Skip to contentSkip to conversation

Fixes for hns3 PMD driver

PATCH In review Wei Hu (Xavier) 2 participants 9 messages Nov 22, 2019 Network drivers

lore
Share & tools
lore
#0 lore raw

cover Fixes for hns3 PMD driver

From: "Wei Hu (Xavier)" <[email protected]>

This series add some fixes for hns3 PMD driver.

Chengchang Tang (1):
net/hns3: fix the error length limit of maiblox response

Hao Chen (1):
net/hns3: fix RSS hardware configuration restore failure

Huisong Li (1):
net/hns3: fix the strategy of getting link status for VF

Min Hu (Connor) (1):
net/hns3: fix duplicate VLAN

Wei Hu (Xavier) (2):
net/hns3: fix VF configuration table entries restore failure net/hns3: fix the failure sending packets less than 60 bytes

 drivers/net/hns3/hns3_ethdev.c    |  5 +++++
 drivers/net/hns3/hns3_ethdev.h    |  1 +
 drivers/net/hns3/hns3_ethdev_vf.c | 22 +++++++++++++++++-----
 drivers/net/hns3/hns3_mbx.c       |  4 +---
 drivers/net/hns3/hns3_mbx.h       |  6 ++++++
 drivers/net/hns3/hns3_rss.c       |  4 +---
 drivers/net/hns3/hns3_rxtx.c      | 24 ++++++++++++++++++++----
 7 files changed, 51 insertions(+), 15 deletions(-)
Signature · 1 line
2.23.0
#1 lore raw

PATCH 1/6 net/hns3: fix RSS hardware configuration restore failure

From: Hao Chen <[email protected]>

This patch fixes the bug that hardware configuration called tc_size doesn't restore to the initial value when starting the app, configuring PFC and then restarting the app, because of the tc_mode didn't initial when rss is disabled.

Fixes c37ca66f2b27 ("net/hns3: support RSS")
---
 drivers/net/hns3/hns3_rss.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/net/hns3/hns3_rss.c +1 -3
index 6a5d63398..b8c20e6d9 100644
@@ -525,10 +525,8 @@ hns3_config_rss(struct hns3_adapter *hns)
 	enum rte_eth_rx_mq_mode mq_mode = hw->data->dev_conf.rxmode.mq_mode;
 
 	/* When there is no open RSS, redirect the packet queue 0 */
-	if (((uint32_t)mq_mode & ETH_MQ_RX_RSS_FLAG) == 0) {
+	if (((uint32_t)mq_mode & ETH_MQ_RX_RSS_FLAG) == 0)
 		hns3_rss_uninit(hns);
-		return 0;
-	}
 
 	/* Configure RSS hash algorithm and hash key offset */
 	ret = hns3_set_rss_algo_key(hw, hash_algo, hash_key);
Signature · 1 line
2.23.0
#2 lore raw

PATCH 2/6 net/hns3: fix VF configuration table entries restore failure

In reply to Wei Hu (Xavier)

From: "Wei Hu (Xavier)" <[email protected]>

When the application using VF device exits abnormally, for example, when it is killed by 'kill -9', kernel PF netdev driver also stores the corresponding configuration table entries of VF device.

This patch fixes it by adding message of deleting VF configuration table entry corresponds to the revision of kernel hns3 netdev driver, the new message is added to notify the kernel PF netdev driver to clean up the VF configuration initialization during VF initialization.

This revision is compatible with the old version of kernel hns3 netdev driver. The old version of kernel pf netdev driver will ignore this message.

Fixes a5475d61fa34 ("net/hns3: support VF")
---
 drivers/net/hns3/hns3_ethdev_vf.c | 14 ++++++++++++++
 drivers/net/hns3/hns3_mbx.h       |  6 ++++++
 2 files changed, 20 insertions(+)
drivers/net/hns3/hns3_ethdev_vf.c +14 -0
index 403674969..2274ac35e 100644
@@ -1096,6 +1096,14 @@ hns3vf_init_hardware(struct hns3_adapter *hns)
 	return ret;
 }
 
+static int
+hns3vf_clear_vport_list(struct hns3_hw *hw)
+{
+	return hns3_send_mbx_msg(hw, HNS3_MBX_HANDLE_VF_TBL,
+				 HNS3_MBX_VPORT_LIST_CLEAR, NULL, 0, false,
+				 NULL, 0);
+}
+
 static int
 hns3vf_init_vf(struct rte_eth_dev *eth_dev)
 {
@@ -1147,6 +1155,12 @@ hns3vf_init_vf(struct rte_eth_dev *eth_dev)
 
 	rte_eth_random_addr(hw->mac.mac_addr); /* Generate a random mac addr */
 
+	ret = hns3vf_clear_vport_list(hw);
+	if (ret) {
+		PMD_INIT_LOG(ERR, "Failed to clear tbl list: %d", ret);
+		goto err_get_config;
+	}
+
 	ret = hns3vf_init_hardware(hns);
 	if (ret)
 		goto err_get_config;
drivers/net/hns3/hns3_mbx.h +6 -0
index ee6e82314..01eddb845 100644
@@ -39,6 +39,8 @@ enum HNS3_MBX_OPCODE {
 	HNS3_MBX_SET_ALIVE,             /* (VF -> PF) set alive state */
 	HNS3_MBX_SET_MTU,               /* (VF -> PF) set mtu */
 	HNS3_MBX_GET_QID_IN_PF,         /* (VF -> PF) get queue id in pf */
+
+	HNS3_MBX_HANDLE_VF_TBL = 38,    /* (VF -> PF) store/clear hw cfg tbl */
 };
 
 /* below are per-VF mac-vlan subcodes */
@@ -58,6 +60,10 @@ enum hns3_mbx_vlan_cfg_subcode {
 	HNS3_MBX_VLAN_RX_OFF_CFG,               /* set rx side vlan offload */
 };
 
+enum hns3_mbx_tbl_cfg_subcode {
+	HNS3_MBX_VPORT_LIST_CLEAR = 0,
+};
+
 #define HNS3_MBX_MAX_MSG_SIZE	16
 #define HNS3_MBX_MAX_RESP_DATA_SIZE	8
 #define HNS3_MBX_RING_MAP_BASIC_MSG_NUM	3
Signature · 1 line
2.23.0
#3 lore raw

Re: [dpdk-dev] [PATCH 2/6] net/hns3: fix VF configuration table entries restore failure

Quoted 16 lines · Wei Hu (Xavier)

From: "Wei Hu (Xavier)" <[email protected]>

When the application using VF device exits abnormally, for example, when it is killed by 'kill -9', kernel PF netdev driver also stores the corresponding configuration table entries of VF device.

This patch fixes it by adding message of deleting VF configuration table entry corresponds to the revision of kernel hns3 netdev driver, the new message is added to notify the kernel PF netdev driver to clean up the VF configuration initialization during VF initialization.

This revision is compatible with the old version of kernel hns3 netdev driver. The old version of kernel pf netdev driver will ignore this message.

Quoted 5 lines
Fixes a5475d61fa34 ("net/hns3: support VF")

Hi Xavier,

We are trying the use unique identifier for same person as much as possible, Above seems your personal email, I will update all occurrences as following, as how git history knows you:
Wei Hu (Xavier) <[email protected]>

Please let us know if there is an objection/concern on this changes.

#4 lore raw

PATCH 3/6 net/hns3: fix the failure sending packets less than 60 bytes

In reply to Wei Hu (Xavier)

From: "Wei Hu (Xavier)" <[email protected]>

Ethernet minimum packet length is 64 bytes. If upper application sends packets with less than 60 bytes in length(no CRC), driver adds padding processing to avoid failure.

Fixes bba636698316 ("net/hns3: support Rx/Tx and related operations")
---
 drivers/net/hns3/hns3_ethdev.h |  1 +
 drivers/net/hns3/hns3_rxtx.c   | 24 ++++++++++++++++++++----
 2 files changed, 21 insertions(+), 4 deletions(-)
drivers/net/hns3/hns3_ethdev.h +1 -0
index 9710e45fb..e9a3fe410 100644
@@ -43,6 +43,7 @@
 #define HNS3_MAX_MTU	(HNS3_MAX_FRAME_LEN - HNS3_ETH_OVERHEAD)
 #define HNS3_DEFAULT_MTU		1500UL
 #define HNS3_DEFAULT_FRAME_LEN		(HNS3_DEFAULT_MTU + HNS3_ETH_OVERHEAD)
+#define HNS3_MIN_PKT_SIZE		60
 
 #define HNS3_4_TCS			4
 #define HNS3_8_TCS			8
drivers/net/hns3/hns3_rxtx.c +20 -4
index 1de238b4c..34cb7faf9 100644
@@ -1598,13 +1598,29 @@ hns3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 		}
 
 		/*
-		 * If the length of the packet is too long or zero, the packet
-		 * will be ignored.
+		 * If packet length is greater than HNS3_MAX_FRAME_LEN
+		 * driver support, the packet will be ignored.
 		 */
-		if (unlikely(tx_pkt->pkt_len > HNS3_MAX_FRAME_LEN ||
-			     tx_pkt->pkt_len == 0))
+		if (unlikely(rte_pktmbuf_pkt_len(tx_pkt) > HNS3_MAX_FRAME_LEN))
 			break;
 
+		/*
+		 * If packet length is less than minimum packet size, driver
+		 * need to pad it.
+		 */
+		if (unlikely(rte_pktmbuf_pkt_len(tx_pkt) < HNS3_MIN_PKT_SIZE)) {
+			uint16_t add_len;
+			char *appended;
+
+			add_len = HNS3_MIN_PKT_SIZE -
+					 rte_pktmbuf_pkt_len(tx_pkt);
+			appended = rte_pktmbuf_append(tx_pkt, add_len);
+			if (appended == NULL)
+				break;
+
+			memset(appended, 0, add_len);
+		}
+
 		m_seg = tx_pkt;
 		if (unlikely(nb_buf > HNS3_MAX_TX_BD_PER_PKT)) {
 			if (hns3_reassemble_tx_pkts(txq, tx_pkt, &new_pkt))
Signature · 1 line
2.23.0
#5 lore raw

PATCH 4/6 net/hns3: fix the error length limit of maiblox response

In reply to Wei Hu (Xavier)

From: Chengchang Tang <[email protected]>

This patch removes the macro 'HNS3_REG_MSG_DATA_OFFSET' which is used to prevent the array from accessing violation and it limits the response data length to be 4. but the limit value is too short to get some longer information such as 6 byte MAC address.

This patch modify the length of response data from mailbox to allows the response data length to be 8. So that the VF driver could get more data from PF drvier by mailbox.

Fixes 463e748964f5 ("net/hns3: support mailbox")
---
 drivers/net/hns3/hns3_mbx.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/net/hns3/hns3_mbx.c +1 -3
index 2bfd97415..c1647af4b 100644
@@ -24,7 +24,6 @@
 #include "hns3_logs.h"
 #include "hns3_intr.h"
 
-#define HNS3_REG_MSG_DATA_OFFSET	4
 #define HNS3_CMD_CODE_OFFSET		2
 
 static const struct errno_respcode_map err_code_map[] = {
@@ -320,8 +319,7 @@ hns3_dev_handle_mbx_msg(struct hns3_hw *hw)
 			resp->resp_status = hns3_resp_to_errno(req->msg[3]);
 
 			temp = (uint8_t *)&req->msg[4];
-			for (i = 0; i < HNS3_MBX_MAX_RESP_DATA_SIZE &&
-			     i < HNS3_REG_MSG_DATA_OFFSET; i++) {
+			for (i = 0; i < HNS3_MBX_MAX_RESP_DATA_SIZE; i++) {
 				resp->additional_info[i] = *temp;
 				temp++;
 			}
Signature · 1 line
2.23.0
#6 lore raw

PATCH 5/6 net/hns3: fix the strategy of getting link status for VF

In reply to Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, port link status is "up" in VF driver after user calling the rte_eth_dev_stop API. This is unreasonable.

Therefore, this patch adjusts the strategy of getting link status from PF driver for VF. VF drvier should stop getting link status from PF by canceling the alarm that VF driver send mailbox message to PF driver, when the rte_eth_dev_stop API is called. And VF driver should restore the alarm when the rte_eth_dev_start API is called.

