From 48f996d4adf15a0a0af8b8184d3ec6042a684ea4 Mon Sep 17 00:00:00 2001
From: Chandramohan Akula <chandramohan.akula@broadcom.com>
Date: Mon, 23 Oct 2023 07:03:23 -0700
Subject: RDMA/bnxt_re: Remove roundup_pow_of_two depth for all hardware queue
 resources

Rounding up the queue depth to power of two is not a hardware requirement.
In order to optimize the per connection memory usage, removing drivers
implementation which round up to the queue depths to the power of 2.

Implements a mask to maintain backward compatibility with older
library.

Signed-off-by: Chandramohan Akula <chandramohan.akula@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1698069803-1787-3-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 include/uapi/rdma/bnxt_re-abi.h | 9 +++++++++
 1 file changed, 9 insertions(+)

(limited to 'include/uapi')

diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 6e7c67a0cca3..a1b896d6d940 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -54,6 +54,7 @@ enum {
 	BNXT_RE_UCNTX_CMASK_HAVE_MODE = 0x02ULL,
 	BNXT_RE_UCNTX_CMASK_WC_DPI_ENABLED = 0x04ULL,
 	BNXT_RE_UCNTX_CMASK_DBR_PACING_ENABLED = 0x08ULL,
+	BNXT_RE_UCNTX_CMASK_POW2_DISABLED = 0x10ULL,
 };
 
 enum bnxt_re_wqe_mode {
@@ -62,6 +63,14 @@ enum bnxt_re_wqe_mode {
 	BNXT_QPLIB_WQE_MODE_INVALID	= 0x02,
 };
 
+enum {
+	BNXT_RE_COMP_MASK_REQ_UCNTX_POW2_SUPPORT = 0x01,
+};
+
+struct bnxt_re_uctx_req {
+	__aligned_u64 comp_mask;
+};
+
 struct bnxt_re_uctx_resp {
 	__u32 dev_id;
 	__u32 max_qp;
-- 
cgit v1.2.3


From d3f4020a213e1cb125eed2363fca372a23f7de7a Mon Sep 17 00:00:00 2001
From: Junxian Huang <huangjunxian6@hisilicon.com>
Date: Thu, 7 Dec 2023 19:42:28 +0800
Subject: RDMA/hns: Response dmac to userspace

While creating AH, dmac is already resolved in kernel. Response dmac
to userspace so that userspace doesn't need to resolve dmac repeatedly.

Signed-off-by: Junxian Huang <huangjunxian6@hisilicon.com>
Link: https://lore.kernel.org/r/20231207114231.2872104-3-huangjunxian6@hisilicon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/hns/hns_roce_ah.c | 7 +++++++
 include/uapi/rdma/hns-abi.h             | 5 +++++
 2 files changed, 12 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/hns/hns_roce_ah.c b/drivers/infiniband/hw/hns/hns_roce_ah.c
index fbf046982374..b4209b6aed8d 100644
--- a/drivers/infiniband/hw/hns/hns_roce_ah.c
+++ b/drivers/infiniband/hw/hns/hns_roce_ah.c
@@ -57,6 +57,7 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 	struct rdma_ah_attr *ah_attr = init_attr->ah_attr;
 	const struct ib_global_route *grh = rdma_ah_read_grh(ah_attr);
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibah->device);
+	struct hns_roce_ib_create_ah_resp resp = {};
 	struct hns_roce_ah *ah = to_hr_ah(ibah);
 	int ret = 0;
 	u32 max_sl;
@@ -97,6 +98,12 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 		ah->av.vlan_en = ah->av.vlan_id < VLAN_N_VID;
 	}
 
+	if (udata) {
+		memcpy(resp.dmac, ah_attr->roce.dmac, ETH_ALEN);
+		ret = ib_copy_to_udata(udata, &resp,
+				       min(udata->outlen, sizeof(resp)));
+	}
+
 err_out:
 	if (ret)
 		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_AH_CREATE_ERR_CNT]);
diff --git a/include/uapi/rdma/hns-abi.h b/include/uapi/rdma/hns-abi.h
index ce0f37f83416..c996e151081e 100644
--- a/include/uapi/rdma/hns-abi.h
+++ b/include/uapi/rdma/hns-abi.h
@@ -125,4 +125,9 @@ struct hns_roce_ib_alloc_pd_resp {
 	__u32 pdn;
 };
 
+struct hns_roce_ib_create_ah_resp {
+	__u8 dmac[6];
+	__u8 reserved[2];
+};
+
 #endif /* HNS_ABI_USER_H */
-- 
cgit v1.2.3


From 07f830ae4913d0b986c8c0ff88a7d597948b9bd8 Mon Sep 17 00:00:00 2001
From: Selvin Xavier <selvin.xavier@broadcom.com>
Date: Thu, 7 Dec 2023 02:47:40 -0800
Subject: RDMA/bnxt_re: Adds MSN table capability for Gen P7 adapters

GenP7 HW expects an MSN table instead of PSN table. Check
for the HW retransmission capability and populate the MSN
table if HW retansmission is supported.

Signed-off-by: Damodharam Ammepalli <damodharam.ammepalli@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1701946060-13931-7-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/bnxt_re/qplib_fp.c   | 67 +++++++++++++++++++++++++++---
 drivers/infiniband/hw/bnxt_re/qplib_fp.h   | 14 +++++++
 drivers/infiniband/hw/bnxt_re/qplib_rcfw.c |  2 +
 drivers/infiniband/hw/bnxt_re/qplib_res.h  |  9 ++++
 include/uapi/rdma/bnxt_re-abi.h            |  1 +
 5 files changed, 87 insertions(+), 6 deletions(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.c b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
index 177c6c185f0c..c98e04fe2ddd 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.c
@@ -982,6 +982,9 @@ int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp)
 	u32 tbl_indx;
 	u16 nsge;
 
+	if (res->dattr)
+		qp->dev_cap_flags = res->dattr->dev_cap_flags;
+
 	sq->dbinfo.flags = 0;
 	bnxt_qplib_rcfw_cmd_prep((struct cmdq_base *)&req,
 				 CMDQ_BASE_OPCODE_CREATE_QP,
@@ -997,6 +1000,11 @@ int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp)
 		psn_sz = bnxt_qplib_is_chip_gen_p5_p7(res->cctx) ?
 			 sizeof(struct sq_psn_search_ext) :
 			 sizeof(struct sq_psn_search);
+
+		if (BNXT_RE_HW_RETX(qp->dev_cap_flags)) {
+			psn_sz = sizeof(struct sq_msn_search);
+			qp->msn = 0;
+		}
 	}
 
 	hwq_attr.res = res;
@@ -1005,6 +1013,13 @@ int bnxt_qplib_create_qp(struct bnxt_qplib_res *res, struct bnxt_qplib_qp *qp)
 	hwq_attr.depth = bnxt_qplib_get_depth(sq);
 	hwq_attr.aux_stride = psn_sz;
 	hwq_attr.aux_depth = bnxt_qplib_set_sq_size(sq, qp->wqe_mode);
+	/* Update msn tbl size */
+	if (BNXT_RE_HW_RETX(qp->dev_cap_flags) && psn_sz) {
+		hwq_attr.aux_depth = roundup_pow_of_two(bnxt_qplib_set_sq_size(sq, qp->wqe_mode));
+		qp->msn_tbl_sz = hwq_attr.aux_depth;
+		qp->msn = 0;
+	}
+
 	hwq_attr.type = HWQ_TYPE_QUEUE;
 	rc = bnxt_qplib_alloc_init_hwq(&sq->hwq, &hwq_attr);
 	if (rc)
@@ -1587,6 +1602,27 @@ void *bnxt_qplib_get_qp1_rq_buf(struct bnxt_qplib_qp *qp,
 	return NULL;
 }
 
