* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
+#include <rte_atomic.h>
+#include <rte_branch_prediction.h>
+#include <rte_common.h>
#include <rte_mbuf.h>
#include <rte_ethdev.h>
#include <rte_malloc.h>
#include <rte_vdev.h>
#include <rte_kvargs.h>
+#include <rte_net.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
+#include <sys/utsname.h>
#include <sys/mman.h>
+#include <errno.h>
+#include <signal.h>
+#include <stdint.h>
+#include <sys/uio.h>
#include <unistd.h>
-#include <poll.h>
#include <arpa/inet.h>
-#include <linux/if.h>
+#include <net/if.h>
#include <linux/if_tun.h>
#include <linux/if_ether.h>
+#include <linux/version.h>
#include <fcntl.h>
+#include <rte_eth_tap.h>
+#include <tap_flow.h>
+#include <tap_netlink.h>
+#include <tap_tcmsgs.h>
+
/* Linux based path to the TUN device */
#define TUN_TAP_DEV_PATH "/dev/net/tun"
#define DEFAULT_TAP_NAME "dtap"
#define ETH_TAP_IFACE_ARG "iface"
#define ETH_TAP_SPEED_ARG "speed"
+#define ETH_TAP_REMOTE_ARG "remote"
#ifdef IFF_MULTI_QUEUE
#define RTE_PMD_TAP_MAX_QUEUES 16
#define RTE_PMD_TAP_MAX_QUEUES 1
#endif
+#define FLOWER_KERNEL_VERSION KERNEL_VERSION(4, 2, 0)
+#define FLOWER_VLAN_KERNEL_VERSION KERNEL_VERSION(4, 9, 0)
+
static struct rte_vdev_driver pmd_tap_drv;
static const char *valid_arguments[] = {
ETH_TAP_IFACE_ARG,
ETH_TAP_SPEED_ARG,
+ ETH_TAP_REMOTE_ARG,
NULL
};
static int tap_unit;
+static volatile uint32_t tap_trigger; /* Rx trigger */
+
static struct rte_eth_link pmd_link = {
.link_speed = ETH_SPEED_NUM_10G,
.link_duplex = ETH_LINK_FULL_DUPLEX,
.link_autoneg = ETH_LINK_SPEED_AUTONEG
};
-struct pkt_stats {
- uint64_t opackets; /* Number of output packets */
- uint64_t ipackets; /* Number of input packets */
- uint64_t obytes; /* Number of bytes on output */
- uint64_t ibytes; /* Number of bytes on input */
- uint64_t errs; /* Number of error packets */
-};
-
-struct rx_queue {
- struct rte_mempool *mp; /* Mempool for RX packets */
- uint16_t in_port; /* Port ID */
- int fd;
-
- struct pkt_stats stats; /* Stats for this RX queue */
-};
-
-struct tx_queue {
- int fd;
- struct pkt_stats stats; /* Stats for this TX queue */
-};
-
-struct pmd_internals {
- char name[RTE_ETH_NAME_MAX_LEN]; /* Internal Tap device name */
- uint16_t nb_queues; /* Number of queues supported */
- struct ether_addr eth_addr; /* Mac address of the device port */
+static void
+tap_trigger_cb(int sig __rte_unused)
+{
+ /* Valid trigger values are nonzero */
+ tap_trigger = (tap_trigger + 1) | 0x80000000;
+}
- int if_index; /* IF_INDEX for the port */
+static int
+tap_ioctl(struct pmd_internals *pmd, unsigned long request,
+ struct ifreq *ifr, int set);
- struct rx_queue rxq[RTE_PMD_TAP_MAX_QUEUES]; /* List of RX queues */
- struct tx_queue txq[RTE_PMD_TAP_MAX_QUEUES]; /* List of TX queues */
-};
+static int tap_intr_handle_set(struct rte_eth_dev *dev, int set);
/* Tun/Tap allocation routine
*
memset(&ifr, 0, sizeof(struct ifreq));
- ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
+ /*
+ * Do not set IFF_NO_PI as packet information header will be needed
+ * to check if a received packet has been truncated.
+ */
+ ifr.ifr_flags = IFF_TAP;
snprintf(ifr.ifr_name, IFNAMSIZ, "%s", pmd->name);
RTE_LOG(DEBUG, PMD, "ifr_name '%s'\n", ifr.ifr_name);
goto error;
}
+ /* Set up trigger to optimize empty Rx bursts */
+ errno = 0;
+ do {
+ struct sigaction sa;
+ int flags = fcntl(fd, F_GETFL);
+
+ if (flags == -1 || sigaction(SIGIO, NULL, &sa) == -1)
+ break;
+ if (sa.sa_handler != tap_trigger_cb) {
+ /*
+ * Make sure SIGIO is not already taken. This is done
+ * as late as possible to leave the application a
+ * chance to set up its own signal handler first.
