On 6/24/2023 4:18 AM, Ilya Maximets wrote:
On 6/19/23 07:05, Chris Mi wrote:Add three sFlow offload test cases:3: offloads - sflow with sampling=1 - offloads enabled ok 4: offloads - sflow with sampling=2 - offloads enabled ok 5: offloads - ping over vxlan tunnel with sflow - offloads enabled ok Signed-off-by: Chris Mi <c...@nvidia.com> Reviewed-by: Roi Dayan <r...@nvidia.com> Acked-by: Eelco Chaudron <echau...@redhat.com> --- Documentation/howto/tc-offload.rst | 24 +++++ tests/system-common-macros.at | 13 +++ tests/system-offloads-traffic.at | 160 +++++++++++++++++++++++++++++ 3 files changed, 197 insertions(+) diff --git a/Documentation/howto/tc-offload.rst b/Documentation/howto/tc-offload.rst index 681dff13e..a50dc3c58 100644 --- a/Documentation/howto/tc-offload.rst +++ b/Documentation/howto/tc-offload.rst @@ -88,6 +88,30 @@ As you can see above the none-offload case reports 140 bytes more, which is 14 bytes per packet. This represents the L2 header, in this case, 2 * *Ethernet address* + *Ethertype*.+Tunnel offload+++++++++++++++ + +Current tunnel offload ignores DF and CSUM flags configuration requested by +the user. TC for now has no way to pass these flags in a flower key and their +masks are set by default. To make tunnel offload work, DF and CSUM flags +are cleared. So please be aware of the following differences. + +Dumping vxlan decap match without offload, it shows:: + + recirc_id(0),tunnel(tun_id=0x4,src=192.168.1.1,dst=192.168.1.2,flags(-df+csum+key)),in_port(vxlan_sys_4789) + +Dumping vxlan decap match with offload, it shows:: + + recirc_id(0),tunnel(tun_id=0x4,src=192.168.1.1,dst=192.168.1.2,tp_dst=4789,flags(+key)),in_port(vxlan_sys_4789) + +Dumping vxlan encap action without offload, it shows:: + + actions:set(tunnel(tun_id=0x4,dst=192.168.1.1,ttl=64,tp_dst=4789,flags(df|key))),vxlan_sys_4789 + +Dumping vxlan encap action with offload, it shows:: + + actions:set(tunnel(tun_id=0x4,dst=192.168.1.64,ttl=64,tp_dst=4789,flags(key))),vxlan_sys_4789 + TC Meter Offload ~~~~~~~~~~~~~~~~diff --git a/tests/system-common-macros.at b/tests/system-common-macros.atindex 0077a8609..55ec0bf1e 100644 --- a/tests/system-common-macros.at +++ b/tests/system-common-macros.at @@ -359,3 +359,16 @@ m4_define([OVS_CHECK_IPROUTE_ENCAP], # OVS_CHECK_CT_CLEAR() m4_define([OVS_CHECK_CT_CLEAR], [AT_SKIP_IF([! grep -q "Datapath supports ct_clear action" ovs-vswitchd.log])]) + +# LOAD_MODULE([name]) +# +# Tries to load specified kernel module and removes it after +# if it wasn't loaded before this call. +# +m4_define([LOAD_MODULE], + [if ! lsmod | grep -q $1; then + on_exit 'modprobe -q -r $1' + fi + AT_CHECK([modprobe $1]) + ] +) \ No newline at end of file diff --git a/tests/system-offloads-traffic.at b/tests/system-offloads-traffic.at index ae302a294..db287a86d 100644 --- a/tests/system-offloads-traffic.at +++ b/tests/system-offloads-traffic.at @@ -93,6 +93,166 @@ AT_CHECK([ovs-appctl upcall/show | grep -E "offloaded flows : [[1-9]]"], [0], [i OVS_TRAFFIC_VSWITCHD_STOP AT_CLEANUP++AT_SETUP([offloads - sflow with sampling=1 - offloads enabled]) +LOAD_MODULE([psample]) +OVS_TRAFFIC_VSWITCHD_START([], [], [-- set Open_vSwitch . other_config:hw-offload=true]) + +on_exit 'kill `cat test-sflow.pid`' +AT_CHECK([ovstest test-sflow --log-file --detach --no-chdir --pidfile 0:127.0.0.1 > sflow.log], [0], [], [ignore])Please, wrap long lines. At least move the error codes and output streams to a new line. Indented to the first argument, of course. It's also possible to break the argument with either 'dnl' or a line continuation '\' (new versions of atotest seems to handle them better).
