diff --git a/providers/bnxt_re/bnxt_re-abi.h b/providers/bnxt_re/bnxt_re-abi.h
index 6382dcb74..f294da2d8 100644
--- a/providers/bnxt_re/bnxt_re-abi.h
+++ b/providers/bnxt_re/bnxt_re-abi.h
@@ -167,6 +167,15 @@ enum bnxt_re_psns_mask {
 	BNXT_RE_PSNS_FLAGS_SHIFT	= 0x18
 };
 
+enum bnxt_re_msns_mask {
+	BNXT_RE_SQ_MSN_SEARCH_START_PSN_MASK	= 0xFFFFFFUL,
+	BNXT_RE_SQ_MSN_SEARCH_START_PSN_SHIFT	= 0,
+	BNXT_RE_SQ_MSN_SEARCH_NEXT_PSN_MASK	= 0xFFFFFF000000ULL,
+	BNXT_RE_SQ_MSN_SEARCH_NEXT_PSN_SHIFT	= 0x18,
+	BNXT_RE_SQ_MSN_SEARCH_START_IDX_MASK	= 0xFFFF000000000000ULL,
+	BNXT_RE_SQ_MSN_SEARCH_START_IDX_SHIFT	= 0x30
+};
+
 enum bnxt_re_bcqe_mask {
 	BNXT_RE_BCQE_PH_MASK		= 0x01,
 	BNXT_RE_BCQE_TYPE_MASK		= 0x0F,
@@ -301,6 +310,11 @@ struct bnxt_re_psns_ext {
 	__u32 rsvd1;
 };
 
+/* sq_msn_search (size:64b/8B) */
+struct bnxt_re_msns {
+	__le64  start_idx_next_psn_start_psn;
+};
+
 struct bnxt_re_sge {
 	__le64 pa;
 	__le32 lkey;
diff --git a/providers/bnxt_re/main.h b/providers/bnxt_re/main.h
index 031db17d2..f51ad1955 100644
--- a/providers/bnxt_re/main.h
+++ b/providers/bnxt_re/main.h
@@ -601,4 +601,6 @@ static inline void bnxt_re_sub_sec_busy_wait(uint32_t nsec)
 			break;
 	}
 }
+
+#define BNXT_RE_HW_RETX(a) ((a)->comp_mask & BNXT_RE_COMP_MASK_UCNTX_HW_RETX_ENABLED)
 #endif
diff --git a/providers/bnxt_re/memory.h b/providers/bnxt_re/memory.h
index 329e14a8e..a3c14631b 100644
--- a/providers/bnxt_re/memory.h
+++ b/providers/bnxt_re/memory.h
@@ -51,6 +51,8 @@ struct bnxt_re_queue {
 	uint32_t head;
 	uint32_t tail;
 	uint32_t stride;
+	void *pad; /* to hold the padding area */
+	uint32_t pad_stride_log2;
 	/* Represents the difference between the real queue depth allocated in
 	 * HW and the user requested queue depth and is used to correctly flag
 	 * queue full condition based on user supplied queue depth.
@@ -62,6 +64,8 @@ struct bnxt_re_queue {
 	uint32_t esize;
 	uint32_t max_slots;
 	pthread_spinlock_t qlock;
+	uint32_t msn;
+	uint32_t msn_tbl_sz;
 };
 
 int bnxt_re_alloc_aligned(struct bnxt_re_queue *que, uint32_t pg_size);
diff --git a/providers/bnxt_re/verbs.c b/providers/bnxt_re/verbs.c
index 709abff56..7d4a8a9fa 100644
--- a/providers/bnxt_re/verbs.c
+++ b/providers/bnxt_re/verbs.c
@@ -47,6 +47,7 @@
 #include <sys/mman.h>
 #include <netinet/in.h>
 #include <unistd.h>
+#include <ccan/ilog.h>
 
 #include <util/compiler.h>
 #include <util/util.h>
@@ -1182,10 +1183,11 @@ static int bnxt_re_alloc_queues(struct bnxt_re_context *cntx,
 	que->depth = nslots;
 	que->diff = (diff * que->esize) / que->stride;
 
+	que->pad = (que->va + que->depth * que->stride);
 	/* psn_depth extra entries of size que->stride */
-	psn_size = qp->cctx->gen_p5_p7 ? sizeof(struct bnxt_re_psns_ext) :
-				      sizeof(struct bnxt_re_psns);
+	psn_size = bnxt_re_get_psne_size(qp->cntx);
 	psn_depth = (nswr * psn_size) / que->stride;
+	que->pad_stride_log2 = (uint32_t)ilog32(psn_size);
 	if ((nswr * psn_size) % que->stride)
 		psn_depth++;
 	que->depth += psn_depth;
@@ -1218,6 +1220,17 @@ static int bnxt_re_alloc_queues(struct bnxt_re_context *cntx,
 			swque[indx].psns = (struct bnxt_re_psns *)psns_ext;
 		}
 	}
+	/* Init and adjust MSN table size according to qp mode */
+	if (!BNXT_RE_HW_RETX(qp->cntx))
+		goto skip_msn;
+	que->msn = 0;
+	que->msn_tbl_sz = 0;
+	if (qp->qpmode & BNXT_RE_WQE_MODE_VARIABLE)
+		que->msn_tbl_sz = roundup_pow_of_two(nslots) / 2;
+	else
+		que->msn_tbl_sz = roundup_pow_of_two(nswr);
+skip_msn:
+
 	qp->cap.max_swr = nswr;
 	pthread_spin_init(&que->qlock, PTHREAD_PROCESS_PRIVATE);
 
