net/tap: use new Tx offloads API
authorMoti Haimovsky <motih@mellanox.com>
Wed, 17 Jan 2018 14:04:33 +0000 (16:04 +0200)
committerFerruh Yigit <ferruh.yigit@intel.com>
Sun, 21 Jan 2018 14:51:52 +0000 (15:51 +0100)
Ethdev Tx offloads API has changed since:
commit cba7f53b717d ("ethdev: introduce Tx queue offloads API")
This commit adds support for the new Tx offloads API.

Signed-off-by: Moti Haimovsky <motih@mellanox.com>
Acked-by: Pascal Mazon <pascal.mazon@6wind.com>
drivers/net/tap/rte_eth_tap.c
drivers/net/tap/rte_eth_tap.h

index be73f93..7357dce 100644 (file)
@@ -24,6 +24,7 @@
 #include <sys/mman.h>
 #include <errno.h>
 #include <signal.h>
+#include <stdbool.h>
 #include <stdint.h>
 #include <sys/uio.h>
 #include <unistd.h>
@@ -346,6 +347,42 @@ end:
        return num_rx;
 }
 
+static uint64_t
+tap_tx_offload_get_port_capa(void)
+{
+       /*
+        * In order to support legacy apps,
+        * report capabilities also as port capabilities.
+        */
+       return DEV_TX_OFFLOAD_IPV4_CKSUM |
+              DEV_TX_OFFLOAD_UDP_CKSUM |
+              DEV_TX_OFFLOAD_TCP_CKSUM;
+}
+
+static uint64_t
+tap_tx_offload_get_queue_capa(void)
+{
+       return DEV_TX_OFFLOAD_IPV4_CKSUM |
+              DEV_TX_OFFLOAD_UDP_CKSUM |
+              DEV_TX_OFFLOAD_TCP_CKSUM;
+}
+
+static bool
+tap_txq_are_offloads_valid(struct rte_eth_dev *dev, uint64_t offloads)
+{
+       uint64_t port_offloads = dev->data->dev_conf.txmode.offloads;
+       uint64_t queue_supp_offloads = tap_tx_offload_get_queue_capa();
+       uint64_t port_supp_offloads = tap_tx_offload_get_port_capa();
+
+       if ((offloads & (queue_supp_offloads | port_supp_offloads)) !=
+           offloads)
+               return false;
+       /* Verify we have no conflict with port offloads */
+       if ((port_offloads ^ offloads) & port_supp_offloads)
+               return false;
+       return true;
+}
+
 static void
 tap_tx_offload(char *packet, uint64_t ol_flags, unsigned int l2_len,
               unsigned int l3_len)
@@ -432,9 +469,10 @@ pmd_tx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
                                rte_pktmbuf_mtod(seg, void *);
                        seg = seg->next;
                }
-               if (mbuf->ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_IPV4) ||
-                   (mbuf->ol_flags & PKT_TX_L4_MASK) == PKT_TX_UDP_CKSUM ||
-                   (mbuf->ol_flags & PKT_TX_L4_MASK) == PKT_TX_TCP_CKSUM) {
+               if (txq->csum &&
+                   ((mbuf->ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_IPV4) ||
+                    (mbuf->ol_flags & PKT_TX_L4_MASK) == PKT_TX_UDP_CKSUM ||
+                    (mbuf->ol_flags & PKT_TX_L4_MASK) == PKT_TX_TCP_CKSUM))) {
                        /* Support only packets with all data in the same seg */
                        if (mbuf->nb_segs > 1)
                                break;
@@ -572,6 +610,17 @@ tap_dev_stop(struct rte_eth_dev *dev)
 static int
 tap_dev_configure(struct rte_eth_dev *dev)
 {
+       uint64_t supp_tx_offloads = tap_tx_offload_get_port_capa();
+       uint64_t tx_offloads = dev->data->dev_conf.txmode.offloads;
+
+       if ((tx_offloads & supp_tx_offloads) != tx_offloads) {
+               rte_errno = ENOTSUP;
+               RTE_LOG(ERR, PMD,
+                       "Some Tx offloads are not supported "
+                       "requested 0x%" PRIx64 " supported 0x%" PRIx64 "\n",
+                       tx_offloads, supp_tx_offloads);
+               return -rte_errno;
+       }
        if (dev->data->nb_rx_queues > RTE_PMD_TAP_MAX_QUEUES) {
                RTE_LOG(ERR, PMD,
                        "%s: number of rx queues %d exceeds max num of queues %d\n",
@@ -648,10 +697,9 @@ tap_dev_info(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
        dev_info->rx_offload_capa = (DEV_RX_OFFLOAD_IPV4_CKSUM |
                                     DEV_RX_OFFLOAD_UDP_CKSUM |
                                     DEV_RX_OFFLOAD_TCP_CKSUM);
-       dev_info->tx_offload_capa =
-               (DEV_TX_OFFLOAD_IPV4_CKSUM |
-                DEV_TX_OFFLOAD_UDP_CKSUM |
-                DEV_TX_OFFLOAD_TCP_CKSUM);
+       dev_info->tx_queue_offload_capa = tap_tx_offload_get_queue_capa();
+       dev_info->tx_offload_capa = dev_info->tx_queue_offload_capa |
+                                   tap_tx_offload_get_port_capa();
 }
 
