maintainers: claim IP fragmentation and ACL
[dpdk.git] / lib / librte_pmd_ixgbe / ixgbe_ethdev.c
index 9401916..b341dd0 100644 (file)
@@ -231,12 +231,6 @@ static int ixgbe_add_syn_filter(struct rte_eth_dev *dev,
 static int ixgbe_remove_syn_filter(struct rte_eth_dev *dev);
 static int ixgbe_get_syn_filter(struct rte_eth_dev *dev,
                        struct rte_syn_filter *filter, uint16_t *rx_queue);
-static int ixgbe_add_ethertype_filter(struct rte_eth_dev *dev, uint16_t index,
-                       struct rte_ethertype_filter *filter, uint16_t rx_queue);
-static int ixgbe_remove_ethertype_filter(struct rte_eth_dev *dev,
-                       uint16_t index);
-static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev, uint16_t index,
-                       struct rte_ethertype_filter *filter, uint16_t *rx_queue);
 static int ixgbe_add_5tuple_filter(struct rte_eth_dev *dev, uint16_t index,
                        struct rte_5tuple_filter *filter, uint16_t rx_queue);
 static int ixgbe_remove_5tuple_filter(struct rte_eth_dev *dev,
@@ -245,6 +239,18 @@ static int ixgbe_get_5tuple_filter(struct rte_eth_dev *dev, uint16_t index,
                        struct rte_5tuple_filter *filter, uint16_t *rx_queue);
 
 static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu);
+static int ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
+                       struct rte_eth_ethertype_filter *filter,
+                       bool add);
+static int ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
+                               enum rte_filter_op filter_op,
+                               void *arg);
+static int ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
+                       struct rte_eth_ethertype_filter *filter);
+static int ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
+                    enum rte_filter_type filter_type,
+                    enum rte_filter_op filter_op,
+                    void *arg);
 
 /*
  * Define VF Stats MACRO for Non "cleared on read" register
@@ -380,12 +386,10 @@ static struct eth_dev_ops ixgbe_eth_dev_ops = {
        .add_syn_filter          = ixgbe_add_syn_filter,
        .remove_syn_filter       = ixgbe_remove_syn_filter,
        .get_syn_filter          = ixgbe_get_syn_filter,
-       .add_ethertype_filter    = ixgbe_add_ethertype_filter,
-       .remove_ethertype_filter = ixgbe_remove_ethertype_filter,
-       .get_ethertype_filter    = ixgbe_get_ethertype_filter,
        .add_5tuple_filter       = ixgbe_add_5tuple_filter,
        .remove_5tuple_filter    = ixgbe_remove_5tuple_filter,
        .get_5tuple_filter       = ixgbe_get_5tuple_filter,
+       .filter_ctrl             = ixgbe_dev_filter_ctrl,
 };
 
 /*
@@ -749,9 +753,16 @@ eth_ixgbe_dev_init(__attribute__((unused)) struct eth_driver *eth_drv,
         */
        if (rte_eal_process_type() != RTE_PROC_PRIMARY){
                struct igb_tx_queue *txq;
-               /* TX queue function in primary, set by last queue initialized */
-               txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
-               set_tx_function(eth_dev, txq);
+               /* TX queue function in primary, set by last queue initialized
+                * Tx queue may not initialized by primary process */
+               if (eth_dev->data->tx_queues) {
+                       txq = eth_dev->data->tx_queues[eth_dev->data->nb_tx_queues-1];
+                       set_tx_function(eth_dev, txq);
+               } else {
+                       /* Use default TX function if we get here */
+                       PMD_INIT_LOG(INFO, "No TX queues configured yet. "
+                                          "Using default TX function.");
+               }
 
                if (eth_dev->data->scattered_rx)
                        eth_dev->rx_pkt_burst = ixgbe_recv_scattered_pkts;
@@ -1469,6 +1480,7 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
        if (status != 0)
                return -1;
        hw->mac.ops.start_hw(hw);
+       hw->mac.get_link_status = true;
 
        /* configure PF module if SRIOV enabled */
        ixgbe_pf_host_configure(dev);
@@ -1483,7 +1495,11 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
                goto error;
        }
 
-       ixgbe_dev_rxtx_start(dev);
+       err = ixgbe_dev_rxtx_start(dev);
+       if (err < 0) {
+               PMD_INIT_LOG(ERR, "Unable to start rxtx queues");
+               goto error;
+       }
 
        if (ixgbe_is_sfp(hw) && hw->phy.multispeed_fiber) {
                err = hw->mac.ops.setup_sfp(hw);
@@ -1502,6 +1518,8 @@ ixgbe_dev_start(struct rte_eth_dev *dev)
        err = ixgbe_check_link(hw, &speed, &link_up, 0);
        if (err)
                goto error;
+       dev->data->dev_link.link_status = link_up;
+
        err = ixgbe_get_link_capabilities(hw, &speed, &negotiate);
        if (err)
                goto error;
@@ -1928,7 +1946,6 @@ ixgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
        if (stats == NULL)
                return;
 
