[PATCH bpf-next] selftests, bpf: Skip MPLS test_tc_tunnel tests if MPLS is unavailable

[PATCH bpf-next] selftests, bpf: Skip MPLS test_tc_tunnel tests if MPLS is unavailable

[Simon Horman]

If MPLS is not available in the kernel then skip MPLS tests.

This avoids the test failing in situations where the test is not
supported by the underlying kernel.

In the case where all tests are run, just skip over the MPLS tests
without altering the exit code of the overall test run - there
is only one exit code in this scenario.

In the case where a single test is run, exit with KSFT_SKIP (4).

In both cases log an informative message.

Signed-off-by: Simon Horman <horms@kernel.org>
---
 tools/testing/selftests/bpf/test_tc_tunnel.sh | 22 ++++++++++++++++++++--
 1 file changed, 20 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/bpf/test_tc_tunnel.sh b/tools/testing/selftests/bpf/test_tc_tunnel.sh
index 7989ec608454..71cddabc4ade 100755
--- a/tools/testing/selftests/bpf/test_tc_tunnel.sh
+++ b/tools/testing/selftests/bpf/test_tc_tunnel.sh
@@ -102,6 +102,20 @@ wait_for_port() {
 	return 1
 }
 
+skip_mac() {
+	if [ "$1" = "mpls" ]; then
+		modprobe mpls_iptunnel || true
+		modprobe mpls_gso || true
+
+		if [ ! -e /proc/sys/net/mpls/platform_labels ]; then
+			echo -e "skip:    mpls tunnel not supported by kernel\n"
+			return # true
+		fi
+	fi
+
+	false
+}
+
 set -e
 
 # no arguments: automated test, run all
@@ -125,6 +139,8 @@ if [[ "$#" -eq "0" ]]; then
 	$0 ipv6 ip6vxlan eth 2000
 
 	for mac in none mpls eth ; do
+		! skip_mac "$mac" || continue
+
 		echo "ip gre $mac"
 		$0 ipv4 gre $mac 100
 
@@ -193,6 +209,10 @@ readonly tuntype=$2
 readonly mac=$3
 readonly datalen=$4
 
+if skip_mac "$mac"; then
+	exit 4 # KSFT_SKIP=4
+fi
+
 echo "encap ${addr1} to ${addr2}, type ${tuntype}, mac ${mac} len ${datalen}"
 
 trap cleanup EXIT
@@ -278,8 +298,6 @@ elif [[ "$tuntype" =~ (gre|vxlan) && "$mac" == "eth" ]]; then
 		  awk '/ether/ { print $2 }')
 	ip netns exec "${ns2}" ip link set testtun0 address $ethaddr
 elif [[ "$mac" == "mpls" ]]; then
-	modprobe mpls_iptunnel ||true
-	modprobe mpls_gso ||true
 	ip netns exec "${ns2}" sysctl -qw net.mpls.platform_labels=65536
 	ip netns exec "${ns2}" ip -f mpls route add 1000 dev lo
 	ip netns exec "${ns2}" ip link set lo up

Re: [PATCH bpf-next] selftests, bpf: Skip MPLS test_tc_tunnel tests if MPLS is unavailable

[Daniel Borkmann]

[ +Alexis ]

On 9/27/24 2:05 PM, Simon Horman wrote:
> If MPLS is not available in the kernel then skip MPLS tests.
>
> This avoids the test failing in situations where the test is not
> supported by the underlying kernel.
>
> In the case where all tests are run, just skip over the MPLS tests
> without altering the exit code of the overall test run - there
> is only one exit code in this scenario.
>
> In the case where a single test is run, exit with KSFT_SKIP (4).
>
> In both cases log an informative message.
>
> Signed-off-by: Simon Horman <horms@kernel.org>

Are you running this as part of net selftests / CI which is why you
need this? (And if yes, why excluding MPLS?)

Alexis is working on converting tests like these into our BPF CI into
test_progs, I'm not sure whether we need to make that more
complex unless there is a compelling reason to exclude MPLS?

Thanks,
Daniel

> ---
>   tools/testing/selftests/bpf/test_tc_tunnel.sh | 22 ++++++++++++++++++++--
>   1 file changed, 20 insertions(+), 2 deletions(-)
>
> diff --git a/tools/testing/selftests/bpf/test_tc_tunnel.sh b/tools/testing/selftests/bpf/test_tc_tunnel.sh
> index 7989ec608454..71cddabc4ade 100755
> --- a/tools/testing/selftests/bpf/test_tc_tunnel.sh
> +++ b/tools/testing/selftests/bpf/test_tc_tunnel.sh
> @@ -102,6 +102,20 @@ wait_for_port() {
>   	return 1
>   }
>   
> +skip_mac() {
> +	if [ "$1" = "mpls" ]; then
> +		modprobe mpls_iptunnel || true
> +		modprobe mpls_gso || true
> +
> +		if [ ! -e /proc/sys/net/mpls/platform_labels ]; then
> +			echo -e "skip:    mpls tunnel not supported by kernel\n"
> +			return # true
> +		fi
> +	fi
> +
> +	false
> +}
> +
>   set -e
>   
>   # no arguments: automated test, run all
> @@ -125,6 +139,8 @@ if [[ "$#" -eq "0" ]]; then
>   	$0 ipv6 ip6vxlan eth 2000
>   
>   	for mac in none mpls eth ; do
> +		! skip_mac "$mac" || continue
> +
>   		echo "ip gre $mac"
>   		$0 ipv4 gre $mac 100
>   
> @@ -193,6 +209,10 @@ readonly tuntype=$2
>   readonly mac=$3
>   readonly datalen=$4
>   
> +if skip_mac "$mac"; then
> +	exit 4 # KSFT_SKIP=4
> +fi
> +
>   echo "encap ${addr1} to ${addr2}, type ${tuntype}, mac ${mac} len ${datalen}"
>   
>   trap cleanup EXIT
> @@ -278,8 +298,6 @@ elif [[ "$tuntype" =~ (gre|vxlan) && "$mac" == "eth" ]]; then
>   		  awk '/ether/ { print $2 }')
>   	ip netns exec "${ns2}" ip link set testtun0 address $ethaddr
>   elif [[ "$mac" == "mpls" ]]; then
> -	modprobe mpls_iptunnel ||true
> -	modprobe mpls_gso ||true
>   	ip netns exec "${ns2}" sysctl -qw net.mpls.platform_labels=65536
>   	ip netns exec "${ns2}" ip -f mpls route add 1000 dev lo
>   	ip netns exec "${ns2}" ip link set lo up
>
>

Re: [PATCH bpf-next] selftests, bpf: Skip MPLS test_tc_tunnel tests if MPLS is unavailable

[Simon Horman]

On Wed, Oct 02, 2024 at 02:07:54PM +0200, Daniel Borkmann wrote:
> [ +Alexis ]
> 
> On 9/27/24 2:05 PM, Simon Horman wrote:
> > If MPLS is not available in the kernel then skip MPLS tests.
> > 
> > This avoids the test failing in situations where the test is not
> > supported by the underlying kernel.
> > 
> > In the case where all tests are run, just skip over the MPLS tests
> > without altering the exit code of the overall test run - there
> > is only one exit code in this scenario.
> > 
> > In the case where a single test is run, exit with KSFT_SKIP (4).
> > 
> > In both cases log an informative message.
> > 
> > Signed-off-by: Simon Horman <horms@kernel.org>
> 
> Are you running this as part of net selftests / CI which is why you
> need this? (And if yes, why excluding MPLS?)

Hi Daniel,

No, this was observed when running the test in a different harness.
I can find out why MPLS is excluded, but my assumption is because
it is not otherwise needed by the distro kernel that is being tested.

I entirely understand if that makes this patch out of scope for upstream.
Although having it upstream would be useful downstream.

> Alexis is working on converting tests like these into our BPF CI into
> test_progs, I'm not sure whether we need to make that more
> complex unless there is a compelling reason to exclude MPLS?

If the MPLS test(s) can somehow be excluded then that may meet the needs of
the test environment where this was observed, but I would need to check.

...