Re: [ovs-dev] [PATCH v2 ovn 1/2] northd: refactor unreachable IPs lb flows

[Mark Michelson]

Hi Lorenzo, I discovered some bugs I missed in my first pass-through. 
Sorry for missing them the first time around. See below.

On 8/25/21 7:58 AM, Lorenzo Bianconi wrote:
Refactor unreachable IPs for vip load-balancers inverting the logic used
during the lb flow creation in order to visit lb first and then related
datapath/ovn_ports. This is a preliminary patch to avoid recomputing
flow hash and lflow lookup when not necessary.
Signed-off-by: Lorenzo Bianconi <lorenzo.bianc...@redhat.com>
---
  northd/ovn-northd.c | 138 ++++++++++++++++++++++++++++++++------------
  1 file changed, 102 insertions(+), 36 deletions(-)

diff --git a/northd/ovn-northd.c b/northd/ovn-northd.c
index af413aba4..f4b5a4a61 100644
--- a/northd/ovn-northd.c
+++ b/northd/ovn-northd.c
@@ -4400,6 +4400,26 @@ do_ovn_lflow_add(struct hmap *lflow_map, struct 
ovn_datapath *od,
  }
  
  /* Adds a row with the specified contents to the Logical_Flow table. */
+static void
+ovn_lflow_add_at_with_hash(struct hmap *lflow_map, struct ovn_datapath *od,
+                           enum ovn_stage stage, uint16_t priority,
+                           const char *match, const char *actions,
+                           const char *io_port, const char *ctrl_meter,
+                           const struct ovsdb_idl_row *stage_hint,
+                           const char *where, uint32_t hash)
+{
+    ovs_assert(ovn_stage_to_datapath_type(stage) == ovn_datapath_get_type(od));
+    if (use_logical_dp_groups && use_parallel_build) {
+        lock_hash_row(&lflow_locks, hash);
+        do_ovn_lflow_add(lflow_map, od, hash, stage, priority, match,
+                         actions, io_port, stage_hint, where, ctrl_meter);
+        unlock_hash_row(&lflow_locks, hash);
+    } else {
+        do_ovn_lflow_add(lflow_map, od, hash, stage, priority, match,
+                         actions, io_port, stage_hint, where, ctrl_meter);
+    }
+}
+
  static void
  ovn_lflow_add_at(struct hmap *lflow_map, struct ovn_datapath *od,
                   enum ovn_stage stage, uint16_t priority,
@@ -4407,24 +4427,14 @@ ovn_lflow_add_at(struct hmap *lflow_map, struct 
ovn_datapath *od,
                   const char *ctrl_meter,
                   const struct ovsdb_idl_row *stage_hint, const char *where)
  {
-    ovs_assert(ovn_stage_to_datapath_type(stage) == ovn_datapath_get_type(od));
-
      uint32_t hash;
  
      hash = ovn_logical_flow_hash(ovn_stage_get_table(stage),
                                   ovn_stage_get_pipeline_name(stage),
                                   priority, match,
                                   actions);
-
-    if (use_logical_dp_groups && use_parallel_build) {
-        lock_hash_row(&lflow_locks, hash);
-        do_ovn_lflow_add(lflow_map, od, hash, stage, priority, match,
-                         actions, io_port, stage_hint, where, ctrl_meter);
-        unlock_hash_row(&lflow_locks, hash);
-    } else {
-        do_ovn_lflow_add(lflow_map, od, hash, stage, priority, match,
-                         actions, io_port, stage_hint, where, ctrl_meter);
-    }
+    ovn_lflow_add_at_with_hash(lflow_map, od, stage, priority, match, actions,
+                               io_port, ctrl_meter, stage_hint, where, hash);
  }
  