+/* Fil the MSN table into the next psn row */
+static void bnxt_qplib_fill_msn_search(struct bnxt_qplib_qp *qp,
+				       struct bnxt_qplib_swqe *wqe,
+				       struct bnxt_qplib_swq *swq)
+{
+	struct sq_msn_search *msns;
+	u32 start_psn, next_psn;
+	u16 start_idx;
+
+	msns = (struct sq_msn_search *)swq->psn_search;
+	msns->start_idx_next_psn_start_psn = 0;
+
+	start_psn = swq->start_psn;
+	next_psn = swq->next_psn;
+	start_idx = swq->slot_idx;
+	msns->start_idx_next_psn_start_psn |=
+		bnxt_re_update_msn_tbl(start_idx, next_psn, start_psn);
+	qp->msn++;
+	qp->msn %= qp->msn_tbl_sz;
+}
+
 static void bnxt_qplib_fill_psn_search(struct bnxt_qplib_qp *qp,
 				       struct bnxt_qplib_swqe *wqe,
 				       struct bnxt_qplib_swq *swq)
@@ -1598,6 +1634,12 @@ static void bnxt_qplib_fill_psn_search(struct bnxt_qplib_qp *qp,
 
 	if (!swq->psn_search)
 		return;
+	/* Handle MSN differently on cap flags  */
+	if (BNXT_RE_HW_RETX(qp->dev_cap_flags)) {
+		bnxt_qplib_fill_msn_search(qp, wqe, swq);
+		return;
+	}
+	psns = (struct sq_psn_search *)swq->psn_search;
 	psns = swq->psn_search;
 	psns_ext = swq->psn_ext;
 
@@ -1706,8 +1748,8 @@ static u16 bnxt_qplib_required_slots(struct bnxt_qplib_qp *qp,
 	return slot;
 }
 
-static void bnxt_qplib_pull_psn_buff(struct bnxt_qplib_q *sq,
-				     struct bnxt_qplib_swq *swq)
+static void bnxt_qplib_pull_psn_buff(struct bnxt_qplib_qp *qp, struct bnxt_qplib_q *sq,
+				     struct bnxt_qplib_swq *swq, bool hw_retx)
 {
 	struct bnxt_qplib_hwq *hwq;
 	u32 pg_num, pg_indx;
@@ -1718,6 +1760,11 @@ static void bnxt_qplib_pull_psn_buff(struct bnxt_qplib_q *sq,
 	if (!hwq->pad_pg)
 		return;
 	tail = swq->slot_idx / sq->dbinfo.max_slot;
+	if (hw_retx) {
+		/* For HW retx use qp msn index */
+		tail = qp->msn;
+		tail %= qp->msn_tbl_sz;
+	}
 	pg_num = (tail + hwq->pad_pgofft) / (PAGE_SIZE / hwq->pad_stride);
 	pg_indx = (tail + hwq->pad_pgofft) % (PAGE_SIZE / hwq->pad_stride);
 	buff = (void *)(hwq->pad_pg[pg_num] + pg_indx * hwq->pad_stride);
@@ -1742,6 +1789,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 	struct bnxt_qplib_swq *swq;
 	bool sch_handler = false;
 	u16 wqe_sz, qdf = 0;
+	bool msn_update;
 	void *base_hdr;
 	void *ext_hdr;
 	__le32 temp32;
@@ -1769,7 +1817,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 	}
 
 	swq = bnxt_qplib_get_swqe(sq, &wqe_idx);
-	bnxt_qplib_pull_psn_buff(sq, swq);
+	bnxt_qplib_pull_psn_buff(qp, sq, swq, BNXT_RE_HW_RETX(qp->dev_cap_flags));
 
 	idx = 0;
 	swq->slot_idx = hwq->prod;
@@ -1801,6 +1849,8 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 					       &idx);
 	if (data_len < 0)
 		goto queue_err;
+	/* Make sure we update MSN table only for wired wqes */
+	msn_update = true;
 	/* Specifics */
 	switch (wqe->type) {
 	case BNXT_QPLIB_SWQE_TYPE_SEND:
@@ -1841,6 +1891,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 						      SQ_SEND_DST_QP_MASK);
 			ext_sqe->avid = cpu_to_le32(wqe->send.avid &
 						    SQ_SEND_AVID_MASK);
+			msn_update = false;
 		} else {
 			sqe->length = cpu_to_le32(data_len);
 			if (qp->mtu)
@@ -1898,7 +1949,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 		sqe->wqe_type = wqe->type;
 		sqe->flags = wqe->flags;
 		sqe->inv_l_key = cpu_to_le32(wqe->local_inv.inv_l_key);
-
+		msn_update = false;
 		break;
 	}
 	case BNXT_QPLIB_SWQE_TYPE_FAST_REG_MR:
@@ -1930,6 +1981,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 						PTU_PTE_VALID);
 		ext_sqe->pblptr = cpu_to_le64(wqe->frmr.pbl_dma_ptr);
 		ext_sqe->va = cpu_to_le64(wqe->frmr.va);
+		msn_update = false;
 
 		break;
 	}
@@ -1947,6 +1999,7 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 		sqe->l_key = cpu_to_le32(wqe->bind.r_key);
 		ext_sqe->va = cpu_to_le64(wqe->bind.va);
 		ext_sqe->length_lo = cpu_to_le32(wqe->bind.length);
+		msn_update = false;
 		break;
 	}
 	default:
@@ -1954,8 +2007,10 @@ int bnxt_qplib_post_send(struct bnxt_qplib_qp *qp,
 		rc = -EINVAL;
 		goto done;
 	}
-	swq->next_psn = sq->psn & BTH_PSN_MASK;
-	bnxt_qplib_fill_psn_search(qp, wqe, swq);
+	if (!BNXT_RE_HW_RETX(qp->dev_cap_flags) || msn_update) {
+		swq->next_psn = sq->psn & BTH_PSN_MASK;
+		bnxt_qplib_fill_psn_search(qp, wqe, swq);
+	}
 queue_err:
 	bnxt_qplib_swq_mod_start(sq, wqe_idx);
 	bnxt_qplib_hwq_incr_prod(&sq->dbinfo, hwq, swq->slots);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_fp.h b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
index 8a6bea201b29..967c6691e413 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_fp.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_fp.h
@@ -338,6 +338,9 @@ struct bnxt_qplib_qp {
 	dma_addr_t			rq_hdr_buf_map;
 	struct list_head		sq_flush;
 	struct list_head		rq_flush;
+	u32				msn;
+	u32				msn_tbl_sz;
+	u16				dev_cap_flags;
 };
 
 #define BNXT_QPLIB_MAX_CQE_ENTRY_SIZE	sizeof(struct cq_base)
@@ -627,4 +630,15 @@ static inline u16 bnxt_qplib_calc_ilsize(struct bnxt_qplib_swqe *wqe, u16 max)
 
 	return size;
 }
+
+/* MSN table update inlin */
+static inline uint64_t bnxt_re_update_msn_tbl(u32 st_idx, u32 npsn, u32 start_psn)
+{
+	return cpu_to_le64((((u64)(st_idx) << SQ_MSN_SEARCH_START_IDX_SFT) &
+		SQ_MSN_SEARCH_START_IDX_MASK) |
+		(((u64)(npsn) << SQ_MSN_SEARCH_NEXT_PSN_SFT) &
+		SQ_MSN_SEARCH_NEXT_PSN_MASK) |
+		(((start_psn) << SQ_MSN_SEARCH_START_PSN_SFT) &
+		SQ_MSN_SEARCH_START_PSN_MASK));
+}
 #endif /* __BNXT_QPLIB_FP_H__ */
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
index 403b6797d9c2..0ea7ccc70679 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
+++ b/drivers/infiniband/hw/bnxt_re/qplib_rcfw.c
@@ -905,6 +905,8 @@ config_vf_res:
 	req.max_gid_per_vf = cpu_to_le32(ctx->vf_res.max_gid_per_vf);
 
 skip_ctx_setup:
+	if (BNXT_RE_HW_RETX(rcfw->res->dattr->dev_cap_flags))
+		req.flags |= CMDQ_INITIALIZE_FW_FLAGS_HW_REQUESTER_RETX_SUPPORTED;
 	req.stat_ctx_id = cpu_to_le32(ctx->stats.fw_id);
 	bnxt_qplib_fill_cmdqmsg(&msg, &req, &resp, NULL, sizeof(req), sizeof(resp), 0);
 	rc = bnxt_qplib_rcfw_send_message(rcfw, &msg);
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_res.h b/drivers/infiniband/hw/bnxt_re/qplib_res.h
index c228870eed29..382d89fa7d16 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_res.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_res.h
@@ -539,6 +539,15 @@ static inline bool _is_ext_stats_supported(u16 dev_cap_flags)
 		CREQ_QUERY_FUNC_RESP_SB_EXT_STATS;
 }
 
