Cache the needed umr_fence and set the wqe ctrl segmennt
accordingly.
Signed-off-by: Max Gurtovoy <maxg@mellanox.com>
Acked-by: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Doug Ledford <dledford@redhat.com>
+static u8 mlx5_get_umr_fence(u8 umr_fence_cap)
+{
+ switch (umr_fence_cap) {
+ case MLX5_CAP_UMR_FENCE_NONE:
+ return MLX5_FENCE_MODE_NONE;
+ case MLX5_CAP_UMR_FENCE_SMALL:
+ return MLX5_FENCE_MODE_INITIATOR_SMALL;
+ default:
+ return MLX5_FENCE_MODE_STRONG_ORDERING;
+ }
+}
+
static int create_dev_resources(struct mlx5_ib_resources *devr)
{
struct ib_srq_init_attr attr;
static int create_dev_resources(struct mlx5_ib_resources *devr)
{
struct ib_srq_init_attr attr;
mlx5_ib_internal_fill_odp_caps(dev);
mlx5_ib_internal_fill_odp_caps(dev);
+ dev->umr_fence = mlx5_get_umr_fence(MLX5_CAP_GEN(mdev, umr_fence));
+
if (MLX5_CAP_GEN(mdev, imaicl)) {
dev->ib_dev.alloc_mw = mlx5_ib_alloc_mw;
dev->ib_dev.dealloc_mw = mlx5_ib_dealloc_mw;
if (MLX5_CAP_GEN(mdev, imaicl)) {
dev->ib_dev.alloc_mw = mlx5_ib_alloc_mw;
dev->ib_dev.dealloc_mw = mlx5_ib_dealloc_mw;
struct mlx5_ib_wq rq;
u8 sq_signal_bits;
struct mlx5_ib_wq rq;
u8 sq_signal_bits;
struct mlx5_ib_wq sq;
/* serialize qp state modifications
struct mlx5_ib_wq sq;
/* serialize qp state modifications
struct mlx5_ib_port *port;
struct mlx5_sq_bfreg bfreg;
struct mlx5_sq_bfreg fp_bfreg;
struct mlx5_ib_port *port;
struct mlx5_sq_bfreg bfreg;
struct mlx5_sq_bfreg fp_bfreg;
};
static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
};
static inline struct mlx5_ib_cq *to_mibcq(struct mlx5_core_cq *mcq)
-static u8 get_fence(u8 fence, struct ib_send_wr *wr)
-{
- if (unlikely(wr->opcode == IB_WR_LOCAL_INV &&
- wr->send_flags & IB_SEND_FENCE))
- return MLX5_FENCE_MODE_STRONG_ORDERING;
-
- if (unlikely(fence)) {
- if (wr->send_flags & IB_SEND_FENCE)
- return MLX5_FENCE_MODE_SMALL_AND_FENCE;
- else
- return fence;
- } else if (unlikely(wr->send_flags & IB_SEND_FENCE)) {
- return MLX5_FENCE_MODE_FENCE;
- }
-
- return 0;
-}
-
static int begin_wqe(struct mlx5_ib_qp *qp, void **seg,
struct mlx5_wqe_ctrl_seg **ctrl,
struct ib_send_wr *wr, unsigned *idx,
static int begin_wqe(struct mlx5_ib_qp *qp, void **seg,
struct mlx5_wqe_ctrl_seg **ctrl,
struct ib_send_wr *wr, unsigned *idx,
static void finish_wqe(struct mlx5_ib_qp *qp,
struct mlx5_wqe_ctrl_seg *ctrl,
u8 size, unsigned idx, u64 wr_id,
static void finish_wqe(struct mlx5_ib_qp *qp,
struct mlx5_wqe_ctrl_seg *ctrl,
u8 size, unsigned idx, u64 wr_id,
- int nreq, u8 fence, u8 next_fence,
- u32 mlx5_opcode)
+ int nreq, u8 fence, u32 mlx5_opcode)
mlx5_opcode | ((u32)opmod << 24));
ctrl->qpn_ds = cpu_to_be32(size | (qp->trans_qp.base.mqp.qpn << 8));
ctrl->fm_ce_se |= fence;
mlx5_opcode | ((u32)opmod << 24));
ctrl->qpn_ds = cpu_to_be32(size | (qp->trans_qp.base.mqp.qpn << 8));
ctrl->fm_ce_se |= fence;
- qp->fm_cache = next_fence;
if (unlikely(qp->wq_sig))
