summaryrefslogtreecommitdiffstats
path: root/VNFs/DPPD-PROX/packet_utils.c
diff options
context:
space:
mode:
authorXavier Simonart <xavier.simonart@intel.com>2020-05-11 21:00:33 +0200
committerXavier Simonart <xavier.simonart@intel.com>2020-05-29 23:45:09 +0200
commitc871c361f9d69a93429ae385e7dbf21a46aa6857 (patch)
tree957ec8a53d2cd0a9b5676e163c22e8a3ae8f6f3e /VNFs/DPPD-PROX/packet_utils.c
parent7c4601f23c526b14a67674782b303663dfaa95af (diff)
Added initial support for NDP (IPv6)
Following messages are now handled by PROX - router_solicitation - neighbour_solicitation - router_advertisement - neighbour_advertisement The following parameters are supported (through the PROX config file) - sub mode=ndp This will enable handling of router and neighbour solicitation and advertisement. - local ipv6=xxxx:xxxx:xxxxx:xxxx:xxxx:xxxx:xxxx:xxxx This will configure the local IPv6 address of the port. This parameter is optional. If not specified, the local IPv6 will be calculated from the EUI. - global ipv6=xxxx:xxxx:xxxxx:xxxx:xxxx:xxxx:xxxx:xxxx This will configure the global IPv6 address of the port. This parameter is optional. If not specified, the global IPv6 will be calculated from the EUI and the router prefix received from the router. - ipv6 router=yes This will cause the core to behave as an IPv6 router i.e. it will generate Router Advertisement messages This is only useful in back to back cases, when no real IPv6 router is present in the setup. - router prefix=xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx The router prefix usedin the router advertisement The prefix will be used by the node to build an IPv6 global address in cases none were configured. "Unsollicited NA" parameter has been added within the core/task section. If set to yes (Unsollicited NA=yes), then an unsollicited neighbour Advertisement is sent at startup A same core/task cannot support both l3 and ndp mode. Those messages will be generated or handled when submode is set to "ndp": - neighbour sollicitation - neighbour advertisement - router sollicitation - router advertisement An example configuration is provided: config/ipv6.cfg in which port 0 / core 1 plays the role of the generator and port 1 / core 2 plays the role of the swap. Change-Id: Id0ab32d384448b4cf767fb4a1c486fc023f4f395 Signed-off-by: Xavier Simonart <xavier.simonart@intel.com>
Diffstat (limited to 'VNFs/DPPD-PROX/packet_utils.c')
-rw-r--r--VNFs/DPPD-PROX/packet_utils.c461
1 files changed, 379 insertions, 82 deletions
diff --git a/VNFs/DPPD-PROX/packet_utils.c b/VNFs/DPPD-PROX/packet_utils.c
index 08178d82..466dd481 100644
--- a/VNFs/DPPD-PROX/packet_utils.c
+++ b/VNFs/DPPD-PROX/packet_utils.c
@@ -31,6 +31,9 @@
#include "prox_lua.h"
#include "hash_entry_types.h"
#include "prox_compat.h"
+#include "prox_cfg.h"
+#include "defines.h"
+#include "prox_ipv6.h"
#include "tx_pkt.h"
static inline int find_ip(struct ether_hdr_arp *pkt, uint16_t len, uint32_t *ip_dst)
@@ -76,15 +79,91 @@ static inline int find_ip(struct ether_hdr_arp *pkt, uint16_t len, uint32_t *ip_
return -1;
}
+static inline struct ipv6_addr *find_ip6(prox_rte_ether_hdr *pkt, uint16_t len, struct ipv6_addr *ip_dst)
+{
+ prox_rte_vlan_hdr *vlan_hdr;
+ prox_rte_ipv6_hdr *ip;
+ uint16_t ether_type = pkt->ether_type;
+ uint16_t l2_len = sizeof(prox_rte_ether_hdr);
+
+ // Unstack VLAN tags
+ while (((ether_type == ETYPE_8021ad) || (ether_type == ETYPE_VLAN)) && (l2_len + sizeof(prox_rte_vlan_hdr) < len)) {
+ vlan_hdr = (prox_rte_vlan_hdr *)((uint8_t *)pkt + l2_len);
+ l2_len +=4;
+ ether_type = vlan_hdr->eth_proto;
+ }
+
+ switch (ether_type) {
+ case ETYPE_MPLSU:
+ case ETYPE_MPLSM:
+ // In case of MPLS, next hop MAC is based on MPLS, not destination IP
+ l2_len = 0;
+ break;
+ case ETYPE_IPv4:
+ case ETYPE_EoGRE:
+ case ETYPE_ARP:
+ l2_len = 0;
+ break;
+ case ETYPE_IPv6:
+ break;
+ default:
+ l2_len = 0;
+ plog_warn("Unsupported packet type %x - CRC might be wrong\n", ether_type);
+ break;
+ }
+
+ if (l2_len && (l2_len + sizeof(prox_rte_ipv6_hdr) <= len)) {
+ prox_rte_ipv6_hdr *ip = (prox_rte_ipv6_hdr *)((uint8_t *)pkt + l2_len);
+ // TODO: implement LPM => replace ip_dst by next hop IP DST
+ memcpy(ip_dst, &ip->dst_addr, sizeof(struct ipv6_addr));
+ return (struct ipv6_addr *)&ip->src_addr;
+ }
+ return NULL;
+}
+
+static void send_unsollicited_neighbour_advertisement(struct task_base *tbase, struct task_args *targ)
+{
+ int ret;
+ uint8_t out = 0, port_id = tbase->l3.reachable_port_id;
+ struct rte_mbuf *mbuf;
+
+ ret = rte_mempool_get(tbase->l3.arp_nd_pool, (void **)&mbuf);
+ if (likely(ret == 0)) {
+ mbuf->port = port_id;
+ build_neighbour_advertisement(tbase->l3.tmaster, mbuf, &prox_port_cfg[port_id].eth_addr, &targ->local_ipv6, PROX_UNSOLLICITED);
+ tbase->aux->tx_ctrlplane_pkt(tbase, &mbuf, 1, &out);
+ TASK_STATS_ADD_TX_NON_DP(&tbase->aux->stats, 1);
+ } else {
+ plog_err("Failed to get a mbuf from arp/ndp mempool\n");
+ }
+}
+
+static void send_router_sollicitation(struct task_base *tbase, struct task_args *targ)
+{
+ int ret;
+ uint8_t out = 0, port_id = tbase->l3.reachable_port_id;
+ struct rte_mbuf *mbuf;
+
+ ret = rte_mempool_get(tbase->l3.arp_nd_pool, (void **)&mbuf);
+ if (likely(ret == 0)) {
+ mbuf->port = port_id;
+ build_router_sollicitation(mbuf, &prox_port_cfg[port_id].eth_addr, &targ->local_ipv6);
+ tbase->aux->tx_ctrlplane_pkt(tbase, &mbuf, 1, &out);
+ TASK_STATS_ADD_TX_NON_DP(&tbase->aux->stats, 1);
+ } else {
+ plog_err("Failed to get a mbuf from arp/ndp mempool\n");
+ }
+}
+
/* This implementation could be improved: instead of checking each time we send a packet whether we need also
to send an ARP, we should only check whether the MAC is valid.
