[RFC net-next v2 1/2] netpoll: support sending over raw IP interfaces

[RFC net-next v2 1/2] netpoll: support sending over raw IP interfaces

[Mark Cilissen]

Currently, netpoll only supports interfaces with an ethernet-compatible
link layer. Certain interfaces like SLIP do not have a link layer
on the network interface level at all and expect raw IP packets,
and could benefit from being supported by netpoll.

This commit adds support for such interfaces by checking if a link layer
header is present using `dev_has_header()`. If that is not the case,
it simply skips adding the ethernet header, causing a raw IP packet to be
sent over the interface. This has been confirmed to add netconsole support
to at least SLIP and WireGuard interfaces.

Signed-off-by: Mark Cilissen <mark@yotsuba.nl>
---
v2:
- Use dev_has_header(netdev) to detect existence of link layer header
  instead of netdev->hard_header_len, restore prior skb allocation sizes
  by always using LL_RESERVED_SPACE() as before
- Add selftest for netconsole

v1: https://lore.kernel.org/netdev/20240313124613.51399-1-mark@yotsuba.nl/T/
---
 Documentation/networking/netconsole.rst |  3 ++-
 net/core/netpoll.c                      | 15 ++++++++-------
 2 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/Documentation/networking/netconsole.rst b/Documentation/networking/netconsole.rst
index d55c2a22ec7a..434ce0366027 100644
--- a/Documentation/networking/netconsole.rst
+++ b/Documentation/networking/netconsole.rst
@@ -327,4 +327,5 @@ enable the logging of even the most critical kernel bugs. It works
 from IRQ contexts as well, and does not enable interrupts while
 sending packets. Due to these unique needs, configuration cannot
 be more automatic, and some fundamental limitations will remain:
-only IP networks, UDP packets and ethernet devices are supported.
+only UDP packets over IP networks, over ethernet links if a
+hardware layer is required, are supported.
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index 543007f159f9..d12aa1b7d79e 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -458,9 +458,7 @@ void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
 		ip6h->saddr = np->local_ip.in6;
 		ip6h->daddr = np->remote_ip.in6;
 
-		eth = skb_push(skb, ETH_HLEN);
-		skb_reset_mac_header(skb);
-		skb->protocol = eth->h_proto = htons(ETH_P_IPV6);
+		skb->protocol = htons(ETH_P_IPV6);
 	} else {
 		udph->check = 0;
 		udph->check = csum_tcpudp_magic(np->local_ip.ip,
@@ -487,14 +485,17 @@ void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
 		put_unaligned(np->remote_ip.ip, &(iph->daddr));
 		iph->check    = ip_fast_csum((unsigned char *)iph, iph->ihl);
 
+		skb->protocol = htons(ETH_P_IP);
+	}
+
+	if (dev_has_header(np->dev)) {
 		eth = skb_push(skb, ETH_HLEN);
 		skb_reset_mac_header(skb);
-		skb->protocol = eth->h_proto = htons(ETH_P_IP);
+		eth->h_proto = skb->protocol;
+		ether_addr_copy(eth->h_source, np->dev->dev_addr);
+		ether_addr_copy(eth->h_dest, np->remote_mac);
 	}
 
-	ether_addr_copy(eth->h_source, np->dev->dev_addr);
-	ether_addr_copy(eth->h_dest, np->remote_mac);
-
 	skb->dev = np->dev;
 
 	netpoll_send_skb(np, skb);
-- 
2.44.0

[RFC net-next v2 2/2] selftests: add netpoll selftest

[Mark Cilissen]

Now that netpoll supports more kinds of network interfaces it benefits
extra from being tested to monitor regressions. This test runs entirely in
userspace, adding a netconsole for the interface-under-test using configfs
and sending a kernel log message that is verified using socat.

To cover a mix of circumstances while keeping host tooling dependences
fairly minimal, the following interface scenarios are tested:
- Ethernet link layer, IPv4 and IPv6 (veth)
- No exposed link layer, IPv4 (gre, ipip, ip6tnl)
- No exposed link layer, IPv6 (gre, sit, ip6tnl)

As a sanity check the test was run on kernel release 6.7.9-200.fc39,
which does not include the expanded netpoll interface support,
and yielded the expected failures:

  PASS: netconsole (eth/ipv4)
  PASS: netconsole (eth/ipv6)
  FAIL: netconsole (ipv4/gre/ipv4): message did not arrive in time (124)
  FAIL: netconsole (ipv4/gre/ipv6): message did not arrive in time (124)
  FAIL: netconsole (ipv6/ip6gre/ipv4): message did not arrive in time (124)
  FAIL: netconsole (ipv6/ip6gre/ipv6): message did not arrive in time (124)
  FAIL: netconsole (ipv4/ipip/ipv4): message did not arrive in time (124)
  FAIL: netconsole (ipv4/sit/ipv6): message did not arrive in time (124)
  FAIL: netconsole (ipv6/ip6tnl/ipv4): message did not arrive in time (124)
  FAIL: netconsole (ipv6/ip6tnl/ipv6): message did not arrive in time (124)

Signed-off-by: Mark Cilissen <mark@yotsuba.nl>
---
 tools/testing/selftests/net/Makefile   |   1 +
 tools/testing/selftests/net/config     |   1 +
 tools/testing/selftests/net/netpoll.sh | 226 +++++++++++++++++++++++++
 3 files changed, 228 insertions(+)
 create mode 100755 tools/testing/selftests/net/netpoll.sh

diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
index 7b6918d5f4af..e9908a976f5e 100644
--- a/tools/testing/selftests/net/Makefile
+++ b/tools/testing/selftests/net/Makefile
@@ -93,6 +93,7 @@ TEST_PROGS += test_bridge_backup_port.sh
 TEST_PROGS += fdb_flush.sh
 TEST_PROGS += fq_band_pktlimit.sh
 TEST_PROGS += vlan_hw_filter.sh
+TEST_PROGS += netpoll.sh
 
