#include <errno.h>
#include <stdlib.h>
#include <string.h>
-#include <poll.h>
#include <netinet/in.h>
#include <net/if.h>
#include <sys/socket.h>
#include <bpf/xsk.h>
#include <rte_ethdev.h>
-#include <rte_ethdev_driver.h>
-#include <rte_ethdev_vdev.h>
+#include <ethdev_driver.h>
+#include <ethdev_vdev.h>
#include <rte_kvargs.h>
#include <rte_bus_vdev.h>
#include <rte_string_fns.h>
#include "compat.h"
+#ifndef SO_PREFER_BUSY_POLL
+#define SO_PREFER_BUSY_POLL 69
+#endif
+#ifndef SO_BUSY_POLL_BUDGET
+#define SO_BUSY_POLL_BUDGET 70
+#endif
+
#ifndef SOL_XDP
#define SOL_XDP 283
#define ETH_AF_XDP_DFLT_NUM_DESCS XSK_RING_CONS__DEFAULT_NUM_DESCS
#define ETH_AF_XDP_DFLT_START_QUEUE_IDX 0
#define ETH_AF_XDP_DFLT_QUEUE_COUNT 1
+#define ETH_AF_XDP_DFLT_BUSY_BUDGET 64
+#define ETH_AF_XDP_DFLT_BUSY_TIMEOUT 20
-#define ETH_AF_XDP_RX_BATCH_SIZE 32
-#define ETH_AF_XDP_TX_BATCH_SIZE 32
+#define ETH_AF_XDP_RX_BATCH_SIZE XSK_RING_CONS__DEFAULT_NUM_DESCS
+#define ETH_AF_XDP_TX_BATCH_SIZE XSK_RING_CONS__DEFAULT_NUM_DESCS
struct xsk_umem_info {
struct pkt_tx_queue *pair;
struct pollfd fds[1];
int xsk_queue_idx;
+ int busy_budget;
};
struct tx_stats {
#define ETH_AF_XDP_QUEUE_COUNT_ARG "queue_count"
#define ETH_AF_XDP_SHARED_UMEM_ARG "shared_umem"
#define ETH_AF_XDP_PROG_ARG "xdp_prog"
+#define ETH_AF_XDP_BUDGET_ARG "busy_budget"
static const char * const valid_arguments[] = {
ETH_AF_XDP_IFACE_ARG,
ETH_AF_XDP_QUEUE_COUNT_ARG,
ETH_AF_XDP_SHARED_UMEM_ARG,
ETH_AF_XDP_PROG_ARG,
+ ETH_AF_XDP_BUDGET_ARG,
NULL
};
struct xsk_umem_info *umem = rxq->umem;
uint32_t idx_rx = 0;
unsigned long rx_bytes = 0;
- int rcvd, i;
+ int i;
struct rte_mbuf *fq_bufs[ETH_AF_XDP_RX_BATCH_SIZE];
+ nb_pkts = xsk_ring_cons__peek(rx, nb_pkts, &idx_rx);
+
+ if (nb_pkts == 0) {
+ rx_syscall_handler(&rxq->fq, rxq->busy_budget, &rxq->fds[0],
+ rxq->xsk);
+ return 0;
+ }
+
/* allocate bufs for fill queue replenishment after rx */
if (rte_pktmbuf_alloc_bulk(umem->mb_pool, fq_bufs, nb_pkts)) {
AF_XDP_LOG(DEBUG,
"Failed to get enough buffers for fq.\n");
+ /* rollback cached_cons which is added by
+ * xsk_ring_cons__peek
+ */
+ rx->cached_cons -= nb_pkts;
return 0;
}
- rcvd = xsk_ring_cons__peek(rx, nb_pkts, &idx_rx);
-
- if (rcvd == 0) {
-#if defined(XDP_USE_NEED_WAKEUP)
- if (xsk_ring_prod__needs_wakeup(fq))
- (void)poll(rxq->fds, 1, 1000);
-#endif
-
- goto out;
- }
-
- for (i = 0; i < rcvd; i++) {
+ for (i = 0; i < nb_pkts; i++) {
const struct xdp_desc *desc;
uint64_t addr;
uint32_t len;
rx_bytes += len;
}