- We should check arp_update_time in the master process. This would also require the generating task to clear its arp ring
+ We should check arp_ndp_retransmit_timeout in the master process. This would also require the generating task to clear its arp ring
to avoid sending many ARP while starting after a long stop.
- We could also check for arp_timeout in the master so that dataplane has only to check whether MAC is available
- but this would require either thread safety, or the exchange of information between master and generating core.
-*/
+ We could also check for reachable_timeout in the master so that dataplane has only to check whether MAC is available
+ but this would require either thread safety, or the the exchange of information between master and generating core.
+ */
-static inline int add_key_and_send_arp(struct rte_hash *ip_hash, uint32_t *ip_dst, struct arp_table *entries, uint64_t tsc, uint64_t hz, uint32_t arp_update_time, prox_next_hop_index_type nh, uint64_t **time)
+static inline int add_key_and_send_arp(struct rte_hash *ip_hash, uint32_t *ip_dst, struct arp_table *entries, uint64_t tsc, uint64_t hz, uint32_t arp_ndp_retransmit_timeout, prox_next_hop_index_type nh, uint64_t **time)
{
int ret = rte_hash_add_key(ip_hash, (const void *)ip_dst);
if (unlikely(ret < 0)) {
@@ -94,26 +173,26 @@ static inline int add_key_and_send_arp(struct rte_hash *ip_hash, uint32_t *ip_ds
} else {
entries[ret].ip = *ip_dst;
entries[ret].nh = nh;
- *time = &entries[ret].arp_update_time;
+ *time = &entries[ret].arp_ndp_retransmit_timeout;
}
- return SEND_ARP;
+ return SEND_ARP_ND;
}
-static inline int update_mac_and_send_mbuf(struct arp_table *entry, prox_rte_ether_addr *mac, uint64_t tsc, uint64_t hz, uint32_t arp_update_time, uint64_t **time)
+static inline int update_mac_and_send_mbuf(struct arp_table *entry, prox_rte_ether_addr *mac, uint64_t tsc, uint64_t hz, uint32_t arp_ndp_retransmit_timeout, uint64_t **time)
{
- if (likely((tsc < entry->arp_update_time) && (tsc < entry->arp_timeout))) {
+ if (likely((tsc < entry->arp_ndp_retransmit_timeout) && (tsc < entry->reachable_timeout))) {
memcpy(mac, &entry->mac, sizeof(prox_rte_ether_addr));
return SEND_MBUF;
- } else if (tsc > entry->arp_update_time) {
+ } else if (tsc > entry->arp_ndp_retransmit_timeout) {
// long time since we have sent an arp, send arp
- *time = &entry->arp_update_time;
- if (tsc < entry->arp_timeout){
+ *time = &entry->arp_ndp_retransmit_timeout;
+ if (tsc < entry->reachable_timeout){
// MAC is valid in the table => send also the mbuf
memcpy(mac, &entry->mac, sizeof(prox_rte_ether_addr));
- return SEND_MBUF_AND_ARP;
+ return SEND_MBUF_AND_ARP_ND;
} else {
// MAC still unknown, or timed out => only send ARP
- return SEND_ARP;
+ return SEND_ARP_ND;
}
}
// MAC is unknown and we already sent an ARP recently, drop mbuf and wait for ARP reply
@@ -154,7 +233,7 @@ int write_dst_mac(struct task_base *tbase, struct rte_mbuf *mbuf, uint32_t *ip_d
if (entry->ip) {
*ip_dst = entry->ip;
- return update_mac_and_send_mbuf(entry, mac, tsc, hz, l3->arp_update_time, time);
+ return update_mac_and_send_mbuf(entry, mac, tsc, hz, l3->arp_ndp_retransmit_timeout, time);
}
// no next ip: this is a local route
@@ -162,30 +241,32 @@ int write_dst_mac(struct task_base *tbase, struct rte_mbuf *mbuf, uint32_t *ip_d
int ret = rte_hash_lookup(l3->ip_hash, (const void *)ip_dst);
if (unlikely(ret < 0)) {
// IP not found, try to send an ARP
- return add_key_and_send_arp(l3->ip_hash, ip_dst, l3->arp_table, tsc, hz, l3->arp_update_time, MAX_HOP_INDEX, time);
+ return add_key_and_send_arp(l3->ip_hash, ip_dst, l3->arp_table, tsc, hz, l3->arp_ndp_retransmit_timeout, MAX_HOP_INDEX, time);
} else {
- return update_mac_and_send_mbuf(&l3->arp_table[ret], mac, tsc, hz, l3->arp_update_time, time);
+ return update_mac_and_send_mbuf(&l3->arp_table[ret], mac, tsc, hz, l3->arp_ndp_retransmit_timeout, time);
}
return 0;
}
// No Routing table specified: only a local ip and maybe a gateway
// Old default behavior: if a gw is specified, ALL packets go to this gateway (even those we could send w/o the gw
