examples/l3fwd: add event em main loop

Add em main loop for handling events based on capabilities of the
event device.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Acked-by: Jerin Jacob <jerinj@marvell.com>
This commit is contained in:
Pavan Nikhilesh 2020-01-28 11:05:03 +05:30 committed by Jerin Jacob
parent 99fc91d180
commit a434a02d5c
7 changed files with 475 additions and 52 deletions

View File

@ -216,6 +216,16 @@ lpm_event_main_loop_tx_q(__attribute__((unused)) void *dummy);
int
lpm_event_main_loop_tx_q_burst(__attribute__((unused)) void *dummy);
int
em_event_main_loop_tx_d(__attribute__((unused)) void *dummy);
int
em_event_main_loop_tx_d_burst(__attribute__((unused)) void *dummy);
int
em_event_main_loop_tx_q(__attribute__((unused)) void *dummy);
int
em_event_main_loop_tx_q_burst(__attribute__((unused)) void *dummy);
/* Return ipv4/ipv6 fwd lookup struct for LPM or EM. */
void *
em_get_ipv4_l3fwd_lookup_struct(const int socketid);

View File

@ -26,6 +26,7 @@
#include <rte_hash.h>
#include "l3fwd.h"
#include "l3fwd_event.h"
#if defined(RTE_ARCH_X86) || defined(RTE_MACHINE_CPUFLAG_CRC32)
#define EM_HASH_CRC 1
@ -699,6 +700,182 @@ em_main_loop(__attribute__((unused)) void *dummy)
return 0;
}
static __rte_always_inline void
em_event_loop_single(struct l3fwd_event_resources *evt_rsrc,
const uint8_t flags)
{
const int event_p_id = l3fwd_get_free_event_port(evt_rsrc);
const uint8_t tx_q_id = evt_rsrc->evq.event_q_id[
evt_rsrc->evq.nb_queues - 1];
const uint8_t event_d_id = evt_rsrc->event_d_id;
struct lcore_conf *lconf;
unsigned int lcore_id;
struct rte_event ev;
if (event_p_id < 0)
return;
lcore_id = rte_lcore_id();
lconf = &lcore_conf[lcore_id];
RTE_LOG(INFO, L3FWD, "entering %s on lcore %u\n", __func__, lcore_id);
while (!force_quit) {
if (!rte_event_dequeue_burst(event_d_id, event_p_id, &ev, 1, 0))
continue;
struct rte_mbuf *mbuf = ev.mbuf;
#if defined RTE_ARCH_X86 || defined RTE_MACHINE_CPUFLAG_NEON
mbuf->port = em_get_dst_port(lconf, mbuf, mbuf->port);
process_packet(mbuf, &mbuf->port);
#else
l3fwd_em_simple_process(mbuf, lconf);
#endif
if (mbuf->port == BAD_PORT) {
rte_pktmbuf_free(mbuf);
continue;
}
if (flags & L3FWD_EVENT_TX_ENQ) {
ev.queue_id = tx_q_id;
ev.op = RTE_EVENT_OP_FORWARD;
while (rte_event_enqueue_burst(event_d_id, event_p_id,
&ev, 1) && !force_quit)
;
}
if (flags & L3FWD_EVENT_TX_DIRECT) {
rte_event_eth_tx_adapter_txq_set(mbuf, 0);
while (!rte_event_eth_tx_adapter_enqueue(event_d_id,
event_p_id, &ev, 1, 0) &&
!force_quit)
;
}
}
}
static __rte_always_inline void
em_event_loop_burst(struct l3fwd_event_resources *evt_rsrc,
const uint8_t flags)
{
const int event_p_id = l3fwd_get_free_event_port(evt_rsrc);
const uint8_t tx_q_id = evt_rsrc->evq.event_q_id[
evt_rsrc->evq.nb_queues - 1];