+ */
+ if (sa.sa_handler != SIG_IGN &&
+ sa.sa_handler != SIG_DFL) {
+ errno = EBUSY;
+ break;
+ }
+ sa = (struct sigaction){
+ .sa_flags = SA_RESTART,
+ .sa_handler = tap_trigger_cb,
+ };
+ if (sigaction(SIGIO, &sa, NULL) == -1)
+ break;
+ }
+ /* Enable SIGIO on file descriptor */
+ fcntl(fd, F_SETFL, flags | O_ASYNC);
+ fcntl(fd, F_SETOWN, getpid());
+ } while (0);
+ if (errno) {
+ /* Disable trigger globally in case of error */
+ tap_trigger = 0;
+ RTE_LOG(WARNING, PMD, "Rx trigger disabled: %s\n",
+ strerror(errno));
+ }
+
if (qid == 0) {
- if (ioctl(fd, SIOCGIFHWADDR, &ifr) == -1) {
- RTE_LOG(ERR, PMD, "ioctl failed (SIOCGIFHWADDR) (%s)\n",
- ifr.ifr_name);
+ struct ifreq ifr;
+
+ if (tap_ioctl(pmd, SIOCGIFHWADDR, &ifr, 0) < 0)
goto error;
+ rte_memcpy(&pmd->eth_addr, ifr.ifr_hwaddr.sa_data,
+ ETHER_ADDR_LEN);
+
+ pmd->if_index = if_nametoindex(pmd->name);
+ if (!pmd->if_index) {
+ RTE_LOG(ERR, PMD,
+ "Could not find ifindex for %s: rte_flow won't be usable.\n",
+ pmd->name);
+ return fd;
+ }
+ if (!pmd->flower_support)
+ return fd;
+ if (qdisc_create_multiq(pmd->nlsk_fd, pmd->if_index) < 0) {
+ RTE_LOG(ERR, PMD,
+ "Could not create multiq qdisc for %s: rte_flow won't be usable.\n",
+ pmd->name);
+ return fd;
+ }
+ if (qdisc_create_ingress(pmd->nlsk_fd, pmd->if_index) < 0) {
+ RTE_LOG(ERR, PMD,
+ "Could not create multiq qdisc for %s: rte_flow won't be usable.\n",
+ pmd->name);
+ return fd;
+ }
+ if (pmd->remote_if_index) {
+ /*
+ * Flush usually returns negative value because it tries
+ * to delete every QDISC (and on a running device, one
+ * QDISC at least is needed). Ignore negative return
+ * value.
+ */
+ qdisc_flush(pmd->nlsk_fd, pmd->remote_if_index);
+ if (qdisc_create_ingress(pmd->nlsk_fd,
+ pmd->remote_if_index) < 0)
+ goto remote_fail;
+ LIST_INIT(&pmd->implicit_flows);
+ if (tap_flow_implicit_create(
+ pmd, TAP_REMOTE_LOCAL_MAC) < 0)
+ goto remote_fail;
+ if (tap_flow_implicit_create(
+ pmd, TAP_REMOTE_BROADCAST) < 0)
+ goto remote_fail;
+ if (tap_flow_implicit_create(
+ pmd, TAP_REMOTE_BROADCASTV6) < 0)
+ goto remote_fail;
+ if (tap_flow_implicit_create(
+ pmd, TAP_REMOTE_TX) < 0)
+ goto remote_fail;
}
-
- rte_memcpy(&pmd->eth_addr, ifr.ifr_hwaddr.sa_data, ETH_ALEN);
}
return fd;
+remote_fail:
+ RTE_LOG(ERR, PMD,
+ "Could not set up remote flow rules for %s: remote disabled.\n",
+ pmd->name);
+ pmd->remote_if_index = 0;
+ tap_flow_implicit_flush(pmd, NULL);
+ return fd;
+
error:
if (fd > 0)
close(fd);
static uint16_t
pmd_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
{
- int len;
- struct rte_mbuf *mbuf;
struct rx_queue *rxq = queue;
uint16_t num_rx;
unsigned long num_rx_bytes = 0;
+ uint32_t trigger = tap_trigger;
+ if (trigger == rxq->trigger_seen)
+ return 0;
+ if (trigger)
+ rxq->trigger_seen = trigger;
+ rte_compiler_barrier();
for (num_rx = 0; num_rx < nb_pkts; ) {
- /* allocate the next mbuf */
- mbuf = rte_pktmbuf_alloc(rxq->mp);
- if (unlikely(!mbuf)) {
- RTE_LOG(WARNING, PMD, "TAP unable to allocate mbuf\n");
+ struct rte_mbuf *mbuf = rxq->pool;
+ struct rte_mbuf *seg = NULL;
+ struct rte_mbuf *new_tail = NULL;
+ uint16_t data_off = rte_pktmbuf_headroom(mbuf);
+ int len;
+
+ len = readv(rxq->fd, *rxq->iovecs,
+ 1 + (rxq->rxmode->enable_scatter ?