 TEST_FILES := settings
 TEST_FILES += in_netns.sh lib.sh net_helper.sh setup_loopback.sh setup_veth.sh
diff --git a/tools/testing/selftests/net/config b/tools/testing/selftests/net/config
index 5e4390cac17e..bc3b6991b1aa 100644
--- a/tools/testing/selftests/net/config
+++ b/tools/testing/selftests/net/config
@@ -100,3 +100,4 @@ CONFIG_NETFILTER_XT_MATCH_POLICY=m
 CONFIG_CRYPTO_ARIA=y
 CONFIG_XFRM_INTERFACE=m
 CONFIG_XFRM_USER=m
+CONFIG_NETCONSOLE=m
diff --git a/tools/testing/selftests/net/netpoll.sh b/tools/testing/selftests/net/netpoll.sh
new file mode 100755
index 000000000000..ab4acd756ca1
--- /dev/null
+++ b/tools/testing/selftests/net/netpoll.sh
@@ -0,0 +1,226 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+#
+# Test netpoll and netconsole in various interface scenarios:
+# ethernet (veth), raw IPv4 (gre, ipip, ip6tnl), raw IPv6 (gre, sit, ip6tnl)
+#
+# +-----------------------------------------+
+# | $NS1                                    |
+# |              $H1 +---+ $T1              |
+# |     192.0.2.1/30 |     192.0.2.65/30    |
+# | 2001:db8:1::1/64 |     2001:db8:2::1/64 |
+# +------------------|----------------------+
+#                    |
+# +------------------|----------------------+
+# | $NS2             |                      |
+# |              $H2 +---+ $T2              |
+# |     192.0.2.2/30       192.0.2.66/30    |
+# | 2001:db8:1::2/64       2001:db8:2::2/64 |
+# +-----------------------------------------+
+# shellcheck disable=SC3043 # allow 'local' keyword
+set -e
+
+readonly ksft_skip=4
+readonly H1A4=192.0.2.1/30
+readonly H1A6=2001:db8:1::1/64
+readonly H2A4=192.0.2.2/30
+readonly H2A6=2001:db8:1::2/64
+readonly T1A4=192.0.2.65/30
+readonly T1A6=2001:db8:2::1/64
+readonly T2A4=192.0.2.66/30
+readonly T2A6=2001:db8:2::2/64
+
+# shellcheck disable=SC2155 # declare and assign separately
+readonly BASE="ksft-$(mktemp -u XXXXXX)"
+readonly NS1="$BASE-n1"
+readonly NS2="$BASE-n2"
+readonly H1="$BASE-h1"
+readonly H2="$BASE-h2"
+readonly T1="$BASE-i1"
+readonly T2="$BASE-i2"
+
+
+# Utilities
+
+setup() {
+	ip netns add "$NS1"
+	ip netns exec "$NS1" sysctl -wq net.ipv4.ip_forward=1
+	ip netns exec "$NS1" sysctl -wq net.ipv6.conf.all.forwarding=1
+	ip netns exec "$NS1" sysctl -wq net.ipv6.conf.default.forwarding=1
+
+	ip netns add "$NS2"
+	ip netns exec "$NS2" sysctl -wq net.ipv4.ip_forward=1
+	ip netns exec "$NS2" sysctl -wq net.ipv6.conf.all.forwarding=1
+	ip netns exec "$NS2" sysctl -wq net.ipv6.conf.default.forwarding=1
+
+	ip link add "$H1" type veth peer name "$H2"
+
+	ip link set dev "$H1" netns "$NS1"
+	ip -n "$NS1" link set dev "$H1" up
+	ip -n "$NS1" addr add "$H1A4" dev "$H1"
+	ip -n "$NS1" addr add "$H1A6" dev "$H1" nodad
+
+	ip link set dev "$H2" netns "$NS2"
+	ip -n "$NS2" link set dev "$H2" up
+	ip -n "$NS2" addr add "$H2A4" dev "$H2"
+	ip -n "$NS2" addr add "$H2A6" dev "$H2" nodad
+
+	grep -q '^netcon' /proc/consoles || modprobe netconsole
+}
+
+cleanup() {
+	set +e
+	for d in /sys/kernel/config/netconsole/"$BASE"*; do
+		[ -d "$d" ] || continue
+		rmdir "$d"
+	done
+	for ns in "$NS1" "$NS2"; do
+		if ip -n "$ns" link show >/dev/null 2>&1; then
+			ip netns del "$ns"
+		fi
+	done
+	# In case we exit before the veth got moved into its namespace
+	if [ -e /sys/class/net/"$H1" ]; then
+		ip link del dev "$H1"
+	fi
+}
+
+trap cleanup EXIT
+
+if ! command -v socat >/dev/null 2>&1; then
+	echo "SKIP: netconsole tests need socat"
+	exit $ksft_skip
+fi
+
+
+test_netconsole() {
+	local name="$1"; shift
