[PATCH 0/3] Add cross-channel support

[PATCH 0/3] Add cross-channel support

[Leon Romanovsky]

From: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

This patchset adds cross-channel support.

The cross-channel feature allows to execute WQEs that involve
synchronization of I/O operations’ on different QPs.

This capability enables to program complex flows with a single
function call, hereby significantly reducing overhead associated
with I/O processing.

Leon Romanovsky (3):
  IB/core: Align coding style of ib_device_cap_flags structure
  IB/core: Add cross-channel support
  IB/mlx5: Add driver cross-channel support

 drivers/infiniband/core/uverbs_cmd.c |  5 ++-
 drivers/infiniband/hw/mlx5/cq.c      |  7 +++-
 drivers/infiniband/hw/mlx5/main.c    |  3 ++
 drivers/infiniband/hw/mlx5/mlx5_ib.h | 12 ++++++
 drivers/infiniband/hw/mlx5/qp.c      | 29 +++++++++++++++
 include/linux/mlx5/qp.h              |  3 ++
 include/rdma/ib_verbs.h              | 71 +++++++++++++++++++++---------------
 7 files changed, 98 insertions(+), 32 deletions(-)

-- 
1.7.12.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 1/3] IB/core: Align coding style of ib_device_cap_flags structure

[Leon Romanovsky]

From: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

Modify enum ib_device_cap_flags such that other patches which add new
enum values pass strict checkpatch.pl checks.

Reviewed-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 include/rdma/ib_verbs.h | 60 ++++++++++++++++++++++++-------------------------
 1 file changed, 30 insertions(+), 30 deletions(-)

diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 9a68a19..bcf40ad 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -105,24 +105,24 @@ enum rdma_link_layer {
 };
 
 enum ib_device_cap_flags {
-	IB_DEVICE_RESIZE_MAX_WR		= 1,
-	IB_DEVICE_BAD_PKEY_CNTR		= (1<<1),
-	IB_DEVICE_BAD_QKEY_CNTR		= (1<<2),
-	IB_DEVICE_RAW_MULTI		= (1<<3),
-	IB_DEVICE_AUTO_PATH_MIG		= (1<<4),
-	IB_DEVICE_CHANGE_PHY_PORT	= (1<<5),
-	IB_DEVICE_UD_AV_PORT_ENFORCE	= (1<<6),
-	IB_DEVICE_CURR_QP_STATE_MOD	= (1<<7),
-	IB_DEVICE_SHUTDOWN_PORT		= (1<<8),
-	IB_DEVICE_INIT_TYPE		= (1<<9),
-	IB_DEVICE_PORT_ACTIVE_EVENT	= (1<<10),
-	IB_DEVICE_SYS_IMAGE_GUID	= (1<<11),
-	IB_DEVICE_RC_RNR_NAK_GEN	= (1<<12),
-	IB_DEVICE_SRQ_RESIZE		= (1<<13),
-	IB_DEVICE_N_NOTIFY_CQ		= (1<<14),
-	IB_DEVICE_LOCAL_DMA_LKEY	= (1<<15),
-	IB_DEVICE_RESERVED		= (1<<16), /* old SEND_W_INV */
-	IB_DEVICE_MEM_WINDOW		= (1<<17),
+	IB_DEVICE_RESIZE_MAX_WR		= (1 << 0),
+	IB_DEVICE_BAD_PKEY_CNTR		= (1 << 1),
+	IB_DEVICE_BAD_QKEY_CNTR		= (1 << 2),
+	IB_DEVICE_RAW_MULTI		= (1 << 3),
+	IB_DEVICE_AUTO_PATH_MIG		= (1 << 4),
+	IB_DEVICE_CHANGE_PHY_PORT	= (1 << 5),
+	IB_DEVICE_UD_AV_PORT_ENFORCE	= (1 << 6),
+	IB_DEVICE_CURR_QP_STATE_MOD	= (1 << 7),
+	IB_DEVICE_SHUTDOWN_PORT		= (1 << 8),
+	IB_DEVICE_INIT_TYPE		= (1 << 9),
+	IB_DEVICE_PORT_ACTIVE_EVENT	= (1 << 10),
+	IB_DEVICE_SYS_IMAGE_GUID	= (1 << 11),
+	IB_DEVICE_RC_RNR_NAK_GEN	= (1 << 12),
+	IB_DEVICE_SRQ_RESIZE		= (1 << 13),
+	IB_DEVICE_N_NOTIFY_CQ		= (1 << 14),
+	IB_DEVICE_LOCAL_DMA_LKEY	= (1 << 15),
+	IB_DEVICE_RESERVED		= (1 << 16), /* old SEND_W_INV */
+	IB_DEVICE_MEM_WINDOW		= (1 << 17),
 	/*
 	 * Devices should set IB_DEVICE_UD_IP_SUM if they support
 	 * insertion of UDP and TCP checksum on outgoing UD IPoIB
@@ -130,18 +130,18 @@ enum ib_device_cap_flags {
 	 * incoming messages.  Setting this flag implies that the
 	 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
 	 */
-	IB_DEVICE_UD_IP_CSUM		= (1<<18),
-	IB_DEVICE_UD_TSO		= (1<<19),
-	IB_DEVICE_XRC			= (1<<20),
-	IB_DEVICE_MEM_MGT_EXTENSIONS	= (1<<21),
-	IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1<<22),
-	IB_DEVICE_MEM_WINDOW_TYPE_2A	= (1<<23),
-	IB_DEVICE_MEM_WINDOW_TYPE_2B	= (1<<24),
-	IB_DEVICE_RC_IP_CSUM		= (1<<25),
-	IB_DEVICE_RAW_IP_CSUM		= (1<<26),
-	IB_DEVICE_MANAGED_FLOW_STEERING = (1<<29),
-	IB_DEVICE_SIGNATURE_HANDOVER	= (1<<30),
-	IB_DEVICE_ON_DEMAND_PAGING	= (1<<31),
+	IB_DEVICE_UD_IP_CSUM		= (1 << 18),
+	IB_DEVICE_UD_TSO		= (1 << 19),
+	IB_DEVICE_XRC			= (1 << 20),
+	IB_DEVICE_MEM_MGT_EXTENSIONS	= (1 << 21),
+	IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1 << 22),
+	IB_DEVICE_MEM_WINDOW_TYPE_2A	= (1 << 23),
+	IB_DEVICE_MEM_WINDOW_TYPE_2B	= (1 << 24),
+	IB_DEVICE_RC_IP_CSUM		= (1 << 25),
+	IB_DEVICE_RAW_IP_CSUM		= (1 << 26),
+	IB_DEVICE_MANAGED_FLOW_STEERING = (1 << 29),
+	IB_DEVICE_SIGNATURE_HANDOVER	= (1 << 30),
+	IB_DEVICE_ON_DEMAND_PAGING	= (1 << 31),
 };
 
