Merge branch 'qed-arfs'

Manish Chopra says:

====================
qed/qede: aRFS support

This series adds support for Accelerated Flow Steering
in qede driver for TCP/UDP over IPv4/IPv6 protocols.

Please consider applying this series to "net-next"
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/qlogic/qed/qed.h

@@ -504,6 +504,8 @@ struct qed_hwfn {
 	u8 dcbx_no_edpm;
 	u8 db_bar_no_edpm;
 
+	struct qed_ptt *p_arfs_ptt;
+
 	/* p_ptp_ptt is valid for leading HWFN only */
 	struct qed_ptt *p_ptp_ptt;
 	struct qed_simd_fp_handler	simd_proto_handler[64];

drivers/net/ethernet/qlogic/qed/qed_cxt.c

@@ -219,9 +219,6 @@ struct qed_cxt_mngr {
 	 */
 	u32				vf_count;
 
-	/* total number of SRQ's for this hwfn */
-	u32 srq_count;
-
 	/* Acquired CIDs */
 	struct qed_cid_acquired_map	acquired[MAX_CONN_TYPES];
 
@@ -237,6 +234,12 @@ struct qed_cxt_mngr {
 	u32 t2_num_pages;
 	u64 first_free;
 	u64 last_free;
+
+	/* total number of SRQ's for this hwfn */
+	u32 srq_count;
+
+	/* Maximal number of L2 steering filters */
+	u32 arfs_count;
 };
 static bool src_proto(enum protocol_type type)
 {
@@ -291,6 +294,9 @@ static void qed_cxt_src_iids(struct qed_cxt_mngr *p_mngr,
 		iids->pf_cids += p_mngr->conn_cfg[i].cid_count;
 		iids->per_vf_cids += p_mngr->conn_cfg[i].cids_per_vf;
 	}
+
+	/* Add L2 filtering filters in addition */
+	iids->pf_cids += p_mngr->arfs_count;
 }
 
 /* counts the iids for the Timers block configuration */
@@ -2007,6 +2013,7 @@ int qed_cxt_set_pf_params(struct qed_hwfn *p_hwfn, u32 rdma_tasks)
 
 		qed_cxt_set_proto_cid_count(p_hwfn, PROTOCOLID_ETH,
 					    p_params->num_cons, 1);
+		p_hwfn->p_cxt_mngr->arfs_count = p_params->num_arfs_filters;
 		break;
 	}
 	case QED_PCI_FCOE:

drivers/net/ethernet/qlogic/qed/qed_hsi.h

@@ -3473,6 +3473,11 @@ void qed_set_geneve_dest_port(struct qed_hwfn *p_hwfn,
 void qed_set_geneve_enable(struct qed_hwfn *p_hwfn,
 			   struct qed_ptt *p_ptt,
 			   bool eth_geneve_enable, bool ip_geneve_enable);
+void qed_set_rfs_mode_disable(struct qed_hwfn *p_hwfn,
+			      struct qed_ptt *p_ptt, u16 pf_id);
+void qed_set_rfs_mode_enable(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+			     u16 pf_id, bool tcp, bool udp,
+			     bool ipv4, bool ipv6);
 
 #define	YSTORM_FLOW_CONTROL_MODE_OFFSET			(IRO[0].base)
 #define	YSTORM_FLOW_CONTROL_MODE_SIZE			(IRO[0].size)
@@ -4862,6 +4867,18 @@ struct eth_vport_tx_mode {
 	__le16 reserved2[3];
 };
 
+enum gft_filter_update_action {
+	GFT_ADD_FILTER,
+	GFT_DELETE_FILTER,
+	MAX_GFT_FILTER_UPDATE_ACTION
+};
+
+enum gft_logic_filter_type {
+	GFT_FILTER_TYPE,
+	RFS_FILTER_TYPE,
+	MAX_GFT_LOGIC_FILTER_TYPE
+};
+
 /* Ramrod data for rx queue start ramrod */
 struct rx_queue_start_ramrod_data {
 	__le16 rx_queue_id;
@@ -4932,6 +4949,16 @@ struct rx_udp_filter_data {
 	__le32 tenant_id;
 };
 
+struct rx_update_gft_filter_data {
+	struct regpair pkt_hdr_addr;
+	__le16 pkt_hdr_length;
+	__le16 rx_qid_or_action_icid;
+	u8 vport_id;
+	u8 filter_type;
+	u8 filter_action;
+	u8 reserved;
+};
+
 /* Ramrod data for rx queue start ramrod */
 struct tx_queue_start_ramrod_data {
 	__le16 sb_id;
@@ -5075,6 +5102,166 @@ struct vport_update_ramrod_data {
 	struct eth_vport_rss_config rss_config;
 };
 