if (l3->gw.ip) {
- if (likely((l3->flags & FLAG_DST_MAC_KNOWN) && (tsc < l3->gw.arp_update_time) && (tsc < l3->gw.arp_timeout))) {
+ if (likely((l3->flags & FLAG_DST_MAC_KNOWN) && (tsc < l3->gw.arp_ndp_retransmit_timeout) && (tsc < l3->gw.reachable_timeout))) {
memcpy(mac, &l3->gw.mac, sizeof(prox_rte_ether_addr));
return SEND_MBUF;
- } else if (tsc > l3->gw.arp_update_time) {
+ } else if (tsc > l3->gw.arp_ndp_retransmit_timeout) {
// long time since we have successfully sent an arp, send arp
- // If sending ARP failed (ring full) then arp_update_time is not updated to avoid having to wait 1 sec to send ARP REQ again
- *time = &l3->gw.arp_update_time;
+ // If sending ARP failed (ring full) then arp_ndp_retransmit_timeout is not updated to avoid having to wait 1 sec to send ARP REQ again
+ *time = &l3->gw.arp_ndp_retransmit_timeout;
+ l3->gw.arp_ndp_retransmit_timeout = tsc + l3->arp_ndp_retransmit_timeout * hz / 1000;
+
*ip_dst = l3->gw.ip;
- if ((l3->flags & FLAG_DST_MAC_KNOWN) && (tsc < l3->gw.arp_timeout)){
+ if ((l3->flags & FLAG_DST_MAC_KNOWN) && (tsc < l3->gw.reachable_timeout)){
// MAC is valid in the table => send also the mbuf
memcpy(mac, &l3->gw.mac, sizeof(prox_rte_ether_addr));
- return SEND_MBUF_AND_ARP;
+ return SEND_MBUF_AND_ARP_ND;
} else {
// MAC still unknown, or timed out => only send ARP
- return SEND_ARP;
+ return SEND_ARP_ND;
}
} else {
// MAC is unknown and we already sent an ARP recently, drop mbuf and wait for ARP reply
@@ -201,17 +282,16 @@ int write_dst_mac(struct task_base *tbase, struct rte_mbuf *mbuf, uint32_t *ip_d
if (likely(l3->n_pkts < 4)) {
for (unsigned int idx = 0; idx < l3->n_pkts; idx++) {
if (*ip_dst == l3->optimized_arp_table[idx].ip) {
- // IP address already in table
- return update_mac_and_send_mbuf(&l3->optimized_arp_table[idx], mac, tsc, hz, l3->arp_update_time, time);
+ return update_mac_and_send_mbuf(&l3->optimized_arp_table[idx], mac, tsc, hz, l3->arp_ndp_retransmit_timeout, time);
}
}
// IP address not found in table
l3->optimized_arp_table[l3->n_pkts].ip = *ip_dst;
- *time = &l3->optimized_arp_table[l3->n_pkts].arp_update_time;
+ *time = &l3->optimized_arp_table[l3->n_pkts].arp_ndp_retransmit_timeout;
l3->n_pkts++;
if (l3->n_pkts < 4) {
- return SEND_ARP;
+ return SEND_ARP_ND;
}
// We have too many IP addresses to search linearly; lets use hash table instead => copy all entries in hash table
@@ -228,16 +308,137 @@ int write_dst_mac(struct task_base *tbase, struct rte_mbuf *mbuf, uint32_t *ip_d
memcpy(&l3->arp_table[ret], &l3->optimized_arp_table[idx], sizeof(struct arp_table));
}
}
- return SEND_ARP;
+ return SEND_ARP_ND;
} else {
// Find IP in lookup table. Send ARP if not found
int ret = rte_hash_lookup(l3->ip_hash, (const void *)ip_dst);
if (unlikely(ret < 0)) {
// IP not found, try to send an ARP
- return add_key_and_send_arp(l3->ip_hash, ip_dst, &l3->arp_table[ret], tsc, hz, l3->arp_update_time, MAX_HOP_INDEX, time);
+ return add_key_and_send_arp(l3->ip_hash, ip_dst, &l3->arp_table[ret], tsc, hz, l3->arp_ndp_retransmit_timeout, MAX_HOP_INDEX, time);
+ } else {
+ // IP has been found
+ return update_mac_and_send_mbuf(&l3->arp_table[ret], mac, tsc, hz, l3->arp_ndp_retransmit_timeout, time);
+ }
+ }
+ // Should not happen
+ return DROP_MBUF;
+}
+
+int write_ip6_dst_mac(struct task_base *tbase, struct rte_mbuf *mbuf, struct ipv6_addr *ip_dst)
+{
+ const uint64_t hz = rte_get_tsc_hz();
+ prox_rte_ether_hdr *packet = rte_pktmbuf_mtod(mbuf, prox_rte_ether_hdr *);
+ prox_rte_ether_addr *mac = &packet->d_addr;
+ struct ipv6_addr *used_ip_src;
+
+ uint64_t tsc = rte_rdtsc();
+ uint16_t len = rte_pktmbuf_pkt_len(mbuf);
+
+ struct ipv6_addr *pkt_src_ip6;
+ if ((pkt_src_ip6 = find_ip6(packet, len, ip_dst)) == NULL) {
+ // Unable to find IP address => non IP packet => send it as it
+ return SEND_MBUF;
+ }
+ struct l3_base *l3 = &(tbase->l3);
+ if (memcmp(&l3->local_ipv6, ip_dst, 8) == 0) {
+ // Same prefix as local -> use local
+ used_ip_src = &l3->local_ipv6;
+ } else if (memcmp(&l3->global_ipv6 , &null_addr, 16) != 0) {
+ // Global IP is defined -> use it
+ used_ip_src = &l3->global_ipv6;
+ } else {
+ plog_info("Error as trying to send a packet to "IPv6_BYTES_FMT" using "IPv6_BYTES_FMT" (local)\n", IPv6_BYTES(ip_dst->bytes), IPv6_BYTES(l3->local_ipv6.bytes));
+ return DROP_MBUF;
+ }
+
+ memcpy(pkt_src_ip6, used_ip_src, sizeof(struct ipv6_addr));
+ if (likely(l3->n_pkts < 4)) {
+ for (unsigned int idx = 0; idx < l3->n_pkts; idx++) {
+ if (memcmp(ip_dst, &l3->optimized_arp_table[idx].ip6, sizeof(struct ipv6_addr)) == 0) {
+ // IP address already in table
+ if ((tsc < l3->optimized_arp_table[idx].arp_ndp_retransmit_timeout) && (tsc < l3->optimized_arp_table[idx].reachable_timeout)) {
+ // MAC address was recently updated in table, use it
+ // plog_dbg("Valid MAC address found => send packet\n");
+ memcpy(mac, &l3->optimized_arp_table[idx].mac, sizeof(prox_rte_ether_addr));
+ return SEND_MBUF;
+ } else if (tsc > l3->optimized_arp_table[idx].arp_ndp_retransmit_timeout) {
+ // NDP not sent since a long time, send NDP
+ l3->optimized_arp_table[idx].arp_ndp_retransmit_timeout = tsc + l3->arp_ndp_retransmit_timeout * hz / 1000;
+ if (tsc < l3->optimized_arp_table[idx].reachable_timeout) {
+ // MAC still valid => also send mbuf
+ plog_dbg("Valid MAC found but NDP retransmit timeout => send packet and NDP\n");
+ memcpy(mac, &l3->optimized_arp_table[idx].mac, sizeof(prox_rte_ether_addr));
+ return SEND_MBUF_AND_ARP_ND;
+ } else {
+ plog_dbg("Unknown MAC => send NDP but cannot send packet\n");
+ // MAC unvalid => only send NDP
+ return SEND_ARP_ND;
+ }
+ } else {
+ // NDP timeout elapsed, MAC not valid anymore but waiting for NDP reply
+ // plog_dbg("NDP reachable timeout elapsed - waiting for NDP reply\n");
+ return DROP_MBUF;
+ }
+ }
+ }
+ // IP address not found in table
+ memcpy(&l3->optimized_arp_table[l3->n_pkts].ip6, ip_dst, sizeof(struct ipv6_addr));
+ l3->optimized_arp_table[l3->n_pkts].arp_ndp_retransmit_timeout = tsc + l3->arp_ndp_retransmit_timeout * hz / 1000;
+ l3->n_pkts++;
+
+ if (l3->n_pkts < 4) {
+ return SEND_ARP_ND;
+ }
+
+ // We have too many IP addresses to search linearly; lets use hash table instead => copy all entries in hash table
+ for (uint32_t idx = 0; idx < l3->n_pkts; idx++) {
+ struct ipv6_addr *ip6 = &l3->optimized_arp_table[idx].ip6;
+ int ret = rte_hash_add_key(l3->ip6_hash, (const void *)ip6);
+ if (ret < 0) {
+ // This should not happen as few entries so far.
+ // If it happens, we still send the NDP as easier:
+ // If the NDP corresponds to this error, the NDP reply will be ignored
+ // If NDP does not correspond to this error/ip, then NDP reply will be handled.
+ plogx_err("Unable add ip "IPv6_BYTES_FMT" in mac_hash (already %d entries)\n", IPv6_BYTES(ip6->bytes), idx);
+ } else {
+ memcpy(&l3->arp_table[ret], &l3->optimized_arp_table[idx], sizeof(struct arp_table));
+ }
+ }
+ return SEND_ARP_ND;
+ } else {
+ // Find IP in lookup table. Send ND if not found
+ int ret = rte_hash_lookup(l3->ip6_hash, (const void *)ip_dst);
+ if (unlikely(ret < 0)) {
+ // IP not found, try to send an ND
+ int ret = rte_hash_add_key(l3->ip6_hash, (const void *)ip_dst);
+ if (ret < 0) {
+ // No reason to send NDP, as reply would be anyhow ignored
+ plogx_err("Unable to add ip "IPv6_BYTES_FMT" in mac_hash\n", IPv6_BYTES(ip_dst->bytes));
+ return DROP_MBUF;
+ } else {
+ memcpy(&l3->arp_table[ret].ip6, ip_dst, sizeof(struct ipv6_addr));
+ l3->arp_table[ret].arp_ndp_retransmit_timeout = tsc + l3->arp_ndp_retransmit_timeout * hz / 1000;
+ }
+ return SEND_ARP_ND;
} else {
// IP has been found
- return update_mac_and_send_mbuf(&l3->arp_table[ret], mac, tsc, hz, l3->arp_update_time, time);
+ if (likely((tsc < l3->arp_table[ret].arp_ndp_retransmit_timeout) && (tsc < l3->arp_table[ret].reachable_timeout))) {
+ // MAC still valid and NDP sent recently
+ memcpy(mac, &l3->arp_table[ret].mac, sizeof(prox_rte_ether_addr));
+ return SEND_MBUF;
+ } else if (tsc > l3->arp_table[ret].arp_ndp_retransmit_timeout) {
+ // NDP not sent since a long time, send NDP
+ l3->arp_table[ret].arp_ndp_retransmit_timeout = tsc + l3->arp_ndp_retransmit_timeout * hz / 1000;
