static struct vr_bridge_entry *
bridge_lookup(uint8_t *mac, struct vr_forwarding_md *fmd)
{
struct vr_route_req rt;
struct vr_bridge_entry *be;
rt.rtr_req.rtr_label_flags = 0;
rt.rtr_req.rtr_index = VR_BE_INVALID_INDEX;
rt.rtr_req.rtr_mac_size = VR_ETHER_ALEN;
rt.rtr_req.rtr_mac = mac;
/* If multicast L2 packet, use broadcast composite nexthop */
if (IS_MAC_BMCAST(rt.rtr_req.rtr_mac))
rt.rtr_req.rtr_mac = (int8_t *)vr_bcast_mac;
rt.rtr_req.rtr_vrf_id = fmd->fmd_dvrf;
be = __bridge_lookup(rt.rtr_req.rtr_vrf_id, &rt);
if (be)
bridge_update_route_req(be, &rt);
if (fmd && (rt.rtr_req.rtr_label_flags & VR_BE_LABEL_VALID_FLAG)) {
vr_forwarding_md_set_label(fmd, rt.rtr_req.rtr_label,
VR_LABEL_TYPE_UNKNOWN);
}
return be;
}
unsigned int
vr_bridge_input(struct vrouter *router, unsigned short vrf,
struct vr_packet *pkt, struct vr_forwarding_md *fmd)
{
struct vr_route_req rt;
struct vr_nexthop *nh;
struct vr_forwarding_md cmd;
char bcast_mac[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
char *mac;
/* First mark the packet as L2 */
pkt->vp_type = VP_TYPE_L2;
mac = (char *)pkt_data(pkt);
rt.rtr_req.rtr_mac_size = VR_ETHER_ALEN;
rt.rtr_req.rtr_mac =(int8_t *) mac;
/* If multicast L2 packet, use broadcast composite nexthop */
if (IS_MAC_BMCAST(mac)) {
rt.rtr_req.rtr_mac = (int8_t *)bcast_mac;
pkt->vp_flags |= VP_FLAG_MULTICAST;
}
rt.rtr_req.rtr_vrf_id = vrf;
nh = vr_bridge_lookup(vrf, &rt, pkt);
if (nh) {
/*
* If there is a label attached to this bridge entry add the
* label
*/
if (rt.rtr_req.rtr_label_flags & VR_RT_LABEL_VALID_FLAG) {
if (!fmd) {
vr_init_forwarding_md(&cmd);
fmd = &cmd;
}
fmd->fmd_label = rt.rtr_req.rtr_label;
}
return nh_output(vrf, pkt, nh, fmd);
}
vr_pfree(pkt, VP_DROP_INVALID_NH);
return 0;
}
/*
* This funciton parses the ethernet packet and assigns the
* pkt->vp_type, network protocol of the packet. The ethernet header can
* start from an offset from vp_data
*/
int
vr_pkt_type(struct vr_packet *pkt, unsigned short offset,
struct vr_forwarding_md *fmd)
{
unsigned char *eth = pkt_data(pkt) + offset;
unsigned short eth_proto;
int pull_len, pkt_len = pkt_head_len(pkt) - offset;
struct vr_vlan_hdr *vlan;
pull_len = VR_ETHER_HLEN;
if (pkt_len < pull_len)
return -1;
pkt->vp_flags &= ~(VP_FLAG_MULTICAST);
/* L2 broadcast/multicast packets are multicast packets */
if (IS_MAC_BMCAST(eth))
pkt->vp_flags |= VP_FLAG_MULTICAST;
eth_proto = ntohs(*(unsigned short *)(eth + VR_ETHER_PROTO_OFF));
while (eth_proto == VR_ETH_PROTO_VLAN) {
if (pkt_len < (pull_len + sizeof(*vlan)))
return -1;
vlan = (struct vr_vlan_hdr *)(eth + pull_len);
if (fmd && (fmd->fmd_vlan == VLAN_ID_INVALID))
fmd->fmd_vlan = vlan->vlan_tag & 0xFFF;
eth_proto = ntohs(vlan->vlan_proto);
pull_len += sizeof(*vlan);
}
pkt_set_network_header(pkt, pkt->vp_data + offset + pull_len);
pkt_set_inner_network_header(pkt, pkt->vp_data + offset + pull_len);
pkt->vp_type = vr_eth_proto_to_pkt_type(eth_proto);
return 0;
}
unsigned int
vr_bridge_input(struct vrouter *router, struct vr_packet *pkt,
struct vr_forwarding_md *fmd)
{
struct vr_route_req rt;
struct vr_forwarding_md cmd;
struct vr_nexthop *nh;
unsigned short pull_len, overlay_len = VROUTER_L2_OVERLAY_LEN;