const uint8_t event_d_id = evt_rsrc->event_d_id;
const uint16_t deq_len = evt_rsrc->deq_depth;
struct rte_event events[MAX_PKT_BURST];
struct lcore_conf *lconf;
unsigned int lcore_id;
int i, nb_enq, nb_deq;
if (event_p_id < 0)
return;
lcore_id = rte_lcore_id();
lconf = &lcore_conf[lcore_id];
RTE_LOG(INFO, L3FWD, "entering %s on lcore %u\n", __func__, lcore_id);
while (!force_quit) {
/* Read events from RX queues */
nb_deq = rte_event_dequeue_burst(event_d_id, event_p_id,
events, deq_len, 0);
if (nb_deq == 0) {
rte_pause();
continue;
}
#if defined RTE_ARCH_X86 || defined RTE_MACHINE_CPUFLAG_NEON
l3fwd_em_process_events(nb_deq, (struct rte_event **)&events,
lconf);
#else
l3fwd_em_no_opt_process_events(nb_deq,
(struct rte_event **)&events,
lconf);
#endif
for (i = 0; i < nb_deq; i++) {
if (flags & L3FWD_EVENT_TX_ENQ) {
events[i].queue_id = tx_q_id;
events[i].op = RTE_EVENT_OP_FORWARD;
}
if (flags & L3FWD_EVENT_TX_DIRECT)
rte_event_eth_tx_adapter_txq_set(events[i].mbuf,
0);
}
if (flags & L3FWD_EVENT_TX_ENQ) {
nb_enq = rte_event_enqueue_burst(event_d_id, event_p_id,
events, nb_deq);
while (nb_enq < nb_deq && !force_quit)
nb_enq += rte_event_enqueue_burst(event_d_id,
event_p_id, events + nb_enq,
nb_deq - nb_enq);
}
if (flags & L3FWD_EVENT_TX_DIRECT) {
nb_enq = rte_event_eth_tx_adapter_enqueue(event_d_id,
event_p_id, events, nb_deq, 0);
while (nb_enq < nb_deq && !force_quit)
nb_enq += rte_event_eth_tx_adapter_enqueue(
event_d_id, event_p_id,
events + nb_enq,
nb_deq - nb_enq, 0);
}
}
}
static __rte_always_inline void
em_event_loop(struct l3fwd_event_resources *evt_rsrc,
const uint8_t flags)
{
if (flags & L3FWD_EVENT_SINGLE)
em_event_loop_single(evt_rsrc, flags);
if (flags & L3FWD_EVENT_BURST)
em_event_loop_burst(evt_rsrc, flags);
}
int __rte_noinline
em_event_main_loop_tx_d(__attribute__((unused)) void *dummy)
{
struct l3fwd_event_resources *evt_rsrc =
l3fwd_get_eventdev_rsrc();
em_event_loop(evt_rsrc, L3FWD_EVENT_TX_DIRECT | L3FWD_EVENT_SINGLE);
return 0;
}
int __rte_noinline
em_event_main_loop_tx_d_burst(__attribute__((unused)) void *dummy)
{
struct l3fwd_event_resources *evt_rsrc =
l3fwd_get_eventdev_rsrc();
em_event_loop(evt_rsrc, L3FWD_EVENT_TX_DIRECT | L3FWD_EVENT_BURST);
return 0;
}
int __rte_noinline
em_event_main_loop_tx_q(__attribute__((unused)) void *dummy)
{
struct l3fwd_event_resources *evt_rsrc =
l3fwd_get_eventdev_rsrc();
em_event_loop(evt_rsrc, L3FWD_EVENT_TX_ENQ | L3FWD_EVENT_SINGLE);
return 0;
}
int __rte_noinline
em_event_main_loop_tx_q_burst(__attribute__((unused)) void *dummy)
{
struct l3fwd_event_resources *evt_rsrc =
l3fwd_get_eventdev_rsrc();
em_event_loop(evt_rsrc, L3FWD_EVENT_TX_ENQ | L3FWD_EVENT_BURST);
return 0;
}
/*
* Initialize exact match (hash) parameters.
*/