+ rxq->nb_rx_desc : 1));
+ if (len < (int)sizeof(struct tun_pi))
break;
- }
- len = read(rxq->fd, rte_pktmbuf_mtod(mbuf, char *),
- rte_pktmbuf_tailroom(mbuf));
- if (len <= 0) {
- rte_pktmbuf_free(mbuf);
- break;
+ /* Packet couldn't fit in the provided mbuf */
+ if (unlikely(rxq->pi.flags & TUN_PKT_STRIP)) {
+ rxq->stats.ierrors++;
+ continue;
}
- mbuf->data_len = len;
+ len -= sizeof(struct tun_pi);
+
mbuf->pkt_len = len;
mbuf->port = rxq->in_port;
+ while (1) {
+ struct rte_mbuf *buf = rte_pktmbuf_alloc(rxq->mp);
+
+ if (unlikely(!buf)) {
+ rxq->stats.rx_nombuf++;
+ /* No new buf has been allocated: do nothing */
+ if (!new_tail || !seg)
+ goto end;
+
+ seg->next = NULL;
+ rte_pktmbuf_free(mbuf);
+
+ goto end;
+ }
+ seg = seg ? seg->next : mbuf;
+ if (rxq->pool == mbuf)
+ rxq->pool = buf;
+ if (new_tail)
+ new_tail->next = buf;
+ new_tail = buf;
+ new_tail->next = seg->next;
+
+ /* iovecs[0] is reserved for packet info (pi) */
+ (*rxq->iovecs)[mbuf->nb_segs].iov_len =
+ buf->buf_len - data_off;
+ (*rxq->iovecs)[mbuf->nb_segs].iov_base =
+ (char *)buf->buf_addr + data_off;
+
+ seg->data_len = RTE_MIN(seg->buf_len - data_off, len);
+ seg->data_off = data_off;
+
+ len -= seg->data_len;
+ if (len <= 0)
+ break;
+ mbuf->nb_segs++;
+ /* First segment has headroom, not the others */
+ data_off = 0;
+ }
+ seg->next = NULL;
+ mbuf->packet_type = rte_net_get_ptype(mbuf, NULL,
+ RTE_PTYPE_ALL_MASK);
/* account for the receive frame */
bufs[num_rx++] = mbuf;
num_rx_bytes += mbuf->pkt_len;
}
+end:
rxq->stats.ipackets += num_rx;
rxq->stats.ibytes += num_rx_bytes;
static uint16_t
pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
{
- struct rte_mbuf *mbuf;
struct tx_queue *txq = queue;
- struct pollfd pfd;
uint16_t num_tx = 0;
unsigned long num_tx_bytes = 0;
- int i, n;
+ uint32_t max_size;
+ int i;
if (unlikely(nb_pkts == 0))
return 0;
- pfd.events = POLLOUT;
- pfd.fd = txq->fd;
+ max_size = *txq->mtu + (ETHER_HDR_LEN + ETHER_CRC_LEN + 4);
for (i = 0; i < nb_pkts; i++) {
- n = poll(&pfd, 1, 0);
+ struct rte_mbuf *mbuf = bufs[num_tx];
+ struct iovec iovecs[mbuf->nb_segs + 1];
+ struct tun_pi pi = { .flags = 0 };
+ struct rte_mbuf *seg = mbuf;
+ int n;
+ int j;
+
+ /* stats.errs will be incremented */
+ if (rte_pktmbuf_pkt_len(mbuf) > max_size)
+ break;
+ iovecs[0].iov_base = π
+ iovecs[0].iov_len = sizeof(pi);
+ for (j = 1; j <= mbuf->nb_segs; j++) {
+ iovecs[j].iov_len = rte_pktmbuf_data_len(seg);
+ iovecs[j].iov_base =
+ rte_pktmbuf_mtod(seg, void *);
+ seg = seg->next;
+ }
+ /* copy the tx frame data */
+ n = writev(txq->fd, iovecs, mbuf->nb_segs + 1);
if (n <= 0)
break;
- if (pfd.revents & POLLOUT) {
- /* copy the tx frame data */
- mbuf = bufs[num_tx];
- n = write(pfd.fd, rte_pktmbuf_mtod(mbuf, void*),
- rte_pktmbuf_pkt_len(mbuf));
- if (n <= 0)
- break;
-
- num_tx++;
- num_tx_bytes += mbuf->pkt_len;
- rte_pktmbuf_free(mbuf);
- }
+ num_tx++;
+ num_tx_bytes += mbuf->pkt_len;
+ rte_pktmbuf_free(mbuf);
}
txq->stats.opackets += num_tx;
}
static int
-tap_link_set_flags(struct pmd_internals *pmd, short flags, int add)
+tap_ioctl(struct pmd_internals *pmd, unsigned long request,
+ struct ifreq *ifr, int set)
{
- struct ifreq ifr;
- int err, s;
+ short req_flags = ifr->ifr_flags;
+ int remote = !!pmd->remote_if_index;
/*
- * An AF_INET/DGRAM socket is needed for
- * SIOCGIFFLAGS/SIOCSIFFLAGS, using fd won't work.
+ * If there is a remote netdevice, apply ioctl on it, then apply it on
+ * the tap netdevice.
*/
- s = socket(AF_INET, SOCK_DGRAM, 0);
- if (s < 0) {
- RTE_LOG(ERR, PMD,
- "Unable to get a socket to set flags: %s\n",
- strerror(errno));
- return -1;
- }
- memset(&ifr, 0, sizeof(ifr));
- snprintf(ifr.ifr_name, IFNAMSIZ, "%s", pmd->name);
- err = ioctl(s, SIOCGIFFLAGS, &ifr);
- if (err < 0) {
- RTE_LOG(WARNING, PMD, "Unable to get %s device flags: %s\n",
- pmd->name, strerror(errno));
- close(s);
- return -1;
+ if (request == SIOCGIFFLAGS && !set) {
+ /*
+ * Special case for getting flags. If set is given,
+ * then return the flags from the remote netdevice only.
