From de8f847a5114ff7cfcdfc114af8485c431dec703 Mon Sep 17 00:00:00 2001
From: Yishai Hadas
Date: Thu, 1 Aug 2024 15:05:16 +0300
Subject: RDMA/mlx5: Add support for DMABUF MR registrations with Data-direct

Add support for DMABUF MR registrations with Data-direct device.

Upon userspace calling to register a DMABUF MR with the data direct bit
set, the below algorithm will be followed.

1) Obtain a pinned DMABUF umem from the IB core using the user input
parameters (FD, offset, length) and the DMA PF device.  The DMA PF
device is needed to allow the IOMMU to enable the DMA PF to access the
user buffer over PCI.

2) Create a KSM MKEY by setting its entries according to the user buffer
VA to IOVA mapping, with the MKEY being the data direct device-crossed
MKEY. This KSM MKEY is umrable and will be used as part of the MR cache.
The PD for creating it is the internal device 'data direct' kernel one.

3) Create a crossing MKEY that points to the KSM MKEY using the crossing
access mode.

4) Manage the KSM MKEY by adding it to a list of 'data direct' MKEYs
managed on the mlx5_ib device.

5) Return the crossing MKEY to the user, created with its supplied PD.

Upon DMA PF unbind flow, the driver will revoke the KSM entries.
The final deregistration will occur under the hood once the application
deregisters its MKEY.

Notes:
- This version supports only the PINNED UMEM mode, so there is no
  dependency on ODP.
- The IOVA supplied by the application must be system page aligned due to
  HW translations of KSM.
- The crossing MKEY will not be umrable or part of the MR cache, as we
  cannot change its crossed (i.e. KSM) MKEY over UMR.

Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Link: https://patch.msgid.link/1f99d8020ed540d9702b9e2252a145a439609ba6.1722512548.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 include/uapi/rdma/mlx5_user_ioctl_cmds.h  | 4 ++++
 include/uapi/rdma/mlx5_user_ioctl_verbs.h | 4 ++++
 2 files changed, 8 insertions(+)

(limited to 'include/uapi')

diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
index 5b74d6534899..106276a4cce7 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_cmds.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
@@ -274,6 +274,10 @@ enum mlx5_ib_create_cq_attrs {
 	MLX5_IB_ATTR_CREATE_CQ_UAR_INDEX = UVERBS_ID_DRIVER_NS_WITH_UHW,
 };
 
+enum mlx5_ib_reg_dmabuf_mr_attrs {
+	MLX5_IB_ATTR_REG_DMABUF_MR_ACCESS_FLAGS = (1U << UVERBS_ID_NS_SHIFT),
+};
+
 #define MLX5_IB_DW_MATCH_PARAM 0xA0
 
 struct mlx5_ib_match_params {
diff --git a/include/uapi/rdma/mlx5_user_ioctl_verbs.h b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
index 3189c7f08d17..7c233df475e7 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_verbs.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -54,6 +54,10 @@ enum mlx5_ib_uapi_flow_action_packet_reformat_type {
 	MLX5_IB_UAPI_FLOW_ACTION_PACKET_REFORMAT_TYPE_L2_TO_L3_TUNNEL = 0x3,
 };
 
+enum mlx5_ib_uapi_reg_dmabuf_flags {
+	MLX5_IB_UAPI_REG_DMABUF_ACCESS_DATA_DIRECT = 1 << 0,
+};
+
 struct mlx5_ib_uapi_devx_async_cmd_hdr {
 	__aligned_u64	wr_id;
 	__u8		out_data[];
-- 
cgit v1.2.3


From ec7ad6530909983c8736c80af46e3529ce7bab55 Mon Sep 17 00:00:00 2001
From: Yishai Hadas
Date: Thu, 1 Aug 2024 15:05:17 +0300
Subject: RDMA/mlx5: Introduce GET_DATA_DIRECT_SYSFS_PATH ioctl

Introduce the 'GET_DATA_DIRECT_SYSFS_PATH' ioctl to return the sysfs
path of the affiliated 'data direct' device for a given device.

Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Link: https://patch.msgid.link/403745463e0ef52adbef681ff09aa6a29a756352.1722512548.git.leon@kernel.org
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/mlx5/std_types.c   | 55 +++++++++++++++++++++++++++++++-
 include/uapi/rdma/mlx5_user_ioctl_cmds.h |  5 +++
 2 files changed, 59 insertions(+), 1 deletion(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/mlx5/std_types.c b/drivers/infiniband/hw/mlx5/std_types.c
index 5c83765a85e0..bdb568411091 100644
--- a/drivers/infiniband/hw/mlx5/std_types.c
+++ b/drivers/infiniband/hw/mlx5/std_types.c
@@ -10,6 +10,7 @@
 #include <linux/mlx5/eswitch.h>
 #include <linux/mlx5/vport.h>
 #include "mlx5_ib.h"
+#include "data_direct.h"
 
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
@@ -204,6 +205,50 @@ static int UVERBS_HANDLER(MLX5_IB_METHOD_QUERY_PORT)(
 					     sizeof(info));
 }
 
+static int UVERBS_HANDLER(MLX5_IB_METHOD_GET_DATA_DIRECT_SYSFS_PATH)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_data_direct_dev *data_direct_dev;
+	struct mlx5_ib_ucontext *c;
+	struct mlx5_ib_dev *dev;
+	int out_len = uverbs_attr_get_len(attrs,
+			MLX5_IB_ATTR_GET_DATA_DIRECT_SYSFS_PATH);
+	u32 dev_path_len;
+	char *dev_path;
+	int ret;
+
+	c = to_mucontext(ib_uverbs_get_ucontext(attrs));
+	if (IS_ERR(c))
+		return PTR_ERR(c);
+	dev = to_mdev(c->ibucontext.device);
+	mutex_lock(&dev->data_direct_lock);
+	data_direct_dev = dev->data_direct_dev;
+	if (!data_direct_dev) {
+		ret = -ENODEV;
+		goto end;
+	}
+
+	dev_path = kobject_get_path(&data_direct_dev->device->kobj, GFP_KERNEL);
+	if (!dev_path) {
+		ret = -ENOMEM;
+		goto end;
+	}
+
+	dev_path_len = strlen(dev_path) + 1;
+	if (dev_path_len > out_len) {
+		ret = -ENOSPC;
+		goto end;
+	}
+
+	ret = uverbs_copy_to(attrs, MLX5_IB_ATTR_GET_DATA_DIRECT_SYSFS_PATH, dev_path,
+			     dev_path_len);
+	kfree(dev_path);
+
+end:
+	mutex_unlock(&dev->data_direct_lock);
+	return ret;
+}
+
 DECLARE_UVERBS_NAMED_METHOD(
 	MLX5_IB_METHOD_QUERY_PORT,
 	UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_QUERY_PORT_PORT_NUM,
@@ -214,9 +259,17 @@ DECLARE_UVERBS_NAMED_METHOD(
 				   reg_c0),
 		UA_MANDATORY));
 