+ if (tsc < l3->arp_table[ret].reachable_timeout) {
+ // MAC still valid => send also MBUF
+ memcpy(mac, &l3->arp_table[ret].mac, sizeof(prox_rte_ether_addr));
+ return SEND_MBUF_AND_ARP_ND;
+ } else {
+ return SEND_ARP_ND;
+ }
+ } else {
+ return DROP_MBUF;
+ }
}
}
// Should not happen
@@ -260,12 +461,22 @@ void task_init_l3(struct task_base *tbase, struct task_args *targ)
.hash_func = rte_hash_crc,
.hash_func_init_val = 0,
};
- tbase->l3.ip_hash = rte_hash_create(&hash_params);
- PROX_PANIC(tbase->l3.ip_hash == NULL, "Failed to set up ip hash table\n");
+ if (targ->flags & TASK_ARG_L3) {
+ plog_info("\tInitializing L3 (IPv4)\n");
+ tbase->l3.ip_hash = rte_hash_create(&hash_params);
+ PROX_PANIC(tbase->l3.ip_hash == NULL, "Failed to set up ip hash table\n");
+ hash_name[0]++;
+ }
+ if (targ->flags & TASK_ARG_NDP) {
+ plog_info("\tInitializing NDP (IPv6)\n");
+ hash_params.key_len = sizeof(struct ipv6_addr);
+ tbase->l3.ip6_hash = rte_hash_create(&hash_params);
+ PROX_PANIC(tbase->l3.ip6_hash == NULL, "Failed to set up ip hash table\n");
+ }
tbase->l3.arp_table = (struct arp_table *)prox_zmalloc(n_entries * sizeof(struct arp_table), socket_id);
- PROX_PANIC(tbase->l3.arp_table == NULL, "Failed to allocate memory for %u entries in arp table\n", n_entries);
- plog_info("\tarp table, with %d entries of size %ld\n", n_entries, sizeof(struct l3_base));
+ PROX_PANIC(tbase->l3.arp_table == NULL, "Failed to allocate memory for %u entries in arp/ndp table\n", n_entries);
+ plog_info("\tarp/ndp table, with %d entries of size %ld\n", n_entries, sizeof(struct l3_base));
targ->lconf->ctrl_func_p[targ->task] = handle_ctrl_plane_pkts;
targ->lconf->ctrl_timeout = freq_to_tsc(targ->ctrl_freq);
@@ -275,36 +486,36 @@ void task_init_l3(struct task_base *tbase, struct task_args *targ)
tbase->l3.task_id = targ->id;
tbase->l3.tmaster = targ->tmaster;
tbase->l3.seed = (uint)rte_rdtsc();
- if (targ->arp_timeout != 0)
- tbase->l3.arp_timeout = targ->arp_timeout;
+ if (targ->reachable_timeout != 0)
+ tbase->l3.reachable_timeout = targ->reachable_timeout;
else
- tbase->l3.arp_timeout = DEFAULT_ARP_TIMEOUT;
- if (targ->arp_update_time != 0)
- tbase->l3.arp_update_time = targ->arp_update_time;
+ tbase->l3.reachable_timeout = DEFAULT_ARP_TIMEOUT;
+ if (targ->arp_ndp_retransmit_timeout != 0)
+ tbase->l3.arp_ndp_retransmit_timeout = targ->arp_ndp_retransmit_timeout;
else
- tbase->l3.arp_update_time = DEFAULT_ARP_UPDATE_TIME;
+ tbase->l3.arp_ndp_retransmit_timeout = DEFAULT_ARP_UPDATE_TIME;
}
void task_start_l3(struct task_base *tbase, struct task_args *targ)
{
- const int NB_ARP_MBUF = 1024;
- const int ARP_MBUF_SIZE = 2048;
- const int NB_CACHE_ARP_MBUF = 256;
const int socket_id = rte_lcore_to_socket_id(targ->lconf->id);
+ const int NB_ARP_ND_MBUF = 1024;
+ const int ARP_ND_MBUF_SIZE = 2048;
+ const int NB_CACHE_ARP_ND_MBUF = 256;
struct prox_port_cfg *port = find_reachable_port(targ);
- if (port && (tbase->l3.arp_pool == NULL)) {
+ if (port && (tbase->l3.arp_nd_pool == NULL)) {
static char name[] = "arp0_pool";
tbase->l3.reachable_port_id = port - prox_port_cfg;
if (targ->local_ipv4) {
- tbase->local_ipv4 = rte_be_to_cpu_32(targ->local_ipv4);
- register_ip_to_ctrl_plane(tbase->l3.tmaster, tbase->local_ipv4, tbase->l3.reachable_port_id, targ->lconf->id, targ->id);
+ tbase->l3.local_ipv4 = rte_be_to_cpu_32(targ->local_ipv4);
+ register_ip_to_ctrl_plane(tbase->l3.tmaster, tbase->l3.local_ipv4, tbase->l3.reachable_port_id, targ->lconf->id, targ->id);
}
if (strcmp(targ->route_table, "") != 0) {
struct lpm4 *lpm;
int ret;
- PROX_PANIC(tbase->local_ipv4 == 0, "missing local_ipv4 will route table is specified in L3 mode\n");
+ PROX_PANIC(tbase->l3.local_ipv4 == 0, "missing local_ipv4 will route table is specified in L3 mode\n");
// LPM might be modified runtime => do not share with other cores
ret = lua_to_lpm4(prox_lua(), GLOBAL, targ->route_table, socket_id, &lpm);
@@ -330,7 +541,7 @@ void task_start_l3(struct task_base *tbase, struct task_args *targ)
// Last but one "next_hop_index" is not a gateway but direct routes
tbase->l3.next_hops[tbase->l3.nb_gws].ip = 0;