View File

@ -5,12 +5,79 @@
#ifndef __L3FWD_EM_H__
#define __L3FWD_EM_H__
static __rte_always_inline uint16_t
l3fwd_em_handle_ipv4(struct rte_mbuf *m, uint16_t portid,
struct rte_ether_hdr *eth_hdr, struct lcore_conf *qconf)
{
struct rte_ipv4_hdr *ipv4_hdr;
uint16_t dst_port;
/* Handle IPv4 headers.*/
ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
sizeof(struct rte_ether_hdr));
#ifdef DO_RFC_1812_CHECKS
/* Check to make sure the packet is valid (RFC1812) */
if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt_len) < 0) {
rte_pktmbuf_free(m);
return BAD_PORT;
}
#endif
dst_port = em_get_ipv4_dst_port(ipv4_hdr, portid,
qconf->ipv4_lookup_struct);
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
#ifdef DO_RFC_1812_CHECKS
/* Update time to live and header checksum */
--(ipv4_hdr->time_to_live);
++(ipv4_hdr->hdr_checksum);
#endif
/* dst addr */
*(uint64_t *)&eth_hdr->d_addr = dest_eth_addr[dst_port];
/* src addr */
rte_ether_addr_copy(&ports_eth_addr[dst_port],
&eth_hdr->s_addr);
return dst_port;
}
static __rte_always_inline uint16_t
l3fwd_em_handle_ipv6(struct rte_mbuf *m, uint16_t portid,
struct rte_ether_hdr *eth_hdr, struct lcore_conf *qconf)
{
/* Handle IPv6 headers.*/
struct rte_ipv6_hdr *ipv6_hdr;
uint16_t dst_port;
ipv6_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
sizeof(struct rte_ether_hdr));
dst_port = em_get_ipv6_dst_port(ipv6_hdr, portid,
qconf->ipv6_lookup_struct);
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
/* dst addr */
*(uint64_t *)&eth_hdr->d_addr = dest_eth_addr[dst_port];
/* src addr */
rte_ether_addr_copy(&ports_eth_addr[dst_port],
&eth_hdr->s_addr);
return dst_port;
}
static __rte_always_inline void
l3fwd_em_simple_forward(struct rte_mbuf *m, uint16_t portid,
struct lcore_conf *qconf)
{
struct rte_ether_hdr *eth_hdr;
struct rte_ipv4_hdr *ipv4_hdr;
uint16_t dst_port;
uint32_t tcp_or_udp;
uint32_t l3_ptypes;
@ -20,58 +87,10 @@ l3fwd_em_simple_forward(struct rte_mbuf *m, uint16_t portid,
l3_ptypes = m->packet_type & RTE_PTYPE_L3_MASK;
if (tcp_or_udp && (l3_ptypes == RTE_PTYPE_L3_IPV4)) {
/* Handle IPv4 headers.*/
ipv4_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv4_hdr *,
sizeof(struct rte_ether_hdr));
#ifdef DO_RFC_1812_CHECKS
/* Check to make sure the packet is valid (RFC1812) */
if (is_valid_ipv4_pkt(ipv4_hdr, m->pkt_len) < 0) {
rte_pktmbuf_free(m);
return;
}
#endif
dst_port = em_get_ipv4_dst_port(ipv4_hdr, portid,
qconf->ipv4_lookup_struct);
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
#ifdef DO_RFC_1812_CHECKS
/* Update time to live and header checksum */
--(ipv4_hdr->time_to_live);
++(ipv4_hdr->hdr_checksum);
#endif
/* dst addr */
*(uint64_t *)&eth_hdr->d_addr = dest_eth_addr[dst_port];
/* src addr */
rte_ether_addr_copy(&ports_eth_addr[dst_port],
&eth_hdr->s_addr);
dst_port = l3fwd_em_handle_ipv4(m, portid, eth_hdr, qconf);
send_single_packet(qconf, m, dst_port);
} else if (tcp_or_udp && (l3_ptypes == RTE_PTYPE_L3_IPV6)) {
/* Handle IPv6 headers.*/
struct rte_ipv6_hdr *ipv6_hdr;
ipv6_hdr = rte_pktmbuf_mtod_offset(m, struct rte_ipv6_hdr *,
sizeof(struct rte_ether_hdr));
dst_port = em_get_ipv6_dst_port(ipv6_hdr, portid,
qconf->ipv6_lookup_struct);
if (dst_port >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port) == 0)
dst_port = portid;
/* dst addr */
*(uint64_t *)&eth_hdr->d_addr = dest_eth_addr[dst_port];
/* src addr */
rte_ether_addr_copy(&ports_eth_addr[dst_port],
&eth_hdr->s_addr);
dst_port = l3fwd_em_handle_ipv6(m, portid, eth_hdr, qconf);
send_single_packet(qconf, m, dst_port);
} else {
/* Free the mbuf that contains non-IPV4/IPV6 packet */
@ -79,6 +98,25 @@ l3fwd_em_simple_forward(struct rte_mbuf *m, uint16_t portid,
}
}
static __rte_always_inline void
l3fwd_em_simple_process(struct rte_mbuf *m, struct lcore_conf *qconf)
{
struct rte_ether_hdr *eth_hdr;
uint32_t tcp_or_udp;
uint32_t l3_ptypes;
eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
tcp_or_udp = m->packet_type & (RTE_PTYPE_L4_TCP | RTE_PTYPE_L4_UDP);
l3_ptypes = m->packet_type & RTE_PTYPE_L3_MASK;
if (tcp_or_udp && (l3_ptypes == RTE_PTYPE_L3_IPV4))
m->port = l3fwd_em_handle_ipv4(m, m->port, eth_hdr, qconf);
else if (tcp_or_udp && (l3_ptypes == RTE_PTYPE_L3_IPV6))
m->port = l3fwd_em_handle_ipv6(m, m->port, eth_hdr, qconf);
else
m->port = BAD_PORT;
}
/*
* Buffer non-optimized handling of packets, invoked
* from main_loop.
@ -108,4 +146,33 @@ l3fwd_em_no_opt_send_packets(int nb_rx, struct rte_mbuf **pkts_burst,
l3fwd_em_simple_forward(pkts_burst[j], portid, qconf);
}
/*
* Buffer non-optimized handling of events, invoked
* from main_loop.
*/
static inline void
l3fwd_em_no_opt_process_events(int nb_rx, struct rte_event **events,
struct lcore_conf *qconf)
{
int32_t j;
/* Prefetch first packets */
for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++)
rte_prefetch0(rte_pktmbuf_mtod(events[j]->mbuf, void *));
/*
* Prefetch and forward already prefetched
* packets.
*/
for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
rte_prefetch0(rte_pktmbuf_mtod(events[
j + PREFETCH_OFFSET]->mbuf, void *));
l3fwd_em_simple_process(events[j]->mbuf, qconf);
}
/* Forward remaining prefetched packets */
for (; j < nb_rx; j++)
l3fwd_em_simple_process(events[j]->mbuf, qconf);
}
#endif /* __L3FWD_EM_H__ */