+DECLARE_UVERBS_NAMED_METHOD(
+	MLX5_IB_METHOD_GET_DATA_DIRECT_SYSFS_PATH,
+	UVERBS_ATTR_PTR_OUT(
+		MLX5_IB_ATTR_GET_DATA_DIRECT_SYSFS_PATH,
+		UVERBS_ATTR_MIN_SIZE(0),
+		UA_MANDATORY));
+
 ADD_UVERBS_METHODS(mlx5_ib_device,
 		   UVERBS_OBJECT_DEVICE,
-		   &UVERBS_METHOD(MLX5_IB_METHOD_QUERY_PORT));
+		   &UVERBS_METHOD(MLX5_IB_METHOD_QUERY_PORT),
+		   &UVERBS_METHOD(MLX5_IB_METHOD_GET_DATA_DIRECT_SYSFS_PATH));
 
 DECLARE_UVERBS_NAMED_METHOD(
 	MLX5_IB_METHOD_PD_QUERY,
diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
index 106276a4cce7..fd2e4a3a56b3 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_cmds.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
@@ -348,6 +348,7 @@ enum mlx5_ib_pd_methods {
 
 enum mlx5_ib_device_methods {
 	MLX5_IB_METHOD_QUERY_PORT = (1U << UVERBS_ID_NS_SHIFT),
+	MLX5_IB_METHOD_GET_DATA_DIRECT_SYSFS_PATH,
 };
 
 enum mlx5_ib_query_port_attrs {
@@ -355,4 +356,8 @@ enum mlx5_ib_query_port_attrs {
 	MLX5_IB_ATTR_QUERY_PORT,
 };
 
+enum mlx5_ib_get_data_direct_sysfs_path_attrs {
+	MLX5_IB_ATTR_GET_DATA_DIRECT_SYSFS_PATH = (1U << UVERBS_ID_NS_SHIFT),
+};
+
 #endif
-- 
cgit v1.2.3


From d8ea645d6984c84a87032063a0941f15a323831f Mon Sep 17 00:00:00 2001
From: Selvin Xavier
Date: Sun, 18 Aug 2024 21:47:26 -0700
Subject: RDMA/bnxt_re: Handle variable WQE support for user applications

User library calculates the number of slots required for user applications
and it can pass that information to the driver.  Driver can use this value
and update the HW directly. This mechanism is currently used only for the
newly introduced variable size WQEs.

Extend the bnxt_re_qp_req structure to pass the Send Queue slot count.
Reorganize the code to get the sq_slots before initializing the Send Queue
attributes.

Link: https://patch.msgid.link/r/1724042847-1481-5-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Hongguang Gao <hongguang.gao@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
---
 drivers/infiniband/hw/bnxt_re/ib_verbs.c | 106 ++++++++++++++++++-------------
 drivers/infiniband/hw/bnxt_re/ib_verbs.h |  16 ++++-
 include/uapi/rdma/bnxt_re-abi.h          |   6 ++
 3 files changed, 82 insertions(+), 46 deletions(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 4dd137b7a5ce..2932db129958 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -1017,20 +1017,15 @@ static int bnxt_re_setup_swqe_size(struct bnxt_re_qp *qp,
 }
 
 static int bnxt_re_init_user_qp(struct bnxt_re_dev *rdev, struct bnxt_re_pd *pd,
-				struct bnxt_re_qp *qp, struct ib_udata *udata)
+				struct bnxt_re_qp *qp, struct bnxt_re_ucontext *cntx,
+				struct bnxt_re_qp_req *ureq)
 {
 	struct bnxt_qplib_qp *qplib_qp;
-	struct bnxt_re_ucontext *cntx;
-	struct bnxt_re_qp_req ureq;
 	int bytes = 0, psn_sz;
 	struct ib_umem *umem;
 	int psn_nume;
 
 	qplib_qp = &qp->qplib_qp;
-	cntx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext,
-					 ib_uctx);
-	if (ib_copy_from_udata(&ureq, udata, sizeof(ureq)))
-		return -EFAULT;
 
 	bytes = (qplib_qp->sq.max_wqe * qplib_qp->sq.wqe_size);
 	/* Consider mapping PSN search memory only for RC QPs. */
@@ -1038,17 +1033,20 @@ static int bnxt_re_init_user_qp(struct bnxt_re_dev *rdev, struct bnxt_re_pd *pd,
 		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) ?
-			    qplib_qp->sq.max_wqe :
-			    ((qplib_qp->sq.max_wqe * qplib_qp->sq.wqe_size) /
-			      sizeof(struct bnxt_qplib_sge));
+		if (cntx && bnxt_re_is_var_size_supported(rdev, cntx)) {
+			psn_nume = ureq->sq_slots;
+		} else {
+			psn_nume = (qplib_qp->wqe_mode == BNXT_QPLIB_WQE_MODE_STATIC) ?
+			qplib_qp->sq.max_wqe : ((qplib_qp->sq.max_wqe * qplib_qp->sq.wqe_size) /
+				 sizeof(struct bnxt_qplib_sge));
+		}
 		if (_is_host_msn_table(rdev->qplib_res.dattr->dev_cap_flags2))
 			psn_nume = roundup_pow_of_two(psn_nume);
 		bytes += (psn_nume * psn_sz);
 	}
 
 	bytes = PAGE_ALIGN(bytes);
-	umem = ib_umem_get(&rdev->ibdev, ureq.qpsva, bytes,
+	umem = ib_umem_get(&rdev->ibdev, ureq->qpsva, bytes,
 			   IB_ACCESS_LOCAL_WRITE);
 	if (IS_ERR(umem))
 		return PTR_ERR(umem);
@@ -1057,12 +1055,12 @@ static int bnxt_re_init_user_qp(struct bnxt_re_dev *rdev, struct bnxt_re_pd *pd,
 	qplib_qp->sq.sg_info.umem = umem;
 	qplib_qp->sq.sg_info.pgsize = PAGE_SIZE;
 	qplib_qp->sq.sg_info.pgshft = PAGE_SHIFT;
-	qplib_qp->qp_handle = ureq.qp_handle;
+	qplib_qp->qp_handle = ureq->qp_handle;
 
