/* vmxnet3 create API */
static int
api_vmxnet3_create (vat_main_t * vam)
{
unformat_input_t *i = vam->input;
vl_api_vmxnet3_create_t *mp;
vmxnet3_create_if_args_t args;
int ret;
u32 x[4];
clib_memset (&args, 0, sizeof (vmxnet3_create_if_args_t));
while (unformat_check_input (i) != UNFORMAT_END_OF_INPUT)
{
if (unformat (i, "%x:%x:%x.%x", &x[0], &x[1], &x[2], &x[3]))
{
args.addr.domain = x[0];
args.addr.bus = x[1];
args.addr.slot = x[2];
args.addr.function = x[3];
}
else if (unformat (i, "elog"))
args.enable_elog = 1;
else if (unformat (i, "bind"))
args.bind = 1;
else if (unformat (i, "rx-queue-size %u", &args.rxq_size))
;
else if (unformat (i, "tx-queue-size %u", &args.txq_size))
;
else if (unformat (i, "num-tx-queues %u", &args.txq_num))
;
else if (unformat (i, "num-rx-queues %u", &args.rxq_num))
;
else
{
clib_warning ("unknown input '%U'", format_unformat_error, i);
return -99;
}
}
M (VMXNET3_CREATE, mp);
mp->pci_addr = clib_host_to_net_u32 (args.addr.as_u32);
mp->enable_elog = clib_host_to_net_u16 (args.enable_elog);
mp->rxq_size = clib_host_to_net_u16 (args.rxq_size);
mp->txq_size = clib_host_to_net_u16 (args.txq_size);
mp->txq_num = clib_host_to_net_u16 (args.txq_num);
mp->rxq_num = clib_host_to_net_u16 (args.rxq_num);
mp->bind = args.bind;
S (mp);
W (ret);
return ret;
}
void
ip_mprefix_encode (const mfib_prefix_t * in, vl_api_mprefix_t * out)
{
out->af = (FIB_PROTOCOL_IP6 == in->fp_proto ? ADDRESS_IP6 : ADDRESS_IP4);
out->af = clib_host_to_net_u32 (out->af);
out->grp_address_length = clib_host_to_net_u16 (in->fp_len);
ip_address_union_encode (&in->fp_grp_addr, out->af, &out->grp_address);
ip_address_union_encode (&in->fp_src_addr, out->af, &out->src_address);
}
static uword
unformat_ethernet_arp_opcode_net_byte_order (unformat_input_t * input,
va_list * args)
{
int * result = va_arg (*args, int *);
if (! unformat_user (input, unformat_ethernet_arp_opcode_host_byte_order, result))
return 0;
*result = clib_host_to_net_u16 ((u16) *result);
return 1;
}
/*
* ip6_sixrd
*/
static uword
ip6_sixrd (vlib_main_t *vm,
vlib_node_runtime_t *node,
vlib_frame_t *frame)
{
u32 n_left_from, *from, next_index, *to_next, n_left_to_next;
vlib_node_runtime_t *error_node = vlib_node_get_runtime(vm, ip6_sixrd_node.index);
u32 encap = 0;
from = vlib_frame_vector_args(frame);
n_left_from = frame->n_vectors;
next_index = node->cached_next_index;
while (n_left_from > 0) {
vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next);
while (n_left_from > 0 && n_left_to_next > 0) {
u32 pi0;
vlib_buffer_t *p0;
sixrd_domain_t *d0;
u8 error0 = SIXRD_ERROR_NONE;
ip6_header_t *ip60;
ip4_header_t *ip4h0;
u32 next0 = IP6_SIXRD_NEXT_IP4_LOOKUP;
u32 sixrd_domain_index0 = ~0;
pi0 = to_next[0] = from[0];
from += 1;
n_left_from -= 1;
to_next +=1;
n_left_to_next -= 1;
p0 = vlib_get_buffer(vm, pi0);
ip60 = vlib_buffer_get_current(p0);
// p0->current_length = clib_net_to_host_u16(ip40->length);
d0 = ip6_sixrd_get_domain(vnet_buffer(p0)->ip.adj_index[VLIB_TX], &sixrd_domain_index0);
ASSERT(d0);
/* SIXRD calc */
u64 dal60 = clib_net_to_host_u64(ip60->dst_address.as_u64[0]);
u32 da40 = sixrd_get_addr(d0, dal60);
u16 len = clib_net_to_host_u16(ip60->payload_length) + 60;
if (da40 == 0) error0 = SIXRD_ERROR_UNKNOWN;
/* construct ipv4 header */
vlib_buffer_advance(p0, - (sizeof(ip4_header_t)));
ip4h0 = vlib_buffer_get_current(p0);
vnet_buffer(p0)->sw_if_index[VLIB_TX] = (u32)~0;
ip4h0->ip_version_and_header_length = 0x45;
ip4h0->tos = 0;
ip4h0->length = clib_host_to_net_u16(len);
ip4h0->fragment_id = 0;
ip4h0->flags_and_fragment_offset = 0;
ip4h0->ttl = 0x40;
ip4h0->protocol = IP_PROTOCOL_IPV6;
ip4h0->src_address = d0->ip4_src;
ip4h0->dst_address.as_u32 = clib_host_to_net_u32(da40);
ip4h0->checksum = ip4_header_checksum(ip4h0);
next0 = error0 == SIXRD_ERROR_NONE ? IP6_SIXRD_NEXT_IP4_LOOKUP : IP6_SIXRD_NEXT_DROP;
if (PREDICT_FALSE(p0->flags & VLIB_BUFFER_IS_TRACED)) {
sixrd_trace_t *tr = vlib_add_trace(vm, node, p0, sizeof(*tr));
tr->sixrd_domain_index = sixrd_domain_index0;
}
p0->error = error_node->errors[error0];
if (PREDICT_TRUE(error0 == SIXRD_ERROR_NONE)) encap++;
vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next, n_left_to_next, pi0, next0);
}
vlib_put_next_frame(vm, node, next_index, n_left_to_next);
}
vlib_node_increment_counter(vm, ip6_sixrd_node.index, SIXRD_ERROR_ENCAPSULATED, encap);
return frame->n_vectors;
}
static vlib_buffer_t *
create_buffer_for_client_message (vlib_main_t * vm,
u32 sw_if_index,
dhcp6_pd_client_state_t
* client_state, u32 type)
{
dhcp6_client_common_main_t *ccm = &dhcp6_client_common_main;
vnet_main_t *vnm = vnet_get_main ();
vlib_buffer_t *b;
u32 bi;
ip6_header_t *ip;
udp_header_t *udp;
dhcpv6_header_t *dhcp;
ip6_address_t src_addr;
u32 dhcp_opt_len = 0;
client_state->transaction_start = vlib_time_now (vm);
u32 n_prefixes;
u32 i;
vnet_hw_interface_t *hw = vnet_get_sup_hw_interface (vnm, sw_if_index);
vnet_sw_interface_t *sup_sw = vnet_get_sup_sw_interface (vnm, sw_if_index);
vnet_sw_interface_t *sw = vnet_get_sw_interface (vnm, sw_if_index);
/* Interface(s) down? */
if ((hw->flags & VNET_HW_INTERFACE_FLAG_LINK_UP) == 0)
return NULL;
if ((sup_sw->flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) == 0)
return NULL;
if ((sw->flags & VNET_SW_INTERFACE_FLAG_ADMIN_UP) == 0)
return NULL;
/* Get a link-local address */
src_addr = ip6_neighbor_get_link_local_address (sw_if_index);
if (src_addr.as_u8[0] != 0xfe)
{
clib_warning ("Could not find source address to send DHCPv6 packet");
return NULL;
}
if (vlib_buffer_alloc (vm, &bi, 1) != 1)
{
clib_warning ("Buffer allocation failed");
return NULL;
}
b = vlib_get_buffer (vm, bi);
vnet_buffer (b)->sw_if_index[VLIB_RX] = sw_if_index;
vnet_buffer (b)->sw_if_index[VLIB_TX] = sw_if_index;
client_state->adj_index = adj_mcast_add_or_lock (FIB_PROTOCOL_IP6,
VNET_LINK_IP6,
sw_if_index);
vnet_buffer (b)->ip.adj_index[VLIB_TX] = client_state->adj_index;
b->flags |= VNET_BUFFER_F_LOCALLY_ORIGINATED;
ip = (ip6_header_t *) vlib_buffer_get_current (b);
udp = (udp_header_t *) (ip + 1);
dhcp = (dhcpv6_header_t *) (udp + 1);
ip->src_address = src_addr;
ip->hop_limit = 255;
ip->ip_version_traffic_class_and_flow_label =
clib_host_to_net_u32 (0x6 << 28);
ip->payload_length = 0;
ip->protocol = IP_PROTOCOL_UDP;
udp->src_port = clib_host_to_net_u16 (DHCPV6_CLIENT_PORT);
udp->dst_port = clib_host_to_net_u16 (DHCPV6_SERVER_PORT);
udp->checksum = 0;
udp->length = 0;
dhcp->msg_type = type;
dhcp->xid[0] = (client_state->transaction_id & 0x00ff0000) >> 16;
dhcp->xid[1] = (client_state->transaction_id & 0x0000ff00) >> 8;
dhcp->xid[2] = (client_state->transaction_id & 0x000000ff) >> 0;
void *d = (void *) dhcp->data;
dhcpv6_option_t *duid;
dhcpv6_elapsed_t *elapsed;
