ethdev: add return values to callback process API
[dpdk.git] / drivers / net / mlx5 / mlx5_ethdev.c
index 7763998..96bccd5 100644 (file)
 #include <net/if.h>
 #include <sys/ioctl.h>
 #include <sys/socket.h>
+#include <sys/utsname.h>
 #include <netinet/in.h>
 #include <linux/ethtool.h>
 #include <linux/sockios.h>
+#include <linux/version.h>
 #include <fcntl.h>
 
 /* DPDK headers don't like -pedantic. */
 #include "mlx5_rxtx.h"
 #include "mlx5_utils.h"
 
+/* Add defines in case the running kernel is not the same as user headers. */
+#ifndef ETHTOOL_GLINKSETTINGS
+struct ethtool_link_settings {
+       uint32_t cmd;
+       uint32_t speed;
+       uint8_t duplex;
+       uint8_t port;
+       uint8_t phy_address;
+       uint8_t autoneg;
+       uint8_t mdio_support;
+       uint8_t eth_to_mdix;
+       uint8_t eth_tp_mdix_ctrl;
+       int8_t link_mode_masks_nwords;
+       uint32_t reserved[8];
+       uint32_t link_mode_masks[];
+};
+
+#define ETHTOOL_GLINKSETTINGS 0x0000004c
+#define ETHTOOL_LINK_MODE_1000baseT_Full_BIT 5
+#define ETHTOOL_LINK_MODE_Autoneg_BIT 6
+#define ETHTOOL_LINK_MODE_1000baseKX_Full_BIT 17
+#define ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT 18
+#define ETHTOOL_LINK_MODE_10000baseKR_Full_BIT 19
+#define ETHTOOL_LINK_MODE_10000baseR_FEC_BIT 20
+#define ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT 21
+#define ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT 22
+#define ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT 23
+#define ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT 24
+#define ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT 25
+#define ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT 26
+#define ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT 27
+#define ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT 28
+#define ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT 29
+#define ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT 30
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_25G
+#define ETHTOOL_LINK_MODE_25000baseCR_Full_BIT 31
+#define ETHTOOL_LINK_MODE_25000baseKR_Full_BIT 32
+#define ETHTOOL_LINK_MODE_25000baseSR_Full_BIT 33
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_50G
+#define ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT 34
+#define ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT 35
+#endif
+#ifndef HAVE_ETHTOOL_LINK_MODE_100G
+#define ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT 36
+#define ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT 37
+#define ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT 38
+#define ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT 39
+#endif
+
 /**
  * Return private structure associated with an Ethernet device.
  *
@@ -180,6 +233,23 @@ try_dev_id:
        return 0;
 }
 
+/**
+ * Check if the counter is located on ib counters file.
+ *
+ * @param[in] cntr
+ *   Counter name.
+ *
+ * @return
+ *   1 if counter is located on ib counters file , 0 otherwise.
+ */
+int
+priv_is_ib_cntr(const char *cntr)
+{
+       if (!strcmp(cntr, "out_of_buffer"))
+               return 1;
+       return 0;
+}
+
 /**
  * Read from sysfs entry.
  *
@@ -207,10 +277,15 @@ priv_sysfs_read(const struct priv *priv, const char *entry,
        if (priv_get_ifname(priv, &ifname))
                return -1;
 
-       MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path,
-             ifname, entry);
-
-       file = fopen(path, "rb");
+       if (priv_is_ib_cntr(entry)) {
+               MKSTR(path, "%s/ports/1/hw_counters/%s",
+                     priv->ctx->device->ibdev_path, entry);
+               file = fopen(path, "rb");
+       } else {
+               MKSTR(path, "%s/device/net/%s/%s",
+                     priv->ctx->device->ibdev_path, ifname, entry);
+               file = fopen(path, "rb");
+       }
        if (file == NULL)
                return -1;
        ret = fread(buf, 1, size, file);
@@ -415,6 +490,30 @@ priv_get_mtu(struct priv *priv, uint16_t *mtu)
        return 0;
 }
 