+static inline bool _is_hw_retx_supported(u16 dev_cap_flags)
+{
+	return dev_cap_flags &
+		(CREQ_QUERY_FUNC_RESP_SB_HW_REQUESTER_RETX_ENABLED |
+		 CREQ_QUERY_FUNC_RESP_SB_HW_RESPONDER_RETX_ENABLED);
+}
+
+#define BNXT_RE_HW_RETX(a) _is_hw_retx_supported((a))
+
 static inline u8 bnxt_qplib_dbr_pacing_en(struct bnxt_qplib_chip_ctx *cctx)
 {
 	return cctx->modes.dbr_pacing;
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index a1b896d6d940..3342276aeac1 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -55,6 +55,7 @@ enum {
 	BNXT_RE_UCNTX_CMASK_WC_DPI_ENABLED = 0x04ULL,
 	BNXT_RE_UCNTX_CMASK_DBR_PACING_ENABLED = 0x08ULL,
 	BNXT_RE_UCNTX_CMASK_POW2_DISABLED = 0x10ULL,
+	BNXT_RE_COMP_MASK_UCNTX_HW_RETX_ENABLED = 0x40,
 };
 
 enum bnxt_re_wqe_mode {
-- 
cgit v1.2.3


From a429ec96c07f3020af12029acefc46f42ff5c91c Mon Sep 17 00:00:00 2001
From: Shun Hao <shunh@nvidia.com>
Date: Wed, 6 Dec 2023 16:01:35 +0200
Subject: RDMA/mlx5: Support handling of SW encap ICM area

New type for this ICM area, now the user can allocate/deallocate
the new type of SW encap ICM memory, to store the encap header data
which are managed by SW.

Signed-off-by: Shun Hao <shunh@nvidia.com>
Link: https://lore.kernel.org/r/546fe43fc700240709e30acf7713ec6834d652bd.1701871118.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/mlx5/dm.c           | 5 +++++
 drivers/infiniband/hw/mlx5/mr.c           | 1 +
 include/linux/mlx5/driver.h               | 1 +
 include/uapi/rdma/mlx5_user_ioctl_verbs.h | 1 +
 4 files changed, 8 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/mlx5/dm.c b/drivers/infiniband/hw/mlx5/dm.c
index 3669c90b2dad..b4c97fb62abf 100644
--- a/drivers/infiniband/hw/mlx5/dm.c
+++ b/drivers/infiniband/hw/mlx5/dm.c
@@ -341,6 +341,8 @@ static enum mlx5_sw_icm_type get_icm_type(int uapi_type)
 		return MLX5_SW_ICM_TYPE_HEADER_MODIFY;
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_PATTERN_SW_ICM:
 		return MLX5_SW_ICM_TYPE_HEADER_MODIFY_PATTERN;
+	case MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM:
+		return MLX5_SW_ICM_TYPE_SW_ENCAP;
 	case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
 	default:
 		return MLX5_SW_ICM_TYPE_STEERING;
@@ -364,6 +366,7 @@ static struct ib_dm *handle_alloc_dm_sw_icm(struct ib_ucontext *ctx,
 	switch (type) {
 	case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM:
+	case MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM:
 		if (!(MLX5_CAP_FLOWTABLE_NIC_RX(dev, sw_owner) ||
 		      MLX5_CAP_FLOWTABLE_NIC_TX(dev, sw_owner) ||
 		      MLX5_CAP_FLOWTABLE_NIC_RX(dev, sw_owner_v2) ||
@@ -438,6 +441,7 @@ struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
 	case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_PATTERN_SW_ICM:
+	case MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM:
 		return handle_alloc_dm_sw_icm(context, attr, attrs, type);
 	default:
 		return ERR_PTR(-EOPNOTSUPP);
@@ -491,6 +495,7 @@ static int mlx5_ib_dealloc_dm(struct ib_dm *ibdm,
 	case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_PATTERN_SW_ICM:
+	case MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM:
 		return mlx5_dm_icm_dealloc(ctx, to_icm(ibdm));
 	default:
 		return -EOPNOTSUPP;
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index 18e459b55746..a8ee2ca1f4a1 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1347,6 +1347,7 @@ struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
 	case MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM:
 	case MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_PATTERN_SW_ICM:
+	case MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM:
 		if (attr->access_flags & ~MLX5_IB_DM_SW_ICM_ALLOWED_ACCESS)
 			return ERR_PTR(-EINVAL);
 
diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h
index d2b8d4a74a30..96cb8845682d 100644
--- a/include/linux/mlx5/driver.h
+++ b/include/linux/mlx5/driver.h
@@ -688,6 +688,7 @@ enum mlx5_sw_icm_type {
 	MLX5_SW_ICM_TYPE_STEERING,
 	MLX5_SW_ICM_TYPE_HEADER_MODIFY,
 	MLX5_SW_ICM_TYPE_HEADER_MODIFY_PATTERN,
+	MLX5_SW_ICM_TYPE_SW_ENCAP,
 };
 
 #define MLX5_MAX_RESERVED_GIDS 8
diff --git a/include/uapi/rdma/mlx5_user_ioctl_verbs.h b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
index 7af9e09ea556..3189c7f08d17 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_verbs.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -64,6 +64,7 @@ enum mlx5_ib_uapi_dm_type {
 	MLX5_IB_UAPI_DM_TYPE_STEERING_SW_ICM,
 	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_SW_ICM,
 	MLX5_IB_UAPI_DM_TYPE_HEADER_MODIFY_PATTERN_SW_ICM,
+	MLX5_IB_UAPI_DM_TYPE_ENCAP_SW_ICM,
 };
 
 enum mlx5_ib_uapi_devx_create_event_channel_flags {
-- 
cgit v1.2.3


From d727d27db536faea7178290c677cc0567f647231 Mon Sep 17 00:00:00 2001
From: Mark Bloch <mbloch@nvidia.com>
Date: Wed, 6 Dec 2023 16:01:38 +0200
Subject: RDMA/mlx5: Expose register c0 for RDMA device

This patch introduces improvements for matching egress traffic sent by the
local device. When applicable, all egress traffic from the local vport is
now tagged with the provided value. This enhancement is particularly useful
for FDB steering purposes.

The primary focus of this update is facilitating the transmission of
traffic from the hypervisor to a VF. To achieve this, one must initiate an
SQ on the hypervisor and subsequently create a rule in the FDB that matches
on the eswitch manager vport and the SQN of the aforementioned SQ.

Obtaining the SQN can be had from SQ opened, and the eswitch manager vport
match can be substituted with the register c0 value exposed by this patch.

Signed-off-by: Mark Bloch <mbloch@nvidia.com>
Reviewed-by: Michael Guralnik <michaelgur@nvidia.com>
Link: https://lore.kernel.org/r/aa4120a91c98ff1c44f1213388c744d4cb0324d6.1701871118.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/mlx5/main.c | 24 ++++++++++++++++++++++++
 include/uapi/rdma/mlx5-abi.h      |  2 ++
 2 files changed, 26 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 650a15b6cfbc..c2b557e64290 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -818,6 +818,17 @@ static int mlx5_query_node_desc(struct mlx5_ib_dev *dev, char *node_desc)
 				    MLX5_REG_NODE_DESC, 0, 0);
 }
 
+static void fill_esw_mgr_reg_c0(struct mlx5_core_dev *mdev,
+				struct mlx5_ib_query_device_resp *resp)
+{
+	struct mlx5_eswitch *esw = mdev->priv.eswitch;
+	u16 vport = mlx5_eswitch_manager_vport(mdev);
+
+	resp->reg_c0.value = mlx5_eswitch_get_vport_metadata_for_match(esw,
+								      vport);
+	resp->reg_c0.mask = mlx5_eswitch_get_vport_metadata_mask();
+}
+
 static int mlx5_ib_query_device(struct ib_device *ibdev,
 				struct ib_device_attr *props,
 				struct ib_udata *uhw)
@@ -1209,6 +1220,19 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 			MLX5_CAP_GEN(mdev, log_max_dci_errored_streams);
 	}
 
