netdev - Re: [PATCH net 2/2] selftests: net: simple selftest for ipvtap

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <16fa929d-db04-48f5-a350-a31d23f8327c@redhat.com>
Date: Sun, 28 Dec 2025 13:14:09 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: Dmitry Skorodumov <dskr99@...il.com>, netdev@...r.kernel.org,
 Simon Horman <horms@...nel.org>, linux-kernel@...r.kernel.org,
 linux-kselftest@...r.kernel.org
Cc: Dmitry Skorodumov <skorodumov.dmitry@...wei.com>,
 "David S. Miller" <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
 Jakub Kicinski <kuba@...nel.org>, Shuah Khan <shuah@...nel.org>
Subject: Re: [PATCH net 2/2] selftests: net: simple selftest for ipvtap

On 12/25/25 7:55 PM, Dmitry Skorodumov wrote:
> diff --git a/tools/testing/selftests/net/ipvtap_test.sh b/tools/testing/selftests/net/ipvtap_test.sh
> new file mode 100755
> index 000000000000..751793f26fed
> --- /dev/null
> +++ b/tools/testing/selftests/net/ipvtap_test.sh
> @@ -0,0 +1,168 @@
> +#!/bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +#
> +# Simple tests for ipvtap
> +
> +
> +#
> +# The testing environment looks this way:
> +#
> +# |------HOST------|     |------PHY-------|
> +# |      veth<----------------->veth      |
> +# |------|--|------|     |----------------|
> +#        |  |
> +#        |  |            |-----TST0-------|
> +#        |  |------------|----ipvlan      |
> +#        |               |----------------|
> +#        |
> +#        |               |-----TST1-------|
> +#        |---------------|----ipvlan      |
> +#                        |----------------|
> +#
> +
> +ALL_TESTS="
> +	test_ip_set
> +"
> +
> +source lib.sh
> +
> +DEBUG=0
> +
> +VETH_HOST=vethtst.h
> +VETH_PHY=vethtst.p
> +
> +NS_COUNT=32
> +IP_ITERATIONS=1024
> +
> +ns_run() {
> +	ns=$1
> +	shift
> +	if [[ "$ns" == "global" ]]; then
> +		"$@" >/dev/null
> +	else
> +		ip netns exec "$ns" "$@" >/dev/null
> +	fi
> +}
> +
> +test_ip_setup_env() {
> +	modprobe -q tap
> +	modprobe -q ipvlan
> +	modprobe -q ipvtap
> +
> +	setup_ns NS_PHY
> +
> +	# setup simulated other-host (phy) and host itself
> +	ip link add $VETH_HOST type veth peer name $VETH_PHY \
> +		netns "$NS_PHY" >/dev/null

It would be better to avoid creating devices in the main netns.

> +	ip link set $VETH_HOST up
> +	ns_run "$NS_PHY" ip link set $VETH_PHY up
> +
> +	for ((i=0; i<NS_COUNT; i++)); do
> +		setup_ns ipvlan_ns_$i
> +		ns="ipvlan_ns_$i"
> +		if [ "$DEBUG" = "1" ]; then
> +			echo "created NS ${!ns}"
> +		fi
> +		if ! ip link add netns ${!ns} ipvlan0 link $VETH_HOST \
> +		    type ipvtap mode l2 bridge; then
> +			exit_error "FAIL: Failed to configure ipvlan link."
> +		fi
> +	done
> +}
> +
> +test_ip_cleanup_env() {
> +	ip link del $VETH_HOST
> +	cleanup_all_ns
> +}
> +
> +exit_error() {
> +	echo "$1"
> +	exit $ksft_fail
> +}
> +
> +rnd() {
> +	echo $(( RANDOM % 32 + 16 ))
> +}
> +
> +test_ip_set_thread() {
> +	ip link set ipvlan0 up
> +	for ((i=0; i<IP_ITERATIONS; i++)); do
> +		v=$(rnd)
> +		ip a a "172.25.0.$v/24" dev ipvlan0 2>/dev/null
> +		ip a a "fc00::$v/64" dev ipvlan0 2>/dev/null
> +		v=$(rnd)
> +		ip a d "172.25.0.$v/24" dev ipvlan0 2>/dev/null
> +		ip a d "fc00::$v/64" dev ipvlan0 2>/dev/null

It's unclear to me why the above tries to remove random addresses
different from the ones just added (possibly not existing)

> +	done
> +}
> +
> +test_ip_set() {
> +	RET=0
> +
> +	modprobe -q tap
> +	modprobe -q ipvlan
> +	modprobe -q ipvtap
> +
> +	trap test_ip_cleanup_env EXIT
> +
> +	test_ip_setup_env
> +
> +	declare -A ns_pids
> +	for ((i=0; i<NS_COUNT; i++)); do
> +		ns="ipvlan_ns_$i"
> +		ns_run ${!ns} bash -c "$0 test_ip_set_thread"&
> +		ns_pids[$i]=$!
> +	done
> +
> +	for ((i=0; i<NS_COUNT; i++)); do
> +		wait "${ns_pids[$i]}"
> +	done

This tests fails quite often in debug build due to timeout, see:

https://netdev.bots.linux.dev/flakes.html?tn-needle=ipvtap

and i.e.

https://netdev-ctrl.bots.linux.dev/logs/vmksft/net-dbg/results/447821/5-ipvtap-test-sh/

You should likely decrease the thread and/or iterations count, at least
for debug builds

/P