void *packet_consumer_thread(void *_i) {
struct dir_info *i = (struct dir_info *) _i;
int tx_queue_not_empty = 0;
if (i->bind_core >= 0)
bind2core(i->bind_core);
while(!do_shutdown) {
if(pfring_zc_recv_pkt(i->inzq, &i->tmpbuff, 0 /* wait_for_packet */) > 0) {
if (unlikely(verbose)) {
#if 1
char bigbuf[4096];
pfring_print_pkt(bigbuf, sizeof(bigbuf), pfring_zc_pkt_buff_data(i->tmpbuff, i->inzq), i->tmpbuff->len, i->tmpbuff->len);
fputs(bigbuf, stdout);
#else
u_char *pkt_data = pfring_zc_pkt_buff_data(i->tmpbuff, i->inzq);
int j;
for(j = 0; j < i->tmpbuff->len; j++)
printf("%02X ", pkt_data[j]);
printf("\n");
#endif
}
i->numPkts++;
i->numBytes += i->tmpbuff->len + 24; /* 8 Preamble + 4 CRC + 12 IFG */
errno = 0;
while (unlikely(pfring_zc_send_pkt(i->outzq, &i->tmpbuff, flush_packet) < 0 && errno != EMSGSIZE && !do_shutdown))
if (wait_for_packet) usleep(1);
tx_queue_not_empty = 1;
} else {
if (tx_queue_not_empty) {
pfring_zc_sync_queue(i->outzq, tx_only);
tx_queue_not_empty = 0;
}
if (wait_for_packet)
usleep(1);
}
}
if (!flush_packet) pfring_zc_sync_queue(i->outzq, tx_only);
pfring_zc_sync_queue(i->inzq, rx_only);
return NULL;
}
static int pfring_zc_daq_acquire_best_effort(void *handle, int cnt, DAQ_Analysis_Func_t callback,
#if (DAQ_API_VERSION >= 0x00010002)
DAQ_Meta_Func_t metaback,
#endif
void *user) {
Pfring_Context_t *context = (Pfring_Context_t *) handle;
int ret = 0, c = 0;
u_int32_t rx_ring_idx = context->num_devices - 1, rx_ring_idx_clone;
u_int32_t mask = 0;
u_char *pkt_buffer;
context->analysis_func = callback;
context->breakloop = 0;
while (!context->breakloop && (cnt <= 0 || c < cnt)) {
#ifdef SIG_RELOAD
if (pfring_zc_daq_reload_requested)
pfring_zc_daq_reload(context);
#endif
while (pfring_zc_daq_in_packets(context, &rx_ring_idx) && !context->breakloop) {
pfring_zc_recv_pkt(context->rx_queues[rx_ring_idx], &context->buffer, 0);
context->buffer->hash = context->ifindexes[rx_ring_idx];
pkt_buffer = pfring_zc_pkt_buff_data(context->buffer, context->rx_queues[rx_ring_idx]);
#ifdef ENABLE_BPF
if (!context->bpf_filter || bpf_filter(context->filter.bf_insns, pkt_buffer, context->buffer->len /* caplen */, context->buffer->len) != 0) { /* accept */
#endif
/* enqueueing pkt (and don't care of no room available) */
mask = 1 << (context->num_devices);
#ifdef ENABLE_BPF
} else {
context->stats.packets_received++;
context->stats.verdicts[DAQ_VERDICT_PASS]++;
}
#endif
mask |= 1 << (rx_ring_idx ^ 0x1);
pfring_zc_send_pkt_multi(context->mq, &context->buffer, mask, 0);
}
rx_ring_idx_clone = rx_ring_idx;
while (!(ret = pfring_zc_daq_in_packets(context, &rx_ring_idx_clone)) && !pfring_zc_queue_is_empty(context->q) && !context->breakloop) {
ret = pfring_zc_recv_pkt(context->q, &context->buffer, 0);
pfring_zc_daq_process(context, context->buffer, user);
c++;
}
if (!ret) {
if (usleep(1) == -1)
if (errno == EINTR)
break;
}
}
return 0;
}
int32_t sysdig_distribution_func(pfring_zc_pkt_buff *pkt_handle, pfring_zc_queue *in_queue, void *user) {
/* NOTE: pkt_handle->hash contains the CPU id */
