Merge branch 'bnxt_en-next'

Michael Chan says:

====================
bnxt_en: Update for net-next.

Three main changes in this series, besides the usual firmware spec
update:

1. Add support for a new firmware communication channel direct to the
firmware processor that handles flow offloads.  This speeds up
flow offload operations.

2. Use 64-bit internal flow handles to increase the number of flows
that can be offloaded.

3. Add level-2 context memory paging so that we can configure more
context memory for RDMA on the 57500 chips.  Allocate more context
memory if RDMA is enabled on the 57500 chips.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/broadcom/bnxt/bnxt.h

@@ -567,7 +567,6 @@ struct nqe_cn {
 #define HWRM_RESP_LEN_MASK		0xffff0000
 #define HWRM_RESP_LEN_SFT		16
 #define HWRM_RESP_VALID_MASK		0xff000000
-#define HWRM_SEQ_ID_INVALID		-1
 #define BNXT_HWRM_REQ_MAX_SIZE		128
 #define BNXT_HWRM_REQS_PER_PAGE		(BNXT_PAGE_SIZE /	\
 					 BNXT_HWRM_REQ_MAX_SIZE)
@@ -585,6 +584,9 @@ struct nqe_cn {
 
 #define HWRM_VALID_BIT_DELAY_USEC	20
 
+#define BNXT_HWRM_CHNL_CHIMP	0
+#define BNXT_HWRM_CHNL_KONG	1
+
 #define BNXT_RX_EVENT	1
 #define BNXT_AGG_EVENT	2
 #define BNXT_TX_EVENT	4
@@ -615,9 +617,12 @@ struct bnxt_sw_rx_agg_bd {
 struct bnxt_ring_mem_info {
 	int			nr_pages;
 	int			page_size;
-	u32			flags;
+	u16			flags;
 #define BNXT_RMEM_VALID_PTE_FLAG	1
 #define BNXT_RMEM_RING_PTE_FLAG		2
+#define BNXT_RMEM_USE_FULL_PAGE_FLAG	4
+
+	u16			depth;
 
 	void			**pg_arr;
 	dma_addr_t		*dma_arr;
@@ -1113,9 +1118,14 @@ struct bnxt_test_info {
 	char string[BNXT_MAX_TEST][ETH_GSTRING_LEN];
 };
 
-#define BNXT_GRCPF_REG_WINDOW_BASE_OUT	0x400
-#define BNXT_CAG_REG_LEGACY_INT_STATUS	0x4014
-#define BNXT_CAG_REG_BASE		0x300000
+#define BNXT_GRCPF_REG_CHIMP_COMM		0x0
+#define BNXT_GRCPF_REG_CHIMP_COMM_TRIGGER	0x100
+#define BNXT_GRCPF_REG_WINDOW_BASE_OUT		0x400
+#define BNXT_CAG_REG_LEGACY_INT_STATUS		0x4014
+#define BNXT_CAG_REG_BASE			0x300000
+
+#define BNXT_GRCPF_REG_KONG_COMM		0xA00
+#define BNXT_GRCPF_REG_KONG_COMM_TRIGGER	0xB00
 
 struct bnxt_tc_flow_stats {
 	u64		packets;
@@ -1183,12 +1193,15 @@ struct bnxt_vf_rep {
 #define PTU_PTE_NEXT_TO_LAST      0x4UL
 
 #define MAX_CTX_PAGES	(BNXT_PAGE_SIZE / 8)
+#define MAX_CTX_TOTAL_PAGES	(MAX_CTX_PAGES * MAX_CTX_PAGES)
 
 struct bnxt_ctx_pg_info {
 	u32		entries;
+	u32		nr_pages;
 	void		*ctx_pg_arr[MAX_CTX_PAGES];
 	dma_addr_t	ctx_dma_arr[MAX_CTX_PAGES];
 	struct bnxt_ring_mem_info ring_mem;
+	struct bnxt_ctx_pg_info **ctx_pg_tbl;
 };
 
 struct bnxt_ctx_mem_info {
@@ -1224,6 +1237,8 @@ struct bnxt_ctx_mem_info {
 	struct bnxt_ctx_pg_info cq_mem;
 	struct bnxt_ctx_pg_info vnic_mem;
 	struct bnxt_ctx_pg_info stat_mem;
+	struct bnxt_ctx_pg_info mrav_mem;
+	struct bnxt_ctx_pg_info tim_mem;
 	struct bnxt_ctx_pg_info *tqm_mem[9];
 };
 