+struct gft_cam_line {
+	__le32 camline;
+#define GFT_CAM_LINE_VALID_MASK		0x1
+#define GFT_CAM_LINE_VALID_SHIFT	0
+#define GFT_CAM_LINE_DATA_MASK		0x3FFF
+#define GFT_CAM_LINE_DATA_SHIFT		1
+#define GFT_CAM_LINE_MASK_BITS_MASK	0x3FFF
+#define GFT_CAM_LINE_MASK_BITS_SHIFT	15
+#define GFT_CAM_LINE_RESERVED1_MASK	0x7
+#define GFT_CAM_LINE_RESERVED1_SHIFT	29
+};
+
+struct gft_cam_line_mapped {
+	__le32 camline;
+#define GFT_CAM_LINE_MAPPED_VALID_MASK				0x1
+#define GFT_CAM_LINE_MAPPED_VALID_SHIFT				0
+#define GFT_CAM_LINE_MAPPED_IP_VERSION_MASK			0x1
+#define GFT_CAM_LINE_MAPPED_IP_VERSION_SHIFT			1
+#define GFT_CAM_LINE_MAPPED_TUNNEL_IP_VERSION_MASK		0x1
+#define GFT_CAM_LINE_MAPPED_TUNNEL_IP_VERSION_SHIFT		2
+#define GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE_MASK		0xF
+#define GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE_SHIFT		3
+#define GFT_CAM_LINE_MAPPED_TUNNEL_TYPE_MASK			0xF
+#define GFT_CAM_LINE_MAPPED_TUNNEL_TYPE_SHIFT			7
+#define GFT_CAM_LINE_MAPPED_PF_ID_MASK				0xF
+#define GFT_CAM_LINE_MAPPED_PF_ID_SHIFT				11
+#define GFT_CAM_LINE_MAPPED_IP_VERSION_MASK_MASK		0x1
+#define GFT_CAM_LINE_MAPPED_IP_VERSION_MASK_SHIFT		15
+#define GFT_CAM_LINE_MAPPED_TUNNEL_IP_VERSION_MASK_MASK		0x1
+#define GFT_CAM_LINE_MAPPED_TUNNEL_IP_VERSION_MASK_SHIFT	16
+#define GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE_MASK_MASK	0xF
+#define GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE_MASK_SHIFT	17
+#define GFT_CAM_LINE_MAPPED_TUNNEL_TYPE_MASK_MASK		0xF
+#define GFT_CAM_LINE_MAPPED_TUNNEL_TYPE_MASK_SHIFT		21
+#define GFT_CAM_LINE_MAPPED_PF_ID_MASK_MASK			0xF
+#define GFT_CAM_LINE_MAPPED_PF_ID_MASK_SHIFT			25
+#define GFT_CAM_LINE_MAPPED_RESERVED1_MASK			0x7
+#define GFT_CAM_LINE_MAPPED_RESERVED1_SHIFT			29
+};
+
+union gft_cam_line_union {
+	struct gft_cam_line cam_line;
+	struct gft_cam_line_mapped cam_line_mapped;
+};
+
+enum gft_profile_ip_version {
+	GFT_PROFILE_IPV4 = 0,
+	GFT_PROFILE_IPV6 = 1,
+	MAX_GFT_PROFILE_IP_VERSION
+};
+
+enum gft_profile_upper_protocol_type {
+	GFT_PROFILE_ROCE_PROTOCOL = 0,
+	GFT_PROFILE_RROCE_PROTOCOL = 1,
+	GFT_PROFILE_FCOE_PROTOCOL = 2,
+	GFT_PROFILE_ICMP_PROTOCOL = 3,
+	GFT_PROFILE_ARP_PROTOCOL = 4,
+	GFT_PROFILE_USER_TCP_SRC_PORT_1_INNER = 5,
+	GFT_PROFILE_USER_TCP_DST_PORT_1_INNER = 6,
+	GFT_PROFILE_TCP_PROTOCOL = 7,
+	GFT_PROFILE_USER_UDP_DST_PORT_1_INNER = 8,
+	GFT_PROFILE_USER_UDP_DST_PORT_2_OUTER = 9,
+	GFT_PROFILE_UDP_PROTOCOL = 10,
+	GFT_PROFILE_USER_IP_1_INNER = 11,
+	GFT_PROFILE_USER_IP_2_OUTER = 12,
+	GFT_PROFILE_USER_ETH_1_INNER = 13,
+	GFT_PROFILE_USER_ETH_2_OUTER = 14,
+	GFT_PROFILE_RAW = 15,
+	MAX_GFT_PROFILE_UPPER_PROTOCOL_TYPE
+};
+
+struct gft_ram_line {
+	__le32 low32bits;
+#define GFT_RAM_LINE_VLAN_SELECT_MASK			0x3
+#define GFT_RAM_LINE_VLAN_SELECT_SHIFT			0
+#define GFT_RAM_LINE_TUNNEL_ENTROPHY_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_ENTROPHY_SHIFT		2
+#define GFT_RAM_LINE_TUNNEL_TTL_EQUAL_ONE_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_TTL_EQUAL_ONE_SHIFT		3
+#define GFT_RAM_LINE_TUNNEL_TTL_MASK			0x1
+#define GFT_RAM_LINE_TUNNEL_TTL_SHIFT			4
+#define GFT_RAM_LINE_TUNNEL_ETHERTYPE_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_ETHERTYPE_SHIFT		5
+#define GFT_RAM_LINE_TUNNEL_DST_PORT_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_DST_PORT_SHIFT		6
+#define GFT_RAM_LINE_TUNNEL_SRC_PORT_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_SRC_PORT_SHIFT		7
