Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma

Pull rdma updates from Jason Gunthorpe:
"Small cycle, with some typical driver updates:

- General code tidying in siw, hfi1, idrdma, usnic, hns rtrs and
bnxt_re

- Many small siw cleanups without an overeaching theme

- Debugfs stats for hns

- Fix a TX queue timeout in IPoIB and missed locking of the mcast
list

- Support more features of P7 devices in bnxt_re including a new work
submission protocol

- CQ interrupts for MANA

- netlink stats for erdma

- EFA multipath PCI support

- Fix Incorrect MR invalidation in iser"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (66 commits)
RDMA/bnxt_re: Fix error code in bnxt_re_create_cq()
RDMA/efa: Add EFA query MR support
IB/iser: Prevent invalidating wrong MR
RDMA/erdma: Add hardware statistics support
RDMA/erdma: Introduce dma pool for hardware responses of CMDQ requests
IB/iser: iscsi_iser.h: fix kernel-doc warning and spellos
RDMA/mana_ib: Add CQ interrupt support for RAW QP
RDMA/mana_ib: query device capabilities
RDMA/mana_ib: register RDMA device with GDMA
RDMA/bnxt_re: Fix the sparse warnings
RDMA/bnxt_re: Fix the offset for GenP7 adapters for user applications
RDMA/bnxt_re: Share a page to expose per CQ info with userspace
RDMA/bnxt_re: Add UAPI to share a page with user space
IB/ipoib: Fix mcast list locking
RDMA/mlx5: Expose register c0 for RDMA device
net/mlx5: E-Switch, expose eswitch manager vport
net/mlx5: Manage ICM type of SW encap
RDMA/mlx5: Support handling of SW encap ICM area
net/mlx5: Introduce indirect-sw-encap ICM properties
RDMA/bnxt_re: Adds MSN table capability for Gen P7 adapters
...

drivers/infiniband/hw/bnxt_re/bnxt_re.h

···
 #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"
···
 #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;
···
 	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)	\

drivers/infiniband/hw/bnxt_re/hw_counters.c

···
 	}
 
 done:
-	return bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx) ?
+	return bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx) ?
 		BNXT_RE_NUM_EXT_COUNTERS : BNXT_RE_NUM_STD_COUNTERS;
 }
 
···
 	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
 	int num_counters = 0;
 
-	if (bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx))
+	if (bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx))
 		num_counters = BNXT_RE_NUM_EXT_COUNTERS;
 	else
 		num_counters = BNXT_RE_NUM_STD_COUNTERS;

drivers/infiniband/hw/bnxt_re/ib_verbs.c

···
 #include <rdma/ib_mad.h>
 #include <rdma/ib_cache.h>
 #include <rdma/uverbs_ioctl.h>
+#include <linux/hashtable.h>
 
 #include "bnxt_ulp.h"
 
···
 	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;
···
 	bytes = (qplib_qp->sq.max_wqe * qplib_qp->sq.wqe_size);
 	/* Consider mapping PSN search memory only for RC QPs. */
 	if (qplib_qp->type == CMDQ_CREATE_QP_TYPE_RC) {
-		psn_sz = bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx) ?
+		psn_sz = bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx) ?
 						   sizeof(struct sq_psn_search_ext) :
 						   sizeof(struct sq_psn_search);
 		psn_nume = (qplib_qp->wqe_mode == BNXT_QPLIB_WQE_MODE_STATIC) ?
···
 }
 
 static int bnxt_re_init_rq_attr(struct bnxt_re_qp *qp,
-				struct ib_qp_init_attr *init_attr)
+				struct ib_qp_init_attr *init_attr,
+				struct bnxt_re_ucontext *uctx)
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
 	struct bnxt_qplib_qp *qplqp;
···
 		/* Allocate 1 more than what's provided so posting max doesn't
 		 * mean empty.
 		 */
-		entries = roundup_pow_of_two(init_attr->cap.max_recv_wr + 1);
+		entries = bnxt_re_init_depth(init_attr->cap.max_recv_wr + 1, uctx);
 		rq->max_wqe = min_t(u32, entries, dev_attr->max_qp_wqes + 1);
 		rq->q_full_delta = 0;
 		rq->sg_info.pgsize = PAGE_SIZE;
···
 	qplqp = &qp->qplib_qp;
 	dev_attr = &rdev->dev_attr;
 
-	if (!bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx)) {
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx)) {
 		qplqp->rq.max_sge = dev_attr->max_qp_sges;
 		if (qplqp->rq.max_sge > dev_attr->max_qp_sges)
 			qplqp->rq.max_sge = dev_attr->max_qp_sges;
···
 
 static int bnxt_re_init_sq_attr(struct bnxt_re_qp *qp,
 				struct ib_qp_init_attr *init_attr,
-				struct ib_udata *udata)
+				struct bnxt_re_ucontext *uctx)
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
 	struct bnxt_qplib_qp *qplqp;
···
 	/* Allocate 128 + 1 more than what's provided */
 	diff = (qplqp->wqe_mode == BNXT_QPLIB_WQE_MODE_VARIABLE) ?
 		0 : BNXT_QPLIB_RESERVED_QP_WRS;
-	entries = roundup_pow_of_two(entries + diff + 1);
+	entries = bnxt_re_init_depth(entries + diff + 1, uctx);
 	sq->max_wqe = min_t(u32, entries, dev_attr->max_qp_wqes + diff + 1);
 	sq->q_full_delta = diff + 1;
 	/*
···
 }
 
 static void bnxt_re_adjust_gsi_sq_attr(struct bnxt_re_qp *qp,
-				       struct ib_qp_init_attr *init_attr)
+				       struct ib_qp_init_attr *init_attr,
+				       struct bnxt_re_ucontext *uctx)
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
 	struct bnxt_qplib_qp *qplqp;
···
 	qplqp = &qp->qplib_qp;
 	dev_attr = &rdev->dev_attr;
 
-	if (!bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx)) {
-		entries = roundup_pow_of_two(init_attr->cap.max_send_wr + 1);
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx)) {
+		entries = bnxt_re_init_depth(init_attr->cap.max_send_wr + 1, uctx);
 		qplqp->sq.max_wqe = min_t(u32, entries,
 					  dev_attr->max_qp_wqes + 1);
 		qplqp->sq.q_full_delta = qplqp->sq.max_wqe -
···
 		goto out;
 	}
 
-	if (bnxt_qplib_is_chip_gen_p5(chip_ctx) &&
+	if (bnxt_qplib_is_chip_gen_p5_p7(chip_ctx) &&
 	    init_attr->qp_type == IB_QPT_GSI)
 		qptype = CMDQ_CREATE_QP_TYPE_GSI;
 out:
···
 				struct ib_udata *udata)
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
+	struct bnxt_re_ucontext *uctx;
 	struct bnxt_qplib_qp *qplqp;
 	struct bnxt_re_dev *rdev;
 	struct bnxt_re_cq *cq;
···
 	qplqp = &qp->qplib_qp;
 	dev_attr = &rdev->dev_attr;
 
+	uctx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
 	/* Setup misc params */
 	ether_addr_copy(qplqp->smac, rdev->netdev->dev_addr);
 	qplqp->pd = &pd->qplib_pd;
···
 	}
 
 	/* Setup RQ/SRQ */
-	rc = bnxt_re_init_rq_attr(qp, init_attr);
+	rc = bnxt_re_init_rq_attr(qp, init_attr, uctx);
 	if (rc)
 		goto out;
 	if (init_attr->qp_type == IB_QPT_GSI)
 		bnxt_re_adjust_gsi_rq_attr(qp);
 
 	/* Setup SQ */
-	rc = bnxt_re_init_sq_attr(qp, init_attr, udata);
+	rc = bnxt_re_init_sq_attr(qp, init_attr, uctx);
 	if (rc)
 		goto out;
 	if (init_attr->qp_type == IB_QPT_GSI)
-		bnxt_re_adjust_gsi_sq_attr(qp, init_attr);
+		bnxt_re_adjust_gsi_sq_attr(qp, init_attr, uctx);
 
 	if (udata) /* This will update DPI and qp_handle */
 		rc = bnxt_re_init_user_qp(rdev, pd, qp, udata);
···
 		goto fail;
 
 	if (qp_init_attr->qp_type == IB_QPT_GSI &&
-	    !(bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx))) {
+	    !(bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx))) {
 		rc = bnxt_re_create_gsi_qp(qp, pd, qp_init_attr);
 		if (rc == -ENODEV)
 			goto qp_destroy;
···
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
 	struct bnxt_qplib_nq *nq = NULL;
+	struct bnxt_re_ucontext *uctx;
 	struct bnxt_re_dev *rdev;
 	struct bnxt_re_srq *srq;
 	struct bnxt_re_pd *pd;
···
 		goto exit;
 	}
 
+	uctx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
 	srq->rdev = rdev;
 	srq->qplib_srq.pd = &pd->qplib_pd;
 	srq->qplib_srq.dpi = &rdev->dpi_privileged;
 	/* Allocate 1 more than what's provided so posting max doesn't
 	 * mean empty
 	 */
-	entries = roundup_pow_of_two(srq_init_attr->attr.max_wr + 1);
+	entries = bnxt_re_init_depth(srq_init_attr->attr.max_wr + 1, uctx);
 	if (entries > dev_attr->max_srq_wqes + 1)
 		entries = dev_attr->max_srq_wqes + 1;
 	srq->qplib_srq.max_wqe = entries;
···
 		qp->qplib_qp.max_dest_rd_atomic = qp_attr->max_dest_rd_atomic;
 	}
 	if (qp_attr_mask & IB_QP_CAP) {
+		struct bnxt_re_ucontext *uctx =
+			rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
+
 		qp->qplib_qp.modify_flags |=
 				CMDQ_MODIFY_QP_MODIFY_MASK_SQ_SIZE |
 				CMDQ_MODIFY_QP_MODIFY_MASK_RQ_SIZE |
···
 				  "Create QP failed - max exceeded");
 			return -EINVAL;
 		}
-		entries = roundup_pow_of_two(qp_attr->cap.max_send_wr);
+		entries = bnxt_re_init_depth(qp_attr->cap.max_send_wr, uctx);
 		qp->qplib_qp.sq.max_wqe = min_t(u32, entries,
 						dev_attr->max_qp_wqes + 1);
 		qp->qplib_qp.sq.q_full_delta = qp->qplib_qp.sq.max_wqe -
···
 		qp->qplib_qp.sq.q_full_delta -= 1;
 		qp->qplib_qp.sq.max_sge = qp_attr->cap.max_send_sge;
 		if (qp->qplib_qp.rq.max_wqe) {
-			entries = roundup_pow_of_two(qp_attr->cap.max_recv_wr);
+			entries = bnxt_re_init_depth(qp_attr->cap.max_recv_wr, uctx);
 			qp->qplib_qp.rq.max_wqe =
 				min_t(u32, entries, dev_attr->max_qp_wqes + 1);
 			qp->qplib_qp.rq.q_full_delta = qp->qplib_qp.rq.max_wqe -
···
 /* 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);
 
···
 int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		      struct ib_udata *udata)
 {
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibcq->device, ibdev);
-	struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
 	struct bnxt_re_cq *cq = container_of(ibcq, struct bnxt_re_cq, ib_cq);
-	int rc, entries;
-	int cqe = attr->cqe;
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibcq->device, ibdev);
+	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;
+	struct bnxt_qplib_chip_ctx *cctx;
 	struct bnxt_qplib_nq *nq = NULL;
 	unsigned int nq_alloc_cnt;
+	int cqe = attr->cqe;
+	int rc, entries;
 	u32 active_cqs;
 
 	if (attr->flags)
···
 	}
 
 	cq->rdev = rdev;
+	cctx = rdev->chip_ctx;
 	cq->qplib_cq.cq_handle = (u64)(unsigned long)(&cq->qplib_cq);
 
-	entries = roundup_pow_of_two(cqe + 1);
+	entries = bnxt_re_init_depth(cqe + 1, uctx);
 	if (entries > dev_attr->max_cq_wqes + 1)
 		entries = dev_attr->max_cq_wqes + 1;
 
···
 	cq->qplib_cq.sg_info.pgshft = PAGE_SHIFT;
 	if (udata) {
 		struct bnxt_re_cq_req req;
-		struct bnxt_re_ucontext *uctx = rdma_udata_to_drv_context(
-			udata, struct bnxt_re_ucontext, ib_uctx);
 		if (ib_copy_from_udata(&req, udata, sizeof(req))) {
 			rc = -EFAULT;
 			goto fail;
···
 	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) {
+				rc = -ENOMEM;
+				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:
···
 		return -EINVAL;
 	}
 
-	entries = roundup_pow_of_two(cqe + 1);
+	uctx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
+	entries = bnxt_re_init_depth(cqe + 1, uctx);
 	if (entries > dev_attr->max_cq_wqes + 1)
 		entries = dev_attr->max_cq_wqes + 1;
 
-	uctx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext,
-					 ib_uctx);
 	/* uverbs consumer */
 	if (ib_copy_from_udata(&req, udata, sizeof(req))) {
 		rc = -EFAULT;
···
 	struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
 	struct bnxt_re_user_mmap_entry *entry;
 	struct bnxt_re_uctx_resp resp = {};
+	struct bnxt_re_uctx_req ureq = {};
 	u32 chip_met_rev_num = 0;
 	int rc;
 
···
 	if (rdev->pacing.dbr_pacing)
 		resp.comp_mask |= BNXT_RE_UCNTX_CMASK_DBR_PACING_ENABLED;
 
+	if (udata->inlen >= sizeof(ureq)) {
+		rc = ib_copy_from_udata(&ureq, udata, min(udata->inlen, sizeof(ureq)));
+		if (rc)
+			goto cfail;
+		if (ureq.comp_mask & BNXT_RE_COMP_MASK_REQ_UCNTX_POW2_SUPPORT) {
+			resp.comp_mask |= BNXT_RE_UCNTX_CMASK_POW2_DISABLED;
+			uctx->cmask |= BNXT_RE_UCNTX_CMASK_POW2_DISABLED;
+		}
+	}
+
 	rc = ib_copy_to_udata(udata, &resp, min(udata->outlen, sizeof(resp)));
 	if (rc) {
 		ibdev_err(ibdev, "Failed to copy user context");
···
 		bnxt_qplib_dealloc_dpi(&rdev->qplib_res, &uctx->dpi);
 		uctx->dpi.dbr = NULL;
 	}
+}
+
+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 */
···
 					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;
···
 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;
+	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);
+	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:
+		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;
+
+	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),
 	{}
 };

drivers/infiniband/hw/bnxt_re/ib_verbs.h

···
 	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 {
···
 	void			*shpg;
 	spinlock_t		sh_lock;	/* protect shpg */
 	struct rdma_user_mmap_entry *shpage_mmap;
+	u64 cmask;
 };
 
 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 {
···
 static inline u16 bnxt_re_get_rwqe_size(int nsge)
 {
 	return sizeof(struct rq_wqe_hdr) + (nsge * sizeof(struct sq_sge));
+}
+
+static inline u32 bnxt_re_init_depth(u32 ent, struct bnxt_re_ucontext *uctx)
+{
+	return uctx ? (uctx->cmask & BNXT_RE_UCNTX_CMASK_POW2_DISABLED) ?
+		ent : roundup_pow_of_two(ent) : ent;
 }
 
 int bnxt_re_query_device(struct ib_device *ibdev,

drivers/infiniband/hw/bnxt_re/main.c

···
 #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"
···
 		dev_info(rdev_to_dev(rdev),
 			 "Couldn't get DB bar size, Low latency framework is disabled\n");
 	/* set register offsets for both UC and WC */
-	res->dpi_tbl.ucreg.offset = res->is_vf ? BNXT_QPLIB_DBR_VF_DB_OFFSET :
-						 BNXT_QPLIB_DBR_PF_DB_OFFSET;
-	res->dpi_tbl.wcreg.offset = res->dpi_tbl.ucreg.offset;
+	if (bnxt_qplib_is_chip_gen_p7(cctx)) {
+		res->dpi_tbl.ucreg.offset = offset;
+		res->dpi_tbl.wcreg.offset = en_dev->l2_db_size;
+	} else {
+		res->dpi_tbl.ucreg.offset = res->is_vf ? BNXT_QPLIB_DBR_VF_DB_OFFSET :
+							 BNXT_QPLIB_DBR_PF_DB_OFFSET;
+		res->dpi_tbl.wcreg.offset = res->dpi_tbl.ucreg.offset;
+	}
 
 	/* If WC mapping is disabled by L2 driver then en_dev->l2_db_size
 	 * is equal to the DB-Bar actual size. This indicates that L2
···
 	struct bnxt_qplib_chip_ctx *cctx;
 
 	cctx = rdev->chip_ctx;
-	cctx->modes.wqe_mode = bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx) ?
+	cctx->modes.wqe_mode = bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx) ?
 			       mode : BNXT_QPLIB_WQE_MODE_STATIC;
 	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)
···
 	ctx->srqc_count = min_t(u32, BNXT_RE_MAX_SRQC_COUNT,
 				attr->max_srq);
 	ctx->cq_count = min_t(u32, BNXT_RE_MAX_CQ_COUNT, attr->max_cq);
-	if (!bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx))
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx))
 		for (i = 0; i < MAX_TQM_ALLOC_REQ; i++)
 			rdev->qplib_ctx.tqm_ctx.qcount[i] =
 			rdev->dev_attr.tqm_alloc_reqs[i];
···
 	memset(&rdev->qplib_ctx.vf_res, 0, sizeof(struct bnxt_qplib_vf_res));
 	bnxt_re_limit_pf_res(rdev);
 
-	num_vfs =  bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx) ?
+	num_vfs =  bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx) ?
 			BNXT_RE_GEN_P5_MAX_VF : rdev->num_vfs;
 	if (num_vfs)
 		bnxt_re_limit_vf_res(&rdev->qplib_ctx, num_vfs);
···
 	if (test_bit(BNXT_RE_FLAG_ERR_DEVICE_DETACHED, &rdev->flags))
 		return;
 	rdev->num_vfs = pci_sriov_get_totalvfs(rdev->en_dev->pdev);
-	if (!bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx)) {
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx)) {
 		bnxt_re_set_resource_limits(rdev);
 		bnxt_qplib_set_func_resources(&rdev->qplib_res, &rdev->rcfw,
 					      &rdev->qplib_ctx);
···
 {
 	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);
 	}
 
 	return 0;
-}
-
-#define BNXT_RE_GEN_P5_PF_NQ_DB		0x10000
-#define BNXT_RE_GEN_P5_VF_NQ_DB		0x4000
-static u32 bnxt_re_get_nqdb_offset(struct bnxt_re_dev *rdev, u16 indx)
-{
-	return bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx) ?
-		(rdev->is_virtfn ? BNXT_RE_GEN_P5_VF_NQ_DB :
-				   BNXT_RE_GEN_P5_PF_NQ_DB) :
-				   rdev->en_dev->msix_entries[indx].db_offset;
 }
 
 static void bnxt_re_cleanup_res(struct bnxt_re_dev *rdev)
···
 	bnxt_qplib_init_res(&rdev->qplib_res);
 
 	for (i = 1; i < rdev->num_msix ; i++) {
-		db_offt = bnxt_re_get_nqdb_offset(rdev, i);
+		db_offt = rdev->en_dev->msix_entries[i].db_offset;
 		rc = bnxt_qplib_enable_nq(rdev->en_dev->pdev, &rdev->nq[i - 1],
 					  i - 1, rdev->en_dev->msix_entries[i].vector,
 					  db_offt, &bnxt_re_cqn_handler,
···
 		ibdev_err(&rdev->ibdev, "Failed to allocate CREQ: %#x\n", rc);
 		goto free_rcfw;
 	}
-	db_offt = bnxt_re_get_nqdb_offset(rdev, BNXT_RE_AEQ_IDX);
+	db_offt = rdev->en_dev->msix_entries[BNXT_RE_AEQ_IDX].db_offset;
 	vid = rdev->en_dev->msix_entries[BNXT_RE_AEQ_IDX].vector;
 	rc = bnxt_qplib_enable_rcfw_channel(&rdev->rcfw,
 					    vid, db_offt,
···
 	bnxt_re_set_resource_limits(rdev);
 
 	rc = bnxt_qplib_alloc_ctx(&rdev->qplib_res, &rdev->qplib_ctx, 0,
-				  bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx));
+				  bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx));
 	if (rc) {
 		ibdev_err(&rdev->ibdev,
 			  "Failed to allocate QPLIB context: %#x\n", rc);
···
 		 */
 		bnxt_re_vf_res_config(rdev);
 	}
+	hash_init(rdev->cq_hash);
 
 	return 0;
 free_sctx:
···
 		return;
 
 	/* Currently enabling only for GenP5 adapters */
-	if (!bnxt_qplib_is_chip_gen_p5(rdev->chip_ctx))
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rdev->chip_ctx))
 		return;
 
 	if (enable) {

drivers/infiniband/hw/bnxt_re/qplib_fp.c

···
 	struct bnxt_qplib_hwq *hwq = &nq->hwq;
 	struct nq_base *nqe, **nq_ptr;
 	int budget = nq->budget;
-	u32 sw_cons, raw_cons;
 	uintptr_t q_handle;
 	u16 type;
 
 	spin_lock_bh(&hwq->lock);
 	/* Service the NQ until empty */
-	raw_cons = hwq->cons;
 	while (budget--) {
-		sw_cons = HWQ_CMP(raw_cons, hwq);
 		nq_ptr = (struct nq_base **)hwq->pbl_ptr;
-		nqe = &nq_ptr[NQE_PG(sw_cons)][NQE_IDX(sw_cons)];
-		if (!NQE_CMP_VALID(nqe, raw_cons, hwq->max_elements))
+		nqe = &nq_ptr[NQE_PG(hwq->cons)][NQE_IDX(hwq->cons)];
+		if (!NQE_CMP_VALID(nqe, nq->nq_db.dbinfo.flags))
 			break;
 
 		/*
···
 		default:
 			break;
 		}
-		raw_cons++;
+		bnxt_qplib_hwq_incr_cons(hwq->max_elements, &hwq->cons,
+					 1, &nq->nq_db.dbinfo.flags);
 	}
 	spin_unlock_bh(&hwq->lock);
 }
···
 	struct bnxt_qplib_hwq *hwq = &nq->hwq;
 	struct bnxt_qplib_cq *cq;
 	int budget = nq->budget;
-	u32 sw_cons, raw_cons;
 	struct nq_base *nqe;
 	uintptr_t q_handle;
+	u32 hw_polled = 0;
 	u16 type;
 
 	spin_lock_bh(&hwq->lock);
 	/* Service the NQ until empty */
-	raw_cons = hwq->cons;
 	while (budget--) {
-		sw_cons = HWQ_CMP(raw_cons, hwq);
-		nqe = bnxt_qplib_get_qe(hwq, sw_cons, NULL);
-		if (!NQE_CMP_VALID(nqe, raw_cons, hwq->max_elements))
+		nqe = bnxt_qplib_get_qe(hwq, hwq->cons, NULL);
+		if (!NQE_CMP_VALID(nqe, nq->nq_db.dbinfo.flags))
 			break;
 