-       memset(stats, 0, sizeof(*stats));
        stats->ipackets = hw_stats->vfgprc;
        stats->ibytes = hw_stats->vfgorc;
        stats->opackets = hw_stats->vfgptc;
@@ -2064,7 +2081,7 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
 {
        struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
        struct rte_eth_link link, old;
-       ixgbe_link_speed link_speed;
+       ixgbe_link_speed link_speed = IXGBE_LINK_SPEED_UNKNOWN;
        int link_up;
        int diag;
 
@@ -2088,6 +2105,12 @@ ixgbe_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
                return 0;
        }
 
+       if (link_speed == IXGBE_LINK_SPEED_UNKNOWN &&
+           !hw->mac.get_link_status) {
+               memcpy(&link, &old, sizeof(link));
+               return -1;
+       }
+
        if (link_up == 0) {
                rte_ixgbe_dev_atomic_write_link_status(dev, &link);
                if (link.link_status == old.link_status)
@@ -2926,6 +2949,7 @@ ixgbevf_dev_start(struct rte_eth_dev *dev)
        PMD_INIT_FUNC_TRACE();
 
        hw->mac.ops.reset_hw(hw);
+       hw->mac.get_link_status = true;
 
        /* negotiate mailbox API version to use with the PF. */
        ixgbevf_negotiate_api(hw);
@@ -3759,125 +3783,6 @@ ixgbe_get_syn_filter(struct rte_eth_dev *dev,
        return -ENOENT;
 }
 
-/*
- * add an ethertype filter
- *
- * @param
- * dev: Pointer to struct rte_eth_dev.
- * index: the index the filter allocates.
- * filter: ponter to the filter that will be added.
- * rx_queue: the queue id the filter assigned to.
- *
- * @return
- *    - On success, zero.
- *    - On failure, a negative value.
- */
-static int
-ixgbe_add_ethertype_filter(struct rte_eth_dev *dev,
-                       uint16_t index, struct rte_ethertype_filter *filter,
-                       uint16_t rx_queue)
-{
-       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-       uint32_t etqf, etqs = 0;
-
-       if (hw->mac.type != ixgbe_mac_82599EB)
-               return -ENOSYS;
-
-       if (index >= IXGBE_MAX_ETQF_FILTERS ||
-               rx_queue >= IXGBE_MAX_RX_QUEUE_NUM)
-               return -EINVAL;
-
-       etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(index));
-       if (etqf & IXGBE_ETQF_FILTER_EN)
-               return -EINVAL;  /* filter index is in use. */
-
-       etqf = 0;
-       etqf |= IXGBE_ETQF_FILTER_EN;
-       etqf |= (uint32_t)filter->ethertype;
-
-       if (filter->priority_en) {
-               if (filter->priority > IXGBE_ETQF_MAX_PRI)
-                       return -EINVAL;
-               etqf |= (uint32_t)((filter->priority << IXGBE_ETQF_SHIFT) & IXGBE_ETQF_UP);
-               etqf |= IXGBE_ETQF_UP_EN;
-       }
-       etqs |= (uint32_t)((rx_queue << IXGBE_ETQS_RX_QUEUE_SHIFT) & IXGBE_ETQS_RX_QUEUE);
-       etqs |= IXGBE_ETQS_QUEUE_EN;
-
-       IXGBE_WRITE_REG(hw, IXGBE_ETQF(index), etqf);
-       IXGBE_WRITE_REG(hw, IXGBE_ETQS(index), etqs);
-       return 0;
-}
-
-/*
- * remove an ethertype filter
- *
- * @param
- * dev: Pointer to struct rte_eth_dev.
- * index: the index the filter allocates.
- *
- * @return
- *    - On success, zero.
- *    - On failure, a negative value.
- */
-static int
-ixgbe_remove_ethertype_filter(struct rte_eth_dev *dev,
-                       uint16_t index)
-{
-       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-
-       if (hw->mac.type != ixgbe_mac_82599EB)
-               return -ENOSYS;
-
-       if (index >= IXGBE_MAX_ETQF_FILTERS)
-               return -EINVAL;
-
-       IXGBE_WRITE_REG(hw, IXGBE_ETQF(index), 0);
-       IXGBE_WRITE_REG(hw, IXGBE_ETQS(index), 0);
-
-       return 0;
-}
-
-/*
- * get an ethertype filter
- *
- * @param
- * dev: Pointer to struct rte_eth_dev.
- * index: the index the filter allocates.
- * filter: ponter to the filter that will be gotten.
- * *rx_queue: the ponited of the queue id the filter assigned to.
- *
- * @return
- *    - On success, zero.
- *    - On failure, a negative value.
- */
-static int
-ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
-                       uint16_t index, struct rte_ethertype_filter *filter,
-                       uint16_t *rx_queue)
-{
-       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
-       uint32_t etqf, etqs;
-
-       if (hw->mac.type != ixgbe_mac_82599EB)
-               return -ENOSYS;
-
-       if (index >= IXGBE_MAX_ETQF_FILTERS)
-               return -EINVAL;
-
-       etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(index));
-       etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(index));
-       if (etqf & IXGBE_ETQF_FILTER_EN) {
-               filter->ethertype = etqf & IXGBE_ETQF_ETHERTYPE;
-               filter->priority_en = (etqf & IXGBE_ETQF_UP_EN) ? 1 : 0;
-               if (filter->priority_en)
-                       filter->priority = (etqf & IXGBE_ETQF_UP) >> 16;
-               *rx_queue = (etqs & IXGBE_ETQS_RX_QUEUE) >> IXGBE_ETQS_RX_QUEUE_SHIFT;
-               return 0;
-       }
-       return -ENOENT;
-}
-
 static inline enum ixgbe_5tuple_protocol
 convert_protocol_type(uint8_t protocol_value)
 {
@@ -4104,6 +4009,219 @@ ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
        return 0;
 }
 