+/**
+ * Read device counter from sysfs.
+ *
+ * @param priv
+ *   Pointer to private structure.
+ * @param name
+ *   Counter name.
+ * @param[out] cntr
+ *   Counter output buffer.
+ *
+ * @return
+ *   0 on success, -1 on failure and errno is set.
+ */
+int
+priv_get_cntr_sysfs(struct priv *priv, const char *name, uint64_t *cntr)
+{
+       unsigned long ulong_ctr;
+
+       if (priv_get_sysfs_ulong(priv, name, &ulong_ctr) == -1)
+               return -1;
+       *cntr = ulong_ctr;
+       return 0;
+}
+
 /**
  * Set device MTU.
  *
@@ -562,6 +661,8 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
        unsigned int max;
        char ifname[IF_NAMESIZE];
 
+       info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+
        priv_lock(priv);
        /* FIXME: we should ask the device for these values. */
        info->min_rx_bufsize = 32;
@@ -592,14 +693,16 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info)
                        (DEV_TX_OFFLOAD_IPV4_CKSUM |
                         DEV_TX_OFFLOAD_UDP_CKSUM |
                         DEV_TX_OFFLOAD_TCP_CKSUM);
+       if (priv->tso)
+               info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO;
+       if (priv->tunnel_en)
+               info->tx_offload_capa |= (DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
+                                         DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
+                                         DEV_TX_OFFLOAD_GRE_TNL_TSO);
        if (priv_get_ifname(priv, &ifname) == 0)
                info->if_index = if_nametoindex(ifname);
-       /* FIXME: RETA update/query API expects the callee to know the size of
-        * the indirection table, for this PMD the size varies depending on
-        * the number of RX queues, it becomes impossible to find the correct
-        * size if it is not fixed.
-        * The API should be updated to solve this problem. */
-       info->reta_size = priv->ind_table_max_size;
+       info->reta_size = priv->reta_idx_n ?
+               priv->reta_idx_n : priv->ind_table_max_size;
        info->hash_key_size = ((*priv->rss_conf) ?
                               (*priv->rss_conf)[0]->rss_key_len :
                               0);
@@ -612,10 +715,10 @@ mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev)
 {
        static const uint32_t ptypes[] = {
                /* refers to rxq_cq_to_pkt_type() */
-               RTE_PTYPE_L3_IPV4,
-               RTE_PTYPE_L3_IPV6,
-               RTE_PTYPE_INNER_L3_IPV4,
-               RTE_PTYPE_INNER_L3_IPV6,
+               RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
+               RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
+               RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
+               RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
                RTE_PTYPE_UNKNOWN
 
        };
@@ -626,15 +729,15 @@ mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev)
 }
 
 /**
- * DPDK callback to retrieve physical link information (unlocked version).
+ * DPDK callback to retrieve physical link information.
  *
  * @param dev
  *   Pointer to Ethernet device structure.
  * @param wait_to_complete
  *   Wait for request completion (ignored).
  */
-int
-mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
+static int
+mlx5_link_update_unlocked_gset(struct rte_eth_dev *dev, int wait_to_complete)
 {
        struct priv *priv = mlx5_get_priv(dev);
        struct ethtool_cmd edata = {
@@ -644,6 +747,8 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
        struct rte_eth_link dev_link;
        int link_speed = 0;
 
+       /* priv_lock() is not taken to allow concurrent calls. */
+
        (void)wait_to_complete;
        if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) {
                WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno));
@@ -689,6 +794,90 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
        return -1;
 }
 
