qed*: Utilize FW 8.37.2.0

This FW contains several fixes and features.

RDMA
- Several modifications and fixes for Memory Windows
- drop vlan and tcp timestamp from mss calculation in driver for
  this FW
- Fix SQ completion flow when local ack timeout is infinite
- Modifications in t10dif support

ETH
- Fix aRFS for tunneled traffic without inner IP.
- Fix chip configuration which may fail under heavy traffic conditions.
- Support receiving any-VNI in VXLAN and GENEVE RX classification.

iSCSI / FcoE
- Fix iSCSI recovery flow
- Drop vlan and tcp timestamp from mss calc for fw 8.37.2.0

Misc
- Several registers (split registers) won't read correctly with
  ethtool -d
Signed-off-by: Ariel Elior <Ariel.Elior@cavium.com>
Signed-off-by: Manish Rangankar <manish.rangankar@cavium.com>
Signed-off-by: Michal Kalderon <Michal.Kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/infiniband/hw/qedr/qedr_hsi_rdma.h

@@ -116,6 +116,7 @@ enum rdma_cqe_requester_status_enum {
 	RDMA_CQE_REQ_STS_TRANSPORT_RETRY_CNT_ERR,
 	RDMA_CQE_REQ_STS_WORK_REQUEST_FLUSHED_ERR,
 	RDMA_CQE_REQ_STS_XRC_VOILATION_ERR,
+	RDMA_CQE_REQ_STS_SIG_ERR,
 	MAX_RDMA_CQE_REQUESTER_STATUS_ENUM
 };
 
@@ -152,12 +153,12 @@ struct rdma_rq_sge {
 	struct regpair addr;
 	__le32 length;
 	__le32 flags;
-#define RDMA_RQ_SGE_L_KEY_MASK      0x3FFFFFF
-#define RDMA_RQ_SGE_L_KEY_SHIFT     0
+#define RDMA_RQ_SGE_L_KEY_LO_MASK   0x3FFFFFF
+#define RDMA_RQ_SGE_L_KEY_LO_SHIFT  0
 #define RDMA_RQ_SGE_NUM_SGES_MASK   0x7
 #define RDMA_RQ_SGE_NUM_SGES_SHIFT  26
-#define RDMA_RQ_SGE_RESERVED0_MASK  0x7
-#define RDMA_RQ_SGE_RESERVED0_SHIFT 29
+#define RDMA_RQ_SGE_L_KEY_HI_MASK   0x7
+#define RDMA_RQ_SGE_L_KEY_HI_SHIFT  29
 };
 
 struct rdma_srq_sge {
@@ -241,18 +242,39 @@ enum rdma_dif_io_direction_flg {
 	MAX_RDMA_DIF_IO_DIRECTION_FLG
 };
 
-/* RDMA DIF Runt Result Structure */
-struct rdma_dif_runt_result {
-	__le16 guard_tag;
-	__le16 reserved[3];
+struct rdma_dif_params {
+	__le32 base_ref_tag;
+	__le16 app_tag;
+	__le16 app_tag_mask;
+	__le16 runt_crc_value;
+	__le16 flags;
+#define RDMA_DIF_PARAMS_IO_DIRECTION_FLG_MASK    0x1
+#define RDMA_DIF_PARAMS_IO_DIRECTION_FLG_SHIFT   0
+#define RDMA_DIF_PARAMS_BLOCK_SIZE_MASK          0x1
+#define RDMA_DIF_PARAMS_BLOCK_SIZE_SHIFT         1
+#define RDMA_DIF_PARAMS_RUNT_VALID_FLG_MASK      0x1
+#define RDMA_DIF_PARAMS_RUNT_VALID_FLG_SHIFT     2
+#define RDMA_DIF_PARAMS_VALIDATE_CRC_GUARD_MASK  0x1
+#define RDMA_DIF_PARAMS_VALIDATE_CRC_GUARD_SHIFT 3
+#define RDMA_DIF_PARAMS_VALIDATE_REF_TAG_MASK    0x1
+#define RDMA_DIF_PARAMS_VALIDATE_REF_TAG_SHIFT   4
+#define RDMA_DIF_PARAMS_VALIDATE_APP_TAG_MASK    0x1
+#define RDMA_DIF_PARAMS_VALIDATE_APP_TAG_SHIFT   5
+#define RDMA_DIF_PARAMS_CRC_SEED_MASK            0x1
+#define RDMA_DIF_PARAMS_CRC_SEED_SHIFT           6
+#define RDMA_DIF_PARAMS_RX_REF_TAG_CONST_MASK    0x1
+#define RDMA_DIF_PARAMS_RX_REF_TAG_CONST_SHIFT   7
+#define RDMA_DIF_PARAMS_BLOCK_GUARD_TYPE_MASK    0x1
+#define RDMA_DIF_PARAMS_BLOCK_GUARD_TYPE_SHIFT   8
+#define RDMA_DIF_PARAMS_APP_ESCAPE_MASK          0x1
+#define RDMA_DIF_PARAMS_APP_ESCAPE_SHIFT         9
+#define RDMA_DIF_PARAMS_REF_ESCAPE_MASK          0x1
+#define RDMA_DIF_PARAMS_REF_ESCAPE_SHIFT         10
+#define RDMA_DIF_PARAMS_RESERVED4_MASK           0x1F
+#define RDMA_DIF_PARAMS_RESERVED4_SHIFT          11
+	__le32 reserved5;
 };
 
