rofl_result_t pop_datapacket_offset(datapacket_dpdk_t *dpkt, unsigned int offset, unsigned int num_of_bytes){
uint8_t *src_ptr = get_buffer_dpdk(dpkt);
uint8_t *dst_ptr = (uint8_t*)rte_pktmbuf_adj(dpkt->mbuf, num_of_bytes);
//NOTE dst_ptr = src_ptr + num_of_bytes
if( NULL==dst_ptr )
return ROFL_FAILURE;
if(false==rte_pktmbuf_is_contiguous(dpkt->mbuf)){
assert(0);
return ROFL_FAILURE;
}
// move first bytes backward
memmove(dst_ptr, src_ptr, offset);
#ifndef NDEBUG
// set now unused bytes to 0x00 for easier debugging
memset(src_ptr, 0x00, num_of_bytes);
#endif
dpkt->clas_state.base = get_buffer_dpdk(dpkt);
dpkt->clas_state.len = get_buffer_length_dpdk(dpkt);
return ROFL_SUCCESS;
}
/*
* Upper layer processing for a received Ethernet packet.
*/
void
ether_demux(struct ifnet *ifp, struct rte_mbuf *m)
{
struct ether_hdr *eh;
int i, isr;
u_short ether_type;
//KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
eh = rte_pktmbuf_mtod(m, struct ether_hdr *);
ether_type = rte_be_to_cpu_16(eh->ether_type);
rte_pktmbuf_adj(m, ETHER_HDR_LEN);
/*
* Dispatch frame to upper layer.
*/
switch (ether_type) {
case ETHER_TYPE_IPv4:
isr = NETISR_IP;
break;
case ETHER_TYPE_ARP:
isr = NETISR_ARP;
break;
#ifdef INET6
case ETHER_TYPE_IPv6:
isr = NETISR_IPV6;
break;
#endif
default:
goto discard;
}
netisr_dispatch(isr, m);
return;
discard:
/*
* Packet is to be discarded. If netgraph is present,
* hand the packet to it for last chance processing;
* otherwise dispose of it.
*/
rte_pktmbuf_free(m);
}
/*
* Parse message from vswitchd and send to appropriate handler
*/
static void
handle_vswitchd_cmd(struct rte_mbuf *mbuf)
{
struct dpdk_message *request = NULL;
request = rte_pktmbuf_mtod(mbuf, struct dpdk_message *);
switch (request->type) {
case FLOW_CMD_FAMILY:
handle_flow_cmd(&request->flow_msg);
rte_pktmbuf_free(mbuf);
break;
case PACKET_CMD_FAMILY:
rte_pktmbuf_adj(mbuf, sizeof(*request));
handle_packet_cmd(&request->packet_msg, mbuf);
break;
default:
handle_unknown_cmd();
rte_pktmbuf_free(mbuf);
}
}
/*
* Reassemble fragments into one packet.
*/
struct rte_mbuf *
ipv6_frag_reassemble(const struct ip_frag_pkt *fp)
{
struct ipv6_hdr *ip_hdr;
struct ipv6_extension_fragment *frag_hdr;
struct rte_mbuf *m, *prev;
uint32_t i, n, ofs, first_len;
uint32_t last_len, move_len, payload_len;
first_len = fp->frags[IP_FIRST_FRAG_IDX].len;
n = fp->last_idx - 1;
/*start from the last fragment. */
m = fp->frags[IP_LAST_FRAG_IDX].mb;
ofs = fp->frags[IP_LAST_FRAG_IDX].ofs;
last_len = fp->frags[IP_LAST_FRAG_IDX].len;
payload_len = ofs + last_len;
while (ofs != first_len) {
prev = m;
for (i = n; i != IP_FIRST_FRAG_IDX && ofs != first_len; i--) {
/* previous fragment found. */
if (fp->frags[i].ofs + fp->frags[i].len == ofs) {
ip_frag_chain(fp->frags[i].mb, m);
/* update our last fragment and offset. */
m = fp->frags[i].mb;
ofs = fp->frags[i].ofs;
}
}
/* error - hole in the packet. */
if (m == prev) {
return NULL;
}
}
/* chain with the first fragment. */
ip_frag_chain(fp->frags[IP_FIRST_FRAG_IDX].mb, m);
m = fp->frags[IP_FIRST_FRAG_IDX].mb;
/* update mbuf fields for reassembled packet. */
m->ol_flags |= PKT_TX_IP_CKSUM;
/* update ipv6 header for the reassembled datagram */
ip_hdr = rte_pktmbuf_mtod_offset(m, struct ipv6_hdr *, m->l2_len);
ip_hdr->payload_len = rte_cpu_to_be_16(payload_len);
/*
* remove fragmentation header. note that per RFC2460, we need to update
* the last non-fragmentable header with the "next header" field to contain
* type of the first fragmentable header, but we currently don't support
* other headers, so we assume there are no other headers and thus update
* the main IPv6 header instead.
