From c44ef998f25eaddcd78924f98e5baed602d933e6 Mon Sep 17 00:00:00 2001
From: Ilya Lesokhin <ilyal@mellanox.com>
Date: Tue, 13 Mar 2018 15:18:48 +0200
Subject: IB/mlx5: Maintain a single emergency page

The mlx5 driver needs to be able to issue invalidation to ODP MRs
even if it cannot allocate memory. To this end it preallocates
emergency pages to use when the situation arises.

This flow should be extremely rare enough, that we don't need
to worry about contention and therefore a single emergency page
is good enough.

Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
---
 drivers/infiniband/hw/mlx5/main.c | 42 +++++++++++++++++++++++++++------------
 1 file changed, 29 insertions(+), 13 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d9474b95d8e5..bd565af30d66 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -92,6 +92,12 @@ static LIST_HEAD(mlx5_ib_dev_list);
  */
 static DEFINE_MUTEX(mlx5_ib_multiport_mutex);
 
+/* We can't use an array for xlt_emergency_page because dma_map_single
+ * doesn't work on kernel modules memory
+ */
+static unsigned long xlt_emergency_page;
+static struct mutex xlt_emergency_page_mutex;
+
 struct mlx5_ib_dev *mlx5_ib_get_ibdev_from_mpi(struct mlx5_ib_multiport_info *mpi)
 {
 	struct mlx5_ib_dev *dev;
@@ -1698,17 +1704,10 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	context->ibucontext.invalidate_range = &mlx5_ib_invalidate_range;
 #endif
 
-	context->upd_xlt_page = __get_free_page(GFP_KERNEL);
-	if (!context->upd_xlt_page) {
-		err = -ENOMEM;
-		goto out_uars;
-	}
-	mutex_init(&context->upd_xlt_page_mutex);
-
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain)) {
 		err = mlx5_ib_alloc_transport_domain(dev, &context->tdn);
 		if (err)
-			goto out_page;
+			goto out_uars;
 	}
 
 	INIT_LIST_HEAD(&context->vma_private_list);
@@ -1785,9 +1784,6 @@ out_td:
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
 		mlx5_ib_dealloc_transport_domain(dev, context->tdn);
 
-out_page:
-	free_page(context->upd_xlt_page);
-
 out_uars:
 	deallocate_uars(dev, context);
 
@@ -1813,7 +1809,6 @@ static int mlx5_ib_dealloc_ucontext(struct ib_ucontext *ibcontext)
 	if (MLX5_CAP_GEN(dev->mdev, log_max_transport_domain))
 		mlx5_ib_dealloc_transport_domain(dev, context->tdn);
 
-	free_page(context->upd_xlt_page);
 	deallocate_uars(dev, context);
 	kfree(bfregi->sys_pages);
 	kfree(bfregi->count);
@@ -5292,13 +5287,32 @@ static struct mlx5_interface mlx5_ib_interface = {
 	.protocol	= MLX5_INTERFACE_PROTOCOL_IB,
 };
 
+unsigned long mlx5_ib_get_xlt_emergency_page(void)
+{
+	mutex_lock(&xlt_emergency_page_mutex);
+	return xlt_emergency_page;
+}
+
+void mlx5_ib_put_xlt_emergency_page(void)
+{
+	mutex_unlock(&xlt_emergency_page_mutex);
+}
+
 static int __init mlx5_ib_init(void)
 {
 	int err;
 
+	xlt_emergency_page = __get_free_page(GFP_KERNEL);
+	if (!xlt_emergency_page)
+		return -ENOMEM;
+
+	mutex_init(&xlt_emergency_page_mutex);
+
 	mlx5_ib_event_wq = alloc_ordered_workqueue("mlx5_ib_event_wq", 0);
-	if (!mlx5_ib_event_wq)
+	if (!mlx5_ib_event_wq) {
+		free_page(xlt_emergency_page);
 		return -ENOMEM;
+	}
 
 	mlx5_ib_odp_init();
 
@@ -5311,6 +5325,8 @@ static void __exit mlx5_ib_cleanup(void)
 {
 	mlx5_unregister_interface(&mlx5_ib_interface);
 	destroy_workqueue(mlx5_ib_event_wq);
+	mutex_destroy(&xlt_emergency_page_mutex);
+	free_page(xlt_emergency_page);
 }
 
 module_init(mlx5_ib_init);
-- 
cgit v1.2.3


From 7672ed33c4c15dbe9d56880683baaba4227cf940 Mon Sep 17 00:00:00 2001
From: Honggang Li <honli@redhat.com>
Date: Fri, 16 Mar 2018 10:37:13 +0800
Subject: IB/mlx5: Set the default active rate and width to QDR and 4X

Before commit f1b65df5a232 ("IB/mlx5: Add support for active_width and
active_speed in RoCE"), the mlx5_ib driver set the default active_width
and active_speed to IB_WIDTH_4X and IB_SPEED_QDR.

When the RoCE port is down, the RoCE port does not negotiate the active
width with the remote side, causing the active width to be zero. When
running userspace ibstat to view the port status, ibstat will panic as it
reads an invalid width from sys file.

This patch restores the original behavior.

Fixes: f1b65df5a232 ("IB/mlx5: Add support for active_width and active_speed in RoCE").
Signed-off-by: Honggang Li <honli@redhat.com>
Reviewed-by: Hal Rosenstock <hal@mellanox.com>
Reviewed-by: Noa Osherovich <noaos@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c | 3 +++
 1 file changed, 3 insertions(+)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index a5a3f0b25608..3408bede0ee5 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -405,6 +405,9 @@ static int mlx5_query_port_roce(struct ib_device *device, u8 port_num,
 	if (err)
 		goto out;
 
+	props->active_width     = IB_WIDTH_4X;
+	props->active_speed     = IB_SPEED_QDR;
+
 	translate_eth_proto_oper(eth_prot_oper, &props->active_speed,
 				 &props->active_width);
 
-- 
cgit v1.2.3


From 61147f391a8b3bdde4c0a631dd132d85d00b90a0 Mon Sep 17 00:00:00 2001
From: Bodong Wang <bodong@mellanox.com>
Date: Mon, 19 Mar 2018 15:10:30 +0200
Subject: IB/mlx5: Packet packing enhancement for RAW QP

Enable RAW QP to be able to configure burst control by modify_qp. By
using burst control with rate limiting, user can achieve best
performance and accuracy. The burst control information is passed by
user through udata.

This patch also reports burst control capability for mlx5 related
hardwares, burst control is only marked as supported when both
packet_pacing_burst_bound and packet_pacing_typical_size are
supported.

Signed-off-by: Bodong Wang <bodong@mellanox.com>
Reviewed-by: Daniel Jurgens <danielj@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c    |  4 ++
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  2 +-
 drivers/infiniband/hw/mlx5/qp.c      | 94 ++++++++++++++++++++++++++++--------
 include/uapi/rdma/mlx5-abi.h         | 19 +++++++-
 4 files changed, 98 insertions(+), 21 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 3408bede0ee5..d06aae9aa600 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -989,6 +989,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 				MLX5_CAP_QOS(mdev, packet_pacing_min_rate);
 			resp.packet_pacing_caps.supported_qpts |=
 				1 << IB_QPT_RAW_PACKET;
+			if (MLX5_CAP_QOS(mdev, packet_pacing_burst_bound) &&
+			    MLX5_CAP_QOS(mdev, packet_pacing_typical_size))
+				resp.packet_pacing_caps.cap_flags |=
+					MLX5_IB_PP_SUPPORT_BURST;
 		}
 		resp.response_length += sizeof(resp.packet_pacing_caps);
 	}
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index f9ba1ea94f0f..aeea74357cbe 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -403,7 +403,7 @@ struct mlx5_ib_qp {
 	struct list_head	qps_list;
 	struct list_head	cq_recv_list;
 	struct list_head	cq_send_list;
-	u32			rate_limit;
+	struct mlx5_rate_limit	rl;
 	u32                     underlay_qpn;
 	bool			tunnel_offload_en;
 	/* storage for qp sub type when core qp type is IB_QPT_DRIVER */
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 6c7b4c2bfaa4..2fb3d9a400d3 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -86,7 +86,9 @@ struct mlx5_modify_raw_qp_param {
 	u16 operation;
 
 	u32 set_mask; /* raw_qp_set_mask_map */
-	u32 rate_limit;
+
+	struct mlx5_rate_limit rl;
+
 	u8 rq_q_ctr_id;
 };
 
@@ -2774,8 +2776,9 @@ static int modify_raw_packet_qp_sq(struct mlx5_core_dev *dev,
 				   const struct mlx5_modify_raw_qp_param *raw_qp_param)
 {
 	struct mlx5_ib_qp *ibqp = sq->base.container_mibqp;
-	u32 old_rate = ibqp->rate_limit;
-	u32 new_rate = old_rate;
+	struct mlx5_rate_limit old_rl = ibqp->rl;
+	struct mlx5_rate_limit new_rl = old_rl;
+	bool new_rate_added = false;
 	u16 rl_index = 0;
 	void *in;
 	void *sqc;
@@ -2797,39 +2800,43 @@ static int modify_raw_packet_qp_sq(struct mlx5_core_dev *dev,
 			pr_warn("%s: Rate limit can only be changed when SQ is moving to RDY\n",
 				__func__);
 		else
-			new_rate = raw_qp_param->rate_limit;
+			new_rl = raw_qp_param->rl;
 	}
 
-	if (old_rate != new_rate) {
-		if (new_rate) {
-			err = mlx5_rl_add_rate(dev, new_rate, &rl_index);
+	if (!mlx5_rl_are_equal(&old_rl, &new_rl)) {
+		if (new_rl.rate) {
+			err = mlx5_rl_add_rate(dev, &rl_index, &new_rl);
 			if (err) {
-				pr_err("Failed configuring rate %u: %d\n",
-				       new_rate, err);
+				pr_err("Failed configuring rate limit(err %d): \
+				       rate %u, max_burst_sz %u, typical_pkt_sz %u\n",
+				       err, new_rl.rate, new_rl.max_burst_sz,
+				       new_rl.typical_pkt_sz);
+
 				goto out;
 			}
+			new_rate_added = true;
 		}
 
 		MLX5_SET64(modify_sq_in, in, modify_bitmask, 1);
+		/* index 0 means no limit */
 		MLX5_SET(sqc, sqc, packet_pacing_rate_limit_index, rl_index);
 	}
 
 	err = mlx5_core_modify_sq(dev, sq->base.mqp.qpn, in, inlen);
 	if (err) {
 		/* Remove new rate from table if failed */
-		if (new_rate &&
-		    old_rate != new_rate)
-			mlx5_rl_remove_rate(dev, new_rate);
+		if (new_rate_added)
+			mlx5_rl_remove_rate(dev, &new_rl);
 		goto out;
 	}
 
 	/* Only remove the old rate after new rate was set */
-	if ((old_rate &&
-	    (old_rate != new_rate)) ||
+	if ((old_rl.rate &&
+	     !mlx5_rl_are_equal(&old_rl, &new_rl)) ||
 	    (new_state != MLX5_SQC_STATE_RDY))
-		mlx5_rl_remove_rate(dev, old_rate);
+		mlx5_rl_remove_rate(dev, &old_rl);
 
-	ibqp->rate_limit = new_rate;
+	ibqp->rl = new_rl;
 	sq->state = new_state;
 
 out:
@@ -2906,7 +2913,8 @@ static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 
 static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 			       const struct ib_qp_attr *attr, int attr_mask,
-			       enum ib_qp_state cur_state, enum ib_qp_state new_state)
+			       enum ib_qp_state cur_state, enum ib_qp_state new_state,
+			       const struct mlx5_ib_modify_qp *ucmd)
 {
 	static const u16 optab[MLX5_QP_NUM_STATE][MLX5_QP_NUM_STATE] = {
 		[MLX5_QP_STATE_RST] = {
@@ -3144,7 +3152,30 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
 		}
 
 		if (attr_mask & IB_QP_RATE_LIMIT) {
-			raw_qp_param.rate_limit = attr->rate_limit;
+			raw_qp_param.rl.rate = attr->rate_limit;
+
+			if (ucmd->burst_info.max_burst_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_burst_bound)) {
+					raw_qp_param.rl.max_burst_sz =
+						ucmd->burst_info.max_burst_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
+			if (ucmd->burst_info.typical_pkt_sz) {
+				if (attr->rate_limit &&
+				    MLX5_CAP_QOS(dev->mdev, packet_pacing_typical_size)) {
+					raw_qp_param.rl.typical_pkt_sz =
+						ucmd->burst_info.typical_pkt_sz;
+				} else {
+					err = -EINVAL;
+					goto out;
+				}
+			}
+
 			raw_qp_param.set_mask |= MLX5_RAW_QP_RATE_LIMIT;
 		}
 
@@ -3332,8 +3363,10 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
 	struct mlx5_ib_qp *qp = to_mqp(ibqp);
+	struct mlx5_ib_modify_qp ucmd = {};
 	enum ib_qp_type qp_type;
 	enum ib_qp_state cur_state, new_state;
+	size_t required_cmd_sz;
 	int err = -EINVAL;
 	int port;
 	enum rdma_link_layer ll = IB_LINK_LAYER_UNSPECIFIED;
@@ -3341,6 +3374,28 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	if (ibqp->rwq_ind_tbl)
 		return -ENOSYS;
 
+	if (udata && udata->inlen) {
+		required_cmd_sz = offsetof(typeof(ucmd), reserved) +
+			sizeof(ucmd.reserved);
+		if (udata->inlen < required_cmd_sz)
+			return -EINVAL;
+
+		if (udata->inlen > sizeof(ucmd) &&
+		    !ib_is_udata_cleared(udata, sizeof(ucmd),
+					 udata->inlen - sizeof(ucmd)))
+			return -EOPNOTSUPP;
+
+		if (ib_copy_from_udata(&ucmd, udata,
+				       min(udata->inlen, sizeof(ucmd))))
+			return -EFAULT;
+
+		if (ucmd.comp_mask ||
+		    memchr_inv(&ucmd.reserved, 0, sizeof(ucmd.reserved)) ||
+		    memchr_inv(&ucmd.burst_info.reserved, 0,
+			       sizeof(ucmd.burst_info.reserved)))
+			return -EOPNOTSUPP;
+	}
+
 	if (unlikely(ibqp->qp_type == IB_QPT_GSI))
 		return mlx5_ib_gsi_modify_qp(ibqp, attr, attr_mask);
 
@@ -3421,7 +3476,8 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		goto out;
 	}
 
-	err = __mlx5_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
+	err = __mlx5_ib_modify_qp(ibqp, attr, attr_mask, cur_state,
+				  new_state, &ucmd);
 
 out:
 	mutex_unlock(&qp->mutex);
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index 1111aa4e7c1e..d2e0d234704f 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -163,6 +163,10 @@ struct mlx5_ib_cqe_comp_caps {
 	__u32 supported_format; /* enum mlx5_ib_cqe_comp_res_format */
 };
 
+enum mlx5_ib_packet_pacing_cap_flags {
+	MLX5_IB_PP_SUPPORT_BURST	= 1 << 0,
+};
+
 struct mlx5_packet_pacing_caps {
 	__u32 qp_rate_limit_min;
 	__u32 qp_rate_limit_max; /* In kpbs */
@@ -172,7 +176,8 @@ struct mlx5_packet_pacing_caps {
 	 * supported_qpts |= 1 << IB_QPT_RAW_PACKET
 	 */
 	__u32 supported_qpts;
-	__u32 reserved;
+	__u8  cap_flags; /* enum mlx5_ib_packet_pacing_cap_flags */
+	__u8  reserved[3];
 };
 
 enum mlx5_ib_mpw_caps {
@@ -362,6 +367,18 @@ struct mlx5_ib_create_ah_resp {
 	__u8	reserved[6];
 };
 
+struct mlx5_ib_burst_info {
+	__u32       max_burst_sz;
+	__u16       typical_pkt_sz;
+	__u16       reserved;
+};
+
+struct mlx5_ib_modify_qp {
+	__u32			   comp_mask;
+	struct mlx5_ib_burst_info  burst_info;
+	__u32			   reserved;
+};
+
 struct mlx5_ib_modify_qp_resp {
 	__u32	response_length;
 	__u32	dctn;
-- 
cgit v1.2.3


From 0ede73bc012c98fba244b33efbc42e48dd23ee9a Mon Sep 17 00:00:00 2001
From: Matan Barak <matanb@mellanox.com>
Date: Mon, 19 Mar 2018 15:02:34 +0200
Subject: IB/uverbs: Extend uverbs_ioctl header with driver_id

Extending uverbs_ioctl header with driver_id and another reserved
field. driver_id should be used in order to identify the driver.
Since every driver could have its own parsing tree, this is necessary
for strace support.
Downstream patches take off the EXPERIMENTAL flag from the ioctl() IB
support and thus we add some reserved fields for future usage.

Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/core/uverbs_ioctl.c         |  5 ++++-
 drivers/infiniband/hw/bnxt_re/main.c           |  1 +
 drivers/infiniband/hw/cxgb3/iwch_provider.c    |  1 +
 drivers/infiniband/hw/cxgb4/provider.c         |  1 +
 drivers/infiniband/hw/hfi1/verbs.c             |  2 +-
 drivers/infiniband/hw/hns/hns_roce_main.c      |  1 +
 drivers/infiniband/hw/i40iw/i40iw_verbs.c      |  1 +
 drivers/infiniband/hw/mlx4/main.c              |  1 +
 drivers/infiniband/hw/mlx5/main.c              |  1 +
 drivers/infiniband/hw/mthca/mthca_provider.c   |  1 +
 drivers/infiniband/hw/nes/nes_verbs.c          |  1 +
 drivers/infiniband/hw/ocrdma/ocrdma_main.c     |  1 +
 drivers/infiniband/hw/qedr/main.c              |  1 +
 drivers/infiniband/hw/qib/qib_verbs.c          |  2 +-
 drivers/infiniband/hw/usnic/usnic_ib_main.c    |  1 +
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c |  1 +
 drivers/infiniband/sw/rdmavt/vt.c              |  3 ++-
 drivers/infiniband/sw/rxe/rxe_verbs.c          |  1 +
 include/rdma/ib_verbs.h                        |  2 ++
 include/rdma/rdma_vt.h                         |  2 +-
 include/uapi/rdma/rdma_user_ioctl_cmds.h       | 24 +++++++++++++++++++++++-
 21 files changed, 48 insertions(+), 6 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/core/uverbs_ioctl.c b/drivers/infiniband/core/uverbs_ioctl.c
index 339b85145044..7016e729f139 100644
--- a/drivers/infiniband/core/uverbs_ioctl.c
+++ b/drivers/infiniband/core/uverbs_ioctl.c
@@ -246,6 +246,9 @@ static long ib_uverbs_cmd_verbs(struct ib_device *ib_dev,
 	size_t ctx_size;
 	uintptr_t data[UVERBS_OPTIMIZE_USING_STACK_SZ / sizeof(uintptr_t)];
 
+	if (hdr->driver_id != ib_dev->driver_id)
+		return -EINVAL;
+
 	object_spec = uverbs_get_object(ib_dev, hdr->object_id);
 	if (!object_spec)
 		return -EPROTONOSUPPORT;
@@ -350,7 +353,7 @@ long ib_uverbs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 			goto out;
 		}
 
-		if (hdr.reserved) {
+		if (hdr.reserved1 || hdr.reserved2) {
 			err = -EPROTONOSUPPORT;
 			goto out;
 		}
diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
index f6e361750466..abe0be8b5ddc 100644
--- a/drivers/infiniband/hw/bnxt_re/main.c
+++ b/drivers/infiniband/hw/bnxt_re/main.c
@@ -619,6 +619,7 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	ibdev->get_hw_stats             = bnxt_re_ib_get_hw_stats;
 	ibdev->alloc_hw_stats           = bnxt_re_ib_alloc_hw_stats;
 
+	ibdev->driver_id = RDMA_DRIVER_BNXT_RE;
 	return ib_register_device(ibdev, NULL);
 }
 
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c
index a578ca559e11..1804b6c4a6ec 100644
--- a/drivers/infiniband/hw/cxgb3/iwch_provider.c
+++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c
@@ -1439,6 +1439,7 @@ int iwch_register_device(struct iwch_dev *dev)
 	memcpy(dev->ibdev.iwcm->ifname, dev->rdev.t3cdev_p->lldev->name,
 	       sizeof(dev->ibdev.iwcm->ifname));
 
+	dev->ibdev.driver_id = RDMA_DRIVER_CXGB3;
 	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto bail1;
diff --git a/drivers/infiniband/hw/cxgb4/provider.c b/drivers/infiniband/hw/cxgb4/provider.c
index 42568a4df3f8..dc4eabd85f54 100644
--- a/drivers/infiniband/hw/cxgb4/provider.c
+++ b/drivers/infiniband/hw/cxgb4/provider.c
@@ -629,6 +629,7 @@ void c4iw_register_device(struct work_struct *work)
 	memcpy(dev->ibdev.iwcm->ifname, dev->rdev.lldi.ports[0]->name,
 	       sizeof(dev->ibdev.iwcm->ifname));
 
+	dev->ibdev.driver_id = RDMA_DRIVER_CXGB4;
 	ret = ib_register_device(&dev->ibdev, NULL);
 	if (ret)
 		goto err_kfree_iwcm;
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 471d55c50066..c8cf4d4984d3 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -1960,7 +1960,7 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 			      i,
 			      ppd->pkeys);
 
-	ret = rvt_register_device(&dd->verbs_dev.rdi);
+	ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_HFI1);
 	if (ret)
 		goto err_verbs_txreq;
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 6e48b1f507cf..83e21f696bbc 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -526,6 +526,7 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	/* OTHERS */
 	ib_dev->get_port_immutable	= hns_roce_port_immutable;
 
+	ib_dev->driver_id = RDMA_DRIVER_HNS;
 	ret = ib_register_device(ib_dev, NULL);
 	if (ret) {
 		dev_err(dev, "ib_register_device failed!\n");
diff --git a/drivers/infiniband/hw/i40iw/i40iw_verbs.c b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
index 60e004d2100e..40e4f5ab2b46 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_verbs.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_verbs.c
@@ -2927,6 +2927,7 @@ int i40iw_register_rdma_device(struct i40iw_device *iwdev)
 		return -ENOMEM;
 	iwibdev = iwdev->iwibdev;
 
+	iwibdev->ibdev.driver_id = RDMA_DRIVER_I40IW;
 	ret = ib_register_device(&iwibdev->ibdev, NULL);
 	if (ret)
 		goto error;
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index b9befda1eb27..d1be3231f4f0 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -2955,6 +2955,7 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	if (mlx4_ib_alloc_diag_counters(ibdev))
 		goto err_steer_free_bitmap;
 
+	ibdev->ib_dev.driver_id = RDMA_DRIVER_MLX4;
 	if (ib_register_device(&ibdev->ib_dev, NULL))
 		goto err_diag_counters;
 
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index d06aae9aa600..6b50711df786 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -4779,6 +4779,7 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 	dev->ib_dev.uverbs_ex_cmd_mask |=
 			(1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) |
 			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
+	dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
 
 	err = init_node_data(dev);
 	if (err)
diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c
index 6fee7795d1c8..541f237965c7 100644
--- a/drivers/infiniband/hw/mthca/mthca_provider.c
+++ b/drivers/infiniband/hw/mthca/mthca_provider.c
@@ -1295,6 +1295,7 @@ int mthca_register_device(struct mthca_dev *dev)
 
 	mutex_init(&dev->cap_mask_mutex);
 
+	dev->ib_dev.driver_id = RDMA_DRIVER_MTHCA;
 	ret = ib_register_device(&dev->ib_dev, NULL);
 	if (ret)
 		return ret;
diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c
index 162475aeeedd..1040a6e34230 100644
--- a/drivers/infiniband/hw/nes/nes_verbs.c
+++ b/drivers/infiniband/hw/nes/nes_verbs.c
@@ -3854,6 +3854,7 @@ int nes_register_ofa_device(struct nes_ib_device *nesibdev)
 	struct nes_adapter *nesadapter = nesdev->nesadapter;
 	int i, ret;
 
+	nesvnic->nesibdev->ibdev.driver_id = RDMA_DRIVER_NES;
 	ret = ib_register_device(&nesvnic->nesibdev->ibdev, NULL);
 	if (ret) {
 		return ret;
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_main.c b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
index 42dc0de54cb8..4547aa28d4ae 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_main.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_main.c
@@ -215,6 +215,7 @@ static int ocrdma_register_device(struct ocrdma_dev *dev)
 		dev->ibdev.destroy_srq = ocrdma_destroy_srq;
 		dev->ibdev.post_srq_recv = ocrdma_post_srq_recv;
 	}
+	dev->ibdev.driver_id = RDMA_DRIVER_OCRDMA;
 	return ib_register_device(&dev->ibdev, NULL);
 }
 
diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index db4bf97c0e15..f865c0991ad9 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -257,6 +257,7 @@ static int qedr_register_device(struct qedr_dev *dev)
 	dev->ibdev.get_link_layer = qedr_link_layer;
 	dev->ibdev.get_dev_fw_str = qedr_get_dev_fw_str;
 
+	dev->ibdev.driver_id = RDMA_DRIVER_QEDR;
 	return ib_register_device(&dev->ibdev, NULL);
 }
 
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index fabee760407e..3977abbc83ad 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -1646,7 +1646,7 @@ int qib_register_ib_device(struct qib_devdata *dd)
 			      dd->rcd[ctxt]->pkeys);
 	}
 
-	ret = rvt_register_device(&dd->verbs_dev.rdi);
+	ret = rvt_register_device(&dd->verbs_dev.rdi, RDMA_DRIVER_QIB);
 	if (ret)
 		goto err_tx;
 
diff --git a/drivers/infiniband/hw/usnic/usnic_ib_main.c b/drivers/infiniband/hw/usnic/usnic_ib_main.c
index f45e99a938e0..aed1ca390e30 100644
--- a/drivers/infiniband/hw/usnic/usnic_ib_main.c
+++ b/drivers/infiniband/hw/usnic/usnic_ib_main.c
@@ -433,6 +433,7 @@ static void *usnic_ib_device_add(struct pci_dev *dev)
 	us_ibdev->ib_dev.get_dev_fw_str     = usnic_get_dev_fw_str;
 
 
+	us_ibdev->ib_dev.driver_id = RDMA_DRIVER_USNIC;
 	if (ib_register_device(&us_ibdev->ib_dev, NULL))
 		goto err_fwd_dealloc;
 
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index d650a9fcde24..4834460e2a0b 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -276,6 +276,7 @@ static int pvrdma_register_device(struct pvrdma_dev *dev)
 		if (!dev->srq_tbl)
 			goto err_qp_free;
 	}
+	dev->ib_dev.driver_id = RDMA_DRIVER_VMW_PVRDMA;
 	spin_lock_init(&dev->srq_tbl_lock);
 
 	ret = ib_register_device(&dev->ib_dev, NULL);
diff --git a/drivers/infiniband/sw/rdmavt/vt.c b/drivers/infiniband/sw/rdmavt/vt.c
index a67b0ddc2230..434199d0bc96 100644
--- a/drivers/infiniband/sw/rdmavt/vt.c
+++ b/drivers/infiniband/sw/rdmavt/vt.c
@@ -730,7 +730,7 @@ static noinline int check_support(struct rvt_dev_info *rdi, int verb)
  *
  * Return: 0 on success otherwise an errno.
  */
-int rvt_register_device(struct rvt_dev_info *rdi)
+int rvt_register_device(struct rvt_dev_info *rdi, u32 driver_id)
 {
 	int ret = 0, i;
 
@@ -831,6 +831,7 @@ int rvt_register_device(struct rvt_dev_info *rdi)
 	rdi->ibdev.node_type = RDMA_NODE_IB_CA;
 	rdi->ibdev.num_comp_vectors = 1;
 
+	rdi->ibdev.driver_id = driver_id;
 	/* We are now good to announce we exist */
 	ret =  ib_register_device(&rdi->ibdev, rdi->driver_f.port_callback);
 	if (ret) {
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index ced79e49234b..5ef8c3333e43 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -1335,6 +1335,7 @@ int rxe_register_device(struct rxe_dev *rxe)
 	}
 	rxe->tfm = tfm;
 
+	dev->driver_id = RDMA_DRIVER_RXE;
 	err = ib_register_device(dev, NULL);
 	if (err) {
 		pr_warn("%s failed with error %d\n", __func__, err);
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 3cc48f34e3e4..2357f2b29610 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -64,6 +64,7 @@
 #include <linux/cgroup_rdma.h>
 #include <uapi/rdma/ib_user_verbs.h>
 #include <rdma/restrack.h>
+#include <uapi/rdma/rdma_user_ioctl.h>
 
 #define IB_FW_VERSION_NAME_MAX	ETHTOOL_FWVERS_LEN
 
@@ -2385,6 +2386,7 @@ struct ib_device {
 						     int comp_vector);
 
 	struct uverbs_root_spec		*specs_root;
+	enum rdma_driver_id		driver_id;
 };
 
 struct ib_client {
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
index 4118324a0310..3f4c187e435d 100644
--- a/include/rdma/rdma_vt.h
+++ b/include/rdma/rdma_vt.h
@@ -538,7 +538,7 @@ static inline void rvt_mod_retry_timer(struct rvt_qp *qp)
 
 struct rvt_dev_info *rvt_alloc_device(size_t size, int nports);
 void rvt_dealloc_device(struct rvt_dev_info *rdi);
-int rvt_register_device(struct rvt_dev_info *rvd);
+int rvt_register_device(struct rvt_dev_info *rvd, u32 driver_id);
 void rvt_unregister_device(struct rvt_dev_info *rvd);
 int rvt_check_ah(struct ib_device *ibdev, struct rdma_ah_attr *ah_attr);
 int rvt_init_port(struct rvt_dev_info *rdi, struct rvt_ibport *port,
diff --git a/include/uapi/rdma/rdma_user_ioctl_cmds.h b/include/uapi/rdma/rdma_user_ioctl_cmds.h
index aa1fffe3620b..40063cf970aa 100644
--- a/include/uapi/rdma/rdma_user_ioctl_cmds.h
+++ b/include/uapi/rdma/rdma_user_ioctl_cmds.h
@@ -64,8 +64,30 @@ struct ib_uverbs_ioctl_hdr {
 	__u16 object_id;
 	__u16 method_id;
 	__u16 num_attrs;
-	__aligned_u64 reserved;
+	__aligned_u64 reserved1;
+	__u32 driver_id;
+	__u32 reserved2;
 	struct ib_uverbs_attr  attrs[0];
 };
 
+enum rdma_driver_id {
+	RDMA_DRIVER_UNKNOWN,
+	RDMA_DRIVER_MLX5,
+	RDMA_DRIVER_MLX4,
+	RDMA_DRIVER_CXGB3,
+	RDMA_DRIVER_CXGB4,
+	RDMA_DRIVER_MTHCA,
+	RDMA_DRIVER_BNXT_RE,
+	RDMA_DRIVER_OCRDMA,
+	RDMA_DRIVER_NES,
+	RDMA_DRIVER_I40IW,
+	RDMA_DRIVER_VMW_PVRDMA,
+	RDMA_DRIVER_QEDR,
+	RDMA_DRIVER_HNS,
+	RDMA_DRIVER_USNIC,
+	RDMA_DRIVER_RXE,
+	RDMA_DRIVER_HFI1,
+	RDMA_DRIVER_QIB,
+};
+
 #endif
-- 
cgit v1.2.3


From 3e44e0ee0893cbea257e585dbd4c7d2ff00f1a6b Mon Sep 17 00:00:00 2001
From: Parav Pandit <parav@mellanox.com>
Date: Sun, 1 Apr 2018 15:08:23 +0300
Subject: IB/providers: Avoid null netdev check for RoCE

Now that IB core GID cache ensures that all RoCE entries have an
associated netdev remove null checks from the provider drivers for
clarity.