 	if (!qp->qplib_qp.srq) {
 		bytes = (qplib_qp->rq.max_wqe * qplib_qp->rq.wqe_size);
 		bytes = PAGE_ALIGN(bytes);
-		umem = ib_umem_get(&rdev->ibdev, ureq.qprva, bytes,
+		umem = ib_umem_get(&rdev->ibdev, ureq->qprva, bytes,
 				   IB_ACCESS_LOCAL_WRITE);
 		if (IS_ERR(umem))
 			goto rqfail;
@@ -1261,14 +1259,15 @@ static void bnxt_re_adjust_gsi_rq_attr(struct bnxt_re_qp *qp)
 
 static int bnxt_re_init_sq_attr(struct bnxt_re_qp *qp,
 				struct ib_qp_init_attr *init_attr,
-				struct bnxt_re_ucontext *uctx)
+				struct bnxt_re_ucontext *uctx,
+				struct bnxt_re_qp_req *ureq)
 {
 	struct bnxt_qplib_dev_attr *dev_attr;
 	struct bnxt_qplib_qp *qplqp;
 	struct bnxt_re_dev *rdev;
 	struct bnxt_qplib_q *sq;
+	int diff = 0;
 	int entries;
-	int diff;
 	int rc;
 
 	rdev = qp->rdev;
@@ -1277,22 +1276,28 @@ static int bnxt_re_init_sq_attr(struct bnxt_re_qp *qp,
 	dev_attr = &rdev->dev_attr;
 
 	sq->max_sge = init_attr->cap.max_send_sge;
-	if (sq->max_sge > dev_attr->max_qp_sges) {
-		sq->max_sge = dev_attr->max_qp_sges;
-		init_attr->cap.max_send_sge = sq->max_sge;
-	}
+	entries = init_attr->cap.max_send_wr;
+	if (uctx && qplqp->wqe_mode == BNXT_QPLIB_WQE_MODE_VARIABLE) {
+		sq->max_wqe = ureq->sq_slots;
+		sq->max_sw_wqe = ureq->sq_slots;
+		sq->wqe_size = sizeof(struct sq_sge);
+	} else {
+		if (sq->max_sge > dev_attr->max_qp_sges) {
+			sq->max_sge = dev_attr->max_qp_sges;
+			init_attr->cap.max_send_sge = sq->max_sge;
+		}
 
-	rc = bnxt_re_setup_swqe_size(qp, init_attr);
-	if (rc)
-		return rc;
+		rc = bnxt_re_setup_swqe_size(qp, init_attr);
+		if (rc)
+			return rc;
 
-	entries = init_attr->cap.max_send_wr;
-	/* Allocate 128 + 1 more than what's provided */
-	diff = (qplqp->wqe_mode == BNXT_QPLIB_WQE_MODE_VARIABLE) ?
-		0 : BNXT_QPLIB_RESERVED_QP_WRS;
-	entries = bnxt_re_init_depth(entries + diff + 1, uctx);
-	sq->max_wqe = min_t(u32, entries, dev_attr->max_qp_wqes + diff + 1);
-	sq->max_sw_wqe = bnxt_qplib_get_depth(sq, qplqp->wqe_mode, true);
+		/* Allocate 128 + 1 more than what's provided */
+		diff = (qplqp->wqe_mode == BNXT_QPLIB_WQE_MODE_VARIABLE) ?
+			0 : BNXT_QPLIB_RESERVED_QP_WRS;
+		entries = bnxt_re_init_depth(entries + diff + 1, uctx);
+		sq->max_wqe = min_t(u32, entries, dev_attr->max_qp_wqes + diff + 1);
+		sq->max_sw_wqe = bnxt_qplib_get_depth(sq, qplqp->wqe_mode, true);
+	}
 	sq->q_full_delta = diff + 1;
 	/*
 	 * Reserving one slot for Phantom WQE. Application can
@@ -1355,10 +1360,10 @@ out:
 
 static int bnxt_re_init_qp_attr(struct bnxt_re_qp *qp, struct bnxt_re_pd *pd,
 				struct ib_qp_init_attr *init_attr,
-				struct ib_udata *udata)
+				struct bnxt_re_ucontext *uctx,
+				struct bnxt_re_qp_req *ureq)
 {
 	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;
@@ -1368,7 +1373,6 @@ static int bnxt_re_init_qp_attr(struct bnxt_re_qp *qp, struct bnxt_re_pd *pd,
 	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;
@@ -1381,8 +1385,7 @@ static int bnxt_re_init_qp_attr(struct bnxt_re_qp *qp, struct bnxt_re_pd *pd,
 		goto out;
 	}
 	qplqp->type = (u8)qptype;
-	qplqp->wqe_mode = rdev->chip_ctx->modes.wqe_mode;
-
+	qplqp->wqe_mode = bnxt_re_is_var_size_supported(rdev, uctx);
 	if (init_attr->qp_type == IB_QPT_RC) {
 		qplqp->max_rd_atomic = dev_attr->max_qp_rd_atom;
 		qplqp->max_dest_rd_atomic = dev_attr->max_qp_init_rd_atom;
@@ -1417,14 +1420,14 @@ static int bnxt_re_init_qp_attr(struct bnxt_re_qp *qp, struct bnxt_re_pd *pd,
 		bnxt_re_adjust_gsi_rq_attr(qp);
 
 	/* Setup SQ */
-	rc = bnxt_re_init_sq_attr(qp, init_attr, uctx);
+	rc = bnxt_re_init_sq_attr(qp, init_attr, uctx, ureq);
 	if (rc)
 		goto out;
 	if (init_attr->qp_type == IB_QPT_GSI)
 		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);
+	if (uctx) /* This will update DPI and qp_handle */
+		rc = bnxt_re_init_user_qp(rdev, pd, qp, uctx, ureq);
 out:
 	return rc;
 }
@@ -1525,14 +1528,27 @@ static bool bnxt_re_test_qp_limits(struct bnxt_re_dev *rdev,
 int bnxt_re_create_qp(struct ib_qp *ib_qp, struct ib_qp_init_attr *qp_init_attr,
 		      struct ib_udata *udata)
 {
-	struct ib_pd *ib_pd = ib_qp->pd;
-	struct bnxt_re_pd *pd = container_of(ib_pd, struct bnxt_re_pd, ib_pd);
-	struct bnxt_re_dev *rdev = pd->rdev;
-	struct bnxt_qplib_dev_attr *dev_attr = &rdev->dev_attr;
-	struct bnxt_re_qp *qp = container_of(ib_qp, struct bnxt_re_qp, ib_qp);
+	struct bnxt_qplib_dev_attr *dev_attr;
+	struct bnxt_re_ucontext *uctx;
+	struct bnxt_re_qp_req ureq;
+	struct bnxt_re_dev *rdev;
+	struct bnxt_re_pd *pd;
+	struct bnxt_re_qp *qp;
+	struct ib_pd *ib_pd;
 	u32 active_qps;
 	int rc;
 