*/
move_len = m->l2_len + m->l3_len - sizeof(*frag_hdr);
frag_hdr = (struct ipv6_extension_fragment *) (ip_hdr + 1);
ip_hdr->proto = frag_hdr->next_header;
ip_frag_memmove(rte_pktmbuf_mtod_offset(m, char *, sizeof(*frag_hdr)),
rte_pktmbuf_mtod(m, char*), move_len);
rte_pktmbuf_adj(m, sizeof(*frag_hdr));
return m;
}
/*
* Reassemble fragments into one packet.
*/
struct rte_mbuf *
ipv4_frag_reassemble(struct ip_frag_pkt *fp)
{
struct ipv4_hdr *ip_hdr;
struct rte_mbuf *m, *prev;
uint32_t i, n, ofs, first_len;
uint32_t curr_idx = 0;
first_len = fp->frags[IP_FIRST_FRAG_IDX].len;
n = fp->last_idx - 1;
/*start from the last fragment. */
m = fp->frags[IP_LAST_FRAG_IDX].mb;
ofs = fp->frags[IP_LAST_FRAG_IDX].ofs;
curr_idx = IP_LAST_FRAG_IDX;
while (ofs != first_len) {
prev = m;
for (i = n; i != IP_FIRST_FRAG_IDX && ofs != first_len; i--) {
/* previous fragment found. */
if(fp->frags[i].ofs + fp->frags[i].len == ofs) {
/* adjust start of the last fragment data. */
rte_pktmbuf_adj(m, (uint16_t)(m->l2_len + m->l3_len));
rte_pktmbuf_chain(fp->frags[i].mb, m);
/* this mbuf should not be accessed directly */
fp->frags[curr_idx].mb = NULL;
curr_idx = i;
/* update our last fragment and offset. */
m = fp->frags[i].mb;
ofs = fp->frags[i].ofs;
}
}
/* error - hole in the packet. */
if (m == prev) {
return NULL;
}
}
/* chain with the first fragment. */
rte_pktmbuf_adj(m, (uint16_t)(m->l2_len + m->l3_len));
rte_pktmbuf_chain(fp->frags[IP_FIRST_FRAG_IDX].mb, m);
m = fp->frags[IP_FIRST_FRAG_IDX].mb;
/* update mbuf fields for reassembled packet. */
m->ol_flags |= PKT_TX_IP_CKSUM;
/* update ipv4 header for the reassmebled packet */
ip_hdr = rte_pktmbuf_mtod_offset(m, struct ipv4_hdr *, m->l2_len);
ip_hdr->total_length = rte_cpu_to_be_16((uint16_t)(fp->total_size +
m->l3_len));
ip_hdr->fragment_offset = (uint16_t)(ip_hdr->fragment_offset &
rte_cpu_to_be_16(IPV4_HDR_DF_FLAG));
ip_hdr->hdr_checksum = 0;
return m;
}
char *rte_pktmbuf_adj_export(struct rte_mbuf *m, uint16_t len) {
return rte_pktmbuf_adj(m, len);
}
/*
* test data manipulation in mbuf
*/
static int
test_one_pktmbuf(void)
{
struct rte_mbuf *m = NULL;
char *data, *data2, *hdr;
unsigned i;
printf("Test pktmbuf API\n");
/* alloc a mbuf */
m = rte_pktmbuf_alloc(pktmbuf_pool);
if (m == NULL)
GOTO_FAIL("Cannot allocate mbuf");
if (rte_pktmbuf_pkt_len(m) != 0)
GOTO_FAIL("Bad length");
rte_pktmbuf_dump(m, 0);
/* append data */
data = rte_pktmbuf_append(m, MBUF_TEST_DATA_LEN);
if (data == NULL)
GOTO_FAIL("Cannot append data");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad data length");
memset(data, 0x66, rte_pktmbuf_pkt_len(m));
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
rte_pktmbuf_dump(m, MBUF_TEST_DATA_LEN);
rte_pktmbuf_dump(m, 2*MBUF_TEST_DATA_LEN);
/* this append should fail */
data2 = rte_pktmbuf_append(m, (uint16_t)(rte_pktmbuf_tailroom(m) + 1));
if (data2 != NULL)