ret = rte_lpm_add(tbase->l3.ipv4_lpm, targ->local_ipv4, targ->local_prefix, tbase->l3.nb_gws++);
- PROX_PANIC(ret, "Failed to add local_ipv4 "IPv4_BYTES_FMT"/%d to lpm\n", IP4(tbase->local_ipv4), targ->local_prefix);
+ PROX_PANIC(ret, "Failed to add local_ipv4 "IPv4_BYTES_FMT"/%d to lpm\n", IP4(tbase->l3.local_ipv4), targ->local_prefix);
// Last "next_hop_index" is default gw
tbase->l3.next_hops[tbase->l3.nb_gws].ip = rte_bswap32(targ->gateway_ipv4);
if (targ->gateway_ipv4) {
@@ -340,15 +551,50 @@ void task_start_l3(struct task_base *tbase, struct task_args *targ)
}
master_init_vdev(tbase->l3.tmaster, tbase->l3.reachable_port_id, targ->lconf->id, targ->id);
+
+ // Create IPv6 addr if none were configured
+ if (targ->flags & TASK_ARG_NDP) {
+ if (!memcmp(&targ->local_ipv6, &null_addr, 16)) {
+ set_link_local(&targ->local_ipv6);
+ set_EUI(&targ->local_ipv6, &port->eth_addr);
+ }
+ plog_info("\tCore %d, task %d, local IPv6 addr is "IPv6_BYTES_FMT" (%s)\n",
+ targ->lconf->id, targ->id,
+ IPv6_BYTES(targ->local_ipv6.bytes),
+ IP6_Canonical(&targ->local_ipv6));
+ memcpy(&tbase->l3.local_ipv6, &targ->local_ipv6, sizeof(struct ipv6_addr));
+
+ if (memcmp(&targ->global_ipv6, &null_addr, sizeof(struct ipv6_addr))) {
+ memcpy(&tbase->l3.global_ipv6, &targ->global_ipv6, sizeof(struct ipv6_addr));
+ plog_info("\tCore %d, task %d, global IPv6 addr is "IPv6_BYTES_FMT" (%s)\n",
+ targ->lconf->id, targ->id,
+ IPv6_BYTES(targ->global_ipv6.bytes),
+ IP6_Canonical(&targ->global_ipv6));
+ }
+ if (targ->ipv6_router)
+ register_router_to_ctrl_plane(tbase->l3.tmaster, tbase->l3.reachable_port_id, targ->lconf->id, targ->id, &targ->local_ipv6, &targ->global_ipv6, &targ->router_prefix);
+ else
+ register_node_to_ctrl_plane(tbase->l3.tmaster, &targ->local_ipv6, &targ->global_ipv6, tbase->l3.reachable_port_id, targ->lconf->id, targ->id);
+ }
+
name[3]++;
- struct rte_mempool *ret = rte_mempool_create(name, NB_ARP_MBUF, ARP_MBUF_SIZE, NB_CACHE_ARP_MBUF,
+ struct rte_mempool *ret = rte_mempool_create(name, NB_ARP_ND_MBUF, ARP_ND_MBUF_SIZE, NB_CACHE_ARP_ND_MBUF,
sizeof(struct rte_pktmbuf_pool_private), rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, 0,
rte_socket_id(), 0);
- PROX_PANIC(ret == NULL, "Failed to allocate ARP memory pool on socket %u with %u elements\n",
- rte_socket_id(), NB_ARP_MBUF);
- plog_info("\t\tMempool %p (%s) size = %u * %u cache %u, socket %d\n", ret, name, NB_ARP_MBUF,
- ARP_MBUF_SIZE, NB_CACHE_ARP_MBUF, rte_socket_id());
- tbase->l3.arp_pool = ret;
+ PROX_PANIC(ret == NULL, "Failed to allocate ARP/ND memory pool on socket %u with %u elements\n",
+ rte_socket_id(), NB_ARP_ND_MBUF);
+ plog_info("\tMempool %p (%s) size = %u * %u cache %u, socket %d (for ARP/ND)\n", ret, name, NB_ARP_ND_MBUF,
+ ARP_ND_MBUF_SIZE, NB_CACHE_ARP_ND_MBUF, rte_socket_id());
+ tbase->l3.arp_nd_pool = ret;
+ if ((targ->flags & TASK_ARG_NDP) && (!targ->ipv6_router)) {
+ plog_info("Sending Router Sollicitation\n");
+ send_router_sollicitation(tbase, targ);
+ }
+ if ((targ->flags & TASK_ARG_NDP) && (targ->flags & TASK_ARG_SEND_NA_AT_STARTUP)) {
+ plog_info("Sending unsollicited Neighbour Advertisement\n");
+ send_unsollicited_neighbour_advertisement(tbase, targ);
+
+ }
}
}
@@ -360,10 +606,10 @@ void task_set_gateway_ip(struct task_base *tbase, uint32_t ip)
void task_set_local_ip(struct task_base *tbase, uint32_t ip)
{
- tbase->local_ipv4 = ip;
+ tbase->l3.local_ipv4 = ip;
}
-static void reset_arp_update_time(struct l3_base *l3, uint32_t ip)
+static void reset_arp_ndp_retransmit_timeout(struct l3_base *l3, uint32_t ip)
{
uint32_t idx;
plogx_dbg("MAC entry for IP "IPv4_BYTES_FMT" timeout in kernel\n", IP4(ip));
@@ -371,9 +617,9 @@ static void reset_arp_update_time(struct l3_base *l3, uint32_t ip)
if (l3->ipv4_lpm) {
int ret = rte_hash_lookup(l3->ip_hash, (const void *)&ip);
if (ret >= 0)
- l3->arp_table[ret].arp_update_time = 0;
+ l3->arp_table[ret].arp_ndp_retransmit_timeout = 0;
} else if (ip == l3->gw.ip) {
- l3->gw.arp_update_time = 0;
+ l3->gw.arp_ndp_retransmit_timeout = 0;
} else if (l3->n_pkts < 4) {
for (idx = 0; idx < l3->n_pkts; idx++) {