  /* Adds a row with the specified contents to the Logical_Flow table. */
@@ -6879,16 +6889,11 @@ build_lswitch_rport_arp_req_flows(struct ovn_port *op,
          /* Check if the ovn port has a network configured on which we could
           * expect ARP requests for the LB VIP.
           */
-        if (ip_parse(ip_addr, &ipv4_addr)) {
-            if (lrouter_port_ipv4_reachable(op, ipv4_addr)) {
-                build_lswitch_rport_arp_req_flow_for_reachable_ip(
-                    ip_addr, AF_INET, sw_op, sw_od, 80, lflows,
-                    stage_hint);
-            } else {
-                build_lswitch_rport_arp_req_flow_for_unreachable_ip(
-                        ip_addr, AF_INET, sw_od, 90, lflows,
-                        stage_hint);
-            }
+        if (ip_parse(ip_addr, &ipv4_addr) &&
+            lrouter_port_ipv4_reachable(op, ipv4_addr)) {
+            build_lswitch_rport_arp_req_flow_for_reachable_ip(
+                ip_addr, AF_INET, sw_op, sw_od, 80, lflows,
+                stage_hint);
          }
      }
      SSET_FOR_EACH (ip_addr, &op->od->lb_ips_v6) {
@@ -6897,16 +6902,11 @@ build_lswitch_rport_arp_req_flows(struct ovn_port *op,
          /* Check if the ovn port has a network configured on which we could
           * expect NS requests for the LB VIP.
           */
-        if (ipv6_parse(ip_addr, &ipv6_addr)) {
-            if (lrouter_port_ipv6_reachable(op, &ipv6_addr)) {
-                build_lswitch_rport_arp_req_flow_for_reachable_ip(
-                    ip_addr, AF_INET6, sw_op, sw_od, 80, lflows,
-                    stage_hint);
-            } else {
-                build_lswitch_rport_arp_req_flow_for_unreachable_ip(
-                    ip_addr, AF_INET6, sw_od, 90, lflows,
-                    stage_hint);
-            }
+        if (ipv6_parse(ip_addr, &ipv6_addr) &&
+            lrouter_port_ipv6_reachable(op, &ipv6_addr)) {
+            build_lswitch_rport_arp_req_flow_for_reachable_ip(
+                ip_addr, AF_INET6, sw_op, sw_od, 80, lflows,
+                stage_hint);
          }
      }
  
@@ -9441,9 +9441,72 @@ build_lrouter_defrag_flows_for_lb(struct ovn_northd_lb *lb,
      ds_destroy(&defrag_actions);
  }
  
+static void
+build_lflows_for_unreachable_vips(struct ovn_northd_lb *lb,
+                                  struct ovn_lb_vip *lb_vip,
+                                  struct hmap *lflows,
+                                  struct hmap *ports,
+                                  struct ds *match)
+{
+    static const char *action = "outport = _MC_flood; output;";
+    bool ipv4 = IN6_IS_ADDR_V4MAPPED(&lb_vip->vip);
+    ovs_be32 ipv4_addr;
+
+    ds_clear(match);
+    if (ipv4) {
+        if (!ip_parse(lb_vip->vip_str, &ipv4_addr)) {
+            return;
+        }
+        ds_put_format(match, "%s && arp.op == 1 && arp.tpa == %s",
+                      FLAGBIT_NOT_VXLAN, lb_vip->vip_str);
+    } else {
+        ds_put_format(match, "%s && nd_ns && nd.target == %s",
+                      FLAGBIT_NOT_VXLAN, lb_vip->vip_str);
+    }
+
+    uint32_t hash = ovn_logical_flow_hash(
+            ovn_stage_get_table(S_SWITCH_IN_L2_LKUP),
+            ovn_stage_get_pipeline_name(S_SWITCH_IN_L2_LKUP), 90,
+            ds_cstr(match), action);
+
+    for (size_t i = 0; i < lb->n_nb_lr; i++) {
+        struct ovn_datapath *od = lb->nb_lr[i];

Let's establish here that od is a logical router. This will be the basis 
for my comments for the rest of this review.

+        if (!od->is_gw_router && !od->n_l3dgw_ports) {
+            continue;
+        }
+
+        for (size_t j = 0; j < od->n_router_ports; j++) {

od->router_ports and od->n_router_ports are only used by logical 
switches. Each logical switch port of type "router" is included in this 
array. In other words, od->router_ports is a list of logical switch 
ports that are connected to logical routers. So I think this means that 
this for loop is a no-op.

I think the appropriate change is to use od->nbr->n_ports here...

+            struct ovn_port *op = ovn_port_get_peer(ports,
+                                                    od->router_ports[j]);

... and use od->nbr->ports[j] here.