GOTO_FAIL("Append should not succeed");
/* append some more data */
data2 = rte_pktmbuf_append(m, MBUF_TEST_DATA_LEN2);
if (data2 == NULL)
GOTO_FAIL("Cannot append data");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_DATA_LEN2)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_DATA_LEN2)
GOTO_FAIL("Bad data length");
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
/* trim data at the end of mbuf */
if (rte_pktmbuf_trim(m, MBUF_TEST_DATA_LEN2) < 0)
GOTO_FAIL("Cannot trim data");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad data length");
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
/* this trim should fail */
if (rte_pktmbuf_trim(m, (uint16_t)(rte_pktmbuf_data_len(m) + 1)) == 0)
GOTO_FAIL("trim should not succeed");
/* prepend one header */
hdr = rte_pktmbuf_prepend(m, MBUF_TEST_HDR1_LEN);
if (hdr == NULL)
GOTO_FAIL("Cannot prepend");
if (data - hdr != MBUF_TEST_HDR1_LEN)
GOTO_FAIL("Prepend failed");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_HDR1_LEN)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_HDR1_LEN)
GOTO_FAIL("Bad data length");
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
memset(hdr, 0x55, MBUF_TEST_HDR1_LEN);
/* prepend another header */
hdr = rte_pktmbuf_prepend(m, MBUF_TEST_HDR2_LEN);
if (hdr == NULL)
GOTO_FAIL("Cannot prepend");
if (data - hdr != MBUF_TEST_ALL_HDRS_LEN)
GOTO_FAIL("Prepend failed");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_ALL_HDRS_LEN)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN + MBUF_TEST_ALL_HDRS_LEN)
GOTO_FAIL("Bad data length");
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
memset(hdr, 0x55, MBUF_TEST_HDR2_LEN);
rte_mbuf_sanity_check(m, RTE_MBUF_PKT, 1);
rte_mbuf_sanity_check(m, RTE_MBUF_PKT, 0);
rte_pktmbuf_dump(m, 0);
/* this prepend should fail */
hdr = rte_pktmbuf_prepend(m, (uint16_t)(rte_pktmbuf_headroom(m) + 1));
if (hdr != NULL)
GOTO_FAIL("prepend should not succeed");
/* remove data at beginning of mbuf (adj) */
if (data != rte_pktmbuf_adj(m, MBUF_TEST_ALL_HDRS_LEN))
GOTO_FAIL("rte_pktmbuf_adj failed");
if (rte_pktmbuf_pkt_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad pkt length");
if (rte_pktmbuf_data_len(m) != MBUF_TEST_DATA_LEN)
GOTO_FAIL("Bad data length");
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
/* this adj should fail */
if (rte_pktmbuf_adj(m, (uint16_t)(rte_pktmbuf_data_len(m) + 1)) != NULL)
GOTO_FAIL("rte_pktmbuf_adj should not succeed");
/* check data */
if (!rte_pktmbuf_is_contiguous(m))
GOTO_FAIL("Buffer should be continuous");
for (i=0; i<MBUF_TEST_DATA_LEN; i++) {
if (data[i] != 0x66)
GOTO_FAIL("Data corrupted at offset %u", i);
}
/* free mbuf */
rte_pktmbuf_free(m);
m = NULL;
return 0;
fail:
if (m)
rte_pktmbuf_free(m);
return -1;
}
void app_main_loop_rx_flow(void)
{
const unsigned lcore_id = rte_lcore_id();
struct rte_mbuf *bufs[RX_BURST_SIZE];
struct rte_mbuf *buf;
