From: Yishai Hadas <yishaih@xxxxxxxxxxxx>
Allow user space applications to accelerate send and receive
traffic which is typically handled by IPoIB ULP by creating
a UD QP with a given source QPN of the IPoIB UD QP.
UD QP with a given source QPN should basically be similar to
RAW QP from point of view of its created resources.
However,
- Its TIS should point to the source QPN.
- Modify can be done only on its state as the transport attributes
are managed by its source QP.
This patch manages below:
- Creating/destroying/modifying UD QP with a given source QPN.
Signed-off-by: Yishai Hadas <yishaih@xxxxxxxxxxxx>
Reviewed-by: Maor Gottlieb <maorg@xxxxxxxxxxxx>
Signed-off-by: Leon Romanovsky <leon@xxxxxxxxxx>
---
drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 +
drivers/infiniband/hw/mlx5/qp.c | 72 ++++++++++++++++++++++++++++--------
2 files changed, 59 insertions(+), 15 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index bdcf25410c99..ee8436033ba2 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -378,6 +378,7 @@ struct mlx5_ib_qp {
struct list_head cq_recv_list;
struct list_head cq_send_list;
u32 rate_limit;
+ u32 underlay_qpn;
};
struct mlx5_ib_cq_buf {
@@ -399,6 +400,7 @@ enum mlx5_ib_qp_flags {
MLX5_IB_QP_CAP_SCATTER_FCS = 1 << 7,
MLX5_IB_QP_RSS = 1 << 8,
MLX5_IB_QP_CVLAN_STRIPPING = 1 << 9,
+ MLX5_IB_QP_UNDERLAY = 1 << 10,
};
struct mlx5_umr_wr {
diff --git a/drivers/infiniband/hw/mlx5/qp.c b/drivers/infiniband/hw/mlx5/qp.c
index ebb6768684de..cc0f6d3b4268 100644
--- a/drivers/infiniband/hw/mlx5/qp.c
+++ b/drivers/infiniband/hw/mlx5/qp.c
@@ -34,6 +34,7 @@
#include <rdma/ib_umem.h>
#include <rdma/ib_cache.h>
#include <rdma/ib_user_verbs.h>
+#include <linux/mlx5/fs.h>
#include "mlx5_ib.h"
/* not supported currently */
@@ -453,7 +454,8 @@ static int set_user_buf_size(struct mlx5_ib_dev *dev,
return -EINVAL;
}
- if (attr->qp_type == IB_QPT_RAW_PACKET) {
+ if (attr->qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) {
base->ubuffer.buf_size = qp->rq.wqe_cnt << qp->rq.wqe_shift;
qp->raw_packet_qp.sq.ubuffer.buf_size = qp->sq.wqe_cnt << 6;
} else {
@@ -1021,12 +1023,16 @@ static int is_connected(enum ib_qp_type qp_type)
}
static int create_raw_packet_qp_tis(struct mlx5_ib_dev *dev,
+ struct mlx5_ib_qp *qp,
struct mlx5_ib_sq *sq, u32 tdn)
{
u32 in[MLX5_ST_SZ_DW(create_tis_in)] = {0};
void *tisc = MLX5_ADDR_OF(create_tis_in, in, ctx);
MLX5_SET(tisc, tisc, transport_domain, tdn);
+ if (qp->flags & MLX5_IB_QP_UNDERLAY)
+ MLX5_SET(tisc, tisc, underlay_qpn, qp->underlay_qpn);
+
return mlx5_core_create_tis(dev->mdev, in, sizeof(in), &sq->tisn);
}
@@ -1229,7 +1235,7 @@ static int create_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
u32 tdn = mucontext->tdn;
if (qp->sq.wqe_cnt) {
- err = create_raw_packet_qp_tis(dev, sq, tdn);
+ err = create_raw_packet_qp_tis(dev, qp, sq, tdn);
if (err)
return err;
@@ -1502,10 +1508,6 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
u32 *in;
int err;
- base = init_attr->qp_type == IB_QPT_RAW_PACKET ?
- &qp->raw_packet_qp.rq.base :
- &qp->trans_qp.base;
-
mutex_init(&qp->mutex);
spin_lock_init(&qp->sq.lock);
spin_lock_init(&qp->rq.lock);
@@ -1587,10 +1589,28 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
qp->wq_sig = !!(ucmd.flags & MLX5_QP_FLAG_SIGNATURE);
qp->scat_cqe = !!(ucmd.flags & MLX5_QP_FLAG_SCATTER_CQE);
+
+ if (init_attr->create_flags & IB_QP_CREATE_SOURCE_QPN) {
+ if (init_attr->qp_type != IB_QPT_UD ||
+ (MLX5_CAP_GEN(dev->mdev, port_type) !=
+ MLX5_CAP_PORT_TYPE_IB) ||
+ !mlx5_get_flow_namespace(dev->mdev, MLX5_FLOW_NAMESPACE_BYPASS)) {
+ mlx5_ib_dbg(dev, "Source QP option isn't supported\n");
+ return -EOPNOTSUPP;
+ }
+
+ qp->flags |= MLX5_IB_QP_UNDERLAY;
+ qp->underlay_qpn = init_attr->source_qpn;
+ }
} else {
qp->wq_sig = !!wq_signature;
}
+ base = (init_attr->qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) ?
