be2net: add multiple RX queue support

This patch adds multiple RX queue support to be2net. There are
upto 4 extra rx-queues per port into which TCP/UDP traffic can be hashed into.
Some of the ethtool stats are now displayed on a per queue basis.

Signed-off-by: Sathya Perla <sathya.perla@emulex.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/benet/be.h

@@ -78,6 +78,8 @@ static inline char *nic_name(struct pci_dev *pdev)
 #define MCC_Q_LEN		128	/* total size not to exceed 8 pages */
 #define MCC_CQ_LEN		256
 
+#define MAX_RSS_QS		4	/* BE limit is 4 queues/port */
+#define BE_MAX_MSIX_VECTORS	(MAX_RSS_QS + 1 + 1)/* RSS qs + 1 def Rx + Tx */
 #define BE_NAPI_WEIGHT		64
 #define MAX_RX_POST 		BE_NAPI_WEIGHT /* Frags posted at a time */
 #define RX_FRAGS_REFILL_WM	(RX_Q_LEN - MAX_RX_POST)
@@ -157,10 +159,9 @@ struct be_mcc_obj {
 	bool rearm_cq;
 };
 
-struct be_drvr_stats {
+struct be_tx_stats {
 	u32 be_tx_reqs;		/* number of TX requests initiated */
 	u32 be_tx_stops;	/* number of times TX Q was stopped */
-	u32 be_fwd_reqs;	/* number of send reqs through forwarding i/f */
 	u32 be_tx_wrbs;		/* number of tx WRBs used */
 	u32 be_tx_events;	/* number of tx completion events  */
 	u32 be_tx_compl;	/* number of tx completion entries processed */
@@ -169,35 +170,6 @@ struct be_drvr_stats {
 	u64 be_tx_bytes_prev;
 	u64 be_tx_pkts;
 	u32 be_tx_rate;
-
-	u32 cache_barrier[16];
-
-	u32 be_ethrx_post_fail;/* number of ethrx buffer alloc failures */
-	u32 be_rx_polls;	/* number of times NAPI called poll function */
-	u32 be_rx_events;	/* number of ucast rx completion events  */
-	u32 be_rx_compl;	/* number of rx completion entries processed */
-	ulong be_rx_jiffies;
-	u64 be_rx_bytes;
-	u64 be_rx_bytes_prev;
-	u64 be_rx_pkts;
-	u32 be_rx_rate;
-	u32 be_rx_mcast_pkt;
-	/* number of non ether type II frames dropped where
-	 * frame len > length field of Mac Hdr */
-	u32 be_802_3_dropped_frames;
-	/* number of non ether type II frames malformed where
-	 * in frame len < length field of Mac Hdr */
-	u32 be_802_3_malformed_frames;
-	u32 be_rxcp_err;	/* Num rx completion entries w/ err set. */
-	ulong rx_fps_jiffies;	/* jiffies at last FPS calc */
-	u32 be_rx_frags;
-	u32 be_prev_rx_frags;
-	u32 be_rx_fps;		/* Rx frags per second */
-};
-
-struct be_stats_obj {
-	struct be_drvr_stats drvr_stats;
-	struct be_dma_mem cmd;
 };
 
 struct be_tx_obj {
@@ -215,10 +187,34 @@ struct be_rx_page_info {
 	bool last_page_user;
 };
 
+struct be_rx_stats {
+	u32 rx_post_fail;/* number of ethrx buffer alloc failures */
+	u32 rx_polls;	/* number of times NAPI called poll function */
+	u32 rx_events;	/* number of ucast rx completion events  */
+	u32 rx_compl;	/* number of rx completion entries processed */
+	ulong rx_jiffies;
+	u64 rx_bytes;
+	u64 rx_bytes_prev;
+	u64 rx_pkts;
+	u32 rx_rate;
+	u32 rx_mcast_pkts;
+	u32 rxcp_err;	/* Num rx completion entries w/ err set. */
+	ulong rx_fps_jiffies;	/* jiffies at last FPS calc */
+	u32 rx_frags;
+	u32 prev_rx_frags;
+	u32 rx_fps;		/* Rx frags per second */
+};
+
 struct be_rx_obj {
+	struct be_adapter *adapter;
 	struct be_queue_info q;
 	struct be_queue_info cq;
 	struct be_rx_page_info page_info_tbl[RX_Q_LEN];
+	struct be_eq_obj rx_eq;
+	struct be_rx_stats stats;
+	u8 rss_id;
+	bool rx_post_starved;	/* Zero rx frags have been posted to BE */
+	u32 cache_line_barrier[16];
 };
 
 struct be_vf_cfg {
@@ -229,7 +225,6 @@ struct be_vf_cfg {
 	u32 vf_tx_rate;
 };
 
-#define BE_NUM_MSIX_VECTORS		2	/* 1 each for Tx and Rx */
 #define BE_INVALID_PMAC_ID		0xffffffff
 struct be_adapter {
 	struct pci_dev *pdev;
@@ -249,21 +244,21 @@ struct be_adapter {
 	spinlock_t mcc_lock;	/* For serializing mcc cmds to BE card */
 	spinlock_t mcc_cq_lock;
 
-	struct msix_entry msix_entries[BE_NUM_MSIX_VECTORS];
+	struct msix_entry msix_entries[BE_MAX_MSIX_VECTORS];
 	bool msix_enabled;
 	bool isr_registered;
 
 	/* TX Rings */
 	struct be_eq_obj tx_eq;
 	struct be_tx_obj tx_obj;
+	struct be_tx_stats tx_stats;
 
 	u32 cache_line_break[8];
 
 	/* Rx rings */
-	struct be_eq_obj rx_eq;
-	struct be_rx_obj rx_obj;
+	struct be_rx_obj rx_obj[MAX_RSS_QS + 1]; /* one default non-rss Q */
+	u32 num_rx_qs;
 	u32 big_page_size;	/* Compounded page size shared by rx wrbs */
-	bool rx_post_starved;	/* Zero rx frags have been posted to BE */
 
 	struct vlan_group *vlan_grp;
 	u16 vlans_added;
@@ -271,7 +266,7 @@ struct be_adapter {
 	u8 vlan_tag[VLAN_GROUP_ARRAY_LEN];
 	struct be_dma_mem mc_cmd_mem;
 
-	struct be_stats_obj stats;
+	struct be_dma_mem stats_cmd;
 	/* Work queue used to perform periodic tasks like getting statistics */
 	struct delayed_work work;
 
@@ -287,6 +282,7 @@ struct be_adapter {
 	bool promiscuous;
 	bool wol;
 	u32 function_mode;
+	u32 function_caps;
 	u32 rx_fc;		/* Rx flow control */
 	u32 tx_fc;		/* Tx flow control */
 	bool ue_detected;
@@ -313,10 +309,20 @@ struct be_adapter {
 
 extern const struct ethtool_ops be_ethtool_ops;
 
-#define drvr_stats(adapter)		(&adapter->stats.drvr_stats)
+#define tx_stats(adapter)		(&adapter->tx_stats)
+#define rx_stats(rxo)			(&rxo->stats)
 
 #define BE_SET_NETDEV_OPS(netdev, ops)	(netdev->netdev_ops = ops)
 
+#define for_all_rx_queues(adapter, rxo, i)				\
+	for (i = 0, rxo = &adapter->rx_obj[i]; i < adapter->num_rx_qs;	\
+		i++, rxo++)
+
+/* Just skip the first default non-rss queue */
+#define for_all_rss_queues(adapter, rxo, i)				\
+	for (i = 0, rxo = &adapter->rx_obj[i+1]; i < (adapter->num_rx_qs - 1);\
+		i++, rxo++)
+
 #define PAGE_SHIFT_4K		12
 #define PAGE_SIZE_4K		(1 << PAGE_SHIFT_4K)
 

drivers/net/benet/be_cmds.c

@@ -71,7 +71,7 @@ static int be_mcc_compl_process(struct be_adapter *adapter,
 	if (compl_status == MCC_STATUS_SUCCESS) {
 		if (compl->tag0 == OPCODE_ETH_GET_STATISTICS) {
 			struct be_cmd_resp_get_stats *resp =
-						adapter->stats.cmd.va;
+						adapter->stats_cmd.va;
 			be_dws_le_to_cpu(&resp->hw_stats,
 						sizeof(resp->hw_stats));
 			netdev_stats_update(adapter);
@@ -754,7 +754,7 @@ int be_cmd_txq_create(struct be_adapter *adapter,
 /* Uses mbox */
 int be_cmd_rxq_create(struct be_adapter *adapter,
 		struct be_queue_info *rxq, u16 cq_id, u16 frag_size,
-		u16 max_frame_size, u32 if_id, u32 rss)
+		u16 max_frame_size, u32 if_id, u32 rss, u8 *rss_id)
 {
 	struct be_mcc_wrb *wrb;
 	struct be_cmd_req_eth_rx_create *req;
@@ -785,6 +785,7 @@ int be_cmd_rxq_create(struct be_adapter *adapter,
 		struct be_cmd_resp_eth_rx_create *resp = embedded_payload(wrb);
 		rxq->id = le16_to_cpu(resp->id);
 		rxq->created = true;
+		*rss_id = resp->rss_id;
 	}
 
 	spin_unlock(&adapter->mbox_lock);
@@ -1259,7 +1260,8 @@ err:
 }
 
 /* Uses mbox */
-int be_cmd_query_fw_cfg(struct be_adapter *adapter, u32 *port_num, u32 *mode)
+int be_cmd_query_fw_cfg(struct be_adapter *adapter, u32 *port_num,
+		u32 *mode, u32 *caps)
 {
 	struct be_mcc_wrb *wrb;
 	struct be_cmd_req_query_fw_cfg *req;
@@ -1281,6 +1283,7 @@ int be_cmd_query_fw_cfg(struct be_adapter *adapter, u32 *port_num, u32 *mode)
 		struct be_cmd_resp_query_fw_cfg *resp = embedded_payload(wrb);
 		*port_num = le32_to_cpu(resp->phys_port);
 		*mode = le32_to_cpu(resp->function_mode);
+		*caps = le32_to_cpu(resp->function_caps);
 	}
 
 	spin_unlock(&adapter->mbox_lock);
@@ -1311,6 +1314,37 @@ int be_cmd_reset_function(struct be_adapter *adapter)
 	return status;
 }
 
+int be_cmd_rss_config(struct be_adapter *adapter, u8 *rsstable, u16 table_size)
+{
+	struct be_mcc_wrb *wrb;
+	struct be_cmd_req_rss_config *req;
+	u32 myhash[10];
+	int status;
+
+	spin_lock(&adapter->mbox_lock);
+
+	wrb = wrb_from_mbox(adapter);
+	req = embedded_payload(wrb);
+
+	be_wrb_hdr_prepare(wrb, sizeof(*req), true, 0,
+		OPCODE_ETH_RSS_CONFIG);
+
+	be_cmd_hdr_prepare(&req->hdr, CMD_SUBSYSTEM_ETH,
+		OPCODE_ETH_RSS_CONFIG, sizeof(*req));
+
+	req->if_id = cpu_to_le32(adapter->if_handle);
+	req->enable_rss = cpu_to_le16(RSS_ENABLE_TCP_IPV4 | RSS_ENABLE_IPV4);
+	req->cpu_table_size_log2 = cpu_to_le16(fls(table_size) - 1);
+	memcpy(req->cpu_table, rsstable, table_size);
+	memcpy(req->hash, myhash, sizeof(myhash));
+	be_dws_cpu_to_le(req->hash, sizeof(req->hash));
+
+	status = be_mbox_notify_wait(adapter);
+
+	spin_unlock(&adapter->mbox_lock);
+	return status;
+}
+
 /* Uses sync mcc */
 int be_cmd_set_beacon_state(struct be_adapter *adapter, u8 port_num,
 			u8 bcn, u8 sts, u8 state)

drivers/net/benet/be_cmds.h

@@ -147,6 +147,7 @@ struct be_mcc_mailbox {
 #define OPCODE_COMMON_READ_TRANSRECV_DATA		73
 #define OPCODE_COMMON_GET_PHY_DETAILS			102
 
+#define OPCODE_ETH_RSS_CONFIG				1
 #define OPCODE_ETH_ACPI_CONFIG				2
 #define OPCODE_ETH_PROMISCUOUS				3
 #define OPCODE_ETH_GET_STATISTICS			4
@@ -409,7 +410,7 @@ struct be_cmd_req_eth_rx_create {
 struct be_cmd_resp_eth_rx_create {
 	struct be_cmd_resp_hdr hdr;
 	u16 id;
-	u8 cpu_id;
+	u8 rss_id;
 	u8 rsvd0;
 } __packed;
 
@@ -739,9 +740,10 @@ struct be_cmd_resp_modify_eq_delay {
 } __packed;
 
 /******************** Get FW Config *******************/
+#define BE_FUNCTION_CAPS_RSS			0x2
 struct be_cmd_req_query_fw_cfg {
 	struct be_cmd_req_hdr hdr;
-	u32 rsvd[30];
+	u32 rsvd[31];
 };
 
 struct be_cmd_resp_query_fw_cfg {
@@ -751,6 +753,26 @@ struct be_cmd_resp_query_fw_cfg {
 	u32 phys_port;
 	u32 function_mode;
 	u32 rsvd[26];
+	u32 function_caps;
+};
+
+/******************** RSS Config *******************/
+/* RSS types */
+#define RSS_ENABLE_NONE				0x0
+#define RSS_ENABLE_IPV4				0x1
+#define RSS_ENABLE_TCP_IPV4			0x2
+#define RSS_ENABLE_IPV6				0x4
+#define RSS_ENABLE_TCP_IPV6			0x8
+
+struct be_cmd_req_rss_config {
+	struct be_cmd_req_hdr hdr;
+	u32 if_id;
+	u16 enable_rss;
+	u16 cpu_table_size_log2;
+	u32 hash[10];
+	u8 cpu_table[128];
+	u8 flush;
+	u8 rsvd0[3];
 };
 
 /******************** Port Beacon ***************************/
@@ -937,7 +959,7 @@ extern int be_cmd_txq_create(struct be_adapter *adapter,
 extern int be_cmd_rxq_create(struct be_adapter *adapter,
 			struct be_queue_info *rxq, u16 cq_id,
 			u16 frag_size, u16 max_frame_size, u32 if_id,
-			u32 rss);
+			u32 rss, u8 *rss_id);
 extern int be_cmd_q_destroy(struct be_adapter *adapter, struct be_queue_info *q,
 			int type);
 extern int be_cmd_link_status_query(struct be_adapter *adapter,
@@ -960,8 +982,10 @@ extern int be_cmd_set_flow_control(struct be_adapter *adapter,
 extern int be_cmd_get_flow_control(struct be_adapter *adapter,
 			u32 *tx_fc, u32 *rx_fc);
 extern int be_cmd_query_fw_cfg(struct be_adapter *adapter,
-			u32 *port_num, u32 *cap);
+			u32 *port_num, u32 *function_mode, u32 *function_caps);
 extern int be_cmd_reset_function(struct be_adapter *adapter);
+extern int be_cmd_rss_config(struct be_adapter *adapter, u8 *rsstable,
+			u16 table_size);
 extern int be_process_mcc(struct be_adapter *adapter, int *status);
 extern int be_cmd_set_beacon_state(struct be_adapter *adapter,
 			u8 port_num, u8 beacon, u8 status, u8 state);

drivers/net/benet/be_ethtool.c

@@ -26,14 +26,16 @@ struct be_ethtool_stat {
 	int offset;
 };
 
-enum {NETSTAT, PORTSTAT, MISCSTAT, DRVSTAT, ERXSTAT};
+enum {NETSTAT, PORTSTAT, MISCSTAT, DRVSTAT_TX, DRVSTAT_RX, ERXSTAT};
 #define FIELDINFO(_struct, field) FIELD_SIZEOF(_struct, field), \
 					offsetof(_struct, field)
 #define NETSTAT_INFO(field) 	#field, NETSTAT,\
 					FIELDINFO(struct net_device_stats,\
 						field)
-#define DRVSTAT_INFO(field) 	#field, DRVSTAT,\
-					FIELDINFO(struct be_drvr_stats, field)
+#define DRVSTAT_TX_INFO(field)	#field, DRVSTAT_TX,\
+					FIELDINFO(struct be_tx_stats, field)
+#define DRVSTAT_RX_INFO(field)	#field, DRVSTAT_RX,\
+					FIELDINFO(struct be_rx_stats, field)
 #define MISCSTAT_INFO(field) 	#field, MISCSTAT,\
 					FIELDINFO(struct be_rxf_stats, field)
 #define PORTSTAT_INFO(field) 	#field, PORTSTAT,\
@@ -51,21 +53,12 @@ static const struct be_ethtool_stat et_stats[] = {
 	{NETSTAT_INFO(tx_errors)},
 	{NETSTAT_INFO(rx_dropped)},
 	{NETSTAT_INFO(tx_dropped)},
-	{DRVSTAT_INFO(be_tx_reqs)},
-	{DRVSTAT_INFO(be_tx_stops)},
-	{DRVSTAT_INFO(be_fwd_reqs)},
-	{DRVSTAT_INFO(be_tx_wrbs)},
-	{DRVSTAT_INFO(be_rx_polls)},
-	{DRVSTAT_INFO(be_tx_events)},
-	{DRVSTAT_INFO(be_rx_events)},
-	{DRVSTAT_INFO(be_tx_compl)},
-	{DRVSTAT_INFO(be_rx_compl)},
-	{DRVSTAT_INFO(be_rx_mcast_pkt)},
-	{DRVSTAT_INFO(be_ethrx_post_fail)},
-	{DRVSTAT_INFO(be_802_3_dropped_frames)},
-	{DRVSTAT_INFO(be_802_3_malformed_frames)},
-	{DRVSTAT_INFO(be_tx_rate)},
-	{DRVSTAT_INFO(be_rx_rate)},
+	{DRVSTAT_TX_INFO(be_tx_rate)},
+	{DRVSTAT_TX_INFO(be_tx_reqs)},
+	{DRVSTAT_TX_INFO(be_tx_wrbs)},
+	{DRVSTAT_TX_INFO(be_tx_stops)},
+	{DRVSTAT_TX_INFO(be_tx_events)},
+	{DRVSTAT_TX_INFO(be_tx_compl)},
 	{PORTSTAT_INFO(rx_unicast_frames)},
 	{PORTSTAT_INFO(rx_multicast_frames)},
 	{PORTSTAT_INFO(rx_broadcast_frames)},
@@ -106,11 +99,24 @@ static const struct be_ethtool_stat et_stats[] = {
 	{MISCSTAT_INFO(rx_drops_too_many_frags)},
 	{MISCSTAT_INFO(rx_drops_invalid_ring)},
 	{MISCSTAT_INFO(forwarded_packets)},
-	{MISCSTAT_INFO(rx_drops_mtu)},
-	{ERXSTAT_INFO(rx_drops_no_fragments)},
+	{MISCSTAT_INFO(rx_drops_mtu)}
 };
 #define ETHTOOL_STATS_NUM ARRAY_SIZE(et_stats)
 
+/* Stats related to multi RX queues */
+static const struct be_ethtool_stat et_rx_stats[] = {
+	{DRVSTAT_RX_INFO(rx_bytes)},
+	{DRVSTAT_RX_INFO(rx_pkts)},
+	{DRVSTAT_RX_INFO(rx_rate)},
+	{DRVSTAT_RX_INFO(rx_polls)},
+	{DRVSTAT_RX_INFO(rx_events)},
+	{DRVSTAT_RX_INFO(rx_compl)},
+	{DRVSTAT_RX_INFO(rx_mcast_pkts)},
+	{DRVSTAT_RX_INFO(rx_post_fail)},
+	{ERXSTAT_INFO(rx_drops_no_fragments)}
+};
+#define ETHTOOL_RXSTATS_NUM (ARRAY_SIZE(et_rx_stats))
+
 static const char et_self_tests[][ETH_GSTRING_LEN] = {
 	"MAC Loopback test",
 	"PHY Loopback test",
@@ -143,7 +149,7 @@ static int
 be_get_coalesce(struct net_device *netdev, struct ethtool_coalesce *coalesce)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
-	struct be_eq_obj *rx_eq = &adapter->rx_eq;
+	struct be_eq_obj *rx_eq = &adapter->rx_obj[0].rx_eq;
 	struct be_eq_obj *tx_eq = &adapter->tx_eq;
 
 	coalesce->rx_coalesce_usecs = rx_eq->cur_eqd;
@@ -167,25 +173,49 @@ static int
 be_set_coalesce(struct net_device *netdev, struct ethtool_coalesce *coalesce)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
-	struct be_eq_obj *rx_eq = &adapter->rx_eq;
+	struct be_rx_obj *rxo;
+	struct be_eq_obj *rx_eq;
 	struct be_eq_obj *tx_eq = &adapter->tx_eq;
 	u32 tx_max, tx_min, tx_cur;
 	u32 rx_max, rx_min, rx_cur;
-	int status = 0;
+	int status = 0, i;
 
 	if (coalesce->use_adaptive_tx_coalesce == 1)
 		return -EINVAL;
 
-	/* if AIC is being turned on now, start with an EQD of 0 */
-	if (rx_eq->enable_aic == 0 &&
-		coalesce->use_adaptive_rx_coalesce == 1) {
-		rx_eq->cur_eqd = 0;
+	for_all_rx_queues(adapter, rxo, i) {
+		rx_eq = &rxo->rx_eq;
+
+		if (!rx_eq->enable_aic && coalesce->use_adaptive_rx_coalesce)
+			rx_eq->cur_eqd = 0;
+		rx_eq->enable_aic = coalesce->use_adaptive_rx_coalesce;
+
+		rx_max = coalesce->rx_coalesce_usecs_high;
+		rx_min = coalesce->rx_coalesce_usecs_low;
+		rx_cur = coalesce->rx_coalesce_usecs;
+
+		if (rx_eq->enable_aic) {
+			if (rx_max > BE_MAX_EQD)
+				rx_max = BE_MAX_EQD;
+			if (rx_min > rx_max)
+				rx_min = rx_max;
+			rx_eq->max_eqd = rx_max;
+			rx_eq->min_eqd = rx_min;
+			if (rx_eq->cur_eqd > rx_max)
+				rx_eq->cur_eqd = rx_max;
+			if (rx_eq->cur_eqd < rx_min)
+				rx_eq->cur_eqd = rx_min;
+		} else {
+			if (rx_cur > BE_MAX_EQD)
+				rx_cur = BE_MAX_EQD;
+			if (rx_eq->cur_eqd != rx_cur) {
+				status = be_cmd_modify_eqd(adapter, rx_eq->q.id,
+						rx_cur);
+				if (!status)
+					rx_eq->cur_eqd = rx_cur;
+			}
+		}
 	}
-	rx_eq->enable_aic = coalesce->use_adaptive_rx_coalesce;
-
-	rx_max = coalesce->rx_coalesce_usecs_high;
-	rx_min = coalesce->rx_coalesce_usecs_low;
-	rx_cur = coalesce->rx_coalesce_usecs;
 
 	tx_max = coalesce->tx_coalesce_usecs_high;
 	tx_min = coalesce->tx_coalesce_usecs_low;
@@ -199,27 +229,6 @@ be_set_coalesce(struct net_device *netdev, struct ethtool_coalesce *coalesce)
 			tx_eq->cur_eqd = tx_cur;
 	}
 
-	if (rx_eq->enable_aic) {
-		if (rx_max > BE_MAX_EQD)
-			rx_max = BE_MAX_EQD;
-		if (rx_min > rx_max)
-			rx_min = rx_max;
-		rx_eq->max_eqd = rx_max;
-		rx_eq->min_eqd = rx_min;
-		if (rx_eq->cur_eqd > rx_max)
-			rx_eq->cur_eqd = rx_max;
-		if (rx_eq->cur_eqd < rx_min)
-			rx_eq->cur_eqd = rx_min;
-	} else {
-		if (rx_cur > BE_MAX_EQD)
-			rx_cur = BE_MAX_EQD;
-		if (rx_eq->cur_eqd != rx_cur) {
-			status = be_cmd_modify_eqd(adapter, rx_eq->q.id,
-					rx_cur);
-			if (!status)
-				rx_eq->cur_eqd = rx_cur;
-		}
-	}
 	return 0;
 }
 
@@ -247,32 +256,25 @@ be_get_ethtool_stats(struct net_device *netdev,
 		struct ethtool_stats *stats, uint64_t *data)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
-	struct be_drvr_stats *drvr_stats = &adapter->stats.drvr_stats;
-	struct be_hw_stats *hw_stats = hw_stats_from_cmd(adapter->stats.cmd.va);
-	struct be_rxf_stats *rxf_stats = &hw_stats->rxf;
-	struct be_port_rxf_stats *port_stats =
-			&rxf_stats->port[adapter->port_num];
-	struct net_device_stats *net_stats = &netdev->stats;
+	struct be_hw_stats *hw_stats = hw_stats_from_cmd(adapter->stats_cmd.va);
 	struct be_erx_stats *erx_stats = &hw_stats->erx;
+	struct be_rx_obj *rxo;
 	void *p = NULL;
-	int i;
+	int i, j;
 