Done.
+AT_CAPTURE_FILE([sflow.log]) +PARSE_LISTENING_PORT([test-sflow.log], [SFLOW_PORT]) + +AT_CHECK([ovs-ofctl add-flow br0 "actions=normal"]) + +ADD_NAMESPACES(at_ns0, at_ns1) +ADD_VETH(p0, at_ns0, br0, "10.1.1.1/24") +ADD_VETH(p1, at_ns1, br0, "10.1.1.2/24") + +AT_CHECK([ovs-vsctl -- --id=@sflow create sflow agent=lo target=\"127.0.0.1:$SFLOW_PORT\" header=128 sampling=1 polling=100 -- set bridge br0 sflow=@sflow], [0], [ignore])Wrap.
Done.
+NS_CHECK_EXEC([at_ns0], [ping -q -c 1000 -i 0.01 -w 12 10.1.1.2 | FORMAT_PING], [0], [dnl +1000 packets transmitted, 1000 received, 0% packet loss, time 0msThis check fails when I run tests with ASAN. It's unable to finish in 12 seconds: ./system-offloads-traffic.at:113: ip netns exec at_ns0 sh << NS_EXEC_HEREDOC ping -q -c 1000 -i 0.01 -w 12 10.1.1.2 | grep "transmitted" | sed 's/time.*ms$/time 0ms/' NS_EXEC_HEREDOC --- - 2023-06-23 18:53:18.127071711 +0000 +++ /root/ovs/tests/system-offloads-testsuite.dir/at-groups/3/stdout 2023-06-23 18:53:18.120000000 +0000 @@ -1,2 +1,2 @@ -1000 packets transmitted, 1000 received, 0% packet loss, time 0ms +750 packets transmitted, 750 received, 0% packet loss, time 0ms Same for the other ping in these tests.
Done.
+]) + +p1_ifindex=$(cat /sys/class/net/ovs-p1/ifindex)It's probably better to use ip utility to get the ifindex. E.g. p1_ifindex=$(ip link show ovs-p1 | head -1 | cut -d ':' -f 1)+m4_define([DUMP_SFLOW], [sed -e "s/used:[[0-9]].[[0-9]]*s/used:0.001s/;s/eth(src=[[a-z0-9:]]*,dst=[[a-z0-9:]]*)/eth(macs)/;s/pid=[[0-9]]*/pid=1/;s/output=$p1_ifindex/output=1/"]) +AT_CHECK([ovs-appctl dpctl/dump-flows type=tc,offloaded | grep "in_port(2)" | grep "eth_type(0x0800)" | grep "actions:userspace" | grep "sFlow" | DUMP_SFLOW], [0], [dnl +recirc_id(0),in_port(2),eth(macs),eth_type(0x0800),ipv4(frag=no), packets:999, bytes:83916, used:0.001s, actions:userspace(pid=1,sFlow(vid=0,pcp=0,output=1),actions),3 +]) + +p0_ifindex=$(cat /sys/class/net/ovs-p0/ifindex) +m4_define([DUMP_SFLOW], [sed -e "s/used:[[0-9]].[[0-9]]*s/used:0.001s/;s/eth(src=[[a-z0-9:]]*,dst=[[a-z0-9:]]*)/eth(macs)/;s/pid=[[0-9]]*/pid=1/;s/output=$p0_ifindex/output=1/"]) +AT_CHECK([ovs-appctl dpctl/dump-flows type=tc,offloaded | grep "in_port(3)" | grep "eth_type(0x0800)" | grep "actions:userspace" | grep "sFlow" | DUMP_SFLOW], [0], [dnl +recirc_id(0),in_port(3),eth(macs),eth_type(0x0800),ipv4(frag=no), packets:999, bytes:83916, used:0.001s, actions:userspace(pid=1,sFlow(vid=0,pcp=0,output=1),actions),2 +])There is no need to define DUMP_SFLOW multiple times. It's also possible to write sed actions on separate lines. E.g.: m4_define([FORMAT_SFLOW], [sed -e " s/used:[[0-9]].