+ &qp->raw_packet_qp.rq.base :
+ &qp->trans_qp.base;
+
qp->has_rq = qp_has_rq(init_attr);
err = set_rq_size(dev, &init_attr->cap, qp->has_rq,
qp, (pd && pd->uobject) ? &ucmd : NULL);
@@ -1741,7 +1761,8 @@ static int create_qp_common(struct mlx5_ib_dev *dev, struct ib_pd *pd,
qp->flags |= MLX5_IB_QP_LSO;
}
- if (init_attr->qp_type == IB_QPT_RAW_PACKET) {
+ if (init_attr->qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) {
qp->raw_packet_qp.sq.ubuffer.buf_addr = ucmd.sq_buf_addr;
raw_packet_qp_copy_info(qp, &qp->raw_packet_qp);
err = create_raw_packet_qp(dev, qp, in, pd);
@@ -1893,7 +1914,7 @@ static int modify_raw_packet_qp(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp,
static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
{
struct mlx5_ib_cq *send_cq, *recv_cq;
- struct mlx5_ib_qp_base *base = &qp->trans_qp.base;
+ struct mlx5_ib_qp_base *base;
unsigned long flags;
int err;
@@ -1902,12 +1923,14 @@ static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
return;
}
- base = qp->ibqp.qp_type == IB_QPT_RAW_PACKET ?
+ base = (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) ?
&qp->raw_packet_qp.rq.base :
&qp->trans_qp.base;
if (qp->state != IB_QPS_RESET) {
- if (qp->ibqp.qp_type != IB_QPT_RAW_PACKET) {
+ if (qp->ibqp.qp_type != IB_QPT_RAW_PACKET &&
+ !(qp->flags & MLX5_IB_QP_UNDERLAY)) {
err = mlx5_core_qp_modify(dev->mdev,
MLX5_CMD_OP_2RST_QP, 0,
NULL, &base->mqp);
@@ -1946,7 +1969,8 @@ static void destroy_qp_common(struct mlx5_ib_dev *dev, struct mlx5_ib_qp *qp)
mlx5_ib_unlock_cqs(send_cq, recv_cq);
spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags);
- if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) {
+ if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) {
destroy_raw_packet_qp(dev, qp);
} else {
err = mlx5_core_destroy_qp(dev->mdev, &base->mqp);
@@ -2702,7 +2726,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
if (is_sqp(ibqp->qp_type)) {
context->mtu_msgmax = (IB_MTU_256 << 5) | 8;
- } else if (ibqp->qp_type == IB_QPT_UD ||
+ } else if ((ibqp->qp_type == IB_QPT_UD &&
+ !(qp->flags & MLX5_IB_QP_UNDERLAY)) ||
ibqp->qp_type == MLX5_IB_QPT_REG_UMR) {
context->mtu_msgmax = (IB_MTU_4096 << 5) | 12;
} else if (attr_mask & IB_QP_PATH_MTU) {
@@ -2799,6 +2824,11 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
u8 port_num = (attr_mask & IB_QP_PORT ? attr->port_num :
qp->port) - 1;
+
+ /* Underlay port should be used - index 0 function per port */
+ if (qp->flags & MLX5_IB_QP_UNDERLAY)
+ port_num = 0;
+
mibport = &dev->port[port_num];
context->qp_counter_set_usr_page |=
cpu_to_be32((u32)(mibport->cnts.set_id) << 24);
@@ -2824,7 +2854,8 @@ static int __mlx5_ib_modify_qp(struct ib_qp *ibqp,
optpar = ib_mask_to_mlx5_opt(attr_mask);
optpar &= opt_mask[mlx5_cur][mlx5_new][mlx5_st];
- if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) {
+ if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) {
struct mlx5_modify_raw_qp_param raw_qp_param = {};
raw_qp_param.operation = op;
@@ -2913,7 +2944,13 @@ int mlx5_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
ll = dev->ib_dev.get_link_layer(&dev->ib_dev, port);
}
- if (qp_type != MLX5_IB_QPT_REG_UMR &&
+ if (qp->flags & MLX5_IB_QP_UNDERLAY) {
+ if (attr_mask & ~(IB_QP_STATE | IB_QP_CUR_STATE)) {
+ mlx5_ib_dbg(dev, "invalid attr_mask 0x%x when underlay QP is used\n",
+ attr_mask);
+ goto out;
+ }
+ } else if (qp_type != MLX5_IB_QPT_REG_UMR &&
!ib_modify_qp_is_ok(cur_state, new_state, qp_type, attr_mask, ll)) {
mlx5_ib_dbg(dev, "invalid QP state transition from %d to %d, qp_type %d, attr_mask 0x%x\n",
cur_state, new_state, ibqp->qp_type, attr_mask);
@@ -4477,9 +4514,14 @@ int mlx5_ib_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
return mlx5_ib_gsi_query_qp(ibqp, qp_attr, qp_attr_mask,
qp_init_attr);
+ /* Not all of output fields are applicable, make sure to zero them */
+ memset(qp_init_attr, 0, sizeof(*qp_init_attr));
+ memset(qp_attr, 0, sizeof(*qp_attr));
+
mutex_lock(&qp->mutex);
- if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET) {
+ if (qp->ibqp.qp_type == IB_QPT_RAW_PACKET ||
+ qp->flags & MLX5_IB_QP_UNDERLAY) {
err = query_raw_packet_qp_state(dev, qp, &raw_packet_qp_state);
if (err)
goto out;
--
2.12.2
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