@@ -1457,20 +1472,25 @@ struct bnxt {
 	u32			msg_enable;
 
 	u32			fw_cap;
-	#define BNXT_FW_CAP_SHORT_CMD	0x00000001
-	#define BNXT_FW_CAP_LLDP_AGENT	0x00000002
-	#define BNXT_FW_CAP_DCBX_AGENT	0x00000004
-	#define BNXT_FW_CAP_NEW_RM	0x00000008
-	#define BNXT_FW_CAP_IF_CHANGE	0x00000010
+	#define BNXT_FW_CAP_SHORT_CMD			0x00000001
+	#define BNXT_FW_CAP_LLDP_AGENT			0x00000002
+	#define BNXT_FW_CAP_DCBX_AGENT			0x00000004
+	#define BNXT_FW_CAP_NEW_RM			0x00000008
+	#define BNXT_FW_CAP_IF_CHANGE			0x00000010
+	#define BNXT_FW_CAP_KONG_MB_CHNL		0x00000080
+	#define BNXT_FW_CAP_OVS_64BIT_HANDLE		0x00000400
 
 #define BNXT_NEW_RM(bp)		((bp)->fw_cap & BNXT_FW_CAP_NEW_RM)
 	u32			hwrm_spec_code;
 	u16			hwrm_cmd_seq;
-	u32			hwrm_intr_seq_id;
+	u16                     hwrm_cmd_kong_seq;
+	u16			hwrm_intr_seq_id;
 	void			*hwrm_short_cmd_req_addr;
 	dma_addr_t		hwrm_short_cmd_req_dma_addr;
 	void			*hwrm_cmd_resp_addr;
 	dma_addr_t		hwrm_cmd_resp_dma_addr;
+	void			*hwrm_cmd_kong_resp_addr;
+	dma_addr_t		hwrm_cmd_kong_resp_dma_addr;
 
 	struct rtnl_link_stats64	net_stats_prev;
 	struct rx_port_stats	*hw_rx_port_stats;
@@ -1672,6 +1692,66 @@ static inline void bnxt_db_write(struct bnxt *bp, struct bnxt_db_info *db,
 	}
 }
 
+static inline bool bnxt_cfa_hwrm_message(u16 req_type)
+{
+	switch (req_type) {
+	case HWRM_CFA_ENCAP_RECORD_ALLOC:
+	case HWRM_CFA_ENCAP_RECORD_FREE:
+	case HWRM_CFA_DECAP_FILTER_ALLOC:
+	case HWRM_CFA_DECAP_FILTER_FREE:
+	case HWRM_CFA_NTUPLE_FILTER_ALLOC:
+	case HWRM_CFA_NTUPLE_FILTER_FREE:
+	case HWRM_CFA_NTUPLE_FILTER_CFG:
+	case HWRM_CFA_EM_FLOW_ALLOC:
+	case HWRM_CFA_EM_FLOW_FREE:
+	case HWRM_CFA_EM_FLOW_CFG:
+	case HWRM_CFA_FLOW_ALLOC:
+	case HWRM_CFA_FLOW_FREE:
+	case HWRM_CFA_FLOW_INFO:
+	case HWRM_CFA_FLOW_FLUSH:
+	case HWRM_CFA_FLOW_STATS:
+	case HWRM_CFA_METER_PROFILE_ALLOC:
+	case HWRM_CFA_METER_PROFILE_FREE:
+	case HWRM_CFA_METER_PROFILE_CFG:
+	case HWRM_CFA_METER_INSTANCE_ALLOC:
+	case HWRM_CFA_METER_INSTANCE_FREE:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static inline bool bnxt_kong_hwrm_message(struct bnxt *bp, struct input *req)
+{
+	return (bp->fw_cap & BNXT_FW_CAP_KONG_MB_CHNL &&
+		bnxt_cfa_hwrm_message(le16_to_cpu(req->req_type)));
+}
+
+static inline bool bnxt_hwrm_kong_chnl(struct bnxt *bp, struct input *req)
+{
+	return (bp->fw_cap & BNXT_FW_CAP_KONG_MB_CHNL &&
+		req->resp_addr == cpu_to_le64(bp->hwrm_cmd_kong_resp_dma_addr));
+}
+
+static inline void *bnxt_get_hwrm_resp_addr(struct bnxt *bp, void *req)
+{
+	if (bnxt_hwrm_kong_chnl(bp, (struct input *)req))
+		return bp->hwrm_cmd_kong_resp_addr;
+	else
+		return bp->hwrm_cmd_resp_addr;
+}
+
+static inline u16 bnxt_get_hwrm_seq_id(struct bnxt *bp, u16 dst)
+{
+	u16 seq_id;
+
+	if (dst == BNXT_HWRM_CHNL_CHIMP)
+		seq_id = bp->hwrm_cmd_seq++;
+	else
+		seq_id = bp->hwrm_cmd_kong_seq++;
+	return seq_id;
+}
+
 extern const u16 bnxt_lhint_arr[];
 