+	ib_pd = ib_qp->pd;
+	pd = container_of(ib_pd, struct bnxt_re_pd, ib_pd);
+	rdev = pd->rdev;
+	dev_attr = &rdev->dev_attr;
+	qp = container_of(ib_qp, struct bnxt_re_qp, ib_qp);
+
+	uctx = rdma_udata_to_drv_context(udata, struct bnxt_re_ucontext, ib_uctx);
+	if (udata)
+		if (ib_copy_from_udata(&ureq, udata,  min(udata->inlen, sizeof(ureq))))
+			return -EFAULT;
+
 	rc = bnxt_re_test_qp_limits(rdev, qp_init_attr, dev_attr);
 	if (!rc) {
 		rc = -EINVAL;
@@ -1540,7 +1556,7 @@ int bnxt_re_create_qp(struct ib_qp *ib_qp, struct ib_qp_init_attr *qp_init_attr,
 	}
 
 	qp->rdev = rdev;
-	rc = bnxt_re_init_qp_attr(qp, pd, qp_init_attr, udata);
+	rc = bnxt_re_init_qp_attr(qp, pd, qp_init_attr, uctx, &ureq);
 	if (rc)
 		goto fail;
 
@@ -4215,7 +4231,7 @@ int bnxt_re_alloc_ucontext(struct ib_ucontext *ctx, struct ib_udata *udata)
 			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;
+			uctx->cmask |= BNXT_RE_UCNTX_CAP_POW2_DISABLED;
 		}
 	}
 
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index e98cb1717338..7c8350fb8aad 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -171,12 +171,26 @@ static inline u16 bnxt_re_get_rwqe_size(int nsge)
 	return sizeof(struct rq_wqe_hdr) + (nsge * sizeof(struct sq_sge));
 }
 
+enum {
+	BNXT_RE_UCNTX_CAP_POW2_DISABLED = 0x1ULL,
+	BNXT_RE_UCNTX_CAP_VAR_WQE_ENABLED = 0x2ULL,
+};
+
 static inline u32 bnxt_re_init_depth(u32 ent, struct bnxt_re_ucontext *uctx)
 {
-	return uctx ? (uctx->cmask & BNXT_RE_UCNTX_CMASK_POW2_DISABLED) ?
+	return uctx ? (uctx->cmask & BNXT_RE_UCNTX_CAP_POW2_DISABLED) ?
 		ent : roundup_pow_of_two(ent) : ent;
 }
 
+static inline bool bnxt_re_is_var_size_supported(struct bnxt_re_dev *rdev,
+						 struct bnxt_re_ucontext *uctx)
+{
+	if (uctx)
+		return uctx->cmask & BNXT_RE_UCNTX_CAP_VAR_WQE_ENABLED;
+	else
+		return rdev->chip_ctx->modes.wqe_mode;
+}
+
 int bnxt_re_query_device(struct ib_device *ibdev,
 			 struct ib_device_attr *ib_attr,
 			 struct ib_udata *udata);
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index e61104f35d73..71140618700a 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -118,10 +118,16 @@ struct bnxt_re_resize_cq_req {
 	__aligned_u64 cq_va;
 };
 
+enum bnxt_re_qp_mask {
+	BNXT_RE_QP_REQ_MASK_VAR_WQE_SQ_SLOTS = 0x1,
+};
+
 struct bnxt_re_qp_req {
 	__aligned_u64 qpsva;
 	__aligned_u64 qprva;
 	__aligned_u64 qp_handle;
+	__aligned_u64 comp_mask;
+	__u32 sq_slots;
 };
 
 struct bnxt_re_qp_resp {
-- 
cgit v1.2.3


From 10a104c0debbb19a1e45193d5670510216e339ff Mon Sep 17 00:00:00 2001
From: Selvin Xavier
Date: Sun, 18 Aug 2024 21:47:27 -0700
Subject: RDMA/bnxt_re: Enable variable size WQEs for user space applications

Add backward compatibility code to enable variable size WQEs only if the
user lib supports it.

Link: https://patch.msgid.link/r/1724042847-1481-6-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Hongguang Gao <hongguang.gao@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
---
 drivers/infiniband/hw/bnxt_re/ib_verbs.c | 5 +++++
 include/uapi/rdma/bnxt_re-abi.h          | 1 +
 2 files changed, 6 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 2932db129958..82444fd748f1 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -4233,6 +4233,11 @@ int bnxt_re_alloc_ucontext(struct ib_ucontext *ctx, struct ib_udata *udata)
 			resp.comp_mask |= BNXT_RE_UCNTX_CMASK_POW2_DISABLED;
 			uctx->cmask |= BNXT_RE_UCNTX_CAP_POW2_DISABLED;
 		}
+		if (ureq.comp_mask & BNXT_RE_COMP_MASK_REQ_UCNTX_VAR_WQE_SUPPORT) {
+			resp.comp_mask |= BNXT_RE_UCNTX_CMASK_HAVE_MODE;
+			resp.mode = rdev->chip_ctx->modes.wqe_mode;
+			uctx->cmask |= BNXT_RE_UCNTX_CAP_VAR_WQE_ENABLED;
+		}
 	}
 
 	rc = ib_copy_to_udata(udata, &resp, min(udata->outlen, sizeof(resp)));
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 71140618700a..6821002931c8 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -66,6 +66,7 @@ enum bnxt_re_wqe_mode {
 
 enum {
 	BNXT_RE_COMP_MASK_REQ_UCNTX_POW2_SUPPORT = 0x01,
+	BNXT_RE_COMP_MASK_REQ_UCNTX_VAR_WQE_SUPPORT = 0x02,
 };
 
 struct bnxt_re_uctx_req {
-- 
cgit v1.2.3


From 181028a0d84cdcc7ac86d05cc49eaa416ce85c8b Mon Sep 17 00:00:00 2001
From: Chandramohan Akula
Date: Thu, 29 Aug 2024 08:34:05 -0700
Subject: RDMA/bnxt_re: Share a page to expose per SRQ info with userspace

Gen P7 adapters needs to share a toggle bits information received
in kernel driver with the user space. User space needs this
info to arm the SRQ.