struct ether_hdr *eth_hdr;
struct ipv4_hdr *ipv4_hdr;
struct ipv6_hdr *ipv6_hdr;
struct tcp_hdr *tcp_hdr;
struct udp_hdr *udp_hdr;
struct pkt_info pktinfo;
int32_t ret;
uint16_t i, n_rx, queueid;
uint8_t port;
port = 0;
queueid = (uint16_t) app.lcore_conf[lcore_id].queue_id;
RTE_LOG(INFO, FLOWATCHER, "[core %u] packet RX & update flow_table Ready\n", lcore_id);
while (!app_quit_signal) {
n_rx = rte_eth_rx_burst(port, queueid, bufs, RX_BURST_SIZE);
if (unlikely(n_rx == 0)) {
port++;
if (port >= app.n_ports)
port = 0;
continue;
}
app_stat[queueid].rx_count += n_rx;
for (i = 0; i < n_rx; i++) {
buf = bufs[i];
pktinfo.timestamp = rte_rdtsc();
pktinfo.pktlen = rte_pktmbuf_pkt_len(buf);
eth_hdr = rte_pktmbuf_mtod(buf, struct ether_hdr *);
/* strip vlan_hdr */
if (eth_hdr->ether_type == rte_cpu_to_be_16(ETHER_TYPE_VLAN)) {
/* struct vlan_hdr *vh = (struct vlan_hdr *) ð_hdr[1]; */
/* buf->ol_flags |= PKT_RX_VLAN_PKT; */
/* buf->vlan_tci = rte_be_to_cpu_16(vh->vlan_tci); */
/* memmove(rte_pktmbuf_adj(buf, sizeof(struct vlan_hdr)), */
/* eth_hdr, 2 * ETHER_ADDR_LEN); */
/* eth_hdr = rte_pktmbuf_mtod(buf, struct ether_hdr *); */
eth_hdr = (struct ether_hdr *) rte_pktmbuf_adj(buf, sizeof(struct vlan_hdr));
}
if (eth_hdr->ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv4)) {
/* IPv4 */
pktinfo.type = PKT_IP_TYPE_IPV4;
ipv4_hdr = (struct ipv4_hdr *) ð_hdr[1];
pktinfo.key.v4.src_ip = rte_be_to_cpu_32(ipv4_hdr->src_addr);
pktinfo.key.v4.dst_ip = rte_be_to_cpu_32(ipv4_hdr->dst_addr);
pktinfo.key.v4.proto = ipv4_hdr->next_proto_id;
switch (ipv4_hdr->next_proto_id) {
case IPPROTO_TCP:
tcp_hdr = (struct tcp_hdr *) &ipv4_hdr[1];
pktinfo.key.v4.src_port = rte_be_to_cpu_16(tcp_hdr->src_port);
pktinfo.key.v4.dst_port = rte_be_to_cpu_16(tcp_hdr->dst_port);
break;
case IPPROTO_UDP:
udp_hdr = (struct udp_hdr *) &ipv4_hdr[1];
pktinfo.key.v4.src_port = rte_be_to_cpu_16(udp_hdr->src_port);
pktinfo.key.v4.dst_port = rte_be_to_cpu_16(udp_hdr->dst_port);
break;
default:
pktinfo.key.v4.src_port = 0;
pktinfo.key.v4.dst_port = 0;
break;
}
rte_pktmbuf_free(buf);
/* update flow_table_v4 */
ret = update_flow_entry(app.flow_table_v4[queueid], &pktinfo);
if (ret == 0)
app_stat[queueid].updated_tbl_v4_count++;
else
app_stat[queueid].miss_updated_tbl_v4_count++;
} else if (eth_hdr->ether_type == rte_cpu_to_be_16(ETHER_TYPE_IPv6)) {
/* IPv6 */
pktinfo.type = PKT_IP_TYPE_IPV6;
ipv6_hdr = (struct ipv6_hdr *) ð_hdr[1];
rte_memcpy(pktinfo.key.v6.src_ip, ipv6_hdr->src_addr, 16);
rte_memcpy(pktinfo.key.v6.dst_ip, ipv6_hdr->dst_addr, 16);
pktinfo.key.v6.proto = ipv6_hdr->proto;
switch (ipv6_hdr->proto) {
case IPPROTO_TCP:
tcp_hdr = (struct tcp_hdr *) &ipv6_hdr[1];
pktinfo.key.v6.src_port = rte_be_to_cpu_16(tcp_hdr->src_port);
pktinfo.key.v6.dst_port = rte_be_to_cpu_16(tcp_hdr->dst_port);
break;
case IPPROTO_UDP:
udp_hdr = (struct udp_hdr *) &ipv6_hdr[1];