@@ -1588,6 +1601,44 @@ static int bnxt_re_build_tx_sge(struct bnxt_re_queue *que, uint32_t *idx,
 	return bnxt_re_put_tx_sge(que, idx, wr->sg_list, wr->num_sge);
 }
 
+static void *bnxt_re_pull_psn_buff(struct bnxt_re_queue *que, bool hw_retx)
+{
+	if (hw_retx)
+		return (void *)(que->pad + ((que->msn) << que->pad_stride_log2));
+	return (void *)(que->pad + ((*que->dbtail) << que->pad_stride_log2));
+}
+
+static void bnxt_re_fill_psns_for_msntbl(struct bnxt_re_qp *qp, uint32_t len, uint32_t st_idx)
+{
+	uint32_t npsn = 0, start_psn = 0, next_psn = 0;
+	struct bnxt_re_msns *msns;
+	uint32_t pkt_cnt = 0;
+
+	msns = bnxt_re_pull_psn_buff(qp->jsqq->hwque, true);
+	msns->start_idx_next_psn_start_psn = 0;
+
+	if (qp->qptyp == IBV_QPT_RC) {
+		start_psn = qp->sq_psn;
+		pkt_cnt = (len / qp->mtu);
+		if (len % qp->mtu)
+			pkt_cnt++;
+		/* Increment the psn even for 0 len packets
+		 * e.g. for opcode rdma-write-with-imm-data
+		 * with length field = 0
+		 */
+		if (len == 0)
+			pkt_cnt = 1;
+		/* make it 24 bit */
+		next_psn = qp->sq_psn + pkt_cnt;
+		npsn = next_psn;
+		qp->sq_psn = next_psn;
+		msns->start_idx_next_psn_start_psn |=
+			bnxt_re_update_msn_tbl(st_idx, npsn, start_psn);
+		qp->jsqq->hwque->msn++;
+		qp->jsqq->hwque->msn %= qp->jsqq->hwque->msn_tbl_sz;
+	}
+}
+
 static void bnxt_re_fill_psns(struct bnxt_re_qp *qp, struct bnxt_re_wrid *wrid,
 			      uint32_t len)
 {
@@ -1810,7 +1861,10 @@ int bnxt_re_post_send(struct ibv_qp *ibvqp, struct ibv_send_wr *wr,
 		bnxt_re_fill_wrid(wrid, wr->wr_id, bytes,
 				  sig, sq->tail, slots);
 		wrid->wc_opcd = bnxt_re_ibv_wr_to_wc_opcd(wr->opcode);
-		bnxt_re_fill_psns(qp, wrid, bytes);
+		if (BNXT_RE_HW_RETX(qp->cntx))
+			bnxt_re_fill_psns_for_msntbl(qp, bytes, *sq->dbtail);
+		else
+			bnxt_re_fill_psns(qp, wrid, bytes);
 		bnxt_re_jqq_mod_start(qp->jsqq, swq_idx);
 		bnxt_re_incr_tail(sq, slots);
 		ring_db = true;
diff --git a/providers/bnxt_re/verbs.h b/providers/bnxt_re/verbs.h
index c42c2926f..52adf28ff 100644
--- a/providers/bnxt_re/verbs.h
+++ b/providers/bnxt_re/verbs.h
@@ -60,6 +60,20 @@ struct bnxt_re_work_compl {
 	struct ibv_wc wc;
 };
 
+static inline uint8_t bnxt_re_get_psne_size(struct bnxt_re_context *cntx)
+{
+	return (BNXT_RE_HW_RETX(cntx)) ? sizeof(struct bnxt_re_msns) :
+					      (cntx->cctx.gen_p5_p7) ?
+					      sizeof(struct bnxt_re_psns_ext) :
+					      sizeof(struct bnxt_re_psns);
+}
+
+static inline uint32_t bnxt_re_get_npsn(uint8_t mode, uint32_t nwr,
+					uint32_t slots)
+{
+	return mode == BNXT_RE_WQE_MODE_VARIABLE ? slots : nwr;
+}
+
 int bnxt_re_query_device(struct ibv_context *context,
 			 const struct ibv_query_device_ex_input *input,
 			 struct ibv_device_attr_ex *attr, size_t attr_size);
@@ -107,4 +121,14 @@ int bnxt_re_destroy_ah(struct ibv_ah *ibvah);
 void bnxt_re_async_event(struct ibv_context *context,
 			 struct ibv_async_event *event);
 
+static inline __le64 bnxt_re_update_msn_tbl(uint32_t st_idx, uint32_t npsn, uint32_t start_psn)
+{
+	/* Adjust the field values to their respective ofsets */
+	return htole64((((uint64_t)(st_idx) << BNXT_RE_SQ_MSN_SEARCH_START_IDX_SHIFT) &
+		 BNXT_RE_SQ_MSN_SEARCH_START_IDX_MASK) |
+		 (((uint64_t)(npsn) << BNXT_RE_SQ_MSN_SEARCH_NEXT_PSN_SHIFT) &
+		 BNXT_RE_SQ_MSN_SEARCH_NEXT_PSN_MASK) |
+		 (((start_psn) << BNXT_RE_SQ_MSN_SEARCH_START_PSN_SHIFT) &
+		 BNXT_RE_SQ_MSN_SEARCH_START_PSN_MASK));
+}
 #endif /* __BNXT_RE_VERBS_H__ */