+	if (offsetofend(typeof(resp), reserved) <= uhw_outlen)
+		resp.response_length += sizeof(resp.reserved);
+
+	if (offsetofend(typeof(resp), reg_c0) <= uhw_outlen) {
+		struct mlx5_eswitch *esw = mdev->priv.eswitch;
+
+		resp.response_length += sizeof(resp.reg_c0);
+
+		if (mlx5_eswitch_mode(mdev) == MLX5_ESWITCH_OFFLOADS &&
+		    mlx5_eswitch_vport_match_metadata_enabled(esw))
+			fill_esw_mgr_reg_c0(mdev, &resp);
+	}
+
 	if (uhw_outlen) {
 		err = ib_copy_to_udata(uhw, &resp, resp.response_length);
 
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index a96b7d2770e1..d4f6a36dffb0 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -37,6 +37,7 @@
 #include <linux/types.h>
 #include <linux/if_ether.h>	/* For ETH_ALEN. */
 #include <rdma/ib_user_ioctl_verbs.h>
+#include <rdma/mlx5_user_ioctl_verbs.h>
 
 enum {
 	MLX5_QP_FLAG_SIGNATURE		= 1 << 0,
@@ -275,6 +276,7 @@ struct mlx5_ib_query_device_resp {
 	__u32	tunnel_offloads_caps; /* enum mlx5_ib_tunnel_offloads */
 	struct  mlx5_ib_dci_streams_caps dci_streams_caps;
 	__u16 reserved;
+	struct mlx5_ib_uapi_reg reg_c0;
 };
 
 enum mlx5_ib_create_cq_flags {
-- 
cgit v1.2.3


From 9b0a7a2cb87d9c430a3588d7d2b6e471200b86ad Mon Sep 17 00:00:00 2001
From: Selvin Xavier <selvin.xavier@broadcom.com>
Date: Wed, 13 Dec 2023 22:31:23 -0800
Subject: RDMA/bnxt_re: Add UAPI to share a page with user space

Gen P7 adapters require to share a toggle value for CQ
and SRQ. This is received by the driver as part of
interrupt notifications and needs to be shared with the
user space. Add a new UAPI infrastructure to get the
shared page for CQ and SRQ.

Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1702535484-26844-2-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/bnxt_re/ib_verbs.c | 105 +++++++++++++++++++++++++++++++
 drivers/infiniband/hw/bnxt_re/ib_verbs.h |   1 +
 include/uapi/rdma/bnxt_re-abi.h          |  26 ++++++++
 3 files changed, 132 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index e7ef099c3edd..758ea02e1d13 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -567,6 +567,7 @@ bnxt_re_mmap_entry_insert(struct bnxt_re_ucontext *uctx, u64 mem_offset,
 	case BNXT_RE_MMAP_WC_DB:
 	case BNXT_RE_MMAP_DBR_BAR:
 	case BNXT_RE_MMAP_DBR_PAGE:
+	case BNXT_RE_MMAP_TOGGLE_PAGE:
 		ret = rdma_user_mmap_entry_insert(&uctx->ib_uctx,
 						  &entry->rdma_entry, PAGE_SIZE);
 		break;
@@ -4254,6 +4255,7 @@ int bnxt_re_mmap(struct ib_ucontext *ib_uctx, struct vm_area_struct *vma)
 					rdma_entry);
 		break;
 	case BNXT_RE_MMAP_DBR_PAGE:
+	case BNXT_RE_MMAP_TOGGLE_PAGE:
 		/* Driver doesn't expect write access for user space */
 		if (vma->vm_flags & VM_WRITE)
 			return -EFAULT;
@@ -4430,8 +4432,111 @@ DECLARE_UVERBS_NAMED_METHOD(BNXT_RE_METHOD_NOTIFY_DRV);
 DECLARE_UVERBS_GLOBAL_METHODS(BNXT_RE_OBJECT_NOTIFY_DRV,
 			      &UVERBS_METHOD(BNXT_RE_METHOD_NOTIFY_DRV));
 
+/* Toggle MEM */
+static int UVERBS_HANDLER(BNXT_RE_METHOD_GET_TOGGLE_MEM)(struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uobject *uobj = uverbs_attr_get_uobject(attrs, BNXT_RE_TOGGLE_MEM_HANDLE);
+	enum bnxt_re_mmap_flag mmap_flag = BNXT_RE_MMAP_TOGGLE_PAGE;
+	enum bnxt_re_get_toggle_mem_type res_type;
+	struct bnxt_re_user_mmap_entry *entry;
+	struct bnxt_re_ucontext *uctx;
+	struct ib_ucontext *ib_uctx;
+	struct bnxt_re_dev *rdev;
+	u64 mem_offset;
+	u64 addr = 0;
+	u32 length;
+	u32 offset;
+	int err;
+
+	ib_uctx = ib_uverbs_get_ucontext(attrs);
+	if (IS_ERR(ib_uctx))
+		return PTR_ERR(ib_uctx);
+
+	err = uverbs_get_const(&res_type, attrs, BNXT_RE_TOGGLE_MEM_TYPE);
+	if (err)
+		return err;
+
+	uctx = container_of(ib_uctx, struct bnxt_re_ucontext, ib_uctx);
+	rdev = uctx->rdev;
+
+	switch (res_type) {
+	case BNXT_RE_CQ_TOGGLE_MEM:
+	case BNXT_RE_SRQ_TOGGLE_MEM:
+		break;
+
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	entry = bnxt_re_mmap_entry_insert(uctx, addr, mmap_flag, &mem_offset);
+	if (!entry)
+		return -ENOMEM;
+
+	uobj->object = entry;
+	uverbs_finalize_uobj_create(attrs, BNXT_RE_TOGGLE_MEM_HANDLE);
+	err = uverbs_copy_to(attrs, BNXT_RE_TOGGLE_MEM_MMAP_PAGE,
+			     &mem_offset, sizeof(mem_offset));
+	if (err)
+		return err;
+
+	err = uverbs_copy_to(attrs, BNXT_RE_TOGGLE_MEM_MMAP_LENGTH,
+			     &length, sizeof(length));
+	if (err)
+		return err;
+
+	err = uverbs_copy_to(attrs, BNXT_RE_TOGGLE_MEM_MMAP_OFFSET,
+			     &offset, sizeof(length));
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int get_toggle_mem_obj_cleanup(struct ib_uobject *uobject,
+				      enum rdma_remove_reason why,
+				      struct uverbs_attr_bundle *attrs)
+{
+	struct  bnxt_re_user_mmap_entry *entry = uobject->object;
+
+	rdma_user_mmap_entry_remove(&entry->rdma_entry);
+	return 0;
+}
+
+DECLARE_UVERBS_NAMED_METHOD(BNXT_RE_METHOD_GET_TOGGLE_MEM,
+			    UVERBS_ATTR_IDR(BNXT_RE_TOGGLE_MEM_HANDLE,
+					    BNXT_RE_OBJECT_GET_TOGGLE_MEM,
+					    UVERBS_ACCESS_NEW,
+					    UA_MANDATORY),
+			    UVERBS_ATTR_CONST_IN(BNXT_RE_TOGGLE_MEM_TYPE,
+						 enum bnxt_re_get_toggle_mem_type,
+						 UA_MANDATORY),
+			    UVERBS_ATTR_PTR_IN(BNXT_RE_TOGGLE_MEM_RES_ID,
+					       UVERBS_ATTR_TYPE(u32),
+					       UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(BNXT_RE_TOGGLE_MEM_MMAP_PAGE,
+						UVERBS_ATTR_TYPE(u64),
+						UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(BNXT_RE_TOGGLE_MEM_MMAP_OFFSET,
+						UVERBS_ATTR_TYPE(u32),
+						UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(BNXT_RE_TOGGLE_MEM_MMAP_LENGTH,
+						UVERBS_ATTR_TYPE(u32),
+						UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD_DESTROY(BNXT_RE_METHOD_RELEASE_TOGGLE_MEM,
+				    UVERBS_ATTR_IDR(BNXT_RE_RELEASE_TOGGLE_MEM_HANDLE,
+						    BNXT_RE_OBJECT_GET_TOGGLE_MEM,
+						    UVERBS_ACCESS_DESTROY,
+						    UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_OBJECT(BNXT_RE_OBJECT_GET_TOGGLE_MEM,
+			    UVERBS_TYPE_ALLOC_IDR(get_toggle_mem_obj_cleanup),
+			    &UVERBS_METHOD(BNXT_RE_METHOD_GET_TOGGLE_MEM),
+			    &UVERBS_METHOD(BNXT_RE_METHOD_RELEASE_TOGGLE_MEM));
+
 const struct uapi_definition bnxt_re_uapi_defs[] = {
 	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(BNXT_RE_OBJECT_ALLOC_PAGE),
 	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(BNXT_RE_OBJECT_NOTIFY_DRV),
+	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(BNXT_RE_OBJECT_GET_TOGGLE_MEM),
 	{}
 };
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index 98baea98fc17..da3fe018f255 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -149,6 +149,7 @@ enum bnxt_re_mmap_flag {
 	BNXT_RE_MMAP_WC_DB,
 	BNXT_RE_MMAP_DBR_PAGE,
 	BNXT_RE_MMAP_DBR_BAR,
+	BNXT_RE_MMAP_TOGGLE_PAGE,
 };
 