pktinfo.key.v6.src_port = rte_be_to_cpu_16(udp_hdr->src_port);
pktinfo.key.v6.dst_port = rte_be_to_cpu_16(udp_hdr->dst_port);
break;
default:
pktinfo.key.v6.src_port = 0;
pktinfo.key.v6.dst_port = 0;
break;
}
rte_pktmbuf_free(buf);
/* update flow_table_v6 */
ret = update_flow_entry(app.flow_table_v6[queueid], &pktinfo);
if (ret == 0)
app_stat[queueid].updated_tbl_v6_count++;
else
app_stat[queueid].miss_updated_tbl_v6_count++;
} else {
/* others */
app_stat[queueid].unknown_pkt_count++;
rte_pktmbuf_free(buf);
continue;
}
}
port++;
if (port >= app.n_ports)
port = 0;
}
RTE_LOG(INFO, FLOWATCHER, "[core %u] packet RX & update flow_table finished\n", lcore_id);
}
static int
paxos_rx_process(struct rte_mbuf *pkt, struct proposer* proposer)
{
int ret = 0;
uint8_t l4_proto = 0;
uint16_t outer_header_len;
union tunnel_offload_info info = { .data = 0 };
struct udp_hdr *udp_hdr;
struct paxos_hdr *paxos_hdr;
struct ether_hdr *phdr = rte_pktmbuf_mtod(pkt, struct ether_hdr *);
parse_ethernet(phdr, &info, &l4_proto);
if (l4_proto != IPPROTO_UDP)
return -1;
udp_hdr = (struct udp_hdr *)((char *)phdr +
info.outer_l2_len + info.outer_l3_len);
/* if UDP dst port is not either PROPOSER or LEARNER port */
if (!(udp_hdr->dst_port == rte_cpu_to_be_16(PROPOSER_PORT) ||
udp_hdr->dst_port == rte_cpu_to_be_16(LEARNER_PORT)) &&
(pkt->packet_type & RTE_PTYPE_TUNNEL_MASK) == 0)
return -1;
paxos_hdr = (struct paxos_hdr *)((char *)udp_hdr + sizeof(struct udp_hdr));
if (rte_get_log_level() == RTE_LOG_DEBUG) {
//rte_hexdump(stdout, "udp", udp_hdr, sizeof(struct udp_hdr));
//rte_hexdump(stdout, "paxos", paxos_hdr, sizeof(struct paxos_hdr));
print_paxos_hdr(paxos_hdr);
}
int value_len = rte_be_to_cpu_16(paxos_hdr->value_len);
struct paxos_value *v = paxos_value_new((char *)paxos_hdr->paxosval, value_len);
switch(rte_be_to_cpu_16(paxos_hdr->msgtype)) {
case PAXOS_PROMISE: {
struct paxos_promise promise = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_promise(proposer, &promise);
break;
}
case PAXOS_ACCEPT: {
if (first_time) {
proposer_preexecute(proposer);
first_time = false;
}
struct paxos_accept acpt = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_accept(proposer, &acpt);
break;
}
case PAXOS_ACCEPTED: {
struct paxos_accepted ack = {
.iid = rte_be_to_cpu_32(paxos_hdr->inst),
.ballot = rte_be_to_cpu_16(paxos_hdr->rnd),
.value_ballot = rte_be_to_cpu_16(paxos_hdr->vrnd),
.aid = rte_be_to_cpu_16(paxos_hdr->acptid),
.value = *v
};
proposer_handle_accepted(proposer, &ack);
break;
}
default:
break;
}
outer_header_len = info.outer_l2_len + info.outer_l3_len
+ sizeof(struct udp_hdr) + sizeof(struct paxos_hdr);
rte_pktmbuf_adj(pkt, outer_header_len);
return ret;
}
static uint16_t
add_timestamps(uint8_t port __rte_unused, uint16_t qidx __rte_unused,
struct rte_mbuf **pkts, uint16_t nb_pkts,
uint16_t max_pkts __rte_unused, void *user_param)
{