 		/*
···
 			cq = (struct bnxt_qplib_cq *)(unsigned long)q_handle;
 			if (!cq)
 				break;
+			cq->toggle = (le16_to_cpu(nqe->info10_type) &
+					NQ_CN_TOGGLE_MASK) >> NQ_CN_TOGGLE_SFT;
+			cq->dbinfo.toggle = cq->toggle;
 			bnxt_qplib_armen_db(&cq->dbinfo,
 					    DBC_DBC_TYPE_CQ_ARMENA);
 			spin_lock_bh(&cq->compl_lock);
···
 				 "nqe with type = 0x%x not handled\n", type);
 			break;
 		}
-		raw_cons++;
+		hw_polled++;
+		bnxt_qplib_hwq_incr_cons(hwq->max_elements, &hwq->cons,
+					 1, &nq->nq_db.dbinfo.flags);
 	}
-	if (hwq->cons != raw_cons) {
-		hwq->cons = raw_cons;
+	if (hw_polled)
 		bnxt_qplib_ring_nq_db(&nq->nq_db.dbinfo, nq->res->cctx, true);
-	}
 	spin_unlock_bh(&hwq->lock);
 }
 
···
 	pdev = nq->pdev;
 	nq_db = &nq->nq_db;
 
+	nq_db->dbinfo.flags = 0;
 	nq_db->reg.bar_id = NQ_CONS_PCI_BAR_REGION;
 	nq_db->reg.bar_base = pci_resource_start(pdev, nq_db->reg.bar_id);
 	if (!nq_db->reg.bar_base) {
···
 		rc = -ENOMEM;
 		goto fail;
 	}
-
+	srq->dbinfo.flags = 0;
 	bnxt_qplib_rcfw_cmd_prep((struct cmdq_base *)&req,
 				 CMDQ_BASE_OPCODE_CREATE_SRQ,
 				 sizeof(req));
···
 			  struct bnxt_qplib_srq *srq)
 {
 	struct bnxt_qplib_hwq *srq_hwq = &srq->hwq;
-	u32 sw_prod, sw_cons, count = 0;
+	u32 count;
 
-	sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
-	sw_cons = HWQ_CMP(srq_hwq->cons, srq_hwq);
-
-	count = sw_prod > sw_cons ? sw_prod - sw_cons :
-				    srq_hwq->max_elements - sw_cons + sw_prod;
+	count = __bnxt_qplib_get_avail(srq_hwq);
 	if (count > srq->threshold) {
 		srq->arm_req = false;
 		bnxt_qplib_srq_arm_db(&srq->dbinfo, srq->threshold);
···
 	struct bnxt_qplib_hwq *srq_hwq = &srq->hwq;
 	struct rq_wqe *srqe;
 	struct sq_sge *hw_sge;
-	u32 sw_prod, sw_cons, count = 0;
+	u32 count = 0;
 	int i, next;
 
 	spin_lock(&srq_hwq->lock);
···
 	srq->start_idx = srq->swq[next].next_idx;
 	spin_unlock(&srq_hwq->lock);
 
-	sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
-	srqe = bnxt_qplib_get_qe(srq_hwq, sw_prod, NULL);
+	srqe = bnxt_qplib_get_qe(srq_hwq, srq_hwq->prod, NULL);
 	memset(srqe, 0, srq->wqe_size);
 	/* Calculate wqe_size16 and data_len */
 	for (i = 0, hw_sge = (struct sq_sge *)srqe->data;
···
 	srqe->wr_id[0] = cpu_to_le32((u32)next);
 	srq->swq[next].wr_id = wqe->wr_id;
 
-	srq_hwq->prod++;
+	bnxt_qplib_hwq_incr_prod(&srq->dbinfo, srq_hwq, srq->dbinfo.max_slot);
 
 	spin_lock(&srq_hwq->lock);
-	sw_prod = HWQ_CMP(srq_hwq->prod, srq_hwq);
-	/* retaining srq_hwq->cons for this logic
-	 * actually the lock is only required to
-	 * read srq_hwq->cons.
-	 */
-	sw_cons = HWQ_CMP(srq_hwq->cons, srq_hwq);
-	count = sw_prod > sw_cons ? sw_prod - sw_cons :
-				    srq_hwq->max_elements - sw_cons + sw_prod;
+	count = __bnxt_qplib_get_avail(srq_hwq);
 	spin_unlock(&srq_hwq->lock);
 	/* Ring DB */
 	bnxt_qplib_ring_prod_db(&srq->dbinfo, DBC_DBC_TYPE_SRQ);
···
 	u32 tbl_indx;
 	int rc;
 
+	sq->dbinfo.flags = 0;
 	bnxt_qplib_rcfw_cmd_prep((struct cmdq_base *)&req,
 				 CMDQ_BASE_OPCODE_CREATE_QP1,
 				 sizeof(req));
···
 
 	/* RQ */
 	if (rq->max_wqe) {
+		rq->dbinfo.flags = 0;
 		hwq_attr.res = res;
 		hwq_attr.sginfo = &rq->sg_info;
 		hwq_attr.stride = sizeof(struct sq_sge);
···
 	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,
 				 sizeof(req));
···
 
 	/* SQ */
 	if (qp->type == CMDQ_CREATE_QP_TYPE_RC) {
-		psn_sz = bnxt_qplib_is_chip_gen_p5(res->cctx) ?
+		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;
···
 	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)
···
 
 	/* RQ */
 	if (!qp->srq) {
+		rq->dbinfo.flags = 0;
 		hwq_attr.res = res;
 		hwq_attr.sginfo = &rq->sg_info;
 		hwq_attr.stride = sizeof(struct sq_sge);
···
 static void __clean_cq(struct bnxt_qplib_cq *cq, u64 qp)
 {
 	struct bnxt_qplib_hwq *cq_hwq = &cq->hwq;
+	u32 peek_flags, peek_cons;
 	struct cq_base *hw_cqe;
 	int i;
 
+	peek_flags = cq->dbinfo.flags;
+	peek_cons = cq_hwq->cons;
 	for (i = 0; i < cq_hwq->max_elements; i++) {
-		hw_cqe = bnxt_qplib_get_qe(cq_hwq, i, NULL);
-		if (!CQE_CMP_VALID(hw_cqe, i, cq_hwq->max_elements))
+		hw_cqe = bnxt_qplib_get_qe(cq_hwq, peek_cons, NULL);
+		if (!CQE_CMP_VALID(hw_cqe, peek_flags))
 			continue;
 		/*
 		 * The valid test of the entry must be done first before
···
 		default:
 			break;
 		}
+		bnxt_qplib_hwq_incr_cons(cq_hwq->max_elements, &peek_cons,
+					 1, &peek_flags);
 	}
 }
 
···
 	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)
···
 
 	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;
 
···
 	flg_npsn = ((swq->next_psn << SQ_PSN_SEARCH_NEXT_PSN_SFT) &
 		     SQ_PSN_SEARCH_NEXT_PSN_MASK);
 
-	if (bnxt_qplib_is_chip_gen_p5(qp->cctx)) {
+	if (bnxt_qplib_is_chip_gen_p5_p7(qp->cctx)) {
 		psns_ext->opcode_start_psn = cpu_to_le32(op_spsn);
 		psns_ext->flags_next_psn = cpu_to_le32(flg_npsn);
 		psns_ext->start_slot_idx = cpu_to_le16(swq->slot_idx);
···
 	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;
···
 	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);
···
 	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;
···
 	}
 
 	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;
···
 					       &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:
···
 						      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)
···
 		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:
···
 						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;
 	}
···
 		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:
···
 		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(hwq, swq->slots);
+	bnxt_qplib_hwq_incr_prod(&sq->dbinfo, hwq, swq->slots);
 	qp->wqe_cnt++;
 done:
 	if (sch_handler) {
···
 	base_hdr->wr_id[0] = cpu_to_le32(wqe_idx);
 queue_err:
 	bnxt_qplib_swq_mod_start(rq, wqe_idx);
-	bnxt_qplib_hwq_incr_prod(hwq, swq->slots);
+	bnxt_qplib_hwq_incr_prod(&rq->dbinfo, hwq, swq->slots);
 done:
 	if (sch_handler) {
 		nq_work = kzalloc(sizeof(*nq_work), GFP_ATOMIC);
···
 		return -EINVAL;
 	}
 
+	cq->dbinfo.flags = 0;
 	hwq_attr.res = res;
 	hwq_attr.depth = cq->max_wqe;
 	hwq_attr.stride = sizeof(struct cq_base);
···
 
 	req.dpi = cpu_to_le32(cq->dpi->dpi);
 	req.cq_handle = cpu_to_le64(cq->cq_handle);
-	req.cq_size = cpu_to_le32(cq->hwq.max_elements);
+	req.cq_size = cpu_to_le32(cq->max_wqe);
 	pbl = &cq->hwq.pbl[PBL_LVL_0];
 	pg_sz_lvl = (bnxt_qplib_base_pg_size(&cq->hwq) <<
 		     CMDQ_CREATE_CQ_PG_SIZE_SFT);
···
 	cq->dbinfo.xid = cq->id;
 	cq->dbinfo.db = cq->dpi->dbr;
 	cq->dbinfo.priv_db = res->dpi_tbl.priv_db;
+	cq->dbinfo.flags = 0;
+	cq->dbinfo.toggle = 0;
 
 	bnxt_qplib_armen_db(&cq->dbinfo, DBC_DBC_TYPE_CQ_ARMENA);
 
···
 {
 	bnxt_qplib_free_hwq(res, &cq->hwq);
 	memcpy(&cq->hwq, &cq->resize_hwq, sizeof(cq->hwq));
+       /* Reset only the cons bit in the flags */
+	cq->dbinfo.flags &= ~(1UL << BNXT_QPLIB_FLAG_EPOCH_CONS_SHIFT);
 }
 
 int bnxt_qplib_resize_cq(struct bnxt_qplib_res *res, struct bnxt_qplib_cq *cq,
···
 		cqe++;
 		(*budget)--;
 skip_compl:
-		bnxt_qplib_hwq_incr_cons(&sq->hwq, sq->swq[last].slots);
+		bnxt_qplib_hwq_incr_cons(sq->hwq.max_elements, &sq->hwq.cons,
+					 sq->swq[last].slots, &sq->dbinfo.flags);
 		sq->swq_last = sq->swq[last].next_idx;
 	}
 	*pcqe = cqe;
···
 		cqe->wr_id = rq->swq[last].wr_id;
 		cqe++;
 		(*budget)--;
-		bnxt_qplib_hwq_incr_cons(&rq->hwq, rq->swq[last].slots);
+		bnxt_qplib_hwq_incr_cons(rq->hwq.max_elements, &rq->hwq.cons,
+					 rq->swq[last].slots, &rq->dbinfo.flags);
 		rq->swq_last = rq->swq[last].next_idx;
 	}
 	*pcqe = cqe;
···
 static int do_wa9060(struct bnxt_qplib_qp *qp, struct bnxt_qplib_cq *cq,
 		     u32 cq_cons, u32 swq_last, u32 cqe_sq_cons)
 {
-	u32 peek_sw_cq_cons, peek_raw_cq_cons, peek_sq_cons_idx;
+	u32 peek_sw_cq_cons, peek_sq_cons_idx, peek_flags;
 	struct bnxt_qplib_q *sq = &qp->sq;
 	struct cq_req *peek_req_hwcqe;
 	struct bnxt_qplib_qp *peek_qp;
···
 	}
 	if (sq->condition) {
 		/* Peek at the completions */
-		peek_raw_cq_cons = cq->hwq.cons;
+		peek_flags = cq->dbinfo.flags;
 		peek_sw_cq_cons = cq_cons;
 		i = cq->hwq.max_elements;
 		while (i--) {
-			peek_sw_cq_cons = HWQ_CMP((peek_sw_cq_cons), &cq->hwq);
 			peek_hwcqe = bnxt_qplib_get_qe(&cq->hwq,
 						       peek_sw_cq_cons, NULL);
 			/* If the next hwcqe is VALID */
-			if (CQE_CMP_VALID(peek_hwcqe, peek_raw_cq_cons,
-					  cq->hwq.max_elements)) {
+			if (CQE_CMP_VALID(peek_hwcqe, peek_flags)) {
 			/*
 			 * The valid test of the entry must be done first before
 			 * reading any further.
···
 				rc = -EINVAL;
 				goto out;
 			}
-			peek_sw_cq_cons++;
-			peek_raw_cq_cons++;
+			bnxt_qplib_hwq_incr_cons(cq->hwq.max_elements,
+						 &peek_sw_cq_cons,
+						 1, &peek_flags);
 		}
 		dev_err(&cq->hwq.pdev->dev,
 			"Should not have come here! cq_cons=0x%x qp=0x%x sq cons sw=0x%x hw=0x%x\n",
···
 			}
 		}
 skip:
-		bnxt_qplib_hwq_incr_cons(&sq->hwq, swq->slots);
+		bnxt_qplib_hwq_incr_cons(sq->hwq.max_elements, &sq->hwq.cons,
+					 swq->slots, &sq->dbinfo.flags);
 		sq->swq_last = swq->next_idx;
 		if (sq->single)
 			break;
···
 	srq->swq[srq->last_idx].next_idx = (int)tag;
 	srq->last_idx = (int)tag;
 	srq->swq[srq->last_idx].next_idx = -1;
-	srq->hwq.cons++; /* Support for SRQE counter */
+	bnxt_qplib_hwq_incr_cons(srq->hwq.max_elements, &srq->hwq.cons,
+				 srq->dbinfo.max_slot, &srq->dbinfo.flags);
 	spin_unlock(&srq->hwq.lock);
 }
 
···
 		cqe->wr_id = swq->wr_id;
 		cqe++;
 		(*budget)--;
-		bnxt_qplib_hwq_incr_cons(&rq->hwq, swq->slots);
+		bnxt_qplib_hwq_incr_cons(rq->hwq.max_elements, &rq->hwq.cons,
+					 swq->slots, &rq->dbinfo.flags);
 		rq->swq_last = swq->next_idx;
 		*pcqe = cqe;
 
···
 		cqe->wr_id = swq->wr_id;
 		cqe++;
 		(*budget)--;
-		bnxt_qplib_hwq_incr_cons(&rq->hwq, swq->slots);
+		bnxt_qplib_hwq_incr_cons(rq->hwq.max_elements, &rq->hwq.cons,
+					 swq->slots, &rq->dbinfo.flags);
 		rq->swq_last = swq->next_idx;
 		*pcqe = cqe;
 
···
 bool bnxt_qplib_is_cq_empty(struct bnxt_qplib_cq *cq)
 {
 	struct cq_base *hw_cqe;
-	u32 sw_cons, raw_cons;
 	bool rc = true;
 
-	raw_cons = cq->hwq.cons;
-	sw_cons = HWQ_CMP(raw_cons, &cq->hwq);
-	hw_cqe = bnxt_qplib_get_qe(&cq->hwq, sw_cons, NULL);
+	hw_cqe = bnxt_qplib_get_qe(&cq->hwq, cq->hwq.cons, NULL);
 	 /* Check for Valid bit. If the CQE is valid, return false */
-	rc = !CQE_CMP_VALID(hw_cqe, raw_cons, cq->hwq.max_elements);
+	rc = !CQE_CMP_VALID(hw_cqe, cq->dbinfo.flags);
 	return rc;
 }
 
···
 		cqe->wr_id = swq->wr_id;
 		cqe++;
 		(*budget)--;
-		bnxt_qplib_hwq_incr_cons(&rq->hwq, swq->slots);
+		bnxt_qplib_hwq_incr_cons(rq->hwq.max_elements, &rq->hwq.cons,
+					 swq->slots, &rq->dbinfo.flags);
 		rq->swq_last = swq->next_idx;
 		*pcqe = cqe;
 
···
 			cqe++;
 			(*budget)--;
 		}
-		bnxt_qplib_hwq_incr_cons(&sq->hwq, sq->swq[swq_last].slots);
+		bnxt_qplib_hwq_incr_cons(sq->hwq.max_elements, &sq->hwq.cons,
+					 sq->swq[swq_last].slots, &sq->dbinfo.flags);
 		sq->swq_last = sq->swq[swq_last].next_idx;
 	}
 	*pcqe = cqe;
···
 		       int num_cqes, struct bnxt_qplib_qp **lib_qp)
 {
 	struct cq_base *hw_cqe;
-	u32 sw_cons, raw_cons;
 	int budget, rc = 0;
+	u32 hw_polled = 0;
 	u8 type;
 
-	raw_cons = cq->hwq.cons;
 	budget = num_cqes;
 
 	while (budget) {
-		sw_cons = HWQ_CMP(raw_cons, &cq->hwq);
-		hw_cqe = bnxt_qplib_get_qe(&cq->hwq, sw_cons, NULL);
+		hw_cqe = bnxt_qplib_get_qe(&cq->hwq, cq->hwq.cons, NULL);
 
 		/* Check for Valid bit */
-		if (!CQE_CMP_VALID(hw_cqe, raw_cons, cq->hwq.max_elements))
+		if (!CQE_CMP_VALID(hw_cqe, cq->dbinfo.flags))
 			break;
 
 		/*
···
 			rc = bnxt_qplib_cq_process_req(cq,
 						       (struct cq_req *)hw_cqe,
 						       &cqe, &budget,
-						       sw_cons, lib_qp);
+						       cq->hwq.cons, lib_qp);
 			break;
 		case CQ_BASE_CQE_TYPE_RES_RC:
 			rc = bnxt_qplib_cq_process_res_rc(cq,
···
 				dev_err(&cq->hwq.pdev->dev,
 					"process_cqe error rc = 0x%x\n", rc);
 		}
-		raw_cons++;
+		hw_polled++;
+		bnxt_qplib_hwq_incr_cons(cq->hwq.max_elements, &cq->hwq.cons,
+					 1, &cq->dbinfo.flags);
+
 	}
-	if (cq->hwq.cons != raw_cons) {
-		cq->hwq.cons = raw_cons;
+	if (hw_polled)
 		bnxt_qplib_ring_db(&cq->dbinfo, DBC_DBC_TYPE_CQ);
-	}
 exit:
 	return num_cqes - budget;
 }
 
 void bnxt_qplib_req_notify_cq(struct bnxt_qplib_cq *cq, u32 arm_type)
 {
+	cq->dbinfo.toggle = cq->toggle;
 	if (arm_type)
 		bnxt_qplib_ring_db(&cq->dbinfo, arm_type);
 	/* Using cq->arm_state variable to track whether to issue cq handler */

drivers/infiniband/hw/bnxt_re/qplib_fp.h

···
 	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)
···
 #define CQE_IDX(x)		((x) & CQE_MAX_IDX_PER_PG)
 
 #define ROCE_CQE_CMP_V			0
-#define CQE_CMP_VALID(hdr, raw_cons, cp_bit)			\
+#define CQE_CMP_VALID(hdr, pass)			\
 	(!!((hdr)->cqe_type_toggle & CQ_BASE_TOGGLE) ==		\
-	   !((raw_cons) & (cp_bit)))
+	   !((pass) & BNXT_QPLIB_FLAG_EPOCH_CONS_MASK))
+
+static inline u32 __bnxt_qplib_get_avail(struct bnxt_qplib_hwq *hwq)
+{
+	int cons, prod, avail;
+
+	cons = hwq->cons;
+	prod = hwq->prod;
+	avail = cons - prod;
+	if (cons <= prod)
+		avail += hwq->depth;
+	return avail;
+}
 
 static inline bool bnxt_qplib_queue_full(struct bnxt_qplib_q *que,
 					 u8 slots)
···
 	bool				resize_in_progress;
 	struct bnxt_qplib_sg_info	sg_info;
 	u64				cq_handle;
+	u8				toggle;
 
 #define CQ_RESIZE_WAIT_TIME_MS		500
 	unsigned long			flags;
···
 #define NQE_PG(x)		(((x) & ~NQE_MAX_IDX_PER_PG) / NQE_CNT_PER_PG)
 #define NQE_IDX(x)		((x) & NQE_MAX_IDX_PER_PG)
 
-#define NQE_CMP_VALID(hdr, raw_cons, cp_bit)			\
+#define NQE_CMP_VALID(hdr, pass)			\
 	(!!(le32_to_cpu((hdr)->info63_v[0]) & NQ_BASE_V) ==	\
-	   !((raw_cons) & (cp_bit)))
+	   !((pass) & BNXT_QPLIB_FLAG_EPOCH_CONS_MASK))
 
 #define BNXT_QPLIB_NQE_MAX_CNT		(128 * 1024)
 
···
 		size = max;
 
 	return size;
+}
+
+/* MSN table update inlin */
+static inline __le64 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__ */

drivers/infiniband/hw/bnxt_re/qplib_rcfw.c

···
 	u32 type, budget = CREQ_ENTRY_POLL_BUDGET;
 	struct bnxt_qplib_hwq *hwq = &creq->hwq;
 	struct creq_base *creqe;
-	u32 sw_cons, raw_cons;
 	unsigned long flags;
 	u32 num_wakeup = 0;
+	u32 hw_polled = 0;
 
 	/* Service the CREQ until budget is over */
 	spin_lock_irqsave(&hwq->lock, flags);
-	raw_cons = hwq->cons;
 	while (budget > 0) {
-		sw_cons = HWQ_CMP(raw_cons, hwq);
-		creqe = bnxt_qplib_get_qe(hwq, sw_cons, NULL);
-		if (!CREQ_CMP_VALID(creqe, raw_cons, hwq->max_elements))
+		creqe = bnxt_qplib_get_qe(hwq, hwq->cons, NULL);
+		if (!CREQ_CMP_VALID(creqe, creq->creq_db.dbinfo.flags))
 			break;
 		/* The valid test of the entry must be done first before
 		 * reading any further.
···
 					 type);
 			break;
 		}
-		raw_cons++;
 		budget--;
+		hw_polled++;
+		bnxt_qplib_hwq_incr_cons(hwq->max_elements, &hwq->cons,
+					 1, &creq->creq_db.dbinfo.flags);
 	}
 
-	if (hwq->cons != raw_cons) {
-		hwq->cons = raw_cons;
+	if (hw_polled)
 		bnxt_qplib_ring_nq_db(&creq->creq_db.dbinfo,
 				      rcfw->res->cctx, true);
-	}
 	spin_unlock_irqrestore(&hwq->lock, flags);
 	if (num_wakeup)
 		wake_up_nr(&rcfw->cmdq.waitq, num_wakeup);
···
 	 */
 	if (is_virtfn)
 		goto skip_ctx_setup;
-	if (bnxt_qplib_is_chip_gen_p5(rcfw->res->cctx))
+	if (bnxt_qplib_is_chip_gen_p5_p7(rcfw->res->cctx))
 		goto config_vf_res;
 
 	lvl = ctx->qpc_tbl.level;
···
 	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 |= cpu_to_le16(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);
···
 	pdev = rcfw->pdev;
 	creq_db = &rcfw->creq.creq_db;
 