+#define MAC_TYPE_FILTER_SUP(type)    do {\
+       if ((type) != ixgbe_mac_82599EB && (type) != ixgbe_mac_X540 &&\
+               (type) != ixgbe_mac_X550)\
+               return -ENOTSUP;\
+} while (0)
+
+static inline int
+ixgbe_ethertype_filter_lookup(struct ixgbe_filter_info *filter_info,
+                       uint16_t ethertype)
+{
+       int i;
+
+       for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
+               if (filter_info->ethertype_filters[i] == ethertype &&
+                   (filter_info->ethertype_mask & (1 << i)))
+                       return i;
+       }
+       return -1;
+}
+
+static inline int
+ixgbe_ethertype_filter_insert(struct ixgbe_filter_info *filter_info,
+                       uint16_t ethertype)
+{
+       int i;
+
+       for (i = 0; i < IXGBE_MAX_ETQF_FILTERS; i++) {
+               if (!(filter_info->ethertype_mask & (1 << i))) {
+                       filter_info->ethertype_mask |= 1 << i;
+                       filter_info->ethertype_filters[i] = ethertype;
+                       return i;
+               }
+       }
+       return -1;
+}
+
+static inline int
+ixgbe_ethertype_filter_remove(struct ixgbe_filter_info *filter_info,
+                       uint8_t idx)
+{
+       if (idx >= IXGBE_MAX_ETQF_FILTERS)
+               return -1;
+       filter_info->ethertype_mask &= ~(1 << idx);
+       filter_info->ethertype_filters[idx] = 0;
+       return idx;
+}
+
+static int
+ixgbe_add_del_ethertype_filter(struct rte_eth_dev *dev,
+                       struct rte_eth_ethertype_filter *filter,
+                       bool add)
+{
+       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct ixgbe_filter_info *filter_info =
+               IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+       uint32_t etqf = 0;
+       uint32_t etqs = 0;
+       int ret;
+
+       if (filter->queue >= IXGBE_MAX_RX_QUEUE_NUM)
+               return -EINVAL;
+
+       if (filter->ether_type == ETHER_TYPE_IPv4 ||
+               filter->ether_type == ETHER_TYPE_IPv6) {
+               PMD_DRV_LOG(ERR, "unsupported ether_type(0x%04x) in"
+                       " ethertype filter.", filter->ether_type);
+               return -EINVAL;
+       }
+
+       if (filter->flags & RTE_ETHTYPE_FLAGS_MAC) {
+               PMD_DRV_LOG(ERR, "mac compare is unsupported.");
+               return -EINVAL;
+       }
+       if (filter->flags & RTE_ETHTYPE_FLAGS_DROP) {
+               PMD_DRV_LOG(ERR, "drop option is unsupported.");
+               return -EINVAL;
+       }
+
+       ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
+       if (ret >= 0 && add) {
+               PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter exists.",
+                           filter->ether_type);
+               return -EEXIST;
+       }
+       if (ret < 0 && !add) {
+               PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
+                           filter->ether_type);
+               return -ENOENT;
+       }
+
+       if (add) {
+               ret = ixgbe_ethertype_filter_insert(filter_info,
+                       filter->ether_type);
+               if (ret < 0) {
+                       PMD_DRV_LOG(ERR, "ethertype filters are full.");
+                       return -ENOSYS;
+               }
+               etqf = IXGBE_ETQF_FILTER_EN;
+               etqf |= (uint32_t)filter->ether_type;
+               etqs |= (uint32_t)((filter->queue <<
+                                   IXGBE_ETQS_RX_QUEUE_SHIFT) &
+                                   IXGBE_ETQS_RX_QUEUE);
+               etqs |= IXGBE_ETQS_QUEUE_EN;
+       } else {
+               ret = ixgbe_ethertype_filter_remove(filter_info, (uint8_t)ret);
