On Monday 18 September 2017 08:26 PM, Ferruh Yigit wrote:
On 9/9/2017 12:21 PM, Shreyansh Jain wrote:
Signed-off-by: Hemant Agrawal <hemant.agra...@nxp.com>
Signed-off-by: Shreyansh Jain <shreyansh.j...@nxp.com>
<...>
+static int dpaa_eth_link_update(struct rte_eth_dev *dev,
+ int wait_to_complete __rte_unused)
+{
+ struct dpaa_if *dpaa_intf = dev->data->dev_private;
+ struct rte_eth_link *link = &dev->data->dev_link;
+
+ PMD_INIT_FUNC_TRACE();
+
+ if (dpaa_intf->fif->mac_type == fman_mac_1g)
+ link->link_speed = 1000;
+ else if (dpaa_intf->fif->mac_type == fman_mac_10g)
+ link->link_speed = 10000;
+ else
+ DPAA_PMD_ERR("invalid link_speed: %s, %d",
+ dpaa_intf->name, dpaa_intf->fif->mac_type);
+
+ link->link_status = dpaa_intf->valid;
+ link->link_duplex = ETH_LINK_FULL_DUPLEX;
+ link->link_autoneg = ETH_LINK_AUTONEG;
Shouldn't this function go and get link information from hardware?
Our currently hardware interfaces don't support these operations
explicitly. For the "fman_mac_1g" and "fman_mac_10g", these are the
default sets which are exposing.
Overtime, we will get more such interfaces exposed from Linux kernel to
Fman library and update this code.
+ return 0;
+}
+
static
int dpaa_eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
uint16_t nb_desc __rte_unused,
@@ -216,6 +238,22 @@ static void dpaa_eth_tx_queue_release(void *txq
__rte_unused)
PMD_INIT_FUNC_TRACE();
}
+static int dpaa_link_down(struct rte_eth_dev *dev)
+{
+ PMD_INIT_FUNC_TRACE();
+
+ dpaa_eth_dev_stop(dev);
Drivers tend to do revers, make link down on device stop. Just to double
check if stop() is intended for link down.
fman_if_disable_rx is equivalent to "link down" as well as stop (because
it flushes the queues). That is why these APIs are linked.
+ return 0;
+}
+
+static int dpaa_link_up(struct rte_eth_dev *dev)
+{
+ PMD_INIT_FUNC_TRACE();
+
+ dpaa_eth_dev_start(dev);
+ return 0;
+}
<...>