+	creq_db->dbinfo.flags = 0;
 	creq_db->reg.bar_id = RCFW_COMM_CONS_PCI_BAR_REGION;
 	creq_db->reg.bar_base = pci_resource_start(pdev, creq_db->reg.bar_id);
 	if (!creq_db->reg.bar_id)

drivers/infiniband/hw/bnxt_re/qplib_rcfw.h

···
 /* Allocate 1 per QP for async error notification for now */
 #define BNXT_QPLIB_CREQE_MAX_CNT	(64 * 1024)
 #define BNXT_QPLIB_CREQE_UNITS		16	/* 16-Bytes per prod unit */
-#define CREQ_CMP_VALID(hdr, raw_cons, cp_bit)			\
+#define CREQ_CMP_VALID(hdr, pass)			\
 	(!!((hdr)->v & CREQ_BASE_V) ==				\
-	   !((raw_cons) & (cp_bit)))
+	   !((pass) & BNXT_QPLIB_FLAG_EPOCH_CONS_MASK))
 #define CREQ_ENTRY_POLL_BUDGET		0x100
 
 /* HWQ */

drivers/infiniband/hw/bnxt_re/qplib_res.c

···
 	hwq->cons = 0;
 	hwq->pdev = pdev;
 	hwq->depth = hwq_attr->depth;
-	hwq->max_elements = depth;
+	hwq->max_elements = hwq->depth;
 	hwq->element_size = stride;
 	hwq->qe_ppg = pg_size / stride;
 	/* For direct access to the elements */
···
 	dpit = &res->dpi_tbl;
 	reg = &dpit->wcreg;
 
-	if (!bnxt_qplib_is_chip_gen_p5(res->cctx)) {
+	if (!bnxt_qplib_is_chip_gen_p5_p7(res->cctx)) {
 		/* Offest should come from L2 driver */
 		dbr_offset = dev_attr->l2_db_size;
 		dpit->ucreg.offset = dbr_offset;

drivers/infiniband/hw/bnxt_re/qplib_res.h

···
 #define CHIP_NUM_57508		0x1750
 #define CHIP_NUM_57504		0x1751
 #define CHIP_NUM_57502		0x1752
+#define CHIP_NUM_58818          0xd818
+#define CHIP_NUM_57608          0x1760
+
+#define BNXT_QPLIB_DBR_VALID		(0x1UL << 26)
+#define BNXT_QPLIB_DBR_EPOCH_SHIFT	24
+#define BNXT_QPLIB_DBR_TOGGLE_SHIFT	25
 
 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 {
···
 	struct bnxt_qplib_hwq	*hwq;
 	u32			xid;
 	u32			max_slot;
+	u32                     flags;
+	u8			toggle;
+};
+
+enum bnxt_qplib_db_info_flags_mask {
+	BNXT_QPLIB_FLAG_EPOCH_CONS_SHIFT        = 0x0UL,
+	BNXT_QPLIB_FLAG_EPOCH_PROD_SHIFT        = 0x1UL,
+	BNXT_QPLIB_FLAG_EPOCH_CONS_MASK         = 0x1UL,
+	BNXT_QPLIB_FLAG_EPOCH_PROD_MASK         = 0x2UL,
+};
+
+enum bnxt_qplib_db_epoch_flag_shift {
+	BNXT_QPLIB_DB_EPOCH_CONS_SHIFT  = BNXT_QPLIB_DBR_EPOCH_SHIFT,
+	BNXT_QPLIB_DB_EPOCH_PROD_SHIFT  = (BNXT_QPLIB_DBR_EPOCH_SHIFT - 1),
 };
 
 /* Tables */
···
 	struct bnxt_qplib_db_pacing_data *pacing_data;
 };
 
+static inline bool bnxt_qplib_is_chip_gen_p7(struct bnxt_qplib_chip_ctx *cctx)
+{
+	return (cctx->chip_num == CHIP_NUM_58818 ||
+		cctx->chip_num == CHIP_NUM_57608);
+}
+
 static inline bool bnxt_qplib_is_chip_gen_p5(struct bnxt_qplib_chip_ctx *cctx)
 {
 	return (cctx->chip_num == CHIP_NUM_57508 ||
···
 		cctx->chip_num == CHIP_NUM_57502);
 }
 
+static inline bool bnxt_qplib_is_chip_gen_p5_p7(struct bnxt_qplib_chip_ctx *cctx)
+{
+	return bnxt_qplib_is_chip_gen_p5(cctx) || bnxt_qplib_is_chip_gen_p7(cctx);
+}
+
 static inline u8 bnxt_qplib_get_hwq_type(struct bnxt_qplib_res *res)
 {
-	return bnxt_qplib_is_chip_gen_p5(res->cctx) ?
+	return bnxt_qplib_is_chip_gen_p5_p7(res->cctx) ?
 					HWQ_TYPE_QUEUE : HWQ_TYPE_L2_CMPL;
 }
 
 static inline u8 bnxt_qplib_get_ring_type(struct bnxt_qplib_chip_ctx *cctx)
 {
-	return bnxt_qplib_is_chip_gen_p5(cctx) ?
+	return bnxt_qplib_is_chip_gen_p5_p7(cctx) ?
 	       RING_ALLOC_REQ_RING_TYPE_NQ :
 	       RING_ALLOC_REQ_RING_TYPE_ROCE_CMPL;
 }
···
 
 int bnxt_qplib_determine_atomics(struct pci_dev *dev);
 
-static inline void bnxt_qplib_hwq_incr_prod(struct bnxt_qplib_hwq *hwq, u32 cnt)
+static inline void bnxt_qplib_hwq_incr_prod(struct bnxt_qplib_db_info *dbinfo,
+					    struct bnxt_qplib_hwq *hwq, u32 cnt)
 {
-	hwq->prod = (hwq->prod + cnt) % hwq->depth;
+	/* move prod and update toggle/epoch if wrap around */
+	hwq->prod += cnt;
+	if (hwq->prod >= hwq->depth) {
+		hwq->prod %= hwq->depth;
+		dbinfo->flags ^= 1UL << BNXT_QPLIB_FLAG_EPOCH_PROD_SHIFT;
+	}
 }
 
-static inline void bnxt_qplib_hwq_incr_cons(struct bnxt_qplib_hwq *hwq,
-					    u32 cnt)
+static inline void bnxt_qplib_hwq_incr_cons(u32 max_elements, u32 *cons, u32 cnt,
+					    u32 *dbinfo_flags)
 {
-	hwq->cons = (hwq->cons + cnt) % hwq->depth;
+	/* move cons and update toggle/epoch if wrap around */
+	*cons += cnt;
+	if (*cons >= max_elements) {
+		*cons %= max_elements;
+		*dbinfo_flags ^= 1UL << BNXT_QPLIB_FLAG_EPOCH_CONS_SHIFT;
+	}
 }
 
 static inline void bnxt_qplib_ring_db32(struct bnxt_qplib_db_info *info,
 					bool arm)
 {
-	u32 key;
+	u32 key = 0;
 
-	key = info->hwq->cons & (info->hwq->max_elements - 1);
-	key |= (CMPL_DOORBELL_IDX_VALID |
+	key |= info->hwq->cons | (CMPL_DOORBELL_IDX_VALID |
 		(CMPL_DOORBELL_KEY_CMPL & CMPL_DOORBELL_KEY_MASK));
 	if (!arm)
 		key |= CMPL_DOORBELL_MASK;
 	writel(key, info->db);
 }
 
+#define BNXT_QPLIB_INIT_DBHDR(xid, type, indx, toggle) \
+	(((u64)(((xid) & DBC_DBC_XID_MASK) | DBC_DBC_PATH_ROCE |  \
+		(type) | BNXT_QPLIB_DBR_VALID) << 32) | (indx) |  \
+	 (((u32)(toggle)) << (BNXT_QPLIB_DBR_TOGGLE_SHIFT)))
+
 static inline void bnxt_qplib_ring_db(struct bnxt_qplib_db_info *info,
 				      u32 type)
 {
 	u64 key = 0;
+	u32 indx;
+	u8 toggle = 0;
 
-	key = (info->xid & DBC_DBC_XID_MASK) | DBC_DBC_PATH_ROCE | type;
-	key <<= 32;
-	key |= (info->hwq->cons & (info->hwq->max_elements - 1)) &
-		DBC_DBC_INDEX_MASK;
+	if (type == DBC_DBC_TYPE_CQ_ARMALL ||
+	    type == DBC_DBC_TYPE_CQ_ARMSE)
+		toggle = info->toggle;
+
+	indx = (info->hwq->cons & DBC_DBC_INDEX_MASK) |
+	       ((info->flags & BNXT_QPLIB_FLAG_EPOCH_CONS_MASK) <<
+		 BNXT_QPLIB_DB_EPOCH_CONS_SHIFT);
+
+	key =  BNXT_QPLIB_INIT_DBHDR(info->xid, type, indx, toggle);
 	writeq(key, info->db);
 }
 
···
 					   u32 type)
 {
 	u64 key = 0;
+	u32 indx;
 
-	key = (info->xid & DBC_DBC_XID_MASK) | DBC_DBC_PATH_ROCE | type;
-	key <<= 32;
-	key |= ((info->hwq->prod / info->max_slot)) & DBC_DBC_INDEX_MASK;
+	indx = (((info->hwq->prod / info->max_slot) & DBC_DBC_INDEX_MASK) |
+		((info->flags & BNXT_QPLIB_FLAG_EPOCH_PROD_MASK) <<
+		 BNXT_QPLIB_DB_EPOCH_PROD_SHIFT));
+	key = BNXT_QPLIB_INIT_DBHDR(info->xid, type, indx, 0);
 	writeq(key, info->db);
 }
 
···
 				       u32 type)
 {
 	u64 key = 0;
+	u8 toggle = 0;
 
-	key = (info->xid & DBC_DBC_XID_MASK) | DBC_DBC_PATH_ROCE | type;
-	key <<= 32;
+	if (type == DBC_DBC_TYPE_CQ_ARMENA || type == DBC_DBC_TYPE_SRQ_ARMENA)
+		toggle = info->toggle;
+	/* Index always at 0 */
+	key = BNXT_QPLIB_INIT_DBHDR(info->xid, type, 0, toggle);
 	writeq(key, info->priv_db);
 }
 
···
 {
 	u64 key = 0;
 
-	key = (info->xid & DBC_DBC_XID_MASK) | DBC_DBC_PATH_ROCE | th;
-	key <<= 32;
-	key |=  th & DBC_DBC_INDEX_MASK;
+	key = BNXT_QPLIB_INIT_DBHDR(info->xid, DBC_DBC_TYPE_SRQ_ARM, th, info->toggle);
 	writeq(key, info->priv_db);
 }
 
···
 	u32 type;
 
 	type = arm ? DBC_DBC_TYPE_NQ_ARM : DBC_DBC_TYPE_NQ;
-	if (bnxt_qplib_is_chip_gen_p5(cctx))
+	if (bnxt_qplib_is_chip_gen_p5_p7(cctx))
 		bnxt_qplib_ring_db(info, type);
 	else
 		bnxt_qplib_ring_db32(info, arm);
···
 	return 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)
 {

drivers/infiniband/hw/bnxt_re/qplib_sp.c

···
 {
 	u16 pcie_ctl2 = 0;
 
-	if (!bnxt_qplib_is_chip_gen_p5(rcfw->res->cctx))
+	if (!bnxt_qplib_is_chip_gen_p5_p7(rcfw->res->cctx))
 		return false;
 
 	pcie_capability_read_word(rcfw->pdev, PCI_EXP_DEVCTL2, &pcie_ctl2);
···
 	 * reporting the max number
 	 */
 	attr->max_qp_wqes -= BNXT_QPLIB_RESERVED_QP_WRS + 1;
-	attr->max_qp_sges = bnxt_qplib_is_chip_gen_p5(rcfw->res->cctx) ?
+	attr->max_qp_sges = bnxt_qplib_is_chip_gen_p5_p7(rcfw->res->cctx) ?
 			    6 : sb->max_sge;
 	attr->max_cq = le32_to_cpu(sb->max_cq);
 	attr->max_cq_wqes = le32_to_cpu(sb->max_cqe);
···
 	attr->max_srq_sges = sb->max_srq_sge;
 	attr->max_pkey = 1;
 	attr->max_inline_data = le32_to_cpu(sb->max_inline_data);
-	attr->l2_db_size = (sb->l2_db_space_size + 1) *
-			    (0x01 << RCFW_DBR_BASE_PAGE_SHIFT);
+	if (!bnxt_qplib_is_chip_gen_p7(rcfw->res->cctx))
+		attr->l2_db_size = (sb->l2_db_space_size + 1) *
+				    (0x01 << RCFW_DBR_BASE_PAGE_SHIFT);
 	attr->max_sgid = BNXT_QPLIB_NUM_GIDS_SUPPORTED;
 	attr->dev_cap_flags = le16_to_cpu(sb->dev_cap_flags);
 
···
 	req->inactivity_th = cpu_to_le16(cc_param->inact_th);
 
 	/* For chip gen P5 onwards fill extended cmd and header */
-	if (bnxt_qplib_is_chip_gen_p5(res->cctx)) {
+	if (bnxt_qplib_is_chip_gen_p5_p7(res->cctx)) {
 		struct roce_tlv *hdr;
 		u32 payload;
 		u32 chunks;

drivers/infiniband/hw/bnxt_re/roce_hsi.h

···
 	__le16	flags;
 	__le16	cookie;
 	u8	resp_size;
-	u8	reserved8;
+	u8	qp_type;
+	#define CMDQ_MODIFY_QP_QP_TYPE_RC            0x2UL
+	#define CMDQ_MODIFY_QP_QP_TYPE_UD            0x4UL
+	#define CMDQ_MODIFY_QP_QP_TYPE_RAW_ETHERTYPE 0x6UL
+	#define CMDQ_MODIFY_QP_QP_TYPE_GSI           0x7UL
+	#define CMDQ_MODIFY_QP_QP_TYPE_LAST         CMDQ_MODIFY_QP_QP_TYPE_GSI
 	__le64	resp_addr;
 	__le32	modify_mask;
 	#define CMDQ_MODIFY_QP_MODIFY_MASK_STATE                   0x1UL
···
 	#define CMDQ_MODIFY_QP_NETWORK_TYPE_ROCEV2_IPV6  (0x3UL << 6)
 	#define CMDQ_MODIFY_QP_NETWORK_TYPE_LAST        CMDQ_MODIFY_QP_NETWORK_TYPE_ROCEV2_IPV6
 	u8	access;
-	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC_REMOTE_READ_REMOTE_WRITE_LOCAL_WRITE_MASK \
-		0xffUL
-	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC_REMOTE_READ_REMOTE_WRITE_LOCAL_WRITE_SFT	\
-		0
-	#define CMDQ_MODIFY_QP_ACCESS_LOCAL_WRITE	0x1UL
-	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_WRITE	0x2UL
-	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_READ	0x4UL
-	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC	0x8UL
+	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC_REMOTE_READ_REMOTE_WRITE_LOCAL_WRITE_MASK 0xffUL
+	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC_REMOTE_READ_REMOTE_WRITE_LOCAL_WRITE_SFT 0
+	#define CMDQ_MODIFY_QP_ACCESS_LOCAL_WRITE   0x1UL
+	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_WRITE  0x2UL
+	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_READ   0x4UL
+	#define CMDQ_MODIFY_QP_ACCESS_REMOTE_ATOMIC 0x8UL
 	__le16	pkey;
 	__le32	qkey;
 	__le32	dgid[4];
···
 	#define CMDQ_MODIFY_QP_VLAN_PCP_SFT 13
 	__le64	irrq_addr;
 	__le64	orrq_addr;
+	__le32	ext_modify_mask;
+	#define CMDQ_MODIFY_QP_EXT_MODIFY_MASK_EXT_STATS_CTX     0x1UL
+	#define CMDQ_MODIFY_QP_EXT_MODIFY_MASK_SCHQ_ID_VALID     0x2UL
+	__le32	ext_stats_ctx_id;
+	__le16	schq_id;
+	__le16	unused_0;
+	__le32	reserved32;
 };
 
 /* creq_modify_qp_resp (size:128b/16B) */
···
 	__le32	reserved32;
 };
 
+/* sq_msn_search (size:64b/8B) */
+struct sq_msn_search {
+	__le64	start_idx_next_psn_start_psn;
+	#define SQ_MSN_SEARCH_START_PSN_MASK 0xffffffUL
+	#define SQ_MSN_SEARCH_START_PSN_SFT 0
+	#define SQ_MSN_SEARCH_NEXT_PSN_MASK 0xffffff000000ULL
+	#define SQ_MSN_SEARCH_NEXT_PSN_SFT  24
+	#define SQ_MSN_SEARCH_START_IDX_MASK 0xffff000000000000ULL
+	#define SQ_MSN_SEARCH_START_IDX_SFT 48
+};
+
 /* sq_send (size:1024b/128B) */
 struct sq_send {
 	u8	wqe_type;
···
 	#define CQ_BASE_CQE_TYPE_RES_UD          (0x2UL << 1)
 	#define CQ_BASE_CQE_TYPE_RES_RAWETH_QP1  (0x3UL << 1)
 	#define CQ_BASE_CQE_TYPE_RES_UD_CFA      (0x4UL << 1)
+	#define CQ_BASE_CQE_TYPE_REQ_V3             (0x8UL << 1)
+	#define CQ_BASE_CQE_TYPE_RES_RC_V3          (0x9UL << 1)
+	#define CQ_BASE_CQE_TYPE_RES_UD_V3          (0xaUL << 1)
+	#define CQ_BASE_CQE_TYPE_RES_RAWETH_QP1_V3  (0xbUL << 1)
+	#define CQ_BASE_CQE_TYPE_RES_UD_CFA_V3      (0xcUL << 1)
 	#define CQ_BASE_CQE_TYPE_NO_OP           (0xdUL << 1)
 	#define CQ_BASE_CQE_TYPE_TERMINAL        (0xeUL << 1)
 	#define CQ_BASE_CQE_TYPE_CUT_OFF         (0xfUL << 1)
 	#define CQ_BASE_CQE_TYPE_LAST           CQ_BASE_CQE_TYPE_CUT_OFF
 	u8	status;
+	#define CQ_BASE_STATUS_OK                         0x0UL
+	#define CQ_BASE_STATUS_BAD_RESPONSE_ERR           0x1UL
+	#define CQ_BASE_STATUS_LOCAL_LENGTH_ERR           0x2UL
+	#define CQ_BASE_STATUS_HW_LOCAL_LENGTH_ERR        0x3UL
+	#define CQ_BASE_STATUS_LOCAL_QP_OPERATION_ERR     0x4UL
+	#define CQ_BASE_STATUS_LOCAL_PROTECTION_ERR       0x5UL
+	#define CQ_BASE_STATUS_LOCAL_ACCESS_ERROR         0x6UL
+	#define CQ_BASE_STATUS_MEMORY_MGT_OPERATION_ERR   0x7UL
+	#define CQ_BASE_STATUS_REMOTE_INVALID_REQUEST_ERR 0x8UL
+	#define CQ_BASE_STATUS_REMOTE_ACCESS_ERR          0x9UL
+	#define CQ_BASE_STATUS_REMOTE_OPERATION_ERR       0xaUL
+	#define CQ_BASE_STATUS_RNR_NAK_RETRY_CNT_ERR      0xbUL
+	#define CQ_BASE_STATUS_TRANSPORT_RETRY_CNT_ERR    0xcUL
+	#define CQ_BASE_STATUS_WORK_REQUEST_FLUSHED_ERR   0xdUL
+	#define CQ_BASE_STATUS_HW_FLUSH_ERR               0xeUL
+	#define CQ_BASE_STATUS_OVERFLOW_ERR               0xfUL
+	#define CQ_BASE_STATUS_LAST                      CQ_BASE_STATUS_OVERFLOW_ERR
 	__le16	reserved16;
-	__le32	reserved32;
+	__le32	opaque;
 };
 
 /* cq_req (size:256b/32B) */
···
 	#define CQ_CUTOFF_CQE_TYPE_SFT    1
 	#define CQ_CUTOFF_CQE_TYPE_CUT_OFF  (0xfUL << 1)
 	#define CQ_CUTOFF_CQE_TYPE_LAST    CQ_CUTOFF_CQE_TYPE_CUT_OFF
+	#define CQ_CUTOFF_RESIZE_TOGGLE_MASK 0x60UL
+	#define CQ_CUTOFF_RESIZE_TOGGLE_SFT 5
 	u8	status;
 	#define CQ_CUTOFF_STATUS_OK 0x0UL
 	#define CQ_CUTOFF_STATUS_LAST CQ_CUTOFF_STATUS_OK
···
 	#define NQ_SRQ_EVENT_TYPE_SFT      0
 	#define NQ_SRQ_EVENT_TYPE_SRQ_EVENT  0x32UL
 	#define NQ_SRQ_EVENT_TYPE_LAST      NQ_SRQ_EVENT_TYPE_SRQ_EVENT
+	#define NQ_SRQ_EVENT_TOGGLE_MASK   0xc0UL
+	#define NQ_SRQ_EVENT_TOGGLE_SFT    6
 	u8	event;
 	#define NQ_SRQ_EVENT_EVENT_SRQ_THRESHOLD_EVENT 0x1UL
 	#define NQ_SRQ_EVENT_EVENT_LAST               NQ_SRQ_EVENT_EVENT_SRQ_THRESHOLD_EVENT

drivers/infiniband/hw/efa/efa.h

···
 /* 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_
···
 	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 {

drivers/infiniband/hw/efa/efa_admin_cmds_defs.h

···
 /* 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_
···
 	 * 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 {
···
 #define EFA_ADMIN_REG_MR_CMD_LOCAL_WRITE_ENABLE_MASK        BIT(0)
 #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)

drivers/infiniband/hw/efa/efa_com_cmd.c

···
 // 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"
···
 
 	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;
 }

drivers/infiniband/hw/efa/efa_com_cmd.h

···
 /* 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_
···
 	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
···
 	 * accessed memory region
 	 */
 	u32 r_key;
+	struct efa_com_mr_interconnect_info ic_info;
 };
 
 struct efa_com_dereg_mr_params {

drivers/infiniband/hw/efa/efa_main.c

···
 // 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>
···
 #include <linux/version.h>
 
 #include <rdma/ib_user_verbs.h>
+#include <rdma/uverbs_ioctl.h>
 
 #include "efa.h"
 