View File

@ -75,6 +75,60 @@ em_get_dst_port_ipv6xN(struct lcore_conf *qconf, struct rte_mbuf *m[],
}
}
static __rte_always_inline void
em_get_dst_port_ipv4xN_events(struct lcore_conf *qconf, struct rte_mbuf *m[],
uint16_t dst_port[])
{
int i;
int32_t ret[EM_HASH_LOOKUP_COUNT];
union ipv4_5tuple_host key[EM_HASH_LOOKUP_COUNT];
const void *key_array[EM_HASH_LOOKUP_COUNT];
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++) {
get_ipv4_5tuple(m[i], mask0.x, &key[i]);
key_array[i] = &key[i];
}
rte_hash_lookup_bulk(qconf->ipv4_lookup_struct, &key_array[0],
EM_HASH_LOOKUP_COUNT, ret);
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++) {
dst_port[i] = ((ret[i] < 0) ?
m[i]->port : ipv4_l3fwd_out_if[ret[i]]);
if (dst_port[i] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[i]) == 0)
dst_port[i] = m[i]->port;
}
}
static __rte_always_inline void
em_get_dst_port_ipv6xN_events(struct lcore_conf *qconf, struct rte_mbuf *m[],
uint16_t dst_port[])
{
int i;
int32_t ret[EM_HASH_LOOKUP_COUNT];
union ipv6_5tuple_host key[EM_HASH_LOOKUP_COUNT];
const void *key_array[EM_HASH_LOOKUP_COUNT];
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++) {
get_ipv6_5tuple(m[i], mask1.x, mask2.x, &key[i]);
key_array[i] = &key[i];
}
rte_hash_lookup_bulk(qconf->ipv6_lookup_struct, &key_array[0],
EM_HASH_LOOKUP_COUNT, ret);
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++) {
dst_port[i] = ((ret[i] < 0) ?
m[i]->port : ipv6_l3fwd_out_if[ret[i]]);
if (dst_port[i] >= RTE_MAX_ETHPORTS ||
(enabled_port_mask & 1 << dst_port[i]) == 0)
dst_port[i] = m[i]->port;
}
}
static __rte_always_inline uint16_t
em_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
uint16_t portid)
@ -187,4 +241,81 @@ l3fwd_em_send_packets(int nb_rx, struct rte_mbuf **pkts_burst,
send_packets_multi(qconf, pkts_burst, dst_port, nb_rx);
}
/*
* Buffer optimized handling of events, invoked
* from main_loop.
*/
static inline void
l3fwd_em_process_events(int nb_rx, struct rte_event **ev,
struct lcore_conf *qconf)
{
int32_t i, j, pos;
uint16_t dst_port[MAX_PKT_BURST];
struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
/*
* Send nb_rx - nb_rx % EM_HASH_LOOKUP_COUNT packets
* in groups of EM_HASH_LOOKUP_COUNT.
*/
int32_t n = RTE_ALIGN_FLOOR(nb_rx, EM_HASH_LOOKUP_COUNT);
for (j = 0; j < EM_HASH_LOOKUP_COUNT && j < nb_rx; j++) {
pkts_burst[j] = ev[j]->mbuf;
rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[j],
struct rte_ether_hdr *) + 1);
}
for (j = 0; j < n; j += EM_HASH_LOOKUP_COUNT) {
uint32_t pkt_type = RTE_PTYPE_L3_MASK |
RTE_PTYPE_L4_TCP | RTE_PTYPE_L4_UDP;
uint32_t l3_type, tcp_or_udp;
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++)
pkt_type &= pkts_burst[j + i]->packet_type;
l3_type = pkt_type & RTE_PTYPE_L3_MASK;
tcp_or_udp = pkt_type & (RTE_PTYPE_L4_TCP | RTE_PTYPE_L4_UDP);
for (i = 0, pos = j + EM_HASH_LOOKUP_COUNT;
i < EM_HASH_LOOKUP_COUNT && pos < nb_rx; i++, pos++) {
rte_prefetch0(rte_pktmbuf_mtod(
pkts_burst[pos],
struct rte_ether_hdr *) + 1);
}
if (tcp_or_udp && (l3_type == RTE_PTYPE_L3_IPV4)) {
em_get_dst_port_ipv4xN_events(qconf, &pkts_burst[j],
&dst_port[j]);
} else if (tcp_or_udp && (l3_type == RTE_PTYPE_L3_IPV6)) {
em_get_dst_port_ipv6xN_events(qconf, &pkts_burst[j],
&dst_port[j]);
} else {
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++) {
pkts_burst[j + i]->port = em_get_dst_port(qconf,
pkts_burst[j + i],
pkts_burst[j + i]->port);
process_packet(pkts_burst[j + i],
&pkts_burst[j + i]->port);
}
continue;
}
processx4_step3(&pkts_burst[j], &dst_port[j]);
for (i = 0; i < EM_HASH_LOOKUP_COUNT; i++)
pkts_burst[j + i]->port = dst_port[j + i];
}
for (; j < nb_rx; j++) {
pkts_burst[j]->port = em_get_dst_port(qconf, pkts_burst[j],
pkts_burst[j]->port);
process_packet(pkts_burst[j], &pkts_burst[j]->port);
}
}
#endif /* __L3FWD_EM_HLM_H__ */

