client.pkt
@@ -0,0 +1,30 @@
+// SPDX-License-Identifier: GPL-2.0
+// Minimal client-side zerocopy test
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+   +0...0 connect(4, ..., ...) = 0
+
+   +0 > S 0:0(0) <mss 1460,sackOK,TS val 0 ecr 0,nop,wscale 8>
+   +0 < S. 0:0(0) ack 1 win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > . 1:1(0) ack 1
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_closed.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_closed.pkt
new file mode 100644
index 000000000000..2cd78755cb2a
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_closed.pkt
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: GPL-2.0
+// send with MSG_ZEROCOPY on a non-established socket
+//
+// verify that a send in state TCP_CLOSE correctly aborts the zerocopy
+// operation, specifically it does not increment the zerocopy counter.
+//
+// First send on a closed socket and wait for (absent) notification.
+// Then connect and send and verify that notification nr. is zero.
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 4
+   +0 setsockopt(4, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = -1 EPIPE (Broken pipe)
+
+   +0.1 recvmsg(4, {msg_name(...)=...,
+                    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE,
+                    msg_control=[]}, MSG_ERRQUEUE) = -1 EAGAIN (Resource temporarily unavailable)
+
+   +0...0 connect(4, ..., ...) = 0
+
+   +0 > S 0:0(0) <mss 1460,sackOK,TS val 0 ecr 0,nop,wscale 8>
+   +0 < S. 0:0(0) ack 1 win 32792 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > . 1:1(0) ack 1
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_edge.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_edge.pkt
new file mode 100644
index 000000000000..7671c20e01cf
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_edge.pkt
@@ -0,0 +1,61 @@
+// SPDX-License-Identifier: GPL-2.0
+// epoll zerocopy test:
+//
+// EPOLLERR is known to be not edge-triggered unlike EPOLLIN and EPOLLOUT but
+// it is not level-triggered either.
+//
+// fire two sends with MSG_ZEROCOPY and receive the acks. confirm that EPOLLERR
+// is correctly fired only once, when EPOLLET is set. send another packet with
+// MSG_ZEROCOPY. confirm that EPOLLERR is correctly fired again only once.
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   +0 fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 epoll_create(1) = 5
+   +0 epoll_ctl(5, EPOLL_CTL_ADD, 4, {events=EPOLLOUT|EPOLLET, fd=4}) = 0
+   +0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 4001:8001(4000) ack 1
+   +0 < . 1:1(0) ack 8001 win 257
+
+// receive only one EPOLLERR for the two sends above.
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 8001:12001(4000) ack 1
+   +0 < . 1:1(0) ack 12001 win 257
+
+// receive only one EPOLLERR for the third send above.
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=2}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_exclusive.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_exclusive.pkt
new file mode 100644
index 000000000000..fadc480fdb7f
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_exclusive.pkt
@@ -0,0 +1,63 @@
+// SPDX-License-Identifier: GPL-2.0
+// epoll zerocopy test:
+//
+// EPOLLERR is known to be not edge-triggered unlike EPOLLIN and EPOLLOUT but
+// it is not level-triggered either. this tests verify that the same behavior is
+// maintained when we have EPOLLEXCLUSIVE.
+//
+// fire two sends with MSG_ZEROCOPY and receive the acks. confirm that EPOLLERR
+// is correctly fired only once, when EPOLLET is set. send another packet with
+// MSG_ZEROCOPY. confirm that EPOLLERR is correctly fired again only once.
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   +0 fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 epoll_create(1) = 5
+   +0 epoll_ctl(5, EPOLL_CTL_ADD, 4,
+		{events=EPOLLOUT|EPOLLET|EPOLLEXCLUSIVE, fd=4}) = 0
+   +0 epoll_wait(5, {events=EPOLLOUT, fd=4}, 1, 0) = 1
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 4001:8001(4000) ack 1
+   +0 < . 1:1(0) ack 8001 win 257
+
+// receive only one EPOLLERR for the two sends above.
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 8001:12001(4000) ack 1
+   +0 < . 1:1(0) ack 12001 win 257
+
+// receive only one EPOLLERR for the third send above.
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=2}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_oneshot.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_oneshot.pkt
new file mode 100644
index 000000000000..5bfa0d1d2f4a
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_epoll_oneshot.pkt
@@ -0,0 +1,66 @@
+// SPDX-License-Identifier: GPL-2.0
+// epoll zerocopy test:
+//
+// This is a test to confirm that EPOLLERR is only fired once for an FD when
+// EPOLLONESHOT is set.
+//
+// fire two sends with MSG_ZEROCOPY and receive the acks. confirm that EPOLLERR
+// is correctly fired only once, when EPOLLONESHOT is set. send another packet
+// with MSG_ZEROCOPY. confirm that EPOLLERR is not fired. Rearm the FD and
+// confirm that EPOLLERR is correctly set.
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   +0 fcntl(4, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 epoll_create(1) = 5
+   +0 epoll_ctl(5, EPOLL_CTL_ADD, 4,
+		{events=EPOLLOUT|EPOLLET|EPOLLONESHOT, fd=4}) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 1:4001(4000) ack 1
+   +0 < . 1:1(0) ack 4001 win 257
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 4001:8001(4000) ack 1
+   +0 < . 1:1(0) ack 8001 win 257
+
+// receive only one EPOLLERR for the two sends above.