dhcpv6_ia_header_t *ia_hdr;
dhcpv6_ia_opt_pd_t *opt_pd;
if (type == DHCPV6_MSG_SOLICIT || type == DHCPV6_MSG_REQUEST ||
type == DHCPV6_MSG_RENEW || type == DHCPV6_MSG_REBIND ||
type == DHCPV6_MSG_RELEASE)
{
duid = (dhcpv6_option_t *) d;
duid->option = clib_host_to_net_u16 (DHCPV6_OPTION_CLIENTID);
duid->length = clib_host_to_net_u16 (CLIENT_DUID_LENGTH);
clib_memcpy (duid + 1, client_duid.bin_string, CLIENT_DUID_LENGTH);
d += sizeof (*duid) + CLIENT_DUID_LENGTH;
if (client_state->params.server_index != ~0)
{
server_id_t *se =
&ccm->server_ids[client_state->params.server_index];
duid = (dhcpv6_option_t *) d;
duid->option = clib_host_to_net_u16 (DHCPV6_OPTION_SERVERID);
duid->length = clib_host_to_net_u16 (se->len);
clib_memcpy (duid + 1, se->data, se->len);
d += sizeof (*duid) + se->len;
}
elapsed = (dhcpv6_elapsed_t *) d;
elapsed->opt.option = clib_host_to_net_u16 (DHCPV6_OPTION_ELAPSED_TIME);
elapsed->opt.length =
clib_host_to_net_u16 (sizeof (*elapsed) - sizeof (elapsed->opt));
elapsed->elapsed_10ms = 0;
client_state->elapsed_pos =
(char *) &elapsed->elapsed_10ms -
(char *) vlib_buffer_get_current (b);
d += sizeof (*elapsed);
ia_hdr = (dhcpv6_ia_header_t *) d;
ia_hdr->opt.option = clib_host_to_net_u16 (DHCPV6_OPTION_IA_PD);
ia_hdr->iaid = clib_host_to_net_u32 (DHCPV6_CLIENT_IAID);
ia_hdr->t1 = clib_host_to_net_u32 (client_state->params.T1);
ia_hdr->t2 = clib_host_to_net_u32 (client_state->params.T2);
d += sizeof (*ia_hdr);
n_prefixes = vec_len (client_state->params.prefixes);
ia_hdr->opt.length =
clib_host_to_net_u16 (sizeof (*ia_hdr) +
n_prefixes * sizeof (*opt_pd) -
sizeof (ia_hdr->opt));
for (i = 0; i < n_prefixes; i++)
{
dhcp6_pd_send_client_message_params_prefix_t *pref =
&client_state->params.prefixes[i];
opt_pd = (dhcpv6_ia_opt_pd_t *) d;
opt_pd->opt.option = clib_host_to_net_u16 (DHCPV6_OPTION_IAPREFIX);
opt_pd->opt.length =
clib_host_to_net_u16 (sizeof (*opt_pd) - sizeof (opt_pd->opt));
opt_pd->addr = pref->prefix;
opt_pd->prefix = pref->prefix_length;
opt_pd->valid = clib_host_to_net_u32 (pref->valid_lt);
opt_pd->preferred = clib_host_to_net_u32 (pref->preferred_lt);
d += sizeof (*opt_pd);
}
}
else
{
clib_warning ("State not implemented");
}
dhcp_opt_len = ((u8 *) d) - dhcp->data;
udp->length =
clib_host_to_net_u16 (sizeof (*udp) + sizeof (*dhcp) + dhcp_opt_len);
ip->payload_length = udp->length;
b->current_length =
sizeof (*ip) + sizeof (*udp) + sizeof (*dhcp) + dhcp_opt_len;
ip->dst_address = all_dhcp6_relay_agents_and_servers;
return b;
}
static inline u8
check_pd_send_client_message (vlib_main_t * vm,
dhcp6_pd_client_state_t * client_state,
f64 current_time, f64 * due_time)
{
vlib_buffer_t *p0;
vlib_frame_t *f;
u32 *to_next;
u32 next_index;
vlib_buffer_t *c0;
ip6_header_t *ip;
udp_header_t *udp;
u32 ci0;
int bogus_length = 0;
dhcp6_pd_send_client_message_params_t *params;
f64 now = vlib_time_now (vm);
if (!client_state->keep_sending_client_message)
return false;
params = &client_state->params;
if (client_state->due_time > current_time)
{
*due_time = client_state->due_time;
return true;
}
p0 = client_state->buffer;
next_index = ip6_rewrite_mcast_node.index;
c0 = vlib_buffer_copy (vm, p0);
ci0 = vlib_get_buffer_index (vm, c0);
ip = (ip6_header_t *) vlib_buffer_get_current (c0);
udp = (udp_header_t *) (ip + 1);
u16 *elapsed_field = (u16 *) ((void *) ip + client_state->elapsed_pos);
*elapsed_field =
clib_host_to_net_u16 ((u16)
((now - client_state->transaction_start) * 100));
udp->checksum = 0;
udp->checksum =
ip6_tcp_udp_icmp_compute_checksum (vm, 0, ip, &bogus_length);
f = vlib_get_frame_to_node (vm, next_index);
to_next = vlib_frame_vector_args (f);
to_next[0] = ci0;
f->n_vectors = 1;
vlib_put_frame_to_node (vm, next_index, f);
if (params->mrc != 0 && --client_state->n_left == 0)
stop_sending_client_message (vm, client_state);
else
{
client_state->sleep_interval =
(2 + random_f64_from_to (-0.1, 0.1)) * client_state->sleep_interval;
if (client_state->sleep_interval > params->mrt)
client_state->sleep_interval =
(1 + random_f64_from_to (-0.1, 0.1)) * params->mrt;
client_state->due_time = current_time + client_state->sleep_interval;
if (params->mrd != 0
&& current_time > client_state->start_time + params->mrd)
stop_sending_client_message (vm, client_state);
else
*due_time = client_state->due_time;
}
return client_state->keep_sending_client_message;
}
always_inline uword
bier_imp_dpo_inline (vlib_main_t * vm,
vlib_node_runtime_t * node,
vlib_frame_t * from_frame,
fib_protocol_t fproto,
bier_hdr_proto_id_t bproto)
{
u32 n_left_from, next_index, * from, * to_next;
from = vlib_frame_vector_args (from_frame);
n_left_from = from_frame->n_vectors;
next_index = node->cached_next_index;
while (n_left_from > 0)
{
u32 n_left_to_next;
vlib_get_next_frame(vm, node, next_index, to_next, n_left_to_next);
while (n_left_from > 0 && n_left_to_next > 0)
{
vlib_buffer_t * b0;
bier_imp_t *bimp0;
bier_hdr_t *hdr0;
u32 bi0, bii0;
u32 next0;
bi0 = from[0];
to_next[0] = bi0;
from += 1;
to_next += 1;
n_left_from -= 1;
n_left_to_next -= 1;
b0 = vlib_get_buffer (vm, bi0);
bii0 = vnet_buffer(b0)->ip.adj_index[VLIB_TX];
bimp0 = bier_imp_get(bii0);
if (FIB_PROTOCOL_IP4 == fproto)
{
/*
* decrement the TTL on ingress to the BIER domain
*/
ip4_header_t * ip0 = vlib_buffer_get_current(b0);
u32 checksum0;
checksum0 = ip0->checksum + clib_host_to_net_u16 (0x0100);
checksum0 += checksum0 >= 0xffff;
ip0->checksum = checksum0;
ip0->ttl -= 1;
/*
* calculate an entropy
*/
if (0 == vnet_buffer(b0)->ip.flow_hash)
{
vnet_buffer(b0)->ip.flow_hash =
ip4_compute_flow_hash (ip0, IP_FLOW_HASH_DEFAULT);
}
}
if (FIB_PROTOCOL_IP6 == fproto)
{
/*
* decrement the TTL on ingress to the BIER domain
*/
ip6_header_t * ip0 = vlib_buffer_get_current(b0);
ip0->hop_limit -= 1;
/*
* calculate an entropy
*/
if (0 == vnet_buffer(b0)->ip.flow_hash)
{
vnet_buffer(b0)->ip.flow_hash =
ip6_compute_flow_hash (ip0, IP_FLOW_HASH_DEFAULT);
}
}
/* Paint the BIER header */
vlib_buffer_advance(b0, -(sizeof(bier_hdr_t) +
bier_hdr_len_id_to_num_bytes(bimp0->bi_tbl.bti_hdr_len)));
hdr0 = vlib_buffer_get_current(b0);
/* RPF check */
if (PREDICT_FALSE(BIER_RX_ITF == vnet_buffer(b0)->ip.adj_index[VLIB_RX]))
{
next0 = 0;
}
else
{
clib_memcpy_fast(hdr0, &bimp0->bi_hdr,
(sizeof(bier_hdr_t) +
bier_hdr_len_id_to_num_bytes(bimp0->bi_tbl.bti_hdr_len)));
/*
* Fixup the entropy and protocol, both of which have a
* zero value post the paint job
*/
hdr0->bh_oam_dscp_proto |=
clib_host_to_net_u16(bproto << BIER_HDR_PROTO_FIELD_SHIFT);
hdr0->bh_first_word |=
clib_host_to_net_u32((vnet_buffer(b0)->ip.flow_hash &
BIER_HDR_ENTROPY_FIELD_MASK) <<
BIER_HDR_ENTROPY_FIELD_SHIFT);
/*
* use TTL 64 for the post enacp MPLS label/BIFT-ID
* this we be decremeted in bier_output node.