 struct bnxt_re_user_mmap_entry {
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 3342276aeac1..9b9eb10cb5e5 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -143,6 +143,7 @@ enum bnxt_re_shpg_offt {
 enum bnxt_re_objects {
 	BNXT_RE_OBJECT_ALLOC_PAGE = (1U << UVERBS_ID_NS_SHIFT),
 	BNXT_RE_OBJECT_NOTIFY_DRV,
+	BNXT_RE_OBJECT_GET_TOGGLE_MEM,
 };
 
 enum bnxt_re_alloc_page_type {
@@ -171,4 +172,29 @@ enum bnxt_re_alloc_page_methods {
 enum bnxt_re_notify_drv_methods {
 	BNXT_RE_METHOD_NOTIFY_DRV = (1U << UVERBS_ID_NS_SHIFT),
 };
+
+/* Toggle mem */
+
+enum bnxt_re_get_toggle_mem_type {
+	BNXT_RE_CQ_TOGGLE_MEM = 0,
+	BNXT_RE_SRQ_TOGGLE_MEM,
+};
+
+enum bnxt_re_var_toggle_mem_attrs {
+	BNXT_RE_TOGGLE_MEM_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	BNXT_RE_TOGGLE_MEM_TYPE,
+	BNXT_RE_TOGGLE_MEM_RES_ID,
+	BNXT_RE_TOGGLE_MEM_MMAP_PAGE,
+	BNXT_RE_TOGGLE_MEM_MMAP_OFFSET,
+	BNXT_RE_TOGGLE_MEM_MMAP_LENGTH,
+};
+
+enum bnxt_re_toggle_mem_attrs {
+	BNXT_RE_RELEASE_TOGGLE_MEM_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+};
+
+enum bnxt_re_toggle_mem_methods {
+	BNXT_RE_METHOD_GET_TOGGLE_MEM = (1U << UVERBS_ID_NS_SHIFT),
+	BNXT_RE_METHOD_RELEASE_TOGGLE_MEM,
+};
 #endif /* __BNXT_RE_UVERBS_ABI_H__*/
-- 
cgit v1.2.3


From e275919d96693c5ca964b20d73a33d52a7e57f04 Mon Sep 17 00:00:00 2001
From: Selvin Xavier <selvin.xavier@broadcom.com>
Date: Wed, 13 Dec 2023 22:31:24 -0800
Subject: RDMA/bnxt_re: Share a page to expose per CQ info with userspace

Gen P7 adapters needs to share a toggle bits information received
in kernel driver with the user space. User space needs this
info during the request notify call back to arm the CQ.

User space application can get this page using the
UAPI routines. Library will mmap this page and get the
toggle bits to be used in the next ARM Doorbell.

Uses a hash list to map the CQ structure from the CQ ID.
CQ structure is retrieved from the hash list while the
library calls the UAPI routine to get the toggle page
mapping. Currently the full page is mapped per CQ. This
can be optimized to enable multiple CQs from the same
application share the same page and different offsets
in the page.

Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://lore.kernel.org/r/1702535484-26844-3-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/bnxt_re/bnxt_re.h   |  3 ++
 drivers/infiniband/hw/bnxt_re/ib_verbs.c  | 61 +++++++++++++++++++++++++++----
 drivers/infiniband/hw/bnxt_re/ib_verbs.h  |  2 +
 drivers/infiniband/hw/bnxt_re/main.c      | 10 ++++-
 drivers/infiniband/hw/bnxt_re/qplib_res.h |  6 +++
 include/uapi/rdma/bnxt_re-abi.h           |  5 +++
 6 files changed, 79 insertions(+), 8 deletions(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index 9fd9849ebdd1..9dca451ed522 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -41,6 +41,7 @@
 #define __BNXT_RE_H__
 #include <rdma/uverbs_ioctl.h>
 #include "hw_counters.h"
+#include <linux/hashtable.h>
 #define ROCE_DRV_MODULE_NAME		"bnxt_re"
 
 #define BNXT_RE_DESC	"Broadcom NetXtreme-C/E RoCE Driver"
@@ -135,6 +136,7 @@ struct bnxt_re_pacing {
 #define BNXT_RE_DB_FIFO_ROOM_SHIFT 15
 #define BNXT_RE_GRC_FIFO_REG_BASE 0x2000
 
+#define MAX_CQ_HASH_BITS		(16)
 struct bnxt_re_dev {
 	struct ib_device		ibdev;
 	struct list_head		list;
@@ -189,6 +191,7 @@ struct bnxt_re_dev {
 	struct bnxt_re_pacing pacing;
 	struct work_struct dbq_fifo_check_work;
 	struct delayed_work dbq_pacing_work;
+	DECLARE_HASHTABLE(cq_hash, MAX_CQ_HASH_BITS);
 };
 
 #define to_bnxt_re_dev(ptr, member)	\
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 758ea02e1d13..7213dc7574d0 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -50,6 +50,7 @@
 #include <rdma/ib_mad.h>
 #include <rdma/ib_cache.h>
 #include <rdma/uverbs_ioctl.h>
+#include <linux/hashtable.h>
 
 #include "bnxt_ulp.h"
 
@@ -2910,14 +2911,20 @@ int bnxt_re_post_recv(struct ib_qp *ib_qp, const struct ib_recv_wr *wr,
 /* Completion Queues */
 int bnxt_re_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 {
-	struct bnxt_re_cq *cq;
+	struct bnxt_qplib_chip_ctx *cctx;
 	struct bnxt_qplib_nq *nq;
 	struct bnxt_re_dev *rdev;
+	struct bnxt_re_cq *cq;
 
 	cq = container_of(ib_cq, struct bnxt_re_cq, ib_cq);
 	rdev = cq->rdev;
 	nq = cq->qplib_cq.nq;
+	cctx = rdev->chip_ctx;
 
+	if (cctx->modes.toggle_bits & BNXT_QPLIB_CQ_TOGGLE_BIT) {
+		free_page((unsigned long)cq->uctx_cq_page);
+		hash_del(&cq->hash_entry);
+	}
 	bnxt_qplib_destroy_cq(&rdev->qplib_res, &cq->qplib_cq);
 	ib_umem_release(cq->umem);
 
