Consider the following flow table: table=0 actions=resubmit(,1),2 table=1 actions=debug_recirc
When debug_recirc triggers recirculation and we later resume processing, only the output to port 2 should be executed, because the effects of "resubmit" have already taken place. However, until now, the "resubmit" was added to the actions to execute post-recirculation, resulting in an infinite loop. Now consider this flow table (as seen in the "MPLS handling" test in ofproto-dpif.at): table=0 actions=pop_mpls(0x0806),resubmit(,1) table=1 ip,nw_dst=1.2.3.4 actions=controller Here, we do want to add the "resubmit" to the actions to execute post-recirculation, since the "resubmit" cannot be processed until after recirculation makes the nw_dst field available. This commit fixes the problem in both cases. Found when testing a feature based on recirculation. Signed-off-by: Ben Pfaff <b...@ovn.org> --- v1->v2: Rewrite after realizing that I broke the "MPLS handling" test. ofproto/ofproto-dpif-xlate.c | 24 +++++++++++++++++++++++- tests/ofproto-dpif.at | 24 ++++++++++++++++++++++++ 2 files changed, 47 insertions(+), 1 deletion(-) diff --git a/ofproto/ofproto-dpif-xlate.c b/ofproto/ofproto-dpif-xlate.c index 4b25bf4..6e928da 100644 --- a/ofproto/ofproto-dpif-xlate.c +++ b/ofproto/ofproto-dpif-xlate.c @@ -4575,8 +4575,30 @@ do_xlate_actions(const struct ofpact *ofpacts, size_t ofpacts_len, break; case OFPACT_RESUBMIT: + /* Recirculation complicates resubmit. There are two cases: + * + * - If mpls_pop has been executed, then the flow table lookup + * as part of resubmit might depend on fields that can only + * be obtained via recirculation, so the resubmit itself + * triggers recirculation and we need to make sure that the + * resubmit is executed again after recirculation. + * Therefore, in this case we trigger recirculation and let + * the code following this "switch" append the resubmit to + * the post-recirculation actions. + * + * - Otherwise, some action in the flow entry found by resubmit + * might trigger recirculation. If that happens, then we do + * not want to execute the resubmit again after + * recirculation, so we want to skip back to the head of the + * loop to avoid that, only adding any actions that follow + * the resubmit to the post-recirculation actions. + */ + if (ctx->was_mpls) { + ctx_trigger_recirculation(ctx); + break; + } xlate_ofpact_resubmit(ctx, ofpact_get_RESUBMIT(a)); - break; + continue; case OFPACT_SET_TUNNEL: flow->tunnel.tun_id = htonll(ofpact_get_SET_TUNNEL(a)->tun_id); diff --git a/tests/ofproto-dpif.at b/tests/ofproto-dpif.at index bfb1b56..aa2f1bb 100644 --- a/tests/ofproto-dpif.at +++ b/tests/ofproto-dpif.at @@ -4203,6 +4203,30 @@ AT_CHECK_UNQUOTED([tail -1 stdout], [0], [Datapath actions: 3 OVS_VSWITCHD_STOP AT_CLEANUP +# This test verifies that "resubmit", when it triggers recirculation +# indirectly through the flow that it recursively invokes, is not +# re-executed when execution continues later post-recirculation. +AT_SETUP([ofproto-dpif - recirculation after resubmit]) +OVS_VSWITCHD_START +ADD_OF_PORTS([br0], [1], [2]) + +AT_DATA([flows.txt], [dnl +table=0 in_port=1 actions=resubmit(,1),2 +table=1 in_port=1 actions=debug_recirc +]) +AT_CHECK([ovs-ofctl add-flows br0 flows.txt]) + +flow="in_port(1),eth(src=50:54:00:00:00:05,dst=50:54:00:00:00:07),eth_type(0x0800),ipv4(src=192.168.0.1,dst=192.168.0.2,proto=1,tos=0,ttl=128,frag=no),icmp(type=8,code=0)" +AT_CHECK([ovs-appctl ofproto/trace ovs-dummy "$flow" -generate], [0], [stdout]) +AT_CHECK_UNQUOTED([tail -1 stdout], [0], [Datapath actions: recirc(0x1) +]) +AT_CHECK([ovs-appctl ofproto/trace ovs-dummy "$flow,recirc_id(1)" -generate], [0], [stdout]) +AT_CHECK_UNQUOTED([tail -1 stdout], [0], [Datapath actions: 2 +]) + +OVS_VSWITCHD_STOP +AT_CLEANUP + # Two testcases below are for the ofproto/trace command # The first one tests all correct syntax: # ofproto/trace [dp_name] odp_flow [-generate|packet] -- 2.1.3 _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev