diff mbox series

[1/2] net/hns3: fix possible copper port initialize fail

Message ID 20250210030113.1154709-2-haijie1@huawei.com (mailing list archive)
State Accepted
Delegated to: Stephen Hemminger
Headers
Series net/hns3: bugfix on hns3 |

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Jie Hai Feb. 10, 2025, 3:01 a.m. UTC 
From: Dengdui Huang <huangdengdui@huawei.com>

The initialization of copper port contains the following two steps.
1. Configure firmware takeover the PHY. The firmware will start an
   asynchronous task to initialize the PHY chip.
2. Configure work speed and duplex.
In earlier versions of the firmware, when the asynchronous task is not
finished, the firmware will return -ENOTBLK in the second step. And this
will lead to driver failed to initialize. Here add retry for this case.

Fixes: 2e4859f3b362 ("net/hns3: support PF device with copper PHYs")
Cc: stable@dpdk.org

Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
---
 drivers/net/hns3/hns3_ethdev.c | 29 ++++++++++++++++++++++++++++-
 1 file changed, 28 insertions(+), 1 deletion(-)
diff mbox series

Patch

diff --git a/drivers/net/hns3/hns3_ethdev.c b/drivers/net/hns3/hns3_ethdev.c
index 72d1c30a7b2e..9f7119b734bc 100644
--- a/drivers/net/hns3/hns3_ethdev.c
+++ b/drivers/net/hns3/hns3_ethdev.c
@@ -4852,7 +4852,7 @@  hns3_get_link_duplex(uint32_t link_speeds)
 }
 
 static int
-hns3_set_copper_port_link_speed(struct hns3_hw *hw,
+hns3_copper_port_link_speed_cfg(struct hns3_hw *hw,
 				struct hns3_set_link_speed_cfg *cfg)
 {
 	struct hns3_cmd_desc desc[HNS3_PHY_PARAM_CFG_BD_NUM];
@@ -4886,6 +4886,33 @@  hns3_set_copper_port_link_speed(struct hns3_hw *hw,
 	return hns3_cmd_send(hw, desc, HNS3_PHY_PARAM_CFG_BD_NUM);
 }
 
+static int
+hns3_set_copper_port_link_speed(struct hns3_hw *hw,
+				struct hns3_set_link_speed_cfg *cfg)
+{
+#define HNS3_PHY_PARAM_CFG_RETRY_TIMES		10
+#define HNS3_PHY_PARAM_CFG_RETRY_DELAY_MS	100
+	uint32_t retry_cnt = 0;
+	int ret;
+
+	/*
+	 * The initialization of copper port contains the following two steps.
+	 * 1. Configure firmware takeover the PHY. The firmware will start an
+	 *    asynchronous task to initialize the PHY chip.
+	 * 2. Configure work speed and duplex.
+	 * In earlier versions of the firmware, when the asynchronous task is not
+	 * finished, the firmware will return -ENOTBLK in the second step. And this
+	 * will lead to driver failed to initialize. Here add retry for this case.
+	 */
+	ret = hns3_copper_port_link_speed_cfg(hw, cfg);
+	while (ret == -ENOTBLK && retry_cnt++ < HNS3_PHY_PARAM_CFG_RETRY_TIMES) {
+		rte_delay_ms(HNS3_PHY_PARAM_CFG_RETRY_DELAY_MS);
+		ret = hns3_copper_port_link_speed_cfg(hw, cfg);
+	}
+
+	return ret;
+}
+
 static int
 hns3_set_autoneg(struct hns3_hw *hw, bool enable)
 {