-/* Memory window type enumeration */
-enum rdma_mw_type {
-	RDMA_MW_TYPE_1,
-	RDMA_MW_TYPE_2A,
-	MAX_RDMA_MW_TYPE
-};
 
 struct rdma_sq_atomic_wqe {
 	__le32 reserved1;
@@ -334,17 +356,17 @@ struct rdma_sq_bind_wqe {
 #define RDMA_SQ_BIND_WQE_SE_FLG_SHIFT        3
 #define RDMA_SQ_BIND_WQE_INLINE_FLG_MASK     0x1
 #define RDMA_SQ_BIND_WQE_INLINE_FLG_SHIFT    4
-#define RDMA_SQ_BIND_WQE_RESERVED0_MASK      0x7
-#define RDMA_SQ_BIND_WQE_RESERVED0_SHIFT     5
+#define RDMA_SQ_BIND_WQE_DIF_ON_HOST_FLG_MASK  0x1
+#define RDMA_SQ_BIND_WQE_DIF_ON_HOST_FLG_SHIFT 5
+#define RDMA_SQ_BIND_WQE_RESERVED0_MASK      0x3
+#define RDMA_SQ_BIND_WQE_RESERVED0_SHIFT     6
 	u8 wqe_size;
 	u8 prev_wqe_size;
 	u8 bind_ctrl;
 #define RDMA_SQ_BIND_WQE_ZERO_BASED_MASK     0x1
 #define RDMA_SQ_BIND_WQE_ZERO_BASED_SHIFT    0
-#define RDMA_SQ_BIND_WQE_MW_TYPE_MASK        0x1
-#define RDMA_SQ_BIND_WQE_MW_TYPE_SHIFT       1
-#define RDMA_SQ_BIND_WQE_RESERVED1_MASK      0x3F
-#define RDMA_SQ_BIND_WQE_RESERVED1_SHIFT     2
+#define RDMA_SQ_BIND_WQE_RESERVED1_MASK        0x7F
+#define RDMA_SQ_BIND_WQE_RESERVED1_SHIFT       1
 	u8 access_ctrl;
 #define RDMA_SQ_BIND_WQE_REMOTE_READ_MASK    0x1
 #define RDMA_SQ_BIND_WQE_REMOTE_READ_SHIFT   0
@@ -363,6 +385,7 @@ struct rdma_sq_bind_wqe {
 	__le32 length_lo;
 	__le32 parent_l_key;
 	__le32 reserved4;
+	struct rdma_dif_params dif_params;
 };
 
 /* First element (16 bytes) of bind wqe */
@@ -392,10 +415,8 @@ struct rdma_sq_bind_wqe_2nd {
 	u8 bind_ctrl;
 #define RDMA_SQ_BIND_WQE_2ND_ZERO_BASED_MASK     0x1
 #define RDMA_SQ_BIND_WQE_2ND_ZERO_BASED_SHIFT    0
-#define RDMA_SQ_BIND_WQE_2ND_MW_TYPE_MASK        0x1
-#define RDMA_SQ_BIND_WQE_2ND_MW_TYPE_SHIFT       1
-#define RDMA_SQ_BIND_WQE_2ND_RESERVED1_MASK      0x3F
-#define RDMA_SQ_BIND_WQE_2ND_RESERVED1_SHIFT     2
+#define RDMA_SQ_BIND_WQE_2ND_RESERVED1_MASK      0x7F
+#define RDMA_SQ_BIND_WQE_2ND_RESERVED1_SHIFT     1
 	u8 access_ctrl;
 #define RDMA_SQ_BIND_WQE_2ND_REMOTE_READ_MASK    0x1
 #define RDMA_SQ_BIND_WQE_2ND_REMOTE_READ_SHIFT   0
@@ -416,6 +437,11 @@ struct rdma_sq_bind_wqe_2nd {
 	__le32 reserved4;
 };
 
+/* Third element (16 bytes) of bind wqe */
+struct rdma_sq_bind_wqe_3rd {
+	struct rdma_dif_params dif_params;
+};
+
 /* Structure with only the SQ WQE common
  * fields. Size is of one SQ element (16B)
  */
@@ -486,30 +512,6 @@ struct rdma_sq_fmr_wqe {
 	u8 length_hi;
 	__le32 length_lo;
 	struct regpair pbl_addr;
-	__le32 dif_base_ref_tag;
-	__le16 dif_app_tag;
-	__le16 dif_app_tag_mask;
-	__le16 dif_runt_crc_value;
-	__le16 dif_flags;
-#define RDMA_SQ_FMR_WQE_DIF_IO_DIRECTION_FLG_MASK	0x1
-#define RDMA_SQ_FMR_WQE_DIF_IO_DIRECTION_FLG_SHIFT	0
-#define RDMA_SQ_FMR_WQE_DIF_BLOCK_SIZE_MASK		0x1
-#define RDMA_SQ_FMR_WQE_DIF_BLOCK_SIZE_SHIFT		1
-#define RDMA_SQ_FMR_WQE_DIF_RUNT_VALID_FLG_MASK		0x1
-#define RDMA_SQ_FMR_WQE_DIF_RUNT_VALID_FLG_SHIFT	2
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_CRC_GUARD_MASK	0x1
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_CRC_GUARD_SHIFT	3
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_REF_TAG_MASK	0x1
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_REF_TAG_SHIFT	4
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_APP_TAG_MASK	0x1
-#define RDMA_SQ_FMR_WQE_DIF_VALIDATE_APP_TAG_SHIFT	5
-#define RDMA_SQ_FMR_WQE_DIF_CRC_SEED_MASK		0x1
-#define RDMA_SQ_FMR_WQE_DIF_CRC_SEED_SHIFT		6
-#define RDMA_SQ_FMR_WQE_DIF_RX_REF_TAG_CONST_MASK	0x1
-#define RDMA_SQ_FMR_WQE_DIF_RX_REF_TAG_CONST_SHIFT	7
-#define RDMA_SQ_FMR_WQE_RESERVED4_MASK			0xFF
-#define RDMA_SQ_FMR_WQE_RESERVED4_SHIFT			8
-	__le32 reserved5;
 };
 
 /* First element (16 bytes) of fmr wqe */
@@ -566,33 +568,6 @@ struct rdma_sq_fmr_wqe_2nd {
 	struct regpair pbl_addr;
 };
 
-/* Third element (16 bytes) of fmr wqe */
-struct rdma_sq_fmr_wqe_3rd {
-	__le32 dif_base_ref_tag;
-	__le16 dif_app_tag;
-	__le16 dif_app_tag_mask;
-	__le16 dif_runt_crc_value;
-	__le16 dif_flags;
-#define RDMA_SQ_FMR_WQE_3RD_DIF_IO_DIRECTION_FLG_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_IO_DIRECTION_FLG_SHIFT		0
-#define RDMA_SQ_FMR_WQE_3RD_DIF_BLOCK_SIZE_MASK			0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_BLOCK_SIZE_SHIFT		1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_RUNT_VALID_FLG_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_RUNT_VALID_FLG_SHIFT		2
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_CRC_GUARD_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_CRC_GUARD_SHIFT	3
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_REF_TAG_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_REF_TAG_SHIFT		4
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_APP_TAG_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_VALIDATE_APP_TAG_SHIFT		5
-#define RDMA_SQ_FMR_WQE_3RD_DIF_CRC_SEED_MASK			0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_CRC_SEED_SHIFT			6
-#define RDMA_SQ_FMR_WQE_3RD_DIF_RX_REF_TAG_CONST_MASK		0x1
-#define RDMA_SQ_FMR_WQE_3RD_DIF_RX_REF_TAG_CONST_SHIFT		7
-#define RDMA_SQ_FMR_WQE_3RD_RESERVED4_MASK			0xFF
-#define RDMA_SQ_FMR_WQE_RESERVED4_SHIFT				8
-	__le32 reserved5;
-};
 
 struct rdma_sq_local_inv_wqe {
 	struct regpair reserved;
@@ -637,8 +612,8 @@ struct rdma_sq_rdma_wqe {
 #define RDMA_SQ_RDMA_WQE_DIF_ON_HOST_FLG_SHIFT	5
 #define RDMA_SQ_RDMA_WQE_READ_INV_FLG_MASK	0x1
 #define RDMA_SQ_RDMA_WQE_READ_INV_FLG_SHIFT	6
-#define RDMA_SQ_RDMA_WQE_RESERVED0_MASK		0x1
-#define RDMA_SQ_RDMA_WQE_RESERVED0_SHIFT	7
+#define RDMA_SQ_RDMA_WQE_RESERVED1_MASK        0x1
+#define RDMA_SQ_RDMA_WQE_RESERVED1_SHIFT       7
 	u8 wqe_size;
 	u8 prev_wqe_size;
 	struct regpair remote_va;
@@ -646,13 +621,9 @@ struct rdma_sq_rdma_wqe {
 	u8 dif_flags;
 #define RDMA_SQ_RDMA_WQE_DIF_BLOCK_SIZE_MASK            0x1
 #define RDMA_SQ_RDMA_WQE_DIF_BLOCK_SIZE_SHIFT           0
-#define RDMA_SQ_RDMA_WQE_DIF_FIRST_RDMA_IN_IO_FLG_MASK  0x1
-#define RDMA_SQ_RDMA_WQE_DIF_FIRST_RDMA_IN_IO_FLG_SHIFT 1
-#define RDMA_SQ_RDMA_WQE_DIF_LAST_RDMA_IN_IO_FLG_MASK   0x1
-#define RDMA_SQ_RDMA_WQE_DIF_LAST_RDMA_IN_IO_FLG_SHIFT  2
-#define RDMA_SQ_RDMA_WQE_RESERVED1_MASK                 0x1F
-#define RDMA_SQ_RDMA_WQE_RESERVED1_SHIFT                3
-	u8 reserved2[3];
+#define RDMA_SQ_RDMA_WQE_RESERVED2_MASK        0x7F
+#define RDMA_SQ_RDMA_WQE_RESERVED2_SHIFT       1
+	u8 reserved3[3];
 };
 
 /* First element (16 bytes) of rdma wqe */

drivers/infiniband/hw/qedr/verbs.c

@@ -3276,7 +3276,7 @@ int qedr_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 				SET_FIELD(flags, RDMA_RQ_SGE_NUM_SGES,
 					  wr->num_sge);
 
-			SET_FIELD(flags, RDMA_RQ_SGE_L_KEY,
+			SET_FIELD(flags, RDMA_RQ_SGE_L_KEY_LO,
 				  wr->sg_list[i].lkey);
 
 			RQ_SGE_SET(rqe, wr->sg_list[i].addr,
@@ -3295,7 +3295,7 @@ int qedr_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 			/* First one must include the number
 			 * of SGE in the list
 			 */
-			SET_FIELD(flags, RDMA_RQ_SGE_L_KEY, 0);
+			SET_FIELD(flags, RDMA_RQ_SGE_L_KEY_LO, 0);
 			SET_FIELD(flags, RDMA_RQ_SGE_NUM_SGES, 1);
 
 			RQ_SGE_SET(rqe, 0, 0, flags);

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

@@ -2792,7 +2792,7 @@ static void qed_hw_info_port_num_bb(struct qed_hwfn *p_hwfn,
 {
 	u32 port_mode;
 
-	port_mode = qed_rd(p_hwfn, p_ptt, CNIG_REG_NW_PORT_MODE_BB_B0);
+	port_mode = qed_rd(p_hwfn, p_ptt, CNIG_REG_NW_PORT_MODE_BB);
 
 	if (port_mode < 3) {
 		p_hwfn->cdev->num_ports_in_engine = 1;

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

@@ -360,6 +360,26 @@ void qed_port_unpretend(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
 	       *(u32 *)&p_ptt->pxp.pretend);
 }
 
+void qed_port_fid_pretend(struct qed_hwfn *p_hwfn,
+			  struct qed_ptt *p_ptt, u8 port_id, u16 fid)
+{
+	u16 control = 0;
+
+	SET_FIELD(control, PXP_PRETEND_CMD_PORT, port_id);
+	SET_FIELD(control, PXP_PRETEND_CMD_USE_PORT, 1);
+	SET_FIELD(control, PXP_PRETEND_CMD_PRETEND_PORT, 1);
+	SET_FIELD(control, PXP_PRETEND_CMD_IS_CONCRETE, 1);
+	SET_FIELD(control, PXP_PRETEND_CMD_PRETEND_FUNCTION, 1);
+	if (!GET_FIELD(fid, PXP_CONCRETE_FID_VFVALID))
+		fid = GET_FIELD(fid, PXP_CONCRETE_FID_PFID);
+	p_ptt->pxp.pretend.control = cpu_to_le16(control);
+	p_ptt->pxp.pretend.fid.concrete_fid.fid = cpu_to_le16(fid);
+	REG_WR(p_hwfn,
+	       qed_ptt_config_addr(p_ptt) +
+	       offsetof(struct pxp_ptt_entry, pretend),
+	       *(u32 *)&p_ptt->pxp.pretend);
+}
+
 u32 qed_vfid_to_concrete(struct qed_hwfn *p_hwfn, u8 vfid)
 {
 	u32 concrete_fid = 0;

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

@@ -244,6 +244,18 @@ void qed_port_pretend(struct qed_hwfn *p_hwfn,
 void qed_port_unpretend(struct qed_hwfn *p_hwfn,
 			struct qed_ptt *p_ptt);
 
+/**
+ * @brief qed_port_fid_pretend - pretend to another port and another function
+ *        when accessing the ptt window
+ *
+ * @param p_hwfn
+ * @param p_ptt
+ * @param port_id - the port to pretend to
+ * @param fid - fid field of pxp_pretend structure. Can contain either pf / vf.
+ */
+void qed_port_fid_pretend(struct qed_hwfn *p_hwfn,
+			  struct qed_ptt *p_ptt, u8 port_id, u16 fid);
+
 /**
  * @brief qed_vfid_to_concrete - build a concrete FID for a
  *        given VF ID

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

@@ -1245,7 +1245,7 @@ void qed_gft_config(struct qed_hwfn *p_hwfn,
 		    bool udp,
 		    bool ipv4, bool ipv6, enum gft_profile_type profile_type)
 {
-	u32 reg_val, cam_line, ram_line_lo, ram_line_hi;
+	u32 reg_val, cam_line, ram_line_lo, ram_line_hi, search_non_ip_as_gft;
 
 	if (!ipv6 && !ipv4)
 		DP_NOTICE(p_hwfn,
@@ -1314,6 +1314,9 @@ void qed_gft_config(struct qed_hwfn *p_hwfn,
 	ram_line_lo = 0;
 	ram_line_hi = 0;
 
+	/* Search no IP as GFT */
+	search_non_ip_as_gft = 0;
+
 	/* Tunnel type */
 	SET_FIELD(ram_line_lo, GFT_RAM_LINE_TUNNEL_DST_PORT, 1);
 	SET_FIELD(ram_line_lo, GFT_RAM_LINE_TUNNEL_OVER_IP_PROTOCOL, 1);
@@ -1337,8 +1340,13 @@ void qed_gft_config(struct qed_hwfn *p_hwfn,
 		SET_FIELD(ram_line_lo, GFT_RAM_LINE_ETHERTYPE, 1);
 	} else if (profile_type == GFT_PROFILE_TYPE_TUNNEL_TYPE) {
 		SET_FIELD(ram_line_lo, GFT_RAM_LINE_TUNNEL_ETHERTYPE, 1);
+
+		/* Allow tunneled traffic without inner IP */
+		search_non_ip_as_gft = 1;
 	}
 
+	qed_wr(p_hwfn,
+	       p_ptt, PRS_REG_SEARCH_NON_IP_AS_GFT, search_non_ip_as_gft);
 	qed_wr(p_hwfn,
 	       p_ptt,
 	       PRS_REG_GFT_PROFILE_MASK_RAM + RAM_LINE_SIZE * pf_id,
@@ -1509,3 +1517,43 @@ void qed_enable_context_validation(struct qed_hwfn *p_hwfn,
 	ctx_validation = CDU_VALIDATION_DEFAULT_CFG << 8;
 	qed_wr(p_hwfn, p_ptt, CDU_REG_TCFC_CTX_VALID0, ctx_validation);
 }
+
+static u32 qed_get_rdma_assert_ram_addr(struct qed_hwfn *p_hwfn, u8 storm_id)
+{
+	switch (storm_id) {
+	case 0:
+		return TSEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    TSTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+	case 1:
+		return MSEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    MSTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+	case 2:
+		return USEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    USTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+	case 3:
+		return XSEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    XSTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+	case 4:
+		return YSEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    YSTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+	case 5:
+		return PSEM_REG_FAST_MEMORY + SEM_FAST_REG_INT_RAM +
+		    PSTORM_RDMA_ASSERT_LEVEL_OFFSET(p_hwfn->rel_pf_id);
+
+	default:
+		return 0;
+	}
+}
+
+void qed_set_rdma_error_level(struct qed_hwfn *p_hwfn,
+			      struct qed_ptt *p_ptt,
+			      u8 assert_level[NUM_STORMS])
+{
+	u8 storm_id;
+
+	for (storm_id = 0; storm_id < NUM_STORMS; storm_id++) {
+		u32 ram_addr = qed_get_rdma_assert_ram_addr(p_hwfn, storm_id);
+
+		qed_wr(p_hwfn, p_ptt, ram_addr, assert_level[storm_id]);
+	}
+}

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

@@ -1159,7 +1159,6 @@ int qed_iwarp_connect(void *rdma_cxt,
 	struct qed_iwarp_info *iwarp_info;
 	struct qed_iwarp_ep *ep;
 	u8 mpa_data_size = 0;
-	u8 ts_hdr_size = 0;
 	u32 cid;
 	int rc;
 
@@ -1218,10 +1217,7 @@ int qed_iwarp_connect(void *rdma_cxt,
 	       iparams->cm_info.private_data,
 	       iparams->cm_info.private_data_len);
 
-	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
-		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
-
-	ep->mss = iparams->mss - ts_hdr_size;
+	ep->mss = iparams->mss;
 	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
 
 	ep->event_cb = iparams->event_cb;
@@ -2337,7 +2333,6 @@ qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
 	u8 local_mac_addr[ETH_ALEN];
 	struct qed_iwarp_ep *ep;
 	int tcp_start_offset;
-	u8 ts_hdr_size = 0;
 	u8 ll2_syn_handle;
 	int payload_len;
 	u32 hdr_size;
@@ -2415,11 +2410,7 @@ qed_iwarp_ll2_comp_syn_pkt(void *cxt, struct qed_ll2_comp_rx_data *data)
 
 	memcpy(&ep->cm_info, &cm_info, sizeof(ep->cm_info));
 
-	if (p_hwfn->p_rdma_info->iwarp.tcp_flags & QED_IWARP_TS_EN)
-		ts_hdr_size = TIMESTAMP_HEADER_SIZE;
-
-	hdr_size = ((cm_info.ip_version == QED_TCP_IPV4) ? 40 : 60) +
-		   ts_hdr_size;
+	hdr_size = ((cm_info.ip_version == QED_TCP_IPV4) ? 40 : 60);
 	ep->mss = p_hwfn->p_rdma_info->iwarp.max_mtu - hdr_size;
 	ep->mss = min_t(u16, QED_IWARP_MAX_FW_MSS, ep->mss);
 

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

@@ -586,6 +586,9 @@ qed_sp_update_accept_mode(struct qed_hwfn *p_hwfn,
 		SET_FIELD(state, ETH_VPORT_RX_MODE_BCAST_ACCEPT_ALL,
 			  !!(accept_filter & QED_ACCEPT_BCAST));
 
+		SET_FIELD(state, ETH_VPORT_RX_MODE_ACCEPT_ANY_VNI,
+			  !!(accept_filter & QED_ACCEPT_ANY_VNI));
+
 		p_ramrod->rx_mode.state = cpu_to_le16(state);
 		DP_VERBOSE(p_hwfn, QED_MSG_SP,
 			   "p_ramrod->rx_mode.state = 0x%x\n", state);

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

@@ -183,6 +183,7 @@ struct qed_filter_accept_flags {
 #define QED_ACCEPT_MCAST_MATCHED        0x08
 #define QED_ACCEPT_MCAST_UNMATCHED      0x10
 #define QED_ACCEPT_BCAST                0x20
+#define QED_ACCEPT_ANY_VNI              0x40
 };
 
 struct qed_arfs_config_params {

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

@@ -1508,11 +1508,8 @@ qed_rdma_register_tid(void *rdma_cxt,
 	case QED_RDMA_TID_FMR:
 		tid_type = RDMA_TID_FMR;
 		break;
-	case QED_RDMA_TID_MW_TYPE1:
-		tid_type = RDMA_TID_MW_TYPE1;
-		break;
-	case QED_RDMA_TID_MW_TYPE2A:
-		tid_type = RDMA_TID_MW_TYPE2A;
+	case QED_RDMA_TID_MW:
+		tid_type = RDMA_TID_MW;
 		break;
 	default:
 		rc = -EINVAL;
@@ -1544,7 +1541,6 @@ qed_rdma_register_tid(void *rdma_cxt,
 			  RDMA_REGISTER_TID_RAMROD_DATA_DIF_ON_HOST_FLG, 1);
 		DMA_REGPAIR_LE(p_ramrod->dif_error_addr,
 			       params->dif_error_addr);
-		DMA_REGPAIR_LE(p_ramrod->dif_runt_addr, params->dif_runt_addr);
 	}
 
 	rc = qed_spq_post(p_hwfn, p_ent, &fw_return_code);

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

@@ -178,7 +178,7 @@
 	0x008c80UL
 #define  MCP_REG_SCRATCH	\
 	0xe20000UL
-#define  CNIG_REG_NW_PORT_MODE_BB_B0 \
+#define  CNIG_REG_NW_PORT_MODE_BB \
 	0x218200UL
 #define  MISCS_REG_CHIP_NUM \
 	0x00976cUL
@@ -1621,6 +1621,7 @@
 #define NIG_REG_TX_EDPM_CTRL_TX_EDPM_TC_EN_SHIFT 1
 
 #define PRS_REG_SEARCH_GFT 0x1f11bcUL
+#define PRS_REG_SEARCH_NON_IP_AS_GFT 0x1f11c0UL
 #define PRS_REG_CM_HDR_GFT 0x1f11c8UL
 #define PRS_REG_GFT_CAM 0x1f1100UL
 #define PRS_REG_GFT_PROFILE_MASK_RAM 0x1f1000UL