+#define GFT_RAM_LINE_TUNNEL_DSCP_MASK			0x1
+#define GFT_RAM_LINE_TUNNEL_DSCP_SHIFT			8
+#define GFT_RAM_LINE_TUNNEL_OVER_IP_PROTOCOL_MASK	0x1
+#define GFT_RAM_LINE_TUNNEL_OVER_IP_PROTOCOL_SHIFT	9
+#define GFT_RAM_LINE_TUNNEL_DST_IP_MASK			0x1
+#define GFT_RAM_LINE_TUNNEL_DST_IP_SHIFT		10
+#define GFT_RAM_LINE_TUNNEL_SRC_IP_MASK			0x1
+#define GFT_RAM_LINE_TUNNEL_SRC_IP_SHIFT		11
+#define GFT_RAM_LINE_TUNNEL_PRIORITY_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_PRIORITY_SHIFT		12
+#define GFT_RAM_LINE_TUNNEL_PROVIDER_VLAN_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_PROVIDER_VLAN_SHIFT		13
+#define GFT_RAM_LINE_TUNNEL_VLAN_MASK			0x1
+#define GFT_RAM_LINE_TUNNEL_VLAN_SHIFT			14
+#define GFT_RAM_LINE_TUNNEL_DST_MAC_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_DST_MAC_SHIFT		15
+#define GFT_RAM_LINE_TUNNEL_SRC_MAC_MASK		0x1
+#define GFT_RAM_LINE_TUNNEL_SRC_MAC_SHIFT		16
+#define GFT_RAM_LINE_TTL_EQUAL_ONE_MASK			0x1
+#define GFT_RAM_LINE_TTL_EQUAL_ONE_SHIFT		17
+#define GFT_RAM_LINE_TTL_MASK				0x1
+#define GFT_RAM_LINE_TTL_SHIFT				18
+#define GFT_RAM_LINE_ETHERTYPE_MASK			0x1
+#define GFT_RAM_LINE_ETHERTYPE_SHIFT			19
+#define GFT_RAM_LINE_RESERVED0_MASK			0x1
+#define GFT_RAM_LINE_RESERVED0_SHIFT			20
+#define GFT_RAM_LINE_TCP_FLAG_FIN_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_FIN_SHIFT			21
+#define GFT_RAM_LINE_TCP_FLAG_SYN_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_SYN_SHIFT			22
+#define GFT_RAM_LINE_TCP_FLAG_RST_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_RST_SHIFT			23
+#define GFT_RAM_LINE_TCP_FLAG_PSH_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_PSH_SHIFT			24
+#define GFT_RAM_LINE_TCP_FLAG_ACK_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_ACK_SHIFT			25
+#define GFT_RAM_LINE_TCP_FLAG_URG_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_URG_SHIFT			26
+#define GFT_RAM_LINE_TCP_FLAG_ECE_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_ECE_SHIFT			27
+#define GFT_RAM_LINE_TCP_FLAG_CWR_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_CWR_SHIFT			28
+#define GFT_RAM_LINE_TCP_FLAG_NS_MASK			0x1
+#define GFT_RAM_LINE_TCP_FLAG_NS_SHIFT			29
+#define GFT_RAM_LINE_DST_PORT_MASK			0x1
+#define GFT_RAM_LINE_DST_PORT_SHIFT			30
+#define GFT_RAM_LINE_SRC_PORT_MASK			0x1
+#define GFT_RAM_LINE_SRC_PORT_SHIFT			31
+	__le32 high32bits;
+#define GFT_RAM_LINE_DSCP_MASK				0x1
+#define GFT_RAM_LINE_DSCP_SHIFT				0
+#define GFT_RAM_LINE_OVER_IP_PROTOCOL_MASK		0x1
+#define GFT_RAM_LINE_OVER_IP_PROTOCOL_SHIFT		1
+#define GFT_RAM_LINE_DST_IP_MASK			0x1
+#define GFT_RAM_LINE_DST_IP_SHIFT			2
+#define GFT_RAM_LINE_SRC_IP_MASK			0x1
+#define GFT_RAM_LINE_SRC_IP_SHIFT			3
+#define GFT_RAM_LINE_PRIORITY_MASK			0x1
+#define GFT_RAM_LINE_PRIORITY_SHIFT			4
+#define GFT_RAM_LINE_PROVIDER_VLAN_MASK			0x1
+#define GFT_RAM_LINE_PROVIDER_VLAN_SHIFT		5
+#define GFT_RAM_LINE_VLAN_MASK				0x1
+#define GFT_RAM_LINE_VLAN_SHIFT				6
+#define GFT_RAM_LINE_DST_MAC_MASK			0x1
+#define GFT_RAM_LINE_DST_MAC_SHIFT			7
+#define GFT_RAM_LINE_SRC_MAC_MASK			0x1
+#define GFT_RAM_LINE_SRC_MAC_SHIFT			8
+#define GFT_RAM_LINE_TENANT_ID_MASK			0x1
+#define GFT_RAM_LINE_TENANT_ID_SHIFT			9
+#define GFT_RAM_LINE_RESERVED1_MASK			0x3FFFFF
+#define GFT_RAM_LINE_RESERVED1_SHIFT			10
+};
+
 struct mstorm_eth_conn_ag_ctx {
 	u8 byte0;
 	u8 byte1;

drivers/net/ethernet/qlogic/qed/qed_init_fw_funcs.c

@@ -961,3 +961,132 @@ void qed_set_geneve_enable(struct qed_hwfn *p_hwfn,
 	qed_wr(p_hwfn, p_ptt, DORQ_REG_L2_EDPM_TUNNEL_NGE_IP_EN,
 	       ip_geneve_enable ? 1 : 0);
 }
+
+#define T_ETH_PACKET_MATCH_RFS_EVENTID 25
+#define PARSER_ETH_CONN_CM_HDR (0x0)
+#define CAM_LINE_SIZE sizeof(u32)
+#define RAM_LINE_SIZE sizeof(u64)
+#define REG_SIZE sizeof(u32)
+
+void qed_set_rfs_mode_disable(struct qed_hwfn *p_hwfn,
+			      struct qed_ptt *p_ptt, u16 pf_id)
+{
+	union gft_cam_line_union camline;
+	struct gft_ram_line ramline;
+	u32 *p_ramline, i;
+
+	p_ramline = (u32 *)&ramline;
+
+	/*stop using gft logic */
+	qed_wr(p_hwfn, p_ptt, PRS_REG_SEARCH_GFT, 0);
+	qed_wr(p_hwfn, p_ptt, PRS_REG_CM_HDR_GFT, 0x0);
+	memset(&camline, 0, sizeof(union gft_cam_line_union));
+	qed_wr(p_hwfn, p_ptt, PRS_REG_GFT_CAM + CAM_LINE_SIZE * pf_id,
+	       camline.cam_line_mapped.camline);
+	memset(&ramline, 0, sizeof(union gft_cam_line_union));
+
+	for (i = 0; i < RAM_LINE_SIZE / REG_SIZE; i++) {
+		u32 hw_addr = PRS_REG_GFT_PROFILE_MASK_RAM;
+
+		hw_addr += (RAM_LINE_SIZE * pf_id + i * REG_SIZE);
+
+		qed_wr(p_hwfn, p_ptt, hw_addr, *(p_ramline + i));
+	}
+}
+
+void qed_set_rfs_mode_enable(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+			     u16 pf_id, bool tcp, bool udp,
+			     bool ipv4, bool ipv6)
+{
+	u32 rfs_cm_hdr_event_id, *p_ramline;
+	union gft_cam_line_union camline;
+	struct gft_ram_line ramline;
+	int i;
+
+	rfs_cm_hdr_event_id = qed_rd(p_hwfn, p_ptt, PRS_REG_CM_HDR_GFT);
+	p_ramline = (u32 *)&ramline;
+
+	if (!ipv6 && !ipv4)
+		DP_NOTICE(p_hwfn,
+			  "set_rfs_mode_enable: must accept at least on of - ipv4 or ipv6");
+	if (!tcp && !udp)
+		DP_NOTICE(p_hwfn,
+			  "set_rfs_mode_enable: must accept at least on of - udp or tcp");
+
+	rfs_cm_hdr_event_id |= T_ETH_PACKET_MATCH_RFS_EVENTID <<
+					PRS_REG_CM_HDR_GFT_EVENT_ID_SHIFT;
+	rfs_cm_hdr_event_id |= PARSER_ETH_CONN_CM_HDR <<
+					PRS_REG_CM_HDR_GFT_CM_HDR_SHIFT;
+	qed_wr(p_hwfn, p_ptt, PRS_REG_CM_HDR_GFT, rfs_cm_hdr_event_id);
+
+	/* Configure Registers for RFS mode */
+	qed_wr(p_hwfn, p_ptt, PRS_REG_SEARCH_GFT, 1);
+	qed_wr(p_hwfn, p_ptt, PRS_REG_LOAD_L2_FILTER, 0);
+	camline.cam_line_mapped.camline = 0;
+
+	/* cam line is now valid!! */
+	SET_FIELD(camline.cam_line_mapped.camline,
+		  GFT_CAM_LINE_MAPPED_VALID, 1);
+
+	/* filters are per PF!! */
+	SET_FIELD(camline.cam_line_mapped.camline,
+		  GFT_CAM_LINE_MAPPED_PF_ID_MASK, 1);
+	SET_FIELD(camline.cam_line_mapped.camline,
+		  GFT_CAM_LINE_MAPPED_PF_ID, pf_id);
+	if (!(tcp && udp)) {
+		SET_FIELD(camline.cam_line_mapped.camline,
+			  GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE_MASK, 1);
+		if (tcp)
+			SET_FIELD(camline.cam_line_mapped.camline,
+				  GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE,
+				  GFT_PROFILE_TCP_PROTOCOL);
+		else
+			SET_FIELD(camline.cam_line_mapped.camline,
+				  GFT_CAM_LINE_MAPPED_UPPER_PROTOCOL_TYPE,
+				  GFT_PROFILE_UDP_PROTOCOL);
+	}
+
+	if (!(ipv4 && ipv6)) {
+		SET_FIELD(camline.cam_line_mapped.camline,
+			  GFT_CAM_LINE_MAPPED_IP_VERSION_MASK, 1);
+		if (ipv4)
+			SET_FIELD(camline.cam_line_mapped.camline,
+				  GFT_CAM_LINE_MAPPED_IP_VERSION,
+				  GFT_PROFILE_IPV4);
+		else
+			SET_FIELD(camline.cam_line_mapped.camline,
+				  GFT_CAM_LINE_MAPPED_IP_VERSION,
+				  GFT_PROFILE_IPV6);
+	}
+
+	/* write characteristics to cam */
+	qed_wr(p_hwfn, p_ptt, PRS_REG_GFT_CAM + CAM_LINE_SIZE * pf_id,
+	       camline.cam_line_mapped.camline);
+	camline.cam_line_mapped.camline = qed_rd(p_hwfn, p_ptt,
+						 PRS_REG_GFT_CAM +
+						 CAM_LINE_SIZE * pf_id);
+
+	/* write line to RAM - compare to filter 4 tuple */
+	ramline.low32bits = 0;
+	ramline.high32bits = 0;
+	SET_FIELD(ramline.high32bits, GFT_RAM_LINE_DST_IP, 1);
+	SET_FIELD(ramline.high32bits, GFT_RAM_LINE_SRC_IP, 1);
+	SET_FIELD(ramline.low32bits, GFT_RAM_LINE_SRC_PORT, 1);
+	SET_FIELD(ramline.low32bits, GFT_RAM_LINE_DST_PORT, 1);
+
+	/* each iteration write to reg */
+	for (i = 0; i < RAM_LINE_SIZE / REG_SIZE; i++)
+		qed_wr(p_hwfn, p_ptt,
+		       PRS_REG_GFT_PROFILE_MASK_RAM + RAM_LINE_SIZE * pf_id +
+		       i * REG_SIZE, *(p_ramline + i));
+
+	/* set default profile so that no filter match will happen */
+	ramline.low32bits = 0xffff;
+	ramline.high32bits = 0xffff;
+
+	for (i = 0; i < RAM_LINE_SIZE / REG_SIZE; i++)
+		qed_wr(p_hwfn, p_ptt,
+		       PRS_REG_GFT_PROFILE_MASK_RAM + RAM_LINE_SIZE *
+		       PRS_GFT_CAM_LINES_NO_MATCH + i * REG_SIZE,
+		       *(p_ramline + i));
+}

drivers/net/ethernet/qlogic/qed/qed_l2.c

@@ -1799,6 +1799,84 @@ void qed_reset_vport_stats(struct qed_dev *cdev)
 		_qed_get_vport_stats(cdev, cdev->reset_stats);
 }
 
+static void
+qed_arfs_mode_configure(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+			struct qed_arfs_config_params *p_cfg_params)
+{
+	if (p_cfg_params->arfs_enable) {
+		qed_set_rfs_mode_enable(p_hwfn, p_ptt, p_hwfn->rel_pf_id,
+					p_cfg_params->tcp, p_cfg_params->udp,
+					p_cfg_params->ipv4, p_cfg_params->ipv6);
+		DP_VERBOSE(p_hwfn, QED_MSG_SP,
+			   "tcp = %s, udp = %s, ipv4 = %s, ipv6 =%s\n",
+			   p_cfg_params->tcp ? "Enable" : "Disable",
+			   p_cfg_params->udp ? "Enable" : "Disable",
+			   p_cfg_params->ipv4 ? "Enable" : "Disable",
+			   p_cfg_params->ipv6 ? "Enable" : "Disable");
+	} else {
+		qed_set_rfs_mode_disable(p_hwfn, p_ptt, p_hwfn->rel_pf_id);
+	}
+
+	DP_VERBOSE(p_hwfn, QED_MSG_SP, "Configured ARFS mode : %s\n",
+		   p_cfg_params->arfs_enable ? "Enable" : "Disable");
+}
+
+static int
+qed_configure_rfs_ntuple_filter(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt,
+				struct qed_spq_comp_cb *p_cb,
+				dma_addr_t p_addr, u16 length, u16 qid,
+				u8 vport_id, bool b_is_add)
+{
+	struct rx_update_gft_filter_data *p_ramrod = NULL;
+	struct qed_spq_entry *p_ent = NULL;
+	struct qed_sp_init_data init_data;
+	u16 abs_rx_q_id = 0;
+	u8 abs_vport_id = 0;
+	int rc = -EINVAL;
+
+	rc = qed_fw_vport(p_hwfn, vport_id, &abs_vport_id);
+	if (rc)
+		return rc;
+
+	rc = qed_fw_l2_queue(p_hwfn, qid, &abs_rx_q_id);
+	if (rc)
+		return rc;
+
+	/* Get SPQ entry */
+	memset(&init_data, 0, sizeof(init_data));
+	init_data.cid = qed_spq_get_cid(p_hwfn);
+
+	init_data.opaque_fid = p_hwfn->hw_info.opaque_fid;
+
+	if (p_cb) {
+		init_data.comp_mode = QED_SPQ_MODE_CB;
+		init_data.p_comp_data = p_cb;
+	} else {
+		init_data.comp_mode = QED_SPQ_MODE_EBLOCK;
+	}
+
+	rc = qed_sp_init_request(p_hwfn, &p_ent,
+				 ETH_RAMROD_GFT_UPDATE_FILTER,
+				 PROTOCOLID_ETH, &init_data);
+	if (rc)
+		return rc;
+
+	p_ramrod = &p_ent->ramrod.rx_update_gft;
+	DMA_REGPAIR_LE(p_ramrod->pkt_hdr_addr, p_addr);
+	p_ramrod->pkt_hdr_length = cpu_to_le16(length);
+	p_ramrod->rx_qid_or_action_icid = cpu_to_le16(abs_rx_q_id);
+	p_ramrod->vport_id = abs_vport_id;
+	p_ramrod->filter_type = RFS_FILTER_TYPE;
+	p_ramrod->filter_action = b_is_add ? GFT_ADD_FILTER : GFT_DELETE_FILTER;
+
+	DP_VERBOSE(p_hwfn, QED_MSG_SP,
+		   "V[%0x], Q[%04x] - %s filter from 0x%llx [length %04xb]\n",
+		   abs_vport_id, abs_rx_q_id,
+		   b_is_add ? "Adding" : "Removing", (u64)p_addr, length);
+
+	return qed_spq_post(p_hwfn, p_ent, NULL);
+}
+
 static int qed_fill_eth_dev_info(struct qed_dev *cdev,
 				 struct qed_dev_eth_info *info)
 {
@@ -2356,6 +2434,59 @@ static int qed_configure_filter(struct qed_dev *cdev,
 	}
 }
 