struct sysdig_event_header *ev = (struct sysdig_event_header*)pfring_zc_pkt_buff_data(pkt_handle, in_queue);
long num_out_queues = (long) user;
return(ev->thread_id % num_out_queues);
}
void* consumer_thread(void* _id) {
long id = (long) _id;
pfring_zc_pkt_buff *b = buffers[id];
bind2core(bind_core[id]);
while(!consumers_stats[id].do_shutdown) {
if(pfring_zc_recv_pkt(outzq[id], &b, wait_for_packet) > 0) {
#if 0
int i;
for(i = 0; i < b->len; i++)
printf("%02X ", pfring_zc_pkt_buff_data(b, outzq[id])[i]);
printf("\n");
#endif
consumers_stats[id].numPkts++;
consumers_stats[id].numBytes += b->len + 24; /* 8 Preamble + 4 CRC + 12 IFG */
}
}
pfring_zc_sync_queue(outzq[id], rx_only);
return NULL;
}
static inline void pfring_zc_daq_process(Pfring_Context_t *context, pfring_zc_pkt_buff *buffer, void *user) {
DAQ_PktHdr_t hdr;
DAQ_Verdict verdict;
u_char *pkt_buffer;
hdr.pktlen = hdr.caplen = buffer->len;
hdr.ts.tv_sec = buffer->ts.tv_sec;
hdr.ts.tv_usec = buffer->ts.tv_nsec/1000;
#if (DAQ_API_VERSION >= 0x00010002)
hdr.ingress_index = buffer->hash;
hdr.egress_index = -1;
hdr.ingress_group = -1;
hdr.egress_group = -1;
#else
hdr.device_index = buffer->hash;
#endif
hdr.flags = 0;
pkt_buffer = pfring_zc_pkt_buff_data(buffer, context->q);
verdict = context->analysis_func(user, &hdr, pkt_buffer);
if (verdict >= MAX_DAQ_VERDICT)
verdict = DAQ_VERDICT_PASS;
switch(verdict) {
case DAQ_VERDICT_BLACKLIST: /* Block the packet and block all future packets in the same flow systemwide. */
/* TODO handle hw filters */
break;
case DAQ_VERDICT_WHITELIST: /* Pass the packet and fastpath all future packets in the same flow systemwide. */
case DAQ_VERDICT_IGNORE: /* Pass the packet and fastpath all future packets in the same flow for this application. */
case DAQ_VERDICT_PASS: /* Pass the packet */
case DAQ_VERDICT_REPLACE: /* Pass a packet that has been modified in-place.(No resizing allowed!) */
case DAQ_VERDICT_BLOCK: /* Block the packet. */
/* Nothing to do really */
break;
case MAX_DAQ_VERDICT:
/* No way we can reach this point */
break;
}
context->stats.packets_received++;
context->stats.verdicts[verdict]++;
}
void recv_packets()
{
int ret;
// Poll for packets
do {
ret = pfring_zc_recv_pkt(pf_recv, &pf_buffer, 0);
if (ret == 0) {
usleep(1000);
}
} while (ret == 0);
// Handle other errors, by not doing anything and logging
if (ret != 1) {
log_error("recv", "Error: %d", ret);
return;
}
// Successfully got a packet, now handle it
uint8_t* pkt_buf = pfring_zc_pkt_buff_data(pf_buffer, pf_recv);
handle_packet(pf_buffer->len, pkt_buf);
}
void *packet_consumer_thread(void *_id) {
struct volatile_globals *g = globals;
bind2core(bind_core);
while(!g->do_shutdown) {
if(pfring_zc_recv_pkt(zq, &buffer, g->wait_for_packet) > 0) {
if (unlikely(g->verbose)) {
u_char *pkt_data = pfring_zc_pkt_buff_data(buffer, zq);
if (buffer->ts.tv_nsec)
printf("[%u.%u] ", buffer->ts.tv_sec, buffer->ts.tv_nsec);
if(g->dump_as_sysdig_event) {
struct sysdig_event_header *ev = (struct sysdig_event_header*)pkt_data;
printf("[cpu_id=%u][tid=%lu][%u|%s]",
buffer->hash, ev->thread_id,