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

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

Signed-off-by: Chandramohan Akula <chandramohan.akula@broadcom.com>
Signed-off-by: Selvin Xavier <selvin.xavier@broadcom.com>
Link: https://patch.msgid.link/1724945645-14989-4-git-send-email-selvin.xavier@broadcom.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/hw/bnxt_re/bnxt_re.h  |  2 ++
 drivers/infiniband/hw/bnxt_re/ib_verbs.c | 34 +++++++++++++++++++++++++++++++-
 drivers/infiniband/hw/bnxt_re/ib_verbs.h |  1 +
 drivers/infiniband/hw/bnxt_re/main.c     |  6 +++++-
 include/uapi/rdma/bnxt_re-abi.h          |  6 ++++++
 5 files changed, 47 insertions(+), 2 deletions(-)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/hw/bnxt_re/bnxt_re.h b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
index 0912d2fa9634..2be9a62d230f 100644
--- a/drivers/infiniband/hw/bnxt_re/bnxt_re.h
+++ b/drivers/infiniband/hw/bnxt_re/bnxt_re.h
@@ -141,6 +141,7 @@ struct bnxt_re_pacing {
 #define BNXT_RE_GRC_FIFO_REG_BASE 0x2000
 
 #define MAX_CQ_HASH_BITS		(16)
+#define MAX_SRQ_HASH_BITS		(16)
 struct bnxt_re_dev {
 	struct ib_device		ibdev;
 	struct list_head		list;
@@ -196,6 +197,7 @@ struct bnxt_re_dev {
 	struct work_struct dbq_fifo_check_work;
 	struct delayed_work dbq_pacing_work;
 	DECLARE_HASHTABLE(cq_hash, MAX_CQ_HASH_BITS);
+	DECLARE_HASHTABLE(srq_hash, MAX_SRQ_HASH_BITS);
 };
 
 #define to_bnxt_re_dev(ptr, member)	\
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index eafbee4af067..f9f944a6094a 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -1707,6 +1707,10 @@ int bnxt_re_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata)
 
 	if (qplib_srq->cq)
 		nq = qplib_srq->cq->nq;
+	if (rdev->chip_ctx->modes.toggle_bits & BNXT_QPLIB_SRQ_TOGGLE_BIT) {
+		free_page((unsigned long)srq->uctx_srq_page);
+		hash_del(&srq->hash_entry);
+	}
 	bnxt_qplib_destroy_srq(&rdev->qplib_res, qplib_srq);
 	ib_umem_release(srq->umem);
 	atomic_dec(&rdev->stats.res.srq_count);
@@ -1811,9 +1815,18 @@ int bnxt_re_create_srq(struct ib_srq *ib_srq,
 	}
 
 	if (udata) {
-		struct bnxt_re_srq_resp resp;
+		struct bnxt_re_srq_resp resp = {};
 
 		resp.srqid = srq->qplib_srq.id;
+		if (rdev->chip_ctx->modes.toggle_bits & BNXT_QPLIB_SRQ_TOGGLE_BIT) {
+			hash_add(rdev->srq_hash, &srq->hash_entry, srq->qplib_srq.id);
+			srq->uctx_srq_page = (void *)get_zeroed_page(GFP_KERNEL);
+			if (!srq->uctx_srq_page) {
+				rc = -ENOMEM;
+				goto fail;
+			}
+			resp.comp_mask |= BNXT_RE_SRQ_TOGGLE_PAGE_SUPPORT;
+		}
 		rc = ib_copy_to_udata(udata, &resp, sizeof(resp));
 		if (rc) {
 			ibdev_err(&rdev->ibdev, "SRQ copy to udata failed!");
@@ -4291,6 +4304,19 @@ static struct bnxt_re_cq *bnxt_re_search_for_cq(struct bnxt_re_dev *rdev, u32 cq
 	return cq;
 }
 
+static struct bnxt_re_srq *bnxt_re_search_for_srq(struct bnxt_re_dev *rdev, u32 srq_id)
+{
+	struct bnxt_re_srq *srq = NULL, *tmp_srq;
+
+	hash_for_each_possible(rdev->srq_hash, tmp_srq, hash_entry, srq_id) {
+		if (tmp_srq->qplib_srq.id == srq_id) {
+			srq = tmp_srq;
+			break;
+		}
+	}
+	return srq;
+}
+
 /* Helper function to mmap the virtual memory from user app */
 int bnxt_re_mmap(struct ib_ucontext *ib_uctx, struct vm_area_struct *vma)
 {
@@ -4519,6 +4545,7 @@ static int UVERBS_HANDLER(BNXT_RE_METHOD_GET_TOGGLE_MEM)(struct uverbs_attr_bund
 	struct bnxt_re_ucontext *uctx;
 	struct ib_ucontext *ib_uctx;
 	struct bnxt_re_dev *rdev;
+	struct bnxt_re_srq *srq;
 	u32 length = PAGE_SIZE;
 	struct bnxt_re_cq *cq;
 	u64 mem_offset;
@@ -4550,6 +4577,11 @@ static int UVERBS_HANDLER(BNXT_RE_METHOD_GET_TOGGLE_MEM)(struct uverbs_attr_bund
 		addr = (u64)cq->uctx_cq_page;
 		break;
 	case BNXT_RE_SRQ_TOGGLE_MEM:
+		srq = bnxt_re_search_for_srq(rdev, res_id);
+		if (!srq)
+			return -EINVAL;
+
+		addr = (u64)srq->uctx_srq_page;
 		break;
 
 	default:
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index 060bf62eceba..b789e47ec97a 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -78,6 +78,7 @@ struct bnxt_re_srq {
 	struct ib_umem		*umem;
 	spinlock_t		lock;		/* protect srq */
 	void			*uctx_srq_page;
+	struct hlist_node       hash_entry;
 };
 