+ * Otherwise return the flags from the tap netdevice.
+ */
+ remote = 0;
}
- if (add)
- ifr.ifr_flags |= flags;
+apply:
+ if (remote)
+ snprintf(ifr->ifr_name, IFNAMSIZ, "%s", pmd->remote_iface);
else
- ifr.ifr_flags &= ~flags;
- err = ioctl(s, SIOCSIFFLAGS, &ifr);
- if (err < 0) {
- RTE_LOG(WARNING, PMD, "Unable to %s flags 0x%x: %s\n",
- add ? "set" : "unset", flags, strerror(errno));
- close(s);
- return -1;
+ snprintf(ifr->ifr_name, IFNAMSIZ, "%s", pmd->name);
+ switch (request) {
+ case SIOCSIFFLAGS:
+ /* fetch current flags to leave other flags untouched */
+ if (ioctl(pmd->ioctl_sock, SIOCGIFFLAGS, ifr) < 0)
+ goto error;
+ if (set)
+ ifr->ifr_flags |= req_flags;
+ else
+ ifr->ifr_flags &= ~req_flags;
+ break;
+ case SIOCGIFFLAGS:
+ if (remote && set)
+ remote = 0; /* don't loop */
+ break;
+ case SIOCGIFHWADDR:
+ /* Set remote MAC on the tap netdevice */
+ if (!remote && pmd->remote_if_index) {
+ request = SIOCSIFHWADDR;
+ goto apply;
+ }
+ break;
+ case SIOCSIFHWADDR:
+ case SIOCSIFMTU:
+ break;
+ default:
+ RTE_LOG(WARNING, PMD, "%s: ioctl() called with wrong arg\n",
+ pmd->name);
+ return -EINVAL;
}
- close(s);
-
+ if (ioctl(pmd->ioctl_sock, request, ifr) < 0)
+ goto error;
+ if (remote--)
+ goto apply;
return 0;
+
+error:
+ RTE_LOG(ERR, PMD, "%s: ioctl(%lu) failed with error: %s\n",
+ ifr->ifr_name, request, strerror(errno));
+ return -errno;
}
static int
tap_link_set_down(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_UP };
dev->data->dev_link.link_status = ETH_LINK_DOWN;
- return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 0);
+ return tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
}
static int
tap_link_set_up(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_UP };
dev->data->dev_link.link_status = ETH_LINK_UP;
- return tap_link_set_flags(pmd, IFF_UP | IFF_NOARP, 1);
+ return tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
}
static int
tap_dev_start(struct rte_eth_dev *dev)
{
+ int err;
+
+ err = tap_intr_handle_set(dev, 1);
+ if (err)
+ return err;
return tap_link_set_up(dev);
}
static void
tap_dev_stop(struct rte_eth_dev *dev)
{
+ tap_intr_handle_set(dev, 0);
tap_link_set_down(dev);
}
return 0;
}
+static uint32_t
+tap_dev_speed_capa(void)
+{
+ uint32_t speed = pmd_link.link_speed;
+ uint32_t capa = 0;
+
+ if (speed >= ETH_SPEED_NUM_10M)
+ capa |= ETH_LINK_SPEED_10M;
+ if (speed >= ETH_SPEED_NUM_100M)
+ capa |= ETH_LINK_SPEED_100M;
+ if (speed >= ETH_SPEED_NUM_1G)
+ capa |= ETH_LINK_SPEED_1G;
+ if (speed >= ETH_SPEED_NUM_5G)
+ capa |= ETH_LINK_SPEED_2_5G;
+ if (speed >= ETH_SPEED_NUM_5G)
+ capa |= ETH_LINK_SPEED_5G;
+ if (speed >= ETH_SPEED_NUM_10G)
+ capa |= ETH_LINK_SPEED_10G;
+ if (speed >= ETH_SPEED_NUM_20G)
+ capa |= ETH_LINK_SPEED_20G;
+ if (speed >= ETH_SPEED_NUM_25G)
+ capa |= ETH_LINK_SPEED_25G;
+ if (speed >= ETH_SPEED_NUM_40G)
+ capa |= ETH_LINK_SPEED_40G;
+ if (speed >= ETH_SPEED_NUM_50G)
+ capa |= ETH_LINK_SPEED_50G;
+ if (speed >= ETH_SPEED_NUM_56G)
+ capa |= ETH_LINK_SPEED_56G;
+ if (speed >= ETH_SPEED_NUM_100G)
+ capa |= ETH_LINK_SPEED_100G;
+
+ return capa;
+}
+
static void
tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
{
dev_info->max_tx_queues = internals->nb_queues;
dev_info->min_rx_bufsize = 0;
dev_info->pci_dev = NULL;
+ dev_info->speed_capa = tap_dev_speed_capa();
}
static void
unsigned int i, imax;
unsigned long rx_total = 0, tx_total = 0, tx_err_total = 0;
unsigned long rx_bytes_total = 0, tx_bytes_total = 0;
+ unsigned long rx_nombuf = 0, ierrors = 0;
const struct pmd_internals *pmd = dev->data->dev_private;
imax = (pmd->nb_queues < RTE_ETHDEV_QUEUE_STAT_CNTRS) ?