+/**
+ * Retrieve physical link information (unlocked version using new ioctl).
+ *
+ * @param dev
+ *   Pointer to Ethernet device structure.
+ * @param wait_to_complete
+ *   Wait for request completion (ignored).
+ */
+static int
+mlx5_link_update_unlocked_gs(struct rte_eth_dev *dev, int wait_to_complete)
+{
+       struct priv *priv = mlx5_get_priv(dev);
+       struct ethtool_link_settings edata = {
+               .cmd = ETHTOOL_GLINKSETTINGS,
+       };
+       struct ifreq ifr;
+       struct rte_eth_link dev_link;
+       uint64_t sc;
+
+       (void)wait_to_complete;
+       if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) {
+               WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno));
+               return -1;
+       }
+       memset(&dev_link, 0, sizeof(dev_link));
+       dev_link.link_status = ((ifr.ifr_flags & IFF_UP) &&
+                               (ifr.ifr_flags & IFF_RUNNING));
+       ifr.ifr_data = (void *)&edata;
+       if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) {
+               DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s",
+                     strerror(errno));
+               return -1;
+       }
+       dev_link.link_speed = edata.speed;
+       sc = edata.link_mode_masks[0] |
+               ((uint64_t)edata.link_mode_masks[1] << 32);
+       priv->link_speed_capa = 0;
+       if (sc & ETHTOOL_LINK_MODE_Autoneg_BIT)
+               priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG;
+       if (sc & (ETHTOOL_LINK_MODE_1000baseT_Full_BIT |
+                 ETHTOOL_LINK_MODE_1000baseKX_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_1G;
+       if (sc & (ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT |
+                 ETHTOOL_LINK_MODE_10000baseKR_Full_BIT |
+                 ETHTOOL_LINK_MODE_10000baseR_FEC_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_10G;
+       if (sc & (ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT |
+                 ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_20G;
+       if (sc & (ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_40G;
+       if (sc & (ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_56G;
+       if (sc & (ETHTOOL_LINK_MODE_25000baseCR_Full_BIT |
+                 ETHTOOL_LINK_MODE_25000baseKR_Full_BIT |
+                 ETHTOOL_LINK_MODE_25000baseSR_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_25G;
+       if (sc & (ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT |
+                 ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_50G;
+       if (sc & (ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT |
+                 ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT))
+               priv->link_speed_capa |= ETH_LINK_SPEED_100G;
+       dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ?
+                               ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX);
+       dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
+                                 ETH_LINK_SPEED_FIXED);
+       if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) {
+               /* Link status changed. */
+               dev->data->dev_link = dev_link;
+               return 0;
+       }
+       /* Link status is still the same. */
+       return -1;
+}
+
 /**
  * DPDK callback to retrieve physical link information.
  *
@@ -700,13 +889,15 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete)
 int
 mlx5_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 {
-       struct priv *priv = mlx5_get_priv(dev);
-       int ret;
-
-       priv_lock(priv);
-       ret = mlx5_link_update_unlocked(dev, wait_to_complete);
-       priv_unlock(priv);
-       return ret;
+       struct utsname utsname;
+       int ver[3];
+
+       if (uname(&utsname) == -1 ||
+           sscanf(utsname.release, "%d.%d.%d",
+                  &ver[0], &ver[1], &ver[2]) != 3 ||
+           KERNEL_VERSION(ver[0], ver[1], ver[2]) < KERNEL_VERSION(4, 9, 0))
+               return mlx5_link_update_unlocked_gset(dev, wait_to_complete);
+       return mlx5_link_update_unlocked_gs(dev, wait_to_complete);
 }
 
 /**
@@ -800,7 +991,6 @@ recover:
                struct rxq *rxq = (*priv->rxqs)[i];
                struct rxq_ctrl *rxq_ctrl =
                        container_of(rxq, struct rxq_ctrl, rxq);
-               int sp;
                unsigned int mb_len;
                unsigned int tmp;
 
@@ -808,10 +998,9 @@ recover:
                        continue;
                mb_len = rte_pktmbuf_data_room_size(rxq->mp);
                assert(mb_len >= RTE_PKTMBUF_HEADROOM);
-               /* Toggle scattered support (sp) if necessary. */
-               sp = (max_frame_len > (mb_len - RTE_PKTMBUF_HEADROOM));
                /* Provide new values to rxq_setup(). */
-               dev->data->dev_conf.rxmode.jumbo_frame = sp;
+               dev->data->dev_conf.rxmode.jumbo_frame =
+                       (max_frame_len > ETHER_MAX_LEN);
                dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame_len;
                if (rehash)
                        ret = rxq_rehash(dev, rxq_ctrl);
@@ -1024,7 +1213,7 @@ static int
 priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev)
 {
        struct ibv_async_event event;
-       int port_change = 0;
+       struct rte_eth_link *link = &dev->data->dev_link;
        int ret = 0;
 