 struct bnxt_re_qp {
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index 31ba89cffe9d..16a84ca1ce48 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -139,8 +139,10 @@ static void bnxt_re_set_drv_mode(struct bnxt_re_dev *rdev)
 	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))
+	if (bnxt_qplib_is_chip_gen_p7(rdev->chip_ctx)) {
 		cctx->modes.toggle_bits |= BNXT_QPLIB_CQ_TOGGLE_BIT;
+		cctx->modes.toggle_bits |= BNXT_QPLIB_SRQ_TOGGLE_BIT;
+	}
 }
 
 static void bnxt_re_destroy_chip_ctx(struct bnxt_re_dev *rdev)
@@ -1771,6 +1773,8 @@ static int bnxt_re_dev_init(struct bnxt_re_dev *rdev)
 		bnxt_re_vf_res_config(rdev);
 	}
 	hash_init(rdev->cq_hash);
+	if (rdev->chip_ctx->modes.toggle_bits & BNXT_QPLIB_SRQ_TOGGLE_BIT)
+		hash_init(rdev->srq_hash);
 
 	return 0;
 free_sctx:
diff --git a/include/uapi/rdma/bnxt_re-abi.h b/include/uapi/rdma/bnxt_re-abi.h
index 6821002931c8..faa9d62b3b30 100644
--- a/include/uapi/rdma/bnxt_re-abi.h
+++ b/include/uapi/rdma/bnxt_re-abi.h
@@ -141,8 +141,14 @@ struct bnxt_re_srq_req {
 	__aligned_u64 srq_handle;
 };
 
+enum bnxt_re_srq_mask {
+	BNXT_RE_SRQ_TOGGLE_PAGE_SUPPORT = 0x1,
+};
+
 struct bnxt_re_srq_resp {
 	__u32 srqid;
+	__u32 rsvd; /* padding */
+	__aligned_u64 comp_mask;
 };
 
 enum bnxt_re_shpg_offt {
-- 
cgit v1.2.3


From 9cbed5aab5aeea420d0aa945733bf608449d44fb Mon Sep 17 00:00:00 2001
From: Chiara Meiohas
Date: Mon, 9 Sep 2024 20:30:24 +0300
Subject: RDMA/nldev: Add support for RDMA monitoring

Introduce a new netlink command to allow rdma event monitoring.
The rdma events supported now are IB device
registration/unregistration and net device attachment/detachment.

Example output of rdma monitor and the commands which trigger
the events:

$ rdma monitor
$ rmmod mlx5_ib
[UNREGISTER]	dev 1 rocep8s0f1
[UNREGISTER]	dev 0 rocep8s0f0

$ modprobe mlx5_ib
[REGISTER]	dev 2 mlx5_0
[NETDEV_ATTACH]	dev 2 mlx5_0 port 1 netdev 4 eth2
[REGISTER]	dev 3 mlx5_1
[NETDEV_ATTACH]	dev 3 mlx5_1 port 1 netdev 5 eth3

$ devlink dev eswitch set pci/0000:08:00.0 mode switchdev
[UNREGISTER]	dev 2 rocep8s0f0
[REGISTER]	dev 4 mlx5_0
[NETDEV_ATTACH]	dev 4 mlx5_0 port 30 netdev 4 eth2

$ echo 4 > /sys/class/net/eth2/device/sriov_numvfs
[NETDEV_ATTACH]	dev 4 rdmap8s0f0 port 2 netdev 7 eth4
[NETDEV_ATTACH]	dev 4 rdmap8s0f0 port 3 netdev 8 eth5
[NETDEV_ATTACH]	dev 4 rdmap8s0f0 port 4 netdev 9 eth6
[NETDEV_ATTACH]	dev 4 rdmap8s0f0 port 5 netdev 10 eth7
[REGISTER]	dev 5 mlx5_0
[NETDEV_ATTACH]	dev 5 mlx5_0 port 1 netdev 11 eth8
[REGISTER]	dev 6 mlx5_0
[NETDEV_ATTACH]	dev 6 mlx5_0 port 1 netdev 12 eth9
[REGISTER]	dev 7 mlx5_0
[NETDEV_ATTACH]	dev 7 mlx5_0 port 1 netdev 13 eth10
[REGISTER]	dev 8 mlx5_0
[NETDEV_ATTACH]	dev 8 mlx5_0 port 1 netdev 14 eth11

$ echo 0 > /sys/class/net/eth2/device/sriov_numvfs
[UNREGISTER]	dev 5 rocep8s0f0v0
[UNREGISTER]	dev 6 rocep8s0f0v1
[UNREGISTER]	dev 7 rocep8s0f0v2
[UNREGISTER]	dev 8 rocep8s0f0v3
[NETDEV_DETACH]	dev 4 rdmap8s0f0 port 2
[NETDEV_DETACH]	dev 4 rdmap8s0f0 port 3
[NETDEV_DETACH]	dev 4 rdmap8s0f0 port 4
[NETDEV_DETACH]	dev 4 rdmap8s0f0 port 5

Signed-off-by: Chiara Meiohas <cmeiohas@nvidia.com>
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Link: https://patch.msgid.link/20240909173025.30422-7-michaelgur@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/core/device.c  |  35 +++++++++++
 drivers/infiniband/core/netlink.c |   1 +
 drivers/infiniband/core/nldev.c   | 124 ++++++++++++++++++++++++++++++++++++++
 include/rdma/rdma_netlink.h       |  12 ++++
 include/uapi/rdma/rdma_netlink.h  |  15 +++++
 5 files changed, 187 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c
index 9e765c79a892..e029401b5680 100644
--- a/drivers/infiniband/core/device.c
+++ b/drivers/infiniband/core/device.c
@@ -1351,6 +1351,29 @@ static void prevent_dealloc_device(struct ib_device *ib_dev)
 {
 }
 
+static void ib_device_notify_register(struct ib_device *device)
+{
+	struct net_device *netdev;
+	u32 port;
+	int ret;
+
+	ret = rdma_nl_notify_event(device, 0, RDMA_REGISTER_EVENT);
+	if (ret)
+		return;
+
+	rdma_for_each_port(device, port) {
+		netdev = ib_device_get_netdev(device, port);
+		if (!netdev)
+			continue;
+
+		ret = rdma_nl_notify_event(device, port,
+					   RDMA_NETDEV_ATTACH_EVENT);
+		dev_put(netdev);
+		if (ret)
+			return;
+	}
+}
+
 /**
  * ib_register_device - Register an IB device with IB core
  * @device: Device to register
@@ -1449,6 +1472,8 @@ int ib_register_device(struct ib_device *device, const char *name,
 	dev_set_uevent_suppress(&device->dev, false);
 	/* Mark for userspace that device is ready */
 	kobject_uevent(&device->dev.kobj, KOBJ_ADD);
+
+	ib_device_notify_register(device);
 	ib_device_put(device);
 
 	return 0;
@@ -1491,6 +1516,7 @@ static void __ib_unregister_device(struct ib_device *ib_dev)
 		goto out;
 
 	disable_device(ib_dev);
+	rdma_nl_notify_event(ib_dev, 0, RDMA_UNREGISTER_EVENT);
 