Fixes a5475d61fa34 ("net/hns3: support VF")
---
 drivers/net/hns3/hns3_ethdev_vf.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
drivers/net/hns3/hns3_ethdev_vf.c +3 -5
index 2274ac35e..b1736e73a 100644
@@ -1246,6 +1246,7 @@ hns3vf_dev_stop(struct rte_eth_dev *eth_dev)
 		hns3_dev_release_mbufs(hns);
 		hw->adapter_state = HNS3_NIC_CONFIGURED;
 	}
+	rte_eal_alarm_cancel(hns3vf_service_handler, eth_dev);
 	rte_spinlock_unlock(&hw->lock);
 }
 
@@ -1265,7 +1266,6 @@ hns3vf_dev_close(struct rte_eth_dev *eth_dev)
 	hns3_reset_abort(hns);
 	hw->adapter_state = HNS3_NIC_CLOSED;
 	rte_eal_alarm_cancel(hns3vf_keep_alive_handler, eth_dev);
-	rte_eal_alarm_cancel(hns3vf_service_handler, eth_dev);
 	hns3vf_configure_all_mc_mac_addr(hns, true);
 	hns3vf_remove_all_vlan_table(hns);
 	hns3vf_uninit_vf(eth_dev);
@@ -1286,8 +1286,6 @@ hns3vf_dev_link_update(struct rte_eth_dev *eth_dev,
 	struct hns3_mac *mac = &hw->mac;
 	struct rte_eth_link new_link;
 
-	hns3vf_request_link_info(hw);
-
 	memset(&new_link, 0, sizeof(new_link));
 	switch (mac->link_speed) {
 	case ETH_SPEED_NUM_10M:
@@ -1352,6 +1350,8 @@ hns3vf_dev_start(struct rte_eth_dev *eth_dev)
 	rte_spinlock_unlock(&hw->lock);
 	hns3_set_rxtx_function(eth_dev);
 	hns3_mp_req_start_rxtx(eth_dev);
+	rte_eal_alarm_set(HNS3VF_SERVICE_INTERVAL, hns3vf_service_handler,
+			  eth_dev);
 	return 0;
 }
 
@@ -1789,8 +1789,6 @@ hns3vf_dev_init(struct rte_eth_dev *eth_dev)
 	}
 	rte_eal_alarm_set(HNS3VF_KEEP_ALIVE_INTERVAL, hns3vf_keep_alive_handler,
 			  eth_dev);
-	rte_eal_alarm_set(HNS3VF_SERVICE_INTERVAL, hns3vf_service_handler,
-			  eth_dev);
 	return 0;
 
 err_rte_zmalloc:
Signature · 1 line
2.23.0
#7 lore raw

PATCH 6/6 net/hns3: fix duplicate VLAN

In reply to Wei Hu (Xavier)

From: "Min Hu (Connor)" <[email protected]>

When setting duplicate vlan, hns3 driver will also add vlan entry to vlan linked list, and this is unreasonable.

This patch adds checking whether the VLAN to be added already exists in the linked list and preventing adding duplicate vlan.

Fixes 411d23b9eafb ("net/hns3: support VLAN")
---
 drivers/net/hns3/hns3_ethdev.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/net/hns3/hns3_ethdev.c +5 -0
index 3435bce26..72315718a 100644
@@ -282,6 +282,11 @@ hns3_add_dev_vlan_table(struct hns3_adapter *hns, uint16_t vlan_id,
 	struct hns3_hw *hw = &hns->hw;
 	struct hns3_pf *pf = &hns->pf;
 
+	LIST_FOREACH(vlan_entry, &pf->vlan_list, next) {
+		if (vlan_entry->vlan_id == vlan_id)
+			return;
+	}
+
 	vlan_entry = rte_zmalloc("hns3_vlan_tbl", sizeof(*vlan_entry), 0);
 	if (vlan_entry == NULL) {
 		hns3_err(hw, "Failed to malloc hns3 vlan table");
Signature · 1 line
2.23.0
#8 lore raw
In reply to Wei Hu (Xavier)
Quoted 13 lines · Wei Hu (Xavier)

From: "Wei Hu (Xavier)" <[email protected]>

This series add some fixes for hns3 PMD driver.

Chengchang Tang (1):
net/hns3: fix the error length limit of maiblox response

Hao Chen (1):
net/hns3: fix RSS hardware configuration restore failure

Huisong Li (1):
net/hns3: fix the strategy of getting link status for VF

Quoted 6 lines

Min Hu (Connor) (1):
net/hns3: fix duplicate VLAN

Wei Hu (Xavier) (2):
net/hns3: fix VF configuration table entries restore failure net/hns3: fix the failure sending packets less than 60 bytes

Series applied to dpdk-next-net/master, thanks.