tap_stats->q_ibytes[i] = pmd->rxq[i].stats.ibytes;
rx_total += tap_stats->q_ipackets[i];
rx_bytes_total += tap_stats->q_ibytes[i];
+ rx_nombuf += pmd->rxq[i].stats.rx_nombuf;
+ ierrors += pmd->rxq[i].stats.ierrors;
tap_stats->q_opackets[i] = pmd->txq[i].stats.opackets;
tap_stats->q_errors[i] = pmd->txq[i].stats.errs;
tap_stats->ipackets = rx_total;
tap_stats->ibytes = rx_bytes_total;
+ tap_stats->ierrors = ierrors;
+ tap_stats->rx_nombuf = rx_nombuf;
tap_stats->opackets = tx_total;
tap_stats->oerrors = tx_err_total;
tap_stats->obytes = tx_bytes_total;
for (i = 0; i < pmd->nb_queues; i++) {
pmd->rxq[i].stats.ipackets = 0;
pmd->rxq[i].stats.ibytes = 0;
+ pmd->rxq[i].stats.ierrors = 0;
+ pmd->rxq[i].stats.rx_nombuf = 0;
pmd->txq[i].stats.opackets = 0;
pmd->txq[i].stats.errs = 0;
struct pmd_internals *internals = dev->data->dev_private;
tap_link_set_down(dev);
+ tap_flow_flush(dev, NULL);
+ tap_flow_implicit_flush(internals, NULL);
for (i = 0; i < internals->nb_queues; i++) {
if (internals->rxq[i].fd != -1)
if (rxq && (rxq->fd > 0)) {
close(rxq->fd);
rxq->fd = -1;
+ rte_pktmbuf_free(rxq->pool);
+ rte_free(rxq->iovecs);
+ rxq->pool = NULL;
+ rxq->iovecs = NULL;
}
}
}
static int
-tap_link_update(struct rte_eth_dev *dev __rte_unused,
- int wait_to_complete __rte_unused)
+tap_link_update(struct rte_eth_dev *dev, int wait_to_complete __rte_unused)
{
+ struct rte_eth_link *dev_link = &dev->data->dev_link;
+ struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = 0 };
+
+ if (pmd->remote_if_index) {
+ tap_ioctl(pmd, SIOCGIFFLAGS, &ifr, 1);
+ if (!(ifr.ifr_flags & IFF_UP) ||
+ !(ifr.ifr_flags & IFF_RUNNING)) {
+ dev_link->link_status = ETH_LINK_DOWN;
+ return 0;
+ }
+ }
+ tap_ioctl(pmd, SIOCGIFFLAGS, &ifr, 0);
+ dev_link->link_status =
+ ((ifr.ifr_flags & IFF_UP) && (ifr.ifr_flags & IFF_RUNNING) ?
+ ETH_LINK_UP :
+ ETH_LINK_DOWN);
return 0;
}
tap_promisc_enable(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_PROMISC };
dev->data->promiscuous = 1;
- tap_link_set_flags(pmd, IFF_PROMISC, 1);
+ tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
+ if (pmd->remote_if_index)
+ tap_flow_implicit_create(pmd, TAP_REMOTE_PROMISC);
}
static void
tap_promisc_disable(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_PROMISC };
dev->data->promiscuous = 0;
- tap_link_set_flags(pmd, IFF_PROMISC, 0);
+ tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
+ if (pmd->remote_if_index)
+ tap_flow_implicit_destroy(pmd, TAP_REMOTE_PROMISC);
}
static void
tap_allmulti_enable(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_ALLMULTI };
dev->data->all_multicast = 1;
- tap_link_set_flags(pmd, IFF_ALLMULTI, 1);
+ tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 1);
+ if (pmd->remote_if_index)
+ tap_flow_implicit_create(pmd, TAP_REMOTE_ALLMULTI);
}
static void
tap_allmulti_disable(struct rte_eth_dev *dev)
{
struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_flags = IFF_ALLMULTI };
dev->data->all_multicast = 0;
- tap_link_set_flags(pmd, IFF_ALLMULTI, 0);
+ tap_ioctl(pmd, SIOCSIFFLAGS, &ifr, 0);
+ if (pmd->remote_if_index)
+ tap_flow_implicit_destroy(pmd, TAP_REMOTE_ALLMULTI);
+}
+
+
+static void
+tap_mac_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
+{
+ struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr;
+
+ if (is_zero_ether_addr(mac_addr)) {
+ RTE_LOG(ERR, PMD, "%s: can't set an empty MAC address\n",
+ dev->data->name);
+ return;