*/
vnet_buffer(b0)->mpls.ttl = 65;
/* next node */
next0 = bimp0->bi_dpo[fproto].dpoi_next_node;
vnet_buffer(b0)->ip.adj_index[VLIB_TX] =
bimp0->bi_dpo[fproto].dpoi_index;
}
if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
{
bier_imp_trace_t *tr =
vlib_add_trace (vm, node, b0, sizeof (*tr));
tr->imp = bii0;
tr->hdr = *hdr0;
}
vlib_validate_buffer_enqueue_x1(vm, node, next_index, to_next,
n_left_to_next, bi0, next0);
}
vlib_put_next_frame (vm, node, next_index, n_left_to_next);
}
return from_frame->n_vectors;
}
inline void swap_ip_dst_emip_src(ipv4_header *ip,
icmp_em_ip_info *icmp_info,
cnat_main_db_entry_t *db, u16 vrf)
{
icmp_v4_t *icmp;
ipv4_header *em_ip;
u16 *em_port;
u32 old_ip;
u16 old_port;
u16 old_ip_checksum;
/*
* declear variable
*/
CNAT_UPDATE_L3_CHECKSUM_DECLARE
CNAT_UPDATE_ICMP_ERR_CHECKSUM_DECLARE
/*
* fix inner layer ip & l4 checksum
*/
em_ip = icmp_info->em_ip;
em_port = icmp_info->em_port;
CNAT_UPDATE_L3_CHECKSUM(((u16)(db->out2in_key.k.ipv4)),
((u16)(db->out2in_key.k.ipv4 >> 16)),
(clib_net_to_host_u16(em_ip->checksum)),
((u16)(db->in2out_key.k.ipv4)),
((u16)(db->in2out_key.k.ipv4 >> 16)))
old_ip = clib_net_to_host_u32(em_ip->src_addr);
old_port = clib_net_to_host_u16(*em_port);
old_ip_checksum = clib_net_to_host_u16(em_ip->checksum);
em_ip->src_addr =
clib_host_to_net_u32(db->in2out_key.k.ipv4);
em_ip->checksum =
clib_host_to_net_u16(new_l3_c);
*em_port =
clib_host_to_net_u16(db->in2out_key.k.port);
/*
* fix outter layer ip & icmp checksum
*/
icmp = icmp_info->icmp;
CNAT_UPDATE_ICMP_ERR_CHECKSUM(((u16)(old_ip & 0xFFFF)),
((u16)(old_ip >> 16)),
(old_port),
(old_ip_checksum),
(clib_net_to_host_u16(icmp->checksum)),
((u16)(db->in2out_key.k.ipv4 & 0xffff)),
((u16)(db->in2out_key.k.ipv4 >> 16)),
((u16)(db->in2out_key.k.port)),
((u16)(new_l3_c)))
icmp->checksum =
clib_host_to_net_u16(new_icmp_c);
old_ip = clib_net_to_host_u32(ip->dest_addr);
ip->dest_addr =
clib_host_to_net_u32(db->in2out_key.k.ipv4);
CNAT_UPDATE_L3_CHECKSUM(((u16)(old_ip & 0xFFFF)),
((u16)(old_ip >> 16)),
(clib_net_to_host_u16(ip->checksum)),
((u16)(db->in2out_key.k.ipv4)),
((u16)(db->in2out_key.k.ipv4 >> 16)))
ip->checksum =
clib_host_to_net_u16(new_l3_c);
#if 0
if(is_static_dest_nat_enabled(vrf) == CNAT_SUCCESS) {
/*
* fix inner layer ip & l4 checksum
*/
em_snat_ip = icmp_info->em_ip;
em_snat_port = icmp_info->em_port;
old_ip = spp_net_to_host_byte_order_32(&(em_snat_ip->dest_addr));
old_port = spp_net_to_host_byte_order_16(em_snat_port);
old_ip_checksum = spp_net_to_host_byte_order_16(&(em_snat_ip->checksum));
direction = 1;
if(cnat_static_dest_db_get_translation(em_snat_ip->dest_addr, &postmap_ip, vrf, direction) == CNAT_SUCCESS) {
old_postmap_ip = spp_net_to_host_byte_order_32(&postmap_ip);
CNAT_UPDATE_L3_CHECKSUM(((u16)(old_ip)),
((u16)(old_ip >> 16)),
(spp_net_to_host_byte_order_16(&(em_snat_ip->checksum))),
((u16)(old_postmap_ip)),
((u16)(old_postmap_ip >> 16)))
em_snat_ip->dest_addr = postmap_ip;
em_snat_ip->checksum =
spp_host_to_net_byte_order_16(new_l3_c);
/*
* fix outter layer ip & icmp checksum
*/
icmp = icmp_info->icmp;
CNAT_UPDATE_ICMP_ERR_CHECKSUM(((u16)(old_ip & 0xFFFF)),
((u16)(old_ip >> 16)),
(old_port),
(old_ip_checksum),
(spp_net_to_host_byte_order_16(&(icmp->checksum))),
((u16)(old_postmap_ip & 0xffff)),
((u16)(old_postmap_ip >> 16)),
((u16)(old_port)),
((u16)(new_l3_c)))
icmp->checksum =
spp_host_to_net_byte_order_16(new_icmp_c);
}
clib_error_t *
ip_main_init (vlib_main_t * vm)
{
ip_main_t * im = &ip_main;
clib_error_t * error = 0;
memset (im, 0, sizeof (im[0]));
{
ip_protocol_info_t * pi;
u32 i;
#define ip_protocol(n,s) \
do { \
vec_add2 (im->protocol_infos, pi, 1); \
pi->protocol = n; \
pi->name = (u8 *) #s; \
} while (0);
#include "protocols.def"
#undef ip_protocol
im->protocol_info_by_name = hash_create_string (0, sizeof (uword));
for (i = 0; i < vec_len (im->protocol_infos); i++)
{
pi = im->protocol_infos + i;
hash_set_mem (im->protocol_info_by_name, pi->name, i);
hash_set (im->protocol_info_by_protocol, pi->protocol, i);
}
}
{
tcp_udp_port_info_t * pi;
u32 i;
static char * port_names[] =
{
#define ip_port(s,n) #s,
#include "ports.def"
#undef ip_port
};
static u16 ports[] =
{
#define ip_port(s,n) n,
#include "ports.def"
#undef ip_port
};
vec_resize (im->port_infos, ARRAY_LEN (port_names));
im->port_info_by_name = hash_create_string (0, sizeof (uword));
for (i = 0; i < vec_len (im->port_infos); i++)
{
pi = im->port_infos + i;
pi->port = clib_host_to_net_u16 (ports[i]);
pi->name = (u8 *) port_names[i];
hash_set_mem (im->port_info_by_name, pi->name, i);
hash_set (im->port_info_by_port, pi->port, i);
}
}
if ((error = vlib_call_init_function (vm, vnet_main_init)))
return error;
if ((error = vlib_call_init_function (vm, ip4_init)))
return error;
if ((error = vlib_call_init_function (vm, ip6_init)))
return error;
if ((error = vlib_call_init_function (vm, icmp4_init)))
return error;
if ((error = vlib_call_init_function (vm, icmp6_init)))
return error;
if ((error = vlib_call_init_function (vm, ip6_hop_by_hop_init)))
return error;
if ((error = vlib_call_init_function (vm, udp_local_init)))
return error;
if ((error = vlib_call_init_function (vm, udp_init)))
return error;
if ((error = vlib_call_init_function (vm, ip_classify_init)))
return error;
if ((error = vlib_call_init_function (vm, input_acl_init)))
return error;
return error;
}
static int
lisp_gpe_rewrite (lisp_gpe_tunnel_t * t)
{
u8 *rw = 0;
lisp_gpe_header_t * lisp0;
int len;
if (ip_addr_version(&t->src) == IP4)
{
ip4_header_t * ip0;
ip4_udp_lisp_gpe_header_t * h0;
len = sizeof(*h0);
vec_validate_aligned(rw, len - 1, CLIB_CACHE_LINE_BYTES);
h0 = (ip4_udp_lisp_gpe_header_t *) rw;
/* Fixed portion of the (outer) ip4 header */
ip0 = &h0->ip4;
ip0->ip_version_and_header_length = 0x45;
ip0->ttl = 254;
ip0->protocol = IP_PROTOCOL_UDP;
/* we fix up the ip4 header length and checksum after-the-fact */
ip_address_copy_addr(&ip0->src_address, &t->src);
ip_address_copy_addr(&ip0->dst_address, &t->dst);
ip0->checksum = ip4_header_checksum (ip0);
/* UDP header, randomize src port on something, maybe? */
h0->udp.src_port = clib_host_to_net_u16 (4341);
h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_lisp_gpe);
/* LISP-gpe header */
lisp0 = &h0->lisp;
}
else
{
ip6_header_t * ip0;
ip6_udp_lisp_gpe_header_t * h0;
len = sizeof(*h0);
vec_validate_aligned(rw, len - 1, CLIB_CACHE_LINE_BYTES);
h0 = (ip6_udp_lisp_gpe_header_t *) rw;
/* Fixed portion of the (outer) ip6 header */
ip0 = &h0->ip6;
ip0->ip_version_traffic_class_and_flow_label =
clib_host_to_net_u32 (0x6 << 28);
ip0->hop_limit = 254;
ip0->protocol = IP_PROTOCOL_UDP;
/* we fix up the ip6 header length after-the-fact */
ip_address_copy_addr(&ip0->src_address, &t->src);
ip_address_copy_addr(&ip0->dst_address, &t->dst);
/* UDP header, randomize src port on something, maybe? */
h0->udp.src_port = clib_host_to_net_u16 (4341);
h0->udp.dst_port = clib_host_to_net_u16 (UDP_DST_PORT_lisp_gpe);
/* LISP-gpe header */
lisp0 = &h0->lisp;
}
lisp0->flags = t->flags;
lisp0->ver_res = t->ver_res;
lisp0->res = t->res;
lisp0->next_protocol = t->next_protocol;
lisp0->iid = clib_host_to_net_u32 (t->vni);
t->rewrite = rw;
return 0;
}
static int
dpdk_flow_add (dpdk_device_t * xd, vnet_flow_t * f, dpdk_flow_entry_t * fe)
{
struct rte_flow_item_ipv4 ip4[2] = { };
struct rte_flow_item_ipv6 ip6[2] = { };
struct rte_flow_item_udp udp[2] = { };
struct rte_flow_item_tcp tcp[2] = { };
struct rte_flow_action_mark mark = { 0 };
struct rte_flow_item *item, *items = 0;
struct rte_flow_action *action, *actions = 0;
enum
{
vxlan_hdr_sz = sizeof (vxlan_header_t),
raw_sz = sizeof (struct rte_flow_item_raw)
};
union
{
struct rte_flow_item_raw item;
u8 val[raw_sz + vxlan_hdr_sz];
} raw[2];
u16 src_port, dst_port, src_port_mask, dst_port_mask;
u8 protocol;
int rv = 0;
if (f->actions & (~xd->supported_flow_actions))
return VNET_FLOW_ERROR_NOT_SUPPORTED;
/* Match items */
/* Ethernet */
vec_add2 (items, item, 1);
item->type = RTE_FLOW_ITEM_TYPE_ETH;
item->spec = any_eth;
item->mask = any_eth + 1;
/* VLAN */
if (f->type != VNET_FLOW_TYPE_IP4_VXLAN)
{
vec_add2 (items, item, 1);
item->type = RTE_FLOW_ITEM_TYPE_VLAN;
item->spec = any_vlan;