int reason;
rt.rtr_req.rtr_label_flags = 0;
rt.rtr_req.rtr_index = VR_BE_INVALID_INDEX;
rt.rtr_req.rtr_mac_size = VR_ETHER_ALEN;
rt.rtr_req.rtr_mac =(int8_t *) pkt_data(pkt);
/* If multicast L2 packet, use broadcast composite nexthop */
if (IS_MAC_BMCAST(rt.rtr_req.rtr_mac))
rt.rtr_req.rtr_mac = (int8_t *)vr_bcast_mac;
rt.rtr_req.rtr_vrf_id = fmd->fmd_dvrf;
nh = vr_bridge_lookup(fmd->fmd_dvrf, &rt);
if (!nh) {
vr_pfree(pkt, VP_DROP_L2_NO_ROUTE);
return 0;
}
if (nh->nh_type == NH_L2_RCV)
overlay_len = VROUTER_OVERLAY_LEN;
if (pkt->vp_type == VP_TYPE_IP || pkt->vp_type == VP_TYPE_IP6) {
if (vif_is_virtual(pkt->vp_if) &&
vr_from_vm_mss_adj && vr_pkt_from_vm_tcp_mss_adj) {
pull_len = pkt_get_network_header_off(pkt) - pkt_head_space(pkt);
if (!pkt_pull(pkt, pull_len)) {
vr_pfree(pkt, VP_DROP_PULL);
return 0;
}
if ((reason = vr_pkt_from_vm_tcp_mss_adj(pkt, overlay_len))) {
vr_pfree(pkt, reason);
return 0;
}
if (!pkt_push(pkt, pull_len)) {
vr_pfree(pkt, VP_DROP_PUSH);
return 0;
}
}
}
/*
* If there is a label attached to this bridge entry add the
* label
*/
if (rt.rtr_req.rtr_label_flags & VR_RT_LABEL_VALID_FLAG) {
if (!fmd) {
vr_init_forwarding_md(&cmd);
fmd = &cmd;
}
fmd->fmd_label = rt.rtr_req.rtr_label;
}
nh_output(pkt, nh, fmd);
return 0;
}
unsigned int
vr_bridge_input(struct vrouter *router, struct vr_packet *pkt,
struct vr_forwarding_md *fmd)
{
int reason, handled;
l4_pkt_type_t l4_type = L4_TYPE_UNKNOWN;
unsigned short pull_len, overlay_len = VROUTER_OVERLAY_LEN;
int8_t *dmac;
struct vr_bridge_entry *be;
struct vr_nexthop *nh = NULL;
struct vr_vrf_stats *stats;
dmac = (int8_t *) pkt_data(pkt);
if (pkt->vp_if->vif_flags & VIF_FLAG_MAC_LEARN) {
if (vr_bridge_learn(router, pkt, fmd)) {
return 0;
}
}
pull_len = 0;
if ((pkt->vp_type == VP_TYPE_IP) || (pkt->vp_type == VP_TYPE_IP6) ||
(pkt->vp_type == VP_TYPE_ARP)) {
pull_len = pkt_get_network_header_off(pkt) - pkt_head_space(pkt);
if (pull_len && !pkt_pull(pkt, pull_len)) {
vr_pfree(pkt, VP_DROP_PULL);
return 0;
}
}
if ((pkt->vp_type == VP_TYPE_IP) || (pkt->vp_type == VP_TYPE_IP6)) {
if (fmd->fmd_dscp < 0) {
if (pkt->vp_type == VP_TYPE_IP) {
fmd->fmd_dscp =
vr_inet_get_tos((struct vr_ip *)pkt_network_header(pkt));
} else if (pkt->vp_type == VP_TYPE_IP6) {
fmd->fmd_dscp =
vr_inet6_get_tos((struct vr_ip6 *)pkt_network_header(pkt));
}
}
} else {
if (fmd->fmd_dotonep < 0) {
fmd->fmd_dotonep = vr_vlan_get_tos(pkt_data(pkt));
}
}
/* Do the bridge lookup for the packets not meant for "me" */
if (!fmd->fmd_to_me) {
/*
* If DHCP packet coming from VM, Trap it to Agent before doing the bridge
* lookup itself
*/
if (vif_is_virtual(pkt->vp_if)) {
if (pkt->vp_type == VP_TYPE_IP)
l4_type = vr_ip_well_known_packet(pkt);
else if (pkt->vp_type == VP_TYPE_IP6)
l4_type = vr_ip6_well_known_packet(pkt);
if (l4_type == L4_TYPE_DHCP_REQUEST) {
if (pkt->vp_if->vif_flags & VIF_FLAG_DHCP_ENABLED) {
vr_trap(pkt, fmd->fmd_dvrf, AGENT_TRAP_L3_PROTOCOLS, NULL);
return 0;
}
}
/*
* Handle the unicast ARP, coming from VM, not
* destined to us. Broadcast ARP requests would be handled
* in L2 multicast nexthop. Multicast ARP on fabric
* interface also would be handled in L2 multicast nexthop.
