ovn/controller/physical.c

   1 /* Copyright (c) 2015 Nicira, Inc.
   2  *
   3  * Licensed under the Apache License, Version 2.0 (the "License");
   4  * you may not use this file except in compliance with the License.
   5  * You may obtain a copy of the License at:
   6  *
   7  *     http://www.apache.org/licenses/LICENSE-2.0
   8  *
   9  * Unless required by applicable law or agreed to in writing, software
  10  * distributed under the License is distributed on an "AS IS" BASIS,
  11  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12  * See the License for the specific language governing permissions and
  13  * limitations under the License.
  14  */
  15
  16 #include <config.h>
  17 #include "physical.h"
  18 #include "match.h"
  19 #include "ofctrl.h"
  20 #include "ofp-actions.h"
  21 #include "ofpbuf.h"
  22 #include "ovn-controller.h"
  23 #include "ovn/lib/ovn-sb-idl.h"
  24 #include "pipeline.h"
  25 #include "simap.h"
  26 #include "vswitch-idl.h"
  27
  28 void
  29 physical_register_ovs_idl(struct ovsdb_idl *ovs_idl)
  30 {
  31     ovsdb_idl_add_table(ovs_idl, &ovsrec_table_bridge);
  32     ovsdb_idl_add_column(ovs_idl, &ovsrec_bridge_col_ports);
  33
  34     ovsdb_idl_add_table(ovs_idl, &ovsrec_table_port);
  35     ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_name);
  36     ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_interfaces);
  37     ovsdb_idl_add_column(ovs_idl, &ovsrec_port_col_external_ids);
  38
  39     ovsdb_idl_add_table(ovs_idl, &ovsrec_table_interface);
  40     ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_name);
  41     ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_ofport);
  42     ovsdb_idl_add_column(ovs_idl, &ovsrec_interface_col_external_ids);
  43 }
  44
  45 void
  46 physical_run(struct controller_ctx *ctx, const struct ovsrec_bridge *br_int,
  47              const char *this_chassis_id, struct hmap *flow_table)
  48 {
  49     struct simap lport_to_ofport = SIMAP_INITIALIZER(&lport_to_ofport);
  50     struct simap chassis_to_ofport = SIMAP_INITIALIZER(&chassis_to_ofport);
  51     for (int i = 0; i < br_int->n_ports; i++) {
  52         const struct ovsrec_port *port_rec = br_int->ports[i];
  53         if (!strcmp(port_rec->name, br_int->name)) {
  54             continue;
  55         }
  56
  57         const char *chassis_id = smap_get(&port_rec->external_ids,
  58                                           "ovn-chassis-id");
  59         if (chassis_id && !strcmp(chassis_id, this_chassis_id)) {
  60             continue;
  61         }
  62
  63         for (int j = 0; j < port_rec->n_interfaces; j++) {
  64             const struct ovsrec_interface *iface_rec = port_rec->interfaces[j];
  65
  66             /* Get OpenFlow port number. */
  67             if (!iface_rec->n_ofport) {
  68                 continue;
  69             }
  70             int64_t ofport = iface_rec->ofport[0];
  71             if (ofport < 1 || ofport > ofp_to_u16(OFPP_MAX)) {
  72                 continue;
  73             }
  74
  75             /* Record as chassis or local logical port. */
  76             if (chassis_id) {
  77                 simap_put(&chassis_to_ofport, chassis_id, ofport);
  78                 break;
  79             } else {
  80                 const char *iface_id = smap_get(&iface_rec->external_ids,
  81                                                 "iface-id");
  82                 if (iface_id) {
  83                     simap_put(&lport_to_ofport, iface_id, ofport);
  84                 }
  85             }
  86         }
  87     }
  88
  89     struct ofpbuf ofpacts;
  90     ofpbuf_init(&ofpacts, 0);
  91
  92     /* Set up flows in table 0 for physical-to-logical translation and in table
  93      * 64 for logical-to-physical translation. */
  94     const struct sbrec_port_binding *binding;
  95     SBREC_PORT_BINDING_FOR_EACH (binding, ctx->ovnsb_idl) {
  96         /* Find the OpenFlow port for the logical port, as 'ofport'.  If it's
  97          * on a remote chassis, this is the OpenFlow port for the tunnel to
  98          * that chassis (and set 'local' to false).  Otherwise, if it's on the
  99          * chassis we're managing, this is the OpenFlow port for the vif itself
 100          * (and set 'local' to true). When 'parent_port' is set for a binding,
 101          * it implies a container sitting inside a VM reachable via a 'tag'.