 int bnxt_alloc_rx_data(struct bnxt *bp, struct bnxt_rx_ring_info *rxr,

drivers/net/ethernet/broadcom/bnxt/bnxt_tc.c

@@ -337,18 +337,21 @@ static int bnxt_tc_parse_flow(struct bnxt *bp,
 	return bnxt_tc_parse_actions(bp, &flow->actions, tc_flow_cmd->exts);
 }
 
-static int bnxt_hwrm_cfa_flow_free(struct bnxt *bp, __le16 flow_handle)
+static int bnxt_hwrm_cfa_flow_free(struct bnxt *bp,
+				   struct bnxt_tc_flow_node *flow_node)
 {
 	struct hwrm_cfa_flow_free_input req = { 0 };
 	int rc;
 
 	bnxt_hwrm_cmd_hdr_init(bp, &req, HWRM_CFA_FLOW_FREE, -1, -1);
-	req.flow_handle = flow_handle;
+	if (bp->fw_cap & BNXT_FW_CAP_OVS_64BIT_HANDLE)
+		req.ext_flow_handle = flow_node->ext_flow_handle;
+	else
+		req.flow_handle = flow_node->flow_handle;
 
 	rc = hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
 	if (rc)
-		netdev_info(bp->dev, "Error: %s: flow_handle=0x%x rc=%d",
-			    __func__, flow_handle, rc);
+		netdev_info(bp->dev, "%s: Error rc=%d", __func__, rc);
 
 	if (rc)
 		rc = -EIO;
@@ -418,13 +421,14 @@ static bool bits_set(void *key, int len)
 
 static int bnxt_hwrm_cfa_flow_alloc(struct bnxt *bp, struct bnxt_tc_flow *flow,
 				    __le16 ref_flow_handle,
-				    __le32 tunnel_handle, __le16 *flow_handle)
+				    __le32 tunnel_handle,
+				    struct bnxt_tc_flow_node *flow_node)
 {
-	struct hwrm_cfa_flow_alloc_output *resp = bp->hwrm_cmd_resp_addr;
 	struct bnxt_tc_actions *actions = &flow->actions;
 	struct bnxt_tc_l3_key *l3_mask = &flow->l3_mask;
 	struct bnxt_tc_l3_key *l3_key = &flow->l3_key;
 	struct hwrm_cfa_flow_alloc_input req = { 0 };
+	struct hwrm_cfa_flow_alloc_output *resp;
 	u16 flow_flags = 0, action_flags = 0;
 	int rc;
 
@@ -527,8 +531,23 @@ static int bnxt_hwrm_cfa_flow_alloc(struct bnxt *bp, struct bnxt_tc_flow *flow,
 
 	mutex_lock(&bp->hwrm_cmd_lock);
 	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
-	if (!rc)
-		*flow_handle = resp->flow_handle;
+	if (!rc) {
+		resp = bnxt_get_hwrm_resp_addr(bp, &req);
+		/* CFA_FLOW_ALLOC response interpretation:
+		 *		    fw with	     fw with
+		 *		    16-bit	     64-bit
+		 *		    flow handle      flow handle
+		 *		    ===========	     ===========
+		 * flow_handle      flow handle      flow context id
+		 * ext_flow_handle  INVALID	     flow handle
+		 * flow_id	    INVALID	     flow counter id
+		 */
+		flow_node->flow_handle = resp->flow_handle;
+		if (bp->fw_cap & BNXT_FW_CAP_OVS_64BIT_HANDLE) {
+			flow_node->ext_flow_handle = resp->ext_flow_handle;
+			flow_node->flow_id = resp->flow_id;
+		}
+	}
 	mutex_unlock(&bp->hwrm_cmd_lock);
 