+ }
+
+ ifr.ifr_hwaddr.sa_family = AF_LOCAL;
+ rte_memcpy(ifr.ifr_hwaddr.sa_data, mac_addr, ETHER_ADDR_LEN);
+ if (tap_ioctl(pmd, SIOCSIFHWADDR, &ifr, 1) < 0)
+ return;
+ rte_memcpy(&pmd->eth_addr, mac_addr, ETHER_ADDR_LEN);
}
static int
pmd->name, qid);
return -1;
}
+ if (qid == 0) {
+ struct ifreq ifr;
+
+ ifr.ifr_mtu = dev->data->mtu;
+ if (tap_ioctl(pmd, SIOCSIFMTU, &ifr, 1) < 0) {
+ close(fd);
+ return -1;
+ }
+ }
}
}
rx->fd = fd;
tx->fd = fd;
+ tx->mtu = &dev->data->mtu;
+ rx->rxmode = &dev->data->dev_conf.rxmode;
return fd;
}
static int
tap_rx_queue_setup(struct rte_eth_dev *dev,
uint16_t rx_queue_id,
- uint16_t nb_rx_desc __rte_unused,
- unsigned int socket_id __rte_unused,
+ uint16_t nb_rx_desc,
+ unsigned int socket_id,
const struct rte_eth_rxconf *rx_conf __rte_unused,
struct rte_mempool *mp)
{
struct pmd_internals *internals = dev->data->dev_private;
+ struct rx_queue *rxq = &internals->rxq[rx_queue_id];
+ struct rte_mbuf **tmp = &rxq->pool;
+ struct iovec (*iovecs)[nb_rx_desc + 1];
+ int data_off = RTE_PKTMBUF_HEADROOM;
uint16_t buf_size;
+ int ret = 0;
int fd;
+ int i;
if ((rx_queue_id >= internals->nb_queues) || !mp) {
RTE_LOG(WARNING, PMD,
return -1;
}
- internals->rxq[rx_queue_id].mp = mp;
- internals->rxq[rx_queue_id].in_port = dev->data->port_id;
+ rxq->mp = mp;
+ rxq->trigger_seen = 1; /* force initial burst */
+ rxq->in_port = dev->data->port_id;
+ rxq->nb_rx_desc = nb_rx_desc;
+ iovecs = rte_zmalloc_socket(dev->data->name, sizeof(*iovecs), 0,
+ socket_id);
+ if (!iovecs) {
+ RTE_LOG(WARNING, PMD,
+ "%s: Couldn't allocate %d RX descriptors\n",
+ dev->data->name, nb_rx_desc);
+ return -ENOMEM;
+ }
+ rxq->iovecs = iovecs;
/* Now get the space available for data in the mbuf */
buf_size = (uint16_t)(rte_pktmbuf_data_room_size(mp) -
RTE_LOG(WARNING, PMD,
"%s: %d bytes will not fit in mbuf (%d bytes)\n",
dev->data->name, ETH_FRAME_LEN, buf_size);
- return -ENOMEM;
+ ret = -ENOMEM;
+ goto error;
}
fd = rx_setup_queue(dev, internals, rx_queue_id);
- if (fd == -1)
- return -1;
+ if (fd == -1) {
+ ret = fd;
+ goto error;
+ }
+
+ (*rxq->iovecs)[0].iov_len = sizeof(struct tun_pi);
+ (*rxq->iovecs)[0].iov_base = &rxq->pi;
+
+ for (i = 1; i <= nb_rx_desc; i++) {
+ *tmp = rte_pktmbuf_alloc(rxq->mp);
+ if (!*tmp) {
+ RTE_LOG(WARNING, PMD,
+ "%s: couldn't allocate memory for queue %d\n",
+ dev->data->name, rx_queue_id);
+ ret = -ENOMEM;
+ goto error;
+ }
+ (*rxq->iovecs)[i].iov_len = (*tmp)->buf_len - data_off;
+ (*rxq->iovecs)[i].iov_base =
+ (char *)(*tmp)->buf_addr + data_off;
+ data_off = 0;
+ tmp = &(*tmp)->next;
+ }
RTE_LOG(DEBUG, PMD, " RX TAP device name %s, qid %d on fd %d\n",
internals->name, rx_queue_id, internals->rxq[rx_queue_id].fd);
return 0;
+
+error:
+ rte_pktmbuf_free(rxq->pool);
+ rxq->pool = NULL;
+ rte_free(rxq->iovecs);
+ rxq->iovecs = NULL;
+ return ret;
}
static int
return 0;
}
+static int
+tap_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
+{
+ struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifreq ifr = { .ifr_mtu = mtu };
+ int err = 0;
+
+ err = tap_ioctl(pmd, SIOCSIFMTU, &ifr, 1);
+ if (!err)
+ dev->data->mtu = mtu;
+
+ return err;
+}
+
+static int
+tap_set_mc_addr_list(struct rte_eth_dev *dev __rte_unused,
+ struct ether_addr *mc_addr_set __rte_unused,
+ uint32_t nb_mc_addr __rte_unused)
+{
+ /*
+ * Nothing to do actually: the tap has no filtering whatsoever, every
+ * packet is received.