Also, ovn_port_get_peer() only works for logical switch ports (it should 
probably be renamed), so this will return NULL every time if a logical 
router port is passed in.

+            if (!op) {
+                continue;
+            }
+
+            if (!od->is_gw_router  && !is_l3dgw_port(op)) {

You already checked od->is_gw_router outside this loop, so you can 
remove the redundant check here.

Also, this code is hurting my brain a bit. In the outer loop, "od" is a 
logical router. We then iterate over od's ports and set "op" to the 
*peer* of the current port. So "op" should presumably be a logical 
switch port. So testing is_l3dgw_port(op) should always return false, right?

+                continue;
+            }
+
+            struct ovn_port *peer = op->peer;
+            if (!peer || !peer->nbsp || lsp_is_external(peer->nbsp)) {
+                continue;
+            }

Now we set peer to op->peer. Since we already know that op is the peer 
of the logical router port, then op->peer should just be the logical 
router port again. Checks for peer->nbsp and lsp_is_external(peer->nbsp) 
will always be false.

+
+            if ((ipv4 && lrouter_port_ipv4_reachable(op, ipv4_addr)) ||
+                (!ipv4 && lrouter_port_ipv6_reachable(op, &lb_vip->vip))) {
+                continue;
+            }
+            ovn_lflow_add_at_with_hash(lflows, peer->od,
+                                       S_SWITCH_IN_L2_LKUP, 90,

And then finally here, since peer is a logical router port, adding the 
lflow to S_SWITCH_IN_L2_LKUP table doesn't work.

I think the issue here is that "op" shouldn't be set to the peer of the 
logical router port. Instead, I think it should be set directly to the 
logical router port. If you do this, then the rest of the logic in this 
loop should work correctly.

+                                       ds_cstr(match), action,
+                                       NULL, NULL, &peer->nbsp->header_,
+                                       OVS_SOURCE_LOCATOR, hash);
+        }
+    }
+}
+
  static void
  build_lrouter_flows_for_lb(struct ovn_northd_lb *lb, struct hmap *lflows,
-                           struct shash *meter_groups,
+                           struct hmap *ports, struct shash *meter_groups,
                             struct ds *match, struct ds *action)
  {
      if (!lb->n_nb_lr) {
@@ -9453,6 +9516,8 @@ build_lrouter_flows_for_lb(struct ovn_northd_lb *lb, 
struct hmap *lflows,
      for (size_t i = 0; i < lb->n_vips; i++) {
          struct ovn_lb_vip *lb_vip = &lb->vips[i];
  
+        build_lflows_for_unreachable_vips(lb, lb_vip, lflows, ports, match);
+
          build_lrouter_nat_flows_for_lb(lb_vip, lb, &lb->vips_nb[i],
                                         lflows, match, action,
                                         meter_groups);
@@ -12778,7 +12843,7 @@ build_lflows_thread(void *arg)
                      build_lrouter_defrag_flows_for_lb(lb, lsi->lflows,
                                                        &lsi->match);
                      build_lrouter_flows_for_lb(lb, lsi->lflows,
-                                               lsi->meter_groups,
+                                               lsi->ports, lsi->meter_groups,
                                                 &lsi->match, &lsi->actions);
                      build_lswitch_flows_for_lb(lb, lsi->lflows,
                                                 lsi->meter_groups,
@@ -12947,8 +13012,9 @@ build_lswitch_and_lrouter_flows(struct hmap *datapaths, 
struct hmap *ports,
                                                   &lsi.actions,
                                                   &lsi.match);
              build_lrouter_defrag_flows_for_lb(lb, lsi.lflows, &lsi.match);
-            build_lrouter_flows_for_lb(lb, lsi.lflows, lsi.meter_groups,
-                                       &lsi.match, &lsi.actions);
+            build_lrouter_flows_for_lb(lb, lsi.lflows, lsi.ports,
+                                       lsi.meter_groups, &lsi.match,
+                                       &lsi.actions);
              build_lswitch_flows_for_lb(lb, lsi.lflows, lsi.meter_groups,
                                         &lsi.match, &lsi.actions);
          }


_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev