@@ -8,10 +8,101 @@
# runner blocked (and silent) until they finish, so ptest-runner kills the
# whole suite. Skip it.
#
+# ptrace:vmaccess and tty:tty_tiocsti_test check for kernel changes that are
+# not in mainline: PTRACE_ATTACH failing with EAGAIN during a multi-threaded
+# exec instead of blocking (the attach test always times out), and TIOCSTI
+# checking the credentials of the file opener rather than the caller. Skip
+# them.
+#
+# landlock:fs_bench only reports timings, and walking its 10000 nested
+# directories takes about 90 seconds in qemu, twice the default timeout.
+SKIP="-S ftrace:ftracetest-ktap -S filesystems:file_stressor \
+ -S ptrace:vmaccess -S tty:tty_tiocsti_test -S landlock:fs_bench"
+
+# sysmap_is_sealed can't tell at runtime whether the kernel seals the system
+# mappings, so skip it when the running kernel's config says it doesn't.
+if [ -r /proc/config.gz ] && \
+ zcat /proc/config.gz | grep -q '^# CONFIG_MSEAL_SYSTEM_MAPPINGS is not set'; then
+ SKIP="$SKIP -S mseal_system_mappings:sysmap_is_sealed"
+fi
+
+# Tests that fail, rather than skip, when the kernel lacks a feature they
+# need (the options are in each collection's "config" file). Skip a test when
+# the running kernel has any of its options neither built in nor as a module.
+NEEDS_CONFIG="
+landlock:audit_test AUDIT
+landlock:fs_test AUDIT
+landlock:net_test AUDIT
+landlock:ptrace_test AUDIT
+landlock:scoped_abstract_unix_test AUDIT
+net:bridge_vlan_dump.sh BRIDGE_VLAN_FILTERING
+net:busy_poll_test.sh NETDEVSIM
+net:cmsg_so_mark.sh IPV6_MULTIPLE_TABLES
+net:cmsg_so_priority.sh NET_ACT_GACT
+net:cmsg_time.sh NET_SCH_ETF
+net:ecmp_rehash.sh NET_ACT_GACT
+net:fdb_notify.sh MACVLAN VXLAN
+net:fib_nexthops.sh VXLAN
+net:fib_rule_tests.sh IPV6_MULTIPLE_TABLES NET_VRF
+net:fq_band_pktlimit.sh NET_SCH_FQ
+net:gre_gso.sh IPV6_GRE
+net:gre_ipv6_lladdr.sh IPV6_MULTIPLE_TABLES
+net:icmp_redirect.sh NET_VRF
+net:ip_local_port_range.sh MPTCP
+net:ipvtap_test.sh IPVTAP
+net:l2tp.sh L2TP_V3 L2TP_ETH L2TP_IP
+net:link_netns.py IPVLAN IPV6_GRE MACSEC MACVLAN MACVTAP NETKIT
+net:netdev-l2addr.sh NETDEVSIM
+net:netns-name.sh NETDEVSIM
+net:nk_qlease.py NETDEVSIM NETKIT
+net:nl_netdev.py NETDEVSIM
+net:protodown.sh MACVLAN VXLAN
+net:rtnetlink.sh IPV6_GRE MACSEC MACVLAN NET_VRF VXLAN
+net:srv6_end_dx4_netfilter_test.sh LWTUNNEL
+net:srv6_end_dx6_netfilter_test.sh LWTUNNEL
+net:tap MACVTAP
+net:test_bpf.sh TEST_BPF
+net:test_neigh.sh NET_ACT_GACT NET_CLS_MATCHALL
+net:test_vxlan_fdb_changelink.sh VXLAN
+net:test_vxlan_vnifilter_notify.sh VXLAN
+net:test_vxlan_vnifiltering.sh VXLAN
+net:tfo_passive.sh NETDEVSIM
+net:tls TLS
+net:tun GENEVE
+net:txtimestamp.sh IFB
+net:udpgro_frglist.sh NET_CLS_BPF
+net:vlan_bridge_binding.sh BRIDGE_VLAN_FILTERING
+net:vrf-xfrm-tests.sh NET_VRF XFRM_INTERFACE
+net:vrf_route_leaking.sh NET_VRF
+"
+if [ -r /proc/config.gz ]; then
+ KCONFIG=$(zcat /proc/config.gz)
+ SKIP="$SKIP $(echo "$NEEDS_CONFIG" | while read -r test options; do
+ for option in $options; do
+ if ! echo "$KCONFIG" | grep -q "^CONFIG_$option=[ym]"; then
+ echo "-S $test"
+ break
+ fi
+ done
+ done)"
+fi
+
+# udpgro.sh and udpgro_fwd.sh set up NAT with iptables, which fails when it
+# uses the legacy backend and the kernel only has nftables.
+if ! iptables -t nat -n -L >/dev/null 2>&1; then
+ SKIP="$SKIP -S net:udpgro.sh -S net:udpgro_fwd.sh"
+fi
+
+# tc-testing's tdc.py needs scapy (in meta-networking) and exits at once
+# without it.
+if ! python3 -c 'import scapy' >/dev/null 2>&1; then
+ SKIP="$SKIP -S tc-testing:tdc.sh"
+fi
+
# Turn the runner's per-test TAP lines ("ok 3 selftests: cgroup: test_core",
# "not ok 4 ... # exit=1", "ok 5 ... # SKIP") into PASS:/FAIL:/SKIP: lines
# that ptest log parsers understand, keeping the original output too.
-/usr/kernel-selftest/run_kselftest.sh -s -S ftrace:ftracetest-ktap -S filesystems:file_stressor 2>&1 | awk '
+/usr/kernel-selftest/run_kselftest.sh -s $SKIP 2>&1 | awk '
{ print; fflush() }
/^(not )?ok [0-9]+ selftests: / {
res = "PASS"
Many tests, most of them in net, fail rather than skip when the kernel lacks a feature they need (vxlan, geneve, macvlan, vrf, netdevsim, TLS, tc actions and qdiscs, audit for the landlock tests, ...). Skip them based on the running kernel's config, so the results show regressions instead of the kernel configuration. Also skip the udpgro tests when iptables can't set up NAT (legacy iptables on a kernel with only nftables), tc-testing when scapy is not installed, landlock's fs_bench, which only reports timings and needs twice the default timeout in qemu, mseal's sysmap_is_sealed when the kernel doesn't seal the system mappings, and the ptrace vmaccess and tty_tiocsti tests, which check for kernel changes that are not in mainline. AI-Generated: Uses Claude Code Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com> --- .../kernel-selftest/kernel-selftest/run-ptest | 93 ++++++++++++++++++- 1 file changed, 92 insertions(+), 1 deletion(-)