struct proposer* proposer = (struct proposer *)user_param;
unsigned i;
uint64_t now = rte_rdtsc();
for (i = 0; i < nb_pkts; i++) {
pkts[i]->udata64 = now;
paxos_rx_process(pkts[i], proposer);
}
return nb_pkts;
}
static inline int
port_init(uint8_t port, struct rte_mempool *mbuf_pool, struct proposer* proposer)
{
struct rte_eth_dev_info dev_info;
struct rte_eth_txconf *txconf;
struct rte_eth_rxconf *rxconf;
struct rte_eth_conf port_conf = port_conf_default;
const uint16_t rx_rings = 1, tx_rings = 1;
int retval;
uint16_t q;
rte_eth_dev_info_get(port, &dev_info);
rxconf = &dev_info.default_rxconf;
txconf = &dev_info.default_txconf;
txconf->txq_flags &= PKT_TX_IPV4;
txconf->txq_flags &= PKT_TX_UDP_CKSUM;
if (port >= rte_eth_dev_count())
return -1;
retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
if (retval != 0)
return retval;
for (q = 0; q < rx_rings; q++) {
retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
rte_eth_dev_socket_id(port), rxconf, mbuf_pool);
if (retval < 0)
return retval;
}
for (q = 0; q < tx_rings; q++) {
retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
rte_eth_dev_socket_id(port), txconf);
if (retval < 0)
return retval;
}
retval = rte_eth_dev_start(port);
if (retval < 0)
return retval;
struct ether_addr addr;
rte_eth_macaddr_get(port, &addr);
rte_eth_promiscuous_enable(port);
rte_eth_add_rx_callback(port, 0, add_timestamps, proposer);
rte_eth_add_tx_callback(port, 0, calc_latency, NULL);
return 0;
}
static void
lcore_main(uint8_t port, __rte_unused struct proposer *p)
{
proposer_preexecute(p);
for (;;) {
// Check if signal is received
if (force_quit)
break;
struct rte_mbuf *bufs[BURST_SIZE];
const uint16_t nb_rx = rte_eth_rx_burst(port, 0, bufs, BURST_SIZE);
if (unlikely(nb_rx == 0))
continue;
uint16_t buf;
for (buf = 0; buf < nb_rx; buf++)
rte_pktmbuf_free(bufs[buf]);
}
}
static __attribute__((noreturn)) int
lcore_mainloop(__attribute__((unused)) void *arg)
{
uint64_t prev_tsc = 0, cur_tsc, diff_tsc;
unsigned lcore_id;
lcore_id = rte_lcore_id();
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_TIMER,
"Starting mainloop on core %u\n", lcore_id);
while(1) {
cur_tsc = rte_rdtsc();
diff_tsc = cur_tsc - prev_tsc;
if (diff_tsc > TIMER_RESOLUTION_CYCLES) {
rte_timer_manage();
prev_tsc = cur_tsc;
}
}
}
static void
report_stat(struct rte_timer *tim, __attribute((unused)) void *arg)
{
/* print stat */
uint32_t count = rte_atomic32_read(&stat);
rte_log(RTE_LOG_INFO, RTE_LOGTYPE_USER8,
"Throughput = %8u msg/s\n", count);
/* reset stat */
rte_atomic32_set(&stat, 0);
/* this timer is automatically reloaded until we decide to stop it */
if (force_quit)
rte_timer_stop(tim);
}
static void
check_timeout(struct rte_timer *tim, void *arg)
{
struct proposer* p = (struct proposer *) arg;
unsigned lcore_id = rte_lcore_id();
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8, "%s() on lcore_id %i\n", __func__, lcore_id);