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 send(4, ..., 4000, MSG_ZEROCOPY) = 4000
+   +0 > P. 8001:12001(4000) ack 1
+   +0 < . 1:1(0) ack 12001 win 257
+
+// receive no EPOLLERR for the third send above.
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+// rearm the FD and verify the EPOLLERR is fired again.
+   +0 epoll_ctl(5, EPOLL_CTL_MOD, 4, {events=EPOLLOUT|EPOLLONESHOT, fd=4}) = 0
+   +0 epoll_wait(5, {events=EPOLLERR|EPOLLOUT, fd=4}, 1, 0) = 1
+   +0 epoll_wait(5, {events=0, ptr=0}, 1, 0) = 0
+
+   +0 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=2}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-client.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-client.pkt
new file mode 100644
index 000000000000..4a73bbf46961
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-client.pkt
@@ -0,0 +1,56 @@
+// SPDX-License-Identifier: GPL-2.0
+// Fastopen client zerocopy test:
+//
+// send data with MSG_FASTOPEN | MSG_ZEROCOPY and verify that the
+// kernel returns the notification ID.
+//
+// Fastopen requires a stored cookie. Create two sockets. The first
+// one will have no data in the initial send. On return 0 the
+// zerocopy notification counter is not incremented. Verify this too.
+
+`./defaults.sh`
+
+// Send a FastOpen request, no cookie yet so no data in SYN
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+   +0 sendto(3, ..., 500, MSG_FASTOPEN|MSG_ZEROCOPY, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+   +0 > S 0:0(0) <mss 1460,sackOK,TS val 1000 ecr 0,nop,wscale 8,FO,nop,nop>
+ +.01 < S. 123:123(0) ack 1 win 14600 <mss 940,TS val 2000 ecr 1000,sackOK,nop,wscale 6, FO abcd1234,nop,nop>
+   +0 > . 1:1(0) ack 1 <nop,nop,TS val 1001 ecr 2000>
+
+// Read from error queue: no zerocopy notification
+   +1 recvmsg(3, {msg_name(...)=...,
+                    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE,
+                    msg_control=[]}, MSG_ERRQUEUE) = -1 EAGAIN (Resource temporarily unavailable)
+
+ +.01 close(3) = 0
+   +0 > F. 1:1(0) ack 1 <nop,nop,TS val 1002 ecr 2000>
+ +.01 < F. 1:1(0) ack 2 win 92 <nop,nop,TS val 2001 ecr 1002>
+   +0 > .  2:2(0) ack 2 <nop,nop,TS val 1003 ecr 2001>
+
+// Send another Fastopen request, now SYN will have data
+ +.07 `sysctl -q net.ipv4.tcp_timestamps=0`
+  +.1 socket(..., SOCK_STREAM, IPPROTO_TCP) = 5
+   +0 fcntl(5, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+   +0 setsockopt(5, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+   +0 sendto(5, ..., 500, MSG_FASTOPEN|MSG_ZEROCOPY, ..., ...) = 500
+   +0 > S 0:500(500) <mss 1460,nop,nop,sackOK,nop,wscale 8,FO abcd1234,nop,nop>
+ +.05 < S. 5678:5678(0) ack 501 win 14600 <mss 1460,nop,nop,sackOK,nop,wscale 6>
+   +0 > . 501:501(0) ack 1
+
+// Read from error queue: now has first zerocopy notification
+   +0.5 recvmsg(5, {msg_name(...)=...,
+                    msg_iov(1)=[{...,0}],
+                    msg_flags=MSG_ERRQUEUE,
+                    msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-server.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-server.pkt
new file mode 100644
index 000000000000..36086c5877ce
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_fastopen-server.pkt
@@ -0,0 +1,44 @@
+// SPDX-License-Identifier: GPL-2.0
+// Fastopen server zerocopy test:
+//
+// send data with MSG_FASTOPEN | MSG_ZEROCOPY and verify that the
+// kernel returns the notification ID.
+
+`./defaults.sh
+ ./set_sysctls.py /proc/sys/net/ipv4/tcp_fastopen=0x207`
+
+// Set up a TFO server listening socket.
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+  +.1 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+   +0 setsockopt(3, SOL_TCP, TCP_FASTOPEN, [2], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+
+// Client sends a SYN with data.
+  +.1 < S 0:1000(1000) win 32792 <mss 1460,sackOK,nop,nop>
+   +0 > S. 0:0(0) ack 1001 <mss 1460,nop,nop,sackOK>
+
+// Server accepts and replies with data.