        /* Read all message and acknowledge them. */
@@ -1032,29 +1221,24 @@ priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev)
                if (ibv_get_async_event(priv->ctx, &event))
                        break;
 
-               if (event.event_type == IBV_EVENT_PORT_ACTIVE ||
-                   event.event_type == IBV_EVENT_PORT_ERR)
-                       port_change = 1;
-               else
+               if (event.event_type != IBV_EVENT_PORT_ACTIVE &&
+                   event.event_type != IBV_EVENT_PORT_ERR)
                        DEBUG("event type %d on port %d not handled",
                              event.event_type, event.element.port_num);
                ibv_ack_async_event(&event);
        }
-
-       if (port_change ^ priv->pending_alarm) {
-               struct rte_eth_link *link = &dev->data->dev_link;
-
-               priv->pending_alarm = 0;
-               mlx5_link_update_unlocked(dev, 0);
-               if (((link->link_speed == 0) && link->link_status) ||
-                   ((link->link_speed != 0) && !link->link_status)) {
+       mlx5_link_update(dev, 0);
+       if (((link->link_speed == 0) && link->link_status) ||
+           ((link->link_speed != 0) && !link->link_status)) {
+               if (!priv->pending_alarm) {
                        /* Inconsistent status, check again later. */
                        priv->pending_alarm = 1;
                        rte_eal_alarm_set(MLX5_ALARM_TIMEOUT_US,
                                          mlx5_dev_link_status_handler,
                                          dev);
-               } else
-                       ret = 1;
+               }
+       } else {
+               ret = 1;
        }
        return ret;
 }
@@ -1074,10 +1258,12 @@ mlx5_dev_link_status_handler(void *arg)
 
        priv_lock(priv);
        assert(priv->pending_alarm == 1);
+       priv->pending_alarm = 0;
        ret = priv_dev_link_status_handler(priv, dev);
        priv_unlock(priv);
        if (ret)
-               _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
+               _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL,
+                                             NULL);
 }
 
 /**
@@ -1089,18 +1275,18 @@ mlx5_dev_link_status_handler(void *arg)
  *   Callback argument.
  */
 void
-mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg)
+mlx5_dev_interrupt_handler(void *cb_arg)
 {
        struct rte_eth_dev *dev = cb_arg;
        struct priv *priv = dev->data->dev_private;
        int ret;
 
-       (void)intr_handle;
        priv_lock(priv);
        ret = priv_dev_link_status_handler(priv, dev);
        priv_unlock(priv);
        if (ret)
-               _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
+               _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL,
+                                             NULL);
 }
 
 /**
@@ -1398,14 +1584,15 @@ void
 priv_select_tx_function(struct priv *priv)
 {
        priv->dev->tx_pkt_burst = mlx5_tx_burst;
-       /* Display warning for unsupported configurations. */
-       if (priv->sriov && priv->mps)
-               WARN("multi-packet send WQE cannot be used on a SR-IOV setup");
        /* Select appropriate TX function. */
-       if ((priv->sriov == 0) && priv->mps && priv->txq_inline) {
+       if (priv->mps == MLX5_MPW_ENHANCED) {
+               priv->dev->tx_pkt_burst =
+                       mlx5_tx_burst_empw;
+               DEBUG("selected Enhanced MPW TX function");
+       } else if (priv->mps && priv->txq_inline) {
                priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw_inline;
                DEBUG("selected MPW inline TX function");
-       } else if ((priv->sriov == 0) && priv->mps) {
+       } else if (priv->mps) {
                priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw;
                DEBUG("selected MPW TX function");
        }