struct paxos_message out;
out.type = PAXOS_PREPARE;
struct timeout_iterator* iter = proposer_timeout_iterator(p);
while(timeout_iterator_prepare(iter, &out.u.prepare)) {
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8,
"%s Send PREPARE inst %d ballot %d\n",
__func__, out.u.prepare.iid, out.u.prepare.ballot);
send_paxos_message(&out);
}
out.type = PAXOS_ACCEPT;
while(timeout_iterator_accept(iter, &out.u.accept)) {
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8,
"%s: Send ACCEPT inst %d ballot %d\n",
__func__, out.u.prepare.iid, out.u.prepare.ballot);
send_paxos_message(&out);
}
timeout_iterator_free(iter);
/* this timer is automatically reloaded until we decide to stop it */
if (force_quit)
rte_timer_stop(tim);
}
int
main(int argc, char *argv[])
{
uint8_t portid = 0;
unsigned master_core, lcore_id;
signal(SIGTERM, signal_handler);
signal(SIGINT, signal_handler);
force_quit = false;
int proposer_id = 0;
if (rte_get_log_level() == RTE_LOG_DEBUG) {
paxos_config.verbosity = PAXOS_LOG_DEBUG;
}
struct proposer *proposer = proposer_new(proposer_id, NUM_ACCEPTORS);
first_time = true;
/* init EAL */
int ret = rte_eal_init(argc, argv);
if (ret < 0)
rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
/* init timer structure */
rte_timer_init(&timer);
rte_timer_init(&stat_timer);
/* load deliver_timer, every 1 s, on a slave lcore, reloaded automatically */
uint64_t hz = rte_get_timer_hz();
/* Call rte_timer_manage every 10ms */
TIMER_RESOLUTION_CYCLES = hz / 100;
rte_log(RTE_LOG_INFO, RTE_LOGTYPE_USER1, "Clock: %"PRIu64"\n", hz);
/* master core */
master_core = rte_lcore_id();
/* slave core */
lcore_id = rte_get_next_lcore(master_core, 0, 1);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER1, "lcore_id: %d\n", lcore_id);
rte_timer_reset(&timer, hz, PERIODICAL, lcore_id, check_timeout, proposer);
/* reset timer */
rte_eal_remote_launch(lcore_mainloop, NULL, lcore_id);
/* stat core */
lcore_id = rte_get_next_lcore(lcore_id , 0, 1);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER1, "lcore_id: %d\n", lcore_id);
rte_timer_reset(&stat_timer, hz, PERIODICAL, lcore_id,
report_stat, NULL);
/* init RTE timer library */
rte_timer_subsystem_init();
mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
NUM_MBUFS, MBUF_CACHE_SIZE, 0,
RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
if (mbuf_pool == NULL)
rte_exit(EXIT_FAILURE, "Cannot create mbuf_pool\n");
/* reset timer */
rte_eal_remote_launch(lcore_mainloop, NULL, lcore_id);
if (port_init(portid, mbuf_pool, proposer) != 0)
rte_exit(EXIT_FAILURE, "Cannot init port %"PRIu8"\n", portid);
lcore_main(portid, proposer);
rte_log(RTE_LOG_DEBUG, RTE_LOGTYPE_USER8, "Free proposer\n");
proposer_free(proposer);
return 0;
}
static void remove_vtep_hdr(struct pg_bench *bench)
{
for (int i = 0; i < 64; ++i) {
rte_pktmbuf_adj(bench->pkts[i], HEADERS_LENGTH);
}
}