- xsk_ring_cons__release(rx, rcvd);
-
- (void)reserve_fill_queue(umem, rcvd, fq_bufs, fq);
+ xsk_ring_cons__release(rx, nb_pkts);
+ (void)reserve_fill_queue(umem, nb_pkts, fq_bufs, fq);
/* statistics */
- rxq->stats.rx_pkts += rcvd;
+ rxq->stats.rx_pkts += nb_pkts;
rxq->stats.rx_bytes += rx_bytes;
-out:
- if (rcvd != nb_pkts)
- rte_mempool_put_bulk(umem->mb_pool, (void **)&fq_bufs[rcvd],
- nb_pkts - rcvd);
-
- return rcvd;
+ return nb_pkts;
}
#else
static uint16_t
struct xsk_ring_prod *fq = &rxq->fq;
uint32_t idx_rx = 0;
unsigned long rx_bytes = 0;
- int rcvd, i;
+ int i;
uint32_t free_thresh = fq->size >> 1;
struct rte_mbuf *mbufs[ETH_AF_XDP_RX_BATCH_SIZE];
if (xsk_prod_nb_free(fq, free_thresh) >= free_thresh)
- (void)reserve_fill_queue(umem, ETH_AF_XDP_RX_BATCH_SIZE,
- NULL, fq);
+ (void)reserve_fill_queue(umem, nb_pkts, NULL, fq);
- if (unlikely(rte_pktmbuf_alloc_bulk(rxq->mb_pool, mbufs, nb_pkts) != 0))
- return 0;
-
- rcvd = xsk_ring_cons__peek(rx, nb_pkts, &idx_rx);
- if (rcvd == 0) {
+ nb_pkts = xsk_ring_cons__peek(rx, nb_pkts, &idx_rx);
+ if (nb_pkts == 0) {
#if defined(XDP_USE_NEED_WAKEUP)
if (xsk_ring_prod__needs_wakeup(fq))
(void)poll(rxq->fds, 1, 1000);
#endif
+ return 0;
+ }
- goto out;
+ if (unlikely(rte_pktmbuf_alloc_bulk(rxq->mb_pool, mbufs, nb_pkts))) {
+ /* rollback cached_cons which is added by
+ * xsk_ring_cons__peek
+ */
+ rx->cached_cons -= nb_pkts;
+ return 0;
}
- for (i = 0; i < rcvd; i++) {
+ for (i = 0; i < nb_pkts; i++) {
const struct xdp_desc *desc;
uint64_t addr;
uint32_t len;
bufs[i] = mbufs[i];
}
- xsk_ring_cons__release(rx, rcvd);
+ xsk_ring_cons__release(rx, nb_pkts);
/* statistics */
- rxq->stats.rx_pkts += rcvd;
+ rxq->stats.rx_pkts += nb_pkts;
rxq->stats.rx_bytes += rx_bytes;
-out:
- if (rcvd != nb_pkts)
- rte_mempool_put_bulk(rxq->mb_pool, (void **)&mbufs[rcvd],
- nb_pkts - rcvd);
-
- return rcvd;
+ return nb_pkts;
}
#endif
static uint16_t
-eth_af_xdp_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+af_xdp_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
{
- nb_pkts = RTE_MIN(nb_pkts, ETH_AF_XDP_RX_BATCH_SIZE);
-
#if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
return af_xdp_rx_zc(queue, bufs, nb_pkts);
#else
#endif
}
+static uint16_t
+eth_af_xdp_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+{
+ uint16_t nb_rx;
+
+ if (likely(nb_pkts <= ETH_AF_XDP_RX_BATCH_SIZE))
+ return af_xdp_rx(queue, bufs, nb_pkts);
+
+ /* Split larger batch into smaller batches of size
+ * ETH_AF_XDP_RX_BATCH_SIZE or less.