+ */
+ return 0;
+}
+
+static int
+tap_nl_msg_handler(struct nlmsghdr *nh, void *arg)
+{
+ struct rte_eth_dev *dev = arg;
+ struct pmd_internals *pmd = dev->data->dev_private;
+ struct ifinfomsg *info = NLMSG_DATA(nh);
+
+ if (nh->nlmsg_type != RTM_NEWLINK ||
+ (info->ifi_index != pmd->if_index &&
+ info->ifi_index != pmd->remote_if_index))
+ return 0;
+ return tap_link_update(dev, 0);
+}
+
+static void
+tap_dev_intr_handler(struct rte_intr_handle *intr_handle __rte_unused,
+ void *cb_arg)
+{
+ struct rte_eth_dev *dev = cb_arg;
+ struct pmd_internals *pmd = dev->data->dev_private;
+
+ nl_recv(pmd->intr_handle.fd, tap_nl_msg_handler, dev);
+}
+
+static int
+tap_intr_handle_set(struct rte_eth_dev *dev, int set)
+{
+ struct pmd_internals *pmd = dev->data->dev_private;
+
+ /* In any case, disable interrupt if the conf is no longer there. */
+ if (!dev->data->dev_conf.intr_conf.lsc) {
+ if (pmd->intr_handle.fd != -1)
+ nl_final(pmd->intr_handle.fd);
+ rte_intr_callback_unregister(
+ &pmd->intr_handle, tap_dev_intr_handler, dev);
+ return 0;
+ }
+ if (set) {
+ pmd->intr_handle.fd = nl_init(RTMGRP_LINK);
+ if (unlikely(pmd->intr_handle.fd == -1))
+ return -EBADF;
+ return rte_intr_callback_register(
+ &pmd->intr_handle, tap_dev_intr_handler, dev);
+ }
+ nl_final(pmd->intr_handle.fd);
+ return rte_intr_callback_unregister(&pmd->intr_handle,
+ tap_dev_intr_handler, dev);
+}
+
+static const uint32_t*
+tap_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
+{
+ static const uint32_t ptypes[] = {
+ RTE_PTYPE_INNER_L2_ETHER,
+ RTE_PTYPE_INNER_L2_ETHER_VLAN,
+ RTE_PTYPE_INNER_L2_ETHER_QINQ,
+ RTE_PTYPE_INNER_L3_IPV4,
+ RTE_PTYPE_INNER_L3_IPV4_EXT,
+ RTE_PTYPE_INNER_L3_IPV6,
+ RTE_PTYPE_INNER_L3_IPV6_EXT,
+ RTE_PTYPE_INNER_L4_FRAG,
+ RTE_PTYPE_INNER_L4_UDP,
+ RTE_PTYPE_INNER_L4_TCP,
+ RTE_PTYPE_INNER_L4_SCTP,
+ RTE_PTYPE_L2_ETHER,
+ RTE_PTYPE_L2_ETHER_VLAN,
+ RTE_PTYPE_L2_ETHER_QINQ,
+ RTE_PTYPE_L3_IPV4,
+ RTE_PTYPE_L3_IPV4_EXT,
+ RTE_PTYPE_L3_IPV6_EXT,
+ RTE_PTYPE_L3_IPV6,
+ RTE_PTYPE_L4_FRAG,
+ RTE_PTYPE_L4_UDP,
+ RTE_PTYPE_L4_TCP,
+ RTE_PTYPE_L4_SCTP,
+ };
+
+ return ptypes;
+}
+
+static int
+tap_flow_ctrl_get(struct rte_eth_dev *dev __rte_unused,
+ struct rte_eth_fc_conf *fc_conf)
+{
+ fc_conf->mode = RTE_FC_NONE;
+ return 0;
+}
+
+static int
+tap_flow_ctrl_set(struct rte_eth_dev *dev __rte_unused,
+ struct rte_eth_fc_conf *fc_conf)
+{
+ if (fc_conf->mode != RTE_FC_NONE)
+ return -ENOTSUP;
+ return 0;
+}
+
static const struct eth_dev_ops ops = {
.dev_start = tap_dev_start,
.dev_stop = tap_dev_stop,
.tx_queue_setup = tap_tx_queue_setup,
.rx_queue_release = tap_rx_queue_release,
.tx_queue_release = tap_tx_queue_release,
+ .flow_ctrl_get = tap_flow_ctrl_get,
+ .flow_ctrl_set = tap_flow_ctrl_set,
.link_update = tap_link_update,
.dev_set_link_up = tap_link_set_up,
.dev_set_link_down = tap_link_set_down,
.promiscuous_disable = tap_promisc_disable,
.allmulticast_enable = tap_allmulti_enable,
.allmulticast_disable = tap_allmulti_disable,
+ .mac_addr_set = tap_mac_set,
+ .mtu_set = tap_mtu_set,
+ .set_mc_addr_list = tap_set_mc_addr_list,
.stats_get = tap_stats_get,
.stats_reset = tap_stats_reset,
+ .dev_supported_ptypes_get = tap_dev_supported_ptypes_get,
+ .filter_ctrl = tap_dev_filter_ctrl,
};
static int
-eth_dev_tap_create(const char *name, char *tap_name)
+tap_kernel_support(struct pmd_internals *pmd)
+{
+ struct utsname utsname;
+ int ver[3];
+
+ if (uname(&utsname) == -1 ||