Reviewed-by: Mark Bloch <markb@mellanox.com>
Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/bnxt_re/ib_verbs.c  |  5 +--
 drivers/infiniband/hw/hns/hns_roce_ah.c   |  8 ++--
 drivers/infiniband/hw/mlx4/ah.c           | 10 ++---
 drivers/infiniband/hw/mlx4/qp.c           |  2 +-
 drivers/infiniband/hw/mlx5/main.c         |  6 ---
 drivers/infiniband/hw/ocrdma/ocrdma_ah.c  |  8 ++--
 drivers/infiniband/hw/ocrdma/ocrdma_hw.c  |  2 +-
 drivers/infiniband/hw/qedr/qedr_roce_cm.c | 10 ++---
 drivers/infiniband/hw/qedr/verbs.c        | 72 +++++++++++++++----------------
 drivers/infiniband/sw/rxe/rxe_verbs.c     |  4 +-
 10 files changed, 54 insertions(+), 73 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 944eb0acbbac..63a0e08dd6fe 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -718,8 +718,7 @@ struct ib_ah *bnxt_re_create_ah(struct ib_pd *ib_pd,
 				grh->sgid_index);
 			goto fail;
 		}
-		if (sgid_attr.ndev)
-			dev_put(sgid_attr.ndev);
+		dev_put(sgid_attr.ndev);
 		/* Get network header type for this GID */
 		nw_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid);
 		switch (nw_type) {
@@ -1697,7 +1696,7 @@ int bnxt_re_modify_qp(struct ib_qp *ib_qp, struct ib_qp_attr *qp_attr,
 		status = ib_get_cached_gid(&rdev->ibdev, 1,
 					   grh->sgid_index,
 					   &sgid, &sgid_attr);
-		if (!status && sgid_attr.ndev) {
+		if (!status) {
 			memcpy(qp->qplib_qp.smac, sgid_attr.ndev->dev_addr,
 			       ETH_ALEN);
 			dev_put(sgid_attr.ndev);
diff --git a/drivers/infiniband/hw/hns/hns_roce_ah.c b/drivers/infiniband/hw/hns/hns_roce_ah.c
index 7dd6a66ea244..d74928621559 100644
--- a/drivers/infiniband/hw/hns/hns_roce_ah.c
+++ b/drivers/infiniband/hw/hns/hns_roce_ah.c
@@ -68,11 +68,9 @@ struct ib_ah *hns_roce_create_ah(struct ib_pd *ibpd,
 		return ERR_PTR(ret);
 	}
 
-	if (gid_attr.ndev) {
-		if (is_vlan_dev(gid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
-		dev_put(gid_attr.ndev);
-	}
+	if (is_vlan_dev(gid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
+	dev_put(gid_attr.ndev);
 
 	if (vlan_tag < 0x1000)
 		vlan_tag |= (rdma_ah_get_sl(ah_attr) &
diff --git a/drivers/infiniband/hw/mlx4/ah.c b/drivers/infiniband/hw/mlx4/ah.c
index 6dee4fdc5d67..9345d5b546d1 100644
--- a/drivers/infiniband/hw/mlx4/ah.c
+++ b/drivers/infiniband/hw/mlx4/ah.c
@@ -101,12 +101,10 @@ static struct ib_ah *create_iboe_ah(struct ib_pd *pd,
 	if (ret)
 		return ERR_PTR(ret);
 	eth_zero_addr(ah->av.eth.s_mac);
-	if (gid_attr.ndev) {
-		if (is_vlan_dev(gid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
-		memcpy(ah->av.eth.s_mac, gid_attr.ndev->dev_addr, ETH_ALEN);
-		dev_put(gid_attr.ndev);
-	}
+	if (is_vlan_dev(gid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(gid_attr.ndev);
+	memcpy(ah->av.eth.s_mac, gid_attr.ndev->dev_addr, ETH_ALEN);
+	dev_put(gid_attr.ndev);
 	if (vlan_tag < 0x1000)
 		vlan_tag |= (rdma_ah_get_sl(ah_attr) & 7) << 13;
 	ah->av.eth.port_pd = cpu_to_be32(to_mpd(pd)->pdn |
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 726a6ae90de0..50af8915e7ec 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -2370,7 +2370,7 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type,
 
 			status = ib_get_cached_gid(&dev->ib_dev, port_num,
 						   index, &gid, &gid_attr);
-			if (!status && gid_attr.ndev) {
+			if (!status) {
 				vlan = rdma_vlan_dev_vlan_id(gid_attr.ndev);
 				memcpy(smac, gid_attr.ndev->dev_addr, ETH_ALEN);
 				dev_put(gid_attr.ndev);
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 6b50711df786..9e6780eadd1e 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -525,9 +525,6 @@ __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
 	if (ib_get_cached_gid(&dev->ib_dev, port_num, index, &gid, &attr))
 		return 0;
 
-	if (!attr.ndev)
-		return 0;
-
 	dev_put(attr.ndev);
 
 	if (attr.gid_type != IB_GID_TYPE_ROCE_UDP_ENCAP)
@@ -547,9 +544,6 @@ int mlx5_get_roce_gid_type(struct mlx5_ib_dev *dev, u8 port_num,
 	if (ret)
 		return ret;
 
-	if (!attr.ndev)
-		return -ENODEV;
-
 	dev_put(attr.ndev);
 
 	*gid_type = attr.gid_type;
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c
index dec650930ca6..3897b64532e1 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_ah.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_ah.c
@@ -193,11 +193,9 @@ struct ib_ah *ocrdma_create_ah(struct ib_pd *ibpd, struct rdma_ah_attr *attr,
 		      __func__, status);
 		goto av_conf_err;
 	}
-	if (sgid_attr.ndev) {
-		if (is_vlan_dev(sgid_attr.ndev))
-			vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev);
-		dev_put(sgid_attr.ndev);
-	}
+	if (is_vlan_dev(sgid_attr.ndev))
+		vlan_tag = vlan_dev_vlan_id(sgid_attr.ndev);
+	dev_put(sgid_attr.ndev);
 	/* Get network header type for this GID */
 	ah->hdr_type = ib_gid_to_network_type(sgid_attr.gid_type, &sgid);
 
diff --git a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
index 12783262eb75..2c260e1c29d1 100644
--- a/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
+++ b/drivers/infiniband/hw/ocrdma/ocrdma_hw.c
@@ -2527,7 +2527,7 @@ static int ocrdma_set_av_params(struct ocrdma_qp *qp,
 
 	status = ib_get_cached_gid(&dev->ibdev, 1, grh->sgid_index,
 				   &sgid, &sgid_attr);
-	if (!status && sgid_attr.ndev) {
+	if (!status) {
 		vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
 		memcpy(mac_addr, sgid_attr.ndev->dev_addr, ETH_ALEN);
 		dev_put(sgid_attr.ndev);
diff --git a/drivers/infiniband/hw/qedr/qedr_roce_cm.c b/drivers/infiniband/hw/qedr/qedr_roce_cm.c
index eedb937d6fa9..0f14e687bb91 100644
--- a/drivers/infiniband/hw/qedr/qedr_roce_cm.c
+++ b/drivers/infiniband/hw/qedr/qedr_roce_cm.c
@@ -412,13 +412,11 @@ static inline int qedr_gsi_build_header(struct qedr_dev *dev,
 		return rc;
 	}
 
-	if (sgid_attr.ndev) {
-		vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
-		if (vlan_id < VLAN_CFI_MASK)
-			has_vlan = true;
+	vlan_id = rdma_vlan_dev_vlan_id(sgid_attr.ndev);
+	if (vlan_id < VLAN_CFI_MASK)
+		has_vlan = true;
 
-		dev_put(sgid_attr.ndev);
-	}
+	dev_put(sgid_attr.ndev);
 
 	has_udp = (sgid_attr.gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP);
 	if (!has_udp) {
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index a523d6f5fef3..a9f494fb892a 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -1093,43 +1093,41 @@ static inline int get_gid_info_from_table(struct ib_qp *ibqp,
 	if (rc)
 		return rc;
 
-	if (gid_attr.ndev) {
-		qp_params->vlan_id = rdma_vlan_dev_vlan_id(gid_attr.ndev);
-
-		dev_put(gid_attr.ndev);
-		nw_type = ib_gid_to_network_type(gid_attr.gid_type, &gid);
-		switch (nw_type) {
-		case RDMA_NETWORK_IPV6:
-			memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
-			       sizeof(qp_params->sgid));
-			memcpy(&qp_params->dgid.bytes[0],
-			       &grh->dgid,
-			       sizeof(qp_params->dgid));
-			qp_params->roce_mode = ROCE_V2_IPV6;
-			SET_FIELD(qp_params->modify_flags,
-				  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
-			break;
-		case RDMA_NETWORK_IB:
-			memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
-			       sizeof(qp_params->sgid));
-			memcpy(&qp_params->dgid.bytes[0],
-			       &grh->dgid,
-			       sizeof(qp_params->dgid));
-			qp_params->roce_mode = ROCE_V1;
-			break;
-		case RDMA_NETWORK_IPV4:
-			memset(&qp_params->sgid, 0, sizeof(qp_params->sgid));
-			memset(&qp_params->dgid, 0, sizeof(qp_params->dgid));
-			ipv4_addr = qedr_get_ipv4_from_gid(gid.raw);
-			qp_params->sgid.ipv4_addr = ipv4_addr;
-			ipv4_addr =
-			    qedr_get_ipv4_from_gid(grh->dgid.raw);
-			qp_params->dgid.ipv4_addr = ipv4_addr;
-			SET_FIELD(qp_params->modify_flags,
-				  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
-			qp_params->roce_mode = ROCE_V2_IPV4;
-			break;
-		}
+	qp_params->vlan_id = rdma_vlan_dev_vlan_id(gid_attr.ndev);
+
+	dev_put(gid_attr.ndev);
+	nw_type = ib_gid_to_network_type(gid_attr.gid_type, &gid);
+	switch (nw_type) {
+	case RDMA_NETWORK_IPV6:
+		memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
+		       sizeof(qp_params->sgid));
+		memcpy(&qp_params->dgid.bytes[0],
+		       &grh->dgid,
+		       sizeof(qp_params->dgid));
+		qp_params->roce_mode = ROCE_V2_IPV6;
+		SET_FIELD(qp_params->modify_flags,
+			  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
+		break;
+	case RDMA_NETWORK_IB:
+		memcpy(&qp_params->sgid.bytes[0], &gid.raw[0],
+		       sizeof(qp_params->sgid));
+		memcpy(&qp_params->dgid.bytes[0],
+		       &grh->dgid,
+		       sizeof(qp_params->dgid));
+		qp_params->roce_mode = ROCE_V1;
+		break;
+	case RDMA_NETWORK_IPV4:
+		memset(&qp_params->sgid, 0, sizeof(qp_params->sgid));
+		memset(&qp_params->dgid, 0, sizeof(qp_params->dgid));
+		ipv4_addr = qedr_get_ipv4_from_gid(gid.raw);
+		qp_params->sgid.ipv4_addr = ipv4_addr;
+		ipv4_addr =
+		    qedr_get_ipv4_from_gid(grh->dgid.raw);
+		qp_params->dgid.ipv4_addr = ipv4_addr;
+		SET_FIELD(qp_params->modify_flags,
+			  QED_ROCE_MODIFY_QP_VALID_ROCE_MODE, 1);
+		qp_params->roce_mode = ROCE_V2_IPV4;
+		break;
 	}
 
 	for (i = 0; i < 4; i++) {
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index f83bbf550ec0..0661c2783b14 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -256,9 +256,7 @@ static int rxe_init_av(struct rxe_dev *rxe, struct rdma_ah_attr *attr,
 
 	rxe_av_from_attr(rdma_ah_get_port_num(attr), av, attr);
 	rxe_av_fill_ip_info(av, attr, &sgid_attr, &sgid);
-
-	if (sgid_attr.ndev)
-		dev_put(sgid_attr.ndev);
+	dev_put(sgid_attr.ndev);
 	return 0;
 }
 
-- 
cgit v1.2.3


From 414448d249d82c9be93b35e61e0303e84ef2f959 Mon Sep 17 00:00:00 2001
From: Parav Pandit <parav@mellanox.com>
Date: Sun, 1 Apr 2018 15:08:24 +0300
Subject: RDMA: Use ib_gid_attr during GID modification

Now that ib_gid_attr contains device, port and index, simplify the
provider APIs add_gid() and del_gid() to use device, port and index
fields from the ib_gid_attr attributes structure.

Signed-off-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/core/cache.c                |  5 ++--
 drivers/infiniband/hw/bnxt_re/ib_verbs.c       | 10 ++++----
 drivers/infiniband/hw/bnxt_re/ib_verbs.h       |  6 ++---
 drivers/infiniband/hw/hns/hns_roce_main.c      | 20 +++++++---------
 drivers/infiniband/hw/mlx4/main.c              | 30 ++++++++++-------------
 drivers/infiniband/hw/mlx5/main.c              | 13 +++++-----
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c | 31 +++++++-----------------
 drivers/infiniband/sw/rxe/rxe_verbs.c          | 10 ++++----
 include/rdma/ib_verbs.h                        | 33 +++++++++++---------------
 9 files changed, 63 insertions(+), 95 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c
index 045ca11fa135..e337b08de2ff 100644
--- a/drivers/infiniband/core/cache.c
+++ b/drivers/infiniband/core/cache.c
@@ -157,7 +157,7 @@ static void del_roce_gid(struct ib_device *device, u8 port_num,
 		 table->data_vec[ix].gid.raw);
 
 	if (rdma_cap_roce_gid_table(device, port_num))
-		device->del_gid(device, port_num, ix,
+		device->del_gid(&table->data_vec[ix].attr,
 				&table->data_vec[ix].context);
 	dev_put(table->data_vec[ix].attr.ndev);
 }
@@ -186,8 +186,7 @@ static int add_roce_gid(struct ib_gid_table *table,
 	}
 
 	if (rdma_cap_roce_gid_table(attr->device, attr->port_num)) {
-		ret = attr->device->add_gid(attr->device, attr->port_num,
-					    ix, gid, attr, &entry->context);
+		ret = attr->device->add_gid(gid, attr, &entry->context);
 		if (ret) {
 			pr_err("%s GID add failed device=%s port=%d index=%d\n",
 			       __func__, attr->device->name, attr->port_num,
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.c b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
index 63a0e08dd6fe..a76e206704d4 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.c
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.c
@@ -314,12 +314,11 @@ int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
 	return rc;
 }
 
-int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, void **context)
+int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context)
 {
 	int rc = 0;
 	struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
 	struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
 	struct bnxt_qplib_gid *gid_to_del;
 
@@ -365,15 +364,14 @@ int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
 	return rc;
 }
 