 	if (rc == HWRM_ERR_CODE_RESOURCE_ALLOC_ERROR)
@@ -544,9 +563,8 @@ static int hwrm_cfa_decap_filter_alloc(struct bnxt *bp,
 				       __le32 ref_decap_handle,
 				       __le32 *decap_filter_handle)
 {
-	struct hwrm_cfa_decap_filter_alloc_output *resp =
-						bp->hwrm_cmd_resp_addr;
 	struct hwrm_cfa_decap_filter_alloc_input req = { 0 };
+	struct hwrm_cfa_decap_filter_alloc_output *resp;
 	struct ip_tunnel_key *tun_key = &flow->tun_key;
 	u32 enables = 0;
 	int rc;
@@ -599,10 +617,12 @@ static int hwrm_cfa_decap_filter_alloc(struct bnxt *bp,
 
 	mutex_lock(&bp->hwrm_cmd_lock);
 	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
-	if (!rc)
+	if (!rc) {
+		resp = bnxt_get_hwrm_resp_addr(bp, &req);
 		*decap_filter_handle = resp->decap_filter_id;
-	else
+	} else {
 		netdev_info(bp->dev, "%s: Error rc=%d", __func__, rc);
+	}
 	mutex_unlock(&bp->hwrm_cmd_lock);
 
 	if (rc)
@@ -633,9 +653,8 @@ static int hwrm_cfa_encap_record_alloc(struct bnxt *bp,
 				       struct bnxt_tc_l2_key *l2_info,
 				       __le32 *encap_record_handle)
 {
-	struct hwrm_cfa_encap_record_alloc_output *resp =
-						bp->hwrm_cmd_resp_addr;
 	struct hwrm_cfa_encap_record_alloc_input req = { 0 };
+	struct hwrm_cfa_encap_record_alloc_output *resp;
 	struct hwrm_cfa_encap_data_vxlan *encap =
 			(struct hwrm_cfa_encap_data_vxlan *)&req.encap_data;
 	struct hwrm_vxlan_ipv4_hdr *encap_ipv4 =
@@ -667,10 +686,12 @@ static int hwrm_cfa_encap_record_alloc(struct bnxt *bp,
 
 	mutex_lock(&bp->hwrm_cmd_lock);
 	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
-	if (!rc)
+	if (!rc) {
+		resp = bnxt_get_hwrm_resp_addr(bp, &req);
 		*encap_record_handle = resp->encap_record_id;
-	else
+	} else {
 		netdev_info(bp->dev, "%s: Error rc=%d", __func__, rc);
+	}
 	mutex_unlock(&bp->hwrm_cmd_lock);
 
 	if (rc)
@@ -1224,7 +1245,7 @@ static int __bnxt_tc_del_flow(struct bnxt *bp,
 	int rc;
 
 	/* send HWRM cmd to free the flow-id */
-	bnxt_hwrm_cfa_flow_free(bp, flow_node->flow_handle);
+	bnxt_hwrm_cfa_flow_free(bp, flow_node);
 
 	mutex_lock(&tc_info->lock);
 
@@ -1246,6 +1267,12 @@ static int __bnxt_tc_del_flow(struct bnxt *bp,
 	return 0;
 }
 
+static void bnxt_tc_set_flow_dir(struct bnxt *bp, struct bnxt_tc_flow *flow,
+				 u16 src_fid)
+{
+	flow->dir = (bp->pf.fw_fid == src_fid) ? BNXT_DIR_RX : BNXT_DIR_TX;
+}
+
 static void bnxt_tc_set_src_fid(struct bnxt *bp, struct bnxt_tc_flow *flow,
 				u16 src_fid)
 {
@@ -1293,6 +1320,9 @@ static int bnxt_tc_add_flow(struct bnxt *bp, u16 src_fid,
 
 	bnxt_tc_set_src_fid(bp, flow, src_fid);
 
+	if (bp->fw_cap & BNXT_FW_CAP_OVS_64BIT_HANDLE)
+		bnxt_tc_set_flow_dir(bp, flow, src_fid);
+
 	if (!bnxt_tc_can_offload(bp, flow)) {
 		rc = -ENOSPC;
 		goto free_node;
@@ -1320,7 +1350,7 @@ static int bnxt_tc_add_flow(struct bnxt *bp, u16 src_fid,
 
 	/* send HWRM cmd to alloc the flow */
 	rc = bnxt_hwrm_cfa_flow_alloc(bp, flow, ref_flow_handle,
-				      tunnel_handle, &new_node->flow_handle);
+				      tunnel_handle, new_node);
 	if (rc)
 		goto put_tunnel;
 