uint32_t ip_dst = l3->optimized_arp_table[idx].ip;
@@ -381,12 +627,12 @@ static void reset_arp_update_time(struct l3_base *l3, uint32_t ip)
break;
}
if (idx < l3->n_pkts) {
- l3->optimized_arp_table[idx].arp_update_time = 0;
+ l3->optimized_arp_table[idx].arp_ndp_retransmit_timeout = 0;
}
} else {
int ret = rte_hash_lookup(l3->ip_hash, (const void *)&ip);
if (ret >= 0)
- l3->arp_table[ret].arp_update_time = 0;
+ l3->arp_table[ret].arp_ndp_retransmit_timeout = 0;
}
return;
}
@@ -413,15 +659,18 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
uint32_t ip, ip_dst, idx, gateway_ip, prefix;
prox_next_hop_index_type gateway_index;
int j, ret, modified_route;
+ uint64_t addr;
+ struct ipv6_addr *ip6, *ip6_dst;
uint16_t command;
prox_rte_ether_hdr *hdr;
struct ether_hdr_arp *hdr_arp;
struct l3_base *l3 = &tbase->l3;
uint64_t tsc= rte_rdtsc();
- uint64_t arp_timeout = l3->arp_timeout * hz / 1000;
+ uint64_t reachable_timeout = l3->reachable_timeout * hz / 1000;
uint32_t nh;
prox_rte_ipv4_hdr *pip;
prox_rte_udp_hdr *udp_hdr;
+ uint8_t port = tbase->l3.reachable_port_id;
for (j = 0; j < n_pkts; ++j) {
PREFETCH0(mbufs[j]);
@@ -434,10 +683,10 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
pip = NULL;
udp_hdr = NULL;
out[0] = OUT_HANDLED;
- command = mbufs[j]->udata64 & 0xFFFF;
+ command = get_command(mbufs[j]);
plogx_dbg("\tReceived %s mbuf %p\n", actions_string[command], mbufs[j]);
switch(command) {
- case ROUTE_ADD_FROM_CTRL:
+ case ROUTE_ADD_FROM_MASTER:
ip = ctrl_ring_get_ip(mbufs[j]);
gateway_ip = ctrl_ring_get_gateway_ip(mbufs[j]);
prefix = ctrl_ring_get_prefix(mbufs[j]);
@@ -457,7 +706,7 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
}
tx_drop(mbufs[j]);
break;
- case ROUTE_DEL_FROM_CTRL:
+ case ROUTE_DEL_FROM_MASTER:
ip = ctrl_ring_get_ip(mbufs[j]);
prefix = ctrl_ring_get_prefix(mbufs[j]);
@@ -471,15 +720,15 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
}
tx_drop(mbufs[j]);
break;
- case UPDATE_FROM_CTRL:
+ case MAC_INFO_FROM_MASTER:
hdr_arp = rte_pktmbuf_mtod(mbufs[j], struct ether_hdr_arp *);
- ip = (mbufs[j]->udata64 >> 32) & 0xFFFFFFFF;
+ ip = get_ip(mbufs[j]);
if (prox_rte_is_zero_ether_addr(&hdr_arp->arp.data.sha)) {
// MAC timeout or deleted from kernel table => reset update_time
// This will cause us to send new ARP request
- // However, as arp_timeout not touched, we should continue sending our regular IP packets
- reset_arp_update_time(l3, ip);
+ // However, as reachable_timeout not touched, we should continue sending our regular IP packets
+ reset_arp_ndp_retransmit_timeout(l3, ip);
return;
} else
plogx_dbg("\tUpdating MAC entry for IP "IPv4_BYTES_FMT" with MAC "MAC_BYTES_FMT"\n",
@@ -494,20 +743,20 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
} else if ((nh = l3->arp_table[ret].nh) != MAX_HOP_INDEX) {
entry = &l3->next_hops[nh];
memcpy(&entry->mac, &(hdr_arp->arp.data.sha), sizeof(prox_rte_ether_addr));
- entry->arp_timeout = tsc + arp_timeout;
- update_arp_update_time(l3, &entry->arp_update_time, l3->arp_update_time);
+ entry->reachable_timeout = tsc + reachable_timeout;
+ update_arp_ndp_retransmit_timeout(l3, &entry->arp_ndp_retransmit_timeout, l3->arp_ndp_retransmit_timeout);
} else {
memcpy(&l3->arp_table[ret].mac, &(hdr_arp->arp.data.sha), sizeof(prox_rte_ether_addr));
- l3->arp_table[ret].arp_timeout = tsc + arp_timeout;
- update_arp_update_time(l3, &l3->arp_table[ret].arp_update_time, l3->arp_update_time);
+ l3->arp_table[ret].reachable_timeout = tsc + reachable_timeout;
+ update_arp_ndp_retransmit_timeout(l3, &l3->arp_table[ret].arp_ndp_retransmit_timeout, l3->arp_ndp_retransmit_timeout);
}
}
else if (ip == l3->gw.ip) {
// MAC address of the gateway
memcpy(&l3->gw.mac, &hdr_arp->arp.data.sha, 6);
l3->flags |= FLAG_DST_MAC_KNOWN;
- l3->gw.arp_timeout = tsc + arp_timeout;
- update_arp_update_time(l3, &l3->gw.arp_update_time, l3->arp_update_time);
+ l3->gw.reachable_timeout = tsc + reachable_timeout;
+ update_arp_ndp_retransmit_timeout(l3, &l3->gw.arp_ndp_retransmit_timeout, l3->arp_ndp_retransmit_timeout);
} else if (l3->n_pkts < 4) {