···
 #define EFA_AENQ_ENABLED_GROUPS \
 	(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,
···
 	dev->ibdev.dev.parent = &pdev->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)

drivers/infiniband/hw/efa/efa_verbs.c

···
 // 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>
···
 #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"
···
 	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;
···
 err_out:
 	atomic64_inc(&dev->stats.reg_mr_err);
 	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)
···
 	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),
+	{},
+};

drivers/infiniband/hw/erdma/erdma.h

···
 
 	atomic_t num_ctx;
 	struct list_head cep_list;
+
+	struct dma_pool *resp_pool;
 };
 
 static inline void *get_queue_entry(void *qbuf, u32 idx, u32 depth, u32 shift)

drivers/infiniband/hw/erdma/erdma_hw.h

···
 	CMDQ_OPCODE_DESTROY_EQ = 1,
 	CMDQ_OPCODE_QUERY_FW_INFO = 2,
 	CMDQ_OPCODE_CONF_MTU = 3,
+	CMDQ_OPCODE_GET_STATS = 4,
 	CMDQ_OPCODE_CONF_DEVICE = 5,
 	CMDQ_OPCODE_ALLOC_DB = 8,
 	CMDQ_OPCODE_FREE_DB = 9,
···
 	u32 qpn;
 	u32 sq_pi;
 	u32 rq_pi;
+};
+
+#define ERDMA_HW_RESP_SIZE 256
+
+struct erdma_cmdq_query_req {
+	u64 hdr;
+	u32 rsvd;
+	u32 index;
+
+	u64 target_addr;
+	u32 target_length;
+};
+
+#define ERDMA_HW_RESP_MAGIC 0x5566
+
+struct erdma_cmdq_query_resp_hdr {
+	u16 magic;
+	u8 ver;
+	u8 length;
+
+	u32 index;
+	u32 rsvd[2];
+};
+
+struct erdma_cmdq_query_stats_resp {
+	struct erdma_cmdq_query_resp_hdr hdr;
+
+	u64 tx_req_cnt;
+	u64 tx_packets_cnt;
+	u64 tx_bytes_cnt;
+	u64 tx_drop_packets_cnt;
+	u64 tx_bps_meter_drop_packets_cnt;
+	u64 tx_pps_meter_drop_packets_cnt;
+	u64 rx_packets_cnt;
+	u64 rx_bytes_cnt;
+	u64 rx_drop_packets_cnt;
+	u64 rx_bps_meter_drop_packets_cnt;
+	u64 rx_pps_meter_drop_packets_cnt;
 };
 
 /* cap qword 0 definition */

drivers/infiniband/hw/erdma/erdma_main.c

···
 {
 	int ret;
 
+	dev->resp_pool = dma_pool_create("erdma_resp_pool", &pdev->dev,
+					 ERDMA_HW_RESP_SIZE, ERDMA_HW_RESP_SIZE,
+					 0);
+	if (!dev->resp_pool)
+		return -ENOMEM;
+
 	ret = dma_set_mask_and_coherent(&pdev->dev,
 					DMA_BIT_MASK(ERDMA_PCI_WIDTH));
 	if (ret)
-		return ret;
+		goto destroy_pool;
 
 	dma_set_max_seg_size(&pdev->dev, UINT_MAX);
 
 	return 0;
+
+destroy_pool:
+	dma_pool_destroy(dev->resp_pool);
+
+	return ret;
+}
+
+static void erdma_device_uninit(struct erdma_dev *dev)
+{
+	dma_pool_destroy(dev->resp_pool);
 }
 
 static void erdma_hw_reset(struct erdma_dev *dev)
···
 
 	err = erdma_request_vectors(dev);
 	if (err)
-		goto err_iounmap_func_bar;
+		goto err_uninit_device;
 
 	err = erdma_comm_irq_init(dev);
 	if (err)
···
 err_free_vectors:
 	pci_free_irq_vectors(dev->pdev);
 
+err_uninit_device:
+	erdma_device_uninit(dev);
+
 err_iounmap_func_bar:
 	devm_iounmap(&pdev->dev, dev->func_bar);
 
···
 	erdma_aeq_destroy(dev);
 	erdma_comm_irq_uninit(dev);
 	pci_free_irq_vectors(dev->pdev);
+	erdma_device_uninit(dev);
 
 	devm_iounmap(&pdev->dev, dev->func_bar);
 	pci_release_selected_regions(pdev, ERDMA_BAR_MASK);
···
 	.driver_id = RDMA_DRIVER_ERDMA,
 	.uverbs_abi_ver = ERDMA_ABI_VERSION,
 
+	.alloc_hw_port_stats = erdma_alloc_hw_port_stats,
 	.alloc_mr = erdma_ib_alloc_mr,
 	.alloc_pd = erdma_alloc_pd,
 	.alloc_ucontext = erdma_alloc_ucontext,
···
 	.destroy_cq = erdma_destroy_cq,
 	.destroy_qp = erdma_destroy_qp,
 	.get_dma_mr = erdma_get_dma_mr,
+	.get_hw_stats = erdma_get_hw_stats,
 	.get_port_immutable = erdma_get_port_immutable,
 	.iw_accept = erdma_accept,
 	.iw_add_ref = erdma_qp_get_ref,

drivers/infiniband/hw/erdma/erdma_verbs.c

···
 
 	ib_dispatch_event(&event);
 }
+
+enum counters {
+	ERDMA_STATS_TX_REQS_CNT,
+	ERDMA_STATS_TX_PACKETS_CNT,
+	ERDMA_STATS_TX_BYTES_CNT,
+	ERDMA_STATS_TX_DISABLE_DROP_CNT,
+	ERDMA_STATS_TX_BPS_METER_DROP_CNT,
+	ERDMA_STATS_TX_PPS_METER_DROP_CNT,
+
+	ERDMA_STATS_RX_PACKETS_CNT,
+	ERDMA_STATS_RX_BYTES_CNT,
+	ERDMA_STATS_RX_DISABLE_DROP_CNT,
+	ERDMA_STATS_RX_BPS_METER_DROP_CNT,
+	ERDMA_STATS_RX_PPS_METER_DROP_CNT,
+
+	ERDMA_STATS_MAX
+};
+
+static const struct rdma_stat_desc erdma_descs[] = {
+	[ERDMA_STATS_TX_REQS_CNT].name = "tx_reqs_cnt",
+	[ERDMA_STATS_TX_PACKETS_CNT].name = "tx_packets_cnt",
+	[ERDMA_STATS_TX_BYTES_CNT].name = "tx_bytes_cnt",
+	[ERDMA_STATS_TX_DISABLE_DROP_CNT].name = "tx_disable_drop_cnt",
+	[ERDMA_STATS_TX_BPS_METER_DROP_CNT].name = "tx_bps_limit_drop_cnt",
+	[ERDMA_STATS_TX_PPS_METER_DROP_CNT].name = "tx_pps_limit_drop_cnt",
+	[ERDMA_STATS_RX_PACKETS_CNT].name = "rx_packets_cnt",
+	[ERDMA_STATS_RX_BYTES_CNT].name = "rx_bytes_cnt",
+	[ERDMA_STATS_RX_DISABLE_DROP_CNT].name = "rx_disable_drop_cnt",
+	[ERDMA_STATS_RX_BPS_METER_DROP_CNT].name = "rx_bps_limit_drop_cnt",
+	[ERDMA_STATS_RX_PPS_METER_DROP_CNT].name = "rx_pps_limit_drop_cnt",
+};
+
+struct rdma_hw_stats *erdma_alloc_hw_port_stats(struct ib_device *device,
+						u32 port_num)
+{
+	return rdma_alloc_hw_stats_struct(erdma_descs, ERDMA_STATS_MAX,
+					  RDMA_HW_STATS_DEFAULT_LIFESPAN);
+}
+
+static int erdma_query_hw_stats(struct erdma_dev *dev,
+				struct rdma_hw_stats *stats)
+{
+	struct erdma_cmdq_query_stats_resp *resp;
+	struct erdma_cmdq_query_req req;
+	dma_addr_t dma_addr;
+	int err;
+
+	erdma_cmdq_build_reqhdr(&req.hdr, CMDQ_SUBMOD_COMMON,
+				CMDQ_OPCODE_GET_STATS);
+
+	resp = dma_pool_zalloc(dev->resp_pool, GFP_KERNEL, &dma_addr);
+	if (!resp)
+		return -ENOMEM;
+
+	req.target_addr = dma_addr;
+	req.target_length = ERDMA_HW_RESP_SIZE;
+
+	err = erdma_post_cmd_wait(&dev->cmdq, &req, sizeof(req), NULL, NULL);
+	if (err)
+		goto out;
+
+	if (resp->hdr.magic != ERDMA_HW_RESP_MAGIC) {
+		err = -EINVAL;
+		goto out;
+	}
+
+	memcpy(&stats->value[0], &resp->tx_req_cnt,
+	       sizeof(u64) * stats->num_counters);
+
+out:
+	dma_pool_free(dev->resp_pool, resp, dma_addr);
+
+	return err;
+}
+
+int erdma_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,
+		       u32 port, int index)
+{
+	struct erdma_dev *dev = to_edev(ibdev);
+	int ret;
+
+	if (port == 0)
+		return 0;
+
+	ret = erdma_query_hw_stats(dev, stats);
+	if (ret)
+		return ret;
+
+	return stats->num_counters;
+}

drivers/infiniband/hw/erdma/erdma_verbs.h

···
 		    unsigned int *sg_offset);
 void erdma_port_event(struct erdma_dev *dev, enum ib_event_type reason);
 void erdma_set_mtu(struct erdma_dev *dev, u32 mtu);
+struct rdma_hw_stats *erdma_alloc_hw_port_stats(struct ib_device *device,
+						u32 port_num);
+int erdma_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,
+		       u32 port, int index);
 
 #endif

drivers/infiniband/hw/hfi1/user_exp_rcv.c

···
 	if (unlikely(tinfo->tidcnt > fd->tid_used))
 		return -EINVAL;
 
-	tidinfo = memdup_user(u64_to_user_ptr(tinfo->tidlist),
-			      sizeof(tidinfo[0]) * tinfo->tidcnt);
+	tidinfo = memdup_array_user(u64_to_user_ptr(tinfo->tidlist),
+				    tinfo->tidcnt, sizeof(tidinfo[0]));
 	if (IS_ERR(tidinfo))
 		return PTR_ERR(tidinfo);
 

drivers/infiniband/hw/hfi1/user_sdma.c

···
 		 * equal to the pkt count. However, there is no way to
 		 * tell at this point.
 		 */
-		tmp = memdup_user(iovec[idx].iov_base,
-				  ntids * sizeof(*req->tids));
+		tmp = memdup_array_user(iovec[idx].iov_base,
+					ntids, sizeof(*req->tids));
 		if (IS_ERR(tmp)) {
 			ret = PTR_ERR(tmp);
 			SDMA_DBG(req, "Failed to copy %d TIDs (%d)",

drivers/infiniband/hw/hns/Makefile

···
 
 hns-roce-objs := hns_roce_main.o hns_roce_cmd.o hns_roce_pd.o \
 	hns_roce_ah.o hns_roce_hem.o hns_roce_mr.o hns_roce_qp.o \
-	hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o hns_roce_srq.o hns_roce_restrack.o
+	hns_roce_cq.o hns_roce_alloc.o hns_roce_db.o hns_roce_srq.o hns_roce_restrack.o \
+	hns_roce_debugfs.o
 
 ifdef CONFIG_INFINIBAND_HNS_HIP08
 hns-roce-hw-v2-objs := hns_roce_hw_v2.o $(hns-roce-objs)

drivers/infiniband/hw/hns/hns_roce_ah.c

···
 	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;
···
 		ret = rdma_read_gid_l2_fields(ah_attr->grh.sgid_attr,
 					      &ah->av.vlan_id, NULL);
 		if (ret)
-			return ret;
+			goto err_out;
 
 		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]);
 
 	return ret;
 }

drivers/infiniband/hw/hns/hns_roce_cmd.c

···
 static int hns_roce_cmd_mbox_post_hw(struct hns_roce_dev *hr_dev,
 				     struct hns_roce_mbox_msg *mbox_msg)
 {
-	return hr_dev->hw->post_mbox(hr_dev, mbox_msg);
+	int ret;
+
+	ret = hr_dev->hw->post_mbox(hr_dev, mbox_msg);
+	if (ret)
+		return ret;
+
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MBX_POSTED_CNT]);
+
+	return 0;
 }
 
 /* this should be called with "poll_sem" */
···
 		return ret;
 	}
 
-	return hr_dev->hw->poll_mbox_done(hr_dev);
+	ret = hr_dev->hw->poll_mbox_done(hr_dev);
+	if (ret)
+		return ret;
+
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MBX_POLLED_CNT]);
+
+	return 0;
 }
 
 static int hns_roce_cmd_mbox_poll(struct hns_roce_dev *hr_dev,
···
 	context->result = (status == HNS_ROCE_CMD_SUCCESS) ? 0 : (-EIO);
 	context->out_param = out_param;
 	complete(&context->done);
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MBX_EVENT_CNT]);
 }
 
 static int __hns_roce_cmd_mbox_wait(struct hns_roce_dev *hr_dev,

drivers/infiniband/hw/hns/hns_roce_cq.c

···
 	struct hns_roce_ib_create_cq ucmd = {};
 	int ret;
 
-	if (attr->flags)
-		return -EOPNOTSUPP;
+	if (attr->flags) {
+		ret = -EOPNOTSUPP;
+		goto err_out;
+	}
 
 	ret = verify_cq_create_attr(hr_dev, attr);
 	if (ret)
-		return ret;
+		goto err_out;
 
 	if (udata) {
 		ret = get_cq_ucmd(hr_cq, udata, &ucmd);
 		if (ret)
-			return ret;
+			goto err_out;
 	}
 
 	set_cq_param(hr_cq, attr->cqe, attr->comp_vector, &ucmd);
 
 	ret = set_cqe_size(hr_cq, udata, &ucmd);
 	if (ret)
-		return ret;
+		goto err_out;
 
 	ret = alloc_cq_buf(hr_dev, hr_cq, udata, ucmd.buf_addr);
 	if (ret) {
 		ibdev_err(ibdev, "failed to alloc CQ buf, ret = %d.\n", ret);
-		return ret;
+		goto err_out;
 	}
 
 	ret = alloc_cq_db(hr_dev, hr_cq, udata, ucmd.db_addr, &resp);
···
 	free_cq_db(hr_dev, hr_cq, udata);
 err_cq_buf:
 	free_cq_buf(hr_dev, hr_cq);
+err_out:
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CQ_CREATE_ERR_CNT]);
+
 	return ret;
 }
 

drivers/infiniband/hw/hns/hns_roce_debugfs.c

···
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Copyright (c) 2023 Hisilicon Limited.
+ */
+
+#include <linux/debugfs.h>
+#include <linux/device.h>
+
+#include "hns_roce_device.h"
+
+static struct dentry *hns_roce_dbgfs_root;
+
+static int hns_debugfs_seqfile_open(struct inode *inode, struct file *f)
+{
+	struct hns_debugfs_seqfile *seqfile = inode->i_private;
+
+	return single_open(f, seqfile->read, seqfile->data);
+}
+
+static const struct file_operations hns_debugfs_seqfile_fops = {
+	.owner = THIS_MODULE,
+	.open = hns_debugfs_seqfile_open,
+	.release = single_release,
+	.read = seq_read,
+	.llseek = seq_lseek
+};
+
+static void init_debugfs_seqfile(struct hns_debugfs_seqfile *seq,
+				 const char *name, struct dentry *parent,
+				 int (*read_fn)(struct seq_file *, void *),
+				 void *data)
+{
+	debugfs_create_file(name, 0400, parent, seq, &hns_debugfs_seqfile_fops);
+
+	seq->read = read_fn;
+	seq->data = data;
+}
+
+static const char * const sw_stat_info[] = {
+	[HNS_ROCE_DFX_AEQE_CNT] = "aeqe",
+	[HNS_ROCE_DFX_CEQE_CNT] = "ceqe",
+	[HNS_ROCE_DFX_CMDS_CNT] = "cmds",
+	[HNS_ROCE_DFX_CMDS_ERR_CNT] = "cmds_err",
+	[HNS_ROCE_DFX_MBX_POSTED_CNT] = "posted_mbx",
+	[HNS_ROCE_DFX_MBX_POLLED_CNT] = "polled_mbx",
+	[HNS_ROCE_DFX_MBX_EVENT_CNT] = "mbx_event",
+	[HNS_ROCE_DFX_QP_CREATE_ERR_CNT] = "qp_create_err",
+	[HNS_ROCE_DFX_QP_MODIFY_ERR_CNT] = "qp_modify_err",
+	[HNS_ROCE_DFX_CQ_CREATE_ERR_CNT] = "cq_create_err",
+	[HNS_ROCE_DFX_CQ_MODIFY_ERR_CNT] = "cq_modify_err",
+	[HNS_ROCE_DFX_SRQ_CREATE_ERR_CNT] = "srq_create_err",
+	[HNS_ROCE_DFX_SRQ_MODIFY_ERR_CNT] = "srq_modify_err",
+	[HNS_ROCE_DFX_XRCD_ALLOC_ERR_CNT] = "xrcd_alloc_err",
+	[HNS_ROCE_DFX_MR_REG_ERR_CNT] = "mr_reg_err",
+	[HNS_ROCE_DFX_MR_REREG_ERR_CNT] = "mr_rereg_err",
+	[HNS_ROCE_DFX_AH_CREATE_ERR_CNT] = "ah_create_err",
+	[HNS_ROCE_DFX_MMAP_ERR_CNT] = "mmap_err",
+	[HNS_ROCE_DFX_UCTX_ALLOC_ERR_CNT] = "uctx_alloc_err",
+};
+
+static int sw_stat_debugfs_show(struct seq_file *file, void *offset)
+{
+	struct hns_roce_dev *hr_dev = file->private;
+	int i;
+
+	for (i = 0; i < HNS_ROCE_DFX_CNT_TOTAL; i++)
+		seq_printf(file, "%-20s --- %lld\n", sw_stat_info[i],
+			   atomic64_read(&hr_dev->dfx_cnt[i]));
+
+	return 0;
+}
+
+static void create_sw_stat_debugfs(struct hns_roce_dev *hr_dev,
+				   struct dentry *parent)
+{
+	struct hns_sw_stat_debugfs *dbgfs = &hr_dev->dbgfs.sw_stat_root;
+
+	dbgfs->root = debugfs_create_dir("sw_stat", parent);
+
+	init_debugfs_seqfile(&dbgfs->sw_stat, "sw_stat", dbgfs->root,
+			     sw_stat_debugfs_show, hr_dev);
+}
+
+/* debugfs for device */
+void hns_roce_register_debugfs(struct hns_roce_dev *hr_dev)
+{
+	struct hns_roce_dev_debugfs *dbgfs = &hr_dev->dbgfs;
+
+	dbgfs->root = debugfs_create_dir(dev_name(&hr_dev->ib_dev.dev),
+					 hns_roce_dbgfs_root);
+
+	create_sw_stat_debugfs(hr_dev, dbgfs->root);
+}
+
+void hns_roce_unregister_debugfs(struct hns_roce_dev *hr_dev)
+{
+	debugfs_remove_recursive(hr_dev->dbgfs.root);
+}
+
+/* debugfs for hns module */
+void hns_roce_init_debugfs(void)
+{
+	hns_roce_dbgfs_root = debugfs_create_dir("hns_roce", NULL);
+}
+
+void hns_roce_cleanup_debugfs(void)
+{
+	debugfs_remove_recursive(hns_roce_dbgfs_root);
+	hns_roce_dbgfs_root = NULL;
+}

drivers/infiniband/hw/hns/hns_roce_debugfs.h

···
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Copyright (c) 2023 Hisilicon Limited.
+ */
+
+#ifndef __HNS_ROCE_DEBUGFS_H
+#define __HNS_ROCE_DEBUGFS_H
+
+/* debugfs seqfile */
+struct hns_debugfs_seqfile {
+	int (*read)(struct seq_file *seq, void *data);
+	void *data;
+};
+
+struct hns_sw_stat_debugfs {
+	struct dentry *root;
+	struct hns_debugfs_seqfile sw_stat;
+};
+
+/* Debugfs for device */
+struct hns_roce_dev_debugfs {
+	struct dentry *root;
+	struct hns_sw_stat_debugfs sw_stat_root;
+};
+
+struct hns_roce_dev;
+
+void hns_roce_init_debugfs(void);
+void hns_roce_cleanup_debugfs(void);
+void hns_roce_register_debugfs(struct hns_roce_dev *hr_dev);
+void hns_roce_unregister_debugfs(struct hns_roce_dev *hr_dev);
+
+#endif

drivers/infiniband/hw/hns/hns_roce_device.h

···
 
 #include <rdma/ib_verbs.h>
 #include <rdma/hns-abi.h>
+#include "hns_roce_debugfs.h"
 
 #define PCI_REVISION_ID_HIP08			0x21
 #define PCI_REVISION_ID_HIP09			0x30
···
 	HNS_ROCE_HW_CNT_TOTAL
 };
 
+enum hns_roce_sw_dfx_stat_index {
+	HNS_ROCE_DFX_AEQE_CNT,
+	HNS_ROCE_DFX_CEQE_CNT,
+	HNS_ROCE_DFX_CMDS_CNT,
+	HNS_ROCE_DFX_CMDS_ERR_CNT,
+	HNS_ROCE_DFX_MBX_POSTED_CNT,
+	HNS_ROCE_DFX_MBX_POLLED_CNT,
+	HNS_ROCE_DFX_MBX_EVENT_CNT,
+	HNS_ROCE_DFX_QP_CREATE_ERR_CNT,
+	HNS_ROCE_DFX_QP_MODIFY_ERR_CNT,
+	HNS_ROCE_DFX_CQ_CREATE_ERR_CNT,
+	HNS_ROCE_DFX_CQ_MODIFY_ERR_CNT,
+	HNS_ROCE_DFX_SRQ_CREATE_ERR_CNT,
+	HNS_ROCE_DFX_SRQ_MODIFY_ERR_CNT,
+	HNS_ROCE_DFX_XRCD_ALLOC_ERR_CNT,
+	HNS_ROCE_DFX_MR_REG_ERR_CNT,
+	HNS_ROCE_DFX_MR_REREG_ERR_CNT,
+	HNS_ROCE_DFX_AH_CREATE_ERR_CNT,
+	HNS_ROCE_DFX_MMAP_ERR_CNT,
+	HNS_ROCE_DFX_UCTX_ALLOC_ERR_CNT,
+	HNS_ROCE_DFX_CNT_TOTAL
+};
+
 struct hns_roce_hw {
 	int (*cmq_init)(struct hns_roce_dev *hr_dev);
 	void (*cmq_exit)(struct hns_roce_dev *hr_dev);
···
 	u32 is_vf;
 	u32 cong_algo_tmpl_id;
 	u64 dwqe_page;
+	struct hns_roce_dev_debugfs dbgfs;
+	atomic64_t *dfx_cnt;
 };
 
 static inline struct hns_roce_dev *to_hr_dev(struct ib_device *ib_dev)

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

···
 	/* Write to hardware */
 	roce_write(hr_dev, ROCEE_TX_CMQ_PI_REG, csq->head);
 
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CMDS_CNT]);
+
 	do {
 		if (hns_roce_cmq_csq_done(hr_dev))
 			break;
···
 	}
 
 	spin_unlock_bh(&csq->lock);
+
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CMDS_ERR_CNT]);
 
 	return ret;
 }
···
 /* Apply all loaded caps before setting to hardware */
 static void apply_func_caps(struct hns_roce_dev *hr_dev)
 {
+#define MAX_GID_TBL_LEN 256
 	struct hns_roce_caps *caps = &hr_dev->caps;
 	struct hns_roce_v2_priv *priv = hr_dev->priv;
 