@@ -2935,10 +2942,11 @@ int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 	struct bnxt_re_ucontext *uctx =
 		rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
 	struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
-	int rc, entries;
-	int cqe = attr->cqe;
+	struct bnxt_qplib_chip_ctx *cctx;
 	struct bnxt_qplib_nq *nq = NULL;
+	int rc = -ENOMEM, entries;
 	unsigned int nq_alloc_cnt;
+	int cqe = attr->cqe;
 	u32 active_cqs;
 
 	if (attr->flags)
@@ -2951,6 +2959,7 @@ int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 	}
 
 	cq->rdev = rdev;
+	cctx = rdev->chip_ctx;
 	cq->qplib_cq.cq_handle = (u64)(unsigned long)(&cq->qplib_cq);
 
 	entries = bnxt_re_init_depth(cqe + 1, uctx);
@@ -3012,22 +3021,32 @@ int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 	spin_lock_init(&cq->cq_lock);
 
 	if (udata) {
-		struct bnxt_re_cq_resp resp;
-
+		struct bnxt_re_cq_resp resp = {};
+
+		if (cctx->modes.toggle_bits & BNXT_QPLIB_CQ_TOGGLE_BIT) {
+			hash_add(rdev->cq_hash, &cq->hash_entry, cq->qplib_cq.id);
+			/* Allocate a page */
+			cq->uctx_cq_page = (void *)get_zeroed_page(GFP_KERNEL);
+			if (!cq->uctx_cq_page)
+				goto c2fail;
+			resp.comp_mask |= BNXT_RE_CQ_TOGGLE_PAGE_SUPPORT;
+		}
 		resp.cqid = cq->qplib_cq.id;
 		resp.tail = cq->qplib_cq.hwq.cons;
 		resp.phase = cq->qplib_cq.period;
 		resp.rsvd = 0;
-		rc = ib_copy_to_udata(udata, &resp, sizeof(resp));
+		rc = ib_copy_to_udata(udata, &resp, min(sizeof(resp), udata->outlen));
 		if (rc) {
 			ibdev_err(&rdev->ibdev, "Failed to copy CQ udata");
 			bnxt_qplib_destroy_cq(&rdev->qplib_res, &cq->qplib_cq);
-			goto c2fail;
+			goto free_mem;
 		}
 	}
 
 	return 0;
 
+free_mem:
+	free_page((unsigned long)cq->uctx_cq_page);
 c2fail:
 	ib_umem_release(cq->umem);
 fail:
@@ -4214,6 +4233,19 @@ void bnxt_re_dealloc_ucontext(struct ib_ucontext *ib_uctx)
 	}
 }
 
+static struct bnxt_re_cq *bnxt_re_search_for_cq(struct bnxt_re_dev *rdev, u32 cq_id)
+{
+	struct bnxt_re_cq *cq = NULL, *tmp_cq;
+
+	hash_for_each_possible(rdev->cq_hash, tmp_cq, hash_entry, cq_id) {
+		if (tmp_cq->qplib_cq.id == cq_id) {
+			cq = tmp_cq;
+			break;
+		}
+	}
+	return cq;
+}
+
 /* Helper function to mmap the virtual memory from user app */
 int bnxt_re_mmap(struct ib_ucontext *ib_uctx, struct vm_area_struct *vma)
 {
@@ -4442,10 +4474,12 @@ static int UVERBS_HANDLER(BNXT_RE_METHOD_GET_TOGGLE_MEM)(struct uverbs_attr_bund
 	struct bnxt_re_ucontext *uctx;
 	struct ib_ucontext *ib_uctx;
 	struct bnxt_re_dev *rdev;
+	struct bnxt_re_cq *cq;
 	u64 mem_offset;
 	u64 addr = 0;
 	u32 length;
 	u32 offset;
+	u32 cq_id;
 	int err;
 
 	ib_uctx = ib_uverbs_get_ucontext(attrs);
@@ -4461,6 +4495,19 @@ static int UVERBS_HANDLER(BNXT_RE_METHOD_GET_TOGGLE_MEM)(struct uverbs_attr_bund
 
 	switch (res_type) {
 	case BNXT_RE_CQ_TOGGLE_MEM:
+		err = uverbs_copy_from(&cq_id, attrs, BNXT_RE_TOGGLE_MEM_RES_ID);
+		if (err)
+			return err;
+
+		cq = bnxt_re_search_for_cq(rdev, cq_id);
+		if (!cq)
+			return -EINVAL;
+
+		length = PAGE_SIZE;
+		addr = (u64)cq->uctx_cq_page;
+		mmap_flag = BNXT_RE_MMAP_TOGGLE_PAGE;
+		offset = 0;
+		break;
 	case BNXT_RE_SRQ_TOGGLE_MEM:
 		break;
 
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index da3fe018f255..b267d6d5975f 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -108,6 +108,8 @@ struct bnxt_re_cq {
 	struct ib_umem		*umem;
 	struct ib_umem		*resize_umem;
 	int			resize_cqe;
+	void			*uctx_cq_page;
+	struct hlist_node	hash_entry;
 };
 
 struct bnxt_re_mr {
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index 7f4f6db1392c..eb03ebad2e5a 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -54,6 +54,7 @@
 #include <rdma/ib_user_verbs.h>
 #include <rdma/ib_umem.h>
 #include <rdma/ib_addr.h>
+#include <linux/hashtable.h>
 
 #include "bnxt_ulp.h"
 #include "roce_hsi.h"
@@ -136,6 +137,8 @@ static void bnxt_re_set_drv_mode(struct bnxt_re_dev *rdev, u8 mode)
 	if (bnxt_re_hwrm_qcaps(rdev))
 		dev_err(rdev_to_dev(rdev),
 			"Failed to query hwrm qcaps\n");
+	if (bnxt_qplib_is_chip_gen_p7(rdev->chip_ctx))
+		cctx->modes.toggle_bits |= BNXT_QPLIB_CQ_TOGGLE_BIT;
 }
 
 static void bnxt_re_destroy_chip_ctx(struct bnxt_re_dev *rdev)
@@ -1206,9 +1209,13 @@ static int bnxt_re_cqn_handler(struct bnxt_qplib_nq *nq,
 {
 	struct bnxt_re_cq *cq = container_of(handle, struct bnxt_re_cq,
 					     qplib_cq);
+	u32 *cq_ptr;
 
 	if (cq->ib_cq.comp_handler) {
-		/* Lock comp_handler? */
+		if (cq->uctx_cq_page) {
+			cq_ptr = (u32 *)cq->uctx_cq_page;
+			*cq_ptr = cq->qplib_cq.toggle;
+		}
 		(*cq->ib_cq.comp_handler)(&cq->ib_cq, cq->ib_cq.cq_context);
 	}
 
@@ -1730,6 +1737,7 @@ static int bnxt_re_dev_init(struct bnxt_re_dev *rdev, u8 wqe_mode)
 		 */
 		bnxt_re_vf_res_config(rdev);
 	}
+	hash_init(rdev->cq_hash);
 
 	return 0;
 free_sctx:
diff --git a/drivers/infiniband/hw/bnxt_re/qplib_res.h b/drivers/infiniband/hw/bnxt_re/qplib_res.h
index 382d89fa7d16..61628f7f1253 100644
--- a/drivers/infiniband/hw/bnxt_re/qplib_res.h
+++ b/drivers/infiniband/hw/bnxt_re/qplib_res.h
@@ -55,6 +55,12 @@ struct bnxt_qplib_drv_modes {
 	u8	wqe_mode;
 	bool db_push;
 	bool dbr_pacing;
+	u32 toggle_bits;
+};
+
+enum bnxt_re_toggle_modes {
+	BNXT_QPLIB_CQ_TOGGLE_BIT = 0x1,
+	BNXT_QPLIB_SRQ_TOGGLE_BIT = 0x2,
 };
 
 struct bnxt_qplib_chip_ctx {
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 9b9eb10cb5e5..c0c34aca90ec 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -102,11 +102,16 @@ struct bnxt_re_cq_req {
 	__aligned_u64 cq_handle;
 };
 
+enum bnxt_re_cq_mask {
+	BNXT_RE_CQ_TOGGLE_PAGE_SUPPORT = 0x1,
+};
+
 struct bnxt_re_cq_resp {
 	__u32 cqid;
 	__u32 tail;
 	__u32 phase;
 	__u32 rsvd;
+	__aligned_u64 comp_mask;
 };
 
 struct bnxt_re_resize_cq_req {
-- 
cgit v1.2.3


From 2307157c85096026043ba11f9ad8393c31515c45 Mon Sep 17 00:00:00 2001
From: Michael Margolin <mrgolin@amazon.com>
Date: Thu, 4 Jan 2024 09:51:55 +0000
Subject: RDMA/efa: Add EFA query MR support

Add EFA driver uapi definitions and register a new query MR method that
currently returns the physical interconnects the device is using to
reach the MR. Update admin definitions and efa-abi accordingly.