ev->event_type, sysdig_event2name(ev->event_type));
} else {
int i;
for(i = 0; i < buffer->len; i++)
printf("%02X ", pkt_data[i]);
}
printf("\n");
}
g->numPkts++;
g->numBytes += buffer->len + 24; /* 8 Preamble + 4 CRC + 12 IFG */
}
}
pfring_zc_sync_queue(zq, rx_only);
return NULL;
}
static int pfring_zc_daq_inject(void *handle, const DAQ_PktHdr_t *hdr,
const uint8_t *packet_data, uint32_t len, int reverse) {
Pfring_Context_t *context = (Pfring_Context_t *) handle;
int i, tx_ring_idx = DAQ_PF_RING_PASSIVE_DEV_IDX;
if (!(context->mode == DAQ_MODE_INLINE || (context->mode == DAQ_MODE_PASSIVE && context->ids_bridge)))
return DAQ_ERROR;
for (i = 0; i < context->num_devices; i++) {
if (context->ifindexes[i] ==
#if (DAQ_API_VERSION >= 0x00010002)
hdr->ingress_index
#else
hdr->device_index
#endif
) {
tx_ring_idx = i ^ 0x1; /* TODO Check this (do we have to send to i or i ^ 0x1?) */
break;
}
}
memcpy(
pfring_zc_pkt_buff_data(context->buffer_inject, context->rx_queues[tx_ring_idx]),
packet_data,
len
);
if (pfring_zc_send_pkt(context->tx_queues[tx_ring_idx],
&context->buffer_inject, 1 /* flush packet */) < 0) {
DPE(context->errbuf, "%s", "pfring_zc_send_pkt() error");
return DAQ_ERROR;
}
context->stats.packets_injected++;
return DAQ_SUCCESS;
}
void *packet_consumer_thread(void *data) {
if (bind_core >= 0)
bind2core(bind_core);
while(!do_shutdown) {
if(pfring_zc_recv_pkt(inzq, &tmpbuff, wait_for_packet) > 0) {
if (unlikely(verbose)) {
char bigbuf[4096];
pfring_print_pkt(bigbuf, sizeof(bigbuf), pfring_zc_pkt_buff_data(tmpbuff, inzq), tmpbuff->len, tmpbuff->len);
fputs(bigbuf, stdout);
}
#if 0
int i;
u_char *pkt_data = pfring_zc_pkt_buff_data(tmpbuff, inzq),
for(i = 0; i < tmpbuff->len; i++)
printf("%02X ", pkt_data[i]);
printf("\n");
#endif
numPkts++;
numBytes += tmpbuff->len + 24; /* 8 Preamble + 4 CRC + 12 IFG */
while (unlikely(pfring_zc_send_pkt(outzq, &tmpbuff, flush_packet) < 0 && !do_shutdown))
usleep(1);
}
}
if (!flush_packet) pfring_zc_sync_queue(outzq, tx_only);
pfring_zc_sync_queue(inzq, rx_only);
return NULL;
}
void *send_traffic(void *user) {
ticks hz, tick_start = 0, tick_delta = 0;
u_int64_t ts_ns_start = 0, ns_delta = 0;
u_int32_t buffer_id = 0;
int sent_bytes;
#ifdef BURST_API
int i, sent_packets;
#endif
if (bind_core >= 0)
bind2core(bind_core);
if(pps > 0) {
if (use_pulse_time) {
ts_ns_start = *pulse_timestamp_ns;
ns_delta = (double) (1000000000 / pps);
} else {
/* cumputing usleep delay */
tick_start = getticks();
usleep(1);
tick_delta = getticks() - tick_start;
/* cumputing CPU freq */
tick_start = getticks();
usleep(1001);
hz = (getticks() - tick_start - tick_delta) * 1000 /*kHz -> Hz*/;
printf("Estimated CPU freq: %lu Hz\n", (long unsigned int) hz);
tick_delta = (double) (hz / pps);
tick_start = getticks();
}
}
#ifdef BURST_API
/****** Burst API ******/
if (use_pkt_burst_api) {
while (likely(!do_shutdown && (!num_to_send || numPkts < num_to_send))) {
if (!num_queue_buffers || numPkts < num_queue_buffers + NBUFF || num_ips > 1) { /* forge all buffers 1 time */
for (i = 0; i < BURSTLEN; i++) {