+	local outif="$1"; shift
+	local outip="$1"; shift
+	local inif="$1"; shift
+	local inip="$1"; shift
+	local ret=0
+
+	local msg="kselftest/net/netpoll: netconsole test message for $name"
+	# Setup logger and background it, it will exit when it receives the message
+	timeout 3s ip netns exec "$NS2" socat -u -S 0 \
+		udp6-recv:6666,reuseaddr,so-bindtodevice="$inif",ipv6only=0 \
+		system:"grep -m 1 -q \'$(printf '%s' "$msg" | sed -e 's/:/\\:/g')\$\'" &
+	# Setup netconsole
+	ip netns exec "$NS1" sh -s <<EOF
+set -e
+mount -t configfs none /sys/kernel/config
+cd /sys/kernel/config/netconsole
+mkdir "$outif" && cd "$outif"
+echo "$outif" > dev_name
+echo "${outip%%/*}" > local_ip
+echo "${inip%%/*}" > remote_ip
+echo 6666 > remote_port
+echo 1 > enabled
+EOF
+
+	# Send message and wait for it to arrive or timeout
+	echo "<1>$msg" > /dev/kmsg
+	if wait %+; then
+		echo "PASS: netconsole ($name)"
+	else
+		ret=$?
+		echo "FAIL: netconsole ($name): message did not arrive in time ($err)"
+	fi
+	rmdir /sys/kernel/config/netconsole/"$outif"
+	return $ret
+}
+
+
+# Test starts here
+
+ret=0
+mark_failed() {
+	# Make sure failure isn't overwritten by skip
+	if [ "$ret" = 0 ] || [ "$ret" = $ksft_skip ]; then
+		ret="$1"
+	fi
+}
+
+setup
+
+# Test ethernet interfaces
+
+test_netconsole eth/ipv4 "$H1" "$H1A4" "$H2" "$H2A4" || mark_failed $?
+test_netconsole eth/ipv6 "$H1" "$H1A6" "$H2" "$H2A6" || mark_failed $?
+
+# Test tunnel interfaces
+
+setup_tunnel() {
+	local type="$1"; shift
+	local h1a="${1%%/*}"; shift
+	local h2a="${1%%/*}"; shift
+	local t1a="$1"; shift
+	local t2a="$1"; shift
+
+	ip -n "$NS1" link add "$T1" type "$type" local "$h1a" remote "$h2a" "$@" \
+		|| return $ksft_skip # Skip test if tunnel type is not available
+	ip -n "$NS1" addr add "$t1a" dev "$T1"
+	ip -n "$NS1" link set "$T1" up
+
+	ip -n "$NS2" link add "$T2" type "$type" local "$h2a" remote "$h1a" "$@" \
+		|| return $ksft_skip # Skip test if tunnel type is not available
+	ip -n "$NS2" addr add "$t2a" dev "$T2"
+	ip -n "$NS2" link set "$T2" up
+}
+
+teardown_tunnel() {
+	ip -n "$NS1" link del dev "$T1"
+	ip -n "$NS2" link del dev "$T2"
+}
+
+test_tunnel_4in4() {
+	local type="$1"; shift
+	local ret=0
+	setup_tunnel "$type" "$H1A4" "$H2A4" "$T1A4" "$T2A4" "$@"
+	test_netconsole "ipv4/$type/ipv4" "$T1" "$T1A4" "$T2" "$T2A4" || ret=$?
+	teardown_tunnel
+	return $ret
+}
+
+test_tunnel_4in6() {
+	local type="$1"; shift
+	local ret=0
+	setup_tunnel "$type" "$H1A6" "$H2A6" "$T1A4" "$T2A4" "$@"
+	test_netconsole "ipv6/$type/ipv4" "$T1" "$T1A4" "$T2" "$T2A4" || ret=$?
+	teardown_tunnel
+	return $ret
+}
+
+test_tunnel_6in4() {
+	local type="$1"; shift
+	local ret=0
+	setup_tunnel "$type" "$H1A4" "$H2A4" "$T1A6" "$T2A6" "$@"
+	test_netconsole "ipv4/$type/ipv6" "$T1" "$T1A6" "$T2" "$T2A6" || ret=$?
+	teardown_tunnel
+	return $ret
+}
+
+test_tunnel_6in6() {
+	local type="$1"; shift
+	local ret=0
+	setup_tunnel "$type" "$H1A6" "$H2A6" "$T1A6" "$T2A6" "$@"
+	test_netconsole "ipv6/$type/ipv6" "$T1" "$T1A6" "$T2" "$T2A6" || ret=$?
+	teardown_tunnel
+	return $ret
+}
+
+test_tunnel_4in4 gre    || mark_failed $?
+test_tunnel_6in4 gre    || mark_failed $?
+test_tunnel_4in6 ip6gre || mark_failed $?
+test_tunnel_6in6 ip6gre || mark_failed $?
+test_tunnel_4in4 ipip   || mark_failed $?
+test_tunnel_6in4 sit    || mark_failed $?
+test_tunnel_4in6 ip6tnl mode ipip6  || mark_failed $?
+test_tunnel_6in6 ip6tnl mode ip6ip6 || mark_failed $?
+
+
+# Test done
+
+cleanup
+exit "$ret"
-- 
2.44.0