 	for (i = 0; i < ETHTOOL_STATS_NUM; i++) {
 		switch (et_stats[i].type) {
 		case NETSTAT:
-			p = net_stats;
+			p = &netdev->stats;
 			break;
-		case DRVSTAT:
-			p = drvr_stats;
+		case DRVSTAT_TX:
+			p = &adapter->tx_stats;
 			break;
 		case PORTSTAT:
-			p = port_stats;
+			p = &hw_stats->rxf.port[adapter->port_num];
 			break;
 		case MISCSTAT:
-			p = rxf_stats;
-			break;
-		case ERXSTAT: /* Currently only one ERX stat is provided */
-			p = (u32 *)erx_stats + adapter->rx_obj.q.id;
+			p = &hw_stats->rxf;
 			break;
 		}
 
@@ -280,19 +282,44 @@ be_get_ethtool_stats(struct net_device *netdev,
 		data[i] = (et_stats[i].size == sizeof(u64)) ?
 				*(u64 *)p: *(u32 *)p;
 	}
+
+	for_all_rx_queues(adapter, rxo, j) {
+		for (i = 0; i < ETHTOOL_RXSTATS_NUM; i++) {
+			switch (et_rx_stats[i].type) {
+			case DRVSTAT_RX:
+				p = (u8 *)&rxo->stats + et_rx_stats[i].offset;
+				break;
+			case ERXSTAT:
+				p = (u32 *)erx_stats + rxo->q.id;
+				break;
+			}
+			data[ETHTOOL_STATS_NUM + j * ETHTOOL_RXSTATS_NUM + i] =
+				(et_rx_stats[i].size == sizeof(u64)) ?
+					*(u64 *)p: *(u32 *)p;
+		}
+	}
 }
 
 static void
 be_get_stat_strings(struct net_device *netdev, uint32_t stringset,
 		uint8_t *data)
 {
-	int i;
+	struct be_adapter *adapter = netdev_priv(netdev);
+	int i, j;
+
 	switch (stringset) {
 	case ETH_SS_STATS:
 		for (i = 0; i < ETHTOOL_STATS_NUM; i++) {
 			memcpy(data, et_stats[i].desc, ETH_GSTRING_LEN);
 			data += ETH_GSTRING_LEN;
 		}
+		for (i = 0; i < adapter->num_rx_qs; i++) {
+			for (j = 0; j < ETHTOOL_RXSTATS_NUM; j++) {
+				sprintf(data, "rxq%d: %s", i,
+					et_rx_stats[j].desc);
+				data += ETH_GSTRING_LEN;
+			}
+		}
 		break;
 	case ETH_SS_TEST:
 		for (i = 0; i < ETHTOOL_TESTS_NUM; i++) {
@@ -305,11 +332,14 @@ be_get_stat_strings(struct net_device *netdev, uint32_t stringset,
 
 static int be_get_sset_count(struct net_device *netdev, int stringset)
 {
+	struct be_adapter *adapter = netdev_priv(netdev);
+
 	switch (stringset) {
 	case ETH_SS_TEST:
 		return ETHTOOL_TESTS_NUM;
 	case ETH_SS_STATS:
-		return ETHTOOL_STATS_NUM;
+		return ETHTOOL_STATS_NUM +
+			adapter->num_rx_qs * ETHTOOL_RXSTATS_NUM;
 	default:
 		return -EINVAL;
 	}
@@ -424,10 +454,10 @@ be_get_ringparam(struct net_device *netdev, struct ethtool_ringparam *ring)
 {
 	struct be_adapter *adapter = netdev_priv(netdev);
 
-	ring->rx_max_pending = adapter->rx_obj.q.len;
+	ring->rx_max_pending = adapter->rx_obj[0].q.len;
 	ring->tx_max_pending = adapter->tx_obj.q.len;
 
-	ring->rx_pending = atomic_read(&adapter->rx_obj.q.used);
+	ring->rx_pending = atomic_read(&adapter->rx_obj[0].q.used);
 	ring->tx_pending = atomic_read(&adapter->tx_obj.q.used);
 }