Reviewed-by: Anas Mousa <anasmous@amazon.com>
Reviewed-by: Firas Jahjah <firasj@amazon.com>
Signed-off-by: Michael Margolin <mrgolin@amazon.com>
Link: https://lore.kernel.org/r/20240104095155.10676-1-mrgolin@amazon.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/efa/efa.h                 | 12 ++++-
 drivers/infiniband/hw/efa/efa_admin_cmds_defs.h | 33 +++++++++++-
 drivers/infiniband/hw/efa/efa_com_cmd.c         | 11 +++-
 drivers/infiniband/hw/efa/efa_com_cmd.h         | 12 ++++-
 drivers/infiniband/hw/efa/efa_main.c            |  7 ++-
 drivers/infiniband/hw/efa/efa_verbs.c           | 71 ++++++++++++++++++++++++-
 include/uapi/rdma/efa-abi.h                     | 21 +++++++-
 7 files changed, 160 insertions(+), 7 deletions(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/efa/efa.h b/drivers/infiniband/hw/efa/efa.h
index 7352a1f5d811..e2bdec32ae80 100644
--- a/drivers/infiniband/hw/efa/efa.h
+++ b/drivers/infiniband/hw/efa/efa.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause */
 /*
- * Copyright 2018-2021 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #ifndef _EFA_H_
@@ -80,9 +80,19 @@ struct efa_pd {
 	u16 pdn;
 };
 
+struct efa_mr_interconnect_info {
+	u16 recv_ic_id;
+	u16 rdma_read_ic_id;
+	u16 rdma_recv_ic_id;
+	u8 recv_ic_id_valid : 1;
+	u8 rdma_read_ic_id_valid : 1;
+	u8 rdma_recv_ic_id_valid : 1;
+};
+
 struct efa_mr {
 	struct ib_mr ibmr;
 	struct ib_umem *umem;
+	struct efa_mr_interconnect_info ic_info;
 };
 
 struct efa_cq {
diff --git a/drivers/infiniband/hw/efa/efa_admin_cmds_defs.h b/drivers/infiniband/hw/efa/efa_admin_cmds_defs.h
index 9c65bd27bae0..7377c8a9f4d5 100644
--- a/drivers/infiniband/hw/efa/efa_admin_cmds_defs.h
+++ b/drivers/infiniband/hw/efa/efa_admin_cmds_defs.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause */
 /*
- * Copyright 2018-2023 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #ifndef _EFA_ADMIN_CMDS_H_
@@ -415,6 +415,32 @@ struct efa_admin_reg_mr_resp {
 	 * memory region
 	 */
 	u32 r_key;
+
+	/*
+	 * Mask indicating which fields have valid values
+	 * 0 : recv_ic_id
+	 * 1 : rdma_read_ic_id
+	 * 2 : rdma_recv_ic_id
+	 */
+	u8 validity;
+
+	/*
+	 * Physical interconnect used by the device to reach the MR for receive
+	 * operation
+	 */
+	u8 recv_ic_id;
+
+	/*
+	 * Physical interconnect used by the device to reach the MR for RDMA
+	 * read operation
+	 */
+	u8 rdma_read_ic_id;
+
+	/*
+	 * Physical interconnect used by the device to reach the MR for RDMA
+	 * write receive
+	 */
+	u8 rdma_recv_ic_id;
 };
 
 struct efa_admin_dereg_mr_cmd {
@@ -999,6 +1025,11 @@ struct efa_admin_host_info {
 #define EFA_ADMIN_REG_MR_CMD_REMOTE_WRITE_ENABLE_MASK       BIT(1)
 #define EFA_ADMIN_REG_MR_CMD_REMOTE_READ_ENABLE_MASK        BIT(2)
 
+/* reg_mr_resp */
+#define EFA_ADMIN_REG_MR_RESP_RECV_IC_ID_MASK               BIT(0)
+#define EFA_ADMIN_REG_MR_RESP_RDMA_READ_IC_ID_MASK          BIT(1)
+#define EFA_ADMIN_REG_MR_RESP_RDMA_RECV_IC_ID_MASK          BIT(2)
+
 /* create_cq_cmd */
 #define EFA_ADMIN_CREATE_CQ_CMD_INTERRUPT_MODE_ENABLED_MASK BIT(5)
 #define EFA_ADMIN_CREATE_CQ_CMD_VIRT_MASK                   BIT(6)
diff --git a/drivers/infiniband/hw/efa/efa_com_cmd.c b/drivers/infiniband/hw/efa/efa_com_cmd.c
index 576811885d59..d3398c7b0bd0 100644
--- a/drivers/infiniband/hw/efa/efa_com_cmd.c
+++ b/drivers/infiniband/hw/efa/efa_com_cmd.c
@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause
 /*
- * Copyright 2018-2023 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #include "efa_com.h"
@@ -270,6 +270,15 @@ int efa_com_register_mr(struct efa_com_dev *edev,
 
 	result->l_key = cmd_completion.l_key;
 	result->r_key = cmd_completion.r_key;
+	result->ic_info.recv_ic_id = cmd_completion.recv_ic_id;
+	result->ic_info.rdma_read_ic_id = cmd_completion.rdma_read_ic_id;
+	result->ic_info.rdma_recv_ic_id = cmd_completion.rdma_recv_ic_id;
+	result->ic_info.recv_ic_id_valid = EFA_GET(&cmd_completion.validity,
+						   EFA_ADMIN_REG_MR_RESP_RECV_IC_ID);
+	result->ic_info.rdma_read_ic_id_valid = EFA_GET(&cmd_completion.validity,
+							EFA_ADMIN_REG_MR_RESP_RDMA_READ_IC_ID);
+	result->ic_info.rdma_recv_ic_id_valid = EFA_GET(&cmd_completion.validity,
+							EFA_ADMIN_REG_MR_RESP_RDMA_RECV_IC_ID);
 
 	return 0;
 }
diff --git a/drivers/infiniband/hw/efa/efa_com_cmd.h b/drivers/infiniband/hw/efa/efa_com_cmd.h
index fc97f37bb39b..720a99ba0f7d 100644
--- a/drivers/infiniband/hw/efa/efa_com_cmd.h
+++ b/drivers/infiniband/hw/efa/efa_com_cmd.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause */
 /*
- * Copyright 2018-2023 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #ifndef _EFA_COM_CMD_H_
@@ -199,6 +199,15 @@ struct efa_com_reg_mr_params {
 	u8 indirect;
 };
 
+struct efa_com_mr_interconnect_info {
+	u16 recv_ic_id;
+	u16 rdma_read_ic_id;
+	u16 rdma_recv_ic_id;
+	u8 recv_ic_id_valid : 1;
+	u8 rdma_read_ic_id_valid : 1;
+	u8 rdma_recv_ic_id_valid : 1;
+};
+
 struct efa_com_reg_mr_result {
 	/*
 	 * To be used in conjunction with local buffers references in SQ and
@@ -210,6 +219,7 @@ struct efa_com_reg_mr_result {
 	 * accessed memory region
 	 */
 	u32 r_key;
+	struct efa_com_mr_interconnect_info ic_info;
 };
 
 struct efa_com_dereg_mr_params {
diff --git a/drivers/infiniband/hw/efa/efa_main.c b/drivers/infiniband/hw/efa/efa_main.c
index 15ee92081118..7b1910a86216 100644
--- a/drivers/infiniband/hw/efa/efa_main.c
+++ b/drivers/infiniband/hw/efa/efa_main.c
@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause
 /*
- * Copyright 2018-2022 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #include <linux/module.h>
@@ -9,6 +9,7 @@
 #include <linux/version.h>
 
 #include <rdma/ib_user_verbs.h>
+#include <rdma/uverbs_ioctl.h>
 
 #include "efa.h"
 
@@ -36,6 +37,8 @@ MODULE_DEVICE_TABLE(pci, efa_pci_tbl);
 	(BIT(EFA_ADMIN_FATAL_ERROR) | BIT(EFA_ADMIN_WARNING) | \
 	 BIT(EFA_ADMIN_NOTIFICATION) | BIT(EFA_ADMIN_KEEP_ALIVE))
 
+extern const struct uapi_definition efa_uapi_defs[];
+
 /* This handler will called for unknown event group or unimplemented handlers */
 static void unimplemented_aenq_handler(void *data,
 				       struct efa_admin_aenq_entry *aenq_e)
@@ -432,6 +435,8 @@ static int efa_ib_device_add(struct efa_dev *dev)
 
 	ib_set_device_ops(&dev->ibdev, &efa_dev_ops);
 