 enum ib_signature_prot_cap {
-- 
1.7.12.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 2/3] IB/core: Add cross-channel support

[Leon Romanovsky]

From: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

The cross-channel feature allows to execute WQEs that involve
synchronization of I/O operations’ on different QPs.

This capability enables to program complex flows with a single
function call, hereby significantly reducing overhead associated
with I/O processing.

Cross-channel operations support is indicated by HCA capability
information.

The queue pairs can be configured to work as a “sync master queue”
or “sync slave queues”.

The added flags are:

1. Device capability flag IB_DEVICE_CROSS_CHANNEL for the
   devices that can perform cross-channel operations.

2. CQ property flag IB_CQ_FLAGS_IGNORE_OVERRUN to disable CQ overrun
   check. This check is useless in cross-channel scenario.

3. QP property flags to indicate if queues are slave or master:
   * IB_QP_CREATE_MANAGED_SEND indicates that posted send work requests
     will not be executed immediately and requires enabling.
   * IB_QP_CREATE_MANAGED_RECV indicates that posted receive work
     requests will not be executed immediately and requires enabling.
   * IB_QP_CREATE_CROSS_CHANNEL declares the QP to work in cross-channel
     mode. If IB_QP_CREATE_MANAGED_SEND and IB_QP_CREATE_MANAGED_RECV are
     not provided, this QP will be sync master queue, else it will be sync
     slave.

Reviewed-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/core/uverbs_cmd.c |  5 ++++-
 include/rdma/ib_verbs.h              | 11 +++++++++++
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/core/uverbs_cmd.c b/drivers/infiniband/core/uverbs_cmd.c
index 94816ae..0e9711f 100644
--- a/drivers/infiniband/core/uverbs_cmd.c
+++ b/drivers/infiniband/core/uverbs_cmd.c
@@ -1843,7 +1843,10 @@ static int create_qp(struct ib_uverbs_file *file,
 		      sizeof(cmd->create_flags))
 		attr.create_flags = cmd->create_flags;
 