···
 		caps->gmv_entry_sz = HNS_ROCE_V3_GMV_ENTRY_SZ;
 
 		caps->gmv_hop_num = HNS_ROCE_HOP_NUM_0;
-		caps->gid_table_len[0] = caps->gmv_bt_num *
-					(HNS_HW_PAGE_SIZE / caps->gmv_entry_sz);
+
+		/* It's meaningless to support excessively large gid_table_len,
+		 * as the type of sgid_index in kernel struct ib_global_route
+		 * and userspace struct ibv_global_route are u8/uint8_t (0-255).
+		 */
+		caps->gid_table_len[0] = min_t(u32, MAX_GID_TBL_LEN,
+					 caps->gmv_bt_num *
+					 (HNS_HW_PAGE_SIZE / caps->gmv_entry_sz));
 
 		caps->gmv_entry_num = caps->gmv_bt_num * (PAGE_SIZE /
 							  caps->gmv_entry_sz);
···
 	return 0;
 
 create_failed_qp:
+	for (i--; i >= 0; i--) {
+		hns_roce_v2_destroy_qp(&free_mr->rsv_qp[i]->ibqp, NULL);
+		kfree(free_mr->rsv_qp[i]);
+	}
 	hns_roce_destroy_cq(cq, NULL);
 	kfree(cq);
 
···
 	struct hns_roce_srq_context *srq_context;
 	struct hns_roce_srq_context *srqc_mask;
 	struct hns_roce_cmd_mailbox *mailbox;
-	int ret;
+	int ret = 0;
 
 	/* Resizing SRQs is not supported yet */
-	if (srq_attr_mask & IB_SRQ_MAX_WR)
-		return -EINVAL;
+	if (srq_attr_mask & IB_SRQ_MAX_WR) {
+		ret = -EOPNOTSUPP;
+		goto out;
+	}
 
 	if (srq_attr_mask & IB_SRQ_LIMIT) {
-		if (srq_attr->srq_limit > srq->wqe_cnt)
-			return -EINVAL;
+		if (srq_attr->srq_limit > srq->wqe_cnt) {
+			ret = -EINVAL;
+			goto out;
+		}
 
 		mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
-		if (IS_ERR(mailbox))
-			return PTR_ERR(mailbox);
+		if (IS_ERR(mailbox)) {
+			ret = PTR_ERR(mailbox);
+			goto out;
+		}
 
 		srq_context = mailbox->buf;
 		srqc_mask = (struct hns_roce_srq_context *)mailbox->buf + 1;
···
 		ret = hns_roce_cmd_mbox(hr_dev, mailbox->dma, 0,
 					HNS_ROCE_CMD_MODIFY_SRQC, srq->srqn);
 		hns_roce_free_cmd_mailbox(hr_dev, mailbox);
-		if (ret) {
+		if (ret)
 			ibdev_err(&hr_dev->ib_dev,
 				  "failed to handle cmd of modifying SRQ, ret = %d.\n",
 				  ret);
-			return ret;
-		}
 	}
 
-	return 0;
+out:
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_SRQ_MODIFY_ERR_CNT]);
+
+	return ret;
 }
 
 static int hns_roce_v2_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
···
 	int ret;
 
 	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
-	if (IS_ERR(mailbox))
-		return PTR_ERR(mailbox);
+	ret = PTR_ERR_OR_ZERO(mailbox);
+	if (ret)
+		goto err_out;
 
 	cq_context = mailbox->buf;
 	cqc_mask = (struct hns_roce_v2_cq_context *)mailbox->buf + 1;
···
 		ibdev_err(&hr_dev->ib_dev,
 			  "failed to process cmd when modifying CQ, ret = %d.\n",
 			  ret);
+
+err_out:
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CQ_MODIFY_ERR_CNT]);
 
 	return ret;
 }
···
 		++eq->cons_index;
 		aeqe_found = IRQ_HANDLED;
 
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_AEQE_CNT]);
+
 		hns_roce_v2_init_irq_work(hr_dev, eq, queue_num);
 
 		aeqe = next_aeqe_sw_v2(eq);
···
 
 		++eq->cons_index;
 		ceqe_found = IRQ_HANDLED;
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CEQE_CNT]);
 
 		ceqe = next_ceqe_sw_v2(eq);
 	}
···
 	/* irq contains: abnormal + AEQ + CEQ */
 	for (j = 0; j < other_num; j++)
 		snprintf((char *)hr_dev->irq_names[j], HNS_ROCE_INT_NAME_LEN,
-			 "hns-abn-%d", j);
+			 "hns-%s-abn-%d", pci_name(hr_dev->pci_dev), j);
 
 	for (j = other_num; j < (other_num + aeq_num); j++)
 		snprintf((char *)hr_dev->irq_names[j], HNS_ROCE_INT_NAME_LEN,
-			 "hns-aeq-%d", j - other_num);
+			 "hns-%s-aeq-%d", pci_name(hr_dev->pci_dev), j - other_num);
 
 	for (j = (other_num + aeq_num); j < irq_num; j++)
 		snprintf((char *)hr_dev->irq_names[j], HNS_ROCE_INT_NAME_LEN,
-			 "hns-ceq-%d", j - other_num - aeq_num);
+			 "hns-%s-ceq-%d", pci_name(hr_dev->pci_dev),
+			 j - other_num - aeq_num);
 
 	for (j = 0; j < irq_num; j++) {
 		if (j < other_num)
···
 
 static int __init hns_roce_hw_v2_init(void)
 {
+	hns_roce_init_debugfs();
 	return hnae3_register_client(&hns_roce_hw_v2_client);
 }
 
 static void __exit hns_roce_hw_v2_exit(void)
 {
 	hnae3_unregister_client(&hns_roce_hw_v2_client);
+	hns_roce_cleanup_debugfs();
 }
 
 module_init(hns_roce_hw_v2_init);

drivers/infiniband/hw/hns/hns_roce_main.c

···
 	struct hns_roce_dev *hr_dev = to_hr_dev(uctx->device);
 	struct hns_roce_ib_alloc_ucontext_resp resp = {};
 	struct hns_roce_ib_alloc_ucontext ucmd = {};
-	int ret;
+	int ret = -EAGAIN;
 
 	if (!hr_dev->active)
-		return -EAGAIN;
+		goto error_out;
 
 	resp.qp_tab_size = hr_dev->caps.num_qps;
 	resp.srq_tab_size = hr_dev->caps.num_srqs;
···
 	ret = ib_copy_from_udata(&ucmd, udata,
 				 min(udata->inlen, sizeof(ucmd)));
 	if (ret)
-		return ret;
+		goto error_out;
 
 	if (hr_dev->pci_dev->revision >= PCI_REVISION_ID_HIP09)
 		context->config = ucmd.config & HNS_ROCE_EXSGE_FLAGS;
···
 
 	ret = hns_roce_uar_alloc(hr_dev, &context->uar);
 	if (ret)
-		goto error_fail_uar_alloc;
+		goto error_out;
 
 	ret = hns_roce_alloc_uar_entry(uctx);
 	if (ret)
···
 error_fail_uar_entry:
 	ida_free(&hr_dev->uar_ida.ida, (int)context->uar.logic_idx);
 
-error_fail_uar_alloc:
+error_out:
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_UCTX_ALLOC_ERR_CNT]);
+
 	return ret;
 }
 
···
 
 static int hns_roce_mmap(struct ib_ucontext *uctx, struct vm_area_struct *vma)
 {
+	struct hns_roce_dev *hr_dev = to_hr_dev(uctx->device);
 	struct rdma_user_mmap_entry *rdma_entry;
 	struct hns_user_mmap_entry *entry;
 	phys_addr_t pfn;
···
 	int ret;
 
 	rdma_entry = rdma_user_mmap_entry_get_pgoff(uctx, vma->vm_pgoff);
-	if (!rdma_entry)
+	if (!rdma_entry) {
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MMAP_ERR_CNT]);
 		return -EINVAL;
+	}
 
 	entry = to_hns_mmap(rdma_entry);
 	pfn = entry->address >> PAGE_SHIFT;
···
 
 out:
 	rdma_user_mmap_entry_put(rdma_entry);
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MMAP_ERR_CNT]);
+
 	return ret;
 }
 
···
 	spin_unlock_irqrestore(&hr_dev->qp_list_lock, flags);
 }
 
+static int hns_roce_alloc_dfx_cnt(struct hns_roce_dev *hr_dev)
+{
+	hr_dev->dfx_cnt = kvcalloc(HNS_ROCE_DFX_CNT_TOTAL, sizeof(atomic64_t),
+				   GFP_KERNEL);
+	if (!hr_dev->dfx_cnt)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void hns_roce_dealloc_dfx_cnt(struct hns_roce_dev *hr_dev)
+{
+	kvfree(hr_dev->dfx_cnt);
+}
+
 int hns_roce_init(struct hns_roce_dev *hr_dev)
 {
 	struct device *dev = hr_dev->dev;
···
 
 	hr_dev->is_reset = false;
 
+	ret = hns_roce_alloc_dfx_cnt(hr_dev);
+	if (ret)
+		return ret;
+
 	if (hr_dev->hw->cmq_init) {
 		ret = hr_dev->hw->cmq_init(hr_dev);
 		if (ret) {
 			dev_err(dev, "init RoCE Command Queue failed!\n");
-			return ret;
+			goto error_failed_alloc_dfx_cnt;
 		}
 	}
 
···
 	if (ret)
 		goto error_failed_register_device;
 
+	hns_roce_register_debugfs(hr_dev);
+
 	return 0;
 
 error_failed_register_device:
···
 	if (hr_dev->hw->cmq_exit)
 		hr_dev->hw->cmq_exit(hr_dev);
 
+error_failed_alloc_dfx_cnt:
+	hns_roce_dealloc_dfx_cnt(hr_dev);
+
 	return ret;
 }
 
 void hns_roce_exit(struct hns_roce_dev *hr_dev)
 {
+	hns_roce_unregister_debugfs(hr_dev);
 	hns_roce_unregister_device(hr_dev);
 
 	if (hr_dev->hw->hw_exit)
···
 	hns_roce_cmd_cleanup(hr_dev);
 	if (hr_dev->hw->cmq_exit)
 		hr_dev->hw->cmq_exit(hr_dev);
+	hns_roce_dealloc_dfx_cnt(hr_dev);
 }
 
 MODULE_LICENSE("Dual BSD/GPL");

drivers/infiniband/hw/hns/hns_roce_mr.c

···
 	int ret;
 
 	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
-	if (!mr)
-		return ERR_PTR(-ENOMEM);
+	if (!mr) {
+		ret = -ENOMEM;
+		goto err_out;
+	}
 
 	mr->iova = virt_addr;
 	mr->size = length;
···
 	free_mr_key(hr_dev, mr);
 err_alloc_mr:
 	kfree(mr);
+err_out:
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MR_REG_ERR_CNT]);
+
 	return ERR_PTR(ret);
 }
 
···
 	unsigned long mtpt_idx;
 	int ret;
 
-	if (!mr->enabled)
-		return ERR_PTR(-EINVAL);
+	if (!mr->enabled) {
+		ret = -EINVAL;
+		goto err_out;
+	}
 
 	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
-	if (IS_ERR(mailbox))
-		return ERR_CAST(mailbox);
+	ret = PTR_ERR_OR_ZERO(mailbox);
+	if (ret)
+		goto err_out;
 
 	mtpt_idx = key_to_hw_index(mr->key) & (hr_dev->caps.num_mtpts - 1);
 
···
 free_cmd_mbox:
 	hns_roce_free_cmd_mailbox(hr_dev, mailbox);
 
-	if (ret)
+err_out:
+	if (ret) {
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_MR_REREG_ERR_CNT]);
 		return ERR_PTR(ret);
+	}
+
 	return NULL;
 }
 
···
 		mtr->kmem = NULL;
 		mtr->umem = ib_umem_get(ibdev, user_addr, total_size,
 					buf_attr->user_access);
-		if (IS_ERR_OR_NULL(mtr->umem)) {
+		if (IS_ERR(mtr->umem)) {
 			ibdev_err(ibdev, "failed to get umem, ret = %ld.\n",
 				  PTR_ERR(mtr->umem));
 			return -ENOMEM;

drivers/infiniband/hw/hns/hns_roce_pd.c

···
 	struct hns_roce_xrcd *xrcd = to_hr_xrcd(ib_xrcd);
 	int ret;
 
-	if (!(hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_XRC))
-		return -EINVAL;
+	if (!(hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_XRC)) {
+		ret = -EOPNOTSUPP;
+		goto err_out;
+	}
 
 	ret = hns_roce_xrcd_alloc(hr_dev, &xrcd->xrcdn);
-	if (ret)
-		return ret;
 
-	return 0;
+err_out:
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_XRCD_ALLOC_ERR_CNT]);
+
+	return ret;
 }
 
 int hns_roce_dealloc_xrcd(struct ib_xrcd *ib_xrcd, struct ib_udata *udata)

drivers/infiniband/hw/hns/hns_roce_qp.c

···
 
 	ret = check_qp_type(hr_dev, init_attr->qp_type, !!udata);
 	if (ret)
-		return ret;
+		goto err_out;
 
 	if (init_attr->qp_type == IB_QPT_XRC_TGT)
 		hr_qp->xrcdn = to_hr_xrcd(init_attr->xrcd)->xrcdn;
···
 	if (ret)
 		ibdev_err(ibdev, "create QP type 0x%x failed(%d)\n",
 			  init_attr->qp_type, ret);
+
+err_out:
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_QP_CREATE_ERR_CNT]);
 
 	return ret;
 }
···
 
 out:
 	mutex_unlock(&hr_qp->mutex);
+	if (ret)
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_QP_MODIFY_ERR_CNT]);
 
 	return ret;
 }

drivers/infiniband/hw/hns/hns_roce_srq.c

···
 
 	ret = set_srq_param(srq, init_attr, udata);
 	if (ret)
-		return ret;
+		goto err_out;
 
 	ret = alloc_srq_buf(hr_dev, srq, udata);
 	if (ret)
-		return ret;
+		goto err_out;
 
 	ret = alloc_srq_db(hr_dev, srq, udata, &resp);
 	if (ret)
···
 	free_srq_db(hr_dev, srq, udata);
 err_srq_buf:
 	free_srq_buf(hr_dev, srq);
+err_out:
+	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_SRQ_CREATE_ERR_CNT]);
 
 	return ret;
 }

drivers/infiniband/hw/irdma/utils.c

···
 			   u32 val)
 {
 	u32 crc = 0;
-	int ret;
-	int ret_code = 0;
 
-	crypto_shash_init(desc);
-	ret = crypto_shash_update(desc, addr, len);
-	if (!ret)
-		crypto_shash_final(desc, (u8 *)&crc);
+	crypto_shash_digest(desc, addr, len, (u8 *)&crc);
 	if (crc != val)
-		ret_code = -EINVAL;
+		return -EINVAL;
 
-	return ret_code;
+	return 0;
 }
 
 /**

drivers/infiniband/hw/mana/cq.c

···
 	struct ib_device *ibdev = ibcq->device;
 	struct mana_ib_create_cq ucmd = {};
 	struct mana_ib_dev *mdev;
+	struct gdma_context *gc;
 	int err;
 
 	mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
+	gc = mdev->gdma_dev->gdma_context;
 
 	if (udata->inlen < sizeof(ucmd))
 		return -EINVAL;
+
+	if (attr->comp_vector > gc->max_num_queues)
+		return -EINVAL;
+
+	cq->comp_vector = attr->comp_vector;
 
 	err = ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen));
 	if (err) {
···
 		return err;
 	}
 
-	if (attr->cqe > MAX_SEND_BUFFERS_PER_QUEUE) {
+	if (attr->cqe > mdev->adapter_caps.max_qp_wr) {
 		ibdev_dbg(ibdev, "CQE %d exceeding limit\n", attr->cqe);
 		return -EINVAL;
 	}
···
 	/*
 	 * The CQ ID is not known at this time. The ID is generated at create_qp
 	 */
+	cq->id = INVALID_QUEUE_ID;
 
 	return 0;
 
···
 	struct mana_ib_cq *cq = container_of(ibcq, struct mana_ib_cq, ibcq);
 	struct ib_device *ibdev = ibcq->device;
 	struct mana_ib_dev *mdev;
+	struct gdma_context *gc;
+	int err;
 
 	mdev = container_of(ibdev, struct mana_ib_dev, ib_dev);
+	gc = mdev->gdma_dev->gdma_context;
 
-	mana_ib_gd_destroy_dma_region(mdev, cq->gdma_region);
+	err = mana_ib_gd_destroy_dma_region(mdev, cq->gdma_region);
+	if (err) {
+		ibdev_dbg(ibdev,
+			  "Failed to destroy dma region, %d\n", err);
+		return err;
+	}
+
+	if (cq->id != INVALID_QUEUE_ID) {
+		kfree(gc->cq_table[cq->id]);
+		gc->cq_table[cq->id] = NULL;
+	}
+
 	ib_umem_release(cq->umem);
 
 	return 0;
+}
+
+void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq)
+{
+	struct mana_ib_cq *cq = ctx;
+
+	if (cq->ibcq.comp_handler)
+		cq->ibcq.comp_handler(&cq->ibcq, cq->ibcq.cq_context);
 }

drivers/infiniband/hw/mana/device.c

···
 	ibdev_dbg(&dev->ib_dev, "mdev=%p id=%d num_ports=%d\n", mdev,
 		  mdev->dev_id.as_uint32, dev->ib_dev.phys_port_cnt);
 
-	dev->gdma_dev = mdev;
 	dev->ib_dev.node_type = RDMA_NODE_IB_CA;
 
 	/*
···
 	dev->ib_dev.num_comp_vectors = 1;
 	dev->ib_dev.dev.parent = mdev->gdma_context->dev;
 
+	ret = mana_gd_register_device(&mdev->gdma_context->mana_ib);
+	if (ret) {
+		ibdev_err(&dev->ib_dev, "Failed to register device, ret %d",
+			  ret);
+		goto free_ib_device;
+	}
+	dev->gdma_dev = &mdev->gdma_context->mana_ib;
+
+	ret = mana_ib_gd_query_adapter_caps(dev);
+	if (ret) {
+		ibdev_err(&dev->ib_dev, "Failed to query device caps, ret %d",
+			  ret);
+		goto deregister_device;
+	}
+
 	ret = ib_register_device(&dev->ib_dev, "mana_%d",
 				 mdev->gdma_context->dev);
-	if (ret) {
-		ib_dealloc_device(&dev->ib_dev);
-		return ret;
-	}
+	if (ret)
+		goto deregister_device;
 
 	dev_set_drvdata(&adev->dev, dev);
 
 	return 0;
+
+deregister_device:
+	mana_gd_deregister_device(dev->gdma_dev);
+free_ib_device:
+	ib_dealloc_device(&dev->ib_dev);
+	return ret;
 }
 
 static void mana_ib_remove(struct auxiliary_device *adev)
···
 	struct mana_ib_dev *dev = dev_get_drvdata(&adev->dev);
 
 	ib_unregister_device(&dev->ib_dev);
+
+	mana_gd_deregister_device(dev->gdma_dev);
+
 	ib_dealloc_device(&dev->ib_dev);
 }
 

drivers/infiniband/hw/mana/main.c

···
 void mana_ib_uncfg_vport(struct mana_ib_dev *dev, struct mana_ib_pd *pd,
 			 u32 port)
 {
-	struct gdma_dev *gd = dev->gdma_dev;
+	struct gdma_dev *gd = &dev->gdma_dev->gdma_context->mana;
 	struct mana_port_context *mpc;
 	struct net_device *ndev;
 	struct mana_context *mc;
···
 int mana_ib_cfg_vport(struct mana_ib_dev *dev, u32 port, struct mana_ib_pd *pd,
 		      u32 doorbell_id)
 {
-	struct gdma_dev *mdev = dev->gdma_dev;
+	struct gdma_dev *mdev = &dev->gdma_dev->gdma_context->mana;
 	struct mana_port_context *mpc;
 	struct mana_context *mc;
 	struct net_device *ndev;
···
 int mana_ib_query_device(struct ib_device *ibdev, struct ib_device_attr *props,
 			 struct ib_udata *uhw)
 {
-	props->max_qp = MANA_MAX_NUM_QUEUES;
-	props->max_qp_wr = MAX_SEND_BUFFERS_PER_QUEUE;
+	struct mana_ib_dev *dev = container_of(ibdev,
+			struct mana_ib_dev, ib_dev);
 
-	/*
-	 * max_cqe could be potentially much bigger.
-	 * As this version of driver only support RAW QP, set it to the same
-	 * value as max_qp_wr
-	 */
-	props->max_cqe = MAX_SEND_BUFFERS_PER_QUEUE;
-
+	props->max_qp = dev->adapter_caps.max_qp_count;
+	props->max_qp_wr = dev->adapter_caps.max_qp_wr;
+	props->max_cq = dev->adapter_caps.max_cq_count;
+	props->max_cqe = dev->adapter_caps.max_qp_wr;
+	props->max_mr = dev->adapter_caps.max_mr_count;
 	props->max_mr_size = MANA_IB_MAX_MR_SIZE;
-	props->max_mr = MANA_IB_MAX_MR;
-	props->max_send_sge = MAX_TX_WQE_SGL_ENTRIES;
-	props->max_recv_sge = MAX_RX_WQE_SGL_ENTRIES;
+	props->max_send_sge = dev->adapter_caps.max_send_sge_count;
+	props->max_recv_sge = dev->adapter_caps.max_recv_sge_count;
 
 	return 0;
 }
···
 
 void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext)
 {
+}
+
+int mana_ib_gd_query_adapter_caps(struct mana_ib_dev *dev)
+{
+	struct mana_ib_adapter_caps *caps = &dev->adapter_caps;
+	struct mana_ib_query_adapter_caps_resp resp = {};
+	struct mana_ib_query_adapter_caps_req req = {};
+	int err;
+
+	mana_gd_init_req_hdr(&req.hdr, MANA_IB_GET_ADAPTER_CAP, sizeof(req),
+			     sizeof(resp));
+	req.hdr.resp.msg_version = GDMA_MESSAGE_V3;
+	req.hdr.dev_id = dev->gdma_dev->dev_id;
+
+	err = mana_gd_send_request(dev->gdma_dev->gdma_context, sizeof(req),
+				   &req, sizeof(resp), &resp);
+
+	if (err) {
+		ibdev_err(&dev->ib_dev,
+			  "Failed to query adapter caps err %d", err);
+		return err;
+	}
+
+	caps->max_sq_id = resp.max_sq_id;
+	caps->max_rq_id = resp.max_rq_id;
+	caps->max_cq_id = resp.max_cq_id;
+	caps->max_qp_count = resp.max_qp_count;
+	caps->max_cq_count = resp.max_cq_count;
+	caps->max_mr_count = resp.max_mr_count;
+	caps->max_pd_count = resp.max_pd_count;
+	caps->max_inbound_read_limit = resp.max_inbound_read_limit;
+	caps->max_outbound_read_limit = resp.max_outbound_read_limit;
+	caps->mw_count = resp.mw_count;
+	caps->max_srq_count = resp.max_srq_count;
+	caps->max_qp_wr = min_t(u32,
+				resp.max_requester_sq_size / GDMA_MAX_SQE_SIZE,
+				resp.max_requester_rq_size / GDMA_MAX_RQE_SIZE);
+	caps->max_inline_data_size = resp.max_inline_data_size;
+	caps->max_send_sge_count = resp.max_send_sge_count;
+	caps->max_recv_sge_count = resp.max_recv_sge_count;
+
+	return 0;
 }

drivers/infiniband/hw/mana/mana_ib.h

···
  */
 #define MANA_IB_MAX_MR		0xFFFFFFu
 
+struct mana_ib_adapter_caps {
+	u32 max_sq_id;
+	u32 max_rq_id;
+	u32 max_cq_id;
+	u32 max_qp_count;
+	u32 max_cq_count;
+	u32 max_mr_count;
+	u32 max_pd_count;
+	u32 max_inbound_read_limit;
+	u32 max_outbound_read_limit;
+	u32 mw_count;
+	u32 max_srq_count;
+	u32 max_qp_wr;
+	u32 max_send_sge_count;
+	u32 max_recv_sge_count;
+	u32 max_inline_data_size;
+};
+
 struct mana_ib_dev {
 	struct ib_device ib_dev;
 	struct gdma_dev *gdma_dev;
+	struct mana_ib_adapter_caps adapter_caps;
 };
 
 struct mana_ib_wq {
···
 	int cqe;
 	u64 gdma_region;
 	u64 id;
+	u32 comp_vector;
 };
 
 struct mana_ib_qp {
···
 struct mana_ib_rwq_ind_table {
 	struct ib_rwq_ind_table ib_ind_table;
 };
+
+enum mana_ib_command_code {
+	MANA_IB_GET_ADAPTER_CAP = 0x30001,
+};
+
+struct mana_ib_query_adapter_caps_req {
+	struct gdma_req_hdr hdr;
+}; /*HW Data */
+
+struct mana_ib_query_adapter_caps_resp {
+	struct gdma_resp_hdr hdr;
+	u32 max_sq_id;
+	u32 max_rq_id;
+	u32 max_cq_id;
+	u32 max_qp_count;
+	u32 max_cq_count;
+	u32 max_mr_count;
+	u32 max_pd_count;
+	u32 max_inbound_read_limit;
+	u32 max_outbound_read_limit;
+	u32 mw_count;
+	u32 max_srq_count;
+	u32 max_requester_sq_size;
+	u32 max_responder_sq_size;
+	u32 max_requester_rq_size;
+	u32 max_responder_rq_size;
+	u32 max_send_sge_count;
+	u32 max_recv_sge_count;
+	u32 max_inline_data_size;
+}; /* HW Data */
 
 int mana_ib_gd_create_dma_region(struct mana_ib_dev *dev, struct ib_umem *umem,
 				 mana_handle_t *gdma_region);
···
 
 void mana_ib_disassociate_ucontext(struct ib_ucontext *ibcontext);
 