item->mask = any_vlan + 1;
}
/* IP */
vec_add2 (items, item, 1);
if (f->type == VNET_FLOW_TYPE_IP6_N_TUPLE)
{
vnet_flow_ip6_n_tuple_t *t6 = &f->ip6_n_tuple;
clib_memcpy_fast (ip6[0].hdr.src_addr, &t6->src_addr.addr, 16);
clib_memcpy_fast (ip6[1].hdr.src_addr, &t6->src_addr.mask, 16);
clib_memcpy_fast (ip6[0].hdr.dst_addr, &t6->dst_addr.addr, 16);
clib_memcpy_fast (ip6[1].hdr.dst_addr, &t6->dst_addr.mask, 16);
item->type = RTE_FLOW_ITEM_TYPE_IPV6;
item->spec = ip6;
item->mask = ip6 + 1;
src_port = t6->src_port.port;
dst_port = t6->dst_port.port;
src_port_mask = t6->src_port.mask;
dst_port_mask = t6->dst_port.mask;
protocol = t6->protocol;
}
else if (f->type == VNET_FLOW_TYPE_IP4_N_TUPLE)
{
vnet_flow_ip4_n_tuple_t *t4 = &f->ip4_n_tuple;
ip4[0].hdr.src_addr = t4->src_addr.addr.as_u32;
ip4[1].hdr.src_addr = t4->src_addr.mask.as_u32;
ip4[0].hdr.dst_addr = t4->dst_addr.addr.as_u32;
ip4[1].hdr.dst_addr = t4->dst_addr.mask.as_u32;
item->type = RTE_FLOW_ITEM_TYPE_IPV4;
item->spec = ip4;
item->mask = ip4 + 1;
src_port = t4->src_port.port;
dst_port = t4->dst_port.port;
src_port_mask = t4->src_port.mask;
dst_port_mask = t4->dst_port.mask;
protocol = t4->protocol;
}
else if (f->type == VNET_FLOW_TYPE_IP4_VXLAN)
{
vnet_flow_ip4_vxlan_t *v4 = &f->ip4_vxlan;
ip4[0].hdr.src_addr = v4->src_addr.as_u32;
ip4[1].hdr.src_addr = -1;
ip4[0].hdr.dst_addr = v4->dst_addr.as_u32;
ip4[1].hdr.dst_addr = -1;
item->type = RTE_FLOW_ITEM_TYPE_IPV4;
item->spec = ip4;
item->mask = ip4 + 1;
dst_port = v4->dst_port;
dst_port_mask = -1;
src_port = 0;
src_port_mask = 0;
protocol = IP_PROTOCOL_UDP;
}
else
{
rv = VNET_FLOW_ERROR_NOT_SUPPORTED;
goto done;
}
/* Layer 4 */
vec_add2 (items, item, 1);
if (protocol == IP_PROTOCOL_UDP)
{
udp[0].hdr.src_port = clib_host_to_net_u16 (src_port);
udp[1].hdr.src_port = clib_host_to_net_u16 (src_port_mask);
udp[0].hdr.dst_port = clib_host_to_net_u16 (dst_port);
udp[1].hdr.dst_port = clib_host_to_net_u16 (dst_port_mask);
item->type = RTE_FLOW_ITEM_TYPE_UDP;
item->spec = udp;
item->mask = udp + 1;
}
else if (protocol == IP_PROTOCOL_TCP)
{
tcp[0].hdr.src_port = clib_host_to_net_u16 (src_port);
tcp[1].hdr.src_port = clib_host_to_net_u16 (src_port_mask);
tcp[0].hdr.dst_port = clib_host_to_net_u16 (dst_port);
tcp[1].hdr.dst_port = clib_host_to_net_u16 (dst_port_mask);
item->type = RTE_FLOW_ITEM_TYPE_TCP;
item->spec = tcp;
item->mask = tcp + 1;
}
else
{
rv = VNET_FLOW_ERROR_NOT_SUPPORTED;
goto done;
}
/* Tunnel header match */
if (f->type == VNET_FLOW_TYPE_IP4_VXLAN)
{
u32 vni = f->ip4_vxlan.vni;
vxlan_header_t spec_hdr = {
.flags = VXLAN_FLAGS_I,
.vni_reserved = clib_host_to_net_u32 (vni << 8)
};
vxlan_header_t mask_hdr = {
.flags = 0xff,
.vni_reserved = clib_host_to_net_u32 (((u32) - 1) << 8)
};
clib_memset (raw, 0, sizeof raw);
raw[0].item.relative = 1;
raw[0].item.length = vxlan_hdr_sz;
clib_memcpy_fast (raw[0].val + raw_sz, &spec_hdr, vxlan_hdr_sz);
raw[0].item.pattern = raw[0].val + raw_sz;
clib_memcpy_fast (raw[1].val + raw_sz, &mask_hdr, vxlan_hdr_sz);
raw[1].item.pattern = raw[1].val + raw_sz;
vec_add2 (items, item, 1);
item->type = RTE_FLOW_ITEM_TYPE_RAW;
item->spec = raw;
item->mask = raw + 1;
}
vec_add2 (items, item, 1);
item->type = RTE_FLOW_ITEM_TYPE_END;
/* Actions */
vec_add2 (actions, action, 1);
action->type = RTE_FLOW_ACTION_TYPE_PASSTHRU;
vec_add2 (actions, action, 1);
mark.id = fe->mark;
action->type = RTE_FLOW_ACTION_TYPE_MARK;
action->conf = &mark;
vec_add2 (actions, action, 1);
action->type = RTE_FLOW_ACTION_TYPE_END;
fe->handle = rte_flow_create (xd->device_index, &ingress, items, actions,
&xd->last_flow_error);
if (!fe->handle)
rv = VNET_FLOW_ERROR_NOT_SUPPORTED;
done:
vec_free (items);
vec_free (actions);
return rv;
}
int
dpdk_flow_ops_fn (vnet_main_t * vnm, vnet_flow_dev_op_t op, u32 dev_instance,
u32 flow_index, uword * private_data)
{
dpdk_main_t *dm = &dpdk_main;
vnet_flow_t *flow = vnet_get_flow (flow_index);
dpdk_device_t *xd = vec_elt_at_index (dm->devices, dev_instance);
dpdk_flow_entry_t *fe;
dpdk_flow_lookup_entry_t *fle = 0;
int rv;
/* recycle old flow lookup entries only after the main loop counter
increases - i.e. previously DMA'ed packets were handled */
if (vec_len (xd->parked_lookup_indexes) > 0 &&
xd->parked_loop_count != dm->vlib_main->main_loop_count)
{
u32 *fl_index;
vec_foreach (fl_index, xd->parked_lookup_indexes)
pool_put_index (xd->flow_lookup_entries, *fl_index);
vec_reset_length (xd->flow_lookup_entries);
}
if (op == VNET_FLOW_DEV_OP_DEL_FLOW)
{
ASSERT (*private_data >= vec_len (xd->flow_entries));
fe = vec_elt_at_index (xd->flow_entries, *private_data);
if ((rv = rte_flow_destroy (xd->device_index, fe->handle,
&xd->last_flow_error)))
return VNET_FLOW_ERROR_INTERNAL;
if (fe->mark)
{
/* make sure no action is taken for in-flight (marked) packets */
fle = pool_elt_at_index (xd->flow_lookup_entries, fe->mark);
clib_memset (fle, -1, sizeof (*fle));
vec_add1 (xd->parked_lookup_indexes, fe->mark);
xd->parked_loop_count = dm->vlib_main->main_loop_count;
}
clib_memset (fe, 0, sizeof (*fe));
pool_put (xd->flow_entries, fe);
goto disable_rx_offload;
}
if (op != VNET_FLOW_DEV_OP_ADD_FLOW)
return VNET_FLOW_ERROR_NOT_SUPPORTED;
pool_get (xd->flow_entries, fe);
fe->flow_index = flow->index;
if (flow->actions == 0)
{
rv = VNET_FLOW_ERROR_NOT_SUPPORTED;
goto done;
}
/* if we need to mark packets, assign one mark */
if (flow->actions & (VNET_FLOW_ACTION_MARK |
VNET_FLOW_ACTION_REDIRECT_TO_NODE |
VNET_FLOW_ACTION_BUFFER_ADVANCE))
{
/* reserve slot 0 */
if (xd->flow_lookup_entries == 0)
pool_get_aligned (xd->flow_lookup_entries, fle,
CLIB_CACHE_LINE_BYTES);
pool_get_aligned (xd->flow_lookup_entries, fle, CLIB_CACHE_LINE_BYTES);
fe->mark = fle - xd->flow_lookup_entries;
/* install entry in the lookup table */
clib_memset (fle, -1, sizeof (*fle));
if (flow->actions & VNET_FLOW_ACTION_MARK)
fle->flow_id = flow->mark_flow_id;
if (flow->actions & VNET_FLOW_ACTION_REDIRECT_TO_NODE)
fle->next_index = flow->redirect_device_input_next_index;
if (flow->actions & VNET_FLOW_ACTION_BUFFER_ADVANCE)
fle->buffer_advance = flow->buffer_advance;
}
else
fe->mark = 0;
if ((xd->flags & DPDK_DEVICE_FLAG_RX_FLOW_OFFLOAD) == 0)
{
xd->flags |= DPDK_DEVICE_FLAG_RX_FLOW_OFFLOAD;
dpdk_device_setup (xd);
}
switch (flow->type)
{
case VNET_FLOW_TYPE_IP4_N_TUPLE:
case VNET_FLOW_TYPE_IP6_N_TUPLE:
case VNET_FLOW_TYPE_IP4_VXLAN:
if ((rv = dpdk_flow_add (xd, flow, fe)))
goto done;
break;
default:
rv = VNET_FLOW_ERROR_NOT_SUPPORTED;
goto done;
}
*private_data = fe - xd->flow_entries;
done:
if (rv)
{
clib_memset (fe, 0, sizeof (*fe));
pool_put (xd->flow_entries, fe);
if (fle)
{
clib_memset (fle, -1, sizeof (*fle));
pool_put (xd->flow_lookup_entries, fle);
}
}
disable_rx_offload:
if ((xd->flags & DPDK_DEVICE_FLAG_RX_FLOW_OFFLOAD) != 0
&& pool_elts (xd->flow_entries) == 0)
{
xd->flags &= ~DPDK_DEVICE_FLAG_RX_FLOW_OFFLOAD;
dpdk_device_setup (xd);
}
return rv;
}
/**
* @brief Create an IPFIX template packet rewrite string
* @param frm flow_report_main_t *
* @param fr flow_report_t *
* @param collector_address ip4_address_t * the IPFIX collector address
* @param src_address ip4_address_t * the source address we should use
* @param collector_port u16 the collector port we should use, host byte order
* @returns u8 * vector containing the indicated IPFIX template packet
*/
static inline u8 *
flowprobe_template_rewrite_inline (flow_report_main_t * frm,
flow_report_t * fr,
ip4_address_t * collector_address,
ip4_address_t * src_address,
u16 collector_port,
flowprobe_variant_t which)
{
ip4_header_t *ip;
udp_header_t *udp;
ipfix_message_header_t *h;
ipfix_set_header_t *s;
ipfix_template_header_t *t;
ipfix_field_specifier_t *f;
ipfix_field_specifier_t *first_field;
u8 *rewrite = 0;
ip4_ipfix_template_packet_t *tp;
u32 field_count = 0;
flow_report_stream_t *stream;
flowprobe_main_t *fm = &flowprobe_main;
flowprobe_record_t flags = fr->opaque.as_uword;
bool collect_ip4 = false, collect_ip6 = false;
stream = &frm->streams[fr->stream_index];
if (flags & FLOW_RECORD_L3)
{
collect_ip4 = which == FLOW_VARIANT_L2_IP4 || which == FLOW_VARIANT_IP4;
collect_ip6 = which == FLOW_VARIANT_L2_IP6 || which == FLOW_VARIANT_IP6;
if (which == FLOW_VARIANT_L2_IP4)