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

@@ -681,7 +681,6 @@ static int qed_roce_sp_modify_requester(struct qed_hwfn *p_hwfn,
 
 static int qed_roce_sp_destroy_qp_responder(struct qed_hwfn *p_hwfn,
 					    struct qed_rdma_qp *qp,
-					    u32 *num_invalidated_mw,
 					    u32 *cq_prod)
 {
 	struct roce_destroy_qp_resp_output_params *p_ramrod_res;
@@ -692,8 +691,6 @@ static int qed_roce_sp_destroy_qp_responder(struct qed_hwfn *p_hwfn,
 	int rc;
 
 	DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "icid = %08x\n", qp->icid);
-
-	*num_invalidated_mw = 0;
 	*cq_prod = qp->cq_prod;
 
 	if (!qp->resp_offloaded) {
@@ -742,7 +739,6 @@ static int qed_roce_sp_destroy_qp_responder(struct qed_hwfn *p_hwfn,
 	if (rc)
 		goto err;
 
-	*num_invalidated_mw = le32_to_cpu(p_ramrod_res->num_invalidated_mw);
 	*cq_prod = le32_to_cpu(p_ramrod_res->cq_prod);
 	qp->cq_prod = *cq_prod;
 
@@ -764,8 +760,7 @@ static int qed_roce_sp_destroy_qp_responder(struct qed_hwfn *p_hwfn,
 }
 
 static int qed_roce_sp_destroy_qp_requester(struct qed_hwfn *p_hwfn,
-					    struct qed_rdma_qp *qp,
-					    u32 *num_bound_mw)
+					    struct qed_rdma_qp *qp)
 {
 	struct roce_destroy_qp_req_output_params *p_ramrod_res;
 	struct roce_destroy_qp_req_ramrod_data *p_ramrod;
@@ -807,7 +802,6 @@ static int qed_roce_sp_destroy_qp_requester(struct qed_hwfn *p_hwfn,
 	if (rc)
 		goto err;
 
-	*num_bound_mw = le32_to_cpu(p_ramrod_res->num_bound_mw);
 
 	/* Free ORQ - only if ramrod succeeded, in case FW is still using it */
 	dma_free_coherent(&p_hwfn->cdev->pdev->dev,
@@ -968,8 +962,6 @@ int qed_roce_query_qp(struct qed_hwfn *p_hwfn,
 
 int qed_roce_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 {
-	u32 num_invalidated_mw = 0;
-	u32 num_bound_mw = 0;
 	u32 cq_prod;
 	int rc;
 
@@ -984,22 +976,14 @@ int qed_roce_destroy_qp(struct qed_hwfn *p_hwfn, struct qed_rdma_qp *qp)
 
 	if (qp->cur_state != QED_ROCE_QP_STATE_RESET) {
 		rc = qed_roce_sp_destroy_qp_responder(p_hwfn, qp,
-						      &num_invalidated_mw,
 						      &cq_prod);
 		if (rc)
 			return rc;
 
 		/* Send destroy requester ramrod */
-		rc = qed_roce_sp_destroy_qp_requester(p_hwfn, qp,
-						      &num_bound_mw);
+		rc = qed_roce_sp_destroy_qp_requester(p_hwfn, qp);
 		if (rc)
 			return rc;
-
-		if (num_invalidated_mw != num_bound_mw) {
-			DP_NOTICE(p_hwfn,
-				  "number of invalidate memory windows is different from bounded ones\n");
-			return -EINVAL;
-		}
 	}
 
 	return 0;
@@ -1010,7 +994,6 @@ int qed_roce_modify_qp(struct qed_hwfn *p_hwfn,
 		       enum qed_roce_qp_state prev_state,
 		       struct qed_rdma_modify_qp_in_params *params)
 {
-	u32 num_invalidated_mw = 0, num_bound_mw = 0;
 	int rc = 0;
 
 	/* Perform additional operations according to the current state and the
@@ -1090,7 +1073,6 @@ int qed_roce_modify_qp(struct qed_hwfn *p_hwfn,
 		/* Send destroy responder ramrod */
 		rc = qed_roce_sp_destroy_qp_responder(p_hwfn,
 						      qp,
-						      &num_invalidated_mw,
 						      &cq_prod);
 
 		if (rc)
@@ -1098,14 +1080,7 @@ int qed_roce_modify_qp(struct qed_hwfn *p_hwfn,
 
 		qp->cq_prod = cq_prod;
 