+static int qed_configure_arfs_searcher(struct qed_dev *cdev, bool en_searcher)
+{
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_arfs_config_params arfs_config_params;
+
+	memset(&arfs_config_params, 0, sizeof(arfs_config_params));
+	arfs_config_params.tcp = true;
+	arfs_config_params.udp = true;
+	arfs_config_params.ipv4 = true;
+	arfs_config_params.ipv6 = true;
+	arfs_config_params.arfs_enable = en_searcher;
+
+	qed_arfs_mode_configure(p_hwfn, p_hwfn->p_arfs_ptt,
+				&arfs_config_params);
+	return 0;
+}
+
+static void
+qed_arfs_sp_response_handler(struct qed_hwfn *p_hwfn,
+			     void *cookie, union event_ring_data *data,
+			     u8 fw_return_code)
+{
+	struct qed_common_cb_ops *op = p_hwfn->cdev->protocol_ops.common;
+	void *dev = p_hwfn->cdev->ops_cookie;
+
+	op->arfs_filter_op(dev, cookie, fw_return_code);
+}
+
+static int qed_ntuple_arfs_filter_config(struct qed_dev *cdev, void *cookie,
+					 dma_addr_t mapping, u16 length,
+					 u16 vport_id, u16 rx_queue_id,
+					 bool add_filter)
+{
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
+	struct qed_spq_comp_cb cb;
+	int rc = -EINVAL;
+
+	cb.function = qed_arfs_sp_response_handler;
+	cb.cookie = cookie;
+
+	rc = qed_configure_rfs_ntuple_filter(p_hwfn, p_hwfn->p_arfs_ptt,
+					     &cb, mapping, length, rx_queue_id,
+					     vport_id, add_filter);
+	if (rc)
+		DP_NOTICE(p_hwfn,
+			  "Failed to issue a-RFS filter configuration\n");
+	else
+		DP_VERBOSE(p_hwfn, NETIF_MSG_DRV,
+			   "Successfully issued a-RFS filter configuration\n");
+
+	return rc;
+}
+
 static int qed_fp_cqe_completion(struct qed_dev *dev,
 				 u8 rss_id, struct eth_slow_path_rx_cqe *cqe)
 {
@@ -2397,6 +2528,8 @@ static const struct qed_eth_ops qed_eth_ops_pass = {
 	.eth_cqe_completion = &qed_fp_cqe_completion,
 	.get_vport_stats = &qed_get_vport_stats,
 	.tunn_config = &qed_tunn_configure,
+	.ntuple_filter_config = &qed_ntuple_arfs_filter_config,
+	.configure_arfs_searcher = &qed_configure_arfs_searcher,
 };
 
 const struct qed_eth_ops *qed_get_eth_ops(void)

drivers/net/ethernet/qlogic/qed/qed_l2.h

@@ -185,6 +185,14 @@ struct qed_filter_accept_flags {
 #define QED_ACCEPT_BCAST                0x20
 };
 
+struct qed_arfs_config_params {
+	bool tcp;
+	bool udp;
+	bool ipv4;
+	bool ipv6;
+	bool arfs_enable;
+};
+
 struct qed_sp_vport_update_params {
 	u16				opaque_fid;
 	u8				vport_id;

drivers/net/ethernet/qlogic/qed/qed_main.c

@@ -883,6 +883,9 @@ static void qed_update_pf_params(struct qed_dev *cdev,
 		params->rdma_pf_params.gl_pi = QED_ROCE_PROTOCOL_INDEX;
 	}
 
+	if (cdev->num_hwfns > 1 || IS_VF(cdev))
+		params->eth_pf_params.num_arfs_filters = 0;
+
 	/* In case we might support RDMA, don't allow qede to be greedy
 	 * with the L2 contexts. Allow for 64 queues [rx, tx, xdp] per hwfn.
 	 */
@@ -926,6 +929,18 @@ static int qed_slowpath_start(struct qed_dev *cdev,
 			goto err;
 		}
 
+#ifdef CONFIG_RFS_ACCEL
+		if (cdev->num_hwfns == 1) {
+			p_ptt = qed_ptt_acquire(QED_LEADING_HWFN(cdev));
+			if (p_ptt) {
+				QED_LEADING_HWFN(cdev)->p_arfs_ptt = p_ptt;
+			} else {
+				DP_NOTICE(cdev,
+					  "Failed to acquire PTT for aRFS\n");
+				goto err;
+			}
+		}
+#endif
 		p_ptt = qed_ptt_acquire(QED_LEADING_HWFN(cdev));
 		if (p_ptt) {
 			QED_LEADING_HWFN(cdev)->p_ptp_ptt = p_ptt;
@@ -1032,6 +1047,12 @@ static int qed_slowpath_start(struct qed_dev *cdev,
 	if (IS_PF(cdev))
 		release_firmware(cdev->firmware);
 
+#ifdef CONFIG_RFS_ACCEL
+	if (IS_PF(cdev) && (cdev->num_hwfns == 1) &&
+	    QED_LEADING_HWFN(cdev)->p_arfs_ptt)
+		qed_ptt_release(QED_LEADING_HWFN(cdev),
+				QED_LEADING_HWFN(cdev)->p_arfs_ptt);
+#endif
 	if (IS_PF(cdev) && QED_LEADING_HWFN(cdev)->p_ptp_ptt)
 		qed_ptt_release(QED_LEADING_HWFN(cdev),
 				QED_LEADING_HWFN(cdev)->p_ptp_ptt);
@@ -1049,6 +1070,11 @@ static int qed_slowpath_stop(struct qed_dev *cdev)
 	qed_ll2_dealloc_if(cdev);
 
 	if (IS_PF(cdev)) {
+#ifdef CONFIG_RFS_ACCEL
+		if (cdev->num_hwfns == 1)
+			qed_ptt_release(QED_LEADING_HWFN(cdev),
+					QED_LEADING_HWFN(cdev)->p_arfs_ptt);
+#endif
 		qed_ptt_release(QED_LEADING_HWFN(cdev),
 				QED_LEADING_HWFN(cdev)->p_ptp_ptt);
 		qed_free_stream_mem(cdev);

drivers/net/ethernet/qlogic/qed/qed_reg_addr.h

@@ -1560,4 +1560,12 @@
 #define NIG_REG_TSGEN_FREECNT_UPDATE_K2 0x509008UL
 #define CNIG_REG_NIG_PORT0_CONF_K2 0x218200UL
 