+ */
+ nb_rx = 0;
+ while (nb_pkts) {
+ uint16_t ret, n;
+
+ n = (uint16_t)RTE_MIN(nb_pkts, ETH_AF_XDP_RX_BATCH_SIZE);
+ ret = af_xdp_rx(queue, &bufs[nb_rx], n);
+ nb_rx = (uint16_t)(nb_rx + ret);
+ nb_pkts = (uint16_t)(nb_pkts - ret);
+ if (ret < n)
+ break;
+ }
+
+ return nb_rx;
+}
+
static void
pull_umem_cq(struct xsk_umem_info *umem, int size, struct xsk_ring_cons *cq)
{
pull_umem_cq(umem, XSK_RING_CONS__DEFAULT_NUM_DESCS, cq);
-#if defined(XDP_USE_NEED_WAKEUP)
- if (xsk_ring_prod__needs_wakeup(&txq->tx))
-#endif
+ if (tx_syscall_needed(&txq->tx))
while (send(xsk_socket__fd(txq->pair->xsk), NULL,
0, MSG_DONTWAIT) < 0) {
/* some thing unexpected */
uint32_t idx_tx;
struct xsk_ring_cons *cq = &txq->pair->cq;
- nb_pkts = RTE_MIN(nb_pkts, ETH_AF_XDP_TX_BATCH_SIZE);
-
pull_umem_cq(umem, nb_pkts, cq);
nb_pkts = rte_ring_dequeue_bulk(umem->buf_ring, addrs,
return nb_pkts;
}
+
+static uint16_t
+af_xdp_tx_cp_batch(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
+{
+ uint16_t nb_tx;
+
+ if (likely(nb_pkts <= ETH_AF_XDP_TX_BATCH_SIZE))
+ return af_xdp_tx_cp(queue, bufs, nb_pkts);
+
+ nb_tx = 0;
+ while (nb_pkts) {
+ uint16_t ret, n;
+
+ /* Split larger batch into smaller batches of size
+ * ETH_AF_XDP_TX_BATCH_SIZE or less.
+ */
+ n = (uint16_t)RTE_MIN(nb_pkts, ETH_AF_XDP_TX_BATCH_SIZE);
+ ret = af_xdp_tx_cp(queue, &bufs[nb_tx], n);
+ nb_tx = (uint16_t)(nb_tx + ret);
+ nb_pkts = (uint16_t)(nb_pkts - ret);
+ if (ret < n)
+ break;
+ }
+
+ return nb_tx;
+}
#endif
static uint16_t
#if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
return af_xdp_tx_zc(queue, bufs, nb_pkts);
#else
- return af_xdp_tx_cp(queue, bufs, nb_pkts);
+ return af_xdp_tx_cp_batch(queue, bufs, nb_pkts);
#endif
}
dev_info->max_mtu = ETH_AF_XDP_FRAME_SIZE - XDP_PACKET_HEADROOM;
#endif
+ dev_info->default_rxportconf.burst_size = ETH_AF_XDP_DFLT_BUSY_BUDGET;
+ dev_info->default_txportconf.burst_size = ETH_AF_XDP_DFLT_BUSY_BUDGET;
dev_info->default_rxportconf.nb_queues = 1;
dev_info->default_txportconf.nb_queues = 1;
dev_info->default_rxportconf.ring_size = ETH_AF_XDP_DFLT_NUM_DESCS;
#endif
rte_free(umem);
- umem = NULL;
}
static int
}
#if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
-static inline uint64_t get_base_addr(struct rte_mempool *mp, uint64_t *align)
+static inline uintptr_t get_base_addr(struct rte_mempool *mp, uint64_t *align)
{
struct rte_mempool_memhdr *memhdr;
- uint64_t memhdr_addr, aligned_addr;
+ uintptr_t memhdr_addr, aligned_addr;
memhdr = STAILQ_FIRST(&mp->mem_list);
- memhdr_addr = (uint64_t)memhdr->addr;
+ memhdr_addr = (uintptr_t)memhdr->addr;
aligned_addr = memhdr_addr & ~(getpagesize() - 1);
*align = memhdr_addr - aligned_addr;
umem->mb_pool = mb_pool;