 	/* Expedite removing unregistered pointers from the hash table */
 	free_netdevs(ib_dev);
@@ -2159,6 +2185,7 @@ static void add_ndev_hash(struct ib_port_data *pdata)
 int ib_device_set_netdev(struct ib_device *ib_dev, struct net_device *ndev,
 			 u32 port)
 {
+	enum rdma_nl_notify_event_type etype;
 	struct net_device *old_ndev;
 	struct ib_port_data *pdata;
 	unsigned long flags;
@@ -2190,6 +2217,14 @@ int ib_device_set_netdev(struct ib_device *ib_dev, struct net_device *ndev,
 	spin_unlock_irqrestore(&pdata->netdev_lock, flags);
 
 	add_ndev_hash(pdata);
+
+	/* Make sure that the device is registered before we send events */
+	if (xa_load(&devices, ib_dev->index) != ib_dev)
+		return 0;
+
+	etype = ndev ? RDMA_NETDEV_ATTACH_EVENT : RDMA_NETDEV_DETACH_EVENT;
+	rdma_nl_notify_event(ib_dev, port, etype);
+
 	return 0;
 }
 EXPORT_SYMBOL(ib_device_set_netdev);
diff --git a/drivers/infiniband/core/netlink.c b/drivers/infiniband/core/netlink.c
index ae2db0c70788..def14c54b648 100644
--- a/drivers/infiniband/core/netlink.c
+++ b/drivers/infiniband/core/netlink.c
@@ -311,6 +311,7 @@ int rdma_nl_net_init(struct rdma_dev_net *rnet)
 	struct net *net = read_pnet(&rnet->net);
 	struct netlink_kernel_cfg cfg = {
 		.input	= rdma_nl_rcv,
+		.flags = NL_CFG_F_NONROOT_RECV,
 	};
 	struct sock *nls;
 
diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 4d4a1f90e484..70b3fa0469f2 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -170,6 +170,7 @@ static const struct nla_policy nldev_policy[RDMA_NLDEV_ATTR_MAX] = {
 	[RDMA_NLDEV_ATTR_DEV_TYPE]		= { .type = NLA_U8 },
 	[RDMA_NLDEV_ATTR_PARENT_NAME]		= { .type = NLA_NUL_STRING },
 	[RDMA_NLDEV_ATTR_NAME_ASSIGN_TYPE]	= { .type = NLA_U8 },
+	[RDMA_NLDEV_ATTR_EVENT_TYPE]		= { .type = NLA_U8 },
 };
 
 static int put_driver_name_print_type(struct sk_buff *msg, const char *name,
@@ -2722,6 +2723,129 @@ static const struct rdma_nl_cbs nldev_cb_table[RDMA_NLDEV_NUM_OPS] = {
 	},
 };
 
+static int fill_mon_netdev_association(struct sk_buff *msg,
+				       struct ib_device *device, u32 port,
+				       const struct net *net)
+{
+	struct net_device *netdev = ib_device_get_netdev(device, port);
+	int ret = 0;
+
+	if (netdev && !net_eq(dev_net(netdev), net))
+		goto out;
+
+	ret = nla_put_u32(msg, RDMA_NLDEV_ATTR_DEV_INDEX, device->index);
+	if (ret)
+		goto out;
+
+	ret = nla_put_string(msg, RDMA_NLDEV_ATTR_DEV_NAME,
+			     dev_name(&device->dev));
+	if (ret)
+		goto out;
+
+	ret = nla_put_u32(msg, RDMA_NLDEV_ATTR_PORT_INDEX, port);
+	if (ret)
+		goto out;
+
+	if (netdev) {
+		ret = nla_put_u32(msg,
+				  RDMA_NLDEV_ATTR_NDEV_INDEX, netdev->ifindex);
+		if (ret)
+			goto out;
+
+		ret = nla_put_string(msg,
+				     RDMA_NLDEV_ATTR_NDEV_NAME, netdev->name);
+	}
+
+out:
+	dev_put(netdev);
+	return ret;
+}
+
+static void rdma_nl_notify_err_msg(struct ib_device *device, u32 port_num,
+				    enum rdma_nl_notify_event_type type)
+{
+	struct net_device *netdev;
+
+	switch (type) {
+	case RDMA_REGISTER_EVENT:
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor register device event\n");
+		break;
+	case RDMA_UNREGISTER_EVENT:
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor unregister device event\n");
+		break;
+	case RDMA_NETDEV_ATTACH_EVENT:
+		netdev = ib_device_get_netdev(device, port_num);
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor netdev attach event: port %d netdev %d\n",
+				     port_num, netdev->ifindex);
+		dev_put(netdev);
+		break;
+	case RDMA_NETDEV_DETACH_EVENT:
+		dev_warn_ratelimited(&device->dev,
+				     "Failed to send RDMA monitor netdev detach event: port %d\n",
+				     port_num);
+	default:
+		break;
+	}
+}
+
+int rdma_nl_notify_event(struct ib_device *device, u32 port_num,
+			  enum rdma_nl_notify_event_type type)
+{
+	struct sk_buff *skb;
+	struct net *net;
+	int ret = 0;
+	void *nlh;
+
+	net = read_pnet(&device->coredev.rdma_net);
+	if (!net)
+		return -EINVAL;
+
+	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+	if (!skb)
+		return -ENOMEM;
+	nlh = nlmsg_put(skb, 0, 0,
+			RDMA_NL_GET_TYPE(RDMA_NL_NLDEV, RDMA_NLDEV_CMD_MONITOR),
+			0, 0);
+
+	switch (type) {
+	case RDMA_REGISTER_EVENT:
+	case RDMA_UNREGISTER_EVENT:
+		ret = fill_nldev_handle(skb, device);
+		if (ret)
+			goto err_free;
+		break;
+	case RDMA_NETDEV_ATTACH_EVENT:
+	case RDMA_NETDEV_DETACH_EVENT:
+		ret = fill_mon_netdev_association(skb, device,
+						  port_num, net);
+		if (ret)
+			goto err_free;
+		break;
+	default:
+		break;
+	}
+
+	ret = nla_put_u8(skb, RDMA_NLDEV_ATTR_EVENT_TYPE, type);
+	if (ret)
+		goto err_free;
+
+	nlmsg_end(skb, nlh);
+	ret = rdma_nl_multicast(net, skb, RDMA_NL_GROUP_NOTIFY, GFP_KERNEL);
+	if (ret && ret != -ESRCH) {
+		skb = NULL; /* skb is freed in the netlink send-op handling */
+		goto err_free;
+	}
+	return 0;
+
+err_free:
+	rdma_nl_notify_err_msg(device, port_num, type);
+	nlmsg_free(skb);
+	return ret;
+}
+
 void __init nldev_init(void)
 {
 	rdma_nl_register(RDMA_NL_NLDEV, nldev_cb_table);
diff --git a/include/rdma/rdma_netlink.h b/include/rdma/rdma_netlink.h
index c2a79aeee113..326deaf56d5d 100644
--- a/include/rdma/rdma_netlink.h
+++ b/include/rdma/rdma_netlink.h
@@ -6,6 +6,8 @@
 #include <linux/netlink.h>
 #include <uapi/rdma/rdma_netlink.h>
 