-int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, const union ib_gid *gid,
+int bnxt_re_add_gid(const union ib_gid *gid,
 		    const struct ib_gid_attr *attr, void **context)
 {
 	int rc;
 	u32 tbl_idx = 0;
 	u16 vlan_id = 0xFFFF;
 	struct bnxt_re_gid_ctx *ctx, **ctx_tbl;
-	struct bnxt_re_dev *rdev = to_bnxt_re_dev(ibdev, ibdev);
+	struct bnxt_re_dev *rdev = to_bnxt_re_dev(attr->device, ibdev);
 	struct bnxt_qplib_sgid_tbl *sgid_tbl = &rdev->qplib_res.sgid_tbl;
 
 	if ((attr->ndev) && is_vlan_dev(attr->ndev))
diff --git a/drivers/infiniband/hw/bnxt_re/ib_verbs.h b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
index e62b7c2c7da6..5c6414cad4af 100644
--- a/drivers/infiniband/hw/bnxt_re/ib_verbs.h
+++ b/drivers/infiniband/hw/bnxt_re/ib_verbs.h
@@ -157,10 +157,8 @@ int bnxt_re_get_port_immutable(struct ib_device *ibdev, u8 port_num,
 void bnxt_re_query_fw_str(struct ib_device *ibdev, char *str);
 int bnxt_re_query_pkey(struct ib_device *ibdev, u8 port_num,
 		       u16 index, u16 *pkey);
-int bnxt_re_del_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, void **context);
-int bnxt_re_add_gid(struct ib_device *ibdev, u8 port_num,
-		    unsigned int index, const union ib_gid *gid,
+int bnxt_re_del_gid(const struct ib_gid_attr *attr, void **context);
+int bnxt_re_add_gid(const union ib_gid *gid,
 		    const struct ib_gid_attr *attr, void **context);
 int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
 		      int index, union ib_gid *gid);
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 76e2e5b41895..9d48bc07a9e6 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -74,12 +74,11 @@ static int hns_roce_set_mac(struct hns_roce_dev *hr_dev, u8 port, u8 *addr)
 	return hr_dev->hw->set_mac(hr_dev, phy_port, addr);
 }
 
-static int hns_roce_add_gid(struct ib_device *device, u8 port_num,
-			    unsigned int index, const union ib_gid *gid,
+static int hns_roce_add_gid(const union ib_gid *gid,
 			    const struct ib_gid_attr *attr, void **context)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(device);
-	u8 port = port_num - 1;
+	struct hns_roce_dev *hr_dev = to_hr_dev(attr->device);
+	u8 port = attr->port_num - 1;
 	unsigned long flags;
 	int ret;
 
@@ -88,21 +87,20 @@ static int hns_roce_add_gid(struct ib_device *device, u8 port_num,
 
 	spin_lock_irqsave(&hr_dev->iboe.lock, flags);
 
-	ret = hr_dev->hw->set_gid(hr_dev, port, index, (union ib_gid *)gid,
-				   attr);
+	ret = hr_dev->hw->set_gid(hr_dev, port, attr->index,
+				  (union ib_gid *)gid, attr);
 
 	spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
 
 	return ret;
 }
 
-static int hns_roce_del_gid(struct ib_device *device, u8 port_num,
-			    unsigned int index, void **context)
+static int hns_roce_del_gid(const struct ib_gid_attr *attr, void **context)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(device);
+	struct hns_roce_dev *hr_dev = to_hr_dev(attr->device);
 	struct ib_gid_attr zattr = { };
 	union ib_gid zgid = { {0} };
-	u8 port = port_num - 1;
+	u8 port = attr->port_num - 1;
 	unsigned long flags;
 	int ret;
 
@@ -111,7 +109,7 @@ static int hns_roce_del_gid(struct ib_device *device, u8 port_num,
 
 	spin_lock_irqsave(&hr_dev->iboe.lock, flags);
 
-	ret = hr_dev->hw->set_gid(hr_dev, port, index, &zgid, &zattr);
+	ret = hr_dev->hw->set_gid(hr_dev, port, attr->index, &zgid, &zattr);
 
 	spin_unlock_irqrestore(&hr_dev->iboe.lock, flags);
 
diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index a31a3edfbf28..8eca09b53fe8 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -246,14 +246,11 @@ static int mlx4_ib_update_gids(struct gid_entry *gids,
 	return mlx4_ib_update_gids_v1(gids, ibdev, port_num);
 }
 
-static int mlx4_ib_add_gid(struct ib_device *device,
-			   u8 port_num,
-			   unsigned int index,
-			   const union ib_gid *gid,
+static int mlx4_ib_add_gid(const union ib_gid *gid,
 			   const struct ib_gid_attr *attr,
 			   void **context)
 {
-	struct mlx4_ib_dev *ibdev = to_mdev(device);
+	struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
 	struct mlx4_ib_iboe *iboe = &ibdev->iboe;
 	struct mlx4_port_gid_table   *port_gid_table;
 	int free = -1, found = -1;
@@ -262,16 +259,16 @@ static int mlx4_ib_add_gid(struct ib_device *device,
 	int i;
 	struct gid_entry *gids = NULL;
 
-	if (!rdma_cap_roce_gid_table(device, port_num))
+	if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
 		return -EINVAL;
 
-	if (port_num > MLX4_MAX_PORTS)
+	if (attr->port_num > MLX4_MAX_PORTS)
 		return -EINVAL;
 
 	if (!context)
 		return -EINVAL;
 
-	port_gid_table = &iboe->gids[port_num - 1];
+	port_gid_table = &iboe->gids[attr->port_num - 1];
 	spin_lock_bh(&iboe->lock);
 	for (i = 0; i < MLX4_MAX_PORT_GIDS; ++i) {
 		if (!memcmp(&port_gid_table->gids[i].gid, gid, sizeof(*gid)) &&
@@ -318,33 +315,30 @@ static int mlx4_ib_add_gid(struct ib_device *device,
 	spin_unlock_bh(&iboe->lock);
 
 	if (!ret && hw_update) {
-		ret = mlx4_ib_update_gids(gids, ibdev, port_num);
+		ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
 		kfree(gids);
 	}
 
 	return ret;
 }
 
-static int mlx4_ib_del_gid(struct ib_device *device,
-			   u8 port_num,
-			   unsigned int index,
-			   void **context)
+static int mlx4_ib_del_gid(const struct ib_gid_attr *attr, void **context)
 {
 	struct gid_cache_context *ctx = *context;
-	struct mlx4_ib_dev *ibdev = to_mdev(device);
+	struct mlx4_ib_dev *ibdev = to_mdev(attr->device);
 	struct mlx4_ib_iboe *iboe = &ibdev->iboe;
 	struct mlx4_port_gid_table   *port_gid_table;
 	int ret = 0;
 	int hw_update = 0;
 	struct gid_entry *gids = NULL;
 
-	if (!rdma_cap_roce_gid_table(device, port_num))
+	if (!rdma_cap_roce_gid_table(attr->device, attr->port_num))
 		return -EINVAL;
 
-	if (port_num > MLX4_MAX_PORTS)
+	if (attr->port_num > MLX4_MAX_PORTS)
 		return -EINVAL;
 
-	port_gid_table = &iboe->gids[port_num - 1];
+	port_gid_table = &iboe->gids[attr->port_num - 1];
 	spin_lock_bh(&iboe->lock);
 	if (ctx) {
 		ctx->refcount--;
@@ -376,7 +370,7 @@ static int mlx4_ib_del_gid(struct ib_device *device,
 	spin_unlock_bh(&iboe->lock);
 
 	if (!ret && hw_update) {
-		ret = mlx4_ib_update_gids(gids, ibdev, port_num);
+		ret = mlx4_ib_update_gids(gids, ibdev, attr->port_num);
 		kfree(gids);
 	}
 	return ret;
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 9e6780eadd1e..bc9eabd95948 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -502,18 +502,19 @@ static int set_roce_addr(struct mlx5_ib_dev *dev, u8 port_num,
 				      vlan_id, port_num);
 }
 
-static int mlx5_ib_add_gid(struct ib_device *device, u8 port_num,
-			   unsigned int index, const union ib_gid *gid,
+static int mlx5_ib_add_gid(const union ib_gid *gid,
 			   const struct ib_gid_attr *attr,
 			   __always_unused void **context)
 {
-	return set_roce_addr(to_mdev(device), port_num, index, gid, attr);
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, gid, attr);
 }
 
-static int mlx5_ib_del_gid(struct ib_device *device, u8 port_num,
-			   unsigned int index, __always_unused void **context)
+static int mlx5_ib_del_gid(const struct ib_gid_attr *attr,
+			   __always_unused void **context)
 {
-	return set_roce_addr(to_mdev(device), port_num, index, NULL, NULL);
+	return set_roce_addr(to_mdev(attr->device), attr->port_num,
+			     attr->index, NULL, NULL);
 }
 
 __be16 mlx5_get_roce_udp_sport(struct mlx5_ib_dev *dev, u8 port_num,
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
index 4834460e2a0b..0be33a81bbe6 100644
--- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c
@@ -62,17 +62,10 @@ static DEFINE_MUTEX(pvrdma_device_list_lock);
 static LIST_HEAD(pvrdma_device_list);
 static struct workqueue_struct *event_wq;
 
-static int pvrdma_add_gid(struct ib_device *ibdev,
-			  u8 port_num,
-			  unsigned int index,
-			  const union ib_gid *gid,
+static int pvrdma_add_gid(const union ib_gid *gid,
 			  const struct ib_gid_attr *attr,
 			  void **context);
-static int pvrdma_del_gid(struct ib_device *ibdev,
-			  u8 port_num,
-			  unsigned int index,
-			  void **context);
-
+static int pvrdma_del_gid(const struct ib_gid_attr *attr, void **context);
 
 static ssize_t show_hca(struct device *device, struct device_attribute *attr,
 			char *buf)
@@ -657,18 +650,15 @@ static int pvrdma_add_gid_at_index(struct pvrdma_dev *dev,
 	return 0;
 }
 
-static int pvrdma_add_gid(struct ib_device *ibdev,
-			  u8 port_num,
-			  unsigned int index,
-			  const union ib_gid *gid,
+static int pvrdma_add_gid(const union ib_gid *gid,
 			  const struct ib_gid_attr *attr,
 			  void **context)
 {
-	struct pvrdma_dev *dev = to_vdev(ibdev);
+	struct pvrdma_dev *dev = to_vdev(attr->device);
 
 	return pvrdma_add_gid_at_index(dev, gid,
 				       ib_gid_type_to_pvrdma(attr->gid_type),
-				       index);
+				       attr->index);
 }
 
 static int pvrdma_del_gid_at_index(struct pvrdma_dev *dev, int index)
@@ -698,17 +688,14 @@ static int pvrdma_del_gid_at_index(struct pvrdma_dev *dev, int index)
 	return 0;
 }
 
-static int pvrdma_del_gid(struct ib_device *ibdev,
-			  u8 port_num,
-			  unsigned int index,
-			  void **context)
+static int pvrdma_del_gid(const struct ib_gid_attr *attr, void **context)
 {
-	struct pvrdma_dev *dev = to_vdev(ibdev);
+	struct pvrdma_dev *dev = to_vdev(attr->device);
 
 	dev_dbg(&dev->pdev->dev, "removing gid at index %u from %s",
-		index, dev->netdev->name);
+		attr->index, dev->netdev->name);
 
-	return pvrdma_del_gid_at_index(dev, index);
+	return pvrdma_del_gid_at_index(dev, attr->index);
 }
 
 static void pvrdma_netdevice_event_handle(struct pvrdma_dev *dev,
diff --git a/drivers/infiniband/sw/rxe/rxe_verbs.c b/drivers/infiniband/sw/rxe/rxe_verbs.c
index 0661c2783b14..08f3e0618b81 100644
--- a/drivers/infiniband/sw/rxe/rxe_verbs.c
+++ b/drivers/infiniband/sw/rxe/rxe_verbs.c
@@ -77,19 +77,17 @@ out:
 	return rc;
 }
 
-static int rxe_add_gid(struct ib_device *device, u8 port_num, unsigned int
-		       index, const union ib_gid *gid,
+static int rxe_add_gid(const union ib_gid *gid,
 		       const struct ib_gid_attr *attr, void **context)
 {
-	if (index >= RXE_PORT_GID_TBL_LEN)
+	if (attr->index >= RXE_PORT_GID_TBL_LEN)
 		return -EINVAL;
 	return 0;
 }
 
-static int rxe_del_gid(struct ib_device *device, u8 port_num, unsigned int
-		       index, void **context)
+static int rxe_del_gid(const struct ib_gid_attr *attr, void **context)
 {
-	if (index >= RXE_PORT_GID_TBL_LEN)
+	if (attr->index >= RXE_PORT_GID_TBL_LEN)
 		return -EINVAL;
 	return 0;
 }
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index dc2541f13d7f..1e3059ce73b6 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -2151,34 +2151,29 @@ struct ib_device {
 	int		           (*query_gid)(struct ib_device *device,
 						u8 port_num, int index,
 						union ib_gid *gid);
-	/* When calling add_gid, the HW vendor's driver should
-	 * add the gid of device @device at gid index @index of
-	 * port @port_num to be @gid. Meta-info of that gid (for example,
-	 * the network device related to this gid is available
-	 * at @attr. @context allows the HW vendor driver to store extra
-	 * information together with a GID entry. The HW vendor may allocate
-	 * memory to contain this information and store it in @context when a
-	 * new GID entry is written to. Params are consistent until the next
-	 * call of add_gid or delete_gid. The function should return 0 on
+	/* When calling add_gid, the HW vendor's driver should add the gid
+	 * of device of port at gid index available at @attr. Meta-info of
+	 * that gid (for example, the network device related to this gid) is
+	 * available at @attr. @context allows the HW vendor driver to store
+	 * extra information together with a GID entry. The HW vendor driver may
+	 * allocate memory to contain this information and store it in @context
+	 * when a new GID entry is written to. Params are consistent until the
+	 * next call of add_gid or delete_gid. The function should return 0 on
 	 * success or error otherwise. The function could be called
-	 * concurrently for different ports. This function is only called
-	 * when roce_gid_table is used.
+	 * concurrently for different ports. This function is only called when
+	 * roce_gid_table is used.
 	 */
-	int		           (*add_gid)(struct ib_device *device,
-					      u8 port_num,
-					      unsigned int index,
-					      const union ib_gid *gid,
+	int		           (*add_gid)(const union ib_gid *gid,
 					      const struct ib_gid_attr *attr,
 					      void **context);
 	/* When calling del_gid, the HW vendor's driver should delete the
-	 * gid of device @device at gid index @index of port @port_num.
+	 * gid of device @device at gid index gid_index of port port_num
+	 * available in @attr.
 	 * Upon the deletion of a GID entry, the HW vendor must free any
 	 * allocated memory. The caller will clear @context afterwards.
 	 * This function is only called when roce_gid_table is used.
 	 */
-	int		           (*del_gid)(struct ib_device *device,
-					      u8 port_num,
-					      unsigned int index,
+	int		           (*del_gid)(const struct ib_gid_attr *attr,
 					      void **context);
 	int		           (*query_pkey)(struct ib_device *device,
 						 u8 port_num, u16 index, u16 *pkey);
-- 
cgit v1.2.3


From 8c84660bb437fe8692e6a2b4e85023ccb874a520 Mon Sep 17 00:00:00 2001
From: Matan Barak <matanb@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:41 +0300
Subject: IB/mlx5: Initialize the parsing tree root without the help of uverbs

In order to have a custom parsing tree, a provider driver needs to
assign its parsing tree to ib_device specs_tree field. Otherwise, the
uverbs client assigns a common default parsing tree for it.
In downstream patches, the mlx5_ib driver gains a custom parsing tree,
which contains both the common objects and a new flags field for the
UVERBS_FLOW_ACTION_ESP_CREATE command.
This patch makes mlx5_ib assign its own tree to specs_root, which
later on will be extended.

Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c    | 38 ++++++++++++++++++++++++++++++++++++
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  1 +
 2 files changed, 39 insertions(+)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index bc9eabd95948..82ad0faf8007 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -60,6 +60,10 @@
 #include "ib_rep.h"
 #include "cmd.h"
 #include <linux/mlx5/fs_helpers.h>
+#include <rdma/uverbs_std_types.h>
+
+#define UVERBS_MODULE_NAME mlx5_ib
+#include <rdma/uverbs_named_ioctl.h>
 
 #define DRIVER_NAME "mlx5_ib"
 #define DRIVER_VERSION "5.0-0"
@@ -4544,6 +4548,24 @@ static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev)
 	mlx5_nic_vport_disable_roce(dev->mdev);
 }
 
+#define NUM_TREES	0
+static int populate_specs_root(struct mlx5_ib_dev *dev)
+{
+	const struct uverbs_object_tree_def *default_root[NUM_TREES + 1] = {
+		uverbs_default_get_objects()};
+	size_t num_trees = 1;
+
+	dev->ib_dev.specs_root =
+		uverbs_alloc_spec_tree(num_trees, default_root);
+
+	return PTR_ERR_OR_ZERO(dev->ib_dev.specs_root);
+}
+
+static void depopulate_specs_root(struct mlx5_ib_dev *dev)
+{
+	uverbs_free_spec_tree(dev->ib_dev.specs_root);
+}
+
 void mlx5_ib_stage_init_cleanup(struct mlx5_ib_dev *dev)
 {
 	mlx5_ib_cleanup_multiport_master(dev);
@@ -4989,11 +5011,21 @@ void mlx5_ib_stage_bfrag_cleanup(struct mlx5_ib_dev *dev)
 	mlx5_free_bfreg(dev->mdev, &dev->bfreg);
 }
 
+static int mlx5_ib_stage_populate_specs(struct mlx5_ib_dev *dev)
+{
+	return populate_specs_root(dev);
+}
+
 int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 {
 	return ib_register_device(&dev->ib_dev, NULL);
 }
 
+static void mlx5_ib_stage_depopulate_specs(struct mlx5_ib_dev *dev)
+{
+	depopulate_specs_root(dev);
+}
+
 void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)
 {
 	destroy_umrc_res(dev);
@@ -5128,6 +5160,9 @@ static const struct mlx5_ib_profile pf_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
 		     NULL,
 		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     mlx5_ib_stage_depopulate_specs),
 	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
 		     mlx5_ib_stage_ib_reg_init,
 		     mlx5_ib_stage_ib_reg_cleanup),
@@ -5173,6 +5208,9 @@ static const struct mlx5_ib_profile nic_rep_profile = {
 	STAGE_CREATE(MLX5_IB_STAGE_PRE_IB_REG_UMR,
 		     NULL,
 		     mlx5_ib_stage_pre_ib_reg_umr_cleanup),
+	STAGE_CREATE(MLX5_IB_STAGE_SPECS,
+		     mlx5_ib_stage_populate_specs,
+		     mlx5_ib_stage_depopulate_specs),
 	STAGE_CREATE(MLX5_IB_STAGE_IB_REG,
 		     mlx5_ib_stage_ib_reg_init,
 		     mlx5_ib_stage_ib_reg_cleanup),
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index aeea74357cbe..0eda960ab8e0 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -740,6 +740,7 @@ enum mlx5_ib_stages {
 	MLX5_IB_STAGE_UAR,
 	MLX5_IB_STAGE_BFREG,
 	MLX5_IB_STAGE_PRE_IB_REG_UMR,
+	MLX5_IB_STAGE_SPECS,
 	MLX5_IB_STAGE_IB_REG,
 	MLX5_IB_STAGE_POST_IB_REG_UMR,
 	MLX5_IB_STAGE_DELAY_DROP,
-- 
cgit v1.2.3


From c6475a0bca30fc2f9e5e4c48935f08973c2780ef Mon Sep 17 00:00:00 2001
From: Aviad Yehezkel <aviadye@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:50 +0300
Subject: IB/mlx5: Add implementation for create and destroy action_xfrm

Adding implementation in mlx5 driver to create and destroy action_xfrm
object. This merely call the accel layer.

A user may pass MLX5_IB_XFRM_FLAGS_REQUIRE_METADATA flag which states
that [s]he expects a metadata header to be added to the payload. This
header represents information regarding the transformation's state.

Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Aviad Yehezkel <aviadye@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c         | 133 +++++++++++++++++++++++++++++-
 drivers/infiniband/hw/mlx5/mlx5_ib.h      |  16 ++++
 include/uapi/rdma/mlx5_user_ioctl_cmds.h  |  44 ++++++++++
 include/uapi/rdma/mlx5_user_ioctl_verbs.h |  43 ++++++++++
 4 files changed, 235 insertions(+), 1 deletion(-)
 create mode 100644 include/uapi/rdma/mlx5_user_ioctl_cmds.h
 create mode 100644 include/uapi/rdma/mlx5_user_ioctl_verbs.h

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 82ad0faf8007..8b16a42f2086 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -60,7 +60,10 @@
 #include "ib_rep.h"
 #include "cmd.h"
 #include <linux/mlx5/fs_helpers.h>
+#include <linux/mlx5/accel.h>
 #include <rdma/uverbs_std_types.h>
+#include <rdma/mlx5_user_ioctl_verbs.h>
+#include <rdma/mlx5_user_ioctl_cmds.h>
 
 #define UVERBS_MODULE_NAME mlx5_ib
 #include <rdma/uverbs_named_ioctl.h>
@@ -3103,6 +3106,122 @@ unlock:
 	return ERR_PTR(err);
 }
 
+static u32 mlx5_ib_flow_action_flags_to_accel_xfrm_flags(u32 mlx5_flags)
+{
+	u32 flags = 0;
+
+	if (mlx5_flags & MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA)
+		flags |= MLX5_ACCEL_XFRM_FLAG_REQUIRE_METADATA;
+
+	return flags;
+}
+
+#define MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED	MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA
+static struct ib_flow_action *
+mlx5_ib_create_flow_action_esp(struct ib_device *device,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dev *mdev = to_mdev(device);
+	struct ib_uverbs_flow_action_esp_keymat_aes_gcm *aes_gcm;
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs = {};
+	struct mlx5_ib_flow_action *action;
+	u64 action_flags;
+	u64 flags;
+	int err = 0;
+
+	if (IS_UVERBS_COPY_ERR(uverbs_copy_from(&action_flags, attrs,
+						MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS)))
+		return ERR_PTR(-EFAULT);
+
+	if (action_flags >= (MLX5_FLOW_ACTION_ESP_CREATE_LAST_SUPPORTED << 1))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	flags = mlx5_ib_flow_action_flags_to_accel_xfrm_flags(action_flags);
+
+	/* We current only support a subset of the standard features. Only a
+	 * keymat of type AES_GCM, with icv_len == 16, iv_algo == SEQ and esn
+	 * (with overlap). Full offload mode isn't supported.
+	 */
+	if (!attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)))
+		return ERR_PTR(-EOPNOTSUPP);
+
+	if (attr->keymat->protocol !=
+	    IB_UVERBS_FLOW_ACTION_ESP_KEYMAT_AES_GCM)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	aes_gcm = &attr->keymat->keymat.aes_gcm;
+
+	if (aes_gcm->icv_len != 16 ||
+	    aes_gcm->iv_algo != IB_UVERBS_FLOW_ACTION_IV_ALGO_SEQ)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	action = kmalloc(sizeof(*action), GFP_KERNEL);
+	if (!action)
+		return ERR_PTR(-ENOMEM);
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+	memcpy(&accel_attrs.keymat.aes_gcm.aes_key, &aes_gcm->aes_key,
+	       sizeof(accel_attrs.keymat.aes_gcm.aes_key));
+	accel_attrs.keymat.aes_gcm.key_len = aes_gcm->key_len * 8;
+	memcpy(&accel_attrs.keymat.aes_gcm.salt, &aes_gcm->salt,
+	       sizeof(accel_attrs.keymat.aes_gcm.salt));
+	memcpy(&accel_attrs.keymat.aes_gcm.seq_iv, &aes_gcm->iv,
+	       sizeof(accel_attrs.keymat.aes_gcm.seq_iv));
+	accel_attrs.keymat.aes_gcm.icv_len = aes_gcm->icv_len * 8;
+	accel_attrs.keymat.aes_gcm.iv_algo = MLX5_ACCEL_ESP_AES_GCM_IV_ALGO_SEQ;
+	accel_attrs.keymat_type = MLX5_ACCEL_ESP_KEYMAT_AES_GCM;
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_TRIGGERED;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ENCRYPT)
+		accel_attrs.action |= MLX5_ACCEL_ESP_ACTION_ENCRYPT;
+
+	action->esp_aes_gcm.ctx =
+		mlx5_accel_esp_create_xfrm(mdev->mdev, &accel_attrs, flags);
+	if (IS_ERR(action->esp_aes_gcm.ctx)) {
+		err = PTR_ERR(action->esp_aes_gcm.ctx);
+		goto err_parse;
+	}
+
+	action->esp_aes_gcm.ib_flags = attr->flags;
+
+	return &action->ib_action;
+
+err_parse:
+	kfree(action);
+	return ERR_PTR(err);
+}
+
+static int mlx5_ib_destroy_flow_action(struct ib_flow_action *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+
+	switch (action->type) {
+	case IB_FLOW_ACTION_ESP:
+		/*
+		 * We only support aes_gcm by now, so we implicitly know this is
+		 * the underline crypto.
+		 */
+		mlx5_accel_esp_destroy_xfrm(maction->esp_aes_gcm.ctx);
+		break;
+	default:
+		WARN_ON(true);
+		break;
+	}
+
+	kfree(maction);
+	return 0;
+}
+
 static int mlx5_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibqp->device);
@@ -4548,13 +4667,23 @@ static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev)
 	mlx5_nic_vport_disable_roce(dev->mdev);
 }
 
-#define NUM_TREES	0
+ADD_UVERBS_ATTRIBUTES_SIMPLE(mlx5_ib_flow_action, UVERBS_OBJECT_FLOW_ACTION,
+			     UVERBS_METHOD_FLOW_ACTION_ESP_CREATE,
+			     &UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS,
+						 UVERBS_ATTR_TYPE(u64),
+						 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
+#define NUM_TREES	1
 static int populate_specs_root(struct mlx5_ib_dev *dev)
 {
 	const struct uverbs_object_tree_def *default_root[NUM_TREES + 1] = {
 		uverbs_default_get_objects()};
 	size_t num_trees = 1;
 
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_DEVICE &&
+	    !WARN_ON(num_trees >= ARRAY_SIZE(default_root)))
+		default_root[num_trees++] = &mlx5_ib_flow_action;
+
 	dev->ib_dev.specs_root =
 		uverbs_alloc_spec_tree(num_trees, default_root);
 
@@ -4796,6 +4925,8 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 	dev->ib_dev.uverbs_ex_cmd_mask |=
 			(1ull << IB_USER_VERBS_EX_CMD_CREATE_FLOW) |
 			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
+	dev->ib_dev.create_flow_action_esp = mlx5_ib_create_flow_action_esp;
+	dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action;
 	dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
 
 	err = init_node_data(dev);
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 0eda960ab8e0..5fe73971425e 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -772,6 +772,16 @@ struct mlx5_ib_multiport_info {
 	bool unaffiliate;
 };
 
+struct mlx5_ib_flow_action {
+	struct ib_flow_action		ib_action;
+	union {
+		struct {
+			u64			    ib_flags;
+			struct mlx5_accel_esp_xfrm *ctx;
+		} esp_aes_gcm;
+	};
+};
+
 struct mlx5_ib_dev {
 	struct ib_device		ib_dev;
 	struct mlx5_core_dev		*mdev;
@@ -895,6 +905,12 @@ static inline struct mlx5_ib_mw *to_mmw(struct ib_mw *ibmw)
 	return container_of(ibmw, struct mlx5_ib_mw, ibmw);
 }
 
+static inline struct mlx5_ib_flow_action *
+to_mflow_act(struct ib_flow_action *ibact)
+{
+	return container_of(ibact, struct mlx5_ib_flow_action, ib_action);
+}
+
 int mlx5_ib_db_map_user(struct mlx5_ib_ucontext *context, unsigned long virt,
 			struct mlx5_db *db);
 void mlx5_ib_db_unmap_user(struct mlx5_ib_ucontext *context, struct mlx5_db *db);
diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
new file mode 100644
index 000000000000..521813d5348c
--- /dev/null
+++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef MLX5_USER_IOCTL_CMDS_H
+#define MLX5_USER_IOCTL_CMDS_H
+
+#include <rdma/ib_user_ioctl_cmds.h>
+
+enum mlx5_ib_create_flow_action_attrs {
+	/* This attribute belong to the driver namespace */
+	MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS = (1U << UVERBS_ID_NS_SHIFT),
+};
+
+#endif
+
diff --git a/include/uapi/rdma/mlx5_user_ioctl_verbs.h b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
new file mode 100644
index 000000000000..8a2fb33f3ed4
--- /dev/null
+++ b/include/uapi/rdma/mlx5_user_ioctl_verbs.h
@@ -0,0 +1,43 @@
+/*
+ * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#ifndef MLX5_USER_IOCTL_VERBS_H
+#define MLX5_USER_IOCTL_VERBS_H
+
+#include <linux/types.h>
+
+enum mlx5_ib_uapi_flow_action_flags {
+	MLX5_IB_UAPI_FLOW_ACTION_FLAGS_REQUIRE_METADATA	= 1 << 0,
+};
+
+#endif
+
-- 
cgit v1.2.3


From 349705c1936269aa6e5f923a0bc663e1addf288e Mon Sep 17 00:00:00 2001
From: Matan Barak <matanb@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:51 +0300
Subject: IB/mlx5: Add modify_flow_action_esp verb

Adding implementation in mlx5 driver to modify action_xfrm object. This
merely call the accel layer. Currently a user can modify only the
ESN parameters.

Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Aviad Yehezkel <aviadye@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c | 49 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 8b16a42f2086..bc46589a904d 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -3201,6 +3201,54 @@ err_parse:
 	return ERR_PTR(err);
 }
 
+static int
+mlx5_ib_modify_flow_action_esp(struct ib_flow_action *action,
+			       const struct ib_flow_action_attrs_esp *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
+	struct mlx5_accel_esp_xfrm_attrs accel_attrs;
+	int err = 0;
+
+	if (attr->keymat || attr->replay || attr->encap ||
+	    attr->spi || attr->seq || attr->tfc_pad ||
+	    attr->hard_limit_pkts ||
+	    (attr->flags & ~(IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			     IB_FLOW_ACTION_ESP_FLAGS_MOD_ESP_ATTRS |
+			     IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)))
+		return -EOPNOTSUPP;
+
+	/* Only the ESN value or the MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP can
+	 * be modified.
+	 */
+	if (!(maction->esp_aes_gcm.ib_flags &
+	      IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED) &&
+	    attr->flags & (IB_FLOW_ACTION_ESP_FLAGS_ESN_TRIGGERED |
+			   IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW))
+		return -EINVAL;
+
+	memcpy(&accel_attrs, &maction->esp_aes_gcm.ctx->attrs,
+	       sizeof(accel_attrs));
+
+	accel_attrs.esn = attr->esn;
+	if (attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW)
+		accel_attrs.flags |= MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+	else
+		accel_attrs.flags &= ~MLX5_ACCEL_ESP_FLAGS_ESN_STATE_OVERLAP;
+
+	err = mlx5_accel_esp_modify_xfrm(maction->esp_aes_gcm.ctx,
+					 &accel_attrs);
+	if (err)
+		return err;
+
+	maction->esp_aes_gcm.ib_flags &=
+		~IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+	maction->esp_aes_gcm.ib_flags |=
+		attr->flags & IB_UVERBS_FLOW_ACTION_ESP_FLAGS_ESN_NEW_WINDOW;
+
+	return 0;
+}
+
 static int mlx5_ib_destroy_flow_action(struct ib_flow_action *action)
 {
 	struct mlx5_ib_flow_action *maction = to_mflow_act(action);
@@ -4927,6 +4975,7 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 			(1ull << IB_USER_VERBS_EX_CMD_DESTROY_FLOW);
 	dev->ib_dev.create_flow_action_esp = mlx5_ib_create_flow_action_esp;
 	dev->ib_dev.destroy_flow_action = mlx5_ib_destroy_flow_action;
+	dev->ib_dev.modify_flow_action_esp = mlx5_ib_modify_flow_action_esp;
 	dev->ib_dev.driver_id = RDMA_DRIVER_MLX5;
 
 	err = init_node_data(dev);
-- 
cgit v1.2.3


From 802c2125689d1ceedd9671a8f728e85eacdac077 Mon Sep 17 00:00:00 2001
From: Aviad Yehezkel <aviadye@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:53 +0300
Subject: IB/mlx5: Add IPsec support for egress and ingress

This commit introduces support for the esp_aes_gcm flow
specification for the Innova device. To that end we add
support for egress steering and some validations that an
IPsec rule is indeed valid.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Aviad Yehezkel <aviadye@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c    | 127 +++++++++++++++++++++++++++++++----
 drivers/infiniband/hw/mlx5/mlx5_ib.h |   2 +
 2 files changed, 117 insertions(+), 12 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index bc46589a904d..119c4c165970 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -51,6 +51,7 @@
 #include <linux/mlx5/port.h>
 #include <linux/mlx5/vport.h>
 #include <linux/mlx5/fs.h>
+#include <linux/mlx5/fs_helpers.h>
 #include <linux/list.h>
 #include <rdma/ib_smi.h>
 #include <rdma/ib_umem.h>
@@ -2321,8 +2322,28 @@ static void set_tos(void *outer_c, void *outer_v, u8 mask, u8 val)
 		   offsetof(typeof(filter), field) -\
 		   sizeof(filter.field))
 
+static int parse_flow_flow_action(const union ib_flow_spec *ib_spec,
+				  const struct ib_flow_attr *flow_attr,
+				  struct mlx5_flow_act *action)
+{
+	struct mlx5_ib_flow_action *maction = to_mflow_act(ib_spec->action.act);
+
+	switch (maction->ib_action.type) {
+	case IB_FLOW_ACTION_ESP:
+		/* Currently only AES_GCM keymat is supported by the driver */
+		action->esp_id = (uintptr_t)maction->esp_aes_gcm.ctx;
+		action->action |= flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS ?
+			MLX5_FLOW_CONTEXT_ACTION_ENCRYPT :
+			MLX5_FLOW_CONTEXT_ACTION_DECRYPT;
+		return 0;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			   u32 *match_v, const union ib_flow_spec *ib_spec,
+			   const struct ib_flow_attr *flow_attr,
 			   struct mlx5_flow_act *action)
 {
 	void *misc_params_c = MLX5_ADDR_OF(fte_match_param, match_c,
@@ -2332,6 +2353,7 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 	void *headers_c;
 	void *headers_v;
 	int match_ipv;
+	int ret;
 
 	if (ib_spec->type & IB_FLOW_SPEC_INNER) {
 		headers_c = MLX5_ADDR_OF(fte_match_param, match_c,
@@ -2482,7 +2504,15 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			       ntohl(ib_spec->ipv6.mask.flow_label),
 			       ntohl(ib_spec->ipv6.val.flow_label),
 			       ib_spec->type & IB_FLOW_SPEC_INNER);
+		break;
+	case IB_FLOW_SPEC_ESP:
+		if (ib_spec->esp.mask.seq)
+			return -EOPNOTSUPP;
 
+		MLX5_SET(fte_match_set_misc, misc_params_c, outer_esp_spi,
+			 ntohl(ib_spec->esp.mask.spi));
+		MLX5_SET(fte_match_set_misc, misc_params_v, outer_esp_spi,
+			 ntohl(ib_spec->esp.val.spi));
 		break;
 	case IB_FLOW_SPEC_TCP:
 		if (FIELDS_NOT_SUPPORTED(ib_spec->tcp_udp.mask,
@@ -2550,6 +2580,11 @@ static int parse_flow_attr(struct mlx5_core_dev *mdev, u32 *match_c,
 			return -EOPNOTSUPP;
 		action->action |= MLX5_FLOW_CONTEXT_ACTION_DROP;
 		break;
+	case IB_FLOW_SPEC_ACTION_HANDLE:
+		ret = parse_flow_flow_action(ib_spec, flow_attr, action);
+		if (ret)
+			return ret;
+		break;
 	default:
 		return -EINVAL;
 	}
@@ -2591,6 +2626,46 @@ static bool flow_is_multicast_only(const struct ib_flow_attr *ib_attr)
 	return false;
 }
 
+enum valid_spec {
+	VALID_SPEC_INVALID,
+	VALID_SPEC_VALID,
+	VALID_SPEC_NA,
+};
+
+static enum valid_spec
+is_valid_esp_aes_gcm(struct mlx5_core_dev *mdev,
+		     const struct mlx5_flow_spec *spec,
+		     const struct mlx5_flow_act *flow_act,
+		     bool egress)
+{
+	const u32 *match_c = spec->match_criteria;
+	bool is_crypto =
+		(flow_act->action & (MLX5_FLOW_CONTEXT_ACTION_ENCRYPT |
+				     MLX5_FLOW_CONTEXT_ACTION_DECRYPT));
+	bool is_ipsec = mlx5_fs_is_ipsec_flow(match_c);
+	bool is_drop = flow_act->action & MLX5_FLOW_CONTEXT_ACTION_DROP;
+
+	/*
+	 * Currently only crypto is supported in egress, when regular egress
+	 * rules would be supported, always return VALID_SPEC_NA.
+	 */
+	if (!is_crypto)
+		return egress ? VALID_SPEC_INVALID : VALID_SPEC_NA;
+
+	return is_crypto && is_ipsec &&
+		(!egress || (!is_drop && !flow_act->has_flow_tag)) ?
+		VALID_SPEC_VALID : VALID_SPEC_INVALID;
+}
+
+static bool is_valid_spec(struct mlx5_core_dev *mdev,
+			  const struct mlx5_flow_spec *spec,
+			  const struct mlx5_flow_act *flow_act,
+			  bool egress)
+{
+	/* We curretly only support ipsec egress flow */
+	return is_valid_esp_aes_gcm(mdev, spec, flow_act, egress) != VALID_SPEC_INVALID;
+}
+
 static bool is_valid_ethertype(struct mlx5_core_dev *mdev,
 			       const struct ib_flow_attr *flow_attr,
 			       bool check_inner)
@@ -2715,13 +2790,17 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 	max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
 						       log_max_ft_size));
 	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
-		if (flow_is_multicast_only(flow_attr) &&
-		    !dont_trap)
+		if (ft_type == MLX5_IB_FT_TX)
+			priority = 0;
+		else if (flow_is_multicast_only(flow_attr) &&
+			 !dont_trap)
 			priority = MLX5_IB_FLOW_MCAST_PRIO;
 		else
 			priority = ib_prio_to_core_prio(flow_attr->priority,
 							dont_trap);
 		ns = mlx5_get_flow_namespace(dev->mdev,
+					     ft_type == MLX5_IB_FT_TX ?
+					     MLX5_FLOW_NAMESPACE_EGRESS :
 					     MLX5_FLOW_NAMESPACE_BYPASS);
 		num_entries = MLX5_FS_MAX_ENTRIES;
 		num_groups = MLX5_FS_MAX_TYPES;
@@ -2808,6 +2887,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	unsigned int spec_index;
 	int err = 0;
 	int dest_num = 1;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
 
 	if (!is_valid_attr(dev->mdev, flow_attr))
 		return ERR_PTR(-EINVAL);
@@ -2824,7 +2904,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	for (spec_index = 0; spec_index < flow_attr->num_of_specs; spec_index++) {
 		err = parse_flow_attr(dev->mdev, spec->match_criteria,
 				      spec->match_value,
-				      ib_flow, &flow_act);
+				      ib_flow, flow_attr, &flow_act);
 		if (err < 0)
 			goto free;
 
@@ -2847,12 +2927,23 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	}
 
 	spec->match_criteria_enable = get_match_criteria_enable(spec->match_criteria);
+
+	if (is_egress &&
+	    !is_valid_spec(dev->mdev, spec, &flow_act, is_egress)) {
+		err = -EINVAL;
+		goto free;
+	}
+
 	if (flow_act.action & MLX5_FLOW_CONTEXT_ACTION_DROP) {
 		rule_dst = NULL;
 		dest_num = 0;
 	} else {
-		flow_act.action = dst ? MLX5_FLOW_CONTEXT_ACTION_FWD_DEST :
-		    MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO;
+		if (is_egress)
+			flow_act.action |= MLX5_FLOW_CONTEXT_ACTION_ALLOW;
+		else
+			flow_act.action |=
+				dst ? MLX5_FLOW_CONTEXT_ACTION_FWD_DEST :
+					MLX5_FLOW_CONTEXT_ACTION_FWD_NEXT_PRIO;
 	}
 
 	if (flow_act.has_flow_tag &&
@@ -3026,6 +3117,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 	struct mlx5_flow_destination *dst = NULL;
 	struct mlx5_ib_flow_prio *ft_prio_tx = NULL;
 	struct mlx5_ib_flow_prio *ft_prio;
+	bool is_egress = flow_attr->flags & IB_FLOW_ATTR_FLAGS_EGRESS;
 	int err;
 	int underlay_qpn;
 
@@ -3034,7 +3126,13 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 
 	if (domain != IB_FLOW_DOMAIN_USER ||
 	    flow_attr->port > dev->num_ports ||
-	    (flow_attr->flags & ~IB_FLOW_ATTR_FLAGS_DONT_TRAP))
+	    (flow_attr->flags & ~(IB_FLOW_ATTR_FLAGS_DONT_TRAP |
+				  IB_FLOW_ATTR_FLAGS_EGRESS)))
+		return ERR_PTR(-EINVAL);
+
+	if (is_egress &&
+	    (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+	     flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT))
 		return ERR_PTR(-EINVAL);
 
 	dst = kzalloc(sizeof(*dst), GFP_KERNEL);
@@ -3043,7 +3141,8 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 
 	mutex_lock(&dev->flow_db->lock);
 
-	ft_prio = get_flow_table(dev, flow_attr, MLX5_IB_FT_RX);
+	ft_prio = get_flow_table(dev, flow_attr,
+				 is_egress ? MLX5_IB_FT_TX : MLX5_IB_FT_RX);
 	if (IS_ERR(ft_prio)) {
 		err = PTR_ERR(ft_prio);
 		goto unlock;
@@ -3057,11 +3156,15 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 		}
 	}
 
-	dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
-	if (mqp->flags & MLX5_IB_QP_RSS)
-		dst->tir_num = mqp->rss_qp.tirn;
-	else
-		dst->tir_num = mqp->raw_packet_qp.rq.tirn;
+	if (is_egress) {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_PORT;
+	} else {
+		dst->type = MLX5_FLOW_DESTINATION_TYPE_TIR;
+		if (mqp->flags & MLX5_IB_QP_RSS)
+			dst->tir_num = mqp->rss_qp.tirn;
+		else
+			dst->tir_num = mqp->raw_packet_qp.rq.tirn;
+	}
 
 	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
 		if (flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP)  {
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 5fe73971425e..2b27ddafc354 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -152,6 +152,7 @@ struct mlx5_ib_pd {
 
 #define MLX5_IB_NUM_FLOW_FT		(MLX5_IB_FLOW_LEFTOVERS_PRIO + 1)
 #define MLX5_IB_NUM_SNIFFER_FTS		2
+#define MLX5_IB_NUM_EGRESS_FTS		1
 struct mlx5_ib_flow_prio {
 	struct mlx5_flow_table		*flow_table;
 	unsigned int			refcount;
@@ -167,6 +168,7 @@ struct mlx5_ib_flow_handler {
 struct mlx5_ib_flow_db {
 	struct mlx5_ib_flow_prio	prios[MLX5_IB_NUM_FLOW_FT];
 	struct mlx5_ib_flow_prio	sniffer[MLX5_IB_NUM_SNIFFER_FTS];
+	struct mlx5_ib_flow_prio	egress[MLX5_IB_NUM_EGRESS_FTS];
 	struct mlx5_flow_table		*lag_demux_ft;
 	/* Protect flow steering bypass flow tables
 	 * when add/del flow rules.
-- 
cgit v1.2.3


From c03faa562d0279e463900c92d4a54c0dc3c806b0 Mon Sep 17 00:00:00 2001
From: Matan Barak <matanb@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:54 +0300
Subject: IB/mlx5: Add information for querying IPsec capabilities

Users should be able to query for IPSec support. Adding a few
capabilities bits as part of the driver specific part in
alloc_ucontext:
MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_REQ_METADATA
	Payload's header is returned with metadata representing the
	IPSec decryption state.
MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_RX
	Support ESP_AES_GCM in ingress path.
MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_TX
	Support ESP_AES_GCM in egress path.
MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_SPI_RSS_ONLY
	Hardware doesn't support matching SPI in flow steering rules
	but just hashing and spreading the traffic accordingly.

Signed-off-by: Aviad Yehezkel <aviadye@mellanox.com>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c | 12 ++++++++++++
 include/uapi/rdma/mlx5-abi.h      | 10 +++++++++-
 2 files changed, 21 insertions(+), 1 deletion(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 119c4c165970..25e70ae0b484 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -1681,6 +1681,18 @@ static struct ib_ucontext *mlx5_ib_alloc_ucontext(struct ib_device *ibdev,
 	resp.response_length = min(offsetof(typeof(resp), response_length) +
 				   sizeof(resp.response_length), udata->outlen);
 
+	if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_DEVICE) {
+		if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_EGRESS))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_REQUIRED_METADATA)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_REQ_METADATA;
+		if (MLX5_CAP_FLOWTABLE(dev->mdev, flow_table_properties_nic_receive.ft_field_support.outer_esp_spi))
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_SPI_STEERING;
+		if (mlx5_accel_ipsec_device_caps(dev->mdev) & MLX5_ACCEL_IPSEC_CAP_TX_IV_IS_ESN)
+			resp.flow_action_flags |= MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_TX_IV_IS_ESN;
+		/* MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_FULL_OFFLOAD is currently always 0 */
+	}
+
 	context = kzalloc(sizeof(*context), GFP_KERNEL);
 	if (!context)
 		return ERR_PTR(-ENOMEM);
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index f7d18fb01771..f60d2659cdb7 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -107,6 +107,14 @@ enum mlx5_user_inline_mode {
 	MLX5_USER_INLINE_MODE_TCP_UDP,
 };
 
+enum {
+	MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM = 1 << 0,
+	MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_REQ_METADATA = 1 << 1,
+	MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_SPI_STEERING = 1 << 2,
+	MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_FULL_OFFLOAD = 1 << 3,
+	MLX5_USER_ALLOC_UCONTEXT_FLOW_ACTION_FLAGS_ESP_AES_GCM_TX_IV_IS_ESN = 1 << 4,
+};
+
 struct mlx5_ib_alloc_ucontext_resp {
 	__u32	qp_tab_size;
 	__u32	bf_reg_size;
@@ -118,7 +126,7 @@ struct mlx5_ib_alloc_ucontext_resp {
 	__u32	max_recv_wr;
 	__u32	max_srq_recv_wr;
 	__u16	num_ports;
-	__u16	reserved1;
+	__u16	flow_action_flags;
 	__u32	comp_mask;
 	__u32	response_length;
 	__u8	cqe_version;
-- 
cgit v1.2.3


From 2d93fc856959bb940b3cf1e7cbad38721d6bea75 Mon Sep 17 00:00:00 2001
From: Matan Barak <matanb@mellanox.com>
Date: Wed, 28 Mar 2018 09:27:55 +0300
Subject: IB/mlx5: Add ability to hash by IPSEC_SPI when creating a TIR

When a Raw Ethernet QP is created, we actually create a few objects.
One of these objects is a TIR. Currently, a TIR could hash (and spread
the traffic) by IP or port only. Adding a hashing by IPSec SPI to TIR
creation with the required UAPI bit.

Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c |  4 ++++
 drivers/infiniband/hw/mlx5/qp.c   | 16 ++++++++++++----
 include/uapi/rdma/mlx5-abi.h      |  1 +
 3 files changed, 17 insertions(+), 4 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 25e70ae0b484..31295e39896c 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -856,6 +856,10 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 						MLX5_RX_HASH_SRC_PORT_UDP |
 						MLX5_RX_HASH_DST_PORT_UDP |
 						MLX5_RX_HASH_INNER;
+			if (mlx5_accel_ipsec_device_caps(dev->mdev) &
+			    MLX5_ACCEL_IPSEC_CAP_DEVICE)
+				resp.rss_caps.rx_hash_fields_mask |=
+					MLX5_RX_HASH_IPSEC_SPI;
 			resp.response_length += sizeof(resp.rss_caps);
 		}
 	} else {
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index c8f01f32ebb4..0a0524f60924 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -1413,6 +1413,7 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	void *tirc;
 	void *hfso;
 	u32 selected_fields = 0;
+	u32 outer_l4;
 	size_t min_resp_len;
 	u32 tdn = mucontext->tdn;
 	struct mlx5_ib_create_qp_rss ucmd = {};
@@ -1543,10 +1544,14 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 		MLX5_SET(rx_hash_field_select, hfso, l3_prot_type,
 			 MLX5_L3_PROT_TYPE_IPV6);
 
-	if (((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
-	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP)) &&
-	     ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
-	     (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))) {
+	outer_l4 = ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_TCP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_TCP)) << 0 |
+		   ((ucmd.rx_hash_fields_mask & MLX5_RX_HASH_SRC_PORT_UDP) ||
+		    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP)) << 1 |
+		   (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI) << 2;
+
+	/* Check that only one l4 protocol is set */
+	if (outer_l4 & (outer_l4 - 1)) {
 		err = -EINVAL;
 		goto err;
 	}
@@ -1577,6 +1582,9 @@ static int create_rss_raw_qp_tir(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
 	    (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_DST_PORT_UDP))
 		selected_fields |= MLX5_HASH_FIELD_SEL_L4_DPORT;
 
+	if (ucmd.rx_hash_fields_mask & MLX5_RX_HASH_IPSEC_SPI)
+		selected_fields |= MLX5_HASH_FIELD_SEL_IPSEC_SPI;
+
 	MLX5_SET(rx_hash_field_select, hfso, selected_fields, selected_fields);
 
 create_tir:
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index f60d2659cdb7..d86a65b993f8 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -327,6 +327,7 @@ enum mlx5_rx_hash_fields {
 	MLX5_RX_HASH_DST_PORT_TCP	= 1 << 5,
 	MLX5_RX_HASH_SRC_PORT_UDP	= 1 << 6,
 	MLX5_RX_HASH_DST_PORT_UDP	= 1 << 7,
+	MLX5_RX_HASH_IPSEC_SPI		= 1 << 8,
 	/* Save bits for future fields */
 	MLX5_RX_HASH_INNER		= (1UL << 31),
 };
-- 
cgit v1.2.3


From 24da00164f7a9c247d2224a54494d0e955199630 Mon Sep 17 00:00:00 2001
From: Ariel Levkovich <lariel@mellanox.com>
Date: Thu, 5 Apr 2018 18:53:27 +0300
Subject: IB/mlx5: Device memory support in mlx5_ib

This patch adds the mlx5_ib driver implementation for the device
memory allocation API.
It implements the ib_device callbacks for allocation and deallocation
operations as well as a new mmap command support which allows mapping
an allocated device memory to a VMA.

The change also adds reporting of device memory maximum size and
alignment parameters reported in device capabilities.

The allocation/deallocation operations are using new firmware
commands to allocate MEMIC memory on the device.

Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/cmd.c         | 106 +++++++++++++++++++++++
 drivers/infiniband/hw/mlx5/cmd.h         |   4 +
 drivers/infiniband/hw/mlx5/main.c        | 143 ++++++++++++++++++++++++++++++-
 drivers/infiniband/hw/mlx5/mlx5_ib.h     |  35 +++++++-
 include/linux/mlx5/mlx5_ifc.h            |  55 ++++++++++++
 include/uapi/rdma/mlx5-abi.h             |   1 +
 include/uapi/rdma/mlx5_user_ioctl_cmds.h |   6 +-
 7 files changed, 347 insertions(+), 3 deletions(-)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/cmd.c b/drivers/infiniband/hw/mlx5/cmd.c
index 6f6712f87a73..55a227cc8609 100644
--- a/drivers/infiniband/hw/mlx5/cmd.c
+++ b/drivers/infiniband/hw/mlx5/cmd.c
@@ -66,3 +66,109 @@ int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *dev,
 
 	return mlx5_cmd_exec(dev, in, in_size, out, sizeof(out));
 }
+
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			  u64 length, u32 alignment)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 num_memic_hw_pages = MLX5_CAP_DEV_MEM(dev, memic_bar_size)
+					>> PAGE_SHIFT;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 max_alignment = MLX5_CAP_DEV_MEM(dev, log_max_memic_addr_alignment);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(alloc_memic_out)] = {};
+	u32 in[MLX5_ST_SZ_DW(alloc_memic_in)] = {};
+	u32 mlx5_alignment;
+	u64 page_idx = 0;
+	int ret = 0;
+
+	if (!length || (length & MLX5_MEMIC_ALLOC_SIZE_MASK))
+		return -EINVAL;
+
+	/* mlx5 device sets alignment as 64*2^driver_value
+	 * so normalizing is needed.
+	 */
+	mlx5_alignment = (alignment < MLX5_MEMIC_BASE_ALIGN) ? 0 :
+			 alignment - MLX5_MEMIC_BASE_ALIGN;
+	if (mlx5_alignment > max_alignment)
+		return -EINVAL;
+
+	MLX5_SET(alloc_memic_in, in, opcode, MLX5_CMD_OP_ALLOC_MEMIC);
+	MLX5_SET(alloc_memic_in, in, range_size, num_pages * PAGE_SIZE);
+	MLX5_SET(alloc_memic_in, in, memic_size, length);
+	MLX5_SET(alloc_memic_in, in, log_memic_addr_alignment,
+		 mlx5_alignment);
+
+	do {
+		spin_lock(&memic->memic_lock);
+		page_idx = bitmap_find_next_zero_area(memic->memic_alloc_pages,
+						      num_memic_hw_pages,
+						      page_idx,
+						      num_pages, 0);
+
+		if (page_idx + num_pages <= num_memic_hw_pages)
+			bitmap_set(memic->memic_alloc_pages,
+				   page_idx, num_pages);
+		else
+			ret = -ENOMEM;
+
+		spin_unlock(&memic->memic_lock);
+
+		if (ret)
+			return ret;
+
+		MLX5_SET64(alloc_memic_in, in, range_start_addr,
+			   hw_start_addr + (page_idx * PAGE_SIZE));
+
+		ret = mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+		if (ret) {
+			spin_lock(&memic->memic_lock);
+			bitmap_clear(memic->memic_alloc_pages,
+				     page_idx, num_pages);
+			spin_unlock(&memic->memic_lock);
+
+			if (ret == -EAGAIN) {
+				page_idx++;
+				continue;
+			}
+
+			return ret;
+		}
+
+		*addr = pci_resource_start(dev->pdev, 0) +
+			MLX5_GET64(alloc_memic_out, out, memic_start_addr);
+
+		return ret;
+	} while (page_idx < num_memic_hw_pages);
+
+	return ret;
+}
+
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length)
+{
+	struct mlx5_core_dev *dev = memic->dev;
+	u64 hw_start_addr = MLX5_CAP64_DEV_MEM(dev, memic_bar_start_addr);
+	u32 num_pages = DIV_ROUND_UP(length, PAGE_SIZE);
+	u32 out[MLX5_ST_SZ_DW(dealloc_memic_out)] = {0};
+	u32 in[MLX5_ST_SZ_DW(dealloc_memic_in)] = {0};
+	u64 start_page_idx;
+	int err;
+
+	addr -= pci_resource_start(dev->pdev, 0);
+	start_page_idx = (addr - hw_start_addr) >> PAGE_SHIFT;
+
+	MLX5_SET(dealloc_memic_in, in, opcode, MLX5_CMD_OP_DEALLOC_MEMIC);
+	MLX5_SET64(dealloc_memic_in, in, memic_start_addr, addr);
+	MLX5_SET(dealloc_memic_in, in, memic_size, length);
+
+	err =  mlx5_cmd_exec(dev, in, sizeof(in), out, sizeof(out));
+
+	if (!err) {
+		spin_lock(&memic->memic_lock);
+		bitmap_clear(memic->memic_alloc_pages,
+			     start_page_idx, num_pages);
+		spin_unlock(&memic->memic_lock);
+	}
+
+	return err;
+}
diff --git a/drivers/infiniband/hw/mlx5/cmd.h b/drivers/infiniband/hw/mlx5/cmd.h
index 78ffded7cc2c..e7206c8a8011 100644
--- a/drivers/infiniband/hw/mlx5/cmd.h
+++ b/drivers/infiniband/hw/mlx5/cmd.h
@@ -33,6 +33,7 @@
 #ifndef MLX5_IB_CMD_H
 #define MLX5_IB_CMD_H
 
+#include "mlx5_ib.h"
 #include <linux/kernel.h>
 #include <linux/mlx5/driver.h>
 
@@ -41,4 +42,7 @@ int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
 			       void *out, int out_size);
 int mlx5_cmd_modify_cong_params(struct mlx5_core_dev *mdev,
 				void *in, int in_size);
+int mlx5_cmd_alloc_memic(struct mlx5_memic *memic, phys_addr_t *addr,
+			 u64 length, u32 alignment);
+int mlx5_cmd_dealloc_memic(struct mlx5_memic *memic, u64 addr, u64 length);
 #endif /* MLX5_IB_CMD_H */
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 31295e39896c..e17eac32394c 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -38,6 +38,7 @@
 #include <linux/pci.h>
 #include <linux/dma-mapping.h>
 #include <linux/slab.h>
+#include <linux/bitmap.h>
 #if defined(CONFIG_X86)
 #include <asm/pat.h>
 #endif
@@ -891,6 +892,11 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		props->raw_packet_caps |= IB_RAW_PACKET_CAP_SCATTER_FCS;
 	}
 
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		props->max_dm_size =
+			MLX5_CAP_DEV_MEM(mdev, max_memic_size);
+	}
+
 	if (mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS))
 		props->device_cap_flags |= IB_DEVICE_MANAGED_FLOW_STEERING;
 
@@ -2014,6 +2020,8 @@ static inline char *mmap_cmd2str(enum mlx5_ib_mmap_cmd cmd)
 		return "best effort WC";
 	case MLX5_IB_MMAP_NC_PAGE:
 		return "NC";
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return "Device Memory";
 	default:
 		return NULL;
 	}
@@ -2172,6 +2180,34 @@ free_bfreg:
 	return err;
 }
 
+static int dm_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
+{
+	struct mlx5_ib_ucontext *mctx = to_mucontext(context);
+	struct mlx5_ib_dev *dev = to_mdev(context->device);
+	u16 page_idx = get_extended_index(vma->vm_pgoff);
+	size_t map_size = vma->vm_end - vma->vm_start;
+	u32 npages = map_size >> PAGE_SHIFT;
+	phys_addr_t pfn;
+	pgprot_t prot;
+
+	if (find_next_zero_bit(mctx->dm_pages, page_idx + npages, page_idx) !=
+	    page_idx + npages)
+		return -EINVAL;
+
+	pfn = ((pci_resource_start(dev->mdev->pdev, 0) +
+	      MLX5_CAP64_DEV_MEM(dev->mdev, memic_bar_start_addr)) >>
+	      PAGE_SHIFT) +
+	      page_idx;
+	prot = pgprot_writecombine(vma->vm_page_prot);
+	vma->vm_page_prot = prot;
+
+	if (io_remap_pfn_range(vma, vma->vm_start, pfn, map_size,
+			       vma->vm_page_prot))
+		return -EAGAIN;
+
+	return mlx5_ib_set_vma_data(vma, mctx);
+}
+
 static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
 {
 	struct mlx5_ib_ucontext *context = to_mucontext(ibcontext);
@@ -2216,6 +2252,9 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 	case MLX5_IB_MMAP_CLOCK_INFO:
 		return mlx5_ib_mmap_clock_info_page(dev, vma, context);
 
+	case MLX5_IB_MMAP_DEVICE_MEM:
+		return dm_mmap(ibcontext, vma);
+
 	default:
 		return -EINVAL;
 	}
@@ -2223,6 +2262,87 @@ static int mlx5_ib_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vm
 	return 0;
 }
 
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs)
+{
+	u64 act_size = roundup(attr->length, MLX5_MEMIC_BASE_SIZE);
+	struct mlx5_memic *memic = &to_mdev(ibdev)->memic;
+	phys_addr_t memic_addr;
+	struct mlx5_ib_dm *dm;
+	u64 start_offset;
+	u32 page_idx;
+	int err;
+
+	dm = kzalloc(sizeof(*dm), GFP_KERNEL);
+	if (!dm)
+		return ERR_PTR(-ENOMEM);
+
+	mlx5_ib_dbg(to_mdev(ibdev), "alloc_memic req: user_length=0x%llx act_length=0x%llx log_alignment=%d\n",
+		    attr->length, act_size, attr->alignment);
+
+	err = mlx5_cmd_alloc_memic(memic, &memic_addr,
+				   act_size, attr->alignment);
+	if (err)
+		goto err_free;
+
+	start_offset = memic_addr & ~PAGE_MASK;
+	page_idx = (memic_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+			     &start_offset, sizeof(start_offset));
+	if (err)
+		goto err_dealloc;
+
+	err = uverbs_copy_to(attrs,
+			     MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+			     &page_idx, sizeof(page_idx));
+	if (err)
+		goto err_dealloc;
+
+	bitmap_set(to_mucontext(context)->dm_pages, page_idx,
+		   DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	dm->dev_addr = memic_addr;
+
+	return &dm->ibdm;
+
+err_dealloc:
+	mlx5_cmd_dealloc_memic(memic, memic_addr,
+			       act_size);
+err_free:
+	kfree(dm);
+	return ERR_PTR(err);
+}
+
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm)
+{
+	struct mlx5_memic *memic = &to_mdev(ibdm->device)->memic;
+	struct mlx5_ib_dm *dm = to_mdm(ibdm);
+	u64 act_size = roundup(dm->ibdm.length, MLX5_MEMIC_BASE_SIZE);
+	u32 page_idx;
+	int ret;
+
+	ret = mlx5_cmd_dealloc_memic(memic, dm->dev_addr, act_size);
+	if (ret)
+		return ret;
+
+	page_idx = (dm->dev_addr - pci_resource_start(memic->dev->pdev, 0) -
+		    MLX5_CAP64_DEV_MEM(memic->dev, memic_bar_start_addr)) >>
+		    PAGE_SHIFT;
+	bitmap_clear(to_mucontext(ibdm->uobject->context)->dm_pages,
+		     page_idx,
+		     DIV_ROUND_UP(act_size, PAGE_SIZE));
+
+	kfree(dm);
+
+	return 0;
+}
+
 static struct ib_pd *mlx5_ib_alloc_pd(struct ib_device *ibdev,
 				      struct ib_ucontext *context,
 				      struct ib_udata *udata)
@@ -4834,13 +4954,22 @@ static void mlx5_ib_cleanup_multiport_master(struct mlx5_ib_dev *dev)
 	mlx5_nic_vport_disable_roce(dev->mdev);
 }
 