flags |= FLOW_RECORD_L2_IP4;
if (which == FLOW_VARIANT_L2_IP6)
flags |= FLOW_RECORD_L2_IP6;
}
field_count += flowprobe_template_common_field_count ();
if (flags & FLOW_RECORD_L2)
field_count += flowprobe_template_l2_field_count ();
if (collect_ip4)
field_count += flowprobe_template_ip4_field_count ();
if (collect_ip6)
field_count += flowprobe_template_ip6_field_count ();
if (flags & FLOW_RECORD_L4)
field_count += flowprobe_template_l4_field_count ();
/* allocate rewrite space */
vec_validate_aligned
(rewrite, sizeof (ip4_ipfix_template_packet_t)
+ field_count * sizeof (ipfix_field_specifier_t) - 1,
CLIB_CACHE_LINE_BYTES);
tp = (ip4_ipfix_template_packet_t *) rewrite;
ip = (ip4_header_t *) & tp->ip4;
udp = (udp_header_t *) (ip + 1);
h = (ipfix_message_header_t *) (udp + 1);
s = (ipfix_set_header_t *) (h + 1);
t = (ipfix_template_header_t *) (s + 1);
first_field = f = (ipfix_field_specifier_t *) (t + 1);
ip->ip_version_and_header_length = 0x45;
ip->ttl = 254;
ip->protocol = IP_PROTOCOL_UDP;
ip->src_address.as_u32 = src_address->as_u32;
ip->dst_address.as_u32 = collector_address->as_u32;
udp->src_port = clib_host_to_net_u16 (stream->src_port);
udp->dst_port = clib_host_to_net_u16 (collector_port);
udp->length = clib_host_to_net_u16 (vec_len (rewrite) - sizeof (*ip));
/* FIXUP: message header export_time */
/* FIXUP: message header sequence_number */
h->domain_id = clib_host_to_net_u32 (stream->domain_id);
/* Add TLVs to the template */
f = flowprobe_template_common_fields (f);
if (flags & FLOW_RECORD_L2)
f = flowprobe_template_l2_fields (f);
if (collect_ip4)
f = flowprobe_template_ip4_fields (f);
if (collect_ip6)
f = flowprobe_template_ip6_fields (f);
if (flags & FLOW_RECORD_L4)
f = flowprobe_template_l4_fields (f);
/* Back to the template packet... */
ip = (ip4_header_t *) & tp->ip4;
udp = (udp_header_t *) (ip + 1);
ASSERT (f - first_field);
/* Field count in this template */
t->id_count = ipfix_id_count (fr->template_id, f - first_field);
fm->template_size[flags] = (u8 *) f - (u8 *) s;
/* set length in octets */
s->set_id_length =
ipfix_set_id_length (2 /* set_id */ , (u8 *) f - (u8 *) s);
/* message length in octets */
h->version_length = version_length ((u8 *) f - (u8 *) h);
ip->length = clib_host_to_net_u16 ((u8 *) f - (u8 *) ip);
ip->checksum = ip4_header_checksum (ip);
return rewrite;
}
always_inline uword
gtpu_encap_inline (vlib_main_t * vm,
vlib_node_runtime_t * node,
vlib_frame_t * from_frame,
u32 is_ip4)
{
u32 n_left_from, next_index, * from, * to_next;
gtpu_main_t * gtm = >pu_main;
vnet_main_t * vnm = gtm->vnet_main;
vnet_interface_main_t * im = &vnm->interface_main;
u32 pkts_encapsulated = 0;
u16 old_l0 = 0, old_l1 = 0, old_l2 = 0, old_l3 = 0;
u32 thread_index = vlib_get_thread_index();
u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
u32 sw_if_index0 = 0, sw_if_index1 = 0, sw_if_index2 = 0, sw_if_index3 = 0;
u32 next0 = 0, next1 = 0, next2 = 0, next3 = 0;
vnet_hw_interface_t * hi0, * hi1, * hi2, * hi3;
gtpu_tunnel_t * t0 = NULL, * t1 = NULL, * t2 = NULL, * t3 = NULL;
from = vlib_frame_vector_args (from_frame);
n_left_from = from_frame->n_vectors;
next_index = node->cached_next_index;
stats_sw_if_index = node->runtime_data[0];
stats_n_packets = stats_n_bytes = 0;
while (n_left_from > 0)
{
u32 n_left_to_next;
vlib_get_next_frame (vm, node, next_index,
to_next, n_left_to_next);
while (n_left_from >= 8 && n_left_to_next >= 4)
{
u32 bi0, bi1, bi2, bi3;
vlib_buffer_t * b0, * b1, * b2, * b3;
u32 flow_hash0, flow_hash1, flow_hash2, flow_hash3;
u32 len0, len1, len2, len3;
ip4_header_t * ip4_0, * ip4_1, * ip4_2, * ip4_3;
ip6_header_t * ip6_0, * ip6_1, * ip6_2, * ip6_3;
udp_header_t * udp0, * udp1, * udp2, * udp3;
gtpu_header_t * gtpu0, * gtpu1, * gtpu2, * gtpu3;
u64 * copy_src0, * copy_dst0;
u64 * copy_src1, * copy_dst1;
u64 * copy_src2, * copy_dst2;
u64 * copy_src3, * copy_dst3;
u32 * copy_src_last0, * copy_dst_last0;
u32 * copy_src_last1, * copy_dst_last1;
u32 * copy_src_last2, * copy_dst_last2;
u32 * copy_src_last3, * copy_dst_last3;
u16 new_l0, new_l1, new_l2, new_l3;
ip_csum_t sum0, sum1, sum2, sum3;
/* Prefetch next iteration. */
{
vlib_buffer_t * p4, * p5, * p6, * p7;
p4 = vlib_get_buffer (vm, from[4]);
p5 = vlib_get_buffer (vm, from[5]);
p6 = vlib_get_buffer (vm, from[6]);
p7 = vlib_get_buffer (vm, from[7]);
vlib_prefetch_buffer_header (p4, LOAD);
vlib_prefetch_buffer_header (p5, LOAD);
vlib_prefetch_buffer_header (p6, LOAD);
vlib_prefetch_buffer_header (p7, LOAD);
CLIB_PREFETCH (p4->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
CLIB_PREFETCH (p5->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
CLIB_PREFETCH (p6->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
CLIB_PREFETCH (p7->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
}
bi0 = from[0];
bi1 = from[1];
bi2 = from[2];
bi3 = from[3];
to_next[0] = bi0;
to_next[1] = bi1;
to_next[2] = bi2;
to_next[3] = bi3;
from += 4;
to_next += 4;
n_left_to_next -= 4;
n_left_from -= 4;
b0 = vlib_get_buffer (vm, bi0);
b1 = vlib_get_buffer (vm, bi1);
b2 = vlib_get_buffer (vm, bi2);
b3 = vlib_get_buffer (vm, bi3);
flow_hash0 = vnet_l2_compute_flow_hash (b0);
flow_hash1 = vnet_l2_compute_flow_hash (b1);
flow_hash2 = vnet_l2_compute_flow_hash (b2);
flow_hash3 = vnet_l2_compute_flow_hash (b3);
/* Get next node index and adj index from tunnel next_dpo */
sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX];
sw_if_index2 = vnet_buffer(b2)->sw_if_index[VLIB_TX];
sw_if_index3 = vnet_buffer(b3)->sw_if_index[VLIB_TX];
hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1);
hi2 = vnet_get_sup_hw_interface (vnm, sw_if_index2);
hi3 = vnet_get_sup_hw_interface (vnm, sw_if_index3);
t0 = >m->tunnels[hi0->dev_instance];
t1 = >m->tunnels[hi1->dev_instance];
t2 = >m->tunnels[hi2->dev_instance];
t3 = >m->tunnels[hi3->dev_instance];
/* Note: change to always set next0 if it may be set to drop */
next0 = t0->next_dpo.dpoi_next_node;
vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;
next1 = t1->next_dpo.dpoi_next_node;
vnet_buffer(b1)->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index;
next2 = t2->next_dpo.dpoi_next_node;
vnet_buffer(b2)->ip.adj_index[VLIB_TX] = t2->next_dpo.dpoi_index;
next3 = t3->next_dpo.dpoi_next_node;
vnet_buffer(b3)->ip.adj_index[VLIB_TX] = t3->next_dpo.dpoi_index;
/* Apply the rewrite string. $$$$ vnet_rewrite? */
vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));
vlib_buffer_advance (b1, -(word)_vec_len(t1->rewrite));
vlib_buffer_advance (b2, -(word)_vec_len(t2->rewrite));
vlib_buffer_advance (b3, -(word)_vec_len(t3->rewrite));
if (is_ip4)
{
ip4_0 = vlib_buffer_get_current(b0);
ip4_1 = vlib_buffer_get_current(b1);
ip4_2 = vlib_buffer_get_current(b2);
ip4_3 = vlib_buffer_get_current(b3);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip4_0;
copy_src0 = (u64 *) t0->rewrite;
copy_dst1 = (u64 *) ip4_1;
copy_src1 = (u64 *) t1->rewrite;
copy_dst2 = (u64 *) ip4_2;
copy_src2 = (u64 *) t2->rewrite;
copy_dst3 = (u64 *) ip4_3;
copy_src3 = (u64 *) t3->rewrite;
/* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header4_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
foreach_fixed_header4_offset;
#undef _
#define _(offs) copy_dst2[offs] = copy_src2[offs];
foreach_fixed_header4_offset;
#undef _
#define _(offs) copy_dst3[offs] = copy_src3[offs];
foreach_fixed_header4_offset;
#undef _
/* Last 4 octets. Hopefully gcc will be our friend */
copy_dst_last0 = (u32 *)(©_dst0[4]);
copy_src_last0 = (u32 *)(©_src0[4]);
copy_dst_last0[0] = copy_src_last0[0];
copy_dst_last1 = (u32 *)(©_dst1[4]);
copy_src_last1 = (u32 *)(©_src1[4]);
copy_dst_last1[0] = copy_src_last1[0];
copy_dst_last2 = (u32 *)(©_dst2[4]);
copy_src_last2 = (u32 *)(©_src2[4]);
copy_dst_last2[0] = copy_src_last2[0];
copy_dst_last3 = (u32 *)(©_dst3[4]);
copy_src_last3 = (u32 *)(©_src3[4]);
copy_dst_last3[0] = copy_src_last3[0];
/* Fix the IP4 checksum and length */
sum0 = ip4_0->checksum;
new_l0 = /* old_l0 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
length /* changed member */);
ip4_0->checksum = ip_csum_fold (sum0);
ip4_0->length = new_l0;
sum1 = ip4_1->checksum;
new_l1 = /* old_l1 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