+struct ib_device;
+
 enum {
 	RDMA_NLDEV_ATTR_EMPTY_STRING = 1,
 	RDMA_NLDEV_ATTR_ENTRY_STRLEN = 16,
@@ -110,6 +112,16 @@ int rdma_nl_multicast(struct net *net, struct sk_buff *skb,
  */
 bool rdma_nl_chk_listeners(unsigned int group);
 
+/**
+ * Prepare and send an event message
+ * @ib: the IB device which triggered the event
+ * @port_num: the port number which triggered the event - 0 if unused
+ * @type: the event type
+ * Returns 0 on success or a negative error code
+ */
+int rdma_nl_notify_event(struct ib_device *ib, u32 port_num,
+			 enum rdma_nl_notify_event_type type);
+
 struct rdma_link_ops {
 	struct list_head list;
 	const char *type;
diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
index 2f37568f5556..5f9636d26050 100644
--- a/include/uapi/rdma/rdma_netlink.h
+++ b/include/uapi/rdma/rdma_netlink.h
@@ -15,6 +15,7 @@ enum {
 enum {
 	RDMA_NL_GROUP_IWPM = 2,
 	RDMA_NL_GROUP_LS,
+	RDMA_NL_GROUP_NOTIFY,
 	RDMA_NL_NUM_GROUPS
 };
 
@@ -305,6 +306,8 @@ enum rdma_nldev_command {
 
 	RDMA_NLDEV_CMD_DELDEV,
 
+	RDMA_NLDEV_CMD_MONITOR,
+
 	RDMA_NLDEV_NUM_OPS
 };
 
@@ -574,6 +577,8 @@ enum rdma_nldev_attr {
 
 	RDMA_NLDEV_ATTR_NAME_ASSIGN_TYPE,	/* u8 */
 
+	RDMA_NLDEV_ATTR_EVENT_TYPE,		/* u8 */
+
 	/*
 	 * Always the end
 	 */
@@ -624,4 +629,14 @@ enum rdma_nl_name_assign_type {
 	RDMA_NAME_ASSIGN_TYPE_USER = 1, /* Provided by user-space */
 };
 
+/*
+ * Supported rdma monitoring event types.
+ */
+enum rdma_nl_notify_event_type {
+	RDMA_REGISTER_EVENT,
+	RDMA_UNREGISTER_EVENT,
+	RDMA_NETDEV_ATTACH_EVENT,
+	RDMA_NETDEV_DETACH_EVENT,
+};
+
 #endif /* _UAPI_RDMA_NETLINK_H */
-- 
cgit v1.2.3


From 12fb1153c53bf9b53e299c9775b84fa7838640f7 Mon Sep 17 00:00:00 2001
From: Chiara Meiohas
Date: Mon, 9 Sep 2024 20:30:25 +0300
Subject: RDMA/nldev: Expose whether RDMA monitoring is supported

Extend the "rdma sys" command to display whether RDMA
monitoring is supported.

RDMA monitoring is not supported in mlx4 because it does
not use the ib_device_set_netdev() API, which sends the
RDMA events.

Example output for kernel where monitoring is supported:
$ rdma sys show
netns shared privileged-qkey off monitor on copy-on-fork on

Example output for kernel where monitoring is not supported:
$ rdma sys show
netns shared privileged-qkey off monitor off copy-on-fork on

Signed-off-by: Chiara Meiohas <cmeiohas@nvidia.com>
Signed-off-by: Michael Guralnik <michaelgur@nvidia.com>
Link: https://patch.msgid.link/20240909173025.30422-8-michaelgur@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
---
 drivers/infiniband/core/nldev.c  | 6 ++++++
 include/uapi/rdma/rdma_netlink.h | 1 +
 2 files changed, 7 insertions(+)

(limited to 'include/uapi')

diff --git a/drivers/infiniband/core/nldev.c b/drivers/infiniband/core/nldev.c
index 70b3fa0469f2..10b1411ac53d 100644
--- a/drivers/infiniband/core/nldev.c
+++ b/drivers/infiniband/core/nldev.c
@@ -1952,6 +1952,12 @@ static int nldev_sys_get_doit(struct sk_buff *skb, struct nlmsghdr *nlh,
 		nlmsg_free(msg);
 		return err;
 	}
+
+	err = nla_put_u8(msg, RDMA_NLDEV_SYS_ATTR_MONITOR_MODE, 1);
+	if (err) {
+		nlmsg_free(msg);
+		return err;
+	}
 	/*
 	 * Copy-on-fork is supported.
 	 * See commits:
diff --git a/include/uapi/rdma/rdma_netlink.h b/include/uapi/rdma/rdma_netlink.h
index 5f9636d26050..39be09c0ffbb 100644
--- a/include/uapi/rdma/rdma_netlink.h
+++ b/include/uapi/rdma/rdma_netlink.h
@@ -579,6 +579,7 @@ enum rdma_nldev_attr {
 
 	RDMA_NLDEV_ATTR_EVENT_TYPE,		/* u8 */
 
+	RDMA_NLDEV_SYS_ATTR_MONITOR_MODE,	/* u8 */
 	/*
 	 * Always the end
 	 */
-- 
cgit v1.2.3