}
}
-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;
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),
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)
{
- struct mlx4_wqe_ctrl_seg *ctrl;
int err;
- int i;
mutex_init(&qp->mutex);
spin_lock_init(&qp->sq.lock);
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;
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)) {
if (err)
goto err_mtt;
- err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context),
- ucmd.db_addr, &qp->db);
- if (err)
- goto err_mtt;
+ if (!init_attr->srq) {
+ err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context),
+ ucmd.db_addr, &qp->db);
+ if (err)
+ goto err_mtt;
+ }
} else {
- err = mlx4_ib_db_alloc(dev, &qp->db, 0);
+ err = set_kernel_sq_size(dev, &init_attr->cap, init_attr->qp_type, qp);
if (err)
goto err;
- *qp->db.db = 0;
+ if (!init_attr->srq) {
+ err = mlx4_ib_db_alloc(dev, &qp->db, 0);
+ if (err)
+ goto err;
+
+ *qp->db.db = 0;
+ }
if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2, &qp->buf)) {
err = -ENOMEM;
if (err)
goto err_mtt;
- for (i = 0; i < qp->sq.max; ++i) {
- ctrl = get_send_wqe(qp, i);
- ctrl->owner_opcode = cpu_to_be32(1 << 31);
- }
-
qp->sq.wrid = kmalloc(qp->sq.max * sizeof (u64), GFP_KERNEL);
qp->rq.wrid = kmalloc(qp->rq.max * sizeof (u64), GFP_KERNEL);
return 0;
err_wrid:
- if (pd->uobject)
+ if (pd->uobject && !init_attr->srq)
mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &qp->db);
else {
kfree(qp->sq.wrid);
mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
err_db:
- if (!pd->uobject)
+ if (!pd->uobject && !init_attr->srq)
mlx4_ib_db_free(dev, &qp->db);
err:
mlx4_mtt_cleanup(dev->dev, &qp->mtt);
if (is_user) {
- mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context),
- &qp->db);
+ if (!qp->ibqp.srq)
+ mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context),
+ &qp->db);
ib_umem_release(qp->umem);
} else {
kfree(qp->sq.wrid);
kfree(qp->rq.wrid);
mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
- mlx4_ib_db_free(dev, &qp->db);
+ if (!qp->ibqp.srq)
+ mlx4_ib_db_free(dev, &qp->db);
}
}
optpar |= MLX4_QP_OPTPAR_PKEY_INDEX;
}
- if (attr_mask & IB_QP_RNR_RETRY) {
- context->params1 |= cpu_to_be32(attr->rnr_retry << 13);
- optpar |= MLX4_QP_OPTPAR_RNR_RETRY;
- }
-
if (attr_mask & IB_QP_AV) {
if (mlx4_set_path(dev, &attr->ah_attr, &context->pri_path,
attr_mask & IB_QP_PORT ? attr->port_num : qp->port)) {
context->pd = cpu_to_be32(to_mpd(ibqp->pd)->pdn);
context->params1 = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28);
+
+ if (attr_mask & IB_QP_RNR_RETRY) {
+ context->params1 |= cpu_to_be32(attr->rnr_retry << 13);
+ optpar |= MLX4_QP_OPTPAR_RNR_RETRY;
+ }
+
if (attr_mask & IB_QP_RETRY_CNT) {
context->params1 |= cpu_to_be32(attr->retry_cnt << 16);
optpar |= MLX4_QP_OPTPAR_RETRY_COUNT;
if (ibqp->srq)
context->srqn = cpu_to_be32(1 << 24 | to_msrq(ibqp->srq)->msrq.srqn);
- if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+ if (!ibqp->srq && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
context->db_rec_addr = cpu_to_be64(qp->db.dma);
if (cur_state == IB_QPS_INIT &&
else
sqd_event = 0;
+ /*
+ * Before passing a kernel QP to the HW, make sure that the
+ * ownership bits of the send queue are set so that the
+ * hardware doesn't start processing stale work requests.
+ */
+ if (!ibqp->uobject && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
+ struct mlx4_wqe_ctrl_seg *ctrl;
+ int i;
+
+ for (i = 0; i < qp->sq.max; ++i) {
+ ctrl = get_send_wqe(qp, i);
+ ctrl->owner_opcode = cpu_to_be32(1 << 31);
+ }
+ }
+
err = mlx4_qp_modify(dev->dev, &qp->mtt, to_mlx4_state(cur_state),
to_mlx4_state(new_state), context, optpar,
sqd_event, &qp->mqp);
qp->rq.tail = 0;
qp->sq.head = 0;
qp->sq.tail = 0;
- *qp->db.db = 0;
+ if (!ibqp->srq)
+ *qp->db.db = 0;
}
out:
*/
wmb();
- if (wr->opcode < 0 || wr->opcode > ARRAY_SIZE(mlx4_ib_opcode)) {
+ if (wr->opcode < 0 || wr->opcode >= ARRAY_SIZE(mlx4_ib_opcode)) {
err = -EINVAL;
goto out;
}
projects / linux-2.6-block.git / blobdiff