length /* changed member */);
ip4_1->checksum = ip_csum_fold (sum1);
ip4_1->length = new_l1;
sum2 = ip4_2->checksum;
new_l2 = /* old_l0 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b2));
sum2 = ip_csum_update (sum2, old_l2, new_l2, ip4_header_t,
length /* changed member */);
ip4_2->checksum = ip_csum_fold (sum2);
ip4_2->length = new_l2;
sum3 = ip4_3->checksum;
new_l3 = /* old_l1 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b3));
sum3 = ip_csum_update (sum3, old_l3, new_l3, ip4_header_t,
length /* changed member */);
ip4_3->checksum = ip_csum_fold (sum3);
ip4_3->length = new_l3;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *)(ip4_0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip4_0));
udp0->length = new_l0;
udp0->src_port = flow_hash0;
udp1 = (udp_header_t *)(ip4_1+1);
new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b1)
- sizeof (*ip4_1));
udp1->length = new_l1;
udp1->src_port = flow_hash1;
udp2 = (udp_header_t *)(ip4_2+1);
new_l2 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b2)
- sizeof (*ip4_2));
udp2->length = new_l2;
udp2->src_port = flow_hash2;
udp3 = (udp_header_t *)(ip4_3+1);
new_l3 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b3)
- sizeof (*ip4_3));
udp3->length = new_l3;
udp3->src_port = flow_hash3;
/* Fix GTPU length */
gtpu0 = (gtpu_header_t *)(udp0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip4_0) - sizeof(*udp0)
- GTPU_V1_HDR_LEN);
gtpu0->length = new_l0;
gtpu1 = (gtpu_header_t *)(udp1+1);
new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b1)
- sizeof (*ip4_1) - sizeof(*udp1)
- GTPU_V1_HDR_LEN);
gtpu1->length = new_l1;
gtpu2 = (gtpu_header_t *)(udp2+1);
new_l2 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b2)
- sizeof (*ip4_2) - sizeof(*udp2)
- GTPU_V1_HDR_LEN);
gtpu2->length = new_l2;
gtpu3 = (gtpu_header_t *)(udp3+1);
new_l3 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b3)
- sizeof (*ip4_3) - sizeof(*udp3)
- GTPU_V1_HDR_LEN);
gtpu3->length = new_l3;
}
else /* ipv6 */
{
int bogus = 0;
ip6_0 = vlib_buffer_get_current(b0);
ip6_1 = vlib_buffer_get_current(b1);
ip6_2 = vlib_buffer_get_current(b2);
ip6_3 = vlib_buffer_get_current(b3);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip6_0;
copy_src0 = (u64 *) t0->rewrite;
copy_dst1 = (u64 *) ip6_1;
copy_src1 = (u64 *) t1->rewrite;
copy_dst2 = (u64 *) ip6_2;
copy_src2 = (u64 *) t2->rewrite;
copy_dst3 = (u64 *) ip6_3;
copy_src3 = (u64 *) t3->rewrite;
/* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header6_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
foreach_fixed_header6_offset;
#undef _
#define _(offs) copy_dst2[offs] = copy_src2[offs];
foreach_fixed_header6_offset;
#undef _
#define _(offs) copy_dst3[offs] = copy_src3[offs];
foreach_fixed_header6_offset;
#undef _
/* Fix IP6 payload length */
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
- sizeof(*ip6_0));
ip6_0->payload_length = new_l0;
new_l1 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)
- sizeof(*ip6_1));
ip6_1->payload_length = new_l1;
new_l2 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b2)
- sizeof(*ip6_2));
ip6_2->payload_length = new_l2;
new_l3 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b3)
- sizeof(*ip6_3));
ip6_3->payload_length = new_l3;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *)(ip6_0+1);
udp0->length = new_l0;
udp0->src_port = flow_hash0;
udp1 = (udp_header_t *)(ip6_1+1);
udp1->length = new_l1;
udp1->src_port = flow_hash1;
udp2 = (udp_header_t *)(ip6_2+1);
udp2->length = new_l2;
udp2->src_port = flow_hash2;
udp3 = (udp_header_t *)(ip6_3+1);
udp3->length = new_l3;
udp3->src_port = flow_hash3;
/* Fix GTPU length */
gtpu0 = (gtpu_header_t *)(udp0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip6_0) - sizeof(*udp0)
- GTPU_V1_HDR_LEN);
gtpu0->length = new_l0;
gtpu1 = (gtpu_header_t *)(udp1+1);
new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b1)
- sizeof (*ip6_1) - sizeof(*udp1)
- GTPU_V1_HDR_LEN);
gtpu1->length = new_l1;
gtpu2 = (gtpu_header_t *)(udp2+1);
new_l2 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b2)
- sizeof (*ip6_2) - sizeof(*udp2)
- GTPU_V1_HDR_LEN);
gtpu2->length = new_l2;
gtpu3 = (gtpu_header_t *)(udp3+1);
new_l3 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b3)
- sizeof (*ip6_3) - sizeof(*udp3)
- GTPU_V1_HDR_LEN);
gtpu3->length = new_l3;
/* IPv6 UDP checksum is mandatory */
udp0->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b0,
ip6_0, &bogus);
if (udp0->checksum == 0)
udp0->checksum = 0xffff;
udp1->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b1,
ip6_1, &bogus);
if (udp1->checksum == 0)
udp1->checksum = 0xffff;
udp2->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b2,
ip6_2, &bogus);
if (udp2->checksum == 0)
udp2->checksum = 0xffff;
udp3->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b3,
ip6_3, &bogus);
if (udp3->checksum == 0)
udp3->checksum = 0xffff;
}
pkts_encapsulated += 4;
len0 = vlib_buffer_length_in_chain (vm, b0);
len1 = vlib_buffer_length_in_chain (vm, b1);
len2 = vlib_buffer_length_in_chain (vm, b2);
len3 = vlib_buffer_length_in_chain (vm, b3);
stats_n_packets += 4;
stats_n_bytes += len0 + len1 + len2 + len3;
/* Batch stats increment on the same gtpu tunnel so counter is not
incremented per packet. Note stats are still incremented for deleted
and admin-down tunnel where packets are dropped. It is not worthwhile
to check for this rare case and affect normal path performance. */
if (PREDICT_FALSE ((sw_if_index0 != stats_sw_if_index) ||
(sw_if_index1 != stats_sw_if_index) ||
(sw_if_index2 != stats_sw_if_index) ||
(sw_if_index3 != stats_sw_if_index) ))
{
stats_n_packets -= 4;
stats_n_bytes -= len0 + len1 + len2 + len3;
if ( (sw_if_index0 == sw_if_index1 ) &&
(sw_if_index1 == sw_if_index2 ) &&
(sw_if_index2 == sw_if_index3 ) )
{
if (stats_n_packets)
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, stats_sw_if_index,
stats_n_packets, stats_n_bytes);
stats_sw_if_index = sw_if_index0;
stats_n_packets = 4;
stats_n_bytes = len0 + len1 + len2 + len3;
}
else
{
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index0, 1, len0);
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index1, 1, len1);
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index2, 1, len2);
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index3, 1, len3);
}
}
if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
{
gtpu_encap_trace_t *tr =
vlib_add_trace (vm, node, b0, sizeof (*tr));
tr->tunnel_index = t0 - gtm->tunnels;
tr->teid = t0->teid;
}
if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED))
{
gtpu_encap_trace_t *tr =
vlib_add_trace (vm, node, b1, sizeof (*tr));
tr->tunnel_index = t1 - gtm->tunnels;
tr->teid = t1->teid;
}
vlib_validate_buffer_enqueue_x4 (vm, node, next_index,
to_next, n_left_to_next,
bi0, bi1, bi2, bi3,
next0, next1, next2, next3);
}
while (n_left_from > 0 && n_left_to_next > 0)
{
u32 bi0;
vlib_buffer_t * b0;
u32 flow_hash0;
u32 len0;
ip4_header_t * ip4_0;
ip6_header_t * ip6_0;
udp_header_t * udp0;
gtpu_header_t * gtpu0;
u64 * copy_src0, * copy_dst0;
u32 * copy_src_last0, * copy_dst_last0;
u16 new_l0;
ip_csum_t sum0;
bi0 = from[0];
to_next[0] = bi0;
from += 1;
to_next += 1;
n_left_from -= 1;
n_left_to_next -= 1;
b0 = vlib_get_buffer (vm, bi0);
flow_hash0 = vnet_l2_compute_flow_hash(b0);
/* Get next node index and adj index from tunnel next_dpo */
sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
t0 = >m->tunnels[hi0->dev_instance];
/* Note: change to always set next0 if it may be set to drop */
next0 = t0->next_dpo.dpoi_next_node;
vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;
/* Apply the rewrite string. $$$$ vnet_rewrite? */
vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));
if (is_ip4)
{
ip4_0 = vlib_buffer_get_current(b0);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip4_0;
copy_src0 = (u64 *) t0->rewrite;
/* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header4_offset;
#undef _
/* Last 4 octets. Hopefully gcc will be our friend */
copy_dst_last0 = (u32 *)(©_dst0[4]);
copy_src_last0 = (u32 *)(©_src0[4]);
copy_dst_last0[0] = copy_src_last0[0];
/* Fix the IP4 checksum and length */
sum0 = ip4_0->checksum;