buffers[buffer_id + i]->len = packet_len;
if (stdin_packet_len > 0)
memcpy(pfring_zc_pkt_buff_data(buffers[buffer_id + i], zq), stdin_packet, stdin_packet_len);
else
forge_udp_packet(pfring_zc_pkt_buff_data(buffers[buffer_id + i], zq), numPkts + i);
}
}
/* TODO send unsent packets when a burst is partially sent */
while (unlikely((sent_packets = pfring_zc_send_pkt_burst(zq, &buffers[buffer_id], BURSTLEN, flush_packet)) <= 0)) {
if (unlikely(do_shutdown)) break;
if (!active) usleep(1);
}
numPkts += sent_packets;
numBytes += ((packet_len + 24 /* 8 Preamble + 4 CRC + 12 IFG */ ) * sent_packets);
buffer_id += BURSTLEN;
buffer_id &= NBUFFMASK;
if(pps > 0) {
u_int8_t synced = 0;
if (use_pulse_time) {
while(*pulse_timestamp_ns - ts_ns_start < numPkts * ns_delta && !do_shutdown)
if (!synced) pfring_zc_sync_queue(zq, tx_only), synced = 1;
} else {
while((getticks() - tick_start) < (numPkts * tick_delta))
if (!synced) pfring_zc_sync_queue(zq, tx_only), synced = 1;
}
}
}
} else {
#endif
/****** Packet API ******/
while (likely(!do_shutdown && (!num_to_send || numPkts < num_to_send))) {
buffers[buffer_id]->len = packet_len;
#if 1
if (!num_queue_buffers || numPkts < num_queue_buffers + NBUFF || num_ips > 1) { /* forge all buffers 1 time */
if (stdin_packet_len > 0)
memcpy(pfring_zc_pkt_buff_data(buffers[buffer_id], zq), stdin_packet, stdin_packet_len);
else
forge_udp_packet(pfring_zc_pkt_buff_data(buffers[buffer_id], zq), numPkts);
}
#else
{
u_char *pkt_data = pfring_zc_pkt_buff_data(buffers[buffer_id], zq);
int k;
u_int8_t j = numPkts;
for(k = 0; k < buffers[buffer_id]->len; k++)
pkt_data[k] = j++;
pkt_data[k-1] = cluster_id;
}
#endif
if (append_timestamp)
buffers[buffer_id]->len = append_packet_ts(pfring_zc_pkt_buff_data(buffers[buffer_id], zq), buffers[buffer_id]->len);
while (unlikely((sent_bytes = pfring_zc_send_pkt(zq, &buffers[buffer_id], flush_packet)) < 0)) {
if (unlikely(do_shutdown)) break;
if (!active) usleep(1);
}
numPkts++;
numBytes += sent_bytes + 24; /* 8 Preamble + 4 CRC + 12 IFG */
buffer_id++;
buffer_id &= NBUFFMASK;
if(pps > 0) {
u_int8_t synced = 0;
if (use_pulse_time) {
while(*pulse_timestamp_ns - ts_ns_start < numPkts * ns_delta && !do_shutdown)
if (!synced) pfring_zc_sync_queue(zq, tx_only), synced = 1;
} else {
while((getticks() - tick_start) < (numPkts * tick_delta))
if (!synced) pfring_zc_sync_queue(zq, tx_only), synced = 1;
}
}
}
#ifdef BURST_API
}
#endif
if (!flush_packet)
pfring_zc_sync_queue(zq, tx_only);
return NULL;
}
static int pfring_zc_daq_acquire(void *handle, int cnt, DAQ_Analysis_Func_t callback,
#if (DAQ_API_VERSION >= 0x00010002)
DAQ_Meta_Func_t metaback,
#endif
void *user) {
Pfring_Context_t *context = (Pfring_Context_t *) handle;
int ret = 0, i = 0, rx_ring_idx = context->num_devices - 1, c = 0;
DAQ_PktHdr_t hdr;
DAQ_Verdict verdict;
u_char *pkt_buffer;
#ifdef DAQ_PF_RING_BEST_EFFORT_BOOST
if (context->mode == DAQ_MODE_PASSIVE && context->ids_bridge == 2)
return pfring_zc_daq_acquire_best_effort(handle, cnt, callback,
#if (DAQ_API_VERSION >= 0x00010002)
metaback,