base_addr = (void *)get_base_addr(mb_pool, &align);
- umem_size = mb_pool->populated_size * usr_config.frame_size +
+ umem_size = (uint64_t)mb_pool->populated_size *
+ (uint64_t)usr_config.frame_size +
align;
ret = xsk_umem__create(&umem->umem, base_addr, umem_size,
return 0;
}
+/* Detect support for busy polling through setsockopt(). */
+static int
+configure_preferred_busy_poll(struct pkt_rx_queue *rxq)
+{
+ int sock_opt = 1;
+ int fd = xsk_socket__fd(rxq->xsk);
+ int ret = 0;
+
+ ret = setsockopt(fd, SOL_SOCKET, SO_PREFER_BUSY_POLL,
+ (void *)&sock_opt, sizeof(sock_opt));
+ if (ret < 0) {
+ AF_XDP_LOG(DEBUG, "Failed to set SO_PREFER_BUSY_POLL\n");
+ goto err_prefer;
+ }
+
+ sock_opt = ETH_AF_XDP_DFLT_BUSY_TIMEOUT;
+ ret = setsockopt(fd, SOL_SOCKET, SO_BUSY_POLL, (void *)&sock_opt,
+ sizeof(sock_opt));
+ if (ret < 0) {
+ AF_XDP_LOG(DEBUG, "Failed to set SO_BUSY_POLL\n");
+ goto err_timeout;
+ }
+
+ sock_opt = rxq->busy_budget;
+ ret = setsockopt(fd, SOL_SOCKET, SO_BUSY_POLL_BUDGET,
+ (void *)&sock_opt, sizeof(sock_opt));
+ if (ret < 0) {
+ AF_XDP_LOG(DEBUG, "Failed to set SO_BUSY_POLL_BUDGET\n");
+ } else {
+ AF_XDP_LOG(INFO, "Busy polling budget set to: %u\n",
+ rxq->busy_budget);
+ return 0;
+ }
+
+ /* setsockopt failure - attempt to restore xsk to default state and
+ * proceed without busy polling support.
+ */
+ sock_opt = 0;
+ ret = setsockopt(fd, SOL_SOCKET, SO_BUSY_POLL, (void *)&sock_opt,
+ sizeof(sock_opt));
+ if (ret < 0) {
+ AF_XDP_LOG(ERR, "Failed to unset SO_BUSY_POLL\n");
+ return -1;
+ }
+
+err_timeout:
+ sock_opt = 0;
+ ret = setsockopt(fd, SOL_SOCKET, SO_PREFER_BUSY_POLL,
+ (void *)&sock_opt, sizeof(sock_opt));
+ if (ret < 0) {
+ AF_XDP_LOG(ERR, "Failed to unset SO_PREFER_BUSY_POLL\n");
+ return -1;
+ }
+
+err_prefer:
+ rxq->busy_budget = 0;
+ return 0;
+}
+
static int
xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
int ring_size)
}
#if defined(XDP_UMEM_UNALIGNED_CHUNK_FLAG)
- if (rte_pktmbuf_alloc_bulk(rxq->umem->mb_pool, fq_bufs, reserve_size)) {
+ ret = rte_pktmbuf_alloc_bulk(rxq->umem->mb_pool, fq_bufs, reserve_size);
+ if (ret) {
AF_XDP_LOG(DEBUG, "Failed to get enough buffers for fq.\n");
goto err;
}
#endif
+
+ if (rxq->busy_budget) {
+ ret = configure_preferred_busy_poll(rxq);
+ if (ret) {
+ AF_XDP_LOG(ERR, "Failed configure busy polling.\n");
+ goto err;
+ }
+ }
+
ret = reserve_fill_queue(rxq->umem, reserve_size, fq_bufs, &rxq->fq);
if (ret) {
xsk_socket__delete(rxq->xsk);
goto err;
}
+ if (!rxq->busy_budget)
+ AF_XDP_LOG(DEBUG, "Preferred busy polling not enabled\n");
+
rxq->fds[0].fd = xsk_socket__fd(rxq->xsk);
rxq->fds[0].events = POLLIN;
.stats_reset = eth_stats_reset,
};
+/** parse busy_budget argument */
+static int
+parse_budget_arg(const char *key __rte_unused,