new_l0 = /* old_l0 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
length /* changed member */);
ip4_0->checksum = ip_csum_fold (sum0);
ip4_0->length = new_l0;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *)(ip4_0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip4_0));
udp0->length = new_l0;
udp0->src_port = flow_hash0;
/* Fix GTPU length */
gtpu0 = (gtpu_header_t *)(udp0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip4_0) - sizeof(*udp0)
- GTPU_V1_HDR_LEN);
gtpu0->length = new_l0;
}
else /* ip6 path */
{
int bogus = 0;
ip6_0 = vlib_buffer_get_current(b0);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip6_0;
copy_src0 = (u64 *) t0->rewrite;
/* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header6_offset;
#undef _
/* Fix IP6 payload length */
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
- sizeof(*ip6_0));
ip6_0->payload_length = new_l0;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *)(ip6_0+1);
udp0->length = new_l0;
udp0->src_port = flow_hash0;
/* Fix GTPU length */
gtpu0 = (gtpu_header_t *)(udp0+1);
new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
- sizeof (*ip4_0) - sizeof(*udp0)
- GTPU_V1_HDR_LEN);
gtpu0->length = new_l0;
/* IPv6 UDP checksum is mandatory */
udp0->checksum = ip6_tcp_udp_icmp_compute_checksum(vm, b0,
ip6_0, &bogus);
if (udp0->checksum == 0)
udp0->checksum = 0xffff;
}
pkts_encapsulated ++;
len0 = vlib_buffer_length_in_chain (vm, b0);
stats_n_packets += 1;
stats_n_bytes += len0;
/* Batch stats increment on the same gtpu tunnel so counter is not
incremented per packet. Note stats are still incremented for deleted
and admin-down tunnel where packets are dropped. It is not worthwhile
to check for this rare case and affect normal path performance. */
if (PREDICT_FALSE (sw_if_index0 != stats_sw_if_index))
{
stats_n_packets -= 1;
stats_n_bytes -= len0;
if (stats_n_packets)
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, stats_sw_if_index,
stats_n_packets, stats_n_bytes);
stats_n_packets = 1;
stats_n_bytes = len0;
stats_sw_if_index = sw_if_index0;
}
if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED))
{
gtpu_encap_trace_t *tr =
vlib_add_trace (vm, node, b0, sizeof (*tr));
tr->tunnel_index = t0 - gtm->tunnels;
tr->teid = t0->teid;
}
vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
to_next, n_left_to_next,
bi0, next0);
}
vlib_put_next_frame (vm, node, next_index, n_left_to_next);
}
/* Do we still need this now that tunnel tx stats is kept? */
vlib_node_increment_counter (vm, node->node_index,
GTPU_ENCAP_ERROR_ENCAPSULATED,
pkts_encapsulated);
/* Increment any remaining batch stats */
if (stats_n_packets)
{
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
node->runtime_data[0] = stats_sw_if_index;
}
return from_frame->n_vectors;
}
always_inline uword
geneve_encap_inline (vlib_main_t * vm,
vlib_node_runtime_t * node,
vlib_frame_t * from_frame, u32 is_ip4)
{
u32 n_left_from, next_index, *from, *to_next;
geneve_main_t *vxm = &geneve_main;
vnet_main_t *vnm = vxm->vnet_main;
vnet_interface_main_t *im = &vnm->interface_main;
u32 pkts_encapsulated = 0;
u16 old_l0 = 0, old_l1 = 0;
u32 thread_index = vm->thread_index;
u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
u32 sw_if_index0 = ~0, sw_if_index1 = ~0;
u32 next0 = 0, next1 = 0;
vnet_hw_interface_t *hi0, *hi1;
geneve_tunnel_t *t0 = NULL, *t1 = NULL;
from = vlib_frame_vector_args (from_frame);
n_left_from = from_frame->n_vectors;
next_index = node->cached_next_index;
stats_sw_if_index = node->runtime_data[0];
stats_n_packets = stats_n_bytes = 0;
while (n_left_from > 0)
{
u32 n_left_to_next;
vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
while (n_left_from >= 4 && n_left_to_next >= 2)
{
u32 bi0, bi1;
vlib_buffer_t *b0, *b1;
u32 flow_hash0, flow_hash1;
u32 len0, len1;
ip4_header_t *ip4_0, *ip4_1;
ip6_header_t *ip6_0, *ip6_1;
udp_header_t *udp0, *udp1;
u64 *copy_src0, *copy_dst0;
u64 *copy_src1, *copy_dst1;
u32 *copy_src_last0, *copy_dst_last0;
u32 *copy_src_last1, *copy_dst_last1;
u16 new_l0, new_l1;
ip_csum_t sum0, sum1;
/* Prefetch next iteration. */
{
vlib_buffer_t *p2, *p3;
p2 = vlib_get_buffer (vm, from[2]);
p3 = vlib_get_buffer (vm, from[3]);
vlib_prefetch_buffer_header (p2, LOAD);
vlib_prefetch_buffer_header (p3, LOAD);
CLIB_PREFETCH (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD);
CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD);
}
bi0 = from[0];
bi1 = from[1];
to_next[0] = bi0;
to_next[1] = bi1;
from += 2;
to_next += 2;
n_left_to_next -= 2;
n_left_from -= 2;
b0 = vlib_get_buffer (vm, bi0);
b1 = vlib_get_buffer (vm, bi1);
flow_hash0 = vnet_l2_compute_flow_hash (b0);
flow_hash1 = vnet_l2_compute_flow_hash (b1);
/* Get next node index and adj index from tunnel next_dpo */
if (sw_if_index0 != vnet_buffer (b0)->sw_if_index[VLIB_TX])
{
sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX];
hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
t0 = &vxm->tunnels[hi0->dev_instance];
/* Note: change to always set next0 if it may be set to drop */
next0 = t0->next_dpo.dpoi_next_node;
}
ASSERT (t0 != NULL);
vnet_buffer (b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;
/* Get next node index and adj index from tunnel next_dpo */
if (sw_if_index1 != vnet_buffer (b1)->sw_if_index[VLIB_TX])
{
sw_if_index1 = vnet_buffer (b1)->sw_if_index[VLIB_TX];
hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1);
t1 = &vxm->tunnels[hi1->dev_instance];
/* Note: change to always set next1 if it may be set to drop */
next1 = t1->next_dpo.dpoi_next_node;
}
ASSERT (t1 != NULL);
vnet_buffer (b1)->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index;
/* Apply the rewrite string. $$$$ vnet_rewrite? */
vlib_buffer_advance (b0, -(word) _vec_len (t0->rewrite));
vlib_buffer_advance (b1, -(word) _vec_len (t1->rewrite));
if (is_ip4)
{
u8 ip4_geneve_base_header_len =
sizeof (ip4_header_t) + sizeof (udp_header_t) +
GENEVE_BASE_HEADER_LENGTH;
u8 ip4_geneve_header_total_len0 = ip4_geneve_base_header_len;
u8 ip4_geneve_header_total_len1 = ip4_geneve_base_header_len;
#if SUPPORT_OPTIONS_HEADER==1
ip4_geneve_header_total_len0 += t0->options_len;
ip4_geneve_header_total_len1 += t1->options_len;
#endif
ASSERT (vec_len (t0->rewrite) == ip4_geneve_header_total_len0);
ASSERT (vec_len (t1->rewrite) == ip4_geneve_header_total_len1);
ip4_0 = vlib_buffer_get_current (b0);
ip4_1 = vlib_buffer_get_current (b1);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip4_0;
copy_src0 = (u64 *) t0->rewrite;
copy_dst1 = (u64 *) ip4_1;
copy_src1 = (u64 *) t1->rewrite;
/* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header4_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
foreach_fixed_header4_offset;
#undef _
/* Last 4 octets. Hopefully gcc will be our friend */
copy_dst_last0 = (u32 *) (©_dst0[4]);
copy_src_last0 = (u32 *) (©_src0[4]);
copy_dst_last0[0] = copy_src_last0[0];
copy_dst_last1 = (u32 *) (©_dst1[4]);
copy_src_last1 = (u32 *) (©_src1[4]);
copy_dst_last1[0] = copy_src_last1[0];
/* Fix the IP4 checksum and length */
sum0 = ip4_0->checksum;
new_l0 = /* old_l0 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
length /* changed member */ );
ip4_0->checksum = ip_csum_fold (sum0);
ip4_0->length = new_l0;
sum1 = ip4_1->checksum;
new_l1 = /* old_l1 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
length /* changed member */ );
ip4_1->checksum = ip_csum_fold (sum1);
ip4_1->length = new_l1;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *) (ip4_0 + 1);
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
sizeof (*ip4_0));
udp0->length = new_l0;
udp0->src_port = flow_hash0;
udp1 = (udp_header_t *) (ip4_1 + 1);
new_l1 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) -
sizeof (*ip4_1));
udp1->length = new_l1;
udp1->src_port = flow_hash1;
}
else /* ipv6 */
{
int bogus = 0;
u8 ip6_geneve_base_header_len =
sizeof (ip6_header_t) + sizeof (udp_header_t) +
GENEVE_BASE_HEADER_LENGTH;
u8 ip6_geneve_header_total_len0 = ip6_geneve_base_header_len;
u8 ip6_geneve_header_total_len1 = ip6_geneve_base_header_len;
#if SUPPORT_OPTIONS_HEADER==1
ip6_geneve_header_total_len0 += t0->options_len;
ip6_geneve_header_total_len1 += t1->options_len;
#endif