+               if (ret < 0)
+                       return -ENOSYS;
+       }
+       IXGBE_WRITE_REG(hw, IXGBE_ETQF(ret), etqf);
+       IXGBE_WRITE_REG(hw, IXGBE_ETQS(ret), etqs);
+       IXGBE_WRITE_FLUSH(hw);
+
+       return 0;
+}
+
+static int
+ixgbe_get_ethertype_filter(struct rte_eth_dev *dev,
+                       struct rte_eth_ethertype_filter *filter)
+{
+       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       struct ixgbe_filter_info *filter_info =
+               IXGBE_DEV_PRIVATE_TO_FILTER_INFO(dev->data->dev_private);
+       uint32_t etqf, etqs;
+       int ret;
+
+       ret = ixgbe_ethertype_filter_lookup(filter_info, filter->ether_type);
+       if (ret < 0) {
+               PMD_DRV_LOG(ERR, "ethertype (0x%04x) filter doesn't exist.",
+                           filter->ether_type);
+               return -ENOENT;
+       }
+
+       etqf = IXGBE_READ_REG(hw, IXGBE_ETQF(ret));
+       if (etqf & IXGBE_ETQF_FILTER_EN) {
+               etqs = IXGBE_READ_REG(hw, IXGBE_ETQS(ret));
+               filter->ether_type = etqf & IXGBE_ETQF_ETHERTYPE;
+               filter->flags = 0;
+               filter->queue = (etqs & IXGBE_ETQS_RX_QUEUE) >>
+                              IXGBE_ETQS_RX_QUEUE_SHIFT;
+               return 0;
+       }
+       return -ENOENT;
+}
+
+/*
+ * ixgbe_ethertype_filter_handle - Handle operations for ethertype filter.
+ * @dev: pointer to rte_eth_dev structure
+ * @filter_op:operation will be taken.
+ * @arg: a pointer to specific structure corresponding to the filter_op
+ */
+static int
+ixgbe_ethertype_filter_handle(struct rte_eth_dev *dev,
+                               enum rte_filter_op filter_op,
+                               void *arg)
+{
+       struct ixgbe_hw *hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+       int ret;
+
+       MAC_TYPE_FILTER_SUP(hw->mac.type);
+
+       if (filter_op == RTE_ETH_FILTER_NOP)
+               return 0;
+
+       if (arg == NULL) {
+               PMD_DRV_LOG(ERR, "arg shouldn't be NULL for operation %u.",
+                           filter_op);
+               return -EINVAL;
+       }
+
+       switch (filter_op) {
+       case RTE_ETH_FILTER_ADD:
+               ret = ixgbe_add_del_ethertype_filter(dev,
+                       (struct rte_eth_ethertype_filter *)arg,
+                       TRUE);
+               break;
+       case RTE_ETH_FILTER_DELETE:
+               ret = ixgbe_add_del_ethertype_filter(dev,
+                       (struct rte_eth_ethertype_filter *)arg,
+                       FALSE);
+               break;
+       case RTE_ETH_FILTER_GET:
+               ret = ixgbe_get_ethertype_filter(dev,
+                       (struct rte_eth_ethertype_filter *)arg);
+               break;
+       default:
+               PMD_DRV_LOG(ERR, "unsupported operation %u.", filter_op);
+               ret = -EINVAL;
+               break;
+       }
+       return ret;
+}
+
+static int
+ixgbe_dev_filter_ctrl(struct rte_eth_dev *dev,
+                    enum rte_filter_type filter_type,
+                    enum rte_filter_op filter_op,
+                    void *arg)
+{
+       int ret = -EINVAL;
+
+       switch (filter_type) {
+       case RTE_ETH_FILTER_ETHERTYPE:
+               ret = ixgbe_ethertype_filter_handle(dev, filter_op, arg);
+               break;
+       default:
+               PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
+                                                       filter_type);
+               break;
+       }
+
+       return ret;
+}
+
 static struct rte_driver rte_ixgbe_driver = {
        .type = PMD_PDEV,
        .init = rte_ixgbe_pmd_init,