// Few packets tracked - should be faster to loop through them thean using a hash table
for (idx = 0; idx < l3->n_pkts; idx++) {
@@ -517,8 +766,8 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
}
if (idx < l3->n_pkts) {
memcpy(&l3->optimized_arp_table[idx].mac, &(hdr_arp->arp.data.sha), sizeof(prox_rte_ether_addr));
- l3->optimized_arp_table[idx].arp_timeout = tsc + arp_timeout;
- update_arp_update_time(l3, &l3->optimized_arp_table[idx].arp_update_time, l3->arp_update_time);
+ l3->optimized_arp_table[idx].reachable_timeout = tsc + reachable_timeout;
+ update_arp_ndp_retransmit_timeout(l3, &l3->optimized_arp_table[idx].arp_ndp_retransmit_timeout, l3->arp_ndp_retransmit_timeout);
}
} else {
ret = rte_hash_add_key(l3->ip_hash, (const void *)&ip);
@@ -526,21 +775,49 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
plogx_info("Unable add ip "IPv4_BYTES_FMT" in mac_hash\n", IP4(ip));
} else {
memcpy(&l3->arp_table[ret].mac, &(hdr_arp->arp.data.sha), sizeof(prox_rte_ether_addr));
- l3->arp_table[ret].arp_timeout = tsc + arp_timeout;
- update_arp_update_time(l3, &l3->arp_table[ret].arp_update_time, l3->arp_update_time);
+ l3->arp_table[ret].reachable_timeout = tsc + reachable_timeout;
+ update_arp_ndp_retransmit_timeout(l3, &l3->arp_table[ret].arp_ndp_retransmit_timeout, l3->arp_ndp_retransmit_timeout);
+ }
+ }
+ tx_drop(mbufs[j]);
+ break;
+ case MAC_INFO_FROM_MASTER_FOR_IPV6:
+ ip6 = ctrl_ring_get_ipv6_addr(mbufs[j]);
+ uint64_t data = ctrl_ring_get_data(mbufs[j]);
+
+ if (l3->n_pkts < 4) {
+ // Few packets tracked - should be faster to loop through them thean using a hash table
+ for (idx = 0; idx < l3->n_pkts; idx++) {
+ ip6_dst = &l3->optimized_arp_table[idx].ip6;
+ if (memcmp(ip6_dst, ip6, sizeof(struct ipv6_addr)) == 0)
+ break;
+ }
+ if (idx < l3->n_pkts) {
+ // IP found; this is a reply for one of our requests!
+ memcpy(&l3->optimized_arp_table[idx].mac, &data, sizeof(prox_rte_ether_addr));
+ l3->optimized_arp_table[idx].reachable_timeout = tsc + l3->reachable_timeout * hz / 1000;
+ }
+ } else {
+ int ret = rte_hash_add_key(l3->ip6_hash, (const void *)ip6);
+ if (ret < 0) {
+ plogx_info("Unable add ip "IPv6_BYTES_FMT" in mac_hash\n", IPv6_BYTES(ip6->bytes));
+ } else {
+ memcpy(&l3->arp_table[ret].mac, &data, sizeof(prox_rte_ether_addr));
+ l3->arp_table[ret].reachable_timeout = tsc + l3->reachable_timeout * hz / 1000;
}
}
tx_drop(mbufs[j]);
break;
- case ARP_REPLY_FROM_CTRL:
- case ARP_REQ_FROM_CTRL:
+ case SEND_NDP_FROM_MASTER:
+ case SEND_ARP_REQUEST_FROM_MASTER:
+ case SEND_ARP_REPLY_FROM_MASTER:
out[0] = 0;
// tx_ctrlplane_pkt does not drop packets
plogx_dbg("\tForwarding (ARP) packet from master\n");
tbase->aux->tx_ctrlplane_pkt(tbase, &mbufs[j], 1, out);
TASK_STATS_ADD_TX_NON_DP(&tbase->aux->stats, 1);
break;
- case ICMP_FROM_CTRL:
+ case SEND_ICMP_FROM_MASTER:
out[0] = 0;
// tx_ctrlplane_pkt does not drop packets
plogx_dbg("\tForwarding (PING) packet from master\n");
@@ -584,6 +861,26 @@ void handle_ctrl_plane_pkts(struct task_base *tbase, struct rte_mbuf **mbufs, ui
tbase->aux->tx_ctrlplane_pkt(tbase, &mbufs[j], 1, out);
TASK_STATS_ADD_TX_NON_DP(&tbase->aux->stats, 1);
break;
+ case IPV6_INFO_FROM_MASTER:
+ // addr = ctrl_ring_get_data(mbufs[j]);
+ ip6 = ctrl_ring_get_ipv6_addr(mbufs[j]);
+ if (memcmp(&l3->global_ipv6 , &null_addr, 16) == 0) {
+ memcpy(&l3->global_ipv6, ip6, sizeof(struct ipv6_addr));
+ plog_info("Core %d task %d received global IP "IPv6_BYTES_FMT"\n", l3->core_id, l3->task_id, IPv6_BYTES(ip6->bytes));
+ } else if (memcmp(&l3->global_ipv6, ip6, 8) == 0) {
+ if (l3->prefix_printed == 0) {
+ plog_info("Core %d task %d received expected prefix "IPv6_PREFIX_FMT"\n", l3->core_id, l3->task_id, IPv6_PREFIX(ip6->bytes));
+ l3->prefix_printed = 1;
+ }
+ } else {
+ plog_warn("Core %d task %d received unexpected prefix "IPv6_PREFIX_FMT", IP = "IPv6_PREFIX_FMT"\n", l3->core_id, l3->task_id, IPv6_PREFIX(ip6->bytes), IPv6_PREFIX(l3->global_ipv6.bytes));
+ }
+ tx_drop(mbufs[j]);
+ break;
+ default:
+ plog_err("Unexpected message received: %d\n", command);
+ tx_drop(mbufs[j]);
+ break;
}
}
}