++.005 accept(3, ..., ...) = 4
+   +0 read(4, ..., 1024) = 1000
+   +0 sendto(4, ..., 1000, MSG_ZEROCOPY, ..., ...) = 1000
+   +0 > P. 1:1001(1000) ack 1001
+ +.05 < . 1001:1001(0) ack 1001 win 32792
+
+// Read from error queue: now has first zerocopy notification
+  +0.1 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                      {cmsg_level=CMSG_LEVEL_IP,
+                       cmsg_type=CMSG_TYPE_RECVERR,
+                       cmsg_data={ee_errno=0,
+                                  ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                  ee_type=0,
+                                  ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                  ee_info=0,
+                                  ee_data=0}}
+                  ]}, MSG_ERRQUEUE) = 0
+
+`/tmp/sysctl_restore_${PPID}.sh`
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_maxfrags.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_maxfrags.pkt
new file mode 100644
index 000000000000..672f817faca0
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_maxfrags.pkt
@@ -0,0 +1,118 @@
+// SPDX-License-Identifier: GPL-2.0
+// tcp_MAX_SKB_FRAGS test
+//
+// Verify that sending an iovec of tcp_MAX_SKB_FRAGS + 1 elements will
+// 1) fit in a single packet without zerocopy
+// 2) spill over into a second packet with zerocopy,
+//    because each iovec element becomes a frag
+// 3) the PSH bit is set on an skb when it runs out of fragments
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [1], 4) = 0
+
+   // Each pinned zerocopy page is fully accounted to skb->truesize.
+   // This test generates a worst case packet with each frag storing
+   // one byte, but increasing truesize with a page (64KB on PPC).
+   +0 setsockopt(3, SOL_SOCKET, SO_SNDBUF, [2000000], 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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   // send an iov of 18 elements: just becomes a linear skb
+   +0 sendmsg(4, {msg_name(...)=...,
+		  msg_iov(18)=[{..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}],
+		  msg_flags=0}, 0) = 18
+
+   +0 > P. 1:19(18) ack 1
+   +0 < . 1:1(0) ack 19 win 257
+
+   // send a zerocopy iov of 18 elements:
+   +1 sendmsg(4, {msg_name(...)=...,
+		  msg_iov(18)=[{..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}],
+		  msg_flags=0}, MSG_ZEROCOPY) = 18
+
+   // verify that it is split in one skb of 17 frags + 1 of 1 frag
+   // verify that both have the PSH bit set
+   +0 > P. 19:36(17) ack 1
+   +0 < . 1:1(0) ack 36 win 257
+
+   +0 > P. 36:37(1) ack 1
+   +0 < . 1:1(0) ack 37 win 257
+
+   +1 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
+
+   // send a zerocopy iov of 64 elements:
+   +0 sendmsg(4, {msg_name(...)=...,
+                  msg_iov(64)=[{..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1},
+			       {..., 1}, {..., 1}, {..., 1}, {..., 1}],
+                  msg_flags=0}, MSG_ZEROCOPY) = 64
+
+   // verify that it is split in skbs with 17 frags
+   +0 > P. 37:54(17) ack 1
+   +0 < . 1:1(0) ack 54 win 257
+
+   +0 > P. 54:71(17) ack 1
+   +0 < . 1:1(0) ack 71 win 257
+
+   +0 > P. 71:88(17) ack 1
+   +0 < . 1:1(0) ack 88 win 257
+
+   +0 > P. 88:101(13) ack 1
+   +0 < . 1:1(0) ack 101 win 257
+
+   +1 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=1,
+                                    ee_data=1}}
+                   ]}, MSG_ERRQUEUE) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_zerocopy_small.pkt b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_small.pkt
new file mode 100644
index 000000000000..a9a1ac0aea4f
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_zerocopy_small.pkt
@@ -0,0 +1,57 @@
+// SPDX-License-Identifier: GPL-2.0
+// small packet zerocopy test:
+//
+// verify that SO_EE_CODE_ZEROCOPY_COPIED is set on zerocopy
+// packets of all sizes, including the smallest payload, 1B.
+
+`./defaults.sh`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_SOCKET, SO_ZEROCOPY, [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>
+   +0 < . 1:1(0) ack 1 win 257
+
+   +0 accept(3, ..., ...) = 4
+
+   // send 1B
+   +0 send(4, ..., 1, MSG_ZEROCOPY) = 1
+   +0 > P. 1:2(1) ack 1
+   +0 < . 1:1(0) ack 2 win 257
+
+   +1 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=0,
+                                    ee_data=0}}
+                   ]}, MSG_ERRQUEUE) = 0
+
+   // send 1B again
+   +0 send(4, ..., 1, MSG_ZEROCOPY) = 1
+   +0 > P. 2:3(1) ack 1
+   +0 < . 1:1(0) ack 3 win 257
+
+   +1 recvmsg(4, {msg_name(...)=...,
+                  msg_iov(1)=[{...,0}],
+                  msg_flags=MSG_ERRQUEUE,
+                  msg_control=[
+                        {cmsg_level=CMSG_LEVEL_IP,
+                         cmsg_type=CMSG_TYPE_RECVERR,
+                         cmsg_data={ee_errno=0,
+                                    ee_origin=SO_EE_ORIGIN_ZEROCOPY,
+                                    ee_type=0,
+                                    ee_code=SO_EE_CODE_ZEROCOPY_COPIED,
+                                    ee_info=1,
+                                    ee_data=1}}
+                   ]}, MSG_ERRQUEUE) = 0