+ sscanf(utsname.release, "%d.%d.%d",
+ &ver[0], &ver[1], &ver[2]) != 3)
+ return 0;
+ if (KERNEL_VERSION(ver[0], ver[1], ver[2]) >= FLOWER_KERNEL_VERSION)
+ pmd->flower_support = 1;
+ if (KERNEL_VERSION(ver[0], ver[1], ver[2]) >=
+ FLOWER_VLAN_KERNEL_VERSION)
+ pmd->flower_vlan_support = 1;
+ return 1;
+}
+
+static int
+eth_dev_tap_create(const char *name, char *tap_name, char *remote_iface)
{
int numa_node = rte_socket_id();
struct rte_eth_dev *dev = NULL;
pmd->nb_queues = RTE_PMD_TAP_MAX_QUEUES;
+ pmd->ioctl_sock = socket(AF_INET, SOCK_DGRAM, 0);
+ if (pmd->ioctl_sock == -1) {
+ RTE_LOG(ERR, PMD,
+ "TAP Unable to get a socket for management: %s\n",
+ strerror(errno));
+ goto error_exit;
+ }
+
/* Setup some default values */
data->dev_private = pmd;
data->port_id = dev->data->port_id;
- data->dev_flags = RTE_ETH_DEV_DETACHABLE;
+ data->mtu = dev->data->mtu;
+ data->dev_flags = RTE_ETH_DEV_DETACHABLE | RTE_ETH_DEV_INTR_LSC;
data->kdrv = RTE_KDRV_NONE;
data->drv_name = pmd_tap_drv.driver.name;
data->numa_node = numa_node;
dev->rx_pkt_burst = pmd_rx_burst;
dev->tx_pkt_burst = pmd_tx_burst;
+ pmd->intr_handle.type = RTE_INTR_HANDLE_EXT;
+ pmd->intr_handle.fd = -1;
+
/* Presetup the fds to -1 as being not valid */
for (i = 0; i < RTE_PMD_TAP_MAX_QUEUES; i++) {
pmd->rxq[i].fd = -1;
pmd->txq[i].fd = -1;
}
+ tap_kernel_support(pmd);
+ if (!pmd->flower_support)
+ return 0;
+ LIST_INIT(&pmd->flows);
+ /*
+ * If no netlink socket can be created, then it will fail when
+ * creating/destroying flow rules.
+ */
+ pmd->nlsk_fd = nl_init(0);
+ if (strlen(remote_iface)) {
+ pmd->remote_if_index = if_nametoindex(remote_iface);
+ snprintf(pmd->remote_iface, RTE_ETH_NAME_MAX_LEN,
+ "%s", remote_iface);
+ if (!pmd->remote_if_index)
+ RTE_LOG(ERR, PMD, "Could not find %s ifindex: "
+ "remote interface will remain unconfigured\n",
+ remote_iface);
+ }
+
return 0;
error_exit:
return 0;
}
+static int
+set_remote_iface(const char *key __rte_unused,
+ const char *value,
+ void *extra_args)
+{
+ char *name = (char *)extra_args;
+
+ if (value)
+ snprintf(name, RTE_ETH_NAME_MAX_LEN, "%s", value);
+
+ return 0;
+}
+
/* Open a TAP interface device.
*/
static int
struct rte_kvargs *kvlist = NULL;
int speed;
char tap_name[RTE_ETH_NAME_MAX_LEN];
+ char remote_iface[RTE_ETH_NAME_MAX_LEN];
speed = ETH_SPEED_NUM_10G;
snprintf(tap_name, sizeof(tap_name), "%s%d",
DEFAULT_TAP_NAME, tap_unit++);
+ memset(remote_iface, 0, RTE_ETH_NAME_MAX_LEN);
if (params && (params[0] != '\0')) {
RTE_LOG(DEBUG, PMD, "paramaters (%s)\n", params);
if (ret == -1)
goto leave;
}
+
+ if (rte_kvargs_count(kvlist, ETH_TAP_REMOTE_ARG) == 1) {
+ ret = rte_kvargs_process(kvlist,
+ ETH_TAP_REMOTE_ARG,
+ &set_remote_iface,
+ remote_iface);
+ if (ret == -1)
+ goto leave;
+ }
}
}
pmd_link.link_speed = speed;
RTE_LOG(NOTICE, PMD, "Initializing pmd_tap for %s as %s\n",
name, tap_name);
- ret = eth_dev_tap_create(name, tap_name);
+ ret = eth_dev_tap_create(name, tap_name, remote_iface);
leave:
if (ret == -1) {
return 0;
internals = eth_dev->data->dev_private;
+ if (internals->flower_support && internals->nlsk_fd) {
+ tap_flow_flush(eth_dev, NULL);
+ tap_flow_implicit_flush(internals, NULL);
+ nl_final(internals->nlsk_fd);
+ }
for (i = 0; i < internals->nb_queues; i++)
if (internals->rxq[i].fd != -1)
close(internals->rxq[i].fd);
+ close(internals->ioctl_sock);
rte_free(eth_dev->data->dev_private);
rte_free(eth_dev->data);