* Unicast ARP packets on fabric interface would be handled
* in plug routines of interface.
*/
if (!IS_MAC_BMCAST(dmac)) {
handled = 0;
if (pkt->vp_type == VP_TYPE_ARP) {
handled = vr_arp_input(pkt, fmd, dmac);
} else if (l4_type == L4_TYPE_NEIGHBOUR_SOLICITATION) {
handled = vr_neighbor_input(pkt, fmd, dmac);
}
if (handled)
return 0;
}
}
be = bridge_lookup(dmac, fmd);
if (be)
nh = be->be_nh;
if (!nh || nh->nh_type == NH_DISCARD) {
/* If Flooding of unknown unicast not allowed, drop the packet */
if (!vr_unknown_uc_flood(pkt->vp_if, pkt->vp_nh) ||
IS_MAC_BMCAST(dmac)) {
vr_pfree(pkt, VP_DROP_L2_NO_ROUTE);
return 0;
}
be = bridge_lookup(vr_bcast_mac, fmd);
nh = be->be_nh;
if (!nh) {
vr_pfree(pkt, VP_DROP_L2_NO_ROUTE);
return 0;
}
stats = vr_inet_vrf_stats(fmd->fmd_dvrf, pkt->vp_cpu);
if (stats)
stats->vrf_uuc_floods++;
/* Treat this unknown unicast packet as multicast */
pkt->vp_flags |= VP_FLAG_MULTICAST;
}
if (be)
__sync_fetch_and_add(&be->be_packets, 1);
if (nh->nh_type != NH_L2_RCV)
overlay_len = VROUTER_L2_OVERLAY_LEN;
}
/* Adjust MSS for V4 and V6 packets */
if ((pkt->vp_type == VP_TYPE_IP) || (pkt->vp_type == VP_TYPE_IP6)) {
if (vif_is_virtual(pkt->vp_if) &&
vr_from_vm_mss_adj && vr_pkt_from_vm_tcp_mss_adj) {
if ((reason = vr_pkt_from_vm_tcp_mss_adj(pkt, overlay_len))) {
vr_pfree(pkt, reason);
return 0;
}
}
if (fmd->fmd_to_me) {
handled = vr_l3_input(pkt, fmd);
if (!handled) {
vr_pfree(pkt, VP_DROP_NOWHERE_TO_GO);
}
return 0;
}
}
if (pull_len && !pkt_push(pkt, pull_len)) {
vr_pfree(pkt, VP_DROP_PUSH);
return 0;
}
nh_output(pkt, nh, fmd);
return 0;
}
unsigned int
vr_bridge_learn(struct vrouter *router, struct vr_packet *pkt,
struct vr_forwarding_md *fmd)
{
int ret = 0, lock, valid_src;
unsigned int trap_reason;
bool trap = false;
struct vr_eth *eth;
struct vr_packet *pkt_c;
struct vr_nexthop *nh = NULL;
struct vr_bridge_entry *be;
eth = (struct vr_eth *)pkt_data(pkt);
if (!eth)
return 0;
if (IS_MAC_BMCAST(eth->eth_smac))
return 0;
be = bridge_lookup(eth->eth_smac, fmd);
if (be) {
nh = be->be_nh;
}
if (!nh) {
be = bridge_lookup((uint8_t *)vr_bcast_mac, fmd);
if (be) {
nh = be->be_nh;
}
if (!nh)
return 0;
lock = bridge_table_lock(pkt->vp_if, eth->eth_smac);
if (lock < 0)
return 0;
be = bridge_add(0, fmd->fmd_dvrf, eth->eth_smac, nh->nh_id);
bridge_table_unlock(pkt->vp_if, eth->eth_smac, lock);
if (!be)
return -ENOMEM;
trap_reason = AGENT_TRAP_MAC_LEARN;
trap = true;
} else {
if (!(be->be_flags & VR_BE_MAC_MOVED_FLAG) && (nh->nh_validate_src)) {
valid_src = nh->nh_validate_src(pkt, nh, fmd, NULL);
if (valid_src != NH_SOURCE_VALID) {
ret = vr_bridge_set_route_flags(be, VR_BE_MAC_MOVED_FLAG);
if (!ret) {
/* trap the packet for mac move */
trap_reason = AGENT_TRAP_MAC_MOVE;
trap = true;
}
ret = 0;
}
}
}
__sync_fetch_and_add(&be->be_packets, 1);
if (trap) {
pkt_c = pkt_cow(pkt, 0);
if (!pkt_c) {
pkt_c = pkt;
ret = -ENOMEM;
}
vr_trap(pkt_c, fmd->fmd_dvrf,
trap_reason, (void *)&be->be_hentry.hentry_index);
}
return ret;
}