@@ -1336,7 +1366,7 @@ static int bnxt_tc_add_flow(struct bnxt *bp, u16 src_fid,
 	return 0;
 
 hwrm_flow_free:
-	bnxt_hwrm_cfa_flow_free(bp, new_node->flow_handle);
+	bnxt_hwrm_cfa_flow_free(bp, new_node);
 put_tunnel:
 	bnxt_tc_put_tunnel_handle(bp, flow, new_node);
 put_l2:
@@ -1397,13 +1427,40 @@ static int bnxt_tc_get_flow_stats(struct bnxt *bp,
 	return 0;
 }
 
+static void bnxt_fill_cfa_stats_req(struct bnxt *bp,
+				    struct bnxt_tc_flow_node *flow_node,
+				    __le16 *flow_handle, __le32 *flow_id)
+{
+	u16 handle;
+
+	if (bp->fw_cap & BNXT_FW_CAP_OVS_64BIT_HANDLE) {
+		*flow_id = flow_node->flow_id;
+
+		/* If flow_id is used to fetch flow stats then:
+		 * 1. lower 12 bits of flow_handle must be set to all 1s.
+		 * 2. 15th bit of flow_handle must specify the flow
+		 *    direction (TX/RX).
+		 */
+		if (flow_node->flow.dir == BNXT_DIR_RX)
+			handle = CFA_FLOW_INFO_REQ_FLOW_HANDLE_DIR_RX |
+				 CFA_FLOW_INFO_REQ_FLOW_HANDLE_MAX_MASK;
+		else
+			handle = CFA_FLOW_INFO_REQ_FLOW_HANDLE_MAX_MASK;
+
+		*flow_handle = cpu_to_le16(handle);
+	} else {
+		*flow_handle = flow_node->flow_handle;
+	}
+}
+
 static int
 bnxt_hwrm_cfa_flow_stats_get(struct bnxt *bp, int num_flows,
 			     struct bnxt_tc_stats_batch stats_batch[])
 {
-	struct hwrm_cfa_flow_stats_output *resp = bp->hwrm_cmd_resp_addr;
 	struct hwrm_cfa_flow_stats_input req = { 0 };
+	struct hwrm_cfa_flow_stats_output *resp;
 	__le16 *req_flow_handles = &req.flow_handle_0;
+	__le32 *req_flow_ids = &req.flow_id_0;
 	int rc, i;
 
 	bnxt_hwrm_cmd_hdr_init(bp, &req, HWRM_CFA_FLOW_STATS, -1, -1);
@@ -1411,14 +1468,19 @@ bnxt_hwrm_cfa_flow_stats_get(struct bnxt *bp, int num_flows,
 	for (i = 0; i < num_flows; i++) {
 		struct bnxt_tc_flow_node *flow_node = stats_batch[i].flow_node;
 
-		req_flow_handles[i] = flow_node->flow_handle;
+		bnxt_fill_cfa_stats_req(bp, flow_node,
+					&req_flow_handles[i], &req_flow_ids[i]);
 	}
 
 	mutex_lock(&bp->hwrm_cmd_lock);
 	rc = _hwrm_send_message(bp, &req, sizeof(req), HWRM_CMD_TIMEOUT);
 	if (!rc) {
-		__le64 *resp_packets = &resp->packet_0;
-		__le64 *resp_bytes = &resp->byte_0;
+		__le64 *resp_packets;
+		__le64 *resp_bytes;
+
+		resp = bnxt_get_hwrm_resp_addr(bp, &req);
+		resp_packets = &resp->packet_0;
+		resp_bytes = &resp->byte_0;
 
 		for (i = 0; i < num_flows; i++) {
 			stats_batch[i].hw_stats.packets =

drivers/net/ethernet/broadcom/bnxt/bnxt_tc.h

@@ -98,6 +98,9 @@ struct bnxt_tc_flow {
 
 	/* flow applicable to pkts ingressing on this fid */
 	u16				src_fid;
+	u8				dir;
+#define BNXT_DIR_RX	1
+#define BNXT_DIR_TX	0
 	struct bnxt_tc_l2_key		l2_key;
 	struct bnxt_tc_l2_key		l2_mask;
 	struct bnxt_tc_l3_key		l3_key;
@@ -170,7 +173,9 @@ struct bnxt_tc_flow_node {
 
 	struct bnxt_tc_flow		flow;
 
+	__le64				ext_flow_handle;
 	__le16				flow_handle;
+	__le32				flow_id;
 
 	/* L2 node in l2 hashtable that shares flow's l2 key */
 	struct bnxt_tc_l2_node		*l2_node;