IB/mlx4: Pass send queue sizes from userspace to kernel

Pass the number of WQEs for the send queue and their size from userspace
to the kernel to avoid having to keep the QP size calculations in sync
between the kernel driver and libmlx4.  This fixes a bug seen with the
current mlx4_ib driver and current libmlx4 caused by a difference in the
calculated sizes for SQ WQEs.  Also, this gives more flexibility for
userspace to experiment with using multiple WQE BBs for a single SQ WQE.

Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 5706f988e2ec79a00fd6cca1d4286c980712d8c8..a824bc5f79fdf5ee60ea4b580cef06819819ab28 100644 (file)
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -188,14 +188,32 @@ static int send_wqe_overhead(enum ib_qp_type type)
        }
 }
 
-static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
-                      enum ib_qp_type type, struct mlx4_ib_qp *qp)
+static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
+                      struct mlx4_ib_qp *qp)
 {
-       /* Sanity check QP size before proceeding */
+       /* Sanity check RQ size before proceeding */
+       if (cap->max_recv_wr  > dev->dev->caps.max_wqes  ||
+           cap->max_recv_sge > dev->dev->caps.max_rq_sg)
+               return -EINVAL;
+
+       qp->rq.max = cap->max_recv_wr ? roundup_pow_of_two(cap->max_recv_wr) : 0;
+
+       qp->rq.wqe_shift = ilog2(roundup_pow_of_two(cap->max_recv_sge *
+                                                   sizeof (struct mlx4_wqe_data_seg)));
+       qp->rq.max_gs    = (1 << qp->rq.wqe_shift) / sizeof (struct mlx4_wqe_data_seg);
+
+       cap->max_recv_wr  = qp->rq.max;
+       cap->max_recv_sge = qp->rq.max_gs;
+
+       return 0;
+}
+
+static int set_kernel_sq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
+                             enum ib_qp_type type, struct mlx4_ib_qp *qp)
+{
+       /* Sanity check SQ size before proceeding */
        if (cap->max_send_wr     > dev->dev->caps.max_wqes  ||
-           cap->max_recv_wr     > dev->dev->caps.max_wqes  ||
            cap->max_send_sge    > dev->dev->caps.max_sq_sg ||
-           cap->max_recv_sge    > dev->dev->caps.max_rq_sg ||
            cap->max_inline_data + send_wqe_overhead(type) +
            sizeof (struct mlx4_wqe_inline_seg) > dev->dev->caps.max_sq_desc_sz)
                return -EINVAL;
@@ -208,12 +226,7 @@ static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
            cap->max_send_sge + 2 > dev->dev->caps.max_sq_sg)
                return -EINVAL;
 
-       qp->rq.max = cap->max_recv_wr ? roundup_pow_of_two(cap->max_recv_wr) : 0;
-       qp->sq.max = cap->max_send_wr ? roundup_pow_of_two(cap->max_send_wr) : 0;
-
-       qp->rq.wqe_shift = ilog2(roundup_pow_of_two(cap->max_recv_sge *
-                                                   sizeof (struct mlx4_wqe_data_seg)));
-       qp->rq.max_gs    = (1 << qp->rq.wqe_shift) / sizeof (struct mlx4_wqe_data_seg);
+       qp->sq.max = cap->max_send_wr ? roundup_pow_of_two(cap->max_send_wr) : 1;
 
        qp->sq.wqe_shift = ilog2(roundup_pow_of_two(max(cap->max_send_sge *
                                                        sizeof (struct mlx4_wqe_data_seg),
@@ -233,16 +246,26 @@ static int set_qp_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
                qp->sq.offset = 0;
        }
 
-       cap->max_send_wr  = qp->sq.max;
-       cap->max_recv_wr  = qp->rq.max;
-       cap->max_send_sge = qp->sq.max_gs;
-       cap->max_recv_sge = qp->rq.max_gs;
+       cap->max_send_wr     = qp->sq.max;
+       cap->max_send_sge    = qp->sq.max_gs;
        cap->max_inline_data = (1 << qp->sq.wqe_shift) - send_wqe_overhead(type) -
                sizeof (struct mlx4_wqe_inline_seg);
 
        return 0;
 }
 
+static int set_user_sq_size(struct mlx4_ib_qp *qp,
+                           struct mlx4_ib_create_qp *ucmd)
+{
+       qp->sq.max       = 1 << ucmd->log_sq_bb_count;
+       qp->sq.wqe_shift = ucmd->log_sq_stride;
+
+       qp->buf_size = (qp->rq.max << qp->rq.wqe_shift) +
+               (qp->sq.max << qp->sq.wqe_shift);
+
+       return 0;
+}
+
 static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
                            struct ib_qp_init_attr *init_attr,
                            struct ib_udata *udata, int sqpn, struct mlx4_ib_qp *qp)
@@ -264,7 +287,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
        qp->sq.head         = 0;
        qp->sq.tail         = 0;
 
-       err = set_qp_size(dev, &init_attr->cap, init_attr->qp_type, qp);
+       err = set_rq_size(dev, &init_attr->cap, qp);
        if (err)
                goto err;
 
@@ -276,6 +299,10 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
                        goto err;
                }
 
+               err = set_user_sq_size(qp, &ucmd);
+               if (err)
+                       goto err;
+
                qp->umem = ib_umem_get(pd->uobject->context, ucmd.buf_addr,
                                       qp->buf_size, 0);
                if (IS_ERR(qp->umem)) {
@@ -297,6 +324,10 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
                if (err)
                        goto err_mtt;
        } else {
+               err = set_kernel_sq_size(dev, &init_attr->cap, init_attr->qp_type, qp);
+               if (err)
+                       goto err;
+
                err = mlx4_ib_db_alloc(dev, &qp->db, 0);
                if (err)
                        goto err;

diff --git a/drivers/infiniband/hw/mlx4/user.h b/drivers/infiniband/hw/mlx4/user.h
index 5b8eddc9fa83e471c36bdf4b5a2764f1a1cd8608..88c72d56368b8a22c256eaf44dd22a8b8a05c809 100644 (file)
--- a/drivers/infiniband/hw/mlx4/user.h
+++ b/drivers/infiniband/hw/mlx4/user.h
@@ -39,7 +39,7 @@
  * Increment this value if any changes that break userspace ABI
  * compatibility are made.
  */
-#define MLX4_IB_UVERBS_ABI_VERSION     1
+#define MLX4_IB_UVERBS_ABI_VERSION     2
 
 /*
  * Make sure that all structs defined in this file remain laid out so
@@ -87,6 +87,9 @@ struct mlx4_ib_create_srq_resp {
 struct mlx4_ib_create_qp {
        __u64   buf_addr;
        __u64   db_addr;
+        __u8   log_sq_bb_count;
+        __u8   log_sq_stride;
+        __u8   reserved[6];
 };
 
 #endif /* MLX4_IB_USER_H */