+ADD_UVERBS_ATTRIBUTES_SIMPLE(mlx5_ib_dm, UVERBS_OBJECT_DM,
+			     UVERBS_METHOD_DM_ALLOC,
+			     &UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET,
+						  UVERBS_ATTR_TYPE(u64),
+						  UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)),
+			     &UVERBS_ATTR_PTR_OUT(MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+						  UVERBS_ATTR_TYPE(u16),
+						  UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
+
 ADD_UVERBS_ATTRIBUTES_SIMPLE(mlx5_ib_flow_action, UVERBS_OBJECT_FLOW_ACTION,
 			     UVERBS_METHOD_FLOW_ACTION_ESP_CREATE,
 			     &UVERBS_ATTR_PTR_IN(MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS,
 						 UVERBS_ATTR_TYPE(u64),
 						 UA_FLAGS(UVERBS_ATTR_SPEC_F_MANDATORY)));
 
-#define NUM_TREES	1
+#define NUM_TREES	2
 static int populate_specs_root(struct mlx5_ib_dev *dev)
 {
 	const struct uverbs_object_tree_def *default_root[NUM_TREES + 1] = {
@@ -4851,6 +4980,10 @@ static int populate_specs_root(struct mlx5_ib_dev *dev)
 	    !WARN_ON(num_trees >= ARRAY_SIZE(default_root)))
 		default_root[num_trees++] = &mlx5_ib_flow_action;
 
+	if (MLX5_CAP_DEV_MEM(dev->mdev, memic) &&
+	    !WARN_ON(num_trees >= ARRAY_SIZE(default_root)))
+		default_root[num_trees++] = &mlx5_ib_dm;
+
 	dev->ib_dev.specs_root =
 		uverbs_alloc_spec_tree(num_trees, default_root);
 
@@ -4925,6 +5058,9 @@ int mlx5_ib_stage_init_init(struct mlx5_ib_dev *dev)
 	INIT_LIST_HEAD(&dev->qp_list);
 	spin_lock_init(&dev->reset_flow_resource_lock);
 
+	spin_lock_init(&dev->memic.memic_lock);
+	dev->memic.dev = mdev;
+
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 	err = init_srcu_struct(&dev->mr_srcu);
 	if (err)
@@ -5087,6 +5223,11 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 			(1ull << IB_USER_VERBS_CMD_CLOSE_XRCD);
 	}
 
+	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
+		dev->ib_dev.alloc_dm = mlx5_ib_alloc_dm;
+		dev->ib_dev.dealloc_dm = mlx5_ib_dealloc_dm;
+	}
+
 	dev->ib_dev.create_flow	= mlx5_ib_create_flow;
 	dev->ib_dev.destroy_flow = mlx5_ib_destroy_flow;
 	dev->ib_dev.uverbs_ex_cmd_mask |=
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 2b27ddafc354..3e9b6548a96b 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -45,6 +45,7 @@
 #include <linux/mlx5/transobj.h>
 #include <rdma/ib_user_verbs.h>
 #include <rdma/mlx5-abi.h>
+#include <rdma/uverbs_ioctl.h>
 
 #define mlx5_ib_dbg(dev, format, arg...)				\
 pr_debug("%s:%s:%d:(pid %d): " format, (dev)->ib_dev.name, __func__,	\
@@ -108,6 +109,16 @@ enum {
 	MLX5_IB_INVALID_BFREG		= BIT(31),
 };
 
+enum {
+	MLX5_MAX_MEMIC_PAGES = 0x100,
+	MLX5_MEMIC_ALLOC_SIZE_MASK = 0x3f,
+};
+
+enum {
+	MLX5_MEMIC_BASE_ALIGN	= 6,
+	MLX5_MEMIC_BASE_SIZE	= 1 << MLX5_MEMIC_BASE_ALIGN,
+};
+
 struct mlx5_ib_vma_private_data {
 	struct list_head list;
 	struct vm_area_struct *vma;
@@ -131,6 +142,7 @@ struct mlx5_ib_ucontext {
 	struct mutex		vma_private_list_mutex;
 
 	u64			lib_caps;
+	DECLARE_BITMAP(dm_pages, MLX5_MAX_MEMIC_PAGES);
 };
 
 static inline struct mlx5_ib_ucontext *to_mucontext(struct ib_ucontext *ibucontext)
@@ -521,6 +533,11 @@ enum mlx5_ib_mtt_access_flags {
 	MLX5_IB_MTT_WRITE = (1 << 1),
 };
 
+struct mlx5_ib_dm {
+	struct ib_dm		ibdm;
+	phys_addr_t		dev_addr;
+};
+
 #define MLX5_IB_MTT_PRESENT (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)
 
 struct mlx5_ib_mr {
@@ -784,6 +801,12 @@ struct mlx5_ib_flow_action {
 	};
 };
 
+struct mlx5_memic {
+	struct mlx5_core_dev *dev;
+	spinlock_t		memic_lock;
+	DECLARE_BITMAP(memic_alloc_pages, MLX5_MAX_MEMIC_PAGES);
+};
+
 struct mlx5_ib_dev {
 	struct ib_device		ib_dev;
 	struct mlx5_core_dev		*mdev;
@@ -830,6 +853,7 @@ struct mlx5_ib_dev {
 	u8			umr_fence;
 	struct list_head	ib_dev_list;
 	u64			sys_image_guid;
+	struct mlx5_memic	memic;
 };
 
 static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
@@ -897,6 +921,11 @@ static inline struct mlx5_ib_srq *to_mibsrq(struct mlx5_core_srq *msrq)
 	return container_of(msrq, struct mlx5_ib_srq, msrq);
 }
 
+static inline struct mlx5_ib_dm *to_mdm(struct ib_dm *ibdm)
+{
+	return container_of(ibdm, struct mlx5_ib_dm, ibdm);
+}
+
 static inline struct mlx5_ib_mr *to_mmr(struct ib_mr *ibmr)
 {
 	return container_of(ibmr, struct mlx5_ib_mr, ibmr);
@@ -1041,7 +1070,11 @@ struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
 						      struct ib_udata *udata);
 int mlx5_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
 bool mlx5_ib_dc_atomic_is_supported(struct mlx5_ib_dev *dev);
-
+struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
+			       struct ib_ucontext *context,
+			       struct ib_dm_alloc_attr *attr,
+			       struct uverbs_attr_bundle *attrs);
+int mlx5_ib_dealloc_dm(struct ib_dm *ibdm);
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev);
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index 13c3bf25753b..a64e59b65a33 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -92,6 +92,8 @@ enum {
 	MLX5_CMD_OP_DESTROY_MKEY                  = 0x202,
 	MLX5_CMD_OP_QUERY_SPECIAL_CONTEXTS        = 0x203,
 	MLX5_CMD_OP_PAGE_FAULT_RESUME             = 0x204,
+	MLX5_CMD_OP_ALLOC_MEMIC                   = 0x205,
+	MLX5_CMD_OP_DEALLOC_MEMIC                 = 0x206,
 	MLX5_CMD_OP_CREATE_EQ                     = 0x301,
 	MLX5_CMD_OP_DESTROY_EQ                    = 0x302,
 	MLX5_CMD_OP_QUERY_EQ                      = 0x303,
@@ -8886,4 +8888,57 @@ struct mlx5_ifc_destroy_vport_lag_in_bits {
 	u8         reserved_at_40[0x40];
 };
 
+struct mlx5_ifc_alloc_memic_in_bits {
+	u8         opcode[0x10];
+	u8         reserved_at_10[0x10];
+
+	u8         reserved_at_20[0x10];
+	u8         op_mod[0x10];
+
+	u8         reserved_at_30[0x20];
+
+	u8	   reserved_at_40[0x18];
+	u8	   log_memic_addr_alignment[0x8];
+
+	u8         range_start_addr[0x40];
+
+	u8         range_size[0x20];
+
+	u8         memic_size[0x20];
+};
+
+struct mlx5_ifc_alloc_memic_out_bits {
+	u8         status[0x8];
+	u8         reserved_at_8[0x18];
+
+	u8         syndrome[0x20];
+
+	u8         memic_start_addr[0x40];
+};
+
+struct mlx5_ifc_dealloc_memic_in_bits {
+	u8         opcode[0x10];
+	u8         reserved_at_10[0x10];
+
+	u8         reserved_at_20[0x10];
+	u8         op_mod[0x10];
+
+	u8         reserved_at_40[0x40];
+
+	u8         memic_start_addr[0x40];
+
+	u8         memic_size[0x20];
+
+	u8         reserved_at_e0[0x20];
+};
+
+struct mlx5_ifc_dealloc_memic_out_bits {
+	u8         status[0x8];
+	u8         reserved_at_8[0x18];
+
+	u8         syndrome[0x20];
+
+	u8         reserved_at_40[0x40];
+};
+
 #endif /* MLX5_IFC_H */
diff --git a/include/uapi/rdma/mlx5-abi.h b/include/uapi/rdma/mlx5-abi.h
index d86a65b993f8..cb4a02c4a1ce 100644
--- a/include/uapi/rdma/mlx5-abi.h
+++ b/include/uapi/rdma/mlx5-abi.h
@@ -430,6 +430,7 @@ enum mlx5_ib_mmap_cmd {
 	MLX5_IB_MMAP_CORE_CLOCK                 = 5,
 	MLX5_IB_MMAP_ALLOC_WC                   = 6,
 	MLX5_IB_MMAP_CLOCK_INFO                 = 7,
+	MLX5_IB_MMAP_DEVICE_MEM                 = 8,
 };
 
 enum {
diff --git a/include/uapi/rdma/mlx5_user_ioctl_cmds.h b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
index 521813d5348c..f7d685ef2d1f 100644
--- a/include/uapi/rdma/mlx5_user_ioctl_cmds.h
+++ b/include/uapi/rdma/mlx5_user_ioctl_cmds.h
@@ -40,5 +40,9 @@ enum mlx5_ib_create_flow_action_attrs {
 	MLX5_IB_ATTR_CREATE_FLOW_ACTION_FLAGS = (1U << UVERBS_ID_NS_SHIFT),
 };
 
-#endif
+enum mlx5_ib_alloc_dm_attrs {
+	MLX5_IB_ATTR_ALLOC_DM_RESP_START_OFFSET = (1U << UVERBS_ID_NS_SHIFT),
+	MLX5_IB_ATTR_ALLOC_DM_RESP_PAGE_INDEX,
+};
 
+#endif
-- 
cgit v1.2.3


From 6c29f57ea4751c4887627521027cd72aba831a97 Mon Sep 17 00:00:00 2001
From: Ariel Levkovich <lariel@mellanox.com>
Date: Thu, 5 Apr 2018 18:53:29 +0300
Subject: IB/mlx5: Device memory mr registration support

Adding mlx5_ib driver implementation for reg_dm_mr callback
which allows registering device memory (DM) as an MR for
local and remote access.

Signed-off-by: Ariel Levkovich <lariel@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
---
 drivers/infiniband/hw/mlx5/main.c    |  1 +
 drivers/infiniband/hw/mlx5/mlx5_ib.h |  9 +++++
 drivers/infiniband/hw/mlx5/mr.c      | 74 ++++++++++++++++++++++++++++++++++++
 3 files changed, 84 insertions(+)

(limited to 'drivers/infiniband/hw/mlx5/main.c')

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index e17eac32394c..4ead79513e3a 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -5226,6 +5226,7 @@ int mlx5_ib_stage_caps_init(struct mlx5_ib_dev *dev)
 	if (MLX5_CAP_DEV_MEM(mdev, memic)) {
 		dev->ib_dev.alloc_dm = mlx5_ib_alloc_dm;
 		dev->ib_dev.dealloc_dm = mlx5_ib_dealloc_dm;
+		dev->ib_dev.reg_dm_mr = mlx5_ib_reg_dm_mr;
 	}
 
 	dev->ib_dev.create_flow	= mlx5_ib_create_flow;
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 3e9b6548a96b..49a1aa0ff429 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -540,6 +540,12 @@ struct mlx5_ib_dm {
 
 #define MLX5_IB_MTT_PRESENT (MLX5_IB_MTT_READ | MLX5_IB_MTT_WRITE)
 
+#define MLX5_IB_DM_ALLOWED_ACCESS (IB_ACCESS_LOCAL_WRITE   |\
+				   IB_ACCESS_REMOTE_WRITE  |\
+				   IB_ACCESS_REMOTE_READ   |\
+				   IB_ACCESS_REMOTE_ATOMIC |\
+				   IB_ZERO_BASED)
+
 struct mlx5_ib_mr {
 	struct ib_mr		ibmr;
 	void			*descs;
@@ -1075,6 +1081,9 @@ struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
 			       struct ib_dm_alloc_attr *attr,
 			       struct uverbs_attr_bundle *attrs);
 int mlx5_ib_dealloc_dm(struct ib_dm *ibdm);
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs);
 
 #ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
 void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev);
diff --git a/drivers/infiniband/hw/mlx5/mr.c b/drivers/infiniband/hw/mlx5/mr.c
index d3f7ce97c3a5..d6350e7a2a47 100644
--- a/drivers/infiniband/hw/mlx5/mr.c
+++ b/drivers/infiniband/hw/mlx5/mr.c
@@ -1229,6 +1229,80 @@ static void set_mr_fileds(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr,
 	mr->access_flags = access_flags;
 }
 
+static struct ib_mr *mlx5_ib_get_memic_mr(struct ib_pd *pd, u64 memic_addr,
+					  u64 length, int acc)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
+	struct mlx5_core_dev *mdev = dev->mdev;
+	struct mlx5_ib_mr *mr;
+	void *mkc;
+	u32 *in;
+	int err;
+
+	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
+	if (!mr)
+		return ERR_PTR(-ENOMEM);
+
+	in = kzalloc(inlen, GFP_KERNEL);
+	if (!in) {
+		err = -ENOMEM;
+		goto err_free;
+	}
+
+	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+
+	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MEMIC & 0x3);
+	MLX5_SET(mkc, mkc, access_mode_4_2,
+		 (MLX5_MKC_ACCESS_MODE_MEMIC >> 2) & 0x7);
+	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+
+	MLX5_SET64(mkc, mkc, len, length);
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET64(mkc, mkc, start_addr,
+		   memic_addr - pci_resource_start(dev->mdev->pdev, 0));
+
+	err = mlx5_core_create_mkey(mdev, &mr->mmkey, in, inlen);
+	if (err)
+		goto err_in;
+
+	kfree(in);
+
+	mr->umem = NULL;
+	set_mr_fileds(dev, mr, 0, length, acc);
+
+	return &mr->ibmr;
+
+err_in:
+	kfree(in);
+
+err_free:
+	kfree(mr);
+
+	return ERR_PTR(err);
+}
+
+struct ib_mr *mlx5_ib_reg_dm_mr(struct ib_pd *pd, struct ib_dm *dm,
+				struct ib_dm_mr_attr *attr,
+				struct uverbs_attr_bundle *attrs)
+{
+	struct mlx5_ib_dm *mdm = to_mdm(dm);
+	u64 memic_addr;
+
+	if (attr->access_flags & ~MLX5_IB_DM_ALLOWED_ACCESS)
+		return ERR_PTR(-EINVAL);
+
+	memic_addr = mdm->dev_addr + attr->offset;
+
+	return mlx5_ib_get_memic_mr(pd, memic_addr, attr->length,
+				    attr->access_flags);
+}
+
 struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata)
-- 
cgit v1.2.3