+int mana_ib_gd_query_adapter_caps(struct mana_ib_dev *mdev);
+
+void mana_ib_cq_handler(void *ctx, struct gdma_queue *gdma_cq);
 #endif

drivers/infiniband/hw/mana/qp.c

···
 	u32 req_buf_size;
 	int i, err;
 
-	mdev = dev->gdma_dev;
-	gc = mdev->gdma_context;
+	gc = dev->gdma_dev->gdma_context;
+	mdev = &gc->mana;
 
 	req_buf_size =
 		sizeof(*req) + sizeof(mana_handle_t) * MANA_INDIRECT_TABLE_SIZE;
···
 	struct ib_rwq_ind_table *ind_tbl = attr->rwq_ind_tbl;
 	struct mana_ib_create_qp_rss_resp resp = {};
 	struct mana_ib_create_qp_rss ucmd = {};
-	struct gdma_dev *gd = mdev->gdma_dev;
+	struct gdma_queue **gdma_cq_allocated;
 	mana_handle_t *mana_ind_table;
 	struct mana_port_context *mpc;
+	struct gdma_queue *gdma_cq;
+	unsigned int ind_tbl_size;
 	struct mana_context *mc;
 	struct net_device *ndev;
+	struct gdma_context *gc;
 	struct mana_ib_cq *cq;
 	struct mana_ib_wq *wq;
-	unsigned int ind_tbl_size;
+	struct gdma_dev *gd;
+	struct mana_eq *eq;
 	struct ib_cq *ibcq;
 	struct ib_wq *ibwq;
 	int i = 0;
 	u32 port;
 	int ret;
 
+	gc = mdev->gdma_dev->gdma_context;
+	gd = &gc->mana;
 	mc = gd->driver_data;
 
 	if (!udata || udata->inlen < sizeof(ucmd))
···
 		return ret;
 	}
 
-	if (attr->cap.max_recv_wr > MAX_SEND_BUFFERS_PER_QUEUE) {
+	if (attr->cap.max_recv_wr > mdev->adapter_caps.max_qp_wr) {
 		ibdev_dbg(&mdev->ib_dev,
 			  "Requested max_recv_wr %d exceeding limit\n",
 			  attr->cap.max_recv_wr);
···
 		goto fail;
 	}
 
+	gdma_cq_allocated = kcalloc(ind_tbl_size, sizeof(*gdma_cq_allocated),
+				    GFP_KERNEL);
+	if (!gdma_cq_allocated) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+
 	qp->port = port;
 
 	for (i = 0; i < ind_tbl_size; i++) {
···
 		cq_spec.gdma_region = cq->gdma_region;
 		cq_spec.queue_size = cq->cqe * COMP_ENTRY_SIZE;
 		cq_spec.modr_ctx_id = 0;
-		cq_spec.attached_eq = GDMA_CQ_NO_EQ;
+		eq = &mc->eqs[cq->comp_vector % gc->max_num_queues];
+		cq_spec.attached_eq = eq->eq->id;
 
 		ret = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_RQ,
 					 &wq_spec, &cq_spec, &wq->rx_object);
-		if (ret)
+		if (ret) {
+			/* Do cleanup starting with index i-1 */
+			i--;
 			goto fail;
+		}
 
 		/* The GDMA regions are now owned by the WQ object */
 		wq->gdma_region = GDMA_INVALID_DMA_REGION;
···
 		resp.entries[i].wqid = wq->id;
 
 		mana_ind_table[i] = wq->rx_object;
+
+		/* Create CQ table entry */
+		WARN_ON(gc->cq_table[cq->id]);
+		gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
+		if (!gdma_cq) {
+			ret = -ENOMEM;
+			goto fail;
+		}
+		gdma_cq_allocated[i] = gdma_cq;
+
+		gdma_cq->cq.context = cq;
+		gdma_cq->type = GDMA_CQ;
+		gdma_cq->cq.callback = mana_ib_cq_handler;
+		gdma_cq->id = cq->id;
+		gc->cq_table[cq->id] = gdma_cq;
 	}
 	resp.num_entries = i;
 
···
 		goto fail;
 	}
 
+	kfree(gdma_cq_allocated);
 	kfree(mana_ind_table);
 
 	return 0;
···
 fail:
 	while (i-- > 0) {
 		ibwq = ind_tbl->ind_tbl[i];
+		ibcq = ibwq->cq;
 		wq = container_of(ibwq, struct mana_ib_wq, ibwq);
+		cq = container_of(ibcq, struct mana_ib_cq, ibcq);
+
+		gc->cq_table[cq->id] = NULL;
+		kfree(gdma_cq_allocated[i]);
+
 		mana_destroy_wq_obj(mpc, GDMA_RQ, wq->rx_object);
 	}
 
+	kfree(gdma_cq_allocated);
 	kfree(mana_ind_table);
 
 	return ret;
···
 	struct mana_ib_ucontext *mana_ucontext =
 		rdma_udata_to_drv_context(udata, struct mana_ib_ucontext,
 					  ibucontext);
+	struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
 	struct mana_ib_create_qp_resp resp = {};
-	struct gdma_dev *gd = mdev->gdma_dev;
 	struct mana_ib_create_qp ucmd = {};
+	struct gdma_queue *gdma_cq = NULL;
 	struct mana_obj_spec wq_spec = {};
 	struct mana_obj_spec cq_spec = {};
 	struct mana_port_context *mpc;
 	struct mana_context *mc;
 	struct net_device *ndev;
 	struct ib_umem *umem;
-	int err;
+	struct mana_eq *eq;
+	int eq_vec;
 	u32 port;
+	int err;
 
 	mc = gd->driver_data;
 
···
 	if (port < 1 || port > mc->num_ports)
 		return -EINVAL;
 
-	if (attr->cap.max_send_wr > MAX_SEND_BUFFERS_PER_QUEUE) {
+	if (attr->cap.max_send_wr > mdev->adapter_caps.max_qp_wr) {
 		ibdev_dbg(&mdev->ib_dev,
 			  "Requested max_send_wr %d exceeding limit\n",
 			  attr->cap.max_send_wr);
···
 	cq_spec.gdma_region = send_cq->gdma_region;
 	cq_spec.queue_size = send_cq->cqe * COMP_ENTRY_SIZE;
 	cq_spec.modr_ctx_id = 0;
-	cq_spec.attached_eq = GDMA_CQ_NO_EQ;
+	eq_vec = send_cq->comp_vector % gd->gdma_context->max_num_queues;
+	eq = &mc->eqs[eq_vec];
+	cq_spec.attached_eq = eq->eq->id;
 
 	err = mana_create_wq_obj(mpc, mpc->port_handle, GDMA_SQ, &wq_spec,
 				 &cq_spec, &qp->tx_object);
···
 	qp->sq_id = wq_spec.queue_index;
 	send_cq->id = cq_spec.queue_index;
 
+	/* Create CQ table entry */
+	WARN_ON(gd->gdma_context->cq_table[send_cq->id]);
+	gdma_cq = kzalloc(sizeof(*gdma_cq), GFP_KERNEL);
+	if (!gdma_cq) {
+		err = -ENOMEM;
+		goto err_destroy_wq_obj;
+	}
+
+	gdma_cq->cq.context = send_cq;
+	gdma_cq->type = GDMA_CQ;
+	gdma_cq->cq.callback = mana_ib_cq_handler;
+	gdma_cq->id = send_cq->id;
+	gd->gdma_context->cq_table[send_cq->id] = gdma_cq;
+
 	ibdev_dbg(&mdev->ib_dev,
 		  "ret %d qp->tx_object 0x%llx sq id %llu cq id %llu\n", err,
 		  qp->tx_object, qp->sq_id, send_cq->id);
···
 		ibdev_dbg(&mdev->ib_dev,
 			  "Failed copy udata for create qp-raw, %d\n",
 			  err);
-		goto err_destroy_wq_obj;
+		goto err_release_gdma_cq;
 	}
 
 	return 0;
+
+err_release_gdma_cq:
+	kfree(gdma_cq);
+	gd->gdma_context->cq_table[send_cq->id] = NULL;
 
 err_destroy_wq_obj:
 	mana_destroy_wq_obj(mpc, GDMA_SQ, qp->tx_object);
···
 {
 	struct mana_ib_dev *mdev =
 		container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev);
-	struct gdma_dev *gd = mdev->gdma_dev;
+	struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
 	struct mana_port_context *mpc;
 	struct mana_context *mc;
 	struct net_device *ndev;
···
 {
 	struct mana_ib_dev *mdev =
 		container_of(qp->ibqp.device, struct mana_ib_dev, ib_dev);
-	struct gdma_dev *gd = mdev->gdma_dev;
+	struct gdma_dev *gd = &mdev->gdma_dev->gdma_context->mana;
 	struct ib_pd *ibpd = qp->ibqp.pd;
 	struct mana_port_context *mpc;
 	struct mana_context *mc;

drivers/infiniband/hw/mlx5/dm.c

···
 		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;
···
 	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) ||
···
 	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);
···
 	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;

drivers/infiniband/hw/mlx5/main.c

···
 				    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)
···
 
 		resp.dci_streams_caps.max_log_num_errored =
 			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) {

drivers/infiniband/hw/mlx5/mr.c

···
 	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);
 

drivers/infiniband/hw/mthca/mthca_cmd.c

···
 
 int mthca_SYS_EN(struct mthca_dev *dev)
 {
-	u64 out;
+	u64 out = 0;
 	int ret;
 
 	ret = mthca_cmd_imm(dev, 0, &out, 0, 0, CMD_SYS_EN, CMD_TIME_CLASS_D);
···
 int mthca_MGID_HASH(struct mthca_dev *dev, struct mthca_mailbox *mailbox,
 		    u16 *hash)
 {
-	u64 imm;
+	u64 imm = 0;
 	int err;
 
 	err = mthca_cmd_imm(dev, mailbox->dma, &imm, 0, 0, CMD_MGID_HASH,

drivers/infiniband/hw/mthca/mthca_main.c

···
 			  struct mthca_init_hca_param *init_hca,
 			  u64 icm_size)
 {
-	u64 aux_pages;
+	u64 aux_pages = 0;
 	int err;
 
 	err = mthca_SET_ICM_SIZE(mdev, icm_size, &aux_pages);

drivers/infiniband/sw/siw/siw.h

···
 };
 
 struct siw_umem {
+	struct ib_umem *base_mem;
 	struct siw_page_chunk *page_chunk;
 	int num_pages;
-	bool writable;
 	u64 fp_addr; /* First page base address */
-	struct mm_struct *owning_mm;
 };
 
 struct siw_pble {
···
 	int skb_offset; /* offset in skb */
 	int skb_copied; /* processed bytes in skb */
 
+	enum siw_rx_state state;
+
 	union iwarp_hdr hdr;
 	struct mpa_trailer trailer;
-
-	enum siw_rx_state state;
+	struct shash_desc *mpa_crc_hd;
 
 	/*
 	 * For each FPDU, main RX loop runs through 3 stages:
···
 	u64 ddp_to;
 	u32 inval_stag; /* Stag to be invalidated */
 
-	struct shash_desc *mpa_crc_hd;
 	u8 rx_suspend : 1;
 	u8 pad : 2; /* # of pad bytes expected */
 	u8 rdmap_op : 4; /* opcode of current frame */
···
 struct siw_qp {
 	struct ib_qp base_qp;
 	struct siw_device *sdev;
+	int tx_cpu;
 	struct kref ref;
 	struct completion qp_free;
 	struct list_head devq;
-	int tx_cpu;
 	struct siw_qp_attrs attrs;
 
 	struct siw_cep *cep;
···
 	} term_info;
 	struct rdma_user_mmap_entry *sq_entry; /* mmap info for SQE array */
 	struct rdma_user_mmap_entry *rq_entry; /* mmap info for RQE array */
-	struct rcu_head rcu;
 };
 
 /* helper macros */
···
 
 static inline int siw_orq_empty(struct siw_qp *qp)
 {
-	return qp->orq[qp->orq_get % qp->attrs.orq_size].flags == 0 ? 1 : 0;
+	return orq_get_current(qp)->flags == 0 ? 1 : 0;
 }
 
 static inline struct siw_sqe *irq_alloc_free(struct siw_qp *qp)

drivers/infiniband/sw/siw/siw_cm.c

···
 
 static void siw_sk_assign_cm_upcalls(struct sock *sk)
 {
-	write_lock_bh(&sk->sk_callback_lock);
-	sk->sk_state_change = siw_cm_llp_state_change;
-	sk->sk_data_ready = siw_cm_llp_data_ready;
-	sk->sk_write_space = siw_cm_llp_write_space;
-	sk->sk_error_report = siw_cm_llp_error_report;
-	write_unlock_bh(&sk->sk_callback_lock);
-}
-
-static void siw_sk_save_upcalls(struct sock *sk)
-{
 	struct siw_cep *cep = sk_to_cep(sk);
 
 	write_lock_bh(&sk->sk_callback_lock);
···
 	cep->sk_data_ready = sk->sk_data_ready;
 	cep->sk_write_space = sk->sk_write_space;
 	cep->sk_error_report = sk->sk_error_report;
+
+	sk->sk_state_change = siw_cm_llp_state_change;
+	sk->sk_data_ready = siw_cm_llp_data_ready;
+	sk->sk_write_space = siw_cm_llp_write_space;
+	sk->sk_error_report = siw_cm_llp_error_report;
 	write_unlock_bh(&sk->sk_callback_lock);
 }
 
···
 	siw_cep_get(cep);
 	s->sk->sk_user_data = cep;
 
-	siw_sk_save_upcalls(s->sk);
 	siw_sk_assign_cm_upcalls(s->sk);
 }
 
···
 	return id->event_handler(id, &event);
 }
 
+static void siw_free_cm_id(struct siw_cep *cep)
+{
+	if (!cep->cm_id)
+		return;
+
+	cep->cm_id->rem_ref(cep->cm_id);
+	cep->cm_id = NULL;
+}
+
+static void siw_destroy_cep_sock(struct siw_cep *cep)
+{
+	if (cep->sock) {
+		siw_socket_disassoc(cep->sock);
+		sock_release(cep->sock);
+		cep->sock = NULL;
+	}
+}
+
 /*
  * siw_qp_cm_drop()
  *
···
 		}
 		siw_dbg_cep(cep, "immediate close, state %d\n", cep->state);
 
-		if (qp->term_info.valid)
-			siw_send_terminate(qp);
+		siw_send_terminate(qp);
 
 		if (cep->cm_id) {
 			switch (cep->state) {
···
 			default:
 				break;
 			}
-			cep->cm_id->rem_ref(cep->cm_id);
-			cep->cm_id = NULL;
+			siw_free_cm_id(cep);
 			siw_cep_put(cep);
 		}
 		cep->state = SIW_EPSTATE_CLOSED;
 
-		if (cep->sock) {
-			siw_socket_disassoc(cep->sock);
-			/*
-			 * Immediately close socket
-			 */
-			sock_release(cep->sock);
-			cep->sock = NULL;
-		}
+		siw_destroy_cep_sock(cep);
 		if (cep->qp) {
 			cep->qp = NULL;
 			siw_qp_put(qp);
···
 {
 	WARN_ON(kref_read(&cep->ref) < 1);
 	kref_put(&cep->ref, __siw_cep_dealloc);
+}
+
+static void siw_cep_set_free_and_put(struct siw_cep *cep)
+{
+	siw_cep_set_free(cep);
+	siw_cep_put(cep);
 }
 
 void siw_cep_get(struct siw_cep *cep)
···
 		/*
 		 * QP scheduled LLP close
 		 */
-		if (cep->qp && cep->qp->term_info.valid)
+		if (cep->qp)
 			siw_send_terminate(cep->qp);
 
 		if (cep->cm_id)
···
 			cep->sock = NULL;
 		}
 		if (cep->cm_id) {
-			cep->cm_id->rem_ref(cep->cm_id);
-			cep->cm_id = NULL;
+			siw_free_cm_id(cep);
 			siw_cep_put(cep);
 		}
 	}
···
 
 		cep->state = SIW_EPSTATE_CLOSED;
 
-		siw_cep_set_free(cep);
-
-		siw_cep_put(cep);
+		siw_cep_set_free_and_put(cep);
 
 	} else if (s) {
 		sock_release(s);
···
 	struct siw_cep *cep = (struct siw_cep *)id->provider_data;
 	struct siw_qp *qp;
 	struct siw_qp_attrs qp_attrs;
-	int rv, max_priv_data = MPA_MAX_PRIVDATA;
+	int rv = -EINVAL, max_priv_data = MPA_MAX_PRIVDATA;
 	bool wait_for_peer_rts = false;
 
 	siw_cep_set_inuse(cep);
···
 
 	if (cep->state != SIW_EPSTATE_RECVD_MPAREQ) {
 		siw_dbg_cep(cep, "out of state\n");
-
-		siw_cep_set_free(cep);
-		siw_cep_put(cep);
-
-		return -ECONNRESET;
+		rv = -ECONNRESET;
+		goto free_cep;
 	}
 	qp = siw_qp_id2obj(sdev, params->qpn);
 	if (!qp) {
 		WARN(1, "[QP %d] does not exist\n", params->qpn);
-		siw_cep_set_free(cep);
-		siw_cep_put(cep);
-
-		return -EINVAL;
+		goto free_cep;
 	}
 	down_write(&qp->state_lock);
-	if (qp->attrs.state > SIW_QP_STATE_RTR) {
-		rv = -EINVAL;
-		up_write(&qp->state_lock);
-		goto error;
-	}
+	if (qp->attrs.state > SIW_QP_STATE_RTR)
+		goto error_unlock;
 	siw_dbg_cep(cep, "[QP %d]\n", params->qpn);
 
 	if (try_gso && cep->mpa.hdr.params.bits & MPA_RR_FLAG_GSO_EXP) {
···
 			"[QP %u]: ord %d (max %d), ird %d (max %d)\n",
 			qp_id(qp), params->ord, sdev->attrs.max_ord,
 			params->ird, sdev->attrs.max_ird);
-		rv = -EINVAL;
-		up_write(&qp->state_lock);
-		goto error;
+		goto error_unlock;
 	}
 	if (cep->enhanced_rdma_conn_est)
 		max_priv_data -= sizeof(struct mpa_v2_data);
···
 			cep,
 			"[QP %u]: private data length: %d (max %d)\n",
 			qp_id(qp), params->private_data_len, max_priv_data);
-		rv = -EINVAL;
-		up_write(&qp->state_lock);
-		goto error;
+		goto error_unlock;
 	}
 	if (cep->enhanced_rdma_conn_est) {
 		if (params->ord > cep->ord) {
···
 			} else {
 				cep->ird = params->ird;
 				cep->ord = params->ord;
-				rv = -EINVAL;
-				up_write(&qp->state_lock);
-				goto error;
+				goto error_unlock;
 			}
 		}
 		if (params->ird < cep->ird) {
···
 				params->ird = cep->ird;
 			else {
 				rv = -ENOMEM;
-				up_write(&qp->state_lock);
-				goto error;
+				goto error_unlock;
 			}
 		}
 		if (cep->mpa.v2_ctrl.ord &
···
 				   SIW_QP_ATTR_ORD | SIW_QP_ATTR_IRD |
 				   SIW_QP_ATTR_MPA);
 	up_write(&qp->state_lock);
-
 	if (rv)
 		goto error;
 
···
 	siw_cep_set_free(cep);
 
 	return 0;
+
+error_unlock:
+	up_write(&qp->state_lock);
 error:
-	siw_socket_disassoc(cep->sock);
-	sock_release(cep->sock);
-	cep->sock = NULL;
+	siw_destroy_cep_sock(cep);
 
 	cep->state = SIW_EPSTATE_CLOSED;
 
-	if (cep->cm_id) {
-		cep->cm_id->rem_ref(id);
-		cep->cm_id = NULL;
-	}
+	siw_free_cm_id(cep);
 	if (qp->cep) {
 		siw_cep_put(cep);
 		qp->cep = NULL;
 	}
 	cep->qp = NULL;
 	siw_qp_put(qp);
-
-	siw_cep_set_free(cep);
-	siw_cep_put(cep);
-
+free_cep:
+	siw_cep_set_free_and_put(cep);
 	return rv;
 }
 
···
 	if (cep->state != SIW_EPSTATE_RECVD_MPAREQ) {
 		siw_dbg_cep(cep, "out of state\n");
 
-		siw_cep_set_free(cep);
-		siw_cep_put(cep); /* put last reference */
+		siw_cep_set_free_and_put(cep); /* put last reference */
 
 		return -ECONNRESET;
 	}
···
 		cep->mpa.hdr.params.bits |= MPA_RR_FLAG_REJECT; /* reject */
 		siw_send_mpareqrep(cep, pdata, pd_len);
 	}
-	siw_socket_disassoc(cep->sock);
-	sock_release(cep->sock);
-	cep->sock = NULL;
+	siw_destroy_cep_sock(cep);
 
 	cep->state = SIW_EPSTATE_CLOSED;
 
-	siw_cep_set_free(cep);
-	siw_cep_put(cep);
+	siw_cep_set_free_and_put(cep);
 
 	return 0;
 }
···
 	if (cep) {
 		siw_cep_set_inuse(cep);
 
-		if (cep->cm_id) {
-			cep->cm_id->rem_ref(cep->cm_id);
-			cep->cm_id = NULL;
-		}
+		siw_free_cm_id(cep);
 		cep->sock = NULL;
 		siw_socket_disassoc(s);
 		cep->state = SIW_EPSTATE_CLOSED;
 
-		siw_cep_set_free(cep);
-		siw_cep_put(cep);
+		siw_cep_set_free_and_put(cep);
 	}
 	sock_release(s);
 
···
 
 		siw_cep_set_inuse(cep);
 
-		if (cep->cm_id) {
-			cep->cm_id->rem_ref(cep->cm_id);
-			cep->cm_id = NULL;
-		}
+		siw_free_cm_id(cep);
 		if (cep->sock) {
 			siw_socket_disassoc(cep->sock);
 			sock_release(cep->sock);
 			cep->sock = NULL;
 		}
 		cep->state = SIW_EPSTATE_CLOSED;
-		siw_cep_set_free(cep);
-		siw_cep_put(cep);
+		siw_cep_set_free_and_put(cep);
 	}
 }
 

drivers/infiniband/sw/siw/siw_main.c

···
 	int num_nodes;
 } siw_cpu_info;
 
+static void siw_destroy_cpulist(int number)
+{
+	int i = 0;
+
+	while (i < number)
+		kfree(siw_cpu_info.tx_valid_cpus[i++]);
+
+	kfree(siw_cpu_info.tx_valid_cpus);
+	siw_cpu_info.tx_valid_cpus = NULL;
+}
+
 static int siw_init_cpulist(void)
 {
 	int i, num_nodes = nr_node_ids;
···
 
 out_err:
 	siw_cpu_info.num_nodes = 0;
-	while (--i >= 0)
-		kfree(siw_cpu_info.tx_valid_cpus[i]);
-	kfree(siw_cpu_info.tx_valid_cpus);
-	siw_cpu_info.tx_valid_cpus = NULL;
+	siw_destroy_cpulist(i);
 
 	return -ENOMEM;
-}
-
-static void siw_destroy_cpulist(void)
-{
-	int i = 0;
-
-	while (i < siw_cpu_info.num_nodes)
-		kfree(siw_cpu_info.tx_valid_cpus[i++]);
-
-	kfree(siw_cpu_info.tx_valid_cpus);
 }
 
 /*
···
 	pr_info("SoftIWARP attach failed. Error: %d\n", rv);
 
 	siw_cm_exit();
-	siw_destroy_cpulist();
+	siw_destroy_cpulist(siw_cpu_info.num_nodes);
 
 	return rv;
 }
···
 
 	siw_cm_exit();
 
-	siw_destroy_cpulist();
+	siw_destroy_cpulist(siw_cpu_info.num_nodes);
 
 	if (siw_crypto_shash)
 		crypto_free_shash(siw_crypto_shash);

drivers/infiniband/sw/siw/siw_mem.c

···
 
 #include <linux/gfp.h>
 #include <rdma/ib_verbs.h>
+#include <rdma/ib_umem.h>
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/sched/mm.h>
···
 #include "siw.h"
 #include "siw_mem.h"
 
+/* Stag lookup is based on its index part only (24 bits). */
+#define SIW_STAG_MAX_INDEX	0x00ffffff
+
 /*
- * Stag lookup is based on its index part only (24 bits).
  * The code avoids special Stag of zero and tries to randomize
  * STag values between 1 and SIW_STAG_MAX_INDEX.
  */
 int siw_mem_add(struct siw_device *sdev, struct siw_mem *m)
 {
-	struct xa_limit limit = XA_LIMIT(1, 0x00ffffff);
+	struct xa_limit limit = XA_LIMIT(1, SIW_STAG_MAX_INDEX);
 	u32 id, next;
 
 	get_random_bytes(&next, 4);
-	next &= 0x00ffffff;
+	next &= SIW_STAG_MAX_INDEX;
 
 	if (xa_alloc_cyclic(&sdev->mem_xa, &id, m, limit, &next,
 	    GFP_KERNEL) < 0)
···
 	return NULL;
 }
 
-static void siw_free_plist(struct siw_page_chunk *chunk, int num_pages,
-			   bool dirty)
+void siw_umem_release(struct siw_umem *umem)
 {
-	unpin_user_pages_dirty_lock(chunk->plist, num_pages, dirty);
-}
-
-void siw_umem_release(struct siw_umem *umem, bool dirty)
-{
-	struct mm_struct *mm_s = umem->owning_mm;
 	int i, num_pages = umem->num_pages;
 
-	for (i = 0; num_pages; i++) {
-		int to_free = min_t(int, PAGES_PER_CHUNK, num_pages);
+	if (umem->base_mem)
+		ib_umem_release(umem->base_mem);
 
-		siw_free_plist(&umem->page_chunk[i], to_free,
-			       umem->writable && dirty);
+	for (i = 0; num_pages > 0; i++) {
 		kfree(umem->page_chunk[i].plist);
-		num_pages -= to_free;
+		num_pages -= PAGES_PER_CHUNK;
 	}
-	atomic64_sub(umem->num_pages, &mm_s->pinned_vm);
-
-	mmdrop(mm_s);
 	kfree(umem->page_chunk);
 	kfree(umem);
 }
···
 {
 	struct siw_device *sdev = to_siw_dev(pd->device);
 	struct siw_mem *mem = kzalloc(sizeof(*mem), GFP_KERNEL);
-	struct xa_limit limit = XA_LIMIT(1, 0x00ffffff);
+	struct xa_limit limit = XA_LIMIT(1, SIW_STAG_MAX_INDEX);
 	u32 id, next;
 
 	if (!mem)
···
 	kref_init(&mem->ref);
 
 	get_random_bytes(&next, 4);
-	next &= 0x00ffffff;
+	next &= SIW_STAG_MAX_INDEX;
 
 	if (xa_alloc_cyclic(&sdev->mem_xa, &id, mem, limit, &next,
 	    GFP_KERNEL) < 0) {
···
 
 	if (!mem->is_mw && mem->mem_obj) {
 		if (mem->is_pbl == 0)
-			siw_umem_release(mem->umem, true);
+			siw_umem_release(mem->umem);
 		else
 			kfree(mem->pbl);
 	}
···
 	return pbl;
 }
 
-struct siw_umem *siw_umem_get(u64 start, u64 len, bool writable)
+struct siw_umem *siw_umem_get(struct ib_device *base_dev, u64 start,
+			      u64 len, int rights)
 {
 	struct siw_umem *umem;
-	struct mm_struct *mm_s;
+	struct ib_umem *base_mem;
+	struct sg_page_iter sg_iter;
+	struct sg_table *sgt;
 	u64 first_page_va;
-	unsigned long mlock_limit;
-	unsigned int foll_flags = FOLL_LONGTERM;
 	int num_pages, num_chunks, i, rv = 0;
-
-	if (!can_do_mlock())
-		return ERR_PTR(-EPERM);
 
 	if (!len)
 		return ERR_PTR(-EINVAL);
···
 	if (!umem)
 		return ERR_PTR(-ENOMEM);
 
-	mm_s = current->mm;
-	umem->owning_mm = mm_s;
-	umem->writable = writable;
-
-	mmgrab(mm_s);
-
-	if (writable)
-		foll_flags |= FOLL_WRITE;
-
-	mmap_read_lock(mm_s);
-
-	mlock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
-
-	if (atomic64_add_return(num_pages, &mm_s->pinned_vm) > mlock_limit) {
-		rv = -ENOMEM;
-		goto out_sem_up;
-	}
-	umem->fp_addr = first_page_va;
-
 	umem->page_chunk =
 		kcalloc(num_chunks, sizeof(struct siw_page_chunk), GFP_KERNEL);
 	if (!umem->page_chunk) {
 		rv = -ENOMEM;
-		goto out_sem_up;
+		goto err_out;
 	}
-	for (i = 0; num_pages; i++) {
+	base_mem = ib_umem_get(base_dev, start, len, rights);
+	if (IS_ERR(base_mem)) {
+		rv = PTR_ERR(base_mem);
+		siw_dbg(base_dev, "Cannot pin user memory: %d\n", rv);
+		goto err_out;
+	}
+	umem->fp_addr = first_page_va;
+	umem->base_mem = base_mem;
+
+	sgt = &base_mem->sgt_append.sgt;
+	__sg_page_iter_start(&sg_iter, sgt->sgl, sgt->orig_nents, 0);
+
+	if (!__sg_page_iter_next(&sg_iter)) {
+		rv = -EINVAL;
+		goto err_out;
+	}
+	for (i = 0; num_pages > 0; i++) {
 		int nents = min_t(int, num_pages, PAGES_PER_CHUNK);
 		struct page **plist =
 			kcalloc(nents, sizeof(struct page *), GFP_KERNEL);
 
 		if (!plist) {
 			rv = -ENOMEM;
-			goto out_sem_up;
+			goto err_out;
 		}
 		umem->page_chunk[i].plist = plist;
-		while (nents) {
-			rv = pin_user_pages(first_page_va, nents, foll_flags,
-					    plist);
-			if (rv < 0)
-				goto out_sem_up;
-
-			umem->num_pages += rv;
-			first_page_va += rv * PAGE_SIZE;
-			plist += rv;
-			nents -= rv;
-			num_pages -= rv;
+		while (nents--) {
+			*plist = sg_page_iter_page(&sg_iter);
+			umem->num_pages++;
+			num_pages--;
+			plist++;
+			if (!__sg_page_iter_next(&sg_iter))
+				break;
 		}
 	}
-out_sem_up:
-	mmap_read_unlock(mm_s);
-
-	if (rv > 0)
-		return umem;
-
-	/* Adjust accounting for pages not pinned */
-	if (num_pages)
-		atomic64_sub(num_pages, &mm_s->pinned_vm);
-
-	siw_umem_release(umem, false);
+	return umem;
+err_out:
+	siw_umem_release(umem);
 
 	return ERR_PTR(rv);
 }

drivers/infiniband/sw/siw/siw_mem.h

···
 #ifndef _SIW_MEM_H
 #define _SIW_MEM_H
 
-struct siw_umem *siw_umem_get(u64 start, u64 len, bool writable);
-void siw_umem_release(struct siw_umem *umem, bool dirty);
+struct siw_umem *siw_umem_get(struct ib_device *base_dave, u64 start,
+			      u64 len, int rights);
+void siw_umem_release(struct siw_umem *umem);
 struct siw_pbl *siw_pbl_alloc(u32 num_buf);
 dma_addr_t siw_pbl_get_buffer(struct siw_pbl *pbl, u64 off, int *len, int *idx);
 struct siw_mem *siw_mem_id2obj(struct siw_device *sdev, int stag_index);

drivers/infiniband/sw/siw/siw_qp.c

···
 /*
  * siw_sq_flush()
  *
- * Flush SQ and ORRQ entries to CQ.
+ * Flush SQ and ORQ entries to CQ.
  *
  * Must be called with QP state write lock held.
  * Therefore, SQ and ORQ lock must not be taken.

drivers/infiniband/sw/siw/siw_qp_rx.c

···
 	return wqe;
 }
 
+static int siw_rx_data(struct siw_mem *mem_p, struct siw_rx_stream *srx,
+		       unsigned int *pbl_idx, u64 addr, int bytes)
+{
+	int rv;
+
+	if (mem_p->mem_obj == NULL)
+		rv = siw_rx_kva(srx, ib_virt_dma_to_ptr(addr), bytes);
+	else if (!mem_p->is_pbl)
+		rv = siw_rx_umem(srx, mem_p->umem, addr, bytes);
+	else
+		rv = siw_rx_pbl(srx, pbl_idx, mem_p, addr, bytes);
+	return rv;
+}
+
 /*
  * siw_proc_send:
  *
···
 			break;
 		}
 		mem_p = *mem;
-		if (mem_p->mem_obj == NULL)
-			rv = siw_rx_kva(srx,
-				ib_virt_dma_to_ptr(sge->laddr + frx->sge_off),
-				sge_bytes);
-		else if (!mem_p->is_pbl)
-			rv = siw_rx_umem(srx, mem_p->umem,
-					 sge->laddr + frx->sge_off, sge_bytes);
-		else
-			rv = siw_rx_pbl(srx, &frx->pbl_idx, mem_p,
-					sge->laddr + frx->sge_off, sge_bytes);
-
+		rv = siw_rx_data(mem_p, srx, &frx->pbl_idx,
+				 sge->laddr + frx->sge_off, sge_bytes);
 		if (unlikely(rv != sge_bytes)) {
 			wqe->processed += rcvd_bytes;
 
···
 		return -EINVAL;
 	}
 
-	if (mem->mem_obj == NULL)
-		rv = siw_rx_kva(srx,
-			(void *)(uintptr_t)(srx->ddp_to + srx->fpdu_part_rcvd),
-			bytes);
-	else if (!mem->is_pbl)
-		rv = siw_rx_umem(srx, mem->umem,
-				 srx->ddp_to + srx->fpdu_part_rcvd, bytes);
-	else
-		rv = siw_rx_pbl(srx, &frx->pbl_idx, mem,
-				srx->ddp_to + srx->fpdu_part_rcvd, bytes);
-
+	rv = siw_rx_data(mem, srx, &frx->pbl_idx,
+			 srx->ddp_to + srx->fpdu_part_rcvd, bytes);
 	if (unlikely(rv != bytes)) {
 		siw_init_terminate(qp, TERM_ERROR_LAYER_DDP,
 				   DDP_ETYPE_CATASTROPHIC,
···
 	mem_p = *mem;
 
 	bytes = min(srx->fpdu_part_rem, srx->skb_new);
-
-	if (mem_p->mem_obj == NULL)
-		rv = siw_rx_kva(srx,
-			ib_virt_dma_to_ptr(sge->laddr + wqe->processed),
-			bytes);
-	else if (!mem_p->is_pbl)
-		rv = siw_rx_umem(srx, mem_p->umem, sge->laddr + wqe->processed,
-				 bytes);
-	else
-		rv = siw_rx_pbl(srx, &frx->pbl_idx, mem_p,
-				sge->laddr + wqe->processed, bytes);
+	rv = siw_rx_data(mem_p, srx, &frx->pbl_idx,
+			 sge->laddr + wqe->processed, bytes);
 	if (rv != bytes) {
 		wqe->wc_status = SIW_WC_GENERAL_ERR;
 		rv = -EINVAL;
···
 	siw_init_terminate(qp, TERM_ERROR_LAYER_DDP, DDP_ETYPE_CATASTROPHIC,
 			   DDP_ECODE_CATASTROPHIC, 0);
 	return rv;
+}
+
+static void siw_update_skb_rcvd(struct siw_rx_stream *srx, u16 length)
+{
+	srx->skb_offset += length;
+	srx->skb_new -= length;
+	srx->skb_copied += length;
 }
 
 int siw_proc_terminate(struct siw_qp *qp)
···
 		goto out;
 
 	infop += to_copy;
-	srx->skb_offset += to_copy;
-	srx->skb_new -= to_copy;
-	srx->skb_copied += to_copy;
+	siw_update_skb_rcvd(srx, to_copy);
 	srx->fpdu_part_rcvd += to_copy;
 	srx->fpdu_part_rem -= to_copy;
 
···
 			   term->flag_m ? "valid" : "invalid");
 	}
 out:
-	srx->skb_new -= to_copy;
-	srx->skb_offset += to_copy;
-	srx->skb_copied += to_copy;
+	siw_update_skb_rcvd(srx, to_copy);
 	srx->fpdu_part_rcvd += to_copy;
 	srx->fpdu_part_rem -= to_copy;
 
···
 
 	skb_copy_bits(skb, srx->skb_offset, tbuf, avail);
 
-	srx->skb_new -= avail;
-	srx->skb_offset += avail;
-	srx->skb_copied += avail;
+	siw_update_skb_rcvd(srx, avail);
 	srx->fpdu_part_rem -= avail;
 
 	if (srx->fpdu_part_rem)
···
 		skb_copy_bits(skb, srx->skb_offset,
 			      (char *)c_hdr + srx->fpdu_part_rcvd, bytes);
 
+		siw_update_skb_rcvd(srx, bytes);
 		srx->fpdu_part_rcvd += bytes;
-
-		srx->skb_new -= bytes;
-		srx->skb_offset += bytes;
-		srx->skb_copied += bytes;
-
 		if (srx->fpdu_part_rcvd < MIN_DDP_HDR)
 			return -EAGAIN;
 
···
 		skb_copy_bits(skb, srx->skb_offset,
 			      (char *)c_hdr + srx->fpdu_part_rcvd, bytes);
 
+		siw_update_skb_rcvd(srx, bytes);
 		srx->fpdu_part_rcvd += bytes;
-
-		srx->skb_new -= bytes;
-		srx->skb_offset += bytes;
-		srx->skb_copied += bytes;
-
 		if (srx->fpdu_part_rcvd < hdrlen)
 			return -EAGAIN;
 	}

drivers/infiniband/sw/siw/siw_qp_tx.c

···
 	return NULL;
 }
 
+static struct page *siw_get_page(struct siw_mem *mem, struct siw_sge *sge,
+				 unsigned long offset, int *pbl_idx)
+{
+	if (!mem->is_pbl)
+		return siw_get_upage(mem->umem, sge->laddr + offset);
+	else
+		return siw_get_pblpage(mem, sge->laddr + offset, pbl_idx);
+}
+
 /*
  * Copy short payload at provided destination payload address
  */
···
 			char *buffer;
 			int pbl_idx = 0;
 
-			if (!mem->is_pbl)
-				p = siw_get_upage(mem->umem, sge->laddr);
-			else
-				p = siw_get_pblpage(mem, sge->laddr, &pbl_idx);
-
+			p = siw_get_page(mem, sge, 0, &pbl_idx);
 			if (unlikely(!p))
 				return -EFAULT;
 
···
 				memcpy(paddr, buffer + off, part);
 				kunmap_local(buffer);
 
-				if (!mem->is_pbl)
-					p = siw_get_upage(mem->umem,
-							  sge->laddr + part);
-				else
-					p = siw_get_pblpage(mem,
-							    sge->laddr + part,
-							    &pbl_idx);
+				p = siw_get_page(mem, sge, part, &pbl_idx);
 				if (unlikely(!p))
 					return -EFAULT;
 
···
 		/*
 		 * Do complete CRC if enabled and short packet
 		 */
-		if (c_tx->mpa_crc_hd) {
-			crypto_shash_init(c_tx->mpa_crc_hd);
-			if (crypto_shash_update(c_tx->mpa_crc_hd,
-						(u8 *)&c_tx->pkt,
-						c_tx->ctrl_len))
-				return -EINVAL;
-			crypto_shash_final(c_tx->mpa_crc_hd, (u8 *)crc);
-		}
+		if (c_tx->mpa_crc_hd &&
+		    crypto_shash_digest(c_tx->mpa_crc_hd, (u8 *)&c_tx->pkt,
+					c_tx->ctrl_len, (u8 *)crc) != 0)
+			return -EINVAL;
 		c_tx->ctrl_len += MPA_CRC_SIZE;
 
 		return PKT_COMPLETE;
···
 				    (char *)&c_tx->pkt.ctrl + c_tx->ctrl_sent,
 			    .iov_len = c_tx->ctrl_len - c_tx->ctrl_sent };
 
-	int rv = kernel_sendmsg(s, &msg, &iov, 1,
-				c_tx->ctrl_len - c_tx->ctrl_sent);
+	int rv = kernel_sendmsg(s, &msg, &iov, 1, iov.iov_len);
 
 	if (rv >= 0) {
 		c_tx->ctrl_sent += rv;
···
 			if (!is_kva) {
 				struct page *p;
 
-				if (mem->is_pbl)
-					p = siw_get_pblpage(
-						mem, sge->laddr + sge_off,
-						&pbl_idx);
-				else
-					p = siw_get_upage(mem->umem,
-							  sge->laddr + sge_off);
+				p = siw_get_page(mem, sge, sge_off, &pbl_idx);
 				if (unlikely(!p)) {
 					siw_unmap_pages(iov, kmap_mask, seg);
 					wqe->processed -= c_tx->bytes_unsent;
···
  * MPA FPDUs, each containing a DDP segment.
  *
  * SQ processing may occur in user context as a result of posting
- * new WQE's or from siw_sq_work_handler() context. Processing in
+ * new WQE's or from siw_tx_thread context. Processing in
  * user context is limited to non-kernel verbs users.
  *
  * SQ processing may get paused anytime, possibly in the middle of a WR
  * or FPDU, if insufficient send space is available. SQ processing
- * gets resumed from siw_sq_work_handler(), if send space becomes
- * available again.
+ * gets resumed from siw_tx_thread, if send space becomes available again.
  *
  * Must be called with the QP state read-locked.
  *

drivers/infiniband/sw/siw/siw_verbs.c

···
 #include "siw_verbs.h"
 #include "siw_mem.h"
 
+static int siw_qp_state_to_ib_qp_state[SIW_QP_STATE_COUNT] = {
+	[SIW_QP_STATE_IDLE] = IB_QPS_INIT,
+	[SIW_QP_STATE_RTR] = IB_QPS_RTR,
+	[SIW_QP_STATE_RTS] = IB_QPS_RTS,
+	[SIW_QP_STATE_CLOSING] = IB_QPS_SQD,
+	[SIW_QP_STATE_TERMINATE] = IB_QPS_SQE,
+	[SIW_QP_STATE_ERROR] = IB_QPS_ERR
+};
+
 static int ib_qp_state_to_siw_qp_state[IB_QPS_ERR + 1] = {
 	[IB_QPS_RESET] = SIW_QP_STATE_IDLE,
 	[IB_QPS_INIT] = SIW_QP_STATE_IDLE,
···
 	entry = to_siw_mmap_entry(rdma_entry);
 
 	rv = remap_vmalloc_range(vma, entry->address, 0);
-	if (rv) {
+	if (rv)
 		pr_warn("remap_vmalloc_range failed: %lu, %zu\n", vma->vm_pgoff,
 			size);
-		goto out;
-	}
-out:
 	rdma_user_mmap_entry_put(rdma_entry);
 
 	return rv;
···
 		goto err_atomic;
 	}
 	/*
-	 * NOTE: we allow for zero element SQ and RQ WQE's SGL's
-	 * but not for a QP unable to hold any WQE (SQ + RQ)
+	 * NOTE: we don't allow for a QP unable to hold any SQ WQE
 	 */
-	if (attrs->cap.max_send_wr + attrs->cap.max_recv_wr == 0) {
-		siw_dbg(base_dev, "QP must have send or receive queue\n");
+	if (attrs->cap.max_send_wr == 0) {
+		siw_dbg(base_dev, "QP must have send queue\n");
 		rv = -EINVAL;
 		goto err_atomic;
 	}
···
 	if (rv)
 		goto err_atomic;
 
