/*
* vr_input is called from linux(host) ingress path. we are not allowed to
* sleep here. return value should indicate whether the router consumed the
* packet or not. if the router did not consume, host will continue with
* its packet processing with the same packet. if the router did consume,
* host will not touch the packet again. a return of 0 will tell the handler
* that router consumed it, while all other return values are passed as is.
* maybe we need a return value to host return mapping, but at a later time ?
*/
unsigned int
vr_input(unsigned short vrf, struct vr_interface *vif, struct vr_packet *pkt)
{
unsigned char *data = pkt_data(pkt);
unsigned char *eth = data;
unsigned char *dmac = ð[VR_ETHER_DMAC_OFF];
unsigned short eth_proto;
struct vr_vlan_hdr *vlan;
struct vrouter *router = vif->vif_router;
struct vr_forwarding_md fmd;
int reason;
if (vif->vif_flags & VIF_FLAG_MIRROR_RX) {
vr_init_forwarding_md(&fmd);
fmd.fmd_dvrf = vif->vif_vrf;
vr_mirror(vif->vif_router, vif->vif_mirror_id, pkt, &fmd);
}
/*
* we will optimise for the most likely case i.e that of IPv4. need
* to see what needs to happen for v6 when it comes
*/
data = pkt_pull(pkt, VR_ETHER_HLEN);
if (!data) {
vif_drop_pkt(vif, pkt, 1);
return 0;
}
eth_proto = ntohs(*(unsigned short *)(eth + VR_ETHER_PROTO_OFF));
while (eth_proto == VR_ETH_PROTO_VLAN) {
vlan = (struct vr_vlan_hdr *)data;
eth_proto = ntohs(vlan->vlan_proto);
data = pkt_pull(pkt, sizeof(*vlan));
if (!data) {
vif_drop_pkt(vif, pkt, 1);
return 0;
}
}
vr_init_forwarding_md(&fmd);
pkt_set_network_header(pkt, pkt->vp_data);
pkt_set_inner_network_header(pkt, pkt->vp_data);
if (eth_proto == VR_ETH_PROTO_IP) {
if (vr_from_vm_mss_adj && vr_pkt_from_vm_tcp_mss_adj &&
(vif->vif_type == VIF_TYPE_VIRTUAL)) {
if ((reason = vr_pkt_from_vm_tcp_mss_adj(pkt))) {
vr_pfree(pkt, reason);
return 0;
}
}
return vr_flow_inet_input(router, vrf, pkt, eth_proto, &fmd);
} else if (eth_proto == VR_ETH_PROTO_ARP)
return vr_arp_input(router, vrf, pkt);
/* rest of the stuff is for slow path and we should be ok doing this */
if (well_known_mac(dmac))
return vr_trap(pkt, vrf, AGENT_TRAP_L2_PROTOCOLS, NULL);
return vr_default_input(pkt);
}
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;
}