+#define PRS_REG_SEARCH_GFT 0x1f11bcUL
+#define PRS_REG_CM_HDR_GFT 0x1f11c8UL
+#define PRS_REG_GFT_CAM 0x1f1100UL
+#define PRS_REG_GFT_PROFILE_MASK_RAM 0x1f1000UL
+#define PRS_REG_CM_HDR_GFT_EVENT_ID_SHIFT 0
+#define PRS_REG_CM_HDR_GFT_CM_HDR_SHIFT 8
+#define PRS_REG_LOAD_L2_FILTER 0x1f0198UL
+
 #endif

drivers/net/ethernet/qlogic/qed/qed_sp.h

@@ -84,6 +84,7 @@ union ramrod_data {
 	struct tx_queue_stop_ramrod_data tx_queue_stop;
 	struct vport_start_ramrod_data vport_start;
 	struct vport_stop_ramrod_data vport_stop;
+	struct rx_update_gft_filter_data rx_update_gft;
 	struct vport_update_ramrod_data vport_update;
 	struct core_rx_start_ramrod_data core_rx_queue_start;
 	struct core_rx_stop_ramrod_data core_rx_queue_stop;

drivers/net/ethernet/qlogic/qede/qede.h

@@ -41,6 +41,9 @@
 #include <linux/mutex.h>
 #include <linux/bpf.h>
 #include <linux/io.h>
+#ifdef CONFIG_RFS_ACCEL
+#include <linux/cpu_rmap.h>
+#endif
 #include <linux/qed/common_hsi.h>
 #include <linux/qed/eth_common.h>
 #include <linux/qed/qed_if.h>
@@ -237,7 +240,10 @@ struct qede_dev {
 	u16				vxlan_dst_port;
 	u16				geneve_dst_port;
 
-	bool wol_enabled;
+#ifdef CONFIG_RFS_ACCEL
+	struct qede_arfs		*arfs;
+#endif
+	bool				wol_enabled;
 
 	struct qede_rdma_dev		rdma_info;
 
@@ -439,6 +445,20 @@ struct qede_fastpath {
 #define QEDE_SP_VXLAN_PORT_CONFIG	2
 #define QEDE_SP_GENEVE_PORT_CONFIG	3
 
+#ifdef CONFIG_RFS_ACCEL
+int qede_rx_flow_steer(struct net_device *dev, const struct sk_buff *skb,
+		       u16 rxq_index, u32 flow_id);
+void qede_process_arfs_filters(struct qede_dev *edev, bool free_fltr);
+void qede_poll_for_freeing_arfs_filters(struct qede_dev *edev);
+void qede_arfs_filter_op(void *dev, void *filter, u8 fw_rc);
+void qede_free_arfs(struct qede_dev *edev);
+int qede_alloc_arfs(struct qede_dev *edev);
+
+#define QEDE_SP_ARFS_CONFIG	4
+#define QEDE_SP_TASK_POLL_DELAY	(5 * HZ)
+#define QEDE_RFS_MAX_FLTR	256
+#endif
+
 struct qede_reload_args {
 	void (*func)(struct qede_dev *edev, struct qede_reload_args *args);
 	union {

drivers/net/ethernet/qlogic/qede/qede_main.c

@@ -225,6 +225,9 @@ static struct pci_driver qede_pci_driver = {
 
 static struct qed_eth_cb_ops qede_ll_ops = {
 	{
+#ifdef CONFIG_RFS_ACCEL
+		.arfs_filter_op = qede_arfs_filter_op,
+#endif
 		.link_update = qede_link_update,
 	},
 	.force_mac = qede_force_mac,
@@ -554,6 +557,9 @@ static const struct net_device_ops qede_netdev_ops = {
 	.ndo_udp_tunnel_del = qede_udp_tunnel_del,
 	.ndo_features_check = qede_features_check,
 	.ndo_xdp = qede_xdp,
+#ifdef CONFIG_RFS_ACCEL
+	.ndo_rx_flow_steer = qede_rx_flow_steer,
+#endif
 };
 
 /* -------------------------------------------------------------------------
@@ -603,7 +609,7 @@ static void qede_init_ndev(struct qede_dev *edev)
 {
 	struct net_device *ndev = edev->ndev;
 	struct pci_dev *pdev = edev->pdev;
-	u32 hw_features;
+	netdev_features_t hw_features;
 
 	pci_set_drvdata(pdev, ndev);
 
@@ -629,6 +635,10 @@ static void qede_init_ndev(struct qede_dev *edev)
 	hw_features |= NETIF_F_GSO_GRE | NETIF_F_GSO_UDP_TUNNEL |
 		       NETIF_F_TSO_ECN | NETIF_F_GSO_UDP_TUNNEL_CSUM |
 		       NETIF_F_GSO_GRE_CSUM;
+
+	if (!IS_VF(edev) && edev->dev_info.common.num_hwfns == 1)
+		hw_features |= NETIF_F_NTUPLE;
+
 	ndev->hw_enc_features = NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
 				NETIF_F_SG | NETIF_F_TSO | NETIF_F_TSO_ECN |
 				NETIF_F_TSO6 | NETIF_F_GSO_GRE |
@@ -798,6 +808,12 @@ static void qede_sp_task(struct work_struct *work)
 		qed_ops->tunn_config(cdev, &tunn_params);
 	}
 