 102          */
 103
 104         int tag = 0;
 105         ofp_port_t ofport;
 106         if (binding->parent_port) {
 107             ofport = u16_to_ofp(simap_get(&lport_to_ofport,
 108                                           binding->parent_port));
 109             if (ofport && binding->tag) {
 110                 tag = *binding->tag;
 111             }
 112         } else {
 113             ofport = u16_to_ofp(simap_get(&lport_to_ofport,
 114                                           binding->logical_port));
 115         }
 116
 117         bool local = ofport != 0;
 118         if (!local) {
 119             if (!binding->chassis) {
 120                 continue;
 121             }
 122             ofport = u16_to_ofp(simap_get(&chassis_to_ofport,
 123                                           binding->chassis->name));
 124             if (!ofport) {
 125                 continue;
 126             }
 127         }
 128
 129         /* Translate the logical datapath into the form we use in
 130          * MFF_METADATA. */
 131         uint32_t ldp = ldp_to_integer(&binding->logical_datapath);
 132         if (!ldp) {
 133             continue;
 134         }
 135
 136         struct match match;
 137         if (local) {
 138             /* Packets that arrive from a vif can belong to a VM or
 139              * to a container located inside that VM. Packets that
 140              * arrive from containers have a tag (vlan) associated with them.
 141              */
 142
 143             /* Table 0, Priority 150 and 100.
 144              * ==============================
 145              *
 146              * Priority 150 is for traffic belonging to containers. For such
 147              * traffic, match on the tags and then strip the tag.
 148              * Priority 100 is for traffic belonging to VMs.
 149              *
 150              * For both types of traffic: set MFF_LOG_INPORT to the
 151              * logical input port, MFF_METADATA to the logical datapath, and
 152              * resubmit into the logical pipeline starting at table 16. */
 153             match_init_catchall(&match);
 154             ofpbuf_clear(&ofpacts);
 155             match_set_in_port(&match, ofport);
 156             if (tag) {
 157                 match_set_dl_vlan(&match, htons(tag));
 158             }
 159
 160             /* Set MFF_METADATA. */
 161             struct ofpact_set_field *sf = ofpact_put_SET_FIELD(&ofpacts);
 162             sf->field = mf_from_id(MFF_METADATA);
 163             sf->value.be64 = htonll(ldp);
 164             sf->mask.be64 = OVS_BE64_MAX;
 165
 166             /* Set MFF_LOG_INPORT. */
 167             sf = ofpact_put_SET_FIELD(&ofpacts);
 168             sf->field = mf_from_id(MFF_LOG_INPORT);
 169             sf->value.be32 = htonl(binding->tunnel_key);
 170             sf->mask.be32 = OVS_BE32_MAX;
 171
 172             /* Strip vlans. */
 173             if (tag) {
 174                 ofpact_put_STRIP_VLAN(&ofpacts);
 175             }
 176
 177             /* Resubmit to first logical pipeline table. */
 178             struct ofpact_resubmit *resubmit = ofpact_put_RESUBMIT(&ofpacts);
 179             resubmit->in_port = OFPP_IN_PORT;
 180             resubmit->table_id = 16;
 181             ofctrl_add_flow(flow_table, 0, tag ? 150 : 100, &match, &ofpacts);
 182
 183             /* Table 0, Priority 50.