+ const char *value, void *extra_args)
+{
+ int *i = (int *)extra_args;
+ char *end;
+
+ *i = strtol(value, &end, 10);
+ if (*i < 0 || *i > UINT16_MAX) {
+ AF_XDP_LOG(ERR, "Invalid busy_budget %i, must be >= 0 and <= %u\n",
+ *i, UINT16_MAX);
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
/** parse integer from integer argument */
static int
parse_integer_arg(const char *key __rte_unused,
static int
parse_parameters(struct rte_kvargs *kvlist, char *if_name, int *start_queue,
- int *queue_cnt, int *shared_umem, char *prog_path)
+ int *queue_cnt, int *shared_umem, char *prog_path,
+ int *busy_budget)
{
int ret;
if (ret < 0)
goto free_kvlist;
+ ret = rte_kvargs_process(kvlist, ETH_AF_XDP_BUDGET_ARG,
+ &parse_budget_arg, busy_budget);
+ if (ret < 0)
+ goto free_kvlist;
+
free_kvlist:
rte_kvargs_free(kvlist);
return ret;
static struct rte_eth_dev *
init_internals(struct rte_vdev_device *dev, const char *if_name,
int start_queue_idx, int queue_cnt, int shared_umem,
- const char *prog_path)
+ const char *prog_path, int busy_budget)
{
const char *name = rte_vdev_device_name(dev);
const unsigned int numa_node = dev->device.numa_node;
internals->rx_queues[i].pair = &internals->tx_queues[i];
internals->rx_queues[i].xsk_queue_idx = start_queue_idx + i;
internals->tx_queues[i].xsk_queue_idx = start_queue_idx + i;
+ internals->rx_queues[i].busy_budget = busy_budget;
}
ret = get_iface_info(if_name, &internals->eth_addr,
eth_dev->data->dev_private = internals;
eth_dev->data->dev_link = pmd_link;
eth_dev->data->mac_addrs = &internals->eth_addr;
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
eth_dev->dev_ops = &ops;
eth_dev->rx_pkt_burst = eth_af_xdp_rx;
eth_dev->tx_pkt_burst = eth_af_xdp_tx;
int xsk_queue_cnt = ETH_AF_XDP_DFLT_QUEUE_COUNT;
int shared_umem = 0;
char prog_path[PATH_MAX] = {'\0'};
+ int busy_budget = -1;
struct rte_eth_dev *eth_dev = NULL;
const char *name;
dev->device.numa_node = rte_socket_id();
if (parse_parameters(kvlist, if_name, &xsk_start_queue_idx,
- &xsk_queue_cnt, &shared_umem, prog_path) < 0) {
+ &xsk_queue_cnt, &shared_umem, prog_path,
+ &busy_budget) < 0) {
AF_XDP_LOG(ERR, "Invalid kvargs value\n");
return -EINVAL;
}
return -EINVAL;
}
+ busy_budget = busy_budget == -1 ? ETH_AF_XDP_DFLT_BUSY_BUDGET :
+ busy_budget;
+
eth_dev = init_internals(dev, if_name, xsk_start_queue_idx,
- xsk_queue_cnt, shared_umem, prog_path);
+ xsk_queue_cnt, shared_umem, prog_path,
+ busy_budget);
if (eth_dev == NULL) {
AF_XDP_LOG(ERR, "Failed to init internals\n");
return -1;
"start_queue=<int> "
"queue_count=<int> "
"shared_umem=<int> "
- "xdp_prog=<string> ");
+ "xdp_prog=<string> "
+ "busy_budget=<int>");
git.droids-corp.org / dpdk.git / blobdiff