diff mbox series

[2/5] net/hns3: fix promiscuous config not clear for PF on uninit

Message ID 1590139278-11046-3-git-send-email-xavier.huwei@huawei.com (mailing list archive)
State Accepted, archived
Delegated to: Ferruh Yigit
Headers
Series misc updates for hns3 PMD driver |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK

Commit Message

Wei Hu (Xavier) May 22, 2020, 9:21 a.m. UTC 
  From: Chengchang Tang <tangchengchang@huawei.com>

Currently, promiscuous mode configuration are not cleared during
unintialziation based on hns3 PF device. The residual entries may cause
unnecessary bandwidth usage.

So, we need clear the PF's promisc mode status during the uninit.

Fixes: a45fd0aa0ea1 ("net/hns3: fix Rx queue search with broadcast packet")
Fixes: d51867db65c1 ("net/hns3: add initialization")
Cc: stable@dpdk.org

Signed-off-by: Chengchang Tang <tangchengchang@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
---
 drivers/net/hns3/hns3_ethdev.c | 53 ++++++++++++++++++++++++++++++++++--------
 1 file changed, 43 insertions(+), 10 deletions(-)
diff mbox series

Patch

diff --git a/drivers/net/hns3/hns3_ethdev.c b/drivers/net/hns3/hns3_ethdev.c
index 06c4886..1c06b8f 100644
--- a/drivers/net/hns3/hns3_ethdev.c
+++ b/drivers/net/hns3/hns3_ethdev.c
@@ -3826,7 +3826,7 @@  hns3_set_promisc_mode(struct hns3_hw *hw, bool en_uc_pmc, bool en_mc_pmc)
 }
 
 static int
-hns3_clear_all_vfs_promisc_mode(struct hns3_hw *hw)
+hns3_promisc_init(struct hns3_hw *hw)
 {
 	struct hns3_adapter *hns = HNS3_DEV_HW_TO_ADAPTER(hw);
 	struct hns3_pf *pf = &hns->pf;
@@ -3834,16 +3834,54 @@  hns3_clear_all_vfs_promisc_mode(struct hns3_hw *hw)
 	uint16_t func_id;
 	int ret;
 
+	ret = hns3_set_promisc_mode(hw, false, false);
+	if (ret) {
+		PMD_INIT_LOG(ERR, "failed to set promisc mode, ret = %d", ret);
+		return ret;
+	}
+
+	/*
+	 * In current version VFs are not supported when PF is driven by DPDK
+	 * driver. After PF has been taken over by DPDK, the original VF will
+	 * be invalid. So, there is a possibility of entry residues. It should
+	 * clear VFs's promisc mode to avoid unnecessary bandwidth usage
+	 * during init.
+	 */
 	for (func_id = HNS3_1ST_VF_FUNC_ID; func_id < pf->func_num; func_id++) {
 		hns3_promisc_param_init(&param, false, false, false, func_id);
 		ret = hns3_cmd_set_promisc_mode(hw, &param);
-		if (ret)
+		if (ret) {
+			PMD_INIT_LOG(ERR, "failed to clear vf:%d promisc mode,"
+					" ret = %d", func_id, ret);
 			return ret;
+		}
 	}
 
 	return 0;
 }
 
+static void
+hns3_promisc_uninit(struct hns3_hw *hw)
+{
+	struct hns3_promisc_param param;
+	uint16_t func_id;
+	int ret;
+
+	func_id = HNS3_PF_FUNC_ID;
+
+	/*
+	 * In current version VFs are not supported when PF is driven by
+	 * DPDK driver, and VFs' promisc mode status has been cleared during
+	 * init and their status will not change. So just clear PF's promisc
+	 * mode status during uninit.
+	 */
+	hns3_promisc_param_init(&param, false, false, false, func_id);
+	ret = hns3_cmd_set_promisc_mode(hw, &param);
+	if (ret)
+		PMD_INIT_LOG(ERR, "failed to clear promisc status during"
+				" uninit, ret = %d", ret);
+}
+
 static int
 hns3_dev_promiscuous_enable(struct rte_eth_dev *dev)
 {
@@ -4186,15 +4224,9 @@  hns3_init_hardware(struct hns3_adapter *hns)
 		goto err_mac_init;
 	}
 
-	ret = hns3_set_promisc_mode(hw, false, false);
-	if (ret) {
-		PMD_INIT_LOG(ERR, "Failed to set promisc mode: %d", ret);
-		goto err_mac_init;
-	}
-
-	ret = hns3_clear_all_vfs_promisc_mode(hw);
+	ret = hns3_promisc_init(hw);
 	if (ret) {
-		PMD_INIT_LOG(ERR, "Failed to clear all vfs promisc mode: %d",
+		PMD_INIT_LOG(ERR, "Failed to init promisc: %d",
 			     ret);
 		goto err_mac_init;
 	}
@@ -4353,6 +4385,7 @@  hns3_uninit_pf(struct rte_eth_dev *eth_dev)
 
 	hns3_enable_hw_error_intr(hns, false);
 	hns3_rss_uninit(hns);
+	hns3_promisc_uninit(hw);
 	hns3_fdir_filter_uninit(hns);
 	hns3_uninit_umv_space(hw);
 	hns3_pf_disable_irq0(hw);