-	num_sqe = attrs->cap.max_send_wr;
-	num_rqe = attrs->cap.max_recv_wr;
 
 	/* All queue indices are derived from modulo operations
 	 * on a free running 'get' (consumer) and 'put' (producer)
 	 * unsigned counter. Having queue sizes at power of two
 	 * avoids handling counter wrap around.
 	 */
-	if (num_sqe)
-		num_sqe = roundup_pow_of_two(num_sqe);
-	else {
-		/* Zero sized SQ is not supported */
-		rv = -EINVAL;
-		goto err_out_xa;
-	}
+	num_sqe = roundup_pow_of_two(attrs->cap.max_send_wr);
+	num_rqe = attrs->cap.max_recv_wr;
 	if (num_rqe)
 		num_rqe = roundup_pow_of_two(num_rqe);
 
···
 	} else {
 		return -EINVAL;
 	}
+	qp_attr->qp_state = siw_qp_state_to_ib_qp_state[qp->attrs.state];
 	qp_attr->cap.max_inline_data = SIW_MAX_INLINE;
 	qp_attr->cap.max_send_wr = qp->attrs.sq_size;
 	qp_attr->cap.max_send_sge = qp->attrs.sq_max_sges;
···
 	struct siw_umem *umem = NULL;
 	struct siw_ureq_reg_mr ureq;
 	struct siw_device *sdev = to_siw_dev(pd->device);
-
-	unsigned long mem_limit = rlimit(RLIMIT_MEMLOCK);
 	int rv;
 
 	siw_dbg_pd(pd, "start: 0x%pK, va: 0x%pK, len: %llu\n",
···
 		rv = -EINVAL;
 		goto err_out;
 	}
-	if (mem_limit != RLIM_INFINITY) {
-		unsigned long num_pages =
-			(PAGE_ALIGN(len + (start & ~PAGE_MASK))) >> PAGE_SHIFT;
-		mem_limit >>= PAGE_SHIFT;
-
-		if (num_pages > mem_limit - current->mm->locked_vm) {
-			siw_dbg_pd(pd, "pages req %lu, max %lu, lock %lu\n",
-				   num_pages, mem_limit,
-				   current->mm->locked_vm);
-			rv = -ENOMEM;
-			goto err_out;
-		}
-	}
-	umem = siw_umem_get(start, len, ib_access_writable(rights));
+	umem = siw_umem_get(pd->device, start, len, rights);
 	if (IS_ERR(umem)) {
 		rv = PTR_ERR(umem);
 		siw_dbg_pd(pd, "getting user memory failed: %d\n", rv);
···
 		kfree_rcu(mr, rcu);
 	} else {
 		if (umem)
-			siw_umem_release(umem, false);
+			siw_umem_release(umem);
 	}
 	return ERR_PTR(rv);
 }

drivers/infiniband/ulp/ipoib/ipoib.h

···
 	struct workqueue_struct *wq;
 	struct delayed_work mcast_task;
 	struct work_struct carrier_on_task;
+	struct work_struct reschedule_napi_work;
 	struct work_struct flush_light;
 	struct work_struct flush_normal;
 	struct work_struct flush_heavy;
 	struct work_struct restart_task;
+	struct work_struct tx_timeout_work;
 	struct delayed_work ah_reap_task;
 	struct delayed_work neigh_reap_task;
 	struct ib_device *ca;
···
 	       struct ib_ah *address, u32 dqpn);
 void ipoib_reap_ah(struct work_struct *work);
 
+void ipoib_napi_schedule_work(struct work_struct *work);
 struct ipoib_path *__path_find(struct net_device *dev, void *gid);
 void ipoib_mark_paths_invalid(struct net_device *dev);
 void ipoib_flush_paths(struct net_device *dev);
···
 void ipoib_ib_dev_flush_light(struct work_struct *work);
 void ipoib_ib_dev_flush_normal(struct work_struct *work);
 void ipoib_ib_dev_flush_heavy(struct work_struct *work);
+void ipoib_ib_tx_timeout_work(struct work_struct *work);
 void ipoib_pkey_event(struct work_struct *work);
 void ipoib_ib_dev_cleanup(struct net_device *dev);
 

drivers/infiniband/ulp/ipoib/ipoib_ib.c

···
 	napi_schedule(&priv->recv_napi);
 }
 
+/* The function will force napi_schedule */
+void ipoib_napi_schedule_work(struct work_struct *work)
+{
+	struct ipoib_dev_priv *priv =
+		container_of(work, struct ipoib_dev_priv, reschedule_napi_work);
+	bool ret;
+
+	do {
+		ret = napi_schedule(&priv->send_napi);
+		if (!ret)
+			msleep(3);
+	} while (!ret && netif_queue_stopped(priv->dev) &&
+		 test_bit(IPOIB_FLAG_INITIALIZED, &priv->flags));
+}
+
 void ipoib_ib_tx_completion(struct ib_cq *cq, void *ctx_ptr)
 {
 	struct ipoib_dev_priv *priv = ctx_ptr;
+	bool ret;
 
-	napi_schedule(&priv->send_napi);
+	ret = napi_schedule(&priv->send_napi);
+	/*
+	 * if the queue is closed the driver must be able to schedule napi,
+	 * otherwise we can end with closed queue forever, because no new
+	 * packets to send and napi callback might not get new event after
+	 * its re-arm of the napi.
+	 */
+	if (!ret && netif_queue_stopped(priv->dev))
+		schedule_work(&priv->reschedule_napi_work);
 }
 
 static inline int post_send(struct ipoib_dev_priv *priv,

drivers/infiniband/ulp/ipoib/ipoib_main.c

···
 		   netif_queue_stopped(dev), priv->tx_head, priv->tx_tail,
 		   priv->global_tx_head, priv->global_tx_tail);
 
-	/* XXX reset QP, etc. */
+
+	schedule_work(&priv->tx_timeout_work);
+}
+
+void ipoib_ib_tx_timeout_work(struct work_struct *work)
+{
+	struct ipoib_dev_priv *priv = container_of(work,
+						   struct ipoib_dev_priv,
+						   tx_timeout_work);
+	int err;
+
+	rtnl_lock();
+
+	if (!test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags))
+		goto unlock;
+
+	ipoib_stop(priv->dev);
+	err = ipoib_open(priv->dev);
+	if (err) {
+		ipoib_warn(priv, "ipoib_open failed recovering from a tx_timeout, err(%d).\n",
+				err);
+		goto unlock;
+	}
+
+	netif_tx_wake_all_queues(priv->dev);
+unlock:
+	rtnl_unlock();
+
 }
 
 static int ipoib_hard_header(struct sk_buff *skb,
···
 
 	ipoib_set_ethtool_ops(dev);
 
-	dev->watchdog_timeo	 = HZ;
+	dev->watchdog_timeo	 = 10 * HZ;
 
 	dev->flags		|= IFF_BROADCAST | IFF_MULTICAST;
 
···
 
 	INIT_DELAYED_WORK(&priv->mcast_task,   ipoib_mcast_join_task);
 	INIT_WORK(&priv->carrier_on_task, ipoib_mcast_carrier_on_task);
+	INIT_WORK(&priv->reschedule_napi_work, ipoib_napi_schedule_work);
 	INIT_WORK(&priv->flush_light,   ipoib_ib_dev_flush_light);
 	INIT_WORK(&priv->flush_normal,   ipoib_ib_dev_flush_normal);
 	INIT_WORK(&priv->flush_heavy,   ipoib_ib_dev_flush_heavy);
 	INIT_WORK(&priv->restart_task, ipoib_mcast_restart_task);
+	INIT_WORK(&priv->tx_timeout_work, ipoib_ib_tx_timeout_work);
 	INIT_DELAYED_WORK(&priv->ah_reap_task, ipoib_reap_ah);
 	INIT_DELAYED_WORK(&priv->neigh_reap_task, ipoib_reap_neigh);
 }

drivers/infiniband/ulp/ipoib/ipoib_multicast.c

···
 		if (test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags))
 			rec.join_state = SENDONLY_FULLMEMBER_JOIN;
 	}
-	spin_unlock_irq(&priv->lock);
 
 	multicast = ib_sa_join_multicast(&ipoib_sa_client, priv->ca, priv->port,
-					 &rec, comp_mask, GFP_KERNEL,
+					 &rec, comp_mask, GFP_ATOMIC,
 					 ipoib_mcast_join_complete, mcast);
-	spin_lock_irq(&priv->lock);
 	if (IS_ERR(multicast)) {
 		ret = PTR_ERR(multicast);
 		ipoib_warn(priv, "ib_sa_join_multicast failed, status %d\n", ret);
 		/* Requeue this join task with a backoff delay */
 		__ipoib_mcast_schedule_join_thread(priv, mcast, 1);
 		clear_bit(IPOIB_MCAST_FLAG_BUSY, &mcast->flags);
-		spin_unlock_irq(&priv->lock);
 		complete(&mcast->done);
-		spin_lock_irq(&priv->lock);
+		return ret;
 	}
 	return 0;
 }

drivers/infiniband/ulp/iser/iscsi_iser.h

···
  *
  * @sg:           pointer to the sg list
  * @size:         num entries of this sg
- * @data_len:     total beffer byte len
+ * @data_len:     total buffer byte len
  * @dma_nents:    returned by dma_map_sg
  */
 struct iser_data_buf {
···
  *
  * @ib_device:     RDMA device
  * @pd:            Protection Domain for this device
- * @mr:            Global DMA memory region
  * @event_handler: IB events handle routine
  * @ig_list:	   entry in devices list
  * @refcount:      Reference counter, dominated by open iser connections
···
  *
  * @mr:         memory region
  * @sig_mr:     signature memory region
- * @mr_valid:   is mr valid indicator
  */
 struct iser_reg_resources {
 	struct ib_mr                     *mr;
 	struct ib_mr                     *sig_mr;
-	u8				  mr_valid:1;
 };
 
 /**
···
  *                    to max number of post recvs
  * @max_cmds:         maximum cmds allowed for this connection
  * @name:             connection peer portal
- * @release_work:     deffered work for release job
+ * @release_work:     deferred work for release job
  * @state_mutex:      protects iser onnection state
  * @stop_completion:  conn_stop completion
  * @ib_completion:    RDMA cleanup completion

drivers/infiniband/ulp/iser/iser_initiator.c

···
 		return -EINVAL;
 	}
 
-	desc->rsc.mr_valid = 0;
+	if (desc->sig_protected)
+		desc->rsc.sig_mr->need_inval = false;
+	else
+		desc->rsc.mr->need_inval = false;
 
 	return 0;
 }

drivers/infiniband/ulp/iser/iser_memory.c

···
 
 	iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask);
 
-	if (rsc->mr_valid)
+	if (rsc->sig_mr->need_inval)
 		iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr);
 
 	ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey));
···
 	wr->access = IB_ACCESS_LOCAL_WRITE |
 		     IB_ACCESS_REMOTE_READ |
 		     IB_ACCESS_REMOTE_WRITE;
-	rsc->mr_valid = 1;
+	rsc->sig_mr->need_inval = true;
 
 	sig_reg->sge.lkey = mr->lkey;
 	sig_reg->rkey = mr->rkey;
···
 	struct ib_reg_wr *wr = &tx_desc->reg_wr;
 	int n;
 
-	if (rsc->mr_valid)
+	if (rsc->mr->need_inval)
 		iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr);
 
 	ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey));
···
 		     IB_ACCESS_REMOTE_WRITE |
 		     IB_ACCESS_REMOTE_READ;
 
-	rsc->mr_valid = 1;
+	rsc->mr->need_inval = true;
 
 	reg->sge.lkey = mr->lkey;
 	reg->rkey = mr->rkey;

drivers/infiniband/ulp/iser/iser_verbs.c

···
 			goto err_alloc_mr_integrity;
 		}
 	}
-	desc->rsc.mr_valid = 0;
 
 	return desc;
 

drivers/infiniband/ulp/rtrs/rtrs-clt.c

···
 				      clt_path->max_pages_per_mr);
 		if (IS_ERR(req->mr)) {
 			err = PTR_ERR(req->mr);
+			pr_err("Failed to alloc clt_path->max_pages_per_mr %d: %pe\n",
+			       clt_path->max_pages_per_mr, req->mr);
 			req->mr = NULL;
-			pr_err("Failed to alloc clt_path->max_pages_per_mr %d\n",
-			       clt_path->max_pages_per_mr);
 			goto out;
 		}
 
···
 		/*
 		 * Device removal is a special case.  Queue close and return 0.
 		 */
+		rtrs_wrn_rl(s, "CM event: %s, status: %d\n", rdma_event_msg(ev->event),
+			    ev->status);
 		rtrs_clt_close_conns(clt_path, false);
 		return 0;
 	default:
···
 			       clt_path->s.dst_addr.ss_family == AF_IB ?
 			       RDMA_PS_IB : RDMA_PS_TCP, IB_QPT_RC);
 	if (IS_ERR(cm_id)) {
-		err = PTR_ERR(cm_id);
-		rtrs_err(s, "Failed to create CM ID, err: %d\n", err);
-
-		return err;
+		rtrs_err(s, "Failed to create CM ID, err: %pe\n", cm_id);
+		return PTR_ERR(cm_id);
 	}
 	con->c.cm_id = cm_id;
 	con->cm_err = 0;

drivers/infiniband/ulp/rtrs/rtrs.c

···
 		cq = ib_cq_pool_get(cm_id->device, nr_cqe, cq_vector, poll_ctx);
 
 	if (IS_ERR(cq)) {
-		rtrs_err(con->path, "Creating completion queue failed, errno: %ld\n",
-			  PTR_ERR(cq));
+		rtrs_err(con->path, "Creating completion queue failed, errno: %pe\n",
+			  cq);
 		return PTR_ERR(cq);
 	}
 	con->cq = cq;

drivers/net/ethernet/mellanox/mlx5/core/eswitch.h

···
 	return esw && MLX5_ESWITCH_MANAGER(esw->dev);
 }
 
-/* The returned number is valid only when the dev is eswitch manager. */
-static inline u16 mlx5_eswitch_manager_vport(struct mlx5_core_dev *dev)
-{
-	return mlx5_core_is_ecpf_esw_manager(dev) ?
-		MLX5_VPORT_ECPF : MLX5_VPORT_PF;
-}
-
 static inline bool
 mlx5_esw_is_manager_vport(const struct mlx5_eswitch *esw, u16 vport_num)
 {

drivers/net/ethernet/mellanox/mlx5/core/lib/dm.c

···
 	unsigned long *steering_sw_icm_alloc_blocks;
 	unsigned long *header_modify_sw_icm_alloc_blocks;
 	unsigned long *header_modify_pattern_sw_icm_alloc_blocks;
+	unsigned long *header_encap_sw_icm_alloc_blocks;
 };
 
 struct mlx5_dm *mlx5_dm_create(struct mlx5_core_dev *dev)
 {
 	u64 header_modify_pattern_icm_blocks = 0;
+	u64 header_sw_encap_icm_blocks = 0;
 	u64 header_modify_icm_blocks = 0;
 	u64 steering_icm_blocks = 0;
 	struct mlx5_dm *dm;
···
 			goto err_modify_hdr;
 	}
 
+	if (MLX5_CAP_DEV_MEM(dev, log_indirect_encap_sw_icm_size)) {
+		header_sw_encap_icm_blocks =
+			BIT(MLX5_CAP_DEV_MEM(dev, log_indirect_encap_sw_icm_size) -
+			    MLX5_LOG_SW_ICM_BLOCK_SIZE(dev));
+
+		dm->header_encap_sw_icm_alloc_blocks =
+			bitmap_zalloc(header_sw_encap_icm_blocks, GFP_KERNEL);
+		if (!dm->header_encap_sw_icm_alloc_blocks)
+			goto err_pattern;
+	}
+
 	support_v2 = MLX5_CAP_FLOWTABLE_NIC_RX(dev, sw_owner_v2) &&
 		     MLX5_CAP_FLOWTABLE_NIC_TX(dev, sw_owner_v2) &&
 		     MLX5_CAP64_DEV_MEM(dev, header_modify_pattern_sw_icm_start_address);
···
 		dm->header_modify_pattern_sw_icm_alloc_blocks =
 			bitmap_zalloc(header_modify_pattern_icm_blocks, GFP_KERNEL);
 		if (!dm->header_modify_pattern_sw_icm_alloc_blocks)
-			goto err_pattern;
+			goto err_sw_encap;
 	}
 
 	return dm;
+
+err_sw_encap:
+	bitmap_free(dm->header_encap_sw_icm_alloc_blocks);
 
 err_pattern:
 	bitmap_free(dm->header_modify_sw_icm_alloc_blocks);
···
 							   log_header_modify_sw_icm_size) -
 				      MLX5_LOG_SW_ICM_BLOCK_SIZE(dev))));
 		bitmap_free(dm->header_modify_sw_icm_alloc_blocks);
+	}
+
+	if (dm->header_encap_sw_icm_alloc_blocks) {
+		WARN_ON(!bitmap_empty(dm->header_encap_sw_icm_alloc_blocks,
+				      BIT(MLX5_CAP_DEV_MEM(dev,
+							   log_indirect_encap_sw_icm_size) -
+				      MLX5_LOG_SW_ICM_BLOCK_SIZE(dev))));
+		bitmap_free(dm->header_encap_sw_icm_alloc_blocks);
 	}
 
 	if (dm->header_modify_pattern_sw_icm_alloc_blocks) {
···
 		log_icm_size = MLX5_CAP_DEV_MEM(dev,
 						log_header_modify_pattern_sw_icm_size);
 		block_map = dm->header_modify_pattern_sw_icm_alloc_blocks;
+		break;
+	case MLX5_SW_ICM_TYPE_SW_ENCAP:
+		icm_start_addr = MLX5_CAP64_DEV_MEM(dev,
+						    indirect_encap_sw_icm_start_address);
+		log_icm_size = MLX5_CAP_DEV_MEM(dev,
+						log_indirect_encap_sw_icm_size);
+		block_map = dm->header_encap_sw_icm_alloc_blocks;
 		break;
 	default:
 		return -EINVAL;
···
 		icm_start_addr = MLX5_CAP64_DEV_MEM(dev,
 						    header_modify_pattern_sw_icm_start_address);
 		block_map = dm->header_modify_pattern_sw_icm_alloc_blocks;
+		break;
+	case MLX5_SW_ICM_TYPE_SW_ENCAP:
+		icm_start_addr = MLX5_CAP64_DEV_MEM(dev,
+						    indirect_encap_sw_icm_start_address);
+		block_map = dm->header_encap_sw_icm_alloc_blocks;
 		break;
 	default:
 		return -EINVAL;

drivers/net/ethernet/microsoft/mana/gdma_main.c

···
 		if (dev_type == GDMA_DEVICE_MANA) {
 			gc->mana.gdma_context = gc;
 			gc->mana.dev_id = dev;
+		} else if (dev_type == GDMA_DEVICE_MANA_IB) {
+			gc->mana_ib.dev_id = dev;
+			gc->mana_ib.gdma_context = gc;
 		}
 	}
 
···
 
 	return 0;
 }
+EXPORT_SYMBOL_NS(mana_gd_register_device, NET_MANA);
 
 int mana_gd_deregister_device(struct gdma_dev *gd)
 {
···
 
 	return err;
 }
+EXPORT_SYMBOL_NS(mana_gd_deregister_device, NET_MANA);
 
 u32 mana_gd_wq_avail_space(struct gdma_queue *wq)
 {

include/linux/mlx5/driver.h

···
 	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

include/linux/mlx5/eswitch.h

···
 #define _MLX5_ESWITCH_
 
 #include <linux/mlx5/driver.h>
+#include <linux/mlx5/vport.h>
 #include <net/devlink.h>
 
 #define MLX5_ESWITCH_MANAGER(mdev) MLX5_CAP_GEN(mdev, eswitch_manager)
···
 static inline bool is_mdev_switchdev_mode(struct mlx5_core_dev *dev)
 {
 	return mlx5_eswitch_mode(dev) == MLX5_ESWITCH_OFFLOADS;
+}
+
+/* The returned number is valid only when the dev is eswitch manager. */
+static inline u16 mlx5_eswitch_manager_vport(struct mlx5_core_dev *dev)
+{
+	return mlx5_core_is_ecpf_esw_manager(dev) ?
+		MLX5_VPORT_ECPF : MLX5_VPORT_PF;
 }
 
 #endif

include/linux/mlx5/mlx5_ifc.h

···
 	u8         log_sw_icm_alloc_granularity[0x6];
 	u8         log_steering_sw_icm_size[0x8];
 
-	u8         reserved_at_120[0x18];
+	u8         log_indirect_encap_sw_icm_size[0x8];
+	u8         reserved_at_128[0x10];
 	u8         log_header_modify_pattern_sw_icm_size[0x8];
 
 	u8         header_modify_sw_icm_start_address[0x40];
···
 
 	u8         memic_operations[0x20];
 
-	u8         reserved_at_220[0x5e0];
+	u8         reserved_at_220[0x20];
+
+	u8         indirect_encap_sw_icm_start_address[0x40];
+
+	u8         reserved_at_280[0x580];
 };
 
 struct mlx5_ifc_device_event_cap_bits {

include/net/mana/gdma.h

···
 	GDMA_DEVICE_NONE	= 0,
 	GDMA_DEVICE_HWC		= 1,
 	GDMA_DEVICE_MANA	= 2,
+	GDMA_DEVICE_MANA_IB	= 3,
 };
 
 struct gdma_resource {
···
 
 #define GDMA_MESSAGE_V1 1
 #define GDMA_MESSAGE_V2 2
+#define GDMA_MESSAGE_V3 3
 
 struct gdma_general_resp {
 	struct gdma_resp_hdr hdr;
···
 
 	/* Azure network adapter */
 	struct gdma_dev		mana;
+
+	/* Azure RDMA adapter */
+	struct gdma_dev		mana_ib;
 };
 
 #define MAX_NUM_GDMA_DEVICES	4

include/uapi/rdma/bnxt_re-abi.h

···
 	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,
+	BNXT_RE_COMP_MASK_UCNTX_HW_RETX_ENABLED = 0x40,
 };
 
 enum bnxt_re_wqe_mode {
 	BNXT_QPLIB_WQE_MODE_STATIC	= 0x00,
 	BNXT_QPLIB_WQE_MODE_VARIABLE	= 0x01,
 	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 {
···
 	__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 {
···
 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 {
···
 
 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__*/

include/uapi/rdma/efa-abi.h

···
 /* 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
···
 	__u16 max_rq_sge;
 	__u32 max_rdma_size;
 	__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 */

include/uapi/rdma/hns-abi.h

···
 	__u32 pdn;
 };
 
+struct hns_roce_ib_create_ah_resp {
+	__u8 dmac[6];
+	__u8 reserved[2];
+};
+
 #endif /* HNS_ABI_USER_H */

include/uapi/rdma/mlx5-abi.h

···
 #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,
···
 	__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 {

include/uapi/rdma/mlx5_user_ioctl_verbs.h

···
 	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 {