[[0-9]]*s/used:0.001s/; s/eth(src=[[a-z0-9:]]*,dst=[[a-z0-9:]]*)/eth(macs)/; s/pid=[[0-9]]*/pid=1/; s/output=$1/output=1/; "]) p1_ifindex=$(ip link show ovs-p1 | head -1 | cut -d ':' -f 1) AT_CHECK([ovs-appctl dpctl/dump-flows --names type=tc,offloaded \ | grep "in_port(ovs-p0)" | grep "eth_type(0x0800)" | grep "sFlow" \ | FORMAT_SFLOW([$p1_ifindex])], [0], [dnl recirc_id(0),in_port(ovs-p0),eth(macs),eth_type(0x0800),ipv4(frag=no), dnl packets:999, bytes:83916, used:0.001s, dnl actions:userspace(pid=1,sFlow(vid=0,pcp=0,output=1),actions),ovs-p1 ]) next time you may use the same macro, but with a different argument.
Done.
+ +hdr="in_ifindex=$p0_ifindex in_format=0 out_ifindex=$p1_ifindex out_format=0 hdr_prot=1 pkt_len=102 stripped=4 hdr_len=98" +count=`grep "$hdr" sflow.log | wc -l`Pease, use $(...) instead of `...`. And grep can count with -c option.
Done.
+AT_CHECK([[[[ $count -ge 996 ]]]])Use test $count -ge 996 instead of []. There will be no need to escape them. You may also just write in a single line: AT_CHECK([test $(grep -c "$hdr" sflow.log) -ge 996 ]) Same logic is applicable to tests below.
Done.
+hdr="in_ifindex=$p1_ifindex in_format=0 out_ifindex=$p0_ifindex out_format=0 hdr_prot=1 pkt_len=102 stripped=4 hdr_len=98" +count=`grep "$hdr" sflow.log | wc -l` +AT_CHECK([[[[ $count -ge 996 ]]]]) + +OVS_TRAFFIC_VSWITCHD_STOP +OVS_APP_EXIT_AND_WAIT([test-sflow]) +AT_CLEANUP + + +AT_SETUP([offloads - sflow with sampling=2 - offloads enabled]) +LOAD_MODULE([psample]) +OVS_TRAFFIC_VSWITCHD_START([], [], [-- set Open_vSwitch . other_config:hw-offload=true]) + +on_exit 'kill `cat test-sflow.pid`' +AT_CHECK([ovstest test-sflow --log-file --detach --no-chdir --pidfile 0:127.0.0.1 > sflow.log], [0], [], [ignore]) +AT_CAPTURE_FILE([sflow.log]) +PARSE_LISTENING_PORT([test-sflow.log], [SFLOW_PORT]) + +AT_CHECK([ovs-ofctl add-flow br0 "actions=normal"]) + +ADD_NAMESPACES(at_ns0, at_ns1) +ADD_VETH(p0, at_ns0, br0, "10.1.1.1/24") +ADD_VETH(p1, at_ns1, br0, "10.1.1.2/24") + +AT_CHECK([ovs-vsctl -- --id=@sflow create sflow agent=lo target=\"127.0.0.1:$SFLOW_PORT\" header=128 sampling=2 polling=100 -- set bridge br0 sflow=@sflow], [0], [ignore]) +NS_CHECK_EXEC([at_ns0], [ping -q -c 1000 -i 0.01 -w 12 10.1.1.2 | FORMAT_PING], [0], [dnl +1000 packets transmitted, 1000 received, 0% packet loss, time 0ms +]) + +p1_ifindex=$(cat /sys/class/net/ovs-p1/ifindex) +m4_define([DUMP_SFLOW], [sed -e "s/used:[[0-9]].