ctrl->signature = wq_sig(ctrl);
if (unlikely(qp->wq_sig))
ctrl->signature = wq_sig(ctrl);
num_sge = wr->num_sge;
if (unlikely(num_sge > qp->sq.max_gs)) {
mlx5_ib_warn(dev, "\n");
num_sge = wr->num_sge;
if (unlikely(num_sge > qp->sq.max_gs)) {
mlx5_ib_warn(dev, "\n");
+ if (wr->opcode == IB_WR_LOCAL_INV ||
+ wr->opcode == IB_WR_REG_MR) {
+ fence = dev->umr_fence;
+ next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
+ } else if (wr->send_flags & IB_SEND_FENCE) {
+ if (qp->next_fence)
+ fence = MLX5_FENCE_MODE_SMALL_AND_FENCE;
+ else
+ fence = MLX5_FENCE_MODE_FENCE;
+ } else {
+ fence = qp->next_fence;
+ }
+
switch (ibqp->qp_type) {
case IB_QPT_XRC_INI:
xrc = seg;
switch (ibqp->qp_type) {
case IB_QPT_XRC_INI:
xrc = seg;
goto out;
case IB_WR_LOCAL_INV:
goto out;
case IB_WR_LOCAL_INV:
- next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
qp->sq.wr_data[idx] = IB_WR_LOCAL_INV;
ctrl->imm = cpu_to_be32(wr->ex.invalidate_rkey);
set_linv_wr(qp, &seg, &size);
qp->sq.wr_data[idx] = IB_WR_LOCAL_INV;
ctrl->imm = cpu_to_be32(wr->ex.invalidate_rkey);
set_linv_wr(qp, &seg, &size);
break;
case IB_WR_REG_MR:
break;
case IB_WR_REG_MR:
- next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
qp->sq.wr_data[idx] = IB_WR_REG_MR;
ctrl->imm = cpu_to_be32(reg_wr(wr)->key);
err = set_reg_wr(qp, reg_wr(wr), &seg, &size);
qp->sq.wr_data[idx] = IB_WR_REG_MR;
ctrl->imm = cpu_to_be32(reg_wr(wr)->key);
err = set_reg_wr(qp, reg_wr(wr), &seg, &size);
- finish_wqe(qp, ctrl, size, idx, wr->wr_id,
- nreq, get_fence(fence, wr),
- next_fence, MLX5_OPCODE_UMR);
+ finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+ fence, MLX5_OPCODE_UMR);
/*
* SET_PSV WQEs are not signaled and solicited
* on error
/*
* SET_PSV WQEs are not signaled and solicited
* on error
- finish_wqe(qp, ctrl, size, idx, wr->wr_id,
- nreq, get_fence(fence, wr),
- next_fence, MLX5_OPCODE_SET_PSV);
+ finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+ fence, MLX5_OPCODE_SET_PSV);
err = begin_wqe(qp, &seg, &ctrl, wr,
&idx, &size, nreq);
if (err) {
err = begin_wqe(qp, &seg, &ctrl, wr,
&idx, &size, nreq);
if (err) {
- next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->wire,
mr->sig->psv_wire.psv_idx, &seg,
&size);
err = set_psv_wr(&sig_handover_wr(wr)->sig_attrs->wire,
mr->sig->psv_wire.psv_idx, &seg,
&size);
- finish_wqe(qp, ctrl, size, idx, wr->wr_id,
- nreq, get_fence(fence, wr),
- next_fence, MLX5_OPCODE_SET_PSV);
+ finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
+ fence, MLX5_OPCODE_SET_PSV);
+ qp->next_fence = MLX5_FENCE_MODE_INITIATOR_SMALL;
num_sge = 0;
goto skip_psv;
num_sge = 0;
goto skip_psv;
- finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq,
- get_fence(fence, wr), next_fence,
+ qp->next_fence = next_fence;
+ finish_wqe(qp, ctrl, size, idx, wr->wr_id, nreq, fence,
mlx5_ib_opcode[wr->opcode]);
skip_psv:
if (0)
mlx5_ib_opcode[wr->opcode]);
skip_psv:
if (0)