ASSERT (vec_len (t0->rewrite) == ip6_geneve_header_total_len0);
ASSERT (vec_len (t1->rewrite) == ip6_geneve_header_total_len1);
ip6_0 = vlib_buffer_get_current (b0);
ip6_1 = vlib_buffer_get_current (b1);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip6_0;
copy_src0 = (u64 *) t0->rewrite;
copy_dst1 = (u64 *) ip6_1;
copy_src1 = (u64 *) t1->rewrite;
/* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header6_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
foreach_fixed_header6_offset;
#undef _
/* Fix IP6 payload length */
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
- sizeof (*ip6_0));
ip6_0->payload_length = new_l0;
new_l1 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)
- sizeof (*ip6_1));
ip6_1->payload_length = new_l1;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *) (ip6_0 + 1);
udp0->length = new_l0;
udp0->src_port = flow_hash0;
udp1 = (udp_header_t *) (ip6_1 + 1);
udp1->length = new_l1;
udp1->src_port = flow_hash1;
/* IPv6 UDP checksum is mandatory */
udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0,
ip6_0,
&bogus);
ASSERT (bogus == 0);
if (udp0->checksum == 0)
udp0->checksum = 0xffff;
udp1->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b1,
ip6_1,
&bogus);
ASSERT (bogus == 0);
if (udp1->checksum == 0)
udp1->checksum = 0xffff;
}
pkts_encapsulated += 2;
len0 = vlib_buffer_length_in_chain (vm, b0);
len1 = vlib_buffer_length_in_chain (vm, b1);
stats_n_packets += 2;
stats_n_bytes += len0 + len1;
/* Batch stats increment on the same geneve tunnel so counter is not
incremented per packet. Note stats are still incremented for deleted
and admin-down tunnel where packets are dropped. It is not worthwhile
to check for this rare case and affect normal path performance. */
if (PREDICT_FALSE ((sw_if_index0 != stats_sw_if_index) ||
(sw_if_index1 != stats_sw_if_index)))
{
stats_n_packets -= 2;
stats_n_bytes -= len0 + len1;
if (sw_if_index0 == sw_if_index1)
{
if (stats_n_packets)
vlib_increment_combined_counter
(im->combined_sw_if_counters +
VNET_INTERFACE_COUNTER_TX, thread_index,
stats_sw_if_index, stats_n_packets, stats_n_bytes);
stats_sw_if_index = sw_if_index0;
stats_n_packets = 2;
stats_n_bytes = len0 + len1;
}
else
{
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index0, 1, len0);
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, sw_if_index1, 1, len1);
}
}
if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
{
geneve_encap_trace_t *tr =
vlib_add_trace (vm, node, b0, sizeof (*tr));
tr->tunnel_index = t0 - vxm->tunnels;
tr->vni = t0->vni;
}
if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED))
{
geneve_encap_trace_t *tr =
vlib_add_trace (vm, node, b1, sizeof (*tr));
tr->tunnel_index = t1 - vxm->tunnels;
tr->vni = t1->vni;
}
vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
to_next, n_left_to_next,
bi0, bi1, next0, next1);
}
while (n_left_from > 0 && n_left_to_next > 0)
{
u32 bi0;
vlib_buffer_t *b0;
u32 flow_hash0;
u32 len0;
ip4_header_t *ip4_0;
ip6_header_t *ip6_0;
udp_header_t *udp0;
u64 *copy_src0, *copy_dst0;
u32 *copy_src_last0, *copy_dst_last0;
u16 new_l0;
ip_csum_t sum0;
bi0 = from[0];
to_next[0] = bi0;
from += 1;
to_next += 1;
n_left_from -= 1;
n_left_to_next -= 1;
b0 = vlib_get_buffer (vm, bi0);
flow_hash0 = vnet_l2_compute_flow_hash (b0);
/* Get next node index and adj index from tunnel next_dpo */
if (sw_if_index0 != vnet_buffer (b0)->sw_if_index[VLIB_TX])
{
sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX];
hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
t0 = &vxm->tunnels[hi0->dev_instance];
/* Note: change to always set next0 if it may be set to drop */
next0 = t0->next_dpo.dpoi_next_node;
}
vnet_buffer (b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;
/* Apply the rewrite string. $$$$ vnet_rewrite? */
vlib_buffer_advance (b0, -(word) _vec_len (t0->rewrite));
if (is_ip4)
{
u8 ip4_geneve_base_header_len =
sizeof (ip4_header_t) + sizeof (udp_header_t) +
GENEVE_BASE_HEADER_LENGTH;
u8 ip4_geneve_header_total_len0 = ip4_geneve_base_header_len;
#if SUPPORT_OPTIONS_HEADER==1
ip4_geneve_header_total_len0 += t0->options_len;
#endif
ASSERT (vec_len (t0->rewrite) == ip4_geneve_header_total_len0);
ip4_0 = vlib_buffer_get_current (b0);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip4_0;
copy_src0 = (u64 *) t0->rewrite;
/* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header4_offset;
#undef _
/* Last 4 octets. Hopefully gcc will be our friend */
copy_dst_last0 = (u32 *) (©_dst0[4]);
copy_src_last0 = (u32 *) (©_src0[4]);
copy_dst_last0[0] = copy_src_last0[0];
/* Fix the IP4 checksum and length */
sum0 = ip4_0->checksum;
new_l0 = /* old_l0 always 0, see the rewrite setup */
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
length /* changed member */ );
ip4_0->checksum = ip_csum_fold (sum0);
ip4_0->length = new_l0;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *) (ip4_0 + 1);
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) -
sizeof (*ip4_0));
udp0->length = new_l0;
udp0->src_port = flow_hash0;
}
else /* ip6 path */
{
int bogus = 0;
u8 ip6_geneve_base_header_len =
sizeof (ip6_header_t) + sizeof (udp_header_t) +
GENEVE_BASE_HEADER_LENGTH;
u8 ip6_geneve_header_total_len0 = ip6_geneve_base_header_len;
#if SUPPORT_OPTIONS_HEADER==1
ip6_geneve_header_total_len0 += t0->options_len;
#endif
ASSERT (vec_len (t0->rewrite) == ip6_geneve_header_total_len0);
ip6_0 = vlib_buffer_get_current (b0);
/* Copy the fixed header */
copy_dst0 = (u64 *) ip6_0;
copy_src0 = (u64 *) t0->rewrite;
/* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
foreach_fixed_header6_offset;
#undef _
/* Fix IP6 payload length */
new_l0 =
clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
- sizeof (*ip6_0));
ip6_0->payload_length = new_l0;
/* Fix UDP length and set source port */
udp0 = (udp_header_t *) (ip6_0 + 1);
udp0->length = new_l0;
udp0->src_port = flow_hash0;
/* IPv6 UDP checksum is mandatory */
udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0,
ip6_0,
&bogus);
ASSERT (bogus == 0);
if (udp0->checksum == 0)
udp0->checksum = 0xffff;
}
pkts_encapsulated++;
len0 = vlib_buffer_length_in_chain (vm, b0);
stats_n_packets += 1;
stats_n_bytes += len0;
/* Batch stats increment on the same geneve tunnel so counter is not
incremented per packet. Note stats are still incremented for deleted
and admin-down tunnel where packets are dropped. It is not worthwhile
to check for this rare case and affect normal path performance. */
if (PREDICT_FALSE (sw_if_index0 != stats_sw_if_index))
{
stats_n_packets -= 1;
stats_n_bytes -= len0;
if (stats_n_packets)
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, stats_sw_if_index,
stats_n_packets, stats_n_bytes);
stats_n_packets = 1;
stats_n_bytes = len0;
stats_sw_if_index = sw_if_index0;
}
if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED))
{
geneve_encap_trace_t *tr =
vlib_add_trace (vm, node, b0, sizeof (*tr));
tr->tunnel_index = t0 - vxm->tunnels;
tr->vni = t0->vni;
}
vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
to_next, n_left_to_next,
bi0, next0);
}
vlib_put_next_frame (vm, node, next_index, n_left_to_next);
}
/* Do we still need this now that tunnel tx stats is kept? */
vlib_node_increment_counter (vm, node->node_index,
GENEVE_ENCAP_ERROR_ENCAPSULATED,
pkts_encapsulated);
/* Increment any remaining batch stats */
if (stats_n_packets)
{
vlib_increment_combined_counter
(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
node->runtime_data[0] = stats_sw_if_index;
}
return from_frame->n_vectors;
}
/* #define UDP_PACKET_DEBUG 1 */
// Temporary debugs which will be suppressed later
#ifdef UDP_PACKET_DEBUG
if (PREDICT_FALSE(udp_inside_packet_dump_enable)) {
printf("\nIn2Out UDP packet before translation");
print_udp_pkt(ip);
}
#endif
//set ip header
ip->src_addr =
clib_host_to_net_u32(db->out2in_key.k.ipv4);
ip->checksum =
clib_host_to_net_u16(new_l3_c);
u16 frag_offset =
clib_net_to_host_u16(ip->frag_flags_offset);
if(PREDICT_FALSE(frag_offset & IP_FRAG_OFFSET_MASK)) {
return; /* No need to update UDP fields */
}
//set udp header
udp->src_port =
clib_host_to_net_u16(db->out2in_key.k.port);
/*
* No easy way to avoid this if check except by using
* complex logic - may not be worth it.
*/