+	dev->ibdev.driver_def = efa_uapi_defs;
+
 	err = ib_register_device(&dev->ibdev, "efa_%d", &pdev->dev);
 	if (err)
 		goto err_destroy_eqs;
diff --git a/drivers/infiniband/hw/efa/efa_verbs.c b/drivers/infiniband/hw/efa/efa_verbs.c
index 0f8ca99d0827..2f412db2edcd 100644
--- a/drivers/infiniband/hw/efa/efa_verbs.c
+++ b/drivers/infiniband/hw/efa/efa_verbs.c
@@ -1,6 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
 /*
- * Copyright 2018-2023 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #include <linux/dma-buf.h>
@@ -13,6 +13,9 @@
 #include <rdma/ib_user_verbs.h>
 #include <rdma/ib_verbs.h>
 #include <rdma/uverbs_ioctl.h>
+#define UVERBS_MODULE_NAME efa_ib
+#include <rdma/uverbs_named_ioctl.h>
+#include <rdma/ib_user_ioctl_cmds.h>
 
 #include "efa.h"
 #include "efa_io_defs.h"
@@ -1653,6 +1656,12 @@ static int efa_register_mr(struct ib_pd *ibpd, struct efa_mr *mr, u64 start,
 	mr->ibmr.lkey = result.l_key;
 	mr->ibmr.rkey = result.r_key;
 	mr->ibmr.length = length;
+	mr->ic_info.recv_ic_id = result.ic_info.recv_ic_id;
+	mr->ic_info.rdma_read_ic_id = result.ic_info.rdma_read_ic_id;
+	mr->ic_info.rdma_recv_ic_id = result.ic_info.rdma_recv_ic_id;
+	mr->ic_info.recv_ic_id_valid = result.ic_info.recv_ic_id_valid;
+	mr->ic_info.rdma_read_ic_id_valid = result.ic_info.rdma_read_ic_id_valid;
+	mr->ic_info.rdma_recv_ic_id_valid = result.ic_info.rdma_recv_ic_id_valid;
 	ibdev_dbg(&dev->ibdev, "Registered mr[%d]\n", mr->ibmr.lkey);
 
 	return 0;
@@ -1735,6 +1744,39 @@ err_out:
 	return ERR_PTR(err);
 }
 
+static int UVERBS_HANDLER(EFA_IB_METHOD_MR_QUERY)(struct uverbs_attr_bundle *attrs)
+{
+	struct ib_mr *ibmr = uverbs_attr_get_obj(attrs, EFA_IB_ATTR_QUERY_MR_HANDLE);
+	struct efa_mr *mr = to_emr(ibmr);
+	u16 ic_id_validity = 0;
+	int ret;
+
+	ret = uverbs_copy_to(attrs, EFA_IB_ATTR_QUERY_MR_RESP_RECV_IC_ID,
+			     &mr->ic_info.recv_ic_id, sizeof(mr->ic_info.recv_ic_id));
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_to(attrs, EFA_IB_ATTR_QUERY_MR_RESP_RDMA_READ_IC_ID,
+			     &mr->ic_info.rdma_read_ic_id, sizeof(mr->ic_info.rdma_read_ic_id));
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_to(attrs, EFA_IB_ATTR_QUERY_MR_RESP_RDMA_RECV_IC_ID,
+			     &mr->ic_info.rdma_recv_ic_id, sizeof(mr->ic_info.rdma_recv_ic_id));
+	if (ret)
+		return ret;
+
+	if (mr->ic_info.recv_ic_id_valid)
+		ic_id_validity |= EFA_QUERY_MR_VALIDITY_RECV_IC_ID;
+	if (mr->ic_info.rdma_read_ic_id_valid)
+		ic_id_validity |= EFA_QUERY_MR_VALIDITY_RDMA_READ_IC_ID;
+	if (mr->ic_info.rdma_recv_ic_id_valid)
+		ic_id_validity |= EFA_QUERY_MR_VALIDITY_RDMA_RECV_IC_ID;
+
+	return uverbs_copy_to(attrs, EFA_IB_ATTR_QUERY_MR_RESP_IC_ID_VALIDITY,
+			      &ic_id_validity, sizeof(ic_id_validity));
+}
+
 int efa_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata)
 {
 	struct efa_dev *dev = to_edev(ibmr->device);
@@ -2157,3 +2199,30 @@ enum rdma_link_layer efa_port_link_layer(struct ib_device *ibdev,
 	return IB_LINK_LAYER_UNSPECIFIED;
 }
 
+DECLARE_UVERBS_NAMED_METHOD(EFA_IB_METHOD_MR_QUERY,
+			    UVERBS_ATTR_IDR(EFA_IB_ATTR_QUERY_MR_HANDLE,
+					    UVERBS_OBJECT_MR,
+					    UVERBS_ACCESS_READ,
+					    UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(EFA_IB_ATTR_QUERY_MR_RESP_IC_ID_VALIDITY,
+						UVERBS_ATTR_TYPE(u16),
+						UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(EFA_IB_ATTR_QUERY_MR_RESP_RECV_IC_ID,
+						UVERBS_ATTR_TYPE(u16),
+						UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(EFA_IB_ATTR_QUERY_MR_RESP_RDMA_READ_IC_ID,
+						UVERBS_ATTR_TYPE(u16),
+						UA_MANDATORY),
+			    UVERBS_ATTR_PTR_OUT(EFA_IB_ATTR_QUERY_MR_RESP_RDMA_RECV_IC_ID,
+						UVERBS_ATTR_TYPE(u16),
+						UA_MANDATORY));
+
+ADD_UVERBS_METHODS(efa_mr,
+		   UVERBS_OBJECT_MR,
+		   &UVERBS_METHOD(EFA_IB_METHOD_MR_QUERY));
+
+const struct uapi_definition efa_uapi_defs[] = {
+	UAPI_DEF_CHAIN_OBJ_TREE(UVERBS_OBJECT_MR,
+				&efa_mr),
+	{},
+};
diff --git a/include/uapi/rdma/efa-abi.h b/include/uapi/rdma/efa-abi.h
index d94c32f28804..701e2d567e41 100644
--- a/include/uapi/rdma/efa-abi.h
+++ b/include/uapi/rdma/efa-abi.h
@@ -1,12 +1,13 @@
 /* SPDX-License-Identifier: ((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) */
 /*
- * Copyright 2018-2023 Amazon.com, Inc. or its affiliates. All rights reserved.
+ * Copyright 2018-2024 Amazon.com, Inc. or its affiliates. All rights reserved.
  */
 
 #ifndef EFA_ABI_USER_H
 #define EFA_ABI_USER_H
 
 #include <linux/types.h>
+#include <rdma/ib_user_ioctl_cmds.h>
 
 /*
  * Increment this value if any changes that break userspace ABI
@@ -134,4 +135,22 @@ struct efa_ibv_ex_query_device_resp {
 	__u32 device_caps;
 };
 
+enum {
+	EFA_QUERY_MR_VALIDITY_RECV_IC_ID = 1 << 0,
+	EFA_QUERY_MR_VALIDITY_RDMA_READ_IC_ID = 1 << 1,
+	EFA_QUERY_MR_VALIDITY_RDMA_RECV_IC_ID = 1 << 2,
+};
+
+enum efa_query_mr_attrs {
+	EFA_IB_ATTR_QUERY_MR_HANDLE = (1U << UVERBS_ID_NS_SHIFT),
+	EFA_IB_ATTR_QUERY_MR_RESP_IC_ID_VALIDITY,
+	EFA_IB_ATTR_QUERY_MR_RESP_RECV_IC_ID,
+	EFA_IB_ATTR_QUERY_MR_RESP_RDMA_READ_IC_ID,
+	EFA_IB_ATTR_QUERY_MR_RESP_RDMA_RECV_IC_ID,
+};
+
+enum efa_mr_methods {
+	EFA_IB_METHOD_MR_QUERY = (1U << UVERBS_ID_NS_SHIFT),
+};
+
 #endif /* EFA_ABI_USER_H */
-- 
cgit v1.2.3