-		rc = qed_roce_sp_destroy_qp_requester(p_hwfn, qp,
-						      &num_bound_mw);
-
-		if (num_invalidated_mw != num_bound_mw) {
-			DP_NOTICE(p_hwfn,
-				  "number of invalidate memory windows is different from bounded ones\n");
-			return -EINVAL;
-		}
+		rc = qed_roce_sp_destroy_qp_requester(p_hwfn, qp);
 	} else {
 		DP_VERBOSE(p_hwfn, QED_MSG_RDMA, "0\n");
 	}

drivers/scsi/qedi/qedi_iscsi.c

@@ -471,14 +471,8 @@ static u16 qedi_calc_mss(u16 pmtu, u8 is_ipv6, u8 tcp_ts_en, u8 vlan_en)
 	else
 		hdrs += IPV4_HDR_LEN;
 
-	if (vlan_en)
-		hdrs += VLAN_LEN;
-
 	mss = pmtu - hdrs;
 
-	if (tcp_ts_en)
-		mss -= TCP_OPTION_LEN;
-
 	if (!mss)
 		mss = DEF_MSS;
 

include/linux/qed/common_hsi.h

@@ -109,8 +109,8 @@
 #define MAX_NUM_LL2_TX_STATS_COUNTERS	48
 
 #define FW_MAJOR_VERSION	8
-#define FW_MINOR_VERSION	33
-#define FW_REVISION_VERSION     11
+#define FW_MINOR_VERSION        37
+#define FW_REVISION_VERSION     2
 #define FW_ENGINEERING_VERSION	0
 
 /***********************/

include/linux/qed/iscsi_common.h

@@ -799,8 +799,8 @@ struct e4_mstorm_iscsi_task_ag_ctx {
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_CONNECTION_TYPE_SHIFT	0
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_EXIST_IN_QM0_MASK		0x1
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_EXIST_IN_QM0_SHIFT		4
-#define E4_MSTORM_ISCSI_TASK_AG_CTX_BIT1_MASK			0x1
-#define E4_MSTORM_ISCSI_TASK_AG_CTX_BIT1_SHIFT			5
+#define E4_MSTORM_ISCSI_TASK_AG_CTX_CONN_CLEAR_SQ_FLAG_MASK	0x1
+#define E4_MSTORM_ISCSI_TASK_AG_CTX_CONN_CLEAR_SQ_FLAG_SHIFT	5
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_VALID_MASK			0x1
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_VALID_SHIFT			6
 #define E4_MSTORM_ISCSI_TASK_AG_CTX_TASK_CLEANUP_FLAG_MASK	0x1
@@ -849,8 +849,8 @@ struct e4_ustorm_iscsi_task_ag_ctx {
 #define E4_USTORM_ISCSI_TASK_AG_CTX_CONNECTION_TYPE_SHIFT	0
 #define E4_USTORM_ISCSI_TASK_AG_CTX_EXIST_IN_QM0_MASK		0x1
 #define E4_USTORM_ISCSI_TASK_AG_CTX_EXIST_IN_QM0_SHIFT		4
-#define E4_USTORM_ISCSI_TASK_AG_CTX_BIT1_MASK			0x1
-#define E4_USTORM_ISCSI_TASK_AG_CTX_BIT1_SHIFT			5
+#define E4_USTORM_ISCSI_TASK_AG_CTX_CONN_CLEAR_SQ_FLAG_MASK     0x1
+#define E4_USTORM_ISCSI_TASK_AG_CTX_CONN_CLEAR_SQ_FLAG_SHIFT    5
 #define E4_USTORM_ISCSI_TASK_AG_CTX_HQ_SCANNED_CF_MASK		0x3
 #define E4_USTORM_ISCSI_TASK_AG_CTX_HQ_SCANNED_CF_SHIFT		6
 	u8 flags1;

include/linux/qed/qed_rdma_if.h

@@ -65,8 +65,7 @@ enum qed_roce_qp_state {
 enum qed_rdma_tid_type {
 	QED_RDMA_TID_REGISTERED_MR,
 	QED_RDMA_TID_FMR,
-	QED_RDMA_TID_MW_TYPE1,
-	QED_RDMA_TID_MW_TYPE2A
+	QED_RDMA_TID_MW
 };
 
 struct qed_rdma_events {
@@ -280,7 +279,6 @@ struct qed_rdma_register_tid_in_params {
 
 	bool dif_enabled;
 	u64 dif_error_addr;
-	u64 dif_runt_addr;
 };
 
 struct qed_rdma_create_cq_in_params {

include/linux/qed/roce_common.h

@@ -43,6 +43,7 @@
 #define ROCE_MAX_QPS			(32 * 1024)
 #define ROCE_DCQCN_NP_MAX_QPS		(64)
 #define ROCE_DCQCN_RP_MAX_QPS		(64)
+#define ROCE_LKEY_MW_DIF_EN_BIT		(28)
 
 /* Affiliated asynchronous events / errors enumeration */
 enum roce_async_events_type {