View File

@ -95,4 +95,30 @@ l3fwd_em_send_packets(int nb_rx, struct rte_mbuf **pkts_burst,
send_packets_multi(qconf, pkts_burst, dst_port, nb_rx);
}
/*
* Buffer optimized handling of events, invoked
* from main_loop.
*/
static inline void
l3fwd_em_process_events(int nb_rx, struct rte_event **events,
struct lcore_conf *qconf)
{
int32_t i, j;
rte_prefetch0(rte_pktmbuf_mtod(events[0]->mbuf,
struct rte_ether_hdr *) + 1);
for (i = 1, j = 0; j < nb_rx; i++, j++) {
struct rte_mbuf *mbuf = events[j]->mbuf;
if (i < nb_rx) {
rte_prefetch0(rte_pktmbuf_mtod(
events[i]->mbuf,
struct rte_ether_hdr *) + 1);
}
mbuf->port = em_get_dst_port(qconf, mbuf, mbuf->port);
process_packet(mbuf, &mbuf->port);
}
}
#endif /* __L3FWD_EM_SEQUENTIAL_H__ */

View File

@ -217,6 +217,12 @@ l3fwd_event_resource_setup(struct rte_eth_conf *port_conf)
[1][0] = lpm_event_main_loop_tx_q,
[1][1] = lpm_event_main_loop_tx_q_burst,
};
const event_loop_cb em_event_loop[2][2] = {
[0][0] = em_event_main_loop_tx_d,
[0][1] = em_event_main_loop_tx_d_burst,
[1][0] = em_event_main_loop_tx_q,
[1][1] = em_event_main_loop_tx_q_burst,
};
uint32_t event_queue_cfg;
int ret;
@ -251,4 +257,7 @@ l3fwd_event_resource_setup(struct rte_eth_conf *port_conf)
evt_rsrc->ops.lpm_event_loop = lpm_event_loop[evt_rsrc->tx_mode_q]
[evt_rsrc->has_burst];
evt_rsrc->ops.em_event_loop = em_event_loop[evt_rsrc->tx_mode_q]
[evt_rsrc->has_burst];
}

View File

@ -1226,7 +1226,10 @@ main(int argc, char **argv)
/* Configure eventdev parameters if user has requested */
if (evt_rsrc->enabled) {
l3fwd_event_resource_setup(&port_conf);
l3fwd_lkp.main_loop = evt_rsrc->ops.lpm_event_loop;
if (l3fwd_em_on)
l3fwd_lkp.main_loop = evt_rsrc->ops.em_event_loop;
else
l3fwd_lkp.main_loop = evt_rsrc->ops.lpm_event_loop;
l3fwd_event_service_setup();
} else
l3fwd_poll_resource_setup();