-	if (attr.create_flags & ~IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK) {
+	if (attr.create_flags & ~(IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK |
+				IB_QP_CREATE_CROSS_CHANNEL |
+				IB_QP_CREATE_MANAGED_SEND |
+				IB_QP_CREATE_MANAGED_RECV)) {
 		ret = -EINVAL;
 		goto err_put;
 	}
diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index bcf40ad..370fbdf 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -139,6 +139,13 @@ enum ib_device_cap_flags {
 	IB_DEVICE_MEM_WINDOW_TYPE_2B	= (1 << 24),
 	IB_DEVICE_RC_IP_CSUM		= (1 << 25),
 	IB_DEVICE_RAW_IP_CSUM		= (1 << 26),
+	/*
+	 * Devices should set IB_DEVICE_CROSS_CHANNEL if they
+	 * support execution of WQEs that involve synchronization
+	 * of I/O operations with single completion queue managed
+	 * by hardware.
+	 */
+	IB_DEVICE_CROSS_CHANNEL		= (1 << 27),
 	IB_DEVICE_MANAGED_FLOW_STEERING = (1 << 29),
 	IB_DEVICE_SIGNATURE_HANDOVER	= (1 << 30),
 	IB_DEVICE_ON_DEMAND_PAGING	= (1 << 31),
@@ -184,6 +191,7 @@ struct ib_odp_caps {
 
 enum ib_cq_creation_flags {
 	IB_CQ_FLAGS_TIMESTAMP_COMPLETION   = 1 << 0,
+	IB_CQ_FLAGS_IGNORE_OVERRUN	   = 1 << 1,
 };
 
 struct ib_cq_init_attr {
@@ -866,6 +874,9 @@ enum ib_qp_type {
 enum ib_qp_create_flags {
 	IB_QP_CREATE_IPOIB_UD_LSO		= 1 << 0,
 	IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
+	IB_QP_CREATE_CROSS_CHANNEL              = 1 << 2,
+	IB_QP_CREATE_MANAGED_SEND               = 1 << 3,
+	IB_QP_CREATE_MANAGED_RECV               = 1 << 4,
 	IB_QP_CREATE_NETIF_QP			= 1 << 5,
 	IB_QP_CREATE_SIGNATURE_EN		= 1 << 6,
 	IB_QP_CREATE_USE_GFP_NOIO		= 1 << 7,
-- 
1.7.12.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH 3/3] IB/mlx5: Add driver cross-channel support

[Leon Romanovsky]

From: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>

Add support of cross-channel functionality to mlx5
driver. This includes ability to ignore overrun for CQ
which intended for cross-channel, export device capability and
configure the QP to be sync master/slave queues.

The cross-channel enabled QP supports combination of
three possible properties:
* WQE processing on the receive queue of this QP
* WQE processing on the send queue of this QP
* WQE are supported on the send queue

Reviewed-by: Sagi Grimberg <sagig-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Leon Romanovsky <leonro-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/mlx5/cq.c      |  7 ++++++-
 drivers/infiniband/hw/mlx5/main.c    |  3 +++
 drivers/infiniband/hw/mlx5/mlx5_ib.h | 12 ++++++++++++
 drivers/infiniband/hw/mlx5/qp.c      | 29 +++++++++++++++++++++++++++++
 include/linux/mlx5/qp.h              |  3 +++
 5 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index 3dfd287..c363b71 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -760,7 +760,7 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
 	int eqn;
 	int err;
 
-	if (attr->flags)
+	if (check_cq_create_flags(attr->flags))
 		return ERR_PTR(-EINVAL);
 
 	if (entries < 0)
@@ -779,6 +779,7 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
 	spin_lock_init(&cq->lock);
 	cq->resize_buf = NULL;
 	cq->resize_umem = NULL;
+	cq->create_flags = attr->flags;
 
 	if (context) {
 		err = create_cq_user(dev, udata, context, cq, entries,
@@ -796,6 +797,10 @@ struct ib_cq *mlx5_ib_create_cq(struct ib_device *ibdev,
 
 	cq->cqe_size = cqe_size;
 	cqb->ctx.cqe_sz_flags = cqe_sz_to_mlx_sz(cqe_size) << 5;
+
+	if (cq->create_flags & IB_CQ_FLAGS_IGNORE_OVERRUN)
+		cqb->ctx.cqe_sz_flags |= (1 << 1);
+
 	cqb->ctx.log_sz_usr_page = cpu_to_be32((ilog2(entries) << 24) | index);
 	err = mlx5_vector2eqn(dev->mdev, vector, &eqn, &irqn);
 	if (err)
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index bdd60a6..be0d994 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -300,6 +300,9 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 	props->odp_caps = dev->odp_caps;
 #endif
 
+	if (MLX5_CAP_GEN(mdev, cd))
+		props->device_cap_flags |= IB_DEVICE_CROSS_CHANNEL;
+
 	return 0;
 }
 