 static int
@@ -1025,21 +1073,46 @@ tap_tx_queue_setup(struct rte_eth_dev *dev,
                   uint16_t tx_queue_id,
                   uint16_t nb_tx_desc __rte_unused,
                   unsigned int socket_id __rte_unused,
-                  const struct rte_eth_txconf *tx_conf __rte_unused)
+                  const struct rte_eth_txconf *tx_conf)
 {
        struct pmd_internals *internals = dev->data->dev_private;
+       struct tx_queue *txq;
        int ret;
 
        if (tx_queue_id >= dev->data->nb_tx_queues)
                return -1;
-
        dev->data->tx_queues[tx_queue_id] = &internals->txq[tx_queue_id];
+       txq = dev->data->tx_queues[tx_queue_id];
+       /*
+        * Don't verify port offloads for application which
+        * use the old API.
+        */
+       if (tx_conf != NULL &&
+           !!(tx_conf->txq_flags & ETH_TXQ_FLAGS_IGNORE)) {
+               if (tap_txq_are_offloads_valid(dev, tx_conf->offloads)) {
+                       txq->csum = !!(tx_conf->offloads &
+                                       (DEV_TX_OFFLOAD_IPV4_CKSUM |
+                                        DEV_TX_OFFLOAD_UDP_CKSUM |
+                                        DEV_TX_OFFLOAD_TCP_CKSUM));
+               } else {
+                       rte_errno = ENOTSUP;
+                       RTE_LOG(ERR, PMD,
+                               "%p: Tx queue offloads 0x%" PRIx64
+                               " don't match port offloads 0x%" PRIx64
+                               " or supported offloads 0x%" PRIx64,
+                               (void *)dev, tx_conf->offloads,
+                               dev->data->dev_conf.txmode.offloads,
+                               tap_tx_offload_get_port_capa());
+                       return -rte_errno;
+               }
+       }
        ret = tap_setup_queue(dev, internals, tx_queue_id, 0);
        if (ret == -1)
                return -1;
-
-       RTE_LOG(DEBUG, PMD, "  TX TAP device name %s, qid %d on fd %d\n",
-               internals->name, tx_queue_id, internals->txq[tx_queue_id].fd);
+       RTE_LOG(DEBUG, PMD,
+               "  TX TAP device name %s, qid %d on fd %d csum %s\n",
+               internals->name, tx_queue_id, internals->txq[tx_queue_id].fd,
+               txq->csum ? "on" : "off");
 
        return 0;
 }
index 829f32f..69a1a08 100644 (file)
@@ -76,6 +76,7 @@ struct rx_queue {
 struct tx_queue {
        int fd;
        uint16_t *mtu;                  /* Pointer to MTU from dev_data */
+       uint16_t csum:1;                /* Enable checksum offloading */
        struct pkt_stats stats;         /* Stats for this TX queue */
 };