 184              * =====================
 185              *
 186              * For packets that arrive from a remote node destined to this
 187              * local vif: deliver directly to the vif. If the destination
 188              * is a container sitting behind a vif, tag the packets. */
 189             match_init_catchall(&match);
 190             ofpbuf_clear(&ofpacts);
 191             match_set_tun_id(&match, htonll(binding->tunnel_key));
 192             if (tag) {
 193                 struct ofpact_vlan_vid *vlan_vid;
 194                 vlan_vid = ofpact_put_SET_VLAN_VID(&ofpacts);
 195                 vlan_vid->vlan_vid = tag;
 196                 vlan_vid->push_vlan_if_needed = true;
 197             }
 198             ofpact_put_OUTPUT(&ofpacts)->port = ofport;
 199             ofctrl_add_flow(flow_table, 0, 50, &match, &ofpacts);
 200         }
 201
 202         /* Table 64, Priority 100.
 203          * =======================
 204          *
 205          * Drop packets whose logical inport and outport are the same. */
 206         match_init_catchall(&match);
 207         ofpbuf_clear(&ofpacts);
 208         match_set_reg(&match, MFF_LOG_INPORT - MFF_REG0, binding->tunnel_key);
 209         match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0, binding->tunnel_key);
 210         ofctrl_add_flow(flow_table, 64, 100, &match, &ofpacts);
 211
 212         /* Table 64, Priority 50.
 213          * ======================
 214          *
 215          * For packets to remote machines, send them over a tunnel to the
 216          * remote chassis.
 217          *
 218          * For packets to local vifs, deliver them directly. */
 219         match_init_catchall(&match);
 220         ofpbuf_clear(&ofpacts);
 221         match_set_reg(&match, MFF_LOG_OUTPORT - MFF_REG0, binding->tunnel_key);
 222         if (!local) {
 223             /* Set MFF_TUN_ID. */
 224             struct ofpact_set_field *sf = ofpact_put_SET_FIELD(&ofpacts);
 225             sf->field = mf_from_id(MFF_TUN_ID);
 226             sf->value.be64 = htonll(binding->tunnel_key);
 227             sf->mask.be64 = OVS_BE64_MAX;
 228         }
 229         if (tag) {
 230             /* For containers sitting behind a local vif, tag the packets
 231              * before delivering them. Since there is a possibility of
 232              * packets needing to hair-pin back into the same vif from
 233              * which it came, push the in_port to stack and make the
 234              * in_port as zero. */
 235             struct ofpact_vlan_vid *vlan_vid;
 236             vlan_vid = ofpact_put_SET_VLAN_VID(&ofpacts);
 237             vlan_vid->vlan_vid = tag;
 238             vlan_vid->push_vlan_if_needed = true;
 239
 240             struct ofpact_stack *stack_action;
 241             const struct mf_field *field;
 242             stack_action = ofpact_put_STACK_PUSH(&ofpacts);
 243             field = mf_from_id(MFF_IN_PORT);
 244             stack_action->subfield.field = field;
 245             stack_action->subfield.ofs = 0;
 246             stack_action->subfield.n_bits = field->n_bits;
 247
 248             struct ofpact_set_field *sf = ofpact_put_SET_FIELD(&ofpacts);
 249             sf->field = mf_from_id(MFF_IN_PORT);
 250             sf->value.be16 = 0;
 251             sf->mask.be16 = OVS_BE16_MAX;
 252         }
 253         ofpact_put_OUTPUT(&ofpacts)->port = ofport;
 254         if (tag) {
 255             /* Revert the tag added to the packets headed to containers
 256              * in the previous step. If we don't do this, the packets
 257              * that are to be broadcasted to a VM in the same logical
 258              * switch will also contain the tag. Also revert the zero'd
 259              * in_port. */
 260             ofpact_put_STRIP_VLAN(&ofpacts);
 261
 262             struct ofpact_stack *stack_action;
 263             const struct mf_field *field;
 264             stack_action = ofpact_put_STACK_POP(&ofpacts);
 265             field = mf_from_id(MFF_IN_PORT);
 266             stack_action->subfield.field = field;
 267             stack_action->subfield.ofs = 0;
 268             stack_action->subfield.n_bits = field->n_bits;
 269         }
 270         ofctrl_add_flow(flow_table, 64, 50, &match, &ofpacts);
 271     }
 272
 273     ofpbuf_uninit(&ofpacts);
 274     simap_destroy(&lport_to_ofport);
 275     simap_destroy(&chassis_to_ofport);
 276 }