[[0-9]]*s/used:0.001s/;s/eth(src=[[a-z0-9:]]*,dst=[[a-z0-9:]]*)/eth(macs)/;s/pid=[[0-9]]*/pid=1/;s/output=$p1_ifindex/output=1/"]) +AT_CHECK([ovs-appctl dpctl/dump-flows type=tc,offloaded | grep "in_port(2)" | grep "eth_type(0x0800)" | grep "actions:sample" | grep "sFlow" | DUMP_SFLOW], [0], [dnl +recirc_id(0),in_port(2),eth(macs),eth_type(0x0800),ipv4(frag=no), packets:999, bytes:83916, used:0.001s, actions:sample(sample=50.0%,actions(userspace(pid=1,sFlow(vid=0,pcp=0,output=1),actions))),3 +]) + +p0_ifindex=$(cat /sys/class/net/ovs-p0/ifindex) +m4_define([DUMP_SFLOW], [sed -e "s/used:[[0-9]].[[0-9]]*s/used:0.001s/;s/eth(src=[[a-z0-9:]]*,dst=[[a-z0-9:]]*)/eth(macs)/;s/pid=[[0-9]]*/pid=1/;s/output=$p0_ifindex/output=1/"]) +AT_CHECK([ovs-appctl dpctl/dump-flows type=tc,offloaded | grep "in_port(3)" | grep "eth_type(0x0800)" | grep "actions:sample" | grep "sFlow" | DUMP_SFLOW], [0], [dnl +recirc_id(0),in_port(3),eth(macs),eth_type(0x0800),ipv4(frag=no), packets:999, bytes:83916, used:0.001s, actions:sample(sample=50.0%,actions(userspace(pid=1,sFlow(vid=0,pcp=0,output=1),actions))),2 +]) + +hdr="in_ifindex=$p0_ifindex in_format=0 out_ifindex=$p1_ifindex out_format=0 hdr_prot=1 pkt_len=102 stripped=4 hdr_len=98" +count=`grep "$hdr" sflow.log | wc -l` +AT_CHECK([[[[ $count -le 550 && $count -ge 450 ]]]]) +hdr="in_ifindex=$p1_ifindex in_format=0 out_ifindex=$p0_ifindex out_format=0 hdr_prot=1 pkt_len=102 stripped=4 hdr_len=98" +count=`grep "$hdr" sflow.log | wc -l` +AT_CHECK([[[[ $count -le 550 && $count -ge 450 ]]]]) + +OVS_TRAFFIC_VSWITCHD_STOP +OVS_APP_EXIT_AND_WAIT([test-sflow]) +AT_CLEANUP + + +AT_SETUP([offloads - ping over vxlan tunnel with sflow - offloads enabled]) +OVS_CHECK_TUNNEL_TSO() +OVS_CHECK_VXLAN() +LOAD_MODULE([psample]) + +OVS_TRAFFIC_VSWITCHD_START([], [], [-- set Open_vSwitch . other_config:hw-offload=true]) +ADD_BR([br-underlay]) + +AT_CHECK([ovs-ofctl add-flow br0 "actions=normal"]) +AT_CHECK([ovs-ofctl add-flow br-underlay "actions=normal"]) + +ADD_NAMESPACES(at_ns0) + +dnl Set up underlay link from host into the namespace using veth pair. +ADD_VETH(p0, at_ns0, br-underlay, "172.31.1.1/24") +AT_CHECK([ip addr add dev br-underlay "172.31.1.100/24"]) +AT_CHECK([ip link set dev br-underlay up]) + +dnl Set up tunnel endpoints on OVS outside the namespace and with a native +dnl linux device inside the namespace. +ADD_OVS_TUNNEL([vxlan], [br0], [at_vxlan0], [172.31.1.1], [10.1.1.100/24]) +ADD_NATIVE_TUNNEL([vxlan], [at_vxlan1], [at_ns0], [172.31.1.100], [10.1.1.1/24], + [id 0 dstport 4789]) + +on_exit 'kill `cat test-sflow.pid`' +AT_CHECK([ovstest test-sflow --log-file --detach --no-chdir --pidfile 0:127.0.0.1 > sflow.log], [0], [], [ignore]) +AT_CAPTURE_FILE([sflow.log]) +PARSE_LISTENING_PORT([test-sflow.log], [SFLOW_PORT]) +AT_CHECK([ovs-vsctl -- --id=@sflow create sflow agent=lo target=\"127.0.0.1:$SFLOW_PORT\" header=128 sampling=1 polling=100 -- set bridge br0 sflow=@sflow], [0], [ignore]) + +dnl First, check the underlay. +NS_CHECK_EXEC([at_ns0], [ping -q -c 3 -i 0.3 -w 2 172.31.1.100 | FORMAT_PING], [0], [dnl +3 packets transmitted, 3 received, 0% packet loss, time 0ms +]) + +dnl Okay, now check the overlay. +NS_CHECK_EXEC([at_ns0], [ping -q -c 1000 -i 0.01 10.1.1.100 | FORMAT_PING], [0], [dnl +1000 packets transmitted, 1000 received, 0% packet loss, time 0ms +]) + +vxlan_sys_ifindex=$(cat /sys/class/net/vxlan_sys_4789/ifindex) +br0_ifindex=$(cat /sys/class/net/br0/ifindex) +br0_mac=$(cat /sys/class/net/br0/address) +vxlan1_mac=$(ip netns exec at_ns0 cat /sys/class/net/at_vxlan1/address) + +dnl Check the encap flow. +match="recirc_id(0),in_port(br0),eth(src=$br0_mac,dst=$vxlan1_mac),eth_type(0x0800),ipv4(tos=0/0x3,frag=no)" +action="actions:userspace(pid=4294967295,sFlow(vid=0,pcp=0,output=$vxlan_sys_ifindex),actions),set(tunnel(dst=172.31.1.1,ttl=64,tp_dst=4789,flags(0))),vxlan_sys_4789" +AT_CHECK([ovs-appctl dpctl/dump-flows --names | grep $match | grep $action > /dev/null]) + +dnl Check the decap flow. +match="recirc_id(0),tunnel(tun_id=0x0,src=172.31.1.1,dst=172.31.1.100,tp_dst=4789,flags(+key)),in_port(vxlan_sys_4789),eth(src=$vxlan1_mac,dst=$br0_mac),eth_type(0x0800),ipv4(frag=no)" +action="actions:userspace(pid=4294967295,sFlow(vid=0,pcp=0,output=$br0_ifindex),actions),br0" +AT_CHECK([ovs-appctl dpctl/dump-flows --names | grep $match | grep $action> /dev/null]) + +dnl Vni is 0, test-sflow doesn't show it. +out_tunnel_hdr="tunnel4_out_protocol=17 tunnel4_out_src=0.0.0.0 tunnel4_out_dst=172.31.1.1 tunnel4_out_src_port=0 tunnel4_out_dst_port=46354" +ifindex="in_ifindex=$br0_ifindex in_format=0 out_ifindex=$vxlan_sys_ifindex" +out_count=`grep "$out_tunnel_hdr" sflow.log | grep "$ifindex" | wc -l` +AT_CHECK([[[[ $out_count -ge 999 ]]]]) + +in_tunnel_hdr="tunnel4_in_protocol=17 tunnel4_in_src=172.31.1.1 tunnel4_in_dst=172.31.1.100" +ifindex="in_ifindex=$vxlan_sys_ifindex in_format=0 out_ifindex=$br0_ifindex" +in_count=`grep "$in_tunnel_hdr" sflow.log | grep "$ifindex" | wc -l` +AT_CHECK([[[[ $in_count -ge 999 ]]]]) + +OVS_TRAFFIC_VSWITCHD_STOP +OVS_APP_EXIT_AND_WAIT([test-sflow]) +AT_CLEANUP + + AT_SETUP([offloads - set ingress_policing_rate and ingress_policing_burst - offloads disabled]) AT_KEYWORDS([ingress_policing]) AT_SKIP_IF([test $HAVE_TC = "no"])
_______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