+#ifdef CONFIG_RFS_ACCEL
+	if (test_and_clear_bit(QEDE_SP_ARFS_CONFIG, &edev->sp_flags)) {
+		if (edev->state == QEDE_STATE_OPEN)
+			qede_process_arfs_filters(edev, false);
+	}
+#endif
 	__qede_unlock(edev);
 }
 
@@ -808,6 +824,9 @@ static void qede_update_pf_params(struct qed_dev *cdev)
 	/* 64 rx + 64 tx + 64 XDP */
 	memset(&pf_params, 0, sizeof(struct qed_pf_params));
 	pf_params.eth_pf_params.num_cons = (MAX_SB_PER_PF_MIMD - 1) * 3;
+#ifdef CONFIG_RFS_ACCEL
+	pf_params.eth_pf_params.num_arfs_filters = QEDE_RFS_MAX_FLTR;
+#endif
 	qed_ops->common->update_pf_params(cdev, &pf_params);
 }
 
@@ -962,9 +981,8 @@ static void __qede_remove(struct pci_dev *pdev, enum qede_remove_mode mode)
 
 	DP_INFO(edev, "Starting qede_remove\n");
 
-	cancel_delayed_work_sync(&edev->sp_task);
-
 	unregister_netdev(ndev);
+	cancel_delayed_work_sync(&edev->sp_task);
 
 	qede_ptp_remove(edev);
 
@@ -1490,6 +1508,18 @@ static int qede_req_msix_irqs(struct qede_dev *edev)
 	}
 
 	for (i = 0; i < QEDE_QUEUE_CNT(edev); i++) {
+#ifdef CONFIG_RFS_ACCEL
+		struct qede_fastpath *fp = &edev->fp_array[i];
+
+		if (edev->ndev->rx_cpu_rmap && (fp->type & QEDE_FASTPATH_RX)) {
+			rc = irq_cpu_rmap_add(edev->ndev->rx_cpu_rmap,
+					      edev->int_info.msix[i].vector);
+			if (rc) {
+				DP_ERR(edev, "Failed to add CPU rmap\n");
+				qede_free_arfs(edev);
+			}
+		}
+#endif
 		rc = request_irq(edev->int_info.msix[i].vector,
 				 qede_msix_fp_int, 0, edev->fp_array[i].name,
 				 &edev->fp_array[i]);
@@ -1871,7 +1901,12 @@ static void qede_unload(struct qede_dev *edev, enum qede_unload_mode mode,
 
 	qede_vlan_mark_nonconfigured(edev);
 	edev->ops->fastpath_stop(edev->cdev);
-
+#ifdef CONFIG_RFS_ACCEL
+	if (!IS_VF(edev) && edev->dev_info.common.num_hwfns == 1) {
+		qede_poll_for_freeing_arfs_filters(edev);
+		qede_free_arfs(edev);
+	}
+#endif
 	/* Release the interrupts */
 	qede_sync_free_irqs(edev);
 	edev->ops->common->set_fp_int(edev->cdev, 0);
@@ -1923,6 +1958,13 @@ static int qede_load(struct qede_dev *edev, enum qede_load_mode mode,
 	if (rc)
 		goto err2;
 
+#ifdef CONFIG_RFS_ACCEL
+	if (!IS_VF(edev) && edev->dev_info.common.num_hwfns == 1) {
+		rc = qede_alloc_arfs(edev);
+		if (rc)
+			DP_NOTICE(edev, "aRFS memory allocation failed\n");
+	}
+#endif
 	qede_napi_add_enable(edev);
 	DP_INFO(edev, "Napi added and enabled\n");
 

include/linux/qed/qed_eth_if.h

@@ -301,6 +301,14 @@ struct qed_eth_ops {
 
 	int (*tunn_config)(struct qed_dev *cdev,
 			   struct qed_tunn_params *params);
+
+	int (*ntuple_filter_config)(struct qed_dev *cdev, void *cookie,
+				    dma_addr_t mapping, u16 length,
+				    u16 vport_id, u16 rx_queue_id,
+				    bool add_filter);
+
+	int (*configure_arfs_searcher)(struct qed_dev *cdev,
+				       bool en_searcher);
 };
 
 const struct qed_eth_ops *qed_get_eth_ops(void);

include/linux/qed/qed_if.h

@@ -178,6 +178,12 @@ struct qed_eth_pf_params {
 	 * to update_pf_params routine invoked before slowpath start
 	 */
 	u16 num_cons;
+
+	/* To enable arfs, previous to HW-init a positive number needs to be
+	 * set [as filters require allocated searcher ILT memory].
+	 * This will set the maximal number of configured steering-filters.
+	 */
+	u32 num_arfs_filters;
 };
 
 struct qed_fcoe_pf_params {
@@ -427,6 +433,7 @@ struct qed_int_info {
 };
 
 struct qed_common_cb_ops {
+	void (*arfs_filter_op)(void *dev, void *fltr, u8 fw_rc);
 	void	(*link_update)(void			*dev,
 			       struct qed_link_output	*link);
 	void	(*dcbx_aen)(void *dev, struct qed_dcbx_get *get, u32 mib_type);