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 6333472..7f9c6bf 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -242,6 +242,9 @@ struct mlx5_ib_cq_buf {
 enum mlx5_ib_qp_flags {
 	MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK     = 1 << 0,
 	MLX5_IB_QP_SIGNATURE_HANDLING           = 1 << 1,
+	MLX5_IB_QP_CROSS_CHANNEL		= 1 << 2,
+	MLX5_IB_QP_MANAGED_SEND			= 1 << 3,
+	MLX5_IB_QP_MANAGED_RECV			= 1 << 4,
 };
 
 struct mlx5_umr_wr {
@@ -284,6 +287,7 @@ struct mlx5_ib_cq {
 	struct mlx5_ib_cq_buf  *resize_buf;
 	struct ib_umem	       *resize_umem;
 	int			cqe_size;
+	u32			create_flags;
 };
 
 struct mlx5_ib_srq {
@@ -662,4 +666,12 @@ static inline int is_qp1(enum ib_qp_type qp_type)
 #define MLX5_MAX_UMR_SHIFT 16
 #define MLX5_MAX_UMR_PAGES (1 << MLX5_MAX_UMR_SHIFT)
 
+static inline u32 check_cq_create_flags(u32 flags)
+{
+	/*
+	 * It returns non-zero value for unsupported CQ
+	 * create flags, otherwise it returns zero.
+	 */
+	return (flags & ~IB_CQ_FLAGS_IGNORE_OVERRUN);
+}
 #endif /* MLX5_IB_H */
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index 307bdbc..54ceba8 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -880,6 +880,21 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 		}
 	}
 
+	if (init_attr->create_flags &
+			(IB_QP_CREATE_CROSS_CHANNEL |
+			 IB_QP_CREATE_MANAGED_SEND |
+			 IB_QP_CREATE_MANAGED_RECV)) {
+		if (!MLX5_CAP_GEN(mdev, cd)) {
+			mlx5_ib_dbg(dev, "cross-channel isn't supported\n");
+			return -EINVAL;
+		}
+		if (init_attr->create_flags & IB_QP_CREATE_CROSS_CHANNEL)
+			qp->flags |= MLX5_IB_QP_CROSS_CHANNEL;
+		if (init_attr->create_flags & IB_QP_CREATE_MANAGED_SEND)
+			qp->flags |= MLX5_IB_QP_MANAGED_SEND;
+		if (init_attr->create_flags & IB_QP_CREATE_MANAGED_RECV)
+			qp->flags |= MLX5_IB_QP_MANAGED_RECV;
+	}
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = MLX5_WQE_CTRL_CQ_UPDATE;
 
@@ -954,6 +969,13 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
 	if (qp->flags & MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK)
 		in->ctx.flags_pd |= cpu_to_be32(MLX5_QP_BLOCK_MCAST);
 
+	if (qp->flags & MLX5_IB_QP_CROSS_CHANNEL)
+		in->ctx.params2 |= cpu_to_be32(MLX5_QP_BIT_CC_MASTER);
+	if (qp->flags & MLX5_IB_QP_MANAGED_SEND)
+		in->ctx.params2 |= cpu_to_be32(MLX5_QP_BIT_CC_SLAVE_SEND);
+	if (qp->flags & MLX5_IB_QP_MANAGED_RECV)
+		in->ctx.params2 |= cpu_to_be32(MLX5_QP_BIT_CC_SLAVE_RECV);
+
 	if (qp->scat_cqe && is_connected(init_attr->qp_type)) {
 		int rcqe_sz;
 		int scqe_sz;
@@ -3110,6 +3132,13 @@ int mlx5_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr, int qp_attr
 	if (qp->flags & MLX5_IB_QP_BLOCK_MULTICAST_LOOPBACK)
 		qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
 
+	if (qp->flags & MLX5_IB_QP_CROSS_CHANNEL)
+		qp_init_attr->create_flags |= IB_QP_CREATE_CROSS_CHANNEL;
+	if (qp->flags & MLX5_IB_QP_MANAGED_SEND)
+		qp_init_attr->create_flags |= IB_QP_CREATE_MANAGED_SEND;
+	if (qp->flags & MLX5_IB_QP_MANAGED_RECV)
+		qp_init_attr->create_flags |= IB_QP_CREATE_MANAGED_RECV;
+
 	qp_init_attr->sq_sig_type = qp->sq_signal_bits & MLX5_WQE_CTRL_CQ_UPDATE ?
 		IB_SIGNAL_ALL_WR : IB_SIGNAL_REQ_WR;
 
diff --git a/include/linux/mlx5/qp.h b/include/linux/mlx5/qp.h
index f079fb1..511e03b 100644
--- a/include/linux/mlx5/qp.h
+++ b/include/linux/mlx5/qp.h
@@ -130,6 +130,9 @@ enum {
 	MLX5_QP_BIT_RWE				= 1 << 14,
 	MLX5_QP_BIT_RAE				= 1 << 13,
 	MLX5_QP_BIT_RIC				= 1 <<	4,
+	MLX5_QP_BIT_CC_SLAVE_RECV		= 1 <<  2,
+	MLX5_QP_BIT_CC_SLAVE_SEND		= 1 <<  1,
+	MLX5_QP_BIT_CC_MASTER			= 1 <<  0
 };
 
 enum {
-- 
1.7.12.4

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html