#endif
user);
#endif
context->analysis_func = callback;
context->breakloop = 0;
while (!context->breakloop && (cnt <= 0 || c < cnt)) {
#ifdef SIG_RELOAD
if (pfring_zc_daq_reload_requested)
pfring_zc_daq_reload(context);
#endif
for (i = 0; i < context->num_devices; i++) {
rx_ring_idx = (rx_ring_idx + 1) % context->num_devices;
ret = pfring_zc_recv_pkt(context->rx_queues[rx_ring_idx], &context->buffer, 0 /* Dont't wait */);
if (ret > 0)
break;
}
if (ret <= 0) {
if (usleep(1) == -1)
if (errno == EINTR)
break;
continue;
}
hdr.pktlen = hdr.caplen = context->buffer->len;
hdr.ts.tv_sec = context->buffer->ts.tv_sec;
hdr.ts.tv_usec = context->buffer->ts.tv_nsec/1000;
#if (DAQ_API_VERSION >= 0x00010002)
hdr.ingress_index = context->ifindexes[rx_ring_idx];
hdr.egress_index = -1;
hdr.ingress_group = -1;
hdr.egress_group = -1;
#else
hdr.device_index = context->ifindexes[rx_ring_idx];
#endif
hdr.flags = 0;
pkt_buffer = pfring_zc_pkt_buff_data(context->buffer, context->rx_queues[rx_ring_idx]);
#ifdef ENABLE_BPF
if (!context->bpf_filter || bpf_filter(context->filter.bf_insns, pkt_buffer, hdr.caplen, hdr.pktlen) != 0) { /* analyse */
#endif
verdict = context->analysis_func(user, &hdr, pkt_buffer);
#ifdef ENABLE_BPF
} else
verdict = DAQ_VERDICT_PASS;
#endif
if (verdict >= MAX_DAQ_VERDICT)
verdict = DAQ_VERDICT_PASS;
if (context->mode == DAQ_MODE_PASSIVE && context->ids_bridge) { /* always forward the packet */
pfring_zc_daq_send_packet(context, context->tx_queues[rx_ring_idx ^ 0x1], hdr.caplen);
} else if (context->mode == DAQ_MODE_INLINE && verdict != DAQ_VERDICT_PASS /* optimisation */ ) {
/* parsing eth_type to forward ARP */
struct ethhdr *eh = (struct ethhdr *) pkt_buffer;
u_int16_t eth_type = ntohs(eh->h_proto);
u_int16_t vlan_offset = 0;
if (eth_type == 0x8100 /* 802.1q (VLAN) */) {
struct eth_vlan_hdr *vh;
vlan_offset = sizeof(struct ethhdr) - sizeof(struct eth_vlan_hdr);
while (eth_type == 0x8100 /* 802.1q (VLAN) */ ) {
vlan_offset += sizeof(struct eth_vlan_hdr);
vh = (struct eth_vlan_hdr *) &pkt_buffer[vlan_offset];
eth_type = ntohs(vh->h_proto);
}
}
if (eth_type == 0x0806 /* ARP */ )
verdict = DAQ_VERDICT_PASS;
}
switch(verdict) {
case DAQ_VERDICT_BLACKLIST: /* Block the packet and block all future packets in the same flow systemwide. */
/* TODO handle hw filters */
break;
case DAQ_VERDICT_WHITELIST: /* Pass the packet and fastpath all future packets in the same flow systemwide. */
case DAQ_VERDICT_IGNORE: /* Pass the packet and fastpath all future packets in the same flow for this application. */
case DAQ_VERDICT_PASS: /* Pass the packet */
case DAQ_VERDICT_REPLACE: /* Pass a packet that has been modified in-place.(No resizing allowed!) */
if (context->mode == DAQ_MODE_INLINE)
pfring_zc_daq_send_packet(context, context->tx_queues[rx_ring_idx ^ 0x1], hdr.caplen);
break;
case DAQ_VERDICT_BLOCK: /* Block the packet. */
/* Nothing to do really */
break;
case MAX_DAQ_VERDICT:
/* No way we can reach this point */
break;
}
context->stats.packets_received++;
context->stats.verdicts[verdict]++;
c++;
}
return 0;
}
